Calmodulin binds to the basolateral targeting signal of the polymeric immunoglobulin receptor.

We have identified a major calmodulin (CaM)-binding protein in rat liver endosomes using I-CaM overlays from two-dimensional protein blots. Immunostaining of blots demonstrates that this protein is the polymeric immunoglobulin receptor (pIgR). We further investigated the interaction between pIgR and CaM using Madin-Darby canine kidney cells stably expressing cloned wild-type and mutant pIgR. We found that detergent-solubilized pIgR binds to CaM-agarose in a Ca-dependent fashion, and binding is inhibited by the addition of excess free CaM or the CaM antagonist W-13 (N-(4-aminobutyl)-5-chloro-2-naphthalenesulfonamide), suggesting that pIgR binding to CaM is specific. Furthermore, pIgR is the most prominent S-labeled CaM-binding protein in the detergent phase of Triton X-114-solubilized, metabolically labeled pIgR-expressing Madin-Darby canine kidney cells. CaM can be chemically cross-linked to both solubilized and membrane-associated pIgR, suggesting that binding can occur while the pIgR is in intact membranes. The CaM binding site is located in the membrane-proximal 17-amino acid segment of the pIgR cytoplasmic tail. This region of pIgR constitutes an autonomous basolateral targeting signal. However, binding of CaM to various pIgR mutants suggests that CaM binding is not necessary for basolateral targeting. We suggest that CaM may be involved in regulation of pIgR transcytosis and/or signaling by pIgR.

The plasma membrane of polarized epithelial cells is divided into apical and basolateral surfaces, each having a distinct protein and lipid composition. Cells use two pathways to deliver proteins to the correct surface. First, newly made proteins are packaged in the trans-Golgi network (TGN) 1 into vesicles that deliver them directly to either the apical or basolateral surface. Second, proteins can be delivered first to one surface and then endocytosed and transcytosed to the opposite surface. At least in the well polarized Madin-Darby canine kidney (MDCK) cell line, targeting to the basolateral surface generally requires a sorting signal located in the cytoplasmic domain of the protein. For a number of basolateral proteins it has been found that mutations in the cytoplasmic domain prevent TGN to basolateral targeting (1). The first basolateral signal to be identified is the 17-amino acid membrane proximal segment (residues 653-670) of the cytoplasmic domain of the polymeric immunoglobulin receptor (pIgR) (2). This protein is normally delivered from the TGN to the basolateral surface, and then endocytosed and transcytosed to the apical surface. Deletion of most of this 17-residue segment prevents TGN to basolateral delivery. Moreover, this segment can be transplanted to a heterologous reporter molecule, which is then retargeted from the apical to the basolateral surface (2). In at least one other case, the low density lipoprotein receptor, autonomous basolateral sorting signals have been identified (3).
The pIgR basolateral targeting signal has been systematically analyzed by alanine scanning mutagenesis (4). Mutation of His-656, Arg-657, or Val-660 to Ala substantially diminishes, but does not eliminate, TGN to basolateral targeting. Mutation of other residues had little or no effect. The structure of a synthetic 17-residue peptide corresponding to this signal has been determined by two-dimensional nuclear magnetic resonance spectroscopy (4). This peptide tends to adopt a putative type 1 ␤-turn, encompassing residues 658 -661, followed by a nascent helical structure. In MDCK cells polarized sorting takes place in both the TGN and after endocytosis. Mutations that diminish TGN to basolateral sorting and increase TGN to apical sorting have a similar effect on sorting in the endocytotic pathway, decreasing recycling to the basolateral surface and increasing transcytosis to the apical surface (5). These data indicate that the basolateral targeting signal also functions as a signal to retrieve the pIgR from the endocytotic pathway to the basolateral surface. Very similar results have been obtained with the two basolateral sorting signals of the low density lipoprotein receptor (6). This suggests that the same sorting machinery operates in both the TGN and the endocytotic pathway, or that sorting in both pathways actually 1 The abbreviations used in this paper are: TGN, trans-Golgi network; pIgR, polymeric immunoglobulin receptor; CaM, calmodulin; d/pIgA, dimeric/polymeric immunoglobulin A; VSVG, vesicular stomatitis virus protein G; BS 3  occurs in a common compartment.
Transcytosis of pIgR is regulated by several mechanisms. Phosphorylation of Ser664, which is part of the basolateral sorting signal, stimulates transcytosis. Mutation of Ser-664 to Ala (S664A) reduces transcytosis, while mutation to Asp (S664D), whose negative charge may mimic a phosphate, stimulates transcytosis (7). Phosphorylation apparently works by weakening the basolateral signal, since the S664D mutant exhibits decreased TGN to basolateral sorting, increased TGN to apical sorting (5), as well as decreased recycling and increased transcytosis after endocytosis.
Binding of the ligand, dimeric IgA (dIgA) to the pIgR also stimulates transcytosis. Ligand binding stimulates transcytosis of the wild-type pIgR, as well as S664D and S664A (8,9). Ligand-dependent stimulation suggests that the pIgR may transduce a signal to the cytoplasm. Indeed, we have recently found that dIgA binding causes tyrosine phosphorylation of a phosphatidylinositol-specific phospholipase C␥1, activation of protein kinase C, and production of inositol 1,4,5-trisphosphate. 2 Transcytosis of pIgR and other molecules is stimulated either by activation of protein kinase C (10) or by increase in intracellular free Ca 2ϩ , 3 so both arms of the phospholipase C signaling pathway may redundantly stimulate transcytosis. Delivery to the apical surface, either by transcytosis or directly from the TGN, is also stimulated by the heterotrimeric G protein, G s (11,12), as well as cAMP and protein kinase A (13,14). Both the G s ␣ and G␤␥ subunits are stimulatory, at least for transcytotic apical delivery (11). Transcytosis in MDCK cells is stimulated by bradykinin, 3 while in animals transcytosis is stimulated by several hormones and neurotransmitters (15)(16)(17)(18). Presumably, these extracellular signals work through the various second messengers mentioned above.
Transcytosis of the pIgR can be divided by both morphological and biochemical assays into at least three steps. Step 1 is internalization via clathrin-coated pits into basolateral early endosomes.
We are interested in molecules that bind to the cytoplasmic domain of the pIgR, particularly its basolateral sorting signal, as such molecules might be involved in basolateral sorting and/or regulation of traffic. The best evidence for a candidate molecule capable of recognizing and deciphering a basolateral targeting signal is a protein of ϳ220 kDa apparent molecular mass that has been cross-linked to the cytoplasmic domain of the vesicular stomatitis virus G protein (VSVG) (23). A peptide corresponding to the cytoplasmic domain of VSVG prevented cross-linking and also inhibited TGN to basolateral transport of VSVG in permeabilized MDCK cells, while a peptide corresponding to a mutant VSVG that is not basolaterally targeted prevented neither cross-linking nor basolateral transport. However, a peptide corresponding to the pIgR 17-amino acid basolateral targeting signal did not inhibit basolateral transport of VSVG (23), suggesting that this protein might not recognize the pIgR basolateral targeting signal.
Studies on MDCK cells using CaM antagonists suggest that CaM is important in endosome function (24,25). We now report that in the presence of Ca 2ϩ , calmodulin (CaM) binds to the basolateral sorting signal of the pIgR. An analysis of mutant pIgRs reveals that the sequence requirements for basolateral targeting of the pIgR and CaM binding are not identical, suggesting that the role of CaM binding is not to target the pIgR to the basolateral surface, but rather may be involved in regulation of pIgR transcytosis.

MATERIALS AND METHODS
Isolation of Rat Liver Subcellular Fractions-Three rat liver endosomal subfractions, designated early endosomes (CURL, compartment for uncoupling of receptors and ligands), late endosomes (MVB, multivesicular bodies), and a receptor recycling compartment (RRC), were isolated from estradiol-treated rats as described previously (26). Endosomal membranes were separated from contents after rupturing the organelles in a French pressure cell (26).
Antibodies-Unfractionated ascites fluid containing the monoclonal antibody SC166, which recognizes the cytoplasmic domain of the rabbit and rat polymeric immunoglobulin receptor (pIgR), was prepared as described elsewhere (27). A sheep polyclonal antibody made against rabbit secretory component (SC) has also been described (28).
Two-dimensional Electrophoresis and Western Blotting-Two-dimensional electrophoresis was carried out as described (29). After electrophoresis, proteins were transferred to nitrocellulose filters at 20 V overnight at 4°C. Filters were incubated in phosphate-buffered saline containing 5% powdered skim milk to block nonspecific sites. The nitrocellulose was then incubated with a 1/1000 dilution of the monoclonal ascites SC-166 directed against the cytoplasmic tail of the pIgR in phosphate-buffered saline containing 0.05% Tween 20, followed by horseradish peroxidase-conjugated anti-mouse IgG (Jackson Immu-noResearch Laboratories, West Grove, PA), and detection with 4-chloro-1-naphthol as described by Burnette (30).
Calmodulin Overlay Technique-For the 125 I-CaM overlay of twodimensional gels, the method described by Bachs and Carafoli (31) was used. Calmodulin, isolated as described (32) or purchased from Sigma, was radioiodinated with the Bolton-Hunter reagent (Amersham Corp.). Endosome fractions enriched in CURL and RRC show similar protein compositions and give similar results in 125 I-CaM overlays. An experiment using RRC is shown under "Results" (Fig. 1).
MDCK Clones Expressing pIgR Constructs-All nucleotide and amino acid positions are according to the original published rabbit pIgR sequence (33). MDCK strain II cells expressing wild-type pIgR (34), a mutant lacking a membrane-proximal 14-amino acid segment of the cytoplasmic tail required for basolateral targeting (pIgR ⌬655-668) (2), and a "tail-minus" construct lacking all but the two membrane-proximal amino acids of the cytoplasmic tail (2, 35) termed R655-STOP (4), have been described previously. An MDCK clone expressing pIgR in which the cytoplasmic domain consists only of the membrane proximal 17-amino acid basolateral targeting signal (T670-STOP) (4), and point mutants in which individual residues in the pIgR 17-amino acid basolateral targeting signal are replaced with alanine in the context of the entire cytoplasmic tail (H656A, R657A, N659A, and V660A) (5) have also been described. These mutants are indicated by the wild-type amino acid, followed by the position and the new amino acid (e.g. H656A designates a mutant where histidine at position 656 has been mutated to alanine). A construct encoding the full-length pIgR molecule with a substitution of alanine for arginine at position 658 was constructed by ligating a DNA fragment encoding the C terminus (residues 661-755) of wild-type pIgR clone to a fragment encoding residues 1-660 derived from the R658A/T670-STOP mutant described previously (4) via a common AvaII site at nucleotide position 2007. The resulting cDNA was ligated into the BglII site of pCB6 (36) and transfected into MDCK strain II cells as described previously (4). Mutants in which serine 664 has been mutated to alanine (S664A) or aspartic acid (S664D) have also been described (7).
Immunoprecipitation-Immunoprecipitation of pIgR was performed using protein G-Sepharose (Sigma) coupled to polyclonal sheep anti-SC IgG, essentially as described (4).
Calmodulin Affinity Chromatography-For CaM-agarose binding assays, MDCK cells expressing pIgR constructs were grown on 24-mm Transwell filters (0.4-m pore size, Costar Corp., Cambridge, MA) metabolically labeled at 37°C for 40 min on an 80-l drop of Dulbecco's modified Eagle's medium lacking cysteine (obtained from the UCSF cell culture facility, supplemented with 5% dialyzed fetal bovine serum) containing 25 Ci of [ 35 S]cysteine (DuPont NEN). The Transwell filters were washed three times with phosphate-buffered saline containing CaCl 2 and Mg 2ϩ , and filters were cut out and placed in 1 ml of ice-cold solubilization buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 containing 1% Triton X-114 (Boehringer Mann-heim) and protease inhibitors (5 g/ml pepstatin, 10 g/ml chymostatin, 5 g/ml leupeptin, 10 g/ml antipain, 500 M benzamidine, and 0.1% Trasylol) and rotated for 15 min at 4°C. Insoluble material was removed by centrifugation in a microcentrifuge for 15 min at 4°C. The supernatant was transferred to a new tube containing 6 l of 0.5 M EGTA (3 mM final). Triton X-114 phase-partitioning was performed by warming to 37°C for 3 min, followed by centrifugation at full speed for 5 min in a microcentrifuge at room temperature. The aqueous phase was discarded, and the detergent phase was resuspended with nine volumes (approximately 0.9 ml) of cold wash buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 3 mM EGTA) and incubated at 4°C for 5 min. The phase-separation procedure was repeated, followed by dilution of the detergent phase with nine volumes of ice-cold binding buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 0.5 mM CaCl 2 ). After 5 min at 4°C, the samples were precleared with 50 l of Sepharose CL-2B at 4°C (34).
The sample (500 l) was combined with 450 l of binding buffer and 50 l of a 20% slurry of CaM-agarose, which had been equilibrated with binding buffer. Samples were incubated for 1 h at 4°C on a rotator. After pelleting the CaM-agarose beads briefly in a cold microcentrifuge and aspirating the liquid, the beads were washed once with 1 ml of binding buffer containing 0.5% Triton X-100 and once with binding buffer containing 0.05% Triton X-100. At this point, a 5-min incubation was carried out at 4°C with binding buffer containing 0.05% Triton X-100, with or without 3 mM EGTA. The liquid was separated from the beads as before, and the beads were washed once with binding buffer with or without EGTA (no detergent). The beads (10 l packed) were boiled in 20 l of sample buffer, and half of the sample (ϳ10 l) was analyzed by SDS-PAGE and fluorography or phosphorimaging. Total pIgR in the starting material for CaM-agarose was determine by immunoprecipitation of a duplicate sample. Starting material, typically 500 l, was diluted with an equal volume of 2.5% Triton dilution buffer (2.5% Triton X-100, 100 mM triethanolamine-HCl, pH 8.6, 100 mM NaCl, 5 mM EGTA, 1% Trasylol, 0.02% NaN 3 ) and analyzed by pIgR immunoprecipitation, followed by SDS-PAGE and fluorography or phosphorimaging.
To correct for differences in pIgR concentrations in the starting material due to differing expression levels among clones or different partitioning behavior of constructs in Triton X-114, unlabeled preparations were analyzed by Western blotting to determine the relative concentration of each construct in the preparations. For clones that yield more pIgR in the starting material, the volume of Triton X-114 preparation added was reduced appropriately so that approximately equal amounts of each pIgR (wild-type or mutant) were added to CaM beads. The total amount of Triton-X114 starting material added was kept at 500 l by adding the appropriate volume of an identical preparation from non-transfected MDCK cells.
Cross-linking of CaM to Solubilized pIgR-Triton X-114 preparations (made as described above, except that Hepes was substituted for Tris) of metabolically labeled proteins from MDCK cells expressing pIgR constructs were incubated in the presence or absence of 1 M exogenous CaM, with or without 3 mM EGTA. After 1 h of incubation at 4°C, 0.5 mM BS 3 (Pierce) was added to the samples indicated. Incubation was continued for an additional 1 h, at which time the cross-linker was quenched by addition of glycine to 100 mM. SDS was added to a final concentration of 1%, samples were boiled for 5 min, diluted with an equal volume of 5% Triton dilution buffer (5% Triton X-100, 100 mM triethanolamine-HCl, pH 8.6, 100 mM NaCl, 5 mM EGTA, 1% Trasylol, 0.02% NaN 3 ), precleared once with 50 l of Sepharose CL-2B, and immunoprecipitated for pIgR. Precipitates were washed and analyzed by SDS-PAGE and fluorography.
Cross-linking of CaM to Crude Membranes-MDCK cells expressing pIgR constructs were grown on 15-cm dishes for 3-5 days until confluent. Cells were harvested by scraping in phosphate-buffered saline, pelleted at 150 ϫ g for 5 min at 4°C, washed once in homogenization buffer (50 mM Hepes, pH 7.4, 250 mM sucrose), and resuspended in 2 ml of homogenization buffer containing protease inhibitors by pipetting with a 1-ml micropipette tip. Cells were homogenized by 12 passes through a 22-gauge needle. A post-nuclear supernatant was prepared by centrifugation at 1400 ϫ g for 15 min at 4°C. EGTA was added to the post-nuclear supernatant to 3 mM, followed by centrifugation at 250,000 ϫ g for 1 h at 4°C (68,000 rpm in a Sorvall RP100-AT4). The pellet, containing crude membranes, was resuspended in 50 mM Hepes, 3 mM EGTA. Glycerol was added to 5%, and the membranes were frozen in liquid nitrogen and stored at Ϫ80°C. Just prior to use, samples were thawed and centrifuged at 250,000 ϫ g for 1 h at 4°C, and pellets were resuspended in HNMC (50 mM Hepes, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 0.5 mM CaCl 2 ). 125 I-CaM (prepared by the iodine monochloride method; Ref. 37), was incubated with crude membranes in HNMC for 1 h at 4°C in the presence or absence of 3 mM EGTA. BS 3 was added to a final concentration of 0.5 mM, and the incubation was continued for 1 h. Samples were quenched with glycine, boiled in SDS, and immunoprecipitated and analyzed as above.

Identification of a Rat Liver Endosome CaM-binding Protein
as pIgR-CaM has been suggested to play several possible roles in membrane traffic in mammalian cells. We and others have examined the effects of inhibitors of CaM, such as W13 and W7, on various steps in membrane traffic (24,25). In MDCK cells expressing pIgR, these drugs inhibited transcytosis of pIgR and recycling of transferrin, and caused all material endocytosed from both surfaces of the cell to be delivered to exceptionally large, abnormal endosomal structures (25), suggesting a role for CaM in polarized sorting in early endosomes. However, the mechanism(s) by which CaM regulates membrane traffic events are poorly understood.
In order to address the mechanism(s) by which CaM regulates endosomal function, we were interested in identifying major CaM-binding proteins in endosomes. Previous studies showed a major CaM-binding protein of approximately 115 kDa could be demonstrated by 125 I-CaM overlay of two-dimensional protein blots of an endosome-rich fraction from rat liver (38). We discovered that this protein can also be labeled by staining with a monoclonal antibody (SC166) against the polymeric immunoglobulin receptor (pIgR) (Fig. 1). Similar results were obtained with a rabbit antiserum raised against rat SC (data not shown).
Characterization of CaM Binding to pIgR-In order to further support the identification of pIgR as the CaM-binding protein, and to map and characterize the CaM binding site on pIgR, we developed a CaM-binding assay as described under "Materials and Methods." MDCK cells expressing the pIgR were metabolically labeled and solubilized with Triton X-114. Hydrophobic, integral membrane proteins partition into the detergent phase during Triton X-114 phase separation induced by warming to 37°C. Surprisingly, when such a preparation is added to CaM-agarose in the presence of Ca 2ϩ , followed by washing to remove nonspecific binding, pIgR is the major protein detected after the CaM beads are eluted with SDS and subjected to SDS-PAGE and fluorography ( Fig. 2A). This result is striking when one considers that pIgR is only a minor protein in the starting material ( Fig. 2A), and that there are numerous CaM-binding proteins that might be present in the Triton X-114-partitioned material. It is similarly striking that pIgR is the major CaM-binding protein detected in the rat liver membrane fraction (Fig. 1).
Binding of pIgR to CaM-agarose is inhibited by addition of EGTA ( Fig. 2A), excess CaM (Fig. 2B), or the CaM antagonist W-13 (N-(4-aminobutyl)-5-chloro-2-naphthalenesulfonamide; data not shown), further supporting the specificity of binding. If CaM binding to pIgR is physiologic, then we would expect the CaM binding site to be on the cytoplasmic domain of pIgR. To test this, we used a pIgR construct encoding the extracellular and transmembrane domains of pIgR, but lacking virtually all of the cytoplasmic tail (pIgR R655-STOP). As shown in Fig. 3, this "tail-minus" construct shows no detectable binding to CaM. We used other mutants to localize the site of CaM binding on the pIgR cytoplasmic tail (Fig. 3). Remarkably, a mutant lacking the membrane-proximal 14-amino acid segment of the cytoplasmic tail required for basolateral targeting (⌬655-668) also fails to bind CaM. Conversely, a mutant in which the cytoplasmic domain consists only of the membrane-proximal 17-amino acid basolateral targeting signal (pIgR T670-STOP) exhibits CaM binding indistinguishable from that of full-length pIgR.
Chemical Cross-linking of CaM to pIgR-CaM binding to pIgR was also demonstrated using chemical cross-linking. Metabolically labeled, Triton X-114-solubilized and partitioned material was incubated with the non-reversible cross-linker, BS 3 , in the presence of CaM. Samples were boiled in SDS, and pIgR was immunoprecipitated and analyzed by SDS-PAGE. As shown in Fig. 4, a significant fraction of wild-type pIgR is converted to a slower migrating species. This species is not observed if CaM, cross-linker, or free Ca 2ϩ are omitted from the assay. Furthermore, the apparent M r of the cross-linked complex is consistent with that expected for a stoichiometric pIgR-CaM complex (ϳ125 kDa). Specificity of the cross-linking as a CaM binding assay is further supported by the fact that mutant pIgRs lacking either the entire cytoplasmic tail (pIgR R655-STOP) or 14 amino acids of the basolateral targeting signal (⌬655-668) could not be cross-linked to CaM (Fig. 4). These results suggest that pIgR binds directly to CaM and that it can bind to soluble CaM as well as CaM-agarose.
In order to determine if CaM can bind to pIgR that is still embedded in the membrane, we performed cross-linking experiments using crude MDCK membranes containing unlabeled pIgR. Radioiodinated CaM is incorporated into a complex of 125 kDa in a Ca 2ϩ -dependent manner with membranes containing wild-type pIgR, but not with pIgR ⌬655-668 (Fig. 5) or cells not expressing pIgR (data not shown). These results attest to the likelihood of physiologic binding of pIgR to CaM in intact cells.
CaM Binding to pIgR Point Mutants Does Not Correlate with Their Basolateral Targeting Phenotype-CaM binding to the basolateral targeting signal suggested that it might play a role in pIgR targeting. In order to determine whether there is a correlation between CaM binding and basolateral targeting, we performed CaM-binding assays on many of our pIgR clones that contain alanine point mutations in the basolateral targeting signal. Typical data from these assays are shown in Fig. 6. One mutant that is not basolaterally targeted shows reduced binding to CaM (R657A), but two others that are not basolat- erally targeted (H656A and V660A) exhibit CaM binding that is comparable to wild-type pIgR. R658A, which is basolaterally targeted as accurately as the wild-type pIgR, does not bind well to CaM, while another mutant that is basolaterally targeted (N659A) does bind CaM. Thus, CaM binding does not correlate with the basolateral targeting phenotype of pIgR mutants.
CaM binding of several proteins is known to be modulated by phosphorylation (39). The pIgR has a major phosphorylation site at Ser-664 in the basolateral targeting signal, and phosphorylation of this residue stimulates transcytosis of pIgR (7,8), presumably by inactivating the basolateral targeting signal (5). We therefore assayed the ability of a non-phosphorylatable mutant, S664A, and a mutant that mimics the presence of a phosphate residue, S664D, to bind to CaM. Both of these mutants bind to CaM as well as wild-type pIgR (data not shown).
Thus phosphorylation of Ser-664 probably does not alter CaM binding.

DISCUSSION
CaM binds to and regulates the function of a wide variety of proteins. In most cases this binding occurs only when CaM is "activated" after binding Ca 2ϩ . Because the affinity of CaM for Ca 2ϩ (ϳ10 Ϫ6 M) is above the resting [Ca 2ϩ ] i (ϳ10 Ϫ7 M), CaM is a sensor for transient increases in [Ca 2ϩ ] i , binding to target proteins only when [Ca 2ϩ ] i is elevated (39). CaM has been suggested to play several possible roles in membrane traffic in mammalian cells. Endocytosis in neurons is dependent on the phosphorylation state of dynamin (40), which is in turn a substrate for calcineurin (phosphatase 2B) (41), a phosphatase that is regulated by CaM binding. Cyclosporin A, an immune FIG. 3. CaM binds to the 17-amino acid basolateral targeting signal of pIgR. CaM binding of pIgR constructs was analyzed as in Fig. 2A. Constructs indicated are as follows: wild-type, wildtype pIgR; pIgR ⌬655-668, a construct in which 14 amino acids of the 17-amino acid basolateral targeting signal have been deleted; pIgR R655-STOP, a construct which contains only the first two amino acids of the cytoplasmic tail; pIgR T670-STOP, a construct which lacks all but the 17-amino acid basolateral targeting segment of the cytoplasmic tail. No binding was detected for pIgR ⌬655-668 or pIgR R655-STOP. The cytoplasmic domain of the protein encoded by each construct is indicated schematically at the right.
FIG. 4. CaM can be chemically cross-linked to detergent-solubilized pIgR. Triton X-114 preparations of metabolically labeled proteins from MDCK cells expressing pIgR constructs (see "Materials and Methods") were incubated in the presence or absence of 1 M exogenous CaM, with or without 3 mM EGTA as indicated, at 4°C for 1 h, followed by incubation for another hour in the presence or absence of BS 3 as indicated. Samples were quenched with 100 mM glycine and boiled in SDS, and pIgR was immunoprecipitated and analyzed by SDS-PAGE. With wild-type pIgR, an additional band of ϳ125 kDa (arrow), which represents pIgR cross-linked to CaM, appears only in the presence of CaM, BS 3 , and free Ca 2ϩ . This treatment does not produce an additional band for pIgR ⌬655-668 or pIgR R655-STOP.
FIG. 5. CaM can be chemically cross-linked to membrane-associated pIgR. Crude membrane preparations from MDCK cells expressing pIgR constructs were incubated with 125 I-CaM in the presence or absence of EGTA. After 1 h at 4°C, BS 3 was added to 0.5 mM, incubation was continued for another hour, glycine and SDS were added to 100 mM and 1%, respectively, samples were boiled for 5 min, and pIgR was immunoprecipitated and analyzed by SDS-PAGE and fluorography. 125 I-CaM was incorporated into a ϳ125-kDa species in the presence of membranes containing pIgR, but not membranes from MDCK cells expressing pIgR ⌬655-668. suppressant that functions by binding calcineurin, inhibits Ca 2ϩ /CaM-dependent secretion from pancreatic acinar cells, thereby implicating calmodulin/calcineurin in regulation of this secretion (42). More generally, [Ca 2ϩ ] i is clearly important in many intracellular membrane traffic events, including both regulated exocytosis of granules (43) and synaptic vesicles (44), as well as classically "constitutive" processes, such as endoplasmic reticulum to Golgi transport (45) and nuclear envelope fusion (46) (at least in in vitro systems). It is not known if [Ca 2ϩ ] i acts on CaM and/or some other target in these events.
We have found that CaM binds to the basolateral targeting signal of the pIgR, making CaM the first identified protein shown to bind specifically to a basolateral sorting signal. Binding is strictly dependent on Ca 2ϩ . The pIgR is the major CaMbinding protein detected in membranes of a highly purified rat liver endosome fraction, and is also the major CaM-binding protein detected in a Triton X-114 detergent phase extract from total metabolically labeled MDCK cells that are transfected with pIgR. These striking results suggest that the interaction of CaM and pIgR is of high specificity. We also detected the CaM-pIgR interaction by cross-linking, both in detergent extracts and, more significantly, in non-solubilized crude membranes prepared from MDCK cells. That CaM can bind to pIgR that is still in its native membrane makes it quite likely that under the appropriate in vivo conditions of elevated [Ca 2ϩ ] i , CaM should bind to pIgR. Capturing the interaction of CaM with pIgR in intact cells may be quite difficult, as elevations of [Ca 2ϩ ] i are generally very transient and spatially localized (47).
Using a series of mutant pIgRs expressed in MDCK cells, we mapped the CaM binding site to the membrane-proximal 17 amino acids of the cytoplasmic domain of the pIgR. This same segment has previously been shown to be necessary and sufficient to direct the pIgR to the basolateral surface from either the TGN or the endocytotic pathway (2). However, an analysis of Ala point mutants in this segment indicates that there is not a precise correlation between the residues needed for basolateral sorting and those needed for CaM binding. The sequence of this 17-residue segment is generally comparable with other known CaM binding sites. CaM often binds to a sequence with a preponderance of basic residues near the N terminus, and hydrophobic residues more concentrated near the C terminus (39). This is largely true of this 17-residue segment of the pIgR. It is remarkable that this short region of the cytoplasmic domain of the pIgR therefore has at least two separable functions, basolateral targeting and CaM binding. This segment also has a third function, as a rather weak signal for endocytosis, centered around tyrosine 668 (48). Based on our knowledge of how sorting occurs in other systems, it seems likely that a specific protein complex recognizes the basolateral sorting signal. This complex would be functionally and perhaps structurally homologous to the adaptor proteins found in clathrin-coated vesicles, and to the coatomer involved in traffic in the early portion of the biosynthetic pathway. A likely candidate for a component of the coat involved in basolateral sorting is the p200 protein found in the TGN (49). Such complexes presumably recognize specific signals on the cytoplasmic domains of integral membrane proteins. In the case of the AP2/HA2 adaptor found in the plasma membrane, the signal contains a type 1 ␤-turn (50 -53), perhaps somewhat similar to the type 1 ␤-turn that is an essential part of the basolateral signal of the pIgR (4).
In contrast, CaM generally does not function as an adaptor, i.e. it does not recognize a feature on one protein, and then bind to another protein. Rather, when CaM binds to a site on a protein, a common consequence is that it simply masks that site, preventing it from interacting with anything else (39). We speculate that one possible role of CaM is to sequester the basolateral signal on the pIgR. In the absence of elevated [Ca 2ϩ ] i , CaM would not bind to the pIgR. The hypothetical complex that recognizes the basolateral signal would bind to the signal and direct the pIgR to the basolateral surface. However, when [Ca 2ϩ ] i is elevated, perhaps due to the binding of dIgA to the pIgR or to another hormone signaling event, CaM would bind to pIgR and thereby prevent the hypothetical complex from binding. By masking the basolateral signal, the pIgR would be allowed to be transcytosed to the apical surface.
In pancreatic epithelial cells, the elevation of [Ca 2ϩ ] i in response to extracellular signals is greatest in the most apical region of the cell (54). With strong stimulation a wave of elevated [Ca 2ϩ ] i can then propagate through the cell. It seems likely that in MDCK cells elevation of [Ca 2ϩ ] i would similarly be concentrated in the most apical region of the cytoplasm. As described in the Introduction, the last known stage of pIgR transcytosis (i.e. step 3) is movement from the apical recycling compartment (located immediately beneath the apical plasma membrane) to the apical surface. It is possible that CaM binds primarily to the pIgR in this compartment and is involved in stimulating this step of transcytosis. Consistent with this hypothesis, we have observed that this is the step of transcytosis that is stimulated by dIgA binding (22), as well as by artificially raising [Ca 2ϩ ] i . 3 This localized increase in [Ca 2ϩ ] i could avoid the binding of CaM to pIgR in other locations, e.g. the TGN, which might lead to inappropriate delivery of pIgR to the apical surface. Exocytosis of synaptic vesicles at nerve terminals represents an extreme example of how a highly localized increase in [Ca 2ϩ ] i leads to membrane traffic at a precise location.
Signaling by dIgA binding to the pIgR or other extracellular FIG. 6. CaM binding to pIgR point mutants does not correlate with their basolateral targeting phenotype. CaM-agarose binding of pIgR mutants, in which alanine is substituted at the indicated positions in the context of fulllength pIgR, was analyzed as for Fig. 2A.
Lanes showing an immunoprecipitation of pIgR from the starting material (IP), and 35 S proteins bound to CaM-agarose in the presence of Ca 2ϩ or EGTA are indicated. The percentage of each construct bound to CaM-agarose, relative to wildtype, and basolateral targeting phenotype (*, determined previously; Ref. 4) of each mutant: (ϩ) normal basolateral targeting; (Ϫ), disrupted basolateral targeting, are indicated at right. Values represent the mean of two experiments (see "Materials and Methods" for details). signals leads to activation of several second messenger pathways, which may redundantly stimulate transcytosis. In particular artificial elevation of [Ca 2ϩ ] i stimulates apical transcytosis of several molecules (e.g. transferrin), although these effects are in general smaller than those observed with pIgR. 3 The molecular mechanisms of these effects are unknown. We suggest that the specific binding of CaM to pIgR provides an additional regulatory mechanism, which amplifies the stimulation of transcytosis of pIgR.
CaM binding to the pIgR could have another, non-mutually exclusive role in regulation of pIgR traffic. When dIgA binds to pIgR, the pIgR initiates a signaling cascade involving tyrosine phosphorylation. 2 Preliminary data indicate that mutations in the 17-residue basolateral targeting signal of the pIgR prevent this signaling, suggesting that this segment has an additional role in signal transduction. Perhaps CaM binding to the pIgR serves to sequester this segment and thereby is part of a negative feedback loop that shuts off signaling by the pIgR.
Taken together our data indicate that the CaM binds to a portion of the pIgR that has multiple functions in regulation of polarized traffic. It will be interesting to learn if other receptors bind to CaM and if this is involved in regulation of their traffic by [Ca 2ϩ ] i .