Multimodal and multikingdom spatial transcriptomics identify bacterial niches implicated in oral cancer development

Background: Oral squamous cell carcinoma (OSCC) of the tongue is the most common head and neck malignancy. OSCC develops from oral precancerous lesions (OPL) such as dysplasias which provide a window for preventive intervention. OSCC pathogenesis is influenced by genetic alterations, host immune responses, and the oral microbiome. Still, how microbial niches and their interactions in the tumor microenvironment (TME) promote oral carcinogenesis remains poorly defined. Materials and Methods: Multimodal and multikingdom (host and bacterial) spatial transcriptomics (ST) analysis using the Visium and Xenium 5K platforms was performed on matched OSCC, dysplastic and normal regions from 16 cross-sectional tumor-normal specimens and from 21 longitudinally transforming dysplasia and OSCC biopsies, and on tissues from carcinogen-driven mouse models of OSCC. Custom add-on panels targeting key oral cancer genes and bacterial probes (16S rRNA, 20 oral taxa including F. nucleatum ) were included which enabled cross-platform mapping of host-microbe interactions. Copy number alterations (CNAs) were spatially inferred and analyzed phylogenetically. Neighborhood analysis was performed to probe proximity of epithelial-microbial-TME interactions. Results: Following quality control, over 120,000 spots from Visium ST and over 3.5 million Xenium-resolved cells were retained which comprised diverse epithelial, immune, and stromal lineages. Trajectory analysis revealed major differentiation states mapping to distinct layers in the oral epithelium. Along the normal to dysplasia to OSCC continuum, basal cells progressively expanded while more differentiated cells (e.g., granular) were reduced, accompanied by increased EMT and inflammatory signatures. Spatial analysis of CNAs showed elevated aneuploidy across the normal-dysplasia-OSCC spectrum, with basal cells among all epithelial layers harboring the highest CNA burdens. Spatial phylogenetic analysis identified shared clones between dysplasias and OSCC, marking early events in OSCC evolution. Multikingdom ST identified bacteria-enriched niches that harbored tumor cells and IL1B high granulocytic myeloid cells. Epithelial cells in these niches displayed genes with roles in host defense ( DEFB4A , LCN2 ) as well as chemokines that promote myeloid cell recruitment (e.g., CXCL1 , CXCL8 ), suggesting epithelial-microbe-immune interactions with both tumor restricting and promoting properties that together tune OSCC progression. These niches and their properties were recapitulated in tumors from 4NQO-exposed mice. Conclusions: Our study suggests that epithelial remodeling, inflammatory signaling, and bacteria-rich niches result in epithelial-microbial-immune interactions that shape the pathogenesis of OPL and OSCC and, thus, harbor high-potential targets that can guide early treatment strategies for OSCC.