The Conserved Immunoglobulin Domain Controls the Subcellular Localization of the Homophilic Adhesion Receptor Protein-tyrosine Phosphatase μ*

The receptor protein-tyrosine phosphatase μ (PTPμ) is a homophilic adhesion protein thought to regulate cell-cell adhesion in the vascular endothelium through dephosphorylation of cell junction proteins. In subconfluent cell cultures, PTPμ resides in an intracellular membrane pool; however, as culture density increases and cell contacts form, the phosphatase localizes to sites of cell-cell contact, and its expression level increases. These characteristics of PTPμ, which are consistent with a role in cell-cell adhesion, suggest that control of subcellular localization is an important mechanism to regulate the function of this phosphatase. To gain a better understanding of how PTPμ is regulated, we examined the importance of the conserved immunoglobulin domain, containing the homophilic binding site, in control of the localization of the enzyme. Deletion of the immunoglobulin domain impaired localization of PTPμ to the cell-cell contacts in endothelial and epithelial cells. In addition, deletion of the immunoglobulin domain affected the distribution of PTPμ in subconfluent endothelial cells when homophilic binding to another PTPμ molecule on an apposing cell was not possible, resulting in an accumulation of the mutant phosphatase at the cell surface with a concentration at the cell periphery in the region occupied by focal adhesions. This aberrant localization correlated with reduced survival and alterations in normal focal adhesion and cytoskeleton morphology. This study therefore illustrates the critical role of the immunoglobulin domain in regulation of the localization of PTPμ and the importance of such control for the maintenance of normal cell physiology.

The physical interactions of individual cells with the local environment is critical to the control of their growth, differentiation, and fate within a multicellular organism. These responses are initiated by cell adhesion receptors, which bind either homophilic or heterophilic ligands on apposing cells or the extracellular matrix. A critical aspect in the cellular response to these interactions is reversible protein tyrosine phosphorylation of cell junction-associated proteins (1,2). In fact, the bulk of the tyrosine-phosphorylated cellular proteins are localized to sites of cell-cell and cell-extracellular matrix adhesion. Several protein-tyrosine kinases and phosphatases have been implicated in the regulation of cell adhesion (3)(4)(5)(6). In particular, receptor protein-tyrosine phosphatases (RPTPs), 1 several of which possess extracellular domains with features characteristic of cell adhesion receptors, linked to a cytoplasmic PTP activity, are uniquely designed to mediate cellular responses to adhesive signals through protein tyrosine dephosphorylation (7,8). Of particular interest is the potential for the extracellular domains of individual RPTPs to mediate interactions with, and control of, either cell-cell or cell-extracellular matrix adhesions. In this way, the targeting of RPTPs to specific junctions may be a key element in controlling the changes in cell junction-actin cytoskeleton interactions that occur as cells transition from a growing, migrating state, in which actin stress fibers are linked to cell-matrix junctions, to contactinhibited cells, in which actin fibers are linked primarily to cell-cell junctions (9).
In this study, we focus on the receptor protein-tyrosine phosphatase PTP. PTP is synthesized as a single polypeptide chain, which is then glycosylated and proteolytically cleaved at a site N-terminal to the transmembrane segment to give two subunits that remain noncovalently associated (10,11). The E-subunit (see Fig. 1A), which constitutes the majority of the extracellular segment, consists of one immunoglobulin (Ig) domain; four fibronectin type III repeats; and a MAM (Mu, A5, and meprin homology) domain, which is a conserved domain unique to PTP, PTP, PTP, and PTP and some cell adhesion molecules (12). The P-subunit (see Fig. 1A) contains a short extracellular segment; the transmembrane segment; and the entire cytoplasmic portion, including two PTP domains. PTP can function as a homophilic adhesion receptor whereby a PTP molecule on one cell can bind to an identical PTP molecule on an apposing cell (13,14). It is this binding that is proposed to direct the localization of PTP to sites of cell-cell contact, where it is thought to interact with the cadherincatenin cell adhesion complex (11,15). Consistent with a role in cell-cell adhesion, the level of PTP protein increases as intercellular contacts are made and cell-cell junctions form (11). This phenomenon is seen with other RPTPs and may provide a means to control protein phosphotyrosine-based signaling in contact-inhibited cells (15)(16)(17)(18). PTP is expressed predominantly in vascular endothelial cells and to a lesser extent in bronchial epithelium and may play a role in endothelium-dependent processes such as the control of vascular permeability and the angiogenic responses to tumor formation and wound healing (19).
Considering the importance of PTPs in the control of signal transduction, it is critical that they be tightly regulated to facilitate protein tyrosine phosphorylation (20). Based on studies with RPTP␣, dimerization of RPTPs, which leads to occlusion of the active site, has been proposed as one mechanism for inhibition of phosphatase activity (21)(22)(23). However, it is not clear whether dimer-induced inhibition is widely applicable, as dimers formed in crystals of RPTP-LAR and RPTP are oriented such that the active sites are unobstructed (24,25). A number of studies indicate that the control of localization and substrate accessibility by intrinsic targeting domains is another important way to modulate cellular dephosphorylation by PTPs (20) such as PTP. In contrast to the structurally related LAR-subtype RPTPs (LAR, PTP, and PTP␦), which can be found in both cell-matrix and cell-cell contacts (26 -28), PTP is thought to function exclusively at sites of cell-cell adhesion, where it has the potential to regulate signaling events specific to contact-inhibited cells. In subconfluent cells, PTP is thought to cycle between the plasma membrane and membrane vesicles in the perinuclear region of the cell (11,15). Although it has been proposed that stable surface expression is dependent upon homophilic binding at cell-cell contacts (11), the mechanism whereby PTP localization is controlled has not been defined.
Previous studies using purified fusion proteins indicate that the conserved Ig domain is necessary and sufficient for homophilic binding of purified PTP fusion proteins as well as binding of the fusion proteins to the surface of cells expressing endogenous levels of PTP (10). Here, we have deleted the Ig domain of PTP and expressed the mutant protein in endothelial and epithelial cell lines to assess the role of this domain in regulating localization of the phosphatase. The results demonstrate the importance of the Ig domain for the localization of PTP to sites of cell-cell contact and also identify a role for this domain in localization of PTP in subconfluent cells in the absence of cell-cell contact.
Construction of Retroviral Expression Plasmids-Human PTP (GenBank TM /EBI accession number NM_002845) DNA constructs were cloned with a C-terminal T7 epitope tag (30) into the retroviral vector pWZL-hygro (provided by S. Lowe, Cold Spring Harbor Laboratory) by PCR with pBSK/hFL (31) as the template. pWZL/PTPWT contains nucleotides 1-4359 of human PTP; pWZL/PTP⌬Ig and pWZL/ PTP⌬E have deletions of nucleotides 568 -930 and 67-1914, respectively, in human PTP. The resulting constructs encoded proteins with a C-terminal extension of AGMASMTGGQQMG, where the underlined amino acids define the T7 epitope tag. All constructs derived from PCR were confirmed by DNA sequencing.
Expression of PTP by Retroviral Infection-LiNX-A retroviral packaging cells (32) were transfected with pWZL/PTP plasmid DNA using calcium phosphate. The resulting supernatants, containing recombinant retrovirus encoding PTP, were used to infect monolayer cultures of GM7372 cells, trBAEC, or MDCK cells. Two days following infection, cells were placed in the appropriate medium supplemented with 100 g/ml (GM7372 cells) or 400 g/ml (trBAEC and MDCK cells) hygromycin B (Invitrogen). The medium was changed every 2-3 days, and cultures were passaged when they reached ϳ70 -90% confluence.
Indirect Immunofluorescence-Cells were plated on glass coverslips, grown to the desired density, and fixed with 2% paraformaldehyde in phosphate-buffered saline for 12.5 min. Cells were either permeabilized in phosphate-buffered saline, 3% normal goat serum, and 0.2% Triton X-100 for 20 min or left intact as indicated. Fixed samples were incubated either with single primary antibodies or, for co-staining, with two antibodies of different isotypes. The primary antibodies were detected with isotype-specific secondary antibodies (Molecular Probes, Inc., Eugene, OR) to mouse IgG1 (labeled with Alexa 594) or to mouse IgG2b or IgG2a (labeled with Alexa 488).
Co-staining with antibodies SK7 and 45C6, both of which are IgG1 isotype, was performed as follows. Fixed cells were first incubated with antibody SK7, followed by a 20-fold excess of Fab fragments of rabbit anti-mouse IgG and then Texas Red-labeled anti-rabbit IgG. Then the cells were incubated with antibody 45C6, followed by fluorescein isothiocyanate-labeled rabbit anti-mouse IgG. A control sample in which antibody 45C6 was omitted did not show staining with the fluorescein isothiocyanate-labeled anti-mouse IgG, indicating that the excess, Fab fragment rabbit anti-mouse IgG, effectively blocked antibody SK7 from reaction with this secondary antibody.
Coverslips were mounted on glass slides using ProLong anti-fade reagent (Molecular Probes, Inc.) and examined by epifluorescence using an Axiophot 2 microscope and digital imaging system with a ϫ63 objective lens (Carl Zeiss Microimaging, Thornwood, NY) unless indicated otherwise. Images were processed using OpenLab software (Improvision Software, Lexington, MA). Confocal images were obtained with a Zeiss LSM510 confocal microscopy system using a ϫ100 objective lens and processed with built-in software and ImageJ software (National Institute of Mental Health, available at rsb.info.nih.gov/ij/).
Immunoblotting and Immunoprecipitation-GM7372 cells were grown to the desired density and solubilized in lysis buffer (30 mM HEPES (pH 7.5), 150 mM NaCl, 1% Triton X-100, 8% glycerol, 1 mM EDTA, 1 mM benzamidine, and 5 g/ml each aprotinin and leupeptin) for 30 min. Lysates were centrifuged at 10,000 ϫ g for 15 min, and the resulting supernatants were assayed for protein concentration with Bradford reagent using bovine serum albumin as the standard. Equal amounts of protein were separated by SDS-PAGE and transferred to nitrocellulose membranes for immunoblotting with anti-PTP monoclonal antibodies SK7 and BK2 as described (15). For immunoprecipitations, equal amounts of protein from GM7372 cell lysates, prepared as described above, were incubated with anti-T7 epitope antibody-conjugated agarose (Novagen) for 2 h at 4°C. Samples were washed four times in lysis buffer, and bound proteins were eluted with Laemmli sample buffer, separated by SDS-PAGE, and analyzed by immunoblotting.
Assay of Cell Proliferation and Apoptosis-Proliferation rates of cell lines were assayed by measuring incorporation of 5-bromo-2-deoxyuridine (BrdUrd). Cells were plated on glass coverslips at a density of 4 ϫ 10 3 cells/cm 2 , and 36 h later, 20 M BrdUrd was added for 40 min. Cells were stained for BrdUrd incorporation using fluorescein isothiocyanate-conjugated anti-BrdUrd antibody (Roche Applied Science) according to the manufacturer's protocol. Samples were co-stained with Hoechst 33342 (Molecular Probes, Inc.) to visualize nuclei of all cells and imaged by epifluorescence using a ϫ10 objective. Total and BrdUrdpositive cells were counted in images using ImageJ software, and the percentage of BrdUrd-positive cells was calculated.
To assay for apoptosis, cells were plated on glass coverslips at 4 ϫ 10 3 cells/cm 2 and incubated at 37°C for 36 h. Cells were fixed as described above for immunofluorescence and stained with Hoechst 33342 according to the supplier's protocol. Cells displaying condensed chromatin and the formation of nuclear blebs containing DNA were counted as apoptotic. The percentage of apoptotic cells in six representative fields of each sample was calculated as described above for the assay of BrdUrd incorporation.

Generation of Stable Lines of GM7372 Cells Expressing PTP Deletion
Constructs-As an initial step in understanding the mechanisms involved in regulating PTP function in endothelial cells, we expressed various mutant constructs in GM7372 cells, a cell line derived from BAEC. GM7372 cells have similar morphology and growth properties compared with primary BAEC and express endothelium-specific cell junction proteins ( Fig. 1) (data not shown) (29). GM7372 cells expressed a similar level of endogenous PTP compared with the parental BAEC ( Fig. 1B). In agreement with previous findings (15,19), PTP was observed as a full-length ϳ200 kDa protein (Fig. 1B, EP) and as processed proteolytic fragments of ϳ110 kDa for the E-subunit (Fig. 1B, E) and P-subunit (data not shown). As described previously (19), at high cell density, PTP co-localized at cell-cell contacts with the endothelial-specific VE-cadherin, and the level of PTP protein was increased (Fig. 1, B and C). In subconfluent cells, PTP was found in a perinuclear vesicular compartment (Fig. 1C).
Ectopic expression of PTPWT, PTP⌬Ig (deletion of amino acids 190 -310), and PTP⌬E (deletion of amino acids 23-638) was achieved by infection of GM7372 monolayers with recombinant retrovirus encoding the T7 epitope-tagged PTP constructs. We utilized a vector, pWZL, containing an internal ribosomal entry site, from which PTP and the selection marker are separately translated from the same mRNA. Stable cells were selected in hygromycin as pooled populations from an ϳ90% confluent monolayer of infected GM7372 cells. Attempts to isolate clonal cell lines of GM7372 expressing PTP from dilute cultures were unsuccessful. We reasoned that any inhibitory effects on growth or survival caused by ectopic expression of PTP might be more pronounced at low cell density, when the normal level of PTP is low, than at high cell density, when expression of the phosphatase is naturally up-regulated.
The use of the pWZL vector resulted in a tight correlation between hygromycin resistance and ectopic expression of PTP, producing cell populations in which 70 -80% of the cells expressed detectable amounts of the specific PTP mutant, as determined by detection with anti-PTP antibodies (data not shown).
Immunoblotting of cell lysates with antibodies specific to the P-subunit (SK7) and the E-subunit (BK2) indicated similar levels of expression from each construct (Fig. 2). Furthermore, the wild-type EP protein migrated as a doublet on SDS-polyacrylamide gel, with the slower migrating band, previously shown to represent the glycosylated form (11), predominating. In contrast, the hypoglycosylated form of PTP⌬Ig predominated, which would contribute to the faster than expected migration of the E-subunit of this mutant on SDS-polyacrylamide gel. In addition, immunoprecipitation of the P-subunit of PTP⌬Ig with the antibody to the C-terminal T7 epitope followed by immunoblotting of the E-subunit with antibody BK3 demonstrated that the two intact segments of the processed protein were associated in the cell, as seen with the WT protein (Fig. 2).
The Ig Domain Is the Major Determinant of Localization of PTP to Sites of Cell-Cell Contact-It has been proposed that Pulse-chase labeling studies suggested that the lower band(s) for the EP protein represents the initial protein product prior to glycosylation and subsequent processing. The migration positions of relative molecular mass markers (in kilodaltons) are shown to the right of each panel. C, subconfluent GM7372 cells were fixed and stained with anti-PTP antibody SK7 (upper panel), and confluent cells were stained sequentially with antibody SK7 (middle panel) and anti-VE-cadherin antibody 45C6 (cad; lower panel) as described under "Experimental Procedures." Scale bar ϭ 10 m.

FIG. 2. Analysis of PTP proteins expressed in GM7372 cells.
Lysates from stable populations of confluent GM7372 cells expressing PTPWT, PTP⌬Ig, PTP⌬E, or the empty pWZL-hygro vector (V) were analyzed by immunoblotting with anti-P-subunit antibody SK7 (left panel) or anti-E-subunit antibody BK2 (right panel, lysate lanes). 18 g of protein were loaded for each lysate. Immunoblotting conditions were chosen to demonstrate the overexpression of the ectopic proteins relative to endogenous PTP. The positions of the non-proteolytically processed protein (EP) and the processed P-subunit (P) and E-subunits of the PTPWT (E(WT)) and PTP⌬Ig (E(⌬Ig)) proteins are indicated to the left of the panels. Pulse-chase labeling studies suggested that the lower band(s) for the EP protein represents the initial protein product prior to glycosylation and subsequent processing. The band at 116 kDa in the PTP⌬Ig samples, which also appears in the vector samples, is from the endogenous PTP E-subunit. Association of the two subunits of PTPWT or PTP⌬Ig was demonstrated by immunoprecipitating the P-subunit from lysates of GM7372 cells with the anti-T7 epitope antibody and detecting the associated E-subunit by immunoblotting the immune complexes with antibody BK3 (right panel, IP lanes). The migration positions of the E-subunits from the PTPWT and PTP⌬Ig proteins are indicated to the left of the right panel. The migration positions of relative molecular mass markers (in kilodaltons) are indicated to the right of each panel.
homophilic binding may be responsible for localization of PTP to cell-cell contacts (10,11). To test this idea in a cellular context, GM7372 cells expressing T7 epitope-tagged PTP mutants were analyzed by indirect immunofluorescence with the anti-T7 epitope antibody to determine the subcellular localization of the phosphatase. Full-length PTP-WTT7 localized to cell-cell contacts as indicated by co-localization with VE-cadherin (Fig. 3, WT panels). Deletion of the Ig domain, containing the homophilic binding site, greatly reduced the localization of PTP to cell-cell contacts (Fig. 3, ⌬Ig panels), resulting in a dispersed pattern consistent with expression at the cell surface. Therefore, the Ig domain, which contains the homophilic binding site (10), is a major determinant for localization to cell-cell contacts, suggesting that homophilic binding is important for this localization. Deletion of the entire E-subunit, containing most of the extracellular domain, completely eliminated localization of PTP to cell-cell contacts (Fig. 3, ⌬E  panels), suggesting that this region may contain sequences in addition to the Ig domain that play a part in this localization. A role has been proposed for the MAM domain in homophilic binding (11); however, we were unable to determine whether the MAM domain may contribute to the localization of PTP because constructs with a deletion of this domain could not be stably expressed at levels comparable with the other mutants (data not shown). This is consistent with studies of the metalloprotease meprin A, which have suggested that the MAM domain may be critical for protein stability (34).
To verify that the localization of PTP⌬Ig in GM7372 cells is a property of the phosphatase and not a phenomenon unique to these cells, we examined the localization the PTP deletion constructs in two additional cell types, a transformed endothelial cell line trBAEC and MDCK epithelial cells, the latter being a well established model for studying cell-cell adhesion. In trBAEC and MDCK cells, as seen in GM7372 cells, deletion of the Ig domain resulted in reduced localization to cell-cell contacts and a more dispersed distribution over the cell compared with the WT protein (Fig. 4). These results confirm that the Ig domain is a major determinant for the localization of PTP to cell-cell contacts in multiple cell types.
The PTP⌬Ig Protein Is Present at the Cell Surface-As described above, deletion of the Ig domain led to a widespread distribution of PTP over what appeared to be the cell surface. To confirm that the PTP⌬Ig protein was present at the cell surface, we used the extracellular domain-specific antibody BK2 to stain GM7372 cells that had not been permeabilized prior to antibody staining. PTPWT was detected primarily at cell-cell contacts in intact non-permeabilized cells, whereas PTP⌬Ig was dispersed over the surface of the cell (Fig. 5). Staining of permeabilized cells revealed additional localization of both PTPWT and PTP⌬Ig in a perinuclear region (Fig. 5) as described for the endogenous protein (Fig. 1). We were unable to determine whether PTP⌬E can reach the cell surface, as it did not react with the available antibodies directed against the extracellular region of PTP.
To analyze the localization of the mutant PTP proteins in FIG. 3. Deletion of the Ig domain reduces localization of PTP to sites of cell-cell contact in GM7372 cells. GM7372 cells expressing the empty pWZL-hygro vector, T7 epitope-tagged PTPWT, PTP⌬Ig, or PTP⌬E were grown to confluence, fixed, and co-stained with antibodies to the T7 epitope and VE-cadherin (cad). Green (left panels) and red (right panels) channels for each field are shown side by side. Arrowheads indicate examples of PTP-T7 and VE-cadherin localization at sites of cell-cell contact. Nuclear staining is seen in all samples stained with the anti-T7 epitope antibody, including cells expressing the empty pWZL vector, and therefore is not representative of PTP-T7 localization. Scale bar ϭ 10 m.

FIG. 4. Localization of ectopic PTP in trBAEC and MDCK cells. Stable lines of trBAEC and MDCK cells expressing the empty pWZL-hygro vector (V)
, T7 epitope-tagged PTPWT, or PTP⌬Ig constructs were grown to confluence, fixed, and stained with the anti-T7 epitope antibody. Nuclear staining is seen in all samples, including cells expressing the empty pWZL vector, and therefore is not representative of PTP-T7 localization. Arrowheads indicate examples of PTP-T7 localization at areas of cell-cell contact (PTPWT and PTP⌬Ig). Scale bars ϭ 10 m.
further detail, we examined MDCK cells, which, in contrast to GM7372 endothelial cells, form a monolayer in culture with sufficient depth to visualize cell morphology in three dimensions using confocal microscopy. Cross-section imaging of MDCK cells expressing full-length PTPWT and PTPWT⌬D2, a mutant lacking the membrane distal catalytic domain, stained with an antibody directed against the P-subunit of PTP, indicated that these proteins, which contain an intact E-subunit, were found concentrated along most of the lateral cell-cell contacts (Fig. 6). Co-staining with antibodies to the tight junction protein ZO-1, a marker for the apical junction, showed that PTPWT was absent from the apical terminal portion of the lateral surface (data not shown). In contrast, the PTP⌬Ig protein was seen in the lateral junction and on the apical and basal surfaces of the monolayer. This diffuse surface staining of PTP⌬Ig observed in GM7372 cells was also seen in an image of MDCK cells from a section in the x-y plane (Fig. 6,  right ⌬Ig panel). In contrast to PTP⌬Ig, the PTP⌬E protein, with the entire E-subunit deleted, was concentrated in heterogeneous membrane structures throughout the cytoplasm and absent from lateral cell-cell contacts and the apical surface. Therefore, the loss of the Ig domain resulted in reduced cell-cell contact localization and diffuse surface expression of the PTP mutant in both GM7372 endothelial cells and MDCK cells.
Deletion of the Ig Domain Affects Localization of PTP in Subconfluent Cells-In contrast to what has been proposed for PTPWT (11), the distribution of PTP⌬Ig in confluent cells (Figs. 3-6) suggested that stable surface expression of this mutant can be maintained independently of localization to cell-cell contacts. Therefore, we examined subconfluent GM7372 cells, which lack cell-cell contacts, to determine whether PTP⌬Ig was, in contrast to the WT protein (Fig. 1), stably expressed at the cell surface. Staining of non-permeabilized cells revealed that PTP⌬Ig was expressed over the surface of subconfluent cells with a high concentration seen at certain regions of the cell periphery (Fig. 7). In contrast, little surface staining could be observed in cells expressing PTPWT (Fig. 7). Permeabilization of cells prior to staining revealed the intracellular localization of both PTPWT and PTP⌬Ig (Fig.   FIG. 5. PTP⌬Ig is present at the cell surface of GM7372 cells. GM7372 cells expressing PTPWT or PTP⌬Ig were grown to confluence and fixed. Cells were either permeabilized as described under "Experimental Procedures" (P) or left intact (NP) and then co-stained with anti-PTP E-subunit antibody BK2 and anti-VE-cadherin antibody 45C6 (VE-cad). Green (left panels) and red (right panels) channels for each field are shown side by side. PTPWT is seen at cell-cell contacts of non-permeabilized cells (first row, arrowheads), whereas in permeabilized cells, additional staining is observed in an intracellular perinuclear region (second row, arrowhead). In cells expressing detectable PTP⌬Ig, the surface-exposed mutant protein can be seen as a diffuse staining pattern in non-permeabilized cells (third row) and permeabilized cells (fourth row). Staining at cell-cell contacts in the PTP⌬Ig samples can be attributed, in part, to endogenous PTP. Samples were viewed with a ϫ40 objective lens. Scale bar ϭ 10 m.
To confirm the presence of epitope-tagged PTP⌬Ig at the cell periphery, we stained permeabilized cells with the anti-T7 epitope antibody in combination with anti-P-subunit antibody SK7. In 68 of 107 cells (64%) observed, PTP⌬Ig showed a striking pattern of expression at regions of the cell periphery (Fig. 8), whereas peripheral localization of PTPWT was seen in only 6 of 102 cells (6%). Deletion of the entire E-subunit eliminated the peripheral localization seen with the PTP⌬Ig mutant, indicating that some features of the extracellular domain are required for targeting PTP⌬Ig to the cell periphery (Fig. 8). Therefore, deletion of the Ig domain allowed PTP to concentrate at the cell periphery in subconfluent cells by a mechanism requiring some other elements of the E-subunit.
Ectopic Expression of PTP Induces Alterations in Focal Adhesion and Actin Cytoskeleton Morphology-The aberrant localization of PTP⌬Ig in subconfluent GM7372 cells suggested that this protein may modulate events that occur at the cell periphery such as the formation of nascent cell-matrix junctions. Consistent with this idea, the morphology of focal adhesions and cytoskeleton components was affected by overexpression of PTP⌬Ig in sparse cultures of GM7372 cells. Vector control cells (Fig. 9, vector panels) showed the characteristic punctate pattern of vinculin-containing focal adhesions (1). In cultures expressing PTP⌬Ig, many cells displayed a broad region of vinculin staining that co-localized with the phosphatase at the cell periphery (Fig. 9, ⌬Ig panels,  arrowheads).
The presence of PTP⌬Ig at the cell periphery correlated with the redistribution of phosphotyrosine proteins from discrete punctate areas, as seen in vector control cells (Fig. 9, vector panels), to continuous broad regions of staining that coincided with the mutant PTP at the cell periphery (⌬Ig panels, arrowheads). Again, the changes in phosphotyrosine localization in PTPWT cells were minor compared with those seen with PTP⌬Ig (Fig. 9). These results, in combination with the redistribution of vinculin, suggest that the aberrant localization of PTP⌬Ig results in specific changes that affect the formation of focal contacts at the cell periphery, rather than the large-scale dephosphorylation of phosphotyrosine proteins in general. The specific PTP activities of PTP⌬Ig and PTPWT immunoprecipitated from lysates of GM7372 cells were similar (data not shown), suggesting that the effects caused by PTP⌬Ig correlate with the unique localization of this mutant and are not caused by a change in activity.
Further analysis by immunoblotting lysates with anti-phosphotyrosine antibodies did not reveal consistent changes in the pattern of protein tyrosine phosphorylation in response to expression of PTPWT or PTP⌬Ig, reinforcing the idea that the ectopic expression of the phosphatase causes limited changes in tyrosine phosphorylation. We tested several potential targets of PTP, including ␤-catenin and p120 ctn (35), as well as the major focal adhesion phosphotyrosine proteins focal adhesion kinase, c-Src, paxillin, and p130 cas , as suggested by the localization of PTP⌬Ig. In the presence of ectopic PTPWT or PTP⌬Ig, we did not detect consistent changes in the state of tyrosine phosphorylation of these proteins by immunoblotting with anti-phosphotyrosine antibodies or phospho-specific antibodies.
To examine further the state of cell-matrix contacts, we visualized the distribution of the focal adhesion protein paxillin (1). In cells expressing empty vector or PTPWT, paxillin was visualized in a punctate staining pattern at regions at the cell-matrix interface characteristic of focal adhesions (Fig. 10). The sizes of the punctate regions observed in WT cells were not as uniformly large as those of the vector control cells. Cells expressing PTP⌬Ig exhibited a diffuse pattern of paxillin localization with some punctate staining at the cell periphery (Fig. 10). Similarly, PTP⌬Ig cells displayed numerous tensincontaining contacts, but they were not as large as those seen with cells expressing empty vector or PTPWT (data not shown). In light of the interdependence of focal adhesions and the actin cytoskeleton, the morphology of F-actin was also FIG. 7. The PTP⌬Ig protein is present at the surface of subconfluent GM7372 cells. GM7372 cells expressing either PTPWT-T7 or PTP⌬Ig were grown to confluence and fixed. Cells were either permeabilized as described under "Experimental Procedures" (perm.) or left intact (non-perm.) and then stained with anti E-subunit antibody BK2. Asterisks indicate a PTPWT-expressing cell displaying little detectable PTP surface expression and a PTP⌬Ig-expressing cell showing elevated levels of PTP surface expression with localized concentration at the periphery (arrowheads). Scale bar ϭ 10 m.
FIG . 8. Localization of PTP-T7 proteins in subconfluent GM7372 cells. GM7372 cells expressing PTPWT-T7, PTP⌬Ig, or PTP⌬E were plated on glass coverslips, fixed, and co-stained with the anti-T7 epitope antibody (green) and anti-PTP P-subunit antibody SK7 (red). Green (left panels) and red (right panels) channels for each field are shown side by side. Arrowheads indicate unique localization of the PTP⌬Ig protein seen in 68 of 107 cells examined. Shown is an example of a typical PTPWT-expressing cell exhibiting no peripheral localization for PTP. Scale bar ϭ 10 m.
examined. PTP⌬Ig cells exhibited poorly defined actin-based stress fibers; instead, F-actin was found at restricted loci at the cell periphery (Fig. 10). Cells expressing empty vector showed well formed stress fibers, whereas PTPWT-expressing cells displayed short stress fibers with numerous branches (Fig. 10,  V and WT panels).
Expression of PTP Constructs Impairs Survival of GM7372 Cells-The defects in cell-matrix junction morphology might be expected to impair cell growth and/or survival (36,37). Furthermore, we would anticipate that these effects would be more pronounced at low cell density, when cells are more dependent upon signals generated from cell-matrix junctions for survival, and could explain the difficulties in generating stable clones of GM7327 cells expressing PTP. To test this hypothesis, GM7372 cells infected with recombinant retrovirus encoding pWZL/PTP mutants were plated at various densities from ϳ50% confluence (1 ϫ 10 4 cells/cm 2 ) to ϳ5% confluence (0.11 ϫ 10 4 cells/cm 2 ) and incubated for 15 days, and the cell number was estimated by staining with crystal violet. Comparison of cells expressing empty vector with those expressing different PTP constructs indicated that at each given plating density, expression of PTP⌬Ig had the greatest inhibitory effect on cell number compared with the PTPWT and PTP⌬E proteins (Fig. 11A). The four cell lines exhibited similar rates of proliferation represented by the percentage of cells progressing through the cell cycle as indicated by the incorporation of BrdUrd (Fig. 11B). However, PTP⌬Ig-expressing cultures exhibited increased levels of apoptosis compared with the other cell lines (Fig. 11B). As cells complete apoptosis, they dissociate from the culture dish and are not counted in the assay we used. As a result, the percentage of apoptotic cells presented in Fig.  11 likely under-represents the rate of apoptosis of the cultures. Therefore, consistent with the changes in focal adhesions and the actin cytoskeleton, expression of PTP⌬Ig reduced the survival of GM7372 cells.

DISCUSSION
The precise consequence of ligand binding is not understood for most RPTPs, but the structural relationship of enzymes such as PTP to cell adhesion molecules suggests that one purpose of the extracellular segment may be to position the phosphatase at sites of cell adhesion. In this regard, it has been suggested that homophilic binding, the engagement of a PTP molecule on one cell by another PTP molecule on an apposing cell, directs localization of PTP to sites of cell-cell contact. At this location, PTP is thought to regulate the cell junction complexes that control cell-cell adhesion (11,15). We examined the role of the conserved Ig domain, which contains the homophilic binding site, in regulating PTP in cultured cells. Recent studies have conveyed the importance of Ig domains in the extracellular segments of RPTP-LAR and PTP, which are in a different structural class from PTP. In Drosophila, the three Ig domains of PTP-LAR have been shown to be important for survival (38). The N-terminal Ig domain of PTP mediates binding to heparin sulfate proteoglycans and may be critical in targeting the phosphatase to sites of cell-matrix interaction (39). Mutations in the Ig domain of the Caenorhabditis elegans RPTP CLR-1 compromise the ability of the phosphatase to control axonal guidance (40). Our results indicate that the Ig domain of PTP not only is involved in localization of the enzyme to cell-cell contacts, but may also control localization in the absence of cell-cell contact.
To explore the function of the Ig domain in PTP, we generated a construct, termed PTP⌬Ig, in which this domain had been deleted. The integrity of this mutant protein is supported by several observations. As in the case with the wild-type enzyme, the mutant protein was processed normally into a P-subunit (containing the intracellular PTP domains), which was associated with the E-subunit, here comprising the mutated extracellular segment. In addition, the enzymatic activity and half-life of PTP⌬Ig were unaffected compared with those of the wild-type enzyme. Furthermore, the data in Fig. 3, showing that regions of the E-subunit other than the Ig domain contribute to the residual localization of PTP⌬Ig to cell-cell contacts, also suggest that some function is retained in either the MAM domain or fibronectin type III repeats when the Ig domain is deleted. In contrast to what has been proposed for PTPWT (11), the PTP⌬Ig protein, which is not localized to sites of cell-cell contact, was observed at the cell surface. As outlined in Fig. 12, PTP is thought to cycle to and from the plasma membrane in subconfluent cells through membrane structures in the perinuclear region of the cell (11,15), thereby limiting its access to other cell compartments, including cellmatrix junctions. When culture density increases and cells make contact, homophilic binding would restrict PTP to sites of cell-cell contact where the ligand, another PTP molecule, would be available on an apposing cell. With the loss of the Ig domain, the normal cycling of PTP was disrupted such that the PTP⌬Ig protein was widely distributed at the cell surface and concentrated at the cell periphery, where it could interact with cell-matrix junctions.
Although our data illustrate that the Ig domain is critical to the control of PTP localization in both confluent and subconfluent cells, its mechanism of action in each case may be different. One possibility is that the extracellular segment, including the Ig domain, mediates a combination of trans-homophilic (apposing cell) and cis-homophilic (with PTP on the same cell) interactions, which may regulate localization of PTP either in the presence or absence of cell-cell contact. There is evidence that some cell adhesion molecules such as cadherin and L1CAM participate in combinations of cis-and trans-homophilic interactions often involving multiple conserved domains, including Ig domains (41,42). One function for the cis-interactions is to mediate the formation of multimeric receptor complexes that increase the avidity of the trans-interactions, thereby strengthening cell-cell adhesion complexes (41). Furthermore, cis-interactions may also affect the transport of adhesion proteins to and from the cell surface and, as a result, may control localization in the absence of cell adhesion (42). In light of these precedents, it is possible that elements of the PTP extracellular domain may participate in both trans-and cis-homophilic interactions, controlling localization of the phosphatase either in the presence or absence of cell-cell contact.
Previous studies have demonstrated that PTP associates with classical cadherins such as E-cadherin, N-cadherin, and cadherin-4 (43) and that PTP can modulate cell adhesion to culture surfaces coated with N-or E-cadherin (44,45). However, PTP⌬Ig did not induce major changes in the localization of VE-cadherin, suggesting that the integrity of cell-cell contacts was maintained (Figs. 3 and 5). Instead, the consequences of PTP⌬Ig expression were most apparent in subconfluent cells in the absence of cadherin-dependent cell-cell adhesion and suggested that this mutant acts on a component of cellmatrix junctions or the associated actin cytoskeleton. Considering that many potential targets of PTP in cell-cell junctions are also found in cell-matrix junctions, the effects caused by the Ig domain deletion mutant may be a result of the mislocalized phosphatase acting on a normal PTP substrate in the wrong context, that of the focal adhesion in subconfluent cells. We were unable to detect changes in the global pattern of tyrosine phosphorylation, implying that the expressed forms of PTP did not act in a broad, nonspecific manner, but instead exhibited selectivity in action, leading to the effects on focal adhesions and cell survival. Although p120 ctn , a phosphotyrosine protein that may function in both cell-cell and cell-matrix adhesions, could be such a substrate (35,46), we did not detect consistent changes in the tyrosine phosphorylation of this or other potential substrates following expression of either the PTPWT or PTP⌬Ig protein. FIG. 11. Expression of PTP constructs in GM7372 cells results in reduced cell survival. A, GM7372 cells infected with recombinant retrovirus encoding PTPWT, PTP⌬Ig, or PTP⌬E or the empty pWZL-hygro vector were plated at 1.0, 0.33, and 0.11 ϫ 10 4 cells/ cm 2 in duplicate and grown in selective medium for 15 days. Culture density was visualized by staining with crystal violet. B, left panel, GM7372 cells expressing the indicated PTP proteins or the empty vector control (V) were assayed for proliferation by measuring the incorporation of BrdUrd into chromatin as described under "Experimental Procedures." The solid bars and numbers represent the percentage of cells that incorporate BrdUrd (BrdU (ϩ)), with each bar representing the mean Ϯ S.E. of three cell counts. Right panel, cells undergoing apoptosis were identified and counted as described under "Experimental Procedures." The solid bars and numbers represent the percentage of apoptotic cells for each culture.
In defining a role for the Ig domain in regulating PTP, we have illustrated the importance of this region in positioning the phosphatase at its proposed site of action, cell-cell junctions. In addition, this analysis has also brought to light the importance of the Ig domain in restricting PTP localization to prevent it from acting on improper targets. In subconfluent GM7372 cells, which normally maintain a low level of endogenous PTP, increasing the amount of the phosphatase by ectopic expression of the wild-type enzyme impaired cell survival. However, expression of the PTP⌬Ig mutant exerted more severe effects on cell survival as well as changes in focal adhesion and cytoskeleton morphology. These effects correlated with the atypical localization of this mutant, rather than the level of protein expression or phosphatase activity. Our analysis therefore highlights the importance of regulation of the localization of PTP in controlling the function of the enzyme and the maintenance of normal cell physiology.