Whole‐exome sequencing for genetic diagnosis of idiopathic liver injury in children

Abstract Genome‐wide approaches, such as whole‐exome sequencing (WES), are widely used to decipher the genetic mechanisms underlying inter‐individual variability in disease susceptibility. We aimed to dissect inborn monogenic determinants of idiopathic liver injury in otherwise healthy children. We thus performed WES for 20 patients presented with paediatric‐onset recurrent elevated transaminases (rELT) or acute liver failure (ALF) of unknown aetiology. A stringent variant screening was undertaken on a manually‐curated panel of 380 genes predisposing to inherited human diseases with hepatobiliary involvement in the OMIM database. We identified rare nonsynonymous variants in nine genes in six patients (five rELT and one ALF). We next performed a case‐level evaluation to assess the causal concordance between the gene mutated and clinical symptoms of the affected patient. A genetic diagnosis was confirmed in four rELT patients (40%), among whom two carried novel mutations in ACOX2 or PYGL, and two had previously‐reported morbid variants in ABCB4 or PHKA2. We also detected rare variants with uncertain clinical significance in CDAN1, JAG1, PCK2, SLC27A5 or VPS33B in rELT or ALF patients. In conclusion, implementation of WES improves diagnostic yield and enables precision management in paediatric cases of liver injury with unknown aetiology, in particular recurrent hypertransaminasemia.


| INTRODUC TI ON
2][3][4][5][6][7] Affected children with poorly functioning liver may need prolonged treatment and/or transplantation, or even die.3][14][15][16][17][18][19][20][21][22][23] However, genetic studies focusing on children and, in particular, patients with different ethnic backgrounds are still limited.The identification of inborn monogenic causes of indeterminate liver diseases holds significant clinical importance, as it may facilitate familial testing for earlier diagnosis, prognosis prediction and more precise management in high-risk family members.Genetic diagnosis streamlines treatment modalities, enhances medical management and clinical decision-making for transplantation.Therefore, in this study, we aimed to identify candidate disease-causing single-gene variants by WES in patients with childhood-onset recurrent idiopathic hypertransaminasemia or with indeterminate paediatric ALF.

| Ethics
This study was conducted in accordance with the institutional, local and national ethical guidelines, and approved by the İhsan Doğramacı Bilkent University Ethics Committee (#2019_11_21_07 and #2020_06_17_01).Clinical history and peripheral blood samples were obtained by the referring physician, with a written informed consent from each participant enrolled in this study and parents if the participant was a minor.

| Patient recruitment
Patient recruitment into this study was performed together with a network of referring physicians in Turkey.Study population included both prospective and retrospective cases with paediatriconset (≤18 years of age) idiopathic liver injury.The evaluation and final diagnosis of the cases were determined entirely at the referring physician's discretion.If the physician could not establish a specific diagnosis for the cause of liver dysfunction, due to lack of supporting evidence, the final diagnosis was considered as idiopathic.Overall, idiopathic cases were enrolled within two diagnostic groups: (i) recurrent elevated liver transaminases (rELT) and (ii) ALF.For rELT, we recruited patients who have had paediatric-onset, at least three repeated episodes of elevated liver transaminases (alanine aminotransferase [ALT] and aspartate aminotransferase AST), equal to or more than twice the upper limit of normal range, with normalization of liver biochemical parameters between crises.[3]

| WES and variant analysis
WES was performed on the genomic DNA (gDNA) isolated from the peripheral blood samples of the participants using commercially available kits.Library preparation, collection of raw sequencing data, alignment with the reference human genome, variant calling and annotations were performed by a WES service provider, Macrogen, Europe or Genoks, Turkey.The allele frequency (AF) values were obtained from the public genome databases: gnomAD, 24 Bravo, 25 and UK Biobank-Allele Frequency Browser (AFB). 26Low quality variants were removed.Only predicted loss-of-function (pLOF) (frameshift indel, stop-gain and essential splicing [± 2 bp from the exon-intron boundary]), start-loss, stop-loss, in-frame indel and missense variants were retained for further analysis.0][31] Regarding CADD, the variant was considered damaging if the CADD score was higher than the MSC value (95% confidence interval) of the mutated gene.Significance of the variants related to the human health were obtained from the ClinVar database (https:// www.ncbi.nlm.nih.gov/ clinv ar/ ).Clinical interpretation and classification of sequence variants were performed manually and using an automated tool such as InterVar, 32 based on the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) guidelines. 33Details of WES analysis and variant filtering are provided in the Supplemental Materials and Methods.

| Sanger sequencing of gDNA
The regions encompassing the target alleles on gDNA were amplified by PCR.Sanger sequencing of the PCR amplicons were performed by a service provider (Macrogen).SnapGene Viewer software (GSL Biotech LLC, USA) was used for sequence analysis.Primers are listed in Table S1.

| Characteristics of the study population
We recruited a total of 20 patients with idiopathic liver injury: 10 patients with recurrent elevated liver transaminases (rELT) and 10 patients with ALF.All cases were sporadic with no familial history.
Female and male patients were nearly equally distributed in both groups.Eight patients, six rELT and two ALF, were born to consanguineous parents.All the rELT patients were alive as of the writing of this article.Among the ALF patients, seven underwent transplantation and two recovered upon medical treatment.Two ALF patients died.WES was performed for all 20 patients, with 13 singletons (five patients with rELT and eight patients with ALF) and seven patientparent trios (five trio designs with rELT and two trio designs with ALF).

| Search for variants in genes predisposing to inherited diseases with liver involvement
We performed a biased WES analysis to determine whether there were any candidate pathogenic variants in genes previously reported to be associated with inherited diseases with liver involvement.We therefore manually curated a liver gene panel by searching for the genes annotated with hepatic and/or biliary phenotype or laboratory finding of elevated liver enzymes/ transaminases in the OMIM database (https:// www.omim.org/ ).A total 380 genes were included in the liver panel, which is listed in Table S2.We analysed the WES data of all patients to search for (i) biallelic (homozygous and compound heterozygous) variants with AF <1% in liver panel genes linked with autosomal recessive (AR) or X-linked recessive (XLR) modes of inheritance and (ii) monoallelic (heterozygous and hemizygous) variants with AF <0.01% in liver panel genes linked with autosomal dominant (AD), X-linked dominant, or XLR modes of inheritance, respectively.Variants listed as benign in the ClinVar were excluded.Missense variants predicted to be benign by all in silico prediction algorithms, CADD, MutationTaster, PolyPhen-2 and SIFT, were also filtered out.Finally, we performed a case-level review to assess the pathogenicity of the gene mutated with regard to its compatibility with the clinical symptoms of the affected patient.Overall, we identified six (five with rELT and one with ALF) of 20 patients carrying rare nonsynonymous variants in nine genes from the liver panel (Table 1; Figure 1).All variants were located on highly-conserved amino acid residues across various species (Figure 2A) and predicted to be damaging by at least three of the algorithms tested, including CADD, MutationTaster, PolyPhen-2 and SIFT, when applicable (Figure 2B).Detailed clinical and laboratory findings of these six patients are provided in Table 2.
Homozygous p.Arg225Gln in Acyl-CoA oxidase 2 (ACOX2) was identified in P1 with rELT (Table 1).Variant's status in patient's father was not assessed as his gDNA was not available, but the mother was heterozygote as confirmed by Sanger sequencing (Figures 1 and S1).
Biallelic mutations in ACOX2 were linked to congenital bile acid synthesis defect (MIM: 617308).A different homozygous missense mutation at the same location, p.Arg225Trp, was reported in patients with congenital bile acid synthesis defect, 34,35 whereas p.Arg225Gln was not listed in the ClinVar or previously associated with a disease.
Overall, rELT, low-normal levels of gamma-glutamyl transferase (GGT) and favourable response to ursodeoxycholic acid treatment in P1 strongly suggested a bile acid synthesis defect (Table 2), probably due to the homozygosity of predicted-to-be deleterious p.Ar-g225Gln in ACOX2 (Figure 2B), albeit being classified as variant of uncertain significance (VUS) based on the ACMG-AMP guidelines (Table 1).
We found a homozygous nonsense variation, p.Glu394*, in Glycogen phosphorylase L (PYGL) in P2 with rELT (Table 1).Sanger sequencing confirmed that this variant was heterozygous in patient's mother, but unknown in father as his gDNA was not available (Figures 1 and S1).Biallelic mutations in PYGL were shown to cause glycogen storage disease VI (MIM: 232700), represented by increased liver glycogen content and hepatomegaly.The p.Glu394*, a novel pLOF variant not listed in any public genome database (Table 1) or ClinVar, was high likely to be morbid in P2, who also had hyperlipidemia and hepatosteatosis (Table 2).In addition, there was a homozygous missense variant, p.Thr308Met, found in Solute Carrier Family 27 Member 5 (SLC27A5) in P2 (Table 1).The p.Thr308Met was heterozygous in patient's mother (Figures 1 and S1).SLC27A5 is mainly expressed in the liver and involved in the regulation of fatty acid and bile acid metabolism. 36It was recently shown that Slc27a5 deficiency led to spontaneous liver fibrosis development in mice. 37homozygous missense mutation (p.His338Tyr) in SLC27A5 has been implicated in Bile acid conjugation defect in a neonate presented with fibrosis and cholestasis in liver biopsy, however there was no experimental evidence provided.38 The p.Thr308Met was not listed in the ClinVar or previously reported to contribute to a disease susceptibility.It was predicted to be damaging by CADD, PolyPhen-2 and SIFT, but it still remains as VUS (Table 1; Figure 2B).Nevertheless, it is possible that homozygosity of p.Thr308Met in SLC27A5 and p.Glu394* in PYGL both contribute to pathogenesis of rELT in P2 (Table 2).Hemizygous p.Arg186Cys was found in Phosphorylase kinase regulatory subunit alpha 2 (PHKA2) in P3 with rELT (Table 1).
Sanger sequencing confirmed that patient's mother was heterozygote for this allele, whereas his father was WT, consistent with XLR inheritance (Figures 1 and S1).Inherited mutations in PHKA2 have been associated with glycogen storage disease (MIM: 306000), which can present with hepatomegaly and elevated liver enzymes.The p.Arg186Cys was listed as likely pathogenic in the ClinVar (VCV000010535.8) and previously implicated in a patient with X-linked liver glycogenosis type 2. 39,40 Therefore, this variant was highly considered to be the genetic cause underlying rELT in P3 (Table 2).
We identified a homozygous missense variant, p.Ala984Thr, in ATP binding cassette subfamily B member 4 (ABCB4) in P4 with rELT (Table 1).Biallelic mutations in ABCB4 have been associated with Gallbladder disease 1 (MIM: 600803) and progressive familial intrahepatic cholestasis type 3 (PFIC3) (MIM: 602347).We confirmed by Sanger sequencing that the familial segregation of the p.Al-a984Thr allele was consistent with an AR mode of inheritance, as both healthy parents were heterozygous for the mutation, whereas the healthy siblings did not carry the mutation (Figures 1 and S1).
The p.Ala984Thr was previously reported at heterozygosity in an adult patient with PFIC3, albeit without any experimental evidence of causality. 41Yet, the homozygosity of ABCB4:p.Ala984Thr, predicted as likely pathogenic (Table 1; Figure 2B), could largely explain TA B L E 1 Patients with rare nonsynonymous variants in liver panel genes.the clinical symptoms of P4, which included elevated GGT, hepatosplenomegaly and cholelithiasis (Table 2).

Patient
A homozygous missense variant, p.Arg649Trp, in Codanin 1 (CDAN1) was found in P5 with ALF (Table 1).Zygosity of p.Ar-g649Trp was not assessed in P5's parents as their gDNA samples were not available (Figures 1 and S1).Biallelic mutations in CDAN1 have been implicated in congenital dyserythropoietic anaemia type Ia (MIM: 224120), in which affected patients may develop severe hepatic injury such as ALF. 42,43Indeed, the p.Arg649Trp with p.Ar-g397Trp at compound heterozygosity were reported in a patient with congenital dyserythropoietic anaemia type Ia. 44However, although this variant was interpreted as likely pathogenic based on the ACMG-AMP classification (Table 1), P5's clinical findings were not compatible with the known phenotypic spectrum of CDAN1 deficiency (Table 2).

| DISCUSS ION
Liver injury of unknown origin still represents a major burden in paediatric hepatology despite diagnostic advances and extensive aetiological workup.This may be attributed in part to nonspecific findings and/or overlapping symptoms, presented by a known disease with primary or secondary liver involvement.Alternatively, it could signify a novel liver disorder with yet uncharacterized clinical manifestations.Thus, there is an urgent necessity for enhanced diagnostic approaches to enable earlier and more precise interventions in affected children.3][14][15][16][17][18][19][20][21][22][23] Herein, we investigated a total of 20 paediatric cases with rELT or ALF of unknown aetiology using WES.While there was no candidate morbid variation found in ALF patients, we established a genetic diagnosis in four out of 10 rELT patients (40%) using a liver gene panel.
). Inherited heterozygous variants in JAG1 are associated with Alagille syndrome 1 (MIM: 118450), which can include liver phenotypes, such as cholestasis and intrahepatic duct deficiency, and laboratory abnormalities including increased conjugated bilirubin, hypercholesterolemia, hypertriglyceridemia, and elevated transaminases.The p.Asn108His in JAG1 was listed as VUS in the ClinVar (VCV001020741.7).Moreover, biallelic mutations in PCK2 have been associated with mitochondrial phosphoenolpyruvate carboxykinase deficiency (MIM: 261650), characterized by hypoglycemia and liver failure.Finally, biallelic mutations in VPS33B have been associated with arthrogryposis, renal dysfunction, and cholestasis (MIM: 208085), progressive familial intrahepatic cholestasis (MIM: 620010), and keratoderma-ichthyosis-deafness syndrome (MIM: 620009), all of which have liver abnormalities reported.However, both p.Gly215Asp in PCK2 (VUS in ClinVar, VCV002713587.2) and p.Leu403Phe in VPS33B were not previously associated with any known diseases.Yet, none of these three genetic variants of uncertain significance matched with the patient's clinical phenotype (Tables

F I G U R E 2
Variant effect predictions.(A) Schemas show conservation of mutated amino acid residues across various species.Asterisk (*), colon (:), and period (.) indicate fully conserved, strongly similar, and weakly similar sites, respectively.Source: Clustal Omega (B) The predicted impact of variants using four algorithms, CADD, MutationTaster, PolyPhen-2 and SIFT, is shown.Red colour indicates damaging, whereas green colour is benign.N/A, Not applicable.TA B L E 2 Clinical and laboratory findings of 6 patients with rare nonsynonymous variants in liver panel genes.

Table 1
). Familial segregation of these three variants with the disease was not assessed as gDNA samples from parents were not available (Figures1 and S1