Central Nervous System Demyelinating Disease Protection by the Human Commensal Bacteroides fragilis Depends on Polysaccharide A Expression

Multiple sclerosis (MS) is a human disease of the CNS that is characterized by an inflammatory process followed by demyelination and axonal loss (1). In MS, autoreactive cells respond to self-Ags by inducing inflammation and attack the CNS, causing the axonal damage. Experimental autoimmune encephalomyelitis (EAE), the most widely used animal model for human MS, has provided significant information suggesting the role of T cells in the induction of this autoimmune disorder (2). It has been hypothesized that encephalitogenic T cells are activated in the periphery and then cross the blood-brain barrier into the CNS, where they encounter the self-Ag, and become reactivated. This reactivation could induce the release of proinflammatory cytokines, provoking the activation of resident macrophage–microglia cells to release NO that is directly involved in the demyelination of the neuronal myelin sheath (3). There is substantial evidence collected in EAE suggesting that Th immune responses might be polarized toward Th1 and Th17 (1). Recent findings have further supported a primary role for IL-17 in the pathogenesis of human MS (4, 5).

Gut commensal microorganisms can modulate immune homeostasis (6–13). Studies in germ-free animals, born and raised in sterile conditions, showed that monocolonization with Bacteroides fragilis was sufficient to stimulate early development of the GALT, to induce normal organogenesis in the spleen and thymus, and balanced immune development (14). Bacteroides species are gram-negative bacteria that compose ∼25% of the microbiota in both humans and other mammals. B. fragilis is symbiotic with the host, but if it reaches sterile then extraluminal sites can be responsible for tissue infection, bacteremia, and abscess formation in the peritoneal cavity, brain, liver, pelvis, or lungs (15, 16).

The administration of a mixture of gut commensal bacteria can protect against EAE (17). The protection observed was found to be IL-10 dependent. We (18) and others (19) recently demonstrated that modification of the bacterial populations of the gut alters the clinical outcome of EAE in mice. Oral treatment of mice with antibiotics reduced EAE severity by diminishing proinflammatory responses and the enhancement of Foxp3⁺T_reg cells that significantly accumulated in mesenteric and cervical lymph nodes (LNs). Adoptive transfer of these IL-10–producing T regulatory (T_reg) cells conferred protection against EAE. In this study, we investigate the effect of oral antibiotic treatment followed by gut reconstitution with a human isolate of B. fragilis that produces the zwitterionic capsular polysaccharide A (PSA), or with an isogenic mutant of B. fragilis deficient in the production of PSA in the development and protection against EAE. PSA has been found to be determinant in the regulatory effect of B. fragilis, restoring the default Th2-immune bias of germ-free animals (14, 16, 20). Moreover, IL-10–producing CD4⁺CD45rb^low T cells induced in response to PSA administration were protective in a Helicobacter hepaticus model of experimental colitis (21), and IL-10–producing T_reg cells protective against EAE (22). In this study, we demonstrate that the absence of PSA production by a human isolate of B. fragilis used to reconstitute disease-resistant mice restores clinical disease susceptibility, whereas the reconstitution with PSA-producing B. fragilis maintains resistance by the induction of highly potent IL-10–producing T_reg cells. Our results suggest a potent regulatory role for this specific bacterial Ag in the control of CNS demyelination in this experimental model of human MS.

Female 6-wk old SJL/J mice were obtained from The Jackson Laboratories (Bar Harbor, ME). All animal care and procedures were in accordance with at Dartmouth College Animal Resources Center institutional policies for animal health and well-being. The animal supplier confirmed that all mice used in the studies were free of exposure to Helicobacter. Dartmouth College Animal Resources Center routinely screens for a wide range of infectious agents, including Helicobacter. Because germ-free housing was not available, mice were maintained in a restricted, access-controlled environment. All water, feed, and cages used were previously sterilized and changed daily. Mice were treated with the following antibiotics dissolved in drinking water for 1wk: ampicillin (1 g/ml), vancomycin (0.5 g/ml), neomycin sulfate (1 g/ml), and metronidazole (1 g/ml) (23). Wild type (WT) B. fragilis (National Collection of Type Culture 9343) and the isogenic mutant of PSA-deficient (ΔPSA) B. fragilis were provided by D.L. Kasper (Harvard Medical School). One week after treatment with antibiotics, mice were infected with 10¹⁰ WT or ΔPSA B. fragilis resuspended in 200 μl sterile PBS by a single-time oral gavage.

Serial dilutions of intestinal and fecal samples were collected 1 wk after the end of treatment with antibiotics and/or bacterial reconstitution and cultured in general bacteriologic agar plates (CDC blood agar; BD Diagnostic Systems, Sparks, MD) and Bacteroides selective media plates (Bacteroides Bile Esculin Agar; BD Diagnostic Systems) for 48 h at 37°C. Plates were cultured in aerobic and anaerobic conditions. Total bacteria per gram of sample was calculated based on the colony forming units counted in each serial dilution.

SJL mice were challenged s.c. with 200 μg PLP_139–151 (Peptides International, Louisville, KY) in 200 μl CFA (Sigma-Aldrich, St. Louis, MO). On days 0 and 2 after challenge, mice received 200 ng Bordetella pertussis toxin i.p. (List Biological Laboratories, Campbell, CA) (24). Mice were monitored and scored daily for disease progression (24).

Luminex and specific cytokine ELISA were used to quantify triplicate sets of supernatants. Cells were cultured in 24-well tissue plates at 2 × 10⁶ cells/ml in the presence of anti-CD3 mAb-coated wells (10 μg/ml; BD Biosciences, San Diego, CA), plus the soluble anti-CD28 mAb (5.0 μg/ml; BD Biosciences) for 3 d, or media (24).

One microgram of QIAgen RNeasy-purified (Qiagen, Valencia, CA) mRNA was reverse transcribed using Multiscribe RT (Amersham Biosciences AB, Uppsala, Sweden). cDNA (200 ng) was amplified using the ×2 SYBR green mix (Applied Biosystems, Foster City, CA) on a Bio-Rad (Hercules, CA) iCycler. Relative expressions were normalized (β-actin sample–β-actin of naive samples) and expressed using the cycle threshold method, where relative mRNA expression = 2^-(ΔCt)*1000.

Single cervical LN lymphocyte preparations were stained using conventional methods. T cell subsets were analyzed using fluorochrome-conjugated mAbs (BD Biosciences) for CD4 and CD25. Intracellular staining for Foxp3 was performed using fluorochrome-labeled anti-Foxp3 mAb (clone FJK-16s; eBioscience, San Diego, CA). Fluorochrome-labeled anti-rat IgG2a (eBioscience) isotype control for Foxp3 expression was used. Dendritic cells (DCs) were analyzed using CD11c, CD11b, and CD103 (BD Biosciences). Bound fluorescence was analyzed with an FACS Canto (BD Biosciences).

CD11c⁺ cells were enriched with magnetic beads (StemCell Technologies, Vancouver, British Columbia, Canada) and then sorted (FACSVantage with Turbo-Sort; BD Biosciences) after staining with FITC-anti–CD103 into CD11c^highCD103^+/− cells. CD4⁺ T cells were obtained with magnetic beads (Dynal Biotech ASA, Oslo, Norway). The enriched CD4⁺ T cells were cell-sorted for CD4⁺CD25⁻ T cells using FITC-anti–CD4 and PE-anti–CD25 mAbs (BD Biosciences).

Cervical LN CD4⁺CD25⁻ T cells were and cultured for 4 d in the presence of anti-CD3/CD28 and IL-2 at different concentrations of TGF-β (0, 0.1, 0.5, and 5 ng/ml). To compare the role of CD11c^highCD103^+/− cells in the conversion of CD4⁺CD25⁻ T cells into Foxp3⁺T_reg cells, 1 × 10⁴ sorted CD103^−/+CD11c^high cells were cocultured with 5 × 10⁴ splenic naive CD4⁺CD25⁻ T cells in the presence of anti-CD3 Ab (plates were precoated with 10 μg/ml anti-CD3 Ab [BD Biosciences]). Cultures were set in the presence of PBS, purified PSA (100 μg/ml), or 4 nM retinoic acid (RA) and 5 ng/ml TGF-β. Foxp3 acquisition by CD4⁺ T cells was measured by flow cytometry. For adoptive transfer experiments, 1 × 10⁵ in vitro Foxp3⁺T_reg converted cells were injected i.v. into naive recipients. EAE was induced 1 d after adoptive transfers.

To inactivate CD25⁺CD4⁺ T cells, mice were given 0.3 mg anti-CD25 mAb (ATCC No. TIB-222, clone PC 61.5.3) on days 4 and 2 before EAE challenge (24). As a control group, treated and naive mice received 0.3 mg purified rat IgG Ab. A separate control group was immunized with PBS 7 d prior to EAE challenge.

Kruskal-Wallis followed by Dunn’s comparison of multiple groups was applied to show differences in EAE clinical scores, cumulative clinical scores, luminex and ELISA detection of cytokines as well as in the flow cytometry of T_reg cell and DC experiments. The p values <0.05 and <0.01 are indicated.

We have reported that oral antibiotic treatment can protect against CNS demyelination in a murine model of EAE (18). Four experimental groups were used throughout the studies: 1) mice treated orally with antibiotics for 1 wk, 2) mice treated with antibiotics and subsequently reconstituted with WT B. fragilis, 3) mice treated with antibiotics and subsequently reconstituted with ΔPSA B. fragilis, and 4) mice sham-treated with PBS. EAE was induced 1 wk after bacterial reconstitution or PBS sham-treated control. Oral antibiotic treatment significantly reduced the total bacterial numbers recovered (Fig. 1A). WT or ΔPSA B. fragilis persist and replicate similarly in vivo (14). The reconstitution of mice with either WT or ΔPSA B. fragilis restored the number of detectable bacteria (Fig. 1A). Bacteriologic culture of fecal samples in Bacteroides selective media showed that the treatment with antibiotics reduced the Bacteroides spp. counts when compared with naive mice whereas oral administration of WT or ΔPSA B. fragilis resulted in effective reconstitution of the gut (Fig. 1B).

We examined whether reconstitution of mice with WT or ΔPSA B. fragilis confers protection against EAE (18, 19). Oral treatment with antibiotics reduces EAE severity and cumulative scores, and it delays the clinical onset (18) (Fig. 2A, Table I). Reconstitution of antibiotic-treated mice with ΔPSA B. fragilis reversed protection by restoring susceptibility to disease. Mice recolonized with the intact strain of B. fragilis were protected against disease and demonstrated reduced clinical severity when compared with the PBS control group or ΔPSA B. fragilis-reconstituted mice (Table I). There was no significant difference in the protection between treatment with antibiotics and antibiotic treatment followed by reconstitution with WT B. fragilis (Table I). In contrast, oral antibiotic treatment followed by reconstitution with ΔPSA B. fragilis resulted in clinically significant disease severity consistent with that observed in the PBS control group.

Cytokine analysis of CNS tissue derived from WT B. fragilis-reconstituted mice versus either control PBS-treated or ΔPSA B. fragilis-reconstituted mice was compared (Fig. 2B). T-box transcription factor TBX21 (T-bet) and IFN-γ, both indicators of Th1 polarization, were reduced in mice treated with antibiotics compared with PBS treated mice, whereas GATA-3 and IL-13 expression were enhanced. There was a reduction in RORγt and IL-17–relative expression in mice treated with antibiotics, and in WT B. fragilis-reconstituted mice when compared with PBS-treated mice. Reconstitution with WT B. fragilis in EAE mice enhanced GATA-3, SMAD-3, and IL-10 compared with PBS control mice. The profile of mice reconstituted with ΔPSA B. fragilis showed no alterations of cytokine relative expression compared with PBS control mice. Enhanced RORγt, IL-17, and T-bet levels and reduced GATA-3, IL-10, and IL-13 levels of expression were detected in ΔPSA B. fragilis-reconstituted mice versus mice treated with antibiotics. When compared with WT B. fragilis, ΔPSA B. fragilis-reconstituted mice showed enhanced RORγt and IL-17 as well as and reduced SMAD-3 and IL-10.

Previous studies from our laboratory have demonstrated changes in cytokine production by cells in the cervical LNs of mice after oral antibiotic treatment (18). IL-13 levels were increased in mice treated with antibiotics compared with mice reconstituted with WT or ΔPSA B. fragilis (Fig. 3). The reconstitution with either WT or ΔPSA B. fragilis induced the production of IFN-γ, compared with mice treated with PBS and with antibiotics. Whereas treatment with antibiotics significantly reduced the levels of IL-17 and IL-6 produced compared with PBS treatment (18), reconstitution with ΔPSA B. fragilis, but not WT B. fragilis, significantly augmented those levels. WT B. fragilis reconstitution induced enhanced levels of IL-10, IFN-γ, IL-12 (p40) in cervical LN cells when compared with mice treated with antibiotics. IL-10 production and GATA3- and SMAD3-relative expressions in cells obtained from WT B. fragilis-reconstituted mice were also enhanced, compared with mice reconstituted with ΔPSA B. fragilis and PBS-treated mice. Although reductions in TNF-α and MCP-1 were observed in mice treated with antibiotics and reconstituted with WT B. fragilis, the differences were found to be not significant compared with PBS control mice and mice reconstituted with ΔPSA B. fragilis (not shown). No significant changes were observed in IL-4, IL-5, MIP-1α, or MIP-1β levels (not shown).

We assessed whether PSA expression in B. fragilis could enhance the percentages of Foxp3⁺ T_reg cells. Oral treatment with antibiotics increased the frequency of Foxp3⁺CD25⁺ in CD4⁺ T cells isolated from the cervical LN (18). No significant differences were observed in the frequency of Foxp3⁺CD25⁺ T_reg cells detected in mice reconstituted with either WT or ΔPSA B. fragilis, compared with mice treated with antibiotics (Fig. 4A, 4B). The frequency of Foxp3⁺ was enhanced in the CD25^high versus the total CD25⁺ fraction of CD4⁺ T cells in all groups (Fig. 4C, 4D). Mice subjected to bacterial reconstitution showed similar frequencies of Foxp3⁺T_reg cell populations in the GALT (Peyer’s patches and mesenteric LN) and spleens to those observed in antibiotic treated mice (Supplemental Fig. 1). In vivo CD25⁺ cell depletion demonstrated a significant effect on the protection after reconstitution with WT B. fragilis and in mice treated with antibiotics, as shown before (18) (Fig. 5, Table II). Disease susceptibility after reconstitution with WT B. fragilis or in mice rendered susceptible after oral antibiotic treatment was enhanced when CD25⁺ T cells were depleted. Of interest was the increased severity of disease following reconstitution with WT B. fragilis compared with mice treated with antibiotics (Table II). As shown in both Fig. 2 and Table I, no significant differences were observed in the severity of EAE in mice reconstituted with ΔPSA B. fragilis and in PBS-treated mice. Depletion of CD25⁺ cells induced equivalent increases in the cumulative disease indexes and mortality of both experimental groups.

We further analyzed the role of PSA in the acquisition of regulatory phenotypes by CD4⁺ T cells. We previously showed that alteration of the gut flora with antibiotics educated CD103⁺DCs to enhance the conversion of CD4⁺ T cells into Foxp3⁺T_reg cells (18). To compare the effect of B. fragilis recolonization in the induction of regulatory phenotypes of T cells, CD103^+/− CD11c^high DCs were sorted from the cervical LNs of mice treated with antibiotics and from mice reconstituted with either WT or ΔPSA B. fragilis. DCs were cocultured in anti-CD3 Ab precoated plates with splenic CD4⁺CD25⁻ (∼10% Foxp3⁺) T cells in the presence of PBS or purified PSA (100 μg/ml; Fig. 6). When CD4⁺ T cells were cultured with CD103⁻ DCs, the conversion of Foxp3⁻CD4⁺ T cells into Foxp3⁺T_reg cells was significantly reduced (data not shown). CD103⁺ DCs from mice reconstituted with ΔPSA B. fragilis exposed to PSA showed significantly reduced conversion to Foxp3⁺T_reg cells compared with DCs from WT B. fragilis-reconstituted mice. The PSA exposure of CD103⁺ DCs from mice treated with antibiotics did not significantly enhance the conversion rates compared with exposure to PBS. Enhanced T_reg conversion was observed when DCs and T cells were cocultured with 4 nM RA and 5 ng/ml TGF-β. However, despite these optimized conditions, the conversion was significantly reduced in cocultures with CD103⁺ DCs from ΔPSA B. fragilis-reconstituted mice, compared with cocultures with DCs from mice treated with antibiotics and mice reconstituted with WT B. fragilis.

We next analyzed whether Foxp3⁻CD4⁺ T cells isolated from the cervical LN of PSA-producing B. fragilis-reconstituted mice could be converted into Foxp3⁺T_reg cells (Fig. 7A, 7B, Supplemental Fig. 2). CD4⁺CD25⁻ (∼10% Foxp3⁺) T cells from WT B. fragilis-reconstituted mice showed enhanced T_reg conversion rates compared with cells obtained from mice reconstituted with ΔPSA B. fragilis, mice treated with antibiotics, and PBS-treated mice, when cultured with 0.5 and 5 ng/ml TGF-β. Conversion rates were significantly enhanced in all groups when TGF-β and RA approached the optimal concentration (Supplemental Fig. 2) (25).

We compared the capacity of these converted Foxp3⁺T_reg cells to protect against EAE (Fig. 7D). Cells cultured with 5 ng/ml TGF-β were collected after 4 d and adoptively transferred into naive recipient mice. Cells converted from WT B. fragilis-reconstituted mice protected against EAE induction, whereas no protection was observed in cells converted from PBS, mice treated with antibiotics, or ΔPSA B. fragilis reconstituted mice. Cells converted from WT B. fragilis-reconstituted mice produced significantly increased levels of IL-10 compared with PBS-treated mice, mice treated with antibiotics, and ΔPSA B. fragilis-reconstituted mice; they produced a modest but significant increase in TGF-β compared with ΔPSA B. fragilis-reconstituted mice (Fig. 7C). No significant differences in the production of IFN-γ, IL-17, IL-6, and IL-13 were observed.

Alterations of the gut commensal bacteria populations by oral treatment with antibiotics can influence the development of EAE (18, 19). We now demonstrate that the reconstitution of mice with B. fragilis deficient in the production of the zwitterionic capsular PSA restores disease susceptibility in mice that had been rendered resistant to disease after treatment with oral antibiotics. Reconstitution with both B. fragilis strains similarly restored the numbers of detectable bacteria, which were significantly reduced after oral treatment with antibiotics. It has been shown that both WT and ΔPSA B. fragilis persist and replicate equally in vivo (14), suggesting that it was PSA and not the number nor strain of B. fragilis that was responsible for disease protection or susceptibility.

Oral treatment with antibiotics enhanced the frequency of Foxp3⁺T_reg cells within the cervical LN and reduced Th17 responses (18). The alterations of Th17 responses upon oral treatment of antibiotics have been recently confirmed by others (26, 27). After antibiotic treatment, reconstitution with either WT or ΔPSA B. fragilis resulted in a similar number and frequency of T_reg cells. In MS, the in vitro conversion rates of CD4⁺ T cells into T_reg cells are significantly reduced in MS patients compared with healthy controls (28). Moreover, functional suppression appears to be impaired (29). CD4⁺ T cells obtained from PSA-producing B. fragilis-reconstituted mice more efficiently converted into Foxp3⁺T_reg cells, with increased IL-10 production and enhanced protective potency after adoptive transfer.

Our results suggest that deficient PSA production could influence the functional role of DCs and Foxp3⁺T_reg cells induced by B. fragilis. Prior studies have demonstrated that CD4⁺ T cell activation by PSA is dependent on the presentation of the Ag by CD11c⁺ DCs (30). Foxp3⁺T_reg cell conversion by CD103⁺ DCs purified from PSA-deficient B. fragilis-reconstituted mice was significantly reduced compared with DCs from PSA-producing B. fragilis. Foxp3⁺T_reg cell conversion studies showed enhanced conversion rates of CD4⁺ T cells obtained from WT B. fragilis-reconstituted mice. SMAD3 was significantly increased in WT B. fragilis-reconstituted mice, and IL-6 was substantially reduced compared with ΔPSA B. fragilis-reconstituted mice. Differentiation of Th17 cells requires TGF-β and IL-6, whereas TGF-β is also required for T_reg cell induction in the absence of IL-6 (31). The differences observed in the conversion rates could be due to the potential capability of WT B. fragilis CD4⁺ T cells to enhance TGF-β in cultures that could facilitate their conversion into Foxp3⁺T_reg cells.

PSA may influence a distinct pathway involved in disease protection, as suggested by our experiments of T_reg neutralization. CD25⁺ cell depletion exacerbated EAE in all groups, but the enhancement of disease severity was significantly higher in mice treated with antibiotics compared with WT B. fragilis-reconstituted mice. Other subpopulations of regulatory cells such as NKT cells may participate in the protection against disease-induced treatment with oral antibiotics (19). Recent observations from our laboratory suggest that B cells could be important in this protective response (32). IL-13 production was reduced in mice reconstituted with WT B. fragilis compared with mice treated with antibiotics, and only IL-10–producing T_reg cells converted from Foxp3⁻CD4⁺ T cells of WT B. fragilis-reconstituted mice protected against EAE. Foxp3⁺T_reg cells that were derived from CD4⁺T cells of PSA-deficient B. fragilis-recolonized mice failed to protect against the disease. We recently demonstrated that a highly purified preparation of PSA is protective against EAE in conventional mice and that this protection is completely abrogated in IL-10–deficient mice, suggesting an important role of this cytokine in the PSA-induced control of the disease (22).

The presence or absence of PSA could determine protective or pathogenic outcomes in EAE. WT or ΔPSA B. fragilis reconstitution induced production of IFN-γ, when compared with mice treated with PBS and with antibiotics; however, IL-10 production was enhanced only after reconstitution with WT B. fragilis, whereas PSA-deficient B. fragilis reconstitution induced enhanced levels of IL-6. The imbalance created by alterations of PSA expression within the gut lumen may lead to peripheral systemic autoimmune disorders, such as EAE or human MS. In the absence of PSA, the human commensal B. fragilis can no longer regulate immune homeostasis, leading to autoimmune disease of the intestine, as described recently (21) and as reported in this study in the CNS. Our previous (18, 22, 32) and present studies suggest that differing compositions of gut microbiota could regulate the balance between protection and disease induction in MS and may offer a novel therapeutic approach for disease intervention.

We thank Dr. Azizul Haque, Dr. Jacqueline Y. Channon-Smith, Marc Christy, Yan Wang, John DeLong, Kathleen Smith, and Alan J. Bergeron for technical support and critical review of the manuscript.

Disclosures The authors have no financial conflicts of interest.