Large-scale multiomic analysis identifies non-coding somatic driver mutations and nominates ZFP36L2 as a driver gene for pancreatic ductal adenocarcinoma

Background The identification and characterisation of somatic cancer driver mutations in the non-coding genome remains challenging. Objective To broadly characterise non-coding driver mutations for pancreatic ductal adenocarcinoma (PDAC). Design Using mutation calls from whole-genome sequence data in PDACs and genome-scale maps of accessible gene regulatory regions in normal-derived and tumour-derived pancreatic samples, we analysed enrichment of non-coding mutations in gene regulatory regions relevant to normal-derived and tumour-derived pancreatic contexts. Functional follow-up of potential driver mutations was performed using chromatin interaction analyses, massively parallel reporter assays (MPRA) and targeted analysis of selected non-coding somatic mutations (NCSMs). Results We first created genome-scale maps of accessible chromatin regions (ACRs) and histone modification marks (HMMs) in pancreatic cell lines and purified pancreatic acinar and duct cells. Integration with whole-genome mutation calls from 506 PDACs revealed 314 ACRs/HMMs significantly enriched with 3614 NCSMs. Chromatin interaction analysis identified 416 potential target genes and MPRA revealed 178 NCSMs impacting reporter activity (19.45% of those tested). Targeted luciferase validation confirmed negative effects on gene regulatory activity for NCSMs near ZFP36L2 and CDKN2A . For the former, CRISPR interference identified ZFP36L2 as a target gene (16.0–24.0% reduced expression, p=0.023–0.0047), and growth inhibition after overexpression of ZFP36L2 (4.1–14.1-fold reduction, p=6.0×10 –4 − 3.2×10 –3 ) implicates a possible tumour suppressor function. Conclusion Our integrative approach provides a catalogue of potential non-coding driver mutations and nominates ZFP36L2 as a novel PDAC driver gene with a likely tumour suppressor function.