A Hepatocyte Nuclear Factor-4α Gene (HNF4A) P2 Promoter Haplotype Linked With Late-Onset Diabetes

RESEARCH DESIGN AND METHODS

Since 1997, physicians have referred subjects to the Norwegian MODY Registry based on at least two of the following criteria: 1) first-degree relative with diabetes, 2) onset of diabetes before age 25 years in at least one family member, 3) low dose of insulin requirement (<0.5 units · kg⁻¹ · day⁻¹), 4) early-onset diabetes defined as diabetes diagnosed between age 25 and 40 years, and 5) subjects with an unusual type 1 diabetes (low-dose insulin requirement, no antibodies, or atypical history). Obviously, the conventional criteria of MODY (1) are not met in all cases. Nevertheless, inclusion of subjects based only on the conventional criteria will exclude some true MODY patients, e.g., those with de novo mutations, age at diagnosis after age 25 years, or limited clinical data on the family history of diabetes.

We sequenced HNF4A in 95 probands recruited from the Norwegian MODY Registry. Twenty-three probands fulfilled conventional MODY criteria (1), whereas the remaining 72 subjects were categorized as having suspected MODY. All probands were negative for mutations in HNF1A (13). Moreover, GCK mutations were excluded in 42 probands who had a MODY2-like phenotype. Standard oral glucose tolerance tests were performed (75 g glucose), and World Health Organization criteria for diabetes were applied. The study was approved by the regional committee for research ethics. We obtained written informed consent from the patients or their parents.

Molecular genetics.

HNF4A was sequenced using previously reported (14) or newly designed (online appendix Table 1 [available at http://diabetes.diabetesjournals.org]) primers. In particular, a new primer pair for exon 8 had to be used in order to avoid allelic drop out observed when using the published strategy (14). For the determination of allele frequencies of the novel variants and the IVS7 nt-88C>T single nucleotide polymorphism (SNP) causing allelic drop out, we sequenced the relevant exon and flanking introns directly in 100 blood donors. The P2 −192C>G SNP was genotyped in 1,006 healthy Norwegian subjects using Sequenom’s MassARRAY system. The allele frequencies of the polymorphisms were estimated in 95 MODY probands.

Linkage analysis.

Power calculation using 1,000 replicates and SLINK/MSIM software yielded maximal average logarithm of odds (LOD) scores of 4.4 and 0.3 in families N88 and N510, respectively. To assess the significance of the P2 variant cosegregating with MODY, we used a previously described parametric LOD score analysis with age-related penetrances (15) and the MLINK/VITESSE (version 1.0) program. To investigate a possibly shared haplotype, we genotyped family members with the microsatellite markers D20S855, D20S96, D20S43, D20S89, D20S119, and D20S1151. Estimation of individual haplotypes in families N88 and N510 were done using the SIMWALK (version 2.89) software, and haplotype frequencies among unrelated controls were calculated with PHASE (version 2.1.1).

Plasmid constructs.

A luciferase reporter gene construct for the P2 promoter studies was kindly provided by Dr. Gerhard Ryffel, Essen, Germany (5). We used specifically designed mutagenesis primers to create P2 promoter constructs containing the −192C>G and −50G>C sequence variants (QuickChange XL kit; Stratagene, La Jolla, CA). Sequences of the constructs were confirmed by DNA sequencing.

Cell culture and transfection studies.

Rat insulinoma INS-1 cells were transfected and assessed as previously described for HeLa cells (13). Each experiment was performed in triplicate on each of 5 consecutive experimental days. Transcriptional activation was measured as luciferase activity generated from reporter constructs transfected into INS-1 cells. We performed transfection of the cells using 4 μl FuGene6 (Roche Molecular Biochemicals, Indianapolis, IN) and a total amount of 2 μg DNA, including 0.025 μg pRLSV40 (Promega, Madison, WI) for normalization of transfection efficiencies and various amounts of wild-type or mutant constructs. The total quantity of plasmids was maintained constant by adding suitable amounts of empty pGL3-Basic (Promega) reporter vector. The cells were harvested 24 h posttransfection, and total cell extracts were prepared. The transcriptional activity was measured using the dual-luciferase reporter assay system (Promega).

Statistical analysis.

We used linear regression to estimate any difference in the luciferase activity for the wild-type and mutant constructs. We chose a significance level of 5% and used Stata (version 7.0; Stata, College Station, TX).

RESULTS

We found four previously undescribed mutations in HNF4A (Figs. 1 and Table 1), three in exon 8 (G326R, T339I, W340X), and one in the P2 promoter (P2 −192C>G). It was not possible to extend the pedigrees N511 and N528 for an evaluation of cosegregation for the G326R and W340X variants, respectively. T339I was a spontaneous mutation, as the healthy parents did not harbor the mutation and the family relationships were confirmed by microsatellite analysis. In addition, we observed the previously reported exon 7 variant V255M (16) in two families in which there was no cosegregation with diabetes (N91 and N557, data not shown). None of the putative pathogenic variants were identified in 200 control chromosomes. We also observed several polymorphisms and intron variants (online appendix Table 2).

During sequencing of exon 8, we observed an allelic drop out caused by an SNP (IVS7 nt-88C>T) in the binding site of the originally published forward primer (14). The frequency of the C allele of the IVS7 nt-88C>T SNP was 59% in 95 MODY probands and 57% in unrelated blood donors.

We observed two novel variants of the P2 promoter, P2 −192C>G (families N88 and N510; Fig. 1) and P2 −50G>C (families N515; data not shown), of which only the former showed a pattern of cosegregation with disease. The P2 −192C>G variant was not found among 1,006 control individuals. We genotyped additional family members from families N88 and N510 for the P2 −192C>G variant. By linkage analysis, we found a maximum two-point LOD score of 3.1 and 0.80, respectively, in the two families (online appendix Table 3). Haplotype determination using microsatellite markers surrounding the HNF4A locus showed that the N88 and N510 families shared a minimal haplotype of <3.7 Mb flanked by the markers D20S855 and D20S89 (Fig. 2A). Furthermore, this haplotype was also shared with two Danish families harboring the same P2 −192C>G variant (families DK-657 and DK-515; Fig. 2A) (17). The same microsatellite marker haplotype was not present among 20 Norwegian blood donors. In contrast, on constructing a haplotype of closely spaced SNPs spanning HNF4A and the P2 promoter region, we found that the P2 −192C>G variant resides on the most common HNF4A SNP haplotype in Norway (Fig. 2B).

An elevated age at onset of diabetes was the most distinguishing clinical feature of subjects with the P2 −192C>G variant (median onset of 45 years) compared with subjects with coding variants. In addition, the subjects were generally lean (Table 2). Since sulfonylurea has been reported to be effective in MODY1 patients (18), we noted that among the subjects with the P2 promoter variant P2 −192C>G, five subjects were treated with sulfonylurea and used on average a dose of 43% of the maximum licensed dose (with a mean level of HbA_1c [A1C] 7.7%). Among the subjects with coding variants, one subject was using sulfonylurea, requiring 8% of the maximum licensed dose (A1C 9.2%). We found no significant difference in transcriptional activity between the wild-type and either of the two P2 promoter variants when they were coupled to a reporter gene and tested in INS-1 cells (online appendix figure). We found that disease-associated variants in the coding and promoter regions of HNF4A account for 9% (2 of 23) of MODY3-negative probands meeting conventional criteria and 4% (3 of 72) among the probands with suspected MODY, based on the putative causal effects of P2 −192C>G, G326R, T339I, and W340X (Table 1).

DISCUSSION

Mutation screening of HNF4A in 95 probands recruited from the Norwegian MODY Registry revealed three coding mutations in exon 8 (G326R, T339I, and W340X) not previously reported, as well as a previously undescribed disease-associated variant of the P2 promoter. Cosegregation was not possible to evaluate for G326R and W340X. The T339I was a de novo mutation and hence the second spontaneous mutation to be reported in MODY1 (18). G326R, T339I, and W340X were all located in a genomic region encoding the transactivating domain of the protein, and they affected residues strictly conserved in the human, mouse, rat, and Drosophila HNF-4α homologues. Moreover, they were not found in 200 control chromosomes. Thus, these variants are likely to cause monogenic diabetes. In our material, variants in the coding and promoter regions of HNF4A causing monogenic diabetes account for 9% of MODY3-negative probands meeting conventional criteria. We therefore consider, in Norway, MODY1 to be much less common than MODY3 and consider MODY2 to be the third most prevalent type (13; unpublished data).

Due to a highly frequent SNP in a primer binding site, we observed an allelic drop out (19) when sequencing exon 8 with the originally suggested primer pairs (14). Resequencing should therefore be considered in patients tested with the previously published primers. Notably, all the three novel coding variants that we observed occurred in exon 8.

We found two single nucleotide substitutions in the P2 promoter region. The P2 −192C>G variant cosegregated with diabetes in two Norwegian families (maximal LOD score 3.1 and 0.80, respectively). The P2 −192C>G variant is located 10 bases upstream of the HNF-1 binding site and is conserved in the corresponding position in mouse. Interestingly, the same variant was also found in five Danish pedigrees (17) but not in 1,006 control subjects. We tested two of the Danish pedigrees and found a shared haplotype block of 3.7 Mb. Other haplotype-specific variants in HNF4A were not found. We believe that the observed maximum two-point LOD score (3.1) for the −192C>G variant at a recombination fraction of 0.10 (and not 0.00) in family N88 results from a slight misspecification of the model that underestimates the true phenocopy rate (subjects IV-1 and IV-19 are phenocopies). Based on the genetic data, we have established that the P2 variant or another variant within the shared 3.7-Mb haplotype is linked to diabetes in the two Norwegian and in two of the Danish pedigrees. We could not support a functional role of the P2 promoter variant by an in vitro transfection assay. This result does not, however, exclude that the P2 promoter variant is biologically relevant. Interestingly, Ek et al. (17) provide in vitro data showing that P2 −192C>G disrupts the binding of an unidentified sequence-specific DNA-binding complex present in human islet extracts.

Most of the subjects with the P2 −192C>G variant were lean and had a later onset of diabetes (median onset age of 45 years) compared with other recently described HNF4A mutation carriers (mean onset age of 23 years) (18), suggesting that it defines a particular subgroup of diabetic subjects. Late-onset monogenic diabetes has also been noted for the coding HNF4A variant V393I (20).

Several studies in various ethnic populations have suggested that haplotypes spanning the regulatory regions of HNF4A associate with type 2 diabetes (7–10,12), while other studies have failed to show this relationship (11), a problem generally noted in association studies (21). One reason for the discrepancy between the studies might be allelic heterogenetity (22). Rare risk variants, like the described P2 −192C>G variant, with modest attributable risk in the population but with major impact in a family, will most likely not be detected by conventional multipopulation association studies (23,24). In such cases, exhaustive sequencing in extended families showing linkage to the region might be an alternative approach.

In conclusion, variants in the coding and promoter regions of HNF4A leading to monogenic diabetes account for ∼9% of MODY3-negative probands meeting conventional criteria, making MODY1 the third most prevalent MODY form in Norway. An HNF4A P2 promoter haplotype was linked with late-onset diabetes, further supporting the role of the HNF4A region in type 2 diabetes.