Introduction

The gene SLC2A2 encodes GLUT2, a facilitative glucose transporter. The pancreas, liver, kidney and intestine all express GLUT2 [1]. Biallelic inactivation of Slca2 in mice leads to severe diabetes with failure to thrive, marked hyperglycaemia with relatively low insulin and high glucagon levels, and death usually at 2–3 weeks of age [2]. The mice can be rescued by exogenous insulin, indicating that the insulin-secretory defect is the major cause of lethality [2]. Glucose-stimulated insulin secretion is normalised when Slc2a2 expression is restored in Slc2a2 null murine beta cells [3].

The role of GLUT2 in man is controversial. Expression of GLUT2 in human beta cells is much lower than in mice [4–6]. Humans with biallelic inactivating SLC2A2 mutations do not normally present with diabetes, but do develop Fanconi–Bickel syndrome (FBS) [7, reviewed in 8]. FBS characteristically involves a renal Fanconi syndrome with glycosuria, galactosuria, aminoaciduria, proteinuria and phosphaturia, short stature, rickets, poor growth, hepatomegaly, and glucose and galactose intolerance. Diagnosis normally occurs in late infancy as clinical features of FBS develop. In some cases galactosaemia screening leads to earlier diagnosis [8].

Monogenic neonatal diabetes mellitus typically presents before 6 months, and may be permanent (PNDM) or transient (TNDM). PNDM is usually due to heterozygous mutations in KCNJ11, ABCC8 or INS. TNDM is most commonly due to an imprinting disorder at chromosome 6q24, and also to mutations in KCNJ11 or ABCC8 [9]. In around one-third of cases the cause has not been identified.

There is one report of a premature (34 weeks’ gestation) infant with neonatal diabetes and hypergalactosaemia who was homozygous for a truncating SLC2A2 mutation [10]. She received intermittent insulin, but it is not clear whether this persisted until her death from hepatic failure with pneumonia at 10 months.

We tested a large series of patients with neonatal diabetes to determine whether SLC2A2 mutations cause neonatal diabetes, which presents before clinical features of FBS are apparent.

Methods

This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the local ethics committee, and written informed consent was obtained from each participant or their parent.

Genetic analysis

We studied SLC2A2 in patients with TNDM (n = 25) or PNDM (n = 79) in whom we had excluded where appropriate the common genetic causes of neonatal diabetes [TNDM: KCNJ11, ABCC8, INS, PLAGL1 (6q24); PNDM: KCNJ11, ABCC8, INS, EIF2AK3]. We sequenced the gene for all patients, who either had TNDM or had PNDM and a possible recessive mutation suggested by a homozygous region >3 cM encompassing the SLC2A2 gene on genome-wide single-nucleotide polymorphism (SNP) analysis using the Affymetrix (Santa Clara, CA, USA) SNP chip 500 K or v6.0.

We amplified the 11 coding exons of the SLC2A2 gene by PCR (primers available on request). PCR products were sequenced using standard methods on an ABI 3730 (Applied Biosystems, Warrington, UK) and were compared with the published sequence NM_000340.1 using Mutation Surveyor v3.2 (SoftGenetics, State College, PA, USA). Where an SLC2A2 mutation was identified, siblings and parents were tested when samples were available.

Results

Molecular genetics

One PNDM patient had a homozygous SLC2A2 mutation c.339del (F114LfsX16) and has been reported previously [11]. Among the 25 TNDM patients, four (16%) had homozygous SLC2A2 mutations, three of which are novel: c.609T>A (S203R), c.1127T>G (M376R), c.963+1G>A (IVS7+1G>A); and one has been reported previously: c.157C>T (p.R53X) [8]. Further details of the mutations are presented in Table 1. All nine parents tested were carriers of the mutations. Patient 2 has a sister homozygous for the same mutation who was not diagnosed with diabetes.

Table 1 Details of homozygous SLC2A2 mutations identified
Full size table

Clinical features

Clinical details of the patients with SLC2A2 mutations are shown in Table 2. Parental consanguinity was reported in four out of five cases. All four patients with birthweight records had intrauterine growth reduction with birthweights at or below the third centile. Diabetes diagnosis was before 6 weeks. Insulin doses are similar to those in other cases of neonatal diabetes [9] and C-peptide values in two patients were low at diagnosis in the presence of marked hyperglycaemia, suggesting the diabetes resulted from relative insulin deficiency.

Table 2 Clinical features of patients with SLC2A2 mutations
Full size table

At the diagnosis of diabetes, only one patient had features suggesting FBS (rickets at 6 weeks). All patients later developed typical clinical features of FBS, which were recognised up to 30 months after diabetes was diagnosed. The older sister of patient 2, who was homozygous for the same mutation, had FBS features but was not diagnosed with neonatal diabetes.

Discussion

There are now six confirmed cases of neonatal diabetes due to SLC2A2 mutations: the five in this paper, plus that reported by Yoo et al [10]. This result is important as it implies a role for GLUT2 in glucose regulation in humans.

All five SLC2A2 mutations are highly likely to be pathogenic as they co-segregated and are either null mutations or missense mutations which affect a highly conserved residue within an important part of the protein (see Table 1). In addition all five patients have developed FBS.

The SLC2A2-positive cases reported here represent 16% of our undiagnosed TNDM cohort. Although a rare cause of neonatal diabetes, our results suggest that it is worth testing when other common causes have been excluded, particularly when the patient is born to related parents and has possible features of FBS, such as persistent glycosuria after resolution of their diabetes.

It is not known why neonatal diabetes is only diagnosed in a minority of patients with SLC2A2 mutations (4%; six in approximately 154 reported FBS cases). It is unlikely that these mutations are functionally different from those found in patients who develop FBS without neonatal diabetes, as the mutation found in patient 4 was previously known in FBS [8], and in one family an older sister with the same homozygous mutation had FBS but was not known to have diabetes. It is not known if the majority of FBS patients who are not diagnosed with neonatal diabetes have mild resolving diabetes that is undetected or, despite the fluctuating glucose levels expected in FBS patients, do not meet diabetes diagnostic criteria. Most of our patients came to medical attention due to intercurrent illness and the diabetes had remitted before the first features of FBS become apparent. It is an interesting possibility that the renal glycosuria seen in FBS may greatly reduce the blood glucose levels despite there being a defect in insulin secretion.

We consider that the neonatal diabetes found in patients with recessive inactivating SLC2A2 mutations is likely to reflect impaired insulin secretion given: (1) the response to insulin therapy in doses similar to other subtypes of neonatal diabetes; (2) the relatively low C-peptide levels at glucose values >20 mmol/l measured at diagnosis in two of our patients; (3) the inappropriately low insulin secretion associated with glucose intolerance reported in many FBS patients [8, 12–15]; and (4) the low birthweight seen in all patients, similar to other causes of neonatal diabetes, and consistent with reduced fetal insulin secretion in utero. We could not find data to assess if birthweight was reduced in FBS patients who are not diagnosed with neonatal diabetes. Further studies are needed on this, as a normal birthweight in FBS without neonatal diabetes would argue against fetal insulin secretion being reduced in all patients with SLC2A2 mutations.

The homozygous recessive inactivating mutations in humans result in less frequent and less severe neonatal diabetes than the biallelic inactivation of Slc2a2 in mice [2]. This is in keeping with GLUT2 being produced in lower amounts and having a less important role in glucose-stimulated insulin secretion in humans than in rodents [4].

While our finding of neonatal diabetes in a subgroup of patients suggests GLUT2 has a role in human insulin secretion, we cannot define the mechanism from our work. Expression studies suggest GLUT1 is the major human beta cell glucose transporter from fetal through to adult life [4–6]. Ninety per cent of beta cells were positive for GLUT1 from 20 weeks’ gestation to adulthood, but only 10–20% were positive for GLUT2 until infants were over 8 months of age [5], and there are lower levels of expression of GLUT2 compared with GLUT1 and GLUT3 in adults [6]. Patients who are haploinsufficient for GLUT1 due to heterozygous SLC2A1 mutations have epilepsy from impaired glucose transport across the blood–brain barrier [16, 17] and do not have diabetes. This may be because in beta cells 50% of GLUT1 is sufficient, or it may be due to compensation by other glucose transporters. Complete deficiency of GLUT1 due to a homozygous mutation would be expected to be fatal in utero. GLUT1 mutations have not been found to cause monogenic diabetes [18]. Glucose phosphorylation, rather than glucose transport, is normally likely to be the rate-limiting step for insulin secretion [19, 20].

It has been proposed that GLUT2 may have a role as a signalling molecule as well as a simple transporter [21]. GLUT2 confers insulin secretion but GLUT1 does not when expression levels are adjusted to ensure a similar glucose flux [22]. The higher K m of GLUT2, compared with GLUT1 or GLUT3 [23], could allow glucose sensing by GLUT2 at physiological concentrations. In the absence of functional GLUT2, failure of an as yet unclear signalling cascade may cause the impaired insulin secretion. A common SNP close to SLC2A2 has been shown to influence fasting glucose [24], but further investigations could not detect a defect in beta cell function using data from fasting and stimulated samples [25]. However, this is a different scenario from having no functional GLUT2, as seen in our patients.

In conclusion, we have shown that SLC2A2 mutations can cause neonatal diabetes that occurs before overt clinical features of FBS appear. FBS should be considered as a cause of neonatal diabetes once other common origins are excluded, particularly when parents are related or when the diabetes is transient. This finding implies a role for GLUT2 in insulin secretion in humans as well as in rodents.