Maturity onset diabetes of the young is an autosomal dominant form of diabetes which accounts for 1–2% of all diabetes [1]. There are currently 13 different MODY subtypes described. With the advent of next-generation sequencing (NGS), it is likely that an increased number of individuals will be re-classified or newly diagnosed with a monogenic form of diabetes [2–4]. As a result, clinicians will need to be aware of these conditions, their natural progression and optimal clinical management. To date, the most common form of MODY in the UK and Ireland is attributed to a mutation in the transcription factor: hepatocyte nuclear factor 1 alpha (HNF1A-MODY) followed by glucokinase (GCK-MODY) [1, 5, 6]. In comparison, HNF4A-MODY is a relatively rare disorder accounting for approximately 6–10 % of all MODY cases in the UK and Ireland [7]. A genetic diagnosis of monogenic diabetes is important as it enables the discontinuation of insulin in a significant proportion of individuals. Patients with HNF1A/HNF4A-MODY are sensitive to sulphonylurea therapy. In contrast, GCK-MODY does not typically require treatment outside of pregnancy [8–14].

HNF4A protein is an orphan member of the nuclear receptor family of ligand-activated transcription factors. Impaired beta cell function is a known feature of both HNF4A-MODY mutation carriers with and without clinically diagnosed diabetes [15, 16]. An annual deterioration of 1–4 % in glucose stimulated insulin secretion in the HNF4A/RW pedigree is reported in the literature [17]. It has also been demonstrated that deletion of HNF4A in the beta cells of animal models results not only in impaired glucose stimulated insulin secretion but interestingly also in fasting hyperinsulinism [18, 19]. Of late, heterozygous mutations in the HNF4A gene have been shown to be a cause of congenital hyperinsulinism [19, 20]. It has been postulated, therefore, that a “biphasic phenotype” exists in a subset of HNF4A-MODY carriers with a period of hyperinsulinaemic hypoglycaemia as a neonate and subsequent diminished insulin secretion and diabetes in later adulthood. In those affected with congenital hyperinsulinism, macrosomia is a consistent finding. The hyperinsulinism attributed to HNF4A-MODY can be transient or persistent. It has a tendency to be responsive to diazoxide treatment with withdrawal of same ranging from hours to 12 years [20–22].

The HNF4A-MODY cohort in the Mater Hospital are diagnosed and clinically managed at a dedicated MODY centre enabling the longitudinal study of the pathophysiology of the mutation. The penetrance of HNF4A-MODY is 95 % by the age of 40 years; therefore, it is recommended that genetically confirmed normoglycaemic HNF4A-MODY mutation carriers have an annual oral glucose tolerance test (OGTT) to detect the development of diabetes [19]. In performing OGTTs on the HNF4A cohort, we have previously noted fasting hypoglycaemia in a proportion of adult mutation carriers. This would suggest that the described so-called biphasic phenotype associated with HNF4A-MODY persists beyond the neonatal period. To this end, in the current study we aimed to establish changes in glycaemia and insulin secretion in response to oral glucose over a 6-year period. We have also sought to elucidate this phenomenon further through the use of a continuous glucose monitoring system (CGMS) to detect glycaemic variability and, in particular, episodes of hypoglycaemia.

Research design and methods

Phenotyping of the HNF4A-MODY cohort

The current study included 14 individuals with HNF4A-DM and 7 HNF4A-MODY individuals with IGT. Further details of the mutations of this cohort are listed in Table 1. A group of 10, age- and BMI-matched normoglycaemic mutation negative family members formed a control group (hereafter referenced to as normal controls) to the IGT group for comparative purposes. The clinical characteristics of all three groups are listed in Table 2. As part of the MODY screening programme, a full medical history and physical examination are performed. Anthropometric measurements including BMI were recorded. A 75 g OGTT is performed on participants with HNF4A-MODY after an overnight fast with measurement of glucose, insulin, C-peptide and glucagon at baseline and 30-min interval to determine degree of glucose intolerance and insulin secretory response. If applicable, oral hypoglycaemic agents are stopped at least 48 h prior to testing whilst long-acting insulin therapy is stopped for 24 h and short-acting insulin for 12 h before the OGTT. The oral glucose insulin sensitivity (OGIS) is calculated as previously described [23]. The glucose: insulin (G:I) index was calculated as previously described [24]. Blood samples are drawn for the measurement of HbA1c, fasting lipids, full blood count, thyroid function, renal and liver profiles, glutamic acid decarboxylase (GAD65) autoantibodies and pancreatic islet cell autoantibodies (ICA). Routine clinical follow-up involves annual OGTTs in the HNF4A-IGT group. Details of the particular class of medication used in the group are listed in Table 2. The majority of the HNF4A-DM group were treated with a sulphonylurea (gliclazide). The median dosage of sulphonylurea (gliclazide) required to maintain euglycaemia in the HNF4A-DM group was 180 mg (40–290) per day. Of the 5 HNF4A-DM individuals who participated in the CGMS arm, 4 were treated with a sulphonylurea and 1 was treated with metformin (intolerant of gliclazide secondary to hypoglycaemic episodes). The 5 HNF4A-IGT individuals were treated with diet alone. The study was approved by the Research Ethics Committee at the Mater Misericordiae University Hospital, Dublin, and all participants gave informed written consent.

Table 1 Mutation details of cohort
Full size table
Table 2 Clinical characteristics of the HNF4A-DM, HNF4A-IGT and Normal control groups
Full size table

Data on continuous glucose monitoring system (CGMS)

A subset of 10 individuals with HNF4A-MODY consented to CGMS profiling. Five HNF4A-DM and five HNF4A-IGT individuals participated in this study arm. A blinded iPro2 continuous glucose monitor (Medtronic Inc.) was inserted by a single researcher. Participants were provided with full instructions on CGM device care and calibration. Participants were instructed to wear the device for a minimum period of 72 h. They were trained on using One Touch Contour glucometer and advised to facilitate calibration of the CGM device by performing a capillary blood glucose check 1 and 3 h following device insertion. Thereafter, a minimum of 4 glucometer recordings ideally pre-meals and bedtime were advised. Participants were encouraged to maintain their current dietary and activity habits for the study duration. A simple log of dietary intake, activity levels, meal times and sleep/wake times were recorded. On completion of the minimum 72-h study period, the participant returned to the metabolic research unit in the Mater Misericordiae University Hospital where the device was removed by the same researcher. Using the iPro2 recorder package, sensor readings were converted into excel format for each subject and each individual trace was analysed.

CGM parameter calculations

The parameters of glycaemic variability utilised were mean blood glucose (MBG), standard deviation of blood glucose (SDBG) and mean amplitude of glycaemic excursions (MAGEs) The MBG was calculated for the arithmetic average of all glucose readings. The SDBG represented the SD calculated for the overall 72-h period of CGM readings. MAGE calculation was made by measuring the arithmetic mean of the difference between consecutive peaks and nadirs provided that the difference was greater than the SD around the mean glucose values. A hypoglycaemic episode was defined as a glucose level of <3.9 mmol/L. We also analysed the data for a glucose cut-off value of 3.5 mmol/L.

Biochemical analysis

All laboratory analyses were performed with commercially available standardised methods. The plasma glucose concentration was measured using Beckman Synchron DXC800 (Beckman Instruments Inc, Brea, USA).HbA1c was determined using high-performance liquid chromatography (Menarini HA81-10, Rome, Italy). Insulin and C-peptide were analysed using Immulite 2000 immunoassay (Siemens Healthcare Diagnostics, Deerfield, IL, USA). Anti-GAD65 antibodies were analysed by ELISA (Euroimmun, Luebeck, Germany). ICA was performed by indirect immunofluorescence test by the supra-Regional Protein Reference Unit and Dept. of Immunology in Sheffield, UK. Glucagon was measured using a radioimmunoassay specific for pancreatic glucagon (Eurodiagnostica, Sweden).

Genetic analysis

Analysis of the HNF4A gene was performed by direct sequencing of the P2 promoter, exon 1d and exons 2-10 (HNF4A sequence accession number NM 175914.3). Sequencing of the HNF4A gene was performed by IntegraGen (Bonn, Germany) in 2006–2007 and the Molecular Genetics Laboratory (Exeter, UK) in 2008–2014.

Statistical analysis

Statistical analyses were performed using Prism/MATLAB and its statistical toolbox (The MathWorks Inc., Natick, USA). Parametric data are given as mean and SEM and compared using Student’s t tests. Nonparametric data are given as median and interquartile range (IQR) and were compared by Mann–Whitney U test. Hypotheses tests were considered statistically significant if P < 0.05.

Results

The clinical characteristics of the HNF4A-DM, HNF4A-IGT and the normal controls are listed in Table 2. The HNF4A-DM group were older with a higher BMI (although still within the normal range) than both the HNF4A-IGT and the normal controls. The median birth weight of the HNF4A-DM cohort was 3840 g (3665–4728) compared to 4280 g (3855–5018) in the HNF4A-IGT group (P = 0.4). Only one participant with HNF4A-IGT had prolonged hypoglycaemia that occurred at birth and lasted 8 days in total.

Insulin and C-peptide responses to oral glucose in the HNF4A-DM, IGT and normal control groups

The glucose, insulin and C-peptide responses to oral glucose in each group studied are shown in Fig. 1. The fasting G:I ratio, reflecting an inappropriately high fasting insulin for a given glucose, was significantly lower in the HNF4A-IGT compared to the normal control group (0.13 vs. 0.24, P = 0.03). (The AUC values for glucose, insulin and C-peptide are contained in the supplementary material).There was, as expected, a significant difference in the AUC glucose between the HNF4A-IGT and normal control groups [29.5 mmol/L/120 min (27–42) vs. 24.5 mmol/L/120 min (21–25), p = 0.001]. Of interest, we noted a lower fasting blood glucose in the HNF4A-IGT group when compared to normal controls [4 mmol/L (3.7–5.4) vs. 4.6 mmol/L (4.3–4.8) P = 0.9]. In addition, there was no difference in the AUC insulin or AUC C-peptide between the HNF4A-IGT and normal controls. In fact, a significantly higher insulin level was noted at 90 min in the HNF4A-IGT group when compared to the normal controls (90 min: 278.5 pmol vs. 207 pmol, P < 0.05). Glucagon levels at baseline were similar in all groups studied.

Fig. 1
figure 1

a–c OGTT profiles demonstrating glucose (a), insulin (b) and C-peptide (c) secretory response to oral glucose in each group studied

Full size image

In accordance with previous publications, we note a mean incremental glucose response (0–120 min) of 10 mmol/L in the HNF4A-DM group [25]. The OGIS value, reflecting insulin sensitivity, was also lower in the HNF4A-DM group than in the HNF4A-IGT group [356 ml/min/m2 (305–427) vs. 503 ml/min/m2 (415–541), P = 0.08]. There was no difference in the OGIS value between the normal control group and the HNF4A-IGT group [499 ml/min/m2 (482–511) vs. 503 ml/min/m2 (415–541), P = 0.9].

Preservation of insulin and C-peptide response on serial oral glucose tolerance testing over a 6-year period

Preservation of insulin and C-peptide secretion is observed over the duration of serial testing in the HNF4A-IGT cohort. Serial oral glucose tolerance testing in the HNF4A-IGT cohort over a period of 6 years is shown in Fig. 2. There was no difference in the serial assessments of AUC for glucose [initial AUC glucose: 29.5 mmol/L/120 min (27–42) vs. AUC glucose @ 6-year follow-up: 34.8 mmol/L/120 min (28.7–38), P = 0.9] or for AUC insulin over the study period [888 pmol/L/120 min (612–1883) vs. 932 pmol/L/120 min (663–1117), P = 0.7]. Likewise, there was no change in AUC C-peptide over the study duration [6043 pmol/L/120 min (3806–10,919) vs. 6043 pmol/L/120 min (3806–10,919), P = 0.8]. Over the study period, there was no difference in HbA1c [initial HbA1c; 5.1 % (4.9–5.3) (32 mmol/mol (30–34)) vs. 6-year follow-up HbA1c; 5.1 % (4.9–5.4) (32 mmol/mol (31–34.5)), P = 0.9].

Fig. 2
figure 2

a–c Serial OGTTs in the HNF4A-IGT cohort over a period of 6 years. a Glucose profile, b Insulin profile and c C-peptide profile

Full size image

Minimal glycaemic variability and prolonged hypoglycaemia is observed amongst HNF4A-MODY mutation carriers using CGMS

Parameters to evaluate the glycaemic variability were MAGE and SD of the arithmetic average of interstitial glucose. The MAGE reference range in a normal non-diabetic population is given as 0.0–2.8 mmol/L [26]. We note minimal glycaemic variability as demonstrated by a low MAGE in both the HNF4A-DM at 3.2 ± 0.5 mmol/L and in the HNF4A-IGT group at 2.2 ± 0.2 mmol/L.

The mean blood glucose recorded using CGMS was 7.5 ± 0.2 mmol/L in the HNF4A-DM group and 4 ± 0.8 mmol/L in the HNF4A-IGT group (P = 0.0007).The averages of 11,729 interstitial glucose values retrieved from CGMS compared well with their corresponding finger stick capillary blood glucose with a positive correlation of r = 0.8 between these two values. Asymptomatic hypoglycaemia is defined by the American Diabetes Association as a plasma glucose of <3.9 mmol/L [27]. In the HNF4A-DM group, the mean duration of hypoglycaemia was 3.2 % ± 0.7 (138 min ± 30.2) over a 72-h period. In the diet-treated HNF4A- IGT group, the mean duration of hypoglycaemia was 10.6 % ± 2.9 (432 min ± 125.5) over a 72-h period. These episodes of hypoglycaemia occurred predominantly in the nocturnal and late post-prandial period (120–180 min post-prandial) and were asymptomatic. A representative CGMS profile of a HNF4A-IGT participant over a 72-h period is shown in Fig. 3. Prolonged episodes of nocturnal hypoglycaemia are clearly illustrated.

Fig. 3
figure 3

Seventy-two hour continuous glucose monitoring profile of a representative individual with HNF4A-IGT. Meal times are denoted by a downward arrow. Blue represents the time spent with blood glucose levels above target. Green represents those blood glucose levels that are on target. Pink represents those blood glucose levels that are below target

Full size image

If a cut-off value of <3.5 mmol/L is used to define hypoglycaemia, the HNF4A-DM group did not experience any hypoglycaemic episodes, whilst in the HNF4A-IGT the mean duration of hypoglycaemia was 2.8 % ± 1.9 (120 min ± 82) over a 72-h period.

Discussion

Previous literature describes a “biphasic phenotype” in individuals with HNF4A-MODY, with an initial period of hyperinsulinaemic hypoglycaemia progressing ultimately to relative insulin deficiency and diabetes in adulthood, with an average age of onset of diabetes at 25 years of age [19]. There is a paucity of the literature available on glycaemic variability in late childhood through adolescence and into early adulthood in non-diabetic HNF4A-MODY carriers. In the current study, using serial OGTT assessment, we note no deterioration in glucose control into adolescence and beyond in a HNF4A-IGT group over a 6-year period. Despite a median age of 21 years, we did not determine the gradual decline in insulin secretion as previously reported nor did we observe the so-called switch point to diabetes in this group [28]. Factors that may explain the lack of progression to overt diabetes may include a lower BMI in the HNF4A-IGT group or perhaps the effect of dietary intervention. However, it is worth noting that diet did not change the BMI of the HNF4A-IGT group over the study duration [BMI at MODY diagnosis; 20 kg/m2 (17–21) vs. current BMI; 21 kg/m2 (20–22.4), P = 0.1]. It is of significant interest that this cohort maintained relatively high fasting insulin for given low fasting blood glucose levels as demonstrated by both the G:I index and the evidence of persistent hyperinsulinaemia in response to oral glucose.

CGMS profiling demonstrated prolonged episodes of hypoglycaemia, in particular, in the diet-only treated HNF4A-IGT cohort. The mean duration of hypoglycaemia over the 72-h period was significantly higher in the HNF4A-IGT group when compared to the HNF4A-DM group (432 vs. 138 min P = 0.04). The studies available on CGMS in a non-diabetic population describe an average time spent in hypoglycaemia of between 30 ± 26 min over a 72-h period and 47 min over a 66-h period [26, 29]. Therefore, the HNF4A-IGT group experienced up to a 13-fold increase in the period of time spent in hypoglycaemia when compared to a non-diabetic population. Both the HNF4A-DM and HNF4A-IGT groups were encouraged to maintain a log of all meal excursions and physical activity. The episodes of asymptomatic hypoglycaemia in the HNF4A-IGT group were not precipitated by the omission of a meal or by a physical activity. The clinical consequence of these episodes of asymptomatic hypoglycaemia in the MODY cohort remains unknown. Measurable neurocognitive deficits attributed to dysglycaemia have been described in individuals with type 1 diabetes. Both acute hyperglycaemia and chronic hypoglycaemia can result in lower IQ scores and in structural white matter change in individuals with type 1 diabetes [30–32]. Although we did not perform formal neurocognitive testing or imaging on our cohort, all participants in the HNF4A-IGT cohort had completed schooling and the majority had received third level education. Neuronal insult can also be attributed to glycaemic variability. [33]. Of importance, it is worth noting that on CGMS monitoring there were minimal glycaemic excursions in the HNF4A-MODY cohort which may afford a potential protective role against the effect of hypoglycaemia. The MAGE values in both groups did not differ to that reported in a normal non-diabetic population [26]. This finding perhaps reflects the preservation of insulin secretion in this cohort or a good response to sulphonylurea in the DM group or to diet therapy in the IGT cohort.

Deletion of HNF4A in the beta cells of mice results in a mildly reduced blood glucose level associated with hyperinsulinaemia in both the fasting and fed state at 3–5 months of age [18]. A study by Pearson et al also demonstrated low blood glucose levels secondary to hyperinsulinism during the earlier neonatal period in a beta cell-specific HNF4A knock out mouse model [19]. In human observational studies, it is apparent that there are dual opposing roles of HNF4A during the lifetime of a mutation carrier [34]. The initial beta cell hypersecretory phase in the foetal and neonatal period results from islet cell hyperplasia and macrosomia. In the subsequent phase, beta cell exhaustion results in altered glucose homoeostasis and ultimately diabetes. The temporal pattern of HNF4A in humans is heterogeneous, with the initial phase of neonatal hypoglycaemia requiring diazoxide for up to 1 year. The longest reported case of neonatal hypoglycaemia requiring treatment in a HNF4A-MODY mutation carrier is 12 years [20]. Our study demonstrates a persistent low fasting blood glucose and HbA1c over time in a HNF4A-IGT cohort with preserved insulin secretion levels and a similar basal glucagon level to a non-diabetic population. Therefore, we propose that the hypoglycaemia demonstrated by CGMS in the HNF4A-IGT cohort may be due to a prolonged hyperinsulinaemic hypoglycaemic phase into adulthood.

The fasting G:I index was significantly lower in the HNF4A-IGT cohort compared to the normal control group. In a previous study, insulin secretion tended to be higher in a HNF4A-IGT group compared to controls at glucose levels <7 mmol/mol [35]. In the current study, the hypoglycaemia noted on both OGTT and CGMS appears to occur in the setting of inappropriately high insulin levels rather than reduced glucagon secretion. Previous studies have demonstrated no difference in fasting basal glucagon levels between the HNF4A-IGT, HNF4A-DM and controls [36, 37]. A diminished glucagon response to hypoglycaemia has been described in a HNF4A-DM group but not in a HNF4A-IGT group. It is possible that the defect in hypoglycaemia-induced glucagon secretion in the HNF4A-DM group is secondary to the mildly diabetic state and not to the presence of a HNF4A mutation per se [37]. Recently, a group has demonstrated the localisation of HNF4A to the alpha cell in human islet studies and has proposed a role for HNF4A in the expression of the SGLT2 (the sodium glucose co-transporter) [38]. These findings would support a role for an alteration in glucagon secretion in the development of asymptomatic hypoglycaemia in this cohort. However, in the current study, in accordance with two previous studies no differences in the basal glucagon profiles were established.

A further postulated mechanism regarding the switch from hypo- to hyperglycaemia in animal models involves a loss of interaction between HNF4A and other nuclear factors such as peroxisome proliferator-activated receptor α (PPARα) [39]. The expression of PPARα is reduced in HNF4A deficient beta cells. PPARα plays a role in the beta oxidation of fatty acids. It has been postulated, therefore, that insulin levels are increased in the HNF4A mutant mouse model as a result of the lipid accumulation in the cytoplasm [40]. Overall, our current findings support those of the animal studies with reduced fasting blood glucose and a prolongation of the hyperinsulinaemic phase beyond the neonatal period into early adulthood.

We recognise that there are a number of limitations to our study. Firstly, the cohort is small; however, given the relative rarity of HNF4A-MODY we believe the sample size to be sufficient. We also recognise that CGMS profiling can read falsely low on occasion; however, it is worth noting that the episodes of asymptomatic hypoglycaemia were not confined to the first 24 h but were recorded throughout the 72-h period.

In conclusion, young HNF4A-IGT adults demonstrate preserved glucose, insulin and C-peptide secretory responses to oral glucose over a 6-year period. In this cohort, utilising CGMS, prolonged periods of hypoglycaemia are evident despite a median age of 21 years. We have proposed mechanistic theories that a prolonged hyperinsulinaemic phase into adulthood is responsible for the notable hypoglycaemic episodes. This study challenges our understanding of the biphasic pathophysiology associated with HNF4A-MODY. The incorporation of such information into the clinical management of HNF4A-MODY enables increased detection and self-awareness of hypoglycaemia and ultimately, improved patient care.