Concordance for Type 1 Diabetes in Identical Twins Is Affected by Insulin Genotype

Epidemiology/Health Services/Psychosocial Research ORIGINAL ARTICLE

1 2 KARL A. METCALFE, MRCP XIAOJIAN HUANG, PHD cated in disease risk, only a few with 1 2 GRAHAM A. HITMAN, MD TIMOTHY STEWART, PHD known function have been identified, in- 1 1 RACHEL E. ROWE, MD R. DAVID G. LESLIE, MD 1 cluding the insulin gene hypervariable re- MOHAMMED HAWA, BSC gion (2–5), the interleukin gene cluster (6,7), CTLA4 (8–10), and the vitamin D receptor (11,12). However, the only consistent association between type 1 diabetes and a non-MHC region is with the OBJECTIVE — Monozygotic twins are usually discordant (only one twin affected) for type 1 insulin gene. diabetes. Discordance for disease between such twins implies a role for nongenetically deter- The study of monozygotic twins con- mined factors but could also be influenced by a decreased load of diabetes susceptibility genes. The aim of this study was to determine whether two susceptibility genes were less prevalent in cordant or discordant for type 1 diabetes discordant twins compared with concordant twins. is a potentially powerful method to determine whether disease penetrance is a ran- RESEARCH DESIGN AND METHODS — We studied 77 monozygotic twin pairs dom effect or influenced by genetic (INS), 40 concordant and 37 discordant, for type 1 diabetes at polymorphism of the insulin gene factors. Monozygotic twins tend to de- region on chromosome 11p and HLA-DQBI. velop type 1 diabetes within 6 years of each other (13), so that twin pairs discor- RESULTS — The disease-associated INS genotype (Hph I) was identified in 87.5% of the ϭ dant for diabetes are likely to remain dis- concordant twins but only in 59.5% (P 0.005) of the discordant twins. Neither DQB1*0201 cordant. We identified such discordant nor DQB1*0302 was seen in 2 of 40 (5%) concordant twins compared with 8 of 37 (22%) twin pairs in a previous HLA phenotype discordant twins (P ϭ 0.04). No statistical differences were seen between concordant and discordant twins at individual alleles of DQB1. Combining insulin and DQ data, 5% of concor- study and showed that they were less dant twins compared with 32.4% of discordant twins had neither DQB1*0201/DQB1*0302 nor likely than concordant pairs to have both the high-risk Hph I INS “ϩϩ” genotype (P ϭ 0.002). HLA-DR3 and HLA-DR4 antigens (14). Subsequent studies indicated that pairs CONCLUSIONS — We conclude that the possession of the high-risk Hph I insulin genotype remaining discordant are more accurately increases the likelihood of identical twins being concordant for type 1 diabetes and that the defined as nondiabetic twins more than “load” of both major histocompatibility complex (MHC) and non-MHC susceptibility genes has 11 years from the diagnosis of the index an impact on the disease penetrance of type 1 diabetes. twin, with normal glucose tolerance and without antibody markers (15–17). Be- Diabetes Care 24:838–842, 2001 cause there are no previous genotype studies of HLA-DQB1 or non–MHC- encoded susceptibility genes to type 1 di- oncordance rates for type 1 diabetes factors also influence susceptibility to dis- abetes in identical twins, we have now are higher in monozygotic than ease. tested discordant as well as concordant C dizygotic twins, which is consistent At least 30% of the genetic suscepti- pairs using this definition of discordance. with a role for genetic factors determining bility to type 1 diabetes can be explained Our hypothesis was that discordance, as the disease (1). Nevertheless, ϳ50% of by an association with the major histo- compared with concordance, for type 1 monozygotic twins are discordant for compatibility complex (MHC) (2). Al- diabetes in monozygotic twins is caused type 1 diabetes, providing strong evi- though a large number of non-MHC by a decreased load of both MHC and dence that nongenetically determined chromosomal regions have been impli- non-MHC susceptibility genes.

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● RESEARCH DESIGN AND From the 1Department of Diabetes and Metabolic Medicine, St. Bartholomew’s and the Royal London School METHODS — We have studied and of Medicine and Dentistry, Queen Mary and Westﬁeld College, University of London, London, U.K.; and 2 compared 40 concordant and 37 discor- Genentech, South San Francisco, California. dant monozygotic twins for type 1 diabe- Address correspondence and reprint requests to Graham A. Hitman, MD, Department of Diabetes and Metabolic Medicine, Medical Unit, The Royal London Hospital, Whitechapel, London E1 1BB, U.K. E-mail: tes at both HLA-DQBI and the insulin [email protected]. gene (INS) loci. Twins studied were se- Received for publication 8 May 2000 and accepted in revised form 21 December 2000. lected blind to those included in our pre- X.H. and T.S. are employed by and hold stock in Genentech. Genentech is engaged in the development vious study 15 years ago of HLA-DR of pharmaceuticals for the treatment of diabetes and its complications. Abbreviations: ICA, islet cell antibody; MHC, major histocompatibility complex. phenotype (14). A total of 54% of twin A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion pairs were common to both studies. factors for many substances. Twins were ascertained from 1967 to

838 DIABETES CARE, VOLUME 24, NUMBER 5, MAY 2001 Metcalfe and Associates

1997 by referral to the British Diabetic Table 1—HLA-DQB1 allele positivity in monozygotic twins with type 1 diabetes Twin Study; monozygosity was established in each twin pair as described pre- % Concordant % Discordant viously (18,19). DNA was not available Allele (n ϭ 40) (n ϭ 37) P value from all twins. All twins were invited to attend the Diabetes Department for blood 0201 82.5 (n ϭ 33) 65 (n ϭ 24) 0.08 sampling, but in the study period, only a 0301 2.5 (n ϭ 1) 5.4 (n ϭ 2) NS fraction were able to do so. We therefore 0302 60 (n ϭ 24) 40.5 (n ϭ 15) 0.09 selected twin pairs who had attended the 0402 — 5.4 (n ϭ 2) NS Diabetes Department consecutively from 0501 — 2.7 (n ϭ 1) NS 1995 to 1997, who were either concor- 0502 13 (n ϭ 5) 5.4 (n ϭ 2) NS dant or discordant for type 1 diabetes. 0601 — 2.7 (n ϭ 1) NS The diabetic index twin from both con- 0602/3 2.5 (n ϭ 1) 5.4 (n ϭ 2) NS cordant and discordant pairs had a com- 0604 — 5.4 (n ϭ 2) NS parable age at diagnosis, and nondiabetic 0201/0302 47.5 (n ϭ 19) 27 (n ϭ 10) 0.06 twins from discordant pairs had a low risk Negative for DQB1*201 and/or *0302 5 (n ϭ 2) 22 (n ϭ 8) 0.04 of developing diabetes. Discordant pairs NS, not significant. likely to remain discordant were identified by being Ͼ11 years from the diagnosis of the index twin and, at 11 years by oped type 1 diabetes but was not part of Nomura et al. (22). DNA was extracted having normal glucose tolerance and the this study) to further evaluate the cutoff from blood (23), and a 241-bp region absence of islet cell antibody (ICA), GAD level for positivity for GAD65 and IA-2ic. from the polymorphic second exon of the Ͼ antibody, or IA-2ic antibody. Levels 3 SD higher than that of a control DQB1 gene was amplified in two separate Of 247 concordant twin pairs, we se- population were considered positive, as de- PCRs, specific for DQwI and DQw2,3,4 lected a consecutive series of 40 pairs at- scribed previously (17). In the latest IA- alleles. Seven DQwI and six DQw2,3,4 tending the Diabetes Center (mean age of 2ic (unpublished) and GAD65 antibody alleles can be identified after restriction diagnosis of index twin 16 Ϯ 9 years, proficiency workshops, our assays had a against panels of seven and five restriction range 1–39; 16 male pairs; median time to sensitivity, specificity, validity, and con- enzymes, respectively (22,24). The di- concordance 3.95 years, range 0.4–25). sistency of 100% in each (21). gested and undigested products were vi- Of 109 discordant twin pairs, we identi- Undiluted sera were screened for sualized by ethidium bromide staining fied a consecutive series of 37 pairs (mean ICA. The presence of ICA was established after electrophoresis in a 4.2% agarose age of diagnosis of index twin 19 Ϯ 11 by testing serum with indirect immuno- gel. years, range 2–59; 20 male pairs) in fluorescence on a fresh group O cryofixed which the index twin had a comparable human pancreas (16). Positive tests were INS Ն age at diagnosis to the index twin of the defined as 4 Juvenile Diabetes Founda- INS was studied at the 1,127 Pst I poly- Ј concordant pairs. Nondiabetic twins un- tion units. The ICA assay assessed in the morphic restriction site in the 3 untrans- derwent oral glucose tolerance testing ENDIT workshop (unpublished) had a lated region of the insulin gene and at the Ϫ (glucose was given as 75 g or 1.75 g/kg, sensitivity and specificity of 100%. 23 Hph I polymorphic restriction site whichever was less) to confirm that they (3). The digested and undigested prod- had neither diabetes nor impaired glucose HLA-DQBI typing ucts were visualized by ethidium bromide tolerance both initially and at intervals HLA-DQBI was studied by a polymerase staining after electrophoresis in a 4.2% thereafter; diagnosis of type 1 diabetes chain reaction restriction fragment– agarose gel. By convention, the disease was made according to standard guide- length polymorphism method combined associated allele at both sites is denoted ϩ lines (20). Sera were stored at –20°C and with group-specific primers adapted from as “ .” analyzed for each of three antibody assays. Analysis was performed on batched Table 2—؊23 Hph I insulin genotypes in monozygotic twins with type 1 diabetes samples by observers blinded to the clin- ical status of the subjects. INS Genotypes Autoantibody assays ϩ/ϩϩ/Ϫ (plus Ϫ/Ϫ) IA-2ic and GAD65 were measured by ra- dioimmunoprecipitation assays as de- Concordant twins (n ϭ 40) 0.875 (n ϭ 35) 0.125 (n ϭ 5) scribed previously (Promega, Madison, Discordant twins (n ϭ 37) 0.595 (n ϭ 22) 0.405 (n ϭ 15) WI) (17). All samples were tested in British caucasoid, ICA-negative controls (n ϭ 63)* 0.54 (n ϭ 34) 0.46 (n ϭ 29) duplicate, including positive and nega- Type 1 diabetes (n ϭ 116)* 0.72 (n ϭ 83) 0.28 (n ϭ 33) tive control standard sera. Each assay for Concordant twins versus discordant twins (P ϭ 0.005); concordant twins versus controls (P ϭ 0.0001); IA-2ic and GAD antibodies included concordant twins versus type 1 diabetes (P ϭ 0.054); discordant twins versus controls (P ϭ 0.68); dis- 65 cordant twins versus type 1 diabetes (P ϭ 0.22); type 1 diabetes versus controls (P ϭ 0.022). *Results serially diluted sera from a patient with taken from ref. 31. Patients with type 1 diabetes were serially selected from the Barts Oxford study, which is stiff-man syndrome and a prediabetic in- a population-based study of type 1 diabetes in the Oxfordshire region. The controls were all ICA-negative dividual (a twin who subsequently devel- and were aged-matched to the subjects with type 1 diabetes. The disease-associated allele is designated (ϩ).

DIABETES CARE, VOLUME 24, NUMBER 5, MAY 2001 839 Genetic susceptibility in twins

Table 3—Hph I high-risk insulin genotype and HLA DQB1*0201/*0302 combinations in erozygous phenotype HLA-DR3 and monozygotic twins with type 1 diabetes HLA-DR4 (linked to HLA-DQB1*0201 and DQB*0302, respectively) in concor- INS negative INS positive INS negative INS positive dant compared with discordant monozy- Twins DQ negative* DQ positive DQ positive DQ negative gotic twins. Several explanations are possible for the discrepant findings. First, Concordant (n ϭ 40) 0.05 (n ϭ 2) 0.40 (n ϭ 16) 0.075 (n ϭ 3) 0.475 (n ϭ 19) there is a need to study a larger sample Discordant (n ϭ 37) 0.324 (n ϭ 12) 0.189 (n ϭ 7) 0.081 (n ϭ 3) 0.405 (n ϭ 15) population to confirm or refute the re- INS, possession of Hph I high-risk genotype (ϩ/ϩ); DQ, possession of HLA DQB1*0201 and/or *0302. sults. The numbers recruited for this Overall Pearson ␹2, P ϭ 0.012; likelihood exact ␹2, P ϭ 0.011. *2 ϫ 2 Pearson ␹2, P ϭ 0.002. study represent the maximum number available at the time of the study; however, prospective sampling has been initi- Statistics 0302 nor the high-risk Hph I INS ϩ/ϩ ated. The discordant twin pairs were All statistical analyses were performed us- genotype (P ϭ 0.002). selected in the previous study because the ing SPSS for Windows software (version index twin had been diagnosed Ͼ5 years 9; Chicago, IL). Differences in frequencies CONCLUSIONS — Our hypothesis prior, so the risk of diabetes developing in of genotypes and alleles were calculat- was that discordance, compared with con- the nondiabetic twin was less than 10% ed by ␹2 analysis; all P values quoted are cordance, for type 1 diabetes in monozy- (19). In the present study, we selected two-sided. Power analysis based on a 31% gotic twins is caused by a decreased load discordant pairs in which the index twin difference of DR3/DR4 heterozygotes be- of both MHC and non-MHC susceptibil- had been diagnosed for more Ͼ11 years, tween concordant and discordant twins, ity genes. From our observations, we con- and the nondiabetic twin had normal glu- as observed in our previous study (14), clude that the “load” of both MHC and cose tolerance and no diabetes-associated indicated that 37 twin pairs in each group non-MHC susceptibility genes does have autoantibodies. In the present study, the would be sufficient to show a difference at an impact on disease penetrance of type risk of the nondiabetic twins developing the 5% level with 80% power. The study 1 diabetes. Thus, the 5Ј Hph I locus of diabetes is calculated as Ͻ2% (16,19). was approved by the ethical committee of the insulin gene was more prevalent in The more precise identification of twin the Barts and the London National Health concordant than discordant twin pairs. pairs likely to remain discordant for dia- Service Trust. Written and informed con- Furthermore, there was an increased fre- betes is an important feature of this sent was received from the twins or their quency of no high-risk HLA-DQB1 alleles present study. Comparison of our discor- parents, as appropriate. in discordant compared with concordant dant twins with a previous study of an twins. HLA-DQ protective alleles were as ICA-negative control population (n ϭ RESULTS prevalent in concordant pairs as in discor- 63) using the same methodology (24) dant pairs. Lastly, only 5% concordant vs. confirms the importance of HLA-DQ in HLA DQBI gene 32.4% discordant twin pairs with type 1 the susceptibility to type 1 diabetes in Results are presented in Table 1. No dif- diabetes possessed neither the HLA- these twins. There was an increase in ferences were found between concor- DQB1*0302/0201 combination nor the DQB1*0302 in discordant twins (65% dant and discordant twins for positivity INS Hph I high-risk genotype. positive) compared with controls (15.9% of DQB1*0302, DQB1*0201, the com- Differences between monozygous positive; P ϭ 0.0004) and a decrease in bination of DQB1*0302/0201, and twins cannot be determined by germ-line the protective DQB1*0602/3 allele (5.4% DQBI*0602/3. In contrast, 2 of 40 (5%) genes. However, there are a number of positive compared with 30.2%; P ϭ concordant twins compared with 8 of 37 ways genes could contribute to such dis- 0.003). This argues against a purely envi- (22%; P ϭ 0.04) discordant possessed ease discordance, including an increase ronmental cause for diabetes, even in neither DQB1*0201 nor DQB1*0302. in either protection or susceptibility to twins discordant for disease. Last, the the disease. In the absence of a compre- HLA-DR study of 1983 and the current INS hensive understanding of the genes that HLA-DQB1 study are not directly compa- Results are presented in Table 2. The type contribute to genetic susceptibility to di- rable with regard to HLA class II typing. 1 diabetes–associated ϩ/ϩ genotype was abetes, the analysis of monozygotic twins Although both HLA-DQB1*0201 and increased for the 5Ј Hph I (P ϭ 0.005; discordant and concordant for diabetes DQB1*0302 are in linkage disequilib- Table 2) polymorphism. Similarly, the 3Ј provides a powerful method to assess the rium with HLA-DR3 and DR4, respec- Pst I INS polymorphism was increased in impact of a particular gene on disease ex- tively, they also form haplotypes with frequency in concordant twins (frequen- pression. other DR antigens. cy of ϩ/ϩ genotype was 87.5% in con- Using DQB1 typing, we have con- No twin studies have assessed the im- cordant twins and 65% in discordant firmed the importance of the total load of pact of non-HLA genes on susceptibility twins; P ϭ 0.035). MHC alleles as a determinant of concor- to type 1 diabetes. We found an increased dance of disease, although we have been frequency of diabetes-associated insulin HLA DQB1 and INS combined unable to confirm our previous results by genotypes in the concordant twins. The Results are presented in Table 3. A total of studying individual DQB1 alleles com- insulin gene is a plausible candidate gene 2 of 40 (5%) concordant twins compared pared with the previous DR serological for type 1 diabetes because of its ␤-cell– with 12 of 37 (32.4%) discordant twins typing (14). In that previous study, we specific expression. It has been suggest- possessed neither DQB1*0201/DQB1* found an increased frequency of the het- ed that differences in the insulin gene

840 DIABETES CARE, VOLUME 24, NUMBER 5, MAY 2001 Metcalfe and Associates transcription could depend on sequence 6. Bergholdt R, Karlsen AE, Johannesen J, in identical twins of patients with type 1 variation in the disease-associated hyper- Hansen PM, Dinarello CA, Nerup J, Po- diabetes. Diabetes 44:1176–1179, 1995 variable region, although different studies ciot F: Danish Study Group of Diabetes in 16. Tun RYM, Peakman M, Alviggi L, Hussain have revealed discrepant results, depend- Childhood: characterisation of polymor- MJ, Lo SS, Shattock M, Pyke DA, Bottazzo ing on the reporter construct and the in phisms of an interleukin-1 receptor type 1 GF, Vergani D, Leslie RD: Importance of gene (IL1R1) promoter region (P2) and persistent cellular and humoral immune vivo/vitro models used (25–30). Inter- their relation to insulin-dependent diabe- changes before diabetes develops: prospec- estingly, whereas twins concordant for tes mellitus (IDDM). Cytokine 7:727–728, tive study of identical twins. BMJ 308: diabetes had a higher frequency of dis- 1995 1063–1068, 1994 ease-associated insulin genotypes than 7. Metcalfe KA, Hitman GA, Pociot F, Berg- 17. Hawa M, Rowe R, Lan MS, Notkins AL, discordant pairs, twins discordant for di- holdt R, Tuomilehto-Wolf E, Tuomilehto Pozzilli P, Christie MR, Leslie RD: Value abetes did not have a greater frequency J, Viswanathan M, Ramachandran A, Ne- of antibodies to islet protein tyrosine than did the islet cell–negative control rup J, The DiMe Study Group: An associ- phosphatase-like molecule in predicting population (31) (Table 2). This supports ation between type 1 diabetes and the type 1 diabetes. Diabetes 48:1270–1275, the hypothesis that there is a decreased interleukin-1 receptor type 1 gene. Hum 1997 load of diabetes susceptibility genes in Immunol 51:41–48, 1996 18. Barnett AH, Eff C, Leslie RDG, Pyke DA: 8. Marron MP, Raffel LJ, Garchon HJ, Jacob Diabetes in identical twins. Diabetologia discordant compared with concordant CO, Serrano-Rios M, Martinez Larrad MT, 20:87–93, 1981 twins. Last, this study would also support Teng WP, Park Y, Zhang ZX, Goldstein 19. Olmos P, A’Hern R, Heaton DA, Millward the interaction of HLA-DQ and insulin DR, Tao YW, Beaurain G, Bach JF, Huang BA, Risley D, Pyke DA, Leslie RD: The sig- gene alleles in determining susceptibility HS, Luo DF, Zeidler A, Rotter JI, Yang nificance in the concordance rate for type and protection to type 1 diabetes. MC, Modilevsky T, Maclaren NK, She 1 (insulin-dependent) diabetes in identi- JX: Insulin-dependent diabetes mellitus cal twins. Diabetelogia 31:747–750, 1988 (IDDM) is associated with CTLA4 poly- 20. National Diabetes Data Group: Classifica- Acknowledgments— This study was sup- morphisms in multiple ethnic groups. tion and diagnosis of diabetes mellitus ported by the Diabetic Twin Research Trust, Hum Mol Genet 6:1275–1282, 1997 and other categories with glucose toler- the Joint Research Board of St. Bartholomew’s 9. Donner H, Rau H, Walfish PG, Braun J, ance. Diabetes 28:1039–1057, 1979 Hospital, and the British Diabetic Association. Siegmund T, Finke R, Herwig J, Usadel 21. Schmidli RS, Colman PG, Bonifacio E: K.A.M. was a British Diabetic Association KH, Badenhoop K: CTLA4 alanine-17 Disease sensitivity and specificity of 52 as- R.D. Lawrence Research fellow during this confers genetic susceptibility to Graves’ says for glutamic acid decarboxylase anti- study. disease and to type 1 diabetes mellitus. bodies: the Second International GADAB We thank the twins and physicians who J Clin Endocrinol Metab 82:143–146, 1997 Workshop. Diabetes 44:636–640, 1995 made these studies possible, as well as Dr. 10. Nistico L, Buzzetti R, Pritchard LE, Van 22. Nomura N, Ota M, Tsuji K, Inoko H: HLA- David Pyke, who initiated the twin study. der Auwera B, Giovannini C, Bosi E, Lar- DQB1 genotyping by a modified PCR- rad MT, Rios MS, Chow CC, Cockram CS, RFLP method combined with group-spe- Jacobs K, Mijovic C, Bain SC, Barnett AH, cific primers. Tissue Antigens 38:53–59, References Vandewalle CL, Schuit F, Gorus FK, Tosi 1991 1. Kaprio J, Tuomilehto J, Koskenvuo M, R, Pozzilli P, Todd JA: The CTLA-4 gene 23. Balnaves ME, Nasioulas S, Dahl H-HM, Romanov K, Reunanen A, Eriksson J, region of chromosome 2q33 is linked to, Forrest S: Direct PCR from CVS and blood Stengard J, Kesaniemi YA: Concordance and associated with, type 1 diabetes. Hum lysates for detection of cystic fibrosis and for type 1 (insulin-dependent) and type 2 Mol Genet 5:1075–1080, 1996 Duchenne muscular dystrophy deletions. (non-insulin-dependent) diabetes melli- 11. McDermott MF, Ramachandran A, Ogun- Nucleic Acids Res 19:115, 1991 tus in a population based cohort of twins kolade BW, Aganna E, Curtis D, Boucher 24. Fennessy M, Hitman GA, Moore RH, Met- in Finland. Diabetologia 35:1060–1067, BJ, Snehalatha C, Hitman GA: Allelic vari- calfe K, Medcraft J, Sinico RA, Mustonen 1993 ation in the vitamin D receptor influences JT, D’Amico G: HLA-DQ gene polymor- 2. Davies JL, Kawaguchi Y, Bennet ST, Cope- susceptibility to IDDM in Indian Asians. phism in primary IgA nephropathy in man JB, Cordell HJ, Pritchard LE, Reed Diabetologia 40:971–975, 1997 three European populations. Kidney Int PW, Gough SC, Jenkins SC, Palmer SM: A 12. Pani MA, Knapp M, Donner H, Braun 49:477–480, 1996 genome-wide search for human type 1 di- J, Baur MP, Usadel KH, Badenhoop K: Vi- 25. Bennett ST, Lucassen AM, Gough SC, abetes susceptibility genes. Nature 371: tamin D receptor allele combinations in- Powell EE, Undlien DE, Pritchard LE, 130–136, 1994 fluence genetic susceptibility to type 1 Merriman ME, Kawaguchi Y, Dronsfield 3. Lucassen AM, Julier C, Berressi J-P, Boi- diabetes in Germans. Diabetes 49:504– MJ, Pociot F, Nerup J, Bouzekri N, Cam- tard C, Froguel P, Lathrop M, Bell JI: Sus- 507, 2000 bon-Thomsen A, Ronningen KS, Barnett ceptibility to insulin dependent diabetes 13. Fava D, Gardner S, Pyke DA, Leslie RDG: AH, Bain SC, Todd JA: Susceptibility to mellitus maps to a 4.1kb segment of DNA Evidence that the age at diagnosis to human type 1 diabetes at IDDM2 is deter- spanning the insulin gene and associated IDDM is genetically determined. Diabetes mined by tandem repeat variation at the VNTR. Nat Genet 4:305–310, 1993 Care 21:925–929, 1998 insulin gene minisatellite locus. Nat Genet 4. Bell GI, Horita S, Karam JH: A polymor- 14. Johnston C, Pyke DA, Cudworth AG, 9:284–292, 1995 phic locus near the human insulin gene is Wolf E: HLA-DR typing in identical twins 26. Kennedy CG, German MS, Rutter WJ: associated with insulin-dependent diabe- with insulin dependent diabetes: a differ- The minisatellite in the diabetes suscepti- tes mellitus. Diabetes 33:176–183, 1984 ence between concordant and discordant bility locus IDDM2 regulates insulin tran- 5. Hitman GA, Tarn AC, Winter RM: Type 1 pairs. BMJ 286:253–255, 1983 scription. Nat Genet 9:293–298, 1995 (insulin dependent) diabetes and a highly 15. Verge CF, Gianani R, Yu L, Pietropaolo M, 27. Lucassen AM, Screaton GR, Julier C, Elliot variable locus close to the insulin gene on Smith T, Jackson RA, Soeldner S, Eisen- TJ, Lathrop M, Bell JI: Regulation of insu- chromosome 11. Diabetologia 28:218–222, barth GS: Late progression to diabetes and lin gene expression by the IDDM associ- 1985 evidence for chronic ␤-cell autoimmunity ated, insulin locus haplotype. Hum Mol

DIABETES CARE, VOLUME 24, NUMBER 5, MAY 2001 841 Genetic susceptibility in twins

Genet 4:501–506, 1995 man thymus and transcription levels cor- by INS VNTR alleles at the IDDM2 locus. 28. Owerbach D, Gabay KH: The search for related with allelic variation at the INS Nat Genet 15:289–292, 1997 IDDM susceptibility genes: the next gen- VNTR-IDDM 2 susceptibility locus for type 31. Siddiqi A, Metcalfe KA, Lai M, Bingley PJ, eration. Diabetes 45:544–551, 1996 1 diabetes. Nat Genet 15:293–297, 1997 Sachs J, Bottazzo GF, Gale EAM, Hitman 29. Pugliese A, Zeller M, Fernandez A Jr, Zalc- 30. Vaﬁadis P, Bennett ST, Todd JA, Nadeau GA: Susceptibility loci in ICA positive berg LJ, Bartlett RJ, Ricardi C, Pietropaolo J, Grabs R, Goodyer CG, Wickramasinghe subjects, IDDM and other auto immune M, Eisenbarth GS, Bennett ST, Patel DD: S, Colle E, Polychronakos C: Insulin ex- disease: implications for disease predic- The insulin gene is transcribed in the hu- pression in human thymus is modulated tion. Diabetologia 37 (Suppl. 1):A55, 1994

842 DIABETES CARE, VOLUME 24, NUMBER 5, MAY 2001