Association of the Tensin N-terminal Protein-tyrosine Phosphatase Domain with the α Isoform of Protein Phosphatase-1 in Focal Adhesions*

Focal adhesions attach cultured cells to the extracellular matrix, and we found endogenous protein phosphatase-1α isoform (PP1α) localized in adhesions across the entire area of adherent fibroblasts. However, in fibroblasts migrating into a scrape wound or spreading after replating PP1α did not appear in adhesions near the leading edge but was recruited into other adhesions coincident in time and space with incorporation of tensin. Endogenous tensin and PP1α co-precipitated from cell lysates with isoform-specific PP1 antibodies. Chemical cross-linking of focal adhesion preparations with Lomant's reagent demonstrated molecular proximity of endogenous PP1α and tensin, whereas neither focal adhesion kinase nor vinculin was cross-linked and co-precipitated with PP1α, suggesting distinct spatial subdomains within adhesions. Transient expression of truncated tensin showed the N-terminal 360 residues, which comprise a protein-tyrosine phosphatase domain, alone were sufficient for isoform-selective co-precipitation of co-expressed PP1α. Human prostate cancer PC3 cells are deficient in tensin relative to fibroblasts and have fewer, mostly peripheral adhesions. Transient expression of green fluorescent protein tensin in these cancer cells induced formation of adhesions and recruited endogenous PP1α into those adhesions. Thus, the protein-tyrosine phosphatase domain of tensin exhibits isoform-specific association with PP1α in a restricted spatial region of adhesions that are formed during cell migration.

Focal adhesions link force-generating elements of the actin cytoskeleton to the extracellular matrix. These adhesions are assembled initially in response to integrin engagement as cells protrude their leading edge in the process of spreading and migration (1)(2)(3). These initial adhesions contain focal adhesion kinase, paxillin, vinculin, and other proteins (4,5). Phosphorylation of these proteins on Tyr residues accompanies the assem-bly of these adhesions and P-Tyr 2 immunostaining is concentrated at these sites (6). Focal adhesions dynamically exchange protein components as they connect to F-actin filaments and develop contractile force for cell locomotion. Adhesions that produce actomyosin-dependent movement of fibronectin receptors have been called "fibrillar adhesions" (7) to distinguish them from adhesions near the leading edge of the cell. These adhesions are relatively enriched in tensin, a protein that has a protein-tyrosine phosphatase domain related to PTEN (phosphatase and tensin1 homologue on chromosome 10), an Src homology 2 (SH2) domain, and a phosphotyrosine binding (PTB) domain (8,9). The overall process of focal adhesion maturation can be followed either in cells migrating on two-dimensional surfaces or after re-plating onto matrix-coated surfaces as cells spread. It is now appreciated that focal adhesions are dynamic structures rapidly exchanging constituent proteins (10). Current work on the dynamics of focal adhesions suggests that Tyr phosphorylation is a major requirement for their turnover. However, results also show Ser/Thr phosphorylation has a role in the assembly and disassembly of focal adhesions (11). Accumulating data show that focal adhesion proteins are phosphorylated at many Ser and Thr residues (Cell Migration Consortium), and understanding the control of and the functional consequences of these modifications is a challenge for future work. One step will be defining which Ser/Thr kinases and phosphatases are in focal adhesions and the basis for their localization.
Our interest in Ser/Thr phosphorylation of myosin and focal adhesion proteins led us some years ago to find protein-Ser/Thr phosphatase-1 (PP1) in focal adhesions of adherent fibroblasts (12). There are three predominant PP1 isoforms, named ␣, ␦ (or ␤), and ␥-1, that are widely expressed in somatic cells, with nearly identical catalytic domains of about 300 residues, plus 20 -30 C-terminal residues that are distinctive between isoforms (13). These C-terminal sequences have been used to prepare isoform-specific antibodies to study the cellular distribution of individual isoforms (14,15). One example of isoform specificity is the binding of the PP1␦ isoform to MYPT1 and G M regulatory subunits (16,17). The PP1␦ isoform also has been reported to associate directly with focal adhesion kinase (FAK), which might account for its localization in focal adhesions (18). A more recent study indicates that PP1 and glycogen synthase kinase 3 react with different Ser phosphorylation sites in FAK (18). In this study we used isoform-specific antibodies to visualize the ␣ isoform of PP1 in focal adhesions. During cell migration or cell spreading, the PP1␣ was not seen in focal adhesions near the leading edge, but it was seen in other adhesions and co-localized with tensin. We co-precipitated tensin and PP1␣ and chemically cross-linked these proteins in preparations of focal adhesions made by detergent extraction of adherent cells. Furthermore, the N-terminal domain of tensin is the domain that resembles a protein-tyrosine phosphatase (8). This protein-tyrosine phosphatase domain is the basis for relationship to PTEN protein and is one of two regions of tensin that mediates binding to adhesions (19). Here we show the tensin protein-tyrosine phosphatase domain is sufficient for isoform-specific association with PP1␣. The results reveal a new spatial domain within cells where PP1␣ is targeted, where it can potentially alter the Ser/Thr phosphorylation of select focal adhesion proteins.

MATERIALS AND METHODS
Reagents-Antibodies against vinculin, paxillin, and tensin were purchased from Sigma, BD Transduction Laboratories (San Diego, CA), and Chemicon-Upstate (Temecula, CA), respectively. Chemicals not otherwise specified were purchased from Fisher. Rabbit polyclonal antibody against a C-terminal 11-residue peptide of human PP1␣ with an N-terminal Cys added was affinity-purified using the SulfoLink Coupling Gel Protocol (Pierce) from serum recovered after immunization with peptide conjugated to keyhole limpet hemocyanin. Anti-PP1␦ antibody was produced in parallel. The cDNAs of PP1␣ and PP1␦ were cloned by the reverse transcription-PCR method using cDNA libraries of human placenta and pig aorta, respectively, and confirmed by automated DNA sequencing. The amino acid sequence of pig PP1␦ is identical to other mammalian PP1␦ (accession number AB016735). GFP-tensin vector was characterized previously (19).
Cells on coverslips were rinsed with phosphate-buffered saline (PBS), fixed, and permeabilized with Ϫ20°C methanol for 2 min, rinsed twice again with PBS, and incubated in a 3% bovine serum albumin in PBS solution for 1 h at room temperature. Primary antibodies diluted in 3% bovine serum albumin-PBS were incubated with the specimen for at least 1 h at room temperature or overnight at 4°C. After washing for 5 min with PBS 3 times, the appropriate secondary antibodies (goat antirabbit rhodamine or goat anti-mouse TRITC) diluted in 3% bovine serum albumin in PBS were incubated with the specimen. Coverslips were then washed as above and mounted onto glass slides with 10 l of Vectashield mounting medium (Vector Laboratories, Burlingame, CA). Staining without primary antibody was done with parallel specimens side-by-side. In each case this yielded a blank image as a negative control. Digital images of fixed cells were captured using OpenLab software (Improvision, Lexington, MA) with a Nikon Microphot-SA epifluorescence microscope equipped with a Nikon Plan Apo 60ϫ/1.4 oil immersion objective, filter sets for fluorescein isothiocyanate, Texas Red, and 4Ј,6-diamidino-2-phenylindole fluorophores, and a Hamamatsu Orca II digital camera. Raw digital images (LIF files) were converted to 8-bit TIFF files in OpenLab and further processed (contrast-enhanced, pseudo colored, and merged) using Photoshop 5.5 software (Adobe).
Cell Migration and Replating-REF52 cells were replated onto fibronectin-coated coverslips and incubated at 37°C for 30 min, 60 min, or 5 h. Cells were fixed with methanol and stained with a mixture of rabbit polyclonal anti-PP1␣ (green) and one of the following mouse monoclonal antibodies to known focal adhesion proteins: anti-tensin, anti-paxillin, and anti-vinculin (red). Cells were visualized, and images were processed as above. For two-dimensional migration REF52 cells were grown to 100% confluence on fibronectin-coated coverslips, and the monolayer was "wounded" with a razor blade to remove cells from a stripe across the culture. Cells were allowed to migrate into the wound for 8 h before fixation with methanol and double staining with rabbit anti-PP1␣ and monoclonal mouse anti-vinculin.
Expression and Immunoblotting of HA-PP1␣ and -␦-COS7 cells in 35-mm dishes were transfected with 3.5 g of HA 3 -PP1␣ or HA 3 -PP1␦ in the pRK7 vector using EscortV transfection reagent (Sigma) following the manufacturer's protocol. After a 24-h transfection, cells were lysed with 50 l of Laemmli sample buffer. The samples were boiled for 10 min and were subjected to 10% SDS-PAGE plus immunoblotting. Triplicate replica membranes were prepared and stained in parallel with anti-HA, anti-PP1␣, and anti-PP1␦.
Immunoprecipitation and Immunoblotting-For immunoprecipitation, a 100-mm dish of confluent rat aorta smooth muscle cells was lysed on ice for 10 min in 1.0 ml of 50 mM MOPS, pH 7.0, 0.1 M NaCl, 1 mM EGTA, 5% glycerol, 0.1% Triton X-100, 0.3% 2-mercaptoethanol, 0.4 mM PefaBloc. The lysate was centrifuged at 20,000 ϫ g for 20 min, and the supernatant was split in half; half was mixed with 20 l of anti-PP1␣ antiserum, the other half was mixed with 20 l of pre-immune serum, and 15 l of a 50% slurry of protein A-agarose beads was added to both samples. After incubation overnight at 4°C, the beads were washed 4ϫ by centrifugation with 0.15 ml of the lysis buffer, and proteins were eluted with SDS sample buffer and separated by SDS-PAGE. Proteins were electrotransferred to nitrocellulose membranes (Bio-Rad 0.22 m) that were blocked with 5% nonfat milk in Tris-buffered saline plus 0.1% Tween 20. Primary and secondary antibodies were diluted with the blocking buffer. The immunocomplexes were detected using SuperSignal West Pico Luminol/Enhancer Solution (Pierce) and x-ray film (Kodak X-Omat).
Cell Permeabilization, Cross-linking, and Immunoprecipitation of PP1 Complexes-REF52 cells were grown to between 75 and 95% confluency in 10-cm dishes, rinsed with PBS, and permeabilized with 0.1% Triton X-100 and 5 mM MgCl 2 in PBS for 0.5 min at room temperature to prepare cytoskeletons. The chemical cross-linker Lomant's reagent, dithiobis(succinimidyl propionate) (Pierce) dissolved in dimethyl sulfoxide (Me 2 SO) was combined with PBS to a final concentration of 4 g/ml and incubated with these cytoskeletons for 30 min at room temperature. Quenching of the reaction used 50 mM NH 4 Cl in PBS for 10 min at room temperature. Samples were scraped from the dish in RIPA buffer (50 mM MOPS, pH 7.5, 1% IGEPAL CA-630 (Nonidet P-40), 0.25% deoxycholate, 0.1% SDS, 150 mM NaCl, 1 mM EDTA, 20 mM ␤-glycerol phosphate, and 1 mM PefaBloc). After sonication for 30 s the suspension was centrifuged for 10 min at 13,800 rpm at 4°C. The pellet was resuspended in RIPA buffer, and aliquots were taken from both pellet and supernatant (RIPA Insoluble and RIPA Extract). The RIPA extract was divided into two aliquots. Rabbit pre-immune or rabbit anti-PP1␣ serum coupled to protein A resin (Sigma P-3476) was incubated with the extract overnight at 4°C. The resin was pelleted by centrifugation at 8000 rpm for 2 min, and the supernatant was taken as the sample for immunoblot analysis. Resin was washed 3ϫ with RIPA buffer, then 2ϫ SDS sample buffer was added to the resin. Samples were boiled for 10 min and subjected to 4 -15% gradient SDS-PAGE using precast gels (Bio-Rad) followed by immunoblotting.
Tensin and PP1 Co-expression and Pulldown-Either COS7 or HEK293 cells were seeded at 30% confluence and plated overnight, then transfected using calcium phosphate with 5 g of pTriEx4 plasmid (Novagen) encoding tensin plus 0.5 g of pKR7 plasmid encoding HA 3 -PP1␣. After 24 -48 h cells were washed with PBS and lysed in buffer containing 50 mM MOPS, pH 7.0, 0.1 M NaCl, 1 mM EGTA, 5% glycerol, 0.1% Tween 20, 0.1% 2-mercaptoethanol, and 0.8 mM PefaBloc. An extract was recovered as the supernatant after 15 min centrifugation at 13,800 rpm, and a sample was reserved before adsorption to S-protein beads (Novagen) for 1 h. The beads were washed three times by centrifugation with the lysis buffer, and the proteins were eluted with SDS sample buffer and boiling. Proteins were resolved by SDS-PAGE using a double-layered gel that was 10% in the bottom half and 6% in the top half to afford resolution of both the tensin proteins and the PP1. Proteins were transferred onto nitrocellulose submerged in transfer buffer by overnight electrophoresis at 50 V and 120 mA. Filters were blocked with 5% nonfat milk and immunoblotted with horseradish peroxidase (HRP)-conjugated S-probe (Novagen) at 1:10,000 or anti-HA at 1:10,000 followed by anti-mouse-HRP at 1:10,000 dilution and developed as described above.

Dynamic Recruitment of PP1 into Focal Adhesions-Rat
embryo fibroblasts (REF52 cells) grown on fibronectin-coated coverslips develop focal adhesions across the entire bottom of the cell, and these adhesions contain both vinculin and PP1␣, based on immunofluorescent staining after fixation (Fig. 1, A and B). By comparison, REF52 fibroblasts showed localization of the PP1␦ isoform primarily along actin stress fibers (Fig. 1, C and D). The specificity of these anti-PP1␣ and anti-PP1␦ antibodies was demonstrated by selective immunostaining of the respective epitope-tagged PP1 isoforms expressed in COS7 cells (Fig. 1E) using anti-HA staining (top panel) to show that equivalent amounts of the PP1 proteins were in each lane. The results demonstrate differential localization of these PP1 isoforms in fibroblasts and isoform-specific targeting of PP1␣ to focal adhesions. The same results using these antibodies were obtained with rat aorta smooth muscle cells (not shown). Thus, differential targeting of these PP1 isoforms occurs in multiple cell types.
Focal adhesions are dynamic structures that undergo reorganization in migrating cells. Polarized migration of cells occurs when confluent monolayers are wounded to scrape away cells. Migrating REF52 cells at the edge of the wound formed protrusions with prominent adhesions behind the leading edge that stained for vinculin (Fig. 2, arrows in the bottom panel) but not PP1␣ (compare Fig. 2, upper panel). PP1␣ was localized in focal adhesions in the same cell but only in those adhesions further away from the front of the cell nearer to the center of the cell (arrows in top panel, Fig. 2). Vinculin served as a marker for FIGURE 1. Differential localization of PP1␣ and PP1␦ in rat embryo fibroblasts. REF52 cells grown on fibronectin-coated coverslips were fixed with methanol and stained with rabbit anti-PP1␣ (A) or anti-PP1␦ (C) and with mouse monoclonal anti-vinculin antibodies (B) or rhodamine-conjugated phalloidin (D). Cells were visualized by fluorescence microscopy with fluorescein-conjugated anti-rabbit and rhodamine-conjugated anti-mouse secondary antibodies. E, the cDNA for PP1␣ and for PP1␦ were transiently expressed in COS7 cells that were dissolved directly in SDS sample buffer and immunoblotted with anti-HA as a loading control. The samples also were immunoblotted with either anti-PP1␣ (center panel) or anti-PP1␦ (lower panel). Differential staining of the PP1␣ and -␦ isoforms demonstrates the specificity of the antibody preparations.
focal adhesions across the entire area of the cell whether or not the adhesions stained for PP1␣ (Fig. 2, merged image). Because PP1␣ was not detected in adhesions near the leading edge but in other adhesions in the same cell, we suspected it was being dynamically recruited into adhesions during cell migration.
To test this hypothesis REF52 cells were taken into suspension and replated onto fibronectin-coated coverslips. At various times cells were fixed, and focal adhesions were visualized by immunofluorescent staining of endogenous proteins (Fig. 3). The three larger images are merged from the smaller images at the upper left of cells stained for PP1␣ (green) and the images at upper right for vinculin, P-Tyr, and tensin (red/orange, top to bottom). Adhesions around the perimeter of the cells nearest to the leading edge contained vinculin and P-Tyr proteins but did not stain for PP1␣, as seen in the outermost red/orange adhesions in the merged images (Fig. 3, A and B). In these same cells other adhesions localized nearer the center of the cells were double-immunostained for vinculin and PP1␣ (Fig. 3A) or separately for P-Tyr and PP1␣ (Fig. 3B). These double-positive adhesions appear yellowish-green in the merged images. Thus, in both cells migrating into a wound (Fig. 2) and cells spreading after replating (Fig. 3) there were two populations of adhesions; peripheral adhesions without PP1␣ and other, more central focal adhesions that contained PP1␣. This differential distribution of PP1␣ relative to vinculin or P-Tyr (or paxillin, data not shown) was a transient phenomenon only seen while cells were actively spreading. At later times after cells were spread and stationary (t Ͼ 4 h) all the focal adhesions were double-stained for both vinculin and PP1␣, as seen in Fig. 1, A and B.
Tensin is a protein previously localized in a particular subset of focal adhesions called fibrillar adhesions that develop from peripheral adhesions in the course of cell migration (7). Prompted by the differential PP1␣ localization in adhesions of migrating and spreading cells, we immunostained for endogenous tensin and found co-localization with PP1␣ in spreading, migrating, and static cells (Fig. 3C). There was full overlap of staining for PP1␣ and tensin even as early as 30 min after replating. Furthermore, throughout the process of cell spreading in specimens fixed and stained at various times from 30 min to 4 h, we observed spatial coincidence of tensin and PP1␣ in adhesions. Similar observations were made during spreading of rat aorta smooth muscle cells after replating (not shown), so the phenomenon was not unique to REF52 fibroblasts. We found tensin and PP1␣ are recruited together at the same time into the same sites during dynamic remodeling of focal adhesions in migrating cells.
Spatial Proximity of Tensin and PP1␣ in Focal Adhesions-Because of the limited spatial resolution of light microscopy, it is not possible to determine whether co-localization of endogenous tensin and PP1␣ by immunofluorescent staining is due to interaction between the proteins. Therefore, we immunoprecipitated PP1␣ from extracts of rat smooth muscle cells and by immunoblotting found co-precipitation of a fraction of tensin from the lysate (Fig. 4). The co-immunoprecipitation of tensin was specific for PP1␣ antiserum relative to an equivalent amount of pre-immune serum as a negative control. As an additional control for specificity, we immunoprecipitated PP1␦ from the same extracts and did not recover any detectable tensin (not shown). The results indicated that endogenous PP1␣ and tensin in fibroblasts were together in a complex that was stable enough to be immunoprecipitated.
For additional evidence of endogenous tensin-PP1␣ association, we optimized conditions to permeabilize REF52 cells without dissociating all the PP1␣ from focal adhesions using minimal time and low concentrations of Triton X-100 detergent (Fig. 5A). Focal adhesions are complex, relatively insoluble protein assemblies like the cytoskeleton itself, so one faces the experimental challenge of solubilizing these assemblies for analysis while attempting to preserve protein-protein interactions between specific constituents. We resorted to chemical cross-linking of the focal adhesion preparations to identify near-neighbor interactions. Different reaction times and a FIGURE 2. Focal adhesions in migrating fibroblasts. REF52 cells were grown to 100% confluence on fibronectin-coated coverslips and wounded to scrape away cells. Cells migrated into the wound area (from left to right) for 8 h before fixation and double staining, as described for Fig. 1, with anti-PP1␣ (top) and anti-vinculin (bottom) antibodies. Arrows highlight different populations of focal adhesions that stained for both proteins (top panel) or those that stained for vinculin but not PP1 (bottom panel). Immunofluorescent images were merged (center), with yellow revealing overlap. We note that these experiments were restricted to methanol fixation. Paraformaldehyde fixation gives higher contrast and better-defined images of focal adhesions but poor immunostaining of endogenous PP1.
range of conditions were tested to establish the highest concentration of Lomant's reagent, dithiobis(succinimidyl propionate) that could be used without diminishing immunostaining of the PP1␣ in the preparations. Using this optimized protocol, we solubilized the PP1␣ complexes from dithiobis(succinimidyl propionate) cross-linked cytoskeleton preparations with a buffer containing both dodecyl sulfate and Triton X-100 and then immunoprecipitated with anti-PP1␣ antibodies. Crosslinks between PP1␣ and its neighbors were cleaved by chemical reduction of the central disulfide bridge in the reagent, and the separate proteins were resolved by SDS-PAGE. The amount of PP1␣ recovered at each individual step of the protocol was assayed by immunoblotting (Fig. 5B) These data demonstrated selective immunoprecipitation of chemically cross-linked PP1␣ complexes that were solubilized from Triton X-100-resistant cytoskeletons of REF52 fibroblasts.
We analyzed these PP1␣ complexes for recovery of focal adhesion proteins by immunoblotting (Fig. 5C). Both PP1␣ and tensin were concentrated and greatly enriched in the cross- Immunofluorescent images were merged (larger panels) to show that the outermost adhesions were stained with vinculin (red in A) or P-Tyr (orange in B) but did not co-stain for PP1, whereas all the focal adhesions co-stained with tensin and PP1 (C).

FIGURE 4. Co-immunoprecipitation (IP) of endogenous tensin with PP1␣.
Extracts of rat aorta smooth muscle cells were precipitated in parallel as described under "Materials and Methods" with rabbit anti-PP1␣ serum or pre-immune rabbit serum as a control, and the precipitates were analyzed by immunoblotting. The tensin and PP1␣ proteins in 2% of the cell extract used for precipitation were identified on the same immunoblots (left lane labeled Input).
linked immunocomplexes relative to the amounts of these respective proteins in the permeabilized cells (Input). Control precipitations with pre-immune serum (Fig. 5C, Preim) did not enrich either tensin or PP1␣, showing the specificity of immunoprecipitation. In contrast, FAK was not detected in the anti-PP1␣ immunoprecipitates but was present in the extracts of the cytoskeletons (Fig. 5C). Some vinculin co-immunoprecipitated with PP1␣, not pre-immune serum; however, this represented only a miniscule fraction of vinculin in the input fraction, where it was very prominently stained. Thus, despite its relative abundance in cytoskeletons, vinculin was not efficiently cross-linked to PP1␣. Our interpretation was that vinculin and PP1␣ do not exist in molecular proximity to one another in focal adhesions.
As an additional control we immunoblotted for actin, but it was not detected in the specific anti-PP1␣ immunoprecipitates (not shown), so binding to F-actin could not account for an indirect recruitment of tensin to PP1␣. The specific co-precipitation after chemical cross-linking showed that endogenous PP1␣ and tensin were in molecular proximity to one another in focal adhesions.
Tensin N-terminal Domain Mediates Association with PP1-In addition to examination of endogenous proteins, we used overexpression of tagged tensin and tagged PP1␣ in COS7 cells to test for which region of tensin was necessary or sufficient to mediate association with PP1␣. Full-length tensin (1735 residues) was fused with a His 6 -S-peptide tag at the N terminus, and from this construct a series of truncated fusion proteins containing tensin residues 1-360, 1-740, and 1-1462 were created by introducing stop codons by site-directed mutagenesis. These proteins were co-expressed with HA-tagged PP1␣, and protein complexes were pulled down from cell extracts using beads conjugated with S-protein. The bound proteins were resolved by SDS-PAGE and analyzed by staining with horseradish peroxidase-conjugated S-protein and immunoblotting for the HA epitope tag. The three truncated tensin His-S tagged proteins were expressed and migrated at the expected sizes in SDS-PAGE. Despite efforts to express nearly comparable amounts of the His-S-tensin proteins, the full-length wild-type tensin was barely detected in pulldown assays, and we suggest this is probably due to a combination of factors, including low level of expression, limited solubilization, and resistance to electrotransfer during immunoblotting. Cells expressing HA-PP1␣ but no S-peptide tensin fusion protein were used to prepare control extracts to show no HA-PP1␣ bound to S-protein beads (Fig. 6A, BLK). Even the trace amounts of the full-length, wild-type tensin effectively pulled down HA-PP1␣ (right lane). Truncations of tensin from the C terminus did not much reduce pulldown of HA-PP1␣. Even the shortest protein tested, which contained only residues 1-360 of tensin, pulled down HA-PP1␣ in this assay. We concluded that the N-terminal residues 1-360 in tensin were sufficient to provide stable association with HA-PP1␣.
Specificity of tensin for the endogenous ␣ versus the ␦ isoform of PP1 was confirmed in pulldown assays with overexpressed proteins (Fig. 6B). A tensin S-peptide fusion protein with residues 1-740 was expressed alone as a control (Fig. 6B,  left lanes) or co-expressed with either HA-PP1␣ (center lanes) or HA-PP1␦ (right lanes). Immunoblots with anti-HA show about equal levels of overexpressed HA-PP1␣ and HA-PP1␦ (Fig. 6B, upper left). Complexes recovered on S-protein beads (right panels) were analyzed for PP1 recovery by anti-HA immunoblotting for the epitope tag on both isoforms (upper panel) and for recovery of the tagged tensin (lower panel). Only HA-PP1␣, not HA-PP1␦, associated with the tensin 1-740 fusion protein (Fig. 6B, upper right panel). These results confirm that the N-terminal half of tensin is sufficient to mediate association with PP1 and dictate isoform specificity.
Binding of PP1 to regulatory subunits involves (R/K)VXF motifs in the subunits that interact with a cleft on the backside of PP1 opposite the active site (20,21). Tensin has 6 VXF motifs, 3 in the N-terminal 360 residues, and 3 in the C-terminal 300 residues. Deletion of the C-terminal region of tensin did not eliminate association with PP1, indicating these VXF motifs were not strictly required. However, there could be multiple sites in tensin for association with PP1; therefore, we further tested the N-terminal tensin domain residues 1-360 versus the C-terminal domain, residues 1401-1735 (Fig. 6C). Single transfection to express HA-PP1␣ served as the control (Fig. 6C, left lanes) compared with dual transfection to express HA-PP1␣ plus S-peptide fusions with tensin N-terminal (center lanes) or C-terminal (right lanes) domains. Immunoblotting the cell extracts (left panels) demonstrated that expression levels of HA-PP1␣ were the same in control and both experimental samples. Detection with S-protein conjugated to horseradish peroxidase showed that both the expression levels and recovery by pulldown of S-peptide tensin (1401-1735) were higher than the levels of S-peptide tensin (1-360) (Fig. 6C, lower panels). Only the N-terminal domain of tensin (residues 1-360) associated with HA-PP1␣, even when higher levels of the C-terminal domain were compared. We have mutated individually two of the VXF motifs in the 1-360 region replacing the Phe with Ala; however, these point mutations did not abrogate pulldown of co-expressed HA-PP1␣ (not shown).
GFP-tensin Recruits PP1 to Focal Adhesions in PC3 Prostate Cancer Cells-Tensin is a key component of focal adhesions, but certain human prostate carcinoma cell lines (PC3) have been reported to be essentially devoid of tensin (22). Indeed, we detected only trace levels of tensin protein by immunoblotting cell extracts of PC3 cells in contrast to intense staining of tensin in an equivalent amount of total protein from REF52 cells on the same blot (Fig. 7A, top panel). As loading controls, we showed that these cell extracts had essentially identical amounts of endogenous PP1␣ and vinculin by immunoblotting with specific antibodies (Fig. 7A). We noticed that PC3 cells had slightly higher levels of FAK relative to REF52 cells (Fig. 7A,  lower panel). With anti-tensin antibody, PC3 cancer cells showed dim and diffuse cytoplasmic staining barely above background (Fig. 7B, upper panel), consistent with the absence of the protein on immunoblots. The same antibodies stained tensin in focal adhesions of HS68 human fibroblasts, showing that the antibodies recognized human as well as rat tensin. Immunostaining for endogenous PP1␣ in these cells showed it was located in the nucleus and was diffuse in the cytosol but not in distinct focal adhesions (Fig. 7B, center panel). There was some staining for PP1␣ along the perimeter of the cell. Focal adhesions in PC3 cells were immunostained with anti-vinculin, which exposed a ring of short, broad adhesions on the outermost edge of the cells (Fig. 7B, bottom panel). The PP1␣ staining did not overlap with these peripheral adhesions. Thus, low levels of endogenous tensin in human PC-3 prostate cancer cells was correlated with only a few focal adhesions at the perimeter of the cell and diffuse cytoplasmic distribution of endogenous PP1␣. These cancer cells provided a near-null background for transient overexpression of ectopic tensin.
Would tensin induce formation of more focal adhesions in human cancer cells and also recruit PP1␣ to those focal adhesions? We tested this idea by transiently transfecting PC3 cells with a plasmid encoding GFP-tensin. The GFP-tensin expressed at low levels (Fig. 7C, right-hand cell) was localized FIGURE 6. Isoform-specific association of PP1␣ with the N-terminal domain of tensin. A, full-length wild-type tensin tagged at the N terminus with His 6 -S-peptide and versions truncated by stop codons after residues 1462, 740, or 360 were transiently expressed in COS7 cells. Control cells were transfected to express HA 3 -PP1␣, and experimental cells were co-transfected to express tensin plus HA 3 -PP1␣. The levels of HA-PP1␣ in the cell extracts were compared by anti-HA immunoblotting (top). Complexes were pulled down on S-protein beads and analyzed for S-peptide tensin and for HA 3 -PP1␣. HRP, horseradish peroxidase. WT, wild type. B, cells were transfected to express His 6 -S-peptide-tensin (1-740) alone or with co-expressed HA 3 -PP1␣ or HA 3 -PP1␦. The levels of the proteins in the extracts are shown by staining for HA or S-peptide (left panels), and the recovery of the proteins pulled down on S-protein beads is shown in the right panels. C, HA 3 -PP1␣ was expressed in COS7 cells alone or with either tensin residues 1-360 or 1401-1735 fused to His 6 -S-peptide. A pulldown assay was conducted and analyzed as described in B. Ctrl, control.
into a few adhesions around the perimeter of the cell, and these same adhesions recruited endogenous PP1␣ (Fig. 7C, lower  panel). The GFP-tensin co-localized with vinculin in these peripheral adhesions, confirmed with immunofluorescent double staining of fixed cells (not shown). At higher levels of expression the GFP-tensin was incorporated into adhesions along the perimeter and also induced the formation of new adhesions across the entire area of the cell (Fig. 7C, left cell). Importantly, endogenous PP1␣ became co-localized with GFPtensin in both the perimeter and interior adhesions (Fig. 7C,  lower panel, arrows). In addition there were green fluorescent foci in the cytoplasm of these high expressing cells that were more diffuse and circular in appearance, which distinguished them from adhesions. These could be merely cytoplasmic protein aggregates but were sites for recruitment of endogenous PP1␣ that was immunostained coincident with the GFP-tensin. Endogenous PP1␦ was not recruited or localized in the adhesions or foci in PC3 cells expressing GFP-tensin (not shown), consistent with the isoform specificity of tensin. If GFP was expressed in these cells as a control, it did not localize into adhesions or induce formation of additional adhesions or cytoplasmic foci and did not result in re-localization of endogenous PP1␣. Cells overexpressing GFP-tensin were tested for GFP emission through the rhodamine filter at the same exposure times as used for staining PP1 to ensure that the images of PP1␣ were not due to fluorescence from GFP-tensin (bleed-through). These experiments demonstrated that expression of GFP-tensin in PC3 cells was sufficient to induce formation of new adhesions and cytoplasmic foci that specifically recruited endogenous PP1␣.

DISCUSSION
In this study we demonstrate that fibroblasts in a woundhealing model or replated on fibronectin to provoke cycling of focal adhesions revealed recruitment of tensin and PP1␣ together in time and space distinct from the recruitment of vinculin and paxillin, well recognized components of focal adhesions. Focal adhesions are now appreciated as dynamic structures, constantly remodeled by a cycle of formation, maturation, and dispersal. Initial adhesions formed at the leading edge of migrating or spreading cells mature and change composition by release and recruitment of protein components. Studies of this process have highlighted the striking differences in composition between early adhesions that are rich in the P-Tyr proteins vinculin and paxillin but lack tensin and adhesions that are tensin-rich but have little detectable P-Tyr (23,24). Our working hypothesis is that tensin recruits PP1 into focal adhesions to catalyze selective dephosphorylation of Ser/ Thr in nearby proteins. It is possible and even likely that tensin itself is a substrate of PP1␣, because it is reported to be phosphorylated on Ser and Thr as well as Tyr residues (25). Which focal adhesion proteins are the substrates of PP1 and how their phosphorylation state affects adhesion dynamics or cell migration are complex unanswered questions. Though many focal adhesion proteins undergo Ser/Thr phosphorylation, the effects of phosphorylation on function is not understood and is generally understudied. One recent idea is that Ser/Thr phos- phorylation regulates rates of focal adhesion assembly disassembly (11), a promising direction for new studies.
Tensin and PP1␣ were selectively co-precipitated from cell extracts and from chemically cross-linked focal adhesions, showing these proteins are associated together in complexes. Chemical cross-linking with Lomant's reagent (dithiobis(succinimidyl propionate)) forms spans ϳ12 Å end to end. This distance is considerably less than the diameter of the crosslinked proteins themselves, such as PP1␣, which is 50 ϫ 35 ϫ 35 Å (26), suggesting that tensin and PP1 are within close molecular proximity in adhesions. It was remarkable that neither FAK nor vinculin nor actin was cross-linked to PP1␣ in focal adhesions considering the abundance of these focal adhesion components and the attachment of actin filaments to tensin. We took the cross-linking results as evidence for distinct spatial domains within focal adhesions. We imagine that tensin-PP1␣ complexes are separate from complexes that contain FAK or vinculin.
We favor the hypothesis that the tensin N-terminal domain directly binds PP1␣ and functions as a regulatory subunit. Many PP1 regulatory subunits act as targeting proteins and directly bind PP1 with a canonical RVXF motif for primary interaction (see 20 -21). Tensin has at least six sequences that conform to this motif, distributed in the N-terminal and C-terminal domains. By making truncated versions of tensin, expressing them in cells, and testing for co-precipitation of coexpressed PP1␣, we found that the N-terminal domain (residues 1-360) alone is sufficient for tensin to associate with PP1. This domain is related in sequence to protein-tyrosine phosphatase and PTEN and alone can target GFP to focal adhesions (22). The multiple VXF motifs in the C terminus of tensin might function to bind PP1 separately from the N terminus, but a fusion protein with the tensin C-terminal domain did not associate with PP1, arguing that the VXF motifs in this region are not used to bind PP1. However, we are mindful that a lack-offunction result in the assay is not conclusive. We do know the N-terminal domain associates with PP1␣, not PP1␦, and mutation of either one of two VXF motifs in this region did not eliminate association in a pulldown assay (not shown). Additional experiments are required to map residues in tensin (1-360) that are required for PP1␣ association and to test for direct binding with purified recombinant proteins. As yet we cannot exclude participation of a scaffold protein that binds tensin and PP1␣ together in a complex, but at least such a protein must be specific for the PP1␣ isoform.
Transiently expressed GFP-tensin appears in focal adhesions and foci in human cancer PC3 cells otherwise deficient in tensin. Our results are consistent with previous reports that GFPtensin localizes into focal adhesions in NIH3T3 cells (19). In the living cells visualized without fixation, these adhesions were well formed, elongated structures as seen in fibroblasts and were evenly distributed across the cell. We noted that after fixation, the peripheral adhesions were unchanged in appearance, but adhesions toward the center of the cell area were lost or became more diffuse foci. This offers a clue that perhaps the stability or organization of the focal adhesions induced by GFPtensin in these cancer cells was different from native focal adhesions in fibroblasts. The sensitivity to fixation was not seen in fibroblasts, even those expressing GFP-tensin, so it was not just because of the overexpression but more likely due to the composition and/or modifications of focal adhesion proteins in the cancer cells. These GFP-tensin-induced focal adhesions did recruit endogenous PP1␣ but not PP1␦, again revealing isoform-specific targeting of the phosphatase in living cells.
It has become evident that focal adhesions physically recruit and are modified by multiple Ser/Thr phosphatases that thereby have effects on cell migration. PP2A is co-precipitated with the focal adhesion protein paxillin, and in metastatic mouse melanoma BL6 cells a truncated version of a PP2A regulatory B subunit impairs targeting of the phosphatase to focal adhesions and prevents specific dephosphorylation of paxillin (27). Reduction in PP2A levels by antisense oligonucleotides increased paxillin phosphorylation and cell motility (28). Cytostatin, a specific chemical inhibitor for protein phosphatase 2A, suppresses cell adhesion to extracellular matrix (29). Other results have shown PP1␦ localizes in focal adhesions of rat smooth muscle cells and HeLa cells, fibroblasts, endothelial cells, and keratinocytes (30). This localization possibly involves direct interaction of the PP1␦ isoform with the C-terminal domain of FAK (FRNK region). PP1␦ is proposed to mediate dephosphorylation of Ser in FAK when cells are released from mitosis (31). Our results may be complementary, not necessarily conflicting with the reported PP1␦-FAK interaction. It is possible PP1␣ and -␦ isoforms may both function in regulation of focal adhesions by assembly into different complexes with different partners, i.e. tensin or FAK. Results here show specific interaction and cross-linking of PP1␣ with tensin but not with FAK. Interaction with regulatory subunits spatially restricts PP1 within cells as well as alters substrate specificity (32) and determines sensitivity to inhibitor proteins (33). Therefore, we presume that PP1␣ binds to the protein-tyrosine phosphatase domain of tensin to selectively regulate a specific subset of phosphoproteins in focal adhesions. Although dozens of possible Ser/Thr phosphorylation sites are predicted and are now found in focal adhesion proteins (34), not much is known about their effects on function or response to signaling, subjects for further study.