Spatial transcriptomics reveals epithelial-immune remodeling preceding malignant transformation of oral premalignant lesions

Oral premalignant lesions (OPLs) are common; however, histopathological grading incompletely identifies lesions destined for cancer. To define tissue ecosystems that precede malignant transformation, we integrated single-cell-resolution Xenium spatial transcriptomics of 20 HPV-negative biospecimens from 16 patients, with independent single-cell RNA sequencing and immunofluorescence data from 38 patients with OPLs after quality control. Progressive OPLs exhibit a coordinated epithelial-immune program comprising MX1/NOTCH3-high KRT14+/KRT15+ basal epithelial states, S100A9-high inflammatory macrophages, TIGIT-high, exhaustion-associated T cells, and altered dendritic cell states. Spatial analyses showed that T cells and dendritic cells were displaced from the basal epithelial interface, and that epithelial neighborhoods were depleted of dendritic cells but enriched for TIGIT-high T cells, revealing immune reorganization not captured by cellular abundance alone. Independent single-cell and protein-level analyses reproduced basal epithelial interferon/stress programs, S100A9-associated myeloid inflammation, TIGIT-high, exhaustion-associated T-cell programs, and dendritic cell redistribution. Ligand-receptor inference suggested convergent myeloid and lymphoid signals linked to epithelial stress, checkpoint regulation, and extracellular matrix remodeling. Functionally, recombinant S100A9 accelerated wound closure in oral epithelial and cancer cells, whereas S100A9 and S100A8/S100A9 enhanced oral cancer cell proliferation. These responses were attenuated by pharmacologic inhibition of TLR4 or RAGE. Together, these data define a spatially organized, myeloid-skewed, checkpoint-enriched ecosystem present before invasion and suggest S100A9-TLR4/RAGE signaling and epithelial-immune geometry as candidate mechanisms and biomarkers for oral cancer interception.