Polarity of endogenous and exogenous glycosyl-phosphatidylinositol-anchored membrane proteins in madin-darby canine kidney cells

In polarized epithelial cells, the apical and basolateral membranes are composed of distinct proteins and lipids (for review, see Simons and Fuller, 1985). Segregation and targeting of proteins into the appropriate compartment or membrane domain provides the basis for cellular polarity, and the establishment and maintenance of polarity is fundamental to the normal development and function of epithelial systems. Differences in membrane composition are maintained by the presence of intercellular tight junctions or zonulae occludentes, which restrict the lateral mobility of membrane components and provide a barrier to intercellular diffusion of ions and macromolecules. In addition, in adherent cells, some apical membrane proteins can be polarized even in the absence of cell-cell contact (Vega-Salas et al. 1987; Ojakian and Schwimmer, 1989). Also, associations with cytoskeletal elements have been suggested to be important in the polarized distribution of basolateral proteins (Nelson, 1989) and for the anchoring of both apical and basolateral proteins in the correct plasma membrane domain (Salas et al. 1988). However, the signal(s) that sort and direct newly synthesized membrane proteins from their common sites of synthesis and processing to their final polar destination remain unclear. Both the cytoplasmic and ectoplasmic portions of membrane proteins have been suggested to contain sorting signals (Compton et al. 1989; McQueen et al. 1986; Mostov et al. 1986; Roth et al. 1987).

Several studies have used the Madin-Darby canine kidney (MDCK) cell line to address the question of sorting and targeting of membrane proteins and glycolipids in epithelial cells (Caplan et al. 1986; McQueen et al. 1986; Nelson and Hammerton, 1989; Rodriguez-Boulan and Sabatini, 1978; van Meer et al. 1987). These cells form monolayers with functional tight junctions and polarized distribution of plasma membranes and lipids, and when grown on permeable supports they form electrically tight monolayers that provide experimental access to both the apical and basolateral sides (Simons and Fuller, 1985). We have used this system to study the polarity of a class of ‘integral’ membrane proteins that are anchored to the membrane not by a hydrophobic amino acid sequence, but by a specialized glycophosyl-phatidylinositol (GPI) tail, which is added to the protein co-translationally after cleavage of a transmembrane polypeptide (for review, see Ferguson and Williams, 1988; Low and Saltiel, 1988). This specialized anchor is of interest in the study of epithelial polarity because, as the polypeptide is exclusively ectoplasmic, any targeting information must be contained in either the ectoplasmic domain or the GPI anchor.

It has been shown that GPI-linked proteins that are endogenous to this polarized epithelial cell type are directed to the apical surface (Lisanti et al. 1988). In addition, a chimeric molecule composed of the ectoplasmic domain of VSV-G and a small portion of the Thy-1 protein along with its GPI anchor was targeted to the apical plasma membrane of transfected MOCK cells (Brown et al. 1989). However, it was not known whether normal nonmutant GPI-linked proteins of non-polarized cells contain the molecular features that would be recognized by polarized cells as an apical targeting signal.

To explore this issue, we transfected MOCK cells with the cDNA encoding mouse Thy-1, a GPI-linked glycoprotein (Low and Kincade, 1985) that is found in non-epithelial mouse thymocytes and neurons. We then utilized selective surface iodination, phosphatidylinositolspecific phospholipase C digestion (PI-PLC), and immunocytochemistry to show that the foreign GPI-linked protein is targeted to the apical plasma membrane. These results suggest that there is a common sorting mechanism for GPI-anchored proteins in epithelial cells.

N,N-bis(2-hydroxyethyl)-2-aminoethane-sulfonic (Bes) was obtained from Calbiochem-Behring Corporation (San Diego, CA), restriction enzymes and modifying enzymes were obtained from Boehringer-Mannheim Biochemicals (Indianapolis, IN), Na¹²⁵I from Amersham Corporation (Arlington Heights, H), Geneticin (G418-sulfate) from Gibco-BRL Ltd Life Technologies (Grand Island, New York), and inositol phosphate-specific phospholipase C from Funakoshi Pharmaceutical Co., Ltd (Tokyo, Japan). All other chemical reagents were obtained from Sigma Chemical Company (St Louis, MO).

The cDNA of Thy-1 was excised from a pGEM-1 vector (Promega Corporation, Madison, WI) by HindHI and BamHI digestion (Fasel et al. 1989; Hedrick et al. 1984), filled in and inserted into a pSV2 vector (Mulligan and Berg, 1980). pSV2-neomycin was used as the selectable marker in the co-transfection experiments (Southern and Berg, 1982).

MDCK cells were plated at a density of 200 000 cells/10 cm dish and incubated overnight. Cells were co-transfected by the method of Chen and Okayama (1987) with 22.5 μg of pSV2-Thy-l and 3 μg of pSV2-neo DNA. Briefly, the transfection was carried out for 17.5h at 37°C under 3% CO₂. After transfection, cells were washed with media, re-fed overnight, trypsinized and split 1:5. Selection was begun the next day using 100 pg ml^-1 of G418. Medium was changed every other day until colony growth was observed. Six G418-resistant colonies were observed, picked by trypsinization with cloning rings, expanded and screened for expression of Thy-1.

MDCK cells (clone I; Fuller et al. 1984) were obtained from Dr Kai Simons at the EMBL and grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (Inotech A.G., Wahlen, Switzerland), 2mM glutamine and penicillin/ streptomycin (Gibco) in a 95% air/5% CO₂ incubator. For polarity experiments, 24 mm polycarbonate membrane filter units with 0.4 μm pores (Costar, Cambridge, MA) were coated with type IV collagen (Sigma, placental type IV) dissolved in 0.017 M acetic acid, air dried, crosslinked with NH4CI vapor, and sterilized with 2.5% glutaraldehyde. After washing with sterile water, filters were pre-incubated overnight with media before plating cells. Cells were plated at a density of 1×10^s cells/filter and used at 6 days growth at which time resistance had reached 5000–10 000 Ohm cm².

Filters were washed three times with ice-cold PBS containing 10 mM glucose and kept on ice. One ml of PBS/lmM CaCl₂/ glucose was added to the side of the filter (apical or basal) that was not being iodinated and 1 ml of enzyme mixture (10 mM glucose, 10 pg ml^-1 lactoperoxidase, and lmCi Na^12eI in PBS) was added to the opposite side of the filter. Filters were incubated for 30 min on ice to allow penetration of reagents and then 20 pl of glucose oxidase (diluted 1:100 in water) was added to the enzyme mixture to start the iodination reaction. The reaction was allowed to proceed for 10 min at 4 °C and was stopped by the addition of 20 pl of 20 % NaN₃. To control for tightness of the epithelium, samples of the apical and basal solutions were taken and counted. There was usually a 1000-fold gradient when monolayers were tight. Filters were washed six times with PBS and then solubilized in buffer containing either Tris-buffered saline, pH 7.4, 1 % Triton X-114 (TX-114), ImM EDTA (PI-PLC digestion), or 50 mM Tris-HCl, pH 7.5, 5mM EDTA, 150 mM NaCl, 1% Triton X-100 (immunoprecipitation) and scraped off the filters; 1 μgml^-1 leu-peptin and antipain, and 0.5 mM PMSF were added to all buffers.

After iodination and solubilization in TX-114-containing buffer (see above), the lysed cells were incubated on ice for 30 min, vortexed every 10 min, and then spun for 3 min at 4°C. Phase separation was carried out at 30°C (Bordier, 1981) and the aqueous phase was discarded. The pellet was washed by two cycles of phase separation and suspended in buffer containing 0.25 M methylmannopyranoside, 0.1M Tris-HCl, pH 7.4, mM EDTA and 20 μl PI-PLC was added. Samples were incubated for 30 min at 30 °C, then an additional 10 μl PI-PLC was added and incubation was continued for 30 min. Detergent and aqueous phases were separated by centrifugation and were washed by two cycles of phase separation. Bovine serum albumin (BSA; 10 pg) was added to each sample as carrier and samples were precipitated with 10% trichloroacetic acid (TCA) before dissolving in sample buffer and loading on a SDS-polyacrylamide gel (Laemmli, 1970).

Monolayers were iodinated as described above and then rinsed with PBS which contained 1 mM-EDTA and 1 % BSA. Monolayers were then incubated for 30 min at 37 °C in 1ml of the same solution with 20 pl PI-PLC. An additional 10 pl of PI-PLC was added and the incubation continued for 30 min. The apical and basolateral media were collected and the cells were solubilized with 0.5 ml lysis buffer (above).

After solubilization in 1 ml of immunoprecipitation buffer (see above), samples were spun briefly to remove nuclei and then preabsorbed with 10 pl of normal rabbit serum for 20 min at 4 °C. Protein A-Sepharose was added and samples were incubated 60 min at room temperature. Protein A—Sepharose was removed and the supernatant was immunoprecipitated overnight at 4 °C with 10 μl of rabbit anti-Thy-1 serum, R191 and R287, a generous gift from Dr Claude Bron (Institute of Biochemistry, University of Lausanne, Switzerland). Protein A-Sepharose was added, incubated for 60 mm at room temperature, spun, and washed three times with immunoprecipitation buffer. Samples were eluted from the beads by boiling in sample buffer.

Collagen-coated filters were seeded as described for iodination. After 6 days of culture the cells were fixed for 3h with a paraformaldehyde-lysine-periodate fixative (McLean and Nakane, 1974). After fixation, filters were washed and cells were scraped up with a rubber policeman, embedded in 6 % gelatin, and frozen in Freon-22 cooled to liquid nitrogen temperature. Sections (1 μm) were cut and processed for immunofluorescence as previously described (Wilson et al. 1987). For immunofluorescence of cells on coverslips, cells were plated on coverslips that had been coated with type IV collagen. Cells were fixed at confluence in paraformaldehyde-lysine-periodate fixative, washed, and incu bated with antibodies according to the same protocol as was used for sections.

After transfections with pSV2-Thy-l and pSV2-neo, the transfected cells continued to grow in the presence of G418. They grew as a monolayer on both solid and permeable supports and were morphologically similar to MDCK cells described in other studies (Fuller et al. 1984). When grown on collagen-coated filters they established high transepithelial electrical resistance (∼5000 Ohm cm²) after seeding.

To determine the surface distribution of endogenous GPI-anchored proteins, transfected MDCK cells grown on permeable supports were iodinated on either the apical or the basolateral side. The epithelial monolayer remained tight for the duration of the iodination as determined by counting the ¹²⁶I activity in samples of apical and basal media collected after termination of the reaction. Apical surface iodination resulted in a twofold higher specific activity than basal iodination, a différence that may be explained by decreased access of the iodination enzymes to the basolateral membrane due to the filter support. Polyacrylamide gel electrophoresis of the iodinated proteins showed the unique protein profiles of the apical and basolateral membranes (Fig. 1, lanes 1 and 2). Digestion of the proteins in the TX-114 detergent pellet with PI-PLC followed by TX-114 phase partitioning showed a striking difference in distribution of GPI-anchored proteins. After PI-PLC digestion, several proteins from the apical membrane were released into the aqueous phase during TX-114 phase separation, indicating cleavage of a phospholipid anchor (Fig. 1, lanes 7 and 8) (Etges et al. 1986). These included seven major bands with apparent molecular weights of 108K, 80K, 72K, 62K, 55K, 36K and 33K (K=10³M_r) (Fig. 2, lane 8). In addition, longer exposure of the film showed many more faint bands, suggesting that a great many minor proteins possess this linkage (not shown). The molecular weights of the major bands were similar to those reported previously (Lisanti et al. 1988), with the exception of the 62K band. This difference may be due to different protein labeling reagents, as the biotinylation reagent used by Lisanti et al. (1988) and the iodination procedure used here would label different protein sites. This disparity may also be due to MDCK cell strain differences (Valentich, 1981). There was no major band at 24K, the molecular weight of Thy-1, indicating that in these transfected cells Thy-1 was a minor component of the plasma membrane. Very little release of labeled protein was observed when the detergent pellets were incubated in the absence of PI-PLC (Fig. 1, lanes 3 and 4). In contrast to the apical membrane, PI-PLC digestion of TX-114 detergent pellets from basolaterally labeled membrane proteins resulted in no detectable PI-PLC-mediated release of proteins into the aqueous phase (Fig. 1, lanes 9 and 10).

To see what proportion of cells were expressing Thy-1, immunofluorescence of non-permeabilized cells on coverslips was performed. Virtually all cells grown on collagen-coated coverslips were expressing Thy-1. Levels of Thy-1 were variable throughout the epithelium, with very brightly staining cells adjacent to cells with very faint staining (Fig. 2). This heterogeneity of expression of transfected proteins has been described by others (Roman and Garoff, 1986). The fluorescence pattern was punctate on the apical cell surface, a distribution similar to that of other apically located proteins in MDCK cells (Gottlieb et al. 1986; Vega-Salas et al. 1987). In addition, there were larger spots of very bright fluorescence distributed randomly on the cell surface. Incubation of untransfected cells with anti-Thy-1 or transfected cells with non-immune serum resulted in no detectable staining (not shown).

To determine if the steady-state distribution of the exogenous GPI-linked protein Thy-1 was polarized in these cells, selective surface iodination followed by immunoprecipitation with rabbit anti-Thy 1 antibodies and SDS-PAGE was performed. Immunoprecipitation from apically iodinated samples resulted in a band of 24K apparent molecular weight, corresponding to the molecular weight of Thy-1 (Fig. 3, lane 3). However, no Thy-1 was immunoprecipitated from samples that were iodinated on the basolateral side of the epithelium (Fig. 3, lane 4), or from untransfected cells iodinated on either surface (not shown), Thy-1 expression was dependent upon the growth substratum, and surface expression was stimulated on collagen-coated filters compared to uncoated filters (Fig. 3, lanes 3–6). Even on collagen-coated filters Thy-1 protein levels remained low and attempts to label Thy-1 metabolically with [^3fiS]methionine were unsuccessful. However, low levels of Thy-1 synthesis may be advantageous for studies on permanently transformed cell lines, as high expression of exogenous GPI-linked proteins may interfere with normal cellular processes. For example, expression of Thy-1 in COS cells and in a mammary cell line was found to perturb cell growth (N. Fasel, unpublished observations).

To establish that the lack of Thy-1 on the basolateral membrane was not due to restricted access of iodination reagents, immunofluorescence of 1 gm frozen sections was carried out. The staining was confined to the apical plasma membrane, and no intracellular or basolateral staining was observed (Fig. 4). The staining of the apical plasma membrane was patchy, likely reflecting the uneven surface staining seen on the coverslips.

We then examined the possibility that processing of Thy-1 in these transfected cells was not normal, specifically that it was not GPI-anchored and retained its cytoplasmic polypeptide tail. Digestion of intact radioiodinated monolayers with PI-PLC followed by immunoprecipitation of Thy-1 from the media and cells resulted in the release of 40 % of the detectable apical Thy-1 into the apical medium (Fig. 5). No Thy-1 was released into the basal medium after basolateral PI-PLC digestion. We can be confident that the lack of release of Thy-1 from the basolateral membrane was not due to a PI-PLC-resistant modification of the GPI linkage, as both immunoprecipitation of basolaterally iodinated monolayers and immunofluorescent staining demonstrated that Thy-1 was not on the basolateral membrane. Lack of total release of apical Thy-1 after PI-PLC digestion does not necessarily indicate that the remaining Thy-1 was not GPI-anchored, as incomplete digestion with PI-PLC has been reported for a variety of GPI-linked proteins, including Thy-1 (Davitz et al. 1986; Low and Finean, 1979; Low and Zilversmit, 1980; Low and Kincade, 1985; Shukla et al. 1980; Thompson et al. 1987).

We have induced expression of mouse Thy-1, an exogenous GPI-linked protein that is normally expressed in thymocytes and neurons, in MDCK. cells and have shown that this glycosyl-phosphatidylinositol-anchored protein is concentrated exclusively in the apical plasma membrane of this polarized cell line. We also confirmed the previous finding (Lisanti et al. 1988) that the steady-state distribution of endogenous GPI-linked proteins is limited to the apical plasma membrane of MDCK cells.

As Thy-1 is a protein that has no transmembrane or cytoplasmic domain, we can begin to address the question of what portion of the molecule contains the signal for targeting to the apical plasma membrane. The basic structure of the GPI anchor is conserved in a variety of GPI-linked proteins (for review, see Low et al. 1986; Low, 1987). Our results indicate that the signal for sorting of these proteins to the apical membrane and/or their stable residence in this domain may be inherent in this specialized linkage, either the lipid-linked glycan or the lipid itself. Alternatively, the signal for sorting of these molecules may reside in the protein or in the protein-linked oligosaccharides. Also, sorting may be conferred by interaction with other membrane proteins.

Does the lipid portion of GPI-anchored proteins determine their distribution? This is unlikely, as the lipid portion of Thy-1 is phosphotidylinositol, and this phospholipid has a non-polarized distribution in MDCK cells (van Meer et al. 1985). However, glycolipids are sorted before insertion into the plasma membrane, are enriched in the apical plasma membrane in MDCK cells (Turner et al. 1985; van Meer et al. 1987) and have been suggested to play a role in the sorting of lipids and proteins (Simons and van Meer, 1988). The glycan moiety of the GPI-anchored proteins has been shown to contain common elements and structural features, and may carry the signal that directs the proteins apically (Howard et al. 1987; Holder and Cross, 1981; Low et al. 1987; Low and Kincade, 1985; Medof et al. 1986; Tse et al. 1985; Brown et al. 1989). On the other hand, there is a degree of microheterogeneity of both content and structure in the various glycans (Ferguson and Williams, 1988; Low, 1987; Low and Saltiel, 1988). Thus, the role of the lipid glycan in sorting is unresolved.

The role of the ectoplasmic domain in the sorting of apically targeted molecules has been examined using truncated and chimeric forms of apically directed proteins. Influenza virus hemaggluttinin is normally targeted to the apical plasma membrane (Gottlieb et al. 1986); however, anchor-minus forms of this protein were secreted either exclusively apically (Roth et al. 1987) or apically and basolaterally (Gonzales et al. 1987). In contrast, chimeric proteins that contain the ectodomain of influenza hemagluttinin, and the transmembrane and cytoplasmic portions of vesicular stomatitis virus G protein (a basolateral protein), were targeted only to the apical plasma membrane (McQueen et al. 1986; Roth et al. 1987) and chimeras with the ectoplasmic domain of VSV G and the cytoplasmic domain of hemagluttinin were targeted to the basolateral domain (Compton et al. 1989). These results suggest that the ectoplasmic domain contains sorting information. On the other hand, a chimera containing the ectoplasmic domain of VSV G with 22 proximal amino acids of Thy-1 and the GPI tail was targeted apically. These results, along with the results reported in this study, suggest that the GPI anchor contains a sorting signal that may dominate other sorting signals.

It is also possible that the sorting and polar distribution or Thy-1 and other GPI-anchored proteins is due to interaction of the lipid or protein moiety with other plasma membrane proteins and the submembrane skeleton. Although the mobility of GPI-anchored molecules in polarized epithelia has not been measured, there is evidence for association of Thy-1 with other membrane components in non-polarized cells derived from fluor escence recovery after photobleaching experiments. These studies have shown that, although a substantial fraction of Thy-1 has a very high mobility, approximately 50% of the Thy-1 in fibroblasts, thymocytes and lymphoma cells is immobile (Ishihara et al. 1987). The GPI-linked protein decay accelerating factor and alkaline phosphatase also have a significant immobile fraction (Noda et al. 1987; Thomas et al. 1987). Additional evidence for interaction of Thy-1 with other membrane and submembrane components was provided by non-ionic detergent extraction of thymocytes and lymphoma cells; Thy-1 and smother GPI-linked protein Qa-2 were resistant to extraction with Triton X-100 (Letarte-Muirhead et al. 1974; Hoessli and Rungger-Brandle, 1985). Also, Thy-1 can be co-purified from lymphoma cells with a band 4.1-like protein (Bourguignon et al. 1986) and it co-caps with actin, myosin and a band 4.1-like protein (Bourguignon et al. 1978, 1986; Bourguignon and Singer, 1977). In epithelial cells, submembrane skeletal proteins may be involved in stabilization of apical components (Mooseker, 1985). In fact, two endogenous membrane proteins in MDCK cells have been shown to be apically polarized even in the absence of tight junctions (Vega-Salas et al. 1987; Ojakian and Schwimmer, 1989). This type of interaction may play a role in the targeting of GPI-anchored proteins.

The physiological significance of the polarized distribution of GPI-linked molecules is unclear. The list of proteins that have this unique lipid attachment is growing rapidly, and many of these proteins have enzymatic activity (Low and Finean, 1978; Low and Zilversmit, 1980; Shukla et al. 1980; Thompson et al. 1987). Alkaline phosphatase is GPI-anchored and is apical in epithelial cells of intestine and kidney (Heidrich et al. 1974; Quaroni and Isselbacher, 1985). Perhaps the polarity of GPI-linked proteins allows rapid removal of these proteins from the apical cell surface.

We thank Dr Claude Bron for his generous gift of anti-Thy-1 antibodies, and Drs Marian Neutra and Karl Matlin for invaluable discussion and comments on the manuscript. We are also grateful to Betty Ann Mclsaac for typing the manuscript.