An evolutionarily conserved salt bridge stabilizes the active site for GTP hydrolysis in Rho GTPases

TL;DR

A conserved salt bridge in Rho GTPases helps stabilize the active site for GTP hydrolysis, offering insights into cancer drug development.

Contribution

The discovery of a conserved salt bridge in Rho GTPases that stabilizes the active site for GTP hydrolysis.

Findings

Removal of R70 in RhoA disrupts active site organization and reduces GTP hydrolysis.

E102 modulates active site conformation and intrinsic hydrolysis when removed.

Epistatic relationships between R70, K98, and D13 coordinate allosteric communication in Rho GTPases.

Abstract

Rho GTPases are members of the Ras superfamily of small GTPases that regulate cell morphology, motility, polarization, and cell cycling. Like members of the Ras subfamily, Rho subfamily GTPases dysregulation is implicated in a range of tumors and can serve as a valid drug target. In this work, we investigate the evolutionary trajectory of Rho GTPases within a region of the protein that has been exploited for cancer drug discovery within the Ras subfamily branch - the “switch II pocket.” Our previous work has illustrated the role of allostery in this region of H-Ras in modulation of intrinsic hydrolysis and effector-binding capacity. Here, we report that a highly conserved salt bridge within the Rho subfamily stabilizes the RhoA GTPase active site in a catalytically favorable conformation. We probed the roles of the Rho salt bridge via X-ray crystallography, accelerated molecular dynamics simulations (aMD), and enzymatic studies. We showed that the removal of a residue within switch II of RhoA, the salt bridge residue R70, can impart catastrophic effects on active site organization and GTP hydrolysis. As expected, removal of the analogous R68 in H-Ras, which is not involved in a salt bridge interaction, results in a structure with changes in the active site and a decrease in GTP hydrolysis rate constant that are more moderate than observed for RhoA. The anionic partner of R70, E102, also modulates active site conformation and, upon removal, decreases intrinsic hydrolysis. Based on aMD simulations, we uncovered evidence of epistatic relationships between the Rho salt bridge, the distal residue K98 and P-loop residue D13 which coordinate allosteric communication from the switch regions directly to the active site. Finally, we describe the functional landscape of switch II pocket in the context of both Rho subfamily evolution and potential for drug discovery.

Figures

Full text

The Ras superfamily of small GTPases encompasses a group of structurally related molecular switches. Members of the family are involved in a variety of signaling outcomes, such as survival, cell morphology, vesicle transport, and proliferation (1). While Ras has major roles in cell proliferation, differentiation and survival (2), Rho proteins are critical for cell motility in processes associated with cytokinesis, plasma membrane repair, actin coating, adhesion spreading, tight junction stretching, to name only a few (3). Throughout these processes, Rho proteins play a major role in cell cortex modulation through F-actin and myosin 2. It has become increasingly clear that Rho function involves complex self-organization with activity patterns regulated by positive and negative feedback mechanisms involving guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) that result in budding, pulsed contractions, traveling waves, junctional homeostasis and other patterns associated with function at the cell cortex (4). GTPase cycling in these processes are spatially and temporally synchronized and require well-timed response to signaling cues. Thus, the integrity of the active site in Rho may have different constraints than those surrounding Ras proteins. Here we focus on intrinsic hydrolysis as correlated to active site stability and show that while Rho is likely to hydrolyze GTP through a similar mechanism to Ras, active site stability is achieved differently in the two subfamilies of GTPases.

Due to the ubiquitous influence of GTPases on cellular events, it is not surprising that many of these proteins are implicated in disease. The Ras subfamily members H-, K-, and N-Ras are well known for their role in cancer and thus have been the subject of extensive structural studies in the context of drug discovery. However, Rho GTPases also play a role in cancer cell metastasis and unchecked cell cycling (5, 6, 7, 8). In addition to the most studied Rho GTPase, RhoA, other members of the Rho subfamily include Rac1, Cdc42, and mitochondrial GTPase Miro (9). Examples in the literature are seen of Rho GTPases promoting tumorigenesis through both downregulation (10, 11) and upregulation (12, 13) of Rho activity in tumors. Wild-type Rho GTPase expression is typically upregulated in various cancer types (7) and Rho plays a major role in cancer invasion and metastasis (14). Interestingly, far fewer oncogenic mutations for Rho GTPases are cataloged than Ras GTPases (15). In our review of the COSMIC database for all Rho subfamily members (see Data Availability), we find only two alleles that have over 100 tumor sample instances, RhoA G17V (n = 314, position G15 in H-Ras) and Rac1 P29S (n = 167, position V29 in H-Ras and P31 in RhoA). Rac1 P29S, found frequently in melanomas (16), have unique GTPase cycling properties, including the ability to “self-activate” and bind to effectors without regulatory proteins (17, 18, 19, 20). Less is known about the mechanism of RhoA G17V toward tumorigenesis. However, this mutation has been found in 50% to 70% of Angioimmunoblastic T-cell lymphoma (21, 22) and has been suggested to be an inactivating mutation and implicated in dysregulated PI3K/MAPK signaling (23, 24).

Rho GTPases are also implicated in the development of neurodegenerative disorders, such as Parkinson’s and Alzheimer’s Disease (25, 26, 27). For example, it was found that inhibiting RhoA GTPase in Parkinson 's-related cell lines led to sustained neurite outgrowth, implying that RhoA can be a druggable target to treat brain injury (28). This role of Rho GTPases is closely tied to their function of actin organization and cell polarity modulation. Given the more recent focus on Rho subfamily members and their roles in cancer and other ailments, it is critical that we increase understanding of Rho structure-function relationships, an important context for how these are impaired in disease. Here, we use what we know about the Ras subfamily to contextualize the unique structure-function relationships in Rho proteins.

Despite the diverse array of functions, all small GTPases share a similar core structure and binary switch mechanism (29). When a small GTPase binds to GTP, switch I (residues 32–38 in H-Ras) and switch II (residues 60–75) are sensitive to γ-phosphate of the nucleotide. Switch residues interact with adjacent parts of the structure to stabilize the active site, as well as with effector proteins. Signaling from the GTPase can be turned off via two different mechanisms. The canonical mechanism involves the binding of a GAP, which significantly increases the rate of GTP hydrolysis upon insertion of a positively charged element into the GTPase active site (30). However, all small GTPases are hydrolytic enzymes that have an intrinsic ability to catalyze the hydrolysis of GTP to GDP and mutations frequently impair GAP-mediated hydrolysis (30, 31). We have hypothesized that, in Ras, the intrinsic pathway is facilitated by allosteric connections that may be promoted under certain instances, such as when the binding affinity of effector proteins is higher than the associated GAP (32). After diffusion of γ-phosphate from the active site, GDP remains bound within the active site, rendering the signal “off.” The switch regions typically gain flexibility with the loss of γ-phosphate mediated stabilization. Upon receipt of a signal for the GTPase to turn ‘on,’ a GEF binds to the switch regions of the GTPase and opens the active site, allowing for abundant GTP to replace GDP (30, 33).

The general architecture of Rho subfamily members is organized in a Rossman fold, well-studied in the minimal Ras G-domain (Fig. 1A). The first half of the Ras G-domain (residues 1–86), the effector lobe, is identical in sequence between the canonical Ras isoforms (H-, K-, and N-Ras), while the second half (residues 87–166), the allosteric lobe, shows about 10% difference between the three Ras proteins. RhoA GTPase carries roughly 30% sequence identity with Ras within the G-domain. The effector lobe of RhoA (residues 1–88) contains 33% sequence identity to H-Ras, while the allosteric lobe bears 32% sequence identity. Rho GTPases contain an additional helix within the allosteric lobe (the “Rho insert”), which has an unknown function (Fig. 1A). Rho and Ras share a common effector binding surface along the switch regions with key sequence differences that modulate effector specificity and function (34). The switch regions are dynamically coupled to canonical Walker motifs, found in all NTPases (29, 35, 36). The Walker A motif is also called the P-loop and interacts with the phosphate region of the bound nucleotide, while the Walker B motif binds to the catalytic magnesium ion through a conserved aspartate residue (D57 in H-Ras). The Rho insert is found in place of Loop 8 in H-Ras (37). Given their structural similarity, several intrinsic mechanisms of regulation seen in Ras may also be present in Rho GTPases.Figure 1**RhoA and H-Ras GTPase structure.**A, crystal structures of RhoA bound to the non-hydrolysable GTP analog guanosine5‘-(β,γ-imido)triphosphate (GppNHp) is aligned with H-Ras bound to GppNHp in grey (PDB code 3K8Y) (32). Within the RhoA structure, major features of the G-domain fold are highlighted. The percent sequence identity (% ID) between H-Ras and RhoA GTPase are calculated for each feature. The effector lobe (green) consists of RhoA residues 1 to 88 and the allosteric lobe (purple) includes the rest of the G-domain (residues 89–181). Within the effector lobe, the Walker A P-loop is colored red. Switch I (residues 34–40) and switch II (residues 62–77) are colored yellow and orange, respectively. Switch II also contains helix 2. Helix 3 and the Rho insert are located in the allosteric lobe. B, the water-mediated network between helix 2/switch II and helix 3 within H-Ras is mapped in sticks, rendered in the state 2 conformation (grey) (32). This hydrogen-bonding network between the effector and allosteric lobes within H-Ras results in the alignment of Y32 and Q61 for the promotion of intrinsic GTP hydrolysis. State 1 of H-Ras is shown in pink (93), illustrating the conformational changes seen in both switch regions, with residues Y32 and T35 moving away from the nucleotide-binding site.

From a variety of studies, it is found that H-Ras and other Ras isoforms can sample several sub-states when bound to GTP (Fig. 1B). Switch I of H-Ras can open away from the active site and prevent effector binding. With F28, Y32, and T35 pulled away from the active site, this “open” conformation is known as state 1 (38). Effector binding stabilizes the closure of switch I over the active site, promoting “state 2.” In this conformation, T35 coordinates to the magnesium ion, and Y32 is within interaction distance of the γ-phosphate. In addition, switch II becomes ordered, stabilizing a catalytically competent active site.

Our hypothesized intrinsic hydrolysis mechanism relies on coordination between Y32 in Switch I and Q61 in Switch II. Our lab has shown that the positioning of the catalytic glutamine is modulated by a distal region known as the allosteric site, located at the pocket formed by loop 7 and helix 3 in the allosteric lobe (39) (Fig. 1A). Communication from the allosteric site to Q61 is established through a water-mediated network formed between switch II helix 2 and helix 3, at the center of which is R68 (Fig. 1B). The movement of the top of helix 3, promoted by interactions with calcium and acetate ions under loop 7 in our crystals, allows helix 3 to properly establish this allosteric network along helix 2 (32, 40). Both State 1 and State 2 are sampled by H-Ras at 37 °C (41) (Fig. 1B), although State 1 is sampled much more prominently in K-Ras (42). Thus, it is likely that signaling through effector interaction is subject to higher order regulation than simply nucleotide state.

Compared to the extensive work on Ras isoforms, there is a substantial gap in understanding of allosteric connections in Rho small GTPases and their potential effects on the intrinsic hydrolysis of GTP in Rho. Rho subfamily GTPases contain all the key residue identities that are implicated in conformational state regulation of the switch regions. In the active site of RhoA GTPase, the switch regions present Q63 (equivalent to Ras Q61, Q61^Ras^) and Y34 (Y32^Ras^). The catalytic magnesium ion is coordinated by T37 (T35^Ras^). The placement of these conserved residues suggests a conserved mechanism of intrinsic hydrolysis shared between both Ras and Rho subfamilies. However, the intrinsic hydrolysis rate constant is observed to be significantly higher in the Rho subfamily than in other small GTPases (43, 44, 45). A particularly compelling study followed the course of continuous and unperturbed GTP hydrolysis in H-Ras and RhoA by monitoring, in real-time, NMR resonance peaks sensitive to the nucleotide-bound state (44). This unbiased approached yielded rate constants, at 20 °C, for RhoA and H-Ras of 22 × 10^−3^ min^−1^ and 8.9 × 10^−3^ min^−1^, respectively. The mutants Q63L in RhoA and Q61L in H-Ras severely impair GTP hydrolysis (46, 47), highlighting the importance of the Q63/Q61 catalytic residue in switch II. Interestingly, it has been seen in molecular dynamics simulations that the mutant Q63L promotes a state 2-type conformation of Cdc42 (48), a similar phenomenon as seen in K-Ras G12D (42), suggesting stabilization of a conformational state favoring interaction with effector proteins in these mutants. RhoA and Cdc42 share 64% sequence identity in switch I (RhoA residues 28–40) and 100% sequence identity in switch II (RhoA residues 61–70). Switch I of Rho subfamily members also exhibit conformational changes that may mimic States 1 and 2 found in Ras. Simulations of switch I conformational states have found that State 2 is more favorable in the GTP-bound RhoA, but State 1 is still energetically possible (49). This observation is consistent with NMR data, suggesting that state two is the dominant conformation in Cdc42 GTPase (50).

In this work, we investigate the role of helix 2 - helix 3 interactions on intrinsic modulation of hydrolysis within the Rho subfamily, using RhoA as a representative Rho subfamily member. We discovered a salt bridge between helices 2 and 3 that is highly conserved within the Rho subfamily. Given the role of this region in switch modulation in Ras subfamily proteins, we hypothesized that the Rho salt bridge would limit helix 2/switch II mobility, potentially favoring a catalytically competent state. To test these ideas, we solved crystal structures of mutants of RhoA that remove the salt bridge and performed accelerated molecular dynamics simulations (aMD) to study the biophysical role of this conserved feature. We found that the salt bridge allosterically controls the conformational population of both switch I and II, with communication also passed to the P-loop of RhoA. These structural observations are consistent with a reduction in intrinsic GTPase activity of RhoA upon removal of the salt bridge residues, particularly the switch II R70 residue. Our crystal structure of the switch II R68 mutant in H-Ras also shows active site perturbations with a correspondingly smaller decrease in GTP hydrolysis rate constant than in RhoA, consistent with the absence of the salt bridge in Ras proteins. We evaluate our findings in the context of the entire Rho subfamily, where sequence evolution has maintained many of the features of allosteric communication found in RhoA.

Results

Salt-bridge interactions between helix 3 and switch II are highly conserved in the Rho subfamily and demonstrate co-evolutionary patterns

In H-Ras GTPase, water-mediated interactions between the switch II helix 2 and helix 3 promote the stabilization of switch II, resulting in the placement of Q61 for catalysis (Fig. 1B). We performed structural alignments of H-Ras against all available Rho GTPase crystal structures deposited in the PDB prior to 2023 and evaluated the propensity for interactions to be formed between these two structural elements within Rho GTPases. We identified a highly conserved salt bridge formed between switch II and helix 3 in nearly all available Rho GTPase crystal structures. Figure 2 shows a diverse set of Rho subfamily members, representing the observed conservation of the salt bridge. Using RhoA numbering, we find that an arginine residue is consistently placed in position 70 of helix 2, interacting with a glutamic acid residue located in position 102, donated by helix 3 (Fig. 2A). In Ras, Q99 is found in the equivalent position to E102 (Fig. 2B). The salt bridge in Rho is found in both GTP analog and GDP-bound structures, regardless of binding to another protein (Fig. S1). The salt bridge is present in a structure without the active site Mg^2+^ ion (Fig. S2A), and in the presence of GEF (Fig. S2B) or GAP (Fig. S2C), albeit in slightly modified arrangements. The active site of Rho-GTP structures in the absence of other binding partners features Y34, T37, and Q63 consistent with a State 2 conformation (Fig. 2C). One clear exception to the conserved salt bridge in the Rho subfamily branch is seen in the Miro GTPases, which feature a hydrophobic patch between helices 2 and 3 (Fig. S3). The orientation of helix 2 relative to helix 3 in Miro GTPases is also divergent from the rest of the Rho structures analyzed. In Miro, helix 2 is tipped towards helix 3 with a significantly shorter helical portion of helix 3 towards its C-terminal end of the helix dipole (PDB IDs: 6D71 (51), 5KSO (52)).Figure 2**The Rho subfamily salt bridge.**A, crystal structures of five Rho subfamily GTPases are aligned on α-carbon atoms (94, 95, 96), and PDB ID 2CLS. Four of the five structures are bound to GTP or a GTP-analog, while the fifth structure is bound to GDP. The panel on the left shows the global alignment of the structures, while the panel on the right zooms onto helices 2 and 3. Residues in RhoA position 70, located on helix 2/switch II (SW II), and 102, located on helix 3, are shown as sticks. A sequence logo constructed of 198 Rho subfamily sequences (see Experimental procedures) is shown for the regions local to positions 70 and 102. B, the crystal structure of RhoA bound to GppNHp (cyan, bold) is aligned to H-Ras bound to GppNHp (green) (32). Residues R70 and E102 in RhoA with equivalent positions in H-Ras shown in sticks. In H-Ras, Y96 hydrogen-bonds to R68 (R70 in RhoA). This residue is a lysine in RhoA. C, the active sites of the aligned Rho GTPases are shown in sticks. All subfamily members feature a tyrosine (Y34 in RhoA) and threonine (T37) within switch I (SW I) to potentially support intrinsic hydrolysis. In the four GTP-bound crystal structures, Y34 (RhoA) is closed over the nucleotide and T37 is interacting with the catalytic magnesium ion. The GDP-bound structure (red) features an open conformation of switch I, where these residues are disengaged from the active site.

Inspired by our findings of the conserved salt bridge from structural analysis, we performed sequence alignments of all reviewed protein sequences identified as Rho GTPases in the UniProt database (n = 198), spanning all available species (53). Of these sequences, 79% contained an arginine residue at position 70 (RhoA numbering) and 84% contained a glutamic acid at position 102 (Fig. 2A, sequence logo). Outlying sequences typically contain hydrophobic residues at these positions. All proteins identified as Miro GTPases lacked this salt bridge and contributed to nearly all outlying sequences. Excluding Miro-type Rho GTPases (n = 30 Miro-type), the sequence conservation of these positions increases to 97.5% (position 70) and 98.7% (position 102). Rho proteins identified in D. discoideum, RacI, RacN, and RacO, all lacked an arginine at position 70 but contained glutamic acid at 102. The evolutionary conservation of this structural feature suggests a functional importance that emerged during the differentiation of Rho GTPases into distinct subfamilies.

Wild-type RhoA GTPase bound to GppNHp displays inter-helix connectivity between switch II and helix 3

To better understand the intramolecular connectivity of the Rho salt bridge as compared to H-Ras bound to the non-hydrolysable GTP analog, guanosine5′-(β,γ-imido)triphosphate (GppNHp), we solved a crystal structure of wild-type RhoA-GppNHp (Fig. 3 and Table 1). This structure was solved to 2.0 Å resolution in a space group with symmetry C 1 2 1 with two molecules of RhoA in the asymmetric unit, which are similar but not identical to each other. Molecule A of the structure is used herein. This structure of RhoA resembles the structure of Q63L RhoA-GppNHp (RMSD = 0.19 Å, PDB ID: 1KMQ), but with less helical character within switch II as compared to the Q63L model. We use our RhoA-GppNHp structure (Molecule A) for a detailed analysis of wild-type RhoA and its comparison with mutant structures containing perturbations within and from the conserved Rho salt bridge.Figure 3**Salt bridge connectivity in RhoA bound to GppNHp****.**Left. The crystal structure of RhoA bound to GppNHp is shown with the Rho salt bridge illustrated. Residues that directly interact with the salt bridge include Y74 and D67. Secondary interactions from this region continue through residues K98 and E64, connecting helix 3 to helix 2 and the active site of RhoA. Right. The electron density of the residues illustrated in the left panel is shown at a sigma level of 1.5.Table 1X-ray crystallography data for wild-type RhoA and mutant structuresWild-type RhoARhoA R70A GTPySRhoA E102AGTPySPDB ID: 9N4CH-Ras R68A GppNHp PDB ID: 10DCGppNHp PDB ID: 9N4APDB ID: 9N4BData Collection Resolution range45.9–2.0 (2.0–2.0)31.9–3.0 (3.1–3.0)44.4–2.3 (2.4–2.3)23.3–2.1 (2.2–2.1) Space groupC 1 2 1P 4_3_ 2_1_ 2P 1P 4_1_ a,b,c (Å)α, β, γ (°)119.1, 60.8, 70.090, 123.7, 9060.0, 60.0, 150.8 90, 90, 9054.2, 60.5, 67.267.8, 78.1, 68.469.9, 69.9, 33.3 90, 90, 90 Unique reflections30,326 (2661)13,211 (1150)31,362 (3053)9861 (1385) Redundancy3.6 (3.3)1.8 (1.7)3.9 (3.8)3.5 (3.4) Completeness (%)98.5 (83.4)94.4 (86.6)82.2 (68.4)100.0 (99.7) R_merge_0.042 (0.276)0.054 (0.237)0.093 (0.246)0.107 (0.413) CC_1/2_0.995 (0.828)0.998 (0.950)0.984 (0.949)0.982 (0.969) Mean (I/σ(I))12.2 (2.5)10.8 (2.4)27.5 (8.9)27.5 (5.6)Refinement R-work0.18 (0.22)0.20 (0.27)0.18 (0.26)0.17 (0.18) R-free0.22 (0.28)0.24 (0.37)0.22 (0.32)0.22 (0.27) Macromolecules (atoms)2776139056441247 Ligands (atoms)805318033 Solvent (atoms)3048596126 RMS(bonds)0.0110.0090.0070.003 RMS(angles)1.211.170.970.56 Ramachandran favored (%)98.093.898.298.7 Ramachandran allowed (%)1.975.651.681.3 Ramachandran outliers (%)0.00.560.140.0 Average B-factor21.2642.7831.6928.14 Macromolecules (B-factor)20.7942.1431.4227.09 Ligands (B-factor)19.1959.3932.6724.26 Solvent (B-factor)26.1444.4733.9139.54Values in parentheses are data resulting from reflections in the highest resolution shell.

In the wild-type RhoA structure, the R70-E102 salt bridge in Rho proteins stabilizes switch II in a helical conformation (helix 2) similar to that found in H-Ras bound to GppNHp with Ca^2+^ and acetate in the allosteric site (PDB ID 3K8Y) (32) (Fig. 1). This leads to nearly identical switch I and switch II conformations, including active site arrangement and position of the catalytic glutamine, despite significant residue differences between the two subfamilies in this region. In H-Ras, Q95, Y96 and Q99 in helix 3 and E62, S65 and R68 in switch II are part of the water-mediated network that stabilizes the switch II helix and the active site. RhoA has a network of ionic interactions, with the E102 (Q99^Ras^) – R70 (R68^Ras^) salt bridge at its center (Fig. 3). E102 is anchored by a hydrogen bond to the side chain hydroxyl group of Y74 (M72^Ras^), while R70 also makes a salt bridge to the side chain of D67 (S65^Ras^). K98 (Q95^Ras^) is nearby, stabilized by W99 (Y96^Ras^) and interacting with E64 (E62^Ras^), extending the network to the active site (Fig. 3). Thus, both residues in the central conserved salt bridge in Rho are anchored by switch II residues in a network of ionic interactions with residues in analogous positions to those involved in the water-mediated network in Ras. In both cases, these alternate networks stabilize the switch II helix from N- to C-termini, including residues 64 to 74 (62–72^Ras^), resulting in an ordered active site with Q63 (Q61^Ras^) over the γ-phosphate of the nucleotide. The position of Q63 is also stabilized by the interaction of switch II residue T60 and the backbone of I10 on the P-loop. The network is underlined by a core of hydrophobic residues between helix 3 and switch II. While similar in outcome, the network of ionic interactions in Rho appears more stable than the water-mediated interactions in Ras. The salt bridge in Rho is always observed in the available crystal structures, whereas in Ras, the water-mediated network is observed in only a small number of structures, particularly those in the presence of ligand binding in the allosteric site (32) or with a binding partner such as NORE1A (54).

Rho salt bridge mutants feature switch destabilization and active site disorganization

To interrogate the structural role of this salt bridge in Rho GTPases, we solved crystal structures of alanine mutants of the residues forming the Rho salt bridge in RhoA GTPase bound to the GTP analog GTPγS. Site-directed mutagenesis of wild-type RhoA was performed to generate the mutant R70A and subject to crystallization with GTPγS. A structure from a crystal with symmetry of the space group P 4_3_ 2_1_ 2_1_ was obtained at a resolution of 3.0 Å, with a single R70A mutated RhoA molecule in the asymmetric unit (Fig. 4A and Table 1). In this structure switch II is in an ordered conformation, with both switch regions stabilized by crystal contacts. This differs from the wild-type structure, which crystallized in a space group that does not include contacts in the switch II region. Despite the ordered nature of switch II, the average B factor of residues 62 to 73 within R70A is higher than that of wild-type RhoA. In contrast, switch I (residues 32–42) has a lower average B factor in the RhoA R70A crystal structure relative to our wild-type RhoA structure, likely due to a more tightly packed environment in the crystal. The electron density of the R70A switch regions is shown in Fig. S4.Figure 4**Salt bridge mutant crystal structures.**A, crystal structure of RhoA mutant R70A is shown in red bound to GTPyS and aligned to wild-type RhoA bound to GppNHp (grey). This structure is resolved to 3.0 Å and contains one molecule in the asymmetric unit. In the inset, the active site is shown in sticks with the salt bridge in view. On the right, two symmetry-related molecules of RhoA R70A are shown packing near switch I and helix 2 regions of the central molecule. B, the four molecules contained in the RhoA mutant E102A asymmetric unit (green, yellow, pink, and blue) are aligned with wild-type RhoA bound to GppNHp (grey). The inset features the active site residues of the five molecules shown as stick. Helix 2 is stabilized by the presence of iodide ions (purple). To the right, this is shown in the context of the G-domain.

The RhoA R70A crystal structure contains several regions of subtle conformational differences relative to our wild-type structure (Fig. 4A). Switch I of R70A is in a more open conformation from residues 32 to 38. While Y34 is shifted away from the active site, its hydroxyl group is directed toward and within 3.5 Å of the γ-phosphate of GTPγS, and T37 coordinates the active site magnesium as expected in a state 2 conformation. Switch II shows more subtle, but nevertheless significant changes. In the absence of R70, the position of E102 remains the same as in the wild type, likely stabilized by the H-bond with Y74. The switch II helix relaxes around A70 and P71 from its position to optimize the R70-E102 salt bridge in the wild type, but otherwise the helix remains in a very similar conformation. In the absence of the arginine, however, the ionic network at the N-terminal end of switch II is undone, with K98 and E64 directed away from one another and Q63 flipped away from the active site. Crystal contacts within the R70A crystal lattice likely help stabilize this ordered switch II conformation and promote some of the differences relative to the wild-type structure in other areas, such as those seen in the Rho insert and β-strand 2.

We also created an alanine mutant of the negatively charged partner in the salt bridge, E102A (Fig. 4B). This mutant crystalized in a P1 space group and diffracted to a resolution of 2.3 Å. Four molecules of RhoA E102A bound to GTPyS are found in the asymmetric unit, each featuring distinct switch conformations. All molecules in the asymmetric unit contain iodide ions from the crystallization conditions, stabilizing the positive charge around R70 in the absence of E102. Two poses of Y34 and T37 are seen in the RhoA E102A mutant crystal structures. Molecules B, C, and D feature Y34 pulled away from the active site but pointed toward T37. Molecule A has a unique conformation of Y34, where the side chain is directed away from the active site. Also located on switch I, T37 typically coordinates with the active site magnesium ion through the hydroxyl group of the side chain. However, in the mutant crystal structures, T37 coordinates to the ion through its backbone carbonyl group. This pose is reminiscent of GDP-bound Rho structures, as shown in Figure 2C. Catalytic Q63 is seen in varying conformations, but all point away from the γ-phosphate of the GTP analog, with molecule B featuring a disordered Q63 residue. In all four molecules in the E102A asymmetric unit, K98 shifts towards the iodide ions, most likely resulting in the observed conformational heterogeneity of E64 within these structures. Overall, RhoA mutant E102A is found in a state 1-like conformation, where switch I is disengaged from the active site and Q63 is destabilized. This state 1-like conformation may be favored by changes in the R70, K98 and E64 interaction near switch I.

H-Ras R68A impairs the active site less severely than its counterpart in RhoA

While R68 in Ras is dynamically involved in a water-mediated network conditionally connecting helix 3 residues to the active site based on ligand binding in the allosteric site (32), the salt bridge in Rho appears to be constitutively present, helping to stabilize switch II and its connections to the active stie. We therefore expect removal of the switch II arginine in the R68A mutant in H-Ras to have a more modest impact on the structure of the active stie and GTP hydrolysis rate constant compared to the R70A mutant in RhoA. We obtained crystals of the mutant H-Ras R68A bound to GppNHp with symmetry of the space group P4_1_ and solved the structure to a resolution of 2.1 Å. The asymmetric unit contains one molecule of Ras R68A with crystal contacts primarily along helices 2 and 4. The Switch II is disordered at its N-terminal end before helix 2, from residues 61 to 65. Switch I, free of crystal contacts, is found in a State 2 conformation with the catalytic magnesium ion in contact with T35 (Fig. 5).Figure 5H-Ras mutant R68A aligned to wild-type H-Ras in theState2conformation. Crystal structure of H-Ras mutant R68A is shown in light green bound to GppNHp and superimposed on wild-type H-Ras bound to GppNHp (dark green, PDB ID: 3K8Y). The mutant structure is resolved to 2.1 Å resolution and contains one molecule in the asymmetric unit. In the inset, the residues typically involved in the coordination between helices 2 and 3 are illustrated in sticks.

Coordination between helix 2 and helix 3 rearranges upon removal of the R68 side chain as it does in the RhoA R70A mutant. Residue Q99 (E102^RhoA^) shifts downwards towards the active site, partially filling the void left by removal of the arginine side chain, and D69 (P71^RhoA^) moves to maintain its H-bonding interaction with Q99. The backbone carbonyl of S65 moves toward D69 and Q99 and this also helps compensate for the absence of the arginine side chain. Residue Y71 (S73^RhoA^), which in wild type H-Ras stacks against the aliphatic portion of R68, flips outward toward the solvent. In the wild-type structure of H-Ras, Y96 (W99^RhoA^) interacts directly with R68 and points towards the backbone carbonyl of A59. With the removal of the R68 side chain, Y96 shifts downwards, towards D92 (N94^RhoA^). We saw in RhoA mutant structures that residue K98 changes conformation with the removal of either part of the salt bridge. In H-Ras R68A, the equivalent residue Q95 is disordered. The catalytic residue Q61 is also disordered and likely flipped out of the active site, in analogy to the RhoA mutant.

Rho salt bridge mutations impact intrinsic GTP hydrolysis

Given the structural consequences of removing the Rho salt bridge, we hypothesized that the intrinsic hydrolysis capabilities of the mutants would also be hindered. We performed single-turnover GTP hydrolysis assays for wild-type RhoA, and RhoA salt bridge mutants R70A and E102A to determine the effect of removing the salt bridge on the rate constants for GTP hydrolysis. All assays were performed in the absence of GTPase activating proteins to observe the effect of the salt bridge on the intrinsic hydrolysis rate constants. We utilized a discontinuous assay based on the production of radiolabeled inorganic phosphate to obtain rate constants for GTP hydrolysis catalyzed by RhoA and its two mutants (see Methods). Kinetic rate constants (Table 2) were determined through curve fitting to a single-step, first-order kinetic model derived by the DynaFit software (Fig. S5). RhoA R70A and E102A both have decreased rate constants compared to wild type RhoA. While E102A shows a modest decrease relative to wild type RhoA, R70A is nearly an order of magnitude slower. The severe impairment of catalysis associated with the R70A mutant points to the importance of R70 for active site integrity.Table 2. Intrinsic hydrolysis rate constants for RhoA and H-Ras GTPases at 37 °CProtein1000 x khyd (min^−1^)RhoA WT28 ± 1RhoA R70A2.5 ± 0.1RhoA E102A20 ± 0.6H-Ras WT25 ± 3H-Ras R68A14 ± 1

To test the importance of R68 in H-Ras, we obtained the GTP hydrolysis rate constants for wild type H-Ras and H-Ras R68A and determined the relative effect of this mutation on H-Ras compared to the R70A mutation on RhoA (Table 2 and Fig. S5). The rate constant for H-Ras R68A is about half of that for WT H-Ras, a 50% decrease. As expected, this is a relatively small decrease compared to the 10-fold or 90% decrease we observed for the RhoA R70A mutant. Interestingly, it is larger than the 30% decrease observed for RhoA E102A relative to WT RhoA, highlighting the importance of the switch II arginine residue in both RhoA and H-Ras. Overall, it is clear that R68 in H-Ras and R70 in RhoA play an important role in active site integrity, with a more pronounced effect in RhoA. This is expected given that the salt bridge in RhoA stabilizes the active site more directly than in H-Ras, where active site integrity is through water-mediated H-bonding interactions.

The real-time NMR experiment, referenced in the introduction, established the prevailing view that the intrinsic GTP hydrolysis rate constant for RhoA is larger than that for H-Ras (44). However, while rate constants are consistently around 22 × 10^−3^ min^−1^ at 20 °C across independently published measurements using NMR or ^32^P-labled GTP for Rho proteins (44, 55, 56), they have been reported to be 8.9 x 10^-3^ min^-1^ for H-Ras at 20 °C by NMR (44) and measured with ^32^P-labled GTP to be either 10 × 10^−3^ min^−1^ at 37 °C (57) or 28 × 10^−3^ min^−1^ at 37 °C (58). Our own laboratory has previously reported a GTP hydrolysis rate constant of 16 × 10^−3^ min^−1^ for WT H-Ras at 37 °C (59), using the same experimental setup and data processing methods that we used here to obtain a rate constant of 25 × 10^−3^ min^−1^ 37 °C. It appears that measurement of GTP hydrolysis rate constants for H-Ras may be more sensitive than for RhoA. Therefore, for the purposes of our analysis in this section, we focus on the decrease in GTP hydrolysis rate constants of the RhoA mutants relative to wild type RhoA and of the H-Ras R68A mutant relative to wild type H-Ras. All experiments within the RhoA set in Table 2 were performed by one person in triplicates, and those of the H-Ras set in Table 2 were performed by another person, also in triplicates. Thus, we are confident in the relative decrease in GTP hydrolysis rate constants for the mutants compared to wild type in each set. To date, the most reliable direct comparison of rate constants between RhoA and H-Ras remains the one performed by NMR (44), where experimental variables can be rigorously controlled.

Rho salt bridge removal impacts global dynamics

To evaluate the global effects of the salt bridge mutations, we performed accelerated Molecular Dynamics (aMD) simulations starting with the wild-type, R70A and E102A RhoA crystal structures (Molecule A of each structure), with GTP substituted for the nucleotide analogue in all cases and iodide ions removed in the case of E102A. We then applied Boltzman-reweighing calculations to account for the boost potentials in our analysis of the simulation trajectories (60, 61). This allows exploring conformational space that cannot be accessed in classical MD simulations due to large energy barriers associated with conformational transitions. We have shown via SAXS-WAXS experiments that aMD trajectories generally reflect a more accurate biophysical portrait of Ras subfamily members than classical MD trajectories (62).

To observe global RhoA dynamics, we superimposed the replicate trajectories (five independent 200 ns simulations per model) and calculated the root mean square fluctuation per residue (RMSF) across the simulations (Figs. 6 and S6). The wild-type RhoA-GTP trajectories point to several expected phenomena (49, 63, 64), including dynamic switch regions and moderate fluctuations of the Rho insert. Switch I, while mobile, does not disengage from the active site magnesium ion, but switch II samples a wider range of conformations that includes catalytic Q63 positioning away from the active site. While the N-terminal end of switch II is highly dynamic, the switch II helix is stabilized by the presence of the R70-E102 salt bridge. The presence of the GTP gamma-phosphate stabilizes the P-loop relative to the dynamic switch regions. A beta-hairpin motif comprising β2-loop3-β3 (residues 42–58) also shows greater fluctuations relative to the rest of the G-domain (Fig. 6A). This region has been described as the ‘interswitch region’ in other GTPases and serves as both an effector docking site (65, 66, 67) and communication modality across the protein (68, 69, 70).Figure 6**RMSF of wild-type, E102A and R70A RhoA-GTP.**A, the root mean square fluctuation (RMSF) per α-carbon is plotted for the averaged and weighted aMD trajectories of wild-type (red), E102A (blue) and R70A (black) RhoA bound to GTP. Major structural features of RhoA are indicated at appropriate positions in the sequence. B, the RMSF value per α-carbon is visualized on the crystal structure of wild-type RhoA bound to GppNHp. Larger RMSF values are colored in warm colors and have a larger radius than smaller RMSF values (cooler colors). RhoA R70A contains the largest range of RMSF values, with the highest fluctuations within the switch regions and the Rho insert.

The trajectories of RhoA mutants R70A and E102A feature altered dynamic landscapes of the switch I, switch II, P-loop, helix 3 and Rho insert regions. Significant dampening of switch I movement is seen within the E102A trajectories, along with reduced Rho insert fluctuations relative to the wild-type trajectories. In contrast, the R70A trajectories have increased mobility throughout the structure (Fig. 6, A and B). It is important to note that for the R70A trajectories, the crystal structure conformation is not maintained, confirming the role of crystal contacts in stabilizing the switch regions in this mutant. All substructures of R70A gain flexibility relative to the wild type, most notably switch I. Both R70A and E102A mutant simulations show increased movement of the P-loop and switch II, with the most dramatic difference seen in R70A trajectories. In addition, helix 3 becomes more flexible in both mutant simulations, but in unique regions. In the E102A trajectories, the N-terminal portion of helix 3 gains flexibility relative to both the wild-type and R70A simulations, while helix 3 within the R70A mutant simulations gains broad structural heterogeneity, especially towards its C-terminal pole. As can be clearly seen in Figure 6B, the E102A mutant shows more subtle changes in dynamics relative to wild-type RhoA than does the R70A mutant. Thus, in the current and following sections, we first discuss the landscape for E102A and then contrast that with the more drastic changes observed for R70A, where switch II becomes untethered from the rest of the G-domain.

Rho salt bridge removal alters conformational states

We sought to understand the impact of removing the Rho salt bridge on the organization of the active site. The aMD trajectories of the wild-type RhoA structure indicates that the protein favors a catalytically competent conformation in which switch I is in state 2 closed over the nucleotide and switch II is positioned such that catalytic residue Q63 has easy access to complete the active site for GTP hydrolysis. We examined the frequency of salt-bridge formation between R70 and E102 in context of active site geometry. For this, we calculated the distance distribution between key active site residues, Gly 14 (Cα) (P-loop) and Q63 (Cγ) (switch II), throughout the simulation trajectory as a rough indication of active site integrity, with smaller distances consistent with a favorable position of Q63 to fully form the active site. A plot of the G14-Q63 distance versus the frequency at which each distance is populated throughout the trajectory shows four distinct peaks (Fig. 7A). We calculated the average structure in the set of conformations found under each peak. For wild-type RhoA, peak 2 is the most prominent, with G14-Q63 distance ranging between 7 and 9 Å (average structure shown in right panel of Fig. 7A). For the trajectory frames within each of the four peaks, we measured the distance between the side chains of R70 and E102 to determine whether the salt bridge is present (Fig. S7). We found that the smaller the G14-Q63 distance, the tighter the salt bridge between R70 and E102 in the wild-type Rho simulations (see average peak 2 structure in Fig. 7A). In the conformation generally associated with the active form of RhoA, the R70-E102 interaction is part of a network of hydrogen bonds that includes the anionic switch II residue E64. Aromatic residues Y66, Y74, and W99 also help to stabilize this conformation and pack tightly against other hydrophobic residues in β1-β3. The Y74 side chain is within hydrogen-bonding distance of E102 (2.5 Å) in the wild-type crystal structure and changes modes of interaction with the salt bridge throughout the aMD trajectory. It appears that the arrangement of this network drives the placement of Q63, primarily through the stabilization of switch II.Figure 7**aMD states of RhoA and mutants as classified by active site conformation.**A, (Left) the distance between switch II residue Q63 (carbon) and the P-loop G14 (α-carbon) is calculated per frame of the averaged weighted (solid line) and unweighted (dashed line) aMD trajectory of wild-type RhoA and plotted as a probability index. Four distinct conformations of the active site are seen, with peaks 1 to 2 containing a ‘closed’ active site and peaks 3 to 4 containing an open active site. The initial crystal structure distance is denoted as a star. (Right) The most dominant conformation, averaged from peak 2, is shown in the grey structure. B, (Left) the distance probability distribution of the Q63-G14 pair is plotted for RhoA E102A as described in panel 6A. (Right) The average structure sampled by peak 2 frames is shown in teal. C, (Left) the distance probability distribution of the P-loop to Q63 is plotted for RhoA R70A as described in panel 6A. (Right) The most dominant conformation of the R70A trajectories (peak 3) is seen in red.

Within the wild-type trajectories, peaks 1 and 2 are associated with Q63-G14 distances between 5 and 9 Å. In these structures, the active site Q63 is tilted toward the nucleotide as exemplified by the average structure taken from peak 2 trajectories (Fig. 7A). The peaks 3 and 4 populations show Q63 and the P-loop G14 12 to 17 Å from each other, as shown for peak 4, the most dominant of the two (Figs. 7A and S8A). This indicates an open, deconstructed active site, and we find that most of this structural population features a broken salt bridge conformation (>4.5 Å distance between residue headgroups, Fig. S7). As Q63 is a catalytic residue, we associate these active site conformations of RhoA with near active and inactive states for GTP hydrolysis, respectively. Overall, we find a clear correlation between the inactive conformation of wild-type RhoA and a disengaged salt bridge.

We interrogated the allosteric consequences of removing the salt bridge using our R70A and E102A crystal structures for aMD simulations. Both sets of mutant simulations feature a deformed active site, with E102A simulations showing a more subtle rearrangement than R70A. Overall, the active site conformations resulting from E102A trajectories resemble the wild-type RhoA, except that the variance of distances between Q63 and the P-loop narrows significantly within the E102A trajectories. Most of the simulation features an active site containing Q63 roughly 9 to 11 Å away from the P-loop (Fig. 7B), whereas in the wild-type the range is greater, from about 5 to 17 Å (Fig. 7A). This wide range of wild-type conformations has distinct population clusters with distances below 10 Å (peaks 1 and 2, Fig. 7A) and above 15 Å (peak 4) between Q63 and the P-loop. This contrasts with a single dominant distance population around 10 Å calculated from the E102A trajectories (peak 2, Fig. 7B), suggesting less conformational heterogeneity in this region with RhoA E102A, in which Q63 less frequent accesses the active site.

The average structure of the dominant distance cluster within the E102A simulations has many hallmarks of a state 1-like conformation. Aside from Q63 moving away from the P-loop, switch I tilts outwards. Residues inferred to be essential for intrinsic hydrolysis, including Y34, point away from the active site. Active site residue T37 no longer coordinates the catalytic magnesium in the average structure. As reflected in the global RMSF profile of the E102A trajectories (Fig. 6A), switch I is relatively immobile, more frequently populating this open conformation. In addition, the P-loop becomes deformed relative to the wild-type structure. This phenomenon is due to allosteric effects resulting from rearranged interhelix structures, discussed later.

On average, R70A trajectories feature a significantly widened distance between Q63 and the P-loop, which remains over 10 Å during most of the simulation. The predominant conformation has a G14-Q63 distance of 11 to 12 Å, compared to around 7 Å within the wild-type RhoA trajectories (Fig. 7C, peak 3). Peak 4 (Fig. S8B) has a G14-Q63 distance near 15 Å, like peak 4 in the wild type, with an open switch I and a salt bridge between K98 on helix 3 and D13 on the P-loop (further discussed below). Although wild-type and both mutants access state 1, R70A is unique in highly increased dynamics of both switch I and switch II regions and the Rho insert. This can be visualized clearly in the superposition of the average structures from each of the G14-Q63 distance peaks for the wild-type (Fig. S9A), E102A (Fig. S9B) and R70A (Fig. S9C). In the R70A trajectories switch I detaches from the active site in all major conformational subpopulations associated with the G14-Q63 distance. In the most dominant average conformation (peak 3) (Fig. 7C), Q63 becomes buried within the core of the GTPase, stabilized away from the P-loop. Unlike the E102A trajectories, the P-loop within R70A simulations does not undergo significant structural rearrangement, despite an increased RMSF profile within this region. As will become evident, there is communication between the switch regions and the P-loop which become pronounced with the dramatic conformational sampling of switch II during the simulation of RhoA R70A.

Packing and coordination through the Rho salt bridge modulate P-loop dynamics

The conformational shifts seen within the active site of RhoA GTP upon removal of the Rho salt bridge reveals widespread communication along the switch regions, the P-loop and the Rho insert. We have located a network of residues lining the interface between helix 3, switch II and the P-loop that may serve as an important site of conformational regulation, with the Rho salt bridge ‘anchoring’ these points of communication.

Removing the salt bridge results in the sequestration of switches I and II away from communication points within the rest of the G-domain. This is best exemplified in the trajectories of E102A. While parts of switch I and switch II appear to be more rigid in the E102A mutant, switch II rearranges frequently to accommodate for the lack of salt bridge. Without E102, R70 is free to interact primarily with negatively charged residues within switch II, isolating the C-terminal portion of switch II from the rest of the G-domain (Fig. 7B). The interaction between R70 and neighboring switch II residue D67 becomes more frequent than in the wild-type trajectories. These two residues interact in wild-type trajectories in the context of the salt bridge between R70 and E102, with D67 and R70 in switch II directed toward helix 3 (Fig. 7A). Thus, in the wild type, the R70-D67 interaction is secondary to the R70-E102 central salt bridge, which anchors switch II to the G-domain. In the absence of E102, the D67 and R70 interact away from the remainder of the G-domain, facing towards the solvent. In the E102A mutant, helix 2 rotates away from the rest of the G-domain, weakening the integrity of switch II/helix 3 interactions. In rarer conformations, this intra-helix interaction breaks, and the switch gains significant structural heterogeneity. Overall, in the E102A mutant, only a few hydrophobic residues, including Y74 and W99, appear to hold the N-terminal region of helix 2 to the beta-sheet core.

With the removal of E102, the architecture of helix 3 rearranges to adopt a more conventional helix structure, without the irregular geometry of the third turn within this helix seen in the wild type, which appears to be due to the placement of K98 and W99 relative to D13 and N94 (Fig. 7, A and B). Without E102, W99 can repack toward the C-terminal end of the helix, facilitated by Y66, and N94 is no longer involved in the H-bonding network between helix 3 and the P-loop. The repacking still allows for interactions of K98 and P-loop residue D13. This conformation is held in all major active site categories as seen in Figure 7B. The changed connectivity to D13 is reflected in the increased RMSF of P-loop residues 13 to 16 within the E102A trajectories relative to the wild-type simulations. Residues 11 and 12 gain structural heterogeneity to facilitate D13 interaction with helix 3, along with the repacking necessary to accommodate W99.

More radical global restructuring of the effector lobe is seen upon mutation of R70. In the most frequently sampled population (peak 3), removal of R70 breaks intramolecular interactions between helix 3 and switch II, with hydrogen bonding between Y74 and the backbone of L103 being the only remaining interhelix interaction (Fig. 7C). This interaction is not observed in either the wild-type or E102A trajectories. With the disorder of switch II, Q63 wedges between W99 and β1, becoming inaccessible to the active site. This new packing, along with charge interactions within the region, maintains a consistent disconnection between switch II and helix 3. Like switch II, helix 3 becomes significantly less structured in the absence of the R70 side chain (Fig. 7C). This is especially seen in the C-terminal end of helix 3, where the helix unwinds. Many of the hydrophobic residues within this region pack against one another, but unlike the E102A trajectory, switch II hydrophobic residues such as Y66 do not interact with hydrophobic residues within the former helix 3. P-loop-interacting residue, K98, turns away from the active site to make close contact with E102. With Q63 buried within the G-domain, E64 and D65 at the N-terminal region of switch II approach D13 in the P-loop and the triphosphate moiety of GTP, respectively, while N94 still interacts with D13 in the absence of K98. Analysis of RMSF of the trajectories shows significantly increased dynamics of the P-loop when compared to the wild-type simulations (Fig. 6A). The simulations clearly show the importance of R70 in maintaining the integrity of switch II and its connection to both helix 3 and the P-loop, as in its absence the active site is dismantled.

The second most dominant conformation of the R70A simulations (peak 2) features several unexpected factors contributing to global RMSF trends observed uniquely in the R70A trajectory (Fig. S10). In this state, K98 again interacts with E102 as a surrogate for R70, as it does in the more frequently sampled peak 3. Q63 is positioned close to the nucleotide, with a sodium ion between the second and third phosphates of the GTP molecule and Q63. This ion is likely stabilized by the inverted position of D13, which now points toward the nucleotide. The lack of both N94 and K98 interactions with D13 promotes this rearrangement, deforming the entire P-loop, and causing the guanine base of the nucleotide to shift out of the active site. The Rho insert loses interactions with the nucleotide essential for its stability and disengages from the rest of the G-domain, resulting in an unexpectedly high RMSF for the Rho insert domain (Figs. 6A and S9C).

Through the aMD simulations of wild-type RhoA bound to GTP, we see that the Rho salt bridge not only maintains structure between switch II and helix 3 but also facilitates the interaction of residues from the top of helix 3 through the active site via the P-loop and catalytic residue Q63. In the absence of the salt bridge, simulations of E102A and R70A show impaired active site features that propagate throughout the Rho G-domain, with global changes in dynamics and sampling of conformational states. Overall, it is clear from the simulations that the R70-E102 conserved salt bridge serves a critical role in active site stability and is important for both structural integrity and communication in RhoA and likely in all of the Rho subfamily members in which the salt bridge is present.

Discussion

The work presented here elucidates a mechanism through which Rho GTPases hydrolyze GTP with active site integrity promoted by the formation of a salt bridge that stabilizes the switch II toward helix 3. This conserved structural element within the Rho subfamily, is divergent from Ras GTPases, where we have shown that allosteric communication at the helix 3/switch II interface is mediated by water molecules (32), leading to a more dynamic active site (71). The salt bridge in the Rho subfamily is unique. However, its interactions that radiate to the active site and beyond, are attained through residues in completely analogous positions to the water-mediated connections that link helix 3 and switch II to the active site in Ras, both centered on R70 (Rho), R68 (Ras) at equivalent positions in switch II. For both RhoA and H-Ras and their respective arginine mutants, the rate constants for intrinsic GTP hydrolysis measured in vitro directly reflect features of the G-domain. Although factors other than the salt bridge (Rho) or water mediated network (Ras) may influence GTP hydrolysis, the clear role that these networks have in stabilizing the active site provides a compelling correlation between the helix3/switch II features and hydrolysis rate constants. Both RhoA and Ras contain catalytic glutamine at the corresponding position (Q63^RhoA^/Q61^Ras^) in switch II, with a similar conformation in the GAP-bound structures suggesting similar mechanisms of hydrolysis. The conservation of active site features that connect switch II and helix 3 to the catalytic glutamine suggests that intrinsic hydrolysis also occurs through similar mechanisms in Ras and Rho. Whereas H-Ras requires coordination of several water molecules to efficiently position the catalytic glutamine (Fig. 1B), RhoA evolution has favored substitution of polar or hydrophobic residues in Ras for charged residues in RhoA (Fig. 8A). Thus, coordination between residues in helix 3 and switch II is required in both subfamilies, with Rho more frequently sampling of the catalytic conformation, stabilized by intramolecular forces rather than relying on solvent molecules to bridge the interactions as is the case for Ras. In addition to the Rho salt bridge, chief among residues responsible for active site integrity are D13 (A11^Ras^) and K98 (Q95^H-Ras^/H95^K-Ras4B^/L95 ^N-Ras^) which connect the P-loop and helix 3, in Rho but not in Ras. Further stabilization of the active site may be promoted through packing of nearby residues M82^RhoA^/C80^Ras^ and W99^RhoA^/Y96^Ras^, with bulkier side chains in RhoA favoring a more tightly packed core. This core helps stabilize the aliphatic portion of the R70^RhoA^/R68^Ras^ residue, while its guanidinium group interacts with charged or polar residues E64^RhoA^/E62^Ras^ or E67^RhoA^/S65^Ras^, E102^RhoA^/Q99^Ras^, and Y74^RhoA^/M72^Ras^ to position Q63^RhoA^/Q61^Ras^ in the active site. In RhoA a series of ionic interactions, centered on the R70-E102 salt bridge, replaces the water-mediated network we have seen in H-Ras linking the switch II/helix 3 regions to the active site (Figs. 3 and 7A). The greater decrease in intrinsic hydrolysis of GTP upon removal of R70 in Rho compared to removal of R68 in Ras is notable. A key feature of active site dismantling observed in Rho R70A is that E102 shifts to interact with K98, contributing to disconnection observed between helix 3 and switch II in this mutant. In Ras R68A, Q99 in helix 3 (Rho^E102^) still maintains contact with D69 (Rho^P71^) in switch II, while Q95 (Rho^K98^) becomes disordered.Figure 8**Rho allosteric network and sequence conservation.**A, allosteric communication from active site residues Q63 and D13 (yellow) to the Rho salt bridge (dark blue) is illustrated within the wild-type RhoA crystal structure. Breaking the salt bridge either due to fluctuations within the wild-type dynamic landscape or artificially through mutation indicates that there are conditional interactions with residues K98, Y74 and D67 (teal). Communication to the P-loop occurs through K98 and D13 (stable interaction). Q63 conformational space is largely influenced by switch II organization, orchestrated in part by R70 and D67 interaction. B, the sequence identity conservation of the Rho (top) and Ras (bottom) subfamilies are mapped onto respective crystal structures.

Overall, it appears that the more dispersed network of polar residues lining switch II conveys flexibility that may have co-evolved with the ability to modulate GTP hydrolysis via the allosteric site in Ras, with coupling between loop 7, helix 3 and helix 4 to the active site (40). This is not possible in Rho due to its more rigid, centralized active site stabilization through the Rho salt bridge. Consistently, in RhoA, loop 7 is significantly shorter, creating a shallower region than the allosteric pocket seen in H-Ras (Fig. 8B). While allosteric modulation in Ras requires a flexible link between switch II and helix 3, the higher active site integrity in Rho proteins could be critical in its ability to respond to the tight feedback mechanisms at the heart of Rho self-organization, allowing quick response to GEFs and GAPs that bind at the switch regions with tightly controlled spatial-temporal constrains (4). We propose that in Rho proteins the introduction of the salt bridge circumvents the role of the allosteric site observed for Ras on active site stabilization associated with intrinsic GTP hydrolysis. Given different constraints associated with signaling outcomes, each subfamily evolved features essential to its unique functional roles.

The molecular dynamics simulations presented here reconcile the differing GTP hydrolysis rate constants observed between wild-type, R70A, and E102A RhoA. First, most of the conformations sampled within the wild-type trajectories places Q63 within favorable distance to participate in intrinsic hydrolysis. A smaller fraction of the simulation is also spent with a disassembled active site, which is indicative of a State 1 conformation. This State 1/State two balance of GTPase conformations impacts catalytic activity and signaling (49). Mutated RhoA GTPases R70A and E102A both display hindered GTP hydrolysis, but to different degrees. The RhoA R70A mutant has a nearly unmeasurable rate of intrinsic hydrolysis, while RhoA E102A still has a significant rate constant, albeit lower than that of wild-type RhoA. This is consistent with the fact that the active site in E102A retains features of the wild-type protein, such as the interaction between K98 and D13 on the P-loop and some connection between switch II and helix 3. This is in contrast with the sustained disruption to the active site organization observed for R70A, likely due to the increased availability of E102 for interaction with K98 and a disconnection between switch II and helix 3. In the wild-type RhoA, K98 typically interacts with both E102 and P-loop residue D13. However, in the R70A mutant, with K98 no longer stabilizing the P-loop and switch II disconnected from the G-domain, the active site becomes highly flexible and essentially unable to catalyze hydrolysis of GTP.

The present work suggests that in RhoA, coordination from switch II towards the active site is funneled, in part, by K98 coordination to the P-loop via D13. D13 is a subfamily-specific residue to Rho GTPases (72) that is 100% conserved in non-Miro Rho GTPases. Miro GTPases contain a glutamate residue in place of the aspartate. Lysine 98 is also well conserved in Rho subfamily lineage, aside from the Miro subgroup, where it is replaced with hydrophobic residues. A handful of other Rho subfamily members (RhoJ, RhoQ, and Rho4) do not contain a lysine at this position. RhoJ and RhoQ contain a glutamate residue, and Rho4 has a leucine residue. All three of these subfamily members contain the conserved salt bridge, potentially suggesting alternative modes of communication to the active site in these proteins. The conservation of K98 within the Rho subfamily illustrates how nodes within the allosteric network between switch II/helix 3 and the Rho active site are functionally relevant towards tuning intrinsic GTPase regulation. Likewise, the entirety of switch II shows remarkable conservation within the Rho subfamily (excluding Miro GTPases), with an average sequence conservation of 88% (±9%, 158 sequences) showing almost no deviation from consensus identity within residues 60 and 75 (Fig. 8B). In contrast, Ras GTPases vary by 74% (±17%, 172 sequences) in sequence identity in the same region. A likely driving factor in this divergence is the fact that the catalytic glutamine residue (Q61^Ras^) is only 74% conserved in the Ras subfamily, compared to the 88% sequence identity conservation in Rho GTPases. Thus, the evolution of this region of the G domain is likely coupled to the distinct modes of intrinsic regulation of GTP hydrolysis and other factors, such as effector binding. While evolutionarily conserved in the Rho subfamily, K98 is replaced by a highly variable position within the Ras subfamily, residue 95^Ras^. While few residues vary between the G-domains of the canonical Ras isoforms (H-, K-, and N-Ras), position 95 features three distinct amino acid identities between the isoforms. We have reported that different residues at this position in Ras influences switch II dynamics (71). However, the impact on the P-loop is indirect if any, as H-, K-, and N-Ras lack an equivalent to D13^Rho^. It is likely that this communication route is entirely unique to the Rho subfamily, resulting from residue coevolution within this region.

In addition to being an evolutionarily conserved feature of Rho GTPases, the Rho salt bridge is also a frequent mutation site within tumor samples, regardless of subfamily member. Oncogenic mutations found at the two salt bridge residues and its associated network in Rho proteins would be expected to impair hydrolysis of GTP due to misplacement of Q63, with increased levels of activated Rho. This would alter the delicate balance of Rho self-organization that must be maintained for proper cell motility and function (4). Out of the 16 Rho subfamily proteins evaluated in the COSMIC database (73), 11 subfamily members have mutations in RhoA position 70. In contrast, P-loop mutations exist in 6 to 8 subfamily members. Within the G-domain of Rho proteins, only RhoA position 64 has a higher frequency of mutation across the subfamily (12 members). Of the 11 Rho subfamily members that contain R70 mutations, six Rho subfamily members have this mutation in multiple tumor samples. Of all the samples, skin melanomas (6 samples), large intestine (5 samples) and lung (4 samples) carcinomas have the highest tissue frequencies of R70 mutations. Genotypes of skin melanoma samples contain mutations within the R70-equivalent position of RhoD (R82Q and R82L), RhoQ (R74L), Rac2 (two instances containing R68W), and Rnd1 (R78C). Rac1 GTPase contains the highest frequency of R70-equivalent mutations (R68 in Rac1), with eight tumor samples containing mutations to serine (one sample, lung carcinoma), cystine (three instances in large intestine, small intestine, and soft tissue tumors), or histidine (two instances in large intestine carcinomas, one instance each in endometrium and lung carcinomas). RhoA mutations to position 70 are recorded in four tumor samples as either serine (two samples, lung carcinoma and pancreatic neoplasm) or lysine (two samples, lung and kidney carcinomas). Likewise, E102-equivalent positions across the subfamily are mutated in cancer cell samples but at a lower frequency. Eight subfamily members contain mutations to E102 within tumor samples, ranging between one to two tumor samples per subfamily member. Skin melanomas are the most represented tumor type containing this mutation (RhoH E101K and two instances of Rnd1 E110K). Other tumors that contain a mutation in the E102 position include stomach carcinomas (RhoA E102K and Rac1 E100D), ovarian carcinoma (RhoB E102K) and endometrium carcinoma (Rac2 E100D). Interestingly, nearly all Rho subfamily members with this type of mutation substitutes the glutamic acid 102 for a lysine. However, Rac-type GTPases show a preference for E102D mutations in cancer tissue. Both types of mutations would disrupt the integrity of the Rho salt bridge, very likely favoring effects captured in our biochemical, structural, and biophysical data to promote tumorigenesis.

The region between switch II and helix 3 within Ras subfamily GTPases has been validated as a drug target, particularly with the advent of covalent drugs targeting oncogenic K-Ras (74, 75, 76, 77), along with close homolog Ral GTPase (78). This druggable “switch II pocket” is also accessible when Ras is bound to GDP, regardless of covalent drug binding mechanisms (79). Our work illustrates that this region in Rho subfamily proteins is vulnerable to mutations that result in the decoupling of the switch regions from the rest of the G-domain. This leads to inhibition of GTP hydrolysis and sustained signaling through Rho GTPases, contributing to important hallmarks of cancer such as metastasis (80). This idea is supported by our aMD simulations of the switch II R70A mutant where we clearly observe destabilization of the nucleotide binding site. Work by Sun et al. in 2020 shows that a covalent inhibitor of RhoA residue C107 disrupts the Rho salt bridge in the GDP-bound state (81). The structures of drug-bound RhoA-GDP are found in a conformation like that represented by peak 4 in our RhoA R70A aMD simulations (RMSD of 1.97 with PDB ID 6KX3) (Fig. S11). This conformation also resembles the RhoA-GDP structure without the bound drug in this study, PDB ID: 6KX2. Excitingly, peak 4 in RhoR70A-GTP represents a minor conformer that resembles the inactive RhoA-GDP, with switch I and Q63 placement particularly striking in similarity. Interestingly, the increased dynamics of the Rho insert within R70A simulations, is also characteristic of GDP-bound Cdc42 GTPase (82). In the alignment, we also see that the Rho inhibitor indirectly displaces network residue Y74. This creates a pocket that bears resemblance to our peak 4 R70A population. We discussed the role of Y74 movement within R70A subpopulations and how this conformational change propagates to the active site. Our simulations support the possibility that the covalent inhibitor (75) could selectively bind the rare conformation represented by peak 4 in GTP-bound Rho-R70 mutants associated with cancers. This would shift the equilibrium of conformational states to significantly populate the GDP-like inactive conformation, potentially rendering the GTP-bound state incapable of signal propagation.

Overall, our work sheds light on the evolutionary differences between Rho and Ras subfamily members, pointing to co-variations of residues near the active site that have an impact on GTP hydrolysis. Our structural discoveries, augmented by the simulation studies on the Rho wild type and salt bridge mutants, add to our fundamental understanding of mechanisms through which these residues collectively contribute to active site organization, and justifies testing an inhibitor of RhoA as a possible drug against mutant Rho GTPases in the fight against cancer. Understanding fundamental mechanisms, such as the allosteric pathways that radiate from the conserved Rho salt bridge to the active site and to distal regions such as the Rho insert is not only important for elucidating co-evolutionary connections with functional relevance but is also critical for guiding specific strategies to targeting Rho proteins associated with disease.

Experimental procedures

Evolutionary sequence conservation analysis

The InterPro server (83) was used to acquire all Rho-like and Ras-like sequences, regardless of taxonomy. Subfamily classifications were defined as proteins falling in the acquisition numbers IPR020849 (Ras subfamily) and IPR003578 (Rho subfamily). Sequences were filtered to those that had been classified as “reviewed” within the Uniprot database. The resulting sequence datasets included 172 Ras-like proteins and 198 Rho-like proteins. Alignments of these two datasets were performed via the Clustal Omega MSA tool provided by EMBL-EBI (84) without customized parameters.

Ancestral tree construction was executed via the Jalview software package (85) running BLOSUM62 alignment calculations. The calculated tree of the Rho subfamily was used to distinguish Miro GTPases from other Rho subfamily GTPases, accounting for 40 Miro sequences. Sequence conservation scores per position were also obtained via Jalview consensus sequence calculations.

Protein expression and purification

Protein constructs of RhoA^WT^, RhoA^R70A^, and RhoA^E102A^ (residues 1–181) on pRSF-1b plasmid containing a kanamycin resistance plasmid were obtained from glycerol stocks of E. coli Rosetta competent cells. Cells were grown in 250 ml culture of 25 g/L Luria Broth solution (LB media) with 50 mg/ml kanamycin and 36 mg/ml chloramphenicol with shaking at 225 rpm at 37 °C for 14 to 16 h. Culture was split into six-one-liter volumes of LB media and grown under the same conditions until a OD_600_ of 0.6 was obtained. Culture was induced with the addition of 525 μM IPTG solution and the temperature lowered to 32 °C for 5 h. Cells were pelleted by ultracentrifugation at 7 Kg for 30 minutes before harvesting the pellet. Cells were stored at −80 °C for storage prior to purification.

All proteins were purified utilizing an Äkta Pure FPLC (Cytiva Life Sciences) through a multi-step process consisting of Q-Sepharose F and size-exclusion chromatography, similar to our standard purification for Ras (86). Buffer A is a standard buffer utilized for all purification experiments 20 mM Tris-HCl, pH 8.0, 10 to 50 mM NaCl, 5 mM MgCl_2_, 5% v/v glycerol. The bacterial pellet was resuspended in a 1 g bacterial mass/10 ml cell resuspension solution (CRS), which is buffer A with the addition of 1 mM dithiothreitol (DTT), 20 μM GDP, and a protease inhibitor cocktail of 0.5 mg benzamidine chloride/10 ml CRS, Pepstatin A, and Leupeptin. Lysis of the cell membrane to release the soluble RhoA constructs resulting from expression is done through ultrasonication of the mixture using a 30 s ON/OFF pulse method at 50%-100% amplitude for 5 min. For anion exchange, buffer B contains the same components as buffer A, but with 1 M NaCl. RhoA is separated on a HIPrep Q-FF 16/10 column (Cytiva Life Sciences) using a flow rate of 3.5 ml/min over 10 to 12.5 CV. RhoA typically elutes at 12% buffer B. Fractions containing RhoA were concentrated to less than 2 ml and applied to a Sephacryl S-200 26/60 size exclusion chromatography (SEC) column and eluted for 1.5 CV in Buffer A. Fractions containing RhoA are applied to a HiTrap QHP column using buffer A and eluted with a gradient of 0 to 20% buffer B for 30 CV. RhoA elutes in roughly 5% buffer B. Final RhoA-containing fractions are concentrated to roughly 20 mg/ml and stored in −80 °C. H-RasR68A was expressed and purified according to previously published methods (86).

Nucleotide exchange of RhoA and H-Ras GTPases

The endogenous nucleotide bound to RhoA and H-Ras GTPases in this study is exchanged for a non-hydrolysable GTP analog, either GppNHp or GTPγS. For all proteins, A NAP-25 desalting column was equilibrated with exchange buffer (50 mM Tris-HCL pH 8.0, 5 mM MgCl_2_, 10 mM EDTA, and 10 mM β-mercaptoethanol). Purified RhoA (1 ml) was applied to the column and eluted with exchange buffer. Either GppNHp or GTPγS was added to the buffer-exchanged GTPase pool (1 mg nucleotide per 10 mg protein) and incubated at 42 °C with rocking for 1 h. The exchange was quenched with addition of 20 mM MgCl_2_. Using another NAP-25 column, GTP-analog-bound GTPase was desalted in stabilization buffer (25 mM Tris-HCl pH 8.0, 2 mM MgCl_2_, and 10 mM β-mercaptoethanol). The resulting protein was concentrated to roughly 10 mg/ml and stored at −80 °C.

Crystallization of RhoA and Ras GTPases

Crystals of WT RhoA bound to the GTP analogue GppNHp were obtained from a fine screen developed from the crystallization conditions listed by Longenecker et al. (47). Hanging drop trays were set with 2 μl RhoA (8 mg/ml) + 2 μl reservoir over a total volume of 500 μl reservoir condition. Square-shaped and rod-shaped crystals appeared within a week of growth at 4 °C. The crystal that provided the resulting structure was grown in 25% dioxane, 17% PEG 8000, and 100 mM HEPES pH 6.9. This crystal was not subject to cryo protection before flash freezing.

Crystals of RhoA mutants bound to GTPγS (8 mg/ml) were obtained after extensive sparse matrix screening using Hampton crystallization screens (PEG Ion Screen I + II and Crystal Screen I + II). After several optimization cycles, good-quality diffraction of RhoA R70A was obtained from crystals grown in 250 mM ammonium sulfate, 1.50 M lithium sulfate, and 100 mM sodium citrate pH 5.6. RhoA E102A crystals were obtained from conditions containing 0.2 M Ammonium Iodide and 20% PEG 3350. Crystals of mutants which provided the best diffraction data were grown in sitting drop trays at 4 °C. Crystals were cryo-protected with 45% glycerol and 55% reservoir condition prior to flash freezing.

H-Ras mutant R68A was crystallized with GppNHp in 0.2 M calcium acetate, 20% w/v PEG 3350 and 0.1% N-octyl-glucoypyranoside (NOG). The sitting drop volume consisted of 2 μl protein and 2 μl well condition, with 500 μl well volume. Crystallization occurred over 1 to 2 weeks at 18 °C. Crystals were cryoprotected using 30% glycerol for data collection.

X-ray diffraction and structure determination

X ray diffraction data for all crystals were collected on a home source Rigaku Micromax 007 HF rotating anode generator coupled with a R-axis IV++ image plate. Collection occurred at 100 K. Diffraction frames were collected in 1° increments over a total phi rotation of 180°. Indexing and scaling operations on the data were performed using HKL 3000 (87). The program Phenix was then used to perform crystal quality analysis (Xtriage), molecular replacement (MR-phaser), and model refinement (phenix.refine) (88). The scattering data of wild-type and mutant RhoA crystals were phased using the crystal structure of RhoA Q63L bound to GTPγS (PDB ID: 1KMQ (27)) via molecular replacement, while H-Ras R68A was solved using wild-type H-Ras bound to GppNHp and calcium acetate (PDB ID: 3K8Y). The phase model was manually altered using COOT to best fit the density calculated from the experimental data (89). Iterative alterations in COOT and refinements in Phenix were performed until optimal R-factors were reached. The placement of iodide ions within the RhoA E102A structure was confirmed using anomalous difference Fourier maps.

Single-turnover GTP hydrolysis assays

The GTP hydrolysis experiments were performed using a protocol that we previously described in detail (60). Briefly, all GTPases (Table 2) were purified in the GDP-bound state using QHP anion exchange chromatography and the protein concentration was determined by the Bradford assay. A protein concentration of 5 μM was used in all hydrolysis experiments. Each Rho-containing mixture was incubated for 5 min at 37 °C with 50 nM GTP-^33^Pγ (GTP containing ^33^P at the gamma phosphate) purchased from American Radiolabeled Chemicals (ARC), 1 mM EDTA and 20 mM HEPES (pH 8.0). The same procedure was used for H-Ras R68A, with GTP-^32^Pγ (50 nM, PerkinElmer), 1 mM EDTA, 20 mM Tris (pH 8.0), and 2 mM DTT. Reaction is started through addition of a buffered solution containing a 5-fold excess of MgCl_2_ (5 mM) in 20 mM HEPES or Tris (pH 8.0 at 37 °C) with 100 mM NaCl and 1 mM DTT. Reaction time-points were taken up to t_max_=420 min at the following timepoints: 0, 5, 10, 15, 20, 25, 30, 35, 40, 60, 75, 90, 120, 150, 180, 210, 240, 300, 360, 420 min. Free-phosphate released by the reaction is captured through organic extraction and the amount of radiation is quantified through liquid scintillation measured using a Hidex 300 SL. All experiments were performed in triplicate. Curve fitting was done to determine the hydrolysis rate constant (k_hyd_) as the best-fit value using the DynaFit software (90) to perform a global fit.

Accelerated molecular dynamics (aMD) simulations and reweighing

The RhoA^WT^, RhoA^R70A^, and RhoA^E102A^ crystal structures were utilized as the starting models for accelerated molecular dynamics simulations (aMD), a form of enhanced sampling that utilizes a non-negative boost potential to bias the simulation and smooth higher energy barriers between transitions (61). All water molecules and ions were removed from the coordinate sets, including the iodide ions found in the RhoA^E102A^ structure. VMD/NAMD was utilized to produce structures protonated using the CHARMM36 topology file and solvated using the TIP3 water model. Neutralization of the solvent box was achieved by 0.15 g/L NaCl. Classical molecular dynamics simulations were performed for 10 ns before the boost potential was applied. The boost factors were determined utilizing the average of the previous 5 ns. The average dihedral energy (DE) and potential energy (PE) values were harvested using data_avg function in VMD for calculation of dual boost parameters. Parameters applied are accelMDTE ( $[eqn]$ ), accelMDTalpha ( $[eqn]$ ), accelMDE ( $[eqn]$ ), and accelMDalpha ( $[eqn]$ ). Each aMD ran for 200 ns simulation time. Temperature was gradually brought to 300 K from a starting point of 0 K utilizing three stepwise processes: 0 K to 50 K at 5 K increments (abeta), 50 K to 100 K at 2 K increments (heat 1), and 100 K to 300 K in increments of 2 K (heat 2). Classical MD simulations ran for 10 ns before applying the boost potential for aMD for the remainder of 200 ns. Five replicates, each with a unique randomly generated velocity, were performed for each structure with periodic boundary conditions and the CHARMM36 parameters. Given the five replicas, each structure was simulated for a total of 1 μs. The five replicas were combined and used jointly for analysis of each RhoA structure.

The boost potential applied to aMD produces a flattened potential energy surface that cannot be used to study the population of conformational states or any property of the system that is energy dependent. To resolve this problem, we applied a Boltzmann re-weighting function to the simulation trajectory as previously published (60, 61). Briefly, the dV values corresponding to each step in the simulation were harvested using python. Transformation of the dV values into weight value utilizing the Boltzmann method (61) utilized R scripting. Distance measurements were calculated using Bio3D (91) and the weights were applied using the R density function. The weighted-average structural coordinates were calculated using the colWeightedMeans function from matrixStats (92).

Data availability

All crystal structures solved in this manuscript have been deposited to the Protein Data Bank: Wild-type RhoA-GppNHp, PDB ID 9N4A; RhoA R70A-GTPγS, PDB ID 9N4B; RhoA E102A-GTPγS, PDB ID 9N4C; and H-Ras R68A-GppNHp, PDB ID 10DC. Sequence alignments used in the analysis of evolutionary trends, tumor sample mutation data obtained from the COSMIC database, and average structures derived from the population distributions as described in Figure 6 can be accessed through Github (https://github.com/MLabAmherst/RhoSaltBridge2026).

Supporting information

Supporting Information Figs. S1–S11 can be found in the accompanying supplemental information document. This article contains supporting information.

Conflicts of interest

The authors declare that they have no conflicts of interest with the contents of this article.