p21-activated Kinase 4 Phosphorylation of Integrin β5 Ser-759 and Ser-762 Regulates Cell Migration*

Modulation of integrin αvβ5 regulates vascular permeability, angiogenesis, and tumor dissemination. In addition, we previously found a role for p21-activated kinase 4 (PAK4) in selective regulation of integrin αvβ5-mediated cell motility (Zhang, H., Li, Z., Viklund, E. K., and Strömblad, S. (2002) J. Cell Biol. 158, 1287–1297). This report focuses on the molecular mechanisms of this regulation. We here identified a unique PAK4-binding membrane-proximal integrin β5-SERS-motif involved in controlling cell attachment and migration. We also mapped the integrin β5-binding site within PAK4. We found that PAK4 binding to integrin β5 was not sufficient to promote cell migration, but that PAK4 kinase activity was required for PAK4 promotion of cell motility. Importantly, PAK4 specifically phosphorylated the integrin β5 subunit at Ser-759 and Ser-762 within the β5-SERS-motif. Point mutation of these two serine residues abolished the PAK4-induced cell migration, indicating a functional role for these phosphorylations in migration. Our results may give important leads to the functional regulation of integrin αvβ5, with implications for vascular permeability, angiogenesis, and cancer dissemination.

Integrins are heterodimers composed of ␣ and ␤ subunits that contain a large extracellular domain with ligand binding capacity, a transmembrane domain, and a short cytoplasmic domain (2). Integrins mediate cell adhesion and thereby play a central role in cell migration. Integrins also participate in bidirectional signaling processes across the plasma membrane where ligand binding to the extracellular matrix generates intracellular signals (outside-in) and where signaling affecting integrin cytoplasmic tails can lead to regulation of the integrin extracellular binding capacity (inside-out) (3)(4)(5)(6).
The membrane-proximal regions of the cytoplasmic domains of the integrin heterodimer interact with each other via a salt bridge. Disruption of this salt bridge can change the integrin affinity state, indicating that the short cytoplasmic domains of integrins are essential for both inside-out and outside-in signaling (3,7). One important mechanism whereby cellular signaling influences cellular behavior is through the interaction and/or phosphorylation of integrin cytoplasmic tails by intracellular proteins (8 -12). Interactions and/or phosphorylation of integrin cytoplasmic tails may also regulate the activation state of integrins, for example, ␤ integrin tyrosine phosphorylation can regulate talin-induced integrin activation (11,(13)(14)(15)(16).
Integrin ␣v␤5 mediates cell attachment and migration on vitronectin (2,6,34,35). Integrin ␣v␤5 is induced in keratinocytes during wound healing and facilitates vascular endothelial growth factor-induced vascular permeability (36 -38). In addition, growth factor activation of integrin ␣v␤5-mediated cell motility has been functionally linked to angiogenesis as well as carcinoma cell dissemination (39 -41). However, how integrin ␣v␤5 itself may be controlled and the role of its cytoplasmic domains are both unclear.
Overexpression of p21-activated kinase 4 (PAK4) 5 can induce localized actin polymerization and filopodia formation (42) and affect cell adhesion and anchorage-independent growth of rodent fibroblasts (43,44). We previously found that PAK4 interacts with the integrin ␤5 cytoplasmic tail and promotes integrin ␣v␤5-mediated cell migration (1). However, whether PAK4 promotes cell motility through its interaction with integrin ␣v␤5 and/or its effects on the actin cytoskeleton remains unknown.
In this study, we used site-directed mutagenesis to map the PAK4-binding site within integrin ␤5 as well as the integrin ␤5-binding site within PAK4. Importantly, we also identified PAK4-mediated phosphorylation of two serine residues in the integrin ␤5 cytoplasmic domain that are involved in the regulation of cell motility. These results provide important information regarding the intracellular regulation of ␣v␤5 activity.
Yeast Two-hybrid Assay-Yeast mating test assay was performed using the DupLEX-A Yeast Two-Hybrid System (OriGen Technologies) as previously described (46). The PAK4 C terminus (CT) (amino acids 396 -591) was subcloned into the EcoRI and XhoI sites of the pJG4 -5 prey vector and transformed into the EGY48 yeast strain, including the pSH18 -34 reporter gene (lacZ) plasmid. The integrin ␤5 cytoplasmic domain (␤5-aa-753-799) and various fragments thereof were subcloned into the EcoRI and XhoI sites of the pEG202 bait vector. Point mutations in the integrin ␤5 tail were generated with the QuikChange site-Directed mutagenesis kit using pEG202-␤5-tail as a template and then transformed into the yeast strain RFY206. The correct reading frames and sequences were verified by sequencing. Yeast mating tests were performed by using the RFY206 strain, including different integrin ␤5-tail constructs with the EGY48 strain, including the PAK4-CT construct.
Cell Transfection-4 -8 g of total DNA was transfected in 10-cm cell culture dishes (80 -90% confluence of COS-7 or MCF-7 cells), using Lipofectamine Plus TM (Invitrogen), according to the manufacturer's protocol, and cells were used 30 -48 h after transfection. For transfection of CS-1 cells, 4 -8 g of total DNA was transfected in 6-well plates (1 ϫ 10 6 cells/ well) using the Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. CS-1 cells stably transfected with integrin ␤5 WT or mutant clones were established in medium containing 600 g/ml G418. Selected cell colonies were transferred to separate culture dishes and were subsequently grown in 500 g/ml G418 medium. Pools of mixed populations of stable transfectant CS-1 cells expressing comparable levels of each of the wild-type or mutant integrins were established by three times consecutive fluorescence-activated cell sorting with anti-integrin ␣v␤5 mAb P1F6 performed over a 3-month period.
Pulldown Assay-For in vitro GST pulldown assays, the integrin ␤1-tail, ␤5-tail (amino acids 753-799), and ␤5-tail mutants were individually expressed as GST fusion proteins using the bacterial expression vector pGEM-1T (Amersham Biosciences). GST fusion proteins were produced and purified using glutathione-Sepharose beads (Amersham Biosciences) according to the manufacturer's protocol. GST pulldown assays were performed as described (1). Briefly, 200 g of lysates from COS-7 cells overexpressing various hPAK4 constructs were incubated with 5 g of GST fusion proteins. The result was visualized by immunoblotting, and band intensities were measured using Kodak one-dimensional image analysis or ImageJ 1.43 software (National Institutes of Health).
Kinase Activity Assay and Phosphopeptide Mapping-Various PAK4 constructs were expressed in COS-7 cells and lysed in kinase lysis buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 1% Nonidet P-40, 10% glycerol 150 mM NaCl) with addition of fresh protease inhibitors (0.5 g/ml leupeptin, 1 mM EDTA, 1 g/ml pepstatin A, 0.2 mM phenylmethylsulfonyl fluoride) and a serine/threonine protein phosphatase inhibitor mixture (Sigma), followed by immunoprecipitation. PAK4 kinase activity was determined in a kinase buffer (50 mM Hepes, pH 7.5, 10 mM MgCl 2 , 2 mM MnCl 2 , 0.2 mM dithiothreitol) in the presence of 30 M cold ATP and 10 Ci of [␥-32 P]ATP (3000 Ci/nM, Amersham Biosciences) and in the presence of 5 g of substrate (MBP, GST, GST-␤1 tail, or GST-␤5 tail) for 30 min at 30°C. Incubation was stopped in Laemmli buffer, and samples were heated at 95°C for 4 min. Phosphorylated proteins were separated by 12.5% SDS-PAGE. The gel was dried and visualized by autoradiography. Phosphorylation sites in GST-␤5 were mapped as described previously (47). Briefly, phosphopeptides were resolved by 10% SDS-PAGE and transferred to nitrocellulose membrane. GST or GST-␤5 corresponding bands were excised and digested with trypsin as described (48). The first dimension electrophoresis was carried out in a pH 1.9 buffer, and the second dimension separation was performed using TLC in isobutyric acid buffer. The chromatography plates were exposed using Fuji Bas Bio-Imaging Analyzer, and radioactive peptides were scraped off the plate, followed by sequencing and phosphoamino acid analysis. For Edman degradation, phosphopeptides were coupled to Sequelon-AA membranes (Millipore) according to the manufacturer's instructions and sequenced on an Applied Biosystems Gas Phase sequencer. The activity in released phenylthiohydantoin derivatives from each cycle was quantified using the Bio-Imaging Analyzer. For phosphoamino acid analysis, peptides were lyophilized and thereafter hydrolyzed in 6 M HCl for 1 h at 110°C, followed by TLC as described (49). To determine PAK4 phosphorylation of the integrin ␤5 subunit in living cells, COS-7 cells transfected with HA-PAK4 underwent a phosphate starvation for 6 h at 40 h post transfection, followed by metabolic labeling with 300 Ci of [␥-32 P]ATP for 2 h at 37°C. Cells were then washed twice with phosphate-free Dulbecco's modified Eagle's medium and lysed in radioimmune precipitation assay buffer. Integrin ␣v␤5 was immunoprecipitated by mAb P1F6, and the phosphorylated ␤5 subunit was visualized by autoradiography.
Cell Migration Assay-Haptotactic cell migration assays were performed using Transwell chambers (Costar Inc.) with 8.0-m pore size as described before (1). Briefly, 1 ϫ 10 5 MCF-7 cells transiently transfected with EGFP, pEGFP-PAK4, or PAK4 mutants were added on top of the Transwell membranes coated with VN on the bottom and allowed to migrate toward VN for 6 h at 37°C in adhesion buffer (1,35). The migrated cells were counted using a fluorescence microscope, and adjustment was made for the transfection rate of each population as determined by flow cytometry or by manual counting using fluorescence microscopy. For CS-1 cell migration assay, 1 ϫ 10 5 CS-1 cells transiently transfected or stable clones with pcDNA3 empty vector, integrin ␤5, or ␤5 mutants were added into the Transwell membranes and allowed to migrate toward VN for 24 h at 37°C. The migrated cells were stained by 0.5% crystal violet and counted using a microscope (10ϫ objective). For CS-1 cells stably expressing integrin ␤5 or mutant ␤5 that were co-transfected to express EGFP or EGFP-PAK4, cell migration was quantified as described above for EGFP/EGFP-PAK4 co-transfected MCF-7 cells.
Flow Cytometry Analysis-Cell surface expression levels of integrins were analyzed by measurement of fluorescein isothiocyanate staining intensity by FACScan flow cytometer using CellQuest software (BD Biosciences) after staining with antiintegrin ␣v␤5 mAb P1F6 and a fluorescein isothiocyanateconjugated goat anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories) essentially as described by Bao and Strömblad (50).
Statistical Analysis-Data were analyzed for statistical significance using an un-paired two-tailed t test.

Identification of an Integrin ␤5 Cytoplasmic Tail SERS-motif
Involved in PAK4 Binding-We previously found that PAK4 binds to the integrin ␤5-subunit and promotes integrin ␣v␤5-mediated cell motility (1). We here mapped the PAK4-binding site within the integrin ␤5-subunit cytoplasmic domain at the amino acid level. Firstly, we analyzed PAK4 interaction in yeast two-hybrid mating tests (a schematic outline is shown in Fig.  1A) of WT ␤5 and 13 deletion mutants or fragments thereof as indicated in Fig. 1B. The smallest ␤5 fragment maintaining full binding strength to PAK4 contained a 10-amino acid N-terminal ␤5-tail sequence (␤5-753-762). Any further deletions of this fragment caused loss or partial loss of PAK4 binding. Further, truncation of this 10-amino acid fragment revealed that a ␤5-SERS motif (␤5-759-762) was required for PAK4 binding. Thus, the results suggest that full PAK4 binding requires a 10-aa region of the integrin ␤5 tail (aa 753-762) and that the membrane-proximal SERS-motif (aa 759 -762) within ␤5 is critical for PAK4 binding. These results were verified in mammalian cells by GST pulldown analyses, where the ␤5-753-762 fragment pulled down the same amount of PAK4 as the ␤5-WT cytoplasmic tail (Fig. 1, C and D). However, deletion of the ␤5-SERS-motif (aa 759 -762) abolished PAK4 binding (Fig.  1D). Among integrins, the SERS amino acid sequence only appears in the integrin ␤5 cytoplasmic tail, although ERS is found at the same position within integrin ␤6 and ER within integrin ␤3 (Fig. 1E).
Using yeast two-hybrid mating tests, we then fine-mapped the ten amino acids within the integrin ␤5 membrane-proximal region involved in PAK4 binding (Fig. 1F). Mutations of ␤5 aa 756, 757, 758, 759, or 762 did not affect its ability to bind PAK4. However, mutation of ␤5-EF 753 and 754 to alanine residues partially disrupted PAK4 binding, and reversing the charges by swapping ER 760 and 761 to RE completely abolished PAK4 binding. However, mutation of aa ER 760 and 761 to alanine residues did not influence PAK4 binding. The effect of the ER to RE swap of ␤5 aa 760 and 761 for binding to PAK4 in mammalian cells was verified by a GST pulldown assay (Fig. 1G). Together, this identified amino acids 753, 754, 760, and 761 in integrin ␤5 as involved in PAK4 binding and the SERS-motif as critical for PAK4 binding to the integrin ␤5 subunit.
PAK4 Regulation of Integrin ␣v␤5-mediated Cell Attachment and Migration May Involve the Integrin ␤5-tail SERS-motif-Overexpression of WT PAK4 in MCF-7 cells stimulated integrin ␣v␤5-mediated cell migration on VN (1). To determine whether the SERS-motif in the integrin ␤5-tail may play a role in PAK4-mediated regulation of cell motility, we co-transfected WT PAK4 with WT integrin ␤5 or with non-PAK4-binding integrin ␤5 SERS-motif mutants in CS-1 cells. These transfected cells were analyzed for the effects on integrin ␣v␤5-mediated cell attachment and migration. When CS-1 cells were transiently transfected with WT FLAG-PAK4 and WT integrin ␤5 using a double-cassette vector, cell attachment decreased and cell migration increased as compared with cells transfected with WT ␤5 alone (Fig. 3, A-D). Consistently, CS-1 cells stably expressing WT ␤5 were markedly affected by EGFP-PAK4 overexpression resulting in decreased cell attachment and increased cell migration on VN as compared with EGFP control-transfected cells (Fig. 3, E and F). Interestingly, CS-1 cells stably expressing ␤5-ER760,761RE or ␤5-⌬SERS displayed similar levels of cell attachment and migration as when WT ␤5 expressing cells were co-transfected with PAK4. However, no significant changes in cell attachment or migration were observed when either of the mutant ␤5 expressing cells were transfected with EGFP-PAK4 as compared with control-transfected cells (Fig. 3, E and F). This suggests that the PAK4-binding ␤5-SERS-motif may be involved in PAK4-mediated regulation of cell attachment and migration.
Mapping of the Integrin ␤5-binding Site within PAK4-To further elucidate the role of the PAK4 to integrin ␤5 interaction, the ␤5-binding site in PAK4 was mapped. The PAK4 C-terminal region (PAK4-CT, aa 396 -591) directly interacts with integrin ␤5 in yeast two-hybrid mating tests and the critical IBD was located within aa 505-533 (1). To analyze if this region of PAK4 was also critical for association with integrin ␣v␤5 in mammalian cells, FLAG-PAK4-WT, a kinase-inactive PAK4 mutant (K350M), FLAG-PAK4-⌬505-533 (⌬IBD), and FLAG-PAK4-⌬69 -221 (Fig. 4A) were transiently expressed in COS-7 cells, and a GST pulldown assay was performed (Fig.  4B). Cell extracts of FLAG-PAK4-WT and ⌬69 -221 were pulled down by GST-␤5 tail fusion protein but not K350M or ⌬IBD. This showed that the PAK4-IBD was required for integrin ␤5-association in mammalian cells. Surprisingly, the kinase-inactive PAK4-K350M that contains the IBD did not associate with integrin ␤5. This is in contrast with our previous results from yeast two-hybrid mating tests where a C-terminal PAK4 fragment with a K350M mutation could still interact with integrin ␤5 (1). This may be explained by conformation differences between the K350M-mutated full-length PAK4 and a mutated PAK4 C-terminal fragment. However, the fact that eliminating the PAK4 kinase activity also blocked its integrin binding capacity raised the question whether PAK4 kinase activity may be correlated to integrin binding. Therefore, the kinase activity of the PAK4 constructs was tested. FLAG-BAP, PAK4-WT, K350M, ⌬IBD, and ⌬69 -221 were transiently transfected into COS-7 cells, and the FLAG-PAK4 immunoprecipitates were used in an in vitro kinase activity assay. WT PAK4 and PAK4-⌬69 -221 displayed autophosphorylation and phosphorylated the substrate MBP, whereas K350M and ⌬IBD did not (Fig. 4C). Together, this indicates that the PAK4-IBD is required for integrin binding and that the PAK4 kinase activity correlates with integrin ␤5 binding in mammalian cells.
To further reveal the relationship between the PAK4 integrin binding capacity and its kinase activity, we sought to separate the kinase activity of PAK4 from its integrin binding capacity by introducing nine different point mutations into PAK4-IBD at amino acids conserved among the PAK family (Fig. 5, A and B), because other PAKs can also bind integrins. 6 To test the effect of these PAK4 mutations on PAK4 association with integrin ␤5 in mammalian cells, a GST pulldown assay was performed with lysates from COS-7 cells overexpressing FLAG-tagged bacterial alkaline phosphatase (BAP) used as negative control, PAK4-WT, PAK4-K350M, and the nine PAK4 point mutations introduced in full-length WT PAK4. As shown in Fig. 5C, the mutants K350M; PP513,514AA; PP519,520AA; and A523P displayed no or weak association with the integrin ␤5-tail in mammalian cells, whereas all the other PAK4 point mutants tested displayed a similar positive ␤5 association as WT PAK4.
The effect of the nine point mutations on PAK4 kinase activity was analyzed using an in vitro kinase assay (Fig. 5D). We observed that the kinase-defective PAK4-K350M as well as three other PAK4 mutants; PP513,514AA; PP519,520AA; and A523P displayed no significant kinase activity. These were the same four PAK4 full-length mutants that also showed no or weak integrin binding capacity. All other mutants displayed equal or stronger PAK4 kinase activity and autophosphorylation compared with WT PAK4. Given that the same four point mutants that lacked kinase activity were also impaired 6 H. Zhang and S. Strö mblad, unpublished observation. in associating with integrin ␤5, we were unable to separate the kinase activity of PAK4 and its integrin binding capacity. Based on this, we hypothesize that the PAK4 integrin-binding site might be located in a substrate-binding pocket of PAK4. Alternatively, PAK4 kinase activity may be required for full-length PAK4 integrin binding.

Effects of PAK4 Mutations on Integrin ␣v␤5-mediated Cell
Motility-We previously demonstrated that overexpression of PAK4 induced integrin ␣v␤5-mediated carcinoma cell migration (1). We now analyzed if the mutations of PAK4-IBD that disrupted integrin binding also affected the capacity of PAK4 to induce integrin ␣v␤5-mediated cell migration (Fig. 5E). Con- Statistically discernable differences compared with PAK4-WT according to t test are indicated (*, p Յ 0.05; **, p Յ 0.01). E, MCF-7 cells transiently transfected with control EGFP, EGFP-PAK4 WT, or EGFP-PAK4 mutants as indicated were analyzed for haptotactic cell motility toward VN. Bar graphs show quantified cell motility relative to the EGFP control (mean value Ϯ S.E. of at least three experiments). Statistically discernable differences compared with EGFP control according to t test are indicated (*, p Յ 0.05; **, p Յ 0.01). sistent with our previous findings, overexpression of WT PAK4 in MCF-7 cells induced integrin ␣v␤5-mediated cell migration on VN ϳ2-3 times above control levels. However, the kinaseinactive PAK4-K350M and the PAK4-⌬IBD mutants failed to stimulate ␣v␤5-mediated cell migration (Fig. 5E). Furthermore, the three point mutations in PAK4-IBD that disrupted the PAK4 integrin binding capacity together with the PAK4 kinase activity did not promote MCF-7 cell migration, whereas WT PAK4 and all other six PAK4-IBD point mutants were able to significantly induce cell motility (Fig. 5E). This suggests that the integrin binding capacity, kinase activity, and the capability to induce cell motility of PAK4 are correlated.
PAK4-mediated Integrin ␤5 Binding Is Not Sufficient to Promote Cell Migration-PAK4-CT, which lacks the PAK4 ATPbinding pocket, strongly interacted with the integrin ␤5 cytoplasmic domain in yeast mating tests (1). We compared PAK4-NT (N-terminal aa 1-322) and two C-terminal truncated mutants, PAK4-KD (kinase domain, aa 323-591) and PAK4-CT (C-terminal, aa 396 -591) (Fig. 6A), for association with integrin ␤5 in mammalian cells by a GST pulldown assay (Fig. 6B). Although PAK4-WT, PAK4-KD, and PAK4-CT were pulled down by the GST-␤5 tail fusion protein, PAK4-NT was not. Thus, deletion of the NT domain or the ATP-binding domain did not affect the binding capacity of PAK4 to integrin ␤5 in mammalian cells. Given that PAK4-CT lacks the PAK4 ATP-binding pocket, it is conceivable that it also lacks kinase activity. To test this, we measured the kinase activity of PAK-CT and compared it to that of PAK4-WT, PAK4-NT, and PAK4-KD (Fig. 6C). Although PAK4-WT and PAK4-KD displayed high kinase activities, PAK4-CT and PAK4-NT did not show any significant kinase activity above background. This way, we identified a PAK-CT fragment that lacked kinase activity, but with an intact integrin binding capacity.
To test if PAK4-CT affected cell motility, we overexpressed PAK-WT, PAK4-NT, PAK-KD, and the PAK4-CT and analyzed integrin ␣v␤5-mediated cell motility of MCF-7 cells. We found that overexpression of PAK4-WT and PAK4-KD consistently enhanced cell motility (Fig. 6D). However, PAK4-CT as well as PAK4-NT failed to induce any substantial cell motility (Fig. 6D). This indicates that PAK4 kinase activity is required for promotion of cell motility, and that PAK4 integrin binding is not sufficient. Taken together, our results indicate that PAK4 kinase activity is critical to promote integrin ␣v␤5-mediated cell motility.
PAK4 Phosphorylates Integrin ␤5 Cytoplasmic Domain in Vitro and in Living Cells-Given the role of PAK4 kinase activity in promotion of cell motility, we examined PAK4-mediated phosphorylation of the integrin ␤5 cytoplasmic tail. Using HA-PAK4 immunoprecipitate prepared from transiently transfected COS-7 cells together with a GST-␤5 cytoplasmic domain fusion protein, we found that the ␤5 cytoplasmic domain was a specific substrate for PAK4 in vitro, whereas GST-␤1 integrin was not (Fig. 7A, top panel). Likewise, purified ␣v␤5 from human placenta could also be phosphorylated by PAK4 on its ␤5 subunit (Fig. 7B). Importantly, we also found that integrin ␤5 was phosphorylated in cells transfected with PAK4, as detected by metabolic 32 P labeling, but not in mock-transfected cells (Fig. 7C).
PAK4 Phosphorylates Integrin ␤5 at the Membrane Proximal Ser-759 and Ser-762-To determine the location of PAK4 phosphorylation within integrin ␤5, we employed phosphopeptide mapping. GST-␤5 tail and GST control proteins were treated with PAK4 in the presence of [␥-32 P]ATP, trypsinized, and subjected to two-dimensional electrophoresis on TLC plates. Two spots (b and c) appeared in the GST-␤5 tail sample that were not found in the GST control (Fig. 8A, upper). Following phosphoamino acid analysis, both spots b and c were identified as serines by comparison with the standard marker (Fig. 8A, middle and lower in boxes). Edman sequencing of the spot b and c peptides by 18 cycles of degradation showed that the spot b peptide contained a high level of 32 P at the first amino acid, whereas the spot c peptide at the third amino acid contained the most radioactivity. Thus, we identified two distinct PAK4 phosphorylation sites at amino acids Ser-759 and Ser-762 using the GST-␤5 cytoplasmic tail as a substrate (Fig. 8, A  and B). One additional spot immediately to the left of spot c was also repeatedly observed and was identified as a serine in the third position, consistent with a phosphorylation of Ser-759 in an incompletely cleaved fragment. However, potential additional phosphorylation site(s) cannot be excluded due to the built-in limitations of the two-dimensional-gel electrophoretic separation. Given that PAK4 is a serine/threonine kinase, we then mutated Ser-759 and Ser-762 to Thr-759 and Thr-762, and prepared GST-␤5-S759T; GST-␤5-S762T, and the double mutant GST-␤5-SS759,762TT to further examine the specificity of the PAK4-mediated phosphorylation sites in the ␤5 tail. In the wild type peptide, spot b contained only phosphorylated serine (Fig. 8C), whereas in the ␤5-S762T and ␤5-SS759,762TT mutants, it contained phosphorylated threonine. Further, spot c displayed only serine phosphorylation in WT ␤5 that was completely reverted to threonine phosphorylation in the ␤5-S759T and ␤5-SS759,762TT mutants. This confirms that both the serine residues 759 and 762 of integrin ␤5 were phosphorylated by PAK4. However, we cannot exclude additional ␤5 phosphorylation sites, as may be suggested by the observation that spot b displayed serine phosphorylation also upon mutation of GST-␤5-S762 to threonine. To further test if the two identified phosphorylation sites were responsible for the observed PAK4 phosphorylation, we generated a mutant with mutation of both Ser-759 and Ser-762 to alanine residues. As shown in Fig. 8D, PAK4 phosphorylated GST-␤5-WT significantly above GST control background, but not GST-␤5-SS759,762AA, indicating that Ser-759 and Ser-762 were responsible for the observed PAK4 phosphorylation. Amino acid sequence alignment of the human ␤ integrin cytoplasmic domains sharing high homology with integrin ␤5 indicates that Ser-759 is unique to ␤5, whereas Ser-762 is conserved between ␤5, ␤6, and ␤8, but with no corresponding residues in ␤1, ␤2, ␤3, or ␤7 (Fig. 1E). This suggests that phosphorylation of serine residues within the membraneproximal PAK4-binding SERS-motif is integrin-selective.
Role of Integrin ␤5 Ser-759 and Ser-762 in PAK4 Regulation of Cell Motility-Mutations of the integrin ␤5 to ␤5-SS759,762AA and ␤5-SS759,762EE did not influence PAK4 binding in yeast two-hybrid mating tests (Fig. 1F). To determine whether the two serines in the integrin ␤5 SERS-motif may play a role in PAK4-mediated regulation of cell motility, we made CS-1 cells stably expressing integrin ␤5-WT, ␤5-SS759,762AA, or ␤5-SS759,762EE. The cell surface expression levels of integrin ␤5 were similar to ␤5-WT, as determined by FACS analysis using anti-integrin ␣v␤5 mAb P1F6 (Fig. 9A). We then analyzed the effects on integrin ␣v␤5-mediated cell migration. When CS-1 cells stably expressing WT integrin ␤5 were transiently co-transfected with EGFP-PAK4-WT, cell migration increased compared with the same cells co-transfected with EGFP control. CS-1 cells expressing ␤5-SS759,762AA or ␤5-SS759,762EE mutants displayed a similar level of cell migration as WT ␤5-expressing CS-1 cells. However, no significant changes in cell migration were observed when the mutant ␤5-expressing cells were co-transfected with WT PAK4, in contrast with the WT ␤5-expressing cells (Fig. 9B). This demonstrates that integrin ␤5 Ser-759 and Ser-762 are critical for PAK4-induced cell motility and thus that PAK4-mediated phosphorylation of ␤5 cytoplasmic tail appears to regulate cell motility. This also indicates that a negative charge at positions 759 and 762 in the ␤5 cytoplasmic tail was not sufficient to promote cell migration.  ). B, in the same manner, 5 g of purified integrin ␣v␤5 was analyzed for phosphorylation by immunoprecipitated PAK4, separated by 7.5% SDS-PAGE, and visualized by autoradiography. C, PAK4 phosphorylates ␤5 subunit in living cells. Cells underwent phosphate starvation and then metabolic labeling as described under "Experimental Procedures." Integrin ␣v␤5 was immunoprecipitated in cells with or without overexpressed HA-PAK4, exposed to SDS-PAGE and autoradiography (upper panel). The lower panel shows the immunoblot for HA-PAK4 expression.

DISCUSSION
Cytoplasmic tails of integrins play key roles in a variety of integrin-mediated events, including adhesion and migration (2). The phosphorylation of integrin cytoplasmic tails has been proposed as a means of regulating integrin functions (10). Our data show that the integrin ␤5 subunit cytoplasmic tail is a substrate of PAK4. This is the first example of phosphorylation of integrin ␣v␤5, and how its extracellular functions may be controlled by intracellular phosphorylation. This also adds to the relatively few known substrates of PAK4 (52,53), thereby extending the knowledge about the immediate signal transduction capacity of PAK4.
The relationship between cell adhesion and migration is complex. Our results showed that, when compared with WT integrin ␤5, the ␤5-SERS mutations resulted in a marked induc-tion of integrin ␣v␤5-mediated cell migration accompanied by a decrease in ␣v␤5-mediated cell attachment. This could potentially be explained by the fact that cell migration compared with attachment strength can follow an approximate bell-shaped curve (54). The rate of cell migration is a function of matrix concentration, integrin abundance, and the integrin activation state. Change in any one of these properties will affect the rate of cell migration in a manner that is dependent on the original position of the cell on the bell-shaped curve (54,55). Further, overexpression of PAK4 mimicked the effect on cell motility and attachment of the SERS mutations in the ␤5-tail. Together, ␣v␤5 in its normal constitution may mediate cell attachment that is too strong for optimal motility, whereas PAK4 phosphorylation or SERS-motif mutation may lower the attachment strength and resultantly increased cell motility. A, PAK4-phosphorylated GST-␤5 was separated by two-dimensional-gel electrophoresis after trypsin digestion. Spots marked with b and c appeared consistently in the same location in different experiments and did not occur in the GST control. These two spots were further analyzed by phosphopeptide mapping and identified as serines 759 and 762 (middle and bottom panels). The inset shows results of phosphoamino acid analysis. B, arrowheads point out the two PAK4-induced phosphorylation sites within the integrin ␤5 membrane-proximal region at serine residues 759 and 762. C, phosphoamino acid content of peptides from spots b and c, analyzed from GST-␤5-WT, GST-␤5-S759T, GST-␤5-S762T, and GST-␤5-SS759,762TT, and GST fusion proteins were phosphorylated in vitro by PAK4. Migrating positions of phospho amino acid markers are shown in the panel to the right. D, the two phosphorylation sites at Ser-759 and Ser-762 were mutated to alanine residues to further elucidate the identity of the two phosphorylatable residues. PAK4 phosphorylation of GST-␤5 was compared with GST and the GST-␤5 alanine mutant.
The conserved ␤-integrin membrane-proximal region where PAK4 binds to and phosphorylates ␤5 is important for integrin oligomerization, inhibition of integrin conformational changes, and tethering of an integrin in the inactive state (56 -58). However, although it is possible that PAK4 binding to and phosphorylation of integrin ␤5 cytoplasmic tail membrane-proximal region may affect the association between integrin ␣v and ␤5 subunits and/or binding between ␣v␤5 and VN, further studies are required to elucidate if the phosphorylation of ␤5-subunit by PAK4 can affect integrin hinge formation or change the conformation of integrin extracellular domain.
Two other interesting proteins, theta-associated protein 20 (TAP20) and the scaffolding protein, receptor for activated C kinase (RACK1) can also interact with the integrin ␤5 cytoplasmic tail. TAP20 binds to the integrin ␤5 tail and reduces integrin ␣v␤5-mediated cell migration by a protein kinase C signaling pathway (59,60). RACK1 interacts with a conserved membrane-proximal region of the integrin ␤ subunit cytoplasmic domain and decreased Chinese hamster ovary cell motility in a manner that may also involve its interaction with protein kinase C (59,61). The RACK1-binding site in integrin ␤5 is overlapping with that of PAK4. However, further experiments are required to elucidate whether TAP20 and/or RACK1 may be functionally related to PAK4 in the regulation of integrin ␣v␤5-mediated cell motility.
Phosphorylation of integrin cytoplasmic tails can have both negative and positive roles in integrin regulation, possibly reflecting the importance of dynamically regulated phosphorylation in the integrin function (10). Integrin ␤5 tail can also be phosphorylated by protein kinase C, but the one or more specific serines to be phosphorylated were not identified (62). In this study, we demonstrate that PAK4 phosphorylation of serines 759/762 in the integrin ␤5 cytoplasmic tail promoted CS-1 cell migration. Among integrins, the SERS amino acid sequence only appears in the ␤5 cytoplasmic tail, although ERS is found at the same position within ␤6, ER within ␤3, and Ser-762 within ␤8. Therefore, PAK4 may also be able to phosphorylate also the integrin ␤6 and ␤8 cytoplasmic domains.
We demonstrate that phosphorylation of serine residues 759 and 762 in the integrin ␤5 cytoplasmic tail is necessary for PAK4-mediated promotion of cell migration. Growth factor stimulation can promote ␣v␤5-mediated cancer cell migration and dissemination as well as induce ␣v␤5-mediated angiogenesis and vascular permeability (35,36,39,41,63). Interestingly, PAK4 kinase can be activated by growth factors, such as Hepatocyte growth factor (64). It will therefore be interesting to elucidate the potential role of PAK4 in growth factor stimulation of angiogenesis, vascular permeability, and cancer cell migration and dissemination and determine if PAK4-mediated phosphorylation of integrin ␤5 serine residues 759 and 762 may play any role. However, PAK4 may also promote cell motility by phosphorylation of additional substrates involved in the control of cell motility, for example by regulation of the actin microfilament system (43).
In summary, we found that a unique membrane-proximal SERS-motif within the integrin ␤5 cytoplasmic domain can be phosphorylated by PAK4 and that this phosphorylation regulates integrin ␣v␤5-mediated carcinoma cell motility. This may also contribute to the understanding of intracellular signaling behind vascular permeability, angiogenesis, and carcinoma cell dissemination (36,39,63).