Plasma proteomic profiling identified prognostic indicators with therapeutic potential for colorectal cancer

Abstract Plasma proteins have been reported as predictors and potential targets for reducing colorectal cancer (CRC) risk. However, their potential roles in CRC prognosis remain unexplored. We measured plasma levels of 367 neuro-related proteins in CRC patients from the West China Hospital (WCH) cohort (N=150, median follow-up=46.72 months) via proximity extension assay. The least absolute shrinkage and selection operator (LASSO) penalized Cox regression identified five overall survival (OS) - and eleven disease-free survival (DFS) - associated proteins, and the multi-protein signature for OS prediction was then validated in the UK Biobank (UKB) cohort (N=1,133). To overcome possible effects from confounders, we then employed Mendelian Randomization analysis leveraging protein quantitative trait loci (pQTLs) to investigate associations between genetically determined protein concentration and OS and cancer-specific survival (CSS) of CRC in the UKB. We found that multi-protein signature developed in the WCH cohort (c-index=0.784, 95% CI=0.713-0.855) showed significant discriminative ability in the external UKB cohort (c-index=0.616, 95%CI=0.559-0.673). A significant association between genetically determined PD-L1 and OS (p=0.043, HR=1.53, 95%CI=1.01-2.29) was observed, although we did not find strong evidence for colocalization. Additionally, single-cell and spatial transcriptome analyses illustrated PD-L1 expression localized predominantly to epithelial cells and immune cells (especially myeloid cells) in CRC tissue. The potential interactions of identified proteins were evaluated in the STRING database. Druggability evaluation also supported PD-L1 as a potential therapeutic target for CRC. Taken together, this study established multi-protein signatures for CRC prognosis and identified plasma PD-L1 as a possible biomarker and therapeutic target.