Abstract PR012: Perioperative liver biopsy captures features of the liver pre-metastatic niche and predicts metastatic outcome after pancreatic cancer resection

Pancreatic cancer (PC) has a high propensity for liver metastasis (LM), a rapidly lethal event, often seen early after resection of the primary tumor. Mounting pre-clinical evidence implicates the establishment of microenvironmental alterations in the liver or other target organs prior to emergence of clinically evident metastases, termed “pre-metastatic niches”, which is a crucial first step in metastatic progression. To determine the translational relevance of these findings, we evaluated liver biopsies from PC patients obtained during pancreatectomy to characterize the cellular, molecular and metabolic features that define the pre-metastatic niche and that may, in turn, be used as biomarkers to stratify the metastatic risk, and as therapeutic targets to interrupt the metastatic cascade. Thus, liver biopsies from 49 patients with localized PC were analyzed by transcriptomics, metabolomics, histopathology, flow cytometry, and in selected patients by single cell RNA sequencing. Patients underwent routine surveillance in a prospective manner to determine which patients developed early (6 months) LM, extrahepatic metastasis, or remained disease-free. Eighteen patients with non-cancerous pancreatic lesions undergoing pancreatectomy were used as controls (non-PC). Single cell transcriptomics analyses and immuno-profiling revealed that, in contrast to non-PC livers, tumor-free, pre-metastatic livers from PC patients exhibited enhanced inflammation evidenced by enrichment of neutrophils, CD11B+ cells, and CD3+ T cells, as well as expansion and activation of a CD11B+ subset of NK cells. Furthermore, immune and metabolic profiling of PC pre-metastatic livers identified features that predicted future patterns of metastasis: high Ki67 proliferation index, neutrophil extracellular trap (NET) formation, and upregulated sortilin-1 gene expression, all of which correlated with earlier LM. Conversely, widespread lobular inflammation, with abundant lobular CD3+ T cells, as well as intact creatine metabolism and steatosis associated with no recurrence, isolated extrahepatic recurrence or later LM. We combined the above variables to develop a decision tree-based prediction model, which performed best for identification of early LM (AUC 0.85). Taken together, these data demonstrate that in patients with apparently localized PC, pre-treatment liver biopsies may confirm the presence of human hepatic pre-metastatic niche and reveal features that correlate w