INTRODUCTION

Enveloped viruses have provided an extremely useful paradigm to understand polarized protein traffic in epithelial cells (1). Viral membrane glycoproteins, even when expressed without other viral constituents, contain information sufficient for selective transport to either the apical or basolateral surface (2-10). Indeed, analyses of both viral and endogenous glycoproteins expressed in epithelial cells, using recombinant DNA approaches, have shown that the information required for polarized transport depends on specific structural determinants contained in the newly synthesized glycoproteins (11); from these analyses, several common themes have emerged.

Specifically, attachment of a glycosylphosphatidylinositol anchor to protein moieties that are otherwise luminal is sufficient for polarized surface delivery (12, 13) that is typically (but not in every case) (14) directed apically. In addition, studies using type I membrane glycoprotein chimeras as well as truncation mutants have suggested that the ectodomains of these membrane proteins possess signals for apical transport (9, 15-18). Most recently, this idea has been extended to suggest that both secretory glycoproteins and the ectodomains of membrane glycoproteins may use N-linked carbohydrates either indirectly (19) or directly (20) for polarized sorting. However, Roth et al. (21) and Green et al. (22) found that complete inhibition of glycosylation did not affect polarized viral glycoprotein expression. On the other hand, information for the basolateral transport of the vesicular stomatitis virus G-protein is contained within its cytoplasmic/transmembrane extension (23), and furthermore, specific motifs in the cytoplasmic tails of several mammalian membrane protein receptors and lysosomal membrane glycoprotein lgp120 are now recognized to be essential for the targeting of these proteins to the basolateral surface (24-28). Of course, it should be noted that most of these studies have employed single-spanning membrane proteins with a type I topology (29), and it has not been excluded that specific basolateral targeting information can also exist in protein luminal domains (9, 23, 30-32).

SV5, which is released by budding from the apical surface of infected epithelial cells (2), possesses two glycoproteins, the hemagglutinin-neuraminidase (HN)¹ and fusion protein. HN is a type II membrane protein with a small N-terminal signal/anchor region and a large C-terminal ectodomain (33, 34). In recent years, chicken muscle pyruvate kinase (PK) has been a favored reporter for protein targeting, as chimeric genes containing specific sorting information can relocate the normally cytosolic PK moiety to the nucleus (35, 36), the membrane of the endoplasmic reticulum (37), or the mitochondria (38). With this in mind, a chimera including the signal/anchor portion of HN was found to redirect PK to the cell surface, suggesting that specific protein targeting information may reside in this region of HN (39). However, it has been unclear if this segment could also account for polarized apical delivery of HN. In this study, we have utilized HN-PK chimeras to examine this question in stably transfected MDCK cells.

EXPERIMENTAL PROCEDURES

Construction of the HN₄₈PK chimera via an in-frame EcoRI linker (5-CGGAATTCC-3, encoding the hydrophilic linker Arg-Asn-Ser) in pSV63 has been described (previously called APK) (39). Intact HN₄₈PK was excised from pSV63 with XhoI and subcloned first into pGEM7Z and then, using HindIII and XbaI sites, into pRC/CMV (Invitrogen). This produced the plasmid pRC/HN₄₈PK, encoding a protein joining HN residues 1-48 to PK residues 17-529 by a 3-amino acid linker (encoding Arg-Asn-Ser), driven by the cytomegalovirus promoter. To construct HN₇₂PK, an oligonucleotide (5-TTGGATCCAAGCTTATGGTTGCAGAAGATGCCCCT-3) containing a HindIII site followed by HN nucleotides 320-340 was used as an upstream primer, and an oligonucleotide (5-GTCGGAATTCCTGCGACAGAGAGAATATT-3) containing HN nucleotides 521-538 followed by an EcoRI linker was used as a downstream primer to amplify the N-terminal fragment 1-72 of the HN gene by polymerase chain reaction as described previously (40). The amplified DNA fragment was isolated after digestion with HindIII and EcoRI, and along with the PK moiety isolated from pRC/HN₄₈PK by digestion with EcoRI and XbaI, a three-way ligation was performed using pRC/CMV vector that had been digested with HindIII and XbaI. Thus, the hybrid protein expressed by the recombinant plasmid pRC/HN₇₂PK encodes HN residues 1-72 joined to PK residues 17-529 by a 3-amino acid linker (encoding Arg-Asn-Ser), driven by the cytomegalovirus promoter. To construct secretory PK, an oligonucleotide (5-ACTGAAGCTTATGGACAGCAAAGGT-3) containing a HindIII site followed by bovine preprolactin nucleotides 68-83 was used as an upstream primer, and an oligonucleotide (5-ATCGGAATTCCGGTGGAGACCACACC-3) containing the sequence of preprolactin nucleotides 145-160 followed by an EcoRI linker was used as a downstream primer to amplify by polymerase chain reaction the preprolactin signal peptide provided in plasmid form (gift of Dr. V. Lingappa, University of California, San Francisco). The amplified DNA fragment was then isolated after digestion with HindIII and EcoRI, and along with the same PK cDNA moiety (isolated from pRC/HN₄₈PK by digestion with EcoRI and XbaI), a three-way ligation was performed using pRC/CMV vector that had been digested with HindIII and XbaI. Thus, the hybrid protein expressed by this recombinant plasmid encodes the N-terminal cleavable prolactin signal peptide residues 1-31 joined to PK residues 17-529 by a 3-amino acid linker (encoding Arg-Asn-Ser), driven by the cytomegalovirus promoter.

MDCK cells (strain II) were transfected with the above recombinant plasmids by the Lipofectin method as described previously (40). Colonies were selected and maintained in the presence of G418 (500 µg/ml; Life Technologies, Inc.) in Dulbecco's modified Eagle's medium supplemented with 10% newborn calf serum. Subclones of transfected MDCK cells were induced with 10 mM sodium butyrate for 16 h (26) and then labeled for 2 h with 100 µCi of [³⁵S]methionine/cysteine in methionine/cysteine-deficient medium. The labeled cells were lysed in 1 ml of TNT lysis buffer (50 mM Tris, 150 mM NaCl, and 1% Triton X-100, pH 7.4) plus 0.1% SDS, and the clarified lysates were incubated overnight at 4 °C with rabbit anti-PK bound to protein A-agarose and analyzed by 8.5% SDS-polyacrylamide gel electrophoresis (SDS-PAGE). At least two colonies expressing the desired protein from each cDNA construct were randomly selected for further analysis.

Immunoprecipitated samples were denatured in 0.5% SDS plus 1% 2-mercaptoethanol at 100 °C for 10 min; cooled; and then incubated for 2.5 h at 37 °C with 10,000 units/ml PNGase F (New England Biolabs Inc., Beverly, MA) in 50 mM sodium phosphate, pH 7.5, supplemented with 1% Nonidet P-40.

Costar Transwell polycarbonate filter units with a pore size of 0.4 µm were used. Cells (2 × 10⁶) were plated and grown for 4 days, and the formation of electrically tight cell monolayers was confirmed by measuring transepithelial resistance with chopstick electrodes (World Precision Instruments, Inc., New Haven, CT). All clones exhibited resistance between 100 and 125 ·cm² prior to their use in experiments. Selective cell-surface biotinylation was performed according to published procedures (41). For measurement of polarized protein distribution at steady state, biotinylated proteins isolated with streptavidin-agarose were resolved by SDS-PAGE and then analyzed by specific immunoblotting.

Filter-grown cells were pulse-labeled with 400 µCi of [³⁵S]methionine/cysteine in methionine/cysteine-deficient medium. After different periods of chase in complete Dulbecco's modified Eagle's medium containing excess methionine and cysteine, the apical and basolateral surfaces were selectively biotinylated as described above. The cells were then lysed in TNT lysis buffer, clarified by centrifugation, and immunoprecipitated with antibodies bound to protein A-Sepharose. Immunoprecipitates were denatured in 1% SDS, diluted 10-fold into TNT lysis buffer, and then re-precipitated after incubation with 10 µl of 4% streptavidin-agarose (Sigma). The radiolabeled biotinylated proteins were then analyzed by SDS-PAGE, and radioactivity was quantified using a PhosphorImager.

For assaying labeled secretion of PK, the transfected MDCK cell lines were either pulse-labeled as described above (see Fig. 4B) or continuously labeled for 8 h with 100 µCi of [³⁵S]methionine/cysteine (see Fig. 4A). Media bathing the labeled cells were collected from either the apical or basolateral chamber, and cellular debris was removed by centrifugation. The clarified media were examined by immunoprecipitation, SDS-PAGE, and fluorography.

[View Larger Version of this Image (56K GIF file)]

Polarized MDCK cells grown on Transwell units were incubated with SV5 at 20 plaque-forming units/cell in Dulbecco's modified Eagle's medium containing 0.5% bovine serum albumin for 1 h at 37 °C. At this time, the cells were washed, and the medium was replaced with fresh Dulbecco's modified Eagle's medium supplemented with 2% newborn calf sera. Infected cells were maintained at 37 °C for 14 h for further analysis.

For immunoblotting, biotinylated samples were analyzed by reducing 8.5% SDS-PAGE. The proteins were transferred to nitrocellulose filters (0.2-µm pore size; Schleicher & Schuell) and incubated for 1 h at room temperature in 5% nonfat dry milk in phosphate-buffered saline and then for 2 h with primary antibodies diluted in the same buffer. Protein-bound antibodies were detected with horseradish peroxidase-conjugated specific secondary antibodies using enhanced chemiluminescence (ECL, Amersham Corp.). Immunoblots were quantified by scanning densitometry. The method was validated by using serial dilutions of protein samples to establish a linear relationship of ECL densities in the range of protein concentrations studied.

Monoclonal antibodies to SV5 HN were kindly provided by Dr. R. E. Randall (42). An antiserum to chicken PK was prepared in rabbits, which received a total of 250 µg of antigen (five × 50 µg). The first intramuscular injection of PK was emulsified in Freund's complete adjuvant, followed by three boosts at weekly intervals emulsified in Freund's incomplete adjuvant. Finally, a single intravenous dose of PK (without adjuvant) was administered 5 days before bleeding.

RESULTS

Previous studies have established that for chimeras that position the normally cytosolic PK moiety in the lumen of the secretory pathway, a major fraction of the expressed chimera undergoes N-linked glycosylation and intracellular transport to the cell surface (39). Since our goal was to attach putative signals for polarized protein targeting to the PK reporter, we first set out to examine the fate of a chimeric protein expressed after ligating a cDNA fragment encoding the N-terminal 31-amino acid signal peptide of preprolactin directly to PK itself ("secretory PK") (Fig. 1). This signal peptide contains the information necessary for normal cleavage by signal peptidase (43, 44), resulting in a threonine residue as the new N terminus of the mature protein. After G418 selection of stably transfected MDCK cells, clones expressing this form of PK were identified by immunoprecipitation and SDS-PAGE analysis.

[View Larger Version of this Image (12K GIF file)]

Unglycosylated secretory PK with an uncleaved signal is predicted to be a polypeptide of 547 amino acids with a molecular mass of ~60 kDa, whereas unglycosylated secretory PK after cleavage of the signal peptide is 517 amino acids with a predicted molecular mass of ~56 kDa. Indeed, when we expressed a leaderless cytosolic form of PK in COS cells, this unglycosylated polypeptide lacking a signal peptide migrated on SDS-PAGE at the predicted size of ~56 kDa (data not shown). For expression in the lumen of the secretory pathway, the PK cDNA sequence contains one potential N-linked glycosylation site (45). With this in mind, we observed that in MDCK cells expressing secretory PK, intracellular immunoreactive forms of PK included different closely migrating species: the majority ran as an ~58-kDa species, whereas a smaller portion ran at ~56 kDa (Fig. 2A, second lane).² To see if the larger form occurred as a consequence of N-linked glycosylation, the secretory PK forms expressed in MDCK cells were treated with PNGase F. In this case, all forms collapsed into a single species of ~56 kDa (Fig. 2A, third lane). Because PNGase F removes N-linked glycans that are added in the lumen of the secretory pathway, these molecular mass data strongly suggest that secretory PK had indeed been translocated into the endoplasmic reticulum lumen and underwent removal of the prolactin signal peptide and that most of the secretory PK received N-glycosylation as has been reported for the membrane-bound form (39). Such N-linked glycosylation has been proposed to have important benefits in favoring protein stability and export through the secretory pathway (46-49).

[View Larger Version of this Image (29K GIF file)]

We therefore examined the stability and export of newly synthesized secretory PK in MDCK cells over a 12-h chase. As expected at the zero chase time, secretory PK was quantitatively intracellular (Fig. 3A). There was intracellular loss of a portion (~50%) of labeled PK during the first 2 h of chase (Fig. 3B) that is not understood; loss was seen for both the 58- and 56-kDa forms (Fig. 3A). Following this, however, between 2 and 12 h of chase, the remaining immunoprecipitable 58-kDa PK was progressively released from the cells to the medium, with t1/2  4 h and with an efficiency of ~90% over this time course (Fig. 3C). As expected, PNGase F treatment converted PK in the medium back to an ~56-kDa species (Fig. 2B). The release of ~58-kDa PK was not due to nonspecific cell leakage or lysis, as the 56-kDa form of PK was retained quantitatively within the MDCK cells (Fig. 3A). Moreover, secretory PK recovered free in the medium was not sedimentable in a 1-h spin at high speed (data not shown). Thus, the data in Figs. 1, 2, 3 indicate that the prolactin signal-PK chimera enters the lumen of the secretory pathway and undergoes N-linked glycosylation. While a portion of the PK protein is apparently degraded intracellularly, the remaining glycosylated 58-kDa form is competent for intracellular transport and behaves as a valid secretory protein marker.

[View Larger Version of this Image (41K GIF file)]

Next, we investigated the polarity of PK secretion from filter-grown MDCK cells. Stable MDCK transfectants expressing secretory PK were continuously labeled for 8 h, and apical and basolateral media were collected and analyzed by immunoprecipitation, SDS-PAGE, and fluorography (Fig. 4A, upper panel). Interestingly, only 20% of secreted PK (~58 kDa) was found on the apical side of transfected cells, with 80% recovered in the basolateral medium (Fig. 4A, lower panel). By contrast, the endogenous MDCK secretory protein gp80/clusterin (50) was released with apical predominance (data not shown). To exclude possible rapid apical secretion of PK followed by re-internalization and transcytosis to the basolateral side, polarized cell monolayers were pulse-labeled with [³⁵S]methionine/cysteine for 5 min, and the kinetics of initial release of 58-kDa PK was followed as a function of chase time (Fig. 4B, left panels). Evidently, from the moment that PK secretion is first detected, the protein is directly delivered with a fixed polarized ratio leading to ~80% in the basolateral medium (Fig. 4B, right panel). Clearly, the glycosylated PK molecule exhibits no sorting signal for apical specific transport in MDCK cells.

We next examined the expression of the HN₄₈PK chimera (Fig. 1), in which the HN N terminus including its signal/anchor is linked to the luminally oriented PK protein (39). Immunoprecipitates with anti-PK serum from cells transfected with pRC/HN₄₈PK revealed two closely migrating species, a major band at ~67 kDa and a slightly less intense band at ~65 kDa (Fig. 5A). When treated with PNGase F to remove N-linked oligosaccharides, the two different species migrated with the same mobility as a nonglycosylated form at ~63 kDa (Fig. 5B), in agreement with previous reports (39). Thus, the luminal domain of HN₄₈PK underwent N-linked glycosylation in MDCK cells, as is the case for secretory PK (Fig. 2) and for HN₄₈PK expression in CV1 cells, where the chimera underwent glycosylation en route to residence at the cell surface (39).

[View Larger Version of this Image (69K GIF file)]

To examine the polarized surface distribution of the HN₄₈PK chimera in filter-grown MDCK cells, we utilized cell-surface biotinylation. As a control, in SV5-infected MDCK cells, ~95% of blottable HN was found on the apical surface (Fig. 6, left). Interestingly, linking the N-terminal region containing the signal/anchor of HN to PK (Fig. 1) caused a change in the targeting of the glycosylated PK reporter (Fig. 5) from a basolateral (Fig. 4) to an apical surface distribution. Specifically, ~75% of HN₄₈PK was now present on the apical surface of transfected cells (Fig. 6, right). The data indicate that the N-terminal region of HN, even when transplanted onto a basolaterally targeted luminal domain, by itself is sufficient for preferential apical distribution, although this steady-state distribution is less completely apical than that found for the full-length HN protein in the setting of expression with other viral components.

[View Larger Version of this Image (32K GIF file)]

To determine whether the apical distribution of HN₄₈PK was accomplished by direct polarized delivery from the trans-Golgi network, we examined the kinetics of first appearance of the newly synthesized protein on the apical and basolateral surfaces. Filter-grown monolayers were pulse-labeled with [³⁵S]methionine/cysteine for 20 min and chased for various times. At each chase interval, replicate cell monolayers were biotinylated on either the apical or basolateral side. As a control, virally infected MDCK cells were examined, and it was found that initial delivery of intact HN was 90% to the apical surface (Fig. 7A), whereas the HN₄₈PK protein arrived at 90 and 180 min with an apical preference in the 80-85% range (Fig. 7B) (and this apical delivery was already detectable at 45 min (data not shown)). The ratio of polarized initial delivery of HN₄₈PK (apical/basolateral  4:1) to the cell surface was slightly higher than the ratio of protein distribution at steady state (apical/basolateral  3:1), just opposite of what would be expected if this protein distribution were derived from basolateral to apical transcytosis. Thus, these results indicate that the polarized sorting of HN₄₈PK occurs largely at the trans-Golgi network.

[View Larger Version of this Image (24K GIF file)]

Although the apical polarity of newly delivered HN₄₈PK was impressive, by 6 h after synthesis, a substantial decline in the level of newly synthesized HN₄₈PK on the cell surface was observed (Fig. 8, dashed line). Since most of the delivery of HN₄₈PK to the cell surface was apical, the subsequent loss of most of the HN₄₈PK from the cell surface also occurred on the apical side (Fig. 7B). No HN₄₈PK was recovered in the bathing medium, indicating that the protein was not shed from transfected cells.³ However, clathrin-coated internalization of HN from the cell surface and its subsequent degradation in the endocytic pathway have been described (51); the kinetics of turnover of newly synthesized HN₄₈PK protein in MDCK cells (Fig. 8, solid line) suggested the possibility of a similar fate for the HN signal/anchor-containing chimera. Although the ordinates for the two graphs plotted on Fig. 8 are not the same, we note that 75% of newly synthesized HN₄₈PK has been reported to reside on the plasma membrane at the 3-h chase time (39). Taken together, the data suggest that despite cellular events that transpire after surface externalization (which may diminish the apparent apical protein distribution), attachment of the HN signal/anchor causes newly synthesized, glycosylated PK protein to be redirected from the basolateral to the apical surface.

[View Larger Version of this Image (19K GIF file)]

Few if any apical sorting signals for membrane proteins have been previously described to exist in cytoplasmic/transmembrane domains, as there has been greater emphasis on the possibility that such signals are contained in the ectoplasmic domain (29). Thus, we constructed HN₇₂PK (Fig. 1) to investigate whether including an increased portion of the HN N terminus, containing more of the juxtamembrane ectoplasmic region, could enhance the efficiency of apical transport. To determine the expression of HN₇₂PK in stable MDCK transfectants, radiolabeled cells were immunoprecipitated with anti-PK serum and analyzed by SDS-PAGE and fluorography. As shown in Fig. 9A, HN₇₂PK was expressed as a predominant band of ~71 kDa, with a minor band at ~69 kDa. When the same sample was treated with PNGase F, the two species migrated with the same mobility as a nonglycosylated form at ~67 kDa, indicating that the HN₇₂PK chimera undergoes N-linked glycosylation. We therefore proceeded to examine the polarized surface distribution of HN₇₂PK in transfected MDCK cells by selective cell-surface biotinylation (Fig. 9B). Quantitatively, ~75% of the HN₇₂PK chimera was apically distributed, with ~25% of the surface protein found basolaterally (Fig. 9C). Evidently, the HN₇₂PK protein behaved similarly to the HN₄₈PK chimera (Fig. 6). Thus, as for the nerve growth factor receptor (16), inclusion or exclusion of luminal juxtamembrane residues has no specific effect on apical protein distribution.

[View Larger Version of this Image (17K GIF file)]

Of note, using the same pRC/CMV expression vector to produce G418-resistant MDCK cells, several attempts to obtain clones stably expressing the full-length HN protein were unsuccessful, possibly because the hemagglutinin-neuraminidase activities produce a selective growth disadvantage in comparison with MDCK cells that do not express this protein. Thus, we cannot exclude that additional information directing apical targeting may reside in more distal portions of the ectoplasmic domain of the HN protein.

DISCUSSION

In SV5-infected MDCK cells, the viral HN protein resides almost exclusively on the apical surface (Fig. 6) (2). Since viral membrane proteins contain autonomous information regulating transport to either the apical or basolateral surface (52), there is reason to think that the HN protein may also contain such signals. In recent years, the cytoplasmic tails of certain membrane proteins have been the focus of attention as sites for basolateral sorting signals (24-28), whereas the ectodomains of membrane proteins have been the focus of attention as sites for apical sorting signals (9, 15-18). Nevertheless, in MDCK cells, truncation of the cytoplasmic tail of the vesicular stomatitis virus G-protein to only 1 amino acid still leads to a predominantly basolaterally targeted protein (23), whereas addition of the cytoplasmic tail with or without the transmembrane domain of influenza hemagglutinin to the truncated vesicular stomatitis virus G-protein may (53) or may not (9, 30) disrupt the basolateral targeting. Moreover, the "rules" that suggest that cytoplasmic/transmembrane domains are relevant only for basolateral targeting may not apply to the polarized sorting for type II membrane glycoproteins since a chimera involving the cytoplasmic/transmembrane domain of influenza virus neuraminidase (normally an apically expressed protein (5)) redirects the ectodomain of the human transferrin receptor (normally a basolaterally expressed protein) to the apical surface (54).

To identify possible polarity signals contained in the N-terminal region of the HN protein including the signal/anchor, we used chimeric constructions with PK, which has been commonly employed by others as a successful reporter for protein targeting studies (35-39). Remarkably, when secretory PK was expressed after attaching a cleavable prolactin signal peptide to its N terminus, in which the "consensus" information for signal peptide cleavage is conserved, the PK protein was secreted from the basolateral surface of transfected MDCK cells with high fidelity, approaching or equal to that of secreted proteins with defined basolateral sorting signals (31, 32). This finding is all the more remarkable in light of the fact that the basolateral secretion of this "neutral reporter" was exclusively restricted to those forms that had undergone N-linked glycosylation, which has recently been suggested to play an indirect (19) or direct (20) role in apical specific trafficking. Thus, considering the earlier studies indicating that complete inhibition of glycosylation did not affect polarized glycoprotein expression (21, 22), it appears that a role for N-glycans in apical specific targeting in MDCK cells exists only in the context of some proteins and not others (55). The 58-kDa form of secretory PK, although glycosylated, exhibits an apparently dominant basolateral sorting signal. With this in mind, the luminal PK domain serves as a useful recipient for testing potential apical targeting information in the N-terminal region of HN.

We expressed two HN-PK chimeras that contained the N-terminal signal/anchor of HN plus increasing segments of the HN ectodomain (Fig. 1). The results with both chimeras were similar: the apical distribution of the proteins was 75% (Figs. 6 and 9). Although one can never exclude the possibility that the conformation of basolaterally secreted PK (i.e. after cleavage from the signal peptide) is subtly different from that of PK tethered to the signal/anchor of HN, there is no evidence to indicate this, suggesting that the N-terminal domain of HN contains signals for apical sorting. We note that, in general, expression of the membrane protein chimeras was not as polarized as that of their endogenous counterparts, which could reflect competing apical and basolateral signals in different protein regions, and this might also be influenced by internalization of the chimeras after delivery to the apical cell surface (Figs. 7B and 8). Thus, we hypothesize that in this type II membrane protein, a dominant apical signal exists in the N-terminal domain containing the signal/anchor, which can account for most if not all of the polarized targeting of the HN protein to the apical surface. Further studies are needed to define more precisely the nature of apical specific targeting signals in this region of HN. The finding of apical specific targeting signals in the cytoplasmic domain would be of special interest in light of recent developments in the identification of new adaptor proteins that may play roles in molecular sorting of apical membrane proteins at the level of the trans-Golgi network (56).