A 4-miRNA signature predicts the therapeutic outcome of glioblastoma

Abstract

Multimodal therapy of glioblastoma (GBM) reveals inter-individual variability in terms of treatment outcome. Here, we examined whether a miRNA signature can be defined for the a priori identification of patients with particularly poor prognosis.FFPE sections from 36 GBM patients along with overall survival follow-up were collected retrospectively and subjected to miRNA signature identification from microarray data. A risk score based on the expression of the signature miRNAs and cox-proportional hazard coefficients was calculated for each patient followed by validation in a matched GBM subset of TCGA. Genes potentially regulated by the signature miRNAs were identified by a correlation approach followed by pathway analysis.A prognostic 4-miRNA signature, independent of MGMT promoter methylation, age, and sex, was identified and a risk score was assigned to each patient that allowed defining two groups significantly differing in prognosis (p-value: 0.0001, median survival: 10.6 months and 15.1 months, hazard ratio = 3.8). The signature was technically validated by qRT-PCR and independently validated in an age- and sex-matched subset of standard-of-care treated patients of the TCGA GBM cohort (n=58). Pathway analysis suggested tumorigenesis-associated processes such as immune response, extracellular matrix organization, axon guidance, signalling by NGF, GPCR and Wnt. Here, we describe the identification and independent validation of a 4-miRNA signature that allows stratification of GBM patients into different prognostic groups in combination with one defined threshold and set of coefficients that could be utilized as diagnostic tool to identify GBM patients for improved and/or alternative treatment approaches.

Keywords: glioblastoma; miRNA; signature.

Figure 1. Extraction of a 4-miRNA signature as independent predictive marker for the overall survival of GBM patients in the exploratory cohort

A. Kaplan-Meier overall survival analyses of high-risk and low-risk GBM patients. High-risk and low-risk patients were stratified based on the risk factors calculated from the cox-proportional hazard coefficients and the miRNA expression levels as measured in the microarray (left panel, 35 patients) or by qRT-PCR analyses (right panel, 19 patients). Hazard ratios and p-values were calculated by log-rank test. B. Overall survival (top panel), hierarchical cluster heat map of miRNA array expression levels (middle panel), and risk factors calculated on the basis of miRNA expression values and cox-proportional hazard coefficients (bottom panel) for all patients. miRNAs hsa-let-7a-5p, hsa-let-7b-5p and hsa-miR-125a-5p in patients of the higher-risk group show a tendency towards lower expression and that of hsa-miR-615-5p a tendency towards higher expression. The median risk factor value was used to classify high-risk and low-risk patients. C. Distribution of age (left panel) and sex (right panel) in high-risk and low-risk GBM patients. Statistical comparison was performed by Student's t-test or Fisher's exact test. The patients of the lower-risk group were statistically significantly older compared with that of the lower-risk group. The differences in the numbers of male and female patients of the lower- and higher-risk groups were not statistically significant.

Figure 2. Evaluation of the prognostic value of the extracted 4-miRNA signature in an age- and sex-matched subgroup of standard-of-care treated patients of the TCGA GBM dataset

A. Age distribution in the exploratory cohort and the TCGA GBM cohort before and after age matching. B. Overall survival (top panel), hierarchical cluster heat map of miRNA expression levels (middle panel), and risk factors for patients of the age- and sex-matched TCGA GBM cohort. The median risk factor value was used to classify high-risk and low-risk patients. miRNAs hsa-let-7a-5p, hsa-let-7b-5p and hsa-miR-125a-5p in patients of the higher-risk group show a slight tendency towards lower expression and that of hsa-miR-615-5p a slight tendency towards higher expression. C. Kaplan-Meier overall survival analyses of high-risk and low-risk standard-of-care treated patients of the age- and sex-matched TCGA GBM cohort. Classification of high-risk and low-risk patients was performed on the basis of the risk factors calculated from the cox-proportional hazard coefficients (Table 2) and the miRNA expression levels. Hazard ratios and p-values were calculated by log-rank test. D. Distribution of age (left panel) and sex (right panel) in high-risk and low-risk patients of the age- and sex-matched TCGA GBM cohort. Student's t-test and Fisher's exact test were employed for statistical comparison as depicted.

Figure 3. Heatmaps of the gene expressions correlating with the 4 miRNAs hsa-let-7b-5p, hsa-miR-125a-5p, hsa-miR-615-5p and hsa-let-7a-5p the age- and sex-matched TCGA GBM cohort of standard-of-care treated patients

Genes whose expression levels statistically significantly correlated (p < 0.01) with the respective miRNA expression levels are shown. Every column represents an individual patient. Data are ordered from left to right by increasing miRNA expression.