Conceived and designed the experiments: BJG AAG JAM. Performed the experiments: BJG AAG. Analyzed the data: BJG AAG. Contributed reagents/materials/analysis tools: BJG AAG JAM. Wrote the paper: BJG AAG JAM.
The authors have declared that no competing interests exist.
Ras mediates signaling pathways controlling cell proliferation and development by cycling between GTP- and GDP-bound active and inactive conformational states. Understanding the complete reaction path of this conformational change and its intermediary structures is critical to understanding Ras signaling. We characterize nucleotide-dependent conformational transition using multiple-barrier-crossing accelerated molecular dynamics (aMD) simulations. These transitions, achieved for the first time for wild-type Ras, are impossible to observe with classical molecular dynamics (cMD) simulations due to the large energetic barrier between end states. Mapping the reaction path onto a conformer plot describing the distribution of the crystallographic structures enabled identification of highly populated intermediate structures. These structures have unique switch orientations (residues 25–40 and 57–75) intermediate between GTP and GDP states, or distinct loop3 (46–49), loop7 (105–110), and α5 C-terminus (159–166) conformations distal from the nucleotide-binding site. In addition, these barrier-crossing trajectories predict novel nucleotide-dependent correlated motions, including correlations of α2 (residues 66–74) with α3-loop7 (93–110), loop2 (26–37) with loop10 (145–151), and loop3 (46–49) with α5 (152–167). The interconversion between newly identified Ras conformations revealed by this study advances our mechanistic understanding of Ras function. In addition, the pattern of correlated motions provides new evidence for a dynamic linkage between the nucleotide-binding site and the membrane interacting C-terminus critical for the signaling function of Ras. Furthermore, normal mode analysis indicates that the dominant collective motion that occurs during nucleotide-dependent conformational exchange, and captured in aMD (but absent in cMD) simulations, is a low-frequency motion intrinsic to the structure.
The Ras family of enzymes mediate signaling pathways controlling cell proliferation and development by cycling between active and inactive conformational states. Mutations that affect the ability to switch between states are associated with a variety of cancers. However, details of how the structural changes occur and how mutations affect the fidelity of this process remain to be determined. Here we employ an advanced computational technique, termed accelerated molecular dynamics, to characterize structural transitions and identify novel highly populated transient conformations. Several spatially distant structural regions were found to undergo correlated motions, highlighting a dynamic linkage between the sites of enzymatic reaction and the membrane-interacting C-terminus. In addition, our results indicate that the major motion occurring during the conformational exchange is a low-frequency motion intrinsic to the structure. Hence, features of the characterized transitions likely apply to a large number of structurally similar but functionally diverse nucleotide triphosphatases. These results provide fresh insights into how oncogenic mutations might modulate conformational transitions in Ras.
Ras proteins are guanine nucleotide-dependent conformational switches that couple cell-surface receptors to signaling pathways that mediate cell proliferation, growth and development
Conformational changes and oncogenic mutations are largely concentrated in the vicinity of the nucleotide binding site, including the so-called switch regions SI (residues 25–40) and SII (residues 57–75). Of particular note are the conserved SI threonine (residue 35) and SII glycine (residue 60) which converge to form hydrogen bonds with the γ-phosphate of GTP (effectively ‘closing’ the nucleotide binding pocket). In the absence of the γ-phosphate (or a suitable analogue such as aluminium fluoride (AlF3)) the switch regions display fewer structural contacts to the nucleotide and reside in a more ‘open’ conformation. This observation has been likened to a loaded spring, where release of the γ-phosphate after GTP hydrolysis allows the switch regions to relax into their ‘open’ GDP-bound conformations
In the present study we employ simulation approaches to perform a detailed characterization of the dynamics of nucleotide-dependent conformational transitions. Previous unbiased molecular dynamics (MD) simulations were restricted to characterizing fluctuations within individual nucleotide states
Simulations were conducted with starting structures corresponding to wild-type GDP, wild-type GTP and mutant GDP states of Ras. Each of these systems was simulated with a bound GDP and GTP. Both classical and accelerated MD (cMD and aMD) simulations were performed with explicit solvent for 60 nanoseconds. In addition to conventional structural analysis, which assessed the stability of Ras during the various simulations (see
Crystallographic GTP conformers are colored red whilst GDP conformers are colored green. The distribution of MD conformers is depicted with density-shaded blue points. Each row corresponds to a single initial conformation, namely: (A–D) wtGTP, (E–H) wtGDP and (I–L) mutantGDP. cMD simulations are depicted in the two left panels (A, B, E, F, I, J) whilst aMD simulations are depicted in the two right panels (C, D G, H, K, L). Simulations were performed with bound GTP (A, C, E, G, I, K) and GDP (B, D, F, H, J, L). Inserts show distances between instantaneous trajectory conformations and the centroids of the main GTP and GDP crystal structure clusters in red and green respectively (see
Interestingly, aMD simulations carried out to further probe the apparent low activation barrier of the G12V variant (
In contrast to the absence of evident transitions during cMD for wild-type Ras, notable transitions are sampled during aMD for systems with a swapped nucleotide (i.e. GTP inserted into a starting GDP structure and
Clustering of trajectory conformers was used to visualize the dominant conformations sampled by each simulation (
Front and back views of representative structures obtained from hierarchical clustering (A, B, E and F). In each case the most populated cluster representative is shown in black (representative of 30.26% and 28.56% of their respective trajectory conformers in each simulation), with subsequent clusters in yellow (23.22% and 27.91%), green (21.3% and 21.02%), pink (19.29% and 16.34%) and red (5.94% and 6.17%). PC projection plots with cluster ellipsoid hulls i.e. the ellipsoid of minimum volume such that points from a given cluster lie inside ellipsoid boundaries (C and G). Trajectory timeline colored according conformational cluster (D and H).
RMSD from Representative Crystal Structure |
|||||
System | Cluster No. | GTP | GDP | A59G | Y32C |
1.32 | 1.31 | 1.32 | 1.38 | ||
1.02 | 1.57 | 0.37 | 1.38 | ||
1.41 | 1.33 | 1.41 | 1.45 | ||
1.38 | 1.01 | 1.37 | 1.52 | ||
1.73 | 0.32 | 1.61 | 1.78 | ||
1.62 | 1.91 | 1.64 | 1.73 | ||
0.94 | 1.49 | 1.06 | 1.26 | ||
1.43 | 1.07 | 1.50 | 1.64 | ||
1.30 | 1.84 | 1.31 | 1.39 | ||
0.82 | 1.58 | 1.06 | 1.19 | ||
0 | 1.64 | 0.94 | 1.04 | ||
1.64 | 0 | 1.62 | 1.81 | ||
0.94 | 1.62 | 0 | 0.68 | ||
1.04 | 1.81 | 0.68 | 0 |
GTP, GDP A59G and Y32C representatives correspond to PDB entries 1qra, 4q21, 1lf0, and 2cl7.
The GTP-to-GDP (backward) transition sampled a number of distinct conformational states (
The temporal evolution of cluster membership in each trajectory (
Analysis of the calculated structures indicates that certain side chain reorientations, diagnostic of GTP and GDP crystallographic states
To examine whether the motions of one residue are related to the motions of another (distant) residue, the correlation of the displacements of all residue pairs were determined (
The extent of correlation for all residue pairs (of Cα atomic displacement) during selected portions of the wild-type GTP (upper triangle) and wild-type GDP (lower triangle) Ras aMD simulations. The color scale runs from pink (for values ranging between −1 to −0.75), through white (−0.25 to 0.25) to cyan (0.75 to 1). Negative values are indicative of displacements along opposite directions, namely anticorrelated motions, whereas positive values depict correlated motions occurring along the same direction. Major secondary structure elements are indicated schematically with helices in black and strands in gray.
Perhaps the most interesting feature of the plot is the pattern of correlation between α2 and α3-loop7 (residues 66 to 74 and 93 to 110). This feature is most evident in GTP-bound simulations and is largely absent in GDP-bound simulations. This pattern is particularly noteworthy as GDP-to-GTP aMD simulations exhibit these correlations only in portions of the trajectory that reside in a GTP like conformation (i.e. after the transition from GDP to GTP, see
The dissection of the catalytic domain into two lobes or subdomains based on the correlated motions of the central β-strands is consistent with the localized nature of sequence variation between Ras isoforms. As previously noted
In an effort to further understand the physical basis of the observed motions upon nucleotide exchange, we analyzed available structures with a simplified elastic-network normal mode method
We have characterized the spontaneous transition between nucleotide-dependent conformational states of wild-type Ras with cMD, aMD and NMA. These functionally important transitions, achieved for the first time for wild-type Ras, are practically impossible to observe with cMD. Furthermore, NMA indicates that the dominant collective motion that occurs during these transitions is a low-frequency motion intrinsic to the structure.
Mapping the reaction path sampled by aMD onto a PCA basis set derived from the distribution of crystallographic structures enabled identification of intermediate structures with unique switch orientations and/or distinct loop3, loop7 and α5 C-terminus conformations. Intriguingly, several of the highly populated intermediates have a close correspondence to known G59A and Y32C crystallographic conformers, both of which have been suggested to be intermediate structures
The pattern of correlated motions revealed by these simulations predicts novel nucleotide-dependent motions of potential significance in the signaling function of Ras. These include correlations of α2 with α3-loop7, loop2 with loop10 and loop3 with α5. Such dynamic linkages between the switching apparatus and the membrane interacting C-terminal region leads us to speculate that residues at each of these sites may be important for nucleotide-dependent modulation of membrane attachment. This is supported by recent experimental evidence for the role of loop3 residues D47 and E49 and α5 residues R161 and R164 in modulating the nucleotide-dependent membrane association of Ras
Finally, low frequency normal modes qualitatively capture the differences between available crystal structure conformations and have high overlap with the eigenvectors obtained from aMD simulations. This result combined with aMD observations suggests that nucleotide-dependent dynamics is facilitated by low frequency, global motions that are intrinsic to the structure and that the nature of the bound nucleotide serves to attenuate these intrinsic low-frequency motions. Furthermore, the significant similarities of aMD, NMA and crystal structure PCA motions highlight the robustness of the observed motions.
We believe that the current advanced simulation and analysis approach is equally applicable to a large number of structurally similar but functionally diverse P-loop NTPases such as kinesin and myosin. Such studies should uncover detailed dynamic behavior and help inform us about general principles and mechanisms underlying nucleotide-dependent conformational changes.
All simulations were performed with the AMBER8 package
Atomic models were prepared from three high-resolution crystal structures (PDB codes: 4Q21, 1QRA and 2Q21). These structures are representative of three distinct conformations highlighted by PCA
All MD simulations were performed using periodic boundary conditions, TIP3P water and charge-neutralizing counter ions, with full particle-mesh Ewald electrostatics. Operational parameters included a 2fs time step and a 10Å cutoff for the truncation of VDW non-bonded interactions. Constant volume heating (to 300 K) was performed over 10ps, followed by constant temperature (300 K), constant pressure (1atm) equilibration for an additional 200ps. Finally, constant pressure constant temperature production dynamics was performed with both classical and accelerated MD implementations. The SHAKE algorithm was used to constrain all covalent bonds involving hydrogen atoms.
Accelerated MD (aMD) extends the accessible time scale of conventional MD simulations by altering the underlying potential energy surface of the system under study. Acceleration stems from the addition of a non-negative boost potential that raises the energy within basins
We employed the coarse-grained AD-ENM normal mode analysis approach developed by Zheng et al.
PCA was employed to aid the interpretation of interconformer relationships. We utilized the previously reported PC basis set obtained from analysis of available Ras crystal structures
Distances between instantaneous trajectory conformations and the centroids of the main GTP and GDP clusters, reported as inserts in
Structures from aMD simulations underwent average-linkage hierarchical clustering according to their pairwise RMSD distance matrix. Inspection of the resulting dendogram was used to partition structures into five dominant groups (ranked according to their populations). The closest structure to the average structure from each cluster, in terms of RMSD, was chosen as a representative for projection onto the PCA basis set described above.
To identify protein segments with correlated atomic motions the cross-correlation coefficient,
Time evolution of Cα atom RMSD from the initial structure of each simulation. Each row corresponds to a single system namely: (A and B) wtGTP, (C and D) wtGDP, (E and F) mutantGDP. Regular MD simulations are depicted in the left panel (A, C, and E,) whilst aMD simulations are depicted on the right (B, D, and F). Simulations with a bound GDP are plotted in green whilst GTP-bound systems are plotted in red. The light green and red lines correspond to the core residues used for superposition.
(2.11 MB TIF)
Principal component based mapping of Ras crystallographic structures. Structures are colored by nucleotide state, triphosphate in red and diphosphate in green and labeled with their PDB code where space permits. Dashed ovals represent the grouping obtained from hierarchical clustering of the projected structures in the PC1 to PC3 planes. Insert: eigenvalue spectrum detailing results obtained from diagonalization of the atomic displacement correlation matrix of Cα atom coordinates. The magnitude of each eigenvalue is expressed as the percentage of the total variance (mean-square fluctuation) captured by the corresponding eigenvector. Labels beside each point indicate the cumulative sum of the total variance accounted for in all preceding eigenvectors (1).
(0.19 MB TIF)
High boost value simulation of mutant GDP system with a bound GDP, see
(2.27 MB TIF)
Heatmap illustrating RMSD clustering of wild-type GDP with bound GTP aMD simulation. See
(1.83 MB TIF)
Contact map of initial wtGTP and wtGDP Ras conformations. Residues are considered in contact when any non-hydrogen atom from a given pair of residues is separated by less than 4Å.
(1.83 MB TIF)
Trajectory averaged contact maps for wtGTP-GTP and wtGDP-GDP simulations. The color scale indicates the fraction of frames in which a given residue-residue contact is present.
(0.05 MB PNG)
Selected time-averaged properties for cMD and aMD simulations†. † Values listed include average Cα atom RMSF along with Cα atom RMSD values for all and core residue subsets during each simulation. System codes are based on the starting structures of the simulations: wtGDP = GDP-bound x-ray structure from the pdb (2), code 4q21; wtGTP = GTP-bound xray structure 1qra; mutGDP = GDP bound G12V structure in 1q21. Note that the time evolution of backbone hydrogen bonds and secondary structure content also remained constant throughout all simulations (not shown).
(0.04 MB DOC)
Comparison of crystal structure and trajectory derived eigenvectors. Inner products between the first five eigenvectors obtained from crystal structure PCA and the first ten eigenvectors obtained from PCA of individual aMD and cMD trajectories.
(0.04 MB PDF)
Supplementary Information Text
(0.10 MB PDF)
Conformational sampling during accelerated molecular dynamics simulation of Ras GDP with a bound GTP.
(2.56 MB MOV)
aMD trajectory snapshots from the 5 to 30 ns portion of GDP with bound GTP trajectory. For reference, the orientation of helix alpha2 in representative GTP (red) and GDP (green) crystal structures are displayed as solid cylinders. See
(27.46 MB MOV)
We thank Drs. Ana Rodrigues and Donald Hamelberg for valuable discussions and acknowledge the NSF Supercomputer Centers and the Center for Theoretical Biological Physics for computational resources.