利用同一切片RNA-蛋白联合检测与荧光H&E对胃癌前期进展进行三维多模态空间分析

Background: Gastric cancer arises through a cascade from intestinal metaplasia to dysplasia and invasive adenocarcinoma, but only a subset of precancerous glands progress. The spatial features that distinguish progressing from non-progressing niches are unknown. We applied a multimodal platform that combines same-slide, subcellular RNA and protein profiling on the G4X Spatial Sequencer with fluorescent H&E (fH&E) and 3D holotomography (HT) to characterize pre-gastric cancer (pre-GC) niches and their transition toward carcinoma. Methods: FFPE gastric tissues from 12 patients (38 regions of interest) were profiled on G4X flow cells using a custom 358-gene gastric cancer/pre-cancer panel, a 16-plex protein panel, and fH&E on the same slide. We used multimodal clustering and classification of spatial transcriptomic, morphology, and protein features to define cell states and associated niches. For matched regions, 3D HT from the same tissues was used to embed 2D molecularly defined states into their gland and mucosal context and to assess whether pre-GC states share or segregate across glandular units, extend through the mucosal depth, and how immune populations are positioned around them in 3D. By aligning subcellular G4X coordinates with HT volumes, this method enables us to visualize the x-y-z localization of transcripts within individual cells, providing a molecular level 3D model of the tumor-stroma-immune microenvironment. Results: Multimodal analysis of integrated RNA, protein, and morphology data resolved distinct epithelial, stromal, and immune compartments in precancerous tissue, whereas cancerous regions showed expansion of epithelial and immune states and altered compartment boundaries. Within the epithelial layer, regions of metaplasia and dysplasia localized to discrete gland-associated zones enriched for intestinal or gastric gene signatures. Same-slide RNA-protein measurements enabled mapping of CD4⁺ helper and CD8⁺ cytotoxic T-cell distributions relative to the gastric epithelium. Incorporating 3D HT enabled us to investigate whether 2D-defined pre-GC foci represented continuous lesions or separate glands and refined estimates of lesion extent and immune proximity along the gland axis in 3D. Conclusions: This 3D-enabled, same-slide G4X platform with fH&E, RNA transcript profiling, and a 16-plex protein panel generates multi-scale maps of pre-GC and gastric cancer niches. In particular, 3D spatial multimodal modeling improves interpretation of pre-cancer epithelial programs and immune positioning and supports discovery of 3D niche features linked to progression risk beyond conventional 2D histology. Generative AI assistance was limited to language editing of this abstract. The scientific content, interpretation, and conclusions are the sole responsibility of the authors, who have reviewed and approved the final version.