IN TYPE 1, insulin-dependent diabetes insulin deficiency

0013-7227/02/$15.00/0 The Journal of Clinical Endocrinology & Metabolism 87(5):2254 2261 Printed in U.S.A. Copyright 2002 by The Endocrine Society The Karlsburg Type 1 Diabetes Risk Study of a Normal Schoolchild Population: Association of -Cell Autoantibodies and Human Leukocyte Antigen-DQB1 Alleles in Antibody-Positive Individuals MICHAEL SCHLOSSER, MARTINA STREBELOW, RALF WASSMUTH, MARIE-LUISE ARNOLD, ISABEL BREUNIG, ILONA RJASANOWSKI, BRIGITTE ZIEGLER, AND MANFRED ZIEGLER Institute of Pathophysiology (M.Sc., M.St., B.Z., M.Z.), Ernst Moritz Arndt University of Greifswald, Karlsburg D-17495, Germany; Surgery Hospital (M.Sc.), Ernst Moritz Arndt University of Greifswald, Greifswald D-17487, Germany; Institute for Clinical Immunology (R.W., M.-L.A., I.B.), Friedrich Alexander University Erlangen-Nürnberg, Erlangen D-91054, Germany; and Centre of Diabetes and Metabolic Disorders (I.R.), Karlsburg D-17495, Germany The intent of this study was to analyze the prevalence of diabetes-associated autoantibodies (AAbs) at or above the 99 th percentile as well as their association with human leukocyte antigen (HLA)-DQB1 alleles in a normal population of 6,337 schoolchildren. AAbs against glutamic acid decarboxylase (GADA), tyrosine phosphatase IA-2 (IA-2A), and/or insulin (IAA) were detected by 125 I-antigen binding and islet cell antibodies (ICA) immunohistochemically in 181 (2.86%) schoolchildren. HLA-DQB1 alleles were analyzed in 178/181 children and subsequently compared with 119 controls. 2.37% (150/ 6,337) possessed only one AAb, whereas 0.49% (31/6,337) had multiple AAbs but at increased levels (P < 0.001). Subjects with GADA, IA-2A, or IAA revealed an increased frequency of IN TYPE 1, insulin-dependent diabetes insulin deficiency occurs as the consequence of an immune-mediated specific destruction of insulin-producing islet cells in genetically predisposed subjects (1). Type 1 diabetes is most often associated with autoantibodies (AAbs) against cell antigens (2) as AAbs to the molecularly defined cell antigens glutamic acid decarboxylase (GADA), protein tyrosine phosphatase (IA-2A), and insulin (IAA), as well as with the heterogeneous islet cell cytoplasmic antibodies (ICA) including GADA and IA-2A, AAbs to gangliosides and to unknown islet antigens (1). Although type 1 diabetes is considered to be a T cell-mediated autoimmune disease, AAbs to cell antigens are at present the most useful markers of cell destruction. AAbs precede the clinical manifestation of diabetes up to 10 yr prior (3, 4). At onset, up to 98% of type 1 diabetic patients have one or more cell AAbs (5, 6). These AAbs (especially AAb combinations) are highly predictive diagnostic markers for developing the disease in both first-degree relatives of diabetic patients (7), and apparently also within the normal population which actually makes up 90% of patients with type 1 diabetes (6, 8). In recent years, several studies have been conducted to identify subjects at risk of developing type 1 Abbreviations: AAb, Autoantibody; GADA, glutamic acid decarboxylase autoantibody; HLA, human leukocyte antigen; IA-2A, IA-2 autoantibody; IAA, insulin autoantibody; ICA, islet cell antibody; JDF, Juvenile Diabetes Foundation. the diabetes-associated HLA-DQB1 alleles *0302 and/or *02 (P 0.001 0.006) as well as a decreased frequency in the protective allele *0602 (P < 0.001 0.022). DQB1*0602 was completely absent within children with multiple AAbs or with GADA, IA2-A, or IAA at or above the 99.9 th percentile. In comparison to children with single AAbs, the frequency of associated/protective alleles of children with multiple AAbs was enhanced/diminished (P 0.004 0.009). The study shows that also in the general population the multiple AAbs or high level single AAbs predict rather certainly a HLA-DQB1-mediated diabetes susceptibility as shown for first degree relatives of type 1 diabetic patients. (J Clin Endocrinol Metab 87: 2254 2261, 2002) diabetes by screening for ICA and AAbs against the biochemically defined cell antigens as well as by human leukocyte antigen (HLA) typing. Follow-up studies in firstdegree relatives have repeatedly shown that the risk of developing type 1 diabetes increases as the number of autoantibodies rises (7, 8). Screening, however, was, for the most part, not carried out using all four available AAb assays and follow-up periods within the general population study group were relatively short. Therefore, the number of individuals at risk of progressing to the point of a diabetic status during the follow-up was still too low to accurately determine the predictive value of ICA and other markers (9 14). Though the complex association and linkage of HLA class II antigens with type 1 diabetes are not yet fully understood, HLA-DRB1 and -DQA1/-DQB1 alleles provide the strongest genetic contribution to the disease (15), accounting for approximately 50% of the genetic risk in type 1 diabetes (16). In Caucasians, DQB1*0302 and DQB1*02 and their linked DR specificities DR4 and DR3 provide disease susceptibility, particularly in heterozygous combination (17, 18). Moreover, various DR4 allelic variants modify DQB1-associated susceptibility to type 1 diabetes (19, 20), whereas dominant protection is conveyed by DQB1*0602, linked to DR2 (21 23). The relationship between HLA markers and the occurrence of AAbs has been previously examined in recent-onset patients as well as in siblings of type 1 diabetic patients indicating positive correlations between the occurrence of 2254

Schlosser et al. Karlsburg Type 1 Diabetes Risk Study J Clin Endocrinol Metab, May 2002, 87(5):2254 2261 2255 diabetes-associated AAbs and high risk HLA markers (23 28). It remains, however, to be shown whether these findings from type 1 patients and their AAb positive relatives can be correlated to AAb positive individuals lacking a family history of type 1 diabetes coming from the general population. Primary screening of 9,419 normal schoolchildren was carried out for the Karlsburg type 1 diabetes risk study. Using capillary blood serum samples, frequencies and levels were screened for the four major type 1 diabetes-associated AAbs, three directed to the well-defined antigens GAD, IA-2, insulin with cut-off at the 98 th percentile as well as ICA with cut-off at 10 Juvenile Diabetes Foundation (JDF) units. Their interaction was also scrutinized as well as the influence of age and sex (6). The purpose of this study was to analyze the occurrence and interactions of AAbs of higher specificity (cut-off at 99 th percentile for GADA, IA-2A, IAA, and at 20 JDF units for ICA) as well as their association with the HLA-DQB1- dependent diabetes susceptibility in schoolchildren taking part in the follow-up study and having retested AAb positive. Subjects and Methods As previously published (6), 9,419 children (4,722 femal/4,697 male) aged 6 17 yr (mean age 11 3, median 11) representing a normal cross population in Mecklenburg-Vorpommern, Germany, were randomly selected and their capillary blood serum samples were primarily screened for the type 1 diabetes-related AAbs GADA, IA-2A, IAA, as well as ICA. A questionnaire was filled out by each child pertaining to a family history of diabetes (type 1 and type 2) as well as thyroid diseases. The cut-off limits for AAb positivity were defined by a control group of 6,877 schoolchildren lacking a family history of either type of diabetes, autoimmune thyroid diseases, or multiple diabetes-associated AAbs. A total of 764 of the 9,419 initially screened children possessed at least one AAb at levels 98 th percentile and/or ICA 10 JDF units (6). A total of 514 of these children returned for follow up reexamination, representing 6,337 probands (9,419 514/764 6,337) of a normal cross-population. Venous blood samples were taken for AAb retesting, and HLA typing was included in reexamination 24 18 wk post primary AAb testing. The present study serum levels of GADA, IA-2A, and IAA at or above the 99 th percentile and of ICA at or above 20 JDF units were defined to be AAb positive. Upon retesting, 34.0% (175/514) of the primarily AAb positive children revealed AAb levels 98 th percentile and/or ICA 5 JDF units. 30.7% (158/514) had AAb levels 98 th and 99 th percentile and/or ICA 5 10 JDF units and were considered as AAb negative for the present study. 35.2% (181/514) had AAb levels 99 th percentile and/or ICA 20 JDF units. HLA-DQB1 genotyping could be performed successfully in 178 (98 female/80 male) of the 181 children with AAb levels 99 th percentile and/or ICA 20 JDF units as well as in 119 healthy controls. The study protocol was authorized by the Ministry of Culture and Education of Mecklenburg-Vorpommern and approved by the ethics committee of the Ernst-Moritz-Arndt-University (Greifswald, Germany). Informed consent was obtained from the parents or legal guardians. Autoantibody assays The autoantibody assays used have been previously described in detail (6). Briefly, GADA and IA-2A were measured by fluid-phase 125 I-antigen binding assays by the use of recombinant human GAD65 (Diamyd Diagnostics AB, Stockholm, Sweden) and recombinant human IA-2ic (BRAHMS Diagnostica GmbH, Berlin, Germany). GADA and IA-2A levels were expressed as arbitrary Karlsburg units (KU/liter) derived from an in-house standard serum pool. The anti-gad assay has a diagnostic sensitivity and specificity of 88% and 96%, respectively. The anti-ia-2 assay achieved 58% diagnostic sensitivity and 100% specificity, using for both cut-off at or above the 98 th percentile of our control group (Tab. 1), in the 1 st Diabetes Antibody Standardization Program proficiency evaluation 2001 of the Immunology of Diabetes Society and the Centers for Disease Control and Prevention. IAAs were determined by a competitive fluid-phase antigen binding assay using A14 mono- 125 I-insulin (kindly donated by Aventis Pharma Deutschland GmbH, Frankfurt a.m., Germany) with and without the addition of unlabeled insulin. The specific binding of 125 I-insulin following precipitation of immunoglobulins with polyethylene glycol was calculated by subtracting the counts in the presence of cold insulin from the counts in the absence of cold insulin (6). This assay achieved an analytical sensitivity and specificity of 100% for both in the 6th IAA Proficiency Test 1997. However, because of the assay improvement by the microassay according to Williams et al. (29) using protein A, all AAb positive sera were retested by this method. The 99 th percentile as cut off was calculated from 991 (537 male/454 female) healthy schoolchildren. The IAA levels obtained with both methods show a significant correlation (Spearmen s nonparametric correlation analysis r s 0.653, P 0.001). Serum concentration of GADA and IA-2A calculated via in-house standard serum is given as Karlsburg units per liter (KU/liter) and IAA analyzed by microassay (29) is given in U/liter (Table 1). ICA were measured by indirect immunofluorescence on cryosections of human pancreas post overnight incubation at 4 C. The detection limit was 5 JDF units. ICA levels at or above 20 JDF units were considered to be positive. The assay achieved an analytical sensitivity and specificity of 100% for both in the 13th ICA Workshop 1998. To distinguish between single or multiple AAb positivity in sera with GADA and ICA only and IA-2A and ICA only, 20 l of serum was preincubated with 10 g human recombinant GAD65 (Diamyd Diagnostics AB, Stockholm, Sweden) or IA-2Aic (kindly provided by Prof. J. F. Elliot, M.D., Ph.D., University of Alberta, Edmonton, Canada) overnight at 4 C, centrifuged at 100,000 g for 15 min followed by testing for non-gad or non-ia-2 ICA on cryosections and for 125 I-antigen binding. HLA-DQB1 typing Genomic DNA was extracted from whole peripheral blood by standard procedures employing a salting-out technique. HLA-DQB1 genotyping was performed by nonradioactive (digoxigenin-2,3 -dideoxyuridine-5 -triphosphate) oligonucleotide hybridization of enzymatically amplified DNA as described previously (30). Because DQB1*0201 and DQB1*0202 differ by a dimorphism at codon 135, located in the third exon and is thus not differentiated by regular DQB1 genotyping, DQ2- positive individuals are given as DQB1*02 (31). The genotyped schoolchildren were stratified into three groups based upon the occurrence of their HLA-DQB1 genotypes: firstly, a group with at least one diabetesassociated allele (*0302 and/or *02, except *0602); secondly, children with the dominant protective allele *0602 (homozygosity or heterozygosity for *0602); and thirdly a group carrying neither associated nor protective HLA-DQB1 alleles (neutral; other). The 178 genotyped children with AAb positive sera were analyzed in comparison to the AAb negative control group of 119 schoolchildren. Statistical analysis Distribution levels of GADA, IA-2A, IAA, and ICA were tested using the Kolmogorov-Smirnow goodness of fit test. Mann Whitney U test was used to analyze skewed distributions of AAb levels in different groups investigated. The relation between different AAb specificities was analyzed by the two-sided Spearmen s nonparametric correlation analysis (r S ). The level of significance between individual groups studied was assessed by explorative two-sided 2 statistics with Yates correction or Fisher s exact test were appropriate. All the statistical analysis and TABLE 1. Percentiles calculated from 125 I-antigen binding of the 6,877 control children and corresponding AAb concentrations for GADA and IA-2A (KU/liter) and for IAA ( U/liter) Percentile GADA (KU/liter) IA-2A (KU/liter) IAA ( U/liter) 98.0 th 2.14 1.4 175.2 99.0 th 3.9 1.85 281.48 99.9 th 22.6 4.2 1598.95

2256 J Clin Endocrinol Metab, May 2002, 87(5):2254 2261 Schlosser et al. Karlsburg Type 1 Diabetes Risk Study calculation of percentiles were performed using the Statistical Package for Social Sciences, version 10.0 (SPSS, Inc., Chicago, IL). A two-tailed P value of 0.05 or less was considered to indicate statistical significance. Results Prevalence and levels of type 1 diabetes-associated AAbs Through follow-up reexamination of 514 schoolchildren who primary tested AAb positive at or above the 98 th percentile and/or ICA at or above 10 JDF units, 2.86% (181/ 6,337) schoolchildren had GADA, IA-2A, IAA at or above the 99 th percentile and/or ICA levels at or above 20 JDF units. 1.5% (95/6,337) of the healthy children had GADA (median level of 8.15 KU/liter, interquartile range 4.96 23.87 KU/ liter), 0.91% (n 58) had IA-2A (median level of 3.22 KU/ liter, range 2.7 11.55 KU/liter), 0.36% (n 23) had IAA (median level of 761.83 U/liter, range 355-1101.13 U/ liter), and 0.95% (n 60) were tested positive for ICA (median level of 40 JDF units, range 20 87 JDF units). The frequencies of GADA were significantly enhanced compared with IA-2A ( 2 8.57, P 0.003), IAA ( 2 43.12, P 0.001), and ICA ( 2 7.55, P 0.006), but also IA-2A and ICA occur significantly more frequent than IAA ( 2 14.36, P 0.001 and 2 15.72, P 0.001, respectively), whereas IA-2A and ICA frequencies did not differ. A total of 2.37% (n 150) had only one AAb and 0.49% (n 31) were tested positive for two or more AAbs. Among children with a single AAb, 1.06% (67/6,337) had GADA, 0.55% (n 35) had IA-2A, 0.2% (n 13) had IAA, and 0.55% (n 35) had ICA. Also among the single AAbs, the frequency of GADA were significantly enhanced compared with IA-2A ( 2 9.5, P 0.002), IAA ( 2 35.34, P 0.001) and ICA ( 2 9.5, P 0.002), but also IA-2A and ICA occur significantly frequent than IAA ( 2 9.22, P 0.002). Among the 31 children with multiple AAbs, 0.44% (28/ 6,337) had GADA, 0.36% (n 23) had IA-2A, 0.38% (n 24) were ICA positive and only 0.16% (n 10) had IAA. All children with multiple AAbs had GADA and/or IA-2A. The frequencies of GADA, IA-2A, and ICA did not differ within this group, but in comparison to IAA, all three occur significantly more frequent ( 2 7.63, P 0.006 for GADA; 2 4.98, P 0.026 for IA-2A; 2 4.37, P 0.036 for ICA). 0.24% (n 15) of the schoolchildren had two, 0.14% (n 9) had three and 0.11% (n 7) were positive tested by all four AAb assays. Eight probands tested positive for three AAb assays had GADA IA-2A ICA, whereas only one had GADA IAA ICA. Combinations of two AAbs occurred for GADA IA-2A (n 5), GADA ICA (n 7), GADA IAA (n 1), IA-2A IAA (n 1), IA-2A ICA (n 2). Following preincubation of sera positive solely for GADA ICA or IA-2A ICA with the human recombinant antigens GAD65 or IA-2Aic respectively, their 125 I-antigen binding was completely inhibited, except for one sera with 60% inhibition of 125 I-GAD binding. The binding as ICA, however, on cryosections was inhibited in only one of the seven GAD ICA positive sera; thus, this proband was listed within the group positive for GADA only. The sera with GADA ICA were therefore reduced to six probands. In four sera, the end-point titer of ICA was diminished by one dilution step. Within the remaining sera, the titer was unaffected. The levels of GADA, IA-2A, IAA, and ICA were significantly higher (P 0.001) in children with multiple AAbs in comparison with those of children only positive for a single AAb (Table 2). Correlation between levels of the different AAbs The AAb levels of GADA positive children (n 95), of IA-2A positive children (n 58) and of ICA positive children (n 60) were significantly correlated (P 0.001) with the levels of the three remaining AAbs, respectively. In contrast, levels of IAA levels in the IAA positive children (n 23) did not correlate with that of any of the remaining AAbs (Table 3). In sera with multiple AAbs (n 31) a significantly positive correlation was found for GADA and ICA (r s 0.585, P 0.001) and for IA-2A, IAA and ICA (r s 0.395, P 0.028; data not shown). Frequencies of type 1 diabetes-associated HLA-DQB1 alleles in AAb positive children HLA-DQB1 alleles were determined in 178/181 of the AAb positive children. AAb positive probands were stratified according to their level of each AAb ( 99 th and 99.9 th percentile for GADA, IA-2A, IAA, and ICA 20 and 160 JDF units) and the occurrence of the diabetes-associated (*0302 and/or *02, *0602 excluded), dominant protective (homozygosity or heterozygosity for *0602), and other HLA- DQB1 alleles (neutral) and compared with the AAb negative control group (n 119; Table 4). DQB1*0302 and/or *02 was seen in 55 of 119 AAb negative children but 32/119 had at least one protective allele. As shown in Table 4, the frequency of the diabetes-associated HLA-DQB1 alleles was significantly increased in children with any AAb at or above the 99 th percentile (n 178; P 0.021) as well as in children with GADA (n 93; P 0.001) or IA-2A (n 57; P 0.005) or IAA (n 23; P 0.006), but not within the ICA positive group (n 60). The strongest associations were seen in probands with GADA (n 24; P 0.001) or IA-2A (n 19; P 0.001) or any one AAb (n 37; P 0.001) at or above the 99.9 th percentile. Further significantly enhanced frequencies of the diabetes-associated alleles were detected in children with ICA titer of 160 JDF units or higher (n 14; P TABLE 2. Comparison of AAb levels of children positive for only one AAb vs. children with multiple AAbs AAb 1 AAb positive (n 150) 1 AAb positive (n 31) P value GADA (KU/liter) 1.48 (0.23 6.18) 26.7 (6.35 146.8) 0.001 IA-2A (KU/liter) 0.28 (0.27 1.28) 6.34 (0.6 43.8) 0.001 IAA ( U/liter) 9.8 ( 24.08 40.75) 118.79 (35.45 354.89) 0.001 ICA (JDF units) 0 (0 10) 40 (20 160) 0.001 Data are median AAb level (interquartile range); Mann-Whitney U test.

Schlosser et al. Karlsburg Type 1 Diabetes Risk Study J Clin Endocrinol Metab, May 2002, 87(5):2254 2261 2257 TABLE 3. Correlation between the levels determined by the four AAb assays and those of the three co-occurring AAbs Subjects positive for n GADA IA-2A IAA ICA GADA 95 0.432 ( 0.001) 0.434 ( 0.001) 0.655 ( 0.001) IA-2A 58 0.664 ( 0.001) 0.463 ( 0.001) 0.628 ( 0.001) IAA 23 0.220 (0.313) 0.209 (0.339) 0.021 (0.925) ICA 60 0.420 (0.001) 0.475 ( 0.001) 0.659 ( 0.001) Data are coefficient of Spearman s nonparametric correlation analysis r 2 (P values). TABLE 4. Frequencies of type 1 diabetes-associated HLA-DQB1 alleles (*0302 and *02) and the protective allele (*0602) of AAb positive children (at or above the 99 th percentile for GADA, IA-2A, and IAA; and at or above 20 JDF units for ICA) compared to those of AAb negative control subjects dependent on AAb level, their specificity, and their occurrence as one of multiple (upper part) or single AAbs (lower part) Children s sera with AAb level at or above percentile/ JDF units n *0302 and/or *02 *0602 Other n n n 2 (P value) 2 (P value) 2 (P value) Any one AAb 99.0/20 178 108 24 46 2 5.45 (0.021) 2 7.53 (0.006) 2 0 (0.948) 99.9/160 37 31 1 5 2 14.62 ( 0.001) 2 8.5 (0.001) 2 2.1 (0.122) GADA 99.0 93 65 12 16 2 10.97 (0.001) 2 5.39 (0.022) 2 2.27 (0.141) 99.9 24 22 0 2 2 14.82 ( 0.001) 2 6.84 (0.002) 2 2.84 (0.065) IA-2A 99.0 57 40 1 16 2 7.97 (0.005) 2 14.38 ( 0.001) 2 0 (1.0) 99.9 19 17 0 2 2 10.61 ( 0.001) 2 6.23 (0.007) 2 1.56 (0.158) IAA 99.0 23 18 1 4 2 6.69 (0.006) 2 4.3 (0.016) 2 0.49 (0.488) 99.9 3 3 0 0 2 1.58 (0.105) 2 0.14 (0.566) 2 0.14 (0.566) ICA 20 60 35 10 15 2 1.88 (0.177) 2 1.79 (0.186) 2 0.01 (0.931) 160 14 11 1 2 2 4.03 (0.026) 2 1.67 (0.187) 2 0.49 (0.517) Only 1 AAb 99.0/20 147 82 24 41 2 2.04 (0.162) 2 3.8 (0.052) 2 0.002 (0.967) 99.9/160 13 10 1 2 2 3.28 (0.043) 2 1.39 (0.184) 2 0.32 (0.513) Only GADA 99.0 65 40 12 13 2 3.36 (0.071) 2 1.21 (0.276) 2 0.74 (0.391) 99.9 6 6 0 0 2 4.63 (0.012) 2 0.99 (0.337) 2 0.99 (0.337) Only IA-2A 99.0 34 20 1 12 2 1.60 (0.206) 2 7.61 (0.002) 2 1.14 (0.289) 99.9 3 2 0 1 2 0.01 (0.598) 2 0.14 (0.556) 2 0 (1.0) Only IAA 99.0 13 8 1 4 2 0.57 (0.384) 2 1.39 (0.184) 2 0 (0.75) 99.9 1 1 0 0 2 0.01 (0.467) 2 0 (1.0) 2 0 (1.0) Only ICA 20 35 14 10 11 2 0.21 (0.652) 2 0 (0.986) 2 0.1 (0.761) 160 3 1 1 1 2 0 (1.0) 2 0 (1.0) 2 0 (1.0) AAb negative control 119 55 32 32 subjects Data are n, 2, and P values calculated in comparison with AAb negative children. 0.026), and in the group of children with any single AAb (n 13; P 0.043) or with GADA as single AAb (n 6; P 0.012). The frequency of the protective HLA-DQB1 allele *0602 was significantly lower in the entire cohort of AAb positive children (n 178; P 0.006) compared with AAb negative children and was strongly reduced in GADA (n 93; P 0.022) or IA-2A (n 57; P 0.001) or IAA (n 23; P 0.016) positive probands and also in children with IA-2A (n 34; P 0.002) as single AAb at or above the 99 th percentile. Only one IA-2A and one IAA positive proband bears the protective allele *0602, but 10 ICA positive children had *0602. Among children with GADA, IA-2A, or IAA at or above the 99.9 th percentile the protective allele was complete absent independent of their occurrence as single or as one of multiple

2258 J Clin Endocrinol Metab, May 2002, 87(5):2254 2261 Schlosser et al. Karlsburg Type 1 Diabetes Risk Study AAbs (Table 4). However, frequencies of associated or protective HLA-DQB1 alleles of all children with only 1 AAb above the 99 th percentile (n 147) as well as frequencies of the protective allele of all children with single AAbs above the 99.9 th percentile or only ICA 160 JDF units (n 13) did not differ from those of AAb negative children. Moreover, there was no association between the occurrence of ICA or IAA as single AAb and the frequency of diabetes-associated or protective alleles. Among six children with ICA at levels 80 JDF units as single AAb only one child bears the diabetesassociated alleles *0302 or *02, but two carry the protective allele (data not shown). Distribution of HLA-DQB1 alleles in children with multiple AAbs None of the 31 children with multiple AAbs carry the protective allele HLA-DQB1*0602, whereas the frequency of the diabetes-associated alleles *0302 and/or *02 was significantly increased (26/31; P 0.001) in comparison to AAb negative children (Table 5). This significant positive association of *0302 and/or *02 was also seen in all children with 3 or 4 AAbs and the negative association of *0602 was present in 15 children positive for 2 AAbs (P 0.021; Table 5). In comparison to children with single AAbs, the occurrence of *0302 and/or *02 was significantly higher in children with 3 AAbs ( 2 5.12, P 0.011) and 4 AAbs ( 2 3.7, P 0.021; Fig. 1). Altogether, children with multiple AAbs have a significantly increased frequency of the diabetes-associated alleles and a reduced frequency of the protective allele compared with children with single AAbs ( 2 7.33, P 0.004 and 2 4.53, P 0.009, respectively; Fig. 1). Discussion The major goal of all screening strategies is to predict individuals at risk with a high sensitivity and specificity to provide an effective, appropriate intervention in the future. Several studies have been carried out in first degree relatives of patients with type 1 diabetes. It is still open whether these results can be extrapolated to the general population, but the source of 90% of all new cases of the disease are without a family history of type 1 diabetes. Aim of the primary AAb screening (6) was to recognize individuals at risk for type 1 diabetes by choosing a high diagnostic sensitivity through the use of low percentiles to define antibody positivity (at or above the 98 th for GADA, IA-2A, IAA, and at or above 10 JDF units for ICA). Screening evaluation using these antibody cut points results in a high antibody prevalence of 8.1% (764 out of 9,419 children tested by the 4 AAb assays), reflecting a low specificity at this antibody cut point as previously published (6). Of 764 children, 514 took part in follow up examination for venous blood drawing for HLA typing and AAb retesting 24 18 wk later. The evaluation of the AAb retesting was carried out with higher diagnostic specificity by the use of higher percentiles to define antibody positivity (at or above the 99 th for GADA, IA-2A, IAA, and at or above 20 JDF units for ICA) resulting in a lower antibody prevalence of 2.86% (181/6,337). The purpose of this study was to analyze the occurrence and interactions of type 1 diabetes-associated autoantibodies and genetic markers in this general schoolchild population. In accordance to a previous report (32), no AAb fluctua- FIG. 1. Frequency of the diabetes-associated HLA-DQB1 alleles *0302 and/or *02 (filled bars), the protective allele *0602 (hatched bars), and other alleles (open bars) of the 147 probands with single GADA, IA-2A or IAA at levels at or above the 99 th percentile and ICA at or above 20 JDF units compared with subjects positive for two, three, or four AAbs as well as to all multiple AAb positive children. The frequencies of associated/protective alleles of children with single AAbs did not differ from those of 119 AAb negative control subjects. Significant differences were calculated by 2 statistics or Fisher s exact test. *, P 0.05; **, P 0.01. TABLE 5. Frequencies of type 1 diabetes-associated HLA-DQB1 alleles (*0302 and *02) and the protective allele (*0602) of children with multiple AAbs (at or above the 99 th percentile for GADA, IA-2A, and IAA, and at or above 20 JDF units for ICA) dependent on the number of AAbs compared with those of AAb negative control subjects Children s sera with n *0302 and/or *02 *0602 Other n n N 2 (P value) 2 (P value) 2 (P value) 1 AAb 31 26 0 5 2 12.56 ( 0.001) 2 9.05( 0.001) 2 1.01(0.251) 2 AAb 15 10 0 5 2 1.49 (0.174) 2 3.92(0.021) 2 0.05(0.557) 3 AAb 9 9 0 0 2 7.65 (0.003) 2 1.95(0.111) 2 1.95(0.111) 4 AAb 7 7 0 0 2 5.65 (0.006) 2 1.3(0.19) 2 1.3(0.19) AAb negative control 119 55 32 32 subjects Data are n, 2, and P values calculated in comparison with AAb negative children.

Schlosser et al. Karlsburg Type 1 Diabetes Risk Study J Clin Endocrinol Metab, May 2002, 87(5):2254 2261 2259 tions were not observed within all of the children with multiple AAbs and high level single AAbs upon reexamination. Among children with single AAbs at low levels, seroconversion to AAb negativity occurred confirming fluctuation for GADA (33), ICA (34), and IAA (13) in other studies. Follow-up of this study performed at higher specificity beginning from the first reexamination will enable us to investigate the AAb persistence and fluctuation. Occurrence of type 1 diabetes-associated AAbs in healthy schoolchildren A total of 2.86% (n 181) of probands aged 6 17 yr were AAb positive, representing a general population of 6,337 schoolchildren. The majority of them, 2.37% (n 150) had only one AAb, and 0.49% (n 31) were positive for more than one AAb. Among the children with single AAbs, GADA occur significantly more frequently than the other AAbs. In this group the frequency of ICA and IA-2A was enhanced compared with that of IAA. In children with multiple AAbs, the frequency of GADA, IA-2A and ICA did not differ, but all three were significantly more frequent than the IAA. The single AAb positivity for GADA, IA-2A, IAA, and ICA was mostly associated with low AAb levels, whereas multiple AAb positivity implies significantly enhanced AAb levels (P 0.001). This is in accordance with studies demonstrating that most of recent-onset type 1 diabetic patients are positive for multiple AAbs, but the presence of any one single AAb appears not to be of predictive value for the disease (5, 6, 8). Furthermore, a recently published prospective study of 15,224 nondiabetic first-degree relatives of probands with type 1 diabetes demonstrates that the risk of diabetes was negligible in probands with single ICA positivity but increased dramatically when two or more AAbs were present (35). Surprisingly, IAA levels did not correlate with any other AAb investigated and occur with the lowest frequency in both, children with single as well as with multiple AAbs. From our data, it is concluded that insulin as the only known cell-specific autoantigen reveals a specific behavior compared with GAD and IA-2. As demonstrated by the German BABYDIAB-Study, IAA most frequently occurs early in life and precedes other AAbs in probands with multiple AAbs (36). Furthermore, a decline of IAA later in life was frequently seen (36), which perhaps explains the low frequency of IAA and failure in correlation to any other AAb in our schoolchild population aged 6 17 yr. The significant correlations observed between AAb levels of GADA or IA-2A and those of ICA demonstrate that both contribute to ICA binding as detected by monoclonal antibodies (37, 38). These various patterns of occurrence and correlations of GADA and IA-2A are in accordance with their higher diagnostic sensitivity and specificity for type 1 diabetes compared with IAA (39). If AAb testing would be performed without measuring ICA, 25.8% (8/31) of the high risk subjects with multiple AAbs (6 positive for GADA ICA, 2 positive for IA-2A ICA) would have been missed in this group. As demonstrated by the preincubation of these sera with the recombinant antigens GAD65 and IA-2ic, their ICA binding was not substantially diminished, although their corresponding 125 I-antigen binding was completely blocked. Thus, it is suggested that these eight probands with two AAbs have beside GAD-ICA or IA-2-ICA also nongad/nonia-2 ICA to gangliosides or still unknown antigens. That means these children belong really to the group with multiple AAbs and might be therefore have an increased risk as also shown by the genetic predisposition. From our study, it is suggested that following primary screening performed on GADA and IA-2A testing, AAb positive sera should be tested also for IAA as well as ICA. Thus, none of our probands with multiple AAbs would be overlooked with this strategy. High level IA-2A, GADA, and IAA implies the absence of the protective HLA-DQB1 allele *0602 The prediction of risk for type 1 diabetes is currently based on the occurrence of autoantibodies, genetic markers and metabolic disorders. Despite the strong genetic contribution of HLA class II genes, the positive predictive value of the highest risk genotype DR3/4, DQB1*0201/*0302 amounts to only 6.3% for development of diabetes by the age of 20 (10). Another recent study investigating genetic and humoral markers for prediction of type 1 diabetes in siblings has reported that among the genetic markers DQB1*02/*0302 was associated with the highest positive predictive value of 22% (40). In our general population-based study, DQB1*02/ *0302 heterozygosity occurred in only 14 out of 178 AAb positive subjects (7.9%). Moreover, 8/14 showed only single AAb positivity. This result might be of importance for population screening strategies based on the occurrence of highest risk HLA-DR or -DQB1 genotypes to avoid overlooking future cases. The analysis of a total of 178 genotyped AAb positive subjects at a level at or above the 99 th percentile and/or at or above 20 JDF units for ICA revealed a significantly increased frequency of *0302 and/or *02 (P 0.021) and a decreased frequency of the protective allele *0602 (P 0.006) compared with the 119 controls (Table 4). The associations were further enhanced at or above the 99.9 th percentile and/or at or above 160 JDF units for ICA (n 37) for both the associated and protective alleles (P 0.001). In children with GADA, IA-2A or IAA levels at or above the 99.9 th percentile a immunogenetic predisposition was discernible as demonstrated by the significantly increased frequency of *0302 and/or *02 for GADA and IA-2A (P 0.001) and by the complete absence of the dominant protective allele *0602 for all three AAbs. The comparison of the four groups with only one AAb supports this finding: children with IA-2A, GADA, or IAA as single AAb at high levels did not bear the protective allele. Thus, the occurrence of *0602 is associated with protection from cell autoimmunity measured by GADA, IA-2A, and IAA, at least in the general schoolchildren population investigated. However, this is in contrast to the observation of an association between DQB1*0602 and high titers of GADA in firstdegree relatives of diabetic patients at very low risk of developing type 1 diabetes (41, 42). In our schoolchild population, subjects with only ICA revealed no significant genetic susceptibility supported by DQB1 genotypes. Only one of six children positive for only ICA at or above 80 JDF

2260 J Clin Endocrinol Metab, May 2002, 87(5):2254 2261 Schlosser et al. Karlsburg Type 1 Diabetes Risk Study units is characterized by DQB1*02 homozygosity, whereas two of them carry the protective allele DQB1*0602. Thus, it is expected that this subgroup of ICA positive probands characterized by the absence of GADA and IA-2A may not progress to type 1 diabetes. This is in accordance with Pugliese et al. (23), demonstrating that occurrence of DQB1*0602 is associated with dominant protection against type 1 diabetes in ICA positive first degree relatives during a follow-up period of 12 yr. Possibly, age at onset may be influenced by DQB1*0602, as this protective allele was absent in patients diagnosed before the age of 10, but increased in frequency with increasing age at onset (43). The majority (82.6%) of AAb positive, HLA-typed children demonstrated only single AAbs. Although the frequency of DQB1*0602 was diminished among children with single AAbs compared with the AAb negative children, this difference was not significant. Furthermore, children with single AAbs at or above the 99 th percentile possessed no increased genetic risk compared with AAb negative children. Thus, single AAb positivity might be of minor diagnostic relevance for type 1 diabetes prediction, but the risk, especially of probands with high level single AAbs should be further differentiated by examination of co-occurrence of genetic risk or protective alleles. The frequency of neutral alleles was not significantly different from those of the control subjects in any of the AAbstratified groups. Strong association of multiple AAbs with the occurrence of HLA-DQB1 risk alleles The children with two or more AAbs bear a significantly increased genetic risk for developing of type 1 diabetes compared with both, children with single AAbs (P 0.004) and AAb negative children (P 0.001). This increased risk is mediated 1) by the absence of the protective allele DQB1*0602 in each of these subjects, and 2) by the increased frequency of the diabetes-associated alleles DQB1*0302 and/or *02. Remarkably, all nine subjects with three AAbs and all seven subjects positive for all four AAbs, had, in addition to the absence of the protective allele, at least one diabetes-associated HLA-DQB1 allele. It should be pointed out that no child involved in this study has had an affected first-degree relative. Therefore, the reliability of the risk assessment based on combined AAb detection in a general schoolchild population seems to be comparable to that as described for first degree relatives identifying 68% of probands with progression to clinical diabetes during 5 yr follow-up by detection of more than 1 AAb (35, 44). In agreement with Bingley et al. (8), we found a strong association of estimated risk for developing diabetes with increasing levels of AAbs and multiple AAb positivity. In conclusion, among subjects with multiple AAbs, especially with three or four AAbs, HLA-DQB1 alleles associated with the highest genetic susceptibility for type 1 diabetes, i.e. DQB1*0302 and/or *02 were seen in all cases, whereas the protective allele DQB1*0602 was completely absent. Thus, the occurrence of DQB1*0602 of AAb positive probands in our general schoolchild population is strongly negatively associated with the occurrence of multiple AAbs as well as with high level GADA, IA-2A, or IAA as single AAb and perhaps therefore provides protection from the destructive cell autoimmunity. HLA-DQB1 genotyping is necessary to differentiate individual risk of subjects positive only for one or two AAbs as well as to identify AAb positive probands carrying protective or other DQB1 alleles for further genetic risk stratification. Acknowledgments We would like to thank all the children participating in the Karlsburg type 1 diabetes risk study, the teachers and physicians supporting the study, to Mrs. Sonja Tietz, Heidi Kenk, Rosemarie Jung, and Christiane Lenth for excellent technical assistance. We also thank Prof. Dr. Åke Lernmark for critically reading the manuscript, which was kindly preedited by English native speaker Mrs. Susan Heidecke. Received April 27, 2001. Accepted February 6, 2002. Address all correspondence to: Dr. Manfred Ziegler, Institute of Pathophysiology, Universitiy of Greifswald, Greifswalder Strasse 11b, Karlsburg D-17495, Germany. E-mail: ziegler@mail.uni-greifswald.de. Address requests for reprints to: Dr. Michael Schlosser, Institute of Pathophysiology, University of Greifswald, Greifswalder Strasse 11b, Karlsburg D-17495, Germany. This study was supported by the Federal Ministry of Research and Technology (BMFT; Grant 07NBL02/D4), by the Ministry of Culture and Education of Mecklenburg-Vorpommern (Grant EMAU16/1995), by the Division of Employment (ABM4576/98), by the Community Medicine Project of the Ernst Moritz Arndt University Greifswald, and by the BRAHMS Diagnostica GmbH, Berlin. Additional support was obtained from the BMFT-funded Center for Interdisciplinary Clinical Research at the Friedrich Alexander University Erlangen-Nürnberg (Grant 01KS9601), by the Deutsche Forschungsgemeinschaft (Grant SFB263), and by the Association for Support of Diabetes Research, Karlsburg, Greifswald e.v. References 1. Castano L, Eisenbarth GS 1990 Type-I diabetes: a chronic autoimmune disease of human, mouse, and rat. Annu Rev Immunol 8:647 679 2. Atkinson MA, Maclaren NK 1990 What causes diabetes? Sci Am 7:62 67 3. Bingley PJ, Bonifacio E, Gale EAM 1993 Can we really predict IDDM? Diabetes 42:213 220 4. Lühder F, Schlosser M, Mauch L, Haubruck H, Rjasanowski I, Michaelis D, Kohnert KD, Ziegler M 1994 Autoantibodies against GAD65 rather than GAD67 precede the onset of type 1 diabetes. Autoimmunity 19:71 80 5. Gorus FK, Goubert P, Semakula C, Vandewalle CL, De Schepper J, Scheen A, Christie MR, Pipeleers DG 1997 IA-2-autoantibodies complement GAD65- autoantibodies in new-onset IDDM patients and help predict impending diabetes in their siblings. Diabetologia 40:95 99 6. Strebelow M, Schlosser M, Ziegler B, Rjasanowski I, Ziegler M 1999 Karlsburg type I diabetes risk study of a general population: frequencies and interactions of the four major type I diabetes associated autoantibodies studied in 9419 schoolchildren. Diabetologia 42:661 670 7. Verge CF, Gianani R, Kawasaki E, Yu L, Pietropaolo M, Jackson RA, Chase HP, Eisenbarth GS 1996 Prediction of type I diabetes in first-degree relatives using a combination of insulin, GAD, and ICA512bdc/IA-2 autoantibodies. Diabetes 45:926 933 8. Bingley PJ, Bonifacio E, Williams AJK, Genovese S, Bottazzo GF, Gale AM 1997 Prediction of IDDM in the general population. Strategies based on combinations of autoantibody markers. Diabetes 46:1701 1710 9. Boehm BO, Manfras B, Seissler J, Schoffling K, Gluck M, Holzberger G, Seidl S, Kuhnl P, Trucco M, Scherbaum WA 1991 Epidemiology and immunogenetic background of islet cell antibody-positive nondiabetic schoolchildren. Ulm-Frankfurt Population Study. Diabetes 40:1435 1439 10. Rewers M, Bugawan TL, Norris JM, Blair A, Beaty B, Hoffman M, McDuffie Jr RS, Hamman RF, Klingensmith G, Eisenbarth GS, Erlich HA 1996 Newborn screening for HLA markers associated with IDDM: diabetes autoimmunity study in the young (DAISY). Diabetologia 39:807 812 11. Schatz D, Krischer J, Horne G, Riley W, Spillar R, Silverstein J, Winter W, Muir A, Derovanesian D, Shah S, Malone J, MacLaren N 1994 Islet cell antibodies predict insulin-dependent diabetes in United States school age children as powerfully as in unaffected relatives. J Clin Invest 93:2403 2407 12. Hagopian WA, Sanjeevi CB, Kockum I, Landin-Olsson M, Karlsen AE, Sundkvist G, Dahlquist G, Palmer J, Lernmark A 1995 Glutamate decarbox-

Schlosser et al. Karlsburg Type 1 Diabetes Risk Study J Clin Endocrinol Metab, May 2002, 87(5):2254 2261 2261 ylase, insulin and islet cell antibodies and HLA typing to detect diabetes in a general population-based study of Swedish children. J Clin Invest 95:1505 1511 13. Lévy-Marchal C, Dubois F, Noel M, Tichet J, Czernichow P 1995 Immunogenetic determinants and prediction of IDDM in French schoolchildren. Diabetes 44:1029 1032 14. Verge CF, Howard NJ, Rowley MJ, Mackay IR, Zimmet PZ, Egan M, Hulinska H, Hulinsky I, Silvestrini RA, Kamath S, Sharp A, Arundel T, Silink M 1994 Anti-glutamate decarboxylase and other antibodies at the onset of childhood IDDM: a population-based study. Diabetologia 37:1113 1120 15. Robinson DM, Holbeck S, Palmer J, Nepom GT 1989 HLA DQ 3.2 identifies subtypes of DR4 haplotypes permissive for IDDM. Genet Epidemiol 6: 149 154 16. Todd JA, Farral M 1996 Panning for gold: genome-wide scanning for linkage in type 1 diabetes. Hum Mol Genet 5:1443 1448 17. Michelsen B, Wassmuth R, Ludvigsson J, Lernmark Å, Nepom GT, Fisher L 1990 HLA heterozygosity in insulin-dependent diabetes is most frequent at the DQ locus. Scand J Immunol 31:405 413 18. Nepom BS, Palmer J, Kim SJ, Hansen JA, Holbeck SL, Nepom GT 1986 Specific genomic markers for the HLA-DQ subregion discriminate between DR4 insulin-dependent diabetes mellitus and DR4 seropositive juvenile rheumatoid arthritis. J Exp Med 164:345 350 19. Cucca F, Muntoni F, Lampis R, Frau F, Argiolas L, Silvetti M, Angius E, Cao A, De Virgiliis S, Congia M 1993 Combinations of specific DRB1, DQA1, DQB1 haplotypes are associated with insulin-dependent diabetes mellitus in Sardinia. Hum Immunol 37:85 94 20. Undlien DE, Friede T, Rammensee HG, Joner G, Dahl-Jorgensen K, Sovik O, Akselsen HE, Knutsen I, Ronningen KS, Thorsby E 1997 HLA-encoded genetic predisposition in IDDM. DR4 subtypes may be associated with different degrees of protection. Diabetes 46:143 149 21. Reijonen H, Ilonen J, Knip M, Kerblom HK 1991 HLA-DQB1 alleles and absence of Asp-57 susceptibility factors of IDDM in Finland. Diabetes 40: 1640 1644 22. Erlich HA, Griffith RL, Bugawan TL, Ziegler R, Aloer C, Eisenbarth GS 1991 Implication of specific DQB1 alleles in genetic susceptibility and resistance by identification of IDDM siblings with novel HLA-DQB1 allele and unusual DR2 and DR1 haplotypes. Diabetes 40:478 481 23. Pugliese A, Gianani R, Moromisato R, Awdeh ZL, Alper CA, Erlich HA, Jackson RA, Eisenbarth GS 1995 HLA-DQB1 *0602 is associated with dominant protection from diabetes even among islet cell antibody-positive firstdegree relatives of patients with IDDM. Diabetes 44: 608 613 24. Vandewalle CL, Decraene T, Schuit FC, De Leeuw IH, Pipeleers DG, Gorus FK 1993 Insulin autoantibodies and high titre islet cell antibodies are preferentially associated with the HLA DQA1 *0301-DQB1 *0302 haplotype at clinical onset of type I (insulin-dependent) diabetes mellitus before age 10 years, but not at onset between age 10 and 40 years. Diabetologia 36:1155 1162 25. Gorus FK, Vandewalle CL, Dorchy H, Van CP, Schuit FC, Pipeleers DG 1994 Influence of age on the association among insulin autoantibodies, islet cell antibodies, and HLA DQA1*0301-DQB1*0302 in siblings of patients with type 1 (insulin-dependent) diabetes mellitus. Belgian Diabetes Registry. J Clin Endocrinol Metab 78:1172 1178 26. Huang W, She JX, Muir A, Laskowska D, Zorovich B, Schatz D, Maclaren NK 1994 High risk HLA-DR/DQ genotypes for IDD confer susceptibility to autoantibodies but DQB1*0602 does not prevent them. J Autoimmun 7: 889 897 27. Delamaire M, Timsit J, Caillat-Zucman S, Maugendre D, Roussely H, Semana G, Yamamoto AM, van Endert P, Boitard C, Genetet B, Bach JF 1995 HLA-associated heterogeneity of the humoral response to islet antigens in insulin-dependent diabetes. J Autoimmun 8:645 657 28. Vandewalle CL, Falorni A, Lernmark, Goubert P, Dorchy H, Coucke W, Semakula C, Van der Auwera B, Kaufman L, Schuit FC, Pipeleers DG, Gorus FK 1997 Association of GAD65- and IA-2-autoantibodies with genetic risk markers in new-onset IDDM patients and their siblings. The Belgian Diabetes Registry. Diabetes Care 20:1547 1552 29. Williams AJK, Bingley PJ, Bonifacio E, Palmer JP, Gale EAM 1997 A novel micro-assay for insulin autoantibodies. J Autoimmun 10:473 478 30. Bugawan TL, Erlich HA 1991 Rapid typing of HLA-DQB1 DNA polymorphism using nonradioactive oligonucleotide probes and amplified DNA. Immunogenetics 33:163 170 31. Hall MA, Lanchbury JS, Lee JS, Welsh KI, Ciclitira PJ 1993 HLA-DQ2 seconddomain polymorphisms may explain increased trans-associated risk in celiac disease and dermatitis herpetifornis. Hum Immunol 38:284 292 32. Kulmala P, Rahko J, Savola K, Vähäsalo P, Veijola R, Sjöroos M, Reunanen A, Ilonen J, Knip M 2000 Stability of autoantibodies and their relation to genetic and metabolic markers of type I diabetes in initially unaffected schoolchildren. Diabetologia 43:457 464 33. Aanstoot HJ, Sigurdsson E, Jaffe M, Shi Y, Christgau S, Grobbee D, Bruining GJ, Molenaar JL, Hofman A, Baekkeskov S 1994 Value of antibodies to GAD 65 combined with islet cell cytoplasmic antibodies for predicting IDDM in a childhood population. Diabetologia 37:917 924 34. Bingley PJ, Bonifacio E, Shattock M, Gillmor HA, Sawtell PA, Dunger DB, Scott RD, Bottazzo GF, Gale EA 1993 Can islet cell antibodies predict IDDM in the general population? Diabetes Care 16:45 50 35. Maclaren N, Lan M, Coutant R, Schatz D, Silverstein J, Muir A, Clare-Salzer M, She JX, Malone J, Crockett S, Schwartz S, Quattrin T, DeSilva M, Vander Vegt P, Notkins A, Krischer J 1999 Only multiple autoantibodies to islet cells (ICA), insulin, GAD65, IA2 and IA-2 predict immune-mediated (type 1) diabetes in relatives. J Autoimmun 12:279 287 36. Ziegler AG, Hummel M, Schenker M, Bonofacio E 1999 Autoantibody appearance and risk for development of childhood diabetes in offspring of parents with type 1 diabetes: the 2-year analysis of the German BABYDIAB Study. Diabetes 48:460 468 37. Ziegler B, Augstein P, Lühder F, Northemann W, Hamann J, Schlosser M, Klöting I, Michaelis D, Ziegler M 1994 Monoclonal antibodies specific to the glutamic acid decarboxylase 65 kda isoform derived from a non-obese diabetic (NOD) mouse. Diabetes Res 25:47 64 38. Augstein P, Ziegler B, Schlosser M, Flassig S, Strebelow M, Ziegler M 1997 Immunohistochemical differentiation of monoclonal GAD antibodies recognizing linear or conformational epitope regions. Pancreas 15:139 146 39. Lermmark Å 1999 Type I diabetes. Clin Chem 45:1331 1338 40. Kulmala P, Savola K, Reijonen H, Veijola R, Vahasalo P, Karjalainen J, Tuomilehto-Wolf E, Ilonen J, Tuomilehto J, Akerblom HK, Knip M 2000 Genetic markers, humoral autoimmunity, and prediction of type 1 diabetes in siblings of affected children. Diabetes 49:48 58 41. Gianani R, Pugliese A, Bonner-Weir S, Shiffrin AJ, Soeldner JS, Erlich H, Awdeh Z, Alper CA, Jackson RA, Eisenbarth GS 1992 Prognostically significant heterogeneity of cytoplasmic islet cell antibodies in relatives of patients with type I diabetes. Diabetes 41:347 353 42. Pugliese A, Gianani R, Eisenbarth GS, De Camilli P, Solimena M 1994 Genetics of susceptibility and resistance to insulin-dependent diabetes in stiffman syndrome. Lancet 344:1027 1028 43. Kockum I, Lernmark Å, Dahlquist G, Falorni A, Hagopian WA, Landin- Olsson M, Li LC, Luthman H, Palmer JP, Sanjeevi CB, Sundkvist G, Ostman IJ 1996 Genetic and immunological findings in patients with newly diagnosed insulin-dependent diabetes mellitus. Horm Metab Res 28:344 347 44. Verge CF, Eisenbarth GS 1996 Strategies for preventing type I diabetes mellitus. West J Med 164:249 255