Wolcott-Rallison Syndrome Is the Most Common Genetic Cause of Permanent Neonatal Diabetes in Consanguineous Families

Abstract

Context and Objective: Mutations in EIF2AK3 cause Wolcott-Rallison syndrome (WRS), a rare recessive disorder characterized by early-onset diabetes, skeletal abnormalities, and liver dysfunction. Although early diagnosis is important for clinical management, genetic testing is generally performed after the full clinical picture develops. We aimed to identify patients with WRS before any other abnormalities apart from diabetes are present and study the overall frequency of WRS among patients with permanent neonatal diabetes.

Research Design and Methods: The coding regions of EIF2AK3 were sequenced in 34 probands with infancy-onset diabetes with a clinical phenotype suggestive of WRS (n = 28) or homozygosity at the WRS locus (n = 6).

Results: Twenty-five probands (73.5%) were homozygous or compound heterozygous for mutations in EIF2AK3. Twenty of the 26 mutations identified were novel. Whereas a diagnosis of WRS was suspected before genetic testing in 22 probands, three patients with apparently isolated diabetes were diagnosed after identifying a large homozygous region encompassing EIF2AK3. In contrast to nonconsanguineous pedigrees, mutations in EIF2AK3 are the most common known genetic cause of diabetes among patients born to consanguineous parents (24 vs. < 2%). Age at diabetes onset and birth weight might be used to prioritize genetic testing in the latter group.

Conclusions: WRS is the most common cause of permanent neonatal diabetes mellitus in consanguineous pedigrees. In addition to testing patients with a definite clinical diagnosis, EIF2AK3 should be tested in patients with isolated neonatal diabetes diagnosed after 3 wk of age from known consanguineous families, isolated populations, or countries in which inbreeding is frequent.

Wolcott-Rallison syndrome is the most common genetic subtype of permanent neonatal diabetes in consanguineous pedigrees. Homozygosity mapping has proved effective to identify candidates for genetic testing before the full clinical diagnosis is evident.

Wolcott-Rallison syndrome (WRS; online inheritance in man no. 226980) is a rare autosomal recessive multisystemic disorder due to biallelic mutations in EIF2AK3, the gene encoding the eukaryotic translation initiation factor-2α kinase 3 (1). This transmembrane enzyme, also known as PKR-like endoplasmic reticulum kinase or Pancreatic eIF-2alpha kinase localizes exclusively in the endoplasmic reticulum, and it is activated by the accumulation of unfolded proteins in the endoplasmic reticulum lumen during stress, resulting in phosphorylation of the α-subunit of the eukaryotic initiation factor 2 at residue Ser51 and down-regulation of protein synthesis (2). Lack of PERK activity leads to cell death by apoptosis in a number of different tissues, as clearly shown in two independent knockout mice (3,4). The high level expression of EIF2AK3 in both β-cells and bone tissue (4) explains the development of early-onset diabetes mellitus and skeletal abnormalities in virtually all patients with WRS, and hence, their association is considered a cornerstone for diagnosing this rare clinical entity (5,6,7). However, the gene is expressed at lower levels in several other tissues, which determines a number of other inconsistently present features (8,9). Importantly, most patients develop chronic or intermittent acute liver dysfunction, sometimes leading to hepatic failure. Genetic testing for this condition is usually delayed until the full clinical picture is evident and a clinical diagnosis of WRS made (9).

Permanent neonatal diabetes mellitus (PNDM) diagnosed within the first 6 months of life is a rare disorder likely to be monogenic rather than autoimmune (10,11). It is estimated to affect approximately one in 215,000 live births (12). Most patients have heterozygous activating mutations in the KCNJ11 and ABCC8 genes encoding potassium ATP channel subunits Kir6.2 (13) and SUR1 (14,15) or heterozygous mutations in the preproinsulin (INS) gene (16,17). In addition, a number of rare recessive causes of PNDM have been described, including INS (18), ABCC8 (19), GCK (20), IPF1 (21), PTF1A (22), GLIS3 (23), SLC2A2 (24), and SLC19A2 (25).

Homozygosity mapping, performed by total genome scan with polymorphic markers in individuals whose parents are related, has been extensively used to identify the gene responsible for a recessive disorder. It assumes that a homozygous mutation in a recessive disease gene segregates twice to the affected child from a common ancestor through both the maternal and the paternal lines and is hence identical by descent. The homozygous mutation will be embedded in a chromosomal segment that has not undergone recombination. Therefore single-nucleotide polymorphism (SNP) markers in that segment will also be homozygous by descent (26). This short segment of homozygosity by descent can be easily detected and will harbor the disease gene. Taking into account that WRS is a recessive disorder in which diabetes is generally the first manifestation to occur, we hypothesized that homozygosity mapping could be a powerful genetic tool to identify candidates for EIF2AK3 sequencing among infants with isolated PNDM and reported parental consanguinity.

We report the successful use of homozygosity mapping for early molecular diagnosis of WRS. In addition, we describe the clinical and genetic findings in the largest international cohort of WRS cases assembled to date.

Patients and Methods

This study was conducted in accordance with the Declaration of Helsinki and informed consent was obtained from all patients, with parental consent given on behalf of children.

Study population

All patients were referred to the Exeter Molecular Genetics Laboratory (United Kingdom) for testing. Clinical information was obtained using a standardized questionnaire from hospital records. Parental consanguinity was defined as parents being second cousins or closer. A total of 34 probands were tested for EIF2AK3 mutations. These included 28 probands with a suspected diagnosis of WRS on the basis of early-onset diabetes (within the first 15 months of age) and either skeletal dysplasia and/or unexplained liver dysfunction, and six consanguineous probands with isolated PNDM at referral in whom the most common genetic causes of diabetes had been previously excluded (including KCNJ11 and ABCC8 in all of them, INS in five, and GCK in four) and a large (9.22–67.64 Mb) region of homozygosity encompassing the EIF2AK3 gene on chromosome 2 had been identified (see below).

Homozygosity mapping

Genotyping was carried out on the Affymetrix human 10K Xba and 50K Hind mapping SNP chips by Medical Solutions Nottingham (formerly GeneService; Nottingham, UK) or the Affymetrix 5.0 mapping chip by ALMAC Diagnostics (Carigavon, Northern Ireland). Processing of genomic DNA was performed as per the Affymetrix protocol, and the mean SNP call rate was 98.7%. In-house Perl scripts were developed to automatically identify genomic homozygous segments for the 10K chip, defined by at least 20 consecutive homozygous SNPs marking a region that exceeded 3 cm (27). These thresholds were empirically extended to the larger chips identifying any region greater than 3 Mb delimited by consecutive homozygous SNP calls, allowing for a maximum of two heterozygous SNPs per 100 calls. All regions for each case were assigned a rank, in descending size order.

EIF2AK3 gene analysis

Genomic DNA was extracted from peripheral leukocytes using standard procedures. The coding exons and the intron-exon boundaries of the EIF2AK3 gene were PCR amplified; primers and conditions are available on request. Sequence specific primers for each amplicon were tagged with 5′ M13 tails to allow sequencing to be performed with a universal M13 primer. Single-strand sequencing was carried out using standard methods on an ABI 3730 (Applied Biosystems, Warrington, UK). Sequences were compared with the published template (accession no. AF110146.1) using Mutation Surveyor (version 3.20; SoftGenetics, State College, PA). Any changes in the sequence were checked against published polymorphisms and mutations and for conservation across species. Sequence variants were tested for their presence in family members whenever a DNA sample was available.

Further molecular testing in probands with EIF2AK3 mutations

In case of proband 3021-1, we used a panel of microsatellites for chromosome 20 (D20S482, D20S851, D20S477, D20S107, D20S481, D20S171) to confirm family relationships. An alternate set of exon 11 primers were also designed to amplify across the original set of exon 11 specific primers, to exclude allelic drop out. We then designed a multiplex ligation-dependent probe amplification assay to quantify the number of copies of EIF2AK3. We used synthetic oligonucleotide probes for EIF2AK3 exons 10–12 with HNF1A and HNF4A as control probes [method previously described by Ellard et al. (28)]. To investigate uniparental isodisomy, a panel of microsatellite markers flanking EIF2AK3 on chromosome 2p11.2-q11.2 (D2S2368, D2S139, D2S2333, D2S388, D2S2216, D2S2181, D2S2154, D2S113, and D2S2264) was used. The same set of microsatellite markers was also used to explore relatedness between probands 3377-1 and 3750-1.

Other genetic testing in patients from consanguineous pedigrees

The common genetic causes of PNDM were tested in 591 patients diagnosed with diabetes within the first 6 months after birth. KCNJ11 (NM_000525), ABCC8 (NM_000352.2), and INS (NM_000207) were screened in all of the patients. GCK (NM_000162.2) testing was limited to those patients born to consanguineous parents in whom a homozygous region encompassing the gene on chromosome 7 had been identified using the mapping chips (see above). Genetic analysis was performed as described above for EIF2AK3.

Statistical analysis

Clinical numeric data are given as median and interquartile range (IQR). sd scores (SDS) for birth weights were calculated by comparing to the Child Growth Foundation LMS data (29). The clinical features of the patients were analyzed using Kruskal-Wallis and Mann Whitney U tests, and Spearman correlation coefficient in the statistical package SPSS version 15.0 (Chicago, IL). A χ2 test was used to compare the frequencies of the different genetic subtypes of PNDM in consanguineous and nonconsanguineous pedigrees.

Results

Molecular genetic findings

We identified 26 different EIF2AK3 mutations in 25 probands (Table 1). Figure 1 displays the pedigrees of the five families with more than one affected individual. Twenty-three probands had a homozygous mutation and two were heterozygous for two different mutations. Six mutations had previously been described and 20 mutations were novel, including nonsense (n = 8), frameshift (n = 7), missense (n = 4), and splicing (n = 1) mutations. Twenty-three probands had private mutations. The same homozygous mutation (R587X) was present in two probands from Turkey, and the possibility of a founder effect could not be excluded by a combination of intragenic SNP and microsatellite analysis.

Table 1.

Mutation data

Proband ID	Reported parental consanguinity	Country of origin	EIF2AK3 mutation (nucleotide)	EIF2AK3 mutation (protein)	Mutation reported previously
1305-1	Yes	Turkey	c.2153delG	S718TfsX6	No
1554-1a	No	Slovakia	c.2694G>T, c.3170T>C	W898C/L1057P	Ref. 9
1945-1	No	UK	c.1894C>T, c.505delA	R632W/S169AfsX31	No
2190-1b	Yes	Qatar	c.1567_1570del	E523X	Ref. 8
2216-1	Yes	UAE	c.1290G>A	W430X	No
2556-1	Yes	Morocco	c.2953G>A	G985R	No
3377-1	No	Turkey	c.1759C>T	R587X	No
3395-1	Yes	UAE	c.2867G>A	G956E	Ref. 31
3576-1	Yes	Saudi Arabia	c.1259delA	N420TfsX14	No
3822-1	Yes	Saudi Arabia	c.3190C>T	R1064X	No
3926-1	Yes	Morocco	c.449delA	K150RfsX2	No
4064-1	No	Saudi Arabia	c.475delG	A159PfsX41	No
4191-1	No	USA	c.1544_1559del	V515GfsX5	No
1229-1	Yes	Pakistan	c.935C>A	S312X	No
1537-1	Yes	Turkey	c.2981+1dupG	IVS14+1dupG	No
2660-1	No	Germany	c.2704C>T	R902X	No
2930-1	Yes	UAE	c.1949T>C	I650T	No
3021-1	No	UK	c.1774T>C	F592L	No
3090-1	Yes	India	c.1274T>A	L425X	No
3169-1	Yes	Libya	c.1032dupT	K345X	Ref. 1
3172-1	Yes	Saudi Arabia	c.1044_1057del	V349SfsX3	No
3750-1	No	Turkey	c.1759C>T	R587X	No
2045-1	Yes	Saudi Arabia	c.1406C>G	S469X	No
3547-1	Yes	India	c.2304_2305del	C768X	No
3705-1	Yes	Turkey	c.1562G>A	W521X	Ref. 9

UK, United Kingdom; UAE, United Arab Emirates. 

^aPatient previously reported by Senée et al. (9). 

^bPatient previously reported by Engelmann et al. (30). 

Figure 1.

Pedigrees for families with more than one affected individual. Solid black filled shapes represent patients with WRS. An arrow indicates the proband. The genotype is shown underneath each symbol. M, Mutant allele; N, normal allele; ID, family identification number.

The predicted effect of the different mutations on the EIF2AK3 protein is shown in Fig. 2. The nonsense and frameshift mutations were distributed throughout the gene. They are predicted to result in truncated proteins missing all or part of the catalytic domain and hence expected to lead to a complete loss of function. Pathogenicity of the four novel missense variants (F592L, R632W, I650T, and G985R) was suggested by: 1) homozygosity or compound heterozygosity for the variant in affected individuals within families; 2) no unaffected family members homozygous or compound heterozygotes for the variant(s); 3) location in one of the two serine/threonine protein kinase domains; and 4) conservation of the amino acid residues involved across species (including chimpanzee, rat, mouse, dog, chicken, Tetraodon, fruitfly, and Caenorhabditis elegans). One patient was homozygous for an intronic variant that changes the splice donor site of exon 14 from GTGAG to GGTGAG and is likely to be pathogenic, either by a direct effect on splicing leading to exon skipping or retention or by incorporating an extra base in exon 14 that would result in a frameshift mutation.

Figure 2.

Summary of the effect of all the mutations identified to date on the EIF2AK3 protein (modified from reference9). The structure of the EIF2AK3 protein is illustrated in the upper part of the chart, with the regulatory (dotted bar) and the two conserved serine/threonine protein kinase subdomains (squared bars) in the catalytic domain indicated. Novel mutations identified in this study are depicted in italics. Missense mutations are grouped directly under the region involved. The bars in the lower part of the chart indicate the extent of the EIF2AK3 mutant proteins, with the black portion representing abnormal amino acid sequence secondary to frameshift mutations. Another three other EIF2AK3 mutations have been identified, including two splicing mutations (c.2981 + 1G>A and c.2981 + 1dupG, the latter in the present study) and a 184-bp deletion in exon15/intron 15.

DNA was available from the parents of 16 probands. All unaffected parents were heterozygous carriers except for the father of 3021-1. The unaffected mother was heterozygous for the F592L mutation. Microsatellite marker analysis confirmed family relationships (results not shown). To determine the genetic mechanism of disease, we excluded allelic drop-out (due to a SNP under the original primers) by resequencing exon 11 with an alternate set of specific primers. A paternally inherited heterozygous deletion was also excluded by multiplex ligation-dependent probe amplification (data not shown). Further analysis using microsatellites flanking EIF2AK3 (chr2p13.3-2q11.2) showed segmental maternal uniparental isodisomy for a minimal approximately 446-kb region encompassing the EIF2AK3 gene (Fig. 3).

Figure 3.

Microsatellite markers analysis in family 3021. The solid black filled circle represents the proband with WRS. The genotype is shown underneath each symbol. M, Mutant allele; N, normal allele. The table shows genotype analyses of family members using nine microsatellite markers around EIF2AK3 on chromosome 2; the disease-relevant mutation is shown in italics. Apparent homozygosity for the informative marker D2S2216 in the proband indicated maternal segmental isodisomy for chromosome 2 (bold).

No mutations in EIF2AK3 were identified in six patients with a suggestive phenotype (three with early onset diabetes and skeletal dysplasia and three with diabetes and liver dysfunction; none of them from consanguineous descent) and in another three consanguineous probands with isolated PNDM.

Clinical features

Twenty-two of the 25 probands with biallelic mutations in EIF2AK3 were selected for genetic testing because of a clinical phenotype suggesting WRS. The remaining three were tested after the identification of a large homozygous region in chromosome 2 encompassing the EIF2AK3 gene (Table 2 and supplemental Table S1, published as supplemental data on The Endocrine Society’s Journals Online web site at http://jcem.endojournals.org). At the time of this study, three probands had died aged between 7 and 14 months (a further affected relative was deceased at 4 yr of age due to an unrelated condition; supplemental Table S1). The median age of the remaining ones was 4.95 yr (range 1.2–32.3). The phenotype of the parents and heterozygous siblings was unremarkable.

Table 2.

Clinical features at time of referral for patients with a proven genetic diagnosis

	n = 29
Age at diagnosis of diabetes (wk)	10.5 (6.0–20.3)
Age at referral for testing (yr)	3.3 (1.3–7.6)
Current age (yr, n = 25)	4.8 (2.8–7.8)
Age at death (yr, n = 4)	0.6–4
Males (%)	48.3
Birth weight (g)	2805 (2515–3173)
Birth weight (SDS)	−1.39 (−2.08 to −0.59)
Skeletal dysplasia at referral (%)	48.3
Liver dysfunction at referral (%)	72.4

Data are generally given as median (IQR). Age at death is given as full range. 

All probands first presented with permanent diabetes mellitus, at a median age of 10.5 wk (IQR 6.0–19.3). The earliest presentation was at 3 wk, and all patients but one were diagnosed within or slightly after the first 6 months of life. The only patient diagnosed after 1 yr of age (14 months) is homozygous for a missense I650T mutation. There are no differences in age at diabetes onset among the different types of mutation (P = 0.46). Although patients with WRS have a reduced birth weight (median −1.4 SDS, IQR −2.1 to −0.6), intrauterine growth retardation defined as a birth weight at or below −2 SDS was present only in seven of 20 patients for whom that information was available. There was no significant correlation between birth weight and age at diagnosis of diabetes (ρ = 0.33, P = 0.16). All patients are currently on full-replacement insulin doses, suggesting endogenous insulin secretion is negligible.

In addition to diabetes, skeletal abnormalities (mostly spondyloepiphyseal dysplasia) were initially reported by the referring clinician in probands from 13 families making the clinical diagnosis of WRS likely. Probands with skeletal abnormalities at referral did not differ from patients without in terms of either birth weight (P = 0.73) or age at diabetes onset (P = 0.87). However, they were significantly older at molecular testing [median age 6.6 yr (IQR 4.0–10.8) vs. 1.4 yr (IQR 0.9–1.8), P = 0.02]. Nine further probands were tested because of the association of early-onset diabetes and liver dysfunction ranging from mild hypertransaminasemia to acute liver failure, requiring a transplant. Although none of them presented clinically evident skeletal abnormalities at referral, they have become evident after molecular diagnosis in at least two cases. Major hepatic dysfunction was often accompanied by acute multiorgan failure (encephalopathy, renal failure, bone marrow failure) and sometimes resulted in the patient’s death. Hepatic and renal functions usually returned to normal in patients who survived. Residual cerebellar signs (ataxia, dysarthria) were found in two patients after recovery from an acute encephalopathic episode. The oldest patient in the series, currently aged 32 yr, is the only one having chronic renal failure and erythropoietin-dependent anemia; the remaining patients are currently 15 yr old or younger. Pancreatic exocrine insufficiency requiring supplemental enzymes was present in two patients, with severe pancreatic hypoplasia reported in one of them.

Three consanguineous patients with isolated PNDM received a molecular diagnosis between 0.8 and 1.6 yr after the identification of a large homozygous region encompassing EIF2AK3. One of them, currently aged 1.6 yr, has not presented any other features of WRS yet. Another one was found to have hepatomegaly and irregular fragmented epiphyses shortly after the molecular diagnosis was made at 1.2 yr. The third one has developed the full clinical picture of WRS by the age of 3 yr, including liver dysfunction identified at 1 yr and skeletal abnormalities at 2 yr. Interestingly, he was also diagnosed with primary hypothyroidism at 1.4 yr, but this may not be related to the EIF2AK3 mutation.

Known parental consanguinity was reported in 17 of the 25 families; affected individuals from all of them have a homozygous EIF2AK3 mutation. Two probands were compound heterozygotes for two different mutations, and a third was homozygous for an EIF2AK3 mutation as a result of segmental uniparental isodisomy of chromosome 2. The remaining five probands were homozygous for an EIF2AK3 mutation. Three originate from countries with high rates of consanguinity (Turkey and Saudi Arabia) (32), and the other two are from relatively isolated populations (Kosovo and South Dakota).

Prevalence of WRS among patients with PNDM

Mutations in EIF2AK3 account for 15 of 63 (23.8%) consanguineous probands with PNDM tested in the Exeter laboratory. WRS is the most common known genetic cause of PNDM in consanguineous pedigrees, followed by recessive mutations in INS (12.7%), GCK (11.1%), and ABCC8 (6.3%). This is in contrast to nonconsanguineous families, in which WRS accounts for only eight of 583 cases (1.4%, P = 7 × 10⁻²⁰). Heterozygous mutations in KCNJ11, ABCC8, and INS, the most common genetic causes of PNDM in nonconsanguineous pedigrees (36.9%), are responsible for only 4.8% of PNDM cases in consanguineous families (P = 3 × 10⁻⁷).

We compared the age at diagnosis of diabetes and birth weight of the 29 patients with EIF2AK3 mutations with the three other most frequent forms of recessive PNDM from the Exeter cohort, including 18 patients with ABCC8 mutations, 14 with INS mutations, and eight with GCK mutations (Fig. 4). There was a strong association of the genotype with both age at diagnosis of diabetes (P = 0.000005) and birth weight adjusted by gestational age (P = 0.000001). Patients with EIF2AK3 mutations were not different from patients with recessive ABCC8 mutations either in terms of age at diabetes onset or birth weight. However, they showed less severe intrauterine growth retardation and were diagnosed later than patients with recessive INS or GCK mutations.

Figure 4.

Age at diabetes onset (left panel) and adjusted birth weight (right panel) in the four most common recessive genetic causes of PNDM. There is no significant difference in age at presentation between EIF2AK3 and ABCC8 (P = 0.60). However, patients with EIF2AK3 mutations are diagnosed with diabetes later than those with INS or GCK mutations (P = 0.001 and P = 0.009, respectively). Similarly, birth weight in EIF2AK3 mutation carriers was slightly reduced and comparable with that of patients with a mutation in ABCC8 (P = 0.63) but significantly higher than birth weight of patients with recessive INS and GCK mutations (P = 0.001 and P = 0.001, respectively).

Discussion

We report the largest series of WRS assembled to date. Overall, a molecular diagnosis of WRS has been confirmed in 25 families. Twenty of the 26 mutations (77%) identified are novel. Most cases were selected for genetic testing after diagnosis of skeletal dysplasia and/or unexplained liver dysfunction in a patient with a previous diagnosis of neonatal or infancy-onset diabetes. However, a homozygosity mapping approach allowed the identification of three patients before any other abnormalities apart from diabetes became evident.

Genetic testing for mutations in EIF2AK3 is usually delayed until the full clinical picture of WRS is evident (9). Because each intercurrent illness can potentially be complicated by acute liver and/or renal failure, this diagnostic delay might be responsible for the poor outcome of some patients who develop unexplained acute multiorgan failure during minor intercurrent illnesses and may explain that WRS syndrome may go unnoticed when the evolution is rapidly fatal before the skeletal involvement is evident (6,33). Early diagnosis of WRS is important because it allows the anticipation of potential complications during concomitant situations such as acute illness, trauma, or major surgery (34). For this purpose, radiological screening for epiphyseal abnormalities in any infant with diabetes has been recommended (34). However, this approach would lead to a high number of unnecessary x-ray surveys in infants with diabetes, among whom WRS is still a relatively rare condition. Moreover, in most spondyloephiphyseal or multiple epiphyseal dysplasias, the bone lesions are discovered only after the first year of life, and sometimes even later (6), which argues against the potential benefit of a radiological screening method for WRS. Homozygosity mapping has been previously used for positional cloning of unknown genes producing a recessive disorder, but this is the first time it has been used for early diagnosis of a recessive disease before the full clinical picture is present.

In contrast to nonsense and frameshift mutations spread throughout the gene, missense mutations have been identified only within or near each of the two Ser/Thr protein kinase domains of the catalytic domain. This underscores the important functional role of these domains. Missense variants in other parts of the gene might be either very well tolerated or contribute to a less severe phenotype with later-onset diabetes or milder skeletal abnormalities. In keeping with this possibility, Senée et al. (9) reported a patient with a homozygous N655K mutation in EIF2AK3 who presented with diabetes at a relatively late age (2.5 yr). Functional studies showed that the mutation is hypomorphic so that the mutated protein still had some residual kinase activity. We identified a proband homozygous for an I650T mutation who also had a late onset of diabetes at 14 months. However, this patient developed acute liver failure requiring liver transplantation at 2 yr, so a later onset of diabetes clearly does not predict a milder phenotype of the disease. Furthermore, skeletal abnormalities are not evident in our patient with the F592L mutation, even though she is 32 yr old and has developed many of the acute complications of WRS. The severity of the different clinical manifestations of WRS is largely independent of the genotype.

The majority of probands (88%) were homozygous for EIF2AK3 mutations and were from known consanguineous pedigrees, isolated populations, or countries in which consanguinity is frequent (32). We report the first case of uniparental isodisomy for an EIF2AK3 mutation in a patient who inherited the mutation only from her mother. This emphasizes the importance of testing parents of patients with recessive disorders to offer an accurate genetic counseling as the risk of recurrence is almost negligible in this case. However, lack of reported consanguinity should not be used as an exclusion criterion for EIF2AK3 testing in the presence of a typical clinical presentation as genetic mechanisms other than homozygosity may be encountered in patients with WRS.

We have shown that birth weight and age at diabetes onset can discriminate between EIF2AK3 and other genetic subtypes of neonatal diabetes in consanguineous probands. Whereas patients with WRS usually have a normal or mildly reduced birth weight and are diagnosed with diabetes after the first 3 wk of life, patients born with severe intrauterine growth retardation (birth weight below −2 SDS for gestational age) or diagnosed with diabetes within the first 3 wk of life are more likely to have biallelic INS or GCK mutations.

In summary, EIF2AK3 mutations are the most common cause of PNDM in consanguineous pedigrees. Besides testing patients with a definite clinical diagnosis of WRS, we recommend that analysis of the EIF2AK3 gene is considered in patients with isolated neonatal diabetes diagnosed after 3 wk of age from known consanguineous pedigrees, isolated populations, or countries in which inbreeding is frequent.