Intestinal microbiota sustains inflammation and autoimmunity induced by hypomorphic RAG defects

Analysis of intestinal pathology in Rag2^R229Q/R229Q (herein referred to as Rag2^R229Q) mice revealed different degrees of spontaneous colitis, wasting diarrhea, and rectal prolapse. The overall incidence of rectal prolapse in the Rag2^R229Q colony was 5% in mice by 24 wk of age. No rectal prolapse was detected in Rag2^+/+ littermates. Nonetheless, 70–80% of all mutant mice showed substantial thickening of the colon throughout their entire length and microscopic evidence of inflammation (Fig. 1 A). Histologically, colonic inflammation is characterized by crypt elongation, epithelial hyperplasia, and a large inflammatory cell infiltrate extending to the colonic lamina propria (LP), with occasional crypt abscesses. Remarkably, such morphological changes also affected the small intestine (SI) of the mutant mice with increased mucosal thickness and histological score and lower villus/crypt ratio (Fig. 1, B and C).

LP infiltrate consisted mainly of T lymphocytes (Fig. 1, A and B; and Fig. S1), particularly CD4⁺, with very few CD8⁺ T cells (Fig. 1 D). Correlating with mucosal thickening, absolute counts were three- to sixfold higher in the mutant mice than controls, in sharp contrast with the lymphoid depletion observed in the peripheral lymphoid organs. The proportion of intestinal Th17 cells was significantly higher in Rag2^R229Q mice than Rag2^+/+ mice, and mutants also had more IFN-γ–producing LP T cells compared with controls (Fig. 1, E and F). Furthermore, IFN-γ IL-17 double-producing T cells, frequently identified in the inflamed LP of patients and murine models of IBD (Annunziato et al., 2007; Ahern et al., 2010), were abundant in Rag2^R229Q mice (Fig. 1, E and F). To comprehensively characterize the gut inflammatory environment in the Rag2^R229Q mice, we measured the expression of several cytokines. Transcript levels of IFN-γ, IL-17A, IL-6, IL-2, IL-1β, IL-22, and TNF were markedly up-regulated in the Rag2^R229Q ileum and colonic tissues (Fig. 1 G). In contrast, IL-10 and IL-12p35 levels were significantly reduced, whereas levels of IL-4, TGF-β, and IL-23p19 were not markedly different (Fig. 1 G). Overall, these results indicate that in Rag2^R229Q mice, gut inflammation is mediated by Th1 and Th17 immune responses.

CD4 T cells from mesenteric LNs (MLNs) of Rag2^R229Q mice showed an activated memory phenotype (CD44^hiCD62L⁻) and expressed the gut-homing receptors CCR9 and α4β7 (Mora and Von Andrian, 2006) at higher frequency compared with the same T cell population from Rag2^+/+ mice (Fig. 2 A). Of note, we detected CCR9 and α4β7 expression in the splenic T effector cell compartment (Fig. 2 A), suggesting that gut inflammation resulted in the circulation of cells activated within the intestine of Rag2^R229Q mice. In this line, CCR9⁺ T lymphocytes are increased in the circulation of IBD patients (Papadakis et al., 2001).

To determine whether Rag2^R229Q CD4 T cells were sufficient to cause intestinal inflammation, total CD4⁺ T cells derived from either Rag2^+/+ or Rag2^R229Q MLNs were transferred into Rag1^−/− recipient mice. Analysis 10 wk afterward revealed signs of colitis in all recipients of mutant CD4⁺ T cells. Rag2^R229Q chimeric Rag1^−/− mice showed significantly greater weight loss, colon shortening, and inflammatory scores than Rag1^−/− mice receiving Rag2^+/+ CD4⁺ T cells (Fig. 2, B and C). Furthermore, MLNs and spleens in recipients of Rag2^R229Q T cells were enlarged (Fig. 2, C and D). Immunophenotype analysis revealed a marked expansion of activated T cells (CD44^hi, CD45RB^low, and CD25⁺) in the MLN (Fig. 2 E) with a Th1/Th17 phenotype, similar to that observed in the Rag2^R229Q mice at steady state (Fig. 2 E). When we examined the cytokine profile in the supernatants from cultured colonic tissue of transplanted mice, we found that the intestinal inflammatory response in the Rag2^R229Q chimeric Rag1^−/− mice is an exacerbated form of the disease described in Rag2^R229Q mice. Indeed, culture supernatants from the colon of Rag2^R229Q chimeras contained two- to fivefold increased levels of IL-2, IFN-γ, IL-17, and TNF than supernatants of corresponding cultures from Rag1^−/− mice transferred with Rag2^+/+ cells (Fig. 2 F). On the contrary, IL-4 and IL-10 levels were significantly reduced (Fig. 2 F). Overall, these results indicate that Rag2^R229Q CD4 T cells are able to transfer the disease into immunodeficient hosts.

We found that Foxp3⁺GITR⁺ T reg cells were abundantly present in the inflamed intestine of Rag2^R229Q mice. In fact, a significant increase in the frequency and number of T reg cells was detected in the SI and colonic LP (Fig. 3 A). In the MLNs, T reg cells were augmented only in frequency (Fig. 3 A), suggesting that T reg cell expansion is unique to the LP compartment. The gut mucosa is considered a primary site for extrathymic T reg cell generation. We therefore analyzed the expression of Neuropilin-1 (Nrp-1) as a marker of inducible T reg (iT reg) cells (Weiss et al., 2012; Yadav et al., 2012) in the gut and in secondary lymphoid organs of Rag2^R229Q mice.

The percentages of Nrp-1^low (iT reg) cells among GITR⁺ T reg cells were comparable between mutant and WT mice in both small and large intestines and higher than in MLNs (Fig. 3 B). The analysis of the spleens revealed that 30% of peripheral T reg cells in Rag2^R229Q mice were constituted by iT reg cells, a significantly higher fraction than in control mice (Fig. 3 B). This result could reflect the poor generation of natural T reg cells in the thymus of mutant mice (Marrella et al., 2007). Alternatively, iT reg cell maintenance could be favored in the periphery under homeostatic conditions.

Mutant T reg cells from MLNs expressed the CD103 marker, indicative of antigen-specific expansion and differentiation, similar to control cells, but displayed enhanced expression of the CCR9 receptor (Fig. 3 C). The analogous effector/memory–like phenotype was observed in the peripheral T reg cell pool of Rag2^R229Q but not of Rag2^+/+ mice, and a major fraction (30–55% vs. 9–15% in the Rag2^+/+ mice; P < 0.05) of these cells retained the CCR9 expression (Fig. 3 C), indicative of their gut tropism. Hence, T reg cells accumulate at the intestinal mucosal areas, although they fail to control the disease in Rag2^R229Q mice. Thus, we assessed the capacity of Rag2^R229Q mice to generate functionally competent T reg cells by the induction of oral tolerance (Dubois et al., 2009; Cassani et al., 2011). We used the well-established model of prevention of experimental allergic encephalomyelitis (EAE) that can be generated by orally supplementing mice with a peptide derived from myelin oligodendrocyte glycoprotein (MOG_35–55; Chen et al., 1994). EAE developed in the Rag2^R229Q mice supplemented orally with PBS, albeit at a lower extent than Rag2^+/+ mice (Fig. 4 A). However, orally administered MOG_35–55 prevented EAE only in Rag2^+/+ but not in Rag2^R229Q mice (Fig. 4 A). Th17 and Th1 cells are involved in EAE pathogenesis, and oral tolerance induction abrogates the generation of these pro-inflammatory T cells (Peron et al., 2010). Consistently, MOG-tolerized Rag2^+/+ mice showed lower frequencies of IL-17– and IFN-γ–producing cells compared with nontolerized controls (Fig. 4 B). On the contrary, mutant mice did not show a decrease in the proportions of Th17 or Th1 cells upon oral MOG_35–55 administration. In agreement, brain lymphocyte infiltration was reduced in the tolerized Rag2^+/+ but not in the mutant animals (Fig. 4 C). These results suggested that Rag2^R229Q iT reg cells could not control autoimmunity in this setting. To substantiate the hypothesis of functional impairment of Rag2^R229Q T reg cells, we assessed their immunosuppressive potential in the severe combined immunodeficiency model of colitis induced by CD4⁺ CD45RB^hi T cells (Powrie et al., 1994). Mice were sacrificed at week four after cell transfer, and IBD was assessed by weight loss (IBD mice: 21.1% ± 4.5; Rag2^R229Q mice: 21.1% ± 2.6; WT mice: 12.6% ± 4.6; n = 6–8) and histological examination of the colon. The disease severity was significantly attenuated when naive CD4 T cells were cotransferred with Rag2^+/+ T reg cells (Fig. 4 D). On the contrary, Rag2^R229Q T reg cells failed to protect lymphopenic mice from colitis development, despite the fact that they were readily detected in the gut tissues as in control mice (Fig. 4 D). In summary, these results demonstrate that T reg cell functionality is compromised in Rag2^R229Q mice.

To address whether the intestinal inflammation in Rag2^R229Q is a consequence of T reg cell functional impairment, we transferred sorted splenic CD25^hiGITR⁺CD4⁺ T reg cells from Rag2^+/+ (CD45.1) donors into Rag2^R229Q (CD45.2) recipients and followed them for 4 wk. Transferred T reg cells expanded well and were readily detectable in the gut LP of Rag2^R229Q mice (Fig. 4 E), indicating that T reg cells efficiently colonized the inflamed intestine. In the T reg cells transfer group, lymphocytic infiltration decreased both in the SI and colon (Fig. 4 F), resulting in a reduced inflammatory score (Fig. 4 G). Accordingly, Th1 and Th17 cell populations lowered in the gut of Rag2^R229Q mice receiving Rag2^+/+ T reg cells (Fig. 4, H and I). However, production of IL-10 and TGF-β did not change (Fig. 4 I). Overall, these data confirm that impaired T reg cell function plays a role in the development of intestinal inflammation in Rag2^R229Q mice.

Intestinal B cells are poorly represented in Rag2^R229Q mice. Peyer’s patches (PPs) are rudimentary, and few LP B220⁺ cells and IgA⁺ plasma cells could be detected (Fig. 5 A). In agreement, fecal IgA and IgM levels were markedly diminished (Fig. 5 B), indicating a general deficiency in B cell function at the intestinal mucosal interface. As a consequence, a significant decrease of IgA-coated bacteria was detected in the large intestine of mutant mice (Fig. 5 C). Mucosal IgA operates immune exclusion of potentially harmful antigens and pathogens and contains the gut microbiota (Macpherson et al., 2000; Macpherson and Uhr, 2004; Palm et al., 2014). Lack of IgA results in augmented serum LPS levels (Shulzhenko et al., 2011). Quantification of bacteria adherent to the ileal mucosa by real-time PCR (RT-PCR; Proietti et al., 2014) showed a higher prokaryotic/eukaryotic DNA ratio in Rag2^R229Q mice (Fig. 5 G) associated with increased serum LPS concentrations (Fig. 5 H), indicating that mucosal B cell deficiency in mutant mice resulted in enhanced mucosal colonization and peripheral translocation of bacterial products. To better evaluate the link between B cell deficiency and gut inflammation, we adoptively transferred naive CD45.1⁺B220⁺IgM⁺ WT B cells into adult Rag2^R229Q recipients (Fig. 5 D). As shown in Fig. 5 E, transferred B cells generated IgA-producing plasma cells rapidly, reconstituting the secretory IgA compartment in recipient mice to the control level. Bacterial translocation and intestinal inflammation were assessed 4 wk after the transfer. In agreement with a previous study reporting the expression of epithelial antimicrobial peptide RegIIIγ in response to mucosal IgA deficiency (Shulzhenko et al., 2011), we found that correction of IgA defect substantially reduced RegIIIγ mRNA level in the gut of Rag2^R229Q recipient mice, suggesting diminished microbial stimulation of the intestinal epithelium (Fig. 5 F). Accordingly, bacterial adherence to ileal mucosa was diminished, although not significantly (Fig. 5 G). These results suggest that low-affinity T cell–independent IgA responses, such as those generated in mutant mice in the absence of T cell help (Cassani et al., 2010a), are not sufficient to contain more invasive bacteria, which tightly adhere to epithelial cells in terminal ileum. Nonetheless, LPS translocation was visibly reduced in the recipients of WT B cells (Fig. 5 H). Interestingly, we found that decreased stimulation of the epithelium attenuated the inflammation in the colonic but not in the SI tissue (Fig. 5 I). Overall, these data suggest that B cells are required but not sufficient to maintain gut homeostasis.

Defects in the adaptive immune compartment have profound impact on gut microbial ecology (Fagarasan et al., 2002; Kawamoto et al., 2014). We therefore evaluated the microbiota composition in Rag2^R229Q mice and their cohoused WT littermates by 16S rRNA gene profiling. Overall, no marked variation was observed in the relative abundance of the dominant intestinal bacterial phyla (Firmicutes, Bacterioidetes, Actinobacteria, and Proteobacteria) between the two sets of mice (Fig. S2 A). However, we found that mutant mice had considerably less diverse bacterial communities at the family and genus levels compared with controls, as determined by the Shannon and Chao1 indexes of α-diversity (Fig. 6 A and not depicted). Moreover, hierarchical cluster analysis based on data of relative abundance of the bacterial families revealed that Rag2^R229Q and Rag2^+/+ mice formed approximate clustering according to their genotype (Fig. 6 B). Then, we observed that most of the differences in the relative abundances of single bacterial genera between Rag2^R229Q and Rag2^+/+ mice was due to a higher abundance in the group of WT mice (Fig. S2 B); in addition, we found that modified taxa belonged most frequently to the phylum Proteobacteria (Fig. S2 B). Finally, to identify the bacterial genera that better describe the difference between genotypes, we performed a partial least squares (PLS1) regression analysis, which takes into consideration the overall composition of the microbiota, using genotype as categorical responses and the relative abundance of bacterial genera as continuous predictors. In accordance to α-diversity analysis, PLS1 results confirmed the overall dearth of the microbiota richness in the Rag2^R229Q mice compared with Rag2^+/+ mice (Fig. 6 C). Furthermore, PLS1 analysis revealed that Rag2^R229Q microbiota is mainly described by several genera of the phylum Proteobacteria (Fig. 6 C), previously associated with chronic inflammatory condition in humans (Manichanh et al., 2012). Of note, segmented filamentous bacteria (Ivanov et al., 2009) were not present in our colonies, thus excluding their role in the Th17 polarizing gut environment of mutant mice (data not shown). Overall, these data indicate that intestinal inflammation in mutant mice is associated with important alterations of the microbial communities in the gut.

To directly explore the role of microbial flora in gut inflammation in Rag2^R229Q mice, we treated mice with broad-spectrum ABXs. A regimen of three ABXs was administered to 8–12-wk-old mice for 4 wk. We did not observe substantial changes in body weight between treated and untreated animals, thus ruling out potential toxicity related to the pharmacological treatment. In ABX-treated Rag2^R229Q mice, we found amelioration of gut inflammation associated with a resolution of lymphocytic mucosal infiltrates (Fig. 7 A) and diminished percentages and absolute numbers of IFN-γ– and IL-17–expressing CD4⁺ T cells in the LP and MLNs (Fig. 7, B and C; and not depicted). On the contrary, a minor reduction of these cell types was observed in the Rag2^+/+-treated group (SI Rag2^+/+ vs. Rag2^+/+ ABX: Th17, 4.6 ± 2% vs. 3.5 ± 2.18%; Th1, 1.01 ± 1.06% vs. 1.27 ± 0.52%; Colon: Th17, 7.4 ± 2.46% vs. 3.1 ± 1.89%; Th1, 4.9 ± 1.99% vs. 3.85 ± 2.06%). Of note, concurrent deficiency of MyD88 (and consequent lack of most TLR signaling) did not protect from bowel inflammation. Rag2^R229Q × MyD88^−/− mice showed waste disease and analogous accumulation of infiltrating inflammatory T cells to Rag2^R229Q littermates (Fig. 8, A and B). Of note, lack of MyD88 abolished Th1 but not Th17 intestinal immune responses in Rag2^R229Q mice, indicating that Th17 cell generation in mutants occurs independently from the TLR-MyD88–dependent sensing of commensal microbiota (Fig. 8, C and D).

Chemokines play a major role in recruiting immune cells to sites of injury. Consistent with the increased tissue infiltration and cytokine skewing, transcript levels of Th1 (Cxcl10 and Cxcl9)- and Th17 (Ccl20)-associated chemokines, as well as those of Ccl2 and Ccl5, were up-regulated in the bowels of Rag2^R229Q mice (Fig. 7 D). However, upon ABX treatment, their expression returned to control values (Fig. 7 D). Notably, analogously to Crohn’s disease (Papadakis et al., 2001), Ccl25 expression was lower in the Rag2^R229Q ileum with respect to control mice (Fig. 7 D), whereas it reached normal levels in ABX Rag2^R229Q mice (Fig. 7 D). Collectively, these results show that reducing microbial load beneficially affects Rag2^R229Q mice.

The same Th cells polarization dominating the intestinal immune response was evident in the spleens of mutant mice (Fig. 9 A). Hence, we examined the possibility that the antiinflammatory effect of ABXs extended beyond the intestinal mucosa. We found that frequencies of Th1 and Th17 cells were considerably reduced in the spleens of ABX Rag2^R229Q mice (Fig. 9 A), whereas their proportion was low and unaffected in ABX Rag2^+/+ mice, indicating a role for microbial flora in the maintenance of circulating, other than intestinal, effector Rag2^R229Q T cell populations. Interestingly, ABX Rag2^R229Q mice also showed significantly decreased frequencies of systemic CCR9⁺ CD4 T cells (Fig. 9 B), indicating that resolution of gut inflammation normalized the circulating pool of these cells.

To confirm the influence of microbial flora on the inflammatory T cell phenotype of Rag2^R229Q mice, independently from the host genetic susceptibility, we recolonized ABX-treated adult Rag2^+/+ mice with fecal materials obtained from mutant mice (Rag2^R229Q → Rag2^+/+). Fecal transplant was repeated twice 1 wk apart, and mice were sacrificed 3 wk later. Recolonized ABX Rag2^+/+ mice showed augmented frequencies of Th1 and Th17 cells in the gut LP as well as spleens with respect to ABX Rag2^+/+ mice receiving feces from Rag2^+/+ mice (Rag2^+/+→ Rag2^+/+; Fig. 10). On the contrary, ABX Rag2^R229Q mice recolonized with Rag2^+/+ feces displayed considerably fewer intestinal and peripheral Th1/Th17 cells compared with controls (Fig. 10). Like effector T cells, intestinal microbiota contributes to local induction of T reg cells (Round and Mazmanian, 2010; Atarashi et al., 2011), and ABX-treated mice had fewer intestinal T reg cells (Rosser et al., 2014). In agreement with these published data, we found that the proportion of GITR⁺ T reg cells was reduced in the LP and MLNs of ABX mice (not depicted). However, a drop of their splenic frequency was observed in the ABX Rag2^R229Q but not Rag2^+/+ mice (Fig. 9 C; Rag2^+/+ vs. Rag2^+/+ ABX: 11.7 ± 1.43% vs. 12.2 ± 5.36%), suggesting that most T reg cells in Rag2^R229Q mice are microbiota dependent. Overall, these findings point to a substantial role of gut microenvironment and commensal bacteria in the OS pathogenesis. Next, we assessed whether the ABX regimen protected against the multisystemic autoimmune-like manifestations of mutant mice. Indeed, Rag2^R229Q mice show important infiltrates of effector T cells in the skin, kidney, lung, and liver, which correlate with abnormal concentration of IFN-γ–induced CXCL10 in the serum (Cassani et al., 2010a; Marrella et al., 2012). Importantly, the decline of circulating inflammatory T cells in ABX Rag2^R229Q mice lowered IFN-γ and TNF serum concentrations (Fig. 9 D). We then examined the CD3⁺ T cell infiltrates in target organs. A decreased number of T cells was observed in the kidney, lung, and liver of ABX Rag2^R229Q mice compared with untreated controls (Fig. 9 E). In line with these results, tissue expression of inflammatory chemokines (Ccl2 and Cxcl10) was lower in ABX mutant mice (Fig. 9 F). On the contrary, skin inflammation was not affected by the ABX treatment (not depicted), suggesting that this district did not benefit from the reduced gut bacterial load and/or that an independent pathogenetic mechanism sustained tissue inflammation.

CD4 T cells have a pivotal role in B cell abnormalities, including plasma cell expansion and high IgE serum levels (Cassani et al., 2010a). Given that reducing microbial load in mutant mice had a clear impact on peripheral T cell phenotype, we asked whether ABX administration could impinge on B cell defects. In Rag2^+/+ mice, microbiota depletion reduced serum IgA levels, whereas other Ig isotypes remained unchanged (mean values of Rag2^+/+ vs. Rag2^+/+ ABX: IgA, 8.5 × 10⁵ vs. 2.2 × 10⁵ ng/ml; IgG1, 5.4 × 10⁵ vs. 3.6 × 10⁵ ng/ml; IgM, 2.2 × 10⁶ vs. 2.6 × 10⁶ ng/ml). Remarkably, in Rag2^R229Q mice, the long-term treatment resulted in a significantly reduced serum IgE (Fig. 9 G). A similar trend was observed for IgA and IgG, but not for IgM (Fig. 9 G). Of note, frequencies and numbers of peripheral B cells were not affected by ABX treatment. Because hyper-IgE constitutes a hallmark of OS, these results indicate that microbiota can constitute a pharmacological target for this pathological trait of the disease.

Here, we investigated the role of intestinal immune responses and gut microbiota in the pathogenesis of OS. A large fraction of Rag2^R229Q mice developed bowel inflammation characterized by marked infiltration of T cells in the LP. We found that lymphocytes are sufficient for disease initiation. In fact, CD4 T cells, either isolated from colitic mice or not, were able to transfer the disease into immunodeficient hosts. Cytokine production profiling of gut infiltrating lymphocytes revealed a dominant mixed Th1/Th17 skewing, frequently implicated in the pathogenesis of IBD and experimental colitis (Powrie et al., 1994; Berg et al., 1996; Fuss et al., 1996; Ahern et al., 2010). An abundant presence of Th17 cells, also producing the IL-22 cytokine, might also contribute to the aberrant expression of antimicrobial peptide RegIIIγ (Liang et al., 2006).

Multiple studies revealed the importance of T reg cells for the maintenance of immune tolerance in the gastrointestinal tract owing to the constant immune stimulation by commensal flora and food antigens (Izcue et al., 2009; Pedros et al., 2016). Impaired intestinal tolerance is believed to contribute to inflammation in response to microflora in IBD (Xavier and Podolsky, 2007). Several experiments in this study addressed the role of T reg cells in the gut homeostasis of Rag2^R229Q mice. We showed that T reg cells accumulate in the intestine but fail to control disease in mutant mice. In the adoptive transfer model of colitis, Rag2^R229Q T reg cells cotransferred with WT naive CD45RB^hi CD4 T cells did not prevent disease development in Rag1^−/− mice. Interestingly, similar failure in controlling ongoing autoimmune responses has been described for Nrp-1^low iT reg cells generated in inflammatory and lymphopenic environment (Yadav et al., 2012). On the contrary, the transfer of a limited number of Rag2^+/+ T reg cells significantly ameliorated intestinal inflammation by dampening Th1 and Th17 immune responses in Rag2^R229Q hosts. Thus, it is conceivable that the absence of functional T reg cells plays a major role in the pathogenesis of intestinal inflammation in Rag2^R229Q mice.

Distinct effector mechanisms from those operating in the Rag2^R229Q T reg cells may be required to control mucosal inflammation. These may include the expression of CD62L and CCR7 (Fu et al., 2004; Schneider et al., 2007), which are necessary for homing to the inductive sites, and production of IL-10, which is essential to limiting the induction of pro-inflammatory Th1 and Th17 cells at the intestinal mucosa interface (Geuking et al., 2011). IL-10 mRNA levels were significantly reduced in the intestines of Rag2^R229Q mice, and IL-10 is also required for oral tolerance induction in different models (Cong et al., 2004; Navarro et al., 2011). Finally, iT reg cells derived from an oligoclonal population of CD4⁺ T cells may be restricted in their ability to effectively suppress autoimmune disease and maintain tolerogenicity to microbiota (Huter et al., 2008; Nishio et al., 2015).

Mucosal T reg cells, together with secretory IgAs, critically preserve the gut barrier and control diversification of commensal species, thus ensuring host–bacterial mutualism (Kawamoto et al., 2014). Conversely, biased expansion of certain bacterial species impairs epithelial integrity and induces hyperactivation of the immune system, also promoting the generation of pro-inflammatory T cell subsets (Kawamoto et al., 2012; Kamada et al., 2013). We demonstrated that the hypomorphic Rag defect in B cells favors microbial access to the LP and circulation and considerably impacts on the diversity of the microbial communities in the gut. Our findings correlate with previous observations in mice harboring immune deficiencies affecting inflammatory responses in adaptive immune cells (Garrett et al., 2007; Kawamoto et al., 2014). We also showed that such altered microbiota from Rag2^R229Q mice was sufficient to promote a skewed Th1/Th17 pro-inflammatory phenotype into Rag2^+/+ mice. This observation indicates that Rag2^R229Q microbiota has a pathogenic potential.

The clinical benefits deriving from long-term ABX dosing formally implicates intestinal microbes in the disease pathogenesis. We showed that both mucosal and circulating pro-inflammatory Th1 and Th17 cell subsets were significantly reduced in ABX-treated mice, establishing that skewed polarization of adaptive immunity in Rag2^R229Q mice was microbiota dependent. Of note, CCR9-expressing T cells were similarly affected by the treatment, consistent with resolution of gut inflammation. However, because CCR9 expression is acquired by T cells upon priming by GALT-derived dendritic cells (Mora, 2008), we propose that its presence on the vast majority of circulating T cells in Rag2^R229Q mice might reflect their original intestinal differentiation.

Although CD4⁺ T cells from Rag2^R229Q mice are able to transfer colitis to immunodeficient hosts, it still remains to be determined whether an aberrant non–T cell population is required for the priming of dysregulated T cells. On this point, it has been shown that both innate and adaptive immune system stimulation by microbiota are required for driving spontaneous T cell proliferation and induction of colitis in immunodeficient mice (Feng et al., 2011). Furthermore, we previously reported perturbed cell trafficking and altered lymphoid distribution of dendritic cells in Rag2^R229Q mice, which tend instead to ectopically accumulate and halt at the environmental interfaces (Maina et al., 2013). It seems therefore logical to envisage a crucial role for innate immune cells in colitis development in mutant mice.

In contrast to previous publications describing its protective effects in different models of IBD induced by immune deficiencies (Rakoff-Nahoum et al., 2006; Rivas et al., 2012), we found that inhibition of MyD88-dependent pathways did not restrain intestinal inflammation in Rag2^R229Q mice. Rather, Rag2^R229Q × MyD88^−/− mice displayed worsening of wasting syndrome and unaffected accumulation of mucosal lymphocytic infiltrates. These results, however, are consistent with the intricate effect of MyD88 deficiency on host–microbial mutualism (Slack et al., 2009; Vaishnava et al., 2011). Interestingly, the absence of MyD88 signaling uncouples, in our model, the molecular requirements for generation of Th1 and Th17 cells in the gut. Despite observations that MyD88 ablation is required for the induction of both Th1 and Th17 cell responses (Fukata et al., 2008; Schenten et al., 2014), our data showed that signaling through MyD88 is crucial for Th1 immune responses yet has a nonessential role for generation of Th17 cells. Therefore, multiple sensors may be involved in the development of these cells in the gut. On this line, it has been suggested that bacterially produced ATP can directly augment intestinal Th17 cell differentiation (Atarashi et al., 2008).

In addition to gut inflammation, reducing intestinal microbial load visibly ameliorated multisystem autoimmunity in Rag2^R229Q mice, confirming the critical role of gut flora in sustaining chronic immune inflammation and autoimmune disease onset/progression in genetically susceptible hosts. The evidence that this inflammatory immune phenotype is transmissible to an intact immune system in a microbiota-dependent manner leads us to speculate that pharmacological gut decontamination might constitute a valid strategy to limit side effects of hematopoietic stem cell transplantation in these patients (Vossen et al., 2014). In contrast, given the importance of the intestinal microbiota, interventions aimed at promoting intestinal diversity may improve immunological outcomes in OS transplant settings. Replenishment of commensal bacterial populations through fecal microbiota transplantation or targeted flora reintroduction may theoretically be effective, although these procedures still need to be carefully evaluated in a transplant context. Alternatively, the use of dietary supplements, which may help to maintain beneficial bacteria (Gerbitz et al., 2004), might be considered to avoid episodes of dominance of harmful bacterial species during posttransplant recovery (Lane et al., 2015) and to promote healthy mucosal responses.

Importantly, our results showed that reducing the microbial load is beneficial for normalizing IgE levels in Rag2^R229Q mice. Because our previously published data indicated that T cells in mutant mice do have a crucial role in B cell abnormalities (Cassani et al., 2010a), diminished IgE levels after ABXs might reflect the lower serum concentration of cytokines promoting class switch recombination. Interestingly, it has been recently proposed that IgE production is strongly dependent on microbial exposure during early life and that a critical level of microbial diversity is required to establish the proper immunoregulatory networks to inhibit IgE production (Cahenzli et al., 2013). Thus, it is conceivable that reduced microbiota diversity in Rag2^R229Q mice also impinges the immune regulation that functions to maintain IgE levels at the baseline.

In summary, our findings might provide novel insights into the pathogenesis of OS. We showed in a murine model that hypomorphic Rag mutation compromises gut barrier integrity and alters microbiota composition. Furthermore, the loss of T cell tolerance to commensal microbiota leads to gut Th17/Th1 inflammation which, in turn, sustains dysbiosis in a well-recognized regulatory loop.

Rag2^R229Q mice represent an important tool to investigate whether intestinal microbiota can be manipulated therapeutically to dampen immune activation and perpetuation of inflammation. Furthermore, they may help to determine the extent to which a limited adaptive immune system also controls microbial community composition in other organ systems, such as the skin, and to clarify whether these other resident communities of indigenous microorganisms shape local immune cell differentiation and similarly influence the host health.

129Sv/C57BL/6 knock-in Rag2^R229Q mice were backcrossed to C57BL/6 strain for four generations. The colony was maintained onsite with heterozygous breeders, and littermates were kept in the same cages until weaning at 4 wk of age. Rag2^R229QMyd88^−/− double-mutant mice and littermates controls were generated from heterozygous–heterozygous crossing and maintained onsite as described for single mutants. All mouse strains were used for experiments between 8 and 13 wk of age, and, unless indicated otherwise, adult littermates were raised in separate cages according to the genotype after weaning. Rag1^−/− (Rag1^tm1Mom/J) mice were obtained from The Jackson Laboratory and C57BL/6 Ly5.1 mice from Charles River. Mice were maintained under specific pathogen–free conditions at the Istituto Clinico Humanitas. To deplete gut flora, adult Rag2^+/+ and Rag2^R229Q mice, housed in separate cages, were treated daily with an ABX cocktail containing 2.5 mg metronidazole, 2.5 mg ampicillin, and 1.25 mg vancomycin in 200-µl/mouse doses by oral gavage for 4 wk. All animal procedures were performed according to protocols approved by the Istituto Clinico Humanitas and the Italian Institutional Animal Care and Use Committee.

Mice were fed daily 0.25 mg MOG peptide (MOG_35–55) via oral gavage for 7 d as described previously (Cassani et al., 2011). Mice were then injected s.c. in the flank with 100 µg MOG_35–55 in 0.15 ml PBS emulsified in an equal volume of CFA containing 4 mg Mycobacterium tuberculosis H37 RA/ml (Difco). Pertussis toxin (200 ng/mouse/injection; List Biological Laboratories) was administered i.p. at the time of immunization and 48 h later. Animals were scored for EAE as follows: 0, no disease; 1, tail paralysis; 2, hind limb weakness; 3, hind limb paralysis; 4, hind limb plus forelimb paralysis; 5, moribund.

CD4⁺ T cells were isolated by pooled spleens and MLNs of 10-wk-old Rag2^+/+ or Rag2^R229Q mice using CD4-specific magnetic beads via negative selection according to the manufacturer’s instructions (Miltenyi Biotec). Enriched CD4 T cells were subsequently sorted into naive CD4⁺CD25⁻CD45RB^hi and regulatory CD4⁺CD45RB⁻CD25^hi populations (>98% purity) using a cell sorter (FACSAria; BD). For colitis induction, 4 × 10⁵ WT CD4⁺CD25⁻CD45RB^hi naive T cells were i.p. transferred into Rag1^−/− mice alone or with 1.5 × 10⁵ CD4⁺CD25^hi T reg cells from Rag2^+/+ or Rag2^R229Q mice. Recipient mice were weighted twice a week and sacrificed at week four after the transfer, when IBD was diagnosed by severe weight loss.

Fecal pellets were suspended in filtered PBS, homogenized, and centrifuged at 700 g for 5 min to remove large particles from bacteria. Then, supernatants were collected and centrifuged at 8,000 g for 10 min to remove unbound Ig’s. The bacterial pellet was suspended in PBS 1% BSA and stained with PE-labeled anti–mouse IgA on ice for 30 min. Bacteria was suspended in 4% formalin overnight at 4°C for fixation and resuspended in DAPI/PBS for acquisition at a cell analyzer (FACSCanto II; BD) using forward scatter (FSC) and side scatter parameters in logarithmic mode. Data were analyzed with FlowJo software (version 7.6.5; Tree Star).

For evaluation of fecal Ig’s, feces were collected, weighed, and resuspended in 1 ml PBS containing 0.1% sodium azide and protein inhibitors. After overnight incubation, fecal pellets were centrifuged at 10,000 g for 15 min, and supernatants were collected. Fecal suspensions were stored at −20°C. Levels of fecal IgA and IgM were measured by a mouse IgA/IgM ELISA quantitation set (Bethyl Laboratories, Inc.) according to the manufacturer’s instructions. Serum levels of IgG1, IgA, and IgM were measured by a multiplex assay kit (Beadlyte Mouse Immunoglobulin Isotyping kit; EMD Millipore) according to the manufacturer’s instructions and run using a Bio-Plex reader (Bio-Rad Laboratories). Levels of IgE were measured in sera by ELISA assay (BD). Cytokine levels in sera were measured using a specific ELISA kit for mouse IFN-γ and TNF (R&D Systems). For cytokine determination, colonic tissues were incubated at 37°C for 48 h. Cultured supernatants were collected after 48 h, and cytokine concentrations were determined by ELISA (R&D Systems) and normalized for tissue weight.

Total RNA was extracted from 1 cm of ileal and colonic tissues. Tissues from intestines, kidney, lung, and liver were homogenized in PureZOL reagent (Bio-Rad Laboratories) using TissueLyser II (QIAGEN). RNA was extracted using the RNeasy Lipid Tissue kit (QIAGEN), and cDNA was synthetized using the High-Capacity cDNA Reverse Transcription kit (Thermo Fisher Scientific). Quantitative RT-PCR was performed on a CFX384 system (Bio-Rad Laboratories) with the SYBR green master mix (Thermo Fisher Scientific). For sorted cells, total RNA was extracted using the RNeasy Plus Micro kit (QIAGEN). RNA was retrotranscribed using the High-Capacity cDNA Reverse Transcription kit followed by a preamplification PCR using TaqMan PreAmp Master Mix (Applied Biosystems). Quantitative RT-PCR was performed using a RT-PCR system (ViiA 7; Thermo Fisher Scientific) with TaqMan probes (Table S1). The relative amounts of mRNAs were calculated as 2^−ΔCT and expressed as arbitrary units. Each sample was analyzed in duplicate, and the relative level of expression was determined by normalization to β-actin (Actb) ribosomal RNA.

The quantification of prokaryotic/eukaryotic ratios was performed as described previously (Proietti et al., 2014). In brief, the latest segment of terminal ileum was collected and digested in lysis buffer (10 mM Tris, 100 mM NaCl, 10 mM EDTA, 0.5% SDS, and 0.4 mg/ml proteinase K). Total DNA was extracted with the isopropanol-ethanol method. The microbial load was determined by RT-PCR (SYBR green) using specific primers for microbial 16s and genomic 18s.

Sera were collected from Rag2^+/+ and Rag2^R229Q mice in endotoxin-free conditions. LPS concentration was determined by Limulus amebocyte lysate assay (Lonza) according to the manufacturer’s instructions.

SIs and colons were flushed to remove contents, opened longitudinally, and cut into 3–5-mm pieces. PPs were removed from SIs. Tissues were washed in HBSS buffer (0.5 M EDTA, 1 mM DTT, 15 mM Hepes, and 10% FCS) under agitation for 20 min at 37°C to remove the epithelial layer. Intestine pieces were then Liberase-digested (0.15 mg/ml; Roche) for 30 min at 37°C. Cells were washed and passed through a 100-µm cell strainer. The cell suspension obtained was loaded into a 44/67% Percoll gradient (GE Healthcare) and centrifuged at 600 g for 20 min at room temperature. Lymphoid fractions were collected at the interphase of the Percoll gradient and used for flow cytometry stainings.

Single-cell suspensions were prepared from pooled MLNs harvested from Rag2^R229Q and Rag2^+/+ mice. Total CD4 T cells were negatively selected using the CD4 T Cell Isolation kit (Miltenyi Biotec) according to the manufacturer’s instructions (purity > 95%). Purified CD4⁺ T cells (4 × 10⁶) were then i.v. injected into Rag1^−/− mice. CD25^hiGITR⁺CD4⁺ T reg cells were sorted from splenic cell suspension derived from WT CD45.1 mice using the FACSAria cell sorter. Sorted cells (5 × 10⁵) were then injected i.v. into Rag2^R229Q mice. The recipient mice were analyzed 4 wk after the transfer. For B cell adoptive transfer, total CD19⁺ B cells were isolated from pooled spleens, MLNs, and PPs derived from WT CD45.1 mice with anti-CD19 magnetic beads via positive selection (Miltenyi Biotec). B220⁺IgM⁺IgA⁻ B cells were then sorted from the enriched CD19⁺ fraction (purity > 98%) using the FACSAria cell sorter. Rag2^R229Q mice were i.v. transferred with 30 × 10⁶ sorted B cells. Recipient mice were monitored for B cell reconstitution and were sacrificed 4 wk after the transfer.

2-µm-thick formalin-fixed and paraffin-embedded tissue sections from liver, lung, kidney, and gut were used for routine hematoxylin and eosin (H&E) staining and immunostains. In brief, sections were dewaxed and rehydrated, endogenous peroxidase activity was blocked by 0.3% H₂O₂/methanol for 20 min, and heat-induced antigen retrieval was obtained by microwave or thermostatic bath treatment using EDTA buffer, pH 8.0, followed by 1-h incubation with primary antibody rabbit anti-CD3 (1:100; Dako) or rat anti-FoxP3 (1:100; eBioscience). Signal was revealed by using Real EnVision Rabbit-HRP (Dako) or Rat-on-Mouse HRP-Polymer (Biocare Medical) detection system followed by diaminobenzidine and nuclei counterstained with hematoxylin. Digital imagines were acquired by a camera (DP70; Olympus) mounted on a microscope (Bx60; Olympus) using Cell^F Imaging software (Soft Imaging System GmbH). Quantification of CD3⁺ cells was performed on representative sections of liver, lung, and kidney acquired by Aperio ScanScope CS Slide Scanner (Aperio Technologies). Total area of the sections and number of CD3⁺ cells were obtained by using Aperio ImageScope Software, and the absolute number of CD3⁺ cells was expressed as the number of CD3⁺ cells/mm². The degree of inflammation of the gut (colon and SI) was double-blind graded using combined scores, including the grade of inflammation (grade 0, no evidence of inflammation; grade 1, low; grade 2, moderate; grade 3, high; grade 4, intense and diffuse level of inflammation), the structural changes of the glands (grade 0, absence; grade 1, focal; grade 2, partial; grade 3, diffuse level of structural changes), and the goblet cell depletion (grade 0, no loss; grade 1, low; grade 2, moderate; grade 3, diffuse level of goblet cell depletion). The cumulative total scores ranged from 0 to 10.

Villous height and crypt depth were randomly measured for 16 well-oriented villi of distal ileum per mouse and averaged. Data were analyzed using commercial image analysis software (ImageJ; National Institutes of Health).

Cell suspensions were obtained from intestines, spleens, and MLNs and stained in PBS 2% FCS with conjugated antibodies to the following markers: TCR-β (eBioscience), TCR-γδ (eBioscience), CD3 (eBioscience), CD45 (eBioscience), CD4 (eBioscience), CD8a (eBioscience), CD62L (eBioscience), CD44 (BD), CD45R (B220; eBioscience), CD357 (GITR; eBioscience), CCR9 (eBioscience), α4β7 (BioLegend), Foxp3 (eBioscience), IgA (eBioscience), IgM (eBioscience), CD25 (eBioscience), CD103 (eBioscience), CD45RB (BioLegend), and Nrp-1 (R&D Systems). Live/dead cell discrimination was obtained using the Aqua Dead Cell Stain kit (Invitrogen). For cytokine production, cells were stimulated for 4 h with 20 ng/ml PMA (Sigma-Aldrich) and 0.5 µg/ml ionomycin (Sigma-Aldrich). Golgi Stop (1,000×; BD) was added during the last 3 h of stimulation. Cells were fixed and permeabilized using intracellular fixation and permeabilization buffer kit (eBioscience) and stained for conjugated anti–IL-17A (eBioscience) and anti–IFN-γ (eBioscience). Flow cytometry data were acquired at FACSCanto II. Data were analyzed with FlowJo software (version 7.6.5).

For the evaluation of intestinal microbiota, weaned Rag2^R229Q and Rag2^+/+ littermates were cohoused in the same cages until analysis. The bacterial microbiota of 20 fecal samples from 10 Rag2^+/+ and 10 Rag2^R229Q mice was investigated by sequencing the V5-V6 hypervariable regions of 16S rDNA gene by using the Illumina MiSeq platform as described previously (Manzari et al., 2015). The obtained 250 × 2 paired-end reads were taxonomically profiled by applying the BioMaS pipeline (Fosso et al., 2015). Paired-end reads of each sample were merged into consensus sequences using Flash (Adobe; Magoč and Salzberg, 2011) and then dereplicated by Usearch (drive5; Edgar, 2010). The unmerged reads were trimmed of low-quality regions (Phred score cutoff of 25), and paired ends containing reads shorter than 50 nt were removed. Both the merged sequences and the unmerged reads were mapped against the RDP II database (release 10.32; Cole et al., 2009) by using Bowtie 2 (Langmead and Salzberg, 2012). The mapping data were filtered according to query coverage (≥70%) and similarity percentage (≥97%) and were stored in a file format suitable for the taxonomic assignment steps. Finally, the Tango tool processed the similarity analysis data to assign the sequences to a taxonomic clade in the NCBI taxonomy (Kim et al., 2013; Alonso-Alemany et al., 2014). Statistically significant differences of the bacterial population in Rag^+/+ and Rag2^R229Q mice were determined at family, genus, and species levels by using the R/Bioconductor DESeq package (Anders et al., 2012).

10–12-wk-old adult male Rag2^R229Q and Rag2^+/+ mice, housed in separated cages since weaning, received long-term (30 d) dosing of ABXs. For the fecal suspension preparation, pellets were collected from adult female Rag2^+/+ and Rag2^R229Q donors, suspended in sterile PBS (three to four pellets in 2 ml PBS), and homogenized. Then, homogenized pooled fecal samples were subsequently introduced by gavage with a flexible plastic tube into the stomachs of each ABX-treated mouse recipient. The fecal suspension was administered twice, 1 wk apart, in 200-µl doses. Transplant recipients were maintained in cages dedicated to mice colonized with the same donor microbiota, Rag2^R229Q or WT, until the sacrifice (n = 5–6 mice per cage per strain per donor microbiota sample per experiment). Recipient mice were analyzed 3 wk after the second microbiota transfer.

Results were analyzed using the nonparametric Mann-Whitney test, Student’s unpaired t test, and two-way ANOVA with Bonferroni post-test analysis. Results are presented as mean ± SEM. Values of P < 0.05 were considered statistically significant.

Fig. S1 shows the experimental strategy for FACS staining of intestinal LP cells. Fig. S2 includes graphs showing distribution of bacterial phyla and classes and heat map showing the relative abundance of gut bacterial genera discriminating Rag2^+/+ and Rag2^R229Q littermates. Table S1 lists PCR primers used for the experiments in this study.

The technical assistance of Enrica Mira Catò is acknowledged. We also thank Teresa De Filippis (University of Bari, Bari, Italy) and Caterina Manzari (Consiglio Nazionale delle Ricerche-Istituto di Biomembrane e Bioenergetica and Molecular Biodiversity Lab, Lifewatch, Bari, Italy) for high-throughput sequencing of metagenomic samples.

This work was supported by Ministry of Education, University, and Research grant MIUR-FIR RBFR12I3UB and Ministry of Health Giovani Ricercatori grant GR-2011-02349759 to B. Cassani, Ministry of Education, University, and Research grant MIUR-FIRB RBAP11H2R9-004 to P. Vezzoni, Ministry of Education, University, and Research grant MIUR PON01_02589 and Consiglio Nazionale delle Ricerche Medicina Personalizzata to G. Pesole and A.M. D’Erchia, San Raffaele Telethon Institute for Gene Therapy (TIGET) core grant A3 from the Italian Telethon Foundation to A. Villa and Fondazione Cariplo grant N-2012-0519 to A. Villa, P.L. Poliani, and F. Grassi, and Ministry of Health Giovani Ricercatori grant GR-2009-1607206 to R. Pedotti. B. Cassani was the recipient of a Gerry Scotti fellowship, and R. Rigoni was the recipient of a short-term European Federation of Immunological Societies fellowship.

The authors declare no competing financial interests.

Author contributions: B. Cassani and A. Villa designed research experiments. R. Rigoni performed most experiments. E. Fontana, V. Maina, M.C. Castiello, G. Pacchiana, and S. Musio performed experiments. E. Fontana and S. Mantero contributed histological analysis. B. Fosso, A.M. D’Erchia, G. Pesole, V. Taverniti, and S. Guglielmetti contributed sequencing and bioinformatic analysis of intestinal microbiota. C. Selmi, J.R. Mora, and R. Pedotti contributed reagents and intellectual input. P. Vezzoni, S. Guglielmetti, P.L. Poliani, and F. Grassi contributed intellectual input and data analysis and revised the paper. R. Rigoni, A. Villa, and B. Cassani analyzed data and wrote the paper.