Loss of Kinase Activity in a Patient With Wolcott-Rallison Syndrome Caused By a Novel Mutation in the EIF2AK3 Gene

The girl came to our attention at 2 months of age, as the only child of nonconsanguineous parents. She presented with severe hypertonic dehydration and diabetic ketoacidosis with two prolonged convulsive episodes. A diagnosis of type 1 diabetes was made and insulin treatment initiated. There was no remission phase. The family history revealed that the mother suffers from a mild form of deafness and that a maternal uncle is affected by type 2 diabetes. After the age of 3 years, mild psychomotor development retardation was recognized. Physical examination showed growth retardation, microcephaly, limited shoulder mobility, and genua valga (Fig. 1A). The anti-islet cell autoantibodies (insulin autoantibody [IAA], insulinoma antigen 2 [IA2], and GAD were negative. Mitochondrial diabetes was excluded. Average glycosylated hemoglobin since diagnosis was elevated (8.7%, normal value <6%). Abdominal ultrasonography showed a hypoplastic pancreas, and skeletal X-ray examination showed typical signs of spondylo-epiphyseal dysplasia (Fig. 1B). Based on these findings, the diagnosis of WRS was made. At age 2.7 years, the girl experienced an episode of atypical prolonged hypoglycemia despite greatly reduced insulin therapy during an acute infection with ataxia and neuromuscular hypotonia. At age 5 years, under similar circumstances, she showed high fever, prolonged hypoglycemia, and highly elevated liver enzymes (aspartate aminotransferase 4,753 units/l, alanine aminotransferase 2,429 units/l, and γ-glutamyl transferase 170 units/l). Hepatitis A, B, and C and Epstein-Barr virus (EBV) were excluded. Spontaneous normalization of clinical and laboratory abnormalities occurred within a few days. So far, no clinical or biochemical evidence of exocrine pancreas insufficiency is present.

Genomic DNA was extracted from peripheral blood leukocytes of the patient, her parents, and 50 normal control subjects using the Qiagen DNA blood and cell culture kit (Qiagen, Hilden, Germany) to perform PCR exonic amplification (Fig. 2A). Sequences of the primers used for amplification of EIF2AK3 (accession no. 16162274) and GNC2 (accession no. NT_010194) genes are available upon request (all primers were purchased from Mycrosynth, Balgach, Switzerland). Direct cycle sequencing of the PCR products was performed using the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit and analyzed by electrophoresis on the ABI Prism 310 Genetic Analyzer (Perkin Elmer Applied Biosystems, Rotkreuz, Switzerland). HphI restriction analysis was performed as suggested by the supplier (New England Biolabs, Allschwil, Switzerland)

To study the functional implications of the mutations found, we established an RT-PCR method using EIF2AK3 mRNA ectopically expressed in peripheral blood leukocytes of normal subjects. Total RNA was extracted from 500 μl whole blood using the RNeasy minikit (Qiagen). RT was performed on 100 ng total RNA using Superscript reverse transcriptase (Invitrogen/Life Technologies, Basel, Switzerland). PCR amplification of EIF2AK3 cDNA was performed using the direct primer 5′-GCGGGAGAGGCAGGCGTCAGT-3′ and the reverse primer 5′-TTGGCTCAAATTAGGTTATGC-3′ under the following cycling conditions: 93°C for 1 min, 44°C for 2 min, and 68°C for 1 min × 40 cycles. The primers contained a BamHI and an EcoRI restriction sites, respectively, at their 5′-end, to facilitate subcloning. The wild-type cDNA was subjected to PCR-mediated site-directed mutagenesis. The wild-type and the mutant EIF2AK3 cDNA were subcloned into a pCMV-Script vector (Stratagene, Basel, Switzerland) and used for functional studies. Confluent COS1 cells (American Type Culture Collection no. CRL 1650) were transiently transfected using 50 μg lipofectamine and 10 μg DNA on a 10-cm plate (Life Technologies). The transfection efficiency, assessed via a β-galactosidase-containing vector (SV-40 β-gal; Promega, Wallisellen, Switzerland), ranged between 10 and 30%. Western blot analysis was performed using standard procedures. The first antibody, used at a titration of 1:200, was a polyclonal goat anti-human αPERK (Santa Cruz Biotechnology, Lucerne, Switzerland) and a rabbit polyclonal phosphorylation site-specific αeIF2α (Biosource International, Lucerne, Switzerland). The second antibody was a rabbit anti-goat and a goat anti-rabbit IgG alkaline phosphatase conjugate (BioRad, Glattbrugg, Switzerland).

Cellular extracts were obtained from transfected COS1 cells (48 h after transfection) as already described (6). The activity of wild-type and mutant EIF2AK3 (PERK) was measured in immune-complexed kinase assay using 1, 3, and 5 μg of purified human eIF2α as substrate (Biosource International) as previously described (4). As negative control, equivalent amounts of the nonphosphorylable eIF2α 51A peptide were used (Biosource International).

GAD, IAA, and IA2 were measured using commercially available kits (Schering, Geneva, Switzerland). Glycosylated hemoglobin levels were determined using the Bayer DCA 2000 method (Bayer, Leverkusen, Germany).

Direct sequencing of PCR fragments amplified from genomic DNA of the patient revealed the presence of a homozygous T to C exchange in exon 13 of the EIF2AK3 gene leading to the missense S877P mutation in the catalytic domain of the EIF2AK3 protein (Figs. 2 and 3A). Serine 877 is conserved in human, mouse, and rat proteins. The rearrangement induces the loss of an HphI restriction site that is instead present in the heterozygous state in the parents of the patient (Fig. 3B) and on both alleles of the 50 normal control subjects (100 alleles; representative experiments, Fig. 3C). These data suggest that the T to C exchange is not a polymorphism. The mutation in the EIF2AK3 gene of our patient was re-created by site-directed mutagenesis and used for functional studies. This rearrangement does not interfere with protein translation and stability, as demonstrated by the Western blot analysis (Fig. 4A and B). The P877 mutated protein is less efficient than the wild-type kinase in phosphorylating itself and is completely unable to phosphorylate eIF2α (Fig. 4C), independently from the amount of substrate (Fig. 4D and E). This loss of kinase activity is likely the pathophysiological basis of the disease in our patient. The lack of an increase in the immunoprecipitable phosphorylated eIF2α (Fig. 4C) between 3 and 5 μg substrate indicates 3 μg substrate as the saturation limit of our system.

Because of the occurrence of hypoglycemia in our patient, which is not present in Perk−/− mice, we also analyzed the primary sequence of the GCN2 gene. GCN2 is thought to be responsible for eIF2α phosphorylation in the liver in response to nutrient depletion, with consequent activation of gluconeogenetic enzymes. No mutation was found in the GCN2 gene of our patient (not shown).

The study of the molecular basis of Mendelian types of diabetes-related disorders, such as maturity-onset diabetes of the young, severe insulin resistance, and WRS, has improved our knowledge of factors involved in the regulation of pancreatic β-cell function. WRS provides further intriguing insights into the relation between glucose homeostasis and protein translation control. In our patient, all the hallmarks of WRS were present, including infancy-onset type 1 diabetes, spondylo-epiphyseal dysplasia, hepatic dysfunction, and growth retardation. The diagnosis was confirmed at the molecular level by the presence of a homozygous mutation of a conserved residue (S877) located in the catalytic domain of the EIF2AK3/PERK protein. The single S877P amino acid exchange completely abolishes the ability of PERK of phosphorylating its physiological substrate eIF2α. This is the first functional demonstration of a link between lack of eIF2α phosphorylation and human inherited disease, although mutations in the EIF2AK3 gene have been demonstrated to be linked to WRS (2). The recent generation of mice bearing targeted disruption of the Perk gene (7) gave new insights in the function of the enzyme and may allow extrapolation for human disease. For instance, in both mice and humans, type 1 diabetes is the main phenotypic feature, and it is caused by loss of β-cells (8). This β-cell depletion does not begin in utero but after birth in Perk−/− mice and in patients with WRS (9 and present work), and it is largely caused by β-cell apoptosis. These observations support the model of the role of PERK in pancreatic β-cell function, in which eIF2α phosphorylation causes a decrease in the number of unfolded proteins in the ER and promotion of β-cell survival. eIF2α phosphorylation might also result in the transcriptional activation of genes that are important for cell survival. In the absence of a functional PERK, the accumulation of unfolded proteins in the ER and the lack of survival factors eventually lead to β-cell apoptosis and diabetes.

However, the WRS and the Perk−/− mice phenotypes display significant differences. For example, the liver involvement and the hypoglycemic crises suffered by our patient are absent in the Perk−/− mice. Severe hypoglycemia caused by defective hepatic gluconeogenesis is rather a feature of the Ser⁵¹ mutant mouse (10), carrying an eIF2α that cannot be phosphorylated. In the liver, eIF2α phosphorylation seems to be mediated by GCN2 because deletion of the other eIF2α kinases (including PERK) does not result in hypoglycemia in mice. We therefore searched for mutations in the GCN2 gene of our patient and found none. It is then unlikely that our patient’s hypoglycemic crises are caused by concomitant GCN2 defect, although we cannot exclude abnormalities in the expression or posttranslational modifications of GCN2. These observations suggest that, unlike in the mouse, PERK might be important for eIF2α phosphorylation in the human liver. In particular, lack of activation of PERK in amino acid depletion states and/or in stress would prevent the physiological activation of gluconeogenetic enzymes, causing blunted stress-related hyperglycemic response and, in concomitance with insulin administration, prolonged but transient hypoglycemia, as our patient also experienced. Any broad conclusions, however, must await the description and characterization of more WRS patients.

Finally, the bone involvement indicates a pleiotropy in humans that might not be present in mice and, more importantly, demonstrates a crucial role of EIFAK3/PERK (and eIF2α phosphorylation) in skeletal development in humans.

This work was supported in part by the Swiss National Science Foundation Grant 32-063629.0.

The authors are grateful to Dr. Torresani for critical discussion of the work. Christian Acklin and Renate Zollinger are acknowledged for their excellent technical skills. We are also indebted to Prof. C.H. Heizmann for his constant support.