INTRODUCTION

The elucidation of the molecular mechanisms how plasma membrane and secretory proteins are sorted in polarized epithelial cells is one of the key challenges in cell biology (1, 2). The study of the transport of viral and cellular membrane proteins to the surface of MDCK¹ and other polarized epithelial cells grown in culture has revealed that the signals for basolateral disposal reside in the cytoplasmic domain, while the structural determinants for apical transport are located in the ectodomain of transmembrane proteins (3). In proteins attached to the membrane via a GPI-anchor the glycolipid tail has been shown to function as an apical sorting signal in the kidney-derived MDCK cells (4, 5). It has been suggested that the sorting of these proteins involves their clustering in the TGN into glycosphingolipid (GSL)-enriched microdomains or rafts (6). These rafts can be isolated from MDCK cells by virtue of their low density and their insolubility in Triton or Chaps extracts in the cold (7, 8). They have been shown to be enriched in the 21-kDa transmembrane protein VIP21/caveolin and to ultrastructurally resemble caveolae (9, 10). On the basis of these results, it has been proposed that lipid and protein inclusion into caveolae and apically targeted vesicles occurs by a similar process (9, 10).

In contrast to the targeting of GPI-anchored proteins, we have very limited information on the molecular mechanisms of apical secretory protein sorting in epithelial cells. It has been shown recently that the carbohydrate moieties of secretory glycoproteins are crucial determinants for their transport to the apical surface in MDCK cells and that the mannose-rich core carbohydrate structure is sufficient for apical delivery (11, 12, 13, 14, 15). On the basis of these results it has been proposed that the interaction with a lectin-like protein represents a key feature in the apical targeting of secretory proteins (12, 15, 16). Given the pivotal role of glycosphingolipids in the correct routing of GPI-anchored proteins to the apical plasma membrane, it has further been suggested that secretory proteins are sorted by a receptor that interacts with glycosphingolipids (1). Alternatively, apical secretory proteins could directly interact with GPI-anchored proteins on their way to the cell surface. In both cases, GPI-anchored proteins and secretory proteins would exploit similar molecular mechanisms to reach the apical cell surface.

There are two independent approaches to test this hypothesis. First, it can be analyzed if a secretory glycoprotein that has been shown to be sorted into the apical exocytic pathway in MDCK cells is included into glycolipid enriched rafts on its way to the cell surface. Second, it can be studied whether in cells in which the sorting of GPI-anchored proteins to the apical surface is perturbated, the targeting of an apical secretory protein is concomitantly affected.

Here we report a study using both approaches to characterize the sorting of gp80 (clusterin, apolipoprotein J) to the apical cell surface in polarized epithelial cells. The biogenesis and apical secretion of this glycoprotein has been well characterized in the kidney-derived MDCK cells as well as in other cell lines (11, 17, 18, 19). The protein is synthesized as a single chain precursor protein of 68 kDa in its high mannose form and of 80 kDa in its terminally glycosylated form. The protein is cleaved intracellularly into subunits of 35 and 45 kDa that are secreted as a disulfide-linked heterodimeric complex. We show that in contrast to GPI-anchored proteins, newly synthesized gp80 is not associated with detergent-insoluble complexes in MDCK cells. Furthermore in FRT cells that sort GPI-anchored proteins preferentially to the basolateral surface (10, 20), the gp80 glycoprotein is secreted at the apical plasma membrane domain. Together these results suggest that this glycoprotein is routed to the apical cell surface independently of GPI-anchored proteins and glycosphingolipids.

EXPERIMENTAL PROCEDURES

FRT (kindly provided by Dr. L. Nitsch, University of Naples, Italy) and MDCK (ATCC CCL 34) cells were grown on plastic dishes or polycarbonate filters as described previously (10, 11). Cells were seeded on filters at a density of 1.5 × 10⁶ cells/filter and used for experiments after 3 (MDCK) or 4 (FRT) days at a density of 3.3 ± 0.3 × 10⁶ cells/filter. The transepithelial electrical resistance of filter-grown monolayers was measured in situ, using the Millicell®-ERS voltohmmeter (Millipore, Eschborn, Germany).

For the analysis of the partitioning in phase-separated Triton X-114 solutions of newly synthesized gp80 and GPI-anchored proteins in MDCK cells, cells were labeled with [³⁵S]methionine for 15 min (11) and then incubated for 1 h at 19.5 °C (21). The cells were lysed in Tris-buffered saline containing 1% (v/v) Triton X-114 (22). All media were supplemented with protease inhibitors (antipain, 1 µg/ml; trypsin inhibitor, 10 µg/ml; and benzamidin, 1.75 µg/ml). Proteins in the clarified cell lysates were subjected to temperature-induced phase separation as described by Lisanti et al. (23). Both phases were analyzed for the presence of gp80 by immunoprecipitation with a polyclonal anti-canine gp80 antibody (11). The re-extracted detergent phase was diluted with 100 mM Tris, pH 7.4, 50 mM NaCl, 1 mM EDTA to a final volume of 500 µl and incubated in the presence or absence of phosphatidylinositol-dependent phospholipase C (PI-PLC, 6 units/ml; obtained from Dr. Martin G. Low, Columbia University, New York) for 1 h at 37 °C under continuous mixing. After PI-PLC treatment, 0.5 ml of Tris-buffered saline containing 2% Triton X-114 was added, and the samples were partitioned into aqueous and detergent phases. Both phases were re-extracted three times and then processed for SDS-polyacrylamide gel electrophoresis (SDS-PAGE) under nonreducing conditions (24, 25).

To analyze the association of gp80 with detergent-insoluble complexes, the method by Brown and Rose was used (7). Briefly, MDCK cells were pulse-labeled for 5 min with [³⁵S]methionine and chased for various times before they were lysed for 20 min on ice in an extraction buffer containing 25 mM Hepes, pH 7.5, 0.15 M NaCl, 1% (v/v) Triton X-100 and protease inhibitors. Lysates and insoluble material were collected and centrifuged for 1 min at 14,000 rpm at 4 °C. The supernatants were harvested and the pellets solubilized in 100 µl of buffer containing 50 mM Tris, pH 8.8, 5 mM EDTA, and 1% SDS. The supernatants and the solubilized pellets were adjusted to 0.1% SDS and used for immunoprecipitation.

To analyze the polarity of gp80 secretion from FRT cells, the cells were grown on polycarbonate filters, pulse-labeled for 30 min with [³⁵S]methionine and chased in nonradioactive medium for either 5 or 180 min. Proteins in the lysates of the cells chased for 5 min and in the media of cells chased for 180 min were immunoprecipitated with a polyclonal anti-rat gp80 antibody, kindly provided by Dr. S. Sylvester (Washington State University, Pullman, WA), separated by SDS-PAGE under nonreducing conditions, and detected by fluorography (24, 25). Specific bands on appropriate exposures were quantitated by densitometric scanning with CS-1 Image Documentation System and analyzed with WINCAM software from CYBERTECH (Berlin/Germany). Quantification is based on the data of at least three independent experiments.

RESULTS

To investigate whether gp80 is associated with GPI-anchored proteins at the site of sorting in the TGN, the partitioning of the proteins in phase-separated Triton X-114 solutions was characterized. MDCK-cells were incubated for 1 h at 19.5 °C to accumulate newly synthesized plasma membrane and secretory proteins in the late Golgi and TGN, lysed, and subjected to Triton X-114 extraction according to the scheme shown in Fig. 1A. In this analysis GPI-anchored proteins were recovered from the detergent pellet. They were identified by the PI-PLC-induced transition into a soluble form that is recovered from the aqueous supernatant (Fig. 1B). In contrast to the GPI-anchored proteins, the gp80 glycoprotein was exclusively detected in the aqueous phase in both the high mannose 68-kDa form as well as in the terminally glycosylated 80-kDa form, indicating that gp80 is not associated with GPI-anchored proteins on its way to the apical plasma membrane.

To investigate whether gp80 is associated with GSL-enriched membrane microdomains by a mechanism other than a direct interaction with GPI-anchored proteins, we analyzed if the protein was present in detergent-insoluble complexes at any stage of its biogenesis. Newly synthesized proteins were metabolically labeled and chased at 37 °C for various times up to 1 h. At this time most of the newly synthesized gp80 has been exported from the cell. At each time the cells were lysed and the proteins subjected to a Triton X-100 extraction in the cold, as described by Brown and Rose (7). Proteins were immunoprecipitated from the soluble fraction and the solubilized pellet and processed for SDS-PAGE (Fig. 2). In this analysis gp80 was detected in the high mannose 68-kDa form up to 10 min of chase (lanes 1-4) and then matured into the 80-kDa terminally glycosylated protein (lanes 5-10). The protein was recovered exclusively from the soluble fraction at all times of chase. These results indicate that in contrast to GPI-anchored proteins the gp80 glycoprotein is not included in detergent-insoluble complexes on its way to the apical surface of MDCK cells.

One caveat of this approach is that an association of gp80 with GPI-anchored proteins to weak to be demonstrated by noncovalent co-association would have escaped our attention. To exclude this possibility we also studied the intracellular transport of newly synthesized gp80 in FRT cells. These cells cluster glycosphingolipids into Triton-insoluble complexes, but fail to include GPI-anchored proteins (10, 20). This defect, which correlates with the lack of VIP21/caveolin, requires that these cells use a fundamentally different mechanism to transport GPI-anchored proteins to the cell surface, which results in the preferential targeting to the basolateral plasma membrane domain. No apically secreted proteins have yet been characterized in FRT cells. If there was any association of gp80 with GPI-ancored proteins that was crucial for its targeting, gp80 should in these cells be secreted predominantly at the basolateral cell surface. As revealed by our analysis, FRT cells endogenously express the rat homologue of the gp80 glycoprotein and process it in a way similar to what has been observed in MDCK and other cells (11, 26). The protein is synthesized as a single chain precursor of approximate 65 kDa in its high mannose form. Upon terminal glycosylation, the precursor is cleaved intracellularly into two subunits with an apparent M_r of approximately 45,000 and 35,000. The two subunits are secreted from FRT cells as a disulfide-linked complex. The analysis of the polarity of gp80 secretion in filter-grown FRT cells revealed that 85% of the newly synthesized gp80 is secreted apically, and only a minor component (15%) is released at the basolateral cell surface (Fig. 3). These results suggest that in FRT cells the gp80 glycoprotein is routed by a mechanism distinct from that used to transport independent GPI-anchored proteins.

DISCUSSION

In MDCK and other epithelial cell lines, GPI-anchored proteins are located predominantly in the apical plasma membrane domain (5, 27). To date only two exceptions are known: FRT cells, which preferentially segregate GPI-anchored proteins into the basolateral plasma membrane domain, and MDCK-ConA cells, which miss-sort those proteins to both cell surface domains (10, 20). In MDCK cells, GPI-anchored proteins become insoluble in nonionic detergent as they pass through the Golgi complex (7). It has been suggested that this insolubility reflects the clustering of these proteins together with apically destined glycosphingolipids in specialized, caveolin-containing membrane microdomains in the TGN (9, 28). It was further proposed that this process constitutes a key event in the sorting of apical proteins (6, 27, 29). In support of this hypothesis it was observed that in FRT cells, which do not express caveolin, GPI-anchored proteins fail to assemble into clusters with glycosphingolipids, but remain detergent-soluble during their entire biogenesis and are preferentially segregated into the basolateral plasma membrane domain (10, 20). In mutant MDCK cells resistant to the lectin concanavalin A, GPI-anchored proteins are miss-sorted to both surfaces, although a fusion GPI-anchored protein, gD1-DAF, was found together with glycosphingolipids in Triton-insoluble complexes during transport to the cell surface (10). These observations indicate that apart from clustering into glycolipid-enriched membrane microdomains additional factors are required for the sorting of GPI-anchored proteins to the apical plasma membrane domain (10).

In order to gain insight into the process of secretory protein sorting, we analyzed if apically secreted proteins are sorted by mechanisms linked to the targeting of GPI-anchored proteins. To this end we characterized the transport of the apical secretory glycoprotein gp80 in MDCK cells and in cells that fail to sort GPI proteins to the apical cell surface. We show that, in contrast to GPI-anchored proteins, gp80 is not associated with detergent-insoluble complexes in MDCK cells. Furthermore, the sorting of gp80 to the apical cell surface is not impaired in FRT cells. In a previous study investigating the role of the carbohydrate moieties in apical sorting, we have shown that gp80 is faithfully targeted to the apical cell surface in MDCK-ConA cells (15). Our results therefore imply that in these three cell lines the sorting of the gp80 glycoprotein to the apical cell surface occurs independently of the sorting of GPI-anchored proteins and does not involve an association with GSL-enriched membrane microdomains. This has also been suggested by a study of Mays et al. (30) who characterized the mechanisms of generating Na/K-ATPase polarity in transfected MDCK cells. They noted that the correct sorting of an 84-kDa apical secretory protein, presumably gp80, was maintained in the presence of fumonisin B₁, a fungal metabolite that inhibits sphingolipid biosynthesis. In order to generalize our findings other well characterized apically secreted proteins need to be analyzed. There are three candidate proteins: a 20-kDa osteopontin-derived protein, erythropoietin, and human corticosteroid binding globulin, all of which have been shown to be released apically in MDCK cells (31, 32, 33). Irrespective of these investigations, however, the present study demonstrating that at least one apically secreted protein is targeted independently of GPI-anchored proteins, and GSL suggests that there is more than one mechanism for the sorting of apical proteins.

It is interesting to note that also apical transmembrane proteins need not necessarily be associated with detergent-insoluble complexes on their way to the cell surface. While influenza virus hemagglutinin, sucrase-isomaltase, aminopeptidase N and A, and dipepdidyl peptidase IV were shown to be partially insoluble in Triton X-100 at low temperature, lactase-phlorizin hydrolase and a number of other brush border transmembrane proteins were found to be fully soluble in detergent (34, 35, 36). Whether this reflects differences in the routing of transmembrane proteins that are specific for the cell type or for the protein remains to be elucidated by further investigations. These results suggest, however, that in epithelial cells distinct mechanisms exist to sort GPI-anchored proteins and transmembrane proteins to the apical surface. It now needs to be investigated whether those apical membrane proteins that are sorted intracellularly and secretory proteins use both the same mechanisms and the same type of transport vesicles to reach the cell surface. Studying the transport of membrane and secretory proteins in MDCK cells, Boll et al. (37) proposed, on the basis of a differential sensitivity toward microtubule disrupting drugs, that distinct pathways exist for the transport of membrane and secretory proteins to the basolateral cell surface. By analogy it could well be that also in the apical exocytic pathway transmembrane and secretory proteins are routed by distinct carrier vesicles and that GPI-anchored proteins use a third class of vesicles to reach the apical plasma membrane domain. Therefore the picture that emerges from all these studies is that in polarized epithelial cells there is not a basolateral and an apical pathway, but that multiple sorting mechanisms and routes are used for different classes of proteins within a single cell type.