Transcriptional regulation of protein synthesis by mediator kinase represents a therapeutic vulnerability in MYC-driven medulloblastoma

Abstract

MYC-driven medulloblastoma (MB) is a highly aggressive brain tumor with poor prognosis and limited treatment options. Through CRISPR-Cas9 screening, we identify the Mediator-associated kinase CDK8 as a critical regulator of MYC-driven MB. Both genetic loss and pharmacological inhibition of CDK8 impair MB tumor growth. Moreover, we find that CDK8 cooperates with MYC to sustain the MYC-mediated translational program, as CDK8 depletion induces pronounced transcriptional changes in translation-associated gene sets, reduces ribosome biogenesis, and impairs protein synthesis. Mechanistically, CDK8 regulates the occupancy of RNA polymerase II at specific chromatin loci, facilitating epigenetic alterations that promote the transcription of ribosomal genes. Furthermore, combined inhibition of CDK8 and mTOR synergistically enhances therapeutic efficacy in vivo, leading to more pronounced tumor growth suppression. Overall, our findings establish a functional link between CDK8-mediated transcriptional regulation and mRNA translation, suggesting a promising therapeutic approach targeting protein synthesis for MYC-driven MB.

Fig. 1. CDK8 is a specific vulnerability in MYC-driven medulloblastoma.

a Log-fold change in gene dependency from CRISPR-Cas9 screens in MB lines with TP53 as a positive control. b Read counts of sgRNAs targeting CDK8. Each dot represents an individual sgRNA. A decrease in read counts after puromycin selection indicates that CDK8 is selected as an essential gene. c DepMap lineage plot of CDK8 DEMETER2 scores across cancer types. n = cell lines per lineage. Lower scores indicate greater dependency. Boxes = Q1, median, Q3; whiskers = 1.5× IQR. Statistics from DepMap. d MB-specific gene-dependency t-statistics from DepMap; values < 0 denote essential genes. e UMAP analysis of cell clusters of single-cell RNA-seq data from G3-MB patient samples. A total of 12,595 cells were plotted after quality-control filtering. f CDK8 immunofluorescence staining. Data points from n = 3 biological replicates. Scale bar, 100 μm. Line indicates median. One-way ANOVA. g Proliferation of shNull/shCDK8 D425 (n = 4) and D458 (n = 5) cells. Technical replicates. h Images of neurosphere size in CDK8 knockdown and control cells are shown. n = 5 technical replicates. Scale bar, 400 μm. Barplot is for day 10. i Sphere formation and self-renewal measured by ELDA in shNull/shCDK8 MB lines. P values: likelihood ratio test. n = 5 biological replicates. j CDK8 knockout reduces D458 proliferation. n = 3 technical replicates. k Representative images of neurospheres formed by CDK8 knockout (n = 5) and control cells (n = 5). Scale bar, 200 μm. l Immunofluorescence staining of CDK8 in xenograft tumors from shNull or shCDK8 D458. White dashed lines = tumor boundary. n = 2 (shNull), n = 3 (shCDK8); one shNull mouse died before scan. Scale bar, 100 μm. Box plot line = mean. Two-sided t-test. m Representative MRI of xenografts. n = 2 (shNull), n = 3 (shCDK8). n H&E shows reduced tumor formation. Three mice per group euthanized at day 20. o Kaplan–Meier survival analysis of shNull (n = 6) and shCDK8 (n = 5) mice. Log-rank test.

Fig. 2. Targeting medulloblastoma with CDK8 inhibitor.

a IC50 determination of various CDK8 inhibitors in MB cell lines. Unit: μmol. b IC50 of RVU120 at 72 h in MB and NHA cells. Experiments were performed in n = 3 independent experiments. All comparisons are to NHA cells. One-way ANOVA. c Dose-dependent proliferation curve of RVU120-treated primary MB cells from a G3-MB patient. n = 5 biological replicates. Mean ± SEM. One-way ANOVA. d Immunofluorescence of CDK8 and DAPI. MB cells were treated with 1000 nM RVU120 for 48 h. Data points from n = 3 biological replicates. Scale bar, 10 μm. The line on the box plot represents the median. Two-sided unpaired t-test. e Immunoblot analysis of p-STAT1 following time-dependent treatment with 1000 nM RVU120 across MB cell lines. Representative of n = 3 experiments. f Immunoblot analysis of p-STAT1 levels following dose-dependent RVU120 treatment at 72 h. Representative of n = 3 experiments. Mean ± SEM. Two-way ANOVA. g Methylcellulose assay in MB cells treated with RVU120. n = 3 biological replicates. Mean ± SEM. Two-way ANOVA. h Annexin V apoptosis assay. MB cells were treated with 1000 nM RVU120 for 48 h. n = 3 biological replicates. Mean ± SEM. Two-way ANOVA. i Identification of the brain tumor-initiating cell fraction in MB cells by ALDH expression demonstrates a decrease in the ALDH⁺ fraction following 1000 nM RVU120 treatment for 48 h. n = 3 biological replicates. Mean ± SEM. Two-way ANOVA. j Representative bioluminescence images of mice treated with RVU120 or vehicle (control, n = 8; RVU120, n = 6). k Kaplan–Meier survival curves of mice treated with vehicle (n = 8) or RVU120 (n = 6). Log-rank test. l Representative MRI of PDX411 xenograft mice treated with RVU120 or vehicle. Mice received the first scan after 14 days of treatment. Asterisks denote spongy tissue texture in mice. An adjusted texture analysis was performed to measure the tumor size. n = 3 mice. One control mouse died before the final scan. Mean ± SEM. Two-way ANOVA.

Fig. 3. CDK8 sustains MYC transcriptional activity in medulloblastoma.

a DEMETER2 scores reveal differing CDK8/19 dependency in high- vs. low-MYC medulloblastoma cells. Two-sides Mann–Whiteny test. b Microarray analysis of CDK8 and CDK19 expression across four subgroups and twelve subtypes of 763 MB patient samples (Cavalli et al.) n = 6 normal cerebellum samples were collected by us. Boxes represent the first, median, and third quartiles, with whiskers extending to 1.5× the interquartile range. Two-sided Wilcoxon test. The p-values for the subtypes are provided in the supplementary information. c Kaplan–Meier survival analysis showing the association between CDK8 expression and overall survival within Group 3 subtypes. Log-rank test. Group 3β: n = 10 high, 17 low; Group 3r: n = 18 high, 13 low; Group 3β + r: n = 50 high, 8 low. d Correlation analysis of CDK8 and CDK19 with MYC or MYCN expression in Group 3 and SHH subgroups; IMPDH2 and MYCNOS were included as positive controls. e IGV tracks showing c-MYC ChIP-seq peaks at the promoters of CDK8 and CDK19; IMPDH2 and ODC1 serve as positive controls with MYC binding peaks. f RNA-seq analysis following knockdown of c-MYC shows no change in CDK8 expression in MB002 cells. n = 2 for each condition. g RNA-seq analysis following CDK8 knockdown with three shRNAs, showing the expression levels of both c-MYC and CDK8 in D458 cells. n = 3 for each condition. P-values were adjusted using the Benjamini–Hochberg FDR and reported as padj (DESeq2). h Volcano plot showing differentially expressed genes in CDK8 knockdown cells compared to shNull control cells. P-values were adjusted using the Benjamini–Hochberg FDR and reported as padj (DESeq2). i Immunoblot showing CDK8 and c-MYC protein levels in MB cells following CDK8 knockdown. n = 3 independent experiments. Mean ± SEM. Two-way ANOVA. j Immunoblot showing CDK8 and c-MYC protein levels in MB cells with CRISPR/Cas9-mediated knockout of CDK8. n = 3 independent experiments. Mean ± SEM. Two-way ANOVA. k Gene set enrichment analysis of hallmark gene sets using RNA-seq data comparing CDK8 knockdown (shCDK8) to control (shNull) cells. n = 3 for each condition.

Fig. 4. CDK8 regulates protein synthesis in MYC-driven medulloblastoma.

a Alterations in GSEA gene sets observed in D458 cells following CDK8 knockdown. FDR from GSEA. b GSVA of patient samples (n = 763) showed enrichment of ribosome biogenesis gene sets in MYC-overexpressing Group 3β/3γ subtypes. c The expression of ribosomal genes was compared to that of all other genes in MB cells using RNA-seq analysis. d RNA-seq analysis demonstrated alterations in the expression of mitochondrial and cytoplasmic ribosomal genes following loss or inhibition of CDK8 in MB cells. n = 3 for each condition. P-values were FDR-adjusted. e RNA-seq of D458 cells: cytosolic ribosomal gene expression in shCDK8, sgCDK8, and control. n = 3 for each condition. f GSEA network showing downregulation of ribosome biogenesis in shCDK8 vs. shNull D458 cells. Node size reflects gene counts; edges indicate shared genes. g Top 10 GO biological processes are shown (n = 3 per condition; FDR-adjusted P values, GSEA). h Polysome profiling of lysates from CDK8 knockout or RVU120-treated (1 μM, 48 h) D458 cells reveals that CDK8 depletion reduces the ratio of polysome to sub-polysome compared to controls. i Immunofluorescence of Y10B and DAPI at 40X. MB cells were treated with the 1000 nM RVU120 for 48 h. Scale bar, 10 μm. Median line. n = 3 independent experiments. Two-sided t-test. j Immunofluorescence staining shows EU-incorporated RNA detected by Click-iT labeling. Scale bar, 100 μm. The line on the box plot represents the mean. n = 3 independent experiments. Kruskal-Wallis test. k CRISPR-mediated CDK8 knockout and control D458 cells were labeled with EU for 1 hour, and detected using the Click-iT reaction. Scale bar, 100 μm. Line indicates mean. n = 3 independent experiments. Two-sided Mann-Whitney test. l Flow cytometry analysis of OPP incorporation in MB cells treated with RVU120 at 1,000 nM versus control, quantified using FlowSight. Scale bar, 10 μm. Mean ± SEM. n = 3 independent experiments. Two-way ANOVA. m OPP assay showing protein synthesis in RVU120-treated MB cells compared to control cells. Scale bar, 100 μm. Line indicates mean. n = 3 independent experiments. Kruskal-Wallis test.

Fig. 5. Interplay between MYC and CDK8 in controlling protein synthesis.

a Top 10 genes showing the highest correlation with MYC expression in Group 3 MB, based on analysis of the Cavalli dataset. n = 763. b GSEA of C5 GO gene sets comparing shMYC versus shNull in MB cells. P-values were adjusted for multiple testing using the Benjamini–Hochberg FDR in GSEA. n = 2 for each condition. c GSEA indicated alterations in GO biological process gene sets (FDR < 0.05) following the knockdown of CDK8, MYC, CDK7, CDK11, HNRNPH1, SOX11, or PLK1. n = 3 for each condition. d OPP assay in D458 and MB002 cells transfected with GFP-shRNA targeting MYC, followed by treatment with two doses of RVU120. Scale bar, 100 μm. The line on the box plot represents the median. n = 3 independent experiments. Kruskal-Wallis test. e OPP assay in D458 cells transfected with RFP-Omomyc, where RFP-positive cells mark Omomyc expression. Scale bar, 100 μm. The line on the box plot represents the median. n = 3 independent experiments. Two-sided Mann-Whitney test. f OPP assay in D458 cells transfected with RFP-Omomyc, followed by treatment with 1,000 nM RVU120. White arrows mark Omomyc-expressing (RFP-positive) cells. Scale bar, 100 μm. The line on the box plot represents the median. n = 3 independent experiments. Kruskal-Wallis test. g OPP assay showing dose-dependent effects of RVU120 treatment on ONS76 cells. Scale bar, 100 μm. The line on the box plot represents the median. n = 3 independent experiments. Kruskal-Wallis test. h OPP assay showing dose-dependent effects of RVU120 treatment on ONS76 cells transfected with c-MYC. Scale bar, 100 μm. The line on the box plot represents the median. n = 3 independent experiments. Kruskal-Wallis test. i IC₅₀ of RVU120 in parental ONS76 and DAOY cells compared to c-MYC–overexpressing cells. Mean ± SD. Two-sided unpaired t-test. Experiments were performed in three independent experiments, each with five biological replicates.

Fig. 6. Chromatin binding profiles of CDK8 in MB cells.

a Heatmaps showing CUT&RUN signals of CDK8, H3K4me3, H3K4me1, H3K27ac, BRD4, and MYC in D458 MB cells. The signals were displayed within a region spanning ± 3 kb around the transcription start site (TSS). n = 2 for each condition. b Pie chart showing CDK8 peaks are localized at promoter and enhancer. n = 2 for each condition. c Pathway enrichment analysis of CDK8 binding genes inferred from CUT&RUN. Translation pathways are enriched in MB cell lines. A total of 11,675, 14,909, and 12,895 genes were identified in D458, D425, and D283 cells, respectively. Statistical significance was assessed using Fisher’s exact test with the total number of genes in the genome as the background, and p-values were adjusted for multiple testing using the Benjamini–Hochberg FDR. d Venn-diagram showing overlapping of CDK8 binding genes associated with mRNA translation pathways. e Heatmaps displaying genome-wide binding CUT&RUN signals of CDK8 in CDK8 knockdown D458 cells compared to control cells. The signals are displayed within a region spanning ± 3 kb around the transcription start site (TSS). n = 2 for each condition. f Heatmaps displaying CUT&RUN signals of CDK8 and H3K4me3 in D458 cells with CDK8 knockdown compared to control cells at promoter regions. n = 2 for each condition. g Pathway enrichment analysis of genes associated with loss of H3K4me3 peaks (1207 genes). Statistical significance was assessed using Fisher’s exact test with the total number of genes in the genome as the background, and p-values were adjusted for multiple testing using the Benjamini–Hochberg FDR. h Heatmaps showing CUT&RUN signals of BRD4, H3K4me1, and MYC in D458 MB cells following CDK8 knockdown. n = 2 for each condition.

Fig. 7. CDK8 transcriptionally regulates the expression of ribosomal genes.

a Heatmaps showing CUT&RUN signals of Pol II and phospho-Pol II in D458 cells with CDK8 knockdown compared to control cells at promoter regions. n = 2 for each condition. b Empirical cumulative distribution function (ECDF) plot shows significant increase in promoter-proximal pausing following CDK8 knockdown. n = 2 for each condition. c Average distribution and heatmaps of H3K4me3, Pol II, and phospho-Pol II signals on ribosomal genes. n = 2 for each condition. d Representative examples of Pol II and phospho-Pol II binding sites on ribosomal genes observed following CDK8 knockdown. n = 2 for each condition. e Enrichment analysis shows mRNA translation pathways enriched among genes with increased Pol II peaks (11,617 genes) or decreased phospho-Pol II peaks (7174 genes) following CDK8 knockdown. Statistical significance was assessed using Fisher’s exact test with the total number of genes in the genome as the background, and p-values were adjusted for multiple testing using the Benjamini–Hochberg FDR. f Immunoblot showing the levels of Pol II and phospho-Pol II in D458 cells following treatment with RVU120. Representative of n = 3 experiments. g Heatmaps showing CUT&RUN signals of RNA Pol II and phospho-RNA Pol II in D458 MB cells treated with 1,000 nM RVU120 for 48 h. n = 2 for each condition. h Average distribution of RNA Pol II and phospho-RNA Pol II peaks showing the alteration of RNA Pol II and phospho-RNA Pol II signals across the gene body following the treatment of RVU120. n = 2. i Average distribution and heatmaps of RNA Pol II and phospho-RNA Pol II signals on cytosolic and mitochondrial ribosomal genes following the treatment of RVU120. n = 2. j Representative examples of RNA Pol II and phospho-RNA Pol II binding sites on ribosomal genes observed following the treatment of RVU120. n = 2 for each condition.

Fig. 8. CDK8 regulates mTOR signaling in MYC-driven medulloblastoma.

a Gene set variation analysis of patient samples (n = 763) revealed that the MYC-overexpressing subtypes Group3β and 3γ were enriched with gene sets of MYC and mTOR signaling. b Multiplex IHC on G3-MB patient samples using CDK8, p-4EBP1, c-MYC, RPS12, p-S6, and p-AKT antibodies. n = 3 patient samples were analyzed. For each antibody, 10 fields of view were captured, and the signal intensities from all points were combined for analysis. Top scale bar, 100 μm. Bottom scale bar, 10 μm. p < 0.0001 in all biomarker groups. Two-sided Mann–Whitney U. c Representative bioluminescence images of mice injected with D425 cells and treated with TAK-228 (1 mg/kg, daily, oral gavage) or vehicle control. Treatment was initiated upon tumor establishment and continued until endpoint. Kaplan–Meier survival curves show extended survival in the TAK-228–treated cohort compared to controls. Statistical significance was determined using the log-rank test. n = 5 for each group. d Representative bioluminescence images and Kaplan–Meier survival analysis of D458 xenograft-bearing mice treated with TAK-228 (1 mg/kg, daily, oral gavage), following the same protocol as in (c). TAK-228 treatment reduced tumor burden and significantly prolonged survival compared to the control group (log-rank test). n = 5 for each group. e GSEA plots of representative gene sets involved in mTOR signaling following CDK8 depletion or inhibition. Normalized enrichment score (NES) and false discovery rate (FDR) are indicated. n = 3 for each condition. f–h Immunoblot showing the levels of p-4EBP1 and p-S6 upon treatment with RVU120 in Group 3 medulloblastoma cells. n = 3 biological replicates. i Immunoblot showing the levels of p-4EBP1 and p-S6 following CDK8 knockout. n = 3 independent experiments. Mean ± SEM. Statistical analysis: two-way ANOVA. j Immunoblot showing the levels of p-mTOR and mTOR upon treatment with RVU120 for 24 h in D458 and MB002 cells. The data are representative of n = 3 independent experiments.

Fig. 9. Synergistic targeting of CDK8 and mTOR in MYC-Driven medulloblastoma.

a Dose-dependent assay of the combined treatment with RVU120 and Torin1 on day 5. Scale bar, 400 μm. n = 3 biological replicates. b Real-time proliferation assay quantifying the combined treatment with RVU120 and Torin1. Control (n = 9); treatment (n = 3) biological replicates. Mean ± SEM. All p < 0.0001 vs. control (*). Two-way ANOVA. c Heatmap representation of the Fraction Affected and the Bliss interaction index across the five-point dose range of RVU120 and Torin1. Mean values of n = 3 biological experiments are shown. d The combination index of RVU120 and Torin1 using Chou-Talalay method. The mean combination index was determined from n = 3 biological replicates. e Apoptosis assay following combined treatment with RVU120 and Torin1. MB cells were treated for 48 h before staining with PI and Annexin V. n = 3 biological replicates. Mean ± SEM. Two-way ANOVA. f, g Effects of the combination of RVU120 and Torin1 on protein synthesis markers, phospho-Pol2 and phospho-STAT1, in MB cells after 48 h of treatment. Representative of n = 3 experiments. h The nude mice injected with D458 cells were treated with vehicle (n = 10), RVU120 (40 mg/kg, n = 6), TAK-228 (1 mg/kg, n = 8), or their combination (n = 7). i Representative Sagittal T2-weighted turboRARE MRI of D458 xenografted mice at 22 days. White arrows indicate tumors. MRI volumetric analysis is shown. n = 2 mice were selected and scanned in each group. j Kaplan–Meier survival analysis of mice treated with vehicle control (n = 10), TAK-228 (n = 8), RVU120 (n = 6), or the combination of TAK-228 and RVU120 (n = 7). Statistical significance was assessed using the log-rank test.