A Fragment of Paxillin Binds the α4Integrin Cytoplasmic Domain (Tail) and Selectively Inhibits α4-Mediated Cell Migration*

The α4 integrins play important roles in embryogenesis, hematopoiesis, cardiac development, and the immune responses. The α4 integrin subunit is indispensable for these biological processes, possibly because the α4 subunit regulates cellular functions differently from other integrin α subunits. We have previously reported that the α4 cytoplasmic domain directly and tightly binds paxillin, an intracellular signaling adaptor molecule, and this interaction accounts for some of the unusual functional responses to α4 integrin-mediated cell adhesion. We also have identified a conserved 9-amino acid region (Glu983-Tyr991) in the α4cytoplasmic domain that is sufficient for paxillin binding, and an alanine substitution at either Glu983 or Tyr991within this region disrupted the α4-paxillin interaction and reversed the effects of the α4 cytoplasmic domain on cell spreading and migration. In the current study, we have mapped the α4-binding site within paxillin using mutational analysis, and examined its effects on the α4tail-mediated functional responses. Here we report that sequences between residues Ala176 and Asp275 of paxillin are sufficient for binding to the α4 tail. We found that the α4 tail, paxillin, and FAT, the focal adhesion targeting domain of pp125FAK, could form a ternary complex and that the α4-binding paxillin fragment, P(Ala176–Asp275), specifically blocked paxillin binding to the α4 tail more efficiently than it blocked binding to FAT. Furthermore, when expressed in cells, this α4-binding paxillin fragment specifically inhibited the α4 tail-stimulated cell migration. Thus, paxillin binding to the α4 tail leads to enhanced cell migration and inhibition of the α4-paxillin interaction selectively blocks the α4-dependent cellular responses.

Integrins are a large family of transmembrane adhesion receptors that each is composed of a ␣ and a ␤ subunit (1)(2)(3). Integrins mediate cell adhesion and cell migration, and regu-late gene expression and cell survival (1,3). The ␣ 4 integrins are primarily expressed on various leukocytes and play important roles in embryogenesis, hematopoiesis, cardiac development, and the immune responses (4 -7). The ␣ 4 integrin subunit is indispensable for these biological processes, possibly because the ␣ 4 subunit regulates cellular functions differently from other integrin ␣ subunits. Indeed, the ␣ 4 integrin promotes increased cell migration and less cell spreading and focal adhesion formation relative to most other ␤ 1 integrins. These unusual functional properties are mediated by the ␣ 4 cytoplasmic domain (8,9) because this region of ␣ 4 markedly stimulates cell migration, and opposes cell spreading and focal adhesion formation when joined to other integrin ␣ subunits (9 -11).
We previously reported that the ␣ 4 cytoplasmic domain directly and tightly binds paxillin, an intracellular signaling adaptor molecule (10,11). The ␣ 4 -paxillin interaction accounts for some of the unusual functional responses to ␣ 4 ␤ 1 integrinmediated cell adhesion, including stimulating cell migration and opposing cell spreading and focal adhesion formation (10,11). We have identified a conserved 9-amino acid region (Glu 983 -Tyr 991 ) in the ␣ 4 cytoplasmic domain that is sufficient for paxillin binding (11), and an alanine substitution at either Glu 983 or Tyr 991 within this region disrupted the ␣ 4 -paxillin interaction and reversed the effects of the ␣ 4 cytoplasmic domain on cell spreading and migration (10,11).
Paxillin is a 68-kDa cytoplasmic protein that is involved in cellular responses to integrin-dependent adhesion (12,13). Paxillin has the structural properties of a signaling adaptor molecule. It contains four C-terminal LIM protein-protein interaction motifs that serve to target it to focal adhesions (12,13), and five N-terminal LD motifs that mediate protein-protein interactions (12)(13)(14). Paxillin directly interacts with several cytoskeletal, intracellular signaling, and adaptor molecules such as Src, PTP-PEST, Crk, p95 PKL, actopaxin, and ILK (15-22, Fig. 1A). Paxillin also interacts with pp125 FAK , a molecule strongly implicated in the regulation of cell migration (23,24), by binding to a C-terminal domain of pp125 FAK termed the focal adhesion targeting (FAT) 1 domain (Fig. 1A). Furthermore, the ␣ 4 cytoplasmic domain markedly enhances activation of pp125 FAK . The enhanced pp125 FAK phosphorylation depends on the integrity of the paxillin-binding site in the ␣ 4 tail (10). These results suggest that the direct association of paxillin with the ␣ 4 cytoplasmic domain might facilitate the rapid re-cruitment and activation of paxillin-binding proteins such as pp125 FAK , thus accounting for some of the unusual biological properties of ␣ 4 integrins. In the current study, we have tested this idea by examining the capacity of the ␣ 4 tail to form ternary complex with paxillin and pp125 FAK . Furthermore, we have mapped the ␣ 4 -binding site to a 100-amino acid fragment within the N-terminal domain of paxillin. This ␣ 4 -binding fragment blocked the binding of paxillin to the ␣ 4 tail to a much greater extent than to FAT. Furthermore, when expressed in cells, this ␣ 4 -binding paxillin fragment inhibited ␣ 4 tail-stimulated cell migration, but not migration mediated by integrin ␣ 5 ␤ 1 . Thus, the binding of paxillin to the ␣ 4 tail leads to enhanced cell migration and specific inhibition of ␣ 4 -paxillin interaction selectively blocks ␣ 4 -dependent cellular responses.

Integrin Cytoplasmic Domain Model Proteins, Recombinant Paxillin Mutant Proteins, and Binding of Paxillin Mutants to Model Proteins-
The design and production of recombinant integrin cytoplasmic domain model proteins have been described (11,25). Each recombinant model protein was expressed in BL21(DE3)pLysS cells (Novagen), isolated by Ni 2ϩ -charged resins, and further purified to Ͼ90% homogeneity using a reverse-phase C18 high performance liquid chromatography column (Vydac). Masses of all proteins were assessed by electrospray ionization mass spectrometry on an API-III quadrupole spectrometer (Sciex, Toronto, Canada) and varied by less than 0.1% from the predicted mass.
The expression and isolation of recombinant glutathione S-transferase (GST) fusion protein of wild-type paxillin, mutants of P⌬(Ile 43 -Gly 60 ), P⌬(Ala 57 -Asn 99 ), P⌬(Gln 101 -Glu 226 ), P(Y31A/Y118A/Y181A), Nterminal domain, P(Met 1 -Gly 315 ), and C-terminal LIM domain, P(Gly 326 -Cys 557 ), have been described previously (26,27). Recombinant full-length paxillin (GST-free) was produced by thrombin digestion of recombinant GST-paxillin fusion protein and purification through glutathione-Sepharose 4B column (Amersham Biosciences). C-terminal truncation mutants of N-terminal domain of paxillin were created by site-directed mutagenesis using the QuikChange kit (Stratagene). Primers for QuikChange reactions were designed so that at each truncation site, the amino acid codon was replaced with a stop codon. Polymerase chain reaction (PCR) was performed using wild-type GSTpaxillin cDNA construct as a template following the manufacturer's instruction. Each site-directed mutation was then confirmed by cDNA sequencing, and expression and isolation of the mutant protein were performed as described (11,26). For construction of other paxillin mutants, PCR was used to generate a BamHI-XhoI fragment for each mutant. Each PCR product was ligated into the pCR vector using a TA cloning kit (Invitrogen). After cDNA sequencing, each fragment was ligated into BamHI-XhoI sites of pGEX-4T-3 vector (Amersham Biosciences), and expression and isolation of each GST fusion protein were performed as described (11,26).
Binding of recombinant paxillin or its mutants to integrin tail model proteins was performed as described (11,25). Briefly, aliquots of recombinant GST fusion protein of paxillin or its mutants were mixed with 300 l of buffer A: 10 mM Pipes, 50 mM NaCl, 150 mM sucrose, 1 mM Na 3 VO 4 , 50 mM NaF, 40 mM sodium pyrophosphate, pH 6.8, plus 0.5% sodium deoxycholate, 1 mM EDTA, 20 g/ml aprotinin, 5 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 0.1% Triton X-100, 3 mM MgCl 2 , and 1 mg/ml bovine serum albumin (BSA), added to 20 l of model protein-loaded resins, and incubated at room temperature with rotation for 2 h. Resins were then washed three times with the same buffer. Bound proteins were extracted with reducing SDS sample buffer, separated on SDS-polyacrylamide gels (PAGE), and detected with antibodies specific for HA-tag or GST followed by ECL (Amersham Biosciences).
Enzyme-linked Immunosorbent Assays-Wells of Ni-NTA HisSorb strips (Qiagen) were coated with 100 l of His-tagged integrin model proteins or His-tagged FAT, a recombinant protein derived from the focal adhesion targeting sequence of focal adhesion kinase (FAK), dissolved in phosphate-buffered saline (PBS) plus 0.2% BSA at 4°C overnight. The next day, wells were washed with PBS three times and blocked with 150 l of 1% (w/v) heat-denatured BSA at room temperature for 1 h. The wells were then washed with PBS three times. 100 l of recombinant paxillin or its mutants at different concentrations dissolved in PBS plus 0.2% (w/v) BSA was added to each well and incubated at room temperature for 1 h. Unbound proteins were washed out with PBS three times. Bound proteins were stained with mouse anti-HA tag (1:1,000 dilution in PBS plus 1% BSA) or mouse anti-GST antibodies (1:1,000 dilution in PBS plus 1% BSA) for 1 h at room temperature, followed by 1 h incubation with horseradish peroxidaseconjugated goat anti-mouse IgG (1:1,000 dilution in PBS plus 1% BSA) (BioSource). After three washes, bound proteins were assayed by measuring peroxidase activity with o-phenylenediamine as a substrate and quantified by reading its optical density at 490 nm. For competition assays using paxillin fragment, P(Ala 176 -Asp 275 ), or recombinant fulllength paxillin (GST-free), the same modified ELISA assays were performed except that different concentrations of this fragment was included in the paxillin solution added to the integrin model protein-or FAT-coated wells. Data were expressed as percentage of inhibition: (1 Ϫ B/B 0 ) ϫ 100%; where B ϭ A 490 in the presence of the competitor and B 0 ϭ A 490 in its absence.
cDNA Construction, Transfection, and Expression of Paxillin Mutants, Immunoprecipitation, and Western Blotting-For construction of mammalian expression vectors encoding paxillin fragments, PCR was used to generate an XhoI-Hind III fragment including Myc tag, EQKLI-SEEDL, sequence at the 3Ј end of each paxillin fragment sequence. Each PCR product was ligated into the pCR vector using a TA cloning kit (Invitrogen). After confirmation by cDNA sequencing, each fragment was ligated into XhoI-Hind III sites of pcDNA3.1(Ϫ) vector (Invitrogen). ␣ IIb ␣ 4 ␤ 3 ␤ 1A -expressing CHO cells were co-transfected with vector encoding each paxillin fragment plus vector encoding GFP (cDNA ratio of paxillin fragment to GFP; 50:1), or a control vector plus GFP at the same ratio, using LipofectAMINE transfection (LipofectAMINE PLUS, Invitrogen) following the manufacturer's instructions. Forty-eight hours after transfection, the cells were trypsinized and resuspended. Aliquots of cells were used for cell adhesion or migration as described below, and other cells were lysed using RIPA buffer. Expression of each paxillin fragment was detected by Western blot analysis on the cell lysate using a monoclonal antibody specific for Myc tag (9E10) or polyclonal antibodies specific for paxillin described previously (28,29). Immunoprecipitation was performed as described previously (10,11). Briefly, for co-precipitation of the ␣ 4 integrin with the ␣ 4 -binding paxillin fragment, cell lysate from ␣ IIb ␣ 4 ␤ 3 ␤ 1A -expressing CHO cells transiently transfected with a P(Ala 176 -Lys 277 ) construct was precipitated using a monoclonal antibody specific for Myc tag (9E10). The precipitated proteins were detected by Western blot analysis using polyclonal antibodies against the ␣ 4 cytoplasmic domain. The same blot was then stripped and blotted with polyclonal antibodies specific for paxillin. For co-precipitation of the ␣ 4 integrin, paxillin, and pp125 FAK , cell lysate from ␣ 4 ␤ 1 -expressing CHO cells was precipitated with polyclonal antibodies specific for pp125 FAK (10). Co-precipitation of the intact ␣ 4 integrin, paxillin, and pp125 FAK were detected using polyclonal antibodies against the ␣ 4 cytoplasmic domain (29) or pp125 FAK (C-20, Santa Cruz), and a monoclonal antibody specific for paxillin (clone 349, Transduction Laboratory), respectively.
Cell Adhesion and Migration Assays-Assays of cell adhesion and migration on fibrinogen or fibronectin (FN) were performed as described previously (10,11). Briefly, for cell adhesion assay, 24-well plates were coated with 10 g/ml fibrinogen or FN in a coating buffer: NaCl, 150 mM; NaH 2 PO 4 , 50 mM; and Na 2 HPO 4 , pH 8.0, at 4°C overnight and blocked with 1% heat-denatured BSA at 37°C for more than 1 h. Equal numbers of ␣ IIb ␣ 4 ␤ 3 ␤ 1A -expressing CHO cells transfected with different cDNA constructs as described were plated on the fibrinogen-or FN-coated wells and incubated in a 37°C incubator for 30 min. At the end of the experiment, unattached cells were washed away with PBS. Attached cells were fixed with 3.7% paraformaldehyde for 15 min at room temperature, washed twice with PBS, and counted under a microscope with high magnification.
For cell migration using Transwell chambers (8 m, Costar), both sides of chambers were coated with 10 g/ml fibrinogen or FN overnight at 4°C. The coated chambers were blocked with 1% heat-denatured BSA. 100 l of transfected cells (1.0 ϫ 10 5 cells) resuspended in Dulbecco's modified Eagle's medium plus 0.5% FBS were added to the upper chamber and 500 l of same medium added to the lower chamber. The cells were then allowed to migrate at 37°C for 4 h. At the end of the experiment, the cells migrated to the lower side were collected and counted either under a microscope with high magnification or counted using fluorescence-activated cell sorting analysis.
For cell migration assay using real time video phase-contrast microscopy, cells (2.0 ϫ 10 4 ) were plated on coverslips coated with 10 g/ml fibrinogen or FN. Dishes for cell migration were prepared as described previously (30). Dishes were placed in an open chamber with atmospheric and temperature control and cell movement viewed with a Nikon DiaPhot Microscope equipped with a SenSys cooled CCD video camera linked to a Silicon Graphics work station running the Inovision ISEE software program. ␣ IIb ␣ 4 ␤ 3 ␤ 1A Expressing CHO cells transfected with vectors encoding paxillin fragment plus GFP, or a control vector plus GFP as described above were detected by immunofluorescence and random cell migration of these cells were assessed by time-lapse imaging beginning 45 min after cell plating, and followed by recoding of every 5-10 min intervals for 5 h. At the end of the experiment, images of cells were outlined and the centroid (cell center) calculated. Displacement of the centroid was then used to determine cell movement over time.

Mapping of the ␣ 4 Integrin Cytoplasmic Domain Binding Site
in Paxillin-Paxillin binding to the ␣ 4 cytoplasmic domain accounts for some of unusual biological properties of the ␣ 4 integrins (10,11,29). We employed integrin tail model protein affinity chromatography to identify the regions of paxillin responsible for binding to the ␣ 4 cytoplasmic domain. The Nterminal half of paxillin, P(Met 1 -Gly 315 ), bound to the ␣ 4 cytoplasmic tail to the same extent as the full-length protein (Fig.  1, B and C). In contrast, the C-terminal half, comprised of four LIM domains, P(Gly 326 -Cys 557 ), failed to bind (Fig. 1, B and C). Thus, the N-terminal region of paxillin is necessary and sufficient for paxillin binding.
The N-terminal half of paxillin contains several regions known to mediate protein-protein interactions. This includes a Pro-rich domain responsible for its interaction with the SH3 domains of Src and Crk family members (31), however, removal of this domain, P⌬(Ile 43 -Gly 60 ) (Fig. 1B), was without effect on binding to the ␣ 4 tail. The N terminus of paxillin also contains 5 LD repeats known to be involved in its interactions with binding partners (14). An internal deletion that disrupts LD repeats LD2 and LD3, P⌬(Gln 101 -Glu 226 ), partially blocked binding to the ␣ 4 tail (Fig. 1, B and C), In contrast, a further N-terminal deletion, P⌬(Ala 57 -Asn 99 ), did not abolish ␣ 4 binding activity (Fig. 1, B and C). In addition, paxillin contains multiple tyrosines that can become phosphorylated to mediate binding to Crk adaptors or Csk kinase (12,13). However, alanine substitutions at these Tyr residues, P(Y31A,T118A, T181A), did not affect the ␣ 4 binding (Fig. 1, B and C). These data indicate that the ␣ 4 binding function of paxillin can be separated from many of its other binding activities and that residues contained in the Gln 101 -Glu 226 interval contribute to this activity.
To further narrow the localization of the paxillin-binding site, we analyzed sequential C-terminal truncation mutants of the N-terminal half of paxillin. P275X bound to the ␣ 4 tail, suggesting that the last 50 amino acid residues of N-terminal region of paxillin are dispensable for the ␣ 4 binding (Fig. 2). However, removal of 50 more residues (P225X) markedly reduced the binding to ϳ25% of that of N terminus. An additional 50-residue truncation (P175X) blocked binding completely (Fig.  2). Thus, these data show that sequences between residues Ala 176 -Asp 275 are required for paxillin to bind the ␣ 4 tail.
The foregoing studies identified a 100-residue sequence required for paxillin binding to the ␣ 4 tail. To determine whether sequences from this region were sufficient for ␣ 4 binding, we assessed the capacity of a fragment containing these residues, P(Ala 176 -Asp 275 ), to bind to the ␣ 4 tail. This fragment bound the ␣ 4 tail, but to a lesser extent than the complete N terminus of paxillin, P(1-315) (Fig. 3). In contrast, smaller fragments P(Glu 226 -Cys 325 ), P(Ala 176 -Glu 225 ), P(Phe 227 -Asp 275 ), and P(Phe 276 -Cys 325 ) were nearly devoid of activity ( Fig. 3 and data not shown). In addition, each individual LD domain, i.e. LD1 to LD5, revealed very weak binding to the ␣ 4 tail that was similar to that of P(Phe 276 -Cys 325 ) (data not shown). Thus, each of the LD repeats may contribute to the binding of paxillin to the ␣ 4 tail, accounting for the reduced affinity of P(Ala 176 -Asp 275 ) relative to the intact protein. However, the residues contained between Ala 176 and Asp 275 are sufficient for detectable paxillin binding to the ␣ 4 tail. ␣ 4 Integrin, Paxillin, and the FAT Domain of pp125 FAK Can Form a Ternary Complex-We previously hypothesized that the ␣ 4 -paxillin and pp125 FAK form a ternary complex leading to the increased membrane targeting and clustering of pp125 FAK and rapid pp125 FAK phosphorylation (10). To directly test this idea, we used affinity chromatography to examine the interactions among the ␣ 4 tail, paxillin, and FAT, a fragment of pp125 FAK which contains its paxillin-binding site (32). Paxillin directly bound to the ␣ 4 tail, whereas FAT did not show detectable direct binding (Fig. 4A). In contrast, in the presence of paxillin, FAT binding was detected. Neither paxillin nor FAT bound to the ␣ IIb tail (Fig. 4A). Thus, FAT does not directly bind to the ␣ 4 tail but it does interact with the ␣ 4 tail through paxillin. Furthermore, the presence of FAT, even at a 100-fold molar excess, did not inhibit paxillin binding or lead to increased FAT binding to the ␣ 4 tail (Fig. 4B and data not shown). In addition, using the ELISA assay that we developed (see "Materials and Methods"), we were also able to demonstrate the formation of a ternary complex of the ␣ 4 tail, paxillin, and FAT (data not shown). Therefore, the ␣ 4 integrin tail, paxillin, and FAT can form a ternary complex. To test whether the intact ␣ 4 integrin, paxillin, and pp125 FAK also form a ternary complex in vivo, we performed co-precipitation experiments. As shown in Fig. 4C, both the ␣ 4 integrin and paxillin co-precipitated with the pp125 FAK . Thus, the intact ␣ 4 integrin, paxillin, and pp125 FAK can also form a ternary complex in cells.

P(Ala 176 -Asp 275 ), an ␣ 4 -Binding Fragment of Paxillin, Blocks Paxillin Binding to the ␣ 4 Tail More Efficiently Than to the Focal Adhesion Targeting
Sequence of pp125 FAK -To further characterize the interactions of paxillin with ␣ 4 and FAT, we developed a quantitative ELISA assay. In the assay, the ␣ 4 tail model protein or FAT were immobilized on Ni 2ϩ -chelated wells through their hexahistidine tags, ensuring a uniform orientation of the immobilized ligand. Binding of full-length paxillin to the immobilized ␣ 4 tail was saturable with an EC 50 of ϳ4 nM (Fig. 5A). Binding was specific because no interaction was detected with immobilized ␣ IIb tail and a recombinant full-length paxillin (GST-free) completely blocked GST-paxillin binding to the ␣ 4 tail (Fig. 5, A and D). Similarly, in the ELISA assays, paxillin binding to FAT was specific and saturable (Fig.  5C). The ␣ 4 binding 100-residue fragment, P(Ala 176 -Asp 275 ), bound with a reduced affinity (EC 50 ϳ27 nM, Fig. 5B). Thus, both paxillin and the P(Ala 176 -Asp 275 ) bind tightly to the ␣ 4 tail. Since P(Ala 176 -Asp 275 ) contains both LD3 and LD4 motifs, in which LD4 motif has been shown to mediate pp125 FAKpaxillin interaction (20), it is possible that this fragment might also interfere with the pp125 FAK -paxillin interaction. To deter-  staining (A, bottom panel). B, recombinant HA-tagged paxillin-GST was added to Ni 2ϩ -charged resins in the absence or presence of FAT at the indicated concentration. Bound fractions were collected and separated on 4 -20% SDS-PAGE under reducing conditions, transferred to a nitrocellulose membrane, and stained with antibody specific for HA-tagged paxillin-GST. C, cell lysate from the ␣ 4 ␤ 1 -expressing CHO cells was immunoprecipitated using antibodies specific for pp125 FAK (␣-FAK) or a control rabbit IgG (IgG) as described under "Materials and Methods." The precipitated proteins were separated on 4 -20% SDS-PAGE and detected with antibodies specific for the ␣ 4 cytoplasmic domain, paxillin, and pp125 FAK , respectively. Bound paxillin was detected with an antibody specific for GST as described under "Materials and Methods." The same experiment was performed using GST as a competitor. Even at 20-fold molar excess, no significant inhibition on GST-paxillin binding to the ␣ 4 tail by GST was observed (data not shown). mine whether P(Ala 176 -Asp 275 ) competes for paxillin binding to ␣ 4 or to FAT, we performed the paxillin-binding ELISA assays in the presence of the P(Ala 176 -Asp 275 ) fragment. The P(Ala 176 -Asp 275 ) fragment effectively competed for paxillin binding to ␣ 4 , producing 50% inhibition in the presence of 10-fold molar excess of the fragment (Fig. 6A). A 20-fold molar excess of P(Ala 176 -Asp 275 ) blocked binding by more than 95% (Fig. 6A). In sharp contrast, P(Ala 176 -Asp 275 ) had a much weaker effect on paxillin binding to FAT. At 10-fold excess, P(Ala 176 -Asp 275 ) did not have a significant inhibitory effect on paxillin binding to FAT (Fig. 6A). Even at 100-fold molar excess, this fragment only inhibited paxillin binding by ϳ30% (Fig. 6A). In contrast, at 6-fold molar excess, the full-length paxillin (GST-free) completely inhibited GST-paxillin binding to FAT (Fig. 6B). Thus, P(Ala 176 -Asp 275 ) specifically blocked paxillin binding to the ␣ 4 tail more efficiently than it blocked binding to FAT.
The ␣ 4 -Binding Fragment of Paxillin Specifically Inhibits ␣ 4 Tail-dependent Cell Migration-The previous finding that a mutation of the ␣ 4 tail that blocks paxillin binding reduces cell migration (10) suggests that inhibitors of paxillin binding to ␣ 4 could perturb ␣ 4 -dependent cell migration. To test this idea, we transfected cells with plasmids encoding P(Ala 176 -Lys 277 ), the fragment that blocked ␣ 4 -paxillin interactions with minimal effects on interactions with pp125 FAK . We used CHO cells expressing a chimeric integrin, ␣ IIb ␣ 4 ␤ 3 ␤ 1A , that contains the ␣ 4 cytoplasmic domains in place of that of ␣ IIb (10, 11). The presence of the ␣ 4 cytoplasmic domain promotes cell migration and inhibits cell spreading when the cells adhere to an ␣ IIb ␤ 3 ligand, fibrinogen (10,11). Expression of P(Ala 176 -Lys 277 ) markedly inhibited cell migration on fibrinogen (Fig. 7A, left  panel). Furthermore, expression of P(Ala 176 -Lys 277 ) fragment also reversed the inhibition of cell spreading by the ␣ 4 -paxillin interaction. 2 In contrast, expression of a fragment of paxillin that failed to bind to the ␣ 4 tail, P(Met 1 -Lys 125 ), did not inhibit cell migration on fibrinogen (Fig. 7A, left panel). Importantly, both fragments were well expressed (Fig. 7B). Interestingly, the inhibitory fragment was expressed at a level only 3-4-fold greater than that of endogenous paxillin, yet it dramatically reduced cell migration. The inhibition of cell migration by the ␣ 4 -binding paxillin fragment was associated with its binding to the integrin, since the ␣ IIb ␣ 4 chimeric integrin co-precipitated with the ␣ 4 -binding paxillin fragment, P(Ala 176 -Asp 275 ) (Fig.  7C). Thus, expression of a paxillin fragment that binds to the ␣ 4 tail and disrupts its interaction with paxillin inhibited cell migration mediated by the ␣ 4 cytoplasmic domain.
As noted above, P(Ala 176 -Asp 275 ) was much less effective at blocking the binding of paxillin to FAT. The paxillin-FAK interaction may be involved in cell migration mediated by many classes of integrins, suggesting that P(Ala 176 -Asp 275 ) would not efficiently inhibit migration mediated by integrins other than ␣ 4 ␤ 1 and ␣ 4 ␤ 7 . To test this idea, we examined the effect of this fragment on cell migration on fibronectin (FN), which is mediated by the endogenous CHO cell integrin, ␣ 5 ␤ 1 (33). Expression of P(Ala 176 -Lys 277 ) or of P(Met 1 -Lys 125 ) had minimal, statistically insignificant effects on cell migration on FN (Fig.  7, lower panel). In addition, expression of these fragments did not affect cell adhesion to either fibrinogen or FN (data not shown). Thus, introduction of an inhibitor of the paxillin-␣ 4 interactions selectively blocked ␣ 4 tail-mediated cell migration.
To further analyze the effect of P(Ala 176 -Lys 277 ) on ␣ 4 -dependent cell migration, we performed cell migration assays using time lapse video microscopy. Cells expressing P(Ala 176 -Lys 277 ) were significantly less motile on fibrinogen, 5.5 ϩ 0.6 mm/h, than those expressing vector control, 12.5 ϩ 2.6 mm/h, or untransfected cells, 14.2 ϩ 2.6 mm/h (Fig. 8, A and C). In contrast, cells expressing this fragment migrated at a signifi-cantly higher rate, 9.4 ϩ 0.9 mm/h, on FN (Fig. 8, B and C). Thus, P(Ala 176 -Lys 277 ) specifically inhibited the ␣ 4 taildependent cell random migration. DISCUSSION In the current study, we have mapped the ␣ 4 integrinbinding region within paxillin and examined its effect on the ␣ 4 cytoplasmic domain-mediated cellular functions. We found that: 1) sequences between residues Ala 176 and Asp 275 of paxillin are sufficient for binding to the ␣ 4 tail; 2) the ␣ 4 integrin tail, paxillin, and FAT can form a ternary complex; 3) the ␣ 4 -binding paxillin fragment, Ala 176 -Asp 275 , specifically blocked paxillin binding to the ␣ 4 tail more efficiently than it blocked binding to FAT; and 4) this fragment specifically blocked cell migration stimulated by the ␣ 4 cytoplasmic domain. Thus, this fragment contains sequences that are required for binding to the ␣ 4 cytoplasmic domain and can function as a dominant negative inhibitor of ␣ 4 integrinmediated cellular functions.
Previously, we suggested that the ␣ 4 , paxillin, and pp125 FAK might form a ternary complex. This complex might increase the membrane targeting and clustering of pp125 FAK and induce the rapid phosphorylation of pp125 FAK , which might account for the increased cell migration in the ␣ 4 -mediated cell adhesion. In the current study, we have provided direct evidence indicating that indeed the ␣ 4 integrin tail, paxillin, and FAT, the focal adhesion targeting domain of pp125 FAK which contains the paxillin-binding region, can form a ternary complex. FAT was unable to bind the ␣ 4 tail directly and also did not inhibit paxillin binding to the ␣ 4 , and it can only interact with the ␣ 4 through paxillin. In addition, our data indicate that the intact ␣ 4 integrin, paxillin, and pp125 FAK can also form a ternary complex in cells. Thus, these data further support the direct association between the ␣ 4 and paxillin and suggest that the ␣ 4 and pp125 FAK might interact with paxillin through different sites. Thus, the ternary complex of the ␣ 4 integrin and paxillin and pp125 FAK might account for the rapid tyrosine phosphorylation and activation of pp125 FAK . This increased activation of pp125 FAK may then contribute to increased ␣ 4 integrin-mediated (10) cell migration because pp125 FAK has been implicated in stimulating cell migration (23,24).
The N-terminal domain of paxillin contains five LD motifs that mediate protein-protein interactions (12)(13)(14). For example, the LD1, LD2, and LD4 motifs mediate paxillin binding to vinculin, the LD2 and LD4 motifs are involved in its interaction with pp125 FAK , and paxillin binds PKL through its LD4 motif (12,13,20). Since P(Ala 176 -Asp 275 ), the ␣ 4 -binding fragment identified in the current study contains both LD3 and Ld4 motifs and the LD4 motif was able to bind pp125 FAK directly (20), we reasoned that it was possible that P(Ala 176 -Asp 275 ) might also interfere paxillin-pp125 FAK interaction. However, our results indicate that the P(Ala 176 -Asp 275 ) fragment only partially (30% of inhibition at 100-fold molar excess, Fig. 6A) blocked paxillin binding to FAT, whereas the full-length paxillin effectively (Ͼ90% inhibition at 6-fold molar excess, Fig.  6B) blocked the binding. In sharp contrast, the same fragment effectively inhibited paxillin binding to the ␣ 4 tail, with 95% inhibition at a 20-fold molar excess. These data indicate that the P(Ala 176 -Asp 275 ) fragment is more potent in inhibiting the ␣ 4 -paxillin interaction than pp125 FAK -paxillin interaction. One possible explanation is that since LD2 and LD4 motifs both can mediate pp125 FAK -paxillin interaction (20), it is possible that in the presence of excess P(Ala 176 -Asp 275 ) fragment, the pp125 FAK -paxillin interaction is most likely mediated by the LD2 motif. Therefore, pp125 FAK can still bind paxillin even though the P(Ala 176 -Asp 275 ) fragment might affect pp125 FAK -LD4 interaction. Using two independent cell migration assays, that is, random cell migration using Transwell chambers and real time video phase-contrast microscopy, we have shown that the ␣ 4binding fragment of paxillin, P(Ala 176 -Lys 277 ), when expressed in the ␣ IIb ␣ 4 ␤ 3 ␤ 1A -expressing CHO cells, effectively blocked the ␣ 4 tail-stimulated cell migration on fibrinogen, whereas it had a minor effect on cell migration on FN, a ligand for the endogenous ␣ 5 ␤ 1 integrin (Figs. 7 and 8). In contrast, P(Met 1 -Lys 125 ), a paxillin fragment that failed to bind the ␣ 4 in vitro, did not have an inhibitory effect on either ␣ 4 -or ␣ 5 -mediated cell migration (Fig. 7). Since pp125 FAK -paxillin interaction may be required for integrin-mediated cell migration (23,24,32), the modest effect of P(Ala 176 -Asp 277 ) on ␣ 5 ␤ 1 -mediated cell migration, suggests that it did not perturb the pp125 FAK -paxillin interaction in vivo. This interpretation is consistent with the modest effects in vitro reported here. Thus, this ␣ 4 -binding paxillin fragment can function as a dominant negative effector for the ␣ 4 integrin-paxillin interaction and specifically inhibit the functions of ␣ 4 integrins.