Atp8b1 deficiency in mice reduces resistance of the canalicular membrane to hydrophobic bile salts and impairs bile salt transport

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Progressive familial intrahepatic cholestasis type 1 (PFIC1, Byler disease, OMIM 211600) is a severe inherited liver disease caused by mutations in ATP8B1. ATP8B1 is a member of the type 4 subfamily of P-type ATPases, which are phospholipid flippases. PFIC1 patients generally develop end-stage liver disease before the second decade of life. The disease is characterized by impaired biliary bile salt excretion, but the mechanism whereby impaired ATP8B1 function results in cholestasis is unclear. In a mouse model for PFIC1, we observed decreased resistance of the hepatocanalicular membrane to hydrophobic bile salts as evidenced by enhanced biliary recovery of phosphatidylserine, cholesterol, and ectoenzymes. In liver specimens from PFIC1 patients, but not in those from control subjects, ectoenzyme expression at the canalicular membrane was markedly deficient. In isolated mouse livers Atp8b1 deficiency impaired the transport of hydrophobic bile salts into bile. In conclusion, our study shows that Atp8b1 deficiency causes loss of canalicular phospholipid membrane asymmetry that in turn renders the canalicular membrane less resistant toward hydrophobic bile salts. The loss of phospholipid asymmetry may subsequently impair bile salt transport and cause cholestasis. (HEPATOLOGY 2006;44:195–204.)

Progressive familial intrahepatic cholestasis type 1 (PFIC1) or Byler disease is caused by mutations in ATP8B1.1, 2 Patients with PFIC1 suffer from chronic intrahepatic cholestasis that progresses to severe, end-stage liver disease, and often require orthotopic liver transplantation during the first or second decade of life.2, 3 PFIC1 is characterized biochemically by low biliary bile salt concentrations, elevated serum bile salt and bilirubin levels, normal cholesterol levels, and normal serum γ-glutamyltranspeptidase (GGT) activity. Electron microscopic analysis shows that canalicular bile has a coarsely granular appearance (termed “Byler bile”) as opposed to the amorphous or finely filamentous bile observed in other forms of cholestasis.4, 5 Thus, absence or severe dysfunction of ATP8B1 affects bile salt homeostasis and bile formation. ATP8B1 is a member of the type 4 subfamily of P-type ATPases.6–8 Proteins of this family translocate phospholipids from the outer to the inner leaflet of biological membranes and are termed flippases.9–12 In the liver, ATP8B1 localizes to the apical membrane of epithelial cells, including hepatocytes and bile duct epithelial cells (cholangiocytes).13–15 Ujhazy and colleagues have provided data to suggest that ATP8B1 is a flippase for phosphatidylserine (PS).15

Bile formation is essential for disposition of lipophilic endobiotics and xenobiotics, for intestinal solubilization and absorption of dietary lipids and fat-soluble vitamins, and for the regulation of cholesterol homeostasis. Primary bile is formed in bile canaliculi, tubular spaces bordered by the apical (or canalicular) membranes of adjacent hepatocytes. A major driving force for bile formation is active secretion of bile salts, mediated by the adenosine triphosphate–dependent bile salt export pump, ABCB11.16 Bile salt excretion induces an osmotic gradient that in turn causes the flow of water and accompanying solutes into the bile canalicular lumen.17–19 Bile salts have detergent properties that enable them to solubilize membrane lipids and are present in millimolar concentrations in the canalicular lumen. However, in the canaliculus, the detergent action of bile salts is decreased by molecular association with phosphatidylcholine (PC). PC is excreted into bile after outward translocation across the canalicular membrane by the flippase ABCB4.20 Absence or severe dysfunction of ABCB11 causes PFIC2,21 a disorder with extensive clinical and biochemical similarities to PFIC1; absence or severe dysfunction of ABCB4 causes PFIC3,22 a disorder having substantially less phenotypic overlap with PFIC1. Although ATP8B1 deficiency has a major impact on bile salt transport, it is unclear by which mechanism impaired ATP8B1 function results in cholestasis. We hypothesized that in the absence of ATP8B1 activity, lipid asymmetry of the canalicular membrane is insufficiently maintained and that this impairs hepatobiliary bile salt transport. We evaluated this hypothesis using Atp8b1G308V/G308V mutant mice, an animal model for PFIC1.23

Our study shows that Atp8b1 deficiency disturbs canalicular membrane phospholipid asymmetry and decreases the resistance of the canalicular membrane to hydrophobic bile salts evidenced by enhanced biliary recovery of PS, cholesterol, and ectoenzymes. In addition, we show in the isolated mouse liver that Atp8b1 deficiency impairs the hepatobiliary transport of hydrophobic bile salts. We propose that phospholipid randomization leads to enhanced extraction of cholesterol from the canalicular membrane, which subsequently may impair hepatobiliary transport of bile salts and causes cholestasis.

Abbreviations

PFIC1, progressive familial intrahepatic cholestasis type 1; Atp8b1G308V/G308V mice, Atp8b1 deficient mice, G308V, glycine-to-valine amino acid substitution; GGT, γ-glutamyl transpeptidase; PC, phosphatidylcholine; PS, phosphatidylserine; TUDC, tauroursodeoxycholate; TC, taurocholate; TDC, taurodeoxycholate; PBS, phosphate-buffered saline.

Materials and Methods

Mice.

Atp8b1G308V/G308V mutant mice (129/Sv strain) harbor a prototypic PFIC1 mutation (G308V) that results in absence of the protein.23 Mice were housed in a pathogen-free animal facility on a 12-hour light–dark cycle and were fed standard rodent chow. One week before experiments, animals were fed a purified semi-synthetic (20% casein) diet (K4068.02, Arie Blok, Woerden, The Netherlands) with or without 0.5% (w/w) cholic acid. In all experiments age-matched (3-6 months) male mice were anesthetized by intraperitoneal injection of 1 mL/kg Hypnorm (fluanisone/fentanyl citrate) and 10 mg/kg diazepam. All animal experiments were approved by the institutional animal care and use committee of the Academic Medical Center.

In Vivo Bile Salt Infusions.

The gallbladder and jugular vein were cannulated. Before bile salt infusion, mice were depleted of endogenous bile salts by bile collection without infusion. At t = 90 minutes TC or tauroursodeoxycholate (TUDC) were infused via the jugular vein. Infusion rates were increased every 30 minutes with 400 nmol/min*100g (400-800-1,200-1,600 nmol/min*100g). Bile samples (collected every 10 minutes) were analyzed for bile salt, choline-containing phospholipid, cholesterol, Alp, and Cd13 content.

Isolated, Single-Pass Liver Perfusions.

The gallbladder, portal vein, and superior vena cava were cannulated. The liver (left in situ) was perfused in orthograde direction (3 mL/h) with carbogen (95%O2/5%CO2)-gassed Krebs-bicarbonate buffer (120 mmol/L NaCl, 24 mmol/L NaHCO3, 1.2 mmol/L KH2PO4, 1.2 mmol/L MgSO4, 1.3 mmol/L CaCl2, 0.9 mg/mL glucose). Taurocholate (TC) and taurodeoxycholate (TDC), containing trace amounts of [3H]TC and [3H]TDC, respectively, were continuously perfused at a 50 nmol/min*100g rate for 90 minutes. Bile and perfusate samples were taken every 5 minutes; at the end of the experiment livers were weighed and part was dissolved in 100% soluene (Packard). Radioactivity was determined by liquid scintillation counting. [3H]TDC/[3H]TC ratios were determined in bile samples by high-pressure liquid chromatography analysis. The fluorescent chenodeoxycholate analog chenodeoxycholyl-(Nϵ-NBD)-lysine24 was perfused for 20 minutes at a 25 nmol/min*100g rate followed by 2 minutes' Krebs buffer; livers were excised and snap-frozen in liquid nitrogen, and liver sections (6 μm) were studied using a fluorescent microscope.

Bile Salts, Phospholipids, and Cholesterol Assays.

Bile salt, choline-containing phospholipids, and cholesterol were determined enzymatically.25 Alkaline phosphatase and aminopeptidase N activities were determined using para-nitrophenylphosphate and L-alanine-β-naphtylamide-HBr as substrates, respectively.25 All measurements were done on a Novostar analyzer (BMG Labtech GmbH, Offenburg, Germany).

High-Performance Thin-Layer Chromatography.

Phospholipids extracted from bile samples26 were dissolved in chloroform/methanol (1:2) and run on silica gel 60 plates (Merck, Darmstadt, Germany).27 Spot densities were quantified via photodensitometric scanning using Quantity One-4.2.3 software (BioRad, Veenendaal, the Netherlands).

Western Blotting of Liver Plasma Membranes.

Mouse liver plasma membranes were isolated as described.28 Ten to fifty micrograms protein was fractionated by 8% SDS-PAGE, blotted overnight, blocked in phosphate-buffered saline (PBS)/2% milk powder/0.05% Tween-20, and incubated with anti-Atp8b1,23 anti-Abcb11,29 anti-Abcg5,30 and anti-Atp1a1.31 Immune complexes were visualized with horseradish peroxidase–conjugated immunoglobulins and detected using enhanced chemiluminescence (Amersham, Buckinghamshire, UK).

Immunofluorescence.

Frozen liver sections (6 μm) were fixed in acetone (100%), blocked with 10% normal goat serum in PBS (block buffer), and incubated with anti-Abcb11.29 Immunoreactivity was visualized with AlexaFluor488-conjugated immunoglobulins (Molecular Probes, Eugene, OR). Sections were mounted in Vectashield/DAPI (Vector Laboratories Inc., Burlingame, CA) and studied in a confocal laser scanning microscope.

Immunohistochemistry.

Liver tissue from persons with liver disease, both cholestatic and non-cholestatic, of various known etiologies (other than ATP8B1 mutation) was obtained at biopsy or hepatectomy during diagnostic or therapeutic interventions. PFIC1 patients (biopsy, 13 weeks, Fig. 4B; hepatectomy, 13 years, Fig. 4D) are homozygotes for a G923T substitution in ATP8B1.1, 32 Tissue was fixed in formalin or in buffered glutaraldehyde/paraformaldehyde and was processed through graded ethanols and xylenes, respectively, into paraffin or into resin. Sections of formalin-fixed, paraffin-embedded material cut at 4 μm were immunostained and hematoxylin-counterstained,33 using anti-GGT,34 anti-CD10 (Novocastra, Newcastle-upon-Tyne, UK), anti-ABCB11,29 anti-ABCC2 (Signet/Bioquote, York, UK), and anti-ABCB4 (Alexis Biochemicals/Axxora, Nottingham, UK). Sections were studied by light microscopy.

Electron Microscopy.

Mouse livers were perfused via the portal vein (3 mL/min) with PBS for 2 minutes, followed by 2% paraformaldehyde/PBS for 5 minutes, and postfixed in McDowell's fixative [4% paraformaldehyde/1% glutaraldehyde in 0.1 mol/L PBS (pH 7.4)]. For standard electron microscopy, samples were postfixed in 1% OsO4, dehydrated in graded ethanols, and embedded in Epon. Thin sections (80 nm) were cut, collected on slot grids, and contrasted with uranyl acetate and lead citrate.35 For freeze fracture, samples were cryoprotected with 2.3 mol/L sucrose in 0.1 mol/L PBS (pH 7.4) and frozen in liquid ethane at 90K. Freeze-fracturing was performed in a BAF300 (BAL-TEC, Balzers, Liechtenstein) at a vacuum of <10−5Pa. Aldehyde fixed samples were fractured at 150K, replicated with 2 nm platinum at 45°, and 20 nm carbon was deposited at 90° to strengthen the replica. Samples were cleansed with household bleach, rinsed with distilled water, and mounted on bare 300 mesh copper grids. Replicas were studied in a Philips EM-420 transmission electron microscope (Philips, Eindhoven, the Netherlands) operated at 100 kV and equipped with a SIS MegaviewII camera.

Statistical Analyses.

Statistical analyses were performed using Student paired t test; a P-value of less than .05 was considered significant. All data were expressed as means ± standard deviation (SD). To determine statistical significance of the in vivo experiments, data were log10(x+1)-transformed and analyzed using SPSS11.5.1 by two-way ANOVA; the significance of the differences between the slopes and between y-intercepts was assayed by covariance analysis. In all cases, values of P < .05 were considered statistically significant.

Results

Atp8b1 Deficiency Leads to Enhanced Biliary Recovery of Lipid and Protein After Infusion of Hydrophobic Bile Salts.

To examine the consequence of Atp8b1 deficiency on bile formation, we challenged wild-type and Atp8b1G308V/G308V mutant mice by intravenous infusion of TC, a mildly hydrophobic bile salt. During bile salt depletion, biliary bile salt output was significantly higher in mutant than in wild-type mice (Fig. 1). Upon infusion of TC, biliary bile salt output was the same in mutant and wild-type mice even at the highest infusion rates. Because excretion of phospholipids and cholesterol is coupled to bile salt excretion, the output of these bile constituents was plotted against bile salt output. Biliary phospholipid excretion was stimulated by TC infusion and was slightly increased in mutants compared to wild-types (Fig. 2A). Remarkably, cholesterol output was stimulated twofold in mutants compared to wild-types (Fig. 2B); intrahepatic cholesterol concentrations were not different [13.2 ± 3.6 vs. 13.8 ± 2.7 μmol/gram dry weight for wild-type and mutant, respectively (n = 6)], and expression levels of the canalicular cholesterol half-transporter Abcg5 were unaffected (Supplemental Fig. 1; Supplementary material for this article can be found on the HEPATOLOGY website (http://interscience.wiley.com/jpages/0270-9139/suppmat/index.html). As a measure of detergent resistance of the membrane, we determined the release of canalicular ectoenzymes into bile, including glycosyl phosphatidylinositol-anchored alkaline phosphatase (Alp) and aminopeptidase N (Cd13), a protein with a single transmembrane domain. Biliary output of Alp and Cd13 was dramatically higher in mutants compared to wild-types (Fig. 2C-D). In contrast, we could not detect any integral membrane proteins such as Abcb11 and Abcc2 in bile (not shown). Elevated enzyme output was not caused by cell damage since no increase in the activity of the cytosolic marker lactate dehydrogenase could be detected in bile. The enhanced recovery of biliary lipids and ectoenzymes in mutant mice depended on bile salt hydrophobicity, as infusion of the hydrophilic bile salt TUDC caused none of these effects (not shown).

Figure 1.

Intravenous infusion of taurocholate (TC) does not affect the bile salt excretion capacity in Atp8b1G308V/G308V mutant mice. Bile salt output is plotted in time. The endogenous bile salt pool was depleted during the first 90 minutes. At t = 90 minutes TC was infused via the jugular vein. Infusion rates were increased stepwise every 30 minutes with 400 nmol/min*100g (400-800-1200-1600 nmol/min*100g). White circle, wild-types; black square, Atp8b1G308V/G308V mutants. Results are expressed as means ± SD of 4 and 5 experiments for wild-types and mutants, respectively; *P < .05.

Figure 2.

Intravenous infusion of taurocholate results in enhanced cholesterol and ectoenzyme output into bile of Atp8b1G308V/G308V mutant mice. (A) Relation between bile salt output and output of choline-containing lipids (PL), (B) cholesterol (CH), (C) alkaline phosphatase (Alp), and (D) aminopeptidase N (Cd13). 1 Unit is defined as 1 nmol or 1 pmol product formed per minute for Alp and Cd13 activity, respectively. Data points in all panels are from four and five experiments for wild-types and mutants, respectively. White circle, wild-types; black circle, Atp8b1G308V/G308V mutants; the difference between wild-types and mutants was significant (P < .05) in panels B, C, and D.

Atp8b1 Deficiency Leads to Extraction of Phosphatidylserine From the Canalicular Membrane by Bile Salts.

We quantified phospholipid species in bile of wild-type and Atp8b1G308V/G308V mutant mice after infusion of TC by high-performance thin-layer chromatography. In both genotypes, the most abundant biliary phospholipid was PC, with small quantities of phosphatidylethanolamine and sphingomyelin also present (Fig. 3). However, bile of mutant mice also contained significant amounts of PS, an aminophospholipid that is normally confined to the inner leaflet of biological membranes and that was completely absent from wild-type bile. This suggests that, in mutant mice, PS is exposed in the outer leaflet of the canalicular membrane, from which TC readily extracts it. The appearance of PS in bile also depended on bile salt hydrophobicity because TUDC infusion did not result in biliary recovery of PS (not shown). PS exposure in the outer leaflet of the canalicular membrane was not caused by apoptosis since we could not detect activated caspase 3–positive cells after TC infusion (not shown). There was no difference in fatty acyl chain saturation in bile samples of wild-types and mutants as determined by gas chromatography/mass spectometry (not shown).

Figure 3.

Atp8b1G308V/G308V mutant mice have significant amounts of phosphatidylserine in bile after infusion of taurocholate. Quantitative analysis of lipid species (expressed as percentage of total phospholipid on a logarithmic scale) in bile of wild-type and Atp8b1G308V/G308V mutant mice after infusion of taurocholate (samples were taken at t = 220 minutes in the experiments described in Figs. 1 and 2). SM, sphingomyelin; PC, phosphatidylcholine; PS, phosphatidylserine; PE, phosphatidylethanolamine. White bars, wild-type; black bars, Atp8b1G308V/G308V mutant. Data represent means ± SD of three samples from three different experiments. *P < .05.

PFIC1 Liver Lacks Immunohistochemically Demonstrable Canalicular Ectoenzymes.

We also analyzed the expression of canalicular ectoenzymes in liver biopsies of PFIC1 patients, including those from patients with a mutation orthologous to that introduced into the mutant mice. We studied the localization of GGT and neutral endopeptidase (CD10), two canalicular ectoenzymes both of which have single transmembrane domains, in human control and PFIC1 liver. We could not detect any immunohistochemical staining of either GGT or CD10 along canaliculi in liver of PFIC1 patients (Fig. 4B,D); however, GGT and CD10 staining was detected in the apical membrane of cholangiocytes. In biopsies from human controls, canalicular GGT and CD10 staining was normally present (Fig. 4A,C). ABCB11, ABCC2, and ABCB4, all integral membrane proteins, were well-expressed at canalicular margins in PFIC1 patients (not shown).

Figure 4.

PFIC1 patients lack immunohistochemical detectable expression of ectoenzymes in the canalicular membrane. (A) Gamma-glutamyl transpeptidase (GGT) staining in human control liver is seen in the apical membrane of cholangiocytes (arrow) and along bile canaliculi throughout the lobule; bar, 100 μm. (B) Immunohistochemical staining of GGT in liver of a PFIC1 patient. GGT marking is absent from the canalicular membrane; however, it is present in the apical membrane of cholangiocytes and along canals of Hering at portal-tract margins (arrow); bar, 100 μm. (C) Neutral endopeptidase (CD10) staining in human control liver is seen in the apical membrane of cholangiocytes (arrow) and along bile canaliculi throughout the lobule; bar, 200 μm. (D) Immunohistochemical staining of CD10 in liver of a PFIC1 patient. CD10 marking is absent from the canalicular membrane; however, it is present in the apical membrane of cholangiocytes (arrow); bar, 200 μm.

Atp8b1G308V/G308V Mutant Mice Accumulate Vesicular Structures in the Canalicular Lumen.

In PFIC1 liver, the canalicular lumen accumulates vesicular material as opposed to livers from patients with other forms of cholestasis.4, 5 To test whether Atp8b1G308V/G308V mutant mice also accumulated vesicular material, we performed transmission EM on livers of wild-type and mutant mice fed a control diet. We have viewed a large number of sections (n = 25), and virtually all canaliculi of mutant livers contained vesicular material, without other histological abnormalities (Fig. 5B). This was in contrast to wild-type livers, in which essentially no canaliculi were seen that contained vesicular content. When mice were fed a cholate-supplemented diet, the appearance of these lipid vesicles was more pronounced and was accompanied by the loss of microvilli (Fig. 5D). Analysis of mutant livers by freeze-fracture EM clearly revealed the multilamellar nature of these lipid vesicles, which were completely absent in livers from control mice (Supplementary Fig. 2).

Figure 5.

Transmission electron microscopic (EM) analysis of liver of Atp8b1G308V/G308V mutant mice displays accumulation of vesicular structures in the canalicular lumen. (A) EM image of wild-type liver (of mice fed a control diet) showing a canaliculus with microvilli without other content. (B) EM image of Atp8b1G308V/G308V mutant liver (of mice fed a control diet) showing a dilated canaliculus containing lipid, without loss of microvilli. (C) EM image of wild-type liver of mice fed a 0.5% cholate-supplemented diet for 1 week showing a canaliculus with shortened microvilli without other content. (D) EM image of Atp8b1G308V/G308V mutant liver of mice fed a 0.5% cholate-supplemented diet for 1 week showing a dilated canaliculus without any microvilli that accumulates (multi)vesicular structures. Bar, 370 nm.

Atp8b1 Deficiency Leads to Impaired Hepatobiliary Bile Salt Transport in the Isolated Liver.

We also investigated bile salt transport in isolated livers from wild-type and Atp8b1G308V/G308V mutant mice. Livers were perfused with radiolabeled TC or TDC (a more hydrophobic bile salt), and excretion of radiolabeled bile salt into bile was measured. In mutant livers, hepatobiliary output of TC was only slightly reduced but TDC excretion was strongly impaired (Fig. 6A,D). Concomitantly, retention of bile salts was much greater in mutant than in wild-type livers and was also stronger for TDC than for TC (Fig. 6E,B). In the perfusions with TDC we determined the extent to which TDC was rehydroxylated to TC (TDC/TC ratio) (Fig. 7). There was a striking, almost 16-fold reduction in the TDC/TC ratio in bile of mutant livers, which indicates that nearly all TDC was converted to TC. Perfusion of a fluorescently labeled chenodeoxycholate analog resulted in bright canalicular staining exclusively in the periportal regions in wild-type livers; in mutant livers, however, fluorescence in the canaliculi was less pronounced with a more diffuse intracellular staining throughout a larger part of the liver lobule (Fig. 8). The biliary output of the fluorescent bile salt analog was 3.5-fold lower in mutant liver compared to wild-type liver. Neither expression levels nor localization of Abcb11 were changed (Supplemental Figs. 1 and 3), suggesting that absence of Atp8b1 affects Abcb11 activity rather than its localization.

Figure 6.

Atp8b1 deficiency affects canalicular bile salt transport in the isolated, single-pass liver perfusion. (A) Biliary output of TC. (B) Hepatic retention of TC. (C) TC that bypasses the liver (perfusate). (D) Biliary output of TDC. (E) Hepatic retention of TDC (F) TDC that bypasses the liver (perfusate). Bile, liver, and perfusate data for [3H]TC and [3H]TDC are expressed as percentage of the administered dose. White bars, wild-type; black bars, Atp8b1G308V/G308V mutant. Results for all panels are expressed as means ± SD of 4 experiments. *P < .05;**P < .005.

Figure 7.

Atp8b1 deficiency results in enhanced hydroxylation of TDC to TC. During liver perfusion with [3H]TDC, the amount of [3H]TDC and [3H]TC was determined in bile samples at t = 30, 60, and 90 minutes by high-pressure liquid chromatography analysis. Data are expressed as the ratio TDC over TC measured at the indicated time points of three independent experiments. White bars, wild-type; black bars, Atp8b1G308V/G308V mutant. Results are expressed as means ± SD; *P < .001.

Figure 8.

Localization of the fluorescent bile salt chenodeoxycholyl-(Nϵ-NBD)-lysine in liver of wild-type and Atp8b1G308V/G308V mutant. (A) Fluorescent dots in wild-type liver represent bile canaliculi in the periportal area of the lobule; dark areas correspond to the centrilobular area; bar, 200 μm. (B) Mutant liver showing a more diffuse fluorescent staining throughout the liver; bar, 200 μm.

Discussion

Here we show that the canalicular membrane of hepatocytes in Atp8b1G308V/G308V mutant mice is more sensitive than normal to the detergent action of hydrophobic bile salts as evidenced by enhanced biliary recovery of cholesterol and ectoenzymes. The presence of PS in bile of mutant mice implies that PS is exposed in the outer leaflet of the canalicular membrane where it is readily extracted by hydrophobic bile salts. In addition, Atp8b1 deficiency impairs the transport of hydrophobic bile salts across the hepatocyte into bile in the isolated perfused liver. Collectively these data suggest that the cholestasis in PFIC1 is caused by phospholipid randomization and subsequent reduced detergent resistance of the canalicular membrane toward hydrophobic bile salts. We hypothesize that phospholipid randomization leads to enhanced extraction of cholesterol by hydrophobic bile salts, which subsequently may impair the activity of ABCB11 and cause cholestasis.

Detergent resistance of biological membranes depends on the relative content of sphingolipids and cholesterol. This has been extensively studied in relation to lipid rafts: rigid, detergent-resistant membrane domains in which lipids are tightly packed (so-called ‘liquid ordered’ phase).36 These domains spontaneously segregate from glycerophospholipid/cholesterol-rich domains (so-called ‘liquid disordered’ phase), which are much more susceptible to detergents. We believe that the bile canalicular membrane is highly asymmetric and that lipids in the outer leaflet largely exist in the ‘liquid ordered’ phase, with high sphingolipid and cholesterol content. In vitro experiments have shown that the combination of these two lipids promotes the formation of detergent-resistant membrane domains.37 Indeed, the canalicular membrane has been shown to be highly enriched in sphingomyelin, which suggests that this membrane is very rigid.38, 39 Membrane phospholipid randomization (either by phospholipid flopping or via fusion of exocytic vesicles with the canalicular membrane) leads to a shift from a more ‘liquid ordered’ to a more ‘liquid disordered’ phase, the latter being less resistant to the detergent action of bile salts. We propose that ATP8B1 activity is crucial in maintaining this ‘liquid ordered’ phase by flipping excess aminophospholipids from the outer to the inner leaflet. The absence of Atp8b1 function in Atp8b1G308V/G308V mice thus reduces lipid ordering in the outer leaflet. Higher sensitivity to extraction of phospholipids, cholesterol, and ectoenzymes follows. Increased cholesterol extraction may reduce the cholesterol content of the outer leaflet, which in turn may impair the activity of the bile salt transporter Abcb11. It has been shown that membrane cholesterol content is important in controlling the activity of other integral membrane proteins such as ABCB1,40 Na+/K+-ATPase,41 and the nicotinic acetylcholine receptor.42

Biliary cholesterol excretion is mediated by ABCG5/ABCG8 heterodimer, which is expressed in the canalicular membrane of the hepatocyte.43 We find an approximate twofold increase in biliary cholesterol output in Atp8b1G308V/G308V mutant mice when challenged with taurocholate, without a change in Abcg5 protein and mRNA expression levels. This suggests that the cholesterol output in mutant mice was enhanced by extraction from the canalicular membrane. Alternatively, Abcg5/g8 activity may be influenced by changes in membrane asymmetry. Furthermore, we believe that the enhanced biliary cholesterol and ectoenzyme levels, as demonstrated in Atp8b1G308V/G308V mutant mice, are derived from the canalicular membrane rather than from the apical membrane of cholangiocytes. This is underscored by the presence of both GGT and CD10 staining in the apical membrane of cholangiocytes in liver of PFIC1 patients. This observation also indicates that the phenotype directly depends on bile salt output and the presence of detergent bile salt micelles in the canalicular lumen. We hypothesize that lipid extraction from the canalicular membrane decreases the detergent action of these micelles, thereby reducing damage to the apical membranes of cholangiocytes.

The origin and nature of the vesicular structures in the canalicular lumen of Atp8b1G308V/G308V mutant mice are unclear, but the structures are very reminiscent of those found in PFIC1 patients.4, 5 Highly similar multilamellar structures have been demonstrated in bile canaliculi of rats in which hepatobiliary cholesterol output was increased by diosgenin/simvastatin feeding.44 The vesicular structures observed in canaliculi of mutant livers are likely formed as a consequence of the elevated biliary cholesterol levels. These structures must not be confused with the vesicles that have been described by Crawford and coworkers45 to support the vesiculation theory, a proposed mechanism for biliary phospholipid excretion. In contrast to the vesicles observed in our study, they described unilamellar vesicles, most of which were adherent to the canalicular membrane. Furthermore, different methods of sample processing were used, therefore the data can not be directly compared.

We show here that the bile salt output in Atp8b1G308V/G308V mice is unaffected both after intravenous infusion of TC (Fig. 1) and in the isolated, single-pass perfused liver (Fig. 6). In line with these results, we previously reported normal biliary bile salt output in mutant mice.23 In contrast, single-pass perfusion of the liver with the more hydrophobic bile salt TDC resulted in a significantly lower biliary output and a concomitant increase in hepatic retention of TDC. Although these results may sound discrepant with our present findings, there are two explanations for this difference: (1) Mutant mice have mild cholestasis only in the presence of hydrophobic bile salts, and (2) The data were obtained by the use of different experimental setups. Our present results have been obtained in the single-pass perfused liver while our previous paper reported studies in intact animals. An important difference between these two experimental systems is that infusion of bile salts in the intact animal leads to accumulation in the bloodstream. If the animals are not strongly cholestatic, all infused bile salt can still be excreted albeit at the expense of higher blood and liver bile salt concentrations and the mobilization of more hepatocytes. The isolated single-pass perfused liver represents a more sensitive setup to measure excretion capacity because excess bile salt is allowed to the liver. This is illustrated by the experiment described in Fig. 6. When radioactive TC was administered to the isolated single-pass perfused liver, more than 90% of the bile salt was recovered in bile and liver, indicating that all the administered bile salt was extracted in a single pass. This was the case in both wild-type and mutant livers, indicating that the mutant livers can almost normally handle this bile salt. When TDC was administered to perfused wild-type livers, 73% and 13% of the bile salt ended up in bile and liver, respectively. Only 14% passed the liver in the perfusate, indicating that these livers are still capable of excreting most of the TDC in a single-pass fashion. In mutant livers, however, 34% of the administered TDC passed the liver, 39% was retained by the liver, and only 27% ended up in bile. Hence, only in the isolated single-pass perfused liver can the cholestasis in mutant mice be properly unmasked. In line with our previous report,23 we observed a strongly increased hydroxylation of the perfused TDC to TC in mutant mice. We assume that this is caused by the retention of TDC in the liver, which causes more extensive exposure to the relevant hydroxylating enzyme in the endoplasmic reticulum. The extensive rehydroxylation of TDC also explains the observation that mutant mice are resistant to TDC-induced cholestasis.23 Because of enhanced hepatic retention, TDC is almost completely converted to TC; since the maximal secretory rate for TC is much higher, mutant animals maintain high bile salt output compared to wild-types (which do become cholestatic).

We conclude that transport of hydrophobic bile salt is significantly impaired in Atp8b1G308V/G308V mutant mice. This may provide an explanation for the striking phenotypic discrepancy between Atp8b1G308V/G308V mutants and PFIC1 patients; there is a strong difference in the composition of the bile salt pool between mouse and man. Mouse bile contains predominantly trihydroxy bile salts (cholate and muricholate), whereas in human bile, the majority of bile salts is dihydroxy bile salts, which are much more hydrophobic.46, 47 These differences underscore the importance of bile salt pool composition in the cause of PFIC1, and also may explain the beneficial effects of bile drainage in PFIC patients.48, 49 This procedure interrupts the enterohepatic circulation and causes a shift of the bile salt composition toward primary, more hydrophilic bile salts.

Apart from liver, ATP8B1 is also expressed in other tissues, e.g., pancreas, small intestine, bladder, stomach, and prostate.1, 13, 14 Indeed, many PFIC1 patients exhibit extrahepatic symptoms, including pancreatitis, watery diarrhea, hearing loss, and elevated sweat salt concentrations, the latter like patients with deficiency in the cystic fibrosis transmembrane conductance regulator.3, 50 Our present results do not explain in any way the additional phenotypes observed in PFIC1 patients. It may be hypothesized, however, that ATP8B1 also plays a role in maintaining apical membrane asymmetry in other tissues, and that its absence causes a pleiotropic malfunction of this membrane domain.

We conclude that the intrahepatic cholestasis in PFIC1 is caused by an impaired aminophospholipid flippase activity that destroys the usual phospholipid asymmetry of the canalicular membrane. As a result the canalicular membrane becomes more sensitive to extraction of cholesterol by bile salts, which impairs ABCB11 activity and, as a consequence, causes cholestasis.

Acknowledgements

The authors thank Drs. T.B.M. Hakvoort and J.M. Ruijter for assistance with statistical analyses, Drs. B. Stieger, L.W. Klomp, J.B. Koenderink, and M.H. Hanigan, respectively, for polyclonal antibodies to Abcb11, Atp8b1, Atp1a1, and γ-glutamyl transpeptidase, and Dr. A.K. Groen for helpful discussions.

Ancillary