Date Published: February 22, 2012
Publisher: Hindawi Publishing Corporation
Author(s): Norio Kanatsuna, George K. Papadopoulos, Antonis K. Moustakas, Åke Lenmark.
Autoimmunity against pancreatic islet beta cells is strongly associated with proinsulin, insulin, or both. The insulin autoreactivity is particularly pronounced in children with young age at onset of type 1 diabetes. Possible mechanisms for (pro)insulin autoimmunity may involve beta-cell destruction resulting in proinsulin peptide presentation on HLA-DR-DQ Class II molecules in pancreatic draining lymphnodes. Recent data on proinsulin peptide binding to type 1 diabetes-associated HLA-DQ2 and -DQ8 is reviewed and illustrated by molecular modeling. The importance of the cellular immune reaction involving cytotoxic CD8-positive T cells to kill beta cells through Class I MHC is discussed along with speculations of the possible role of B lymphocytes in presenting the proinsulin autoantigen over and over again through insulin-carrying insulin autoantibodies. In contrast to autoantibodies against other islet autoantigens such as GAD65, IA-2, and ZnT8 transporters, it has not been possible yet to standardize the insulin autoantibody test. As islet autoantibodies predict type 1 diabetes, it is imperative to clarify the mechanisms of insulin autoimmunity.
The pancreatic islets constitute about 2-3% of the pancreas weight that is about 100 grams in adults . The islets represent the endocrine portion of the pancreas and are present as more than a million well-defined cellular clusters throughout the pancreas [2, 3]. Each pancreatic islet (Figure 1) is composed of about 54% beta cells, 35% alpha cells, and 11% delta cells in addition to connective tissue and capillary cells . Proinsulin, converted to insulin (Figure 2), is the major hormone produced in the beta cells while glucagon and GLP-1 are produced by the alpha cells, somatostatin by the delta cells, and pancreatic polypeptide by the PP cells. Pancreatic islet cells are also reported to produce ghrelin , apelin [6, 7], and CART [8–10]. These polypeptide hormones may be coexpressed with insulin in the beta cells or with other hormone-producing cells . PP cells are more often seen in the head of the pancreas, while alpha cells dominate the tail [11, 12]. Insulin is the life-saving hormone for people suffering from type 1 and at times type 2 diabetes (see what follows). More beta cells are available than necessary to main blood glucose at normal levels. However, loss of insulin has catastrophic consequences. It has been estimated that 50% of the pancreas may be removed by surgery without a development of diabetes [13, 14]. Type 1 diabetes (T1D) is an autoimmune disease leading to a progressive loss of beta cells as they are attacked by the patients’ own immune system (for reviews see [15–18]). T1D has a prodromal stage of islet autoimmunity. Children who develop islet autoantibodies against all four autoantigens: insulin, GAD65, IA-2, or ZnT8 (Table 1), before 3–5 years of age, tend to have a shorter prodrome prior to the clinical onset than older children, young adults, or adults . These individuals may have multiple islet autoantibodies for years before the clinical onset of the disease . GAD65, not insulin, autoantibodies characterize patients with latent autoimmune diabetes in adults (LADA) [15–18]. It has been estimated that although an individual may be positive for islet autoantibodies for months to years, the clinical onset does not occur until 80–90% of the beta cells have been killed . Hence, T1D appears due to the selective autoimmune destruction of the pancreatic beta cells [16, 22]. The major genetic factor for T1D is the HLA-DQ locus on chromosome 6p21 . Recent reviews can be found in [24, 25]. The association between the HLA Class II genes and T1D is well established and several HLA-DQ genotypes have been used to randomize newborn children to follow up investigations of the development of islet autoantibodies [26–30]. All over the world, the majority (80–90%) of newly diagnosed T1D children do not have a first-degree relative (father, mother, or sibling) already affected by the disease. The presence of certain HLA-DQ already at birth confers the genetic risk for T1D (Table 2). The highest risk is conferred by the HLA-DQ2/8 genotype. The risk for T1D with this genotype is highest in the young but is markedly decreasing with increasing age [31, 32]. Affected sib-pairs with T1D share HLA alleles more often than expected, and alleles at the Class II DR and DQ loci are not only associated with susceptibility to but also negatively associated with T1D and therefore offer at least partial protection . In a large population-based study the HLA DQ A1*01:02-B1*06:02 (DQ6.2) was rarely found among T1D children below the age of 10; however, the negative association was decreased with increasing age and lost at 30 years of age . It is noted that other HLA genotypes, often with somewhat similar physicochemical properties confer T1D risk in other populations such as in Japan and China (Table 2) [35–39]. As indicated the risk for T1D conferred by HLA-DQ is dependent on age. It is therefore important that autoantibodies against insulin are not only present particularly in young children at the time of clinical diagnosis of T1D but also prior to the clinical onset [17, 40, 41]. As will be reviewed the autoimmune reaction against insulin in T1D has been mapped in terms of both cellular [42, 43] and humoral [17, 44] recognition. However, insulin is a target not only in T1D but also in other autoimmune conditions. In Hirata’s disease insulin autoantibodies are detected in association with hypoglycemia in the patient . This disease is also associated with HLA Class II (Table 2) [46, 47]. The detailed mechanisms by which patients recognize their own insulin as an autoantigen may therefore have vastly different consequences for the patient and these differences will be discussed in the present paper. The reader is referred to the following reviews where insulin autoimmunity in T1D [17, 48, 49] or in the insulin autoimmune syndrome [47, 50, 51] has previously been reviewed.
T1D may be viewed as a two-step disease. The first step is the initiation of islet autoimmunity; the second step is precipitation of diabetes when islet autoimmunity has caused a major β-cell loss (>80%), and insulin deficiency becomes clinically manifest. The pancreatic beta cells are destroyed in an aggressive autoimmune process. The immunopathogenesis of T1D is associated with T-lymphocyte autoimmunity, and the disease is often referred to as a T-cell-mediated disease [79–81]. This is somewhat self-evident as an immune response cannot be initiated without the help from CD4+ positive T-helper cells. Also it is rare that an immune response does not engage all cells in the immune system as cytotoxic CD8+ T cells are not able to develop without the help from CD4+ T-helper cells. These cells are also critical for the activation of B cells to differentiate into autoantibody-producing plasma cells. The importance of both T and B cells in the pathogenesis of T1D is illustrated in recent clinical trials [82, 83]. Monoclonal antibody therapeutics, depleting T cells (CD3 antibodies) or B cells (CD20 antibodies; Rituximab), had similar effects to transiently inhibit the progression of beta-cell loss after the clinical onset of T1D measured as residual beta-cell function [82–86].
Most patients with T1D have islet autoantibodies at the time of clinical diagnosis. Autoimmune diabetes rather than T1D would therefore be a more appropriate designation of the disorder. Several authors have reported that about 10–15% of newly diagnosed patients with diabetes classified with T1D have no islet autoantibodies at the time of clinical onset. The question will then arise if patients have had islet autoantibodies, which disappeared prior to the clinical onset. The use of autoantibody tests against ICA, insulin, GAD65, IA-2, and the three variants of ZnT8 as well as islet cell antibodies (ICAs) by indirect immunofluorescence, in more 600 newly diagnosed T1D children, indicates that only 5% did not have any of the seven different types of autoantibodies . It was not possible to determine if these children have had autoantibodies and lost them prior to diagnosis. However, in children born to mothers with islet autoantibodies during pregnancy, it was found that such children tended to be negative at the time of clinical onset . Presence of islet autoantibodies at birth may explain why some T1D children are islet autoantibody negative at clinical diagnosis .
Insulin autoantibodies are primarily detected in children below the age of 5 years [15–17]. In Kappa statistics of agreement there was a moderate to fair agreement between any pairs of autoantibodies against GAD65, IA-2, or ZnT8 (W,R,Q) (Table 1), while insulin autoantibodies showed only a slight agreement with any combination . It is often observed that insulin autoantibodies are the first to appear, at least in children younger than 3–5 years of age. However, it has been difficult to dissect the sequence of events that leads to the formation of insulin autoantibodies in very young children. One could envisage the following scenario. Beta cells would be killed, perhaps lysed by a virus infection. The dead beta cells or remnants thereof would be engulfed by APC. These cells are activated and migrate through the lymphatic system to the lymph nodes that drain the pancreas. Antigen presentation to CD4+ T cells would take place in the lymph node. It is possible that the antigen presentation is particularly effective in small children leading to an early insulin autoantibody response . This hypothetical mechanism is consistent with studies in experimental animals [132, 133].
CD8+ T-cytotoxic cells directed against insulin peptides expressed on MHC Class I have been described both in the NOD mouse and in man [100, 101, 141, 142]. Remarkably, the two NOD mouse epitopes InsB15-23 and InsB25-C34, respectively, bind either very weakly or very strongly to the restriction element, H2-Kd [70, 142]. The preproinsulin epitopes to several HLA-A/B alleles (including the most frequent allele in the Caucasian population, -A2) (Figures 3(b)–3(d)) span the entire molecule, and the frequency of reactivity to the different epitopes varies [69, 70, 89–91]. Remarkably, occasional high responses to certain peptides are also seen in controls (SI > 4), with no other sign of autoimmunity . In a pioneering study on in situ reactivity of persons at onset of type 1 diabetes and patients with long-standing disease, it has been shown that CTLs in HLA-A2+ individuals showed reactivity to single epitopes from 6 different autoantigens (preproinsulin included, epitope 15–23). There was an inversely proportional staining of pancreases with HLA-A2 tetramers with respect to age from diagnosis. In fact, no such reactivity was detected in any patients with over 10 years time from the date of onset of type 1 diabetes .
An alternative pathway to the formation of IAA is illustrated in Figure 4. This pathway remains to be fully explored in humans. The clinical trial with Rituximab (CD20 monoclonal antibody) in newly diagnosed T1D children demonstrates that depletion of B lymphocytes was associated with a significant preservation of mixed meal-stimulated C-peptide . The contribution of B lymphocytes to T1D pathogenesis may have been overlooked. As illustrated in Figure 4, B lymphocytes with an antigen receptor recognizing insulin would take up the insulin and process it to be presented on HLA-DR, -DQ, or both. The trimolecular complex with insulin would next be recognized by a TCR on the surface of a mature, matching CD4+ T-helper cell. Upon the cell-to-cell contact the CD4+ T-helper cell is activated to produce cytokines (such as IL-4 or IL-10). These cytokines would help the B cell to differentiate, replicate, and mature into an IAA producing B lymphocyte, and eventually turning into a plasma cell. It is important to note that Rituximab treatment appeared to reduce antibody formation to new antigens such as the bacteriophage PhiX174 . It was suggested that Rituximab decreased both antibody production and isotype switching . However, at the same time as residual C-peptide was preserved , Rituximab suppressed IA but not the levels of postdiagnosis GADA, IA-2A, and ZnT8A . In the European-Canadian cyclosporine trial, it was demonstrated that cyclosporine reduced the formation of insulin antibodies in response to the regular insulin therapy given to all the participating T1D patients . It is of interest in this regard that Rituximab-treated patients were thought to be able to develop immunological tolerance to bacteriophage PhiX174 . In Stiff Person Syndrome, Rituximab was reducing GADA in some  but not in all  patients. Further studies are warranted to determine the interaction between APC, T-helper cells, and B lymphocytes. It needs to be established to what extent B lymphocytes may be acting as APC that either initiate, maintain, or both, the autoimmune response to insulin in children.
Measurement of IAA was initially limited by the large serum volume required for the early immunoprecipitation assays, which used polyethylene glycol to separate immune complexes . The first IAA assay required one milliliter serum or plasma . Insulin was labeled by 125I in an approach similar to that which had been used for both regular insulin radioimmunoassays as well as for insulin-receptor-binding experiments . Later it was found that labeling of multiple tyrosine residues compromised both antibody—as well as receptor binding . These observations resulted in the now established use of only insulin that is monoiodinated at position A14 . The improvement in insulin iodination procedures [154, 155] made it possible to develop alternative radiobinding assays that required less serum. This type of microassay allowed a major reduction of the amount of serum used and has improved assay specificity [156, 157].
The IAA radioimmunoassay was first tested in serum or plasma samples from siblings to first-degree relatives with T1D. These siblings, including monozygotic twins or triplets, were followed longitudinally for the appearance of IAA and other islet autoantibodies [163, 164], in larger prospective studies such as BABYDIAB [87, 165], DIPP [29, 166], and DAISY [167, 168]. IAA was reported to show an association between levels and risk for T1D, which was not observed for GADA or IA-2A . These observations seemed also to be corroborated in studies of children at genetic risk for T1D based on HLA typing rather than having a first-degree relative with the disease . In the Diabetes Prevention Trial-1 (DPT-1) , GADA, IA-2A were measured along with ICA and IAA . No subjects with IAA as single autoantibodies developed T1D . When a second autoantibody appeared, any other autoantibody except IAA was added significantly to the prediction of T1D . In the DIPP study of children born with high-risk HLA, IAA tended to be the first autoantibody to appear . It is therefore possible that the initiation of the T1D disease process may involve insulin itself or proinsulin, perhaps also preproinsulin [20, 41, 87, 171]. However, most authors suggest that the number of islet autoantibodies is the strongest predictor of clinical onset of T1D . It can however not be excluded that IAA affinity may be a better predictor for T1D in children with multiple autoantibodies [74, 172]. Indeed, high-affinity cell surface antibody on B lymphocytes readily promotes their differentiation and proliferation upon antigen binding, in contrast to low-affinity antibody .
There is a paucity of detailed investigations to clarify IAA epitopes of proinsulin and insulin (Table 3). There is a lack of information to what extent HLA-DQ or DR are associated with IAA binding to either A chain, B chain, or proinsulin autoantibody epitopes. Similar to HLA Class I peptide binding (Figures 3(b)–3(d)), IAA was reported to recognize the A8–A10 (13) epitope [74, 76, 77]. It is not clear why IAA would recognize the same epitope as might be presented on HLA Class I molecules. The B1–B3  as well as the B3 position [72, 73], both presented on HLA Class II molecules (Table 2) may also be recognized by IAA. Studies with systematic site-directed mutagenesis of the preproinsulin cDNA may prove useful to map the IAA binding site more carefully in relation to the HLA-DQ and DR genotypes of newly diagnosed, non-insulin-treated T1D patients, or IAA-positive nondiabetic subjects. Such studies are also important as it has been suggested that the IAA levels may be the best predictor of clinical onset in young children  as well as in children born to mothers with T1D in the BABY DIAB study . Further studies are also needed to determine epitope specificity in relation to the apparent polyclonal nature of IAA and their similarity to the insulin antibodies (IA) detected after insulin therapy has been initiated [174, 175].
The diagnostic sensitivity of IAA for T1D is on the average only about 30%  but varies with the age at onset. In children below the age of 3 years the diagnostic sensitivity may be as high as 50–60%  but decreases to about 10% in T1D patients diagnosed after 20 years of age. It was estimated that IAs produced in response to the insulin treatment appear after about 7 days . When comparing binding characteristics between IAA and IA, it was found that the two antibody types were comparable in several affinity tests . The authors therefore concluded that both IAA and IA were polyclonal in nature and that both developed in response to insulin as the antigen . In some individuals, it is therefore possible that insulin itself is able to break the immunological tolerance to allow the formation of IAA. It follows that it cannot be excluded that insulin treatment itself may induce a T1D disease process. For example, insulin given to type 2 diabetes patients in Japan was thought to induce T1D . In these patients insulin antibodies (IAs) of high titer were detected at or after the development of insulin deficiency. These IAs were characterized by an extremely high-affinity and a very low-binding capacity. The characteristics of these insulin-treatment-induced IA were thought to be similar to the IA found in the insulin autoimmune syndrome . The insulin aspart had comparable immunogenicity to human insulin, and antibodies developing in response to either insulin seemed to cross-react [177–180]. High titer insulin antibodies requiring immunosuppression have been reported . Epitope-specific insulin antibodies may develop in some patients who showed benefit when one insulin analogue was replaced by another [181–183]. Similar to animal insulins, also insulin analogues such as Lispro insulin may cause insulin allergy .
The insulin gene (INS) region is an established T1D susceptibility locus. The variable nucleotide tandem repeat (VNTR) in the promoter region of the insulin gene may contribute to T1D possibly by mechanisms of central tolerance . The INS VNTR is composed of 14 to 15 bp variant repeats. The shortest (Class I) variable number of tandem repeat (VNTR) alleles was found to increase, whereas the longest (Class III) alleles were observed to decrease in the patients in comparison to the controls . The possible role of central tolerance is illustrated by the observation that IAAs in newly diagnosed T1D patients was found to be associated with the INS VNTR polymorphism in some [34, 186] but not all studies . In children born to mothers with T1D, it was reported that the combination of genotyping for high-risk HLA-DQ (e.g., HLA-DQ2/8 and 8/8) and INS VNTR identified a minority of children with an increased T1D risk . One study compared the INS VNTR polymorphism between Finland and Sweden . The T1D risk genotypes (Class I/I and I/III) were significantly more common in Finland than in Sweden, both among patients and controls. Class III homozygous genotypes showed varying degrees of protective effect due to polymorphisms within Class III. These observations suggest that heterogeneity between protective Class III lineages could exist.
The insulin autoimmune syndrome (IAS, Hirata disease) is characterized by a combination of fasting and sometimes postprandial hypoglycemia, high serum concentrations of total immunoreactive insulin, and presence in the serum of polyclonal autoantibodies against native human insulin [45, 215].
Humoral autoimmunity against insulin, first described in 1983 is established but needs to be better defined. The current IAA radioimmunoassay continues to perform poorly in the DASP [158, 159]. The current interlaboratory variation is simply too large to allow valuable comparisons between laboratories throughout the world. The autoreactivity against (pro)insulin also needs to be better defined. It will be necessary to clarify the way by which (prepro)insulin released from dead or damaged beta cells is taken up by APC, processed and finally presented on HLA-DR and -DQ molecules. Is antigen-presentation by—DQ more critical than—DR to induce autoreactivity? Once IAAs have been formed, it will be critical to define the role of IAA-producing B cells and plasma cells. What is the role of B cells as APC in the disease process? Both CD4+ T-helper and regulatory T cells specific for (pro)insulin need also to be identified in humans at risk for T1D; the former have been shown to exist at the population but not at the clonal level. Insulin-specific CD8+ T cells may be critical to identify in children at increased risk for T1D as such children tend to develop IAA early, while CD8+ Tregs may be able to control the action of diabetogenic self-reactive immunocytes. It will be important to compare the insulin autoantibodies IAS with the IAA in T1D. The IAS is characterized by a combination of fasting hypoglycemia, high concentration of total serum immunoreactive insulin, and presence of autoantibodies to native human insulin in serum . The release of insulin from the IAS insulin autoantibodies may cause hypoglycemia, and further studies are needed to explain why this type of autoantibodies may be related to hypoglycemia with no apparent loss of beta cells.