Canalicular ABC transporters and liver disease

Abstract

Bile is a complex mixture that includes bile salts, the membrane phospholipid phosphatidylcholine (PC), cholesterol and various endobiotic and xenobiotic toxins, each of which is secreted across the canalicular membrane of the hepatocyte by different ATP-binding cassette (ABC) transporters. The bile salts are essential for the emulsification of dietary fat and lipophilic vitamins. They are synthesized from cholesterol in the hepatocyte and their secretion by the bile salt export pump (BSEP or ABCB11) drives bile flow and is the starting point for the enterohepatic cycle. The detergent nature of bile salts that is key to their physiological role also means that they are inherently cytotoxic, and failure to secrete bile (intraheptic cholestasis) can precipitate severe liver disease and mortality. Such progressive familial intrahepatic cholestasis (PFIC) comes in three types of autosomal recessive disease. PFIC2 is caused by mutation to ABCB11. PFIC3 is caused by mutation of a closely related ABC transporter, ABCB4, which flops PC into the outerleaflet of the canalicular membrane. The flopped PC is extracted by the bile salts in the canaliculus to form a mixed micelle that reduces bile salt detergent activity. The third protein that is essential for bile flow from the hepatocyte is a member of a different class of transporter protein, a P-type ATPase, ATP8B1. Mutation of ATP8B1 causes PFIC1, but ATP8B1 does not transport a component of bile into the canaliculus. Data from different laboratories, published this year, suggests two different roles for ATP8B1 in the hepatocyte: a lipid flippase, that counterbalances the deleterious effects of ABCB4 on barrier function of the canalicular membrane; and an anchor of the actin cytoskeleton necessary to form the microvilli of the brush border. These latest discoveries are described, along with a spectrum of cholestatic disorders whose aetiologies lie in these and other transporters of the canalicular membrane. Copyright © 2011 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Introduction

Bile salts are essential for the solubilization of dietary fats and lipophilic vitamins. They also interact with a number of membrane-bound and nuclear receptors to stimulate a variety of signalling pathways. They are synthesized in the hepatocyte and secreted across the canalicular membrane into the biliary tree, from where they drain into the gall bladder. From the gall bladder they are expelled into the small intestine in response to food intake to fulfil their primary role. They are then reabsorbed very efficiently by enterocytes, and secreted into the portal vein for return to the liver to complete the enterohepatic cycle. The detergent nature of bile salts that is necessary for their physiological role also means that they are inherently cytotoxic. Both the synthesis and enterohepatic cycling of bile salts is therefore under tight control at the protein and genetic level to limit their intrahepatic and serum concentrations. In the biliary tree, high concentrations of bile salts are inevitable, so we secrete phospholipid to form a mixed micelle with the bile salt, thus reducing its detergent activity; and we have mechanisms to ensure that the apical membranes of the cell types lining the biliary tree are maximally resistant to detergent solubilization.

Bile salts constitute approximately 50% of the organic constituents of the bile dry weight. The other major constituent, at 25%, is the membrane phospholipid phosphatidylcholine (PC), and the remainder is a complex mixture of cholesterol, bilirubin, glutathione, electrolytes, pigments, drugs and plant sterols (bile flow itself is an important route for excretion of endobiotic and xenobiotic toxins).

The anabolic enzymes, the transporters, receptors and transcription factors involved in bile salt homeostasis are now largely identified, and these are described below. However, from the perspective of human cholestatic morbidity, three of the hepatocyte canalicular transporters are key. The bile salt export pump (ABCB11) and the PC floppase (ABCB4), efflux the two main components of bile (bile salts and PC) and both are members of the ATP binding cassette (ABC) superfamily of transport proteins. In humans there are 42–44 ABC transporters, many of which have important roles in normal physiology and morbidity 1. The third protein necessary for bile flow from the hepatocyte is a P-type ATPase, ATP8B1 (also known as FIC1). ATP8B1 appears to flip a different lipid, phosphatidylserine (PS), in the opposite direction to PC flop by ABCB4 (ie it flips PS inwardly). Why this activity is necessary is only just becoming apparent and current theories are explored herein. The endobiotics and xenobiotics are transported by a variety of other pumps (mostly other members of the ABC transporter family; see below) but it is the flux of bile salts across the canalicular membrane that drives bile flow, and this review therefore focuses on our current understanding of the activities of ABCB11, ABCB4 and ATP8B1.

Bile salt synthesis

The bile salt pool (Figure 1) comprises a mixture of primary bile salts that are synthesized de novo in the hepatocyte and secondary and tertiary bile salts, which result from bacterial catabolism of the primary bile salts in the gut. The two primary bile salts in humans, cholic acid (CA) and chenodeoxycholic acid (CDCA), are synthesized from cholesterol 2. This is the major route of cholesterol catabolism and involves at least 14 different enzyme activities. It begins with a key hydroxylation event catalysed by the cytochrome P-450 enzyme, CYP7A1. The expression of CYP7A1 is induced by cholesterol (and oxy-sterol metabolites) and repressed by bile salts, providing tight control of bile salt synthesis at this rate-limiting step 3. The main product of the secondary metabolism of the symbiotic gut microflora is deoxycholic acid (DCA). CA, CDCA and DCA comprise about 95% of the bile salt pool in humans. The remainder is mostly lithocholic acid (LCA) and the tertiary bile salt ursodeoxycholic acid (UDCA; the product of the secondary bile salt 7-keto-LCA, which is further modified in the liver). Ninety-nine per cent of all human bile salts are conjugated in the hepatocyte at carbon-24 with either the amino acid glycine or taurine, a derivative of cysteine (in a ratio of 3 : 1, respectively 4, 5). Amidation reduces the pKa such that most bile salts are ionized (ie they are deprotonated bile acids) at the low pH present in the bile duct, gallbladder, duodenum and small intestine. Amidation thus increases their hydrophilicity, making them more effective emulsifying agents, and decreases their toxicity 6. It also significantly reduces their ability to diffuse passively across cell membranes, and influx and efflux systems have evolved to ensure efficient transcellular flux through enterocytes and hepatocytes.

Figure 1. Chemical structure of the major bile salts comprising the human bile salt pool. The 24-carbon primary bile acids CA and CDCA are synthesized in the liver from cholesterol. The rate-limiting step is the hydroxylation of C7 of cholesterol, catalysed by the microsomal CYP7A1. Secondary bile acids (DCA and LCA) are formed in the gut through the dehydroxylation of primary bile acids by bacterial enzymes. Together, CA, CDCA and DCA form 95% of the total bile salt pool. Epimerization of the C7 hydroxyl of CDCA results in the formation of 7-ketolithocholic acid, which is further modified in the liver to form the more polar and hydrophilic tertiary bile acid UDCA. The carboxyl group (C24) of 99% of bile salts is amidated with either glycine or taurine in a ratio of 3 : 1, respectively. CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; LCA, lithocholic acid; UDCA, ursodeoxycholic acid; BS, bile salt; R, functional group

A note on the use of model animals

Here, it is important to note that the composition of the bile salt pool varies in different higher organisms. This is therefore an appropriate juncture to consider the impact and challenges of animal studies in this area of research. While the study of model organisms (rodents primarily) have provided many insights, it has become clear that direct extrapolation of model animal data (or biochemical data derived from animal transporters in vitro), to human physiology can be difficult. For example, the transporters in different species appear to have co-evolved with anabolic pathways (both the endogenous and symbiotic) that are the basis of the different bile salt pools in rodents and man, to minimize loss of bile salts during enterohepatic cycling. The phenotype of knockout animals can also be surprisingly different (generally milder) to the disease found in transporter-deficient human patients. This is likely because the rodent bile salt pool is less toxic and also because compensatory mechanisms appear to be up-regulated in rodent models. These phenomena are described in more detail below as germane to understanding the function of the individual transporters.

The enterohepatic cycle

The enterohepatic cycle (Figure 2) is very efficient at recovering bile salts; approximately 95% are re-absorbed, with the remaining 5% that is lost in the feces replaced by de novo synthesis in the liver. Thus, ∼500 mg of primary bile salts are synthesized daily and added to the 2–3 g that circulate three or four times per day. The cycle begins at the canalicular membrane of the hepatocyte, where newly synthesized bile salts are effluxed into the canaliculus by the bile salt export pump ABCB11. ABCB11, and its cognate transporters, ABCB4 and ATP8B1, which are necessary for bile flow, are described in detail below. The bile collects in the gall bladder, which contracts in response to the peptide hormone cholecystokinin that is secreted by the mucosal epithelium of the small intestine in response to the presence of chyme. To exit the gastrointestinal tract and return to the hepatocyte, the bile salts must cross three membranes; the apical and basolateral of the ileocyte, and the basolateral of the hepatocyte 7–9.

Figure 2. The enterohepatic recycling of bile salts (BSs). Approximately 95% of the secreted BSs are re-absorbed in the enterohepatic cycle. De novo synthesis from cholesterol (chol) in the liver replaces the 5% that is lost in the faeces. Newly synthesized BSs are effluxed into the bile canaliculus by the bile salt export pump ABCB11, where they form micelles with phosphatidylcholine (PC) flopped to the outer leaflet of the canalicular membrane by ABCB4. The action of the phosphatidylserine (PS)-flippase ATP8B1 is thought be involved in the maintenance of lipid asymmetry at the canalicular membrane. Once they are secreted into the canaliculus, BSs collect in the gall bladder and are episodically released into the small intestine in response to hormonal signals. In the distal ileum, apical sodium-dependent bile salt transporter (ASBT) imports BSs into the ileocyte in a Na⁺-dependent manner. At the basolateral membrane of the ileocyte, BSs are exported by the heterodimeric organic solute transporter OSTα/β into the portal vein and thence to the liver. At the basolateral membrane of the hepatocyte, BSs are imported mainly via the Na⁺-dependent taurocholate co-transporting polypeptide (NTCP). Members of the organic anion transporting polypeptides (OATPs) also import BSs via either facilitated diffusion or an antiport mechanism in exchange for bicarbonate or reduced glutathione

Trans-ileocyte transport of bile salts

The major transporters involved in bile salt recycling through the ileocyte are: the apical sodium-dependent bile salt transporter [ASBT; also known as the ileal bile acid transporter (IBAT)]; the cytosolic ileal bile acid binding protein (IBABP); and the basolateral organic solute transporter (a heterodimer of α- and β-subunits; OSTα/β).

ASBT (SLC10A2) is a member of the SLC10 family of solute carriers 10. It is expressed on the apical membrane of the ileocytes lining the distal ileum 11, 12. ASBT concentrates bile salts in the ileocyte by a symport mechanism driven by the sodium gradient across the apical membrane. Sodium is thus cotransported with the bile salt (the intra-ileocyte sodium concentration is kept low by the Na⁺K⁺ ATPase on the basolateral membrane). Translocation of bile salts through the ileocyte to the basolateral membrane is thought to involve IBABP, which can bind to the intracellular face of ASBT 13, 14. At the basolateral membrane, OSTα/β exports the bile salts into the portal vein. OSTα/β was first described in the skate, Leucoraja erinacea15, before the human homologue was cloned in 2003 16. In humans it is expressed at high levels in a variety of tissues, including liver and small intestine, and is the primary exporter of bile salts into the systemic circulation 16, 17. OSTα/β works by facilitated diffusion, transporting bile salts (and other organic solutes and steroids) down their concentration gradients. After secretion into the portal blood, bile salts reach the liver, where they diffuse into the perisinusoidal space of Disse and come into contact with the basolateral membrane of the hepatocyte.

Trans-hepatocyte transport of bile salts

Across the basolateral membrane of the hepatocyte, there is both Na⁺-dependent and -independent uptake of bile salts 8, 9, 18, 19. The Na⁺-dependent taurocholate co-transporting polypeptide [NTCP (SLC10A1), another SLC10 family transporter] is the most important, contributing 80% of the transport capacity. NCTP transports both conjugated and unconjugated bile salts (bile salts are commonly deamidated in the intestine by bacteria) into the hepatocyte 10, 20–24, again by symport with sodium ions (with a stoichiometry of 2 × Na⁺ ions/bile salt 25).

The Na⁺-independent system is mediated by the organic anion transporting polypeptides (OATPs), which belong to the SLCO (SLC21) family of solute carriers 9, 19, 26, 27. The main human OATPs (OATP1A2, OATP1B1, OATP1B3 and OATP2B1) are expressed at the basolateral membrane of the hepatocyte and primarily transport unconjugated bile salts 19, 27–29. OATP1A2 is unique in its substrate specificity, in that it can also transport organic cations (similar to OSTα/β) 30. The OATPs were thought to be secondary active transporters, as described for leukotriene uptake by rat Oatp1a1 in exchange for glutathione ( 31), or taurocholate uptake in exchange for bicarbonate 32. More recently however, human OATPB1B1, and to a lesser extent OATPB13, have been shown to be electrogenic, suggesting that they facilitate diffusion of bile salts down their electrochemical gradients (although anion exchange resulting in a net influx of anions was not ruled out 33). In the same study, rat Oatp1a1 was characterized as electroneutral, consistent with the earlier data, which suggests that different members of the OATP family really do have different mechanisms of action.

Following uptake, the mechanism of translocation through the hepatocyte is not well understood but most likely involves carrier proteins. Several cytoplasmic proteins, including glutathione S-transferases, steroid dehydrogenases and fatty acid binding proteins, have been shown to have affinity for bile salts but no definitive evidence for a role in intra-hepatocyte transport has been reported (reviewed in 34). In the hepatocyte, the recycled bile salts are added to the pool synthesized de novo and delivered to ABCB11 for export across the canalicular membrane.

Transport across the canalicular membrane

Our current understanding of this key activity in the intrahepatic cycle stems from 1969, when a rare progressive familial intrahepatic cholestatic (PFIC) disease was first described by Clayton et al35 (the condition is still referred to as Byler disease, after the eponymous Amish kindred in which it was described). It is now clear that there are at least three primary active transport proteins that are necessary for bile flow across the canalicular membrane, and each gives rise to a different type of autosomal recessive PFIC. All three present similarly as cholestasis early in life, often within the first year (Table 1) 35. Affected individuals absorb dietary fats and fat-soluble vitamins inefficiently, which leads to growth restriction and progressive liver damage caused by increased hepatic and serum levels of bile salts. Affected individuals also commonly have jaundice and intractable pruritus. The condition progresses to end-stage liver disease in the first or second decade. Characterization of these patients has allowed the genetic aetiology of PFIC to be determined and the three transporters that are key to bile flow across the canalicular membrane to be identified and characterized. PFIC types 2 and 3 are caused by mutations to the transporters that efflux the two major components of bile; bile salts via ABCB11, and the lipid PC via ABCB4. PFIC1 (Byler disease) is caused by mutation of ATP8B1, which transports PS in the opposite direction (from the outer to the inner leaflet of the membrane). While PFIC3 patients may be differentially diagnosed by raised serum γ-glutamyl transferase (γ-GT), and immunohistochemistry can be invaluable to establish whether specific transporters are implicated in disease aetiology 36, definitive clinical diagnosis generally requires genetic testing. It is becoming increasingly apparent that there is a spectrum of cholestatic disease caused by mutations in ABCB11, ABCB4 and ATP8B1. This ranges from the severe phenotype of PFIC to milder, intermittent forms of cholestasis, eg benign recurrent intrahepatic cholestasis (BRIC), drug-induced cholestasis (DIC) or intrahepatic cholestasis of pregnancy (ICP) (Table 1).

Table 1. Diseases associated with mutations in ATP8B1, ABCB11 and ABCB4

		ATP8Bl (OMIM_602397)	ABCBll (OMIM_603201)	ABCB4 (OMIM_171060)
PFIC, progressive familial intrahepatic cholestasis; PS, phosphatidylserine; PC, phosphatidylcholine; γ-GT, γ-glutamyl transferase; ALT, alanine transaminase; BRIC1/2/3, benign recurrent intrahepatic cholestasis type 1/2/3; ICP, intrahepatic cholestasis of pregnancy; LPAC, low phospholipid associated cholelithiasis; DIC, drug-induced cholestasis; CIC, contraceptive-induce cholestasis; HC, hepatocellular carcinoma.
Alternative names		FIC1	BSEP, SPGP	MDR3
Function		PS-flippase/microvillus formation	Bile salt export	PC-floppase
Associated PFIC		Type 1	Type 2	Type 3
Cholestatic onset		1st year—adolescence	1st year—adolescence	1st year—early childhood
Biliary bile acid		Very low, PS present	Absent	Normal but devoid of PC
Serum	Bile salts	High	High	Normal
	γ-GT	Normal	Normal	Raised
	ALT	Mild elevation	Five times normal levels	Five times normal levels
	Cholesterol	Sometimes elevated	Often elevated	Normal
Liver histology		Canalicular cholestasis	Giant cell hepatitis	Bile duct proliferation and fibrosis
		Granular bile
Pruritus		Severe	Severe	Moderate
Extrahepatic symptoms		diarrhoea, pancreatitis, hearing loss	None	None
Conditions associated with partial function		BRICl, ICP	BRIC2, JCP, DIC, CIC, HC	(BRIC3?), ICP, DIC, LPAC, CIC

ABCB11: The bile salt exporter and cause of PFIC2

In 1991, Nishida et al., described an ATP-dependent system for the transport of taurocholate in canalicular membrane vesicles derived from the rat liver 37. This observation was soon confirmed by three other groups 38–40 and a low stringency hybridization screen on pig liver cDNA identified a novel gene with high sequence identity to the human drug efflux pump ABCB1 (described in more detail below). This new gene was eventually renamed Abcb11 [earlier names include sister of Pgp (spgp) and bile salt export pump (Bsep)] 41. Successful cloning of the full-length rat homologue showed that Abcb11 could function in vitro as a bile salt efflux transporter in the presence of ATP, and northern blot analysis revealed that it was primarily expressed in the liver 42. Positional cloning of the human orthologue was then correlated with a set of mutations linked to PFIC2 43. This established ABCB11 as the main bile salt export pump in the human liver and mapped the ABCB11 gene to chromosome 2q24 43. Expression of the 140–150 kDa human ABCB11 in insect cells confirmed its ability to transport bile salts in an ATP-dependent manner, with an apparent affinity for different bile salts that reflects the composition of the bile salt pool in humans 4, 44. Knock-out mice deficient for Abcb11 develop intrahepatic cholestasis, but the phenotype is less severe than in PFIC2 patients 45. This is considered to be due to the less toxic nature of the murine bile salt pool and the compensatory up-regulation of other ABC transporters. In Abcb11^−/− mice there is a dramatic decrease in the export of cholic acid but the total output of bile salts is only slightly reduced 45. This unexpected observation was later shown to be due to the up-regulation of Abcb1a, which could apparently function as bile salt transporter in the absence of Abcb1146. However, no compensatory increase in ABCB1 mRNA has been observed in PFIC2 patients 47.

ABCB11 mutations have been reported in patients with PFIC2 and benign recurrent intrahepatic cholestasis type 2 or BRIC2 43, 48–52. Biochemically, PFIC2 is characterized by normal γ-GT and normal cholesterol levels in the serum but very low biliary bile salts 53, 54. BRIC2 develops usually after the first decade and can progress into PFIC2, although it is normally episodic and benign 51, 55, 56. Mutations in ABCB11 have also been linked to intrahepatic cholestasis of pregnancy (ICP) 57, drug-induced cholestasis (DIC) 58–61 and contraceptive-induced cholestasis (CIC) 62. The association of genetic variation in ABCB11 with ICP and CIC prompted research into the mechanisms by which reproductive hormones can impair ABCB11 function. Two sulphated progesterone metabolites have been demonstrated to trans-inhibit efflux of bile acids by ABCB11 in an oocyte model system 63. Furthermore, oestrogen 17β-glucuronide has also been demonstrated to trans-inhibit ABCB11 in an SF9 cell vesicle system 64. Many of the ABCB11 SNPs described have since been shown to affect mRNA splicing, protein stability and/or protein function in vitro52. ABCB11 deficiency is also associated with hepatocellular carcinoma in young children 65.

ABCB4: The PC flopper and cause of PFIC3

ABCB4 was first cloned in 1987 by Van der Bliek et al., from a human liver cDNA library and named MDR3, due to its high degree of similarity with the third hamster multidrug resistance (Mdr) gene 66. The ABCB4 gene locus spans 74 kb on chromosome 7q21.1 and gives four alternatively spliced transcripts 67. The high level of expression in the canalicular membrane suggested an important role for ABCB4 in the liver 68, which was later confirmed by the generation of Abcb4^−/− mice 69. Mice deficient in Abcb4 developed severe liver disease characterized by a complete absence of PC from the bile 69. Analysis of fibroblasts from transgenic mice showed that Abcb4 flops PC from the inner to the outer leaflet of the plasma membrane 70. The human orthologue was cloned and also shown to translocate PC specifically, across the plasma membrane 71 and, when knocked-in to Abcb4^−/− mice, human ABCB4 rescued PC excretion into the bile 72. In the absence of ABCB4, the detergent activity of bile salts transported into the bile by ABCB11 can solubilize cell membranes, resulting in biliary toxicity 73. Homozygous mutation of ABCB4 causes PFIC3, which presents within the first few years of age with pruritus, jaundice and growth restriction 61, 74. Liver histology reveals portal fibrosis, which can progress to cirrhosis and ductular proliferation that eventually leads to end-stage liver disease 53, 54. Biochemically, PFIC3 can be distinguished from PFIC1 and PFIC2 by high levels of γ-GT in the serum and complete absence of PC from the bile 75.

It is becoming increasingly apparent that mutations of ABCB4 are associated with a spectrum of cholestatic diseases. These overlap with ABCB11 phenotypes and include ICP 76, DIC 77, CIC and low phospholipid-associated cholelithiasis (LPAC, a form of cholesterol gallstone disease 78). Together, these conditions suggest that ABCB4-dependent hepatobiliary disease is common. ICP (also known as obstetric cholestasis) presents in the third trimester of pregnancy, when oestrogen levels reach their peak 57, 76, 79–81. Symptoms include severe pruritus and abnormal bile flow in the mother, which resolve rapidly after delivery. ICP carries a significant increased risk of adverse pregnancy outcome and intrauterine death, and affects 0.6% of births in the British Caucasian population. In LPAC, female gender and parity (in agreement with the influence of oestrogens in bile flow, as suggested for ICP) are among the major risk factors. LPAC patients often also present with raised serum γ-GT, further implicating ABCB4.

In 2004, 18 mutations in ABCB4 had been associated with PFIC3 and 30 mutations were linked to induced cholestatic disease, but often only in single or few families 58. These include 11 and 16 single-nucleotide polymorphisms (SNPs), respectively, which encode non-synonymous changes in the protein sequence. The effects of these SNPs on protein trafficking, function and response to inducing agents have not been characterized, so the causal conjunction between ABCB4 SNPs remains to be definitively proved for some hepatobiliary diseases (particularly the induced forms with a complex aetiology).

Some insights into the functional consequences of ABCB4 mutation have been obtained in our laboratory 82 and, more recently, by the group of Michelle Maurice 83, by mimicking clinically relevant mutations in the closest relative of ABCB4, the multidrug resistance ABCB1 and measuring the effect on ABCB1 expression and drug efflux. However, this approach is limited and only relevant if the particular amino acid is conserved in both proteins and performs the same function. This rules out the study of amino acid changes that directly influence PC binding or efflux, or responses to hormones of pregnancy, contraceptives or other drugs. Biochemical characterization of ABCB4 SNPs directly has, hitherto, been difficult for two reasons: the technical difficulty of measuring the flux of PC from the inner to the outer leaflet of the plasma membrane; and because expression of functional PC floppase, transiently in vitro, appears to be deleterious to cells and leads to cell death. We now believe that we have overcome this limitation and have coincidentally shed some light on why ATP8B1 is needed in the canalicular membrane (see below 84).

ATP8B1 (FIC1): the PS flipper and cause of PFIC1

Surprisingly, mutation of neither ABCB11 nor ABCB4, which transport the two major components in bile, could explain the original Byler family PFIC pedigree. The third transport protein that appears to be necessary for bile flow across the canalicular membrane was not discovered until 1998, when Laura Bull identified the ATP8B1 gene on chromosome 18q21 by positional cloning 85. ATP8B1 (also known as FIC1) is a P-type ATPase which functions in complex with the accessory protein CDC50 86. ATP8B1 moves a different membrane phospholipid, PS, in the opposite direction to the transport of PC by ABCB4. PS is generally restricted to the inner leaflet of the plasma membrane and so is not normally exposed at the cell surface. There are three lines of evidence in support of PS-flippase activity for ATP8B1: (a) loss of ATP8B1 activity reduces bile flow but causes the appearance of PS in the residual flow; this is also true of Atp8b1-deficient mice following taurocholate infusion 87; (b) fluorescently-labelled PS analogues are internalized by UPS-1 cells following transient expression of ATP8B1 86, 88; and (c) the same cells show reduced binding of annexin-V when added extracellularly (annexin-V has a high affinity for native PS when the PS headgroup is exposed on the cell surface).

ATP8B1-deficiency is linked to several cholestatic liver diseases, PFIC1 and BRIC1 85 and ICP 89, 90. PFIC1 is by far the most severe 91. Scanning electron micrographs of bile from patients reveals coarse granules (sometimes called ‘Byler bile’ 35). Histological examination of early liver biopsies reveals relatively mild liver damage, compared to PFIC2 and PFIC3, with lobular architecture largely preserved and an absence of giant cells 35. Signs of cholestasis are, in fact, largely confined to the canaliculi 35. In keeping, serum levels of hepatic enzymes γ-GT and ALT, are only mildly elevated in patients with PFIC1. Nevertheless, the common outcome is cirrhosis and end-stage liver disease within the second decade. The presence of ATP8B1 at apical membranes of pancreatic acinar cells, enterocytes of the ileum and jejunum, and cochlear hair cells of the inner ear means that PFIC1 is a systemic condition. Extrahepatic symptoms include intractable diarrhoea 92, sensorineural hearing loss 93 and, in a few cases, pancreatitis 94.

BRIC1 is also associated with ATP8B191. Affected individuals present with episodic bouts of cholestasis, pruritus and extrahepatic symptoms such as diarrhoea, which may last days, weeks or months but is not progressive. It is considered benign because no lasting liver damage occurs and during remission patients are biochemically normal. However, as is the case with PFIC1, pruritus can be very severe during cholestatic episodes and some patients have elected for liver transplantation in order to improve quality of life 95. Initial onset generally occurs early in childhood and episodes can occur multiple times within a single year, although there is considerable variation and some patients remain in remission for a decade or more 94. Triggers of cholestatic episodes are not well defined but there is anecdotal correlation with viral infection and use of the contraceptive pill 94, 96. BRIC1 is less severe than PFIC1 and it is presumed that the mutations associated with BRIC1 have a milder impact on protein function, although this has not been demonstrated definitively.

Treatment of PFIC/BRIC disorders and how their success, or failure, relates to the pathophysiologcal process

Various therapeutic interventions have been used to treat PFIC1-3 and their BRIC counterparts, with mixed success. As mentioned earlier, UDCA is more hydrophilic, making it less hepatotoxic relative to the more abundant primary bile acids (CA and CDCA). When administered at high dose, UDCA competes for re-uptake in the small intestine and on return to the hepatocyte down-regulates de novo synthesis of primary bile acids, effectively supplanting them within the enterohepatic cycle 97. Such treatment has been temporarily successful in cases in which the detergent resistance of the canalicular membrane is likely to be impaired, such as PFIC1 and BRIC1 98 and, longer term, in cases of PFIC3 75. However, patients with PFIC2 and BRIC2, in whom the primary defect directly impinges on the canalicular efflux of bile salts, are not responsive to this treatment 99.

Rifampicin is also commonly used to promote bile salt efflux from hepatocytes into the blood for elimination in urine. Rifampicin up-regulates CYP34A, which hydroxylates bile salts, increasing their affinity for the basolateral transporter, ABCC4 100. In so doing, the net movement of bile salts through hepatocytes into the biliary tree is reduced and the canalicular membrane is exposed to a lower dose of detergent. Rifampicin can give temporary respite to PFIC1 patients and can also abort cholestatic episodes in both BRIC1 and BRIC2 patients 99, 101. However, it is largely ineffective in the treatment of PFIC2.

Finally, cholestyramine, a negatively charged resin that binds bile salts in the small intestine and promotes fecal excretion 102, has been tested in a number of different trials for the treatment of PFIC and BRIC diseases. Whilst it has been successful in reducing pruritus in some cases, positive changes in hepatic pathology are rare 99.

Surgical intervention has also been used with some success. In some cases of PFIC1, early non-transplant surgery may be considered and is also an option in other cases, where medicinal approaches have failed. This predominantly takes two forms, partial biliary diversion (PBD) and ileal bypass (IB). PBD is performed using a jejunal conduit between the gall bladder and either the abdominal wall for external drainage (partial external biliary drainage; PEBD), or the colon to promote fecal excretion (partial internal biliary drainage; PIBD). IB involves shortening the ileum and thereby reducing the small intestine's capacity for bile salt uptake. Both approaches reduce the total bile salt pool by promoting fecal excretion. In effect, this has some similarity with cholestyramine treatment; however, the outcome is generally much more successful. These approaches have been particularly successful in the treatment of low γ-GT diseases (ie PFIC1 and PFIC2), as indicated by improved liver function tests and reduced pruritus 99, 103, especially when performed early in the disease course before portal hypertension and cirrhosis have developed. For episodic cases, nasobiliary biliary drainage may be more appropriate, since it does not involve permanent change and it has been successful in aborting BRIC1 cholestasis 104.

In many cases the final, or only, option is liver transplant, which carries the obvious complication of potential rejection that requires long-term immunosuppression. In cases of PFIC1 and PFIC2, other complications have also been reported. As discussed above, ATP8B1 is expressed in several extrahepatic tissues, therefore ATP8B1-deficiency causes extrahepatic symptoms, such as diahrroea, that persist and in some cases worsen after liver transplant. This is possibly because when liver function resumes in PFIC1 transplant patients, bile flows into the small intestine, which is still devoid of ATP8B1 (and therefore detergent-sensitive). In keeping with this theory, the diarrhoea can be ameliorated by cholestryamine 105. Furthermore, the transplanted liver can develop steatosis, leading to further complications, such as cirrhosis 106. The precise cause of these reactions is currently unknown, although it has been speculated that there may be some correlation between mutation severity and post-operative outcome. Finally, in some cases of PFIC2, where the causative mutation completely abrogates ABCB11 expression at the plasma membrane, ABCB11 on the transplanted liver may raise an immune response, thereby renewing PFIC2 symptoms and liver pathology 107.

Why do we need ATP8B1 to flip PS at the canalicular membrane?

Two possible explanations have recently been put forward to explain the need for ATP8B1 at the canalicular membrane. These hypotheses may not be mutually exclusive.

The PS flippase is needed to control membrane fluidity and bile salt resistance

It is well established that the phospholipids in the plasma membrane are arranged asymmetrically. For example, the sphingolipids and PC are enriched in the outer leaflet, while phosphatidylethanolamine, phosphatidylinositol and PS are enriched in the inner leaflet (Figure 3). This lipid asymmetry and the high concentration of sphingolipids and cholesterol in the outer leaflet is thought to influence the lipid packing, fluidity and barrier function 108, 109 and is particularly important at the canalicular membrane, which is essentially under attack from bile salts that are effluxed into the canaliculus by ABCB11. ABCB4 flops excess PC into the outer leaflet, which might be expected to destabilize the bilayer but is extracted into the bile by the detergent activity of the bile salts. ABCG5 and ABCG8 also efflux cholesterol into the bile (see below), presumably for the same purpose—to form a bile salt/lipid mixed micelle and reduce the detergent nature of the bile salt. If a PS flippase is important for bile flow, one must expect PS to appear in the outer leaflet. The lipid and bile salt traffic at the canalicular membrane means that it is in a continuous state of flux and the ensuing instability may allow PS to flop spontaneously into the outer leaflet. Alternatively, it is possible that PS is flopped by ABCB4 directly, although there is no evidence for this, as analogues of PS were not tested in the most comprehensive study of ABCB4 substrate specificity 71. Whichever the explanation, the presence of PS in the outer leaflet will, along with the excess of PC flopped by ABCB4, dilute the relative concentration of sphingolipids and cholesterol, thus preventing the leaflet from forming a liquid crystalline phase. This would render the canalicular membrane more sensitive to detergent solubilization. We postulate, therefore, that ATP8B1 is required to re-internalize the flopped PS to maintain the liquid crystalline order of the canalicular membrane. In the ATP8B1-deficient state this does not happen and the membrane is progressively damaged by the liver's own metabolites.

Figure 3. Maintenance and role of lipid asymmetry at the canalicular membrane. Lipid asymmetry at the plasma membrane is known to be important for barrier function. Bile salts are exported into the bile canaliculus by ABCB11. The PC floppase ABCB4 flops PC into the outer leaflet of the canalicular membrane, where it is extracted into the bile by bile salts. Flopping of excess PC destabilizes the canalicular membrane, causing PS to flop spontaneously into the outer leaflet (alternatively, ABCB4 may flop PS directly). The dilution of sphingomyelin in the outer leaflet induces a liquid disordered phase and renders the canalicular membrane more sensitive to detergent solubilization. The PS flippase ATP8B1 is therefore required to flip PS back into the inner leaflet of the canalicular membrane to maintain its liquid crystalline order

Is there any evidence to support this theory? Data from our own laboratory and from that of our collaborator Ronald Oude Elferink strongly suggest that ATP8B1 is critically important to cell survival when ABCB4 is present. We observe that transient expression of ABCB4 in HEK293T cells is deleterious to the integrity of the membrane and cell survival [cells become leaky and release the cytoplasmic enzyme lactate dehydrogenase (LDH) into the culture medium; Figure 4A 84]. ABCB4 protein level in the transiently transfected cells is therefore low (Figure 4B). However, LDH release can be attenuated and ABCB4 expression rescued when ATP8B1 is co-expressed. These cells co-expressing the PC floppase and the PS flippase are even able to secrete PC in a bile salt-dependent manner (Figure 4C). Consistent with this observation, double knock-out mice that lack expression of Abcb4 and Atp8b1 have a milder phenotype than Atp8b1-deficient mice, suggesting that the phenotype of the latter is at least in part due to the cytotoxic effect of unchecked Abcb4 activity on the integrity of the canalicular membrane 84.

Figure 4. ATP8B1 is necessary for functional expression of ABCB4. (A) ABCB4 expression alone in HEK293T cells is cytotoxic causing lactate dehydrogenase (LDH) to be released from the cells. This is rescued by co-expression of ATP8B1 and CDC50. (B) Western analysis showing that transient expression of the ABCB4 PC floppase is only tolerated when the ATP8B1/CDC50 PS flippase is co-expressed. (C) Cells co-expressing the PC floppase and PS flippase can efflux PC to added taurocholate (for experimental details, see 84)

ATP8B1 is needed for microvillus formation at the canalicular membrane

A cellular morphological role for ATP8B1 has also been proposed recently by the groups of Leo Klomp and Joost Holthuis 110; 90% depletion of ATP8B1 by stable RNAi knockdown in polarized Caco2 cells had little effect on cellular PS-flipping activity (although this could be explained by the activity of the remaining 10%). On the other hand, a strong effect on cell morphology was observed. The organization of the actin cytoskeleton was deranged, leading to the loss of microvilli at the apical membrane and a reduction in expression of apical membrane proteins (Figure 5). While rescue of the phenotype by heterologous expression of an RNAi-insensitive ATP8B1 to control for off-site effects was not performed, the simplest conclusion is that ATP8B1 is necessary to either anchor the cortical actin cytoskeleton at the brush border, or, by flipping PS, to form the membrane invaginations around the microvilli. Microvilli are lost from the hepatocytes of Atp8b1-deficient mice 84, but this is also true of Abcb4-deficient mice 111, 112 and may simply be the downstream effect of cholestasis and ensuing liver damage.

Figure 5. ATP8B1 is required for brush border formation. Transmission electron microscope images of Caco-2 cell lines transfected with empty vector (A, C) or with a vector encoding a short hairpin RNA to knock-down ATP8B1 expression (B, D). The position of tight junctions and desmosomes are labelled with arrowheads and arrows, respectively, indicating that the ATP8B1-depleted cells remained polarized. A, apical membrane; BL, basolateral membrane; M, mitochondria; MV, microvilli; N, nucleus. Scale bars: (A, B) 1 mm; (C, D) 500 nm. Modified from 110

Regulation of bile salt homeostasis in the hepatocyte

Control of bile salt homeostasis at the genetic level is achieved through a set of nuclear hormone receptors: the farnesoid-X receptor (FXR), the retinoid-X receptor (RXR), the liver receptor homologue-1 (LRH-1), the small heterodimer partner (SHP) and the liver-X receptor (LXR) 3, 113. The rate-limiting step in bile salt synthesis is catalysed by CYP7A1 (cytochrome P450 cholesterol 7α-hydroxylase) and its function is regulated by the level of bile salts in the hepatocyte 114. Oxysterols in the liver bind RXR : LXRα heterodimers, which in turn up-regulate CYP7A1 expression, resulting in an increase in bile salt synthesis 115. Increased bile salts in the hepatocyte bind FXR : RXR heterodimers, activating FXR 116, which subsequently induces the transcription of SHP 117, 118. SHP down-regulates the expression of CYP7A1 and SHP by binding and inactivating LRH-1 117, 118. ABCB4 and ABCB11 are also targeted by FXR and their induction leads to increased export of bile salts and PC from the hepatocyte 119, 120.

Other ABC transporters of clinical importance at the canalicular membrane

ABCG5/G8: the cause of sitosterolaemia

ABGC5 and ABCG8 are ‘half-transporters’ that heterodimerize 121 to efflux both plant and animal sterols from hepatocytes into the bile (and also from enterocytes into the intestinal lumen). The function of ABCG5/G8 is therefore antagonistic to the activity of the sterol uptake transporter NPC1L1 122. Net sterol absorption results from the kinetic activities of these two transporters. In healthy individuals, net uptake of plant sterols is very low compared to animal sterols (<5% versus 55%, respectively 123–127). However, in sitosterolaemic patients, in whom either ABCG5 or ABCG8 is inactive 121, 128–130, the level of plant sterols in the plasma (β-sitosterol being the most abundant) is increased up to 100-fold 131–133.

Tendon xanthoma (most commonly on the Achilles tendon) is often the earliest clinical presentation and can develop within the first decade. Tendon xanthoma is also a feature of other lipid storage disorders, such as familial hypercholesterolaemia and cerebral tendon xanthomatosis, therefore misdiagnosis is a possibility. The defining feature of sitosterolaemic tendon xanthoma is the abundance of plants sterols within the lipid mass; however, to distinguish between plant and animal sterols in the laboratory, highly sensitive separation procedures, such as gas-phase chromatography, are required 134, 135. Several other pathologies are also commonly found in cases of sitosterolaemia, including arthritis, haemolysis, thrombocytopenia, splenomegaly and premature atherosclerosis 132, 136–141. It remains unclear whether elevated plant sterols or cholesterol is responsible. For example, in studies of patients with sitosterolaemia and in mice lacking either Abcg5 or Abcg8, there was no correlation between the level of plant sterols in the plasma and atherosclerotic severity 142. However, plant sterols have been reported to promote macrophage cell death, an important event in the latter stages of atherosclerosis 143. The current treatment of choice is ezetimibe, which inhibits sterol uptake via NPC1L1 and has been shown to successfully regress xanthomatosis and resolve cardiovascular disease 140.

ABCC2: the cause of Dubin–Johnson syndrome and role in clinical pharmacokinetics

The Groningen yellow/transporter-deficient (TR⁻) rat, a spontaneous mutant from the Wistar strain, is defective in excretion of conjugated bilirubin and amphiphilic organic anions into bile, and was first described in 1985 by Jansen and co-workers 144. The TR⁻ rats had lost an ATP-dependent transporter from their canalicular membrane as shown by comparison of inside-out vesicles from TR⁻ and normal rats 145. The transporter was subsequently identified as Abcc2 146. Cloning of the full-length rat cDNA was achieved in 1996 147, and its human homologue ABCC2 was cloned from a liver cDNA library the following year and shown to be expressed in the liver using immunohistochemistry 148. Various mutations in the ABCC2 gene (including missense and nonsense mutations, as well as deletions 149, 150) have been shown to cause Dubin–Johnson syndrome. First described in 1954, Dubin–Johnson syndrome is an autosomal recessive disorder characterized by hyperbilirubinaemia and abnormal urinary coproporphyrin excretion 151, 152. Serum bile salt levels are unaffected in these patients and the syndrome is generally regarded as relatively benign 153. Characterization of Abcc2^−/− knock-out mice revealed a similar phenotype, with elevated serum and urinary bilirubin levels, and reduced hepatobiliary excretion of conjugated bilirubin and glutathione 154. More recent studies using these mice have shown an increase in erythromycin metabolism, thought to result from increased hepatocyte residence time due to reduced biliary excretion of the drug 155. Furthermore, cross-breeding of Abcc2^−/− mice with other ABC transporter or cytochrome P450 knock-out animals is beginning to shed light upon the complex interplay involved in the detoxification and elimination of xenobiotics such as methotrexate 156, 157, lopinavir 158, docetaxel 159 and ezetimibe 160.

ABCB1 and ABCG2: role in clinical pharmacokinetics

In contrast to the other transporters described in this review, ABCB1 (previously known as MDR1 and P-glycoprotein) and ABCG2 (previously known as BCRP and MXR) are predominantly involved in the transport of exogenous xenobiotics. First identified as conferring multidrug resistance in colchicine-resistant Chinese hamster ovary cells in 1976 161, ABCB1 has since been estimated to recognize around half of all marketed pharmaceutical compounds. Similarly, ABCG2, which was first cloned from the adriamycin-resistant MCF-7/AdrVp human breast cancer cell line that lacked expression of ABCB1 and ABCC1 162, is known to produce a distinct but overlapping pattern of drug resistance to ABCB1 163. Expression of both ABCB1 and ABCG2 is localized to barrier sites and detoxifying organs, such as the luminal surface of intestinal epithelia and renal proximal tubule 164–166, blood–brain barrier 167, 168, ovaries, testes, placenta and stem cells 165, 169–171. In addition, both transporters are located on the hepatocyte bile canalicular membrane 164, 165, where they act to efflux drugs into the bile. Due to the inherent difficulty in obtaining bile drainage in healthy individuals, the true extent of biliary excretion of drugs in humans is difficult to assess. Instead, pioneering research, particularly from the laboratory of Alfred Schinkel 172–175, has used knock-out mice to demonstrate the importance of ABCB1 and ABCG2 in hepatobiliary excretion. In one example, cumulative biliary excretion of the fluoroquinolone antibiotics ciprofloxacin, grepafloxacin, ofloxacin and ulifloxacin in Abcg2^−/− mice were reduced to 86%, 50%, 40% and 16% of that seen in wild-type mice 176. Rodents express two ABCB1 paralogues, Abcb1a and Abcb1b, with overlapping but not identical patterns of expression. Abcb1a is highly expressed in brain and intestinal epithelium, whereas Abcb1b is highly expressed in the adrenal gland and ovaries 177. Both isoforms are expressed in the liver, kidneys, heart, lung, and spleen.

Biliary secretion of doxorubicin administered intravenously to Abcb1a^−/− mice was reduced by 80% versus wild-type animals 178, whereas in combined Abcb1a^−/−/1b^−/− knockout mice it was reduced by 90% 179. Inter-individual differences in the level of ABCB1 or ABCG2 expression within the liver, or polymorphisms affecting the protein sequence or folding 180, are likely to influence drug retention times and have important clinical consequences for therapy. Further research on hepatic drug transporters remains an important pharmacological and clinical objective and may have a contribution to make to personalized therapy.

Perspectives

Impairment of bile flow causes a wide range of hepatic disease, from the benign and episodic to the malign and fatal. The synergistic efforts of clinicians, geneticists, molecular biologists, biochemists and animal physiologists have provided insight into the enterohepatic cycle and have defined the pathways for biliary excretion of metabolites and xenobiotic compounds. The liver is replete with ABC transporters that are key to the flow of these compounds, particularly across the canalicular membrane. Two (ABCB4 and ABCB11) are critical, along with a P-type ATPase (ATP8B1), for bile flow across the canalicular membrane. We do not yet fully understand how the many mutations described in these genes contribute to the aetiology of the wide spectrum of liver diseases with which they are associated. This is particularly true of the more complex disorders. The type of in vitro analysis that has already been performed in Richard Thompson's laboratory for ABCB11 52 shows the way forward and has begun to define the relationship between transporter genotype, protein function and patient phenotype. The promise of therapeutic intervention tailored to the individual is thus a step closer.

Author contributions

All of the authors contributed to writing the manuscript.