IDENTIFYING DISTINCT INFLAMMATORY FIBROBLAST STATES BY SPATIAL TRANSCRIPTOMICS TO PREDICT COLECTOMY IN PEDIATRIC ACUTE SEVERE COLITIS

Abstract BACKGROUND Studies suggest approximately between 10-20% of patients with pediatric Ulcerative Colitis (UC) will require surgical intervention with colectomy. Although prior studies have attempted to find predictors for the need of colectomy in pediatric UC, biomarkers for surgical interventions are still lacking. We set out to identify transcript and protein signatures of medication-refractory colitis through spatial single-cell transcriptomics and multiplex protein staining to predict the need for colectomy. METHODS Patients were identified from 2019 to 2023 who required inpatient hospitalization for acute severe colitis (ASC) at diagnosis. Deidentified intestinal biopsy and intestinal resection samples were obtained through ISTAT study (Seattle Children’s IRB Study ID 00001308) from pediatric patients with ASC at diagnosis, on treatment, and at the time of colectomy. We profiled diagnostic human biopsies from the colon of treatment-naïve pediatric patients with ASC at the time of diagnosis (n = 12) in two cohorts of patients [colectomy (n = 6) and no colectomy (n = 6)] (Figure 1A and 1C). Data was obtained using 10x Xenium with 475 transcript probes on formalin fixed paraffin embedded (FFPE) samples (Figure 2Ai+ii) Patients in both groups had similar clinical presentation in age, PUCAI scores, Mayo scores, and laboratory parameters (CRP, ESR, Stool calprotectin) at time of diagnosis (Figure 1B). RESULTS Using 10x Xenium platform, we identified cell types from 5 major cell type classes (Stromal cells, Epithelial cells, T cells, B cells and Myeloid cells) via 326 million spatially-localized transcripts from 1.1 million cells. We identified shared cellular neighborhoods (termed zones 0 through 8) to define sub-anatomic regionalized networks of disease activity and to use as a basis for comparison across samples. We mapped neighborhood structure onto tissue samples (Figure 2Ai) and visualized cell-type clustering via uniform manifold approximation projection (UMAP) (Figure 2Aii). Cell types were identified using reference single-cell RNA-seq datasets available from previous single-cell studies available in the literature. Using data from naïve-to-treatment colonic biopsies at time of diagnosis, we identified an expansion of inflammatory fibroblasts in the cohort that went on to need colectomy (Figure 2Aiii). Further analysis reveals divergent transcript signatures in inflammatory fibroblasts in the two cohorts (Figure 2B), along with differences in fibroblast proximity to immune cells (T cells) and fibroblast location within the lamina propria. CONCLUSION Taken together, this suggests that there exist distinct inflammatory microenvironments driving recalcitrant disease states. Further analysis is underway to identify specific pathways and understand cell-cell network pathogenesis.