The maintenance and generation of membrane polarity in hepatocytes.

Abstract

Hepatocytes, the major epithelial cells in the liver, are highly polarized. Their plasma membranes are separated by tight junctions into sinusoidal–basolateral and canalicular–apical domains, which contain distinct sets of proteins and lipids. A normal membrane polarity is vital for hepatocytes to perform many diverse functions, such as canalicular bile secretion and simultaneous sinusoidal secretion of large quantities of serum proteins into blood. Hepatocyte polarity is lost in many diseases like cholestasis. A complete understanding of the molecular mechanisms involved in hepatocyte polarity therefore is of considerable significance to both liver cell biology and the pathogenesis of liver diseases. To maintain the distinct distribution of proteins and lipids in the canalicular and sinusoidal membranes, hepatocytes have developed complex polarized trafficking and retention mechanisms. The establishment of membrane polarity in hepatocytes begins during liver embryogenesis. During hepatocyte differentiation, specific routes and mechanisms are defined for the delivery of plasma membrane proteins. How this polarity is maintained in adult hepatocytes and is generated during development are two major unanswered questions in this field. In this review, we first give an overview of the current knowledge about how membrane polarity is maintained in hepatocytes with the focus on recent developments in polarized protein trafficking. We then present a hypothetical model about how membrane polarity may be generated during liver development, and finally, we present an example of how membrane polarity is altered in cholestasis. Readers are directed to recent reviews for discussion of lipid trafficking in hepatocytes.1, 2 TGN, trans-Golgi network; DPP IV, dipeptidyl peptidase IV; pIgA-R, polymeric immunoglobulin A receptor; GPI, glycosyl-phosphatidylinositol; 5′NT, 5′-nucleotidase; MDCK, Madin-Darby canine kidney cells; ATP, adenosine triphosphate; Mdr1, multidrug resistance protein 1; Bsep, bile salt export pump; GFP, green fluorescent protein; Mrp2, multidrug resistant associated protein; LDL-R, low density lipoprotein receptor; AP, adaptor protein; SNARE, the soluble NSF attachment protein receptor; ERM, ezrin-radixin-moesin; TLCA, taurolithocholic acid. Three major intracellular trafficking pathways are involved in the polarized expression of plasma membrane proteins in hepatocytes. Through biogenesis, a plasma membrane protein is first synthesized in the endoplasmic reticulum, modified in Golgi, and finally sorted at the trans-Golgi network (TGN). From the TGN, they traffic along the post-Golgi biosynthetic pathway to the canalicular or sinusoidal membrane domain. After a protein reaches the plasma membrane, it can be either retained or internalized (endocytosed). After endocytosis, some proteins are transported to the opposite membrane domain of hepatocytes, a process known as transcytosis. During the past two decades, a key quest in liver cell biology has been to understand the trafficking routes and molecular mechanisms that are used by various canalicular and sinusoidal proteins as they journey along these pathways. Previous studies have demonstrated that proteins destined for the sinusoidal membrane in hepatocytes traffic from the Golgi complex to their destination directly (Fig. 1A).3 However, most canalicular proteins traffic by an indirect route, first to the sinusoidal membrane, followed by transcytosis through basolateral early endosomes and a subapical compartment, before final targeting to the canalicular domain (Fig. 1A). These canalicular proteins include single transmembrane proteins (such as dipeptidyl peptidase IV [DPP IV], aminopeptidase N, and polymeric immunoglobulin A receptor [pIgA-R]) and the glycosyl-phosphatidylinositol (GPI)-anchored proteins like 5′-nucleotidase (5′NT). Until recently, the indirect transcytotic pathway was thought to be the only route by which canalicular proteins reach their destination in hepatocytes. In contrast, most apical proteins traffic directly from TGN to the apical membrane in Madin-Darby canine kidney (MDCK) cells, a polarized kidney cell line, whereas both direct and indirect pathways operate in the retinal pigment and intestinal epithelial cells.4, 5 The post-Golgi biosynthetic trafficking and endocytic routes in hepatocytes. A hepatocyte couplet is shown with the basolateral–sinusoidal and apical–canalicular membranes separated by tight junctions (TJ). (A) Sinusoidal membrane proteins (blue circles) take a direct route (blue arrows) from the TGN to the basolateral membrane. Most canalicular proteins (green patches) are targeted initially to the basolateral membrane and then undergo transcytosis (green arrows) through the basolateral early endosome (BEE) and a subapical compartment (SAC) before reaching the canalicular membrane. (B) In contrast, Mdr1 (solid red patch) traffics directly (red arrow) from the TGN to the canalicular membrane. Bsep (hatched red patch) may traffic first to a subapical compartment (also termed as SAC) and then to the canalicular membrane. It remains unclear if Bsep and the transcytosed proteins traffic through the same SAC. (C) Some sinusoidal proteins recycle between the basolateral membrane, BEE, and SAC (dashed blue lines). The SAC functions as a common sorting station for both transcytosed and endocytosed proteins.65 Some sinusoidal proteins are targeted to late endosome (LE) and lysosomes (L) for degradation after endocytosis. At the canalicular membrane, most canalicular proteins are very slowly endocytosed with little recycling (dashed thin green lines). When an apical phosphoinositide 3 kinase is inhibited, some canalicular proteins (green patches) accumulate in lysosomes.46 (D) Mdr1 and Bsep both may recycle (dashed red lines) between the canalicular membranes and subapical compartment(s). The exact endocytic routes for Mdr1 and Bsep remain to be determined (indicated by ?). Possible intracellular compartments en route may include the apical early endosome (AEE), SAC, the apical recycling endosome (ARE), and lysosome for degradation. BC, bile canaliculi, N, nucleus. Recently, two canalicular adenosine triphosphate (ATP)–binding cassette transporters, the multidrug resistance protein 1 (Mdr1) and the bile salt export pump (Bsep), were shown to traffic directly from the Golgi to the canalicular domain (Fig. 1B). These findings were based on in vivo pulse-chase labeling in rats6 and expression of green fluorescent protein (GFP) fusion constructs in WIF-B cells, a polarized hepatic hybrid cell line.7 In addition to Mdr1 and Bsep, newly synthesized sphingolipids also reach the canalicular membrane directly from the TGN in hepatocytes.8 However, it remains unclear whether another major canalicular ATP-binding cassette transporter, the multidrug resistant-associated protein (Mrp2), which contains PDZ domains (a motif originally identified in postsynaptic protein PSD-95, Drosophila tumor suppressor Dlg and tight junction protein ZO-1; and used extensively for forming molecular scaffold as described below), also traffic directly from the Golgi to canalicular membrane. Indeed, a recent study demonstrated that Mrp2 resides in the same intracellular vesicles as pIgA-R, a canalicular protein that undergoes transcytosis.9 Much of our knowledge of the molecular mechanisms involved in polarized trafficking comes from work on epithelial cells other than hepatocytes.10 Direct basolateral targeting often involves cytoplasmic sorting signals, which can be either dileucine or tyrosine-based motifs.11 Two sinusoidal membrane proteins, the low-density lipoprotein receptor (LDL-R) and the sodium-taurocholate cotransporter, contain two tyrosine-based sorting signals.12, 13 Mutation of a critical glycine adjacent to the high-capacity tyrosine signal in LDL-R results in the mistargeting of LDL-R to the canalicular membrane in the WIF-B cells.14 Interestingly, β-turn structures, which are common to tyrosine-based basolateral sorting and endocytosis signals, are not present in the two tyrosine signals of sodium-taurocholate cotransporters.13 Basolateral sorting signals are believed to be recognized at the TGN by adaptor protein (AP) complexes, which are tetrameric protein complexes involved in both biosynthetic and endocytic sorting.15 Hepatocytes do not express μ1B, one important subunit of the AP-1B complex.16 μ1B is indispensable in the sorting of many basolateral proteins in kidney epithelial cells.17 It is not clear if hepatocytes use an unknown isoform of μ1 or a distinct AP complex to recognize the sinusoidal protein cargos. After they are bound to the AP complex, the sinusoidal proteins may be targeted and fused to the sinusoidal membrane, a process mediated in some epithelial cells by the soluble NSF (N-ethyl Maleimide-sensitive fusion protein) attachment protein receptor (SNARE) and an octomeric protein complex, which in yeast is called the exocyst-sec6/8 complex.18 The exact composition of the basolateral SNARE and exocyst complexes in hepatocytes and their roles in sinusoidal targeting have not been established. The canalicular transcytosing protein pIgA-R also takes a transcytotic route in MDCK and intestinal epithelial cells. Transcytosis of pIgA-R is stimulated by binding of immunoglobulin A and regulated by phosphorylation.19 pIgA-R contains a series of targeting signals within a cytoplasmic domain of 103 amino acids, which include a basolateral sorting signal and two tyrosine-based endocytosis signals.19 Notably, the basolateral sorting signal of pIgA-R is not dileucine or tyrosine based, and hence is different from the direct sinusoidal sorting signals discussed above. Different mechanisms other than the AP complex may determine the basolateral sorting of pIgA-R. The sorting signals in other single transmembrane proteins that undergo transcytosis are less well defined.20 Compared with direct basolateral sorting, the signals and machinery responsible for apical sorting are less well understood. Thus far, three types of apical sorting signals have been described: (1) GPI anchors, (2) transmembrane domains, and (3) N-glycans.10 In addition, the sphingolipid- and cholesterol-enriched domains, also known as lipid rafts, function as sorting platforms for apical cargos in the TGN.21 Lipid rafts are characterized by insolubility to detergent extraction at cold temperature. The enrichment of cholesterol and sphingolipid in canalicular membrane of hepatocytes has long been noted.22 Two recent independent studies demonstrate that lipid rafts mediate both transcytosis and direct apical trafficking of canalicular proteins in hepatocytes.23, 24 Slimane et al.23 showed that a GPI-anchored model protein, GPI-GFP, reaches the apical membrane of hepatic HepG2 cells by transcytosis and largely is present in lipid rafts, which are insoluble in both Triton X-100 and Lubrol WX at 4°C (Triton rafts). In contrast, two direct trafficking apical proteins, Mdr1-GFP and ATP7B (a P-type copper transporting ATPase), are contained in a distinct type of raft (Lubrol rafts), which are insoluble in Lubrol WX but are soluble in the stronger detergent, Triton X-100. Pulse-chase labeling indicated that both Triton and Lubrol rafts are formed at the TGN in HepG2 cells. Cholesterol depletion moderately increased the solubility of GPI-GFP to Triton X-100 without affecting its apical sorting. However, the same treatment led to missorting of Mdr1-GFP to the basolateral domain in HepG2 cells. Triton and Lubrol rafts also have been described in MDCK cells and segregate the GPI-anchored proteins from multispanning transmembrane proteins in the apical membrane.25 Nyasae et al.24 examined the solubility of a number of WIF-B cell apical proteins to Triton X-100 at 4°C. Two GPI-anchored proteins, 5′NT and CD59, are virtually insoluble to Triton X-100. In contrast, the single transmembrane proteins DDP IV and pIgA-R are completely soluble. Thus, not all of these canalicular proteins are contained in rafts, although they all use the indirect transcytotic pathway. Other mechanisms in addition to rafts must determine their indirect apical sorting. Moreover, the detergent-insoluble apical proteins seem to enter lipid rafts at different stages in the transcytotic pathway and with different kinetics.24 In this study, cholesterol depletion after 60 minutes severely impaired transcytosis of both raft-associated (CD59) and nonassociated proteins (DDP IV), presumably by blocking transcytotic flux from basolateral endosomes.24 Interestingly, cholesterol depletion did not affect the apical trafficking of GPI-GFP after 180 minutes in HepG2 cells in the study by Slimane et al.23 It is conceivable that the model protein GPI-GFP may not reflect faithfully the apical sorting of endogenous GPI-anchored canalicular proteins like CD59 and 5′NT. Despite this discrepancy, these two studies23, 24 together show that lipid rafts mediate the apical trafficking of some canalicular proteins in hepatocytes and that cholesterol is probably required for both Triton and Lubrol rafts. Furthermore, these two studies emphasize the complexity of membrane microdomains in the plasma membrane of hepatocytes. Judged by the sole criteria of detergent insolubility, the canalicular membrane consists of at least two types of rafts and a pool of single transmembrane proteins, some of which may be partly raft associated.23, 24 Lipid rafts also exist in the sinusoidal membrane of hepatocytes because some transcytosing proteins may have been incorporated into rafts before their transit to the sinusoidal membrane. In addition, caveolae, a special type of raft, have been detected at microvilli beneath the sinusoidal membranes of hepatocytes26 and contain raft-resident signaling proteins such as Ras and protein kinase C.27 Taken together, these studies suggest that both the canalicular and sinusoidal membranes can be viewed as a mosaic of raft-associated and nonassociated microdomains. A comprehensive analysis of detergent solubility of all canalicular and sinusoidal proteins will be required to appreciate fully the diversity of microdomains in the plasma membrane of hepatocytes and their functional implication. It also would be interesting to examine directly rafts and other microdomains in live hepatic cells using high-resolution fluorescence microscopy, which has been used successfully to examine raft-resident proteins in several cell types.28 Although it is not yet certain how raft mechanisms mediate the apical sorting of canalicular proteins, the work by Slimane et al.23 suggests that Lubrol rafts may segregate multispanning transmembrane proteins and may influence their direct apical sorting in HepG2 cells. It remains to be determined if other ATP-binding cassette proteins that traffic directly to the canalicular domain, such as Bsep, reside in the same type of lipid rafts. As discussed above, most GPI-anchored and single transmembrane proteins traffic directly from the TGN to the apical membrane in MDCK cells but indirectly via transcytosis to the canalicular membrane in hepatocytes. Why is there such a difference? Presumably, the plasma membrane and sorting mechanisms are organized differently between these two cell types. For example, secretory forms of DPP IV and pIgA-R (lacking the transmembrane regions) are secreted across the apical membrane in MDCK cells,29, 30 but across the basolateral domain in hepatic WIF-B cells.31 In hepatocytes, most proteins are secreted primarily across the sinusoidal domain. Canalicular proteins that undergo transcytosis seem to be sorted similarly at the TGN as the sinusoidal secreted proteins, that is, they are initially targeted to the sinusoidal domain. Basolateral endosomes also have been proposed to function as a sorting station for proteins that undergo transcytosis to the apical membrane in hepatocytes.31 In contrast, Mdr1 and Bsep must be sorted differently because they traffic directly to the canalicular domain. How these proteins exactly are sorted through the TGN remains to be determined. Transcytosis is not a phenomenon unique to hepatocytes because a number of apical proteins are delivered partly by transcytosis and partly by direct targeting to the apical membrane in intestinal epithelial cells.32 Transcytosis also seems to be regulated during epithelial differentiation. In Fisher rat thyroid cells, transcytosis functions as a major apical sorting mechanism for DPP IV in early epithelial differentiation, but is replaced by direct apical targeting in more differentiated epithelial cells.33 It is worth noting that few mammalian cells have robust protein secretion across the basolateral membrane like hepatocytes. Therefore, sinusoidal protein secretion, which is established early in hepatocyte differentiation and persists in adult liver, may be the reason why transcytosis is a major apical sorting mechanism for most canalicular proteins in adult hepatocytes. Interestingly, Bsep, a direct trafficking canalicular protein, is detected in rat liver only after birth,34 consistent with the notion that direct apical targeting may be established in hepatocytes at a later stage of differentiation. In contrast, Mrp2 is detected at the bile canaliculi before birth,34 consistent with the possibility that Mrp2 may be sorted via the transcytotic rather than the direct pathway. To understand fully the mechanistic difference in polarized trafficking between hepatocytes and other epithelial cells, a comprehensive comparison of the molecules and complexes involved in protein sorting will be required. For example, soluble NSF (N-ethyl maleimide-sensitive fusion protein) attachment protein-23, a key component of the SNARE complex, is present at both the apical and basolateral membranes of many epithelial cells but is absent from the canalicular membrane of hepatocytes.3 The apical SNARE complex mediates the docking and fusion of Triton rafts (containing GPI-anchored proteins) to the apical membrane in MDCK cells.35 It remains to be determined if the absence of soluble NSF attachment protein-23 in the canalicular membrane contributes to the preference of transcytosis as a major apical delivery mechanism in hepatocytes. Several sinusoidal membrane proteins, such as LDL-R, the transferrin receptor, and the asialoglycoprotein receptor, have a dynamic residence at the plasma membrane and undergo endocytosis (Fig. 1C).36 What about canalicular proteins? Tuma et al.37 recently demonstrated that several indirect pathway proteins (DPP IV, aminopeptidase N, and 5′NT) as well as MRP2 largely are retained at the apical membrane in differentiated hepatic WIF-B cells (Fig. 1C). However, these proteins recycle between plasma membrane and a novel intracellular compartment in less differentiated Fao and Clone 9 cells, which contain only would-be apical proteins. These data suggest that apical retention mechanisms are established during hepatocyte differentiation. In contrast, Mdr1-GFP and Bsep-GFP (which presumably reflect Mdr1 and Bsep) undergo dynamic endocytosis and recycle between the canalicular membrane and a subapical region when expressed in WIF-B cells (Fig. 1D).7, 38 Why is there such a difference in the dynamics of plasma membrane proteins in hepatocytes? First, the actin-based cytoskeleton is thought to be responsible for retention of domain-specific proteins.39, 40 Proteins like ezrin-radixin-moesin (ERM) bind to both actin and single transmembrane proteins directly41 and interact with multispanning proteins like Mrp2 via their PDZ domains.42 Therefore, ERM proteins mediate the formation of a molecular scaffold between apical membrane and actin cytoskeleton, which leads to the retention of these canalicular proteins. The degree of retention for single transmembrane proteins and Mrp2 may be different because disruption of actin filaments affected their localization differently in WIF-B cells.37 In addition, knockout of radixin in mice led to a significant decrease in the canalicular membrane expression of Mrp2 but not DDP IV, Mdr1, or Bsep.43 At present, it is not known how ERM or other retention mechanisms operate underneath the sinusoidal membrane. Second, dynamic membrane proteins like the transferrin receptor and raft-resident proteins are endocytosed by the clathrin- and caveolae-mediated pathways, respectively.44, 45 At present, it is not known which pathways mediate the fast apical endocytosis of Mdr1 and Bsep and the slow apical endocytosis of indirect pathway proteins and Mrp2. Several mechanisms may explain why Mdr1 and Bsep recycle rather than being retained in the canalicular membrane like Mrp2: (1) Mdr1 and possibly Bsep reside in distinct lipid rafts,23 which may confer different lateral diffusion properties, and (2) Mdr1 and Bsep do not posses PDZ motifs and presumably do not bind to tethering ERM proteins. Despite a tight apical retention mechanism, slow endocytosis and subsequent targeting to lysosomes may still function as a means to remove and degrade the indirect pathway proteins like aminopeptidase N (2%–3% degraded per hour37) and Mrp2 from the canalicular membrane. Inhibition of Vps34p, an apical phosphoinositide 3 kinase by wortmannin or injection of inhibitory Vps34p antibodies, led to accumulation of apical proteins like 5′-NT and MRP2 in lysosomal vacuoles in WIF-B cells.46, 47 Phosphoinositide 3 kinase generates secondary lipid messengers from inositol, which in turn regulates many steps of both biosynthetic and endocytic sorting. Thus far, we have considered how hepatic membrane polarity is maintained. How is such a complex polarity generated in hepatocytes during liver development and in particular how are the transcytotic delivery and apical retention mechanisms established? A number of electron microscopic studies have documented the ultrastructural changes in hepatocyte progenitor cells and hepatocytes at different stages of liver development.48-50 In chick embryos, the earliest liver parenchymal cells are not polarized.48 As the liver bud invades the perihepatic mesoderm, the parenchymal cells become polarized at day 5 when canaliculi-specific antigens are first detected. By day 7, the liver rudiment shows ultrastructural characteristics of bile canaliculi, and a canalicular network can be seen.48 From this point on, hepatocytes are polarized, but the bile canaliculi do not attain a fully mature structure until several days after birth.49, 50 Interestingly, in fetal hepatocytes, intracellular bile canaliculi have been observed50, 51 that may represent premature canalicular domains and may be related to the intracellular apical compartment in Fao cells described by Tuma et al.37 The development of membrane polarity is integral to the process of hepatocyte differentiation. To understand fully the molecular events involved in the generation of hepatocyte polarity, a systematic analysis of the trafficking machinery, cytoskeleton, and signaling molecules at different stages of hepatic differentiation are required. Based on recent work from other epithelial cells,52 we speculate that the genesis of hepatic polarity likely involves the events illustrated in Fig. 2. In nonpolarized hepatocyte progenitor cells, would-be sinusoidal and canalicular membrane proteins are sorted at the TGN into distinct transport vesicles by basolateral sorting signals and raft mechanisms (Fig. 2A). Because nonpolarized cells lack the necessary spatial segregation mechanisms, these proteins may be segregated into the microdomains, but these microdomains are mixed with each other in the plasma membrane. Some intracellular compartments may be formed and selectively may contain the would-be canalicular proteins, representing an intermediate stage before formation of the canalicular membrane. As the liver bud invades the perihepatic mesoderm, the polarization of hepatocytes is initiated by cell–cell or cell–extracellular matrix contact (Fig. 2B), or both, the former a process mediated by cell adhesion molecules like E-cadherin.53 Exocysts may be recruited to the sites of cell–cell contact and subsequently may contribute to the growth of domain-specific membranes.52 Cell–cell contact, cell–extracellular matrix contact, or both establish a membrane asymmetry or a spatial cue that in turn may trigger a number of events (Fig. 2C). As suggested by Yeaman et al.39 and Nelson,52 these include the establishment of apical junction complexes (tight junction and cadherin adherens junction), the restriction of signaling events, and polarized attachment of actin and microtubules, which then lead to the reorganization of the cytoskeleton in the cell. Microdomains carrying would-be canalicular and sinusoidal proteins form distinct patches on the plasma membrane, followed by the growth of apical and basolateral domains. Finally, the mature canalicular and basolateral domains are formed in hepatocytes (Fig. 2D). Microtubule and actin cytoskeleton also are established in a polarized orientation, shown in Fig. 2D. A hypothetical model for the genesis of membrane polarity during hepatocyte differentiation. Several stages of polarization occur during hepatocyte differentiation. (A) Initially, hepatocyte progenitor cells are not polarized. The would-be sinusoidal (blue circles) and would-be canalicular proteins (green and red patches for indirect and direct pathway proteins, respectively) are sorted at the TGN and are mixed in the plasma membrane. Some intracellular intermediate apical compartments may be formed that contain canalicular proteins. (B) As the liver bud invades the perihepatic mesoderm, polarization of hepatocytes is initiated by cell–cell adhesion. Cytosolic exocyst complexes may be recruited to the cell–cell contact sites, contributing to the subsequent growth of domain-specific membranes. (C) Cell–cell contacts define a membrane asymmetry, which triggers the formation of apical junctional complexes. The actin network (AN) attaches to the junctional complexes, leading to the reorganization and polarization of the cytoskeleton in hepatocytes. Would-be canalicular and sinusoidal proteins traffic to the plasma membrane (indicated by arrows) and form the canalicular and sinusoidal domains. (D) Finally, membrane polarity is fully established in mature hepatocytes. The plus and minus ends of microtubules (MT) are indicated. The subapical actin cytoskeleton network (AN) attaches to tight junctions and the canalicular membrane, forming as a bundle circumscribing the canalicular domain. BC, bile canaliculi; TJ, tight junctions. During this process, specific factors that induce liver differentiation may influence the expression and distribution of specific membrane targeting molecules (like soluble NSF attachment protein-23), retention molecules (like the ERM proteins), and the organization of cytoskeleton (like the subapical actin network). Collectively, these events establish the hepatocyte-specific mechanisms for polarized protein expression. For example, hepatocyte nuclear factor 4α and retinoic acid trigger the formation of tight junctions during visceral endodermal differentiation54, 55 and may function similarly in liver differentiation. Addition of chenodeoxycholic acid to cultures of Fao cells induces these nonpolarized cells to develop bile canaliculi and to manifest functional evidence of cell polarity.56 Thus bile salts, which are secreted only at the later stage of liver development, also may affect hepatocyte membrane polarity and the structure of bile canaliculi, possibly by signaling through nuclear hormone receptors. To assess better these complicated events at the molecular level, investigators will need to develop in vitro experimental models. Comparison of the gene expression profile, trafficking, and signaling events between the chenodeoxycholic acid-treated and untreated Fao cells56 at different time points during development of cell polarity may yield interesting information about when and how transcytosis and direct apical trafficking is established in relation to hepatocyte polarity. Loss of hepatocyte polarity is a characteristic feature of cholestatic liver injury. Cholestasis induced by bile-duct ligation or injection of ethinyl estradiol leads to a redistribution of canalicular proteins such as pIgA-R and DPP IV to the sinusoidal–lateral domain and accumulation in subapical cytoplasm (Fig. 3).57-60 Bile-duct ligation and ethinyl estradiol treatment also decreases the motility of transcytotic vesicles, concomitant with an increase in both the number and size of transcytotic vesicles within the subapical cytoplasm. Multiple intracellular lumina (pseudocanaliculi [PC] in Fig. 3) develop, which may result from fusion of transcytotic vesicles and may resemble bile canaliculi both morphologically and functionally.60 These intracellular pseudocanaliculi seem similar to the intermediate apical compartments observed in less differentiated hepatic cells and neonatal liver, as discussed above. In cholestatic liver, the tight junctions are also impaired, thereby increasing the permeability between bile and blood.61 Taken together, these observations suggest that cholestasis significantly impairs transcytosis of canalicular proteins and leads to a loss or even reversal of membrane polarity. The mechanisms responsible for altered cell polarity in cholestasis are incompletely understood. However, the ability of taurocholate to stimulate the trafficking of Bsep to the canalicular domain is likely to be impaired.62 Taurolithocholic acid (TLCA) increases in hepatocytes in cholestatic conditions and can reduce both transcytosis and the expression of Mrp2 in the canalicular membrane.63, 64 The anticholestatic reagent, ursodeoxycholic acid, reverses these inhibitory effects of TLCA. TLCA and ursodeoxycholic acid stimulate the translocation of PKC ε and PKC α, respectively, to the canalicular membrane, which may in turn affect the trafficking of Mrp2. It is conceivable that bile acids also may directly or indirectly affect the trafficking of other plasma membrane proteins in hepatocytes. Much further work is needed. Changes in membrane polarity and polarized trafficking in cholestatic hepatocytes. Cholestasis leads to the redistribution of canalicular proteins (green patches) to the sinusoidal and lateral domains, slowed movement of transcytotic vesicles (small red arrows), and the appearance of intracellular pseudocanaliculi (PC). Tight junctions also are disturbed, resulting a decreased permeability between the blood and bile. Intracellularly elevated taurolithocholic acids (TLCA) reduce (the ⟂ sign) the targeting and insertion of transcytosed proteins and Mrp2 to the canalicular membrane, concomitant with the translocation of PKC ε to the canalicular membrane. Ursodeoxycholic acids reverse (the ⟂ sign) the inhibitory effect of TLCA and stimulate translocation of PKC α to the canalicular membrane. Recent studies described in this review have provided significant insights into how sinusoidal and canalicular proteins traffic to and reside in the plasma membrane. However, the molecular mechanisms by which hepatocytes establish and maintain membrane polarity still remain incompletely understood. Clearly, more detailed characterization of the polarized sorting machinery is required. Perhaps by exploiting the models of hepatocyte differentiation, we may gain further insight into how hepatocyte polarity is generated and how the preference of hepatocytes to deliver proteins to the apical canalicular membrane by transcytosis is established. Historically, studies of polarized trafficking in hepatocytes have been difficult largely because of technical obstacles. Continuing developments such as those described in this review should lead the way toward a better understanding of this complex yet vital property of hepatocytes to develop and maintain their polarity in health and disease. The authors thank Drs. Dick Hoekstra and Pam Tuma for sharing unpublished information during the writing of this review and Drs. Carol Soroka and Kim Ng for helpful comments. We apologize to those whose primary papers are not cited because of lack of space.