A stochastic epigenetic Mendelian oligogenic disease model for type 1 diabetes

Abstract

The incidence of type 1 diabetes (T1D) and some other complex diseases is increasing. The cause has been attributed to an undefined changing environment. We examine the role of the environment (or any changing non-genetic mechanism) in causing the rising incidence, and find much evidence against it: 1) Dizygotic twin T1D concordance is the same as siblings of patients in general; 2) If the environment is responsible for both the discordance among identical twins of patients with T1D and its rising incidence, the twin concordance rate should be rising, but it is not; 3) Migrants from high-to low-incidence countries continue to have high-incidence children; 4) TID incidence among the offspring of two T1D parents is identical to the monozygotic twin rate. On the other hand, genetic association studies of T1D have revealed strong susceptibility in the major histocompatibility complex and many optional additive genes of small effect throughout the human genome increasing T1D risk. We have, from an analysis of previously published family studies, developed a stochastic epigenetic Mendelian oligogenic (SEMO) model consistent with published observations. The model posits a few required recessive causal genes with incomplete penetrance explaining virtually all of the puzzling features of T1D, including its rising incidence and the specific low T1D incidence rates among first-degree relatives of patients. Since historic selection against any causal gene could prevent T1D, we postulate that the rising incidence is because of increasing population mixing of parents from some previously isolated populations that had selected against different causal genes.

Keywords: Disease association; Epigenetics; Genetics; HLA; Monozygotic twins; Type 1 diabetes.

Figure 1.

MZT and DZT concordance for T1D over the last several decades and in various parts of the world. T1D MZT (•) and DZT (♦) pairwise concordance (calculated as in reference (ref.) 13) and the trends of each as a function of time (—, MZT; - -,DZT) are shown. Data are from ref. and publications cited in refs. , and , except that: a) ref. is a more recent study of the Finnish T1D twins than the one reported in ref. (and therefore the earlier one is not included here); and, b) ref. is the corrected reference (although the same data) as that cited in ref. . The Finnish result from ref. is labeled “F” and the Japanese MZT concordance result from ref. is labeled “J” (both results displayed as o).

Figure 2.

Schematic presentation of the genetic component of the SEMO model for T1D involving three required susceptibility loci, of which two are recessive and one is dominant. For T1D, all three genes may be recessive. The population frequency (p) of aggregate susceptibility alleles at each of three hypothetical genetic loci (X, Y and Z) is used to calculate population susceptible frequencies (circles with equations) for each locus. Where the three susceptibility circles overlap (D), individuals are genetically susceptible to the disease. Thus, the frequency of population susceptibles (D) is the product of the frequencies of potential susceptibility frequencies at each gene. (Modified version with permission from ref. .)

Figure 3.

Postulated disease allele distribution for a recessive 3-locus model in a family with one child with T1D (II-7) (dark square). Susceptibility alleles are “D” and protective alleles are “N.” We have assumed three required recessive loci for T1D susceptibility, 1,2 and 3, that interact multiplicatively with each other [6]. Note that individual II-2 (cross-hatched circle) is also genetically susceptible to T1D but disease susceptibility is not penetrant.

Figure 4.

Predicted frequencies of 2, 1 and 0 alleles/haplotypes shared for recessive (a) and dominant (b) inheritance of an MHC susceptibility gene at various aggregate allele frequencies of D. Observed allele/haplotype distributions for MS and T1D are marked. Note that for MS, the distribution fits either form of inheritance. There is no dominant solution for TID. (Reproduced with permission from refs. and .)

Figure 5.

Hypothetical susceptibility gene frequencies and disease susceptibility in previously separate and currently admixed populations. Before admixture, populations A, B and C have similar low susceptibility to disease (A = 0.64%, B and C = 0.49%). Populations A and B have selected against different susceptibility genes and thus have inversely proportional frequencies of X and Y (Fig. 5a), whereas populations A and C have selected against the same susceptibility genes and have similar frequencies of X and Y (Fig. 5b). After genetic mixing in equal parts, the disease susceptibility in the offspring of A + B is five times or more than that of A + C or that of the parental populations, A, B or C. (Modified version with permission from ref. .)