The Rising Incidence of Type 1 Diabetes Is Accounted for by Cases With Lower-Risk Human Leukocyte Antigen Genotypes

Subjects with type 1 diabetes were from the Australian Type 1 Diabetes DNA Repository (11,12). They had been diagnosed by a specialist physician on the basis of hyperglycemia requiring immediate and ongoing insulin therapy. All subjects were of European descent and lived in metropolitan Melbourne or the wider state of Victoria. They were recruited in an unselected manner from diabetes clinics, including the Royal Children's Hospital, which until recently followed up the majority of children and teenagers with type 1 diabetes in Victoria. Virtually all individuals with type 1 diabetes in Victoria attend diabetes clinics or are seen by specialists with attachments to these clinics. Subjects (n = 462) diagnosed before 18 years of age between 1950 and 2005 (91% after 1975) were typed for HLA-DRB1 alleles.

HLA-DRB1 genotyping was performed by PCR-based sequence-specific oligonucleotide hybridization, modified from the 11th International Histocompatibility Workshop protocol to accommodate subsequently described sequence polymorphisms (13). Briefly, PCR primers were designed to provide either generic or allele-specific amplification of exon 2 of the DRB1 allele. The PCR-amplified DNA was rendered single stranded by exposure to sodium hydroxide, immobilized onto nylon membranes, and hybridized with biotinylated oligonucleotide probes designed to detect sequence polymorphisms between DRB1 alleles. Stringent washing was performed in the presence of 3 mol/l tetramethylammonium chloride and the hybridized product detected using a streptavidin/alkaline phosphatase conjugate and chemiluminescent substrate CDP-star (Roche Diagnostics, Castle Hill, NSW, Australia). Allele assignments were made by comparing patterns of hybridization to those predicted from published sequences.

Subjects were grouped by decade of birth or diagnosis (1950–1969, combined due to small numbers; 1970–1979; 1980–1989; 1990–1999; and 2000–2005) and from the highest (DR3,4) to intermediate (DR3,3 or DR4,4; DR4,X or DR3,X) to the lowest (DRX,X, where X denotes a non-3 or non-4 allele) HLA genotypes for type 1 diabetes risk. Genotypes between decade groups were compared by the χ² test for trend. Mean age at diagnosis across decades was analyzed by ANOVA. Statistics were performed with GraphPad Prism version 3.0 software (La Jolla, CA).

The mean age at diagnosis of the total cohort of type 1 diabetes subjects was (means ± SD) 8.5 ± 4.5 years, and this did not differ across decades of diagnosis (P = 0.54, ANOVA). DRB1 genotypes were first analyzed according to decade of diagnosis. The proportion of the highest-risk genotype, DR3,4, decreased significantly over time from 79% in 1950–1969 to 28% in 2000–2005 (P < 0.0001) (Table 1) (Fig. 1). On the other hand, the proportion of the heterozygous intermediate-risk genotypes, DR4,X and DR3,X, increased significantly over this period (Table 1); taken together, they increased from 20 to 48% (P = 0.0002) (Fig. 1). Homozygosity for DR4, but not DR3, increased over time, and the lowest-risk genotype, DRX,X, was consistently low (≤3%) (Table 1).

Matched for DRB1 genotype and decade of birth, subjects were analyzed for age at diagnosis. There was no change across time in age at diagnosis for the high-risk DR3,4 genotype, but for the intermediate risk genotypes (DR4,4; DR3,3; DR4,X; and DR3,X) age at diagnosis decreased over time (Fig. 2).

The incidence of childhood-onset type 1 diabetes in Australia has doubled in the last 20 years, from 11.3 cases per 100,000 person-years in 1985 to 23.2 in 2002 (14). On the assumption that the DRB1 genotype frequencies were representative of the total population of individuals with type 1 diabetes, the contribution of DRB1 genotypes to the increasing incidence of type 1 diabetes was examined. The numbers of incident cases (rev. in 14) were related to specific DRB1 genotype proportions. As shown in Table 2, the population incidence of type 1 diabetes in subjects with the highest-risk genotype (DR3,4) has remained unchanged. In contrast, the number of cases with intermediate-risk genotypes (DR3,3 or DR4,4 and DR3,X or DR4,X) has increased, accounting for the increase in disease incidence.

HLA class II profiling demonstrates that the genetic contribution in individuals diagnosed with type 1 diabetes has changed over the last five decades. The contemporary increase in disease incidence is accounted for by individuals with lower-risk HLA class II genotypes who, in previous eras, would not have developed diabetes in childhood. It would appear, therefore, that changing environmental conditions have increased the penetrance of these lower-risk genotypes. Our findings are similar to those reported in the northern hemisphere (6,7) but are more likely to be free of selection bias based on survival, as 91% of our subjects were diagnosed after 1975. Subjects were unselected and predominantly of European Anglo-Celtic origin, with no change over the decades to suggest dilution by immigrants of other ethnic backgrounds. Nevertheless, the possibility that they were not entirely representative of the total population of individuals with type 1 diabetes across the decades remains a potential limitation. With this caveat, environmental conditions operating in both hemispheres therefore appear to be permissive for the increasing incidence of type 1 diabetes among individuals without the classic high-risk DR3,4 genotype. Furthermore, the mean age at diagnosis has decreased in children with lower-risk genotypes, representing further evidence for the impact of environment.

The temporal change in the genetic contribution to type 1 diabetes raises several points for discussion. First, studies of type 1 diabetes may need to consider stratification by year of birth and diagnosis to dissect the relative influences of genes and environment. This concept is also more broadly relevant to other human diseases in which polygenetic susceptibility and environment interact. Second, the HLA profile of classic, childhood-onset type 1 diabetes has shifted and is now similar to that of adults with autoimmune diabetes, who are characterized as having lower-risk HLA genes (15,16). However, adults diagnosed today with autoimmune diabetes were born >30 years ago when the contribution of environment was presumably much less. Born today, would they reach adulthood before developing type 1 diabetes? Third, it appears that the shift in the distribution of HLA genotypes began in the 1980s. A variety of environmental factors including infectious exposure in early life, the quantity and composition of food intake, and the amount of exercise and sleep (with concomitant obesity) have been implicated in promoting the rising incidence of type 1 diabetes. Since the 1980s, the incidence of type 1 diabetes in Australia has more then doubled (14). A change in individuals diagnosed in the 1980s could reflect the impact of environmental factors acting in early life in the 1970s. Two candidate factors may be worthy of comment. First, up until the mid-1970s newborns in Victorian hospitals were housed together in nurseries for up to a week, but this practice ceased when the high prevalence of diarrhea associated with rotavirus infection was found to be significantly reduced by “rooming-in” with the mother (17). The possibility that rotavirus infection in the neonate could protect against type 1 diabetes is supported by findings in the nonobese diabetic (NOD) mouse model of the disease (18); interestingly, though, rotavirus infection postweaning in the NOD mouse did not protect but rather increased disease incidence (18). The latter result is consistent with our finding in children followed longitudinally in whom rotavirus infection was positively associated with evidence of islet autoimmunity (19). Second, the prevalence of obesity, and presumably insulin resistance, in Australian children began to accelerate in the 1970s (20). We have shown that insulin resistance is an independent risk factor for development of type 1 diabetes in children with islet autoimmunity (21). Therefore, insulin resistance promoted by changing environmental conditions may contribute to the pathogenesis and rising incidence of type 1, as well as type 2, diabetes.

The stable incidence of DR3,4 across time suggests that this genotype is resistant to the influence of environment. One interpretation is that adaptive T-cell–mediated immunity responsible for β-cell destruction is optimal with DR3,4 but not with lower-risk genotypes. However, the latter can be complemented by innate immunity promoted (together with insulin resistance) by a proinflammatory environment (22). Our findings illustrate that the contribution of HLA genes to type 1 diabetes has changed but not lessened over time. Indeed, genetic susceptibility to type 1 diabetes is no less relevant as more individuals fall under the shadow of a “diabetogenic” environment.

S.F. was a postgraduate scholar supported by the Australian National Health and Medical Research Council (NHMRC). G.M. and L.C.H. are, respectively, principal and senior principal research fellows of the NHMRC. Catherine McLean provided valuable secretarial assistance.