The Adaptor Protein Nck Links Receptor Tyrosine Kinases with the Serine-Threonine Kinase Pak1*

Nck is an adaptor protein composed of a single SH2 domain and three SH3 domains. Upon growth factor stimulation, Nck is recruited to receptor tyrosine ki- nases via its SH2 domain, probably initiating one or more signaling cascades. In this report, we show that Nck is bound in living cells to the serine-threonine kinase Pak1. The association between Nck and Pak1 is mediated by the second SH3 domain of Nck and a pro-line-rich sequence in the amino terminus of Pak1. We also show that Pak1 is recruited by activated epidermal growth factor (EGF) and platelet-derived growth factor receptors. Moreover, Pak1 kinase activity is increased in response to EGF in HeLa cells transfected with hu- man Pak1, and the kinase activity was enhanced when Nck was co-transfected. It is concluded that Nck links receptor tyrosine kinases with Pak1 and is probably involved in targeting and regulation of Pak1 activity. The Nck protein (1) belongs to the class of signaling molecules termed adaptors which, like Grb2, are composed solely of SH2 1 and SH3 domains, with no intrinsic catalytic activity. These proteins exert their biological function by coupling up-stream signals, usually those initiated by activation of receptor tyrosine kinases, to downstream elements in the cell (2). The role of the adaptor protein Grb2 in the Ras signaling pathway has well established

The Nck protein (1) belongs to the class of signaling molecules termed adaptors which, like Grb2, are composed solely of SH2 1 and SH3 domains, with no intrinsic catalytic activity. These proteins exert their biological function by coupling upstream signals, usually those initiated by activation of receptor tyrosine kinases, to downstream elements in the cell (2). The role of the adaptor protein Grb2 in the Ras signaling pathway has been well established (3).
Nck associates directly with activated PDGF-␤-receptor via its SH2 domain (4) and is recruited to EGF receptor upon EGF stimulation (5)(6)(7). Nck contains three SH3 domains and could potentially bind to at least three effector proteins. However, the proteins which form a complex with Nck in vivo to transmit downstream signals are largely unknown. It has been reported that upon overexpression Nck can bind to the guanine nucleo-tide exchange factor Sos (8), and that Nck can interact with p120 cbl (9) and the Wiskott-Aldrich syndrome protein WASP (10). The physiological relevance of these interactions remains uncertain. Recently, the Drosophila homologue of Nck has been identified (termed dock) in a genetic screen for mutants that disrupt photoreceptor guidance and targeting (11), thus providing a potential clue for a biological role of Nck in vertebrates. It was proposed that the Drosophila Nck may play a role in linking receptor tyrosine kinases with changes in organization of the actin cytoskeleton.
Pak (p21-associated kinase) proteins are serine-threonine kinases (12) homologous to the yeast Ste20 kinase, an enzyme involved in linking pheromone-activated G protein-coupled receptors to a MAP kinase cascade (13). Several distinct members of the Pak family have been identified (12, 14 -17). It has been shown that Pak proteins can be activated in vitro by the small GTP-binding proteins Cdc42 and Rac1. In addition, Pak proteins appear to participate in the activation of both the JNK and p38 MAP kinase signaling pathways (18 -21). Physiological substrates for Pak proteins have not been identified.
In a recent report, Bagrodia et al. (16) noted the existence of proline-rich sequences similar to canonical SH3 domain binding regions at the amino terminus of Pak3; similar sequences are also present in other Pak proteins. It was demonstrated that GST fusion proteins containing the SH3 domains of Nck and PLC␥ can bind to Pak3 in an in vitro binding assay. In addition, a serine-threonine kinase activity was shown to be associated with Nck in a variety of cell types (22). The second SH3 domain of Nck was sufficient for association with the kinase activity. This kinase activity did not appear to be modulated by a variety of mitogenic stimuli, and it appeared to be localized exclusively in the particulate fraction. The molecular mass of the Nck-associated kinase was estimated to be ϳ65 kDa. We note that this molecular mass is similar to that of Pak proteins.
In this report we show that Nck and Pak1 are constitutively complexed in living cells, and that their association is mediated through the second SH3 domain of Nck and the first prolinerich sequence of Pak1. In addition, we demonstrate that the Nck/Pak1 complex is recruited to EGF and PDGF receptors upon growth factor stimulation. We also show that Nck modulates Pak1 activity upon EGF stimulation in vivo.

EXPERIMENTAL PROCEDURES
Plasmids-For mammalian expression, cDNA encoding wild-type human Pak1 (21,23) was subcloned into the epitope tag expression vector pJ3M (24), then subcloned into pCMV6. For GST fusion proteins, cDNA constructs encoding individual SH3 domains of Nck and the full-length protein were amplified by polymerase chain reaction (Perkin-Elmer) using oligonucleotides with BamHI/EcoRI linkers to facilitate subcloning into pGEX-2T (Pharmacia Biotech Inc.). The constructs were sequenced on both strands using the Sequenase Quick Denature kit (U. S. Biochemical Corp.).
Antibodies-Rabbit polyclonal antibodies raised in rabbits against a GST fusion protein comprising full-length Nck were used for Nck immunoprecipitation. To facilitate visualization of Nck by SDS-PAGE (since it runs very close to the IgG heavy chain), anti-Nck antibodies were covalently cross-linked to protein A-Sepharose beads (Zymed) using dimethyl pimelimidate (Pierce), following the protocol described in Ref. 25. For Nck immunoblotting, we used either the antibodies described above or affinity-purified polyclonal antibodies (SC-290, Santa Cruz Biotechnology). For Pak1 immunoprecipitation and immunoblotting, affinity purified polyclonal antibodies (SC-881, Santa Cruz Biotechnology) raised against the last 19 amino acids of human or rat Pak1 (12) were used. myc-Pak1 was immunoprecipitated and immunoblotted using a monoclonal anti-myc antibody (9E10, Santa Cruz Biotechnology). For PDGF receptor blotting, we used polyclonal antibodies anti-PR4 (26), which were obtained from Dr. Sara Courtneidge. For detection of tyrosine-phosphorylated proteins, polyclonal anti-phosphotyrosine antibodies previously described in Ref. 27 were used.
Cell Lines and Stimulation with Growth Factors-L6 (rat myoblasts), 293T (human kidney), and HeLa (human cervix epitheloid carcinoma) cells were grown in DMEM (Cellgro) supplemented with 10% FBS (Life Technologies, Inc.), glutamine, and antibiotics. L6 cells were starved in DMEM supplemented with 0.2% FBS before stimulation with PDGF (Intergen) at 30 ng/ml for 5 min, or insulin (Sigma) at 0.1 mM for 5 min. Transiently transfected HeLa cells were starved in DMEM with 0.1% FBS and stimulated with EGF (100 ng/ml) for different times. Cell starvation was for 16 -24 h, and all stimulations were performed at 37°C.
Cell Lysis, Immunoprecipitation, and Immunoblotting-Cells were washed twice in cold phosphate-buffered saline and lysed in lysis buffer containing phosphatase inhibitors as described (27). Cell extracts were precleared by centrifugation. For Nck immunoprecipitation, precleared cell extracts were incubated with anti-Nck antibodies cross-linked to Protein A-Sepharose beads in a nutator at 4°C for 3 h. For Pak1 or myc-Pak1 immunoprecipitation, cell extracts were incubated with 15 g of anti-Pak1 or anti-myc affinity-purified antibodies for 90 min, then recovered on 30 l of protein A-or protein G-Sepharose beads for 90 min at 4°C. Immunocomplexes were washed three times with lysis buffer and eluted in SDS sample buffer. Upon SDS-PAGE (Bio-Rad), gels were transferred to nitrocellulose (MSI), incubated with TBS-5% bovine serum albumin (Intergen) for 2 h at room temperature or overnight at 4°C, followed by incubation with primary antibodies for 1 h at room temperature. Following extensive washes in TBS-0.1% Triton X-100, filters were incubated for 1 h with secondary antibodies (protein A or G conjugated to horseradish peroxidase) in TBS supplemented with 5% non-fat dry milk. Proteins bound to filters were visualized by enhanced chemiluminescence (Renaissance, DuPont NEN).
Transient Transfections-293T cells were transfected by the calcium precipitation method (28) using 10 g of total DNA per 10-cm dish. HeLa cells were transfected with 6 g of DNA per 6-cm dish using LipofectAMINE (Life Technologies, Inc.), following the instructions of the manufacturer. In the experiment shown in Fig. 3, each time point corresponds to an individually transfected dish.
Stable Transfection of L6 Cells-To overexpress human Nck in L6 cells, Nck cDNA was subcloned into the retroviral vector SR␣. Helperfree infectious retrovirus was produced by transient transfection into the retroviral packaging cell line BOSC 23 using calcium phosphate. Nck-containing retroviruses were used to infect L6 cells, and G418resistant pools were selected. Expression of Nck was confirmed by immunoblotting.
Pak Activity Assay-Pak1 immunoprecipitates on beads were washed four times in lysis buffer, followed by two washes in 2 ϫ kinase buffer (23). Kinase reactions were carried out in 30 l of a solution containing 2 g of MBP, 20 M ATP, and 5 Ci of [␥-32 P]ATP (6000 Ci/mmol) in kinase buffer. After incubation for 20 min at 30°C, reactions were stopped by addition of 15 l of 3 ϫ SDS sample buffer followed by boiling the samples for 4 min. Results were visualized by SDS-PAGE and autoradiography.
Interaction Trap Binding Assay-Point mutations in human Pak1 were introduced by a unique site elimination protocol (30). Wild-type and mutant forms of Pak1 were subcloned into the bait vector pEG202 (31). Human Nck and Saccharomyces cerevisiae RNA polymerase subunit RPB4 (32) were each subcloned into the activation domain plasmid pJG4-5 (31). Bait vectors, activation domain vectors, and a lacZ reporter were co-transformed into EGY48, and transformants were selected on dextrose-containing media lacking uracil, leucine, and histidine. Three independent colonies from each transformation were analyzed for reporter activation. The colonies were replica-plated to galactose-containing medium to induce production of the bait protein, and the colonies were assayed for ␤-galactosidase production (33).

RESULTS AND DISCUSSION
We first used parental and Nck-overexpressing L6 cells to determine whether Pak1 and Nck form a complex in vivo.
Lysates from quiescent, PDGF-, or insulin-stimulated cells were prepared and subjected to immunoprecipitation and immunoblotting with anti-Nck or anti-Pak1 antibodies. Pak1 was detected in Nck immunoprecipitates from cells which express endogenous Nck (Fig. 1A) and at a higher level in immunoprecipitates derived from cells that overexpress Nck (Fig. 1B). The amount of Pak1 complexed with Nck does not appear to vary upon cell stimulation with PDGF or insulin (also with EGF, see below). These results argue for the presence of a preformed Pak1/Nck complex in vivo, and that complex formation is limited by the amount of Nck in the cells.
Upon stimulation of cells with EGF or PDGF, Nck is recruited by activated EGF and PDGF receptors by means of its SH2 domain (4 -7). This therefore raises the question of the involvement of Pak1 in the Nck-mediated signal transduction pathway(s) initiated by growth factor stimulation. Serumstarved L6 cells stably overexpressing Nck were PDGF-or mock-stimulated, and lysates were immunoprecipitated with anti-Pak1 or anti-Nck antibodies and probed with anti-PDGF receptor antibodies. PDGF receptor was detected in PDGFstimulated Pak1 immunoprecipitates (Fig. 2A, lane 4), demonstrating that Pak1 is able to associate with activated PDGF receptors. The same blot was probed with anti-phosphotyrosine antibodies (Fig. 2D). Pak1 was not detectably tyrosine-phosphorylated upon PDGF stimulation; Pak1 does not appear to be a substrate for PDGF receptor. However, two constitutively tyrosine-phosphorylated proteins of ϳ90 kDa and ϳ150 kDa, were present in the Nck (pp90 and pp150) and Pak1 (pp90) immunoprecipitates. Tyrosine-phosphorylated proteins of the same apparent molecular mass were observed in Nck and Pak1 immunoprecipitates from A431 and HER14 (NIH 3T3 overexpressing human EGF receptors) cells (data not shown). The identity of pp90 and pp150 is currently unknown.
The results discussed above suggest a role for Pak1 in a PDGF receptor signal transduction pathway. Since EGF also triggers association between activated EGF receptor and Nck (5-7), we repeated this experiment with A431 and HER14 cells to examine the possible involvement of Pak1 in EGF receptor signaling. The results were essentially the same as in Fig. 2, with Pak1 detected in anti-Nck immunoprecipitates (stimulated or nonstimulated) and activated EGF receptor detected in Nck and Pak1 immunoprecipitates. 2 Whether or not the recruitment of Pak1 to activated growth factor receptors was associated with activation of Pak1 kinase activity was explored in the next set of experiments. We tested whether EGF activates Pak1 in vivo. HeLa cells were transiently transfected with a myc-tagged human Pak1 expression 2 M. L. Galisteo and J. Schlessinger, unpublished results.
FIG. 1. Nck co-immunoprecipitates with the serine-threonine kinase Pak1. A, lysates from mock (Ϫ), PDGF-or insulin-stimulated L6 cells were immunoprecipitated with anti-Nck or anti-Pak1 antibodies. The co-precipitating proteins were resolved by SDS-PAGE and transferred to a nitrocellulose membrane. The membrane was divided in two parts. The lower part was probed with anti-Nck antibodies, and the upper part was probed with anti-Pak1 antibodies. B, as in A, but the lysates are from an L6 clone which stably overexpresses human Nck.
Nck Recruits Pak1 to Activated RTKs 20998 vector, human Nck, or both expression vectors. Cells were stimulated with EGF for different times, Pak1 was immunoprecipitated from cell lysates with anti-myc antibody, and in vitro kinase assays using MBP as an exogenous substrate were performed. The experiment presented in Fig. 3A shows that Pak1 activity was increased as a function of time of stimulation with EGF. The rate of stimulation of Pak1 activity lagged the rate of EGF receptor autophosphorylation (Fig. 3C), presumably reflecting the time required for recruitment of the Nck-Pak1 complex and activation of Pak1. Pak1 activation upon EGF stimulation was much more pronounced when Nck was co-transfected with Pak1 (Fig. 3A, compare lanes 1-4 with 5-8). To verify that the activity measured in Fig. 3 was due to Pak1, we repeated the experiment with a kinase-deficient Pak1 mutant (21). In this case, no increase in kinase activity was detected. 2 In line with the Nck immunoprecipitation data, these kinase data again suggest that Nck is limiting and that overexpression of Nck brings additional Pak1 molecules into an activation-competent complex. Thus, Pak1 activation is triggered by EGF stimulation, and Nck appears to modulate the magnitude of this increase in vivo.
To determine the region of Nck responsible for binding to Pak1, we performed in vitro binding experiments using GSTfusion proteins, each containing either one of the three individual SH3 domains of Nck or full-length protein. 293T cells were transiently transfected with a myc-tagged human Pak1 construct for mammalian expression. Transfected cells were lysed and extracts were incubated with the different fusion proteins coupled to Sepharose beads. Immunoblot analysis using anti-Pak1 antibodies (Fig. 4) showed that full-length Nck associated with Pak1, as well as the second SH3 domain. These results demonstrate that the interaction between the two proteins is mediated by the second SH3 domain of Nck.
We next determined the region in Pak1 responsible for binding to Nck. Human Pak1 contains three proline-rich motifs that fit the consensus for SH3 domain binding. The most amino-terminal of these (KPPAPPMR) is highly conserved in all three human Pak proteins and is also present in the Drosophila homologue of Pak1. In order to determine which, if any, of these motifs mediate binding to Nck, we carried out an interactiontrap analysis in S. cerevisiae, using as bait wild-type Pak1 or point mutants in each of the three proline-rich regions, respectively (Fig. 5A). The interaction between Pak1 and Nck is easily detectable in this system, resulting in the strong activation of the reporter ␤-galactosidase (Fig. 5B). Replacement of proline 13 (located in the first proline-rich domain) with alanine completely abolishes binding to Nck, whereas similar mutations within the second and third proline-rich regions are without effect. Pak1 does not detectably interact with an irrelevant partner, RPB4 (32), indicating that the interaction with Nck is specific.
It was recently demonstrated that a GST-fusion protein containing the three SH3 domains of Nck was able to bind to Pak3 in an in vitro "pull-down" assay, as well as to a GST protein containing the SH3 domain of PLC␥ (16). The second SH3   (7). We have examined the possibility that Pak1 binds to PLC␥ in vivo. We were unable to detect association between PLC␥ and Pak1 by co-immunoprecipitation experiments in either unstimulated or EGF-stimulated HER14 cells (data not shown).
In this report we demonstrate that Nck and Pak1 form a complex in living cells. Nck is constitutively bound by means of its second SH3 domain to a proline-rich region in Pak1. Cell fractionation experiments demonstrated that the Pak1 protein that co-immunoprecipitated with Nck is exclusively associated with the particulate fraction. 2 Thus, the Pak1 protein may represent the serine-threonine kinase activity that was shown to be associated with Nck (22). In addition, we have shown that, upon growth factor stimulation, an Nck-Pak1 complex is recruited to activated EGF or PDGF receptors. We have demonstrated that EGF stimulation leads to Pak1 activation in vivo. It is possible that the physiological targets of Pak1 are localized in the plasma membrane, and Nck may function as a mediator that recruits Pak1 to specific sites on the cell surface by virtue of its ability to bind to tyrosine-phosphorylated receptors through its SH2 domain.
Pak proteins have been shown to bind to and become activated by Cdc42 and Rac1 (12). The roles of Rho family members in inducing changes in cell morphology and motility (34,35), and in activating the JNK and p38 signaling pathways (18,19), have been well established. Pak proteins have been shown to control the JNK (19,20) and p38 MAP kinase signaling pathways (21). However, involvement of Pak proteins in regulation of the cytoskeleton has not been reported.
WASP has been shown to be a mediator between Cdc42 and the actin cytoskeleton (35,36), and it has been reported recently to complex with Nck in hematopoietic cells (10). These results are consistent with the proposal that the Drosophila homologue of Nck (dock), may link tyrosine kinase receptor activation with intracellular signaling pathways that regulate cytoskeletal changes in the growth cone (11). By analogy, the Nck adaptor protein may act as a link between tyrosine kinase receptors and changes in the actin cytoskeleton in mammalian cells, and its function may be modulated by the action of Rholike GTPases that are present at the plasma membrane. FIG. 5. Mapping the sequence in Pak1 responsible for binding to Nck. A, schematic map of human Pak1 and point mutations introduced at different proline-rich stretches. PBD, p21 binding domain. B, LexA fusion protein baits in pEG202 were co-transformed with the indicated activation domain fusion vectors, plus the lacZ reporter pSH18 -34 into S. cerevisiae EGY48. Protein expression was induced by growing the transformants on galactose-containing medium and confirmed by immunoblot with anti-LexA antibodies. Interactions were detected by assaying for ␤-galactosidase production, as indicated by formation of blue cells. Three independent colonies were assayed for each construct.