Diabetes Mellitus: A Fundamental and Clinical Text
3rd Edition

Although the contribution of heredity is well recognized (1,2), progress toward an understanding of the genetic basis of type 2 diabetes mellitus (DM) has been largely restricted to a few distinct rare monogenic syndromes with predictable modes of inheritance, including autosomal-dominant maturity-onset diabetes of the young (MODY) (3), autosomal-recessive syndromes of extreme insulin resistance (4), and maternally inherited diabetes and deafness (5), among others. By contrast, the common forms of type 2 DM are the result of a pool of mutant genes, each of which contributes modestly and conspires with the environment and aging to manifest the disease. Individually, these genes are more accurately termed susceptibility genes and are likely to be those involved in pancreatic function, insulin signaling, energy expenditure/obesity, and appetite behavior. Identifying these genes and defining their relative contributions will be highly relevant in determining the cause of type 2 DM, as well as developing new strategies for prevention and treatment.

There are two general approaches to elucidating genetic defects that cause or predispose to disease: genome-wide linkage analysis/positional cloning and candidate genes (6). Genome-wide linkage analysis/positional cloning uses polymorphic markers distributed throughout the genome to search for specific chromosomal regions that are shared more often in affected members of pedigrees than would be expected (6,7,8). Once a specific chromosomal region is localized, the disease gene is identified by linkage disequilibrium mapping or the sequencing of genes in the region. The disease gene may be one not previously known and may require additional studies to determine its function and how the specific defect leads to its pathophysiologic consequences. Although the genome-wide approach is reasonably straightforward for highly penetrant monogenic diseases, this method is much more difficult for complex disorders such as type 2 DM, in which the genetic basis is likely to be heterogeneous in the population, several genes in the same individual/family each contribute modestly, and environmental factors (e.g., physical inactivity and caloric excess) influence risk. A detailed review of genome-wide strategies for the identification of type 2 DM susceptibility genes is discussed in other chapters.

Over the past three decades, the genetic basis for many monogenic diseases (e.g., sickle cell anemia) has been elucidated through candidate gene approaches. Based on the role of a given protein in a pathway thought to be involved in the disease, molecular biologic methods are applied to elucidate specific mutations in the gene. Candidate gene approaches are now being applied to polygenic and heterogeneous genetic disorders such as asthma, hypertension, Alzheimer disease, and type 2 DM. This chapter reviews (a) the theory of candidate gene approaches, (b) methods used to identify gene variants, and (c) progress to date on several candidate genes for type 2 DM. Several candidate genes are reviewed, including protein phosphatase type 1 (intestinal fatty acid–binding protein), β-cell adenosine triphosphate sensitive potassium channel, β₃-adrenergic receptor, and peroxisome proliferator–activated receptor-γ (PPAR-γ). Studies of other type 2 DM candidate genes are summarized in Table 68.1. Insulin receptor substrate-1, MODY, and mitochondrial DNA–encoded genes are discussed in other chapters. Finally, although diabetes is necessary for the development of macrovascular and microvascular complications, susceptibility to these complications is under substantial genetic control, presumably due to genetic variants distinct from those that predispose to diabetes itself. A discussion of diabetes complication susceptibility genes is beyond the scope of this chapter.

Once a gene is identified as a candidate for Type II DM, there are several approaches that may be applied to determine whether it is indeed defective. The most straightforward approach is to sequence the gene of interest from individuals with diabetes, and compare the sequence to that from unaffected individuals.

DNA sequence from a moderate number of individuals (at least 50–100 alleles) are required for complex disorders such as type 2 DM since a mutation in a given gene may contribute to the disease in only a subset of subjects. Furthermore, the gene of interest must be studied in several ethnic groups because the relative contribution of a given diabetes susceptibility gene to risk may differ among ethnic populations. In addition to studying coding regions, other functionally important regions should be studied, including exon-intron splice junctions and regulatory regions. Identifying mutations in regulatory regions may be difficult because, for most genes, all regulatory regions are not known and may be many kilobases away from the coding region. Additionally, these regions tend to be more polymorphic, that is, more likely to have sequence variation without biologic consequences. Minimally, the immediate 5′-flanking region of the gene (∼1,000 bp) should be screened for mutations.

DNA sequence analysis is commonly used to identify sequence variation in candidate genes. Because most subjects will be heterozygous for a given gene mutation, a high-quality sequence, usually of both strands, is necessary to reliably detect a heterozygous base change. Other rapid and economical alternatives to DNA sequencing have been developed, including electrophoretic methods and denaturing high-pressure liquid chromatography (dHPLC). Electrophoretic methods include single-stranded conformational polymorphism analysis (9), denaturing gradient gel electrophoresis (10), temperature gradient gel electrophoresis (11), and heteroduplex analysis (12). These approaches generally involve polymerase chain reaction (PCR) amplification of a segment of the candidate gene, followed by gel electrophoresis under special conditions. Altered mobility through the gel when compared with the normal sequence may indicate sequence variation, which then may be confirmed by DNA sequence analysis. dHPLC uses a progressive chemical gradient through a temperature-controlled reverse-phase HPLC column to resolve PCR products that differ by as little as a single nucleotide. An important caveat of the above methods is that they will not detect heterozygotes for large deletions or duplications because the normal allele will be amplified and detected as normal. To detect large deletions and duplications, restriction fragment length polymorphism (RFLP)/Southern blot analysis or gene dosage–PCR (13) must be performed.

Once a sequence variant in a candidate gene is discovered, it is necessary to determine if it is a normal variant (i.e., polymorphism), or whether it contributes to the development of Type II DM. This determination is a critical step since most sequence variation that is discovered as a result of mutation screening will not be relevant to the disease process. Many rapid genotyping methods have been developed, including PCR-RFLP, allele-specific oligonucleotide hybridization, ligation assays, fluorescent primer extension, mass spectrometry, and pyrosequencing. Once genotypes from many affected and unaffected individuals are obtained, statistical approaches may be used to determine whether the gene variant is present more frequently in individuals with diabetes than in those without diabetes (i.e., whether it is associated with type 2 DM). Alternatively, association of the gene variant with quantitative traits related to type 2 DM (e.g., higher glucose concentration, lower insulin sensitivity) can be sought using, for example, analysis of variance or analysis of covariance. Unfortunately, subtleties of ascertainment, particularly stratification bias, can result in false-positive (or false-negative) findings, and as such, population-based association studies must be interpreted with caution. Thus, analysis of gene variants should be performed in several populations. Confirmatory findings among more than one population are supportive of a pathophysiologic role of a given gene variant. However, because findings may be influenced by ascertainment methods, genetic (ethnic) background, phenotypic characterization, statistical methods, and allele frequency/power, replication of associations of genotype with phenotypes may not be universal. An alternative to population-based association analysis is family-based linkage or association analysis. Although these approaches are less prone to stratification bias, family collections are more difficult to obtain. Furthermore, linkage analysis in families is less sensitive than population-based case-control designs in detecting modest effects of gene variants.

Demonstration of a functional defect in the gene variant is an important step toward validating that a gene variant is indeed pathogenic. Association of genotype with phenotype in the absence of evidence of a functional defect suggests that the variant may be in linkage disequilibrium with a yet-to-be discovered mutation in the gene, or a nearby gene, or that the association is a false-positive result. Alternatively, the gene variant may have a subtle functional defect that is difficult to detect in vitro.

An alternative to searching directly for mutations in candidate genes is to genotype polymorphic markers—that is, single nucleotide polymorphisms (SNPs) or short tandem repeat (STR) markers—within or near the candidate gene to determine whether specific alleles cosegregate with diabetes or other related phenotypes in pedigrees (linkage analysis) or are present at a greater frequency in unrelated subjects with type 2 DM than in those without type 2 DM (association analysis). SNPs have the advantage of being easier to assay and are much more frequent in the genome than STRs (approximately 1 per 1,000 base pairs). With the sequencing of the human genome, several million SNPs have been discovered and have been cataloged in public databases (see http://www.ncbi.nlm.nih.gov/SNP/). The rationale behind this approach is an SNP that is close to the pathogenic mutation is likely to be in linkage disequilibrium with it; therefore, the SNP should provide information about the presence (or absence) of a nearby pathogenic mutation. Thus, association between phenotype and a single SNP or haplotype consisting of several closely spaced SNPs may provide evidence that a pathogenic (mutant) form of the candidate gene exists. DNA sequence analysis of the candidate gene may then be undertaken to identify the mutation.

Recent insights into the molecular mechanisms of pancreatic development, insulin secretion, insulin signaling, and body weight regulation, and in pathophysiologic changes in these and other pathways thought to contribute to diabetes, have resulted in an explosion in the number of potential candidate genes for type 2 DM (14,15,16). Almost as rapidly as new molecules are identified, they are studied as candidate genes for diabetes. With only a few exceptions, these studies have been negative.
However, due to the relatively small number of subjects studied, usually in a single ethnic population and limited in scope to coding regions, these negative studies should not be overinterpreted. In this section, we focus on a few candidate genes for which the genetic variants have been demonstrated to be functional and shown in more than one study to associate with type 2 DM or related phenotypes.

Using linkage analysis, Prochazka and co-workers identified a region on chromosome 4q that was linked to fasting insulin levels and insulin resistance in Pima Indians (17). Among the several genes known to be encoded within this chromosomal region was intestinal fatty acid binding protein type 2 (FABP2). FABP2 is expressed only in the villus epithelial cells of the small intestine and plays a role in absorption of fatty acids from the gut. Abnormal fatty acid metabolism is associated with insulin resistance (18); therefore, FABP2 was considered a candidate gene for type 2 DM. Molecular scanning of FABP2 identified a missense mutation (Ala54Thr) (19). The Thr54 variant is prevalent in the Pima Indians with an allele frequency of 0.29 and is also found in a number of other populations [allele frequencies: Japanese 0.34 (20), U.S. whites 0.31 (21), Finns 0.28 (22), aboriginal Canadians 0.14 (23)]. Pima Indians with the Ala54Thr variant had higher fasting insulin values (p < 0.05) and decreased glucose uptake during a euglycemic hyperinsulinemic clamp (p < 0.04), indicating greater insulin resistance (19). They also had higher rates of fat oxidation. Japanese subjects homozygous for the Thr54 allele demonstrated higher basal (p < 0.05) and 2-hour insulin levels (p < 0.002) on an oral glucose tolerance test (OGTT) compared with other genotypes, again suggesting an association of greater insulin resistance with the Thr54 allele (24). Similar findings were observed in postmenopausal white women (25). In contrast, studies in other Japanese (20,26), white (22,27), and African-American (28) cohorts have not found an association of type 2 DM, insulin levels, or obesity with the Thr54 variant.

To better understand the pathophysiology of the Ala54Thr FABP2 substitution, it was expressed in vitro and found to bind long-chain fatty acids with twofold greater affinity (19) than the wild-type receptor. Additionally, Thr54 FABP2 had increased fatty acid transport in Caco-2 cells compared with Ala54 FABP2 (29). Consistent with this, Agren et al. (30) found that Thr54 homozygotes have a significantly greater increase in triglyceride levels after a fat test meal than Ala54 homozygotes. Thus, Thr54 FABP2 may be more efficient in binding fatty acids, which may result in increased absorption of dietary free fatty acids and in turn contribute to insulin resistance.

Glycogen synthase is the key enzyme that regulates the synthesis and storage of glycogen from glucose in muscle and liver (31). Insulin activates glycogen synthase phosphatase [protein phosphatase type 1 (PP1)], which in turn dephosphorylates glycogen synthase, creating the active form of the enzyme. PP1 is composed of a catalytic subunit (PPP1CB) and a regulatory subunit (PPP1R3) that regulates the interaction of PP1 with glycogen-bound substrates. A decrease in the ability of insulin to stimulate glycogen synthesis may be a very early defect in the progression to hyperglycemia and type 2 DM (32,33). The regulatory subunit of protein phosphatase 1 (PPP1R3) is located on chromosome 7q31.1-q31.2. Mutations in PPP1R3 have the potential to inactivate PP1, thereby decreasing the levels of active glycogen synthase and in turn decreasing glucose incorporation into glycogen. An Asp905Tyr missense mutation (allele frequency = 0.11), identified in a Danish cohort, did not associate directly with type 2 DM (34). However, in healthy Danish volunteers, the Tyr905 allele was associated with decreased insulin-stimulated nonoxidative glucose metabolism (glycogen synthesis) (p < 0.04) and increased basal glucose oxidation (p < 0.04). During an intravenous glucose tolerance test (IVGTT), obese nondiabetic Tyr905 heterozygotes had a 42% higher first-phase insulin response compared with obese Asp905 homozygous subjects (34). Possibly, the increased insulin levels help overcome the decreased glycogen synthase activity that results from the mutation in PPP1R3.

Xia et al. (35) identified a 5-bp length insertion/deletion in the 3′-untranslated region (UTR) that affects the distance between two ATTTA motifs. ATTTA motifs are the smallest consensus motifs of messenger RNA (mRNA)-destabilizing AU(AT)-rich elements (35). In Pima Indians (allele frequency = 0.56), those homozygous for the deletion had a mean PPP1R3 mRNA concentration that was 44% lower than those homozygous for the insertion, with heterozygotes having intermediate levels of PPP1R3 mRNA (20% decrease) (35). Those with the deletion had higher fasting plasma insulin, lower insulin-mediated glucose uptake, and a higher prevalence of type 2 DM. In a Japanese study, the 3′-UTR variant was in complete linkage disequilibrium with the Asp905 allele. The allele frequency of the 3′-UTR variant was significantly higher in subjects with type 2 DM (0.34 vs. 0.26, p < 0.02) (36). The 3′-UTR variant has been identified in other populations, but has not been consistently associated with type 2 DM and insulin resistance (37,38).

The β-cell adenosine triphosphate (ATP)-sensitive potassium channel plays a critical role in insulin secretion. The channel is composed of two subunits: the sulfonylurea receptor-1 (SUR1) and an inward rectifying potassium channel (Kir6.2). Both components are encoded on chromosome 11p15.1 about 4.5 kb apart. SUR1 is a transmembrane protein with two intracellular nucleotide binding folds that bind ATP. Binding sites for sulfonylurea and meglitinides are most likely located at the amino terminus of the protein. Kir6.2 has two membrane-spanning domains and most likely homodimerizes to form the potassium channel. Mutations in both SUR1 and Kir6.2 have been identified in subjects with persistent hyperinsulinemic
hypoglycemia of infancy (39,40,41), thus pointing out the importance of the K_ATP channel in the regulation of insulin secretion.

Several polymorphisms have been identified in the SUR1 gene, including exon 18 (ACC→ACT, Thr759Thr) and in the intron, 5′ of exon 16 (t→c change located at position –3 of the exon 16 splice acceptor site) [numbering of SUR1 exons according to Hansen et al. (42)], which associated with type 2 DM in two white populations of European descent (Utah Mormons, United Kingdom) (43). Similarly, a study in French whites showed an association of type 2 DM with the exon 18 variant; however, the association appeared to depend on body mass index (BMI) (44). No association with diabetes was found for the intron variant. Although all of these variants are silent or within introns, the association with type 2 DM is most likely the result of being in linkage disequilibrium with a yet-to-be discovered pathogenic mutation in the SUR1 gene or another nearby gene.

Molecular scanning of the Kir6.2 gene has identified three missense mutations: Glu23Lys, Leu270Val, and Ile337Val (45,46,47). Glu23Lys is in linkage disequilibrium with Ile337Val (47). The Lys23 variant was significantly associated with type 2 DM in French whites (allele frequency for Lys23 allele 27% vs. 14%, p = 0.015) (48) and in UK whites [odds ratio for Lys23 heterozygotes = 1.23 (1.12–1.36), p = 0.000015; and for Lys23 homozygotes = 1.65 (1.34–2.02), p = 0.000002] (49). A metanalysis of four white populations (one French, two from UK, one Danish) (type 2 DM n = 521, control n = 367) demonstrated an association of the Lys23 variant with type 2 DM (p = 0.0016; corrected for multiple comparisons p < 0.01) (48). In several other studies, the Kir6.2 variants have not been directly associated with type 2 DM; however, the variants have been associated with traits related to diabetes (45). Young healthy Danish subjects heterozygous for Val270 and homozygous for Lys23/Val337 had 62% higher insulin sensitivity on an IVGTT compared with normal homozygotes (Leu270, Glu23/Ile337) (45). Nondiabetic German subjects with the Lys23 allele had less suppression of glucagon secretion in response to hyperglycemia; however, there were no differences in acute insulin response or glucose effectiveness (50). Functional studies of the Glu23Lys allele suggest that this variant may produce an overactive K_ATP channel, which in turn may decrease insulin secretion (51). Further studies are needed to determine the significance of these findings in the development of type 2 DM.

The marked susceptibility of populations to obesity and thus type 2 DM has been hypothesized to be due to a putative “thrifty genotype” (52). In traditional populations, efficient energy storage and utilization provided a survival advantage. However, in an environment of assured access to a diet abundant in calories, and a more sedentary lifestyle, this thrifty genotype becomes disadvantageous, leading to obesity and its comorbidities, including type 2 DM (52). The β₃-adrenergic receptor gene (ADRB3) is expressed in visceral adipose tissue. Stimulation of the receptor by β-agonists (or the sympathetic nervous system) activates adenylyl cyclase, which increases intracellular cyclic adenosine monophosphate concentrations and results in increased lipolysis and thermogenesis. Walston and co-workers (53) identified a missense mutation (Trp64Arg) in Pima Indians (allele frequency = 0.31). Subsequently, the variant was found in all populations studied (53,54,55,56,57) with the exception of the Nauruans of the South Pacific (55). In Pima Indians, subjects homozygous for the Arg64 ADRB3 variant had a significantly earlier onset of type 2 DM and tended to have a higher, albeit not statistically significant, prevalence of type 2 DM (53). They also tended to have lower resting metabolic rates and increased BMIs (53). In Finns, the variant (allele frequency = 0.12), even in its heterozygous form, was associated significantly with an earlier onset of type 2 DM, elevated 2-hour glucose and insulin levels during an OGTT, decreased glucose disposal rates, increased waist-to-hip ratio, and increased diastolic blood pressure, suggesting a contribution of the Arg64 variant to the insulin resistance syndrome (54). In a French cohort of obese subjects, the variant was associated with increased capacity to gain weight over time (58). Since these original reports, over 100 association or linkage studies of the Trp64Arg substitution have been published, and its role in the development of type 2 DM and obesity continues to be controversial. Studies have shown both associations (59,60,61,62,63,64) and the absence of associations (59,65,66,67,68) with insulin resistance syndrome–related traits. Metanalyses support a small effect of the Arg64 ADRB3 variant on BMI (69,70,71). Several more recent studies suggest that the Arg64 ADRB3 variant may interact with other gene variants within the same or converging biochemical pathways [e.g., variants in genes for FABP-2 (72), uncoupling protein-1 (73), α₂-adrenergic receptor (74), PPAR-γ (75), and type 2 deiodinase (76)] to increase susceptibility to insulin resistance syndrome–related traits.

Several studies investigated the functional consequence of the Trp64Arg ADRB3 substitution. Two studies in the Pima Indians examined the relationship of ADRB3 genotype and lipolysis in vivo in subcutaneous fat and did not find significant differences in lipolysis among the three genotypes (77,78). However, these studies used nonspecific β-agonists. Li et al. measured lipolysis in vitro in isolated visceral white adipose cells from Trp64 homozygotes and Arg64 heterozygotes and initially did not find significant differences in lipolysis between the two groups (79). However, when the number of subjects was increased, the median effective concentration (EC₅₀) for CGP12177, a β3-receptor selective agonist, was tenfold higher in Arg64 heterozygotes (EC₅₀ = –8.8 ± 1.3 vs. –7.8 ± 1.2; p = 0.01) (80), suggesting a decrease in Arg64 ADRB3 function. The results of this study were confirmed by Umekawa and co-workers (81). Studies of ligand binding, adenylyl cyclase activation, and desensitization in Chinese hamster ovary (CHO) cells overexpressing either the normal Trp64 ADRB3 or the Arg64 variant receptor failed to show any differences between the two receptors (82). However, another study in CHO cells suggests that the Arg64 ADRB3 variant may have lower basal activity and decreased maximal activation of adenylyl cyclase (83). Recently, we showed that Arg64 ADRB3 fails to associate with G-proteins and acts as a dominant negative receptor when coexpressed with the normal Trp64 ADRB3 (84).

In summary, the Arg64 ADRB3 increases susceptibility to several features of the insulin resistance syndrome as well as an
earlier (accelerated) onset of type 2 DM (53,54). Although the Arg64 variant does not have a major influence, it is relatively common and is likely to play a role as a modifier in the typical (polygenic) forms of type 2 DM and obesity in humans.

PPAR-γ is a nuclear receptor that is encoded on chromosome 3p25. As a result of differential splicing, two isoforms are generated, γ₁ and γ₂. The γ₁ isoform is expressed in many tissues, including muscle and adipose tissue, whereas the γ₂ isoform is expressed only in adipose tissue. Both isoforms have been implicated in the regulation of adipocyte differentiation as well as insulin sensitivity and lipid metabolism, making them promising candidate genes for both diabetes and obesity. A proline to alanine substitution at codon 12 in the γ₂-specific exon (Pro12Ala PPAR-γ₂) was identified (85) and is common in diverse populations (allele frequency 0.10–0.19 in white populations, 0.09 in Pima Indians, and 0.02 in African Americans) (86). Functional studies reveal lower affinity to the PPAR-γ response element and decreased activation of the Ala12 PPAR-γ₂ variant by PPAR-γ agonists compared with the normal Pro12 allele (87,88).

The Ala12 variant is found less commonly in Japanese (89,90), Finnish (91,92), and Oji-cree women (93) with type 2 DM. However, this is not the case in all studies (94,95,96,97). A metanalysis including more than 3000 subjects revealed a significantly lower risk for type 2 DM in subjects carrying the Ala12 allele (RR for allanineallele = 0.79; p = 0.00007) (98). Consistent with this finding is association of the Ala12 variant with greater insulin sensitivity in nondiabetic subjects as well as lower fasting insulin levels (99,100,101,102). In contrast, nondiabetic German whites who were first-degree relatives of subjects with type 2 DM with the Ala12 variant allele had lower glucose uptake and insulin sensitivity index on euglycemic hyperinsulinemic clamps (103). However, when a subgroup of subjects with BMIs over 30 kg/m² were analyzed separately, an association with increased insulin sensitivity was noted (103). The association with improved insulin sensitivity could not be confirmed in Japanese (104), Italians (105), or Koreans (106). The association of this variant with BMI is less well established. Several studies have shown association between Ala12 carriers and increased BMI (107,108), whereas only one study has shown lower BMI in Ala12 carriers (109). Differences in the direction and magnitude of associations with BMI and insulin sensitivity may be due in part to interaction of the Ala12 variant with dietary fat intake (110) or with obesity itself (111,112). For example, in a weight loss study, postmenopausal white women with the Ala12 allele lost similar amounts of weight as did those homozygous for the Pro12 allele (112). However, those with the Ala12 allele became significantly more insulin sensitive after weight loss and regained their weight more rapidly after the weight loss intervention ended.

In summary, the Ala12 PPAR-γ₂ variant appears to be protective against the development of insulin resistance and susceptibility to type 2 DM, and at the same time increases susceptibility to obesity and weight gain. It may thus be considered a “thrifty gene” (52).

Type 2 DM is a complex genetic disorder that is both polygenic and heterogeneous. Genes that increase susceptibility to type 2 DM are likely to include those involved with pancreatic development, insulin secretion, and insulin action, and to overlap with those that increase susceptibility to obesity. Most likely, mutations in some genes will be distinct for a given population, whereas others may occur more broadly in several populations. Furthermore, the phenotypic expression of a given mutation may vary markedly among populations (and even within individuals of the same population) depending on genetic background and differences in environment and behavior. Due to the inherent difficulties in elucidating susceptibility genes for polygenic and heterogeneous diseases such as type 2 DM, early positive results for any gene or locus must be regarded with suspicion until its authenticity is supported by studies in the same or other populations, and until evidence for a functional defect in the mutant gene product is demonstrated. Despite these difficulties and caveats, the next several years will prove to be exciting in the search for diabetes susceptibility genes. With these new discoveries, fundamental insights into the molecular basis of diabetes will lead to improved early diagnostics and to the development of novel interventions (pharmacologic as well as the prospect of gene therapy) for the prevention, delay, and treatment of diabetes.