Colonic biopsy-associated microbial signatures are predictive of response to anti-TNFα biological therapy in Crohn’s disease

Abstract

Introduction Crohn’s disease (CD) is commonly treated with biologic therapies, including anti-TNFα agents, vedolizumab (VDZ), and ustekinumab (USTE), yet only a subset of patients respond to these treatments. This study aimed to evaluate the potential of the gut microbiome to predict treatment response. Methods Adult CD patients initiating anti-TNFα (infliximab or adalimumab), VDZ or USTE were enrolled. Pre-treatment ileal and/or colonic biopsies were collected endoscopically. Treatment response after 26–52 weeks was defined by ≥50% reduction in the simple endoscopic score for CD and either a corticosteroid-free clinical response (≥3-point HBI decrease or remission [HBI ≤4] without systemic steroids) or a biochemical response (≥50% or ≤5 mg/L CRP reduction and ≥50% or ≤250 μg/g faecal calprotectin reduction) versus baseline. Mucosal microbiota was profiled by 16S rRNA gene sequencing of biopsies. Machine learning models predicting treatment response were trained using ASV-level count data. The impact of heat-killed bacteria on anti-TNFα–induced CD14 + CD206 + macrophages was tested in mixed lymphocyte reactions (MLRs). Results A total of 125 patients were included: 39 on anti-TNFα, 47 on VDZ, and 39 on USTE. Clinical features were similar between responders and non-responders, aside from sex (USTE-colon) and CRP (USTE-ileum). No major microbial differences were observed in VDZ, USTE ileal or colon samples. However, in colonic biopsies, anti-TNFα responders had significantly higher pre-treatment α-diversity, and 3.9% of β-diversity variation associated with response. Among six models, the anti-TNFα colonic model performed significantly better than random (AUC = 0.90) to predict response. Mediterraneibacter gnavus ASVs associated with non-response, whereas Blautia ASVs associated with response, to anti-TNFα. When tested in MLRs, pretreatment with M. gnavus and B. luti led to a reduction in macrophage polarization, with a significantly stronger effect observed for M. gnavus compared with B. luti . Discussion Taken together, this study demonstrates that the colonic mucosal microbiome prior to anti-TNFα treatment can distinguish responders from non-responders in CD, supporting its potential as a predictive biomarker.

Keywords

Identifiers

DOI

10.3389/fcimb.2026.1741002

Journal

Frontiers in Cellular and Infection Microbiology

Year

2026