Chapter 7 (Updated 10/26/12)  

Type 1 Diabetes Mellitus of Man: Genetic Susceptibility and Resistance    

A.   K. Steck¹, A. Pugliese² and G.S. Eisenbarth¹

1.    Barbara Davis Center for Childhood Diabetes, University of Colorado at Denver and Health Sciences Center

2.    Diabetes Research Institute, University of Miami

INTRODUCTION

            Though there is heterogeneity for type 1A in age of diabetes onset¹, age at which islet autoantibodies first appear, rate of progression to diabetes² and even completeness of beta cell destruction³, overall the genetic determinants are similar¹.  Even latent Autoimmune Diabetes of adults appears to be a variant of type 1A diabetes upon genetic analysis⁴.  This homogeneity is reflected in the islet autoantibodies expressed, specific beta cell destruction within islets^{5, 6} and HLA associations.  We believe that type 1A diabetes is driven primarily by CD4 and CD8 T lymphocyte targeting of the molecule insulin (or proinsulin) leading to the specific beta cell destruction⁷.  Thymic deletion of insulin reactive T cells for both man and animal models⁸ is a critical determinant and likely relates to diabetogenicity of the insulin gene VNTR and mutations of the AIRE gene^{9, 10}.  HLA alleles determine how and which islet peptides are recognized by specific T cell receptors¹¹ including the register in which such peptides can be recognized by analogy with the NOD mouse^{12, 13}.  Multiple additional genetic loci with smaller effects combine to determine the probability of maintaining tolerance¹⁴ and thus patients with type 1A diabetes are at risk for other autoimmune disorders related to both their specific HLA alleles (e.g. DQ2 for type 1A and celiac disease) and less characterized abnormalities of tolerance.  Already at onset of diabetes a third of the patients have multiple autoimmune disorders¹⁵.

Insulin-dependent diabetes mellitus (IDDM), or type 1 diabetes, is a chronic disease usually characterized by the autoimmune destruction (Type 1 A) of pancreatic ß-cells and severe insulin deficiency ^16-18. Completion of multiple large scale genome wide association studies^19-22 has provided a clearer understanding of the genetic architecture of Type 1A diabetes^{14, 22, 23}.  In particular the overwhelming genetic determinants of Type 1A diabetes are in the major histocompatibility²⁰ complex ^{14, 24}.  This is followed by insulin gene polymorphisms, the T cell receptors signaling molecule PTPN22, and the multiple (>40) loci with very small effects. Of note, there appears to be little or no overlap between loci for Type 2 and Type 1 diabetes²⁵.  Type 1B diabetes refers to insulin dependent diabetes not of immune etiology, is not the subject of this chapter and has been difficult to diagnose.  It has been suggested that fulminant diabetes, found almost exclusively in Japan, represents type 1B diabetes, but even these patients that lack anti-islet autoantibodies, have HLA alleles associated with type 1 diabetes ²⁶.  An increasing number of “monogenic” forms of diabetes are now recognized, some of which result in severe beta cell loss (e.g. neonatal diabetes with insulin gene mutations²⁷ while others create forms of diabetes that require no therapy (e.g glucokinase mutations) or are better treated with sulfonylureas rather than insulin including mutations of the sulfonylurea receptor²⁸ and HNF1alpha mutations^{27, 29}.  Monogenic forms of diabetes occur in approximately 1.5% of children developing diabetes.   Thus defining whether a patient has the more common form of diabetes in children, namely immune mediated diabetes has assumed greater importance as correct genetic diagnosis can alter therapy.  Testing of new onset children with an inclusive series of anti-islet autoantibody assays (assays for GAD65, IA-2(ICA512), insulin and ZnT8 autoantibodies) can now identify more than 90% of children with type 1A diabetes, and can aid in defining (negative autoantibodies) a subgroup of children with new onset diabetes with both monogenic (including insulin gene mutations³⁰)  and particularly for teenagers, children with type 2 diabetes. It is estimated that 10% of autoantibody negative children have monogenic forms of diabetes.  Recent studies of the pancreas of the NPOD program indicate that a significant proportion of African American and Hispanic American individuals with childhood onset diabetes have islet pathology that is very different from classical Type A diabetes pathology with pseudoatrophic islets (islets lacking all insulin producing beta cells)³¹.  The etiology of this form of Type 1 diabetes is unknown but is not associated with islet autoantibodies or HLA DR3 and DR4 alleles and may be related to poorly characterized ketosis prone diabetes, “Flatbush” or “Type 1.5” diabetes ^32-34.

Type 1A diabetes frequently develops in children, adolescents and young adults, but approximately half of individuals developing type 1A diabetes first present as adults³⁵. The disease is quite heterogeneous in its clinical expression and it can be confused with type 2 diabetes, especially in those patients who develop diabetes at a later age ^{36, 37}. Inherited genetic factors influence both susceptibility to and resistance to the disease. Although a significant proportion of patients with type 1A diabetes lack a first degree family history for the disease(>85%), there is significant familial clustering with an average prevalence of approximately 6% diabetic for siblings compared to 0.4% in the US Caucasian population. The familial clustering (λs) can be calculated as the ratio of the risk to siblings over the disease prevalence in the general population, and thus λs = 6/0.4 = 15 ^{38, 39}.

One’s genetic susceptibility depends on the degree of genetic identity with the proband. The risk of diabetes in family members has a non-linear correlation with the number of alleles shared with the proband. The highest risk is observed in monozygotic twins (100% sharing) followed by first, second and third degree relatives (50%, 25%, 12.5% sharing, respectively). Based on such estimates of observed risk, it has been suggested that diabetes susceptibility may be linked to a major locus and that several other minor loci may contribute to diabetes risk in an epistatic way. This model generates the risk curve that best parallels the risk curve obtained from observed risk estimates ⁴⁰. The moderate disease concordance observed even amongst identical twins (usually 30-50%, 70% in studies with longest follow-up) implies that inherited genes provide increased susceptibility ^41-45  with dizygotic twins having a risk not appreciably different from siblings⁴⁶.

Much technological progress has facilitated the study of the genome to map disease susceptibility genes for multi-factorial diseases, including the increasing availability of microsatellite markers, single nucleotide polymorphisms (SNPs), automated typing technology ⁴⁷, and recently whole genome SNP analysis ⁴⁸. In the case of type 1 diabetes, genome scans for IDDM susceptibility loci have been facilitated by the availability of large collections of families with affected sib-pairs, including those in the Human Biological Data Interchange (HBDI), the British Diabetic Association (BDA)-Warren repositories ⁴⁹ and recently the Type 1 Diabetes Genetics Consortium (T1DGC).  During the last decade many loci and recently genes have been linked to diabetes and there is evidence for epistatic interactions, suggesting that type 1 diabetes is a polygenic disorder, with loci within the major histocompatibility complex providing the bulk of genetic susceptibility ^{50, 51}.  These loci are discussed in detail in this chapter.

It is also possible that a subset of the disease is genetically heterogeneous, with different  loci determining disease risk in different families. Genetic heterogeneity has been demonstrated in most of the genome wide scans performed to date. The genetic heterogeneity can also be demonstrated with the study of groups of monozygotic twins.  When the first twin of a twin-pair develops type 1 diabetes after age 25, the risk of the second monozygotic twin developing type 1 diabetes is less than 5% with long-term follow up ⁴⁴, while approximately 60% of initially discordant twins whose twin mate developed diabetes prior to age 6 have progressed to diabetes (by life table analysis with 40 years of follow-up).  For monozygotic twins of patients with type 1 diabetes, expression of anti-islet autoantibodies directly correlates with progression to overt diabetes.  Essentially all such twins who express “biochemical” anti-islet autoantibodies (to GAD, IA-2/ICA512, insulin, measured by radioimmunoassays) progress to diabetes, some after decades of follow-up ⁵².  In contrast, dizygotic twins have a low risk of expressing anti-islet autoantibodies, a risk that is essentially identical to that of siblings.  These risk estimates have been validated through the exchange of sera ⁵³ and confirmed by a large study of the DPT-1 (Diabetes Prevention Trial – Type 1) cohort of at-risk relatives ⁴⁴. Similar results were obtained studying a population-based twin cohort of 22,650 twin pairs from Finland, the country with the highest disease incidence in the world ⁵⁴. 

Figure 7.1 Diabetes-free survival analysis of the combined Great Britain and United States cohorts, by age at diagnosis in the index twin: Ages 0-24 years (n=150) in solid line, 25 years and older (n=37) in dashed line.

Besides inherited alleles, other mechanisms regulating gene expression including epigenetic and parent-of-origin effects may influence susceptibility by modifying the transmission and transcription of inherited genes.  It is also an intriguing possibility that additional epigenetic factors or their expression may be acquired after birth, perhaps through environmental exposures.  Thus, a variety of genetic mechanisms may influence the autoimmune responses leading to ß-cell destruction. This chapter will review the current knowledge about the genetics of type 1 diabetes in humans.

Figure 7.2.  Odds ratios for a series of identified “genes/genetic loci” from recent genome screens and replication studies. In most cases the association is with a locus and not proven for the genes indicated (Concannon et al NEJM).

INHERITED SUSCEPTIBILITY LOCI

Both association studies and linkage analysis using various analytical methods have been used to identify IDDM susceptibility loci.  These are conventionally noted using the abbreviation IDDM and a number, e.g. IDDM1, IDDM2, etc., with the number usually reflecting the order in which such loci were reported (Table 7.1 and Figure 7.2). Many of the early IDDM loci appear at present to have been “false positives” and are generally being replaced by more recent GWAS studies and in a few instances identified genes (figure 7.2).  Using the candidate gene approach, association studies provided evidence for the first two susceptibility loci, the HLA region (IDDM1) and the insulin gene (INS) locus (IDDM2). These two loci contribute the great majority of known familial clustering (Figure 7.2). One estimate is that the MHC alone contributes 41% of the familial clustering of type 1 diabetes of the 48% estimated to be accounted for with all known genes ⁵⁰.  The next most potent locus for type 1 diabetes of man, after the insulin gene, was also discovered using a candidate gene approach, namely the PTPN22 (LYP) gene with an odds ratio of approximately 1.7 for a “missense” mutation that creates susceptibility to multiple autoimmune disorders ^55-57.  Figure 7.2 illustrates odds ratio for multiple loci summarized for GWAS studies.  The ratio of differences in frequencies, except for PTPN22 are relatively small (Figure 7.3), making it unlikely that the other indicated loci will contribute to the genetic prediction of type 1A diabetes, except through combinatorial analysis^{58, 59}, in contrast to the HLA and insulin region genes.  For instance the HLA DR3/4-DQ2/8 genotype is present in 2.3% of newborns in Colorado, but more than 30% of children developing diabetes, providing “extreme” risk, as will be discussed subsequently.  Compared to a population prevalence of type 1 diabetes of approximately 1/300, DR3/4-DQ2/8 newborns from the general population have a 1/20 genetic risk ⁶⁰.  As will be discussed subsequently additional loci within or linked to the MHC (Major Histocompatibility Complex) can increase this risk for first degree relatives of DR3/4-DQ2/8 newborns to as high as 80% ⁵¹.  Such extreme risk, suggests that for this major subgroup of children, the bulk of familial aggregation is determined by alleles of genes within or linked to the classic MHC, and the search for additional (non-DR and DQ) genetic determinants in this region is underway ^61-66.

Figure 7.3.  Allele frequencies for case versus control association studies with “significant” associations outside of the major histocompatibility complex.

Prior to the whole genome SNP analyses that have recently been reported, a number of genome-wide studies of families and affected sibling-pairs have been performed since the mid 1990’s in an attempt to identify susceptibility loci using linkage analysis ⁶⁷. Linkage analysis confirmed linkage with IDDM1 (HLA) and IDDM2 (insulin gene) and further provided evidence for the existence of approximately 20 susceptibility loci. Many of these loci show modest linkage and linkage is often not confirmed in all genome scans. Sample size and composition, genetic heterogeneity and analytical methods underlie much of the variability observed in these studies. A coordinated effort to investigate the genetics of the disease, the Type 1 Diabetes Genetics Consortium (T1DGC) (www.t1dgc.org), involves the study of patients and their families from around the world. In 2005 the consortium published its first report, with combined linkage analysis of four datasets, three previously published genome scans, and a new dataset of 254 families. This analysis included 1,435 families with 1,636 affected sibling pairs, representing one of the largest linkage studies ever performed for any common disease and involving families from the U.S., U.K. and Scandinavia ⁶⁸. Given the average map information content (67%, >400 polymorphic microsatellite markers in each scan), this dataset had ~95% power to detect a locus with _S 1.3 and p= 10^-4. With this analytical power, more than 80% of the genome was found not to harbor susceptibility genes of modest effect that could be detected by linkage. The study confirmed linkage with IDDM1 (nominal P = 2.0 x 10^–52). Moreover, nine non–HLA-linked regions showed some evidence of linkage (nominal P < 0.01), including three at (or near) genome-wide significance (P < 0.05): 2q31-q33, 10p14-q11, and 16q22-q24. In addition, after taking into account the linkage at the 6p21 (HLA) region, there was evidence of linkage with the 6q21 region (IDDM15). The published literature on these loci is discussed in detail in the following paragraphs. A comprehensive list of these initial susceptibility loci is shown in Table 7.1 with LOD scores and _S from the 2005 T1DGC scan ⁶⁸. 

Table 7.1. Susceptibility Loci for Type 1 Diabetes as of “2005”

            The recent whole genome screens, with increasing power suggest as indicated above that many of the prior loci are either false positives, have such small effects that they were not detected in the genome screens, or are related to only specific populations, as for instance is suggested for the SUMO4 gene for only Asian patients ⁶⁹.  Table 7.2 summarizes “significant” regions for the whole Wellcome Trust case control study using the combined “control” reference population of 7,670 controls compared to 2,000 patients with type 1 diabetes (The locus for IFIH1 did not reach “significance” in this Wellcome whole genome analysis with the SNPs analyzed, but is included in Table 7.2 related to a follow-up study⁵⁰).

Table 7.2. Post-whole genome screens (2007) -Susceptibility Loci for Type 1 Diabetes; Bolded loci <10(-7); underline <10(-5) WTCCC analysis

The Major Histocompatibility Complex, IDDM1

The major locus for type 1 diabetes susceptibility⁵⁹