An investigation of the association between HLA-DQ heterodimers and type 1 (insulin-dependent) diabetes mellitus in five racial groups

An Investigation of the Association Between HLA-DQ Heterodimers and Type 1 (Insulin-Dependent)Diabetes Mellitus in Five Racial Groups Michelle A. Penny, Catherine H. Mijovic, David A. Cavan, Karen H. Jacobs, David Jenkins, Jeremy Fletcher, and Anthony H. Barnett

ABSTRACT: The association between HLA-DQ c~Arg52-HLA-DQ/3 non-Asp57 heterodimers and type 1 (insulin-dependent) diabetes was compared in Japanese, Chinese, Caucasian, North Indian Asian, and AfroCaribbean patients to determine their importance in disease susceptibility. The potential to encode four Arg52-non-Asp57 DQ heterodimers, two in cis and two in trans, was significantly associated with increased risk of type 1 diabetes in all races except the Japanese. The possibility of encoding two Arg52-non-Asp57 heterodimers was also significantly associated with increased risk of the disease in all races except the Japanese. The possibility of

encoding one heterodimer was not significantly associated with type 1 diabetes in any of the races studied. Heterogeneity testing revealed significant differences in RR values for four, two, and one heterodimers in all races except the Japanese and significant differences in RR for four and two heterodimers when compared across the races. This, together with the lack of an association between Arg52-non-Asp57 heterodimers and type 1 diabetes in the Japanese, suggests that, assuming the same genetic basis for disease in all races, the heterodimer is unlikely to be of primary importance in susceptibility to the disease. Human Immunology 38, 179-183 (1993)

ABBREVIATIONS

Arg Asp CL HLA IDDM

arginine aspartate confidence limit human leukocyte antigen insulin-dependent diabetes mellitus

MHC NS RR SSO

major histocompatibility complex not significant relative risk sequence-specific oligonucleotide

INTRODUCTION Associations between genes in the major histocompatibility complex (MHC) and inherited susceptibility to type 1 (insulin-dependent) diabetes mellitus (IDDM) are well documented [1]. It has been suggested that the amino acids at position 57 of the DQ/3 chain and position 52 of the DQo~ chain in the D Q molecule may have

From the Department of Medicine, University of Birmingham, Birmingham, England. Address reprint requests to Dr. M. A. Penny, Department of Medicine, Clinical Research Block, Queen Elizabeth Hospital, Edgbaston, Birmingham, BI5 2TH England. Received ;'E) December 17, 1992," acceptedJuly 8, 1993. Human Immunology 38, 179-183 (1993) © American Society for Histocompatibility and Immunogenetics, 1993

a role to play in susceptibility to type 1 diabetes. The presence of aspartate (Asp) at position 57 of the DQ/3 chain has been correlated with protection from the disease [2] and the presence of arginine (Arg) at position 52 of the DQo~ chain has been implicated in susceptibility [3]. It has been proposed that the combination of an Arg52 D Q ~ chain and a non-Asp57 DQ/~ chain results in a "diabetogenic D Q heterodimer" [3]. Recently, Gutierrez-Lopez et al, [4] concluded that susceptibility to I D D M correlated quantitatively with expression of "diabetogenic" H L A - D Q ~ A r g 5 2 - H L A - D Q / 3 non-Asp57 alleles in a study o f Spanish patients. In that study, the possibility of encoding four heterodimers gave the high179 0198-8859/93/$6.00

180

est absolute risk for diabetes and one heterodimer the lowest. Analysis of HLA associations with IDDM is complicated by strong linkage disequilibrium within the class II region of the MHC. Recombination events during evolution have generated combinations of MHC alleles that vary between populations [5]. A transracial approach is, therefore, valuable in the investigation of the importance of the HLA-DQ oeArg52-DQ3 non-Asp57 heterodimer in susceptibility to IDDM. If the presence of the Arg52-non-Asp57 heterodimer is of primary importance in disease susceptibility, a similar effect should be seen in all racial groups, despite linkage disequilibrium. We have, therefore, examined the frequencies of the possible Arg 52-non-Asp 57 heterodimers among diabetic patients and control subjects in five distinct racial groups. MATERIALS AND METHODS The subjects recruited comprised 49 diabetic and 59 control Japanese subjects; 40 diabetic and 83 control southern Chinese subjects; 149 diabetic and 101 control Caucasian subjects; 42 diabetic and 85 control North Indian Asian subjects; and 37 diabetic and 82 control Afro-Caribbean subjects. All diabetic patients were diagnosed according to strict clinical criteria. All were ketosis prone and either presented with frank ketoacidosis or severe symptoms of short duration. All were continuously dependent on insulin from diagnosis. Controls subjects were racially matched and had no history of type 1 diabetes. Patients and control subjects were HLA-DQ typed using sequence-specific oligonucleotide (SSO) gene probing [6-10]. The DQA 1- and DQB 1-typing results were used to determine the potential to encode DQ~ Arg52-DQ/~ non-Asp57 heterodimers. For example, an individual heterozygous for alleles DQAI*0501DQB 1"0201/DQA 1"0301-DQB 1"0302 has the potential to encode four different DQ~ Arg52-DQ/3 nonAsp57 heterodimers, two in cis and two in trans. An individual homozygous for DQAI*0301-DQBI*0302 also has the potential to encode four DQoe Arg52-DQ/3 non-Asp57 heterodimers. These four heterodimers would be identical molecules. No distinction was made between whether the heterodimers were different or identical. They were, however, identical with respect to DQoe Arg52-DQ/3 non-Asp57 status. To test the hypothesis that DQc~ Arg52-DQ/3 nonAsp57 heterodimers are directly involved in susceptibility to type 1 diabetes, the case where no Arg52-nonAsp57 heterodimer can be encoded was assigned the relative risk (RR) value of 1.0. The effect of encoding four, two, or one possible Arg52-non-Asp57 heterodi-

M.A. Penny et al.

mers compared with none was analyzed in diabetic and control subjects by the X2 test or Fisher's exact test where appropriate. RR values were calculated by the method of Woolf, with Haldane's modification for small numbers [11]. Tests of heterogeneity for the RR values obtained for four, two, and one heterodimers within each of the racial groups and between the racial groups were also calculated by the method of Woolf [11]. RESULTS There was no association between possession of four different versus four identical Arg52-non-Asp57 heterodimers and IDDM in any of the races. They segregated as follows: Japanese, none with four heterodimers; Chinese, six (60%) of ten different versus four (40%) of ten the same; white Caucasians, 56 (58%) versus 36 (36%) of 96; North Indian Asians, ten (40%) versus 15 (60%) of 25; and Afro-Caribbeans, two (13%) versus two (13%) of 15. No distinction was, therefore, made between whether the heterodimers were different or identical. They were, however, identical with respect to D Q a Arg52-DQ3 non-Asp57 status. The frequencies of four, two, one and no possible Arg52-non-Asp57 heterodimers among diabetic and control subjects in each of the five racial groups are listed in Table 1. The possibility of encoding four Arg52-non-Asp57 heterodimers compared with none was significantly associated with increased risk of type 1 diabetes in all races except the Japanese (Chinese, p < 10 5, RR = 15.4; Caucasians, p < 10 -6, RR = 52; North Indian Asians, p < 10 -6, RR = 73.7; and AfroCaribbeans, p < 10 -6, RR = 221). Encoding two possible Arg52-non-Asp57 heterodimers compared with none was also significantly associated with increased risk of type 1 diabetes in all races except the Japanese (Chinese, p < 0.025, RR = 3.0; Caucasians, p < 10 -6, RR = 9.2; North Indian Asians, p < 5 × 10 -4, RR = 10.7; and Afro-Caribbeans, p < 10 -4, RR = 9.4). The frequency of one Arg52-non-Asp57 heterodimer was not significantly associated with increased risk of the disease in any of the races studied. Figure 1 shows RR plotted against possession of four, two, one and no possible Arg52-non-Asp57 heterodimers in each of the races studied. Tests of heterogeneity showed that the RR values calculated for the potential to encode four and two heterodimers in each of the racial groups were significantly different from one another in all races except the Japanese (p < 10 -6 in all races). Tests of heterogeneity also showed that the RR values calculated for four, two, and one heterodimers in each of the racial groups were significantly different from one another in all races except the Japanese (Chi-

DQ Heterodimers and Susceptibility to Type 1 Diabetes

TABLE 1

181

Frequencies of HLA-DQ~ Arg 52-DQ/3 non-Asp 57 heterodimers possible in diabetic and control subjects in five distinct racial groups

Possible heterodimers

Diabetics n (%)

Controls n (%)

Japanese 4 2 1 0

N = 49 0 (0) 3 (6) 12 (25) 34 (69)

N = 59 1 (2) 2 (4) 13 (22) 43 (73)

NS NS NS

Southern Chinese 4 2 1 0

N = 40 10 (25) 13 (32.5) 4 (10) 13 (32.5)

N = 83 2 (2.4) 16 (19.3) 16 (19.3) 49 (59)

<10 5 <0.025 NS

15.4 (3.6-61) 3 (1.2-7.6) 1 (0.3-3.2) 1.0

White Caucasians 4 2 I 0

N = 149 96 (64.4) 40 (26.8) 5 (3.4) 8 (5.4)

N = 101 9 (8.9) 22 (21.8) 27 (26.7) 43 (42.6)

<10 6 < 1 0 -6 NS

52 (19-134) 9.2 (3.7-22) 1 (0.3-3.1) 1.0

N o r t h Indian Asians 4 2 1 0

N = 42 25 (59.5) 9 (21.4) 5 (11.9) 3 (7.1)

N = 85 4 (4.7) 11 (12.9) 25 (29.4) 45 (52.9)

<10 6 < 5 × 1 0 -4 NS

Afro-Caribbeans 4 2 1 0

N = 37 15 (40.5) 12 (32.4) 4 (10.8) 6 (16.2)

N = 82 0 (0) 9 (10.9) 27 (32.9) 46 (56.1)

< 1 0 -6 < 1 0 -4 NS

p

RR (95% CL)

0.4 (0.04-5) 1.8 (0.4-8.6) 1.2 (0.5-2.9) 1.0

73.7 ( 1 8 - 2 9 9 ) 10.7 (2.7-37) 2.8 (0.7-11) 1.0 221 ( 2 5 - 1 9 4 0 ) 9.4 (2.9-28) 1.2 (0.3-4.2) 1.0

p, p value uncorrected; NS, not significant; and RR (95% CL), relative risk (95% confidence limits).

nese, p < 0.005; Caucasian, p < 10-6; North Indian Asian, p < 10-3; and Afro-Caribbean, p < 10-4). In addition, the RR values calculated for two and four possible heterodimers were significantly different between the racial groups (two heterodimers, p < 0.025; and four heterodimers, p < 10-4). The difference in RR observed for two heterodimers became nonsignificant on removal of the Japanese data from the heterogeneity calculation. Removal of Japanese and Chinese data sets from the heterogeneity test on four heterodimers was necessary before the differences observed in RR across the races became nonsignificant. DISCUSSION Our aim was to test the hypothesis that HLA-DQ~ Arg52-DQ/~ non-Asp57 heterodimers contribute to susceptibility to type 1 diabetes. The direct contrast of the associations with Arg52-non-Asp57 heterodimers and type 1 diabetes in the Japanese compared with the other races and the heterogeneity of the RR values calculated for two and four heterodimers between the ra-

cial groups suggests that these heterodimers alone are unlikely to be critical in determining susceptibility to type 1 diabetes. A transracial approach to the investigation of MHC susceptibility to IDDM relies on the assumption that the disease is determined by the same mechanisms and has the same genetic basis in all races studied. This assumption is based on the following evidence. Diabetes was diagnosed according to the same strict clinical criteria in all subjects for this study. IDDM is a polygenic disease and at least one component of genetic susceptibility to the disease is MHC encoded; associations between IDDM and MHC alleles have been detected in all races studied [1]. It has been suggested that racial differences in MHC associations with IDDM may reflect disease heterogeneity [12]. Islet cell antibodies have been identified in Japanese diabetic subjects [13] and there is no evidence to suggest that type 1 diabetes is determined by different mechanisms in Oriental subjects compared with other races. It remains a possibility that the environmental factors that trigger IDDM are absent from or differ in Oriental populations. The inci-

182

M.A. Penny et al.

250

200

150

~

JAPANESE

m SOUTHERNCHINESE

>~

[ ~ WHITECAUCASIANS m NORTHINDIANASIANS

100

~

AFRO-CARIBBEANS

50 (

0

0

1

!

2

4

possiblenumberof Arg52-nonAsp57heterodimers

RR values plotted against number of Arg52nonAsp57 heterodimers in five distinct racial groups. Heterogeneity tests revealed significant differences in RR for two and four heterodimers between races (* p < 0.025, and ****p < 10-4). The difference in RR observed for two heterodimers became nonsignificant on removal of the Japanese data from the heterogeneity calculation. Removal of both the Japanese and Chinese data from the heterogeneity test on four hetrodimers was required for the observed differences in RR to become nonsignificant. The 95% CLs calculated for each of the RR values are given in parentheses at the top of each bar. FIGURE 1

dence of I D D M varies greatly in different racial groups [14]. Whether the differences in incidence of the disease are a result of environmental or genetic differences is unknown. The hypothesis that aspartate at position 57 of the DQ/3 chain is of critical importance in determining disease susceptibility [2] is refuted by studies in Japanese subjects [15] and by animal data [16]. The only common Arg52-positive DQA1 alleles in Caucasians are those that occur on DR3 and DR4 haplotypes [3]. The apparent disease associations with both Arg52 and nonAsp57 may therefore merely reflect the well-documented high risk o f disease associated with DR3/4 heterozygosity [ 1]. Indeed, the risk associated with DR3/4

heterozygosity is consistently greater than that with DR3/3 or DR4/4 [ 17], even though each of these genotypes should theoretically produce four A r g 5 2 - n o n Asp57 heterodimers. The proposed heterodimer model is, thus, too simple to explain the excess risk associated with DR3/4 heterozygosity. Combinations of D Q A 1 and D Q B 1 alleles have also been considered to confer resistance to IDDM: in particular the combinations D Q A I * 0 1 0 2 - D Q B I * 0 6 0 2 , DQAI*0103-DQBI*0603 or DQBI*0601, and D Q A 1*0501-DQB 1"0301. DQ-associated resistance to I D D M has been suggested to be partly dominant over DQ-associated susceptibility [ 18]. The protective effect of certain DQA1 and DQB1 alleles has not been considered in this investigation. An individual heterozygous at both the DQA1 and DQB1 loci can, theoretically, produce four distinct heterodimers. This assumes that D Q molecules encoded in both cis and trans are expressed. The hypothesis that susceptibility to type 1 diabetes correlates quantitatively with expression of "diabetogenic D Q heterodimers" also relies on the assumption that such molecules are expressed to the same extent and that each specificity of DQc~ and DQ/3 chains is equally likely to combine. Studies in murine class II transfectants show that molecules encoded in cis are expressed more efficiently than those encoded in trans [19]. It has also been demonstrated that certain combinations of DQc~ and DQ/3 chains are unable to form D Q molecules [20, 21]. Recent studies have suggested that the amino acid residues 6 0 - 9 1 of the DQ/~ chain may be important in determining stable cell-surface formation of certain preferential combinations of DQc~ and DQ/~ chains [22]. Further functional studies are required to validate these assumptions. It is thus unwise to draw firm conclusions on the effect of heterodimers on susceptibility to type 1 diabetes by using genotyping data alone [4]. In conclusion, these data do not support the hypothesis that susceptibility to type 1 diabetes directly relates to the frequency of HLA-DQc~ A r g 5 2 - D Q / 3 nonAsp57 heterodimers. The correlation observed in some races is probably due to the high frequency of DR3 and DR4 among diabetic subjects. Using heterodimer frequencies derived from D Q genotypes is a limited way of testing the hypothesis that quantitative expression of "diabetogenic heterodimers" determines susceptibility to type 1 diabetes. Functional studies into the expression of D Q molecules and their effect on T-cell function are required. ACKNOWLEDGMENTS

M.A.P. is funded by the West Midlands Regional Health Authority, C.H.M. by the Wellcome Trust, D.A.C. by the Medical Research Council, and K.H.J. by the British Diabetic Association. We gratefully acknowledge Dr. C. S. Cockram and

DQ Heterodimers and Susceptibility to Type 1 Diabetes

Dr. V. T. F. Yeung, the Chinese University of Hong Kong, for collection of the Chinese samples and Drs. Y. Uchigata, Y. Hirata, and T. Otani of the Tokyo Women's Medical College, Toyko, for collection of the Japanese samples. We thank Eli Lilly (UK), the Medical Research Council, the Wellcome Trust, and the British Diabetic Association for additional financial support.

REFERENCES 1. Jenkins D, Mijovic CH, Fletcher J, Jacobs KH, Bradwell AR, Barnett AH: Identification of susceptibility loci for type 1 (insulin-dependent) diabetes by trans-racial gene mapping, Diabetologia 33:387, 1990. 2. Dorman J, Laporte R, Stone R, Trucco M: Worldwide differences in the incidence of type 1 diabetes are associated with amino acid variation at position 57 of the HLADQ beta chain. Proc Natl Acad Sci USA 87:7330, 1991. 3. Khalil I, D'Auriol L, Gobet M, Morin L, Lepage V, Deschamps I, Park MS, Degos L, Galibert F, Hors J: A combination of HLA-DQ~ Asp 57 and HLA-DQce Arg 52 confers susceptibility to insulin-dependent diabetes mellitus. J Clin Invest 85:1315, 1990. 4. Gutierrez-Lopez MD, Bertera S, Chantres MT, Vavassori C, Dorman JS, Truuco M, Seranno-Rios M: Susceptibility to type 1 (insulin-dependent) diabetes mellitus in Spanish patients correlates quantitatively with expression of HLA-DQ~x Arg 52 and HLA-DQ/3 non-Asp 57 alleles. Diabetologia 35:583, 1992. 5. Baur MP, Neugebauer M, Albert ED: Reference tables of two-locus haplotype frequencies for all MHC marker loci. In Albert ED, Baur MP, Mayr WR (eds): Histocompatibility Testing 1984. Berlin, Springer, 1984. 6. Mijovic CH, Jenkins D, Jacobs KH, Penny MA, Fletcher J, Barnett AH: HLA-DQA1 and -DQB1 alleles are associated with genetic susceptibility to IDDM in a black population. Diabetes 40:748, 1991. 7. Jenkins D, Mijovic CH,Jacobs KH, Penny MA, Fletcher J, Barnett AH: Allele specific gene probing supports the DQ molecule as a determinant of inherited susceptibility to type 1 (insulin-dependent) diabetes. Diabetologia 34:109, 1991. 8. Jacobs KH, Jenkins D, Mijovic CH, Penny M, Uchigata Y, Cavan D, Hirata Y, Otani T, Fletcher J, Barnett AH: An investigation of Japanese subjects maps susceptibility to type 1 (insulin-dependent) diabetes mellitus close to the DAQ1 gene. Hum Immunol 33:24, 1992. 9. Penny MA, Jenkins D, Mijovic CH, Jacobs KH, Cavan DA, Yeung VTF, Cockram CS, Hawkins BR, Fletcher J, Barnett AH: Susceptibility to insulin-dependent diabetes mellitus in a Chinese population: the role of HLA class II alleles. Diabetes 41:914, 1992. 10. Cavan DA, Jacobs KH, Penny MA, Kelly MA, Mijovic CH, Jenkins D, Fletcher J, Barnett AH: Both DQA1 and DQB 1 alleles are implicated in HLA-associated protec-

183

tion from type 1 (insulin-dependent) diabetes mellitus in a British Caucasian population. Diabetologia 36:252, 1993. 11. Mathews JD: Statistical aspects of immunogenetic associations with disease. In Simons MJ, Tait BD (eds): Practical Methods in Clinical Immunology: Detection of Immune-Associated Genetic Markers of Human Disease, vol 7. Edinburgh, Churchill-Livingstone, 1984, p 82. 12. Giphart MJ, Roep BO, Drabbels J, D'Amaro J, Bruining J, Abdulkadir J, Verduyn W: Relative contribution of HLA-DQA and -DQB alleles to predisposition to insulindependent diabetes mellitus. Hum Immunol. 34:142, 1992. 13. Kobayashi T, Sugimoto T, Itoh T, Kosaka K, Tanaka T, Suwa S, Sato K, Tsuji K: The prevalence of islet cell antibodies in Japanese insulin-dependent and non-insulin-dependent diabetic patients studied by indirect immunofluorescence and by a new method. Diabetes 35: 335, 1986. 14. Diabetes Epidemiology Research International Group: Geographic patterns of childhood insulin-dependent diabetes mellitus. Diabetes 37:1113, 1988. 15. Awata T, Kuzuya T, Matsuda A, Iwamoto Y, Kanazawa Y, Okuyama M, Juli T: High frequency of aspartic acid at position 57 of HLA-DQ/3 chain in Japanese IDDM patients and nondiabetic subjects. Diabetes 39:266, 1990. 16. Miyazaki T, Uno M, Uehira M, Kikutani H, Kishimoto T, Kimoto M, Nishimoto H, Miyazaki J, Yamamura K: Direct evidence for the contribution of the unique I-AN°D to the development of insulitis in non-obese diabetic mice. Nature 345:722, 1990. 17. Thompson G, Robinson WP, Kuhner MK, Joe S, MacDonald MJ, Gottschall JL, Barbosa J, Rich S, Bertrams J, Buar MP, Partanen J, Tait BD, Schober E, Mayr WR, Ludvigsson J, Lindbolm B, Farid NR, Thompson C, Deschamps I: Modes of inheritance, and risk estimates for a joint study of Caucasians with insulin-dependent diabetes mellitus. Am J Hum Genet 43:799, 1988. 18. Thorsby E, Ronningen KS: Particular HLA-DQ molecules play a dominant role in determining susceptibility or resistance to type 1 (insulin-dependent) diabetes mellitus. Diabetologia 36:371, 1993. 19. Germain RN, Bentley DM, Quill H: Influence of allelk polymorphism on the assembly and surface expression of class II MHC (Ia) molecules. Cell 43:233, 1985. 20. Kwok WW, Mickelson E, Masewicz, Milner ECB, Hansen J, Nepom GT: Polymorphic DQce and DQ/3 interactions dictate HLA class II determinants of allo-recognition. J Exp Med. 171:85, 1990. 21. Nepom BS, Schwarz D Palmer JP, Nepom GT: Transcomplementation of HLA genes in IDDM. Diabetes 36:114, 1987. 22. Kwok WW, Kovats S, Thurtle P, Nepom GT: HLA-DQ allelic polymorphisms constrain patterns of heterodimer formation. J Immunol 150:2263, 1993.