Introduction:Chronic graft-versus-host disease (GVHD) is a leading cause of late non-relapse mortality following allogeneic hematopoietic cell transplantation, yet spatial biomarkers capable of predicting outcomes remain lacking. Minor salivary glands (MSGs) are anatomically accessible organs that can be easily accessed by fast and minor invasive biopsies and reflect systemic immune perturbations, making them ideal for spatial profiling of immune diseases. We applied STARComm, a spatial multiomics framework that identifies Multicellular Interaction Modules (MCIMs), which represent tissue microenvironments where multiple cell types engage in coordinated signaling. MCIMs capture clinically relevant hubs of immune communication and tissue remodeling, enabling spatial resolution of pathogenic processes in GVHD.

Methods:Thirty-six MSG tissues were collected from patients enrolled between 2018 and 2022 at the Hospital das Clínicas, University of São Paulo. All patients had undergone allogeneic bone marrow transplantation and subsequently received MSG biopsies as part of clinical follow-up for chronic GVHD assessment. The cohort included healthy controls (n=8), GVHD survivors (n=18), and patients who died from GVHD-related complications (n=10), with detailed clinical metadata. Spatial multiomic profiling was performed on the same histological section using a dual-modality approach: spatial transcriptomics with Xenium (10x Genomics) using a targeted 280-gene panel, and spatial proteomics with Phenocycler Fusion (Akoya Biosciences) using a 36-antibody multiplex panel. MCIMs were computed via STARComm by integrating local communication density maps across spatial grids. Cell types were annotated with TACIT, and differences in MCIM composition were compared across groups. Prognostic associations were modeled using Cox regression. To prioritize therapeutic targets, we applied SpatialDrug2Cell, a single-cell-based pipeline for inferring druggable dependencies, to spatially segmented cell populations, adapting it for high-resolution in situ data.

Results:A total of 842,154 spatially resolved cells were analyzed across all samples. Receptor–ligand interactions were inferred from 17 curated gene pairs selected for biological relevance to GVHD and compatibility with both transcriptomic and proteomic detection, based on expression specificity and spatial proximity. Nine MCIMs were identified. In healthy MSGs, MCIMs reflected structured stromal, epithelial, and vascular signaling. GVHD samples exhibited marked reorganization, with MCIMs-0, -2, and -6 significantly enriched (p<0.001), particularly within TLS-like periductal niches. MCIM-0, dominant in fatal cases, exhibited dense B and T cell, T and T cell, and stromal interactions, with high spatial density of CXCL12-CXCR4 and CCL5-SDC4 signaling, primarily transcribed by fibroblasts and pericytes. In contrast, MCIM-3, enriched in fibroblast–endothelial interactions and B cell regulatory networks, was more abundant in survivors. In multivariate analysis, MCIMs-0 and -6 were independently associated with increased mortality (HR=2.08 and 2.83), while MCIM-3 showed a protective trend (HR=0.51). SpatialDrug2Cell, applied to MCIM-localized cell populations and spatially restricted therapeutic targets, highlighting candidate agents such as rituximab, atezolizumab, and plerixafor.Conclusions:Our analysis establishes spatially resolved MCIMs as prognostic biomarkers in chronic GVHD. The emergence of high-risk communication modules in MSGs, particularly in patients with fatal outcomes, reveals tissue-localized immune architectures associated with disease progression. By integrating STARComm with adapted drug inference tools like SpatialDrug2Cell, we identify clinically relevant, spatially anchored therapeutic targets in routinely sampled tissues. This framework may enable mechanism-based stratification and therapeutic guidance in GVHD and other immune-mediated diseases

This content is only available as a PDF.
2025
Sign in via your Institution