The Genetic Basis of Type 2 Diabetes

Cytokines, Adipokines and Inflammation

Insulin stimulated glucose uptake into muscle cells and adipocytes depends mostly on the glucose transporter GLUT4, skeletal muscle being the principal site for this glucose uptake. Downregulation of GLUT4 expression in adipose tissue is a common feature of insulin-resistant states, including obesity, T2DM, and the metabolic syndrome (26). Given that skeletal muscle is the major site for glucose disposal, the hyperglycemia associated with obesity and T2DM cannot be explained by the decreased uptake of glucose into adipose tissue due to downregulation of GLUT4 in adipocytes. Furthermore, adipocyte-selective knockout of GLUT4 (adipose-Glut4 −/−) in mice resulted in systemic insulin resistance similar to that induced by muscle-selective GLUT4 knockout mice (muscle-Glut4 −/−, 27,28). These studies suggest that adipocyte GLUT4 deficiency results in circulating factors that are responsible for cross-organ communication (“crosstalk”). Yang et. al (29) recently identified retinol binding protein-4 (RBP4) as one candidate for that crosstalk in the adipose tissue of adipose-specific Glut4 knockout mice. Furthermore, serum RBP4 protein levels were elevated in insulin-resistant mice and obese and diabetic humans. Transgenic overexpression of the human RBP4 gene or injection of recombinant RBP4 protein in normal mice caused insulin resistance, whereas Rbp4 knockout mice had enhanced insulin sensitivity (29). Increased serum RBP4 protein levels induced hepatic expression of the gluconeogenic enzyme phosphoenolpyruvate carboxykinase (PEPCK) and impaired insulin signaling in muscle.

Investigators have proposed other potential mediators of inter-organ crosstalk to explain T2DM pathogenesis. Obesity predisposes to insulin resistance and T2DM. Adipocytes, both visceral and peripheral, secrete a plethora of factors which may alter systemic insulin action and hepatic glucose production, including adiponectin, resistin, leptin, cytokines IL-6 and TNFα, visfatin, retinol binding protein 4 (RBP4), as well as free fatty acids (FFA) (30–32). In recent years, the concept that obesity and elevated cytokine secretion results in an inflammatory state that is itself the cause of insulin resistance has emerged (32–34). The inhibition of signaling downstream of the insulin receptor may be a primary mechanism through which this inflammatory state causes insulin resistance. Exposure of cells to TNF-α or elevated levels of free fatty acids stimulates phosphorylation of serine residues of IRS-1. This phosphorylation reduces tyrosine phosphorylation of IRS-1 in response to insulin, and thus signaling downstream of IRS-1 (35–38). The inflammatory state is increased by the accumulation of macrophages in adipose tissue with obesity (39;40). Thus, both adipose tissue and the macrophages within adipose tissue serve in both endocrine and paracrine fashion to promote inflammation and decrease insulin sensitivity (32,33).

Endoplasmic Reticulum Stress Response and Diabetes

Metabolic and inflammatory stresses of obesity may also disrupt cell metabolism, particularly the processing of cellular proteins in the secretory cells of the liver, pancreas, and adipocyte. The response to this increased demand is the endoplasmic reticulum stress response pathway, which activates an extensive cascade which shuts down further translation, increases the transcription of chaperone proteins, and if not successful, causes cellular apoptosis (41–43). Among the activated proteins are IRE1 and XBP1. Özcan and colleagues (42) showed recently that the endoplasmic reticulum stress pathway is active in adipose tissue and in the liver from obese mice, leading to increased activation of c-jun N terminal kinase (JNK), which in turn phosphorylates IRS-1 on serine residues, thus suppressing insulin action and downstream signaling pathways. Similar data also support a role for endoplasmic reticulum stress pathways in β-cell dysfunction and apoptosis (44).

Endoplasmic reticulum stress is a molecular link between obesity, decreased insulin sensitivity and type 2 diabetes. Additionally, increased glucose metabolism can lead to a rise in mitochondrial production of reactive oxygen species (ROS). Inflammatory pathways can be initiated by extracellular mediators such as cytokines and lipids or by intracellular stresses such as endoplasmic reticulum stress or excess ROS production by mitochondria. Signals from all of these mediators converge on inflammatory signaling pathways, including the kinases JNK and IKK. These pathways lead to the production of additional inflammatory mediators through transcriptional regulation as well as to the direct inhibition of insulin signaling (45,46).

Mitochondria and Reactive Oxygen Species

Reactive Oxygen Species (ROS) are another factor suggested to have a role in insulin resistance. ROS production is elevated in obesity, which in turn causes enhanced activation of inflammatory pathways (47,48). Several lines of evidence support the association of oxidative stress markers with diabetes (47–49). Houstis et al (50) recently connected these mechanisms by showing that treatment of 3T3-L1 adipocyte with either TNFα (an inflammatory cytokine that is elevated in obesity) or dexamethasone (an anti-inflammatory agent that causes insulin resistance) increased the ROS level and resulted in decreased insulin action. Antioxidant molecules or transgenes encoding ROS scavenging enzymes both ameliorated the insulin resistance of TNFα or dexamethasone treated 3T3-L1 adipocytes to varying degrees. Furthermore, antioxidant molecules improved insulin sensitivity and glucose homeostasis in obese, insulin resistant ob/ob mice (50). Oxidative stress is also likely to be involved in progression of pancreatic β-cell dysfunction in T2DM.

Reduced mitochondrial oxidative phosphorylation has recently been demonstrated in several studies, using both microarrays to identify downregulated genes involved in oxidative phosphorylation (OXPHOS genes). PGC1 appears to be at least partially responsible for this defect (51,52). Supportive studies in humans have suggested a role for defective mitochondrial fatty acid oxidation, mitochondrial dysfunction, and reduced numbers of skeletal muscle mitochondria in T2DM pathogenesis, and have suggested that increased intramyocellular lipid content was associated with defects in mitochondrial activity (53–56). Morino et al (56) found 38% less mitochondrial density in muscle of insulin resistant (IR) individuals compared to controls, which was associated with a 50% increase in serine phosphorylation of IRS-1 and a 60% reduction in insulin stimulated Akt activation (a measure of insulin signaling). Decreased mitochondrial fatty acid oxidation, caused by either mitochondrial dysfunction and/or reduced mitochondrial numbers, produces increased levels of intracellular fatty acyl CoA and diacylglycerol. These molecules activate a novel protein kinase C, which in turn activates a serine kinase cascade, again leading to increased serine phosphorylation of IRS-1. As described above, serine phosphorylation blocks IRS-1 tyrosine phosphorylation and inhibits downstream signaling, including recruitment of GLUT4 to the plasma membrane and insulin-mediated glucose uptake in skeletal muscle (53).

Oxidative Metabolism and the Pancreatic β-Cell

Insulin secretion by the pancreatic β-cell is modulated by multiple stimuli. Oxidative mitochondrial metabolism and adenosine triphosphate (ATP) generation is essential to glucose stimulated insulin secretion. The increased ratio of ATP to adenosine diphosphate (ADP) in the β-cell triggers a series of events: inhibition of the cell’s ATP/ADP-regulated potassium channel (Katp, encoded by genes KCNJ11 and ABCC8), plasma membrane depolarization, opening of a voltage-gated calcium channel, calcium influx, and transport and binding of insulin granules to the cell surface (53). The ATP/ADP ratio is in turn altered by UCP2, an integral mitochondrial membrane protein that permits protons to leak across the mitochondrial inner membrane, thus uncoupling of glucose oxidative metabolism from ATP production (57). By decreasing the amount of ATP generated from glucose, UCP2 expression negatively regulates glucose-stimulated insulin secretion. Over-expression of UCP2 in β-cells in vitro decreases glucose-stimulated insulin secretion (58), whereas targeted inactivation of the UCP2 gene in mice has the opposite effect (59). Importantly, heterozygosity for a null UCP2 allele produces an effect that is intermediate between those observed in wild-type and homozygous mice, indicating that even relatively small changes in UCP2 expression may have a substantial impact on glucose-stimulated insulin secretion.

Glucose Homeostasis: The Central Nervous System

Levels of glucose in the blood are regulated by a complex interplay between the appearance of glucose from both intestinal absorption and hepatic production and its disappearance through insulin-dependent and insulin independent glucose uptake in a variety of tissues. After the overnight fast, glucose is largely produced by glycogen breakdown and gluconeogenesis. About 80% of this glucose released by liver is metabolized independent of insulin by brain and other insulin–independent tissues (gut, red cells). The medial hypothalamus, a major integrator of nutritional and hormonal signals, plays a pivotal role not only in the regulation of energy balance but also in the modulation of liver glucose output. Diminished responses of first-order neurons, such as those in the arcuate nucleus, may contribute to the pathogenesis of T2DM (60–62). The enzyme carnitine palmitoyltransferase-1 (CPT1) regulates long-chain fatty acid (LCFA) entry into mitochondria, where they undergo β-oxidation. Obici et al (63) observed that the selective reduction of lipid oxidation in the hypothalamus by inhibiting or decreasing the activity of CPT1, substantially diminished food intake and endogenous glucose production. These results indicated that changes in the rate of lipid oxidation in selective hypothalamic neurons signaled nutrient availability to the hypothalamus, which in turn modulated the exogenous and endogenous inputs of nutrients into the circulation.

A recent study by Pocai et al (64) revealed an important role for circulating LCFA and insulin levels, and a sensing role of hypothalamic (arcuate nucleus) Katp channels. Insulin infused into the region of the arcuate nucleus potently reduced hepatic glucose production from the liver. This effect was mediated by changing the level of activity of the hepatic branch of the vagus nerve that provided the major neural input into the liver. Katp channel inhibitors infused into the arcuate nucleus completely blocked the effects of CNS-infused insulin to reduce hepatic glucose production. Consistent with these results, mice lacking the SUR1 (ABCC8) subunit of the Katp channel were resistant to the inhibitory action of insulin on gluconeogenesis. Thus, activation of hypothalamic Katp channels normally restrains hepatic gluconeogenesis. Alterations within this central nervous system/hepatic circuit may contribute to diabetic hyperglycaemia (64). Other peripheral signals to the hypothalamic centers include leptin and insulin (61).

Glucose Homeostasis: Liver

Glycogenolysis and gluconeogenesis are the two known source of hepatic glucose production (HGP). HGP is controlled by the integrated metabolite fluxes to and from the liver, hormonal cues, and as reviewed above, neurotransmitter release. Fasting promotes glucose production through cAMP-dependent mechanisms, whereas feeding inhibits glucose production through insulin. In the absorptive state, insulin silences hepatic PGC-1α transcription, and suppresses gluconoegenic enzymes PEPCK and G6Pase through the insulin signaling pathway (IRS2-Akt-mediated phosphorylation, 65). Glucagon opposes these actions by stimulating the transcription of gluconeogenic genes via the cyclic AMP-inducible factor CREB (cAMP response element–binding protein), thus increasing HGP. Recent findings (66–69) have begun to elucidate the role of additional hormone-regulated transcription factors that control gluconeogenesis. LKB1, a tumor-suppressor gene which encodes a protein-threonine kinase, phosphorylates and activates AMPK (adenosine monophosphate activated protein kinase) and SIK2 (Salt-inducible kinase) in liver (67). The deletion of LKB1 in the liver of adult mice resulted in a nearly complete loss of AMPK activity and hyperglycemia with increased gluconeogenic and lipogenic gene expression. LKB1 acts through TORC2 (transducer of regulated CREB activity 2), a transcriptional coactivator of CREB which in turn acts on PGC-1α to increase gluconeogenesis. Downregulation of TORC2 reduced PGC-1α expression and normalized blood glucose levels in mice with liver specific LKB1 inactivation, indicating that TORC2 is a critical target of LKB1/AMPK signaling (67).

The β-Cell and Type 2 Diabetes

Impairment of insulin secretion from pancreatic β-cell is also a major component of T2DM pathogenesis. Analysis of mutations involved in six different maturity onset diabetes of the young (MODY) genes have revealed the important role of transcription factors in the insulin secretion. In a recent gene expression analysis of pancreatic islets isolated from T2DM and NGT individuals, Gunton et al (70) found a 90% decrease of the transcription factor, arylhydrocarbon nuclear receptor translocator (ARNT) in T2DM islets compared with nondiabetic individuals. Mice lacking ARNT in β-cells exhibited abnormal glucose tolerance, impaired insulin secretion and changes in islet gene expression similar to human T2DM (70). ARNT and its partner proteins are members of the basic helix-loop-helix Per/AhR/ARNT/Sim (bHLH-PAS) family of transcription factors. ARNT interacting partners include HIF1α, HIF2α and AhR, which together regulate the transcription of genes required for optimal glucose-responsive insulin secretion by β-cells. ARNT-HIF dimers may directly regulate the expression of several glycolytic enzymes. Oxygen and NO may also regulate insulin secretion through this ARNT-related pathway. ARNT is an obligate partner of several transcription factors involved in the response to toxins and hypoxic stress, and may integrate the genetic and environmental insults leading to T2DM pathogenesis (71).

Many mechanisms contributing to T2DM may trigger β-cell apoptosis and reduced β-cell mass or ability to compensate for insulin resistance (72). Mechanisms include endoplasmic reticulum stress (44), chronic hyperglycemia (glucotoxicity, 73,74), chronic hyperlipidemia (lipotoxicity) (75;76), oxidative stress (77), and inflammatory cytokines (78). Decreased IRS-2 expression, an essential β-cell growth factor, may lead to spontaneous β-cell apoptosis (79). Because several mechanisms relevant to pathogenesis of T2DM may increase IRS-2 serine/threonine phosphorylation (80,81) with resultant IRS-2 ubiqutination and proteosomal degradation, defects in insulin signaling and insulin secretion may be coupled.

Multiple pathways may lead to the T2DM phenotype. This review of the pathophysiology has covered only a fraction of the genes and pathways that may be implicated in T2DM pathogenesis, and that may harbor genetic variants. Based on the intertwined and complex physiologic pathway, a complicated genetic picture can be well anticipated. Indeed, such a picture is now emerging and is reviewed below.

Chromosome 2q

The first locus for T2DM was mapped by Hanis et. al. in 1996 (83) to the telomeric region of chromosome 2q (2q33–q37) in 330 Mexican American sib pairs and called NIDDM1. Further fine mapping placed NIDDM1 near D2S140 (263.6 cM; multipoint lod score=4.03) with a 1-lod support interval from 257 to 269 cM, a 12-cM region that included marker D2S125. In subsequent studies, Cox and colleagues (84) used a two–locus conditional search strategy to show that NIDDM1 interacted with a chromosome 15 locus near CYP19, with an increase in the evidence for linkage from a logarithm of odds ratio of linkage to chance segregation (LOD) of 1.3 to 4.0 when NIDDM1 was taken into account. Similarly, evidence for linkage at NIDDM1 increased from 4.03 to 7.28 by conditioning on chromosome 15 linkage and decreased the 1-lod support interval (the size of the probable region harboring the susceptibility gene) from 12 cM to 7 cM (from 257–269 cM to 259–266 cM). Studies to identify a susceptibility gene in this region (calpain 10) are reviewed below. Other studies, including those from our laboratory (85) and the original Mexican American linkage study (83), have found suggestive evidence for linkage of T2DM to chromosome 2q in a second, more centromeric region from the CAPN10 gene. Other studies with some evidence for linkage have included Han Chinese (86) at marker D2S126 (221 cM), Japanese families selected for a BMI< 30 kg/m² at 2q34 (223 cM) (87), and indigenous Australian families at 2q24.3 (177 cM; LOD score 3.9, 88). The gene or geness responsible for these observations are unknown.

Chromosome 20

Chromosome 20 was among the first regions with evidence for linkage in multiple studies (89–91). A large genome scan in 716 affected sib pairs from 477 Finnish families (The Finland–U.S. Investigation of Non-Insulin- Dependent Diabetes Mellitus Genetics or FUSION study) provided considerable additional evidence (92), and subsequent studies have added to the replication in Ashkenazy Jewish families (93) and in African sib pairs (94). In the large FUSION study, the evidence for linkage was observed on both the long and short arms, with multipoint lod scores of 3.08 on 20p (19.5cM) and 2.06 on 20q (57cM). Further high resolution mapping and conditioning on a chromosome 2p locus (8.5cM) provided further support for a locus on 20q at 69 cM (lod 5.50, 95). Linkage in the 472 sib pairs of Ashkenazy Jewish descent was similarly in the region form 51 cM to 62 cM, with a much weaker signal on 20p (2 cM, 93). A genome wide scan on Han Chinse families also found evidence for linkage near marker D20S196 on 20q in the subset of families with low BMI (86), and a similar subset analysis in Japanese families with BMI <30kg/cm² provided additional evidence for T2DM linkage at 20q12–q13 (D20S119 at 61.77 cM; MLS 2.32, 87). In 343 sibling pairs from West Africa, highly suggestive linkage was found at 96 cM (D20S171, lod 2.63) (94). At least two strong candidate genes (PTPN1 and HNF4A) show evidence for an association with T2DM, and are reviewed below. However, additional susceptibility genes seem likely to explain this well replicated linkage across a broad region of chromosome 20.

Chromosome12

A T2DM susceptibility loci on chromosome 12q near HNF1α was reported first by Mahtani et al (96) in a subset analysis of families from Western Finland that had the lowest mean insulin levels. In an extension of this initial study by using 32 additional families, for a total of 54 multiplex families, Lindgren et al replicated the initial findings at chromosome 12q24 (97), and support a locus on 12q near 130–150cM. In the West African study of affected sibling pairs, a T2DM susceptibility locus was also reported at 12q24 (D12S2070 at 125.31cM; LOD=1.92, 94), and Ehm et al reported a locus on chromosome 12q in the Caucasian subset of families collected under the American Diabetes Association’s Genetics of NIDDM (GENNID) study, albeit somewhat centromeric to the earlier findings at 95 cM (lod 2.81, 98). Other reports of linkage provide additional support, albeit many findings are to a more centromeric region on chromosome 12 (50–100cM, 89,99–102). These findings suggest substantial replication to chromosome 12, but the gene or genes responsible remain unknown and the actual number of loci is uncertain.

Chromosome 1q

The best replicated locus at 1q21–q23 was initially identified independently by our laboratory in North European Caucasians (85) and in Pima Indians with onset before age 25 years (103). Subsequent replications of this region included Amish (104), French (105), United Kingdom (106), US Caucasians (Framingham, 107), two Chinese populations (108,109) and Arkansas African-American families (unpublished). In contrast to chromosomes 20 and 12, the chromosome 1 linkage peaks overlap in a relatively narrow region extending from 165 cM to 200 cM (110). This well replicated region of T2DM linkage is characterized by an extraordinary gene density, including a plethora of strong candidate genes. We recently reported that this region encompasses at least two and possibly three linkage peaks (111). We have reported that two genes in this region, liver pyruvate kinase (PKLR, 112) and calsequestrin1 (CASQ1, 113), were associated with T2DM. Both have been observed in other populations (114), and support a model of multiple susceptibility loci accounting for the chromosome 1 linkage signal. Work is ongoing to identify additional susceptibility genes in this region as an international collaborative effort.

Chromosome 18p

A locus on 18p (0–20cM) is likewise now well replicated. This region was suggestive in a scan from our laboratory (85). A subgroup analysis of the 20% of the most obese subjects from a study of 480 Finnish-Swedish Caucasian affected sib-pairs showed linkage to 18p11 (lod= 3.82 at D18S976 – D18S391, 115). Two subsequent studies from the Netherlands also suggested a locus that interacted with obesity. Aulchenko et al (116) reported a lod score of 2.3 at marker D18S63. When association was tested, the most significant findings were for subjects with high BMI. In a study from van Tilburg and colleagues (117), the linkage was again for the 20% of the most obese pedigrees at 18p11, between markers D18S471 and D18S843. Hence, the putative T2DM susceptibility locus at 18p11 is replicating well, but may be of most importance in interaction with obesity.

Chromosomes 3q, 11q and 8p

Several other loci show reasonable evidence for replication, and many cross ethnic groups. A locus at 3q22–q27 near the adiponectin locus (165–220cM) has replicated well, and adiponectin itself may be partially responsible (88,105,118). A broad region of 11q has been implicated in several studies, and obesity also maps to this region (95,103,119,120). In Pima Indians, the strongest 11q linkage was for a bivariate trait of obesity (BMI) and T2DM (103). Finally, studies in British families, Utah families of European descent, and indigenous Australian populations suggest an 8^th possible locus at 8p21–p23 (85,88,106). With the exception of adiponectin, genes accounting for these linkage signals have not been identified.

Genes Identified by Linkage Studies

Horikawa Y and colleagues (121) used a relatively sparse linkage disequilibrium map of SNPs in the region of linkage on chromosome 2q to identify 3 common intronic variants of a previously unknown gene. The gene, calpain 10 (CAPN10), belonged to the family of calcium-activated, non-lysosomal thiol proteinases. Both a single intronic variant (UCSNP-43: G to A) and a specific haplotype combination defined by three polymorphisms (UCSNP-43, -19, and -63) were associated with T2DM in Mexican Americans, with lesser evidence of an association in a Northern European population from the Botnia region of Finland (121). Surprisingly, individuals with a combination of two different haplotypes (121/112) were at the highest risk of T2DM. In further support of CAPN10 as a susceptibility gene, Baier and colleagues (122) showed altered gene transcription and reduced muscle mRNA levels in muscle biopsies from Pima Indians with T2DM. Recent studies provide further support for a role of CAPN10 in T2DM, even among European derived populations (123). SNP-44 (designated “CAPN10-g4841TtoC”; minor allele frequency 16%), located in intron 3 and 11 bp from SNP-43, was independently associated with T2DM in several populations, and was in linkage disequilibrium with the missense mutation Thr504Ala and two 5′-UTR variants (SNP-134 and SNP-135). An initial meta-analysis also supported an association with T2DM (124), but a subsequent meta-analysis of over 5000 cases and 5000 controls suggested an effect size that was clearly less than suggested in the first studies (125) and raised questions about the role in T2DM. Nonetheless, we and others have found support for a role of CAPN10 in the association with intermediate, quantitative traits (126;127). Furthermore, CAPN10 variants were valuable along with traditional risk factors in predicting the onset of T2DM in a prospective study of Botnian Finnish individuals (128).

Although initial studies of the HNF4α gene, an obvious candidate on chromosome 20, failed to find any association with T2DM (92), subsequent studies based on the discovery of an upstream, β-cell specific promoter and first exon prompted a reevaluation. Two studies found evidence for association with T2DM that could partially explain the linkage signal (129,130). HNF4A has a complex expression pattern, which includes elaborate alternative splicing, and is expressed in many tissues, including the liver and pancreas. Three of the isoforms are transcribed by an alternative P2 promoter, located ~46 kb upstream of the P1 promoter and the rest of the coding exons. Transcripts from both the P1 and P2 promoters have been detected in pancreatic β-cells, but the P2 promoter is suggested to be the primary transcription start site in these cells. In total, 4 SNPs flanking the P2 promoter were associated with T2DM in both populations. Subsequent replication has been inconsistent. Vaxillaire et al failed to replicate in a French case control study (131). Hansen and colleagues found only modest evidence for one of 4 SNPs originally reported (p=0.02, OR=1.15) in a large Danish study (132), and Weedon et al found a similar effect in a large British study (5256 subjects; p=0.02, OR=1.15, 133). Damcott et al reported replication in the Old Order Amish (134), and Muller and colleagues reported replication in Pima Indians with OR 1.3 (135). Winckler and colleagues, on the other hand, found no replication in 7883 Caucasians from several populations (136). Although these SNPs appear to have a role in T2DM susceptibility, the effect is likely much smaller than originally reported, and likely similar to that of CAPN10.

The PTPN1 gene (chromosome 20q13) codes for protein tyrosine phosphatase 1B (PTP1B), which negatively regulates insulin signaling by dephosphorylating the phosphotyrosine residues of the insulin receptor kinase activation domain. Multiple noncoding SNPs in the PTPN1 gene (PTP1B) on chromosome 20q13 were recently implicated in T2DM in Caucasian and Mexican-American populations (137,138). The association was observed between multiple SNPs and T2DM, with the most consistent evidence for association occurring with SNPs spanning the 3′ end of intron 1 of PTPN1 through intron 8 (p values ranging from 0.043 to 0.004 in one case-control set and 0.038–0.002 in a second case-control set). All of the associated SNPs were present in a single 100-kb haplotype block that encompassed the PTPN1 gene. Haplotype frequencies were significantly different between T2DM case and control subjects (p=0.0035–0.0056), with a single common haplotype (36%) contributing strongly to the evidence for association, with OR 1.3. Furthermore, the same haplotypes were associated with glucose homeostasis measures in 811 Hispanic subjects (138). However, Florez et al failed to replicate the findings in a large Caucasian study (139). Further study is needed to determine the role of the strong candidate in a region of replicated linkage.

The search for diabetes susceptibility genes on most other chromosomes is ongoing. On chromosome 1, we initially described an association of polymorphisms of the liver pyruvate kinase gene with T2DM (112); subsequent preliminary reports have confirmed that association in Older Order Amish (140) and in other populations with chromosome 1 linkage (141). The region has extended linkage disequilibrium, such that several potential candidate genes are included within the two associated haplotypes. Work is ongoing in our laboratory and by others to identify the responsible variants. Closer to the most prominent chromosome 1 linkage peak, we and Fu et al independently showed an association of noncoding variants of calsequestrin gene (CASQ1) with T2DM in populations of European ancestry (113,114). The effect size is likely similar to that of CAPN10 or HNF4A. The role of this muscle gene, which is not an obvious candidate for T2DM, must await replication in other populations, however. Adiponectin is a strong candidate for T2DM given the clear role of plasma adiponectin levels in insulin sensitivity and the fall in adiponectin levels with obesity and T2DM. Adiponectin maps to the region of replicated linkage on chromosome 3q. A very large number of studies have examined adiponectin variants with metabolic traits, T2DM, and risk of conversion of impaired glucose tolerance to T2DM. Multiple studies found evidence that promoter variants in the APM1 gene were associated with T2DM in French and Swedish Caucasian and Japanese populations, and some data support a role of the APM1 locus in controlling adiponectin levels (142). However, APM1 variants did not appear to account for the 3q27 linkage in French families (142). Zacharova and colleagues showed that APM1 variation at positions 45 (G allele) in exon 2 and 276 in intron 2 (T allele) were associated with a 4.5 fold increased risk of converting from impaired glucose tolerance to T2DM in the STOP-NIDDM trial (143). On the other hand, no association was observed in Pima Indians (144). Larger studies are needed to determine the magnitude of risk imparted by APM1 variation.

Grant and colleagues (145) searched for a T2DM susceptibility gene under the suggestive linkage peak on chromosome 10q in an Icelandic population. They identified a 6 allele simple tandem repeat (STR) polymorphism in intron 3 of the transcription factor 7 like 2 gene (TCF7L2, formerly TCF4). Impressive associations were found in a Danish population and a United States Caucasian population, with a relative risk of 1.45 for heterozygotes and 2.41 in homozygotes. Although the exact role of TCF7L2 is unknown, it appears to act through the Wnt pathway to regulate proglucagon gene expression in enteroendocrine cells. The authors calculated a relatively high 27% population attributable risk (the proportion of T2DM that would not be present if the gene could be removed from the population). Two additional noncoding SNPs were closely correlated with the simple tandem repeat composite risk allele. Additional replication is needed to determine the real role of this novel gene.

Despite initial identification through a linkage signal, which would be expected to select relatively strong effects, each of these variants reviewed above has only modest individual effect (Odds Ratio, OR<1.4). Indeed, each with further study has shown much smaller effect than in the initial report – a phenomenon named the “winner’s curse”. Available data suggest that 1) the effects of individual variants are likely small; 2) multiple variants probably contribute to replicated linkage signals, as we have demonstrated on chromosome 1q; and 3) most variants are noncoding and regulatory rather than altering protein structure. We predict that each linkage region will in fact harbor multiple disease genes of small effect, and that no single variant will explain the linkage peak.

Genes Discovered by the Candidate Gene Approach

Extensive studies have been conducted of genes identified through physiologic pathways. Consistent with the very large numbers of studies, many have shown modest associations in single populations, but these associations have not yet been replicated. We make no attempt to provide a comprehensive review of the hundreds of genes examined by investigators around the world, and instead focus on genes for which reasonable replication has been achieved already. In each case, like those genes identified through linkage, the effect of these variants is relatively small, with OR<1.4 and in most cases approaching 1.2 or less. Perhaps as a bias of candidate gene studies, these variants are more often coding.

The peroxisome proliferator activated receptor gamma (PPARγ) is a member of nuclear hormone receptor superfamily of transcription factors, and the target of the widely used class of insulin sensitizers, the thiazolidinediones. Multiple studies have examined a common amino acid change in the PPARγ2 isoform in which alanine is substituted for proline at position 12 (P12A). Although smaller studies provided inconsistent results, larger studies have shown a consistently protective effect of the rare, alanine allele (A/A) when compared with the common proline allele. Hence, the common P12 allele consistently increases T2DM risk by an OR of 1.2–1.3 in cross sectional studies (146). Furthermore, in the Nurses Health Study, the OR of carrying the Ala12 allele was 0.72, again suggesting a protective role (147). Several, but not all studies suggest that the rare “A” allele protects against insulin resistance and obesity, but the association with obesity has been inconsistent. In studies of families from Utah, we found an association with several “metabolic syndrome” traits, including BMI, serum total cholesterol levels, triglyceride levels, systolic and diastolic blood pressures, and glucose concentrations. Gene-environment interactions have been suggested, and may explain the many inconsistent results. Because the common allele is also the risk allele for T2DM, and this allele has a frequency of at least 75% in most populations, the population attributable risk is again extremely high at 25%. Hence, the small individual effect of this variant translates into a large effect in the context of the population.

The β-cell potassium channel comprises two subunits, the potassium channel encoded by the gene KCNJ11, and the regulatory subunit (SUR1 encoded by ABCC8) that binds sulfonylureas and ATP. Variants in both subunits have been associated with T2DM. Two variants in ABCC8 were initially associated in smaller studies (148–150), and larger subsequent studies appeared to confirm the association (151,152). Other studies have not confirmed the association, however (153,154). We (155) and others (156) have shown altered insulin secretion associated with these variants. The adjacent KCNJ11 gene was initially noted to have multiple nonsynonymous coding variants (157), but none were associated with T2DM in this or other early studies (146,158). However, a coding glutamine to lysine at position 23 in the single exon (E23K) was shown to lower the sensitivity of the potassium channel to ATP and thus reduce insulin secretion in vitro (159). Gloyn et al subsequently showed a modest association of the E23K variant with T2DM in nearly 2500 British case-control subjects (153), and in a meta-analysis confirmed association with an OR of 1.16. More recently Florez and colleagues (154) thoroughly examined the two subunits (KCNJ11, ABCC8), and tested sufficient variants to completely tag all variation in this region. Again, the E23K variant was associated with T2DM in 3,413 subjects, and the association was confirmed in a meta-analysis that included over 5000 T2DM subjects and 4747 controls (p<10⁻⁵, OR 1.15). The E23K variant is in nearly complete linkage disequilibrium (completely correlated) with a Serine to Alanine change at position 1369 in exon 33 of the sulfonylurea receptor, and thus S1369A and E23K cannot be distinguished genetically. Nonetheless, E23K appears to have a role comparable to the PPARG P12A polymorphism. The appropriate genetic model is close to recessive, in which the relatively rare K23 homozygotes have the highest risk, but E23K heterozygotes are at some increased risk. Some (160) but not all (161) studies have found an effect of the K allele on insulin secretion.

Other variants in candidate genes have been more difficult to establish. The insulin gene variable tandem repeat polymorphism, which has clearly been associated with type 1 diabetes, and been controversial for over 20 years in T2DM (162). Whereas considerable evidence supports an association with obesity and birth weight (163–165), the association with T2DM is inconsistent (165–167). Similarly, both physiologic and genetic data are inconsistent for the glycine to arginine (G972R) variant of the insulin receptor substrate 1(IRS1) gene (168,169), now more correctly referred to as G971R. Despite evidence that this variant alters insulin action on IRS1 (169), the association with T2DM has been inconsistently replicated. Two recent large studies have revisited this variant. Zeggini and colleagues (170) examined 971 cases and 1257 controls from the United Kingdom, and found no association. However, a meta-analysis of 31 studies that included 5104 cases and 7,418 controls showed a modest association (p=0.025) with an OR of 1.15, suggesting that the effect size was similar to or smaller than that of KCNJ11 and PPARG. Florez et al (171) three large Caucasian samples with 9000 total individuals, including 4279 cases and 3532 controls, and found no evidence for an association with T2DM. If the IRS1 G972R (G971R) variant has any role in T2DM, clearly the role is quite small and of minimal clinical significance. Similarly, the association of a glycine to serine variant at position 482 of the PGC1α?gene has been reported by several groups, but not yet well replicated. PGC1α??as reviewed above, plays key roles in liver gluconeogenesis and was implicated in array studies of diabetic and control muscle. Positive associations were reported in JapaneseT2DM (172) and Danish cases and controls (173), and subsequently some replication was seen in populations from India (174) and Korea (175). Other associations were reported with insulin resistance (176) elevated glucose in nondiabetic offspring of diabetic parents (177), and reduced insulin secretion (178), but not with metabolic syndrome (179). Additional studies are required to determine the real role of this variant in diabetes susceptibility.

Variants in the gene ectonucleotide pyrophosphatase phosphodiesterase 1 (ENPP1), also known as plasma cell glycoprotein 1, were shown recently to be associated with both childhood and adult obesity in a French population. A moderate excess of the risk haplotype was also seen in adults with T2DM compared with controls (10.7% vs 7.1%, OR 1.44), and the association was replicated in an Austrian cohort with similar frequencies. A Mantel-Haenszel analysis of 2,569 European subjects supported the association, with OR 1.56 and p=0.00002 (180). The nonsynonymous variant lysine 212 to glutamine (K121Q) has been best studied, and also shown to be associated with obesity but not diabetes in Caucasian and African American subjects ascertained from the New York Cancer Project (181). Other investigators found an association of K121Q with earlier T2DM onset and coronary disease among T2DM patients (182). The association of K121Q with T2DM was replicated in three relatively small populations, two of South Asian ancestry and one Caucasian (183). Although the K121Q variant appears to associate with T2DM and obesity, additional work is needed to determine the magnitude of the risk, the populations affected, and to confirm a role comparable to KCNJ11 or PPARG.