Single-cell spatial transcriptomics of formalin-fixed, paraffin-embedded biopsies reveals colitis-associated cell networks
Elvira Mennillo, Madison L. Lotstein, Gyehyun Lee, Julian H. Hou, Vrinda Johri, Donna E. Leet, Christina A. Ekstrand, Jessica Tsui, Jun Yan He, Uma Mahadevan, Walter Eckalbar, Ryan M. Gill, Christopher J. Bowman, David Y. Oh, Gabriela K. Fragiadakis, Michael G. Kattah, Alexis J. Combes
Elvira Mennillo, Madison L. Lotstein, Gyehyun Lee, Julian H. Hou, Vrinda Johri, Donna E. Leet, Christina A. Ekstrand, Jessica Tsui, Jun Yan He, Uma Mahadevan, Walter Eckalbar, Ryan M. Gill, Christopher J. Bowman, David Y. Oh, Gabriela K. Fragiadakis, Michael G. Kattah, Alexis J. Combes
View: Text | PDF
Research Article Gastroenterology Inflammation

Single-cell spatial transcriptomics of formalin-fixed, paraffin-embedded biopsies reveals colitis-associated cell networks

Abstract

Imaging-based, single-cell, spatial transcriptomics (iSCST) of FFPE tissue enables comprehensive analysis of archived specimens while preserving spatial context, critical to an understanding of ulcerative colitis (UC) pathology. Here, we deployed a robust framework for applying iSCST to clinical FFPE mucosal biopsies from patients with UC or immune checkpoint inhibitor–induced colitis, as well as patients serving as healthy controls. iSCST using custom Xenium gene panels enabled precise detection of diverse cell subsets and disease-specific genes. We mapped transcriptionally distinct fibroblast subsets within mucosal niches, including inflammation-associated fibroblasts (IAFs), and identified colitis-specific neighborhoods formed by IAFs, monocytes, and neutrophils. Transcriptional signatures and spatial neighborhoods uncovered through iSCST were associated with vedolizumab (VDZ) response, with nonresponders exhibiting either an innate IAF-monocyte-neutrophil signature or adaptive gut-associated lymphoid tissue signature, while responders showed enrichment of an epithelial cellular neighborhood. These signatures were validated in an internal and an external dataset, supporting the existence of 2 distinct archetypes of treatment resistance to VDZ in UC. This iSCST framework provides a powerful approach for analyzing FFPE tissues, offering insights into colitis-associated cellular networks and identifying biomarkers to enhance patient risk stratification in routine clinical workflows.

Authors

Elvira Mennillo, Madison L. Lotstein, Gyehyun Lee, Julian H. Hou, Vrinda Johri, Donna E. Leet, Christina A. Ekstrand, Jessica Tsui, Jun Yan He, Uma Mahadevan, Walter Eckalbar, Ryan M. Gill, Christopher J. Bowman, David Y. Oh, Gabriela K. Fragiadakis, Michael G. Kattah, Alexis J. Combes

×

Figure 5

Cross-dataset validation of pretreatment response and nonresponse signatures and identification of a consensus gene list.

Options: View larger image (or click on image) Download as PowerPoint
Cross-dataset validation of pretreatment response and nonresponse signat...
(A) Schematic of the response datasets (top) (created with BioRender). Mean expression of each gene signature per patient across Datasets 1 and 3 in HC, UC PRE R, and UC PRE NR (bottom). The box-and-whisker plots depict the minimum and maximum values (whiskers), the upper and lower quartiles, and the median. (B) Spatial scatterplot of representative cores for UC PRE R and UC PRE NR pseudocolored by CellCharter spatial cluster annotation (top) and proportion of cells assigned to each annotated CellCharter cluster in UC PRE R and UC PRE NR (bottom). The HS34 core shown here was also used in Figure 4E. (C) Schematic of the 3 datasets used for validation (created with BioRender). (D) GSEA NES of Xenium signatures in 3 datasets comparing UC PRE NR and UC PRE R with significance determined by *Padj < 0.1 calculated using the Benjamini-Hochberg method. (E and F) Leading-edge genes in IEC_R (E), IAF-monocyte-neutrophil NR, and GALT-B-DC-S4_fibroblast NR signatures (F) across different datasets; consensus genes across datasets are labeled in blue (E) and red (F). (G) List of consensus genes associated with treatment responsiveness. For A, pairwise 2-sided Mann-Whitney U tests were performed between groups within each gene set, with Benjamini-Hochberg FDR correction applied across all comparisons; *Padj < 0.05, **Padj < 0.01, and ***Padj < 0.001.