A novel CAF-cancer cell crosstalk-related gene prognostic index based on machine learning: prognostic significance and prediction of therapeutic response in head and neck squamous cell carcinoma.

Background: Cancer-associated fibroblast (CAF)-cancer cell crosstalk (CCCT) plays an important role in tumor microenvironment shaping and immunotherapy response. Current prognostic indexes are insufficient to accurately assess immunotherapy response in patients with head and neck squamous cell carcinoma (HNSCC). This study aimed to develop a CCCT-related gene prognostic index (CCRGPI) for assessing the prognosis and response to immune checkpoint inhibitor (ICI) therapy of HNSCC patients.

Methods: Two cellular models, the fibroblast-cancer cell indirect coculture (FCICC) model, and the fibroblast-cancer cell organoid (FC-organoid) model, were constructed to visualize the crosstalk between fibroblasts and cancer cells. Based on a HNSCC scRNA-seq dataset, the R package CellChat was used to perform cell communication analysis to identify gene pairs involved in CCCT. Least absolute shrinkage and selection operator (LASSO) regression was then applied to further refine the selection of these gene pairs. The selected gene pairs were subsequently subjected to stepwise regression to develop CCRGPI. We further performed a comprehensive analysis to determine the molecular and immune characteristics, and prognosis associated with ICI therapy in different CCRGPI subgroups. Finally, the connectivity map (CMap) analysis and molecular docking were used to screen potential therapeutic drugs.

Results: FCICC and FC-organoid models showed that cancer cells promoted the activation of fibroblasts into CAFs, that CAFs enhanced the invasion of cancer cells, and that CCCT was somewhat heterogeneous. The CCRGPI was developed based on 4 gene pairs: IGF1-IGF1R, LGALS9-CD44, SEMA5A-PLXNA1, and TNXB-SDC1. Furthermore, a high CCRGPI score was identified as an adverse prognostic factor for overall survival (OS). Additionally, a high CCRGPI was positively correlated with the activation of the P53 pathway, a high TP53 mutation rate, and decreased benefit from ICI therapy but was inversely associated with the abundance of various immune cells, such as CD4+ T cells, CD8+ T cells, and B cells. Moreover, Ganetespib was identified as a potential drug for HNSCC combination therapy.

Conclusions: The CCRGPI is reliable for predicting the prognosis and immunotherapy response of HSNCC patients and may be useful for guiding the individualized treatment of HNSCC patients.