Multivariate analysis reveals differentially expressed genes among distinct subtypes of diffuse astrocytic gliomas: diagnostic implications

Diagnosis and classification of gliomas mostly relies on histopathology and a few genetic markers. Here we interrogated microarray gene expression profiles (GEP) of 268 diffuse astrocytic gliomas—33 diffuse astrocytomas (DA), 52 anaplastic astrocytomas (AA) and 183 primary glioblastoma (GBM)—based on multivariate analysis, to identify discriminatory GEP that might support precise histopathological tumor stratification, particularly among inconclusive cases with II–III grade diagnosed, which have different prognosis and treatment strategies. Microarrays based GEP was analyzed on 155 diffuse astrocytic gliomas (discovery cohort) and validated in another 113 tumors (validation set) via sequential univariate analysis (pairwise comparison) for discriminatory gene selection, followed by nonnegative matrix factorization and canonical biplot for identification of discriminatory GEP among the distinct histological tumor subtypes. GEP data analysis identified a set of 27 genes capable of differentiating among distinct subtypes of gliomas that might support current histological classification. DA + AA showed similar molecular profiles with only a few discriminatory genes overexpressed (FSTL5 and SFRP2) and underexpressed (XIST, TOP2A and SHOX2) in DA vs AA and GBM. Compared to DA + AA, GBM displayed underexpression of ETNPPL, SH3GL2, GABRG2, SPX, DPP10, GABRB2 and CNTN3 and overexpression of CHI3L1, IGFBP3, COL1A1 and VEGFA, among other differentially expressed genes.