NOD mice have a common structure to their gut microbiome that is independent of animal facility
It is well known that the gut microbiota of mice varies between animal facilities. Despite this, genetic effects dependent on host background can also be observed [19]. To provide evidence that genetics may influence the microbiome in T1D, we sought to determine whether commonalities existed in the gut microbiome of the T1D-susceptible NOD inbred mouse strain independent of the animal facility. To do this, we collected fecal samples from NOD mice bred in four facilities and from two NOD substrains (NOD/Lt and NOD/MrkTac originally derived from Jackson Laboratories and Taconic Farms, respectively). In addition, samples were collected from C57BL/6 (B6) mice bred in two facilities, and microbial profiling was performed by 16S rRNA gene sequencing. Differences in alpha diversity were observable between animal facilities; however, a consistent strain effect was not observed (Additional file 1: Figure S1A, B). A distinct difference between the microbiota composition in mice from the two genetic backgrounds was seen (Fig. 1a). Dimension reduction analysis by supervised sparse partial-least squares discriminant analysis (sPLS-DA) showed substantial overlap of all of the NOD groups which were clearly distinct from the B6 mice (Fig. 1b, PERMANOVA test p = 0.001). B6 mice were dominated by members of the genus Allobaculum while NOD strains by Lactobacillus (Fig. 1c), along with many significant changes in other less abundant taxa (Additional file 1: Figure S2). These data were used to identify seven genera which were the common members of the NOD microbiome present across ≥ 80% of the NOD mice (Additional file 1: Figure S3). By comparison, 15 genera were present in ≥ 80% of the B6 strain (Additional file 1: Figure S3). Although, it is possible that a mismatched sample size (n = 62 NOD and n = 25 B6) mice contributed to the difference in the number of “core” genera present in ≥ 80% of mice. We concluded that the NOD genetic background is associated with a distinct fecal microbiota to B6 mice.
As differences in the microbiota can be influenced by founder effects driven by caging, litter, or colony drift, we performed a cohousing experiment to test whether the NOD gut microbiota could be substantially altered by early exposure to a different microbiota. We housed pregnant NOD/Lt (BRF) mothers with two B6 females (BRF), which remained with the mother and litter until the pups were weaned. This allowed the pups to be exposed to a B6 microbiota during the critical period of initial colonization after birth. The NOD/Lt-cohoused mice were then separated at weaning and 16s rRNA gene sequencing carried out on fecal samples collected at 12 weeks of age. While we saw an effect of increased alpha diversity in cohoused NOD mice compared with non-cohoused controls (Additional file 1: Figure S1), the microbial profile remained remarkably similar to the non-cohoused NOD/Lt mice (Fig. 1d). The abundance of Allobaculum, the dominant genus in the B6 strain, did not significantly increase following cohousing (FDR = 0.706 control NOD vs cohoused NOD). Similarly, the abundance of the seven most dominant genera associated with the NOD genetic background did not significantly change following cohousing (Lactobacillus FDR = 0.301, S24-7 FDR = 0.301, Clostridiales FDR = 0.184, Lachnospiraceae FDR = 0.523, Ruminococcaceae FDR = 0.301, Oscillospira FDR = 0.080, and Ruminococcus FDR = 0.812 control NOD vs cohoused NOD). We concluded from these data that although the microbiota composition of NOD mice is influenced by housing conditions and early-life exposures to a degree, the overall community composition is constrained by the NOD genetic background.
NOD mice carrying protective alleles at T1D susceptibility loci Idd3 and Idd5 have a distinct microbiota to wildtype NOD mice
Our findings so far indicate that the gut microbiota composition of NOD mice from various colonies and housing conditions is restricted by genetic background. While many genetic loci may influence the microbiota profile, we were interested to test whether specific loci known to profoundly impact disease susceptibility contributed to shaping the gut microbiota. To do this, we compared wildtype NOD mice with congenic NOD mice carrying protective alleles at specific T1D susceptibility loci. Congenic NOD.H-2b mice have the NOD H-2g7 MHC alleles replaced with H-2b alleles and are completely protected from diabetes development [16]. Likewise, NOD mice carrying protective alleles of both the Idd3 (IL2) and Idd5 (Ctla4, Slc11a1, and Acadl) loci are also nearly completely protected from disease, whereas mice carrying the individual Idd3 and Idd5 regions are partially protected [17, 18]. Microbial profiling of the congenic NOD strains showed that while they still had similarity to the wildtype NOD strains compared to the highly distinct B6 microbiota, Idd3/5 mice had significant alterations in their microbiota composition (Fig. 2a). sPLS-DA multivariate analysis showed that NOD and Idd3/5 groups clustered distinctly (Fig. 2b, PERMANOVA p = 0.001). The microbiota composition differences included significant alterations in 14 taxa (Additional file 1: Figure S4) with the major drivers of differences between NOD/Mrk and Idd3/5 mice shown in Fig. 2c. Partially disease-protected Idd3 and Idd5 strains had smaller but also significant changes compared with control NOD mice, with Idd3 having two and Idd5 having ten taxa with FDR < 0.05 (Additional file 1: Figure S4). The Idd3/5 fecal microbiota had 11 genera common to ≥ 80% of mice with a switch in the relative abundance of Lactobacillus and S24-7 compared to NOD along with the addition of Bacteriodes, Parabacteroides, Prevotella, and 5-7N15 (Additional file 1: Figure S3A).
Introduction of protective MHC alleles in NOD.H-2b mice resulted in relatively small overall changes to the microbiota (Fig. 2a). sPLS-DA analysis (Fig. 2d) separated NOD.H-2b and NOD mice with the variance predominately explained by a significant decreased abundance of Ruminococcus (FDR = 0.04) and non-significant alterations in a number of other taxa (Additional file 1: Figure S5). The overall difference in variance between NOD and NOD.H-2b mice was not significant by PERMANOVA test (p = 0.3). The NOD.H-2b fecal microbiota had six members common to ≥ 80% of mice and was very similar to NOD (Additional file 1: Figure S3). We also compared the common genera from the NOD, B6, Idd3/5, and NOD.H-2b strains together by sPLS-DA (Additional file 1: Figure S3B). This clearly showed complete overlap between the NOD and NOD.H-2b while B6 and Idd3/5 separate from NOD concordantly on component 1, with the drivers shown in Additional file 1: Figure S3C. We concluded that Idd3/5 risk alleles contributed to maintaining the common gut microbiota features of NOD mice while the MHC alleles had only a minor effect.
NOD mice have increased inflammation in the ileum and colon compared with disease-protected Idd3/5, NOD.H-2b, and C57BL/6 mice
We hypothesized that the changes we observed in the gut microbiota of disease-protected congenic NOD mice were due to genetically driven changes in the gut immune environment. To this end, we examined histological sections of the ileum, cecum, and colon for signs of inflammation. NOD mice did not have any signs of overt tissue damage or colitis. However, the colon and ileum lamina propria of NOD mice contained an increased frequency of polymorphonuclear leukocytes (PMNs), neutrophils, and lymphocytes compared with disease-protected B6 mice (p = 0.0001, Fig. 3a, b). This infiltration was significantly reduced relative to NOD in both NOD.H-2b (p = 0.0001) and Idd3/5 mice (p = 0.0001). Partially disease-protected Idd3 and Idd5 strains had a similar overall inflammation score to NOD mice (not significant, Fig. 3b). We concluded that autoimmune-susceptible NOD mice have low-level inflammatory infiltration of the colon and ileum.
Goblet cell mucous production is reduced in NOD mice compared with disease-protected Idd3/5 and C57BL/6 mice
Goblet cell mucous production is essential for maintaining an effective barrier between the host and gut bacteria [20]. Immune cell-produced regulatory factors such as IL-10 and IL-22 promote mucous production and protect from colitis [21, 22]. The mean area of goblet cell mucous staining was quantified in the colon (Fig. 3a, c). Compared with disease-protected B6 mice, NOD mice had far fewer periodic acid-Schiff (PAS)-stained goblet cells (p = 0.0001), and these did not extend as far down the length of the crypt. Although both Idd3 and Idd3/5 mice had reduced goblet cell area compared to B6 mice, it was increased compared with NOD mice (p = 0.0112 and p = 0.0032). Goblet cell area in NOD.H-2b mice and Idd5 mice was not significantly different to wildtype NOD mice. Therefore, susceptibility alleles at the Idd3 (IL-2) locus are associated with decreased goblet cell mucous production.
Paneth cell production of antimicrobial peptides is reduced in NOD mice compared with disease-protected NOD.H-2b and C57BL/6 mice
Compared with B6 mice, Paneth cell morphology in NOD mice was irregular (Fig. 4a). In B6 mice, Paneth cell granules were small, numerous, and densely packed. In contrast, in NOD mice, the granules were large, vacuolar, and fewer in number. Idd3/5 Paneth cells looked similar to NOD; however, Paneth cell granules in NOD.H-2b mice were smaller and more numerous than NOD. In order to determine whether the production of antimicrobial peptides by Paneth cells was impaired in NOD mice, we stained for lysozyme P (Fig. 4b, c), confirming that Paneth cells were reduced in number and function in NOD mice compared with both NOD.H-2b (p = 0.0487) and B6 mice (p < 0.0001). Lysozyme P and cryptdin expression by Paneth cells were also significantly lower in NOD mice compared with B6 mice (p = 0.0199 and p < 0.0001). NOD.H-2b mice had intermediate cryptdin expression (p = 0.0061), though lysozyme P gene expression did not reach significance (Fig. 4d, e). We concluded that Paneth cell number and function is reduced in NOD mice, and the introduction of protective MHC alleles results in a partial recovery of Paneth cell function.
Defective regulatory environment within the NOD intestine
To investigate the immune cell types infiltrating the NOD colon, we performed FACS analysis of the lamina propria. The NOD colon had a significant increase in the proportion of CD45+ immune cells within the lamina propria compared with both B6 (p = 0.0065) and Idd3/5 mice (p = 0.0151, Fig. 5a). Within the CD45+ population, the proportion of T cells was similar (Fig. 5b). CD103+Foxp3+ regulatory T cells (Tregs) were significantly enriched within B6 compared with NOD CD4+ cells (p = 0.0054, Fig. 5c), although CD103− Tregs were not significantly altered (now shown). Within the antigen presenting cell populations, F4/80+ inflammatory macrophages were significantly increased in NOD mice compared with B6 (p = 0.0002, Fig. 5d). Two key regulatory cytokines that promote immune regulation and gut homeostasis are IL-10 and IL-22. Gene expression of both these cytokines was reduced in the ileum and trended to a decrease in the colon of NOD in comparison with B6 mice (p = 0.0268 and p = 0.0495, Fig. 5e, f). Idd3/5 mice had an intermediate level of IL-10 expression but similar IL-22 expression compared with NOD mice. These findings confirmed that NOD mice have increased infiltration of immune cells with an inflammatory rather than a regulatory phenotype in the colon compared with disease-protected B6 mice.
IL-2 therapy reduces the gut histological inflammation score in NOD mice and alters the gut microbiota
NOD alleles at the Idd3 locus lead to a deficiency in IL-2 production in NOD mice, resulting in defects in Treg function and loss of T cell tolerance [17, 23]. We have previously shown that NOD mice have a generalized reduction in the frequency of Tregs in the lymph nodes, which is contributed to both by Idd3 and Idd5 alleles [24]. To test whether long-term treatment with IL-2 would expand Tregs within the gut and reduce inflammation, NOD mice were treated with IL-2c or PBS once a week from 4 to 10 weeks of age. Treatment with IL-2c extends the in vivo half-life of IL-2 and favors expansion of Tregs [25]. In agreement with other reports showing that low-dose IL-2 treatment protects NOD mice from diabetes development, insulitis was significantly reduced in the IL-2c treated mice (p < 0.001, Fig. 6a). Treg expansion in the IL-2c-treated mice was confirmed in the spleen, mesenteric LN (MLN), PcLN, and colon lamina propria (p < 0.0001, Fig. 6b). Consistent with this, IL10 but not IL22 expression was increased in the lamina propria (Additional file 1: Figure S6A). The overall gut inflammation score of the IL-2c-treated mice was significantly reduced compared to PBS-treated mice (p < 0.0001, Fig. 6c), which was mainly due to decreased infiltration of PMN cells. Consistent with this, the absolute number of CD45+ hematopoietic cells within the colon lamina propria was also decreased (p = 0.0075, Fig. 6d). Goblet cell area was also notably increased in IL-2c-treated mice (p < 0.0001, Fig. 6e, f). Paneth cell morphology was not altered by IL-2c treatment (Additional file 1: Figure S6B).
In order to determine whether IL-2c treatment also altered the microbiota composition, fecal samples were collected at the end of the treatment period. sPLS-DA analysis showed that IL-2c- and PBS-treated mice had differences in their microbiota composition (Fig. 6g, PERMANOVA p = 0.041). The changes in the gut microbiota induced by the IL-2c treatment that was consistent across three experiments were driven by a significant reduction in Bacteroidales (FDR = 0.01) and Oscillospira (FDR = 0.04) and non-significant decreases in Rikenellaceae, Turicbacter, S24-7, and Mucispirillum and an increase in Bifidobacterium (Fig. 6h). In one of these experiments, the fecal microbiota was assessed both before treatment (at 4 weeks of age) as well as at the end of treatment (at 10 weeks of age). This analysis demonstrated that the microbiota profile of the two groups was identical before treatment and only diverged following IL-2c treatment (Additional file 1: Figure S6C). Together, these data indicate that IL-2 expands regulatory immune cell populations within the gut tissues, controlling underlying inflammation and influencing the composition of the microbiota.
T1D-protective alleles linked to the IL-2 signaling pathway are associated with alterations in Bacteroides, Lachnospiraceae, and Ruminococcaceae family members in a healthy human cohort
To explore whether similar effects to those we observed from Idd3/5 alleles on the gut microbiota could be mediated by T1D susceptibility alleles in humans, we utilized a cohort of healthy human samples from the TwinsUK study [26]. This study included matched stool microbiome and SNP genotyping data from 1392 individuals. We looked for associations between specific taxa and genotype at a number of T1D-associated SNPs linked to the IL-2 signaling pathway and Treg function. A combined genetic risk score was developed for SNPs in IL-2 interacting genes. The genes included were selected from IL-2 interacting genes determined by STRING network analysis (IL2, IL2RA, IL10, IL27, IL7R, and DGKA region SNPs); CTLA4 was included due to its overlap with Idd5 and role in Treg function and PTPN2 due to the described effects of T1D-associated PTPN2 SNPs on IL-2 signaling [27]. A combined risk score involving all SNPs within these genes was calculated for each subject based on their genotype at these loci and weighted using the odds ratio of each loci. The weighted risk score was then used to perform a linear regression to determine associations with the operational taxonomic unit (OTU) identified in these individuals. Although we were underpowered to identify significant associations after adjusting for false discovery rate, 38 OTUs were identified with an unadjusted p value < 0.01 (Fig. 7a, b and Additional file 2: Table S1). This analysis suggests that susceptibility T1D alleles within the IL-2/Treg pathway may be associated with a decreased abundance of a number of members of the Clostridiales, Bacteroides, Lachnospiraceae, Ruminococcaceae, and Rikenellaceae families, while only a few OTUs were positively associated with T1D risk. These associations overlapped with the effects of Idd3/5 alleles in the NOD mouse, with Lachnospiraceae members including Coprococcus, Bacteroides, and unclassified Clostridiales associated with protection. We then repeated this analysis using a risk score calculated from all known T1D-associated SNPs. This analysis found very similar associations with a decreased abundance of members of the Clostridiales, Bacteroides, Lachnospiraceae, and Ruminococcaceae families associated with disease risk including Faecalibacterium prausnitzii, which has known protective associations in other autoimmune diseases (Fig. 7c, d and Additional file 3: Table S2). These data indicate that protective alleles at T1D-associated risk loci may modify the composition of the gut microbiota in humans.