Type 1 Diabetes: Cellular, Molecular & Clinical Immunology

Chapter 10 - Humoral Autoimmunity
Liping Yu, MD and George S. Eisenbarth, MD, PhD.

Updated 6/09, slides updated 4/08, new slides #1 - #4 Click to download Powerpoint slide set
Underlined references are clickable links and will take you to the reference on the PubMed website.

Introduction
General Assay Methodology
Islet Cell Autoantibodies(ICAs)
Sequenced Autoantigens
            Glutamic Acid Decarboxylase
            ICA512(IA-2) and IA-2β (phogrin)
            Glycolipid Autoantigens
            Insulin
            ZnT8
            ICA69
            Carboxypeptidase H
            ICA12 (SOX13)
Partially Characterized Autoantigens
Other Autoantigens
T Cell Reactivity
Islet Transplantation
Conclusions
References      

Introduction
A little more than a decade ago, screening for autoantibodies associated with Type 1A (immune mediated) diabetes was limited to measuring cytoplasmic islet cell autoantibodies (ICAs) and insulin antibodies (1;2). Currently, multiple sequenced autoantigens have been defined and recombinant autoantibody assays (3;4)  are available (5;6) and include insulin (7), glutamic acid decarboxylase (GAD) (8), ICA512/IA-2 (9,10,11), I-A2 beta (phogrin) (12;13), the islet zinc transporter, ZnT8 (14;15)  and carboxypeptidase H (16). In addition, there are many other autoantigens in a variety of stages of characterization, including molecules with characterized sequences but without fully developed assays and proteins with only known molecular weights of 155 kd (17;18), 52 kd (19), and molecules recognized by T lymphocytes for which autoantibodies have not been described such as chromagranin A (oral presentation-Haskins et al: IDS 2009 Target of BDC2.5 T cell NOD clone) and IGRP (20;21) and RegII (22). Other molecules have been described but their association with Type 1A diabetes has either not been evaluated in man, has not been substantiated, or follow-up studies have not been published. These molecules include anti-bovine serum albumin antibodies (anti-BSA) (23), antibodies reacting with ICA69 (24,25,26), anti-insulin receptor antibodies (27), antibodies to heat shock proteins (28,29,30), anti-topoisomerase II (31) anti-ganglioside (32), lysophospholipids (33) and GLIMA38 a membrane glycoprotein (34) and antibodies to a series of autoantigens identified by screening islet libraries such as ICA12 (35;36). Finally, a subset of autoantibodies termed anti-islet cell surface antibodies are currently rarely measured as most recent studies have failed to demonstrate disease specificity (37). Recently described autoantigens include osteopontin (38), importin (39), antibodies reacting with “peri-islet Schwann cells/ GFAP/S100beta (glial fibrillary acidic protein)” surrounding islets (40), densin and filtrin (41) and antibodies reacting CD38 (42).  There are certainly additional specificities awaiting discovery.

Figure 10.1. Development of diabetes is greatly reduced if maternal autoantibodies are not present during pregnancy, presumably due to a lack of transplacental autoantibodies. HEL=Transgenic producing antibodies to Hen Egg Lysozyme; KO=IgM knockout; DBA/2 strain of mouse used as foster mother; SCID=Severe Combined Immunodeficient Mother.

General Assay Methodology
There are multiple assay formats for the detection of disease associated autoantibodies, and similar to many fields of autoimmunity, the first useful assay for islet autoantibodies utilized indirect immunofluorescence with frozen sections of human pancreas as substrate (the ICA assay-see below) (1). The other major format for determination of islet autoantibodies consists of variations on the theme of fluid phase radioassays. In fact the methodology used for the discovery of the radioassay by Berson and Yalow, utilizing insulin antibodies produced by patients treated with insulin, was the basis for the discovery of insulin autoantibodies (2,49). For most autoimmune disorders and for basic immunologic research ELISA assays employing plate bound antigen are the standard. Multiple workshops utilizing sera from patients with type 1 diabetes and multiple mouse strains have demonstrated that standard ELISA formats lack both sensitivity and specificity compared to the fluid phase radioassay formats. For insulin autoantibodies, the ELISA formats were able to detect high capacity antibodies following insulin immunization, but not the insulin autoantibodies of prediabetic individuals (7). This inability of human insulin autoantibodies to bind to plate bound insulin is specific to human autoantibodies as the insulin autoantibodies of NOD mice can readily be detected in a plate bound ELISA format and we have described a highly sensitive and specific assay for such autoantibodies, equivalent to radioassays (50). We believe insulin bound to plastic plates obscures the limited key determinant(s) recognized by human anti-insulin autoantibodies.
In addition when one analyzes thousands of individuals with ELISA formats there is always a subset of sera that react with the wells of the plate itself or unique epitopes created by plate bound proteins. A small percentage of such sera (e.g. 1%) in studies such as DAISY (Diabetes Autoimmunity Study of the Young) where individuals are sampled on multiple occasions over time would result in enough false positives to invalidate efforts to discover environmental, immunogenetic, and diabetes predictive parameters (51,52). It is possible to develop modified ELISA like assays that utilize fluid phase reactivity of antigen and antibody (see discussion of method below), but to date in international workshops only an assay for GAD65 autoantibodies (company RSR developed assay and distributed by Kronus) has attained a level of specificity and sensitivity equivalent to the radioassays, though in the most recent 2009 DASP workshop multiple laboratories using this modified ELISA had excellent sensitivity but only 95% specificity (Mueller oral communication). Increasing the threshold for calling positives for such ELISAs might result in 99% specificity with a presumed small decrease in sensitivity, but for many applications 5% false positives would be problematic. 
A disadvantage of the current fluid phase radioassays is that they utilize low levels of radiation (usually I125, S35-methionine, and 3H-leucine). They are however often performed in a format in 96-well format that is as convenient as the ELISA format (53,54,55). It is likely that many of the clinical assays (but not all, e.g. transglutaminase) that utilize ELISA formats for other autoimmune disorders are utilizing assays with compromised specificity and sensitivity, and the assay format development has just not been optimized and directly compared with modern fluid phase radioassays (e.g. Farr assay versus, protein A based assay, versus ELISA assay for anti-DNA autoantibodies) (56).

Figure 10.2. General outline of fluid phase 96-well plate assays for autoantibodies. In above example I¹²⁵ -insulin utilized, but assay format identical for GAD65, ICA512 assays.

Most investigators assay GAD65 and IA-2 anti-islet autoantibodies in a high throughput 96-well format where labeled antigen is incubated with patient sera, and then both are placed in 96-well filtration plates, where a "bead" (e.g. sepharose) with coupled protein A and or protein G is added, and bound from free radioactivity is separated by filtration washing, and then scintiallation fluid is added directly to the 96-well filtration plates, and counting is performed on multichannel beta counters able to handle the plates (44). Even the assay for insulin autoantibodies utilizing I125 insulin is performed in the same manner, detecting with beta counting emission from I125 (Figure 10.2). A major advance is the simple production of labeled autoantigen by in vitro transcription and translation of cDNAs for given autoantigens to produce the label for the fluid phase radioassay. For example we utilize a combined GAD65 and ICA512bdc radioassay in which GAD65 is labeled with 3H-leucine and ICA512 (IA-2) is labeled with S35-methionine. It has been our experience in setting up multiple such autoantibody assays that approiximately 2/3 of the time an assay using such in vitro produced labeled autoantigen works, and if the assay does not work on the first try, it is unlikely to be modified to work. At present we routinely determine GAD65, ICA512bdc, IA-2ic, IA-2 full length, 21-hydroxylase, and transglutaminase autoantibodies with this methodology (57,58,59,60,61). Such assays are not useful if post-translational modifications, or particular folding of the protein is essential that is not reproduced in the in vitro production of the antigen. With this in vitro translation and transcription methodology minimal protein preparation is needed and following the kit generation of the labeled product we simply perform size separation to produce the labeled antigen.
There are a number of modifications proposed and implemented for the determination of defined islet autoantibodies that can detect autoantibodies with various degrees of sensitivity and specificity relative to the best standard fluid phase radioassays (62,63,64,65). In particular the GAD65 autoantibody described by Smith and coworkers (64) utilizes a novel ELISA format in which a low concentration of the GAD antigen on the plate captures the autoantibody, and then biotinylated GAD in the fluid phase is added and is captured by the second binding site of the autoantibody, and it is the biotinylated GAD65 that is detected to produce the non-isotopic signal. This assay performed well in the previous Immunology of Diabetes/CDC DASP workshop, but a similar format for IA-2 autoantibodies did not have an equivalent sensitivity to the standard fluid phase radioassays (oral discussion at IDS-Cambridge and in current 2009 DASP specificity for multiple ELISA assays was only approximately 95%). There is a need for a "point of service" anti-islet autoantibody assays as at present a major portion of the expense of screening for anti-islet autoantibodies relates to handling of the serum specimen and communication with families. Screening at the point of service (e.g. primary care office), with follow up laboratory confirmatory assays would enhance the ability to test large populations in a Preventive Health mode, that will likely be necessary if successful/safe prevention of type 1A diabetes is achieved.
Islet Cell Autoantibodies (ICAs)
Studies by groups associated with Blizzard, Nerup, and Irvine in the early 1970s demonstrated that Type 1A diabetes was specifically associated with idiopathic Addison's disease (66). This was in contrast to tuberculous Addison's disease. Idiopathic Addison's disease with lymphocytic infiltrates and destruction of the adrenal cortex was considered an organ-specific autoimmune disorder. In addition, insulitis in the pancreatic islets of new-onset diabetic patients who died shortly after diagnosis was reported. These two findings and discovery of human leukocyte (HLA) alleles associated with Type 1 but not Type 2 diabetes led to the hypothesis that in a subset of patients with Type 1 diabetes, the disease resulted from autoimmune beta cell destruction. Patients with Addison's disease, important for recognizing the association of Type 1A diabetes with autoimmunity, provided the sera for the discovery of cytoplasmic islet cell autoantibodies (1). Patients with polyendocrine autoimmunity had high titers of antibodies that reacted with frozen sections of human pancreas. It is believed such patients with Type 1A diabetes have ICAs that frequently persist for decades, in contrast to the majority of patients with Type 1A diabetes where the ICA autoantibodies slowly disappear. The discovery of cytoplasmic islet cell antibodies in this population facilitated their identification in patients with the more common form of Type 1A diabetes. Nevertheless, the utilization of immunohistochemical tests that are dependent upon human pancreas (namely the ICA test) has contributed to debate as to whether anti-islet autoantibodies are predictive of diabetes in a general population and whether such antibodies frequently appear and disappear in relatives of patients with Type 1A diabetes (67,68). We believe these tests can in general be replaced with "biochemical" autoantibody assays even though there almost certainly remain ICA autoantigens that are not identified with current "biochemical" islet autoantibody radioassays (Antigen X, Figure 10.3). This is an area of some debate in that presence of cytoplasmic ICA antibodies increases risk when biochemical autoantibodies are present  (69) though this is often associated with higher titers of the “biochemical” autoantibodies. A recent study, though with a very insensitive assay for insulin autoantibodies illustrates the increased progression to diabetes with ICA in addition to biochemical autoantibodies for first degree relatives of patients with type 1 diabetes (70).
In addition to utility in predicting type 1A diabetes, islet autoantibodies aid in the diagnosis of type 1A diabetes. Approximately 7% of patients with initially non-insulin dependent adult onset diabetes have GAD autoantibodies (71,72,73). These GAD positive individuals almost always become insulin dependent  (71) and they are referred to as having LADA (Latent Autoimmune Diabetes of Youth) diabetes (71,74).

Figure 10.3. GAA (GAD autoantibodies) and IA-2 autoantibodies contribute to ICA reactivity, but insulin autoantibodies do not.

Recent data from the large DPT-1 (Diabetes Prevention Trial) study demonstrated that ICA autoantibodies, as well as GAD autoantibodies and IA-2 autoantibodies, in relatives were relatively stable with 75-85% of all positivity (90 to 95% with high levels) were confirmed during follow up. Approximately 5% of DPT-1 patients without biochemical autoantibodies did progress to diabetes, emphasizing the presence of additional autoantigens.
Since the discovery of cytoplasmic ICA multiple workshops have been held and a Proficiency Testing Service established for laboratories measuring cytoplasmic islet cell autoantibodies (75). The Immunology of Diabetes Society (IDS) workshops rapidly demonstrated that there existed large variation between laboratories in terms of sera called ICA positive. A standard for ICA was established utilizing dilutions of sera and the units assigned were termed JDF units (named for the Juvenile Diabetes Foundation, which sponsored the meetings). With such a standard, differences in assay sensitivity could be rationalized and it became apparent that greater than 20 JDF units of cytoplasmic islet cell autoantibodies were highly predictive of diabetes among first-degree relatives. Laboratories attempting to measure cytoplasmic islet cell autoantibodies and"biochemical" anti-islet autoantibodies are strongly encouraged to take part in IDS workshops (76,77), in the program termed DASP (Diabetes Antibody Standardization Program). As demonstrated with multiple workshops, though the majority of laboratories have excellent assays for GAD65 and IA-2 autoantibodies many laboratories have unacceptable sensitivity and specificity in determining insulin autoantibodies, especially when evaluating in blind the Centers for Disease Control DASP panel of 100 control sera and 50 sera from patients with type 1A diabetes of recent onset. The Centers for Disease Control have evaluated filter paper collection of sera for determination of autoantibodies with the finding that autoantibodies can readily be measured following elution from such blood spots (78).
Limits of the ICA test should be appreciated. Current ICA testing measures greater binding of immunoglobulin to islet versus acinar pancreas. Thus any antigen that is in greater density within islets, or that for a variety of reasons survives the preparation of frozen sections of pancreas better within islets than within acinar pancreas, may be recognized by antibodies and contribute to ICA positivity. Despite this caveat, in appropriate populations high titers of ICA are a very important risk factor for the development of Type 1A diabetes (79). Prognostically significant heterogeneity of ICA has been appreciated and there exist first-degree relatives of patients with Type 1A diabetes or patients with polyendocrine autoimmunity with high titers of ICA with a very low risk of progression to overt diabetes (80,81).
To date, four autoantigens have been defined that contribute to the ICA positivity (Table 10.1, Figure 10.3). The best-characterized subset of ICA autoantibodies is unusual (high titers and unique determinants) anti-GAD autoantibodies (80,81,82). Such antibodies when present as the only autoantibody appear to be present in a small minority of individuals actually developing Type 1A diabetes but are found in a significant proportion of ICA-positive relatives or polyendocrine patients followed for more than 5 years without developing diabetes. Such ICA antibodies have been termed restricted/selective in that they fail to react with mouse islets and give a beta cell-specific pattern of staining. These ICA antibodies can be absorbed with GAD and are similar to the ICA antibodies found in patients with stiff-man syndrome. It is likely that theses GAD-ICA autoantibodies are representative of a more general phenomenon. Presence of autoantibodies reacting with a single islet autoantigen are associated with relatively low diabetes risk (see below) (83, 84, 85).
Other defined antigens accounting for a subset of ICA staining are antibodies reacting with ICA512/IA-2 (or I-A2 β), and ZnT8. Along with GAD65 autoantibodies, these antibodies appear to be the bulk of ICA reactivity in individuals progressing to diabetes. It is also likely that the above autoantigens do not account for all ICA reactivity and in particular human-specific antibodies exist that fail to react with mouse or rat pancreas (Table 10.1) and that are not absorbed by GAD, ICA512. IA-2 and IA-2beta autoantibodies are the most disease specific and associated with the highest risk of progressing to diabetes (86). There are two major splice variants of IA-2 that are commonly utilized in fluid phase autoantibody radioassays (IA-2ic [IA-2 intra-cytoplasmic fragment] and ICA512bdc [IA-2 molecule lacking the transmembrane domain, termed ICA512 Barbara Davis Center]) (87,88). In a minority of new onset patients or prediabetics only one or the other autoantibody is present, while most patients produce anti-IA-2 autoantibodies reacting with both constructs.  DASP workshops indicate that of the two variants assays IA-2ic radioassays have a higher sensitivity.
The standard "biochemical" autoantibody assays measure antibodies reacting with insulin, GAD65, and IA-2 (also termed ICA512). It has been much easier to standardize assays for GAD65 and IA-2 as compared to insulin autoantibodies (89) which probably relates to the much stronger signal/noise for GAD65 and IA-2 autoantibodies (90). Receiver Operator Curves demonstrate a very small difference in signal between prediabetic/diabetics and the bulk of normal control sera for insulin autoantibodies. Thus this assay requires meticulous attention to detail, in contrast to GAD65 and IA-2 assays that are more "forgiving". With the Immunology of Diabetes Society (IDS) and Centers for Disease Control DASP workshop program standard sera are now available and help with assay implementation (see http://www.immunologyofdiabetessociety.com/).
Sequenced Autoantigens
Glutamic Acid Decarboxylase (GAD)
Glutamic Acid Decarboxylase (GAD)
The discovery that anti-GAD autoantibodies accounted for an important component of anti-64 kd autoantibodies and a subset of ICA began with the discovery that patients with stiff-man syndrome expressed autoantibodies that gave rise to ICA staining and had high titers of anti-GAD autoantibodies (82). Glutamic acid decarboxylase (GAD) is the enzyme responsible for GABA synthesis within the nervous system and islet cells (91). GAD is expressed within all human islet cells (e.g. alpha,beta,delta). The protein is post-translationally modified with palmitoylation and is targeted to synaptic vesicle membranes of neurendocrine cells (92). For rat islets GAD is beta cell specific and levels of GAD are extremely low in mouse islets (93,94). In that only beta cells are destroyed in man, and insulitis of man and animal models disappears from islets once all the beta cells are destroyed, it is possible that anti-GAD autoantibodies and the autoimmune response to GAD is of secondary rather than of primary pathogenic importance (93,95). In addition to islets, the two forms of GAD, termed GAD65 and GAD67, are synthesized in testes, ovary, and neurons. The anti-GAD autoantibodies of patients with stiff-man syndrome (SMS) (a rare neurological disorder characterized by spasms of muscular rigidity) are unusual in that they react with fixed sections of the brain and react on Western blots with purified GAD (82). The levels of anti-GAD autoantibodies in SMS do not correlate with disease and there is no clear evidence that the autoantibodies are causative rather than a marker of the disease process. Prior to studies of SMS, Baekkeskov and coworkers (96,97) had described anti-64 kd autoantibodies utilizing an immunoprecipitation assay of metabolically labeled islet cells. Although a high percentage of patients with new-onset diabetes had anti-GAD autoantibodies with this assay, these antibodies do not Western blot GAD or show immunohistochemical reactivity with neurons. Furthermore, anti-GAD stiff-man antibodies recognized different epitopes in contrast to the anti-GAD antibodies usually found in new-onset diabetic patients (98). In 1990, the groups of DeCamilli and Baekkeskov (8) jointly discovered that the bulk of anti-64 kd autoantibodies reacted at low titer (relative to stiff-man syndrome antibodies) with GAD and that a small subset of the anti-GAD autoantibodies of new-onset diabetics were similar to anti-GAD stiff-man syndrome antibodies. In 1991, Gianani and coworkers (80) (for relatives of patients with Type 1A diabetes) and Genovese and coworkers (81) (for polyendocrine patients) discovered that individuals with high titers of anti-GAD antibodies reacted on sections and Western blot similar to SMS patients anti-GAD antibodies. These antibodies were termed "restricted" ICA or "selective" ICA in that the antibodies failed to react with islets of mouse pancreas and on rat islets stained only beta cells. This restricted reactivity is apparently due to the almost eight-fold less expression of GAD by mouse islets (93). When ICA reactivity is due solely to autoantibodies reacting with GAD, absorption of sera with affinity purified GAD, recombinantly produced GAD, or absorption with a brain extract can be used to identify such antibodies (Table 10.1) (99). In studies utilizing brain extract, the specificity for GAD absorption is usually evaluated by preclearing the extract with a monoclonal anti-GAD antibody. Utilization of beta cell-specific staining to aid identification of restricted ICA is a problem with human pancreas in that GAD is expressed by a subset of non-beta cells of human pancreas. In contrast, within rat islets GAD expression is beta cell specific (94).
Several different radioassays for anti-GAD autoantibodies have been developed and multiple Immunology of Diabetes Workshops (IDW, now IDS (Immunology of Diabetes Society)) for measuring such antibodies was held (89). Four formats for determining anti-GAD autoantibodies include: (1) determination of antibody precipitation of GAD enzymatic activity, (2) radioassays utilizing affinity purified porcine brain GAD that has been labeled with I125 (3), radioassays utilizing endogenously labeled GAD produced by in vitro transcription and translation of GAD cDNA (100), and (4) ELISA assay formats. These assay formats are likely to give very different results depending upon the population studied. In particular, the immunoenzymatic assay appears to be less sensitive than either radioassay. Radioassays utilizing in vitro transcribed and translated GAD65 or I125-labelled recombinant human GAD65 can give equivalent results with high sensitivity and specificity. The Combinatorial Autoantibody International workshop of the Immunology of Diabetes Society indicated that most laboratories (but not all) concordantly measured anti-GAD65 autoantibodies (89). All attempts at standard ELISA assays have failed to achieve the sensitivity and specificity exhibited by most of the fluid phase radioassays. Recently a modified ELISA assay where fluid phase reaction of antibody and biontinylated GAD65 with capture of the complex on a plate gave results equivalent to the radioassays (oral report Cambridge IDS workshop) (64).
The Immunology of Diabetes Society (IDS) and Centeres for Disease Control (CDC) have organized multiple DASP workshops. The majority of participating laboratories utilize fluid phase radioassays. A standard calibration sample (World Health Organization) was distributed and used in these workshops and results for GAD65 and IA-2 autoantibodies can be reported in WHO units, or as an index based upon a standards reactivity.  Despite the reporting of GAD autoantibodies in WHO units, the utilization by each laboratory of their own standard samples, quantitative concordance between labortories can be limited, and it is likely to obtain better correlation the same standard sera might need to be utilized in each laboratory.  This is not possible except for limited consortia, and thus though reported levels of autoantibodies correlate, exact levels reported by different laboratories differ.
The quantitation of anti-GAD autoantibodies can aid in the detection of extremely high titers of such antibodies associated with restricted ICA. The highest levels of anti-GAD autoantibodies that we have observed among ICA-positive relatives occur in relatives with restricted ICA independent of the GAD autoantibody assay utilized.
Among ICA-positive first-degree relatives, the levels of anti-GAD autoantibodies are remarkably constant with up to a decade of prospective evaluation. Harrison and coworkers (101) have reported that there is an inverse correlation between levels of GAD autoantibodies and a T cell proliferative response to GAD. Thus it is hypothesized that high levels of GAD autoantibodies may be associated with a diminished anti-islet T cell response (Th1 rather than Th2) and therefore less or no beta cell destruction. Given perhaps the difficulties of T cell assays this observation has not been confirmed and in general dividing islet autoantibodies into quartiles, the higher the level of the islet autoantibody the greater the risk of progression to diabetes for at risk relatives (87,102).
Two isomers of GAD, GAD65 and GAD67, share an identical exon-intron structure and are 76% identical in amino-acid sequences with only 22% identity in N-terminal region and over 90% identity in C-terminal regions (103). Antibodies to GAD65 were detected in over 80% of the patients with type 1A diabetes (83) versus only 11-18% having antibodies to GAD67 (104,105). The autoantibodies binding to GAD67 in IDDM seem to represent antibodies to shared epitopes with GAD65 (104) although there was a report that the patients with Grave's disease have been found carrying specific GAD67 antibodies in the absence of GAD65 antibodies (106).
GAD67 and GAD65 have similar protein conformations that are important for antibody binding. GAD65/GAD67 chimeric proteins were widely utilized in most of laboratories to define the epitope of GAD65 while maintaining the overall GAD protein conformation. The precise mapping of GAD epitopes targeted by patients with type 1A diabetes is difficult because serum contains polyclonal antibodies with multiple specificities.  Monoclonal antibodies derived from individuals with new-onset diabetes have been useful in mapping GAD epitopes in type 1A diabetes (107,108,109). Two major epitope regions targeted by human diabetic sera have been identified (110) and confirmed by other laboratories.  One epitope is in the middle part of GAD65 between amino acids 240 - 435 (termed IDDM-E1) and one is at the C-terminal region of GAD65 between amino acids 451 - 570 (termed IDDM-E2). Recently, Baekkeskov, et. al. (111) used homolog-scanning mutagenesis to identify GAD65-specific amino acid residues which form autoreactive antibody epitopes in this molecule. Detailed mapping of 13 conformational epitopes, recognized by human monoclonal antibodies derived from patients, together with two and three dimensional structure predictions led to a model of the GAD65 dimer. The results revealed a remarkable spectrum of human autoreactivity to GAD65, targeting almost the entire surface of the molecule and reactivity to the multiple epitopes are present both before and after onset of type 1A diabetes (112,113). This suggests that native folded GAD65 is the immunogen for autoreactive B-cells in IDDM.
Comparisons of the autoimmune repertoires to GAD65 in type 1 diabetes, Stiff Man Syndrome (SMS), and Autoimmune Polyendocrine Syndrome (APS) might provide insight into why some individuals develop one disease and other individuals develop the other disease. Most type 1 sera only target GAD65 protein epitopes dependent on protein conformation and do not bind GAD65 protein fragments or react with denatured GAD65 protein with immunoblotting (8,110). In contrast, SMS sera have antibodies that bind GAD65 protein fragments consisting of N-terminal, middle, and C-terminal regions  (110;114,115,116). Characteristically, SMS sera target an epitope contained in the N-terminal first eight amino acids and neither IDDM and APS sera have such antibodies. Despite clear differences in the GAD65 antibody profile in SMS and IDDM, an important similarity exists. Both SMS and IDDM sera contain antibodies that bind a conformation-dependent region in the middle part of the protein and a second set of antibodies that binds a conformation-dependent epitope in the C-terminal regions of the protein (110). Antibodies specific for these two regions of GAD65 are present in approximately equivalent titers in most sera although a single specificity may predominant. Like SMS, APS1 sera have antibodies that inhibit the enzymatic activity of GAD65, while this is not a property of type 1 diabetes autoantibodies (116). The similarity and difference in the GAD65 antibody repertoire in the Autoimmune Polyendocrine Type 2 Syndrome (APS2) and diabetes has been studied (117). Neither diabetes nor APS2 sera bind the N-terminal third of GAD65, but instead target the C-terminal two-thirds of GAD65, IDDM-E1 and IDDM-E2. More detailed mapping has found differences for APS2 sera reactivity in the IDDM-E2 region (117). There are multiple reports of GAD65 autoantibodies associated with interferon alpha therapy, and in particular with therapy for viral hepatitis. Schories and coworkers described de novo development of GAD65 autoantibodies in 2/74 patients treated with interferon-alpha for hepatits C infection and recommended testing for such autoantibodies prior to a second course of therapy, given the progression to overt diabetes of one of their GAD65 positive patients (118,119,120).
Hampe and coworkers have developed a novel assay for GAD epitopes utilizing competition with Fab fragments of monoclonal anti-GAD antibodies (121). Their group recently reported the detection of anti-idiotypic antibodies in the sera of normal individuals which they believe mask the presence of GAD autoantibodies in all normal individuals (122). They contend that patients expressing GAD autoantibodies lack these anti-idiotypic antibodies and thus are positive with current GAD assays.  This is a very novel suggestion and will require more detailed study.  In particular the assay used to analyze anti-idiotypic antibodies involves incubating serum with a human anti-GAD monoclonal covalently bound to beads and then centrifuging the mixture to remove “anti-idiotypic” antibodies. There is the possibility for the monoclonal anti-GAD antibody to be released from the beads and thus give the appearance of anti-idiotypic antibodies. The claim that positive diabetic sera does not have anti-idiotypic antibodies is primarily that such sera gives signals with or without absorption with the beads (122).  
ICA512 (IA-2) and IA-2 β (phogrin)
Following the identification that a major component of the 64 kd antigen is GAD, many groups reported an association between Type 1A diabetes and anti-GAD antibodies. Evidence for heterogeneity of anti-64 kd autoantibodies came from the work of Christie  (123), who found that antigenic tryptic fragments were generated after mild trypsinization of 64 kd molecule(s). In particular, precipitation of human islets with sera from nondiabetic polyglandular failure individuals, followed by trypsinization, yielded a 50 kd fragment whereas a 38-40 kd fragment was precipitated by new-onset and prediabetic sera. Of note, mild trypsinization of the GAD molecule yields a 50 kd fragment. Reports (123,124) indicated that autoantibodies to the 37-38 kd fragment had a higher positive predictive value for diabetes than anti-GAD antibodies. It appeared that there existed two different 64 kd autoantigens in human islets, one being GAD and another two molecules that after trypsinization yield a fragment of 37-40 kd. Christie reported that the 40 kd fragment is related to the protein tyrosine phosphatase IA-2 (ICA512) (125) and the 37Kd fragment to the molecule IA-2 β (13).
The autoantigens ICA512 (IA-2) and subsequently IA-2 β (phogrin) were isolated independently by several investigators probably reflecting the importance of these related molecules. Rabin and coworkers (9) screened an islet expression library with sera from patients with type 1A diabetes. The molecule he isolated was termed ICA512. The same molecule has been termed IA-2. The original sequence in GenBank from Rabin and coworkers lacked several nucleotides out of approximately 2,000, but with a frame shift predicted a shortened protein, which was originally utilized in autoantibody assays. A related protein with homology in a putative tyrosine phosphatase region (though to date no enzymatic activity has been demonstrated for either molecule) was termed IA-2 β by Notkins and colleagues (12), and phogrin (phosphatase of granules of rat insulinoma) by Hutton and coworkers (13). Both molecules are most homologous in their C-terminal intracytoplasmic domains to which essentially all of the autoantibodies are directed. Almost all autoantibodies which react with IA-2 β also react with IA-2, while approximately 10% of patients developing type 1A diabetes have autoantibodies reacting with IA-2 but not with IA-2 β. Thus we do not routinely assay for IA-2 β autoantibodies.
There is additional heterogeneity of the autoantibodies which react with IA-2 (ICA512). A large number of epitopes (>10) within the intracytoplasmic C-terminus of the molecule are targets of autoantibodies (126). In addition islet cells express differentially spliced messenger RNA for ICA512, with one form lacking exon 13 which includes the transmembrane region of the molecule. In our laboratory by screening an islet expression library we obtained an ICA512 clone (ICA512bdc) which was used to develop an autoantibody radioassay (127,128). This clone lacks exon 13 [construct termed ICA512bdc (bdc for Barbara Davis Center)]. Approximately 10% of patients developing type 1 diabetes have autoantibodies reacting with only one or the other of full-length ICA512 or ICA512bdc. Thus to fully define all individuals expressing these autoantibodies one needs to apply both assays. It appears that with exon 13 missing, unique epitopes recognized by autoantibodies are created with the joining of exon 12 to exon 14. Other patients have autoantibodies to the transmembrane/juxtamembrane region of the molecule which is lacking in ICA512bdc. With multiple autoantigens tested (e.g. insulin, GAD65 and ICA512) very few patients developing type 1A diabetes have autoantibodies to only a single construct of ICA512 and lack all other autoantibodies. Thus, testing for only ICA512bdc or ICA512 autoantibodies is usually sufficient. Full length IA-2 (ICA512) has N-terminal determinants that increase somewhat false positive (normal controls) reactivity and thus to measure both forms of IA-2/ICA512 investigators usually measure ICA512bdc (amino acids 256-556, 630-979) or IA-2ic (intracytoplasmic = amino acids 605-979). If one has to choose a single IA-2 form, we would recommend IA-2ic, given higher sensitivity and preserved specificity in DASP workshops.
Two epitopes of particular interest within the ICA512 molecule are a mini-epitope described by Dotta and coworkers at the C-terminus of the molecule (129) and a juxtamembrane epitope described by Bonifacio and coworkers (45) (Figure 10.4). The mini-epitope is of interest in that despite in vitro transcribing and translating only 51 amino acids of the ICA512 molecule, approximately 56% of patients who have ICA512 autoantibodies react with this epitope (130). Bonifacio and coworkers have evidence that autoantibodies to the juxtamembrane region (amino acids) are amongst the first to appear in individuals developing type 1 diabetes. In our studies of young infants we have not been able to confirm this finding but the number of such children we have studied developing antibodies de novo is small.

Figure 10.4. In Vitro transcription and translation of fragments of alternative splice variants of ICA512 and smaller fragments recognized by ICA512 autoantibody positive sera.

Glycolipid Autoantigens
The initial studies suggesting that antibodies directed against islet glycolipids may be a component of anti-islet autoantibodies came from evaluation of monoclonal antibodies that reacted with neuronal gangliosides and that also react with islets. Monoclonal antibodies A2B5 (131), 3G5 (132), R2D6 (133), and tetanus toxin all react with complex gangliosides and all give an ICA pattern of staining that involves all cells within islets (108,110). In contrast to anti-islet monoclonal antibodies that react with proteins (e.g., HISL19 (134,135)), the above anti-ganglioside antibodies were not species specific (136). In that gangliosides are composed of a ceramide group connected to a complex polysaccharide with at least one sialic acid group, gangliosides are amphiphilic. Thus, organic solvents such as methanol remove gangliosides from tissue sections, while gangliosides are stable to nonpolar solvents such as acetone. Stability to acetone but removal by methanol was a quality found for most ICA. Gangliosides can readily be extracted from tissues with organic solvents and then separated from nonpolar lipids by partitioning with chloroform. Extraction of human pancreas followed by the above Folch partition, followed by chromatography that separates gangliosides based on their hydrophobicity (thin-layer chromatography, TLC) or number of sialic acid residues (DEAE chromatography) allowed the isolation of ganglioside preparations that, when incubated with ICA-positive sera, removed all reactivity (32). It is noteworthy that the absorption assay utilized mouse pancreas (32). It is likely (in retrospect) that the choice of mouse pancreas allowed absorption to be demonstrated even if sera contained high-titer anti-GAD autoantibodies that are not absorbed by glycolipid preparations (and that do not give ICA-positive staining with mouse islets; e.g., restricted ICA).
Which glycolipids are autoantigens and which are the target autoantigens for ICA? Three candidate molecules are GT3 (136), sulphatides (137), and what has been termed a GM2-1 (138,139,140) ganglioside (on thin-layer chromatography it has a mobility between GM2 and GM1 standards).
Insulin
Insulin is a 51-amino acid disulphide-linked heterodimer that is specifically produced by beta cells of islets (Figure 10.5). Proinsulin which is processed within the secretory granule is the precursor to insulin. To date insulin and proinsulin are the only known beta cell-specific autoantigens within human islets (all other putative autoantigens are produced by human non-beta islet cells). Of note, both in mouse and man, proinsulin messenger RNA and proinsulin are present in the thymus (141,142,143,144,145,146). Transgenic mice with transgenes with the rat insulin promoter coupled to several proteins indicates that thymic transcription of the insulin gene is possible (147). Insulin in islet beta cells is produced as a preprohormone that is processed to pro-insulin and then, with removal of connecting peptide (C peptide), to mature insulin. Processing occurs within beta cell secretory granules, where insulin is packaged in a crystal form (148). Equimolar concentrations of insulin and C peptide are secreted by beta cells.  In the thymus processing is likely to be very different and thus the epitopes presented. 

Figure 10.5. Sequence in pre-proinsulin with the insulin B chain (B:9-23) dominant T cell epitope highlighted.

In the 1950s, Berson and Yallow developed the first radioimmunoassay utilizing sera with insulin antibodies from patients treated with bovine insulin (149). Bovine insulin differs from human insulin at 3 of 51 amino acid and porcine insulin differs from human insulin by one amino acid (at the terminus of the B chain). For most patients treated with subcutaneous insulin, the presence of anti-insulin antibodies does not interfere with insulin therapy. A subset of insulin-treated patients with extremely high levels of insulin antibodies are insulin resistant with mean insulin binding capacities greater than 216 nM (30,000 microunits of insulin/ml serum) (150). For such patients, the species of insulin used for therapy is usually changed (e.g., from bovine to human insulin) or to humalogue insulin and occasionally to sulfated insulin. Anti-insulin antibodies during pregnancy may facilitate transplacental passage of animal insulins.
A very rare, but interesting syndrome, is the Insulin Autoimmune Syndrome, also termed Hirata syndrome (151;152).  Patients with this syndrome (almost all with DRB1*0406) after exposure to sulfhydryl containing medications (e.g. methimizole, penicillamine) develop extremely high levels of insulin autoantibodies (153). These patients usually present with hypoglycemia which resolves with discontinuation of the medication. In addition to this MHC restricted syndrome, some patients have monoclonal insulin autoantibodies produced by B lymphocyte tumors (152).
The production of antibodies to animal insulin was not unexpected. It was, however, discovered that individuals treated with human insulin also produce anti-human insulin antibodies (though at a lower concentration) (150). The development of anti-insulin antibodies following subcutaneous therapy with human recombinant insulin has been ascribed to the possible presence of "denatured" insulin molecules. It is, however, very likely that self proteins when administered subcutaneously, especially in a depot form (e.g., complexed with the positively charged protein NPH (neutral protein Hagedorn), or zinc, as in the two most common pharmacologic long-acting insulin preparations), will induce antibodies. Antibodies also develop following therapy within a number of other recombinant proteins including human growth hormone and factor VIII. To test the generality of production of insulin antibodies following subcutaneous insulin administration, we transplanted a pituitary cell line producing rat insulin (which has the same sequence as mouse insulin) into histoincompatible mice. With rejection of the tissue, anti-insulin "autoantibodies" were induced (154).
In 1983 Palmer and coworkers (2) discovered the presence of anti-insulin antibodies in patients with new-onset Type 1A diabetes prior to the administration of exogenous insulin. Subsequently a large number of studies have demonstrated that anti-insulin antibodies are present for years before the development of Type 1A diabetes (155). Initially, two different assay formats were utilized to detect such antibodies (an ELISA format with insulin immobilized on plates and fluid phase radioassays). The two different formats gave very different results relative to the positive predictive value for development of Type 1A diabetes. A series of international workshops and sera exchanges led to the observation that the two formats measure different antibodies, and that only the antibodies detected with the radioassay formats were associated with risk for Type 1A diabetes (7). Both assay formats detected antibodies following insulin therapy and rarely were positive in normal controls.  We have evidence that the inability of standard LISA assays with insulin bound to plates is specific to human insulin autoantibodies as we have developed ELISA format assays that readily detect the insulin autonatibodies of the NOD mouse (50).
As more has been learned concerning insulin autoantibodies, the inability of standard ELISA assays to detect disease-relevant antibodies is understandable and probably relates to obscuring a key epitope of insulin when bound to a solid phase. In particular, the anti-insulin autoantibodies associated with diabetes risk are all of extremely high affinity and, most important, of very low capacity (10 -12 M), making their detection with plate-binding assays problematic (7). The epitope of the insulin molecule recognized with fluid phase assays appears to be homogeneous (156). In particular, the antibodies from all antibody-positive relatives we have studied react with a conformational epitope (do not react with either the A or B chain) and react equally well with insulin and proinsulin (52). To date, all insulin autoantibodies of at-risk relatives react with des B23 to B30 insulin lacking the terminal eight amino acids of the B chain. Des B23 to B30 insulin fails to bind to the insulin receptor. The A13 leucine is in the base of a pocket of the insulin molecule, and if this leucine is replaced by a tryptophan that extends out of the pocket, insulin autoantibody reactivity is abrogated. The reason for this marked specificity of insulin autoantibodies reacting with the face opposite the insulin-binding domain is unknown. It is, however, very likely that the antibody response to insulin is "antigen-driven", giving rise to high-affinity autoantibodies. Insulin may react with B cells while bound to insulin receptors, either of neighboring immunocytes or perhaps following its transport on insulin receptors through the endothelium. The ability to produce insulin analogues that are recognized by insulin autoantibodies but not by the insulin receptor suggests that one can selectively target B cells producing such antibodies. Whether elimination of such B cells which are probably extremely efficient at presenting insulin to T cells, would influence beta cell destruction is unknown. Autoantibodies of prediabetics which react with proinsulin are all absorbed with insulin, and assays for insulin autoantibodies are reported to have higher disease specificity compared to proinsulin autoantibodies.  We have recently developed a modified insulin autoantibody ELISA assay for the autoantibodies of the NOD mouse that is as sensitive as our standard radioassay while retaining specificity, but to date the same assay format is not sufficient for human anti-insulin autoantibodies (50).  A portion of the difficulty we believe relates the marked epitope specificity of human autoantibodies with plate binding interfering with this reaction as well as additional difficulties created by variable background binding of different human sera to plates +/- antigen.
Levels of insulin autoantibodies, similar to GAD autoantibodies, appear to be regulated at different levels over long periods of time in prediabetic first-degree relatives (Figure 10.6). Of interest is a report indicating that in a subset of children who are negative for insulin autoantibodies at onset, insulin autoantibodies are detectable in their IgG serum fraction, suggesting blocking of reactivity by immune complexes (157).

Figure 10.6. Development of GAD and insulin autoantibodies in child from the general population followed to the development of diabetes. This child did not express ICA512 autoantibodies.

The levels of insulin autoantibodies correlate inversely with the age at which type 1 diabetes develops. Thus levels greater than 2000 nU/ml are almost exclusively found in patients who progress to Type 1A diabetes prior to age 5, and less than half of individuals developing Type 1A diabetes after age 15 have levels of anti-insulin autoantibodies distinguished from controls. The levels of such antibodies are to some extent genetically influenced, are associated with DR4 (158) and DQ8 (52), and are also associated with other haplotypes with DQA1 alleles of lineage 2, DQA1*0102, 0201,0301, 0401 (158). Presence of anti-insulin autoantibodies and their levels appear to be associated with such lineage 2 alleles in a dominant manner. In particular, individuals developing type 1A diabetes who are homozygous for DQA1*0501/DQB1*0201 (DR3 homozygotes) either are negative for anti-insulin autoantibodies or have very low levels. There are suggestive data that the levels of anti-insulin autoantibodies among ICA-positive first-degree relative correlate with the rate at which individuals progress to overt diabetes (159). Relatives, however, who only express anti-insulin autoantibodies infrequently progress to overt diabetes (52). A high proportion of such anti-insulin autoantibody-positive, ICA-negative relatives under the age of 10 convert to ICA positivity.
Williams and coworkers modified the insulin autoantibody assay by utilizing protein A to precipitate autoantibodies (160) rather than polyethylene glycol used in standard insulin autoantibody assays (2). The assay (termed micro-insulin autoantibody assay) was modified such that rather than 600ul of sera/sample, the assay utilized approximately 25ul of sera per sample for duplicate determination with and without competition with non-labeled insulin. Not only did the assay utilize less sera, it also eliminated two artifacts apparently related to the use of polyethylene glycol (polyethylene glycol precipitates many proteins based on a protein's solubility). Both cord blood and hemolyzed sera have factors, which bind labeled insulin and are precipitated by polyethylene glycol while both types of sera are negative with assays utilizing protein A precipitation (161,162). We have further modified the Williams assay by utilizing a 96-well plate format with membrane filtration to separate bound from free autoantibodies, and direct counting on a beta counter of the precipitated I125-insulin. The counting in a 96-well plate, and handling in 96-well plates allows semi-automated determination of insulin autoantibodies in a format similar to that for the GAD65 and ICA512 autoantibody assays (Figure 10.2).

Figure 10.7. ROC curves of autoantibodies, Denver Laborabory DASP 2002.

Unlike GAD65 and IA-2 autoantibody assays, mIAA assays have a poor correlations between laboratories of the DASP Immunology of Diabetes/CDC workshops including the most recent 2009 DASP workshop (Mueller oral presentation). In the DASP workshops to date, assay sensitivities have been relatively poor for the majority of laboratories (<30%), with only a handful of laboratories having sensitivies (with preserved specificity) exceeding 50%. The sera samples of the patients with diabetes in the DASP workshops are predominantly obtained from older individuals with type 1A diabetes given the volumes needed for the workshops. Since adolescents and adults have markedly lower levels of insulin autoantibodies compared to younger children, difficulty with measuring these low-titer anti-insulin autoantibodies is not surprising. As illustrated (Figure 10.7) by the ROC curves above, the signal to noise ratio for insulin autoantibodies (mIAA: micro-insulin autoantibodies) of DASP workshop samples is very small (lower right panel) compared to the curves for GAD65, full length IA-2 and ICA512bdc (IA-2 variant lacking exon (transmembrane) 13).
Insulin autoantibodies are usually the first autoantibody to appear in young children developing type 1 diabetes (52,163,164). This is particularly true for infants less than 1 year of age who begin to express autoantibodies.  Achenbach and coworkers have analyzed the affinity of anti-insulin autoantibodies for children followed prospectively in the BabyDiab study (Figure 10.8). As illustrated in the figure below a high percentage of the children who went on to develop multiple anti-islet autoantibodies or to progress to diabetes express high affinity autoantibodies (>10⁹ l/mol). In addition the high risk, high affinity autoantibodies differed from the autoantibodies of children who failed to develop additional autoantibodies (remained IAA positive only) or had transient insulin autoantibodies in that the majority reacted well with proinsulin (52). It is likely that most of the lower affinity proinsulin non-reactive insulin autoantibodies are "false positive" autoantibodies relative to diabetes risk.

Figure 10.8. Insulin autoantibodies of BabyDiab children.

Isotypes of insulin autoantibodies have been evaluated in the BabyDiab study and in studies from Finland (87,165) with the observation that a broader response to insulin (including IgG3 autoantibodies) and strong IgG1 responses is associated with a somewhat greater risk of progression to diabetes.
In addition the micro-IAA assay readily detects insulin autoantibodies in NOD mice and the early expression of insulin autoantibodies of NOD mice correlates with the early development of type 1 diabetes. Approximately 90% of NOD mice expressing insulin autoantibodies at 8 weeks of age develop diabetes by 16 weeks of age. Figure 10.9. Insulin autoantibodies can be rapidly induced in normal Balb/c mice with the administration of the autoantigenic peptide, residues B:9-23 of insulin. These antibodies do not react with the immunizing peptide but with intact insulin (166). This surprising finding clearly indicates that normal mice have autoreactive T and B cells able to respond to the peptide (T cell) and to intact insulin (B cell). Further studies using the B:9-23 peptide and poly-IC (viral mimic and activator of the innate immune system) in Balb/c mice have generated insulitis and in Balb/c mice with transgene induced B7.1 islet expression, diabetes is induced. Thus immunization with this single peptide can lead to the destruction of islet beta cells (167,168,169,170,171).

Figure 10.9. Development of insulin autoantibodies in individual NOD mice. Mice which expressed insulin autoantibodies at 8 weeks of age developed diabetes prior to 16 weeks, while mice expressing insulin autoantibodies after 20 weeks developed diabetes later. Yu et al. PNAS 97:1701, 2000.

Two workshops have evaluated insulin, GAD65, and ICA512 (IA-2) autoantibodies in NOD mice. The workshop reports conclude that insulin autoantibodies measured by a sensitive radioassay are strongly associated with autoimmunity while neither GAD65 nor ICA512 (IA-2) specific autoantibodies were demonstrated (172).
Thomas and coworkers have produced a series of monoclonal insulin autoantibodies and recently transgenic mice producing an insulin autoantibody (173). These mice are tolerant to insulin, suggesting that even low levels of insulin can induce B cell tolerance. Proinsulin is present in thymus and lymph node. Knocking out of the insulin 2 gene (the insulin of mice expressed within the thymus) greatly accelerates the development of diabetes and increases levels of insulin autoantibodies (167,174,175,176). In contrast an insulin 1 gene knockout on the NOD background prevents the development of diabetes in the majority of mice but has relatively little effect upon insulin autoantibodies (167).   Knocking out both the insulin 1 and the insulin 2 gene, replacing insulin with a mutated preproinsulin gene (B16:A reather than B16:Y) completely prevents the development of NOD diabetes and greatly diminishes insulin autoantibodies (176,177).  We believe the insulin peptide B:9-23 is a primary autoantigenic determinant of the NOD mouse.
ZnT8
The fourth major confirmed islet autoantigen recognized by human autoantibodies in the islet specific zinc transporter ZnT8. It was discovered to be an autoantigen by Hutton and coworkers based on algorithms searching for beta cell specific highly expressed molecules (14). The ZnT8 transporter is one of a large family of zinc transporters, but this one is associated with the membrane of secretory granules of islet beta cells.  Zinc within secretory granules of beta cells is complexed with insulin, forming a storage crystal composed of zinc and insulin. 
An initial fluid phase radioassay for anti-ZnT8 autoantibodies utilizing the full length molecule had a relatively low sensitivity.  Most of the full length molecule is hydrophobic as it is within the membrane of secretory granules. When Hutton and coworkers developed assays utilizing the C-terminus of the molecule highly specific and sensitive assays were achieved.  Autoantibodies reacting with ZnT8 are present in the majority of patients with Type 1 diabetes and assay specificity and sensitivity is similar to those for GAD65 autoantibodies in DASP workshops (2009 IDS: Mueller oral presentation). Of note there are two major polymorphic variants of ZnT8, one with a tryptophan and the other an arginine at position 325 of the molecule (178).  The majority of patients have autoantibodies recognizing both variants but a subset have autoantibodies recognizing only one variant. Those recognizing the specific arginine variant are patients homozygous for the arginine variant, and vice versa for the tryptophane variant. This is a unique demonstration that islet autoimmunity in terms of epitopes recognized is truly autoimmune, with patients recognizing their own sequence. If ZnT8 were discovered prior to GAD65 or IA-2 autoantibodies it would be part of the primary panel of autoantibodies. At present we measure ZnT8 autoantibodies in diabetic patients negative for the other biochemical autoantibodies in whom we attempting to diagnose Type 1A diabetes.  For at risk populations such as relatives of patients with Type 1 diabetes and individuals with high risk HLA alleles from the general population we measure ZnT8 autoantibodies in those expressing a single anti-islet autoantibody. If ZnT8 autoantibodies are present these individuals would have >=2 anti-islet autoantibodies and have much higher risk of progressing to diabetes (14). 
ICA69
Pietropaolo and coworkers identified a novel islet protein termed ICA69 (24) through the screening of a lambda gt11 human islet expression library with ICA-positive sera (179). This protein, which on SDS gel migrates at 69 kd, has been sequenced and found to have in molecular weight of 54,600 (the aberrant gel migration is probably explained by the presence of several highly charged region in the molecule, as ICA 69 is not glycosylated). ICA69 is present, at least at the levels of messenger RNA, in brain, lung, kidney, and heart, with high levels in islet and other neuroendocrine tissues. Pietropaolo and coworkers have reported that ICA69 expression in the thymus of NOD mice is reduced (180). In the rat, ICA69 is beta cell specific (181). ICA69 by confocal microscopy is associated with the Golgi-complex and immature (lesser extent) insulin secretory granules with essentially no ICA69 evident on secrtory granules near the plasma membrane (26). A knockout of ICA69 in C. elegans compromised neurotransmission (182). A knockout of the ICA69 gene (ica-1) was bred onto the NOD mouse and insulitis and diabetes developed normally in these knockout mice (183).  
ICA69 is identical in sequence to the cow's milk-related protein p69 described by Dosch and coworkers (184). It has been reported that diabetic individuals have antibodies to bovine serum albumin and there are a number of epidemiological studies indicating that the neonatal ingestion of cow's milk increases the development of Type 1A diabetes. ICA69 and bovine serum albumin have two short regions (five amino acids) of identity and in the region of the ABBOS albumin peptide (potential T cell epitope), four of nine amino acids are identical. Such identity may be enough to stimulate cross-reactive T cell clones.
In the initial assay format, antibodies to ICA69 were measured by Western blot utilizing recombinant ICA69. The Western blot assay format relative to radioassays for antibodies reacting with insulin, GAD65, and ICA512 is inadequate, with more than 5% of normal controls reacting on Western blots. Anti-ICA69 autoantibodies are diabetes-related autoantibodies, but are present also in rheumatoid arthritis patients (185). Further studies are needed to elucidate the role of this molecule in the pathogenesis of Type 1A diabetes. A report by Atkinson and coworkers questions the association with Type 1A diabetes of anti-albumin antibodies and of T cell responses to albumin (186). Dosch and coworkers however in studies of mice can accelerate or inhibit development of diabetes with a peptide of the mouse molecule (187).
Carboxypeptidase H
Carboxypeptidase H is another autoantigen discovered by the screening of islet expression libraries with prediabetic sera. Within islets, carboxypeptidase H cleaves carboxyterminal amino acids during the processing of proinsulin to insulin. This molecule is not islet specific, being present also in bovine adrenal, pituitary, and kidney (148). In beta cells, carboxypeptidase H is localized in the insulin secretory granule, where it exists both in a membrane bound (52 kd) and soluble form (50 kd). Carboxypeptidase H is probably the most abundant islet protein after proinsulin-insulin. A radioassay or an ELISA for anticarboxypeptidase antibodies has not been developed. In the original assay format utilizing reactivity with recombinant Escherichia coli plaques, one of four prediabetics had anti-carboxypeptidase H antibodies.
ICA12
In the screening of islet expression libraries with patient autoantibodies Rabin and coworkers identified a molecule they termed ICA12 (9) in addition to ICA512, discussed above. The complete sequence of ICA12 is now known and radioassays have been developed with in vitro transcription and translation of ICA12. ICA12 is the transcription factor SOX13. Only a small percentage of patients with new onset diabetes express ICA12 autoantibodies (35). It is not clear whether the percentage of patients positive for ICA12 autoantibodies, compared to controls with other autoimmune disorders, is high enough for ICA12 autoantibodies to be truly type 1 diabetes related, including increased levels in primary biliary cirrhosis (188,189).
Partially Characterized Autoantigens
52 kd
A 52 kd antigen distinct from carboxypeptidase has been identified by Western blotting of human islet extracts with diabetic sera (19;190). Interestingly, this antigen that appears to be islet cell-specific shares an antigenic determinant with the protein PC2, a component of the rubella virus capsid. The rubella virus is as yet to the only virus that has been clearly associated with the development of diabetes and only congenital rubella infection is associated with diabetes. The 52 kd antigen appears to be expressed within beta cell granules (190).
37-38 kd
A number of groups have reported antibodies to 38 kd molecules in patients with Type 1A diabetes and their first-degree relatives. In the original description of antibodies precipitating the 64 kd molecule form islets, it was shown that some sera also precipitated a 38 kd molecule. Honeymann and coworkers (191) identified a 38 kd autoantigen as the nuclear transcription protein jun-B from an islet expression library. The relevance of this antigen to autoantibodies of prediabetics is unclear. An additional 38 kd molecule has been described by Pak and coworkers (192). Following their observation that some new-onset diabetic patients have cytomegalovirus infection, they immunized BALB/c mice with cytomegalovirus and obtained islet-reactive monoclonal antibodies that by Western blot recognize a 38 kd human islet protein.
Roep and coworkers (193) have identified a 38 kd islet granule autoantigen to which T cells react. The relationship between this antigen recognized by T cells and autoantibodies to 38 kd molecules is unknown.
155 kd
In 1990, McEvoy and coworkers used a novel method to detect diabetes-associated autoantigens. The developed a panel of mouse monoclonal antibodies reacting with the rat insulinoma cell line RIN5F and displaced the binding of the monoclonals to the tumor cells with diabetic sera. The binding of one of these monoclonals, termed 1A2, was selectively displaced by sera from diabetic children. The 1A2 monoclonal binds specifically to the insulinoma cell line RIN5F but not the other rat or human tissues. In a subsequent communication (193), the authors reported that monoclonal 1A2 recognizes by Western blot a 150 kd membrane-pound protein DAP1 (diabetes associated protein 1). Several sera from IDDM patients also identified a 150 kd band by Western blotting of rat brain homogenates. Using an assay based on displacement of the radiolabeled monoclonal binding to rat insulinoma cells, 159 of 169 sera of children with IDDM versus 10 of 351 age-matched control sera were positive for autoantibodies to DAP1. Autoantibodies to this antigen are, however, also present in a large percentage of nondiabetic first-degree relatives (> 60%).
GLIMA38
Baekkeskov and coworkers have immunoprecipitated a glycosylated islet cell membrane antigen which they have termed GLIMA 38 (Glycated Islet cell Membrane-Associated protein). The molecule of 38 Kd was immunoprecipitated with sera from 19% of patients with new onset diabetes (34). Winnock and coworkers analyzed 100 new onset patients and 23 prediabetic siblings and found GLIMA 38 autoantibodies (by precipitation) in 38 and 35% respectively versus 0% in control subjects (194). In that almost all of the positive patients expressed another anti-islet autoantibody (e.g. IA-2 autoantibodies) they concluded that GLIMA 38 assay did not enhance the prediction or classification of diabetes.
Osteopontin and Importin beta
Following the screening of a random peptide library antibodies were found that reacted with the molecule Osteopontin. Osteopontin is reported to be expressed in somatostatin producing islet cells. By radioassay patients with new onset diabetes did not differ from controls in the percentage of individuals with higher levels of osteopontin antibodies, with the ELISA assay finding more positives (38). In a similar manner the molecule Importin beta was associated with autoimmunity with 60% of type 1 patients with ELISA assay and 30% of patients with other autoimmune disorders having antibodies (195). With a fluid phase radioassay 6.3% of patients exceeded 99th percentile of normal controls with relatively low counts precipitated.
Nephrin, Densin and Filtrin
A number of molecules are expressed in the kidney and in the islets including nephrin (196), densin and filtrin. Given the presence of the strong association of nephropathy with type 1A diabetes, sera from patients with type 1A diabetes were evaluated for autoantibodies to these molecules (41) using fluid phase radioimmunoprecipitation assay. Densin autoantibodies were detected in 33% of patients versus 2% of controls. Antibodies to filtrin were present in 11% versus 3% of controls.
CD38:ADP-ribosyl cyclase/cADPR hydrolase
Several groups have described antibodies to CD38 in patients with type 2 diabetes and type 1 diabetes using either Western blot assays, or an enzymatic immunoassay on immobilized recombinant CD38 (42,197,198). Prevalence of the reported autoantibodies are modest (e.g. 14% for type 2 patients versus 1% for controls) with lower positivity for type 1 diabetes.
SOX13(ICA12)
With an ELISA assay antibodies to SOX13, a transcription factor which is ICA12 (36) was found in 18% of patients with type 1 diabetes, and a similar percentage of patients with hepatic autoimmunity (188,199). The full length clone of SOX13 has been expressed and a series of patients evaluated for autoantibodies reacting with SOX13. 7.6% of new onset pediatric patients had anti-SOX13 antibodies exceeding the 99th percentile of normal controls, while no adult patient or adult control had such antibodies (35).
Other Autoantigens
Reports from Elias and coworkers (27) suggested a role for T cell clones reacting to the heat shock protein (HSP)65 in the pathogenesis of diabetes in the NOD mouse (200). In particular, these authors identified a peptide within the sequence of human HSP that was recognized by T cell clones that could transfer a transient form of diabetes and hyperglycemia (201). The administration of irradiation-attenuated HSP-specific T cells, or of the HSP peptide p277, prevented diabetes in the NOD mouse (controversial) or attenuated diabetes in the BB rat with a hydrolysed casein diet (202,203). Despite initial reports, HSP65 is apparently not an autoantigen of humoral autoimmunity in man (204), though this again is controversial (205). Studies of heat shock protein have progressed from animal studies (206) to phase I/phase II clinical trials with a report of positive results in a trial of the peptide p277 in adult patients with new onset type 1 diabetes but apparently there was no effect in children (29).
In 1990, Johnson and coworkers (207,208) suggested that an islet-specific glucose transporter is an autoantigen in Type 1A diabetes. The evidence supporting this conclusion derived from the ability of IgG fractions from diabetic patients to inhibit glucose uptake from rat islet cells. This inhibitory activity could be removed by incubation of the IgG fraction with islet cell and hepatocyte membranes (which express two types of glucose transporter, GLUT-1 and GLUT-2), but not with erythrocytes or brush border membranes (which express a only GLUT-2). Sera from 26 of 27 patients with IDDM inhibited the rat islet glucose uptake versus 0 of 5 sera from patients with NIDDM. Interestingly, the uptake of L-leucine was not affected by the IgG fractions. Though no direct biochemical data are provided for reaction of antibodies with GLUT-2, a publication indicates that introduction of the GLUT-2 gene into cell lines results in detection of surface autoantibodies (207), implying that GLUT-2 is recognized by autoantibodies of diabetic patients.
T Cell Reactivity
Many of the target molecules of autoantibodies described in this chapter have been reported to drive proliferation or cytokine secretion of T cells, or produce tetramer positive T cells from animal models and patients with Type 1A diabetes or from islet autoantibody positive individuals.  
In animal models a number of T cell assays, utilizing T cells obtained from spleen, pancreatic or peripheral lymph nodes, readily detectable autoreactivity. For the autoantigen (IGRP) T cells are detectable in peripheral blood (209) and even can be imaged in vivo (210). Islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) was recently identified as the Beta-cell specific antigen targeted by a highly prevalent and pathogenic population of CD8+ T cells of NOD mice (211,212). The IGRP peptide, amino acids 206-214, is naturally processed and presented and a tetramer with a high affinity peptide analogue was developed. A higher percentage of T lymphocytes in the circulation of NOD mice reacting with this NRP peptide (presented by K^d) is associated with progression to diabetes (213,214). A modified high affinity peptide termed NRP-V7 of the IGRP major CD8 autoantigen has been utilized to produce a class I tetramer.  Despite this prominence knocking out the IGRP gene (Hutton oral communication) or inducing tolerance to IGRP (215,216) does not alter progression to diabetes of NOD mice.  Thus in some ways the immune response to IGRP may mark islet autoimmunity of the NOD similar to presence of insulin autoantibodies and not be a primary driver of disease.  Nevertheless IGRP peptides can be utilized to drive CD8 T cells that can prevent diabetes (IDS 2009 Santamaria oral presentation).  To date in neither the NOD mouse nor man have anti-IGRP autoantibodies been found (Hutton et al, unpublished observations). 
A putative autoantigen for T cells of the NOD mouse is dystrophia myotonica kinase, a widely distributed molecule that may be the target for AI4 cells (217).  In that the molecule has been defined through studies primarily of mimotopes and it is widely distributed, one hypothesis is that the relevant islet autoantigen might not yet be defined, and DMK has a mimotope sequence.
Another well studied CD8 epitope is insulin peptide B:15-23 that appears to be an early eptiope in insulitic lesions but not nearly as prominent as IGRP reactive T cells (218). In addition to the tetramer CD8 assays, assays utilzing ELISPOT responses and proliferation, especially after peptide immunization give clear CD4 responses. Wegmann and coworkers have found that a large percentage of CD4 T cells infiltrating islets of NOD mice strongly proliferate to insulin and a B-chain peptide of insulin (B:9-23 (239,240)) and the same peptide can be used to prevent diabetes (241). In contrast, a minimal response of splenocytes to insulin is found. A subset of these T cell clones are able to transfer diabetes into immunodeficient NOD mice and thus are sufficient for disease pathogenesis. A response to insulin is seen as early as 4 weeks of age. This study in NOD mice, where islets are available for analysis, highlights a potential difficulty of studies in man utilizing peripheral lymphocytes. The T cell clones of Wegmann and coworkers recognizing insulin B-chain peptide B:9-23 recognize one of two distinct peptides, B:9-16 or B:9-23 (242). The T cell clones utilized a dominant Vα chain, Vα13.3 (also termed TRAV5D-4*04) (243) and T cells with this alpha chain as a Calpha only transgenic or retrogenic are able to induce insulin autoimmunity (177,244). Alleva and coworkers have analyzed stimulation of human lymphocytes with the B:9-23 insulin peptide (identical sequence in mouse and man insulin) and find reactivity amongst patients but not controls (245). Kent and coworkers isolated T cell clones reacting with an insulin A chain determinant from pancreatic lymph nodes of man (246). In the NOD mouse model with an insulin peptide tetramer (B chain amino acids 15 to 23) autoreactive T cells can be detected particularly within islets (247). In addition to insulin peptides, proinsulin peptides and in particular a peptide spanning the B chain C-peptide junction is a target of autoimmunity (248,249,250,251).
In addition tetramers have been developed utilizing the mimotopes of the BDC2.5 clone that was just reported to respond to a chromagranin peptide (Haskins and coworkers oral communication IDS 2009) (219). This clone and transgenic have provided a wealth of data concerning the pathogenesis of diabetes of the NOD mouse (220,221,222).
Atkinson and coworkers (228,229) studied the reactivity of peripheral blood mononuclear cells to recombinant glutamic acid decarboxylase in newly diagnosed type 1 patients, ICA-positive and ICA-negative relatives of individuals with diabetes, and in healthy controls. A higher proportion of subjects gave a positive proliferative response to GAD among the newly diagnosed patients than among the controls. Interestingly, nondiabetic ICA-positive relatives were also more likely to express T cell reactivity to GAD (63% versus 11%). Harrison and coworkers (230) reported proliferation of peripheral blood T cells in presence of the central region of GAD67 (aa 208-404) in 38% of newly diagnosed IDDM patients and in 41% of ICA-positive relatives (> 20 JDF units). Only 4% of HLA-matched controls responded to GAD67. In a subsequent report, Harrison and coworkers described an assay utilizing fluorescent DNA labeling (fluorescent dye 5,6-carboxylfluorescein diacetate succinimidyl ester [CFSE]) (231).
Assays for reactivity of peripheral blood lymphocytes with GAD give relatively low stimulation indices (232). In the NOD mouse, utilizing spleen cells, Tisch and coworkers (233) reported that T cell proliferative responses to GAD precede responses to other antigens. The stimulation indices again are low but of interest and T cell clones to GAD have been derived (234). Nepom and coworkers have isolated a peptide of GAD65 and have developed T cell tetramer (DRB1*0401) assays (235) for reactivity with this epitope (236), and tetramer assays have evaluated responses post pancreatic transplantation (237).
The study to date of T cell responses of man to islet autoantigens has been particularly difficult with relatively low levels of stimulation, when stimulation is observed. This may relate to a very low frequency of autoreactive T cells outside of the islets. In an international workshop with multiple laboratories studying a series of "blinded" islet antigens and peptides none of the laboratories could distinguish a small panel of patients with diabetes from control individuals (223,224). We believe it is likely that the assay formats will have to be greatly improved to be able to reliably detect T cell responses and there has been recent progress in particular for CD8 T lymphocytes (225,226,227).
Testing proliferation to multiple autoantigens by Dosch and coworkers and proliferation to multiple bands on polyacrylamide gel chromatography of human islets by Brooks-Worrell and coworkers indicated that the combined responses of each assay significantly distinguished between “workshop” provided control and patient samples (238). The general approach of adding responses to multiple different antigens/antigenic determinants allows distinction between control and patient samples, though with specificities usually far below that of most autoantibody assays (238).     
Characterization of T cell reactivity has been particularly difficult in man and we believe this results from lack of ability to obtain T lymphocytes from the site of pathology, namely pancreatic islets. In an effort to obtain access to such T cells we have initiated a program where cadaveric organ donors are screened in real time for the expression of anti-islet autoantibodies. Between 1/100 and 1/300 of such donors will express multiple anti-islet autoantibodies (252,253) and we predict that the pancreas from such individuals (that can be obtained at the time of organ donation), will harbor relevant T cell clones.

Figure 10.10. Development of anti-islet autoantibodies following islet transplantation.

Islet Transplantation
With the initial success of the "Edmonton Protocol" transplantation of islets has increased, especially in patients with severe recurrent hypoglycemia on insulin therapy for type 1A diabetes (254). With followup most patients unfortunately eventually require resumption of insulin therapy. For both islet and pancreatic transplantation there is evidence that recurrent anti-islet autoimmunity may contribute to failure of grafts to reverse hyperglycemia or for patients with stable grafts to lose function and in particular lose insulin independence. As illustrated in Figure 10.10 a subset of patients (approximately 10%) with islet transplantation can have a dramatic rise in islet autoantibodies and this has been associated with the loss of the graft (254). Stduies of the influence of anti-islet autoantibodies (with determination of biochemical anti-islet autoantibodies) on graft function were sponsored by the Immune Tolerance Network as part of the evaluation of the Edmonton protocol (255).
Conclusions
During the past several years, investigators have defined a growing family of islet autoantigens recognized by autoantibodies of prediabetics. The ability to produce many of the antigens with recombinant technology (insulin, GAD, ICA512) will facilitate the prediction of Type 1A diabetes. In addition, availability of such autoantigens will be important to elucidate the pathogenesis of Type 1A diabetes. It is likely that such a large number of autoantibodies are secondary to beta cell destruction and the presentation of normal antigens in an inflammatory environment. Nevertheless, studies in the NOD mouse suggest that influencing the immune response to several autoantigens (e.g., oral insulin, intravenous GAD, subcutaneous insulin) will prevent diabetes. At present, the primacy in disease pathogenesis of any of the defined autoantigens is unknown in man, with considerable evidence that insulin may a primary autoantigen of the NOD mouse (176,215).

Figure 10.11. GAD65 and IA-2(ICA512bdc) autoantibodies of the DPT-1 (Diabetes Prevention Trial-1 study.

Despite uncertainty about pathogenic importance, the ability to biochemically detect autoantibodies to a series of autoantigens greatly facilitate the prediction of Type 1A diabetes (Figure 10.11,10.12). Utilizing only four radioassays for autoantibodies (insulin, GAD, IA-2 and ZnT8) 95% of prediabetic or recent-onset diabetic children express one or more antibodies (versus only 4 out of 100 controls). More than 80% express two or more autoantibodies, versus "no" controls. Approximately 25% of diabetics or prediabetics express all of the above autoantibodies. The expression of autoantibodies assorts independently in prediabetic individuals and thus, by utilizing the three assays, a high positive predictive value for disease is possible while retaining disease sensitivity (Combinatorial Autoantibody Prediction (83,256,257)).

Figure 10.12. Combinatorial Autoantibody prediction (Modified from Verge et al. Diabetes 47:1857-1866, 1998).

"Biochemical" autoantibody screening will likely have its largest impact for the prediction of Type 1A diabetes in the general population where the positive predictive value of ICA positivity is less than 10%, while presence of two or more of four autoantibodies should give positive predictive values greater than 70%. Screening of the general population for diabetes risk is likely to be a major importance if trials for the prevention of Type 1A diabetes among first-degree relatives are successful. Approximately 90% of individuals who develop Type 1A diabetes have no first-degree relatives with the disease and thus such screening may be essential for the large-scale prevention of Type 1A diabetes.
The figure below (Figure 10.13) illustrates a number of caveats of screening for anti-islet autoantibodies in "low" risk populations, in this case, relatives of patients with type 1 diabetes and HLA characterized individuals from the general population with increased risk (e.g. DR3/DR4-DQ8) of diabetes (more than 30,000 newborns from the general population were HLA typed from cord blood in the DAISY study to identify those with high risk genotypes) (51). Overall both groups (HLA general population identified and first degree relatives without HLA typing)  have an approximate risk of diabetes of 1/20 compared to a Denver population risk of 1/300. As is apparent from the figure a rule of 1/3 captures much of the data. One third are false positive (50/162= 31%) and though found positive for one of the three anti-islet autoantibodies on initial assay (of GAD65, IA-2 or insulin autoantibodies, despite having repeated the assay in duplicate at least twice), were found to be negative when two blinded aliquots of the same sample were re-analyzed in the laboratory (51) The cutoff for positivity for each assay was set at the 99 th percentile of normal controls, and as expected those samples just above this cutoff had a much higher probability of the sample not confirming. One third of all individuals whose first sample confirmed and had a second independent blood sample positive (22/74=30%) had transient autoantibodies. None of the individuals with transient autoantibodies has developed diabetes to date. Of the remaining group with persistent anti-islet autoantibodies, 1/3 (24/74=32%) have already progressed to diabetes, and we suspect the great majority of the remaining 1/3 with persistent anti-islet autoantibodies will progress to diabetes. The DAISY study follows children from birth, and the expression of anti-islet autoantibodies at an early age is a very high risk characteristic, but a significant number of individuals have autoantibodies that do not confirm or that are transient. In epidemiologic and certainly interventional studies and clinical practice these characteristics of screening low risk populations are essential to consider, and only expression of multiple islet autoantibodies provides a high positive predictive value.

Figure 10.13. Confirmation and Persistance of islet autoantibodies (GAD65,IA-2,insulin).

We are at a somewhat unusual phase in the application of our knowledge of humoral autoimmunity of type 1A diabetes. Namely we can predict the disease, identifying varying degrees of diabetes risk, such that trials for the prevention of type 1A diabetes are feasible. Unfortunately at present we do not have a proven and safe therapy to delay or prevent type 1A diabetes. As analyzed in a number of studies increased anxiety is associated with screening for serious disorders (258,259), but most of the studies that have addressed this issue have found relatively mild and transient increases in such anxiety with islet autoantibody screening (260,261,262). Balancing such an increase in anxiety at present, it is very clear that screening for anti-islet autoantibodies coupled with metabolic and close monitoring dramatically decreases the development of ketoacidosis at diabetes diagnosis, such that only one child of thirty in the DAISY study required hospitalization verus 40% of children in the general population presenting with diabetes (263). This dramatic difference is almost certainly due to earlier diagnosis and many of the children from the general population had glucoses greater than 1,000 mg% while none of the DAISY children had glucose greater than 400 mg% at diagnosis. Though there is clinical benefit in earlier diagnosis the major impetus for identifying at risk individuals will be driven by trials for disease prevention. It is likely that islet autoantibodies will always be a useful, though incomplete reflection of pathogenesis, and assays for autoreactive pathogenic T lymphocytes are essential to accelerate the pace of clinical trials. International networks are now available to foster translational research in the area of type 1A diabetes, with Trialnet and the Immune Tolerance Network, both of which have web pages and an open solicitation for novel trials and in the case of the Immune Tolerance Network, novel assays.

Reference List - links to PubMed available in Reference List.

Click to download Powerpoint slide set - Updated 4/08, new slides #1 - #4

For comments, corrections or to contribute teaching slides, please contact Dr. Eisenbarth at: george.eisenbarth@ucdenver.edu

Back to Type 1 Diabetes Table of Contents