Spatial niche organization shapes clinical heterogeneity in Ewing sarcoma

Single-cell transcriptomics can identify cell populations but often loses spatial context because it requires recovering intact, viable cells; spatial transcriptomics, in contrast, resolves high-dimensional gene transcription in situ ^[1]. Integrating the two can further reveal tissue regions, restricted cell distributions, and co-enrichment relationships ^[3]. Tumor spatial architecture may carry clinical significance: colorectal cancer liver metastases show spatial reprogramming of immunosuppressive cells ^[2]; in liver cancer, a tumor immune barrier composed of SPP1+ macrophages and cancer-associated fibroblasts is associated with response to anti-PD-1 therapy ^[4]. Thus, the question is not only cell identity, but also how neighboring cells assemble into niches and relate to outcomes. Ewing sarcoma (EwS), a childhood cancer driven by FET::ETS fusions, is particularly striking: it has a monomorphic histological appearance and few additional driver mutations, yet patient outcomes are highly variable.

In 135 patients, the study analyzed EwS and its microenvironment at single-cell and spatial resolution, covering more than 3 million cells. EwS cells converged on 5 recurrent transcriptional states, including fusion activity, proliferation, interferon response, and hypoxia states; the microenvironment comprised 12 stromal and immune cell types. Recurrent combinations of adjacent tumor cells and microenvironmental cells defined 8 niches: 4 of these each featured distinct EwS tumor cell states, while the other 4 were characterized primarily by microenvironmental composition. Co-localization and mutual segregation unfolded along a perivascular–necrotic gradient or were organized around focal immune infiltration zones, revealing the spatial structure of EwS growth.

Based on niche composition, patients were classified into 4 EwS types showing distinct overall survival patterns; this classification was further validated by bulk transcriptomics in 2 independent patient cohorts. Immune Reactive EwS contained interferon-activated tumor cells and patients had better survival, suggesting that tumor-intrinsic interferon signaling may serve as a new prognostic axis. Fusion Active had better outcomes, TME-enriched showed intermediate survival, and Proliferative had the worst survival. Thus, the study advanced heterogeneity analysis from single-cell states to reproducible spatial niches, directly linking them to patient outcomes.

However, the current results mainly show that spatial niches are prognostically associated and that the classification has been validated; causal effects have not been demonstrated, nor has the efficacy of selecting treatment according to the classification been evaluated. Whether niches can form prospective clinical decision tools and whether tumor-intrinsic interferon signaling can become an actionable target remain to be answered; moreover, processed spatial transcriptomics data are currently available only upon reviewer request, with raw data to be deposited publicly after publication, so external verification and re-analysis remain limited.