Mice (female, mixed mice from different breeders chosen randomly) used in this study were kept in specific pathogen-free (SPF) facilities with a 12-hour-light/dark (light: 7 am - 7 pm) cycle at the Yale Animal Resource Center (YARC). B-cell–specific TLR9-deficient (Tlr9^{f l/fl} Cd19-Cre⁺, KO group) and the control (Tlr9 ^fl/fl Cd19-Cre⁻, Ctr group) NOD mice were generated as previously reported (8). NOD mice, Rag1 ^-/- NOD mice and BDC2.5⁺ NOD mice were originally obtained from the Jackson Laboratory and maintained at SPF animal room at YARC. All the mice received autoclaved food (Global 2018S, Envigo), bedding and hyperchlorinated (4–6 ppm) water ad libitum. Germ-free (GF) NOD mouse breeders were generously provided by Dr. Kathy McCoy (University of Calgary, Canada) and have been bred and maintained at the gnotobiotic facility of YARC on 12-h light/dark cycles. The use of mice in this study was approved by the Institutional Animal Care and Use Committee at Yale University.

Total splenocytes from BDC2.5⁺ NOD mice were injected (i.v.) into Rag1 ^-/- NOD mice (5 × 10⁶/mouse). The recipients were monitored for glycosuria every 3 days, for 75 days. To circumvent the circadian rhythmic effect on blood glucose in mice, tail vein blood was measured daily at 9 a.m. using a glucometer (Bayer Diastix). Hyperglycemia was confirmed by two blood glucose measurements, 24 h apart, of over 250 mg/dl (>13.9 mmol/L) (26, 27).

Intestinal permeability was assessed based on the permeability to 4kD-FITC-dextran in plasma (DX-4000-FITC) (46944; Sigma-Aldrich, St. Louis, Missouri, USA) as previously described (19). Briefly, after fasting for 12 h, the mice were administered with DX-4000-FITC (500 mg/kg body weight, 125 mg/mL) by oral gavage. After 4 h, blood samples were collected, from the retro-orbital vein at different time points, diluted with an equal volume of PBS (pH 7.4) and analyzed for DX-4000-FITC concentration with a fluorescence spectrophotometer (HTS-7000 Plus-plate-reader; Perkin Elmer, Wellesley, Massachusetts, USA) at an excitation wavelength of 485 nm and emission wavelength of 535 nm. A standard curve was obtained by diluting known concentrations of FITC-dextran with normal plasma diluted with PBS (1:3 v/v).

Rag1 ^-/- NOD mice were transplanted with the fecal microbiota of Tlr9 ^fl/fl Cd19-Cre⁺ (or control) NOD mice prior to transferring diabetogenic splenocytes. Pancreata were harvested from diabetic recipients and formalin-fixed and embedded in paraffin. Tissue sections were stained with hematoxylin and eosin. Insulitis was scored under light microscopy. Approximately 130 islets from 7 mice were individually scored blindly. The degree of lymphocyte infiltration was evaluated using a light microscope (Olympus, Ballerup, Denmark) and graded as follows: 0, no infiltration; 1, intact islets but with a few mononuclear cells surrounding the islets; 2, peri-insulitis; 3, islet infiltration <50%; 4, islet infiltration >50%.

Murine IL-10 from the serum samples and splenocyte culture supernatants were measured using the Mouse ELISA kit (BioLegend), following the manufacturer’s instructions. The serum samples were diluted 1x with PBS (with 5% BSA), whereas the culture supernatants were measured without dilution.

For depletion of endogenous commensal microbiota, Tlr9 ^fl/fl Cd19-Cre⁺ mice, the control (Tlr9 ^fl/fl Cd19-Cre⁻) NOD mice and Rag1 ^-/- NOD mice were treated with an antibiotic cocktail (containing 0.5 g/L vancomycin, 1 g/L ampicillin, 1 g/L metronidazole, and 1 g/L neomycin) added in drinking water for 3 weeks. To mask the bitter taste of the added antibiotic cocktail, we added 0.1% energy-free sweetener to the drinking water.

The solid luminal contents from the mouse small intestine (jejunum and ileum) and colon were gently squeezed out followed by carefully flushing the intestine with 10 mL of precooled PBS (pH:7.2). After vortex, the suspensions were centrifuged at 8000 g for 15 minutes at 4°C. To minimize protein degradation, the clear supernatant was immediately transferred to cryovials and stored at -80°C. Immunoglobulin assessement was performed promptly to ensure sample integrity.

Gut flush and serum samples were collected from the mice studied at termination. Antibody concentrations were determined by ELISA, using reagents purchased from Southern Biotech (Birmingham, AL), following the protocol previously described (28). Samples were diluted (gut flush 1:2–1:10, serum 1:100) before the antibody (IgA, IgG1, IgG2a and IgM) measurements. Antibody concentrations were converted according to standard curves.

Fresh fecal samples collected from the mice were resuspended in 300 μl Tris-EDTA buffer (10 mM Tris and 1 mM EDTA, pH8) containing 7.5 μl 0.5% SDS and 3 μl Proteinase K (200 μg/ml). The samples were then incubated at 37°C for 1 h followed by homogenization in a solution containing one volume of phenol/chloroform/isoamyl alcohol (25:24:1), 200 μl 20% SDS and 0.3 g zirconium silica beads, with a mini-bead-beater (BioSpec) for 2 minutes. Phenol/chloroform/isoamyl alcohol was then added to the samples prior to centrifugation (4°C, 12000 g, 15 mins). The upper aqueous layer, containing DNA, was transferred to a new tube. Bacterial DNA was subsequently precipitated with isopropanol, washed with 70% ethanol, air-dried, and resuspended in 100 μl of nuclease-free water.

A 466 bp length sequence located in 16S rRNA gene was used for quantitative PCR (qPCR) with the primers Uni331F (TCCTACGGGAGGCAGCAGT) and Uni797R (GGACTACCAGGGTATCTAATCCTG TT) (29). Briefly, qPCR was performed with an iCycler qPCR machine (BIO-RAD, Geneva, Switzerland) using iQ SYBR Green Supermix (170-8882AP, BIO-RAD, CA, USA). The qPCR conditions were 95°C for 3 min, followed by 40 cycles of 95°C for 15 s, 60°C for 60 s and the plates were read for 5 s at 80°C. The bacterial load in the fecal samples was determined by a two color Real-Time PCR detection system (MyiQ2, BIO-RAD, USA).

The V4 region of the bacterial 16S rRNA gene was amplified from each DNA sample by PCR using barcoded broadly-conserved primer pairs (5’-GTGCCAGCMGCCGCGGTAA-3’) and (5’-GGACTACHVGGGTWTCTAAT-3’). The PCR products were purified using gel extraction kits (QIAGEN) and quantified by a Nanodrop spectrophotometer. Equimolar amounts of each sample were pooled for pyrosequencing using the Ion Torrent Personal Genome Machine (PGM) sequencing system (Life Technologies). The sequencing results were analyzed with the Quantitative Insights Into Microbial Ecology (QIIME) software package (version 1.8) and UPARSE pipeline (version 7.0). Taxonomy assignment was performed at various levels using representative sequences of each operational taxonomic unit (OTU). All the samples were rarefied to 10,000 reads per sample for downstream analysis. Principal coordinate analysis (PCoA) of OTUs based on Bray-Curtis distance was performed using QIIME. The statistical significance was subsequently assessed using permutational multivariate analysis of variance (PERMANOVA) with 9,999 permutations and P values were adjusted for multiple comparison with Benjamini-Hochberg method (30).

Sparse partial least-squares discriminant analysis (sPLS-DA) models (31) were established to identify specific OTUs that contributed to the segregation of gut microbial structure in control and KO mice using “mixOmics (v6.3.1)” package (32) in R (v3.4.4). Centered log ratio (CLR) transformations of the relative abundance of OTUs were implemented in sPLS-DA models. The optimal classification performances of the sPLS-DA models were estimated by the perf function using 5-fold cross-validation with the smallest balanced error rate.

Fresh fecal samples, collected from 6~8-week-old Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice and the Tlr9 ^fl/fl Cd19-Cre⁻ NOD mice, were resuspended in sterile PBS (1 g/ml), and homogenized by vortexing vigorously for 30 secs. The samples were filtered through a 100 μm filter to remove large debris. Bacteria were re-suspended in sterile PBS and co-cultured overnight (10⁵∼10⁹ CFU) with 2.5 million total splenocytes from NOD mice. Stimulated splenocytes were further analyzed by flow cytometry after staining with different monoclonal antibodies conjugated with various flurochomes; all from BioLegend including CD1d (Clone: 1B1), B220 (Clone: RA3-6B2), CD5 (Clone: 53-7.3).

RNA from ileum and colon was extracted using Trizol reagent and a RNeasy mini plus kit (QIAGEN). After quantification, RNA was used for cDNA synthesis using the iScript cDNA synthesis kit (Invitrogen). Samples were analyzed on an iCycler qPCR machine (Bio-Rad). Gene expression level was determined using the 2^−ΔΔCt method and normalized with the reference gene, Gapdh. Primer sequences are listed in Table 1 .

Immune cells were incubated with an Fc-blocking antibody at 4°C for 15 min. Post- Fc-receptor blocking, cells were stained for surface markers using antibodies (conjugated with different fluorochromes, all from BioLegend) against CD1d (Clone: 1B1, Cat: 123524, BioLegend), CD40 (Clone: 3/23, Cat: 124626, BioLegend), B220 (Clone: RA3-6B2, Cat: 103224, BioLegend), CD5 (Clone: 53-7.3, Cat: 100624, BioLegend) and a viability dye (Zombie dye, BioLegend), for 30 min at 4°C. For intracellular cytokine staining, cells were incubated at 37°C in the presence of PMA (50 ng/mL, Sigma), Ionomycin (0.5 μg/mL, Sigma) and Golgi Plug (BD) for 4 h prior to washing and surface staining as outlined above. After surface staining, the cells were fixed (20 min, room temperature) and permeabilized (buffers purchased from BD Bioscience) and incubated with an Fc-blocking antibody at 4°C for 15 min prior to staining with anti-cytokine antibodies (30 min, 4°C) and washing. The cells were analyzed on a BD LSR II flow cytometry followed by analysis using Flowjo software.

Statistical analysis was performed using GraphPad Prism 9 software. Diabetes incidence was compared using log-rank test. In vitro assays were analyzed with Student’s t test or ANOVA, and P < 0.05 was considered to be statistically significant. The significance of beta diversity in gut microbiota was assessed using PerMANOVA. Differences in insulitis between two groups were evaluated using the chi-square test. The log-rank (Mantel-Cox) test was employed to examine differences in the incidence of type 1 diabetes (T1D) between two groups of mice.

We reported previously that B-cell-specific TLR9-deficient NOD mice were protected from autoimmune diabetes development (8). However, the role that the gut microbiota play in the protection is not known, given the fact that B cells are a major type of immune cells in the MALT. We first assessed gut permeability of B-cell-specific TLR9-deficient NOD mice (Tlr9 ^fl/fl Cd19-Cre⁺,designated KO group) and control Tlr9 ^fl/fl Cd19-Cre^- NOD mice (designated control (Ctr) group) with fluorescein isothiocyanate (FITC)-dextran test. Our results showed that Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice had a significant increase in gut permeability compared to the control mice ( Figure 1A ). Next, we evaluated the expression of 8 different genes related to gut barrier integrity in different segments of intestine and found that approximately one-half (9/16) of the mRNA expression levels were significantly downregulated in the ileum and colon of mice in the KO group ( Figures 1B, C ). To investigate whether the gut microbiota contribute to the gut barrier disruption in the Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice, we treated the two groups of mice with a cocktail of antibiotics for 3 weeks to deplete the gut microbiota followed by assessement of gut permeability in vivo and intestinal barrier associated gene expression in vitro. With depletion of gut microbiota, Tlr9 ^fl/fl Cd19-Cre⁺ (ABX_KO) mice showed an indistinguishable level of gut permeability compared to that of Tlr9 ^fl/fl Cd19-Cre^- (ABX_Ctr) NOD mice ( Figures 1D-F ). The change of gut microbiome is likely related to the specific knockout of TLR9 in B cells, which affects the level of immunoglobulins represented by IgA in the intestine where B cells are abundantly present ( Supplementary Figure 1A ). In addition, different IgG isotypes were all reduced in the serum ( Supplementary Figures 1B–D ). Similar to the immunoglobulin levels in the intestine, circulating IgA and some isotypes of IgG were also reduced in Tlr9 ^fl/fl Cd19-Cre⁺ mice ( Supplementary Figures 2A–E ). Unlike the intestinal IgM, circulating IgM was markedly reduced in in Tlr9 ^fl/fl Cd19-Cre⁺ mice ( Supplementary Figure 2E ). Thus, the depletion of the gut microbiota altered the gut barrier in the Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice, indicating that the gut microbiota are important in maintaining homeostasis of intestinal permeability.

To explore whether the altered mucosal immune milieu in Tlr9^fl/fl Cd19-Cre⁺ NOD mice affects the gut bacterial load, we quantified bacteria in fecal samples. Compared to the control mice, there was a significant increase in the bacterial load in the fecal samples from KO mice ( Figure 2A ). The maintenance of homeostasis of bacterial load in the intestine is influenced by different molecules, including immunoglobulins, mucins and antimicrobial peptides, all of which are related to intestinal mucosal immunity (33–35). Next, we quantified the gene expression of 7 different antimicrobial peptides in the colon, and found that the expression of Cramp, Reg3b and CRP was significantly down-regulated in the KO mice compared with the control mice ( Figures 2B–H ). We also observed the downregulation of Cramp in the ileum of the KO mice ( Supplementary Figure 3A ). Interestingly, under the gut microbiota-depleted conditions, treated by the antibiotic combination, there was no difference in the gene expression of all the antimicrobial peptides tested between the KO mice and the control mice in both colon and ileum ( Figures 2I–O , Supplementary Figure 4 ). Thus, gut microbiota are required for regulating antimicrobial peptides in the intestinal tract of B cell-specific TLR9-deficient NOD mice.

To futher dissect the structure of the gut microbiota, we performed high-throughput sequencing of the 16S rRNA genes in the fecal samples of the mice from the two study groups. Although we did not find significant differences in alpha diversity and richness between the two groups of mice ( Figures 3A–C ), the structure of gut microbiota in the KO mice was significantly different from the control mice as seen in the principal coordinate analysis (PCoA) plot of Bray-Curtis distance (P=0.0001 with permutational multivariate analysis of variance (PERMANOVA) test ( Figure 3D ). Thus, in addition to altering bacterial load ( Figure 2A ), TLR9 deficiency in B cells resulted in significant changes in the gut microbiota in NOD mice.

To identify key members of the gut microbiota enriched in the KO mice that may contribute to the increased intestinal permeability and decreased antimicrobial peptides, we assessed bacteria load and found that there were significant differences between the Tlr9 ^fl/fl Cd19-Cre⁺ and control mice in the operational taxonomic units (OTUs) of the gut microbiota using sparse partial least squares discriminant analysis (sPLS-DA) ( Figure 3E ; Supplementary Figure 5 ). Notable changes were observed in the family Lachnospiraceae (phylum Firmicutes), a major but understudied mammalian taxon (36). Specifically, 7 OTUs in this family were enriched, while 3 OTUs were reduced ( Figure 3E ). Taken together, our results revealed that Lachnospiraceae were significantly enriched in the Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice compared to the control mice ( Figure 3F ). Lachnospiraceae are known to have mucin-degrading properties (37); thus, the enriched Lachnospiraceae are likely to be related to the increased intestinal permeability in the KO mice.

To test if the altered structure of gut microbiota in the KO mice was responsible for the gut mucosal changes seen in the KO mice, we transferred fresh feces from Tlr9 ^fl/fl Cd19-Cre⁺ and the control Tlr9 ^fl/fl Cd19-Cre^- NOD mice to wild-type germ-free (GF) NOD mice. The recipient mice were designated GF_KO and GF_Ctr mice, respectively. We found that the structures of gut microbiota in the GF_KO and GF_Ctr mice were close to their donors as early as one week after the fecal transplantation ( Figure 4A ; Supplementary Figures 6A–C ). Quantification of the composition distance between the samples further confirmed this finding ( Figures 4B, C ). Next, we tested gut permeability of the receipient GF_KO and GF_Ctr mice with the FITC-dextran assay and observed a significant increase in gut permeability in GF_KO mice compared to the GF_Ctr mice ( Figure 4D ). The increased gut permeability was maintained in GF_KO mice one month after the fecal transplantation ( Supplementary Figure 7 ). Our results indicated that the gut microbiota of Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice had a legacy effect on increasing intestinal permeability. Similar to the phenotype of the donor KO mice, there was a significant downregulation in the mRNA expression levels of five out of eight genes associated with gut barrier integrity in the colon of the GF_KO mice ( Figure 4E ). Muc1 and Cldn1 genes were also significantly downregulated in the ileum of GF_KO mice compared to the GF_Ctr mice ( Supplementary Figures 8A, G ). Among the detected genes related to the expression of antimicrobial peptides, Cramp, β-defen and Reg3β were significantly down-regulated in the the colon of GF_KO mice compared to those of GF_Ctr mice ( Figure 4F ), and the expression of Reg3γ appeared to be down-regulated ( Figure 4F ). Thus, our results showed the legacy effects of the gut microbiota on the regulation of gut barrier and antimicrobial peptides in the colon and ileum of the secondary hosts ( Supplementary Figures 8I–X ) and gut barrier related gene expression was further reduced with time in ileum ( Supplementary Figures 8Q–X ).

We have previously reported that TLR9 deficiency in B cells reduces pro-inflammatory signaling pathways by upregulating interleukin-10 (IL-10) secretion and this reduces the incidence of T1D in NOD mice (8). Considering that the microbiota modulate immune responses, directly or indirectly, we hypothesize that the gut microbiota of Tlr9 ^fl/fl CD19-Cre⁺ NOD mice may directly promote IL-10 secretion. To test this hypothesis, we assessed circulating IL-10 of the recipient GF mice. Consistent with the donor Tlr9 ^fl/fl CD19-Cre⁺ mice in our previous report (8), GF NOD mice that received gut microbiota from Tlr9 ^fl/fl Cd19-Cre+ mice (GF_KO) also had a higher levels of IL-10 in serum compared with GF_Ctr group ( Figure 4G ). Moreover, GF_KO mice showed a significantly increased proportion of IL-10-expressing Breg cells (regulatory B cells) in both the spleen and pancreatic lymph nodes (PLNs) ( Figures 4H, I ). Thus, the gut microbiota from Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice induced increased intestinal permeability, decreased expression of intestinal antimicrobial peptides and increased IL-10-expressing Bregs together with circulating IL-10.

To further explore the direct effect of the microbiota of Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice on IL-10 related immune cells, we co-cultured splenocytes of wild type NOD mice with different concentrations of filtered gut bacteria for 10 hours (38). The culture supernatants were tested for secreted IL-10 by ELISA. Supporting the in vivo data from ex-GF NOD mice, we found that splenocytes stimulated with fecal bacteria from Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice promoted higher levels of IL-10 secretion in a dose dependent manner ( Figure 5A ). Notably, the proportion of Breg cells, assessed by flow cytometry, was in line with the profile of secreted IL-10 but at a higher bacterial dose ( Figures 5B, D ). Further, there were increased IL-10-secreting Breg cells ( Figures 5C, E ) whereas there was no difference in IL-10-produing T cells (data not shown). Thus, our results provided evidence that the fecal microbiota from Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice promoted IL-10-producing Breg cells directly.

Our previous work showed that TLR9 deficiency in B cells delayed the onset of autoimmune diabetes in NOD mice (8). However, the role of the gut microbiota in diabetes protection was not known. To test whether the microbiota of Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice play a role in the disease protection, we performed an immune cell adoptive transfer approach using Rag1 ^-/- NOD mice that were depleted of endogenous gut microbiota with antibiotic treatment(“pseudo germ-free” mice). The pseudo germ-free Rag1 ^-/- NOD mice were colonized with gut microbiota from Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice (KO group) or and Tlr9 ^fl/fl Cd19-Cre^- NOD mice (Ctr group), respectively, followed by adoptively transferring total splenocytes (5 × 10⁶/mouse) from BDC2.5⁺ TCR transgenic NOD mice ( Figure 6A ) (39). The recipient mice were denoted Rag^-/- _KO and Rag1^-/- _Ctr mice, respectively. One week after the adoptive transfer, we assessed random blood glucose levels of the recipient mice. The results indicated that Rag1 ^-/- KO mice colonized with microbiota from Tlr9^fl/fl Cd19-Cre⁺ donors (Rag1^-/- _KO) exhibited significantly lower blood glucose levels compared to the Rag1 ^-/- mice that were colonized with gut microbiota from Tlr9^fl/fl Cd19-Cre^- control donors (Rag1^-/- _Ctr, Figure 6B ). Moreover, the incidence of diabetes in the Rag1^-/- _KO mice was significantly delayed and reduced compared with the Rag1^-/- _Ctr mice ( Figure 6C ). Supporting the findings related to diabetes development, Rag1^-/- _KO mice also had significantly less severe insulitis compared to the Rag1^-/- _Ctr mice ( Figures 6D, E ). Next we assessed the anti-inflammatory cytokine IL-10 in the circulation and found much higher IL-10 levels in the serum samples from the Rag1^-/- _KO mice compared with the Rag^-/- _Cre mice ( Figure 6F ). Thus, our results indicated that the microbiota of Tlr9 ^fl/fl Cd19-Cre⁺ NOD mice are directly involved in diabetes protection by modulating the host immune tolerance through promoting the secretion of immune-regulatory cytokine IL-10.