Spatial analysis of IPMNs defines a paradoxical KRT17-positive, low-grade epithelial population harboring malignant features

Background & Aims: Intraductal papillary mucinous neoplasms (IPMNs) are pancreatic cysts that represent one of the few radiologically identifiable precursors to pancreatic ductal adenocarcinoma (PDAC). Though the IPMN-bearing patient population represents a unique opportunity for early detection and interception, current guidelines provide insufficient accuracy in determining which patients should undergo resection versus surveillance, resulting in a sizable fraction of resected IPMNs only harboring low-grade dysplasia, suggesting that there may be overtreatment of this clinical entity. Methods: To investigate the transcriptional changes that occur during IPMN progression, we performed spatial transcriptomics using the Nanostring GeoMx on patient samples containing the entire spectrum of IPMN disease including low-grade dysplasia, high-grade dysplasia, and IPMN-derived carcinoma. Single cell RNA sequencing was performed on side branch and main duct IPMN biospecimens. Results: We identified a subpopulation of histologically low-grade IPMN epithelial cells that express malignant transcriptional features including KRT17, S100A10 and CEACAM5, markers that are enriched in PDAC. We validated this high-risk gene signature in both single-cell RNA sequenced samples and an external ST dataset containing a larger number of IPMN samples including non-tumor bearing IPMN (i.e. low-grade IPMN in isolation). Immunofluorescence staining of a large cohort of patient tissues confirmed the presence of KRT17-positive cells, which were found to comprise a small subset of epithelial cells within histologically low-grade IPMN in a patchy distribution. Conclusions: Our study demonstrates that KRT17 marks a distinct transcriptional signature in a subpopulation of epithelial cells within histologically low-grade IPMN. This population of cells likely represents a transitional state of histologically low-grade epithelial cells undergoing progression to a higher grade of dysplasia and thus may represent a higher risk of progression to carcinoma. Overall design: A total of 3 slides containing tissue sections from pancreatic cysts, associated malignancy, and/or adjacent normal tissue was submitted for spatial transcriptomics. The Solid Tumor TME Morphology Kit was utilized according to manufacturer instructions to detect epithelial cells (pan-cytokeratin [CK]) together with DAPI for nuclei detection. 310 ROIs were captured for whole transcriptome profiling.