Aminophospholipids Have No Access to the Luminal Side of the Biliary Canaliculus

About 95% of the bile phospholipids are phosphatidylcholine. Although the fractions of phosphatidylcholine and of both aminophospholipids phosphatidylserine and phosphatidylethanolamine in the canalicular membrane are in the same order of about 35% of total lipids, both aminophospholipids are almost absent from the bile. To rationalize this observation, we studied the intracellular uptake of various fluorescent phospholipid analogues and their subsequent enrichment in the bile canaliculus (BC) of HepG2 cells. Diacylaminophospholipid analogues but not phosphatidylcholine analogues became rapidly internalized by an aminophospholipid translocase (APLT) activity in the plasma membrane of HepG2 cells. We observed only low labeling of BC by diacylaminophospholipids but extensive staining by phosphatidylcholine analogues. In the presence of suramin, known to inhibit APLT, a strong labeling of BC by diacylaminophospholipid analogues was found that declined to a level observed for control cells after removal of suramin. Unlike diacylphosphatidylserine, diether phosphatidylserine analogue, which is not an appropriate substrate of APLT, accumulated in the BC. The correlation between low labeling of BC and an APLT-mediated transbilayer movement suggests the presence of an APLT activity in the canalicular membrane that prevents exposure of aminophospholipids to the bile.

A major component of bile fluid are phospholipids. Their uptake from the canalicular membrane into the BC requires the solubilizing activity of bile salts (1,2). Phosphatidylcholine (PC) accounts for about 95% of the bile phospholipids (3), whereas it constitutes only 35% of the canalicular membrane phospholipids (4). In agreement with the composition of mammalian plasma membranes, the aminophospholipids phosphatidylethanolamine (PE) and phosphatidylserine (PS) represent about 24 and 11%, respectively (4), of the canalicular membrane lipids. However, PS is virtually absent from the bile, and PE represents only 4.5% of biliary phospholipids. Because previous work has shown that the interaction of bile salts with phospholipids is independent of the phospholipid head group (5), other mechanisms that prevent specific phospholipids located in the apical membrane from being solubilized into BC must ensure the characteristic phospholipid composition of the bile fluid.
We hypothesized that the absence of the aminophospholipids PS and PE from bile is a consequence of their almost exclusive localization on the cytoplasmic leaflet of the apical membrane (6,7). This asymmetric localization could be the result of the activity of aminophospholipid translocase (APLT), a protein shown to be ubiquitous in the plasma membrane of mammalian cells. APLT rapidly transports the aminophospholipids PS and PE from the exoplasmic to the cytoplasmic leaflet at the expense of ATP (8 -10), a process first shown for the erythrocyte membrane (8,11,12). In contrast, PC and sphingomyelin are not recognized by APLT. Employing spin-labeled lipid analogues, the half-time of APLT-mediated inward transport in red blood cells was found to be in the order of 2-5 min and 30 min for PS and PE, respectively (13). Although the molecular identity of APLT is awaiting, it is very likely to be a member of the P-type ATPase family (14,15).
To address the presence of an APLT activity in the canalicular membrane we studied the enrichment of fluorescent phospholipid analogues in the BC of HepG2 cells. This human hepatoma cell line is able to polarize and form a biliary vacuole that resembles the tubular BC of the liver (16,17). Although we suggested recently the presence of an APLT activity in the plasma membrane of nonpolarized HepG2 cells by employing spin-labeled phospholipid analogues (18,19), this approach did not allow us to address whether the canalicular membrane itself harbors an APLT activity. Here, using a panel of fluorescent phospholipid analogues with differing affinities toward APLT, we found that analogues that are efficiently transported by APLT are essentially excluded from the BC, whereas those that are not transported by APLT were enriched in the BC. These observations provide strong evidence for an APLT activ-ity in canalicular membrane which prevents aminophospholipids from accumulating on the luminal side of BC.
Cell Culture-HepG2 cells were grown in Dulbecco's modified Eagle's medium containing 4.5 g/liter glucose supplemented with 10% heatinactivated fetal bovine serum and penicillin/streptomycin. Cells were routinely passaged in 25-cm 2 plastic culture flasks coated with collagen A; the medium was changed every 3-4 days. For preparation of cell suspensions, cells were cultured in 175-cm 2 plastic culture flasks coated with collagen A; the medium was changed on the day before the experiment. Cell suspensions were prepared in that cell monolayers were rinsed twice with HBSS and harvested by incubation with 0.05% trypsin and 0.02% EDTA in PBS for 5 min at 37°C. Subsequently, cells were resuspended in 50 ml of culture medium. After keeping the suspension for 30 min on ice the cells were washed twice with PBS and resuspended in PBS ϩ (PBS supplemented with 20 mM glucose, 1 mM sodium pyruvate, and 10 mM HEPES to prevent ATP depletion and pH shift) for labeling.
For microscopic observations, cells were grown on microcover glasses (18 ϫ 18 mm, BRAND, Germany) coated with poly-L-lysine (21). After 3 days, where the highest degree of polarization was obtained (16,22), cells were used for experiments.
Cell Labeling-Polarized cells on cover glasses were incubated with the label (final concentration 4 M in HBSS ϩ ) for 20 min on ice, washed with HBSS ϩ , and incubated for 30 min at 37 or 14°C (7). Taking into account the less efficient incorporation of diacyl-NBD-PE into membranes compared with the diacyl analogues of PS and PC (20,23), labeling with the analogues was performed by incubating the cells for 30 min with the analogue at 37°C followed by washing with HBSS ϩ and further incubation for 30 min at 37°C.
For measurements in suspensions, cells (4 ϫ 10 7 cells) were incubated with 600 l of PBS ϩ containing the respective lipid analogue. Labeling was performed for 5 min with 12 nmol of diacyl analogues or 15 min with 24 nmol of diether analogues on ice to achieve comparable extent of labeling. To prevent hydrolysis of the diacyl lipid analogues 5 mM DFP was added in parallel and was present in all following steps. Subsequently, cells were washed with PBS.
Double Labeling of Polarized Cells-HepG2 cells were double labeled with the green fluorescent NBD-PS analogue and the red fluorescent diacyl-␤-BODIPY-PC. For double labeling, polarized cells on cover glasses were incubated with 4 M diacyl-NBD-PS for 20 min on ice or 4 M diether NBD-PS for 30 min at 37°C. After removing the NBD-PS analogue the cells were labeled with 10 M ␤-BODIPY-PC bound to BSA for 10 min at 37°C, washed, and incubated further for 30 min at 37°C. Because of its higher hydrophobicity compared with NBD-labeled analogues, ␤-BODIPY-PC was bound to fatty acid-depleted BSA before labeling the cells (7).
Inhibition of Internalization by Suramin-Inhibition of APLT by suramin was performed as described earlier (15). Briefly, cells were incubated with 200 M suramin in respective incubation buffer for 30 min at 37°C prior to labeling. The inhibitor was present during all following steps. For adherent cells, in some cases, cells were labeled and incubated for 30 min at 37°C to allow internalization of analogues and subsequently incubated with suramin for 30 min at 37°C.

Measurements of Analogue Internalization in Cell
Suspensions-Labeled cells were incubated at 37 or 14°C. After various times aliquots were transferred into a fluorescence cuvette containing 2.4 ml of icecold PBS ϩ . NBD fluorescence was monitored at 540 nm ( ex ϭ 470 nm) (Aminco Bowman Series 2 spectrofluorometer) at 4°C while continuously stirring of the suspension. Dithionite was added from a freshly prepared 1 M stock solution in 100 mM Tris (pH 9.5) to give a final concentration of 50 mM as described earlier (24,25). Dithionite quenches the fluorescence by chemical reaction with the NBD group. Because dithionite permeates very slowly across membranes at low temperature (25; see "Results"), only the fluorescence of analogues on the exoplasmic leaflet is quenched. Indeed, upon addition of dithionite, we observed for labeled HepG2 cells an initial rapid decline of fluorescence intensity corresponding to reduction of the analogues on the exoplasmic leaflet (not shown). Subsequently, only a very slow fluorescence decline was observed which, very likely, is because of slow permeation of dithionite and reduction of analogues on the cytoplasmic side. Reduction of intracellular localized analogues did not exceed 2% during dithionite assay measurements, which regularly took about 5 min. After the fluorescence intensity was reduced by dithionite to a plateau value, Triton X-100 was added to a final concentration of 2%, making the NBD-phospholipids on the cytoplasmic side accessible to dithionite. The amount of phospholipids on the cytoplasmic side (PL i ) was determined according to: with F p being the fluorescence of the plateau after dithionite reduction, F e the background fluorescence after the addition of Triton X-100, and F s the fluorescence intensity before the addition of dithionite. Fluorescence Microscopy-Labeled cells grown on cover glasses were analyzed with an inverted Axiovert 100 standard epifluorescence microscope (Carl Zeiss, Inc. Oberkochen, Germany) equipped with a Pla-nApo 100ϫ/1.3 numerical aperture objective and a green fluorescence filter set (BP 450 -490 nm excitation filter, FT 510 nm dichroic mirror, and LP 515 nm emission filter) (Carl Zeiss) as described previously (7). Canalicular vacuoles (BC) were identified by phase-contrast microscopy. The percentage of BC containing the fluorescent lipid analogue was quantified by counting labeled and nonlabeled BC. BC having a fluorescence intensity in the BC as low as cellular autofluorescence levels were defined as nonlabeled BC.
Localization of the fluorescent lipid analogues in the BC was confirmed by the addition of dithionite, which can diffuse into BC and react with NBD-labeled analogues in the lumen and on the luminal membrane leaflet of BC (22). The remaining fluorescence of the canalicular membrane originates of phospholipid analogues located on the cytoplasmic side of this membrane. Cells were covered with HBSS supplemented with 20 mM HEPES (pH 7.4) to prevent any pH shift and were incubated with 30 mM sodium dithionite (added from a fresh 1 M stock solution in 100 mM Tris, pH 9.5) for 10 min on ice (7).
Photographs were taken using Kodak Ektachrome Panther P1600 films, which were push-processed to 3200 ASA and scanned using a CanoScan 2700F scanner (Canon, Tokyo, Japan) or by a cooled CCD camera (Coolsnap fx, Visitron Systems, Puchheim, Germany) using Metamorph software (Universal Imaging, Media, PA). Image analysis was carried out using Metamorph software (Universal Imaging).
Enrichment of Fluorescent Bile Salt Analogues in the BC-After washing HepG2 cell layers three times with HBSS ϩ , cells were labeled with 10 M fluorescent bile salt analogues CGamF or NBD-cholan by the addition of respective amounts of 10 mM stock solutions of analogues in ethanol (7). After incubation for 15 min at 4 or 37°C, cells were washed on ice with HBSS ϩ and observed by fluorescence microscopy.
Measurement of Metabolism of Diacyl-NBD Analogues-To measure hydrolysis of cell-associated lipid analogues, cells in suspensions or in monolayers were labeled as described above. Subsequent to labeling, suspensions were incubated for 60 min, monolayers for 30 min at 37°C (see above and "Results"). For lipid extraction, monolayers were scraped from the dish and resuspended in 1 ml of HBSS.
Lipids were extracted as described previously (26). Briefly, 3.2 ml of methanol/chloroform (2.2:1) was added to 1 ml of cells and incubated for 30 min at room temperature. After phase separation by adding 1 ml of chloroform and 1 ml of 40 mM acetic acid, lipids in the chloroform phase were collected. The aqueous phase was washed with 1 ml of chloroform, the two combined chloroform phases were dried under a nitrogen stream, and the lipids were resuspended in a small volume of chloroform/methanol (1:1) and applied to a TLC plate. Plates were developed in two dimensions using chloroform:methanol:ammonia (13:5:1) as basic solvent and aceton:chloroform:methanol:glacial acetic acid:water (8:6:2:2:1) as acidic solvents. Spots on the dried TLC plates were analyzed using a Fluorescence Image Analyzer FLA-3000 (Raytest Isotopenmessgerä te GmbH, Germany) and AIDA image analysis software.

RESULTS
To identify and characterize APLT activity in the canalicular membrane, we used diacyl and diether analogues of PS and PC and a diacyl analogue of PE, each bearing the fluorescent NBD moiety at the short sn2 fatty acid chain (C6). These analogues differ in their affinity to and transport by APLT which is primarily determined by the head group as well as the glycerol backbone of phospholipids (13). As shown for various mammalian cells, PC analogues are not transported by APLT (20,23,25,27). In contrast to PC, diacyl-NBD-PS and, to a lesser extent diacyl-NBD-PE, are transported by APLT. However, the transport of the diether analogue of PS by APLT is very slow compared with diacyl-NBD-PS (20).
First, we characterized the transbilayer redistribution of these analogues across the plasma membrane of suspended HepG2 cells to verify that the respective analogues behave in a manner similar to that found for other mammalian cells (see above). Subsequently, we studied the enrichment of aminophospholipid analogues in BC of polarized HepG2 cells in monolayer cultures.
Internalization of Diacyl-Lipid Analogues in Suspended HepG2 Cells-In Fig. 1 the kinetics of internalization of NBD-lipid analogues in suspended HepG2 cells is shown. To differentiate between uptake by endocytic activity and by transbilayer movement from the exo-to the cytoplasmic leaflet, internalization was measured not only at 37°C (Fig. 1, A and B), but also at 14°C (Fig. 1, C and D), where endocytosis is strongly reduced (28,29). During the labeling procedure a low amount of NBD analogues was lost from the exoplasmic leaflet, indicated by the offset at t ϭ 0 of the respective plots.
At 37°C the internalization of diacyl-NBD-PS was rapid, about 75% of the analogue was internalized with a half-time of about 5 min (Fig. 1A). We also observed a rather rapid uptake of the analogue with respect to the other analogues at 14°C (Fig. 1B). Although the internalization of diacyl-NBD-PE was slower compared with diacyl-NBD-PS, it was still faster with respect to the PC analogue (see below). The internalization of diacyl-NBD-PS and -PE across the plasma membrane was diminished by preincubation of the cells with suramin, an inhibitor of APLT (Fig. 1, A and C); the plateau of redistribution kinetics was about 3-4-fold lower compared with that in the absence of suramin. These results are consistent with an APLT activity in HepG2 cells.
The fraction of diacyl-NBD-PC localized in the exoplasmic leaflet of the plasma membrane was much higher compared with that of the diacylaminophospholipid analogues (Fig. 1A). At 37°C only about 20% were found nonaccessible to dithionite after 1 h of incubation. Again, the internalization was significantly reduced upon lowering the temperature (cf. Fig. 1C). At 14°C, about 10% of PC analogues redistributed to cytoplasmic side within 1 h. Preincubation of cells with suramin did not affect the transbilayer dynamics of diacyl-NBD-PC (data not shown).

Internalization of Diether Lipid Analogues in Suspended
HepG2 Cells-Internalization kinetics of diether NBD-PC was very similar to that of the diacyl analogue (Fig. 1). However, the internalization of diether NBD-PS was slower with respect to the diacyl analogue (Fig. 1, B and D), which is consistent with our previous observation on fibroblasts and red blood cells, indicating a very low affinity of APLT for diether NBD-PS. In agreement with this, suramin only had a marginal influence on the inward reorientation of this analogue (data not shown). However, the amount of internalized diether NBD-PS was still higher compared with both PC analogues (Fig. 1, B  and D).
Enrichment of Fluorescent NBD-Phospholipid Analogues to the BC of Polarized HepG2 Cells-We next compared the enrichment of various fluorescent lipid analogues in BC of polarized HepG2 cells by fluorescence microscopy. Subsequent to labeling of the basolateral membrane on ice, the accumulation of lipid analogues in the BC was monitored after incubation of cells for 30 min at 37°C (Fig. 2) or at 14°C (see Fig. 4). For diacyl-NBD-PC, an extensive labeling of BC was found at 37°C (Fig. 2, A and B). About 80% of the BC were labeled with the analogue (Fig. 3). The punctuate intracellular staining observed originates from endocytic uptake of the analogue (7,30). We found the same pattern of intracellular labeling for diether NBD-PC, in particular the enrichment of the analogue in the BC (Fig. 2, C and D). The amount of labeled BC was similar to that of diacyl-NBD-PC (Fig. 3).
The pattern of intracellular fluorescence was very different between the PC and diacylaminophospholipid analogues. For the latter, we observed a diffuse distribution in the cytoplasm (Fig. 2, E and F; only shown for diacyl-NBD-PS), rather than the punctuate staining of endocytic vesicles as detected for the PC analogues. The amount of BC labeled by diacylaminophospholipid analogues was low (Fig. 2, E and F). Only about 20% of the BC were labeled with diacyl-NBD-PS after 30 min at 37°C (Fig. 3). In these BC the fluorescence intensity was lower than that of those labeled with PC analogues. This indicates a reduced enrichment of the PS analogue compared with PC. Similarly, diacyl-NBD-PE was excluded from the BC but less rigorously as diacyl-NBD-PS. Enrichment of the PE analogue was found in about 30% of the BC (see Fig. 3).
Unlike diacyl-NBD-PS, diether NBD-PS became enriched in the BC (Fig. 2, G and H). The percentage of labeled BC was almost identical to that of PC analogues (Fig. 3). However, compared with PC analogues, the punctuate staining within the cytoplasm was less pronounced, and we noted a somewhat diffusive fluorescence in the cytoplasm.
Endocytosis and transcytosis of vesicles from the basolateral to the apical membrane in polarized cells are temperature-dependent and are strongly reduced at 14°C (31). However, also under this condition bright labeling exclusively of the BC was found for both PC analogues and diether NBD-PS (Fig. 4). In contrast, BC were not labeled by diacyl-NBD-PS. Similar to our observation at 37°C, we found a bright and diffuse cytoplasmic staining for the PS analogue.
To confirm the difference in localization between diacyl-NBD-PS and diether NBD-PS as well PC analogues, cells were double labeled with NBD-PS analogues and a diacyl-␤-BODIPY-PC analogue. We have shown previously that ␤-BODIPY-PC is specifically enriched in BC (7). In agreement with that, ␤-BODIPY-PC redistributed rapidly to the lumen of the BC (Fig. 5), which was not observed for diacyl-NBD-PS. In Confocal laser scanning microscopy of cells chased at 14°C with diacyl-NBD-PS provided additional evidence for the absence of this analogue from the BC lumen (Fig. 6), while the canalicular membrane of BC was labeled. Because dithionite did not affect the fluorescence of diacyl-NBD-PS-labeled cells (see "Experimental Procedures"), we conclude that fluorescence of the canalicular membrane is caused by PS analogues localized in the cytoplasmic leaflet of the membrane. This is in contrast to the labeling pattern of BC by diacyl-NBD-PC. As we have reported previously and confirmed here (not shown), CLSM revealed that the PC analogue enriches in the lumen of BC (7).
In the presence of suramin, we found a significant enrichment of diacyl-NBD-PS in BC. Suramin treatment increased the amount of BC labeled by diacyl-NBD-PS to about 50% (Fig.  3). Interestingly, punctuate intracellular staining was observed, indicating that endocytic pathways contribute significantly to the uptake of the PS analogue under those conditions. When suramin was washed out, the amount of labeled BC decreased to a value found for cells that had not been treated with the inhibitor (Fig. 7). The half-time of the process was in the order of about 15 min. A comparable effect of suramin was observed for diacyl-NBD-PE (data not shown). Suramin affected neither the number of labeled BC with diacyl-NBD-PC and with diether NBD analogues nor the punctuate staining seen for PC analogues.
Diacyl-NBD-PS was also enriched in BC when cells were treated with suramin after internalization of the PS analogue (see "Experimental Procedures"). To this end, cells were labeled with diacyl-NBD-PS and incubated for 30 min at 37°C to allow internalization of the analogue. At this point there was only low labeling of BC (see above). Subsequent addition of suramin led to a bright labeling of BC (data not shown).
As shown by CLSM, diacyl-NBD-PS became enriched in the lumen of BC after suramin treatment (Fig. 8, A-E). The labeling of BC was very similar to that observed for diacyl-NBD-PC in (nontreated) HepG2 cells (see above). When the inhibitor was washed out and cells were incubated for 30 min at 37°C, a labeling pattern of BC similar to the one seen in Fig. 6 was observed (not shown).
Metabolism of the Lipid Analogues-The intracellular hydrolysis of the diacyl-NBD analogues and their metabolic conversion to other fluorescent lipids are summarized in Table I. For cells in suspension, between 10 and 20% of diacyl analogues were hydrolyzed after incubation at 37°C for 60 min. In monolayers of polarized cells, 6 -8% of diacyl-NBD-PE and -PC and about 17% of diacyl-NBD-PS were hydrolyzed after incubation for 30 min at 37°C. In the presence of DFP, hydrolysis was reduced to a level of 2-4% of total analogues. We were not able to treat polarized cells with DFP to study intracellular localization of analogues because of impaired labeling of BC in the presence of DFP.
For both cells in suspension and polarized cells a metabolic conversion of diacyl-NBD-PS to PE and PA was observed which was not affected by DFP (Table I).
As expected, no breakdown of diether analogues was observed.
Transport of Fluorescent Bile Salts to the BC of Monolayer Cultures of Polarized HepG2 Cells-To confirm that HepG2 cells were functionally polarized we examined whether cells are able to transport bile salts into the BC. To this end, we used the fluorescent bile salt analogues CGamF and NBD-cholan. Although HepG2 cells do not express all bile salt transporter proteins identified in the basolateral membrane of hepatocytes (32), we observed bright staining of BC after labeling of the cells for 15 min at 37°C (Fig. 9, A and B; only shown for NBD-cholan). About 80% of BC were labeled by the bile salt analogues (Fig. 9C). Incubation of the cells at 4°C significantly reduced the number of labeled BC (Fig. 9). Preincubation of cells with the inhibitor PSC 833 (5 M) according to Cantz et al. (33) reduced the percentage of labeled BC in either case to less than 30% (Fig. 9). These results indicate a vectorial transport of fluorescent bile acid analogues from the basolateral membrane toward the BC in HepG2 cells. The inhibition by PSC 833 suggests that ABC proteins are involved in the enrichment of the analogues in the BC. It has been shown that PSC 833 inhibits the ABC transporter MDR1 Pgp, MRP2, and BSEP (bile salt export pump) (34). Taken together, these data prove the functional polarity of HepG2 cells used in our experiments. The identification of the transporters involved in bile salt analogue enrichment in HepG2 cells is beyond the aim of this study and warrants further investigations. DISCUSSION The present study was aimed at providing experimental evidence for a mechanism that explains the absence of aminophospholipids from the bile. In particular, we investigated whether the access of the aminophospholipids PS and PE to the luminal side of the BC is prevented by an APLT activity in the canalicular membrane by analyzing the redistribution of various fluorescent analogues to the BC lumen. The main result of our study is that aminophospholipid analogues that are not (efficiently) transported by APLT are enriched in BC, whereas analogues that represent suitable substrates for APLT are exported from the lumen of the canalicular membrane, where they accumulated only when APLT activity is inhibited. We conclude that the canalicular membrane harbors an APLT activity essential for preserving the specific phospholipid composition of the bile. This APLT activity is sufficient to explain the virtual absence of aminophospholipids from the bile.
It may be asked whether the fluorescent analogues used are faithful reporters of natural phospholipid movements because they may locally perturb the lipid bilayer. The great differences in translocation for various lipid analogues demonstrate selectivity to the head group/glycerol backbone and not the fluorescent NBD moiety. Indeed, the active transport of aminophospholipids in red cells first discovered using spin-labeled lipids (8) was confirmed with fluorescent probes (35) as used here as well as with radiolabeled long chain lipids (12). Thus, the same conclusions could be drawn employing different families of lipid probes. Therefore, we are confident that in HepG2 cells study- ing phospholipid translocation based on trafficking of fluorescent probes is a valid approach. This is also strongly supported by the fact that the labeling of BC by fluorescent diacylphospholipid analogues reflects the specific phospholipid composition of the bile; although aminophospholipids are barely found, PC is enriched in the bile.
Mechanisms of Phospholipid Analogue Internalization in HepG2 Cells-Because most of the fluorescent analogues used in this study have not been tested in HepG2 cells previously, we characterized the internalization of the analogues from the exoplasmic leaflet to the cytoplasmic side on suspended cells. We found a much higher accumulation of diacyl-NBD-PS and diacyl-NBD-PE on the cytoplasmic side with respect to diacyl-NBD-PC, which is consistent with observations on other mammalian cells (for review, see Ref. 10) and points to the presence of an APLT activity in HepG2 cells. The higher rate of diacyl-PS internalization with respect to PE is in agreement with a lower transport affinity of APLT for PE, in particular for diacyl-NBD-PE (9,23). The rapid disappearance of the aminophospholipid analogues does not originate from an endocytic uptake because we found a rapid inward redistribution of the PS analogue even at 14°C where endocytosis is reduced and transcytosis blocked (28,29). Furthermore, if endocytic activity was the main component of analogue uptake, we would not expect such great differences among the various analogues. Only after inhibition of APLT activity by suramin, endocytosis became relevant in the case of diacylaminophospholipids at 37°C as indicated by punctuate intracellular staining.
Compared with the amount of internalized diether NBS-PS the enrichment of the respective diacyl analogue was significantly lower. This confirms previous studies on the plasma membrane of human fibroblasts and red blood cells showing a low transport activity of APLT for diether PS (20). It is also consistent with the fact that the glycerol backbone of phospholipids affects transport of lipids by APLT (13). Compared with diacyl-NBD-PS and -PE, internalization of diacyl-NBD-PC and diether NBD-PC from the exoplasmic leaflet was much lower. Nevertheless, we have shown recently that diacyl-NBD-PC is internalized in HepG2 cells at 37°C by two routes, by endocytic uptake and by transbilayer movement facilitated by a yet unknown transporter (7). Upon endocytosis diacyl-NBD-PC was also transported into a recycling compartment containing transferrin (7). Because of subsequent exposure of PC analogues to the plasma membrane, the fraction of analogues internalized at 37°C is underestimated.
Taken together, we found that the inward redistribution of various NBD-lipid analogues for the plasma membrane of HepG2 cells is very similar to that of other mammalian cells. In particular, a high affinity of APLT to diacyl-PS, a slightly lower affinity to diacyl-PE, and a very low affinity of the transporter to diether NBD-PS were observed.
Presence of an APLT Activity in the Canalicular Membrane-Having established this pattern of inward redistribution, we studied the labeling of BC by fluorescence microscopy after incorporation of fluorescent analogues into the basolateral membranes of polarized HepG2 cells. We found a strong correlation between a low degree of BC labeling and APLT-mediated internalization of the analogue (Table II). Analogues that were not or only inefficiently transported by APLT such as PC analogues and diether NBD-PS were rapidly enriched in the BC. For the aminophospholipid analogues diacyl-NBD-PS and -PE that were shown to be transported efficiently by an APLT activity in HepG2 cells (see above), only a low percentage of labeled BC was found (see bold entries in Table II). Importantly, the PE analogue was less rigorously excluded from the BC. This can be explained by the lower affinity of APLT to PE (see above), which can also rationalize the small amount of PE found in the bile fluid (4).
Labeling of BC by diacyl-NBD-PS increased upon treatment of cells with suramin an inhibitor of APLT. Suramin did not influence the labeling pattern of PC analogues and their enrichment in BC, indicating that the transport pathways of PC (7) are not affected by the inhibitor. These results strongly suggest that low labeling of BC by diacyl-NBD-aminophospholipids is related to an APLT activity in the canalicular membrane pumping aminophospholipids from the luminal to the cytoplasmic leaflet and that inhibition of this APLT activity, e.g. by suramin, leads to labeling of BC by the analogues.
One may wonder whether the rapid inward movement of diacyl-NBD-PS and -PE by an APLT activity on the basolateral  membrane may specifically prevent the access of the analogues to the BC. First of all, we would like to emphasize that tight junctions prevent lateral diffusion of analogues from the basolateral to the apical membrane on the exoplasmic leaflet and thereby access of the analogues to the BC by extra-cellular aqueous space (7,36,37). Therefore, analogues have to be delivered to the BC via intracellular pathways or at least by accessing the cytoplasmic leaflet of the plasma membrane. We found a bright diffuse intracellular staining for both diacyl-NBD-PS and -PE but no labeled endocytic vesicles. This shows that the major route of intracellular uptake of these analogues is the rapid transport from the exo-to the cytoplasmic leaflet by APLT, whereas the endocytic pathway plays only a minor role. Once on the inner leaflet of the plasma membrane, both analogues equilibrate rapidly with the cytoplasmic leaflet of subcellular membranes presumably by monomer diffusion (9). Thus, these analogues have access to the canalicular/apical membrane by lateral diffusion in the cytoplasmic leaflet of the plasma membrane and/or by diffusion through the cytoplasm. Indeed, CLSM confirmed the localization of aminophospholipid analogues to the canalicular membrane as shown for diacyl-NBD-PS. The diffuse intracellular labeling observed in the case of diether NBD-PS suggests that this analogue may have access to the canalicular membrane in a similar manner.
It might be argued, that enrichment of diacylaminophospholipid analogues in BC upon treatment with suramin is related solely to an inhibition of APLT activity in the basolateral membrane but does not argue for an APLT activity in the canalicular membrane. This concern is ruled out by the following observations. First, when cells were treated with suramin after internalization of diacyl-NBD-PS or -PE, the original low labeling of BC (see above) changed to a bright labeling of BC, indicating strongly that APLT activity in the canalicular membrane was inhibited. Second, bright labeling of BC by diacylaminophospholipid analogues of suramin-treated HepG2 cells was reversed upon removal of suramin, suggesting that the restored APLT activity in the canalicular membrane caused a redistribution of the analogues from the luminal to the cytoplasmic leaflet of the BC.
Notably, the repartition of PC analogues, of diether NBD-PS and, in the presence of suramin, of diacyl-NBD-aminophospholipids to the BC lumen is in agreement with our previous observation that solubilization of analogues by bile salts is independent of their head group (5,6). Thus, low labeling of BC by diacylaminophospholipid analogues in the absence of suramin (control) cannot be explained by a failure of bile components to solubilize these analogues from the luminal leaflet.
Does Labeling of BC Correlate with Metabolic Conversion of Analogues?-Our data preclude that labeling of BC is caused by products of metabolic conversion of analogues, but not by the analogues themselves. In particular, we can rule out that labeling of BC corresponds to the enrichment of the hydrolyzed  sn2 fatty acid residue of the analogues carrying the NBD moiety. First, if the latter was the case, the highest degree of BC labeling would be expected when using diacyl-NBD-PS (see Table I). However, the opposite was found. We can also discard that fluorescence in BC arises from NBD-labeled lysolipids. We never observed a metabolic conversion of analogues to fluorescent lysolipids. Second, when incubating polarized cells with the free NBD-labeled fatty acid according to the protocol used for lipid analogues we did not find a stable enrichment of the fatty acid in the BC. Although BC-associated fluorescence in the case of PC and diether analogues or of diacyl-NBD-PS upon treatment with suramin was not affected by washing cell monolayers, for free NBD-labeled fatty acid, BC-associated fluorescence was removed after washing (not shown). This is consistent with a rapid transbilayer movement of this short NBDlabeled fatty acid chain as probed by the dithionite assay in liposomes (not shown). Third, the strong labeling of BC by diether analogues, which are resistant to hydrolytic cleavage of the sn2 chain (e.g. by endogenous phospholipases; see "Results"), argues also against such an explanation. We can eliminate the possibility that the metabolic conversion of diacyl-NBD-PS (Table I) accounts for the low degree of BC labeling. Even after incubation for 30 min at 37°C, about 70% of internalized diacyl-NBD-PS were not metabolized. Furthermore, addition of suramin at this point of incubation lead to a strong labeling of BC. CONCLUSIONS We have provided evidence for an APLT activity in the canalicular/apical membrane of polarized HepG2 cells. In the light of the finding, that solubilization of phospholipids from the membrane by bile salts is independent of the phospholipid head group, and, thus, does not offer a mechanism to ensure the specific phospholipid composition of the bile fluid (see Introduction), the presence of an APLT activity in the canalicular membrane may provide an alternative explanation for the absence of aminophospholipids in bile fluid. This activity excludes aminophospholipids from the luminal leaflet of the BC in a very efficient manner. Remarkably, this activity is even sufficient to prevent BC enrichment of diacyl-NBD-PS, which can be solubilized by bile salts much more efficiently than endogenous PS having two long fatty acid chains (6).
Further studies are warranted to clarify which proteins of the canalicular membrane are involved in the secretion of diether NBD-PC and -PS, and, visible upon suramin treatment, of diacyl-NBD-PS and -PE. The Fic1 gene, which is mutated in patients with progressive familial intrahepatic cholestasis 1, was shown to code for a P-type ATPase (38), and hence it was supposed to encode for APLT in the canalicular membrane. Recently, Ujhazy et al. (39) have demonstrated a Fic1-mediated PS translocation in transfected cells. However, because the authors did not investigate the translocation of non-aminophospholipids, it remains to be established whether Fic1 is a transporter specific for aminophospholipids. Also, several ABC transporters are localized in the canalicular membranes (40). ABC transporters, e.g. MDR1 Pgp and MDR3 (see above), have been shown to transport phospholipids including PC and PE as well as ether lipids (31, 41,42). Recently, we found that MDR1 Pgp is also able to transport diacyl-NBD-PS and endogenous PS (43). However, the transport activity of MDR1 Pgp was much lower compared with that of APLT.