MIF and CXCL8 shape inflammatory myeloid–epithelial crosstalk in cervical squamous cell carcinoma: a machine learning and integrative transcriptomic investigation

Background Progression from high-grade squamous intraepithelial lesion (HSIL) to cervical squamous cell carcinoma (SCC) is accompanied by extensive epithelial remodeling and changes in the inflammatory immune microenvironment. However, how stage-specific epithelial states interact with inflammatory myeloid responses, spatial immune organization, and patient prognosis remains unclear. This study investigated these changes and identified inflammation-related molecular features associated with clinical outcome. Methods Single-cell RNA sequencing, spatial transcriptomics, bulk transcriptomic and survival cohorts, and drug-response data were integrated. Stage-associated cell states, epithelial trajectories, gene co-expression programs, and predicted cell–cell communication were analyzed across normal cervical tissue, HSIL, and squamous cell carcinoma. Machine-learning survival models were developed in TCGA-CESC and evaluated in GSE39001, GSE52903, and CGCI-HTMCP-CC. Pathway enrichment and immune deconvolution were used to assess biological and immune features associated with the prognostic model. Drug-response prediction, molecular docking, 100-ns molecular dynamics simulations, and immunohistochemistry provided additional computational and tissue-level evaluation. Results HSIL was characterized by a proliferative epithelial state, whereas SCC was enriched for an epithelial state associated with keratinization stress, interferon responses, and inferred copy-number abnormalities, accompanied by increased inflammatory myeloid populations. Spatial transcriptomic analysis showed stronger immune and myeloid programs at the invasive front than in the tumor core. Predicted epithelial–myeloid communication in SCC prominently involved MIF–CD74 receptor complexes and CXCL8–CXCR2 signaling. A five-gene prognostic model consisting of MIF, SPP1, CXCL8, CALML5, and CD74 achieved C-indices of 0.748, 0.696, and 0.691 in TCGA-CESC, GSE39001, and GSE52903, respectively. Higher risk scores were associated with hypoxia, glycolysis, TGF-β/EMT signaling, stromal remodeling, and reduced adaptive immune programs. MK-2206 showed favorable predicted interactions with MIF and CXCL8, with greater modeled stability for the CXCL8 complex. Immunohistochemistry confirmed progressively increased CXCL8 and MIF expression from normal cervix to HSIL and SCC. Conclusion Progression from HSIL to SCC was associated with coordinated epithelial remodeling, inflammatory myeloid activation, and spatial reorganization of the immune microenvironment. MIF–CD74 and CXCL8–CXCR2 signaling were implicated in epithelial–myeloid communication, while the five-gene inflammatory model showed reproducible prognostic performance across independent cohorts. These results support further investigation of MIF- and CXCL8-related inflammatory signaling in cervical cancer progression and prognosis.