IA2 and anti-SOX13

Diabetes Research and Clinical Practice 52 (2001) 205– 211 www.elsevier.com/locate/diabres

Antibodies to diabetes-associated autoantigens in Indian patients with Type 1 diabetes: prevalence of anti-ICA512/IA2 and anti-SOX13 S. Fida a, M. Myers a, I.R. Mackay a, P.Z. Zimmet a,b, V. Mohan c, R. Deepa c, M.J. Rowley a,* a

Department of Biochemistry and Molecular Biology, Monash Uni6ersity, Wellington Road, Clayton, Victoria 3168, Australia b International Diabetes Institute, Kooyong Road, Caulfield, Victoria 3162, Australia c M.V. Diabetes Specialities Centre and Madras Diabetes Research Foundation, 35 Conron Smith Road, Chennai 600086, India Received 3 October 2000; received in revised form 20 November 2000; accepted 16 January 2001

Abstract We ascertained frequencies of autoantibodies to a suite of islet cell antigens including ICA512/IA2 and SOX13 in Asian Indians with Type 1 diabetes and in other forms of diabetes. Autoantibodies to ICA512/IA2 and SOX13 were tested by radioimmunoprecipitation assay, and results were amalgamated with previous data on antibodies to glutamic acid decarboxylase (GAD) and to islet cell cytoplasmic antigens (ICA). The frequency of anti-SOX13 was higher in Asian Indians than in Europids. Overall, the combined frequency for all autoantibodies to diabetes-associated antigens in Type 1 diabetes in Indians approached the frequency reported for Europids. There was an unexpectedly high frequency of autoantibody reactions to any one of the autoantigens tested (24%) in fibrocalculous pancreatic diabetes, however, individual autoantibody frequencies were relatively low. Our data indicate that, whatever the population studied, testing for multiple autoantigenic reactivities is more informative than more limited testing, and that there may be regional (presumably ethnically based) differences in levels of particular autoantibodies in cases of Type 1 diabetes. © 2001 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Type 1 diabetes; Type 2 diabetes; Fibrocalculous pancreatic diabetes (FCPD); ICA512/IA2; SOX13

1. Introduction Type 1 diabetes is more prevalent among individuals of Northern European descent than * Corresponding author. Tel.: + 61-3-99051438; fax: +613-99054699. E-mail address: [email protected] (M.J. Rowley).

among various other ethnic groups. This applies particularly to Asians and African populations including Chinese, Koreans, Malaysians, Thais and Tanzanians (see Discussion). The disease in Europids is attributed to autoimmune destruction of the islet b-cells of the pancreas, as judged by the frequency of multiple autoantibodies, in aggregate nearly 100% in acute-onset cases of child-

0168-8227/01/$ - see front matter © 2001 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 1 6 8 - 8 2 2 7 ( 0 1 ) 0 0 2 3 0 - 3

206

S. Fida et al. / Diabetes Research and Clinical Practice 52 (2001) 205–211

hood [1,2]. Among non-Europid populations the prevalence of Type 1 diabetes and the frequency of autoantibodies in incident cases are lower than among Europid populations [3]. The explanation may be that autoantibodies disappear more quickly among non-Europid populations, or that there are autoantibodies to islet-cell autoantigens other than those usually measured, e.g. ICA and anti-GAD, or that ‘Type 1 diabetes’ in Asians includes a higher frequency of non-autoimmune types of disease. Mohan et al. (1998) studied antibodies to glutamic acid decarboxylase (antiGAD) and islet cell cytoplasmic antigens (ICA) in various categories of diabetes among Southern Indians and ascertained lower frequencies in Type 1 diabetes, 47 and 54% respectively, than frequencies reported for Europids of around 70% [2,5– 7]. We have extended the earlier study of Mohan et al. to include the frequency of autoantibodies to the tyrosine phosphate-like molecule ICA512/IA2 and to a novel autoantigen initially described as ICA12 [8,9]. ICA12 has now been identified as a member of the SOX family of transcription factors and the human equivalent of mouse Sox13 [10]. We have previously reported frequencies of anti-ICA512/IA2 and anti-SOX13 among Australian Europids with Type 1 diabetes of up to 60– 70% and 20% respectively [11], dependent on age and duration of diabetes.

2. Patients and methods The subjects included in this study have been described previously [4] and included 43 with Type 1 diabetes, 71 with Type 2 diabetes, 57 with fibrocalculous pancreatic diabetes (FCPD), 48 first-degree relatives of patients with Type 1 diabetes and 50 normal subjects who were healthy Indians residing in Chennai. The diagnosis of types of diabetes was based on standard criteria. For Type 1 (insulin dependent) diabetes mellitus, these included an abrupt onset of diabetes; susceptibility to ketosis in the basal state or documented episodes of ketoacidosis in the past; requirement of insulin for control of hyperglycemia from the onset of disease and Cpeptide levels less than 0.01 pmol/ml in the fasting

state and 0.02 pmol/ml in the stimulated state. The diagnosis of Type 2 diabetes mellitus was based on the WHO study group report criteria [12], and included response to diet or diet and oral hypoglycemic agents, and absence of ketosis. FCPD was diagnosed as described previously [4]; all FCPD patients required insulin for control of diabetes but none had ketoacidosis or ketonuria. Patients with Type 1 diabetes were subclassified according to the age of onset of the disease into children (B 20 years) or adults (\ 20 years), and according to whether the disease at the time of testing for autoantibodies was of short duration (B 12 months) or long duration (\ 12 months). Of the 43 patients with Type 1 diabetes, 24 had a childhood onset and 19 had an adult onset. We also added data on the few of the Type 1 diabetic patients that had not been reported earlier. Anti-ICA512/IA2 and anti-SOX13 were measured by radioimmunoprecipitation (RIP) using 35 S-labeled antigen prepared by in vitro transcription and translation using rabbit reticulocyte lysate [1]. A full length cDNA for human ICA512/IA2 (amino acids 1–979) and a cDNA for human SOX13 were transcribed in vitro followed by translation in rabbit reticulocyte lysate (Promega, Maddison, WI) in the presence of 35S express label mix (Dupont, Boston, MA). The 35S labeled protein was separated from unincorporated label by gel filtration on Sephadex G25 columns (NAP 5, Pharmacia, Uppsala, Sweden) before use in the RIP assay. The expressed ICA512/IA2 protein migrated as a 102 kDa component, and the SOX13 protein migrated as a 62 kDa component by autoradiography after separation by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE). RIP assays were performed in square-well titer plates with 2 ml capacity per well (Beckman, USA), as described previously [13]. Sera from patients from each group, and controls, were tested in the same assay and each assay run included a positive reference serum (PR) and a negative control serum (NC). The results for test sera (TS) were expressed as an index calculated from the counts per minute as follows: index= [(TS− NC)/(PR − NC)] × 100. A positive result was 10 units or greater, representing the mean plus three standard

S. Fida et al. / Diabetes Research and Clinical Practice 52 (2001) 205–211

deviations (S.D.) for healthy Indian controls. The calibrations were as follows, the mean (range) cpm precipitated for anti-ICA512/IA2 by the PR serum was 1336 (1150– 1591), and by the NC serum 99 (83–179), in 11 separate assays. The inter-assay coefficient of variation (CV) was 12% for the PR serum, and 16% for the NC serum. The mean (range) counts per minute precipitated for anti-SOX13 by the PR serum was 3591 (3320– 3778), and by the NC serum 298 (294– 335), in 10 separate assays. The inter-assay coefficient of variation (CV) was 7% for the PR serum and 5% for the NC serum. The assays for ICA and anti-GAD on these sera have been described previously [4]. The frequencies of autoantibody specificities were examined individually and in combination.

207

though the groups tested were too small for statistical significance. Anti-ICA512/IA2 was rare in adults at all ages (Table 1). Considering all four autoantibodies in combination, one or more was detected in 74% of patients with Type 1 diabetes, 10% of Type 2 diabetes, 6% of healthy relatives, and 6% of

3. Statistical analysis Differences between the frequencies of autoantibodies were compared using the Chi-squared test with Yate’s correction using Statistica for Windows, version 4.5 (StatSoft, Tulsa, OK).

4. Results The frequencies of diabetes autoantibodies of four different specificities, ICA, GAD, ICA512/ IA2 and SOX13, in all the groups tested, and the frequencies in Type 1 diabetes according to youth or adult onset and short or long duration of disease, are shown in Table 1. The levels of autoantibodies are shown in Fig. 1. The data for ICA and anti-GAD are those reported previously [4], and comparative data for Europids included in Table 1 also have been reported previously [11]. As shown in Table 1 anti-ICA512/IA2 were detected in 23% of patients with Type 1 diabetes (P= B0.003 vs. controls), anti-SOX13 in 30% (P= B0.0006), anti-GAD in 51% (P = 0.001), and ICA in 53% (P=0.001). No other group differed significantly from the controls. Among children, but not adults, the frequency of each autoantibody declined according to duration of disease, assessed as B1 year or \ 1 year, al-

Fig. 1. Frequencies and levels of diabetes-related autoantibodies in categories of diabetes, Type 1 diabetes (D1), Type 2 diabetes (D2), FCPD (see text), relatives (REL) and healthy controls (CONT). RU is reference units for anti-ICA512/IA2, anti-SOX13 and anti-GAD and JDFU is Juvenile Diabetes Foundation Units for ICA. Dotted line shows upper normal limit (mean + 3SD) of controls. NT: not tested.

208

Group

Mean age (years 9S.D.)

M.F.

Mean duration (months9 S.D.)

Type 1 diabetes Recent onset in children: n =15 Recent onset in adults: n = 7 Long duration in children: n =9 Long duration in adults: n = 12 All Type 1 diabetes: n =43

10+ 6 5 23+ 6 4 11+ 6 6 35+ 6 8 19+ 6 12

6:9 7:0 6:3 5:7 24:19

5+ 6 5 6+6 3 41+ 6 12 42+ 6 16 23+ 6 21

Type 2 diabetes Diabetes: n = 71 FCPD: n= 57 Relatives: n= 48 Healthy subjects: n= 50

48+ 6 11 37+ 6 12 38+ 6 8 29+ 6 8

48:23 39:18 24:24 23:27

52+ 6 38 NA NR NR

a

Frequency of autoantibodies n(%) [% Europids c] ICA512

SOX 13

ICA*

GAD*

6(40)[78] 1(14)[31] 1(11)[11] 2(17)[3] 10(23)[30]

7(47)[6] 2(28)[23] 1(11)[14] 3(25)[31] 13(30)[18]

9(82)[72] 3(43)[23] 2(22)[7] 7(58)[14] 23(53)[30]

10(67)[81] 4(57)[46 2(22)[50] 6(50)[72] 22(51)[66]

1(1) 4(7) 0 0

3(4) 4(7) 2(4) 1(2)

7(9) 3(6) NA 2(4)

4(5) 4(7) 1(2) 0

At least one Ab

14(93)[90] 5(71)[77] 5(55)[68] 8(66)[86] 32(74)[81]

7(10) 14(24) 3(6) 3(6)

SD, standard deviation; M:F, male:female ratio; NA, not available; NR, not relevant; n, number; (%), percentage.*Frequencies of ICA and anti-GAD for Indian subjects were previously reported [4].[ c] Frequencies of autoantibodies for Australian Europid subjects as previously reported [11].

S. Fida et al. / Diabetes Research and Clinical Practice 52 (2001) 205–211

Table 1 Frequency of antibodies to four diabetes-associated autoantigens according to category of diabetesa

S. Fida et al. / Diabetes Research and Clinical Practice 52 (2001) 205–211

healthy control subjects. The combined frequency was highest in children with newly diagnosed Type 1 diabetes (93%). When the data for frequencies of autoantibodies in combination were analyzed in Type 1 diabetes, there were 32 (74%) seropositive for at least one autoantibody and 26 (60%) with autoantibodies of more than one specificity; eight were seropositive for three of the four autoantibodies tested. Among all other groups, only one patient with Type 2 diabetes, was seropositive for more than one autoantibody. The frequency of autoantibodies in FCPD was unexpectedly high, since 14 of 57 patients (24%) were seropositive to one or another of the panel of autoantigens; however, there were no multiple reactivities, and levels of reactivity albeit positive were mostly low (Fig. 1). 5. Discussion There are many non-Europid populations for whom Type 1 diabetes has a much lower prevalence [1–3], and for whom the frequency of diabetes-associated autoantibodies in incident cases appears lower. The present study extends data on serological testing of autoantibodies to diabetesrelevant autoantigens in Southern Indians with Type 1 diabetes, by including data for antibodies to ICA512/IA2 and SOX13. Among Australian Europids with Type 1 diabetes, the reported frequencies of anti-ICA512/IA2 and anti-SOX13 overall are 30 and 18% respectively; the frequency of anti-SOX13 is higher in adults than in children (23 vs. 6%), and appears not to vary greatly according to duration of disease [11]. We previously reported anti-ICA in 53% and anti-GAD in 47% of Asian Indians with Type 1 diabetes, higher than previously reported for other Asian populations but lower than reported in Europids (60–80%). The frequency of antiGAD reported among incident cases of Type 1 diabetes was 35% for Thais [14], 30% for Hong Kong Chinese [15], 39% for Singapore Chinese [16], 35% for young Malaysians of Malay, Chinese or Indian origin [17] and 30% for Koreans [3]. Data for Japanese vary, with cited frequencies of anti-GAD ranging from 35% [18] to 55% [19], and up to 70% in selected groups [20].

209

We re-analyzed the results for anti-ICA and anti-GAD among the Asian Indians according to the duration and age of disease onset. The highest prevalence of anti-ICA and anti-GAD (72 and 67%) in this Asian Indian cohort was observed for patients with recent disease onset, which is comparable to the results for Europids (80–90% and 81%). Anti-ICA positivity declined with increasing disease duration, similar to previous reports in both the populations [21,22]. Furthermore our study indicated similar frequencies of anti-GAD in recent or long duration adult patients (57 vs. 50%). This is in line with the view that anti-GAD declines more slowly than other autoantibodies [23]. In children, however the anti-GAD frequency was 67 versus 22% in recent and long duration cases respectively. The possible explanation may be the small sample size of our study or a more rapid disappearance of autoantibodies in some cases, which needs further validation in a longitudinal study. Antibodies to ICA512/IA2 have been reported to be a poor marker for Type 1 diabetes in Asian populations [22] and our observations indicated a frequency of 40% in children with recent onset compared with 50–80% in Europids, but only 11% in children with long duration of the disease. In Indians the frequency of anti-ICA512/IA2 declines with longer duration of disease similar to the Europids [23]. The frequency of antibodies to SOX13 in Asian Indians was higher than we have reported for Europids with Type 1 diabetes (30 vs. 18%) and was similar to anti-GAD in their persistence during disease duration. However a slightly higher prevalence was observed in children with recent onset (47%). This observation may have important implications for disease prediction in the Asian Indian population. When antibody combinations were analyzed for patients with Type 1 diabetes the frequency with any one antibody in our cohort of Asian Indians was similar to the Europids (74 vs. 81%). If this result is restricted to ICA and GAD alone 62% cases were identified, hence combined antibody testing may be necessary for complete ascertainment in Asians. There are few reports describing testing for more than two antibodies in Asian

210

S. Fida et al. / Diabetes Research and Clinical Practice 52 (2001) 205–211

populations. The data presented here suggest that different autoantibodies may be prominent among different populations. In particular anti-SOX13 is more prominent in Asian Indians than in Europids. The present study on Asian Indians with Type 1 diabetes indicates lower frequency for individual autoantibodies but a similar frequency for combined autoantibodies compared with Europid populations. Hence, Type 1 diabetes among Asian Indians, like Europids, may predominantly be of autoimmune origin. The lower frequencies for individual autoantibodies may indicate a more rapid disappearance of autoantibodies, or an occurrence of alternative and as yet unidentified sets of autoantibodies. Alternatively, for genetic or other reasons, autoantibodies may not be expressed, noting that there is minimal expression of autoimmune serological markers in NOD mice [24], where b-cell destruction does occur by autoimmune processes. Present knowledge is insufficient for us to prioritize the foregoing possibilities, except that the frequency of particular autoantibodies expressed in Type 1 diabetes can vary according to the age of onset and the duration of disease [6,25]. In this latter context we note that, in the present study, the frequency of one or more diabetes autoantibodies in a small group (15 subjects) of youth onset cases with short duration Type 1 diabetes was 93%, similar to the figure for Europids [1]. It is well recognized that ICA and anti-ICA512/IA2 tend to disappear with increasing duration [23,26], in contrast to persistence of anti-GAD [5] and anti-SOX13 (unpublished data) in Europeans. However, a recent study in our laboratory of a community-based rather than a clinic-based sample of Type 1 diabetes (Australian Europids) mostly of long duration, the frequency of antiGAD was only 48%, suggesting that anti-GAD had disappeared in a proportion of these cases [27]. Moreover, we have shown that HLA-genotypes affect both the frequency and persistence of antibodies to GAD in patients with Type 1 diabetes [28,29]. Hence comparisons of frequencies of autoantibodies across different populations requires that case samples be very carefully matched.

Finally, a comment is needed on the high frequency (24%) of autoantibodies in FCPD. Among possible explanations would be the inclusion in this group of cases of mixed pathogenesis, or the occurrence of a secondary low-level autoimmune response consequential to the pathology of FCPD itself, noting that the levels of the autoantibodies in the seropositive cases of FCPD were just above the cut-off. In conclusion, an extension of studies on diabetes-associated autoantibodies in Indians with Type 1 diabetes indicates: (a) a slightly lower frequency of autoantibodies to ICA and GAD, and a much lower frequency to ICA512/IA2 than among Europids, recognizing difficulties in matching with other cross-sectional studies; (b) a similar frequency of combined autoantibodies as in Europids; (c) a notably higher frequency of reactivity with a newly-identified reactant SOX13; (d) an unexpectedly high frequency of autoantibodies in FCPD where islet cell damage is consequent on inflammation that is not primarily autoimmune in nature. Acknowledgements We thank Dr. Daniel Rabin, Bayer Research Centre, West Hayen, CT for providing the ICA12 and ICA512 cDNA. The work was supported by a grant from Bayer Corporation, USA. References [1] S.J. Feeney, M.A. Myers, I.R. Mackay, et al., Evaluation of ICA512As in combination with other islet cell autoantibodies at the onset of IDDM, Diabetes Care 20 (1997) 1403 – 1407. [2] P.J. Bingley, E. Bonifacio, A.J.K. Williams, S. Genovese, G.F. Bottazzo, E.A.M. Gale, Prediction of IDDM in the general population, Diabetes 46 (1997) 1701 – 1710. [3] T. Tuomi, P. Zimmet, M.J. Rowley, et al., Differing frequency of autoantibodies to glutamic acid decarboxylase among Koreans, Thais, and Australians with diabetes mellitus, Clin. Immunol. Immunopathol. 74 (1994) 202 – 206. [4] V. Mohan, D.R. Bhatia, A.K. Singh, et al., Antibodies to pancreatic islet cell antigens in diabetes seen in southern India with particular reference to fibrocalculous pancreatic diabetes, Diabetic Med. 15 (1998) 156 – 159.

S. Fida et al. / Diabetes Research and Clinical Practice 52 (2001) 205–211 [5] M.J. Rowley, I.R. Mackay, Q.Y. Chen, W.J. Knowles, P.Z. Zimmet, Antibodies to glutamic acid decarboxylase discriminate major types of diabetes mellitus, Diabetes 41 (1992) 548 – 551. [6] C.F. Verge, N.J. Howard, M.J. Rowley, et al., Antibodies to glutamic acid decarboxylase and other antibodies at the onset of childhood IDDM: a population based study, Diabetologia 37 (1994) 1113 –1120. [7] H.J. Anstoot, E. Sigurdsson, M. Jaffe, et al., Value of antibodies to GAD65 combined with islet cell cytoplasmic antibodies for predicting IDDM in a childhood population, Diabetologia 37 (1994) 917 –924. [8] D.U. Rabin, S.M. Pleasic, R. Palmer-Crocker, J.A. Shapiro, Cloning and expression of IDDM specific human autoantigens, Diabetes 41 (1992) 183 – 186. [9] M.J. Rowley, M.A. Myers, S. Fida, et al., Antibodies to ICA12, a new autoantigen in IDDM (abstract), Diabetes 47 (1998) A202. [10] H. Kasimiotis, M.A. Myers, A. Argentaro, et al., The SRY-related protein, SOX13, is a diabetes autoantigen expressed in pancreatic islets, Diabetes 49 (2000) 555 – 561. [11] H. Kasimiotis, S. Fida, M. Rowley, et al., 2001. Autoantibodies to SOX13 (ICA12) are associated with Type 1 diabetes, Autoimmunity, (in press). [12] World Health Organization Study Group on Prevention of Diabetes Mellitus, 1994. Technical Report Series No. 844, WHO, Geneva. [13] S. Fida, M.J. Rowley, 1998. Radioimmunoprecipitation assay for antibodies to glutamic acid decarboxylase and other autoantigens in microplates, Topcount Topics, Packard Publications, TCA-032. [14] C. Rattarasarn, M. Aguilar-Diosdado, S. Soonthornpun, N. Patarakijvanich, S. Jaruratanasirikul, GAD antibodies in IDDM in Thailand, Diabetes Care 19 (1996) 674 –675. [15] J.C. Chan, V.T. Yeung, C.C. Chow, et al., Pancreatic beta cell function and antibodies to glutamic acid decarboxylase (anti-GAD) in Chinese patients with clinical diagnosis of insulin-dependent diabetes mellitus, Diabetes Res. Clin. Pract. 32 (1996) 27 –34. [16] A.C. Thai, W.Y. Ng, K.Y. Loke, W.R. Lee, K.F. Lui, J.S. Cheah, Anti-GAD antibodies in Chinese patients with youth and adult-onset IDDM and NIDDM, Diabetologia 40 (1997) 1425 –1430. [17] W.M. Wan Nazaimoon, I. Faridah, M. Singaraveloo, et al., Prevalence of glutamic acid decarboxylase antibodies amongst young Malaysian diabetics, Diabetes Res. Clin. Pract. 43 (1999) 59 –66. [18] A. Tsuruoka, I. Matsuba, T. Toyota, G. Isshiki, S. Nagataki, Y. Ikeda, Antibodies to GAD in Japanese diabetic

.

[19]

[20]

[21]

[22]

[23]

[24]

[25]

[26]

[27]

[28]

[29]

211

patients: a multicenter study, Diabetes Res. Clin. Pract. 28 (1995) 191 – 199. K. Yamada, X. Yuan, C. Inada, et al., Combined measurements of GAD65 and ICA512 antibodies in acute onset and slowly progressive IDDM, Diabetes Res. Clin. Pract. 35 (1997) 91 – 98. H. Takino, E. Kawasaki, M. Yano, et al., The heterogeneity of whole islet cell cytoplasmic antibodies evidenced by absorption test with glutamic acid decarboxylase, J. Autoimmunity 7 (1994) 389 – 398. Diabetes Epidemiology Research International Group, 1988. Geographic patterns of childhood insulin-dependent diabetes mellitus. Diabetes 37, 1113-1119. M.A. Kelly, N.S. Alvi, N.J. Croft, et al., Genetic and immunological characteristics of Type 1 diabetes mellitus in an Indo-Aryan population, Diabetologia 43 (2000) 450 – 456. M.A. Myers, D.U. Rabin, M.J. Rowley, Pancreatic islet cell cytoplasmic antibody in diabetes is represented by antibodies to islet cell antigen 512 and glutamic acid decarboxylase, Diabetes 44 (1995) 1290 – 1295. I.R. Mackay, A. Bone, T. Tuomi, et al., Lack of autoimmune serological reactions in rodent models of insulin dependent diabetes mellitus, J. Autoimmunity 9 (1996) 705 – 711. R.S. Schimidli, H.J. De Aizpurua, L.C. Harrison, P.G. Colman, Antibodies to glutamic acid decarboxylase in at-risk and clinical insulin-dependent diabetic subjects: relationship to age, sex and islet cell antibody status, and temporal profile, J. Autoimmun. 7 (1994) 55 – 66. M.R. Christie, S. Genovese, D. Cassidy, et al., Antibodies to islet 37k antigen, but not to glutamic acid decarboxylase, discriminate rapid progression to IDDM in endocrine autoimmunity, Diabetes 43 (1994) 1254 – 1259. T.M.E. Davis, P.Z. Zimmet, W.A. Davis, D.G. Bruce, S. Fida, I.R. Mackay, 2000. Autoantibodies to glutamic acid decarboxylase in diabetic patients from a multi-ethnic urban Australian community: The Fremantle Diabetes Study, Diabetic Med. 17 (2000) 667 – 674. S.W. Serjeantson, J. Court, I.R. Mackay, et al., HLA-DQ Genotypes are associated with autoimmunity to glutamic acid decarboxylase in insulin-dependent diabetes mellitus patients, Human Immunol. 38 (1993) 97 – 104. S.W. Serjeantson, M.R.J. Kohonen-Corish, M.J. Rowley, I.R. Mackay, W. Knowles, P. Zimmet, Antibodies to glutamic acid decarboxylase are associated with HLA-DR genotypes in both Australians and Asians with type 1 (insulin-dependent) diabetes mellitus, Diabetologia 35 (1992) 996 – 1001.