Solution Structure and Backbone Dynamics of the Non-receptor Protein-tyrosine Kinase-6 Src Homology 2 Domain*

Human protein-tyrosine kinase-6 (PTK6, also known as breast tumor kinase (Brk)) is a member of the non-receptor protein-tyrosine kinase family and is ex-pressed in two-thirds of all breast tumors. To under-stand the structural basis of PTK6 function, we have determined the solution structure and backbone dynamics of the PTK6-Src homology 2 (SH2) domain using multidimensional NMR spectroscopy. The solution structure clearly indicates that the SH2 domain of human PTK6 contains a consensus (cid:1) / (cid:2) -fold and a Tyr(P) peptide binding surface, which are common to other SH2 domains. However, two of the (cid:1) -helices ( (cid:1) A and (cid:1) B) are located on opposite faces of the central (cid:2) -sheet. In addition, the topological arrangement of a central four-stranded antiparallel (cid:2) -sheet (strands (cid:2) A, (cid:2) B, (cid:2) C, and (cid:2) D) differs from that of other Src family members. Backbone dynamics and Tyr(P) peptide titration experiments revealed that the putative ligand binding sites of the PTK6-SH2 domain undergo distinctive internal motions when compared with other regions of the protein. Surface plasmon resonance analysis showed that the Tyr(P) peptide had a dissociation constant of about 60 (cid:3) M , which is substantially weaker binding than previously by solid-phase Fmoc ( N -(9-fluorenyl)methoxycarbonyl) chemistry (17). After trifluoroacetic acid cleavage, the peptide was eluted by analytical HPLC with the pH value of 3–4. The purity of synthetic peptide was confirmed by preparative HPLC and matrix-assisted laser desorption ionization time-of-flight mass spectrometry. NMR Spectroscopy— All NMR experiments were performed at 298 K on a Bruker DRX500 or DRX600 or Varian Unity INOVA 500-MHz spectrometer equipped with a triple resonance probe and shielded triple axis gradients. The 1 H chemical shifts were referenced to internal sodium 4,4-dimethyl-4-silapentane-1-sulfonate. The 13 C and 15 N chemical shifts were referenced indirectly using the 1 H/ 13 C or 1 H/ 15 N fre- quency ratios of the zero point: 0.101329118 ( 15 N) and 0.251449536 ( 13 C) (18). The 1 H- 15 N HSQC (19–20) experiment was acquired using a uniformly 15 N-labeled sample. Pulsed-field gradient techniques with a WATERGATE pulse sequence (21) were used for all H 2 O experiments, resulting in a good suppression of the solvent signal. An HCCH-TOCSY (22) experiment with 12-ms mixing time and two sets of three-dimen- sional 15 N-edited TOCSY-HSQC (21, 23–25) experiments with 45- and 77.75-ms mixing times were performed. 15 N-Edited NOESY-HSQC (21, 24, 25) experiments with mixing times of 150 and 200 ms in 90% H 2 O, 10% D 2 O solution and a 13 C-edited NOESY experiment (21, 24–26) in D 2 O solution with a mixing time of 100 ms were collected. Two-dimen- sional 1 H- 15 N heteronuclear multiple quantum coherence-J (27), three-dimensional HNHA (28), and a series of 1 H- 15 N HSQC experiments were performed. The three-dimensional triple resonance experiments, force field on the DISCOVER3 module. To obtain the lowest energy conformation, restrained molecular dynamics was performed for 250 ps at 298 K. The lowest energy conformers were sampled every 100 steps during the molecular dynamic calculation. The restrained energy minimization procedure using the Broyden-Fletcher-Goldfard method was performed for the final structure until its energy gradient deviation decreased to 0.001 kcal (cid:2) mol (cid:3) 1 .

through mutation, rearrangement, or gene amplification. The human PTK6 (also known as breast tumor kinase (Brk)) is a non-receptor protein-tyrosine kinase that was identified from a partial cDNA clone during a survey of PTK mRNAs in human melanocytes (1). It has been found that PTK6 is overexpressed in breast carcinomas (2) and colon tumors (3). Since the expression of PTK6 in human mammary epithelial cells resulted in an activation of the mitogenic response of the cells to epidermal growth factor (4), its expression serves as a probe of the tumor development process. PTK6 is considered to be most closely related to the Frk family proteins; however, it differs from Src family kinases both in lacking an N-terminal myristoylation site (5) and in having low sequence identity with members of this family. In addition, PTK6 showed an evolutionary divergence in the genomic structure (6). Although mouse Sik (a mouse homologue of PTK6) did not phosphorylate GTPaseactivating protein-associated p65 protein, a phosphotyrosine residue of GTPase-activating protein-associated p65 interacts with and activates Sik through its Src homology 2 (SH2) domain in mouse keratinocytes (7). Derry et al. (8) reported that PTK6 phosphorylated Sam68, a nuclear RNA-binding protein.
Recently a protein, BKS, encoded by cDNA that was isolated using a yeast two-hybrid method was found to contain an N-terminal pleckstrin homology-like domain connected to an SH2 domain, suggesting that BKS protein would be a strong candidate for phosphorylation mediated by the kinase domain of PTK6 (9).
Analysis of the primary structure of PTK6 showed that it is composed of SH3, SH2, and catalytic domains (5,6). For the case of the non-receptor protein-tyrosine kinases, the SH2 domain is typically involved in negative regulation of kinase activity by binding to a phosphorylated tyrosine residue near to the C terminus. A recent report suggested that the inactive form of Src was further stabilized by interactions between the SH3 domain and a polyproline linker near the kinase domain (10). The SH2 domain has been determined as an independent folded module involved in a number of intracellular signaling pathways ( Fig. 1) that binds to specific amino acid sequences in growth factor receptors and other molecules with phosphotyrosine residues. A number of reports have revealed that, in most cases, the SH2 domain adopts a common ␣/␤-folding motif with two binding pockets for the phosphotyrosine peptide (11)(12)(13)(14)(15). The structure of ligand peptide has also been determined as an extended or turnlike conformation, which enables the anchoring of its phosphotyrosyl group to the ligand binding pocket of the SH2 domain.
It was recently proposed that the C-terminal sequence of PTK6 (PTSpYENPT where pY is phosphotyrosine) might be a self-ligand for the SH2 domain. The solution structure of PTK6-SH2 revealed a distinct binding surface, providing evidence for an interdomain interaction of PTK6. Therefore, the structural data together with the binding mechanism of the ligand peptide derived from the C-terminal sequence of PTK6 would add to our understanding of the biochemical data and the biological activity of PTK6. In this report, we present the high resolution solution structure of the human PTK6-SH2 domain together with its ligand binding mode determined using heteronuclear multidimensional NMR spectroscopy and surface plasmon resonance analysis.

EXPERIMENTAL PROCEDURES
Protein Expression and NMR Sample Preparation-The cDNA segment encoding the PTK6-SH2 (residues 75-174) was amplified using a primer pair (5Ј-cGGgaTCcGAACCGTGGTTCTTTGGC-3Ј and 5Ј-ggaat-tcaCTCGTGCTTCCGGCAGG-3Ј where lowercase letters indicate nucleotides that were introduced to generate restriction enzyme sites and a termination codon) by PCR from the full-length PTK6 cDNA (6) and cloned into the BamHI-EcoRI sites of pGEX 4T-3 (Amersham Bio-sciences), an Escherichia coli expression vector. The recombinant protein contained a glutathione S-transferase (GST) tag encoded from the vector sequence in its N terminus. This vector was used to transform the E. coli strain BL21(DE3)pLysS for expression of the fusion protein. Uniformly 13 C/ 15 N-and 15 N-labeled protein samples were prepared by growing cells in M9 minimal media containing 15 NH 4 Cl, either with or without 13 C 6 -D-glucose. The purified GST-PTK6-SH2 was digested using bovine thrombin (Amersham Biosciences), and the final purification of PTK6-SH2 was performed using fast protein liquid chromatography with a Superdex 75 HR 10/30 column in 2 mM dithiothreitol, 50 mM potassium phosphate buffer at pH 6.5. The NMR samples were concentrated and transferred to a 5-mm symmetrical microcell (Shigemi). The protein-peptide complex solution was prepared by titrating Tyr(P) peptide into PTK6-SH2 with a final molar ratio of 1:2 (protein:peptide).
Preparation of the Tyr(P) Peptide Sample-The peptide (Ac-SSFTS-pYENPT-COOH) derived from the C-terminal region of PTK6 (residues 442-451) was synthesized commercially (Anygen Co., Kwangju, Korea) by solid-phase Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry (17). After trifluoroacetic acid cleavage, the peptide was eluted by analytical HPLC with the pH value of 3-4. The purity of synthetic peptide was confirmed by preparative HPLC and matrix-assisted laser desorption ionization time-of-flight mass spectrometry.
NMR Spectroscopy-All NMR experiments were performed at 298 K on a Bruker DRX500 or DRX600 or Varian Unity INOVA 500-MHz spectrometer equipped with a triple resonance probe and shielded triple axis gradients. The 1 H chemical shifts were referenced to internal sodium 4,4-dimethyl-4-silapentane-1-sulfonate. The 13 C and 15 N chemical shifts were referenced indirectly using the 1 (29,32,34), and HCACO (35), were collected for backbone resonance assignment. All NMR data were processed on a Silicon Graphics Indigo 2 work station using NMRPipe/NMRDraw software (Biosym/Molecular Simulations, Inc.). SPARKY 3.95 (37) and XEASY programs (38) were used for spectral analysis.
Structure Calculations-The solution structures were calculated by the distance geometry and dynamic simulated annealing (SA) procedures with the CNS 1.0 program on an Silicon Graphics Indigo 2 work station. Structural calculations were carried out following methods published previously (39 -41). The target functions for the molecular dynamics and energy minimization consisted of the covalent structure, van der Waals repulsion, NOE, and torsion angle terms. All restraints for the structure calculations were derived from the distance, torsion angle, and hydrogen bond information. Based on cross-peak intensities of NOESY spectra, distance constraints were classified as strong, medium, and weak corresponding to the upper distance bound of 2.7, 3.3, and 5.0 Å, respectively. Appropriate pseudoatom corrections were also applied to non-stereospecifically assigned methylene, methyl groups, and ring protons (42). Hydrogen bond constraints were used by defining two upper limit distances for each hydrogen bond: Angle constraints for the torsion angles were set to Ϫ120 Ϯ 50°for 3 J HNa Ͼ 8 Hz and Ϫ60 Ϯ 35°for 3 J HNa Ͻ 6 Hz, respectively. The 20 lowest energy structures (SA k ) from 50 simulated annealing conformers were selected for further structural analysis. None of the 20 SA k structures had any constraint violation greater than 0.5Å for distance and 3°for angle constraints. Structural analysis of the final 20 SA k structures was done using the MOLMOL (36) program.
Backbone Dynamics-NMR pulse sequences for backbone dynamic experiments were used as reported previously (43). To eliminate the effects of cross-correlation between 15 N-1 H dipolar and 15 N chemical shift anisotropy relaxation mechanisms, 1 H 180°pulses were inserted during the relaxation times (44 -46). Longitudinal (T 1 ) and transverse (T 2 ) relaxation data for the backbone amide 15 N nuclei were recorded as 2048 ϫ 64 data sets with 64 scans per point using 1 s of relaxation recovery time. Seven different relaxation times for T 1 and T 2 experiments were recorded with T 1 ϭ 5, 65, 145, 246, 366, 527, and 757 ms and T 2 ϭ 8.3, 25.1, 41.8, 58.6, 75.3, 108.8, and 142.3 ms, respectively. For the estimation of noise levels, duplicate spectra were recorded for values of T 1 ϭ 246 ms and T 2 ϭ 56.8 ms. The steady-state heteronuclear NOE (43, 47) experiment was obtained using a relaxation delay of 5 s. All data were processed on Silicon Graphics Indigo 2 work stations using the NMRPipe software. The standard errors of T 1 and T 2 values were calculated from the non-linear least square fit. The overall tumbling motion of the SH2 domain was calculated from the T 1 /T 2 ratios of N-H groups using the program Quadric Diffusion 1.11 (A. G. Palmer, Columbia University).
Surface Plasmon Resonance Analysis-Surface plasmon resonance measurements were performed by the BIACORE biosensor and BIAevaluation version 3.1 interactive software (Biocore) (48,49). GST-PTK6-SH2 was immobilized at a flow rate of 10 l/min using phosphate-buffered saline buffer solution (pH 6.5 with 2 mM dithiothreitol). Modeling of the PTK6-SH2⅐Tyr(P) Peptide Complex-The energyminimized average structure of PTK6-SH2 was used as a template to generate a model of the protein⅐peptide complex. The template structures of the peptides were derived from the p56lck-peptide (Protein Data Bank accession code 1LCJ) and Grb2-peptide (Protein Data Bank accession code 1QG1) structure. The initial structure of the PTK6-SH2⅐Tyr(P) complex was further optimized using consistent valence  force field on the DISCOVER3 module. To obtain the lowest energy conformation, restrained molecular dynamics was performed for 250 ps at 298 K. The lowest energy conformers were sampled every 100 steps during the molecular dynamic calculation. The restrained energy minimization procedure using the Broyden-Fletcher-Goldfard method was performed for the final structure until its energy gradient deviation decreased to 0.001 kcal⅐mol Ϫ1 .

Resonance Assignments and Secondary Structures-
The backbone resonance assignments were completed by data from HNCACB, CBCA(CO)NH, HNCO, and HCACO spectra. For the assignment of the residues preceding proline, HCACO data proved useful (50). Most of the side chain assignments were accomplished by three-dimensional HCCH-TOCSY and 15 Nedited TOCSY-HSQC experiments. The secondary structures were determined from the chemical shift indices (51), NOEs, 3 J HN␣ coupling constant, and slowly exchanging amide hydrogen information (Fig. 2). PTK6-SH2 is composed of two ␣-helices, six ␤-strands, and seven flexible loops.
Three-dimensional Structure of PTK6-SH2-A total of 1532 distance constraints, including 250 intraresidue, 374 sequential, 271 medium range, and 637 long range constraints, were derived from NOESY spectra. For 23 backbone hydrogen bonds, 46 corresponding interatomic distance constraints were used for the structure calculation. Fig. 3 summarizes backbonebackbone, side chain-side chain, and backbone-side chain NOEs. The structural statistics associated with the 20 SA k structures are listed in Table I four-stranded ␤-sheet (␤A, ␤B, ␤C, and ␤D), two ␣-helices (␣A and ␣B), and a small ␤-sheet (␤E and ␤F) near the second helix. The structure of PTK6-SH2 is relatively rigid, and a central ␤-barrel is well defined (Fig. 4). The topology of PTK6-SH2 is similar to that of other SH2 domains (52); however, it shows some differences from a canonical SH2 domain. A spatial rearrangement of the secondary structures especially in the central ␤-barrel region due to different length of ␤-strands was clearly observed (Fig. 5). In addition, the C-terminal ␤-strand (␤G), which is involved in not only ␤-barrel formation but also in ligand binding, is not observed in PTK6-SH2. This implies that PTK6-SH2 might have a unique binding feature for its specific ligand peptide.
Backbone Dynamics-Of the 96 backbone amide protons, the relaxation parameters for 81 residues have been reliably quantified. The experimental T 1 , T 2 , and steady-state 15 N-{ 1 H} heteronuclear NOE data are summarized in Fig. 6. The average values of T 1 and T 2 were determined to be 365.9 and 131.6 ms, respectively. The 15 N-{ 1 H} heteronuclear NOEs for most residues were observed to be 0.8 -1.0. The profile of heteronuclear NOEs shows that the secondary structural regions are relatively rigid. The overall rotational correlation time ( m ) determined from the T 1 and T 2 values and NMR structure is 6.5 ns (Fig. 7). The calculated order parameters (S 2 ) were in the range 0.7-0.95; however, small order parameters were found for the residues in the loops and the terminus of the secondary structural regions. The effective correlation time for internal motions, e , was determined as 55 ps. Interestingly most residues of the protein involved in Tyr(P) peptide binding exhibited smaller 15 N-{ 1 H} steady-state heteronuclear NOEs and order parameters as well as longer internal correlation times than those of residues in the structured regions, implying that the ligand binding site could be intrinsically dynamic in the absence of its ligand molecule (Figs. 6 and 7).
Chemical Shift Change upon Ligand Binding-The ligand binding pocket was predicted from the backbone dynamics together with sequence alignment data. To determine residues involved in ligand binding, a phosphorylated peptide derived from the C-terminal region (spanning residues 442-451) of PTK6 was used for NMR titration experiments. It has been observed that the autophosphorylated oligopeptide at Tyr-447 of PTK6, but not the unphosphorylated peptide, was successfully bound to PTK6-SH2 (16). A series of 1 H-15 N HSQC spectra of PTK6-SH2 were collected with different concentrations of phosphopeptide. Fig. 8A shows a superposition of the 1 H-15 N HSQC spectra of free PTK6-SH2 and peptide-bound PTK6-SH2 showing the chemical shift perturbations upon ligand binding more graphically. Most of the observed chemical shift perturbations occurred at residues involved in phosphopeptide binding. Chemical shift changes (⌬␦ tot ) between free SH2 and complexed SH2 were calculated using the equation, where ␦ i is the chemical shift of nucleus i, and W i denotes its weight factor (W HN ϭ 1, W N ϭ 0.2) (53). A single resonance peak was observed for each residue of PTK6-SH2 during titration, showing that the PTK6-SH2⅐Tyr(P) peptide is in fast exchange on the NMR time scale (Fig. 8A). The chemical shift perturbations upon Tyr(P) peptide binding were plotted as a function of residue number (Fig. 8B). The model structure of PTK6-SH2⅐Tyr(P) peptide agrees well with this data, showing that the peptide ligand interacts mainly with residues comprising a continuous patch of both hydrophobic and hydrophilic surfaces, which are ␣A (Arg-85 and Val-89), ␤B (Arg-105), BC loop (Glu-108), ␤D (His-126, Tyr-127, and Lys-128), DE loop (Ala-133), ␤⌭ (Leu-139), EF loop (Glu-141), ␣B (Tyr-154), and BG loop (Leu-160 and Leu-164) (Fig. 8C). Surface Plasmon Resonance Analysis-The binding affinity of the Tyr(P) peptide to the SH2 domain was measured by surface plasmon resonance spectroscopy using GST-PTK6-SH2. Fig. 9 illustrates a surface plasmon resonance sensorgram for the ligand peptide binding mode of PTK6-SH2. A rapid response curve indicates that the binding of peptide occurs with a fast association rate (k on ) and dissociation rate (k off ). Based on surface plasmon resonance data, an apparent dissociation constant (K d ϭ k on /k off ) for the interaction between PTK6-SH2 and Tyr(P) peptide is 60 M. Since the relative affinities for various SH2-Tyr(P) peptide interactions have been reported as K d values from 10 M to 0.1 nM, the binding affinity of Tyr(P) peptide to PTK6-SH2 is lower than those of previously reported phosphopeptide-SH2 binding (48). However, this weak binding is clearly supported by previous biological data showing that the enzyme activity of PTK6 increased about 1.4-fold with 1 mM Tyr(P) peptide, and no activity change has been observed within the micromolar range of peptide concentration (16). DISCUSSION Although the molecular topology of PTK6-SH2 is similar to that of the SH2 domains, the spatial arrangement of a central four-stranded antiparallel ␤-sheet (strands ␤A, ␤B, ␤C, and ␤D) clearly differs due to the missing ␤G strand, which is commonly found in most SH2 structures. The structural superposition with other SH2 domains of p56lck, Hck, and Blk clearly supports the differences; the pairwise backbone (N, C ␣ , and CЈ) root mean square deviations for these structures are 2.016 Å (p56lck), 2.196 Å (Hck), and 3.271 Å (Blk), respectively. Several notable structural differences were clearly observed in the ordered secondary structural regions, and high root mean square deviation values in the BC, EF, and BG loops between PTK6-SH2 and other SH2 domains were also observed. For example, a dramatic difference was found in loops; PTK6-SH2 has a shorter CD loop and longer AB loop than those of the canonical SH2 domains (Fig. 5A), suggesting that the structural rearrangement of the central ␤-barrel due to different loop length might enable its ligand specificity for PTK6 signaling.
A surface charge distribution of PTK6-SH2 calculated with the Delphi program suggests that a cluster of positively charged surface residues would be a ligand binding site for the negatively charged phosphotyrosine peptide, providing a strongly favorable electrostatic interaction between protein and ligand molecules. A continuous stretch of the ligand binding surface between BC and ␤G loop was also observed (Fig.  5B). Since it is well known that one of the major interactions between Tyr(P) peptide and SH2 domain is a hydrophobic interaction, it is expected that the hydrophobic patch of PTK6-SH2 should play a critical role in ligand binding. Fig. 5B clearly shows that the hydrophobic amino acids (Leu-139 and Leu-160) of PTK6-SH2 contact the proline residue of the Tyr(P) peptide, which is the Tyr(P) ϩ3 position of the ligand. The addition of Tyr(P) peptide to PTK6-SH2 also supports the structural data based on chemical shift perturbation of the residues involved in Tyr(P) peptide binding. For example, several charged residues in ␣A (Arg-85) and ␤D (His-126 and Lys-128) undergo large chemical shift perturbation upon ligand binding (Fig. 10). A significant change of the NMR resonance intensities was also observed for residues Val-89 (␣A) and Tyr-127 and Ile-129 (␤D). Most of the residues in the BC, EF, and BG loops experience chemical shift perturbations upon ligand binding.
Recently Qiu and Miller (16) suggested that the phosphorylation of Tyr-447 of PTK6 might play an important role in its negative regulation of the enzyme activity. They have shown that a single mutation at Tyr-447 increased SH2 domain accessibility to a synthetic ligand and resulted in 2.5-fold higher catalytic activity than that of the wild type. In this study, Tyr(P) peptide derived from the C-terminal region of PTK6 including Tyr(P)-447 successfully bound to the SH2 domain, supporting the autoinhibitory role of PTK6 protein. The apparent dissociation constant (K d ) between Tyr(P) peptide and PTK6-SH2 was determined as 60 M, which is much weaker than that of other SH2 domains. However, this might be ex-  10. Ligand binding pocket and a structure of Tyr(P) peptide⅐PTK6-SH2 complex generated by Insight II program (Biosym/Molecular Simulations, Inc.). A, protein surface representation of PTK6-SH2. The residues involved in Tyr(P) peptide binding are displayed. B, the structure of PTK6-SH2⅐Tyr(P) peptide complex. The structure was generated from solution structure of PTK6-SH2 (dark blue), and the ligand structure was derived from Tyr(P) peptide⅐p56lck-SH2 and Tyr(P) peptide⅐Grb2-SH2 complexes. plained by the fact that the local concentration of the phosphorylated C-terminal region of PTK6 is high in vivo because the actual binding will occur as an intramolecular interaction. Our data indicate that the structural motif of PTK6-SH2 would be of importance for specific recognition of ligand because the binding of phosphorylated C-terminal region to SH2 domain would be highly specific for the optimum autoregulation of PTK6 activity. In addition, it could be possible that the enzymatic activity of PTK6 would be tightly regulated through an autophosphorylation process at Tyr-447. Although a recent report suggests that PTK6 phosphorylates two other proteins, Sam68 and BKS in vivo (54), the downstream signaling of PTK6 is still not known in detail. Further structural study of PTK6 together with its interacting molecules will elucidate the signaling network and negative autoregulation process of this family of proteins at the atomic scale.