The cytoplasmic domain of CEACAM1-L controls its lateral localization and the organization of desmosomes in polarized epithelial cells

CEACAM1 (carcinoembryonic antigen-related cell adhesion molecule 1), previously known as C-CAM, BGP or CD66a, is an immunoglobulin superfamily cell adhesion molecule that plays important roles in cell signaling and signal regulation (Öbrink, 1997; Beauchemin et al., 1999). It is abundantly expressed in epithelia, vessel endothelia, trophoblasts, platelets and neutrophilic granulocytes (Odin et al., 1988; Sawa et al., 1997), and is also present in many cell types of the immune system such as B cells, T cells, NK cells, macrophages and dendritic cells (Boulton and Gray-Owen, 2002; Kammerer et al., 1998; Kammerer et al., 2001; Markel et al., 2002; Morales et al., 1999; Singer et al., 2002). Because CEACAM1 regulates signal transduction and is expressed in so many different cell types, it affects a large variety of biological functions including cell proliferation (Singer et al., 2000), tumor growth (Hsieh et al., 1995; Kunath et al., 1995), angiogenesis (Ergün et al., 2000), apoptosis (Huang et al., 1999; Kirshner et al., 2003), epithelial cell polarization and lumen formation in glandular structures (Huang et al., 1999; Kirshner et al., 2003), lymphocyte activation (Boulton and Gray-Owen, 2002; Kammerer et al., 1998; Kammerer et al., 2001; Greicius et al., 2003), and insulin receptor internalization (Poy et al., 2002).

CEACAM1 is a homophilic cell-cell adhesion molecule, and the only known well-characterized physiological extracellular ligand for CEACAM1 is CEACAM1 itself. This suggests that homophilic transbinding of CEACAM1 molecules on adjacent cells plays an important role in CEACAM1-mediated regulation of cellular signaling processes. It has indeed been demonstrated that CEACAM1-mediated regulation of B-cell proliferation, T-lymphocyte activation, and NK-cell cytotoxic activity is governed by homophilic transbinding of CEACAM1 on adjacent cells (Greicius et al., 2003; Donda et al., 2000; Markel et al., 2002). On the cytoplasmic face of the plasma membrane, several different proteins were demonstrated to interact with CEACAM1 (Öbrink, 1997). Two major CEACAM1 splice isoforms occur, CEACAM1-L and CEACAM1-S, which differ in their cytoplasmic domains. CEACAM1-L has a 71-73-residue cytoplasmic domain containing two phosphorylatable Tyr residues, Tyr488 and Tyr515 (the numbering of the amino acid residues refer to the mouse sequence), whereas CEACAM1-S has a cytoplasmic domain containing only ten amino acids and lacking phosphorylatable Tyr residues. There is high sequence homology in CEACAM1 from different species, especially in their cytoplasmic domains, which are almost identical in rat and mouse (Öbrink, 1997). Signaling and signal regulation by CEACAM1 involve recruitment and activation of SH2-domain-containing protein tyrosine kinases (Brümmer et al., 1995) and protein tyrosine phosphatases (Huber et al., 1999) to the phosphorylated Tyr residues of CEACAM1-L. An interesting feature of CEACAM1 is that it can affect signal transduction in positive and negative manners, and we have suggested that variations in the balance of kinase and phosphatase recruitment to the CEACAM1-L cytoplasmic domain might be a reason (Öbrink et al., 2002).

In unpolarized cells and stratified epithelial cells CEACAM1 is distributed throughout the cell surface and/or in cell contact areas, whereas in polarized simple epithelial cells it is concentrated in the apical microvilli (Odin et al., 1988). Recently, we demonstrated that CEACAM1-S localizes exclusively to the apical cell surface of polarized epithelial cells, whereas CEACAM1-L occurs both at the lateral and the apical surfaces (Sundberg and Öbrink, 2002). We also showed that maintenance of the lateral localization of CEACAM1-L required homophilic transbinding of adjacent cells. This suggests that laterally localized CEACAM1-L is involved in signal regulation between polarized epithelial cells. It was therefore of great interest to investigate how the lateral localization of CEACAM1-L is regulated and which cellular processes CEACAM1-L might control. In this report, we demonstrate that several amino acids in the cytoplasmic domain are important for the lateral localization of CEACAM1-L, of which Tyr515 and Ser503 play prominent roles. We also show that drugs affecting Tyr and Ser phosphorylation cause disappearance of CEACAM1-L from the lateral surfaces, and that phosphatidylinositol 3-kinase (PI3K) is instrumental for the disappearance after induction of Tyr phosphorylation. Finally, we demonstrate that the cytoplasmic domain of CEACAM1-L influences the organization of desmosomes and cytokeratin filaments in the lateral cell surface domains.

Previously described rabbit polyclonal antibody against rat CEACAM1 (αCC16) (Sundberg and Öbrink, 2002) and mouse CEACAM1 (antisera 231 and 2457) (McCuaig et al., 1992) were used in immunofluorescence and immunoblotting analyses. Other primary antibodies used: occludin, clone OC-3F10 (Zymed Laboratories, San Francisco, USA); ZO-1, MAB 1520 (Chemicon, Temecula, USA); E-cadherin, C20820; β-catenin, C19220; EEA1, E41120 (all BD Transduction Laboratories, Erembodegem, Belgium); desmoplakin 1 and 2, DP2.15/2.17/2.20; cytokeratin 5+8, C22 (both Progen Biotechnik, Heidelberg, Germany); AP-1 (α-adaptin γ), A36120; AP-2 (α-adaptin α), A43920 (both BD Transduction Laboratories); antibody against phosphorylated-Tyr PY 99 (Santa Cruz Biotechnology, Santa Cruz, USA); rabbit polyclonal antibody against LAMP-1 (a kind gift of Sven Carlsson, Department of Medical Biochemistry and Biophysics, University of Umeå, Sweden). Secondary antibodies used: Alexa488-conjugated goat-anti-mouse Fab2; Alexa488-conjugated goat-anti-rabbit Fab2; Alexa546-conjugated goat-anti-rabbit Fab2; Alexa546-conjugated goat-anti-mouse Fab2 (all Molecular Probes, Leiden, The Netherlands); Cy3-conjugated anti-rabbit IgG (Rockland, Gilbertsville, USA); horseradish peroxidase-labeled swine anti-rabbit antibodies (Dako, Solna, Sweden).

The rat CEACAM1-L and CEACAM1-S cDNA constructs in the plasmid pRAX vector were described in (Olsson et al., 1995). The mouse CEACAM1-L cDNA in the retroviral vector pLXSN were described in (Kunath et al., 1995), and deletions and point mutants of mouse CEACAM1-L used in this study were specified in (Huber et al., 1999; Sadekova et al., 2000).

Madin-Darby canine kidney (MDCK) II cells were grown in a 5% CO₂-humidified atmosphere at 37°C in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 100 U/ml penicillin and 100 μg/ml streptomycin. MDCK cells stably transfected with rat CEACAM1-L and CEACAM1-S were described previously (Sundberg and Öbrink, 2002); clones F6H12 (CEACAM1-L) and E10H8 (CEACAM1-S) were used in the present investigation. MDCK cells were transfected with wild-type and mutant forms of mouse CEACAM1-L by using the calcium phosphate precipitation method (10 μg of cDNA of the pLXSN vector per 10 cm Petri dish). This was followed by selecting transfected cells in G418 (geneticin) and cloning by limiting dilution as described previously (Sundberg and Öbrink, 2002). Complete cell polarization was achieved by culturing the cells for 3 days on permeable filter inserts [PET filters: pore size 0.45 μm, catalogue number 3090 (Falcon, Labora, Stockholm, Sweden)] that separated the apical and basolateral compartments. To monitor cell-layer permeability ³H-inulin was added to the upper compartment (Sundberg and Öbrink, 2002). Only cell layers that permitted passage of less than 1% of the inulin from the upper to the lower compartment were used for experiments. Partially polarized cell layers were obtained by culturing cells on coverslips in tissue culture Petri dishes.

Confluent, CEACAM1-expressing MDCK cells were incubated with pervanadate, genistein, phorbol-12-myristate-13-acetate (PMA) or staurosporine as described in the Results. When not otherwise specified, the following conditions were used: 100 μM pervanadate for 5 minutes, 80 μM genistein for 50 minutes, 20 nM PMA for 15 minutes or 20 nM staurosporine for 120 minutes. Cell viability and intactness of the polarized cell layers were determined both by inulin permeability and by confocal microscopy of unpermeabilized cells. The effects of wortmannin were tested by incubating the cells in 50 nM wortmannin for 30 minutes at 37°C before the addition of other drugs.

To determine CEACAM1-L Tyr phosphorylation, confluent cells were solubilized at 4°C for 60 minutes with 1% Triton X-100 or 1% Brij 58 in a solution containing 150 mM NaCl, 25 mM HEPES pH 7.4, 5 mM EGTA, 10 mM Na₄P₂O₇, 50 mM NaF, 1 mM Na₃VO₄, 0.5 mM AEBSF (aminoethylbenzenesulfonyl fluoride), 1000 kIU (kallikrein inhibitory units)/ml aprotinin, 1 μg/ml leupeptin, 1 μg/ml pepstatin. Lysates were centrifuged and the protein concentrations of the supernatants were determined with the Micro BCA protein determination kit (Pierce, Rockford, USA). Supernatants were immunoprecipitated with anti-CEACAM1 IgG at 4°C overnight and immune complexes were collected on protein A-Sepharose for 1 hour. Immunoprecipitates were solubilized in sodium dodecyl sulfate (SDS) sample buffer (Laemmlie, 1970), reduced with 50 mM dithiothreitol (DTT), heated at 100°C for 5 minutes, and electrophoresed on 7% polyacrylamide gels (Laemmlie, 1970). The separated proteins were transferred to nitrocellulose membranes which were blocked with 5% nonfat milk powder in TRIS buffered saline (TBS) with 0.05% Tween 20 and incubated at room temperature for 1 hour with antibody against CEACAM1 or antibody against phosphorylated-Tyr, followed by appropriate horseradish-peroxidase-labeled secondary antibodies and developed by enhanced chemiluminescence (ECL). Membranes incubated with antibody against phosphorylated-Tyr were stripped in 2% SDS solution, reincubated with antibody against CEACAM1 and further developed by ECL. Developed films were scanned in a gel documentation equipment (Herolab, Stockholm, Sweden) or a Fuji Film detector system for quantitative analysis.

To determine total levels of cellular CEACAM1, filters with confluent MDCK cells were extracted with 150 μl of SDS sample buffer. Protein extracts were centrifuged, reduced with 50 mM DTT, heated at 100°C and subjected to SDS-PAGE on pre-cast NuPAGE 10% Bis-Tris gels (Invitrogen Life Technologies). Separated proteins were transferred to nitrocellulose membranes and CEACAM1 was quantified by ECL as described above. CEACAM1 expression levels of different cellular populations were analyzed by unpaired Student's t-test and compared. The ratio of apically (ap) to laterally plus intracellularly (l+i) expressed CEACAM1-L (ap:l+i) was determined in the following way. The apical phase of the polarized cell layer was first incubated with antibody against CEACAM1 (25 μg IgG/ml; 12.5 μg IgG per filter) for 60 minutes at 4°C. Cell layers were then washed, solubilized and immune complexes (representing the apically expressed CEACAM1-L) were collected by centrifugation at 13,000 rpm for 5 minutes after the addition of protein A-Sepharose. Antibodies against CEACAM1 were added to the supernatants and immune complexes were collected by protein A-Sepharose, as described above. This fraction represented the l+i expressed CEACAM1-L. The ap and the l+i fractions were run on pre-cast NuPAGE gels and quantified by ECL as described above. Ratios ap:l+i were determined in triplicate samples of untreated cells, or of cells treated with either pervanadate for 5 minutes or staurosporine for 2 hours. The results are given as the numerical average±s.e.m.

At least two independent clones for each CEACAM1 construct were analyzed. Confluent monolayers of fully or partially polarized CEACAM1-L-expressing MDCK cells were fixed with 3% paraformaldehyde for 30 minutes at room temperature. Cells were permeabilized with methanol for 2 minutes at room temperature (Sundberg and Öbrink, 2002) and reactive aldehyde groups were quenched in 50 mM NH₄Cl for 15 minutes at room temperature. Cells were then incubated overnight at 4°C with primary antibodies, followed by appropriate Alexa-conjugated secondary antibodies for 1 hour at room temperature and examined with a Zeiss LSM 510 scanning module fitted to an Axiovert 100 M microscope using a 63×oil immersion objective with a numerical aperture of 1.4. Specimens were analyzed at a range of different resolutions; all localization and colocalization results that are reported in this study were performed with volume elements (voxels) of the dimension 0.05×0.05×0.3 (x,y,z) μm³ or 0.07×0.07×0.3 (x,y,z) μm³. Data files were processed by using LSM software (Zeiss, Jena, Germany) and transferred to PowerPoint (Microsoft, www.microsoft.com) for reproduction.

Desmosomal plaques in the lateral surfaces of confluent, completely polarized cells were quantified in the following way. Three-dimensional reconstructions of cell layers stained for desmoplakin were created with LSM software. Tilted projections were then used to identify desmosomal plaques in the lateral surfaces and the number of plaques was counted in 5×5 μm squares. For each of the analyzed cell clones four to 16 different 25 μm² squares were counted. Desmosomal plaque density in the different clones was compared with the plaque density in untransfected cells and was analyzed by unpaired Student's t-test.

Although maintenance of CEACAM1-L on the lateral surfaces of confluent polarized MDCK cells requires homophilic transbinding of the extracellular domains (Sundberg and Öbrink, 2002), the signals directing CEACAM1-L to the lateral surfaces must reside in its cytoplasmic domain. CEACAM1-S (which differs from CEACAM1-L only in its cytoplasmic domain) does not appear on the lateral cell surfaces. In order to identify amino acids and motifs crucial for the lateral localization of CEACAM1-L we produced a series of deletions and mutations within the mouse CEACAM1-L cytoplasmic domain. The amino acid sequences of these constructs and of the entire wild-type cytoplasmic domain of mouse CEACAM1-L are shown in Fig. 1. MDCK cells were stably transfected with these mutants. The CEACAM1 surface localization in transfected cells was determined by confocal microscopy, and the results are shown in Fig. 2 and Table 1. Five deletion mutants (Δ518, Δ510, Δ495, Δ472, Δ461), which had an increasing part of the C-terminal end of the cytoplasmic domain removed, were analyzed. Like wild-type CEACAM1-L, Δ518 showed both lateral and apical localization, whereas the other four deletion mutants localized exclusively to the apical surface. This identified a motif for lateral localization within amino acid residues 511-518.

Different amino acid sequences have been identified as lateral target signals in various integral membrane proteins, and in several proteins Tyr-containing motifs play crucial roles. The cytoplasmic domain of CEACAM1-L has two Tyr residues, Tyr488 and Tyr515, the latter being within the lateral localization motif of residues 511-518. We mutated these Tyr residues to Phe and investigated the localization of the mutants Y488F, Y515F and of the double mutant Y488,515F. Whereas Y488F expression remained lateral, both Y515F and Y488,515F localized exclusively to the apical surface. This identified Tyr515, but not Tyr488, as being important for the lateral localization of CEACAM1-L. To characterize further the lateral Tyr-dependent localization motif within residues 511-518, Val518 was mutated to Ala. This did not alter the lateral localization of CEACAM1-L. However, when investigating phosphorylation dependence (see below) we found that Val518 is important for steady state lateral localization. The deletion of the three C-terminal Lys residues (Δ518) did not cause the disappearance of CEACAM1-L from the lateral side, which suggests that they are not an important part of the localization motif. However, were the Lys residues mutated to three Ala residues a complete re-localization of CEACAM1-L to the apical surface occurred. These three Lys residues therefore play an important role in the steady state lateral localization of CEACAM1-L.

Another important amino acid for CEACAM1-L intracellular transport and function, is Ser503 (Choice et al., 1999; Fournès et al., 2001) and thus we investigated its role in lateral localization of CEACAM1-L by changing it to Ala. This mutation resulted in the complete disappearance of CEACAM1-L from the lateral surface. Ser503, therefore, plays an important role in steady state lateral localization, although it itself is not part of the C-terminal lateral localization motif.

Because both Ser503 and Tyr515 are crucial for the lateral localization of CEACAM1-L, we asked whether protein phosphorylation plays a role. To find out, confluent, polarized MDCK cells expressing rat CEACAM1-L were treated with drugs that affect protein phosphorylation, and CEACAM1-L localization was investigated by confocal microscopy. We found that pervanadate (a protein tyrosine phosphatase inhibitor) and staurosporine (a protein serine/threonine kinase inhibitor) caused the complete disappearance of CEACAM1-L from lateral surfaces (Fig. 3), whereas genistein (a protein tyrosine kinase inhibitor) and PMA (a protein kinase C stimulator) did not alter localization. CEACAM1-L remained on the apical cell surfaces after treatment with all four drugs, although the apical staining appeared sometimes weaker following treatment with staurosporine (compare Figs 3 and 5).

The disappearance of CEACAM1-L from the lateral surface caused by pervanadate and staurosporine was not because of toxic effects. Cell layers were still intact and were impermeable even after the complete disappearance of CEACAM1-L from the lateral surfaces [as demonstrated by transepithelial passage of inulin (data not shown) and by confocal microscopy of nonpermeabilized cells (Fig. 3)]. These results imply that pervanadate- and staurosporine-mediated changes in phosphorylation influence the surface localization of CEACAM1-L. We found that the time course in which CEACAM1-L disappears from the lateral surfaces following pervanadate treatment (Fig. 3) was identical to the Tyr phosphorylation of its cytoplasmic domain (Fig. 4), which indicates that direct Tyr phosphorylation of CEACAM1-L might regulate its surface localization. However, staurosporine did not indirectly induce any Tyr phosphorylation of CEACAM1-L (data not shown), which demonstrates that the lateral disappearance of CEACAM1-L induced by staurosporine was not because of direct Tyr phosphorylation of CEACAM1-L. Because appropriate antibodies against phosphorylated-Ser were not available, the kinetics of Ser phosphorylation of CEACAM1-L was not analyzed.

Ser/Thr and Tyr phosphorylation status, as well as Ser503 and Tyr515 residues and the sequence around this last Tyr residue are crucial for the lateral localization of CEACAM1-L. We therefore investigated the correlation of phosphorylation and lateral localization in more detail. The effects of pervanadate and staurosporine on the localization of some mouse CEACAM1-L cytoplasmic mutants are shown in Table 1. As described above for rat CEACAM1-L, wild-type mouse CEACAM1-L and the mutant Y488F disappeared from the lateral cell surfaces after treatment with pervanadate or staurosporine. The exclusively apical localization of Δ510, Δ495, Y515F, S503A and 3K-3A was not changed by the drug treatment. Lateral localization of V518A, however, was affected by staurosporine only, not by pervanadate, whereas lateral localization of Δ518 was not affected by either pervanadate or staurosporine. Thus, Val518 plays a regulatory role in the lateral localization of CEACAM1-L.

To investigate more directly whether the effect on CEACAM1-L localization is because Tyr phosphorylation is altered in the CEACAM1-L cytoplasmic domain itself, we determined the relative efficiency of Tyr phosphorylation in the CEACAM1-L mutants after pervanadate treatment. Only constructs that contained both Tyr488 and Tyr515 (i.e. wild-type CEACAM1-L, Δ518, S503A, V518A and 3K-3A) were analyzed to allow quantitative comparison. To have enough cells to determine CEACAM1-L phosphorylation levels accurately, cells in this experiment were grown in tissue culture dishes. Cultured this way, cells exhibited CEACAM1-L on lateral and apical surfaces, although they were not as polarized as cells grown on permeable filters. However, separate experiments showed no significant difference in the extent of Tyr phosphorylation of lateral and apical wild-type CEACAM1-L in cells polarized completely (data not shown), justifying the use of partially polarized cells in these experiments. Confluent cells were treated with pervanadate for 5 minutes, then solubilized in Triton X-100 and the lysates were immunoprecipitated with antibody against mouse CEACAM1. The immunoprecipitates were analyzed in western blots, first, with antibodies against phosphorylated-Tyr and, second, after stripping the membranes, with antibodies against CEACAM1; intensities of the CEACAM1 bands were quantified. The ratio of anti-phosphorylated-Tyr labeling to anti-CEACAM1 labeling was calculated for each CEACAM1 construct; ratios were normalized so that the phosphorylated-Tyr:wild-type CEACAM1 ratio equaled 1. Results are given in Table 2. A striking difference in the extent of Tyr phosphorylation was observed in different CEACAM1-L variants. Δ518 and V518A were significantly less phosphorylated than wild-type CEACAM1-L, whereas S503A and 3K-3A showed strikingly increased phosphorylation levels. These differences in Tyr phosphorylation correlated well with the cell surface localization in polarized cells: the less efficiently phosphorylated variants showed lateral localization with and without pervanadate treatment, whereas the more effectively phosphorylated variants localized exclusively to the apical surface even without pervanadate treatment.

Tyr515 and Val518 constitute a YxxΦ motif (in which Φ is a hydrophobic amino acid), which is known to be an important sorting signal for membrane proteins (Robinson and Bonifacino, 2001). Therefore, the less efficient Tyr phosphorylation of V518A and the unaffected lateral localization of this mutant by pervanadate strongly suggest that direct phosphorylation of Tyr515 is an important factor in the mutant's disappearance from the lateral surface. Thus, all the collected data demonstrate that Tyr515 and its phosphorylation status play a key role in the steady state lateral localization of CEACAM1-L, which suggests that the unphosphorylated state results in lateral localization, whereas phosphorylation leads to the disappearance of CEACAM1-L from the lateral surfaces.

To characterize further the disappearance of CEACAM1-L induced by pervanadate and staurosporine, we turned our interest towards PI3K, which has been implicated in vesicular trafficking and plays a key role in endocytosis as well as basolateral to apical transcytosis in MDCK cells (Hansen et al., 1995; Céfai et al., 1998). PI3K is inhibited specifically by nanomolar concentrations of wortmannin (Hansen et al., 1995). In polarized MDCK cells that were incubated with 50 nM of wortmannin, the rapid pervanadate-induced disappearance of CEACAM1-L from the lateral cell surface was blocked completely (Fig. 5). Thus, PI3K might play a key role in the lateral disappearance of CEACAM1-L induced by Tyr-phosphorylation of its cytoplasmic L-domain. Wortmannin did not affect the slower disappearance of CEACAM1-L from the lateral surface that is mediated by staurosporine (Fig. 5), which indicates that Ser/Thr phosphorylation regulates the lateral localization of CEACAM1-L by a mechanism that is distinct from the one regulated by Tyr phosphorylation.

Because phosphorylation-regulated disappearance of CEACAM1-L from the lateral cell surfaces might be important for controlling CEACAM1-mediated signaling, we were interested in the mechanisms of pervanadate- and staurosporine-induced disappearance of lateral CEACAM1-L. Confocal microscopy showed that pervanadate and staurosporine caused a transient and a more prolonged increase, respectively, of the intracellular, vesicular pool of CEACAM1-L (Figs 3, 5). To find out whether this might reflect transcytosis of CEACAM1-L from the lateral to the apical cell surface we determined ap:l+i CEACAM1-L expression. ap CEACAM1-L was determined by adding antibody against CEACAM1 to the apical surfaces before cell solubilization followed by collecting the immune complexes. The remaining CEACAM1-L, representing l+i CEACAM1-L, was then collected by immunoprecipitation from the cleared lysates. Ratios were quantified by immunoblotting and ECL. The ap:l+i ratio of untreated cells was 3.25±1.61. The ap:l+i ratios of cells treated with pervanadate for 5 minutes and staurosporine for 2 hours were 1.26±0.41 and 1.87±1.76, respectively. Thus, although a considerable variability was observed within the groups, neither pervanadate nor staurosporine induced a net lateral-to-apical transcytosis of CEACAM1-L, because this would have resulted in an ap:l+i ratio larger than that of untreated cells. Instead, the data indicate that treatment with pervanadate and staurosporine increased the intracellular pool of CEACAM1-L.

Next, we determined by quantitative immunoblotting whether any net degradation was induced by pervanadate or staurosporine. No statistically significant changes of the total cellular levels of CEACAM1-L or CEACAM1-S were observed after 15 minutes of pervanadate incubation (Fig. 6A) and no significant decrease of the total cellular levels of CEACAM1-L or CEACAM1-S during a 2 hour incubation with staurosporine was observed (Fig. 6A). If anything, 2 hours of staurosporine treatment caused a barely significant (P<0.05) increase in the expression level of CEACAM1-L (Fig. 6A). Thus, during these relatively short incubation periods no degradation of CEACAM1-L was observed. We did not extend the quantitative analyses of cellular CEACAM1 to later time points because of putative toxic effects at prolonged incubation with pervanadate and staurosporine.

To collect more data concerning the fate of CEACAM1-L after drug treatment, we double-stained the cells for CEACAM1-L and either EEA-1, the marker for early endosomes (Naslavsky et al., 2003), or LAMP-1, the marker for late endosomes/lysosomes (Eskelinen et al., 2003). Notice that endosomal vesicles are ∼1 μm in length whereas the pixel size in the confocal scanning procedure was 0.07×0.07 μm² in these determinations. Therefore, localization of CEACAM1-L to endosomal vesicles is defined as occurrence of both red (CEACAM1-L) and green (endosomal marker) pixels as well as yellow pixels within an endosomal vesicle. Because of the small pixel size (high resolution) colocalization does not necessarily show up as yellow pixels unless the different proteins are very close together in the endosomal membrane. By these criteria we could identify vesicular structures in untreated cells, in which CEACAM1-L colocalized with EEA-1 (Fig. 6B) as well as LAMP-1 (Fig. 6C). Pervanadate-treatment did not alter the extent of colocalization of CEACAM1-L and EEA-1, whereas staurosporine-treatment increased the amount of structures that were positive for both CEACAM1-L and EEA-1 (Fig. 6B). A striking increase in vesicular structures that were positive for CEACAM1-L and LAMP-1 was caused by both pervanadate and staurosporine, the effect of staurosporine being more prominent (Fig. 6C).

To characterize vesicular, intracellular CEACAM1-L further we focused on the adaptor protein complexes AP-1 and AP-2, components of clathrin-coated vesicles. AP-1 is associated with sorting to and from the trans-Golgi network and AP-2 mediates internalization of coated vesicles at the plasma membrane (Boehm and Bonifacino, 2001; Robinson and Bonifacino, 2001). First, we investigated partially polarized cells that were grown on coverslips. On the plasma membrane and in AP-2-containing vesicles close to the plasma membrane CEACAM1-L occurred together with AP-2, whereas in central and perinuclear regions it was associated with AP-1-containing vesicles (Fig. 7A). This indicated that CEACAM1-L is subjected to continuous intracellular traffic, in which it is sorted in AP-1-containing vesicles at the trans-Golgi network and internalized in AP-2-containing vesicles at the plasma membrane. We then analyzed completely polarized cells that were treated with pervanadate or staurosporine. Pervanadate treatment did not change colocalization of AP-2-containing vesicles (data not shown). However, a striking accumulation in AP-1-containing vesicles was seen, that was most pronounced 2 minutes after adding the drug (Fig. 7B). The opposite was observed after treating the cells with staurosporine. No accumulation of CEACAM1-L was seen in AP-1-positive vesicles (data not shown), but clusters of vesicles with closely associated CEACAM1-L and AP-2 occurred 60 minutes after the addition of staurosporine (Fig. 7B). Together with the observed EEA-1 and LAMP-1 staining patterns, these results suggest that pervanadate causes an increased routing of CEACAM1-L from the trans-Golgi network to late endosomes and lysosomes, and that staurosporine stimulates internalization of CEACAM1-L from the plasma membrane to endosomes and lysosomes.

Because CEACAM1-L participates in homophilic cell-cell adhesion at the lateral surfaces of polarized MDCK cells, it was of interest to investigate how it relates to and possibly affects other adhesive molecules and cell junctions in the lateral membranes. We therefore analyzed the organization of the following adhesion and adhesion-related proteins and their colocalization with CEACAM1-L: the tight junction proteins occludin and ZO-1; E-cadherin and β-catenin, which occur in the lateral cell surfaces and are concentrated in adherens junctions; desmoplakin, a component of desmosomal plaques; and cytokeratin filaments, which insert into desmosomal plaques.

Overexpression of CEACAM1-L did not perturb the localization or organization of the tight junction proteins occludin and ZO-1. Both proteins were present in well-organized tight junctions that ran like a belt around the cell periphery in the apical parts of confluent, polarized cells. There was no staining for CEACAM1-L within the regions that stained for occludin and ZO-1 (Fig. 8), and we therefore conclude that CEACAM1-L was not localized to the tight junctions.

Localization and organization of E-cadherin and β-catenin in confluent, polarized MDCK cells were the same in untransfected cells and in cells that were stably transfected with CEACAM1-L. E-cadherin and β-catenin occurred close together over the entire lateral cell surface all the way from the tight junctions to the basal surface, with a higher concentration towards the apical part in the region of adherens junctions (Fig. 8). CEACAM1-L was also distributed over the entire lateral cell surface (Fig. 8). In the lateral surfaces below the adherens junctions, E-cadherin and β-catenin occurred in discrete clusters that were separated from CEACAM1-L. However, in the apical regions of the lateral membranes CEACAM1-L was located so close to E-cadherin and β-catenin that a large proportion of the voxels (0.07×0.07×0.3 μm³) were double-stained (yellow) (Fig. 8). We conclude that CEACAM1-L localized together with E-cadherin and β-catenin to the adherens junctions, but not to the rest of the lateral surfaces.

In contrast to the unaffected tight junctions and adherens junctions, CEACAM1-L perturbed the organization of desmosomes in the lateral surfaces of confluent, polarized MDCK cells (Fig. 9A). Desmosomes in untransfected cells, as visualized by staining for desmoplakin, were seen as discrete spot-like entities in the lateral surfaces. However, they were not randomly distributed but appeared as two populations. One population was found just beneath the adherens junction, where they appeared as a circumferential band of dense and closely associated spots. The other population consisted of scattered spots, randomly distributed in the remainder of the lateral surface. Abundance and organization of desmosomal plaques were not perturbed by expression of CEACAM1-S but the number of desmosomes decreased significantly in cells transfected with both rat and mouse CEACAM1-L (Fig. 9A). The population of randomly distributed desmosomes decreased dramatically, whereas the apical belt of desmosomes remained although they appeared to be smaller and less distinct than those in untransfected cells (Fig. 9A). The perturbation of desmosomal abundance and organization was observed only in the lateral surfaces of adjacent cells that expressed CEACAM1-L and did not occur in lateral contact areas of CEACAM1-L-expressing and non-expressing cells. No close colocalization of CEACAM1-L and desmoplakin was observed in the remaining desmosomal plaques (Fig. 9A).

The desmosomal plaques in the randomly distributed lateral population were quantified by counting several different 5×5 μm squares in the lateral surfaces. The number of desmosomal plaques decreased from an average of 19 plaques per 25 μm² in untransfected and CEACAM1-S-transfected cells to an average of 4.5 plaques per 25 μm² in cells expressing rat or mouse CEACAM1-L (Fig. 9B). These differences were highly significant. To learn more about the regulation of desmosomal plaque organization by CEACAM1-L, we investigated several of the deletion and mutant variants of mouse CEACAM1-L for their effects on desmosomal plaque abundance. This was done by quantifying the number of desmosomal plaques in the lateral surfaces and is shown in Fig. 9B. By increasing the deletion of parts of the cytoplasmic domain, the desmosomal regulating sequence was mapped to the C-terminal part distal of amino acid 483. Within this sequence Ser503 plays some role, because compared with wild-type CEACAM1-L the amount of desmosomal plaques was significantly less reduced when this amino acid was mutated. However, the most striking effect was seen after the mutation of Tyr515, causing the complete disappearance of the desmosomal perturbation effect of CEACAM1-L. Mutation of Tyr488 had no effect, demonstrating that Tyr515, but not Tyr488, can perturb desmosomal organization. The three C-terminal Lys residues, Lys519, Lys520 and Lys521, did not play any role in the desmosomal perturbation sequence, because construct 3K-3A was as effective as wild-type CEACAM1-L in reducing the number of desmosomal plaques. The effective perturbation by this construct is interesting because it demonstrates that CEACAM1-L does not need to be localized in the lateral surface domain to exert its effect on desmosomal organization.

Desmosomes are attached to cytokeratin filaments on the cytoplasmic face of the plasma membrane. Therefore, we analyzed the structure and organization of cytokeratin filaments in cells expressing various forms of CEACAM1. Stained by a pan-cytokeratin antibody the cytokeratins in untransfected cells occurred as distinct bundles of filaments running parallel to the plasma membrane in the basolateral regions (Fig. 9C). Little cytokeratin staining was seen in the more central regions of the cytoplasm. The same pattern was seen in polarized CEACAM1-S-transfected MDCK cells. However, in CEACAM1-L-transfected cells the submembraneous cytokeratin bundles disappeared and increased unstructured staining for cytokeratin deeper in the cytoplasm was observed. Thus, CEACAM1-L caused a disorganization of the submembraneous cytokeratin filament bundles in the polarized epithelial cells.

We demonstrated previously that CEACAM1-L and CEACAM1-S are localized to different regions of MDCK cells. CEACAM1-S was exclusively localized to the microvilli of the apical cell surfaces, whereas CEACAM1-L was localized to the apical microvilli and to the lateral cell surfaces of confluent, polarized cells (Sundberg and Öbrink, 2002). Because CEACAM1-L and CEACAM1-S differ only in their cytoplasmic domains, this suggested that the signals for lateral localization of CEACAM1-L reside in its cytoplasmic domain. However, we also found previously that the maintenance of lateral localization of CEACAM1-L required homophilic transbinding of CEACAM1-L molecules presented by adjacent cells (Sundberg and Öbrink, 2002) demonstrating that transfected rodent CEACAM1-L does not bind to endogenous canine CEACAM1, and that lateral targeting signals in the cytoplasmic domain are necessary but not sufficient for steady state lateral localization of CEACAM1-L. In the present investigation we characterized cytoplasmic domain amino acid motifs that are crucial for lateral targeting. In addition to identifying such motifs, we also found that changes in the phosphorylation state correlated with the rapid disappearance of CEACAM1-L from the lateral surfaces, even when close intercellular contacts were maintained. Hence, the steady state lateral localization of CEACAM1-L is the result of a balance between transport to the lateral cell surface because of targeting signals in the cytoplasmic domain, maintenance of the lateral localization because of homophilic binding of neighboring cells, and transport to other cellular locations regulated by phosphorylation.

A large range of lateral localization signals have been found (He et al., 2002; Moll et al., 2001; Fultz et al., 2001; Martens et al., 2000; Rodionov et al., 2000; Nemer et al., 1999; Simonsen et al., 1999; Le Gall et al., 1997; Sheikh and Isacke, 1996) and some proteins carry more than one lateral targeting signal (He et al., 2002; Rodionov et al., 2000). Two of the more common lateral localization signals contain Tyr residues (Moll et al., 2001; Fultz et al., 2001; Martens et al., 2000; Rodionov et al., 2000) and di-Leu motifs (Rodionov et al., 2000; Nemer et al., 1999; Simonsen et al., 1999). Using CEACAM1-L variants that differed in their cytoplasmic domain, we identified a lateral targeting sequence in the C-terminal part of the CEACAM1-L cytoplasmic domain, in which Tyr515 has a crucial role. The other Tyr residue, Tyr488, did not contribute to lateral targeting. Tyr515 occurs in the amino acid sequence YxxΦ, which was found to be an important signal for sorting proteins into clathrin-coated vesicles (Robinson and Bonifacino, 2001).

Tyr phosphorylation induced by pervanadate treatment led to the rapid disappearance of CEACAM1-L from lateral surfaces, suggesting that one important factor for the role of Tyr515 in CEACAM1-L sorting is its phosphorylation state. This assumption was supported by the findings that pervanadate treatment did not abolish the lateral localization when Val518 (which is part of the YxxΦ motif) was mutated to Ala, and that this mutation significantly reduced the Tyr phosphorylation of CEACAM1-L.

Mutation of Ser503 and of the three C-terminal Lys residues also caused the disappearance of CEACAM1-L from the lateral cell surfaces. The Ser and Lys mutations strongly facilitated Tyr phosphorylation of CEACAM1-L, suggesting that the disappearance of these CEACAM1-L mutants from the lateral cell surface might be mediated by effects on the Tyr515 motif.

Interference of Ser/Thr phosphorylation by staurosporine also regulated the lateral localization of CEACAM1-L. However, the finding that low concentrations of wortmannin completely blocked pervanadate-induced but not staurosporine-induced disappearance of CEACAM1-L from the lateral surfaces, indicated that these two drugs affected different parts of the sorting machinery for CEACAM1-L. Moreover, staurosporine did not indirectly stimulate Tyr phosphorylation of CEACAM1-L. We could not link the staurosporine effect to the phosphorylation of specific Ser/Thr residues in CEACAM1-L, although the lateral disappearance induced by mutation of Ser503 indicated that phosphorylation of this Ser residue might be a potential lateral targeting signal. However, the C-terminal deletion mutant Δ510, lacking Tyr515 but having Ser503 intact, localized exclusively apically, suggesting that the effect of Ser503 on the CEACAM1-L surface localization might result from its influence on the Tyr phosphorylation of CEACAM1-L, as discussed above.

The finding that CEACAM1-L occurred in centrally localized AP-1-associated vesicles and in AP-2-associated vesicles at the plasma membrane of confluent cells indicated that CEACAM1-L is subject to a rapid, continuous, clathrin-mediated transport to and from the plasma membrane. This traffic was perturbed by pervanadate and staurosporine, both of which increased the abundance of vesicular CEACAM1-L. However, as already suggested these two drugs apparently acted on different parts of the CEACAM1-L intracellular trafficking. Staurosporine increased the association of CEACAM1-L with early endosomes and AP-2-containing vesicles. Pervanadate increased the association of CEACAM1-L with AP-1-containing vesicles but not with early endosomes. Both drugs caused an accumulation of CEACAM1-L in late endosomes and/or lysosomes but no increased degradation was observed during the time periods studied. Thus, pervanadate presumably affects sorting at the level of the trans-Golgi network, resulting in decreased transport to the cell surface and increased routing to late endosomes/lysosomes. Staurosporine, however, seemed to increase the uptake from the plasma membrane. No net transcytosis of CEACAM1-L from the lateral cell surfaces to the apical cell surfaces was observed.

CEACAM1 functions both in cell adhesion and signal transduction. This might reflect two sides of the same coin, because more and more data show that CEACAM1-mediated homophilic adhesion can control CEACAM1-mediated signaling (Donda et al., 2000; Markel et al., 2002; Greicius et al., 2003). This might also be the case in polarized epithelial cells because laterally localized CEACAM1-L is involved in homophilic transbinding (Sundberg and Öbrink, 2002). Important trigger of CEACAM1 signaling pathways are hormone and growth factor receptor-mediated phosphorylation of CEACAM1. The present findings that changes in Tyr and Ser/Thr phosphorylation result in decreased expression of CEACAM1-L at the lateral surfaces point to yet another way of controlling the signaling activities of CEACAM1-L.

CEACAM1 not only mediates cell-cell adhesion by trans-homophilic binding but also regulates other adhesive systems, e.g. integrin-mediated adhesion in granulocytes (Skubitz et al., 2000) and B lymphocytes (Greicius et al., 2003). Therefore, it was of interest to investigate the relationship of laterally localized CEACAM1-L to other epithelial cell adhesion systems, particularly intercellular junctions. CEACAM1-L was excluded from tight junctions and did not affect their integrity or alter the localization of occludin or ZO-1. This was in contrast to its topographical relationship to E-cadherin and β-catenin. Both E-cadherin and β-catenin were distributed over the entire lateral surface of polarized MDCK cells and at high concentrations in the adherens junctions. CEACAM1-L did not affect the localization of E-cadherin or β-catenin. CEACAM1-L occurred in the entire lateral surface and showed close colocalization with E-cadherin and β-catenin in the region of the adherens junction. It was found previously that epithelial adherens junctions also contain the homophilic immunoglobulin superfamily adhesion molecule nectin (Takahashi et al., 1999). The finding of CEACAM1-L in adherens junctions now adds another immunoglobulin superfamily molecule to the group of proteins that contribute to intercellular adhesion in this specialized part of the lateral surfaces of polarized epithelial cells. Like E-cadherin, CEACAM1-L was also found to associate with actin filaments (Schumann et al., 2001).

The unperturbed organization of occludin, ZO-1, E-cadherin and β-catenin by CEACAM1-L contrasted with its effects on desmosomes. Interestingly, CEACAM1-L did not have to be localized to the lateral cell surfaces in order to perturb desmosomal organization, because some of the mutants that exhibited exclusive apical localization were as effective desmosome disorganizers as wild-type CEACAM1-L. The most probable explanation of the decreased abundance of lateral desmosomes and concomitant disorganization of cytokeratin filaments is that the cytoplasmic domain of CEACAM1-L competes for proteins that are important for desmosomal and cytokeratin organization. The binding site for these putative proteins was mapped to amino acids 484-518 in the C-terminal part of the cytoplasmic domain, among which the Tyr515 residue played a crucial role.

A reciprocal relationship of the expression levels of CEACAM1-L and desmosomes was found previously in other cell types, e.g. in the bladder-derived epithelial NBT-II cells. Confluent NBT-II cells cultured with fetal calf serum abundantly express desmosomes and moderate levels of CEACAM1. When these cells were cultured with a serum-substitute that causes epithelial-to-mesenchymal transition, both desmosomal internalization (Boyer et al., 1989) and increased expression of CEACAM1-L (Hunter et al., 1994) were observed. At the same time, the cells became more motile but still kept together in loosely associated groups (Hunter and Öbrink, unpublished observations). In vivo, a similar situation was found in the junctional epithelium of the gingiva (Heymann et al., 2001). This epithelium, which is directly associated with the teeth, distinguishes itself from the rest of the gingival epithelium. It consists of loosely associated cells that have significantly less desmosomes than the surrounding epithelium. At the same time they express abundant CEACAM1, which is not at all expressed in the surrounding epithelium (Heymann et al., 2001). A continuous migration of leukocytes from the blood to the oral cavity occurs in the loosely connected epithelial cells of the junctional epithelium. Thus, it seems that increased CEACAM1-L expression, at the expense of decreased abundance of desmosomes, promotes a more dynamic type of intercellular adhesion allowing increased cell motility. Notice, that not only CEACAM1 but also intercellular junctions are dynamic structures that are constantly turned over. However, the balance of CEACAM1 expression and expression of other cell adhesion proteins might be important to control cellular motility within an epithelium. Further studies of the reciprocal interactions of CEACAM1 and desmosomes, and also of the putative role of CEACAM1-L in adherens junctions, will shed more light on these cellular processes.

This work was supported by grants from the Swedish Research Council (project no 05200), the Swedish Cancer Foundation (project no 4720), the Canadian Institutes of Health Research (grant no 13911), the Karolinska Institutet, and Polysackaridforskning AB.