Trans-Golgi network and subapical compartment of HepG2 cells display different properties in sorting and exiting of sphingolipids.

In HepG2 cells, the subapical compartment (SAC) is involved in the biogenesis of membrane polarity. By contrast, direct apical transport originating from the trans-Golgi network (TGN), which may contribute to polarity establishment, has been poorly defined in these cells. Thus, although newly synthesized sphingolipids can be directly transported from the TGN to the apical membrane, numerous apical resident proteins are traveling via the transcytotic route. Here, we developed an in vitro transport assay and compared the molecular sorting of 6-[N-(7-nitrobenz-2-oxa-1,3 diazol-4-yl)amino] hexanoyl-sphingomyelin (C(6)NBD-SM) and C(6)NBD-glucosylceramide (C(6)NBD-GlcCer) in TGN and SAC. SM is released from both TGN and SAC in the lumenal leaflet of transport vesicles. This holds also for GlcCer released from the SAC but not for a substantial fraction that departed from the Golgi. Distinct transport vesicles, enriched in either SM or GlcCer are released from SAC, consistent with their rigid sorting in this compartment. Different vesicle populations could not be recovered from TGN, although in situ experiments reveal that GlcCer is preferentially transported to the apical membrane, reflecting different transport mechanisms. The results indicate that in HepG2 cells sphingolipids are mainly sorted in the SAC membrane and that the release of SM from SAC and TGN is differentially regulated.

Polarized cells have distinct plasma membrane domains, i.e. the apical and basolateral membranes each displaying a specific protein and lipid composition. These differences are generated and maintained, despite a continuous membrane flux between these domains and intracellular organelles, sorting and vesicular transport being instrumental in these events (1).
Two major pathways have been implicated in polarized sorting. De novo synthesized proteins and lipids are sorted to either plasma membrane in the trans-Golgi network (TGN) 1 (2,3). In addition, molecular redistribution can also be mediated by the endosomal sorting machinery (4). Thus, proteins and lipids internalized from the basolateral and apical membrane meet in an apically localized common endosome (5), defined as the subapical compartment (SAC) in HepG2 cells (6). Here, molecules are sorted and either recycled to the original membrane or transcytosed to the opposite membrane domain (7), thereby securing and maintaining a polarized distribution. The existence of two such pathways raise intriguing questions concerning the relative contributions of SAC and TGN to polarity generation and maintenance and the molecular nature of the sorting machinery involved in either pathway.
Detergent-insoluble membrane domains, called rafts (8), have been implicated in sorting (9), although raft formation per se is probably not sufficient for apical targeting (10). (Glyco)sphingolipids, together with cholesterol, are the main lipid constituents of rafts and the finding that inhibition of sphingolipid synthesis, but not cholesterol depletion, interferes with the apical sorting of detergent-insoluble GPI-anchored proteins in Fischer rat thyroid cells (11) indicates that sphingolipids are especially important for the sorting of apical cargo. In many epithelial cells newly synthesized proteins are sorted in the TGN and directly transported to either the basolateral or apical membrane (3). In contrast, hepatocytes transport most (12,13), but not all (14,15) newly synthesized apical proteins first to the basolateral membrane, prior to reaching the apical (bile canalicular (BC)) membrane via transcytosis. Yet, in these cells, newly synthesized analogs of SM and GlcCer can be transported directly from the TGN to the BC membrane (16), although it has not been determined whether, similar to MDCK and Caco-2 cells (17,18), sphingolipids are actually sorted in the TGN. In the latter studies, it was shown that de novo synthesized C 6 NBD-labeled sphingolipid analogs can be sorted in the TGN, directing C 6 NBD-GlcCer preferentially to the apical membrane, whereas SM distributes about equally over apical and basolateral domains.
In MDCK cells, it has been demonstrated that the common recycling endosome is enriched in raft lipids (19). This could imply that by analogy the formation of sphingolipid-enriched membrane domains may be important for polarized sorting in the SAC as well. In this context, we have recently demonstrated that the SAC is intimately involved in regulating the distribution of both of these lipids during transcytotic transport (20,21). At steady state, SM is transported predominantly from the SAC to the basolateral membrane, whereas GlcCer is recycling between the SAC and the BC membrane (20). Circumstantial evidence suggests that the exiting of both lipid analogs from SAC is vesicle-mediated (20). However, de novo synthesized GlcCer analogs (22,23) as well as the natural counterpart (24) may at least in part be transported as monomers, likely from an early Golgi compartment, where its synthesis occurs at the cytosolic surface (25,26). Following arrival at the inner leaflet of the plasma membrane, transmembrane translocation of the GlcCer analogs has been shown to occur via ABC transporter proteins (23,27).
To understand the mechanisms that govern these various sorting processes, the regulation of sphingolipid transport in the biosynthetic and endocytotic pathways, and how vesicular transport derived from SAC and TGN contribute to the biogenesis and maintenance of polarity, the isolation of the putative transport vesicles and the reconstitution of vesicular transport of both proteins and lipids in permeabilized cells are desirable. Thus far, most studies focusing on these issues have been restricted to the analysis of vesicular protein transport (28 -30), whereas only a few in vitro studies have been developed to characterize the vesicular transport of lipids (31,32). Here, we developed an in vitro assay using permeabilized HepG2 cells to study the exit of fluorescently tagged SM and GlcCer from the SAC and TGN. The purpose was to reveal the occurrence and nature of sorting in either compartment, thereby improving our knowledge of where and how membrane domains enriched in sphingolipids may contribute to the sorting of apical and basolateral proteins. We demonstrate that both sphingolipids can exit SAC and TGN in transport vesicles. However, whereas SAC-derived trafficking is exclusively vesicle-mediated, a fraction of the newly synthesized GlcCer exits from Golgi by a mechanism that does not involve its packaging into the lumenal side of transport vesicles. Moreover, evidence is presented that SM and GlcCer are released in distinct vesicles from the SAC, highlighting the prominent role of this compartment as a sorting station in the transcytotic pathway.
Cell Culture-HepG2 cells were grown in 25-cm 2 cell culture flasks (Costar) in Dulbecco's modified Eagle's medium containing 4500 mg of glucose/liter, supplemented with 10% fetal calf serum and 2 mM Lglutamine. For the experiments, the cells were used 3 days after plating when they were maximally polarized (33).
Labeling of HepG2 Cells with C 6 NBD-Sphingolipids-To analyze sphingolipid transport of intact HepG2 cells, they were labeled with 4 M of C 6 NBD-sphingolipids. For in vitro studies 8 M of sphingolipids were used. To label the TGN, the cells were incubated with C 6 NBDceramide in Hanks' balanced salt solution for 90 min at 18°C. Noninternalized ceramide was removed from the plasma membrane by backexchange with 5% BSA at 4°C in Hanks' balanced salt solution (2 ϫ 30 min). The cells were incubated for an additional 90 min at 18°C in presence of 5% BSA. To label the SAC, the cells were incubated with C 6 NBD-SM and -GlcCer in Hanks' balanced salt solution for 30 min at 37°C. Noninternalized sphingolipids were removed by back-exchange, and the sphingolipids were chased into the SAC for 2 h at 18°C in presence of BSA. Sphingolipids that had been transported to the BC membrane were quenched for 10 min with 30 mM sodium dithionite at 4°C. At these conditions, the agent does not permeate the membrane and exclusively abolishes NBD fluorescence that is localized in the external leaflet of the BC (20).
Transport of Sphingolipids in Intact Cells-To study basolateral transport after labeling of the TGN, HepG2 cells were incubated for various time intervals at 37°C in presence of 5% BSA, acting as a scavenger for basolaterally arrived lipid. The incubation medium and the cells were collected separately, and the lipids of both fractions were extracted and analyzed by TLC as described previously (34). To study apical transport the cells were grown on coverslips, and the TGN was labeled as described above. The cells were incubated for various time intervals at 37°C in back-exchange medium. Transport to the BC membrane was analyzed by fluorescence microscopy. BC structures were identified in phase contrast and characterized as C 6 NBD-positive or -negative after switching to epifluorescence, as has been described in detail elsewhere (33). Images were scanned and processed with Paint Shop Pro.
In Vitro Budding Assay-After loading either TGN or SAC of HepG2 cells with C 6 NBD-sphingolipids, the cells were scraped from the culture dish into 1 ml of transport buffer (140 mM KCl, 0.5 mM KH 2 PO 4 , 20 mM Hepes-KOH, pH 7.3). The cells were permeabilized by pressing them four times through a 25-gauge syringe, and cell debris was removed by a brief centrifugation step (10 min, 2000 ϫ g). The permeabilized cells were treated for 30 min at 4°C with 5% BSA, washed once in transport buffer, and incubated at 37°C in the presence of cytosol (2 mg/ml), ATP, an ATP-regenerating system, and 1% BSA. After the incubation, the cells were removed by centrifugation at 2000 ϫ g, and the supernatant was centrifuged for 2 h at 100,000 ϫ g to recover membrane-associated sphingolipids, which were protected from BSA. The lipids of all fractions were extracted and analyzed by TLC. To calculate the release of C 6 NBD-sphingolipids in the lumenal leaflet of vesicles, the amount recovered in the 100,000 ϫ g pellet was correlated to the total amount of this sphingolipid prior to the incubation (2000 ϫ g pellet ϩ 100,000 ϫ g supernatant ϩ 100,000 ϫ g pellet). The amount of sphingolipids that became accessible to BSA (100,000 ϫ g supernatant) was determined similarly.
Fractionation of Vesicles by Gradient Centrifugation-The vesicular supernatant obtained from the in vitro budding assay was immediately loaded onto a continuous gradient of 20 -50% sucrose (w/w) in transport buffer. The gradient was centrifuged for 16 h at 100,000 ϫ g in an SW50 rotor (Beckman) and fractionated. Immediately before lipid extraction the obtained gradient fractions were treated for 5 min with 30 mM sodium dithionite at 4°C to quench any NBD fluorescence associated with sphingolipids that were extracted by BSA and/or exposed in the outer leaflet of the membranes.
Preparation of Cytosol-HepG2 cells were grown for 3 days on 165cm 2 dishes, washed and scraped into 0.5 ml of transport buffer. The cells of 5-10 dishes were collected and homogenized with a tight Dounce homogenizer. The homogenate was centrifuged for 10 min at 1000 ϫ g to remove cell debris and nuclei, and the post nuclear supernatant was centrifuged two times at 100,000 ϫ g. The final supernatant was used in the in vitro experiments.
Transmission Electron Microscopy-Gradient fractions containing vesicles enriched in C 6 NBD-sphingolipids were collected and adjusted with transport buffer to a sucrose concentration of 10% (w/w). The vesicles were pelleted for 3 h at 100,000 ϫ g. After fixation for 15 min with 0.5% glutaraldehyde in transport buffer containing 10% sucrose, the pellets were preincubated for 15 min in the dark with 0.15% (w/w) diaminobenzidine (DAB). For photoconversion of DAB by the NBDlabeled sphingolipids, the pellets were illuminated for 30 min at a wavelength of 460 nm (for experimental details see Ref. 36). Thereafter the pellets were incubated for 60 min at room temperature with 1% OsO 4 in cacodylate buffer (pH 7.4) and embedded in Epon. After sectioning, the structures containing osmiophilic precipitates were analyzed using a Philips CM100 electron microscope at 80 kV. The micrographs were scanned and processed with Paint Shop Pro.
Determination of Galactosyltransferase Activity-To assay for galactosyltransferase (GTase) activity, the samples were incubated for 1 h at 37°C in 50 l of a mixture containing (final concentrations) 100 mM cacodylic acid (pH 7.3), 20 mM N-acetylglucosamine, 20 mM MnCl 2 , 4 mM ATP, 0.5% Triton X-100, and 0.7 mM UDP-galactose (including 100,000 dpm [ 14 C]UDP-galactose). The incubation was stopped by the addition of 450 l of cold H 2 O, and [ 14 C]N-acetyllactosamine was separated from [ 14 C]UDP-galactose by a Dowex 1X8 -200 column. The values obtained in the presence of N-acetylglucosamine were corrected for control values (containing H 2 O), and the specific activity (dpm/g protein) was determined. Finally the amount of GTase in each sample was determined by multiplying the specific activity with the protein content.

RESULTS
Both Golgi-derived transport of de novo synthesized sphingolipids and SAC-derived glycosphingolipid transport contribute to the biogenesis and maintenance of polarized membrane domains in HepG2 cells. Although direct sphingolipid transport from Golgi to the BC membrane can occur in liver cells (16), the mechanism of transport has not been determined, and it is also unknown whether sorting of the lipid species occurs as observed in other epithelial cells (17,18). Accordingly, we first determined the mechanism by which newly synthesized GlcCer and SM reach the basolateral and apical membrane in optimally polarized HepG2 cells. This flux was then compared with the transport of both these lipids from the SAC, involved in membrane dynamics related to endocytotic events at either membrane surface, to appreciate the relative contribution of either pathway to membrane polarity development in liver cells.
Direct Transport of Newly Synthesized C 6 NBD-SM and -GlcCer from TGN to BC Membrane-To analyze the exit of newly synthesized sphingolipids from the TGN in HepG2 cells, the cells were labeled for 90 min with C 6 NBD-ceramide at 18°C. After a wash, the sphingolipids were chased for an additional 90 min at 18°C in the presence of 5% BSA. The latter will preclude the re-entry of basolaterally arrived lipids into the cell. After the chase, the perinuclear region of the cells was brightly labeled (Fig. 1, A and B), which is typical for a Golgi localization in these cells (16). 50% of the ceramide was metabolized to C 6 NBD-SM and -GlcCer (in a molar ratio of 3:1, Fig. 1C), the only products that could be detected. The accumulation of the newly synthesized lipids in the Golgi is reflected by the observation that only approximately 10% of both sphingolipids were transported from the Golgi, presumably exiting from the TGN, to the basolateral plasma membrane during the incubation at 18°C (Fig. 1D).
During the incubation at 18°C, no significant labeling of the BC could be detected (Fig. 1B), implying that sphingolipid transport from Golgi to BC was impeded, whereas evidently C 6 NBD-Cer did not acquire access to the BC. However, when the temperature was subsequently raised to 37°C in the presence of 5% BSA, 50 -60% of the BCs were C 6 NBD-positive after a 30-min incubation period (Fig. 2, A-C). Treatment with sodium dithionite after the incubation reduced the amount of labeled BCs to control levels ( Fig. 2C), providing evidence that the C 6 NBD-labeled sphingolipids were transported directly to the exoplasmic leaflet of the BC membrane. After longer incubation times (30 -60 min) the amount of positive BCs decreased again, indicating that both sphingolipids are transcytosed from the apical to the basolateral membrane (33), where they are captured by BSA (Fig. 2C).
Because the BC membrane is not accessible to BSA, the direct quantification of the apically transported sphingolipid analogs is not possible. To discriminate between the fate of either newly synthesized C 6 NBD-GlcCer or -SM, the cells were therefore incubated with 10 M PDMP, which completely inhibited the synthesis of C 6 NBD-GlcCer but did not affect synthesis of SM (data not shown). At these conditions, the fraction of the total pool of SM that was transported over a 30-min time interval to the basolateral membrane was not affected (Fig.  2D), but the number of labeled BCs after 30 min of transport was reduced by approximately 40% (Fig. 2D). Because only 25% of the total newly synthesized sphingolipid fraction (C 6 NBD-GlcCer plus C 6 NBD-SM) at control conditions consists of GlcCer, this result indicates that, in relative amounts, C 6 NBD-GlcCer is the major sphingolipid that is transported to the BC membrane. Yet, evidently, (a fraction of) C 6 NBD-SM is also transported from the Golgi to the BC membrane.
Mechanism of Sphingolipid Transport from TGN to Basolateral and Apical Membranes-Interestingly, C 6 NBD-GlcCer transport from the Golgi to the basolateral membrane was much faster than that of C 6 NBD-SM. After 15 min at 37°C, 32% of the pool of C 6 NBD-GlcCer and 18% of that of C 6 NBD-SM had reached the basolateral membrane (Fig. 3, A  and B). These different transport kinetics to the basolateral membrane of HepG2 cells are reminiscent of the fast transport of natural GlcCer compared with other sphingolipids in Chinese hamster ovary cells (24). With increasing incubation times the difference between SM and GlcCer with regard to the fraction of the total lipid pools that reached the basolateral membrane diminished, amounting to 75-80% of the pools of both sphingolipids after 60 min. As shown above (Fig. 2C) the percentage of labeled BCs decreased concomitantly, emphasizing the integration of the newly synthesized sphingolipids in a dynamic transport process between apical and basolateral membrane. This is not the case after short term incubations (30 min; Fig. 2C). Accordingly, this would therefore suggest that the rapid transfer of GlcCer from the Golgi to basolateral membrane after 15 min involves a direct transport step and, given its pool size compared with that of SM, may be related to a mechanism that differs from that of SM.
GlcCer is synthesized at the cytoplasmic leaflet of the cis-Golgi (25) and its transport from the Golgi surface to the cytoplasmic leaflet of the plasma membrane, either by monomeric flow or in the cytoplasmic leaflet of transport vesicles, FIG. 1. C 6 NBD-ceramide is metabolized at 18°C, but only a minor part of the newly synthesized sphingolipids is transported to the cell surface. The cells were labeled with C 6 NBD-ceramide as described under "Experimental Procedures." A and B, the perinuclear area is brightly labeled in the fluorescence microscope (B), but BC structures (arrows) that were identified in phase contrast microscope (A) appear unlabeled. C, analysis of C 6 NBD-ceramide metabolism by TLC; 100% corresponds to the total amount of cell-associated NBD. D, basolateral sphingolipid transport during the 18°Cstep; 100% corresponds to the total amount of each sphingolipid recovered in the cells and the medium. In C and D a representative TLC analysis as well as the mean values and standard deviations of three independent experiments are shown. Bar, 20 m.
has been described (24). In contrast to its natural counterpart, the short chain fluorescent GlcCer analog can be translocated to the exoplasmic leaflet by the basolaterally localized multidrug resistance protein MRP1 to become accessible for BSA (23,37). Indeed, incubation with the MRP1 inhibitor MK571 significantly reduced the amount of C 6 NBD-GlcCer that became accessible to BSA but had no effect on the basolateral transport of C 6 NBD-SM (Fig. 3, A and B). The effect of MK571 on GlcCer transport was time-dependent, showing 60% inhibition of lipid translocation after 15 min, which diminished to approximately 35% after 30 min, whereas the effect had largely disappeared after 60 min. These data thus indicate that early (i.e. up to 15-30 min) after the onset of sphingolipid biosynthesis in HepG2 cells, the transport mechanism of SM and GlcCer from Golgi to the basolateral membrane at least partly differs. However, after longer incubation times, an increasing pool of the C 6 NBD-GlcCer is transported directly to the exoplasmic leaflet of the plasma membrane via a (slower) MRP-independent pathway that originates from an intracellular site that differs from the cytoplasmic surface of the Golgi. In contrast to trafficking of de novo synthesized GlcCer from Golgi to the basolateral membrane, GlcCer (and SM) exiting from SAC, after specific accumulation in this compartment following endocytic internalization (see "Experimental Procedures"), reaches the outer leaflet of the (apical) plasma membrane, entirely independent of multidrug resistance protein 1 (MDR1) activity (20).
To analyze how newly synthesized C 6 NBD-GlcCer and -SM FIG. 2. Sphingolipid transport to the exoplasmic leaflet of the BC membrane is reduced by PDMP. The cells were labeled with C 6 NBD-ceramide and incubated at 37°C in presence of BSA. BC membranes (arrows) were identified by phase contrast microscopy (A) and analyzed for C 6 NBD labeling by switching to epifluorescence (B). C, quantification of transport to the BC membrane. In control cells (white bars) the amount of labeled BCs reached a maximum after 30 min at 37°C and decreased after prolonged incubation. Treatment of the cells with sodium dithionite after the transport reduced the amount of labeled BCs to control levels ( are transported from the Golgi to the apical membrane, we examined the effect of CSA, an inhibitor of the MDR1 protein that is localized in the BC membrane of HepG2 cells (37). Incubation with 5 M CSA, which completely abolished the translocation of rhodamine 123 into the BCs (data not shown) neither affected the number of labeled BCs nor the accessibility of the C 6 NBD-labeled sphingolipids to sodium dithionite (Fig.  3C). In conjunction with the observation that upon inhibition of GlcCer biosynthesis BC labeling is reduced by almost 40% (see above), the results imply that the major part of the C 6 NBD-GlcCer (and -SM) is transported directly to the exoplasmic leaflet of the BC membrane, i.e. the lipid must have been localized in the lumenal leaflet of transport vesicles. To obtain direct evidence for this notion, we subsequently developed an assay to isolate the putative transport vesicles, released from TGN and SAC. C 6 NBD-SM and -GlcCer Can Be Recovered in the Lumenal Leaflet of Transport Vesicles Released from the TGN and the SAC-HepG2 cells were mechanically permeabilized to induce holes in the cell surface that are large enough for the release of transport vesicles. Because of cell detachment, the filter stripping method described previously (28,32) was not suitable for HepG2 cells, and permeabilization was therefore carried out by pressing the cells through a 25-gauge syringe. After this treatment 80 -90% of the cells were permeable for trypan blue and had largely lost their plasma membrane as revealed by phase contrast microscopy (not shown). Using this protocol, 15-20% of the total cell associated C 6 NBD-sphingolipids (ceramide, SM, and GlcCer) were released from the cells following the Golgi labeling procedure, whereas 25-30% of the total pool of C 6 NBD-SM and -GlcCer were released from SAC-labeled cells (Fig. 4). To remove any sphingolipids that remained exposed on the cytoplasmic leaflet of intracellular membranes following permeabilization, the cell pellet was washed with BSA. 30% of both C 6 NBD-ceramide and -GlcCer was localized in the cytoplasmic leaflet of the Golgi membrane (Fig. 4A). In contrast, only 10% of the newly synthesized C 6 NBD-SM was accessible to BSA. After labeling the SAC of HepG2 cells with C 6 NBD-SM and -GlcCer, 10 -15% of both sphingolipids were accessible to BSA after permeabilization (Fig. 4B).
After these treatments, the remaining cell-associated fractions of C 6 NBD-Cer and newly synthesized GlcCer and SM, associated with the Golgi, as well as the association of both sphingolipids with SAC, displayed a largely localized appearance, as reflected by bright fluorescent spots within the cellular lumen (not shown). Under these conditions, C 6 NBD-GlcCer and -SM should be localized solely in the lumenal leaflet of the SAC or the TGN/Golgi (Ref. 33; see also "Discussion"). Accordingly, if transport vesicles exclusively mediate subsequent transport and assuming that the topology is maintained, the lipids should now be recovered in the lumenal leaflet of such vesicles after their release and thus must be inaccessible to BSA and pelletable at high speed. After an incubation for 30 min at 37°C in presence of ATP and cytosolic proteins, 8% of either newly synthesized sphingolipid was recovered in the high speed pellet. In contrast, at 4°C or in the absence of ATP and cytosol only 2-4% of C 6 NBD-GlcCer and -SM were recovered (Fig. 5A). Comparably low vesicle budding was obtained when ATP or cytosol alone was omitted from the assay (data not shown). Essentially the same requirements applied for the release of C 6 NBD-SM and -GlcCer from the SAC. Again 2-3% of C 6 NBD-SM and -GlcCer were released at 4°C or at 37°C in the absence of ATP and/or cytosol. However, in presence of ATP and cytosol the vesicular release of both sphingolipids increased 2-3-fold, amounting to 7-8% of the initial cell-associated lipid fraction (Fig. 5B). The results thus show that the release of sphingolipid-containing vesicles from TGN and SAC in this assay is ATP-and cytosol-dependent and requires incubation at an elevated (37°C) temperature.
Prior to monitoring vesicle release, the permeabilized cells were extensively washed to ascertain that only lipid release in newly formed vesicles was monitored. Nevertheless, to exclude the possibility that the release of sphingolipids was due to nonspecific fragmentation of intracellular membranes, we monitored the activity of the Golgi-resident enzyme GTase in the cell lysate and the various fractions obtained during vesicle isolation (Table I). The specific activity of GTase in the membrane fractions that were released during permeabilization and the subsequent washing steps was similar to the specific activity in the cell lysate. 17% of the GTase was released during the first permeabilization step, whereas only a residual release was observed during the subsequent washing steps. More than 75% of the GTase activity was still intracellular prior to the final incubation step. During the incubation at 4°C, a very minor amount of protein was released. At 37°C more proteins were released, but the specific activity of GTase was much lower than in the other fractions, indicating that the membranes that were released in a temperature-dependent manner contained almost exclusively cargo proteins, whereas the amount of resident Golgi proteins was very low (Table I).
In principle, the transport vesicles that were released could have acquired NBD-labeled lipids that were transferred to the cell surface of intact cells, still present in the sample, as well as from plasma membrane fragments of the permeabilized cells. Although most cells were permeabilized during the preparation, an effort was made to examine this possibility and to estimate the contribution. Therefore, 1% BSA was added during the various incubation conditions. Only a minor fraction of C 6 NBD-SM and -GlcCer (3%) became accessible to BSA after labeling the SAC and a subsequent incubation at 4°C. The amount of sphingolipids was increased to 7-8% after incubation for 30 min at 37°C, irrespective of the addition of ATP and cytosol (Fig. 5D, hatched versus black bars). The ATP independ- ence demonstrates that the sphingolipids were released from cells that were left intact during the preparation. Although newly synthesized C 6 NBD-SM became accessible to BSA in significant amounts only after incubation at 37°C, this was not the case for C 6 NBD-ceramide and C 6 NBD-GlcCer. Even at 4°C 20% of the C 6 NBD-GlcCer and 10% of the C 6 NBD-ceramide became accessible to BSA (Fig. 5C). Incubation at 37°C did not markedly change the relative amount of released C 6 NBD-ceramide and -GlcCer.
Distinct C 6 NBD-SM and -GlcCer-containing Vesicles Are Released from the SAC but Not from the TGN-To analyze whether C 6 NBD-SM and -GlcCer are sorted in the SAC-and/or TGN membrane, which might be reflected by distinct vesicle populations enriched in either lipid, we fractionated the vesicle-containing supernatants by sucrose gradient centrifugation. The fractions were treated on ice for 5 min with sodium dithionite to ensure that only C 6 NBD-sphingolipids in the lumenal leaflet of transport vesicles were analyzed. In SACderived vesicles, C 6 NBD-SM was clearly enriched over C 6 NBD-GlcCer in vesicles obtained at a density corresponding to 30 -33% sucrose (w/w), whereas C 6 NBD-GlcCer was enriched at a sucrose density between 40 and 44% (Fig. 6A). Although neither lipid was exclusively retrieved in a particular vesicle fraction, it is evident that distinct "preferential" transport vesicles from the SAC for either sphingolipid can be identified.
To further characterize the potentially different nature of the carriers responsible for the exit of C 6 NBD-SM and -GlcCer from the SAC, we took advantage of the ability of NBD to photoconvert DAB into an osmiophilic precipitate that can be visualized by electron microscopy (36). Prior to incubation with DAB and processing for electron microscopy, the gradient fractions corresponding to 30 -33 and 40 -44% sucrose, respectively, were adjusted to 10% sucrose to avoid possible morphological alterations caused by osmotic effects. The 30 -33% fraction contained labeled and unlabeled spherical vesicles of different diameters. Especially small vesicles of approximately 50 nm in diameter were heavily labeled with DAB and are most likely those containing C 6 NBD-SM (arrows in Fig. 7, A and B). Such vesicles were only occasionally observed in the 40 -44% fraction, in which the main DAB-positive structures consisted of larger, cup-like vesicles with a diameter of 80 -100 nm (arrowheads in Fig. 7B). No DAB-positive, osmiophilic structures were obtained when the pellets were kept in the dark, although faint vesicular structures could be identified (Fig. 7, C and D).
The analysis of TGN-derived vesicles was hindered by the different amounts of the sphingolipids, because HepG2 cells synthesize much more C 6 NBD-SM than C 6 NBD-GlcCer. As observed for the SAC-derived vesicles, C 6 NBD-SM was found predominantly at a sucrose concentration of 30 -33% (Fig. 6B). C 6 NBD-GlcCer was recovered at slightly lower sucrose concentrations, in contrast to its location at a higher density seen for SAC-derived vesicles. However, the difference in localization between SM and GlcCer, present in the TGN-derived vesicles, was usually not large enough to support the view that sorting  3 and black bars). The cells were removed, and the supernatant was centrifuged at 100,000 ϫ g for 2 h. The sphingolipids in the high speed pellet (A and B) and in the supernatant (C and D) were extracted and analyzed by TLC. For each sample a representative TLC plate, and the mean values and standard deviations of at least three independent experiments are shown.

TABLE I Galactosyltransferase activity in fractions obtained during permeabilization and vesicle formation
To determine GTase activity unlabeled HepG2 cells were used in parallel experiments. BSA was omitted during the preincubation step to allow a correct protein determination. After the wash step the cell pellet was divided and incubated at either 4 or 37°C. The cell lysate, the supernatants obtained during permeabilization (S1), preincubation (S2), and wash step (S3), and the released vesicles were centrifuged at 100,000 ϫ g for 2 h. The protein content of the cell lysate was set to 100, and the protein contents of all other samples were correlated to this value. The GTase activity of the cell lysate was set to 1, and the other values were correlated to it. Finally the amount of GTase in each sample was determined by multiplying the specific activity with the protein content. From the values obtained for the vesicular pellets at 37 and 4°C, the temperature-dependent release of GTase was calculated. The means and standard deviations of three independent experiments are shown. of the two sphingolipids into distinct transport vesicles occurred.
The Release of C 6 NBD-SM-containing Vesicles from the SAC and the TGN Is Differently Affected by the Calmodulin Antagonist TFP-We have previously shown that C 6 NBD-SM and -GlcCer are located in different domains in the SAC. Among others, the evidence relied on the selective regulation of their exiting to the apical and basolateral domain upon treatment of the cells with the calmodulin antagonist TFP (38). The preferential localization of SM and GlcCer in distinct SAC-derived transport vesicles (Fig. 6A) is therefore fully consistent with this notion. To further corroborate these observations, the present assay system provides the possibility to investigate in a direct manner whether TFP modulates vesicular exiting of the sphingolipids.
In intact cells, TFP inhibited the transport of C 6 NBD-SM and -GlcCer from the SAC to the basolateral membrane, whereas it had no effect on the recycling of C 6 NBD-GlcCer between SAC and BC membrane (38). When 20 M TFP was included during the BSA step after the permeabilization and the budding assay, the temperature-dependent release of C 6 NBD-SM from the SAC was inhibited by 50%, whereas there was no effect on the release of C 6 NBD-GlcCer (Fig. 8A). At higher concentrations of TFP (50 M), the release of C 6 NBD-SM was inhibited by 70%, and a slight effect on the release of C 6 NBD-GlcCer was observed (20% inhibition; not shown).
Surprisingly, incubation with 20 M TFP increased the vesicular release of C 6 NBD-SM from the TGN by 30%, whereas it had no effect on the release of C 6 NBD-GlcCer (Fig. 8B). It is unlikely that this increased release of C 6 NBD-SM was due to nonspecific fragmentation of the TGN, because higher concentrations of TFP (100 M) slightly inhibited the release of C 6 NBD-SM and abolished the release of C 6 NBD-GlcCer (Fig.  8B). Interestingly, the data thus suggest that distinct mechanistic features are involved in mediating transport vesicle assembly and/or release from SAC on the one hand and Golgi on the other. DISCUSSION The present study demonstrates that the mode of sorting of SM and GlcCer during their exit from Golgi and SAC, the major sorting compartments in the biogenesis and maintenance of polarized membrane domains, differs. This is reflected by the recovery of transport vesicles, distinctly enriched in these sphingolipids. In addition, sphingolipid sorting in HepG2 cells is mainly associated with the SAC, highlighting its prominent role in polarity development of these cells.
In Vitro Budding System and Sphingolipid Release from SAC and TGN-To appreciate the involvement of SAC and TGN, including their hierarchy, in the polarized transport itinerary of hepatocytes, the isolation and characterization of the respective transport vesicles will be imperative. In fact, a variety of in vitro systems have been developed to analyze vesicular transport of proteins from the TGN to the plasma membrane (28 -30). Also in hepatocytes distinct constitutive pathways from the TGN to the basolateral plasma membrane have been described (39,40). However, although lipids, especially sphingolipids, play an important role in the polarized sorting of apical and basolateral proteins in epithelial cells (11), only very few in vitro systems have been developed to study the polarized sorting of de novo synthesized sphingolipids in the context of the assembly of transport vesicles (31,32). Therefore, FIG. 6. Distinct C 6 NBD-SM and C 6 NBD-GlcCer-containing vesicles are released from the SAC but not from the TGN. HepG2 cells were incubated with a mixture of C 6 NBD-SM and -GlcCer (A) or C 6 NBD-ceramide (B). After permeabilization and BSA treatment the cells were incubated at 37°C in the presence of BSA, ATP, and cytosol. The cells were removed, and the supernatant was layered on top of a sucrose gradient consisting of 20 -50% sucrose (w/w) in transport buffer. The gradient was centrifuged for 16 h at 100,000 ϫ g and fractionated. Fraction 1 corresponds to the bottom of the gradient. The fractions were treated with sodium dithionite for 5 min on ice, and the sphingolipids were extracted and analyzed by TLC. In A the relative amount of C 6 NBD-SM (f) and C 6 NBD-GlcCer (Ⅺ) in each fraction is calculated to demonstrate the enrichment of one sphingolipid over the other at a particular sucrose concentration. In B the distribution of C 6 NBD-SM and -GlcCer over the whole gradient is shown. For both TGN-and SAC-derived vesicles a representative TLC plate as well as the mean and extreme values of two gradients are shown. as a first approach, an in vitro assay was developed to characterize transport and sorting of sphingolipids from either compartment. To this end, we used mechanically permeabilized HepG2 cells that had been labeled with C 6 NBD-tagged sphingolipid analogs. Our data reveal that the release of C 6 NBD-SMand -GlcCer-containing vesicles from both compartments was temperature-dependent and required the addition of ATP and cytosolic proteins. A significant contribution of nonspecific fragmentation of the organelles to the observed sphingolipid transport is excluded, because no significant release of the Golgi marker GTase occurred during vesicle formation, whereas the TGN-derived vesicles (Fig. 5A) did not contain significant amounts of the Golgi-localized C 6 NBD-ceramide.
From an experimental point of view it is relevant to emphasize that, although after internalization at the plasma membrane a fraction of the fluorescent sphingolipid analogs may be transported to the Golgi, as occurs in undifferentiated cells (35,41,42), in polarized HepG2 cells both C 6 NBD-GlcCer and -SM are specifically accumulated in the SAC. As shown previously (20,33), localization in this compartment can be readily distinguished from the Golgi apparatus, e.g. in its sensitivity to monensin and nocodazole. Indeed, the present results demonstrate that, prior to their vesicular release, newly synthesized and internalized sphingolipid analogs are located in the intended "target" compartments, i.e. Golgi and SAC. This is, for example, apparent from the distinct effects of TFP on the release of SM (Fig. 8), present in either compartment, or the differences in accessibility of GlcCer to BSA at 4°C (Fig. 5,  compare C and D).
In the context of the latter observation, it is interesting that only de novo synthesized C 6 NBD-GlcCer, after its translocation to the lumenal leaflet in the Golgi, became accessible to BSA at 4°C. Because all accessible sphingolipid analogs had been re-moved from the cytoplasmic leaflet of internal membranes prior to the incubation, this implies that a temperature-independent (and energy-independent) activity exists in the Golgi membrane that translocates C 6 NBD-GlcCer but not C 6 NBD-SM from the lumenal to the cytoplasmic leaflet. Because such a translocation was not observed in the SAC-labeled membranes, the data reveal that the compartments display different properties concerning C 6 NBD-GlcCer translocation and emphasize Golgi specificity rather than a nonspecific translocation of the lipid caused by the C 6 NBD group.
Finally, although these analogs resemble the properties of natural sphingolipids in many aspects (42,43), a direct comparison would be preferable. However, to specifically label SAC with sphingolipids or to accumulate newly synthesized sphingolipids in Golgi, the use of exchangeable or modifiable (dithionite quenching) sphingolipids is essential. In this manner, "contamination" of SAC-or Golgi-derived transport vesicle fractions with (natural) sphingolipid-containing vesicle fractions of similar density, but derived from different sources (plasma membrane, endocytic pathway), is avoided. Besides, a lack of suitable markers currently precludes further purification of the recovered vesicle fractions. It is most relevant, however, that the analogs used are sorted in SAC in a manner that depends on the progress of polarity (20,21). This emphasizes their validity as markers for apically and basolaterally directed transport from SAC (or TGN) and for characterization of the respective transport vesicles (see also below).
Sorting of SM and GlcCer in SAC and TGN Prior to Vesicle Release-The present data reveal that despite the transcytotic processing of apical proteins (12), the Golgi of hepatic cells does display (direct) sphingolipid sorting capacity. This fact gains considerable significance by recent observations that several hepatic apical proteins may reach the apical membrane by direct transport (14,15,44), which might link sphingolipid sorting to raft assembly. In this regard, questions have been raised concerning the ability of NBD lipids to partition into rafts, triggered in artificial liposomal systems, given their relative water solubility (45). However, in biological systems it has been shown that exogenous addition of C 6 NBD-SM to de- In some experiments cells labeled with C 6 NBD-ceramide were incubated with 100 M TFP (lane 3 and black bars in B). The cells were removed, and the supernatant was centrifuged at 100,000 ϫ g. The sphingolipids in the vesicle pellet were extracted and analyzed by TLC. To calculate the temperature-dependent release of sphingolipids, the values obtained at 4°C were subtracted from those obtained at 37°C, and the resulting values for the control cells were set as 100. Representative TLC plates, showing the release at 37°C, and the mean values and standard deviations of at least three independent experiments are shown. veloping neurons triggered raft domain formation and the apical sorting of the Thy-1 receptor to axonal domains (46). Other, but similar, short chain fluorescent (BODIPY) sphingolipid analogs are sorted in a cholesterol-dependent, i.e. a potentially raft-mediated mechanism, in fibroblasts derived from patients suffering from various sphingolipid storage diseases (47). Nevertheless, further work will be needed to directly clarify these issues.
The present data are schematically summarized in Fig. 9. During the C 6 -NBD-ceramide labeling at 18°C (Fig. 9A, steps 1  and 2), not only C 6 NBD-SM but also C 6 NBD-GlcCer accumulated in the Golgi. This differs from LLC-Pk1 cells (23) and indicates that in liver cells, C 6 NBD-GlcCer release from the TGN is also largely vesicular. Inhibition of MRP1 activity confirmed previous findings (23) that transport of C 6 NBD-SM to the plasma membrane is exclusively in the lumenal leaflet of transport vesicles (Fig. 9A, step 3a). Because C 6 NBD-GlcCer is partly transported to the cytoplasmic leaflet of the basolateral membrane, prior to its translocation to the exoplasmic leaflet (Fig. 9A, step 4), the temperature dependence of this process (Ͼ18°C) may suggest involvement of transport vesicles (Fig.  9A, step 3b). Nevertheless, monomeric transport, possibly mediated by a glycosphingolipid transfer protein (48), cannot yet be excluded (Fig. 9A, step 3c). Interestingly, even in presence of MK571, the majority of the C 6 NBD-GlcCer became accessible to BSA after prolonged incubation. These data indicate that only in an early phase (15-30 min) after de novo synthesis of GlcCer, may vesicle-independent transport occur. At later stages, when the lipid has acquired its correct topological transmembrane distribution, transport of GlcCer is predominantly of a vesicular nature.
Interestingly, inhibition of the apically localized MDR1 had no effect on the number of labeled BCs and the accessibility of the lipids to sodium dithionite, indicating that the sphingolipid analogs were transported directly to the exoplasmic leaflet of the BC membrane. Furthermore, inhibition of GlcCer synthesis reduced the number of labeled BCs, in spite of the fact that the pool of C 6 NBD-SM, synthesized in control cells, is more than 3-fold in excess of that of GlcCer. Together, these results imply that (i) C 6 NBD-GlcCer is, in relative amounts, the major sphingolipid that is transported to the BC membrane, whereas C 6 NBD-SM is predominantly transported to the basolateral plasma membrane (indicated in Fig. 9A, steps 3a and 3d) and (ii) the majority of apically transported C 6 NBD-GlcCer is localized in the lumenal leaflet of transport vesicles. This is in contrast to studies on LLC-Pk1 cells, where MDR1 activity was required for the transport of C 6 NBD-GlcCer to the outer leaflet of the apical membrane (23,27), thus indicating cell type-dependent differences in sphingolipid sorting in polarized cells.
By using the in vitro assay, direct evidence demonstrated that C 6 NBD-GlcCer and -SM, after their accumulation in the SAC, are released in distinct transport vesicles (Fig. 9B). Thus, TFP, which blocks the exit of C 6 NBD-SM but not that of C 6 NBD-GlcCer from the SAC in intact cells (38), inhibited only the release of SM-containing vesicles but had no effect on the release of C 6 NBD-GlcCer (Fig. 8). This shows that the vesicular release of C 6 NBD-SM and -GlcCer is differentially regulated and indicates that both sphingolipids are localized in distinct membrane domains of the SAC, presumably accomplished by lateral sorting in the luminal leaflet (38). Intriguingly the sucrose fractions enriched in the apically sorted C 6 NBD-GlcCer contained DAB-positive cup-shaped vesicles. These are reminiscent of the vesicles derived from the common recycling endosome of MDCK cells that contain predominantly the apically directed polymeric immunoglobulin receptor (7). Because the fractions of the 30 -33% sucrose contain virtually no C 6 NBD-GlcCer, we assume that the small DAB-positive vesicles found there contain exclusively C 6 NBD-SM, which is predominantly transported to the basolateral membrane under our experimental conditions. Similarly, in MDCK cells the transferrin receptor is released from recycling endosomes predominantly in small vesicles of 60-nm diameter (7). Further work should clarify whether the C 6 NBD-SM identified in the heavier gradient fractions is localized only in the small vesicles that were occasionally seen or whether it is also localized in the larger cup-shaped vesicles. If so, another issue would be whether C 6 NBD-GlcCer and -SM are present in the same vesicles or whether they are sorted into distinct but morphologically similar vesicles.
Using different epithelial cells, it has been shown that de novo synthesized sphingolipid analogs can be sorted into distinct transport vesicles at the Golgi (31,32). However, a more pronounced sorting capacity of these lipids in the SAC of HepG2 cells is in line with the fact that in hepatocytes most newly synthesized apical proteins are transported to the basolateral membrane, prior to their transport to the apical membrane via transcytosis. In WIF-B cells, transcytosis via an endosomal subapical compartment has been demonstrated (49), implying that a heavy burden may be placed on the sorting capacity of the SAC. Cell type-dependent differences in the relative contribution of distinct sorting sites (SAC versus TGN) may also explain that sorting of sphingolipid analogs was not observed in the transcytotic pathway in MDCK cells (50). Indeed the initial sorting observed for de novo synthesized sphingolipids was abolished after prolonged incubation times, indicating that the SAC equivalent in these cells (i.e. the common endosome) has a lower sorting capacity than in HepG2 cells and underlining the importance to investigate vesicular transport in different cell types to fully understand polarity establishment. The current approach may provide an excellent opportunity to analyze the mechanisms that mediate the sorting of sphingolipids and proteins in the SAC (and the TGN), by careful analysis of the molecular mechanisms that underlie the formation of the transport vesicles and the regulation of their release.