Deleterious variants in ABCC12 are detected in idiopathic chronic cholestasis and cause intrahepatic bile duct loss in model organisms

Abstract

Background and Aims:

The etiology of cholestasis remains unknown in many children. We surveyed the genome of children with chronic cholestasis for variants in genes not previously associated with liver disease and validated their biological relevance in zebrafish and murine models.

Method:

Whole-exome (n=4) and candidate gene sequencing (n=89) was completed on 93 children with cholestasis and normal serum γ-glutamyl transferase (GGT) levels without pathogenic variants in genes known to cause low GGT cholestasis such as ABCB11 or ATP8B1. CRISPR/Cas9 genome editing was employed to induce frameshift pathogenic variants in the candidate gene in zebrafish and mice.

Results:

In a one-year-old female patient with normal GGT cholestasis and bile duct paucity, we identified a homozygous truncating pathogenic variant (c.198delA, p.Gly67Alafs*6) in the ABCC12 gene (NM_033226). Five additional rare ABCC12 variants, including a pathogenic one, were detected in our cohort. ABCC12 encodes MRP9 that belongs to the ATP-binding cassette transporter C family with unknown function and no previous implication in liver disease. Immunohistochemistry and Western blotting revealed conserved MRP9 protein expression in the bile ducts in human, mouse, and zebrafish. Zebrafish abcc12 null mutants were prone to cholangiocyte apoptosis, which caused progressive bile duct loss during juvenile stage. MRP9-deficient mice had fewer well-formed interlobular bile ducts and higher serum alkaline phosphatase levels compared to WT mice. They exhibited aggravated cholangiocyte apoptosis, hyperbilirubinemia, and liver fibrosis upon cholic acid challenge.

Conclusion:

Our work connects MRP9 with bile duct homeostasis and cholestatic liver disease for the first time. It identifies a potential therapeutic target to attenuate bile acid-induced cholangiocyte injury.

Lay Summary:

ABCC12 encodes MRP9 protein that is important for maintaining cholangiocyte health in zebrafish and mice. Pathogenic variants in this gene are associated with intrahepatic bile duct loss in patients with chronic cholestasis.

Graphical Abstract

Introduction

Neonatal cholestasis is a rare condition of impaired bile flow presented with conjugated hyperbilirubinemia in the neonatal period. Retention of bile acids contributes to pruritus and liver fibrosis, while lack of luminal bile can lead to malnutrition and fat-soluble vitamin deficiency. Neonatal cholestasis may arise from acquired conditions such as infections, or from inherited diseases resulting in impaired hepatocellular bile acid synthesis and trafficking, diminished canalicular bile excretion, or inadequate elimination by a hypoplastic biliary tree ¹.

In acquired cholestatic conditions such as extrahepatic biliary atresia and primary sclerosing cholangitis (PSC), serum levels of γ-glutamyl transferase (GGT), an enzyme expressed by cholangiocytes, are elevated. In inherited conditions of intrahepatic cholestasis with defective canalicular bile acid transport and diminished bile flow, the serum GGT levels are typically low or within the normal range ² Genetic investigations in kindreds with low GGT progressive familial intrahepatic cholestasis (PFIC) have associated the disease with pathogenic variants in ATP8B1 encoding for phospholipid flippase FIC1 ³, ABCB11 encoding for canalicular bile salt export pump BSEP ⁴, and NR1H4 encoding for FXR that is the master regulator of bile acid synthesis and transport ⁵ Pathogenic variants in tight junction genes TJP2, LSR, and USP53 have also been found in rare cases of low/normal GGT inherited cholestasis ^6–9

As opposed to a plethora of genes linking hepatocellular function to intrahepatic cholestasis, little is known about the roles of the molecular determinants of bile duct development and cholangiocyte health in PFIC. Discovering new genes associated with inherited biliary diseases may identify novel genetic causes for PFIC and genetic modifiers for acquired cholestatic liver diseases ¹⁰ Pathogenic variants in VPS33B and KMT2D were detected in rare cases of bile duct paucity and low GGT cholestasis in arthrogryposis-renal dysfunction-cholestasis and Kabuki syndrome, respectively ^{11, 12} Variants in genes governing bile duct development, including JAG1 (Alagille syndrome), KDM6A (Kabuki syndrome), GPC1, and ADD3, have been associated with biliary atresia ^{11, 13–15}.

We hypothesized that whole-exome and targeted sequencing in infants with idiopathic, low GGT chronic cholestasis syndromes will identify novel genetic causes for neonatal cholestasis. We identified a homozygous, nonsense pathogenic variant in ABCC12 encoding an ATP-binding cassette (ABC)-transporter, Multidrug Resistance-associated Protein 9 (MRP9), in a patient with bile duct paucity and low GGT cholestasis. We explored the role of ABCC12 pathogenic variants as genetic cause for liver disease in a cohort of 93 patients and delineated its biological function in two model organisms.

Methods

Please refer to Supplemental Material for comprehensive materials and methods.

HUMAN Subjects

Whole-exome sequencing (WES) was performed on four subjects from the Middle East with idiopathic low GGT cholestasis. Targeted Sanger or next-generation sequencing (NGS) of ABCC12 were conducted on additional 89 deidentified subjects with low GGT chronic cholestasis without disease causing pathogenic variants in ATP8B1 or ABCB11 from either the Molecular Genetics Laboratory at Cincinnati Children’s Hospital Medical Center (CCHMC) (n=34) or the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)–supported Childhood Liver Disease Research Network (ChiLDReN) (n=55). Subjects enrolled into either of two prospective observational CHiLDReN cohort studies (A Prospective Database of Infants with Cholestasis [PROBE; NCT00061828] or Longitudinal Study of Genetic Causes of Intrahepatic Cholestasis [LOGIC; NCT00571272]) with GGT levels <100 IU/L at the time of enrollment were eligible for inclusion. This study protocol was approved by the Institutional Review Boards at CCHMC (Exome Sequencing in liver disease [IRB #2013–0086], Genetic variants in children with chronic cholestasis [IRB#2012–1311]), or the ChiLDReN ancillary study committee, respectively, and conformed to the ethical guidelines of the 1975 Declaration of Helsinki. Informed consent was obtained from parents/guardians or participants 18 years or older, and assent was obtained from participants >7 years of age, per local guidelines. Limited clinical data were obtained for all 93 subjects enrolled in this study.

WES, Targeted sequencing, variant annotation, and analysis

WES, Targeted sequencing, and analysis protocols were developed and validated by the CCHMC Molecular Genetics Laboratory of the Division of Human Genetics. Detailed protocols are provided in the Supplemental Material.

Zebrafish

Fish lines used in our experiments included: Wild type (WT), abcc12^ci203+/−, Tg(fabp10a:dsRed)^gz4, Tg(EPV. Tp1-Mmu.Hbb:EGFP)^um14//Tg(Tp1:GFP), Tg(−3.5ubb:loxP-EGFP-loxP-mCherry)^cz1701, and Tg(EPV.Tp1-Mmu.Hbb:CreERT2,cryaa:mCherry)^s959/Tg(Tp1:CreERT2). Generation of abcc12 mutant zebrafish was described in the Supplemental Material. Zebrafish were maintained under standard conditions ¹⁶ in accordance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health publication 86–23, revised 1985) and approved by the Institutional Animal Care and Use Committee at CCHMC. Animals of both genders were studied.

Lineage tracing experiments

WT and abcc12 mutant zebrafish expressing Tg(EPV.Tp1-Mmu.Hbb:CreERT2;cryaa:mCherry;−3.5ubb:loxP-EGFP-loxP-dsRed) transgene were treated with 6.7 μM 4-Hydroxytamoxifen (4-OHT) (Sigma-Aldrich, St. Louis, MO) or Ethanol (EtOH) for 32 hours. To test the specificity and efficacy of lineage labeling, five EtOH-treated control and five 4-OHT-treated larvae were fixed at 48 hours post treatment and stained with Alcam antibody that labels the intrahepatic bile ducts ¹⁷ and DAPI that marks the nuclei. Lineage-labeling efficiency was calculated based on the number of DsRed+ cholangiocytes per total number of cholangiocytes. The remaining animals were harvested at 5 month of age. Their livers were dissected and cryosectioned into 50 μm sections, and stained with Anxa4 and DsRed antibodies and DAPI ¹⁸ For each liver, six to ten confocal z-stacks were collected from all three liver lobes. Lineage-labeling efficiency was calculated based on the number of DsRed+ Anxa4+ double positive cells per total number of Anxa4+ cells.

Generation of Abcc12 mutant mice using CRISPR/Cas9-based genome editing Abcc12 mutant mice were generated in the C57Bl6/J background by the CCHMC Transgenic Animal and Genome Editing Core. Two sgRNAs (CRISPR #1: 5’-GGCAGCCCAGCATCATCCACAGG-3’; CRISPR #2: 5’-GAGTCATAAGGAGACAGTGGGGG-3’) were designed by using CRISPRscan (http://www.crisprscan.org) to target exon 3 of the Abcc12 transcript (NM_172912.4) ¹⁹ The mutant allele harboring a 100 bp deletion (c.139_238del, p.Asp47Cysfs*22) could be genotyped by PCR amplification using the following primers: VS5198: 5’-CCCCAGGCTCCATATGTAGC-3’ and VS5199: 5’-CCTCAGTCCAAGCCACATCTC-3’. The WT amplicon was 408 bp and the mutant fragment was 308 bp.

In situ hybridization, immunofluorescence, and histology

Whole-mount in situ hybridization on zebrafish larvae was performed as described ²⁰. To generate DNA templates to make the abcc12 anti-sense RNA probe, a 879 bp fragment of the open reading frame were amplified from the total cDNA of 5-day-old WT larvae using the following primers: 5’-CGGTTCTAGATTACAGGGAACCAG-3’ and 5’-GATAAACGAGTTCGCACCATGAATG-3’. The fragment starts in exon 3 and ends in exon 14, which includes the CRISPR targeting site located in exon 4.

Immunofluorescence of whole-mount larval and juvenile zebrafish was performed as described ²¹. The stained juvenile zebrafish were cleared by FocusClear (Cedarlane Laboratories, Burlington, Ontario, Canada) at room temperature for 4 hours ²². Immunofluorescence on paraffin sections was performed as described ²³ Tris-EDTA buffer (pH9.0) (MRP9) or a high pH buffer (H3301, Vector Laboratories, Burlingame, CA) (GFP, Anxa4, and Bhmt) was used for antigen retrieval under high pressure and heat for 10 minutes. All confocal images were taken on Nikon Ti-E Inverted Microscope, A1R Confocal, GaAsp PMTs, Andor iXon 888 EMCCD Widefield Camera (Nikon Instruments Inc., Melville, NY). Post image processing and data analyses were conducted using Imaris (Bitplane, Concord, MA). Hematoxylin and eosin (H&E) stain and chromogenic immunohistochemistry were conducted on paraffin sections as described ^{22, 24} Primary and secondary antibodies are listed in Table S1.

Bile duct quantification

Number and type of bile duct in the hilar region were determined per 100 portal triads in liver sections from 4 WT and 3 Abcc12−/− mice at 6 months. Bile ducts were labeled by anti-PanCK antibody and classified into three categories according to a previous report ²⁵ Well-formed bile ducts referred to those with one layer of epithelial cells and a round lumen. Functional bile ducts had one or more layers of biliary cells and a discernible lumen. Clusters referred to groups of biliary cells with no discernible lumen.

Statistical Analysis

Student two-tailed t-test and ANOVA were performed using GraphPad Prism (GraphPad, La Jolla, CA). P values of less than .05 were considered statistically significant.

Results

WES identified a homozygous frameshift variant in ABCC12 in a child with PFIC phenotype

To identify novel genes associated with PFIC-like symptoms, we conducted WES in four children from the Middle East with low GGT chronic cholestasis without genetic diagnoses. Bi-allelic rare variants in PFIC genes TJP2 or ABCB11 were detected in three of the children. The fourth child did not carry any variants in known genes associated with low GGT cholestasis, including ATP8B1, ABCB11, TJP2, USP53, and LSR. Instead, a homozygous nonsense pathogenic variant c.198delA (p.Gly67Alafs*6) in the gene ABCC12 was identified. ABCC12 encodes an ABC transporter MRP9 with unknown function.

The female index patient presented with seizure and subdural hematoma at 6 months of age, which was suspected to be the result of coagulopathy from cholestatic liver disease. At 12 months, her serum alkaline phosphatase (ALP) and aspartate aminotransferase (AST) levels were mildly elevated at 322 IU/L and 41 IU/L, respectively. GGT, direct bilirubin, and ALT levels were within the normal range. She subsequently developed recurrent attacks of severe itching, fat-soluble vitamin deficiency, and hepatomegaly. During an episode of pruritus at 15 months of age, her serum total bile acid levels were highly elevated at 278 μmol/L (normal <8.2 μmol/L), direct bilirubin, ALP, and AST levels rose to 83 μmol/L, 463 IU/L, and 62 IU/L, respectively, while GGT remained normal at 10 IU/L. Pruritus was partially responsive to treatment with ursodeoxycholic acid, rifampicin, and cholestyramine. Currently at 9 years of age her itching is greatly improved. She underwent a liver biopsy at 12 months of age. Histomorphological analysis of H&E-stained liver sections revealed mild periportal inflammation and stage 1 fibrosis (Figure 1A). CK7 immunohistochemistry showed poorly formed interlobular bile ducts, prominent ductular reaction, and prevalent bi-phenotypic hepatocytes (Figure 1B). Bile duct paucity was confirmed by CK19 immunofluorescence demonstrating that only 50% of the portal triads contained a lumenized interlobular bile duct (Figure 1C–E).

Figure 1. Biallelic pathogenic variants in ABCC12 were detected in the index patient with low GGT cholestasis and bile duct paucity.

(A) H&E staining of the liver biopsy from the index patient. (B) CK7 staining showed extensive ductular reaction and lack of lumenized bile ducts in some portal triads. (C-D) Confocal single plane images of representative portal triads in the control and patient livers stained with cholangiocyte marker CK19 (red) and nuclei marker DAPI (blue). (E) The ratio of lumenized bile duct to portal triad numbers quantified based on CK19 immunofluorescence. (F) Racial background of the 93 subjects enrolled into the three cohorts for genomic studies. (G) Putative topological models of WT human MRP9 protein and sites of pathogenic (circled) and VOUS variants discovered by WES or targeted NGS of ABCC12. Scale bars: (A-B) 50 μm; (C-D) 20 μm.

A subset of children with idiopathic chronic cholestasis harbored rare variants in ABCC12

To explore the prevalence of rare variants in ABCC12 among children with PFIC-like symptoms, we performed Sanger Sequencing or NGS of all 29 exons and intron-exon boundaries of the ABCC12 gene on DNA samples from additional 89 children belonging to two cohorts. Cohort 1 consisted of 34 subjects with low GGT jaundice who had JaundiceChip (Affymetrix-Thermo Fisher Scientific, Santa Clara, CA) re-sequencing of ATP8B1, ABCB11, and other cholestasis genes performed at the CCHMC Molecular Genetics Laboratory ²⁶. None of the patients were found to have disease-causing variants in any of the genes re-sequenced. Cohort 2 of 55 subjects were enrolled into the Cholestasis Long-Term Observation Study of ChiLDReN and had GGT levels <100 IU/L at enrollment. They were not known to harbor disease-causing variants in established cholestasis genes.

The 93 subjects enrolled into the WES study and from the two cohorts were of diverse racial background (Figure 1F). Six rare variants with the highest minor allele frequency (MAF) <1% were detected in the coding region of ABCC12, of which two were interpreted as pathogenic variants predicted to cause premature truncation, and four as variants of uncertain significance (VOUS; Table 1). Three of the variants, including the two truncating pathogenic variants, have not been reported in the gnomAD database. The full-length MRP9 protein is predicted to contain two nucleotide binding domains in the intracellular part and two core membrane-spanning domains ^{27, 28} The two truncating pathogenic variants were located at the N-terminal intracellular domain (circled in Figure 1G). Missense variants c.1159A>G (p.Met387Val) and c.3215T>A (p.Val1072Glu) were predicted to be localized in the transmembrane domains. Missense variants c.3417C>G (p.Ser1139Arg) and c.3907T>C (p.Phe1303Leu) were localized intracellularly. Including the index patient, we found one subject with putative homozygous pathogenic variants in ABCC12 and five other patients with heterozygous variants in the study cohort. We did not identify additional copy number variants in patients with heterozygous changes.

Table 1:

Main characteristics of the subjects with rare variants in ABCC12

Subject ID#	Gender	Race/Ethnicity	cDNA	Protein	Zygosity	MAF (%)	CADD	Interpretation
1Δ	F	Arabic	c.198delA	p.Gly67Alafs*6	Hom	NA	NA	Pathogenic
2	F	Arabic	c.3907T>C	p.Phe1303Leu	Het	0.02	24.1	VOUS
3	F	Multi-racial	c.1159A>G	p.Met387Val	Het	NA	23.2	VOUS
4	F	Asian	c.38delinsTT	p.Asp13Valfs*14	Het	NA	NA	Pathogenic
5	F	Unknown	c.3215T>A	p.Val1072Glu	Het	0.03	24.1	VOUS
6	M	African American	c.3417C>G	p.Ser1139Arg	Het	0.10	11.58	VOUS

Bi-allelic variants were detected in Subject 1, and single variants were found in Subject 2–6. Δ indicates the variants in the index patient surveyed by whole exome sequencing. F – female, M - male, NA – not available or not reported, Hom – homozygous, Het – heterozygous, MAF – minor allele frequency in population with highest MAF (GnomAD), CADD – Combined Annotation Dependent Depletion, VOUS – variant of uncertain significance. Interpretation was made following the ACMG guidelines.

MRP9 was expressed in the cholangiocytes in human, mouse, and zebrafish

ABCC12 is located on the human chromosome 16q12.1. Its mRNA expression is abundant in the brain and testis (https://www.proteomicsdb.org/ and http://www.proteinatlas.org/). Four different splice variants for ABCC12 were reported in a variety of tissues ²⁸ and we detected at least two of them in the liver (Supplemental Figure S1). Immunohistochemistry revealed MRP9 protein in large and small intrahepatic bile ducts and zone 1 hepatocytes in the control liver tissue from a child without liver disease (Figure 2A,B). Immunofluorescence showed that MRP9 was modestly enriched on the apical membrane of the interlobular bile ducts (Figure 2C). The human MRP9 protein is 85% and 52% identical to its mouse and zebrafish orthologs, respectively (Supplemental File 1). In zebrafish, abcc12 mRNA was expressed in the brain, liver, and intestine at the larval stage, and its expression persisted in the digestive organs through adulthood (Supplemental Figure S2A–B). We performed immunofluorescence on the larval and juvenile zebrafish that expressed Tg(Tp1:GFP) transgene in the intrahepatic bile ducts (Figure 2D–F and Supplemental Figure S2C–E) ^{29, 30} At both stages, Mrp9 was enriched in the intrahepatic bile ducts. In mice, western blot analysis detected MRP9 protein in the testis and purified primary cholangiocytes by (Figure 2G). In summary, MRP9 showed evolutionarily conserved expression in the cholangiocytes in three vertebrate species.

Figure 2. MRP9 protein was expressed in cholangiocytes in human, zebrafish, and mouse.

(A-B) Chromogenic detection of MRP9 protein in the control liver of a child without known liver disease. (C) Confocal single plane image showing expression of MRP9 protein (green) in the intrahepatic bile duct in the control human liver. (D-F) Confocal three-dimensional projections showing Mrp9 protein expression (D) and Tg(Tp1:GFP) transgene expression (E) in the intrahepatic bile ducts in WT zebrafish at 5 days post fertilization (dpf). (F) Merged image of (D) and (E). Ventral views, anterior is to the top. (G) Immunoblotting of protein lysates from mouse testis (positive control), purified primary intrahepatic cholangiocytes, and hepatocytes with antibodies against CK19, MRP9, and loading control β-actin. Molecular weights (MW) are listed. Scale bars: (A) 20 μm; (B) 10 μm; (C-E) 30 μm.

Zebrafish abcc12 mutants exhibited hepatocellular injury at the adult stage

To validate the biological consequences of MRP9 loss, we generated zebrafish abcc12 mutants using CRISPR/Cas9-mediated genome editing. We targeted exon 4 of the abcc12 gene to match the location of the pathogenic variant found in the index patient. The abcc12^ci203 allele harbors a 7 bp and a 38 bp insertions that result in a frameshift variant, creating a 14 amino acids insertion and a premature stop codon (c.262_263ins7 and c.264_265ins38, p.Trp88Thrfs*15) (Figure 3A). The mutant protein was predicted to retain only the N-terminus intracellular domain and lack all the transmembrane helices and nucleotide binding domains (Figure 3B). Whole-mount in situ hybridization and qPCR both showed that abcc12 transcript expression was reduced in the mutants compared to WT siblings (Figure 3C–D).

Figure 3. Zebrafish abcc12 adult mutant males exhibited liver injury.

(A) Schematic representation of the zebrafish abcc12 locus along with the CRISPR target site (orange), PAM motif (red), and WT and indel sequences. 5’ and 3’ UTRs are shown in gray. Nucleotide changes are highlighted in yellow and the premature stop codon is marked. (B) Putative topological models of WT and mutant Mrp9 protein in zebrafish. (C) Expression of abcc12 transcripts in WT (top) and abcc12 mutant (bottom) larvae at 4 dpf as revealed by whole-mount in situ hybridization. (D) qPCR analysis comparing abcc12 transcripts in the livers of 1-year-old WT and abcc12 adult mutant siblings. (E-H) Body length (E), body weight (F), spleen/body weight ratio (G), and total bile salt/liver weight ratio (H) of WT and abcc12 mutant males at 4.5 months of age (mean±s.e.m). Each dot represents the result from individual fish. (I) H&E staining of the liver sections in 5-month-old WT (n=5) and abcc12 mutant males (n=5). The black arrows point to the areas with vascular congestion. The red arrowheads mark necrotic cells with condensed eosinophilic cytoplasm. (J) qPCR analysis comparing expression of abcc12 transcripts in the livers of adult WT males and females. Male 1, 2 and Female 1, 2 are siblings, and Male 3, 4 and Female 3, 4 are siblings. Statistical significance in (D, E-H) was calculated by two-tailed student’s t-test. *, p<0.05; **, p<0.01; ****, p<0.0001. Scale bars: (C) 150 μm; (I) 20 μm.

abcc12 mutants did not show any obvious defects in the number of cholangiocytes or the morphology of intrahepatic biliary network at the larval stage (Supplemental Figure S3A–B). Hepatic bile excretion and biliary bile flow were preserved in the mutants (Supplemental Figure S3C–D). They survived to adulthood and were viable. However, the male mutants were smaller (Figure 3E–F), and showed splenomegaly (Figure 3G) and increased total liver bile salt levels (Figure 3H). H&E stain of the male mutant liver showed cell death and vascular congestion without apparent inflammation or fibrosis (Figure 3I). Confocal time-lapse imaging revealed reduced blood flow in the mutant hepatic vein, consistent with the splenomegaly and vascular congestion phenotypes (Supplemental Figure S4). The female mutants did not exhibit growth delay, hepatic injury, or splenomegaly (Supplemental Figure S3E–H and data not shown). By qPCR, we detected higher expression of abcc12 in the adult WT male livers compared to the female livers, which may contribute to the male-biased liver phenotypes (Figure 3J).

abcc12 mutant zebrafish developed bile duct paucity during the juvenile stage

Mrp9 expression in the cholangiocytes prompted us to examine the intrahepatic bile ducts in the adult mutant zebrafish based on either Tg(Tp1:GFP) transgene expression or immunostaining with the cholangiocyte marker Annexin A4/Anxa4 ³¹. Compared to WT, the male mutant livers showed a drastic decrease in the number of bile ducts (Figure 4A–B and Supplemental Figure S5A–B). Such phenotype was not seen in the female mutants (Supplemental Figure S3H–I). To determine the onset of the bile duct phenotype in the mutants, we performed time-course experiments by harvesting WT and abcc12 mutant zebrafish every week starting at 1 week of age. At each time point, we counted the total number of Tg(Tp1:GFP)-positive cholangiocytes in the liver, and normalized it to the liver volume that was measured based on Tg(fabp10a:Dsred) transgene expression ³² The earliest stage when we observed a statistically significant reduction in the cholangiocyte number in abcc12 mutants was four weeks (Figure 4C), which coincided with sexual differentiation in zebrafish ³³ The liver volumes and hepatocyte numbers were comparable between WT and mutants at this stage (Supplemental Figure S5C–D). The zebrafish liver contains two dorsolateral lobes (left lobe and right lobe) and one ventral lobe that develop sequentially within the first four weeks of life ³⁴ (Supplemental Figure S5E–F). At four weeks of age, only the left lobe, which grew first, showed a decrease in cholangiocytes in the mutants (Figure 4D and Supplemental Figure S5G). In adults, patchy bile duct deficiency was seen in all the liver lobes in the mutants. Taken together, abcc12 mutants displayed bile duct loss at the juvenile stage, with the more matured left lobe to exhibit the phenotype first. The defect progressed into adulthood and displayed male sex predilection.

Figure 4. abcc12−/− zebrafish mutants showed intrahepatic bile duct deficiency caused by cholangiocyte apoptosis.

(A,B) Confocal single plane images of WT and abcc12 mutant liver sections. Tg(Tp1:GFP) expression (green) marks the intrahepatic bile ducts and DAPI staining (blue) labels the nuclei. (C) Numbers of cholangiocyte normalized to liver volume from 1 week to 6 weeks (wk) of age (mean±s.e.m). (D) Numbers of cholangiocyte normalized to hepatocyte numbers in different liver lobes at 4 weeks (mean±s.e.m). In (C,D), the cholangiocytes and hepatocytes were recognized by the expression of Tg(Tp1:GFP) and Tg(fabp10a:DsRed) transgenes, respectively. (E) Schematic illustration of the lineage-tracing analysis. (F,G) Confocal single plane images of 4-OHT-treated WT and abcc12 mutant adult males. DsRed expression (red) marks the lineage-labeled cells. Anxa4 antibody staining (green) labels the cholangiocytes. GFP expression (blue) marks the non lineage-labeled cells. (H) Percentages (mean±s.e.m) of cholangiocytes that incorporated EdU during the indicated incubation periods. (I) Percentages (mean±s.e.m) of TUNEL-positive cholangiocytes at different stages. In (H,I), the cholangiocytes were recognized by Tg(Tp1:GFP) expression. (J) Percentages (mean±s.e.m.) of TUNEL-positive hepatocytes at 3 weeks. (K,L) Transmission electron microscopy images of the cholangiocytes in 5-month-old WT and abcc12 mutant males. N, nucleus; l, lumen. 4 WT and 4 mutant livers, 6 thin sections per liver were examined. In (L), black arrowhead points to the opening of the bile duct lumen. White arrow marks the debris likely resulting from leakage of cytoplasmic contents. Scale bars: (A-B, F-G) 50 μm; (K-L) 2 μm. Statistical significance in (C-D, H-J) was calculated by two-tailed student’s t-test. *, p<0.05; **, p<0.01. In all graphs, each dot represents the result from individual liver.

Bile duct loss in abcc12 mutants was due to increased cholangiocyte apoptosis.

Three plausible mechanisms could cause intrahepatic bile duct deficiency in abcc12 mutants: 1) failure to maintain cholangiocyte cell identity; 2) decreased cholangiocyte proliferation; and 3) increased cholangiocyte cell death or any combination. To determine if abcc12 mutant cholangiocytes preserved their identity during the juvenile stage, we performed lineage-tracing experiment using Tg(TP1:CreERT2;−3.5ubb:loxp-EGFP-loxP-DsRed) transgenic zebrafish (Figure 4E) ^{35, 36} Upon treatment with 4-OHT, the existing cholangiocytes and their descendants switched off EGFP expression and turned on DsRed expression, whereas the rest of the body continued expressing EGFP. We treated WT and mutant zebrafish with either EtOH as the vehicle control or 6.7 μM 4-OHT from 96 hours post fertilization (hpf) to 128 hpf, after differentiation of hepatocytes and cholangiocytes had concluded ³⁷ We harvested some zebrafish at 48 hours post treatment and validated the specificity and efficiency of Cre recombination (Supplemental Figure S6A–B). At 5 months of age, we cryosectioned the livers from the 4-OHT-treated zebrafish and stained them with Anxa4 (Figure 4F–G). The DsRed-positive cells made up 85.6 ± 13.0% and 91.7 ± 13.0% of the Anxa4-positive cholangiocytes in WT (5 zebrafish, 17,726 Anxa4-positive cells) and abcc12 mutant (6 zebrafish, 51,319 Anxa4-positive cells) livers, respectively. All DsRed-positive cells were positive for Anxa4. Therefore, the mutant cholangiocytes maintained their identity and the bile duct paucity was not due to a switch of the cholangiocyte cell fate.

To assess cholangiocyte proliferation, we incubated WT and mutant zebrafish with the replication marker EdU at various stages and analyzed EdU incorporation in the cholangiocytes. There was no significant difference in the proliferation indices between WT and abcc12 mutants at 5 days, 9 days, and 5 weeks of age (Figure 4H).

To study apoptosis, we harvested WT and abcc12 mutant zebrafish every week starting at 1 week of age and performed TUNEL assay or anti-activated Caspase 3 immunostaining. Both assays showed that the mutant cholangiocytes had increased apoptosis compared to WT starting at 3 weeks post fertilization (Figure 4I and Supplemental Figure S6C). Meanwhile, the mutant hepatocytes had similar apoptotic index as WT, suggesting that the increased apoptosis was restricted to the cholangiocytes (Figure 4J). Consistent with increased cell death, abcc12 mutant cholangiocytes exhibited ultrastructural features of cellular injury, including leakage of cytoplasmic content, pleomorphic mitochondria, and opened bile duct lumen (Figure 4K–L). We concluded that Mrp9 function is required for survival and integrity of intrahepatic cholangiocytes through protection from apoptosis.

MRP9-deficient mice developed bile duct paucity

To validate zebrafish findings in mammal, we generated MRP9-deficient mice by deleting 100 bp in exon 3 using CRISPR/Cas9-genome editing (Figure 5A). The variant (c.139_238del, p.Asp47Cysfs*22) resulted in premature termination and loss of MRP9 protein expression (Figure 5B). Adult mutant mice showed significantly elevated serum ALT- and ALP-levels, indicating hepatocellular and biliary injury, respectively (Figure 5C,D). Similar to the patients, Abcc12−/− mutant mice had normal serum GGT levels (Figure 5E). H&E showed no significant steatosis, fibrosis, or necroinflammatory activity in the mutant livers at low magnification. At higher magnification, there was patchy ductular reaction surrounding some of the portal triads in the mutant livers (Supplemental Figure S7). We subjected liver sections to anti-PanCK immunohistochemistry and classified each bile duct profile as either well-formed, functional, or cell cluster without bile duct resemblance as described previously ²⁵ The average number of well-formed bile ducts per portal triad was significantly decreased, while the number of cell clusters was increased in Abcc12−/− mice (Figure 5F). Testing the hypothesis that MRP9 expression in biliary epithelial cells is regulated by primary bile acids present in bile, we incubated neonatal cholangiocytes from WT mice with cholic acid (CA) and detected a dose-dependent increase of Abcc12 mRNA and MRP9 protein expression in these cells in culture (Figure 5G–J).

Figure 5. Abcc12−/− adult male mice exhibited intrahepatic bile duct paucity.

(A) Schematic of the murine Abcc12 locus along with the two CRISPR target sites in exon 3 (orange), PAM motifs (red), and WT and indel sequences. 5’ and 3’ UTRs are shown in gray. Nucleotide changes are highlighted in yellow. (B) Western blotting confirmed absence of MRP9 protein in the mutant mice testes. The antibody recognizes amino acids 690 to 734 of human MRP9, which has high homology to amino acid 692 to 733 of mouse MRP9. (C-E) Serum levels of ALT, ALP, and GGT in 6-month-old male mice (mean±s.e.m.). (F) Liver sections from 6-month-old male mice were subjected to immunohistochemistry with antibody against pan-CK and analyzed for bile duct profiles. The number of each type per portal triad (mean±s.e.m.) was displayed for 3–4 mice/group. (G) qPCR analyses of Abcc12 transcripts in neonatal WT cholangiocytes cultured in water as vehicle control or in various concentrations of cholic acid (CA) for 12 hours. (H) Western blotting showing MRP9, CK19, and β-actin protein expression in neonatal WT cholangiocytes cultured in vehicle or CA for 48 hours. (I,J) The levels of MRP9 and CK19 were quantified by densitometry and normalized to β-actin. Statistical significance was calculated by two-tailed student’s t-test in (C-F,G,I,J). *, p<0.05; **, p<0.01; ***, p<0.001. In (C-F), each dot represents result from individual animal.

Lack of MRP9 conferred susceptibility to bile acid-induced injury in mice

Mouse models of FIC1- and BSEP-deficiency display an attenuated phenotype which is attributed to the less toxic bile acids in mice compared with humans ^{38, 39} Since CA, an abundant bile acid in humans, accentuates the PFIC phenotypes in these models ^{40, 41} and increased MRP9 expression in the cultured cholangiocytes, we fed adult male WT and Abcc12−/− mice with 1% CA-supplemented chow for 7 or 14 days. Following 7 days of CA feeding, weight loss was significantly higher in Abcc12−/− mice and accompanied by higher serum ALT and total bilirubin levels compared with WT (Supplemental Figure S8A–C). CK19 immunohistochemistry and Sirius Red stain indicated more ductular reaction and fibrosis in Abcc12−/− mice compared to WT mice following CA challenge (Supplemental Figure S8D–I). Given that several ABCC12 heterozygous variants were identified in our patients, we included Abcc12+/− mice in the 14-day CA challenge to investigate if partial loss of MRP9 affected the liver. Both Abcc12+/− and Abcc12−/− mice had more weight loss and higher liver/body weight ratio compared to WT mice after CA challenge (Figure 6A). Serum total bilirubin levels were significantly increased in both mutant strains compared with WT mice (Figure 6A). Serum GGT levels were comparable among all three genotypes (Figure 6A). The number of well-formed bile duct profiles per portal triad was significantly reduced in Abcc12−/− mutants compared with WT (Figure 6B). Exploring if the loss of bile duct profiles was related to bile acid-induced cell death, we isolated primary cholangiocytes from neonatal WT, Abcc12+/−, and Abcc12−/− mice and incubated the cells with CA. The viability of cholangiocytes from both mutant strains was significantly lower at baseline compared to WT (Supplemental Figure S9A). The heterozygous cholangiocytes exhibited a more drastic reduction in viability in response to rising CA concentrations compared to WT. Based on the zebrafish study, we investigated apoptosis as a cause for cholangiocyte death in this system by using flow cytometric assay to measure Annexin V binding to phosphatidylserines which are translocated to the outer leaflet of the cell membrane during apoptosis. CA at 500μM concentration caused apoptosis in WT cholangiocytes, which was further exacerbated in Abcc12+/− and Abcc12−/− cholangiocytes (Supplemental Figure S9B). To determine the impact of Abcc12 on bile acid induced cholangiocyte injury in vivo, we isolated cholangiocytes from mice after 14 days of CA feeding. The rate of apoptosis by Annexin V staining was significantly increased in biliary epithelial cells from the heterozygous and homozygous mutants compared with WT mice (Figure 6C). Thus, targeted deletion of Abcc12 in mice supports the phenotype of MRP9 loss as observed in the index patient and zebrafish mutant. Partial and complete loss of MRP9 renders bile duct epithelium susceptible to bile acid-induced apoptosis and results in bile duct loss.

Figure 6. MRP9 deficiency conferred susceptibility to bile acid-induced cholangiocyte injury in mice.

(A) Adult WT, Abcc12+/−, and Abcc12−/− male mice were fed with chow containing 1% CA for 14 days. Weight change from baseline, end of treatment liver weights, liver/ body weight ratios, serum total bilirubin and GGT levels are displayed. (B) Liver sections post CA treatment were subjected to immunohistochemistry with antibody against CK19 and analyzed for bile duct profiles. Well-formed bile ducts per portal triad (arrows) were enumerated for 820 portal triads in WT (n=6 mice), 335 in Abcc12+/− mice (n=3), and 407 in Abcc12−/− mice (n=3). (C) Cholangiocytes were isolated from mice after 14 days of CA feeding by percoll gradient centrifugation followed by EpCam positive selection. Apoptotic cells were enumerated by flow cytometry based on positive staining for Annexin V and 7AAD. Statistical significance in (A-C) was calculated by one-way ANOVA with Tukey’s multiple comparison test and adjusted p value of *, p<0.05; **, p<0.005; ***, p<0.0001. In (A) and (C), each dot represents result from individual animal.

Discussion

Applying WES and targeted gene sequencing to 93 pediatric patients with low GGT chronic cholestasis, we connect the cholestatic phenotype to pathogenic variants in ABCC12 for the first time. We find its protein, the ABC transporter MRP9, to be expressed in the cholangiocytes in human, mice, and zebrafish. Collective evidence from a patient with MRP9 deficiency and two animal models demonstrates that lack of MRP9 confers susceptibility to loss of intrahepatic bile duct epithelium. Analyses of the zebrafish and mouse mutants associate bile duct paucity with increased apoptotic and bile acid-induced cell death.

Chronic cholestasis due to bile duct loss or cholangiocyte injury, such as Alagille syndrome ⁴² and MDR3 deficiency ⁴³, is normally associated with elevated serum GGT levels. In contrast, in conditions like BSEP- and FIC1-deficiency that result in decreased canalicular export of bile acids from hepatocytes, cholangiocytes are not injured in the patients and serum GGT levels do not typically rise above normal limits or are even lower than in healthy controls ^{2, 5}. Our patient and mouse studies, however, demonstrate that bile duct paucity also occurs in low GGT cholestasis. ABCC7 encodes CFTR that is another member of the ABCC superfamily and also expressed in cholangiocytes. Pathogenic variants in ABCC7 are associated with cystic fibrosis liver disease, progression of which is often accompanied by no or mild elevation of GGT ⁴⁴. We propose that MRP9 deficiency is a cholangiopathy leading to programmed cell death of cholangiocytes. Investigation into the impact of MRP9 deficiency on bile composition and homeostasis may provide insights into why it presents with low GGT cholestasis.

How the heterozygous missense variants identified in our study contribute to the liver phenotype in each subject requires further characterization. We show that CA feeding results in significantly higher serum bilirubin levels in Abcc12+/− mice compared with WT mice. Cholangiocytes from Abcc12+/− mice display increased susceptibility to CA-induced apoptosis compared to WT cholangiocytes. These results suggest haploinsufficiency for MRP9. It is noteworthy that in addition to the full-length transcript, we detected a second ABCC12 transcript in the control human liver that was predicted to produce two smaller proteins. Determining the cell type-specific expression of different ABCC12 transcripts is crucial to delineate the pathogenicity of each non-synonymous variant.

Our animal studies provide important clues on the mechanisms driving liver disease in MRP9 deficiency. In zebrafish, the onset of bile duct loss coincides with sex differentiation and the liver phenotypes are only observed in the male abcc12 mutants. The higher expression of abcc12 transcript in the male liver compared to the female liver may contribute to the male sex predilection for biliary injury in abcc12 mutants. Murine studies previously showed that female sex hormones promote proliferation of estrogen-receptor expressing cholangiocytes ^{45, 46}, representing a plausible mechanism to compensate for the loss of biliary cells from MRP9 deficiency. We demonstrate that primary bile acids found in high concentrations in bile induce MRP9 expression in murine cholangiocytes in vitro, raising the possibility that MRP9 is part of the cellular machinery defending biliary epithelial cells against the cytotoxic effects of hydrophobic bile acids. Both zebrafish and mouse studies indicate that MRP deficiency renders cholangiocytes susceptible to apoptosis. We hypothesize that it could be due to accumulation of toxic compounds that require biliary excretion via MRP9, or compromised cholangiocyte cell function or cell membrane integrity due to requirement of MRP9 unrelated to its transporter function.

There are several implications of our experimental findings and related questions for future research: 1) we identify ABCC12 as a novel candidate gene for chronic cholestasis syndromes. It will be important to validate our findings in an external cohort of both low and high GGT cholestasis to estimate the contribution of pathogenic variants of ABCC12 to chronic liver disease with various phenotypes and perform phenotype-genotype correlation studies. 2) Since our patients were all diagnosed with liver disease prior to onset of puberty in a cross-sectional study, it is impossible to determine at this point whether the sex dismorphism observed in zebrafish is relevant to the human disease. Based on the information from our model organisms, we predict MRP9-associated liver disease to improve after onset of puberty in females, and adult males presenting with idiopathic cholestasis to be more likely to harbor deleterious variants in ABCC12 compared with females. 3) Studies in animal models linked MRP9 deficiency to cholangiocyte apoptosis. Future studies may examine the role of MRP9 as genetic modifier or therapeutic target in hepatobiliary diseases in which cholangiocyte apoptosis is dysregulated, for instance, in cholangiocarcinoma, vanishing bile duct syndrome following drug-induced liver injury, or in the apoptosis-resistant phenotype of senescent cholangiocytes in PSC ⁴⁷

What You Need to Know:

BACKGROUND AND CONTEXT:

We surveyed cohorts of children with idiopathic chronic cholestasis with whole exome and candidate gene sequencing to identify novel genetic susceptibility factors for cholestatic liver diseases.

NEW FINDINGS:

A homozygous deleterious variant in ABCC12, encoding the ABC-cassette protein MRP9, was detected in a patient with low-GGT cholestasis and bile duct paucity. MRP9 deficiency renders cholangiocytes susceptible to bile acid-induced apoptosis in zebrafish and mouse models.

LIMITATIONS:

The genetic studies were restricted to children with low GGT cholestasis. Future studies ought to examine the role of MRP9 in patients with acquired cholangiopathies like biliary atresia or primary sclerosing cholangitis.

IMPACT:

ABCC12 represents a novel candidate gene for inherit chronic cholestasis and a potential therapeutic target for progressive intrahepatic bile duct loss.

Abbreviations used in the paper:

4-OHT

4-Hydroxytamoxifen

ABC

adenosine triphosphate-binding cassette

ACMG

American College of Medical Genetics

ALP

alkaline phosphatase

ALT

alanine aminotransferase

Annexin A4

Anxa4

AST

aspartate aminotransferase

BSEP

bile salt export pump

CA

cholic acid

CADD

Combined Annotation Dependent Depletion

CCHMC

Cincinnati Children’s Hospital Medical Center

CGamF

cholylglycylamidofluorescein

ChiLDReN

Childhood Liver Disease Research Network

dpf

days post fertilization

EdU

5’-ethynyl-2’-deoxyuridine

EtOH

ethanol

GGT

γ-glutamyl transferase

H&E

Hematoxylin and eosin

hpf

hours post fertilization

MAF

minor allele frequency

MRP

multidrug resistance-associated protein

MW

molecular weight

NGS

next generation sequencing

NIDDK

National Institute of Diabetes and Digestive and Kidney Diseases

PFIC

progressive familial intrahepatic cholestasis

PSC

primary sclerosing cholangitis

qPCR

quantitative real-time PCR

TUNEL

Terminal deoxynucleotidyl transferase dUTP nick end labeling

VOUS

variant of uncertain significance

WES

whole-exome sequencing

wk

week

wt

wild-type