Interactions of the AP-1 Golgi Adaptor with the Polymeric Immunoglobulin Receptor and Their Possible Role in Mediating Brefeldin A-sensitive Basolateral Targeting from the trans -Golgi Network*

We provide morphological, biochemical, and functional evidence suggesting that the AP-1 clathrin adaptor complex of the trans- Golgi network interacts with the polymeric immunoglobulin receptor in transfected Madin-Darby canine kidney cells. Our results indicate that immunofluorescently labeled g -adaptin subunit of the adaptor complex and the polymeric immunoglobulin receptor partially co-localize in polarized and semi-po-larized cells. g -Adaptin is co-immunoisolated with membranes expressing the wild-type receptor. The entire AP-1 adaptor complex could be chemically cross-linked to the receptor in filter-grown cells. g -Adaptin could be co-immunoprecipitated with the wild-type receptor, with reduced efficiency with receptor mutant whose basolateral sorting motif has been deleted, and not with receptor lacking its cytoplasmic tail. Co-immunopre-cipitation of g -adaptin was inhibited by brefeldin A. Mutation of cytoplasmic serine 726 inhibited receptor interactions with AP-1 but did not abrogate the fidelity of its basolateral targeting from the trans -Golgi network. However, the kinetics of receptor delivery to the basolateral cell surface were slowed by the mutation. Although surface delivery of the wild-type receptor was inhibited by brefeldin A, the delivery of the mutant receptor antibody was purified on protein-A Sepharose and used at 20 m g/ml concentration to label the pIgR. The 100/3 (Sigma Immunochemicals, Rehovot, Israel) antibody was used at 1:200 dilution to label the g -adap- tin. The R40.76 anti-ZO-1 rat monoclonal antibody (48) was used at 1:250 dilution to label the tight junction-associated protein ZO-1. In some experiments, cells were treated with 10 m g/ml BFA for 15 min at 37 °C prior to fixation. Anti-sheep secondary antibodies conjugated to either fluorescein isothiocyanate (FITC), anti-rat-conjugated tetramethylrhodamine isothiocyanate (TRITC), or anti-mouse coupled to Texas Red were obtained from Jackson ImmunoResearch and used at recommended dilutions as described (24). It should be noted that ac- cording to the manufacturer’s comments, secondary antibodies were tested for minimal cross-reactivity for IgG and serum proteins of other species. In several experiments, the co-localization between g -adaptin and internalized IgA was examined. In these experiments IgA was internalized into apical recycling endosome either from the basolateral or the apical plasma membrane as described previously (24). Cells on filters were then fixed and stained with anti-human IgA antibodies (Sigma), and appropriate secondary antibodies were conjugated to FITC. g -Adaptins were stained with 100/3 monoclonal antibodies and anti-mouse coupled to Texas Red. Confocal Microscopy and Co-localization— A Bio-Rad MRC-1024 confocal scanhead coupled to a Zeiss Axiovert 135M inverted microscope was used to acquire images of the stained cells, with a 63 3 oil objective (numerical aperture 1.4). Excitation light was provided by a 100-milli- watt air-cooled argon ion laser run in the multi-line mode. The excitation wavelength was 514 nm and was selected with a suitable interfer- ence filter. The relative excitation power level was set to 10% with a neutral density filter. The images presented in this paper were obtained by accumulating boiled for 5 min, and diluted with 1 ml of TDB containing 2.5% Triton X-100. The pIgR was re-immunoprecipitated using sheep anti-rabbit SC coupled to protein G-Sepharose. Samples were analyzed on 10% gels, and dried gels were autoradiographed using Kodak Bio-Max x-ray film for 72 h with intensifying screen (for 32 P–pIgR re-immunoreprecipi- tates), or for 2.5 h (for total 32 P-pIgR immunoprecipitation). The total amount of radiolabeled pIgR and the amount of radiolabeled pIgR that has been co-immunoprecipitated with g -adaptins was determined by autoradiogram scanning and imaging with the NIH Image program. Determination of the Rate of Traffic of Newly Synthesized pIgR to the Cell Surface by Protease-based Assay for Cell-surface Delivery— A pro- tease-based delivery assay was used to determine quantitatively the fraction of newly synthesized pIgR targeted from the Golgi to either the apical or basolateral surface of MDCK cells. The methodology has been described in detail, and the data obtained by this assay were consistent with results achieved by the independent cell-surface biotinylation-based delivery assay (19, 20, 42). In some experiments cells were treated with 10 m g/ml BFA (EpiCenter, stored as a 10 mg/ml stock solution in Me 2 SO at 2 20 °C), as described (42).

Polarized epithelial cells possess two surface domains as follows: the apical plasma membrane that faces the external environment and the basolateral plasma membrane that is in contact with internal milieu. The apical and basolateral plasma membranes have very different protein and lipid compositions. It has been proposed that sorting and targeting events of membrane proteins and lipids largely contribute to the polarized phenotype of the cell (1)(2)(3). Although in recent years we have learned a great deal about relatively small stretches of amino acids encoding sorting signals that mediate polarized trafficking of membrane proteins (4 -7), almost nothing is known about the sorting machinery that decodes these signals and confers polarized trafficking to a specific surface domain (1, 2, 6 -8).
Polarized trafficking steps are modulated by signal transduction processes (4,7,9,10), whose activity is thought to be superimposed on the activity of the sorting machinery, but the mechanisms through which these processes facilitate protein incorporation into a particular pathway are largely unknown.
Presently, we know four distinct features shown to determine sorting of apical proteins to the apical domain as follows: first is the glycosylphosphatidylinositol anchor of membrane proteins (11,12); the second is the mannose-rich core of Nglycans present in the luminal portion of proteins (13); the third is O-glycosylated stalk domain of transmembrane protein (14); and the fourth is a proteinaceous signal encoded by either the transmembrane domain or the ectodomain (15). Many studies agree that basolateral sorting of plasma membrane proteins is mediated by the presence of relatively short but specific cytoplasmic sorting motifs (reviewed in Refs. 4 and 5). Extensive mutagenesis studies have uncovered two general types of basolateral sorting motifs. First there are basolateral sorting signals for localization to clathrin-coated membranes that rely either on a critical tyrosine residue, such as those found in the low density lipoprotein receptor (low density lipoprotein receptor proximal determinant, (16)), the vesicular stomatitis virus G protein (17), or on a di-leucine motif (18). The second are basolateral sorting signals unrelated to localization to clathrincoated membranes. These signals can rely either on a Tyr motif, e.g. the distal determinant in the low density lipoprotein receptor, or rely on non-aromatic residues such as found for the polymeric immunoglobulin receptor (pIgR 1 (4,19)). Interestingly, either the same or a closely related basolateral sorting signal can mediate basolateral sorting from the TGN and recycling from endosomes, after endocytosis from the basolateral plasma membrane (20,21).
We use the pIgR expressed in Madin-Darby canine kidney (MDCK) cells as a model system for studying the mechanisms that regulate polarized trafficking of membrane proteins in epithelial cells. Previous studies have intensively analyzed the intracellular trafficking of the pIgR in MDCK cells. According to the simplest model, the pIgR is synthesized in the endoplasmic reticulum (ER) as a single spanning type I membrane protein. It is then targeted from the ER to the Golgi apparatus and from the trans-Golgi network (TGN) pIgR molecules are directed to the basolateral surface. However, the exact pathway that newly synthesized membrane proteins take to the basolateral cell surface is currently not understood. Although in the simplest model, basolateral proteins are selectively sorted (packed) into basolateral transport vesicles that are subsequently directly delivered to the basolateral surface, recent studies have provided evidence that targeting to the cell surface may be indirect, i.e. involving passage through an endosomal compartment prior to cell-surface delivery (22,23). Whether a similar indirect route exists in MDCK cells is currently unknown nor is the mechanism of sorting to these endosomes known. Upon arrival at the basolateral cell surface, polymeric Igs (e.g. dimeric IgA; dIgA) specifically bind to the pIgR. The pIgR⅐dIgA complexes are endocytosed via clathrin-coated pits, delivered to basolateral endosomes, and then delivered to the opposite pole of the cell via vesicular intermediates in a process termed transcytosis. Transcytotic pIgR⅐dIgA complexes are not directly routed from basolateral endosomes to the apical surface. Rather, pIgR⅐Ig complexes are first targeted to, and subsequently accumulate at, a pericentriolar subapical endosomal compartment that is active in recycling of apical and perhaps basolateral membrane-bound ligands (24,25). This compartment has been termed the "apical recycling endosome." The final step of transcytosis involves pIgR targeting from the apical recycling endosome to the apical surface. There, the pIgR is cleaved by an endogenous protease, and the extracellular ligand-binding domain (i.e. the secretory component, SC) is released together with dIgA to external (mucosal) secretions, such as milk, saliva, bile, tears, and intestinal fluids, where they form the first immunological response against infections. The pIgR is also transcytosed constitutively (i.e. in the absence of the ligand), and this process is regulated by phosphorylation of cytoplasmic Ser-664 (26).
Mutagenesis studies revealed that the cytoplasmic domain of the pIgR contains several discrete sorting signals that mediate as follows: its targeting from the TGN to the basolateral surface; endocytosis; avoidance of lysosomes; and transcytosis (4,6). Basolateral sorting from the TGN is mediated by an autonomous and dominant 17-residue membrane-proximal basolateral sorting signal (27), whose activity depends mainly on three amino acids (His-656, Arg-657, and Val-660) contained within the 17-residue signal (19). Structural studies of oligopeptides corresponding to the 17-residue basolateral sorting signal provided evidence for a ␤-turn secondary structure, which might constitute a general feature of endocytotic and other sorting signals (7). Residues involved in basolateral sorting of the pIgR from the TGN also control polarized sorting in endosomes (20), suggesting that common mechanisms regulate polarized trafficking in the TGN and in endosomes. The pIgR thus appears to contain multiple cytoplasmic signals that mediate distinct intracellular transport events; these signals are probably decoded by a cytoplasmic sorting machinery located at specific compartments through which the pIgR passes en route to a target organelle.
Coat proteins such as clathrin and clathrin adaptor proteins (AP), AP-1 and AP-2, are thought to be involved in membrane protein sorting into coated membrane domains and promote the budding of transport vesicles from the trans-Golgi network (TGN) and from the plasma membrane, respectively (for recent reviews see Refs. 28 and 29). The AP-1 and AP-2 adaptor complexes are composed of two large subunits (␥ and ␤1 for AP-1 or ␣ and ␤ for AP-2), a medium-sized subunit (1 or 2), and a small subunit (1 and 2). Although in a few cases the interactions between AP-binding and membrane proteins have been resolved (for a recent review see Ref. 30), the mechanistic relations between coat binding and membrane protein targeting has not been fully elucidated. One common feature is that many of the coat proteins involved in protein sorting are sensitive to the action of the fungal metabolite brefeldin A (BFA). This drug is thought to act, in part, by blocking the binding to Golgi membranes of at least four coat proteins as follows: coatomer (e.g. COP-I) involved in ER-Golgi transport and in endocytosis (8,29,31); the TGN clathrin-coated vesicle-associated protein AP-1 adaptor complex (32) involved in sorting of the mannose 6-phosphate receptor (MPR) to endosomes; p200, a protein of 200,000 daltons that is now known to be type II myosin (33,34); and the recently discovered AP-3 adaptor protein (35,36) whose association with clathrin coats in the Golgi is controversial (36,37). Both, AP-1 and COP-I bind to target membranes in association with a small GTPase ARF1 (for ADP-ribosylation factor 1) (38). GDP-GTP exchange on ARF occurs concomitant with binding; inhibition of exchange by BFA blocks the assembly of these coat proteins. Do coat proteins participate in polarized membrane trafficking? Recent in vitro binding studies revealed that AP-1 adaptors interact with tyrosine-based basolateral sorting motif and with a dileucine signal artificially introduced into the influenza hemagglutinin's (HA) C-terminal domain (39), but the function of these interactions in basolateral trafficking of the HA mutant is unknown.
Basolateral targeting of the pIgR from the TGN is significantly inhibited by BFA (40 -42), suggesting that BFA-sensitive coat proteins regulate pIgR exocytosis from the TGN to the basolateral surface. The cytoplasmic tail of pIgR contains a phosphorylated Ser, Ser-726, that functions in rapid internalization of the pIgR in the basolateral plasma membrane via clathrin-coated pits (43). This Ser-based motif resides in a putative CKII/PKA phosphorylation site upstream to a dileucine motif with yet undefined function (see Fig. 1A and Ref. 43). In this respect this motif is interesting as it resembles to the putative CKII/PKA motif present in the cytoplasmic tail of cation-dependent mannose 6-phosphate receptor (CD-MPR), shown to serve as an AP-1-binding site (44,45). These observations led us to propose that AP-1 adaptors bind to the Ser-726 motif of pIgR and that these interactions may facilitate certain steps in basolateral exocytosis of the receptor. The goal of the work reported herein was to identify these interactions and to examine their role in pIgR trafficking from the TGN to the basolateral cell surface in polarized MDCK cells. We find that the AP-1 adaptor complex associates with the wild-type pIgR cytoplasmic tail in vivo and that the interactions are diminished when Ser-726 is mutated to Ala. The interactions with AP-1 seem to commence in the TGN, persist, and/or are even enhanced in post-TGN compartments. Interestingly, the TGN to basolateral delivery of pIgR-Ser-726 to Ala mutant appears to be significantly slower than that of the wild-type receptor. In addition, unlike the wild-type receptor, the basolateral pathway of this pIgR mutant is completely insensitive to the action of BFA. Our results indicate that AP-1 interactions play a regulatory role in prompting efficient basolateral transport of the pIgR from the TGN, and they also suggest the existence of multiple mechanisms that direct membrane proteins from the TGN to the basolateral surface in MDCK cells.

EXPERIMENTAL PROCEDURES
Characterization of MDCK Cells Stably Expressing the pIgR-MDCK cells expressing the wild-type or mutant receptors were maintained for up to 10 passages in minimal Eagle's medium (MEM, Biological Industries Co, Beit Haemek, Israel) supplemented with 5% (v/v) fetal bovine serum (Biological Industries Co, Beit Haemek, Israel), 100 units/ml penicillin, and 0.1 mg/ml streptomycin in 5% CO 2 , 95% air. In our experiments we used three MDCK cell lines that stably express the wild-type pIgR ("pIgR-WT"). One previously described cell line generated by cell transfection with the retroviral pWE vector has been used (26). Two additional cell lines were generated by transfecting MDCK cells with the wild-type pIgR cDNA subcloned into the BglII sites of cytomegalovirus-based pCB6 or pCB7 vectors, as described previously (19,46). In some cases, MDCK cells expressing pIgR mutants at levels comparable to those of the wild-type receptor were generated by transfecting the cells with cDNA encoding mutant receptors subcloned into the pCB6 vector. Polarized MDCK cells expressing receptors were isolated and characterized as described (19,20). Two clones expressing the "pIgR R654stop" mutant, where all but two residues of the cytoplasmic tail have been deleted (47), have been isolated and their exocytic transport characteristics were analyzed and found to be identical to those described for the earlier characterized expressors (47). The pIgR-⌬ 655-668 mutant whose 15 of the 17 residues comprising the basolateral sorting signal have been deleted in-frame is delivered from the TGN directly to the apical surface (27). Newly isolated pIgR-⌬655-668 expressing MDCK clones revealed identical exocytic transport properties to the originally described clone (27). The previously described MDCK clones expressing pIgR S726A, whose Ser at position 726 in the cytoplasmic tail was mutated to Ala, have been used. This mutation does not abrogate pIgR sorting from the TGN to the basolateral surface, but it inhibits its endocytosis from that surface (43). Expression level was determined by SDS-PAGE analysis of equal protein amounts derived from SDS-cell lysates followed by immunoblotting and probing with polyclonal sheep anti-SC antibodies and appropriate horseradish peroxidase-labeled secondary antibodies. Protein bands were detected using the SuperSignal® chemiluminescent reagent (Pierce), according to the manufacturer's protocol. Autoradiograms were scanned at 300 dpi resolution using the HP ScanJet IIcx scanner, and band intensity was quantified by the NIH image 1.61 software. Mounting of figures was performed using the Adobe Photoshop TM 3.05 (Adobe Systems, Inc., Mountain View, CA) and Aldus Freehand 5.02, Macromedia Inc. Autoradiograms showing the expression of selected pIgR mutants relative to the wild-type pIgR are depicted in Fig. 1B.
Fixation and Fluorescent Labeling of Cells-MDCK cells expressing the wild-type receptor were grown on Transwell filters (Costar, 0.4 m) for 3-4 days. Cells on filters were fixed for 10 min in methanol at Ϫ20°C, blocked, immunostained for pIgR and ␥-adaptin, mounted, and stored as described previously (24). The primary sheep anti-rabbit SC  (27) are indicated with an underline. The two main phosphorylation sites of the cytoplasmic tail are contributed by Ser-664 and Ser-726 (marked with an asterisk); these residues have been shown to play a role in constitutive transcytosis and endocytosis, respectively (26,43). Ser-726 resides in a putative CKII/PKA phosphorylation site, located upstream to a double leucine-based motif (underlined). Typical examples of CKII and PKA phosphorylation motifs are SXXE and KRXS, where X denotes any amino acid, and are indicated in parentheses. In the pIgR mutant, pIgR-⌬655-668, an internal deletion (⌬) has been constructed from Arg-655 to Tyr-668, leaving the C-terminal 89 amino acids intact and fused to Ala-654 (27). A stop codon has been inserted instead of arginine 655 to produce a pIgR mutant lacking all but two residues of the cytoplasmic tail (pIgR R654stop (47)). In the pIgR-S726-A mutant, serine at position 726 has been substituted with an alanine (43). The Ser-726 to Ala mutation has been shown to inhibit pIgR endocytosis, without affecting its basolateral delivery from the TGN (43). B, pIgR expression level was determined by SDS-PAGE and immunoblotting analyses of equal protein amounts derived from cell lysates as indicated under "Experimental Procedures." The expression levels of pIgR mutants were compared with those of the pIgR-WT. Cell clones expressing comparable or somewhat higher expression levels than the wild-type receptor were chosen for analysis of interactions with ␥-adaptin.
antibody was purified on protein-A Sepharose and used at 20 g/ml concentration to label the pIgR. The 100/3 (Sigma Immunochemicals, Rehovot, Israel) antibody was used at 1:200 dilution to label the ␥-adaptin. The R40.76 anti-ZO-1 rat monoclonal antibody (48) was used at 1:250 dilution to label the tight junction-associated protein ZO-1. In some experiments, cells were treated with 10 g/ml BFA for 15 min at 37°C prior to fixation. Anti-sheep secondary antibodies conjugated to either fluorescein isothiocyanate (FITC), anti-rat-conjugated tetramethylrhodamine isothiocyanate (TRITC), or anti-mouse coupled to Texas Red were obtained from Jackson ImmunoResearch and used at recommended dilutions as described (24). It should be noted that according to the manufacturer's comments, secondary antibodies were tested for minimal cross-reactivity for IgG and serum proteins of other species. In several experiments, the co-localization between ␥-adaptin and internalized IgA was examined. In these experiments IgA was internalized into apical recycling endosome either from the basolateral or the apical plasma membrane as described previously (24). Cells on filters were then fixed and stained with anti-human IgA antibodies (Sigma), and appropriate secondary antibodies were conjugated to FITC. ␥-Adaptins were stained with 100/3 monoclonal antibodies and anti-mouse coupled to Texas Red.
Confocal Microscopy and Co-localization-A Bio-Rad MRC-1024 confocal scanhead coupled to a Zeiss Axiovert 135M inverted microscope was used to acquire images of the stained cells, with a 63ϫ oil objective (numerical aperture 1.4). Excitation light was provided by a 100-milliwatt air-cooled argon ion laser run in the multi-line mode. The excitation wavelength was 514 nm and was selected with a suitable interference filter. The relative excitation power level was set to 10% with a neutral density filter. The images presented in this paper were obtained by accumulating (summing) three scans. In order to detect three dyes (FITC, TRITC, and Texas Red) simultaneously, the three detection channels were configured as follows.
The fluorescence emission was first split between two channels by a dichroic mirror (555DRLP, 50% point at 550 nm). The low wavelength side of the dichroic was followed by an HQ535/20 (535 nm Ϯ 10 nm) band pass filter. The iris aperture was 2.5-3.0 mm. This channel detected FITC. However, when the TRITC emission was much stronger than the FITC emission, TRITC emission from the tight junctions was also detected in this channel. In post-processing, the output of the TRITC channel was subtracted from the FITC channel to eliminate the contribution of TRITC to this channel. This was particularly effective because the TRITC only stained tight junctions and was therefore well localized and distinct from the FITC emission.
The long wavelength side of the first dichroic mirror was split by a second dichroic mirror (605DRHP, 50% point at 605 nm). The short wavelength side of this dichroic mirror was followed by an HQ570/30 (570 nm Ϯ 15 nm) band pass filter. The iris aperture was 0.7-1.7 mm. This channel detected TRITC, as well as FITC. When necessary, the FITC signal as detected in the short wavelength channel described above was subtracted from the output of the TRITC channel. This subtraction was done using the mixer controls on the Bio-Rad MRC-1024. In some cases the TRITC emission was much stronger than the FITC emission, and this subtraction was not required.
The long wavelength side of the second dichroic mirror was followed by an HQ655/90 (655 nm Ϯ 45 nm) band pass filter. The iris aperture was 3.5 mm, and this channel was optimized for Texas Red. All filters were from Chroma Technology Corp. (Brattleboro, VT). Since this channel also detected TRITC and some FITC, worst case cross-talk from each of these sources was estimated and then subtracted from this channel, as follows.
Areas on the images where FITC was dominant were used in order to establish an upper limit to any such bleed-through. Similarly, the tight junctions were used to determine bleed-through of TRITC to the other two channels. The assumption that such areas are due only to one fluorophore, and that appearance of a signal at that point in the other channels is caused entirely by bleed-through, sets an upper limit on the bleed-through. Once a bleed-through factor ␣ was determined between two channels, the bleed-through was eliminated by applying Equation 1.
where C A is the channel to be corrected and C B is the channel from which the bleed-through occurs. Note that the areas used for correction were obtained in the linear portion of the PMT response (gray level under 200). Co-localization of the FITC and Texas Red is visualized using the following processing steps. First, a 3 ϫ 3 median filter was applied to remove point noise. Then, cross-talk was eliminated as described above. Finally, the image contrast was stretched. The contrast-enhanced images were then merged into a single true color image (green for FITC, red for Texas Red, blue for TRITC). Areas of overlap of Texas Red and FITC appear yellow. There is no attempt to quantitate the relative strengths of the FITC and Texas Red emission.
Immunoisolation of MDCK Membranes Expressing the pIgR-A postnuclear supernatant (PNS) fraction was prepared from pIgR expressing MDCK cells as follows: confluent MDCK culture grown on a 150-mm culture dish was washed twice with ice-cold PBS and once with homogenization buffer (HB, 250 mM sucrose, 3 mM imidazole, pH 7.4). Cells were then scraped from the dish with a rubber policeman, and a mixture of protease inhibitors (25 g/ml pepstatin, 50 g/ml chymostatin, 25 g/ml leupeptin, 50 g/ml antipain, 2.5 g/ml aprotinin) was added. Cells were homogenized through a 21-gauge needle connected to a 1-ml syringe. A PNS fraction was obtained after centrifugation at 600 ϫ g for 15 min (4°C). Streptavidin-coated magnetic beads (30 l, DynaBeads M-280, Dynal, Oslo, Norway) were reacted first with biotinylated rabbit anti-mouse IgG (3 g, 1 h, 4°C; Sigma), washed to remove unbound antibodies, and then incubated (16 h, 4°C) with 3 g of protein G-Sepharose-purified monoclonal antibody SC166 directed against the cytoplasmic tail of the pIgR (49). Immunobeads were extensively washed of unbound antibodies and then incubated with PNS for 16 h at 4°C. Immunobeads were then washed three times with cold HB buffer, and bound (immunoisolated) membranes were extracted with 1% Triton X-100 in HB buffer (10 min at 22°C). One volume of 2ϫ concentrated SDS sample buffer (1ϫ sample buffer: 15 mM Tris-HCl, pH 6.8, 25 mM EDTA, pH 7.0, 0.25% urea mixed with bromphenol blue, 6% v/v glycerol, 65 mM dithiothreitol) was added, and the SDS/Triton mixture was subjected to SDS-PAGE analysis on 10% acrylamide gels, using the Hoefer Mighty Small II gel system (Hoefer Scientific Instruments, San Francisco). After SDS-PAGE, proteins were electrophoretically transferred onto a nitrocellulose membrane for 2 h at 90 mA per single gel. Membranes were then incubated in blocking buffer (PBS containing 0.5% Tween, 10% glycerol, 0.025% w/v BSA, 0.01% w/v dry milk, 40 mM glucose, 0.01% sodium azide) for 1 h at room temperature and subsequently incubated for 16 h at 4°C with sheep anti-pIgR serum diluted 1:1000 in PBS supplemented with 0.05% Tween and 5% milk. After washings the membranes were incubated for 1 h in PBS containing 0.05% Tween and 1% milk and appropriate horseradish peroxidaselabeled secondary antibodies (Jackson ImmunoResearch). Horseradish peroxidase-labeled secondary antibodies were detected using the Su-perSignal ® chemiluminescent reagent (Pierce), according to the manufacturer's protocol. The signal on blots was then quenched by incubating the nitrocellulose sheets in PBS containing 0.1% sodium azide for 30 min, and the same nitrocellulose sheet was re-probed with 100/3 mouse antibodies (Sigma) followed by horseradish peroxidase-labeled antimouse antibodies (Jackson ImmunoResearch) for the detection of ␥-adaptin. Autoradiograms were scanned and band intensity was quantified by NIH image.
Detection of pIgR⅐AP-1 Complexes by Chemical Cross-linking-MDCK cells expressing the wild-type pIgR were cultured for 3-5 days on a 100-mm Transwell porous filter and rinsed twice with ice-cold TGH, and the basolateral surface of the cells was selectively permeabilized by placing the filter onto 1 ml of TGH buffer (50 mM Hepes-NaOH, pH 7.4, 1 mM EGTA-NaOH, 10 mM MgCl 2 , 150 mM NaCl, 1 mM sodium orthovanadate, 10% v/v glycerol) containing 50 g/ml digitonin (Sigma) for 20 min on ice, as described (50). Cells were washed with ice-cold TGH, and the permeabilized surface was exposed to 2 mM DTSSP (3,3Ј-Dithiobis[sulfosuccinimidyl-propionate, from Pierce), diluted from a 100 mM stock in N,N-dimethylformamide into Hepes buffer (25 mM Hepes, pH 7.4, 1 mM MgCl 2 , 0.25 M sucrose, 1 mM sodium orthovanadate). Following incubation for 120 min at 4°C with the cross-linker, the activity of the cross-linker was quenched by incubation with 150 mM glycine, pH 7.0, in Hepes buffer for 20 min at 22°C. The buffer was aspirated leaving the cells on filter as dry as possible, and cells were then scraped off the filter using a rubber policeman. Detached cells were resuspended into 1 ml of 2.5% Triton dilution buffer (TDB: 100 mM triethanolamine chloride, pH 8.6, 100 mM NaCl, 5 mM EDTA, 0.025% w/v NaN 3 ) containing 2.5% w/v Triton X-100, phenylmethylsulfonyl fluoride and a mixture of protease inhibitors, homogenized through a 1-ml pipette tip, and cleared with CL-2B (50% slurry, Pharmacia, Uppsala, Sweden) as described (19). The pIgR was immunoprecipitated, and co-immunoprecipitated adaptor subunits were analyzed by SDS-PAGE and immunoblotting. The 100/3 monoclonal antibody was used for detecting ␥-adaptins; monospecific rabbit antibodies directed against ␤Ј-adaptin, -1, or -1 subunits kindly provided by Prof. Margaret S. Robinson (Cambridge, UK) were used as documented (51).
Blots were incubated with horseradish peroxidase-labeled secondary antibodies, and protein visualization was obtained by the SuperSignal® chemiluminescent reagent.
Co-immunoprecipitation Detected by Electrophoresis and Western Blotting-The co-immunoprecipitation protocol is principally adapted from Sorkin and Carpenter (52), with the modifications applied by Fire et al. (53), who co-immunoprecipitated the AP-2 adaptors with the epidermal growth factor receptor, or influenza HA, respectively. All the following steps were performed at 4°C. Confluent MDCK cell culture grown on a 100-mm Petri dish was split 1:10 onto three 150-mm dishes, 3-4 days prior to the experiment. Cells in each dish were rinsed twice with ice-cold PBS and once with TGH buffer. The buffer was then aspirated, leaving the cell monolayer as dry as possible. Cells were scraped into 300 l of ice-cold TGH containing freshly diluted Triton X-100 (1% w/v), phenylmethylsulfonyl fluoride (5 mM), and a mixture of protease inhibitors. Cell lysate from each dish was pooled into a single Eppendorf tube, and the combined lysate was gently homogenized through a 1-ml pipette tip. Extracts were centrifuged (13,200 rpm, 15 min, 4°C; IEC Microfuge RF) to remove nuclei and detergent-insoluble membranes, and the resultant supernatant was further spun (75,000 rpm, 30 min, 4°C; TL-100 Beckman centrifuge) to remove any remaining Triton-insoluble materials. The protein concentration of the high speed supernatant was determined to be 7-10 mg/ml, using the bicinchoninic acid (BCA) assay kit from Pierce. The pIgR was immunoprecipitated from this supernatant by incubation with sheep anti-rabbit SC covalently coupled to protein A-Sepharose for 3 h at 4°C (19,54). Immunocomplexes were washed of unbound proteins with cold TGH/ Triton X-100, and pelleted beads were boiled for 5 min in 2ϫ sample buffer. About one-tenth of the sample buffer volume was excluded for pIgR analysis and the remainder for the detection of ␥-adaptins. Samples were analyzed by SDS-PAGE and Western blotting, and signals contributed by the pIgR and ␥-adaptins were quantified as above.
Co-immunoprecipitation of 35 S-Labeled pIgR with ␥-Adaptin-MDCK cells expressing the wild-type pIgR were cultured for 3 days on 24 mm Transwell filters prior to the experiment. To co-immunoprecipitate pIgR located in the ER, cells were starved for 10 min at 37°C in Dulbecco's modified Eagle's medium lacking L-Cys and L-Met (Sigma) but supplemented with Hanks' balanced salts and 20 mM Hepes, pH 7.4, 5% dialyzed fetal bovine serum, and pulse-labeled for 8 min with 4.5 mCi/ml [ 35 S]Met-Cys (Pro-Mix, Amersham Corp., Buckinghamshire, UK). To co-immunoprecipitate pIgR accumulated in the TGN (i.e. after TGN block), cells after the pulse were incubated for 120 min at 18°C with MEM/BSA (MEM containing Hanks' balanced salt solution (Sigma), 20 mM Hepes, pH 7.4, 0.6% BSA) present on the apical and basolateral chambers of the Transwell; after the 18°C chase the pIgR becomes insensitive to digestion by endoglycosidase H (41). To coimmunoprecipitate pIgR chased into post-TGN compartments, cells subjected to TGN block incubation were further incubated in warm (37°C) MEM/BSA for 10, 30, or 60 min. Cells were then lysed in ice-cold TDB containing 1.25% Triton X-100, protease inhibitors, and 1 mM sodium orthovanadate, pre-cleared once with Sepharose CL-2B (19) (Pharmacia, Uppsala, Sweden), and ␥-adaptin was immunoprecipitated at 4°C for 16 h with the 100/3 antibodies bound to protein A-Sepharose. Beads were washed three times with "mixed micelle" buffer (MMB: 20 mM triethanolamine Cl, pH 8.6, 150 mM NaCl, 5 mM EDTA, pH 8.0, 8% w/v sucrose, 0.1% NaN 3 , 1% w/v Triton X-100, 0.2% SDS), once with "Final Wash" buffer (MMB lacking Triton X-100 and SDS), and bead pellet was boiled in 80 l of 1% SDS. 35 S-pIgR was re-immunoprecipitated from the released proteins with anti-pIgR antibodies coupled to protein A-Sepharose. The total amount of radiolabeled pIgR was determined by pIgR immunoprecipitation from cells that have been metabolically labeled under identical conditions. Samples were analyzed on 10% SDS gels; dried gels were exposed to Fuji Imaging Plates, and pIgR radioactivity levels were determined by the FUJIX BAS 1000 imaging system. Gels onto which co-immunoprecipitated pIgR was analyzed were exposed for 3-days, whereas gels with total pIgR were exposed for 15 h.
Co-immunoprecipitation of 32 P-Labeled pIgR with ␥-Adaptin-Cells were grown on 35-mm dishes and labeled with 0.5 mCi/ml [ 32 P]orthophosphate for 3 h. Cells were lysed in cold TDB containing 1.25% Triton X-100, a mixture of protease inhibitors, and 1 mM vanadate. Lysates were precleared with Sepharose CL-2B and incubated with anti-␥-adaptin 100/3 antibodies and protein G-Sepharose for 16 h at 4°C with continuous end-to-end rotation. Immunoprecipitates were washed three times with ice-cold MMB buffer and one time with "Final Wash" buffer. Immunoprecipitates were treated with 80 l of 1% SDS, boiled for 5 min, and diluted with 1 ml of TDB containing 2.5% Triton X-100. The pIgR was re-immunoprecipitated using sheep anti-rabbit SC coupled to protein G-Sepharose. Samples were analyzed on 10% gels, and dried gels were autoradiographed using Kodak Bio-Max x-ray film for 72 h with intensifying screen (for 32 P-pIgR re-immunoreprecipitates), or for 2.5 h (for total 32 P-pIgR immunoprecipitation). The total amount of radiolabeled pIgR and the amount of radiolabeled pIgR that has been co-immunoprecipitated with ␥-adaptins was determined by autoradiogram scanning and imaging with the NIH Image program.
Determination of the Rate of Traffic of Newly Synthesized pIgR to the Cell Surface by Protease-based Assay for Cell-surface Delivery-A protease-based delivery assay was used to determine quantitatively the fraction of newly synthesized pIgR targeted from the Golgi to either the apical or basolateral surface of MDCK cells. The methodology has been described in detail, and the data obtained by this assay were consistent with results achieved by the independent cell-surface biotinylationbased delivery assay (19,20,42). In some experiments cells were treated with 10 g/ml BFA (EpiCenter, stored as a 10 mg/ml stock solution in Me 2 SO at Ϫ20°C), as described (42).

␥-Adaptin Partially Co-localizes with the pIgR in Polarized
and Semi-polarized MDCK Cells-To provide qualitative analysis of the interactions between pIgR and ␥-adaptin, we analyzed the co-localization between these molecules in MDCK cells grown on filter supports (i.e. in polarized cells, Fig. 2A) or in cells grown on coverslips (i.e. in semi-polarized cells, Fig. 2B) using confocal immunofluorescence microscopy. Clear and dense overlapping fluorescent signals (yellow in the merged panel) contributed by the pIgR (green) and ␥-adaptin (red) are seen primarily in the center of the apical region of many cells (an example is given in Fig. 2A). In addition to the trans-Golgi network, ␥-adaptin has been recently reported to be localized to basolateral endosomal structures (50); thus, the co-localization of ␥-adaptin and pIgR could represent co-localization of the two proteins in endosomal membranes. However, the ␥-adaptinpositive structures observed here are apical, and, in doublelabeling experiments, dIgA internalized into apical endosomes from either the apical or basolateral membrane does not significantly co-localize with ␥-adaptin. 2 Thus, we conclude that the majority of co-localization of ␥-adaptin and pIgR is associated with the trans-Golgi network. In addition, in immunostained MDCK cells grown on coverslips, significant co-localization between ␥-adaptin and pIgR could be observed in perinuclear structures, a distribution characteristic of the trans-Golgi network in semi-polarized cells (32).
The ␥-Adaptin Subunit of the AP-1 Complex Is Co-immunoisolated with Membranes Expressing the pIgR-We have used MDCK monolayers that express the wild-type pIgR and various pIgR mutants to identify the association between the pIgR and ␥-adaptin by immunoisolation of membranes from MDCK cells. A PNS fraction was prepared from MDCK cells expressing the wild-type pIgR, and the crude membrane preparation was incubated with purified antibodies directed against the cytoplasmic tail of the pIgR (49) coupled to magnetic beads. Immunoisolated membranes were solubilized and analyzed for the presence of pIgR and ␥-adaptin by immunoblotting. Data presented in Fig. 3A demonstrate that ␥-adaptin is co-immunoisolated with membranes bearing the wild-type pIgR. The antibodies specifically immunoisolate membranes expressing an intact cytoplasmic tail since only residual amounts of pIgR are detected upon incubation of immunobeads with membranes containing a pIgR mutant which lacks all but two cytoplasmic amino acid residues (pIgR-R654stop, Fig. 3B,  upper panel). ␥-Adaptin did not bind to immunobeads exposed to MDCK cells that do not express the pIgR (Fig. 3A), further suggesting that the coat protein is specifically co-isolated with pIgR-expressing membranes. To evaluate the relative extent of ␥-adaptin association with membranes expressing pIgR mu-tants, the ␥-adaptin/pIgR ratio was determined in the immunoisolated membranes, and the value obtained was normalized to that determined for the wild-type receptor (Fig. 3B, lower  panel). The results reveal a significant reduction in ␥-adaptin/ pIgR ratio in membranes isolated from MDCK cells expressing pIgR-⌬655-668, or pIgR-S726-A. Note that although the expression levels of pIgR mutants were similar to those of the wild-type receptor (Fig. 1B), membranes expressing pIgR mutants were more efficiently immunoisolated since compared with the expression of the wild-type pIgR, higher pIgR levels are associated with them (compare band intensity of pIgR-⌬655-668 or pIgR-S726-A to band intensity of pIgR-WT in Fig.  3B, upper panel). The apparently inefficient isolation of membranes expressing the wild-type receptor might be attributed to inaccessibility of the antibody to the cytoplasmic tail by associated cytosolic proteins, e.g. by ␥-adaptin, which may be more efficiently recruited to the cytoplasmic tail of the wild-type pIgR but not to its mutants, making these membranes less accessible to the anti-tail antibodies. These results suggest that ␥-adaptin associates more extensively with membranes containing the wild-type pIgR than those with the indicated pIgR mutants. Our next aim was to characterize the interaction between ␥-adaptin and the pIgR.
AP-1 Adaptors Interact with the Wild-type pIgR-We employed cross-linking to examine the interactions between pIgR and AP-1 adaptors in filter-grown cells. The basolateral plasma membrane of pIgR-expressing MDCK cells was permeabilized with digitonin, and the permeabilized surface was then incubated with the hydrophilic cross-linker DTSSP as described under "Experimental Procedures." Cells were solubilized, and pIgR was immunoprecipitated. Immunoprecipitates were analyzed on Western blots for the presence of ␥-adaptin, ␤1-adaptin, 1, and 1 subunits of the AP-1 adaptor complex. The results presented in Fig. 4 demonstrate that all four subunits of the AP-1 complex co-precipitate with the pIgR in the crosslinked cells but not in cells that were not exposed to the crosslinker. Neither pIgR nor adaptor chains were co-precipitated with plain protein A. These results hence suggest that the pIgR is complexed with the entire AP-1 adaptor in polarized MDCK cells.
To confirm that the putative interaction between AP-1 and pIgR demonstrated by cross-linking is a reflection of in vivo interactions, we next developed a co-immunoprecipitation approach to study the interactions between the two molecules. This methodology has been successfully used by others (52,53) to demonstrate the interactions between endocytic membrane proteins and AP-2 adaptors. MDCK cells expressing the wildtype pIgR were solubilized in buffer containing Triton X-100, and the pIgR was immunoprecipitated with an anti-ectodomain antibody (anti-SC). Immunoprecipitates were analyzed by immunoblotting for co-immunoprecipitating ␥-adaptin. Negligible levels of pIgR and ␥-adaptin were associated with plain protein

FIG. 2. Confocal immunofluorescence microscopy analysis of co-localization of ␥-adaptin with transfected pIgR in filter-grown (A) or coverslip-grown (B) MDCK cells.
MDCK cells expressing the wild-type pIgR were cultured for 4 days on 12.5-mm diameter filter supports or for 3 days on a coverslip. Cells were triple-labeled for pIgR (purified sheep anti-rabbit pIgR followed by FITC-conjugated anti-sheep immunoglobulin G), ␥-adaptin (100/3 monoclonal antibody followed by Texas Red-conjugated anti-mouse antibodies), and for the tight-junction ZO-1 protein (R40.76 anti-ZO-1 rat monoclonal antibodies followed by TRITC-conjugated anti-rat antibodies) (blue). In the overlay images obtained for pIgR (green) and ␥-adaptin (red), yellow indicates areas of co-localization. Arrows point to structures where pIgR and ␥-adaptins co-localize.

Functional Interactions of AP-1 Adaptors with IgA Receptor
A-Sepharose beads or with protein A-Sepharose coupled to irrelevant sheep IgG (Fig. 5A, upper panel). In contrast, only when the pIgR was immunoprecipitated by anti-SC antibodies were significant amounts of ␥-adaptin co-immunoprecipitated (Fig. 5A, upper panel). The amount of co-immunoprecipitated ␥-adaptin correlated with the amount of pIgR that had been immunoprecipitated from cell lines expressing different levels of pIgR (not shown), further suggesting that AP-1 interacts with the wild-type pIgR in MDCK cells.
We compared the efficiency with which ␥-adaptin was coimmunoprecipitated with the pIgR after cross-linking versus after co-immunoprecipitation. The cross-linker covalently links the AP-1 adaptor with the pIgR, thus higher efficiency of ␥-adaptin co-precipitation is expected to be observed in the cross-linking experiment. Indeed, data presented in Fig. 5A (lower panel) are consistent with this expectation. The amount of co-immunoprecipitated ␥-adaptin normalized to the level of precipitated pIgR in a typical cross-linking experiment was about 20 times higher than the ratio determined for a coimmunoprecipitation experiment. This result suggests that coimmunoprecipitation experiments are limited in their ability to provide a quantitative estimation on pIgR fraction that interacts with ␥-adaptin in the cell. Flexible and weak interactions probably mediate pIgR association with proteins involved in its trafficking, and the majority of these interactions are likely to be disrupted upon cell solubilization under conditions of coimmunoprecipitation. Even the cross-linking approach cannot provide such information as it is impossible to determine the absolute efficiency at which AP-1 complexes are crosslinked to the receptor. Nonetheless, since co-immunoprecipitation is relatively easy to perform, and as it enables the detection of specific and physiological interactions, further analysis of ␥-adaptin-pIgR interactions was performed using this approach.  3. Immunoisolation of pIgRcontaining membranes. A PNS fraction was prepared from the following: MDCK cells that do not express the pIgR; MDCK cells that express the wild-type receptor (A); or from MDCK cells that express the pIgR mutants pIgR-⌬655-668, pIgR-S726-A, or pIgR-R654stop (B). Purified anti-tail monospecific antibodies were reacted with streptavidin-coupled magnetic beads pre-coated with biotinylated rabbit anti-mouse IgG. Immunobeads were exposed to equal protein amounts of PNS in the cold and washed to remove unbound membranes, and bound (immunoisolated) membranes were solubilized in Triton X-100-containing buffer, mixed with SDS sample buffer, and subjected to SDS-PAGE followed by immunoblotting with anti-pIgR or anti-␥-adaptin antibodies. Signals were visualized by enhanced chemiluminescence. For quantitative determination of band intensity, exposures were always in the linear range, and the mean band density was quantified using the NIH Image program. Intensity levels contributed by ␥-adaptins were divided by those contributed by the pIgR, and the ratio determined for each pIgR mutant was calibrated to the ratio measured for the wild-type receptor (B, lower panel). Two different MDCK clones expressing each pIgR construct were analyzed in three separate experiments. Results are mean Ϯ S.E.
BFA Inhibits AP-1 Association with the pIgR-To characterize further the specificity and physiological relevance of pIgR-AP-1 adaptor interactions, we examined whether cell treatment with BFA will reduce the efficiency of ␥-adaptin coimmunoprecipitation. Cells expressing the wild-type pIgR were either not treated (ϪBFA) or treated with BFA (ϩBFA) and were subsequently subjected to pIgR immunoprecipitation. Approximately equal amounts of pIgR have been immunoprecipitated from each preparation (as determined by quantitative immunoblotting; see Fig. 5B, upper panel). The amount of co-immunoprecipitated ␥-adaptin was quantified as well and normalized to the amount of precipitated pIgR (Fig. 5B, lower  panel). Compared with untreated cells, a reduction of approximately 50% in co-immunoprecipitated ␥-adaptin was observed in BFA-treated cells (Fig. 5B, lower panel). BFA treatment has no effect on total ␥-adaptin levels in the cells, indicating that the reduction in co-immunoprecipitated coat protein is not due to fluctuations in expression level of endogenous ␥-adaptin. The BFA-resistant interactions could be explained by previous data suggesting that even after long periods of exposure to BFA, a small amount of ␥-adaptin remains concentrated in the perinuclear region of MDCK cells, possibly representing ␥-adaptin that did not dissociate from the Golgi apparatus (32). 2 Nevertheless, the reduction in ␥-adaptin co-immunoprecipitation by BFA argues that BFA-sensitive interactions between pIgR and ␥-adaptin are detected by this approach, reflecting specific and physiologically relevant interactions between these molecules in the cell.  ϩ cross-linking). In one experiment cells were not treated with the cross-linker (Ϫ cross-linking). After quenching with glycine, cells were solubilized in buffer containing Triton X-100, and the pIgR was immunoprecipitated from the lysate using sheep anti-pIgR antibodies. In a control experiment, cells exposed to the cross-linker were subsequently incubated with plain protein A-Sepharose (Protein A-Seph). Immunoprecipitates were separated by SDS-PAGE, immobilized on a nitrocellulose membrane, and probed with sheep anti-pIgR antibodies, mouse monoclonal anti-␥-adaptin 100/3 antibodies, and rabbit monospecific antibodies directed against peptides derived from the 1, ␤1, and 1 subunits of the AP-1 complex.
FIG. 5. ␥-Adaptin co-immunoprecipitates (Co-IPs) with the pIgR in a BFA-sensitive fashion. A, MDCK cells expressing the wild-type pIgR were solubilized in the cold with buffer containing Triton X-100. Triton-insoluble materials were removed by high speed centrifugation. Solubilized proteins were subjected to immunoprecipitation (IP) with anti-pIgR antibodies. Negative controls include lysates exposed to protein A-Sepharose (Protein A Seph) or to protein A-Sepharose coupled to non-relevant sheep IgG. Immunocomplexes were analyzed by SDS-PAGE and immunoblotting for the presence of pIgR (Anti-pIgR) and ␥-adaptins (Anti-␥-adaptin) (upper panel). Note that immunoprecipitated pIgR often appears as two major bands, one at 110 kDa represents the intact receptor and the other at 85 kDa represents cell-associated SC. In the lower panel, the relative efficiencies of coimmunoprecipitation of ␥-adaptin with pIgR in the absence (Co-IP) and presence of chemical cross-linker (Cross-linking) is also shown. B, MDCK cells expressing the wild-type pIgR were either treated (ϩ) or not treated (Ϫ) with 10 g/ml BFA for 1 h at 37°C. Cells were lysed in ice-cold Triton X-100 containing buffer, and subjected to pIgR immunoprecipitation. Immunoprecipitates and 5-l sample from each lysate were subjected to SDS-PAGE and Western blotting, which were subsequently probed with anti-pIgR (Anti-pIgR, upper panel) or anti-␥-adaptin antibodies (Anti-␥-adaptin, lower panel). Approximately equal amounts of pIgR were immunoprecipitated by protein A-Sepharose coupled to sheep anti-pIgR IgG, as judged by quantifying pIgR band intensity (pIgR level, arbitrary units (a.u.), upper panel). A reduction of approximately 50% in the amount of co-immunoprecipitated ␥-adaptin was observed in BFA-treated cells relative to untreated cells, after normalizing the amount of co-precipitated adaptins to the level of precipitated pIgR (lower panel). This result is the mean of three experiments. Representative results are shown. of the pulse-labeled receptor form to digestion with endoglycosidase H (41). Upon a subsequent chase at 18°C, pulse-labeled receptors accumulate in the TGN and become resistant to digestion by endoglycosidase H (41). To chase metabolically labeled receptor into post-TGN compartments, cells were pulselabeled and chased for 30 min at 37°C or cells were pulselabeled, chased at 18°C, and subsequently chased for 10, 30, or 60 min at 37°C. Cells expressing 35 S-pIgR in the ER, in the TGN, or in post-TGN compartments were solubilized, and ␥-adaptin was immunoprecipitated. Immunoprecipitates were analyzed for the presence of co-precipitated 35 S-pIgR. Fig. 6A shows that pulse-labeled receptors are not co-immunoprecipitated with ␥-adaptin, confirming that newly synthesized receptors in the ER are not associated with AP-1 adaptors. In contrast, pIgR accumulated in the TGN appears to coimmunoprecipitate with AP-1 adaptors. Similarly, 35 S-pIgR chased into post-TGN compartments co-immunoprecipitates with ␥-adaptin. These results suggest that the coat protein interacts with the pIgR after its delivery to the TGN and that these interactions persist after pIgR exit from the TGN and/or are either re-initiated in or persist in post-TGN compartments. After 30 and 60 min chase, the majority (75%) of the 35 S-pIgR cohort reaches the basolateral surface (19,20,27), 2 and a significant fraction of the surface-arriving molecules is likely internalized into endosomal and transcytotic elements. Note that the signal contributed by 35 S-pIgR co-immunoprecipitated following 30 min chase at 37°C is significantly greater than the signal contributed by 35 S-pIgR after "TGN block" (Fig. 6A), suggesting that pIgR-␥-adaptin interactions could be augmented when 35 S-pIgR is present in post-TGN compartments. An alternative explanation is that incubation at 37°C increases the affinity of pIgR interactions with the adaptor com-plex or merely enhances the efficiency of pIgR labeling with the radioactive amino acids.
To address this point, cells were pulse-labeled and incubated at 18°C, or pulse-labeled, incubated at 18°C and then subsequently chased for 10, 30, or 60 min at 37°C. After 10 min of chase, metabolically labeled pIgR accumulated in the TGN barely reaches the basolateral cell surface, yet the temperature shift did not result in significant changes in the amount of co-precipitated 35 S-pIgR compared with pIgR blocked in the TGN (Fig. 6B, upper panel), thus diminishing the possibility that higher co-immunoprecipitation levels observed for 35 S-pIgR after chase at 37°C is due to more efficient metabolic labeling of pIgR at 37 versus 18°C. Interestingly, however, after incubation at 18°C and chase for 30 or 60 min at 37°C, higher levels of 35 S-pIgR appeared to co-immunoprecipitate with the coat protein (Fig. 6B, lower panel). Co-immunoprecipitation of the pIgR and ␥-adaptin seems to be specific since metabolically labeled receptors were not immunoprecipitated if protein A-Sepharose coupled to irrelevant IgG2b (the 100/3 monoclonal antibody is of mouse IgG2b isotype) were used (see Control in Fig.  6A). Since after 30 or 60 min chase at 37°C, the majority of labeled receptors likely reached the basolateral surface and endocytic/post-endocytic/compartments, these data indicate that pIgR-AP-1 interactions are enhanced in these compartments.
Finally, it is important to note that only a small fraction (Ͻ10%) of metabolically labeled pIgR has been co-immunoprecipitated with ␥-adaptin. We believe that this figure does not necessarily indicate that only a small fraction of pIgR interacts with AP-1 adaptors. Rather, inefficient co-immunoprecipitation may reflect a limitation of this experimental protocol in detecting the real fraction of pIgR molecules that interacts with ␥-adaptin, as previously discussed.

␥-Adaptin Interacts with Reduced Efficiency with Apically
Targeted pIgR and with pIgR Mutated in Ser-726 -We employed co-immunoprecipitation to examine directly the level of interactions between pIgR-⌬655-668 (Fig. 7A) or pIgR-S726-A (Fig. 7B) and ␥-adaptin. The amount of co-precipitated adaptins was normalized to the levels of immunoprecipitated pIgR, and the relative association of ␥-adaptin with pIgR mutants was quantitatively determined with respect to the wildtype receptor (Fig. 7C). To confirm the specificity of ␥-adaptin co-immunoprecipitation with pIgR mutants, co-immunoprecipitation from MDCK cells that do not express the pIgR (MDCK), or from cells that express the wild-type pIgR (pIgR-WT), or from cells expressing the pIgR-R654stop mutant was conducted in parallel as negative and positive controls. ␥-Adaptin could not be efficiently co-immunoprecipitated with pIgR-R654stop or with pIgR-S726-A mutants. The amount of ␥-adaptin present in cell lysates was equal, indicating that the reduction in co-immunoprecipitated coat protein is not due to fluctuations in endogenous expression levels of ␥-adaptin in the different clones. This latter result, which is in agreement with the co-immunoisolation data shown above (Fig. 3), argues that apically targeted pIgR-R654stop or basolaterally targeted receptors with inactivated Ser-726 largely avoid interactions with AP-1 adaptors. Yet, pIgR-⌬655-668, which is also apically targeted, seems to maintain higher levels of interaction with ␥-adaptin than the other mutant pIgRs. This result supports the contention that Ser-726 is involved in AP-1 association since pIgR-⌬655-668 contains an intact Ser-726 motif that is potentially free to interact with AP-1 complexes. Previous in vitro binding experiments indeed demonstrated that unlike the wild-type influenza HA (an apical protein in MDCK cells), which does not efficiently interact with AP-1 adaptors, a mutant HA bearing C-terminal di-leucine sequence is capable of interacting with purified AP-1 adaptors (39). Hence, apically targeted pIgR-⌬655-668 may still maintain significant degree of interactions with the coat protein, which suggests that AP-1 FIG. 6. 35 S-pIgR is co-immunoprecipitated with ␥-adaptins after its accumulation in the TGN. A, filter-grown MDCK cells expressing the wild-type pIgR were either pulse-labeled for 8 min at 37°C with [ 35 S]cysteine and methionine alone, pulse-labeled and subsequently chased for 120 min at 18°C (TGN Block), or pulsed and chased for 30 min at 37°C. Cells on filters were lysed in cold buffer containing Triton X-100, and ␥-adaptin was immunoprecipitated using the 100/3 monoclonal antibodies coupled to protein A-Sepharose. Immunoprecipitated complexes were released from the beads by boiling the beads in SDS, and the pIgR was re-immunoprecipitated from the released protein mixture by anti-pIgR antibodies coupled to protein A-Sepharose. In the control experiment, lysates prepared from cells that have been pulse-labeled and chased for 30 min at 37°C were incubated with irrelevant mouse IgG2b that does not precipitate ␥-adaptin and protein A-Sepharose. Immunoprecipitates were analyzed by SDS-PAGE, and autoradiography of co-immunoprecipitated pIgR is shown. B, PIgRexpressing MDCK cells were pulse-labeled and subsequently incubated under conditions that block the pIgR in the TGN. Cells were either not chased (0 min) or chased for 10, 30, or 60 min at 37°C. Cells were then subjected to immunoprecipitation of ␥adaptin and analyzed for co-immunoprecipitated 35 S-labeled pIgR as above. Total amounts of radiolabeled pIgR after each chase period were immunoprecipitated from parallel samples and analyzed by SDS-PAGE, and autoradiography is presented. The amount of 35 S-pIgR co-immunoprecipitated after 10 min chase at 37°C is identical to 35 S-pIgR co-precipitated after TGN block (upper panel). In contrast, compared with the amount of 35 S-pIgR coprecipitated with ␥-adaptin after TGN block, the level of co-precipitated 35 S-pIgR is about 3.5-fold greater after TGN block and chase for 30 or 60 min at 37°C (lower panel). Note that the levels of total 35 S-pIgR immunoprecipitation after 30-and 60-min chase periods is reduced compared with the amount of 35 S-pIgR immunoprecipitated after TGN block. This reduction is attributed to 35 S-pIgR transcytosis and arrival at the apical surface where it is cleaved to soluble SC. adaptors may still be involved in certain steps of apical targeting of the pIgR mutant.
Phosphorylated pIgR Co-immunoprecipitates with ␥-Adaptin-If phosphorylation of Ser-726 is important for binding to AP-1 adaptors, one would expect that phosphorylated receptors would associate with the coat protein. To address this point, cells were labeled with [ 32 P]orthophosphate to steady-state levels and lysed in Triton-containing buffer, and AP-1 adaptors were immunoprecipitated via ␥-adaptin. The amount of co-immunoprecipitated 32 P-labeled pIgR was assessed by SDS-PAGE and autoradiography quantified by densitometric analysis after normalizing the amount of co-immunoprecipitated 32 P-pIgR to the total amount of 32 P-pIgR immunoprecipitated from the cell lysates. 32 P-pIgR co-immunoprecipitated with ␥-adaptin (Fig. 8, upper panel). Less than 5% of total labeled pIgR coprecipitated with ␥-adaptin. Beads coupled to irrelevant mouse FIG. 7. ␥-Adaptin interacts with the cytoplasmic tail of the pIgR, possibly through the Ser-726 motif. A, the pIgR was immunoprecipitated from the following: MDCK cells; from MDCK cells that express the wild-type receptor; from cells that express the pIgR-⌬655-668 mutant; or from cells that express the pIgR-R654stop mutant. B, the pIgR was immunoprecipitated from cells that express the wild-type pIgR or cells that express the pIgR-S726-A mutant. In all cases comparable amounts of the pIgR were immunoprecipitated under conditions described in Fig. 4 and under "Experimental Procedures." Immunoprecipitated pIgR and coimmunoprecipitated ␥-adaptin were visualized by Western blotting followed by enhanced chemiluminescence, and representative exposures are presented. The experiment was performed at least six times on three different MDCK clones expressing the wild-type pIgR and on two different cell lines expressing each of the indicated pIgR mutants. Band intensity of ␥-adaptin was normalized to the intensity of precipitated pIgR. C, the ␥-adaptin/ pIgR ratio obtained for each mutant receptor is calibrated to that of the wildtype receptor. Data are presented as mean Ϯ S.E.
IgG2b did not immunoprecipitate ␥-adaptin nor co-precipitate the pIgR (not shown). BFA treatment significantly reduced the level of co-precipitated 32 P-pIgR (Fig. 8). 32 P-pIgR-␥-adaptin interactions with the coat protein also seem to be inhibited by mutational inactivation of Ser-726 as a relatively smaller fraction of 32 P-labeled pIgR-S726-A co-immunoprecipitated with ␥-adaptin (Fig. 8). Together, the results indicate that ␥-adaptin interact with phosphorylated wild-type pIgR in a BFA-sensitive manner and that Ser-726 is involved in these interactions.
However, the persistence of a fraction of phosphorylated S726-A mutant pIgR that co-immunoprecipitates with ␥-adaptin suggests that whereas Ser-726 is apparently critical for the association of AP-1 adaptors with the pIgR, it may not be the sole factor in this interaction. Since Ser-664, the residue that regulates the constitutive transcytosis of the pIgR (26), is likely to be the phosphorylated residue in the S726-A mutant pIgR, these data raise the possibility that phosphorylated pIgR in the constitutive transcytotic pathway may also interact with AP-1 adaptors.

The pIgR S726-A Mutant Is Delivered to the Basolateral Surface with Slower Kinetics Than the Wild-type pIgR and via a BFA-insensitive
Route-It has been previously demonstrated that mutation of Ser-726 to Ala does not impair basolateral targeting of the pIgR (43). Since the same mutation seems to impair pIgR interactions with ␥-adaptin, we conclude that the interactions are not involved in basolateral targeting of the molecule. We have reasoned, however, that interactions with ␥-adaptin may play a regulatory role in exocytosis of pIgR from the TGN. Thus, the pIgR-S726-A might display basolateral delivery kinetics that differ from those exhibited by wild-type pIgR. The results in Fig. 9, A and B, indeed reveal that pIgR-S726A is directly delivered to the basolateral surface with slower kinetics than the wild-type receptor. In addition, unlike the case for the wild-type pIgR, whose biosynthetic pathway is inhibited by BFA, basolateral delivery of newly synthesized pIgR-S726-A is not inhibited by the drug. These results further emphasize the importance of Ser-726 in mediating the interactions with AP-1 adaptors and that interactions with AP-1 adap- FIG. 8. ␥-Adaptin interacts with 32 P-pIgR. MDCK cells expressing the wildtype pIgR or pIgR-S726-A were labeled with [ 32 P]orthophosphate; equal amounts of ␥-adaptin were immunoprecipitated, and pIgR was re-precipitated from the immunocomplexes as described in Fig. 5 and under "Experimental Procedures." Total amount of 32 P-pIgR was estimated by immunoprecipitating the entire 32 P-pIgR population from the lysates. In one experiment, cells that express the wild-type receptor were treated with 10 g/ml BFA for 1 h at 37°C prior to immunoprecipitation. Immunoprecipitates were analyzed by SDS-PAGE, and representative autoradiograms are presented in the upper panel. Note that band intensity of co-precipitated pIgR-WT appears to be greater than the corresponding band of total 32 P-pIgR in cell lysate. This apparent discrepancy is due to much shorter autoradiogram exposure times for total radiolabeled pIgR as indicated under "Experimental Procedures." Band intensity was quantitatively determined in five independent experiments. The signal of co-immunoprecipitated 32 P-pIgR was normalized to the signal contributed by total 32 P-pIgR, and the ratio obtained for each experiment was calibrated with respect to pIgR-WT (lower panel). Results are mean Ϯ S.E. tors could elicit a BFA-sensitive exocytic step that concentrates pIgR molecules into AP-1/clathrin-coated areas of the TGN. This process may result in efficient delivery of pIgR molecules from the TGN to the basolateral surface.
The data presented in Fig. 9B also indicate that apical delivery of pIgR-⌬655-668 and pIgR-R654stop is significantly inhibited by BFA. These results are consistent with previous findings that, whereas BFA may not inhibit exocytosis of some basolaterally targeted membrane proteins, in all cases reported so far, BFA inhibited exocytosis of apical membrane proteins. In addition, this result is intriguing as it suggests that, particularly in the case of pIgR-R654stop, interactions between pIgR cytoplasmic tail and BFA-sensitive coat proteins are not required for mediating apical targeting. Apical transport of transmembrane proteins occurs either by proteinaceous determinants (15) or is mediated by widely distributed carbohydrate determinants (13,14,55,56). It is hence possible that BFAsensitive coat proteins are general regulators of trafficking of apical membrane proteins. An alternative scenario is that unlike basolateral membrane proteins, transmembrane proteins destined to the apical surface all pass through a common compartment whose sorting activity is impaired by BFA. DISCUSSION Evidence from several laboratories now strongly argues that both AP-2 and AP-1 clathrin adaptors recognize tyrosine-based signals conforming to the YXXØ type cytoplasmic signals (where X is any amino acid and Ø is an amino acid with bulky hydrophobic group), which mediate endocytosis from the plasma membrane and may also confer intracellular sorting in the TGN (30). Intact AP-1 and AP-2 adaptors have been shown to interact in vitro with an artificially created Tyr-based basolateral sorting signal that also functions in endocytosis (39). Nonetheless, the functional role of these interactions in polarized sorting remains speculative. Other sorting signals, such as the acidic cluster comprising the CKII site (ESEER) juxtaposed to a di-leucine motif in the CD-MPR cytoplasmic tail is also essential for high affinity binding of AP-1 adaptors in vitro (45) and therefore probably acts as a dominant determinant controlling CD-MPR sorting in the TGN. The cytoplasmic tail of FIG. 9. Kinetics and BFA sensitivity of delivery of wild-type pIgR, pIgR-S726-A, pIgR-⌬655-668, and pIgR-R654stop from the TGN to the basolateral and apical cell surface. A, delivery of nascent receptors to the respective cell surface was measured over 15-, 30-, and 60-min chase periods at 37°C using the protease-based delivery assay, as described under "Experimental Procedures" and elsewhere (19,20). B, targeting of receptors to the apical and basolateral surfaces was measured after 60 min chase at 37°C in the absence (Ϫ) or presence (ϩ) of BFA. In BFA-treated cells, BFA (10 g/ml) was included in the media bathing both the apical and basolateral surfaces of the cells throughout the starvation, pulse, and chase steps (42). Results are mean Ϯ S.E. of six independent experiments performed on two different clones expressing each pIgR construct. the pIgR appears to possess putative AP-1-binding motifs that resemble the known Tyr-based motifs and the CKII/di-leucine motifs (Fig. 1). In this study we investigated the interactions between ␥-adaptin and the pIgR, and we asked whether a cytoplasmic Ser-726, residing in a putative CKII/PKA phosphorylation site upstream to a di-leucine motif, is important for these interactions. Ser-726 is required for rapid internalization of the pIgR from the basolateral plasma membrane of MDCK cells but not for basolateral targeting from the TGN (43). We investigated whether Ser-726 plays a role in the mechanism of pIgR exocytosis from the TGN to the basolateral cell surface by virtue of its interactions with the AP-1 clathrin adaptor.
We have used four different methodologies (co-localization, co-immunoisolation, cross-linking, and co-immunoprecipitation) to address the possibility that the pIgR interacts with the AP-1 adaptor complex in MDCK cells. Our data suggest that the entire AP-1 adaptor complex interacts directly, or indirectly (namely via other proteins), with the cytoplasmic tail of the pIgR. These interactions are inhibited by BFA, suggesting that pIgR-␥-adaptin interactions might be regulated by the small GTPase ARF1 (57). Hence, like in the case of MPRs bound to lysosomal enzymes, specific determinants in the cytoplasmic tail, perhaps more particularly a CKII phosphorylation site, bind AP-1 adaptors, a process that leads to pIgR sorting in the TGN into clathrin-coated vesicles.
Immunofluorescent analysis of pIgR and ␥-adaptin suggests that at steady-state the pIgR partially co-localized with ␥-adaptin, possibly on the surface of the trans-Golgi network. Our co-immunoprecipitation data further suggest that the pIgR interacts with ␥-adaptin and that these interactions are initiated in the Golgi since metabolically labeled pIgR could be co-immunoprecipitated with ␥-adaptin only after pIgR accumulation in the TGN. Interestingly, however, the co-immunoprecipitation experiments also suggest that ␥-adaptin may interact with pIgR located in post-TGN compartments, possibly in basolateral endosomes or transcytotic compartments. This interpretation is consistent with previous studies showing the association of ␥-adaptin with endosomes in non-polarized cells (58) and with basolateral endosomes containing internalized transferrin or dIgA in polarized cells (50).
With respect to the regulation of the putative interaction of pIgR with AP-1 adaptors, we currently do not know whether Ser-726 phosphorylation occurs in the TGN or in post-TGN compartments, but our data clearly suggest that Ser-726, and perhaps its phosphorylated form, is an important element contributing to the establishment of pIgR-␥-adaptin interactions, in addition to regulating the interaction of pIgR with endocytotic machinery (38). Phosphorylation of Ser-726 may therefore serve as a bifunctional biological switch that prompts the interactions with either AP-1 adaptors in the TGN or AP-2 adaptors at the basolateral membrane. Alternatively, the detection of pIgR-AP-1 interactions in the TGN presented here may be a consequence of "cross-talk" with a degenerate sorting signal for AP-2 adaptors regulated by Ser-726 or vice versa. Even if phosphorylated Ser-726 is not directly involved in the binding of adaptors, phosphorylation may cause conformational changes in the cytoplasmic tail which expose other motifs, such as the downstream di-leucine motif, for interactions with the coat protein, as was recently proposed for the CD3␥ chain of the T cell receptor (59).
Interactions between the wild-type cytoplasmic domain and the 1 chain could not be resolved at the resolution provided by the yeast two-hybrid assay. 3 Thus, a distinct subunit of the AP-1 complex may interact with the cytoplasmic tail. Although 2 and 1 chains recognize Tyr-based signals (60) and some leucine-based signals (61,62), the adaptor subunit that interacts with Ser-based determinants has not yet been identified. Recent studies have even suggested the N-terminal trunk domain of the ␥-subunit of membrane bound AP-1 adaptors to mediate the interactions with the bovine papillomavirus E6 protein (63), indicating that adaptin subunits other than medium chains may mediate interactions with signal motifs. Of course, an alternative scenario is that other proteins may contribute to the interactions between AP-1 and the cytoplasmic domain of pIgR, such as the recently cloned PACS family of TGN-associated coat proteins involved in the trafficking of furin and MPRs (64).
BFA inhibits the binding of the coat protein and basolateral delivery of the pIgR (41,42). BFA treatment does not cause receptor mistargeting to the apical surface, nor does the Ser-726 to Ala mutation, which also inhibits receptor association with the coat protein. Thus, basolateral sorting of the pIgR is probably determined by other cytosolic factors that may interact with the 17-residue basolateral sorting signal. Consistent with this are recent studies by Distel et al. (65), demonstrating that deletion of the AP-1-binding site in a chimeric protein made of the luminal domain of the influenza virus hemagglutinin and the cytoplasmic tail of CD-MPR does not affect its polarized sorting to the basolateral surface in MDCK cells. The basolateral sorting machinery may decode the basolateral sorting signal and incorporate the pIgR into the basolateral pathway irrespective of the interactions of pIgR with AP-1. Another possibility is that basolateral sorters interact with the basolateral sorting signal of pIgR in post-TGN compartments, for example after possible targeting of exocytic receptors from the TGN to endosomes (see below). Consistent with this hypothesis is the characterization of the mammalian Sec6/Sec8 homologues involved in the regulation of basolateral targeting of membrane proteins in MDCK cells (66).
The analogy to the MPR raises the hypothesis that the putative CKII/PKA phosphorylation site and the juxtaposed dileucine motif mediates pIgR targeting from the TGN to endosomes in the course of exocytosis. As indicated before, classical plasma membrane proteins have been suggested to pass through endosomes en route to the cell surface in non-polarized cells (22,23), but the involvement of AP-1 binding or cytoplasmic motifs in this process has not been investigated. When Ser-726 is inactivated by mutation to Ala, interactions with the AP-1 are compromised, and receptor exocytosis is probably mediated by alternative mechanisms that do not involve AP-1 and targeting to endosomes. In this context it is worth pointing out that exposure of cells to high BFA concentrations did not result in complete inhibition of pIgR delivery to the basolateral cell surface (42). This result suggests that the pIgR utilizes parallel biosynthetic routes to reach the basolateral cell surface; one route depends on AP-1 binding and is BFA-sensitive (and possibly involves passage through endosomes), and the other route is ␥-adaptin independent and BFA-insensitive (and possibly bypasses the endosomal system).
The physiological significance of multiple redundant pathways is not clear, but it may provide means for regulation of protein expression by a given organelle. The existence of mechanistically distinct alternative pathways may also serve as a salvage mechanism to maintain membrane trafficking to the cell surface in cases where one of the pathways is paralyzed, e.g. as a result of cell exposure to toxins. It is worth noting that two distinct transport pathways for soluble and membranebound proteins from the Golgi to basolateral plasma membrane of liver cells have been reported by Boll et al. (67), although the coat proteins regulating these pathways have not been identi-fied. Another interesting case of two parallel pathways has been described for the sorting of hydrolases to the vacuole in yeast; one of these pathways is regulated by the recently discovered adaptor-like AP-3 coat protein (68). Multiple redundant pathways for protein delivery thus appear to exist in higher as well as in lower eukaryotes.
Finally, our results provide significant insight into physiological mechanisms regulating Golgi-to-basolateral surface transport, as they suggest that interaction of membranes and cargo with AP-1 adaptors might be a target for cellular signals regulating constitutive vesicle formation and membrane traffic in polarized cells. Signal transduction processes are suggested to be involved in controlling clathrin adaptor and non-clathrin coat recruitment on membranes (69 -72). For example, phospholipase D, a phospholipid-hydrolyzing enzyme whose activation has been implicated in signal transduction pathways, cell growth, and membrane trafficking (73), stimulates the release of nascent secretory vesicle budding from the TGN (70). ARF1 has been shown to stimulate endogenous phospholipase D activity, a process that correlated with enhancement of vesicle budding. A possible model to explain these data suggests that hydrolysis of phosphatidylcholine in the TGN mediated by ARF-activated phospholipase D results in the production of high local concentration of phosphatidic acid; this process could alter local composition and physical properties of the lipid bilayer in a manner that facilitates the recruitment of AP-1 adaptors or other coats, resulting in the budding of nascent secretory vesicles from the TGN (71). Superimposed upon this regulatory mechanism, modulation of phosphorylation levels of Ser-726 may determine the extent of pIgR sequestration into coated areas and consequently determine the kinetics of its appearance on the basolateral surface. This process can indirectly affect its subsequent transcytosis. Further characterization of the regulation of the trafficking of the pIgR in other cell types, including those found in secretory organs, may provide additional insight into the physiological role(s) of Ser-726 in pIgR trafficking.