Introduction

While environmental risk factors for multiple sclerosis (MS) have been extensively studied, the most strongly associated determinants, such as low vitamin D status and prior infection with Epstein-Barr Virus, are not sufficient to cause MS on their own. It is likely they depend on other factors1,2. Modifiable factors such as diet and the gut microbiota are of particular interest as they can exhibit immunomodulatory and neuroprotective effects of relevance in MS3,4. The Mediterranean diet has been consistently associated with better health outcomes in the general population, including a reduced risk of mortality and lower risk of developing several chronic diseases, including some neurodegenerative conditions5. A pilot study suggested that a Mediterranean diet may alleviate MS symptoms, such as fatigue6, but whether the diet is associated with a reduced risk of MS is unclear7,8. Studies evaluating early life diet are required to determine whether populations consuming a Mediterranean diet have a lower risk of MS. A few observational studies have found a potential relationship between specific components of the diet, including higher intakes of fruits, vegetables, and fish, as well as some nutrients (e.g., iron) and lower MS risk9,10,11,12. Combined, these early findings are encouraging, with the suggestion that aspects of diet could alter MS susceptibility.

Emerging evidence suggests that a perturbed gut microbiota may be linked with MS susceptibility3. For example, animal models have demonstrated that a transplanted stool sample from a person with MS can exacerbate, as well as lower the threshold of inducing, inflammatory CNS demyelination13,14. Key differences in the abundance of various gut bacteria and archaea, such as Methanobrevibacter, have also consistently been observed between individuals with and without MS15,16,17,18,19.

The relationship between diet and gut microbiota is of growing interest as the microbiota may facilitate several health benefits3,4. Interestingly, the gut microbiota’s gene composition (metagenome) in youth with pediatric-onset MS reflects a diet depleted in fiber, whole-grains, and iron15. However, the interplay between diet and the microbiota in people with MS is largely unexplored, especially in younger individuals early in their disease course3. Children and youth with MS have fewer exposures acting across their life course compared to adults, which may allow for more insight into the role of environmental factors in MS.

Here, we investigated the trivariate relationship between diet, the gut microbiota and pediatric-onset MS (the main outcome) in a case-control study. Our specific measures of diet included the Mediterranean diet score and components of that diet, as well as intakes of fiber, whole-grains, and iron. In addition, we also tested for microbiota taxa that may mediate the association of dietary intake with MS susceptibility. We observe that a higher Mediterranean diet score and higher nutrient intakes such as fiber, are associated with a lower odds of having MS, and that the gut microbiota might mediate this potential protective relationship.

Methods

Participant selection and data sources

Participant enrolment, data sources and procurement of stool samples and DNA sequencing have been described in detail elsewhere15,19. Briefly, MS cases and controls were enrolled through the Canadian Pediatric Demyelinating Disease Network study, which included participants from seven sites—six Canadian and one USA. All MS participants were <18 years old at the time of MS symptom onset, met the 2017 McDonald criteria and, when testing was available, were negative for serum anti–myelin oligodendrocyte glycoprotein antibodies using live, cell-based methods (Oxford, UK)20,21. Unaffected controls had no known neurological or immune-related conditions, although conditions such as headache/migraine, asthma and allergies were permissible. Informed consent was obtained from all participants and/or their parent/legal guardian. The University of British Columbia (UBC) Clinical Research Ethics Board (number: H15-03330) and the Hospital for Sick Children’s (SickKids’) Research Ethics Board (number: 1000051098) approved the study.

The Diet group

Enrolled participants who had completed a Block Kids Food Screener (BKFS) before reaching 22 years of age were potentially eligible for the current study and formed the ‘Diet group.’ The BKFS is a validated 41-item food frequency questionnaire (FFQ) developed by NutritionQuest (Berkeley, CA, USA) to evaluate the habitual dietary intake of foods and nutrients in children and adolescents9. Participants or their caregivers were asked to recall the frequency and quantity of foods and beverages consumed by the participant during the previous 7 days. Participants with 15 or more missing answers on the FFQ or with implausibly low or high total energy intake (<500 or >3500 kcal/day) were excluded22.

Assessment of dietary intake and the Mediterranean diet scores

From the FFQ, the average daily intakes of food groups (grams/day) were estimated and were used to derive the widely used alternative Mediterranean diet score (aMED); considered an improved scale from the original with modifications to the dietary components based on associations with lower chronic disease risk and validated against blood biomarkers of inflammation and endothelial dysfunction23. We calculated the aMED scores based on sex-specific median intake from the controls for each of seven of the nine original dietary components, i.e., fruit, vegetables, legumes, whole-grain foods, fish, processed and red meat combined, and the ratio of monounsaturated fatty acids to saturated fatty acids (MUFA:SFA). Nuts and ethanol components were excluded as the FFQ did not capture them. One point was assigned if intake was >median, or, for processed and red meat only, if <median, and no points if =median or criteria not met (see Supplementary Table 1). Total possible scores for aMED ranged from 0–7, indicating the lowest to highest resemblance to the Mediterranean diet.

The Diet-microbiota subgroup

The ‘Diet-microbiota subgroup’ included only those participants from the Diet group who provided a stool sample within +/−90 days of completing the FFQ (representing a time period during which major dietary patterns were observably relatively stable24,25), and who had not taken antibiotics or corticosteroids before 30 days of stool collection. If a participant had more than one available stool sample and FFQ, then the earliest sample procured and FFQ were included in the analyses.

Participant characteristics

Participant characteristics, captured via questionnaires or standardized forms completed by site investigators and/or participants, included: sex, self-reported race (White, non-White) and the following characteristics at the time of the FFQ completion: age- and sex-standardized body mass index (BMI) percentiles (and overweight/obese categories)26,27, history of comorbidities, and cigarette smoking (ever, never). At the time of stool sample procurement, participants completed a 7-point Bristol stool scale, which was categorized as hard (types 1–2), medium (types 3–5), or loose (types 6–7). Site investigators captured age at MS symptom onset, disease duration, and disease-modifying drug exposure status (ever or naïve) for the MS cases (Table 1). Further details on these characteristics and their categorization are in Supplementary Table 2.

Table 1 Characteristics of the pediatric-onset MS cases and controls at the time of food frequency questionnaire completion (unless otherwise stated)
Full size table

Stool sample collection, sequencing, and bioinformatics

Participants’ stool samples were collected using a provided collection kit, and these were shipped with ice to the research laboratories and stored in a −80 °C freezer. DNA was extracted from stool samples using the Zymo Quick-DNA™ Fecal/Soil Microbe Miniprep Kit. The 16 S rRNA gene V4 region was amplified and sequenced at both ends up to 300 bp by the Illumina MiSeq platform. The paired-end reads were trimmed to 252 bp, merged, and denoised into amplicon sequence variants (ASVs) using Deblur (v.1.1.0), all via the QIIME2 (Quantitative Insights Into Microbial Ecology; v.2019.4) platform. We curated putative erroneous ASVs using LULU (v. 0.1.0)28. See Supplementary Methods for further details.

Statistical analyses

For the Diet group, we assessed the association between individual dietary measures, which served as explanatory variables, including the aMED dietary pattern scores (0–7), nutrient intakes: whole-grains (oz equiv./day), fiber (g/day), and iron (mg/day) with MS, i.e., the primary outcome variable, categorized as “MS” or “control,” all using logistic regression adjusted for age at time of FFQ completion, sex and total energy intake. Whole-grains (oz equiv./day) was calculated by determining the whole-grain component of each food item from the FFQ.

Diet-microbiota subgroup analyses

For all downstream analyses, the microbial taxon counts were centered log-ratio transformed by taking the logarithm of the ratio of each taxon to the geometric mean of all taxa from the same sample, with a pseudo-count of 0.5 added to avoid computing the logarithm of zeros29. Only those genera/ASVs with a minimum relative abundance of 0.01% in at least 10% of all samples were further assessed. We assessed the association between the gut microbiota taxa, which served as the explanatory variables, and the odds of MS, i.e., the outcome variable. All taxa were standardized to Z-scores and included the relative abundance of 201 genera and 770 ASVs, as well as the presence of the genus Methanobrevibacter. We did this using logistic regression adjusting for library size, measured as total abundance of species sequenced from an individual’s stool sample, as well as the Bristol Stool Scale group, age and sex.

We also examined the associations between dietary measures, which served as the explanatory variables, and each taxon, i.e., the outcome variable in this case. In this specific analysis, we standardized only the dietary measures to Z-scores. These included the aMED score, the aMED seven dietary components (g/day), whole-grains (oz equiv./day), fiber (g/day), and iron (mg/day) intake. We used linear regression, additionally adjusted for total energy intake. To prevent influence from outliers, all regression analyses involving taxa abundances mentioned earlier were performed using generalized linear regressions based on Mallow’s type robust quasi-likelihood estimator via the R package robustbase (v. 0.93-9, method equal to”Mqle,” default settings)30,31.

Next, we examined the possible role of the microbiota to mediate the association between diet—specifically the aMED score and fiber intake—and MS odds by conducting a mediation analysis within the potential outcomes framework via the R package mediation, v. 4.5.032,33. We chose to evaluate fiber intake alongside the aMED score given its established influence on gut microbiota composition3. In the Diet-microbiota subgroup, the abundances of gut microbiota taxa were individually tested as candidate mediators. This was only performed for taxa that exhibited significant associations (P < 0.05) with the aMED or fiber, and with odds of MS. The indirect effect, within this framework, represents the association between diet and MS odds, i.e., the outcome variable, through the mediating variable, namely an individual gut microbiota taxa. It quantifies the difference in the odds of MS when comparing the mediator variable values of 0 and 1, while holding the dietary value constant. Conversely, the direct effect represents the unmediated association of diet on the odds of MS, and quantifies the difference in odds when comparing the dietary values of 0 and 1, while holding the mediator variable constant. These two effects sum up to the total effect32. The odds ratios and their 95% confidence intervals were calculated using the Quasi-Bayesian approximation with 10,000 Monte Carlo draws. Mediation was hypothesized when the indirect effect reached a significance level of P < 0.05.

Results were reported as adjusted odds ratios (ORs) or beta coefficients, as appropriate, with 95% confidence intervals (CI) and two-sided P-values. The correlations, detailed in the complementary analysis, and beta coefficients were reported as having a strong association if the absolute value was ≥0.70 (arbitrarily assigned). High-dimensional tests were adjusted for multiple comparisons using the Benjamini–Hochberg method and reported as Q-values. All statistical analyses were conducted using R, v. 4.0.434.

Complementary analyses

We conducted complementary analyses, including, for the Diet group only: assessing the association between the seven aMED dietary component intakes (g/day) with MS. The component intakes were also dichotomized to binary scores of 0 or 1 based on the sex-specific median intakes of controls, see Supplementary Table 1, and their associations were also assessed. We also assessed the interactions of age and sex with the associations between aMED scores or nutrients and MS odds. We also examined the pairwise associations between the dietary pattern scores, their individual dietary components intakes and scores, and nutrients, using Spearman’s correlation coefficient (ρ). We also adjusted the main diet-related findings for overweight/obese status, as well as self-reported race, recognizing that race is a social construct and structural racism has adverse outcomes on many health indicators35.

For the Diet-microbiota subgroup only, we investigated the sources of inter-individual variation in the gut microbiota composition and then assessed the relationship of the diet-associated variation of the microbiota with MS risk using multivariate analyses. The details are in the Supplementary Methods.

We conducted additional analyses: first, we compared the relative abundance of microbial taxa by vitamin D supplementation and disease-modifying drug exposure status among MS cases using the Wilcoxon rank sum test; second, we re-examined the associations between dietary measures and each taxon among those participants who completed the FFQ within 6 days of stool sample collection, using linear regression.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Ninety-five participants fulfilled our inclusion criteria for the Diet group, consisting of 44 pediatric-onset MS cases and 51 controls. Among these, 59 were included in the Diet-microbiota subgroup, comprising 27 pediatric-onset MS cases and 32 controls. A flowchart depicting the participant selection process is provided in Supplementary Fig. 1. There were no significant differences in the characteristics of the MS cases and controls, as shown in Table 1, all P < 0.050. However, MS cases were, on average, two-years older than controls at the time of completion of the food frequency questionnaire (FFQ) (Wilcoxon rank-sum test, P < 0.037). Females predominated, representing 75% of MS cases and 63% of controls. All MS participants had a relapsing-remitting disease course. Of the 44 MS cases, 41% (18) were disease-modifying drug naïve. The median disease duration from MS symptom onset to completion of the FFQ was <1 year (Table 1).

Diet-only related results

A higher alternative Mediterranean diet score (aMED) score and higher intakes of fiber (g/day), whole-grains (oz equiv./day), and iron (mg/day) were each associated with a lower likelihood of having MS, albeit whole-grains did not reach significance (Fig. 1, Supplementary Fig. 2). For example, a 1-point increase in the aMED score was associated with 34% lower odds of MS (adjusted odds ratio [ORadj] 0.66, 95% CI 0.47–0.90, P = 0.011), and a 1-gram increase in fiber was associated with 13% lower odds (ORadj 0.87, 95% CI 0.78–0.96, P = 0.011).

Fig. 1: The association between the Mediterranean diet score or nutrient intake and odds of MS.
figure 1

aMED alternate Mediterranean diet score, OR odds ratio, CIs confidence intervals. The associations between the aMED score, whole-grains (oz equiv./d), fiber (g/d), and iron (mg/d) intakes and multiple sclerosis (MS) risk are represented as odds ratios adjusted for total energy intake (kcal/d), age (in years) at food frequency questionnaire completion, and sex, their 95% CIs, and P-values. A total of n = 44 MS cases and n = 51 controls were assessed. Results were similar when reassessed for the subgroup of participants with only a stool sample (n = 27 MS cases and n = 32 controls), albeit associations with iron did not reach significance (P > 0.05). The x-axes units are in log 10 scale. An OR < 1 is a “protective” association, meaning that an increase in the dietary intake is associated with a lower odds of MS, and an OR > 1 is a ‘detrimental’ association, meaning that an increase in the dietary intake is associated with a higher odds of MS. For example, an OR of 0.66 for aMED means that a 1-point increase in the aMED score for an individual is associated with a 34% lower odds of MS. Source data are provided as a Source Data File.

Full size image

Diet-microbiota related results

In the Diet-microbiota subgroup, most participants, 53 out of 59, completed their FFQ within a shorter 30-day period relative to stool collection, with the overall median interval just 6 days between FFQ completion and stool sampling. When comparing disease status, MS cases vs controls, significant differences were found for 7 genera and 12 amplicon sequence variants (ASVs), all P < 0.050, although all Q > 0.68, see Supplementary Data 1. ASVs are taxonomic units at the species-level. The relative abundances of all taxa were standardized to Z-scores and served as the explanatory variable. Among the genera examined, Methanobrevibacter abundance was the most enriched in MS cases relative to controls and a 1-standard deviation increase in its relative abundance was associated with a 5-fold higher odds of MS, ORadj 5.7, P = 0.032. Moreover, Methanobrevibacter was detected in 33% of MS cases (9/27) compared to a just 3.1% of controls (1/32). Its presence corresponded to a 23-fold higher odds of MS, albeit confidence intervals were wide, ORadj 23, 95%CI 3.0–540, P = 0.011, logistic regression. The prevalence of all other genera and species within each group, i.e., the MS cases vs controls, can be found in Supplementary Data 1. In contrast, Ruminococcaceae NK4A214 group genus was the most depleted in MS and a 1-standard deviation increase in its relative abundance was associated with a 68% lower odds of MS (ORadj 0.32, P = 0.0034). Several taxa that reached statistical significance at the ASV-level were also significant at the genus-level findings, including ASVs housed in Variovarax, Eggerthella, Lactococcus and Ruminococcaceae NK4A214 group.

Dietary measures, serving as standardized explanatory variables, exhibited a strong relationship with 3 of the 7 (43%) genera and 3 of the 12 (25%) ASVs that were identified as statistically different by disease status, i.e., differed between the MS participants and controls. These dietary measures included intakes of fruits, the ratio of monounsaturated fatty acids to saturated fatty acids (MUFA:SFA), whole-grains, fiber, and iron. The absolute beta coefficients, βadj, were ≥0.70, as shown in Fig. 2, Supplementary Data 2. The aMED score was inversely associated with the following taxa identified to be more abundant in the MS participants relative to the controls: Methanobrevibacter, Eggerthella sp. and Lactococcus sp., with βadj ranging from −0.34 to −0.52, P < 0.019. The aMED score was positively associated with the following taxa which also had lower abundance in the MS participants relative to controls: Clostridiales vadin BB60 group and Ruminococcaceae NK4A214 group sp. (βadj > 0.32, P < 0.025), the latter being most strongly associated with iron intake (βadj 1.6, P = 2.9 × 10−13). Compared to the aMED, fiber intake was more strongly associated with all of the aforementioned taxa (Fig. 2, Supplementary Data 2).

Fig. 2: The Diet-microbiota subgroup: heatmap of the relationship between dietary measures, microbial taxa, and disease status.
figure 2

MS multiple sclerosis, aMED alternate Mediterranean diet score, P&R meat processed and red meat, MUFA:SFA ratio of monounsaturated to saturated fat, sp. unknown species. a displays the 7 genera and b displays the 12 ASVs that were significantly associated with MS (P < 0.050). The association of each dietary measure Z-score with taxa relative abundance (the outcome variable) was represented as beta coefficients. The association of each taxa Z-score with disease status (n = 27 MS cases vs n = 32 controls, the outcome variable in this case) was represented as log odds ratios (far most right column). All results were adjusted for library size, Bristol Stool Scale group, age (in years) at the food frequency questionnaire completion and sex while beta coefficients were also adjusted for total energy intake. Dietary measures and taxa were standardized to Z-scores only when modeled as an independent variable. Actual estimates and associated P and Q values are available in Supplementary Table. Heatmap was created using the R package pheatmap (v. 1.0.12). Source data are provided as a Source Data File.

Full size image

A total of 4 genera and 6 ASVs, were associated with both diet (the aMED or fiber intake) and disease status (Fig. 1, Supplementary Data 2); therefore, we further tested these microbiota measures as candidate mediators of the association of aMED (and fiber intake) with MS risk. In total, seven taxa were significant potential mediators (P-values of the indirect [mediated] effect <0.045, Fig. 3), of which, Methanobrevibacter was the strongest mediator by proportion mediated. Interestingly, Anaerovoracaceae Family XIII AD3011 group sp. was a significant competitive mediator meaning that it may partially offset the protective association of aMED and fiber intake with MS risk.

Fig. 3: The Diet-microbiota subgroup: forest plot of mediation analysis for diet, potential gut microbiota mediators, and MS odds.
figure 3

aMED alternative Mediterranean diet score, sp. unknown species Participants assessed include n = 27 MS cases and n = 32 controls. The “indirect effect” (blue circle) is the association of the dietary measure with MS risk mediated by the microbiota. Mediation by the microbiota is hypothesized when the indirect effect is significant. The “direct effect” (yellow circle) is the non-mediated association of diet with MS risk. The “total effect” (gray circle) is the combined mediated (indirect effect) and non-mediated (direct effect) association of the dietary intake (aMED or fiber intake) with MS risk. Because the total effect is dependent on the mediator, the total effect changes when the mediator is different. Results are represented as odds ratio (OR) adjusted for total energy intake, library size, the Bristol Stool Scale group, age at the FFQ completion and sex. An OR < 1 is a “protective” association and an OR > 1 is a ‘detrimental’ association. Only PCs and taxa significantly associated with both diet (aMED or fiber) and MS risk were displayed and tested for mediation. All Q-values of the indirect effect were >0.10 for aMED and >0.069 for fiber. The circles and error bars represent the mean and 95% confidence intervals of the OR, respectively. Thick error bars represent a significant association with MS risk (P < 0.05). Source data are provided as a Source Data File.

Full size image

Complementary analyses results

The complementary analyses did not change our interpretation of findings. For the Diet group, although most participants reported consuming few to no legumes, an increase in the intake of this component by one-standard deviation was associated with 55% lower odds of MS (ORadj 0.45, 95%CI 0.18–0.85, P = 0.041). While the median daily intakes in grams were lower in MS cases, relative to controls, for fruits, vegetables, and whole-grain foods, and higher in processed and red meat, none reached statistical significance, all P > 0.17, see Supplementary Table 3a. However, only 25% of MS cases scored a point for whole-grain foods compared to 47% of controls (ORadj 0.40, 95% CI 0.16–0.97, P = 0.047, Supplementary Table 3b).

Age and sex did not significantly interact with any dietary measures in logistic regression models of MS risk (P > 0.067). As expected, fiber intake strongly correlates with the aMED score and fruit intake (both ρ = 0.71, Supplementary Fig. 3b). Adjusting for race and overweight/obese status strengthened the protective associations of the aMED and nutrients with MS risk, as shown in Supplementary Table 4. For example, a 1-unit increase in the aMED score was associated with an additional 5% (i.e., 0.05) lower odds of MS after additional adjustments (ORadj 0.61, 95% CI 0.43–0.84, P = 0.0040), as was consumption of whole-grains (ORadj 0.40 95%CI 0.14–0.94, P = 0.058).

For the Diet-microbiota subgroup, a trivariate relationship between MS, the overall gut microbiota and diet was identified in a stepwise approach, shown in Supplementary Figs. 4 and 5. When comparing sources of inter-individual variation in the overall gut microbiota at the genus-level, race explained the most (5.7%), followed by fruit and then fiber intakes and then the aMED score (3.0%–4.3%, all P < 0.0050, PERMANOVA), while MS itself, relative to controls, explained only 1.7%, P = 0.41, see Supplementary Fig. 4a. We generated constrained principal components (PC) that represented the inter-individual variance of the gut microbiota composition explained (that is, constrained) by the aMED or fiber intake, defined as PC-aMED and PC-Fiber, respectively. The PC-aMED or PC-fiber scores were associated with lower odds of MS (ORadj 0.41–0.45, P < 0.020, Supplementary Fig. 4b, c). The results at the ASV-level were consistent with those at the genus-level and support the same interpretation, see Supplementary Fig. 5.

Our analysis found no evidence that the association of specific microbial taxa with MS and dietary items was explained by either vitamin D supplementation or disease-modifying drug exposure status (Supplementary Tables 5 and 6).

In total, 32 participants (10 MS cases and 22 controls) had completed the FFQ within 6 days of stool sample collection. Our findings were largely consistent with the original analysis. For example, fiber was strongly associated with Eggerthella, Lactococcus, Clostridiales vadin BB60 group and Ruminococcaceae NK4A214 group (the absolute beta coefficients, βadj, were ≥0.90, P < 0.00050). However, Methanobrevibacter was no longer associated with aMED or fiber intake (P > 0.05), likely attributable to its limited prevalence as it was detected in only 3 of the 32 remaining participants, all of which had MS.

Discussion

A higher Mediterranean diet score and higher intakes of specific nutrients were associated with a lower likelihood of having pediatric-onset MS. We demonstrated that the gut microbiota might facilitate this protective association. Key findings included that a higher alternative Mediterranean diet score (aMED) and higher intakes of fiber and iron were each associated with a lower likelihood of having pediatric-onset MS. Consistent with previous observations, relative to controls, MS cases were enriched with the archaeon Methanobrevibacter15,36,37,38—a gut-dominant methane-producer associated with constipation39, a common gastrointestinal complaint in individuals with MS40—and Eggerthella41,42. Intriguingly, the presence of Methanobrevibacter in our cohort was associated with a significantly higher likelihood of having MS, albeit the 95%CIs were wide (ORadj 23, 95%CI 3.0–540). Furthermore, our MS cases were depleted of Clostridiales vadinBB60 group and Ruminococcaceae NK4A214 group19, the latter is a genus from a short-chain fatty acid (SCFAs)-producing family43. Immunomodulatory SCFAs are prominent bacterial products of fiber fermentation44. Our study is unique in that it links dietary measurements such as the aMED and dietary fiber to these MS-associated taxa. Notably, these particular taxa were significant potential mediators of the association of both the aMED and fiber intake with the risk of MS. Fiber intake and the aMED score were significant sources of inter-individual variation in the overall gut microbiota, while age, sex, and MS (vs controls) were not, indicating that diet may have had an important impact on the gut microbiota composition. Remarkably, the specific aMED- and fiber-associated variations of the microbiota in turn were associated with lower risk of MS. Together, findings suggest the interplay between the host, microbiota and diet are likely relevant in MS.

We found just two other published studies investigating the Mediterranean diet with MS risk. Both studies involved adults exclusively and neither included the gut microbiota7,8. Consistent with our findings, a cross-sectional study from Iran compared 69 MS cases and 140 controls (82% women) and reported that the highest versus lowest tertile on the Mediterranean diet score was associated with a 77% reduced risk of MS risk7. However, a large longitudinal study of over 185,000 US female nurses, including 480 individuals with MS, did not observe an association between the aMED scores (assessed as quintiles) and the risk of developing MS8. This US-based study was the only one to include alcohol as a (healthy) component of the Mediterranean diet, while recent evidence suggests that it is not healthy45. Emerging evidence suggests that the Mediterranean diet may help modify MS course and related symptoms. A randomized controlled pilot trial and two cross-sectional studies reported a higher Mediterranean diet score among MS cases was associated with reduced disease severity, disability, fatigue and depression6,46,47. Together, these findings warrant further investigations into the potential health benefits of a Mediterranean diet for individuals with or at risk for MS.

Motivated by our prior research conducted on a considerably smaller cohort from the same source population15,48, we assessed the nutritional intakes of fiber, whole-grains, and iron. We observed that consumption of these nutrients were lower in the diets of MS cases (relative to controls) and higher intakes were associated with reduced likelihood (odds) of having MS. Consistent with these findings, lower nutritional intakes of fiber and iron by MS cases compared to controls were observed in two studies: a US multi-center study of 312 pediatric-onset MS cases and 546 controls and an Israel-based study of 63 adult MS cases and 83 controls9,49. Similarly, an Iran-based study also observed a lower dietary fiber intake among 77 MS cases compared to 148 controls50. While no other study has assessed the intake of whole-grains as a food component in MS cases, the consumption of whole foods that are primarily composed of whole-grain—a healthy component of the aMED—has been previously evaluated. Although the direction of their findings was similar to ours, none reached significance, at least when adults were asked to recall their diet from ages 6 to 20 years11. We were able to show a significant relationship with a high intake of whole-grain foods associated with a 60% lower odds of MS. Interestingly, whole-grain food consumption was found to have an inverse relationship with MS disability in a cohort of participants enrolled in the North American Research Committee on MS (NARCOMS) study51. Combined, evidence suggests that a diet rich in fiber, whole-grains, and adequate iron—found in a Mediterranean diet—may be protective for MS onset.

Consistent with previous observations, the consumption of other healthy components of the aMED were also associated with a reduced risk of MS in our study, such as legumes11,50, fruit7,11, and vegetables7,50, but only legumes reached significance. Surprisingly, MS participants from the International Multiple Sclerosis Microbiome Study (iMSMS) consumed more fruits and vegetables, and scored better overall on a healthy eating index, when compared with household controls, who were predominately their spouses18. These counterintuitive and seemingly contradictory findings may be attributed to reverse causation, as those affected by a disease such as MS may later decide to adopt a healthier diet consisting of fruits and vegetables52. Participants in the study averaged 50 years of age and had a long MS disease duration (averaging 14 years). Sex differences between cases and controls may also contribute, as most MS cases were women, while most household controls were men. Women are reported to consume more nutritious diets than men53.

Diet is an important determinant of microbiota composition, and diet-induced shifts in gut microbiota composition have been linked to health benefits54. Findings from our correlation and mediation analyses indicate that dietary habits may explain differences between MS cases and controls for several gut taxa. These findings also concur with the dietary intervention studies in animals and humans. For example, animals diets high in fiber such as resistant starch and grains were associated with a reduced abundance of Eggerthella55, and Methanobrevibacter56,57. These genera were also enriched in our MS cases. Interestingly, reduced carbohydrate and fiber consumption rapidly increased the abundances of Eggerthella and Lactococcus58. Similarly, a short-term, 1–3 day reduction of carbohydrate and fiber intake resulted in an increased abundance of Eggerthella and Lactococcus in a study of ten participants58. Studies in animals have shown that Ruminococcaceae NK4A214 group abundance—depleted in our MS cases—increased after a high grain diet in cows and decreased after a low iron diet in mice59,60. Interestingly, we observed that an ASV from the R. NK4A214 bacterial group was associated most strongly with dietary iron followed by fiber intake. Also consistent with our findings, a high fat, high animal protein diet in rats was associated with an increased abundance of the Anaerovoracaceae Family XIII AD3011 group61. Intriguingly, our mediation analysis suggests that an ASV from the A. Family XIII AD3011 group, which was depleted in our MS cases (relative to controls), may partially offset the protective association of the aMED or fiber intake with MS risk. These findings warrant further investigation into the potential of the gut microbiome to mediate the effects of diet on MS risk.

Others have also identified putative inter-relationship(s) between diet, the gut microbiota and MS, but primarily using correlation and network analysis3,18,36,62,63. For example, a study linked meat servings with lower gut Bacteroides thetaiotaomicron abundance (a fiber digesting bacterium), higher circulating T-helper 17 cell and higher abundance of meat-associated blood metabolites in 185 MS cases relative to 330 controls62. Also, the iMSMS study observed an overlap between the gut microbiota taxa associated with MS and intakes from several food groups, especially fruits18. Although many of these studies are modest in size or provide indirect evidence, together with our present work, they suggest that diet-associated alterations of the gut microbiota may play a role in MS pathogenesis.

Pediatric-onset MS is a rare disease, with most studies reporting an estimated annual incidence of <1 per 100,0000 population. Hence although our study size was modest, it represents a valuable and unique cohort that could be highly informative64,65. Nonetheless, cases and controls were demographically well balanced and similar for stool consistency, an important confounder in gut microbiota studies and seldom reported in the MS literature66. Furthermore, our MS cases had a short disease duration, averaging <1 years from symptom onset (at time of FFQ completion), and thus would have incurred relatively few competing (or confounding) exposures, such as comorbidities, polypharmacy and other lifestyle changes that occur when living with chronic disease, especially when compared to adults with MS. Our study had limitations. First, the study was cross-sectional in design and data collection occurred after disease onset, and therefore, a causal relationships between diet, the microbiota, and MS could not be determined. Second, guided by the literature, we allowed a +/− 90 day window between completion of the FFQ and stool sample procurement as substantial dietary changes are unlikely to have occurred during this time24,25. While it is possible that a shorter time window may be preferable, most participants (90%) completed their FFQ within a narrower 30-day period. Nonetheless, even when participants with the shorter window of +/− 6 days between the FFQ and stool sample collection were examined, findings remained largely consistent and a strong fiber-microbiota correlation remained. Third, we cannot exclude the possibility that individuals with MS changed their dietary habits after symptom onset or diagnosis, although, this is unlikely as disease duration (from symptom onset) was short. Fourth, we lacked information on nuts and alcohol intake and therefore could not compute the aMED score based on all of its original healthy components. However, alcohol use in our young cohort would be expected to be low45. Fifth, we did not have access to participants’ physical activity levels, sunlight exposure or serum vitamin D levels which prevents us from evaluating their possible impact on the relationship between diet, the gut microbiota and MS risk. Nonetheless, to our knowledge, no study has demonstrated a link between physical activity and subsequent risk of developing pediatric-onset MS. Sixth, our modest cohort size limited our ability to draw robust conclusions, especially for specific microbial taxa, such as Methanobrevibacter, which was infrequently detected. This size limitation could reduce our ability to uncover true relationships. It also remains possible that some of our findings could have occurred by chance alone. These constraints collectively necessitate a cautious interpretation of the findings. Despite these limitations, this study provides valuable insight into the diet-microbiome-MS relationship. In particular, we identified key potential dietary and microbial risk factors for MS which generate intriguing hypotheses regarding their underlying mechanisms. Furthermore, based on our mediation analysis, we hypothesize that the potential protective effect of a fiber-rich nutritious diet against MS may be attributed to its potential influence on the gut microbiota. To help determine if this hypothesis reflects a causal relationship, conducting a large longitudinal study would be of value.

To conclude, a higher score on the alternative Mediterranean dietary metric, and higher intakes of fiber and iron, were significantly associated with a reduced likelihood of pediatric-onset MS. Furthermore, a protective relationship was identified between the aMED- and fiber-associated variation in the gut microbiota, in terms of both individual taxa abundance and overall composition, and MS risk. Combined, our findings suggest that an interaction between the host, the gut microbiota and a healthy diet rich in fiber such as the Mediterranean diet, is likely important in MS. Further work is warranted to characterize the diet-microbiome-MS trivariate relationship.