TY - JOUR T1 - Quantitative single cell heterogeneity profiling of patient derived tumor initiating gliomaspheres reveals unique signatures of drug response and malignancy JF - bioRxiv DO - 10.1101/2020.01.14.900506 SP - 2020.01.14.900506 AU - Michael Masterman-Smith AU - Nicholas A. Graham AU - Ed Panosyan AU - Jack Mottahedeh AU - Eric E. Samuels AU - Araceli Nunez AU - Sung Hyun Lim AU - Tiffany Phillips AU - Meeryo Choe AU - Koppany Visnyei AU - William H. Yong AU - Thomas G. Graeber AU - Ming-Fei Lang AU - Harley I. Kornblum AU - Jing Sun Y1 - 2020/01/01 UR - http://biorxiv.org/content/early/2020/01/14/2020.01.14.900506.abstract N2 - Background Glioblastoma is a deadly brain tumor with median patient survival of 14.6 months. At the core of this malignancy are rare, highly heterogenous malignant stem-like tumor initiating cells. Aberrant signaling across the EGFR-PTEN-AKT-mTOR signal transduction pathways are common oncogenic drivers in these cells. Though gene-level clustering has determined the importance of the EGFR signaling pathway as a treatment indicator, multiparameter protein-level analyses are necessary to discern functional attributes of signal propagation. Multiparameter single cell analyses is emerging as particularly useful in identifying such attributes.Methods Single cell targeted proteomic analysis of EGFR-PTEN-AKT-mTOR proteins profiled heterogeneity in a panel of fifteen patient derived gliomaspheres. A microfluidic cell array ‘chip’ tool served as a low cost methodology to derive high quality quantitative single cell analytical outputs. Chip design specifications produced extremely high signal-to-noise ratios and brought experimental efficiencies of cell control and minimal cell use to accommodate experimentation with these rare and often slow-growing cell populations. Quantitative imaging software generated datasets to observe similarities and differences within and between cells and patients. Bioinformatic self-organizing maps (SOMs) and hierarchical clustering stratified patients into malignancy and responder groups which were validated by phenotypic and statistical analyses.Results Fifteen patient dissociated gliomaspheres produced 59,464 data points from 14,866 cells. Forty-nine molecularly defined signaling phenotypes were identified across samples. Bioinformatics resolved two clusters diverging on EGFR expression (p = 0.0003) and AKT/TORC1 activation (p = 0.08 and p = 0.09 respectively). TCGA status of a subset showed genetic heterogeneity with proneural, classical and mesenchymal subtypes represented in both clusters. Phenotypic validation measures indicated drug responsive phenotypes to EGFR blocking were found in the EGFR expressing cluster. EGFR expression in the subset of drug-treated lines was statistically significant (p<.05). The EGFR expressing cluster was of lower tumor initiating potential in comparison to the AKT/TORC1 activated cluster. Though not statistically significant, EGFR expression trended with improved patient prognosis while AKT/TORC1 activated samples trended with poorer outcomes.Conclusions Quantitative single cell heterogeneity profiling resolves signaling diversity into meaningful non-obvious phenotypic groups suggesting EGFR is decoupled from AKT/TORC1 signalling while identifying potentially valuable targets for personalized therapeutic approaches for deadly tumor-initiating cell populations.TBD ER -