Development of a putative liquid biopsy gene panel for predicting liver metastasis in castration-resistant prostate cancer.

Abstract

226 Background: Liver metastases (LM) in prostate cancer (PCa) while considered rare, are associated with poor clinical outcomes. At autopsy, LM are present in 25% of PCa patients. However, they are rarely detected until late in the disease, as they are often asymptomatic until well-established. Understanding their biological drivers may inform treatment stoppages and could lead to better detection by means such as liquid biopsy. A meaningful panel of genes with altered expression in LM is needed to address these unmet and urgent needs. Methods: To develop a putative gene expression-based panel we acquired 17 samples of single cell RNA seq from primary PCa tumors and distal metastases from PCa from GSE210358 and GSE264573, including 5 from LMs. After quality control and cell type determination via Seurat, epithelial tumor cells were isolated. Using FindMarkers, gene expression was compared in cancer cells from LMs to all other metastatic sites. Significantly differentially expressed (DE, adjusted p-val < 0.05 and log2 fold change >1 or <-1) genes were filtered by keeping only those with a gene encoding surface protein (GESP) score 5 in The Cancer Surfacesome Atlas (TCSA) and that are found in extracellular vesicles (EVs) in either Vesiclepedia or Exocarta. Genes that passed these criteria were queried on enrichR to determine controlling transcription factors (TFs), queried with the terms “liver metastasis” and “prostate cancer” on NCBI’s search API, and validated by their expression on the Human protein atlas (HPA) IHC staining on tissue microarrays in PCa. Results: Comparing LM to all other metastatic sites, 677 DE genes were found. Of these, 69 upregulated genes were confirmed surface proteins in TCSA and found in EVs in Vesiclepedia or Exocarta, and 45 of these were confirmed present in PCa, pancreatic, or colon cancer from the HPA. REST and SUZ12 were the dominant TFs controlling these genes. Twelve genes were found in conjunction with the term “liver metastasis” and 32 with the term “prostate cancer” more than once in NCBI-listed publications, with MET, KIT, and MUC1 being found in 211, 142, and 25 publications with “liver metastasis” and 2633, 805, and 111 publications with “prostate cancer” respectively. Additionally, two genes (DLL3 and GPC3) have been confirmed in plasma samples from PCa with LM via click chemistry mediated PCR. In total, 13 genes (ENO1, MUC1, ASPH, MET, PLXND1, ITGB6, CA9, KIT, DLL3, TGFA, SPHK1, LGR5, and GPC3) either meet all our criteria for being a putative liquid biopsy marker for LM in PCa, or have been confirmed in plasma samples from PCa with LM. Conclusions: Using this unique approach we have identified a candidate gene signature for detection of LM in advanced PCa using platforms such as EV-based liquid biopsy. Further studies are underway to confirm the utility these genes using biobanked plasma samples.