Transcriptome and spatial analyses of tumor microenvironment in addressing colorectal cancer racial and ethnic disparities; and ethical evaluation of results communication in polygenic risk score-based prediction tool

Abstract

Substantial and persistent racial and ethnic differences exist in colorectal cancer (CRC) incidence and mortality, while the underlying biology remains incompletely understood. To bridge this knowledge gap, I leveraged tumor RNA sequencing data from 548 CRC patients from African American, Alaska Native, Hispanic, and non-Hispanic White groups. Within each racial and ethnic group, a nested case-control design was employed to match patients who died of CRC to those who did not, on key clinical factors. I observed that Alaska Native and Hispanic patients exhibited significantly higher mean T cell-inflamed gene expression profile scores than the other two groups, which is indicative of an immunologically "hot" tumor, with active T cell infiltration, IFN-gamma pathway activation, and immune checkpoint expression. I identified 14 genes, multiple pathways, and 4 immune cell types that significantly differed between patients with and without CRC-specific death. Specifically, 6 genes reported to be associated with CRC death for the first time. I developed a gene expression-based model to predict CRC-specific death with a better predictive performance compared to existing signatures. With the inclusion of diverse patient populations, we identified novel prognostic biomarkers, which also provide potential biological explanations for observed racial and ethnic disparities in CRC outcomes. To further characterize the spatial organization within the tumor microenvironment, I investigated the spatial proteomics profile of 97 Alaska Native CRC patients using Akoya Biosciences' PhenoCycler-Fusion system. I included 35 patients who died of CRC and 62 matched patients who did not die of CRC based on key clinical factors. I classified 2.35 million cells into 16 different cell types and eight cellular neighborhoods with distinct local cell composition. I found that patients with a higher proportion of M2 macrophages were more likely to die of CRC. While no significant associations between cellular neighborhoods and CRC-specific death, I observed a higher proportion of podoplanin-expressing stromal cells and M1 macrophages within the tumor epithelial compartment among patients who died of CRC compared to patients who did not, which emphasizes the value of cell locations. I calculated the spatial proximity between different cell types and found colocalizations of stromal and immune cells (e.g., podoplanin-expressing stromal cells and CD8+ T cells) were associated with CRC-specific deaths. These results reveal the prognostic value and biological relevance of colocalized stromal and immune cell types, suggesting that spatial patterns between cells may serve as a prognostic biomarker to reduce CRC mortality. In the era of precision medicine, polygenic risk scores (PRS) demonstrate strong performance in risk prediction for multiple diseases, and are increasingly returned in direct-to-consumer genetic testing. I summarized current approaches to communicating PRS and the associated perceptual, psychological and behavioral changes observed, which provided advice for standardizing PRS communication metrics and the optimization of report design to facilitate further implementation research.