Identification of commensal gut microbiota signatures as predictors of clinical severity and disease progression in multiple sclerosis

Abstract

Multiple sclerosis (MS) is a chronic autoimmune disease of the central nervous system and a leading cause of neurological disability in young adults. Clinical presentation and disease course are highly heterogeneous. Typically, disease progression occurs over time and is characterized by the gradual accumulation of disability. The risk of developing MS is driven by complex interactions between genetic and environmental factors, including the gut microbiome. How the commensal gut microbiota impacts disease severity and progression over time remains unknown. In a longitudinal study, disability status and associated clinical features in 58 MS patients were tracked over 4.2 ± 0.98 years, and the baseline fecal gut microbiome was characterized via 16S amplicon sequencing. Progressor status, defined as patients with an increase in Expanded Disability Status Scale (EDSS), were correlated with features of the gut microbiome to determine candidate microbiota associated with risk of MS disease progression. We found no overt differences in microbial community diversity and overall structure between MS patients exhibiting disease progression and non-progressors. However, a total of 41 bacterial species were associated with worsening disease, including a marked depletion in Akkermansia, Lachnospiraceae, and Oscillospiraceae, with an expansion of Alloprevotella, Prevotella-9, and Rhodospirillales. Analysis of the metabolic potential of the inferred metagenome from taxa associated with progression revealed enrichment in oxidative stress-inducing aerobic respiration at the expense of microbial vitamin K₂ production (linked to Akkermansia), and a depletion in SCFA metabolism (linked to Oscillospiraceae). Further, as a proof of principle, statistical modeling demonstrated that microbiota composition and clinical features were sufficient to predict disease progression. Additionally, we found that constipation, a frequent gastrointestinal comorbidity among MS patients, exhibited a divergent microbial signature compared with progressor status. These results demonstrate a proof of principle for the utility of the gut microbiome for predicting disease progression in MS in a small well-defined cohort. Further, analysis of the inferred metagenome suggested that oxidative stress, vitamin K₂, and SCFAs are associated with progression, warranting future functional validation and mechanistic study.

Keywords: Akkermansia; Bacterial metabolism; Gut microbiota; Multiple sclerosis; Progression; Vitamin K.

Figure 1

MS disease progression is not associated with differences in baseline gut microbiome alpha and beta diversity. (a) Schematic of clinical study design depicting participant binning strategy based on longitudinal disease progression where a non-progressor = change in EDSS of ≤0.5 and a progressor = change in >0.5. Bar graphs, line diagrams, and scatter plots of selected study metadata as outlined in Table 1, segregated by participant progressor status, including: (b) progressor severity (change in EDSS over study period), (c) non-progressor and (d) progressor initial and final EDSS, (e) MS-subtype, (f) sex (g) self-identified race, (h) age, (i) BMI, (j) collection DMT status (k) DMT prior history. p-values represent a two-tailed binomial test of significant deviation from expected outcome of a 50% division of between progressors and non-progressors for categorical data, or standard t-test for numerical data, with significance at p ≤ 0.05. Complete statistical analysis of subject metadata is included in Tables 1 and 2. (l) Alpha (Fisher, Shannon and Simpson) and (m) beta (Bray–Curtis dissimilarity, Unifrac, wUnifrac) diversity analysis in progressors and non-progressors, as analyzed using Wilcoxon rank sum non-parametric or Adonis testing, respectively. Stacked bar plots depicting microbiota composition at bacterial phylum (n) and top 10 most abundant genera (o), as relative proportion of total 16S V4 amplicon reads within each taxonomic rank.

Figure 2

MS disease progressors exhibit a unique baseline gut microbial signature. Differentially abundant taxa between subjects with or without disease progression by (a) phylum, (b) class, (c) order, (d) family, (e) genus, and (f) top 30 ASVs represented as taxonomic best-hit, as determined by DESeq2 analysis, using a cutoff of p_adj ≤ 0.05. Log2 fold-change reflects increased abundance in progressors when positive, and decreased abundance when negative. Heatmaps and bar graphs of prevalence as well as center log ratio (CLR) transformed abundance graphs are aligned to the right. (e) Taxonomic association heat tree of differentially abundant microbiota determined by total sum scaling (TSS) log2 linear regression through the level of genus. Shown on the right is a phylogenetic key for the heat tree shown on the left. Data was filtered at a minimum prevalence of 0.1, significance determined at a cutoff of p_adj ≤ 0.05. Node size indicates proportional prevalence, significant nodes are represented as open circles (or red font in the key), warmer colors are enriched in progressors, with cooler colors indicating depletion.

Figure 3

Abundance and prevalence of progressor-associated microbiota. Association of (a) genera and (b) ASVs with progressor status and closely related subject metadata, as determined by Spearman rank correlation ≥|0.2|, at p_adj ≤ 0.05. Taxa are sorted from high to low rho-value within each metadata group top to bottom on y-axes, where warmer colors are indicative of positive association (increased abundance) and cooler colors represent negative association (decreased abundance). Metadata binning strategy is listed in Table 1. CLR transformed abundance (c–j) and % prevalence at each taxonomic rank (k–r) are shown for each genera associated with progressor status. Kaplan–Meier curves predicting probability of disease progression over subject disease duration (yrs) using cohorts stratified based on abundance of (s) Sutterella (low/high), (t) Alloprevotella (low/high), (u) Rhodospirillales (low/high), (v) Bilophila (low/higher) and (w) Eubacterium ventriosum (low/high). Significance represents log-rank p-value ≥ 0.05 for differential probability of disease progression survival between microbiota driven strata.

Figure 4

Disease progressors exhibit distinct functional gut microbial metabolic potential. (a) Schematic of the computational approach used to infer the functional potential of differentially abundant microbiota associated with disease progression. (b) Volcano plot of ASVs (represented as taxonomic best-hits) associated with disease progression as determined by Spearman rank correlation ≥|0.2|, at p_adj ≤ 0.05. (c) Differentially abundant enzymes and (d) pathways in the inferred metagenome of progressor status associated ASVs. Total metagenomic potential was inferred with PICRUSt2 and differential abundance analyzed for the subset of ASVs associated with progressor status using DESeq2 at p_adj ≤ 0.05. (e) Top 50 differentially abundant enzymes plotted as gene count by ASV (taxonomic best-hit) colored by phylum, as determined by log2 fold-change at p_adj ≤ 0.05. Ubiquinol (f), menaquinol (g), and propanediol degradation (h) pathway schematics mapped using MetaCyc and annotated with differentially abundant enzymes and their originating ASVs.

Figure 5

Microbiota composition and patient clinical features are sufficient to predict patient progressor status. (a) X–Y scatter plot reflecting total variables of importance by mean decrease in accuracy and mean decrease in Gini coefficient as determined by a Random Forest classifier trained on total ASV level 16S abundance data, with ROC curve analysis shown in (b). (c) Top 10 clinical features as variables of importance by mean decrease in Gini coefficient as determined by a Random Forest classifier trained on patient metadata available at study baseline, with associated ROC curve shown in (d) for clinical data alone and (e) when combined with total ASV 16S data. Top 10 variables of importance (f), ROC curve (g) and class error (h) are shown for a Random Forest classifier trained on only the ASVs correlated with progressor status as in Fig. 4B. Inclusion of patient baseline clinical metadata in classifier training of progressor associated microbiota is shown in (i). All Random Forest classifiers were bootstrapped 1000 times using a balanced bagging approach with leave-one-out cross validation.

Figure 6

Constipation is associated with a unique microbial signature that is divergent from that of disease progression. (a) Metadata bubble plot correlation matrix using a Spearman rank correlation with a significance cutoff at p_adj ≤ 0.05, with significance of association denoted by increasing size. (b) PERMANOVA of Bray–Curtis dissimilarity with the percent of variance explained by each metadata feature as R² on x-axis and colored by FDR, as determined by the adonis test. (c) Alpha (Shannon) and (d) beta (Bray–Curtis dissimilarity) diversity analysis in subjects not reporting or reporting constipation, as analyzed using Wilcoxon rank sum non-parametric or adonis tests, respectively. Differentially abundant taxa by (e) genus and (f) ASVs represented as taxonomic best-hit between subjects with or without constipation, as determined using DEseq2 using a cutoff of p_adj ≤ 0.05. Log2 fold-change reflects increased abundance in subjects experiencing constipation when positive and decreased abundance when negative. (g) Venn diagrams of shared or divergent ASVs between disease progressors and subjects experiencing constipation as determined by DESeq2 (p_adj ≤ 0.05) and Spearman rank correlation ≥|0.2|, at p_adj ≤ 0.05. ASVs associated with both progressor status and constipation are plotted as CLR transformed abundance for shared ASVs in (h). (i) Differentially abundant enzymes and (j) pathways from the inferred metagenome of constipation associated ASVs. Total metagenomic potential was inferred with PICRUSt2 and differential abundance analyzed for the subset of ASVs associated with constipation using DESeq2 at p_adj ≤ 0.05.

Figure 7

Summary of gut microbiota related mechanisms associated with MS disease progression. Our data and analysis revealed a significant enrichment in oxidative stress-inducing aerobic respiration at the expense of microbial vitamin K2 production (linked to Akkermansia), and a depletion in SCFA metabolism (linked to Lachnospiraceae and Oscillospiraceae) associated with MS progression.