The ATP-sensitive potassium (KATP) channel plays a central role in glucose-stimulated insulin secretion from the pancreatic beta cell. The channel consists of two types of essential subunit: the pore-forming subunit Kir6.2 and the regulatory subunit sulfonylurea receptor 1 (SUR1), which is the target for sulfonylureas. Heterozygous activating mutations in KCNJ11, the gene for the Kir6.2 subunit, can cause both permanent and transient neonatal diabetes, which—in a minority of cases—have neurological features such as developmental delay and epilepsy [15]. The identification of a KCNJ11 mutation can have a major impact on a patient’s treatment. Many patients have transferred from insulin injections to sulfonylurea tablets with an improvement in glycaemic control [1, 69], making the diagnosis of a KCNJ11 mutation clinically important.

The prevalence of diabetes due to a mutation in KCNJ11 is uncertain, with studies reporting mutations in between 34 and 64% of patients with permanent neonatal diabetes [1, 4, 6, 10]. This may, in part, reflect differences in the definition of neonatal diabetes. Strictly speaking, the neonatal period is the first month of life; however, recent definitions of neonatal diabetes have included cases diagnosed before 3 months [11, 12]. Massa et al. suggest that aetiological classifications do not follow this definition and propose the study of patients diagnosed in the first year as a new subgroup of permanent diabetes mellitus of infancy (PDMI) [4]. Only two studies to date have investigated patients aged over 6 months at diagnosis [4, 13], and the latter only looked at six patients diagnosed between the ages of 3 months and 1 year. More information is needed from large series of patients to determine the prevalence of KCNJ11 mutations in different age groups.

Initial reports suggest a possible relationship between the site of the mutation and the phenotype of the patient. The first report of activating KCNJ11 mutations [1] described three patients with severe neurological features (developmental delay and epilepsy) and neonatal diabetes, now known as the DEND syndrome [5, 14]. These patients had mutations not seen in patients with isolated diabetes. Only three of 16 probands with a mutation at residue R201 had neurological symptoms, and all three patients had the less common R201C mutation [1, 4, 6, 7, 9, 10, 13, 15, 16]. Some patients with the V59M mutation have permanent neonatal diabetes with developmental delay but no epilepsy (intermediate DEND [I-DEND] syndrome) [1, 4, 6, 10], whereas others have no reported neurological features [1, 4]. Functional studies support a relationship between the site of the mutation and the ability of the channel to close in the presence of ATP. Disease severity is correlated with the extent of reduction in ATP sensitivity; Kir6.2 mutations associated with DEND syndrome are less sensitive to ATP than those that cause isolated diabetes [14, 17, 18]. The extent to which the clinical phenotype reflects the mutation requires further study.

We report the results of KCNJ11 sequencing in a consecutive series of 239 patients from 21 countries, who had diabetes diagnosed in infancy or early childhood. Studying this large cohort enabled us to gain better insight into the prevalence of KCNJ11 mutations at different ages of diagnosis and allowed further examination of genotype–phenotype relationships.

Subjects and methods


We studied 239 patients with permanent diabetes (without remission) diagnosed before 2 years of age; 54% of the cohort were male. The majority were recruited following a request for referrals to the International Society of Paediatric and Adolescent Diabetes (ISPAD). The patients were recruited from 21 different countries across five continents (Fig. 1). Informed consent was obtained from all participants and the study was conducted in accordance with the Declaration of Helsinki as revised in 2000.

Fig. 1
figure 1

World map showing the 21 countries of origin of patients studied. Numbers in brackets represent the total number of probands referred from each country

KCNJ11 genetic analysis

Genomic DNA was extracted from peripheral lymphocytes using standard procedures. The single exon of KCNJ11 was amplified in three fragments by PCR. All primers included a 5′ M13 tail, and the sequences for fragments 1–3 were: 1F 5′-CCG AGA GGA CTC TGC AGT GA-3′, 1R 5′-TAG TCA CTT GGA CCT CAA TGG AG-3′, 2F 5′-CTG CTG AGC CCT GTG TCA CC-3′, 2R 5′-CAC GCC TTC CAG GAT GAC GAT-3′, 3F 5′-CTA CCA TGT CAT TGA TGC-3′ and 3R 5′-CCA CAT GGT CCG TGT GTA-3′. Sequencing was performed in both directions using universal M13 primers and a BigDye Terminator Cycler Sequencing Kit v1.1 (Applied Biosystems, Warrington, UK), and reactions were analysed on an ABI 3100 Capillary sequencer (Applied Biosystems). Sequences were compared to the published sequence (NM_000525.3) using Sequence Navigator (Applied Biosystems) or Staden analysis software (, last accessed in February 2006). Novel mutations were tested for co-segregation with diabetes in other family members and in 200 normal chromosomes of UK white origin.

Microsatellite analysis

Where possible, family relationships were confirmed using a combination of the following six microsatellites on chromosome 11: D11S902, D11S419, D11S1397, D11S1901, D11S921 and D11S1888.

Clinical studies

All patients underwent routine developmental assessment by a trained paediatrician. The assessment nearest to the age at the time of the testing for Kir6.2 was used for classification. A paediatric neurologist confirmed any abnormal neurological features. The following definitions were used:

  1. 1.

    Developmental delay—this was a marked delay of motor, language and cognitive abilities such that functional age in two or all of these parameters was at least 25% lower than the child’s chronological age.

  2. 2.

    Muscle weakness—clear objective evidence of symmetrical muscle weakness, usually involving the legs.

  3. 3.

    Epilepsy—clinically observed seizures with generalised abnormal activity on an EEG.

These features were only used in the classification of patients’ KCNJ11 disease if there was not an alternative explanation (e.g. severe cerebral oedema following the treatment of diabetic ketoacidosis). Patients were classified into full DEND syndrome when the patient had severe developmental delay (functional age <50% chronological age) and generalised epilepsy diagnosed in the first 12 months of life, and I-DEND syndrome when there was developmental delay but the functional age was >50% of chronological age and the patients did not have epilepsy.

Locally measured HbA1c results were requested from the physician at the time of referral. All assays were said to be DCCT aligned, but precise data on normal reference ranges and methods of analysis cannot be given because of the many different laboratories involved.


Heterozygous activating mutations were identified in 31 of 239 probands with permanent diabetes (Fig. 2). Fourteen of these families have previously been reported [1, 2, 7, 9, 13, 15]. The majority of probands with a KCNJ11 mutation were white (n=24), but mutations were also found in patients of the following ethnicities: Asian (n=2), black African (n=1), Hispanic (n=1), White/African American (n=1), Hispanic/Asian (n=1), Black African/White/American Indian (n=1). We found 14 different missense mutations (Fig. 2), of which seven are novel: H46Y (c.136C>T), R50Q (c.149G>A), G53D (c.158G>A), L164P (c.491T>C), C166Y (c.497G>A), K170T (c.509A>C), Y330S (c.989A>C c.990C>T). These novel mutations were not found in 200 normal chromosomes, and within families these mutations were only present in subjects with permanent, young-onset diabetes. All the mutations affect residues that are conserved in dogs, mice, rats, chickens, Fugu and zebrafish. Three mutations occurred in more than one proband, namely, R201H (n=11), R201C (n=4) and V59M (n=5).

Fig. 2
figure 2

a Illustration of the four Kir6.2 subunits of the channel, showing the 14 mutated residues. The illustration is based on the crystal structure of the potassium channel KirBac1.1 [25]. Each of the four subunits is represented by a different coloured ribbon structure (red, blue, lime, green). The mutated residues in patients with neonatal diabetes are shown in yellow. Residues have been labelled and colour coded according to the region of the protein in which they are located (orange text, residues located in the ATP binding site; grey text, residues located in the slide helix; green text, residue located in the cytosolic NH2 terminal domain; black text, residue located in the domain involved in the interaction of the Kir6.2 and SUR1 subunits; purple text, residues lying in the outer transmembrane helix; and blue text, residue located in the permeation pathway at the mouth of the transmembrane pore).The shadowed band represents the cell membrane and the labels ‘inside’ and ‘outside’ refer to areas outside and inside the cell. b Schematic representation of KCNJ11 showing the 14 mutations identified in 31 unrelated probands. Numbers in brackets represent the total number of probands found to carry the mutation. The seven novel mutations are shown in grey italic print

In all 31 families there was an affected proband born to two unaffected parents, although in four families diabetes was then transmitted to a child or children in the next generation. No families had more than two affected generations. DNA samples were available from both unaffected parents of 22 probands allowing us to establish that these were de novo mutations, as the mutations were not present in the parents, and microsatellite analysis confirmed the family relationships.

The prevalence of KCNJ11 mutations was dependent upon the age at diagnosis (Fig. 3). All subjects with a KCNJ11 mutation were diagnosed with diabetes at the age of 6 months or under; the median age of diagnosis was 5 weeks. The detection rate was 26% before 6 months of age and 0% after. Patients with KCNJ11 mutations were more commonly diagnosed with diabetes in the first 3 months than in the second 3 months (24 vs 7), but the percentage of cases that had a mutation was similar (24 vs 39% p=0.17) in the two age groups. The age at diagnosis for one previously reported patient (I296L, ISPAD 43 [1]) has changed, as further clinical information showed that analysis of glucose in a blood spot used for screening for congenital hypothyroidism on day 7 was 20 mmol/l, which is diagnostic of diabetes.

Fig. 3
figure 3

Frequency of KCNJ11 mutations in probands grouped by age at diagnosis. The number of patients tested is shown in the white columns, with age of diagnosis grouped from 0 to 2 years. The number of KCNJ11 mutations identified within each group is shown by the grey columns and the percentage of positive tests is stated above

The clinical characteristics of the probands and the six additional affected family members with KCNJ11 mutations are shown in Table 1. In all patients diabetes was treated with insulin and most required doses consistent with a full replacement dose. Low birthweight was a consistent feature, with the median birthweight being the third centile and 67% being classified as small for gestational age as their weights were below the 10th centile.

Table 1 Clinical and biochemical characteristics of the 37 patients with KCNJ11 mutations from 31 families

In addition to diabetes, a spectrum of neurological features were present in 11 of 37 (30%) patients with a KCNJ11 mutation. Five of these patients (with the Q52R, G53D, V59G, C166Y and I296L mutations) had the DEND syndrome with all three features of developmental delay, epilepsy and neonatal diabetes, two of which are previously unreported. The first unreported case (C166Y) had profound developmental delay and microcephaly and had been diagnosed with West syndrome at 3 months of age. The second case (G53D) had developmental delay and seizures during infancy and an abnormal EEG. In addition, there were two patients with developmental delay (both patients had the R201C mutation) and four patients with developmental delay and muscle weakness (all four patients had the V59M mutation), but none of these six patients had epilepsy. We describe this as I-DEND syndrome [5]. The patients with neurological features did not have a more severe beta cell phenotype, as shown by birthweight (2,550 vs 2,548 g, p=0.63) and age at diagnosis (5 vs 6 weeks, p=0.82).

We found evidence to support a genotype–phenotype relationship for the common KCNJ11 mutations (see Table 2). The most common mutation, R201H, was associated with isolated diabetes in all of the 14 patients, in contrast to V59M where four of five patients had neurological features (I-DEND syndrome) (p=<0.001). Two of the seven patients (both probands) with the R201C mutation had developmental delay in addition to diabetes.

Table 2 Phenotypes of patients with the three most common mutations


We report on 31 probands with 14 different heterozygous activating mutations in KCNJ11, identified by mutation screening 239 patients with permanent diabetes before the age of 2 years. The detection of KCNJ11 mutations is increasingly important, as many cases have been able to discontinue insulin injections and achieve better glycaemic control on sulfonylurea tablets [1, 69]. We show that in this large series, mutations in this gene are only detected in those diagnosed with diabetes in the first 6 months of life, and provide further evidence for a genotype–phenotype relationship.

We describe seven novel mutations (H46Y, R50Q, G53D, L164P, C166Y, K170T, Y330S). Although functional studies have not been performed for all of these, we are confident that these mutations are pathogenic since (1) none of the mutations were found in 200 normal chromosomes; (2) all seven residues are conserved in other species; and (3) all seven mutations were shown to be spontaneous mutations. In five of the seven cases, different mutations at the same amino acid have previously been reported: (R50P [4], G53S[2], G53R[2], G53N[10], K170R [4], K170N [4], C166F [19], Y330C [6, 10]). The finding of multiple mutations at the same codon highlights the functional importance of these residues. Functional studies have already established an altered response to ATP by KATP channels with mutations in Kir6.2 at the following residues: R50, Q52, G53, V59, R201, L164, C166 or I296 [1, 2, 14, 17, 18, 20, 21].

In all families there was an affected patient with a KCNJ11 mutation whose parents were unaffected. This is consistent with a spontaneous mutation in the affected patient. Where DNA was available from both parents (n=22), we were able to confirm a de novo mutation. The high prevalence of spontaneous mutations is consistent with other series where familial cases were rare or not found [4, 6, 10]. We have previously reported a family in which the unaffected father of two half-siblings with a heterozygous KCNJ11 mutation showed no evidence of the mutation in leukocyte DNA [2], suggesting that he was a germline mosaic carrier of the mutation. The possibility of germline mosaicism means that the risk of a further sibling being affected is significantly higher than the very low chance of a second spontaneous mutation.

In this large series, mutations in KCNJ11 were found exclusively in those diagnosed with diabetes within the first 6 months of life. Interestingly, although patients were more likely to have been diagnosed in the first 3 months (Fig. 2), the percentage of cases that had a mutation was similar in those diagnosed in the first 3 months (24%) and in those diagnosed between 3 and 6 months (39%). Only those diagnosed in the first 3 months have previously been classified as having neonatal diabetes, and this result confirms that a broader definition is needed when considering KCNJ11 mutations [4]. A cut-off point at 6 months is consistent with previous work looking at autoantibodies and HLA suggesting that type 1 diabetes is rare in those aged below 6 months [22]. Our study would not support testing for KCNJ11 mutations in patients diagnosed after 6 months, although relatives of probands with transient neonatal diabetes have been diagnosed at 3, 5, 22 and 26 years [2, 3]. The detection rate of 26% in our series was lower than the 34–64% previously reported [1, 4, 6, 10], although the 95% confidence limits of all series overlap. This may be because in this international series, detailed phenotypic assessment was not performed on a longitudinal basis before testing. Hence, patients who remitted after DNA was analysed will not have been excluded, and phenotypic selection is likely to be stricter in the long-term longitudinal series established in France and Italy [4, 10].

We have identified 11 patients who have neurological features as well as permanent diabetes. This provides further evidence that some patients with a KCNJ11 mutation have neurological features as a result of the mutated KATP channel. Five patients have full DEND syndrome with the key characteristics of developmental delay, epilepsy and neonatal diabetes. There is a sixth French patient with severe neurological deficit and generalised epilepsy with the characteristics of West syndrome [19]. Functional studies have suggested that the mutations associated with DEND syndrome are more severe, with the channel less likely to close in the presence of ATP compared with mutations associated with isolated neonatal diabetes [14]. The other six patients have the I-DEND syndrome (neonatal diabetes with moderate developmental delay and/or muscle weakness but not epilepsy), bringing the total number of reported cases of this phenotype to 14 [4, 6, 10]. The most common mutation associated with I-DEND syndrome is V59M (ten of 15 cases), and functional studies have shown that the V59G mutation found in a patient with DEND syndrome has a more severe effect than V59M [23].

Our series provides further evidence for a genotype–phenotype relationship for KCNJ11 mutations. The most common mutation, R201H, has now been reported in 20 patients [6, 8, 10, 16], with none of the patients having neurological features. In contrast, of the 13 patients with the V59M mutation, ten (77%) have intermediate DEND syndrome [1, 4, 6, 10]. The mutations causing the full DEND syndrome have not been described in patients with isolated diabetes. It is interesting that three of nine (33%) patients with the R201C mutation have intermediate DEND [4, 10, 13, 15], which is not seen in those with the R201H (0 of 20) mutation at the same residue. This has only become clear with larger numbers and is in keeping with the greater severity of mutation seen in vitro [23]. It is interesting that the V59M and R201C mutations are not associated with a consistent neurological phenotype. This may, in part, reflect differences in assessment (e.g. developmental delay is difficult to detect in the first year of life), but it also is likely that there is phenotypic variation between subjects with the same mutation. Variation in severity of beta cell dysfunction has been described in subjects with the R201H mutation [1, 7, 16].Therefore, no genotype–phenotype relationship is absolute and there must be modification by other genetic and/or environmental factors.

Localisation of the mutated residues in structural models of the channel show that residues associated with isolated diabetes (R50, L164, R201, Y330) cluster around the ATP binding site [17, 20], thus decreasing ATP sensitivity by reducing ATP binding. Residues associated with the more severe phenotype (Q52, G53, V59, C166, I296) frequently occur at a distance from the ATP binding site, either within the pore-forming domain or the slide helix area of the protein [2, 14, 17, 20]. Functional analysis of these mutations shows that the mutated residue within the Kir6.2 protein correlates with the observed phenotype [14, 17]. All mutations causing permanent diabetes, tested to date, have been shown to lower the sensitivity of the channel to ATP, reducing the channel’s ability to close in response to elevated levels of ATP. The mutations associated with DEND syndrome show the most profound reduction in response to ATP, with mutations causing isolated permanent diabetes, having a reduced response compared with those associated with transient diabetes [2, 14, 17].

Our study does have limitations. First, we did not test any patients diagnosed over the age of 2 years, and therefore we cannot make conclusions about this age group. Secondly, developmental delay was not assessed by formal criteria at a set age by a single physician as a result of the age distribution and international location of patients. This is likely to have resulted in a lack of clinical detection of some of the neurological characteristics or inconsistent reporting, especially as those patients thought not to have problems will not have been referred to specialist neurologists. Patients with KCNJ11 mutations were aged from 6 days to 56 years old at the time of this study. In younger patients, the neurological phenotype may not have been clinically detectable. The third limitation is that some of the younger patients may not have had ‘permanent’ diabetes: although they were diabetic when the study was performed they were still young (<6 months) and may remit later and so have transient diabetes. The final limitation is that this study is a referred series rather than an epidemiological sample, so those sent for screening may be subject to selection bias. Population-based studies would establish if this were the case.

In conclusion, we have shown that the majority of patients with mutations in KCNJ11 are diagnosed with diabetes before the age of 6 months but are not confined to those with a formal diagnosis of neonatal diabetes. We provide further evidence for discrete neurological phenotypes (DEND syndrome and I-DEND syndrome) as a consequence of a mutated Kir6.2 channel. Our work supports a clear genotype–phenotype relationship, supporting the hypothesis that the phenotype observed is predominantly related to the nature of the mutation within the protein. The major therapeutic implications for those found to have a KCNJ11 mutation highlights the importance of screening all patients who are diagnosed with permanent diabetes in the first 6 months of life. Our study does not support routine testing after this age.