The gut microbiota in pediatric multiple sclerosis and demyelinating syndromes

Abstract Objective To examine the gut microbiota in individuals with and without pediatric‐onset multiple sclerosis (MS). Methods We compared stool‐derived microbiota of Canadian Pediatric Demyelinating Disease Network study participants ≤21 years old, with MS (disease‐modifying drug [DMD] exposed and naïve) or monophasic acquired demyelinating syndrome [monoADS] (symptom onset <18 years), and unaffected controls. All were ≥30 days without antibiotics or corticosteroids. V4 region 16S RNA gene‐derived amplicon sequence variants (Illumina MiSeq) were assessed using negative binomial regression and network analyses; rate ratios were age‐ and sex‐adjusted (aRR). Results Thirty‐two MS, 41 monoADS (symptom onset [mean] = 14.0 and 6.9 years) and 36 control participants were included; 75%/56%/58% were female, with mean ages at stool sample = 16.5/13.8/15.1 years, respectively. Nine MS cases (28%) were DMD‐naïve. Although microbiota diversity (alpha, beta) did not differ between participants (p > 0.1), taxa‐level and gut community networks did. MS (vs. monoADS) exhibited > fourfold higher relative abundance of the superphylum Patescibacteria (aRR = 4.2;95%CI:1.6–11.2, p = 0.004, Q = 0.01), and lower abundances of short‐chain fatty acid (SCFA)‐producing Lachnospiraceae (Anaerosporobacter) and Ruminococcaceae (p, Q < 0.05). DMD‐naïve MS cases were depleted for Clostridiales vadin‐BB60 (unnamed species) versus either DMD‐exposed, controls (p, Q < 0.01), or monoADS (p = 0.001, Q = 0.06) and exhibited altered community connectedness (p < 10−9 Kruskal–Wallis), with SCFA‐producing taxa underrepresented. Consistent taxa‐level findings from an independent US Network of Pediatric MS Centers case/control (n = 51/42) cohort included >eightfold higher abundance for Candidatus Stoquefichus and Tyzzerella (aRR = 8.8–12.8, p < 0.05) in MS cases and 72%–80% lower abundance of SCFA‐producing Ruminococcaceae‐NK4A214 (aRR = 0.38–0.2, p ≤ 0.01). Interpretation Gut microbiota community structure, function and connectivity, and not just individual taxa, are of likely importance in MS.


Introduction
The human microbiome's combined genetic load surpasses that of human genes with bacterial protein-coding genes estimated as being over 300 times more abundant. Most (>90%) of the human microbiota reside in the gastrointestinal tract. 1 Alterations in the gut microbiota may be influential in neurological diseases, including multiple sclerosis (MS). 2 The gut microbiota regulates the immune system and contributes to the maturation and modulation of the CNS, including myelination, via multiple complex mechanisms. 3 MS is considered an immune-mediated and neurodegenerative disease, with the CNS being the primary target. While both genetic and early-life environmental exposures are implicated in triggering MS, current knowledge surrounding these exposures remains incomplete. Animal models of CNS demyelination provide a proof-ofprinciple that the gut microbiota influence CNS-directed immune responses. 4,5 Studies involving persons with MS, while still limited in size, suggest that compared with controls, subtle differences in key gut microbial taxa exist. 6 The concept of a "period of risk" during which the inciting biology is triggered is important when considering risk factors for chronic disease. A symptomatic prodromal period, possibly extending for years before clinical MS onset in adults, has recently been recognized. 7 Further, childhood and adolescence are key periods of risk exposure for MS. As such, analysis of the gut microbiota in pediatric-onset MS patients represents a unique opportunity to examine pathological processes closer to actual risk acquisition. Children and youth have accrued fewer confounding exposures, such as medications and medical comorbidities compared to adults, permitting a unique window into the native gut microbiota. 8 We compared the gut microbiota from stool samples of well-characterized persons with pediatric-onset MS and unaffected controls in a case-control study, taking into consideration any prior disease-modifying drug (DMD) exposure, and capturing key features seldom considered in MS studies, including other medications, dietary supplements, and stool consistency (the Bristol Stool Scale). 8,9 In addition, we included another disease group--participants with monophasic acquired demyelinating syndromes (monoADS)--to serve as an additional comparator to the chronic disease, MS. Generalizability of main findings was sought in an independent case-control cohort of pediatric-onset MS and unaffected controls.

Study design and participants
This case-control study was embedded within two larger prospective North American studies of pediatric-onset MS and related demyelinating diseases. Participants ≤21 years old who provided a stool sample and had monoADS or MS (McDonald criteria, 2017) and symptom onset (first clinical attack) <18 years or were an unaffected control were eligible. MonoADS was defined as an initial acute clinical episode of symptoms involving the CNS, with evidence of inflammatory demyelination and no new clinical or MRI findings of recurrent demyelination (median observation from first symptom onset = 9.1 years, range = 3.1-12.9 years). 10 Unaffected controls had no known neurological or (auto) immune-related condition (headache/migraine, asthma, and allergies were permissible) and were recruited using a mixed-methods approach (e.g., via general pediatric clinic posters, and web-based advertising), with the aim of enrolling age, sex, race, and geographical location representative individuals.
Informed assent/consent were obtained from participants/guardians. Ethical approval was obtained from each institution's research ethics board.
The main and complementary analyses were conducted for the 'Canada-USA cohort' which comprised MS cases, monoADS, and unaffected controls enrolled from four Canadian and one USA site (Children's Hospital of Philadelphia), between 11/2015 and 03/2018 through the Canadian Pediatric Demyelinating Disease Network. A second, independent "USA-only cohort" comprised MS cases and unaffected controls enrolled from eight USA sites, between 06/2012 and 03/2018 through the US Network of Pediatric MS Centers was used to test generalizability of findings.
Cohort characteristics were captured for participants primarily through standardized forms and questionnaires administered to the participant/caregiver by trained coordinators at stool sample collection (details of data sources and categorization of variables are in Supplementary Methods). Briefly, these included demographics: age, sex, country of birth/residence, and race (white, non-white); clinical: comorbidities, body mass index (BMI = height (kg)/weight(m) 2 ), cigarette smoking (active or passive), medication use (any prior DMD use for MS, and, in the 30 days pre-stool sample, any other medication/dietary  11 and, for the Canada-USA cohort, the Bristol Stool Scale, 9 adapted for children. The validated food screener captured the prior week's diet, 11 reported as the percentage caloric intake of protein, fat, and carbohydrate and total grams of fiber. The seven-point ordinal Bristol Stool Scale captures stool consistency, considered a useful reflection of the gut ecosystem, and is associated with gut microbiota composition 9 and was categorized into: hard (types 1-2); medium (3)(4)(5); or loose (6-7).

Stool sample collection, sequencing, and bioinformatics
A common protocol was used for stool sample collection. The following were not permitted: antibiotics or corticosteroids within 30 days pre-stool sample; any history of cytotoxic immunosuppressant use or major bowel-related comorbidity (e.g., inflammatory bowel disease, IBD). The same collection kits were used for all participants, with stool shipped on ice before À80°C storage in the central laboratories (University of Manitoba IBD Clinical/ Research Centre, Winnipeg, Canada or UCSF, USA), with all sequencing performed together (batched) at the National Microbiology Laboratory, Winnipeg. Dry ice was used for cross-border shipping (USA to Canada) to prevent thawing.
Alpha and beta-diversity were examined as evenness, richness (Shannon, Margalef's index, Chao1), and weighted UniFrac. 15 Gut microbiota network analyses (genus-level) used the R-package SPIEC-EASI (SParse InversE Covariance Estimation for Ecological Association Inference, neighbourhood mode), when present in ≥80% of samples. 16 Network connectivity were quantified as degrees and betweenness. 17 The five most connected taxa were annotated and described. Predicted metagenome functions were generated using the validated Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2) algorithm, summarized as metabolic pathways (MetaCyc database). 18,19 Statistical analyses Cohort characteristics were described. The gut microbiota metrics were compared by disease, and then DMD status (grouped as three categories: MS, controls, monoADS; then four: MS [DMD-na€ ıve, exposed], controls, mono-ADS). Alpha-diversity, network metrics (connectivity and betweenness), and the metabolic pathway relative abundances were compared between groups using nonparametric tests (KruskalÀWallis [KW] rank sum test, Holmadjusted [adj .] p-values). Beta-diversity was similarly explored using permutational multivariate analysis of variance (PERMANOVA). The relative abundance of individual ASVs was compared between groups at the phylum, genus, and species-level, using sex and age at stool sample (continuous) adjusted negative binomial models. Findings were expressed as crude and adjusted-rate ratios (aRR) and 95% confidence intervals (95%CI), along with p and Q-values (false discovery rate adjusted p-values).
To guide future studies, complementary analyses were performed for the Canada-USA cohort, with alpha and beta-diversity compared by: sex, age at stool sample, race, country of residence, Bristol Stool Scale, BMI, dietary intake (protein, carbohydrate, fiber, and fat), and other medications/dietary supplements), categorized as shown in the Supplementary Methods. Finally, key main analyses (alpha-, beta-diversity and genus, and species-level comparisons) were performed using a similar approach for the pediatric-onset MS cases (DMD-na€ ıve and exposed) and unaffected controls within the independently acquired USA-only cohort. Statistical analyses were performed using R (V.4.0.2).
All cases had relapsing-remitting MS, and the mean disease duration (from symptom onset) at stool sample = 30.1 months (SD:35.1) in the Canada-USA cohort and 16.2 months (SD:13.3) in the USA-only. Approximately one-third of MS cases had never used a DMD prior to stool sample procurement [nine (28%) Canada-USA and 18 (35%) USA-only]. Beta-interferon or glatiramer acetate were most commonly used (Tables 1 and 2).
For the Canada-USA cohort, the mean number of non-DMD medication/supplement classes used in the previous 30 days = 2.0 for MS, 1.1 for monoADS and 0.8 for control. The most common were vitamins/dietary supplements, with >80% (n = 26) of MS cases, 66% (n = 27) of monoADS and 28% (n = 10) of controls taking ≥one. Atopy was common, affecting 32 (29%) of the participants, but only 4 (4%) had any other comorbidity ( Table 1).

Gut diversity and taxa-level findings by disease and DMD status
Alpha and beta-diversity did not differ by disease status (across the three groups compared: MS monoADS, controls) or by DMD status (four groups: MS DMD-exposed/ na€ ıve, monoADS, controls), all p > 0.1. Figure 1 depicts richness, all other diversity metrics are shown in Table 3.
Other genera appeared particularly relevant in differentiating monoADS participants from the other groups-both Actinomyces and Bacteroides differed versus the MS cases or controls (p < 0.05, although not all Q < 0.05), while no differences emerged when the MS cases and controls were directly compared (p > 0.05). Shared features for both disease groups were observed: versus controls, both Ruminococcaceae-NK4A214 group and (Eubacterium) eligens were lower in MS and monoADS (p,Q < 0.05), while the two disease groups did not differ (p > 0.05). Finally, for the remaining three genera, findings were largely driven by differences in relative abundance between MS and monoADS participants. For example, both Lachnospiraceae and Ruminococcaceae were lower in MS (and the DMD-exposed subgroup) versus monoADS, while the latter were enriched versus controls (all p,Q < 0.05). Conversely, Enterorhabdus was higher in MS versus monoADS (aRR = 26.2;95%CI:4.6-149.3, p,Q < 0.02), with the direction of findings consistent irrespective of the MS cases' DMD-exposure (p < 0.01). Further, the monoADS participants were depleted versus controls (p,Q < 0.05).

Gut microbiota network analysis by disease and DMD status
From network analyses, the MS cases, monoADS, and controls' genus-level gut microbiota did not differ by degree of connectivity or betweenness (p > 0.1, Figs. 3  and 4). However, findings differed by DMD status; the na€ ıve MS cases exhibited a visually distinct gut microbial network (Fig. 3) and had a higher connectivity (betweenness) versus the other three groups (DMD-exposed cases, monoADS, and controls, all p < 0.00007, Fig. 4). Further, annotation of the five most connected taxa suggested distinct patterns across groups (Fig. S1A-E  Four-groups compared: multiple sclerosis (DMD-na€ ıve and exposed), ADS, and controls. (D) Fourgroups compared: multiple sclerosis (DMD-na€ ıve and exposed), ADS, and controls, overlaid with a hierarchical cluster analysis. ADS = monophasic acquired demyelinating syndrome, DMD = disease-modifying drug, MS = pediatric-onset multiple sclerosis; MS DMD-na€ ıve = MS case has never been exposed to a DMD at the time of the stool sample. Each Panel summarizes age and sex-adjusted RRs derived from a single negative binomial regression model for each genus (two models in total, one for three group comparison, and another for four group comparisons), with only the RRs reaching nominal significance (p < 0.05) for at least one group comparison within a genus shown (see Tables S1-S3)  (2) MS DMD-exposed versus control; (3) MS DMD-exposed versus na€ ıve; (4) ADS versus MS DMD-exposed; (5) ADS versus MS DMD-exposed; (6) ADS versus control. (B and D) Findings are ordered according to the hierarchical cluster analysis (R package pheatmap). Briefly, each taxon is assigned to its own cluster, then the algorithm proceeds iteratively, at each stage joining the two most similar clusters, continuing until there is a single cluster. At each stage distances between clusters are recomputed using the Lance-Williams dissimilarity update formula. 22 Biological relevance is inferred from the clusters (rather than being directly assessed).

Metagenomic predictions by disease and DMD status
After filtering for sparsity, 193 of 399 identified pathways were assessed, and 50 differed between the MS, mono-ADS, and controls (p < 0.05, KW, Fig. 5 Fig. 5A and B).
The latter pathway was also lower in monoADS versus controls (adj.p = 0.0410). While the direction of these findings was consistent regardless of whether MS cases were DMD-na€ ıve or exposed (unadj.p < 0.05), others differed, although none remained significant after multiple comparison adjustments. For example, both the reductive tricarboxylic acid cycle I and glycolysis pathways were enriched for the DMD-na€ ıve MS cases (unadj.p < 0.05, Fig. 5C).

Complementary analyses
For all Canada-USA participants combined, those exposed (versus unexposed) to other medications/supplements exhibited lower alpha-diversity (richness, Chao1 p < 0.02). Beta-diversity also differed by other medication use, race, country of residence, and dietary intake (fiber), but all associations were small, explaining 3-8% of the variability in the gut microbiota (all p < 0.02). No other findings reached significance (all p > 0.05; Table S5). Comparisons with the independent USA-only cohort

Diversity
Consistent with the Canada-USA cohort, alpha and beta gut microbiota diversity did not differ for the MS cases (DMD-na€ ıve or exposed) versus the unaffected controls, all p > 0.1 (data not shown).

Taxa-level findings
While consistency in the direction of effect was observed across several genera and species for the MS cases and controls within each cohort, none reached both p and Q < 0.05 in both cohorts (Fig. 6 also lower in MS in both cohorts (aRR < 1, but p > 0.05), and all were housed in the phylum Firmicutes, including Veillonella and three within the order Clostridiales (Roseburia, Ruminococcaceae UCGÀ003). In addition, some genera were higher in MS versus controls in both cohorts, such as: Candidatus Stoquefichus (also identified as highly connected in the Canada-USA network analyses) and Tyzzerella aRR Canada-USA 10.1-12.8; aRR USA 8.8-9.1, all p < 0.05). The genera Clostridium sensu stricto 1 and Turicibacter specifically differed by the MS cases' DMD status in both cohorts, being higher for the DMDexposed versus na€ ıve (aRR Canada-USA 4.1-5.5 and aRR USA 2.6-2.9, all p < 0.05). Further, compared to controls, the DMD-na€ ıve cases exhibited a significantly higher relative abundance of the species Clostridium innocuum group (genus Erysipelotrichaceae; aRR exceeded 10, with p < 0.03 in both cohorts) and lower abundance of an unnamed species within the genus Ruminococcaceae UCGÀ003 (aRR<0.4, p < 0.047 in both cohorts). Conversely, two of the genus-level findings differed across cohorts; both were higher for the MS cases (vs. controls) in the Canada-USA but lower in the USA-only cohort (all p < 0.05). These comprised Pseudomonas (Proteobacteria phylum) (aRR Canada-USA 5.5;95%CI:1.5-20.2, p = 0.01; aRR USA 0.16;95%CI:0.03-0.85, p = 0.03) and an unnamed taxon within the functionally diverse Lachnospiraceae family (aRR Canada-USA 2.6;95%CI:1.4-5.1, p = 0.004; aRR USA 0.57;95%CI:0.37-0.88, p = 0.001). The latter similarly differed for the DMD-exposed cases, Figure 4. Gut microbiota network analysis (genus-level): box plots show node (taxa) connectivity (betweenness) for the unaffected controls, monoADS, and pediatric-onset MS cases (DMD-exposed and na€ ıve). ADS = monophasic acquired demyelinating syndrome; MS = multiple sclerosis (pediatric-onset). Summary: node (taxa) connectivity (betweenness) differed between the 4 groups: controls, monoADS, DMD-exposed and na€ ıve MS cases (p < 10-9 Kruskal-Wallis) The DMD na€ ıve MS cases also differed from each of the groups -controls, monoADS, & DMD-exposed MS cases (all p < 0.00007, Holm adjusted for multiple pairwise comparisons). Additional information and related analyses: The degree of connectivity did not differ between groups (all p > 0.05, data not shown). On completion of the pre-planned analyses, post-hoc principal component analysis (PCA) plots (genus-level) were generated to explore the potential for bias for the following elements: for the Canada-USA cohort, presence of atopy, and use of supplements (all participants combined, then separated by disease status), for both cohorts combined, shipping-related (use of additional dry ice shipment [yes, no]) and cohort (Canada-USA, USA- Figure 6. Heatmap summarizing gut microbiota genus-level findings (ASV counts): Comparisons between the Canada-USA and USA-only cohorts for the multiple sclerosis cases and controls. Adjusted RRs were ordered from highest to lowest for the multiple sclerosis cases versus controls, first, for the Canada-USA cohort, then for the USA-only. Each Panel summarizes age and sex-adjusted RRs derived from negative binomial regression models for each genus, with only the RRs reaching nominal significance (p < 0.05) in at least one of the cohorts within a genus shown (see Tables S2 and S6) for unadjusted and adjusted models). For each comparison, the control group forms the reference. RRs are colored from high (red) to low (blue); gray shading-not available. *p < 0.05, **p < 0.01, ***p < 0.0001, ***+p < 0.0001 and Q < 0.05. relative abundance(s) of pathways were lower for the MS cases relative to unaffected controls, with the direction of findings remaining consistent regardless of DMD exposure (A and B) or differed by DMD exposure status, being lower for the DMD-exposed relative to na€ ıve MS cases (C). All predicted metagenomic findings are shown in Table S4. Box plots: Thick black horizontal line = median; horizontal edges of box depict Q1 and Q3 (interquartile range); the ends of the whiskers represent one and a half times the interquartile range (1.5*IQR); circles = individual outliers. Y axis represents the mean relative pathway abundance from metagenomic predictions, derived from a validated algorithm via PICRUSt2 (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) and summarized as metabolic pathways using MetaCyc (database of metabolic pathways and enzymes from all domains of life). ADS = monophasic acquired demyelinating syndromes; MS = multiple sclerosis (pediatric onset); DMD = disease modifying drug; DMD+/DMDÀ = DMD disease modifying drug exposed/na€ ıve (never exposed); KW = Kruskal-Wallis rank sum test; adj.p = multiple comparisons adjusted p-values (derived from the Dunn Kruskal-Wallis test with Holm adjustment for multiple comparisons). Bolded p-values indicate <0.05 (reached nominal significance). only). No remarkable patterns or clustering of individuals based on these elements were observed (all PCA plots available upon request from the authors).

Discussion
While overall gut microbiota diversity did not differ between individuals with pediatric-onset MS (DMD-na€ ıve or exposed), monoADS, and unaffected controls, important taxa-level differences emerged. Compared to mono-ADS, MS cases exhibited a fourfold higher relative abundance of the recently identified superphylum Patescibacteria (aRR = 4.2; p,Q < 0.05), and a lower abundance of several butyrate-producing genera and species within Lachnospiraceae (e.g., Anaerosporobacter) and Ruminococcaceae families (p,Q < 0.05). An unnamed species within the genus Clostridiales vadin BB60 differentiated the DMD-na€ ıve MS cases from all others, being depleted versus either the DMD-exposed cases, controls (p,Q < 0.01) or monoADS participants (p = 0.001, Q > 0.05). In addition, for all MS cases (either DMDexposed or na€ ıve), gut metagenomic predictions suggested depletion in the tryptophan-related nicotinamide adenine dinucleotide salvage and SCFA-producing pyruvate fermentation pathways versus controls (Holm-adjusted p < 0.03). These pathways were recently found of importance in persons with MS and MS animal models. 23 Finally, network analyses revealed a distinct gut microbiota community structure for the MS cases versus mono-ADS and controls, with an overrepresentation of highly connected opportunistic pathogens, and an under representation of SCFA-producing taxa. Together, findings suggest that the gut microbiota community structure, function and connectivity, and not just individual taxa, are of likely importance in MS.
We were able to compare our main findings with an independent cohort of pediatric-onset MS cases and unaffected controls. Consistent across both cohorts, and with most other studies, 4,6,24-27 neither alpha nor beta gut microbiota diversity differed significantly for the MS cases (DMD-na€ ıve or exposed) versus controls. At the individual ASV-level, these included a lower relative abundance of several SCFA-producing taxa for the MS cases versus controls, such as Ruminococcaceae-NK4A214 group which was 61%-79% lower for the cases (aRR:0.39-0.21) and 72%-90% lower for the DMD-na€ ıve cases (aRR:0.28-0.10) across both cohorts, all p < 0.02. Depletion of this genus has been reported in MS and other chronic diseases, 24,28,29 as has depletion of other SCFA-producing taxa in MS (versus controls), including for example, Butyricimonas 27 and Clostridia clusters XIVa and IV. 25 Christensenellaceae R-7 group, a member of the highly inheritable Christensenellaceae family, 30 was lower for the MS cases versus controls across both cohorts, by 42%-52% at the genus-level (aRRs:0.58-0.44) and 75-87% at the species-level (aRRs:0.25-0.13), reaching nominal significance (p < 0.05) in the USA-only cohort at the genuslevel, and in the Canada-USA cohort at the species-level. Christensenellaceae appears depleted in other immunemediated conditions, such as IBD, as well as MS, suggesting that diminution of these genera may be important for systemic T-cell dysregulation disorders. 6,30 A large metaanalysis, comprised 3,048 individuals, identified Christensenellaceae R-7 group, along with Ruminococcaceae UCG-005, as among the top five genera enriched in controls versus IBD-participants. 31 Considered as potential biomarkers of a healthy gut, 31 these observations broadly concurred with ours, including, for example, our Canada-USA cohort's gut community network analyses; Ruminococcaceae UCG-005 genus was among the top five most connected taxa for the unaffected controls. Authors of the same meta-analyses identified members of the family Erysipelotrichaceae as potential markers of gut inflammation. 31 Relatedly, we observed >10-fold higher relative abundance of Clostridium innocuum group sp., housed within this family for our MS cases in both cohorts (p < 0.03). Interestingly, both of the genus-level taxa that differed between our cohorts (families Pseudomonas and Lachnospiraceae) house functionally diverse species, such that inconsistencies in findings across studies and diseases are not unexpected. 32 Intriguingly, within our main Canada-USA cohort, the DMD-na€ ıve MS cases' gut communities were more connected (betweenness) versus all others (DMD-exposed cases, monoADS, or controls), perhaps indicative of more resilient pathogenic or pro-inflammatory microbial communities in the native MS gut. Findings warrant further investigation in larger populations, although accessing sizable groups of individuals with MS who are entirely DMD-na€ ıve is challenging in today's therapeutic era. We found just one other study assessing gut community structures in adults with MS (aged 20-63 years). 33 They also observed differences in network connectivity between DMD-na€ ıve MS cases (n = 45) and controls (n = 44). 33 While the degree of connectivity was lower for the MS cases in this older population, inferences are similar to ours; gut microbiota communities and network structure are of likely importance, not just individual taxa. We annotated our gut microbiota community networks, observing overrepresentation in MS cases of opportunistic pathogens, for example, Actinomyces (Actinobacteria phylum), Gemella, and Leuconostoc (Firmicutes phylum). In contrast, for controls, SCFA-producing taxa dominated, including Ruminococcaceae family members, Anaerostipes and Veillonella (Firmicutes phylum). Other members of the Veillonellaceae family are also reported as higher in controls (versus DMD-na€ ıve MS case), 5 and higher in DMD-na€ ıve versus exposed MS cases. 34 SCFAs represent key microbial metabolites, can exert anti-inflammatory effects, and may facilitate beneficial microbiota-gut-brain interactions. 35 Gut metagenomic predictions suggested depletion of the nicotinamide adenine dinucleotide salvage and pyruvate fermentation pathways in MS (DMD-exposed or na€ ıve) versus controls. Both pathways are implicated in MS pathogenesis, possibly mediated via SCFA-production (acetate and lactate), neuronal mitochondrial damage and energy depletion. [36][37][38][39] Modification of these pathways may has potential neuroprotective and immunomodulatory effects in MS. [36][37][38][39] Our findings concur with the broader MS literature; for example, dietary precursors of the former nicotinamide salvage pathway include tryptophan and niacin (vitamin B 3 ), and a higher abundance of gut-derived tryptophan metabolites were associated with a lower risk of pediatriconset MS and subsequent disease activity. 23 Two other pathways-reductive tricarboxylic acid cycle I and glycolysis -differed by DMD status, being lower for our DMDexposed versus na€ ıve MS cases. Both are considered central hubs for energy metabolism, 40,41 and others report upregulation within active MS lesions, and a relationship of these pathways with disease severity. 42,43 Further, pathways related to energy metabolism (e.g., methane) and metabolism of other vitamins (e.g., retinol, vitamin A) have been shown to differ by DMD status. 34 These predicted pathways provide mechanistic insights warranting future direct functional characterization of the microbiota.

Strengths and limitations
Pediatric-onset MS remains relatively rare, such that our study size was modest. Nonetheless, our participants were well phenotyped; our MS cases, controls, and monoADS participants were similar for important metrics rarely captured, such as stool consistency (via the Bristol Stool Scale). This is considered an important confounder in gut microbiota studies. 9 Household controls (typically spouses, sometimes siblings) can be another approach, but are not well suited to a rare pediatric disease for several reasons. 2,44,45 Adults are not suitable controls for children, and declining fertility rates renders it impractical to enroll a household sibling of similar age and sex (North American women average <2 children). 46 Overmatching and misclassification also pose a threat 44,47 ; an unaffected sibling is genetically predisposed to be at a higher risk of MS, but may not develop MS until some (unknown) time in the future.
Our MS cases had a relatively short disease duration, averaging 2.5 years from symptom onset in our primary Canada-USA cohort, and had thus accrued few comorbidities (aside from atopy), and had a low medication burden, aside from dietary supplement use. Nonetheless, our complementary analyses suggested some modest differences in gut diversity based on broad cohort characteristics, including race, country of residence, fiber intake and other medication, and supplement use. These modest differences, and others, may become relevant in larger cohorts and warrant further consideration, especially as comorbidity and chronic medication/supplement use are common, increase with age and disease duration, 48 and most MS microbiota studies have included older adults with very long disease durations. 6 This raises the possibility that these seldom reported exposures may have a profound impact on the adult MS gut microbiota, including, for example, the commonly used antipsychotics. 16,17 Further, supplement use (e.g., vitamin D) is common in persons with MS of all ages; our findings suggest that future work is needed to establish its full potential impact on the MS gut microbiota. Our inclusion of pediatric-onset MS cases who were never exposed to a DMD and use of an additional comparator group (monoADS) may be of value to advance understanding of which differences in the gut microbiota are specific to MS, or are common to other neurological or demyelinating diseases. We also compared our main findings with an independent cohort of pediatric-onset MS cases and unaffected controls, with all samples collected in a similar manner and sequenced in the same central facility. 49 Finally, we were able to assign taxonomy using the newer ASVs 50 rather than the operational taxonomic unit system employed previously in MS. 6 ASVs are considered advantageous in achieving greater resolution and may have enhanced our ability to detect previously unrecognized taxa of importance in MS.

Conclusions
Gut microbiota diversity was similar for pediatric-onset MS cases versus either monoADS or unaffected controls. However, at the taxa-and gut-community-network-level, differences were observed. MS cases, irrespective of prior DMD exposure, exhibited an overrepresentation of highly connected opportunistic pathogens, and an under representation of SCFA-producing taxa. Further, several SCFAproducing taxa, such as Ruminococcaceae NK4A214, and Christensenellaceae R-7 group, identified as possible universal makers of gut health, 31 were consistently lower for the pediatric-onset MS cases versus unaffected controls across two independent North American cohorts. Together, findings suggest that commonality in the gut microbiota composition can be found across different MS cohorts, and that disruptions in key taxa may contribute to MS pathogenesis. Further, findings suggest that the gut microbiota community structure, function, and connectivity, and not just individual taxa, are of likely ª 2021 The Authors. Annals of Clinical and Translational Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association importance in MS. Further work is warranted to delineate the likely bi-directional relationship between the gut microbiota and MS.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Data S1. Phenotyping participants and relevant data sources.