Microbiome may hold key to methotrexate response in RA Save

Methotrexate (MTX) remains the cornerstone of rheumatoid arthritis (RA) treatment, yet it remains unclear as to how to predict who will respond to it. Two studies presented at ACR (Abstracts 2639 and 2643) suggest the answer may be hiding in the gut.

In the first study (Abstract 2639), Vinod Gupta and colleagues analysed stool samples from treatment-naïve RA patients before they began MTX. Using whole-genome metagenomic sequencing, the team compared the microbiomes of responders and non-responders, defined by EULAR response criteria, after three months of treatment.

Five microbial species and nine metabolic pathways differed significantly between the two groups, including Bacteroides fragilis and Eubacterium species, that were strongly enriched among responders. The presence of any of the top ten differentially prevalent microbes increased the odds of MTX response tenfold. Interestingly, genes encoding dihydrofolate reductase, the enzyme directly targeted by MTX, were less abundant in responders, an observation that may point to functional interactions between host and microbial folate metabolism.

To test clinical relevance, the Mayo team trained a machine learning model that combined baseline microbiome and clinical data. The model predicted MTX response with impressive accuracy (AUC 0.83), outperforming clinical data alone (AUC 0.72). Even when tested in an external validation cohort, the model still worked (AUC 0.77). Microbiome profiling could therefore guide treatment selection before MTX failure occurs.

A second, larger study (Abstract 2643) independently confirmed that the pretreatment gut microbiome differs between MTX responders and non-responders, and that these differences are reproducible across diverse patient populations.

Rahul Bodkhe and colleagues analysed three independent cohorts of treatment-naïve RA patients (n=86) using metagenomic sequencing, harmonising the data across studies to identify consistent microbial and functional signatures of methotrexate response. They found 95 microbial strains that differed between responders and non-responders, with non-responders showing marked depletion of key commensals including Faecalibacterium, Clostridium, and Bacteroides.

Non-responders had reduced abundance of 231 microbial genes involved in pyrimidine and purine metabolism, histidine synthesis and folate-related pathways, many of which overlap with MTX’s known mechanisms of action. These microbial “footprints” were then used to train a predictive model that successfully classified patients by response status, underscoring the biological consistency of these gut signatures.

Taken together, these two studies move us a step closer to precision prescribing in RA. Instead of waiting months to see if MTX will work, clinicians could one day send a stool sample for microbiome profiling and receive a probability score of likely response.

The findings also raise fascinating mechanistic questions. Could altering the microbiome improve MTX efficacy? Would probiotics, dietary interventions or even antibiotics modify treatment outcomes?

The concept of using microbial biomarkers to tailor DMARD therapy fits within the growing movement toward data-driven, individualised rheumatology care. With large, multi-cohort validation and the integration of machine learning, the next step is to test these models prospectively and determine whether microbiome-guided treatment decisions improve outcomes.

Both Abstracts 2639 and 2643 exemplify a new generation of translational research that blurs the boundary between bench and bedside. MTX may still be the anchor drug, but the gut microbiome could become the anchor biomarker.