Stool host transcriptome reveals gut epithelial cell states and immune dynamics for gastrointestinal diagnosis and monitoring

Gastrointestinal diagnosis and monitoring rely on colonoscopy, which requires anesthesia and a long wait, and non-invasive alternatives are lacking. Here we show that whole-transcriptome sequencing of stool host RNA, recovering a median of 6,300 protein-coding genes across 285 libraries spanning colorectal cancer (CRC) and inflammatory bowel disease (IBD), precancerous lesions (APL), and normal samples, reads a census of the cells shed into the lumen. In normal samples, epithelium supplies 81% and immune cells supply 11% of stool host mRNA transcripts; the immune cells’ share rises in CRC and IBD, and B- and T-cell receptor clonotypes rise with it. In this work, we show that CRC, IBD, and normal samples carry distinct transcriptome profiles, which a three-class model separates with an area under the curve of 0.94–0.96, and that within inflammatory bowel disease a 15-gene neutrophil score tracks endoscopic activity better than fecal calprotectin in the same stool.