Tumor Heterogeneity in Gliomas – a Histopathology-Targeted Proteomic Pilot Study

Abstract

Intra- and inter-tumor heterogeneity underlies both tumor evolution and response to treatment. Our aim was to explore the protein expression of infiltrative gliomas (grades II to IV) with high throughput protein-based mass spectrometry (MS), comparing different histopathologic areas of glioblastomas (GBM) and lower grade gliomas (LGG), to further characterize gliomas pathways, and identify druggable targets and protein signatures to be used as diagnostic and prognostic biomarkers. We studied 12 patients with diffuse gliomas (1 diffuse astrocytoma, 1 anaplastic astrocytoma, 3 anaplastic oligodendrogliomas, and 7 glioblastomas). Each tumor was reclassified according to the 2016 WHO classification of tumors of CNS (IDHmutant-1p19q Codeleted; IDHmutant-1p19q intact; IDH wild type glioma), and mapped using H&E stained slides to identify 2-3 areas with different histopathologic phenotypes, different cellularity, and perinecrotic areas. The selected areas with at least 80% of tumor cells were cored with 1.0mm needle using a semi-automated tissue microarray platform, in a total of 35 samples – hemorrhagic and necrotic portions of the tumors were avoided in all cases. Total protein was extracted from the FFPE tissue and analyzed with 4 hour liquid chromatography tandem MS (LC-MS-MS) for label-free expression proteomics. We compared the average of protein expression in different tumor areas from each patient to identify associations with Wilcoxon rank-sum test. The results were further interpreted following a top-bottom approach and compared in silico with TCGA expression data (mRNA). The statistical analysis was performed using R software. Overall, 9,222 peptides were mapped to 1758 non-redundant proteins in the samples, 320 of which had a significant differential expression in GBM versus LGG. The PCA analysis showed that according to the molecular signatures, the samples clustered well by disease and by patient. Samples also clustered by IDH1 status and WHO subgroup. One patient had non-neoplastic brain adjacent to a GBM available for analysis and this sample clustered with the LGGs. Among the top 35 proteins most differentially expressed in the 2 groups (raw p-value= 0.00252; FDR=0.12684), 14 had higher expression in GBM, and 21 in LGG. This included one protein previously explored by our group in previous studies (TAGLN2), validating our findings with different approach. Our results showed that LC-MS-MS analysis of morphologically different glioma areas is feasible and was able to identify clusters with different patterns of protein expression. Further correlation of proteome profiles with clinical data, histological features, pathway analysis, and druggable target analysis are ongoing. FUNDING: R01CA108633, R01CA169368, RC2CA148190, U10CA180850-01 (NCI), Brain Tumor Funders Collaborative Grant, and the Ohio State University CCC (all to AC).