2 Genes and environment

2 Genes and environment J A N I C E S. D O R M A N RONALD E. LAPORTE MASSIMO TRUCCO

One of the most important epidemiological contributions to the study of chronic diseases has been the documentation of the worldwide variation in incidence (Hutt and Burkitt, 1986). Insulin dependent diabetes mellitus (IDDM) is one such disorder for which the global patterns of incidence have been extremely well characterized. Through the establishment of standardized population based registries, it has become apparent that the incidence of IDDM varies dramatically across racial groups and countries (LaPorte et al, 1985; Diabetes Epidemiology Research International Group, 1988; Rewers et al, 1988a). The geographic difference in disease risk is more than 30-fold (Figure 1), with the age-adjusted rates being very high in the Scandinavian countries, such as Finland (28.6/100 000 per year), but very low in the

Q Q

o Ilc

ev

u~

Figure 1, Geographic differences in age-adjusted annual IDDM incidence rates. From LaPorte et al (1985).

229 Bailli~re's Clinical Endocrinology and Metabolism-Vol. 5, No. 2, June 1991 Copyright © 1991, by Bailli6re Tindall ISBN 0-7020-1490-7 All rights of reproduction in any form reserved

230

J. S. D O R M A N ET A L

Oriental populations, such as Korea and Japan (0.8/100000 per year) (Diabetes Epidemiology Research International Group, 1988). Among Caucasians there is also marked variability in the annual incidence of disease. Rates are much lower in parts of central and southern Europe, such as France (4.4/100000 per year), compared with northern European countries (Diabetes Epidemiology Research International Group, 1988; Rewers et al, 1988a). In addition, major racial differences in incidence within a population have also been reported; Caucasians consistently have higher rates of IDDM than Blacks from the same geographically defined area. With the development of IDDM incidence registries across the world (LaPorte et al, 1985; Diabetes Epidemiology Research International Group, 1988), population based resources are available for more in-depth studies of the epidemiology and aetiology of the disease. Comparative descriptive studies using IDDM registries have revealed the existence of similar epidemiological patterns in areas of high compared with low disease incidence (Tajima et al, 1985). These data suggest that the differences in IDDM incidence across populations are due to variation in the prevalence of common aetiological risk factors (Diabetes Epidemiology Research International Group, 1987), such as viruses (Banatvala et al, 1985; Yoon and Ray, 1985), nutrition (Borch-Johnsen et al, 1984; Helgason and Jonasson, 1985), or other life-style habits (LaPorte et al, 1981a; Robinson and Fuller, 1985). Major genetic variations also exist between racial groups and countries (Tiwari and Terasaki, 1985). A next step in the evolution of comparative IDDM registry research is to begin to test specific aetiological hypotheses regarding the causes of the international patterns of disease incidence (Dorman and LaPorte, 1990). In this chapter, we will outline an epidemiological approach for testing the hypothesis that the geographic differences in IDDM incidence are due to population variation in host susceptibility. This research model is based upon the assumption that any factor causally related to IDDM should not only explain the risk of disease within a population, but also account for differences between populations. The current example represents an interface between molecular biology and population based epidemiology, and has broad applications extending to investigations of host-environmental interactions in the aetiology of IDDM, as well as other autoimmune and chronic diseases. HLA AND DISEASE ASSOCIATIONS

The H L A region of the short arm of chromosome 6 contains genes known to be related to susceptibility to autoimmune diseases, including IDDM (Figure 2) (Trucco and Dorman, 1989). These genes encode class I (HLA-A, B, C), class II (HLA-DR, DQ, DP) and class Ill molecules, as well as tumour necrosis factor and complement. Class I and II molecules are highly polymorphic and consist of an a and a 13chain. The ~ chain of a class I molecule comprises three domains, and is non-covalently associated with 132-microglobulin, which is encoded by a gene located on chromosome 15

231

GENES AND ENVIRONMENT

CHROMOSOME 6

DP

f

",.

MHC

\

DQ

DR

f

B

C

"~ f

A

\

DPB2DPA2 DPB1 DPAI D Z A DOB DQB2 DQA.2 DQBI DQAI DRB1 DRB2 DRB3 DRA

Figure 2. The major histocompatibility complex in humans. From Trucco and Dorman (1989).

(Grey et al, 1987). Class I molecules are expressed on the surface of most nucleated cells and function in normal immune response by presenting processed antigens to cytotoxic T lymphocytes (Trucco and Dorman, 1989). The ~ and [3 chains of class II molecules, which are both encoded by genes within the H L A complex, each consist of two domains. Class II molecules also function immunologically by presenting antigens to helper T lymphocytes, but they are expressed primarily on B lymphocytes, macrophages and activated T cells. Crystallography studies have revealed that the outer domains of class I molecules form a cleft in which polypeptides from processed antigens can bind (Bjorkman et al, 1987). Class II molecules appear to have a structure similar to that of class I molecules (Brown et al, 1988). The shape of the cleft is determined by hypervariable regions which are particularly important in forming an antigen combining site for the molecule (Garrett et al, 1989). These critical hypervariable sequences can be recognized at the DNA level (Malissen et al, 1982), and are the basis for the highly polymorphic nature of the H L A system. They are responsible for the serological and cellular determination of distinct specificities at each locus. The H L A cleftpolypeptide complex is known to be important in T cell recognition and in determining an individual's T cell repertoire (Marrack and Kappler, 1987). Thus, the shape of the H L A molecule contributes to the ability of the immune system to respond to the presence of foreign (or self) antigens. From an epidemiological perspective, there are three unique features of the H L A system which make it an excellent marker for studies of host susceptibility: (1) the high degree of polymorphism at each locus, particu-

232

J. S. D O R M A N ET A L

larIy within the B and D regions; (2) linkage disequilibrium between alleles at various loci within the complex; and (3) geographic and racial variation in the frequencies of specific H L A antigens. The polymorphic nature of the H L A system has permitted the identification of genetic variation between individuals that appears to be important in determining susceptibility or resistance to specific diseases. Such variation has been detectable at the protein level through serological and cellular H L A typing (Tiwari and Terasaki, 1985). However, with the recent advances in molecular biology, it is apparent that there is an even greater degree of polymorphism in the H L A region than was evident just a few years ago. Using molecular techniques, such as the polymerase chain reaction (Trucco et al, 1989), it is now possible to define precisely these genetic differences between individuals at the D N A level. Linkage disequilibrium is another important characteristic of the H L A system for epidemiological research of host susceptibility. Under the assumption of Hardy-Weinberg equilibrium, the probability that two alleles from two loci exist on the same haplotype is the product of their individual gene frequencies (Tiwari and Terasaki, 1985). However, particular haplotypes are much more common in a given population than would be expected based upon the joint frequency for that combination. This phenomenon has been suggested as one potential explanation for H L A disease associations, with the reported population associations for some conditions being the result of linkage disequilibrium between H L A antigens and the 'true' disease susceptibility genes. Whether specific H L A antigens are directly involved in disease causation or, instead; serve as genetic markers for the condition, they enable researchers to identify 'susceptible' individuals for further aetiological studies of the disease. Finally, there are major geographic and racial variations in the frequency of particular H L A antigens across populations (Tiwari and Terasaki, 1985). HLA-DR3 is common among Caucasians, approximating 20% in Europe and North America, but is rare in American Indians (6%) and the Japanese (3%). HLA-DR4 is highly prevalent among the Japanese (41%) and American Indians (48%), less common in Caucasian populations (18-27%) and rare among American Blacks (10%). It is not known whether the distribution of particular antigens between racial groups and countries is the result of migration or natural selection. However, population variation in H L A antigen frequencies may have a major influence on the strength of disease associations within populations, and affect the geographic distribution of autoimmune diseases such as IDDM (Tiwari and Terasaki, 1985). This has provided a rationale for cross-cultural studies, as described in this review, to test epidemiological hypotheses that variations in genetic factors can account for differences in disease incidence within and across populations. IDDM susceptibility genes

Associations between H L A and IDDM began to be documented in the mid-1970s when it was observed that individuals with the disease were

GENES AND ENVIRONMENT

233

significantly more likely to be positive for HLA-B8 and B15 antigens than non-diabetic controls (Nerup et al, 1974; Cudworth and Woodrow, 1976). With rapid changes in H L A serology and the discovery of class II molecules, associations between antigens at the D R locus and IDDM became apparent. These studies revealed stronger relationships between IDDM and DR3 and/or DR4 than with the B locus antigens, which are in linkage disequilibrium with DR3 and DR4, respectively (Wolf et al, 1983; Bertram and Baur, 1984). It was concluded from these investigations that approximately 95% of IDDM cases had either DR3, DR4 or were DR3/DR4 heterozygotes; and that individuals with both DR3 and DR4 were particularly susceptible for developing the disease compared with those with neither high-risk antigen. With advances in molecular biology, studies of the associations between H L A and IDDM began to be conducted at the DNA level. Restriction fragment length polymorphism (RFLP) analyses revealed the existence of associations to the DQ locus, which appeared even stronger than those observed for H L A - D R (Schreuder et al, 1986; Owerbach et al, 1988). Studies of D N A sequences of the DQ o~and [3chain genes confirmed these results. In particular, it became apparent that the presence of D N A sequences coding for an amino acid other than aspartic acid in the 57th position of the DQ [3 chain (non-Asp 57) is highly associated with susceptibility to IDDM, whereas an aspartic acid in this position (Asp 57) appears to confer resistance to the disease (Todd et al, 1987; Morel et al, 1988). When aspartic acid is at position 57 of the DQ [3 chain molecules, rather than a non-charged amino acid (alanine, valine, serine), a salt bridge is formed with the arginine residue in position 79 of the DQ a chain (Figure 3) (Trucco and Dorman, 1989). Thus, the peptide binding ability of the cleft is affected, which in turn influences the recognition of the HLA-molecule complex by particular T cell clones. This type of modification in the groove may offer a logical explanation for the importance of specific amino acid sequences in determining susceptibility or resistance to certain autoimmune diseases such as IDDM. A study of 27 multiple case families identified from the IDDM registries in Allegheny County, PA was conducted to evaluate the relationship between IDDM and specific DQw alleles, defined at the DNA level, and to those determined serologically at the DR locus (Morel et al, 1988). Two of the non-Asp 57 genes (DQw2:33 vs. 18%; and DQw3.2:46 vs. 7%) were increased in frequency among the diabetic haplotypes, but the difference was statistically significant only for DQw3.2. Three of the Asp 57 DQw genes ( D Q w l . 2 : 0 vs. 25%; DQw3.1:6 vs. 18%; DQw3.3:0 vs. 15%) were significantly negatively associated with the disease. Pooling the data for all Asp 57 or non-Asp 57 alleles revealed that 94% of the diabetic haplotypes contained non-Asp 57, in comparison to 39% in haplotypes which were observed only among non-diabetic individuals ( P < 0.0001). Stratified analysis comparing the relative contributions of the D R and DQ locus to IDDM susceptibility indicated that the presence of non-Asp 57 was essential in determining susceptibility to IDDM when DR3 or DR4 was absent from the haplotype (Morel et al, 1988). Among the diabetic nonDR3, non-DR4 haplotypes, the prevalence of non-Asp 57 was 87.5%,

234

J. S. DORMAN ET AL

compared with only 32.4% among non-diabetic haplotypes without DR3 or DR4 (P = 0.001). To estimate the relative risks associated with the H L A - D R serological and D Q molecular markers, phenotype frequencies for the I D D M probands in these multiplex families were compared with those from unrelated nondiabetic controls (Morel et al, 1988). As illustrated in Table l, odds ratios for the high-risk serological markers (i.e. DR3/DR4) were consistent with previous reports in the literature (Wolf et al, 1983; Bertram and Baur, 1984). However, the associations between the DQ molecular markers and I D D M were extremely strong. Ninety-six per cent of the probands were

Amino Acid 57

J Non-Asp

Asp Salt Bridge

Figure 3. The possible role played by amino acid 57 of the HLA-DQ 13chain in IDDM susceptibility. From Trucco and Dorman (1989).

235

GENES AND ENVIRONMENT

Table 1. HLA-DR and DQ phenotype frequencies among probands from multiplex families and unrelated non-diabetic controls. Phenotype DR (Serology) DR3/DR4 D R3/DR3 DR3/DRX DR4/DR4 DR4/DRX DRX/DRX DQ (ASO probes) N/N N/A A/A

Diabetics (n = 27)

Non-diabetics (n = 123)

0.33 0.07 0.07 0.26 0.22 0.04

0.056 0.008 0.138 0.000 0.162 0.634

8.29* 9.76 0.05 -1.47 0.02

0.96 0.04 0.00

0.195 0.463 0.341

107.25* 0.04* 0.00"

Odds ratio

X, non-DR3 or non-DR4; N, non-Asp 57; A, Asp 57. * P<0.01. From Morel et al (1988). homozygous for non-Asp 57, as compared with 19.5% among the controls; with an odds ratio of 107 relative to those with at least one Asp 57 allele (Morel et al, 1988). These results confirm the importance of position 57 of the D Q [3 chain molecule in determining susceptibility or resistance to the disease.

Racial variation in IDDM susceptibility Although most studies of I D D M susceptibility have been conducted in Caucasian populations, there have been a n u m b e r of evaluations of the associations between H L A antigens and I D D M in other racial groups (Reitnauer et al, 1981; B e r t r a m and Baur, 1984; M a c D o n a l d et al, 1986; Hawkins et al, 1987). The results of these investigations were generally consistent with those observed for Caucasians, with a decrease in susceptibility for individuals with D R 2 (Reitnauer et al, 1981; B e r t r a m and Baur, 1984; M a c D o n a l d et al, 1986), and an increase in risk for those with H L A D R 3 (Reitnauer et al, 1981; M a c D o n a l d et al, 1986; Hawkins et al, 1987), D R 4 (Bertram and Baur, 1984; Reitnauer et al, 1981) or both D R 3 / D R 4 (Reitnauer et al, 1981) c o m p a r e d with those with alternative phenotypes. A m o n g the Blacks and Chinese, associations to D R 7 (Reitnauer et al, 1981) and/or D R 9 (Hawkins et al, 1987) were also reported. In the Japanese, D R 3 occurs with a very low frequency, but D R 4 and D R 9 were highly related to I D D M susceptibility (Bertram and Baur, 1984). Interestingly, heterozygous individuals i.e. D R 4 / D R 9 in the Japanese (Bertram and Baur, 1984), D R 3 / D R 9 in the Chinese (Hawkins et al, 1987) and D R 3 / D R 4 , D R 3 / D R 7 and D R 4 / D R 7 in the Blacks (Reitnauer et al, 1981) were at greatest risk of developing the disease relative to individuals from the same racial group who were homozygous for one, or had neither of the high-risk alleles. Because H L A - D R a p p e a r e d to be a consistent m a r k e r of I D D M susceptibility in all ethnic groups studied, researchers speculated that the racia!

236

J.S.

DORMAN ET AL

differences in IDDM incidence could be explained by population variation in the prevalence of high-risk serologically defined H L A antigens (Rotter and Hodge, 1980; Reitnauer et al, 1981; LaPorte et al, 1985, 1986; Tajima et al, 1985; MacDonald et al, 1986; Hawkins et al, 1987). However, the results of these investigations provided little support that the hypothesis was correct. For example, a comparative study between Allegheny County, PA and Japan (Tajima et al, 1985) found that although DR3 was very rare in Japan compared to the USA (3.2 vs. 22.2%), the prevalence of DR4 in Japan was higher than in the USA (41.1 vs. 27.3%). Thus, the proportion of 'susceptible' individuals, based on the prevalence of high-risk H L A - D R antigens, was virtually identical in these two racial groups. These data suggest that variation in the environment was a more likely explanation for the 20-fold difference in incidence between these two populations. Another ecological approach was employed, correlating the prevalence of HLA-DR2, DR3 and DR4 and IDDM incidence in 13 countries and racial groups (LaPorte et al, 1985). No significant associations were found between IDDM incidence and the frequency of these antigens across populations, either alone (DR2, r-- -0.1; DR3, r--0.16; DR4 r-- -0.36) or in combination (DR3/DR4, r - - - 0 . 1 2 ) . Despite the strong association between serological markers and IDDM within populations, there was little evidence that geographic variation in the distribution of H L A antigens was related to the global patterns of IDDM incidence. However, serologically defined H L A antigens may be imprecise markers of 'true' IDDM susceptibility. To investigate accurately the proposed hypothesis, susceptibility genes must be studied at the D N A level. Recently, molecular studies of the associations between the H L A - D R and DQ loci and IDDM have been conducted in several racial groups (Aparicio et al, 1988; Fletcher et al, 1988; Bao et al, 1989; Dunston et al, 1989; Ronningen et al, 1989; Todd et al, 1989, 1990; Yamagata et al, 1989; Awata et al, 1990; Dorman et al, 1990). The results of these investigations are consistent with earlier reports for Caucasians (Todd et al, 1987; Morel et al, 1988) and indicate that in Black and Oriental populations, IDDM susceptibility is conferred by the non-Asp 57 allele. However, the incomplete association between non-Asp 57 and IDDM suggests that other genetic and/or environmental factors also contribute to the development of the disease (Bao et al, 1989; Yamagata et al, 1989; Todd et al, 1989, 1990; Ronnfngen et al, 1989; Awata et al, 1990; Dorman et al, 1990). A study of Chinese cases and controls revealed a high prevalence of heterozygosity (72%) and Asp 57 homozygosity (22%) among the diabetics (Bao et al, 1989). Only 6% were non-Asp 57 homozygotes, indicating that the majority of Chinese diabetics were not at high genetic risk, as defined by non-Asp 57 homozygosity. Eleven of the 13 heterozygous Chinese patients were positive for the DQw3.1 (an Asp 57 allele), which has been considered 'neutral' in terms of its diabetogenic effect (Todd et al, 1987). Therefore, susceptibility to IDDM does not appear to be an 'all or none' phenomenon, but related to the number and type of non-Asp 57 alleles that an individual carries (Kwok et al, 1990). A high prevalence of Asp 57 has also been reported among Japanese with

GENES AND ENVIRONMENT

237

IDDM (Yamagata et al, 1989; Awata et al, 1990; Todd et al, 1990). Results of three independent studies were quite consistent, indicating that few diabetic cases (0-3%) were non-Asp 57 homozygotes, approximately 50% were heterozygous and the remaining 50% were homozygous for Asp 57. However, corresponding frequencies among Japanese non-diabetics varied considerably, with estimates of Asp 57 homozygosity ranging from 35 to 100% in the samples tested. Although methodological differences are a likely explanation for these inconsistencies, the low prevalence of non-Asp 57 homozygotes among the Oriental IDDM cases suggests that DQA1 or DRB1 alleles may also contribute to IDDM susceptibility. Further evaluation of the DQ region in the Japanese has revealed a strong association with the DQA1 allele A3 gene (Todd et al, 1990). This allele is in linkage disequilibrium with DR4 and DR9 in Caucasians and Blacks, and has been shown to be significantly associated with disease susceptibility (Todd et al, 1989). Caucasian DR9 haplotypes also encode an Asp 57 DQB1 allele, and they appear more neutral in their association to IDDM than black DR9 haplotypes, which contain DQw2 (non-Asp 57). The A3 allele is also present on Black, but not Caucasian, DR7 haplotypes both of which encode DQw2 (Todd e t al, 1989). Caucasian DR7 haplotypes contain the DQAI*0201 allele, which appears to be less diabetogenic. From these molecular studies, it is apparent that other amino acids of the HLA-DQ c~ and [3 chain molecules (Todd et al, 1987, 1990; Bao et al, 1989; Kwok et al, 1990), or specific combinations of DQ [3 and DR [3 sequences in the same haplotype (Erlich et al, 1990) are important in determining IDDM susceptibility. However, at the present time, the class of alleles defined by non-Asp 57 represent the strongest single marker of susceptibility to IDDM. Because of the methodological differences, including variations in laboratory procedures and the selection of IDDM cases and non-diabetic controls, it has been very difficult to generalize or compare results across studies. In the Japanese reports, case groups were small in size and not homogeneous, with a considerable proportion having an older age of IDDM onset (Awata et al, 1990). Controls were restricted to DR4 and/or DR9 positive individuals in one investigation (Yamagata et al, 1989), and the remaining studies reported large differences in DQ genotype frequencies, suggesting that neither was representative of the population at risk. To determine the association between non-Asp 57 or any genetic marker and disease, it is essential to evaluate accurately the distribution of IDDM susceptibility markers in the general population of the area. Thus, it is necessary to formally evaluate the relationship between non-Asp 57 and IDDM in a standardized manner in high, intermediate and low-risk areas where the incidence of disease is known. EPIDEMIOLOGICAL STUDIES OF HOST SUSCEPTIBILITY

To begin to investigate the association between non-Asp 57 and IDDM incidence across populations, IDDM cases and non-diabetic controls from China, the Black and White populations of Allegheny County, PA, Norway

238

J. S. DORMAN ET AL

and Sardinia were evaluated and compared (Dorman et al, 1990). The annual age-adjusted IDDM incidence rates for individuals less than 15 years of age in these populations were 0.7/100000 per year (Bao et al, 1989), 11.8/100000 per year (Diabetes Epidemiology Research International Group, 1988), 15.8/100 000 per year (LaPorte et al, 1981b), 19.7/100 000 per year (Diabetes Epidemiology Research International Group, 1988) and 24/100 000 per year (Diabetes Epidemiology Research International Group, 1990), respectively. For each population, comparative groups of randomly selected IDDM cases who were on insulin therapy and less than 15 years of age at the time of first insulin administration, as well as unrelated nondiabetic controls from the area were studied. The HLA-DQ genotype distribution for the Chinese (Bao et al, 1989) and the Norwegian samples (Ronningen et al, 1989) were previously published. Table 2. H L A - D Q genotype and n o n - A s p 57 gene frequencies a m o n g diabetics and nondiabetics controls in five populations. Diabetics

Non-diabetics

Country

n

N/N

N/A

A/A

N

n

N/N

N/A

A/A

N

Sardinia Norway US Whites US Blacks China

30 52 49 30 18

1.00 0.81 0.61 0.73 0.06

0.00 0.16 0.39 0.27 0.72

0.00 0.04 0.00 0.00 0.22

1.00 0.89 0.81 0.87 0.42

60 187 123 51 25

0.38 0.27 0.20 0.18 0.00

0.47 0.51 0.46 0.37 0.08

0.15 0.22 0.34 0.45 0.92

0.62 0.53 0.43 0.36 0.04

N, n o n - A s p 57; A, A s p 57. From D o r m a n et al (1990).

As demonstrated in Table 2, the HLA-DQ t3 genotype distributions among the IDDM case groups were significantly different ( P < 0.001) across populations, with an increase in non-Asp 57 homozygosity and a decrease in heterozygosity and Asp 57 homozygosity among IDDM cases in high incidence areas (Dorman et al, 1990). If non-Asp 57 is operative in all populations, then the observed distribution of alleles in the diabetic patients should reflect that for the corresponding control groups. Thus, one would expect to find a higher prevalence of non-Asp 57 homozygotes among non-diabetics from high risk areas, and an increase in heterozygotes and Asp 57 homozygotes in the low risk populations. The genotype distributions among the non-diabetic control populations also differed significantly (P<0.001) as expected (Table 2), with a high prevalence of non-Asp 57 homozygotes in Sardinia and Norway compared with the proportions found in the Blacks and Chinese. These results indicate that there is extraordinary genetic variation with respect to the prevalence of IDDM susceptibility genes across populations; with a 15-fold increase in the frequency of nonAsp 57 in Sardinia compared to China. An international case control study of host susceptibility is now being formalized as part of the WHO Multinational Study of Childhood Diabetes to test the hypothesis in high, moderate and low-risk populations from various racial groups (Dorman et al, 1990). This investigation will be based

GENES AND ENVIRONMENT

239

upon standardized epidemiological methods for the selection of: (1) comparable samples of I D D M cases from population based registries in areas where the incidence of disease has been established; and (2) representative samples of non-diabetic controls from the general populations at risk. The case control methodology is a very powerful epidemiological design for hypothesis-testing research. In addition to investigating the association between non-Asp 57 and I D D M , this international effort will provide the framework for evaluations of other potential aetiological factors and hostenvironmental interactions to disease incidence within populations and across the world. Assessment of genotype-specific IDDM incidence rates A major advantage of using molecular epidemiological studies for aetiological research is the possibility of being able to link the relative risk estimates from the case control analyses to the absolute risks of developing the disease (Dorman et al, 1990). For example, the overall incidence of I D D M in a given population can be expressed as a weighted average of the genotype-specific incidence rates, where the weights are the proportions of the general population exhibiting each genotype (Breslow and Day, 1980). These weights can be calculated from the genotype distributions for the controls. The genotype-specific incidence rates are then estimated using odds ratios to establish the relationships between the incidence rates for each genotype, and represent the absolute risk for developing the disease in genetically homogeneous groups (Dorman et al, 1990). To illustrate this approach, data from the comparative study of five populations was employed (Dorman et al, 1990). If a specific allele, such as non-Asp 57, were a strong marker of I D D M susceptibility in all populations studied, then one would also expect that the odds ratios associated with the presence of one or more non-Asp 57 alleles to be consistently elevated. As shown in Table 3, the odds ratios for I D D M associated with non-Asp 57 homozygosity relative to Asp 57 homozygosity were significantly increased in all populations, ranging from 14 to 111. The exact test for homogeneity was non-significant (P = 0.26). The odds ratios for heterozygous individuals Table 3. Odds ratios for IDDM associatedwith nonAsp 57 homozygosityand heterozygosityrelative to Asp 57 homozygosity. Country Sardinia Norway US White US Blacks China

Odds ratio (N/N vs. A/A) 24.7* 13.7" 105.8" 111.3" 15.7t

Odds ratio (N/Avs. A/A) NE 1.5 NS 28.8* 20.5* 28.2*

N, non-Asp 57; A, Asp 57. *P<0.0001; +P<0.05; NS, not significant, P>0.05; NE, not estimated. From Dorman et al (1990).

240

J.S. DORMAN ET AL

compared with Asp 57 homozygotes were statistically significant for Allegheny County Whites, Blacks and the Chinese, and non-significantly elevated for the Norwegians. None of the IDDM cases from Sardinia were heterozygous, so the odds ratio was not estimated. There was evidence for heterogeneity of these odds ratios (P = 0.03), primarily due to Norway. Using these data, genotype-specific IDDM incidence rates were estimated for Allegheny County Caucasians, as described. The risk was highest for non-Asp 57 homozygotes (47.6/100 000 per year), intermediate for heterozygous individuals (13.0/1.00 000 per year), and lowest for Asp 57 homozygotes (0.45/100 000 per year) (Dorman et al, 1990). These findings are suggestive of a dose-response relationship to IDDM susceptibility and further indicate that this genetic marker appears to play a critical r01e in the aetiology of the disease. If the geographic differences in disease incidence were due to variations in host susceptibility, then one would predict that the genotype-specific incidence rates would be similar across populations. Because the statistical properties of the genotype-specific rates are currently under investigation, this issue was addressed indirectly by applying the point estimates for Allegheny County Caucasians to the other four populations for which preliminary data were obtained to predict the overall IDDM incidence rate for each area. As shown in Figure 4, each of the predicted rates fell within

m predicted

30-

O

observed Sardinia

+

o oo 20o

Norway

O

i

0A

@ US Whites US Blacks

~ 10-

•

China I

0.1

I

I

I

I

I

I

0.2 0.3 0.4 0.5 0.6 0.7 Gene Frequency of n0n-Asp-57

I

0.8

Figure 4. Association between actual and predicted age-adjusted annual IDDM incidence rates. From Dorman et al (1990).

GENES AND ENVIRONMENT

241

the 95% confidence intervals for the rate established through IDDM registries (Dorman et al, 1990). In the future, it will be possible to directly evaluate the geographic variation in IDDM incidence among genetically homogeneous subgroups. In this manner, one can determine whether susceptible individuals from China, as defined by non-Asp 57 homozygosity, have a similar risk of developing IDDM as a Finnish child with the same genotype. The incidence of disease among less susceptible individuals can also be contrasted across racial groups and countries. Through these comparative analyses, it will be possible to quantify the contribution of non-Asp 57 to the global Patterns of IDDM incidence and begin to investigate other potential aetiological factors. Host-environmental interactions in IDDM incidence

Although it appears that population differences in the frequency of non-Asp 57 may be responsible for the geographic patterns of IDDM incidence, environmental and other genetic factors clearly play a role in the aetiology of the disease. Many non-Asp 57 homozygous individuals fail to develop IDDM (Dorman et al, 1990), and the diabetes concordance rate among monozygotic twins is typically low, ranging from 30 to 50% (Barnett et al, 1981). Thus, non-Asp 57 provides only a partial explanation for the occurrence of disease. One strategy for investigating the role of other aetiological factors within and across populations is to first control for host susceptibility (Dorman and LaPorte, 1990). The effects of other genetic and environmental factors would be best evaluated after the strong influence of non-Asp 57 had been accounted for. This can be achieved by designing studies that focus on high-risk individuals (i.e. non-Asp 57 homozygotes) or genetically related populations, as described below. Despite the 30-fold difference in IDDM incidence worldwide, preliminary data analyses suggest that genotype-specific incidence rates will be less variable when compared across racial groups and countries. Geographic differences in genotype-specific IDDM incidence rates are likely to be the result of population variation in the environment or other IDDM susceptibility genes. Although the epidemiological approach outlined in this review was developed to test a genetic hypothesis within and across populations, it can easily be applied to genetically homogeneous subgroups to investigate other aetiological determinants of the disease. Future international case control studies could focus on genetically susceptible individuals (i.e. nonAsp 57 homozygous) or high-risk groups for evaluations of viruses, nutritional factors, etc., to determine their associations to the disease within geographically defined areas or racial groups. Variation in the prevalence of these factors can then be assessed across populations, as previously described. However, the focus of these analyses would be the contribution of these factors to geographic differences in genotype-specific risk, rather than to the overall global patterns of disease incidence. By evaluating only those at high genetic risk, it is likely that the effects of environmental determinants of the disease, which were difficult to assess previously, will become apparent.

242

J.s.

D O R M A N ET A L

Another approach to investigating potential diabetogenic risk factors while controlling for host susceptibility is to study ethnically-related populations in areas which differ in IDDM incidence. A study from Montreal, Canada revealed that age-adjusted IDDM incidence rates among French Canadian (8.2/100 000 per year) and Jewish children (17.2/100 000 per year) living in Montreal were approximately twice as high as those reported from France (3.7/100000 per year) and Israel (6.8/100000 per }'ear), respectively (Siemiatycki et al, 1988). Although one cannot rule out a significant contribution of genetic admixture, the common ethnic heritage within various subgroups suggests that changes in environmental exposure may be contributing to the observed increase in IDDM incidence among the French and Jewish migrants. By conducting heritage studies of populations which have a common ethnic background (Diabetes Epidemiology Research International Group, 1989), one can more easily control for the influence of genetic factors and evaluate potential environmental determinants of the disease. Other natural experiments which provide an opportunity to evaluate the influence of aetiological factors in genetically homogeneous populations are the reported epidemics of IDDM (Akerblom and Reunanen, 1985; Rewers et al, 1988b). These rapid temporal changes in risk are almost certainly related to sudden variations in the prevalence of potential environmental determinants of the disease, and are associated with approximately a 2- to 3-fold increase in IDDM incidence over time. Through disease surveillance or retrospective studies of the potential causes of reported epidemics, one can control for host susceptibility by design, and accurately investigate the contributions of environmental agents, such as viruses, nutritional factors, etc., to the temporal patterns of disease incidence.

SUMMARY

Many other autoimmune and chronic diseases exhibit marked geographic variation in incidence, which has been attributed to environmental differences across populations (Hutt and Burkitt, 1986). The results of our international IDDM research have provided evidence for the importance of large genetic variations in the frequency of HLA susceptibility genes between racial groups and countries. One may speculate that differences in the prevalence of susceptibility genes for other chronic diseases exist and significantly contribute to the geographic patterns of incidence of these disorders. Other autoimmune diseases are known to have epidemiological features similar to those described for IDDM. Although they are also characterized by an underlying HLA-related susceptibility, environmental factors are known to play an important aetiological role (Tiwari and Terasaki, 1985). DNA polymorphisms of the DR, DO and DP locus antigens are associated with various autoimmune diseases (Todd et al, 1988; Thorsby et al, 1989). These molecular variations are similar to those described for IDDM, in that they are typically related to the hypervariable regions of the molecule and,

GENES AND ENVIRONMENT

243

thus, affect the p e p t i d e b i n d i n g ability of the antigen. B a s e d o n the e v i d e n c e for I D D M , p o p u l a t i o n differences in the f r e q u e n c y of o t h e r H L A susceptibility genes are likely to be m a j o r d e t e r m i n a n t s of the geographic distrib u t i o n of diseases such as r h e u m a t o i d arthritis a n d m u l t i p l e sclerosis. T h e e p i d e m i o l o g i c a l a p p r o a c h o u t l i n e d in this review is, thus, applicable to o t h e r a u t o i m m u n e diseases a n d will significantly c o n t r i b u t e to o u r k n o w l e d g e of the aetiology of these disorders. T h e e m e r g i n g field of m o l e c u l a r e p i d e m i o l o g y r e p r e s e n t s a n e w research a p p r o a c h which will lead to a b e t t e r u n d e r s t a n d i n g of the relationships b e t w e e n specific risk factors a n d the aetiology of c h r o n i c diseases within p o p u l a t i o n s a n d across the world.

Acknowledgements We thank Drs Lewis Kuller, Allan Drash and Mark Sperling for their thoughtful comments, the members of the Diabetes Research Center in Pittsburgh, the Diabetes Epidemiology Research International Study Group and members of the WHO Multinational Study of Childhood Diabetes (the DIAbetes MONDiale study) for their collaboration and support. This work was supported by NIH Grants RO1 DK24021, RO1 DK39125, RO1 AI23963 and the American Diabetes Association.

REFERENCES Akerblom HK & Reunanen A (1985) The epidemiology of insulin-dependent diabetes mellitus (IDDM) in Finland and Northern Europe. Diabetes Care 8 (supplement 1): 10-16. Aparicio JMR, Wakisaka A, Takada A, Matsuura N & Aizawa M (1988) HLA-DQ system and insulin-dependentdiabetes mellitus in Japanese: does it contribute to the development of IDDM as it does in Caucasians? Immunogenetics 38" 240-246. Awata T, Kuzuya T, Matsuda A et al (1990) High frequency of aspartic acid at position 57 of HLA-DQ beta chain in Japanese IDDM patients and non-diabetic subjects. Diabetes 39: 266-269. Banatvala JE, Schernthaner G, Schober E et al (1985) Coxsackie B, mumps, rubella, and cytomegalovirus specific IgM responses in patients with juvenile-onset insulin-dependent diabetes mellitus in Britain, Austria and Australia. Lancet i: 1409-1411. Bao MZ, Wang JX, Dorman JS & Trucco M (1989) HLA-DQ beta non-Asp-57 allele and incidence of diabetes in China and the USA. Lancet ii: 497-498. Barnett AH, Eft C, Leslie RDG & Pyke DA (1981) Diabetes in identical twins: a study of 200 pairs. Diabetologia 20: 87-93. Bertram J & Baur M (1984) Insulin-dependentdiabetes mellitus. In Histocompatibility Testing 1984, pp 348-368. Heidelberg: Springer-Verlag. Bjorkman P, Saper M, Samraoui W, Bennett W, Strominger J & Wiley D (1987) Structure of the human class I histocompatibility antigen, HLA-A2. Nature 329: 512-517. Borch-Johnsen K, Mandrup-Poulsen T, Zachau-Christiansen Bet al (1984) Relation between breast-feeding and incidence rates of insulin-dependent diabetes mellitus. Lancet ii: 1083-1086. Breslow NE & Day NE (1980) Statistical Methods in Case-Control Studies. Lyon: International Agency for Research on Cancer. Brown J, Jardetzky T, Saper M, Samraoui B, Bjorkman P & Wiley D (1988) A hypothetical model of the foreign antigen binding site of class II histocompatibility molecules. Nature 332: 845-851. Cudworth AG & Woodrow JC (1976) Genetic susceptibility in diabetes mellitus: analyses of the HLA association. British Medical Journal 2: 846-848. Diabetes Epidemiology Research International Group (1987) Preventing insulin-dependent diabetes mellitus: the environmentalchallenge. British Medical Journal 295: 479-481.

244

J . s . DORMAN ET AL

Diabetes Epidemiology Research International Group (1988) Geographic patterns of childhood insulin-dependent diabetes mellitus. Diabetes 37: 1113-1119. Diabetes Epidemiology Research International Group (1989) Evaluation of epidemiology and immunogenetics of IDDM in Spanish and Portuguese-heritage registries: a key to understanding the etiology of IDDM? Diabetes Care 12: 487-493. Diabetes Epidemiology Research International Group (1990) The epidemiology and immunogenetic in Italian-heritage populations. Diabetes Metabolism Reviews 6: 63-69. Dorman JS & LaPorte RE (1990) IDDM epidemiology: next generation of research. Diabetes Care 13:184--185 (letter). Dorman JS, LaPorte RE, Stone RA & Trucco M (1990) Worldwide differences in the incidence of Type 1 diabetes are associated with amino acid variation at position 57 of the HLA-DQ beta chain. Proceedings of the National Academy of Sciences, USA 87: 7370-7374. Dunston GM, Henry LW, Christian J, Ofosu MD & Callender CO (1989) HLA-DR3, DQ heterogeneity in American Blacks is associated with susceptibility and resistance to insulin-dependent diabetes mellitus. Transplantation Proceedings 21: 653-655. Erlich HA, Bugawan TL, Scharf S, Nepom GT, Tait B & Griffith RL (1990) HLA-DQ beta sequence polymorphism and genetic susceptibility to IDDM. Diabetes 36: 96-103. Fletcher J, Mijovic C, Odugbesan O, Jenkins D, Bradwell AR & Barnett AH (1988) Transracial studies implicate HLA-DQ as a component of genetic susceptibility to Type 1 (insulin-dependent) diabetes. Diabetologia 31: 864-870. Garrett TP, Saper MA, Bjorkman P J, Strominger JI & Wiley DC (1989) Specificity pockets for the side chains of peptide antigens in HLA-Aw68. Nature 342: 692-696. Grey A, Kubo R, Colon S e t al (1987) The small subunit of HLA antigens is beta 2 microglobulin. Journal of Experimental Medicine 138: 1608-1616. Hawkins BR, Lam KS, Ma JT et al (1987) Strong associations of HLA-DR3/DRw9 heterozygosity with early-onset insulin-dependent diabetes mellitus in Chinese. Diabetes 36: 1297-1300. Helgason T & Jonasson MR (1985) Evidence for a food additive as a cause of ketosis-pronc diabetes. Lancet ii: 716-720. Hutt MSR & Burkitt DP (1986) The Geography of Non-Infectious Disease, pp 1-6. Oxford: Oxford University Press. Kwok WW, Mickelson E, Masewicz S, Milner ECB, Hansen J & Nepom GT (1990) Polymorphic DQ alpha and DQ beta interactions dictate HLA Class II determinants of allo-recognition. Journal of Experimental Medicine 171: 85-95. LaPorte RE, Orchard TJ, Kuller LH et al (1981a) The Pittsburgh insulin-dependent diabetes mellitus registry: the relationship of insulin-dependent diabetes mellitus incidence to social class. American Journal of Epidemiology 114: 379-384. LaPorte RE, Fishbein HA, Drash AL et al (1981b) The Pittsburgh IDDM registry: the incidence of IDDM in Allegheny County, PA (1965-1976). Diabetes 30: 279-284. LaPorte RE, Tajima N, Akerblom HK et al (1985) Geographic differences in the risk of insulin-dependent diabetes mellitus: the importance of registries. Diabetes Care 8 (supplement 1): 101-107. LaPorte RE, Tajima N, Dorman JS et al (1986) Differences between blacks and whites in the epidemiology of insulin-dependent diabetes mellitus in Allegheny County, PA. American Journal of Epidemiology 123: 592-603. MacDonald M J, Famuyiwa O, Nwabueo IA et al (1986) HLA-DR associations in black Type I diabetics in Nigeria: further support for models of inheritance. Diabetes 35: 583-589. Malissen M, Malissen B & Jordan BR (1982) Exon/intron organization and complete nucleotide sequence of an HLA gene. Proceedings of the National Academy of Sciences of the USA 79: 506-510. Marrack P & Kappler J (1987) The T cell receptor. Science 238: 1073-1079. Morel PA, Dorman JS, Todd JA, McDevitt HO & Trucco M (1988) Aspartic acid at position 57 of the HLA-DQ beta chain protects against Type I diabetes. Proceedings of the National Academy of Sciences of the USA 85: 8111-8116. Nerup J, Platz P, Anderson O et al (1974) HLA antigens and diabetes mellitus. Lancet ii: 864-867. Owerbach D, Gunn S, Ty G, Wible L & Gabbay KH (1988)Oligonucleotide probes for HLA-DQA and DQB genes define susceptibility to type 1 (insulin-dependent) diabetes mellitus. Diabetologia 31: 751-757.

GENES AND ENVIRONMENT

245

Reitnauer PJ, Roseman JM, Barger BO, Murphy CC, Kirk KA & Acton RT (1981) HLA associations with insulin-dependent diabetes mellitus in a sample of the American black population. Tissue Antigens 17: 286-293. Rewers M, LaPorte RE, King HOM & Tuomilehto J (1988a) Insulin-dependent diabetes mellitus in childhood: international patterns and trends. World Health Statistics Quarterly 41: 171-190. Rewers M, LaPorte RE, Walczak M, Dmochowski K & Bogaczynska E (1988b) Apparent epidemic of insulin-dependent diabetes mellitus in Midwestern Poland. Diabetes 37: 1113-1119. Robinson N & Fuller JH (1985) Role of life events and difficulties in the onset of diabetes mellitus. Journal of Psychosomatic Research 29: 583-591. Ronningen LS, Halstensen TS, Spurkland A & Thorsby E (1989) The amino acid at position 57 of the HLA-DQ beta chain and susceptibility to develop insulin-dependent diabetes mellitus. Human Immunology 26: 215-225. Rotter JI & Hodge SE (1980) Racial differences in juvenile-type diabetes are consistent with more than one mode of inheritance. Diabetes 29: 115-118. Schreuder G, Tilanus M, Bontrop R et al (1986) HLA-DQ polymorphism associated with resistance to Type I diabetes detected with monoclonal antibodies, isoelectric point differences, and restriction length fragment polymorphism. Journal of Experimental Medicine 164: 138-150. Siemiatycki J, Colle E, Campbell S, Dewar R, Aubert D & Belmonte MM (1988) Incidence of IDDM in Montreal by ethnic group and by social class and comparisons with ethnic groups living elsewl'mre. Diabetes 37: 1096-1102. Tajima N, LaPorte RE, Hibi I, Kitagawa T, Fujita H & Drash AL (1985) A comparison of the epidemiology of youth-onset insulin-dependent diabetes mellitus between Japan and the United States (Allegheny County, PA). Diabetes Care 8 (supplement 1): 17-23. Thorsby E, Lundin KEA, Ronningen KS, Sollid LM & Vartdal F (1989) Molecular basis and functional importance of some disease-associated HLA polymorphisms. Tissue Antigens 34: 39-49. Tiwari JL & Terasaki PI (1985) HLA and Disease Associations. New York: Springer-Verlag. Todd JA, Bell JL & McDevitt HO (1987) HLA-DQ beta gene contributes to susceptibility and resistance to insulin-dependent diabetes mellitus. Nature 329: 559-604. Todd JA, Acha-Orbea H, Bell JI et al (1988) A molecular basis for MHC class II-associated autoimmunity. Science 240: 1003-1009. Todd JA, Mijovic C, Fletcher J, Jenkins D, BradweU AR & Barnett AH (1989) Identification of susceptibility loci for insulin-dependent diabetes mellitus by trans-racial gene mapping. Nature 338: 587-589. Todd JA, Fukui Y, Kitagawa T & Sasazuki T (1990) The A3 allele of the HLA-DQAI locus is associated with susceptibility to Type 1 diabetes in Japanese. Proceedings of the National Academy of Sciences of the USA 87: 1094-1099. Trucco M & Dorman JS (1989) Immunogenetics of insulin-dependent diabetes mellitus in humans. Critical Reviews in Immunology 9: 201-245. Trucco G, Fritsch R, Giorda R & Trucco M (1989) Rapid detection of IDDM susceptibility using HLA-DQ beta alleles as markers. Diabetes 38: 1617-1622. Wolf E, Spencer KM & Cudworth A G (1983) The genetic susceptibility to Type I (insulindependent) diabetes: analyses of the HLA-DR associations. D iabetologia 23: 224. Yamagata K, Nakajima H, Hanafusa T et al (1989) Aspartic acid at position 57 of the DQ B chain does not protect against Type I (insulin-dependent) diabetes mellitus in Japanese subjects. Diabetologia 32: 762-764. Yoon JW & Ray U R (1985) Perspectives on the role of viruses in insulin-dependent diabetes. Diabetes Care 8 (supplement 1): 39-44.