Insulin Mutation Screening in 1,044 Patients With Diabetes: Mutations in the INS Gene Are a Common Cause of Neonatal Diabetes but a Rare Cause of Diabetes Diagnosed in Childhood or Adulthood

We studied five groups of patients with permanent diabetes at the time of referral (Table 1). The PND group consisted of 141 patients with diabetes diagnosed in the first 6 months of life who were on insulin treatment at the time of referral. Mutations in KCNJ11 had been excluded by sequencing in all 141 patients, and ABCC8 mutations had been excluded in the 129 cases where there was sufficient DNA. Patients diagnosed outside the first 6 months of life were divided into those diagnosed in infancy (between 27 and 52 weeks of life, n = 86) and those diagnosed in early childhood (between 53 and 104 weeks of life, n = 58). We also screened 296 cases who met minimal diagnostic criteria for MODY (at least two generations affected and at least one subject diagnosed before 25 years of age). Mutations in the HNF1A, HNF4A, and GCK genes had been excluded by sequencing in 189, 81, and 56 cases, respectively. Finally we screened 463 nonobese patients with young type 2 diabetes diagnosed before 45 years of age. Of these, 45% had a diabetic first-degree relative. Mutations in HNF1A, HNF4A, and GCK were excluded by sequencing in 180, 76, and 87 cases, respectively.

Patients diagnosed between 0 and 2 years of age (n = 285) were predominantly recruited in Exeter through the International Society of Pediatric and Adolescent Diabetes (ISPAD) group. These were supplemented by 39 patients from the Barts and Oxford (BOX) study and the British Diabetes Association 1972–1981 cohort, as previously described (2). In addition, we also screened 49 patients with isolated hyperinsulinism, diagnosed at a median age of 7 weeks (0–728 weeks), where mutations in the K_ATP channel genes (KCNJ11 or ABCC8) had been excluded by sequence analysis.

Clinical characteristics are presented as median (range) and comparative analysis using a Kruskal Wallis or Mann-Whitney U test. The summary clinical details of these groups are shown in Table 1.

Mutations were tested for cosegregation with diabetes in family members, and the INS gene was sequenced in 111 U.K. population Caucasian control subjects. Clinical characteristics were obtained from hospital notes with assistance from the referring clinician.

Genomic DNA was extracted from peripheral leukocytes using standard procedures. Coding exons 2 and 3 of the INS gene were amplified by PCR. Sequence-specific primers for each exon (10) were tagged with 5′ M13 tails to allow sequencing to be performed with a “universal” M13 primer. A new primer for the first coding exon was used: INS_Exon2F-TGTAAAACGACGGCCAGTTGGCTGGGCTCGTGAAG and INS_Exon2R CAGGAACACGCTATGACCCCCTTCTGCCCATGCTG.

Single-strand sequencing was carried out using standard methods on an ABI 3730 (Applied Biosystems, Warrington, U.K.). Sequences were compared with the published sequence (NM_000207) using Mutation Surveyor v2.61. Changes in the sequence were checked against published polymorphisms and mutations and for conservation across species. We used a panel of microsatellites for chromosome 20q (16) or chromosome 11p (17) to confirm family relationships for sporadic cases of diabetes where neither parent was shown to carry the mutation found in their child.

Heterozygous INS mutations were found in 33 of 141 (23%) probands diagnosed before 6 months, 2 of 86 (2%) between 6 and 12 months, and none of 58 diagnosed between 12–24 months of age. Only 1 of 296 (0.3%) MODY and 1 of 463 (0.2%) young type 2 diabetic probands had possible mutations identified by sequencing. No mutations were found in the 49 patients with hyperinsulinism. The details of the mutations found and the associated clinical characteristics are discussed below.

We identified 16 different heterozygous mutations in 35 probands with diabetes diagnosed in the first year of life (Fig. 1). Of the 35 families, 12 have been reported previously (10). Of the 16 mutations found, 10 have previously been described (10): A24D (five families), G32S (two families), G32R (two families), C43G, G47V (two families), F48C (five families), R89C (six families), G90C, C96Y (three families), and Y108C. There are six novel mutations: H29D (c.85C>G, p.His29Asp), L35P (c.104T>C, p.Leu35Pro), G84R (c.250G>A, p.Gly84Arg), C96S (c.287G>C, p.Cys96Ser), S101C (c.302C>G, p.Ser101Cys in two families), and Y103C (c.308A>G, p.Tyr103Cys). These residues are conserved across species (human, mouse, dog, horse, lizard, and Xenopus), with the exception of Y103 (F103 in Xenopus) and G84 (R84 in platypus). These mutations were not seen in 222 U.K. Caucasian control chromosomes.

Three mutations (A24D, F48C, and R89C) account for 16 of 35 (46%) of cases, but only R89C occurs at a CpG dinucleotide, where the spontaneous deamination of methylcytosine is frequent. Of the 16 mutations, 9 either abolish a cysteine residue or create an additional cysteine.

Six families demonstrated autosomal dominant inheritance of diabetes that cosegregated with the INS mutation, including one previously published family ISPAD 155 (10) (Fig. 2). An additional 11 family members were shown to have an INS mutation. Six were diagnosed before 6 months, three between 6 and 12 months, 1 at 7 years, and 1 at 22 years of age. In 29 of 35 (83%) of probands, an affected child was born to unaffected parents, representing probable de novo mutations. In 17 families, DNA was available from both unaffected parents, and microsatellite analysis established that these were de novo mutations (including 11 previously published families [10]). In family ISPAD180, two affected brothers had the S101C mutation, but the mutation was not detected in leukocyte DNA from either parent (Fig. 2). This pattern of inheritance is consistent with germline mosaicism. The clinical features of this family have previously been described (18).

Clinical characteristics are provided for all INS mutation carrying probands diagnosed under 1 year of age and their affected family members (Table 2). The median age at diagnosis of the INS mutation carriers was 11 weeks. Presentation was either with symptomatic hyperglycemia (41%) or diabetic ketoacidosis (59%). All patients were treated with insulin replacement therapy and most with a full replacement dose (>0.5 units · kg⁻¹ · day⁻¹) (median 0.7 units · kg⁻¹ · day⁻¹ [range 0.3–1.9]). Glycemic control was variable (A1C median 7.95% [range 4.6–13.8]). Autoantibodies, when measured, were not detected in any cases, although the two affected siblings from family ISPAD180 were previously reported with autoantibodies to the exocrine pancreas but not the islet at diagnosis (18).

Birth weights were reduced, consistent with in utero growth retardation due to reduced insulin secretion. The median birth weight was 2.7 kg (range 1.7–3.9), corresponding to the 6th percentile (range <1st–87th). There was no difference in the age at diagnosis for male INS mutation carriers compared with female carriers (median 10 weeks [range 0–1,144] for male vs. 13 weeks [1–35] for female carriers; P = 0.9), but birth weight was lower for male carriers (median centile birth weight <1 [range <1–87] for male vs. 11th centile [<1–83] for female carriers; P = 0.048).

Figure 3 illustrates the penetrance of INS mutations in families where the proband was diagnosed in infancy. The percentage of patients with diabetes increases from 24% in the first month to 85% at 6 months and to 96% at 1 year.

The clinical characteristics of patients with an INS, KCNJ11, or ABCC8 gene mutation were compared (Table 3). No difference in sex, birth weight, or gestational age was observed between the three groups, but patients with an INS mutation were diagnosed later (median 11 weeks [range 0–1,144]) than those with a K_ATP channel mutation (median 8 weeks [range 0–40], P < 0.01).

The 33 INS mutation carriers diagnosed before 6 months of age are from a cohort of patients with PND referred to Exeter. This cohort includes 87 patients with KCNJ11 mutations, 29 with ABCC8 mutations, and 22 with mutations in the GCK, EIF2AK3, or FOXP3 genes (10, 8, and 4 cases, respectively). There are also 108 subjects in whom no mutations were found after sequencing the KCNJ11, ABCC8, and INS genes. We were unable to complete sequencing of these three genes in 21 cases due to insufficient DNA. The prevalence of the different genetic subtypes within the 279 fully tested cohort members is shown in Fig. 4. Sequence analysis for GCK was performed only for consanguineous pedigrees, and EIF2AK3/FOXP3 mutation testing was guided by extra-pancreatic features and/or consanguinity; thus, these etiologies may be underestimated.

Two different missense mutations were identified in two families: one in a proband with MODY (1 of 296) and one in a proband with young type 2 diabetes (1 of 463). The novel mutation, R6C (c.16C>T, p.Arg6Cys), was found in a U.K. MODY family and was not present in 222 U.K. Caucasian control chromosomes. The novel mutation L68M (c.202C>A, p.Leu68Met) was identified in a family of Turkish origin with young type 2 diabetes and was not seen in 170 Turkish control chromosomes.

The R6C mutation cosegregated with diabetes, as it was found in the proband, the proband's mother, and the maternal grandmother, who were diagnosed with diabetes at 15, 15, and 65 years of age, respectively. The clinical features in the three affected family members were consistent with a diagnosis of MODY: they were nonobese (BMIs 24.1, 26.9, and 29.3 kg/m²), and their diabetes was noninsulin dependent. The proband had 10 years on diet and then oral agents before starting low-dose insulin (0.2 units · kg⁻¹ · day⁻¹), her mother was treated with diet for 40 years before starting oral agents at the age of 55 years, and the maternal grandmother has been on diet treatment for the 8 years since diagnosis. The L68M mutation was found in a proband diagnosed with diabetes at the age of 30 years (BMI 17 kg/m²). She was treated with diet for 1 year and started oral agents but soon required insulin treatment. Her father was diagnosed at age 31 years and was insulin treated; he died from diabetes complications, and DNA from her unaffected mother was not available (Fig. 5). Mutations in the GCK, HNF1A, or 4A genes were excluded in both probands by sequence and dosage analysis.

The R6 residue is conserved across mammalian species (Rhesus monkey, mouse, dog, and horse) and Xenopus but not conserved in lizard (S6) or platypus (G6). The L68 residue is also conserved in mammals (Rhesus monkey, mouse, dog, and horse) but absent from Xenopus, lizard, and platypus. The SIFT analysis software (19) predicts that L68M is tolerated but R6C is not. PolyPhen (20) predicts that the L68M variant is benign. While leucine and methionine are both small hydrophobic amino acids, the R6C substitution creates an additional cysteine residue and the likelihood of disrupted signal peptide localization and/or function.

In addition to these nonsynonymous mutations, we also detected the common intronic single nucleotide polymorphisms rs689 and rs3842752, a synonymous A12A variant (rs3842744) in 1 of 759 MODY cases, and a synonymous P21P variant (rs11564720) in 1 of 759 MODY cases and 1 of 222 Caucasian controls.

We report the largest series of INS gene mutations identified by screening 1,044 patients diagnosed with diabetes from birth to 45 years of age. Heterozygous missense mutations were found in 37 probands of whom 35 were diagnosed during infancy (the first 12 months of life) and 33 had permanent neonatal diabetes diagnosed before 6 months of age. INS mutations were common in patients diagnosed with diabetes in the first 6 months of life (23% of cases tested) but far less frequent in patients diagnosed in childhood or early adulthood (two novel mutations, 0.3% of cases). We have also excluded INS mutations as a common cause of congenital hyperinsulinism.

The majority of patients with an INS gene mutation are sporadic cases that result from de novo mutations. Dominant inheritance was observed in 7 of 35 (20%) families. A similar proportion of KCNJ11 mutations are spontaneous (21), hence PND is primarily a sporadic form of diabetes.

We identified 16 different INS mutations in the 35 patients with permanent diabetes diagnosed in infancy (the first 12 months of life). Ten of these mutations have previously been reported (10), including the three most common mutations, A24D, F48C, and R89C, that account for 16 of 35 (46%) of cases. Most mutations are located within the A or B chains of insulin, and nine are predicted to interfere with the formation of disulphide bridges between cysteine residues either by replacing a cysteine residue or introducing an additional cysteine residue. In vitro experimental studies like those described by Liu et al. (22) are required to explore the effect of such mutations upon insulin processing and secretion.

Six novel missense mutations (H29D, L35P, G84R, C96S, S101C, and Y103C) were identified in seven families. Three of these mutations involve cysteine residues within the insulin A chain and are therefore predicted to affect the normal folding of the proinsulin molecule. S101C was identified in two families: in ISPAD180 the finding that both children are heterozygous for the mutation but neither parent is a carrier suggests that the mutation has arisen de novo in one parent who must be a germline mosaic (Fig. 2). In the second family, the mutation was not present in the unaffected father, but DNA was not available from the mother.

The novel C96S mutation occurs at the site of a previously reported mutation (C96Y) identified in a patient with PND (10) and in the Akita mouse model (11). The same amino acid substitution (Cys>Ser) is present at the adjacent residue (C95S) in the Munich mouse model (23). Studies of these mice indicate that mutant proinsulin is trapped and accumulated in the ER, leading to induction of the ER stress response, inhibition of protein synthesis, and ultimately β-cell death (11,24–26).

The other three novel mutations, H29D (residue B5), L35P (B11), and G84R, do not involve cysteine residues. The H29D mutation had arisen de novo and affects a residue containing an imidazole ring, which packs closely to the C96-C31 inter-disulfide bridge between the A and B chains (27). This mutation has been shown to disrupt the folding and conformation of proinsulin in vitro, decreasing the thermodynamic stability of the protein in the ER (28).

The L35P mutation affects a residue located in the B chain of the insulin molecule. The B chain contains two β turns at C31-H34 and G44-G47. These turns are highly conserved and guide the folding of the protein to form the inter-disulfide bond with the A chain (29). Residue L35 is highly conserved across all insulins and IGFs and is located adjacent to β turn 1. Due to the importance of this region in forming the C96-C31 interchain disulfide bond, we predict that a substitution of this leucine by proline would introduce prohibited steric contacts with C96, resulting in impaired disulfide bond formation.

The pathogenicity of the mutation G84R is less certain. Neither parent is affected with diabetes, but DNA samples were not available for testing. G84 is located within the C-peptide, a 31–amino acid linker between the A and B chains of proinsulin that is not highly conserved throughout evolution. The G84 residue is conserved from human to Xenopus, but an arginine (R) residue is present at this position in platypus. The patient with the G84R mutation is Vietnamese, and the possibility that this is a rare Asian polymorphism of no clinical significance cannot be excluded. In vitro studies have shown that the proinsulin molecule requires at least five linker amino acids (acting as the C-peptide) to produce the correct orientation of the molecule for the critical disulfide bond formation between the A and B chains (30), but it is not known which are the crucial amino acids within the C-peptide. Functional studies may help to determine the significance of this variant on insulin biosynthesis in vivo.

Two novel mutations were identified within the cohort of patients with MODY (R6C) and young-onset type 2 diabetes (L68M). R6C affects a residue within the signal peptide, while L68 is located within the C-peptide, which shows a lower level of evolutionary conservation. The sum of the genetic, evolutionary, and structural evidence suggests that while R6C is probably a pathogenic MODY mutation, L68M is probably a rare nonfunctional variant. Functional studies will be required to investigate these mutations further.

All mutation carriers identified through the infancy-onset cohort presented with symptomatic hyperglycemia or diabetic ketoacidosis and were treated with insulin. The majority (85%) of patients with an INS mutation were diagnosed before 6 months of age, and 96% had diabetes at 1 year, but two family members were diagnosed at 7 and 22 years of age (Fig. 3). In contrast, none of the three family members with the R6C mutation or the proband heterozygous for L68M was initially treated with insulin, and their age at diagnosis ranged from 15 to 65 years.

Our initial study showed that the age at onset of diabetes can vary within families (10), and in this larger series it has become clear that variable penetrance is common not only within families but also between families with the same mutation. For example, the four affected individuals from family ISPAD166 with the G32R mutation were diagnosed between 5 months and 22 years of age, whereas the proband from another family was diagnosed at 4 weeks of age. The age at diagnosis for A24D mutations carriers ranged from 4 weeks to 7 years. These findings suggest that other genetic or environmental factors may affect penetrance.

Both the mouse models show a more severe phenotype in male than in female mice, with more severe diabetes symptoms in the male Akita mice (11) and lower birth weight in the Munich mice (23). While we saw no difference in the age at diagnosis of diabetes, the insulin dose, or glycemic control, and birth weight was lower for male than female INS mutation carriers, suggesting a more severe insulin secretory defect in utero. The reasons underlying this sex-specific phenotypic variability are not known, either in man or mouse.

No difference in birth weight was observed between patients with INS mutations and those with activating KCNJ11 or ABCC8 mutations. Although patients with INS gene mutations are diagnosed later (median 11 vs. 8 weeks, P < 0.01), the ranges overlap and hence patients diagnosed within the first 6 months of life with permanent diabetes require molecular genetic testing to confirm the genetic subtype. We suggest that newly diagnosed patients with neonatal diabetes (aged <6 months) should be referred for testing of the most common forms of transient neonatal diabetes and PND to determine the genetic etiology. This would include analysis for chromosome 6q24 abnormalities as well as KCNJ11 and INS mutations, followed by ABCC8 if these tests are negative.

INS mutations accounted for 23% (33 of 141) of neonatal diabetes diagnosed before 6 months of life, when patients with known etiological mutations were excluded. These patients were selected from the Exeter collection of permanent diabetes diagnosed before 6 months of age. In this series, mutations in the KCNJ11, ABCC8, and INS genes are the cause in 31, 10, and 12% of cases, respectively, while mutations in GCK, EIF2AK3, or FOXP3 genes account for a further 8% of cases. Thus, mutations in known genes account for ∼60% of all cases, indicating that additional genes for this form of diabetes remain to be identified.

Both KCNJ11 and INS mutations are a rare cause of permanent diabetes diagnosed between 6 and 12 months of age, but INS gene mutations may be more frequent in this group (five INS mutation carriers identified vs. three KCNJ11 mutations in our patient cohort).

In summary, we report the largest series of INS mutation carriers. These mutations are the second most common cause of PND but a rare cause of diabetes in childhood or adulthood, including MODY. We recommend that the INS gene should be screened for mutations in all children diagnosed with diabetes in the first year of life.

Since submission and acceptance of this article, the L68M variant indentified in the proband of family DUK2193 has been shown to be inherited from her unaffected mother. This result suggests that L68M is unlikely to be the cause of early-onset diabetes in this family.

The following investigators also participated in this study: S. Amemiya, Faculty of Medicine, Pediatrics, Saitama Medical University, Saitama, Japan; K. Azad, Pediatric Diabetic Team BIRDEM, Dhaka, Bangladesh; L. Barak, 1st Dept. Pediatrics, Comenius University School of Medicine, Bratislava, Slovakia; T. Barrett, Institute of Child Health, University of Birmingham, Birmingham, U.K.; C. Costigan, Our Lady's Hospital for Sick Children, Dublin, Ireland; D. Darko, Jeffrey Kelson Centre, Central Middlesex Hospital, London, U.K.; S. Diamantopoulos, Miller School of Medicine, University of Miami, Miami, Florida; D. Doyle, DuPont Hospital for Children, Wilmington, Delaware; M. Densriwiwat, Faculty of Medicine, Siriraj Hospital, Mahidol University, Bangkok, Thailand; R.P.F. Dullaart, Department of Endocrinology, University Medical Center Groningen, Groningen, The Netherlands; I. Dzivite, Centre for Child and Adolescent Endocrinology, University Children's Hospital, Riga, Latvia; J.A. Edge, Department of Pediatric Endocrinology and Diabetes, Oxford Children's Hospital, Oxford, U.K.; K. Ekstrom, Karolinska University Hospital Solna, Stockholm, Sweden; G. Forsander, Division of Diabetes, The Queen Silvia Children's Hospital, Sahlgrenska/Ostra University Hospital, Göteborg, Sweden; D. Gasperikova, Institute of Experimental Endocrinology & DIABGENE, Slovak Academy of Sciences, Bratislava, Slovakia; V. Hakeem, Barnet General Hospital, London, U.K.; J.P. Hamilton-Shield, Pediatric Endocrinology, Bristol Children's Hospital, Bristol, U.K.; M.L. Hofstra, Klaver 1, Eelde, The Netherlands; S.-A. Ivarsson, Dept. of Pediatrics, Malmo University Hospital, University of Lund, Malmo, Sweden; I. Klimes, Institute of Experimental Endocrinology & DIABGENE, Slovak Academy of Sciences, Bratislava, Slovakia; M. Kocova, Department Endocrinology and Genetics, Skopje, Republic of Macedonia; O. Kordonouri, Diabetes Centre for Children and Adolescents, Children's Hospital at the Bult, Hannover, Germany; A.R.A. Lafferty, Pediatrics and Child Health, ANU Medical School, The Canberra Hospital, Canberra ACT, Australia; S. Likitmaskul, Faculty of Medicine Siriraj Hospital, Mahidol University, Bangkok, Thailand; L. Liu, Guangzhou Children's Hospital, Guangzhou, China; A. Mayo, Royal Aberdeen Children's Hospital, Scotland, U.K.; T. Milenkovic, Mother and Child Healthcare Institute of Serbia “Dr. Vukan Cupic,” Belgrade, Serbia; W. Mlynarski, Department of Pediatrics, Medical University of Lodz, Poland; F. Mohsin, Pediatric Diabetic Team BIRDEM, Dhaka, Bangladesh; A. Noczynska, Department of Endocrinology for Children and Adolescents, Medical University, Wroclaw, Poland; J. Odrezin, Greenvale Pediatrics, Birmingham, Alabama; J. Porter, Dept. Endocrinology and Diabetes, Birmingham Children's Hospital, Birmingham, U.K.; A. Roeleveld, Department of Pediatrics, Catharina Hospital, Eindhoven, The Netherlands; J. Sanchez, Miller School of Medicine, University of Miami, Miami, Florida; M. Schebek, Kinderkrankenhaus Park Schönfeld, Kassel, Germany; A. Schumacher, Kinderklinik Klinik Am Eichert, Göppingen, Germany; D. Segal, Centre for Diabetes and Endocrinology, Johannesburg, South Africa; J. Stanik, 1st Dept. Pediatrics, Comenius University School of Medicine and Institute of Experimental Endocrinology & DIABGENE, Slovak Academy of Sciences, Bratislava, Slovakia; Y. Tomita, Tokai University Hachioji Hospital, Tokyo, Japan; S. Wentworth, Danville, Indiana.

This work was funded by the European Union (Integrated Project EURODIA LSHM-CT-2006-518153 in the Framework Programme 6 of the European Community), the Wellcome Trust, and the Sir Graham Wilkins studentship (to S.E.F.). A.T.H. is a Wellcome Trust Research Leave Fellow. Funding for the work carried out in Chicago was provided by U.S. Public Health Service Grants DK-13914, DK-20595, DK-44752, DK-73541, and DK-77489 and a gift from the Kovler Family Foundation.

We thank Jon Locke, Piers Fulton, Noaman Hasan, and Mindy Kendrick for technical assistance. Kathleen Gillespie, Paul Lambert, Neil Raymond, Peter Swift, Polly Bingley, and Edwin Gale collected the DNA samples and clinical data for the Bart's Oxford (BOX) study and the British Diabetes Association 1972–1981 cohort, funded by Diabetes U.K. and the Wellcome Trust. Mark Strachan and Jill Little assisted with recruitment of the MODY cohort.