EZH2 knockout in oral cavity basal epithelia causes more invasive squamous cell carcinomas

Abstract

Oral squamous cell carcinoma (oral SCC) is an aggressive disease and despite intensive treatments, 5-year survival rates for patients have remained low in the last 20 years. Enhancer of zeste homolog 2 (EZH2), part of polycomb repressive complex 2 (PRC2), is highly expressed in human oral SCC samples and cell lines and has been associated with greater epithelia-to-mesenchymal transition (EMT), invasion and metastasis. Here, we developed a tamoxifen-regulated, transgenic mouse line (KcEZH2) in which EZH2 is selectively knocked out (KO) in some tongue epithelial basal stem cells (SCs) in adult mice. EZH2 KO SCs do not show the H3K27me3 mark, as assessed by double-label immunofluorescence. We used this mouse line to assess EZH2 actions during oral tumorigenesis with our immunocompetent 4-nitroquinoline 1-oxide model of oral SCC. We report that higher percentages of mice with invasive SCCs and high-grade neoplastic lesions are observed in mice containing EZH2 KO SCs (KcEZH2-2TΔ and KcEZH2-5TΔ mice). Moreover, EZH2 expression does not correlate with the expression of markers of invasive SCCs. Finally, EZH2 KO cells that are E-cadherin+ are present at invasion fronts infiltrating underlying muscle tissue. Our findings indicate that the knockout of EZH2 in basal SCs of tongue epithelia results in more aggressive carcinomas, and this should be considered when targeting EZH2 as a therapeutic strategy.

Graphical Abstract

A model for the induction of invasive squamous cell carcinomas in KcEZH2-5TΔ mice. Arrows depict the effect of EZH2 KO on these processes during 4-NQO-induced tumorigenesis in KcEZH2-5TΔ mice (compared to KcEZH2 mice).

Figure 1.

Tamoxifen-induced knockout of EZH2 in the tongue epithelial stem cells of Transgenic K14-CreER^TAM; EZH2^F/F (KcEZH2) mice. (A) Schematic illustrating the tamoxifen-inducible EZH2 KO system in KcEZH2 Transgenic mice. (B) Verification of tamoxifen-induced excision of Exons 16 and 17 in 5-day treated (KcEZH2-5TΔ) transgenic mice by PCR. Genomic DNA was isolated from tongue epithelia and livers of two KcEZH2-5TΔ mice 24 weeks after treatments. Two age-matched untreated (KcEZH2) mice were used as negative controls. 36B4 was used as a loading control. (C) EZH2 IF staining in tongue and liver samples in KcEZH2 and KcEZH2-5TΔ mice 4 weeks after tamoxifen treatment (200×; scale bar: 100 μm; N = 2 mice/group, 5 fields/mouse; representative fields are shown). EZH2 KO SCs are indicated by arrows. The ratios of the percentages of SCs expressing EZH2 in tongue epithelia per condition are shown on the right. Data graphed denotes the mean ± standard deviation of the mean (SD). Statistical significance was determined using Welch’s t-test. *0.01 < P < 0.05, **0.001 < P < 0.01, ***0.0001 < P < 0.001, ****P < 0.0001.

Figure 2.

EZH2 KO changes H3K27 methylated epigenetic marks but does not affect differentiation of basal SCs in tongue epithelia. (A) Timeline outlining the different treatment groups for this experiment. All samples were collected at 30 weeks. (B) EZH2 and Ki67 IF staining in tongue samples from all experimental groups. (C) EZH2 and H3K27me3 IF staining in KcEZH2 and KcEZH2-5TΔ groups. (D) H3K27me2 and Ki67 IF staining in KcEZH2 and KcEZH2-5TΔ groups. (E) EZH2 and K14 IF staining in KcEZH2 and KcEZH2-5TΔ groups. (F) Loricrin and Ki67 IF staining in KcEZH2 and KcEZH2-5TΔ groups. For (B–F), N = 3 mice/group (N = 2 mice for WT group) and 5 fields/mouse were analyzed, and representative fields are shown. EZH2 KO SCs are indicated by arrows (200×; scale bar: 100 μm).

Figure 3.

Tumor development and pathological classification during 4-NQO-induced carcinogenesis. (A) Timeline outlining the different treatment groups for this experiment. All mice shown in Figures 3–6 were sacrificed at 30 weeks. (B) Representative whole tongue images displaying epithelial lesions graded from least (1) to most (6) severe from 4-NQO treated mice (10× magnification; 2 mm scale bar). Arrows highlight tongue lesions and progression with increasing grade in size and number. (C, D) Graphs of the distribution (%) of tongue lesions by tumor grade (C) and lesion number (D) from KcEZH2 (N = 9), KcEZH2-2TΔ (N = 8), and KcEZH2-5TΔ (N = 9) mice. (E) Representative images of H&E stained sections for normal tongue epithelium, dysplasia, in-situ squamous cell carcinoma (in-situ SCC), and invasive SCC (100× magnification; 100 μm scale bar). (F) Distribution of the most severe 4-NQO-induced lesion observed in each mouse by pathological classification, with KcEZH2 (N = 9), KcEZH2-2TΔ (N = 8), and KcEZH2-5TΔ (N = 9) mice. For C, D and F, statistical significance was determined with the chi-square test. *0.01 < P < 0.05, **0.001 < P < 0.01, ***0.0001 < P < 0.001, ****P < 0.0001.

Figure 4.

EZH2 expression does not correlate with expression of phospho-STAT3, a marker of invasive SCCs. (A) H&E stained sections from KcEZH2-5TΔ mice with invasive SCCs (35×; 0.5 mm scale bar. Inset: 100×; 100 μm scale bar). (B) EZH2, STAT and p-STAT IF staining in normal, dysplasia and invasive samples from mouse KcEZH2-5TΔ # 2. 5 out of 6 fields that were obtained at the invasion fronts of KcEZH2-5TΔ #2 contained invading SCs with no EZH2 expression and high p-STAT3 expression. A representative “Invasive” field is shown. Same lesion area is circled in “Invasive” panels for comparison (200×; scale bar: 100 μm). The field in “Invasive” panels corresponds to arrow in H&E section of KcEZH2-5TΔ # 2 in (A).

Figure 5.

EZH2 KO SCs do not show H3K27me3 marks and have reduced K14 expression in pre-neoplastic/neoplastic tongue lesions. (A) EZH2 and H3K27me3 IF staining in the 4-NQO-treated KcEZH2 and KcEZH2-5TΔ groups. (B) H3K27me2 and Ki67 IF staining in the 4-NQO-treated KcEZH2 and KcEZH2-5TΔ groups. (C) EZH2 and K14 IF staining in the 4-NQO-treated KcEZH2 and KcEZH2-5TΔ groups. (D) Loricrin and Ki67 IF staining in the 4-NQO-treated KcEZH2 and KcEZH2-5TΔ groups. Areas containing EZH2 KO SCs are circled in A and C. Representative areas containing Loricrin expression at edges of lesions are circled in (D) (200×; scale bar: 100 μm). We report ratios of integrated density of proteins of interest (measured by Fiji) over integrated density of Hoechst signal in the same area (N = 3 mice/group, 8–11 fields/mouse, representative fields are shown). All data graphed denotes the mean ± standard deviation of the mean (SD). Statistical significance was determined using Welch’s t-test. *0.01 < P < 0.05, **0.001 < P < 0.01, ***0.0001 < P < 0.001, ****P < 0.0001.

Figure 6.

EZH2 KO SCs that are E-cadherin+ are present at invasion fronts of 4-NQO-induced lesions. (A) EZH2 and E-cadherin IF staining in KcEZH2 and KcEZH2-5TΔ groups (no 4-NQO). N = 3 mice/group and 5 fields/mouse were analyzed, and representative fields are shown. EZH2 KO SCs are indicated by arrows. (B) EZH2 and E-cadherin IF staining in the 4-NQO-treated KcEZH2 and KcEZH2-5TΔ groups. (C) EZH2 and Twist IF staining in the 4-NQO-treated KcEZH2 and KcEZH2-5TΔ groups. In (B–C), N = 3 mice/group and 8 to 11 fields/mouse were analyzed, and representative fields are shown. Areas containing EZH2 KO SCs in 4-NQO-induced lesions are circled in white. In (B), areas containing SCs with low E-cadherin expression in 4-NQO-induced lesions are circled in yellow. (D) EZH2 and E-cadherin IF staining in invasive fields from mouse KcEZH2-5TΔ #1. All 6 fields that were obtained at the invasion fronts of KcEZH2-5TΔ #1 contained invading SCs with no EZH2 expression and high E-cadherin expression. Two representative “Invasive” fields are shown. Corresponding H&E stained sections are shown on the right. All images are 200×; scale bar: 100 μm.