Poly-Ligand Profiling differentiates pancreatic cancer patients according to treatment benefit from gemcitabine+placebo versus gemcitabine+evofosfamide and identifies candidate targets

Introduction: The accumulation of a multitude of subtle molecular aberrations during tumor progression limit the efficacy of anti-cancer drugs. A vast array of these variations can be assessed with Poly-Ligand Profiling (PLP), which is utilizing libraries of trillion unique ssDNA with aptamer binding properties. The aims of this study were to develop a PLP library that differentiate pancreatic cancer patients who can benefit from gemcitabine+evofosfamide (GE) or gemcitabine+placebo (G) and identify its molecular targets. Methods: Patients: locally advanced or metastatic pancreatic cancer patients randomized to G vs GE in the unsuccessful phase III MAESTRO trial (Threshold Pharmaceuticals, Merck KgaA). FFPE tissues of patients with good (OS > 13 mos) or poor (OS < 7 mos) outcome from GE were used for PLP library development. Affinity maturation and testing of library for binding FFPE tissue is done with IHC-like protocol. Assay conditions and algorithm were locked based on the training set (n = 12) and used for testing assay performance in the blinded set (n = 172, primary and metastatic sites). PLP-assay performance metrics from blinded test set served to estimate the impact on the MAESTRO study (n = 693) by performing 1000 simulations. For target ID, FFPE tissue of patients with poor outcome, stained with enriched library, was recovered, lysed, underwent affinity-based pull-downs, purified with PAGE gel and subjected to high resolution mass-spectrometry (MS). Results: 1,000 simulations of projected PLP-positive patients from MAESTRO study revealed a median OS increase of 37.6% (mean) in G+E cohort, compared to G (17.4% OS increase in MAESTRO) with mean Hazard Ratio (HR) 0.72 (0.84 in MAESTRO). 96.9% of simulated trials achieved statistical significance. For primary tumor samples the median OS increase for G+E patients was 53.4% with mean HRs of 0.64 with 100% of trials exhibiting log-rank p < 0.05. MS reliably detected 20 proteins, 11 of which have reported associations with pancreatic cancer and 6 have been associated with resistance to gemcitabine: vimentin (VIM), pyruvate kinase (PKM), endoplasmic reticulum chaperone BiP (HSPA5), heat shock protein HSP 90-alpha (HSP90AA1), Histone H3-1 (HIST1H3A), heat shock protein beta-1 (HSPB1). Vimentin is a mesenchymal marker whose expression increases during epithelial–to-mesenchymal transition (EMT) and tumor progression. EMT results in the suppression of human equilibrative/concentrative nucleoside transporter and protects tumor cells from gemcitabine. GRP78 overexpression confers resistance to gemcitabine and its knockdown sensitizes tumor cells to drug treatment. Alternative splicing of PKM promotes gemcitabine resistance in pancreatic cancer cells most likely by boosting glycolysis-fueled proliferation. Heat shock proteins regulate multiple tumor survival and progression pathways and their inhibition attenuates resistance of cancer cells to gemcitabine. Pancreatic tumors demonstrate increased histones acetylation, which was correlating with increased protection against gemcitabine. Further characterization of these candidate targets is ongoing. Conclusion: PLP is a novel platform for classifying pancreatic cancer patients according to their benefiting from GE treatment. MS of the PLP library pull-downs reveals targets associated with gemcitabine resistance. In principle, the novel PLP platform could be applied to different therapeutic regimen for the development of urgently needed companion diagnostic tests in cancer and other diseases.