Novel Mechanism of Interaction of p85 Subunit of Phosphatidylinositol 3-Kinase and ErbB3 Receptor-derived Phosphotyrosyl Peptides*

Ligand-activated and tyrosine-phosphorylated ErbB3 receptor binds to the SH2 domain of the p85 subunit of phosphatidylinositol 3-kinase and initiates intracellular signaling. Here, we studied the interactions between the N- (N-SH2) and C- (C-SH2) terminal SH2 domains of the p85 subunit of the phosphatidylinositol 3-kinase and eight ErbB3 receptor-derived phosphotyrosyl peptides (P-peptides) by using molecular dynamics, free energy, and surface plasmon resonance (SPR) analyses. In SPR analysis, these P-peptides showed no binding to the C-SH2 domain, but P-peptides containing a phospho-YXXM or a non-phospho-YXXM motif did bind to the N-SH2 domain. The N-SH2 domain has two phosphotyrosine binding sites in its N- (N1) and C- (N2) terminal regions. Interestingly, we found that P-peptides of pY1180 and pY1241 favored to bind to the N2 site, although all other P-peptides showed favorable binding to the N1 site. Remarkably, two phosphotyrosines, pY1178 and pY1243, which are just 63 amino acids apart from the pY1241 and pY1180, respectively, showed favorable binding to the N1 site. These findings indicate a possibility that the pair of phosphotyrosines, pY1178-pY1241 or pY1243-pY1180, will fold into an appropriate configuration for binding to the N1 and N2 sites simultaneously. Our model structures of the cytoplasmic C-terminal domain of ErbB3 receptor also strongly supported the speculation. The calculated binding free energies between the N-SH2 domain and P-peptides showed excellent qualitative agreement with SPR data with a correlation coefficient of 0.91. The total electrostatic solvation energy between the N-SH2 domain and P-peptide was the dominant factor for its binding affinity.

Ligand-activated and tyrosine-phosphorylated ErbB3 receptor binds to the SH2 domain of the p85 subunit of phosphatidylinositol 3-kinase and initiates intracellular signaling. Here, we studied the interactions between the N-(N-SH2) and C-(C-SH2) terminal SH2 domains of the p85 subunit of the phosphatidylinositol 3-kinase and eight ErbB3 receptor-derived phosphotyrosyl peptides (P-peptides) by using molecular dynamics, free energy, and surface plasmon resonance (SPR) analyses. In SPR analysis, these P-peptides showed no binding to the C-SH2 domain, but P-peptides containing a phospho-YXXM or a non-phospho-YXXM motif did bind to the N-SH2 domain. The N-SH2 domain has two phosphotyrosine binding sites in its N-(N1) and C-(N2) terminal regions. Interestingly, we found that P-peptides of pY1180 and pY1241 favored to bind to the N2 site, although all other P-peptides showed favorable binding to the N1 site. Remarkably, two phosphotyrosines, pY1178 and pY1243, which are just 63 amino acids apart from the pY1241 and pY1180, respectively, showed favorable binding to the N1 site. These findings indicate a possibility that the pair of phosphotyrosines, pY1178-pY1241 or pY1243-pY1180, will fold into an appropriate configuration for binding to the N1 and N2 sites simultaneously. Our model structures of the cytoplasmic C-terminal domain of ErbB3 receptor also strongly supported the speculation. The calculated binding free energies between the N-SH2 domain and P-peptides showed excellent qualitative agreement with SPR data with a correlation coefficient of 0.91. The total electrostatic solvation energy between the N-SH2 domain and P-peptide was the dominant factor for its binding affinity.
Overexpression or mutation of ErbB receptors (ErbB1; epidermal growth factor receptor, ErbB2; Neu, ErbB3, and ErbB4) is implicated as a cause of various human cancers (1)(2)(3)(4). ErbB3 is a kinase-impaired receptor; nevertheless ligand binding to the receptor can cause heterodimerization with another ErbB receptor and induces activation and transphosphorylation of the ErbB3 receptor in cells (5)(6)(7). Once activated and tyrosinephosphorylated, the ErbB3 binds to the Src homology 2 (SH2) 1 domain of p85 (a regulatory subunit of phosphatidylinositol 3-kinase (PI3K)) at its binding sites within the C-terminal regulatory region. Six binding sites have been identified for p85 binding; accordingly the ErbB3 receptor is considered as a scaffold protein for PI3K (2, 4, 8 -9). Studies on insulin receptor substrate-1 (IRS-1) (10 -12), platelet-derived growth factor receptor (13), and ErbB receptors (4,8) revealed that the phospho-YXXM (pYXXM) motif is a specific binding site for SH2 domain of the p85, to which binding is necessary for activation of the p110 catalytic subunit of PI3K (12,14). On the other hand, the p85 is reported to bind to a pYVXV motif in the hepatocyte growth factor receptor (15), and another study also suggested that there was no binding specificity in any binding motifs for SH2 domains of p85, PLC␥, RasGAP, and Shc (16).
The p85 has a tandem SH2 region; two SH2 domains at N-terminal (N-SH2) and C-terminal (C-SH2) domains linked by the inter-SH2 (IS) region ( Fig. 1). Generally, the existence of both N-SH2 and C-SH2 domains induces high affinity interaction with phosphotyrosine and activation of PI3K; however, binding of individual SH2 domains to the phosphotyrosyl peptides is quite specific. Phosphotyrosyl peptides derived from IRS-1 (17) and the platelet-derived growth factor receptor (13) preferred binding to the p85 C-SH2, and the ones derived from the ErbB3 receptor preferred the N-SH2 (4,8). Furthermore, an NMR study of p85 N-SH2 and a doubly phosphorylated peptide derived from a polyomavirus middle T antigen (PvMT) complex revealed double phosphotyrosine binding in N-SH2 (18). Thus, the interaction mechanism of p85 SH2 and phosphotyrosyl peptides is quite complex. To study this complexity, we analyzed the interaction of ErbB3 receptor-derived phos-photyrosyl peptides and SH2 domains of p85 using SPR analysis and molecular dynamics (MD) simulation.
MD simulation has been acknowledged as a powerful method to investigate protein-protein, protein-ligand, and protein-DNA interactions at the atomic level. The molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) method has been recently proposed for calculating free energies of macromolecules in solution (19,20). This method combines MD simulations in explicit solvent with implicit solvation model, Poisson-Boltzmann (PB) analysis (21)(22)(23)(24), and solvent accessible surface area (25) for estimating binding free energies. Successful estimations of binding free energy using the MM-PBSA method have been reported in previous studies (26 -34).
In this study, we first investigated the interactions between the tandem SH2 (N-SH2, IS, and C-SH2 domains), N-SH2 and C-SH2 domains of the p85 subunit of the PI3K, and eight phosphotyrosyl peptides derived from the ErbB3 receptor (Ppeptides: pY1035, pY1178, pY1180, pY1241, pY1243, pY1257, pY1270, and pY1309) using SPR and cellular interaction analyses. Positive interactions observed for the N-SH2 domain and P-peptides were further analyzed by MM-PBSA and MD on a dedicated MD computer, the Molecular Dynamics Machine (MDM) (34 -39).

SPR Analysis
P-peptide Synthesis-Human ErbB3 receptor-derived P-peptides in which the N terminus was biotinylated were synthesized as previously described (34). The amino acid sequences of P-peptides are listed in Table I.
Interaction Analysis-BIAcore 3000 system (Biacore AB) was used to experimentally determine the kinetic constants of SH2-P-peptides interactions as previously described (34). The association rate constant (k a ) and the dissociation rate constant (k d ) were determined with a nonlinear least square method using BIAevaluation 3.0 software. K D was obtained by calculating k d /k a . Three independent measurements were performed for each set, and the average was obtained.

Western Blot Analysis
MCF-7 human breast cancer cells that endogenously express high amounts of ErbB3 receptor (40,41) were obtained from the American Type Culture Collection and routinely maintained in Dulbecco's modified Eagle's medium (Invitrogen, Life Technologies, Inc.) supplemented with 10% bovine calf serum. For heregulin ␤176 -246 (HRG, a ligand for the ErbB3 receptor) treatment, cells were serum-starved overnight and stimulated with or without 1 nM HRG (R&D Systems) for 5 min. MCF-7 cells were then washed with PBS and homogenized in cell lysis buffer containing 5 mM sodium phosphate (pH 7.5), 10 mM thioglycerol, 10% glycerol, 1 mM Na 3 VO 4 , and protease inhibitors. After centrifugation, clear supernatant was used for peptide binding analysis.
Biotinylated P-peptides and the corresponding non-phosphopeptides were dissolved in a buffer containing 50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM EDTA, 2 mM aprotinin, and 4 mM dithiothreitol at a final concentration of 1 M. 10 -15 l of streptavidin agarose slurry (Oncogene Research Product) was incubated in the above peptide solution (100 l) for 1 h at 4°C and washed three times with PBS. Each immobilized peptide bead was incubated with the MCF-7 cell lysate (1 mg each) for 3 h at 4°C. The beads were washed three times with PBS and applied to Western blot. Dissolved protein bands were detected with an anti-p85 antibody (Santa Cruz Biotechnology).

System Setup for MD Simulation
The solution structure of the p85 N-SH2 domain (amino acids 321-431)/P-peptide (derived from PvMT: amino acids 312-326) complex (Supplemental Fig. 1) was taken from the Protein Data Bank (PDB ID: 1FU5) (18). The p85 N-SH2 has two phosphotyrosine binding sites in its N-(N1) and C-terminal (N2) regions (Arg-340 and Lys-423, respectively) ( Fig. 1 and Supplemental Fig. 1). Eight complex structures of P-peptides (Table I) with the p85 N-SH2 used in this study were constructed based on the solution structure using MOE (Chemical Computing Group, Inc.). The modeling procedures were as follows: (i) amino acid residues were added, deleted, or replaced with appropriate residue manually; and (ii) the structure was energy-minimized in vacuum with an all-atom position fixing of the p85 N-SH2. As the initial structure for MD simulation of free p85 N-SH2, the p85 N-SH2 was extracted from the solution structure. Eight P-peptides were also extracted from each complexed structure as the initial structure for MD simulations of the free P-peptides. The N and C termini of the P-peptides were capped by acetyl and Nmethyl groups, respectively, for neutralization of the charge. For MD simulations, the systems of eight p85 N-SH2/P-peptide complexes (P-peptides bound with the N1 site; N1 simulation), another eight of those complexes (P-peptides bound with the N2 site; N2 simulation), a free p85 N-SH2, and eight free P-peptides were prepared (total of 25 systems). These systems were spherically surrounded by TIP3P water molecules (42). The size of the sphere was chosen so that the distance of atoms in the protein from the wall was longer than 15.0 Å. Then the fully solvated systems were energy-minimized.

Equilibration
All MD simulations were carried out using modified Amber 6.0 (43) for MDM on a PC (Intel Pentium4, 2.53 GHz) equipped with an MDGrape-2 board, which is a main component of the MDM (4 chips; 64 Gflops) (35). The parm96 force field (44) was adopted, and the time step was set at 1 fs. All non-bonded interactions, van der Waals, and Coulomb forces and energies were calculated using the MDM accurately. The bond lengths involving hydrogen atoms were constrained to equilibrium lengths using the SHAKE method (45). The temperature of each system was gradually heated to 300 K during the first 100 ps period, and additional 900 ps MD simulations were performed for data collection. The temperature was kept constant at 300 K using the method of Berendsen et al. (46) coupled to a temperature bath with coupling constants of 0.2 ps.

Free Energy Calculations
The production MD trajectory was collected for a 300 ps period (from 700 to 1,000 ps) with each snapshot saved every 10 ps. The set of structures obtained by an MD trajectory was sampled for use in estimating binding free energies. In the analysis of binding free energies,   pY1035  QSLLSPSSGpYMPMNQGNL  pY1178  TEEEDEDEEpYEYMNRRRR  pY1180  TEEEDEDEEYEpYMNRRRR  pY1241  PTAGTTPDEDpYEYMNRQR  pY1243  PTAGTTPDEDYEpYMNRQR  pY1257  RDGGGPGGDpYAAMGACPA  pY1270  GACPASEQGpYEEMRAFQG  pY1309 LEATDSAFDNPDpYWHSRLF the water molecules were replaced with implicit solvation models. The procedures for estimating the binding free energy have been described in detail in the previous report (34). Briefly, the Delphi program (47,48) was used to solve the PB equation. The PARSE (49) (van der Waals radii) and parm96 charges were used in the PB calculation, and the grid spacing used was 0.5 Å. The dielectric constants inside and outside the molecule were 1.0 and 80.0, respectively. Normal mode analysis was employed to estimate the conformational entropy (50).

Modeling of Cytoplasmic C-terminal Domain of ErbB3 Receptor Structure
To elucidate the structural model of the cytoplasmic regulatory Cterminal region of the ErbB3 receptor (amino acids 966 -1342), we performed the FORTE search (51). The FORTE is a method used to identify structural similarities between proteins through profile-profile comparisons. To build a backbone model from the alignment derived from the FORTE search, we used the AL2TS program 2 that converts sequence structure alignment into standard PDB format. All side chains were generated through the LEAP module of Amber 6.0, and the side-chains were energy-minimized with position fixing of all backbone atoms. The model structure of the ErbB3 receptor and p85 N-SH2 were docked manually to reproduce the N1-pY1178 and N2-pY1241, and N1-pY1243 and N2-pY1180 interactions, and the energy of the docking structures was minimized.

RESULTS AND DISCUSSION
SPR and Cellular Interaction Analyses-Binding of tandem SH2, N-SH2, and C-SH2 domains of p85 and the ErbB3 receptor-derived eight P-peptides were analyzed (Table II). Results showed specific binding of tandem SH2 to the P-peptides that include the pYXXM motif (pY1035, pY1178, pY1241, pY1257, and pY1270). None of the C-SH2, or SH3 and BCR domains of p85 showed detectable binding to P-peptides (data not shown). These results on tandem SH2 binding were consistent with the earlier study describing that pYXXM is a specific binding motif for p85 SH2 (4, 8, 10 -13). However, in our study, p85 N-SH2 bound to most of the P-peptides including pY1180 and pY1243, except for pY1309. The pY1309 has been known as a Shc binding site and pY1180 and pY1243 as growth factor receptorbinding protein 7 (Grb7) binding sites (9,53). The interaction analysis of the p85 with P-peptides and non-phosphopeptides in HRG-stimulated MCF-7 cells clearly supported the SPR data on tandem SH2 (Fig. 2). Our data suggested a difference in binding patterns between the p85 N-SH2 and the tandem SH2 or p85 subunit (SH3, BCR and tandem SH2). To understand the binding mechanism of p85 N-SH2 to the P-peptides, MD simulation and structure modeling were performed.
MD Simulations-We performed 25 series of 1-ns MD simulations for estimating binding free energies between the p85 N-SH2 and P-peptides. From the backbone heavy atom root mean square deviations (r.m.s.d.) of the p85 N-SH2 and Ppeptides from the initial structures (Supplemental Fig. 2), the structures of the p85 N-SH2 were stable through the period of 700 -1,000 ps in the N1, N2, and the free N-SH2 simulations (3.13 Ϯ 0.12, 3.05 Ϯ 0.11 and 2.26 Ϯ 0.09 Å, respectively). Because the initial structures of all P-peptides in our MD simulations were modeled, the r.m.s.d. of P-peptides in the N1, N2, and free P-peptide simulations greatly fluctuated (7.64 Ϯ 0.82 Å) (Supplemental Table I).
Binding Free Energies between p85 N-SH2 and P-peptides-To examine whether the calculated energies can evaluate binding free energies precisely, fluctuations of the calculated energies were examined. The average values and the standard deviations of calculated enthalpy H (sum of the internal energy of protein (E MM ), the electrostatic solvation energy (G PB ) and non-polar solvation energy (G ASA )) and entropic contribution TS were checked. Each standard deviation of the H of the p85 N-SH2/P-peptides complexes and the free p85 N-SH2 were smaller than 1.0%, and those of the free P-peptides were also smaller than 3.0% (Supplemental Table II). These results indicate that the calculated energies are stable during the simulations; thus, these values can be utilized for evaluating the binding free energy, although the r.m.s.d. of the p85 N-SH2/P-peptide complexes and the free P-peptides fluctuated to some extent (Supplemental Fig. 2). As a result of analysis, the calculated binding free energies on P-peptides except for pY1309 and p85 N-SH2 obtained from MM-PBSA method showed excellent qualitative agreement with those obtained from SPR analysis (correlation coefficient was 0.91) (Table III  and Fig. 3). Unexpectedly, the binding free energy of the p85 N-SH2/pY1309 complex obtained from N1 and N2 simulations showed no good agreement with that obtained from experimental analysis. The r.m.s.d. of the pY1309 extracted from the p85 N-SH2/pY1309 complex was unstable, increased monotonously, and greatly fluctuated from 700 to 1,000 ps (Supplemental Fig. 2 and Supplemental Table I). The root mean square fluctuations (r.m.s.f.) of phosphotyrosines in all P-peptides highly correlated with the binding free energies obtained from experimental analysis (correlation coefficient was 0.92). The r.m.s.f. of the pY1309 showed the largest value among all P-peptides tested (2.46 Å). In addition, residues around the phosphotyrosine, pY1309, (residues from pY-3 to pY) fluctuated more than others. From the theoretical study, K D is very sensitive to the fluctuations of molecules that form the complex; that is, a large fluctuation of molecules results in low binding affinity (54). Thus the inconsistency in the binding free energies obtained from computational and experimental analyses may be caused by these fluctuations of the residues around the pY1309.
Interestingly, the result of calculated binding free energies showed that pY1180 and pY1241 favored the N2 site in the p85 N-SH2, though all other P-peptides favored the N1 site. There are two tyrosines, Tyr-1178 and Tyr-1243, which are just 63 amino acids apart from pY1241 and pY1180, respectively, and these two tyrosines showed favorable binding to the N1 site when they are phosphorylated. These findings indicate the possibility that the pair of phosphotyrosines, pY1178-pY1241 or pY1243-pY1180, will fold into appropriate configuration for binding to N1 and N2 sites at once. Our model structures of the cytoplasmic C-terminal domain of the ErbB3 receptor strongly supported this speculation (Fig. 4).
Component Analysis of Binding Free Energies-Next, the dominant factor for binding affinity (with the exception of pY1309) was identified by component analysis of the binding free energies. The regression coefficients between the calculated binding free energies and its components indicate contribution of the energetic component to the binding affinity (Table IV). Since the absolute values of the regression coefficients of the electrostatic energy ⌬E ele (regression coefficient of Ϫ5.26) and the electrostatic solvation energy ⌬G PB (regression efficient of 6.25) were larger than that of the other components, these components seem to greatly contribute to binding affinity.
The ⌬G PB and experimental binding free energy ⌬G exp showed a strong positive correlation coefficient of 0.80. In contrast, the ⌬E ele and the ⌬G exp showed negative correlations of Ϫ0.79. Interestingly, total electrostatic solvation energy ⌬G ele , which is the sum of the ⌬E ele and the ⌬G PB and the ⌬G exp showed the highest correlation among all energy components (correlation coefficient of 0.84). This result indicates that the ⌬G PB and the ⌬E ele compensate for each other, and the ⌬G ele between the p85 N-SH2 and the P-peptides is the dominant factor for binding affinity. To clarify the reason for the negative correlation between the ⌬E ele and the ⌬G exp , the hydrogen bonds between p85 N-SH2 and the P-peptides were investigated. The number of hydrogen bonds and the ⌬G exp showed high correlation (correlation coefficient of 0.82), and these results indicate that a lesser number of hydrogen bonds induces higher binding affinity (Supplemental Fig. 3). This fact also explains the strong negative correlation between ⌬E ele and the FIG. 2. Interaction of p85 and P-peptides in MCF-7 cells. Synthetic biotinylated P-peptides were immobilized with streptavidin beads and incubated with HRG stimulated or non-stimulated MCF-7 cell lysates (A). P-peptides (pY) or the corresponding non-phosphopeptides (Y) were also incubated with HRG-treated MCF-7 cell lysate (B) and quantified by densitometer. Specific binding of p85 with P-peptides was obtained by subtracting non-phosphopeptide binding (C). Note that pY1180 and pY1243 share the same non-phosphopeptide sequences of Tyr-1178 and Tyr-1241, respectively. Peptide-protein complexes were then analyzed by Western blot. PI3K p85 was detected with a specific antibody against p85. ⌬G exp (Table IV). This means that loss of hydrogen-bonded energy between p85 N-SH2 and P-peptides are compensated by the solvation of the p85 N-SH2 and P-peptides. CONCLUSION Binding of p85 to the tyrosine-phosphorylated proteins is an essential step for the activation of the p110 catalytic subunit of PI3K. Our SPR analysis showed that p85 N-SH2 mainly contributes to the interaction with the ErbB3 receptor. From the computational analysis, we found that pY1180 and pY1241 favored the N2 site in the p85 N-SH2 for binding, though all other P-peptides favored the N1 site. There are two tyrosines, Tyr-1178 and Tyr-1243, just 63 amino acids apart from the pY1241 and pY1180, respectively. These tyrosines are also targets for phosphorylation and these P-peptides (pY1178 and pY1243) favored the N1 site for binding. The model structure of the cytoplasmic C-terminal region of the ErbB3 receptor indicates that the pair of phosphotyrosines, pY1178-pY1241 or pY1243-pY1180, will act for simultaneous binding of "double phosphotyrosines." Actually, pY1180 and pY1243 have been recognized as Grb7 binding sites (53). However, those phosphopeptides corresponding to Grb7 binding sites showed a slight rise in PI3K activity (14) or an affinity for p85 (9) in earlier studies, indicating that the non-pYXXM motif is also capable of binding to p85 SH2. Our SPR and computational analyses suggest that the amino acid sequence is not the sole determinant for specificity of the phosphotyrosyl peptide-SH2 interaction.
In general, binding to double phosphotyrosines enhances interaction and enzymatic activity of PI3K than binding to a single phosphotyrosine, and bivalent bindings of p85 at two tyrosine residues have been reported at 13 amino acids apart (pY1257 and pY1270) for the ErbB3 receptor, 7 amino acids apart for PvMT (18), and 20 amino acids apart for IRS-1 (12). Optimal amino acid distance for the p85 bivalent binding is not known. However, since different growth factor ligands induce discriminative phosphorylation patterns at multiple tyrosine sites within the ErbB receptor (55), those binding sites and patterns may differ in response to the various kinds of stimuli, and double phosphotyrosines at a greater distance may facilitate stabilizing the association with p85 at N1 and N2 sites.
Our experimental and computational analyses showed that the p85 N-SH2 domain itself is capable of binding to the non-pYXXM motif; however, the tandem SH2 and cellular p85 subunit did not bind to the same peptides. We estimated that the steric constraint of the p85 structure might alter the binding property. We made a homology model of p85 tandem SH2 based on the structure of ZAP-70 tandem SH2 (56); however, such constraint was not observed for P-peptide binding (data not shown). For another reason, the presence of the IS domain in p85 raises the binding affinity of the SH2 domain to that of ErbB3-derived phosphopeptides (8), and conformation of the IS domain is changed after N-SH2 binding to phosphopeptide (52). Accordingly the IS domain might induce a conformational change and contribute to the specificity of p85 binding to the pYXXM motif.
As for MD simulation, the calculated binding free energies showed excellent qualitative agreement with experimental data with the correlation coefficient of 0.91, although they showed no quantitative agreement with experimental data. The binding free energies obtained from computational analysis were smaller than those obtained from experiment. This fact may be caused by the short time scale of our MD simulations for equilibration of the small peptide (34). However, our present study shows that the current method for estimating binding free energy is effective for in silico screening.
The electrostatic solvation contribution and electrostatic interaction compensate each other, and the sum of these compo-  nents, total electrostatic solvation energies ⌬G ele , is the dominant factor in binding affinity. In the case of the interaction between the SH2 domain of Grb2 and P-peptides derived from ErbB1 and ErbB4 receptors, the dominant factor for its binding affinity was the van der Waals interaction (34). The difference between two dominant factors for binding affinity indicates that the Grb2 recognizes the "molecular shape" of the ErbB1 and ErbB4 receptor, and the binding of the p85 subunit of PI3K and the ErbB3 receptor is controlled by an "electrostatic field," and may be very important for receptor-adaptor recognition.