Ketone Body Metabolism in Cancer-Associated Fibroblasts and FSTL3 as Prognostic and Therapeutic Determinants of Chemoradiotherapy Response in Cervical Cancer

Background: Concurrent chemoradiotherapy (CCRT) is a cornerstone of treatment for locally advanced cervical cancer, although many patients show a poor response. The tumor microenvironment (TME) is a key determinant of therapeutic resistance and prognosis, with the metabolic reprogramming within specific cellular subsets playing a particularly important role. However, the cellular sources of metabolic heterogeneity in cervical cancer and their clinical implications have yet to be firmly established. Methods: An integrated analysis of single-cell RNA sequencing (scRNA-seq) datasets from eight patients cervical cancer, together with bulk RNA sequencing data retrieved from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) datasets, was conducted. Cellular heterogeneity was characterized using the Seurat and Harmony, while metabolic activity was assessed by integrating five algorithms (AUCell, UCell, singscore, ssGSEA, and AddModuleScore). The Least Absolute Shrinkage and Selection Operator (LASSO)-Cox regression was used to construct a prognostic model, which was subsequently validated in external cohorts. The functional role of the key gene FSTL3 was examined through both in vitro and in vivo experiments. Results: scRNA-seq analyses identified 10 major cell types within the cervical cancer TME. Among these, cancer-associated fibroblasts (CAFs) showed the greatest enrichment of ketone body metabolism-related gene signatures. CAFs with high enrichment of these signatures exhibited enhanced cell–cell communication and were associated with activation of the AP-1 transcription factor network. Analysis of these CAFs led to the development of a nine-gene prognostic signature (KLF4, SERPINF1, CD74, PRPS2, ERRFI1, FSTL3, SOD2, JUN, and CXCL3). A risk model based on these genes predicted survival in the TCGA training and validation sets and in independent GEO cohorts (disease-free survival in GSE44001 and overall survival in GSE52903). A nomogram combining the risk score with clinical variables showed improved predictive performance compared with either the risk score or clinical variables alone. Knockdown of FSTL3 suppressed tumor growth and enhanced sensitivity to irradiation and cisplatin in both in vitro and in vivo models, supporting its functional role in cervical cancer progression and treatment response. Conclusion: These analyses identify a CAF population enriched in ketone body metabolism-related gene expression that may be associated with cervical cancer progression. In addition to establishing a prognostic model for further evaluation, this study identifies FSTL3 as a potential therapeutic target that warrants further mechanistic and pharmacological investigation.