Baseline microbiome and metabolome are associated with response to ITIS diet in an exploratory trial in patients with rheumatoid arthritis

Changes in diet might modify the faecal microbiome and metabolomic profile, affecting pain in rheumatoid arthritis (RA). We examined the effect of an anti-inflammatory “ITIS” diet1 on clinical outcomes, gut microbiome, and metabolome in RA patients, and found that baseline faecal microbiome and metabolome composition were associated with the pain response. A prospective, open-label pilot trial was conducted to evaluate a 2-week isocaloric ITIS diet (Figure 1A) in patients with active RA. The study was approved by the Institutional Board Review. Change in pain (assessed on a visual analogue scale from 0 to 10) was the primary outcome. Patients were classified as responders (N = 7) or non-responders (N = 13), based on the achievement of a 50% improvement in pain. Amplicon sequencing was used for microbiome profiling and untargeted metabolomics for metabolite analysis. Additional methods are included in the supporting information. Twenty patients finalized the trial. Demographics and disease characteristics are summarized in Figure 1B. A diet score (212 = gold-standard) was designed to characterize patients’ baseline diet (Figure 1C and Tables S1 and 2). Patients with higher baseline disease activity had lower anti-inflammatory food scores, specifically fruit, probiotics, and anti-inflammatory spices (Figure 1D). Dietary intervention was well tolerated, and overall adherence based on the self-reported diaries was approximately 70%, except for plant protein and probiotics, with final average scores less than 60% of the gold standard (Figure 1C,E and Table S3). We also assessed adherence using a reference data-driven metabolomics approach. In large, the foods recommended increased while forbidden foods decreased post-intervention (https://assets.researchsquare.com/files/rs-654519/v1/bc74ec0e-1d08-4c67-ad53-98a73e03e3ff.pdf?c=1631886103 and Figure 1F). Outcomes significantly improved post-2-weeks of the ITIS diet (Figure 1G and Tables S4 and 5). Pain improved from 3.89 ± 1.9 before versus 2.45 ± 2.4 after diet, p < .01 (Figure 1G). No significant change in BMI was observed (Figure S1A). Although obese patients (BMI ≥ 30) had higher disease activity, outcomes scores decreased in all patients (Figure S1B,C). There were no significant BMI changes in responders and non-responders (Figure S1D). Baseline pain was similar in both groups (Figure S1E). Yet, patients that reached remission had lower DAS28CRP (Figure S1F). Total diet scores before and after intervention were not different between responders and non-responders (Figure S2A). Yet, responders had a higher baseline anti-inflammatory score than non-responders (Figure S2B). Responders also had a less negative proinflammatory score than non-responders after diet (meaning responders ate less forbidden ingredients than non-responders) (Figure S2C). Additionally, patients with a higher baseline intake of whole grains, berries, enzymatic fruits, and unsaturated fat responded better to diet (Figure S2D–F). Yet, these scores were similar in responders and non-responders post-diet. We next evaluated changes in microbiome and metabolome post-intervention. Faecal microbiome and plasma and faecal metabolome alpha-diversity didn't change over time (Figure 2A–C and Figure S3B). Changes in microbiome trajectories were very discrete, while they were more pronounced in both faecal and plasma metabolome (Figure S4). Different microbial and metabolic features increased or decreased post-diet (Figure 2D–F). Interestingly, some metabolites are microbial metabolism products, including phenylacetylglutamine, bile acids (BA), and tryptophan/kynurenine. We also evaluated if baseline microbiome or metabolome were associated with response. Baseline microbiome but not metabolome alpha-diversity was significantly higher in responders (Figure 3A–C and Figure S3A), possibly reflecting the baseline dietary differences between the two groups (Figure S2D–F). Diet explains over 25% of the microbial structural variations in humans,2 and our data suggest that response to diet might be dependent on prior diet and microbiome. The microbiome and faecal metabolome beta-diversity also showed differences between responders and non-responders (Figure 3D–F), likely reflecting the baseline differences in diet and gut microbiome. Baseline log ratios of gut microbes and metabolites, as well as plasma metabolites, were different in responders and non-responders (Figure 3G–I), suggesting their potential as predictive biomarkers of response to the ITIS diet. The log ratios Akkermansia to Collinsella/Eubacterium, Lachnospira to Blautia/Collinsella/Eubacterium and Alistipes to Blautia/Collinsella/Eubacterium were higher in R than NR. Akkermansia breaks down mucins converting them into short-chain fatty acids, with anti-inflammatory properties, whereas Blautia and Dorea were associated with other inflammatory diseases. These specific microbes might also be reflecting the baseline diet since Lachnospira was associated with vegetable intake.3 We further evaluated the relationship between the microbiomes and metabolomes changes post-intervention with the pain response. Microbiome and metabolome alpha-diversity did not change post-ITIS diet in both responders and non-responders (Figure 4A–C and Figure S3C). Gut microbiota is unique to each individual and relatively stable throughout life, yet it could be due to the short intervention of the current trial. Whether prolonged dietary changes can induce permanent alterations in the gut microbiota is unknown (review).4 Plasma metabolome beta-diversity was different at day+14 between responders vs non-responders (Figure 4D–F), suggesting that several circulating metabolites might be associated with pain post-diet. Importantly, we identified microbiome changes at the genus level (Figure 4G). Since overall adherence was similar between responders and non-responders, either the baseline microbiome of responders and/or some of the species that changed post-intervention in responders metabolized the new ingredients and shifted the plasma metabolome to a new pool of circulating anti-inflammatory metabolites. Anti-inflammatory metabolites5-7 such as BA, l-carnitine and acetyl-carnitine, kynurenine, increased more in responders than non-responders post-diet (Figure 4G–I). Moreover, another BA, deoxycholic acid, co-occurred with Akkermansia and was associated with response to pain (Figure 4J,K). Since BA and tryptophan are gut microbial products,8 microbiome differences throughout the trial seem to be associated with pain response in RA patients. Bile acids have been described to be involved in the regulation of immune cells9 and our findings require further studies to characterize their role in RA. Finally, we also detected differences in faecal and plasma abundances of some drugs that may reflect the effect of gut microbiome variations on their bioavailability.10 To our knowledge, this is the first study to describe metabolomic and microbiome changes in RA after diet intervention. Several limitations including the number of patients, the lack of a control group, the short dietary intervention, and the lack of metagenomics to better understand the faecal and plasma metabolome, warrant further studies. In addition, metabolic tracing experiments in animal models will demonstrate the transfer of microbial (supporting information) anti-inflammatory metabolites to the blood. All the authors declare no conflict of interest. Roxana Coras and Monica Guma were supported by the Krupp Endowed Fund. Roxana Coras was also supported by a UCSD Rheumatic Diseases Research Training Grant from the NIH/NIAMS (T32AR064194). Figure S1. Improvement in clinical scores is independent of changes in BMI Figure S2. Diet scores and the relation with the pain response Figure S3. Different microbiome alpha-diversity indexes (faith, evenness and observed features) in relation to pain response Figure S4. Trajectories of the microbiome (A), faecal (B) or plasma metabolome (C) between the timepoints for each patient Table S2. Baseline diet scores Table S3. Change in diet scores after diet Table S4. Clinical outcomes across the three timepoints Table S5. Number of responder/non-responder patients by different outcomes Table S6. Summary of dietary recommendations Table S7. Proposed meal organization for the 2 weeks of the intervention Table S8. Feasibility outcomes of the trial Table S9. Demographic and clinical characteristics of R and NR Table S1. Diet score calculation Table S2. Baseline diet scores Table S3. Change in diet scores after diet Table S4. Clinical outcomes across the three timepoints Table S5. Number of responder/non-responder patients by different outcomes Table S6. Summary of dietary recommendations Table S7. Proposed meal organization for the 2 weeks of the intervention Table S8. Feasibility outcomes of the trial Table S9. Demographic and clinical characteristics of responders and non-responders Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.