The Type and the Position of HNF1A Mutation Modulate Age at Diagnosis of Diabetes in Patients with Maturity-Onset Diabetes of the Young (MODY)-3

This study includes 356 unrelated patients (87% Euro-Caucasians, 60% women) who had been referred for genetic testing from 1998 to 2007 and in whom a HNF1A mutation was identified. All patients gave written informed consent.

Several criteria were used to ascertain that the novel mutations identified in the present study were pathogenic: nature of the amino acid change, conservation of the residue across species, absence of the mutation in 300 control subjects of Euro-Caucasian origin, and cosegregation of the mutation with young-onset diabetes, when relatives were available.

The molecular spectrum of HNF1A mutations was analyzed according to three criteria. First, mutations were classified into two groups according to their predicted functional consequences. One group included missense mutations resulting in amino acid changes, and the second included nonsense, small insertions/deletions, or splicing mutations, predicted to generate premature stop codons (referred to as “truncating mutations”). Second, mutations were analyzed according to the three HNF1A functional domains: NH₂-terminal dimerization domain (amino acids 1–32), DNA-binding domain (amino acids 91–281), and COOH-terminal transactivation domain (282–631) (Fig. 1A) (17). Third, mutations were analyzed according to the affected isoforms of HNF1A. The HNF1A(A) isoform is the full-length transcript comprising the 10 exons, whereas HNF1A(B) and HNF1A(C) isoforms result from alternative splicing and contain the first seven and first six exons, respectively (Fig. 1B). Three groups of mutations were considered: mutations located in exons 1–6, affecting the three isoforms; mutations located in exon 7, affecting isoforms HNF1A(A) and (B); mutations located in exons 8–10, involving only the HNF1A(A) isoform.

Age at diagnosis is reported as median and range. Data were compared with the Mann-Whitney and the Kruskall-Wallis tests. Statistical analyses were performed with GraphPad InStat (GraphPad Software, San Diego, CA).

Among the 356 unrelated patients, 169 HNF1A mutations were identified. Fifty-one have previously been reported, whereas 118 are novel. The 51 known mutations were detected in 200 patients (supplementary Table [available at http://dx.doi.org/10.2337/db07-0859]). By contrast, the 118 novel mutations were mainly private mutations (97 of 118, 82%) (Table 1).

There were 201 (56.5%) missense and 155 (43.5%) truncating mutations (103 small insertion/deletion mutations, 31 nonsense mutations, and 21 splicing defects).

The same numbers of mutations affected the dimerization and DNA-binding domains (the two structurally major domains) and the transactivation domain (179 and 177 cases, respectively). However, missense mutations were much more frequent than truncating ones in the dimerization and DNA-binding domains (74% were missense mutations), while the opposite was noted in the transactivation domain (62% were truncating mutations) (Fig. 1A). The distribution was the same when considering only distinct mutations (not shown).

A large majority (83%) of the mutations were located in exons 1–6, thus affecting the three HNF1A isoforms; 13% of the mutations, located in exons 8–10, specifically affected the HNF1A(A) isoform; and 4% were located in exon 7, affecting isoforms HNF1A(A) and (B). This distribution within the isoforms was very similar when considering either missense or truncating mutations and when considering all or distinct mutations (Fig. 1B).

Age at diagnosis of diabetes was available for 352 patients. Median age at diagnosis was lower by 4 years in patients with truncating mutations than in those with missense mutations (18 vs. 22 years respectively, P = 0.005).

There was no difference in the age at diagnosis according to the location of the mutation within the dimerization/DNA-binding or transactivation domains (19 and 21.5 years, respectively) (Table 2). However, when both the type of the mutation and its position within the functional domains were considered, marked differences appeared. First, truncating mutations were associated with a lower age at diagnosis than missense ones when they affected the transactivation domain (19 vs. 30 years, P < 10⁻⁴). By contrast, age at diagnosis was similar for truncating and missense mutations of the dimerization/DNA binding domain (18 and 20 years, respectively). Second, missense mutations affecting the dimerization/DNA-binding domains were associated with a lower age at diagnosis than missense mutations affecting the transactivation domain (20 vs. 30 years, P = 10⁻⁴).

We then analyzed the age at diagnosis according to the isoforms affected by the mutation. Patients carrying a mutation affecting the three HNF1A isoforms had a younger age at onset of diabetes (19 years) than those with a mutation affecting isoforms A and B (29 years) or a mutation affecting only the HNF1A(A) isoform (24 years, P = 0.03 by ANOVA). No difference in the age at diagnosis was observed in patients with truncating mutations, regardless of the affected isoforms. By contrast, patients with missense mutations affecting the three isoforms were much younger at diagnosis (20 years) than those with missense mutations altering one or two isoforms (31 and 33 years, respectively, P = 0.006 by ANOVA).

Because the functional domains and the isoform structure are overlapping within the first six exons (Fig. 1), we compared age at diagnosis associated with missense mutations located in the dimerization/DNA-binding domains (amino acids 1–281) with that associated with missense mutations located in the part of the transactivation domain common to the three isoforms (amino acids 282–437). Age at diagnosis was lower in the former than in the latter (20 vs. 26.5 years, respectively, P = 0.015).

This large series of HNF1A mutations in 356 unrelated MODY3 patients emphasizes the high allelic heterogeneity of HNF1A. Among the 169 distinct mutations, 118 were not reported in a recent update (14). The large majority (82%) of the novel mutations were private. The type of mutations differed markedly within functional domains: in the dimerization/DNA-binding domain, 74% of the mutations were missense, whereas in the transactivation domain, truncating mutations were predominant (62%). A similar distribution of HNF1A mutations has previously been reported (14). Some missense mutations may have mild functional consequences on the protein, and their clinical expression may depend on the functional importance of the affected domain. Thus, some missense mutations of the transactivation domain may not be associated with overt diabetes or lead to a milder phenotype suggesting type 2 diabetes. In patients with truncating mutations, the mean age at diagnosis of diabetes was 18. This is similar to that previously reported in a large series of MODY3 patients (16) and suggests that truncating mutations have similar functional consequences. Nonsense-mediated decay may be the common mechanism leading to this homogenous phenotype through haplo-insufficiency (18). In patients with missense mutations, diabetes was diagnosed later (by 4 years on average) than in those with truncating mutations. This is in contrast with previous results that did not show relationship between the type of the mutations and age at onset (16). However, we only studied probands, while in the study by Harries et al., 55% of the MODY3 patients were relatives (16). We suggest that analyzing relatives together with the probands may introduce a bias toward the inclusion of young subjects through family screening. This would decrease the median age at diagnosis. Moreover, our diagnosis criteria are less restrictive than those often used to raise the diagnosis of MODY3, since we included probands with an age of onset of diabetes above 25 years.

Further analysis combining the type of the mutation and its location within the functional domains revealed striking differences in the age at diagnosis of diabetes. Diabetes was revealed 10 years earlier in patients carrying missense mutations located in the dimerization/DNA-binding domains than in those with a missense mutation in the transactivation domain. We hypothesize that missense mutations affecting the dimerization/DNA-binding domain have more severe functional consequences, such as impaired DNA-binding and protein stability (19).

Recently, it has been shown that the age at onset of diabetes may be influenced by the position of the mutation relative to HNF1A isoforms. Missense mutations located in exons eight to 10, that are specific of the HNF1A(A) isoform, were associated with an older age of onset (16). The authors suggested that this was due to differences in the expression level of the various isoforms in fetal and adult pancreas. We found that patients harboring missense mutations located in exon 7 or in exons 8–10 were diagnosed more than 10 years later than those with mutations in exons 1–6. However, since exons 1–6 include the dimerization and DNA-binding domains, the observed effect on age at diagnosis could be due either to involvement of the three isoforms or to the position of the mutation within the dimerization/DNA-binding domains (Fig. 1). To distinguish between these two possibilities, we compared mutations affecting the dimerization/DNA-binding domain to that affecting the first part of the transactivation domain, and we observed a younger age at onset in the former than in the latter. Thus, the location of the mutation within a domain crucial for the function of the protein overcomes the fact that the mutation affects the three isoforms of HNF1A. This was confirmed by a multivariate analysis (not shown).

The wide variability of MODY3 phenotype has suggested the role of modifier genes. However, such genes have not been identified yet (12). We have shown that truncating mutations, as compared with missense mutations, have an effect on the clinical expression of the disease. Moreover, in patients with missense mutations, which represent more than half of the cases, the position of the mutation relative to the functional domains of HNF1A also plays a role in the severity of the disease. We suggest that these parameters should be considered in the studies aiming at the identification of other factors that may influence the clinical expression of MODY3.

We thank all participants of the French Study Group of MODY for referring the patients and communication of clinical data. We thank Sandrine Beaufils, Florence Bellanger, Séverine Clauin, Sandrine Gobrecht, Gwendoline Leroy, and Christelle Vaury of the Molecular Genetics Laboratory for MODY genetic screening.