Deconvolution of cell-type-associated markers predictive of response to neoadjuvant radiotherapy

Abstract

Tumor microenvironent contains prognostic molecular markers and therapeutic targets from different cellular sources, which are still not fully revealed in the resistance and recurrence after radiotherapy for rectal cancer. By integrating the scRNA-seq data, we deconvoluted the bulk transcriptomics of rectal cancer collected before preoperative neoadjuvant radiotherapy (nRT) into fractions and gene expression of the six cell types. The inferred cell-type-associated DEGs, abbreviated as caDEGs, of myeloid and stromal cells were enriched for overlapping yet unique biological processes including immunity, angiogenesis, and metabolism, respectively. Ecotyper analysis indicates that the caDEGs reflects cell states and ecotypes in association with nRT response. By mapping the caDEGs onto the context-free and newly built ligand-receptor and collagen-integrin lists from scRNA-Seq data, respectively, we inferred 297 cell-type-specific trans- and/or cis-collagen-integrin and 219 heterotypic ligand-receptor interactions potentially associated with nRT response, including interactions between stromal-associated COL1A2/COL6A1/COL6A2 and stromal or CMS1-associated ITGA1/B1, between epithelial-associated JAG1 and stromal-associated NOTCHs, between CMS2 epithelial-associated CCL15 and proliferating myeloid-associated CCR1, between myeloid-associated CCL4/CD86 and lymphatic endothelial-associated ACKR2, and between myeloid-associated TNFS13B and B cell-associated TNFRSF13B/C, etc. Intriguingly, results suggest a greater number of down-regulated cell-type-related markers in resistant cancers to nRT. Favorable myeloid-associated CD14, epithelial-associated DYM, stromal-associated COL1A2 and COL3A1, and unfavorable epithelial-associated CELSR3 and KCNH8 markers were inferred at least from two independent nCRT datasets of GSE119409, GSE35452, and GSE45404. The results provide insights into roles of the stromal and immune cells beside epithelial cells in resistance to radiotherapy for rectal cancers. The proposed approach can be applicable to other diseases as well. Codes and additional data are available at https://github.com/Xueling21/rectalNRT_deconv.