Abstract

Objective:

Cholestasis has diverse causes, with genetic factors playing a key role. Diagnosis is challenging due to varied presentations and overlapping genetic conditions. Next-generation sequencing cholestasis gene panels enable faster, more accurate identification of genetic causes. This study summarizes results from more than 10,000 tests, highlighting their clinical utility.

Methods:

We analyzed aggregate data from a 77-gene cholestasis panel used between 2016 and 2022. Eligible patients had unexplained cholestasis or chronic liver disease. DNA sequencing utilized custom capture libraries (SureSelect 2016–2021 and PGxome® 2021–2022). Variants were classified per ACMG/AMP guidelines. Definitive diagnoses required biallelic pathogenic/likely pathogenic variants in autosomal recessive genes or a single pathogenic/likely pathogenic variant in autosomal dominant JAG1 or NOTCH2. Potential diagnoses involved one pathogenic/likely pathogenic variant in autosomal recessive genes plus one variant of uncertain significance.

Results:

Of 10,894 samples analyzed, 51.1% were from patients less than 1 year old and 9.2% from those 18 years of age or older. Overall, 2917 patients carried one or more pathogenic or likely pathogenic variant(s). Diagnostic yield was 6.8% for definitive and 2.2% for potential diagnoses. Definitive findings were most common in JAG1, SERPINA1, ABCC2, ABCB11, CFTR, POLG, and NOTCH2. Potential diagnoses commonly involved ABCC2, ABCB4, ABCB11, CFTR, and PKHD1. Monoallelic variants were frequent in SERPINA1, CFTR, DHCR7, ABCB4, and PKHD1.

Conclusions:

These cholestasis gene panel results reinforce their value in diagnosing and identifying complex genetic causes of cholestasis, especially in infants less than 1 year old. Early detection supports timely intervention, and the panel provides clearer insight to support accurate diagnoses and inform potential therapeutic strategies.

1 Introduction

Cholestasis, defined by impaired bile formation or flow, arises from diverse hepatobiliary disorders with both environmental and genetic causes. In neonatal cholestasis, genetic etiologies account for about one-quarter of cases, reinforcing the importance of early diagnosis (1–3). These include multisystem syndromes like Alagille syndrome (ALGS) (JAG1 or NOTCH2 variants) and isolated hepatic conditions such as alpha-1-antitrypsin deficiency (A1ATD) (SERPINA1). Other inherited causes include bile acid synthesis disorders (BASD) due to single enzyme defects (HSD3B7, AKR1D1, CYP27A1), progressive familial intrahepatic cholestasis (PFIC)-related canalicular transport defects (ATP8B1, ABCB11, ABCB4, TJP2, NR1H4, MYO5B, SLC51A, USP53, KIF12, ZFYVE19), and systemic disorders such as cystic fibrosis (CF) (CFTR) and autosomal recessive polycystic kidney disease (PKHD1) (2, 4–9).

Although often considered a neonatal disorder, cholestasis can present in childhood or adulthood, reflecting a lifelong phenotypic expression. For example, ABCB11 variants are associated with benign recurrent intrahepatic cholestasis (BRIC), and ABCB4 variants with intrahepatic cholestasis of pregnancy (ICP) and other adult-onset forms. Genetic susceptibility also contributes to complex hepatobiliary diseases, including gallstones, fatty liver, and drug-induced liver injury (DILI), demonstrating the broad influence of inherited factors on hepatic physiology (10–14).

Despite major advances, pinpointing the genetic cause of cholestasis remains challenging. Overlapping phenotypes, variable penetrance, and numerous variants of uncertain significance (VUS) complicate diagnosis and management. Next-generation sequencing (NGS) has improved the diagnostic landscape by enabling simultaneous analysis of many genes implicated in bile acid metabolism, transport, and signaling (8, 15–18).

This study presents a large-scale analysis of cholestasis gene panel testing in children and young adults, offering updated insight into diagnostic yield and the evolving genetic landscape. With a dataset nearly five times larger than earlier reports (8), it refines diagnostic yield estimates and highlights newly implicated genes as well as trends in variant classification. The findings affirm both the clinical utility and current interpretive challenges of broad NGS-based testing in cholestatic disorders.

2 Materials and methods2.1 Study design and population

This retrospective analysis evaluated patients with a clinical diagnosis or history of cholestasis who underwent NGS using a targeted cholestasis gene panel. The panel expanded from 57 to 77 genes in 2018–2019 (Supplementary Table 1). Testing was provided at no cost through the Genetic Cholestasis Panel Testing Program, sponsored by Retrophin, Travere Therapeutics, and Mirum Pharmaceuticals, Inc., to facilitate early diagnosis and management of genetic cholestasis.

Eligible patients had intrahepatic cholestasis or chronic liver disease of unknown cause. Exclusions included extrahepatic disorders (e.g., biliary atresia, choledochal cyst, large-duct primary sclerosing cholangitis [PSC]) and cholestasis secondary to total parenteral nutrition. Blood or saliva samples were collected by healthcare providers and sent to Eurofins/Emory Genetics Laboratory (EGL; 2016–2021) and PreventionGenetics (PG; 2021–2022) with a completed test requisition form and signed informed consent. The study was approved by the Institutional Review Board of Johns Hopkins University School of Medicine (IRB00299237) with a waiver of informed consent. All research was conducted in accordance with the Declarations of Helsinki and Istanbul.

2.2 Genetic testing and variant classification

DNA was extracted and sequenced using hybrid-capture enrichment (SureSelect [EGL] or PGxome® [PG]) followed by NGS. Sequence reads were aligned to GRCh37/UCSC hg19 reference genome. Variant interpretation followed ACMG/AMP guidelines, and variants were classified as pathogenic (P), likely pathogenic (LP), variant of uncertain significance (VUS), likely benign (LB), or benign (B). Copy number variants (CNVs) were assessed using validated read-depth-based algorithms and are summarized separately in Supplementary Table 2.

2.3 Definition of diagnostic categories

A definitive molecular diagnosis was defined as either:

Two pathogenic/likely pathogenic (P/LP) single nucleotide variants (SNVs; homozygous or compound heterozygous) in an autosomal recessive gene, or

One P/LP SNV in an autosomal dominant gene (JAG1 or NOTCH2 in ALGS).

A potential diagnosis was assigned to cases with one P/LP variant and one VUS in the same autosomal recessive gene. Monoallelic findings in autosomal recessive genes were summarized separately.

Copy number variants were analyzed as part of the testing workflow but were evaluated independently from SNVs and were not incorporated into diagnostic-yield calculations.

The diagnostic yield was calculated as the number of positive results (definitive + potential diagnoses) divided by the total number of completed tests. Demographic metadata (age and ethnicity) were available for all samples, whereas sex data were only available for those analyzed by PG. Clinical data were not available for the present analysis.

2.4 Subanalysis of PFIC-associated genes

A targeted subanalysis evaluated heterozygous and compound heterozygous variants in ATP8B1, ABCB11, ABCB4, TJP2, NR1H4, MYO5B, and SLC51A—genes implicated in PFIC that frequently generate partial or ambiguous results that complicate clinical interpretation. Other PFIC-associated genes, including USP53, KIF12, and ZFYVE19, were not included in the current panel.

This subanalysis aimed to:

Quantify the frequency of monoallelic pathogenic variants in each gene.

Identify cases with a potential diagnosis (P/LP + VUS); and

Explore possible cross-gene interactions, where monoallelic pathogenic variants in one PFIC gene co-occurred with VUS in another.

Cases meeting these criteria were extracted from the larger dataset and analyzed separately to assess potential digenic or modifier effects.

2.5 Enrichment analysis of VUS for genes of interest

To evaluate the significance of VUS in this cohort, their frequency in a clinically enriched cholestatic disorder population was compared with population-based reference data from the Genome Aggregation Database (gnomAD). Heterozygous VUS (as classified by the diagnostic laboratories) were extracted across 11 genes of interest: JAG1, NOTCH2, ATP8B1, ABCB11, ABCB4, TJP2, NR1H4, MYO5B, HSD3B7, AKR1D1, and CYP27A1. Corresponding variant frequencies from gnomAD v4.1 encompassing all population groups were matched by gene symbol and MANE Select transcript consequence. The number of variants analyzed per gene was 105, 200, 113, 195, 164, 186, 22, 110, 57, 34, and 99, respectively. Statistical comparison between the two groups was conducted using a two-proportion z-test, and P values were adjusted for multiple testing using the Benjamini–Hochberg (BH) method. Results were considered statistically significant for adjusted P values <0.05.

3 Results3.1 Cohort characteristics

A total of 10,894 samples were analyzed between February 2016 and February 2022 across two diagnostic laboratories. Panel size expanded over time: ∼4000 samples used the original 57-gene panel (2/2016–11/2018), ∼4600 the 66-gene panel (11/2018–2019), and ∼2400 the final 77-gene panel (2022), with each sample evaluated using the version active at testing (Table 1 and Supplementary Table 1). Patients ranged in age from birth to 88 years, with infants and young children predominating. Over half (51.1%) were under one year old, while 9.2% were adults. Specific results from the adult subset will be analyzed and reported separately (Table 2). Infants aged 2 to 3 months represented the largest age group (13.0%), followed by 1-year-olds (12.6%).

Cholestasis-related genetic disordersGenesBASD due to single enzyme defects and CTXAKR1D1; AMACR; BAAT; CYP27A1; CYP7A1; CYP7B1; DHCR7; HSD3B7; SLC27A5Peroxisomal disorders, including ZSDPEX1; PEX10; PEX11B; PEX12; PEX13; PEX14; PEX16; PEX19; PEX2; PEX26; PEX3; PEX5; PEX6; PEX7PFICABCB11; ABCB4; ATP8B1; MYO5B; NR1H4; TJP2ALGSJAG1; NOTCH2A1ATDSERPINA1CFCFTRPolycystic kidney diseasePKHD1Other genetic causes of cholestasisABCC2; ABCG5; ABCG8; ACOX2; AKR1C4; ALDOB; CC2D2A; CLDN1; DCDC2; DGUOK; EHHADH; FAH; GNAS; GPBAR1; HNF1B; HSD17B4; INVS; KMT2D; LIPA; MKS1; MPV17; NPC1; NPC2; NPHP1; NPHP3; NPHP4; PKD1L1; POLG; SCP2; SLC10A1; SLC10A2; SLC25A13; SLC51A; SLC51B; SLCO1B3; SMPD1; TALDO1; TMEM216; TRMU; UGT1A1; UTP4; VIPAS39; VPS33B

Genes included in the cholestasis sequencing panel.

A1ATD, alpha-1-antitrypsin deficiency; ALGS, Alagille syndrome; BASD, bile acid synthesis disorders; CF, cystic fibrosis; CTX, cerebrotendinous xanthomatosis; PFIC, progressive familial intrahepatic cholestasis; ZSD, Zellweger spectrum disorders.

CharacteristicsTotal (N = 10,894)Age range, y0–88Age categories, n (%) <1 y5570 (51.1) 1–10 y2850 (26.1) 11–17 y1477 (13.6) >18 y997 (9.2)Sex,an (%) Female1281/3275 (39.1) Male1994/3275 (60.9)Ethnicity, n (%) Caucasian/Northwestern European3681 (33.8) Hispanic2010 (18.5) African American1621 (14.9) Asian658 (6.0) Native American104 (1.0) Other122 (1.1) Unknown2080 (19.1) Mixed/multiple618 (5.7)

Key demographics and baseline characteristics.

a

Sex data were only provided by PG. PG, PreventionGenetics.

Among patients with available sex data (n = 3275), 60.9% were male and 39.1% female. Ethnicity was reported for all samples. Larger self-identified groups were Caucasian/Northwestern European (33.8%), Hispanic (18.5%), African American (14.9%), and Asian (6.0%). Smaller groups included Native American (1.0%) and Other (1.1%), while 19.1% reported unknown and 5.7% mixed or multiple ethnic backgrounds (Table 2).

3.2 Overall variant distribution

Across all genes analyzed, 2591 patients had at least one P variant and 435 had at least one LP variant, totaling 2917 patients (26.8%) with a reportable P/LP variant. Of these, 81.7% (n = 2384) carried a single P/LP variant, 16.5% (n = 482) carried two, and 1.6% (n = 48) carried three; only three patients (0.1%) harbored more than three reportable variants (Figure 1A).

(A) distribution of reported pathogenic and likely pathogenic variants in patients with ≥1 variant across genes included in the panel. (B) Frequency of genes with pathogenic and likely pathogenic variants.

Overall, P variants were identified in 57 genes and LP variants in 57 partially overlapping genes, with a combined total of 64 unique genes harboring at least one P or LP variant (Figure 1B).

3.3 Copy number variants

Copy number variants were identified in over 50 genes (Supplementary Table 2) but were most frequently observed in SLCO1B3 (n = 26), SLCO1B1 (n = 9), ABCG5 (n = 7), ABCG8 (n = 6), and ABCC2 (n = 6). Several CNVs were classified as P or LP, including recurrent deletions in NPHP1, NPHP4, HNF1B (as part of multigene events), ABCB11, ATP8B1, AMACR, and SLC51A. The majority of CNVs were VUS, most prominently involving ABCG5, ABCG8, PEX11B, DCDC2, and SCP2.

3.4 Diagnostic yield

Based on variant classification and zygosity, 745 patients (6.8%) met the criteria for a definitive molecular diagnosis, while 241 (2.2%) had potential diagnoses, with an overall diagnostic yield of 9.0% (Table 3). The remainder of the cohort had nondiagnostic findings, which encompassed VUS, B and LB variants, reduced function or at-risk alleles, and negative genetic results (Table 4). Definitive diagnoses were most frequently attributed to JAG1 (n = 192), SERPINA1 (n = 142), ABCC2 (n = 72), ABCB11 (n = 45), CFTR (n = 38), POLG (n = 31), and NOTCH2 (n = 25). Potential diagnoses most often involved ABCC2 (n = 49), ABCB4 (n = 40), ABCB11 (n = 38), CFTR (n = 24), and PKHD1 (n = 17) (Table 3; Figure 2).

GeneaDefinite diagnosisPotential diagnosisMonoallelic variantsABCB11453881ABCB42140100ABCC2724979ABCG59ABCG81240ACADM1ACOX26AKR1D1614ALDOB13183AMACR8ATP8B114722BAAT1CC2D2A31CFTR3824360CYP27A19465CYP7B1224DCDC21214DGUOK7210DHCR751157FAH9GNAS1HNF1B10HSD17B410HSD3B775INVS111JAG1192KMT2D4LIPA3233MKS1127MPV178216MYO5B35NOTCH225NPC115634NPC25NPHP111NPHP32628NPHP4225NR1H416PEX111232PEX1011PEX11B3PEX122111PEX1911PEX226PEX26115PEX34PEX56PEX6418PEX721PKD1L1110PKHD121789POLG31671SCP213SERPINA11424569SLC10A15SLC25A137218SMPD13239TALDO12TJP28410TMEM2164TRMU14UGT1A19636VIPAS3936VPS33B9110

Frequency of patients with >1 P and/or LP variant with definitive and potential diagnoses and monoallelic variantsa.

LP, likely pathogenic; P, pathogenic.

a

The following genes from the 77 gene panel had no P or LP variants: AKR1C4, CLDN1, CYP7A1, EHHADH, GPBAR1, PEX13, PEX14, PEX16, SLC10A2, SLC27A5, SLC51A, SLC51B, SLCO1B3, UTP4.

GeneHOM P/LP, nTwo alleles P/LP, nTwo alleles 1 VUS + 1 P/LP, nOne allele P/LP, nHOM VUS, n1 or 2 HET alleles VUS, nOther reportable, nB/LB, nNegative sample, nABCB11a21243881144591133ABCB4a1474010025487130ABCC2a3042497921582226ABCG5a929614ABCG8a12402381831ACADMb1