Promotion of Autoimmune Diabetes by Cereal Diet in the Presence or Absence of Microbes Associated With Gut Immune Activation, Regulatory Imbalance, and Altered Cathelicidin Antimicrobial Peptide

Animal studies were approved by the local animal care committees. BBdp or control BB rats (BBc) were maintained in SPF or GF conditions and weaned onto the AIN93G HC-based diet or a cereal diet (NTP-2000 fed to SPF rats or irradiated C.R. Rodent 18% diet [PMI Nutrition International, Brentwood, MO] fed to GF rats) beginning at 23 days. GF status was confirmed by microbiological and immunological testing and by PCR analysis. Animals were monitored biweekly for T1D, defined as ≥11 mmol/L fasting blood glucose. Rats were killed within 24 to 48 h of T1D diagnosis or after ∼134 days without T1D development.

Pancreas and washed gut tissues were fixed in Bouin’s fixative for histological evaluation or were frozen for RNA isolation and gene expression analysis. Additional studies were performed using rats that were nonlymphopenic or that had in addition the Gimap5 (lymphopenia) gene inserted (19). Except where noted, all analyses were performed on pancreas or jejunum of BBdp rats that remained asymptomatic until ≥134 days.

Jejunum biopsies were obtained from nondiabetic control subjects (patients with short stature, failure to thrive, gastroesophageal reflux disease, recurrent abdominal pain, anemia, n = 14 males) and 8 male patients with T1D aged between 8 and 16 years of age, with informed consent (20). All biopsy samples (range 1–7 per subject) on a single slide were analyzed (17).

Antibodies to the following markers were used: CD3 (Abcam), CD8α (BD Biosciences), CD4 (Abcam), Foxp3 (eBioscience), CD163 (rat; Santa Cruz Biotechnology), CD163 (human; Abcam), cathelicidin antimicrobial peptide (CAMP; Abcam), CD68 (AbD Serotec), and CD14 (Santa Cruz Biotechnology). Coded sections were analyzed using a Zeiss Axioplan2 microscope and Northern Eclipse morphometry software (Empix Imaging) (17). The numbers of CD3⁺, CD8α⁺, and CD4⁺ intraepithelial lymphocytes (IEL) were expressed as the number of IELs per 100 epithelial cells or lamina propria lymphocytes (LPL) per mm² mucosal area. Foxp3⁺ cells, CD163⁺ cells, and CD68⁺ cells were expressed as the cell number per mm² mucosal area. Corresponding biotinylated secondary antibodies (DAKO Diagnostics) were applied, followed by incubation with streptavidin-conjugated horseradish peroxidase (DAKO Diagnostics) and 0.06% diaminobenzidine and 0.03% H₂O₂ as substrates; counterstain was hematoxylin. Immunofluorescence was performed using a Zeiss LSM 510 Meta confocal microscope and Alexa-488, Cy-3, and Cy-5-conjugated secondary antibodies (Life Technologies). Hoechst (Sigma) was used for nuclear staining.

RNA was extracted from frozen jejunum using TRIzol (Invitrogen) or Nucleospin RNA II RNA Isolation Kit (Macherey-Nagel). Quantitative PCR studies were performed using TaqMan Gene Expression Assays (Life Technologies). An ABI Prism 7000 Sequence Detection System was used to obtain threshold cycle (Ct) values; results are shown as fold-change (2^-ΔΔCt). Changes in gene expression are presented in relation to the low-antigen HC-GF group, which was arbitrarily set to one. For PCR array profiling, gene expression was analyzed using Innate and Adaptive Immune Response PCR arrays (SABiosciences/Qiagen) on samples from HC-SPF and cereal-SPF BBdp rats (n = 4 per group). Ct values were normalized with the average Ct values of five housekeeping genes.

DNA was isolated from individual fecal pellets by grinding in sterile sand and liquid nitrogen and purified using the QIAamp DNA stool mini kit (Qiagen, Inc., Mississauga, ON). Quality was assessed using 0.8% agarose gels, and samples were stored frozen at −20°C. Fecal bacterial community profiling compared the V2-3 region of the 16S rRNA genes and was carried out by denaturing gradient gel electrophoresis (DGGE). Gels were digitized and then analyzed using BioNumerics software (BioNumerics, Brussels, Belgium). Band cross-sectional areas were expressed as relative intensity and analyzed by cluster analysis followed by non-metric multidimensional scaling analysis using Statistica software (Tulsa, OK).

Survival analysis was performed using the log-rank test. Comparisons among sample populations were made using the Student t test or one-way ANOVA and least significant difference post hoc testing to determine the significance of differences between multiple means. The Mann-Whitney U test was performed for human samples. P values < 0.05 were considered statistically significant.

As we previously reported in SPF BBdp rats (2), feeding an HC diet increased age at onset and inhibited T1D development compared with a cereal diet (29 vs. 65%, P < 0.005; Fig. 1A). The HC diet was even more protective in GF BBdp rats, where T1D was almost completely prevented (7 vs. 53% in cereal-fed GF rats, P < 0.005). The main effect of microbes on T1D incidence was observed in HC-fed rats, in which disease was promoted by microbial exposure (29% HC-SPF vs. 7% HC-GF, P < 0.05). Cereal-fed rats housed in GF conditions displayed a delay in age at onset, but T1D incidence was not significantly different compared with cereal-fed SPF rats. Thus, in the absence of an antigenic diet (HC feeding), microbes can promote T1D in BBdp rats whereas microbial exposure did not significantly enhance T1D progression in cereal-fed animals. Overall, our results demonstrate that microbes were not required for T1D development and cereal antigens were the strongest environmental promoters of T1D in the BBdp model.

Body mass, pancreatic mass, and islet inflammation (insulitis) were not modified by microbes or diet in 130-day asymptomatic rats (data not shown). Total islet number was increased in GF rats compared with SPF rats (Fig. 1B), consistent with the delay in T1D onset in GF rats, regardless of diet. Under SPF conditions, rats fed an HC diet had a significantly increased β-cell mass compared with cereal-fed animals, consistent with our previous report (17). Cereal-fed rats raised in GF conditions displayed a larger β-cell mass than SPF cereal-fed rats. Among HC-fed rats, β-cell mass was not different between those raised in GF compared with SPF conditions. β-cell mass was lowest in cereal-SPF rats, and this was overcome by feeding the protective HC diet (HC-SPF). Serum insulin concentration was not different among the groups (data not shown).

Denaturing gradient gel electrophoresis of fecal samples from BBc and BBdp rats was performed to ascertain whether the bacterial communities were affected by diet. Nonmetric multidimensional scaling analysis revealed that BBdp rats fed a protective HC diet had a distinct bacterial profile compared with BBdp rats and BBc rats fed a standard cereal diet (Fig. 1C).

Because gut inflammation has been reported in BBdp rats, NOD mice, and T1D patients, we evaluated the effect of diet and/or microbes on key gut immune cell populations. The highest numbers of CD3⁺ and CD8α⁺ IEL and LPL occurred in cereal-SPF rats. HC-SPF rats displayed decreased CD3⁺ and CD8α⁺ cell numbers compared with cereal-SPF rats (Fig. 2A and B). No significant diet-related differences in the numbers of CD3⁺ or CD8α⁺ IEL or LPL were observed in GF animals. Among cereal-fed rats, the numbers of CD3⁺ and CD8α⁺ IEL and LPL were larger in SPF rats compared with GF rats. Also, more CD3⁺ cells were found in HC-SPF than in HC-GF animals. Thus, CD3⁺ and CD8α⁺ lymphocytic infiltration was promoted by microbial exposure. The number of CD4⁺ cells was higher than expected and the overall pattern differed from CD3 and CD8α labeling, possibly due to other cell populations being labeled (Fig. 2C).

Foxp3 is a master transcription factor that directs the differentiation and function of regulatory T cells (Foxp3⁺ T_reg) (21). Foxp3⁺ cells were exclusively localized to the lamina propria (Fig. 3A), and confocal microscopy confirmed nuclear expression in CD3⁺ lymphocytes (data not shown). Unlike total CD3⁺ cells, the number of Foxp3⁺ cells did not decline under GF conditions. The most protective low-antigen combination of GF housing and HC feeding resulted in the highest number of Foxp3⁺ cells (Fig. 3B), and this difference became significant when expressed as a percentage of jejunal CD3⁺ (Fig. 3C) or CD4⁺ LPL (Fig. 3D). A much lower percentage of Foxp3⁺ cells was observed in the most diabetes-promoting, high-antigen situation of SPF housing and/or cereal feeding (Fig. 3C and D). Thus, the dietary influence on Foxp3⁺ T_reg proportion was apparent under sterile conditions, with protective HC feeding resulting in enrichment of regulatory lymphocytes.

The proinflammatory T_H1 cytokine gene, Ifng was upregulated by the cereal diet in GF rats (Fig. 4A). Microbes increased Ifng expression in HC-SPF rats versus HC-GF rats, but expression was similar in cereal-SPF and cereal-GF rats. There were no differences in Il17a gene expression among the four groups (Fig. 4B), and Il15 expression was largest in cereal-GF rats (Fig. 4C). Under GF conditions, the cereal diet stimulated upregulation of the hallmark T_H2 cytokine, Il4, compared with the HC diet (Fig. 4D). The ratio of Ifng/Il4 was highest in cereal-SPF rats and was significantly different from the HC-SPF and cereal-GF groups (Fig. 4G), reflecting enhanced cereal-associated T_H1 polarization in the gut immune system, as reported previously (7). The highest expression of Il10 and Tgfb1 was in SPF compared with GF rats, regardless of diet (Fig. 4E and F). The ratio of Ifng/Il10 was highest in cereal-GF rats and low in the other three groups (Fig. 4H). The ratio of Ifng/Tgfb1 was highest in cereal-GF rats and was relatively low in the other three groups (Fig. 4I). The data suggest that overall, antigenic stimulation by microbial exposure and/or cereal feeding upregulated cytokine gene expression, with Ifng expression predominating over various counterregulatory cytokines in cereal-fed rats.

Tissue-resident CD163⁺ (M2) macrophages are immunosuppressive, have lower levels of major histocompatibility complex (MHC) class II and other markers of classically activated macrophages (22), and contribute to maintenance of a hyporesponsive state. We observed CD163⁺ cells mainly in the lamina propria and also in a smaller subset in the epithelial compartment (Fig. 5A). In a previous analysis of cereal-fed SPF BBdp rats, we observed a deficiency in gut CD163⁺ macrophages compared with BBc rats (Fig. 5B). In the current study of BBdp rats only, feeding the protective HC diet increased the number of CD163⁺ cells compared with BBdp rats fed a cereal-based diet (Fig. 5C). No dietary differences were observed under GF conditions. Gene expression analysis was consistent with these data (Fig. 5D).

The Hmox1 gene for the rate-limiting cytoprotective heme oxygenase-1 (HO-1) enzyme in the CD163 pathway showed highest expression in the most protected animals lacking antigen exposure, HC-GF (Fig. 5E). The number of CD68⁺ macrophages was unaffected by diet or microbial status (data not shown). These data suggest the influence of the environment on macrophages occurs primarily through changes in the M2 subset of macrophages, in keeping with the HC-associated decrease in the Ifng/Il4 ratio (Fig. 4G), which favors polarization of M2 macrophages (23). To investigate the potential relevance of these results to patients, we performed preliminary analysis of gut biopsy specimens from a small group of males newly diagnosed with T1D. This analysis revealed lower numbers of CD163⁺ macrophages in the jejunum compared with control individuals (Fig. 5F and G).

To gain a better understanding of dietary modulation of immune factors under standard housing conditions, we screened for diet-modifiable immune factors in the jejunum of conventionally housed SPF rats using focused immune-associated PCR arrays. Only candidates that were differentially expressed ± twofold with P < 0.05 were explored; results are presented as a volcano plot (Fig. 6A). Il1f6 expression was 2.78-fold upregulated in HC-fed rats compared with cereal-fed rats (P = 0.004; Fig. 6B). Il1f6 encodes the cytokine interleukin 1 family member 6 (IL-1F6). The IL-1F6 cytokine has been implicated in cutaneous inflammation (24) and induced antimicrobial expression, including β-defensins and CAMP (25). Interestingly, the other significantly upregulated gene in HC-fed BBdp rats was Camp. Camp expression was 2.37-fold upregulated in HC-fed rats compared with cereal-fed rats (P = 0.03; Fig. 6B). CAMP is a multifunctional antimicrobial effector and immunomodulatory host defense factor (26). In addition, consistent with increased CD3⁺ and CD8α⁺ cells in gut of cereal-fed rats (Fig. 2), the expression of Lck, a T-cell signaling molecule, was 2.43-fold downregulated in HC-fed rats compared with cereal-fed rats (P = 0.003; Fig. 6B). CAMP⁺ cells were detected in the epithelium and lamina propria of the SPF jejunum (Fig. 6C). CAMP⁺ cells were also present in gut of sterile embryos and GF BBdp rats (Fig. 6C), suggesting CAMP is not only antimicrobial but also has additional functions. Additional confocal analyses revealed that CAMP colocalized with CD163⁺ and CD14⁺ cells in the epithelium and lamina propria (Fig. 6D), but not with the macrophage marker CD68 (Fig. 6C). This suggests that CAMP could be an effector peptide or product of M2 macrophages. These findings reveal that feeding a protective HC diet is associated with antimicrobial gene upregulation in CD163⁺ macrophages in the small intestine.

We previously reported that BBdp rats display enteropathy that precedes T1D and is not present in BBc rats (6). To determine whether BBdp enteropathy is linked to major BBdp diabetes risk loci, we evaluated the presence of enteropathy in various congenic rat strains. Jejunal sections of congenic ACI.1^u and ACI.1^ulyp/lyp rats were analyzed at 49, 60, 82, and 259 days. No distinct morphological abnormalities were observed in the gut tissue from these congenic animals (Fig. 7). Inflammation was absent in nonlymphopenic rats congenic for the diabetes risk MHC II allele, RT1^u (Wistar Furth, BB control, ACI.1^u) nor was it present in ACI.1^u.lyp/lyp rats that possess the RT1^u allele and the Gimap5 mutation. Therefore, the presence of enteropathy in BBdp rats was not attributable to Gimap5-associated lymphopenia or RT1^u, the rat diabetes risk MHC.

The relationship among diet, microbiota, gut immunity, and diabetes is poorly understood and, at times, contradictory. Some studies report no relation between early infections or vaccinations and risk of T1D (27), and a large nationwide study of Danish children found no association with antibiotic use (28). MyD88-deficient NOD mice were protected from T1D under SPF conditions but not GF conditions, further illustrating the complex role of microbes in T1D (16).

In the present study, a cereal-based diet was the major diabetes-promoting factor in both SPF and GF conditions. This clearly illustrates that despite dietary differences in gut bacterial profiles (Fig. 1C), diet has a stronger influence than microbes on T1D development in these animals. An earlier report also found that microbes were not essential for development of T1D in the cereal-fed BB rat (29). The nearly complete protection afforded by the HC diet in GF conditions was partially inhibited in the presence of microbes (SPF-HC). Thus, microbes promote T1D, but to a lesser extent than cereal diets. Among HC-fed rats, microbes also promoted infiltration of T cells (Fig. 2) and decreased the proportion of T_regs (Fig. 3). It is possible that any effect of bacteria would be most prominent in the colon (30). Immune system changes in the jejunum may only be part of the picture, and a more detailed examination of the influence of microbial and dietary antigenic load along the entire length of the gastrointestinal tract would also be informative.

Snell-Bergeon et al. (18) suggest that diet-induced gut inflammation could be a prerequisite for gut pathogen–induced islet autoimmunity. Interestingly, we observed more overlap between the bacterial profiles of cereal-fed BBdp rats and cereal-fed BBc rats compared with HC-fed BBdp rats. This suggests that HC feeding did not afford protection from T1D in SPF rats by shifting the bacterial profile toward the control strain profile. Nonetheless, the microbiota are different in BBdp rats that develop diabetes (31–33), and administering protective bacteria can prevent T1D (34). When BBdp Wor/Gro rats were treated with antibiotics and fed an HC diet from weaning, T1D was completely prevented (31), in agreement with the present results (Fig. 1) and consistent with the proposition that gut-derived environmental antigenic load is a key determinant of T1D promotion. The HC diet can change the gut microbiota (33), and this may contribute to the diabetes-protective effect. Thus, diet and microbes could have synergistic effects on gut immunity, islet homeostasis, and T1D incidence.

Under SPF conditions, there were increases in the numbers of CD3⁺ and CD8α⁺ LPL and IEL in rats fed a cereal diet compared with an HC diet, demonstrating that under normal conditions, cereal-feeding drives T-cell infiltration in the small intestine of BBdp rats. Although the cytotoxic nature of IEL in BBdp rats is unclear, patients with celiac disease, Crohn’s disease, and ulcerative colitis are known to have increased densities of these cells that may mediate tissue damage in the small bowel (35), potentially increasing gut leakiness. When small intestinal biopsy samples from children with T1D were cultured with gliadin, the numbers of CD3⁺ IEL and activated lamina propria CD25⁺ cells were increased in T1D gut (20). These data indicate that inflammation is present in the jejunum of patients with T1D and suggest that the small intestine of subjects with T1D is immunologically more responsive to cereal antigens, consistent with our findings.

Ifng was upregulated by microbes and cereal diet, whereas Il10 and Tgfb were upregulated primarily by microbes. In GF rats, the cereal diet upregulated Il15, Il4, Ifng/Il10, and Ifng/Tgfb1. IL-15 is a proinflammatory cytokine that is implicated in celiac disease (35) and has recently been shown to promote T1D in NOD mice (36). In the HC-GF group, which had essentially no exposure to environmental antigens, gene expression for Ifng, Il4, and Il10 was low, corresponding with the decreased number of T cells in this group (Fig. 2). A combined effect of microbes and cereal-feeding on proinflammatory T_H1 polarization was evident from the Ifng/Il4 ratio, which was highest in the cereal-SPF group but low in the cereal-GF group. This ratio was decreased by HC feeding, further demonstrating an inhibition of the usual T_H1 bias in the gut of BBdp rats (37). Thus, cereal antigens promote T_H1-biased gene expression in the small intestine, as suggested previously (37). The induction of small intestinal T-cell signaling in the cereal-SPF group was further reflected by increased expression of Lck, the most strongly expressed cereal-induced gene on the PCR array. Lck encodes tyrosine kinase/p56, a lymphocyte-specific protein involved in the initiation of T-cell activation (38). The involvement of T_H17 in BB diabetes is controversial (39,40). However, our data demonstrate no diet- or microbe-induced differences in jejunum Il17a gene expression (Fig. 4B).

Foxp3⁺ T_reg cells play a central role in the inhibition of autoimmunity and suppression of physiological immune responses (41). In GF conditions, the proportion of Foxp3⁺ T_reg cells was decreased in cereal-fed rats compared with HC-fed rats. Strikingly, the proportion of Foxp3⁺ T_reg cells was lowest in the cereal-SPF rats and highest in the HC-GF rats, reflecting the highest and lowest external antigen load and the highest and lowest T1D incidences, respectively. In addition, HC-GF rats displayed a significantly increased proportion of Foxp3⁺ T_reg cells compared with HC-SPF rats, which paralleled the delay in age of onset in GF rats. Consistent with this, Peyer’s patches of BALB/c mice fed a gluten-containing diet displayed a smaller proportion of Foxp3⁺ T cells than mice fed a gluten-free formula (42). The decreased proportion of Foxp3⁺ T_reg cells in cereal-GF rats compared with HC-GF rats could be an important permissive factor contributing to proinflammatory reactivity to cereal antigens under sterile conditions. The relative increase in Foxp3⁺ T cells in HC-GF rats was not accompanied by increased gene expression for Tgfb1 and Il10. However, the lowest number of lamina propria CD4⁺ T cells was observed in HC-GF rats (Fig. 2C), by a factor of two- to threefold, yet the levels of Tgfb1 in these animals were not different from cereal-GF rats and only half that of both SPF groups. We speculate that on a cell-per-cell basis, the synthesis of Tgfb1 was increased in the HC-GF rats, further consistent with the high proportion of Foxp3⁺ T cells being the source of Tgfb1 (Fig. 3).

The largest reservoir of tissue-resident regulatory macrophages in the body occurs in the lamina propria of the intestine (43). These cells suppress inflammation (44) and express CD163, a scavenger receptor for haptoglobin-hemoglobin complexes. CD163 is inducible by IL-4 and IL-10 (45), and differentiation of regulatory CD163⁺ macrophages is promoted by Foxp3⁺ T_reg (46). Thus, development and maintenance of an immunoregulatory state in the gut depends on a constant interplay between tolerogenic subsets of antigen-presenting cells and T cells. The number of resident CD163⁺ M2 macrophages was inversely related to T1D risk, highest in control, diabetes-resistant BBc and lowest in overt diabetic BBdp rats and was increased in the diet-protected HC-SPF BBdp rats. The HC diet–induced increase in CD163⁺ cells required the presence of microbes. CD163 expression was highest when the ratio of Ifng/Il4 was low, consistent with a downregulation of T_H1 inflammation by M2 macrophages. In addition, the Hmox1 gene was downregulated in antigen-exposed groups and was highest in the most protected (low-antigen) HC-GF group, suggesting additional environmental sensitivity of the CD163/HO-1 pathway (47). These findings are in keeping with a recent report that IL-4/IL-10/TGF-β−induced M2 macrophages prevented T1D in NOD mice (48). Furthermore, our preliminary data suggest there can be fewer CD163⁺ cells in the gut of some newly diagnosed T1D patients, a finding that requires confirmation.

The increase in small intestinal Camp gene expression in HC diet-protected BBdp rats and colocalization of CAMP and CD163 proteins suggests that antimicrobial peptide production is a function of M2 macrophages that could promote an anti-inflammatory state in the gut. Rats and humans possess only one cathelicidin gene (49). CAMP has been reported to be produced by neutrophils, monocytes, macrophages, and epithelial cells in various tissues (50). The most extensively characterized function of CAMP is microbicidal activity, but additional roles have been described, including regulation of inflammation, growth, chemotaxis, tissue repair, wound healing, apoptosis regulation, and angiogenesis (49). CAMP has been implicated in inflammatory bowel diseases, as decreased CAMP expression has been reported in patients with Crohn’s disease (51). CAMP-deficient mice displayed a more pronounced form of colitis compared with control mice treated with dextran sodium sulfate (52). In addition, intracolonic treatment with synthetic CAMP prevented dextran sodium sulfate colitis, an effect associated with a decreased number of fecal bacteria, enhanced mucin production, and suppression of neutrophil infiltration (53). Thus, decreased Camp expression, partly from CD163⁺ M2 macrophages, could favor chronic inflammation induced by environmental antigens.

Therefore, in BBdp rats, a cereal diet promoted T1D in the presence or absence of microbes. Cereal feeding was associated with increased T-cell infiltration, a T_H1 cytokine bias, and deficiencies in anti-inflammatory CD163⁺ M2 macrophages that were partly corrected in SPF animals fed a protective HC diet. PCR array screening of SPF BBdp rat jejunum further revealed Camp as a novel HC diet-upregulated factor. CAMP colocalization with CD163 suggests it may be a new marker of M2 macrophages and highlights a potential role for this antimicrobial peptide in T1D modulation. Importantly, the upregulation of CD163 in HC-SPF animals suggests that this protective effect required the presence of microbes, further emphasizing the complex interaction between diet and microbes in the gut of diabetes-prone animals. We speculate that diet-induced T1D depends on an increased load of antigens from diet and/or microbes crossing a leaky gut epithelial barrier and activating β-cell–specific immunity. This immune activation was dampened when animals were fed an essentially antigen-free HC diet, particularly under GF conditions. Microbes could modify this process because the HC diet resulted in a different bacterial profile in parallel with increased CD163 macrophages that produce antimicrobial CAMP. This study highlights that development of T1D is a heterogeneous, environmentally driven process, particularly in humans, and is likely to occur by several different pathways.

These studies were supported by the Juvenile Diabetes Research Foundation, Canadian Diabetes Association, Canadian Institutes of Health Research, Ontario Research and Development Challenge Fund, and Canada Foundation for Innovation and Health Canada. C.P. was the recipient of Queen Elizabeth II Graduate Scholarship in Science and Technology, Ontario Graduate Scholarship in Science and Technology, and Ontario Graduate Scholarships. D.E.L. and B.S. were the recipients of Ontario Graduate Scholarships, and B.S. also received a scholarship from the Fonds de la recherche en Santé du Québec.

No potential conflicts of interest relevant to this article were reported.

C.P., G.-S.W., D.E.L., and J.A.C. designed and performed the experiments, analyzed the data, and wrote the manuscript. B.S., C.E., M.Mo., S.P.J.B., and M.L.K. designed and performed the experiments and analyzed the data. C.R.K. contributed materials and analysis tools. M.Ma. and R.T. contributed human samples and helped write the manuscript. P.P. contributed congenic animals, assisted in the design of the study, and helped write the manuscript. F.W.S. conceived and designed the study, analyzed the data, and wrote the manuscript. F.W.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Parts of this work were published in abstract form in Diabetes 2007;56(Suppl. 1):A668 and Diabetes 2008;57(Suppl. 1):A51 and were presented at the 11th Immunology of Diabetes Society Meeting, Seoul, Korea, 31 October–3 November 2010 and at the 12th Immunology of Diabetes Society Meeting, Victoria, British Columbia, Canada, 15–19 June 2012.

The authors acknowledge the important contribution of Dr. Claude Gardell (formerly Health Canada), veterinarian, and Normand Turcotte in establishing the GF colony and the excellent technical assistance of Ahmed Zafer (OHRI) and the following at Health Canada: Jocelyn Souligny, Douglas Parks, Charles Martineau, Joann Clausen, Dominique Patry, Bill Pearce, and Svetlana Popovic for animal care. The authors thank Dr. William White and Dr. Jocelyn Goyer at Charles River Laboratories for assistance with the GF protocol and serological analyses of viral antibodies. The authors thank Caroline Vergette of the StemCore Laboratories at the Ottawa Hospital Research Institute for technical assistance and are grateful to Dr. Alexander Strom (German Diabetes Center, Dűsseldorf, Germany) for helpful discussions and technical assistance.