Genetic Modifiers of the Age at Diagnosis of Diabetes (MODY3) in Carriers of Hepatocyte Nuclear Factor-1α Mutations Map to Chromosomes 5p15, 9q22, and 14q24

Maturity-onset diabetes of the young (MODY) is a form of familial diabetes characterized by an early age of onset and an autosomal dominant pattern of inheritance (1). Six MODY genes have been identified to date (2–8). One MODY locus corresponds to the enzyme glucokinase, which is an essential component of the β-cell glucose sensor (2,3). The other five MODY genes are transcription factors (hepatocyte nuclear factor [HNF]-1α, HNF-4α, HNF-1β, insulin promoter factor 1 (IPF1), and NEUROD1) that are involved in islet cell differentiation and metabolism as well as insulin secretion (4–8). Most MODY cases in the U.S. are attributable to mutations in HNF-1α (9). This form of diabetes, also known as MODY3, is characterized by impaired insulin secretion in response to glucose (10,11). While HNF-1α mutations are highly penetrant, there is wide variability in the severity of the disease. There is also wide variation in the age at which MODY becomes clinically manifest. Indeed, only 50–60% of HNF-1α mutation carriers develop diabetes before age 25 years (the traditional criterion for a MODY diagnosis) (9). Such variability does not appear to be related to the type (truncated versus missense) or to the domain localization of the HNF-1α mutations (9).

In a large panel of HNF-1α mutation carriers, we found that inheritance of the mutation from the mother and exposure to maternal diabetes in utero are associated with an early onset of MODY (9). In addition, we found the age of diabetes onset to be highly heritable (h² = 0.47), and when the parent-of-origin of the HNF-1α mutation was taken into account, the heritability was even greater, with familial (presumably genetic) factors accounting for 91% of the residual variability (9). Thus, the clinical manifestation of MODY3 appears to be more complex than previously thought, being influenced by both modifying genes that are inherited independently of HNF-1α as well as environmental factors. Similar phenotypic complexity has been proposed for other “simple” Mendelian disorders such as Gaucher disease (12).

To identify loci that influence MODY3 onset, we conducted a genome screen of 13 extended families in which diabetes segregates with an HNF-1α mutation. The clinical characteristics of these families have been previously described (9). Included in the screen were 87 mutation carriers with diabetes and 9 with impaired glucose tolerance (IGT). The number of individuals per family and the nature of the HNF-1α mutation segregating with diabetes in each family are summarized in Table 1. All 96 study subjects were genotyped by the Mammalian Genotyping Service (Marshfield, WI) for 377 autosomal markers corresponding to a 9.3 cM average spacing. Linkage with age at onset of diabetes was evaluated by variance component analysis using SOLAR (13). Because of its significant effect on MODY onset (9), parent-of-origin of the HNF-1α mutation was included as a covariate in the analysis. Results are reported in Fig. 1 and summarized in Table 2. The locus with the highest multipoint logarithm of odds (LOD) score for age at onset of diabetes was on chromosome 14q24 at 82 cM from p-ter on the Marshfield map (D14S588; LOD = 2.58, empirical P = 0.0004) (Table 2). The LOD-1 support interval extends from 68 to 102 cM from p-ter. Two other locations supported linkage, with LOD scores >2. One site occurred on 5p15 (D5S817; LOD = 2.44; P = 0.0004) and the other on 9q22 (D9S910; LOD = 2.02; P = 0.0018), with the LOD-1 support interval extending from 9 to 41 cM and from 93 to 123 cM for chromosomes 5 and chromosome 9, respectively. Less strong linkage evidence was observed at 3p24 (LOD = 1.44, P = 0.008), 7q21 (LOD = 1.20, P = 0.016), and 16q23 (LOD = 1.51, P = 0.007) (Table 2). To account for the fact that the diagnosis of IGT may precede the diagnosis of diabetes by several years, linkage analysis was repeated after increasing the age at diagnosis of the nine IGT subjects by 5–10 years, as described in research design and methods. The evidence of linkage remained unmodified at 14q24 (LOD = 2.61, P = 0.0003), increased at 9q22 (LOD = 2.42, P = 0.0009), and decreased at 5p15 (LOD = 1.78, P = 0.0043). There were no major changes in the LOD scores at 3p24, 7q21, and 16q23.

In the November 2002 Genome Assembly (www.genome.ucsc.edu), the LOD-1 support interval for the diabetes age-at-onset locus on 14q24 corresponds to 27 Mb (from 57 to 84 Mb). This region partially overlaps with the type 1 diabetes locus (IDDM11, 80–84 Mb) that was identified by Field et al. (14) in a subset of families that had no evidence of linkage to HLA. A type 2 diabetes locus has also recently been described on 14q in an Amish population, but this is placed >50 cM centromeric to our peak (15). According to the latest annotation of the human genome sequence, the 14q24 LOD-1 support interval contains at least 128 genes, many of which could potentially affect glucose metabolism and, in turn, modify the age at onset of MODY. Among these, one interesting candidate is SEL1L, a gene that is almost exclusively expressed in the pancreas and islets of Langerhans (16,17). The product of SEL1L is a negative regulator of the Notch signaling pathway (17), which controls cell fate decisions in many different settings during development (18). In the pancreas, the Notch pathway moderates islet formation by inhibiting the expression of the HNF-1α-activated pro-endocrine factor neurogenin 3 (Ngn3) (19). Thus, a genetic defect of the Notch inhibitor SEL1L would result in increased activity of the Notch pathway and worsen the Ngn3 deficiency caused by HNF-1α mutations. Presumably, the result of this defect would be a more profound deficit of β-cell mass (or function) leading to an earlier development of diabetes. Two nonsynonymous polymorphisms are in the dbSNP database for SEL1L (rs2371800 [E182K] and rs1051193 [V714I]). However, after genotyping all 96 HNF-1α mutation carriers for these two single nucleotide polymorphisms, only one heterozygote for rs1051193 (and none for rs2371800) was observed. Both exons and 5′ flanking sequence of the gene are being screened to identify additional polymorphisms that may be associated with a young age at onset of MODY in these families.

The peak on chromosome 9q22 corresponds to a locus linked with 2-h oral glucose tolerance test (OGTT) insulin concentrations in Pima Indians (20). The 2-h insulin level is a widely accepted index of insulin resistance in target organs such as skeletal muscle, adipose tissue, and liver. In the presence of insulin resistance alleles at 9q22, carriers of HNF-1α mutations would require more insulin from β-cells to maintain normoglycemia and may therefore develop MODY earlier in life. The LOD-1 support interval at this location spans 27 Mb (from 82.5 to 109.5). Of the 140 genes included in this region, two are known to be directly involved in glucose metabolism and insulin action. One gene codes for fructose-1,6 bisphosphatase, a rate limiting enzyme in gluconeogenesis. The other gene is the glycolytic enzyme aldolase B. Deletion and nonsense mutations in this enzyme are a cause of hereditary fructose intolerance (21). Perhaps less severe polymorphisms can influence insulin sensitivity and therefore MODY onset. An effect on insulin sensitivity might also be at play for the locus on 5p15, since this exact location has been linked with circulating levels of the insulin sensitizer adiponectin (22). The LOD-1 support interval spans 27.6 Mb (from 4.1 to 31.7 Mb) and includes 32 known genes, none of which has obvious actions on glucose metabolism.

It must be emphasized that none of the LOD scores obtained in this genome screen meet the criteria for genome-wide significance. In this regard, it must be noted that for effects of the magnitude observed at chromosomes 5, 9, and 14, our family collection provided 75% power to detect linkage with LOD scores >2.5, but only 56% power to detect LOD scores >3.00. The power to detect linkage at genome-wide significance levels may have been further reduced by the fact that age at diagnosis is not a perfect surrogate of the age at onset of diabetes.

In summary, these data suggest that multiple loci, inherited independently of MODY, are modifiers of the MODY3 phenotype, although this remains to be substantiated by replication in other studies. Some of these genes may exert their effect at the level of the β-cell, enhancing the insulin secretion deficit caused by HNF-1α mutations, while others may influence the onset of MODY3 by affecting the amount of insulin that peripheral tissues require to maintain euglycemia. Identification of these genes will offer new insights to the pathophysiology of MODY that may, in turn, increase our understanding of the cellular events underlying more common forms of diabetes, particularly with regard to the role of gene-gene interactions in the development of hyperglycemia. At the same time, knowledge of these genetic modifiers will provide new tools to evaluate more accurately the prognosis for carriers of HNF-1α mutations.