Spatial 3D and multi-omics mapping of diffuse gastric cancer evolution from preinvasive to invasive lesions in CDH1 mutation carriers

Background: Hereditary diffuse-type gastric cancer (DGC) lacks a defined precursor sequence. Germline CDH1 loss-of-function mutations confer a high lifetime risk, and prophylactic gastrectomies in asymptomatic carriers frequently reveal numerous occult intramucosal signet-ring cell (SRC) foci that can remain indolent for years. The biological mechanisms underlying the progression from these indolent SRC foci to invasive DGC remain largely unknown. Methods: We profiled primary DGCs from 26 patients, sampling the full histologic continuum from SRC foci to invasive lesions and overtly invasive fronts. Whole-slide sections were pathologist-annotated for tumor subtype and key microenvironmental features. Serial sections were assayed with complementary spatial multi-omics platforms: GeoMx DSP, Visium, Xenium Prime 5K with post hoc Codex (IO60), and CosMx (WTX). Selected specimens also underwent high-plex spatial multi-omics sequencing (transcriptome/protein/TCR/BCR) using Singular G4X, as well as 3D modeling with both Singular G4X and holotomography (HT-X1). Cross-platform datasets were co-registered to common coordinates; cell segmentation, phenotype annotation, and neighborhood/niche mapping quantified transcriptomic dynamics and tumor-microenvironment interactions across regions spanning preinvasive to invasive lesions. Integration of Singular G4X and Label-free holotomography provided 3D context. Results: Spatial transcriptomic profiling revealed invasion-associated transcriptional programs along the SRC-to-invasive continuum. The meta-programs (MPs) associated with discrete stages of progression were defined. Histology-integrated analyses highlighted invasion-depth-dependent shifts in cell-cycle, EMT, and stress-metabolism pathways, and trajectory/differentiation analyses supported a continuous invasion axis. Copy-number inference indicated two evolution trajectories, linear and branched. Single-cell-resolution spatial multi-omics refined tumor cell-state signatures and their spatial neighborhoods, revealing lymphocyte-dominant microenvironmental states at early lesions and stromal/myeloid remodeling with deeper invasion, corroborated by multiplex proteomics (IO60). Cross-section registration with label-free 3D holotomography generated volumetric reconstructions mapping coherent clonal sectors and concordant TME remodeling along the invasion axis. Conclusions: We present a spatial multi-omic atlas of DGC progression. The work delineates subtype- and region-specific gene programs, clonal architectures, and TME remodeling at high spatial and molecular resolution, modeling disease progression in 3D. Our findings identify candidate biomarkers of progression, risk stratification, and potential therapeutic interception of SRC tumorigenesis in CDH1 mutation carriers.