M-Ras distinct activation scenarios: A mechanistic outlook and targeting

Abstract

The conformational states of canonical Ras (H-, K-, and N-Ras) GTPases define their nucleotide-exchange and effector binding capabilities. M-Ras, whose mutational variants can cause cancer directly, and indirectly through their complexes, appear to be one exception. Unlike canonical Ras, the GTP-bound M-Ras is mostly in the inactive state. The active state is stabilized in the holophosphatase ternary complex, which includes SHOC2 scaffolding protein and protein phosphatase-1 (PP1). PP1 dephosphorylates Raf’s inhibitory site, promoting Raf activation and MAPK signaling and contributing to Noonan syndrome phenotype. Activating Q71R variant (M-Ras^Q71R) exhibits higher affinity than the wild type to holophosphatase. With allosteric drug discovery benefitting from insight into allosteric mechanisms, which are unsurprisingly distinct between M-Ras and canonical Ras, we explored M-Ras and M-Ras^Q71R conformational dynamics in GTP- and GDP-bound states by all-atom molecular dynamics simulations. We show that M-Ras and M-Ras^Q71R exhibit differential GTP/GDP loading. GTP-bound M-Ras and M-Ras^Q71R display distinct conformational dynamics in their switch regions although both preferentially assume the inactive conformations. The conserved nucleotide-coordinating asparagine residue in M-Ras G4-loop is the weakest link in the nucleotide-binding coordination, offering mechanistic insights into the GDP release mediated by the guanine nucleotide exchange factor of canonical Ras. Pharmacologically, the occlusion of the Switch II pocket due to the highly mobile Switch II region indicates that it may be infeasible to target this pocket. Targeting M-Ras^Q71R binding interface with optimized K-Ras inhibitor and cyclophilin A appears an alternative approach. Collectively, Ras allosteric mechanistic scenarios shape their personalities, function, and likely drug discovery.

Graphical Abstract

Conceptual model linking M-Ras conformational propensities to RASopathy phenotypes. Our MD simulations show that M-Ras^Q71R-GTP increases the population of active conformations. This conformational shift enhances Ras/MAPK and PI3K/AKT/mTOR signaling and thus contributes to RASopathy-associated phenotypes.

1. Introduction

M-Ras is distinct from its canonical relatives. Its GTP-bound state populates the inactive state. On its own, the active conformation has not yet been captured experimentally, likely due to its relative instability. Although M-Ras can activate MAPK directly, like other Ras family isoforms, activation is weak, due to this active state sparsity. The switch toward active GTP-bound M-Ras conformation takes place through its allosteric stabilization by the holophosphatase. Mechanistically, the minor population of GTP-bound M-Ras active state binds scaffolding protein SHOC2, albeit with low affinity. PP1 phosphatase, whose active conformation binds SHOC2 adjacent to M-Ras, on the concave SHOC2 surface, stabilizes the active M-Ras/SHOC2 complex indirectly, through SHOC2 [1]. Thus, PP1 binding can allosterically, cooperatively, strengthen active M-Ras/SHOC2 interactions in the ternary complex, shifting the M-Ras ensemble toward the now stabilized active state. The population of the stabilized active, GTP- and SHOC2 bound, M-Ras, increases, driving the activation of the GTP-bound M-Ras, and Raf activation, through the cooperative actions of M-Ras and PP1. As in other scaffolding proteins, the SHOC2 concave surface is critical in the formation of the M-Ras/SHOC2/PP1 assembly, driving strong allosteric cooperativity, which we believe is its raison d′être. As to PP1, it is already active, and the increased stability of active GTP-bound M-Ras does not make it more catalytically efficient. Instead, active M-Ras, like other active Ras isoforms, can act by guiding PP1’s localization on Raf and specificity [2], [3]. Alternatively, recently it was suggested that rather than the M-Ras being in a GTP-loaded state, effector binding could be due to sampling of an activated conformation in the GDP-bound state, including the nucleation of the SHOC2/PP1 holophosphatase complex [4], a scenario resembling our proposed sampling of active-like state by oncogenic K-Ras^G12V-GDP [5]. Here, we explore Ras allosteric mechanisms, focusing on M-Ras activation and stabilization, and possible implications to drug discovery.

R-Ras subfamily is closely related to small Ras GTPases (H-Ras, K-Ras, and N-Ras), sharing ∼55 % sequence similarity with the canonical Ras proteins [6]. R-Ras subfamily includes three members: R-Ras, R-Ras2 (TC21), and R-Ras3 (M-Ras) [7]. M-Ras was suggested to evolve independently of other R-Ras isoforms and constitute a subfamily among the Ras family GTPases [8]. Unlike canonical Ras proteins that promote oncogenesis with activating mutations at three classic hotspots (Gly12, Gly13, and Q61) and many other non-hotspots [9], the activating mutations in three R-Ras species are rarely found in human cancers, but occasionally identified in Noonan syndrome [6], [10], [11], a RASopathy that is associated with the dysregulation of Ras/MAPK pathway [12], [13]. RASopathy, like other neurodevelopmental disorders, was suggested to involve weaker mutations thus signaling, exiting the cell cycle prior to the S, DNA replication stage [14], [15], [16]. On the other hand, the stronger M-Ras interaction with PI3K [17], and consequently the likely strong PI3K/AKT/mTOR signaling, could relate to M-Ras function in the muscle and cytoskeleton, requiring robust cell growth.

Canonical Ras proteins are molecular switches, cycling between inactive GDP-bound and active GTP-bound states [18], [19]. The intrinsically low GTP-hydrolysis is stimulated by GTPase-activating proteins (GAPs), and the nucleotide exchange is promoted by the guanine nucleotide exchange factors (GEFs) [20], [21], [22], [23]. Early studies showed that R-Ras and R-Ras2 were regulated by the same GEFs and GAPs, whereas M-Ras used the same regulatory proteins as the canonical Ras [24]. The binding affinity of R-Ras to downstream effectors differs. M-Ras binds to Raf with a lower affinity [25], but R-Ras2 and M-Ras bind to PI3Kα with a higher affinity than canonical Ras [17]. M-Ras binds to SHOC2 and the catalytic subunit of protein phosphatase-1 (PP1c) to form the SHOC2/M-Ras/PP1c (SMP) complex that dephosphorylates a conserved inhibitory phosphorylation site on Raf and thus facilitates Raf activation [1], [3], [26], [27], [28]. The SMP complex is also responsible for the adaptive resistance to K-Ras^G12C inhibitors induced by mis-localization of an apical-basal polarity protein, Scribble [29]. A recent study suggested that M-Ras is deficient in GTP loading, likely due to its relative instability owing to the impaired interactions described below, and unable to easily switch between GDP-bound and GTP-bound states in solution and membrane [4]. The crystal structure of GppNHp (a nonhydrolyzable GTP analog)-bound mouse M-Ras (96 % sequence identity with human M-Ras) indicates that it exhibits the inactive state I conformation, characterized by impaired intramolecular interaction between Thr45 (Thr35 of K-Ras) and the γ-phosphate of GppNHp [30]. However, the GppNHp-bound M-Ras Q71R and Q71L mutants display the active state II conformation in the SMP complexes, suggesting that these mutants drive M-Ras-GTP into the active conformation similar to the canonical Ras oncogenic Q61 mutants [1], [27]. NMR data suggest that the use of GTP analogs can bias the conformational populations of H- and K-Ras [31], [32], altering Ras conformational propensities [14], [33], [34]. Unlike canonical Ras proteins, the conformational states of nucleotide-bound M-Ras are not as explored, and the structural basis for nucleotide binding and exchange remains unclear. Targeting M-Ras inhibits the formation of the SMP complex, offering therapeutic benefit for Noonan syndrome, which could be aided by detailed structural information of M-Ras and its mutants.

In this study, we employ all-atom molecular dynamics (MD) simulations to explore the populations of GTP/GDP-bound M-Ras and M-Ras Q71R mutant (M-Ras^Q71R). Distinct from K-Ras-GTP whose active conformation predominantly populates the ensemble, we observe that M-Ras-GTP chiefly samples various inactive conformations, which are distinguished by their dynamics in both switch regions. The population of GTP-loaded M-Ras is lower than that of GDP-loaded state. In contrast, M-Ras^Q71R is primarily GTP-loaded. Like wild-type M-Ras-GTP, M-Ras^Q71R-GTP preferentially populates the inactive conformations. However, formation of the phosphatase complex with SHOC2 and PP1c promotes M-Ras Q71R switching to the active, effector binding-competent state, contributing to Raf activation and RASopathy. We provide structural evidence showing that Asp129 is the weakest nucleotide-binding site due to the vulnerable 3₁₀-helix in the G4 motif of M-Ras. Our results suggest that rather than inhibitors targeting the Switch II pocket, cyclophilin A and K-Ras inhibitors, which are independent of the pocket, could be more productive.

2. Material and methods

2.1. Structural modeling of M-Ras-GTP and M-Ras-GDP

The full-length M-Ras comprises 208 amino acids (Fig. 1A). Compared with K-Ras4B, M-Ras contains an extended N-terminal tail and a longer C-terminal hypervariable region (HVR). Like K-Ras4B, M-Ras has a polybasic region in its HVR for the plasma membrane association. The cysteine in the C-terminal CVIM motif of K-Ras4B is modified by farnesylation, while the cysteine in the C-terminal CVIL motif of M-Ras undergoes geranylgeranylation [25]. To model M-Ras in cytoplasm, we took the active conformation of GppNHp-bound M-Ras^Q71R from the crystal structure of SMP complex (PDB ID: 7TXH) [1]. We constructed the wild-type GTP-bound M-Ras (6—178) (refer to as M-Ras-GTP in this study) by mutating Arg71 back to Gln71 and replacing GppNHp with GTP. We used the crystal structure of GDP-bound M-Ras (PDB ID: 9C1A) [35] to model the initial conformation of M-Ras-GDP (Fig. 1B). In addition to wild type M-Ras, we also modeled M-Ras^Q71R-GTP and M-Ras^Q71R-GDP.

Fig. 1.

Sequence and structures of M-Ras-GTP and M-Ras-GDP. (A) Sequence and domain structures of M-Ras. In the sequence, the non-polar, basic, acidic, and polar residues are colored black, blue, red, and green, respectively. For comparison, the sequence of K-Ras4B is aligned with that of M-Ras. (B) The model structures of M-Ras-GTP and M-Ras-GDP. The starting conformation of M-Ras-GTP was constructed based on the active conformation of M-Ras^Q71R-GppNHp from the crystal structure of the SHOC2/M-Ras^Q71R/PP1c complex (PBD ID: 7TXH). Thr45 in the Switch I region, and Gly70 and Gln71 in the Switch II region are labeled. For clarity, the hydrogen atoms of GTP/GDP are not shown.

2.2. All-atom MD simulations

We performed all-atom MD simulations to explore the conformational ensemble of the GTP/GDP-bound M-Ras using NAMD 2.14 package [36], [37]. We represented the model structures using the updated and modified version of the CHARMM36m force field parameters [38] and solved the protein in a cubic water box filled with TIP3P water molecules. The minimum distance between protein and the edge of the water box was 15 Å. We neutralized each system by adding 100 mM NaCl. We applied the Nosé-Hoover Langevin piston pressure control and Langevin temperature control to maintain the pressure at 1 atm and the temperature at 310 K, respectively [39], [40]. We calculated the long-range electrostatic interactions using the Particle Mesh Ewald (PME) method, with the PME grid spacing of 1.0 Å and PME interpolation order of 6 [41]. We calculated the van der Waals interactions using a switching function with a twin cutoff of 10 Å and 12 Å and constrained the motion of bonds involving hydrogen atoms using the SHAKE algorithm. The integration time step is 2 fs. Each system was first minimized for 5000 steps and then equilibrated in the NVT ensemble (constant volume and temperature at 310 K) for 5 ns with Cα atoms restrained using harmonic restraints. We carried out 2 µs production simulations in the NPT ensemble (constant pressure at 1 atm and temperature at 310 K) without any restraints. For each system, we ran three independent replicas and excluded the first 500 ns trajectory to avoid the effect of the initial structures on the conformational populations. We calculated M-Ras root-mean-square deviations (RMSDs) and root-mean-square fluctuations (RMSFs) of the Cα atoms. As shown in Figs. S1 and S2, the conformational dynamics of the Switch I (residues 40—48) and N-terminal regions significantly contribute to the RMSDs and RMSFs. Especially, we calculated the RMSFs at two equal time intervals and found that they were well overlapped, suggesting that good convergence of each simulation was achieved. We further assessed the influence of the CHARMM36m parameters on GTP/GDP coordination by analyzing the distances between GTP/GDP and their coordinating residues. CHARMM36m parameters well reproduce the GTP coordination sites in both GTP-bound M-Ras and M-Ras^Q71R. However, in the GDP-bound M-Ras and M-Ras^Q71R, the phosphate–magnesium interactions appear more sensitive to the force field parameters. Both systems exhibited a more flexible GDP-binding site (Fig. S3).

3. Results

3.1. Asp129 is the weakest nucleotide-binding site in M-Ras

A recent experimental study showed that formation of M-Ras-GTP is challenging [4]. Binding of GppNHp with M-Ras required phosphatase to remove all GDPs [4], [30]. To examine the GTP-binding dilemma and GDP-binding preference at a molecular level, we first checked the stability of GTP/GDP binding with M-Ras. Five regions designated G1 to G5 loops of Ras have long been recognized as being critical for guanine nucleotide binding and exchange [42] (Fig. 1A). Residues Gly13 and Lys16 (K-Ras4B sequence numbering) in G1 coordinate to α- and β-phosphates of GTP/GDP; in addition to GTP/GDP, Thr35 in G2, as well as Ser17 in G1 coordinate to Mg²⁺; Gly60 in G3 forms a hydrogen bond with the γ-phosphate of GTP; Asp119 in G4 forms hydrogen bonds with the guanine ring of GTP/GDP; and Ala146 in G5 directly contacts with GTP. In our prior simulation study of K-Ras4B, all these interactions were well maintained in the absence and presence of the GAP-related domain of neurofibromin [43]. However, not all interactions were maintained in the present simulations of M-Ras. Given the sequence identity between K-Ras4B and M-Ras, we focused on the structural changes in the G4 and G5 regions, as M-Ras has an additional residue in both regions compared to K-Ras4B.

We calculated the distance between Asp129 (Asp119 of K-Ras4B) and GTP/GDP and observed that failure to establish hydrogen bonds between Asp129 and the guanine ring of GTP/GDP led to unstable/improper binding of GTP/GDP (Fig. S4). In the crystal structures of GppNHp/GDP-bound K-Ras4B, the corresponding distance varies between 4.0—4.2 Å. Also, in our prior simulations of K-Ras4B-GTP, the calculated distance is 4.1 ± 0.1 Å [43]. Thus, we used the threshold of 4.2 Å to distinguish the conformations with relatively stable binding of GTP/GDP (≤ 4.2 Å) from those conformations with unstable GTP/GDP binding (> 4.2 Å). We combined all trajectories from three replica simulations to obtain the relative populations of GTP/GDP-bound M-Ras (Fig. 2A). We then presented representative conformations for stable (Fig. 2B) and unstable (Fig. 2C) GTP/GDP binding. We found that the populations of stable GTP- and GDP-binding for wild-type M-Ras are 28 % and 39 %, respectively. This indicates that wild-type M-Ras binds poorly to GTP/GDP; however, the population of M-Ras-GDP is slightly higher than that of M-Ras-GTP. In contrast, the Q71R mutation substantially increased the population of stable GTP-bound M-Ras^Q71R to 58 % and significantly reduced the amount of stable GDP-bound M-Ras^Q71R to 2 %, thus our simulation results imply that M-Ras^Q71R tends to remain GTP-bound.

Fig. 2.

M-Ras is deficit in loading GTP/GDP. The presence or absence of hydrogen bonding between Asp129 and guanine of GTP/GDP determines stable or unstable binding of GTP/GDP to M-Ras. (A) The populations of conformations of M-Ras with stable binding of GTP and GDP in the wild type (WT) and Q71R mutant. Trajectories from three replica simulations were combined and categorized into stable or unstable GTP/GDP binding based on the presence or absence of the interaction between Asp129 and the guanine of nucleotide. The standard deviations were calculated in terms of three replicate simulations (represented as black dots). The Q71R mutation significantly increases the population of M-Ras^Q71R-GTP but decreases the population of M-Ras^Q71R-GDP. (B and C) Representative conformations of wild-type M-Ras with stable GTP/GDP binding (B) and unstable GTP/GDP binding (C). Compared to the nucleotide-coordinating residues in other G-loops, Asp129 in the G4-loop is determinant for nucleotide binding.

3.2. Molecular determinant of stable nucleotide binding to M-Ras

We have shown that the interaction of Asp129 with GTP/GDP is prerequisite for GTP/GDP binding to M-Ras. In our initial structure model of M-Ras-GTP, a 3₁₀-helix was formed by residues Met131, His132, and Leu133 (Fig. 3A). Polar solutions, such as water used in our simulations, tend to destabilize 3₁₀-helices [44]. To further elucidate the structural basis of nucleotide binding pitfall, we characterized the changes in the secondary structures of this 3₁₀-helix in the G4-loop of M-Ras (Fig. 3B). The percentages of each secondary structure (3₁₀-helix, bend, turn, and coil) were averaged over the three residues (M131—Leu133), and the calculations were performed separately for those conformations with stable and unstable GTP/GDP binding. As expected, the conformations of M-Ras with stable GTP/GDP binding exhibit a higher percentage of 3₁₀-helix (32—42 %) in the G4 region, indicating that this 3₁₀-helix stabilizes the interaction of Asp129 with GTP/GDP. In addition to the 3₁₀-helix structure, these residues also display a high percentage of turn structures (40—52 %). In contrast, the conformations of M-Ras with unstable GTP/GDP binding exhibit less 3₁₀-helix (8—18 %) and fluctuate among turn (32—74 %), bend (6—34 %), and coil (5—26 %) structures. The highest percentage of turn structure (74 %) shown in the conformations with unstable binding of GTP to M-Ras suggests that 3₁₀-helix is necessary for stable binding of GTP to M-Ras. We speculate that the formation of this 3₁₀-helix adjacent to Asp129 restrains the flexibility of Asp129 to stabilize its interaction with the guanine ring of GTP/GDP. We then calculated the root-mean-square fluctuation (RMSF) of the G4 region for those conformations of M-Ras with stable and unstable GTP binding (Fig. 3C). Indeed, the RMSF value of Asp129 decreased from 2.4 Å in the unstable GTP-binding conformations to 0.7 Å in the stable GTP-binding conformations. In addition, the presence of the 3₁₀-helix maintains low conformational dynamics in the G4 region, retaining stable binding of GTP (Fig. 3D). In contrast, we observed high conformational dynamics in the G4 region when the 3₁₀-helix disappeared, resulting in loss of interaction between Asp129 and GTP. Together, these results indicate that the 3₁₀-helix in the G4 region functions as a molecular determinant for stable binding of GTP/GDP to M-Ras.

Fig. 3.

The 3₁₀-helix structure adjacent to Asp129 in the G4-loop restrains the conformational flexibility of Asp129, enabling stable binding of Asp129 with GTP/GDP. (A) The position and composition of the 3₁₀-helix in the G4-loop of wild-type M-Ras. Met131, His132, and Leu133 form this 3₁₀-helix structure. (B) The populations of secondary structures in those conformations of M-Ras with stable and unstable GTP/GDP binding. M-Ras-GTP and M-Ras-GDP exhibit relatively high percentage of the 3₁₀-helix structure, compared to those conformations with unstable GTP/GDP binding. (C) The root-mean-square fluctuations (RMSFs) calculated for the G4-loop of those conformations with stable and unstable nucleotide binding of wild-type M-Ras. The stable GTP binding restrains the conformational dynamics of the G4-loop, especially the nucleotide-coordinating Asp129. (D) Representative conformational dynamics of the G4-loop. The stable interaction between Asp129 and GTP leads to low flexibility in the G4-loop whereas the loss of such interaction results in a highly mobile G4-loop in wild-type M-Ras.

3.3. M-Ras^Q71R-GTP populates effector-binding-competent conformations

Although deficient in GTP/GDP binding, the conformational states of GTP/GDP-bound M-Ras merit further investigation. Two regions, Switch I and Switch II, are of particular importance as they play crucial roles in interactions with effector proteins like Raf and PI3K and regulatory proteins like GAPs and GEFs [21]. We calculated the distances between Thr45 and GTP/GDP, and between Gly70 and GTP/GDP to characterize the conformational dynamics of the Switch I and Switch II regions of M-Ras, respectively. We then constructed the free energy landscape by calculating the potential of mean force ΔG_PMF for each system in terms of the populations of these two distances. In addition, we also performed principal component analysis (PCA) to characterize the free energy landscape, and comparable results were obtained (Figs. S5 and S6). We found that wild-type M-Ras-GTP predominantly assumes the inactive state I conformation by loss of interactions between Thr45 and Mg²⁺/GTP (Fig. 4A and Fig. S7). The Thr45-GTP distance spans a wide range (6.5 ∼ 23.5 Å), exhibiting a highly mobile conformation of the Switch I loop, compared to the Switch II region. The representative conformation corresponding to the lowest basin on the free energy landscape of M-Ras-GTP (conformation 1) is well superimposed with the crystal structure of GppNHp-bound, mouse M-Ras in the inactive state I conformation (Fig. S8). Since the crystal structure of wild-type M-Ras in the active state II conformation is unavailable, we used the active conformation of GppNHp-bound M-Ras^Q71R (taken from the SMP complex) as a reference and found that our current simulations did not sample the active state II conformation (marked by “+”). The conformational ensemble represented by conformation 5 resembles the active-like conformation, with the conformations of both Switch regions close to the active state. We propose that the active state II conformation of M-Ras-GTP is more unstable, thus less populated. Even though we used it as the starting conformation in our simulations, it quickly shifted to inactive state I conformation, suggesting a higher energy barrier is expected between the inactive and active state conformations. Capturing such short-lived conformations remains challenging to experimental techniques and computational simulations. Different from M-Ras-GTP, the conformational dynamics of the two switch regions of M-Ras-GDP are mainly restricted in two inactive states (Fig. 4B). The more populated conformations (represented by conformation 1) display more flexible dynamics in the Switch I region than the conformations in the other state (represented by conformation 2). A high energy barrier separates these two states, rendering conformational transition between these two states difficult.

Fig. 4.

Free energy landscapes of wild-type M-Ras-GTP and M-Ras-GDP reveal that M-Ras-GTP predominantly assumes the inactive state I conformation with large dynamics in the Switch I region. (A) The potential of mean force ΔG_PMF serves as a free energy landscape for M-Ras-GTP as a function of the distance between the selected residues and GTP/GDP. The distance between Cα atom of Thr45 and β-phosphate of GTP/GDP is used to characterize the conformational dynamics of the Switch I region. The distance between Cα atom of Gly70 and β-phosphate of GTP/GDP is used to characterize the conformational dynamics of the Switch II region. Representative conformations (conformations 1–5) in different energy basins of ΔG_PMF are shown. The “+ ” symbol indicates the location of the active state II conformation according to the structure of M-Ras^Q71R-GppNHp (PDB ID: 7TXH). (B) The same for M-Ras-GDP. Two representative conformations (conformations 1 and 2) in the energy basins of ΔG_PMF are shown.

In contrast to wild-type M-Ras-GTP (Fig. 4A), M-Ras^Q71R-GTP demonstrates a more mobile Switch II region than the Switch I region (Fig. 5A). The distance of Gly70–GTP varies between 6.5 and 13 Å whereas the distance of Thr45–GTP narrows to 7 ∼ 9 Å. Like wild-type M-Ras-GTP, M-Ras^Q71R-GTP primarily exhibits the inactive state I conformation as Thr45 is still far away from Mg²⁺/GTP. The two representative conformations (1 and 2) correspond to the lowest and second lowest energy basin of the free energy landscape, respectively, with conformation 2 closer to the active state of M-Ras^Q71R-GTP. We then superimposed these two conformations with the active state II conformation of GppNHp-bound M-Ras^Q71R in the SMP complex and found that only conformation 2 has no steric clash with SHOC2 and PP1c, suggesting that the reduced flexibility in the Switch I region allows M-Ras^Q71R-GTP to populate active-like conformations that can bind with SHOC2 and PP1c (Fig. S9). Since conformation 2 lies in the second lowest basin, our result indicates that binding to SHOC2 and PP1c shifts the population of M-Ras^Q71R-GTP from the largely inactive conformation (in the lowest energy basin) toward the active-like conformations that are competent for effector binding. For M-Ras^Q71R-GDP, only one energy basin exists (Fig. 5B). Considering the likely high barrier encountered by GDP binding to M-Ras^Q71R (Fig. 2A), the conformational state of M-Ras^Q71R-GDP may be less relevant for biological function. The higher propensity of M-Ras^Q71R-GTP to populate the effector-binding conformation than the wild type is consistent with its increased binding affinity to SHOC2 in the surface plasmon resonance assay [1], contributing to its pathological phenotype in Noonan syndrome.

Fig. 5.

Free energy landscapes of M-Ras^Q71R-GTP and M-Ras^Q71R-GDP reveal that M-Ras^Q71R-GTP preferentially assume the inactive state I conformation but could sample the active-like conformations. (A) The potential of mean force ΔG_PMF serves as a free energy landscape for M-Ras^Q71R-GTP as a function of the distance between the selected residues and GTP/GDP. The distance between Cα atom of Thr45 and β-phosphate of GTP/GDP is used to characterize the conformational dynamics of the Switch I region. The distance between Cα atom of Gly70 and β-phosphate of GTP/GDP is used to characterize the conformational dynamics of the Switch II region. Two representative conformations (conformations 1 and 2) in the energy basins of ΔG_PMF are shown. Compared to M-Ras-GTP, the conformational dynamics of the Switch I region in M-Ras^Q71R-GTP significantly reduced. The “+ ” symbol indicates the location of the active state II conformation according to the structure of M-Ras^Q71R-GppNHp (PDB ID: 7TXH). (B) The same for M-Ras^Q71R-GDP. One representative conformation (conformation 1) in the energy basin of ΔG_PMF is shown.

3.4. Targeting M-Ras-GTP by remodeling the surface of cyclophilin A with tri-complex K-Ras inhibitor

Targeting the oncogenic K-Ras^G12C-GDP led to the development of two FDA-approved drugs sotorasib and adagrasib for treatment of patients with non-small cell lung cancers (NSCLCs) [45], [46]. These drugs bind to the cryptic pocket formed by the Switch II region and the α3-helix of K-Ras [47]. This targetable pocket (termed Switch II pocket) also exists in many other GTPases such as H-Ras^G12C, N-Ras^G12C and N-Ras^Q61R, M-Ras^G22C, and RhoA^G14C [48], [49]. Covalent inhibitors targeting both GTP- and GDP-bound K-Ras^G12C were also developed recently [32], [50], [51]. In addition, non-covalent inhibitors targeting K-Ras^G12D-GDP bind to this conserved Switch II pocket [52]. Compared to canonical Ras, M-Ras, especially Q71R mutant, displays a rather flexible Switch II region (Figs. 4A and 5A). In the crystal structure of GppNHp-bound mouse M-Ras, residues Ala69—Glu73 in the Switch II region were disordered (unresolved) [30]. The lack of a reactive residue adjacent to the Switch II pocket like M-Ras^G22C, along with the flexible Switch II regions in the wild-type and Q71R-mutated M-Ras may preclude the formation of the Switch II pocket. To provide structural evidence, we superimposed the crystal structures of K-Ras^G12C-GppNHp and K-Ras^G12C-GDP in complex with the covalent inhibitor BBO-8520 [50] onto the representative conformations of M-Ras-GTP and M-Ras-GDP (Fig. 6A). The Switch II region moves closer to the α3-helix of M-Ras, occluding the inhibitor-binding pocket, compared to the binding modes of BBO-8520 in K-Ras^G12C (Fig. S10). Similarly, the non-covalent inhibitor seems unable to bind to M-Ras (Fig. 6B). One alternative strategy to target the active state of Ras is to design small molecules that associate Ras with ubiquitously expressed immunophilin proteins such as cyclophilin A (CypA) and FKBP12 [53], [54], [55], [56]. This approach does not require the Switch II pocket, and the binding surface between small molecules and CypA/Ras can be remodeled to increase the binding affinity with the active state of wild-type and mutant Ras [55]. We superimposed the crystal structure of tri-complex formed by K-Ras-GppNHp, inhibitor RMC-7977, and CypA [54] with the representative conformation of M-Ras-GTP (Fig. 6C). Key interactions of RMC-7977 with K-Ras, including Ala59, Gln61, Tyr64, and Met67, are preserved when M-Ras replaces K-Ras. In particular, K-Ras Tyr64 forms π-π stacking interactions with the pyridine and indole groups of RMC-7977, and after superimposition, the equivalent M-Ras Tyr74 can also form π-π stacking with RMC-7977. These interactions were also observed when we replaced K-Ras with M-Ras^Q71R-GTP (Fig. S11). The above results indicate that the binding of M-Ras with the binary complex of CypA/inhibitor could be more feasible than binding to the Switch II pocket. The increasing number of crystal structures of K-Ras and CypA in complex with inhibitors suggests that evaluating their binding affinity with M-Ras is feasible [54]. Optimization may further enhance the affinity of the M-Ras/inhibitor/CypA tri-complex, a potential avenue of future exploration. In addition, inhibitors that directly target the interactions of SHOC2 with N-Ras^Q61R-GTP were also reported recently [57], meriting studies.

Fig. 6.

Targeting M-Ras-GTP via tri-complex formation with an optimized K-Ras inhibitor and cyclophilin A (CypA) seems more feasible than targeting the Switch II pocket. (A) The covalent inhibitor of K-Ras^G12C BBO-8520 is unable to bind to the Switch II pocket formed between the Switch II region and α3-helix of M-Ras-GTP and M-Ras-GDP. The binding positions of BBO-8520 in M-Ras are obtained by structural superimposition of the representative conformations of M-Ras-GTP and M-Ras-GDP (conformation 1 in Fig. 4) with the crystal structures of BBO-8520-bound K-Ras^G12C-GppNHp (PDB ID: 8V39) and K-Ras^G12C-GDP (PDB ID: 8V3A). The compound BBO-8520 is designed to inhibit both GTP- and GDP-bound K-Ras^G12C. The binding of BBO-8520 locks K-Ras^G12C-GppNHp in the state I conformation. The highly dynamic Switch II region of M-Ras-GTP occludes BBO-8520 from occupying the Switch II pocket in M-Ras. (B) The non-covalent inhibitor of K-Ras^G12D-GDP MRTX-1133 is unable to bind to the Switch II pocket in M-Ras-GDP. The binding position of MRTX-1133 in M-Ras is obtained by structural superimposition of the representative conformation of M-Ras-GDP (conformation 1 in Fig. 4B) with the crystal structure of MRTX-1133-bound K-Ras^G12D-GDP (PDB ID: 7RPZ). Similarly, the dynamic Switch II region of M-Ras-GDP occludes MRTX-1133 from occupying the Switch II pocket in M-Ras-GDP. (C) The tri-complex formed by M-Ras-GTP, the inhibitor RMC-7977, and CypA preserves key interactions, including the π-π stacking between Tyr74 and RMC-7977. The binding position of M-Ras-GTP is obtained by structural superimposition of the representative conformation of M-Ras-GTP (conformation 1 in Fig. 4A) with the crystal structure of the tri-complex K-Ras-GppNHp/RMC-7977/CypA (PDB ID: 8TBF). The π-π stacking of Tyr64 of K-Ras with the pyridine and indole groups of RMC-7977 are critical to stabilize the binding interface. The tri-complex formation is independent on the Switch II pocket of M-Ras.

4. Discussion

Classical Ras-GTP assumes inactive state I and active state II conformations [58], [59], [60]. State I, with dissociation of Thr35 and Gly60 from the γ-phosphate of GTP, is deficient in GTPase activity. Oncogenic mutation G12V mutation locks K-Ras-GTP in active state by reducing flexibility in both Switch I and Switch II regions [61], and enables K-Ras4B-GDP to sample the active state II conformation [5]. Tumor-derived RhoA mutants A161P and A161V also populate the active state II conformations and interact with effectors in the GDP-bound states [62]. Here, we sampled the conformational states of M-Ras-GTP and found that M-Ras-GTP predominantly assumes the inactive state conformation (Fig. 4A), consistent with previous experimental study showing that M-Ras-GppNHp populated 93 % state I conformation [63]. Recent Gaussian accelerated MD simulations also suggested that M-Ras-GTP populates the inactive state I conformation due to the dynamics of the Switch I region [64]. We also carried out accelerated MD simulations on M-Ras-GTP and M-Ras^Q71R-GTP by applying a dihedral potential boost, an enhanced sampling technique we used on K-Ras [5]. Consistently, both GTP-bound M-Ras and M-Ras^Q71R primarily sampled the inactive conformation, with M-Ras^Q71R-GTP exhibiting significantly reduced flexibility in the Switch I region relative to the wild type (Fig. S12). This distinct conformational propensity was also supported by chemical shift data (Fig. S13), which shows that the ¹H and ¹ ⁵N nuclear environments of GTP-bound M-Ras and K-Ras4B are only moderately similar, suggesting considerable differences in their structures and dynamics. Two crystal structures demonstrate state I conformations of GppNHp-bound M-Ras (PDB IDs: 1X1S [30] and 9B4R [17]). State II conformation of M-Ras-GTP was not observed in our present simulations, probably due to the high energy barrier between these two states, and its lower stability. State II of GppNHp-bound M-Ras was crystalized by introducing triple mutations (P40D/D41E/L51R). Single mutations P40D and D41E, as well as double mutations P40D/D41E, increased state II population but still favored state I conformation [63]. M-Ras-GTP may assume the state II conformation when bound with effector proteins or with mutations. In the crystal structure of M-Ras^Q35A-GppNHp in complex PI3Kα, M-Ras^Q35A exhibits active state II conformation [17]. In the SHOC2/M-Ras^Q71R/PP1c complex, M-Ras^Q71R also assumes the state II conformation [1]. Like M-Ras-GTP, M-Ras^Q71R-GTP preferentially adopts the state I conformation, but with greatly reduced conformational dynamics in the Switch I region, enabling M-Ras^Q71R-GTP to populate active-like conformations and bind to effector proteins (Fig. 5A). The Q71R mutation substantially increases the population of M-Ras^Q71R-GTP (Fig. 2A), which drives the formation of SHOC2/M-Ras^Q71R/PP1c holophosphatase complex and promotes Raf activation and MAPK signaling, thus contributing to RASopathy [2]. The distinct conformational propensities between the wild-type and Q71R mutant provide the structural basis for the association of M-Ras mutants with Noonan syndrome pathogenesis, underscoring the role of protein conformations in the prediction of cell phenotypes [33], [34]. Fig. 7 plots the distinct free energy landscapes of M-Ras and its mutant M-Ras^Q71R.

Fig. 7.

Distinct free energy landscape of wild-type M-Ras and M-RasQ^71R. (A) K-Ras-GTP prefers active state II conformation but can assume the inactive state I conformation characterized by impair interaction between Thr35 and γ-phosphate of GTP. K-Ras-GDP predominantly adopts inactive conformation. The free energy landscape of K-Ras is shown as a reference. (B) Different from K-Ras-GTP, the wild-type M-Ras-GTP adopts multiple inactive conformations (①–④) and less populates the active conformation (⑤). The multiple inactive conformations are characterized by distinct conformational dynamics of the Switch I and Switch II of M-Ras-GTP. The active conformation of M-Ras-GTP resembles that of K-Ras-GTP but displays more flexible Switch II. Like K-Ras-GDP, M-Ras-GDP assumes inactive conformations (① and ②). (C) M-Ras^Q71R-GTP primarily adopts inactive conformation (①) but populates more the active conformation (②) that is competent for effector binding. M-Ras^Q71R-GDP predominantly samples the inactive conformation (①). The representative conformations of ①–⑤ are shown in ​Fig. 4, Fig. 5​. Compared to K-Ras, M-Ras exhibits more flexible Switch I and II regions, contributing to its distinct free energy landscape.

A recent experimental study showed that M-Ras functions in the GDP-bound state and is unable to exchange nucleotides in vitro or in cells [4]. Although our results suggest that the population of M-Ras-GDP is higher than M-Ras-GTP, the difference (11 %) is not significant (Fig. 2A). Our structural analyses point to the essential role of Asp129 in stabilizing GTP/GDP binding. Early studies of the equivalent Asp119 in H-Ras showed that D119A reduced the affinity of H-Ras for GDP and GTP by a factor of 20 [65]. H-Ras^D119N reduced the affinity for GTP and GDP by a factor of 100 and 2000, respectively [66], [67], [68]. The higher dissociation rate for GDP allows H-Ras^D119N to more readily bind GTP (which is abundant in cells), leading to constitutive activation—the hallmark of an oncogenic Ras mutation [67]. In addition, H-Ras^D119N switched the nucleotide specificity from guanine to xanthine but had little or no effect on binding of effector and regulator proteins [69]. Subsequent studies further showed that H-Ras^D119A substantially destabilized H-Ras by weakening nucleotide (GDP and GTP) binding [70], [71]. Mutational variants on other G-loops of H-Ras such G12V (G1-loop), G13D (G1-loop), and Q61L (G3-loop) are more stable than K117N (G4-loop) and D119A (G4-loop) variants [71]. Mutations in K16 (G1-loop) and G60 (G3-loop) of K-Ras are likely to favor the GDP-bound state [72]. In addition, H-Ras^T35S-GppNHp (Thr35 in G3-loop) favors state I conformation [63]. These mutational studies indicate that compared to other mutations in nucleotide-coordinating residues such as Gly13, Lys16, and Gly60, Asp119 in the G4-loop appears to affect the most vulnerable spot for nucleotide binding to canonical Ras. We observed that loss of interaction between Asp129 of M-Ras and GTP/GDP led to unstable nucleotide binding (Figs. 2B and 2C), which is more severe in M-Ras^Q71R-GDP (Fig. 2A). The highly unstable M-Ras^Q71R-GDP may increase GDP dissociation rate, resulting in spontaneous nucleotide exchange that is independent of GEFs. This can explain why M-Ras^Q71R-GTP is more abundant and populates more active-like conformations capable of effector binding.

The interaction of Asp in the G4-loop with the guanine ring of GTP/GDP is conserved among different families of GTPases, which stabilize the interaction in distinct ways. Compared to K-Ras4B, M-Ras G4-loop is one residue longer (Fig. 1A), and formation of a short 3₁₀-helix adjacent to Asp129 restrains the conformational dynamics of this loop, allowing stable hydrogen bonding with GTP/GDP. Our results reveal that transitions between the 3₁₀-helix and turn structures can maintain the binding of Asp129 with GTP/GDP. However, the shorter 3₁₀-helix and the more abundant turn or coil structures abolish the binding of Asp129 with GTP/GDP (Fig. 3). Nonpolar environments tend to stabilize 3₁₀-helices [44]. SHOC2 features a hydrophobic surface on its concave side where it binds M-Ras-GTP and interacts with PP1c to form the holophosphatase complex [73].

Ras-associated protein (Rap), another member of Ras superfamily of GTPases, has high sequence similarity with Ras. Rap1 can compete with Ras to bind Raf, and thus antagonize Ras activity [74], [75], or reduce the effective local concentration of oncogenic Ras molecules in the nanocluster [76]. Like Rap1, Rap2 can also bind to Raf-1 but is less sensitive to the same Rap1 GAPs [77]. We aligned the sequence of Rap1A and Rap2A with that of M-Ras and found that their G4-loops have the same length, and conserved Asp119 (Fig. 8 A). The G4-loop also forms a 3₁₀-helix structure in both GTP-bound structures (Fig. 8B). Unlike Ras and Rap, the Rho (Ras homolog) family of Ras GTPases contains a unique insert region following the 3₁₀-helix, which forms α-helical structure (Fig. 8C). The conserved Asp is in the 3₁₀-helix in each member of Rho GTPases. In our prior study, we showed that the inset region in the GDP-bound Cdc42 exhibits larger conformational fluctuations than in the GTP-bound state, contributing to the deficiency of Cdc42-GDP in effector binding [78]. Prior studies showed that Cdc42-GTP binds and activates the phospholipase D1 (PLD1), and deletion of the insert region fails to stimulate the PLD1 activity [79]. Alterations (deletion or mutation) in the insert region of Rac1-GTP decrease its binding affinity to NADPH oxidase [80]. In addition, the insert region of RhoA-GTP is required to induce cellular transformation [81]. These studies indicate that the insert region may participate in stabilizing GTP-bound Rho GTPases. The presence of the 3₁₀-helix and α-helical structure of the insert region enhances the hydrogen bonding interactions of Asp with the nucleotide by reducing the flexibility of this region, promoting GTP hydrolysis. This is consistent with the observation that the intrinsic GTP hydrolysis rate of Rho (especially Rac) is much faster than that of H-Ras [82]. Such a fast process can be mediated by the Rho GDP dissociation inhibitor (RhoGDI) that inhibits Rac-catalyzed GTP hydrolysis [83], adding another level of regulation of Rho GTPases by RhoGDI [84].

Fig. 8.

The 3₁₀-helix structure is conserved in Rap and Rho GTPase. K-Ras GEF SOS1 may mediate GDP release via disruption of Asp129–GDP interaction. (A) Sequence alignment between the G4-loop of M-Ras and Rap1A/Rap2A and three members of Rho GTPase. Asp119 of Rap1A/Rap2A is equivalent to Asp129 of M-Ras. In addition to Asp120 in RhoA and Asp118 in Cdc42/Rac1, Rho GTPases have a unique insert region that forms an α-helix. (B) Structural superimposition of Rap1A-GTP with Rap2A-GTP. The conserved 3₁₀-helix and Asp119 are highlighted. (C) Structural superimposition of three members of Rho GTPase. The 3₁₀-helix and Asp120 in RhoA, and Asp118 in Cdc42/Rac1 are highlighted. The 3₁₀-helix, along with the insert region, stabilizes GTP binding, yielding faster intrinsic GTP hydrolysis than H-Ras. (D) The crystal structure of H-Ras with SOS1 (PDB ID: 1BKD). The α-helix of SOS1 that inserts into H-Ras is highlighted in green. The Asp119 of H-Ras is labeled (top). Structural superimposition of unstable GDP-bound M-Ras with the above H-Ras (bottom). The representative conformation of the unstable GDP-bound M-Ras is obtained by clustering analysis of those conformations lacking Asp129–GDP interaction. For clarity, GDP is not shown. Thr35 and Gly60 of H-Ras correspond to Thr45 and Gly70 of M-Ras, respectively. The loss of Asp129–GDP interaction results in the opening of the nucleotide-binding site, and significant conformational changes in the switch regions, similar to the effects of SOS-mediated GDP release.

To provide further structural evidence that Asp129 of M-Ras is vulnerable, we superimposed the unstable M-Ras-GDP with the complex of H-Ras bound with Ras guanine-nucleotide-exchange-factor region of the Son of sevenless 1 (SOS1) protein (Fig. 8D). To facilitate GDP release, SOS1 inserts an α-helix to H-Ras and opens the nucleotide-binding site [85]. The conformational changes in H-Ras largely resemble those in the conformation of unstable M-Ras-GDP in which Asp129 loses interactions with GDP (Fig. 8D). Such a result indicates that the association of SOS1 with H-Ras could primarily disrupt the interactions of Asp119 with GDP. The conformational changes in the switch regions of H-Ras seem to be caused by the release of GDP rather than direct interactions with SOS1. Our results here offer mechanistic insights into the GEF-stimulated release of GDP-bound Ras, implying a link between Asp119 of H-Ras and GEF-mediated nucleotide exchange.

Collectively, our results support experimental studies and provide conformational details for M-Ras activation and functional scenarios. Whereas M-Ras can activate MAPK directly, activation is weak, due to its relatively unstable GTP-bound active state detailed here, thus low population time. It is stabilized allosterically in the holophosphatase by PP1, via SHOC2, where both interact on its concave SHOC2 surface [1]. As we observed, GDP-bound inactive M-Ras is more stable, with the relative populations of stable GTP- versus GDP-binding for wild-type M-Ras being 28 % and 39 %, respectively, increasing to 58 % in the mutant. Binding of the unstable, low populated active conformation to SHOC2-bound PP1, stabilizes it, shifting the M-Ras ensemble toward the active state. The increase in the population of the stabilized active, GTP- and SHOC2 bound M-Ras, permits a more significant Raf and MAP activation through M-Ras binding Raf, cooperatively promoting PP1 dephosphatase catalytic action. Further, M-Ras^Q71R showed 2-fold increases in the population of the GTP-bound state compared to the wild-type (Fig. 2A). Our results showed that M-Ras^Q71R-GTP tends to sample active-like conformational states that favor effector binding (Fig. 5A), in qualitative agreement with the 3-fold increases in affinity to SHOC2 [1]. Consequently, M-Ras^Q71R-GTP elevated basal MAPK activity in RASopathy phenotypes. However, this mediating activation scenario may not be as strong as the direct activation in canonical Ras isoforms. The weaker signaling may not be as consequential in cancer, but more so in neurodevelopmental disorders [86]. Signaling strength can be pivotal in determining the pathological phenotypes—cancer and (or) developmental disorders. Stabilization of active M-Ras through the indirect cooperative scenario or mutations, such as Q71R, which still does not reach that of other Ras isoforms, can clarify the mode and impact of the outcome.

Like canonical Ras proteins, active GTP-bound M-Ras can engage the Ras/Raf/MEK/ERK and PI3K/AKT/mTOR cascades [87]. Previous studies have shown that M-Ras interacts with both Raf-1 and the p110 subunit of PI3K [17], [25], although the relative efficiencies of these interactions vary by cellular context [88]. The binding affinity of M-Ras to Raf-1 was ∼10-fold lower than that of H-Ras [30], implying that M-Ras-driven Raf activation is preferred through the M-Ras/SHOC2/PP1c complex. Compared to canonical Ras proteins (H-Ras, N-Ras, and K-Ras), M-Ras displayed a ∼3—5-fold higher binding affinity to PI3Kα [17]. Our simulations indicate that M-Ras^Q71R-GTP stabilizes conformations competent for effector binding, suggesting a bias toward MAPK and especially PI3K activation. On the other hand, negative feedback of upstream M-Ras activation may degrade PI3K/AKT signaling and contribute to M-Ras-associated RASopathy [16].

5. Conclusions

Our analysis of the conformational ensembles of wild-type M-Ras and M-Ras^Q71R observed distinct populations of GTP-bound M-Ras, with the latter being associated with Noonan syndrome phenotype. We spotlighted the role of the conserved Asp in the G4-loop in maintaining the stable binding of GTP/GDP to M-Ras, and in the GDP release upon interaction with GEFs. We observed that M-Ras-GTP predominantly assumes the inactive state I conformation, and its switch regions are highly mobile, challenging the strategy of targeting the Switch II pocket of M-Ras. We suggested an alternative approach to inhibit M-Ras-GTP by targeting the binding interface of the binary complex of CypA and K-Ras inhibitor. Most importantly, by harnessing a dynamic free energy landscape conceptual outlook, we not only supported the experimental results but also provided a thorough and detailed rationale.

CRediT authorship contribution statement

Ruth Nussinov: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization. Yonglan Liu: Software, Methodology, Data curation. Hyunbum Jang: Writing – review & editing, Methodology, Investigation, Conceptualization. Liang Xu: Writing – review & editing, Writing – original draft, Software, Methodology, Investigation, Formal analysis, Data curation.

Declaration of Generative AI and AI-assisted technologies in the writing process

During the preparation of this work the author(s) did not use any generative AI and AI-assisted technologies.

Declaration of Competing Interest

The authors declare no competing financial interest.

Appendix A. Supplementary material

Supplementary material

Data availability

The research data presented in this study are documented in the paper. The structure files and representative trajectories generated from MD simulations were deposited to Zenodo and can be accessed via the link: https://doi.org/10.5281/zenodo.17475543.