Targeting Self-Binding Peptides as a Novel Strategy To Regulate

Mar 27, 2017 - SH2 domain in the crystal structure and then the SIP-CT peptide was ... heating, and initially equilibrated using a NF5280M4 server equ...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/jcim

Targeting Self-Binding Peptides as a Novel Strategy To Regulate Protein Activity and Function: A Case Study on the Proto-oncogene Tyrosine Protein Kinase c‑Src Zhengya Bai,† Shasha Hou,† Shilei Zhang,† Zhongyan Li,† and Peng Zhou*,†,‡,§ †

Center for Informational Biology, School of Life Science and Technology, ‡Center for Information in BioMedicine, and §Key Laboratory for Neuroinformation of the Ministry of Education, University of Electronic Science and Technology of China (UESTC), Chengdu 610054, China ABSTRACT: Previously, we have reported a new biomolecular phenomenon spanning between protein folding and binding, termed as self-binding peptides (SBPs), where a short peptide segment in monomeric protein functions as a molecular switch by dynamically binding to/unbinding from its cognate domain in the monomer (Yang et al. J. Chem. Inf. Model. 2015, 55, 329−342). Here, we attempt to raise the SBP as a new class of druggable targets to regulate the biological activity and function of proteins. A case study was performed on the proto-oncogene nonreceptor tyrosine kinase, c-Src, which contains two SBPs that bind separately to SH3 and SH2 domains of the kinase. State-of-the-art molecular dynamics (MD) simulations and post binding energetics analysis revealed that disrupting the kinase-intramolecular interactions of SH3 and SH2 domains with their cognate SBP ligands can result in totally different effects on the structural dynamics of c-Src kinase architecture; targeting the SH2 domain unlocks the autoinhibitory form of the kinasethis is very similar to the pTyr527 dephosphorylation that functionally activates the kinase, whereas targeting the SH3 domain can only release the domain from the tightly packed kinase but has a moderate effect on the kinase activity. Subsequently, based on the cognate SBP sequence we computationally designed a number of SH2-binding phosphopeptides using a motif grafting strategy. Fluorescence polarization (FP) assay observed that most of the designed phosphopeptides have higher binding affinity to SH2 domain as compared to the native SBP segment (Kd = 53 nM). Kinase assay identified a typical dose−response relationship of phosphopeptides against kinase activation, substantiating that disruption of SH2-SBP interaction can mimic c-Src dephosphorylation and activate the kinase. Two rationally designed phosphopeptides, namely EPQpYEEIEN and EPQpYEELEN, were determined as strong binders of SH2 domain (Kd = 8.3 and 15 nM, respectively) and potent activators of c-Src kinase (EC50 = 3.2 and 41 μM, respectively).

1. INTRODUCTION Proteins are very important molecular entities in cells. They are involved in virtually all cell functions. Targeting functional proteins has long been established as an efficient approach to studying diverse biological phenomena and, more importantly, development of therapeutic strategies against various diseases. Proteins can be targeted using a variety of molecular ligands, including small organic compounds, peptide regulators, nucleic acid aptamers, and protein partners. A region to which the ligands recognize and bind in target protein is usually a small cavity or narrow cleft such as enzyme active site,1 but sometimes the region consists of a large biomolecular interface of protein−protein or protein−nucleic acid complex,2,3 which is normally difficult to perturb chemically and biologically, thus making it unfavorable for designing ligand mediators to regulate particular biological processes.4 Over the past decade, peptide-mediated interactions (PMIs) have been recognized as attractive druggable targets due to their weak and transient nature.5,6 PMIs are commonly involved in many important protein−protein interactions, particularly those that are transient, low-affinity, or related to post-translational modification events like phosphorylation, © XXXX American Chemical Society

which are mediated by the binding of a globular domain in one protein to a short peptide stretch in another.7 Similar to PMIs, we have recently described a novel sort of peptide interactions, termed as self-binding peptides (SBPs),8,9 which represent those short peptide segments in monomeric proteins that perform biological functions by dynamically binding to/ unbinding from their cognate domains in the same monomer. In an SBP system, the peptide segment, on the one hand, specifically recognizes and interacts with its cognate target to establish a dynamic balance between the bound adduct and unbound state; on the other hand, the segment is integrated to the target in primary sequence via a flexible polypeptide linker (Figure 1). A typical SBP should have three features:8 (i) the SBP is connected to its target in primary sequence and binds independently to the target in advanced structure, (ii) the SBP−target recognition is highly specific so that the SBP can regulate specific biological function, and (iii) the SBP can bind to/unbind from its target under a controllable mechanism such as membrane voltage change, phosphorylation, or ligand Received: November 2, 2016 Published: March 27, 2017 A

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

Figure 1. Schematic representation of a self-binding peptide (SBP) system. A monomeric protein contains three functional domains as well as an SBP which is connected to domain 3 via a flexible linker. The protein function is regulated through the SBP switching between bound and unbound states with domain 2, where the specific binding site of SBP can be found.

complex can further induce PPII binding to SH3 domain and, consequently, the catalytic domain of TK is locked in an autoinhibitory form.12 Dephosphorylation of pTyr527 activates the kinase system by destabilizing SH2−CT and then SH3− PPII interactions.

regulation. An SBP can locate at either terminus (as shown in Figure 1) or middle region of a protein. In previous works, we showed that SBPs are structurally and energetically almost a binding phenomenon, although they are traditionally considered as folding owing to their sequence connectivity with target.8 We also proposed a two-step binding model for SBP− domain recognition to successfully explain the binding dynamics of the protein−intramolecular interaction.9 Considering the important functionality of SBPs in their parent proteins and, like PMIs, the weak and transient nature that make SBPs vulnerable to be targeted by chemical drugs and biological agents, we herein proposed the SBPs as a new and promising class of druggable targets to which ligand mediators can recognize and bind to regulate the biological activity and f unction of parent proteins. In this study, a typical SBP-containing protein system, c-Src, was adopted as a paradigm to illustrate ligand regulation by targeting SBPs. Human c-Src is a nonreceptor tyrosine kinase encoded by the protooncogene SRC, which has been closely associated with a variety of malignant solid tumors such as colon, liver, lung, breast, and pancreas.10,11 As shown in Figure 2, c-Src is a monomeric protein consisting of two

2. MATERIALS AND METHODS 2.1. c-Src Structure Models Used. Four structure models (one crystal structure plus three artificial structures modified from the crystal structure) were separately used to represent four different conditions of c-Src protein: (a) Model 1 is the c-Src in inactive form, where the Tyr527 residue is phosphorylated (pTyr527) and the CT and PPII are bound to SH2 and SH3 domains, respectively; the c-Src holds in a tightly locked conformation (Figure 3A). The model is represented by crystal structure of inactive c-Src retrieved from the PDB database (PDB: 2SRC).13 (b) Model 2 is a dephosphorylated version of model 1; we manually removed the phosphate moiety from pTyr527 residue in the crystal structure to obtain its dephosphorylated counterpart (Figure 3B). (c) Model 3 describes the CT-binding site of c-Src SH2 domain attacked by a self-inhibitory phosphopeptide14 which mimics the 13-mer core sequence of phosphorylated CT segment (SIP-CT peptide, TSTEPQpYQPGENL). Here, the CT was manually moved away from SH2 domain in the crystal structure and then the SIP-CT peptide was modeled to the CT-binding pocket of the domain (Figure 3C), that is, the SH2-bound conformation of phosphorylated CT segment in crystal structure was adopted as a counterpart of the binding mode of SIP-CT peptide. This is reasonable if considering that the SH2−peptide binding system is basically independent of rest of the kinase and thus splitting the CT segment from the kinase as a peptide would not influence its binding mode significantly, and model 3 is just an initial form that would be further subjected to long-term MD simulations for relaxing and refining the domain−peptide system. (d) Model 4 denotes the PPII-binding site of c-Src SH3 domain occupied by a self-inhibitory peptide that mimics the 10-mer core sequence of PPII segment (SIP-PPII peptide, TSKPQTQGLA). Here, the SH3 domain was manually moved away from PPII in the crystal structure, and then, the SIP-PPII peptide was modeled as a counterpart of crystallographic PPII segment to complex with SH3 domain (Figure 3D). 2.2. Molecular Dynamics Simulation and Binding Free Energy Analysis. The phosphate parameters for phosphotyr-

Figure 2. Three-dimensional crystal structure of human c-Src nonreceptor tyrosine kinase. c-Src is a monomeric protein consisting of three domains (a tyrosine kinase domain TK and two peptiderecognition domains SH3 and SH2) and two SBPs (the first SBP polyproline-II helical peptide [PPII]; the second SBP C-terminal [CT] tail). Phosphorylation of Tyr527 residue (pTyr527) forces CT binding to SH2 domain, which further induces SH3−PPII interaction. In this way, the catalytic domain of TK is locked in an autoinhibitory form.

peptide-recognition domains SH3 and SH2 as well as a tyrosine kinase domain TK. The protein also contains two SBPs: one is a polyproline-II helical peptide (PPII) between the SH2 and TK domains (the f irst SBP), and another is the C-terminal (CT) tail which complexes with SH2 domain once its Tyr527 residue is phosphorylated (pTyr527) (the second SBP). The B

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

Figure 3. Four c-Src models used to perform MD investigations. (A) c-Src in inactive form: the Tyr527 residue in CT is phosphorylated (pTyr527) to promote the intramolecular interactions of SH2 and SH3 domains with self-binding peptides CT and PPII, respectively. This model is represented by crystal structure (PDB: 2SRC). (B) Dephosphorylated c-Src: the phosphate moiety is manually removed from pTyr527 residue in the crystal structure to make a dephosphorylated counterpart. (C) Targeting CT-binding site of SH2 domain: the phosphorylated CT is manually moved away from SH2 domain in the crystal structure, and then a SIP-CT peptide mimicking the phosphorylated CT is modeled as a counterpart of crystallographic phosphorylated CT segment to the CT-binding pocket of SH2 domain. (D) Targeting PPII-binding site of SH3 domain: the SH3 domain is manually moved away from PPII in the crystal structure, and then a SIP-PPII peptide mimicking the PPII is modeled as a counterpart of crystallographic PPII segment to the PPII-binding site of SH3 domain.

were performed using AMBER ff03 force field.19 The simulations were first submitted to energy minimization, heating, and initially equilibrated using a NF5280M4 server equipped with two Intel Xeon E5−2620 v2 CPUs (SANDER in AMBER11 package). Subsequently, the simulations were transferred to and accelerated by NVIDIA Tesla K20 GPU SimCluster for further equilibrium and production (the GPU version of PMEMD in AMBER14 package). Considering that the numerical results obtained by hybrid single/double precision (SPDP) model are comparable with the double precision DPDP model but at significantly reduced computational cost,20 the SPDP was used in the GPU implementation of PMEMD, which combined single precision for calculation and double precision for accumulation. The binding free energy ΔG of peptides to c-Src domains was calculated based on the structure snapshots extracted over the MD equilibrium phase using molecular mechanics Poisson−Boltzmann surface area (MM-PBSA) method:21

osine were derived from the AMBER parameter database (http://research.bmh.manchester.ac.uk/bryce/amber/). A truncated octahedral box of TIP3P water molecules15 was added with a 12 Å buffer around the c-Src structures. Counterions of Na+ were placed based on the Coulombic potential to keep the whole systems electroneutral. Consequently, the four c-Src systems investigated in this work contain ∼7000 atoms for nonsolvent components and ∼20 000 water molecules (in total ∼70 000 atoms). The steepest descent and conjugate gradient algorithm energy minimizations were in turn conducted to remove bad contacts between c-Src and water molecules. After the minimizations, systems were heated to 300 K over 500 ps followed by constant temperature equilibration at 300 K for 1 ns. Subsequently, MD simulations (0.5 to 10 μs time scale) were carried out in an isothermal isobaric ensemble with periodic boundary conditions, during the last production phase (equilibrium simulations) a total of 500 snapshots were evenly saved for subsequent energetic analysis. The particle mesh Ewald (PME) method16 was employed to analyze the long-range electrostatic energy of a unit cell in a macroscopic lattice of repeating images and a cutoff distance of 10 Å was considered to calculate the shortrange electrostatics and van der Waals interactions. The SHAKE strategy17 was used to restrain all covalent bonds involving hydrogen atoms, and an integration step of 2 fs was applied. Each simulation was coupled to a 300 K thermal bath at 1.0 atm through the Langevin algorithm.18 All simulations

ΔG = ΔE int + ΔGslv − T ΔS = (ΔEelc + ΔEvdW ) + (ΔGplr + ΔGnplr) − T ΔS (1)

where ΔEint is the nonbonded interaction energy between domain and peptide, which can be divided into electrostatic (ΔEelc) and van der Waals (ΔEvdW) potentials and was calculated with molecular mechanics (MM) approach. ΔGslv is solvent effect associated with the interaction, which is C

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

Figure 4. Average structures of four c-Src model systems derived over the MD production phase. (A) Model 1: the inactive c-Src with phosphorylation of Tyr527 residue. The c-Src structure is in tightly locked, autoinhibitory state. (B) Model 2: the active c-Src with dephosphorylation of Tyr527 residue. The SH2-SH3 diad moves away from TK domain to form a splitting conformation of c-Src. (C) Model 3: targeting CT-binding site of SH2 domain with self-inhibitory peptide SIP-CT. Similar to the model 2, the c-Src shows a splitting conformation. (D) Model 4: targeting PPII-binding site of SH3 domain with self-inhibitory peptide SIP-PPII. The SH3 domain separates from SH2 and TK domains. D

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

denatured enolase, 5 μM ATP and 5 μCi [γ-32P]ATP. The mix was incubated with phosphotyrosyl peptides at different concentrations on ice for 30 min followed by incubation at 30 °C for 1 h. Reaction was stopped by the addition of 20 μL 2× SDS loading buffer and boiling for 5 min. The mixture was then briefly centrifuged and the supernatant was resolved on 10% SDS-PAGE gel. Phosphorylated enolase was detected by autoradiography and plotted against the concentration of peptides to determine activating potency EC50 for the peptides against c-Src.

contributed from polar (ΔGplr) and nonpolar (ΔGnplr) desolvations; the polar aspect was calculated by numerical solution of the nonlinear Poisson−Boltzmann (PB) equation,21 while the nonpolar facet was described using surface model (SA) as ΔGnplr = β + γA, where A is solvent accessible surface area (SASA) computed with a solvent-probe radius of 1.4 Å, and β and γ are coefficient terms that were assigned to 0.92 and 0.00542 kcal/(mol·Å2), respectively, as suggested by Kollman and co-workers.22 The grid size for the PB calculations was set to 0.5 Å, and the interior and exterior dielectric constants were 2 and 78, respectively. −TΔS is the conformational free energy due to entropy penalty of the binding, which was computed by normal-mode analysis (NMA).23 Frequencies of the vibrational modes were computed at 300 K for structure snapshots and using a harmonic approximation of the energies. Considering the high computational demand, only 50 snapshots for each system were utilized to estimate the −TΔS. Here, the MMPBSA and NMA calculations were implemented using mmpbsa and nmode programs, respectively. 2.3. Experimental Section. Peptide Synthesis and Protein Expression. The peptide TSTEPQYQPGENL and phosphopeptides TSTEPQpYQPGENL, EPQpYQPGEN, EPQpYEEIEN, EPQpYNELEN, EPQpYEELEN, and EPQpYEDLEN were synthesized and purified commercially. Glutathione S-transferase (GST)-tagged protein of human cSrc SH2 domain (residues 141−250) was cloned into a pETllb expression vector and expressed in Escherichia coli BL21(DE3) cells. Cell lysis was in a buffer containing 20 mM HEPES (pH 7.5), 5 mM DTT, 5 μM APMSF, 10 μM pepstatin, 10 μM leupeptin, and 1 μM aprotinin.24 The expressed fusion proteins were isolated by a procedure described elsewhere.25 The fulllength, Tyr527-phosphorylated c-Src protein was supplied in 50 mM Tris-HCl, pH 7.5, with 150 mM NaCl, 0.25 mM DTT, 0.1 mM EGTA, 0.1 mM EDTA, 0.1 mM PMSF, and 25% glycerol. Fluorescence Polarization (FP) Assay. The FP assays were performed using a protocol reported previously.26,27 Briefly, the synthetic phosphopeptides were labeled with N-terminally conjugated fluorescein rhodamine and diluted to a concentration of 10 nM. Titrations were performed by monitoring the FP as a function of increasing amounts of SH2 proteins (0.1− 1000 nM) added to the rhodamine-labeled peptides in an assay buffer 20 mM HEPES (pH 7.8), 100 mM KCl, 0.1% Tween-20 and 5 mM DTT. The FP was measured using PerkinElmer LS55 luminescence spectrometer at 25 °C with excitation and emission wavelengths of 530 and 555 nm, respectively. The FP value with no SH2 domain present was used as a background. Each assay was read three times, and the values were averaged prior to analysis. Raw fluorescence intensity was also read at the same wavelengths for each assay. The dissociation constants (Kd) were determined by fitting titration curves to the equation: FP =

FP0 + FPmax ([SH2]/Kd) 1 + ([SH2]/Kd)

3. RESULTS AND DISCUSSION 3.1. Structural, Energetic, and Dynamic Analysis of Targeting c-Src SBP Sites. The four c-Src model systems were separately subjected to long-term MD simulations to reconstruct their dynamics trajectories from initial artificial structure to the final equilibrium state. Time scales were set to 0.5, 10, 10, and 5 μs for the simulations of model systems 1, 2, 3, and 4, respectively; the settings were carefully selected to ensure the four systems to reach at equilibrium. Here, average structures for the four model systems were derived over MD production phase and shown in Figure 4. Model 1. It is known that phosphorylation of Tyr527 residue in c-Src would promote the second SBP binding of CT to SH2 domain, which then induces the first SBP interaction between SH3 domain and PPII.12 This can be well reflected in the dynamics equilibrium structure of model 1 (Figure 4A), where, as expected, the CT and PPII are properly bound to the SBPbinding sites of SH2 and SH3 domains, respectively, resulting in a tightly locked, autoinhibitory state of c-Src kinase. Since the final equilibrium conformation of model 1 is very close to its initial crystallographic form, the 0.5-μs MD simulations are sufficient to reconstruct the complete dynamics trajectory of cSrc converting from the initial crystal structure to final equilibrium state. Model 2. Dephosphorylation of pTyr527 residue would cause a dramatic conformational change in the kinase architecture. As shown in Figure 4B, after 10-μs MD simulations the CT and PPII are completely dissociated from their cognate binding sites in SH2 and SH3 domains, respectively, thus unlocking the kinase. Visual examination of the MD trajectory revealed two dynamics steps of the c-Src conformational change, that is, first, the CT dissociates from SH2 domain at ∼1 μs and then the PPII slowly moves away from SH3 domain after ∼3 μs, suggesting that the pTyr527 dephosphorylation can induce a cascade effect of destabilizing the two SBP interactions, that is, (i) the SH2−CT dissociation, and (ii) then the SH3−PPII dissociation. It is worth noting that the two successive phases are not distinct; they are largely overlapping to each other. This is consistent with previous reports that Tyr527 phosphorylation promotes SH2-CT binding, which then induces SH3-PPII interaction to lock the kinase in an autoinhibitory form.12 In addition, the linker regions between TK and SH2 domains and between SH2 and SH3 domains are highly flexible and thus the whole kinase system exhibits a large conformational fluctuation during the equilibrium phase of model 2. Model 3. In order to disrupt the native SBP interaction between SH2 domain and its cognate CT ligand, a 13-mer selfinhibitory peptide SIP-CT (TSTEPQpYQPGENL) derived from the core sequence of CT was employed to target the SH2 domain, where pTyr527 is kept in phosphorylation state. The whole c-Src kinase system in complex with SIP-CT was

(2)

where [SH2] is the protein concentration, FP is the observed value at the given protein concentration, FP0 is the polarization value of free peptide, and FPmax is the maximal polarization value saturated with protein. In vitro Kinase Assay. The full-length, Tyr527-phosphorylated c-Src protein were diluted in 20 μL kinase assay buffer containing 20 mM HEPES, 1 mM MnCl2, 0.2 mM sodium orthovanadate, and 1 mM DTT as well as 2.5 μg substrate acidE

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

Figure 5. (A) Superposition of the activation loops (A-loops) of four models in MD equilibrium phase. The A-loops of model 1 (Tyr527 phosphorylation) and model 4 (targeting PPII) are in DFG-out state that inactivates the kinase, whereas model 2 (pTyr527 dephosphorylation) and model 3 (targeting CT) are in DFG-in state that activates the kinase. The A-loops of models 2 and 3 are unstructured, while the loops of models 1 and 4 are highly and partially structured, respectively. The Tyr416 residue of models 2 and 3 is exposed to solvent that can be readily phosphorylated. In contrast, this residue of model 1 is fully buried that prevents it from phosphorylation, while model 2 is partially buried. (B) Superposition of the αC-helices of four models in MD equilibrium phase. The αC-helices of models 1 and 4 are displaced from the activated state as that of models 2 and 3. The Glu310 residues of models 2 and 3 points to the active site of the kinase, while the residues of models 1 and 4 are removed from the active site.

are in active DFG-in (Figure 5A). The A-loops of models 2 and 3 are intrinsically disorder (helical rate = 17.2% and 20.7%, respectively), while the loop regions of models 1 and 4 are highly and partially structured, respectively (helical rate = 51.7% and 36.4%, respectively) (Table 1). The Tyr416 residue

subjected to 10-μs MD simulations, and resulting equilibrium structure is shown in Figure 4C. It is seen in the structure that the TK domain is separated by a large distance (>10 Å) from the SH2-SH3 diad; this is just like that in model 2. However, the SH3 domain seems not to fully strip from the PPII region, although the SH3-PPII interaction strength is impaired substantially, given the considerable distance fluctuation between SH3 and PPII during the equilibrium simulations. In this respect, it is suggested that targeting the CT-binding site of SH2 domain using a peptide ligand could induce a similar effect with pTyr527 dephosphorylation on the global structure of cSrc kinase; both of them block the binding of CT to SH2 domain and destabilize the SH3-PPII interaction. Thus, targeting the SH2 domain, like pTyr527 dephosphorylation, is considered as a potential approach to activating the kinase. Model 4. We also performed 5-μs MD simulations of targeting the PPII-binding site of SH3 domain with a 10-mer self-inhibitory peptide SIP-PPII (TSKPQTQGLA) derived from the core sequence of PPII. As can be seen in Figure 4D, the peptide can disrupt the cognate SBP interaction between SH3 domain and PPII, and then induce the SH3 separating from rest of the kinase. However, the phosphorylated CT tail seems to keep in bound state with SH2 domain, and no obvious loosing of the SH2-CT adduct can be observed during the whole simulation procedure. Therefore, the SH2/ PPII/TK/CT section of the kinase still keeps in packing state when the SH3 domain is targeted by SIP-PPII peptide, where only the SH3 moves away from the kinase ensemble upon the peptide binding, suggesting that the kinase cannot be fully activated by using SIP-PPII to target SH3 domain. Next, we visually examined the local structures of activation loop (A-loop) and αC-helix of c-Src kinase in the four model systems. The A-loop is close to kinase’s active site and related directly to catalytic activity of the kinase, which contains a DFG motif that can flip between two conformational states, i.e. active DFG-in and inactive DFG-out.28 In addition, the A-loop possesses a phosphorylable Tyr416 that plays a key role in regulation of the loop conformation by its phosphorylation and dephosphorylation.29 Superposition between the four models of dynamics equilibrium phase revealed that the A-loops of model 1 (Tyr527 phosphorylation) and model 4 (targeting the first SBP) hold in inactive DFG-out, whereas the model 2 (pTyr527 dephosphorylation) and model 3 (targeting the second SBP)

Table 1. Summary of the Four Model Systems in MD Equilibrium Phase binding free energy ΔG (kcal/mol)

model model model model model

1 2 3 4

helical rate of A-loopa

SASA of Tyr416 (Å2)b

RMSD of Glu310 (Å)c

SH2 domain

SH3 domain

51.7% 17.2% 20.7% 36.4%

23.4 158.3 174.2 79.6

0 7.8 6.1 1.6

−8.37d −0.23d −7.45e −8.10d

−6.52f −4.87f −4.34f −5.76g

a

The ratio of helical residues to all residues in A-loop. The helix can be α, 310, and π, assigned with the DSSP program.32 bThe SASA of Tyr416 residue in A-loop, calculated using the MSMS program.33 cThe RMSD of Glu310 from that of model 1. dThe MM-PBSA derived free energy for CT fragment binding to SH2 domain. eThe MM-PBSA derived free energy for SIP-CT peptide binding to SH2 domain. fThe MM-PBSA derived free energy for PPII fragment binding to SH3 domain. gThe MM-PBSA derived free energy for SIP-PPII peptide binding to SH3 domain.

of models 2 and 3 is exposed to solvent that can be readily phosphorylated (to induce DFG-in and to active the kinase) (SASA = 158.3 and 174.2 Å2, respectively). In contrast, this residue of model 1 is deeply buried that prevents it from phosphorylation, while model 4 is partially buried (SASA = 23.4 and 79.6 Å2, respectively) (Table 1). The αC helix plays a molecular switcher role in the conformational transition of c-Src kinase;30 the αC helix displacement serves as a general approach for allosteric modulation of the kinase.31 Superposition of the αC helices of four model systems in MD equilibrium phase is shown in Figure 5B, from which it is evident that the displacement of four αC helices can be classified into two sets: the models 1 and 4 (displacement RMSD = 0 and 1.6 Å, respectively) as one set and the models 2 and 3 (displacement RMSD = 7.8 and 6.1 Å, respectively) as other set (Table 1). The αC helix in models 1 F

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

Figure 6. Binding free energy change upon the truncation of SIP-CT peptide to SH2 domain. The truncation is performed separately at N- and Ctermini of the peptide, one residue at a time, until it reached to the phosphorylated tyrosine residue pY.

and 4 is located far away from kinase’s active site; the conserved helical residue Glu310 points oppositely from the active site, which indicates the kinase in inactive form. In models 2 and 3 the αC helix rotates close to the active site, which alters the saltbridging interaction of Glu310 with another conserved residue Lys295 and then rearranges the A-loop conformation into an active DFG-in state. Binding free energy analysis (Table 1) of the four model systems further solidifies the findings obtained from MD simulations. For SH2 domain, the CT tail can tightly bind to SH2 domain in models 1 and 4 with ΔG values of −8.37 and −8.10 kcal/mol, respectively, indicating that targeting the SH3PPII interaction (the first SBP) would not influence the SH2CT interaction (the second SBP) substantially. However, dephosphorylation of pTyr527 residue can totally eliminate the binding capability of the CT tail to SH2 domain (model 2, ΔG = −0.23 kcal/mol). In addition, given the similar binding potency of self-inhibitory peptide SIP-CT (model 3, ΔG = −7.45 kcal/mol) and the intact CT tail (models 1 and 4, ΔG = −8.37 and −8.10 kcal/mol, respectively), the protein context is suggested to has only a minor effect on the CT binding to SH2 domain. For SH3 domain, both the pTyr52 dephosphorylation (model 2) and targeting SH2-CT interaction (model 3) have a similar effect on PPII binding to SH3 domain, with ΔG decrease from −6.52 kcal/mol (model 1) to −4.87 kcal/mol (model 2) and to −4.34 kcal/mol (model 3). However, splitting of the PPII segment from kinase protein would not alter its interaction with SH3 domain significantly (model 4, ΔG = −5.76 kcal/mol). 3.2. Rational Design and Optimization of Peptide Ligands to Target the Second SBP. Above structural, energetic and dynamics investigations of the four c-Src model systems revealed that targeting SH2 domain can result in the biological effect analogous to Tyr527 phosphorylation, i.e. unlocking the autoinhibitory form of c-Src kinase and activating the kinase, but targeting SH3 domain cannot. Therefore, we herein attempted to computationally design potent peptide ligands of SH2 domain that can competitively disrupt the cognate SBP interaction of SH2 domain with CT and then activate the kinase. The self-inhibitory peptide SIP-CT (TSTEPQpYQPGENL) was used as the start to perform the design and optimization. In order to determine the shortest sequence of SIP-CT required for domain−peptide binding, we generated a series of truncated versions of the SIP-CT peptide

around the phosphotyrosine residue pY and calculated their binding free energies to the SH2 domain using MM-PBSA analysis. First, the SH2 domain−SIP-CT complex system was stripped from the equilibrium structure of model 3. Second, the truncation was carried out separately at N- and C-termini of the SIP-CT peptide, one residue at a time, until reached to the pY, to generate a series of truncated peptides in complex with SH2 domain. Third, an additional 50 ns MD simulation was performed on each of the domain−peptide complexes for equilibrium and snapshot collection for binding energetic analysis. Consequently, 14 truncated peptides were obtained systematically, and their binding free energy changes (ΔΔG) relative to the native peptide SIP-CT are shown in Figure 6. As can be seen, the truncation at both sides of SIP-CT is energetically unfavorable to the peptide binding. However, the three N-terminal residues (ΔTST) and the one C-terminal residue (ΔL) seem to contribute very limitedly to the binding; removal of these residues only causes a slight free energy loss (ΔΔG < 0.05 kcal/mol). A further truncation at both the sides would considerably reduce the peptide binding energy. For example, removal of all residues N-terminal or C-terminal to pY (ΔTSTEPQ or ΔQPGENL) would largely impair SH2− peptide interaction (ΔΔG = 2.41 or 3.07 kcal/mol). In particular, further removal of the anchor pY, either from N- or C-terminus (ΔTSTEPQpY or ΔpYQPGENL), can fully block the peptide binding (ΔΔG = 6.08 or 5.19 kcal/mol). In this respect, we obtained a shortened version of SIP-CT peptide, namely SIP-CT(cΔTST-nΔL) (EPQpYQPGEN), which is absent of three N-terminal residues and a C-terminal residue as compared to SIP-CT but exhibits a similar binding behavior (ΔG = −6.8 kcal/mol) with the SIP-CT (ΔG = −7.5 kcal/ mol). In order to improve the binding affinity of SIP-CT(cΔTSTnΔL) to c-Src SH2 domain, we herein employed a grafting strategy to optimize the sequence and structure of the peptide. The SH2-binding phosphopeptides possess a common core binding motif, including the phosphotyrosine residue pY and its subsequent residue triad X+1X+2X+3.34 Here, a number of peptide sequences extracting from known SH2-binding sites of protein partners were collected from the literature.35−38 These sequences are aligned at pY and shown in Figure 7, in which the core binding motif pYX+1X+2X+3 is highlighted. As can be seen, the first two residues X1 and X2 of the triad prefer to be occupied by negatively charged aspartic acid (Asp) and G

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

energy between SH2 domain and these peptides are very favorable (ΔEint = ∼−100 kcal/mol), which, however, would largely be counteracted by the unfavorable desolvation penalty (ΔGslv = ∼60 kcal/mol) and entropy loss (−TΔS = ∼30 kcal/ mol), thus exhibiting only a moderate binding potency (ΔG > −10 kcal/mol for most peptides). In addition, the entropy loss varies modestly over these peptides, although its absolute values seem to be significant. It is known that entropy can significantly contribute unfavorably to protein−peptide binding due to the large flexibility of peptide ligands. Previously, our group has systematically investigated entropic effect in peptide binding and found that entropy penalty is a primary contribution to the indirect readout energy in protein−peptide recognition instead of deformation energy which is the main source of indirect readout energy in classical biomolecular interaction.40 In addition, crystal structure-based survey of a large number of protein−peptide interactions unraveled that most peptides do not induce conformational changes on their partner upon binding, thus minimizing the entropic cost of binding.41 As might be expected, the self-inhibitory peptide SIP-CT can bind tightly to SH2 domain (ΔG = −7.5 kcal/mol), and its shortened version SIP-CT(cΔTST-nΔL) can also interact efficiently with the domain (ΔG = −6.8 kcal/mol). However, the binding energy of dephosphorylated peptide SIP-CT(Δp) decreases considerably relative to the intact SIP-CT (ΔG changes from −7.5 to −2.4 kcal/mol), suggesting that the phosphorylated pY is required for the domain−peptide recognition and interaction. The 11 mutant peptides exhibit different binding potency toward the domain, with ΔG ranging between −4.6 and −11.5 kcal/mol. In particular, most of these mutants (7/11) have increased potency as compared to SIP-CT and SIP-CT(cΔTST-nΔL), demonstrating that the grafting strategy is successful and could be used to generate potent binders of SH2 domain. In order to substantiate the computational findings, four mutant peptides (i.e., mutant 2 EPQpYEEIEN, mutant 5 EPQpYNELEN, mutant 7 EPQpYEELEN, and mutant 8 EPQpYEDLEN, where the grafted residue triad is underlined) as well as the native peptides SIPCT, SIP-CT(Δp), and SIP-CT(cΔTST-nΔL) were synthesized, purified, and confirmed, and their binding affinity toward the recombinant protein of human c-Src SH2 domain was

Figure 7. Summary of the known binding sites of c-Src SH2 domain.35−38 The core binding motif pYX1X2X3 is highlighted. The residue triad X1X2X3 in these motif includes AEI, EEI, NTI, TGL, NEL, EGL, EEL, EDL, EEM, EDM, and EGM.

glutamic acid (Glu), while the third residue X3 tends to present nonpolar amino acids such as leucine (Leu), isoleucine (Ile), and methionine (Met). A total of 11 different triads were extracted from these sequences, including AEI, EEI, NTI, TGL, NEL, EGL, EEL, EDL, EEM, EDM, and EGM, which were separately grafted into the SIP-CT(cΔTST-nΔL) using a SCWRL4 rotamer-based virtual mutagenesis method,39 resulting in 11 peptide mutants. Here, the binding energetic data (binding free energy ΔG, interaction energy ΔEint, desolvation penalty ΔGslv and entropy loss −TΔS) of the 11 mutants as well as the SIP-CT(cΔTST-nΔL), SIP-CT and the dephosphorylated version of SIP-CT [i.e., SIP-CT(Δp)] to c-Src SH2 domain are tabulated in Table 2. The intermolecular interaction

Table 2. Calculated Binding Energetics As Well As Experimental Affinity and Activation Potency of SIP-CT and Its Derivative Peptides to c-Src SH2 Domain binding energetics (kcal/mol)

a

peptide

sequence

ΔEint

ΔGslv

−TΔS

ΔG

Kd (nM)

EC50 (μM)

SIP-CT SIP-CT(Δp) SIP-CT(cΔTST-nΔL) mutant 1 mutant 2 mutant 3 mutant 4 mutant 5 mutant 6 mutant 7 mutant 8 mutant 9 mutant 10 mutant 11

TSTEPQpYQPGENL TSTEPQYQPGENL EPQpYQPGEN EPQpYAEIEN EPQpYEEIEN EPQpYNTIEN EPQpYTGLEN EPQpYNELEN EPQpYEGLEN EPQpYEELEN EPQpYEDLEN EPQpYEEMEN EPQpYEDMEN EPQpYEGMEN

−126.2 −88.7 −107.5 −98.0 −118.6 −103.4 −95.4 −101.6 −105.2 −114.0 −105.7 −103.9 −100.4 −95.7

79.3 50.1 68.5 58.3 75.3 67.2 58.8 58.3 67.4 70.3 67.5 65.2 60.5 56.8

39.4 36.2 32.2 34.6 31.8 29.5 32.0 34.2 30.1 33.5 28.4 30.1 31.7 32.8

−7.5 −2.4 −6.8 −5.1 −11.5 −6.7 −4.6 −9.1 −7.6 −10.2 −9.8 −8.6 −8.2 −6.1

53 ± 8 nda 107 ± 18 ntb 8.3 ± 2.1 ntb ntb 78 ± 13 ntb 15 ± 4 32 ± 7 ntb ntb ntb

580 ± 96 ntb ntb ntb 3.2 ± 0.6 ntb ntb ntb ntb 41 ± 7 ntb ntb ntb ntb

Not detectable. bNot tested. H

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

Figure 8. Fluorescence polarization (FP) curves of different peptide ligands in the presence of varying concentrations of the c-Src SH2 domain.

Figure 9. (A) Three-dimensional structure of c-Src SH2 domain in complex with the high-affinity mutant 2 peptide (EPQpYEEIEN). The complex structure was computationally modeled via virtual mutagenesis, MD equilibrium, and energy minimization. (B) Schematic representation of the nonbonded interactions across the complex interface. The nonbonded interactions were identified using in-house program 2D-GraLab.42

determined by using in vitro fluorescence polarization (FP) assays. The FP curves of different peptide ligands in the presence of varying concentrations of the domain protein are shown in Figure 8, and the measured dissociation constants (Kd values) are listed in Table 2. A good consistency between the calculated energy and experimental affinity can be observed; the SIP-CT and SIP-CT(cΔTST-nΔL) can bind to SH2 domain with a moderate affinity (Kd = 53 and 107 nM, respectively), while no binding was detected for the dephosphorylated SIPCT(Δp) (Kd = nd). Three (i.e., mutants 2, 7, and 8) out of the four tested mutant peptides have increased affinity as compared to the native SIP-CT, in which mutant 2 was measured as the strongest binder of SH2 domain in all tested peptides, which can bind tightly to the domain at nanomolar level (Kd = 8.3 nM). Subsequently, we employed an in-house program 2DGraLab42 to examine the SH2−mutant 2 complex structure modeled computationally via virtual mutagenesis, MD equilibrium and energy minimization (Figure 9A), identifying a complicated network of nonbonded interactions across the complex interface (Figure 9B). As can be seen, the phosphotyrosine residue pY at position 0 (pY0) of the peptide can form three salt bridges and a hydrogen bond with the positively charged residues Arg155, His201, and Lys203 and

polar residue Thr179 of SH2 domain, which defines an anchor site for the domain−peptide interaction (His201 residue was predicted to be protonated by using H++ server43 based on the modeled complex structure of c-Src SH2 domain with mutant 2 peptide; it means that the residue carries one positive formal charge in the system). The MD equilibrium conformations of SIP-CT and mutant 2 peptide as well as the intrinsic CT in crystal structure were superposed in the peptide-binding pocket of c-Src SH2 domain (Figure 10). Note that the intrinsic CT and SIP-CT share a common sequence (TSTEPQpYQPGENL), while the mutant 2 peptide (EPQpYEEIEN) is a truncated and mutated version, which misses three N-terminal residues and a C-terminal residue relative to CT and SIP-CT. As can be seen, the three peptides exhibit a basically consistent conformation, with only a moderate structural variation observed in their N-terminus. This is expected if considering that all peptide ligands prefer to adopt a similar bind mode to their common domain receptor, although the side chains and local atom locations may vary considerably over different peptides. According to fluorescence polarization (FP) assays, mutant 2 peptide has a high affinity (Kd = 3.2 μM) relative to SIP-CT (Kd = 580 μM); mutant 2 peptide possess two negatively charged Glu residues (E+1 and E+2) succeeding the pY0 residue, which can form better I

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

tightly packed kinase, but influence the kinase activity moderately. We also carried out rational design of peptide mediators to competitively bind with SH2 domain, which was then substantiated using in vitro FP and kinase assays. A truncated and mutated peptide was obtained, which exhibits a high-affinity to the SH2 domain and strong activation potency for the kinase.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: +862883202351. ORCID

Peng Zhou: 0000-0001-5681-9937 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by the National Natural Science Foundation of China (Nos. 31671361 and 31200993), the Science and Technology Project of Sichuan Province (No. 2015JY0252), and the Fundamental Research Funds for the Central Universities of China (No. ZYGX2015Z006).

Figure 10. Superposition of the MD equilibrium conformations of SIP-CT and mutant 2 peptide as well as the intrinsic CT in crystal structure complexed with c-Src SH2 domain.



electrostatic complementarity with the SH2 pocket that contains a number of positively charged residues such as Arg155, Arg175, Lys200, His201, and Arg205. In the residue triad of peptide core binding motif, the negatively charged E+1 and E+2 interact effectively with the domain via short-range salt bridges and long-range electrostatic forces, while the nonpolar I+3 is in hydrophobic contact with domain residues Ile214 and Leu237. Thus, the grafted residue triad (EEI) of mutant 2 peptide can confer high stability and specificity to the complex system. In addition, the N-terminal E−3 and C-terminal E+4/N+3 of the peptide ligands can also form a number of polar chemical interactions with the domain. This molecular evidence well explains the measured high affinity of mutant 2 peptide, and we therefore consider that the peptide can be used as a potential activator of c-Src kinase by competitively targeting the SBP binding site of SH2 domain and then disrupting the cognate SH2-CT interaction. To verify this hypothesis, the activation potency of mutant 2 peptide and, for comparison purpose, SIP-CT and mutant 7 peptide was determined using in vitro kinase assays. As expected, these peptides can efficiently release the autoinhibitory state of c-Src kinase (with Tyr527 phosphorylated); the activation potency against the kinase increases in following order: SIP-CT < mutant 7 < mutant 2 (EC50 = 580, 41, and 3.2 μM, respectively). This is well consistent the measured binding affinity of the three peptides to SH2 domain (Kd = 53, 15, and 8.3 μM, respectively), confirming that the activation potency is retrieved from peptide targeting the domain.

REFERENCES

(1) Drews, J. Drug Discovery: A Historical Perspective. Science 2000, 287, 1960−1964. (2) Wells, J. A.; McClendon, C. L. Reaching for High-hanging Fruit in Drug Discovery at Protein−Protein Interfaces. Nature 2007, 450, 1001−1009. (3) Pommier, Y.; Cherfils, J. Interfacial Inhibition of Macromolecular Interactions: Nature’s Paradigm for Drug Discovery. Trends Pharmacol. Sci. 2005, 26, 138−145. (4) Neduva, V.; Russell, R. B. Peptides Mediating Interaction Networks: New Leads at Last. Curr. Opin. Biotechnol. 2006, 17, 465− 471. (5) Stein, A.; Aloy, P. Novel Peptide-mediated Interactions Derived from High-resolution Three-dimensional Structures. PLoS Comput. Biol. 2010, 6, e1000789. (6) Vanhee, P.; van der Sloot, A. M.; Verschueren, E.; Serrano, L.; Rousseau, F.; Schymkowitz, J. Computational Design of Peptide Ligands. Trends Biotechnol. 2011, 29, 231−239. (7) Petsalaki, E.; Stark, A.; García-Urdiales, E.; Russell, R. B. Accurate Prediction of Peptide Binding Sites on Protein Surfaces. PLoS Comput. Biol. 2009, 5, e1000335. (8) Yang, C.; Zhang, S.; He, P.; Wang, C.; Huang, J.; Zhou, P. Selfbinding Peptides: Folding or Binding? J. Chem. Inf. Model. 2015, 55, 329−342. (9) Yang, C.; Zhang, S.; Bai, Z.; Hou, S.; Wu, D.; Huang, J.; Zhou, P. A Two-step Binding Mechanism for the Self-binding Peptide Recognition of Target Domains. Mol. BioSyst. 2016, 12, 1201−1213. (10) Yeatman, T. J. A Renaissance for SRC. Nat. Rev. Cancer 2004, 4, 470−480. (11) Wheeler, D. L.; Iida, M.; Dunn, E. F. The Role of Src in Solid Tumors. Oncologist 2009, 14, 667−78. (12) Hubbard, S. R. Src Autoinhibition: Let Us Count the Ways. Nat. Struct. Biol. 1999, 6, 711−714. (13) Xu, W.; Doshi, A.; Lei, M.; Eck, M. J.; Harrison, S. C. Crystal Structures of c-Src Reveal Features of Its Autoinhibitory Mechanism. Mol. Cell 1999, 3, 629−638. (14) London, N.; Raveh, B.; Movshovitz-Attias, D.; Schueler-Furman, O. Can Self-inhibitory Peptides be Derived from the Interfaces of Globular Protein-protein Interactions? Proteins: Struct., Funct., Genet. 2010, 78, 3140−3149. (15) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.; Klein, M. L. Comparison of Simple Potential Function for Simulating Liquid Water. J. Chem. Phys. 1983, 79, 926−935.

4. CONCLUSIONS In this study, SBPs have been successfully established as a new class of druggable targets to regulate protein activity and function. An oncogene nonreceptor tyrosine kinase c-Src was adopted as a paradigm to perform case study of the targetability of SBPs by small peptide ligands. Long-term MD simulations and free energy analysis revealed that the binding of peptide ligands to SH2 domain in the kinase system can result in a similar effect of structural dynamics and energetics with the dephosphorylation of C-terminal pTyr527 residue; both of them unlock the kinase from an autoinhibitory form, whereas targeting SH3 domain can only release the domain from the J

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX

Article

Journal of Chemical Information and Modeling

SH2 Domains Recognize Specific Phosphopeptide Sequences. Cell 1993, 72, 767−778. (36) Payne, G.; Shoelson, S. E.; Gish, G. D.; Pawson, T.; Walsh, C. T. Kinetics of p56lck and p60src Src Homology 2 Domain Binding to Tyrosine-phosphorylated Peptides Determined by a Competition Assay or Surface Plasmon Resonance. Proc. Natl. Acad. Sci. U. S. A. 1993, 90, 4902−4906. (37) Ladbury, J. E.; Lemmon, M. A.; Zhou, M.; Green, J.; Botfield, M. C.; Schlessinger, J. Measurement of the Binding of Tyrosyl Phosphopeptides to SH2 Domains: A Reappraisal. Proc. Natl. Acad. Sci. U. S. A. 1995, 92, 3199−3203. (38) Lindfors, H. E.; Drijfhout, J. W.; Ubbink, M. The Src SH2 Domain Interacts Dynamically with the Focal Adhesion Kinase Binding Site as Demonstrated by Paramagnetic NMR Spectroscopy. IUBMB Life 2012, 64, 538−544. (39) Krivov, G. G.; Shapovalov, M. V.; Dunbrack, R. L. Improved Prediction of Protein Side-chain Conformations with SCWRL4. Proteins: Struct., Funct., Genet. 2009, 77, 778−795. (40) Yu, H.; Zhou, P.; Deng, M.; Shang, Z. Indirect Readout in Protein−Peptide Recognition: A Different Story from Classical Biomolecular Recognition. J. Chem. Inf. Model. 2014, 54, 2022−2032. (41) London, N.; Movshovitz-Attias, D.; Schueler-Furman, O. The Structural Basis of Peptide−Protein Binding Strategies. Structure 2010, 18, 188−199. (42) Zhou, P.; Tian, F.; Shang, Z. 2D Depiction of Nonbonding Interactions for Protein Complexes. J. Comput. Chem. 2009, 30, 940− 951. (43) Gordon, J. C.; Myers, J. B.; Folta, T.; Shoja, V.; Heath, L. S.; Onufriev, A. H++: A Server for Estimating pKa’s and Adding Missing Hydrogens to Macromolecules. Nucleic Acids Res. 2005, 33, W368− W371.

(16) Darden, T.; York, D.; Pedersen, L. Particle mesh Ewald: An N· log(N) Method for Ewald sums in large systems. J. Chem. Phys. 1993, 98, 10089−10092. (17) Ryckaert, J. P.; Ciccotti, G.; Berendsen, H. J. C. Numerical Integration of the Cartesian Equations of Motion of a System with Constraints: Molecular Dynamics of n-Alkanes. J. Comput. Phys. 1977, 23, 327−341. (18) Wu, X. W.; Brooks, B. R. Self-guided Langevin Dynamics Simulation Method. Chem. Phys. Lett. 2003, 381, 512−518. (19) Duan, Y.; Wu, C.; Chowdhury, S. S.; Lee, M. C.; Xiong, G. M.; Zhang, W.; Yang, R.; Cieplak, P.; Luo, R.; Lee, T. S.; Caldwell, J.; Wang, J. M.; Kollman, P. A Point-charge Force field for Molecular Mechanics Simulations of Proteins. J. Comput. Chem. 2003, 24, 1999− 2012. (20) Götz, A. W.; Williamson, M. J.; Xu, D.; Poole, D.; Le Grand, S.; Walker, R. C. Routine Microsecond Molecular Dynamics Simulations with AMBER on GPUs. 1. Generalized Born. J. Chem. Theory Comput. 2012, 8, 1542−1555. (21) Homeyer, N.; Gohlke, H. Free Energy Calculations by the Molecular Mechanics Poisson-Boltzmann Surface Area Method. Mol. Inf. 2012, 31, 114−122. (22) Kollman, P. A.; Massova, I.; Reyes, C.; Kuhn, B.; Huo, S.; Chong, L.; Lee, M.; Lee, T.; Duan, Y.; Wang, W.; Donini, O.; Cieplak, P.; Srinivasan, J.; Case, D. A.; Cheatham, T. E. Calculating Structures and Free Energies of Complex Molecules: Combining Molecular Mechanics and Continuum Models. Acc. Chem. Res. 2000, 33, 889− 897. (23) Case, D. Normal Mode Analysis of Protein Dynamics. Curr. Opin. Struct. Biol. 1994, 4, 285−290. (24) Gilmer, T.; Rodriguez, M.; Jordan, S.; Crosby, R.; Alligood, K.; Green, M.; Kimery, M.; Wagner, C.; Kinder, D.; Charifson, P.; Hassell, A. M.; Willard, D.; Luther, M.; Rusnak, D.; Sternbach, D. D.; Mehrotra, M.; Peel, M.; Shampine, L.; Davis, R.; Robbins, J.; Patel, I. R.; Kassel, D.; Burkhart, W.; Moyer, M.; Bradshaw, T.; Berman, J. Peptide Inhibitors of Src SH3-SH2 Phosphoprotein Interactions. J. Biol. Chem. 1994, 269, 31711−31719. (25) Frangioni, J. V.; Neel, B. G. Solubilization and Purification of Enzymatically Active Glutathione S-transferase (pGEX) Fusion Proteins. Anal. Biochem. 1993, 210, 179−187. (26) Ayrapetov, M. K.; Nam, N. H.; Ye, G.; Kumar, A.; Parang, K.; Sun, G. Functional Diversity of Csk, Chk, and Src SH2 Domains due to a Single Residue Variation. J. Biol. Chem. 2005, 280, 25780−25787. (27) Hause, R. J.; Leung, K. K.; Barkinge, J. L.; Ciaccio, M. F.; Chuu, C. P.; Jones, R. B. Comprehensive Binary Interaction Mapping of SH2 Domains via Fluorescence Polarization Reveals Novel Functional Diversification of ErbB Receptors. PLoS One 2012, 7, e44471. (28) Treiber, D. K.; Shah, N. P. Ins and Outs of Kinase DFG Motifs. Chem. Biol. 2013, 20, 745−746. (29) Chen, M. L.; Chai, C. Y.; Yeh, K. T.; Wang, S. N.; Tsai, C. J.; Yeh, Y. T.; Yang, S. F. Crosstalk between Activated and Inactivated cSrc in Hepatocellular Carcinoma. Dis. Markers 2011, 30, 325−333. (30) Huang, H.; Zhao, R.; Dickson, B. M.; Skeel, R. D.; Post, C. B. αC Helix as a Switch in the Conformational Transition of Src/CDKlike Kinase Domains. J. Phys. Chem. B 2012, 116, 4465−4475. (31) Palmieri, L.; Rastelli, G. αC Helix Displacement as a General Approach for Allosteric Modulation of Protein Kinases. Drug Discovery Today 2013, 18, 407−414. (32) Kabsch, W.; Sander, C. Dictionary of Protein Secondary Structure: Pattern Recognition of Hydrogen-bonded and Geometrical Features. Biopolymers 1983, 22, 2577−2637. (33) Sanner, M. F.; Olson, A. J.; Spehner, J. C. Reduced Surface: an Efficient Way to Compute Molecular Surfaces. Biopolymers 1996, 38, 305−320. (34) Gan, W.; Roux, B. Binding Specificity of SH2 Domains: Insight from Free Energy Simulations. Proteins: Struct., Funct., Genet. 2009, 74, 996−1007. (35) Zhou, S.; Shoelson, S. E.; Chaudhuri, M.; Gish, G.; Pawson, T.; Haser, W. G.; King, F.; Roberts, T.; Ratnofsky, S.; Lechleider, R. J. K

DOI: 10.1021/acs.jcim.6b00673 J. Chem. Inf. Model. XXXX, XXX, XXX−XXX