Molecular mechanistic explanation for the spectrum of cholestatic disease caused by the S320F variant of ABCB4


  • Edward J. Andress,

    1. Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary, University of London, London, UK
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Michael Nicolaou,

    1. Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary, University of London, London, UK
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Marta R. Romero,

    1. Laboratory of Experimental Hepatology and Drug Targeting, CIBERehd, IBSAL, University of Salamanca, Spain
    Search for more papers by this author
  • Sandhia Naik,

    1. Department of Paediatric Gastroenterology, Barts and the London Children's Hospital, London, UK
    Search for more papers by this author
  • Peter H. Dixon,

    1. Women's Health Academic Centre, King's College London, London, UK
    Search for more papers by this author
  • Catherine Williamson,

    1. Women's Health Academic Centre, King's College London, London, UK
    Search for more papers by this author
  • Kenneth J. Linton

    Corresponding author
    1. Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary, University of London, London, UK
    • Address reprint request to: Professor Kenneth J. Linton, Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary, University of London, 4 Newark Street, Whitechapel, London UK E1 2AT. E-mail:; fax +44 (0)20 7882-7172.

    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

  • E.J.A. was funded by Barts and the London Charity award 458/1495. M.N. was funded by a Medical Research Council centenary award. M.R.R. was supported by funding from the Spanish Ministry of Science (Grant SAF2010-15517). The groups of K.J.L. and C.W. are supported by the Medical Research Council, UK (MC_U120088463) and Imperial College Healthcare NHS trust biomedical research centre, respectively.


ABCB4 flops phosphatidylcholine into the bile canaliculus to protect the biliary tree from the detergent activity of bile salts. Homozygous-null ABCB4 mutations cause the childhood liver disease, progressive familial intrahepatic cholestasis, but cause and effect is less clear, with many missense mutations linked to less severe cholestatic diseases. ABCB4S320F, in particular, is described in 13 patients, including in heterozygosity with ABCB4A286V, ABCB4A953D, and null mutants, whose symptoms cover the spectrum of cholestatic disease. We sought to define the impact of these mutations on the floppase, explain the link with multiple conditions at the molecular level, and investigate the potential for reversal. ABCB4S320F, ABCB4A286V, and ABCB4A953D expression was engineered in naïve cultured cells. Floppase expression, localization, and activity were measured by western blot, confocal microscopy, and lipid transport assays, respectively. ABCB4S320F was fully active for floppase activity but expression at the plasma membrane was reduced to 50%. ABCB4A286V expressed and trafficked efficiently but could not flop lipid, and ABCB4A953D expressed poorly and was impaired in floppase activity. Proteasome inhibition stabilized nascent ABCB4S320F and ABCB4A953D but did not improve plasma membrane localization. Cyclosporin-A improved plasma membrane localization of both ABCB4S320F and ABCB4A953D, but inhibited floppase activity. Conclusion: The level of ABCB4 functionality correlates with, and is the primary determinant of, cholestatic disease severity in these patients. ABCB4S320F homozygosity, with half the normal level of ABCB4, is the tipping point between more benign and potentially fatal cholestasis and makes these patients more acutely sensitive to environmental effects. Cyclosporin-A increased expression of ABCB4S320F and ABCB4A953D, suggesting that chemical chaperones could be exploited for therapeutic benefit to usher in a new era of personalized medicine for patients with ABCB4-dependent cholestatic disease. (Hepatology 2014;59:1921–1931)


ATP binding cassette




drug-induced cholestasis


intrahepatic cholestasis of pregnancy


low phospholipid-associated cholelithiasis




progressive familial intrahepatic cholestasis

Bile flow across the canalicular membrane of hepatocytes is dependent on three primary-active transport proteins, the ATP binding cassette (ABC) transporters ABCB11[1, 2] and ABCB4[3] and the P-type ATPase ATP8B1[4] (formerly known as BSEP, MDR3, and FIC1, respectively). The emerging picture is that ABCB4 and ATP8B1 (which functions in complex with its accessory protein CDC50) function to protect the hepatocyte (and cholangiocytes that line the bile ducts) from the deleterious detergent activity of bile salts.[5] ABCB4 is thought to flop the membrane lipid phosphatidylcholine (PC) from the inner to the outer leaflet of the canalicular membrane,[3, 6, 7] from where it is extracted into the canaliculus by bile salts transported by ABCB11 to form mixed micelles. This reduces the detergent activity of the bile salts.[3, 8] In contrast, ATP8B1/CDC50 flips the aminophospholipids, phosphatidylserine, and phosphatidylethanolamine from the outer to the inner leaflet of the membrane.[9] This activity is thought to maintain lipid asymmetry of the canalicular membranes of hepatocytes and cholangiocytes, to resist the detergent activity of bile salts.[5, 10] All three transporters are known to be critical for bile flow because null mutations in ATP8B1, ABCB11, and ABCB4 cause progressive familial intrahepatic cholestasis (PFIC) types 1, 2, and 3, respectively. These monogenetic, autosomal recessive, diseases are characterized by intrahepatic cholestasis within the first year of life which leads to severe growth retardation. Progression to liver failure typically occurs within the first two decades and is often only treatable by orthotopic liver transplantation.[7]

Dysfunction of each of the transporters is also considered causative in less severe but nevertheless debilitating complex liver disorders. Of relevance to the current study, ABCB4 single nucleotide polymorphisms (SNPs) have been linked to drug-induced cholestasis (DIC[11]), intrahepatic cholestasis of pregnancy (ICP[12]), and low phospholipid-associated cholelithiasis (LPAC[13]). Recently, ABCB4 SNPs have also been suggested to predispose to cholangiocarcinoma.[14] In the rare type-3 PFIC the link between mutation and disease is clear, but the majority of cholestatic liver diseases are complex conditions and a causative link with ABCB4 SNPs that are typically private to individual families is suspected but not proved, with critical evidence at the functional and mechanistic level of protein expression and floppase activity lacking.

Biochemical characterization of the effect of ABCB4 SNPs directly has, hitherto, been challenging for several reasons: it is technically difficult to measure translocation of PC from the inner to the outer leaflet of the plasma membrane (this has been achieved previously but interpretation is complicated by the use of fluorescently tagged PC analogs,[6, 15] and expression of functional PC floppase, transiently in vitro, is deleterious to cultured cells.[5] Some insight has been obtained by mimicking mutations in the multidrug resistance transporter ABCB1, the closest homolog of ABCB4.[16, 17] However, this approach is only relevant if the particular amino acid is conserved in both proteins and performs the same function, ruling out the study of mutations that directly influence PC binding, or responses to hormones of pregnancy, contraceptives, and other drugs. Recently, we described the transient expression of active ABCB4 in naïve HEK293T cells by coexpressing it with ATP8B1/CDC50 to maintain the integrity of the plasma membrane.[5] This ameliorates the cytotoxic effect of ABCB4 function. We now report the use of this expression system to characterize variant ABCB4 and describe the relationship between genotype and phenotype in a cohort of nonfamilial patients who share the S320F mutation and who presented with a spectrum of hepatobiliary disease (Table 1).

Table 1. Missense Mutations in 14 Patients With Cholestatic Liver Disease Who Carry ABCB4S320F and Related Variants
Patient1st allele2nd alleleSexAdditional MutationsDiseaseReference
  1. Patient 12 is nulligravid and also carries an SNP in the FXR gene which results in the missense mutation M1V and lowers the abundance of FXR by two-thirds.42 The bile acid sensor FXR controls transcription of ABCB4 and ABCB11, therefore low levels of FXR in patient 12[29] is likely to lower ABCB4 abundance further, but in parallel with ABCB11.

  2. *,Siblings.

  3. Drugs in these cases refer to prescription of the contraceptive pill or progesterone fertility treatment. NK: not known.

1S320FY279XFNKPFIC3Degiorgio et al.[25]
2S320FA286VFNKLPAC to PFIC3Degiorgio et al.[25]
3*S320FA953DMABCB11-V444A+/-LPAC to PFIC3Poupon et al.[27]
4*S320FA953DFABCB11-V444A+/-LPAC to PFIC3Poupon et al.[27]
5S320FS320FMNKPFIC3Colombo et al.[23]
6S320FS320FNKNKLPACRosmorduc et al.[38]
7S320FS320FFNKICPRosmorduc et al.[13]
8S320FS320FFNKDIC and ICPRosmorduc et al.[13]
9S320FS320FFNKICPPauli-Magnus et al.41
10S320FWTFNKICPBacq et al.[28]
11S320FWTFABCB11-wt+/+LPAC and ICPZimmer et al.[29]
12S320FWTFABCB11-V444A+/-LPACZimmer et al.[29]
13S320FWTFABCB11-V444A+/+DIC and ICPKeitel et al.[30]
14A953DA953DNKNKPFIC3Keitel et al.[24]

In the current study, we focus on ABCB4S320F and other mutations (ABCB4A286V and ABCB4A953D) that are found in heterozygosity with ABCB4S320F, and characterized their impact on protein expression, plasma membrane localization, and floppase function. The S320F variant is particularly interesting because it is described in 13 case studies and is linked with the development of cholestatic disorders which unusually cover the whole spectrum of severity including several ICP, LPAC DIC, and PFIC3 cases. The distinct in vitro phenotypes correlate with the severity of disease confirming that ABCB4 genotype is the primary determinant of cholestatic pathophysiology. The phenotypes of ABCB4S320F and ABCB4A953D, but not ABCB4A286V, suggest a translational path to therapeutic intervention.

Materials and Methods


Plasmids pcDNA3-ABCB4, pcDNA3-ABCB4E558Q, pCIneo-ATP8B1, and pCIneo-CDC50 were described previously.[5] Site-directed mutagenesis to introduce the SNPs c.857C>T (p.A286V), c.959C>T (p.S320F), and c.2858C>A (p.A953D) into the wild-type ABCB4 complementary DNA (cDNA)[18] was performed using QuikChange-II (Stratagene, La Jolla, CA). Mutagenesis was verified by sequencing of the entire cDNA and promoter region. Mutagenic oligonucleotides (5′-3′): A286V, GAAAGGTATCAGAAACATTTAGAAAATGTCAAAGAGATTGGAATTAAAAAAGCTATT; S320F, CTGGCCTTCTGGTATGGATTCACTCTAGTCATATCAAA; A953D, GCATTTATGTATTTTTCCTATGACGGTTGTTT TCGATTTGGTGCA.

Mammalian Cell Culture

Human embryonic kidney (HEK293T) cells were cultured and transiently transfected as described previously.[5] Cells were seeded 24 hours prior to transfection at a density of 4.6 × 104/cm2 and triple-transfected with 2.5 μg each of plasmids encoding ABCB4, ATP8B1, and CDC50 using polyethyleneimine (Sigma, Gillingham, UK). Where ATP8B1 and CDC50 were omitted, DNA concentration was maintained constant with 5 μg of empty pCIneo. Where applied, MG132 or cyclosporin A (CsA) solubilized in dimethylsulphoxide (DMSO) was added 24 hours posttransfection. ABCB4 localization and function was analyzed 48 hours posttransfection.

Western Analysis

Cells were washed twice with phosphate-buffered saline (PBS) and harvested in 150 mM NaCl, 20 mM HEPES pH 7.4, 1% SDS, 1× EDTA-free complete protease inhibitor cocktail from Roche, 1 mM PMSF. Crude protein lysates (1 μg) were denatured in Laemmli sample buffer (70°C for 5 minutes), separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and transferred to polyvinylidene difluoride membrane (Millipore, Billerica, MA). Blots were probed with mouse anti-ABCB4 monoclonal P3II-26 (Sigma), diluted 1:2,000, and mouse anti-β-tubulin (SourceBioscience, Nottingham, UK) diluted 1:2,000. Primary antibodies were detected using goat antimouse secondary antibody, conjugated to horseradish peroxidase diluted 1:2,000 (Santa Cruz, Dallas, TX) and visualized by enhanced chemiluminescence (GE Healthcare, Amersham, UK), or goat antimouse, conjugated to IRDye 680 diluted 1:20,000 (LI-COR, Lincoln, NE), for visualization by the Odyssey scanning system. ABCB4 abundance normalized against β-tubulin was analyzed using ImageJ software.

Confocal Microscopy

Cells were seeded onto 12-mm glass coverslips coated with poly-L-lysine, 24 hours prior to triple-transfection. Cells were fixed 48 hours posttransfection in ice-cold 10% acetone (in ethanol) for 20 minutes, blocked in 1% bovine serum albumin (BSA; in PBS) for 1 hour, and incubated with primary antibodies, mouse anti-ABCB4 (P3II-26; Sigma), and rabbit anti-Na+/K+ ATPase (H-300; Santa Cruz) at 1:50 dilution in PBS + 1% BSA for 1 hour. The cells were washed 3× in PBS and incubated with secondary antibody (goat antimouse Alexa568 or goat antirabbit Alexa488; Invitrogen, Paisley, UK), diluted 1:500 in PBS + 1% BSA for 1 hour. The labeled cells were washed 3× in PBS, mounted on glass slides, and imaged using a Zeiss 710 confocal microscope. Images were obtained using a plan-apochromat 63× oil immersion objective with a numerical aperture of 1.4.

Phosphatidylcholine Efflux Assay

PC extraction was measured as described previously.[5] Briefly, cells were triple-transfected in poly-L-lysine-coated 6-well dishes and fed 2 μCi (∼24 pmol) 3H-choline chloride (Perkin-Elmer, Waltham, MA) 24 hours posttransfection. The cells were washed 24 hours later in prewarmed Dulbecco's modified Eagle's medium (DMEM) and incubated with 2 mM sodium taurocholate hydrate (TC) in DMEM. Media was recovered and cellular lysates prepared. Radionuclide content was measured by liquid scintillation. PC extraction is calculated as radionuclide detected in the media as a percentage of the total (media plus lysate). The data were normalized against PC extraction from cells expressing the nonfunctional E558Q Walker B mutant of ABCB4.


Expression Levels of Variant ABCB4

Mutations were introduced into the ABCB4 cDNA to mimic the SNPs observed in patients and to encode the A286V, S320F, and A953D variant proteins. HEK293T cells were transiently transfected with either a single plasmid encoding an ABCB4-variant or three plasmids encoding the ABCB4-variant plus ATP8B1 and CDC50 (the pCIneo-ABCB4 plasmids were kept constant at 2.5 μg throughout). The transfection conditions were optimized to achieve a confluent monolayer of cells 48 hours posttransfection. Sixty percent of the cells in the population were transfected and 98% of the transfected cells took up all three plasmids (Supporting Fig. 1). Expression of each variant protein was compared to wild-type ABCB4 and a nonfunctional mutant carrying a point mutation in the Walker B motif of the first nucleotide-binding domain, ABCB4E558Q (Fig. 1). The wild-type transporter expressed poorly in single-transfected cells likely due to the deleterious effect of flopping excess PC into the outer leaflet of the plasma membrane, but expressed well when triple-transfected with ATP8B1/CDC50.5 In contrast, the nonfunctional Walker B mutant ABCB4E558Q expressed equally well under both conditions. The three variants had distinct expression profiles. ABCB4A286V was similar to the Walker B mutant and expressed to high levels in both the single and triple-transfected cells. ABCB4A953D expressed poorly in both conditions and to only 16.3% of the wild-type level in the triple-transfected cells. ABCB4S320F also expressed to similar levels in the single- and triple-transfected cells, but to 50.1% of the level achieved by the wild-type transporter in the triple-transfected cells.

Figure 1.

Analysis of expression of ABCB4 missense variants. (A) Western blot of the expression of wild-type and variant ABCB4 in HEK293T cells in the absence and presence of ATP8B1/CDC50 as indicated (black vertical line indicates samples were run on the same gel but were noncontiguous). The Walker B mutant E558Q was used as a control for expression of a nonfunctional protein. (B) Expression levels of the mature 160-kDa form of the ABCB4 variants in different biological replicates were normalized to wild-type ABCB4 expression expressed contemporaneously in the absence of ATP8B1/CDC50. Mean expression level ± SEM are plotted. Statistical analysis was by unpaired Student t test (n ≥ 3; ***P < 0.005; ns, not significantly different).

ABCB4S320F and ABCB4A286V Traffic to the Plasma Membrane While ABCB4A953D Is Largely Retained Intracellularly

Triple-transfected cells were fixed, stained, and analyzed by confocal microscopy to determine whether the variant proteins could traffic to the plasma membrane (Fig. 2). Wild-type ABCB4 traffics efficiently to reside in the plasma membrane as shown by the substantial colocalization with the Na+/K+-ATPase. Significant levels of ABCB4A286V and ABCB4S320F were also detected in the plasma membrane. ABCB4A953D, despite a low level of expression, was also detected, but predominantly in the intracellular compartment with only a small fraction at the plasma membrane. This is consistent with the low level of expression of the mature form of ABCB4A953D observed by western analysis.

Figure 2.

Trafficking of variant ABCB4 to the plasma membrane. Wild-type and variant ABCB4 were expressed in HEK293T cells in the presence of ATP8B1/CDC50. Cells were fixed, permeabilized, and stained for ABCB4 (magenta) and the plasma membrane protein, Na+/K+-ATPase (green). Nuclei were stained with DAPI (blue) and the cells imaged by confocal microscopy. Scale bar = 20 μm.

ABCB4S320F Is Fully Active, ABCB4A953D Is Partially Active, and ABCB4A286V Is Inactive for the Efflux of PC

Triple-transfected cells were fed radiolabeled choline to convert into 3H-PC.[19] The cells were incubated with taurocholate to extract the flopped PC from the outer leaflet of the plasma membrane. More 3H-PC is extracted from cells that express wild-type ABCB4 than those that express the catalytically inactive ABCB4E558Q (Fig. 3). 3H-PC extracted from cells expressing ABCB4S320F was reduced to 49% that of cells expressing wild-type ABCB4, while the level mediated by cells expressing ABCB4A286V or ABCB4A953D was very low. After correction for ABCB4 expression level (Fig. 3, inset) the data indicate that ABCB4S320F is fully active as a floppase, while ABCB4A286V is unable to flop PC. ABCB4A953D retains partial functionality with a floppase activity 29.4% that of the wild-type level after correction for the low level of mature protein expressed.

Figure 3.

Phosphatidylcholine efflux by ABCB4 variants in the presence of ATP8B1/CDC50. PC floppase activity was tested in HEK293T cells expressing wild-type ABCB4 (WT), ABCB4S320F, ABCB4A286V, or ABCB4A953D in the presence of ATP8B1/CDC50. 3H-PC extracted in the presence of 2 mM TC was calculated as the percentage of total cellular radioactivity after subtraction of the background level from cells expressing the Walker B mutant ABCB4E558Q. The data were analyzed by Student t test (n ≥ 3; *P < 0.01, ***P < 0.005). Inset shows percentage mean PC floppase activity after correction for ABCB4 variant mean expression levels.

Proteasome Inhibition Causes Accumulation of Immature ABCB4S320F and ABCB4A953D but Does Not Improve Trafficking to the Plasma Membrane

Membrane proteins that fold inefficiently are targeted for degradation by the proteasome by way of the endoplasmic reticulum-associated degradation (ERAD) pathway.[20, 21] Inhibition of the proteasome could therefore, potentially, slow the rate of degradation and allow the variant proteins to adopt the appropriate fold. As a test of this hypothesis, triple-transfected cells expressing either wild-type ABCB4, ABCB4S320F, or ABCB4A953D were treated with the proteasome inhibitor MG132 (10 μM) from 24 hours posttransfection. Western analysis showed that these cells accumulated a smaller molecular weight species of ABCB4 (Fig. 4A). This species is not glycosylated and migrates with the same electrophoretic mobility as deglycosylated ABCB4 (Fig. 4B); therefore, most likely represents full-length, immature protein which accumulates in the ER (Fig. 5A-C). Inhibition of the proteasome appears to have little impact on maturation of ABCB4 because there is no increase in abundance of the mature, fully glycosylated protein. These data indicate that ABCB4S320F, ABCB4A953D, and also wild-type ABCB4 are subject to ERAD but that inhibition of the proteasome, by itself, does not improve abundance at the plasma membrane in vitro and is not likely to improve ABCB4 density in the canalicular membrane of patients.

Figure 4.

Effects of potential modulators of ABCB4 expression and function. (A) Cells were triple-transfected to express wild-type ABCB4 (WT), ABCB4S320F, or ABCB4A953D. Treatment with the proteasome inhibitor MG132 (10 μM) 24 hours posttransfection induced accumulation of a smaller molecular weight form of the floppase, detected by western analysis. (B) Whole-cell lysates treated with PNGase F shows that deglycosylated mature protein migrates with the same mobility as the smaller molecular weight form produced by treatment with MG132 (loading was adjusted to ensure similar levels of variant and wild-type protein). (C) Representative gels and bar graph showing that treatment of triple transfected cells with CsA 24 hours posttransfection reproducibly causes the mature wild-type ABCB4 and ABCB4S320F to accumulate, but has no effect on the abundance of ABCB4A953D. Expression levels of the mature protein are normalized against β-tubulin expression. (D) CsA inhibits the PC floppase activity of wild-type ABCB4. Triple-transfected cells were incubated with 0-1 μM CsA 24 hours posttransfection for the indicated times. The media was washed from the cells and replaced with fresh media containing TC. Data were analyzed by Student t test (n ≥ 3; *P < 0.01, ***P < 0.005). Black vertical line(s) on panels B and C indicate that samples were run on the same gel but were noncontiguous.

Figure 5.

Effect of MG132 and CsA on ABCB4 expression and trafficking. (A) Wild-type ABCB4. (B) ABCB4S320F. (C) ABCB4A953D. Cells were cultured on glass coverslips, triple-transfected, and treated 24 hours posttransfection with the proteasome inhibitor MG132 (10 μM) or the chemical chaperone CsA (1 μM) as indicated. The cells were fixed, stained, and imaged by confocal microscopy 48 hours posttransfection. Staining as per Fig. 2. Scale bar = 20 μm.

Cyclosporin-A Improves Folding of ABCB4 but Inhibits PC Efflux

CsA can improve the folding kinetics of the drug efflux pump ABCB1.22 Incubation of triple-transfected cells with CsA 24 hours posttransfection caused a marked increase in the expression of ABCB4S320F and also wild-type ABCB4 (Fig. 4C). Even at low micromolar concentrations of CsA, ABCB4S320F can attain the expression level observed for the wild-type protein in untreated cells. In cells expressing wild-type ABCB4 and ABCB4S320F the improved level of protein expression resulted in increased localization at the plasma membrane (Fig. 5A,B, lower panels). The expression level of ABCB4A953D was not increased by CsA (Fig. 4C), but the treatment did improve localization of this variant to the plasma membrane (Fig. 5C). Unfortunately, CsA treatment from 24 hours posttransfection was found to inhibit ABCB4 PC floppase activity (Fig. 4D). This was apparent even if the treated cells were washed extensively 48 hours posttransfection and CsA was omitted from the transport buffer.


We describe the characterization of three missense variants of ABCB4 that are linked to a spectrum of cholestatic liver disease in 14 case studies. The variants were expressed in HEK293T cells which do not polarize but offer two key advantages: they do not express endogenous ABCB4 and they are highly transfectable, which is necessary for efficient triple-transfection with ATP8B1/CDC50 to avoid negative selective pressure acting on PC floppase activity. Taurocholate was added to the cells to extract the 3H-PC flopped by ABCB4 (we showed previously that there is no cellular efflux in the absence of an extracellular acceptor[5]). ABCB4A286V expressed and localized efficiently to the plasma membrane in vitro but could not efflux PC. Expression of ABCB4S320F and ABCB4A953D was reduced to 50.1% and 16.3% of the wild-type floppase, respectively, and while ABCB4S320F trafficked to the plasma membrane, a large fraction of ABCB4A953D was retained intracellularly. Efflux of PC from cells expressing ABCB4S320F and ABCB4A953D (respectively, 49% and 5% of the level from cells expressing wild-type ABCB4), which when corrected for floppase abundance at the plasma membrane indicates that ABCB4S320F is fully functional while ABCB4A953D retains only 29.4% floppase activity. The lower density of this functional ABCB4S320F in the plasma membrane, and its insensitivity to the ATP8B1/CDC50 status of the cells suggests that HEK293T cells have a threshold level of PC floppase activity that they can tolerate. This is not surprising because the lipids of all cells are asymmetrically arranged in the plasma membrane and therefore all cells are likely to have flippases and floppases for specific lipids to maintain asymmetry. We expect that this homeostatic mechanism is capable of withstanding the imbalance caused by ABCB4S320F expression. The in vitro data appear consistent with the available expression and immunohistochemistry data from clinical samples (albeit from chronically diseased liver). Patient 5 (Table 1), who is homozygous for ABCB4S320F is reported to express 50% of the normal level of ABCB4,[23] and patient 14, who is homozygous for ABCB4A953D, shows a marked reduction in expression (although this was not quantified by the authors[24]). This indicates that our in vitro system is a good model of the in vivo situation and can be used to characterize the impact of missense SNPs on ABCB4 protein expression, localization, and function.

The in vitro phenotype of the variant floppases correlates with the severity of disease, confirming that ABCB4 genotype is the primary determinant of cholestatic pathophysiology in this cohort. The ABCB4S320F variant is common to the first 13 patients described in Table 1. In patients 1 and 2,[25] ABCB4S320F is heterozygous with either ABCB4Y279X (which, truncated to only a quarter of the polypeptide, would be unable to form a functional ABC transporter[26]) or the nonactive ABCB4A286V, and therefore ABCB4S320F would be the only functional ABCB4 PC floppase present in the canalicular membrane. Extrapolating from the in vitro data indicates that less than one-quarter of the normal activity of the PC floppase is unable to prevent damage to canalicular membranes and results in the severe childhood disease PFIC3. The critical effect of the level of ABCB4 expression and function is also supported by the two siblings described by Poupon et al.[27] (patients 3 and 4; Table 1), who are heterozygous for ABCB4S320F and ABCB4A953D and so are likely to have only 27% of the PC floppase activity of a normal individual. These patients presented with LPAC characterized by cholesterol gallstones which precipitate from the bile due to the low solubility of cholesterol in mixed micelles that are low in PC.[13] Unusually for LPAC patients, they developed PFIC3 in the third decade of life, suggesting that even a small increase in PC-floppase activity (compared to patients 1 and 2) can delay the onset of this progressive disease.

In cases 10-13 (Table 1), ABCB4S320F is heterozygous with the wild-type allele and these patients present with LPAC, DIC or ICP.[28-30] In DIC and ICP, the cholestasis is induced and limited by drugs (the contraceptive pill and fertility treatment in the two cases described) or pregnancy hormones which are thought to inhibit the transporters at the canalicular membrane directly[31] and/or reduce their levels of expression by direct inhibition of the farnesoid-X receptor.[32] The allelic combination of patients 10 to 13 (where ABCB4S320F is combined with the wild-type allele) is likely to encode three-quarters of the normal level of ABCB4 floppase activity, which we would argue is sufficient to protect against childhood PFIC3 but insufficient to prevent LPAC, DIC, or ICP. It has been suggested previously that relation of genotype to phenotype in patients 11 to 13 may be complicated by additional variation at the ABCB11 locus. The ABCB11V444A variant is reported to express poorly compared to the wild-type allele both in vitro and in vivo[30, 33] and is likely, therefore, to reduce bile salt efflux across the canalicular membrane. Population studies have identified ABCB11V444A as a risk factor for several forms of cholestasis.[30, 34-37] However, reduced bile salt efflux would be likely to reduce the toxicity of the bile in the canaliculae and therefore may be expected to alleviate the pathology of disease in patients with low levels of ABCB4. Patient 11 is homozygous for the wild-type ABCB11 allele, patient 12 is heterozygous, and patient 13 is homozygous for ABCB11V444A, but all present with biliary conditions of similar severity, suggesting that the ABCB11 genotype is of little relevance in these cases.

The tipping point for development of PFIC3 in childhood would appear to be half the normal level of functional ABCB4. Patients 5 to 9 are homozygous for ABCB4S320F and so would be expected to express half the normal level of ABCB4. These patients presented with the full spectrum of cholestatic disease, including LPAC (patient 6),[38] ICP (patients 7 and 9),[13, 39] DIC and ICP (patient 8),[13] and one patient (patient 5)[23] who presented with PFIC3 aged 8. Patient 5 responded to treatment with the less toxic bile acid, ursodeoxycholate, but developed liver fibrosis and gallstones by age 10. At the tipping point of ABCB4 functionality, these patients are likely to be more acutely sensitive to environmental affects such as therapeutic drug intake, hormones, and diet which may modulate liver disease severity. A similar threshold effect has also been postulated for rodent models of ABCB4-dependent cholestatic disease.[40]

None of these SNPs are close to an exon/intron boundary and while an effect on message stability cannot be ruled out, the close correlation of the in vitro cell biological data with patient symptoms suggest that the phenotypic effects are mediated at the protein level. This implies that relatively subtle changes in ABCB4 abundance and activity can affect bile composition and the progression of disease. The corollary, that relatively subtle improvement in PC flopping is likely to have significant therapeutic benefit, make ABCB4 an attractive target for intervention. The defects associated with ABCB4S320F and ABCB4A953D but not ABCB4A286V may be correctable by agents that improve the ABCB4 protein level at the canalicular membrane. Treatment with CsA produced a marked improvement in abundance and plasma membrane localization of both wild-type ABCB4 and ABCB4S320F, and also in the localization of ABCB4A953D at the plasma membrane, but it was found to inhibit PC floppase activity. CsA is thought to act as a chemical chaperone for the drug efflux pump ABCB1 by binding directly to the nascent protein and improving folding kinetics by induced fit.[22] The effects of CsA on the abundance, localization, and floppase activity suggests a similarly direct interaction with ABCB4. Despite the inhibitory effect, the CsA data provide proof-of-principle that the density of ABCB4S320F and ABCB4A953D at the plasma membrane can be manipulated, and that there is scope for development of drugs to target increased expression of the PC floppase at the canalicular membrane. To date, 127 nonsynonymous SNPs in ABCB4 have been identified ( In vitro characterization will be key to stratify these missense variants of ABCB4 to clarify the relationship between genotype and phenotype, and to identify those disease-causing variants that could be targeted for therapeutic intervention.


We thank Capucine Grandjean and Theodore Sanders for technical assistance with triple fluorescence flow cytometry.