Senescence Induced by BMI1 Inhibition Is a Therapeutic Vulnerability in H3K27M-Mutant DIPG

Abstract

Diffuse intrinsic pontine glioma (DIPG) is an incurable brain tumor of childhood characterized by histone mutations at lysine 27, which results in epigenomic dysregulation. There has been a failure to develop effective treatment for this tumor. Using a combined RNAi and chemical screen targeting epigenomic regulators, we identify the polycomb repressive complex 1 (PRC1) component BMI1 as a critical factor for DIPG tumor maintenance in vivo. BMI1 chromatin occupancy is enriched at genes associated with differentiation and tumor suppressors in DIPG cells. Inhibition of BMI1 decreases cell self-renewal and attenuates tumor growth due to induction of senescence. Prolonged BMI1 inhibition induces a senescence-associated secretory phenotype, which promotes tumor recurrence. Clearance of senescent cells using BH3 protein mimetics co-operates with BMI1 inhibition to enhance tumor cell killing in vivo.

Keywords: BH3 mimetics; BMI1; DIPG; H3K27M mutant; H3WT; PTC 028; RNAi screen; SASP; senescence.

Graphical abstract

Figure 1

Complimentary Functional shRNA and Drug Screens Targeting Epigenetic Genes to Identify Crucial Regulators of the H3K27M-Mutant DIPG Tumor Cell Growth (A) Schematic of the epigenetic shRNA and drug screen performed. (B) Fold changes in depleted shRNAs (blue), unchanged shRNAs (black), and enriched shRNAs (red) targeting epigenetic genes from the shRNA pool. Blue dots show shRNAs corresponding to the set of genes when knockdown inhibits SU-DIPG04 cell proliferation, black dots represent shRNAs with no change, and red dots identify tumor-suppressor genes that, when inhibited, promote cell proliferation. The x axis shows shRNA representation for each gene. (C) Percentage of cell viability of SU-DIPG04 cells after indicated individual drug treatments from the drug screen pool for 5 days at 3 μM compared with DMSO-treated cells (100%). (D) Secondary validation using individual shRNAs of the top hits identified from the primary pooled-shRNA screen using SF8628 cells. Data represent mean ± SEM. (E) Venn diagram summarizes the number of genes with the highest fold change (FDR = 0.5) that were depleted in the two complementary screens using SU-DIPG04 cells. (F) Overlap of seven sets of genes significantly depleted in both screens. See also Figure S1 and Table S1.

Figure 2

IHC and ChIP-Seq Analysis of Indicated Proteins in H3K27M-DIPGs and Normal Cells (A) DIPG patient tumors and normal pons tissue section staining for H&E, BMI1, H3K27M, H3K27me3, and H2Aub. Positive control is the SU-DIPGXIII cell pellet sections stained for different proteins. Scale bar, 50 μm. (B and C) Scoring of BMI1, ^∗∗∗p < 0.003 DIPG versus pons (B), and H2Aub, ^∗p < 0.04 DIPG versus pons (C), from the IHC slides of pons and DIPG patient samples. Data represent mean ± SEM. (D) Western blot of BMI1 protein in DIPG cell lines. (E) Genome tracks of H3K27ac at the BMI1 promoter in OPCs, in different brain regions, and in H3K27M-DIPGs. The x axis represents the BMI1 genomic location, and the y axis represents the RPM (reads per million) value. See also Figure S2 and Table S2.

Figure 3

Impact of H3K27M Mutation on BMI1 Demonstrated Using Isogenic Model Cell Lines (A–C) Expression of BMI1 mRNA (A), BMI1 protein (B), and histone proteins (C). (D) Cell proliferation measured by XCELLigence. (E) Enrichment plot from GSEA for BMI1- and Mel18-related gene sets in H3.3K27M-mutant overexpression in HSJD-GBM01 (left) and in H3.3 WT overexpression in HSJD-DIPG-007 (right). (F) H3K27ac, H3K27me3, and total H3 ChIP-seq tracks at the BMI1 promoter in HSJD-DIPG-007 and H3.3 WT transduced cells. See also Figure S2D. (G) Effect of CRISPR-CAS9-mediated H3K27M-KO in SU-DIPGXIII on the expression of (i) BMI1 mRNA, (ii) BMI1 protein, (iii) histones in the parental and H3K27M-KO cells, and (iv) genome browser tracks showing changes in H3K27ac and H3K27me3 marks at the BMI1 promoter in these cells. Pairwise comparisons; ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001 by Student’s t test. See also Figures S2D and S2E. Data represent mean ± SEM.

Figure 4

In Vitro and In Vivo Anti-tumor Effects of BMI1 Inhibition in DIPGs (A) Growth inhibitory IC₅₀ values of BMI1 small-molecule inhibitors (i) in DIPG cell lines using PTC209 and PTC028 and (ii) in H3K27M-mutant-modified lines (+H3K27M is mutant transduced, and H3K27M-KO is the CRISPR-CAS9 KO of the mutant) using PTC028. See also Figure S5B. (B) Neurosphere size was measured by live-cell imaging with DIPG cells treated with PTC drugs, using Incucyte at day 12 after treatment. By ANOVA, DMSO versus PTC (three groups), p = 0.0034 (SU-DIPGXIII) and p = 0.0048 (SF8628). Pairwise comparison; ^∗p < 0.01, ^∗∗p < 0.002, ^∗∗∗p < 0.0001; DMSO versus PTC treatments by Student’s t test. See also Figure S5C. Data represent mean ± SEM (C) Cell viability measured using a primary DIPG patient-derived cell, UPN1285, treated with 100 nM of PTC028 for 3 days. Pairwise comparison; ^∗∗∗p < 0.004; ^∗DMSO versus PTC treatments by Student’s t test. Data represent mean ± SEM. (D) (i) Tumor implant and treatment protocol, (ii) Kaplan-Meier survival plot of the animals treated with either vehicle or PTC028 (n = 7 each group), and (iii) IHC analysis of p16, p21, and GLB1 expression of the treated tumor tissue. See also Figure S5D. (E) Volcano plot, with a heatmap adjacent to it, showing the differentially expressed genes detected by RNA-seq in PTC028-treated SF8628 cells compared with DMSO control. Genes highlighted in orange are expressed with an FDR = 0.05 and are associated with cell proliferation, differentiation, or tumor-suppressor pathways. See also Figure S5E. (F) GSEA from the RNA-seq data obtained from SF8628 cells treated with PTC028 (IC₅₀) compared with DMSO-treated cells annotated to multiple signaling pathway gene sets. See Figure S5F.

Figure 5

BMI1 and H2Aub Genomic Distribution and Changes in Its Enrichment at Specific Genomic Regions with Chemical Inhibition of BMI1 (A and B) ChIP-seq heatmap analysis of BMI1-enriched genomic loci (A) and KEGG pathway enrichment for BMI1 peaks (B) in SU-DIPG04 and SF8628 cells. (C) ChIP-seq heatmaps of BMI1 and H2Aub in SF8628 cells treated with DMSO (blue) or PTC028 (orange). (D and E) Average profiles of BMI1 peaks (D) and H2Aub (E) in SU-DIPG04 cells treated with DMSO (green) or PTC028 (orange). (F) KEGG pathway changes in BMI1 peak enrichment in SU-DIPG04 cells after treatments. (G) Genome browser views representing the results of ChIP-seq for BMI1 and H2Aub and the results of RNA-seq for the genes GLUL, PTGDS, and SFRP2 in SU-DIPG04 cells. All ChIP-seq peaks are relative to TSS (±3 kb). Asterisks indicate pathways of differentiation. See also Figure S6.

Figure 6

Senescence and Senescence-Associated Gene Expressions and Secretory Phenotype (SASP) with BMI1 Inhibition and Impact of BH3 Mimetics on BMI1-Inhibition-Induced Cellular Senescence and Apoptosis (A and B) Representative pictures of SA-β-gal (senescence associated beta galactosidase) staining with genetic inhibition of BMI1 (A) and with BMI1 inhibitors (B) in SF8628. Adjacent graphs show the quantification of the number of cells with blue staining normalized to the total number of cells. By ANOVA, the p value within the three groups was p < 0.0001 (A) and p = 0.0037 (B). (C) Western blot analysis of p16 and p21 in SU-DIPG04 cells transduced with shBMI1. (D) Western blot analysis of p16 and p21 in SF8628 cells treated with PTC028. (E–G) Concentration of SASP factors released by DIPG cells after treatment with either PTC209 (E) or PTC028 (F) and after genetic knockdown of BMI1 (G). (H) Changes in BMI1 and H2Aub genome tracks at the specific gene locations indicated in SU-DIPG04 cells after PTC028 treatment. The x axis represents genome locations, and the y axis represents the RPM for the indicated genes. For RNA-seq tracks, the increase in gene expression with PTC028 treatment is visualized by greater transcript tag density over the exons of the indicated genes (exons are the horizontal bars, and the direction of transcription is indicated by arrows shown below ChIP-seq tracks). (I) Representative pictures of neurospheres and sub-spheres (red arrows) formed in short- and long-term exposure of DIPG cells to PTC028, with quantification of size and number of neurospheres formed shown adjacently. (J) SASP factors measured in SU-DIPG04 with long-term exposure to PTC209. (K) Western blot showing changes in anti-apoptotic proteins after short- and long-term exposure of cells to PTC drugs. (L) Dose-response curve on cell viability (MTS (viability asassay endpoint) of SF8628 cells treated with PTC028 and ABT263 either alone or in combination. MTS assay is a viability assay based on cell metabolic activity. (M) Real-time measurement of proliferation of HSJD-DIPG-007 shBMI1 or shNull cells with ABT263 (250 nM). (N) Representative pictures of SA-β-gal staining of BT245 cells, with different treatment conditions shown. (O) Flow cytometry analysis of annexin V-PI/7AAD as a measure of apoptosis in HSJD-DIPG-007 cells treated with PTC028 and obatoclax. By ANOVA, ^∗∗∗∗p < 0.0001. Pairwise comparisons; ^∗p < 0.05, ^∗∗p < 0.012, ^∗∗∗∗p < 0.001; control versus treatment by Student’s t test. Data represent mean ± SEM. See also Figures S7 and S8.

Figure 7

In Vivo Effects of PTC028 and Obatoclax Either Alone or in Sequential Combination in DIPG Xenografts, and Overall Results from This Study (A) Schematics illustrating the experimental in vivo drug treatment schedule. After tumor establishment in mouse pons, the animals were randomized into four treatment groups: (i) vehicle alone, (ii) PTC028 alone, (iii) obatoclax alone, and (iv) sequential treatment with PTC028 and obatoclax. (B) Representative axial MRI of intracranial pons tumor in mice at the 2^nd week of the indicated treatment and at the 4^th week (a week after treatment was stopped). X denotes those animals that reached the endpoint. (C) Quantification of tumor volume from the MRI. Data represent mean ± SEM. (D) Representative in vivo bioluminescence images of the Luc-expressing BT245 tumor cells before, after, and during the treatments specified. The scale bar adjacent to the image displays bioluminescence counts (photons per second). (E) Kaplan-Meier survival plot of the animals exposed to the different drug treatments specified. (F) Graphical representation highlighting the overall outcome of this study. See also Figure S9.