Resetting microbiota by Lactobacillus reuteri inhibits T reg deficiency–induced autoimmunity via adenosine A_2A receptors

Regulatory T (T reg) cells maintain immune homeostasis and play a pivotal role in immune tolerance. Forkhead box protein 3 (Foxp3) is a major transcription factor that is associated with T reg cell development and function (Ouyang et al., 2010). Mutations or deletions of the Foxp3 gene result in IPEX syndrome (immunodysregulation, polyendocrinopathy, and enteropathy, with X-linked inheritance) in humans. IPEX syndrome is an autoimmune disease associated with eczema, severe enteropathy, type I diabetes, thyroiditis, hemolytic anemia, and thrombocytopenia (Bennett et al., 2001). Recently, a Foxp3 mutation has been also identified in a two-generation family with inflammatory bowel disease (IBD; Okou et al., 2014). several other gene defects that affect the function of T reg cells give rise to IPEX-like syndromes, including mutations or deficiency in the α-chain of the IL-2 receptor (CD25), signal transducer and activator of transcription 5b (STAT5b), Itchy E3 ubiquitin protein ligase (ITCH), STAT1, cytotoxic T-lymphocyte–associated protein 4 (CTLA4), and Wiskott-Aldrich syndrome (WAS; Massaad et al., 2013; Verbsky and Chatila, 2013).

The scurfy (SF) mouse, which bears a mutation in the Foxp3 gene, displays a similar clinical phenotype, with early onset dermatitis, progressive multiorgan inflammation, and death within the first month of life caused by a lymphoproliferative syndrome (Godfrey et al., 1991a,b; Sharma et al., 2009). The lethal lymphoproliferative syndrome has been shown to be predominately mediated by CD4⁺ Th1 and Th2 cell–induced pathology (Blair et al., 1994; Kanangat et al., 1996; Sharma et al., 2009, 2011; Suscovich et al., 2012), similar to what is seen in patients with IPEX syndrome (Zennaro et al., 2012; Baris et al., 2014). To date, treatment with immunosuppressive drugs in combination with supportive care, such as total parental nutrition (TPN) and blood transfusion, may help to palliate clinical manifestations (Hannibal and Torgerson, 2011). Transplantation of donor T reg cells and stem cells is promising, but the procedure is limited by the availability of a suitable donor; and the ultimate outcome can be fatal or associated with chronic health problems (Rao et al., 2007; Seidel et al., 2009; Burroughs et al., 2010; Nademi et al., 2014).

The intestinal microbiota drives host immune homeostasis by regulating the differentiation and expansion of T reg cells (Round and Mazmanian, 2010; Arpaia et al., 2013; Furusawa et al., 2013; Weng and Walker, 2013; Belkaid and Hand, 2014). Intestinal microbial dysbiosis can develop as a result of an abnormal diet, infection, inflammation, and altered host genetics (Lupp et al., 2007; David et al., 2014; Goodrich et al., 2014; Lukens et al., 2014). Gut microbial dysbiosis can lead to autoimmune diseases, including IBD, autoimmune arthritis, type I diabetes, and experimental autoimmune encephalomyelitis (EAE; Cerf-Bensussan and Gaboriau-Routhiau, 2010; Wu and Wu, 2012; Markle et al., 2013). However, the host–microbiota interactions in monogenic autoimmune diseases largely remain unknown. Therefore, we hypothesized that (i) T reg cell deficiency caused by a Foxp3 mutation disrupted the gut microbiota; and (ii) T reg cell deficiency–mediated autoimmune disease can be treated by targeting gut microbiota.

Probiotics have the capacity to not only induce large-scale changes in the host microbiota composition but also modulate the global metabolic function of intestinal microbiomes (Hemarajata and Versalovic, 2013; Sanders et al., 2013). Lactobacillus reuteri (L. reuteri) demonstrates beneficial attributes caused by a mutualistic relationship between microbe and host (Walter et al., 2011). L. reuteri DSM17938 is effective in treating and preventing diseases that affect infants and children, including necrotizing enterocolitis, diarrhea, and infantile colic (Urbańska and Szajewska, 2014). L. reuteri modulates the abnormal microbial communities associated with these diseases (Mai et al., 2006; Rhoads et al., 2009; Torrazza and Neu, 2013; Patel and Denning, 2015).

In this study, we characterize the dynamic changes of autoimmunity and gut microbial dysbiosis over the short lifespan of the Foxp3⁺ T reg cell–deficient SF mouse. We demonstrate a shift in resident microbiota after treatment with probiotic L. reuteri in SF mice, which consequently inhibits the degree of autoimmunity. Protective effects of a microbiota-modulated metabolite, inosine, against T reg cell deficiency–induced autoimmunity were identified. Our study highlights the immunomodulatory mechanism by which remodeling microbiota and inosine counteract T reg cell deficiency–induced autoimmunity.

Because human genetic factors can shape the gut microbiome (Goodrich et al., 2014), we initially explored whether gut microbiota would be altered by the Foxp3 gene mutation. We conducted a dynamic study of autoimmunity and gut microbiota in Foxp3⁺ T reg cell–deficient scurfy (SF) mice, comparing WT littermates to SF mice at day 8 (d8), d15, and d22 of age. We observed that SF mice develop chronic inflammation in several tissues (liver, lung, ear, and tail; Fig. S1, A–E) and acquire an increase the frequency of IFN-γ–producing CD4⁺ T (Th1) and IL-4–producing CD4⁺ T (Th2) cells in the spleen and mesenteric LNs (MLNs; Fig. 1 A and Fig. S1 G). Also, plasma levels of IFN-γ and IL-4 were increased as early as d8, at a time when a SF clinical phenotype had not been demonstrated (Fig. 1 B). However, Th17 cells were not increased in spleen and MLN of SF mice from d15 to d22 (Fig. S1, F and G), suggesting that Th1 and Th2 responses are the main drivers of autoimmunity in T reg cell–deficient SF mice.

Gut microbial dysbiosis developed over the first 22 d of life in SF mice (compared with WT littermates), as revealed by stool 16S rRNA gene sequencing analysis at d8, d15, and d22. First, Shannon α diversity was significantly decreased in SF mice compared with WT littermates at d22 (Fig. 1 C). Second, weighted UniFrac-based three-dimensional principal coordinates analysis (PCoA) revealed a distinctly shifted gut microbial composition between SF and WT (Fig. 1 D). Longitudinal analysis of the relative abundance of gut microbial communities in these mice indicated broad population changes from the phylum to the genus levels in SF compared with WT mice, suggesting that Foxp3⁺ T reg cell deficiency shapes gut microbial community structure (Fig. 1 E and Fig. S1 H). For example, the relative abundance of Lactobacillus was significantly lower in SF stool compared with that in WT stool at d8 (P < 0.001; Fig. 1 F). Bacteroides was significantly higher abundance in SF than in WT stool at d22 (P < 0.001; Fig. 1 F). In summary, whereas sampling across time, we observed that development of autoimmunity was accompanied by microbial dysbiosis.

We further examined whether gut microbiota shaped by T reg cell deficiency can be reprogrammed by oral administration of probiotic L. reuteri. Stool microbiota were analyzed when L. reuteri was orally fed by gavage, 10⁷ CFU/day, daily, starting on d8 for 2 wk (early treatment) or on d15 for 1 wk (late treatment). We found that L. reuteri given either early or later produced similar effects in resetting gut microbial dysbiosis in SF mice. The decreased Shannon α diversity associated with T reg cell deficiency was reversed by L. reuteri early or late treatment (P < 0.05; Fig. 2 A and Fig. S2 F). A three-dimensional PCoA and nonparametrical multiple dimensional scaling analysis revealed that SF mice with L. reuteri treatment displayed a shift in microbial community composition, which was distinct from either WT or SF populations (Fig. 2, B and C; and Fig. S2 G). According to our evaluation of predominant bacteria from the phylum to genus level, L. reuteri specifically increased the relative abundance of the phylum Firmicutes and the genera Lactobacillus and Oscillospira, and decreased the relative abundance of the phylum Tenericutes and the genus Bacteroides (Fig. 2, D–F; and Fig. S2, H and I). Further analysis at the species level of Lactobacillus showed that the relative abundances of L. reuteri and L. others significantly increased after oral feeding of L. reuteri to SF mice compared with that of SF mice without L. reuteri treatment (Fig. 2 G). These results indicated that gut microbial dysbiosis induced by T reg cell deficiency could be reprogramed by oral administration of L. reuteri.

Gut microbiota play an important role in regulating host immune homeostasis (Wu and Wu, 2012; Belkaid and Hand, 2014). To determine whether remodeling gut microbiota by L. reuteri could globally impact T reg cell deficiency–associated autoimmunity, the survival and inflammation of SF mice were evaluated after orally feeding L. reuteri, starting on d8 (early treatment, to infinity for survival; and to 2 wk for tissue analysis; Fig. 3 A). SF mice (as expected) died between d21 and d30 as a result of lymphoproliferative syndrome and severe multiorgan inflammation (Fig. 3, B–G). However, the survival rate of SF mice with L. reuteri early treatment was significantly increased to at least 125 d (Fig. 3 B). In addition, L. reuteri significantly reduced inflammatory infiltration measured by the mean area of lymphocyte infiltration in liver and lung (P < 0.001; Fig. 3, C and D). The frequency of IFN-γ–producing CD4⁺ T (Th1) and IL-4–producing CD4⁺ T (Th2) cells in the spleen and MLNs of SF mice were significantly reduced by L. reuteri treatment, respectively (Fig. 3, E and F). Furthermore, the levels of IFN-γ and IL-4 in plasma of SF mice were significantly decreased by L. reuteri treatment (Fig. 3 G).

We also tested the therapeutic effects of L. reuteri in SF mice, beginning once they demonstrated overt clinical symptoms, on d15; we continued L. reuteri or Lactobacillus acidophilus DDS (La DDS) as a control for 2 wk (late treatment; Fig. S2 A). La DDS has been shown to lack the properties of adherence to epithelial cells, induction of mucin expression by intestinal epithelial cells, inhibition of enteropathogenic E. coli epithelial cell adherence, and inhibition of NF-κB (Mack et al., 2003; Liu et al., 2012). We found that L. reuteri significantly increased the survival of SF mice (P < 0.0001; Fig. S2 B). Interestingly, the survival of SF mice was improved even after terminating oral feeding of L. reuteri on d29, indicating that L. reuteri may induce a persistent shift of gut microbiota in SF mice. In contrast, La DDS had a lesser effect on SF survival than L. reuteri (P = 0.0149 vs. SF mice; P = 0.0024 vs. SF+L. reuteri; Fig. S2 B). L. reuteri but not La DDS late treatment reduced inflammatory infiltration in liver and lung, inhibited Th1/Th2 cells in the spleen, and reduced plasma levels of IFN-γ and IL-4 (Fig. S2, C–E). Altogether, these results demonstrate that remodeling gut microbiota by L. reuteri suppresses autoimmunity mediated by T reg cell deficiency.

Given that metabolites of commensal bacterial metabolism play a key role in microbe–host interactions (Nicholson et al., 2012; Dorrestein et al., 2014; Lee and Hase, 2014), we analyzed metabolomic profiles of plasma and feces obtained from WT, SF, and SF with L. reuteri given before overt symptoms (early treatment; SFL). We measured 525 metabolites in the plasma of WT, SF, and SFL mice (Table S1). The PCoA and heat map of the metabolites indicate that T reg cell deficiency affected plasma metabolomic profiles, and that L. reuteri treatment had a significant impact on plasma metabolome associated with T reg cell deficiency (Fig. 4, A and B). T reg cell deficiency led to significant alterations in 51% (269/525) of all detected metabolites in plasma, whereas L. reuteri treatment changed 5.5% (29/525) of all detected metabolites compared with SF (Fig. 4 C and Table S1). We further studied 29 plasma metabolites that were significantly altered by T reg cell deficiency and focused specifically on which metabolites were restored to WT control levels by L. reuteri (Fig. 4 D). We observed that the purine metabolite inosine was decreased fivefold in SF, but was completely restored by L. reuteri treatment (Fig. 4, D and E). Other metabolites involved in inosine metabolism, including adenosine, hypoxanthine, and xanthine, were altered by T reg cell deficiency, but were only partially restored by L. reuteri treatment (Fig. 4 E and Fig. S3 A).

A total of 657 metabolites in feces of WT, SF, and SFL mice could be detected (Fig. S3 B and Table S2). We found that 11% (76/657) of all detected metabolites in feces were altered by T reg cell deficiency compared with WT, whereas L. reuteri treatment changed 6% (42/657) of all detected metabolites compared with SF (Fig. S3 C and Table S2). Notably, the levels of metabolites of inosine metabolism, including adenosine, hypoxanthine, and xanthine, were not changed by either T reg cell deficiency or L. reuteri treatment. In feces, adenosine levels trended to be low and inosine was significantly decreased in L. reuteri–fed mice (Fig. S3 D). Because plasma inosine levels reverted to normal, we postulated that the reduced inosine in feces might be related to the increased absorption. To address this hypothesis, we measured the villus height and crypt depth, and we examined the expression levels of the intestinal nucleoside transporters equilibrative nucleoside transporter 1 (ENT1) and concentrative nucleoside transporter 2 (CNT2; Ward et al., 2000; Okada et al., 2006). Indeed, we found that L. reuteri treatment was associated with improvement of villus height and increased expression of intestinal nucleoside transporters ENT1 and CNT2 (Fig. S4, A–D). The changes of inosine in plasma associated with T reg cell deficiency and reversed by L. reuteri suggest that the inosine levels modulated by gut microbiota may play an important role in the mechanism of action of this probiotic.

Inosine has been reported to have immunomodulatory effects on immune cells, such as macrophages (Haskó et al., 2004). It is necessary to know the effects of inosine on the differentiation and proliferation of T cells, because T reg cell deficiency results in an early immunopathology driven by Th1/Th2 cells and their associated cytokines in SF mice (Fig. 1, A and B; and Fig. S1 G). Naive CD4⁺ T cells isolated from spleens of C57BL/6J mice can differentiate into Th1 and Th2 cells in vitro in the presence of specific stimulatory/inhibitory cytokines (Zhou et al., 2009). Inosine was added during the process of cell differentiation to test whether inosine affects this process. Results indicated that the relative mRNA expression of IFN-γ was increased in naive T cells with Th1 differentiation media, from 10.3 ± 0.8 at 24 h to 78.3 ± 5.1 at 72 h culture, compared with Th0 in a medium without Th1 differentiation antibodies and cytokines (Fig. 5 A). Increased mRNA expression of IFN-γ was inhibited by inosine (Fig. 5 A). Inosine also inhibited mRNA expression of IL-4 in naive T cells with Th2 differentiation media, at both 24 and 72 h (Fig. 5 B). The frequency of IFN-γ–producing CD4⁺ Th1 cells and IL-4–producing CD4⁺ Th2 cells derived from naive CD4⁺ T cell differentiation and analyzed by flow cytometry also indicated that inosine inhibited Th1/Th2 differentiation (Fig. 5, C and D). However, it appears that inosine had no effects on the proliferation of B- and T-lymphocytes in vitro (Fig. S5, A–C).

Inosine is an agonist of adenosine receptors, which are required for the protective effects of inosine in vivo (Gomez and Sitkovsky, 2003; Haskó et al., 2004; Nascimento et al., 2010; Rahimian et al., 2010; da Rocha Lapa et al., 2013; Muto et al., 2014; Welihinda et al., 2016). One of the major pathways for T reg cells controlling T effector cells (Th1/Th2) is the adenosine A_2A pathway (Deaglio et al., 2007; Csóka et al., 2008; Antonioli et al., 2013). We explored which isoforms of adenosine receptors contribute to the inhibition of inosine on Th1/Th2 differentiation. We initially examined the expression levels of 4 adenosine receptors (A₁, A_2A, A_2B, and A₃) on CD4⁺ T cells, and we showed that Th cells mainly express A₁ and A_2A receptors. Receptor levels could not be changed by inosine treatment (Fig. S5, D and E). To further examine which adenosine receptors specifically mediate the effect of inosine, we isolated naive CD4⁺ T cells from mice genetically deficient in the A₁, A_2A, A_2B, or A₃ receptor (A₁^−/−, A_2A^−/−, A_2B^−/−, and A₃^−/− mice, respectively). Findings indicated that the inhibition by inosine of Th1/Th2 differentiation was absent in mice with A_2A receptor deficiency, but not with deficiency of A₁, A_2B, or A₃ receptors (Fig. 5, C and D). Thus, we concluded that the immunomodulatory effect of inosine was mediated by A_2A receptors.

We further confirmed that inosine increased the downstream mediator of A_2A signaling, intracellular cyclic adenosine monophosphate (cAMP). We found that inosine enhanced the cAMP level in WT Th1/Th2 cells, but not in A_2A^−/− Th1/Th2 cells (Fig. 5, E and F), further supporting the concept that inosine activates adenosine the A_2A receptor in Th1/Th2 cells. Altogether, these results provide strong evidence that inosine inhibits Th1/Th2 differentiation in an adenosine A_2A receptor–dependent manner.

T reg cells control Th1/Th2 mediated by A_2A receptors (Deaglio et al., 2007; Csóka et al., 2008; Antonioli et al., 2013). In this T reg cell–deficient model, inosine’s inhibition of Th1/Th2 by A_2A shown in vitro (Fig. 5) may have therapeutic value in T reg cell deficiency–mediated autoimmunity. We orally fed inosine daily to determine its effects on survival of SF mice (Fig. 6 A). We found that SF mice when fed inosine had significantly prolonged survival to a maximal lifespan of 125 d, compared with 30 d in SF mice without inosine treatment (Fig. 6 B). Tissue inflammation was significantly decreased by inosine treatment, as measured by mean area of lymphocyte infiltration in liver and lung (Fig. 6, C and D). Inosine reduced the frequency of IFN-γ–producing CD4⁺ T (Th1) and the frequency of IL-4–producing CD4⁺ T (Th2) cells in the spleen, as well as plasma levels of IFN-γ and IL-4 (Fig. 6, E and F).

To further determine the in vivo role of adenosine A_2A receptors, as a mediator of the beneficial effects of inosine, we treated SF mice with inosine and A_2A receptor–specific antagonist SCH58261 for 2 wk (SFIS) and examined the inflammatory biomarkers, and compared with the groups of WT mice with inosine (WTI) or inosine+SCh58261 (WTIS), and SF mice with inosine (SFI), or SCH58261 (SFS) or inosine+SCH58621 (SFIS). Our results indicated that in SF mice, SCH58261 itself has no effects on inflammation in SF, but SCH58261 reversed the antiinflammatory effects of inosine in liver and lung (Fig. 6, C and D), the percentage of Th1 or Th2 cells in the spleen, as well as plasma cytokine levels of IFN-γ and IL-4 (Fig. 6, E and F). Our data thus confirmed that the A_2A receptor is required for inhibition of inosine on autoimmunity in SF mice.

To further explore whether the adenosine A_2A receptors also play a critical role in beneficial effects of the L. reuteri–remodeled microbiota on autoimmunity, we fed SF mice with L. reuteri in combination with A_2A receptor–specific antagonist SCH58261 (Fig. 7 A). We found that the A_2A antagonist SCH58261 reversed the beneficial effects of L. reuteri–remodeled microbiota on autoimmune damage in liver and lung, the frequencies of Th1 and Th2 cells in the spleen, as well as plasma cytokine levels of IFN-γ and IL-4 in SF mice (Fig. 7, B and E), indicating that the adenosine A_2A receptor mediates a substantial proportion of the protection of L. reuteri–remodeled microbiota against autoimmunity in this model.

Functional T reg cells are of critical importance for the establishment and maintenance of self-tolerance and immune homeostasis. To control inflammation or allergy, T reg cells exert a dominantly negative regulation of other Th subsets (Th1/Th2), which are proinflammatory (Rudensky, 2011). In the absence or malfunction of T reg cells, T effector cell immunity drives inflammation and lethality in humans with IPEX syndrome or IPEX-like syndromes, and in Foxp3-deficient SF mice. Clinical management has not been updated since 1993 (Hannibal and Torgerson, 2011). The life expectancy of patients with IPEX syndrome without hematopoietic stem cell transplant rarely extends beyond infancy. This study aimed to define new therapeutic strategies by targeting microbiota to T reg cell deficiency–induced autoimmunity.

The gut microbiota is not only essential for the development and maturation of the immune system, but also is shaped by the complex host immune system. A critical role for T reg cell–mediated control of inflammation has been studied by using germ-free (GF) mice compared with specific pathogen–free (SPF) mice (Chinen et al., 2010). Their study demonstrated that T reg cell development and suppressive function showed little dependence on gut microbiota. However, in a T reg cell–depleted model (Foxp3-DTR), inflammation in the small intestine of SPF mice was more severe than in GF mice, as shown by significantly increased gut lymphocyte infiltration, decreased body weight, and increased % of IFN-γ–producing T helper cells, indicating that T reg cell deficiency–induced inflammation is related to gut microbiota (Chinen et al., 2010). Reduced diversity of gut microbiota has been shown in immunodeficient mice including mice lacking both B and T cells (Rag1^−/−) and in mice lacking B cells only (Ighm^−/−) or T cells only (Cd3e^−/−; Kawamoto et al., 2014). Indeed, our results also showed reduced gut microbial diversity and altered bacterial composition in T reg cell–deficient SF mice, compared with WT mice. It was noted that the microbiota in SF mice could be shifted by oral administration of L. reuteri to produce a distinct signature without reversion to that seen in WT mice. The mechanisms by which L. reuteri regulates the intestinal microbiota may be multifactorial, including observations that L. reuteri produces antimicrobial agents (e.g., reuterin) and other metabolites that suppress the growth of other microorganisms (Spinler et al., 2008); L. reuteri also competes for receptors and binding sites with other intestinal microbes on the intestinal mucosa (Collado et al., 2007). The gut microbiota drives host immune homeostasis by regulating the differentiation and expansion of T cells (Hooper et al., 2012). We found that the development of Th1/Th2-drived autoimmunity was accompanied by microbial dysbiosis over the lifespan of SF mice. When this gut microbial dysbiosis was remodeled by L. reuteri, the autoimmunity was inhibited, as indicated by prolonged survival, reduced multiorgan inflammation, and decreased Th1/Th2 cytokines in SF mice. We have previously observed the same effects of L. reuteri in the setting of experimental necrotizing enterocolitis (Liu et al., 2012, 2013).

Metabolites produced by bacteria promote or suppress immune cell functions (Roelofsen et al., 2010; Arpaia et al., 2013; Furusawa et al., 2013). Microbiota-modulated metabolites accompanying the introduction of L. reuteri into T reg cell–deficient mice may play a critical role in regulating immune responses. We discovered that the purine metabolite inosine is reduced in plasma by T reg cell deficiency and is completely restored by L. reuteri treatment. L. reuteri does not generate large amounts of purine/inosine in culture (unpublished data). Based on the evidence of recovery of plasma levels of inosine to the levels similar to WT, in association with decreased levels in stool of SF mice, we postulate that L. reuteri may promote inosine absorption in the intestine by improving overall gut health through multiple mechanisms (for example by improving villus length) and/or by modulating the gut microbial community. As shown, we measured the small intestinal villi in SF mice compared with SF mice after oral feeding and showed that orally feeding L. reuteri improves the length of villi and depth of crypts. The increased expression of ENT transporters after L. reuteri feeding would also be predicted to correlate with improved absorption.

In addition, the complete genome of L. reuteri contains the tRNA-specific adenosine deaminase gene (tadA), which is involved in the biological process of tRNA wobble adenosine-to-inosine editing (UniProt accession no. F8DLR1), and the gene LPXTG-motif cell wall anchor domain protein (HMPREF0538_20056) that belongs to 5′-nucleotidase family which could participate in the conversion of inosine monophosphate into inosine (UniProt accession no. F8DRN6). Therefore, we could not rule out that the gut environment activates these enzymatic functions to help generate more inosine to be absorbed (Muzny et al., 2011). Identity of enzymatic activities related to and direct measurements of inosine absorption will be further explored.

Previous studies demonstrated that inosine treatment reduces levels of inflammatory cytokines produced by LPS-stimulated macrophages in murine models of endotoxic shock (Haskó et al., 2000). Inosine also attenuates the course of chronic autoimmune inflammatory diseases including type I diabetes (Mabley et al., 2003b) and experimental colitis (Mabley et al., 2003a; Rahimian et al., 2010), in association with a reduction of the production of proinflammatory cytokines and chemokines. Recently, investigators demonstrated antiinflammatory effects of inosine in mouse models of pleurisy (da Rocha Lapa et al., 2012) and allergic lung inflammation (da Rocha Lapa et al., 2013). Our results confirm that inosine is sufficient to inhibit Th1/Th2 differentiation in vitro and autoimmunity in vivo by reducing Th1/Th2 cells and the associated cytokines in SF mice.

However, mechanisms of the immunomodulatory effects of inosine are poorly understood. We herein demonstrate that the inhibition of Th1/Th2 differentiation by inosine requires A_2A receptors. Mechanistically, the key to T reg cell suppression of T effector cells is the interaction between adenosine produced by T reg cells (mediated by a CD39–CD73 pathway) and the A_2A receptor (expressed on T effector cells; Antonioli et al., 2013). It has been reported that lymphocytes predominately express A_2A receptors (Gomez and Sitkovsky, 2003). However, during T reg cell deficiency in SF mice or human IPEX syndrome, T effector cells lose regulation by adenosine-A_2A mediated signaling. Our findings highlight the possibility that inosine or other A_2A agonists could be used therapeutically to control T effector cell-mediated autoimmunity during T reg cell deficiency conditions.

Fredholm et al. (2001) reported that adenosine is the natural ligand at all four adenosine receptors. However, inosine was a potent agonist for A₁ and A₃ receptors, but not for A₂ receptors. This group used in vitro techniques with CHO cells stably transfected with the human A₁, A_2A, A_2B, and A₃ receptors. The inosine concentrations they tested were between 0.03 and 30 µM. Jin et al. (1997) concluded that physiologically significant concentrations (10–50 µM) of inosine selectively activate A₃, but not A₂, receptors in mast cells. However, several other findings indicate the protective effects of inosine are mediated by interaction with adenosine A_2A receptors (Gomez and Sitkovsky, 2003; Mabley et al., 2008; Rahimian et al., 2010; da Rocha Lapa et al., 2012, 2013; Muto et al., 2014). Welihinda et al. (2016) recently reported that inosine is a functional agonist of the A_2A receptor. They aimed to solve why the immunomodulatory effects of inosine in vivo, which at least in part, are mediated via A_2A, seem to differ from levels needed for in vitro pharmacological effects at the A_2A receptor. The research group used a combination of label-free, cell-based, and membrane-based functional assays in conjunction with an equilibrium agonist-binding assay. It provided evidence for inosine engagement of the A_2A receptor and subsequent activation of downstream signaling events. EC₅₀ values were 12 (4–33) nM for adenosine and 259.9 (104–651) µM for inosine.

Compared with adenosine, inosine has a lower affinity interaction with the A_2A receptor, probably because the structure of inosine contains fewer hydrogen bonds than adenosine, which are responsible for the binding to agonist-binding pocket of A_2A receptors (Lebon et al., 2011; Xu et al., 2011; Welihinda et al., 2016). Therefore, a high concentration (approximate millimolar) of inosine was previously required in vitro for inosine to inhibit Th1/Th2 differentiation and to activate A_2A receptors and increase intracellular cAMP levels (Welihinda et al., 2016). Previous studies showed that inosine has no capacity to activate A_2A or displace the binding of a high-affinity agonist to A_2A (Fredholm, 1995; Jin et al., 1997; Fredholm et al., 2001), which may be caused by its low affinity for the A_2A receptor, to low levels of A_2A expression, or to differential cell types that affect the binding between inosine and A_2A receptors when studied using insufficient concentrations of inosine for those in vitro experiments. Methods focusing on target identification of bioactive small molecules will further provide tools to elucidate the molecular details of inosine–A_2A interaction (Futamura et al., 2013; Ziegler et al., 2013).

Our current study focused on the effects of inosine in T cells. T cells differ from mast cells, platelets, or macrophages in that T cells express A_2A as their main adenosine receptor (Csóka et al., 2008). A_2A knockout mice data strongly suggest the effects of inosine are dependent on the A_2A receptor on T cells. We believe our studies using knockout models provide more solid evidence than in vitro studies with A_2A antagonists.

We could not rule out that inosine modulates A_2A activation via the salvage pathway in vivo (Kolassa et al., 1977). In this pathway, inosine is converted to hypoxanthine, which can be converted into inosine 5-monophosphate by hypoxanthine-guanine phosphoribosyltransferase, to adenylosuccinate, and then to adenosine 5-monophosphate (AMP), finally generating adenosine. Adenosine could activate A_2A at very low concentrations, and then rapidly degrade into inosine due to its short biological half-life (∼10 s). Adenosine would be rapidly cleared in plasma in vivo (in ∼30 s (Möser et al., 1989), whereas inosine has a much longer in vivo half-life than adenosine (∼15 h; Viegas et al., 2000).

In clinical trials in humans, investigators have concluded that inosine, when given orally at 3,000 mg per day, was safe and well-tolerated (Schwarzschild et al., 2014). The dosages of oral inosine used in mouse models varied in different disease models, with the highest dose being 750 mg/kg/bw/day (Mabley et al., 2003b; Hou et al., 2007; Muto et al., 2014), and toxic effects were not demonstrated. We used a dosage of inosine of 800 mg/kg/day for our survival study because the phenotype and manifestations of IPEX in SF mice were severe. Drug dosages used in mouse studies can be translated into human dosages based on surface area (Reagan-Shaw et al., 2008); accordingly, the dosage of 800 mg/kg/bw/day used in mice would be approximately equivalent to 65 mg/kg/bw/day in humans. Our dosage is thus much lower than the dosage in the aforementioned safety clinical trial (Schwarzschild et al., 2014). Moreover, the inosine plasma level in SF mice is significantly lower than its level in WT mice, according to our plasma metabolomics analysis.

In conclusion, our work reveals that microbial dysbiosis may be implicated in the pathogenesis of T reg cell deficiency–induced autoimmunity and that microbiota remodeling greatly influences outcome in autoimmune diseases via inosine–A_2A receptor signaling, and we have summarized our findings in Fig. 8. Even though the data do not formally rule out noninosine-dependent effects of L. reuteri in suppressing inflammation in SF mice, this study may broaden the concept of how immunodeficiency diseases evolve during early development, with key contributions by the microbiota and their metabolites acting through receptors on immune cells. This proof-of-principle study will spur future application of probiotic L. reuteri, inosine, and A_2A receptor agonists in other autoimmune diseases, such as IPEX syndrome, many of which are linked to inborn immune defects (Uhlig et al., 2014).

WT C57BL/6 and heterozygous B6.Cg-Foxp3^sf/J mice were purchased from BioGaia AB and allowed to acclimatize for 2 wk before experimentation. SF mice with hemizygous B6.Cg-Foxp3^sf/Y were generated by breeding heterozygous B6.Cg-Foxp3^sf/J female to C57BL/6J male mice. Because the Foxp3 gene is on the X chromosome, in each litter of breeding pairs, all males are either SF used as the experiments or WT littermates as the controls. Adenosine receptor–deficient mice with B6 background, including A₁^−/−, A_2A^−/−, A_2B^−/−, and A₃^−/− knockout mice, were provided by M.R. Blackburn (The University of Texas Health Science Center at Houston McGovern Medical School, UT Health, Austin, TX). Animal numbers used in each group of different experiments are indicated in the figures and figure legends. All mice were housed in animal facility at UT Health. All experimental procedures were approved by the IACUC (protocol number: AWC-14-056).

Human breast milk–derived Lactobacillus reuteri DSM17938 (L. reuteri) was provided by BioGaia AB. Lactobacillus acidophilus DDS (La DDS) was provided by D.R. Mack (Children’s Hospital of Eastern Ontario, Ontario, Canada). L. reuteri was prepared as previously described (Liu et al., 2014). In brief, L. reuteri was anaerobic cultured in deMan-Rogosa-Sharpe (MRS) medium at 37°C for 24 h, and then plated in MRS agar at specific serial dilution and grown anaerobically at 37°C for 48–72 h. Quantitative analysis of bacteria in culture media was performed by comparing absorbance at 600 nm of cultures at known concentrations, using a standard curve of bacterial CFU/ml grown on MRS agar. Each male mouse was given either MRS as control, or L. reuteri (WTL or SFL) or La DDS (WTDDS or SFDDS) in cultured media (10⁷ CFU/day) by gavage, daily, starting from 8 d of age (d8; early treatment) or d15 (late treatment), until the date as indicated in Fig. 3 A and Fig. S2 A, respectively. For treatment with L. reuteri in combination with A_2A receptor antagonist SCH58261 (Sigma-Aldrich) to examine the immunological biomarkers, each SF mouse was orally administered 10⁷ CFU of L. reuteri and i.p. injected 2 mg/kg of SCH58261 daily (SFLS) or was i.p. injected 2 mg/kg of SCH58261 daily (SFS) from d8 to the date as indicated in Fig. 3 A, Fig. S2 A, and Fig. 7 A, respectively.

Inosine (Sigma-Aldrich) was dissolved in sterilized water at the concentration of 40 mg/ml. For determining the effect of inosine on autoimmunity in SF mice, 800 mg/kg of inosine per day was orally administered to WT (WTI), SF mice (SFI) from d8 until the date as indicated in Fig. 6 A. For treatment with inosine in the combination with SCH58261 to examine the immunological biomarkers, WT (WTIS) SF (SFIS) mice were orally administered 800 mg/kg of inosine and 2 mg/kg of SCH58261 i.p. once daily from d8 to the date as indicated in Fig. 6 A.

All tissues were fixed and processed by the Cellular and Molecular Morphology Core Lab (Texas Medical Center Digestive Diseases Center, Houston, TX) and stained with hematoxylin and eosin (H&E) for histological evaluation. The area of lymphocyte infiltration of liver and lung and the villus height and crypt depth of small intestine were measured using ImageJ (National Institutes of Health) morphometry software.

Single-cell suspensions from the spleen and MLNs were obtained by gently fragmenting and filtering the tissues through 40-µm cell strainers (BD) into MACS buffer consisting of PBS, 0.5% BSA (Hyclone Laboratories), and 2 mM EDTA (Lonza). For in vitro stimulation, cells were stimulated with PMA (50 ng/ml) and ionomycin (1 µg/ml) in the presence of Brefeldin A (5 µ/ml) for 4 h to analyze IFN-γ–producing (Th1) and IL-4–producing (Th2) CD4⁺ T cells by flow cytometry.

Naive CD4⁺ T cells were isolated from the spleens of 6–8-wk-old C57BL/6J or adenosine receptor knockout mice by magnetic cell sorting, using a naive CD4⁺ T cell isolation kit (MACS; Miltenyi Biotec). For Th1 differentiation, naive T cells (5 × 10⁵ cells/well) were plated in 24-well plates containing 1 µg/ml anti-CD3, 2 µg/ml anti-CD28, 20 ng/ml IL-2, 10 µg/ml anti–IL-4–neutralizing antibody, and 20 ng/ml recombinant mouse IL-12 in RPMI-1640 complete medium at 37°C for 5 d with or without 2 mM of inosine. For Th2 differentiation, naive T cells were cultured in the presence of 1 µg/ml anti-CD3, 2 µg/ml anti-CD28, 20 ng/ml IL-2, 10 µg/ml anti–IFN-γ neutralizing antibody and 10 ng/ml recombinant mouse IL-4 in RPMI-1640 complete medium at 37°C for 5 d with or without 2 mM of inosine. At day 5, the cells were stimulated with PMA and ionomycin in the presence of Brefeldin A. The detailed antibodies and cytokines used are listed in Table S3.

For evaluation of the purity of naive CD4⁺ T cells, after sorting, cells were stained using fluorescein-labeled CD44, CD45RB, CD4, and CD62L. For characterization of Th1 and Th2 cells, cells were surface-stained by fluorescein labeled–CD4 and intracellularly stained with IFN-γ for Th1 and IL-4 for Th2. Intracellular staining was performed with a fixation/permeabilization kit, according to the manufacturer’s protocol (eBioscience). The data from all samples were acquired on FACSCalibur (BD) and analyzed using FlowJo software (Tree Star). The detailed antibodies used are listed in Table S3.

Total RNA was extracted from treated cells and animal tissues by using TRIzol (Sigma-Aldrich) and RNAeasy Mini kit (QIAGEN), according to the manufacturer’s protocol. RNA (2 µg) was reverse transcribed using amfiRivert Platinum ONE cDNA Synthesis Master Mix (GEnDEPOT). Quantitative RT-PCR was performed using amfiQSYBR Green PCR Master Mix (GenDEPOT) with CFX96 RT-PCR detection system (Bio-Rad Laboratories). All qPCR primers used are listed in Table S4.

Splenic lymphocytes were seeded into 96-well plates at an initial density of 2 × 10⁴ cells per well and were incubated with different doses of inosine under control, 40 µg/ml of LPS or 10 µg/ml of phytohemagglutinin (PHA), respectively. After 96 h, cell viability was measured by TACS XTT cell proliferation assay kit (Trevigen, Inc.).

Differentiated Th1 or Th2 cells from naive CD4⁺ T cells isolated from WT and adenosine A_2A receptor knockout mice were seeded (10⁴ cells/well) into 96-well plates. Cells were incubated in the absence or presence of inosine (2 mM) or adenosine A_2A receptor agonist CGS21680 (300 nM) at 37°C. The cAMP levels in Th1 and Th2 cells were measured by cAMP-Glo assay kit (Promega) after treatment for 15 min.

Plasma cytokine levels of IFN-γ, IL-4, IL-2, IL-1β, and IL-10 were assessed using a mouse multi-spot proinflammatory panel kit from Meso Scale Discovery (MSD), according to the manufacturer’s protocol.

Feces from cecum to rectum of mice were collected. Stool DNA was extracted by using Quick Stool DNA Isolation kit (QIAGEN). The composition of the stool microbiota was analyzed using high-throughput sequencing analysis of PCR-amplified 16s rRNA genes as previous described (Gupta et al., 2013). Bacterial diversity, species composition, and abundance were assessed using QIIME (Caporaso et al., 2010).

Plasma and stool metabolites were measured by Metabolon Inc. A total of 657 metabolites in stool and 525 metabolites in plasma were detected by a nontargeted metabolomic analysis platform, including UPLC-MS/MS and GC/MS, respectively. The metabolomic data were analyzed by pattern recognition analyses (unsupervised principal component analysis and Heat-map), revealing the biochemical perturbations induced by T reg cell deficiency or L. reuteri treatment, as previously described (He et al., 2015).

Data are presented as mean ± SEM. Significance was determined using one-way ANOVA corrected for multiple comparisons with Tukey and Dunnett’s posttests, or two-way ANOVA for multiple comparisons with a Bonferroni test. Kaplan-Meier survival curves were graphed, and the comparison was analyzed using Logrank with χ² test. The statistical analysis was performed using Prism version 4.0 (GraphPad Software). Differences were noted as significant at P < 0.05.

Fig. S1 shows the dynamic changes of autoimmunity and gut microbiota over 22 d of life in SF mice. Fig. S2 shows that L. reuteri late treatment prolongs survival, inhibits autoimmunity, and modulates gut microbiota in SF mice, but L. acidophilus DDS does not. Fig. S3 shows that L. reuteri early treatment reprograms fecal metabolomic profiles in SF mice. Fig. S4 shows that L. reuteri early treatment improves villus height and decreases crypt depth, and increases the expression of nucleoside transporters in small intestine of SF mice. Fig. S5 shows that inosine does not inhibit the proliferation of B and T cells or change the expression of adenosine receptors in Th1/Th2 cells. Table S1 shows 525 plasma metabolites and their relative quantification in WT, SF, and SF with L. reuteri treatment (SFL) mice. Table S2 shows 657 fecal metabolites and their relative quantification in WT, SF and SF with L. reuteri early treatment (SFL) mice. Table S3 lists antibodies and cytokines used in the study. Table S4 lists sequences of qPCR primers used in the study. Tables S1–S4 are available as Excel files.

We thank Drs. Elizabeth Donnachie and Miguel Escobar (Gulf States Hemophilia Center at the University of Texas) for generously providing access to their BD FACSCalibur; Pamela Parsons (Cellular and Morphology Core Lab at Texas Medical Center Digestive Diseases Center) for histological technical assistance; Dr. Guoyao Wu (Animal Science at Texas A&M University) and Dr. Steve Stirdivant (Metabolon Inc.) for their assistance with the metabolomics data; Dr. Eammon Connolly (Biogaia AB, Stockholm, Sweden) for providing Lactobacillus reuteri DSM 17938; and Dr. David R. Mack (Children’s Hospital of Eastern Ontario, Ontario, Canada) for providing Lactobacillus acidophilus DDS.

This work was supported by National Institutes of Health/National Center for Complementary and Integrative Health grant R01AT007083, and, in part, by BioGaia AB (Sweden) Investigator Research Grant.

B. He, S. Roos, J.M. Rhoads, and Y. Liu have a patent application pending on use of inosine and A_2A agonists in IPEX syndrome. The authors declare no additional competing financial interests.

Author contributions: B. He, J.M. Rhoads, and Y. Liu conceived the project, designed the experiments, and wrote the manuscript. B. He, T.K. Hoang, T. Wang, F. Luo, J.G. Molina, T.H. Gomez, and Y. Liu. performed all experiments and analysis. J. Zhou and N. Tatevian performed and assisted the pathological analysis of tissue inflammation. M. Ferris, C.M. Taylor, X. Tian, and M. Luo performed stool microbiota analysis. X. Tian and S. Roos contributed to metabolomics analysis. J.M. Rhoads, M.R. Blackburn, D.Q. Tran, and Y. Liu guided experimental design and data interpretation and edited the manuscript.