3132 Background: Spatial interactions between tumor cells and the tumor microenvironment (TME) influence disease progression, yet current diagnostic profiling lacks spatial resolution. A clinically implementable framework enabling cross-cancer spatial characterization using pretreatment clinical specimens is needed. SCRUM-Japan MONSTAR-SCREEN-3 (MONSTAR3) is an ongoing prospective clinical screening platform enrolling over 3,200 patients with solid tumors, integrating clinical annotation with multi-omic. We report an intermediate analysis validating the feasibility of extracting reproducible spatial phenotypes from pretreatment specimens at clinical scale. Methods: Spatial transcriptomics (Xenium 5K) was performed on 331 pretreatment FFPE tumor specimens from 309 patients representing ≥15 solid cancer types enrolled in MONSTAR3. Specimens obtained at diagnosis or prior to systemic therapy included unresectable advanced and curative-intent settings. Analyses examined spatial organization of tumor–immune interactions at the TME level, cancer cell functional states, neighborhoods, and tertiary lymphoid structures. Spatial features were systematically compared across cancer types and disease settings without treatment outcomes. Results: We generated a spatial atlas comprising approximately 20 million cells with high-confidence annotation across over 15 cancer types, representing one of the largest clinically annotated pan-cancer spatial transcriptomic resources to date. The atlas captures tumor-microenvironment organization, cancer-intrinsic functional heterogeneity, conserved cellular niches, and specialized immune structures including tertiary lymphoid structures with defined maturation states. Spatial tumor architecture varies substantially across cancer types despite similar cellular composition, underscoring value of spatially-resolved profiling beyond conventional cell-type quantification. Analysis across disease stages revealed that advanced tumors exhibit higher spatial compartmentalization of cancer functional states, with immunogenic and proliferative cancer populations showing consistent spatial separation across cancer types. Pan-cancer niche analysis identified conserved cellular neighborhoods, a subset of which showed stage-specific enrichment or prognostic associations. We also performed detailed spatial characterization of tertiary lymphoid structures, which demonstrated association with immunotherapy response in an independent cohort. Conclusions: We present a large-scale pan-cancer spatial atlas integrated within a nationwide precision oncology framework. This resource enables systematic investigation of spatial tumor biology and nominates spatially-defined features as candidate biomarkers for patient stratification and therapeutic response studies.
Imai et al. (Wed,) studied this question.