Enhancement of Mdr2-mediated phosphatidylcholine translocation by the bile salt taurocholate. Implications for hepatic bile formation.

Expression of the Mdr2-protein in secretory vesicules (SVs) from the yeast mutant sec6-4 causes a time- and temperature-dependent enhancement of phosphatidylcholine (PC) translocation from the outer to the inner leaflet of the SV lipid bilayer. We show that this activity is independent of changes either in the membrane potential or the pH gradient (inside positive) generated in these SVs by the yeast proton-translocating PMA1 ATPase. However, loading of the SVs with the primary bile salt taurocholate results in an apparent enhancement of Mdr2-mediated PC translocation activity. Reducing the intravesicular taurocholate (TC) concentration by dissipating the electrochemical potential across the SV membranes eliminates the enhancing effect of TC. Three lines of evidence suggest that the enhanced Mdr2-mediated PC translocation activity is not caused by a regulatory effect of TC on Mdr2 but rather reflected the formation of TC/PC aggregates or micelles in the lumen of SVs. First, significantly higher detergent concentrations are required to reveal the fluorescence of (7-nitro-2-1,3-benzoxadiazol-4-yl)amino-PC molecules translocated in Mdr2-SV under conditions of TC stimulation than under control conditions; second, the nonmicelle-forming bile salt taurodehydrocholate does not cause enhancement of PC translocation in Mdr2-SVs; third, enzyme marker studies indicate that TC behaves as a potent lipid solubilizer directly extracting PC molecules out of the bilayer without causing leakage. This results in the formation of intravesicular aggregates or mixed micelles, and provokes the apparent stimulation of Mdr2 activity. These data demonstrate a unique relationship between Mdr2, PC, and TC in the process of bile formation and secretion.

Expression of the Mdr2-protein in secretory vesicules (SVs) from the yeast mutant sec6 -4 causes a time-and temperature-dependent enhancement of phosphatidylcholine (PC) translocation from the outer to the inner leaflet of the SV lipid bilayer. We show that this activity is independent of changes either in the membrane potential or the pH gradient (inside positive) generated in these SVs by the yeast proton-translocating PMA1 ATPase. However, loading of the SVs with the primary bile salt taurocholate results in an apparent enhancement of Mdr2-mediated PC translocation activity. Reducing the intravesicular taurocholate (TC) concentration by dissipating the electrochemical potential across the SV membranes eliminates the enhancing effect of TC. Three lines of evidence suggest that the enhanced Mdr2-mediated PC translocation activity is not caused by a regulatory effect of TC on Mdr2 but rather reflected the formation of TC/PC aggregates or micelles in the lumen of SVs. First, significantly higher detergent concentrations are required to reveal the fluorescence of (7-nitro-2-1,3-benzoxadiazol-4-yl)amino-PC molecules translocated in Mdr2-SV under conditions of TC stimulation than under control conditions; second, the nonmicelle-forming bile salt taurodehydrocholate does not cause enhancement of PC translocation in Mdr2-SVs; third, enzyme marker studies indicate that TC behaves as a potent lipid solubilizer directly extracting PC molecules out of the bilayer without causing leakage. This results in the formation of intravesicular aggregates or mixed micelles, and provokes the apparent stimulation of Mdr2 activity. These data demonstrate a unique relationship between Mdr2, PC, and TC in the process of bile formation and secretion.
Analysis of the biochemical basis of multidrug resistance (MDR) 1 led to the discovery of a group of membrane proteins, called P-glycoproteins or P-gps (1), which become overex-pressed in MDR cells (for review, see Refs. 2 and 3). P-gps are encoded by a small family of highly homologous mdr genes composed of three members in rodents, mdr1 (mdr1b), mdr2, and mdr3 (mdr1a) (4 -7), and two members in humans, MDR1 and MDR2 (8,9). Transfection experiments have shown that P-gps can be separated into two groups: the first one, which includes Mdr1 and Mdr3, and human MDR1 can directly confer drug resistance to drug-sensitive cells (6,10,11), while the second group, which includes mouse Mdr2 and human MDR2, fails to do so (5,6,11,12). Although several additional functions have been proposed for P-gp (13)(14)(15), recent controlled studies of P-gps expressed in the membrane of yeast secretory vesicles (16,17) or intact cells (18) strongly suggest that it functions as an energy-dependent drug transporter.
The various P-gp isoforms are expressed in normal tissues in a highly restricted cell-specific manner and are normally found only in apical domains of cells from polarized epithelia (19 -22). It was suggested that P-gps may fulfill a possible physiological role at these sites being involved either by protecting cells against attacks by environmental xenobiotics or by transporting yet to be identified normal cellular products. The normal physiological substrate(s) of P-gps has only been recently identified in the case of the Mdr2 isoform. Indeed, the phenotype of a mouse mutant bearing a null allele at the mdr2 locus is mainly restricted to pathological changes in the liver, which included severe necrotic damages of hepatocytes, strong portal inflammation, and as a consequence, proliferation and destruction of the canalicular and small bile ductular tracts (23). Interestingly, those changes were linked to a complete absence of phospholipids in the bile of those mutants (23). Consequently, Smit et al. suggested that the Mdr2 P-gp may act as a flippase for major phospholipid in normal bile transferring PC molecules from the inner to the outer leaflet of the canalicular membrane.
We tested this hypothesis directly and recently confirmed that Mdr2 can indeed function as a PC translocator (24). For these experiments, we used a heterologous P-gp expression system provided by the yeast mutant strain sec6 -4, which upon induction of a temperature-sensitive mutation in the normal biosynthetic secretory pathway, accumulates large amounts of unfused SVs (25,26). To study Mdr2-mediated phospholipid movements within a membrane bilayer, we established an assay that uses a fluorescent PC analog containing a fluorescent (7-nitro-2-1,3-benzoxadiazol-4-yl)amino (NBD) moiety (24). This compound was chosen because this NBD group can be chemically reduced by dithionite to a nonfluorescent PC derivative. Since the reducing anion dithionite is membrane impermeable, it was possible to determine accurately the asymmetric distribution of the labeled lipid (27). Using this experimental system, we observed in SVs from Mdr2-expressing yeast cells a time-and temperature-dependent translocation of the marker lipid from the outer to the inner leaflet of the SV membrane. This enhanced activity was strictly ATP-and magnesium-dependent and completely abolished by the addition of vanadate or verapamil. This activity was specific to the Mdr2 isoform of P-gps and was not detected in SVs from yeast cells expressing Mdr3. Finally, the Mdr2 activity was specific for the phospholipid PC since neither phosphatidylethanolamine (PE) nor phosphatidylserine (PS) were substrates for Mdr2.
PC is normally found in the bile complexed with bile salts (e.g. taurocholate, glycocholate) and cholesterol (for review, see Refs. 28 and 29). The complexing of bile salts by PC and cholesterol into mixed micelles is thought to greatly reduce the otherwise cytotoxic detergent effect of bile salts traveling within the biliary tree (30). However, the observation that the rate of secretion of bile salts and biliary lipids are intimately related (29) has suggested that bile salts may also play a direct positive regulatory role in the transhepatic movement of biliary lipids (31). The mechanism responsible for such a promoting effect of bile salts remains unclear, but it has alternatively been proposed to involve direct solubilization of PC from the canalicular membrane (31,32), modulation of either phospholipid biosynthesis (33) or of intracellular trafficking of PC-rich vesicles from the endoplasmatic reticulum or the Golgi apparatus (34,35), or finally through direct activation of a PC transporter (Mdr2) in the canalicular membrane (23,24 Yeast Strain and Culture Conditions-Cells of the Saccharomyces cerevisiae strain SY1 (MATa, ura3-52, leu2-3, 112, his4 -619, sec6 -4 GAL) were transformed with the plasmid pVT (control cells) or pVT containing full-length cDNAs from mouse Mdr2 (pVT-mdr2) as described previously (17). Transformants were selected on supplemented minimal plates containing 2% glucose lacking uracil. Several independent clones were pooled and grown at 25°C in YPD (1% yeast extract, 2% Bactopeptone, 2% glucose) up to midlogarithmic phase (A 600 of 1.5-2.0). Subsequently, the temperature was shifted to 37°C, and cellular accumulation of unfused SVs was allowed to proceed by a further 2-h growth period. After addition of 10 mM NaN 3 , the cells were collected by centrifugation (8,275 ϫ g, 10 min), washed once in 10 mM Tris-HCl, pH 7.5, 5 mM NaN 3 , and the cell pellets were stored at Ϫ70°C in 3-5-g aliquots.
Preparation of Secretory Vesicles containing NBD-PC. SVs from control and Mdr2-expressing cells were isolated exactly as described previously (17). The pelleted SVs were resuspended in the appropriate vesicle buffer (VB) (17) to obtain a relative fluorescence emission signal of about 20 -22 relative fluorescent units in a 100 l of SV aliquot.
Phospholipid Translocation Assay-Briefly, an aliquot of NBD-PC containing SVs from control and Mdr2-expressing cells in the appropriate VB supplemented with creatine phosphokinase (3 g/ml) was mixed with an equal aliquot of the appropriate VB containing 5 mM NaATP, 10 mM Mg 2ϩ gluconate, and 20 mM creatine phosphate (24). Where indicated, 5 M TC or TDHC were added. After immediate removal of a 100-l aliquot for the 0 min time point, lipid translocation was initiated by keeping the SV suspension in a 37°C water bath. After the appropriate incubation time, another 100-l aliquot was removed and placed into a fluorescence cuvette containing 1.9 ml of the appropriate VB (25°C) supplemented with 2.5 mM EDTA and 10 M verapamil. Fluorescence emission by NBD-PC associated with the outer and inner leaflet of the membrane bilayer of SV was then recorded at a wavelength of 540 nm using an excitation wavelength of 470 nm. Solutions in the cuvettes were continuously mixed with a magnetic stirrer. After an initial stable fluorescence emission was obtained, 25 l of a solution containing 1 M Na 2 S 2 O 4 in 1 M Tris, pH 10, was added to chemically reduce the fluorescent NBD associated with outer leaflet of the SV bilayers. The decrease of the fluorescence signal was monitored for 5-6 min until a new stable emission base line was recorded (usually not shown in the figures). Finally, NBD-PC fluorescence associated with the inner leaflet of the SV bilayers was revealed after disrupting the vesicles by addition of 80 l of a 12.5% (w/v) Triton X-100 stock solution followed by further monitoring the disappearance of the fluorescence signal. The amount of NBD-PC translocated (pmol/mg of protein) was calculated as described previously (24).
Analytical Methods-[ 3 H]Tetraphenylphosphonium ([ 3 H]TPP) uptake was measured as described previously (17) in the presence or absence of 5 M TC or TDHC, respectively. Briefly, 10-l aliquots of SVs from Mdr2-expressing cells resuspended in the appropriate VB supplemented with creatine phosphokinase (3 g/ml) were incubated with 90 l of prewarmed (37°C) transport buffer (transport buffer/Glu containing 50 mM sucrose, 100 mM potassium gluconate, 10 magnesium sulfate, 10 mM creatine phosphate, 2.5 mM NaATP, 10 mM Tris-HEPES, pH 7.5, whereas in the transport buffer/NO 3 buffer, gluconate was replaced by 100 mM potassium nitrate) containing [ 3 H]TPP (20 Ci/mmol) at a final concentration of 1 M. Where indicated, transport buffers were supplemented with 5 M TC or TDHC. After a 15-min incubation period at 37°C, transport was terminated by the addition of 2 ml of an ice-cold stop solution consisting of 200 mM sucrose, 10 mM Tris-HEPES, pH 7.5. The vesicles were collected by filtration through a 0.65-m nitrocellulose filter (Type HA, Millipore). After two additional washes with 2 ml of cold stop solution, vesicle-associated radioactivity was determined by liquid scintillation counting. Similarly, we determined [ 3 H]TC uptake as described previously (37). Invertase activity was determined in 25-l aliquots for 10 min at 37°C according to Goldstein and Lampen (25); units of activity are mol of glucose/min. Relative protein concentrations in SV fractions were determined according to Bradford in the presence of 5% formic acid (36). RESULTS We have used SVs purified from the yeast S. cerevisiae mutant sec6 -4 to study and characterize the mechanism of action of various members of the mouse P-gp family that we successfully expressed in the membrane of these vesicles (17). Also inserted in the membrane of these SVs are newly synthesized PMA1 H ϩ /ATPase molecules, which in the presence of ATP and magnesium, concentrate protons into the lumen of SVs (17,26). In a buffer system (at physiological pH) containing a membrane impermeable anion such as gluconate (VB/Glu), this proton translocation results in the formation of a strong pH gradient across the SV membrane of ϳ 0.6 -0.8 pH units and consequently, causes a significant polarization of the luminal membrane surface (17). We have observed that these SVs can readily accumulate amphiphilic anions, such as bile salts, when included at low concentrations (5 M) in vesicular buffer. Under normal conditions (VB/Glu buffer), the primary bile salt TC was concentrated by a factor of 10 -13-fold reaching, within a time period of 10 min, intravesicular concentrations of ϳ55-65 M (37). Replacing in the buffer system the membraneimpermeable anion gluconate by the free permeable anion nitrate (VB/N0 3 buffer), the intravesicular concentration of TC was significantly reduced (ϳ22 M). The addition of nigericin in VB/N0 3 buffer did not further reduce intravesicular accumulation of TC, although it completely dissipated the electrochemical potential (17). The remaining ⌬ H ϩ-insensitive component of TC accumulation in SVs was found to be strictly ATP-dependent, indicating that yeast plasma membranes may possess a potential-insensitive, ATP-dependent organic anion transport system (37).
The specific lipid flippase activity of the Mdr2 P-gp isoform is measured in SV preparations using a fluorescent PC analog containing a NBD group, which can be chemically reduced to the nonfluorescent 7-amino-2-1,3-benzoxadiazol-4-yl compound by dithionite (27). Dithionite is membrane impermeant, and thus, it is possible to determine the asymmetric distribution of labeled lipid molecules within the SV bilayer. Upon addition to intact SVs, dithionite will reduce only those NBD groups of labeled PC molecules, which are present in the outer leaflet of the bilayer while leaving unaffected those either associated with the inner leaflet or present within the luminal space (27). Once this first reaction is completed (i.e. no further decrease in fluorescence emission over time), the vesicles are disrupted with Triton X-100, and further loss of remaining fluorescence emission (second reaction) is measured. The detergent-sensitive fraction of total NBD fluorescence directly reflects the amount of translocated NBD-PC associated with the inner leaflet of the bilayer.
We took advantage of this experimental system, in which both the intravesicular concentration of TC can be manipulated and the Mdr2 lipid flippase activity can be measured, to monitor the effect of the bile salt TC on Mdr2-mediated PC translocation. As shown in Fig. 1A, preloading SVs with TC was without effect on the amount of unspecific NBD-labeled PC translocation observed in SVs isolated from control cells. In contrast, the addition of TC to SVs from Mdr2-expressing cells resulted in a significant increase in the amount of fluorescent NBD-PC molecules remaining in the detergent-sensitive fraction of these SVs (Fig. 1B). Since the amount of dithionite was sufficient to chemically reduce up to 5 times more NBD moieties than routinely used in these experiments (24), these findings suggested a 2-3-fold increase in the number of NBDlabeled PC molecules translocated within the bilayer and thus associated with the luminal side of the membranes of Mdr2expressing SVs. The enhanced NBD-PC translocation activity noticed in the presence of TC did not affect the integrity of the SV membranes since, after reduction of all NBD groups present in the outer leaflets, the emission profiles reached a new stable base line (Fig. 1B). After disruption of the SVs membranes by the addition of detergent, the total fluorescent signal could be completely abolished (Fig. 1B). Taken together, these observations indicate that a direct or indirect interaction may exist between bile salts and Mdr2-mediated PC translocation.
Further experiments were carried out to establish the biochemical basis of the identified activation of Mdr2-mediated PC translocation by TC. Experiments summarized in Fig. 2 indicated that the enhancing effect of TC was dependent on sufficient concentration of TC within the lumen of SVs. Indeed, when the VB/NO 3 buffer system was used in the translocation assays, the addition of TC had no effect on the PC translocation activity of Mdr2. Since this latter condition significantly decreases the intravesicular accumulation of TC (37), while leaving unaffected the lipid translocation capacity of Mdr2 (Fig. 2), these results indicate that sufficient amounts of TC are required in the luminal space of SVs to cause the apparent enhancement of Mdr2-mediated NBD-PC translocation activity.
Two possible mechanisms can be put forward to explain this phenomenon. The first one would be based on a direct regulatory effect of TC on the activity of Mdr2 through some interaction at the trans, intravesicular side of Mdr2. The second would be through a more indirect "trapping" effect of TC reflecting its micelle-forming properties. Three experiments were carried out to distinguish between these two hypotheses. First, NBD-PC translocation (in the presence or absence of TC) was induced in Mdr2-expressing SVs, and the fluorescent moieties of lipid molecules incorporated in the outer leaflets were reduced by addition of dithionite as described above (Fig. 1). The characteristics of the remaining emission signal were then analyzed by stepwise additions of increasing concentrations of Triton X-100 (Fig. 3). In control Mdr2-SVs unexposed to TC, the addition of 0.0625% Triton X-100 was sufficient to allow complete reduction of the SV-associated fluorescence (Fig. 3, left  panel). This clearly demonstrated that the presence of detergent in subsolubilizing concentration was in fact sufficient to make the Mdr2-SVs leaky, and consequently, the reducing agent could modify the labeled lipid molecules in the inner leaflets. In contrast, the addition of equivalent amount of detergent to the TC-treated Mdr2-SVs caused only a 50% reduction in the fluorescent signal, which then reached a new stable base line within 1 min (Fig. 3, right panel). Increasing the detergent concentration to 0.125% resulted in a further reduction of the fluorescence signal, which reached yet a new stable base line significantly above background emission. Finally, raising the Triton X-100 concentration to 0.5% completely eliminated the fluorescence signal. These observations suggest that the increased amount of NBD-PC translocated by Mdr2 under conditions of TC stimulation is in a physical state (either within the membrane or in the luminal space of Mdr2-SVs), which requires higher amounts of detergent to expose the fluorescent NBD-moiety to the reducing action of dithionite.
To rule out the possibility that the presence of high intrave-sicular concentrations of TC may alter unspecifically the membrane properties of SVs, additional control experiments were carried out (Fig. 4). NBD-PC containing SVs from control and Mdr2-expressing cells were preloaded with TC in VB/Glu buffer, and the effect of increasing concentrations of Triton X-100 on the integrity of SV membranes was estimated by monitoring the appearance of the luminal enzyme marker invertase in supernatant fractions of treated SVs. We also measured the effect of increasing Triton X-100 concentrations on the amount of total proteins associated with the SVs, and on the NBD-PC fluorescence of the SVs (Fig. 4). While at a low concentration (0.01%) of detergent, a significant increase in invertase activity was detectable in the medium fraction, a maximal release of enzymatic activity was reached only at detergent concentrations of 0.05-0.1%. These results suggest that in the concentration range of 0.01-0.1%, Triton X-100 only induces leakiness of the vesicles, since the amount of total protein as well as the recovery of labeled PC in the SV fractions both remained unchanged. The addition of higher amounts of detergent (up to 0.25%) caused solubilization of the SV membranes as demonstrated by a progressive loss of total protein and NBP-PC fluorescence associated the pellet fractions. Identical effects of Triton X-100 on SV membranes integrity as measured by total protein loss from the pellet and appearance of invertase in the supernatant were observed from Mdr2-SVs prepared in the presence (Fig. 4) or absence of TC (data not shown). These results suggest that TC itself or the enhanced NBD-PC translocation detected in Mdr2-SVs do not affect the integrity of the SV membrane. They also indicate the that the amount of 0.0625% Triton X-100 concentration used in the experiments shown in Fig. 3 is completely sufficient to render the Mdr2-SVs leaky (invertase loss). Thus, one likely explanation for the results shown in Figs TC-dependent NBD-PC translocation in SV from Mdr2-expressing cells was carried out for 15 min at 37°C, as described under "Experimental Procedures." Subsequently, 250-l aliquots (containing ϳ200 g of protein) were diluted into 10 ml of cold VB/Glu-buffer and supplemented with increasing concentrations (0 -0.25%) of Triton X-100. After a 10-min incubation at 0°C, the samples were centrifuged (100,000 ϫ g, 30 min), the supernatants collected, and the resulting pellets resuspended in 10 ml of cold VB/Glu-buffer. Invertase activity was measured in 50 l of the each supernatant fraction, whereas in each resuspended SV pellet fraction, the total protein content as well as the relative fluorescence emission of NBD-PC was determined. Sample designated as B represents an untreated aliquot, which was used for comparison and standardization (protein values are expressed as percent of this sample).
tures to finally expose the labeled lipids to dithionite.
To further test this hypothesis, we substituted TC with the known nonmicelle-forming triketo TC analog, TDHC (38,39). As shown in Fig. 5, when analyzed under normal as well as clamped conditions (VB/NO 3 ), the addition of 5 M TDHC to Mdr2-SVs labeled with NBD-PC did not cause the enhanced PC translocation effect previously observed with TC (Fig. 1). Likewise, TDHC had no effect on NBD-PC translocation in SVs from control cells (Fig. 5). Since the TC-mediated enhanced PC translocation in Mdr2-SVs is strictly dependent of a high intravesicular concentration of TC (Fig. 2), it was important to determine whether TDHC accumulation in the lumen of SVs was similar to that obtained with TC under identical experimental conditions. As radiolabeled analog of TDHC is not commercially available for direct uptake measurements, two indirect approaches have been used to assess the degree of intravesicular accumulation of TDHC. First, we monitored the effect of TC and TDHC on the relative distribution of TPP under different experimental conditions. TPP is a lipophilic cation that is membrane permeable and readily equilibrates across biological membranes in response to membrane potential (40). We have previously used this compound as a membrane potential marker to characterize the membrane potential across yeast SV membranes (17). In these studies, we showed that under polarized conditions, [ 3 H]TPP was incorporated into the SV membranes in low amounts, since it is repulsed by the highly positively charged inner membrane surface. Dissipating the membrane potential by replacing gluconate by the freely permeable anion nitrate resulted in increased [ 3 H]TPP uptake by a factor of 4 -5-fold (17). Since we postulated that TC uptake under polarized conditions was caused by a strong compensating effect provoking the amphiphilic anion to accumulate in SVs, we determined whether this could also be verified using [ 3 H]TPP distribution as a probe. As shown in Fig. 6A, the addition of TC to a VB/Glu-buffer system caused a 2-3-fold increased uptake of [ 3 H]TPP in Mdr2-SVs (second column, VB/Glu ϩ TC) compared with control conditions in the absence of TC (first column). The addition of TDHC to Mdr2-SVs caused a similar increase in [ 3 H]TPP accumulation (third column, VB/Glu ϩTDHC). Abolishing the membrane potential by replacing gluconate by the permeant anion nitrate in the buffer system (VB/NO 3 ) resulted in a further slight increase of [ 3 H]TPP uptake (Fig. 6A, columns 4 and 5). Secondly, we determined whether the uptake of the tracer [ 3 H]TC could be uncoupled by excess of TDHC. As shown in Fig. 6B, in the presence of 5 M TDHC, TC accumulation was significantly reduced compared with control conditions (in the presence of 5 M TC). Taken together, these results indicate that under polarizing conditions TC, and the non-micelle-forming bile salt TDHC both quench positive charges in the lumen of SVs to same extent, in agreement with a comparable accumulation of both bile salts in the lumen of SVs. This similar degree of accumulation in Mdr2-SVs together with the lack of TDHC stimulation of Mdr2-mediated PC translocation strongly support the proposition that the TC-enhancing effect of PC translocation is due to the active formation of TC/NBD-PC aggregates or micelles.
We postulate that TC could specifically enhance removal of NBD-labeled PC molecules from the inner lipid layer followed by effective trapping through the formation of some kind of NBD-PC/TC aggregates or micelles. One would expect a similar effect occurring at the cis side of SVs if TC molecules were added to the buffer system. To test this prediction, NBD-PClabeled SV were incubated in the absence of ATP and magnesium cations and increasing concentrations of TC. We monitored the relative amounts of the remaining fluorescent signals of each sample after centrifugation through a 10% Ficoll cushion (Fig. 7). Up to a TC concentration of about 40 M in the external medium, only a small loss of NBP-PC from the pelleted SV fraction was observed. Above 45 M, the recovery of labeled NBD-PC molecules incorporated in SV membranes started to decline sharply, reaching a new plateau after a loss of approximately 100 -110 pmol of NBD-PC molecules/mg of protein, corresponding to about 25% of the initial fluorescence. Interestingly, comparable amounts of labeled lipid molecules were removed by TC from the inner leaflet of SV bilayers. We determined that the extent of depletion was approximately 40 -85% using for the calculations for the internal TC concentration either 65 or 50 M. Furthermore, the observed depletion of the labeled lipids from SVs seemed to be caused by TC as opposed to a nonspecific effect of membrane permeability. Indeed, the addition of TC in the 0 -100 M range did not affect the integrity of the membranes since no increase in invertase activity was detectable under these conditions (Fig. 7). In addition, TDHC added at 75 M to the Mdr2-SVs produced only a minor loss of NBD-PC from the membranes (less than 6% of the initial amount) (Fig. 7A). These data further support the hypothesis that the primary bile salt TC in concentration as low as 50 M can cause removal of NBD-PC molecules from the exposed lipid layer of the membrane resulting in the formation of NBD-PC/TC aggregates. DISCUSSION A primary function of hepatic bile is the emulsification of dietary fat mediated by the detergent action of its bile salt constituents. Bile salts are secreted by hepatocytes by specific transport systems (41) into bile canaliculi where they quickly form mixed micelles with PC and cholesterol (28,29). This physical state of bile salts prevents deleterious detergent action on membrane of epithelial cells of the canalicular and ductular tract they have to travel through (30). The process of PC transport across the canalicular membrane has remained poorly understood until recently, and a direct or indirect role of bile salts in this event has been proposed but remains controversial. The study of a mouse mutant bearing a null allele at the mdr2 locus initially suggested that the P-gp isoform encoded by this gene may be implicated in PC transport across the canalicular membrane (23). We subsequently demonstrated that Mdr2 encodes a PC translocase (flippase) that creates an asymmetric distribution within the canalicular membrane (24). Similar results were subsequently reported by Smith et al. (42).
The aim of the current study was to analyze possible func-tional relationships between the primary bile salt taurocholate and PC translocation mediated by Mdr2. The biochemical function of Mdr2 was identified and characterized in SV from the temperature-sensitive yeast mutant sec6 -4. Using the same SV preparations, we recently observed that in the presence of a strong electrochemical potential the primary bile salts tauro-or glycocholate actively accumulated intravesicularly, reaching a final concentration of 55-65 M (37). We combined these two observations to study in the same experimental system possible effects of bile salts on Md2-mediated NBD-PC translocation. We observed that TC caused a significant increase in the amount of NBD-PC translocated by Mdr2. Intravesicular concentrations of TC above 50 M were required to enhance PC translocation, as clamping the membrane potential significantly decreased the intravesicular TC accumulation (37) and completely abolished enhanced PC translocation (Fig. 2). Control experiments indicated that neither the increased amount of NBD-PC translocated to the inner leaflet of the SV bilayer nor the intravesicular accumulated TC affected overall membrane integrity of the SVs (Fig. 5). Several lines of evidence suggest that the enhancing effect of TC on Mdr2-mediated PC translocation was not due to modulation of the enzymatic activity of Mdr2 but rather reflected the formation of mixed micelles or larger aggregates within the SV lumen (Fig. 3). First, we observed that a significant portion (about 50%) of labeled lipid NBD-PC molecules translocated by Mdr2 under TC stimulation could only be reduced by dithionite with higher concentrations of Triton X-100 (Fig. 3), compared with control conditions (absence of TC) where total translocated NBD-PC fluorescence was completely revealed at a subsolubilizing Triton X-100 concentration (0.0625%). In addition, under TC-activated conditions the initial loss of the fluorescence signal caused by 0.0625% Triton X-100 was similar to that observed in control conditions (no TC). Second, the stimulating effect of TC on PC translocation by Mdr2 was specific to the micelle-forming activity of TC since the known non-micelle-forming bile salt TDHC had no enhancing effect on lipid translocation. Third, we observed that increasing concentrations of TC in the medium solublized NBD-PC out of the outer leaflets of the bilayers of SVs without affecting the membrane integrity. After separation of TC-treated SVs through a Ficoll cushion, we determined in the supernatant fraction the appearance of both NBD-PC lipids but also some type of aggregates at the interface of the gradient (data not shown). In the presence of the non-micelleforming TDHC, neither a loss of NBD-PC from the SV fraction nor appearance of labeled lipid in the supernatant overlying the Ficoll cushion was observed. Our findings are compatible with the possibility that the high concentration of TC causes a constant and selective removal of NBD-PC molecules from the inner lipid leaflet of the SVs followed by the incorporation into the preformed bile salt aggregates. According to this proposal, the amount of NBD-PC in the inner membrane leaflet would be kept in a subsaturating concentration and, consequently, would result in an apparent activation of Mdr2. The total amount of NBD-PC translocated and/or possibly trapped by TC in the vesicular lumen would be above the maximal levels normally reached within 10 -15 min in Mdr2-expressing SVs under control conditions (24).
An important role of bile salts in induction of bile flow and biliary lipid secretion has been established (28,29). Although the secretion of bile salts, biliary phospholipids, and cholesterol occur via separate processes (39), the former is required for the latter to occur (31). As all amphiphilic molecules, bile salts tend to self-associate in water with increasing concentration (44). But in contrast to typical anionic detergents, which aggregate abruptly at a characteristic critical micellar concentration (CMC), the self-aggregations of bile salts occurs over a broad concentration range (28,45). For example, the CMC value of cholate is approximately 13 mM (45), but even at lower concentrations, cholate forms aggregates as small as dimers or oligomers (46). Another important property of bile salts is to solubilize phospholipids and sterols, a property that itself causes a dramatic decrease of their CMC value (28). For example, mixing diacylphosphatidylcholines (lecithin) with cholate in a molar ration of 1:27 results in a dramatic change of the CMC of cholate, which is further reduced at near saturation (molar ratio of 0.6:1) to a CMC value of 200 M. These unique physicochemical properties of bile salts are of paramount importance for 1) their physiological role as powerful detergents in digestion and absorption of dietary lipids but also 2) in the formation of aggregates with biliary phospholipids and cholesterol to prevent their cytotoxic detergent effect toward cells of the hepatobiliary system directly exposed to the bile fluid.
The underlying molecular mechanism by which bile salts induce biliary lipid secretion is poorly understood and the issue of an on-going debate. Bile salts are transported across the canalicular membrane as individual anionic molecules by two independent transport systems (41). In the canalicular lumen, they reach concentrations of at least 1-2 mM and most likely start to form simple aggregates and micelles (28, 29, 44 -46). The proposed sequence of events leading to the formation of mixed micelles or vesicles containing phospholipid and cholesterol is more speculative, and several models have been proposed. One of them is the fusion-budding model proposed by Coleman et al. (29) (see Fig. 8). This model suggests that as bile salts reach micelle-forming concentrations, they may intercalate into microdomains of the canalicular domain with enhanced membrane fluidity. These domains have been proposed to represent the fusion points within the canalicular membrane into which the membranes of vesicles fused bringing biliary lipids from the cell (47). Such supply vesicles are thought to derive directly from the Golgi and contain unusually high amounts of PC (29). As a consequence of the increased intercalation of bile salts with such areas, these regions of the membrane would curve outward, eventually pinching off as vesicles. In agreement with this model is the observation that colchicine, which inhibits microtubule assembly and blocks intracellular vesicle movements, abolishes bile salt-dependent biliary phospholipid secretion (39). Also in support of this model is the observation that, in contrast to TC, retrograde injection of the non-micelle-forming bile salt TDHC back up the biliary tree does not release biliary lipids (32).
However, the recent demonstration that the Mdr2-Pgp isoform actively translocates PC within the membrane bilayer (24) together with the data presented in this report suggest an alternative mechanism for the initial step of hepatic bile formation in which the Mdr2 flippase may play a fundamental role. This flippase would actively translocate lipid molecules within the canalicular membrane bilayer creating an asymmetric gradient with a higher concentration in the outer leaflet facing the canalicular lumen, therefore facilitating the extraction/removal of PC by bile salts. The physiological importance of the formation of such a gradient can be observed in the mdr2 Ϫ/Ϫ null mutant where it is absent (23). The almost exclusive selection of PC as the major biliary phospholipid (95% of the total phospholipids in the bile (29)) may further optimize this biological system. Due to their physicochemical properties, PCs have a high tendency to interact with bile salts, causing a significant decrease of their CMC values. It is noteworthy that aminophospholipids (PE or PS) do not have this property. Clearly, this effect would make it possible for otherwise insoluble PC molecules concentrated at the site by the Mdr2 flippase to be efficiently extracted out of a membrane domain and thus, be provided in high enough amounts to the bile. Consistent with this model is our recent findings that the Mdr2-mediated lipid translocation is specific for PC, whereas PE and PS are not recognized as substrates.
Our data are consistent with a model in which high concentration of intravesicular TC may cause the formation of micelles containing NBD-PC and as a consequence of an apparent enhancement of Mdr2-mediated NBD-PC translocation. The formation of such aggregates did not affect the overall integrity of the membranes of SVs, which suggests that TC molecules possess the potent capacity to solely extract NBD-PC molecules  (47). These vesicles contain unusually high amounts of PC and thus, their biliary lipid composition differs considerably from that of the canalicular membrane (29). Since PC molecules in these vesicles are concentrated in the outer leaflet of the bilayer (48), they would be preferentially incorporated into the cytosolic leaflet of the canalicular membrane to form so-called "PC-rich microdomains." According to the model proposed by Coleman et al. (29) (model A), bile salts would then intercalate into these PC-rich microdomains initiating the budding-off of small vesicles containing whole membrane parts. Alternatively, Higgins (49) (model B) suggests that Mdr2 translocates PC molecules across the membrane, but without a compensating movement of lipids in the opposite direction. Such an event could cause small parts of the outer membrane leaflet to pinch off in micellular form, which subsequently would fuse to larger aggregates or vesicles. Our data support an alternative mechanism shown in model C. Bile salts are concentrated in the canalicular lumen where they form simple aggregates that possess an enormous potency to solubilize individual PC molecules out of the membrane bilayer. The Mdr2 flippase activity would increase the PC content in the outer leaflet toward the lumen of the canaliculus, where PC molecules would be available to form specifically mixed aggregates with bile salts. This very effective, and constant removal of PC by bile salts would cause an indirect positive regulation of the Mdr2 transport system to translocate more PC molecules through the bilayer. Consequently, as more bile salts are secreted, more PC molecules are translocated creating a enhanced lipid secretion, which is a fundamental prerequisite for bile formation and flow. out of the SV membranes. In addition, results from control experiments suggested that under normal conditions, most of the TC accumulated in SVs was located in the luminal space rather than found incorporated into SV membranes. (Indeed, results in Fig. 1 of the studies of St. Pierre et al. (37) indicated that under clamped conditions, only a small amount of TC associated nonspecifically with the membrane of SVs). The absence of TC-mediated enhancement of PC translocation by Mdr2 under clamped conditions suggested that this amount of TC associated with the bilayer was insufficient to solubilize NBD-PC, and that only free and/or aggregated TC in the luminal space could mediate the enhancing effect. In our experimental system, it is difficult to envision a scenario in which bile salt molecules would be intercalated in the membrane and subsequently would induce the formation of micelles (see Fig.  8, model A or B) without severely affecting the integrity of the SV membrane (distribution of the luminal marker enzyme invertase).
We would like to speculate on an additional possible mechanism for the initial step of hepatic bile formation (Fig. 8,  model C). Bile salts are concentrated in the canalicular lumen, where they can form simple aggregates, which would directly solubilize PC molecules out of the outer leaflet of the canalicular membrane of hepatocytes. PC molecules are derived from internal pools, most likely from the Golgi, trafficking in vesicular form to the apical domain, where they fuse into the canalicular domain (47). Following such a route, PC would be present in higher concentrations in the inner leaflet of the canalicular membrane facing the cytosol (48) (Fig. 8, left panel). The Mdr2 P-gp would act at that site to reverse this gradient and increase the PC content in the outer leaflet, lining the lumen of the canaliculus, where PC molecules would be available to form mixed aggregates with bile salts. We interpret our current data as suggesting that the TC/PC interaction does not occur within the membrane. We rather speculate that the high concentration and enormous solubilization capacity of bile salts for PC may directly force PC molecules to translocate out of the canalicular membrane into preformed bile salt aggregates. Since only PC molecules, but not aminophospholipids, are extracted by this process, the overall integrity of the canalicular membrane would not be affected. The constant removal of PC would then cause an indirect positive regulation of the Mdr2 transport system to translocate more PC molecules through the bilayer. As more bile salts are secreted, more PC molecules are translocated, creating an enhanced lipid secretion, which is a fundamental prerequisite for bile formation and flow.