BMI1 Inhibition Eliminates Residual Cancer Stem Cells after PD1 Blockade and Activates Antitumor Immunity to Prevent Metastasis and Relapse

Abstract

PD1 blockade-based combination therapy has been approved as a first-line treatment for head and neck squamous cell carcinoma (HNSCC). However, the response rate remains relatively low, and patients with HNSCC eventually relapse. Here, we show that the combination treatment of anti-PD1 and cisplatin enriched BMI1⁺ CSCs in HNSCC while inhibiting HNSCC growth. In contrast, the pharmacological and genetic inhibition of BMI1 eliminated BMI1⁺ CSCs and enabled PD1 blockade therapy, resulting in the inhibition of metastatic HNSCC and prevention of HNSCC relapses. BMI1 inhibition strongly induced tumor cell-intrinsic immune responses by recruiting and activating CD8⁺ T cells in addition to eliminating BMI1⁺ CSCs. Mechanistically, BMI1 inhibition induced CD8⁺ T cell-recruiting chemokines by stimulating IRF3-mediated transcription and erasing repressive H2A ubiquitination. Our results suggest that targeting BMI1 may enable immune checkpoint blockade to inhibit metastatic tumor growth and prevent tumor relapse by activating cell-intrinsic immunity, in addition to purging CSCs.

Keywords: BMI1; CD8+ T cells; HNSCC; PD1 blockade; antitumor immune response; cancer stem cells; immunotherapy; invasive growth; metastasis; squamous cell carcinoma.

Figure 1.. Enrichment of BMI1⁺ CSCs after the combination treatment of anti-PD1 and cisplatin.

(A) Schematic diagrams show the treatment and lineage tracing of primary HNSCC in Bmi1^CreER;Rosa^tdTomato mice. Tamoxifen (Tam) was administered 1 day prior to sacrificing (Sac) mice in order to label BMI1⁺ CSCs. (B) Representative image of tongue visible lesions in different treatment groups. Black dashed lines demark lesion areas. Scale bar, 2 mm. (C) Quantification of HNSCC lesion area from mice with treatment as indicated. Values are mean ± SD from the pool of two independent experiments. n = 10. *p < 0.05 and **p < 0.01 by one-way ANOVA. (D) Representative H&E staining of HNSCC from mice with treatment as indicated. Scale bar, 200 μm. Enlarged images are shown in the lower panels. Scale bar, 50 μm. (E) Quantification of HNSCC number and area from mice with treatment as indicated. Values are mean ± SD from the pool of two independent experiments. n = 10. *p < 0.05 and **p < 0.01 by one-way ANOVA. (F) Quantification of HNSCC invasion grades from mice with treatment as indicated. Data was pooled from two independent experiments. n = 10. *p < 0.05 by Cochran-Armitage test. (G) Immunostaining of metastatic cells in cervical lymph nodes using anti-PCK. Scale bar, 200 μm. (H) Percentage of metastatic lymph nodes from mice with treatment as indicated. Number of metastatic lymph nodes in each group is indicated in the figure. Data was pooled from two independent experiments. *p < 0.05 by Chi-square test. (I) Quantification of metastatic area in lymph nodes from mice with treatment as indicated. Values are mean ± SEM from the pool of two independent experiments. *p < 0.05 by one-way ANOVA. (J) Immunofluorescent images for CD8⁺ T from mice with treatment as indicated. Scale bar, 10 μm. (K) Quantifications of CD8⁺ T cells in HNSCC from mice with treatment as indicated. Values are mean ± SD from the pool of two independent experiments. n = 10, **p < 0.01 by one-way ANOVA. (L) Representative images of Tomato⁺ BMI1⁺ CSCs in HNSCC from mice with treatment as indicated. White dashed lines demark tumor-stromal junction. Scale bar, 10 μm. (M) Quantification of the percentage of Tomato⁺ cells in HNSCC from mice with treatment as indicated. Values are mean ± SD from the pool of two independent experiments. n = 10. *p < 0.05 and **p < 0.01 by one-way ANOVA. Also see Figure S1

Figure 2.. PTC-209 eliminates BMI1⁺ CSCs and collaborates with anti-PD1 to suppress HNSCC growth and metastasis by recruiting CD8⁺ Cells.

(A) Representative image of tongue visible lesions in different treatment groups as indicated. Black dashed lines demark lesion areas. Scale bar, 2 mm. (B) Quantification of HNSCC lesion areas from mice. Values are mean ± SD from the pool of two independent experiments. n = 12, *p < 0.05 and **p < 0.01 by one-way ANOVA. (C) Representative H&E staining of HNSCC from mice with treatment as indicated. Scale bar, 200 μm. Enlarged images are shown in the lower panels. Scale bar, 50 μm. (D) Quantification of HNSCC number and area. Values are mean ± SD from the pool of two independent experiments. n = 12, *p < 0.05 and **p < 0.01 by one-way ANOVA. (E) Quantification of HNSCC invasion grades. *p < 0.05 and **p < 0.01 by Cochran-Armitage test. (F) Representative images of active caspase3 (Ac-casp3, green) in HNSCC. Nuclei were stained with DAPI (blue). White dashed lines demark tumor-stromal junction. Scale bar, 10 μm. (G) Percentage of Ac-Casp3⁺ cells in HNSCC from mice with indicated treatments. Values are mean ± SD from the pool of two independent experiments. n = 12, **p < 0.01 by one-way ANOVA. (H) Immunostaining of metastatic cells in cervical lymph nodes by anti-PCK. Scale bar, 200 μm. (I) Quantification of percentage of metastatic lymph nodes from mice with treatment as indicated. Number of metastatic lymph nodes in each group is indicated in the figure. *p < 0.05 and **p < 0.01 by Chi-square test. (J) Quantification of metastatic areas in lymph nodes from mice with treatment as indicated. Values are mean ± SEM from the pool of two independent experiments. *p < 0.05 and **p < 0.01 by one-way ANOVA. (K) Representative immunofluorescent images for CD8 (red) and PCK (green) of HNSCC from mice with treatment as indicated. Nuclei were visualized by DAPI (blue). Scale bar, 10 μm. (L) Quantifications of CD8⁺ T cells percentage from mice with treatment as indicated. Values are mean ± SD from the pool of two independent experiments. n = 12, **p < 0.01 by one-way ANOVA. (M) Representative images of Bmi1⁺ cell-driven lineage tracing in HNSCC from mice with treatment as indicated. White dashed lines demark tumor-stromal junction. Scale bar, 10 μm. (N) Quantification of the percentage of Tomato⁺ cells in HNSCC from mice with treatment as indicated. Values are mean ± SD from the pool of two independent experiments. n = 12, *p < 0.05 and **p < 0.01 by one-way ANOVA. Also see Figures S2 and S3

Figure 3.. Depletion of intratumoral CD8⁺ T cells reverse PTC209 plus anti-PD1-mediated antitumor immunity.

(A) Representative immunofluorescent images for CD8 (red) and PCK (green) in HNSCC from mice with indicated treatments. Nuclei were visualized by DAPI (Blue). Scale bar, 10 μm. (B) Quantifications of percentage of CD8⁺ T cells in HNSCC. Values are mean ± SD from the pool of two independent experiments. n = 8, *p < 0.05 and **p < 0.01 by one-way ANOVA. (C) Representative image of tongue visible lesions. Scale bar, 2 mm. (D) Quantification of HNSCC lesion areas. Values are mean ± SD from the pool of two independent experiments. n = 8, **p < 0.01 by one-way ANOVA. (E) Representative H&E staining of HNSCC. Scale bar, 200 μm. Enlarged images are shown in the lower panels. Scale bar, 50 μm. (F) Quantification of HNSCC number and area. Values are mean ± SD from the pool of two independent experiments. n = 8, *p < 0.05 and **p < 0.01 by one-way ANOVA. (G) Quantification of HNSCC invasion grades. *p < 0.05 by Cochran-Armitage test. (H) Representative images of Ac-casp3 (red) and PCK (green) in HNSCC. Nuclei were visualized by DAPI (Blue). Scale bar, 10 μm. (I) Percentage of Ac-casp3⁺ apoptotic cells in all tumor cells. Values are mean ± SD from the pool of two independent experiments. n = 8, **p < 0.01 by one-way ANOVA. (J) Immunostaining of metastatic cells in cervical lymph nodes using anti-PCK. Scale bar, 200 μm. (K) Percentage of metastatic lymph nodes from mice. Number of metastatic lymph nodes in each group is indicated in the figure. *p < 0.05 and **p < 0.01 by Chi-square test. (L) Quantification of metastatic areas in cervical lymph nodes. Values are mean ± SEM from the pool of two independent experiments. *p < 0.05 and **p < 0.01 by one-way ANOVA.

Figure 4.. Epithelial deletion of BMI1 collaborates with anti-PD1 to suppress HNSCC growth and metastasis by recruiting CD8⁺ cells.

(A) Experimental design for Bmi1 knockout in tumor cells and anti-PD1 treatment in vivo. Three administrations of Tam were given to tumor-bearing mice. Mice were randomly divided into four experimental groups (n = 14 per group from two independent experiments): BMI1^f/f with IgG isotype, BMI1^f/f with anti-PD1, K14Cre;BMI1^f/f with IgG isotype, and K14Cre;BMI1^f/f with anti-PD1. (B) Representative image of tongue visible lesions. Black dashed lines demark lesion area. Scale bar, 2 mm. (C) Quantification of lesion areas from mice treated with different conditions as indicated. Values are mean ± SD from the pool of two independent experiments. *p < 0.05 and **p < 0.01 by two-way ANOVA. (D) Representative H&E staining of HNSCC from mice treated with different conditions as indicated. Scale bar, 200 μm. Enlarged images are shown in the lower panels. Scale bar, 50 μm. (E) Quantification of HNSCC area and number from mice treated with different conditions as indicated. Values are mean ± SD from the pool of two independent experiments. *p < 0.05 and **p < 0.01 by two-way ANOVA. (F) Quantification of HNSCC invasion grades from mice treated with different conditions as indicated. *p < 0.05 and **p < 0.01 by Cochran-Armitage test. (G) Represent immunostaining of metastatic cells in cervical lymph nodes by anti-PCK. Scale bar, 200 μm. (H) Quantification of percentage of metastatic lymph nodes. Number of metastatic lymph nodes in each group is indicated in the figure. *p < 0.05 and **p < 0.01 by Chi-square test. (I) Quantification of metastatic area in lymph nodes from mice treated with different conditions as indicated. Values are mean ± SEM from the pool of two independent experiments. *p < 0.05 and **p < 0.01 by two-way ANOVA. (J) Representative immunofluorescent images for CD8 (red) and PCK (green) in HNSCC. Nuclei were visualized by DAPI (blue). Scale bar, 10 μm. (K) Quantification of the percentage of CD8⁺ T cells from mice treated with different conditions as indicated. Values are mean ± SD from the pool of two independent experiments. **p < 0.01 by two-way ANOVA. ##p < 0.01 treatment x genotype interaction. Also see Figure S4.

Figure 5.. BMI1 Inhibition induces expression of effector T cell attracting chemokines in SCC cells by activating cGAS-STING-IRF3 signaling and erasing repressive H2AUb on their promoters.

(A) Heatmap from RNA-sequencing data showing the differentially expressed genes related to chemokines-mediated signaling in SCC23 cells upon PTC209 or BMI1 knockdown. Blue rectangles indicate the genes related to IFN-regulated chemokines. (B) qRT-PCR showed that the expression of CCL5, CXCL9, CXCL10, and CXCL11 in SCC23 cells were induced by PTC209 or BMI1 knockdown. Means ± SD were shown. **p < 0.01 by unpaired Student’s t test. (C) Immunofluorescent staining of pH2A.X (green) in 4NQO-induced HNSCC by PTC209 or BMI1 knockout and their quantifications. Nuclei were stained with DAPI (blue). Scale bar, 10 μm. Means ± SD were shown (n = 8). **p<0.01 by unpaired Student’s t test. (D) Confocal images showing cytosolic DNA accumulations and their quantifications in SCC23 cells upon PTC209 or shBMI1 treatment. Double strand DNA (dsDNA) was stained by Picogreen (green). Mitochondria and nuclei were respectively stained with Mito-tracker (Red) and DAPI (blue). White arrows indicate cytosolic dsDNA. Scale bar, 10 μm. More than 100 cells were analyzed per group. Means ± SD were shown. **p < 0.01 by one-way ANOVA. (E) Induction of phosphorylation of STING (S366), TBK1 (S172) and IRF3 (S396) in SCC23 cells by PTC209 or shBMi1 treatment. (F) qRT-PCR showing the induction of IFNβ mRNA expression in SCC23 cells by PTC209 or shBMI1 treatment. **p < 0.01 by unpaired Student’s t test. (G) The reduction of BMI1 occupied on the promoters of CCL5, CXCL9, CXCL10, and CXCL11 in SCC23 cells by PTC209. (H) The reduction of H2AUb levels on the promoters of CCL5, CXCL9, CXCL10, and CXCL11 in SCC23 cells by PTC209. (I) The reduction of BMI1 occupied on the promoters of CCL5, CXCL9, CXCL10, and CXCL11 in SCC23 cells by shBMI1. (J) The reduction of H2AUb levels on the promoters of CCL5, CXCL9, CXCL10, and CXCL11 in SCC23 cells by shBMI1. n = 3, means ± SD are shown. *p < 0.05 and **p < 0.01 by unpaired Student’s t test. Also see Figures S5, S6 and S7.

Figures 6.. Inhibition of chemokine signaling impairs PTC209 plus anti-PD1-mediated antitumor immunity.

(A) Representative immunofluorescent images for CD8 (red) and PCK (green) in HNSCC from mice with indicated treatments. Nuclei were visualized by DAPI (Blue). Scale bar, 10 μm. (B) Quantifications of percentage of CD8⁺ T cells in HNSCC. Values are mean ± SD from the pool of two independent experiments. n = 8, *p < 0.05 and **p < 0.01 by one-way ANOVA. (C) Representative image of tongue visible lesions. Scale bar, 2 mm. (D) Quantification of HNSCC lesion areas. Values are mean ± SD from the pool of two independent experiments. n = 8, *p < 0.05 and **p < 0.01 by one-way ANOVA. (E) Representative H&E staining of HNSCC. Scale bar, 200 μm. Enlarged images are shown in the lower panels. Scale bar, 50 μm. (F) Quantification of HNSCC number and area. Values are mean ± SD from the pool of two independent experiments. n = 8, *p < 0.05 and **p< 0.01 by one-way ANOVA. (G) Quantification of HNSCC invasion grades. n = 8, *p < 0.05 by Cochran-Armitage test. (H) Representative images of Ac-casp3 (red) and PCK (green) in HNSCC. Nuclei were visualized by DAPI (Blue). Scale bar, 10 μm. (I) Percentage of Ac-casp3⁺ apoptotic cells in all tumor cells. Values are mean ± SD from the pool of two independent experiments. n = 8, *p < 0.05 and **p < 0.01 by one-way ANOVA. (J) Immunostaining of metastatic cells in cervical lymph nodes using anti-PCK. Scale bar, 200 μm. (K) Percentage of metastatic lymph nodes from mice. Number of metastatic lymph nodes in each group is indicated in the figure. *p < 0.05 and **p < 0.01 by Chi-square test. (L) Quantification of metastatic areas in cervical lymph nodes. Values are mean ± SEM from the pool of two independent experiments. *p < 0.05 and **p < 0.01 by one-way ANOVA.

Figures 7.. The combination treatment of anti-PD1 and PTC209 prevents BMI1⁺ CSC-mediated tumor relapse.

(A) Experimental design for BMI1⁺ CSCs lineage tracing in HNSCC after treatment with anti-PD1 plus cisplatin or anti-PD1 plus PTC209 (n = 7 per group). After treatment, mice were injected with Tam and maintained for 4 additional weeks. (B) Representative images of Tomato⁺ tumor cells (red) derived from BMI1⁺ CSCs one month after treatment. Nuclei are stained with DAPI (blue). White dashed lines demark tumor-stromal junction. Scale bar, 10 μm. (C) Quantification of the percentage of Tomato⁺ tumor cells in HNSCC. Values are mean ± SD from the pool of two independent experiments. ns, not significant, *p < 0.05 and **p < 0.01 by one-way ANOVA. (D) Quantification of HNSCC lesion areas. Values are mean ± SD from the pool of two independent experiments. ns, not significant, *p < 0.05 and **p < 0.01 by one-way ANOVA. (E) Quantification of HNSCC number and area from mice with treatment as indicated. Values are mean ± SD from the pool of two independent experiments. ns, not significant, *p < 0.05 and **p< 0.01 by one-way ANOVA. (F) Quantification of HNSCC invasion grades. ns, not significant, *p < 0.05 by Cochran-Armitage test. (G) Experimental design for examining HNSCC relapse after treatment. Bmi^CreER;Rosa^tdTomato mice with 4NQO induced HNSCC were randomly divided into three experimental groups (n = 8). After treatment, mice were maintained for 8 additional weeks for the tumor relapse. Tamoxifen (Tam) was administered 1 day prior to sacrificing (Sac) the mice in order to label BMI1⁺ CSCs. (H) Representative image of tongue visible lesions. Black dashed lines demark lesion area. Scale bar, 2 mm. (I) Quantification of HNSCC lesion areas. Mean ± SD from the pool of two independent experiments. ns, not significant, *p < 0.05 and **p < 0.01 by one-way ANOVA. (J) H&E staining of HNSCC. Scale bar, 200 μm. Enlarged images are shown in the lower panels. Scale bar, 50 μm. (K) Quantification of HNSCC number and area. Mean ± SD from the pool of two independent experiments. ns, not significant, *p < 0.05 and **p< 0.01 by one-way ANOVA. (L) Quantification of HNSCC invasion grades. ns, not significant, *p < 0.05 by Cochran-Armitage test. (M) Immunostaining of metastatic cells in cervical lymph nodes by anti-PCK. Scale bar, 200 μm. (N) Percentage of metastatic lymph nodes in HNSCC. Number of metastatic lymph nodes in each group is indicated. ns, not significant, **p < 0.01 by Chi-square test. (O) Quantification of metastatic areas in lymph nodes. Mean ± SEM from the pool of two independent experiments. ns, not significant, *p < 0.05 and **p < 0.01 by one-way ANOVA. (P) Representative images of Tomato⁺ BMI1⁺ CSCs in HNSCC after treatment. White dashed lines demark tumor-stromal junction. Scale bar, 10 μm. (Q) Quantification of the percentage of Tomato⁺ BMI1⁺ CSCs in HNSCC after treatment. Mean ± SD from the pool of two independent experiments. *p < 0.05 and **p < 0.01 by one-way ANOVA.