Modulation of multiple sclerosis risk and pathogenesis by the gut microbiota: complex interactions between host genetics, bacterial metabolism, and diet

SUMMARY

Multiple sclerosis (MS) is an autoimmune disease of the central nervous system, affecting nearly 2 million people world-wide. The etiology of MS is multifactorial: approximately 30% of the MS risk is genetic, which implies that the remaining ~70% is environmental, with a number of factors proposed. One recently implicated risk factor for MS is the composition of the gut microbiome. Numerous case-control studies have identified changes in gut microbiota composition of people with MS (pwMS) compared with healthy control individuals, and more recent studies in animal models have begun to identify the causative microbes and underlying mechanisms. Here, we review some of these mechanisms, with a specific focus on the role of host genetic variation, dietary inputs, and gut microbial metabolism, with a particular emphasis on short-chain fatty acid and tryptophan metabolism. We put forward a model where, in an individual genetically susceptible to MS, the gut microbiota and diet can synergize as potent environmental modifiers of disease risk and possibly progression, with diet-dependent gut microbial metabolites serving as a key mechanism. We also propose that specific microbial taxa may have divergent effects in individuals carrying distinct variants of MS risk alleles or other polymorphisms, as a consequence of host gene-by-gut microbiota interactions. Finally, we also propose that the effects of specific microbial taxa, especially those that exert their effects through metabolites, are highly dependent on the host dietary intake. What emerges is a complex multi-faceted interaction that has been challenging to disentangle in human studies, contributing to the divergence of findings across heterogeneous cohorts with differing geography, dietary preferences, and genetics. Nonetheless, this provides a complex and individualized, yet tractable, model of how the gut microbiota regulate susceptibility to MS, and potentially progression of this disease. Thus, we conclude that prophylactic or therapeutic modulation of the gut microbiome to prevent or treat MS will require a careful and personalized consideration of host genetics, baseline gut microbiota composition, and dietary inputs.

INTRODUCTION

Multiple sclerosis (MS) is a complex multifactorial disease of the central nervous system (CNS). Affecting nearly 2 million people worldwide and approximately one million people in the United States alone, MS is the most common cause of neurological disability in young adults^1,2. Disease etiology stems from environmental risk factors, including Epstein-Barr virus infection, reduced levels of vitamin D3, low UV radiation exposure, cigarette smoking, obesity, and diet, contributing approximately 70% disease-risk, which in a genetically susceptible host, conspire together to elicit CNS autoimmunity^3–8. A recently proposed environmental risk factor for MS is the gut microbiome, which is a complex ecosystem composed of approximately 100 trillion microbes, including bacteria, fungi, archaea, and viruses, encoding up to 3 million genes and producing thousands of unique metabolites within the gastrointestinal (GI) tract^9–11. Notably the GI tract represents the largest immune organ in the body, housing a diversity of immune cell types, all in close proximity to the gut luminal microbiota¹². Consequently, commensal gut microbiota have an established role in regulating the development, homeostasis, and function of our innate and adaptive immune systems, both in the gut and systemically¹². Moreover, bacteria have a known role in regulating host digestion, and in producing nutrients and metabolites from host dietary inputs, including bile acids, lipids, amino acids, vitamins, and short-chain fatty acids (SCFAs)¹³. More than 10% of host systemic circulating metabolites are bacterial-derived, at times reaching micromolar concentrations in the blood. These metabolites have far reaching effects, including modulation of blood-brain barrier (BBB) integrity and/or directly crossing the BBB to impact CNS resident neuronal and immune cells^14,15. As a result, the gut microbiota are proposed to modulate disease in CNS autoimmunity in two major ways: 1) through direct interactions with cells of the immune system, modulating their maintenance and function, and 2) via indirect mechanisms through the metabolites they produce.

Host genetics have a clear role in mediating susceptibility to MS, as evidenced by familial linkage analyses and genome wide association studies (GWAS). Moreover, the genetics of the host is known to alter the compositional nature of the gut microbiota and ultimately shape microbial interaction with and impact on the host¹⁶. Further, dietary inputs from the host have a direct impact on gut microbiome composition and its metabolic output, both of which are also influenced by host genetics. Understanding the complex interactions between genetics, the gut microbiome, and diet as influencing CNS autoimmunity remains challenging. Here, we summarize the recent insights from human studies, as revealed by GWAS, as well as large-scale case-controlled studies detailing the specific changes in the gut microbiota in the MS disease state. Furthermore, we review the mechanistic knowledge gained from basic science research studies, wherein the gut microbiota, diet, and genetics can be carefully controlled, which have begun to reveal context-specific molecular mechanisms influencing both predisposition to initiation of disease, as well as factors contributing to disease progression. Future efforts should further elucidate such mechanisms in reductionist models, incorporating multi-omics approaches to define the host genetic determinants, bacterial species, metabolites, and relevant immunological timeframes. These efforts will enhance targeted development of therapeutics aimed at modifying the gut microbiota to improve disease outcome.

Genetic determinants and heritability of MS

The genetic basis for MS is well established through studies assessing familial risk, disease prevalence among ancestral groups, and more recently, detailed GWAS, defining the risk loci and genomic regions of interest to garner insight into the functional consequences pertinent to disease etiology. Early studies showed clustering of MS-affected individuals occurring within families, with a 2–5% increased lifetime risk of developing disease among relatives of pwMS¹⁷. Furthermore, there is a high degree of disease concordance among monozygotic twins at a striking 25–30%, which is reduced to 6% for dizygotic twins, dropping to 2–4% among siblings¹⁸.

The strongest genetic associations with MS risk are polymorphisms within human leukocyte antigen (HLA) genes encoding major histocompatibility complex (MHC) class I, involved in antigen presentation to CD8+ T cells, and MHC class II, involved in antigen presentation to CD4+ T cells, located on the short arm of chromosome 6¹⁹. The most prominent variant associated with MS, is HLA DRB1*15:01, resulting in a 3-fold increase in disease risk²⁰. Initial characterization of variants outside the MHC identified polymorphisms predominantly related to both the innate and adaptive immune system. These include those within interleukin-2 receptor alpha gene (IL2RA; reducing IL-2 receptor signaling)^21,22, interleukin-7 receptor alpha gene (IL7R; altering the ratio of the soluble and membrane bound receptor)²³, leukocyte function-associated antigen 3; LFA-3 (CD58; modifying regulatory T cell function), interferon regulatory factor 8 (IRF8; modulating the expression of type I interferon related genes), and tyrosine kinase 2 (TYK2; modulating T helper (Th)17 differentiation^24–29, among others.

The recent MS GWAS conducted by the International Multiple Sclerosis Genetics Consortium included 47,429 pwMS and 68,374 healthy controls expanding our understanding of the genetic factors controlling MS susceptibility³⁰. Over 200 risk loci, in 156 distinct genomic regions were identified (including replication of the early candidate genes described above), as well as a single variant on chromosome X, and 32 variants in the extended MHC. Combined with epidemiological evidence, these association studies, bolstered by the basic science mechanistic studies they prompted, underscore the importance of genetics in the etiology of MS, and strongly implicate genetic variants that generally promote an activated state of the immune system as a major predisposing factor to CNS autoimmunity (see Fig. 1).

Figure 1. Complex interactions between host diet and the gut microbiota modulate CNS autoimmunity in a genetically susceptible host.

Schematic depicting major genetic drivers of gut microbiome composition (top left) and genetic risk factors for MS (top right), as well as key commensal microbes of the MS gut microbiome (middle left), the metabolites they are known to produce (center) derived from host dietary substrates (bottom) to ultimately modulate the host immune system and drive CNS autoimmunity.

Microbiome in MS: lessons from case-control studies

There have been numerous case-controlled observational reports cataloguing the compositional changes present in the gut microbiome of pwMS. While initial studies often consisted of relatively small heterogenous cohorts, several alterations in the MS gut microbiome were concordant between studies. Importantly, most studies report no broad differences in overall community structure or diversity in MS, while differences among individual taxa were observed. Perhaps one of the more consistent findings in the MS gut microbiome is a depletion in SCFA-producing microbiota, although the specific taxa contributing to this functional difference often differs depending on the study cohort. SCFA-producing Bacteroidaceae, with varying degree of taxonomic resolution, are diminished in pwMS and are further linked to a reduction in pathogenic Th17 T cells^31–34. Interestingly, glatiramer acetate treatment restores abundance of Bacteroidaceae to levels observed in healthy controls, suggesting that disease modifying therapy (DMT) usage may positively impact the composition of the gut microbiome³³. Additional SCFA-producing microbiota displaying altered abundance in pwMS include both Ruminococcaceae and Clostridial clusters XIVa and IV^31,33,35,36. Notably, Clostridia have also been associated with fatigue and increased disease severity in pwMS, suggesting that functional differences among specific clades of Clostridia may divergently impact CNS autoimmunity³⁷.

Altered Akkermansia muciniphila (A. muciniphila) abundance is another hallmark of the MS gut microbiome^33,38–44. Although increased abundance of A. muciniphila is consistently observed in multiple studies, it has been suggested that rather than functioning as a disease driver, this may represent a compensatory mechanism whereby the host selects for enrichment of beneficial microbiota in autoimmune disease³⁷. Prevotella as a genus is consistently diminished in pwMS, inversely correlating with disease severity^{31,32,35,38,39}. Moreover, two independent studies identified the same Prevotella species, Prevotella copri, to be depleted in pwMS, and DMT treatment was sufficient to restore Prevotella abundance^31,35. This suggests that Prevotella, as a genus, may serve an anti-inflammatory or otherwise protective role in MS. Some studies have also reported alterations in Parabacteroides, Blautia, and Dorea, as well as increases in Actinobacteria or its genus member Bifidobacterium, Streptococcus, Desulfovibrionaceae, and Coriobacterium, all of which have been implicated in the maintenance, maturation, and inflammatory polarization of immune cell subsets^{31,32,35,36,38,41,45–48}. Notably, in addition to limited cohort size, the lack of consistency between initial studies characterizing the gut microbiome may stem from differences in sample collection, processing, and/or analysis, divergent demographics, MS subtypes, DMT usage captured between cohorts, and/or inability to adequately account for additional environmental factors, including diet, that impact the gut microbiota and/or CNS autoimmunity.

To reconcile divergent findings between initial studies of the gut microbiome and pwMS, the International Multiple Sclerosis Microbiome Study (iMSMS) consortium conducted the largest study to date, including 576 pwMS and an equivalent number of genetically unrelated household controls⁴⁰. Consistent with most other studies of the gut microbiome and CNS autoimmunity, species richness (i.e. alpha diversity) was not different in the context of MS. Notably, a modest but significant difference in overall bacterial community composition (i.e. beta diversity) was observed between pwMS and controls, both within the entire cohort and when segregating by DMT usage or among MS subtypes. This suggests that prior studies may not have been sufficiently powered (or properly controlled) to detect small but consistent alterations in gut microbiome community structure that a well-powered study was able to capture. Importantly, the iMSMS cohort also evaluated the impact of MS subtype, disease severity, and DMT usage on the gut microbiota.

Comparing untreated pwMS to healthy controls, major alterations among specific taxa included increases in A. muciniphila and Ruminococcus with a decrease in Blautia, Clostridium, and Faecalibacterium prausnitzii (F. prausnitzii). Interestingly, bacterial metabolic pathway analysis linked increased A. muciniphila abundance to overrepresentation of the phytate degradation pathway leading to production of the immunomodulatory metabolite, myo-inositol, which was previously reported to be increased in pwMS⁴⁹. While the data require further functional study, they suggest a potential mechanism whereby increased A. muciniphila abundance exerts a protective role in MS. Moreover, depletion of Faecalibacterium in MS has been previously reported, with vitamin D supplementation sufficient to restore its levels to that observed in healthy controls³³. When stratifying pwMS into disease subtype specific groups, although similar trends were observed, there were no significant changes between individuals with relapsing-remitting disease and healthy controls, although this may be due to the heterogeneity of disease course present in relapsing-remitting MS (RRMS). Notably, in progressive disease a more significant increase of Ruthenibacterium lactatiformans, Hungatella hathewayi, and Eisenbergiella tayi with a decrease of Fusicatenibacter saccharivorans, and F. prausnitzii was evident. These data suggest that while there may be overlap between the gut microbiota that are altered between MS-subtypes, unique compositional changes are also present and may represent taxa that are specifically associated with divergent inflammatory mechanisms reflective of each disease course, disease specific symptomology, and/or disease progression.

Correlating alterations in the gut microbiome with disease severity revealed significant associations in both remitting and progressive disease, including a consistent negative association between abundance of SCFA-producing taxa and disease severity in both RRMS and progressive MS. A profound impact of DMT usage was also reported in people with relapsing-remitting disease, promoting restoration of A. muciniphila and Parabacteroides species to levels observed in healthy controls, as has been reported previously^33,39. While these changes may reflect DMT as a confounder in evaluating the gut microbiota in pwMS, they may also reflect a return to baseline or healthy control abundance levels with appropriate therapeutic intervention.

To assess the role of diet, as a major potential influence on the gut microbiota, a food frequency questionary was also administered to participants in the iMSMS study and used to calculate a Healthy Eating Index (HEI-2015) as a quantifiable metric of overall diet quality. While diet influenced numerous microbial taxa, only a limited number of these were MS-associated including an enrichment in Ruminococcus torques, which was negatively correlated with salt intake, and F. prausnitzii, which was positively correlated with fruit. However, sodium intake was consistent between pwMS and healthy controls, and fruit intake was higher in pwMS. Other observed associations included expansion of Eubacterium eligens with provision of dietary fiber, and an association between higher abundance of Faecalibacterium, Eubacterium and Blautia species and intake of whole grains as well as a correlation between Alistipes abundance and an overall healthier diet. While none of these associations were specific to MS, they do highlight the prominent role of host diet in modulating the composition of the gut microbiome. Collectively, along with prior studies, the iMSMS study extends our knowledge of the compositional changes that occur in the MS gut microbiome, highlighting those alterations that appear to be dependent on diet, disease course, and/or treatment status (see Fig. 1).

Importantly, these studies have two major caveats. The first, which is inherent to their case-control design, is the inability to segregate changes in the gut microbiota that are secondary to disease progression from those that are causative of disease initiation and/or progression. To address this question, longitudinal or prospective studies are necessary: either longitudinal studies in pwMS in early stages of disease progression, or prospective studies in a healthy population at high risk for developing MS. While the former are highly resource intensive, cohort studies of this type have been initiated^50,51. As to the latter, a few smaller longitudinal studies have recently begun to identify microbial changes associated with disease progression in MS, revealing interesting insights that are in partial agreement with prior case-control studies, such as the finding that increased A. muciniphila abundance is associated with lower risk of progression^52–54. The second major caveat is a lack of examination of host genetic variation as a co-factor. The iMSMS study highlighted a major role for geographical location, which may at least in part reflect distinct genetics at each geographical site. In principle, host genetic variation can be assessed using genome-wide genotyping arrays as in GWAS studies and integrated as another co-factor in the previous or new case-control microbiome studies. The other approaches to account for host genetics have included a twin study design and using ethnicity as a proxy for genetic variation, both of which are discussed in detail in subsequent sections.

Towards causation and mechanism: a perspective on the interpretation of microbiome studies using animal models

While association studies in humans generate ample hypotheses, confirmation of causation and determination of mechanism require basic research and animal models. The primary autoimmune animal model of MS, experimental autoimmune encephalomyelitis (EAE), represents a tractable disease model in which to interrogate host genetics, the gut microbiome, dietary influences, their interactions as predisposing factors to MS risk, and to discern their impact on disease severity and/or progression. Notably, some of the first evidence for a role of the gut microbiome in autoimmune disease originates from such mouse models. Germ-free mice, bred and maintained in isolators, preventing all exposure to and colonization by microbiota, are immunologically immature. They display a marked reduction in proinflammatory Th17 cells with a bias towards Th2 responses⁵⁵. Moreover, germ-free mice exhibit reduced EAE severity as compared to mice colonized with a normal complement of gut microbiota, corresponding with lower levels of proinflammatory cytokines, including interferon (IFN)-γ and IL-17, in the intestine and spinal cord, with a concomitant increase in regulatory T cells^56,57. Broad spectrum antibiotic treatment, depleting the gut microbiota, also reduces EAE disease severity and the levels of proinflammatory cytokines⁵⁸. Perhaps most importantly, gut microbiota from pwMS, when transplanted into germ-free transgenic mice expressing myelin reactive T cells, increase the incidence of spontaneous EAE, demonstrating sufficiency of gut microbiome compositional changes specific to MS, to drive CNS autoimmunity⁴².

Similar to the MS gut microbiome in humans, studies in animal models often have low concordance regarding the compositional nature of the gut microbiota in the disease state, as well frequently reporting divergent functional impacts of specific microbes on disease pathogenesis. These discrepancies likely stem from three main factors, including: 1) timing and/or mode of exposure to the gut microbiota, 2) the baseline composition of the gut microbiome, and 3) host genetics. An important distinction during hypothesis testing is whether a particular microbe is considered as a: A) commensal factor contributing to MS risk, B) commensal factor impacting MS progression, or C) potential exogenous therapeutic. This information is critical when designing an animal study. Importantly, while case-control studies in humans should identify putative commensal microbes that fit the profile of A) or B), most microbiota studies in animals involve continuous exogenous treatment with large quantities of what are often lab-grown probiotic strains of bacteria, rather than true commensal isolates, suitable only for profile C). As an illustrative example of such contextual differences in modulating CNS autoimmunity, we highlight here the seemingly contradictory findings concerning the role of Limosilactobacillus reuteri (L. reuteri) and other Lactobacillaceae family members.

In our own studies, we identified that stable colonization with the commensal and putative probiotic species, L. reuteri, is sufficient to exacerbate disease severity in a genetically susceptible host (discussed in more detail in the subsequent sections). Importantly, we used stable one-time colonization with a murine commensal isolate strain of L. reuteri, together with vertical transmission to experimental offspring, such that immune system development and, importantly, disease initiation, occurs in the context of commensal colonization. More recently, another group demonstrated that the gut microbiome context within which L. reuteri colonization occurs impacts overall outcomes in CNS autoimmunity. While stable monocolonization of germ-free mice with L. reuteri was insufficient to modulate EAE pathogenesis, co-colonization with an Erysipelotrichaceae family member resulted in greater disease severity than did either microbial species alone⁵⁹, in agreement with our findings. In contrast, oral daily high dose (10⁸ – 10⁹ CFU per day) administration of probiotic strains of L. reuteri to B6 mice starting at the time of EAE induction was shown to ameliorate disease severity, with a concomitant reduction in proinflammatory T cells^60,61.

While these seemingly disparate results may appear difficult to reconcile, differential impact on disease severity likely stems from mode of administration (continuous oral gavage versus commensal colonization), timing of exposure (during disease induction or lifelong, as influencing immune system development and maturation), interactions with differing baseline gut microbiota composition, and/or the bacterial strain used and its origin (lab-grown probiotic vs. true gut commensal). Notably, these factors are all at play in the context of human autoimmune disease and are likely to affect perceived associations between drivers of CNS pathology and the gut microbiota. Interestingly, the abundance of specific Lactobacillaceae taxa has in fact shown significant positive associations with disease severity in progressive MS³⁷.

Importantly, clinical trials using probiotic mixes containing Lactobacillaceae (in some cases including L. reuteri) are already being explored as potential MS therapeutic strategies. Some studies have shown improvement from an immunological standpoint, with administration of probiotics eliciting an increase in regulatory T cells and diminishing levels of inflammatory Th1/17 cells^43,62–64, although results of these studies have been somewhat variable⁶⁵. The impact of probiotics on MS disease course remains to be rigorously evaluated in large RCTs. To date, one small RCT study assessed disability status and found no significant impact of probiotic supplementation on EDSS⁶³. A series of three different RCTs from a different group reported partial clinical responses in either relapse rate or disability status with probiotic treatment^66–68. We note that there has been an editorial expression of concern published for one of these three studies⁶⁹. Overall, a difficulty with interpreting these human trials mechanistically is that the probiotic interventions often contain several types of bacteria that are taxonomically disparate (in some cases from different phyla), in addition to the above mentioned caveats regarding species and strain-specific effects. Taken together, better designed further mechanistic studies in animal models to define microbial relationships and determine the immunological window within which they serve as modifiers of CNS autoimmunity will likely help to resolve apparent discrepancies in both human and mouse studies of the gut microbiome and MS, as well as inform future efforts to modify interactions between the gut microbiota and the host to ameliorate disease.

HOST GENETICS CONTROL GUT MICROBIOME COMPOSITION TO INFLUENCE CNS AUTOIMMUNITY

Host genetics modify the composition the gut microbiota: evidence from human studies

Initial attempts to causally link human genetic variation to compositional changes in the gut microbiota yielded numerous loci putatively influencing the abundance or presence of specific microbial taxa and their interactions with the host^70–74. Despite this fact, associations were relatively weak with low concordance between studies, owing to insufficient sample size, disparities in the collection, isolation, and analytical strategies employed, potential environmental influences differing among cohorts, and/or underlying genetic differences between cohorts dictating true compositional changes in the gut flora. Consequently, the international consortium, MiBioGen, was established to undertake a large-scale, genome-wide association analysis of human genetic variation and the gut microbiome, enrolling 18,340 participants from 24 independent and highly diverse cohorts from numerous countries^16,75. Perhaps not surprisingly, no association was found between specific genetic polymorphisms and broad metrics of species diversity or global community structure of the gut microbiota. However, at the level of individual genera, 27 taxa were associated with 20 distinct loci, including those impacting the abundance of Bifidobacterium, Ruminococcaceae, Faecalibacterium, Lachnospiraceae, Bacteroidales, and Peptococcaceae. The strongest association was found between Bifidobacterium abundance and a genomic block including the well-known LCT gene encoding lactase, which contains polymorphisms (frequent in European populations) that allow this enzyme to be persistently expressed during adulthood. Fittingly, Bifidobacterium abundance increased in study participants reporting milk intake only in the context of a low functioning haplotype of the LCT gene, suggesting a gene-by-diet interaction impacting gut microbiome composition. Phenome-wide association of the LCT:Bifidobacterium linkage suggested a positive association with lactose intolerance and type 2 diabetes, with a negative association with obesity. Moreover, abundance of Actinobacteria (including the Bifidobacterium genus) was linked to reduced risk of ulcerative colitis. Abundance of both Ruminococcus torques and gnavus (Lachnospiraceae members) was associated with a locus containing the FUT1 and FUT2 genes, encoding fucosyltransferase 1 and 2, involved in secretion of fucosylated mucus glycans in the GI lumen, which microbiota, including Ruminococcaceae, are known to degrade as a nutrient source. Dietary fruit intake enhanced the Ruminococcus:FUT1/FUT2 association, again suggesting a gene-by-diet interaction controlling microbial abundance in the gut. Phenotypically, the Rumminococus:FUT1/FUT2 linkage was positively associated with fish intake and negatively associated with risk of cholelithiasis, Crohn’s disease, alcohol intake, and high cholesterol.

Notably, diversity analyses in the MiBioGen study suggested that sampling, of cohort size or depth in sequencing of the microbiota, was insufficient to capture full diversity and/or all genetic variation contributing to gut microbiota compositional differences. Despite this fact, linkages that were observed were bolstered by prior studies, including three which had previously documented the association between Bifidobacterium and the LCT gene, as well as twin studies, which also show heritability of the microbiota^42,76,77. Moreover, the incorporation of dietary information in association analyses strongly supports the notion that gene-by-diet interactions contribute to compositional differences in the microbiota. Such interactions may account for the lack of concordance in prior association analyses, and could explain the divergent results in the study of the gut microbiome in autoimmunity, requiring further work to determine the scope of their impact on disease pathogenesis in MS. Altogether, these studies confirm that host genetics has subtle but significant effects on gut microbiota composition (see Fig. 1), which in human studies may be challenging to detect due to variability in environmental influences.

Host genetics is a modulator of microbiome effects in MS: evidence from human studies

Evidence from twin studies supports a role for host genetic control of gut microbial community structure. In one of the first studies to show sufficiency of the gut microbiome in triggering autoimmune disease, Berer et al. compared the gut microbiota from a series 34 monozygotic twin pairs discordant for MS of mixed geographic and genetic backgrounds⁴². Gut microbial composition between discordant twins was more similar than unrelated individuals when segregated by disease status, suggesting a clear genetic component for control of the gut microbiota even when accounting for autoimmune status. While there was no overt difference between discordant twins in overall community richness, structure, or individual microbial constituents colonizing the gut, upon stratifying by treatment status, an increase in A. muciniphila was observed in MS-affected untreated twins, compared with both treated and unaffected twins. Importantly, fecal microbiome transplantation from 5 twin pairs into germ-free transgenic mice expressing myelin-specific T cells, as a spontaneous model of MS, demonstrated that fecal microbiota from MS-affected individuals was sufficient to trigger spontaneous disease at a higher penetrance compared with microbiota from unaffected twins. Among the observed differences in the gut microbiota in MS-transplant recipients was a higher abundance Ruminococcus and a lower abundance of Adlercreutzia, Tannerella, and most notably Sutterella. Functionally, while MS microbiota transplant recipients displayed no major differences in splenic or small intestinal lamina propria immune cell subsets, skewing of serum anti-myelin IgG was evident, with increased IgG1 over IgG2a. Further, lower IL-10 production by in vitro activated splenocytes isolated from MS microbiota transplant recipients was echoed by lower IL-10 production from PBMCs isolated directly from MS fecal donors themselves⁴². Taken together, these data represent one of the few efforts to isolate and control for genetic variance as an experimental variable in the study of the MS gut microbiome, and to show sufficiency in driving disease for those unique compositional changes in the gut microbiota that occur in individuals with CNS autoimmunity.

An alternative approach used to interrogate the interplay between host genetic diversity, gut microbiome compositional changes, and CNS autoimmunity, is the inclusion of individuals from diverse ethnic backgrounds in studies profiling the MS gut microbiome. In a case-control study, Ventura et al. profiled the gut microbiota of 45 treatment-naïve pwMS of Caucasian, Hispanic American, and African American descent and matched healthy controls⁴⁴. Similar to monozygotic twin studies, no overt differences in community richness or compositional changes were observed in the gut microbiota between ethnic groups or as a function of disease, with the exception of a significant difference in community structure as measured by beta diversity between MS and control subjects of Hispanic American descent. Shifts in individual taxa were compared between ethnicities and between pwMS and matched controls. Interestingly, a single genus, Clostridia, was consistently altered in MS, with a marked increase in all three ethnicities. Compositional changes that segregated by ethnicity in pwMS included an increase in A. muciniphila in Caucasians, increased Adlercreutzia in Hispanic and African Americans, increased Blautia, Holdemania, and Dorea with a decrease in Prevotella, Slackia, Lachnospira, and Dialister in Hispanic Americans, and increased Butyricicoccus in African Americans. Importantly, in other studies, ethnic background is not always reported as a component of cohort demographics, and this is the only study to date which has formally compared the gut microbiota of pwMS between different ethnicities in the same cohort. Taken together, these results suggest that MS-specific microbiota profiles differ across ethnicity, and hence possibly by host genetic background. However, stratification by ethnicity can also add a number of non-genetic confounding variables, such as socioeconomic status and diet, and hence ethnicity may be a poor proxy for genetic differences^78,79. Future studies of a similar design that include genome-wide or targeted genotyping can address this issue directly.

Diverse geographic location can serve as a proxy for incorporating genetic diversity, which is potentially more robust than ethnicity⁷⁸. The recent iMSMS study enrolled individuals exclusively of Caucasian or Hispanic American descent, but recruited from seven sites in San Francisco, Boston, New York, Pittsburgh, Buenos Aires, Edinburgh, and San Sebastián⁴⁰. Interestingly, differences in gut microbiome alpha diversity were only observed between geographical locations rather than as a function of disease status. Moreover, geographical location had the most significant effect on microbial community structure, accounting for a much higher percentage of variance than did disease status. MS-associated alterations among specific taxa in the iMSMS study shared some similarities with changes reported by Ventura et al.⁴⁴ specifically in individuals of Caucasian or Hispanic American ethnicity, including an increase in A. muciniphila and Blautia (positively correlating with disease severity in RRMS). By contrast, Prevotella and Clostridial abundance displayed contradictory patterns between these studies. The contribution of diet to gut microbial composition was also accounted for in the iMSMS study, as discussed above. While microbial diversity did correlate with a higher HEI-2015 score, this was inconsistent between geographical locations. Although additional environmental factors differing between geographical locations may account for this disparity, genetic differences between recruitment sites may also contribute to this finding, and future studies should include genotyping to directly assess this.

Lastly, a recent study directly assessed the effect of MS genetic risk variants (as defined by MS GWAS, quantified together as a single polygenic risk score), on the composition of the gut microbiota⁸⁰. Microbiome community structure analysis demonstrated clustering of individuals into distinct microbiota profiles, which correlated to some extent with the MS polygenic risk score, suggesting that the cumulative effect of MS genetic risk variants could affect gut microbiota composition. Taken together, these studies provide some suggestive evidence that the composition of the gut microbiota in MS differs by host genetic background, and/or that host genetic background can modify the effect of the gut microbiota on MS risk or progression (see Fig. 1), although more definitive studies are needed.

Host genetics is a modulator of microbiome effects in MS: evidence from animal studies

There is also evidence from mouse models of MS for the interplay between host genetics and the gut microbiota in CNS autoimmunity. Leveraging transgenic mice, Shahi et al. demonstrated that both gut microbiota composition and EAE severity are impacted by polymorphisms in HLA class II gene⁸¹. Specifically, transgenes for human HLA-DQ8 and/or HLA-DR3 were introduced into AE-KO mice, lacking endogenous mouse MHC class II. Importantly, the DQ8 and DR3 alleles have both been implicated as risk factors for MS^82–85. Interestingly, while mice harboring HLA-DQ8 were resistant to EAE, HLA-DR3 mice were susceptible. Surprisingly, double transgenic mice expressing both the DR3 and DQ8 polymorphisms, displayed more severe disease than was observed with either allele alone, suggesting a synergist effect in modulating CNS pathogenesis. The authors next compared the fecal gut microbiota of AE-KO mice to all HLA transgenic strains. While overall species diversity increased in the presence of any particular HLA allele, compositionally, the gut microbiota of the disease-resistant HLA-DR8 mice was more similar to that of the AE-KO strain. Moreover, the gut microbiota of the double-transgenic HLA-DR3.DQ8 mice more similar to HLA-DQ8 than to mice expressing DR3 allele alone. Predictive modeling indicated that Turicibacter, Roseburia, and Candidatus Saccharimonas were key features of the gut microbiota discriminating among transgenic strains. These data suggest that, in the context of DR3 expression, microbiota controlled by the DQ8 haplotype maybe enhance disease severity. An important future functional study would be needed to determine whether the observed changes in microbiota correlated with EAE severity and if HLA variants are causative for the disease phenotype vs. simply co-correlated. Together, by focusing on allelic variation known to directly impact MS risk, this work suggests that host genetic-driven differences in gut microbiota composition may impact autoimmune disease severity.

One of the major limitations to the study of the gut microbiota and genetic diversity in mouse models in general is the over-reliance on standard laboratory inbred strains, such as C57BL/6J (B6) or SJL/J, which are closely related and thus fail to capture the genetic diversity present in humans. In our own studies, we have previously leveraged wild-derived PWD/PhJ (PWD) mice, which are genetically divergent from standard in-bred laboratory strains of mice, including B6^86,87. Analysis of the gut microbiota from B6 and PWD mice demonstrated that they display markedly divergent gut microbiome compositions, with a contraction of diversity in the PWD strain, presumptively due to higher selective genetic pressure in the wild-derived host⁸⁸. Among numerous changes in the abundance of different taxa, while PWD mice were marked by an increase in Lactobacillus and Alistipes genera, B6 mice displayed high inter-individual variation in A. muciniphila abundance. Importantly, these mouse strains also displayed divergent susceptibility to disease in the EAE model, where B6 mice were susceptible and PWD mice were resistant^88,89.

To formally assess the role of PWD genetic variation in modulating gut microbial composition and CNS autoimmunity, we leveraged the B6.Chr^PWD consomic model, which consists of 27 strains of mice on the B6 background carrying single or partial chromosomes from the wild-derived PWD mouse⁸⁶. PWD-derived genetic variation in this model dictated differential susceptibility to CNS autoimmunity, with a number of consomic strains demonstrating higher or lower susceptibility compared with the background B6 strain. Likewise, each of the 27 consomic strains displayed unique and divergent gut microbiota profiles that spanned the continuum between the B6 and PWD parental strains, suggesting that host genetics modulates gut microbiota composition. Integrating gut microbiota composition with EAE clinical scores revealed associations of specific gut bacterial taxa with disease severity. While Acetatifactor muris, Clostridium leptum, Turicibacter sanguinis species, as well as operational taxonomic units (OTUs; proxy for individual species or taxa) belonging to the Clostridium cluster XlVa, Erysipelotrichaceae, and Lactobacillus families were positively associated with CDS, A. muciniphila and Clostridium viride, Clostridium sensu stricto and members of Desulfovibrio displayed a negative association⁸⁸. Importantly, the estimated contribution of any single OTU to the overall observed differences in EAE severity was modest, at 0.5–1.3%, suggesting that the impact of any individual species may be of less import than the complex interactions among the gut microbiota themselves, and/or microbiota interactions with the host genetic background in modulating CNS autoimmunity. These data, in part, may help explain the discordant results between human studies characterizing the gut microbiota in MS among genetically disparate cohorts and individuals.

To isolate the gut microbiota as an experimental variable and to demonstrate causation, we transplanted divergent intestinal microbiota derived from B6 and PWD mice into a host of the same genetic background: B6 germ-free mice. PWD transplant recipients displayed enhanced disease severity compared with B6 transplant recipients, which was surprising, given the resistance of PWD mice to EAE. These data suggest that divergent gut microbiota within genetically identical susceptible hosts (here B6 germ-free mice) can differentially impact susceptibility to CNS autoimmunity, and highlight a need to consider both host genetics and baseline gut microbiome composition in any therapeutic application aimed at manipulation of the gut microbiota in MS. In an effort to identify causative microbes, we went on to show that L. reuteri, a mouse and human gut commensal and putative probiotic species, was a key bacterial feature of the PWD gut microbiota, discriminating it from the divergent B6 microbiome. Stable colonization of germ-free B6 mice harboring the B6 microbiota with L. reuteri isolated from PWD microbiome was sufficient to exacerbate disease, enhancing encephalitogenic T cell responses. Importantly, in the context of the normal PWD host, the presence of endogenous L. reuteri was not sufficient to confer susceptibility to EAE, suggesting that genetically determined EAE resistance is not easily overridden by microbial factors. Taken together, these data suggest the gut microbiota requires a genetically susceptible host to modulate disease pathogenesis in MS⁹⁰, and that specific “keystone” gut commensal microbes can modulate disease susceptibility in those hosts (see Fig. 1).

MICROBIAL METABOLISM OF DIETARY SUBSTRATES AS A MODULATOR OF CNS AUTOIMMUNITY

SCFA as modulators of physiological and immunological mechanisms of MS

The beneficial role of SCFAs in MS is supported by a number of studies. The production of SCFAs, including acetate, propionate, and butyrate, occurs through fermentation of host dietary fiber and resistant starch by the gut microbiota in the small and large intestines⁹¹. Classically, the particular types of SCFA produced segregate by bacterial genera. Clostridium clusters IV and XIVa, Eubacterium, Ruminococcus and Faecalibacterium, are established butyrate producers, while acetate and propionate are associated with Bacteroidota phylum members^92–95. Notably propionate can also be produced by Lachnospiraceae and Verrucomicrobia (Fig. 2A)⁹⁶. However, recent studies indicate that up to 19% of the gut microbiota encode the metabolic pathways necessary for butyrate production, suggesting that SCFA production may be more widely conserved⁹⁷.

Figure 2. Major microbial metabolic pathways postulated to impact multiple sclerosis risk and severity.

(A) Diagram illustrating host dietary fiber as substrate for major SCFA-producing gut microbiota in MS. (B) Schematic of bacterial and host tryptophan metabolic pathways. (C) Proposed metabolic mechanisms of A. muciniphila-driven impact on CNS autoimmunity, including production of vitamin K, myo-inositol, and SCFAs.

Functionally, SCFAs serve as an energy source for both the host and the gut microbiota. Specifically, in the gut, intestinal epithelial cells leverage SCFAs as an available energy source, with excess being transported to the liver, not only to serve as an energy source but also as substrate for cholesterol and fatty acid biosynthesis. Importantly, SCFAs can enter host circulation and can cross the BBB, enabling far reaching effects from their original point of synthesis including a broad anti-inflammatory effect, with a role in maintaining critical barrier integrity^98–100. In the peripheral immune system, SCFAs can enhance regulatory T cell differentiation and promote IL-10 production, inhibit inflammatory signaling in macrophages, dendritic cells, and T cells, and downregulate tumor necrosis factor alpha production by neutrophils and peripheral blood mononuclear cells^101–107. In the CNS, fatty acid receptors are expressed on BBB epithelial cells, astrocytes, oligodendrocytes and microglia, promoting their maturation and function^108–113.

Some of the most concrete lines of evidence for a microbiota-driven role of SCFA in CNS autoimmunity comes from basic research using mouse models. Concentrations of SCFA in germ-free mice are 100-fold less than in conventional mice, where the gut microbiota produce between 500–600 mmol of SCFA each day reaching 20–140 mM in the lumen of the intestine^91,114–116. Further, while germ-free mice have decreased tight junctional proteins on endothelial cells of the BBB and increased barrier permeability^91,117, colonization with known SCFA-producers or direct treatment of germ-free mice with SCFAs is sufficient to restore BBB integrity. In the context of CNS autoimmunity, treatment with acetate, propionate, butyrate, mixtures thereof, or provision of high fiber, as the pre-biotic substrate for gut microbiota SCFA production, have all been shown to reduce disease severity in the EAE model of MS^{103,118–122}. We note that at least one study has also shown pro-encephalitogenic effects of SCFA treatment on T cells in an adoptive transfer model, as well as documenting EAE resistance in mice deficient for SCFA receptors¹¹⁸, suggesting possible context-dependent effects of SCFA that may not be uniformly anti-inflammatory. As above, an important factor not considered in these therapeutic studies is the baseline gut microbiota composition in the host, which may be predictive of a response (or lack thereof) to treatment with SCFAs or dietary fiber.

Direct evidence from pwMS for the importance of SCFAs also exists. Fecal, serum and plasma levels of acetate, propionate, butyrate, isobutyrate, valerate, and isovalerate have all been observed in pwMS^{98,118,123–125}. Further, diminished butyrate abundance has been correlated with markers of increased intestinal permeability and regulatory T cells in the peripheral blood¹²⁴ and depletion of both isobutyrate and isovalerate correlate with worsening Expanded Disability Status Scale (EDSS)^123,125. Moreover, known SCFA-producing gut microbiota are reduced in MS, including an observed deceased in the abundance of Butyricimonas, Bacteroides, Lachnospira, and Eubacterium^{44,98,118,124,125}.

Notably, demographics, including ethnic background, are not fully reported in the majority of studies, nor has the genetic control of SCFA production been evaluated. With regard to country-specific cohorts, in a Chinese cohort, fecal acetate, propionate, and butyrate were all decreased in pwMS, correlating with circulating levels of regulatory T cells and a reduction in SCFA-producing microbiota¹²⁶. In four Spanish cohorts, variable outcomes were observed. While total levels of fecal SCFA were comparable between pwMS and healthy controls, higher disability was associated with increased levels of acetate and diminished abundance of butyrate, with no detectable difference in propionate. Notably, analysis of the gut microbiome in pwMS identified an increase of Bacteroidaceae, known propionate producers, which may account for the lack of measurable difference in this metabolite¹²⁷. In a second Spanish cohort, only the serum ratio of propionate to acetate was found to be diminished, while a separate study surprisingly reported higher plasma acetate, positively correlating with EDSS, and inversely correlating with both CD4+ T cells and IL-17 producing CD8+ T cells in PBMCs¹²⁸. These data suggest environmental, geographical, ethnic, and/or genetic diversity in the control of SCFA-producer abundance and their metabolites; however, additional studies are needed to define the relationship between host genetics and the interplay between host diet and gut bacterial SCFA metabolism.

Nonetheless, there is increasing interest in evaluating the safety and efficacy of SCFA and dietary fiber administration in the treatment of MS. In a clinical trial including a large well-defined cohort of 303 pwMS and 68 controls, Duscha et al. evaluated SCFA supplementation, in conjunction with standard DMT usage, to demonstrate the therapeutic potential of propionate⁹⁸. Consistent with previous studies, baseline propionate was lower in pwMS; however, no differences in fecal or serum acetate or butyrate were identified. Further, analysis of the gut microbiota revealed lower SCFA-producing gut microbiota, including Butyricimonas and Collinsella. In both the treatment-naïve and in the context of traditional DMT usage, daily propionate supplementation increased regulatory T cells, with a reduction in Th1/Th17 cells in whole blood. Retrospective analysis of relapse rate, disability status, and progression, as well as brain atrophy, all showed improvement in a subset of study participants who consistently adhered to propionate usage for a full year⁹⁸. Taken together, this study serves as a proof-of-principle evaluating the safety and possible efficacy of propionate treatment in pwMS, and suggests that restoration of SCFA levels or microbiota that produce SCFAs may indeed be beneficial in MS.

Collectively, evidence from observational clinical studies demonstrating reduced abundance of SCFA and the microbiota that produce them in pwMS, coupled with functional studies in mouse models, establish the importance of these metabolites in CNS autoimmunity. Further study is required to establish the precise host-microbe interactions relevant to therapeutic applications and the correct intervention window (during disease initiation or therapeutically aimed at slowing disease progression) as the most likely to prove beneficial.

Dietary tryptophan metabolism by the host and the gut microbiota: a brief overview

Changes in metabolism of tryptophan have also been extensively studied in MS. Tryptophan is acquired through digestion of dietary protein in the small intestine where it is absorbed through the epithelium to enter host circulation^129–131. Tryptophan is an essential amino acid, used by the mammalian host in protein synthesis and for the production of serotonin (5-hydroxytryptamine) and kynurenine^132–134. 90% of tryptophan is metabolized along the kynurenine pathway using one of two rate-limiting enzymes, indoleamine 2,3 dioxygenase (IDO) in the brain, gut, and liver, or tryptophan 2,3 dioxygenase (TDO) in the liver (Fig. 2B)^133,135. Kynurenine, which can cross the BBB, is metabolized into either kynurenate, with known neuroprotective properties, or quinolinic acid, which is neurotoxic^135–138. Unlike kynurenine, serotonin cannot cross the BBB, with 90% produced by enterochromaffin cells in the gut and local production in the CNS^134,139. Consequently, pools of peripheral and CNS serotonin, produced by specific isoforms of tryptophan hydroxylase expressed in each anatomical location, are considered distinct.

The gut microbiota compete with the host for dietary tryptophan as a limited resource. Microbial metabolism of tryptophan in the large and small intestine occurs through four major pathways, as dictated by distinct rate-limiting enzymes within particular clades of microbiota (Fig. 2B). While Clostridium, Ruminococcus, Blautia, and Lactobacilli metabolize tryptophan along the tryptamine pathways, using tryptophan decarboxylase, Escherichia coli, Clostridium, and Bacteroides utilize tryptophanase to produce indole^140–142. In Clostridium, Bacteroides, and Bifidobacterium, tryptophan is instead converted to indole-acetamide using tryptophan monooxygenase and both Lactobacilli and Bifidobacterium conserve the aromatic amino acid aminotransferase for indole-3-lactcate production^{141,143–146}. Importantly, the gut microbiota can also modulate host metabolites of tryptophan either directly or indirectly. For example, Lactococcus, Lactobacillus, Streptococcus, Escherichia coli, and Klebsiella can directly produce serotonin, while spore-forming microbiota have been shown to indirectly promote host production of serotonin in the gut^133,147,148. Altogether, there are a number of unique tryptophan-derived bacterial metabolites that have the potential to impact host physiology in the context of CNS autoimmune disease, as discussed in more detail below.

Signatures of host and microbial tryptophan metabolism dysregulation in pwMS

The significance of tryptophan in MS is underscored by the observed alteration in the abundance of bacterial or host tryptophan associated metabolites in the serum, plasma, CSF, or urine of pwMS. Notably, given the range of diversity in cohort geographical location, patient demographics, MS subtype and sample type, there is a high degree of variability in findings between studies. However, three independent studies did note a marked depletion of serum or plasma tryptophan^149–151. Further, depleted serum levels of 5-hydroxytryptophan, kynurenine, kynurenate, quinolinate, xanthurenate, and the bacterial metabolites, indole-3-lactate and indole-3-propionate, as well as an increase in 3-hydroxyanthranilate, have all been reported^{149,151–153}. In the urine, both kynurenine and the kynurenine to tryptophan ratio are decreased, correlating with EDSS¹⁵⁴.

MS subtype-specific differences in tryptophan metabolism may also exist. In confirmed cases of clinically isolated syndrome (CIS), directly following a first attack, tryptophan is depleted in the cerebral spinal fluid (CSF)¹⁵⁵. Interestingly, in the CSF from people with progressive MS, N-acetyl-serotonin is increased and elevated bacterial indole-3-acetate is associated with MRI lesions and disease duration¹⁵⁶. Moreover, cohorts exclusively enrolling RRMS patients consistently report elevated levels of kynurenate^157,158, and the neurotoxic bacteria-derived metabolites of tryptophan, indoxyl-sulfate and p-cresol-sulfate, were elevated in the CSF of RRMS patients compared to those with progressive MS¹⁵⁹. By contrast, in pediatric cases, higher serum levels of bacterial indole-3-lactate were associated with lower disability and higher processing speeds, while increased kynurenine was associated with increased rate of relapse¹⁶⁰. Notably, DMT usage appears to impact host and/or bacterial tryptophan metabolism as well. A 6-month course of dimethyl fumarate treatment was sufficient to reduce CSF and serum levels of indoxyl-sulfate but notably did not significantly impact levels of indole-3-acetate or kynurenate, which were increased or decreased at baseline respectively, relative to healthy controls¹⁵⁹. Taken together, these data suggest that in pwMS, host and gut microbial tryptophan metabolism is dysregulated, with a general trend of increased abundance of specific bacterial tryptophan metabolites in the blood and the CSF, although significant study-to-study variability exists. However, identifying key context-specific alterations relative to disease progression as possible targets for therapeutic intervention remains challenging.

Pre-clinical evidence for an impact of microbial tryptophan metabolism on autoimmunity

Some of the most comprehensive direct evidence for the role of dietary tryptophan in CNS autoimmunity originates from the EAE model, where the levels of tryptophan, as a substrate for bacterial and host metabolism, as well as downstream metabolites, can be experimentally manipulated. Sonner and colleagues reported a complete suppression of disease in mice fed a tryptophan depleted diet, administered prophylactically, starting on the day of immunization¹⁶¹. Restriction of dietary tryptophan was sufficient to reduce plasma tryptophan levels by 50–60% and further, to abolish BBB disruption, leukocyte infiltration, and demyelination in the CNS. Flow cytometric analyses and adoptive transfer studies went on to show that while dietary tryptophan restriction did not impair autoreactive T cell induction, Th1/Th17 skewing and impaired migratory capacity were observed. Interestingly, suppression of EAE by dietary tryptophan depletion was partially lost in germ-free mice compared with mice colonized with a conventional microbiome, suggesting a microbiota-dependent effect. Importantly, dietary tryptophan restriction resulted in changes in the gut microbiota, with contraction of known tryptophan-catabolizing bacteria, including Lactobacilli, with an expansion in microbiota that synthesize tryptophan de novo. However, supplementation with the host tryptophan metabolite, serotonin, or bacterial tryptophan metabolite, indole-3-aldehyde, was insufficient to restore CNS autoimmunity in tryptophan deficient mice, suggesting that additional metabolites and/or combinatorial effects may be involved. These data suggest that tryptophan metabolism by the host and/or gut microbiota may promote CNS autoimmunity, exerting an inflammatory or immunomodulatory effect.

In our own studies, building on our initial observation that L. reuteri colonization was sufficient to exacerbate CNS autoimmunity⁸⁸, we went on to interrogate the role of dietary tryptophan as a potential substrate for L. reuteri¹⁶². In a dietary intervention model similar to that used by Sonner et al., we randomized mice, harboring a specific microbiota configuration colonized or not by L. reuteri, to either a low or high tryptophan diet 1-week prior to EAE induction. Consistent with Sonner et al., we observed a near complete abrogation of disease in mice fed a low tryptophan diet. Importantly, we demonstrated that L. reuteri-dependent exacerbation of disease, while robust in mice fed a high tryptophan diet, was abrogated upon dietary tryptophan depletion, demonstrating that L. reuteri requires host dietary tryptophan to enhance EAE severity and inflammatory T cell infiltration in the CNS. Interestingly, we found that in mice colonized with a divergent (PWD) microbiome, naturally harboring L. reuteri, EAE exacerbation by specific microbiota occurred in the absence or presence of dietary tryptophan, suggesting the presence of other microbes that are capable of modulating EAE in a tryptophan-independent manner. Metabolomic profiling of serum revealed that dietary tryptophan depletion resulted in a profound downregulation of both bacterial and host-derived tryptophan metabolites, including several indoles, and, surprisingly, bacterial tyrosine metabolites, p-cresol-sulfate and p-cresol-glucuronide, which were among the topmost significantly affected metabolites. Moreover, L. reuteri colonization modulated several host and bacterial metabolites of dietary tryptophan, unexpectedly including some within the classically mammalian kynurenine pathway, as well as p-cresols and several imidazoles. These findings were substantiated by in vitro monoculture experiments, whereby L. reuteri produced a variety of indoles, imidazoles, and p-cresols, as well as actively depleting kynurenine. Several of these indole metabolites were demonstrated to activate the AhR in vitro and potentiate IL-17 production by T cell cultures, providing a potential mechanism by which L. reuteri promotes EAE, which is also supported by a recent study demonstrating microbiota-dependent CD4+ T cell-intrinsic AhR signaling promoting encephalitogenic responses in EAE¹⁶³. Taken together, these data causally link a singular keystone species of the gut microbiota to tryptophan-dependent exacerbation of CNS autoimmunity and emphasize context-specific effects of dietary tryptophan and its bacterial metabolites. Importantly, as noted above, both p-cresol-sulfate and indoxyl-sulfate were found to be elevated in the CSF of pwMS¹⁵⁹, supporting the notion that tryptophan-derived metabolites can promote autoimmunity.

The immunomodulatory effects of tryptophan metabolites are postulated to occur, at least in part, through binding to the intracellular receptor and transcription factor, the aryl hydrocarbon receptor (AhR). Quintana and colleagues showed that a tryptophan-depleted diet when administered therapeutically, staring on day 22 post-disease induction, prevented recovery from EAE, with oral supplementation of tryptophan sufficient to reverse this effect¹⁴⁹, although we note that the diet described in this study appears to be deficient in ligands for the AhR, but sufficient in tryptophan (discussed below). A similar EAE exacerbation was observed with therapeutic administration of ampicillin, which could be reversed by provision of the host-modified bacterial tryptophan metabolite, indoxyl-3-sulfate, or other bacterial metabolites of tryptophan, including indole, indole-3-propionate, or indole-3-aldehyde. By contrast, antibiotic treatment with vancomycin did not exert the same impact on EAE pathogenesis, suggesting that specific members of the gut microbiota exert a tryptophan-dependent impact on CNS autoimmunity. Consistent with this, the abundance of L. reuteri, a known tryptophan-catabolizing member of the mouse and human gut microbiota, was reduced upon ampicillin treatment. The role of the AhR in regulation of CNS autoimmunity was also examined, defining an AhR-dependent protective role for tryptophan metabolites in regulating astrocyte and microglial crosstalk to ultimately regulate inflammation and neurodegeneration in the CNS^149,164. Altogether, these findings suggest that bacterial tryptophan metabolites, via glial AhR activation, play a protective role in the chronic phase of CNS autoimmunity. We note that these findings may appear to be in contrast to those of Sonner et al. and our own (described above), and we attempt to reconcile these findings collectively at the end of this section.

In an alternative approach, additional studies have evaluated the therapeutic utility of individual tryptophan metabolites on modulating disease pathogenesis in the EAE model. Dopkins et al. found that tryptamine, when administered intraperitoneally starting on day one post-immunization, ameliorates EAE severity, leading to a reduction in CD4+ T cell infiltration and IL-17 production in the CNS with an increase in regulatory T cells. Further, adoptive transfer studies, demonstrated that tryptamine mediated EAE suppression was dependent on the AhR and was sufficient to modify the gut microbiome composition to support expansion of butyrate-producing microbiota¹⁶⁵. Interestingly, the plant derived indolic, indole-3-carbinol, was also shown to reduce EAE pathogenesis when administered prophylactically or therapeutically¹⁶⁶. Lastly, a recent study indicated that oral gavage with indole-3-lactate, likely derived from Ligilactobacillus murinus endogenously, can also ameliorate EAE¹⁶⁷.

The impact of dietary tryptophan availability on autoimmune disease has also been explored in other disease models. In a model of systemic lupus erythematosus (SLE), Morel and colleges demonstrated that dietary tryptophan restriction is sufficient to ameliorate disease, while high dietary tryptophan exacerbates autoimmunity in a genetically susceptible host in a microbiome-dependent manner¹⁶⁸. Further, in a model of rheumatoid arthritis, Kuhn and colleagues also found that dietary tryptophan restriction reduced disease severity and incidence, while provision of the bacterial tryptophan metabolite, indole, returned arthritis severity to comparable levels as observed in mice receiving a tryptophan sufficient diet¹⁶⁹. Taken together, these data suggest that the role of tryptophan in promoting autoimmunity is not limited to MS, and may be more broadly conserved^170,171.

Reconciling such disparate findings among studies interrogating the role of tryptophan in autoimmune disease will better inform any potential therapeutic applications. Several key aspects differed among such studies that are likely to impact the outcomes and interpretations. Perhaps most importantly, the timing and specific formulation of dietary intervention differed between studies. Dietary tryptophan restriction consistently suppressed clinical disease when administered prophylactically, before or at EAE induction in two different studies^161,162, while interfering with the recovery phase of disease when introduced during the chronic phase of EAE in a different study¹⁴⁹. Similarly, in lupus-prone mice, tryptophan restriction ablated the autoimmune response early in disease, but was insufficient to prevent age-associated accumulation of disease when given as a therapeutic within this model¹⁶⁸. In showing tryptophan-dependent exacerbation of autoimmune arthritis, dietary intervention began on the day of immunization, indicating that the observed effects of both tryptophan restriction and indole supplementation function in disease initiation or onset, rather than in progression of disease¹⁶⁹. Besides timing, baseline gut microbial composition differs among genetic backgrounds of mice and between vivaria, which is likely to impact the interplay between host and gut microbial metabolism of tryptophan as a limited resource in the gut^88,172. Taken together, these data collectively suggest that tryptophan and its associated metabolites, may exert divergent effects through the course of autoimmune disease, with a pathogenic role in disease susceptibility/onset, and a possible protective role during the chronic phase of the disease. This suggests that lowering tryptophan intake might diminish MS onset risk in individuals at high risk for MS, but that tryptophan restriction could be detrimental if administered therapeutically to pwMS.

Importantly, defining the precise dietary formulations across pre-clinical studies is essential to interpreting impact on the gut microbiota, systemic metabolites, and ultimately disease pathogenesis. The recommended level of tryptophan is 0.2% in general mouse chow, and 0.1% is required for adequate growth in young animals¹⁷³. In our own studies, a high tryptophan diet exceeded this recommendation, at 0.8%, and thus was defined as provision of excess dietary tryptophan¹⁶². This was also the case in dietary invention in models of SLE^168,174, where a high tryptophan diet was formulated at 1.19%. The remainder of studies, including two in the EAE models^149,161 and the collagen-induced arthritis model¹⁶⁹, used a range of 0.18–0.25% tryptophan sufficient diets, as opposed to high dietary tryptophan formulations. Even more pronounced variation in the degree of tryptophan restriction occurred between studies. In general, tryptophan restriction ranged at 0.02–0.08%. In the autoimmune arthritis model and lupus-prone mice, in order to avoid the deleterious effects of complete tryptophan ablation, a 0% tryptophan diet was rotated with a tryptophan sufficient diet, such that over a 7 day period, levels where cumulatively held at 0.05% and 0.08%, respectively^168,169. In our own studies, we used a fixed 0.02% tryptophan deficient diet to provide a low-level steady state without significant adverse effects¹⁶². Interestingly, another EAE study did choose to completely withhold tryptophan from the diet and did not see an impact indicative of abnormal stress or fasting response¹⁶¹. All of the above studies reported suppression of autoimmune disease by tryptophan restriction. In contrast, the other study reporting EAE exacerbation by tryptophan restriction during the chronic phase¹⁴⁹ opted to use an AhR-ligand low diet, as the tryptophan-depleted diet, which was previously characterized in the context of intraepithelial lymphocyte maintenance and function¹⁷⁵. While this diet is indeed deficient in components that are likely to activate the immunomodulatory AhR, such as plant-derived indole-3-carbinol, it also contains 0.2% tryptophan, which is quite comparable to the standard chow described above. Since this tryptophan-deficient diet was compared to a control 0.25% tryptophan standard feed diet, it is unclear whether dietary effects in this model are tryptophan-dependent, although this does not negate the findings concerning the role of the AhR itself. Importantly, these data are frequently interpreted as exacerbation of EAE in the context of tryptophan restriction, rather than during AhR-ligand deficiency, and therefore appear in direct contradiction to the numerous other studies demonstrating tryptophan restriction dependent suppression of autoimmunity (as outlined above). Moreover, in lupus prone mice, Morel and colleagues used tryptophan-modified synthetic chows that differ only by their tryptophan content (ranging at 0.08, 0.19, 0.3, or 1.19%) to demonstrate that while host-derived tryptophan metabolites including kynurenine and serotonin increase systemically in a tryptophan dependent manner, the relationship with metrics of disease severity is not always linear¹⁶⁸. In fact, there appeared to be an immunological tipping point near 0.19% tryptophan, at which tryptophan and associated metabolites were sufficiently depleted so as to impact systemic autoimmune disease, with more pronounced amelioration of disease at 0.08% dietary tryptophan. Taken together, these data suggest that tryptophan restriction ameliorates autoimmune disease in a tryptophan concentration-dependent fashion.

Additional emerging diet-dependent microbial metabolic regulators of CNS autoimmunity

Isoflavones are a class of phytoestrogens found in legumes, including soybeans, that exert a broadly anti-inflammatory effect. Furthermore, the gut microbes Adlercreutzia equolifaciens and Parabacteroides distasonis, which are reduced in pwMS³⁸, metabolize isoflavones to the potent nonsteroidal estrogen, S-equol, with posited health benefits. To interrogate the role of phytoestrogens in CNS autoimmunity, Mangalam and colleagues modulated the availability of isoflavones in the diet in the EAE model of MS^176,177. Compared with a phytoestrogen-free diet, an isoflavone-rich diet provided protection against disease. Importantly, the authors used bacterial colonization and direct equol supplementation to demonstrate that EAE protection by isoflavones was dependent on the presence of isoflavone-metabolizing bacteria and their estrogenic metabolite, equol. Moreover, fecal LPS extracted from isoflavone fed mice, enhanced anti-inflammatory cytokine production in macrophages, suggesting that isoflavones could modulate LPS biosynthesis by the gut microbiota to impart an anti-inflammatory response, thereby reducing CNS autoimmunity. These results suggest that targeting phytoestrogens or bacteria that metabolize them could represent an interesting therapeutic approach in MS, although this is complicated by the large number of putative phytoestrogens and their dietary sources, as well as potential unwanted estrogenic effects on the host, which can be pleiotropic¹⁷⁸.

Bile acids are cholesterol metabolites mainly produced in the liver, which are then modified by the host in the gall bladder to produce primary bile acids conjugated with glycine or taurine. Upon secretion in the gut, they are further modified by the microbiota to produce secondary bile acids, which can enter host systemic circulation, with far reaching effects, including within the CNS^179,180. Importantly, host-derived primary and/or bacterial-derived secondary bile acids are reduced in the plasma of pwMS and in the EAE model of MS^181,182. Moreover, bile acids have been shown to exert an anti-inflammatory effect through binding to one of two receptors, the nuclear farnesoid X receptor (FXR) and the cell surface G protein-coupled bile acid receptor (GPBAR1)¹⁸². FXR agonists reduce disease severity in the EAE model, and GPBAR1 activation exerts an anti-inflammatory effect on myeloid, glial, and innate immune cells^182,183. Notably, Calabresi and colleagues demonstrated a therapeutic role of microbial-modified bile acids themselves in the EAE model. Administration of the secondary bile acid, tauroursodeoxycholic acid (TUDCA), was sufficient to ameliorate CNS pathogenesis, leading to a reduction in demyelination, astrocytosis, and infiltration of myeloid cells into the CNS¹⁸². Further, treatment of astrocytes and microglial with TUDCA exerted anti-inflammatory and neuroprotective effects. Taken together, these data suggest that bile acid metabolism is dysregulated in pwMS and that microbial-derived secondary bile acids may have a therapeutic role in CNS autoimmunity. In support of this, a recent clinical study from the same group has established the safety and tolerability of TUDCA supplementation in pwMS, with favorable changes in circulating bile acids¹⁸⁴.

In more generalized approaches to modify diet, both fat content and caloric intake were also shown to modulate disease in the EAE model, and are supported by evidence directly from pwMS. Epidemiological studies suggest a correlation between disease severity and intake of fatty acids, while cholesterol metabolism is dysregulated in pwMS and in the EAE model, with a causative role in disease progression^185–189. Interestingly, provision of a high fat diet has also been shown to increase EAE severity, elevating immune cell infiltration and inflammation in the CNS¹⁹⁰. High fat diet mediated increase in CNS autoimmunity was further associated with activation of the renin angiotensin system (RAS), implicated in exerting an immunomodulatory role in the CNS^191,192. Similarly, caloric restriction or intermittent fasting exert a broadly anti-inflammatory and neuroprotective role in preclinical models that is posited to be beneficial in pwMS^193–196. Further, 40% caloric restriction in mice led to a reduction in EAE severity with reduced inflammation, demyelination, and axonal damage in the CNS¹⁹⁷. Taken together, these data suggest that alternative dietary modifications such as low fat or reduced calorie diets, or specific fasting regimes, may be beneficial in MS.

As discussed in previous sections, a key microbe that appears to be a hallmark of MS in multiple cohorts is A. muciniphila, which makes it an attractive therapeutic target. While it is yet unclear how this enigmatic microbe modulates the immune system in MS, some tantalizing evidence regarding the role of its metabolism has begun to emerge (Fig. 2C). In the iMSMS study, the abundance of this microbe was linked to bacterial degradation of phytate from the diet⁴⁰, which could lead to production of the immunomodulatory metabolite, myo-inositol, previously reported to be increased in pwMS⁴⁹. As its name suggests, A. muciniphila also utilizes host mucins as a major nutrient source. Provision of mucin as a sole carbon source leads to in vitro propionate production by A. muciniphila, and some studies suggest that host mucin degradation can also lead to the production of propionate in vivo^96,198, or, alternatively, to the production of additional metabolites that cross-feed other gut microbiota, including Alistipes, which can directly produce indoles to promote EAE¹⁹⁹. Notably, A. muciniphila itself can produce a variety of indoles, can synthesize tryptophan de novo, and has been shown to modulate host tryptophan metabolism, suggesting complexity in the metabolic consequences of A. muciniphila colonization^200,201. Furthermore, a recent study demonstrated an EAE-promoting role for commensal A. muciniphila in the context of defined minimal gut bacterial consortium, which was associated in increased cecal γ-aminobutyric acid (GABA) concentration and major changes in gene expression by other gut microbes²⁰². The former finding is supported by the evidence that A. muciniphila can metabolize glutamate to produce GABA in vitro²⁰³. Lastly, in our own prospective study in pwMS, we found that A. muciniphila abundance was negatively associated with disease progression, and positively linked to bacterial production of vitamin K⁵², which is known to be depleted in MS²⁰⁴, and predicted from microbiota analysis to occur in EAE⁶¹. Taken together, these results suggest that while A. muciniphila has the potential to modulate MS pathogenesis, this could occur in a bi-directional manner, as is dictated by factors such as host diet and the other members of the gut microbiota.

CONCLUSIONS, PERSPECTIVES, AND UNANSWERED QUESTIONS

Overall, we propose a model where, in an individual genetically susceptible to MS, the gut microbiota and diet can act as potent environmental modifiers of disease risk and possibly progression, with diet-dependent gut microbial metabolites serving as a key mechanism (Fig. 1). We also propose that specific microbial taxa may have divergent effects in individuals carrying distinct variants of MS risk alleles, as a result of host gene-by-gut microbiota interactions. Finally, we also propose that the effects of specific microbial taxa, especially those that exert their effects through metabolites, are highly dependent on the host dietary intake and the abundance of other microbes. What emerges is a complex multifaceted interaction that becomes challenging to disentangle in human studies, explaining the divergence of findings across heterogeneous cohorts with differing geography, dietary preferences, and genetics.

Nonetheless, while such divergent results concerning the particular microbial and metabolic contributors to CNS autoimmunity have hampered our cohesive understanding of the drivers of disease etiology and pathogenesis, some common themes have emerged. Alterations in A. muciniphila, Faecalibacterium species, Prevotella, Blautia, and SCFA-producing microbiota, with an accompanying depletion in SCFAs, and an imbalance in tryptophan metabolites are all hallmarks of MS. While association studies have expanded our knowledge of the genetic risk factors for MS, there appear to be more modest contributions to the genetic heritability of gut microbiome composition in heterogenous cohorts, suggesting that host genetic susceptibility determines risk for MS likely via directly modulating the immune system, rather than acting indirectly through modifying the gut microbiota. However, additional studies are needed to isolate environmental confounders, expand cohort sizes, and/or increase depth, to characterizing the genetic contributors to gut microbiome composition. Moreover, large-scale disease-specific association studies to integrate the genetic risk loci associated with CNS autoimmunity and compositional changes in the gut microbiota that occur over the course of disease (and as also influenced by diet), are likely to garner insight that may not have been readily apparent within the general population. Notably, expanding such integrated association studies to include more diverse ethnic backgrounds, differentiate among MS subtypes, and incorporate immunological and metabolomic phenotyping would likely be informative. Indeed, while challenging, such approaches are beginning to emerge, as highlighted by a recent study from Cantoni et al. that identified interactions between meat intake, abundance of a specific gut bacterial species, bacterial meat metabolites, and Th17 cell activity in pwMS²⁰⁵. An orthogonal needed approach is one of reductionist basic science, whereby specific variables such as host genetics, diet, and microbiota can be manipulated in a controlled fashion. In this regard, more studies are needed that identify and rigorously test specific microbial mechanisms that contribute to CNS autoimmunity, for example through genetic manipulation of responsible microbes, as well as studies that thoroughly consider microbe-microbe interactions in the context of a complete or simplified gut microbial ecosystem.

While host genetics may represent a fixed risk factor for MS, both dietary inputs and the composition of the gut microbiome are modifiable. Indeed, the gut microbiome represents a potentially tractable tool to influence host physiological processes, immune responses, and metabolic functions, with implications for personalized medicine and therapeutic intervention strategies. Interestingly, pwMS frequently express interest in diet as an additional means to improve disease course²⁰⁶. Moreover, as a group, pwMS were found to have a higher healthy eating index than household controls⁴⁰, suggesting that dietary modification is already being undertaken in pwMS, and consequently represents a feasible intervention strategy. Despite this fact, there is limited evidence on the precise dietary formulations that might be most beneficial in treatment of autoimmune disease²⁰⁶. Moreover, integrating such evidence to account for heterogenous populations of varying genetic background and to incorporate compositional and functional impacts of “baseline” gut microbiota to modulate CNS autoimmunity remains challenging. One approach to leverage the heterogeneity reflected in pwMS is to adopt an adaptive platform or crossover design to identify responders and non-responders upon dietary intervention as built into clinical trial design²⁰⁷. Further, recent studies suggest that dietary intake can be accurately predicted from fecal samples by constructing a metagenomic food database, which could be leveraged in tandem with or in lieu of standard food questionnaires that are frequently inaccurate, incomplete, and labor-intensive to execute²⁰⁸. The process of iteratively identifying the subset of pwMS who most benefit from any particular dietary protocol, within large-scale studies, coupled with integrated immunological, metabolomic, microbiomic, and genetic characterization, is likely to identify the underlying mechanisms at play and potential biomarkers to personalize dietary recommendations in the future.

The combined effect of diet and microbiome present numerous options for targeted intervention strategies, both as prophylactics and possibly therapeutics. The gut microbiota itself can be directly manipulated, with provision of individual beneficial species (probiotics), defined consortia, or complete fecal microbiome transplantation (FMT). Notably, while FMT has been shown to be both safe and possibly efficacious in small, sometimes single case studies, the outcome in larger heterogenous cohorts is likely to be highly variable^209–211. Moreover, in contrast to FMT, therapeutic administration of individual microbes or defined microbial consortia may have improved safety profiles with enhanced reproducibility, and such microbes can be engineered to exert defined immunological and metabolomic impacts within the host. The efficacy of defined consortia is exemplified by their application in treatment of recurrent Clostridium difficile (C. difficile). Traditional FMT is highly efficacious for treatment of C. difficile infection; however, both the underlying mechanisms involved and long-term consequences to recipients are less clear^212–214. Consequently, various formulations of defined consortia have been developed, including a recent phase 2, randomized, double-blind, placebo-controlled, dose-ranging study using 8 commensal strains of Clostridia in the treatment of recurrent C. difficile infection, reducing recurrence rate by a third²¹⁵.

Notably a major limitation of such approaches is the overall resiliency of the baseline gut microbiome, leading to colonization resistance or requirement for continual supplementation to avoid washout²¹⁶. In contrast to administration of single species, defined consortia are also thought to improve colonization in the gut, with greater stability over time²¹⁷. Interestingly, the vast majority of studies thus far have focused on administration of general probiotic bacteria in CNS autoimmunity, such as laboratory-adapted probiotic strains of Bifidobacterium and Lactobacillaceae, with mostly variable results⁶⁵. Importantly, recent studies have suggested that such probiotic bacteria are unlikely to stably colonize the gut, and in fact may disrupt the normal gut microbiome^218,219. This suggests that a more targeted and rationally designed approach to define the particular microbial species relevant to MS and their functional consequences would improve outcome. In addition to provision of probiotics, prebiotic substrates, such as dietary fiber to boost SCFA production by the microbiota, or postbiotics, including metabolites determined to directly benefit disease, should be explored further²²⁰. In tandem approaches, provision of so-called synbiotics, wherein both prebiotic dietary substrate and probiotic species or consortium are provided together, may also prove useful²²¹. Genetically tractable probiotic bacterial species can also be engineered to produce defined metabolites or exert tailored immunological effects, to enhance safety and efficacy of their application as therapeutics and are already being explored in models of MS^222,223. In all such approaches, defined functional and mechanistic study informed by basic research to elucidate the particular physiological target, outcomes relevant to MS, and potential therapeutic window are essential. Advances in techniques to manipulate the microbiota both in vitro and in vivo within animal models will continue to be indispensable to move the field away from correlative studies and towards mechanistic understanding. Reductionist model microbiomes wherein defined consortia are manipulated can be coupled with dietary manipulation. Human microbiome-associated mice and cultivation of fecal isolates or complete communities can be utilized to add translational relevance to such studies as well. Leveraging these tools to define the context-specific functional consequences of individual species, metabolites, and impact on the immune system in basic research will enhance likelihood of successful therapeutic outcome.

The combined power of large scale multiomic analyses of the gut microbiome and diet in pwMS, coupled with association studies defining the genetic loci contributing to disease risk and composition of the microbiota, has garnered significant insight into the underlying drivers of CNS autoimmune disease. Further incorporation of such approaches in basic, translational, and clinical study of MS, is a necessity to define the drivers of this heterogenous multi-factorial autoimmune disease. Moreover, a bench-to-bedside and back again mentality to iteratively assess and refine applications of diet and the microbiome to treat disease will be required to establish a tailored therapeutic regimen incorporating MS-subtype and underlying genetic drivers of disease.