Background
Despite widespread use, the Marsh classification for celiac disease (CeD) has poor discrimination for CeD severity, inter-observer variability, and poor correlation with clinical outcomes. Intestinal spatial profiling of gene expression can enhance resolution of cellular and molecular architecture over histology alone, which may improve CeD diagnostic classification, identify underlying susceptibility pathways, and inform clinical outcome predictions.
Methods
We performed imaging-based in situ spatial transcriptomics on duodenal biopsies from clinically-indicated endoscopies in 111 adults with CeD across the Marsh score spectrum and 29 healthy controls. An experienced pathologist assigned a Marsh score and marked representative areas on the biopsy. We constructed tissue microarrays using 1mm cores with two samples per participant, which were then run on a Xenium analyzer using a custom 481 gene panel capturing cell identities, broad immune pathways, and genes relevant to CeD in prior studies. We performed cell type-specific pseudobulk differential expression and Cellcharter-based spatial neighborhood analyses.
Results
After quality control filtering, our dataset included 1.8 million cells from 47 cell types across a spectrum of Marsh scores. Spatial neighborhood analysis identified early compositional changes in the lamina propria (false discovery rate [FDR]<0.1 for Marsh 0-3) and later detectable changes in epithelial neighborhoods (FDR<0.05 for Marsh 1-3). Cell type-specific pseudobulk differential expression showed substantial increase (>3 fold) in interferon and cytotoxic cellular programs in intraepithelial lymphocytes along with loss of coordinated metabolic and differentiation programs in enterocytes in Marsh 3A-C vs controls.
Conclusions
In a large-scale duodenal spatial transcriptomics study in CeD, we show significant upregulation of interferon and cytotoxic programs in intraepithelial lymphocytes with concurrent reduction of metabolic and differentiation pathways in the epithelial lineage. We spatially localize early CeD changes to the lamina propria, with progressive changes across all neighborhoods with increasing severity. These findings could help improve diagnostic capabilities and more personalized prognoses.