Gene signatures associated with prognosis and chemotherapy resistance in glioblastoma treated with temozolomide

Abstract

Background: Glioblastoma (GBM) prognosis remains extremely poor despite standard treatment that includes temozolomide (TMZ) chemotherapy. To discover new GBM drug targets and biomarkers, genes signatures associated with survival and TMZ resistance in GBM patients treated with TMZ were identified. Methods: GBM cases in The Cancer Genome Atlas who received TMZ (n = 221) were stratified into subgroups that differed by median overall survival (mOS) using network-based stratification to cluster patients whose somatic mutations affected genes in similar modules of a gene interaction network. Gene signatures formed from differentially mutated genes in the subgroup with the longest mOS were used to confirm their association with survival and TMZ resistance in independent datasets. Somatic mutations in these genes also were assessed for an association with OS in an independent group of 37 GBM cases. Results: Among the four subgroups identified, subgroup four (n = 71 subjects) exhibited the longest mOS at 18.3 months (95% confidence interval: 16.2, 34.1; p = 0.0324). Subsets of the 86 genes that were differentially mutated in this subgroup formed 20-gene and 8-gene signatures that predicted OS in two independent datasets (Spearman's rho of 0.64 and 0.58 between actual and predicted OS; p < 0.001). Patients with mutations in five of the 86 genes had longer OS in a small, independent sample of 37 GBM cases, but this association did not reach statistical significance (p = 0.07). Thirty-one of the 86 genes formed signatures that distinguished TMZ-resistant GBM samples from controls in three independent datasets (area under the curve ≥ 0.75). The prognostic and TMZ-resistance signatures had eight genes in common (ANG, BACH1, CDKN2C, HMGA1, IFI16, PADI4, SDF4, and TP53INP1). The latter three genes have not been associated with GBM previously. Conclusion: PADI4, SDF4, and TP53INP1 are novel therapy and biomarker candidates for GBM. Further investigation of their oncologic functions may provide new insight into GBM treatment resistance mechanisms.

Keywords: chemotherapy resistance; glioblastoma; oncogene; somatic variant; temozolomide.