Differentiation Stage Predicts Radiosensitivity in Mesenchymal-Like Pancreatic Cancer.

Abstract

To derive a genomic classifier to predict radiosensitivity of pancreatic cancer cell lines and patients with pancreatic cancer to allow genomic-guided radiation therapy. We collected a comprehensive data set of full clonogenic cell survival curves of 45 pancreatic cancer cell lines irradiated with clinical photon and proton beams. We derived classifiers based on data from human embryonic and fetal pancreas single-cell RNA-sequencing to distinguish between epithelial and mesenchymal cells and to predict pancreas cell-line differentiation stage. Independent testing was done with an embryonic mouse pancreas single-cell RNA-sequencing data set. We then used bulk RNA-seq profiles from the Cancer Cell Line Encyclopedia to classify our pancreatic cancer cell lines using our epithelial-mesenchymal and differentiation stage classifiers. We then correlated the differentiation stage classifier with the radiosensitivity of the pancreatic cancer cell lines as well as with pancreatic cancer patient data from The Cancer Genome Atlas. We found wide variability in radiosensitivity to both photons and protons among pancreatic cancer cell lines. We showed that the differentiation stage is predictive of radiosensitivity of mesenchymal pancreatic cancer cell lines but not epithelial pancreatic cancer cell lines. We found that chromatin compaction is associated with the differentiation stage and showed that the less differentiated mesenchymal pancreatic cancer cell lines tend to be radioresistant and with more compact chromatin than the radiosensitive differentiated cell lines. Patients with more differentiated tumors exhibit better overall survival. We found that mesenchymal-like undifferentiated pancreatic cancer cell lines are more radioresistant than mesenchymal-like differentiated ones and that patients with pancreatic cancer with mesenchymal-like undifferentiated tumors treated with radiation therapy tend to have lower overall survival compared with patients with mesenchymal-like differentiated tumors. We show that it is feasibility to use the differentiation stage of mesenchymal pancreatic cancer cells to predict tumor specific radiosensitivity.