Comprehensive analysis of single-cell and bulk RNA sequencing data unveils antigen-presenting and processing fibroblasts and establishes a predictive model in gastric cancer.

Background: Antigen-presenting and processing fibroblasts (APPFs) have emerged as pivotal regulators of antitumor immunity. However, the predictive value of APPF-related genes (APPFRGs) in the prognosis and tumor immune status of gastric cancer (GC) remains largely unexplored.

Methods: Bioinformatics analysis was conducted using single-cell and bulk RNA sequencing datasets of GC retrieved from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases. The APPFs were identified using AUCell algorithm based on APP-associated genes obtained from the InnateDB database. CellChat algorithm was utilized to evaluate interactions between cells. The non-negative matrix factorization (NMF) clustering analysis was performed to identify APPF-related subgroups based on TCGA‑stomach adenocarcinoma cohort. LASSO and multivariate Cox regression analysis were conducted to establish the predictive model. Immunohistochemistry of GC tissue microarrays was performed to validate the model.

Results: Compared to non-APPFs, APPFs exhibited more interactions with myeloid cells, endothelial cells, and lymphocytes via MHC-II signaling network. The two APPF-related subgroups clustered by NMF demonstrated significant differences in prognosis and immune cell infiltration. Five APPFRGs (CPVL, ZNF331, TPP1, LGALS9, TNFAIP2) were identified to establish the predictive model and stratify GC patients based on risk score. The prognosis was significantly different between the two risk groups and was validated using GEO datasets. A nomogram that efficiently predicted the overall survival of GC patients was established by integrating the risk score with age, T stage, N stage, and M stage. Furthermore, the high-risk group exhibited reduced infiltration of activated CD4⁺ T cell and increased infiltration of Treg cells, higher resistance to chemotherapy and immunotherapy, and lower tumor mutation burden. Finally, the immunohistochemical results of GC tissue microarrays revealed higher expression of CPVL, ZNF331, and TPP1, and lower expression of LGALS9 and TNFAIP2 in GC compared to adjacent normal tissues. Additionally, higher risk score in GC samples was relevant with poor differentiation, positive nerve invasion, advanced T and TNM stages, and higher expression of FOXP3.

Conclusions: APPFs may play an important role in the regulation of tumor immune microenvironment in GC and warrant further exploration. The predictive model based on APPFRGs effectively predicts the prognosis and tumor immune status of GC.