SH2/SH3 Adaptor Proteins Can Link Tyrosine Kinases to a Ste20-related Protein Kinase, HPK1*

Ste20-related protein kinases have been implicated as regulating a range of cellular responses, including stress-activated protein kinase pathways and the control of cytoskeletal architecture. An important issue involves the identities of the upstream signals and regulators that might control the biological functions of mammalian Ste20-related protein kinases. HPK1 is a protein-serine/threonine kinase that possesses a Ste20-like kinase domain, and in transfected cells activates a protein kinase pathway leading to the stress-activated protein kinase SAPK/JNK. Here we have investigated candidate upstream regulators that might interact with HPK1. HPK1 possesses an N-terminal catalytic domain and an extended C-terminal tail with four proline-rich motifs. The SH3 domains of Grb2 bound in vitro to specific proline-rich motifs in the HPK1 tail and functioned synergistically to direct the stable binding of Grb2 to HPK1 in transfected Cos1 cells. Epidermal growth factor (EGF) stimulation did not affect the binding of Grb2 to HPK1 but induced recruitment of the Grb2·HPK1 complex to the autophosphorylated EGF receptor and to the Shc docking protein. Several activated receptor and cytoplasmic tyrosine kinases, including the EGF receptor, stimulated the tyrosine phosphorylation of the HPK1 serine/threonine kinase. These results suggest that HPK1, a mammalian Ste20-related protein-serine/threonine kinase, can potentially associate with protein-tyrosine kinases through interactions mediated by SH2/SH3 adaptors such as Grb2. Such interaction may provide a possible mechanism for cross-talk between distinct biochemical pathways following the activation of tyrosine kinases.

In the budding yeast, Saccharomyces cerevisiae, the Ste20 serine/threonine protein kinase regulates a mitogen-activated protein kinase (MAPK) 1 pathway, composed of the Ste11 (MEK kinase), Ste7 (MEK), and Fus3/Kss1 (MAPK) protein kinases (1). Ste20 controls the mating response of haploid cells and is positioned by genetic epistasis analysis between heterotrimeric G proteins activated by the ␣ factor mating pheromone and the MAPK cassette (2). In addition to its role in the mating pathway, yeast Ste20 participates in the development of nitrogendeprived diploid cells into a filamentous pseudohyphal form and in the ability of haploid cells to invade an agar substrate (3,4). Ste20 specifically associates through an N-terminal regulatory sequence with the small GTPase Cdc42 (2), an interaction that results in the localization of cytosolic Ste20 to specific membrane sites. Of interest, the interaction of Ste20 with Cdc42 is essential for cell invasion and pseudohyphal formation, as well as for an effect of Ste20 in stabilizing septincontaining microfilaments during cytokinesis, but is dispensable for the mating response (3). Thus Ste20 appears to have pleiotropic functions in the yeast cell and to have both multiple upstream activators and downstream biological outputs.
Recent work has identified a number of mammalian kinases whose catalytic domains are closely related in sequence to that of yeast Ste20 (5). Among these, members of the PAK family share the same overall structure as Ste20, consisting of a C-terminal kinase domain, and an N-terminal regulatory region that contains binding motifs for the small GTPases Cdc42 and Rac1 (6), as well as for the second of the three SH3 domains of the adaptor protein Nck. Interactions with these signaling proteins may both relocalize PAK1 and PAK3 to the membrane and stimulate their kinase activity (7,8). PAK kinases have been implicated in the regulation of a MAPK cassette that responds to stressful stimuli and results in activation of the stress-activated protein kinases (SAPK/JNK) and p38 (5,6,9). PAK1 may also be involved in the control of actin stress fiber disassembly and in reorganization of focal complexes (10,11).
A distinct family of Ste20-related mammalian protein kinases including GC kinase, HPK1, KHS, NIK, and MST1 and -2 consists of an N-terminally located kinase domain and a C-terminal tail, which is proposed to have a regulatory function (6,(12)(13)(14)(15)(16)(17). Like PAK kinases, GC kinase, HPK1, NIK, and KHS have the capacity to activate the SAPK pathway (12)(13)(14)17), although they may well have additional biological functions. The SAPKs, as well as the related kinases of the p38/ CSBP/RK family are of considerable interest, as they are activated in response to numerous stimuli, including inflammatory cytokines, ischemia, ultraviolet and ionizing radiation, DNA-damaging chemotherapeutic drugs, and heat shock (5). Although the physiological consequences of SAPK and p38 activation are not well understood, they have been linked to growth arrest and the induction of apoptosis, compatible with a role for stress pathways in inducing cellular repair or in triggering cell death (18 -21). However, tumor necrosis factor ␣-induced apoptosis was shown to be independent of tumor necrosis factor ␣-induced SAPK/JNK activation (22)(23).
HPK1 is a serine/threonine protein kinase that is widely expressed during embryonic development and becomes restricted to hematopoietic organs in the adult (13,14). HPK1 interacts with a downstream kinase, MLK3/SPRK1, which has an N-terminal SH3 domain. MLK3, in turn, can stimulate SEK (MKK4), thereby leading to activation of SAPK (14,24,25). Here we have investigated potential mechanisms through which cell surface receptors might be linked to HPK1. HPK1 contains four proline-rich motifs in its C-terminal tail, suggesting that it might associate with cytoplasmic signaling proteins that possess SH3 domains. Here we show that HPK1 binds SH2/SH3 adaptor proteins, notably Grb2, in vivo. In a transfection model, Grb2 bound constitutively through its SH3 domains to HPK1. Upon EGF stimulation the Grb2⅐HPK1 complex was recruited to the autophosphorylated EGF receptor and to the Shc docking protein, and HPK1 was detectably phosphorylated on tyrosine. These and related data suggest that SH2/SH3 adaptors such as Grb2 can link tyrosine kinases to HPK1, a Ste20-related serine/threonine protein kinase. Grb2, which couples tyrosine kinase to the Ras GTPase, may thereby provide a molecular bridge between different biochemical pathways and consequently a mechanism for interpathway cross-talk.

Recombinant Plasmids
Yeast Expression Vectors-Varying cDNA fragments encoding the proline-rich region of HPK1 as indicated in Fig. 2, top, were cloned into pACTII (53). SH3 domains from different sources were cloned into the pASI vector as described previously (50).
Mammalian Expression Vectors-HPK1 wild type and kinase dead expression vectors were described previously (14). pCDNA3-HA-tagged vectors of the wild type and mutated forms of human HA-Grb2 cDNA were subcloned into a modified pCDNA3 expression vector (Invitrogen Inc.) from the pEBB plasmid. pEBB plasmids containing Grb2 cDNA with specific mutations (31) were a generous gift from Dr. Bruce J. Mayer. v-Fps (p130 gag-fps ) expression vector was described previously (54).
Yeast Two-hybrid Assays pASI vectors, encoding fusions between the DNA binding domain of Gal4 and the different SH3 domains, and pACTII vectors, encoding the transcriptional activation domain of Gal4 fused to different fragments of the HPK1 proline-rich region, were transformed into S. cerevisiae strains Y153 and Y187, respectively. Yeast transformation was performed by the lithium acetate method (55) except that 10% dimethyl sulfoxide (Me 2 SO) was included during 42°C heat shock. Co-expression of pASI and pACTII vectors was achieved by mating Y153 and Y187 yeast strains, each containing the appropriate vector. ␤-Galactosidase activity was detected on 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside plates after permeabilizing the yeast by liquid nitrogen treatment.

Synthesis of Proline-rich Peptides and Inhibition Analysis Using Surface Plasmon Resonance Detection
Peptides were synthesized on an Applied Biosystems Inc. ABI 431A synthesizer using standard 9-fluorenyl methoxycarbonyl (Fmoc) chemistry with 2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate activation on p-alkoxybenzyl alcohol (HMP) resin as outlined by the manufacturer. Cleavage of the peptide from the resin and deprotection was achieved through a 90-min incubation at room temperature in trifluoroacetic acid (10 ml) and a scavenger mixture composed of water/thioanisole/1,2-ethanedithiol (5.0:0.2:0.1% by volume) and solid phenol (75 mg). The product was precipitated with cold t-butyl ethyl ether, collected by centrifugation, and purified using reverse phase high pressure liquid chromatography. The authenticity of the phosphopeptides was confirmed by amino acid analysis and mass spectrometry.
Surface plasmon resonance (SPR) analysis carried out to quantitate peptide binding to Grb2 in a competition assay was performed using a BIAcore apparatus (Pharmacia Biosensor). A GST fusion protein containing residues 291-518 of HPK1 resuspended in 50 mM sodium acetate, 150 mM NaCl, pH 4.0, was immobilized onto the surface of a CM5 sensor chip surface following conditions described previously for immobilization of mSos1 (56). Solutions (50 l) of purified baculovirus-expressed Grb2 (5 M) (a generous gift of Dr. Terance Kubiseski) in 50 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 2 mM DTT, pH 7.5, and a quantity of synthetic peptide were injected across the surface containing the immobilized GST-HPK1 surface, and the SPR signal was monitored. The amount of Grb2 bound was estimated from the steady-state surface plasmon resonance signal at a fixed time during the injection and the percentage bound relative to injection of Grb2 alone calculated. After each injection the GST-HPK1 surface was regenerated using 0.1 M NaOH and 1 M NaCl.

Antibody Preparation
Anti-HPK1 N-terminal antibodies-(1-18) were raised against the peptide MALYDPDIFNKDPREHYD, corresponding to the first 18 amino acids of HPK1. 1-18 antibodies did not interfere with HPK1-Grb2 complex formation and were used for co-immunoprecipitation experiments described in Figs. 4 -6.
The production of anti-Nck antibodies was described previously (57).

Preparation and Purification of GST Fusion Proteins
Bacterial cultures were induced with 1 mM isopropylthiogalactopyranoside. The cell pellets were homogenized by sonication in bacterial lysis buffer (phosphate-buffered saline containing 1% Triton X-100, 1% Tween 20, 2 mM DTT, 10 g/ml aprotinin, 10 g/ml leupeptin, 2 mM phenylmethylsulfonyl fluoride). The lysates were subjected to centrifugation at 28,000 ϫ g for 30 min. The supernatants were then mixed with glutathione-agarose (Sigma), and the beads were washed 4 times with the bacterial lysis buffer, stored at 4°C in phosphate-buffered saline containing 0.05% sodium azide and used for in vitro binding assays.

In Vitro Binding Assays
Cell lysates from HPK1-transfected cells were incubated with approximately 1 g of GST fusion protein coupled to glutathione-agarose (Sigma) in lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl 2 , 10% glycerol, 1% Triton X-100, 1.5 mM MgCl 2 , 1 mM EGTA, 10 mM sodium pyrophosphate, 100 mM sodium fluoride, 200 M sodium orthovanadate, 10 g/ml aprotinin, 15 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride). The beads were washed three times and treated at 95°C for 5 min with SDS sample buffer. Proteins were separated by SDS-polyacrylamide gel electrophoresis and transferred to a nitrocellulose filter. Associated HPK1 was detected by Western blot analysis with anti-HPK1 antisera.

In Vivo Complex Formation
Cos1 cells were transiently transfected with wild type or mutated forms of the HA-Grb2 pCDNA3 expression vector alone or in combination with an HPK1 expression vector. After 48 h of growth in Dulbecco's modified Eagle's medium (high glucose) supplemented with 10% fetal bovine serum, the cells were placed in serum-free Dulbecco's modified Eagle's medium contain 20 mM HEPES, pH 7.5, for 18 h prior to EGF stimulation. After treatment, the cells were rinsed three times in icecold phosphate-buffered saline and lysed with lysis buffer A (50 mM Tris, pH 7.5, 120 mM NaCl 2 , 1% Nonidet P-40, 5 mM DTT, 10 mM sodium pyrophosphate, 25 mM sodium fluoride, 200 M sodium orthovanadate, 10 g/ml aprotinin, 5 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride). HPK1 was immunoprecipitated by rabbit anti-HPK1 1-18 antibodies, and the complexes were subjected to Western blot analysis as indicated under "Results."

Tyrosine Phosphorylation of HPK1
Cos1 cells were transiently transfected with HPK1 expression vector alone or in combination with expression vectors for tyrosine kinases, as indicated. The cells were starved in serum-free medium prior to EGF and PDGF stimulation. HPK1 was immunoprecipitated using specific anti-HPK1 antibodies directed to the proline-rich region of HPK1. Tyrosine phosphorylation of HPK1 was analyzed by blotting the immunoprecipitated protein with affinity purified rabbit anti-phosphotyrosine antibodies or with monoclonal anti-phosphotyrosine antibodies (4G10 from UBI Inc.).

Multiple Proline-rich Motifs in the HPK1 C-terminal Region
Bind SH3 Domains-A cluster of four proline-rich sequences (P1-P4) are located in the C-terminal tail of HPK1, in the region following the kinase domain (Fig. 1A). Three of these proline-rich elements (P1, P2, and P4) display high sequence homology with motifs located in the tail of mSos1 that bind Grb2 SH3 domains ( Fig. 1B) (26 -28). These HPK1 motifs conform to the sequence consensus for SH3 binding sites in the class II orientation (Fig. 1B) (29,30). Structural analysis has shown that such elements adopt a left-handed polyproline type II helix, in which two pairs of aliphatic-proline residues occupy relatively conserved hydrophobic pockets on the SH3 domain, while the P Ϫ3 residue binds a more variable specificity pocket. The SH3 domains of adaptors such as Grb2 preferentially recognize a basic residue at the P Ϫ3 position, as found in the HPK1 proline-rich motifs. We initially used the yeast two-hybrid system to analyze the ability of the proline-rich regions of HPK1 to bind specific SH3 domains. A fragment of HPK1 encompassing all four proline-rich sites (construct 1234, Fig. 2) bound to both N-terminal and C-terminal SH3 domains of Grb2, to a polypeptide containing the three Nck SH3 domains, to the SH3 domain of the MLK3/SPRK1 serine kinase, and more weakly to the Abl SH3 domain. A more detailed analysis of these interactions using shorter HPK1 polypeptides with one, two, or three of the proline-rich motifs showed some differences in binding specificity. For example, the Grb2 SH3-C domain associated with a polypeptide containing only the P2 site, whereas the Grb2 SH3-N domain did not. The Grb2 SH3-N domain showed an interaction with P1, whereas both Grb2 SH3 domains bound a polypeptide containing sites P3 and P4. The ability of the individual Grb2 SH3 domains to bind HPK1 with distinct specificities suggests that both SH3 domains of Grb2 might simultaneously recognize HPK1. Furthermore, in the yeast twohybrid assay, the three Nck SH3 domains interacted with all of the HPK1 proline-rich polypeptides, indicating that Nck might also recognize multiple HPK1 proline motifs.
We have shown previously (14) that HPK1 binds MLK3, which in turn can activate SEK and thereby stimulate SAPK. The MLK3 SH3 domain only interacted with HPK1 polypeptides containing the P3 and P4 sites.
To test these interactions directly, immobilized GST fusion proteins containing specific SH3 domains were incubated with cell lysates from HPK1-transfected cells. The GST fusion proteins and any associated polypeptides were then harvested using glutathione-agarose beads and probed for the presence of HPK1 using Western blot analysis (Fig. 3A). The individual Grb2 SH3 domains each formed a complex with HPK1, although in this assay the N-terminal SH3 domain interacted more strongly. A GST fusion protein containing full-length Grb2 bound HPK1 more tightly than either SH3 domain alone, indicating that the two SH3 domains of Grb2 may bind HPK1 in a bidentate fashion, as suggested by the two-hybrid analysis. A GST fusion containing the three Nck SH3 domains also FIG. 1. A, schematic diagram of HPK1. The relative locations of the HPK1 kinase domain and the proline-rich motifs are indicated. The sequence of HPK1 residues 291-490 is shown, with the proline-rich motifs underlined. B, sequence similarity between proline-rich regions of HPK1 and SH3 domain binding sites of mSos1/mSos2. The sequences of HPK1 proline-rich motifs P1, P2, and P4 are compared with sequences in the C-terminal tails of mouse Sos1 and -2. The P3 motif of HPK1 does not closely match sequences in mSos1 or mSos2. ⌽, hydrophobic residue.
FIG. 2. Specific proline-rich regions of HPK1 interact with SH3 domains. Specific fragments of the HPK1 proline-rich region were tested for their interaction with indicated SH3 domains in the yeast two-hybrid system. The relative level of ␤-galactosidase activity identified by staining the yeast with 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside is indicated. ϩϩϩϩ, relatively very strong interaction (appearance of blue color in less than 20 min). ϩϩϩ, relatively strong interaction (appearance of blue color in less than 1 h). ϩϩ, intermediate interaction (blue color in less than 3 h). ϩ, weak interaction (appearance of blue color in less than 6 h). Ϫ, no detectable binding (no trace of blue color by 20 h). Abl W99R is a mutant of the Abl SH3 domain that lacks specific binding to proline-rich sequences (45). The mSos1 tail was used as a positive control for interactions with Grb2 and Nck SH3 domains.
bound HPK1, although more weakly than GST-Grb2. Neither GST alone, nor fusion proteins containing the Abl or phospholipase C-␥1 SH3 domains, bound detectable amounts of HPK1 in vitro, suggesting that the in vitro interactions of HPK1 with the Grb2 and Nck SH3 domains are relatively specific.
To probe further the regions of HPK1 that might interact with Grb2, we prepared a series of peptides corresponding to the proline-rich regions of HPK1 and tested their ability to inhibit Grb2 binding to the tail of HPK1 as monitored using surface plasmon resonance (SPR) technology. In this SPR assay a purified GST-HPK1 fusion protein containing amino acids 291-518 was immobilized to a Biosensor chip surface (BIAcore, Pharmacia Biosensor), and the signal obtained following injection of purified full-length Grb2 in the absence and presence of peptide was observed. As shown in Fig. 3B, peptide P2 (NIPP-PLPPKPKFR) displayed the strongest inhibition (IC 50 ϭ 24 M) suggesting that this region may be a primary site for Grb2 binding on HPK1. Peptides P1 (EPEPPPAIPRRIR) and P4 (PGQPPLVPPRKEK) also competed for binding, yielding IC 50 values of 90 and 120 M, respectively. Only peptide P3 (PP-PRTPRPGPPPA) did not inhibit Grb2 binding to HPK1. In addition, we have found that a proline-rich peptide derived from the Drosophila Sos tail (amino acids YRAVPPPLPPRR) inhibits the Grb2/GST-HPK1 interaction to a similar extent as the HPK1 peptide P2 (data not shown). These data are rela-tively consistent with previously reported results for the binding of Grb2 to proline-rich peptides from mSos1 (Fig. 1B) (26,28) and suggest that Grb2 may bind in a similar fashion to both HPK1 and mSos.
By having demonstrated the formation of an HPK1⅐Grb2 complex in vitro and identified specific proline-rich motifs in HPK1 that can mediate this interaction through recognition of Grb2 SH3 domains, we investigated whether such a complex can form in vivo. For this purpose an HA-tagged version of wild type Grb2 was transiently expressed in Cos1 cells in combination with full-length HPK1. After immunoprecipitation of HPK1 with anti-HPK1-(1-18) antiserum, we probed for the presence of co-precipitated HA-tagged Grb2 by Western blot analysis (Fig. 4). No HA-Grb2 was precipitated with anti-HPK1 serum in the absence of co-transfected HPK1. However, coexpression of both HPK1 and HA-Grb2 resulted in the formation of a readily detectable complex, indicating that HPK1 and Grb2 can associate in mammalian cells. The role of the Grb2 SH3 domains in binding to HPK1 in Cos1 cells was tested using HA-Grb2 SH3 mutants in which an invariant Trp, essential for SH3 domain binding activity, was substituted with Lys in either the N-terminal or C-terminal SH3 domains, or in both SH3 domains together (31). Inactivation of either Grb2 SH3 domain resulted in a striking reduction in the HPK1⅐Grb2 complex, and mutation of both SH3 domains abrogated complex formation. These results suggest that the SH3 domains of Grb2 bind synergistically to HPK1 in mammalian cells.
At present the extracellular signals controlling HPK1 activity are unknown. When transfected into Cos1 cells HPK1, like several other Ste20-related kinases, is constitutively active as a potent stimulator of the SAPK pathway (13,14). This finding has hindered the analysis of signals that might regulate HPK1. However, the interaction of HPK1 with the SH3 domains of Grb2 and Nck suggested that HPK1 might be modulated by phosphotyrosine-containing proteins, such as activated growth factor receptors or the Shc adaptor, that bind the Grb2 SH2

FIG. 3. Association between HPK1 and Grb2 in vitro.
A, GST fusion proteins containing the indicated SH3 domains or full-length Grb2, or GST alone, were incubated with lysates from NIH 3T3 cells stably transfected with an HPK1 expression vector. After washing, proteins were separated by SDS-polyacrylamide gel electrophoresis, and associated HPK1 was visualized by immunoblotting with anti-HPK1 antibodies. As a control HPK1 was immunoprecipitated with anti-HPK1 1-18 antibodies (␣HPK1). B, peptide inhibition of Grb2 binding to HPK1. Surface plasmon resonance was used to monitor the ability of proline-rich peptides derived from HPK1 to inhibit binding of Grb2 to an immobilized GST-HPK1 fusion protein containing residues 291-518. Solutions (50 l) of purified baculovirus-expressed Grb2 (5 M) in 50 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 2 mM DTT, pH 7.5, containing the designated quantity of synthetic peptides were injected across a sensor chip surface containing immobilized GST-HPK1-(291-518), and the SPR signal was monitored using a BIAcore instrument (Pharmacia Biosensor). The amount of Grb2 bound was estimated from the steady-state SPR signal at a fixed time during the injection, and the percentage bound relative to injection of Grb2 alone was calculated. The signal obtained in the absence of competing peptide was assigned a value of 100%. domain through Tyr(P)-X-Asn-X motifs (32,33). To test this possibility we have used the EGF receptor as a model to study the possible impact of receptor tyrosine kinase activation on HPK1.
Grb2 Links HPK1 to the Activated EGF Receptor and Shc-Upon EGF stimulation of fibroblastic cells, a preexisting complex containing Grb2 and mSos1 is recruited to the autophosphorylated EGF receptor (26,27) through recognition of the activated receptor by the Grb2 SH2 domain. Since HPK1 contains C-terminal proline-rich motifs closely related to those found in the tail of mSos1, and like mSos1 binds constitutively to Grb2, we asked whether EGF stimulation would induce binding of the HPK1⅐Grb2 complex to the autophosphorylated EGF receptor. EGF stimulation of Cos1 cells co-expressing HPK1 and HA-tagged Grb2 did not affect the binding of HPK1 to Grb2 (Fig. 5A). However, following stimulation of these cells with EGF, both HA-Grb2 and HPK1 inducibly associated with the EGF receptor as judged by co-precipitation of both HPK1 and HA-Grb2 with the activated receptor (Fig. 5B). This EGFdependent association between HPK1 and the EGF receptor was also detected after immunoprecipitation of the EGF receptor and Western blot analysis with anti-HPK1 antibodies (Fig.  5C). The interaction was not dependent on HPK1 kinase activity, since a kinase-deficient HPK1 mutant (K46E, Ref. 14) complexed with the EGF receptor equally well as wild type HPK1 (Fig. 5C).
To investigate whether Grb2 might mediate the association of HPK1 with the EGF receptor, we analyzed the effect of overexpression of Grb2 on formation of the HPK1⅐EGF receptor complex. The amount of the EGF receptor that co-precipitated with HPK1 was significantly enhanced in Cos1 cells transfected with HA-Grb2 and therefore overexpressing Grb2, as compared with cells containing only endogenous Grb2 (Fig. 6,  top panel), suggesting that Grb2 can mediate the interaction between the EGF receptor and HPK1.
In addition to a direct interaction with the EGF receptor, Grb2 binds through its SH2 domain to other tyrosine-phosphorylated proteins in EGF-stimulated cells, notably the Shc adaptor (34,35). Consistent with the notion that Grb2 might bridge on interaction between phosphotyrosine-containing proteins and HPK1, EGF stimulation induced the formation of a complex containing Shc proteins and HPK1, which was further enhanced by overexpression of Grb2 (Fig. 6). These data suggested that EGF stimulates the formation of a phospho-Shc⅐Grb2⅐HPK1 ternary complex.
Data obtained from the yeast two-hybrid analysis (Fig. 2) as well as in vitro binding experiments with GST-SH3 fusion proteins (Fig. 3A) suggested that Nck adaptor might also have the capacity to interact with HPK1 in vivo. Western blots of anti-HPK1 immunoprecipitates from unstimulated or EGF-stimulated Cos1 cells that had been transfected with HPK1 were therefore probed with antibodies against Nck. Nck was found to co-precipitate with HPK1 from transfected Cos1 cells. However, unlike the Grb2⅐HPK1 complex, the interaction between Nck and HPK1 was apparently stimulated by EGF. In addition, overexpression of Grb2 significantly decreased the interaction between Nck and HPK1, suggesting that these two SH2/SH3 adaptors can compete in vivo for HPK1 binding.
HPK1 Is Tyrosine-phosphorylated in Response to EGF, PDGF, and in v-Src-or v-Fps-transfected Cells-The recruitment of the HPK1⅐Grb2 complex to the activated EGF receptor might potentially result in phosphorylation of HPK1 on tyrosine. To test this possibility, HPK1 was precipitated from Cos1 cells that had been transfected with HPK1 and HA-Grb2 and subsequently stimulated with EGF. Tyrosine phosphorylation of HPK1 was then assayed by Western blotting with anti-phosphotyrosine antibodies (Fig. 7). A tyrosine-phosphorylated protein corresponding in size to HPK1 was specifically detected in anti-HPK1 immunoprecipitates from cells that had been transfected with HPK1. No similar phosphotyrosine signal was detected in unstimulated transfected cells or in EGF-stimulated cells that had not been transfected with HPK1 and stimulated with EGF. The phosphotyrosine signal was strongly enhanced if the HPK1-transfected cells were pretreated with FIG. 5. HPK1 associates with the activated EGF receptor. A, Cos1 cells were transfected with wild type or mutant (W33,193K) HAtagged Grb2 together with HPK1. Cells were serum-deprived or stimulated with EGF as indicated. Cell lysates were immunoprecipitated with anti-HPK1-(1-18) antibodies and blotted with anti-HA antibodies to detect the HA-Grb2⅐HPK1 complex (top panel). Whole cell lysates were blotted with anti-HA to detect HA-Grb2 expression (middle panel) or blotted with anti-Tyr(P) (anti pTyr) antibodies to detect autophosphorylated EGF receptor (bottom panel). The two right-hand lanes represent untransfected controls. EGF stimulation did not affect formation of the HA-Grb2⅐HPK1 complex. B, Cos1 cells were co-transfected with HA-Grb2 and HPK1 and either serum-deprived (Ϫ) or stimulated with EGF (ϩ). Cell lysates were immunoprecipitated with antibodies to the EGF receptor (␣EGFR), HA (␣HA), or HPK1 (␣HPK1), and the immunoprecipitates were blotted with antibodies to the EGF receptor. EGF stimulation induced the association of both HA-Grb2 and HPK1 with the EGF receptor. C, Cos1 cells were co-transfected with HA-Grb2 and either wild type (wt) or kinase-deficient (kd, K46E) HPK1 and either serum-deprived (Ϫ) or stimulated with EGF (ϩ). Cell lysates were immunoprecipitated (IP) with anti-EGF receptor antibodies and blotted with antibodies to HPK1. Total cell lysates (TCL) were also examined by anti-HPK1 blotting for HPK1 expression. the tyrosine phosphatase inhibitor sodium orthovanadate prior to EGF induction. These results strongly suggest that HPK1 can be tyrosine-phosphorylated in EGF-stimulated cells and that this modification is regulated by a tyrosine phosphatase activity in vivo.
To test whether tyrosine phosphorylation of HPK1 could be mediated by additional growth factor receptors, HPK1 was co-expressed with the ␤ receptor for platelet-derived growth factor (PDGF). PDGF stimulation of these cells resulted in HPK1 tyrosine phosphorylation (Fig. 8A). In addition, we found that HPK1 was tyrosine-phosphorylated in cells that had been co-transfected with v-Src or with v-Fps, activated forms of cytoplasmic tyrosine kinases (36) whose normal counterparts are expressed in hematopoietic cells (37, 38) (Fig. 8, B and C). This result suggests that HPK1 is potentially phosphorylated by a variety of tyrosine kinases. DISCUSSION We have investigated possible mechanisms through which the Ste20-related protein kinase HPK1 might be coupled to cell surface receptors. Our results have indicated that proline-rich motifs in the tail of HPK1 can bind in vivo to the SH3 domains of the adaptor proteins Grb2 and Nck. Using a well-documented model system, we have established that Grb2 can physically couple HPK1 to the activated EGF receptor. These data establish a basis from which to explore the molecular regulation of HPK1 in hematopoietic cells and suggest that adaptor proteins such as Grb2 may bridge distinct biochemical signaling pathways involving protein serine/threonine kinases.
Grb2 Binds HPK1, a Ste20-related Protein Kinase, in Vivo-SH2/SH3 adaptor proteins such as Grb2, Crk, and Nck physically couple activated growth factor receptors and phosphorylated cytoplasmic docking proteins to downstream targets with proline-rich motifs. A combination of genetic and biochemical evidence has indicated that an important biological function of FIG. 7. HPK1 is phosphorylated on tyrosine in EGF-stimulated cells. Cos1 cells were transiently transfected with HA-Grb2 alone or in combination with HPK1. The cells were incubated in the presence or absence of sodium vanadate for 90 min, followed by EGF stimulation as indicated. HPK1 was precipitated with antibodies directed to the HPK1 proline-rich region. Tyrosine phosphorylation of HPK1 was analyzed by blotting the immunoprecipitations (IP) with affinity purified rabbit anti-phosphotyrosine (anti pTyr) antibodies. The same filter was stripped and reblotted with anti-HPK1 antibodies to determine the amount of HPK1 in each lane.

FIG. 8. Multiple tyrosine kinases induce HPK1 phosphorylation in transfected cells.
A, Cos1 cells were transiently co-transfected with 1 g of pMT2-HPK1 alone or in combination with 160 or 480 ng of pMT-PDGFR. Cells were placed in serum-free Dulbecco's modified Eagle's medium for 18 h prior to PDGF stimulation. Cell lysates were immunoprecipitated with anti-HPK1 antibodies and blotted with antibodies for phosphotyrosine (pTyr) (4G10) or for HPK1. B, Cos1 cells were transfected with HPK1, HA-Grb2, and v-Src expression vectors alone or in combination as indicated. Tyrosine phosphorylation of HPK1 was analyzed by anti-HPK1 immunoprecipitation followed by antiphosphotyrosine Western blot. C, 1 g of wild type (wt) or kinase dead (kd) HPK1 was transfected into Cos1 cells alone or with combination with a v-Fps expression vector. Tyrosine phosphorylation of HPK1 was analyzed as described above. mammalian Grb2 and its invertebrate homologues is to link tyrosine kinases to Sos guanine nucleotide exchange factors (26 -28) and thereby to activate the Ras GTPase and a downstream MAPK cassette composed of the Raf, Mek, and extracellular signal-regulated kinase protein kinases (for reviews see Refs. 39 and 40).
Although the only defined function of Grb2, based on invertebrate genetics, is to activate the Ras-MAPK pathway, biochemical evidence suggests that the Grb2 SH3 domains have additional binding partners and functions and that multiple pools of Grb2 may exist in mammalian cells. For example, interactions of the Grb2 SH3-C domain with the GTPase dynamin may regulate receptor endocytosis (41,42). In addition to mSos1, mSos2, and dynamin the Grb2 SH3 domains have been shown to bind other proteins with proline-rich sequences (43)(44)(45)(46). However, neither dynamin nor these other potential Grb2 binding partners have multiple SH3 binding motifs highly related to those in mSos1 or mSos2. Nevertheless, these results suggest that adaptors such as Grb2 may mediate the formation of multiple signaling complexes in a dynamic fashion and thereby orchestrate a network of signaling events downstream of tyrosine kinases.
Here we present data indicating that HPK1, a Ste20-related kinase, binds stably to Grb2 SH3 domains, both in vitro and in vivo. The topology of HPK1 is rather similar to that of mSos1, with a more N-terminal catalytic domain and an extended C-terminal region containing multiple proline-rich motifs (Fig.  1A). Even more striking, three of the proline-rich motifs in HPK1 are very similar to known Grb2 SH3-binding sites in mSos1 (Fig. 1B). Consistent with the notion that these HPK1 proline-rich motifs might constitute Grb2-binding sites, synthetic peptides modeled in these HPK1 sequences bound Grb2 in vitro (Fig. 3B). Furthermore, each of the Grb2 SH3 domains, expressed individually as GST fusion proteins, complexed with HPK1 from mammalian cell lysates (Fig. 3A). Similarly, isolated Grb2 SH3 domains associated with specific proline-rich HPK1 sequences in the yeast two-hybrid system, although with distinct specificities (Fig. 2). Most important, Grb2 bound stably to HPK1 in Cos1 cells. Efficient association of Grb2 with HPK1 required that both Grb2 SH3 domains be intact (Fig. 4), consistent with the notion that the two SH3 domains may preferentially recognize distinct sites and thereby bind synergistically to HPK1. HPK1 apparently differs from Sos proteins in its interactions with the Grb2 SH3-C domain, which binds only weakly to Sos proline-rich sequences and is dispensable for signaling to the Ras pathway in Drosophila (28,47). This may potentially be explained by the residue at the P Ϫ3 specificity position in each of the HPK1 and Sos proline-rich motifs. The Grb2 SH3-N domain binds preferentially to sites with Arg at P Ϫ3 as found in all of the Sos proline-rich motifs (48), and indeed the P1 and P4 sites of HPK1 that associate with SH3-N have Arg at this position. The Grb2 SH3-C domain, in contrast, bound the P2 site in HPK1, which has a Lys at P Ϫ3 followed by a Pro and additional Lys at P Ϫ4 and P Ϫ5 . These results are consistent with the hypothesis that the SH3-N domain binds either to the P1 or P4 sites in HPK1, whereas the SH3-C domain binds preferentially to the P2 site.
Although HPK1 appears to bind preferentially to Grb2 in vivo, it can also associate with Nck SH3 domains both in vitro and in vivo (Figs. 2, 3A, and 6). The observation that Nck binding was reduced by overexpression of Grb2 suggests that these adaptors may compete for binding to the same sites on HPK1, as suggested by the yeast two-hybrid analysis (Fig. 2). The HPK1 complexes formed in vivo will therefore likely reflect the local concentrations of specific adaptors and the affinities of their SH3 domains for HPK1 proline motifs. A distinction be-tween Grb2 and Nck binding to HPK1 is that the interaction of HPK1 with Grb2 appears constitutive, whereas its association with Nck is stimulated by EGF stimulation (Figs. 5A and 6). A similar observation of inducible binding has been made for the association of Grb2 with mSos1 in activated T cells (49). The molecular basis for the inducible interaction between Nck and HPK1 is unknown but might involve tyrosine phosphorylation of Nck or modification of HPK1. For example, we have previously found that tyrosine phosphorylation of the SH2/SH3 adaptor Crk-II can positively influence SH3-mediated interactions (50).
Biochemical Consequences of HPK1 Association with Grb2-The observation that Grb2 binds through its SH3 domains to HPK1 led us to investigate whether Grb2 could serve as a bridge to link tyrosine kinases to HPK1. For this purpose we employed the EGF receptor as the initial test system, since this receptor tyrosine kinase possesses Tyr(P)-X-Asn-X motifs known to bind the Grb2 SH2 domain with high affinity and also phosphorylates the Shc docking protein at Tyr(P)-X-Asn-X sites which consequently bind Grb2 (34,35). Indeed, EGF stimulation of transfected Cos1 cells induced the recruitment of HPK1 into a stable complex with both the autophosphorylated EGF receptor and tyrosine-phosphorylated Shc. The association of HPK1 with the EGF receptor was not dependent on HPK1 kinase activity but was potentiated by increased Grb2 expression (Fig. 5, B and C), suggesting that Grb2 inducibly couples HPK1 to specific phosphotyrosine-containing proteins. Of interest, HPK1 itself became tyrosine-phosphorylated in vivo following EGF stimulation, consistent with the view that HPK1 is recruited into a complex with the activated receptor. HPK1 was also phosphorylated on tyrosine in Cos1 cells expressing activated forms of the ␤PDGF receptor or the Src or Fps/Fes cytoplasmic tyrosine kinases, which are normally expressed in hematopoietic cells (37,38). Thus, it is possible that hematopoietic receptor tyrosine kinases and cytokine receptors, which recruit Grb2 either directly or through Shc phosphorylation, may interact with HPK1. Indeed, it is of interest to note that cytokines such as interleukin-3 and granulocyte colony stimulating factor can induce activation of the SAPK/JNK pathway in hematopoietic cell lines, which potentially might involve Ste20-like kinases such as HPK1. In the case of the granulocyte colony stimulating factor receptor, a specific Tyr at residue 763 is required for SAPK/JNK activation but is in part dispensable for Ras activation (51).
These results suggest a model in which SH2/SH3 adaptors may recruit HPK1 to phosphotyrosine-containing proteins in hematopoietic cells and thereby modulate HPK1-regulated signaling pathways. While this scheme is attractive, there are several caveats that should be noted. First, transfected HPK1 is constitutively active in stimulating the SAPK pathway, and we have therefore been unable to investigate whether EGF stimulation modifies the effect of HPK1 on the SAPK pathway. It seems likely that HPK1, like Ste20 itself, may have multiple inputs and outputs, and it is not yet clear whether SH3-mediated interactions or HPK1 tyrosine phosphorylation control HPK1 regulation of the stress-activated MAPK cassette. It is also possible that in hematopoietic cells, HPK1 may bind SH3containing proteins other than Grb2. In this context it is of interest that a hematopoietic-specific SH2/SH3 adaptor protein closely related to Grb2 has recently been described (52). The data we have obtained concerning the in vitro and in vivo interactions of the HPK1 proline-rich motifs with SH3-containing proteins can now serve as the basis for establishing the physiological binding partners and biological activators of HPK1 in hematopoietic cell types.
In summary, we have found that the Ste20-related kinase HPK1 has a C-terminal tail with proline-rich motifs that are closely related to those of mSos1/mSos2 and bind synergistically to the Grb2 SH3 domains in vitro. In transfected cells, HPK1 is recruited to activated EGF receptor and Shc and becomes tyrosine-phosphorylated following EGF stimulation. These observations raise the possibility that Grb2 might couple both to the Ras pathway and to a Ste20-related kinase, suggesting the possibility of cross-talk between pathways controlled by distinct serine/threonine kinases. The biological partners and in vivo targets of HPK1 are under investigation.