Structural characterization of autoinhibited c-Met kinase produced by coexpression in bacteria with phosphatase

Abstract

Protein kinases are a large family of cell signaling mediators undergoing intensive research to identify inhibitors or modulators useful for medicine. As one strategy, small-molecule compounds that bind the active site with high affinity can be used to inhibit the enzyme activity. X-ray crystallography is a powerful method to reveal the structures of the kinase active sites, and thus aid in the design of high-affinity, selective inhibitors. However, a limitation still exists in the ability to produce purified kinases in amounts sufficient for crystallography. Furthermore, kinases exist in different conformation states as part of their normal regulation, and the ability to prepare crystals of kinases in these various states also remains a limitation. In this study, the c-Abl, c-Src, and c-Met kinases are produced in high yields in Escherichia coli by using a bicistronic vector encoding the PTP1B tyrosine phosphatase. A 100-fold lower dose of the inhibitor, Imatinib, was observed to inhibit the unphosphorylated form of c-Abl kinase prepared by using this vector, compared to the phosphorylated form produced without PTP1B, consistent with the known selectivity of this inhibitor for the unactivated conformation of the enzyme. Unphosphorylated c-Met kinase produced with this vector was used to obtain the crystal structure, at 2.15-Å resolution, of the autoinhibited form of the kinase domain, revealing an intricate network of interactions involving c-Met residues documented previously to cause dysregulation when mutated in several cancers.

Sequencing of the human genome indicates there are >500 different protein kinase genes expressed in man (1). Many of these are already known to play important roles in biology, and all could potentially be important as targets for pharmaceutical intervention in medicine. Conservation in the active site residues within the protein kinase gene family makes the development of selective kinase inhibitors challenging. Structural biology offers valuable information useful in the design of new inhibitors (2), but a limitation in its application to kinases can often be the inability to produce highly purified proteins in amounts suitable for cocrystallography. Inhibitor binding sometimes can be sensitive to the specific conformation state of a kinase (3), or to changes in the kinase sequence caused by mutations, such as those occurring during cancer progression (4–7). These pose further barriers to the implementation of structural approaches to drug design, as there can be a need to produce the target kinase in several different forms.

Through efforts to create a robust system to produce protein kinases, we discovered that, contrary to common belief, it is possible to produce many kinases in bacteria, including catalytic domains of receptor tyrosine kinases (8). We discovered that good production systems can be developed by using Escherichia coli by a simple strategy involving testing many different N- and C-terminal boundaries for optimal expression (8). Such analyses were previously difficult because of the expense of oligonucleotide PCR primers, but these now are readily manageable. We found that kinase samples produced in bacteria can be heterogeneously autophosphorylated during expression in bacteria, but that coexpression with different phosphatases works to produce kinases in an unphosphorylated form (8). In the current study, we describe in detail the production of the c-Abl, c-Src, and c-Met kinases using such a system.

c-Met is the membrane receptor for hepatocyte growth factor (HGF), and is important for liver development and regeneration (ref. 9, and references therein). A link between c-Met and cancer was made when it was first cloned as an oncogene, later found to be a truncated protein fused to the translocated promoter region locus as the result of a gene translocation (ref. 10, and references therein). Further links to cancer have been documented through the identification of germline mutations in the c-Met gene in the majority of hereditary papillary renal carcinomas (11, 12), and in gastric cancer (13). Somatic mutations in the c-Met gene have been identified in sporadic papillary renal carcinomas (14), small cell lung cancer (15), squamous cell cancer of the oropharynx (16), hepatocellular carcinomas (17), and lung and lymph node metastases (18, 19). Such truncated and mutated forms of c-Met are found to transform cells in culture (18, 20), as well as to cause tumor formation in transgenic mice (21). When c-Met expression is expressed at high levels in mice, it loses its dependence on HGF stimulation (22). However, in the majority of cancers where c-Met plays a role, it is thought to be through a modest induction of c-Met expression levels, and it has been demonstrated that hypoxia can up-regulate the c-Met gene (23–25). Even with activating point mutations, the oncogenic actions of c-Met typically still require increased expression levels (26, 27), and remain dependent on HGF stimulation (28). Strategies to reduce c-Met activity include targeting both the extracellular receptor domain in addition to the intracellular tyrosine kinase domain (23–25, 29–31).

The c-Met receptor is composed of an extracellular alpha chain and a transmembrane beta chain, products of a single gene that become proteolytically cleaved but that remain associated through a disulfide bond (see ref. 32 for review). Crystal structures have been reported for the extracellular c-Met Sema domain (33), as well as a mutated form of the intracellular tyrosine kinase domain (34, 35). Signaling through c-Met is thought to occur upon HGF binding through dimerization in the membrane (23), leading to activation of the autoinhibited receptor through transphosphorylation. Once phosphorylated, the intracellular domains intiate a cascade of signaling by binding to several other proteins at a multifunctional docking site linked to the C terminus of the kinase domain (36). The juxtamembrane residues linked to the N terminus of the kinase participate in modulation of the signaling cascade through the recruitment of phosphatases (37) and ubiquitination complexes (38). Within the kinase domain itself, activation of the wild-type c-Met involves the required phosphorylation of two tyrosines in the activation loop, occurring stepwise, first at Tyr 1235 and following at Tyr 1234 (39). For activation of c-Met harboring oncogenic point mutations, the requirement for phosphorylation at Tyr 1234 can become lost (40, 41). Such regulation likely relate to specific structural features of the kinase domain.

In this study we present the crystal structure of unphosphorylated c-Met kinase, revealing how this kinase can exist in an autoinhibited form when not activated by phosphorylation. The structure reveals the likely mechanism for the oncogenic activities of particular mutations found in patients with hereditary forms of renal cancer. Also in this study an analysis is made of c-Abl and c-Src kinases produced through coexpression with phosphatase in E. coli. Phosphorylation-dependent differences are documented for bacterially expressed c-Abl in the sensitivity to the inhibitor, Imatinib, shown previously to inhibit preferentially the unphosphorylated form of c-Abl (3). Thus the use of the phosphatase coexpression system can facilitate the development of kinase inhibitor therapeutics that target different protein conformation states.

Results

The bicistronic pET-N6 BI-PTP plasmid (Fig. 1A) is a convenient vector for bacterial expression of unphosphorylated tyrosine kinases. We have engineered several kinases into both this vector and the parent pET-N6 vector lacking the PTP (Fig. 1B). Thus, c-Abl and c-Src coding sequences were engineered into these vectors, choosing boundaries similar to ones described in earlier structure determination reports. c-Met kinase domain also was engineered into these vectors, but the boundaries had previously been determined empirically by testing several different N and C termini for optimal expression (8). Although all three kinases have previously been produced by using baculovirus systems, all three were found to express well in E. coli, yielding amounts >1 mg per liter culture, convenient for making preparations for crystallography.

Fig. 1.

Kinase expression in bacteria. (A) Sequence of the N-terminal HIS-tag and the NdeI and SalI polylinker region from pET-N6 BI-PTP, the bicistronic expression vector used for coexpression of protein tyrosine kinases with the catalytic fragment of the tyrosine phosphatase, PTP1B. (B) Kinase-encoding sequences are ligated into vectors without (pET-N6) or with (pET-N6 BI-PTP) the phosphatase-encoding sequences, for production of the phosphorylated or the unphosphorylated proteins, respectively. (C) (Top) Coomassie-stained SDS/PAGE of 1 μg of the HIS-tagged c-Abl, c-Src, or c-Met kinases after expression without (−) or with (+) phosphatase coexpression, and after purification by metal affinity chromatography. (Middle) Western blot detection of phospho-Tyr present in 10 ng of the same kinases separated by SDS/PAGE as in Top. (Bottom) Western blot detection of phospho-Tyr present in 10 μg of the unpurified soluble protein extract from the E. coli cultures used to produce the kinases in Top and Middle.

The successful bacterial expression of these soluble kinases did not require the coexpression of a phosphatase; each showed high enrichment after only the metal affinity purification (Fig. 1C Upper). However, the coexpression of the phosphatase allowed the preparation of kinases that were minimally phosphorylated, as determined by Western blotting with an anti-phosphotyrosine antibody. Whereas 10 ng of the metal affinity-purified kinases gave a clear phosphotyrosine signal when no PTP was expressed, this signal was abolished when the same kinases were coexpressed with PTP (Fig. 1C, middle panel). Similarly, when 10 μg of unpurified soluble proteins from the extracted bacteria were analyzed by Western blotting, all three kinases were shown to have phosphorylated the bacterial proteins, but that the coexpression of the PTP could mostly reverse this (Fig. 1C Lower). These experiments demonstrate that all three of these kinases are active during the culturing, but that coexpression of active PTP can reverse the phosphorylation.

All three kinases produced in bacteria show activitiy in vitro after the extraction and purification, as demonstrated for c-Abl (Fig. 2). When the phosphorylated c-Abl made without PTP was compared with the unphosphorylated c-Abl produced by PTP coexpression, there was a marked difference in sensitivity to the inhibitor Imatinib; the IC₅₀ concentrations differed by 100-fold. This result confirms the difference in sensitivies observed previously with dephosphorylated and phosphorylated c-Abl preparations derived from baculovirus cultures (3). This result demonstrates that these kinases made in bacteria are useful for activity analyses, and are especially good starting points for investigations of conformation-dependent activities or inhibitor selectivities.

Fig. 2.

Comparison of dose-dependent inhibition of c-Abl kinase activities by Imatinib. The inhibition of the unphosphorylated (UP) c-Abl produced in E. coli using coexpression with phosphatase occurs with an IC₅₀ of 28 nM (±5 nM), compared to an IC₅₀ of 3.3 μM (±1.1 μM) for the phosphorylated c-Abl produced in E. coli without phosphatase. Points are duplicates normalized to 100% for the uninhibited kinase, with error bars representing the standard deviation of the mean.

We obtained the crystal structure of apo c-Met tyrosine kinase domain at 2.15-Å resolution with the wild-type, unphosphorylated protein produced in bacteria by using PTP coexpression. The c-Met protein used for previous structures used material from baculovirus cultures in which unphosphorylated material was obtained by mutating the tyrosines of the activation loop that otherwise would become phosphorylated during the production (34, 35), and therefore the mechanisms for autoinhibition of c-Met kinase have not yet been fully revealed structurally. By using the wild-type, unphosphorylated protein, we investigated the conformation that underlies the autoinhibited state. The refinement statistics are shown in Table 1.

Table 1.

Statistics of crystallographic data and refinement

Crystallization and data collection
Unit cell dimensions, Å	a = 103.8
Space group	P213
Solvent content, %	54.7
Resolution range, Å	20–2.15
Unique reflections (highest shell)*	19,665 (1,643)
Data redundancy (highest shell)	5.4 (5.1)
Completeness (highest shell), %	99.2 (100)
I/σ(I) (highest shell)	5.3 (0.7)
Rsym (highest shell), %	12.5 (55.9)
Refinement Statistics
σ cut off	None
Total non-hydrogen atoms	2546
Average B factor, Å2	35.7
Rcryst/Rfree, %	22.3/25.9
rms deviation
Bond lengths, Å	0.007
Bond angles, °	1.097

R_sym = Σ|I_avg − I_j|/Σ I_j. R_cryst = Σ | F_o − F_c |/ΣF_o, where F_o and F_c are observed and calculated structure factors, respectively. R_free was calculated from a randomly chosen 5% of reflections excluded from the refinement, and R_cryst was calculated from the remaining 95% of reflections. The rms deviation values are the rms deviation from ideal geometry.

*Highest shell resolution, 2.21–2.15 Å.

All of the residues of the apo c-Met kinase domain are visible in the structure, including those of the P-loop and the activation loop. As expected, the structure shows the canonical bilobed protein kinase fold (Fig. 3A), but with an extra helix at the N terminus, and conformations of the activation loop and P-loop that are dramatically different from the reported c-Met structure (35). The kinase is in an inactive conformation, which causes the residues important for catalysis to be too far apart for activity. Thus, the distance between Lys 1110 (the conserved residue that stabilizes ATP phosphate binding; ref. 42) and Glu 1127 (the conserved residue from the αC helix that should stabilize the orientation of Lys 1110 after kinase activation, ref. 42) is 10.3 Å, and Asp 1222 (the conserved residue from the “DFG” motif known to function in stabilizing Mg binding during catalysis; ref. 42), is pointed away from the blocked active site. The space between Lys 1110 and Glu 1127 is occupied by the leading residues of the activation loop, with Phe 1223 from the DFG motif projecting inwardly to form van der Waals contacts with residues Met 1131, Phe 1134, Val 1139 and Phe 1200, and with Leu 1225 also forming buried van der Waals contacts with residues Phe 1124 and Val 1155. As the activation loop main chain becomes exposed to the surface, residue Arg 1227 forms a salt bridge with Glu 1127 (Fig. 3B), thus underscoring the inactive aspect of this kinase conformation.

Fig. 3.

Ribbon cartoon images of the autoinhibited c-Met kinase domain. (A) The entire kinase domain, colored by rainbow from the N terminus to the C terminus, with a white rectangle circumscribing the region of the activation loop that blocks the active site. (B) A close-up of the autoinhibitory activation loop, with a selection of the side chains shown as stick figures, depicting the more buried interactions stabilizing the inhibitory conformation. (C) A close-up with a different selection of side chains, shown as stick figures, depicting stabilizing interactions occurring in a more superficial layer.

In most respects, these features of the leading residues of the activation loop are similar to those reported for a structure of c-Met bound to the alkaloid inhibitor, K252 (Protein Data Bank code 1ROP; ref. 35). However, the subsequent residues of the activation loop in the current structure adopt a conformation that differs strikingly, by forming a series of autoinhibitory interactions with the P-loop, the αC helix, and with itself. Thus, Met 1229 projects into the ATP-binding pocket, sandwiched between the P-loop residue, Phe 1089, and the side chain of Lys 1110. The P-loop conformation is further stabilized through interactions with the main chain of Tyr 1230 and the side chain of Lys 1232, as the activation loop turns in the direction of the αC helix. Two activation loop residues that are required to become phosphorylated for activation, Tyr 1234 and Tyr 1235, project in opposite directions, with Tyr 1234 pointing inward and forming a hydrogen bond with catalytic residue, Glu 1127, and with Tyr 1235 pointing to solvent, with its phenyl side chain sitting buried in a contour formed by the subsequent residues of the activation loop. Before the activation loop traverses to the lower lobe, yet another layer of autoinhibitory stabilizing interactions are formed through salt bridges of Lys 1240 with Asp 1228 and Asp 1231 (Fig. 3C). The remainder of the activation loop is involved in crystal packing contacts, which may contribute to stabilizing the loop in the observed conformation. However, the extensive interactions between the activation loop with the inside of the active site strongly suggest this conformation could be adopted in solution.

Discussion

From the experience described here with c-Abl, c-Src, and c-Met, as well as from several other kinases we have tried, we know that phosphatase coexpression systems offer tremendous utility in the bacterial production of many kinases, as well as other proteins. However, we have found some kinase domains, such as those for ZAP70 and c-KIT, remain a challenge to overexpress in E. coli, most likely due to the presence of surface hydrophobic groups that make it difficult to maintain sufficient solubility. Variations of the system described here using other phosphatases have also worked successfully to produce unphosphorylated enzyme. When optimal N- and C-terminal boundaries are chosen, active c-Abl, c-Src, and c-Met kinases can be produced in E. coli even without the use of phosphatases, indicating that E. coli is not catastrophically sensitive to toxic effects of these kinase activities. This is at odds with a description using yeast (43), which might be explained trivially by differences in how the kinase boundaries were selected, what fusion tags were used, or the culture conditions selected, but may indicate that one or more essential eukaryotic yeast proteins are more highly sensitive to phosphorylation compared to E. coli.

Several features of c-Met autoregulation can be understood from the autoinhibited structure presented here. In the quiescent state, neither of Tyr 1234 or 1235 are phosphorylated. This is certainly logical for Tyr 1234, because phosphorylation of this buried residue would be incompatible with the autoinhibited conformation of the activation loop. By contrast, the phenolic hydroxyl of Tyr 1235 is solvent-exposed and predicted to be more available for phosphorylation. However, phosphorylation at Tyr 1235 should require some motion of the activation loop to allow this residue enough freedom to serve as a substrate in a transphorylation reaction. Such a model is consistent with the observation that Tyr 1235 becomes phosphorylated ahead of Tyr 1234 during the activation process (39). However, phosphorylation at Tyr 1235 should only partially destabilize the autoinhibited conformation, because Tyr 1234 could still remain buried; this is consistent with the finding that phosphorylation at Tyr 1234 is required in addition to that at Tyr 1235 for full activation to occur (39).

The effects of some of the mutations found in cancer can also be rationalized from the autoinhibited structure (Fig. 4). Mutation of Tyr 1235 to Asp should be more destabilizing to the autoinhibited conformation than even phosphorylation at this site, because this mutation would bring the anionic charge into the hydrophobic space normally occupied by the Tyr phenyl side chain. This mutation was incorporated into the protein crystallized previously (34, 35), with the effect of causing the local section of the activation loop to adopt a conformation that extends away from the active site. Mutation of Tyr 1230 to His or Cys, found sporadically in cancers, should destabilize the inhibitory interactions made by the activation loop with the P-loop. Mutations of Asp 1228 to either Asn or His, found in hereditary renal cancer, should disrupt the salt bridge with Lys 1240, and therefore partially destabilize the autoinhibited conformation; this is consistent with the finding that such mutations, when present in the germ line, predispose the individual to cancer that still requires decades to appear (44), an increase in c-Met expression (26), and stimulation by HGF (28, 39).

Fig. 4.

Location of mutations identified in human cancers. The ribbon cartoon image is colored teal, with the activation loop shown in yellow. The P-loop and aC helix are indicated by arrows. Residues K1110, E1127, and D1222 are displayed as sticks. Residues found to be mutated in human cancers are displayed as spheres at their CA atoms. Mutations include V1092I (12, 14); H1094L,Y,R (14); H1106D (14); G1119V (51); M1131T (20); T1173I (17); V1188L (20); L1195V (28); V1220I (20); D1228H,N (28, 40); Y1230H,C,D (14, 19); Y1235D (16, 18, 19); K1244R (17); and M1250T,I (17, 28, 40).

A pattern emerges within the set of mutations found in cancer of changes that cause a partial instability of the autoinhibited state, with no mutations found to date that would completely disrupt the autoinhibition, such as would be expected for a mutation of Tyr 1234. From the current and previous structures, some of the mutations found in cancer patients, including Met 1250 to Thr or Ile, can be mapped to locations where they conceivably could indirectly destabilize the P-loop or activation loop from their autoinhibited conformations. From the current structure all these would be predicted to allow some of the control to remain intact, as has been observed. Certainly, however, there exist subtleties about the structures of mutated forms of c-Met that require additional studies to explain. Thus, individual mutations are found to vary in the types of cancer they promote (21). Furthermore, individual mutations can vary in their affinities for small molecule inhibitors (6, 7). The latter observations suggest that a subset of cancer patients treated with c-Met kinase inhibitors may become resistant, as observed with inhibitors of c-Abl and EGF kinase inhibitors (4, 5). However, from the current structure, it appears possible that some of the known mutations might facilitate the search for new inhibitors, because they may mimic conformation states that exist during the normal activation process. If such alternative states can be targeted effectively, the mutated forms may provide a source of material conducive for identifying novel inhibitory compounds.

It has been reported that the kinase inhibitor, Imatinib, selectively inhibits the unphosphorylated form of c-Abl expressed in mammalian cell cultures (3). Our in vitro biochemical assays of bacterial c-Abl confirm these findings. This establishes the utility of the phophatase coexpression system for identifying new inhibitors that act selectively on the nonactivated, nonphosphorylated forms of such kinases. In conjunction with studies of mutated kinases and kinases activated in vitro, the expression system described here should help in the discovery of kinase inhibitors for many clinical needs.

Materials and Methods

Plasmid Engineering.

The vector (pET-N6) used for bacterial expression of tyrosine kinases was engineered as a derivative of pET28 (Novagen), by using synthetic oligonucleotides to replace the pET28 HIS-tag and polylinker with a noncleavable N-terminal HIS-tag placed upstream of an NdeI-SalI polylinker (Fig. 1A). A bicistronic vector (pET-N6 BI-PTP) that encodes the first 283 catalytic residues of the human protein tyrosine phosphatase, PTP1B [National Center for Biotechnology Information (NCBI) accession no. NM_002827), on a bicistronic mRNA was engineered by using PCR to amplify the sequences encoding PTP1B and inserting them with a second ribosome binding site into the pET-N6 SalI polylinker site (Fig. 1A).

The DNA fragments encoding the kinases analyzed here were obtained by PCR of cDNA made from human tissues (Invitrogen) by using synthetic oligonucleotides as primers. The final vectors encoded Gly 1056 through Gly 1364 of c-Met (NCBI NM_000245), Gly 227 through Val 515 of c-Abl (NCBI NM_005157), or Val 86 through Leu 536 of c-Src (NCBI NM_005417). Each PCR was engineered to be flanked with NdeI and SalI restriction sites for ligation into the pET-N6 and pET-N6 BI-PTP vectors (Fig. 1B).

Kinase Protein Expression.

N-terminal HIS-tagged kinases for small-scale analysis were produced by using E. coli strain BL21(DE3) (RIL) (Stratagene) in 100 ml of 2YT media with antibiotics, and induced with 0.5 mM IPTG overnight at 18°C. Centrifuged culture pellets were extracted by sonication in 50 mM Tris (pH between 7.0 and 8.0 depending on protein pI), 250 mM NaCl, 0.1% Triton X-100, 0.04 mM PMSF, 0.02% monothioglycerol, and 150 μg/ml lysozyme. Extracts were clarified by centrifugation for 30 min at 17,000 rpm (41,800 × g) in a SA600 rotor (Sorvall) to yield the soluble extract. HIS-tagged proteins present in the soluble extracts were purified with Talon resin (BD Bioscience), with elution in 50 mM Tris (pH between 7.0 and 8.0), 100 mM NaCl, 10% glycerol, 0.04 mM PMSF, and 0.02% monothioglycerol.

For detection of tyrosine phosphorylation of proteins from E. coli, 10 μg of the soluble extract or 10 ng of the affinity-purified HIS-tagged kinase was analyzed by Western blotting using anti-phosphotyrosine mouse monoclonal antibody (PY100; Cell Signaling Technology) with horseradish peroxidase (HRP)-goat anti-mouse secondary antibody (Jackson ImmunoResearch) and detection using ECL Plus (Amersham Pharmacia).

c-Met Expression Scale-Up and Purification.

c-Met kinase was produced in 30-liter Bioreactor cultures of E. coli strain BL21(DE3) RIL (Stratagene) using Terrific Broth, with 15-h induction at 12°C using 1 mM IPTG. Frozen cell pastes suspended with 40 ml of lysis buffer (100 mM potassium phosphate, pH 8.0/250 mM NaCl/0.1% Igepal/5% glycerol/25 mM imidazole/2 mM PMSF) per liter of cells were lysed by using a microfluidizer (Microfluidics M-110H) at 18,000 psi, and clarified by centrifugation at 25,000 × g at 4°C for 1 h. Supernatants were fractionated by using Ni-Chelating Sepharose FF (GE Healthcare), washing with lysis buffer containing 50 mM imidazole, and by 10-column volumes of 50 mM Tris (pH 8.8), 150 mM NaCl, and 25 mM imidazole, and eluted with 500 mM imidazole in the same buffer. Eluted protein was fractionated with 20 ml SP and Q FF columns (GE Healthcare) in 20 mM Tris (pH 7.5), 5% glycerol; protein flow through from SP FF column was diluted into 20 mM CHES (pH 9.5) and reloaded over Q FF column. Eluted protein was concentrated and fractionated with Superdex200 26/60 SEC (GE Healthcare) in 20 mM Tris (pH 8.8), 150 mM NaCl, 5% glycerol, and 10 mM DTT. The c-Met protein behaved monomerically and was concentrated to 20 mg/ml and stored at −80°C.

c-Met Crystallization, Data Collection, and Structure Determination.

Crystals of human c-Met kinase domain were grown by the sitting drop vapor diffusion method at 4°C. Protein solution at 16 mg/ml containing 20 mM Tris·HCl (pH 8.5), 100 mM NaCl, and 14 mM 2-mercaptoethanol, was mixed with an equal volume of reservoir solution containing 1.0 M diammonium hydrogen phosphate, 0.2 M sodium chloride, 0.1 M citrate (pH 5.0), and 7.5% glycerol. Triangular crystals grew within 8 days to a size of 0.15-× 0.15-× 0.15 mm.

A complete x-ray diffraction data set of a c-Met crystal was collected at the Advanced Photon Source (APS) COM-CAT beam line under cryogenic temperature. The diffraction data were integrated and scaled by using elves (45), mosflm, and scala (46) (Table 1).

The structure determination with data up to 4.0 Å was successful by molecular replacement (MR) using amore (47) with homology models from earlier kinase structures. The final MR calculations used a model derived from the structure of the insulin receptor kinase in its inactive form (Protein Data Bank ID code 1IRK). The space group was P2₁3 with one molecule in the asymmetric unit. The initial electron density map revealed an extra alpha-helix near the N terminus, and significant differences in the P-loop and activation loop regions compared to the initial model, which was rebuilt by using O (48). The rebuilt model was refined to completion (Table 1) by using cnx (49) and refmac5 (50) against 2.15-Å data with least squares refinement, individual B-factor refinement, and TLS refinement protocols.

Biochemical Protein Phosphorylation Assays.

Kinase activities were determined in a reaction buffer of 50 mM Hepes (pH 7.2), 5 mM MgCl₂, 5 mM MnCl₂, 0.02% BSA, 0.01% Nonidet P-40, 5% DMSO, and 10 μM ATP (all reagents from Sigma). The final enzyme (c-Abl, c-Met, or c-Src) concentration was ≈1 μg/ml. A biotinylated substrate peptide (Biotin-EEEEYEEEEYEEEEYEEEE, from New England Peptide) was used at a final concentration of 200 μg/ml. Twenty-microliter reactions at room temperature were initiated by ATP addition. After 30 min, the reactions were stopped by adding 5 μl per well of stop buffer (50 mM Hepes, pH 7.2/20 mM EDTA/0.02% BSA/0.01% Nonidet P-40). The extent of phosphorylation was measured by using AlphScreen PY20 kits (PerkinElmer). For inhibition of c-Abl, Imatinib (Novartis) was purchased and dissolved in water before dilution into the assay.

Abbreviation

HGF

hepatocyte growth factor.