Evidence for Apical Endocytosis in Polarized Hepatic Cells: Phosphoinositide 3-Kinase Inhibitors Lead to the Lysosomal Accumulation of Resident Apical Plasma Membrane Proteins

Male Sprague-Dawley rats (125–150 g) were supplied by Charles River Breeding Laboratories. Wortmannin, cycloheximide, cytochalasin D, and nocodazole were purchased from Sigma Chemical Co. Wortmannin, cytochalasin D, and nocodazole were stored at −20°C as 10-, 1-, or 16.5-mM stock solutions, respectively, in DMSO. Cycloheximide was prepared as a 10-mg/ml stock solution in 5% ethanol and used directly. LY294002 was purchased from Calbiochem Corp. and stored at −20°C as a 10-mM solution in DMSO. Cell culture media and FBS were purchased from GIBCO BRL. Fluorescently labeled secondary antibodies were obtained from Jackson ImmunoResearch Laboratories. Anti–β-tubulin antibodies were obtained from Sigma Chemical Co. Texas red–conjugated phalloidin was purchased from Molecular Probes Inc., and stored at −20°C as a 200-U/ml stock in methanol. The antibodies recognizing the 120-kD lysosomal glycoprotein (LGP-120), mannose 6-phosphate receptor (M6P-R), 5′nucleotidase (5′NT), and endogenous canine plasma membrane antigens (3F2 and G12) were kindly provided by W. Dunn (University of Florida, Gainesville, FL), Peter Nissley (National Institutes of Health, Bethesda, MD), Paul Luzio (Cambridge University, Cambridge, UK), and George Ojakian (State University of New York, Oswego, New York), respectively. Antibodies against aminopeptidase N (APN), CE9, HA4, asialoglycoprotein receptor (ASGP-R), HA321, syntaxin 3, and endolyn-78 were all prepared by the Hubbard laboratory and have been described elsewhere (Bartles et al., 1985; Scott and Hubbard, 1992; Barr and Hubbard, 1993; Ihrke et al., 1993).

WIF-B cells were grown in a humidified 7% CO₂ incubator at 37°C as described (Shanks et al., 1994). In brief, cells were grown in modified Ham's F12 medium, pH 7.0, supplemented with HAT (10 μM hypoxanthine, 40 nM aminoterpin, 1.6 μM thymidine) and 5% FBS. MDCK cells were grown at 37°C in a 5% CO₂ humidified incubator as described (Weisz et al., 1992). For indirect immunofluorescence experiments, cells were seeded onto glass coverslips at 1.3 × 10⁴ cells/cm² and cultured for 8–12 d (WIF-B) or 4–5 d (MDCK) until they reached maximal density and polarity (Shanks et al., 1994).

To examine the effects of wortmannin or LY294002 on the steady state distributions of various proteins, cells were incubated at 37°C up to 3 h in their respective serum-free culture medium buffered with either 20 mM Hepes, pH 7.0 (for WIF-B cells), or 44 mM NaHCO₃, pH 7.0 (for MDCK cells), in the presence or absence of either agent (see Results or figure legends for details). After treatment, cells were rinsed briefly in PBS and placed on ice, fixed with chilled PBS containing 4% paraformaldehyde for 1 min, and permeabilized with methanol (also chilled) for 10 min (Ihrke et al., 1993). Cells were rehydrated in PBS by three washes of 5 min each. Cells were further processed for single- or double-labeled indirect immunofluorescence according to previously published methods (Ihrke et al., 1993) with the following primary antibodies: anti–HA321, –LGP-120 and –APN (rabbit polyclonals, 1:100, 1:200, and 1:300, respectively), anti– 5′NT, –endolyn and –HA4 (mouse monoclonal ascites, 1:300, 1:500, and 1:100, respectively). MDCK cells were processed for indirect immunofluorescence using anti–3F2 and –G12 (hybridoma culture supernatants, 1:10). The secondary antibodies (FITC or Cy3 goat anti–rabbit or anti–mouse) were used at 5–10 μg/ml.

To assess the effects of microtubule, actin, or protein synthesis disruption on redistribution, cells were pretreated for 1 h at 37°C with nocodazole (33 μM), cytochalasin D (1 μM), or cycloheximide (25 μg/ml), respectively. Cells were incubated an additional 3 h at 37°C in the presence of wortmannin, and the continued presence of either nocodazole, cytochalasin D, or cycloheximide. The treatments were stopped by fixation and the cells were processed for indirect immunofluorescence. Anti–β-tubulin antibodies (mouse monoclonal) were diluted to 1:500. Texas red–conjugated phalloidin was diluted to 5 U/ml.

Livers from fasted (12–18 h) rats (125–200 g) were surgically removed and perfused in a recirculating balanced salt solution as described (Dunn et al., 1983) in the absence or presence of 2 μM wortmannin and 50 μM leupeptin. Biopsies (0.25–1.0 g) were removed at the indicated times and fixed by immersion in Krebs buffer containing 4% PFA for at least 1 h at room temperature. The intact liver was fixed by perfusion with 200–300 ml of the same fixative. After fixation, the biopsies and intact liver were processed as described (Dunn et al., 1983) and semithin sections (0.5 μm) cut with a Reichert Jung Ultracut E microtome. The liver sections were processed for immunofluorescence according to previously published methods (Hubbard et al., 1985) using the following primary antibodies: anti–APN, –HA321 (rabbit polyclonals, 1:300 and 1:100, respectively), anti–HA4 and –endolyn (mouse monoclonal ascites, 1:100 and 1:300, respectively). The secondary antibodies (FITC or Cy3 goat anti–rabbit or anti–mouse) were used at 5–10 μg/ml.

The transcytosis assays were performed as described (Ihrke et al., 1998). WIF-B cells were placed in Hepes-buffered, serum-free medium and treated at 37°C with the agents as specified in Results or the figure legends. The cells were placed on ice for 5 min to cool to 4°C. Antigens present at the basolateral cell surface were labeled (those also present at the apical plasma membrane are excluded from labeling) at 4°C for 15 min with anti–APN (rabbit polyclonal, 1:50) or purified mouse monoclonal antibodies against 5′NT (purified IgG fraction, 20–50 μg/ml) or CE9 (mouse monoclonal ascites, 1:25). Total IgG was recovered from mouse ascites using EZ Sep (Pharmacia LKB Biotechnology Inc.) according to the instructions from the manufacturer. All antibodies were diluted in buffered, serum-free medium containing 2 mg/ml BSA in the absence or presence of the various agents. After labeling, the cells were washed extensively in buffered, serum-free medium containing 2 mg/ml BSA to get rid of excess and nonspecifically associated antibodies. Cells were then returned to 37°C and incubated for the desired times of chase in the absence or presence of the different agents (see Results or figure legends). Treatments were stopped by fixation and permeabilization as described above. The trafficked antibodies were labeled with Cy3- or FITC-conjugated goat anti–rabbit or anti–mouse secondary antibodies (5–10 μg/ml) and visualized by indirect immunofluorescence.

The transcytosis assay was performed as described above with anti–5′NT antibodies (20–50 μg/ml) directly conjugated to 5-nM gold particles (Slot and Slot, 1985). The cells were fixed and processed in situ (on the glass coverslip) using standard Epon embedding techniques (Ihrke et al., 1998). Ultrathin sections were cut parallel to the growth substrate, stained with lead citrate, visualized, and photographed with an EM10 transmission electron microscope (Carl Zeiss, Inc.).

The cells and perfused liver sections were visualized by epifluorescence (Axioplan Universal Microscope; Carl Zeiss, Inc.). Pictures were taken and processed using standard photographic techniques. Scanned images of the photographic prints were compiled, and figures prepared using Adobe Photoshop and Microsoft PowerPoint software. For Fig. 4, WIF-B immunofluorescence was analyzed using a laser scanning confocal microscope (LSCM 410; Carl Zeiss, Inc.).

Our initial approach to examine the effects of the selective inhibitors, wortmannin and LY294002, on apical plasma membrane dynamics was to determine the steady state distributions of apical membrane proteins in WIF-B cells in treated cells. As observed for other cell types, these agents led to extensive vacuolation (compare Fig. 1 a with c, e, and g). By indirect immunofluorescence, we observed that the apical membrane proteins, 5′NT and APN, accumulated in vacuoles after treatment with 100 nM wortmannin or 200 μM LY294002 (compare Fig. 1 b with d, f, and h). Similar patterns of redistribution were observed for dipeptidyl peptidase IV, another apical membrane protein (data not shown). Vacuole size in cells treated with LY294002 was consistently smaller than observed in wortmannin-treated cells for all markers examined. This may reflect differences in the inhibitory mechanisms of these agents or the decreased potency of LY294002 (Vlahos et al., 1993; Ui et al., 1995).

Vacuole formation and accumulation of the apical proteins into vacuoles were both time dependent, with vacuole formation preceding redistribution (Fig. 2, a and b). By phase microscopy, vacuoles were observed within 15 min of treatment and, by 45 min, all cells contained numerous, large vacuoles (Fig. 2, a and b). In contrast, apical proteins were first faintly detected in vacuoles after 45 min of treatment and did not reach their peak redistribution until after 120–180 min, when all cells were positive for vacuole staining (Fig. 2, a and b). The vacuoles containing apical proteins increased in diameter as wortmannin treatment was prolonged. At 45 min, apical protein-positive vacuoles were ∼1.9 ± 0.5 μm in diameter, whereas by 180 min, the diameter had increased to an average of 3.1 ± 0.9 μm (Table I). Both the increase in vacuole size and gradual accumulation of apical protein staining suggest the inhibitors allowed forward traffic into vacuoles, but blocked transport from them. Alternatively, the inhibitors may have altered the relative kinetics of membrane traffic to and from the vacuole.

The wortmannin concentration used in this and many other published studies was 100 nM. At higher concentrations, wortmannin can inhibit other enzymes, including myosin light chain kinase and PI 4-kinase (Ui et al., 1995). To determine whether the effects described here were due to specific inhibition of PI 3-kinase isoforms, we incubated WIF-B cells with varying concentrations of wortmannin. The extents and rates of accumulation of the apical proteins and vacuole formation were similar in cells treated with wortmannin concentrations ranging from 10 nM to 6 μM (data not shown). Also, increased times of treatment with either agent (up to 10 h) did not lead to visibly greater extents of accumulation (see Discussion).

Not all apical membrane proteins accumulated in vacuoles. Although the cells were heavily vacuolated after 180 min of treatment, both HA4 (Fig. 3 a) and syntaxin 3 (data not shown) maintained their apical distributions. This suggests that wortmannin and LY294002 were selectively perturbing protein trafficking rather than generally disrupting apical membrane integrity. Interestingly, the distributions of the resident basolateral plasma membrane proteins HA321 (Fig. 3 b) and CE9 (data not shown) did not change in the presence of either inhibitor, indicating the effect was specific to proteins at the apical surface. In addition, the staining patterns of the tight junction component, ZO-1, were not changed in the presence of wortmannin (data not shown), suggesting tight junctions remained intact during treatment. Together, these data indicate that plasma membrane polarity was not compromised in treated cells.

We next determined what other molecules were present in the vacuoles containing apical proteins. The distributions of TGN-38, p115, and albumin (a major hepatic secretory protein) were unaltered by wortmannin treatment (data not shown). This is in agreement with findings from other cell types that wortmannin does not affect organelles or transport steps of the biosynthetic pathway (Brown et al., 1995; Hansen et al., 1995; Martys et al., 1996; Reaves et al., 1996; Shpetner et al., 1996). However, disruption of endocytic compartments has been reported in nonpolarized cells. In particular, the formation of large intracellular vacuoles derived from early and late endosomes and lysosomes has been observed (Brown et al., 1995; Reaves et al., 1996; Shepherd et al., 1996; Shpetner et al., 1996; Spiro et al., 1996; Fernandez-Borja et al., 1999). We examined whether wortmannin induced similar morphological changes in WIF-B cells. For these experiments, cells were treated for 180 min with 100 nM wortmannin and the steady state distributions of three recycling membrane proteins, ASGP-R (an early endosomal membrane protein), M6P-R (a late endosomal membrane protein) and endolyn-78 (a lysosomal membrane protein), were examined by confocal microscopy. The most notable vacuolation was observed when cells were stained for endolyn-78 (Fig. 4, a, c, and e) or LGP-120 (data not shown). Unlike reports from other cell types, few vacuoles contained early or late endosomal markers (here, ASGP-R or M6P-R; Fig. 4, d and f). Rather, these markers were primarily found in intracellular structures that did not significantly overlap with the endolyn-78 positive structures. These results indicate that wortmannin was disrupting mainly lysosomal compartments in WIF-B cells.

Comparison of APN and endolyn-78 staining patterns revealed striking colocalization (Fig. 4, a and b). After 180 min of wortmannin treatment, all cells contained vacuoles that were positive for both APN and endolyn-78 (Fig. 4, a and b). Similar results were obtained when 5′NT distributions were compared with those of another resident lysosomal membrane protein, LGP-120 (data not shown). The kinetics of appearance of endolyn or LGP-120 into vacuoles generally paralleled those of vacuole formation itself and preceded the appearance of apical plasma membrane proteins (Fig. 2, a and b). By 45 min, the majority of cells examined contained vacuoles that were positive for endolyn-78 and LGP-120 and, by 120 min, redistribution was complete. The appearance of LGP-120 in vacuoles consistently trailed the appearance of endolyn-78. Whether this reflects differences in their cellular itineraries or kinetics (Ihrke et al., 1998) is not known. From these data, we conclude that the apical plasma membrane proteins accumulated in lysosomally derived vacuoles in the presence of PI 3-kinase inhibitors.

The apical proteins found in vacuoles could have originated from three possible sources: from the biosynthetic pathway, the transcytotic pathway, or from the apical plasma membrane itself. To rule out a biosynthetic origin, WIF-B cells were pretreated in the absence or presence of 25 μg/ml cycloheximide for 60 min to inhibit protein synthesis. Cells were further incubated for 180 min with 100 nM wortmannin in the continued absence or presence of cycloheximide. Cells treated with wortmannin alone or with cycloheximide and wortmannin achieved the same levels of vacuole formation (100% positive cells) and 5′NT vacuole redistribution (92 and 93% positive cells, respectively). Cells treated with cycloheximide alone were not observed to vacuolate, nor did apical plasma membrane protein distributions change. Similar results were obtained for APN, further indicating that newly synthesized apical proteins were not the source of apical molecules accumulating in vacuoles.

We next investigated the possibility that the apical membrane proteins observed in vacuoles originated from the transcytotic pathway. For these experiments, WIF-B cells were pretreated in the absence or presence of 100 nM wortmannin for 15 min at 37°C. Transcytosis of apical proteins was examined after their labeling with specific antibodies at 4°C. Since the antibodies were excluded from the apical plasma membrane by tight junctions, only those molecules en route from the basolateral plasma membrane were labeled (Ihrke et al., 1998; and see Fig. 5, a and d). After extensive washing, the cells were warmed to 37°C and the antibody–antigen complexes chased for the indicated times. The cells were fixed and permeabilized and the trafficked antibody–antigen complexes were visualized with secondary antibodies.

After 90 min in control cells, the majority of 5′NT staining was observed at or near the apical plasma membrane and, after 180 min, nearly all transcytosing 5′NT was at the apical surface signaling its successful delivery (Fig. 5, a–c, and see the ultrastructural analysis, g and h). Similarly, in wortmannin-treated cells, 5′NT was also successfully delivered to the apical plasma membrane by 180 min. However, we observed more intracellular accumulation of the trafficked molecule after 90 min (Fig. 5 e). Similar results were obtained for APN and HA4 in both control and wortmannin- or LY294002-treated cells (data not shown). We also monitored the dynamics of the resident basolateral protein, CE9, using the antibody-labeling and trafficking approach. In wortmannin-treated cells, CE9 maintained its basolateral distributions (data not shown), further indicating that PI 3-kinases differentially regulate resident proteins at each domain.

Because only a 15-min wortmannin pretreatment was used in these experiments, we needed to rule out the possibility that transcytosing molecules were transported past a potential point of diversion before vacuoles were fully formed. Therefore, we pretreated cells for 120 min with wortmannin before antibody labeling of live cells. As shown in Fig. 6, transcytosis remained fully functional. We observed no further impairment of 5′NT delivery to the apical surface and no increased intracellular staining relative to cells pretreated for only 15 min. Importantly, we also determined the steady state distributions of APN in the same cells to determine if transcytosing 5′NT molecules appeared in vacuoles. Comparison of the staining patterns in Fig. 6, a and b, and c and d, reveals that the structures containing trafficking 5′NT were distinct from vacuoles. Some 5′NT molecules were found in discrete structures surrounding the APN-positive vacuoles, but the patterns never overlapped.

These results revealed that basolateral-to-apical transcytosing molecules in wortmannin-treated cells did not accumulate in vacuoles, but seemed to traverse similar membrane compartments (albeit displaced or delayed) as in control cells. Therefore, we turned to ultrastructural analysis of antibody-antigen complexes in wortmannin-treated cells to further examine the compartments through which the molecules were traveling. In particular, we were interested in determining whether 5′NT traveled through the SAC, a known intermediate of the transcytotic pathway in hepatocytes (Barr and Hubbard, 1993; Barr et al., 1995; Ihrke et al., 1998). As shown in Fig. 5, gold-conjugated anti–5′NT antibodies were detected at the apical surface in both control and treated cells after 90 min of trafficking. As predicted, gold particles were also observed just adjacent to the apical cell surface in the SAC. Examination of the SAC in control and treated cells revealed no significant differences in its placement, size, or number (Table II). These data indicate that the SAC remained functional during treatment and further show that the intracellular structures containing trafficking 5′NT in wortmannin-treated cells were not an altered SAC. The identity of the compartment containing trafficking 5′NT in treated cells is not known, but may likely be an altered early endosomal compartment, another known intermediate of the transcytotic pathway in hepatocytes (Barr and Hubbard, 1993; Barr et al., 1995; Ihrke et al., 1998). Preliminary studies revealed that these structures stained positive for ASGP-R (data not shown). Also, when 5′NT and ASGP-R were trafficked together and visualized, both molecules traversed a similar compartment (data not shown), presumably an early endosome.

To directly determine whether the resident apical plasma membrane proteins in vacuoles originated from the apical cell surface, we examined the trafficking of antigen–antibody complexes that were first delivered to the apical surface. For this analysis, APN or 5′NT molecules were labeled at the basolateral surface at 4°C, chased to the apical membrane for several hours at 37°C, chased an additional 2 h in the presence or absence of wortmannin, and labeled with secondary antibodies for immunofluorescent detection. As shown in Fig. 7, b and h, nearly all the labeled APN or 5′NT molecules were at the apical surface in control cells. After the final 2 h chase in the presence of wortmannin, APN and 5′NT staining was observed in vacuoles (Fig. 7, d, f, and j). These data indicate that the apical membrane proteins found in vacuoles originate from the apical surface. This is the first direct evidence for internalization of membrane proteins from the hepatic apical cell surface.

To further characterize the mechanism(s) by which resident apical membrane proteins were internalized and accumulated intracellularly, we examined the effect of cytoskeletal disruption on redistribution. Microtubules were depolymerized by addition of 33 μM nocodazole for 1 h at 37°C (as assayed by β-tubulin staining, data not shown). The cells were further incubated for 180 min with wortmannin in the continued absence or presence of nocodazole. As shown in Fig. 8, the addition of nocodazole prevented vacuole formation, further implying that vacuolar compartments were enlarged due to accumulated cargo and membrane received from vesicular intermediates. Consistent with this conclusion, microtubule disruption also prevented any detectable intracellular accumulation of APN (Fig. 8 d). Similar experiments were performed using 1 μM cytochalasin D to disrupt the actin cytoskeletal network. Although actin filaments were depolymerized, no changes in the kinetics of vacuole formation or apical protein redistribution were observed (data not shown).

Because the effects of PI 3-kinase inhibitors on protein trafficking patterns and cellular morphology have been examined only in cultured cells, it was important to rule out the possibility of a tissue culture–induced effect by the inhibitors. Therefore, we treated hepatocytes in isolated perfused livers with 2 μM wortmannin for 180 min. We observed hepatocyte vacuolation with positive staining for APN (Fig. 9 b') and 5′NT (data not shown). As observed for WIF-B cells, the vacuoles also contained the resident lysosomal membrane proteins, endolyn-78 (Fig. 9 b') and LGP-120 (data not shown). HA321 and HA4 distributions did not change in treated livers, indicating that similar wortmannin-sensitive transport steps were operating in intact hepatocytes as in WIF-B cells (Fig. 9, d and f). These results indicate that regulation of apical internalization by wortmannin-sensitive molecules occurs in the intact liver.

To determine whether vacuolar accumulation of apical membrane proteins is a phenomenon specific to hepatocytes, we examined the effects of wortmannin and LY294002 on the distributions of membrane proteins in MDCK cells. As shown in Fig. 10, wortmannin induced the formation of large vacuoles that also stained positive for the endogenous resident apical membrane protein, 3F2 (compare Fig. 10, c with d). The distribution of the basolateral resident membrane protein, G12, remained unaltered by wortmannin treatment (data not shown). The kinetics of vacuolar formation and apical protein redistribution in MDCK cells were similar to those observed in WIF-B cells. Within 30 min, all cells contained numerous large vacuoles (Fig. 10 e). Vacuolar 3F2 staining was observed after 30 min of treatment in ∼50% of cells and, by 120 min, nearly all cells examined contained vacuoles positive for 3F2 (Fig. 10 e). These results indicate that PI 3-kinase is regulating the dynamics of resident plasma membrane proteins in MDCK cells and suggest that similar processes may be operating in other polarized epithelial cell types.

Using a pharmacological approach, we have documented for the first time the internalization of apical membrane proteins from the canalicular surface of hepatocytes. When the PI 3-kinase inhibitors, wortmannin and LY294002, were applied to WIF-B cells or intact liver hepatocytes, large lysosomal vacuoles formed that contained apical plasma membrane proteins. Residents of other organelles, including the basolateral plasma membrane and early or late endosomes were completely or largely absent from these structures. The source of the apical membrane proteins was the apical plasma membrane itself.

Our study has several important implications. First, our finding that PI 3-kinase inhibitors affect apical endocytosis differently from basolateral endocytosis in both MDCK and hepatic cells adds to the list of differences between endocytic properties of the two plasma membrane domains in epithelial cells. Secondly and unexpectedly, the SAC, which we identified both in vivo and in WIF-B cells as an intermediate in the basolateral-to-apical transcytosis of apical proteins, is not involved in apical endocytosis. We know this because the plasma membrane proteins found in vacuoles did not traverse the SAC. Furthermore, PI 3-kinase inhibitors have little or no effect on SAC's basolateral-to-apical function or morphology. Thus, the results of this study support our earlier hypothesis that apical plasma membrane proteins present in SAC are delivered exclusively to the apical membrane even though other SAC cargo (i.e., endolyn-78) can be sorted off to other destinations (Ihrke et al., 1998). Finally, since our results show that the life cycles of selected resident apical membrane proteins include delivery to lysosomes, we propose that degradation in this organelle is the normal fate of such membrane proteins.

Mammalian cells, including hepatocytes, encode at least three different classes of PI 3-kinase isoforms (for review, see Carpenter and Cantley, 1996; Shepherd et al., 1996; Toker and Cantley, 1997; Vanhaesebroeck et al., 1997). Class I includes the catalytic 110-kD subunits complexed with other accessory proteins. Class II PI 3-kinases include higher molecular weight kinases that contain C2 domains, and class III kinases share the highest sequence similarity with the sole isoform identified in yeast, Vps34p. All mammalian PI 3-kinase isoforms are sensitive to wortmannin and LY294002 at the concentrations used in the studies described here and elsewhere (Shepherd et al., 1996; Vanhaesebroeck et al., 1997; Corvera and Czech, 1998). This has led to ambiguity in distinguishing the roles that specific PI 3-kinases play in membrane transport.

Despite the present uncertainties in a pharmacological approach, a picture emerging from studies in multiple cell types identifies the lysosomal system as the principal target of PI 3-kinase regulation. Yeast strains deficient in the sole PI 3-kinase isoform, Vps34p, missort lysosomal hydrolases from the trans-Golgi network to the extracellular medium (Herman and Emr, 1990; Herman et al., 1992). Recently, a mammalian homologue of this enzyme has been implicated in both transferrin recycling and delivery of activated PDGF receptor to lysosomes, pointing to its role in endosomal sorting/retrieval (Siddhanta et al., 1998). However, others have reported that a p110p85 isoform (Class I), traditionally viewed as a mediator of signal transduction (Carpenter and Cantley, 1996; Toker et al., 1997; Vanhaesebroeck et al., 1997), also influences transferrin recycling and PDGF receptor delivery to lysosomes (Siddhanta et al., 1998; Joly et al., 1994). Equally intriguing is the report that Vps34p, and by inference membrane transport, may play a role in mitogenesis (Siddhanta et al., 1998). Clearly, we need specific probes to dissect these complex and interacting pathways.

Our understanding of how PI 3-kinases regulate vesicle trafficking in mammalian cells is also in its infancy. Several proteins with roles in membrane transport have been shown to bind PI 3-kinase catalytic products (phosphatidylinositol 3 phosphate (PtdIns(3)P), PtdIns(3,4)P₂, and PtdIns(3,4,5)P₃) in vitro including dynamin, the clathrin adaptor protein, AP2, GRP1, cytohesin-1, EEA1 and ARNO (De Camilli et al., 1996; Klarlund et al., 1997; Patki et al., 1997; Corvera and Czech, 1998). One possibility is that the inhibition of the formation of specific lipid species by wortmannin in turn inhibits the recruitment of these factors to cellular membranes where they function in membrane transport. In hepatocytes, the roles these proteins play (if any) in apical membrane internalization have not been identified. Alternatively, the lipids themselves may contribute to alterations in membrane physical properties that allow membrane transport to occur (e.g., fluidity, fusibility). Further experimentation is clearly needed to discriminate between these possibilities.

Our finding that PI 3-kinase inhibitors differentially affect the behaviors of apical and basolateral membrane proteins in both WIF-B and MDCK cells is consistent with other pharmacological studies of endocytosis in polarized epithelial cells. For example, in MDCK cells, addition of calmodulin antagonists (W7 and trifluoperazine) or phorbol esters activated apical endocytosis of ricin, while no changes were seen at the basolateral domain (Llorente et al., 1996; Shurety et al., 1996). In Caco2 and MDCK cells, addition of cytochalasin D inhibited apical uptake of fluid phase markers and no changes were observed from the basolateral domain (Gottlieb et al., 1993; Jackman et al., 1994; Shurety et al., 1996). These differences, as well as those observed with PI 3-kinase inhibitors, may reflect the specialized functions and/or environments of each membrane domain.

Differential effects of wortmannin administration were also seen among apical plasma membrane proteins. APN and 5′NT (as reported here) and dipeptidyl peptidase IV and polymeric IgA receptor (Tuma, P.L., data not shown) all accumulated in lysosomal vacuoles in treated cells, whereas HA4 and syntaxin 3 maintained their apical distributions. No overt sequence or structural motifs are apparent that group the proteins in the two classes, nor are the half-lives of the proteins significantly different. Differences have been determined in the kinetics of delivery of newly synthesized proteins to the apical surface (Bartles et al., 1987; Schell et al., 1992). In particular, HA4 was found to be the slowest molecule with only 15–20% detected at the apical membrane after 2.5 h. Whether slower delivery also implies slower rates of internalization from the apical membrane and, by extension, decreased vacuolar accumulation, is not known. Further examination of these molecules in terms of lipid binding properties and specific internalization mechanisms (e.g., vesicle types, signals) are required before we understand the reasons for the selective regulation by PI 3-kinases.

Although wortmannin and LY294002 have been shown to impair transcytosis of dIgA and ricin in MDCK cells (Cardone and Mostov, 1995; Hansen et al., 1995), the effect of these agents on cellular morphology or the distributions of apical membrane proteins had not been examined before this report. Thus, our finding that MDCK cells mirror WIF-B cells in the regulation by PI 3-kinases leads us to predict that many of the apical regulatory mechanisms described for other polarized epithelial cells may be operating at the hepatocyte canalicular domain.

Based on the results of this study, we propose that apical membrane proteins are degraded in lysosomes as part of their normal life cycle. While such a fate might have been expected, this study is the first to document accumulation of resident apical membrane proteins in a lysosomal compartment of polarized hepatic cells. These results are consistent with limited examples from polarized intestinal cells (Caco-2; Matter et al., 1990). Our proposal raises several questions. For example, why are the levels of apical proteins in the vacuoles so low? From fluorescent micrographs, we estimate that <5% of the apical proteins are found there after 120 min treatment with wortmannin. The subtlety of the effect is probably not due to degradation of a portion of the population since addition of lysosomal proteinase inhibitors, such as leupeptin, together with longer exposure of cells to wortmannin, did not significantly increase the signal. Rather, we favor an explanation that takes into account the apical proteins' relatively long half-lives, which are between 3 and 6 d in vivo (Scott and Hubbard, 1992). That is, if such long-lived molecules remain at the apical membrane until they are retrieved for degradation, only 1–3% of the total APN or dipeptidyl peptidase IV would be sorted to lysosomes in 120 min, a value consistent with our estimations. Despite the good agreement, we cannot rule out the formal possibility that wortmannin induced endocytosis at the apical domain. Interestingly, this model assumes that the apical proteins do not recycle through an intracellular compartment(s) as part of their normal life cycle. Alternatively, the recycling compartment is not sensitive to wortmannin.

Another question raised by our results is why there is any expression of apical proteins if the vacuoles are degradative. The ability to detect apical proteins in the vacuoles suggests that they have decreased degradative potential, consistent with their identification in other cells as prelysosomal intermediates. In treated WIF-B cells, cathepsin D (an acid hydrolase) staining was not observed in vacuoles, but in smaller, punctate structures, implying that mature lysosomes are not altered by wortmannin (Tuma, P.L., data not shown). Reports from NRK and human melanoma (Mel JuSo) cells have documented that dense core (and presumably end-stage) lysosomes were not changed by wortmannin with regard to their morphology (Brown et al., 1995; Reaves et al., 1996; Spiro et al., 1996; Fernandez-Borja et al., 1999) or their ability to receive internalized proteins and lipid dyes (Reaves et al., 1996; Shpetner et al., 1996; Fernandez-Borja et al., 1999). In addition, previous studies have shown that acid hydrolases are missorted to the constitutive secretory pathway in wortmannin-treated cells (Brown et al., 1995; Davidson, 1995; Martys et al., 1996). Thus, similar missorting may occur in hepatic cells such that the vacuoles have decreased hydrolytic activity.

A final puzzle focuses on the vacuoles themselves. If entry of apical membrane into this prelysosomal compartment continues but subsequent events are blocked by PI 3-kinase inhibition, resulting in an enlarged vacuole, what is the “normal” mechanism for maintaining the steady state dimensions of the compartment in untreated cells? A tempting hypothesis is that membrane microdomains invaginate into the luminal space and pinch off to form multivesicular endosomes/lysosomes. This scenario would accomplish both degradation of membrane components and maintenance of compartment size. However, there is limited experimental evidence for such an idea (Haigler et al., 1979; van Deurs et al., 1993; Futter et al., 1996).

Our recent studies of plasma membrane biogenesis have focused on the “indirect” route taken by newly synthesized apical membrane proteins and identified SAC as a subapical compartment in that route (Barr and Hubbard, 1993; Barr et al., 1995; Ihrke et al., 1998). It is important to emphasize that transcytosing apical proteins moved through the SAC and were successfully delivered to the apical surface in wortmannin-treated cells. Ultrastructural analysis confirmed these results and further revealed that the morphology, placement, and number of the subapical organelles were not changed by wortmannin. Thus, the function of the SAC as an intermediate in the transcytotic pathway was not changed. Most importantly, in the presence of wortmannin, the resident apical proteins did not travel through the SAC on their way to lysosomal vacuoles. This indicates that the SAC is not an intermediate of the apical endocytic pathway and is consistent with work in hepatocytes that biochemically identified the SAC and found it contained no recycling resident populations of apical membrane proteins (Barr and Hubbard, 1993; Barr et al., 1995). Previously, we proposed that the SAC is a “one-way” sorting station to the apical domain and the results presented here continue to support our hypothesis.

What is the entry point for those apical membrane residents that were subsequently found in the lysosomal vacuoles? Endosomal structures have been identified in other polarized cells that receive proteins internalized from the apical surface (Bomsel et al., 1989; Hughson and Hopkins, 1990; Apodaca et al., 1994; Barroso and Sztul, 1994). Morphologically, these compartments resemble the SAC in hepatic cells (Parton et al., 1989; Apodaca et al., 1994; Knight et al., 1995; Odorizzi et al., 1996). Since we are limited in our ability to access the apical surface, we have been unable to directly test whether the hepatic SAC acts similarly. However, there are recent reports of a subapically located structure in HepG2 cells receiving fluorescently labeled glycosphingolipids (C6-NBD-glucosylceramide and C6-NBD-sphingomyelin) that have been first accumulated at the apical surface (van IJzendoorn et al., 1997; van IJzendoorn and Hoekstra, 1998). Whether this compartment is identical to the SAC or if it receives apically internalized proteins is not currently known. Future experiments are needed to distinguish among the various possibilities and to identify the putative intracellular compartment that receives incoming cargo from the apical membrane.

We thank W. Dunn, P. Nissley, P. Luzio, and G. Ojakian for generously providing antibodies. We also thank the members of the Hubbard laboratory, Carolyn Machamer, and Anne Kenworthy for critically reading the manuscript and for many helpful comments.

This work was supported by the National Institutes of Health grants GM29185 and DK44375 awarded to A.L. Hubbard and fellowship DK09620 awarded to P.L. Tuma.

5′NT

5′nucleotidase

APN

aminopeptidase N

ASGP-R

asialoglycoprotein receptor

LGP-120

120-kD lysosomal glycoprotein

M6P-R

mannose 6-phosphate receptor

PI

phosphoinositide

SAC

subapical compartment