Natural history of Wolcott‐Rallison syndrome: A systematic review and follow‐up study

To systematically review the literature for reports on Wolcott‐Rallison syndrome, focusing on the spectrum and natural history, genotype‐phenotype correlations, patient and native liver survival, and long‐term outcomes.


| INTRODUC TI ON
Wolcott-Rallison Syndrome (WRS) is a rare autosomal recessive disorder first described by Wolcott and Rallison in 1972. 1 It is caused by biallelic mutations in eukaryotic translation initiation factor 2α kinase 3 (EIF2AK3), which encodes the protein kinase RNA (PKR)like endoplasmic reticulum (ER) kinase (PERK).At the ER, PERK is a sensor and regulator of the unfolded stress response, which ensures proper function of protein synthesis and ER homeostasis. 2e reported WRS features include endocrine pancreas insufficiency, intellectual deficits, hypothyroidism, neutropenia, skeletal dysplasia, liver and kidney failure. 3Common to all reported WRS patients is the insufficiency of the endocrine pancreas (neonatal diabetes mellitus). 3Interestingly, patients, even siblings, sharing the same EIF2AK3 mutation show varying disease manifestation. 4,5However, with the exception of neutropenia, 5 no clear genotype-phenotype correlation has been postulated so far.
Another common manifestation of WRS is multiple epiphyseal dysplasia.In particular, long bones, vertebrae and pelvis are affected.Bone mineralisation is aberrant as well. 6Moreover, liver impairment marked by acute episodes of hepatic dysfunction with elevated liver enzymes and jaundice are frequent in individuals with WRS.Often, spontaneous remission of acute liver injury is observed, yet acute liver failure is a known complication.The triggers for these episodes can be viral infections (e.g.infections of the upper airway).
A recent study showed that 85% of WRS patients identified suffered from liver failure with high morbidity. 7e management of WRS depends on the clinical manifestation of the particular patient.Neonatal diabetes is often treated successfully by subcutaneous insulin while recurrent bone fractures and skeletal dysplasia require orthopaedic management.
The increased risk of liver complications needs to be highlighted, as specific treatment is not available, and patients need to be referred to a paediatric hepatology centre at earliest convenience.In acute liver failure, liver transplantation (LT) can be the only treatment option.
Despite a growing body of case reports and series of WRS, a systematic analysis of the natural history of WRS, phenotypic and genotypic spectrum and outcome is lacking.This systematic review tries to close this gap.

| MATERIAL S AND ME THODS
This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guidelines. 8equent clinical feature was neonatal diabetes in all patients, followed by liver impairment in 73%, impaired growth in 72%, skeletal abnormalities in 59.8%, the nervous system in 37.6%, the kidney in 35.4%, insufficient haematopoiesis in 34.4%, hypothyroidism in 14.8% and exocrine pancreas insufficiency in 10.6%.Episodes of acute liver failure were frequently reported.Liver transplantation was performed in six, combined liver-pancreas in one and combined liver-pancreas-kidney transplantation in two individuals.Patient survival was significantly better in the transplant cohort (p = .0057).One-, five-and ten-year patient survival rates were 89.4%, 65.5% and 53.1%, respectively.Liver failure was reported as the leading cause of death in 17.9% of cases.Overall survival was better in individuals with missense mutations (p = .013).

Key points
• We include 189 individuals, including 30 patients not reported previously.
• Liver failure was reported as the leading cause of death.
Episodes of acute liver failure were frequently reported.
Albeit liver transplantation was only done in nine patients, survival was significantly better in the transplant cohort (p = .0057).

| Ethics statement
This study was conducted in accordance with the guidelines of the Institutional Review Board of the Medical University of Innsbruck and the 1975 Declaration of Helsinki. 9Participants gave written informed consent for genetic investigations according to local regulations.All papers were categorised and original studies were extracted.If an article was considered eligible by either of the two reviewers, the full text was retrieved where available.

| Search strategy and data extraction
After categorisation, a second search was performed on full text articles to check for method quality and fulfilment of inclusion and exclusion criteria.Two investigators (CB and GFV) extracted the data elements from all pre-selected studies into an Excel spreadsheet.These data included authorship, year of publication, type of study, patient characteristics, presence and onset of symptoms and genotype.Phenotypes were classified using the human phenotype ontology (https:// hpo.jax.org/ ).In cases of incomplete data, we tried to contact the authors for further study information.Data were controlled, cross-checked and confirmed by three reviewers (CB, DA and GFV).Based on tabulated data, studies were again assessed for inclusion eligibility and data consistency by both reviewers.Missing data on a substantial level led to study exclusion.In case of any disagreement, the issue was solved by discussion.

| Inclusion and exclusion criteria
The prerequisite for study inclusion was a diagnosis of WRS based on clinical or genetic grounds.Observational studies, case series and case reports were included.Reviews and commentaries were excluded.

| Follow-up
For follow-up inquiries, study authors were contacted by e-mail to extend the individual data based on the data extraction criteria initially applied.Follow-up data were merged with baseline data from study extraction.

| Study quality and bias assessment
Data extracted from the systematic literature search and the additionally recruited patients were combined for further analyses.
Quality assessment of included studies was performed with the Joanna Briggs Institute critical appraisal tools for cohort studies, case series and case reports (PMID 26262566).Additionally, the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for conference abstracts was used if no full text article was retrievable. 8For abstracts of case studies certain items of the STROBE checklist were considered as not applicable.Studies were considered with low, moderate or serious risk of bias if they fulfilled more than 80%, between 60 and 80% or less than 60% of the requested items respectively.

| Outcome measures and genotype categorisation
Baseline data, including age, sex, age at diagnosis, onset of symptoms, genotype (as annotated in the respective publication), phenotype, therapeutic measures, cause of death and time of follow-up, were extracted.

| Data synthesis and statistical analysis
The hepatic phenotype was subdivided into 'acute liver failure' (ALF) according to the paediatric acute liver failure study group 10 and 'acute liver injury' (ALI) without coagulopathy.Data harmonisation was conducted in the following instance: if recurrent episodes of acute liver injury or failure were mentioned without exact numbers given, at least two episodes were assumed.All statistical analyses were performed using ggplot, rcompanion (Cramer's V), survival and survminer packages in R statistical software. 11For Kaplan-Meier estimator analyses, Cox log rank test was used.A p-value ≤.05 was considered significant.Bar and density plots were generated via the R ggplot2 package.The world map was generated using the R gglot2 and maps packages.Schematics and figures were compiled using Illustrator CC 2019 (Adobe).

| In-silico modelling of EIF2AK3 missense variant pathogenicity scores
3][14] ) were annotated for all biologically possible EIF2AK3 missense variants and mapped onto a linearised representation of the PERK protein, as previously demonstrated. 15 generated all biologically possible base substitutions in the EIF2AK3 coding sequence (transcript: NM_004836.7)and used the Mutalyzer Position Converter to match the resulting variant call format file to the GRCh37/hg19 reference genome.Scores were annotated using the Ensembl variant effect prediction tool.
Predictions were generated using the aRgus webtool. 16A generalised additive model was built using the geom_smooth function of the R 11 ggplot2 package to plot a smoothened line and confidence interval.

| Study cohort
The study cohort comprises 159 individuals identified through a systematic literature search from 62 studies 4,5,7,17-75 (Figure 1) and an additional 30 individuals who were collected via contacting physicians and study authors (Supplemental Table S1).Additional followup data were acquired on 11 individuals already reported 7,[24][25][26]28,45 via direct contact with the original study authors. Overa 46 case reports, 12 case series and 4 cohort studies were included in the final analysis.We identified 26 studies with low risk, 30 with moderate risk and 4 with serious risk of bias (Supplemental Table S3).

F I G U R E 1
Preferred reporting items for systematic reviews and meta-analysis flowchart depicting the screening and selection process of studies included in the systematic review.

Identification of studies via databases and registers
Finally, a total of 189 individuals worldwide were included in this study, of whom 51.3% were female (97/189, 16 not available (NA)) (Figure 2A, Supplemental Table S2).Consanguineous parents were reported for 138 individuals (16 NA).Diabetes, which was the initial disease presentation in all patients, was recognised at a median age of 2.5 months (interquartile range (IQR) 2) and genetic diagnosis was confirmed at a median age of 2.5 months (IQR 4).The median duration of follow-up was 48 months (IQR 62.5 months, max. 35 years) with a total duration of follow-up of 945.75 years.In the entire cohort, 29.6% (56/189, 7 NA) were reported deceased.The median age at death was 36 months (IQR 50.75).One-, five-and ten-year patient survival was 89.4%, 65.5% and 53.1%, respectively (Figure 2B).Liver failure was cited as the leading cause of death (17.9% cases), followed by unknown causes (6.3%, 12 cases), multiorgan failure (4.8%, nine cases), metabolic derangement (3.2%, six cases), pneumonia (one case), cerebral oedema (one case) and renal failure (one case) (Figure 2C).

| Neonatal diabetes
The median age at diagnosis was 2.5 months (IQR 2).Eight of the 189 patients (4%) did not fulfil the neonatal diabetes criteria with age at manifestation older than 6 months (range 6.4-30 months).
No correlation between disease onset and different genotypes

| Liver impairment
The hepatic phenotype is manifested by episodes of ALF and or ALI.ALF episodes were reported at a median of one (IQR 1) per individual, and ALI episodes at a median of two (IQR 1) (Figure 4B).
Trigger events were not systematically reported.Recurrent episodes were identified in 71 individuals but triggers were not systematically reported.Both ALF and ALI episodes occurred early in the first two years of life, with a median age at the first episode of 10 months (IQR 8).No correlation between genotypes PERK1-3 and ALF (Cramer's V 0.107) or ALI (Cramer's V 0.0378) was seen (Supplemental Table S4).Patient survival was significantly better in the group without ALI episodes than in the group with ALI (p = .014)(Figure 4C).No significant difference in survival was seen in patients with ALF compared to patients without ALF (p = .1).LT was carried out in 9 individuals (4.8%) at a median age of 42 months (IQR 72).LT was done in six individuals (one living donor, two deceased donor and three NA), combined liver-pancreas transplantation in one and combined liver-pancreas-kidney transplantation in two individuals.The median duration of follow-up after LT was 24 months (IQR 627 months, max.14 years), with a total duration of follow-up of 36.25 years.All transplanted individuals were reported to be alive.Overall, one-, five-and ten-year native liver survival was 99.4%, 96.1% and 89.5% (Figure 4D).Survival for patients with episodes of ALF or ALI was significantly better in the transplant cohort (p = .0057)(Figure 4E).

| Nervous system
Developmental delay without any further specification was reported in 18.5% (35/189) of cases.In an additional ten cases, isolated neuromotor delay was seen, whereas mild intellectual disability was present in 20 (10.6%) and severe intellectual disability in three cases.Seizures were reported in 18 cases (9.5%), ataxia in two cases and nystagmus in one case.Microcephaly was reported in nine cases and hydrocephalus in two cases.Other features were cerebellar atrophy (two cases), leukodystrophy (one case) and lissencephaly (one case).

| Renal involvement
Renal dysfunction was reported in 29 (15.3%),renal insufficiency in 22 (11.6%),renal tubular acidosis in six (3.2%), multicystic dysplastic kidneys in two and enlarged kidneys in two cases.One child had an ectopic kidney and another child had acute tubular nectrotic injury.

| DISCUSS ION
WRS is a rare multisystem genetic disorder with a broad clinical spectrum.It is recognised as the most frequent cause of neonatal diabetes in patients with consanguineous parents. 76Other manifestations vary in frequency and severity, including involvement of the liver, skeletal, renal, nervous and haematopoietic systems.Various genotype-phenotype correlations were explored but could neither be confirmed nor refuted due to the small number of cases. 3,4This systematic review provides a comprehensive report on the natural history of the disease in 189 individuals, highlights genotype-phenotype correlations and provides insight into survival rates and outcomes.
The main clinical feature was insulin-requiring diabetes in all of the 189 cases, occurring in the first months of life except for 8 individuals with later onset (>6 months).It was previously postulated that mutations that have been associated with a relatively milder phenotype (later onset or longer survival) appear to be all missense mutations located in the first kinase domain. 3However, we could not verify this hypothesis in our study as no specific genotype was associated with late-onset presentation.Both missense and truncating mutations at different regions of the gene were observed in these cases.With one exception, a residual kinase activity was reported (N655K), which may explain the comparable late onset (30 months) in this specific case. 4However, in general, it should not be overlooked that access to medical care is not comparable all around the world and that diagnosis might be delayed due to limited awareness and resources rather than reflecting the natural course of the disease.
We found overall survival to be poor, with a median age at death of 2.5 years.Interestingly, there was a clear survival advantage for individuals with missense mutations in both or one allele over individuals with two loss-of-function mutations.A possible explanation might be residual protein function retained in patients with missense variants, but studies exploring this hypothesis are lacking.
Strikingly, liver impairment was reported in over 70% of the cases.However, no correlation between specific genotypes and the hepatic phenotype could be identified.Recurrent liver injury, Considering all this, LT should be aimed for, especially after preceding liver crises.As severe neuro-cognitive impairment was rare (3 out of 189 patients), maximal therapy with the best possible outcome should be opted.The next question that arises is whether to pursue isolated LT, combined liver-pancreas transplantation, or even multiorgan, including kidney transplantation, as in previously reported cases. 35,53Combined liver-pancreas transplantation for difficult-tocontrol neonatal-onset diabetes is an obvious choice, provided the necessary resources are available.Also, the potential exocrine pancreatic insufficiency could be remedied with it.Renal involvement is less frequent than liver disease, with 35.4%, but the further course of kidney disease is unclear.It is possible that renal involvement may become significant much later, and therefore longer follow-up data would be necessary to answer this question sufficiently.
Management of diabetes in WRS patients can be a challenge not only due to the young age of onset but also because WRS individuals have a tendency for recurrent hypoglycaemic episodes, most likely resulting from liver dysfunction with impaired gluconeogenesis. 30erefore, insulin pump therapy, especially in small infants, is recommended in order to optimise the prevention of severe hypoglycaemia.However, such therapy should be considered based on the local facilities and resources of the treating centre.
Of note, both liver failure and metabolic derangements are frequently triggered by infections. 3Arguably, the increased secretory load to the hepatocytes' UPR associated with viral infections cannot be resolved due to the underlying PERK deficiency and might trigger apoptosis.As this is still poorly understood, a therapeutic recommendation cannot be derived yet.Neutropenia was common, reported in 22.2% of cases.Affected patients tend to suffer from frequent infections and are therefore at high risk for liver and metabolic crises.The once-postulated correlation between neutropenia and a specific genotype 4 could not be confirmed in our study.
In addition to frequent pancreatic and hepatic involvement, somewhat less frequent renal involvement and hematologic abnormalities, a major feature of the underlying disease was growth restriction and skeletal dysplasia of long bones, pelvis and vertebrae affecting gait and in some cases, bone abnormalities even led to spinal cord compression. 44Multidisciplinary care, including orthopaedic-paediatric management, is therefore strongly recommended.
Overall, there was poor concordance between genotype and the variable features of the disease, suggesting that other genetic or environmental factors may be influential.Triggers leading to

:
Wolcott-Rallison syndrome has variable clinical courses.Overall survival is better in individuals with missense mutations.Liver-or multi-organ transplantation is a feasible treatment option to improve survival.K E Y W O R D S acute liver failure, ER stress, neonatal-onset diabetes, paediatric, skeletal dysplasia

For
study selection, a systematic literature research was performed using the following bibliographic databases: PubMed, Livivo, Google Scholar, Scopus and Web of Science.The search terms 'Wolcott-Rallison syndrome', 'Wolcott-Rallison', 'WRS', 'Permanent neonatal diabetes', 'Neonatal insulin-dependent diabetes mellitus' or 'EIF2AK3 mutation' were used.The date of the last search was May 2023.Only English articles were considered for further analysis.Two investigators (AMK and GFV) independently performed screening of the mentioned databases via abstract and title search.
Records identified from database searching: PubMed, Livivo, Google Scholar, Web of Science (n = 998) Total records before removal of duplicates (n = 1001) Records screened via title and abstract (n = 489) Additional studies from other sources (n = 3) Duplicate records removed (n = 512) Records excluded due to irrelevance (n = 367) Screening Included Identification Records excluded with reason: (n = 44) Lacking information or study error (n = 2) Thorough assessment for eligibility (n = 114) Articles included (n = 62) No cases reported (n = 42)

A
genetic diagnosis was confirmed in 182 individuals, and relevant information was missing in seven.In total, 26 individuals with two missense mutations (PERK-1), five with one non-sense mutation (PERK-2) and 151 with two loss-of-function mutations (PERK-3) were identified.All EIF2AK3 variants reported in this study are shown in Figure 3A.The frequency of EIF2AK3 variants reported in this study with respect to the affected protein domains is shown in Figure 3B and the in silico pathogenicity prediction, using REVEL scoring, of all potential EIF2AK3 missense variants is shown in Figure 3C.Overall survival was better in F I G U R E 2 (A) World map representing the origin of individuals included in the study.(B) Overall survival shown with the Kaplan-Meier estimator.(C) Causes of death with reported individual numbers are shown with a bar plot.++ + + + + + + + + + + + + + + + + + + + + + + + + + + ++ + ++ + + + + + + ++ + + + + + mutations in both or one allele over individuals with two loss-of-function mutations (p = .013)(Figure3D).
often triggered by benign viral infections, putatively triggering an ER-stress crisis and hepatocytic apoptosis, occurred early in the first two years of life and significantly impacted patient survival, but not every acute liver injury resulted in liver failure.These are important facts when trying to deduce treatment strategies for acute liver injury or failure in WRS patients: for example, how long should we watch and wait and when is the right time for liver or even combined multi-organ transplantation?Only nine out of 189 WRS patients had undergone liver or combined liver-pancreas transplantation in our cohort, with patient survival being significantly better in the transplant cohort.As we identified liver failure as the main cause of death in WRS individuals, care at specialised hepatologic centres seems paramount.Individuals with recurrent ALI or ALF had a worse outcome compared to those who had not experienced a recurrent acute liver injury.Without access to LT centres, specialised care, including current strategies for the management of ALF episodes (e.g.access to renal replacement therapies and plasma exchange),77 should be considered to bridge WRS patients to recovery throughout the first years of life.Particularly in WRS-associated ALF treatment with Nacetylcysteine as a scavenger for reactive oxygen species might be considered, but systematic data is lacking.78