THE AUTOIMMUNE destruction of pancreatic cells

0021-972X/00/$03.00/0 Vol. 85, No. 1 The Journal of Clinical Endocrinology & Metabolism Printed in U.S.A. Copyright 2000 by The Endocrine Society Autoantibody Recognition of COOH-Terminal Epitopes of GAD65 Marks the Risk for Insulin Requirement in Adult-Onset Diabetes Mellitus* ALBERTO FALORNI, GIOVANNI GAMBELUNGHE, FRANCESCA FORINI, GEORGIA KASSI, ANNA COSENTINO, PAOLA CANDELORO, GEREMIA B. BOLLI, PAOLO BRUNETTI, AND FILIPPO CALCINARO Department of Internal Medicine and Endocrine and Metabolic Sciences, University of Perugia, Perugia, Italy ABSTRACT Some type 2 diabetic subjects develop secondary failure to sulphonylurea treatment and require insulin therapy. To test the diagnostic sensitivity and specificity of epitopes of GAD65 autoantibodies (GAD65Ab) for insulin requirement, in patients with latent autoimmune diabetes of the adult, we studied 569 adult subjects with a clinical diagnosis of type 2 diabetes mellitus. All the patients had been initially treated with hypoglycemic agents and/or diet for at least 1 yr. The presence of GAD65Ab (61/569, 10.7%) depended on insulin therapy (P 0.0001), low BMI (P 0.0001), and low basal C-peptide (P 0.01). The majority of GAD65Ab-positive subjects (47/61, 77%) had antibodies directed to both middle (GAD65-MAb) and COOH-terminal (GAD65-CAb) epitopes. However, GAD65-CAb were more frequent in insulin-treated subjects (92% of GAD65Ab individuals) than in subjects treated with hypoglycemic agents and/or diet (18.2% of GAD65Ab individuals), while the exclusive presence of GAD65- THE AUTOIMMUNE destruction of pancreatic cells that causes type 1 (insulin-dependent) diabetes mellitus is made evident by the appearance of circulating islet autoantibodies (1). The islet autoantigens, targets of autoantibodies, include insulin (2), the enzyme glutamic acid decarboxylase (GAD65) (3), the tyrosine phosphatase-like proteins IA-2 and IA-2 (4), and a long series of other less characterized autoantigens (5). Animal studies have shown that GAD65 plays a key role in the pathogenesis of autoimmune diabetes, at least in these models (6 8). While the prevalence of insulin autoantibodies (IAA) and of IA-2 autoantibodies (IA-2Ab) shows an inverse correlation with the age at onset of type 1 diabetes (9, 10), GAD65 autoantibodies (GAD65Ab) are frequently found in both diabetic children and adults, and the presence of this marker has a high diagnostic sensitivity in the 20 39-yr age group (11). The presence of GAD65Ab in subjects diagnosed as having type 2 diabetes mellitus identifies a subgroup at high risk for Received June 29, 1999. Revision received August 30, 1999. Accepted September 3, 1999. Address correspondence and requests for reprints to: Alberto Falorni, M.D., Ph.D., Department of Internal Medicine and Endocrine & Metabolic Sciences, Via E. Dal Pozzo, 06126 Perugia, Italy. E-mail: falorni@dimisem.med.unipg.it. * This study was supported in part by a grant from the Juvenile Diabetes Foundation International. The financial support of Telethon (Grant E.C787) is also gratefully acknowledged. MAb was more frequent in subjects treated with hypoglycemic agents and/or diet (81.8% vs. 8%) (P 0.0001). The presence of GAD65-CAb had a diagnostic specificity for insulin requirement as high as 99.4% (compared with 96.9% of GAD65Ab as measured in the traditional radiobinding assay) and identified a subgroup of patients with low BMI, low basal C-peptide values, and a need for insulin therapy. Subjects carrying only GAD65-MAb were phenotypically indistinguishable from GAD65Ab-negative patients. Patients positive for GAD65-M CAb, but not those positive for GAD65-MAb only, showed an increased risk for thyroid autoimmunity, as revealed by the presence of thyroid peroxidase autoantibodies. Our study demonstrates that the use of epitope-specific antibody assays improves the diagnostic specificity of GAD65Ab, and that the presence of GAD65Ab binding to COOH-terminal epitopes is strongly associated with a need for insulin requirement. (J Clin Endocrinol Metab 85: 309 316, 2000) insulin dependency (12, 13). In Japan, the secondary failure to respond to sulphonylurea treatment is retrospectively classified as slowly progressive type 1 diabetes (14, 15), while in western countries the presence of GAD65Ab identifies the so-called latent autoimmune diabetes of the adult (LADA) (16 21). Because of the high diagnostic sensitivity for autoimmune diabetes, the presence of GAD65Ab is currently used to identify subjects at high risk for the disease (22, 23). However, the not infrequent occurrence of this marker in healthy subjects and in patients with other autoimmune diseases not necessarily associated with type 1 diabetes, such as the rare neurological stiff-man-syndrome (24), the autoimmune polyendocrine syndrome type I (25), or Graves disease (26), limits its diagnostic specificity. In adult diabetic subjects, the concomitant presence of islet cell antibodies (ICA) improves the predictive value of GAD65Ab for insulin requirement (19). The analysis of the variable region of human monoclonal autoantibodies revealed that the production of diabetes-associated GAD65Ab is the result of an autoantigen-driven process, which selects high-affinity autoantibodies (27). During the natural history of childhood type 1 diabetes, autoantibodies directed to highly conserved epitopes located in the 240 360 (middle: M) and 451 570 (carboxyterminal: C) amino acid regions of the autoantigen (28 32) are selected. Production of epitope-specific GAD65Ab may require a permissive genetic background as GAD65Ab found in stiff-man- 309

310 FALORNI ET AL. JCE&M 2000 Vol 85 No 1 syndrome or in the autoimmune polyendocrine syndrome type I, both of which have different genetic associations than type 1 diabetes, are directed to a different and broader spectrum of epitopes (31, 33, 34). Thus, the mechanisms of GAD65Ab production may be different in different diseases. Preliminary data suggest that an increase in GAD65Ab to COOH-terminal epitopes distinguishes diabetic from healthy children (32). Although it has been proposed that LADA-associated GAD65Ab would recognize both a middle and a carboxyterminal epitope (35), it is still unclear what is the diagnostic sensitivity and specificity of GAD65Ab epitopes for insulin requirement in adult diabetic subjects with slowly progressive type 1 diabetes. During long-term chronic autoimmunity, either spreading or restriction of autoantibody epitopes may occur. Because LADA is the result of a long-term, chronic autoimmune process, the accurate localization of GAD65Ab epitopes associated with this form of diabetes would provide important information to our understanding of the molecular mechanisms of diabetesassociated autoantibody production. In addition, the identification of epitope-specific GAD65Ab with high diagnostic specificity for insulin requirement may be instrumental in the accurate diagnosis of LADA subjects. The major aim of our study was to estimate the diagnostic sensitivity and specificity for insulin requirement of GAD65Ab epitopes in subjects with adult-onset diabetes. We took advantage of cdnas for chimeric molecules generated by substitution of regions of human GAD65 with homologous regions of GAD67, which have been previously used to estimate the diagnostic accuracy of GAD65Ab epitopes for childhood type 1 diabetes (32). We first identified a group of GAD65Ab-positive subjects from a large population of adult patients diagnosed with and treated for type 2 diabetes mellitus. We then tested the diagnostic sensitivity and specificity of epitopes of GAD65Ab for insulin requirement. Finally, the risk for thyroid autoimmunity, as made evident by the presence of thyroid peroxidase (TPO) autoantibodies, was evaluated in relation to the occurrence of epitope-specific GAD65Ab. Although our initial screening of type 2 diabetic subjects was cross-sectional, follow-up samples from some of the GAD65Ab-positive subjects were available and were used to study the time-related modifications of levels of GAD65Ab epitopes. Subjects Subjects and Methods A total of 569 Italian adult diabetic subjects were studied in the initial screening for GAD65Ab. Subjects were randomly selected among those consecutively attending the diabetes unit of our department between October 1997 and March 1999. Inclusion criteria were: 1) an age of 25 yr or more at the time of the clinical diagnosis of diabetes mellitus, 2) classification as Type 2, noninsulin-dependent diabetes mellitus, and 3) treatment with diet and/or hypoglycemic agents for at least 1 yr after the initial diagnosis. The clinical features of the patients enrolled in our study are summarized in Table 1. Fasting serum samples were obtained from each subject and used to analyze the levels of GAD65Ab, GAD65Ab epitopes, and basal C-peptide. A total of 23 additional follow-up samples from 21 GAD65Abpositive subjects were also available. This follow-up material included retrospective samples, collected from 7 subjects before the beginning of the insulin treatment. An informed consent was obtained from all the study participants. GAD65Ab assays GAD65Ab were determined using a radiobinding assay with in vitro translated recombinant human 35 S-GAD65 (36, 37). The cdna for human GAD65 (38, 39) was a kind gift of dr. Å Lernmark (University of Washington, Seattle, WA). GAD65Ab levels were expressed as a relative index (GAD65 index), using one positive standard serum from a type 1 diabetic subject and two negative control sera from healthy subjects, in each assay. Our autoantibody assay was validated by participation in international workshops for the standardization of islet autoantibody determinations (40 42), and the upper level of normal was 0.035, as calculated as the mean 3 sd of the GAD65 index observed in healthy controls. Using this cut-off value, the diagnostic sensitivity and specificity measures of our GAD65Ab assay, as evaluated in the 1995 Combined Autoantibody Workshop (lab code AT) (42), were 85% and 100%, respectively. To localize the GAD65Ab epitopes associated with LADA, we used radiobinding antibody assays with in vitro translated 35 S-GAD65/ GAD67 chimeric molecules, generated by substituting GAD65 regions with homologous regions of rat GAD67 (32). The cdna for rat GAD67 (43) was a kind gift of dr. B. Michelsen (Hagedorn Research Institute, TABLE 1. Clinical features of the 569 adult diabetic subjects subdivided for GAD65Ab epitopes Total population (n 569) GAD65Abpositive subjects (n 61) GAD65Abnegative subjects (n 508) GAD65-MAb only (n 13) Both GAD65-M and -CAb (n 47) Males/Females ratio 299/270 27/34 272/236 7/6 20/27 Age at diagnosis (yr) 50 (25 66) 46 (25 62) 50 (25 66) 45 (25 62) 46 (29 60) BMI at diagnosis 27 (17.5 45) 24 (17.5 32) 27.2 (18.8 45) 26 (18.4 31.2) 22.8 (17.5 32.4) a Age at screening (yr) 59 (31 72) 55 (31 66) 59 (31 72) 55 (34 66) 55 (31 66) Time after diagnosis (yr) 7 (1 46) 8 (1 46) 7 (1 35) 7 (1 46) 8 (1 21) Insulin therapy 217/569 (38.1%) 50/61 (82.0%) 168/508 (33.1%) 4/13 (30.8%) 45/47 (95.7%) b Prevalence of TPOAb c nd 15/61 (24.6%) d 9/174 (5.2%) 1/13 (7.7%) 14/47 (29.8%) d Basal C-peptide (nmol/l) 0.78 (0 8.23) 0.063 (0 2.0) e 0.86 (0 8.23) 0.46 (0 2.0) 0.023 (0 0.67) f Values of age, time after diagnosis, BMI, and basal C-peptide are median (range). Prevalence of subjects in insulin therapy is shown. One subject was found positive for GAD65-CAb in the absence of GAD65-MAb and was excluded from this analysis. a P 0.025 vs. GAD65-MAb only. b P 0.0001 vs. GAD65-MAb only. c TPOAb levels were determined in all the GAD65Ab-positive patients and in a control group of 174 GAD65Ab-negative diabetic subjects. nd not determined. d P 0.0001 vs. GAD65Ab-negative. e P 0.01 vs. GAD65-Ab-negative. f P 0.01 vs. GAD65-MAb only.

GAD65Ab EPITOPES IN LADA 311 Gentofte, Denmark). The DNA coding for a chimeric molecule containing the NH 2 -terminal region of rat GAD65 (aa. 1 83) in fusion with the central-cooh terminal regions of rat GAD67 (aa. 89 593) (44) (GAD65- N-67) was a kind gift of dr. P. De Camilli (Yale University, New Haven, CT). GAD65-M was generated by substituting the amino acid region 361 585 of human GAD65 with the region 369 593 of rat GAD67 (GAD65 1 360 /GAD67 369 593 ), and it contained the middle epitopes of GAD65Ab; GAD65-C was generated by substituting the amino acid region 1 436 of human GAD65 with the region 1 445 of rat GAD67 (GAD67 1 445 /GAD65 437 585 ) and contained the COOH-terminal epitopes of GAD65Ab (32). The rabbit antiserum R7309 (45), specific for the region 4 22 amino acids of human GAD65, was used as positive standard serum in the GAD65-M and GAD65-N-67 antibody assays, and the rabbit antiserum R10266 (45), specific for the region 2 19 amino acids of rat GAD67, was used as positive standard serum in the GAD67 and GAD65-C antibody assays. Epitope-specific GAD65Ab levels were expressed as relative indices (GAD67 index, GAD65-N-67 index, GAD65-M index, and GAD65-C index) using 2 negative control sera from healthy individuals. Using the mean 3 sd of the results obtained with the sera from 100 healthy subjects, the upper level of normal of the GAD67Ab assay was 0.02, that of the GAD65-N-67Ab assay was 0.02, that of the GAD65-MAb assay was 0.02, and that of the GAD65-CAb assay was 0.025. Sera from all the GAD65Ab-positive adult diabetic subjects identified in our study were separately tested in all the antibody assays using GAD67 or the three forms of GAD65/GAD67 chimeric molecules. To correct for the inter-assay variation, all the follow-up samples were analyzed in a single assay. The intra-assay coefficients of variation were 5% for the GAD65Ab assay, 7% for the GAD67Ab assay and for the GAD65-N-67Ab assays, 6% for the GAD65-MAb assay, and 8% for the GAD65-CAb assay. Thyroid peroxidase antibody assay Thyroid peroxidase (TPO) autoantibodies (TPOAb) were determined using an immunoradiometric assay with recombinant human 125 I-TPO (Radim, Angleur, Belgium). Results were expressed as international units per milliliter of bound TPO, and the upper level of normal was 130 IU/mL. C-peptide assay Basal plasma C-peptide values were evaluated using a commercially available RIA kit (Diagnostics Systems Laboratories, Inc., Webster, TX). Sensitivity of the assay was 0.003 nmol/l. Statistical analysis Differences in frequency of autoantibodies were tested by the -square test with Yates correction, whenever appropriate, or by the Fisher s exact test. Differences in median and range were tested by the nonparametric Mann-Whitney test. Differences in autoantibody levels during the follow-up were tested by the nonparametric Wilcoxon test for paired samples. Autoantibody levels were not normally distributed in diabetic subjects and were therefore logarithmically transformed. The relationship between autoantibody levels and the duration of diabetes was analyzed by linear regression. The dependence of the dichotomous GAD65Ab positivity on several other variables age, gender, disease duration, type of therapy, BMI and basal C-peptide was assessed by logistic regression analysis. A p value less than 0.05 was considered significant. All statistical tests were performed by the Statistical Package for Social Sciences for Windows for personal computers (SPSS, Inc., Chicago, IL, USA). Sensitivity and specificity of GAD65Ab and GAD65Ab epitopes for insulin requirement were percentage of antibody-positive subjects among insulin treated patients (true positives) and percentage of antibody-negative subjects among noninsulin treated patients (true negatives), respectively. Results Prevalence of GAD65Ab in adult subjects with a diagnosis of type 2 diabetes mellitus Using in vitro translated human autoantigen, GAD65Ab were found in 61/569 (10.7%) Italian subjects with an initial diagnosis of type 2 diabetes mellitus. A total of 50/61 (82%) GAD65Ab-positive subjects had already been converted to a treatment with insulin at the time of the study. Using logistic regression analysis, the presence of GAD65Ab in adult diabetic subjects depended on (in order of decreasing strength) insulin therapy (P 0.0001), low BMI (P 0.0001), and low basal C-peptide values (P 0.01) (Table 1). When the prevalence of GAD65Ab was evaluated in relation to the type of therapy, GAD65Ab were found in 50/217 (23.0%) subjects in treatment with insulin and in only 11/352 (3.1%) subjects in treatment with oral hypoglycemic agents and/or diet. In autoantibody-positive subjects, GAD65Ab levels were significantly higher in insulin-treated subjects (GAD65 index median: 0.992; range: 0.12 3.02) than in noninsulin treated ones (GAD65 index median: 0.258; range: 0.066 1.45) (Fig. 1) (P 0.002). Furthermore, GAD65Ab levels correlated positively with the duration of diabetes (r 0.29, P 0.025). In the presence of GAD65Ab, no statistical significant difference in disease duration (as estimated in years after the initial diagnosis of type 2 diabetes mellitus) was observed between the insulin-treated (median: 8 yr, range: 1 29 yr) and the noninsulin-treated (median: 6 yr, range: 1 46 yr) patients. In summary, the presence of GAD65Ab identified a subgroup of type 2 diabetic subjects with features of a slowly progressive type 1 diabetes and high probability of need for insulin requirement. GAD65Ab epitopes in LADA Based on the results of the initial screening for the presence of GAD65Ab, we subsequently tested the diagnostic sensitivity and specificity of epitopes of GAD65Ab for insulin requirement. A total of 20/61 (32.8%) GAD65Ab-positive subjects had a GAD67 index above the upper level of normal. The presence of antibodies reacting with GAD67 was associated with high levels of GAD65Ab (GAD65 index median: 1.42, range: 0.124 2.59). Antibodies to the GAD65-N-67 chimera were found in 16/61 (26.2%) GAD65Ab-positive subjects. A total of 15 of the 16 samples found positive in the assay using the GAD65-N-67 chimera were found positive also for GAD67Ab. Because the majority of GAD65Ab-positive samples were negative for both GAD67Ab and GAD65-N-67Ab and because the presence of GAD65-N-67Ab was strongly associated with that of GAD67Ab, we concluded that the GAD65Ab detected in our LADA patients was primarily directed to epitopes located in the middle and COOHterminal regions of GAD65. A total of 48/61 (78.7%) and 60/61 (98.4%) GAD65Abpositive adult diabetic subjects demonstrated positive results in the assays using GAD65-C or GAD65-M chimeras, respectively. The simultaneous presence of GAD65-M and GAD65-C autoantibodies was demonstrated in 47/61 (77%) cases. All of the 20 samples found positive for GAD67Ab, as well as all of the 16 samples found positive for GAD65-N- 67Ab, were also positive for both GAD65-MAb and GAD65- CAb. We found GAD65-MAb in 13 subjects (21.3%) negative

312 FALORNI ET AL. JCE&M 2000 Vol 85 No 1 for GAD65-CAb. Conversely, GAD65-CAb were present in only 1 subject (1.6%) in the absence of GAD65-MAb. We then analyzed the presence of epitope-specific GAD65Ab in relation to the type of therapy at the time of the study (Fig. 1 and Table 2). Out of the 50 GAD65Ab-positive subjects in treatment with exogenous insulin, 45 (90%) were positive for both GAD65-MAb and GAD65-CAb; 1 (2%) was positive only for GAD65-CAb; and 4 (8%) were positive only for GAD65-MAb. Out of the 11 subjects in treatment with hypoglycemic agents and/or diet, 2 (18.2%) were positive for both GAD65-M- and GAD65-CAb, and 9 (81.8%) were positive only for GAD65-MAb in the absence of GAD65-CAb. The frequency of GAD65-CAb was significantly higher in insulin-treated subjects (46/50, 92%) than in noninsulin treated (2/11, 18.2%) subjects (P 0.0001). Conversely, the presence of GAD65-MAb in the absence of GAD65-CAb was significantly lower in insulin-treated (4/50, 8%) than in noninsulin-treated (9/11, 81.8%) subjects (P 0.0001). Furthermore, the clinical features of subjects positive for GAD65- MAb only (Table 1) were indistinguishable from those of GAD65Ab-negative individuals. On the contrary, the presence of autoantibodies directed to both the middle and the carboxyterminal epitope regions identified a subgroup of patients with low BMI and need for insulin requirement (Table 1). Basal C-peptide values were significantly lower in the subjects positive for both GAD65-M and -CAb (median 0.023 nmol/l, range 0 0.67) than in the subjects positive for GAD65-MAb only (median 0.46 nmol/l, range 0 2.0) (P 0.01) (Table 1). The levels of both GAD65-MAb and GAD65-CAb were higher in insulin-treated subjects than in noninsulin-treated subjects (Fig. 1) and correlated with those of GAD65Ab (r 2 0.74, P 0.001 and r 2 0.76, P 0.001, respectively). In addition, levels of GAD65-MAb and GAD65-CAb correlated with each other (r 2 0.71, P 0.001). Overall, the samples positive for both GAD65-M and GAD65-CAb had GAD65Ab levels (GAD65 index median: 1.102, range:0.12 3.021) higher than those observed in the samples positive for GAD65-MAb only (GAD65 index median:0.342, range:0.066 0.587) (P 0.01). Furthermore, the two samples from noninsulin-treated subjects found positive for both GAD65-M and GAD65-CAb were those with the highest GAD65 index (1.357 1.445) (Fig. 1). However, when samples with a GAD65 index equal to or lower than 0.587 (the highest GAD65 index associated with the presence of GAD65-MAb only) were taken into consideration, 11/15 (73.3%) insulin-treated subjects and 0/9 noninsulin-treated subjects were positive for GAD65-CAb (P 0.0006). The sensitivity of islet autoantibodies for insulin requirement ranged from 1.8% of GAD65-MAb only to 23% of GAD65Ab (Table 3). The specificity of GAD65-CAb (99.4%) was higher than that of either GAD65Ab or GAD65-MAb only (Table 3). FIG. 1. Levels of GAD65Ab (panel A), GAD65-CAb (panel B), and GAD65-MAb (panel C) in relation to the type of treatment. Subjects positive for GAD65-MAb only, in the absence of GAD65-CAb, are identified by an open square. Dotted lines are the upper level of normal in each assay. Note that the three antibody assays used different positive control sera and the y-axis scales are different.

GAD65Ab EPITOPES IN LADA 313 TABLE 2. Prevalence of epitope-specific GAD65Ab in 61 antibody-positive adult diabetic subjects subdivided for type of treatment Total population Insulin therapy Oral hypoglycemic agents and/or diet GAD65-CAb only 1/61 (1.6%) 1/50 (2%) 0/11 GAD65-MAb only 13/61 (21.3%) 4/50 (8%) 9/11 (81.8%) Both GAD65-CAb and GAD65-MAb 47/61 (77%) 45/50 (90%) 2/11 (18.2%) Numbers are frequencies in GAD65Ab-positive subjects. TABLE 3. Sensitivity and specificity of GAD65Ab epitopes for insulin requirement in type 2 diabetic subjects Number of positive subjects Follow-up of GAD65Ab epitopes Sensitivity (%) Specificity (%) GAD65Ab 61 23.0 96.9 Presence of M-GAD65Ab 60 22.6 96.9 Presence of C-GAD65Ab 48 21.2 99.4 Both C- and M-GAD65Ab 47 20.7 99.4 M-GAD65Ab only 13 1.8 97.4 Figure 2 shows the levels of GAD65Ab, GAD65-MAb, and GAD65-CAb in 16 insulin-treated subjects during follow-up. In 12/16 subjects, the levels of GAD65Ab increased during the follow-up period (P 0.045) (Fig. 2, panel A). In addition, the levels of GAD65-CAb increased during the follow-up period in 14/16 subjects (P 0.01) (Fig. 2, panel B). On the contrary, levels of GAD65-MAb did not increase significantly at follow-up (Fig. 2, panel C). Retrospective serum samples were available from seven subjects treated with insulin. These samples included one collected at the time of clinical diagnosis and six collected 6 months to 2.5 yr before the conversion to insulin therapy. In six of these seven latter cases, both GAD65-M and GAD65-C antibodies were detected in both the retrospective samples collected before the conversion to insulin therapy and the sample collected at the time of the screening for GAD65Ab. The seventh subject was positive only for GAD65-MAb in both the retrospective sample (collected 1 yr before the conversion to insulin therapy) and the sample collected at the time of the screening for GAD65Ab. TPO antibodies in LADA The presence of TPOAb was evaluated in 61 GAD65Abpositive subjects and in a group of 174 diabetic subjects (92 males and 82 females) of similar age, but negative for GAD65Ab. TPOAb occurred more frequently in GAD65Abpositive (15/61, 24.6%) (median levels: 555 UI/mL, range: 162-1766) than in GAD65Ab-negative (9/174, 5.2%) (median levels: 909 UI/mL, range: 144-2556) subjects (P 0.0001) (Table 1). In the presence of GAD65Ab, 14/15 (93.3%) TPOAb-positive subjects were in treatment with insulin at the time of our study. Similarly, 14/15 (93.3%) TPOAb-positive subjects had antibodies directed to both GAD65-M and GAD65-C. Accordingly, TPOAb were present in 14/47 (29.8%) subjects positive for both GAD65-M and C-Ab (P 0.0001 vs. GAD65Ab-negative subjects), but in only 1/13 (7.7%) with GAD65-MAb only. Discussion It is well known that GAD65Ab can be found in approximately 10% of type 2 diabetic subjects (12, 13, 16 21). Although the presence of GAD65Ab has a high predictive value for future insulin requirement, some subjects do not seem to progress towards insulin dependency in spite of the presence of markers of an islet autoimmune process. In a large study (19), performed by the United Kingdom Prospective Diabetes Study Group (UKPDS) on over 3,600 type 2 diabetic subjects, the specificity of GAD65Ab for insulin requirement resulted lower than that of ICA. Thus, the diagnostic specificity may be a limiting factor in the use of GAD65Ab to identify LADA subjects. In our study, we addressed the problem of the selection of GAD65Ab epitopes in adult subjects initially diagnosed with and treated for type 2 diabetes mellitus. We observed that GAD65Ab was present in 10.7% of the studied population from central Italy and that the presence of autoantibodies binding to the COOH-terminal region of GAD65 was strongly associated with the need for insulin therapy. The relatively high frequency of GAD65Ab in type 2 diabetic subjects is highly relevant for the understanding of the pathogenesis of autoimmune diabetes and for the classification of adult-onset diabetes mellitus. It must be noted that our study was carried out at a WHO collaborating center for the improvement of quality of diabetes care and that our GAD65Ab assay had a very high diagnostic sensitivity and specificity. Accordingly, it is unlikely that the results of our study were influenced by an inaccurate clinical diagnosis at the time of the first visit or by an inadequate accuracy of our assay. Rather, our GAD65Ab frequency of 10.7%, similar to those recorded in several other studies, confirms that slowly progressive type 1 diabetes mellitus has been so far underdiagnosed. On the other hand, it must be noted that 38% of our type 2 diabetic subjects were in insulin treatment at the time of our study. Thus, the majority of insulin-treated subjects was GAD65Ab-negative. The positive correlation between GAD65Ab levels and disease duration supports the hypothesis that a long-term, chronic, autoantigen-specific process is sustained for several years in adult subjects with latent and slowly progressive autoimmune diabetes. This conclusion is also supported by our data on follow-up samples. Thus, GAD65Ab can be considered a sensitive marker of chronic islet autoimmunity. However, the mechanisms responsible for maintaining a GAD65-specific autoimmune process with slowly progressive -cell destruction have not yet been clarified. The presence of GAD67Ab in type 1 diabetic subjects is associated with high levels of GAD65Ab and seems to be the result of a cross-reactivity of the two isoforms of the enzyme with some GAD65Ab (32, 36). Thus, chimeric GAD65/ GAD67 molecules can be used in epitope-specific assays. This strategy has previously proven successful in localizing conformation-dependent GAD65Ab epitopes (29 32). The strong association between GAD65-N-67Ab and GAD67Ab

314 FALORNI ET AL. JCE&M 2000 Vol 85 No 1 FIG. 2. Levels of GAD65Ab (panel A), GAD65-CAb (panel B) and GAD65-MAb (panel C) in 16 insulin-treated subjects at follow-up. Note that the three antibody assays used different positive control sera and the y-axis scales are different. SEM of each value is shown. observed in our study suggests that LADA-associated GAD65Ab is not directed to the N-terminal end of the autoantigen. This is in line with similar data obtained when using samples from type 1 diabetic children (32). Our finding of a high association between the appearance of COOHterminal epitope-specific GAD65Ab and insulin requirement, with a specificity as high as 99.4%, can be interpreted to indicate that the selective production of epitope-specific autoantibodies marks the need for insulin treatment. A previous study (32) showed that GAD65Ab binding to both the middle and the COOH-terminal region of the autoantigen is present in more than 80% of antibody-positive type 1 diabetic children. Furthermore, a selective increase in COOH-terminal specific autoantibodies distinguished diabetic from healthy children (32). Our data on LADA subjects show that the concomitant presence of GAD65-M and GAD65-C Ab (GAD65-M CAb) is a feature also of adultonset autoimmune diabetes. In addition, detection of GAD65-M C Ab in samples collected years before the development of insulin dependency supports the hypothesis that this immune marker can predict insulin requirement. GAD65-MAb only has been found in 7% of type 1 diabetic children (32), a frequency similar to that now documented in LADA subjects in insulin therapy (8%). In addition, we identified a group of subjects positive only for GAD65-MAb in treatment with oral hypoglycemic agents and/or diet. At this stage, it is not yet clear whether middle antibody epitopes associated with childhood diabetes are identical to those observed in adult patients. Nevertheless, our data are consistent with the existence of a subgroup of GAD65Ab-positive adult-onset diabetic subjects with phenotypic characteristics similar to those of the typical type 2, GAD65Abnegative, subjects. This subgroup is identified by the exclusive presence of GAD65Ab binding to the middle region of the autoantigen (GAD65-MAb only). Several hypotheses can be formulated to explain the existence of two subgroups of GAD65Ab-positive subjects. The presence of GAD65-MAb, in the absence of GAD65-CAb, might be the result of a different sensitivity of the two epitope assays. However, it must be noted that, in the previous study on 155 type 1 diabetic children (32) that used the same chimeric molecules, no difference in diagnostic sensitivity was observed between the two epitope assays. In our present study, the presence of GAD65-MAb only could not be explained simply by a low GAD65 index and by a higher sensitivity of this assay, compared with that for GAD65-CAb. In fact, when we analyzed adult subjects with low levels of GAD65Ab (to standardize for the sensitivity of the GAD65- MAb assay), the presence of GAD65-CAb could be detected only in insulin-treated subjects but not in patients in treatment with oral hypoglycemic agents and/or diet. Thus, the observed differences in GAD65Ab epitope selection are more likely due to a biological phenomenon than to a methodological artifact. At the present stage, we cannot rule out the possibility that the selective production of GAD65-MAb represents an early phase of the GAD65-specific autoimmune process. Future prospective studies will address this specific question. However, several lines of evidence tend to favor the hypothesis that GAD65-MAb identifies a distinct subgroup of patients

GAD65Ab EPITOPES IN LADA 315 with low risk for insulin requirement. Firstly, disease duration was similar in the two subgroups of GAD65Ab-positive subjects (GAD65-MAb only and GAD65-M CAb). Second, we did not observe any switch in epitope specificity in our limited follow-up samples. None of the 5 subjects positive for GAD65-MAb only, of whom follow-up serum samples were available, became positive for GAD65-CAb during a follow-up period of 10 21 months. Third, and more important, we observed the exclusive presence of GAD65-MAb in two subjects still in treatment with oral hypoglycemic agents after 22 yr and 46 yr of disease duration, respectively. Finally, the prevalence of thyroid autoantibodies was significantly increased only in patients positive for GAD65 M CAb, but not in those positive for GAD65-MAb only, as compared to GAD65Ab-negative subjects. This supports the hypothesis that only patients positive for GAD65-M CAb have an increased risk for endocrine autoimmunity. If the selective presence of GAD65-MAb only is not directly related to an islet autoimmune process, it may be the result of a crossreactivity with other, unrelated antigens. In conclusion, our study demonstrates that, in adult subjects diagnosed with type 2 diabetes, the presence of GAD65Ab is associated with characteristics typical of a slowly progressive type 1 diabetes with high probability of need of exogenous insulin. Based on epitope specificity, two subgroups of subjects with GAD65Ab have been identified with different phenotypic characteristics and different risk for insulin requirement. The use of epitope-specific assays can improve the diagnostic specificity of GAD65Ab for insulin requirement. Acknowledgments We thank Martina Enggrön for editorial assistance, Dr. Å. Lernmark (University of Washington, Seattle, WA) for providing the human GAD65 cdna and the R7309 and R10266 rabbit antisera, Dr. B. Michelsen (Hagedorn Research Institute, Gentofte, Denmark) for providing the rat GAD67 cdna, and Dr. P. De Camilli (Yale University, New Haven, CT) for providing the DNA of the GAD65-N-67 chimera. References 1. Lernmark Å, Falorni A. 1997 Immunology of insulin-dependent diabetes mellitus. In: Pickup J, Williams G, eds. Textbook of diabetes. Oxford, UK: Blackwell Science Ltd.; 15.1 15.23. 2. Palmer JP, Asplin CM, Clemons P, et al. 1983 Insulin antibodies in insulindependent diabetics before insulin treatment. Science. 222:1337 1339. 3. Sanjeevi CB, Falorni A, Robertson J, Lernmark Å. 1996 Glutamic acid decarboxylase (GAD) in insulin-dependent diabetes mellitus. Diab Nutr Metab. 9:167 182. 4. Bonifacio E, Christie MR. 1996 Tyrosine phosphatase-like proteins as autoantigens in insulin-dependent diabetes mellitus: the targets for 37/40k antibodies. Diab Nutr Metab. 9:183 187. 5. Atkinson MA, Maclaren NK. 1993 Islet cell autoantigens in insulin-dependent diabetes. J Clin Invest. 92:1608 1616. 6. Kaufman DL, Clare-Salzler M, Tian J, et al. 1993 Spontaneous loss of T-cell tolerance to glutamic acid decarboxylase in murine insulin-dependent diabetes. Nature. 366:69 72. 7. Tisch R, Yang X-D, Singer SM, et al. 1993 Immune response to glutamic acid decarboxylase correlates with insulitis in non-obese diabetic mice. Nature. 366:72 75. 8. Won Yoon J, Yoon C-S, Lim HW, et al. 1999 Control of autoimmune diabetes in NOD mice by GAD expression or suppression in cells. Science. 284:1183 1187. 9. Vardi P, Ziegler AG, Matthews JH, Dib S, Keller RJ, et al. Concentration of insulin autoantibodies at onset of type I diabetes: inverse log-linear correlation with age. Diabetes Care. 9:736 739. 10. Seissler J, de Sonnaville JJ, Morgenthaler NG, et al. 1998 Immunological heterogeneity in type I diabetes: presence of distinct autoantibody patterns in patients with acute onset and slowly progressive disease. Diabetologia. 41:891 897. 11. Vandewalle CL, Falorni A, Svanholm S, et al. 1995 High diagnostic sensitivity of glutamate decarboxylase autoantibodies in insulin-dependent diabetes mellitus with clinical onset between age 20 and 40 years. J Clin Endocrinol Metab. 80:846 851. 12. Hagopian WA, Karlsen AE, Gottsaeter A, et al. 1993 Quantitative assay using recombinant human islet glutamic acid decarboxylase (GAD65) shows that 64K autoantibody positivity at onset predicts diabetes type. J Clin Invest. 91:368 374. 13. Tuomi T, Groop LC, Zimmet PZ, Rowley MJ, Knowles W, Mackay IR. 1993 Antibodies to glutamic acid decarboxylase reveal latent autoimmune diabetes mellitus in adults with a non-insulin-dependent onset of disease. Diabetes. 42:359 362. 14. Kobayashi T, Itoh T, Kosaka K, Sato K, Tsuji K. 1987 Time course of islet cell antibodies and -cell function in non-insulin-dependent stage of Type I diabetes. Diabetes. 36:510 517. 15. Kobayashi T, Kajio H, Tamemoto K, et al. 1993 Immunogenetic and clinical characterization of slowly progressive IDDM. Diabetes Care. 16:780 788. 16. Zimmet PZ, Tuomi T, Mackay IR, et al. 1994 Latent autoimmune diabetes mellitus in adults (LADA): the role of antibodies to glutamic acid decarboxylase in diagnosis and prediction of insulin dependency. Diabet Med. 11:299 303. 17. Gottsaeter A, Landin-Olsson M, Lernmark Å, Fernlund P, Sundkvist G, Hagopian WA. 1995 Glutamic acid decarboxylase antibody levels predict rate of B-cell decline in adult-onset diabetes. Diabetes Res Clin Pract. 27:133 140. 18. Niskanen LK, Tuomi T, Karjalainen J, Groop LC, Uusitupa MIJ. 1995 GAD antibodies in NIDDM. Diabetes Care. 18:1557 1565. 19. Turner R, Stratton I, Horton V, et al. 1997 UKPDS 25:autoantibodies to islet-cell cytoplasm and glutamic acid decarboxylase for prediction of insulin requirement in type 2 diabetes. Lancet. 350:1288 1293. 20. Tuomi T, Carlsson A, Li H, et al. 1999 Clinical and genetic characteristics of type 2 diabetes with and without GAD antibodies. Diabetes. 48:150 157. 21. Littorin B, Sundkvist G, Hagopian W, et al. 1999 Islet cell and glutamic acid decarboxylase antibodies present at diagnosis of diabetes predict the need for insulin treatment. A cohort study in young adults whose disease was initially labeled as type 2 or unclassifiable diabetes. Diabetes Care. 22:409 412. 22. Verge CF, Gianani R, Kawasaki E, Yu L, et al. Prediction of type I diabetes in first-degree relatives using a combination of insulin, GAD and ICA512bdc/ IA-2 autoantibodies. Diabetes. 45:926 933. 23. Falorni A, Kassi G, Murdolo G, Calcinaro F. 1998 Controversies on humoral immune markers of insulin-dependent diabetes mellitus. J Ped Endocrinol Metab 11 (Suppl 2):307 317. 24. Solimena M, Folli F, Aparisi R, Pozza G, De Camilli P. 1990 Autoantibodies to GABA-ergic neurons and pancreatic cells in stiff-man syndrome. N Engl J Med. 322:1555 1560. 25. Tuomi T, Björses P, Falorni A, et al. 1996 Antibodies to glutamic acid decarboxylase and insulin-dependent diabetes in patients with APECED (autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy). J Clin Endocrinol Metab. 81:1488 1494. 26. Hallengren B, Falorni A, Landin-Olsson M, Lernmark Å, Papadopoulos KI, Sundkvist G. 1996 Islet cell and glutamic acid decarboxylase antibodies in hyperthyroid patients at diagnosis and following treatment. J Intern Med. 239:63 68. 27. Richter W, Jury KM, Loeffler D, Manfras BJ, Eiermann TH, Boehm BO. 1995 Immunoglobulin variable gene analysis of human autoantibodies reveals antigen-driven immune response to glutamate decarboxylase in type 1 diabetes mellitus. Eur J Immunol. 25:1703 1712. 28. Richter W, Shi Y, Baekkeskov S. 1993 Autoreactive epitopes defined by diabetes-associated human monoclonal antibodies are localized in the middle and C-terminal domains of the smaller form of glutamate decarboxylase. Proc Natl Acad Sci USA. 90:2832 2836. 29. Ujihara N, Daw K, Gianani R, Boel E, Yu L, Powers AC. 1994 Identification of glutamic acid decarboxylase autoantibody heterogeneity and epitope regions in type I diabetes. Diabetes. 43:968 975. 30. Daw K, Powers A. 1995 Two distinct glutamic acid decarboxylase autoantibody specificities in IDDM target different epitopes. Diabetes. 44:216 220. 31. Daw K, Ujihara N, Powers MA. 1996 Glutamic acid decarboxylase autoantibodies in Stiff-Man Syndrome and insulin-dependent diabetes mellitus exhibit similarities and differences in epitope recognition. J Immunol. 156:818 825. 32. Falorni A, Ackefors M, Carlberg C, et al. 1996 Diagnostic sensitivity of immunodominant epitopes of glutamic acid decarboxylase (GAD65) autoantibodies in childhood IDDM. Diabetologia. 39:1091 1098. 33. Björk E, Velloso LA, Kämpe O, Karlsson FA. 1994 GAD autoantibodies in IDDM, Stiff-Man Syndrome and autoimmune polyendocrine syndrome type I recognize different epitopes. Diabetes. 43:161 165. 34. Butler MH, Solimena M, Dirkx JR, Hayday A, De Camilli P. 1993 Identification of a dominant epitope of glutamic acid decarboxylase (GAD65) recognized by autoantibodies in Stiff-Man Syndrome. J Exp Med. 178:2097 2106. 35. Ziegler B, Strebelow M, Rjasanowski I, Schlosser M, Ziegler M. 1998 A monoclonal antibody-based characterization of autoantibodies against glu-

316 FALORNI ET AL. JCE&M 2000 Vol 85 No 1 tamic acid decarboxylase in adults with latent autoimmune diabetes. Autoimmunity. 28:61 68. 36. Falorni A, Grubin CE, Takei I, et al. 1994 Radioimmunoassay detects the frequent occurrence of autoantibodies to the Mr 65,000 isoform of glutamic acid decarboxylase in Japanese insulin-dependent diabetes. Autoimmunity. 19:113 125. 37. Falorni A, Örtqvist E, Persson B, Lernmark Å. 1995 Radioimmunoassays for glutamic acid decarboxylase (GAD65) and GAD65 autoantibodies using 35 Sor 3 H recombinant human ligands. J Immunol Methods. 186:89 99. 38. Karlsen AE, Hagopian WA, Grubin CE, et al. 1991 Cloning and primary structure of a human islet isoform of glutamic acid decarboxylase from chromosome 10. Proc Natl Acad Sci USA. 88:8337 8341. 39. Grubin CE Daniels T Toivola B, et al. 1994 A novel radioligand binding assay to determine diagnostic accuracy of isoform-specific glutamic acid decarboxylase antibodies in childhood IDDM. Diabetologia. 37:344 350. 40. Schmidli RS, Colman PG, Bonifacio E, Bottazzo GF, Harrison LC, participating laboratories. 1994 High levels of concordance between assays for glutamic acid decarboxylase autoantibodies: the first international glutamic acid decarboxylase antibody workshop. Diabetes. 43:1005 1009. 41. Schmidli RS, Colman PG, Bonifacio E, participating laboratories. 1995 Disease sensitivity and specificity of 52 assays for glutamic acid decarboxylase antibodies. The second international GADAb workshop. Diabetes. 44:636 640. 42. Verge CF, Stenger D, Bonifacio E, et al. 1998 Combined use of autoantibodies (IA-2 autoantibody, GAD autoantibody, insulin autoantibody, cytoplasmic islet cell antibodies) in type 1 diabetes: Combinatorial Islet Autoantibody Workshop. Diabetes. 47:1857 1866. 43. Michelsen B, Petersen JS, Boel E, Møldrup A, Dyrberg T, Madsen OD. 1991 Cloning, characterization and autoimmune recognition of rat islet glutamic acid decarboxylase in insulin-dependent diabetes mellitus (IDDM). Proc Natl Acad Sci USA. 88:8754 8758. 44. Solimena M, Aggujaro D, Muntzel C, et al. 1993 Association of GAD-65, but not of GAD-67, with the Golgi complex of transfected Chinese hamster ovary cells mediated by the N-terminal region. Proc Natl Acad Sci USA. 90:3073 3077. 45. Li L, Jiang J, Hagopian WA, et al. 1995 Differential detection of rat islet and brain glutamic acid decarboxylase (GAD) isoforms with sequence-specific peptide antibodies. J Histochem Cytochem. 43:53 59.