Микробиота гута и метаболиты связаны с развитием болезней в множестве склерозов.

Introduction

Multiple studies have identified abnormalities in the gut microbiome in multiple sclerosis (MS). 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 Intestinal microbes and their metabolites are known to influence a wide variety of processes related to both health and disease. 13 , 14 , 15 , 16 , 17 In terms of MS, a major question is the manner in which intestinal microbes and their metabolites are related to different stages of the disease and to the development of progressive MS from relapsing remitting forms. Studies in relapsing MS have identified increased Akkermansia muciniphila and decreased butyrate producers. 4 , 12 , 18 , 19 In a cross-sectional study of progressive MS, we identified microbes associated with clinical measurements of disease including disability, brain MRI, and quality of life 12 and found beneficial associations with Akkermansia. A major unanswered question is the degree to which the gut microbiome and its metabolites are associated with both longitudinal changes in clinical and radiological measures of disease severity and in the transition to progressive MS.

Microbial metabolites influence health and disease by affecting both immunologic and neural pathways. Serum and stool metabolites are altered in MS, including lipids, fatty acid metabolites, 20 secondary bile acids, 21 and tryptophan derivatives. 22 Several of these metabolites are known to be produced by the gut microbiota, and secondary bile acids are made exclusively by intestinal microbes. Individuals with MS have decreased stool levels of microbially derived metabolites providing a rationale for clinical trials with these metabolites. 13 , 21 , 23

Disease progression in MS can be measured by change in disease severity, MRI measures, quality of life measures (e.g., fatigue), biomarkers of degeneration, and most importantly by the transition from the relapsing to the progressive form of MS. The Comprehensive Longitudinal Investigation of MS at the Brigham and Women’s Hospital (CLIMB study) comprises a cohort of patients followed longitudinally with regular clinical assessment, MRI, and blood and stool collections, and we have previously demonstrated the utility of fluid biomarkers including neurofilament light and glial fibrillary acidic protein in predicting disease worsening using CLIMB samples. 24 , 25 The CLIMB study thus provides a unique opportunity to understand how the gut microbiota and its metabolites are associated with disease progression in MS. Using the CLIMB cohort, we identified microbes and metabolites that decreased in subjects who worsened clinically vs. those who remained stable. In addition, we identified microbes associated with brain MRI measures of disease and quality of life. Our findings identify pathways by which the gut microbiome and its metabolites influence progressive MS.

Results

Identification of microbiota linked to disease worsening

To identify microbes linked to clinical outcomes, we investigated subjects with MS whose expanded disability status scale (EDSS) worsened over a 2-year period, as defined by an increase of ≥1, and compared them to subjects with MS with an EDSS change of ≤1 point or improved ( Figure 1 A). This is an EDSS change routinely defined as the minimally clinically important difference. 26 Subjects with a baseline EDSS between 1.5 and 3.5 and who remained on the same disease-modifying therapy (DMT) during the 2-year follow-up period were selected for analysis. At baseline, there were no differences in EDSS between groups. As expected, there was a significant increase in EDSS in the worsening group ( Figure 1 B; p  < 0.0001). Of the subjects who worsened, 20.8% had a relapse during the 2-year follow-up, whereas relapses were observed in 8.2% of subjects without EDSS worsening ( Figure 1 C). Age, BMI ( Figure 1 D), DMT, or the percentage of male vs. female subjects was not different between stable and worsened groups ( Table S1 ), and there was no difference in 2-year EDSS related to DMT ( Figure 1 E). α-diversity metrics including Evenness, Faith’s phylogenetic diversity (PD), Shannon’s index, or richness were not different between EDSS groups ( Figure 1 F). We visualized effect size of potential confounders associated with variation in microbiota via a weighted Adonis R 2 and did not find any significant contribution of these factors on permutational multivariate ANOVA ( Figure 1 G). β-diversity was not altered in stable vs. worsened EDSS groups as measured by weighted and unweighted UniFrac distances ( Figure 1 H). Microbiome multivariable associations with linear models (MaAsLin2) 27 was used to associate microbial amplicon sequence variants (ASVs) with a 2-year change in EDSS controlled for covariates including age, BMI, DMT, and the percentage of male and female subjects. We found four ASVs negatively associated with a 2-year change in EDSS and three ASVs positively associated ( Figure 1 I, p  < 0.05). ASVs were initially identified using the SILVA taxonomic database in QIIME2. For further identification, we also used the basic local alignment search tool (BLAST, Table S2 ). We found that Alistipes ( A. onderdonkii , 100% BLAST identity) and 2 Bacteroides ASVs ( B. xylanisolvens , 100% BLAST identity and Phocaeicola vulgatus , 100% BLAST identity) were associated with an increased 2-year change in EDSS and Streptococcus ( S. thermophilus , 100% BLAST identity), Eubacterium hallii ( A. hallii , 100% BLAST identity), Butyricicoccus ( A. butyriciproducens , 100% BLAST identity), and Eggerthella ( E. bacterium , 100% BLAST identity) were associated with a decreased 2-year change in EDSS. Consistent with this, we found that subjects who worsened had decreased Streptococcus ASV -0179bed8fd230300765694433b3858c9 and increased Bacteroides ASV-024a4a4e09775389a6cd47c02a336ddb at baseline ( Figure 1 J).

Identification of microbiota linked to disease worsening

Schematic of experimental workflow (A). EDSS changes between 0 and 2 years in subjects who were “stable” (no change or improved) or “worsened” (B). Pie charts showing the percentage of subjects with a clinical relapse during the 2-year follow-up by group (C). Age and BMI across groups (D). Change in EDSS between 0 and 2 years grouped by DMT (E). α-diversity measures for Evenness, Faith’s PD, Shannon’s, and Richness across groups (F). Bar plot of effect size (weighted Adonis R 2 ) of potential confounders associated with microbial variation (G). β-diversity as depicted by unweighted and weighted UniFrac distances (H). MaAsLin2 controlled for age, the percentage of male and female subjects, and BMI, and processed using a pseudo-count, shows microbial taxa significantly associated with a change in EDSS at the 2-year follow-up (I) and representative plots of relative abundances for the highest and lowest taxa associated with EDSS change (J). Silva vs. GreenGenes2 taxonomic assignments and associated BLAST results are in Table S2 n  = 124 “stable/improved” and n  = 68 “worsened” subjects, full demographics in Table S1 . Unadjusted ∗ p  < 0.05, ∗∗ p  < 0.01, and ∗∗∗ p  < 0.001. Data in (C), (D), and (J) are presented as mean ± SEM, and (E) and (F) are violin plots where width is distribution density and lines are median and upper/lower quartiles.

Use of a DMT may shift gut microbiota composition. 1 , 12 We investigated the relationship between microbiota in subjects on one of eight common DMTs vs. untreated subjects. We found multiple Blautia genera and Eubacterium hallii elevated in subjects on anti-CD20 therapy compared to untreated subjects. Furthermore, anti-CD20 therapy was negatively associated with Erysipelotrichaceae and Bacteroides dorei ( Figure S1 ).

Gut microbes correlate with 2-year changes in 3T MRI

Brain atrophy and lesion volume are well-established measures of disease severity in MS. We investigated the association between the microbiota and longitudinal MRI changes. When we compared untreated subjects with those receiving DMTs, we found no differences in changes in total deep gray matter (DGM) volume ( Figure 2 A), whole brain parenchymal volume ( Figure 2 B), and whole brain T2 white matter lesion volume (T2LV; Figure 2 C). In addition, MRI changes were not correlated with a 2-year change in EDSS, age, or BMI ( Figure S2 ). Microbial taxa correlated with changes in seven brain volumetric measures. We found multiple bacteria that correlated with changes in whole brain volume, DGM, thalamus, caudate, putamen, and globus pallidus volumes, including Lachnospiraceae and Alistipes genera members ( Figure 2 D). We also found microbes that correlated with T2LV, including A. onderdonkii and Streptococcus ( Figure 2 E) . T2LV and atrophy represent different disease processes in MS. T2LV represents white matter foci of inflammatory demyelination whereas atrophy is associated with widespread neurodegeneration of gray and white matter. We found unique microbes associated with T2LV or atrophy, suggesting that the microbiome influences T2LV vs. atrophy by different mechanisms.

Gut microbes correlate with 2-year changes in 3T MRI

Changes in DGM (A), whole brain parenchymal volume (B), and T2LV stratified by DMT (C). Spearman’s correlations demonstrate microbial taxa significantly positively (red) or negatively (blue) correlated with 2-year changes in 3T volumetric measures (D). Spearman’s correlation demonstrates taxa positively (red) or negatively (blue) correlated with T2LV (E). n  = 60 subjects with RRMS and n  = 14 subjects with ProgMS, full demographics in Table S1 .

Unadjusted ∗ p  < 0.05, ∗∗ p  < 0.01, and ∗∗∗ p  < 0.001. Data are presented as truncated violin plots where width is distribution density and lines are median and upper/lower quartiles.

Gut microbes associated with cognitive changes in MS

It is well recognized that cognitive changes occur in MS. 28 Cognitive testing was performed in a subset of subjects ( Table S1 ) within 6 months of microbiome sampling, and 2 years later. We investigated subjects whose symbol digit modality test (SDMT) increased, was stable, or decreased ( Figure 3 A), defined as a change of ±4 points as previously described. 29 SDMT outcomes were not associated with age or EDSS ( Figures 3 B and 3C). Three taxa were correlated with worsening of SDMT, including a Ruminococcus gauvreauii group member, Lachnoclostridium , and Eubacterium ( Figure 3 D). Similar to microbes associated with less brain atrophy ( Figure 2 D), a Ruminococcus genus member and a Peptostreptococcaceae member were associated with an improved SDMT score ( Figure 3 D). Another measure of cognition in MS is the California verbal learning test (CVLT-II), which measures speed of language processing. 30 A change in the CVLT-II was defined as an increase or decrease of 3 points over a 2-year period as previously defined 30 , 31 ( Figure 3 E). A decreased CVLT-II score was associated with increasing age ( Figure 3 F, p  = 0.04; Spearman’s rho = −0.243, p  = 0.022), but not with EDSS ( Figure 3 G). We identified microbes that correlated with a 0–2 year worsening on the CVLT-II score, including Clostridium sensu stricto and Turicibacter . We also identified multiple Lachnoclostridium and Bacteroides genera associated with an improvement in CVLT-II, suggesting a beneficial role ( Figure 3 H).

Gut microbes associated with cognitive changes in MS

Line plots show SDMT trajectories of subjects by group from 0 to 2 years (A). Age of subjects and correlation of age with a 2-year change in SDMT (B). EDSS stratified by SDMT group and correlation of SDMT over 2 years with EDSS (C). Heatmap of microbial taxa correlated with 0 to 2-year change in SDMT, with positive correlation in red and negative in blue (D). Line plots show CVLT-II trajectories of subjects by group from 0 to 2 years (E). Age of subjects and correlation of age with a 2-year change in CVLT-II (F). EDSS stratified by CVLT-II group and correlation of CVLT-II over 2 years with EDSS (G). Heatmap of microbial taxa correlated with 0 to 2-year change in CVLT-II, with positive correlation in red and negative in green (H). n  = 95 subjects for SDMT and CVLT-II, full demographics in Table S1 .

Unadjusted ∗ p  < 0.05, ∗∗ p  < 0.01, and ∗∗∗ p  < 0.001. Data in (B) and (F) are presented as mean ± SEM, and (C) and (G) are truncated violin plots where width is distribution density and lines are median and upper/lower quartiles.

Gut microbes associated with quality-of-life changes in MS

The Quality of Life in Neurological Disorders (Neuro-QoL) survey assesses health-related quality of life across 11 physical, mental, and social domains. 32 Neuro-QoL domain T scores between subjects with relapsing-remitting MS (RRMS) and progressive MS (ProgMS) were stable across 0–2 years ( Figure 4 A). Similar to microbes associated with EDSS worsening, A. onderdonkii was negatively correlated with change in the lower extremity function Neuro-QoL domain ( Figure 4 B). Furthermore, there were multiple beneficial associations with Eubacterium hallii including positive correlations with change in upper extremity function, ability to participate in social roles and activities, satisfaction with social roles and activities, and cognitive function ( Figure 4 B).

Gut microbes associated with quality-of-life changes in MS

Change in Neuro-QoL domains across 0–2 years grouped by disease state (A). Microbial taxa correlated with the 11 Neuro-QoL domains show taxa positively (red) and negatively (blue) correlated with a 2-year worsening in the Neuro-QoL domain (B). n  = 88 subjects, full demographics in Table S1 .

Unadjusted ∗ p  < 0.05, ∗∗ p  < 0.01, and ∗∗∗ p  < 0.001. Data are presented as box and whisker plots with box indicating the interquartile range, the line as the median, and dots indicate points beyond 1.5x the interquartile range.

Serum metabolites linked to the development of progressive MS

To identify both microbially and host-produced metabolites related to disease progression, we performed global metabolomics on serum and stool. We compared samples from an independent cohort ( Table S3 ) with groups: healthy controls (HCs), subjects with RRMS who remained stable (RR → RR), subjects with RRMS who transitioned to secondary progressive MS (SPMS; RR → SP), and subjects with SPMS who were matched based on age, the percentage of male and female subjects, race/ethnicity, and DMT usage ( Table S3 ). Subjects who transitioned to SPMS (RR → SP) were classified by their clinician’s diagnosis based on continued worsening of neurological exam and increasing EDSS. All diagnoses were independently confirmed by a second MS clinician through chart review. Subjects who transitioned (RR → SP) had higher EDSS at the 2-year follow-up vs. baseline ( Figure 5 A) and longer 25-foot walking times than stable RR → RR subjects ( Figure 5 B). Principal component analysis showed clustering of serum metabolites by disease category and the percentage of male and female subjects ( Figure 5 C). We identified hundreds of metabolites that differed between groups, with the largest group differences detected between ProgMS and HC and the smallest differences detected between the RR → SP and RR → RR groups ( Figure 5 D). To determine if there were metabolites uniquely associated with MS progression, we identified metabolites that were altered in both subjects with RR → SP and SPMS vs. HCs but not in RR → RR vs. HCs. We found an increase in 8 lipids and 1 xenobiotic and a decrease in 6 lipids and 6 xenobiotics ( Figure 5 E). Specifically, we found elevated palmitoleate (16:1n7) and p-cresol sulfate and a decrease in the secondary bile acids iso- and ursodeoxycholate ( Figures 5 E and 5F). Furthermore, we investigated metabolites correlated with a change in EDSS and found 5 metabolites linked with worsening and 26 with stability/improvement, a majority of which were lipids or xenobiotics ( Figure S3 A).

Serum metabolites linked to development of progressive MS

Change in EDSS over 2 years in RR → RR and RR → SP subjects (A). Increased 25-foot walking times (s) in RR → SP subjects vs. RR → RR (B). Principal component analysis shows clustering of overall serum metabolites by disease status and the percentage of male and female subjects (C). Total biochemicals detected that were significantly different based on ANCOVA contrasting in serum across groups (D). All serum metabolites that were significantly altered in RR → SP and SP groups but not RR → RR group compared to healthy controls (E). Bar graphs with individual values of biologically relevant metabolic alterations by disease status (F). Volcano plot of metabolites specifically altered in RR → SP serum compared to RR → RR serum (G). Pathway analysis using pathway enrichment scores demonstrates the 17 significantly enriched pathways in RR → SP subject’s serum compared to RR → RR subjects (H). n  = 87 subjects, full demographics in Table S3 .

Unadjusted ∗ p  < 0.05, ∗∗ p  < 0.01, ∗∗∗ p  < 0.001, and ∗∗∗∗ p  < 0.0001, only metabolites with ANCOVA q  < 0.10 are plotted. Data are presented as mean ± SEM.

We next investigated metabolites linked with a recent transition to progressive MS by comparing RR → RR vs. RR → SP subjects. Volcano plot shows altered metabolites with a log 2 fold change cutoff of ±1 ( Figure 5 G). In total, 27 metabolites were altered in RR → SP compared with RR → RR subjects, with 18 being elevated and 9 decreased in RR → SP based on analysis of covariance (ANCOVA) contrasts ( Figure 5 D). Pathway enrichment score (PES) analysis was performed to identify metabolic pathways associated with the transition to progressive MS. We found 17 pathways that were enriched (PES > 1) in RR → SP subjects, with vitamin B6, fatty acid synthesis, glycerolipid metabolism, and xenobiotic respiratory drugs being the most enriched ( Figure 5 H).

Stool metabolites linked to the development of progressive MS

We next performed global metabolomics on stool and found that principal component analysis showed clustering of metabolites by disease category and the percentage of male and female subjects ( Figure 6 A). Consistent with our serum metabolomics, the greatest number of differences was found between SPMS and HCs. In addition, SPMS vs. RR → SP subjects had the fewest differences, suggesting a unique metabolomic signature associated with disease progression in MS ( Figure 6 B). We found that metabolites linked to disease progression were decreased vs. HCs ( Figure 6 C), including nicotinate, protoporphyrin IX, phenyllactate, and N1,N12-diacetylspermine ( Figures 6 C and 6D). A total of 85 metabolites were altered in subjects with RR → SP vs. RR → RR MS, with all but 5 being decreased in RR → SP ( Figure 6 E). A majority of these alterations were in amino acids, lipids, nucleotides, or xenobiotics. We identified fecal metabolic pathways that were altered in RR → SP vs. stable RR → RR MS and found an increase in 35 pathways related to sugar metabolism, polypeptides, and purine/pyrimidine metabolism ( Figure 6 F). We next investigated metabolites linked to a change in EDSS and found 7 metabolites correlated with worsening and 65 with stability/improvement, a majority of which were amino acids, lipids, nucleotides, or peptides ( Figure S3 B).

Stool metabolites linked to development of progressive MS

Principal component analysis shows clustering patterns of all metabolites by disease status and the percentage of male and female subjects (A). Total biochemicals significantly altered by group as measured by ANCOVA contrasts (B). All metabolites significantly altered in both RR → SP and SP but not in RR → RR subjects compared to healthy controls (C) and bar graphs of biologically relevant metabolites from this group (D). Volcano plot of metabolites significantly altered in RR → SP subjects compared to RR → RR subjects (E). Pathway analysis using pathway enrichment scores shows the 35 significantly enriched pathways in RR → SP subject’s stool compared to RR → RR subjects (F). n  = 87 subjects, full demographics in Table S3 .

Unadjusted ∗ p  < 0.05, ∗∗ p  < 0.01, and ∗∗∗ p  < 0.001, only metabolites with ANCOVA q  < 0.10 are plotted. Data are presented as mean ± SEM.

To investigate the relationship between serum and stool metabolites, we identified metabolites that were altered in both matrices. The majority of metabolites altered in MS were increased or decreased in both serum and stool. However, multiple lipids that were elevated in serum were decreased in stool including palmitoleate and N-oleoyltaurine ( Figure S4 ). Furthermore, we observed altered secondary bile acids associated with MS worsening in serum ( Figures S5 A and S5B) and stool ( Figures S5 C and S5D). Stool levels of short-chain fatty acids (SCFAs) including acetate, butyrate, valerate, propionate, isobutyrate, and isovalerate were not associated with disease worsening or DMT status ( Figures S5 E–S5G).

Microbiota predicted to produce metabolites linked to MS progression

Using MetOrigin analysis, 33 we identified 193 stool metabolites that were derived only from microbiota. We then used the MetOrigin microbiota-predicted metabolite pathway analysis and found that the majority of metabolic pathways were driven by metabolites comprising a combination of host and microbial factors ( Figure 7 A). We next used MMvec 34 to investigate which members of the microbiota were capable of producing stool metabolites that were altered in RR → SP vs. RR → RR subjects ( Figures 6 E and 7 B). We found that stool metabolites, which correlated with the transition to progressive MS, were associated with the production of metabolites including pyridoxamine and nicotinate by Akkermansia , xanthine and erythronate by multiple Lachnospiraceae , and sphingomyelin by Eubacterium hallii ( Figure 7 B). Furthermore, when we investigated all 85 stool metabolites ( Figure S6 ) and a subset of serum metabolites ( Figure S7 ) that were altered in RR → SP vs. RR → RR, we found that similar microbes were implicated in the production of these metabolites.

Microbiota predicted to produce metabolites linked to MS progression

Analysis of metabolite origins using MetOrigin shows host vs. microbiota-derived metabolite and pathway counts (A). Heatmap from MMvec showing log conditional probabilities of microbes producing metabolites that were altered in RR → SP vs. RR → RR subject’s stool (B). n  = 87 subjects, full demographics in Table S3 .

Discussion

Although the gut microbiome is altered in RRMS and ProgMS, 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 specific microbes and metabolites associated with disease worsening and the transition to progressive disease have not been identified. Using our longitudinal CLIMB cohort, we investigated subjects who transitioned from RRMS to progressive disease. This provided a unique opportunity to identify microbial metabolic signaling pathways involved in the clinical transition from relapsing to progressive disease, which is the major cause of MS disability.

We found microbiota that were associated with changes in EDSS over a 2-year period. Specifically, species from the genera Butyricoccus and Eubacterium hallii were negatively correlated with a 2-year change in EDSS, suggesting a beneficial role. Consistent with other follow-up studies, we found an increase in Bacteroides and a decrease in multiple Lachnospiraceae associated with disease worsening. 35 , 36 These genera are producers of SCFAs including butyrate. 37 , 38 Nonetheless, we did not observe altered SCFA profiles. In subjects who transitioned from relapsing to progressive MS, multiple Lachnospiraceae family members were decreased. These included Blautia species, microbes that play a regulatory role in the gut. 39 Conversely, two Bacteroides genera and one Alistipes were positively associated with a 2-year change in EDSS, suggesting a role in driving disease worsening. Alistipes has been associated with other CNS diseases including Parkinson’s 40 and Alzheimer’s 41 and is associated with an increased risk of colonic cancer 42 , 43 and major depression. 44 Of note, Alistipes obesi was reported to be elevated in MS vs. healthy controls in one study, 12 but not in another. 1 Elevated abundance of a microbe can correlate with disease worsening; however, this does not necessitate a detrimental effect. Transferring MS-derived microbiomes to mice can trigger spontaneous experimental autoimmune encephalomyelitis (EAE), 2 and MS-associated microbes like Akkermansia can worsen EAE. 4 However, we have shown that Akkermansia strains isolated from subjects with MS can ameliorate disease, 12 and studies from other groups have shown improvements in subjects with MS treated with propionate 13 and in EAE mice treated with bile acids. 21

We identified gut microbes that were correlated both with MRI and clinical measures of disease such as cognition and quality of life. Microbes from the genera Erysipelatoclostridium , Romboutsia , and Tyzzerella were positively correlated with a 2-year change in whole brain volume, indicating less atrophy in subjects with higher levels of these microbes. Erysipelatoclostridium has been associated with decreased lesion burden in pediatric subjects with MS. 45
Romboutsia and members of the Peptostreptococcaceae family are reported to be elevated in MS 12 , 46 and negatively correlated with healthy dietary interventions including the Mediterranean diet. 46
Tyzzerella is elevated in pediatric MS, 47 but has not been associated with disease worsening. We also found that a member of the Peptostreptococcaceae family positively correlated with a 2-year change in speed of processing as measured by SDMT, suggesting a beneficial role. Furthermore, members of the Blautia genus were negatively correlated with a 2-year worsening in Neuro-QoL domains, as was the Eubacterium hallii genus. Taken together, our data identify specific microbes with beneficial associations that are decreased in subjects with MS with disease progression.

Clinical progression and microbiota composition in MS may be affected by DMT. 3 , 12 To eliminate the confounder of treatment, we restricted our analyses to subjects who received the same treatment during the 2-year follow-up period. Furthermore, when we examined changes in worsening stratified by DMT, we found no differences in EDSS change, atrophy, or T2 lesion size. We did identify specific microbes that were altered in subjects treated with DMTs compared to untreated subjects. These included an increase in Blautia and a decrease in Bacteroides dorei in anti-CD20-treated subjects.

To identify metabolites associated with MS progression, we performed global metabolomics on subjects with MS who transitioned from relapsing to progressive disease during the 2-year follow-up. Serum metabolites that were altered in stable vs. RR → SP subjects consisted of xenobiotics, lipids, and amino acids. Several serum metabolites were increased in subjects that transitioned to progressive MS, including palmitoleate, a lipid shown to be elevated in subjects with MS with an EDSS > 3, 48 and p-cresol sulfate, a myelin basic protein mimic that has been shown to be elevated only in progressive MS. 49 Secondary bile acids are modified by microbes and have important roles in health and disease. 50 We found that two microbially derived bile acids in the serum, ursodeoxycholate and isoursodeoxycholate, were decreased in subjects who worsened. A decrease of these microbially derived bile acids suggests a potential beneficial role as demonstrated in mouse models. 21 , 51

When we investigated stool metabolites, we found that most were decreased in subjects who progressed, including nicotinate (vitamin B3), nicotinate ribonucleoside, and protoporphyrin IX, which are cofactors/vitamins that have a beneficial role in the EAE mouse model of MS. 52 , 53 The decrease in these beneficial metabolites in subjects who progressed could be explained by the production of nicotinate and other vitamin B metabolites by microbial taxa largely absent in these subjects, including Akkermansia. 54 Altered sphingolipids have been detected in chronic MS lesions 55 and are a target of therapy. 56 We found that multiple sphingolipids were decreased in the stool of subjects who progressed. These included sphingosine and sphingomyelin, which are predicted to be made by multiple Lachnospiraceae including Eubacterium hallii . We identified numerous metabolites associated with stable disease that were predicted to be made by microbes that were downregulated in subjects with SPMS and RR → SP.

Our study provides a comprehensive investigation of the microbiome and metabolome and how they are related to multiple measures of disease progression in MS over time. The use of longitudinal EDSS, brain MRI, and quality of life measures in a tightly controlled population enabled us to identify consistent beneficial associations with Blautia , Eubacterium , and Akkermansia and detrimental associations with Alistipes and Bacteroides across multiple measures of disease progression. In addition, we identified microbial metabolites in the serum and stool associated with the transition from relapsing to progressive MS. Taken together, our results identify an altered microbiota and metabolome in subjects with worsening disease and provide targets for the treatment of MS by modulation of the microbiome.

Limitations of the study

Our study presents important data on the impact of altered microbiota and metabolites on MS disease progression across 2 years; however, there are limitations. We included a relatively small number of study subjects ( Table S1 ), namely due to the demanding nature of the clinical and imaging studies that we performed. We were also limited by the use of 16S sequencing for microbial classification, and future work will benefit from similar longitudinal designs incorporating shotgun metagenomic methodologies. Taken together, our investigations provide testable targets for modifying the microbiome and its metabolites within the context of MS and identify the need for larger scale, longitudinal investigations that harness multi-site cohorts.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Howard L. Weiner ( hweiner@bwh.harvard.edu ).

Materials availability

This study did not generate new unique reagents.

Data and code availability

De-identified raw data for microbiome sequencing are accessible at the NCBI Sequence Read Archive: SUB14526550 and BioProject: PRJNA1123247. Metabolomics data are deposited at the National Metabolomics Data Repository: fecal #4916, serum #4918.

No custom code was generated in this paper. To improve reproducibility of our results, workflows used in this study are available at GitHub: https://github.com/tobylanser/gut_microbiota_metabolites_progMS .

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We would like to thank Dr. Lynn Bry and Mary Delaney at the Massachusetts Host-Microbiome Center for their expert assistance in SCFA processing and analysis. Funding was provided by Postdoctoral Fellowship, NMSS, grant #FG-2207-40162 (to L.A.S.), Nancy Davis Race to Erase MS Young Investigator Award (to L.M.C.), and Water Cove Charitable Foundation, the Clara E. and John H. Ware Jr. Foundation, and NIH grant #1R01NS087226-01A1 (to H.L.W.). The graphical abstract and Figure 1 A were generated in part with BioRender.

Author contributions

L.A.S. conceived the study, performed the experiments, analyzed and interpreted the data, wrote the manuscript, and acquired funding. F.M. analyzed and interpreted the data and edited the manuscript. T.B.L. performed data analysis and curation and edited the manuscript. M.N.E. performed the experiments and edited the manuscript. J.Z., S.T., R.C., B.I.G., and R.B. curated data and edited the manuscript. T.C. curated data and subject biorepository and edited the manuscript. L.M.C. and H.L.W. conceived the study, interpreted the data, edited the manuscript, acquired funding, and oversaw the entirety of the project.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE SOURCE IDENTIFIER Biological samples Human healthy control and MS subject stool This Paper CLIMB Biorepository, Brigham and Women’s Hospital Human healthy control and MS subject serum This Paper CLIMB Biorepository, Brigham and Women’s Hospital Critical commercial assays DNAeasy PowerLyzer Microbiome DNA extraction kits Qiagen, Inc. CAT# 12255-50 Metabolomics on stool and serum Metabolon Inc. N/A SCFA analysis This Paper Massachusetts Host-Microbiome Center Deposited data NCBI Sequence Read Archive This Paper SUB14526550 BioProject This Paper PRJNA1123247 National Metabolomics Data Repository (NMDR): fecal This Paper #4916 National Metabolomics Data Repository (NMDR): serum This Paper #4918 GitHub repository of statistical workflows This Paper https://github.com/tobylanser/gut_microbiota_metabolites_progMS Oligonucleotides Earth Microbiome Project’s barcoded primers 57 Thompson et al. 57 https://earthmicrobiome.org/protocols-and-standards/16s/ Software and algorithms Prism 10 Graphpad version 10.2.2 Qiime2 This Paper version 2023.5 MaAsLin2 package Mallick et al. 27 version 1.15.1 MMvec Morton et al. 34 version 1.0.5 Qiime2 environment This Paper version 2020.6 metOrigin Yu et al. 33 N/A Spearman Correlations R version 4.3.3 (2024-02-29) N/A BioRender BioRender https://biorender.com/n08n589

Experimental model and study participant details

Subjects were originally recruited from the CLIMB study at the Brigham and Women’s Hospital MS center. Diagnosis of RRMS or Progressive MS, willingness to participate, and meeting the below inclusion and exclusion criteria were required for participation. All protocols were approved by the BWH institutional review board (IRB Protocol #: 2017P001169) and informed consent was obtained from every study subject. Inclusion criteria was a diagnosis of MS according to the latest McDonald criteria, with secondary and primary progressive MS being grouped together. Progression independent of relapse activity (PIRA) was not quantified in these subjects. Exclusion criteria were current pregnancy, gastrointestinal surgical history, gastrointestinal disease, or antibiotics within 3 months prior to sample. Healthy control samples were garnered from healthy subjects from the PhenoGenetic Project, and processing procedures for these samples were identical to those from MS participants. After being enrolled, subjects were asked about their sex (male or female) but not gender (e.g., man, woman, nonbinary). Subjects who participated in SDMT, CVLT-II, or Neuro-QoL measurements underwent testing within 6 months of stool and blood donation. Subjects who transitioned from relapsing to progressive disease were classified according to their clinician’s diagnosis based on continued worsening of standard neurological exam measures and increasing EDSS. Subjects did not overlap between those included in Tables S1 and S3 . Dietary information was not collected at the time of sampling.

Method details

MRI methods

People with MS in our CLIMB study undergo annual 3T MRI of the brain using a prescribed high-resolution protocol. We applied automated post-processing pipelines to measure T2LV, 58 and whole brain and DGM volumes, after appropriate normalization. 59 , 60 Volumetric data was collected within +/− 6 months of stool sample donation.

Microbiome processing and analysis

DNA was extracted from ∼200 mg stool using DNAeasy PowerLyzer Microbiome DNA extraction kits (Qiagen, Hilden, Germany). The V4-5 region of the 16S rRNA gene was amplified per the updated Earth Microbiome Project’s barcoded primers. 57 Paired-end sequencing was performed at the Harvard Biopolymers facility on a MiSeq and all samples were sequenced on the same run. Sequence analysis was done using QIIME2. Data was denoised and filtered using DADA2 and paired samples were subsequently merged. The table was filtered to remove samples from analysis that had less than 1,000 reads, appeared in only 1 sample, or were from known contaminate sequences. Taxonomic assignment was performed using a trained SILVA classifier that uses sequences from the November 2021 release of the SILVA rRNA database as a reference set. 61 Statistical testing for α-diversity differences across disease subtype was done using a Kruskal-Wallis test, and β-diversity was analyzed by PERMANOVA with correction for multiple comparison testing. Species with less than 10% prevalence in any disease category were removed from statistical analyses. Viewing differences in composition across multi-variable samples, including longitudinal, age, EDSS, and disease status was performed in R using the Microbiome Multivariate Associations with Linear Models tool (MaAsLin2 27 ). MaAsLin2 analysis was performed while adjusting for the percentage of male and female subjects, age, and BMI. Spearman’s correlations between individual bacterial taxa linked with 3T MRI volumetric measures, cognitive tests, and Neuro-QoL domains were performed in R.

Metabolomics

Global metabolomics were performed by Metabolon Inc, (Durham, NC, USA) as previously described. 62 Serum or stool samples, collected within +/− 45 days of each other and previously frozen at −80°C were processed with an 80% methanol extraction and metabolites detected via ultra-high performance liquid chromatography-tandem mass spectrometry. Raw data was processed by Metabolon Inc, as described previously. 63 Briefly, data was extracted from the mass spectrometer, quality-controlled, and aligned across individuals. Compounds were identified using an internal Metabolon library containing more than 5,400 purified standards. Samples were compared for retention times and mass/charge ratios to the purified standards. Retention time indices and mass spectral profiles were quality controlled to filter biochemicals with ranges outside of Metabolon’s database standards of +/− 10 parts/million of a library entry. For each identified metabolite, ion current area under the curve was used to generate raw abundances which were then log2-transformed and median-scaled by dividing the metabolite’s raw abundance by the median raw abundance of that metabolite across the entire dataset. Imputation for missing metabolites per sample was performed using the minimum median scaled abundance. For each metabolite, fold differences were determined between groups by dividing a metabolites average median scaled abundance in one group by another. Groupings contrasts were performed on diagnostic group (HC, RR → RR, RR → SP, or SP). Pathway enrichment scores (PES) were calculated to assess the overall contribution of a pathway to metabolite variance between diagnostic groups. PES were calculated using the following equation:

Where “k” represents the number of metabolites reaching statistical significance in a given pathway, “m” is the total number of metabolites in said pathway, “n” is the total count of significant metabolites in all pathways, and “N” is all detected metabolites. A PES >1 was considered a contributor to group differences. GraphPad Prism Version 10.1.1 (San Diego, CA, USA) was used to visualize PES. Metabolite origins were ascertained using metOrigin 33 and identified sources were plotted using Venn-diagrams. Microbial correlations with identified metabolites was performed using MMvec qiime2 plugin as previously described. 34

SCFA quantification

Stool samples were placed into a sterile Eppendorf tube prior to analysis of SCFAs via gas chromatography-mass spectrometry at the Massachusetts Host-Microbiome Center. Stool was weighed and between 0.1 and 0.4 g of stool was diluted in 1mL HPLC grade water prior to analysis for the following volatile acids: acetate, propionate, isobutyrate, butyrate, isovalerate, valerate, isocaproate, caproate, and heptanoate. Detectable levels were observed in a majority of samples for 6 SCFAs: acetate, propionate, butyrate, valerate, isobutyrate, and isovalerate. Values were normalized to the original input mass of stool sample prior to analysis.

Bioinformatics tools

Microbiome analyses were performed using Qiime2 (version 2023.5). R scripts were done in R (version 4.3.3) using R Studio (version 2023.12.1 + 402). MaAsLin2 was performed in R using a publicly available MaAsLin2 package 27 (version 1.15.1) and controlled for covariates as described in pertinent experimental results. To control for situations where taxa were completely absent in one group but present in another, MaAsLin2 was adjusted to include a relative abundance pseudo-count of 0.0058, derived from the mean relative abundance when all zeroes were removed from the table. Spearman’s correlations were performed in R and Spearman’s rho and unadjusted p value are reported, except for CVLT-II analysis which were adjusted for age. MMvec 34 (version 1.0.5) analysis on microbiota – metabolomic interactions was performed using a command line script in a Qiime2 environment (version 2020.6) and log conditional probabilities are reported.

Quantification and statistical analysis

Statistical analyses, except for specific bioinformatic tools, were performed using Prism 10 (Graphpad, v 10.2.2). All data where applicable are presented as mean ± standard error of the mean (SEM) and n -values represent number of subjects. Demographic comparisons (e.g., Figures 1 B and 1C) were done via Student’s t test without a Welch’s correction. Change in EDSS ( Figure 1 A) was analyzed via two-way ANOVA by time-point and disease category with a Sidak’s multiple comparison’s post-hoc test. Metabolomic analyses of variance were performed using a one-way ANCOVA to account for covariance factors in our dataset including age and BMI. For all metabolomic heatmaps, Log 2 (fold change) is reported with significance denoted by ANCOVA p -values when the false discovery statistic was q < 0.10. Metabolomic bar charts are expressions of batch normalized scaled intensity values for individual biochemicals, and significance annotations are based on ANCOVA results.

Published: April 3, 2025

Footnotes

Supplemental information can be found online at https://doi.org/10.1016/j.xcrm.2025.102055 .

Contributor Information

Laura M. Cox, Email: lcox@bwh.harvard.edu.

Howard L. Weiner, Email: hweiner@bwh.harvard.edu.

Supplemental information

Document S1. Figures S1–S7 and Tables S1–S3

Document S2. Article plus supplemental information

References

Associated Data

Supplementary Materials

Document S1. Figures S1–S7 and Tables S1–S3

Document S2. Article plus supplemental information

Data Availability Statement

De-identified raw data for microbiome sequencing are accessible at the NCBI Sequence Read Archive: SUB14526550 and BioProject: PRJNA1123247. Metabolomics data are deposited at the National Metabolomics Data Repository: fecal #4916, serum #4918.

No custom code was generated in this paper. To improve reproducibility of our results, workflows used in this study are available at GitHub: https://github.com/tobylanser/gut_microbiota_metabolites_progMS .

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.