Single-cell and spatial transcriptomic analyses identify a PPIA-associated malignant epithelial state linked to SPP1⁺ macrophage remodeling in head and neck squamous cell carcinoma

Intratumoral heterogeneity in head and neck squamous cell carcinoma (HNSCC) contributes to disease progression and poor outcomes, but the malignant epithelial cell states underlying adverse prognosis remain unclear. Here, we integrated single-cell RNA sequencing data from 39 HNSCC patients to construct a comprehensive cellular atlas. Malignant epithelial cells were identified by copy number variation inference and further resolved by unsupervised clustering. By linking single-cell profiles with TCGA-HNSC transcriptomic and survival data using Scissor, we identified a prognosis-associated malignant epithelial subpopulation, termed C1, characterized by epithelial–mesenchymal transition and extracellular matrix remodeling programs. Integrative analyses combining Scissor-defined prognostic genes, TCGA survival-associated genes, C1 marker genes, and hdWGCNA co-expression modules converged on PPIA as a putative core regulator of this adverse cell state. Transcription factor network analysis further revealed selective activation of HMGA2 in C1 cells. Cell–cell communication analysis showed that PPIA-high C1 cells preferentially interact with SPP1-positive macrophages through the PPIA–BSG/CD147 ligand–receptor axis, and spatial transcriptomic analysis confirmed significant colocalization of these two populations within the tumor microenvironment. Functional experiments demonstrated that PPIA promotes HNSCC cell proliferation, migration, tumor growth, and M2-like macrophage polarization. Together, our study identifies an EMT-like malignant epithelial subpopulation associated with poor prognosis in HNSCC and reveals a PPIA-mediated communication network between malignant epithelial cells and SPP1-positive macrophages, highlighting PPIA as a potential prognostic biomarker and therapeutic target.