DeCAF redefines fibroblast states uncovering multidimensional tumor-stroma relationships driving clinical tumor progression and immunotherapy response [scRNA-Seq]

Cancer-associated fibroblast (CAF) heterogeneity within pancreatic ductal adenocarcinoma (PDAC) has been previously assessed through single-cell RNA sequencing (scRNAseq) to discover distinct CAF populations such as myCAFs and iCAFs. While useful as a biological framework, no studies have conclusively and robustly demonstrated a correlation of CAF subpopulations with clinical prognosis or therapy response. We define restraining (rest) and promoting (pro) CAFs in patient samples that are both prognostic and predictive of therapy response in disparate tumor types. Using a single sample classifier, DeCAF, we uncover distinct clinical and spatial interactions between pro- and restCAF subtypes and basal-like and classical tumor subtypes that support tumor stroma crosstalk where basal-like tumors are enriched in proCAFs and classical tumors are enriched in restCAFs. Finally, we find striking differences in the immune contexture of pro- and restCAF tumors where restCAF dominant tumors are enriched in patients who respond to immune checkpoint inhibition (ICI) in two large clinical trials, and proCAF dominant tumors are enriched in patients who respond to myeloid inhibition. This work is the first to define CAF subtypes that are clinically robust, prognostic, predictive of immunotherapy response, and associate with tumor molecular subtype.