Eukaryotic protein kinases (PKs) are a large family of proteins critical for cellular response to external signals, acting as molecular switches. PKs propagate biochemical signals by catalyzing phosphorylation of other proteins, including other PKs, which can undergo conformational changes upon phosphorylation and catalyze further phosphorylations. Although PKs have been studied thoroughly across the domains of life, the structures of these proteins are sparsely understood in numerous groups of organisms, including plants. In addition to efforts towards determining crystal structures of PKs, research on human PKs has incorporated molecular dynamics (MD) simulations to study the conformational dynamics underlying the switching of PK function. This approach of experimental structural biology coupled with computational biophysics has led to improved understanding of how PKs become catalytically active and why mutations cause pathological PK behavior, at spatial and temporal resolutions inaccessible to current experimental methods alone. In this review, we argue for the value of applying MD simulation to plant PKs. We review the basics of MD simulation methodology, the successes achieved through MD simulation in animal PKs, and current work on plant PKs using MD simulation. We conclude with a discussion of the future of MD simulations and plant PKs, arguing for the importance of molecular simulation in the future of plant PK research.

Introduction

Eukaryotic cells are able to rapidly respond to biotic and abiotic changes in their environment, allowing them to follow chemical gradients [1], alter their growth cycle in response to changes in chemical signal concentrations [2], or co-ordinate their behavior with other cells [3], among numerous other abilities. Though the mechanisms for transduction of external signals into cellular response vary in detail across different organisms, cell types, and signal types, eukaryotic protein kinases (PKs) are central to many of the responsible signaling pathways [4]. The proximate function of PKs is to catalyze the transfer of phosphoryl groups, typically donated by ATP molecules, to the Tyr, Thr, or Ser residues of other proteins. Phosphorylation can alter the physical and catalytic properties of proteins, including PKs themselves. The propagation of a signal through successive phosphorylation of PKs by other PKs is often referred to as a kinase cascade and is an important component of the eukaryotic cellular circuitry [5]. Ultimately, kinase cascades are often coupled to gene regulation machinery, completing the connection between the environment of a cell and its gene expression.

As a consequence of their importance in such a wide variety of cellular processes, PKs with abnormal function or expression levels are often implicated in disease, perhaps most notably in many cancer types [68]. For this reason, PKs attract considerable attention from researchers in a biomedical context. A large portion of this effort has been directed toward understanding how PKs are structured and how their structures change upon phosphorylation or interaction with other proteins [9]. Owing to their primary role in information relay, PKs have evolved to act as highly sensitive switches rather than catalytic workhorses rapidly churning out products [10], meaning that understanding the structure and conformational dynamics of PKs is central to understanding their function and dysfunction within cells. Further supporting the importance of a structural understanding of PKs to medicine is the rise of rational PK inhibitor design using computational and experimental models, which depend on the detailed knowledge of protein structure [11,12].

In this review, we aim to highlight the importance of MD simulations for understanding the function and regulation of plant PKs. In the first section, we discuss the state of plant PK structural biophysics and describe general PK structure. Next, we give a brief overview of molecular dynamics (MD) simulation, emphasizing the scientific questions motivating its development. We describe past work on non-plant eukaryotic PKs and the unique biological insights gained through combining MD simulation and experimental results. We then review the few examples of MD simulation applied to plant PKs. Finally, we discuss the possible role of MD simulation in the study of plant PKs and argue that MD simulation, complementing structural experimental studies, will be critical to further progress in the field.

Plant PK structure and dynamics: where few have gone before

The basic anatomy of a typical PK is shown in Figure 1a, with some of the most important features highlighted. The αC helix, shown in gray, must be folded and positioned close to the remainder of the kinase domain for catalysis. The Lys-Glu bond is an ionic interaction between a conserved Lys-Glu pair, where the Glu is part of the αC helix, which must be formed in catalytically competent conformations [19]. While ATP is not present in this structure, ATP must be bound at the labeled ATP-binding site with the Gly-rich loop closing over it, while the Phe in the DFG motif must be positioned outward, as in the displayed structure. Proper positioning of the Phe in the DFG motif and of the αC helix completes a contiguous stack of hydrophobic residues, known as the regulatory spine (R-spine) [10]. The activation loop (A-loop) must be unfolded flat over the top of the C-lobe, allowing for substrate binding and exposing the catalytic residues in the catalytic loop (C-loop) to the substrate. Positioning the A-loop for activation is often accomplished through phosphorylation and accounts for the activation of PKs by other PKs.

Structural overview of plant protein kinases.

Figure 1.
Structural overview of plant protein kinases.

(a) A typical active eukaryotic PK structure, consisting of an N-lobe and a C-lobe with an ATP-binding site in between. Several important regulatory domains are highlighted. In the N-lobe, the αC helix, shown in gray, must be properly positioned for activity, with a conserved Glu forming the Lys-Glu ionic interaction with a conserved Lys above the ATP-binding pocket. Additionally, the Gly-rich loop, shown in blue, closes over the ATP-binding site in the presence of ATP, holding it in place. In the C-lobe, the C-loop contains a critical Asp which directly participates in catalysis. The DFG motif contains an additional Asp involved in holding the phosphate groups of ATP in place for catalysis, as well as a Phe which can flip into the ATP-binding site, preventing ATP binding. Finally, the A-loop serves as a phosphorylation-response motif, in several PKs folding in such a way as to block catalysis and binding of the substrate until phosphorylation when it unfolds and lays at on the C-lobe, as in the shown structure. The protein structure (PDBID: 3TL8, chain A) [13] image was prepared using VMD 1.9.2. [14]. (b) Coverage of the human and A. thaliana kinomes within the PDB with time. All structures of proteins containing PK domains from both species were identified, and only one structure per UniProt accession number was counted. The total kinome sizes [15,16] were used to determine percent kinome coverage. The plot of PDB structures [17] was prepared using Matplotlib 1.5.3 [18].

Figure 1.
Structural overview of plant protein kinases.

(a) A typical active eukaryotic PK structure, consisting of an N-lobe and a C-lobe with an ATP-binding site in between. Several important regulatory domains are highlighted. In the N-lobe, the αC helix, shown in gray, must be properly positioned for activity, with a conserved Glu forming the Lys-Glu ionic interaction with a conserved Lys above the ATP-binding pocket. Additionally, the Gly-rich loop, shown in blue, closes over the ATP-binding site in the presence of ATP, holding it in place. In the C-lobe, the C-loop contains a critical Asp which directly participates in catalysis. The DFG motif contains an additional Asp involved in holding the phosphate groups of ATP in place for catalysis, as well as a Phe which can flip into the ATP-binding site, preventing ATP binding. Finally, the A-loop serves as a phosphorylation-response motif, in several PKs folding in such a way as to block catalysis and binding of the substrate until phosphorylation when it unfolds and lays at on the C-lobe, as in the shown structure. The protein structure (PDBID: 3TL8, chain A) [13] image was prepared using VMD 1.9.2. [14]. (b) Coverage of the human and A. thaliana kinomes within the PDB with time. All structures of proteins containing PK domains from both species were identified, and only one structure per UniProt accession number was counted. The total kinome sizes [15,16] were used to determine percent kinome coverage. The plot of PDB structures [17] was prepared using Matplotlib 1.5.3 [18].

In contrast with human and, more generally, mammalian PKs, plant PK research has focused far less on structure and conformational dynamics; perhaps, the moral and financial imperatives driving human PK research at such a high level of spatio-temporal detail are not as readily evident for plants. Regardless of the reasons for these differences, plant PK research has made significant strides toward understanding PK function within a variety of signaling pathways exemplified by research on brassinosteroid signaling [2024]. Still, the dearth of plant PK structures determined by any experimental method (11 entries in total), quantified by the number of unique Arabidopsis thaliana kinase domains with at least one structure on the Protein Data Bank (PDB) [17], has limited a detailed understanding of plant signaling pathways (Figure 1b). We analyzed the availability of PK structures from the model organism A. thaliana rather than any plant of any direct relevance to agriculture due to the even further scarcity of PK structures from any other plant, where we found two entries from Zea mays (maize) and one entry from Solanum pimpinellifolium (tomato).1

The present and projected disastrous effects of global climate change on agricultural growing conditions in regions with high levels of food insecurity [25], coupled with a growing global population predicted to outgrow current food production projections [26], make the need for crop innovation increasingly clear. The importance of PKs in plant growth and drought resistance could make them prime targets for genetic engineering or agrochemical development [27]. For example, changes in the expression of several PKs have been found to improve plant drought resistance [2831], while expression of constitutively active PKs in plants has been shown to improve stress tolerance [32], increase sensitivity to stress signaling hormones [33], and improve leaf growth [34]. A detailed understanding of PK structure and dynamics is invaluable for guiding the introduction of point mutations or more extensive modifications to PK sequences in order to tune function and ultimately improve crop yield. Just as structural information guides rational design of inhibitors for mammalian PKs involved in cancers [11,12], agrochemicals could be rationally designed to target plant PKs given sufficient knowledge of their structure, providing another avenue for the manipulation of plant growth and stress resistance.

The study of plant PKs from a structural–dynamics perspective also has the potential to uncover previously unknown modes of PK regulation. Considering the extensive growth of plant kinomes through the history of evolution partially owing to adaption to environmental stresses [35,36] in addition to the broad repertoire of post-translational modifications (PTMs) utilized by plants to alter protein function [37,38], it seems likely that unique modes of PK regulation await discovery within the kinomes of plants. Additionally, a structural understanding of how PKs fit into signaling pathways, in terms of their protein–protein interactions, is critical for insights into the biological function of PKs in the cellular context. For example, a dimeric crystal structure was found for subfamily 2 Snfl-related kinase 6 (SnRK2.6) and a type 2C protein phosphatase (HAB1) in A. thaliana [39]. The binding interface of HAB1 and SnRK2.6 closely resembles that of HAB1 and the abscisic acid (ABA) receptor PYL2 when ABA is bound, illuminating how ABA signaling is coupled with SnRK2.6 enzymatic control. With structural models of protein–protein interactions involving PKs, it could be possible to rationally engineer PKs through mutagenesis of interfacial residues to specifically target proteins in one downstream signaling pathway over another, providing fine control over cellular signaling. One potential target for this mode of study is BRI1-ASSOCIATED KINASE 1 (BAK1), a receptor-like kinase involved in both plant growth and immune signaling, where it is already known that differential phosphorylation of BAK1 leads to differential specificity in its phosphotransfer activity [40]. As is the case with many plant PKs, the structural details needed to explain the underlying mechanisms of these changes in specificity are missing.

If we are to understand the structural and dynamic basis for plant PK function, how can the successes of the human PK research program be translated into advances in plant biology? While the human kinome is far from fully characterized on a structural–dynamics level, great progress has been made by combining experimental methods, such as X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy, with computational methods such as MD simulation. While the information provided by X-ray crystallography primarily concerns the static ‘native’ structure of a protein, NMR experiments can provide the timescales of interconversion between different protein conformations along with some structural details. Additionally, computational structure prediction methods, ranging from homology modeling using the extensive library of protein folds in experimentally solved structures [4144] to physics-based modeling guided by evolutionary information [4549], have great potential to extend static structural knowledge of plant kinomes and compensate for the lack of experimental structures. MD simulation, however, enables researchers to gain unique insights into the movement of each atom in a protein over relatively long periods of time (nanoseconds to milliseconds).

Molecular dynamics simulation

The basics of classical MD simulations

MD simulation was initially developed by Fermi et al. [50] in order to study the dynamics of relatively simple multiparticle systems with no closed-form mathematical solutions. From then on, MD simulations of gaseous argon [51], liquid water [52], and eventually proteins [53] were performed as the required software and hardware improved, and as protein crystal structures began to be solved [54]. MD simulation is often likened to a ‘computational microscope’ [55,56] in that it can give a detailed, atom-by-atom view of how proteins and nucleic acids move with time [57].

Modern MD simulation of proteins requires several sources of input. First, an initial set of protein co-ordinates is needed from which to start simulations. These initial co-ordinates are typically taken from X-ray crystallography experiments or are inferred through homology modeling, which uses proteins with similar sequences and known structures to predict a possible structure. In principle, one could construct a computational model of an unfolded protein from sequence information alone and use MD simulation to observe the folding process [58,59], but this approach is usually impractical due to the enormous amount of computing time required to fold most proteins of biological interest.

The second input needed for MD simulation is the set of mathematical functions defining forces between atoms in the protein. One must decide both the forms of the functions used to calculate forces, approximating more accurate physical theory, as well as the parameters tuning those functions to match with experimental quantities and quantum calculations (Figure 2). In practice, for simulation of a protein consisting of the 20 standard amino acids, one must only choose from one of the many available biomolecular force fields [6063], which provide a functional form for force calculations as well as a set of parameters carefully tuned to reproduce certain experimental results.

Interatomic interactions in classical MD force fields.

Figure 2.
Interatomic interactions in classical MD force fields.

(a) A typical eukaryotic PK structure (PDBID: 3TL8, chain A) [13] prepared using VMD 1.9.2. [14]. The inset shows an amino acid with labeled atoms used to demonstrate force field terms. (b–f) Potential energy terms in MD force fields, with diagrammatic representations of the interactions and plots showing the relationship between the variables in the interaction diagrams and potential energy (V), shown on arbitrary, unitless scales. (b) Covalent bonding is approximated by adding a spring between bonded atoms following Hooke's law. The potential energy behaves quadratically with interatomic distance (dij). The parameter k determines how stiff the ‘spring’ is, where a high k value yields low-amplitude vibrations and a low k value yields high-amplitude vibrations. (c) Angles between bonds on the same atom are also approximated with springs, where the potential energy instead depends on the angle between the two bonds (θijk). (d) Dihedral angles are the angles (ϕijkl) between the planes on which the first and last three atoms of four sequentially bonded atoms fall. The potential energy is periodic with the value of a dihedral angle, where the v parameter determines how high the potential energy barriers are, the n parameter determines how many stable angles there are, and the ϕ0 parameter determines which angles are stable. (e) Electrostatic interactions are approximated using Coulomb's law, depending on interatomic distance (dij), while qi and qj are the electric charges of the two atoms and ε0 is the vacuum permittivity, which is replaced by εEnv depending on the environment in which the interaction takes place. The electrostatic potential energy between positively and negatively charged atoms decreases proportionally to 1/dij as dij decreases, but other force field terms account for repulsive interactions preventing two atoms from acquiring arbitrarily small distances from one another. (f) Other non-bonded interactions are represented by the Lennard–Jones potential, depending on interatomic distance (dij). The σ parameter determines the relative stability of the lowest potential energy distance, while A and B determine the balance between attractive [(B/dij)6] and repulsive [(A/dij)12] terms in the potential. The repulsive term provides atoms in MD simulations with their space-filling features, as the potential energy between two atoms increases so rapidly as they come closer together past a certain point that the atoms effectively act like billiards balls.

Figure 2.
Interatomic interactions in classical MD force fields.

(a) A typical eukaryotic PK structure (PDBID: 3TL8, chain A) [13] prepared using VMD 1.9.2. [14]. The inset shows an amino acid with labeled atoms used to demonstrate force field terms. (b–f) Potential energy terms in MD force fields, with diagrammatic representations of the interactions and plots showing the relationship between the variables in the interaction diagrams and potential energy (V), shown on arbitrary, unitless scales. (b) Covalent bonding is approximated by adding a spring between bonded atoms following Hooke's law. The potential energy behaves quadratically with interatomic distance (dij). The parameter k determines how stiff the ‘spring’ is, where a high k value yields low-amplitude vibrations and a low k value yields high-amplitude vibrations. (c) Angles between bonds on the same atom are also approximated with springs, where the potential energy instead depends on the angle between the two bonds (θijk). (d) Dihedral angles are the angles (ϕijkl) between the planes on which the first and last three atoms of four sequentially bonded atoms fall. The potential energy is periodic with the value of a dihedral angle, where the v parameter determines how high the potential energy barriers are, the n parameter determines how many stable angles there are, and the ϕ0 parameter determines which angles are stable. (e) Electrostatic interactions are approximated using Coulomb's law, depending on interatomic distance (dij), while qi and qj are the electric charges of the two atoms and ε0 is the vacuum permittivity, which is replaced by εEnv depending on the environment in which the interaction takes place. The electrostatic potential energy between positively and negatively charged atoms decreases proportionally to 1/dij as dij decreases, but other force field terms account for repulsive interactions preventing two atoms from acquiring arbitrarily small distances from one another. (f) Other non-bonded interactions are represented by the Lennard–Jones potential, depending on interatomic distance (dij). The σ parameter determines the relative stability of the lowest potential energy distance, while A and B determine the balance between attractive [(B/dij)6] and repulsive [(A/dij)12] terms in the potential. The repulsive term provides atoms in MD simulations with their space-filling features, as the potential energy between two atoms increases so rapidly as they come closer together past a certain point that the atoms effectively act like billiards balls.

Once the initial atomic co-ordinates have been acquired and the force field has been chosen, MD simulation can begin.2 At their core, MD programs operate using the same physical principles that Isaac Newton developed over 300 years ago to describe the motion of far larger objects, hence the word ‘classical’ in the title of this section. At the same time, methodological advances continue to improve the accuracy, efficiency, and capability of MD simulation [56].

The motion of each atom is calculated in time steps, producing a sequence of protein conformations describing its motion with time. To give a simple illustrative example of the process, we consider a single atom moving along a straight line (Figure 3a). The position of the atom on the line at a certain time is denoted as x(t) with a velocity v(t) and acceleration a(t). Just as one might calculate the motion of an object in a secondary school physics problem, the position of the atom after a certain amount of time, which we call Δt, can be calculated according to3x(t + Δt) ≈ x(t) + v(tt + (1/2m)f(tt2. The initial position and velocity of the particle are provided at the start of the simulation. From Newton's second law of motion, f = ma, we know that the acceleration of an object is proportional to the force that it feels, meaning that the force fields described above will provide the a(t) term [where a(t) = f(t)/m]. However, our system has only one particle with no forces acting on it and will perpetually continue motion in one direction (Figure 3b), which is not particularly useful for our purposes. If we add another atom so that the two atoms are attached by a spring and the force between them is f(x1, x2) = −k|x1 − x2| (Figure 3c), where | · | is the absolute value of ‘·’, we will have to calculate the motion of both atoms, taking into account the force that they impart on one another (Figure 3d). If we instead consider a system with many interacting atoms each moving in three dimensions, the dynamics can become very complicated and the number of force calculations between each pair of atoms will increase rapidly with the total number of atoms in the system.

Two examples of simple systems with dynamics calculated using Newtonian mechanics.

Figure 3.
Two examples of simple systems with dynamics calculated using Newtonian mechanics.

(a) A single particle with mass m in one dimension has an initial position, x1(t1), and velocity, v1(t1). (b) With no forces acting upon the particle, its motion can be calculated exactly according to x1(t2) = x1(t1) + (t1 − t2) v1(t1) and v1(t2) = v1(t1). (c) For a one-dimensional system with two particles interacting through a spring potential, we must also consider the force that the two particles impart on one another. In this case, the initial positions of the two atoms, x1(t1) and x2(t1), are known along with the initial velocities v1(t1) and v2(t1). The two atoms feel an equal and opposite force, according to f(x1, x2) = −k|x1 − x2|, depending only on the positions. (d) The new positions of the particles can be calculated as x1(t2) ≈ x1(t1) + (t2 − t1)v1(t1) + (1/2m)(t2 − t1)2f1(t1), where x2(t2) is calculated similarly, and new velocities as v1(t2) ≈ v1(t1) + (1/m) (t2 − t1)f1(t1). While an analytical solution to this problem can easily be found, we emphasize that this example is meant to demonstrate the numerical MD simulation method.

Figure 3.
Two examples of simple systems with dynamics calculated using Newtonian mechanics.

(a) A single particle with mass m in one dimension has an initial position, x1(t1), and velocity, v1(t1). (b) With no forces acting upon the particle, its motion can be calculated exactly according to x1(t2) = x1(t1) + (t1 − t2) v1(t1) and v1(t2) = v1(t1). (c) For a one-dimensional system with two particles interacting through a spring potential, we must also consider the force that the two particles impart on one another. In this case, the initial positions of the two atoms, x1(t1) and x2(t1), are known along with the initial velocities v1(t1) and v2(t1). The two atoms feel an equal and opposite force, according to f(x1, x2) = −k|x1 − x2|, depending only on the positions. (d) The new positions of the particles can be calculated as x1(t2) ≈ x1(t1) + (t2 − t1)v1(t1) + (1/2m)(t2 − t1)2f1(t1), where x2(t2) is calculated similarly, and new velocities as v1(t2) ≈ v1(t1) + (1/m) (t2 − t1)f1(t1). While an analytical solution to this problem can easily be found, we emphasize that this example is meant to demonstrate the numerical MD simulation method.

Analysis and extensions of MD

Once trajectories are completed, useful information has to be extracted in order to provide biological insight. One common method of analyzing MD trajectories is to calculate a free energy landscape (FEL) over protein structural features (Figure 4). A FEL provides information about the relative stability of different protein conformations as well as the free energy barriers between stable states, which control how fast a protein can switch between different conformations. For example, if two conformations of a protein are compared, one can use the free energy difference between the two to determine their relative stabilities which can be translated into the relative amount of time a given protein will spend in either conformation. A glaring limitation of the FEL is that proteins require a large number of variables to fully describe the motion of all atoms therein; one must choose several ‘important’ variables to calculate an FEL which can be visualized. Deciding what makes a variable important and how to identify such variables is a challenge and remains an active area of research.

Timescales in protein dynamics.

Figure 4.
Timescales in protein dynamics.

(a) A generic free energy landscape. Arrows represent transitions between conformations, with the associated rate (k) and time scale (t). Protein-free energy landscapes are hierarchical, with slow transitions between large conformational states (k1) and faster transitions (k2 and k3) between conformational states nested within the larger states. (b) Characteristic length and time scales of conformational changes in eukaryotic PKs. The blue rectangle represents the current capabilities of MD simulations, reflecting the largest (on the order of 100 nm) and longest (on the order of 1 ms) simulations completed to date. Kinase cascades span the scales characteristic of PK–PK interaction to signaling networks across the cytoplasm of a cell.

Figure 4.
Timescales in protein dynamics.

(a) A generic free energy landscape. Arrows represent transitions between conformations, with the associated rate (k) and time scale (t). Protein-free energy landscapes are hierarchical, with slow transitions between large conformational states (k1) and faster transitions (k2 and k3) between conformational states nested within the larger states. (b) Characteristic length and time scales of conformational changes in eukaryotic PKs. The blue rectangle represents the current capabilities of MD simulations, reflecting the largest (on the order of 100 nm) and longest (on the order of 1 ms) simulations completed to date. Kinase cascades span the scales characteristic of PK–PK interaction to signaling networks across the cytoplasm of a cell.

Beyond the basic MD simulations we have described above, often referred to as unbiased MD for reasons which will become clear shortly, many other computational methods for studying protein conformational dynamics exist. One class of these methods is biased MD simulation, where the potential energy function determining the motion of a protein is altered in some way from the realistic approximate energy function of unbiased MD, promoting conformational changes that could otherwise take prohibitively long times to occur [64]. Because the bias added to the potential energy function is a known quantity, one can account for the bias in the analysis stage and produce the same FEL that might have come from unbiased simulation, but with less computational time required. Despite the potential for increased efficiency in biased simulations, it can be a challenge to determine how exactly to add the biasing energy to a simulation in order to actually improve efficiency [65]. Another method that can be used involves creating coarse-grained models of a protein structure, where multiple atoms are combined into representative units and the motion of these units is calculated [6668]. This strategy provides several advantages, decreasing the number of calculations a computer must do and thus allowing simulation of larger systems for longer periods of time, while sacrificing the detailed description of dynamics provided by all-atom MD [68].

MD simulation and structural biology: a successful approach for animal PKs

We have discussed the general benefits of MD simulation in the previous section, but how has MD contributed to the study of PKs? For animal PKs, an approach combining structural biology experiments with MD simulation has uncovered the dynamic nature of PK structure for many different specific proteins [69]. MD simulations of PKs have uncovered the effects of phosphorylation and ATP binding, including induction of conformational change along activation pathways [7077], functional mechanisms of activating mutations [7880], binding properties of inhibitors [8185], as well as allosteric interactions between different structural domains [86,87]. All of these properties are difficult to access experimentally but are critical for fully understanding, and eventually manipulating, PKs as the molecular switches that they are. It is precisely because of these unique abilities that MD simulation has played such an important role in animal PK research, and will very likely continue to do so for years to come. Although MD simulation provides unique insights into PK dynamic behavior, the successes of MD as a method of studying PKs have come through integration with experimental results. Information can flow either way; MD simulation is capable of providing a wide range of experimentally testable predictions, while experimental results can be used to improve the accuracy and efficiency of simulations. For example, observables from NMR spectroscopy experiments can be compared with the same quantities calculated from MD simulations or used to restrain atoms in simulations in such a way as to ensure agreement with experimental values [88,89]. The idea behind introducing NMR restraints is to account for inaccuracies in the empirical force fields, with the intent of improving the predictive power of simulations for quantities not included in the restraints. As opposed to the quantitative physical properties measured in NMR, MD simulations can also be used to make qualitative predictions, such as which residues are important for protein function or stability [9092]. This type of prediction can be tested through site-directed mutagenesis coupled with functional assays. Inversely, deep mutational scanning [93] can be used to experimentally identify functionally important residues in a protein, and MD simulation can be used to understand why a particular residue is important [94].

Activation pathways and post-translational modifications

Out of the PK properties typically studied by MD simulation, the activation pathway, the most probable sequence of conformations taken to travel from the inactive to active state, is perhaps of the most fundamental importance. Characterizing how phosphorylation or release of an inhibiting domain leads to PK activation is analogous to understanding how gears and levers allow the flip of a mechanical switch to perform its ultimate function. However, PK activation pathways are difficult to characterize due to the long timescales over which conformational changes take place, meaning that large amounts of computer time are needed to observe an activation event, let alone other modes of PK motion [71,95].

Still, many groups have had success either through clever use of enhanced sampling techniques or through brute force simulation. Shukla and co-workers [71] used extensive MD simulations performed on the Folding@home distributed computing platform [96] to study the activation pathway of human c-Src kinase, describing the conformational path taken between crystal structures of the active and inactive states (Figure 5). While this transition is driven in vivo by phosphorylation, the authors utilized a sophisticated algorithm, called the string method in collective variables [97,98], in order to identify a likely transition path for unphosphorylated c-Src. The structures along this pathway were then used to start simulations which were, in turn, used to create a Markov state model (MSM). MSMs are approximate kinetic models created by grouping structurally similar conformations into discrete states and calculating rates of interconversion between states from the original MD trajectories [99,100]. Through this approach, they were able to predict the presence of intermediate conformational states, one of which was found to be similar to an inhibitor-bound X-ray crystal structure of cyclin-dependent kinase 2 (CDK2) [101]. Further simulations of c-Src bound to the CDK2 inhibitor revealed that the inhibitor stabilized the c-Src inactive state, demonstrating the presence of an inhibitor-binding pocket present in the conformations between the active and inactive states.

Activation of human c-Src kinase.

Figure 5.
Activation of human c-Src kinase.

(a) A free energy landscape of c-Src, where the x-axis is the root mean-squared distance (RMSD) of the A-loop from the inactive crystal structure, and the y-axis is the difference in the distances between the conserved Glu in the αC helix from an Arg in the A-loop and the same Glu from the conserved Lys in the N-lobe. The inactive and active structures, representing known crystal structures, are shown together with the pathway between the two and two intermediate states identified through MD simulation. (b) Images of the four conformations indicated on the free energy landscape. Adapted with permission from Shukla et al. [71].

Figure 5.
Activation of human c-Src kinase.

(a) A free energy landscape of c-Src, where the x-axis is the root mean-squared distance (RMSD) of the A-loop from the inactive crystal structure, and the y-axis is the difference in the distances between the conserved Glu in the αC helix from an Arg in the A-loop and the same Glu from the conserved Lys in the N-lobe. The inactive and active structures, representing known crystal structures, are shown together with the pathway between the two and two intermediate states identified through MD simulation. (b) Images of the four conformations indicated on the free energy landscape. Adapted with permission from Shukla et al. [71].

The study of PK response to PTMs and protein–protein interaction is intimately linked to the study of activation pathways and is another integral part of the structural research on PKs. Past research has focused on the effects of PTMs or interacting proteins on both the protein conformational FEL [7476] and directly on the efficiency of the phosphoryl transfer reactions catalyzed by PKs [72]. In an example of the former case, Kuzmanic and co-workers examined the effects of A-loop phosphorylation and ATP binding to the human PK p38α using an enhanced MD sampling method called parallel tempering metadynamics [76,102]. p38α was simulated when unphosphorylated, doubly phosphorylated, doubly phosphorylated with ATP bound, and doubly phosphorylated with ATP and a docking peptide bound in order to study the effects of these factors on the p38α FEL. They found that in order for a stable state representing the active-like crystal structure to occur, phosphorylation, ATP binding, and docking peptide binding must all occur, even though only phosphorylation is present in the crystal structure. However, this apparent contradiction was resolved by showing that in simulations of a crystal super cell, the crystal contacts stabilized the active-like state leading to the discrepancies between X-ray crystal structures and MD simulation together with prior NMR results.

Given the high levels of PK structural conservation across the eukaryotes [10], we expect that the insights into PK activation gained from MD simulation of animal PKs to be of great use from understanding plant PK activation. The principles of R-spine alignment, including αC helix placement and DFG positioning, and A-loop unfolding together with other conserved motifs mentioned above, can be used as starting points for the study of plant PKs, although the possibility for new regulatory motifs remains enticing given the large size of plant kinomes [36].

Understanding PK dysfunction: oncogenic mutations and drug design

While a mechanistic understanding of normal PK function is of great scientific value, the importance of abnormally functioning PKs in cancer brings the focus of biomedical research toward understanding how mutations lead to pathological PK function and how inhibitors might be designed to treat specific cancers by blocking the function of overactive PKs. Through MD simulations, researchers can correlate differences in residue interactions upon mutation with shifts in conformational dynamics in order to understand how mutations alter PK function. Shan et al. [78] took this approach to study how an oncogenic mutation in the human epidermal growth factor receptor (EGFR) kinase domain leads to pathogenic activation. EGFR is a transmembrane receptor tyrosine kinase which forms homodimers upon binding of EGF [103], although the L834R mutation in the kinase domain causes constitutive EGFR activation [104]. Using MD simulations validated by biochemical experiments, Shan and co-workers showed that the αC helix of the wild-type EGFR kinase domain displayed a high degree of disorder in monomeric form, while dimerization suppressed this disorder leading to activation. Simulation of the L834R mutant revealed suppressed disorder in the αC helix, promoting dimerization and therefore activation of EGFR.

Oncogenic PKs are a common target for the development of cancer drugs [105]. Numerous studies have used MD simulation in order to characterize inhibitor binding to PKs, including attempts to explain differences in inhibitor affinity to different PKs [81,83,106] and to discover binding pathways [82]. Binding pathways, in particular, are uniquely accessible by MD simulation, where alternative binding sites and important protein residue–inhibitor interactions can be observed [82]. Both approaches to understanding the impact of cancer-causing mutations on animal PKs and designing inhibitory drugs could be extended to genetic engineering of plant PKs and the development of agrochemicals targeting plant PKs. In terms of agrochemical development, the established field of plant chemical genetics [107], where the focus of research is on screening small molecules for biological activity and identifying target proteins, could be complemented by the knowledge of plant PK structure and dynamics. Aside from fundamental differences in the goals of applying methodologies previously used to understand and treat cancer (increasing rather than inhibiting growth), the basic principles appear almost wholly transferrable to problems in plant cellular signaling and crop engineering.

MD simulation and structural biology: a successful approach for plant PKs?

Despite the successes achieved for animal PKs through marrying structural experiments and MD simulation, there has been relatively little progress on either front for plant PKs. Of the 940 known PKs in the A. thaliana kinome, only 27 have an entry in the PDB, 11 of which contain any part of a kinase domain.4 Compared with the human kinome, the structural features of plant kinomes are largely uncharacterized (Figure 1b), while details of conformational dynamics are practically unknown. In this section, we review what progress has been made on understanding the structure and dynamics of plant PKs.

In terms of molecular simulation, very little work was focused on plant PKs, let alone plant proteins in general. Two studies [108,109] have focused on calculating binding enthalpy of members of the brassinosteroid plant hormone group to their primary receptors, BRASSINOSTEROID-INSENSITIVE 1 (BRI1) and BAK1, both leucine-rich repeat receptor-like kinases. As brassinosteroids bind to the extracellular domains of BRI1 and BAK1 across the cell membrane from the cytoplasmic kinase domains (Figure 6), these studies provide important insights into brassinosteroid signaling but are of little relevance to the study of the plant PKs. As far as we are aware, the only MD simulation studies, to date, on plant PK kinase domains are our own, focusing on the conformational dynamics of BRI1 and BAK1 [110,111]. In our first paper, we sought to characterize the long-term conformational dynamics of the BRI1 and BAK1 kinase domains in their active phosphorylation states. In a second paper, we examined the structural mechanism of BAK1 deactivation by S-glutathionylation, which was observed from in vitro experiments [112]. The remainder of this section will be dedicated to reviewing the results of these two papers.

A cartoon representation of the current state of structural knowledge for brassinosteroid signaling initiation.

Figure 6.
A cartoon representation of the current state of structural knowledge for brassinosteroid signaling initiation.

A brassinosteroid hormone (brassinolide, abbreviated as BL) first binds to the BRI1-binding pocket (left), inducing BAK1 association with BRI1 (center), and/or activating a preformed BRI1–BAK1 dimer (not shown). The BRI1 and BAK1 kinase domains in the complex then phosphorylate one another (right), leading to downstream signaling and ultimately changes in gene expression relating to plant growth and development. The question marks denote several aspects of the signaling process that remain unknown from a structural–dynamics point of view. First, it is unclear how BL binds to BRI1 (left) and how this leads to activation of BRI1 and BAK1 (right). Relatedly, it is also unclear how BRI1 and BAK1 interact in their transmembrane and kinase domains.

Figure 6.
A cartoon representation of the current state of structural knowledge for brassinosteroid signaling initiation.

A brassinosteroid hormone (brassinolide, abbreviated as BL) first binds to the BRI1-binding pocket (left), inducing BAK1 association with BRI1 (center), and/or activating a preformed BRI1–BAK1 dimer (not shown). The BRI1 and BAK1 kinase domains in the complex then phosphorylate one another (right), leading to downstream signaling and ultimately changes in gene expression relating to plant growth and development. The question marks denote several aspects of the signaling process that remain unknown from a structural–dynamics point of view. First, it is unclear how BL binds to BRI1 (left) and how this leads to activation of BRI1 and BAK1 (right). Relatedly, it is also unclear how BRI1 and BAK1 interact in their transmembrane and kinase domains.

BRI1 and BAK1 conformational dynamics

Despite the importance of brassinosteroid signaling through BRI1 and BAK1 for plant growth and development [40,113,114], the atomic-level details of early BRI1/BAK1 signaling events are largely unknown. In the generally accepted model of brassinosteroid signaling initiation, BRI1 binds to a single BL molecule, promoting association with BAK1 and reciprocal phosphorylation of the BRI1 and BAK1 kinase domains (Figure 6) [20]. In addition to kinase domain phosphorylation, the largely unstructured BRI1 kinase inhibitor 1 is phosphorylated and subsequently released [115]. The activated BRI1 kinase domain then phosphorylates downstream targets, ultimately leading to activation of the transcription factors BRASSINOZOLE-RESISTANT 1 and BRI1-EMS-SUPPRESSOR 1 [116] and alteration of gene expression controlling plant growth and development. While this qualitative model has been pieced together through years of experiments [20,117119], many questions remain concerning the precise mechanisms of BRI1 and BAK1 activation. For example, it is currently unknown how the BRI1–BAK1 transmembrane helices and kinase domains interact with one another and how BL binding activates BRI1 and BAK1. Improved understanding of the molecular basis for plant growth signaling will not only mark a significant advance in fundamental plant science, but also could inform rational genetic engineering efforts and agrochemical development targeting brassinosteroid signaling pathways to improve crop yield and stress resistance.

To address this knowledge gap, we sought to characterize the conformational dynamics of the BRI1 and BAK1 kinase domains in ATP-bound and fully phosphorylated forms by performing MD simulations on both kinase domains separately. We found that much like p38α [76], phosphorylation and ATP binding were insufficient for either protein to maintain active-like structures similar to their crystal structures. For BAK1, we found that the αC helix had a high propensity for disorder (Figure 7a), much like EGFR [78], while the BRI1 αC helix adopted both disordered and Src-like [121] inactive conformations (Figure 7b). To test whether the high degree of BAK1 αC helix disorder was due to an intrinsically disordered sequence or the surrounding BAK1 residues, we substituted the stable αC helix of human c-Src into our BAK1 computational model and performed further MD simulations. The BAK1–Src chimera exhibited an enhanced level of αC helix stability, suggesting that intrinsic disorder in the native BAK1 αC helix sequence is at least partly responsible for its unfolding in simulations. We also performed circular dichroism experiments on the BRI1 αC helix peptide in order to characterize its propensity for disorder in isolation, finding that the BRI1 αC helix is less disordered than the EGFR αC helix, but more disordered than the c-Src αC helix.

BRI1 and BAK1 kinase domain dynamics.

Figure 7.
BRI1 and BAK1 kinase domain dynamics.

(a) A free energy landscape of the BRI1 kinase domain, with the α helical content of the αC helix shown on the x-axis and the distance between the conserved Glu on the αC helix and the conserved Lys above the ATP-binding site shown on the y-axis. Structures representing the indicated regions of the free energy landscape are shown. (b) A free energy landscape of the BAK1 kinase, with the same axes as for BRI1. (c) A histogram of predicted disorder propensity in the putative αC helix regions of all PKs in the A. thaliana kinome together with human EGFR and c-Src, for reference. Disorder propensity was predicted using the PONDR-FIT webserver [120]. (d) A phylogenetic tree of the A. thaliana kinome, with predicted αC helix disorder propensity shown for each PK. Adapted with permission from Moffett et al. [110].

Figure 7.
BRI1 and BAK1 kinase domain dynamics.

(a) A free energy landscape of the BRI1 kinase domain, with the α helical content of the αC helix shown on the x-axis and the distance between the conserved Glu on the αC helix and the conserved Lys above the ATP-binding site shown on the y-axis. Structures representing the indicated regions of the free energy landscape are shown. (b) A free energy landscape of the BAK1 kinase, with the same axes as for BRI1. (c) A histogram of predicted disorder propensity in the putative αC helix regions of all PKs in the A. thaliana kinome together with human EGFR and c-Src, for reference. Disorder propensity was predicted using the PONDR-FIT webserver [120]. (d) A phylogenetic tree of the A. thaliana kinome, with predicted αC helix disorder propensity shown for each PK. Adapted with permission from Moffett et al. [110].

Finding disordered αC helices in BRI1 and BAK1 led us to ask how common this feature is across entire plant kinomes. Accordingly, we used the PONDR-FIT disorder-prediction webserver [120] to predict disorder in putative αC helix regions of all 940 known A. thaliana PKs [16], finding that αC helix disorder could be a common attribute in plant kinomes (Figure 7c,d). As phosphorylation was insufficient in our simulations to keep the kinase overwhelmingly in an active-like state, we proposed that some other factor not included in our simulations may play a role in BRI1 and BAK1 activation by suppressing αC helix disorder, possibly the actual physical interaction between the two kinase domains. In any case, as αC helix disorder is predicted to be widespread in the A. thaliana kinome, it may be a widely used regulatory mechanism in plant PKs.

BAK1 deactivation by S-glutathionylation

In addition to activation by phosphorylation, BAK1 can be deactivated through S-glutathionylation [112]. S-glutathionylation is a post-translational modification where glutathione, a ubiquitous tripeptide shown to be critical for normal function of several plant and animal species [122,123], forms a disulfide bond with a Cys residue. While Bender and co-workers identified three possible S-glutathionylation sites on BAK1, it was unclear which site or which combinations of sites were responsible for BAK1 deactivation as well as how, at the molecular level, S-glutathionylation was able to deactivate BAK1. To understand the mechanism of BAK1 deactivation by S-glutathionylation, we performed MD simulations of BAK1 singly modified at each possible site for a total of three simulation systems. Using the unglutathionylated BAK1 simulations as a reference, we investigated the effects of BAK1 S-glutathionylation on its FEL. We calculated the Kullback–Leibler divergence, a measure of similarity between probability distributions, between the distributions of residue conformations in S-glutathionylated and unglutathionylated BAK1 simulations, finding that modification of one site, C408, had a large effect on BAK1 conformation across the kinase domain (Figure 8a,b). Further supporting C408 as an inhibitory S-glutathionylation site is the shift of the BAK1 FEL away from the active-like structure with C408 S-glutathionylation, while modification of other sites yielded only slightly changed landscapes (Figure 8c). While C408 S-glutathionylation induced long-range allosteric effects across the BAK1 kinase domain, αC helix unfolding is promoted through direct interaction with glutathione.

Deactivation of BAK1 by S-glutathionylation.

Figure 8.
Deactivation of BAK1 by S-glutathionylation.

(a and b) The local Kullback–Liebler divergence (KLD) of S-glutathionylated BAK1 residues with respect to non-glutathionylated BAK1 from MD simulations. (a) Local KLD of the three S-glutathionylated BAK1 systems displayed using the PyMOL B-factor putty representation [124]. The color scale shows KL divergence on separate scales for each simulation set. (b) Local KLD plotted for each S-glutathionylated BAK1 system by residue. (c) Free energy landscapes of BAK1 in non-glutathionylated and S-glutathionylated forms, over the inverse distance between Lys and Glu in the conserved Lys-Glu ionic interaction and αC helix helical content. Adapted with permission from Moffett et al. [111].

Figure 8.
Deactivation of BAK1 by S-glutathionylation.

(a and b) The local Kullback–Liebler divergence (KLD) of S-glutathionylated BAK1 residues with respect to non-glutathionylated BAK1 from MD simulations. (a) Local KLD of the three S-glutathionylated BAK1 systems displayed using the PyMOL B-factor putty representation [124]. The color scale shows KL divergence on separate scales for each simulation set. (b) Local KLD plotted for each S-glutathionylated BAK1 system by residue. (c) Free energy landscapes of BAK1 in non-glutathionylated and S-glutathionylated forms, over the inverse distance between Lys and Glu in the conserved Lys-Glu ionic interaction and αC helix helical content. Adapted with permission from Moffett et al. [111].

The future of MD and the plant kinome

Toward the lofty goal of characterizing the dynamics of any large set of proteins, it will be necessary to remain on the cutting edge of molecular simulation methodology. While the analogy of a computational microscope paints a vivid picture of the capabilities of MD simulation, it is important not to take the analogy too literally since the accuracy of simulation results is limited by the error inherent to numerical simulation, the approximate nature of calculated interatomic forces, and the high computational cost necessary for simulations on biologically relevant timescales [56]. Despite all of the advances in MD methodology, there is much room for improvement. Perhaps, most clear among possible strategies for improving MD simulation are further integration of experimental results into simulations [88,125], benefitting both the experimentalist and the computational scientist. Additionally, successes of coarse-grained [126] and quantum [72] molecular modeling applied to animal PKs, allowing the study of larger and smaller spatio-temporal scales than classical all-atom MD, respectively, demonstrate the potential of methodological developments in other simulation methods for the study of plant PKs.

For the use of physics-based computational models to find success in the study of plant PKs, and plant proteins in general, it will be necessary for future generations of plant biologists to be trained in quantitative thinking and methods [127]. As we have pointed out, very little structural information exists concerning plant PKs despite their enormous number within plant species [36]. Researchers with strong backgrounds in physics, mathematics, and computational science will be necessary to uncover the atom-level details of plant PKs and in doing so, guide future crop innovation. We emphasize that this directive should not be limited to the PKs; the study of plant proteins, in general, will greatly benefit from realization of the vision outlined by many leading plant biologists [127].

Flowering plants tend to have large kinomes, ranging from 600 to 2500 members among analyzed species [36]. In contrast, the human kinome consists of ∼500 members [15]. A possible explanation is that plants have such extensive kinomes due to their need to respond to constantly changing external stimuli without the benefits of motility or a central nervous system [36]. Regardless of the reasons for the size of plant kinomes, it is clear that plant PKs are underrepresented in the PDB with 296 kinase domain entries from unique Homo sapiens proteins (55.0% of the human kinome) when compared with 27 entries from unique A. thaliana proteins (2.9% of the A. thaliana kinome), 2 entries from Z. mays, and 1 entry from S. pimpinellifolium. While the focus on human PKs is understandable due to their importance in disease, we argue that a research program is needed for plant PKs mirroring that of human PKs. First, on scientific grounds, plant kinomes represent fertile ground for new discoveries, given the large number of PKs with no structural or dynamic characterization. While the catalytically active conformations of PKs are necessarily similar to one another, the inactive conformations of animal PKs are far more diverse, reflecting the effectively infinite number of ways a protein can be catalytically incompetent for a particular reaction [121,128]. Second, on practical grounds, understanding how plant PKs function could greatly improve genetic engineering and rational agrochemical design attempts to improve crop growth or stress resistance. For both of these reasons, it is time for the era of plant PK structural and dynamic proteomics.

Abbreviations

     
  • ABA

    abscisic acid

  •  
  • A-loop

    activation loop

  •  
  • BAK1

    BRI1-ASSOCIATED KINASE 1

  •  
  • BRI1

    BRASSINOSTEROID-INSENSITIVE 1

  •  
  • CDK2

    cyclin-dependent kinase 2

  •  
  • C-loop

    catalytic loop

  •  
  • EGFR

    epidermal growth factor receptor

  •  
  • FEL

    free energy landscape

  •  
  • MD

    molecular dynamics

  •  
  • MSM

    Markov state model

  •  
  • NMR

    nuclear magnetic resonance

  •  
  • PDB

    Protein Data Bank

  •  
  • PK

    protein kinase

  •  
  • PTMs

    post-translational modifications

  •  
  • R-spine

    regulatory spine

  •  
  • SnRK2.6

    2 Snfl-related kinase 6

Acknowledgments

We thank Shukla research group members for their critical reading of this manuscript. We also thank Steven C. Huber for the discussion of the topics covered in this review. A.S.M. and D.S. acknowledge the University of Illinois at Urbana-Champaign for the support.

Competing Interests

The Authors declare that there are no competing interests associated with the manuscript.

*

The number of unique entries in the PDB for plant PK domains is higher than reported here, as we noted that the enzyme classification for at least several entries was misannotated in downloaded reports, classifying protein kinases as non-protein kinases. For the sake of simplicity, we have reported the numbers according to these reports.

For the sake of brevity, we avoid discussion of several crucial steps in between securing input information and actual MD simulation as they are not important for the overall description of MD simulation. At the same time, the omitted steps are oftentimes among the most challenging in the entire process and cannot be ignored in a more thorough discussion.

In practice, this simple algorithm for calculating trajectories is not used due to the high degree of error it incurs. This error arises due to the necessarily non-vanishing values of Δt used in computer simulations, where v(t) and a(t) are approximated as constants over the period of this time step.

§As mentioned earlier for maize and tomato PK structures, this number could be an underestimate if the misannotation of enzyme classifications is a recurring issue in the PDB.

References

References
1
Swaney
,
K.F.
,
Huang
,
C.H.
and
Devreotes
,
P.N.
(
2010
)
Eukaryotic chemotaxis: a network of signaling pathways controls motility, directional sensing, and polarity
.
Annu. Rev. Biophys.
39
,
265
289
2
Zhang
,
W.
and
Liu
,
H.T.
(
2002
)
MAPK signal pathways in the regulation of cell proliferation in mammalian cells
.
Cell Res.
12
,
9
18
3
Tirziu
,
D.
,
Giordano
,
F.J.
and
Simons
,
M.
(
2010
)
Cell communications in the heart
.
Circulation
122
,
928
937
4
Schenk
,
P.W.
and
Snaar-Jagalska
,
B.E.
(
1999
)
Signal perception and transduction: the role of protein kinases
.
Biochim. Biophys. Acta, Mol. Cell Res.
1449
,
1
24
PMID:
[PubMed]
5
Kholodenko
,
B.N.
(
2006
)
Cell-signalling dynamics in time and space
.
Nat. Rev. Mol. Cell Biol.
7
,
165
176
6
Lahiry
,
P.
,
Torkamani
,
A.
,
Schork
,
N.J.
and
Hegele
,
R.A.
(
2010
)
Kinase mutations in human disease: interpreting genotype-phenotype relationships
.
Nat. Rev. Genet.
11
,
60
74
7
Leicht
,
D.T.
,
Balan
,
V.
,
Kaplun
,
A.
,
Singh-Gupta
,
V.
,
Kaplun
,
L.
,
Dobson
,
M.
et al. 
(
2007
)
Raf kinases: function, regulation and role in human cancer
.
Biochim. Biophys. Acta, Mol. Cell Res.
1773
,
1196
1212
8
Antal
,
C.E.
,
Hudson
,
A.M.
,
Kang
,
E.
,
Zanca
,
C.
,
Wirth
,
C.
,
Stephenson
,
N.L.
et al. 
(
2015
)
Cancer-associated protein kinase C mutations reveal kinase's role as tumor suppressor
.
Cell
160
,
489
502
9
Endicott
,
J.A.
,
Noble
,
M.E.
and
Johnson
,
L.N.
(
2012
)
The structural basis for control of eukaryotic protein kinases
.
Annu. Rev. Biochem.
81
,
587
613
10
Taylor
,
S.S.
and
Kornev
,
A.P.
(
2011
)
Protein kinases: evolution of dynamic regulatory proteins
.
Trends Biochem. Sci.
36
,
65
77
11
Liu
,
Y.
and
Gray
,
N.S.
(
2006
)
Rational design of inhibitors that bind to inactive kinase conformations
.
Nat. Chem. Biol.
2
,
358
364
12
Zhang
,
J.
,
Yang
,
P.L.
and
Gray
,
N.S.
(
2009
)
Targeting cancer with small molecule kinase inhibitors
.
Nat. Rev. Cancer
9
,
28
39
13
Cheng
,
W.
,
Munkvold
,
K.R.
,
Gao
,
H.
,
Mathieu
,
J.
,
Schwizer
,
S.
,
Wang
,
S.
et al. 
(
2011
)
Structural analysis of Pseudomonas syringae AvrPtoB bound to host BAK1 reveals two similar kinase-interacting domains in a type III effector
.
Cell Host Microbe
10
,
616
626
14
Humphrey
,
W.
,
Dalke
,
A.
and
Schulten
,
K.
(
1996
)
VMD: visual molecular dynamics
.
J. Mol. Graphics
14
,
33
38
15
Manning
,
G.
,
Whyte
,
D.B.
,
Martinez
,
R.
,
Hunter
,
T.
and
Sudarsanam
,
S.
(
2002
)
The protein kinase complement of the human genome
.
Science
298
,
1912
1934
16
Zulawski
,
M.
,
Schulze
,
G.
,
Braginets
,
R.
,
Hartmann
,
S.
and
Schulze
,
W.X.
(
2014
)
The Arabidopsis kinome: phylogeny and evolutionary insights into functional diversification
.
BMC Genomics
15
,
548
17
Berman
,
H.M.
,
Westbrook
,
J.
,
Feng
,
Z.
,
Gilliland
,
G.
,
Bhat
,
T.
,
Weissig
,
H.
et al. 
(
2000
)
The protein data bank
.
Nucleic Acids Res.
28
,
235
242
18
Hunter
,
J.D.
(
2007
)
Matplotlib: a 2D graphics environment
.
Comput. Sci. Eng.
9
,
90
95
19
Ozkirimli
,
E.
and
Post
,
C.B.
(
2006
)
Src kinase activation: a switched electrostatic network
.
Protein Sci.
15
,
1051
1062
20
Wang
,
X.
,
Kota
,
U.
,
He
,
K.
,
Blackburn
,
K.
,
Li
,
J.
,
Goshe
,
M.B.
et al. 
(
2008
)
Sequential transphosphorylation of the BRI1/BAK1 receptor kinase complex impacts early events in brassinosteroid signaling
.
Dev. Cell
15
,
220
235
21
Yan
,
L.
,
Ma
,
Y.
,
Liu
,
D.
,
Wei
,
X.
,
Sun
,
Y.
,
Chen
,
X.
et al. 
(
2012
)
Structural basis for the impact of phosphorylation on the activation of plant receptor-like kinase BAK1
.
Cell Res.
22
,
1304
1308
22
Wang
,
Y.
,
Li
,
Z.
,
Liu
,
D.
,
Xu
,
J.
,
Wei
,
X.
,
Yan
,
L.
et al. 
(
2014
)
Assessment of BAK1 activity in different plant receptor-like kinase complexes by quantitative profiling of phosphorylation patterns
.
J. Proteomics
108
,
484
493
23
Bojar
,
D.
,
Martinez
,
J.
,
Santiago
,
J.
,
Rybin
,
V.
,
Bayliss
,
R.
and
Hothorn
,
M.
(
2014
)
Crystal structures of the phosphorylated BRI1 kinase domain and implications for brassinosteroid signal initiation
.
Plant J.
78
,
31
43
24
Wang
,
J.
,
Jiang
,
J.
,
Wang
,
J.
,
Chen
,
L.
,
Fan
,
S.L.
,
Wu
,
J.W.
et al. 
(
2014
)
Structural insights into the negative regulation of BRI1 signaling by BRI1-interacting protein BKI1
.
Cell Res.
24
,
1328
1341
25
Wheeler
,
T.
and
Von Braun
,
J.
(
2013
)
Climate change impacts on global food security
.
Science
341
,
508
513
26
Long
,
S.P.
,
Marshall-Colon
,
A.
and
Zhu
,
X.G.
(
2015
)
Meeting the global food demand of the future by engineering crop photosynthesis and yield potential
.
Cell
161
,
56
66
27
Marshall
,
A.
,
Aalen
,
R.B.
,
Audenaert
,
D.
,
Beeckman
,
T.
,
Broadley
,
M.R.
,
Butenko
,
M.A.
et al. 
(
2012
)
Tackling drought stress: receptor-like kinases present new approaches
.
Plant Cell
24
,
2262
2278
28
Saijo
,
Y.
,
Hata
,
S.
,
Kyozuka
,
J.
,
Shimamoto
,
K.
and
Izui
,
K.
(
2000
)
Over-expression of a single Ca2+-dependent protein kinase confers both cold and salt/drought tolerance on rice plants
.
Plant J.
23
,
319
327
29
Umezawa
,
T.
,
Yoshida
,
R.
,
Maruyama
,
K.
,
Yamaguchi-Shinozaki
,
K.
and
Shinozaki
,
K.
(
2004
)
SRK2C, a SNF1-related protein kinase 2, improves drought tolerance by controlling stress-responsive gene expression in Arabidopsis thaliana
.
Proc. Natl Acad. Sci. U.S.A.
101
,
17306
17311
30
Shou
,
H.
,
Bordallo
,
P.
and
Wang
,
K.
(
2004
)
Expression of the Nicotiana protein kinase (NPK1) enhanced drought tolerance in transgenic maize
.
J. Exp. Bot.
55
,
1013
1019
31
Umezawa
,
T.
,
Fujita
,
M.
,
Fujita
,
Y.
,
Yamaguchi-Shinozaki
,
K.
and
Shinozaki
,
K.
(
2006
)
Engineering drought tolerance in plants: discovering and tailoring genes to unlock the future
.
Curr. Opin. Biotechnol.
17
,
113
122
32
Kovtun
,
Y.
,
Chiu
,
W.L.
,
Tena
,
G.
and
Sheen
,
J.
(
2000
)
Functional analysis of oxidative stress-activated mitogen-activated protein kinase cascade in plants
.
Proc. Natl Acad. Sci. U.S.A.
97
,
2940
2945
33
Gong
,
D.
,
Zhang
,
C.
,
Chen
,
X.
,
Gong
,
Z.
and
Zhu
,
J.K.
(
2002
)
Constitutive activation and transgenic evaluation of the function of an Arabidopsis PKS protein kinase
.
J. Biol. Chem.
277
,
42088
42096
34
Oh
,
M.H.
,
Sun
,
J.
,
Oh
,
D.H.
,
Zielinski
,
R.E.
,
Clouse
,
S.D.
and
Huber
,
S.C.
(
2011
)
Enhancing Arabidopsis leaf growth by engineering the BRASSINOSTEROID INSENSITIVE1 receptor kinase
.
Plant Physiol.
157
,
120
131
35
Hanada
,
K.
,
Zou
,
C.
,
Lehti-Shiu
,
M.D.
,
Shinozaki
,
K.
and
Shiu
,
S.H.
(
2008
)
Importance of lineage-specific expansion of plant tandem duplicates in the adaptive response to environmental stimuli
.
Plant Physiol.
148
,
993
1003
36
Lehti-Shiu
,
M.D.
and
Shiu
,
S.H.
(
2012
)
Diversity, classification and function of the plant protein kinase superfamily
.
Philos. Trans. R Soc. B. Biol. Sci.
367
,
2619
2639
37
Ytterberg
,
A.J.
and
Jensen
,
O.N.
(
2010
)
Modification-specific proteomics in plant biology
.
J. Proteomics
73
,
2249
2266
38
Friso
,
G.
and
van Wijk
,
K.J.
(
2015
)
Posttranslational protein modifications in plant metabolism
.
Plant Physiol.
169
,
1469
1487
39
Soon
,
F.F.
,
Ng
,
L.M.
,
Zhou
,
X.E.
,
West
,
G.M.
,
Kovach
,
A.
,
Tan
,
M.E.
et al. 
(
2012
)
Molecular mimicry regulates ABA signaling by SnRK2 kinases and PP2C phosphatases
.
Science
335
,
85
88
40
Wang
,
W.
,
Bai
,
M.Y.
and
Wang
,
Z.Y.
(
2014
)
The brassinosteroid signaling network — a paradigm of signal integration
.
Curr. Opin. Plant Biol.
21
,
147
153
41
Vitkup
,
D.
,
Melamud
,
E.
,
Moult
,
J.
and
Sander
,
C.
(
2001
)
Completeness in structural genomics
.
Nat. Struct. Mol. Biol.
8
,
559
566
42
Zhang
,
Y.
and
Skolnick
,
J.
(
2005
)
The protein structure prediction problem could be solved using the current PDB library
.
Proc. Natl Acad. Sci. U.S.A.
102
,
1029
1034
43
Skolnick
,
J.
,
Zhou
,
H.
and
Brylinski
,
M.
(
2012
)
Further evidence for the likely completeness of the library of solved single domain protein structures
.
J. Phys. Chem. B
116
,
6654
6664
44
Khafizov
,
K.
,
Madrid-Aliste
,
C.
,
Almo
,
S.C.
and
Fiser
,
A.
(
2014
)
Trends in structural coverage of the protein universe and the impact of the Protein Structure Initiative
.
Proc. Natl Acad. Sci. U.S.A.
111
,
3733
3738
45
Marks
,
D.S.
,
Colwell
,
L.J.
,
Sheridan
,
R.
,
Hopf
,
T.A.
,
Pagnani
,
A.
,
Zecchina
,
R.
et al. 
(
2011
)
Protein 3D structure computed from evolutionary sequence variation
.
PLoS ONE
6
,
e28766
46
Morcos
,
F.
,
Pagnani
,
A.
,
Lunt
,
B.
,
Bertolino
,
A.
,
Marks
,
D.S.
,
Sander
,
C.
et al. 
(
2011
)
Direct-coupling analysis of residue coevolution captures native contacts across many protein families
.
Proc. Natl Acad. Sci. U.S.A.
108
,
E1293
E1301
47
Marks
,
D.S.
,
Hopf
,
T.A.
and
Sander
,
C.
(
2012
)
Protein structure prediction from sequence variation
.
Nat. Biotechnol.
30
,
1072
1080
48
Ovchinnikov
,
S.
,
Kinch
,
L.
,
Park
,
H.
,
Liao
,
Y.
,
Pei
,
J.
,
Kim
,
D.E.
et al. 
(
2015
)
Large-scale determination of previously unsolved protein structures using evolutionary information
.
eLife
4
,
e09248
49
Ovchinnikov
,
S.
,
Park
,
H.
,
Varghese
,
N.
,
Huang
,
P.S.
,
Pavlopoulos
,
G.A.
,
Kim
,
D.E.
et al. 
(
2017
)
Protein structure determination using metagenome sequence data
.
Science
355
,
294
298
50
Fermi
,
E.
,
Pasta
,
J.
,
Ulam
,
S.
and
Tsingou
,
M.
(
1955
)
Studies of Nonlinear Problems. I
,
Las Alamos National Laboratory
51
Rahman
,
A.
(
1964
)
Correlations in the motion of atoms in liquid argon
.
Phys. Rev.
136
,
A405
A411
52
Rahman
,
A.
and
Stillinger
,
F.H.
(
1971
)
Molecular dynamics study of liquid water
.
J. Chem. Phys.
55
,
3336
3359
53
McCammon
,
J.A.
,
Gelin
,
B.R.
and
Karplus
,
M.
(
1977
)
Dynamics of folded proteins
.
Nature
267
,
585
590
54
Kendrew
,
J.
,
Bodo
,
G.
,
Dintzis
,
H.
,
Parrish
,
R.
,
Wyckoff
,
H.
and
Phillips
,
D.
(
1958
)
A three-dimensional model of the myoglobin molecule obtained by x-ray analysis
.
Nature
181
,
662
666
55
Lee
,
E.H.
,
Hsin
,
J.
,
Sotomayor
,
M.
,
Comellas
,
G.
and
Schulten
,
K.
(
2009
)
Discovery through the computational microscope
.
Structure
17
,
1295
1306
56
Dror
,
R.O.
,
Dirks
,
R.M.
,
Grossman
,
J.
,
Xu
,
H.
and
Shaw
,
D.E.
(
2012
)
Biomolecular simulation: a computational microscope for molecular biology
.
Annu. Rev. Biophys.
41
,
429
452
57
Leach
,
A.R.
(
2001
)
Molecular Modelling: Principles and Applications
,
2nd ed.
Pearson Education
,
Harlow, U.K.
58
Voelz
,
V.A.
,
Bowman
,
G.R.
,
Beauchamp
,
K.
and
Pande
,
V.S.
(
2010
)
Molecular simulation of ab initio protein folding for a millisecond folder NTL9 (1-39)
.
J. Am. Chem. Soc.
132
,
1526
1528
59
Lindorff-Larsen
,
K.
,
Piana
,
S.
,
Dror
,
R.O.
and
Shaw
,
D.E.
(
2011
)
How fast-folding proteins fold
.
Science
334
,
517
520
60
Schmid
,
N.
,
Eichenberger
,
A.P.
,
Choutko
,
A.
,
Riniker
,
S.
,
Winger
,
M.
,
Mark
,
A.E.
et al. 
(
2011
)
Definition and testing of the GROMOS force-field versions 54A7 and 54B7
.
Eur. Biophys. J.
40
,
843
856
61
Shi
,
Y.
,
Xia
,
Z.
,
Zhang
,
J.
,
Best
,
R.
,
Wu
,
C.
,
Ponder
,
J.W.
et al. 
(
2013
)
Polarizable atomic multipole-based AMOEBA force field for proteins
.
J. Chem. Theory Comput.
9
,
4046
4063
62
Maier
,
J.A.
,
Martinez
,
C.
,
Kasavajhala
,
K.
,
Wickstrom
,
L.
,
Hauser
,
K.E.
and
Simmerling
,
C.
(
2015
)
ff14SB: improving the accuracy of protein side chain and backbone parameters from ff99SB
.
J. Chem. Theory Comput.
11
,
3696
3713
63
Huang
,
J.
,
Rauscher
,
S.
,
Nawrocki
,
G.
,
Ran
,
T.
,
Feig
,
M.
,
de Groot
,
B.L.
et al. 
(
2017
)
CHARMM36m: an improved force field for folded and intrinsically disordered proteins
.
Nat. Methods
14
,
71
73
64
Bernardi
,
R.C.
,
Melo
,
M.C.
and
Schulten
,
K.
(
2015
)
Enhanced sampling techniques in molecular dynamics simulations of biological systems
.
Biochim. Biophys. Acta, Gen. Subj.
1850
,
872
877
65
Li
,
W.
and
Ma
,
A.
(
2014
)
Recent developments in methods for identifying reaction coordinates
.
Mol. Simul.
40
,
784
793
66
Chebaro
,
Y.
,
Pasquali
,
S.
and
Derreumaux
,
P.
(
2012
)
The coarse-grained OPEP force field for non-amyloid and amyloid proteins
.
J. Phys. Chem. B
116
,
8741
8752
67
de Jong
,
D.H.
,
Singh
,
G.
,
Bennett
,
W.D.
,
Arnarez
,
C.
,
Wassenaar
,
T.A.
,
Schfer
,
L.V.
et al. 
(
2013
)
Improved parameters for the martini coarse-grained protein force field
.
J. Chem. Theory Comput.
9
,
687
697
68
Kmiecik
,
S.
,
Gront
,
D.
,
Kolinski
,
M.
,
Wieteska
,
L.
,
Dawid
,
A.E.
and
Kolinski
,
A.
(
2016
)
Coarse-grained protein models and their applications
.
Chem. Rev.
116
,
7898
7936
69
D'Abramo
,
M.
,
Besker
,
N.
,
Chillemi
,
G.
and
Grottesi
,
A.
(
2014
)
Modeling conformational transitions in kinases by molecular dynamics simulations: achievements, difficulties, and open challenges
.
Front. Genet.
5
,
128
70
Shan
,
Y.
,
Arkhipov
,
A.
,
Kim
,
E.T.
,
Pan
,
A.C.
and
Shaw
,
D.E.
(
2013
)
Transitions to catalytically inactive conformations in EGFR kinase
.
Proc. Natl Acad. Sci. U.S.A.
110
,
7270
7275
71
Shukla
,
D.
,
Meng
,
Y.
,
Roux
,
B.
and
Pande
,
V.S.
(
2014
)
Activation pathway of Src kinase reveals intermediate states as targets for drug design
.
Nat. Commun.
5
,
3397
72
Ojeda-May
,
P.
,
Li
,
Y.
,
Ovchinnikov
,
V.
and
Nam
,
K.
(
2015
)
Role of protein dynamics in allosteric control of the catalytic phosphoryl transfer of insulin receptor kinase
.
J. Am. Chem. Soc.
137
,
12454
12457
73
Meng
,
Y.
,
Shukla
,
D.
,
Pande
,
V.S.
and
Roux
,
B.
(
2016
)
Transition path theory analysis of c-Src kinase activation
.
Proc. Natl Acad. Sci. U.S.A.
113
,
9193
9198
74
Karp
,
J.M.
,
Sparks
,
S.
and
Cowburn
,
D.
(
2017
)
Effects of FGFR2 kinase activation loop dynamics on catalytic activity
.
PLoS Comput. Biol.
13
,
e1005360
75
Li
,
Y.
and
Nam
,
K.
(
2017
)
Dynamic, structural and thermodynamic basis of insulin-like growth factor 1 kinase allostery mediated by activation loop phosphorylation
.
Chem. Sci.
8
,
3453
3464
76
Kuzmanic
,
A.
,
Sutto
,
L.
,
Saladino
,
G.
,
Nebreda
,
A.R.
,
Gervasio
,
F.L.
and
Orozco
,
M.
(
2017
)
Changes in the free-energy landscape of p38α MAP kinase through its canonical activation and binding events as studied by enhanced molecular dynamics simulations
.
eLife
6
,
e22175
77
Perdios
,
L.
,
Lowe
,
A.R.
,
Saladino
,
G.
,
Bunney
,
T.D.
,
Thiyagarajan
,
N.
,
Alexandrov
,
Y.
et al. 
(
2017
)
Conformational transition of FGFR kinase activation revealed by site-specific unnatural amino acid reporter and single molecule FRET
.
Sci. Rep.
7
,
39841
78
Shan
,
Y.
,
Eastwood
,
M.P.
,
Zhang
,
X.
,
Kim
,
E.T.
,
Arkhipov
,
A.
,
Dror
,
R.O.
et al. 
(
2012
)
Oncogenic mutations counteract intrinsic disorder in the EGFR kinase and promote receptor dimerization
.
Cell
149
,
860
870
79
Sutto
,
L.
and
Gervasio
,
F.L.
(
2013
)
Effects of oncogenic mutations on the conformational free-energy landscape of EGFR kinase
.
Proc. Natl Acad. Sci. U.S.A.
110
,
10616
10621
80
Meng
,
Y.
and
Roux
,
B.
(
2016
)
Computational study of the W260A activating mutant of Src tyrosine kinase
.
Protein Sci.
25
,
219
230
81
Aleksandrov
,
A.
and
Simonson
,
T.
(
2010
)
Molecular dynamics simulations show that conformational selection governs the binding preferences of imatinib for several tyrosine kinases
.
J. Biol. Chem.
285
,
13807
13815
82
Shan
,
Y.
,
Kim
,
E.T.
,
Eastwood
,
M.P.
,
Dror
,
R.O.
,
Seeliger
,
M.A.
and
Shaw
,
D.E.
(
2011
)
How does a drug molecule find its target binding site?
J. Am. Chem. Soc.
133
,
9181
9183
83
Lin
,
Y.L.
,
Meng
,
Y.
,
Huang
,
L.
and
Roux
,
B.
(
2014
)
Computational study of Gleevec and G6G reveals molecular determinants of kinase inhibitor selectivity
.
J. Am. Chem. Soc.
136
,
14753
14762
84
Lovera
,
S.
,
Morando
,
M.
,
Pucheta-Martinez
,
E.
,
Martinez-Torrecuadrada
,
J.L.
,
Saladino
,
G.
and
Gervasio
,
F.L.
(
2015
)
Towards a molecular understanding of the link between imatinib resistance and kinase conformational dynamics
.
PLoS Comput. Biol.
11
,
e1004578
85
Verkhivker
,
G.
(
2016
)
Molecular dynamics simulations and modelling of the residue interaction networks in the BRAF kinase complexes with small molecule inhibitors: probing the allosteric effects of ligand-induced kinase dimerization and paradoxical activation
.
Mol. Biosyst.
12
,
3146
3165
86
McClendon
,
C.L.
,
Kornev
,
A.P.
,
Gilson
,
M.K.
and
Taylor
,
S.S.
(
2014
)
Dynamic architecture of a protein kinase
.
Proc. Natl Acad. Sci. U.S.A.
111
,
E4623
E4631
87
Tse
,
A.
and
Verkhivker
,
G.M.
(
2015
)
Molecular dynamics simulations and structural network analysis of c-Abl and c-Src kinase core proteins: capturing allosteric mechanisms and communication pathways from residue centrality
.
J. Chem. Inf. Model.
55
,
1645
1662
88
Pastor
,
N.
and
Amero
,
C.
(
2015
)
Information flow and protein dynamics: the interplay between nuclear magnetic resonance spectroscopy and molecular dynamics simulations
.
Front. Plant Sci.
6
,
306
89
Boomsma
,
W.
,
Ferkinghoff-Borg
,
J.
and
Lindorff-Larsen
,
K.
(
2014
)
Combining experiments and simulations using the maximum entropy principle
.
PLoS Comput. Biol.
10
,
e1003406
90
Kasson
,
P.M.
,
Ensign
,
D.L.
and
Pande
,
V.S.
(
2009
)
Combining molecular dynamics with Bayesian analysis to predict and evaluate ligand-binding mutations in influenza hemagglutinin
.
J. Am. Chem. Soc.
131
,
11338
11340
91
Koukos
,
P.I.
and
Glykos
,
N.M.
(
2014
)
Folding molecular dynamics simulations accurately predict the effect of mutations on the stability and structure of a vammin-derived peptide
.
J. Phys. Chem. B
118
,
10076
10084
92
Latallo
,
M.
,
Cortina
,
G.
,
Faham
,
S.
,
Nakamoto
,
R.
and
Kasson
,
P.
(
2017
)
Predicting allosteric mutants that increase activity of a major antibiotic resistance enzyme
.
Chem. Sci.
8
,
6484
6492
93
Fowler
,
D.M.
and
Fields
,
S.
(
2014
)
Deep mutational scanning: a new style of protein science
.
Nat. Methods
11
,
801
807
94
Koenig
,
P.
,
Lee
,
C.V.
,
Walters
,
B.T.
,
Janakiraman
,
V.
,
Stinson
,
J.
,
Patapoff
,
T.W.
et al. 
(
2017
)
Mutational landscape of antibody variable domains reveals a switch modulating the interdomain conformational dynamics and antigen binding
.
Proc. Natl Acad. Sci. U.S.A.
114
,
E486
E495
95
Xiao
,
Y.
,
Liddle
,
J.C.
,
Pardi
,
A.
and
Ahn
,
N.G.
(
2015
)
Dynamics of protein kinases: insights from nuclear magnetic resonance
.
Acc. Chem. Res.
48
,
1106
1114
96
Shirts
,
M.
and
Pande
,
V.S.
(
2000
)
COMPUTING: screen savers of the world unite!
.
Science
290
,
1903
1904
97
Maragliano
,
L.
,
Fischer
,
A.
,
Vanden-Eijnden
,
E.
and
Ciccotti
,
G.
(
2006
)
String method in collective variables: minimum free energy paths and isocommittor surfaces
.
J. Chem. Phys.
125
,
024106
98
Pan
,
A.C.
,
Sezer
,
D.
and
Roux
,
B.
(
2008
)
Finding transition pathways using the string method with swarms of trajectories
.
J. Phys. Chem. B
112
,
3432
3440
99
Prinz
,
J.H.
,
Wu
,
H.
,
Sarich
,
M.
,
Keller
,
B.
,
Senne
,
M.
,
Held
,
M.
et al. 
(
2011
)
Markov models of molecular kinetics: generation and validation
.
J. Chem. Phys.
134
,
174105
100
Shukla
,
D.
,
Hernández
,
C.X.
,
Weber
,
J.K.
and
Pande
,
V.S.
(
2015
)
Markov state models provide insights into dynamic modulation of protein function
.
Acc. Chem. Res.
48
,
414
422
101
Betzi
,
S.
,
Alam
,
R.
,
Martin
,
M.
,
Lubbers
,
D.J.
,
Han
,
H.
,
Jakkaraj
,
S.R.
et al. 
(
2011
)
Discovery of a potential allosteric ligand binding site in CDK2
.
ACS Chem. Biol.
6
,
492
501
102
Bussi
,
G.
,
Gervasio
,
F.L.
,
Laio
,
A.
and
Parrinello
,
M.
(
2006
)
Free-energy landscape for β hairpin folding from combined parallel tempering and metadynamics
.
J. Am. Chem. Soc.
128
,
13435
13441
103
Schlessinger
,
J.
(
2000
)
Cell signaling by receptor tyrosine kinases
.
Cell
103
,
211
225
104
Zhang
,
X.
,
Gureasko
,
J.
,
Shen
,
K.
,
Cole
,
P.A.
and
Kuriyan
,
J.
(
2006
)
An allosteric mechanism for activation of the kinase domain of epidermal growth factor receptor
.
Cell
125
,
1137
1149
105
Cohen
,
P.
(
2002
)
Protein kinases–the major drug targets of the twenty-first century?
Nat. Rev. Drug Discov.
1
,
309
315
106
Lin
,
Y.L.
,
Meng
,
Y.
,
Jiang
,
W.
and
Roux
,
B.
(
2013
)
Explaining why Gleevec is a specific and potent inhibitor of Abl kinase
.
Proc. Natl Acad. Sci. U.S.A.
110
,
1664
1669
107
Hicks
,
G.R.
and
Raikhel
,
N.V.
(
2014
)
Plant chemical biology: are we meeting the promise?
Front. Plant Sci.
5
,
108
Lei
,
B.
,
Liu
,
J.
and
Yao
,
X.
(
2015
)
Unveiling the molecular mechanism of brassinosteroids: insights from structure-based molecular modeling studies
.
Steroids
104
,
111
117
109
Nakamura
,
A.
,
Tochio
,
N.
,
Fujioka
,
S.
,
Ito
,
S.
,
Kigawa
,
T.
,
Shimada
,
Y.
et al. 
(
2017
)
Molecular actions of two synthetic brassinosteroids, iso-carbaBL and 6-deoxoBL, which cause altered physiological activities between Arabidopsis and rice
.
PLoS ONE
12
,
e0174015
110
Moffett
,
A.S.
,
Bender
,
K.W.
,
Huber
,
S.C.
and
Shukla
,
D.
(
2017
)
Molecular dynamics simulations reveal the conformational dynamics of Arabidopsis thaliana BRI1 and BAK1 receptor-like kinases
.
J. Biol. Chem.
292
,
12643
12652
111
Moffett
,
A.S.
,
Bender
,
K.W.
,
Huber
,
S.C.
and
Shukla
,
D.
(
2017
)
Allosteric control of a plant receptor kinase through S-glutathionylation
.
Biophys. J.
113
,
2354
2363
112
Bender
,
K.W.
,
Wang
,
X.
,
Cheng
,
G.B.
,
Kim
,
H.S.
,
Zielinski
,
R.E.
and
Huber
,
S.C.
(
2015
)
Glutaredoxin AtGRXC2 catalyses inhibitory glutathionylation of Arabidopsis BRI1-associated receptor-like kinase 1 (BAK1) in vitro
.
Biochem. J.
467
,
399
413
113
Zhu
,
J.Y.
,
Sae-Seaw
,
J.
and
Wang
,
Z.Y.
(
2013
)
Brassinosteroid signalling
.
Development
140
,
1615
1620
114
Wang
,
Z.Y.
,
Bai
,
M.Y.
,
Oh
,
E.
and
Zhu
,
J.Y.
(
2012
)
Brassinosteroid signaling network and regulation of photomorphogenesis
.
Annu. Rev. Genet.
46
,
701
724
115
Wang
,
X.
and
Chory
,
J.
(
2006
)
Brassinosteroids regulate dissociation of BKI1, a negative regulator of BRI1 signaling, from the plasma membrane
.
Science
313
,
1118
1122
116
Belkhadir
,
Y.
and
Jaillais
,
Y.
(
2015
)
The molecular circuitry of brassinosteroid signaling
.
New Phytol.
206
,
522
540
117
Caño-Delgado
,
A.
,
Yin
,
Y.
,
Yu
,
C.
,
Vafeados
,
D.
,
Mora-García
,
S.
,
Cheng
,
J.C.
et al. 
(
2004
)
BRL1 and BRL3 are novel brassinosteroid receptors that function in vascular differentiation in Arabidopsis
.
Development
131
,
5341
5351
118
Santiago
,
J.
,
Henzler
,
C.
and
Hothorn
,
M.
(
2013
)
Molecular mechanism for plant steroid receptor activation by somatic embryogenesis co-receptor kinases
.
Science
341
,
889
892
119
Sun
,
Y.
,
Han
,
Z.
,
Tang
,
J.
,
Hu
,
Z.
,
Chai
,
C.
,
Zhou
,
B.
et al. 
(
2013
)
Structure reveals that BAK1 as a co-receptor recognizes the BRI1-bound brassinolide
.
Cell Res.
23
,
1326
1329
120
Xue
,
B.
,
Dunbrack
,
R.L.
,
Williams
,
R.W.
,
Dunker
,
A.K.
and
Uversky
,
V.N.
(
2010
)
PONDR-FIT: a meta-predictor of intrinsically disordered amino acids
.
Biochim. Biophys. Acta
1804
,
996
1010
121
Jura
,
N.
,
Zhang
,
X.
,
Endres
,
N.F.
,
Seeliger
,
M.A.
,
Schindler
,
T.
and
Kuriyan
,
J.
(
2011
)
Catalytic control in the EGF receptor and its connection to general kinase regulatory mechanisms
.
Mol. Cell
42
,
9
22
122
Rouhier
,
N.
,
Lemaire
,
S.D.
and
Jacquot
,
J.P.
(
2008
)
The role of glutathione in photosynthetic organisms: emerging functions for glutaredoxins and glutathionylation
.
Annu. Rev. Plant Biol.
59
,
143
166
123
Noctor
,
G.
,
Mhamdi
,
A.
,
Chaouch
,
S.
,
Han
,
Y.
,
Neukermans
,
J.
,
Marquez-Garcia
,
B.
et al. 
(
2012
)
Glutathione in plants: an integrated overview
.
Plant Cell Environ.
35
,
454
484
124
Schrödinger
,
L.LC.
(
2015
)
The PyMOL Molecular Graphics System, Version 1.8 https://pymol.org/2/support.html
125
Papaleo
,
E.
(
2015
)
Integrating atomistic molecular dynamics simulations, experiments, and network analysis to study protein dynamics: strength in unity
.
Front. Mol. Biosci.
2
,
28
126
Koland
,
J.G.
(
2014
)
Coarse-grained molecular simulation of epidermal growth factor receptor protein tyrosine kinase multi-site self-phosphorylation
.
PLoS Comput. Biol.
10
,
e1003435
127
Friesner
,
J.
,
Assmann
,
S.M.
,
Bastow
,
R.
,
Bailey-Serres
,
J.
,
Beynon
,
J.
,
Brendel
,
V.
et al. 
(
2017
)
The next generation of training for Arabidopsis researchers: Bioinformatics and quantitative biology
.
Plant Physiol.
175
,
1499
1509
128
Noble
,
M.E.
,
Endicott
,
J.A.
and
Johnson
,
L.N.
(
2004
)
Protein kinase inhibitors: insights into drug design from structure
.
Science
303
,
1800
1805