A novel patient stratification strategy to enhance the therapeutic efficacy of dasatinib in glioblastoma

Abstract

Background: Glioblastoma is the most common primary malignancy of the central nervous system with a dismal prognosis. Genomic signatures classify isocitrate dehydrogenase 1 (IDH)-wildtype glioblastoma into three subtypes: proneural, mesenchymal, and classical. Dasatinib, an inhibitor of proto-oncogene kinase Src (SRC), is one of many therapeutics which, despite promising preclinical results, have failed to improve overall survival in glioblastoma patients in clinical trials. We examined whether glioblastoma subtypes differ in their response to dasatinib and could hence be evaluated for patient enrichment strategies in clinical trials.

Methods: We carried out in silico analyses on glioblastoma gene expression (TCGA) and single-cell RNA-Seq data. In addition, in vitro experiments using glioblastoma stem-like cells (GSCs) derived from primary patient tumors were performed, with complementary gene expression profiling and immunohistochemistry analysis of tumor samples.

Results: Patients with the mesenchymal subtype of glioblastoma showed higher SRC pathway activation based on gene expression profiling. Accordingly, mesenchymal GSCs were more sensitive to SRC inhibition by dasatinib compared to proneural and classical GSCs. Notably, SRC phosphorylation status did not predict response to dasatinib treatment. Furthermore, serpin peptidase inhibitor clade H member 1 (SERPINH1), a collagen-related heat-shock protein associated with cancer progression, was shown to correlate with dasatinib response and with the mesenchymal subtype.

Conclusion: This work highlights further molecular-based patient selection strategies in clinical trials and suggests the mesenchymal subtype as well as SERPINH1 to be associated with response to dasatinib. Our findings indicate that stratification based on gene expression subtyping should be considered in future dasatinib trials.

Keywords: SERPINH1; SRC; dasatinib; glioblastoma; glioblastoma stem-like cells.

Fig. 1

The glioblastoma MES subtype is more sensitive to dasatinib treatment than the PN and CL subtypes in vitro. A: Histograms of cell cycle assay (left) with cell cycle phases (right) upon SRC knockdown in MES and PN GSCs. *P < 0.05. B: Dose-response curves to 96 h of dasatinib treatment in nine patient-derived GSC line of MES, PN, and CL subtypes. Error bars denote the standard error of the mean (SEM). IC₅₀ was determined as the concentration required to bring the curve down to the point halfway between the top and bottom plateaus of the curve. ANOVA was used to compare groups *P < 0.05, **P < 0.01, (n = 3). C: Propidium iodide staining with flow cytometry after 96 h of dasatinib inhibition at different concentrations in NCH711d (MES) and NCH644 (PN) GSCs. n.s.: not significant, *P < 0.05, (n = 3). D: EdU proliferation assay showing the effect of dasatinib treatment (5 µM) on proliferation of NCH711d (MES) GCSs compared to NCH644 (PN) GCSs after 48 h. **P < 0.01, (n = 3). E: Invasion assay with representative images showing the invasion of NCH711d (MES) or NCH644 (PN) cells through a collagen matrix over 5 days with increasing concentrations of dasatinib. Scale bar = 200 µm. Abbreviations: CL, classical; GSC, glioblastoma stem-like cell; MES, mesenchymal; PN, proneural; ANOVA, analysis of variance.

Fig. 2

Phosphorylation status is not sufficient for determining SRC activation. A: Scheme of the SRC enzyme based on literature. Generally, autophosphorylation at the Y-416 position correlates with SRC activation and phosphorylation at Y-527 with inactivation. B: Western blot showing levels of total and phosphorylated (Y-416 and Y-527) SRC in the four GSC lines. α-Tubulin was used as a loading control. Numbers below indicate the quantified ratio of Y-416/Y-527. Phospho bands were normalized to total SRC and α-tubulin. C: Left: Reverse phase protein array (RPPA) data showing SRC protein expression across different subtypes. Right: Ratio of the Y-416 SRC phosphorylation/Y-527 SRC phosphorylation using RPPA data of the TCGA to determine SRC activation.

Fig. 3

SRC substrate analysis and an expression-based signature reveal higher SRC activation in the MES subtype than PN. A: Western blot showing levels of phosphorylated p130cas (Y-410) compared with total p130cas in NCH705 (MES) and NCH644 (PN) GSCs treated with 5 µM dasatinib. β-Actin was used as a loading control. B: Western blot showing levels of phosphorylated p130cas (Y-410) compared with total p130cas in and NCH644, NCH421k (PN), NCH705, and NCH711d (MES) GSCs. β-Actin was used as a loading control. Numbers below indicate quantified levels of P-p130cas. Phospho bands were normalized to total p130cas and α-tubulin. C: Heatmap with unsupervised clustering analysis of the TCGA GBM expression data patient cohort of IDH-WT tumors reclassified according to the Wang et al. subtype signatures (n = 265), showing two main clustering groups upon calling genes of the expression-based SRC activation signature: a PN-led and an MES-led group. Patient samples are distributed horizontally, genes from gene set are listed vertically. D: Single-sample Gene Set Enrichment Analysis (ssGSEA) scores of TCGA glioblastoma samples calculated based on the gene set applied to 2A (two-sided t-tests ****P < 0.0001). E: ssGSEA SRC activation score of microarray gene expression data of the patient-derived GCSs used for IC₅₀ determination of dasatinib (see Figure 1B). F: ssGSEA scores for single-cell RNA-Seq datal, reclassified according to the Wang subtype signatures. G: p416/p527 phosphorylation ratio (activation) plotted against the SRC pathway activation scores of TCGA patient samples.

Fig. 4

Patients with higher SRC activation have a significantly shorter survival rate. Survival analysis of n = 38 patients from the NCT Neuro Master Match (N²M²) pilot trial. Patients were classified into three groups according to the quartile of SRC signature expression scores (Q1, Q2–3, Q4). Significance testing was performed using a multivariate cox proportional hazards model (significance at P < 0.05).

Fig. 5

Serpin peptidase inhibitor clade H member 1 (SERPINH1) correlates with sensitivity to dasatinib in GSCs. A: Plot showing relationship of SERPINH1 expression with log(IC₅₀) of GSCs with dasatinib based on a regression analysis (Blue = MES, purple = CL, green = PN). B: Box plots showing SERPINH1 expression across the three glioblastoma subtypes in The Cancer Genome Atlas (TCGA) dataset (one-way ANOVA, Tukey’s multiple comparisons test, ***P < 0.001, ****P < 0.0001). C: SERPINH1 expression in glioblastoma tumor samples compared to normal brain tissue, data from (unpaired t-test, P < 0.0001). D: Western blot showing SERPINH1 protein levels in MES and PN patient-derived GSC lines. α-Tubulin was used as a loading control. E: SERPINH1 correlation with single-sample gene set enrichment analysis (ssGSEA) SRC activation scores for each TCGA sample. Box plots of plotted data in different subtypes are provided on both axes. F: Immunohistochemistry (IHC) showing SERPINH1 protein levels in tumor samples with low and high SRC activation scores from the N2M2 clinical trial. Scale bar = 50 µm.