Host–microbiota interplay in mediating immune disorders

Age-dependent disease in mouse models

Airway exposure to bacteria can suppress allergic airway inflammation in adult mice.¹³ However, studies indicate that such treatments could be greatly optimized by administering the therapy at an optimal age window, because age is an important factor in immunoregulation of the lung. For example, an earlier report showed that exposure to endotoxin during the first 10 days of life, but not during adulthood, can cause the formation of tertiary lymphoid tissues in the lung.¹⁴ Later, it was shown that colonization by airway microbiota early in life induces T regulatory (T_reg) cells, a T cell type that plays a major role in maintaining immune tolerance and homeostasis, which contributes to reduced susceptibility to airway hyperresponsiveness in adulthood (Fig. 1).¹⁵ This effect is associated with an increase in the lung bacterial load and a shift in bacterial phyla from a predominance of Gammaproteobacteria and Firmicutes towards Bacteroidetes after the first 2 weeks of life. The increased accumulation of T_reg cells requires the interaction of programmed death 1 (PD-1) on T cells and programmed death-ligand 1 (PD-L1) on dendritic cells. Thus, the maturation of the lung microbiota during this 2-week window is a key factor for the host to develop protection against airway hyperresponsiveness.

Aside from the direct impact of lung microbiota on lung disease, recently a strong interest has emerged in characterizing the role of gut microbiota in lung disease. There is a clear gut–lung axis of communication, exemplified by gut microbiota’s impact on immune disorders, such as asthma; lung diseases, such as chronic obstructive pulmonary disease (COPD); and respiratory infections.^16,17 However, little is known mechanistically regarding how commensals in the gut modulate another mucosal site in the lung. Two reports point toward a tissue-specific effect of gut microbiota in regulating lung and intestinal inflammation. Invariant natural killer T (iNKT) cells recognize lipid antigens presented by CD1d.¹⁸ Initial findings showed that, compared with specific pathogen–free (SPF) mice, germ-free (GF) mice have an increase in iNKT cells in mucosal tissues, including colon and lung. They display an enhanced inflammatory response to chemical-induced colitis and exhibit airway hyperresponsiveness.^19,20 The augmented inflammation in response to chemical assaults in the intestine can be dampened in GF mice by colonization with a type of gut commensal, Bacteroides fragilis, or oral gavage with B. fragilis–derived sphingolipid during the first 2 weeks of life, but not afterward (Fig. 1). Conversely, iNKT cell accumulation in the lung and susceptibility to airway hyperresponsiveness can be halted in GF mice by colonization with the general microbiota community—but not by monocolonization with B. fragilis—during the first 2 weeks of life. It is particularly interesting that B. fragilis monocolonization only protects against intestinal but not lung inflammation. This tissue-specificity may be due to the local immunoregulatory effect of B. fragilis and/or B. fragilis–derived sphingolipid.

Skin also exhibits a similar age- and microbiota-dependent T_reg regulation. Thus, Staphylococcus epidermidis colonization on neonatal (postnatal day 7) but not adult skin leads to immune tolerance characterized by an increase in commensal-specific T_reg cells in skin and skin-draining lymph nodes, which is accompanied by reduced commensal-specific CD4⁺ T effector cells and diminished tissue inflammation (Fig. 1).²¹ An age-dependent effect is also present in another skin disease: psoriasis. A general antibiotic treatment in adult mice reduces the severity of psoriasis, with a decrease in IL-17– and IL-22–producing T cells. However, antibiotic treatment in neonatal mice induces long-term dysregulation of the microbial composition of the skin, an effect lasting until adulthood, which ultimately leads to an increased susceptibility to psoriasis development that is accompanied by an increase in IL-22–producing γδ⁺ T cells (Fig. 1).²² In summary, age-dependent interactions of the host and microbiota in lung, intestine, and skin can cause beneficial and/or detrimental effects on the host. This knowledge presents a unique window of opportunity for future microbial-based therapies for immune disorder–related diseases.

Human diseases: early-life perturbations of microbiota lead to lifelong consequences

In humans, there is much evidence suggesting that a change in microbiota composition in early life may affect later susceptibility to disease as an adult. For example, children exposed to farm environments have a decreased risk for the development of allergic disease.²³ In an animal model of HDM-induced allergies, low-dose endotoxin or farm dust protected mice from developing allergies by suppressing the activation of epithelial cells and dendritic cells through induction of the ubiquitin-modifying enzyme A20 (encoded by Tnfaip3).²⁴ Combined with recent GWAS data that identified several SNPs in the TNFAIP3-interacting protein (TNIP-1) as being associated with asthma, the authors further discovered that TNFAIP3 SNPs are linked to asthma in children growing up on farms. Most recently, an elegant study took advantage of two similar yet distinct U.S. agricultural populations, the Amish and the Hutterites, to elucidate the role of a farm environment in asthma risk.²⁵ Despite their similar genetic ancestries and lifestyle, one major difference between the two groups is that the Amish use traditional farming practices, whereas the Hutterites use industrialized farming practices. Children of Hutterites develop asthma 4–6 times more often than Amish children. The dust from Amish but not Hutterite homes protected mice from developing airway hyperreactivity and eosinophilia in a murine model of asthma. Mechanistically, the authors reveal that MYD88- and TRIF-dependent innate immune pathways are required for the protection mediated by Amish dust. Finally, the critical window for asthma control in early life, demonstrated first in animal models, has now been confirmed in human infants. Comparing the gut microbiota of 319 subjects, Arrieta et al. showed that infants at risk of asthma exhibited transient gut microbial dysbiosis during the first 100 days of life.²⁶ Thus, children at risk of asthma display reduced bacterial genera: Lachnospira, Veillonella, Faecalibacterium, and Rothia. Inoculation of GF mice with these four bacterial taxa ameliorated airway inflammation, demonstrating a causative effect of these bacteria on reducing asthma development.

With such strong evidence showing that a healthy/normal gut microbiota composition in early life is critical for lowering the incidence of immune disorders, it is not surprising to find that exposure to antibiotics can harm healthy microbiota and render the host more susceptible to many diseases. For example, antibiotic use within the first 6 months of life is associated with an increased susceptibility to allergy and asthma in later childhood.²⁷ Studies have also shown that antibiotic exposure during the first year of life is associated with the development of childhood wheezing and eczema, as well as inflammatory bowel disease (IBD).^28,29 In the IBD study, those receiving antibiotics were 2.9 times (95% CI: 1.2–7.0) more likely to develop IBD.²⁹ Another means of perturbation of the normal microbiota composition occurs in children born by cesarean section. Studies have shown that these children are more susceptible to type 1 diabetes mellitus, multiple sclerosis, allergy, and asthma later in life.^30–32 The mechanistic basis for how these microbial perturbations affect disease susceptibility is unclear. However, it is clear that antibiotics can either transiently or permanently alter the composition of healthy microbiota, usually via depletion of one or several taxa.³³ Understanding this mechanism could help define the pathogenesis of dysbiosis-related immune disorders and offer major diagnostic and therapeutic opportunities.

The innate immune system and the adjuvant effect

Infection has long been linked to the onset of autoimmune disease.^2,55 Neutrophils are a key innate immune cell type at the front line of fighting infectious diseases.⁵⁶ Neutrophils are increasingly being recognized as integral players in the cross talk between microbiota and the host immune system, as they respond to signals from the microbiota and control the expansion of the microbiome. Oral microbiota affect neutrophil phenotype, as shown by the results of a report studying Aggregatibacter actinomycetemcomitans infection.⁵⁷ This oral pathogen releases a pore-forming toxin—leukotoxin A (LtxA)—which activates peptidylarginine deiminase enzymes in neutrophils and leads to hypercitrullination of self-proteins. Hypercitrullination is associated with rheumatoid arthritis (RA) development, and citrullination patterns induced by LtxA largely recapitulate those seen in RA, a finding further supported by the high concordance of anti-LtxA antibodies with RA development in human patients. These findings suggested a potential role for A. actinomycetemcomitans infection in triggering RA.

Another study compared the microbiome of three groups of infants in Northern Europe.⁵⁸ Early-onset autoimmune diseases are common in Finland and Estonia but are less prevalent in Russia. This study showed that Finnish and Estonian infants both have a greater proportion of Bacteroides species than Russia infants.⁵⁸ Therefore, the Finnish and Estonian infants are exposed to lipopolysaccharide (LPS) primarily derived from Bacteroides, which is structurally and functionally distinct from the LPS of E. coli. Unlike E. coli LPS, which is a potent innate immune activator, Bacteroides LPS does not inhibit innate immune signaling and endotoxin tolerance. The study further showed that, unlike LPS from E. coli, LPS from B. dorei, a species of Bacteroides, does not decrease the incidence of autoimmune diabetes in NOD mice. Interestingly, B. dorei has been shown to be associated with T1D pathogenesis.⁵⁹ Early colonization with immunologically silencing microbiota may thus preclude some aspects of immune education.

LPS is the ligand for Toll-like receptor 4 (TLR4), and another study along similar lines revealed the contribution of TLR4 in ameliorating autoimmune diabetes in NOD mice.⁶⁰ TRIF is a downstream signaling molecule of TLR4 and TLR3. TRIF gene deletion causes an increased incidence of autoimmune diabetes in NOD mice, suggesting a protective role for TRIF against autoimmune diabetes.⁶⁰ Mice with gene knockout of TLR4 but not TLR3 in an SPF facility have accelerated autoimmune diabetes,^61,62 and TLR4 gene knockout mice in a GF facility have normal T1D incidence, suggesting that microbiota-dependent TLR4/TRIF signaling protects against T1D. Conversely, TLR2 gene deletion reduces the incidence of autoimmune diabetes, which is reversed when TLR2 gene knockout NOD mice are re-derived into the GF environment.⁶⁰ These findings suggest that microbiota deliver a prodiabetic signal through TLR2. Together, these results suggest that differences in microbiota-derived LPS among individuals may prevent their immune tolerance education, leading to autoimmune disease. These studies also indicate that microbiota can promote or inhibit autoimmunity by signaling through different receptors.

The IL-2 signaling pathway potently inhibits T follicular helper (T_FH) cell differentiation by decreasing Bcl-6 expression.^63,64 We discovered that SFB drive the T_FH response via a DC-dependent adjuvant effect that is independent of TCR specificity.⁴⁷ Using a DC transfer model, we found that SFB-mediated Bcl-6 upregulation corresponds with SFB-mediated, DC-dependent IL-2Rα⁺ CD4⁺ suppression that occurs only in Peyer’s patches (PPs). Consistent with other findings suggesting that IL-2 functions as a potent inhibitor of T_FH differentiation, anti-IL-2 treatment boosts the T_FH response, which leads to arthritis enhancement in SFB⁻ K/BxN mice, and SFB colonization does not further enhance these effects in the anti-IL-2–treated mice. These data suggest that the mechanism of SFB-triggered T_FH cell differentiation overlaps with anti-IL-2 treatment, which depends on suppression of the IL-2 signaling pathway. Our findings indicate that T cells can differentiate into T_FH cells upon receiving an SFB-mediated, DC-derived bystander signal that restrains IL-2 signaling, regardless of TCR Ag-specificity. Although molecular mimicry is a prominent mechanism whereby microbial infections trigger autoimmunity (see below), it is very unlikely that SFB induce differentiation of autoimmune T cells into T_FH cells by molecular mimicry of the self-antigen glucose-6-phosphate isomerase in the K/BxN model, because our data show that SFB also induce T_FH differentiation in NZB/NZW F1 mice, a non-transgenic murine SLE model.⁴⁷ Our finding that SFB mediate bystander activation of GPI-specific T_FH cells is supported by a recent report showing that SFB induced vigorous GC formation but a very low percentage of SFB-specific IgA in PPs compared with a commensal strain of E. coli.⁴⁴ Previous reports demonstrated an autoimmune-enhancing effect of SFB via the induction of T_H17 cells.^42,65 Chappert et al. showed that, in a T_H1 cell–driven murine model of autoimmune arthritis, SFB modulate the activation threshold of self-reactive T cells and enhance autoimmune arthritis.⁶⁶ Interestingly, they found that SFB enhance disease by inducing T_H1 but not T_H17 cells. In the local microenvironment of gut-associated lymphoid tissues, inflammatory cytokines—particularly IL-12 elicited by the commensal flora—preferentially enhanced the T cell response to self-antigen that was otherwise tuned down.

Molecular mimicry

Molecular mimicry is a well-known mechanism for bacterial or viral infection–triggered autoimmunity.^1,67 However, whether gut commensal bacteria can also trigger autoimmunity by molecular mimicry is poorly understood. Two studies have recently shed light on this topic (Fig. 2). The first report studied autoimmune uveitis, a major cause of blindness. Autoreactive retina-specific T cells were hypothesized to first become activated at non-eye sites and induce autoimmune uveitis by breaking through the blood–retinal barrier, given that their cognate antigens are sequestered within the immune-privileged eye.⁶⁸ To test this hypothesis, Horai et al. used a TCR transgenic mouse model that recognizes interphotoreceptor retinoid-binding protein (IRBP), a self-antigen, and demonstrated that gut commensals are required for the development of autoimmune uveitis. Retina-specific T cells are activated in a host that does not express self-antigen IRBP, and their activation is also independent of bacterial superantigens. Signaling through the retina-specific TCR by a protein extract derived from commensals is critical for the activation of retina-specific T cells, which can be blocked by anti-MHC class II antibodies, suggesting that the activation signals require TCR–MHC interactions. Although a molecular mimicry epitope was not identified in this study, it is conceivable that retina-specific TCRs can be activated by crossreactive products derived from commensals independent of the IRBP self-antigen.

The second study used islet-specific glucose-6-phosphatase catalytic subunit–related protein (IGRP)-specific CD8 TCR transgenic NOD mice to demonstrate that the gut microbiota strongly modulates CD8⁺ T cell–mediated T1D development (Fig. 2).⁶⁹ The authors further discovered that one type of commensal, fusobacteria, expresses peptides that share significant homology with IGRP and promote diabetes development. Thus, molecular mimicry between commensal antigens and an islet self-antigen provokes autoimmune diabetes. These studies raise a critical warning that activation of auto-reactive TCRs by commensal bacteria might be a more frequent trigger of autoimmune diseases than is currently appreciated.

Metabolites: short-chain fatty acids

Short-chain fatty acids (SCFAs), including butyric acid, propionic acid, and acetic acid, are the main metabolic products of undigested carbohydrates and have broad effects on the host immune system.⁷⁰ Recently, one group reported that long-chain fatty acids (LCFAs) enhanced the differentiation and proliferation of T_H1 and/or T_H17 cells, while SCFAs expanded gut T_reg cells.⁷¹ Using experimental autoimmune encephalomyelitis (EAE) as a model for multiple sclerosis, Haghikia et al. showed that LCFAs decreased SCFAs in the gut, leading to exacerbated EAE by expanding pathogenic T_H1 and/or T_H17 cells in the small intestine. Treatment with SCFAs ameliorated EAE by inducing lamina propria–derived T_reg cells. SCFAs can also indirectly regulate autoimmunity through other molecules, such as antimicrobial peptides (AMPs), expressed by epithelial and immune cells for their defense function against invading microbes.⁷² Interestingly, β cells have been shown to produce the cathelicidin-related antimicrobial peptide (CRAMP), and this production was defective in female but not male NOD mice.⁷³ Systemic CRAMP administration to female NOD mice induced T_reg cells in the pancreatic islets and reduced the incidence of autoimmune diabetes. Mechanistically, the production of CRAMP by β cells was controlled by SCFAs produced by the gut microbiota. Fecal transplant of gut microbiota from male NOD mice to female NOD recipients increases CRAMP production and reduces the incidence of diabetes, indicating a causative effect of male gut microbiota in controlling autoimmune diabetes.

Metabolites: retinoic acid

Retinoic acid, a metabolite of vitamin A, has a wide range of biological activity, including regulating immune responses.⁷⁴ A major part of retinoic acid's anti-inflammatory effects depends on the inhibition of T_H17 and promotion of FOXP3⁺ Treg responses.^74,75 AM80 is a synthetic retinoic acid that is characterized by higher stability and fewer potential adverse effects compared with all-trans-retinoic acid, one of the most active physiological retinoid metabolites.^{76, 77} It has been reported that retinoic acid and AM80 ameliorate many autoimmune responses, including experimental autoimmune myositis, experimental autoimmune encephalitis, and collagen-induced arthritis.^78–81 We recently showed that oral administration of AM80 inhibits autoimmune disease in joints as well as in lung.⁸² We elucidated a novel mechanism whereby AM80 suppresses the autoimmune pathology at both lung and joints by inhibiting T_FH and T_H17 responses. Specifically, AM80 increased the expression of the gut-homing integrin α₄β₇ on T_FH cells, which diverted T_FH cells from systemic (non-gut) inflamed sites, such as lung, into the gut (the non-immunopathological site) and thus reduced the systemic auto-Abs.

Metabolites: uric acid

It is worth mentioning that, although most studies have focused on the bacterial microbiota, the gut microbial communities also include fungi and viruses. One study has reported that Saccharomyces cerevisiae exacerbated intestinal disease in a mouse model of colitis and increased gut barrier permeability by enhancing host purine metabolism, leading to an increase in uric acid production.⁸³ Thus, fungi in the gut may be able to potentiate metabolite production that worsens the development of IBD.

Microbiota-mediated epigenetic modifications

One of the mechanisms by which the microbiota affects the host is by triggering epigenetic changes in host cells. This often occurs through histone acetylation controlled by microbial metabolites, such as SCFAs.^84–86 Administration of propionate rescued the T_reg population in GF mice, restoring percentage and numbers comparable to those in SPF mice, as well as increasing T_reg expression of FOXP3, TGF-β, and IL-10 in vitro. These effects were mediated by inhibition of histone deacetylases (HDACs) 9 and 6, which led to increased histone acetylation.⁸⁶ Similarly, butyrate was found by multiple groups to increase histone acetylation in order to promote peripheral or colonic T_reg differentiation.^84,85 Naïve CD4⁺ T cells cultured under T_reg-inducing conditions in the presence of butyrate exhibited increased histone acetylation at the promoter and conserved noncoding sequence 3 (CNS3) compared with those cultured without butyrate, and Rag1^−/− mice fed butyrate were protected against T cell–induced colitis in a T_reg-dependent manner.⁸⁵ In another study, Arpaia et al. found that butyrate inhibited HDACs in order to increase histone acetylation in DCs, enhancing their ability to promote T_reg differentiation. This effect was dependent on peripheral T_reg differentiation, as mice lacking CNS1—an intronic enhancer required for extrathymic but not thymic T_reg differentiation—in FOXP3-expressing cells failed to respond to butyrate.⁸⁴ These studies demonstrate that the commensal microbiota mediates protection against autoimmunity in part through histone modifications that promote T_reg proliferation and differentiation. It is interesting to note that these modifications occur in multiple cell populations, including T_reg cells themselves as well as DCs, demonstrating the wide target range of microbiota by-products.⁸⁴

The microbiota causes other epigenetic modifications, such as alterations in DNA or histone methylation. The previously mentioned iNKT-enhanced inflammatory response to chemical-induced colitis and airway hyperresponsiveness in GF mice were due to increased DNA methylation at five CpG sites in the gene for CXCL16, which results in a significant increase in iNKT cells numbers.²⁰ Another group found that NK cells were actually less reactive to stimuli than their counterparts from SPF mice.⁸⁷ Furthermore, TLR ligand–induced type I IFN production, as well as IL-6, TNF-α, IL-12 and IL-18, were systemically impaired in DCs of GF mice or mice treated with broad-spectrum antibiotics, which was connected to the absence of trimethylation on histone H3, lysine 4 (H3K4) in the regulatory regions of these cytokine genes in DCs, presumably impairing their ability to prime NK cells. Thus, the microbiota epigenetically regulates gene expression to dampen autoimmune inflammation but increase the ability to respond to outside challenges.