Noninsulin-dependent diabetes mellitus (NIDDM)² is strongly inherited, as evidenced by a high concordance in identical twins and strong familial aggregation. In unbiased studies, the concordance in identical twins is ~70%, whereas the lifetime risk to siblings is only half this rate (Rewers and Hamman 1995). A history of NIDDM in a first-degree relative doubles the risk of diabetes. Offspring of two diabetic parents have an 80% lifetime risk of diabetes (Kenny et al. 1995, Rewers and Hamman 1995). Further evidence for a genetic role is suggested by the wide variation in incidence and prevalence among different ethnic groups. Thus, Pima Indians have a nearly 50% prevalence that is marked by a degree of insulin resistance not seen in Caucasians, and both Hispanics and African-Americans are also at high risk (Kenny et al. 1995). These facts and the development of new molecular genetic methodologies have made the etiology of NIDDM a focus for geneticists and diabetologists. The reader desiring a detailed analysis of the challenges and progress toward the goals of defining genetic susceptibility to NIDDM is referred to the many excellent reviews (Elbein et al. 1994a). This review will define the current state of the search for human NIDDM susceptibility genes and how obesity might relate both to the genetic susceptibility and to the phenotypic expression of independent NIDDM susceptibility genes.

The physiology of NIDDM has provided some clues to possible defects. Insulin resistance is nearly ubiquitous. The offspring of two diabetic parents, who historically are at particularly high risk for future NIDDM, are insulin resistant many years before NIDDM development (Warram et al. 1990). In Pima Indians, San Antonio Hispanics and Utah Caucasians, insulin resistance is inherited in an autosomal fashion, suggesting that a single gene might influence this intermediate phenotype (Bogardus et al. 1989, Schumacher et al. 1992, Stern et al. 1996). Studies of patients with NIDDM and insulin resistance have demonstrated a primary defect in insulin-mediated glucose uptake into muscle, and specifically nonoxidative uptake into glycogen (DeFronzo et al. 1992, Rothman et al. 1995, Vaag et al. 1992 ). Insulin sensitivity is reduced 50% on average in NIDDM, and glycogen synthesis and storage are reduced by 50-70% (DeFronzo et al. 1992). Like insulin resistance in general, this specific defect in glycogen synthesis is present in first-degree family members of an NIDDM proband (Vaag et al. 1992).

Individuals at risk for NIDDM also show various forms of pancreatic -cell dysfunction (Polonsky et al. 1996). Those with both impaired glucose tolerance (IGT) and NIDDM demonstrate loss of first-phase insulin release in response to an IV glucose load (S. C. Elbein, unpublished data). Although this defect may result from glucotoxicity (Rossetti et al. 1990), more subtle defects of insulin secretion are present in family members at risk. Thus, the normal cyclic patterns of insulin release are lost in both NIDDM subjects and in relatives at risk (Polonsky et al. 1996). An increased proinsulin:insulin ratio is present in NIDDM, the intermediate and high risk stage of IGT, and in euglycemic relatives at risk (Kahn et al. 1995). Among Pima Indians, the risk of future NIDDM is highest among those with both insulin resistance and a defect in insulin secretion (Lillioja et al. 1993). Defects of both insulin secretion and insulin sensitivity are apparent in offspring of NIDDM parents in Utah Caucasian pedigrees ascertained for multiple NIDDM siblings when insulin secretion is appropriately normalized for the degree of insulin resistance (SCE, unpublished data). Thus inherited defects of both insulin sensitivity and -cell function are likely to contribute to NIDDM susceptibility.

In contrast to defects that are at least partially inherited, NIDDM is also characterized by increased hepatic glucose production from both glycogenolysis and gluconeogenesis (DeFronzo et al. 1992). This defect appears relatively late in the course of NIDDM development in most studies and thus is less likely to result directly from genetic susceptibility. On the other hand, like the defect in insulin sensitivity, this defect may be closely related to obesity. Both peripheral insulin resistance and the increased hepatic glucose output may result from high levels of circulating free (nonesterified) fatty acids (FFA). These high levels of FFA in turn appear to relate to the amount of visceral fat (Boden 1997). Thus, inherited susceptibility to obesity and particularly visceral obesity may contribute to both the hepatic and peripheral defects seen in NIDDM. Finally, both individuals with NIDDM and offspring of two NIDDM parents have reduced glucose effectiveness, i.e., the ability of glucose to mediate its own uptake independent of insulin. The potentially complex model of environmental and genetic susceptibility to NIDDM is summarized in Figure 1.

The role of obesity in the etiology of NIDDM is complex. In epidemiologic studies, obesity alone increases the risk of NIDDM twofold but interacts with a family history of NIDDM to increase the risk for obese patients with a family history of NIDDM fourfold ( Kenny et al. 1995). Although obesity appears to exert its effects by reducing insulin sensitivity, this relationship is complex. When obesity is measured as body mass index (BMI), a nonlinear relationship is seen between BMI and the insulin sensitivity index (S_I), a measure of insulin action. As BMI increases, S_I decreases in a nonlinear fashion, but variance also decreases (Kahn et al. 1993). Furthermore, as stated in S. C. Elbein's unpublished data, the insulin secretory response when normalized for S_I was low in obese (BMI > 30 kg/m²) members of families ascertained for two NIDDM siblings. Thus, obesity as measured by BMI may contribute to both the insulin resistance and the insulin secretory effects seen in NIDDM. The role of visceral fat in these measures is less well studied.

Although it is clear that obesity affects NIDDM susceptibility, the interaction between NIDDM susceptibility loci and obesity is uncertain. Obesity may mark a unique subset of NIDDM patients with more profound insulin resistance and a relatively mild pancreatic -cell defect. Alternatively, obesity may act equally in all susceptible individuals to increase the penetrance of genetic loci. The latter appears to be the case in some single-gene disorders such as glucokinase mutations. Finally, obesity may decrease the age of onset of NIDDM in susceptible family members in a manner analogous to pregnancy and gestational diabetes. Evidence exists for all three roles.

The genetics of NIDDM has been studied by several methods. Until recently, most studies compared the frequency of a polymorphic marker near or within a gene of potential interest in diabetic and nondiabetic individuals. These studies have implicated several loci, including the muscle glycogen synthase gene in Finnish diabetics (Groop et al. 1993), an amino acid variant of the glucagon receptor gene in French and Sardinian diabetics (Hager et al. 1995), and the insulin receptor substrate 1 (IRS-1) (Elbein et al. 1994a). Similar studies were used to suggest an association of the -3 adrenergic receptor gene with obesity and earlier onset of NIDDM (Clement et al. 1995, Walston et al. 1995, Widen et al. 1995). Recent studies from our laboratory, in collaboration with the laboratories of Permutt and Turner, found an association of a silent change in the sulfonylurea receptor gene and NIDDM (Inoue et al. 1996). The well-described artifacts of association due to population stratification constitute a potential weakness of association studies. A second problem with published studies is the tendency to search for multiple associations from multiple laboratories without appropriate correction in reporting significance levels. In general, these studies have at best defined weak susceptibility loci that have been difficult to confirm. Furthermore, in the case of glycogen synthase (Orho et al. 1995) and the sulfonylurea receptor gene (Inoue et al. 1997), among others, the molecular defect that might explain the association has not been found.

Because studies of candidate genes for NIDDM have been largely negative (see below), investigators have begun using the human linkage map to search for unknown susceptibility genes. Several statistical and design methods have been applied to families with multiple diabetic siblings. All use highly polymorphic markers that do not themselves cause disease and that are not usually within genes. The sharing of alleles designated by these markers is examined among affected individuals, and significant deviation from chance sharing is interpreted as evidence for a susceptibility gene in close proximity to the genetic marker. Although most studies of NIDDM have applied this method to sharing between affected siblings (affected sibling pair analysis, or ASP), we and others have applied this method to extended families using multiple single-gene models that might approximate true NIDDM inheritance. Another method of study is to examine quantitative measures of NIDDM risk among individuals. In quantitative trait locus (QTL) analysis of humans, as in animals, a locus is implicated by the significant correlation of alleles shared at a locus between siblings with the square of the difference in the trait values between those siblings. Each of these methods has been applied to candidate genes, i.e., those genes that are in pathways suspected to influence glucose homeostasis. However, greater success has derived from studies of anonymous markers placed at regular intervals throughout the human gene map. These studies, known as the "genome-wide scan," test markers at set (usually 10- to 15-cM) intervals on all human autosomes. All current studies use markers with multiple alleles determined by short DNA sequence variations and detected by polymerase chain reaction. Most recently, statistical methods and computer programs have permitted these adjacent markers to be considered simultaneously (multipoint analysis) with the potential for considerably increased power to find susceptibility loci (Kruglyak and Lander 1995, Kruglyak et al. 1996).

Most forms of NIDDM do not follow simple Mendelian inheritance, and thus are considered among the complex genetic diseases. In contrast, NIDDM with onset before age 25 is unique in that many families show autosomal dominant inheritance. Unlike most NIDDM, these individuals are generally not obese and they have predominant defects in insulin secretion with normal insulin sensitivity. Three loci have been defined. MODY1, localized to chromosome 20 by a genome-wide search in a single large family, has now been traced to a nonsense mutation (Q268X) in the hepatocyte nuclear transcription factor 4  (NF4) (Yamagata et al. 1996a). This rare mutation in a member of the steroid/thyroid hormone receptor superfamily causes severe diabetes with all of the complications of typical NIDDM. Presumably the HNF4 acts through the regulation of a downstream transcription factor, hepatocyte nuclear transcription factor 1 alpha (HNF1) (see below), but the precise mechanism is unknown. MODY2 was identified by linkage studies of a candidate gene, glucokinase (GCK). Glucokinase mutations account for 50% of maturity onset diabetes of the young (MODY) in France and have been found in English families (Froguel et al 1993). They are relatively rare elsewhere and do not contribute to late onset NIDDM (Elbein et al. 1993 and 1994b). Mutations are found throughout the gene and appear to alter the glucose set point for insulin secretion. Insulin secretin rates in response to glucose infusion are diminished (Polonsky et al. 1996). Clinically, patients have mild fasting hyperglycemia and little risk of complications, although diabetes may be more severe in the presence of obesity.

MODY3 may be the most important single-gene cause of NIDDM. The locus was initially mapped to chromosome 12q by linkage analysis of nonglucokinase-linked French MODY families (Vaxillaire et al. 1995). It accounts for 25% of French MODY diabetes. Recently Bell and colleagues identified the locus as HNF1 (Yamagata et al. 1996b). A large number of mutations have been identified in English, French, German, Finnish and American families. Unlike glucokinase, the same mutations have occurred in several unrelated families, and mutations appear to be clustered in several exons (Froguel 1996). The role of this locus in more typical NIDDM is unclear, but late onset of NIDDM has been noted in some family members who carry the mutation. Furthermore, HNF1 mutations may be present in individuals with early onset (before age 40) NIDDM who do not meet the usual criteria for MODY (Dussoix et al. 1997, Kaisaki et al. 1997).

Studies of typical NIDDM have focused heavily on candidate genes implicated from the known physiology of NIDDM. These can be divided into studies of genes for insulin action and genes that control pathways of insulin secretion. As discussed above, candidate genes have been studied by association and linkage studies. Many candidate genes also have been examined directly by molecular screening techniques. Although a detailed discussion is beyond the intent of this review, a nonexhaustive list of genes examined is shown in Table 1. In general, no single locus that can explain the genetic propensity to either NIDDM or insulin resistance has been identified among known candidate genes. Although there is some controversy regarding the roles of IRS1 and fatty acid binding protein 2 (FABP2; see below), there is no consistent evidence to implicate these loci in NIDDM pathogenesis when all studies are considered. There is more evidence for a role of genes for insulin secretion (Table 2); however, no locus has been definitely demonstrated to be an important cause of typical late onset NIDDM, although the role of MODY3 mutations in some subsets of NIDDM is uncertain (Mahtani et al. 1996). The sulfonylurea receptor has been implicated in association studies in Caucasian populations from Utah and the United Kingdom (Inoue et al. 1996), but is not linked to NIDDM in several family studies (Elbein and Hoffman 1996, Stirling et al. 1995), and no disease-causing mutation has been identified in either the sulfonylurea receptor or the contiguous potassium channel gene (Inoue et al. 1996 and 1997).

Table 1. Candidate genes for insulin action [View Table]

Table 2. Candidate genes for insulin secretion [View Table]

Although success has been limited in finding a role for known loci in NIDDM susceptibility, genetic subgroups of NIDDM have been defined. Although rare, insulin gene mutations causing uncleaved proinsulin or fully cleaved insulin with reduced activity have been described in a small number of families (Steiner et al. 1990). Interestingly, diabetes is a variable feature of these autosomal dominant mutations, suggesting a relatively low penetrance of NIDDM in homozygotes. Both autosomal dominant and recessive forms of insulin receptor mutations have contributed to diabetes in some families, although again the dominant feature is hyperinsulinemia (Taylor et al. 1990). Mutations causing decreased insulin binding to the receptor or diminished tyrosine kinase activity are an unusual cause of insulin-resistant diabetes. Mitochondrial DNA mutations result in insulin-deficient diabetes in up to 1% of NIDDM in some populations (Elbein and Hoffman 1996, Elbein et al. 1994a). This disorder, usually resulting from a mutation of mitochondrial leucine tRNA, has been associated with maternal transmission and sensorineural hearing loss.

Genome-wide screens of families with multiple NIDDM siblings are in progress in several laboratories, including ours. Hanis and colleagues initially reported linkage to a marker near the telomere of chromosome 2q with NIDDM (NIDDM1) in Hispanics from Starr County, Texas (Hanis et al. 1996). Other published studies (Mahtani et al. 1996, Stern et al. 1996) and unpublished work in Pima Indians and in our laboratory have failed to confirm evidence for linkage, however. Subsequently, Mahtani et al. (1996) reported a second locus at or near MODY3 that they identified in a small subgroup of Botnian Finnish families in the lowest quartile for insulin secretion. They found no evidence for a major NIDDM locus in the remaining 75% of 26 Botnian families. This analysis has not been repeated, although most groups find no evidence for linkage when families are not stratified by insulin secretion. A study of 32 Hispanic families in San Antonio (Stern et al. 1996) used somewhat different methods based on glucose as a quantitative trait. They reported suggestive linkage to regions of chromosome 6 and 11, but no linkage to NIDDM1 or NIDDM2.

Our laboratory has completed studies of 19 Caucasian families (469 family members) from Utah. Our families were selected for at least two NIDDM siblings with onset before age 65. All available siblings and offspring of diabetic individuals were screened by a 75-g oral glucose tolerance test if not known to be diabetic. We have estimated marker allele frequencies from 100 unrelated spouses. Regions of possible linkage were examined in a total of 42 families; more recently we have entered an additional 20 families into the study. We also found no linkage to NIDDM1 or NIDDM2, and no region has met current genome-wide proposals for significant linkage. We have analyzed our linkage data using both sibling pair and parametric ("lod score") linkage methods, and both two-point and multipoint analysis. To date, our highest two-point significance level (lod score) is for linkage to a region of chromosome 7 also reported to be linked to NIDDM in Pima Indians, but our recent analyses suggest that this locus, if real, is a relatively weak susceptibility locus. Our multipoint analysis provided suggestive evidence for linkage near the apolipoprotein A2 region of chromosome 1. These results have not been reported in other studies, however. Work is in progress in our laboratory to extend these findings.

Because NIDDM may represent the final outcome of several converging intermediate traits, many groups have attempted to study the intermediate traits directly as QTL. Among the traits studies are insulin sensitivity, insulin levels, fasting and postchallenge glucose levels and measures of obesity. In Pima Indians, insulin resistance as measured by insulin clamp studies was linked to the intestinal fatty acid binding protein locus (FABP2) on chromosome 4 (Prochazka et al. 1993). Subsequently, this group identified an amino acid variant of FABP2 that may partially account for insulin resistance in this population (Baier et al. 1995). The same group mapped a locus for acute insulin response to glucose to a marker on chromosome 1p (D1S198; Thompson et al. 1995), although these findings are without published confirmation. In preliminary work from our laboratory, we were unable to identify a locus for fasting or postchallenge insulin levels despite evidence for Mendelian inheritance of both fasting and 1-h postchallenge insulin in Utah families (Schumacher et al. 1992). However,we did map fasting glucose to a locus on chromosome 9 using a quantitative multipoint approach. This locus is not linked to either diabetes or postchallenge glucose.

A number of laboratories have examined the role of obesity and obesity genes in NIDDM. We and others have tested the human homologs of mouse obesity genes for linkage to NIDDM in humans (Table 3), but these studies have been uniformly negative. Among diabetes-prone populations, there has been evidence for linkage of tumor necrosis factor  (TNF) to obesity in Pima Indians (Norman et al. 1995) but not in our population. Recently, both leptin levels and fat mass were reported to be linked to a marker on chromosome 2p in 10 large Hispanic families from San Antonio (Comuzzie et al. 1997), which explained 47% of the leptin variance. The nature of this locus is unknown. In 283 Pima Indian sibling pairs, percentage of body fat was linked using similar QTL methods to chromosome 11 (D11q21-22), although this study did not meet genome-wide significance levels for proof of linkage (Norman et al. 1997). Neither study has localized obesity loci to known candidate genes. Despite initial reports for an association of a common amino acid polymorphism of the -3 adrenergic receptor gene with insulin resistance in Finns (Widen et al. 1995), earlier onset NIDDM in Pima Indians (Walston et al. 1995) and increased BMI in France (Clement et al. 1995), these associations have been difficult to confirm. We examined family members of NIDDM probands and were unable to document any role for this variant in the context of high risk families (Elbein et al. 1996).

Table 3. Mouse and human obesity loci not involved in NIDDM by linkage or molecular screening [View Table]

In collaboration with Sandy Hasstedt (Human Genetics, University of Utah), we have examined the role of obesity in our Caucasian families ascertained for at least two NIDDM siblings. We initially tested for segregation of BMI in 616 members of 42 families. We found evidence for two recessive obesity loci, one for extreme obesity with mean BMI 39 and a gene frequency of 0.42, and a second moderate obesity locus with mean BMI 32 and a gene frequency of 0.28. Together, these loci account for 68% of the variance in BMI. Homozygosity at both loci resulted in a mean BMI of 47. Both loci raised the fasting insulin and postchallenge glucose levels and resulted in earlier onset of NIDDM. Thus, the obesity loci appear to modulate the expression of the underlying NIDDM susceptibility loci. We used this model to test for linkage to a large number of candidate genes for obesity, including leptin, the leptin receptor, lipoprotein lipase, hepatic lipase, TNF, glycogen synthase, the agouti region, the Prader-Willi region, the -3 adrenergic receptor locus and the apolipoprotein genes. We found no evidence for linkage of these candidate genes to obesity under the segregation model or with the use of nonmodel-based methods.

In summary, many laboratories are currently applying genetic tools to the search for NIDDM susceptibility loci in humans. To date, success in identifying a cause for the bulk of NIDDM is limited. Initial studies from multiple ethnic groups suggest that NIDDM will result from a complex interaction of several susceptibility genes and the environment. Obesity, both genetic and environmental, will play an important role in modulating the expression of these loci. However, it seems less likely that obesity loci will define a unique subgroup of NIDDM or that unique obesity loci will be found among NIDDM families.

ACKNOWLEDGMENTS