All experimental procedures used in this study were approved by the University of Vermont’s Animal Care and use Committee. Mice were maintained under barrier conditions with sterilized caging, fed irradiated diets (Prolab IsoPro RMH 3000), and handled minimally in a structured sequence to avoid cross-contamination and the introduction of new microbes. Donor male C57BL/6J (B6) and PWD/PhJ (PWD) mice used for cecal microbiota transplant were purchased from Jackson Laboratories (Bar Harbor, Maine, USA) and housed within a single vivarium room at the Larner College of Medicine at the University of Vermont for 2–4 generations. Gut microbial transplantation from cecal donors was performed as previously described (75); ceca from donor B6 and PWD mice were collected and transferred to an anaerobic chamber where the contents were flushed out and mixed to a final concentration of 20% glycerol in Hungate tubes, flash frozen, and stored at -80 °C in single use aliquots. PWDβ cecal stocks were generated by collecting and combining cecal contents of ex-germ free B6 mouse recipients of the PWD microbiome, and pups of ex-germ free B6 mouse recipients of the PWD microbiome, performed as above.

Germ-free (GF) 7–9 week-old B6 mice from the National Gnotobiotic Rodent Resource Center at the University of North Carolinae School of Medicine (USA) were shipped in sterile crates. GF B6 mice were opened under a laminar flow hood and immediately inoculated by gastric gavage with 100-200 µl of cryopreserved stocks from B6 or PWD ceca, generating B6.Gut^B6 mice (denoted as “B6”) and B6.Gut^PWD mice (denoted as “PWD”), respectively. Colonization by A. muciniphila was achieved in B6 and PWD microbiome-colonized mice with a series of 3 gastric gavages of 1.65×10⁷ CFU/200 µl A. muciniphila every other day, generating B6.Gut^B6+Akk mice (denoted as “B6+Akk”) and B6.Gut^PWD+Akk mice (denoted as “PWD+Akk”). B6.Gut^PWDβ mice (“PWDβ”) were generated using a single 200 µl gastric gavage containing 100 µl of cryopreserved stocks from PWDβ cecal stock and 100 µl of anerobic 50% glycerol in PBS. B6.Gut^PWDβ+Akk counterparts (denoted as “PWDβ+Akk”) were generated using a single 200 µl gastric gavage containing 100 µl of cryopreserved PWDβ cecal stock and 8.26x10⁶ CFU/100 µl of A. muciniphila. The resulting B6, B6+Akk, PWD, PWD+Akk, PWDβ, and PWDβ+Akk mice were paired for breeding to generate male and female pups with vertically transmitted gut microbiota to be used for subsequent experiments.

For dietary fiber intervention experiments, mice were randomized by microbiome to receive low fiber (TD.180916, 0% fermentable fiber) or high fiber (TD.220544, 20% inulin, 10% pectin) for 2 weeks prior to EAE induction. Low and high fiber chow (Inotiv, USA) were vacuum packed, irradiated, and stored at 4°C. Chow was refreshed daily.

EAE was induced in 6-11-week-old pups of specific microbiota-colonized ex-germ-free (ex-GF) breeders using the 2× MOG_35-55/CFA protocol as previously described (76); mice were injected subcutaneously with 100 µl of emulsion containing 100 µg myelin oligodendrocyte glycoprotein 35-55 (MOG_35-55; New England Peptide, USA) in 50% complete Freund adjuvant (CFA; Sigma, USA) supplemented with an additional 4 mg/ml Mycobacterium tuberculosis H37Ra (Difco, USA). A first set of injections was completed on each lower flank (50 µl/side), and a 2^nd set of injections was completed 7 days later into each upper flank (50 µl/side) of the mouse. On days 10 through 30, mice were scored for ascending paralysis as follows: 0 – asymptomatic, 1 – loss of tail tone, 2 – loss of tail tone and hind limb weakness, 3 – hind limb paralysis, 4 – hind limb paralysis with incontinence, and 5 – moribund/quadriplegic. Cumulative disease score was calculated as the sum of all daily scores. Cages that included at least one mouse exhibiting hind limb paralysis received 2–3 chow pellets and approximately 2cm³ of Napa nectar (Systems Engineering, USA) on the cage floor, both refreshed daily.

Fecal samples were collected by transferring individual mice to empty cages without bedding and waiting for them to defecate at least one fecal pellet. Collection cages were only used within-microbiome group and replaced for each experiment. Once collected, pellets were kept on ice before being stored at -80 °C until later use. DNA was extracted from fecal pellets using QIAamp PowerFecal Pro DNA extraction kits (Qiagen, USA); DNA quality and quantity was assessed via Nanodrop. A. muciniphila relative abundance was quantified using species-specific primers (forward sequence, CAGCACGTGAAGGTGGGGAC; reverse sequence, CCTTGCGGTTGGCTTCAGAT) normalized to a pan-bacteria eubacteria primer set (forward sequence, ACTCCTACGGGAGGCAGCAG; reverse sequence, ATTACCGCGGCTGCTGG). Quantification by qPCR was achieved with Dynamo ColorFlash SYBR Green (Thermo Fisher Scientific, USA) on a Quant Studio 3 or 5 Real-Time PCR machine (Thermo Fisher Scientific, USA) with annealing temperatures of 66 °C for A. muciniphila and 60 °C for eubacterial primers and 30 PCR cycles. Any B6+Akk, PWD+Akk, and PWDβ+Akk mice whose fecal samples did not show amplification for A. muciniphila prior to EAE induction were assumed to be unsuccessfully colonized and thus excluded from the analysis. A. muciniphila-free B6, PWD, and PWDβ feces were also tested for A. muciniphila, which was never observed in these samples. Melting curves were also used to confirm the presence of a single peak at 87 °C for detecting A. muciniphila.

Fecal slurries were prepared from frozen fecal samples. Fecal pellets were weighed and transferred to a screw-top 2 mL tube filled approximately 1/5 full of 1.0 mm diameter silicon carbide sharp particles (BioSpec, USA). Cold phosphate-buffered saline (PBS) with 0.01% Tween 20 was added to achieve 50 mg feces/mL. Tubes were homogenized by vortexing with a tube adapter for 10 minutes and pelleted at 15,000 × g for 10 minutes. The supernatant was transferred to a clean tube and stored for use. Lipocalin-2 was quantified using ELISA reagent kits per manufacturers procedure (R&D Systems, USA). Optical density was measured at 450 nm with background subtraction at 570 nm wavelength and concentrations were calculated using a standard curve. PBS with 0.01% Tween 20 was used for washing, and PBS with 1% bovine serum albumin (Sigma, USA) was used for blocking and reagent dilutions.

Fecal DNA, extracted as described above, was utilized for full length 16S ribosomal RNA gene amplicon sequencing at the University of Illinois W.M. Keck Center for Comparative and Functional Genomics. 16S amplicons were generated with barcoded primers from PacBio targeting the entire 16S gene (forward sequence, AGRGTTYGATYMTGGCTCAG; reverse sequence, RGYTACCTTGTTACGACTT) and Roche KAPA HiFi Hot Start Ready Mix. The amplicons were converted into a sequencing library with a SMRTbell Express Template Prep Kit 3.0, and the library was then sequenced on a single SMRTcell 8M using a PacBio Sequel IIe platform in Circular Consensus Sequencing (CCS) mode and a 15-hour movie run time. CCS libraries were analyzed for read quality and demultiplexed on SMRTLink version 11.1, creating raw sequencing reads as FASTQ files for each fecal sample.

The raw FASTQ sequencing files were analyzed in R Studio (R version 4.2.1) with preprocessing using the DADA2 package (version 1.26.0) (77). Following primer removal, reads were filtered to include only those with a nucleotide length between 1000 and 1600, a minimum quality score of 3, a maximum number of expected errors (maxEE) of 2. Dereplicated data was used to generate and apply an error model, followed by denoising and then removal of chimeric sequences by the consensus method, generating amplicon sequencing variants (ASVs). ASV sequences were reference against SILVA database (version 138.1) for taxonomic assignment (78, 79). Construction of a phylogenetic tree was done using the DECIPHER package (version 2.26.0) for alignment of unaligned sequences (80) and the phangorn package (version 2.11.1) for building the tree using the neighbor-joining method with pairwise computed distances, a general time-reservable model, the nearest neighbor interchange, and rooted using a randomly selected ASV (81). The resulting phylogenetic tree, ASV table, ASV sequences, taxonomic assignments, and imported metadata were combined in phyloseq (version 1.48.0) to create a phyloseq object for downstream analyses (82).

Alpha diversity, determined by Shannon index, was graphed and calculated using the phyloseq functions plot_richness and estimate_richness and compared by Wilcox test with Holm-Bonferroni correction for multiple comparisons. For subsequent analyses, ASVs within the phyloseq object were first agglomerated using the tip_glom function to a height threshold of 0.05; using the metagMisc package (version 0.0.4), ASVs were then filtered to only include those with a prevalence of at least 3% and a mean abundance threshold of 2. Graphical representations of 16S data, including the relative abundance of bacterial taxa across samples, were generated using the constructed phyloseq object and ggplot2 (version 3.5.1). Beta diversity was calculated using the ordinate and plot_ordination functions of phyloseq, represented using PCoA plots, and compared across microbiome groups by permutation multivariate ANOVA using the adonis function from the vegan package (version 2.6-4) (83) and the pairwise adonis function from the pairwiseAdonis package (version 0.4.1). Differential abundance was calculated with the DESeq2 package (version 1.38.3) using Wald significance testing and an adjusted p <0.05 (70). Composition networks of microbiome structures were created using the NetCoMi package (version 1.1.0) using the Semi-Parametric Rank-based approach for INference in Graphical (SPRING) model that included only the 50 most abundant ASVs and with a sparsity parameter (lambda) of 20 and filtering to the 50 most frequent reads (84).

Inferred functional metagenomic pathways of the entire gut microbiota of each sample was mapped using PICRUSt2 (version 2.2.0_b) in Python (version 3.6.7) (85). The abundance of inferred pathways was then applied to a new phyloseq object as a wrapper and analyzed using DESeq2 as done with ASVs described above. ASV contributions to individual enzyme commission (EC) numbers were generated in PICRUSt2 to analyze individual ASV contributions to butyrate production as previously described (86). Butyrate producers were defined as bacteria conserving any of the following terminal enzymes in butyrate-producing pathways: butyryl-CoA:4-hydroxybutyrate CoA transferase (4Hbt; EC:2.8.3.-), butyryl-CoA:acetoacetate CoA transferase (Ato; EC:2.8.3.9), butyryl-CoA: acetate CoA transferase (But; EC:2.8.3.8) or butyrate kinase (Buk; EC:2.7.2.7) (87). Statistical differences in relative abundance of all butyrate producer ASVs between comparisons was determined by Wilcoxon rank sum non-parametric testing.

A. muciniphila murine strain YL44/DSM26127 was provided by Dr. Adam Sateriale (currently at the Francis Crick Institute, UK) and was originally obtained from DSMZ (Leibniz Institute, Germany). A. muciniphila culture used for gastric gavages was grown anaerobically in brain heart infusion broth (BHI; Sigma, USA) supplemented with 15% of fetal bovine serum (Thermo Fisher Scientific, USA), 5 g/L yeast extract (Sigma, USA), 0.2 mL/L vitamin K solution (Sigma, USA; made as 1% vitamin K1 in 100% ethanol), 0.5 mL/L of hemin solution (Sigma, USA; 0.5g/L dissolved in 1% NaOH solution), 0.5 g/L cysteine (Sigma, USA). Cultures were grown in 5 mL volumes of liquid BHI with 100 µL of 5% type 3 porcine stomach mucin (Sigma, USA; 50 g/L mucin mixed in deionized water). A. muciniphila was grown without shaking to an OD₆₀₀ of 0.833, used as high-density stocks and for breeding pair mice (B6+Akk, PWD+Akk, and PWDβ+Akk) and an OD₆₀₀ of 0.215, then diluted 1:10 using bacteria-free media, for low-density stocks. CFU was calculated using an OD₆₀₀ 0.8 equivalent of 1x10⁸ CFU/mL of A. muciniphila after subtracting the OD₆₀₀ of the growth media (88). Anaerobic conditions were maintained in an anerobic chamber (Coy Labs, USA) with 5% carbon dioxide, 2% hydrogen, and 91% nitrogen at 37 °C. Individual cultures were pooled prior to being cryopreserved for future use as gavage stocks, which were made as 50% liquid culture, 25% sterile glycerol (Sigma, USA), and 25% sterile PBS. For treatment of control group SPF B6 mice ( Figure 1A ), a vehicle containing 50% media, 25% sterile glycerol, and 25% sterile PBS was used.

Post-EAE mice were anaesthetized with isoflurane (Piramal, USA) and perfused transcardially with 40 mL of cold PBS. Each spinal cord was dissected, homogenized using a Dounce glass homogenizer, and filtered through a 70 µm strainer. The resulting single-cell suspension was mixed with 37% Percoll (Cytiva, USA) and loaded above 70% Percoll to achieve a Percoll gradient that isolated mononuclear cells from the interphase following centrifugation. Isolated cells were stimulated with 5 ng/mL PMA, 250 ng/mL ionomycin, and brefeldin A (GolgiPlug; BD Bioscience, USA) for 4 hours. Cells were then labeled with UV-Blue Live/Dead (Thermo Fisher Scientific, USA), stained with surface antibodies: CD45, CD11b, CD19, CD4, CD8, TCRβ, TCRγδ (BioLegend, USA), fixed and permeabilized with 0.2% saponin (Sigma, USA), then labeled with intracellular antibodies: IL-17A and IFNγ (BioLegend, USA). Stained cells were analyzed in the Harry Hood Bassett Flow Cytometry and Small Particles Detection facility (RRID: SCR_022147) at the Larner College of Medicine using a Cytek Aurora and SpectroFlo software (versions 2.2-3.3; Cytek Biosciences, USA) using spectral unmixing with single-color controls and autofluorescence correction from unstained cells. Data was analyzed using FlowJo software (version 10.10.0) (BD Biosciences).

Gas chromatography coupled with mass spectrometry (GC-MS) was used to quantify SCFA levels in fecal samples, including acetate, propionate, butyrate, and isobutyrate. Fecal samples were directly weighed into a 2 mL microcentrifuge tube, and 700 mg silica beads and 600 µL distilled and deionized water were added to each. Samples were homogenized using a bead beater. After homogenization, samples were centrifuged at 10,000 rpm for 10 minutes and the liquified sample in the supernatant was extracted. Solvent extraction of SCFAs was performed by combining 200 µL of supernatant, 20 µL of hydrochloric acid (HCl), 100 µL of potassium bisulfate (KHSO4), and 1 mL of dimethyl carbonate (DMC) into a 2 mL microcentrifuge tube (89). Upon addition of HCl, the anion forms of SCFAs are protonated such that they readily partition into the DMC. The resulting mixture was vortexed for 10 seconds and centrifuged at 3800 rpm for 10 minutes. The supernatant organic phase was then transferred into a GC-MS vial for analysis. For the quantification, a Shimadzu Nexus GS 2030 couple to a TQ8040NX mass spectrometer was used. The GC-MS autosampler used a 2 μL smart syringe to inject 0.8 μL of liquid sample. A DB-FATWAX UI column, with 30 m length, 0.25 μm thickness, and 0.25 mm diameter was used. Upon injection of the sample, the oven temperature was held at 80 °C for 1 minute and then increased by 15 °C/minute until it reached 115 °C. The oven temperature was held at 115 °C for 3 minutes. Then, the oven temperature ramped again at a rate of 3 °C/minute until it reached 130 °C. The temperature was then increased at a rate of 15 °C/minute until it reached 230 °C. The oven was held at 230 °C for 3 min. Acquisition mode was Q3 scan, with ion source temperature and interface temperature at 280 °C and 250 °C, respectively. Concentrations were determined using a calibration curve generated from external standards that underwent the same extraction protocol as samples. Concentrations were then used to calculate mg of each SCFA per gram of fecal matter.

To assess a potential causative role for A. muciniphila in modulating EAE severity, we sought to modulate its abundance by colonization via oral gavage, first using commercially available SPF C57BL/6J (B6) mice from the Jackson Laboratory (Jax). B6 Jax mice were assigned to gavage treatment groups and inoculated by a single oral gavage with 200µl of A. muciniphila culture from either low (1.07×10⁶ CFU) or high-density (1.65×10⁸ CFU) anaerobically grown stocks, and a 3^rd treatment group received two gavages of high-density culture ( Figure 1A ), based on previous studies suggesting that 1× and 2× oral gavage of A. muciniphila is sufficient to modulate its abundance in SPF mice (90, 91). Sterile vehicle control was administered as a 2^nd gavage for single-gavage-treated groups and control groups. EAE was induced at 7 or 14 days post-treatment (see Figure 1A ) by immunization with MOG_35–55 in CFA, as previously described.

Analysis of A. muciniphila treatment effect on EAE disease course revealed no significant differences across gavage groups or any dose-dependent trends ( Figure 1B ). Similarly, analysis of cumulative disease score demonstrated no significant differences between control and any of the A. muciniphila-treated groups ( Figure 1C ). We next assessed the baseline and post-treatment levels of A. muciniphila, using species-specific qPCR on fecal DNA. Unexpectedly we found that all controls and treatment groups already carried high levels of A. muciniphila at baseline ( Figure 1D ), and oral gavage of A. muciniphila failed to modulate the relative abundance of A. muciniphila among the treatment groups ( Figure 1D ). Kinetic analysis of A. muciniphila abundance found a significant overall decline (p < 0.0001) in relative abundance observed between arrival day and all subsequent fecal collection timepoints, with post-hoc testing demonstrating significant declines over time across all treatment groups ( Figures 1D, E ). Given the relatively wide variation in A. muciniphila abundance among individual mice ( Figure 1D ), we tested for its association with EAE severity. We found no significant association between EAE cumulative disease score and the relative abundance of A. muciniphila at any collection timepoint ( Figure 1E ; Supplementary Figures 1A–E ). Taken together, these results demonstrate that commercially available SPF B6 Jax mice carry high levels of endogenous A. muciniphila, which may limit the ability to experimentally manipulate abundance of A. muciniphila by oral gavage and hence assess its effect on EAE outcomes in this model.

Given that direct administration of A. muciniphila to commercially-available mice with an established gut microbiome replete with endogenous A. muciniphila failed to appreciably elevate its abundance levels or impact CNS autoimmunity, we pivoted to utilize our previously established compositionally defined microbiome transplantation and vertical transmission model (75, 92). This model capitalizes on cryopreserved cecal microbiota harvested from a set of B6 and genetically divergent Prague wild-derived D (PWD) mice in our own vivarium, allowing the transplantation of these two distinct microbiota contexts into genetically identical germ-free B6 hosts. Using our previously published 16S rRNA DNA sequencing data (75), we first assessed the abundance of endogenous A. muciniphila among a large number of B6 and PWD mice in our colony at the time that cryopreserved cecal microbiota stocks were established. We found that A. muciniphila (the only representative of the phylum Verrucomicrobiota in these mice) was present in 34% of B6 mice and, surprisingly, completely absent in PWD ( Figure 2A ). We next assessed the abundance of A. muciniphila in the cryopreserved B6 and PWD microbiomes, which had been established from a small subset of donor mice from our colony. Serendipitously, both B6 and PWD donor microbiota stocks lacked A. muciniphila, which was also recapitulated in the colonized ex-germ-free cecal microbiota transplant recipients ( Figure 2B ), thus providing two divergent A. muciniphila-free microbiomes/ecological contexts for our studies.

As previously published (75), we colonized new germ-free B6 recipient mice via oral gavage with stocks of cecal B6 or PWD microbiota, establishing two distinct stably colonized microbiomes, referred to here simply as B6 and PWD, that recapitulated the microbiomes of their respective donor mice and retained the strong divergence in composition between the original donor B6 and PWD microbiomes ( Figure 2C ). Pilot experiments demonstrated that while by 1× gavage or enema of A. muciniphila culture was not sufficient to stably colonize A. muciniphila-free mice, a 3× gavage regimen was successful ( Supplementary Figures 2A, B ). Thus, we inoculated a subset of B6 and PWD microbiome-colonized ex-germ-free (ex-GF) mice with high-density (1.65×10⁸ CFU) stocks of A. muciniphila via oral gavage, successfully colonizing mice following a series of three 200 µl gavages, establishing B6+Akk and PWD+Akk microbiome-colonized ex-GF mice ( Figure 2D ). B6, PWD, B6+Akk, and PWD+Akk microbiome-colonized ex-GF mice were set up as breeding pairs to allow for vertical transmission of their distinct microbiota to their offspring, which were used as experimental animals, thus circumventing potential effects of underdeveloped gut immune-microbe interfaces in ex-germ-free animals (93, 94). Fecal samples from gavage recipient breeders and their offspring were tested for A. muciniphila abundance over time, demonstrating stable colonization and vertical transmission of A. muciniphila ( Figures 2E, F ). Thus, we were able to generate two distinct microbiome contexts in which we could isolate the effects of stable A. muciniphila colonization, while avoiding the confounding effects caused by continuous gavage stress (95). Interestingly, PWD+Akk pups maintained a significantly greater relative abundance of A. muciniphila compared to B6+Akk pups, allowing us to leverage microbiome contexts that have both distinct ecological contexts and differing but stable levels of A. muciniphila ( Figures 2C, F ).

Using our B6, B6+Akk, PWD, and PWD+Akk microbiome-colonized mice, we induced EAE using MOG_35-55/CFA immunization, as previously described (76). No significant difference was observed between B6+Akk and B6 microbiome mice ( Figures 3A, B ). In contrast, A. muciniphila exacerbated EAE severity in PWD+Akk microbiome mice compared to their A. muciniphila-free counterparts ( Figures 3C, D ). These data indicate that the effect of A. muciniphila on EAE severity is highly dependent on the overall composition of the microbiome into which it is introduced.

To assess whether A. muciniphila-mediated EAE exacerbation was associated with augmented CNS-directed immune responses, we isolated immune cells infiltrating into the spinal cord of PWD and PWD+Akk microbiome mice at day 30 post EAE induction for immune phenotyping ( Figure 3E ). Flow cytometry revealed differences in immune profiles in A. muciniphila-colonized mice, including a lower frequency of CD11b⁺ myeloid cells and a greater frequency of CD11b^-CD19^- cells, representing additional CNS-infiltrating immune cells, including T cells ( Figure 3F ). A. muciniphila-colonized mice also showed a greater frequency of IL-17-producing CD4⁺ T cells, a lower frequency of IFNγ-producing TCRγδ cells, and a trend toward a lower frequency of IFNγ-producing CD4⁺ T cells ( Figure 3G ), which was further substantiated by a significant increase in the ratio of IL-17:IFNγ-producing CD4⁺ T cells ( Figure 3H ). Taken together, these results demonstrate A. muciniphila colonization in the PWD microbiome mice leads to a shift towards a pro-inflammatory Th17 phenotype relevant to autoimmune responses and EAE severity.

While PWD mice are not naturally colonized with endogenous A. muciniphila ( Figures 2A, B ), our PWD+Akk microbiome mice harbor a high level of A. muciniphila (compared to our B6+Akk microbiome mice) ( Figures 2D, E ). Considering that elevated abundance in A. muciniphila has been shown in some contexts to induce intestinal inflammation via the degradation of mucin (22, 34), we assessed whether the high abundance of A. muciniphila in PWD+Akk microbiome mice was modulating intestinal inflammation prior to and during EAE progression, which by itself has been shown to trigger intestinal inflammation or permeability (96–98). We used fecal lipocalin-2 levels as a sensitive surrogate marker of gut inflammation, as in our previous IBD studies (99) and in EAE studies by others (96). Consistent with previous studies, quantification of lipocalin-2 from fecal samples by ELISA demonstrated significant increases following EAE ( Figure 3I ). However, no significant differences in fecal lipocalin-2 levels between PWD+Akk and PWD microbiome mice were observed either before or following EAE induction ( Figure 3I ). Moreover, no association was found between lipocalin-2 (chronic EAE timepoint) and cumulative disease score ( Supplementary Figure 3A ). Taken together, these results suggest that EAE exacerbation by A. muciniphila colonization is not accompanied by significant changes in gut inflammation.

Our finding that exacerbation of EAE by A. muciniphila is highly dependent on the ecological context of the microbiome suggested interactions with other members of the microbiome. We therefore sought to explore how the B6 and PWD microbiomes differ and respond uniquely to colonization by A. muciniphila, using full length 16S DNA sequencing to assess gut microbial composition across the 4 microbiomes. We focused our analysis on fecal samples collected prior to EAE induction, in order to avoid potentially confounding effects of EAE progression on gut microbiome composition (35, 57), especially given the differences in EAE severity among groups of interest ( Figures 3A–D ). In agreement with relative abundance determined by qPCR ( Figure 2F ), 16S analysis indicated that A. muciniphila represents 6.3% of the total of gut bacterial reads in PWD+Akk microbiome mice compared with 0.64% in B6+Akk microbiome mice ( Figure 4A ; Supplementary Figure 4A ), confirming successful colonization in both contexts, with relative abundances reflective of those found in human microbiome analyses (23, 100), and indicating a possible difference in niche occupancy across microbiome contexts. Shannon diversity, a metric that considers both species richness and evenness, demonstrated a reduction in alpha diversity driven by A. muciniphila colonization only in the PWD microbiome ( Figure 4B ). Statistical analyses of overall microbial community structure (beta diversity) demonstrated a highly significant difference by PermANOVA (adonis2) between base microbiomes (B6 vs PWD; R ² = 0.64, p = 0.001), as expected from our previous studies (75). This analysis also revealed more modest yet significant effects of A. muciniphila colonization (R ² = 0.052, p = 0.004) and of base microbiome × A. muciniphila interactions (R ² = 0.049, p = 0.002), suggesting that A. muciniphila colonization alters the composition of the microbiome in a context-dependent manner ( Figure 4C ). Further, using pairwise comparisons on our A. muciniphila-free and A. muciniphila-colonized microbiomes (83), we found that in the PWD microbiome context, A. muciniphila accounted for 6.7% of the Bray-Curtis dissimilarity, but only 0.8% of the dissimilarity in the B6 microbiome context ( Supplementary Figure 4B ). Together, these data suggest that A. muciniphila reshapes the PWD gut microbiome to a greater extent than it does the B6 gut microbiome.

In order to identify changes in specific bacterial taxa caused by A. muciniphila colonization unique to each of these two different microbiome contexts, we assessed differential abundance of amplicon sequence variants (ASVs) between A. muciniphila-colonized mice and their A. muciniphila-free counterparts. Analysis of differentially abundant (p_adj < 0.05) ASVs between PWD+Akk and PWD microbiome mice revealed a reduction in the abundance of a large number (40) of ASVs, while only 3 ASVs were increased in abundance, with A. muciniphila among them, as expected ( Figure 4D ). Examination of ASVs decreased in abundance by A. muciniphila colonization in the context of the PWD microbiome revealed that these ASVs predominantly belonged to the Clostridia class, including a number of Lachnospiraceae, Ruminococcaceae, and Oscillospiraceae family members ( Figure 4D ; Supplementary Figure 4C ; Supplementary Table 1b ), all well-known SCFA producers (52). Collapsing the ASVs to the taxonomic level of class confirmed a global reduction in Clostridia ( Figure 4E ). These changes suggest that colonization with A. muciniphila in the context of the PWD microbiome, where it selectively increases EAE severity, drives a depletion in numerous members of the SCFA-producing Clostridia class, prominently including ASVs of the Lachnospiraceae family.

We next performed a differential abundance analysis comparing B6+Akk and B6 microbiomes, a context where A. muciniphila colonization does not exacerbate EAE. In contrast to changes induced by A. muciniphila colonization in the PWD microbiome, this revealed more subtle effects, with fewer (21) total differentially abundant ASVs, with only 3 of these decreased in abundance ( Figure 4F ). The little (5 ASVs) overlap among differentially abundant ASVs between comparisons of B6 ± Akk and PWD ± Akk microbiomes, included the expected increased abundance of A. muciniphila, and 4 other ASVs that were decreased in abundance in the PWD ± Akk comparison but increased in the B6 ± Akk comparison ( Figure 4G ). Examination of the differentially abundant ASVs driven by A. muciniphila in the B6 context revealed that several Lachnospiraceae (11) and/or Clostridia (14) ASVs were in fact increased by A. muciniphila colonization, while only 2 Clostridia ASVs (3 total ASVs) were decreased ( Figures 4F, G ). Taken together, these results demonstrate that colonization by A. muciniphila in the context of the B6 microbiome exerts a highly divergent effect compared with colonization in the PWD microbiome, lacking the antagonistic effect on Clostridia.

To get at the basis of ASV-specific depletion of Clostridia ASVs in the PWD and not the B6 microbiome, we next examined the abundance of the numerous ASVs downregulated by A. muciniphila colonization in the PWD microbiome ( Figure 4D ), in the B6 microbiome with or without A. muciniphila colonization. We found that many of these ASVs (which were predominantly Clostridia) were abundant in the PWD microbiome but either absent (0 mapped reads) or at low abundance in the B6 microbiome with or without A. muciniphila ( Figure 4H ), suggesting that in the PWD microbiome, A. muciniphila colonization may have unique negative interactions with specific members of the Clostridia class, whereas many of these interactions and/or bacteria are absent in the B6 microbiome. Altogether, these findings suggest that A. muciniphila colonization exerts highly divergent effects on B6 and PWD microbiome composition, in parallel with divergent effects on EAE severity, suggesting that these two phenotypes are functionally linked.

Because gut microbiota represent a highly interconnected ecosystem, we utilized an approach to visualize and quantify bacterial interaction networks. We used the semi-parametric ranked-based approach for inference in graphical model (SPRING) approach using the NetCoMi package to generate networks of our PWD ± Akk and B6 ± Akk microbiomes, to identify differences in microbial associations and network connectivity. The B6 network had an overall modest connectivity, with a relative largest connected component (LCC) size of 0.44 ( Figure 5A ), compared with the well-connected PWD network, with a relative LCC of 0.86 ( Figure 5B ). Moreover, the B6 network included 18 disconnected subnetwork components compared to 7 components in the PWD network ( Figures 5A, B ), suggesting more cohesion in the microbial community structure of the PWD microbiome. No connectivity between A. muciniphila and Clostridia was observed in the B6 network ( Figure 5A ). In contrast, the PWD network included numerous negatively weighted edges connecting A. muciniphila and various nodes belonging to Clostridia (e.g. Lachnospiraceae), with the latter also forming many connections with other members of the network ( Figure 5B ). Taken together, these results suggest that the Clostridia component of the PWD microbiome represents a well-integrated part of the microbial ecosystem that is particularly vulnerable to disruption by A. muciniphila colonization.

To analyze the functional consequences imparted by A. muciniphila colonization across our microbiome models, we employed PICRUSt2 ( 85) to infer functional microbial gene content from the taxonomic data derived from our full-length 16S analysis. Considering the unique reduction in Clostridia-mapped ASVs in the PWD+Akk microbiome, we predicted a shift among pathways pertaining to SCFA production. Indeed, pathway enrichment analysis predicted a reduction in 5 pathways in the PWD+Akk microbiome compared to the PWD microbiome, 2 related to butyrate production and 3 related to both butyrate and acetate production ( Figure 6A ; Supplementary Table 2b ). Only two SCFA-related pathways, one related to acetate consumption and the other related to propionate production, were enriched in the PWD+Akk microbiome ( Figure 6A ). The latter is consistent with the known role for A. muciniphila itself in propionate production (22). These results suggest that the selective depletion of Clostridia by A. muciniphila colonization in the PWD microbiome drives a net depletion in microbial pathways related to SCFA production.

Given these pathway enrichment analysis results, and the reported ability for butyrate to dampen autoimmune responses like those seen in MS (58, 62), we quantified the abundance of ASVs contributing specifically to butyrate production across our microbiome models. As previously published (86), we defined butyrate-producing microbiota as those ASVs whose metagenomes encode terminal butyrate-producing enzymes (87). The relative abundance of total butyrate-producing microbiota was significantly decreased in the PWD+Akk microbiome compared to A. muciniphila-free counterparts, whereas butyrate producers showed no such reduction in our B6 ± Akk microbiomes ( Figures 6B, C ). These results confirm the selective depletion of predicted butyrate-producers by A. muciniphila colonization in the PWD microbiome.

To determine if the depletion of SCFA-producing microbiota contributed to measurable changes in SCFA levels, we quantified SCFAs in day-30 fecal samples from PWD and PWD+Akk mice via GC-MS. PWD+Akk microbiome mice exhibited a significant depletion of acetate (C2) compared to their A. muciniphila-free counterparts ( Figure 6D ), consistent with inferred metagenomic analyses identifying an enrichment in acetate-consuming and a depletion in acetate-producing pathways in the PWD+Akk microbiome ( Figure 6A ). Combined isobutyrate and butyrate levels were also decreased, although not significantly, in PWD+Akk microbiome mice ( Figure 6E ). Total SCFA levels were also significantly decreased in PWD+Akk microbiome mice ( Figure 6F ). Collectively, these data suggest that A. muciniphila colonization causes a net decrease in SCFA production in a microbiome composition-dependent manner, which is linked to EAE exacerbation.

In effort to generate additional B6 mice colonized with the original compositionally well-characterized PWD microbiome using limited original cecal inoculum stocks, we colonized new ex-GF mice using a cecal inoculum from a PWD microbiome-colonized B6 mouse (B6.Gut^PWD) donor rather than an original PWD donor ( Figure 7A ), generating mice designated as PWDβ microbiome-colonized. To generate A. muciniphila-colonized counterparts for these new mice, A. muciniphila was combined with the initial cecal inoculum (100µL cecal inoculum + 100µL A. muciniphila culture) to generate 3 PWDβ and PWDβ+Akk microbiome breeding pairs whose experimental offspring mice used for EAE ( Supplementary Figure 5A ).

Surprisingly, unlike the effect of A. muciniphila seen in the original PWD microbiome-colonized mice, EAE severity did not differ between PWDβ and PWDβ+Akk microbiome mice ( Figures 7B, C ), suggesting that A. muciniphila colonization of the original PWD microbiome may have contributed to unique changes in the gut microbiota that were not recapitulated in the context of the new PWDβ+Akk microbiome. To better understand these differences, we analyzed full-length 16S sequencing data from fecal samples of PWDβ and PWDβ+Akk microbiome mice, as above. Surprisingly, A. muciniphila represented only 0.20% of the total mapped reads in PWDβ+Akk microbiome mice, more similar to the abundance seen in B6+Akk microbiome mice ( Figure 7D ; Supplementary Figure 5B ). Comparing the Shannon diversity across all 6 microbiomes demonstrated a significant reduction in microbiome diversity only between PWD and PWD+Akk microbiomes ( Figure 7E ), suggesting that A. muciniphila is less disruptive to microbiome diversity in PWDβ compared to the PWD microbiome. To identify differences in the microbial composition among PWD and PWDβ microbiomes, we focused our beta diversity analysis on these microbiomes and their A. muciniphila-colonized counterparts ( Figure 7F ). Using PermANOVA, we found a significant effect of base microbiome (PWD vs. PWDβ; R ² = 0.23, p = 0.001), a significant effect of A. muciniphila (R ² = 0.14, p = 0.001), and a significant A. muciniphila × base microbiome interaction (R ² = 0.052, p = 0.002). These results suggest that the effect of A. muciniphila colonization on the overall microbiome composition differs between the PWD and PWDβ microbiomes, with a disparate and muted effect seen in the latter context.

We next performed differential abundance analysis comparing PWDβ and PWDβ+Akk microbiome mice. Compared with the above-described effect in the PWD microbiome ( Figure 4F ), we observed far fewer (9) significantly differentially abundant ASVs ( Figure 7G ) and minimal changes at other taxonomic levels besides Akkermansia itself ( Supplementary Figures 5C–E ), confirming a limited effect of A. muciniphila colonization on the PWDβ microbiome. In contrast to A. muciniphila colonization in the PWD+Akk microbiome ( Figure 4D ), a reduction of the Clostridia class was not observed ( Supplementary Figure 5C ), with 2 Lachnospiraceae members increased and 2 Lachnospiraceae members decreased with A. muciniphila colonization ( Figure 7G ). Furthermore, network analysis of the PWDβ ± Akk microbiomes using previously used parameters ( Figures 5A, B ) failed to find any connectivity for A. muciniphila ( Figure 7H ), suggesting a lack of major changes to the gut microbiota caused by A. muciniphila colonization in the PWDβ microbiome.

To determine the inferred functional effect of A. muciniphila in the context of the PWDβ microbiome, we performed pathway enrichment analyses of inferred gene content and quantification of butyrate producer abundance, as above ( Figures 6A–C ). Consistent with the lack of change in overall Clostridia abundance between PWDβ and PWDβ+Akk microbiomes, we observed no significant enrichment in pathways related to SCFA metabolism ( Supplementary Table 4C ), with few changes in any functional pathways driven by A. muciniphila colonization. Additionally, we observed no significant decrease in the relative abundance of butyrate producers ( Figure 7I ). Taken together, these results demonstrate limited effects of A. muciniphila colonization on the composition of the PWDβ microbiome. This suggests the existence of differences between the PWD and PWDβ ecological contexts that limit the ability for A. muciniphila colonization to restructure the PWDβ microbiome and contribute to increased EAE severity.

To identify such ecological differences, we compared the base PWD and PWDβ microbiomes by analysis of differentially abundant ASVs, inferred functional pathways, and relative abundance of butyrate producers, as described above. Members of the Clostridia class dominated differentially abundant ASVs, with 23 Clostridia ASVs underrepresented and 4 Clostridia ASVs overrepresented in PWDβ compared to PWD ( Figure 7J ). Moreover, inferred metagenomic analyses identified a significant decrease in in the relative abundance of butyrate producers between these microbiome contexts, emphasizing decreased butyrate production potential in the PWDβ context ( Figure 7K ). Finally, pathway enrichment analysis found 4 SCFA-related pathways, all predicting a reduction in butyrate production potential in the PWDβ microbiome compared with the PWD microbiome ( Figure 7L ). Taken together, these results demonstrate that compared with the PWD microbiome, the derivative but distinct PWDβ microbiome exhibits lower levels of Clostridia and SCFA production potential, offering a potential mechanism for the lack of effect of A. muciniphila colonization and EAE exacerbation.

Because the effect of A. muciniphila on EAE exacerbation and Clostridia depletion was specific to the PWD microbiome and lacking in the B6 microbiome, we asked whether there was a baseline difference in abundance of Clostridia and their associated functional pathways between B6 and PWD microbiomes prior to A. muciniphila colonization. Of a total of 136 differentially abundant ASVs, 92 were Clostridia members overrepresented in PWD and only 13 Clostridia members were underrepresented in PWD ( Figure 8A ; Supplementary Table 1E ). Notably, when comparing the base PWD and B6 microbiomes for pathways related to SCFA metabolism, 6 pathways related to production of butyrate were significantly overrepresented in the PWD microbiome ( Figure 8B ). Consistent with this, we found butyrate producers to be significantly and dramatically more abundant in the PWD microbiome compared to the B6 microbiome ( Figure 8C ). These results demonstrate that, compared with the B6 microbiome, the baseline PWD microbiome contains more SCFA-producing Clostridia, suggesting a major contribution of SCFA production to the immunological homeostasis conferred by PWD microbiota.

To test the high capacity for, and potential reliance on, SCFA metabolism by gut Clostridia in the PWD microbiome in order to maintain immunological homeostasis and prevent autoimmunity, we utilized modulation of dietary fermentable fiber, as a prebiotic substrate for SCFA production. We hypothesized that chow enriched with soluble dietary fiber, which Clostridia metabolize to produce SCFAs (101), would attenuate subsequent EAE severity in PWD microbiome mice to a greater extent than in B6 microbiome mice, which have lower abundance of SCFA-producers. PWD and B6 microbiome mice were randomized to high (20% pectin and 10% inulin) and low (0%) fiber diets 2 weeks prior to EAE induction, which were maintained throughout the course of EAE. High fiber diet failed to suppress EAE in B6 microbiome mice ( Figures 8D, E ). In contrast, high fiber diet significantly attenuated EAE severity in PWD microbiome mice ( Figures 8F, G ). Taken together, these results provide functional confirmation that the high SCFA production potential of the PWD microbiome results in a more effective suppression of EAE when excess dietary fiber is provided. They also highlight the importance of baseline microbiome composition in modulating the therapeutic response to prebiotic intervention.

The exact mechanisms by which A. muciniphila depletes Clostridia and drives CNS autoimmunity in a context-dependent manner remain unclear. Previous studies showed bacterial antagonism by A. muciniphila has been demonstrated against the Clostridia member Ruminococcus via the release of potentially antibacterial metabolites (129). The intestinal mucus layer is a unique niche for gut bacteria, where A. muciniphila and Clostridia species both reside and interact intimately with each other, potentially in competition for resources like dietary and host polysaccharides and niche occupancy. Such competition could be heightened in microbiome contexts where these mucosal microbiota are more abundant, such as our Clostridia-rich PWD microbiome, in which A. muciniphila colonizes at a higher abundance ( Figure 2F ). Although we cannot definitively say that the higher abundance of A. muciniphila or depletion of Clostridia is required for EAE exacerbation, these phenotypes appear to be linked, at least across the three distinct microbiomes examined in our study. This could be addressed in future studies by experimentally increasing the abundance of Clostridia in the B6 microbiome prior to A. muciniphila colonization. Moreover, it is formally possible that the high abundance of A. muciniphila, rather than the microbial ecology per se, drives effects on EAE severity; although the ecology clearly dictates A. muciniphila abundance. We also do not know why A. muciniphila colonizes the PWD microbiome at a higher level, but this is clearly associated with the baseline abundance of Clostridia, and hence potential metabolic crosstalk or even species-specific production of antimicrobial compounds. Nonetheless, research investigating A. muciniphila and cancer immunotherapy has suggested a bimodal effect of A. muciniphila abundance, where A. muciniphila-colonized individuals outperformed A. muciniphila-free counterparts, however A. muciniphila overabundance (exceeding 4.8%) was associated with reduced overall survival (100). This parameter would describe our PWD+Akk microbiome as having an overabundance ( Figure 6C ; 6.3%) of A. muciniphila that may contribute to disease worsening. However, this does not pinpoint how this overabundance contributes to the depletion of Clostridia, although it is interesting to note that other classes of gut bacteria are not depleted. To address this, future studies may be able to leverage a reduction in oxygen-sensitive Clostridia. In the PWDβ microbiome, this is likely an effect of the secondary cecal microbiota transplant model ( Figure 6A ), contributing to the concurrent loss of EAE responsiveness of this microbiome to A. muciniphila colonization. Inter-microbiota interactions remain a complex challenge for elucidating gut microbiota associations with disease, where the use of additional microbiota contexts may become valuable.

While we showed that the effects of A. muciniphila on EAE are correlated with a depletion in Clostridia, we did not demonstrate that depletion of Clostridia drives EAE outcomes, e.g. by rescuing EAE exacerbation seen with A. muciniphila colonization by supplementation of Clostridia. Moreover, we did not demonstrate that mice colonized with the PWD microbiome and A. muciniphila lose responsiveness to high fiber supplementation and SCFA production. These experiments were precluded by the limited quantity of the original PWD microbiome inoculum available to generate new mice for these experiments.

Multiple challenges remain in elucidating the relationship between gut microbiota and MS. The human gut microbiome includes substantial variability across individuals that is poorly captured in animal studies, although our use of divergent gut microbiomes representing two distinct ecological contexts emphasizes the importance in considering microbiota baseline composition in experimental designs. Similarly, changes in bacterial taxonomy and strain differences within bacterial species, especially across host organisms, limits the translatability and reproducibility of gut microbiome research. In this regard, even the full length 16S rRNA sequencing used in our study may underperform in specifying the genera and species of some ASVs compared to shotgun metagenomic methods (130) and only infers gene content and pathways via indirect approaches (131). A. muciniphila, previously described as the only species of the Verrucomicrobiota phylum to colonize the human gut, now shares its genus with several other Akkermansia species, and research suggests that A. muciniphila strain variation also contributes differentially towards disease (18). Additionally, the ability for dietary fiber to ameliorate EAE in our A. muciniphila-colonized microbiome models remains underexplored, although we would expect blunted amelioration due to the depletion of Clostridia seen in PWD+Akk microbiome mice. Because modulating dietary fiber is a highly translatable intervention, interactions between dietary fiber treatment and A. muciniphila should be further explored.