Mutational Analysis of the Nucleotide Binding Site of the Epidermal Growth Factor Receptor and v-Src Protein-tyrosine Kinases

Tyrosine kinases differ from serine/threonine kinases in sequences located at the active site where ATP and substrate bind. In the structure of cyclic AMP-depend- ent protein kinase, the catalytic loop contains the sequence Lys-Pro-Glu where the Lys residue contacts the (cid:103) -phosphate of ATP and the Glu residue contacts a basic residue located in the peptide substrate. In tyrosine ki- nases, the analogous sequence is Ala-Ala-Arg in the receptor tyrosine kinase subfamily and Arg-Ala-Ala in the Src tyrosine kinase subfamily. To deduce the role of these residues in tyrosine kinase function, site-directed mutations were prepared in the epidermal growth factor receptor (EGFR) and in v-Src and effects on ATP binding and kinase activity were determined. Changing Arg to either Lys or Ala dramatically reduced activity of both tyrosine kinases and this correlated with loss of ATP binding. Changing the orientation of this sequence impaired activity of EGFR to a greater extent than that of v-Src but did not change substrate specificity of the two enzymes. These results support the hypothesis that Arg functions to coordinate the (cid:103) -phosphate of ATP. Analysis of sequence inversions in the catalytic loop indicate that the active site of v-Src exhibits greater flexibility than that of EGFR. Protein kinases are divided into two major classes based on specificity for substrate hydroxyl-amino acid: serine/threonine (PSK) 1 and tyrosine (PTK) (1, 2). Sequence homologies within the core domains of all protein kinases and three-dimensional structures available for several members (3–11) indicate a similar overall conformation. However, amino acids assigned to the catalytic loop or deletion mutant cDNAs from pXER (20) into Xba I and Hin dIII sites of a pRC/CMV vector (Invitrogen) that was modified to contain a polylinker region derived from pBSK (cid:49) (Strat- agene). A polymerase chain reaction (PCR) protocol was used to prepare site-specific mutations. Three oligonucleotide primers were employed. EGFR v-Src— The activities of EGFR and were determined using (Boehringer Affinity-purified and various mutants were incubated (cid:57) -fluorosulfonylbenzoyl adenosine (FSBA) FSBA, kinases (25). FSBA-derivatized by blotting using polyclonal as probe. Kinase Assays and Kinetic using Reaction and different Reactions out facilitate After 32 by Kinetic data with the program Inc.). The concentration of EGFR and v-Src was determined by quantitative Western blotting. The ECL Western blot was measured by densitometry scanning and compared with a standard curve prepared from purified EGFR and v-Src.

Protein kinases are divided into two major classes based on specificity for substrate hydroxyl-amino acid: serine/threonine (PSK) 1 and tyrosine (PTK) (1,2). Sequence homologies within the core domains of all protein kinases and three-dimensional structures available for several members (3)(4)(5)(6)(7)(8)(9)(10)(11) indicate a similar overall conformation. However, amino acids assigned to the catalytic loop and to the Pϩ1 peptide binding loop regions differ between PSK and PTK and these differences distinguish the two major classes (see Fig. 1). Two subfamilies of PTK are distinguished by sequences located in the catalytic loop. The receptor tyrosine kinase subfamily contains Ala-Ala-Arg in the catalytic loop, while the Src subfamily contains this sequence in the reverse orientation Arg-Ala-Ala (1,2). The analogous position in the PSK protein kinase A is Lys 168 -Pro-Glu and corresponds to a loop containing amino acids important for catalysis and recognition of peptide substrate (4 -6). Lys 168 coordinates to the ␥-phosphate of ATP, and Glu 170 coordinates to a positively charged amino acid located at P-2 in the peptide substrate. Replacement of either residue with Ala markedly impaired catalytic activity of the Saccharomyces cerevisiae protein kinase A (12).
The Arg residue in the catalytic loop of PTK is predicted to function analogously to Lys in PSK coordinating the ␥-phosphate of ATP (13). In the crystal structure of the tyrosine kinase domain of the insulin receptor (InsR) Arg 1136 of the Ala-Ala-Arg sequence is hydrogen-bonded to the hydroxyl group of Tyr 1162 and makes other contacts to the carboxylate groups of Asp 1132 and Asp 1161 and with the indole ring Trp 1175 (11) Because ATP was not present in this structure, the relation of the Arg residue in the catalytic loop to ATP could not be compared to residues of the catalytic loop in the ternary complex of protein kinase A that contains ATP and a peptide inhibitor (5). Tyr 1162 is an autophosphorylation site and thus represents a substrate structure; Asp 1132 is the proposed catalytic base. The essential function of Arg was demonstrated by a Glu substitution in the human PTK Bruton's tyrosine kinase that resulted in X-linked agammaglobulinemia (14,15).
Although it appears likely that Arg in the catalytic loop of PTK contacts the ␥-phosphate of ATP analogous to the function of Lys in the catalytic loop of PSK (13,16), it has also been proposed to interact with substrate analogous to Glu 170 in protein kinase A. In the present studies, we have investigated the catalytic loop of PTK using site-directed mutagenesis to change Arg to Lys or Arg to Ala. Additionally, the orientation of this sequence in the catalytic loop of EGFR and v-Src has been reversed to examine the specificity of sequence that distinguishes the two PTK subfamilies. Results indicate that mutation of Arg dramatically reduced ATP binding and catalytic activity of both PTKs. Reversing the orientation of the Ala-Ala-Arg sequence severely impaired EGFR activity, but reversing the Arg-Ala-Ala sequence had lesser effects on v-Src, suggesting greater flexibility in the active site of v-Src.
Site-specific Mutagenesis-pRC/CMV-EGFR and the C-terminal deletion mutants pRC/CMV-cЈ991 EGFR and pRC/CMV cЈ957 EGFR were constructed by cloning EGFR or deletion mutant cDNAs from pXER (20) into XbaI and HindIII sites of a pRC/CMV vector (Invitrogen) that was modified to contain a polylinker region derived from pBSK ϩ (Stratagene). A polymerase chain reaction (PCR) protocol was used to prepare site-specific mutations. Three oligonucleotide primers were employed. * These studies were supported in part by National Institutes of Health Grant DK13149. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Supported by Postdoctoral Fellowship 1FB-0319 from the California Breast Cancer Research Program.
§ To whom correspondence should be addressed: Dept. of Medicine, University of California San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0650. Tel.: 619-534-4310; Fax: 619-534-0871. 1 The abbreviations used are: PSK, protein-serine kinase; PTK, protein-tyrosine kinase; EGF, epidermal growth factor; EGFR, EGF receptor; InsR, insulin receptor; protein kinase A, cyclic AMP-dependent protein kinase; PCR, polymerase chain reaction; FSBA, 5Ј-fluorosulfonylbenzoyl adenosine; WT, wild type. Primer 1 (GGTACCATGGAACGCCACATCGTTCGG) and primer 2 (ATATAAGCTTAAATGACAAGGTAGCG) each contained a restriction site for cloning purposes; primer 3 contains the desired mutation. The first PCR reaction mixture contained 100 pmol of primer 2 and 3, 20 ng of pRC/CMV-cЈ991 EGFR (with EGFR cDNA truncated at nucleotide 3231 (21)), 200 M dNTP, 10% (v/v) Me 2 SO, and 2.5 units of Pfu polymerase (Stratagene, San Diego, CA). Reaction conditions were: 1.5 min at 95°C to denature the template DNA, 2 min at 45°C to allow primer to anneal, and 3 min at 72°C for DNA extension. This reaction was cycled 30 times, and the DNA fragment containing the mutated base was generated. The first PCR product (0.6 pmol) was gel-purified and incubated with 20 ng of pRC/CMV cЈ957 EGFR (with EGFR cDNA truncated at nucleotide 3089) and extended to full length with five cycles of PCR. Then primers 1 and 2 were added directly to this second PCR mixture and run for 30 cycles. This reaction amplified the mutated EGFR kinase domain selectively. Mutations were confirmed by DNA sequencing, and the final PCR product was digested with restriction endonucleases ApaLI and BstEII and cloned into the plasmid pRC/CMV-EGFR.
v-Src cDNA was cloned into the plasmid pRSET-B (Invitrogen) to introduce a polyhistidine tag at the 5Ј end of v-Src cDNA. The polyhistidine-v-Src cDNA was excised from pRSET-v-Src with XbaI and Hin-dIII, and the resulting DNA fragment was gel-purified and recloned into the plasmid pRC/CMV. v-Src mutants were constructed with the Chameleon double-stranded site-directed mutagenesis kit (Stratagene) and verified by sequencing.
Protein Expression and Purification-Human embryonic kidney fibroblasts (293 cells) were grown in Dulbecco's modified Eagle's medium containing 10% calf serum. One day prior to transfection, 1.5 ϫ 10 5 cells were seeded into each well of a six-well dish. Transfections were carried out using 4 g of plasmid DNA by the calcium phosphate procedure (22), and cells were harvested 48 h later. Wild-type and mutant EGFR were isolated by immunoprecipitation with the monoclonal anti-EGFR antibody, 13A9. Transfected cells were washed twice with cold phosphate-buffered saline containing 1 mM EDTA and 1 mM EGTA and lysed in buffer containing 20 mM Hepes, pH 7.4, 1 mM EDTA, 6 mM ␤-mercaptoethanol, 0.05% Triton X-100, 10% glycerol, and 130 mM NaCl. EGFR were selectively precipitated with 13A9-protein G Sepharose (23). Recombinant v-Src and mutants were purified by Ni 2ϩ -affinity chromatography. Transfected 293 cells were harvested and homogenized in five volumes of Buffer A (10 mM Hepes, pH 7.8, 10 mM NaCl, 6 mM ␤-mercaptoethanol, 4 mM benzamidine, 2 g/ml aprotinin, 2 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, and 1% glycerol) containing 0.1% Triton X-100. The lysate was centrifuged at 115,000 ϫ g for 1 h, NaCl and imidazole were added to the supernatant to a final concentration of 300 mM and 8 mM, respectively, and this was loaded on a 1-ml Ni 2ϩ -NTA-agarose column (Qiagen). The column was washed with a linear gradient of imidazole (8 -40 mM) in Buffer A, and recombinant v-Src proteins were competitively eluted with 100 mM imidazole in Buffer A.
Western Blot Analysis of Phosphorylated EGFR and v-Src-Transfected 293 cells were stimulated with 30 ng/ml EGF for 10 min at 37°C. After stimulation, cells were lysed with 200 l of hot SDS sample buffer and proteins were separated on 8% SDS-polyacrylamide gels and transferred electrophoretically to polyvinylidene difluoride filters. For autophosphorylation analysis, polyvinylidene difluoride filters were incubated with 5% bovine serum albumin in TBST (20 mM Tris, pH 7.5, 150 mM NaCl, 0.02% Tween 20), incubated for 2 h with the anti-phosphotyrosine monoclonal antibody PY-20, and incubated for another hour with peroxidase-conjugated goat anti-mouse antibody. Immunoblots were developed using the ECL system (Amersham Corp.), and multiple exposures were obtained to ensure linearity. The autophosphorylation activity of v-Src was determined using a similar protocol but without EGF stimulation. Following the initial immunoblotting analysis, polyvinylidene difluoride filters were incubated with stripping buffer (62.5 mM Tris-HCl, pH 6.7, 100 mM ␤-mercaptoethanol, and 2% SDS) for 30 min at 50°C. After washing two times with TBST buffer, the filters were reprobed with anti-EGFR polyclonal antibody 1964 that is specific for the EGFR C terminus (24) or with the monoclonal anti-v-Src antibody 2-17. After washing three times with TBST, the filters were incubated with peroxidase-conjugated secondary antibodies and developed with ECL reagents.
ATP Binding to EGFR and v-Src-The ATP-binding activities of EGFR and v-Src were determined using an ATP-binding protein detection kit (Boehringer Mannehim). Affinity-purified EGFR, v-Src, and various mutants were incubated with 1 mM 5Ј-fluorosulfonylbenzoyl adenosine (FSBA) at room temperature for 30 min. FSBA, an ATP analog labeling reagent, binds covalently to the nucleotide binding sites of kinases (25). FSBA-derivatized proteins were detected by Western blotting techniques using polyclonal anti-FSBA antibody as probe.
Kinase Assays and Kinetic Analysis-Phosphate incorporation into the various peptide substrates was determined using the phosphocellulose paper binding assay (26). Reaction mixtures contained 10 mM Hepes, pH 7.4, 10 mM MnCl 2 , 100 mM Na 3 VO 4 , 10 M [␥-32 P]ATP, 100 ng of purified recombinant protein, and different concentrations of peptide in a final volume of 40 l. Reactions were carried out for 10 min at room temperature and terminated by the addition of 17 l of 24% trichloroacetic acid. Bovine serum albumin (100 g) was added to facilitate precipitation of the kinase domain. After centrifugation, 32 P incorporation into the peptide substrate was determined by spotting the supernatant onto phosphocellulose paper disks. The phosphocellulose disks were washed three times in 75 mM phosphoric acid, dried, and radioactivity measured. Kinetic data were analyzed with the program k-cat (BioMetallic, Inc.). The concentration of EGFR and v-Src was determined by quantitative Western blotting. The ECL Western blot signal was measured by densitometry scanning and compared with a standard curve prepared from purified EGFR and v-Src. provide a useful system for analysis of the activities of mutant EGFR. To examine the features of the catalytic loop required for kinase activity two point mutations: Arg 817 3 Ala (R817A), Arg 817 3 Lys (R817K), and a reversal of sequence mutation Ala-Ala-Arg 817 to Arg-Ala-Ala 817 (AAR 3 RAA) were prepared.

Effects of Changes in the Catalytic
Activity was first measured in vivo as EGF-dependent autophosphorylation. As shown in Fig. 2, EGF strongly stimulated autophosphorylation of WT EGFR. In contrast, mutations that changed the Arg residue to either Ala or Lys (R817A, R817K) abolished kinase activity. Retaining an Arg residue but reversing its position in the catalytic loop sequence to resemble that found in v-Src greatly impaired EGFR kinase activity to the range of Х1% of ligand-activated WT EGFR. Autophosphorylation of EGFR was not detected in untransfected 293 cells, indicating that measured activities reflected those of transfected EGFR (data not shown). In an immunocomplex kinase assay using a variety of peptide substrates, no activity of the EGFR mutants could be detected (see Table II).
To determine whether the severe reductions in tyrosine kinase activity exhibited by mutations of Arg 817 were due to decreased binding of ATP, wild-type and mutant EGFR were immunoisolated and incubated with FSBA, a suicide inhibitor of ATP. As shown in Fig. 3A, R817A and R817K mutant EGFR exhibited severe reductions in FSBA binding. The sequence reversal mutant AAR 3 RAA had greatly reduced but detectable FSBA binding in the range of 1% that of WT EGFR. The reduction in ATP binding of EGFR mutants corresponded closely to the reduction in tyrosine kinase activity, suggesting that Arg 817 is essential for ATP binding.
Effect of Changes in the Catalytic Loop of v-Src-Analysis of mutations in the catalytic loop of v-Src revealed that mutation of Arg 388 to Ala (R388A) or to Lys (R388K) abolished tyrosine kinase activity measured as enzyme autophosphorylation (Fig.  4). As observed for EGFR, an Arg residue in the catalytic loop was necessary for enzymatic activity. However, reversing the orientation of the catalytic loop to that found in EGFR had only a small effect on v-Src tyrosine kinase activity. ATP binding activity paralleled tyrosine kinase activity. As shown in Fig.  3C, the R388A and R388K v-Src mutants did not bind detectable amounts of the ATP analog FSBA, whereas FSBA binding by the loop sequence reversal mutant RAA 3 AAR approached that of wild-type v-Src. In v-Src, as in EGFR, an Arg residue in the catalytic loop is essential for ATP binding and enzymatic activity. Positioning of the Arg residue in the catalytic loop is, however, much more constrained in EGFR than in v-Src.
The kinase activities of wild-type and mutant EGFR and v-Src were further analyzed using in vitro kinase assays. Immunoprecipitated EGFR had a K m for AII peptide of 2.2 mM. The V max value for EGFR was 1.7 mol of phosphate min Ϫ1 mol Ϫ1 EGFR using AII peptide as a substrate. These data are comparable with previously reported kinetic parameters for EGFR tyrosine kinase activity (30). The activity of EGFR mutants was below the sensitivity limit of the in vitro kinase assay and could not be determined. Use of Mg 2ϩ in place of Mn 2ϩ did not reveal tyrosine kinase activity of mutant EGFR.
v-Src and the reverse orientation mutant RAA 3 AAR both showed detectable kinase activity in vitro as well as in vivo. Although the extent of autophosphorylation in vivo appeared similar for wild-type and RAA 3 AAR v-Src (Fig. 4), kinetic analysis in vitro indicated decreased activities for the RAA3 AAR mutant (Table I)  sequence reversal mutant of v-Src was approximately 2-fold higher than that of wild-type v-Src, and the V max was significantly reduced. The K m for both the Cdc2- (6 -20) and Src optimal peptide substrates were 5-fold or more higher for the reverse mutant compared to wild-type v-Src. The catalytic efficiency K cat /K m was correspondingly less for the RAA 3 AAR loop sequence reversal mutant v-Src.
Comparison of Substrate Preferences for Mutant EGFR and v-Src-Using selection from highly degenerate peptide libraries, Songyang et al. (31) deduced optimal tyrosine-containing peptide substrates for several PTKs. The crystal structure of InsR suggested that the Arg residue corresponding to EGFR Arg 817 interacted with a Tyr residue that is autophosphorylated in the activated InsR (11). The Arg in the catalytic loop may thus coordinate to both ATP and the peptide substrate. The sequence of the catalytic loop in EGFR and v-Src thus had the potential to alter substrate specificity. To determine whether sequence orientation in the catalytic loop affected substrate specificity, a variety of peptide substrates were tested using wild-type EGFR and v-Src and the catalytic loop sequence reversal mutants of each. As shown in Table II, while the loop reversal mutant of v-Src was in general less active than wild-type v-Src, in no case did the mutant have altered substrate preference. Specifically, no preference for the EGFR optimal peptide was acquired. Reversal of the loop sequence in EGFR reduced activity to undetectable levels, and no preference for v-Src kinase preferred substrates was acquired. DISCUSSION Mutation of the Arg residue in the catalytic loop of either EGFR or v-Src abolished ATP binding and phosphotransfer activity. These results support the proposal that, analogous to Lys 168 in protein kinase A, Arg coordinates to the ␥-phosphate of ATP providing charge neutralization. Neither Ala, which eliminated the side chain beyond the ␤-carbon without imposing severe constraints on conformation (32), nor Lys, which retains the ability to neutralize charge, could replace Arg. Although homology modeling of EGFR suggested that the amino group of lysine should be able to interact with the ␥-phosphate of ATP (13), Arg is the conserved residue in the catalytic loop of the extended PTK family (1,2). This suggests that the ␦-guanidinium group of Arg interacts with additional residues, as observed in the InsR tyrosine kinase structure (11). In InsR Arg 1136 makes a hydrogen bond with the catalytic base Asp 1132 , whereas in protein kinase A the catalytic base was hydrogen-bonded to the Thr 201 . A need to simultaneously interact with additional residues such as the catalytic base and residues in the Pϩ1 site can only be provided by Arg, which is proposed to play an essential role in both ATP binding and catalysis.
Reversing the orientation of the catalytic loop had differential effects on members of the two subfamilies of PTK. Arg is located 4 and 2 residues from the catalytic base in WT EGFR and v-Src, respectively (1). Reversal of the orientation of the sequence in the catalytic loop in EGFR seriously impaired both ATP binding and catalytic activities. In an attempt to understand these effects, a homology model of the active site of EGFR was prepared using the coordinates of the InsR kinase core. As shown in Fig. 5, reversing the orientation of the loop resulted in the ␦-guanidinium group pointing away from the catalytic base Asp 813 . This orientation also disrupted interaction with the Pϩ1 site. Because ATP was not present in the InsR structure, these changes that result from catalytic loop sequence reversal do not address additional effects on coordination to ATP.
The orientation of this sequence has less effect on v-Src. ATP binding was only slightly reduced as assessed with FSBA binding, and the K m for ATP of RAA3 AAR v-Src was increased only about 2-fold. This suggests that Arg coordination to the  The kinase reaction was performed as described under ''Experimental Procedures.'' Recombinant protein (100 ng) was used in each assay, and the final concentration of each peptide is stated in the table. Kinase activity was measured as the initial velocity of the phosphorylation reaction and was expressed as pmol of PO 4 incorporated into the substrate peptide per min. ND, not detectable.
␥-phosphate of ATP is maintained. However, the catalytic efficiency (k cat /K m ) of the reverse mutant of v-Src was about 25fold lower than WT v-Src using Cdc2- (6 -20) and Src optimum peptide substrates. This suggests that, analogous to Arg 1136 interactions in InsR, Arg 388 in v-Src interacts with the catalytic base and Pϩ1 site. Reversal of sequence in the catalytic loop likely impairs these functions of the Arg residue. The flexibility at the active site of v-Src appears greater than that of EGFR and is consistent with the observation that Src tyrosine kinase can phosphorylate a wide spectrum of peptide substrates including substrates with different chirality and chain length of alcohol (33).
No change in peptide substrate preference was observed with catalytic loop sequence reversals. Neither PTK acquired the peptide substrate preference of the other. These results suggest that the orientation of the catalytic loop is not a major structural determinant of substrate preference. Rather these sequences are essential to ATP binding and catalysis.