The epithelial mucin MUC1 contains at least two discrete signals specifying membrane localization in cells.

The MUC1 gene product (PEM, polymorphic epithelial mucin) is a cell-associated glycoprotein expressed on the apical surface of most simple secretory epithelia. The transmembrane and cytoplasmic domains of MUC1 have been shown to be highly conserved between mammalian species, and it has been shown that this molecule interacts with the actin cytoskeleton. Apical targeting signals in polarized cells have yet to be defined. The mechanism by which MUC1 is targeted and maintained on the apical surface is not known; correct localization, however, would be predicted to be crucial for function. In order to determine which domains of MUC1 were important for this localization, mutational analysis of the protein was undertaken. Using cytoplasmic tail deletion mutants, fusion proteins of MUC1 and CD2, and site-directed mutagenesis, it could be shown that MUC1 appeared to contain at least two motifs involved in apical localization. The first was located in the extracellular domain and was sufficient to confer apical localization on the fusion protein. The second was the Cys-GlnCys (CQC) motif at the junction of the cytoplasmic and transmembrane domains. This sequence was necessary for surface expression. These results suggest that MUC1 contains two discrete motifs important in its apical localization.

The MUC1 gene product (PEM, polymorphic epithelial mucin) is a cell-associated glycoprotein expressed on the apical surface of most simple secretory epithelia. The transmembrane and cytoplasmic domains of MUC1 have been shown to be highly conserved between mammalian species, and it has been shown that this molecule interacts with the actin cytoskeleton. Apical targeting signals in polarized cells have yet to be defined. The mechanism by which MUC1 is targeted and maintained on the apical surface is not known; correct localization, however, would be predicted to be crucial for function. In order to determine which domains of MUC1 were important for this localization, mutational analysis of the protein was undertaken. Using cytoplasmic tail deletion mutants, fusion proteins of MUC1 and CD2, and site-directed mutagenesis, it could be shown that MUC1 appeared to contain at least two motifs involved in apical localization. The first was located in the extracellular domain and was sufficient to confer apical localization on the fusion protein. The second was the Cys-Gln-Cys (CQC) motif at the junction of the cytoplasmic and transmembrane domains. This sequence was necessary for surface expression. These results suggest that MUC1 contains two discrete motifs important in its apical localization.
The MUC1 gene product (also designated PEM for polymorphic epithelial mucin, episialin, and DF3) is a high molecular weight, membrane glycoprotein expressed on the apical surface of most simple secretory epithelia. This protein has also been shown to be expressed abundantly in many breast and other carcinomas where it is aberrantly glycosylated (1)(2)(3)(4)(5)(6). Normal MUC1 is expressed on the luminal surface of cells lining epithelial ducts and glands, where it may play a role in protecting and lubricating the epithelium or in modulating cell adhesion, cell recognition, or tumor progression (7)(8)(9). The mechanism by which MUC1 is targeted and maintained on the apical surface is not known; correct localization, however, would be predicted to be crucial for function. Because of the important role of mucin glycoproteins in cancer, more information regarding the expression and trafficking of this molecule is needed.
The MUC1 gene encodes a protein that consists of three distinct domains, namely a large extracellular domain and transmembrane and cytoplasmic domains (10 -13). The extracellular domain consists predominantly of variable numbers of a 20-amino acid tandem repeat, making the gene encoding MUC1 highly polymorphic, with each allele encoding a product containing different numbers of repeats (10,14). Each tandem repeat contains several potential sites for O-linked glycosylation, and the mature molecules expressed by the mammary gland have been estimated to be composed of 50 -60% carbohydrate (10). The extensive glycosylation of MUC1, combined with the presence of many proline residues, contributes to the extended and rigid structure of MUC1 on the cell surface (15,16).
It is thought that the cytoskeleton has an important role in maintaining the cytoarchitecture of epithelia and in defining membrane specializations such as junctional complexes and microvilli. Previous work showing that actin-disrupting drugs selectively led to a disruption of the apical localization of MUC1 suggested that interaction with the actin cytoskeleton may be important for its localization (17). Moreover, MUC1 appears to localize to microvilli, and it is possible that it may play a role in organizing the cytoskeleton (17). This is also suggested from the observation that MUC1 is expressed early in developing epithelia before the glands are active, while morphogenesis is still occurring (18). In this paper we attempt to determine if a particular domain of MUC1 contains a signal to specify its apical localization. Discrete domains in the cytoplasmic region of some proteins have been shown to target these proteins to the basolateral domain, although such signals have not yet been identified for apical targeting (19,20). Because of the high degree of sequence conservation in the cytoplasmic tail of MUC1 in different mammalian species (21)(22)(23), we postulated that the cytoplasmic tail is functionally important, and we chose to dissect this domain carefully. One possible function for the cytoplasmic tail is to target MUC1 to the apical membrane. Our results indicated that the cytoplasmic domain was not involved in the apical localization; however, two cysteine residues at the end of the transmembrane/beginning of the cytoplasmic tail appeared to be important for membrane localization. Using chimeric molecules composed of different domains of CD2 and MUC1, we have demonstrated that the extracellular domain of MUC1 can confer apical localization on the CD2 molecule. The results suggest that MUC1 contains two discrete motifs important in its apical localization.

EXPERIMENTAL PROCEDURES
Materials-The cell lines MDCK 1 and Panc-1 were from the ATCC. The monoclonal antibodies HMFG-2, OKT11, and anti-MHC Class I W6/32 were from the Imperial Cancer Research Fund, London, United Kingdom, and 12C10 was a kind gift from Bruce Acres, Transgene, Strasbourg, France. The anti-transferrin receptor antibody was purchased from Boehringer Mannheim. G418-geneticin was purchased from Life Technologies, Inc.
Tissue Culture Techniques-Cell lines were cultured at 37°C in 10% CO 2 in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Cells were passaged every 2-3 days, and early passages were frozen and stored in liquid nitrogen. Cells were grown in flasks and transferred to 3-cm plastic dishes (Nunc) or 16-mm glass coverslips for further processing.
Transfections-MDCK cells were transfected by calcium phosphate precipitation. Briefly, cells were grown to 70% confluence in 10-cm dishes. Twenty g of plasmid DNA was precipitated with calcium phosphate and added to the growth medium. Cells were incubated with above for 4 h at 37°C, followed by incubation with 15% glycerol for 2 min before returning them to normal growth conditions. After 72 h, 0.7 mg/ml G418 was added to the growth medium. Panc-1 cells were transfected by DEAE-dextran precipitation. Cells at 70% confluence were incubated with 20 g of DNA and 50 g/ml DEAE-dextran in Trisbuffered saline for 30 min at room temperature. Cells were osmotically shocked with glycerol as above and returned to normal growth conditions. At 72 h post-transfection, cells were grown in selection medium containing 0.6 mg/ml G418. From both cell lines, discrete colonies were isolated and expanded for further testing.
Immunofluorescence Staining-Cells were grown to confluence on 3-cm plastic dishes or glass coverslips. Cells were either fixed for 10 min in methanol:acetone at Ϫ20°C or fixed with 3.5% paraformaldehyde at room temperature. For "live" staining, cells were left unpermeabilized and unfixed and only fixed with methanol:acetone as above at the end of the antibody-labeling procedures. Cells that were solubilized with 0.5% Nonidet P-40 were grown previously in 3-cm plastic dishes for 3-4 days. Cells were incubated on ice for 15 min with 0.5% Nonidet P-40 in 10 mM Hepes, 0.17 M NaCl, 1 mM CaCl 2 . After a brief washing, cells were fixed with 3.5% paraformaldehyde on ice for 15 min and then further fixed with methanol at Ϫ20°C for 15 min. Immunofluorescent staining was as described (23). Stained cells were analyzed using a Zeiss Axiophot microscope or a Nikon Optiphot Confocal Microscope.
FACS Analysis-Cells were trypsinized, and 1 ϫ 10 6 cells were processed in microtiter plates for FACS analysis as described (5). Five thousand cells were analyzed on a Becton Dickinson FACS scan.
Electron Microscopy-Cells from a 75-cm 2 flask were scraped gently into a tube and pelleted by centrifugation. For pre-embedding labeling, cells were either unfixed or lightly fixed in 1% monomeric glutaraldehyde in phosphate-buffered saline. After washing in phosphate-buffered saline and quenching in 50 mM glycine, cells were blocked with 10% normal goat serum and 1% bovine serum albumin in phosphatebuffered saline for 30 min at 4°C. After washing in phosphate-buffered saline with 0.1% bovine serum albumin, cells were incubated with antibody at 10 -50 mg/ml for 30 min at 4°C. Cells were washed and incubated with colloidal gold conjugated goat anti-mouse Ig (1:20) for 60 min at 4°C. After washing, cells were fixed in 2.5% glutaraldehyde, and silver enhancement was carried out using Intensive M. Silver Enhancement Kit (Amersham). Cells were postfixed in osmium tetroxide in Sorenson buffer at room temperature for 2 h, dehydrated through graded ethanol solutions (70 -100%), followed by propylene-oxide, and then impregnated with propylene-oxide:Epon at decreasing concentrations of propylene-oxide until embedding in pure Epon overnight. Ultrathin (80 -100 nm) sections were cut, stained with uranyl acetate, and viewed with a Zeiss 10C microscope. For postembedding labeling, cells were fixed in 1% monomeric glutaraldehyde, dehydrated as above, and infiltrated with Lowicryl rather than Epon. Ultrathin sections were taken as above and then reacted with antibodies as described.
Oligonucleotides-The sequence of oligonucleotides used in the generation of MUC1 cytoplasmic tail deletion constructs, CQC mutations, and MUC1-CD2 chimeric constructs are shown below.
Oligonucleotides used in the construction of MUC1 cytoplasmic tail deletion (numbers based on MUC1 sequence published in Ref. Development of MUC1 Constructs-A BamHI fragment containing the entire coding sequence of MUC1 including 30 tandem repeats was ligated into the BamHI site of pBS (Stratagene) from which the polylinker sites between SalI and KpnI had been removed by digestion, filled in with Klenow, and religated (creating pBS⌬AA). This construct had unique AccI and ApaI sites in the MUC1 cDNA located 5Ј and 3Ј of the transmembrane and CT region (see Fig. 1A for position of restriction sites). MUC1 cDNAs with deletions in the cytoplasmic tail (CT) were created as follows: PCR fragments were generated from the full-length cDNA with a 5Ј primer spanning the AccI site and 3Ј primers located at various positions within the CT. The 3Ј primers contained a translation termination codon and a 3Ј ApaI site. PCR products were subcloned into pBS⌬AA MUC1 using AccI and ApaI. The cDNAs with different length CTs were transferred to pH␤APr-1-neo using BamHI (25).
Site-directed mutagenesis of the cysteines proximal to the cytoplasmic tail region to alanine was performed using a two-stage PCR method. Overlapping primers containing the necessary nucleotide substitutions were made in both orientations. Using these in combination with a primer at the AccI site (primer A) and a primer at the stop codon (MUC-D), two separate PCR reactions were performed. The PCR products formed the template for the second PCR reaction using the AccI and stop codon primers. The cysteines were mutated singly (using oligonucleotides CQA M1, CQA M2, or AQC M1, AQC M2) and together (using oligonucleotides AQA M1, AQA M2). After verifying the sequence by sequencing, the PCR product was digested with AccI and BamHI and ligated into similarly cut MUC1 to replace the existing region. All constructs were transferred to pH␤APr-1-neo vector.
Full-length CD2 was subcloned into pH␤APr-1-neo. Constructs containing domains from both CD2 and MUC1 were generated by PCR. Restriction sites were included at the 5Ј of the PCR primers to allow coligation of domains from the two molecules without changing the amino acid sequence. All constructs were then ligated into the pH␤APr-1-neo expression vector.
To create NOTR and 3TR, an XmnI to MscI fragment of the MUC1 cDNA encompassing the translational start and the entire tandem repeat domain was ligated into the SmaI site of pBS in which the AccI site had been destroyed previously (creating pBSTR). Following partial digestion of pBSTR with SmaI (which cuts in every tandem repeat), plasmids were recircularized by ligation. Resulting plasmids were analyzed for the number of remaining tandem repeats. DNA fragments containing 3 or 0 tandem repeats were removed by digestion with HindIII and AccI and ligated into pBS⌬MUC1 replacing the original tandem repeat-containing fragment. Full-length MUC1 cDNAs with 0 or 3 repeats were removed from pBS-⌬AA by digestion with HindIII and BamHI and ligated into pH␤APr-1-neo expression vector. All constructions were verified by extensive restriction enzyme analysis, and fragments generated by PCR were verified by sequence analysis.

Mutant MUC1 Molecules with Deletions of the Cytoplasmic
Tail Are Expressed on the Cell Surface-To determine whether the cytoplasmic domain of MUC1 plays an important role in targeting the molecule to the cell surface, truncated forms of the gene were created (Fig. 1A). Stop codons were inserted at specific locations in the cytoplasmic domain coding sequence as outlined under "Experimental Procedures." The wild type (WT) construct contained the entire extracellular domain including 30 tandem repeats, the transmembrane region, and all 69 amino acids of the cytoplasmic tail. The deletion mutants contained all of the extracellular and transmembrane sequences, and 45 (CT45), 33 (CT33), 18 (CT18), or 3 (CT3) amino acids of the cytoplasmic tail. The three amino acids were the charged residues Arg-Arg-Lys (RRK) which are believed to represent a stop transfer sequence and would therefore be necessary for transmembrane retention of the molecule during biosynthesis. All constructs were subcloned into the pH␤APr-1-neo vector and transfected into canine MDCK and human Panc-1 cells, both of which express very low levels of the MUC1 gene or homologue. After antibiotic selection, multiple independent clones of positive transfectants were initially selected by staining methanol:acetone-fixed cells with the monoclonal antibody HMFG-2 (26) by indirect immunofluorescence. HMFG-2 is reactive with an epitope contained within the extracellular tandem repeat domain (27).
The surface expression of MUC1 was analyzed on the transfected Panc-1 cell lines (Fig. 1B) and MDCK cell lines (data not shown) by FACS after immunofluorescence staining of live cells with HMFG-2. In both cell lines, high levels of the wild type MUC1 (WT) were expressed on the cell surface. Cells transfected with the cytoplasmic tail deletion constructs CT45, CT33, CT18, and CT3 also expressed MUC1 on the cell surface. MDCK cells that had been transfected with the empty expression vector (V) did not express MUC1, while Panc-1 cells transfected with this vector expressed low levels comparable with those detected in the parental cell line. There were, however, no differences in the levels of expression of the wild type and different mutant forms, other than clonal differences in levels of expression. Panc-1 cells transfected with MUC1 appeared to express higher levels of the glycoprotein than the MDCK cells, although in both cell lines, expression of MUC1 in a clonal population of cells appeared fairly heterogeneous.
In cultured epithelial cells which are polarized, MUC1 is generally detected on the apical domain corresponding to the distribution observed in normal epithelial cells in vivo (6,17). The membrane localization of MUC1 in the transfected cells was determined by immunofluorescence analysis of stably transfected clones using the antibody HMFG-2 directed to the tandem repeat domain. The cells were grown on glass coverslips for several days in order to form tight monolayers and then stained for immunofluorescence without prior permeabilization and analyzed by confocal laser microscopy. Analysis of the sequence of the XY planes (horizontal sections) of the cells transfected with wild type MUC1 clearly showed segregation of the label to the apical domain of these cells. Vertical XZ sections showed that, in the MDCK cells, which form tight monolayers, the immunofluorescent signal was excluded from the basolateral domain and was seen only in the apical domain (Fig. 2, A and B). In the Panc-1 cells, staining was seen in the apicolateral domain. Panc-1 cells do not form tight monolayers as seen in the MDCK cells and polarize to form apicolateral and basal domains, where the apicolateral domain encompasses much of the lateral region of the cell. MUC1 is restricted to this domain in these cells and is excluded from the basal domain. Similar analysis of both cell lines expressing the deletion mutants CT45, CT33, CT18 (data not shown), and CT3 (Fig. 2, C and D; shown here for Panc-1 CT3 cells) indicated that the mutant MUC1 was expressed on the apical (or apicolateral) cell surface in the same way as the wild type molecule. These results indicated that in these transfected cells the MUC1 protein was targeted to the apical domain and the cytoplasmic tail was not important for apical expression.
Wild Type and Mutant MUC1 Show the Same Ultrastructural Localization-The characteristic punctate pattern of MUC1 observed by indirect immunofluorescence staining and the ultrastructural studies of Parry et al. (17) suggested that this molecule was associated with microvilli. Moreover, data obtained following treatment of cells with actin-disrupting drugs suggested that MUC1 is associated with the actin cytoskeleton (17), which is well represented in microvilli rich in actin. Electron microscopic immunocytochemistry was carried out on the transfected cells in order to confirm the confocal microscopy results and to determine whether the products of the transfected MUC1 gene and the MUC1 mutants showed the same ultrastructural localization. Ultrastructural studies by Parry et al. (17) have shown that MUC1 in an endogenously expressing cell line was localized to apical microvilli.
Cells were grown to confluence on plastic dishes, scraped from the dishes, and fixed as pellets. Panc-1 cells (which are not as well polarized as MDCK cells) were fixed and embedded prior to staining and immunolabeled with HMFG-2 and goldconjugated goat anti-mouse IgG, whereas the MDCK cells were fixed and embedded following immunolabeling. This latter type of procedure generally prevents the entry of any labeled anti- body into intracellular compartments, and only antigen available on the membrane is labeled. Also, since the cells form tight monolayers, they retained their cell-cell attachments during processing, and it was possible to orientate them and to determine the apical and basolateral surfaces, both of which were accessible to labeled antibody. In the MDCK cells, the wild-type MUC1 and CT3 mutant were expressed only on the apical surfaces and appeared to be localized to microvilli as illustrated for CT3 in Fig. 3A. No surface expression was seen in MDCK cells transfected with the vector alone (data not shown). Similar results were observed with the Panc-1 cells. In the cells transfected with empty expression vector, MUC1 was expressed at very low levels both in intracellular vesicles and on the cell surface (Fig. 3B). In the MUC1-transfected Panc-1 cells, a high degree of labeling was observed in intracellular vesicles and on the cell surface (Fig. 3C) where labeling appeared to localize to microvilli. No differences could be detected between the localization of ImmunoGold labeling in transfected Panc-1 cells expressing MUC1 WT and the mutants CT45, CT33, CT18 (data not shown), or CT3 (Fig. 3D) MUC1 proteins. These results demonstrated that the transfected MUC1 was localized to apical microvilli as has been shown for endogenously expressed MUC1 (17), and, similarly, that the CT3 mutant MUC1 was also expressed on microvilli. This suggests that this localization is not dependent on the cytoplasmic tail.
Membrane Retention of MUC1 Is Not Dependent on the Cytoplasmic Tail-It has been suggested that interactions between MUC1 and the actin cytoskeleton are important for the apical localization of MUC1. Results presented above with the tailless CT3 mutant which is unlikely to be able to interact directly with the actin cytoskeleton suggested that this was not the case. In order to determine whether interactions of the cytoplasmic tail with the cytoskeleton were important for anchoring MUC1 on the cell surface, detergent extractions of cells transfected with MUC1 and the MUC1 mutants were carried out. Extraction of MUC1 from a breast epithelial cell line, MCF-7, with 0.5% Nonidet P-40 for 15 min on ice and subsequent separation by centrifugation of the insoluble and soluble fractions, resulted in approximately 50% of the cellular MUC1 fractionating into the insoluble pellet fraction, as determined by Western blotting. 2 To determine whether interactions of the cytoplasmic tail of the MUC1 gene product with the cytoskeleton were conferring the detergent insolubility on the MUC1 molecule, cells transfected with the wild type and truncated genes were solubilized with 0.5% Nonidet P-40. This concen- tration of nonionic detergent has been reported to solubilize most other membrane proteins.
Panc-1 cells were extracted with 0.5% Nonidet P-40 as outlined under "Experimental Procedures." After washing away the solubilized proteins, the detergent-insoluble fraction was fixed using paraformaldehyde and Ϫ20°C methanol and processed for indirect immunofluorescence staining using the monoclonal antibody HMFG-2. In the Panc-1 cells expressing transfected MUC1 WT, the pattern of staining of the MUC1 protein remaining in the insoluble fraction appeared as a fibrillar lattice (Fig. 4). This probably represented MUC1 associating with the cortical cytoskeleton. Under the same conditions, it was possible to show that the transferrin receptor (Fig. 4) and HLA Class 1 (data not shown) were completely solubilized from the cell membrane. Comparison of the levels and patterns of immunofluorescent labeling of detergent-insoluble MUC1 between cells expressing the wild-type MUC1 and those expressing the mutants CT45, CT33 (data not shown), CT18 (data not shown), and CT3 showed no differences (Fig. 4). These data suggested that the insolubility of this fraction of MUC1 was not dependent on interactions of the cytoplasmic tail with the underlying cytoskeleton. Interactions mediated via the transmembrane or extracellular domains of MUC1 may be important for anchoring the molecule in the membrane.
Cytoplasmic Cysteine Residues Are Important for Membrane Expression-The terminal three amino acids of the transmembrane domain adjacent to the cytoplasmic domain are the residues Cys-Gln-Cys (CQC). Previously, investigators have arbitrarily demarcated the CQC as the end of the transmembrane domain after studying hydrophobicity plots of the molecule (10). However, the transmembrane domain of MUC1 is longer than the most other single spanning domains, and, without fine mapping of this domain, it is not clear whether the CQC residues are actually within the lipid bilayer or within the cytosol. The cysteines in this tripeptide motif have been postulated to be involved in MUC1 protein complex formation (28,29). In order to determine whether the cysteines in this tripeptide motif were important for MUC1 expression on the apical sur-face, the cysteine residues were either singly or collectively mutated to alanines using PCR-based site-directed mutagenesis as outlined under "Experimental Procedures." The constructs transfected into MDCK cells were identical with the MUC1 WT construct used in earlier experiments except that one or two cysteine to alanine substitutions (CQA, AQC, or AQA) were encoded (Fig. 5). After transfection and expression in the MDCK cell line, the mutant MUC1 expressing cells were analyzed by indirect immunofluorescence using the antibody HMFG-2. Cells were analyzed after methanol:acetone fixation and live, without any prior permeabilization. In the live stained cells, no immunofluorescence was detected in either the CQA, AQC, or AQA mutants, suggesting that MUC1 was not present on the cell surface (Fig. 6, Column 1). In the fixed cells, however, strong immunofluorescence could be detected in a diffuse pattern in the cytoplasm, with some signal localizing to the lateral regions of the cell (Fig. 6, Column 2). These results demonstrated that single amino acid substitution of either cysteine of the CQC tripeptide results in a loss in surface expression and suggested that this domain may be important for either targeting or retaining MUC1 on the cell surface.
Analysis of the Domains of MUC1 Important for Apical Expression-As no portion of the cytoplasmic tail had been demonstrated to be responsible for the apical targeting and retention of MUC1, it was possible that, in addition to the CQC motif, the transmembrane or extracellular domains contained signals that were important for apical localization. To analyze this further, chimeric proteins were constructed as described under "Experimental Procedures." Fusion proteins consisting of the lymphocyte-specific glycoprotein CD2 and specific regions of the MUC1 protein were constructed to determine which domain of MUC1 conferred apical localization on CD2. CD2 was chosen as a reporter molecule as it is expressed only on nonpolarized lymphoid cells (30) and would not be expected to contain strong targeting signals.
Three chimeric proteins were made (Fig. 7). The construct called TMCT consisted of the CD2 extracellular domain and the MUC1 transmembrane and cytoplasmic domains. 30 tandem repeats fused to the transmembrane and cytoplasmic tail domains of CD2. The entire coding region of CD2 was also cloned into the expression vector. The three chimeric proteins and the wild type CD2 and MUC1 proteins were expressed in stable transfectants of MDCK cells. After transfection and selection of individual clones, the cells were processed for immunofluorescence staining and analyzed by laser scan-ning confocal microscopy. The MDCK cells expressing wildtype CD2, TM, and TMCT were stained with the antibody OKT11 reactive with an epitope in the extracellular domain of CD2. Cells were fixed with methanol:acetone prior to labeling, since the staining was much stronger following fixation, which presumably allowed access of the antibody to its epitope. The distribution of the protein was visualized in the XY (horizontal, Fig. 8, Column 1) and the XZ (vertical, Fig. 8, Column 2) planes. In the horizontal plane, the labeling pattern appeared as a sharp profile outlining cell-cell contact areas, in a characteristic basolateral pattern. Small amounts of stained intracellular material were also visible. In the XZ plane, the label appeared to be restricted to the lateral surfaces of the cells (Fig. 8, A, B,  and E-H). Therefore, in these cells, the wild type CD2 protein was expressed exclusively on the basolateral domain of the cells. Comparison of at least three clones for each of the CD2, TM, and TMCT transfectants showed that no significant differences in the staining patterns could be detected; all the proteins were expressed basolaterally. Similar results were obtained when the cells were analyzed without prior fixation, although the staining was weaker.
The MDCK cells expressing the wild type MUC1 and MEX proteins were not fixed prior to labeling and were labeled with the monoclonal antibody HMFG-2 for confocal analysis. The staining pattern for wild type MUC1 transfectants was as described in Fig. 3 with the characteristic apical staining seen in the XY section (Fig. 8, C and D). In the cells expressing the MEX chimeric protein, the staining pattern was the same, with punctate dots appearing over the apical surface of the cell (Fig.  8, I and J). In the vertical sections, both the wild type MUC1 and the MEX chimera exhibited labeling that was restricted to the apical domain (Fig. 8, D and J).
These results suggested that the extracellular domain was the region important for the apical targeting of the fusion proteins. Therefore, we asked whether the tandem repeat domain was determining this localization, since this domain comprises much of the extracellular portion of the MUC1 protein. Two constructs were made and expressed in MDCK cells (Fig.  9). The first, called NOTR, consisted of the MUC1 protein containing no tandem repeats but with the flanking degenerate 5Ј and 3Ј repeat units intact. The second construct, designated 3TR, consisted of the entire MUC1 sequence with three tandem repeats of 20 amino acids in the extracellular region. Following transfection of MDCK cells and selection of individual clones, immunofluorescence analysis was performed as before. Since  7. MUC1 and CD2 chimeric constructs. A schematic representation of chimeric constructs made by PCR using sequences from MUC1 (gray shading, hatched area represents tandem repeat domain) and CD2 (black shading). The extracellular, transmembrane, and cytoplasmic domains are indicated, and names shown are also those used for cell lines transfected with these constructs. 3TR and NOTR represent MUC1 constructs that encode for MUC1 with only 3 (3TR) or no (NOTR) tandem repeats. Both constructs include the remainder of the MUC1 coding sequence. the HMFG-2 antibody did not label these mutant forms of MUC1 very strongly, the antibody 12C10, recognizing an extracellular epitope outside the tandem repeat domain, was utilized (31). Immunofluorescence analysis of these cells confocally showed that the mutant MUC1 proteins lacking the tandem repeat or with three repeats present were expressed on the apical cell surface (Fig. 9). Comparison of the XY and XZ series showed that the staining patterns were similar to MUC1 WT. Thus, the tandem repeat sequences do not appear to contain a signal for apical localization. DISCUSSION MUC1 gene product is detected on the apical surface of most simple glandular epithelia. The precise function of this protein is not well understood, although it is probable that apical localization is important for its function. In this paper we have analyzed the contribution of the various domains to the apical targeting of the molecule by transfecting into polarized epithelial cells mutated, deleted, or chimeric forms of the MUC1 gene. Deletions of the cytoplasmic tail of MUC1 appeared to have no effect on the apical localization and, surprisingly, MUC1 with virtually no cytoplasmic tail was shown not only to be correctly localized to the apical microvilli, but also retained in the detergent insoluble fraction. In contrast to the lack of effect seen by removing the entire cytoplasmic domain, the substitution of a single transmembrane/cytoplasmic cysteine by an alanine led to the protein being mislocalized, with no MUC1 detected on the cell surface. However, while these cysteines appear to be important for membrane localization of the full-length MUC1, their presence in the fusion proteins discussed below is not sufficient to confer apical localization in cells transfected with these hybrid genes.
Hybrid molecules containing the extracellular domain of CD2 and the transmembrane or transmembrane and cytoplasmic domains of MUC1, although expressed on the cell surface, were not localized to the apical domain but to the basolateral domain. These results indicated that both the cytoplasmic and transmembrane domains lacked dominant signals for apical targeting. It was possible that CD2 contained a basolateral targeting signal in the extracellular domain; however, this seems unlikely as it is endogenously expressed on nonpolarized cells, and all basolateral signals identified to date have been cytoplasmically located (19,20). In contrast, a fusion protein consisting of the extracellular domain of MUC1 and the transmembrane and cytoplasmic domains of CD2 was targeted to the apical surface, suggesting that the extracellular domain of MUC1 was the domain required for apical targeting.
The transmembrane and cytoplasmic domains of MUC1 are conserved between mammalian species with an amino acid identity of 80 -90% (22,23), suggesting that this domain is functionally important in the MUC1 molecule as a whole. The cytoplasmic tail, which has been described as being phosphorylated, contains 7 tyrosines, 6 of which are conserved in other species (22,29). MUC1 has been shown to be recycled through the trans-Golgi network from the apical surface several times (32), a process that is accompanied by further sialylation of the O-glycans. A specific endocytic determinant has not been shown in MUC1, but, by analogy with other proteins, one of the cytoplasmic tyrosines could be involved (33). Although the cytoplasmic tail may be involved in endocytosis, our results suggest that this function is not necessary for maintaining the apical localization of MUC1.
The association of MUC1 with the actin cytoskeleton is implied from the capping effect of cytochalasin D (17), and the cytoplasmic domain would be the most likely candidate. We postulated that this interaction may be important for the correct localization of MUC1. However, as our results showed clearly that the cytoplasmic domain was not necessary for apical localization nor for the detergent insolubility and retention of MUC1 in the membrane, it is unlikely that MUC1cytoskeletal interactions are important for this localization. The detergent insolubility was most likely not due to direct cytoskeletal interaction. Thus, it seems likely that some other domain of the MUC1 molecule is important for anchoring it on the apical surface. Similar results using a mutant version of CD44 with a deleted cytoplasmic domain demonstrated that a significant fraction of this mutant also partitions into the detergent insoluble fraction (34). Interestingly, a cytoplasmic deletion mutant of MUC1 has been shown to be able to be capped using cross-linked antibodies (8). It is possible that MUC1 is forming large detergent insoluble complexes either with itself or another molecule, perhaps via transmembrane or extracellular interactions.
The cysteines in the tripeptide motif CQC have been postulated to be involved in the formation of protein complexes (28). Studies involving the small molecule MUC1/Y showed these complexes were detected only under nonreducing conditions (29,35). This suggests that disulfide bond formations could be involved, although this is unlikely as these cysteines are cytosolic and not exposed to the lumen of the ER. A similar CXC motif in the cytoplasmic domain of CD4 has been shown to be involved in complex formation (36). The authors suggested these cysteines were involved in forming protein complexes involving heavy metals such as zinc. Further studies are needed to distinguish more precisely the mechanism by which the CQC motif is functioning.
Although all basolateral targeting signals investigated to date have been shown to comprise short motifs of several amino acids, usually including tyrosines located within the cytoplasmic domain, similar sorting determinants for apical proteins have not been identified (19,20). The identification of the Drosophila protein crumbs as an apical protein that promotes the development of the apical membrane provides direct evidence of an apical targeting signal (37). One group of specialized proteins that are anchored in the membrane by a glycosylphosphatidylinositol anchor are expressed apically, and the glycosylphosphatidylinositol anchor has been shown to confer apical localization on these proteins (38). For some viral proteins detected on the apical membrane of cells after infection, the sorting determinant has been shown to be in the ectodomain (39). Other proteins that are usually localized to the apical membrane are selectively secreted from the apical surface when expressed as truncated proteins lacking a transmembrane or cytoplasmic domain, also suggesting that there may be a sorting determinant in the ectodomain (40,41). Many investigators have suggested that the transport machinery for apical and basolateral proteins differs. Antibodies against NSF and S-nitrose-DL-penicillamine (SNAP) affect only the basolateral pathway (42). Some studies have suggested a differential requirement for heterotrimeric G proteins and small GTPbinding proteins in the sorting and fusion of apically and basolaterally bound vesicles (43,44). Prevention of GDP exchange on Rab proteins using a Rab-GDI affects only the basolateral pathway (42), while others have shown that the kinases, cAMP-dependent protein kinase and protein kinase C, selectively stimulate the apical transport pathway (45). It has also been demonstrated that the transport of basolaterally and apically destined vesicles has a differential requirement for microtubule-based motors (46). Apical and basolateral transport is mediated by distinct sets of transport vesicles (47), and apical transport vesicles themselves have been shown to be enriched in different subsets of proteins such as annexin XIIIb (48). It is likely that proteins carried in these vesicles destined for the apical domain possess some determinant to mark them as apical so that enrichment in the correct vesicle and engagement of the correct machinery can occur. As discussed above, it is possible that this signal is contained in the ectodomain of the protein. If so, it may depend on conformation, since homology has not been observed in the ectodomains of apically targeted proteins. Therefore, the inclusion of a determinant in the extracellular domain such as we have described for MUC1 is consistent with this idea.
As the construct lacking the tandem repeat domain (NOTR) was expressed on the apical membrane, it appears that the signal for apical localization and retention in the apical membrane is not contained within the tandem repeats. The tandem repeat domain is a scaffold for the majority of the O-linked carbohydrate that is attached to MUC1 (10). It is, therefore, probable that this O-linked carbohydrate is not playing a role in the localization either. Unpublished data from our laboratory indicate that inhibition of addition of the O-linked carbohydrate does not affect the localization of MUC1. 2 However, it has been suggested by Fiedler and Simons (49) that N-linked glycans may affect the apical localization of some proteins, and there are a number of sites for both N-and O-linked glycosylation outside the tandem repeat domain. N-Linked glycosylation of MUC1 has been demonstrated (50). Studies are in progress to determine the significance of N-linked carbohydrate in apical trafficking of MUC1.
Comparison of the extracellular domain from different species shows a much lower level of homology than is observed in the transmembrane and extracellular domains (21,22). Analysis of partial MUC1 clones from several mammalian species that include the extracellular domains proximal to the membrane indicated that, although homology overall was low, several short motifs were well conserved, and it is possible these may represent sequences important in the apical localization of MUC1 (22).
In summary, although the cytoplasmic domain is highly conserved, it is not important in the localization and retention of MUC1 at the apical cell surface. In contrast, the molecule appears to contain at least two motifs involved in this localization. One is at the junction of the transmembrane and cytoplasmic domains that is necessary for surface expression, and the other is in the extracellular domain (outside of the tandem repeat region). A more detailed analysis of the stalk region of the extracellular domain which contains sequences conserved across species is in progress.