Normal Mode Analysis of KRas4B Reveals Partner Specific Dynamics

Abstract

Ras GTPase interacts with its regulators and downstream effectors for its critical function in cellular signaling. Targeting the disrupted mechanisms in Ras-related human cancers requires understanding the distinct dynamics of these protein–protein interactions. We performed normal mode analysis (NMA) of KRas4B in wild-type or mutant monomeric and neurofibromin-1 (NF1), Son of Sevenless 1 (SOS1) or Raf-1 bound dimeric conformational states to reveal partner-specific dynamics of the protein. Gaussian network model (GNM) analysis showed that the known KRas4B lobes further partition into subdomains upon binding to its partners. Furthermore, KRas4B interactions with different partners suppress the flexibility in not only their binding sites but also distant residues in the allosteric lobe in a partner-specific way. The conformational changes can be driven by intrinsic residue fluctuations of the open state KRas4B-GDP, as we illustrated with anisotropic network model (ANM) analysis. The allosteric paths connecting the nucleotide binding residues to the allosteric site at α3-L7 portray differences in the inactive and active states. These findings help in understanding the partner-specific KRas4B dynamics, which could be utilized for therapeutic targeting.

Graphical Abstract

INTRODUCTION

Ras proteins are a family of small guanosine triphosphatases (GTPases) and have a conserved catalytic domain (G-domain) for activity. They are active when GTP is bound and inactive when guanosine diphosphate (GDP) is bound.¹ The GTPase activating proteins (GAPs) speed the intrinsically slow GTP hydrolysis reaction and take Ras to its inactive state.² Guanosine exchange factors (GEFs) activate Ras proteins by catalyzing the removal of GDP from the protein’s nucleotide-binding site, resulting in GTP binding.² GTP-bound Ras activates various signaling pathways crucial to cell proliferation and survival by interacting with effectors like Raf, phosphoinositide 3-kinase (PI3K), and Ral guanine nucleotide dissociation stimulator (RalGDS) at their Ras binding domains (RBDs).³ Mutant Ras proteins with constitutively active state promote tumorigenesis by hyperactivation of the downstream signal pathways.^3,4 Human RAS isoforms are HRAS, KRAS, and NRAS. KRAS oncogenes make up about 82% of the Ras nonsynonymous mutations seen in human cancer according to data from COSMIC.⁵ The clinical importance of KRas protein makes the dominant splice variant KRas4B the focus of our study.

Figure 1A shows the KRas4B structure in active state together with its secondary structure elements. The G-domain (residues 1–166) is divided into two lobes based on function. The effector lobe (residues 1–86) is the N-terminal region of the catalytic domain: P-loop (residues 10–17) binds GTP/GDP with a Mg²⁺; Switch-I (residues 30–38) and Switch-II (residues 60–76) regions make up the main protein binding sites.^1,6 The allosteric lobe (residues 87–166) contains α-helical dimerization interface and nucleotide base binding motifs and relays information through the protein (about the interactions with the nucleotide, proteins, and membrane).^6,7 The hypervariable region (HVR, residues 167–188) is post-translationally modified and binds to the plasma membrane.^6,8

Figure 1.

KRas4B structure, sequence, and protein interactions. (A) Detailed structure of active state KRas4B (6GOD_A). The α-helices, β-strands, and loops are colored red, yellow, and green, respectively. KRas4B sequence (UniProt code P01116) and the corresponding secondary structures are shown in the lower panel. (B) Representative GEF, GAP, and effector interactions of KRas4B (red) with SOS1 (blue, 6EPL_RS), NF1 (orange, 6OB2_AB), and Raf-1 (pink, 6XI7_AB). KRas4B–SOS1 binding interface residue indices are 17, 21, 30–32, 34, 40, 54–57, 59–61, 63–71, 73, 102, and 105 according to HotRegion. NF1 binding interface residue indices are 21, 25, 31–34, 36–41, 61–64, 66, 67, and 88. Raf-1 binding interface residue indices are 21, 23–26, 29, 31, 33, 36–48, 56, 149, 153, and 157. Structure images were prepared using the PyMOL Molecular Graphics System, version 1.2r3pre, Schrödinger, LLC.

KRas4B can form heterodimers with its regulators such as GEFs and GAPs or with its effectors such as Raf and PI3K. Figure 1B highlights KRas4B in complex with Son of Sevenless 1 (SOS1, Ras GEF), neurofibromin 1 (NF1, Ras GAP), and Raf-1 specifically. These partner proteins bind through different but partially overlapping regions on KRas4B. Switch-I, β2-L3-β3 are on the KRas4B–Raf-1 interaction interface.⁹ Both Switch regions are on the KRas4B–SOS1 and KRas4B–NF1 interaction interfaces.^10,11 SOS1 inserts an α-helical wedge structure to the nucleotide-binding site of GDP-bound Ras, moves Switch-I away from GDP, and modifies Switch-II conformation to facilitate the nucleotide exit.¹⁰ NF1 does the opposite; it interacts with Switch regions to catalyze the GTP hydrolysis reaction.¹¹ Although GTP-bound active KRas4B can interact with GAPs and effectors; Switch-I residue positions and orientations selectively recruit them.^12,13

The unique structural and biophysical properties of different KRas4B conformations affect its dynamics. There are various studies on the GTP-bound and GDP-bound states of Ras proteins that analyze the nucleotide-switching and hydrolysis reactions.⁷ A shift in α3-L7 toward the allosteric lobe enables Switch-II ordering for GTP hydrolysis.¹³ The reaction occurs basically through proton shuffling from a catalytic water molecule to residues Q61 and Y32 via the nucleotide’s γ-phosphate and the bridging water molecule. This culminates in the remaining hydroxide ion’s attack on the γ-phosphate causing formation of GDP and an inorganic phosphate. This allosteric modulation of Switch-II enables GTP hydrolysis.¹³ GDP-bound inactive state is associated with higher Switch-region flexibilities, while GTP binding introduces new hydrogen bonds and increases stiffness of the structure.^14–16 T35 and G60 make hydrogen bonds with GTP’s γ-phosphate oxygen atoms, Y32 resides over the nucleotide with a “bridging” water molecule, and also S17 and T35 side chains coordinate Mg²⁺ with two water molecules and the nucleotide’s β- and γ-phosphates.^16,17 Hydrogen bonding network is altered after GTP hydrolysis: T35 and G60 no longer interact with the nucleotide, Y32 side chain orientation changes, and T35 does not coordinate Mg²⁺.^16,18 Therefore, Ras is overall more flexible in the GDP-bound “open” state whereas the GTP-bound state is associated with less flexibility, especially in P-loop, β4, β5, and β6.^14,15 Mutations can also lead to conformational and allosteric changes, affecting protein interactions.^11,19–23 Therefore, comprehending the characteristics of KRas4B dynamics in different conformations can shed light on its oncogenic activity in various signal pathways. Most Ras dynamics studies are performed by atomistic molecular dynamics (MD) simulations. However, MD simulations are computationally expensive to study the overall dynamics of KRas4B with its partners. Normal mode analysis (NMA) can help in overcoming this challenge as it is computationally efficient and provides the large-scale collective motions.²⁴ NMA was utilized in previous studies investigating the nucleotide-switch mechanism of Ras superfamily,¹⁵ the intrinsic dynamics of KRas4B,^14,15 and also effects of mutations on KRas4B dynamics.¹⁹ However, there is not a comprehensive study that shows the dynamic differences when KRas4B is bound to its regulators and effectors, i.e., KRas4B partner-specific dynamics. This could help with understanding the undruggable KRas4B dynamics toward drug design.

Gaussian network model (GNM)²⁵ and anisotropic network model (ANM)²⁶ are NMA-based methods and give accurate estimates on global protein motions at slow modes and local motions at fast modes.²⁷ Residues with high fluctuations in slow modes correspond to more flexible regions, while the residues with lower fluctuations form the hinge sites critical for the protein activity.^27,28 Peak residues in fast modes comprise “kinetically hot residues” constrained in the structure and with roles in folding or binding.^29,30 Moreover, changes in residue motions over all modes provide insight into protein’s function and allosteric regulation.^31–33 These network models can be further modified by incorporating residue specificity³⁴ or used in concert with other methods such as perturbation-response scanning.^35,36

In this study, we perform NMA of KRas4B complexed with SOS1, NF1, and Raf-1 to analyze its partner specific conformational dynamics. We use GNM to investigate the global and local KRas4B intrinsic residue fluctuations in the complexes. The modules of residues with correlated fluctuations may indicate functional subdomains in the structure. We employ ANM to highlight the transition from inactive “open” state of wild-type KRas4B to bound “closed” states of KRas4B with its partners, calculating theoretical residue displacements. We use a Monte Carlo path (MCPath) generation approach³⁷ and a computationally efficient network-based method, Ohm,³⁸ to find allosteric pathways that might differentiate between these conformations. As a summary, we aim to reveal how binding to different partners affects residue flexibilities, introduces new intramolecular residue communications, and forms functional subdomains in KRas4B.

METHODS

Protein Structures.

We selected seven KRas4B representative structures from Protein Data Bank (PDB) with high resolution and no missing sequences based on their mutation status, bound nucleotide type, and the interacting proteins (Table 1). KRas4B proteins in all the structures were modified to start and end at the same residue, from T2 to H166.

Table 1.

List of Analyzed KRas4B Structures and the Associated Data

PDB codea	type	resolution (Å)	binding status	main KRas chain	mutation status	nucleotideb
4OBE A	X-ray	1.24	unbound	A	WT	GDP
6GOD A	X-ray	1.71	unbound	A	WT	GNP
4EPR A	X-ray	2.00	unbound	A	G12D, C118S	GDP
6GOF B	X-ray	1.98	unbound	A	G12D	GNP
6EPL RS	X-ray	2.55	SOS1 (REM-CDC25) bound	R	G12C, C118S, DTK126ESR
6OB2 AB	X-ray	2.85	NF1 (GRD) bound	A	WT	GNP
6XI7 AB	X-ray	1.95	Raf-1 (RBD-CRD) bound	B	WT	GNP

^aSelected chain identifiers are provided with the PDB codes.

^bGNP (or GppNHp) is a nonhydrolyzable GTP analog.

Gaussian Network Model (GNM) with All Cα in the Structure and the Nucleotide Atoms.

Normal mode analysis was performed with ProDy.³⁹ The parameters we used for analysis with ProDy are provided in Supporting Information (Table S1). We applied GNM analysis for the seven KRas4B structures and considered the complex forms of the proteins (see Table 1 first column for chain identifiers) while also including the nucleotides in the structures. Besides all Cα atoms, the nucleotide atoms C1′, C4′, Pα, Pβ, and Pγ were used to define the contact maps. We set the cutoff distance to 7.5 Å and obtained N – 1 nonzero normal modes (N = Cα and nucleotide atom number). The squared fluctuations over the weighted 10 slowest modes and the individual three fastest modes were calculated for each structure to compare their global and local motions, respectively. The cross-correlations between residue fluctuations were determined from overall GNM modes. We prepared the graphics with Matplotlib⁴⁰ using the normalized results. The HotRegion⁴¹ results of the complexes were obtained to determine the regions on the interaction interfaces. We also analyzed the structures in terms of the node centrality with network analysis of protein structures (NAPS)⁴² using default parameters in the Web server.

Experimental Distance Vectors.

All structures were superimposed on the reference wild-type KRas4B-GDP structure at their common α-helix and β-strand residues (rigid parts, 109 residues in total). The distance vector ΔD = [D₂, D₃, D₄, …, D₁₆₆] contains the residue displacements given by Cα displacements

$$D_i=\sqrt{{({C_i^{\mathrm{A}}}_x-{C_i^{\mathrm{B}}}_x)}^2+{({C_i^{\mathrm{A}}}_y-{C_i^{\mathrm{B}}}_y)}^2+{\left({C_i^{\mathrm{A}}}_z-{C_i^{\mathrm{B}}}_z\right)}^2}$$

between two conformations of the protein (B and A).⁴³ Here, the Cα atom coordinates are represented by $C_i^{\mathrm{A}}$ and $C_i^{\mathrm{B}}$, where B is the reference structure coordinates and i is the residue number (from 2 to 166).

Anisotropic Network Model (ANM) with Only Reference Structure Cα Atoms.

We further performed ANM analysis of the inactive state of KRas4B (GDP-bound) which corresponds to the “open” conformation. A cutoff of 12.0 Å was chosen with a default spring constant of 1.0. NumPy⁴⁴ was used to compute the Pearson correlations between the ANM global modes and the experimental distance vectors. We visualized the protein motions with the NMWiz tool in VMD.^39,45

Allosteric Pathways.

The MCPath Web server was utilized to find allosteric pathways without a set target node.³⁷ We generated paths with 10 residue length starting from KRas4B mutational hotspot 12th residue (Gly in wild-type, Asp in G12D and Cys in G12C mutants). Furthermore, we looked for paths between common interface residue Y32 and known allosteric site residue Q99. The Ohm Web server was utilized with the default parameters to perform this allosteric path search.³⁸

RESULTS AND DISCUSSION

KRas4B Interactions Lead to Decreased Flexibility in Global Motions Not Only in the Binding Regions but Also in the Allosteric Lobe.

Protein binding is accomplished by mostly new noncovalent bond formation between the interacting residues from two proteins. Moreover, the interactions can affect residues in distal sites by disrupting previous residue communication paths and introducing new ones. This modulation can drastically change the usual residue communication network compared to monomeric states. Such alterations can be revealed by GNM analysis as evidenced with previous NMA studies.^31,46,47 Here, we applied NMA to reveal the unique dynamics of KRas4B when in complex with an effector or one of the two regulators. We used the monomer KRas4B structures as the reference. The slow modes from GNMs provide the collective residue motions; therefore, the alterations in both interface and distal sites upon binding can be better explained with the top slowest modes. The mean-square fluctuations of the weighted 10 slowest modes are shown in Figure 2 (mode frequencies are provided in Table S2). We identified the hinge sites based on the local minima in the mean-square fluctuation profiles of the proteins which mostly coincide with β4, β5, and β6 in the monomeric KRas4B (Figure 2A). These core β-strands contribute to protein stability. Centrality of a residue represents its topological importance in the residue–residue contact network.⁴⁸ Our further topological analysis of KRas4B residues shows that KRas4B residues with highest degree centrality (high number of contacts) overlap with the hinge regions in the monomeric KRas4B (Figure S1A, Table S3).

Figure 2.

Global protein motions from the slowest 10 modes of GNMs. The weighted squared fluctuations of the unbound KRas4B (A) and protein bound KRas4B (B) are plotted after normalizing the results. The common α-helix and β-sheet residues are indicated above the residue numbers. P-loop, Switch-I, and Switch-II are labeled on the plots along with the main protein interaction interfaces.

We find that functionally important residues are affected similarly by partner-binding across all complexes compared to their counterparts in the monomeric state. Protein binding causes the residues in Switch-I, β2-L3-β3, and L8 (P121) to become less flexible and the residues in α4 to become more flexible in all bound conformations (Figure S2). Switch-I is a common interface for all partner proteins. This might explain the decrease in its flexibility. β2-L3-β3 is termed the “interswitch region” and has potential roles in both nucleotide-switching and GTP-hydrolysis mechanisms.^14,49 On the other hand, α4 residues take part in allosteric regulation of Switch-II for GTP hydrolysis with α3-L7 residues.¹³ Overall, our approach, using the global motions, can catch functionally important residues of which a subset is already known in the KRas4B dynamics.

We clearly observe that the residue fluctuation changes are at not only the binding interface but also distal sites. Most of the regions with decreased flexibility upon binding correspond to specific interaction interfaces. However, the residues with altered fluctuations in the allosteric lobe are mostly distal to the interacting proteins. Interestingly, the G12C KRas4B–SOS1 complex has the most distinct fluctuation profile. SOS1 binding decreases the residue fluctuations in Switch-II and α3-L7 significantly, whereas it increases the residue fluctuations in β4-L6, β5, and β6-L10-α5 (Figure 2B). These results imply that the absence of a nucleotide may potentially lead to a more flexible/open conformation of the protein even compared to the GDP-bound monomer structure. Allosteric shift of α3-L7 enables GTP hydrolysis reaction to take place by Switch-II ordering as explained.¹³ The decreased flexibility in the α3-L7 region is unique to KRas4B–SOS1 interaction. This suggests that the regulatory role of GEFs depends on distinctive dynamics to allow nucleotide exchange.

NF1 inserts an Arg finger to nucleotide binding-site for correct Q61 positioning with P-loop and Switch-I residues around the nucleotide to catalyze the GTP hydrolysis.^11,50 The obvious reduction in fluctuations of Switch-I and G60-Y64 residues of NF1-bound KRas4B-GNP is most likely a result of NF1 binding. Lu et al.²³ also found reduced fluctuations when GAP was bound to KRas4B (compared to the unbound KRas4B) using MD simulations. They report Switch and interswitch region fluctuations were suppressed, and we report similar but more pronounced differences for the KRas4B-GNP–NF1 complex. Also, KRas4B bound to NF1 has the highest fluctuations in α1 and interswitch region and lowest in G60-Y64 residues compared to Raf-1 or SOS1 bound structures.

Raf-1 bound KRas4B has the lowest fluctuations in α1-L2 and interswitch regions but the highest in Switch-II region. Higher flexibility in Switch-II was also reported for Raf-1 RBD bound HRas complex compared to monomer HRas.⁵¹ Tran et al.⁹ recently published a KRas4B–Raf-1 structure which included the cysteine-rich domain (CRD). Although the main role of Raf-1 CRD is membrane anchorage, their study showed it is also important for the KRas–Raf-1 interaction. Raf-1 RBD in this complex interacts with α1-L2 residues in KRas4B, whereas its CRD interacts with the interswitch region. We also used this structure in this study. Differential binding of Raf-1 domains to KRas4B can explain the reduced flexibility of the KRas4B interface residues. Baussand and Kleinjung⁵² investigated HRas interactions with different effectors by MD simulations. They found “rigidification” at binding interfaces, which also affected distal sites, and an increase in flexibility at Switch-II, similar to our findings for KRas4B–Raf-1. Moreover, KRas4B in Raf-1 bound conformation has the lowest fluctuations in β4 and β5 hinge sites but the highest in α3-L7 and α4, even higher than the GAP bound structure (Figure 2B). This is interesting because Raf is also suggested to activate the intrinsic GTP-hydrolysis reaction in Ras proteins.¹³

Local Motions Reveal Common Hotspot Residues in P-Loop, β4, β5-L8, and L10.

Besides the global motions, we further investigated the fastest three individual modes of KRas4B in both monomer and complex states (Figure 3). The residues with high fluctuations (peak sites) at the fast modes are potential hotspots. In general, hotspots are critical residues for the stability and functionality of proteins.⁵³ We considered a residue as a potential hotspot if it corresponds to a peak site based on the mean-square fluctuation profile calculated from the fastest three modes (Table S5). Among them, some potential hotspots are consistently present in all conformational states. G10-A18 (P-loop), C80-F82 (β4), V114-K117 (β5-L8), and T144-K147 (L10) are regions with potential hotspots in most conformations. β4 is at the core of the protein separating the effector and allosteric lobes, and its residues have the highest betweenness (the highest impact in the information flow) in the monomeric structures (Figure S1B, Table S4). Most of the other hotspot residues are involved in nucleotide binding. In addition to P-loop, the nucleotide base binding motifs NKxD (116–119) and ExSAK (143–147)⁵⁴ are populated with hotspots in all conformations. These residues not only are highly fluctuating in the fast modes but also are highly central in the residue–residue contact network. For example, N116 is consistently the most central residue across all conformations. The monomeric KRas4B residues at peak sites also correlate with the residues with high stiffness.¹⁴ Therefore, we propose these residues as folding hotspots and nucleotide-binding hotspots. Furthermore, the most frequent mutations in KRAS mutant tumors occur at the codons G12, G13, Q61, and A146 (COSMIC data) which are known cancer-driver hotspots. And the hotspots we find either coincide with these residues or reside in their proximity in all structures.

Figure 3.

Local protein motions from the three fastest modes of GNMs. The mean-square fluctuations of the individual modes are plotted for each KRas4B structure. The prominent peak site residues are indicated by labels. Refer to Table S5 for the exact residue lists.

The local motions also provide some differences across multiple KRas4B conformations. Peaks at Y32-T35 are only present in KRas4B-GNP monomers and KRas4B-GNP–NF1 complex. This site was proposed as a potential binding pocket previously.⁵⁵ In addition, these Switch-I residues were proposed to be on a Raf RBD and RasGAP binding hotspot cluster.¹⁷ Although GNM analysis did not reveal hotspots at this site in Raf-1 RBD-CRD bound KRas4B-GNP, it correctly predicted them for NF1 GAP related domain (GRD) bound KRas4B. Local motions further show that G12D KRas4B-GDP, G12C KRas4B–SOS1, and KRas4B-GNP–Raf-1 have no peaks at D57-T58 while the others have. GDP-bound monomeric KRas4B and nucleotide-free KRas4B–SOS1 have higher fluctuations in their allosteric lobes compared to the other KRas4B conformations. In fact, the majority of the peaks in SOS1-bound KRas4B are located in the allosteric lobe.

Protein Binding Introduces New Domain Partitioning.

We constructed the residue–residue cross-correlation maps shown in Figure 4 using all modes from GNMs. We determine how the correlations between different residue fluctuations are altered in KRas4B to rewire the residue communications upon binding to its partners. The cross-correlation maps allow a fair comparison between three KRas4B conformations in their complexed states and signify the subdomain regions (like lobes) introduced upon protein binding. Residues at or near the binding interfaces correlate more strongly with the partner protein residues. This leads to a further partitioning of KRas4B lobes at partner-specific locations: both Switch regions in KRas4B–SOS1, Switch-II in KRas4B-GNP–NF1, and the interswitch region in Raf-1 bound Kras4B-GNP. These residues connecting the lobes have also high betweenness centrality which represents a high impact in the information flow in a residue contact network (Figure S1B, Table S4). Additionally, the regulator bound KRas4B complexes have a distinct subdomain formation in the allosteric lobe with a border at D108.

Figure 4.

Cross-correlations of protein bound KRas4B. Residue motion correlations were mapped using all GNM normal modes obtained. The interacting protein residues are given on the x-axis. Color scale ranges from −0.5 for highly anticorrelated motions (blue) to 0.0 for low or no correlation (green) and to 0.5 for high correlation (red).

Although the binding of the partners results in increased concerted motions overall (see Figure S3 for cross-correlations of the monomers), the binding interfaces show low intramolecular residue correlations. This might suggest that there is little communication between the Switch regions and the rest of the KRas4B when bound to the regulators. Indeed, the NF1 and SOS1 interactions are to facilitate the hydrolysis and nucleotide exchange reactions.² Also, the KRas4B residues other than in these regions show highly anticorrelated motions with the regulator residues. In the case of KRas4B-GNP–NF1, high intermolecular cross-correlations point to communications involving the nucleotide binding residues G12, Y32, and G60 in KRas4B. These residues and the bound nucleotide have high positive correlations with the Arg finger residues in NF1 GRD. Additionally, Switch-II residues show correlation with close-by NF1 residues in the complex. These communications may potentially play roles in the GTP hydrolysis reaction as well. SOS1-bound complex in general has KRas4B residues communicating with CDC25 domain (CDC25 homology domain). KRas4B residues in P-loop, Switch-I, Switch-II, and α3 show the highest correlation with CDC25. These are mostly on the binding interface. However, Switch-I residues in KRas4B also show correlation with SOS1 Ras-exchanger motif (REM) residues, despite not being in proximity to each other. Active state Ras proteins have been noted to bind SOS1 in the REM domain and allosterically regulate the GEF activity of SOS1 in a positive-feedback manner for activation of another Ras.^56,57 Despite absence of an active Ras protein in the complex we analyzed, it is noteworthy to find such a communication between REM and KRas4B in light of this mechanism.

In contrast to the regulator bound KRas4B complexes, the communication between the allosteric and effector lobes is still strong in Raf-1 bound KRas4B. This might coordinate the membrane interactions required for Ras–effector binding. These are also mediated by KRas4B allosteric lobe residues.⁵⁸ And on that note, the correlation map of Raf-1 bound KRas4B-GNP indicates communications with both RBD and CRD domains of Raf-1. KRas4B Switch-I and interswitch region residues show the most correlated motions with Raf-1, especially with residues in CRD. Raf-1 RBD has the most correlated motions with KRas4B α1 and β2. These also support both previous studies⁹ and our deductions from 10 slowest mode analysis.

KRas4B-GDP Intrinsic Fluctuations Are Correlated with Experimental Conformational Changes.

After the detailed analysis of each KRas4B conformational states, we observe how the global motions drive the change to the bound states. HRas deformation modes describing the transition between the “on” and “off” states were previously analyzed in detail, providing information on the nucleotide-switching mechanism in the Ras superfamily.¹⁵ On the basis of the GNM analysis, we propose that the inactive “open” state KRas4B-GDP modes may describe the transitions to the different protein-bound “closed” conformations. We generated experimental distance vectors from KRas4B-GDP to the other KRas4B structures (see Table S6 for RMSDs) and obtained the residue displacements required for the transitions to their conformations. Since ANM analysis gives the directions of residue motions as well, we used the 10 slowest modes from KRas4B-GDP ANM that could drive these conformational changes (the eigenvalues and mode weightsof the modes analyzed are provided in Table S7). The theoretical residue displacements were compared to the distance vectors. In general, significant correlation was observed as illustrated in Figure 5 (the dark red colors).

Figure 5.

Correlations between the KRas4B-GDP modes from ANM and the experimental distance vectors to the six other structures’ main KRas4B chains. The absolute values were used to plot the results, with darker red colors pointing to higher correlation.

Distinct Modes Drive KRas4B to Its Bound Conformational States.

Experimental distance vectors are most correlated with mode 1 and then with mode 5 in all except mutant KRas4B-SOS1. This complex has the highest correlation with modes 1 and 4, despite being relatively low. One of the global GNM/ANM modes can successfully drive the open conformations to closed conformations; however, the opposite is not always correct. A combination of several modes, from the most global to intermediate modes, are needed to open an already closed structure owing to the requirement of noncollective localized residue motions.⁵⁹ KRas4B-GDP is an open structure compared to all other structures studied here except KRas4B–SOS1. The mutant KRas4B in SOS1 complex is free of a nucleotide, and its Switch-I is disconnected from the P-loop with the SOS1 interaction that resulted in GDP-exit from the nucleotide-binding site.¹⁰ Its more “open” conformation would explain why KRas4B-GDP slow modes could not describe the transition to SOS1-bound conformation.

In Figure 6A, the KRas4B-GDP distance vectors to the other six conformations are shown in detail with their most correlated modes’ residue fluctuations. The peaks coincide with the most prominent fluctuating regions except for the Switch-I region residues Y32 and T35-I36. Slow modes fail to explain these Switch-I region fluctuations because these residues interact with Mg²⁺ and γ-phosphate of GTP in the closed structures.^12,16,60 And probably their fluctuations are more localized and driven by GTP.

Figure 6.

KRas4B-GDP ANM modes most correlated with the experimental distance vectors. (A) KRas4B-GDP distance vectors and their most correlated modes’ residue fluctuations are plotted. Mode 1 is drawn on the structure (B), and cross-correlations are mapped (C). Mode 5 is also depicted similarly in (D) and (E). Two perspectives are provided as indicated by the axis orientations (and the rotation). These structure images were rendered with Tachyon ray-tracing in VMD. Blue to gray and red coloring in these represents low to high mobility (color scale was narrowed to highlight the mobile sites), and the arrows indicate the arbitrary motion direction and amplitude. Maps in (C) and (E) have the same color-scale shown on the right; red color means correlated residue motions and blue anticorrelated.

KRas4B-GDP Switch-I Shows Anticorrelated Motions with Switch-II and α4 in Mode 1.

We mapped the direction of motions derived from eigenvectors of mode 1 onto the KRas4B-GDP structure, together with a color representation based on the amplitude of residue fluctuations (Figure 6B,C). The Switch-II motions are dominant in mode 1, but P-loop, Switch-I, α3 (residues 86–105), part of L8 (residues 120–125), α4 (residues 126–136), and L10 (between β6 and α5) also show some mobility (Figure 6B). We predict that this mode drives a loop closing movement that results in the “closed” conformations of KRas4B. High Switch-II motions correlated with the allosteric lobe, and antiparallel to Switch-I and P-loop motions enable this conformational change. Meanwhile, α4 and a portion of α3 (until residue 98) move in parallel causing the allosteric lobe to get near the effector lobe. The residue cross-correlations for this mode further show the correlations between these motions in detail (Figure 6C). Switch-I and Switch-II are anticorrelated in this map. Switch-II residues have highly correlated motions with allosteric lobe β6-L10-α5 (residues 140–159). While the first part of the effector lobe (residues T2-Q61) displays highly concerted motions as a whole, E37 displays an anticorrelated motion in this part. It rather makes correlated motions with E76-A146 in the allosteric lobe. E37 and D38 normally interact with HVR,⁷ and this could be the reason for the communication between E37 and residues in the allosteric lobe.

Interswitch Region Mobility Is High, and Switch Region Motions Are Correlated in Mode 5.

We similarly show mode 5 in Figure 6D,E. The major difference between mode 5 and mode 1 is the high mobility of the interswitch region (β2-L2-β3, residues 38–56) and lower flexibilities of P-loop and L6-α3. We explained how the interswitch region was affected in both GAP and effector bound KRas4B complexes, and here we see that its motions are also required to describe the conformational changes. Cross-correlations of mode 5 (Figure 6E) reveal that the interswitch region motions are correlated most strongly with each other, A11-V14 (P-loop), T35-D38 (Switch-I), A66 (Switch-II), L6-α3, and P121-P140 (in L8-α4-L9) residues. Switch-II residues after Y64 show anticorrelated motions with the rest of Switch-II residues. Overall residue fluctuations in the Switch-regions are correlated. They move collectively to reach the closed conformations. Furthermore, the boundary between the allosteric and effector binding lobes is more apparent in this mode.

Allosteric Paths Differ between “Open” and “Closed” Conformations.

We further inspected the communication networks in KRas4B structures by analyzing the allosteric paths. The most frequent KRAS oncogenic mutation occurs at the 12th codon. Therefore, we investigated paths starting at this position using MCPath and found residues that bind the nucleotide in the predicted results. The most populated pathways are given in Supporting Information (Table S8). MCPath finds interaction probabilities between any residues with pairwise calculations of their all-atom interaction energies.³⁷ Given a perturbation at the 12th residue of KRas4B, without setting a target node, the signal was propagated to Switch-I in most conformations. The only exceptions are KRas4B-GDP and KRas4B–SOS1, in which the paths remain strictly at the allosteric lobe (L8 and β4). G60 on the other hand, which binds γ-phosphate of GTP, appears only in the paths of monomeric KRas4B-GNP. KRas4B-GNP–NF1 path goes to an NF1 residue in proximity to Arg finger. Overall, these paths differentiate between the open structures where the P-loop is not strongly coupled to Switch regions but to the L8 region.

We also investigated paths involving the known KRas4B allosteric site at α3-L7.¹³ We utilized the Ohm server in order to include the bound nucleotide and cofactors. Ohm calculates allosteric paths by using a perturbation propagation algorithm whereby residues are assumed to be in contact with each other based on a chosen cutoff distance.³⁸ Y32 is a common interface residue and critical in GTP-hydrolysis as we explained; it forms a hydrogen-bond network around GTP with water molecules, T35 and Q61. There are four proposed allosteric-binding pockets in KRas4B that were validated with experimental research: pocket 1 is a hydrophobic cavity between interswitch and Switch-II; pocket 2 is near Switch-II and α3; pocket 3 is between L7 and α5; and pocket 4 is behind Switch-I.⁶¹ The Switch-I hotspots we predicted are at pocket 4. Q99 is located in the pocket 2 and propagates signals to Switch-II via a hydrogen-bonding network centered on R68.¹³ Small molecules were found to bind the pocket at this site and inhibit G12C KRas4B.⁶² Therefore, we specifically looked for paths between residues Y32 and Q99. The critical residues found from the most important 10 paths are shown as networks in Figure 7 (most important path nodes are provided in Table S9). These paths include A59, G60, R68, and Y96 residues from the known allosteric pocket and the hydrogen-bond network around it.^13,17,62 There are also common residues from our MCPath results; connections between P-loop and Switch-I residues of the closed KRas4B conformations are present in Ohm as well. All structures except KRas4B–SOS1 complex have paths through the nucleotide and Mg²⁺ nodes as expected, and the importance (weights) of these nodes are higher in the GNP-bound active conformations. This “importance” simply means they appear more often in the paths found (shown with larger nodes colored red in the networks). On the other hand, residues that are seen in only one or few paths are designated with low importance (shown with small nodes colored green). KRas4B-GDP paths alternate between different nodes that are of low importance, including the nucleotide. The signal from the nucleotide is relayed to G60 in the KRas4B-GNP paths. Interestingly, G60 is also a relatively important node in the mutant KRas4B-GDP path; it is connected to Q61, A59, and E37 in this result. Moreover, G60 comes up as a highly important residue only in the Raf-1 bound KRas4B-GNP pathway and it is of low importance in the Kras4B-GNP–NF1 pathway. In addition, the signal propagated to SOS1 residues in the GEF-bound KRas4B structure. P-loop, Switch-I, and Switch-II are not connected to a nucleotide in this structure, and SOS1 residues appear to make up for the absence of these connections. Therefore, SOS1 might facilitate the signal propagation between active site and allosteric site in this KRas4B conformation. To summarize, these paths connecting the active site residues to the allosteric site α3-L7 show differences in the open and closed conformations.

Figure 7.

Critical residues in the allosteric pathways found between KRas4B residues Y32 and Q99. The most important 10 pathways were included to obtain the network graphs provided from the Ohm Web server. The radius and color of each node represent its importance in the signal propagation; red colored big nodes are the most important, while green colored small nodes are the least important nodes. Node labels are in the form of “chain ID/residue index”. Chains A and R are KRas4B in the structures analyzed in this study. Start and target nodes, GDP-GNP and Mg²⁺ nodes are additionally labeled on the networks. Refer to Table S9 for the most important paths for each conformation.

CONCLUSION

We explored dynamics of KRas4B in SOS1, NF1, or Raf-1 bound conformational states by normal mode analysis (NMA). KRas4B is known to have two lobes in its G-domain. GNM analysis revealed that these lobes can be further partitioned into subdomains upon binding to its partners. Interface residues with high degree betweenness appear to mediate the sectioning of the effector lobe. These new lobe divisions can be critical in modulating the interaction dynamics with small molecules. Furthermore, KRas4B interactions with different partners suppress the flexibility in not only their binding sites but also distant residues in the allosteric lobe in a partner-specific way. Interestingly, some of these distal residues have known roles in nucleotide-switching and the intrinsic GTPase activity. In addition, we identified nucleotide binding residues as hotspots in all the structures. ANM analysis of the open state KRas4B-GDP revealed that conformational changes can be driven by the intrinsic residue fluctuations. Moreover, our allosteric path investigation starting at the mutation hotspot G12 showed that this residue was strongly coupled to nucleotide binding residues in either the allosteric lobe or effector lobe in the open and closed states, respectively. The paths reside in Switch-I in the closed stateswhile going to the allosteric lobe in the open states. Paths connecting the common interface residue Y32 to known allosteric site residue Q99 further highlighted differences between the conformations. Overall, NMA helps us in understanding the partner-specific KRas4B dynamics comprehensively.