Microbiome Signature May Help Predict Infliximab Response in Crohn’s Disease

A substantial proportion of patients with Crohn's disease don't respond to infliximab, and a new study suggests the gut microbiome may tip investigators off before the first dose. Published in the Journal of Advanced Research, the prospective cohort study identifies baseline Bifidobacterium as a predictor of infliximab response and links it with bile acid metabolism and downstream immune regulation.

The findings point to 2 related possibilities: stool-based microbial signatures could eventually help identify patients at risk for primary non-response, and the microbiome itself may contribute to treatment sensitivity rather than simply reflecting disease activity.

Baseline Microbiota Distinguishes Treatment Response

The study included 100 treatment-naïve adults with active Crohn’s disease initiating infliximab and 49 healthy controls. Patients received standard induction therapy, and response was assessed at week 14 using a multi-parameter clinical approach integrating trough concentrations and anti-drug antibodies. Researchers performed longitudinal 16S rRNA sequencing and untargeted metabolomics on fecal samples pre- and post-treatment, then evaluated baseline features using 12 machine-learning algorithms.

Overall microbial richness did not distinguish eventual responders from non-responders at baseline, but community composition did. Non-responders showed depletion of beneficial bacteria Bifidobacteriaceae, Lachnospiraceae, and Ruminococcaceae and enrichment of potentially pathogenic bacteria Enterobacteriaceae and Streptococcaceae. At the genus level, Bifidobacterium and Blautia were enriched among responders, whereas Escherichia-Shigella was more abundant among non-responders.

The results suggest that response may depend less on how diverse the microbiome is than on which organisms—and potentially which microbial functions—are present.

Linking the Microbiome to Immune Response

Metabolomic profiling identified 179 metabolites that differed between responders and non-responders at baseline. Non-responders showed alterations in glycerophospholipid metabolism and deficiencies in primary bile acid biosynthesis, with non-responders showing lower levels of taurochenodeoxycholic acid (TCDCA). Integration of the microbiome and metabolome data added another layer. Using Model-based Integration of Metabolite Observations and Species Abundances 2 (MIMOSA2), the researchers identified Bifidobacterium catenulatum as a contributor to TCDCA production.

The investigators then explored whether differences in the microbiome could influence infliximab response experimentally. Fecal microbiota from 3 clinical responders or 3 non-responders were transplanted into antibiotic-pretreated mice with 2,4,6-trinitrobenzenesulfonic acid-induced colitis receiving infliximab. Mice receiving non-responder microbiota developed greater inflammatory injury, including more weight loss, higher disease activity, shorter colons, and worse histopathology than mice receiving responder microbiota.

The immune and metabolic findings offered a possible explanation for that difference. Non-responder FMT was associated with increased M1 macrophages and neutrophils and fewer regulatory T (Treg) cells. These mice also had lower TCDCA levels, and TCDCA was significantly associated with Treg populations. Taken together with the MIMOSA2 findings, these results support a proposed microbiome–metabolite–immune pathway in which differences in microbial composition may alter bile acid metabolism and, in turn, the immune environment associated with infliximab response.

Microbiome Data Outperformed More Complex Models

The microbiome-based linear discriminant analysis model achieved a test area under the receiver operating characteristic curve (AUC) of 0.805, compared with 0.634 for the best metabolomics-only approach and 0.779 for the best integrated multi-omics model.

Despite the relatively strong AUC, classification performance was mixed, identifying 14 of 20 non-responders, for 70.0% sensitivity, but only 10 of 30 responders, corresponding to 33.3% specificity. The authors note that multicenter validation will be needed before these microbial signatures can be translated into clinical practice.

Model interpretation nevertheless reinforced the biological findings. Bifidobacterium was the dominant microbial predictor, with greater abundance associated with a lower predicted probability of treatment failure.

Study Limitations and Clinical Implications

The study provides converging evidence for a relationship between microbial ecology and infliximab response. Still, the clinical cohort came from a single center, the predictive models used relatively small training and test samples, and 16S sequencing provides limited taxonomic resolution. Untargeted metabolomics also requires confirmation with targeted assays.

For now, neither Bifidobacterium nor TCDCA are ready to guide infliximab selection. Larger, external multicenter cohorts will be needed to determine whether these signatures generalize and improve treatment decision-making beyond established clinical and pharmacologic predictors. But the study nonetheless provides a testable framework: baseline microbial function, particularly the Bifidobacterium-bile acid axis, may be one factor shaping whether anti-TNF therapy achieves its intended immune effect.

Reference:
Liu L, Liang L, Liang H, et al. Microbiome-metabolome generated bile acids gatekeep infliximab efficacy in Crohn’s disease by licensing M1 suppression and Treg dominance. J Adv Res. 2026;83:789-806. doi:10.1016/j.jare.2025.08.017.