Purpose: We aimed to develop a diagnostic platform to capture the transcriptomic resemblance of individual adult diffuse gliomas of WHO grades II to IV to neural development and the genomic signature associated with glioma progression.
Experimental design: Based on the EM/PM classification scheme, we designed a RT-PCR-based TaqMan low-density array (TLDA) containing 44 classifier and 4 reference genes. Samples of a training dataset (GSE48865), characterized by RNA-sequencing, were utilized to optimize the TLDA design and to develop a support vector machine (SVM)-based prediction model. Complemented with Sanger sequencing for IDH1/2, and low coverage whole genome sequencing (WGS), the TLDA and SVM prediction model were tested in a validation (31 gliomas) and a test (121 gliomas) dataset.
Results: Independent of morphologically defined subtypes and grades, gliomas can be individually assigned into the EM and PM glioma subtypes with the respective areas under ROC curves at 0.86 and 0.85 in the validation dataset. The EM gliomas showed a medium overall survival (OS) of 15.6 months, whereas the medium OS for PM gliomas was not reached (HR = 3.55; 95% confidence interval, 1.96-6.45). The EM and PM gliomas showed distinct patterns of genomic alterations, with IDH mutation and 1p19q codeletion in the PM gliomas and gain of chromosome 7/loss of chromosome 10 in the EM gliomas. Extensive chromosomal abnormalities marked the progression of PM gliomas.
Conclusions: The integration of EM/PM subtyping, IDH sequencing, and low coverage WGS may improve the risk stratification and selection of treatment regimens for patients with glioma.
©2019 American Association for Cancer Research.