Polyspecificity of autoimmune responses in type 1 (autoimmune) diabetes

Autoimmunity is, by definition, supposed to be antigen specific [1], yet for almost 50 years, associations have been recognized between different autoimmune diseases [2–5]. For example, the increased incidence of gastric and thyroid autoimmunity in type 1 diabetes was first reported in 1963 [6]. These associations are in addition to the occasional cases of the autoimmune polyglandular syndromes [7] and are found independently of chronic mucocutaneous moniliasis, which is characteristic of the Type I syndrome (APECED), and Addison's disease which invariably occurs in the Type II syndrome [8].

In the 1980s, Riley et al. [9] reported that about 5% of young patients with type 1 diabetes had achlorhydria (lack of gastric acid) secondary to autoimmune gastritis. About twice that number had gastric parietal cell autoantibodies, while the frequency of gastric parietal cell autoantibodies in healthy children the same age was only 2%. In a separate study, thyroid function tests of young diabetic patients revealed that about 1% had Grave's disease (autoimmune hyperthyroidism) and about 7% had Hashimoto's disease (autoimmune thyroiditis). Again, another 10% of patients tested positive for thyroid microsomal autoantibodies without concurrent evidence of clinical disease [10]. This frequency of antithyroid autoantibodies is about five times that found in healthy children of a similar age [11]. Thyroid microsomal autoantibodies were even more frequent in patients with both type 1 diabetes and autoimmune gastritis, affecting 46% of Caucasian, and 25% of black subjects [9].

In this issue of the Journal, de Block et al. [12] update these findings. Their studies of 399 Belgian patients with type 1 diabetes identified parietal cell (PC) autoantibodies in 18% of the subjects and antithyroid peroxidase (TPO) autoantibodies in 22%, with 7% receiving treatment for hyper- or hypothyroidism. The concordance between these figures and the studies completed almost 20 years ago is extraordinary. The authors also identified a strong association between thyroid or gastric autoimmunity and the presence of anti-GAD (glutamic acid decarboxylase-65) autoantibodies in diabetic patients. While 25% of type 1 diabetic patients with anti-GAD autoantibodies had anti-TPO antibodies, only 15% of those without them did. Similarly, while 21% with GAD antibodies had anti-PC antibodies, only 12% of those without them did. Thus, the presence of GAD antibodies was associated with an almost two-fold greater risk of reactivity to each of these antigens than for type 1 diabetes in the absence of GAD seroreactivity.

There are at least two potential classes of explanation for this association; it may result from either downstream or upstream causality. In the former case, one may hypothesize that autoimmune reactivity to pancreatic islet components can contribute to the risk of developing additional autoreactivity and disease. The thyroid and stomach may share certain antigens with the pancreatic islets of Langerhans, and it is perhaps in this context that the relationship with anti-GAD autoantibodies can be understood. GAD-65 and -67 are the enzymes that catalyse the conversion of glutamate to gamma-aminobutyric acid (GABA), which acts as an inhibitory neurotransmitter. The presence of GAD-65 in the islets therefore reflects these organs' role as neuroendocrine tissue. GAD also plays a significant role in the stomach, as it has been identified in the myenteric plexus, the circular muscular layer, the submucosa and the lamina propria of the mucosa [13]. Although the major autoantigen in the stomach is the H/K-ATPase (proton pump) of parietal cells [14], and GABA induces acid secretion by this molecule [15], one must still postulate the induction of gastric tissue damage, release of sequestered antigen, local priming and antigen spreading of the immune response to explain this shift in the predominant immune reactivity. In the case of the thyroid gland, GABA regulates thyroid hormone secretion [16] and is concentrated in follicle cells [17], suggesting that similar sequelae may occur in this organ. It is therefore possible that once a T cell response to GAD has been primed in the pancreas or its draining lymph nodes, the resulting activated T cells can then initiate damage of other neuroendocrine tissues containing the same or similar enzyme(s).

Alternatively, upstream causality as an explanation for the increased frequency of other autoimmune diseases in patients with type 1 diabetes may exist in the form of a common aetiological factor – most probably genetic. Consistent with this hypothesis, Becker et al. [18,19] found that approximately 65% of human linkages for different autoimmune diseases mapped nonrandomly into 18 distinct clusters. For example, Grave's disease has been mapped to the same region on chromosome 14q as the diabetes susceptibility gene IDDM11 [20], as well as to the same region of chromosome 2q as IDDM12 [21]. To date, autoimmune gastritis has not been mapped in humans, although we have found that susceptibility to gastritis in the BALB/c mouse model maps to two loci, Gasa1 and Gasa2 on distal chromosome 4, which colocalize with the NOD mouse diabetes susceptibility genes Idd9 and Idd11, respectively [22]. Indeed, because of the practical difficulties involved in trying to dissect such associations in clinical studies, it may be that the issue of shared genetic origins for multiple autoimmune diseases can only be adequately tested in animal models.

In addition to type 1 diabetes, NOD mice are susceptible to spontaneous thyroiditis [23], haemolytic anaemia [24] and Sjögren's syndrome (autoimmune destruction of lacrimal and salivary glands) [25,26], as well as to the induction of experimental allergic encephalomyelitis (EAE, an experimental model of multiple sclerosis) [27,28] and systemic lupus erythematosus [29–31]. The coexistance of susceptibility to Sjögren's syndrome, haemolytic anaemia and lupus in the NOD mouse is consistent with the observation that Sjögren's syndrome in humans frequently occurs secondary to lupus [2]. Similarly, thyroiditis occurs in about one fifth of patients with Sjögren's syndrome [32], and is at least three times more common in women with multiple sclerosis than in female controls [33].

Perhaps one of the most impressive associations occurring in NOD mice is that exposure to mycobacteria prevents diabetes while precipitating lupus in the same animals [34]. This model may therefore represent an example of an environmental switch between two clinical expressions of an underlying tendency to autoimmune disease in which susceptibility to both diabetes and lupus are encoded by the same genes. With this ‘common gene’ hypothesis in mind, we mapped the genetic segments conferring susceptibility to mycobacterial-induced lupus in NOD mice, in order to compare the locations of lupus susceptibility genes to those known to confer susceptibility to type 1 diabetes [35]. Surprisingly, linkage for diabetes and lupus only coincided at the MHC. None of the other lupus susceptibility loci identified were linked to known diabetes susceptibility genes, suggesting that these two diseases were independently inherited in this model. In contrast, Encinas et al. [36] found that both diabetes and EAE were suppressed in congenic NOD mice carrying a 0·15-cm chromosome 3 segment derived from the disease resistant C57BL/6 strain. The difference between these two experimental systems may reflect the much stronger associations that EAE and multiple sclerosis have to autoimmune diabetes [28], compared to that of lupus.

Thus the reasons for the strong associations observed between some autoimmune diseases currently remain unresolved. It seems likely, however, that these associations reflect important aetiological relationships, and that improved understanding of their natures will significantly contribute to our ability to predict and prevent autoimmune diseases.