Pre-Surgical Identification of Diagnostic, Prognostic and Predictive DNA Methylation-Based Markers in Serum (Liquid Biopsy) of Patients Harboring Gliomas
Noushmehr H, Sabedot TS, Malta TM, Nelson KK, Snyder J, Wells M, Mosella MS, DeCarvalho AC, Asmaro K, Scarpace L, Robin AM, Rosenblum ML, Mikkelsen T, Rock J, Walbert T, Lee I, Poisson LM, Kalkanis SN, and Castro AV. Pre-Surgical Identification of Diagnostic, Prognostic and Predictive DNA Methylation-Based Markers in Serum (Liquid Biopsy) of Patients Harboring Gliomas. Cancer Res 2019; 79(13).
Central nervous system-related tumors release tumoral material into circulating blood and the cerebrospinal fluid (e.g. cell free DNA). The sampling of these biofluids, i.e. liquid biopsy (LB), may offer an opportunity for diagnosis, prognostication and response prediction in a constantly evolving and biologically and prognostically heterogeneous tumor, such as glioma, in real-time. In glioma-tumor tissue, genome-wide DNA methylation profiling has shown that epigenetic abnormalities play significant biological and clinical roles, making DNA methylome profiling attractive for LB application in these tumors. Thus far, studies of epigenetic LB (eLB) focused on targeted markers which have shown low sensitivity; however, this can be potentially circumvented by a comprehensive genome-wide CpG methylation profiling. Herein, we profiled the genome-wide CpG methylation landscape of matching serum/tissue from 22 patients who received surgical resection for a glioma diagnosis (15 IDH-mutant and 7 IDH-wildtype) and 4 who received surgical resection of the brain for a non-tumor brain related disease. We identified 199 glioma specific DNA methylation-serum based markers (Wilcoxon Rank Sum test, p-value < 0.001) that differentiated glioma from non-tumor brain tissues (diagnostic eLB). These eLB diagnostic markers were found to be enriched for CpG islands and depleted for open seas and shores. Interestingly, CpG methylation of MYC and CD34 promoters, previously described in the tissue, were detectable in the serum of glioma patients as part of the 199 CpG eLB signature. We also identified 987 eLB markers (Wilcoxon Rank Sum test, p-value < 0.01) that discriminated patients with IDH-mutant from IDH-wildtype (prognostic eLB). Among the initial cohort, comprised by 4 MGMT-unmethylated and 18 MGMT-methylated gliomas, we also identified 428 specific eLB markers that discriminated the MGMT status among the patients (predictive eLB). Harnessing DNA methylation data of The Cancer Genome Atlas (TCGA) consortium, derived from 10,000 primary and untreated tumor tissue samples, spanning 33 cancer types, we found our three eLB signatures (diagnostic, prognostic and predictive) to be highly specific to gliomas. Our results suggest that serum eLB profiling may be useful as a surrogate or complementary for tissue-based approach for diagnosis, prognostication and treatment prediction of gliomas. In addition, our eLB signatures can be applied as a real-time non-invasive approach to improve detection of glioma progression and recurrence. Once validated, the application of the eLB panels discovered in this study have the potential to significantly and positively improve the pre- and post-surgical quality of care for patients harboring gliomas.