K-Ras at Anionic Membranes: Orientation, Orientation…Orientation. Recent Simulations and Experiments

Main Text

In this issue of Biophysical Journal, the article by Prakash et al. (1) provides a detailed view of the potential influence of the membrane on the function of the K-Ras4B GTPase, a powerful example of the application of large-scale molecular dynamics simulation to the study of dynamics of a protein-membrane complex. The K-Ras orientational dynamics presented offers a new perspective on Ras-regulatory protein interactions as well as Ras-Ras protein associations for GTPase cluster formation at membranes. However, these simulations and the recent experimental work are just the beginning, not the end. Several issues need to be studied further, such as the diffusion and clustering of both wild-type and oncogenic K-Ras, the nucleotide dependence of K-Ras behavior as well as the GTPase's coordination with other partner proteins at different membrane environments.

Similar to most other Ras superfamily GTPases, K-Ras4B, becomes membrane anchored when farnesylated at its C-terminus but the GTPase also has a strong electrostatic attraction with negatively charged membranes due to several lysine residues in the C-terminal hypervariable region (HVR) of the protein. However, whether these interactions would give rise to orientational preferences for K-Ras at the membrane was unknown. Recent experiments demonstrated that the bound GTP or GDP nucleotides as well as mutations of K-Ras could change the preferred orientation of the GTPase relative to the membrane (2, 3). These experiments also suggested that the orientation of the catalytic domain could be correlated with the function of K-Ras. A thorough understanding of this correlation calls for a detailed investigation of membrane-anchored K-Ras at the atomic level, especially considering that several regions of the protein—the switches as well as the HVR segments—tend to be rather dynamic. Continuous increases in computing power have greatly improved the application of molecular dynamics simulations to biological macromolecules, recently enabling relatively accurate calculations on protein-membrane systems.

The article by Prakash et al. (1) investigates the interaction of the active oncogenic mutant K-Ras4B G12D with an anionic membrane (80% POPC and 20% POPS lipids) by carrying out extensive classical all-atom molecular dynamics simulations totaling 7.5 μs. The most important finding of the work is that the orientation of a single K-Ras protein is not random but, indeed, presents the GTPase in mainly two dominant orientations relative to the membrane. These orientation states, denoted OS1 and OS2, use two different regions of the K-Ras catalytic domain (α3, α4, and β1–β3, α2, respectively) to contact the membrane (Fig. 1). Furthermore, the simulations indicate how the two dominant orientation states could interconvert via an intermediate orientation state. Similar to the contacts with the membrane, interactions between the HVR and the core domain are also rather transient in both orientation states. This apparently contrasts with simulation and experimental work on H-Ras (3, 4) and with a recent report based on computational modeling and solution NMR experiments, which suggest that HVR-core domain contacts could be an integral aspect of K-Ras function (5).

The finding that a single K-Ras could sample multiple different orientation states expands the view of K-Ras dynamics at the membrane. Biologically, the dynamical behavior of K-Ras-membrane complexes likely influences the function of K-Ras in several respects: For example, analysis of the two major configurations shows that only one of the membrane orientations (OS1) is available for further binding of downstream partners such as effector protein Raf (see Fig. 1). Whereas in the other conformation (OS2), partner proteins are excluded from the effector protein binding region—switch S1 and switch S2—as this region is too close to the membrane. Thus, the simulation results imply that the membrane may hinder the binding of partner proteins to K-Ras and hence repress the activity of the oncogenic GTPase. This point will likely inspire the future design of Ras inhibitors for the treatment of cancers. In addition to effector protein binding, the dynamics of K-Ras at the membrane may also affect the dimerization or lateral clustering of the GTPase on the membrane surface. For example, recent molecular modeling and early NMR experiments in solution suggested that the K-Ras could form a homodimer through a β-sheet populated interface (β-homodimer), or alternatively an interface involving α3 and α4 (α-homodimer) (6). Presently, the dimer models are largely computational and the interpretation of the experimental data/conditions are preliminary (e.g., only K-Ras loaded with GTPγS forms dimers). It is not clear how they would translate to a membrane environment, although H-Ras dimerization (and clustering) has been reported (7). Nevertheless, by visual inspection of the two suggested dimers (6), it appears that the OS2 orientation state would, indeed, be compatible with the formation of an α-homodimer (see Fig. 1). By contrast, while the OS1 orientation state is available for effector protein binding, it may not form either α- or β-homodimers unless a considerable rotation occurs (>30° for β-dimer). To complement the simulation results, Prakash et al. (1), carried out electron microscopy and fluorescence lifetime imaging microscopy-fluorescence resonance energy transfer experiments to show that the K-Ras orientation could have an influence on its clustering at a membrane. Specifically, it is thought that Arginine residue 73 of K-Ras is involved in stabilizing the OS2 orientation at the membrane. Reversing the charge of this residue by mutation to Glutamic acid is expected to lessen the binding affinity of OS2 to the membrane and may shift the population to the OS1 state. As OS2 is more favorable for the formation of dimers and clusters than OS1, the mutation will lead to a decreased probability to form K-Ras clusters. This, in fact, is observed experimentally in the report with the contribution of John Hancock’s laboratory (1).

Comparison between simulations and experiments is an important reality-check (that increasingly works both ways). The closest comparison of this study is with the recent experimental work of Mazhab-Jafari et al. (3) using NMR and spin-labeled lipid bilayer nanodisks. Two orientation states were derived for K-Ras4B GDP and GTP (using the nonhydrolyzable analog GMPPNP). The GDP form favors an α-interface with K-Ras helices α4 and α5 at the membrane and has the switch regions exposed, whereas the GTP form has a β-interface with the membrane similar to OS2. The differences between OS1 and the GDP-favored orientation (3) could arise from several differences between experiments and simulation. From the experimental side, only a limited number of distance restraints were available from the spin label to Ile methyl groups. Undoubtably, additional restraints to other methyl groups will be used to refine these orientation states, if not suggest new ones, in the future. Surprisingly, it is the GDP favored-, OS1 equivalent state of Mazhab-Jafari et al. (3) that has the switch regions accessible, whereas the active/GTP form seems to favor the switch-region occluded state (as if to prolong the lifetime of the GTP state/shielding it from GTPase-activating proteins). The experimental finding that oncogenic mutants shift the equilibrium toward the open, OS1-like state suggests a mechanism of K-Ras hyperactivation in cancer and is consistent with the simulations, reported here, that are carried out with such a mutant (G12D) and appear to sample OS1 and OS2 similarly. Thus, overall the agreement between simulations and experiments is very good, also considering the observation in both studies (1, 3) that K-Ras core-membrane interactions are rather transient by nature.

There are several fundamental questions that need to be considered in future simulation as well as experimental studies. The topic of GTPase conformational or dynamic allostery has not yet been extensively addressed in the context of membrane or HVR-core interactions, possibly because these effects are small. However, they could be more significant upon complex formation with effector proteins (e.g., Bouguet-Bonnet and Buck (8) and Fetics et al. (9)). It was supposed that the formation of Ras dimers could promote activation of the effector protein Raf (4), while the simulations by Prakash et al. (1) imply that a dimer formed by the OS2 orientation state may not be accessible for effector binding. The next series of simulations should include effector protein domains such as Raf-RBD at the membrane. Mazhab-Jafari et al. (3) suggest that different effector domains may interact differently with membranes, some not interacting with the bilayer while others interact directly. In the latter case, a common orientation of the Ras and effector protein will likely be stabilized in a way that is also compatible with their formation of the protein-protein complex. Whether such an orientation state is similar to those characterized so far for isolated Ras is unclear. At least the population of the orientation states could be shifted very significantly. Finally, neither experiments nor simulations yet provide quantitative information about the kinetics of interconversion and the energy barriers between the orientation states of K-Ras. The functional experiments, such as the decreased clustering of a R73E mutant of K-Ras reported here by Prakash et al. (1), suggest that these barriers could be significant, but further work is needed.

Author Contributions

Z.L. carried out analysis, and Z.L., S.C., and M.B. wrote the article.

Editor: Michael Feig.

Note Added in Proof

Very recently, another study on K-Ras anionic membrane interactions was published: Jang, H., A. Banerjee, …, R. Nussinov. 2015. The higher level of complexity of K-Ras4B activation at the membrane. FASEB J. 2015 Published online December 30, 2015. fj.15-279091. The simulations were carried out with wild-type K-Ras4B GDP and GTP bound proteins and were started with lipidated but solvated states that were then placed near an anionic membrane.

Intriguingly, only one of four of the GDP and three of four of the GTP states inserted the farnesyl group into the membrane within the 200 ns all-atom simulations. This, as well as the fact that wild type rather than the G12D mutant protein was used, may explain why there is a limited correspondence of membrane bound states with those of Prakash et al. (1) and Mazhab-Jafari et al. (3). Thus, the overall theme emerging from these and other studies is that the behavior of Ras near the membrane is isoform, nucleotide, and mutant specific.