Background

Hepatocyte nuclear factor 1A (HNF1A) is a transcription factor that has a role in the development and function of pancreas ß-islet cells. In the developmental stage, both endocrine and exocrine cells of the pancreas have HNF1A expression [1]. HNF1A is necessary for insulin secretion in response to glucose [2,3,4]. The HNF1A gene has been identified in both monogenic and polygenetic diabetes. Rare mutations of the HNF1A gene cause a monogenic form of diabetes as Type 3 maturity-onset-diabetes of the young (MODY3) [1]. The HNF1A gene contributes to the pathogenesis of Type 2 diabetes mellitus (T2DM). HNF1A gene single nucleotide polymorphisms (SNPs) were modestly associated with Type 2 diabetes mellitus (T2DM) and glycemic features in different populations [5,6,7]. HNF1A SNPs were associated with impaired insulin secretion [8, 9]. HNF1A gene SNPs (p.I27L, p.A98V and p.S487 N) were inconsistently associated with impaired glucose tolerance and having diabetes [8,9,10,11,12,13,14]. Some young people with diabetes have atypical features such as insulin resistance or a need for insulin treatment. However, these features are not similar to T2DM. These non-obese adults have early-onset, MODY-like diabetes. Monogenic MODY has not been confirmed in patients with early-onset, MODY-like diabetes through genetic analysis [15]. The genetic basis of early-onset, non-monogenic diabetes is not yet known. The aim of this study was to obtain the effects of HNF1A gene SNPs on developing MODY-like diabetes. This is the first study to investigate the association between HNF1A gene SNPs rs1169288 (encoding HNF1A p.Ile27Leu), rs1800574 (encoding HNF1A p.Ala98Val) and rs2464196 (encoding HNF1A p.Ser486Asn), and having early-onset, MODY-like diabetes in the Turkish population.

Methods

Patients

In our study, none of the control subjects (n = 263) had diabetes. All patients with diabetes (n = 486) met the criteria for the diagnosis of MODY. Subjects with a clinical suspicion of MODY [16] (diagnosis of diabetes age below 25 years, positive family history including autosomal dominant inheritance in at least 2-3 generations, residual insulin secretion with normal C-peptide concentration and absence of B-cell autoimmunity) who presented to our hospital between 2012 and 2015 were included in the study. The inclusion criteria were as follows; patients with T2DM with C-peptide concentrations ≥0.3 nmol/L, negative anti-GAD antibodies, and age-at-onset below 25 years [2]. Patients with suspected MODY did not need insulin treatment for at least first 2 years after diagnosis and had no family history of T1DM [17]. Early or late-onset diabetes was identified by using age 45 years as a cut-off, as described in previous studies [2, 17]. If we selected control subjects from those whose mean age was below 28 years, some of these subjects would develop diabetes later in life. As a way of reducing the possibility of recruiting control subjects who might later develop T2DM, healthy-normoglycemic subjects with fasting glucose below 100 mg/dL and glycated hemoglobin (Hb1Ac) < 5.7%, who were aged ≥45 years and had no first-degree relatives or grandparents with T2DM were included in the control group [17]. Healthy controls without chronic disease such as diabetes, hypertension, renal and hepatic disease, were recruited from the outpatient clinic. Subjects with genetically confirmed MODY or T1DM were excluded [2]. Genetic analysis was performed for all patients (n = 565) in order to diagnose MODY. After genetic analysis, patients with diabetes (n = 46) who had HNF1A, HNF1B, HNF4A, GCK gene mutations were diagnosed as having MODY3, MODY5, MODY1, and MODY2 respectively. Thirty patients with diabetes had HNF1B, HNF4A, had GCK gene SNPs. Patients with diabetes with MODY and HNF1B, HNF4A, and GCK SNPs were excluded from the study. Finally, subjects without HNF1A, HNF1B, HNF4A, and GCK gene mutations and HNF1B, HNF4A, and GCK SNPs (n = 486) and non-diabetic healthy controls (n = 263) were included this study.

Measurements

Fasting glucose, postprandial glucose, creatinine, HbA1c, triglycerides (TG), cholesterol, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), C-peptide, high-sensitivity (hs-CRP), anti-glutamic acid decarboxylase (GAD) antibody, anti-insulin antibody, anti-islet antibody, and urinary microalbuminuria concentrations were measured. Age and symptoms at onset of diabetes, diabetes treatment, and parental history of diabetes (first-degree relatives, mother or father) were recorded from all patients with diabetes. Body mass index (BMI) was calculated as weight (kg) / height (m2). BMI ≥ 30 kg/m2 was diagnosed as obesity. T2DM was diagnosed when plasma fasting glucose concentrations were > 125 mg/dL, casual or postprandial glucose levels were > 200 mg/dL or in the presence of current treatment with a hypoglycemic agent, according to the American Diabetes Association criteria [17]. Informed consent was obtained from all participants. This study was approved by Diskapi Yildirim Beyazit Training and Research Hospital Local Ethics Committee.

Genotyping and Statistical analysis is presented in Additional file 1.

Results

The percentage of women (51.5% vs. 58.2%) and BMI value (27.88 ± 5.72 vs. 27.01 ± 3.29 kg/m2) was similar between the diabetics and controls (p > 0.05). The mean age of the controls was 49.18 ± 3.38 years. The mean age at onset of diabetes was 24.08 ± 4.82 years. The mean C-peptide concentration of the patients with diabetes was 2.47 ± 1.79 nmol/L. Fasting glucose, HbA1c, TG, cholesterol, LDL-C concentrations were higher among the diabetics compared with the controls (p < 0.05, Table 1). HNF1A gene p.I27L rs1169288 and p.A98V rs1800574 SNPs were consistent with the Hardy-Weinberg equilibrium (HWE), and p.S487 rs2464196 were not consistent with the HWE (Table 2). HNF1A genotypes are shown in Table 3. Thefrequency of p.S487 N SNPs was similar between the diabetics and controls in the codominant model and dominant model and recessive model (p > 0.05, each). p.A98V SNPs were similar between the diabetics and controls in the dominant model and recessive model (p > 0.05, each). The p.A98V TT genotype was higher in diabetics in the codominant model compared with the controls (TT vs. CC, OR = 1.35, 95% CI: [0.95–3.54]; p = 0.027). HNF1A gene p.I27L TT genotype was increased in diabetes (TT vs. GG, OR = 1.71, 95% CI: [1. 25-3.46]; p = 0.024) compared with the controls in the codominant model. p.I27L GT/TT carriers had increased 1.68 odds of having diabetes (GT/TT vs. GG, OR = 1.68, 95% CI: [1. 21-2.13]; p = 0.035) in the dominant model. p.I27L TT carriers had 1.56-fold increased odds of having diabetes (GT/GG vs. TT, OR = 1.56, 95% CI: [1. 14-2.57]; p = 0.048) in the recessive model. Clinical and biochemical characteristics did not differ between patients with diabetes with p.I27L, p.S487 N, and p.A98V SNPs and diabetics without the SNPs (p > 0.05). Onset of diabetes was 26.17 ± 7.4 years in p.S487 N, 25.58 ± 2.7 years in p.A98V, and 24.57 ± 5.2 years in p.I27L (p > 0.05). Parent diabetes (mother or father) was higher in the p.I27L TT genotype compared with the GG genotype (78.5 vs. 98.7%, p = 0.035). Diabetics with p.I27L TT genotype had higher triglyceride concentrations compared with diabetics with the GG genotype (p = 0.041) (Table 4). HNF1A gene p.I27L and p.A98V haplotypes were within Linkage Disequilibrium. p.I27L SNPs was modestly associated with having diabetes after adjusting for BMI and age (β = 1.45, 95% CI: [1. 2-4.2]; p = 0.036).

Table 1 Characteristics of subjects
Full size table
Table 2 Minor allele frequency of HNF1A gene SNPs
Full size table
Table 3 Genotype analysis of HNF1A gene SNPs
Full size table
Table 4 HNF1A gene p.I27L SNPs and clinical features in diabetics patients
Full size table

Discussion

This case-control study showed that the HNF1A gene p.I27L SNP was modestly associated with having early-onset, MODY-like diabetes in the Turkish population. Family inheritance of diabetes was significantly more common in patients with the p.I27L TT genotype. The HNF1A gene p.I27L SNP might contribute to age at diabetes diagnosis and family inheritance.

In this study, we suggest that polygenic T2DM may show differences in age-related and family inheritance transmission for an associated monogenic form of diabetes. This is the first study to show the effect of the p.I27L genotype on modifying age at diagnosis in the Turkish population. We excluded monogenic diabetes modifier genes, which often include mutations, because we aimed to examine the influence of variations on polygenic diabetes. In our study, subjects with diabetes were non-obese and the onset of diabetes was early. A previous study showed that non-obese patients with early-onset diabetes were more susceptible to β-cell dysfunction as compared with old and obese individuals [18]. A modest association was found between HNF1A missense SNPs (p.I27L, p.A98V, and p.S487 N) and having late-onset T2DM in the European population [2,3,4]. In European ancestry, no association was shown between HNF1A gene SNPs and having late-onset T2DM [19], but a robust association was found when p.A98V SNPs were included [20]. Similar to our study, European ancestry reported that p.I27L and p.A98V SNPs were associated with having late-onset T2DM [12]. p.I27L GT/TT carriers had 1.68-fold increased odds of having diabetes, and p.I27L TT carriers had 1.5 6-fold increased odds of having diabetes in our study. Only the p.I27L variant was modestly associated with having diabetes and this relationship continued after adjusting BMI and age. There was no association between p.S487 N and p.A98V SNPs and early-onset T2DM. Similar to our report, a modest association was shown between p.I27L, p.S487 N, and p.A98V and having T2DM in the European population [19]. In agreement with our report, p.I27L was associated with having T2DM in non-obese French [21] and Finnish subjects [13]. A Chinese and Japanese meta-analysis reported that p.I27L was associated with having T2DM [18]. HNF1A gene p.I27L was associated with having late-onset T2DM in Brazilian [22] and Western Indian [23] overweight/obese subjects aged 51–60 years; however, this association was found in normal-weight Japanese subjects [11]. In line with our report, Holmkvist et al. determined that p.I27L was associated with having late-onset T2DM in overweight Scandinavian subjects aged over 60 years, and p.A98V was reported to decrease in vivo glucose-responsive to insulin secretion [2]. Chi et al. demonstrated that p.I27L has a modest role in β-cell dysfunction [10] and in insulin resistance [8, 24]. Consistent with our study, European population studies found a modest association between only p.A98V and having T2DM [3, 4]. A Danish study of Caucasians found p.A98V to be associated with decreased insulin secretion in healthy individuals [9], but this effect was balanced by increased insulin sensitivity [25]. HNF1A gene p.A98V was associated with having early-onset T2DM in Scandinavian [26] and Asian-Indian [27] individuals. HNF1A p.A98V was associated with having late-onset T2DM in Finnish but not in Chinese individuals [14]. Our study reported that the p.A98V TT genotype was higher compared with the GG genotype in diabetics, nevertheless, with no association.

Early-onset diabetes (19 years) was observed in Chinese p.I27L + p.S487 N carriers [28]. Yorifuji et al. reported that patients who were MODY-mutation–positive were younger and had a lower BMI percentile at diagnosis compared with mutation-negative patients in Japan [29]. A German-Austrian study reported that age at onset of diabetes (10.9 years) was found to be younger in p.I27L + p.S487 N ± p.A98V carriers, as compared with HNF1A mutation (14 years). Locke et al. reported that each p.I27L allele was associated with a 1.6-year decrease in age at diagnosis in patients with HNF1A-MODY [30]. Our study reported early onset of diabetes (24.08 ± 4.82 year) with no differences between HNF1A gene SNPs. Similar to our report, paternal diabetes was higher in HNF1A gene SNP carriers [31]. This study found that diabetes was higher in first-degree relatives (mother or father) of p.I27L homozygous TT carriers, suggesting a probability of significant familial transmission.

The HNF1A locus p.I27L is localized in the dimerization domain, p.S487 N is localized in the transactivation domain, and the p.A98V is localized in the DNA-binding domain [1, 22, 28]. HNF1A gene p.I27L, p.A98V, and p.S487 N variants reduce transcriptional activities of genes that have a role in glucose metabolism [2]. It was reported that p.I27L + p.A98V variations decreased transactivation activity on GLUT2 in HeLa cells more than p.I27L alone and p.A98V alone [2]. Decreased insulin secretion and ß-cell dysfunction was observed in p.I27L coexisting with p.487 N carrier (when p.A98V carrier included). This leads to developing diabetes [1, 2, 4, 24, 25, 31]. HNF1A controls ß-cell function by regulating target genes such as glucose transporter 2 (GLUT2), HNF 4A, collectrin, liver pyruvate kinase, and hepatocyte growth factor activator. HNF1A activity dysfunction causes a reduction β-cell mass and induces onset of diabetes [1]. Gene expression regulation among diabetic subjects with HNF1A variation can be explained by environmental factors together with epigenetic factors [22, 31].

This study had a case-control design and small sample size. p.I27L and p.A98V were consistent with the HWE whereas p.S487 was not consistent with HWE. HNF1A gene p.I27L and p.A98V haplotypes were within LD. These are the limitations of this study.

Conclusions

We report a genetic modifier of the HNF1A gene age at diagnosis that shows an effect of genetic variation on diabetes phenotype. The HNF1A variant p.I27L was associated with having early-onset, MODY-like diabetes in the Turkish population. Enlightening the role of HNF1A in β-cells would be helpful in understanding the molecular mechanism of both T2DM and MODY and would guide new therapeutic approaches.