We have recently shown that permanent neonatal diabetes can be caused by activating mutations in KCNJ11 that encode the Kir6.2 subunit of the β-cell ATP-sensitive K+ channel. Some of these patients were diagnosed after 3 months of age and presented with ketoacidosis and marked hyperglycemia, which could have been diagnosed as type 1 diabetes. We hypothesized that KCNJ11 mutations could present clinically as type 1 diabetes. We screened the KCNJ11 gene for mutations in 77 U.K. type 1 diabetic subjects diagnosed under the age of 2 years. One patient was found to be heterozygous for the missense mutation R201C. She had low birth weight, was diagnosed at 5 weeks, and did not have a high risk predisposing HLA genotype. A novel variant, R176C, was identified in one diabetic subject but did not cosegregate with diabetes within the family. In conclusion, we have shown that heterozygous activating mutations in the KCNJ11 gene are a rare cause of clinically defined type 1 diabetes diagnosed before 2 years. Although activating KCNJ11 mutations are rare in patients diagnosed with type 1 diabetes, the identification of a KCNJ11 mutation may have important treatment implications.
We have recently shown that heterozygous mutations in the Kir6.2 (KCNJ11) gene, on chromosome 11p15.1, accounts for approximately one-third of cases of permanent neonatal diabetes mellitus (PNDM) (1). The gene encodes a subunit of the β-cell ATP-sensitive K+ channel (KATP channel) known as the inwardly rectifying K+ channel (Kir6.2). The KATP channel plays a critical role in coupling metabolism to membrane electrical events, which bring about closure of the KATP channel and initiate insulin secretion. Activating mutations in KCNJ11 cause familial and sporadic PNDM by decreasing β-cell insulin secretion through reduced ATP sensitivity of the KATP channel (1). Subjects presented with marked hyperglycemia and ketoacidosis at a mean age of 7 weeks (range 0–26), with 15% presenting outside 3 months. We hypothesized that KCNJ11 activating mutations could present in young children with a clinical diagnosis of type 1 diabetes.
The KCNJ11 gene, encoding Kir6.2, is located close to the large IDDM2 linkage peak (logarithm of odds [LOD] = 4.28) at 11p15.5 seen in the combined analysis of the U.S. and the U.K. type 1 diabetes genome-wide scans (2). IDDM2 is primarily explained by the insulin gene variable number tandem repeat (INS VNTR) region, located upstream of the insulin gene. However, after allowing for the INS VNTR, the resulting LOD score for chromosome 11 in this region was still suggestive of linkage (LOD = 2.53), and it was suggested by Cox et al. (2) that there could be a second type 1 diabetes locus in this region. A subset of families linked to KCNJ11 might explain this. A recent large study did not find any association with type 1 diabetes of the single nucleotide polymorphism E23K in KCNJ11 (3).
Type 1 diabetes is defined by an absolute deficiency of insulin secretion (4). The most common form is immune-mediated diabetes resulting from autoimmune destruction of the pancreatic β-cells. Pancreatic auto-antibodies are present in 85–90% of type 1 diabetic patients at diagnosis (4). Another indicator of the autoimmune nature of type 1 diabetes is the presence of the high risk HLA haplotypes (DRB1*03-DQA1*0501-DQB1*0201 and DRB1*04-DQA1*0301-DQB1*0302) in the majority of Caucasian type 1 diabetic subjects, while the protective haplotype DRB1*15-DQA1*102/DQB1*0602 is rarely found (5). In contrast to type 1 diabetes, patients with KCNJ11 mutations do not have pancreatic auto-antibodies (1). HLA studies have not been performed in patients with KCNJ11 mutations, but children diagnosed with permanent diabetes before 6 months have the same frequency of high risk HLA haplotypes as that found in the normal population (6).
Recent studies have shown that some patients with a clinical diagnosis of type 1 diabetes have maturity-onset diabetes of the young due to HNF-1α mutations (7–10). We previously reported a patient with an HNF-1α mutation who was GAD antibody negative and homozygous for a protective HLA haplotype, not supporting the diagnosis of type 1 diabetes. The identification of a specific genetic etiology can have implications for treatment; in this case, an affected family member was taken off insulin and is now treated with a sulfonylurea (7).
To test whether KCNJ11 activating mutations could present outside the neonatal period and be mistaken for clinical type 1 diabetes, we analyzed samples from 77 individuals clinically diagnosed with type 1 diabetes before the age of 2 years from two U.K.-based cohorts. The KCNJ11 gene was sequenced in 19 subjects from the Barts & Oxford (BOX) study (11) and 58 subjects from the British Diabetic Association (BDA) 1972–1981 cohort (12).
We identified possible heterozygous mutations in two cases from the BDA 1972–1981 cohort and no mutations in subjects from the BOX study. In addition, we identified several previously recognized polymorphisms (E23K, A190A, L267L, L270V, I337V, K381K, and S385C). The potential mutations were the substitution of arginine by cystine at codon 201 (CGT>TGT, R201C) and the substitution of arginine by cystine at codon 176 (CGC>TGC, R176C). Neither of these variants was identified in 100 normal chromosomes, and both residues are conserved in rat, mouse, and bullfrog. R201 and R176 both lie within the COOH-terminal of the Kir6.2 (Fig. 1). The arginine residue encoded by codon 201 is also conserved across 10 members of the Kir channel family, supporting a critical role for this residue in channel function.
R201C is likely to be an etiological mutation, as it has previously been reported in a subject with PNDM (1). The amino acid position R201 has been reported as the most common mutated residue (R201H) in KCNJ11 PNDM subjects. The residue R201 includes a CpG dinucleotide, which is a mammalian “hot spot” for gene mutations. Functional studies have shown that mutating this residue to histidine reduces the ATP sensitivity by ∼40-fold in the homozygous state (1). The R201 residue is thought to influence ATP binding indirectly by acting as an allosteric modifier. The R201C mutation has not been studied, but other substitutions at this site have also shown a reduction of ATP sensitivity (13). DNA was not available from other members of this family, but the absence of diabetes in the parents is consistent with a spontaneous mutation in the proband, as previously reported with this mutation (1).
R176C has not previously been found to be mutated in PNDM. It occurs at a residue that is involved in PIP2 (phosphatidylinositol 4,5 bisphosphate) binding to the KATP channel, which causes the channel to open (14). This arginine residue has been extensively analyzed by functional studies (14–16). The R176C variant was shown to reduce PIP2 binding, which would tend to result in channel closure and increased insulin release (16). Family studies showed that the R176C variant did not cosegregate with diabetes; her diabetic father did not have the variant, but the proband’s two unaffected sisters (one of which is her nonidentical twin) and mother were heterozygous carriers of the variant. Hyperinsulinism is the clinical phenotype associated with mutations that increase channel closure, but there were no clinical features to suggest this in the R176C carriers. The discrepancy between the functional studies and the clinical phenotype may reflect that in the functional studies all of the channels were mutated, but in the family members, as the mutation was heterozygous, only 50% of Kir6.2 channel subunits are mutated. We therefore conclude that R176C, despite being rare, conserved across species and in a functionally important position is a noncausal variant.
The clinical characteristics of the subjects with R201C and R176C are shown in Table 1. The subject with the R201C KCNJ11 mutation has similar characteristics to previously described patients with activating KCNJ11 mutations (1): she was born with a low birth weight (2.67 kg) and diagnosed with diabetes at 5 weeks of age, and GAD and IA-2 antibody testing performed at age 26 years were negative. This was the only patient in the cohort diagnosed before 6 months of age. According to current criteria, this patient would be diagnosed with PNDM (17), but at the time of diagnosis in 1977, neonatal diabetes was defined as presentation during the 1st month of life (the neonatal period). In contrast, the subject with the R176C variant did not have low birth weight, was diagnosed at 17 months, and is likely to have type 1 diabetes despite having a rare KCNJ11 variant.
HLA typing of the patient with the R201C KCNJ11 mutation identified her genotype as DRB1*02-DQB1*0502/DRB1*03-DQB1*0201. This genotype has recently been shown in a U.K. cohort (BOX) to have a neutral risk of type 1 diabetes despite carrying one disease-associated haplotype (15). The patient with the R176 variant had the neutral genotype DRB1*04-DQB1*0301/DRB1*07-DQB1*02. Ninety percent (69 of 77) of the diabetic patients had high risk HLA genotypes for type 1 diabetes, with 48% having the highest risk HLA genotype DRB1*03-DQB1*0201/DRB1*04-DQB1*0302, as expected in a young type 1 diabetic cohort. Only 10% (8 of 77) of individuals were positive for protective or neutral haplotypes, and it was from this group that the KCNJ11 mutation was identified. HLA analysis may help to identify a subset of individuals who are more likely to have KCNJ11 mutations. However, in this and our previous study (1), age of diagnosis (before 6 months) is more specific. Further studies are needed to assess whether age of diagnosis and HLA testing are helpful in identifying patients likely to have KCNJ11 mutations.
In summary, we have identified a heterozygous activating KCNJ11 mutation in a subject classified as having type 1 diabetes diagnosed under the age of 2 years. We also report a novel rare variant R176C, which family studies show to be unlikely to be a pathogenic mutation illustrating the importance of confirming cosegregation of a variant with diabetes in the family. There is difficulty differentiating this genetic subtype from type 1 diabetes on clinical criteria at presentation because patients can present with marked hyperglycemia and ketoacidosis. This and other studies (1,6) suggest that age of diagnosis (<6 months) and possibly HLA genotype (nonpredisposing) may help select subjects for testing for Kir6.2 mutations. Additional studies are required to test this further. The identification of a single subject (1 of 77, 1.3%) diagnosed at 5 weeks does not support an important role of KCNJ11 mutations in subjects diagnosed with type 1 diabetes under the age of 2 years. However, if, as suggested by physiological studies (1), sulfonylurea therapy proves effective in patients with activating KCNJ11 mutations, then this low pick-up rate may still warrant genetic testing of subjects with a clinical diagnosis of type 1 diabetes, particularly when diagnosed before 6 months of age.
RESEARCH DESIGN AND METHODS
All 77 subjects (42 male) had been diagnosed with diabetes before the age of 2 years and were considered by their clinicians to have type 1 diabetes on the basis of clinical features. A total of 58 subjects were from the BDA 1972–1981 cohort, a collection of individuals diagnosed under the age of 2 years established from the BDA register (12). The clinical data were collected from questionnaires sent out to the affected individual, the general practitioner, and the consultant. These data included HbA1c, creatinine, cholesterol, HDL, and LDL. Blood sampling was carried out for GAD and IA-2 antibody testing and DNA collection. An additional cohort of 19 subjects from the BOX study were selected because they were diagnosed before the age of 2 years (11,18). The mean age at diagnosis was 16.2 months; 1 (1.3%) child was diagnosed before 6 months, 12 (15.6%) between 6 and 12 months, and 64 (83%) over 12 and under 24 months. Twenty-one children (27.2%) had a first-degree relative with diabetes.
The entire coding region and intron-exon boundaries of the KCNJ11 gene were amplified by PCR using genomic DNA and sequence specific primers as previously described (1). Both strands were sequenced in forward and reverse using the Big Dye Terminator Cycler Sequencing Kit (Applied Biosystems, Warrington, U.K.) according to the manufacturer’s instructions. Reactions were analyzed on an ABI 3100 Capillary DNA sequencer (Applied Biosystems). Sequences were compared with the published sequence (NM_000525) using Sequence Navigator Software (Applied Biosystems). Changes in the sequence were checked against published polymorphisms and mutations.
Family relationships were confirmed using a panel of six microsatellites (19).
HLA class II genotyping.
Pancreatic auto-antibody analysis.
Antibodies to GAD65 and IA-2ic were measured by radioimmunoassay as previously described (11).
E.L.E. and A.L.G. contributed equally to this work.
The Wellcome Trust and Diabetes U.K. funded this study in Exeter and Bristol. A.T.H. is a Wellcome Trust Clinical Research Leave Fellow.
We are grateful to Rachel Dix for excellent technical assistance.