Index of Cancer-Associated Fibroblasts Is Superior to the Epithelial–Mesenchymal Transition Score in Prognosis Prediction

In many solid tumors, tissue of the mesenchymal subtype is frequently associated with epithelial-mesenchymal transition (EMT), strong stromal infiltration, and poor prognosis. Emerging evidence from tumor ecosystem studies has revealed that the two main components of tumor stroma, namely, infiltrated immune cells and cancer-associated fibroblasts (CAFs), also express certain typical EMT genes and are not distinguishable from intrinsic tumor EMT, where bulk tissue is concerned. Transcriptomic analysis of xenograft tissues provides a unique advantage in dissecting genes of tumor (human) or stroma (murine) origins. By transcriptomic analysis of xenograft tissues, we found that oral squamous cell carcinoma (OSCC) tumor cells with a high EMT score, the computed mesenchymal likelihood based on the expression signature of canonical EMT markers, are associated with elevated stromal contents featured with fibronectin 1 (Fn1) and transforming growth factor-β (Tgfβ) axis gene expression. In conjugation with meta-analysis of these genes in clinical OSCC datasets, we further extracted a four-gene index, comprising FN1, TGFB2, TGFBR2, and TGFBI, as an indicator of CAF abundance. The CAF index is more powerful than the EMT score in predicting survival outcomes, not only for oral cancer but also for the cancer genome atlas (TCGA) pan-cancer cohort comprising 9356 patients from 32 cancer subtypes. Collectively, our results suggest that a further distinction and integration of the EMT score with the CAF index will enhance prognosis prediction, thus paving the way for curative medicine in clinical oncology.