p21-activated protein kinase 4 (PAK4) interacts with the keratinocyte growth factor receptor and participates in keratinocyte growth factor-mediated inhibition of oxidant-induced cell death.

Keratinocyte growth factor (KGF), a member of the fibroblast growth factor (FGF) family (also known as FGF-7), is an important protective factor for epithelial cells. The receptor for KGF (also called FGFR2-IIIb), which has intrinsic tyrosine kinase activity, is expressed specifically on epithelial cells and in the lung epithelium. Administration of KGF has been shown to protect the lung from various insults, but the mechanism of protection is not well understood. To understand the mechanism by which KGF exerts protective functions on epithelial cells, we used the yeast two-hybrid assay to identify proteins that interact with the KGF receptor (KGFR). Here we show that the cytoplasmic domain of KGFR interacts with p21-activated protein kinase (PAK) 4, which is a new member of the PAK family. The PAKs are regulated by the Rho-family GTPases Rac and Cdc42. PAK4 is the most divergent member of the PAK family of proteins and may have distinct functions. However, stimuli that regulate PAK4 activity are not known. Our data show that PAK4 can associate with the KGFR, which is dependent on KGFR tyrosine kinase activity. We show that a dominant negative mutant of PAK4 blocks KGF-mediated inhibition of caspase-3 activation in epithelial cells subjected to oxidant stress. Our data demonstrate that PAK4 is an important mediator of the anti-apoptotic effects of KGF on epithelial cells.

Fibroblast growth factors (FGFs) 1 are a large family of growth factors that control cell proliferation, differentiation, and organ development (1)(2)(3)(4)(5). The activities of FGFs are me-diated by binding to receptors that have tyrosine kinase activity. To date, four FGF receptors have been described, namely FGFR1, FGFR 2, FGFR 3, FGFR 4, and their splice variants (3)(4)(5). Each FGF receptor contains an extracellular domain composed of 2-3 immunoglobulin (Ig)-like loops, a transmembrane segment, and an intracellular domain with tyrosine kinase activity. In the case of FGFR1-3, the third Ig loop is encoded by two exons, an invariant exon called IIIa that is spliced to either exon IIIb or IIIc, yielding receptors with distinct ligand binding abilities. FGFR2-IIIb binds FGF1, FGF3, FGF7 (also called keratinocyte growth factor or KGF), and FGF10 (6,7). It is expressed specifically on epithelial cells and is diffusely expressed in the day 11 lung (8). Gene-targeting studies have revealed an indispensable role for this receptor in mouse lung development. Transgenic mice expressing a dominant negative mutant of FGFR2-IIIb under the control of a lung-specific promoter exhibit embryonic lethality with grossly abnormal lung development with only two primordial epithelial tubes and no branching morphogenesis (9). Similarly, mice with homozygous mutations in FGFR2-IIIb are viable until birth but show severe defects with no evidence of lung formation (10). A key observation in these mice is that although the growth of organs such as lungs and limbs is initiated, further development is not sustained in the face of extensive apoptosis in the tissues (10).
Interactions between FGFR2-IIIb and its ligands are just as important in adults. For example, FGF7/KGF has been shown to be essential in wound reepithelialization after skin injury (11). The administration of KGF has been shown to protect the lungs of animals from a variety of insults, including oxygen (12), radiation, chemotherapy (bleomycin) (13,14), and acid (15). However, the detailed downstream signaling pathways that contribute to KGF-FGFR2-IIIb-mediated cell protection are not well understood.
Here we show that FGFR2-IIIb interacts with a protein that shares 86% identity with a member of the PAK family of kinases, namely PAK4, previously identified in human cells (16). PAK proteins were originally described as downstream targets for the Rho GTPases Rac and Cdc42 (17,18). The PAK proteins can be essentially divided into two categories, Groups I and II, based on their structures (19). In Group I are the proteins PAK1, PAK2, and PAK3. Each of these proteins has an amino-terminal regulatory domain and a carboxyl-terminal kinase domain. The regulatory domain also includes the GTPase-binding domain, which mediates binding to Rac and Cdc42. These PAK members also bind to SH3 domain-containing adapter proteins such as Nck through the proline residues in the regulatory domain (17,18). The carboxyl terminus in-teracts with yeast G protein ␤␥-subunits, which suggests that these PAK proteins may be regulated by heterotrimeric G proteins in mammalian cells. In Group II are PAK4 (16), PAK5, and PAK6 (19). Although overall human PAK4 resembles other PAK members in that it also contains an aminoterminal GTPase-binding domain and a carboxyl-terminal kinase domain, it lacks a G protein ␤␥-binding domain or the ability to bind to Nck (16). Even within the GTPase-binding domain and the kinase domain, it shares only 50% similarity with other PAK members (16). Although a role for PAK proteins in the regulation of cytoskeletal organization is well described, PAK proteins, particularly those belonging to Group II, have been implicated in other distinct cellular processes (19). For example, PAK6 has been shown to bind to the androgen receptor and repress androgen receptormediated gene transcription in the presence of the ligand androgen (20). The different members of the PAK family have been shown to have different roles in the apoptotic response of cells. Although PAK2 promotes apoptosis (21,22), PAK1 has been shown to protect cells from apoptosis induced by growth factor withdrawal (23,24). Recently, PAK4 was shown to protect cells from apoptosis induced by multiple stimuli, including serum withdrawal, tumor necrosis factor-␣ (TNF␣) treatment, or UV irradiation (25). Although PAK4 has been shown to be a target for Cdc42, extracellular stimuli that regulate endogenous PAK4 activity have not been identified. Here, using the yeast two-hybrid assay, we have identified an association between the KGFR and PAK4. Our stud-ies suggest that this association may play an important role in the protective effects of KGF on epithelial cells.

EXPERIMENTAL PROCEDURES
Interaction Cloning of Murine PAK4 -Duplex-A TM yeast two-hybrid system (OriGene Technologies, Inc., Rockville, MD) was used to identify KGFR-interacting proteins. The KGFR cytoplasmic domain containing the tyrosine kinase domain was cloned into the yeast expression plasmid pEG202-NLS, and the resulting bait construct was named pNEGKGFRc. Expression of the cytoplasmic domain of the KGFR in the yeast strain EGY194 led to autophosphorylation of the receptor as judged by immunoprecipitation with an anti-phosphotyrosine antibody and Western blotting with an anti-KGFR antibody. We first ensured that the bait construct, pNEGKGFRc containing the LexA-KGFRc fusion, did not activate reporter genes due to autoactivation. Autoactivation of the reporter gene lacZ was checked by co-transforming EGY194 with pNEGKGFRc and the reporter plasmid pSH18 -34. No activation of lacZ was observed with the LexA-KGFRc fusion protein. To identify KGFR-interacting proteins, EGY194 was first transformed with pNEGKGFRc and pSH18 -34, and the pretransformed EGY194 was next transformed with a cDNA expression library constructed from cDNA derived from a 19-day-old post-coital mouse embryo fused to the B42 activation domain HA-tagged expression vector pJG4 -5. After selecting clones potentially expressing interacting proteins, the specificity of the interaction was confirmed by a yeast-mating assay. The expression plasmid isolated from the putative positive clone was introduced into EGY194 (a strain), and a combination of the bait plasmid (pNEGKGFRc) and reporter plasmid (pSH18 -34) was introduced into EGY40 (␣-strain) and used in the mating test. Appropriate controls were used in the mating assay. Among the positive clones, three independent clones contained sequences that matched with the sequence of human PAK4. Full-length cDNA of mPAK4 was cloned by 5Ј-RACE (Clontech). The mouse lung cDNA library (Clontech) was applied as a template in touchdown PCR using PfuTurbo (Stratagene), which amplified the full-length mPAK4 gene. The PCR product was subcloned into a TOPO vector (Invitrogen). SP6 and T7 primers were used for sequencing and, after sequence confirmation, the full-length cDNA was named mPAK4.
Immunoprecipitation and Western Blotting-To express the protein(s) of interest, the constructs were either transfected individually or co-transfected into HEK293 cells. 36 h post-transfection, cells were starved in serum-free Dulbecco's modified Eagle's medium for 3 h and then stimulated with or without KGF (100 ng/ml, Roche Molecular Biochemicals) or NGF (100 ng/ml, Roche Molecular Biochemicals), depending on whether we wished to stimulate the endogenous receptor or the chimeric receptor in trk-KGFR-transfected cells. Cells were then lysed in immunoprecipitation (IP) buffer containing protease inhibitors (20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerolphosphate, 1 mM Na 3 VO 3 , and the complete protease-inhibitor mixture from Roche Molecular Biochemicals, catalog number 1697498). Protein concentration in the cell extracts was detected and measured using the DC TM protein assay kit (Bio-Rad, catalog number 500-0112). In general, the appropriate polyclonal antibody coupled with Sepharose was used for immunoprecipitation, and either unconjugated mouse monoclonal antibody or enzyme-conjugated antibody from the same species was used for immunoblotting. Cell extracts and the appropriate antibody coupled to Sepharose beads were incubated together at 4°C for 12-16 h. Immunoprecipitates were collected by centrifugation and washed fours times with IP buffer. The immunoprecipitates were resolved by SDS-PAGE, transferred to polyvinylidene difluoride membrane (Millipore), and probed with the appropriate antibodies. The antibodies used were anti-HA affinity matrix and anti-HA-peroxidase high affinity antibodies, catalog numbers 1815016 and 2013819 respectively (Roche Molecular Biochemicals), anti-Bek (KGFR) antibody (SC122, Santa Cruz Biotechnology), anti-Grb2 rabbit polyclonal antibody (SC-255AC, Santa Cruz Biotechnology), anti-Grb2 monoclonal antibody (SC8034, Santa Cruz Biotechnology), anti-phosphotyrosine (pY) antibody pY99 (SC7020 AC, Santa Cruz Biotechnology), and anti-TrkA antibody against the extracellular domain of TrkA (06574, Upstate Biotechnology). Blots were visualized by enhanced chemiluminescence. Cell extracts were immunoprecipitated with control Ig (con. IG) or anti-KGFR, and immunoprecipitates were analyzed by immunoblotting with anti-human PAK4 (Cell Signaling). The blot was stripped and reprobed with anti-KGFR. Cell extracts were also directly analyzed with anti-PAK4 to assess PAK4 expression in the cells. B, association of HA-murine PAK4 (HA-mPAK4) with endogenous KGFR. HEK293 cells were transfected with HA-tagged mPAK4 (pHA-mPAK4). 36 h posttransfection, cells were stimulated with KGF (100 ng/ml) for 30 min after 3 h of serum deprivation. Cells were lysed, and 0.4 mg of protein in the lysates was used for IP with anti-HA-coupled agarose (Roche Molecular Biochemicals). The immunoprecipitates were analyzed by Western blotting techniques using anti-KGFR (Santa Cruz Biotechnology). The membrane was stripped and reprobed with anti-HA to demonstrate co-immunoprecipitation of HA-mPAK4. C, Co-immunoprecipitation of trk/KGFR and mPAK4. HEK293 cells were transiently cotransfected with ptrk/KGFR (1.25 or 2.5 g) and pHA-mPAK4. 0.4 mg of cell extract protein was used for immunoprecipitation with anti-HA (anti-HA was coupled to agarose beads, Roche Molecular Biochemicals). The immunoprecipitates were analyzed by immunoblotting with anti-trkA (Upstate Biotechnology). The membranes were stripped and reprobed with anti-HA (Roche Molecular Biochemicals).

RESULTS
Cloning of Murine PAK4 -To identify proteins that interact with FGFR2III-b/KGFR, we used the yeast two-hybrid system to screen for proteins that interact with the kinase (cytoplasmic) domain of KGFR. A murine embryonic cDNA library was screened for possible interactions with a chimeric bait comprising LexA and the cytoplasmic domain of KGFR. Among the several positive clones identified in this assay, three independent clones contained DNA fragments with open reading frames that were highly homologous to the carboxyl terminus of human PAK4. The corresponding full-length cDNA was cloned using 5Ј-RACE. Fig. 1A shows the predicted amino acid sequence of the full-length cDNA. Domain search revealed the presence of a p21 Rho-binding domain (PBD) and a protein kinase domain (PKD) in hPAK4 (GenBank TM Accession number NM_005884). The amino acid sequence was found to share 86% identity with that of hPAK4 (Fig. 1B) and less than 50% identity, only in the PKD domain, with the corresponding domain present in murine PAK1-3. We therefore consider this to be murine PAK4 (mPAK4; GenBank TM Accession number AY217016).
PAK4 Associates with KGFR in Mammalian Cells-We investigated whether endogenous KGFR and PAK4 associate with each other in HEK293 epithelial cells. HEK293 cell extracts were immunoprecipitated with anti-KGFR antibody or control immunoglobulin, and analysis of the immunoprecipitates by immunoblotting with anti-human PAK4 revealed the presence of PAK4 in immunoprecipitates obtained with the anti-PAK4 but not the control Ig ( Fig. 2A). The association between the two proteins was detected in the absence of KGF and was not appreciably augmented upon treatment of the cells with KGF. Interaction between these two proteins was also detected when HA-tagged murine PAK4 was used (Fig. 2B). To further explore interactions between KGFR and PAK4, we expressed a chimeric receptor containing the extracellular ligand-binding domain of NGF (Trk) and the cytoplasmic domain of KGFR, which we termed trk-KGFR. The rationale for using the chimeric receptor was that, because epithelial cells do not express the NGF receptor, treatment of cells with NGF would allow us to specifically monitor the activity of the hybrid receptor without interference from the endogenous KGF receptor. Also, in the absence of an effective anti-KGFR antibody that could be used for immunoprecipitation, the anti-TrkA antibody was particularly useful. As shown in Fig. 2C, at a lower level of trk-KGFR expression, NGF-stimulation was required for coimmunoprecipitation of trk-KGFR and PAK4 (Fig. 2C). However, at higher levels of expression of the chimeric receptor, the association was readily detectable with (Fig. 2C) or without (not shown) NGF stimulation. Because receptor tyrosine kinases are known to undergo ligand-independent activation (autophosphorylation) due to dimerization when the protein is overexpressed, it is possible that phosphorylation augments or stabilizes association between PAK4 and KGFR. We also found that the tyrosine residues 542 and 543, corresponding to tyrosine residues 653 and 654 in FGFR1 (which are autophosphorylation sites of KGFR), are critical for the association with PAK4 (Fig 3A). However, mutations of tyrosine residues 352 and 655 corresponding to tyrosine residues 463 and 766 in FGF receptor 1, the putative binding sites for Shc and PLC-␥, respectively, with FGF receptor 1 (26 -28), did not affect association with PAK4. It is to be noted that, compared with the wild type receptor, these mutant receptors displayed reduced basal autophosphorylation, which may be due to reduced receptor dimerization when these residues are mutated.
Grb2 Participates in the Association between mPAK4 and KGFR-Grb2 is an adapter protein that associates with receptor tyrosine kinases and couples the activated receptor to downstream signaling molecules such as Ras (28,29). The PAK family members have been shown to associate with the adapter protein Nck, which has similarities to Grb2 (30,31). However, PAK4, the sequence of which is the most divergent of all PAK proteins, does not associate with Nck (16). We therefore investigated whether PAK4 has the ability to associate with Grb2. Because Nck and PAK1 have been shown to form a complex in the absence of growth factor stimulation (32), we investigated whether PAK4 exists in a complex with Grb2 in unstimulated cells. As shown in Fig. 4A, anti-Grb2 antibody-immunoprecipitated PAK4 and KGF had no effect on Grb2-PAK4 association. We also investigated whether endogenous KGFR associates with Grb2. Immunoprecipitation of HEK293 cell extracts with anti-Bek but not control Ig led to co-immunoprecipitation of Grb2 as revealed by immunoblotting of the immunoprecipitates with anti-Grb2 (Fig. 4B). If anti-Bek and anti-Grb2 have comparable affinities for their respective proteins, then a comparison of the amount of Grb2 co-immunoprecipitated with KGFR with the net Grb2 obtained by immunoprecipitation with monoclonal anti-Grb2 shows that a small fraction of Grb2 is complexed with KGFR in the cells, which probably reflects the level of KGFR expression in the cells. Also, we detected similar levels of Grb2 in the immunoprecipitate whether cells were left with or without KGF. To determine the importance of specific phosphorylation sites for KGFR-Grb2 association, HEK293 cells were transfected with expression vectors for trk-KGFR (wild type or individual mutants). We uniformly treated all cells with NGF to achieve maximum receptor activation, because overexpression of the wild type receptor consistently resulted in basal autophosphorylation and, therefore, only small differences between stimulated and unstimulated cells were expected (Fig. 3). As shown in Fig. 4C, the wild type receptor associated with Grb2 and the autophosphorylation site of the receptor was critical for this association. Again, the other tyrosine residues, Tyr 352 and Tyr 655 , did not affect association with Grb2 and, similar to what was observed with PAK4, the mutation of Tyr 352 actually augmented association of Grb2 with the receptor. The receptor-Grb2 complex also contained PAK4 (Fig. 4D). Because the endogenous KGFR was found to associate with both PAK4 and Grb2 in the absence of KGF stimulation of cells, and because for both associations the autophosphorylation sites in the KGFR were found to be critical for association, it is possible that the endogenous KGFR (at least a fraction of the receptor) in HEK293 cells exists in an autophosphorylated state. Also, because Grb2 is widely expressed in many species, including yeasts, it is possible that the association between KGFR and PAK4 is indirect and mediated by Grb2, because Grb2 has been shown to bind to receptor tyrosine kinases in yeast two-hybrid assays. Also, we were unable to co-immunoprecipitate KGFR and PAK4 when the recombinant proteins were expressed in bacteria, which reinforces our suspicion that Grb2 mediates association between the receptor and PAK4.
PAK4 Is Phosphorylated by Ligand-activated trk/KGFR-The association between PAK4 and KGFR ( Fig. 1) prompted us to investigate whether PAK4 undergoes tyrosine phosphorylation in response to receptor activation. As shown in Fig. 5, activation of the hybrid receptor with NGF induced tyrosine phosphorylation of PAK4, which was clearly evident in cells that expressed higher levels of the chimeric receptor. The phosphorylation status paralleled that of the chimeric receptor. The bottom panel of Fig. 5 demonstrates equivalent expression of HA-PAK4 in all cells.

Dominant Negative Mutant of PAK4 (PAK4/NT) Prevents KGF-mediated Inhibition of Oxidant-induced Poly(ADP-ribose)
Polymerase (PARP) Cleavage-Because our data suggested an important role for PAK4 in cellular protection by KGF, we investigated whether interference of PAK4 function would influence anti-apoptotic functions induced by KGF in HEK293 cells. For this purpose, we treated HEK293 cells with H 2 O 2 , which has been shown to induce pro-apoptotic pathways in these cells (33). Because apoptosis is associated with caspase-3 activation, a focal point of different pro-apoptotic pathways, we investigated the state of protein PARP, a substrate of caspase-3, in the cells under different conditions of treatment. Also, to inhibit the functions of endogenous PAK4, we overexpressed the N-terminal fragment of PAK4 lacking the kinase domain, the inactivation of which imparts dominant negative functions to PAK4 (16,34). As shown in Fig. 6, treatment of cells with H 2 O 2 promoted cleavage of the PARP protein. Cleavage of PARP was inhibited in the presence of the caspase-3 inhibitor DEVD-fluoromethylketone or KGF. However, when the PAK4 dominant negative mutant was expressed in the cells, KGF was unable to inhibit PARP cleavage. Furthermore, Fig. 6C shows that the PAK4 mutant is able to interact with the KGF receptor. By binding to the receptor, the mutant is probably able to disrupt KGFR signaling to downstream pathways. DISCUSSION In this study, we have identified PAK4 as a KGFR-interacting protein using a yeast two-hybrid assay. We show that this interaction has functional relevance, because expression of a dominant negative mutant of PAK4 prevented the inhibition of oxidant-induced caspase-3 activation by KGF in epithelial cells. It is interesting to note that just as mice deficient in the KGF receptor show organ malformation with evidence of enhanced apoptosis of developing organs such as the lung (9, 10), Drosophila lacking the protein Mushroom Body Tiny (MBT), the closest known homolog of PAK4, also shows defects in organ development with a reduced number of Kenyon cells in a structure in the brain called the mushroom body (35).
Human PAK4 was identified as the most divergent member of the PAK family of proteins (16). The mouse PAK4 we cloned shares 86% homology with hPAK4, with greatest identity in the p21 Rho-binding domain and the protein kinase domain. In addition to its role as a cytoskeletal regulatory protein, PAK4 has been shown to induce anchorage-independent growth and promote cell survival in response to different stimuli such as tumor necrosis factor ␣ and UV irradiation. Although important functions of PAK4 have been identified, factors that stimulate PAK4 activity have heretofore not been identified. The KGFR-PAK4 association, as identified in this study, provides a link between extracellular stimuli and endogenous PAK4 activity. We also show that the KGFR-PAK4 complex is dependent on the tyrosine kinase activity of the KGFR and that PAK4 undergoes tyrosine phosphorylation upon receptor activation. In previous studies, tyrosine phosphorylation of PAK1 was reported in v-ErbB-transformed cells, and dephosphorylation of PAK led to a reduction in PAK kinase activity (36). Interestingly, tyrosine phosphorylation of PAK was not Ras-, Rac-or Cdc42-dependent, which led to the suggestion that this activity of PAK1 may be distinct from growth factor-induced mitogenic activity (36). Similarly, PAK4 was also shown to promote anchorage-independent growth (34), and both PAK1 and PAK4 have been shown to possess anti-apoptotic functions and cause phosphorylation and inactivation of the death-promoting protein Bad (23)(24)(25).
It has been reported previously that PAK1 is recruited to growth factor receptors such as the epidermal growth factor (EGF) and the platelet-derived growth factor (PDGF) receptors through association with Nck (37,38). However, unlike other PAK proteins, PAK4 does not interact with Nck. We show that Grb2, an adaptor protein similar to Nck, associates with PAK4. Grb2 is recruited to ligand-activated growth factors, which results in targeting of the Ras guanylnucleotide exchange factor SOS to the plasma membrane location of Ras, which, in turn, promotes Raf-1 recruitment and activation of the MAP kinase pathway (28, 29, 39 -41). In our studies, we show that Grb2 associates with PAK4 and the chimeric trk-KGFR. The association between Grb2 and PAK4 may involve the SH3 domains in Grb2 and the proline-rich PXXP residues in the p21-binding domain in PAK4. Thus, although independent studies have implicated both molecules, KGF and PAK4, in protection from cell death, our study connects them in a novel signaling axis that may have critical protective functions during organ development in the embryonic stage and during injury in adults.