Mutations in ATP-Sensitive K⁺ Channel Genes Cause Transient Neonatal Diabetes and Permanent Diabetes in Childhood or Adulthood

We studied 97 patients with diabetes diagnosed in the first 6 months of life whose diabetes remitted before the age of 5 years. Sixty-four have been previously reported (1,3,4,12–15). The patients were referred from 14 different countries across four continents and were either recruited following a request for referrals to the International Society of Pediatric and Adolescent Diabetes rare diabetes collection or following referral to the Wessex Regional Genetics Laboratory for 6q24 abnormality testing. The study was conducted in accordance with the Declaration of Helsinki, as revised in 2000. Informed consent was obtained from all patients, with parental consent given on behalf of children.

Genomic DNA was extracted from peripheral lymphocytes using standard procedures. In all patients, analysis of the 6q24 locus was undertaken using previously described methods to detect duplications, uniparental isodisomy and methylation abnormalities (1,13). In patients for whom no 6q24 abnormality was identified, the single exon of KCNJ11 was amplified in three overlapping fragments, as previously described (16). The 39 exons of ABCC8 were analyzed in all patients in whom no KCNJ11 mutation was identified. The ABCC8 gene was amplified in 38 fragments using previously described primers (7). PCR products were sequenced using standard methods on an ABI 3100 or ABI 3730 (Applied Biosystems, Warrington, U.K.). Sequences were compared with the published sequences ([KCNJ11 − NM_000525.3] [ABCC8 − NM_000352.2]) using Staden analysis or Mutation Surveyor version 2.61. Mutations were tested for cosegregation with diabetes in other family members and in 200 normal chromosomes from U.K. Caucasians. Where possible, family relationships were confirmed using a panel of six microsatellite markers on chromosome 11p (16). Clinical characteristics were obtained from the patient's hospital records with assistance from their physician. Clinical characteristics are presented as median (range), and comparative statistics predominantly used the Mann-Whitney U test.

We have determined the genetic etiology in 94 of 97 (97%) probands diagnosed with diabetes before the age of 6 months whose diabetes had subsequently entered remission (Fig. 1). The majority of patients (69 of 97) had an abnormality at the 6q24 locus. Of these patients, 27 (39%) had paternal uniparental isodisomy, 27 (39%) had a duplication of the paternal allele, and 15 (22%) had a methylation defect. For the 28 remaining patients, in whom no 6q24 aberration was identified, sequencing of KCNJ11 and ABCC8 demonstrated that 25 probands were heterozygous for K_ATP channel mutations.

Twelve of the probands had a heterozygous mutation in the KCNJ11 gene; six of these patients (with five different mutations) have been reported previously (4,12). We found eight different KCNJ11 mutations; R34C (c.100C>T), G53R (c.157G>C), G53S (c.157G>A), E179A (c.536A>C), I182V (c.544A>G), E227K (c.679G>A), E229K (c.685G>A), and R365H (c.1094G>A). Two mutations were identified in more than one proband (E227K, n = 2, and E229K, n = 4). The mutations R34C, E179A, and R365H are novel. We did not observe these mutations in 200 normal chromosomes, and all mutations affect residues that are conserved in human, mouse, rat, and dog (http://genome.ucsc.edu).

Ten different ABCC8 gene mutations were identified in 13 probands: D209E (c.627C>A), D212N (c.634G>A), D212I (c.634 G>A 635A>T), V324M (c.970G>A), L451P (c.1352T>C), R826W (c.2476C>T), R1183W (c.3547C>T), R1183Q (c.3548G>A), R1380C (c.4138C>T), and R1380H (c.4139G>A). All were identified in one proband, with the exception of R1183W, which was identified in four probands. All mutations were novel except for R1380C and R1183Q (8). We did not observe these mutations in 200 normal chromosomes, and all affect residues that are conserved in human, mouse, rat, and dog (http://genome.ucsc.edu).

Seven (28%) probands had de novo mutations as confirmed by the absence of the mutation in the unaffected parents (KCNJ11, n = 3, and ABCC8, n = 4). Microsatellite analysis was used to verify the family relationships in all but one case with an R1380C ABCC8 mutation, in whom parental DNA was unavailable. In the remaining 17 probands, the mutation had been inherited from a parent (partial pedigrees are shown in Fig. 2). In these families, 17 of 26 parents or siblings with the K_ATP channel mutation had diabetes; in contrast, only 1 family member without the mutation was affected (P < 0.001). However there was not complete cosegregation of the mutations with diabetes because in eight probands the mutation was inherited from a nondiabetic parent. Neonatal diabetes (diagnosed in the first 6 months) was described in five siblings and four parents. Two mothers (identical twins, Fig. 2) were reported to have had diabetes in the neonatal period, although information regarding their ages at diagnosis and remission was not available. Nonpenetrance was seen in families with eight different mutations, suggesting that it was not a feature of specific mutations.

The characteristics of diabetes in the probands with KCNJ11 or ABCC8 mutations are shown in Table 1. There was no significant difference in clinical characteristics between these two groups. Reduced birth weight was a consistent feature of both KCNJ11 and ABCC8 mutations (25th vs. 11th percentile; P = 0.26). Diabetes was most frequently diagnosed between 0 and 8 weeks (86%), with all cases diagnosed before 17 weeks. None of the patients with an ABCC8 mutation had relapsed at the time of this study; however, this may be due to the younger age of these patients compared with those with KCNJ11 mutations.

In addition to diabetes, neurological features were identified in four probands (16%) and were associated with both KCNJ11 and ABCC8 mutations. Two probands and one family member with a KCNJ11 mutation had neurological features. The first proband with an E229K KCNJ11 mutation has reported speech delay and was diagnosed with autistic spectrum disorder at the age of 3 years. Interestingly, the affected sibling and unaffected mother who both carry the same mutation have no reported speech problems, although the sibling is currently 2 years of age and still in the early stages of language development. The E229K mutation has been identified in five affected patients from three further families, with no reports of speech or developmental delay in any of the mutation carriers. Mild learning difficulties were observed in a second proband and his mother who both have the G53R KCNJ11 mutation. This mutation has not been identified in any other families to date; therefore, it is difficult to conclude whether these complications are a feature of the mutation or whether environmental and/or other genetic factors are involved.

Neurological features were also reported in three patients with an ABCC8 mutation. Muscle weakness was identified in two cousins with a D212I mutation. The proband had muscle hypotonia until the age of 8 months, and her cousin has been diagnosed with motor developmental delay, although no muscle weakness was reported in either of their affected mothers. The most severe neurological phenotype was identified in a proband with a de novo ABCC8 mutation (R1183W). This patient had one episode of tonic posturing with right facial involvement following admission to the hospital with diabetic ketoacidosis. He subsequently had two episodes of generalized seizures shortly after diagnosis of diabetes. No further seizures have been reported since insulin therapy was started, and it is likely that these seizures were a result of the severe hyperglycemia rather than as part of the DEND (Developmental Delay, Epilepsy, and Neonatal Diabetes) syndrome (17). Neurological features have not been reported in any of the five other patients with the R1183W mutation.

The clinical characteristics of the probands with K_ATP channel mutations were markedly different when compared with subjects with 6q24 abnormalities (Table 2). Patients with a 6q24 abnormality presented with a more severe clinical phenotype, as shown by their lower percentile birth weight (<1st vs. 12th; P < 0.001) and earlier age of diagnosis (0 vs. 4 weeks; P < 0.001). For patients with a K_ATP channel mutation, the initial episode of diabetes was longer, with diabetes remitting significantly later than in patients with a 6q24 abnormality (35 vs. 13 weeks; P < 0.001). There is a trend toward an earlier age of relapse in patients with a K_ATP channel mutation (P = 0.07), and this may become more evident as the patients are studied over a longer period of time.

The characteristics of the diabetes varied according to the age at which it was first diagnosed (see Table 3). Remission of diabetes occurred in 32 of 35 (91%) patients diagnosed before 6 months but in none of those diagnosed after 6 months (P < 0.001) (Table 3 and Fig. 3). The median age at remission was 35 weeks (range 2–208). The three patients diagnosed before 6 months who had not entered remission were 1, 2, and 43 years of age at mutation screening.

All patients were treated with insulin during their initial episode of diabetes and during any subsequent relapse of diabetes, except for one proband and his father with a KCNJ11 mutation (E227K), a father with a KCNJ11 mutation (E229K), and a mother with an ABCC8 mutation (D209E), who were treated with sulfonylureas. Two probands and four family members have now attempted to transfer off insulin and onto sulfonylureas using a protocol similar to that described for KCNJ11 mutations (18). Transfer was successful in five cases (83%) but unsuccessful in a mother with a G53R mutation. Her son, who carries the same mutation, has successfully transferred onto sulfonylureas, suggesting that the observed lack of response to sulfonylureas is not a feature of the mutation.

We were able to determine the genetic etiology in all but 3 of 97 probands diagnosed with TNDM. Interestingly, all three patients had an atypical phenotype. The first patient was born at 30 weeks’ gestation, and his transient hyperglycemia was possibly a consequence of the premature birth. Early gestational age was not a consistent feature in patients with either a 6q24 anomaly or a K_ATP channel mutation. The second patient stopped insulin treatment at the age of 4 years, but this was recommenced 1 year later. He also developed epilepsy at the age of 9 years, which has not been reported in any other patients in our series. The third patient has had four episodes of remission and relapse, a unique phenotype that has also not been described in any other patients included in this study. In addition, the coding sequence of the glucokinase gene was normal in all three patients. This phenotypic variation suggests further rare distinct etiologies for TNDM.

We have shown that mutations in the KCNJ11 and ABCC8 genes that encode the Kir6.2 and SUR1 subunits of the K_ATP channel are a major cause of TNDM, accounting for 29% of all cases and 89% of non-6q24 TNDM. As a result, a genetic diagnosis could be made in 97% of patients with neonatal diabetes that remitted. These estimates of the relative prevalence of the three major genetic subgroups are likely to be robust, as this study of a consecutive series of 97 TNDM patients is more than twice the size of the largest previous study. In our series, KCNJ11 (n = 12) and ABCC8 (n = 13) mutations were equally represented, which is in marked contrast to PNDM, for which there are >60 KCNJ11 mutations and only 3 ABCC8 mutations reported (7,8). The only previous comprehensive study is from France, which showed that a genetic diagnosis was possible for 33 of 44 patients with TNDM (8). This is in contrast to the current study, in which the genetic etiology remains unknown in only 3 of 97 patients. The reason for this discrepancy is unknown; it could result from differences in the patients studied or in the detection of K_ATP channel mutations. KCNJ11 mutations have now been identified in a total of 18 patients with TNDM (5,6,19). Our study emphasizes that both KCNJ11 and ABCC8 mutations are an important cause of TNDM.

There is strong evidence that the mutations detected in KCNJ11 or ABCC8 are the cause of TNDM in these patients. All of the mutations occur at residues that are conserved across multiple species, alter the coding sequence (an essential requirement for an activating channel mutation), and are not found in normal control subjects or patients with PNDM. The strongest evidence for an etiological role is the de novo mutations, as spontaneous mutations are rare, even in genes as large as ABCC8 (8.5 × 10⁻⁵; 20). However, de novo mutations are only seen for 4 of the 14 novel mutations (29%) in contrast to PNDM, in which 84% of cases represent spontaneous mutations (21). For the majority of cases (68%), the mutation is inherited from a parent, and although cosegregation of TNDM is rare, all cases of TNDM in the families have inherited the mutation. Fifty percent of parents have diabetes diagnosed before 45 years, which is rare in the general population (<0.2%) (calculated from 22) and is not seen in any of the non–mutation-carrying parents. The final evidence that these mutations are pathogenic is that functional studies for all mutations performed to date (G53R, G53S, I182V, E227K, and E229K in KCNJ11) have shown altered response to ATP in the presence of magnesium ions (4,8,19). The presence of de novo mutations identified in five probands strongly supports an etiological role for the I182V, E229K, R1380C, and R1183Q mutations. Therefore, these mutations are highly likely to be causative of TNDM, even in the absence of complete cosegregation.

We have identified a number of commonly mutated residues in patients with TNDM within both K_ATP channel genes. KCNJ11 mutations in our TNDM series cluster with residue G53 mutated in 2 of 12 probands and E227 and E229 residues mutated in 6 of 12 probands. Structural analysis has shown that G53 lies in a region linking the ATP binding site to the transmembrane domain (4), whereas E227 and E229 lie distant to the ATP binding site at the interface between Kir6.2 subunits and are thought to influence channel gating (19).

The ABCC8 mutations are located throughout the SUR1 protein. The residue R1183, which is located at a position involved in joining transmembrane domain 2 to nucleotide binding domain 2, was mutated in five probands with two different residues, tryptophan and glutamine substituting arginine. The second most commonly affected residue was R1380, which is located in nucleotide binding domain 2 and is mutated in two probands. There are a cluster of mutations (D209E, D212N, and D212I) in the intracellular region that links the transmembrane domain with the gatekeeper module (8).

Functional studies of channels with the TNDM-associated KCNJ11 mutations G53S, G53R, and I182V showed a moderate reduction in activity in the presence of ATP, which is less marked than the reduction seen with PNDM mutations (4). This suggested clear association of the genotype with functional effect as well as clinical phenotype, but further studies have shown that there is some overlap between the magnitude of the K_ATP channel currents in TNDM- and PNDM-associated mutations (19). Furthermore, the V252A and R201H mutations have been described in both patients with PNDM and TNDM (6,19), so the genotype/phenotype correlation is not absolute.

We have shown that patients with K_ATP channel mutations, when diagnosed within the neonatal period, have similar clinical characteristics irrespective of whether KCNJ11 or ABCC8 is mutated (Table 1). The TNDM probands with K_ATP channel mutations show a biphasic course in which diabetes is diagnosed before age 6 months. They then typically go into remission between age 6 and 12 months and are likely to relapse during adolescence or early adulthood. None of the patients diagnosed with diabetes after 6 months of age had a remission of diabetes, even if they had the same mutation as patients who remitted. Interestingly, birth weight, a reflection of insulin-mediated growth and, hence, insulin secretion in utero, is reduced to a similar extent whether the patient is diagnosed before 6 months or not. These findings imply that the K_ATP channel mutations have a biphasic course, and patients diagnosed later may have had a period of hyperglycemia that was undetected in the neonatal period. This could explain the apparent phenotypic heterogeneity with the same mutation. Importantly, this means that K_ATP channel mutations can present outside the neonatal period, and, if the parents in our families had not had a child with TNDM, they would have been diagnosed as having type 1 or type 2 diabetes. It will be important to test cohorts with a clinical diagnosis of type 1 diabetes to see how prevalent these mutations are, especially because it will have implications for treatment.

The explanation of the biphasic course in both types of K_ATP channel mutations is not known. A fixed-channel defect would be expected to produce a fixed phenotype, as seen with K_ATP channel mutations causing PNDM. The change seen in glucose tolerance during remission and subsequent relapse likely reflects either changes in insulin requirements or changes in the number of β-cells or both. As insulin levels are very low in nondiabetic neonates (23), this would suggest that the initial diabetes does not reflect increased insulin requirements at this time, although the relapse around puberty may reflect the associated increase in insulin resistance. The biphasic course seen in 6q24 TNDM is still not fully understood, even with the creation of a good mouse model (24).

Neurological complications were present in addition to diabetes in six patients with an ABCC8 or KCNJ11 mutation. In all cases, the affected individual had been diagnosed with diabetes before the age of 6 months. In one family, both mutation carriers had learning difficulties. In all other families, neurological features did not cosegregate with the mutation, suggesting that neurological phenotype is not a consistent feature of these mutations.

Our series provides strong evidence that clinical characteristics differ markedly between patients with a K_ATP channel mutation and those with a chromosome 6q24 anomaly. In addition to diabetes, seven patients with a 6q24 abnormality (30%) had macroglossia and two patients had umbilical hernia (9%). These abnormalities were not present in any of the patients with a K_ATP channel mutation. Patients with K_ATP channel TNDM have a higher birth weight (2,570 vs. 1,950 g), are diagnosed later (4 vs. 0 weeks), and enter remission later (35 vs. 13 weeks) than patients with 6q24 abnormalities. Although these clinical characteristics can aid in the identification of the cause of TNDM, there is overlap in the range of values (Table 2). Therefore, while clinical features can guide the order in which genetic tests are performed, a molecular genetic diagnosis is still required.

There is a clear genotype/phenotype relationship. None of the K_ATP channel mutations identified within this series have been found in patients with PNDM. Only one report, to date, has identified R201H, a well-characterized PNDM mutation, in a patient with TNDM (6). Between the TNDM and PNDM groups, there are no significant differences between the percentile birth weights and age of diagnosis (12th vs. 3rd percentile and 4 vs. 5 weeks) (16). Distinction between the two subgroups can be made only by the presence of a period of remission in the TNDM patients. One difference between the groups is the mode of inheritance. In our series, a greater number of probands (68%) have a familial mutation. This is in contrast to PNDM, in which 84% of all mutations occur de novo (21).

Identification of a K_ATP channel mutation has important implications for patient clinical management. Recent reports (19,25,26) have shown that most patients with a KCNJ11 mutation and PNDM are able to transfer from insulin onto sulfonylurea treatment with an improvement in glycemic control in all cases reported. Babenko et al. (8) reported two patients with an ABCC8 mutation who were successfully treated with sulfonylureas after their TNDM relapsed. In our series, 3 of the 14 patients who developed diabetes outside the initial 6 months (either as a relapse or initial diagnosis) were initially treated with sulfonylureas. Since the diagnosis was made, a further five patients have successfully transferred onto sulfonylureas from insulin using a similar protocol to that described for the transfer of patients with PNDM and a KCNJ11 mutation (18). For the remaining patients, transfer is either in progress or the patients are currently in remission.

In conclusion we have shown that a genetic diagnosis is possible for 97% of patients diagnosed with TNDM. Mutations of the K_ATP channel genes, ABCC8 and KCNJ11, account for the majority (89%) of non-6q24 TNDM. Diabetes associated with these K_ATP channel mutations has a biphasic course, although neonatal diabetes that enters remission may not be detected. Patients therefore may present with permanent diabetes diagnosed in childhood or adulthood. In addition, not all mutation carriers may be diabetic when the mutation is detected in a proband. The therapeutic implications for those patients found to have a K_ATP channel mutation highlights the importance of screening all patients with non-6q24 TNDM for mutations in ABCC8 and KCNJ11.

We acknowledge funding from the Sir Graham Wilkins studentship to S.E.F. The study was funded by the Wellcome Trust and was supported by the European Union (Integrated Project EURODIA LSHM-CT-2006-518153 in the Framework Programme 6 of the European Community). A.T.H. is a Wellcome Trust Research Leave fellow. A.L.G. is a Diabetes UK R.D. Lawrence Research Fellow.

The authors thank all of the families for participating in this study. Our thanks go to the many referring clinicians who include Ijaz Ahmad, Jesús Argente, Chandar Batra, Mark Bone, Jill Challener, Ethel Codner, Elizabeth Davis, Dorothy Deiss, Jan Fairchild, Peter Fowlie, Mathias Herr, Sabine Hofer, Mary Jetha, Olga Kordonouri, Michael MacDonald, Kathryn Noyes, Laurent Legault, Derek Sandeman, Annabelle Slingerland, Zdenek Sumnick, Peter Swift, Charalambos Theodoridis, Verena Wagner, and Esko Wiltshire. We also thank Andrew Parrish for his technical assistance and Dr. Beverley Shields for statistical analysis.