Allosteric site variants affect GTP hydrolysis on Ras

Abstract

RAS GTPases are proto‐oncoproteins that regulate cell growth, proliferation, and differentiation in response to extracellular signals. The signaling functions of RAS, and other small GTPases, are dependent on their ability to cycle between GDP‐bound and GTP‐bound states. Structural analyses suggest that GTP hydrolysis catalyzed by HRAS can be regulated by an allosteric site located between helices 3, 4, and loop 7. Here we explore the relationship between intrinsic GTP hydrolysis on HRAS and the position of helix 3 and loop 7 through manipulation of the allosteric site, showing that the two sites are functionally connected. We generated several hydrophobic mutations in the allosteric site of HRAS to promote shifts in helix 3 relative to helix 4. By combining crystallography and enzymology to study these mutants, we show that closure of the allosteric site correlates with increased hydrolysis of GTP on HRAS in solution. Interestingly, binding to the RAS binding domain of RAF kinase (RAF‐RBD) inhibits GTP hydrolysis in the mutants. This behavior may be representative of a cluster of mutations found in human tumors, which potentially cooperate with RAF complex formation to stabilize the GTP‐bound state of RAS.

1. INTRODUCTION

Understanding how proto‐oncogenes regulate cell signaling is critical to understanding their role in disease. For instance, KRAS, NRAS, and HRAS (collectively RAS) are peripheral membrane hub proteins that activate many different signaling pathways, including the RAF/MEK/ERK signaling pathway, which is essential for cell differentiation, proliferation and survival (Fernandez‐Medarde & Santos, 2011). RAS proteins regulate cell signaling through the controlled cycling of GTP‐ and GDP‐bound states. When bound to GTP, RAS is in a signaling active state that interacts with downstream effectors, such as RAF kinase (Johnson & Mattos, 2013; Lu et al., ²⁰¹⁶). When Ras is bound to GDP, it is in its inactive state that cannot bind effectors. Point‐mutations in KRAS, NRAS and HRAS act to stabilize the GTP bound state and are found in 16%, 2.6%, and 0.9% of all cancers, respectively (Consortium APG, 2017). Unfortunately, despite the recent approval of KRAS‐specific inhibitors Sotorasib and Adagrasib which target G12C mutations, the vast majority of KRAS, NRAS, and HRAS mutants remain elusive to therapeutic intervention (Johnson et al., 2022). Therefore, more work is needed to understand the molecular mechanisms through which RAS and its effectors work.

Inactivation of RAS in cells is driven by hydrolysis of GTP to GDP. This can occur through GTPase‐activating proteins (GAPs) which bind to the active site of RAS (Scheffzek et al., 1997). This mechanism is widely viewed as critical to maintaining the normal functions of RAS, because many GAPs (e.g., NF1, p120GAP) are found inactivated in tumors (Consortium APG, 2017). However, regulation of the GTP bound state of RAS by GAPs alone is difficult to reconcile for several reasons. The p120GAP (also called RASA1) protein cannot compete with Ras effectors, including RAF, under physiological‐like conditions (Smith & Ikura, 2014), and it is well documented that RAF blocks GAP activity by binding to RAS (Moodie et al., 1995; Scheffler et al., ¹⁹⁹⁴; Smith & Ikura, ²⁰¹⁴; Warne et al., ¹⁹⁹³; Zhang et al., ¹⁹⁹³). Furthermore, the spectrum of cancers that frequently show NF1 and p120GAP inactivation weakly overlap with those with activating RAS mutations (Consortium APG, 2017). Thus, we propose that an alternative means to inactivate RAS may rely on enhancing intrinsic hydrolysis when RAS is bound to RAF (Buhrman et al., 2010). While intrinsic hydrolysis occurs slowly under most studied conditions (0.006–0.019 min⁻¹; Johnson et al., ²⁰¹⁷), we put forth a model whereby intrinsic hydrolysis could be enhanced in a regulated manner from an allosteric site via a network of water‐mediated H‐bonding interactions to the RAS active site (Buhrman et al., 2010; Buhrman et al., ²⁰¹¹). In this model (Figure 1a, left), stabilization of the active site into a closed conformation (e.g., state 2) by RAF kinase, or another effector, increases dependency of the intrinsic hydrolysis reaction on the conformation of switch II (Figure 1a, right). Structural analysis suggests that movement of helix 3 toward a pocket on the other side of the protein (allosteric site, Figure 1a) allows switch II to become ordered in the active site, placing catalytic residues near GTP. We hypothesize that these ground state changes in conformation poise RAS for accessing the intrinsic GTP hydrolysis transition state.

FIGURE 1.

Mutations at R97 induce increased packing in the allosteric site. (a) Model for the regulation of intrinsic hydrolysis by allosteric site transitions. (b) Residue interactions of the allosteric site when HRAS is in the R‐ and T‐states.

The active site of HRAS undergoes several conformational transitions once it becomes bound to GTP (Lu et al., 2016; Volmar et al., ²⁰²²). Here, we differentiate the sub‐states of state 2 relevant to our model of hydrolysis as the inactive “T‐state” (gray, Figure 1a) and the catalytically active “R‐state” (green, Figure 1a; Johnson & Mattos, ²⁰¹³). In our previous crystallography experiments, the transition of HRAS from the T‐state to the R‐state was accomplished by either growing or soaking protein crystals with Ca(OAc)₂. Calcium and acetate are likely ligand mimetics as they bind weakly to the allosteric site in solution (O'Connor & Kovrigin, ²⁰¹²). In the presence of bound Ca²⁺ and acetate, helix 3 (residues 93–104) and loop 7 (residues 105–108) shift toward helix 4, allowing switch II (residues 60–76) to stabilize the R‐state conformation (green, Figure 1b). This promotes a pre‐transition state where Y32 from switch I (residues 28–40) and Q61 from switch II (residues 60–76) interact with a “bridging” water molecule that also forms an H‐bond with the γ‐phosphate of GTP. This water molecule, which carries a partial positive charge, neutralizes the accumulation of negative charge at the β‐ and γ‐phosphates of GTP during the hydrolysis reaction (Du & Sprang, 2009). The bridging water molecule is thus positioned to lower the transition state energy associated with GTP hydrolysis similar to the arginine finger introduced by GAPs (Buhrman et al., 2010; Knihtila et al., ²⁰¹⁵).

In the present work, we explicitly studied the relationship between allosteric and active sites to understand the mechanism through which the disorder to order transition in switch II completes the active site for GTP hydrolysis. In the wild type T‐state, R97 is partially buried in a relatively large hydrophobic pocket at the base of the allosteric site (dashed oval, Figure 1b). In the wild type R‐state, R97 transitions out of the hydrophobic pocket and forms a salt‐bridge with the acetate molecule, which also interacts with the bound calcium ion. With the help of an H‐bond between Y137 (helix 4) and H94 (helix3), the bound calcium and acetate ions promote a shift in helix 3 toward helix 4 (Figure 1b). We hypothesized that hydrophobic mutants of R97 would similarly promote the shift of helix 3 toward helix 4, by packing in the hydrophobic pocket, thereby facilitating the R‐state transition in a ligand‐independent manner.

To test the relationship between the R‐state and intrinsic GTP hydrolysis, we crystallized seven R97 mutants in different crystal forms to determine the extent to which each mutation promoted conformations associated with the R‐state, and then correlated these changes to measurements of GTP hydrolysis rate constants. By combining our structure and hydrolysis data, we show that closure of the allosteric site (e.g., shift of helix 3 toward helix 4) correlates with an increase in GTP hydrolysis. Unexpectedly, RAF inhibits hydrolysis to varying degrees in a mutation‐specific manner. This potentially explains the presence of allosteric site mutations of RAS found in some human tumors. Our work is a proof‐of‐principle demonstration showing that manipulation of the allosteric site leads to changes in hydrolysis, paving the way for the development of pro‐hydrolysis RAS inhibitors.

2. RESULTS AND DISCUSSION

2.1. R‐ and T‐state classification of crystalized HRAS R97 mutants

HRAS crystals with R32 space group symmetry have switch II and the C‐terminal end of helix 3 away from crystal contacts, allowing for transition between R‐ and T‐states in the crystals (Buhrman et al., 2010; Holzapfel et al., ²⁰¹²). Thus, this crystal form is ideal for assessing the effects of the R97 mutants on the switch II/helix 3/allosteric site conformational states (Table S1). Mutants of HRAS in the R‐state showed stabilization of the bridging water molecule in the active site, stabilization of switch II into an ordered helix, and movement of helix 3 toward helix 4 (Figure 2a). The nucleophilic water was present in each of our crystal structures (Figure 2a,b).

FIGURE 2.

Classification of R32 protein crystals from HRAS R97 mutants. (a) R97V, R97L, R97M, and R97I conserve all features of the R‐state except the conformation of Q61. Inset shows wildtype HRAS in the R‐state (PDB code 3K8Y) (Buhrman et al., 2010). The catalytic residues Y32 and Q61, and the bridging and nucleophilic waters of wildtype HRAS are shown in black for reference. (b) R97A, R97G, and R97F have active site features similar to the T‐state. Inset shows Q61 and nucleophilic water in black for wildtype HRAS in the T‐state (PDB code 2RGE) (Buhrman et al., 2007). (c) Comparison helix 3 and switch II in each of the R97 mutants. Wildtype reference structures are included in bold for the R‐state, T‐state and the T‐state when HRAS is in the “anti‐catalytic” conformation (PDB code 4DLZ) (Holzapfel et al., 2012). In light gray is an overlay of the R‐state. The black arrow denotes the location of Y96. The dashed line in (c) of R97F refers to its disordered portion of switch II.

The R‐state favoring mutants R97V, R97I, R97L, and R97M, stabilized the bridging water molecule in the active site via an H‐bond with Y32 (Figure 2a, yellow dashes). Likewise, switch II of these mutants was well ordered and formed an H‐bond network with helix 3 consistent with the network described for wildtype HRAS in the R‐state (Buhrman et al., 2010), including participation of Y96 which was shifted toward switch II in these mutants (arrow, Figure 2c). However, despite the overall features of the R‐state, none of the R97V, R97I, R97L nor R97M structures captured Q61 in a conformation that allowed H‐bonding with the bridging water molecule (Figure 2a,c).

The R97A, R97F, and R97G structures showed active site and switch II features more like the T‐state, including absence of the bridging water molecule and a direct interaction between the side chain of Y32 and the γ‐phosphate of GTP (Figure 2b). Typically switch II in the T‐state is disordered (residues 61–69), however it can be stabilized in an “anti‐catalytic” conformation that is overall similar to small GTPase RAN in complex with IMPORTIN‐β (PDB code 1IBR) (Vetter et al., 1999), which we previously also observed for the oncogenic mutant HRAS Q61L (Buhrman et al., 2007; Buhrman et al., ²⁰¹⁰). While R97A adopts the “anti‐catalytic” conformation with helix 3 in the T‐state, R97F is in the T‐state with an extended disorder of switch II, from residues 61 to 72 (Figure 2c). In R97G, electron density supports a switch II helical structure similar to the R‐state (Figure 2c). Consistent with these features, Y96 in R97G is in an R‐state position, whereas Y96 in R97A and R97F are in a T‐state position (Figure 2c). Thus, while the active site of HRAS R97G adopts the T‐state, its helix 3 adopts an R‐state like conformation, indicating a decoupling of the two sites in the HRAS R97G variant.

The overall structural features observed for the allosteric site mutants support our initial hypothesis that mutation of R97 to hydrophobic residues with side chains of moderate size induces a shift of helix 3 to promote the R‐state independent of bound Ca²⁺ and acetate. This is consistent with our expectation of a functional connection between the allosteric and active sites of HRAS. As shown in Table S2, there was no apparent correlation between crystallization condition (i.e., Ca(AOc)₂, PEG) and whether the crystallized proteins favored the R‐ or T‐states, indicating that the R97 mutation is the main determinant. However, the observation that only a subset of these mutations induces this shift suggests that both the extent of side chain packing, and the relative position of those amino acid residues in the allosteric site packing interaction, play a role in the transitions between R‐ and T‐states. This is exemplified by our R97A, R97F and R97G mutants. R97F is too large to pack in the hydrophobic cluster in the allosteric site to stabilize helix 3 in the R‐state, while the smallest side chain at R97 (i.e., R97A) must not pack at all and leaves helix 3 shifted toward switch II in the T‐state. Increasing helix 3 flexibility by removing the R97 side chain altogether (i.e., R97G) appears to decrease the barrier between states, resulting in a mixture of the R‐ and T‐states instead of the more discrete states seen in the other mutants.

2.2. Allosteric site mutants stabilize the R‐state through compensatory interactions

We next focused on the details of packing in the allosteric site to understand the mechanism through which each R97 mutant favors the active site in the R‐state or T‐state. In our canonical R‐state structure of HRAS, Ca²⁺ and acetate bind in the allosteric site to stabilize helix 3 toward helix 4, with backbone interactions between Y137 of helix 4, R97 in helix 3, and D107 of loop 7 (see inset, Figure 3a). Although Ca(AOc)₂ is present in the crystallization mother liquor in all R32 crystals, calcium and acetate did not bind in the allosteric site in any of the structures and instead we observed a single water molecule in each of the R‐state favoring structures (“allosteric site water”, Figure 3a). This water molecule forms H‐bonds with the backbone carbonyl groups of both Y137 and D107, replacing the interactions observed for Ca²⁺ in the wild type HRAS structure. Simultaneously, the side chain of Y137 shifts slightly to optimize packing with the various hydrophobic side chains at residue 97. Furthermore, K101 in R97V, R97L, and R97M was shifted toward the solvent to overlap with the position occupied by R97 in the wild type HRAS in the R‐state. In the R‐state mutants, the conformation of K101 is stabilized by proximity to D107 (Figure 3a). Thus, in addition to the water molecule between helix 4 and loop 7, the interaction between the side chains of D107 and K101 appears to compensate for the loss of the acetate‐R97 interaction. R97I is an exception, as K101 is disordered in that structure due to D107 taking on an alternate conformation (magenta structure, Figure 3a).

FIGURE 3.

R97 mutants promote compensatory interactions in the allosteric site and hydrophobic pocket. (a) R‐state favoring mutants show a water molecule (allosteric site water) in the allosteric site that replaces Ca²⁺ while K101 compensates for loss of the guanidinium group of R97. Inset shows the allosteric site of wildtype HRAS in the R‐state (PDB code 3K8Y) (Buhrman et al., 2010). Y137, R97, K101, D107, acetate and water molecules coordinated to Ca²⁺ (green) are shown in black. (b) With the exception of R97G, the T‐state favoring mutants do not show compensation by K101, nor closure of the allosteric site. Inset shows the allosteric site of wild type HRAS in the T‐state (PDB code 2RGE) (Buhrman et al., 2007), with Y137, D107, R97, and K101 in black for reference. (c) and (d) are the hydrophobic pockets of R97 mutants crystallized in the R32 space group. (e) and (f) are the mutants crystallized in the P3₂21 space group, with wildtype HRAS shown in black for reference (PDB code 1CTQ) (Scheidig et al., 1999).

Consistent with our active site observations, loop 7 and helix 3 in the T‐state favoring mutants R97A and R97F were shifted away from helix 4. However, the allosteric site water molecule was present in the crystal structure of R97F (Figure 3b), indicating that binding of this water molecule is not sufficient to drive the T‐ to R‐state shift. As expected, K101 in R97A and R97F is oriented away from the allosteric site, as it is in the wild type HRAS in the absence of Ca(AOc)₂ (see inset, Figure 3b), suggesting that this position is a characteristic of the T‐state. In contrast, R97G favored K101 placement into the allosteric site, and overall showed allosteric site features like the R‐state favoring mutants (cyan structures, Figures 2c and 3b).

The overall residue conformations in the hydrophobic pocket for the R97 variants remain similar to those observed in the wild type structures, as they are likely to be important for the stability of the protein core. However, conformational changes do occur for residues I93 and Y137 adjacent to the mutation site (Figure 3c,d). In both the R‐ and T‐states of wild type HRAS, the tyrosyl ring of Y137 interacts with the hydrocarbon moiety of R97 (see inset, Figure 3a,b). In R97M, R97L, and R97I, Y137 also packs against the mutant 97 residue, but is rotated 90° (Figure 3c). The I93 side chain in these structures is found in a different rotamer to make room for the bulky hydrophobic side chains at residue 97. The smaller V97 side chain preserved the conformation of both Y137 and I93 as in the wild type structure. Notably, the rotated tyrosyl group of Y137 H‐bonds with helix 3 residues H94 and E98 in the R‐state mutants, while this is not observed for the T‐state favoring mutants. Though subtle, these compensatory packing alterations demonstrate the influence of the hydrophobic cluster at the base of the allosteric site on the global conformation of HRAS. Likewise, since the R97V structure revealed minimal changes in allosteric site interactions, its conformation of switch II and helix 3 were the most consistent with wild type HRAS in the R‐state (Figure 2c).

In R97F, R97G, and R97A, rotation of Y137 is only seen in the presence of F97, while the conformation of I93 was unaltered in each of these structures (Figure 3d). Thus, changes in the conformation of I93 and Y137 correlate to some degree with the R‐ and T‐states. However, the nature of their influence on this global transition was not immediately obvious. Coincidentally, the HRAS R97 mutants, except for R97V and R97A, also crystallized with P3₂21 symmetry in the same conditions that yielded crystals with R32 symmetry (Table S3). Unlike R32 crystals of HRAS, P3₂21 crystals show more extensive crystal contacts at the allosteric site, helix 3 and switch II (Buhrman et al., 2003; Scheidig et al., ¹⁹⁹⁹). Furthermore, helix 3 in this crystal form is stabilized in the R‐state conformation due to crystal packing. Thus, P3₂21 crystals provide a view of the allosteric effects of the R‐state conformation on I93 and Y137, regardless of whether that mutant favored the R‐ or T‐states in the R32 crystal form. As expected, we observed the same series of packing interactions for the R‐state mutants in the P3₂21 crystals as were seen for their R32 crystals. This is because the observed changes in I93 and Y137 are necessary to accommodate the bulky side chains at residue 97 within the more closely packed allosteric site associated with the R‐state (Figure 3e). Interestingly, while the R97A, R97G, and R97F mutants retain their I93 side chain conformations as observed in the wild type HRAS structures in their R32 crystals, both the bulky R97F side chain and non‐side chain mutant R97G in P3₂21 crystals show I93 side chain conformations as in the R‐state mutants (Figure 3f). While Y137 is near the allosteric site surface with room to adjust its conformation, the I93 residue is packed in the protein core and is located on helix 3 opposite the active site. Thus, the R‐state promoted by crystal packing requires adjustment of I93 for bulky side chain mutations of R97. Conversely, changes in the I93 side chain conformation could affect the active site with a potential impact on GTP hydrolysis.

A role of I93 in the global RAS dynamics has been suggested before from ¹⁵N NMR experiments (Chao et al., 2022; O'Connor & Kovrigin, ²⁰⁰⁸), consistent with our structural observations of changes in conformations of this side chain in response to mutation of R97. Likewise, Y137 is more dynamic than suggested by the crystal structures alone, as it is a site of phosphorylation in HRAS by ABL kinase (Ting et al., 2015). Changes in the conformations of I93 and Y137 revealed in P3₂21 crystals of R97G suggest increased conformational plasticity in the allosteric site in the absence of a residue 97 side chain. In these crystals, I93 is in its alternate rotamer, despite no obvious cause due to packing, and Y137 is in the alternate rotated conformation (Figure 3f), as observed only for the R‐state mutants in R32 crystals. Overall, it appears that Y137 and I93 are connected to the active site in a manner that correlates with the R‐state and T‐state conformations. R‐state mutants require a change in packing at the adjacent hydrophobic cluster, while T‐state mutants accommodate the hydrophobic cluster as in the wild type due to a more spacious allosteric site with helix 3 shifted toward switch II. In crystals where the active site is stabilized in an R‐state like conformation with helix 3 shifted toward helix 4 (i.e., P3₂21 crystals) the conformational changes in I93 and Y137 are observed to accommodate this shift in the context of the bulky mutant side chains as expected. However, adjustment of I93 in response to active site R‐state stabilization without an obvious cause as seen in the R97G mutant suggests that I93 participates in the communication between the allosteric and active sites in a way that is not fully understood.

Under our experimental conditions, neither R97V nor R97A crystalized in the P3₂21 space group. In the case of R97V, this may be due to this mutant favoring an active site most like wild type HRAS in the R‐state (Figure 2c). In contrast, R97A stabilizes ordering of switch II into the anti‐catalytic T‐state conformation, which is inconsistent with crystal packing of P3₂21 (Figure 2c). Given the small side chain of residue A97, it was expected to behave like the R97G mutant, with electron density maps showing a mixture of R‐ and T‐state conformations. However, crystals of HRAS R97A grown at 18°C yielded a structure with full occupancy of the anti‐catalytic T‐state conformation without calcium or acetate bound in the allosteric site (Figures 2 and 3). Interestingly, we did observe the R‐state when crystals of HRAS R97A were grown under the same solution conditions but at 4°C (Figure 4). It appears that a decrease in crystal growth temperature stabilizes the R97A R‐state, with switch II making extensive intramolecular contacts in a more compact structure. At 4°C, R97A adopts a more packed allosteric site, with the water molecule present and both K101 and helix 3 shifted toward the site (Figure 4a). As expected, the conformations of I93 and Y137 were unchanged in this structure, as the alanine side chain allows for the wild type conformations associated with the hydrophobic cluster adjacent to the allosteric site. Likewise, Y96 was shifted toward switch II, and switch II was fully ordered in an R‐state conformation (Figure 4a). The active site was also stabilized in the R‐state (Figure 4b), as observed in all the R‐state favoring R97 mutants (Figure 2a). It is not surprising that a mutation at R97 which has a propensity for both R‐ and T‐state conformations, would favor the more packed R‐state structure at the lower crystal growth temperature and the more disordered T‐state at room temperature, given increased thermal motions.

FIGURE 4.

R97A crystals stabilize the R‐state when grown at 4°C. (a) Crystals of R97A grown at 4°C (beige) stabilize the allosteric site in the R‐state conformation (b) 4°C crystals of R97A show an active site consistent with the R‐state conformation.

2.3. The R‐state enhances GTP hydrolysis by HRAS

Our experiments indicate that the transition between R‐ and T‐states is dynamic and that the balance of conformational states between them can vary both with point mutations and with temperature. To correlate these conformational shifts with biochemical outcomes at physiologically relevant temperature, we performed single‐turnover GTP hydrolysis assays at 37°C (Table S4) and compared the first‐order rate constants of these reactions to allosteric site packing in each of the variants. Since the placement of Y137 and M111 hardly shifts within our current structures, and R102 moves with helix 3, we used Heron's rule to calculate the triangular area between the C_αs of Y137, M111, and R102 in the R32 crystal structures as a way of measuring the allosteric site closure indicative of tighter packing between helices 3 and 4.

Comparisons between allosteric site closure and GTP hydrolysis are shown as orange dots in Figure 5a. Closure of the allosteric site correlates with a small increase in the intrinsic hydrolysis rate constant of GTP, reflecting an increase in R‐state populated HRAS as observed in our crystal structures of the R‐state R97 mutants (R97V, R97L, R97I, and R97M in Figure 5a). Even R97G, which favored R‐state features outside the active site, showed a subtle increase in its hydrolysis rate constant (orange circle, Figure 5a). Likewise, the T‐state favoring mutants R97F and R97A showed slightly lower rate constants than wild type HRAS. For R97A, these hydrolysis data were consistent with the observed anti‐catalytic T‐state conformation at the higher crystallization temperature (red circle, Figure 5a).

FIGURE 5.

GTP hydrolysis is sensitive to both allosteric site mutation and binding to RAF1‐RBD (a) Correlation of allosteric site closure and intrinsic hydrolysis. Orange dots reflect first‐order rate constants for intrinsic GTP hydrolysis of the R97 mutants. Black dots reflect first‐order rate constants for intrinsic GTP hydrolysis of the R97 mutants in the presence of excess purified RAF1‐RBD. (b) Single‐turnover reaction for HRAS R97G alone (red line) or in the presence of excess RAF1‐RBD (blue line). Each dot represents at least triplicate measurements of radioactive P_i at the indicated time point. The initial rates (Pi per second) for the reactions involving HRas R97G in the absence (red) and presence (blue) of RAF1‐RBD are shown to the right of the reaction curves.

Part of the allosteric model for intrinsic hydrolysis involves stabilization of switch I in a closed conformation by the RAS effector RAF (step 1, Figure 1a). Since RAF binds switch I but not switch II, the T‐state mutants are expected to retain a disordered switch II and RAF binding should not alter GTP hydrolysis, as is the case for wild type HRAS (2RGE in Table S4). Indeed, binding of RAF1‐RBD had no effect on GTP hydrolysis for the T‐state mutants (e.g., R97F and R97A) (Table S5). Our model predicts that addition of RAF1‐RBD should synergize with R‐state favoring mutants of R97, stabilizing switch I and switch II to enhance GTP hydrolysis. However, the hydrolysis rate constant was reduced to varying degrees in the R‐state mutants (black dots, Figure 5a). Of special note was GTP hydrolysis by R97G in the presence of RAF1‐RBD, which was too slow to model in any significant way (Figure 5b, left), although a comparison of initial reaction rates showed a 5‐fold decrease in GTP hydrolysis when HRAS R97G is in the presence of RAF‐RBD (Figure 5b, right). Clearly, the ordering of switch II in the presence of RAF1‐RBD with a stabilized switch I is not sufficient to promote GTP hydrolysis. Binding of RAF1‐RBD quenches switch I dynamics in HRAS while simultaneously increasing motions in switch II (Fetics et al., 2015; Li et al., ²⁰²²). Perhaps the alternative conformations of I93 and Y137 we observe in our R97 mutants help to decouple the cooperative motions that must occur between the allosteric site, helix 3, and switch II, particularly regarding the placement of catalytic residue Q61 for hydrolysis of GTP. Taken together, our data show unequivocally that the allosteric site is in communication with the active site of HRAS, where it influences the rate of GTP hydrolysis on HRAS.

3. CONCLUSIONS

Here, we establish a functional link between the active and allosteric sites in HRAS by showing that modulation of the allosteric site by mutagenesis affects the rate constants for GTP hydrolysis. The small change that we observe for each mutant reflects a shift in the balance of conformational states for that mutant, whereas ligand binding to the wild type would be expected to more robustly stabilize the R‐state and result in larger changes in rate constants. None‐the‐less the effects are significant enough to reveal that changes in the allosteric site impact activity. Allosteric modulation of the active site was previously proposed based on structural analysis of HRAS in a crystal form where switch I is in the conformation observed in the RAS/RAF1‐RBD complex and switch II is free of crystal contacts (PDB ID 3K8Y) (Buhrman et al., 2010). An H‐bonding network linking the allosteric site to the active site was clearly observed, leading us to predict that measuring hydrolysis rate constants in the presence of Ca(AcO)₂ should reveal enhanced intrinsic hydrolysis of GTP on RAS, given that the binding of calcium and acetate to the allosteric site was expected to order the active site in solution as it does in the crystals. This turned out not to be the case, consistent with NMR experiments that aimed at probing the metal binding properties of the allosteric site in solution. The NMR experiments revealed weak and non‐selective binding of metal ions (O'Connor & Kovrigin, ²⁰¹²), contrary to what was observed in the crystals, where Ca²⁺ was clearly selected over Mg²⁺ (Buhrman et al., 2010). Since these early experiments, we have shown that RAF1‐RBD promotes robust dimerization of RAS on supported lipid bilayers and to a smaller extent in solution (Packer et al., 2021). The crystal structure of the HRAS/RAF1‐RBD complex (PDB ID 4G0N) shows that RAS forms a dimer through helices α4 and α5 (the α4‐α5 dimer), generated through a 2‐fold crystallographic symmetry axis (Fetics et al., 2015). This dimer appeared again in a structure of KRAS/RAF‐RBD‐CRD (PDB ID 6XI7) (Tran et al., 2021). The α4‐α5 dimer is also present in crystals of HRAS (PDB ID 3K8Y) where we first observed the connection between the allosteric and active sites (Buhrman et al., 2010). As the allosteric site is adjacent to the dimer interface, it is possible that the Ca²⁺ binding site is stabilized in the RAS dimer, providing increased affinity and specificity for Ca²⁺. This would explain the solution NMR experiments under conditions where RAS is entirely in its monomeric form (Kovrigina et al., 2015; O'Connor & Kovrigin, ²⁰¹²) and the lack of response to Ca²⁺ in terms of GTP hydrolysis rate constants in monomeric RAS. As the Ca²⁺ binding issue remains unresolved, site directed mutagenesis was used in the current study to systematically perturb the allosteric site and show that the proposed R‐state modulated by a shift in helix 3 toward helix 4 indeed correlates with hydrolysis of GTP on RAS.

The results presented here suggest two critical points that are generally relevant to small GTPases. First, more attention to allostery is needed to properly understand the intrinsic GTP hydrolysis reaction performed by different small GTPases. While we focused on HRAS here, the role of allostery in regulating the function of oncogenic mutants of NRAS, and even more so KRAS, are also needed. These two isoforms are well conserved between HRAS (95% sequence similarity), yet some amino acid substitutions exist near the allosteric site and network (Johnson et al., 2017). These isoforms differ subtly from HRAS in their dynamics and ability to promote GTP hydrolysis (Harrison et al., 2016; Johnson et al., ²⁰¹⁷; Killoran & Smith, ²⁰¹⁹; Parker et al., ²⁰¹⁸; Volmar et al., ²⁰²²). Whether the allosteric site plays a role in these functional differences, and how the allosteric site interacts with oncogenic mutations in the active site, are necessary next steps to better understand these critically important enzymes. Second, the study of small GTPase allostery will likely provide a more complete picture of their evolution as signaling proteins. RAS, as well as many other small GTPases, are unlike other well‐studied enzymes in that they have evolved to have poor catalytic efficiency on their own (Cherfils & Zeghouf, 2013), and thus could be greatly susceptible to allosteric modulation of GTP hydrolysis. For instance, despite structural similarities between ATPases and GTPases, such as the well conserved P‐loop, enzymes of these two families can show up to six‐orders of magnitude differences in their phosphoryl‐transfer capabilities (Wittinghofer, 2016; Wittinghofer & Vetter, ²⁰¹¹). While the experiments described here looked at mutations of a single residue, they demonstrate that the allosteric site could be an underappreciated region in this evolution.

While the overall changes in GTP hydrolysis rate constants are small, taken together our experiments show a functional connection between the allosteric site and the active site. However, the role of RAF in our hydrolysis experiments shows more complexity than implied in our previously published mechanism of intrinsic hydrolysis (Buhrman et al., 2010), where RAF primarily played the role of stabilizing switch I in a closed conformation (Figure 1). At that point we had not yet considered the effect of binding RAF on increased dynamics of switch II, where binding of RAF to RAS alters the balance of R‐ and T‐ conformational states toward the more disordered T‐state (Fetics et al., 2015). Thus, the increase in switch II dynamics seen for HRAS‐RAF complexation represents a mechanism to destabilize the R‐state (Fetics et al., 2015; Li et al., ²⁰²²), which potentially synergizes, to different degrees, with the allosteric site mutants in our current study. This synergy appears to be important, as the dynamic change in switch II has little effect on intrinsic GTP hydrolysis for wild type HRAS in complex with RAF‐RBD. Alternatively, in the presence of RAF‐RBD, the R‐state favoring allosteric site mutants may stabilize the complex in the ground state, thus impairing HRAS progress toward the transition state for GTP hydrolysis.

Interestingly, the work presented here may lend some insight to a cluster of poorly studied RAS amino acid substitutions (AAS) found in human tumors and rasopathies (Figure 6) (Nseuronet Database Web Site, n.d.; Cerami et al., ²⁰¹²; Forbes et al., ²⁰¹⁴; Gao et al., ²⁰¹³; Karczewski et al., ²⁰²⁰). From our hydrolysis experiments with RAF1‐RBD, we can infer that some of these AAS may subtly alter the amplitude and frequency of MAPK pathway signaling by inhibiting GTP hydrolysis and promoting the activated state of RAS in the presence of RAF kinase, thereby promoting cell growth. For instance, it is likely that RAF1 will cooperate with NRAS R97G and KRAS R97I in cells to decrease GTP hydrolysis (Figure 6). Some of the allosteric site AAS may destabilize the R‐state by impairing the cooperative interactions that occur between helix 3, loop7 and helix 4, or interrupting binding with a putative allosteric ligand, thereby increasing the overall activated state of RAS in the cell (orange residues, Figure 6). In the case of I93V, the valine substitution will essentially mimic the preferred rotamer of I93 in our R‐state favoring structures of HRAS, indicating that it could influence the T‐ to R‐state transition via destabilizing hydrophobic packing beneath the allosteric site. Other AAS found in this pocket may have similar effects (yellow residues, Figure 6). The majority of the AAS shown in Figure 6 appear in tumor samples or as pathological germline substitutions (red, purple and * AAS vs. orange AAV), which supports these AAS as promoting aberrant cell growth. Tumorigenesis, and the penetrance and severity of rasopathy phenotypes, require optimal levels of MAPK signaling pathway activation. Single activating mutations in KRAS or other genes that regulate ERK signaling (e.g., strongly activating mutations in cancer, weakly activating mutations in rasopathies) are the most commonly observed mechanisms of pathological ERK activation (Johnson et al., 2022; Li et al., ²⁰¹⁸). However, combinatorial mechanisms (e.g., co‐occurring BRAF and RAS mutations) to pathologically activate the MAPK signaling pathway have also been identified in human tumors, and experimentally verified (Andreadi et al., 2012; Heidorn et al., ²⁰¹⁰). Thus, mutations in the allosteric site may cooperate with other oncogenes that regulate MAPK signaling (e.g., EGFR, BRAF, CRAF) to drive pathological activation of the MAPK signaling pathway. While experimental studies will be necessary to confirm the cellular effects of these allosteric AAS in a disease context, dissection of the allosteric site will provide a better understanding of disease mutations that have been difficult to characterize (Johnson et al., 2022).

FIGURE 6.

Amino acid substitutions (AAS) in the allosteric pocket. Boxes represent AAS found at a given residue for each of the RAS isoforms. Residue colors are explained in the text. AAS present in tumors identified in either the AACR project GENIE or COSMIC databases are presented in red lettering (Cerami et al., 2012; Forbes et al., ²⁰¹⁴; Gao et al., ²⁰¹³). Germline AAS identified in either gNOMAD or NSEuroNET are represented with orange lettering (Nseuronet Database Web Site, n.d.; Karczewski et al., ²⁰²⁰). AAS that are present in both tumor and germline samples are shown with purple lettering, and AAS associated with rasopathies are identified with an asterisk.

Finally, this work revitalizes the possibility of utilizing intrinsic hydrolysis to attenuate or block the oncogenic activities of RAS. While the most common oncogenic mutations in KRAS target GAP‐mediated hydrolysis, many (if not most) oncogenic mutations retain some form of intrinsic hydrolysis (Johnson et al., 2022). Thus, targeting the allosteric site may be a potential means to activate intrinsic hydrolysis. Indeed, we previously showed that the allosteric site can be utilized to drive G12V mutants of HRAS into the R‐state when crystals of this mutant are soaked in Ca(OAc)₂ (Buhrman et al., 2011). More studies on the allosteric site are needed to fully understand cellular regulation of RAS and to leverage it against RAS driven cancers.

4. MATERIALS AND METHODS

4.1. Experimental design to explore allostery in Ras

The allosteric site consists of a ring of polar (Y137, H94) and charged residues (D107, K101, E98) surrounding R97, which is poised over a hydrophobic pocket formed by the sidechains of I93, C80, M111, L113, L133, Y137, I139 and the aliphatic carbon chain of R97 (Figures 1b and 6). A potential space for packing in the hydrophobic pocket was observed during initial analyses and R97 was mutated to hydrophobic residues. This could drive the R‐state transition by packing of the mutated side chain in the core, thereby shifting the helix 3 balance of conformational states toward helix 4. Moderate (i.e., R97V, R97L, R97I) and large (i.e., R97M, R97F) packing mutations were tested, as well as different side‐chain configurations (i.e., R97L vs. R97I, R97M vs. R97F). Moderately sized groups were expected to favor the R‐state and the larger hydrophobic groups to sterically prevent the motion of helix 3 toward helix 4, which is necessary for the R‐state. Increased flexibility (R97G) and loss of packing (R97A) mutations were also tested. The large hydrophobic mutants were expected to favor the T‐state due to steric interactions within the allosteric stie, and the R97A and R97G mutants were expected to favor the T‐state due to increased dynamics of both helix 3 and switch II.

4.2. Mutagenesis, protein purification and enzymology

All mutant protein for crystallization and hydrolysis used residues 1–166 comprising the catalytic G‐domain of HRAS (EC 3.6.5.2). Residue 97 mutations to glycine, alanine, valine, isoleucine, methionine and phenylalanine of HRAS were made using a two‐stage mutagenesis protocol (Wang & Malcolm, 2001), based on QuikChange parameters. The RAS proteins were purified as previously described (Johnson et al., 2016) and so was C‐RAF1‐RBD (RAF1‐RBD; EC 2.7.11.1) (Ting et al., 2015). In order to crystallize HRAS mutants in their active form (i.e., GTP bound), we exchanged GDP for a non‐hydrolyzable GTP‐analogue guanylyl‐5‐imidodiphosphate (GppNHp) using an established protocol (Johnson et al., 2016). Protein for hydrolysis was kept bound to GDP, and both the GDP and the GppNHp bound proteins, were transferred into stabilization buffer (20 mM HEPES, 50 mM NaCl, 20 mM MgCl₂, 1 mM DTT, pH 7.5). Immediately after buffer exchange, protein was flash frozen in liquid nitrogen, in 20–100 μl aliquots, and stored at −80°C. Protein purity was determined using SDS‐PAGE and concentration was determined to be 12–20 mg/ml using the Bradford assay (Bradford, 1976).

Single turnover hydrolysis reactions for the R97 mutants were performed as previously published (Kearney et al., 2014), with small modifications described here. All reactions were performed at 37°C. GDP‐bound RAS (5 μM) was preloaded with ³²P‐γ‐GTP (50 nM) from Perkin Elmer, for 5 min in exchange buffer (20 mM Tris, 1 mM EDTA, 2 mM DTT, pH 8.0) in a 100 μl reaction. Once nucleotide was exchanged, initiation of hydrolysis was done by adding 4 μl of nucleotide exchange reaction, to 16 μl of pre‐warmed hydrolysis buffer (20 mM Tris, 100 mM NaCl, 5 mM MgCl₂, 2 mM DTT, and pH 8.0), to produce a 5‐fold dilution of RAS and nucleotide concentration. Sixteen total reactions were performed for each single‐turnover experiment, corresponding to 16 time points at 0, 2, 4, 6, 8, 10, 12, 14, 16, 20, 25, 30, 40, 50, 60, 80 and 100 min. Amount of ³²P_i formed for each time point was determined by organic extraction (Bollag & McCormick, 1995; Shacter, ¹⁹⁸⁴), and detection of β‐emission was done using a HIDEX liquid scintillation counter. Recorded CPM and TDCR measurements were used to determine the fmols of ³²P_i formed during the reaction using a published procedure (Kearney et al., 2014). Reactions were then converted into concentration (nM) of ³²P_i before analysis.

4.3. Kinetic modeling

Dyanfit4 was used to determine the rate constants for intrinsic hydrolysis (k _hyd) for each of the HRAS R97 mutants (Kuzmic, 1996; Kuzmic, ²⁰⁰⁹), using protocols that we previously published in detail for RAS proteins (Johnson et al., 2017). For each reaction, two parameters were determined: k _hyd and the starting concentration of RAS‐GTP. Fitting of RAS‐GTP was performed to mitigate errors in the estimation of protein concentration, titration errors, and differences in GTP concentration due to radioactive decay. Since the association and dissociation rates for binding of RAF1‐RBD to HRAS are significantly faster than the intrinsic hydrolysis of RAS (Sydor et al., 1998), we assumed that the association and dissociation rates were also significantly higher for the HRAS mutants and so were not included in the parameter model. We were successful in using a first‐order kinetic mechanism for wild type HRAS and all the allosteric site mutants except for R97G in the presence of RAF1‐RBD, where the reaction was too slow to measure (Figure 5b). Rates and rate constants determined by Dynafit4 can be found in Tables S4 and S5.

4.4. Protein crystallization

All protein for these crystallization experiments were prepared in the same way as the protein used for our hydrolysis assays. Crystallization was performed using the sitting drop and vapor diffusion methods using various concentrations of Ca(OAc)₂ and PEG 3350 for each mutant around the original hit condition 28 from the Hampton Research PEG Ion Screen. We obtained crystals of HRAS R97G, R97I, R97L, R97M, and R97F mutants with symmetries of both R32 and P3₂21 space groups, while the R97A and R97V mutants were only crystallized with symmetry of the R32 space group. The crystals were grown in 2 μl by 2 μl protein to mother liquor drops. Prior to X‐ray diffraction and data collection, crystals were briefly soaked in mother liquor containing 30% glycerol for cryoprotection. X‐ray diffraction data for R97F crystallized with P3₂21 symmetry, and all of the HRAS R97G, R97A, R97V, R97I, R97L, and R97M bound to GppNHp data were collected on a home source instrument (Rigaku MicroMax 007 R‐AxisIV⁺⁺). Data for the R97F and R97A crystals with R32 symmetry were collected on the ID‐22 SER‐CAT beamline at the Advanced Photon Source (Argonne National Laboratory). Diffraction data were collected at 100 K for both the home and synchrotron X‐ray sources.

Indexing, integration, scaling and post‐refinement were performed on HKL3000 (Otwinowski & Minor, 1997), and data collection statistics can be found in Tables S1 and S3. Molecular replacement was performed using the default settings in the auto‐MR program in the PHENIX suite of programs (Adams et al., 2010); the PDB model 1CTQ was used as a phasing model for the P3₂21 crystals, and PDB model 3K8Y was used as a phasing model for the R32 crystals—ligands were not included in molecular replacement. After molecular replacement, starting models were edited for the correct mutation, and ligands (i.e., GppNHp‐Mg), were set to full occupancy. PHENIX was then used to perform a single round of simulated annealing to remove model bias. Further refinement and model building was performed by alternating use of PHENIX and the model‐building program COOT (Emsley & Cowtan, 2004). Refinement statistics can be found in Tables S1 and S3.

AUTHOR CONTRIBUTIONS

Carla Mattos conceived of the project. Susan K. Fetics crystallized and solved the structure of R97L in the P3₂21 and R32 space groups. Kathleen P. Davis crystallized the R97A mutant, in both R‐ and T‐state R32 space groups. Jose A. Rodrigues crystallized the R97G mutant in the R32 space group. Christian W. Johnson crystallized the remaining allosteric site mutants, designed experiments, performed all GTP hydrolysis experiments. Carla Mattos and Christian W. Johnson designed experiments and wrote the paper.

Supporting information

TABLE S1. Crystal statistics for R32 protein crystals

TABLE S2. Crystallization conditions for HRAS R97 mutants

TABLE S3. Crystal statistics for P3221 protein crystals

TABLE S4. Initial rate and rate constants for HRAS R97 mutants alone

TABLE S5. Initial rate and rate constants for HRAS R97 mutants with RAF1‐RBD

Description of supplementary material: 2 tables describing crystal statistics, 1 table describing crystallization conditions, and 2 tables describing rates and rate constants for GTP hydrolysis reactions by H‐Ras R97 mutants.

Johnson CW, Fetics SK, Davis KP, Rodrigues JA, Mattos C. Allosteric site variants affect GTP hydrolysis on Ras. Protein Science. 2023;32(10):e4767. 10.1002/pro.4767