Heterozygous Missense Mutations in the Insulin Gene Are Linked to Permanent Diabetes Appearing in the Neonatal Period or in Early Infancy

A Report From the French ND (Neonatal Diabetes) Study Group

• ISPAD, International Society for Paediatric and Adolescent Diabetes
• K_ATP channel, ATP-sensitive K⁺ channel
• PND, permanent neonatal diabetes
• TND, transient neonatal diabetes

Diabetes in childhood, including so-called “neonatal diabetes,” is unrelated to autoimmune destruction of the pancreatic β-cells in the vast majority of cases (1). Neonatal diabetes is a rare (∼1:300,000 newborns) but potentially devastating metabolic disorder characterized by mild to severe hyperglycemia with low levels of circulating insulin within the first months of life (1). The presentation of the disease can be transient (transient neonatal diabetes [TND]) or permanent (permanent neonatal diabetes [PND]), which differ in the duration of insulin dependence early in the first months/years of life and in the molecular mechanisms responsible for the severe insulin secretion defect. More than half of TND cases are associated with abnormalities of an imprinted region on chromosome 6q24 (2), whereas mutations in the two subunits (Kir6.2 and SUR1) of the ATP-sensitive K⁺ channel (K_ATP channel) of the pancreatic β-cell have been characterized as a common cause of both TND and PND (3–5). In addition, a few cases of PND were attributed to mutations in the genes encoding the glycolytic enzyme glucokinase (GCK) (6) and insulin promoter factor-1 (IPF-1) (7), and syndromic cases including neonatal diabetes are caused by rare mutations in PTF1A, FOXP3, GLIS3, TCF2, and EIF2AK3 (1).

Our previous studies from the French ND (neonatal diabetes) case series, and others from the International Society for Paediatric and Adolescent Diabetes (ISPAD) cohort, have demonstrated that dominant mutations in both subunits of the K_ATP channels expressed in the neuroendocrine cells lead to a range in the severity of the disease. These phenotype specificities depend on the mutation in Kir6.2 or SUR1: from mild transient hyperglycemia to PND for SUR1 mutations (4,8,9) and from PND to more severe phenotypes associated with developmental delay and epilepsy usually for Kir6.2 and for some SUR1 mutations (10,11). In the patients referred to the French ND Study Group, over 50 cases presenting with very-early-onset diabetes are not yet defined for their molecular origin (8). This suggests that other gene defects are involved in key mechanisms regulating insulin processing and secretion and/or the survival process of insulin-producing β-cells.

Stoy et al. (12), using a positional cloning approach in a pedigree including four diabetic members (age at diagnosis from 13 to 52 weeks), identified the insulin (INS) gene as responsible for diabetes in this family and replicated their findings in unrelated neonatal diabetic patients and in the ISPAD cohort, showing 10 heterozygous missense mutations in 16 probands.

In the present study, 38 PND patients from the French ND cohort and 1 child with nonautoimmune early-infancy diabetes, who are negative for a mutation in GCK, KCNJ11, and ABCC8 (4,8,10), were screened for the INS gene by direct sequencing. The H family (Table 1) was referred to us to sequence ABCC8, as mutations in this gene may be present in patients with diabetes from the neonatal period to young adulthood (4,8). As GCK (MODY2) and HNF1A (MODY3) sequences were also found to be normal in this family, the proband HA was included in the current study, despite diabetes appearing like nonautoimmune early-onset type 1 rather than bona fide neonatal diabetes.

The three exons of INS were amplified in two fragments using PCR (supplementary Table 1 [available in an online appendix at http://dx.doi.org/10.2337/db07-1358), and both strands were sequenced by using a standard protocol (4) and 3730xl DNA analyzer (Applied Biosystems, Foster City, CA). Three missense mutations—A24D (c.71 C>A in exon 2), R89C, and C96Y (c.265 C>T and c.287 G>A in exon 3)—were identified in four probands diagnosed with a permanent form of diabetes. A detailed description of these patients is shown in Table 1. The same three mutations were found in PND patients from the original report (12). We also investigated whether the identified mutations were inherited from a parent or arose de novo by sequencing a DNA sample of the parents when available. In two families (H and B), the R89C mutation was inherited in a dominant manner from the mother. The sister of proband HA who presented with early-onset diabetes was also found with the R89C mutation (Table 1). The A24D and C96Y mutations were not inherited from a parent in families F and G. Such family relationships were confirmed by genotyping a panel of six microsatellite markers as previously used (4). Eleven patients with TND from the French cohort, in which chromosome 6 anomalies had been excluded (1), were also screened for the INS gene, but no mutation was identified in these patients.

All three mutations are located in critical regions of the preproinsulin molecule and may affect the proteolytic processing of insulin precursors or disrupt insulin biosynthesis and induce endoplasmic reticulum stress, as proposed in the previous studies (12,13). The two amino acid changes at the signal peptide and A-chain C-peptide cleavage sites (A24D and R89C) are present in several unrelated neonatal diabetic patients (in three and four patients, respectively, when considering the current and previously reported cases [12]); this likely reflects the potential deleterious effect of these preproinsulin cleavage mutants on the folding and secretion of the normal insulin peptide. Importantly, the C96Y mutation found in two unrelated patients (this study and the previous report [12]) was also shown to act in a dominant manner in the Akita diabetic mouse and consequently to activate the unfolded protein response (13).

The clinical features of four proband children, of one relative child (sister of proband HA), and of the two affected mothers (except for data at birth and at presentation, which were not available for the two mothers) are shown in Table 1. Diabetes was diagnosed in the children at a median age of 8.5 months (range 25 days to 4.25 years). One child presented with mild hyperglycemia (7.2 mmol/l), whereas the four others had marked hyperglycemia (mean plasma glucose at diagnosis 27.3 mmol/l, range 7.2–49.5). One case (FZ-A24D) presented with severe ketoacidosis, one (HA-R89C) presented with hyperglycemia and ketoaciduria, and three (HL-R89C, BG-R89C, and GI-C96Y) were diagnosed based on polyuria and polydipsia. The age at diagnosis in family H was consistent with familial type 1 diabetes. However, autoantibodies associated with type 1 diabetes (islet cell antibody, GAD65, and insulinoma-associated protein 2 antigen) were all negative in the two siblings, as well as in the other three younger patients. Pancreas ultrasonography was normal when performed (in four of five of the children cases). Two of five of the young cases were small for gestational age (<3rd centile), including one of the offspring of a mother with diabetes (BG-R89C), suggesting a certain degree of insulin deficiency in utero. Initial insulin was required for those five patients and could never be stopped. At the last follow-up examination (age range 22 months to 9.8 years), insulin dose ranged from 0.42 to 0.70 units · kg⁻¹ · day⁻¹ (median 0.60) with good metabolic control in each of the patients. These are relatively low insulin requirements, especially for the children and mother in family H, compared with what is required in type 1 diabetic patients in the same time frame (usually ∼1 unit · kg⁻¹ · day⁻¹). Consistent with a partially preserved β-cell secretory function, C-peptide levels in those two children were detectable, and a normal response to a glucagon test was observed when they were 7.5 and 3.4 years of age, i.e., 3 and 1 years, respectively, after diabetes onset (Table 1). Interestingly, in patient HL, before the start of insulin therapy (at 4 years and 3 months), 2-h postprandial glucose level was moderately high (9 mmol/l) and insulinemia was high (126 mUI/l), with a C-peptide level of 4.4 ng/ml, which proves an endogenous insulin secretion in this affected child at the age of 3 years and 5 months, an unusual situation in type 1 autoimmune diabetes.

Three patients (among whom a brother and a sister) have inherited the R89C mutation from their diabetic mother. In family H, the mother (HB) was discovered to have diabetes at 3 years of age and has a good metabolic control at age 38 years with a low dose of insulin (∼0.35 units · kg⁻¹ · day⁻¹). In contrast, the mother in family B, who was discovered at 3 months of age to have diabetes, had a poor metabolic control over the years and developed severe retinopathy, neuropathy, and macroangiopathy; at present (aged 35 years), she had to undergo amputation of both feet.

Clinically, two main differences were found by comparing in our cohort the children with an INS mutation and those with early-onset diabetes due to K_ATP channel mutations: the disease starts later with a median age of 8.5 months compared with 57 and 33 days in the cases with a mutation of KCNJ11 or ABCC8, respectively (Table 2); no neuropsychological or neuromotor dysfunctions are present in children with an INS mutation in contrast to the children with a K_ATP channel mutation, where 6 of 17 patients with an ABCC8/SUR1 mutation and 4 of 18 patients with a KCNJ11/Kir6.2 mutation in our cohort do have such developmental anomalies. From our findings, diabetes caused by an INS mutation was found to be isolated in contrast to what has been reported in some patients previously described (12). Indeed, none of the patients from our cohort had acanthosis nigricans, a finding suggestive of insulin resistance, whereas four patients in the previous study have such a skin lesion.

In the current study, the three mutations identified were present in five children and two mothers with childhood-onset diabetes and not found in the transient form in accordance with the original report (12). In our cohort, the INS mutations account for 13% of neonatal diabetes cases where the genetic etiology was unknown and for ∼10% of all permanent neonatal diabetes cases, compared with 4% for ABCC8 and 35% for KCNJ11 mutations (4,10; A.D., H.C., M.V., unpublished data). A higher prevalence of PND due to INS mutations was reported in the ISPAD cohort (∼20%), which could be explained by more strict inclusion criteria in the French cohort, particularly age of diagnosis and presentation at diagnosis for the patients who were mostly referred as newborns. The exception of family H is of great importance, as children with an INS mutation can present clinically as type 1 diabetic patients. This is of special relevance, as the age of diagnosis of type 1 diabetes has decreased over the years with many children diagnosed before or around 4 years of age (14), and highlights the need of knowing the status for autoimmune markers in those children; if negative, then a monogenic form of diabetes should be searched for. Another biological feature that may help distinguish diabetes linked to INS mutations from autoimmune type 1 diabetes is a higher insulin level at presentation.

Highly variable penetrance and clinical presentation of diabetes, both within and between families, is indeed documented from our study cohort for the patients carrying an INS mutation. Furthermore, previous works have reported familial hyperproinsulinemia with normal to mild glucose intolerance in adults caused by INS mutations at residues R89P/H (15,16) and H34D (17). Cellular biology experiments are needed to further understand these contrasting disease phenotypes. A clinical variability in the age at onset of diabetes in early infancy was also reported in patients with known Wolcott-Rallison syndrome, another disease linked to endoplasmic reticulum stress (18). This severe condition is due to mutations in the EIF2AK3 gene that encodes the pancreatic eukaryotic initiation factor 2α kinase or PERK. Severe defects in fetal/neonatal β-cell proliferation and differentiation were observed in the PERK-deficient mice, which results in low β-cell mass, defects in proinsulin trafficking, and abrogation of insulin secretion and culminate in PND (19).

A better understanding of the mechanism(s) leading to β-cell dysfunction in the patients with diabetes caused by an INS mutation will be crucial to define new treatments. As in the Akita mouse model, it may relate to the general concept of protein toxicity (20). The protein toxicity and overload in the endoplasmic reticulum are molecular defects already known to be involved in congenital diabetes insipidus, where the disease begins after the neonatal period. Indeed, this delay represents the time to destroy vasopressin-secreting neurons (21). The INS mutations associated with permanent early-infancy diabetes extend this concept demonstrated in diabetes insipidus to diabetes. Strategies to increase insulin secretion by upregulating the normal allele of the INS gene may be applicable, as long-term residual β-cell function may exist in some patients with diabetes due to an INS mutation.

In conclusion, heterozygous INS gene mutations are a cause of isolated permanent early-onset diabetes and should be searched for in neonatal as well as in childhood diabetes appearing like type 1, when autoimmune markers are absent, and when there is a strong autosomal dominant pattern of inheritance.

• Received September 24, 2007.
• Accepted December 22, 2007.

• DIABETES