Carbonic Anhydrase Inhibition Sensitizes Group 3 Medulloblastoma to Radiotherapy

Abstract

Group 3 (G3) medulloblastoma constitutes the most aggressive molecular subgroup, and nearly all patients present with metastases upon recurrence. Treatment for newly diagnosed medulloblastoma relies on a combination of maximal safe surgical resection, followed by chemotherapy and ionizing radiation, and no therapies have been shown to confer a survival benefit at the time of recurrence. Given the limited therapeutic options available for patients with medulloblastoma, especially at recurrence, and the incomplete understanding of the molecular mechanisms underlying resistance to treatment, we sought to uncover actionable targets and biomarkers that could help refine patient selection and treatment of newly diagnosed medulloblastoma to reduce the risk of recurrence. In clinically relevant mouse models of G3 medulloblastoma, CT-guided fractionated radiotherapy extended overall survival and induced the clonal selection of radioresistant subpopulations of tumor cells that drove medulloblastoma recurrence. Comparison of recurrent tumors with treatment-naïve newly diagnosed tumors revealed a gene expression signature that was found to be a biomarker of radioresistance and poor prognosis. This prognostic gene signature was shown to be subgroup specific in a large patient cohort. Recurrent tumors had elevated expression of carbonic anhydrase 4, and genetic and pharmacologic modulation of carbonic anhydrase 4 could promote or reduce resistance to radiotherapy. These data suggest that the FDA-approved carbonic anhydrase inhibitor acetazolamide may be a useful radiosensitizer to improve the efficacy of the treatment of newly diagnosed G3 medulloblastoma that could reduce the risk of tumor recurrence and improve survival in pediatric patients.

Significance: G3 medulloblastoma features a prognostic subgroup-specific gene expression signature and can be targeted with a carbonic anhydrase inhibitor to enhance radiosensitivity, reducing the risk of recurrence and improving survival.

Figure 1.

Modeling radiation resistance in G3 medulloblastoma. A, Schematic depicting the protocol for CSI prescribing 24 Gy or 36 Gy at 2 Gy per fraction in mice engrafted with G3 medulloblastoma PDXs. B–E, Kaplan–Meier survival curves for control sham- (black), 24 Gy (blue)–, and 36 Gy (red)–irradiated mice engrafted with GFP-luciferase–tagged Med-114FH, Med-211FH, and Med-411FH PDXs. MB002SU was used to assess efficacy in a treatment-refractory tumor. Comparisons of survival among treated (36 Gy and 24 Gy) and sham groups were performed using log-rank (Mantel–Cox) tests. The number of mice per group (n) is indicated below each graph. Significance levels are represented as P < 0.0001. F, Uniform Manifold Approximation and Projection (UMAP) of single-cell expression data from sham- and 24 Gy–irradiated PDX tumors Med-114FH, Med-211FH, and Med-411FH. Object consists of greater than 50,000 cells harvested from six mice per replicate. G, Heatmap depicting the top 50 commonly differentially expressed genes when comparing sham- to 24 Gy–irradiated PDXs.

Figure 2.

Derivation of a G3-specific recurrent transcriptional signature. A, Venn diagram depicting the overlap of genes significantly upregulated in the 24 Gy–treated recurrent tumors of mice implanted with Med-114FH, Med-211FH, and Med-411FH. The five overlapping genes that constitute a G3 recurrent signature are listed. B, Violin plots depicting the normalized gene expression of PRDX1, PRDX4, UBE2S, UCHL1, and NPTX1 across the different PDXs for both sham and 24 Gy–irradiated mice. A Student t test was performed, with P value <2 × 10⁻¹⁶. C, Normalized expression of the G3 recurrent signature across sham- and 24 Gy–irradiated PDXs is depicted. P values for intergroup comparisons are presented. A Student t test was performed, with P value <2 × 10⁻¹⁶. D, Violin plot depicting the expression of the G3 recurrent signature gene score across different medulloblastoma subgroups and normal cerebellum (CB). A Wilcoxon test was performed. P values for intergroup comparisons are presented and listed above each comparison group. E, Kaplan–Meier survival curves of different medulloblastoma subgroups stratified according to the expression level of the G3 recurrent signature. A high signature score is only associated with poor overall survival in G3 patients, demonstrating the specificity of this prognostic measure. A log-rank (Mantel–Cox) test was performed. F, Schematic of the CRISPR-mediated transcriptional activation (CRISPRa) workflow. D425-Med Cas9-VPR⁺ cells were transduced using a custom 435 sgRNA library and engrafted followed by a low-dose irradiation schedule (3 × 2 Gy) delivered focally to the primary tumor mass to permit tumor recurrence and the emergence of radioresistant cells for next-generation sequencing. G, Plot depicting the enrichment of guides from the CRISPRa screen. The y-axis indicates the log₂-fold change (LFC) of guides enriched in irradiated mice relative to sham controls, and the x-axis shows the FDR. Genes are labeled for the most enriched guides, and positively selected genes constituting the G3 recurrent signature are labeled in orange. BLI, bioluminescence imaging; MOI, multiplicity of infection; SHH, sonic-hedgehog; WNT, wingless.

Figure 3.

Integrative pathway enrichment analysis and druggability of significantly upregulated genes. A, Enrichment map of biological pathways and processes found to be upregulated in 24 Gy–treated samples using ActivePathways and visualized using Cytoscape. Nodes in the network represent similar pathways with gene interactions visualized through connecting lines. Nodes are colored by the PDX for which the pathway is enriched. Those pathways which are significantly enriched through the integration of multiple PDXs are depicted in green. B, Volcano plot showing the log₂-fold change (x-axis) and FDR (y-axis) of significantly upregulated genes with overlapping enrichment. Fold change was calculated as the mean across all three PDXs and the FDR is derived from the adjusted P value after Brown P value integration. The druggability of genes was determined using the DrugBank and is depicted by the red nodes. C, Table showing the top significantly upregulated and druggable genes. Genes are ranked by FDR and mean fold change. Corresponding approved drugs for each gene are indicated. D, Pearson correlations of CA4 and the G3 recurrent signature for sham- and 24 Gy–irradiated PDXs. E, Experimental design for orthotopic in vivo xenotransplantation studies using a CA4-overexpressing construct in G3 medulloblastoma PDX cell lines. Mice were intracranially engrafted with equal parts BFP control andmCherry-CA4–overexpressing cells and focally irradiated at the primary tumor site using a low-dose fractionated treatment schedule (3 × 2 Gy). F, Flow cytometry analysis of xenotransplanted Med-411FH cells expressing BFP^+ve control or mCherry^+veCA4 overexpression constructs. G and H, Tumor cell fractions 2 weeks after engraftment are depicted showing the significant enrichment of CA4-overexpressing cells after RT (sham, n = 3; RT, n = 6; G) and insignificant change using an mCherry^+ve control harvested 2 days after RT (H). A Student t test was performed, with P values listed above each comparison group. I and J, Repeating this approach for D425-Med revealed the same enrichment for CA4-overexpressing cells (sham, n = 3; RT, n = 5), and the use of an enzymatically defective CA4isoform (R219S) abolished this enrichment (sham, n = 2; RT, n = 3; K). A Student t test was performed, with P values listed above each comparison group.

Figure 4.

Radiosensitization of G3 medulloblastoma through the inhibition of CA4. A, Design of in vivo orthotopic xenotransplantation studies and treatment schedule for CA inhibitors. Mice were administered using fractionated (6 × 2Gy) CSI and daily drug treatments (two doses of 50 mg/kg) of a select inhibitor or a polyethylene glycol control. The specificity of ACTZ and METHA is depicted by the concentration of drug required to achieve the targeted inhibition of each CA isozyme. B, Quantification of the average bioluminescence (BLI) signal from each treatment group. Relative luminescence as quantified through normalization to pre-irradiated (day 0) mice for Med-114FH, Med-211FH, Med-411FH, and MB002SU PDX cell lines. A two-sided one-way ANOVA followed by the Tukey post hoc test was used, with P values listed where significant. C, Representative image of the bioluminescence signal observed at the start of treatment (day 0) and after the last treatment (day 15) in a responder and nonresponder PDX cell line. D, Violin plot quantifying the G3 recurrent signature across treatment-naïve PDX lines after RT with vehicle (Veh.) or ACTZ treatment. A Wilcoxon test was performed, with P values listed above each comparison group. E, Biological pathways found to be enhanced in both CA4-high PDX cell lines (MB002SU and Med-114FH) after RT with ACTZ- vs. vehicle-treated samples. Nodes are colored by the PDX in which the pathway is enriched or labeled as integrated for pathways detected from the merged signal of both PDXs but neither alone.

Figure 5.

Inhibition of CA4 diminishes RT resistance. A, Kaplan–Meier survival curves of G3 engrafted mice using different therapeutic regimens. Overall survival is presented for Med-411FH treated with vehicle + RT (n = 15), ACTZ + RT (n = 19), METHA + RT (n = 13), vehicle (n = 15), ACTZ (n = 13), and METHA (n = 9) without RT. B, Overall survival for Med-211FH treated with vehicle + RT (n = 10), ACTZ + RT (n = 10), METHA + RT (n = 10), vehicle (n = 5), ACTZ (n = 5), and METHA (n = 4) without RT. C, Overall survival for Med-114FH treated with vehicle + RT (n = 10), ACTZ + RT (n = 10), METHA + RT (n = 6), vehicle (n = 7), ACTZ (n = 7), and METHA (n = 7) without RT. D, MB002SU treated with vehicle + RT (n = 10), ACTZ + RT (n = 10), METHA + RT (n = 10), vehicle (n = 8), ACTZ (n = 8), and METHA (n = 8) without RT. A log-rank (Mantel–Cox) test was performed, with P values listed where significant. E, Uniform Manifold Approximation and Projection (UMAP) of single-cell expression data from irradiated (12 Gy) mice treated with ACTZ or vehicle contrasting CA4-high and -low PDXs (Med-114FH and Med-411FH, respectively). Comparison of the treated clusters reveals a distinct transcriptional change after ACTZ treatment with variable clustering between CA4-high– and CA4-low–expressing tumors. F, Heatmap reporting the expression of discriminative gene sets between CA4-high (Med-114FH) and -low (Med-411FH) PDXs filtered using vehicle controls. Distinct transcriptional regulatory programs are apparent after ACTZ treatment, with similar baseline expression noted in the vehicle controls. G, Schematic summarizing the transcriptional changes that occur in G3 medulloblastoma after RT and the inhibition of CA4 as a novel therapeutic strategy. After irradiation, G3 medulloblastoma expresses a subgroup-specific gene signature accompanied by an increase in CA4promoting radioresistance. Treatment using the FDA-approved drug ACTZ decreases tumor burden and increases survival for CA4-high–expressing tumors, presenting CA4 as a novel predictive biomarker. SHH, sonic-hedgehog; WNT, wingless.