Allosteric KRas4B Can Modulate SOS1 Fast and Slow Ras Activation Cycles

Abstract

Membrane-anchored Ras family proteins are activated by guanine nucleotide exchange factors such as SOS1. The CDC25 domain of SOS1 catalyzes GDP-to-GTP exchange, thereby activating Ras. Here, we aim to decipher the activation mechanism of KRas4B, a significantly mutated oncogene. We perform large-scale molecular dynamics simulations on 12 SOS1 systems, scrutinizing each step in two possible KRas4B activation cycles, fast and slow. To activate KRas4B at the CDC25 catalytic site, the allosteric site in the Ras exchanger motif (REM) domain of SOS1 needs to recruit a (nucleotide-bound) KRas4B molecule. Our simulations indicate that KRas4B-GTP interacts with the REM allosteric site more strongly than with the CDC25 catalytic site, consistent with its allosteric role in the GDP-to-GTP exchange. In the fast cycle, the allosteric KRas4B-GTP induces conformational change at the catalytic site. The conformational change facilitates loading KRas4B-GDP at the catalytic site and opening the KRas4B nucleotide-binding site for GDP release and GTP binding. GTP binding reduces the affinity of KRas4B-GTP to the CDC25 catalytic site, resulting in its release. By contrast, in the slow cycle, KRas4B-GDP binds at the allosteric REM site. The limited, altered conformational change that it induces prevents the exact alignments of switch I and II of KRas4B. The increasing binding strength at both binding sites due to interactions of regions other than switch I and II retards GDP release from the catalytic KRas4B, thus KRas4B activation. The accelerated activation cycle supports a positive feedback loop with allosteric signals communicating between the two Ras molecules and is the predominant, native function of SOS. SOS1 activation details may help drug discovery to inhibit Ras activation.

Introduction

Small GTPase Ras proteins anchor to the plasma membrane through their C-terminal hypervariable regions (HVRs) (1). Active GTP-bound Ras binds and activates downstream effectors such as Raf kinase, phosphatidylinositide 3-kinase, NORE1A (RASSF5) (2), and Ral guanine nucleotide dissociation stimulator (RalGDS). Among the three HRas, NRas, and KRas isoforms, KRas is the most frequently mutated in RAS-driven cancers (3, 4). The wild-type KRAS gene has two splicing protein isoform products, KRas4A and KRas4B; thus, mutations in KRas will be observed in both KRas4A and KRas4B. Mutant KRas4B is the most abundant in RAS-driven cancers. KRas isoforms have 98% sequence identity at the G-domain (residue 1–166) but differ in their C-terminal HVR (residue 167–189). The HVRs of both KRas4A and KRas4B are highly positively charged and avidly interact with the anionic membrane; however, that of KRas4B is more so than KRas4A’s. KRas4A has two states: in state 1, the HVR is only farnesylated, but in state 2, it is also palmitoylated. The KRas4B HVR is only farnesylated. Anchored, prenylated HVR mediates the assembly of active Ras molecules into nanoclusters (5, 6), promoting signaling through the mitogen-activated protein kinase (MAPK, Raf/MEK/ERK) pathway and leading to cell proliferation (7, 8). Son of sevenless 1 (SOS1) is a guanine nucleotide exchange factor (GEF) that specifically activates Ras proteins by exchange of GDP to GTP (9, 10). SOS1 is a large multidomain protein with atomic mass of ∼153 kDa (11, 12). It consists of the membrane-interacting N-terminal and Ras-activating C-terminal regions. The N-terminal region contains histone-like fold (residues 1–198), Dbl-homology (DH, residues 200–390), and pleckstrin-homology (PH, residues 444–548) domains. The C-terminal catalytic region is composed of Ras exchanger motif (REM, residues 567–741) and CDC25 (residues 780–1019) domains, followed by a C-terminal SH3 binding motif tail (residues 1020–1333). The N-terminal regulatory domains act in SOS1 recruitment to the plasma membrane (13), which is mediated by an epidermal growth factor (EGF) receptor phosphorylated tyrosine motif (14). The C-terminal catalytic region contains two Ras binding sites, one located at the REM and the other at the CDC25 domain, corresponding to the allosteric and catalytic Ras binding sites, respectively (Fig. 1).

Figure 1.

SOS1 sequence and structure. (A) The amino acid sequence of the C-terminal catalytic region of SOS1 is shown. In the sequence, hydrophobic, polar/glycine, positively charged, and negatively charged residues are colored black, green, blue, and red, respectively. The cyan and blue underlines highlight the REM (residues 567–741) and CDC25 (residues 780–1019) domains of SOS1. (B) A crystal structure of the C-terminal catalytic region of SOS1 (PDB: 4NYJ) is shown. The REM and CDC25 domains are colored cyan and blue, respectively. To see this figure in color, go online.

The solved Ras-SOS1-Ras ternary complex (15, 16) revealed that the switch I (residues 30–38) and II (residues 60–76) regions of the two Ras proteins are involved in both allosteric and catalytic sites’ SOS1 binding. Based on the ternary complex, the mechanism of Ras activation postulated that Ras binding to the REM allosteric site is crucial for activating GDP-bound Ras at the CDC25 catalytic site (hereafter, Ras interacting with SOS at the allosteric and catalytic sites refers to as allosteric and catalytic Ras, respectively). GTP-bound Ras at the allosteric site is known to facilitate GDP → GTP exchange at the catalytic site (17, 18, 19). The allosteric Ras-GTP elicits conformational changes of the tandem REM-CDC25 domains of SOS1, leading to displacement of the helix-hairpin motif formed by αF-βA-βB-αG at the CDC25 catalytic site (16, 20, 21) (domain structures shown in Fig. 1). The large movement of the helix-hairpin motif causes the αF helix to sterically interfere in the switch I region of Ras-GDP at the CDC25 catalytic site, exposing the nucleotide-binding site of Ras. The switch I open conformation of Ras-GDP facilitates GDP’s exit from the nucleotide-binding site, thus permitting loading cytosolic GTP. The crystal structures of the ternary complex depict nucleotide-free Ras at the CDC25 catalytic site with widely opened switch I, implicating a snapshot conformation caught in the midst of an exchange event. Fully active SOS1 promotes heteronucleotide exchanges rather than homonucleotide exchanges (19, 22) because the cytosolic cellular concentration of GTP is 10 times higher than GDP (23). Upon exchange by GTP, switch I shifts to the closed state, completing Ras activation. The weakened interaction with SOS of the activated Ras culminates in Ras release. Thus, accelerated Ras activation requires Ras-GTP at the allosteric SOS1 site. This is a fast activation cycle via a positive feedback loop (18, 19).

SOS1 was crystallized in complex with HRas (15, 16). Structural data relating to the highly oncogenic KRas are currently unavailable, and details of SOS1 conformational changes in KRas activation are missing. Here, we studied the wild-type KRas4B interacting with SOS1. Because the HVR is long, disordered, and not involved in the Ras-SOS interactions, it is not considered in our work. Because the G-domains of the KRas isoforms are almost identical, our observations can be applicable to both KRas4A and KRas4B. Using all-atom molecular dynamics (MD) simulations, we examine the mechanism of KRas4B activation to clarify how the allosteric site in the REM domain selectively accommodates GDP- and GTP-bound KRas4B. For comprehensive analysis, we constructed SOS1 systems with conformational ensembles representing the sequence of steps in Ras activation cycles. We investigated SOS1 systems, including the SOS1-KRas4B dimers and the KRas4B-SOS1-KRas4B ternary complexes modeled with GDP- and GTP-bound and nucleotide-free KRas4B. Using complementary techniques of conformational analysis, free-energy calculations, and allosteric pathway elucidation, we show that allosteric KRas4B-GTP supports the fast cycle with positive feedback activation of SOS1, which mediates accelerated activation. By contrast, allosteric KRas4B-GDP impedes SOS1 in Ras activation and regulates a limited, slow activation cycle. The GTP-bound Ras at the REM allosteric site promotes allosteric signals that propagate through the REM-CDC25 tandem domains, shifting the SOS1 landscape, with the resulting conformational changes stimulating KRas4B activation. Clarifying how the KRas4B-SOS-KRas4B ternary complex influences SOS1 catalytic action may help the development of small-molecule drugs to inhibit allosteric activation (15, 24, 25, 26, 27, 28).

Materials and Methods

Generating initial configurations of Ras-SOS complex

To construct initial configurations, we obtained the crystal structures of SOS1 (PDB: 4NYJ) (15), GDP-bound KRas4B^C118S (PDB: 4EPT) (24), and GTP-bound KRas4B^Q61H (PDB: 3GFT) from the Protein Data Bank (PDB). In 4NYJ, both Ras binding sites (allosteric and catalytic) in SOS1 are occupied by HRas. At the allosteric site, the Ras protein is a GNP-bound HRas, and it is a nucleotide-free HRas at the SOS1 catalytic site. After replacing the mutants with the wild-type residues, KRas4B was superimposed onto HRas in complex with SOS1, generating the coordinates for the KRas4B-SOS1 complex. At the catalytic site of SOS1, the switch I region of both GDP- and GTP-bound KRas4B clashed with the αF helix because of the switch I open conformation of the nucleotide-free Ras. To avoid this steric clash, we re-modeled the switch I region using the Modeler server (29), ensuring that the modeling of switch I does not affect the positions of GDP, GTP, and Mg²⁺. A total of 12 SOS1 systems were constructed: an apo-SOS1 monomer, SOS1⁰⁰; four dimeric systems, SOS1^D0, SOS1^T0, SOS1^0D, and SOS1^0T; four ternary systems, SOS1^DD, SOS1^DT, SOS1^TD, and SOS1^TT; and four ternary systems in the exchange event, SOS1^TD∗, SOS1^TR, and SOS1^TT∗ (details of the notations are given in the Results).

Atomistic MD simulations

The initial configurations were subject to MD simulations in an aqueous environment. The modified TIP3P water model (30) was used to create the isomeric unit cell box containing the Ras-SOS complex. The initial systems were neutralized by adding counterions and soaked additional Na⁺ and Cl⁻ to satisfy a total ion concentration near 100 mM. The updated CHARMM all-atom additive force field (31) was used to construct the set of starting points and to relax the systems to a production-ready stage, closely following the same protocol as in our previous works (32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42).

A series of minimization steps using steepest decent and the adopted-basis Newton-Raphson methods and dynamic relaxation cycles were performed for the solvent around the harmonically restrained complex. At the final pre-equilibrium stage, the SOS1 systems were gradually relaxed by removing the harmonic restraints through dynamic cycles with the full particle mesh Ewald electrostatics calculation. In the production runs, the constant temperature at 310 K was maintained by the Langevin temperature control, and the pressure at 1 atm was sustained by the Nosé-Hoover Langevin piston pressure control. A 400 ns production run was performed using the NAMD parallel computing code (43) on a Biowulf cluster at the National Institute of Health (Bethesda, MD). Analysis was performed with the CHARMM programming package (44).

Binding free energy calculation

The binding free energy was estimated by the combination of molecular mechanics combined with the generalized Born (GB) and surface area continuum solvation (MMGBSA). In the calculation, we closely followed the protocol reported in our previous studies (37, 39, 41, 42). The average binding free energy was calculated as a sum of the gas phase contribution, the solvation energy contribution, and the entropic contribution,

$$\langle \mathit{\Delta }{G}_b\rangle =\langle \mathit{\Delta }{G}_{gas}\rangle +\langle \mathit{\Delta }{G}_{sol}\rangle -T\mathit{\Delta }S,$$ (1)

where $\langle \rangle$ denotes an average along the MD trajectory. The gas phase contribution to the binding free energy is a sum of the internal energy, the vdW interaction, and the electrostatic energy,

$$\mathit{\Delta }{G}_{gas}=\mathit{\Delta }{H}_{intra}+\mathit{\Delta }{H}_{vdW}+\mathit{\Delta }{H}_{elec}.$$ (2)

The solvation contribution is a sum of the electrostatic and nonpolar contribution,

$$\mathit{\Delta }{G}_{sol}=\mathit{\Delta }{G}_{sol}^{elec}+\mathit{\Delta }{G}_{sol}^{nonpolar}.$$ (3)

The solvation free energy was obtained from the GB calculation using the GBSW module (45) of the CHARMM program (44). The entropy term can be divided into the translational, rotational, and vibrational contribution,

$$T\mathit{\Delta }S=T\mathit{\Delta }{S}_{trans}+T\mathit{\Delta }{S}_{rot}+T\mathit{\Delta }{S}_{vib}.$$ (4)

The translational and rotational contributions were obtained from the calculation of principal moment of inertia. The vibrational entropy was obtained from the quasiharmonic mode calculation in the VIBRAN module of the CHARMM program (44). The number of vectors (mode) to calculate the vibrational analysis was set to NATOM × 3, where NATOM denotes the number of atoms. Finally, the change in binding free energy due to the complex formation was calculated using the equation,

$$\mathit{\Delta }{G}_b=G_b^{complex}-\left(G_b^{SOS1}+G_b^{Ras}\right).$$ (5)

Results

The C-terminal catalytic region of SOS1 exhibits conformational changes upon binding to KRas4B

To decipher how SOS1 conformations with different Ras-binding modes delineate the activation cycle, we performed all-atom MD simulations on the SOS1 N-terminal-truncated, C-terminal-catalytic region (hereafter referred to as SOS1) in complex with KRas4B in solution. The simulations were performed for SOS1 systems including a Ras-free apo-SOS1 monomer, four KRas4B-SOS1 dimers, and seven KRas4B-SOS1-KRas4B ternary complexes. To abbreviate the system notations throughout the text, we introduce two superscripted letters on SOS1. For instance, SOS1^TD denotes GTP- and GDP-bound KRas4B interacting with SOS1 at the REM allosteric and CDC25 catalytic sites, respectively. Here, “T” and “D” denote GTP and GDP, with the former and latter superscripts corresponding to Ras-interacting sites at REM and CDC25, respectively. Thus, for an apo-SOS1 monomer, we have SOS1⁰⁰, where “0” denotes Ras free. For the dimeric systems, we have SOS1^D0, SOS1^T0, SOS1^0D, and SOS1^0T. For the ternary systems, we have SOS1^DD, SOS1^DT, SOS1^TD, and SOS1^TT. To express ternary systems in the exchange event, we introduce SOS1^TD∗, SOS1^TR, and SOS1^TT∗, where “R” denotes nucleotide-free Ras and “^∗” highlights Ras with the switch I open conformation as in the crystal structure. A total of 12 SOS1 systems were constructed (Fig. S1). During the simulations, we observed that SOS1 exhibits significant conformational adjustments compared to the initial structures (Fig. S1), especially the REM domain, depending on the KRas4B binding modes with respect to both allosteric and catalytic sites (Fig. S2). The interdomain interface between REM and CDC25 can be divided into two regions: region 1 is formed by α2/αG and β2/βB interfaces, and region 2 is formed by the α5/αA interface (Fig. 1). Hydrophobic contacts drive the tandem domain-domain interaction in region 1 and salt bridge/hydrogen-bond (H-bond)/polar interactions in region 2 (Table S1). All SOS1 systems yield generally similar hydrophobic interacting residue pairs at the interdomain interface (Fig. 2 A), whereas the residue pairs of the salt bridge/H-bond/polar interactions vary (Fig. 2 B), indicating that allosteric KRas4B binding mostly affects region 2 rather than region 1 of the interface.

Figure 2.

Domain-domain interactions of SOS1. The interacting residue pairs with high occurrence rate (>50%) at the REM-CDC25 domain-domain interface for (A) the hydrophobic and (B) salt bridge/H-bond/polar interactions are shown. The sticks with labels highlight the interacting residues. Table S1 summarizes a list of the residue pairs in the intramolecular interaction. To see this figure in color, go online.

SOS1 function relies on the conformational change of the CDC25 helix-hairpin motif, which is induced by REM movement (20, 46). To observe the REM movement, we compare the entire SOS1 conformation by superimposing the average structure of CDC25 of apo-SOS1 (Fig. S3), projecting the protein backbone trace onto the two-dimensional xy plane (Fig. 3 A). In the absence of REM-bound allosteric KRas4B, SOS1 barely activates the catalytic KRas4B at CDC25. SOS1 conformational ensembles for this rare activation cycle follow SOS1⁰⁰ → SOS1^0D → SOS1^0T. For SOS1^0D and SOS1^0T, the center of mass (COM) of REM is projected vertically below the reference position of Leu-670 (top row of Figs. 3 A and S3), which is opposite to SOS1 systems containing allosteric KRas4B. Although the REM structure in SOS1^0D deviates from that in SOS1⁰⁰, SOS1^0T presents tandem domain conformations as in SOS1⁰⁰. The SOS1 system with catalytic KRas4B-GDP represents a configuration in the beginning of the nucleotide exchange, and with catalytic KRas4B-GTP, it implicates the completion of nucleotide exchange. The exit of GTP-bound Ras from the catalytic site indicates that SOS1 conformation with catalytic Ras-GTP can be similar to that with a Ras-free catalytic site. For SOS1 systems with allosteric KRas4B-GDP, the COM of REM is projected slightly above the reference position because of REM dynamics (middle row of Figs. 3 A and S3). However, in the presence of allosteric KRas4B-GTP, with large conformational changes in SOS1, REM’s COM is projected high above the reference position (bottom row of Figs. 3 A and S3). To compare the region 2 interface between SOS1^TD and SOS1^DD (Fig. 2 B), the allosteric KRas4B-GTP breaks down more interacting residue pairs and weakens the interdomain interaction, allowing the flexible movement of REM. By contrast, the allosteric KRas4B-GDP strengthens the interdomain interaction at the region 2 interface, suggesting that the allosteric KRas4B-GDP induces insufficient conformational change in SOS1 to modulate the catalytic CDC25 conformation. We hypothesize that SOS1 may delay KRas4B activation in the limited activation cycle through SOS1^D0 → SOS1^DD → SOS1^DT. By contrast, with allosteric KRas4B-GTP, the catalytic activity of SOS1 can be enhanced, leading to an accelerated activation cycle through SOS1^T0 → SOS1^TD → SOS1^TT. To describe sequentially the event of KRas4B activation in the accelerated activation cycle, the allosteric KRas4B-GTP in SOS1^T0 can induce large conformational change in the REM-CDC25 tandem domains, enabling the CDC25 catalytic site to recruit KRas4B-GDP. This is in line with the nucleotide exchange event for conversion of the catalytic Ras to the GTP-bound state beginning with SOS1^TD and completing with SOS1^TT. These represent the SOS1 systems before and after the nucleotide exchange event. The similarity of the least inclination angle between REM and CDC25 suggests that SOS1^TT resembles the conformation in SOS1⁰⁰ (Fig. 3 A). This indicates that the “backward” conformation change in SOS1 facilitates GTP-bound Ras exit from the catalytic site. With allosteric KRas4B-GDP, the hydrophilic interaction at the tandem domains interface increases, preventing REM’s dynamics, whereas the hydrophilic interaction at the interface with the allosteric KRas4B-GTP decreases, allowing the movement of REM (Fig. 2 B). To quantitate the conformational change in the tandem domains, we projected the protein backbone trace onto the yz plane (middle and right panels of Fig. 3 B) and calculated the interaction energy between REM and CDC25 domains as a function of the orientation angle between the two Ras proteins (left panel of Fig. 3 B). In the projection, the relative orientation of both Ras proteins due to the movement of REM is clearly visible. With allosteric KRas4B-GDP, the interaction between REM and CDC25 is relatively stronger (Fig. S4), and the orientation angle between the two Ras proteins is relatively smaller than for the corresponding systems with allosteric KRas4B-GTP. The strong domain-domain interaction between REM and CDC25 may impede SOS1’s catalytic activity.

Figure 3.

Conformational analysis. (A) The projection of protein backbone traces onto the two-dimensional xy plane for average structures of different SOS1 systems is shown. In each panel, the purple dots from the left-hand side correspond to the projected positions of the COM for the residues Leu-670, REM, and CDC25, respectively. All SOS1 systems were oriented with respect to CDC25 of SOS1⁰⁰ as a reference, with y = 0 for all dots. The arrows indicate directions of the relative movement of the COM along the y axis. The solid and dotted arrows denote large and small movements, respectively. (B) The interaction energy between REM and CDC25 domains correlates to the allosteric and catalytic KRas4B orientation (left panel). The least active SOS1 (SOS1^DD) has the strongest domain-domain interaction, and the two KRas4B are closely bound. By contrast, the most active SOS1^TT relaxes REM and CDC25 as well as KRas4B. Therefore, the decrease of REM-CDC25 interaction promotes KRas4B activation. The rotation of the projected protein backbone trace into the YZ plane for the SOS1 systems on the limited (SOS1^DD → SOS1^DT) and accelerated (SOS1^TD → SOS1^TT) activation cycles (middle and right panels) is shown. In the trace, the REM and CDC25 domains are colored cyan and blue, respectively. GTP- and GDP-bound KRas4B molecules are colored green and red. To see this figure in color, go online.

Competing interactions of KRas4B at the SOS1 binding sites

The interaction of KRas4B with SOS1 mainly involves the switch I (residues 30–38) and II (60–76) regions. The SOS1-KRas4B interacting residue pairs depend on the KRas4B states at the SOS1 binding sites (Table S2). For example, comparison of the catalytic KRas4B in SOS1^TD and SOS1^TT points to a similar number of polar interactions, whereas the hydrophobic interaction and salt bridges vary. We observed that there are more hydrophobic interactions between CDC25 and switch I at the catalytic site of SOS1^TT than SOS1^TD. Also, more switch I and II regions’ residues of the allosteric KRas4B interact with SOS1 compared to catalytic KRas4B (Fig. S5). To quantify their contributions to the interaction, we calculated the binding free energies of the complex using molecular mechanics combined with the GB and MMGBSA. The generally lower binding-free-energy values indicate that the allosteric KRas4B binds SOS1 more strongly than the catalytic KRas4B (Fig. 4). The relative changes in the binding free energies among different SOS1 systems suggest that allosteric KRas4B enables recognition of other Ras binding at the catalytic site. At the catalytic site, KRas4B-GDP with low values of the binding free energy interacts with SOS1 more strongly than KRas4B-GTP, reflecting the status of the SOS1 conformations before and after the nucleotide exchange. For example, in SOS1^TD, the strong interaction of the catalytic KRas4B-GDP with SOS1 promotes nucleotide exchange, whereas in SOS1^TT, the weak and unstable interactions of catalytic KRas4B-GTP with SOS1 cause the catalytic Ras to be released. The highest average binding free energy of catalytic KRas4B-GTP in in SOS1^TT shows the weakest interaction, and its wide standard deviation represents the instability. The relative changes in the binding free energies for catalytic KRas4B among the different possible activation cycles also suggest an allosteric communication between the two Ras binding sites. When SOS1 is recruited to the plasma membrane, the allosteric site provides higher binding affinity for KRas4B than the catalytic site does, suggesting that KRas4B first binds to the allosteric site. At the allosteric site, KRas4B-GTP with high binding affinity easily competes with KRas4B-GDP, as observed in the binding free energies for SOS1^T0 and SOS1^D0. When SOS1 first recruits allosteric KRas4B-GTP, it enters the accelerated activation cycle with a positive feedback loop (18, 19). However, in the case of SOS1 with allosteric KRas4B-GDP, the induced catalytic SOS1 conformation prevents exact alignments of switch I and II of KRas4B in the interaction. As a result, the increasing binding strength at both binding sites due to interactions of KRas4B regions that are not involved with switch I and II retard the SOS1 system activation cycle.

Figure 4.

Relative changes in binding free energy. The binding free energy of KRas4B interacting with SOS1 at the (A) allosteric and (B) catalytic sites is shown. The binding free energy was calculated using Eq. 1. The MMGBSA method was used to calculate the gas and solvation contributions, and the quasiharmonic mode analysis was used to calculate the entropic contribution. The mean and median values are denoted by red squares and red horizontal lines, respectively. The blue boxes represent the quartiles, and the vertical dashed lines are determined by the 95% confidence intervals. To see this figure in color, go online.

Conformational changes of KRas4B at the SOS1 catalytic site during activation

During the nucleotide exchange, catalytic KRas4B dramatically changes its conformation as the exchange event progresses. To reflect the conformational change of the catalytic KRas4B during the exchange, we refined the SOS1 systems in the accelerated activation cycle to include the exchange event SOS1^T0 → SOS1^TD → SOS1^TD∗ → SOS1^TR → SOS1^TT∗ → SOS1^TT. Based on the crystal structures of the Ras-SOS1-Ras ternary complex with the switch I open conformation of the nucleotide-free Ras at the catalytic site (15, 16), we can propose sequential changes of the catalytic KRas4B conformation during the exchange: 1) at the beginning of the nucleotide exchange, SOS1 “begins to pull” the switch I loop of the catalytic KRas4B-GDP (SOS1^TD); 2) this exposes the nucleotide binding pocket of KRas4B-GDP with switch I in an open conformation (SOS1^TD∗); 3) GDP now escapes, and the catalytic Ras becomes nucleotide free (SOS1^TR); 4) GTP fills in the void in the binding pocket, and the catalytic Ras turns into a GTP-bound Ras with switch I in an open conformation (SOS1^TT∗); 5) finally, the catalytic Ras closes the switch I loop and prepares to leave SOS1 catalytic site (SOS1^TT). To grade the conformational ensembles of catalytic KRas4B during the nucleotide exchange, we calculated the binding free energy of the catalytic KRas4B with SOS1 (Fig. 5 A, left panel). We observed that catalytic KRas4B with switch I in the open conformation (SOS1^TD∗, SOS1^TR, and SOS1^TT∗) binds SOS1 more strongly than with switch I in the closed conformation (SOS1^TD and SOS1^TT). At the catalytic site, the nucleotide-free Ras has a high binding affinity with the dissociation constant, K_D, 1000 times smaller than the nucleotide-loaded Ras (16, 19, 47). By contrast, catalytic KRas4B with switch I in the closed conformation binds the nucleotides GDP and GTP more strongly than with the switch I open conformation. Nucleotide binding to catalytic KRas4B is strongly correlated to the catalytic KRas4B binding to SOS1, with a correlation coefficient of r = −0.98 (Fig. 5 A, right panel), suggesting that the increasing binding strength of KRas4B to SOS1 compensates for the decrease in nucleotide interaction with the Ras nucleotide-binding site. Strong binding of GTP to Ras at the catalytic site suggests that GTP easily competes with GDP when the Ras nucleotide-binding site is exposed. To activate Ras, SOS1 uses its αF helix to intrude into the switch I loop of KRas4B-GDP generating the open conformation (Fig. 5 B). For KRas4B in the switch I closed state (SOS1^TD and SOS1^TT), relatively large fluctuations in the switch II region can be observed (Fig. S6). However, for KRas4B in the switch I open state, the fluctuations in the switch II region disappear, and instead, large fluctuations in the G2 region (residues 26–37) containing switch I can be observed (SOS1^TD∗, SOS1^TR, and SOS1^TT∗). Essentially, the switch II of KRas4B acts as a pivot when SOS1 opens switch I (16, 18, 47, 48); however, for KRas4B in the nucleotide-free state (SOS1^TR), the fluctuations in the G4 region (residues 117–126) are due to the lack of stability that is established by the guanine group of the nucleotide.

Figure 5.

Binding free energy and conformation of catalytic KRas4B in the sequence of event. (A) The binding free energy of KRas4B at the SOS1 catalytic site during the nucleotide exchange for the systems on the accelerated activation cycle is shown (left panel). The correlation of the binding free energy between the catalytic KRas4B-SOS1 and KRas4B-nucleotide interactions is shown (right panel). (B) Average structures of catalytic KRas4B with the CDC25 helix-hairpin motif depicting intervention of the motif in the switch I region during the exchange event for the SOS1 systems on the accelerated activation cycle are shown. To see this figure in color, go online.

Allosteric signaling from KRas4B at the allosteric site to that at the catalytic site in SOS1 activation

We hypothesized that KRas4B binding to the allosteric site may affect the conformation of the catalytic site allosterically, promoting the activation of KRas4B at the catalytic site. To test this hypothesis, the weighted implementation of suboptimal paths can identify the signal propagation pathways through the protein by calculating the correlated motion among residues (49). We calculated a number of optimal and suboptimal pathways between two selected residues on KRas4B, one at the allosteric site and the other at the catalytic site (Table S3), to obtain the best pathways through the REM-CDC25 tandem domains of SOS1. The allosteric signals arise from switch I and II of the allosteric KRas4B, propagate through the SOS1 domains, and terminate at switch I and II of the catalytic KRas4B. The switch I and II regions are involved in the binding interface when KRas4B interacts with SOS1. For the accelerated activation cycle (Fig. 6 A), the allosteric signals mostly propagate through the REM domain (via α1, α2, and β1). In CDC25, the most frequently occurring residues (Thr-940, Ser-959, and Lys-960) are in the helix-hairpin motif formed by αF-βA-βB-αG at the catalytic site. Allosteric dislocation of the helix-hairpin motif results in the αF helix pushing against the switch I region of the catalytic KRas4B, inducing the open conformation. For the limited activation cycle (Fig. 6 B), although SOS1^DD shows allosteric signals transmitting through the REM domain to the helix-hairpin motif, SOS1^DT shows allosteric signals passing through the αD and αE regions of CDC25, which are far from the helix-hairpin motif at the catalytic site.

Figure 6.

Allosteric pathway analysis. The allosteric pathways (yellow lines) between the allosteric and catalytic KRas4B propagating through the REM-CDC25 tandem domains for the SOS1 systems on the (A) accelerated (SOS1^TD → SOS1^TT) and (B) limited (SOS1^DD → SOS1^DT) activation cycles are shown. In tube representation of SOS1, the REM and CDC25 domains are colored cyan and blue, respectively. For KRas4B, the GTP- and GDP-bound molecules are colored green and red. The blue beads on the pathways represent high-occurrence residues (>50%) in the pathway calculation. Table S3 summarizes a list of the residues involved in the optimal pathways. To see this figure in color, go online.

Discussion

Here, we aimed to decipher the structural complexity of KRas4B activation by SOS1. To efficiently activate KRas4B at the CDC25 catalytic site, the allosteric site in the REM domain needs to recruit an active KRas4B (16, 47, 50). Our simulations illustrate that KRas4B-GTP binding to the REM allosteric site induces movement of the REM domain, translocating the CDC25 helix-hairpin motif (Fig. 3), which causes displacement of the switch I region of KRas4B-GDP at the catalytic site, yielding a switch I open state (16, 20, 21). Allosteric pathway analyses show that at the allosteric site, KRas4B-GTP generates signals propagating through SOS1’s residues in the helix-hairpin motif (Fig. 6). These allosteric signals trigger large conformational changes in the tandem domains that accelerate SOS1 activation with positive feedback through binding cycles of the allosteric active Ras (Fig. 7 A). By contrast, in the limited activation cycle, the allosteric KRas4B-GDP does not induce proper allosteric signals, impeding SOS1 activation of KRas4B. The binding free energies of the allosteric GDP- and GTP-bound KRas4B are relatively similar. However, the binding free energies of the catalytic KRas4B vary based on the types of nucleotide and the allosteric effects. Due to the allosteric signals, the catalytic KRas4B-GTP becomes less stable when the allosteric KRas4B is free, GDP bound, and GTP bound. For all three different activation cycles, the catalytic KRas4B-GDP is more stable than the catalytic KRas4B-GTP. The relative change in the binding free energy of the catalytic KRas4B between SOS1^TD → SOS1^TT is the largest, followed by SOS1^DD → SOS1^DT and SOS1^0D → SOS1^0T.

Figure 7.

Schematic diagrams of SOS1 activation. (A) The KRas4B activation cycles by SOS1 are shown. The dotted lines denote the rare activation cycle, SOS1^0D → SOS1^0T, in the absence of allosteric Ras. The thin solid lines represent the limited activation, SOS1^D0 → SOS1^DD → SOS1^DT. The thick solid lines denote the accelerated activation cycle, SOS1^T0 → SOS1^TD → SOS1^TT, with high level of Ras activation. (B) A schematic diagram illustrates SOS1 activity at the plasma membrane. SOS1 activates KRas4B, and active KRas4B leads to cell proliferation regulating through the MAPK pathway. To see this figure in color, go online.

Our simulations elucidate the mechanism of KRas4B activation by SOS1 in structural detail. They indicate that KRas4B interacts with the REM allosteric site more strongly than with the CDC25 catalytic site, suggesting that SOS1 first recruits allosteric Ras, then the catalytic Ras. The binding free energy confirms that apo-SOS1 first loads an active KRas4B-GTP at the allosteric site, which primes SOS1 for an inactive KRas4B-GDP at the catalytic site (Fig. 4). The interaction of KRas4B-GTP with SOS1 at the allosteric site induces a local conformation change at the catalytic site, facilitating the accommodation of the inactive Ras. The conformational change further involves opening of the nucleotide-binding site of GDP-bound Ras. During this feedback process, the interaction between SOS1 and the catalytic Ras gets tighter, whereas the interaction between Ras and the nucleotide gets weaker (Fig. 5). The nucleotide-free Ras shows the strongest binding strength with the SOS1 catalytic site, indicating a widely opened conformation of the nucleotide-binding site. GTP binds Ras more strongly than GDP and thus is easily loaded to the nucleotide-free Ras. GTP-loaded Ras closes its nucleotide-binding site, becoming the active form. SOS1 releases the active Ras-GTP because of weak interaction at the catalytic site, resetting for the next activation cycle.

The CDC25 domain is found in the family of GEFs for Ras-like small GTPases. Proteins containing this domain include Kndc1, RalGDS, RapGEF1-6, and SOS1-2, which have both REM and CDC25 domains. Our work, which shows how multiple Ras GTPases communicate with each other via forming a complex with GEF’s tandem domain, can be applied to most Ras-like small GTPases interacting with GEF. For example, the basic subunits of GEF for the Rho family proteins (RhoGEF) are DH and PH domains: the DH domain catalyzes Rho, Rac, and Cdc42 by exchanging GDP to GTP, and the PH domain mainly interacts with the plasma membrane. However, one of the RhoGEF members, Dbs protein, forms a dimer through its DH-PH interface (51). Two catalytic Cdc42 may also transmit allosteric signals via a Dbs dimer. Consequently, the PH domain participates in the DH-Cdc42 association and may induce conformational changes promoting the nucleotide exchanges. Even though there is no allosteric Cdc42 binding to Dbs, this is a similar case—but more complex than our model—in which the PH domain may play a role similar to the REM domain in RasGEF.

Oncogenic KRas4B mutants in the GTP-bound state (KRas4B^mut-GTP) present a conformation with tightly closed switch I and II regions, whereas the GDP-bound mutants have large fluctuations in both switch regions, showing more widely open nucleotide-binding sites than wild-type KRas4B-GDP (34, 35). With distinct switch I and II conformations, mutant KRas4B interacting with SOS1 can shift the equilibrium of Ras activation. Because oncogenic KRas4B mutants block GTP hydrolysis by GTPase-activating protein (GAP), there is a large population of active KRas4B^mut-GTP. The highly populated, active KRas4B mutant can occupy the allosteric site of SOS1, sending the activation signal to the catalytic site. SOS1 catalytic site does not favor the tightly closed switch I and II conformation of active KRas4B^mut-GTP. Instead, inactive KRas4B^mut-GDP conformers with preopened switch I and II conformation can easily select and associate with the SOS1 catalytic site. Mutant KRas4B highly accelerates the action of SOS1 in the Ras activation cycle, whereas GTP hydrolysis by GAP to produce the inactive GDP-bound form is abolished or extremely slow. The imbalanced Ras activation/deactivation cycle due to the extremely fast activation by SOS1 and extremely slow or halted deactivation by GAP culminates in a highly populated active mutant KRas4B-GTP state in cancer.

The major Ras activation cycle is fast and is the predominant, native function of SOS. To regulate the MAPK pathway (Fig. 7 B), EGF first binds to the EGF receptor promoting phosphorylation. The phosphorylated EGF receptor recognizes and interacts with the SH2 domain of growth-factor-receptor-bound protein-2 (GRB2). Subsequently, two SH3 domains of GRB2 bind the proline-rich SOS1 C-terminal tail (52, 53, 54, 55), recruiting SOS1 to the cell membrane. The PH domain in the N-terminal region of SOS1 is responsible for the membrane anchorage, and the activity of the SOS1 catalytic unit is enhanced when interacting with membrane-anchored Ras (56). The REM allosteric site of SOS1 first binds membrane-anchored GTP-bound KRas4B with lower binding free energy than the CDC25 site binds the GDP-bound Ras. It was suggested that Ras-GTP binds the allosteric site 10-fold tighter than Ras-GDP (47). We also observe tighter binding. The binding of KRas4B-GTP to the allosteric site induces the dynamics of REM, with allosterically promoted conformational changes displacing the CDC25 helix-hairpin motif, resulting in the switch I open conformation of KRas4B-GDP at the SOS1 catalytic site. A grasp of the allosteric motion of REM-CDC25 tandem domains and identification of allosteric pathways through the SOS1 domains are crucial for therapeutics to control Ras activation and signaling in cancer. Our comprehensive simulations clarify SOS1 allosterically linked conformational change, delineate the key steps in KRas4B activation cycle, and provide detailed allosteric signal propagations upon associating with different forms of KRas4B.

Author Contributions

T.L., H.J., and R.N. conceived and designed the study. T.L. performed MD simulations. T.L. and H.J. analyzed the data. T.L., H.J., and R.N. prepared and wrote the manuscript. All authors edited and approved the manuscript.

Editor: David Sept.