Type 1 diabetes mellitus (T1DM) is a T cell-mediated autoimmune disease characterized by lymphocytic infiltration of the pancreatic islets of Langerhans with subsequent destruction of the insulin-producing beta cells 1. Non-obese diabetic (NOD) female mice, a murine model for T1DM, spontaneously develop diabetes by 30 weeks-of-age, with infiltrating cells appearing around the pancreatic islets as early as at 3–4 weeks-of-age 2.

T1DM susceptibility in the NOD mouse is linked to I-A^g7, the murine MHC class II gene that encodes a histidine at position 56 and a serine at position 57 in the β chain, in place of the more frequent proline 56β and aspartic acid 57β 3. The development of diabetes is prevented in NOD.PD mice (which are NOD mice with I-A^g7) that carry a β chain transgene with site-specific mutations that restore proline and aspartic acid at positions 56β and 57β, respectively 4. Furthermore, because of the two amino acid changes in the additional (transgenic) MHC class II allele β chain in NOD.PD mice, NOD.PD mice recognize three additional peptide epitopes in the glutamic acid decarboxylase 65 (GAD65) autoantigen 5.

Among beta-cell autoantigens, GAD65 is an important initial target of the immune response that results in beta-cell destruction and diabetes, in both humans and NOD mice 6789. While both humoral and cellular responses to GAD65 occur as early as 4 weeks of age in NOD mice 8, there is considerable evidence that beta-cell-specific T_H1 cells are the effectors of T1DM, whereas T_H2 cells appear to have a protective role 10. Accordingly, a shift of the autoimmune response from T_H1 to T_H2 predominance has represented a promising strategy for prevention of diabetes and other T_H1-mediated autoimmune diseases.

For example, administration of GAD65 to young NOD mice has been shown to prevent insulitis and diabetes 89, apparently via induction of CD4+ regulatory T cells with a T_H2 phenotype 10. Similarly, treatment with immunodominant peptides of myelin can prevent or reverse experimental autoimmune encephalomyelitis (EAE), a T_H1-associated inducible "autoimmune" disorder that is widely used as a model for human multiple sclerosis 111213.

Unfortunately, recent work indicates that the application of strategies to shift autoimmune responses from T_H1 to T_H2 predominance is not without risk. Thus, some of us recently showed that administration of two self peptides that can induce EAE, myelin proteolipid protein peptide 139 to 151 (PLP139-151) or myelin oligodendrocyte glycoprotein peptide 35–55 (MOG35-55), can result in severe anaphylactic reactions 14. This result clearly indicated that severe allergic reactions to self peptides can occur in mice that have been induced to express pathology (i.e., EAE) related to "autoimmunity" to these peptides. However, it was initially unclear whether anaphylactic reactivity also could be elicited to self peptides that have been implicated in the development of a spontaneous autoimmune disorder.

In the present study, we show that anti-peptide autoantibodies and fatal anaphylactic reactions can be elicited by immunodominant GAD65 peptides in NOD mice that have been injected with these peptides intraperitoneally in incomplete Freund's adjuvant (IFA), as part of an attempt to induce "tolerance" and prevent the spontaneous development of T1DM. Moreover, while this manuscript was in review, Liu and colleagues reported that anti-peptide autoantibodies and fatal anaphylaxis can be induced in NOD mice that have been immunized with insulin B chain peptides B:9–23 or B:13–23 15. However, in the Liu et al. study, the peptides were administered subcutaneously in saline without adjuvant. As reviewed in Liu et al., 15 several lines of evidence indicate that amino acids 9–23 of the insulin B chain also represent a major target of anti-islet autoimmunity in T1DM. Taken together with the findings reported herein, this work indicates that anaphylactic reactions can be elicited in mice that have been immunized with pancreatic islet-associated self-peptides that also represent significant targets of autoimmunity in T1DM.

This study clearly demonstrates that i.p. immunization of NOD mice with preparations of GAD65 self peptides in IFA can cause a marked shift towards a T_H2 like response, as reflected by high levels of IgG1. Similarly, Liu et al. recently demonstrated that strong IgG1 responses can be induced in NOD mice that have been immunized subcutaneously with insulin B chain peptides administered in physiological saline 15. However, both studies showed that anaphylaxis can be induced in such mice upon subsequent re-challenge with preparations of the peptides used for immunization 15. Moreover, the anaphylactic reactions in mice that had been immunized and challenged with G7 peptides were severe, with reductions in body temperature that were very similar to those observed in mice exhibiting IgE-dependent passive systemic anaphylaxis and with a very high fatality rate (12/14 mice, or 86%). Anaphylaxis also developed in some NOD mice that had been immunized and challenged with preparations of PD peptides (that are not immunodominant in NOD mice), although both the drop in body temperature and the death rate in these mice were significantly less than those observed in the mice immunized and challenged with G7 peptides.

There were both similarities and differences between our findings in the NOD mouse model of T1DM and those in the EAE model 14. Expression of EAE requires specific immunization with self peptides (e.g., PLP 139–151 or MOG35-55), and these peptides generally are administered in complete Freund's adjuvant (CFA). By contrast, T1DM develops spontaneously in NOD mice. On the other hand, induction of anaphylactic reactivity in NOD mice appeared to require immunization of the mice with GAD65 peptides (in this model, in IFA), as naïve NOD mice challenged with G7 peptides exhibited no detectible reactions, and none of them died (Table 1). Thus, in both the EAE model 14 and the NOD T1DM model (this study, and that of Liu et al., 15) some form of artificial "immunization" with a self peptide preparation appears to be required for the development of anaphylactic reactivity to "self". This of course is not a surprising result. Indeed, it is challenging to conceive of any possible selective advantage that would be conferred by a propensity to develop, under "natural" conditions, potentially fatal allergic reactions to components of self. It remains to be determined whether self peptide immunization protocols that induce anaphylactic reactivity do so simply because of the manner in which they present large amounts of self peptides to the immune system, or because of other factors, such as the presence in the peptide preparations of aggregates or other components beside self peptide monomers.

Whatever the underlying reason(s) for the development of anaphylactic reactivity to these self peptide preparations, in both the EAE and the NOD T1DM models, anaphylactic reactions occurred in mice that had developed strong IgG1 responses to the relevant self peptides, with only modest changes in total IgE levels. In humans, antigen-specific anaphylactic reactivity is thought to be mediated solely (or primarily) by IgE antibodies, whereas it has long been known that either IgE or IgG1 antibodies can mediate anaphylaxis in mice (reviewed in 1415 and 17). However, it has been reported that IgG1-dependent anaphylaxis in the mouse is associated with substantially less histological evidence of mast cell degranulation than is observed in IgE-dependent anaphylaxis in that species 17. In neither of the models of "autoimmunity" that we have studied (i.e., EAE, T1DM in NOD mice) was anaphylaxis associated with histological evidence of substantial mast cell degranulation 14 (data not shown). Taken together, these findings suggest that IgG1 antibodies contribute importantly to the development of anaphylaxis in both of these models. On the other hand, we can not rule out some role for IgE antibodies in these reactions.

Indeed, Liu et al. 15 found that, in NOD mice that had been immunized with peptide B:9–23, treatment with both anti FcγRII/RIII and anti-IgE monoclonal antibodies was required to prevent anaphylaxis upon challenge with the peptide. Interestingly, however, Liu et al. 15 did not detect IgE antibodies to B:9–23 or B:13–23 in the serum of their NOD mice. By contrast, mice that had been immunized with B:9–23 peptide at 10 or 100 μg/dose exhibited a robust and dose-dependent IgG1 antibody response to the peptide 15. Thus, in both B:9–23 peptide-associated anaphylaxis (Liu et al. 15) and GAD65 peptide-associated anaphylaxis (our study), anti-peptide IgG1 antibodies contribute to the response. However, IgE antibodies also appear to contribute to anaphylaxis to B:9–23 peptides 15, and may also be involved in our model.

One point not yet clarified by the comparison of the present results, those of Liu et al. 15, and those of Pedotti et al.14 is whether the influence of thymic expression of the self peptide on the propensity to develop anaphylactic reactivity differs in the EAE and NOD T1DM models. In the study by Pedotti et al. 14, it was noted that the two self peptides that induced anaphylactic reactivity, MOG 35–55 and PLP139–151, are not expressed in the thymus, whereas the two peptides tested that did not induce anaphylactic reactivity, PLP_p 178–191 and MBPA_C1–11, are expressed at that site. However, both GAD65 and GAD67 mRNA can be detected in the thymic medullary epithelial cells in mice 24. Thus, despite thymic expression of GAD65 and GAD67 at the level of mRNA, NOD mice spontaneously develop autoreactivity to these islet (and brain) expressed proteins, and re-challenge of mice that have been immunized with peptides from GAD65 results in severe anaphylactic reactions. On the other hand, expression of GAD65 or GAD67 protein in the thymus has not yet been reported. Similarly, as reviewed in Liu et al. 15, although several lines of evidence indicate that insulin is present in the thymus of mice and humans, it is possible that the specific insulin peptides that induced anaphylaxis in their study are not ordinarily present in that site. As a result, it has not yet been demonstrated that anaphylactic reactions can develop to self peptides that are expressed in the thymus.

It should be emphasized that NOD mice have a partial defect in thymic negative selection 2526, a defect in FcγRIIB (that can negatively regulate anaphylactic reactions 2728), and perhaps other genetic polymorphisms that may result in immunological hyperresponsiveness. The same is likely to be true in at least some patients with type 1 diabetes, and in patients in the pre-diabetic phase. Therefore, because of the risk of induction of anaphylactic sensitization, extreme caution needs to be used in developing any type of antigen-specific immunosuppressive therapy for the prevention or treatment of T1DM. This caution probably should be extended to all attempts to shift immune responses to self or foreign antigens from a T_H1 to a T_H2 response. Indeed, in a recent phase II clinical trial, 9% of MS patients given an altered peptide ligand (APL) of a myelin basic protein epitope developed immediate hypersensitivity reactions after multiple injections of the APL 29. Thus, it would appear that great care must be taken when injecting preparations of putative "tolerogens" in attempts to suppress T_H1-mediated autoimmune diseases.

Rosetta Pedotti and Maija Sanna participated in the design of the experiments and performed the peptide immunizations and challenges in the mice, measurements of anaphylactic responses and ELISA immunoassays for antibodies. Rosetta Pedotti and Maija Sanna contributed equally to this study, including collaborating in writing the first draft of the manuscript. Mindy Tsai participated in the design and execution of the study and the drafting of the manuscript. Jason DeVoss performed some of the ELISA immunoassays for antibodies. Lawrence Steinman, Hugh McDevitt, and Stephen J. Galli participated in experimental design, interpretation of the results, and revision of the manuscript. All authors read and approved the final version of the manuscript.