Document resource
Background Plasma cell infiltration is a hallmark of refractory IBD, yet whether pre-treatment mucosal plasma cell burden independently predicts anti-TNF response remains uncharacterised. We developed a transcriptomic Plasma Cell Score (PCS), evaluated its predictive utility for infliximab response, and benchmarked it against complementary biomarkers in a multi-score comparison.Methods Transcriptomic data from 616 colonic mucosal biopsies across four GEO cohorts (GSE75214, GSE179285, GSE16879, GSE36807) were integrated using ComBat batch correction and subjected to unsupervised consensus clustering identifying three molecular subtypes. A multi-gene PCS was derived from plasma cell differentiation and effector markers (MZB1, XBP1, PRDM1, IRF4, CD38, TNFRSF17, CD79A, CD27, and five additional markers). PCS was benchmarked against a 47-gene Fibrosis Risk Score (FRS) and 13-cell-type Total Immune Burden Score (TIBS) for infliximab response prediction in pre-treatment biopsies (GSE16879; n=61) by ROC analysis, including a combined panel.Results Consensus clustering identified three subtypes with markedly distinct plasma cell burdens the Inflammatory-EMT subtype showing highest PCS (0.726±0.522), while Hypoxia-Metabolic (-0.398) and Metabolic-Absorptive (-0.293) subtypes were plasma cell-depleted ( IDDF2026-ABS-0265 Figure 1. Plasma cell score distribution by molecular subtype). Plasma cell marker expression confirmed this subtype stratification, with MZB1 as the strongest individual non-responder marker (log2FC=0.640, p<0.001), followed by CD38, IRF4, and TNFRSF17 (IDDF2026-ABS-0265 Figure 2. Plasma cell marker expression heatmap by molecular subtype and infliximab response). PCS projected continuously across molecular subtype space on UMAP, highest in the inflammatory cluster (IDDF2026-ABS-0265 Figure 6. Plasma cell score projected onto UMAP of all 616 samples).Pre-treatment PCS significantly predicted infliximab non-response (AUC=0.676, p=0.009; IDDF2026-ABS-0265 Figure 4. Pre-treatment plasma cell score in infliximab responders vs non-responders), with non-responders carrying significantly higher pre-treatment plasma cell burden (p=0.0086). However, head-to-head benchmarking identified FRS as the strongest individual predictor (AUC=0.829), followed by TIBS (AUC=0.805) and PCS (AUC=0.676) (DDF2026-ABS-0265 Figure 3. Comparative ROC curves for PCS, FRS, TIBS and combined score for infliximab non-response prediction). Critically, combining all three scores yielded no improvement over FRS alone (combined AUC=0.801), confirming fibrotic programming as the dominant non-redundant driver of treatment resistance. At the combined score threshold, 85% of high-score patients were non-responders versus 31% of low-score patients, a 4.5-fold difference in response rates (IDDF2026-ABS-0265 Figure 5. Infliximab response rates by combined score group).Conclusions Plasma cell infiltration is a significant but insufficient standalone predictor of anti-TNF non-response, with MZB1 as the single most informative marker. Fibrotic programming outperforms both plasma cell and immune burden scores, and score combination yields no additive benefit positioning the Fibrosis Risk Score as the optimal pre-treatment transcriptomic biomarker for biological therapy selection in IBD.Abstract IDDF2026-ABS-0265 Figure 1Abstract IDDF2026-ABS-0265 Figure 2Abstract IDDF2026-ABS-0265 Figure 3Abstract IDDF2026-ABS-0265 Figure 4Abstract IDDF2026-ABS-0265 Figure 5Abstract IDDF2026-ABS-0265 Figure 6