Anti-Islet Autoantibodies Trigger Autoimmune Diabetes in the Presence of an Increased Frequency of Islet-Reactive CD4 T Cells

3A9TCR-transgenic (T-cell receptor [TCR]+) (28), insulin promoter-driven hen egg lysozyme (HEL)-transgenic (HEL+, originally called ILK3) (29), MD4 (IgMD^HEL) (30), γγ4 (HyHEL10-IgG1) (31), and Ptprc^SJL (CD45.1) mice were backcrossed seven or more generations to B10.BR. Roquin^san C57BL/6 mice were backcrossed to TCR+HEL+ mice with a mixed (CBA × B10.BR) background, fixing H2^k. Bim-deficient mice have been described (32), and Cd79a^ken is an N-ethyl-nitrosourea–induced null allele resulting from a premature stop codon in exon 2 that abolishes CD79a mRNA and results in a completely nonleaky block in B-cell development and antibody formation (33). Mice in Fig. 4 were third, fourth, and fifth generation backcross of the TCR and HEL transgenes from B10.BR to CBA. The Animal Ethics and Experimentation Committee of the Australian National University approved all procedures.

HEL-binding IgG was measured by ELISA on Nunc MaxiSorp plates coated with 5 μg/mL HEL (Fluka). Serial serum dilutions were applied followed by alkaline phosphatase-conjugated goat anti-mouse IgG or IgG2a (Southern Biotechnology) and then Phosphatase Substrate (Sigma). Plates were read at 405 nm with a Thermomax Microplate Reader (Molecular Devices). Anti-HEL IgG titers were interpolated from a standard curve made using serum from HyHEL10-IgG1 mice (31) analyzed on the same plate, whereas pooled serum from Roquin^san/san TCR+HEL+ mice was used to generate a standard curve for anti-HEL IgG2a.

HEL-immune serum was collected from mice immunized intraperitoneally 4 weeks earlier with 100 μg HEL protein in 4.5% alum, or 4.5% alum alone for control serum. Alternatively, a 50 μL emulsion containing 50 μg HEL or OVA mixed 1:1 with complete Freund's adjuvant (CFA; Sigma) was injected subcutaneously in each flank. To purify IgG, serum was clarified by centrifugation, diluted 20-fold with binding buffer (300 mM NaCl, 100 mM Tris/HCl, pH 8.0) and filtered through a 0.45 mm Millipore (Billerica) membrane. Diluted serum aliquots of 20 mL were applied to HiTrap Protein G Sepharose column (GE Healthcare). IgG eluted with 0.1 M glycine/HCl (pH3) was collected in 1.0 mL fractions buffered with 30 μL of 3.0 M Tris/HCl (pH8).

Neonates were injected intraperitoneally with 17 μg purified anti-HEL or anti-OVA IgG per gram of body weight or 50 μL serum on days 1, 3, and 5 after birth. Mice were injected intraperitoneally with 20 μg of monoclonal FcγR-blocking antibody (2.4G2; BioXCell) or rat IgG2b isotype control (LTF2; BioXCell) on days 1, 3, 5, 8, 11, 14, and 17 after birth. These neonates were also injected intraperitoneally with 50 μL anti-HEL immune serum on days 2, 4, and 6 after birth. Adult mice were injected with 200 μL serum per dose.

Urine glucose was tested using Uristix (Bayer) twice per week. In mice with two successive positive urine glucose tests, blood glucose was measured, and mice with blood glucose >10 mmol/L were called diabetic. In the Fig. 4 cohort, urine glucose was tested from individual mice when their cage was observed to be wet, and mice with glucosuria >111 mmol/L (++++ on Uristix) were called diabetic. For histology, the pancreas was fixed in 10% formalin, paraffin embedded, and stained with hematoxylin and eosin.

To isolate islets, the pancreas of CBA mice was infused via the common bile duct with 3 mL of Hanks' buffered saline solution (HBSS) containing 3 mg/mL collagenase P (Roche, 11 213873 001), 10 μg/mL DNAse I (Roche, 0104159), 0.15% w/v bovine serum albumin plus penicillin, and streptomycin. Infused pancreata were dissected free and incubated for 15 min at 37°C before replacement of the medium with HBSS containing 10% bovine fetal serum. Islets were separated by two rounds of “hand-picking” using a drawn glass pipette under a dissecting microscope. For transplantation, two depots of 100–125 islets each were incorporated into clots of recipient blood and placed under the kidney capsule, and blood glucose was measured weekly. Eighty-five percent of transplanted diabetic TCR+HEL+ animals became euglycemic within 1–2 weeks.

Lymphocytes were stained as described previously (34) using the following antibodies: 1G12 anticlonotype (mouse IgG1, detecting the 3A9 αβ TCR [TCR^HEL]) (35) culture supernatant followed by rat anti-mouse IgG1-APC; anti–CD8a-PerCP; anti–CD4-FITC-PerCP or -PE, anti–CD3-PE, anti-CXCR5-biotin, anti–CD69-PE, and anti-B220-APCCy7 (BD Biosciences); anti–CD5-FITC and anti–CD25-PE (Caltag); anti-CD45.2-PacificBlue and anti-CD45.1-Alexa700 (BioLegend); and anti–PD-1-PE (eBioscience). An LSR II flow cytometer (BD Biosciences) was used for data acquisition, and FlowJo software (Treestar) was used for analysis.

Spleen and lymph node single-cell suspensions containing 1 × 10⁵ cells in tissue culture medium (RPMI containing 5% fetal calf serum, 2 mM l-glutamine, penicillin, streptomycin, and nonessential amino acids [Gibco], 1 mM sodium pyruvate plus 10 mM Hepes buffer [Sigma]) were placed into flat-bottomed 96-well plates with or without 50 μg/mL HEL protein for 5 days; cell supernatants were then collected and stored at −20°C. Cytokines in supernatants were measured using the Cytometric Bead Array kit (BD Biosciences).

Whole splenocytes from TCR+ mice were labeled with 2.5 µM carboxyfluorescein succinimidyl ester (CFSE) and injected intravenously into recipients at a TCR^HEL+ CD4+ cell dose of ∼3 × 10⁵ per mouse.

Diabetes incidences were compared with the log-rank test using GraphPad Prism. Otherwise, Mann-Whitney U tests were used except for the “ratio t test” used in Figs. 6B and 6C, the rationale of which is described in GraphPad Prism Statistics Guide (http://www.graphpad.com).

A point mutation in the Roquin gene (Rc3h1, Roquin^san) exaggerates inducible T-cell costimulator expression and formation of T follicular helper (Tfh) cells, and causes systemic autoimmunity (36,37). To investigate its effect on islet-specific CD4 T–cell tolerance mechanisms, Roquin^san mice were crossed with TCR+HEL+ double-transgenic mice that have an increased frequency of islet-reactive CD4+ T cells. The HEL transgene encodes HEL under the insulin gene promoter, and mirrors the pattern of insulin expression with high expression in islet β-cells, nanomolar concentrations in serum, and Aire-dependent expression in thymic medullary epithelial cells (29,38). The TCR transgene encodes the T-cell receptor from clone 3A9, which recognizes a dominant HEL peptide (46–61) presented by I-A^k (28).

The resulting TCR+HEL+ mice make it possible to measure cellular mechanisms of CD4 T–cell tolerance. They are, on the one hand, an experimentally contrived model because of their increased thymic production of islet-reactive CD4+ T cells recognizing a single, experimentally expressed autoantigen that mirrors but may not exactly replicate insulin itself. On the other hand, only ∼20% of TCR+HEL+ animals progress to β-cell destruction and diabetes on the diabetes-resistant B10BR strain background due to the combined action of thymic deletion, Treg function, and peripheral anergy. Perturbations found to precipitate diabetes in TCR+HEL+ mice—such as inherited defects in AIRE, Bim-induction, interleukin (IL)-2 signaling, and Cbl-b—have revealed cellular mechanisms that are highly relevant to human type 1 diabetes and to antigenically complex models such as the NOD mouse strain or the KDP rat strain, in which mechanisms regulating islet-reactive CD4 cells are difficult or impossible to measure (29,34,39–41). Hence we used this model to investigate how dysregulated formation of Tfh cells affects the regulation of islet-reactive CD4 T cells.

The Roquin^san mutation dramatically increased progression to type 1 diabetes in TCR+HEL+ mice such that 100% of Roquin^san/san homozygotes developed diabetes by 8 weeks of age (Fig. 1A). The median onset of type 1 diabetes in these mice was earlier than TCR+HEL+ mice with mutations in Aire, Il2 or NOD non-MHC diabetes-susceptibility genes (Fig. 1B) (34,38,40). The Roquin^san mutation had no discernable effect on thymic deletion of islet-reactive CD4 T cells bearing the 3A9 TCR (TCR^HEL) (Supplementary Fig. 1), but the frequency of these cells was increased in the pancreatic lymph node (PLN) of Roquin^san/san mice (Fig. 1C and Supplementary Fig. 2). TCR^HEL+CD4+ cells from Roquin^san/san mice divided extensively ex vivo in response to HEL (not shown) and produced elevated levels of γ-interferon (Fig. 1D and E), indicating that they were not anergic, unlike their counterparts from wild-type TCR+HEL+ mice (29). There was no detectable increase in secreted IL-4 or IL-5, suggesting that the islet-reactive cells in Roquin^san/san TCR+HEL+ mice are predominantly Th1 cells. Thus, Roquin^san/san TCR+HEL+ animals provide an experimental model of spontaneous, rapidly developing diabetes that stems from increased frequency of islet-reactive CD4 cells (due to TCR and HEL transgenes) and breakdown in peripheral tolerance (due to Roquin mutation).

B cells were of particular interest in the accelerated diabetes caused by the Roquin^san mutation because accumulation of Tfh cells causes spontaneous germinal centers and IgG autoantibodies that characterize nontransgenic Roquin^san/san mice (36,42). Roquin^san/san TCR+HEL+ mice had elevated Tfh cell frequencies in PLN compared with those in Roquin^san/+ TCR+HEL+ mice, which reached 10-fold when the heightened CD4+ T–cell frequency was taken into account (Fig. 2A). Germinal center B cells were also detected at increased frequencies in PLN of Roquin^san/san TCR+HEL+ mice (Fig. 2B). Roquin^san/san TCR+HEL+ mice examined 11 days after birth exhibited lymphocytic insulitis, whereas Roquin^+/+ TCR+HEL+ littermates did not (Fig. 2C). Regardless of Roquin genotype, nearly all islets in adult TCR+HEL+ mice exhibited peri-insulitis or insulitis (Fig. 2D), but focal to generalized exocrine pancreatitis was only observed in Roquin^san/san mice (Fig. 2D).

Roquin^san/san TCR+HEL+ mice also developed high titers of anti-HEL IgG autoantibodies in serum (Fig. 3A), including antibodies of the IgG2a isotype (Fig. 3B) characteristic of Th1-driven immune responses. Autoantibody titers correlated with progression to diabetes in heterozygous Roquin^san/+ TCR+HEL+ mice (Fig. 3A), and were characteristic of type 1 diabetes progression in Roquin^san/san TCR+HEL+ mice, as high titers were not observed in diabetic TCR+HEL+ mice with mutations in Aire or Il-2 (Fig. 3C). Remarkably, the diabetes incidence in Roquin^san/san TCR+HEL+ mice rendered B cell deficient by a null mutation in Cd79a (Cd79a^ken) was no greater than in B cell–deficient Roquin^+/+ TCR+HEL+ mice (Fig. 3D).

The requirement for B cells in Roquin^san-accelerated autoimmune diabetes could reflect the function of B cells acting as antigen-presenting cells or the function of secreted IgG autoantibodies. To investigate this, we generated Roquin^san/san TCR+HEL+ mice in which high frequencies of B cells expressed HEL-specific IgM and IgD transgenes (MD4 [30]). Surprisingly, inheritance of IgMD^HEL decreased the diabetes incidence in Roquin^san/san TCR+HEL+ mice so that diabetes incidence was no greater than in Roquin^+/+ TCR+HEL+ mice (Fig. 3E). This result demonstrates that intact antigen presentation by B cells is not sufficient to drive Roquin^san-accelerated diabetes when the B cells cannot secrete islet-specific IgG.

Next, to test whether Roquin^san/san serum was sufficient to trigger Roquin^san-accelerated diabetes, we transferred serum from diabetic Roquin^san/san TCR+HEL+ mice or control serum pooled from diabetic Roquin^san/+ and Roquin^+/+ TCR+HEL+ mice into nondiabetic Roquin^san/+ TCR+HEL+ mice. Four of six recipients of Roquin^san/san serum developed diabetes within 25 days, whereas six recipients of the control serum remained nondiabetic (Fig. 3F).

We also investigated the role of anti-islet IgG in contexts other than the Roquin^san mutation. A unique feature of the breeding strategy used to produce the TCR+HEL+ animals represented in Fig. 4 was that one parent was TCR+HEL+ and had previously developed diabetes that was cured by transplanting nontransgenic islets lacking the HEL autoantigen. Once euglycemia was restored, the mouse was bred with a nontransgenic partner. Fifty-five percent of TCR+HEL+ offspring of diabetes-cured TCR+HEL+ mothers were diabetic by 150 days old, compared with 20% of otherwise matched offspring of diabetes-cured fathers (Fig. 4A). The increase in diabetes incidence occurred in both male and female offspring but was more severe in males (Fig. 4B). This demonstrates that there is a naturally occurring diabetogenic factor that is transmitted from mother to offspring throughout the breeding life of a formerly diabetic mother. Furthermore, this maternal factor discloses an unexpected sex dimorphism in susceptibility to diabetes in the TCR+HEL+ offspring. Anti-HEL IgG was present in the serum of diabetes-cured mothers around the time of transplantation and also >200 days after transplantation (Fig. 4C), indicating that their offspring were open to vertical transmission of anti-HEL IgG (43).

To test whether a maternally transmitted cofactor could sensitize the offspring to diabetes in a setting where mothers were themselves diabetes free, wild-type nontransgenic females were immunized with HEL antigen in CFA to elicit HEL-specific IgG antibodies and then mated with TCR+HEL+ males (Fig. 5A). The neonates had high titers of anti-HEL IgG, which declined by 70 days after birth, confirming their maternal origin (Fig. 5B). All of the TCR+HEL+ offspring developed diabetes 3–6 weeks after birth, whereas littermates lacking either or both transgenes did not become diabetic (Fig. 5C). Diabetes was not increased when the mothers were immunized with a control antigen OVA (Fig. 5D).

To determine whether anti-HEL IgG was sufficient to trigger diabetes, neonatal TCR+HEL+ mice from unmanipulated breeders were given serum, or in a separate experiment purified IgG, from mice immunized with HEL/CFA. Diabetes developed at 3–4 weeks of age in all neonates receiving HEL-immune serum and by 15 days of age in neonates receiving purified anti-HEL IgG but not in controls receiving anti-OVA serum or IgG (Fig. 5E and F).

That diabetes did not occur in HEL+ single-transgenic offspring suggests that the antibodies do not cause direct lysis of islets but rather act indirectly by enhancing the activation of islet-reactive CD4+ T cells. To test antibody-mediated effects on CD4+ cells, splenocytes from CD45.1 TCR+ transgenic donors were labeled with CFSE and injected into CD45.2 HEL+ or HEL– recipients. HEL+ recipients were given a course of injections of HEL-immune or control serum, and the fate of the HEL-specific T cells was tracked by flow cytometry of recipient spleen (Fig. 6A) using the ratio of donor CD4+ cells to donor B cells to normalize any mouse-to-mouse differences in the number of cells injected. Treatment of recipients with HEL-immune serum increased the numbers of HEL-reactive CD4+ cells present at days 3, 5, and 7 (Fig. 6B) and their CFSE dilution at each time-point analyzed (Fig. 6C).

To determine whether the HEL-immune serum enhances CD4+ cell proliferation and/or survival, the experiment was repeated using donor splenocytes that were Bim-deficient and therefore resistant to cell death (32). Interestingly, although the Bim-deficient T cells were triggered to proliferate specifically in the HEL+ recipients as expected, CFSE dilution within Bim-deficient T cells was equivalent regardless of whether mice were treated with HEL-immune or control serum (Fig. 6D). This suggests that HEL-immune serum does not enhance cell proliferation, but rather it inhibits the death of cells that have proliferated.

Antibody-mediated activation of T cells would appear to occur through antigen-presenting cells expressing receptors for the Fc portion of IgG (FcγR), given that CD4⁺ T cells themselves do not express FcγR. To test this, we injected neonates with three doses of anti-HEL serum on days 2, 4, and 6 after birth and either a monoclonal FcγR-blocking antibody or rat IgG2b isotype control on days 1, 3, and 5 and twice a week thereafter until day 17. FcγR blockade delayed the onset and reduced the incidence of diabetes (Fig. 6E). Thus, in the presence of islet-reactive CD4 T cells, anti-islet autoantibodies trigger diabetes via a mechanism that requires antibody recognition by FcγR+ cells.

Here we study the role of B cells in type 1 diabetes and show that anti-islet IgG autoantibodies accelerate CD4 T cell–driven diabetes. The anti-islet antibodies were insufficient to precipitate diabetes on their own, arguing against a direct antibody-mediated destruction of β-cells. Instead, they acted in an FcγR-mediated manner to enhance accumulation of islet-reactive CD4+ T cells and promote diabetes in animals that were genetically predisposed because of increased frequency of islet-reactive CD4+ T cells. The accelerated progression to type 1 diabetes in Roquin^san TCR+HEL+ animals with high-titer anti-islet antibodies demonstrates a principle of positive feedback between dysregulation of autoreactive helper T cells and the resulting autoantibodies.

Although the experimental evidence that B cells contribute to diabetes induction in NOD mice is strong, whether autoantibodies play a prominent role is unclear. Evidence for a contribution by autoantibodies came from reports of diabetes suppression when maternal antibody transfer was abrogated (22,23). New data have raised the question of whether the epigenetic factor is islet-specific autoantibody transmitted from mother to offspring or altered colonization of the offspring by microbial commensals whose interaction with the innate immune system influences diabetes incidence in NOD mice (44). Furthermore, maternal hyperglycemia can perturb fetal islet development (45), which could affect tolerance to the islets. Evidence against a role for antibodies came from the observation that fortnightly injections of 70 μg purified Ig from diabetic NOD mice could not induce diabetes in B cell–deficient NOD.Ighm^null mice (10), although it is possible that the concentration of active antibody achieved was suboptimal. Furthermore, anti-insulin VH IgM transgene (VH125Tg) restored diabetes in NOD.Ighm^null mice in the absence of measurable anti-insulin antibodies (12). Given the lack of diabetes enhancement observed here with an anti-HEL IgM transgene and dramatic enhancement by anti-HEL IgG antibody, these collective experimental findings imply that antigen presentation by islet-specific B cells and secretion of anti-islet IgG may both serve as positive feedbacks compounding the dysregulation of islet-reactive T cells but with varying significance in different individuals.

Cross-presentation of islet autoantigen by dendritic cells (DCs) to islet-specific CD8+ cells was enhanced by coinfusion of rabbit anti-islet antibodies, and this promoted diabetes (27). Our results reinforce and complement these findings by demonstrating that spontaneously produced anti-islet autoantibodies, either made by the animal itself or passively transferred as maternal antibody, can overcome thymic and peripheral tolerance mechanisms to trigger CD4+ cell–mediated autoimmune diabetes in an FcγR-dependent manner. Although the inhibition of HEL-immune serum-induced diabetes by anti-FcγR (2.4G2) reveals a crucial role for FcγRs, the specific FcγR+ cell types that are pivotal in unleashing the autoimmune cascade remain unclear. 2.4G2 binds to both the inhibitory FcγRIIb (CD32) and the activatory FcγRIII (CD16) molecules (46), but its immunosuppressive effect is presumably due to its binding to FcγRIII. Blockade of FcγRIII expressed by antigen-presenting cells would inhibit the uptake of antigen-antibody complexes and would prevent antigen-presenting cell maturation, but the blockade may also inhibit the activation of immune effector cells such as macrophages, natural killer cells, and granulocytes (47). Indeed, reinstating FcγR+ cells in either the afferent (purified DCs) or the efferent (purified natural killer cells) arm of the immune response was sufficient to increase diabetes in NOD mice lacking the FcR common γ chain (26). Although we cannot exclude a direct action on effector cells, increased autoantigen uptake by and maturation of DCs is likely to underlie the enhancement of islet-specific CD4+ T–cell survival by HEL-immune serum.

The role of B cells and autoantibodies in human type 1 diabetes is complex and poorly understood because although anti-islet autoantibodies are the best current predictors of future type 1 diabetes, the observational studies that followed multiple cohorts of subjects with either mothers or fathers with type 1 diabetes have revealed paradoxical protection associated with transmission of anti-islet autoantibodies—albeit dependent on HLA-DR status. Genome-wide association studies and studies of HLA polymorphisms indicate that the pathogenesis of human type 1 diabetes is likely to be heterogeneous, and hence it is conceivable that the diabetes-promoting role of B cells and anti-islet antibodies may not be uniform. Islet autoantibodies may be active cofactors in human type 1 diabetes pathogenesis once islet-reactive CD4+ T cells have escaped thymic or peripheral tolerance. In our transgenic mouse studies, this occurs congenitally because of the high frequency of islet-specific CD4 cells formed in the thymus even before birth, whereas it may only occur years after birth in humans, for example because of inherited gene variants of HLA-DR, CTLA4, or AIRE. The findings here, by defining mechanisms by which anti-islet IgG autoantibodies cooperate with incomplete thymic deletion to promote clonal expansion of islet-reactive CD4 T cells in vivo and progression to diabetes, provide an experimental foundation for future studies in humans and more complex rodent models where it is more difficult to trace the regulation of islet-specific CD4 cells.

The FcγR-blocking study also shows that inhibiting immune complex recognition by FcγR+ cells is attractive as a therapy. The first B cell–depleting trials with rituximab (rituxan, monoclonal CD20-specific antibody) in type 1 diabetes have reported responsiveness to therapy broadly comparable with anti-CD3 monoclonal antibody, highlighting an important role for B cells or antibody in type 1 diabetes in humans (9). Rituximab is highly effective in depleting recirculating and memory B cells and causes a rapid decrease in autoantibodies in a number of autoimmune diseases (48), although its effect on anti-islet antibodies in the recent phase 2 clinical trial was not described. Our study provides experimental evidence that it will be valuable both to measure anti-islet antibodies as an internal surrogate end point and pathogenic cofactor in future B cell–depleting trials, and to investigate clinical interventions that more specifically target autoantibody secretion or its effects on T-cell activation.

This work was funded by a Juvenile Diabetes Research Foundation (JDRF) regular research grant to C.G.V. and D.G.S., National Health and Medical Research Council program grant to C.G.V. and C.C.G., an A*STAR grant to K.-P.L. and C.C.G., a JDRF Program Grant, and a JDRF regular research grant to C.C.G. C.G.V. is supported by a Viertel Senior Medical Research Fellowship, and C.C.G. is an ARC Federation Fellow.

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

D.G.S. researched data. S.R.D. researched data and helped write the manuscript. J.H., S.K.L., C.E.T., and D.Y.H. conducted experiments. K.-P.L. helped plan the study. C.C.G. and C.G.V. planned the study and wrote the manuscript.

The authors thank Peter Milburn and the Biomolecular Resource facility at the John Curtin School of Medical Research for assistance with serum antibody purification.