Monoubiquitination of KRAS at Lysine104 and Lysine147 Modulates Its Dynamics and Interaction with Partner Proteins

Abstract

KRAS, a 21 kDa guanine nucleotide-binding protein that functions as a molecular switch, plays a key role in regulating cellular growth. Dysregulation of this key signaling node leads to uncontrolled cell growth, a hallmark of cancer cells. KRAS undergoes post-translational modification by monoubiquitination at various locations, including at lysine104 (K104) and lysine147 (K147). Previous studies have suggested that K104 stabilizes helix-2/helix-3 interactions and K147 is involved in nucleotide binding. However, the impact of monoubiquitination at these residues on the overall structure, dynamics, or function of KRAS is not fully understood. In this study, we examined KRAS monoubiquitination at these sites using data from extensive (12 μs aggregate time) molecular dynamics simulations complemented by nuclear magnetic resonance spectroscopy data. We found that ubiquitin forms dynamic nonspecific interactions with various regions of KRAS and that ubiquitination at both sites modulates conformational fluctuations. In both cases, ubiquitin samples a broad range of conformational space and does not form long-lasting noncovalent contacts with KRAS but it adopts several preferred orientations relative to KRAS. To examine the functional impact of these preferred orientations, we performed a systematic comparison of the dominant configurations of the ubiquitin/KRAS simulated complex with experimental structures of KRAS bound to regulatory and effector proteins as well as a model membrane. Results from these analyses suggest that conformational selection and population shift may minimize the deleterious effects of KRAS ubiquitination at K104 and K147 on binding to some but not all interaction partners. Our findings thus provide new insights into the steric effects of ubiquitin and suggest a potential avenue for therapeutic targeting.

Graphical Abstract

INTRODUCTION

Human KRAS4b (hereafter KRAS), a member of the Ras superfamily of small GTPases, plays a key role in regulating cell growth and proliferation. KRAS cycles between a guanine triphosphate (GTP)-bound active and a guanine diphosphate (GDP)-bound inactive conformational state.^1–3 This process is tightly regulated by GTPase activating (GAP) proteins that inactivate KRAS by stimulating its GTPase activity and by guanine nucleotide exchange factors (GEFs) that help release GDP and promote binding of GTP.⁴ The GTPase activity of KRAS is restricted to its catalytic domain (residues 1–166), which comprises an effector lobe or lobe 1 (residues 1–86) and allosteric lobe or lobe 2 (residues 87–166).⁵ The effector lobe, as the name suggests, harbors the effector-binding surface as well as the GTP-binding site defined by the P-loop (residues 10–17), switch-I (SW-I; residues 30–37), and switch II (SW-II; residues 60–76, which includes helix-2 (residues 66–74)).³ The occasionally membrane-proximal allosteric lobe^6–9 harbors the NKXD (residues 116–119) and EXSAK (residues 143–147) motifs that are involved in nucleotide specificity and binding.¹⁰ The concerted interaction between the two lobes results in the switch regions (switch I and II) adopting different conformations based on the bound nucleotide, with effector proteins selectively binding to active GTP-bound KRAS to activate downstream signaling pathways.

Oncogenic KRAS mutants are impaired in GAP regulation, which leads to defective GTP hydrolysis and an accumulation of active GTP-bound KRAS. The consequent increase in downstream signaling causes uncontrolled cell growth and differentiation, a hallmark of cancer cells.⁴ RAS proteins, which include NRAS and HRAS in addition to KRAS, are one of the most frequently mutated proteins in human cancers, accounting for nearly 20% of all cancers¹¹ KRAS mutations account for approximately 85% of all RAS mutations and are found at high frequency in pancreatic, colorectal and lung cancers, which, together, account for nearly 40% of all cancer-attributed deaths in the US.¹² However, direct targeting of KRAS by small-molecule inhibitors presents a unique set of challenges, and thus, KRAS was considered undruggable until recent successes in the development of inhibitors for a specific mutant that contains a reactive cysteine (KRAS G12C).^4,13–15 While these inhibitors are a promising step forward in efforts to combat Ras-driven cancers, multiple mechanisms of resistance against inhibition of KRAS signaling exist.¹⁶ Moreover, these covalent inhibitors do not target the most prevalent KRAS mutants. Hence, there is a need for new avenues of inhibiting KRAS by identifying novel mechanisms of its regulation.¹⁷ Post-translational modification (PTM) of KRAS may present one such avenue for therapeutic intervention.^16,18

KRAS undergoes PTM via ubiquitination at several lysine residues including K104 and K147.¹⁹ Located at the C-terminus of helix-3 (H3) and proximal to the GEF-binding surface, K104 is also a prominent site for other PTMs including acetylation.²⁰ Lysine104 forms stabilizing electrostatic interactions with R73 and G75 in helix-2 (H2) within the key SW-II region. Loss of these interactions leads to partial unfolding of H2 and defects in GEF and GAP interactions.²⁰ Intriguingly, KRAS monoubiquitinated at K104 (mUbKRAS¹⁰⁴) retains GEF and GAP activity.²¹ To better understand how KRAS ubiquitination at position 104 disrupts internal interactions with H2 yet retains interactions with regulatory proteins, we recently employed computational, biochemical, and nuclear magnetic resonance (NMR) approaches to identify a network of interactions involving R102 and D69 that compensate for the loss of K104 interactions with R73/G75; these newly formed ubiquitin-dependent interactions help maintain the H2/H3 interface and thus retain GEF activity.²² However, the structural basis for the retention of GAP activity was not examined.

KRAS K147, another major site of monoubiquitination, lies on the loop connecting β6 and H5 and is part of the EXSAK motif involved in nucleotide binding.¹⁰ Earlier studies of KRAS monoubiquitinated at K147 (mUbKRAS¹⁴⁷) indicated that this PTM mimics activation by oncogenic KRAS mutants that are defective in GTP hydrolysis and enhance the active GTP-bound population, resulting in upregulated KRAS signaling and tumorigenesis through the RAF and PI3K pathways.¹⁹ This increase in the GTP-bound population was found to be due to defective GAP action.²³ In the same study, computational modeling of mUbKRAS¹⁴⁷ suggested that ubiquitin impairs GAP action likely by steric occlusion, consistent with a recent report on the impaired binding of mUbKRAS¹⁴⁷ to the p120 GAP catalytic domain.²⁴ Further, compared with the wild-type KRAS (WTKRAS), GTP-bound mUbKRAS¹⁴⁷ showed lower affinity to the Ras-binding domains (RBD) of C-Raf and Ral-GDS as well as the catalytic subunit of phosphoinositide 3-kinase gamma (PI3Kγ).²⁴ In contrast, GDP-bound mUbKRAS¹⁴⁷ has a higher affinity for C-Raf RBD, suggesting a potential role for nucleotide-independent mUbKRAS¹⁴⁷ signaling.²⁴ However, the structural changes mediating these effects were not fully understood.

In this study, we employed molecular dynamics (MD) simulations and NMR spectroscopy to characterize the effect of K104 or K147 monoubiquitination on the structure, dynamics, and regulator- or effector-interactions of KRAS. We found that, while its interactions with KRAS are nonspecific and dynamic, ubiquitin adopts preferred orientations relative to KRAS, which is consistent with experimental data. The simulations also suggest that mUbKRAS samples a large configurational spaces but interactions with regulators and/or effectors may bias the population toward certain configurations. These observations provide deeper insights into the structural and dynamic changes in KRAS upon monoubiquitination at K104 and K147.

MATERIALS AND METHODS

Simulation Setup.

To check if K104 and K147 are accessible for modification by ubiquitin, we downloaded 9 GDP- and 5 GTP-bound WTKRAS structures from the Protein Data Bank (total of 16 and 6 unique chains, respectively) and calculated the average solvent accessible surface area (SASA) of K104 and K147 using Visual Molecular Dynamics (VMD) and a probe radius of 0.14 nm. The SASA of K104 is 37.8 ± 26 and 18.4 ± 23.7 nm² in GTP- and GDP-bound WTKRAS structures. The corresponding values for K147 are 67.4 ± 32.1 and 73.1 ± 28.2 nm². For comparison, we obtained SASA values of 0.03 ± 0.1 and 0.2 ± 0.5 nm² for the fully buried K16. Clearly, K104 and K147 are solvent exposed and amenable for ubiquitin ligation. We then used a structure of wild-type KRAS described in previous studies^22,25 to build four structural models of KRAS monoubiquitinated at K104 (mUbKRAS¹⁰⁴) and four models of KRAS monoubiquitinated at K147 (mUbKRAS¹⁴⁷). Ubiquitin (Ub) was linked to K104 or K147 via an isopeptide bond as described previously²² and placed in different orientations relative to KRAS in the four different starting configurations of both mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷. GTP and Mg²⁺ were added to mUbKRAS as per the template X-ray structure. Each starting model was then placed in a TIP3P (transferable intermolecular potential with three points²⁶) water box containing NaCl at a final concentration of 150 mM. The box size was set such that the edge of the box was 20 Å away from the protein (typically 10 Å), due to the potential for large conformational changes in mUbKRAS. The resulting systems were energy-minimized using the steepest decent algorithm until the highest F_max between two atom pairs was less than 1000 kJ/mol/nm. The energy-minimized systems were equilibrated under the canonical (constant number of particles, volume, and temperature or NVT) ensemble for 100 ps at 310 K and 1 bar. During this step, a harmonic restraint of force constant 1000 kJ/mol/nm² was applied on the Mg²⁺ ion and heavy atoms of mUbKRAS and GTP. The restraint was relaxed during three subsequent 100 ps equilibration steps under the isothermal–isobaric NPT (constant number of particles, pressure, and temperature) ensemble by reducing the force constant to 200, 40, and 0 kJ/mol/nm². The equilibrated systems were used to commence the production phase, which was performed with a time step of 2 fs for 1 and 1.5 μs each for the four mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ systems, respectively, totaling 10 μs of aggregate simulation time. A constant pressure of 1 bar and constant temperature of 310 K were maintained using the Parrinello–Rahman barostat²⁷ and the Nose–Hoover thermostat.²⁸ The particle-mesh Ewald method²⁹ (with a cutoff of 12 Å) was used to calculate long-range electrostatic interactions, with covalent bonds involving hydrogen atoms restrained using the LINCS algorithm.³⁰ The simulations were performed using GROMACS (v2016.3)³¹ and the CHARMM36³² force field. As reference, we also performed two simulations of GTP-bound wild-type (WT) KRAS for 1 μs each, under identical conditions as the mUbKRAS systems.

Trajectory Analysis.

The trajectories were analyzed using GROMACS tools and in-house VMD (v1.9.3)³³ and Python scripts. Two reaction coordinates were used to evaluate the dynamics of Ub in mUbKRAS: the distance and orientation of Ub relative to KRAS. Since the center of mass (CoM) distance between KRAS and Ub is proportional to the angle defined by the CoM positions of KRAS, LYQ104 or LYQ147 Cα atom, and ubiquitin we used this angle as a measure of the Ub–KRAS distance (LYQ represents ubiquitin-ligated lysine). To calculate the orientation of ubiquitin, we first defined and then aligned a vector from KRAS:CoM to LYQ104/147:Cα on the z-axis and calculated the angle defined by Q129:Cα, KRAS:CoM and Ub:CoM in the x and y coordinates. This angle effectively measures the rotation of Ub around KRAS on the x–y plane. The 2D density plots of the two metrics (or angles) were used to perform structure-based clustering analysis of the simulated conformers using DBSCAN.³⁴ We characterized ubiquitin–KRAS interfaces based on the frequency of pairwise interresidue contacts between ubiquitin and KRAS, where two residues are considered to be in contact with each other if any heavy atom of one residue is within 4 Å of any heavy atom of the other residue. The contact frequencies were normalized by the total number of conformations in each cluster or the entire ensemble. To visualize KRAS residues involved in interactions with effector or regulator proteins, we overlaid representative structures from each cluster (i.e., the cluster center) of mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ with the crystal structure of WTKRAS (GMPPNP-bound) in complex with the protein of interest (e.g., GAP-related domain (GRD) of neurofibromin (RCSB ID: 6OB2)). The overlays were performed by aligning the backbone atoms of the nonswitch regions of the simulated mUbKRAS and the KRAS crystal structure. Finally, we assessed the internal dynamics of KRAS or Ub using backbone root-mean-square deviations (RMSDs) and residue-based root-mean-square fluctuations (RMSFs). RMSDs were calculated over the entire simulation, while ensemble-averaged RMSF values were computed using the ensemble of conformers within clusters or the entire simulation data. We evaluated the absolute change in the fluctuation of KRAS upon monoubiquitination as ΔRMSF = |RMSF^WTKRAS – RMSF^mUbKRAS|. Visualization and image rendering were performed in VMD and UCSF Chimera.³⁵

Protein Ligation and NMR Spectroscopy.

The detailed protocol for protein expression and purification of KRAS and preparation of mUbKRAS has been described previously.²² The human KRAS4B (C118S) cDNA sequence containing a K104C mutation encoding the G-domain (guanine nucleotide-binding domain, residues 1–169) was generated using site-directed mutagenesis with protein expression in E. coli BL21 (DE3). Similarly, the full-length ubiquitin G76C mutant was expressed in E. coli BL21 (pLysS). Both the proteins were purified as described²² with protein purity validated using SDS-PAGE. Ubiquitin G76C was ligated to KRAS K104C using a chemical ligation strategy adapted from Baker et al.²³ NMR sample preparation and data collection were conducted as described previously.²² The NMR spectra were recorded using Bruker Avance III 700 and 850 spectrometers at 25 °C.

RESULTS AND DISCUSSION

In this study, we examined the effect of monoubiquitination at K104 and K147 on KRAS structure, dynamics, and protein interactions using atomistic MD simulations and NMR. The MD simulations were started from four different configurations of ubiquitin attached to GTP-bound KRAS via an isopeptide bond between G76 of ubiquitin and K104 (systems mUbKRAS¹⁰⁴:GTP or mUbKRAS¹⁰⁴ for short) or K147 (systems mUbKRAS¹⁴⁷:GTP or mUbKRAS¹⁴⁷) of KRAS (Figure 1A and Table 1). In each of the starting configurations, ubiquitin was linked to the lysine in an extended manner and arranged in different orientations around KRAS to reduce bias from the initial configuration. For reference, we simulated WTKRAS:GTP in two copies (see Table 1). Each simulation was run for 1 μs except for the mUbKRAS¹⁴⁷ systems, which were extended to 1.5 μs to check for any potential conformational changes that might not occur within 1 μs. The backbone RMSD plots in Figure 2B show that all simulations equilibrated quickly (well within 1 μs), and extending the runs to 1.5 μs did not lead to additional conformational changes in KRAS. Moreover, the RMSD plots indicate that monoubiquitination at K104 or K147 does not significantly affect the global structure of KRAS. To ensure comparable statistics among the three sets of simulations, we used the last 800 ns data from each simulation for analysis, and the results are discussed in subsequent sections.

Figure 1.

Ubiquitin linked to KRAS at K104 and K147 is dynamic but samples certain configurations during unbiased MD simulations. (A) The starting configurations of mUbKRAS used in this study. KRAS is depicted as an orange cartoon model with SW-I and SW-II highlighted in green and yellow, respectively. The GTP nucleotide is shown in licorice. Ubiquitin ligated to KRAS at K104 and K147 is depicted as a cartoon model in hues of black and blue, respectively. (B) Time evolution of the backbone RMSD of KRAS (excluding the switch regions) for each of the four mUbKRAS¹⁰⁴ (top) and mUbKRAS¹⁴⁷ (bottom) simulations, using the starting structure as a reference. (C) Time-course of the angle between the CoM of KRAS, LYQ (lysine ligated to ubiquitin) and ubiquitin (top), and the rotation of ubiquitin around KRAS (bottom) for mUbKRAS¹⁰⁴ (left) and mUbKRAS¹⁴⁷ (right). (D) 2D density plot of the distribution of KRAS–LYQ–Ub and ubiquitin rotation angles, based on the entire simulated data for mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷. The conformations within the clusters are depicted in blue, red, green, and purple. (E) Cluster identity (i.e., cluster number) versus simulation number for mUbKRAS¹⁰⁴ (top) and mUbKRAS¹⁴⁷ (bottom).

Table 1.

Summary of the Simulations Performed in This Work

	number of atoms
system	Conf1	Conf2	Conf3	Conf4	simulation length (μs) × # of runs
mUbKRAS104:GTP	81 629	101 932	76 658	83 120	1 × 4
mUbKRAS147:GTP	80 076	83 162	95 072	119 148	1.5 × 4
WTKRAS:GTP	20 526	20 526			1 × 2
					aggregate time = 12 μs

Figure 2.

Interactions of ubiquitin with KRAS are dynamic and nonspecific. (A) Normalized contact frequency of ubiquitin residues with KRAS for each of the dominant conformational clusters in mUbKRAS¹⁰⁴ (black) and mUbKRAS¹⁴⁷ (blue). (B) Contact frequency of ubiquitin residues with KRAS mapped onto the 3D structure of ubiquitin (1UBQ).

Ubiquitin Linked to KRAS at K104 or K147 Samples a Wide Range of Conformations, yet Shows Preference for Certain Orientations without Making Stable Contacts with KRAS.

While KRAS itself does not undergo large conformational changes (Figure 1B), the overall conformation of mUbKRAS evolves significantly over the course of the simulations. In fact, the orientation of ubiquitin around KRAS is quite dynamic and samples a large region of conformational space. To quantitate this in the mUbKRAS simulations, we defined two parameters: the angle between KRAS:CoM, K104/K147:Cα and Ub:CoM, and the rotation of ubiquitin around KRAS. The former is effectively a measure of the displacement of ubiquitin relative to KRAS, while the latter quantitates the rotation of ubiquitin around KRAS (see Materials and Methods). Figure 1C shows significant fluctuations of the mUbKRAS systems in terms of these measures, despite some dependence on the initial configuration. For example, ubiquitin is freely rotating in all simulations of mUbKRAS¹⁰⁴, while it is stable in two of the four simulations of mUbKRAS¹⁴⁷ (Figure 1C). Moreover, in both sets of simulations, ubiquitin’s rotational dynamics is less pronounced than its translation (i.e., ubiquitin moves toward and away from KRAS more frequently than it rotates around it). While this could in part be due to a restricted torsional flexibility of the isopeptide bond, a lack of stable interactions between ubiquitin and KRAS also contributes. In other words, the noncovalent interactions between Ub and KRAS, which involve only a limited set of ubiquitin and KRAS residues (see below), are largely nonspecific and transient. As a result, ubiquitin remains free to sample the broad rotational space around KRAS, aided in part by the starting configurations and by the fact that ubiquitin is tethered to KRAS only by a single isopeptide linkage.

However, we observed that there are certain conformations of mUbKRAS that are sampled more frequently in both sets of simulations, with mUbKRAS also residing in these conformations for longer periods of time during the simulations. To quantify this, we pooled together the four simulations for each of the mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ systems and conducted cluster analysis (Figure 1D). This analysis yielded four clusters of conformers for mUbKRAS¹⁰⁴ and three clusters for mUbKRAS¹⁴⁷, representing 61.6 and 57.9% of the data, respectively. While some degree of dependence on the starting configuration is apparent, most of the simulations sampled conformations belonging to one of the dominant clusters (Figure 1E). Results from these analyses indicate that ubiquitin has access to a large conformational space around KRAS but prefers certain orientations that are potentially relevant for function. Therefore, we used these preferred configurations or clusters for subsequent analyses.

The regulation of the proliferating cell nuclear antigen (PCNA) by monoubiquitination is well-established.^36,37 In this protein, ubiquitin stays in an extended conformation linked to PCNA only by the isopeptide bond.³⁸ In contrast, a previous modeling of mUbKRAS¹⁰⁴ observed that ubiquitin interacts with KRAS more extensively albeit in a dynamic and nonspecific manner.²¹ The current simulations corroborate this finding by showing that ubiquitin forms transient contacts with KRAS in all clusters in both sets of simulations (Figure 2). Irrespective of its ligation site, ubiquitin contacts KRAS primarily via its C-terminus (residue 72–76), β1, the loop connecting β1 and β2 (residues 7–11), and β3–β4–β5 (residues 34–54) (Figure 2). Although there are some variations among the clusters, roughly the same ubiquitin interface residues interact with different KRAS residues depending only on its orientation. Importantly, the ubiquitin interface identified here is consistent with the ubiquitin interface identified for mUbKRAS¹⁴⁷ using ¹H–¹⁵N 2D HSQC NMR in an earlier study.²³

Monoubiquitination at K104 and K147 Affects KRAS Structural Dynamics Differently.

Monoubiquitination can alter KRAS structural dynamics either by directly interacting with KRAS residues or allosterically altering the structure. K104 lies at the C-terminus of H3 and interacts with R73 and G75 at SW-II,²¹ whereas K147 is located within the nucleotide-binding SAK motif.³⁹ Therefore, ubiquitin ligated to either K104 or K147 has the potential to directly interact with the switch regions or allosterically alter the switch regions’ conformation or dynamics to modulate RAS function.⁴⁰ To check for these effects, we calculated the change in KRAS backbone structural fluctuations (ΔRMSF), and KRAS-ubiquitin residue contact frequencies (Figures 3 and 4). Reflecting the fact that K104 and K147 are located distal from each other, the KRAS surface patches interacting with ubiquitin during the simulations of mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ significantly differ (Figures 3 and 4). In mUbKRAS¹⁰⁴, ubiquitin interacts with three sites on KRAS (excluding the site of covalent linkage at 104): near SW-II, the N-terminus, and the C-terminal region. Ubiquitin interacts with residues throughout SW-II (clusters 1 and 2) or just a few residues at the C-terminus of H2 plus residues at the N- and C-terminal regions (clusters 3 and 4); in cluster 1, ubiquitin also interacts with H3 residues near K104 (Figure 3A).

Figure 3.

Ubiquitin at K104 forms transient interactions with KRAS and affects the dynamics of SW-I, SW-II, and the H2/H3 interface during MD simulations and NMR analysis. (A) Absolute normalized root-mean-square fluctuation differences (ΔRMSF) of KRAS residues between mUbKRAS¹⁰⁴ and WTKRAS simulations (left y-axis), and frequency of KRAS–ubiquitin residue contacts (right y-axis) in red line and blue bar, respectively. Data is shown for each cluster and the entire ensemble (complete conformational ensemble). (B) Chemical shift perturbations in KRAS calculated based on weighted average chemical shift of WTKRAS and mUbKRAS¹⁰⁴ NH peaks in ¹H–¹⁵N 2D HSQC NMR spectra of ¹⁵N-enriched WTKRAS:GMPPNP (top), and the ratio of the backbone NH peak intensity of mUbKRAS¹⁰⁴ to WTKRAS obtained from the ¹H–¹⁵N 2D HSQC spectrum (bottom). The red dotted lines indicate cutoffs for significance (CSP > 0.05 and I_ub/I_wt > 0.75). The two switch regions are highlighted in orange. (C) Residues with significant ΔRMSF (>0.1; blue), those in contact with ubiquitin (yellow), or both (red) mapped on to the WTKRAS structure. K104, the site of ubiquitination, is indicated in a purple stick model.

Figure 4.

Ubiquitin at K147 forms dynamic interactions with KRAS and affects the dynamics of the effector lobe. (A) ΔRMSF of KRAS residues between mUbKRAS¹⁴⁷ and WTKRAS simulations (left y-axis), and the frequency of KRAS–ubiquitin residue contacts (right y-axis) in red line and blue bar, respectively. Data is shown for each cluster and the entire esnsemble. The switch regions are highlighted in orange. (B) Residues with significant ΔRMSF (>0.1; blue), those in contact with ubiquitin (yellow), or both (red) mapped on to the WTKRAS structure. K147, the site of ubiquitination, is indicated in a purple stick model.

To experimentally probe the structural alterations resulting from monoubiquitination of GTP-bound KRAS at K104, we collected 2D ¹H–¹⁵N HSQC NMR spectra. 2D ¹H–¹⁵N HSQC spectra allow for the detection of protons directly bonded to a ¹⁵N nucleus, including both backbone and side-chain resonances. Because an NH resonance can be detected for every residue except for proline, the spectrum contains a “fingerprint” of the protein backbone conformation. While the ubiquitin disulfide linkage associated with GMPPNP-bound mUbKRAS¹⁰⁴ was stable for short-term NMR experiments, we were unable to collect 3D data needed for backbone assignments. Hence, we compared 2D ¹H–¹⁵N HSQC resonance differences between WTKRAS:GMPPNP (BMRB: 17785) and GMPPNP-bound mUbKRAS¹⁰⁴. Of note, NH resonances associated with residues 33–41, residues 54–64, and residues 68–72 are missing in the original assignment deposited in BMRB (BMRB: 17785) due to dynamic properties of the switch regions in the GTP-bound state of RAS. As shown in Figure 3B, chemical shift perturbations (CSPs) in the HSQC spectrum of mUbKRAS¹⁰⁴ were compared with WTKRAS. Despite the missing assignments for most of the switch regions, the end of the SW-II region (residues 74–76) and H3, particularly the area proximal to K104, exhibit significant CSP (>0.5 ppm). This differs from our previous study on GDP-bound mUbKRAS¹⁰⁴, in which the majority of H3 was not perturbed.²² CSPs were also observed in β1 and the C-terminus of the protein. These chemical shift changes correlate well with the residues observed to interact with ubiquitin in our simulations (Figure 3A). The switch residues are still undetected in the ubiquitin ligated sample, suggesting that these regions retain dynamics. While no single cluster from the mUbKRAS¹⁰⁴ simulations can explain all the CSPs, in combination, the clusters account for most of the residues perturbed in the NMR spectrum.

As the NH peak intensity in the ¹H–¹⁵N HSQC spectrum is dependent on the relaxation properties of amides, changes in the peak intensity can also be used to assess the changes in dynamics. Consistent with the CSPs, peak intensity decreases are observed for residues in the region neighboring K104 (residues 102–107), indicating changes in the backbone dynamics (Figure 3B). Similarly, we calculated the change in the backbone root-mean-square fluctuation (ΔRMSF) relative to WTKRAS and observed that residues in SW-I and SW-II and those proximal to K104 have significantly altered local dynamics (Figure 3A). The dynamics of the switch regions is supported by the absence of the I_ub/I_wt signal in the NMR results (Figure 3B), while the lack of peak intensity decrease in most of the nonswitch regions in the NMR results is consistent with the nearly zero values of ΔRMSF and contact frequencies in the MD results (Figure 3A). The MD simulations in combination with NMR data thus suggest that monoubiquitination of KRAS at K104 causes perturbations in switch regions and alters the helix-3 dynamics. Taken together, our findings suggest that mUbKRAS¹⁰⁴ adopts certain preferred configurations/conformations while sampling a large conformational space.

Similar contact analyses of the mUbKRAS¹⁴⁷ simulations revealed that ubiquitin’s transient contacts with KRAS primarily involve the P-loop, SW-I, and SW-II (Figure 4). Except in cluster 2, ubiquitin also interacts with KRAS at loop 8, which connects β5 to α4 and harbors the NKxD motif (residues 116–119) that is critical for nucleotide recognition. Unsurprisingly, KRAS residues in mUbKRAS¹⁴⁷ that interact with ubiquitin are spatially proximal to K147. Similar to mUbKRAS¹⁰⁴, ubiquitination at K147 affected the dynamics of SW-I and SW-II residues and parts of H3, but unlike the former, fluctuations of parts of L7 are also modulated by ubiquitination at K147. These changes in backbone flexibility of KRAS are consistent with the interactions of ubiquitin with the switch regions and with an earlier study of mUbKRAS¹⁴⁷:GDP,²³ where residues in the switch regions were observed using NMR to undergo a change in their backbone dynamics upon ubiquitination at K147.

Monoubiquitination at K104 and K147 Alters KRAS Internal Interaction Networks.

It is currently unclear why ubiquitination at K104 does not significantly affect KRAS interactions with regulators despite the loss of stabilizing electrostatic interactions of K104 with SW-II residues. In an earlier study, we identified a compensatory interaction between D69 and R102 that forms in mUbKRAS¹⁰⁴:GDP in the absence of K104–R73/G75 interactions, which stabilizes the H2/H3 interface to potentially facilitate GEF binding.²² We observed similar D69–R102 interactions in clusters 1 and 2 of our mUbKRAS¹⁰⁴:GTP simulations, as well as interactions between R73 and V103 that further stabilize the H2/H3 interface. However, cluster 3 has a low frequency of D69–R102 contacts and overall insignificant interactions between H2 and H3; likewise, cluster 4 did not show significant interactions between H2 and H3. It is notable that in clusters 1 and 2, but not in clusters 3 and 4, ubiquitin interacts with SW-II, suggesting a potential role for dynamic ubiquitin–SW-II interactions in stabilizing the H2/H3 interface that is important for GEF binding.^41,42 A similar analysis of the mUbKRAS¹⁴⁷ simulations revealed that, as in WTKRAS, K104 interacts with G75 in all three clusters (Supp. Table S1), as well as with E76 in clusters 2 and 3. Perhaps surprisingly, a D69–R102 salt bridge is also observed in mUbKRAS¹⁴⁷, further stabilizing the H2/H3 interface. Although the CHARMM36 force field used here tends to overestimate the interaction energy between the Arg side chain and carboxylates, it is interesting to note that the D69–R102 interaction was observed only in one of the two WTKRAS simulations (Supp. Table S1). This interaction has been observed previously only in a few crystal structures of WT²⁵ and G13D⁴³ KRAS and during MD simulations of Q61H KRAS,⁴⁴ but it appears to be facilitated by ubiquitination.

Ubiquitination of KRAS at K104 and K147 Affects Interaction with GAPs, Effectors, and Membrane through Steric Occlusion: Potential Role for Conformational Selection and Population Shift.

Regulators and effectors interact with KRAS primarily via the effector lobe, with SW-I and SW-II playing an important role. While the conformations sampled by the mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ simulations are potentially accessible in cells as well, not all of them would be conducive to regulator or effector binding due to occlusion by ubiquitin. Additionally, recent studies have indicated an important role of KRAS membrane orientation^9,45 and dimerization^46,47 in its functions. Therefore, we examined the contacts of ubiquitin in mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ with KRAS residues involved in interactions with GAPs, effectors and the membrane, as well as residues predicted to be involved in dimerization.

To examine occlusion-related effects of K104 and K147 monoubiquitination on GAP binding, we plotted the contact frequencies of the KRAS–GAP interface residues⁴⁸ with ubiquitin (Figure 5A). For mUbKRAS¹⁰⁴, KRAS–GAP interface residues did not interact with ubiquitin in clusters 3 and 4, while some of the residues in clusters 1 and 2 have partial interactions (Figures 5A and 6A). To visualize this, we overlaid a representative structure from the clusters on the crystal structure of WTKRAS:GMPPNP in complex with the GAP-related domain (GRD) of neurofibromin (NF1) (Figure 5B). As expected, we observe that ubiquitin in clusters 1 and 2 occludes the GRD, thereby potentially inhibiting Ras–GAP interactions. In contrast, ubiquitin in clusters 3 and 4 is oriented away from the GRD and would not interfere in the Ras–GAP interaction (Figure 5B). This correlates with the experimental data, wherein mUbKRAS¹⁰⁴ has only a partial defect in GAP activity.²¹ Similarly, a recent study showed that monoubiquitination of KRAS at K147 leads to defective GAP activity by impairing its binding to KRAS.²⁴ Here, too, occlusion effects are expected, because K147 is located proximal to the nucleotide-binding site, which is part of the GAP–RAS interaction interface that primarily consists of residues in SW-I and SW-II regions. Consistent with this expectation, all of the mUbKRAS¹⁴⁷ clusters show a high frequency of ubiquitin contact for the KRAS–GAP interface residues (Figures 5A and 6A). Further, comparing the overlay of mUbKRAS¹⁴⁷ clusters with the X-ray structure of GRD-bound WTKRAS:GMPPNP, we see that all the clusters significantly occlude the GRD (Figure 5B). This result provides a structural basis for the GAP defect previously observed²³ in mUbKRAS¹⁴⁷.

Figure 5.

GAP interactions are sensitive to monoubiquitination. (A) Ubiquitin-contact frequency of KRAS residues located at the GAP-binding interface in mUbKRAS¹⁰⁴ (black) and mUbKRAS¹⁴⁷ (blue). (B) Representative structures from the different mUbKRAS clusters overlaid on the crystal structure of GMPPNP-bound WTKRAS in complex with the GAP-related domain (GRD) of neurofibromin (NF1) (RCSB code 6OB2). KRAS is shown only for one of the mUbKRAS structures in gray ribbon. Ubiquitin from the different clusters is depicted in blue, red, green, and purple. GTP is in licorice, and the GRD is the gray transparent surface. mUbKRAS¹⁴⁷ has three clusters, and mUbKRAS¹⁰⁴ has four.

Figure 6.

Ubiquitin potentially occludes regulator and effector binding and affects membrane interaction. Contact frequencies of ubiquitin with KRAS residues at the (A) GAP-binding interface, (B) RAF RBD- and CRD-binding interfaces, (C) dimerization interfaces, and (D) membrane-binding interfaces in two distinct orientation states. Data is shown for each cluster and the entire ensemble of mUbKRAS¹⁰⁴ (left) and mUbKRAS¹⁴⁷ (right).

While data is lacking regarding the role of ubiquitination of KRAS at K104 on effector binding, ubiquitination of GTP-loaded KRAS at K147 has been shown to cause a significant disruption of binding to C-Raf RBD.²⁴ Analyzing the contact frequencies of ubiquitin with the KRAS C-Raf RBD interface residues in the mUbKRAS¹⁴⁷ simulations, we see that Ub interacts with some of these interface residues in two of the three clusters (Figure 6B). Therefore, Ub likely occludes and thereby prevents C-RAF RBD binding to KRAS, consistent with experiments. We also analyzed the potential impact of ubiquitination at K147 on the interaction of KRAS with C-RAF CRD (Figure 6B and Suppl. Figure S1A,B). Figure 6B and shows that CRD-interacting residues of KRAS do not make direct contact with ubiquitin, suggesting that they are available for interaction with RAF CRD in mUbKRAS¹⁴⁷. Similarly, the simulations predict that ubiquitin modification of GTP-loaded KRAS at K104 has a minimal effect on RBD or CRD recognition (Figure 6B and Suppl. Figure S1A,B).

In addition to regulators and effectors, KRAS signaling is also regulated by its organization and dynamics on the plasma membrane, which can be potentially modulated by ubiquitination. This issue has not been studied before. Therefore, we first checked if ubiquitination at K104 or K147 impacts the two previously reported surface patches involved in KRAS dimerization.^46,47,49 These dimer interfaces are predicted to involve surface residues from helices 3 and 4 (interface-1) or 4 and 5 (interface-2).⁴⁷ Both interfaces are relatively unaffected in the simulated GTP-bound mUbKRAS¹⁴⁷ (Figure 6C and Suppl. Figure S1F). In contrast, ubiquitin in mUbKRAS¹⁰⁴ forms significant contacts with several residues at dimer interface-1, especially in clusters 3 and 4, suggesting a potential occlusion of this interface by ubiquitin (Figure 6C and Suppl. Figure S1E). We next looked at the effect of monoubiquitination on KRAS orientation at the membrane. Previous studies from our lab identified two dominant orientation states of KRAS at the membrane.^9,50,51 In orientation state 1 (OS1), surface residues from helices 3 and 4 directly interact with the membrane, whereas in orientation state 2 (OS2), residues from β-strands 1–3 and helix-2 are in contact with the membrane (similar results were obtained by others (e.g., refs 52–56)). We calculated the contact frequencies of these residues with ubiquitin in the simulated mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ conformers. We found that residues in the OS2 interface interact significantly with ubiquitin in clusters 1 and 2 of mUbKRAS¹⁰⁴, while the membrane interfaces are relatively unaffected in clusters 3 and 4 (Figure 6D and Suppl. Figure S1C). In mUbKRAS¹⁴⁷, ubiquitin does not make significant contact with any of the KRAS–membrane interface residues (Figure 6D and Suppl. Figure S1D). This makes sense, since the site of ubiquitination, K147, is located distal from the membrane-binding interfaces. We conclude that ubiquitination at K104, but not K147, potentially affects both membrane reorientation and dimerization of KRAS.

In summary, our results suggest a model of KRAS–ubiquitin interactions involving certain preferred configurations of mUbKRAS¹⁰⁴ and mUbKRAS¹⁴⁷ that exist within a large conformational ensemble. We predict that some of these conformers affect interactions with regulator and effector proteins or membrane. Conversely, interactions with activators or regulators may bias the mUbKRAS population toward certain configurations. Biasing the mUbKRAS populations toward conformations that hinder interactions with regulators, effectors, or the membrane in a regulated manner may provide a new avenue for targeting KRAS for therapeutic purposes. This approach has been successfully demonstrated in an earlier study, where a small-molecule ligand was shown to stabilize KRAS in an orientation where the membrane occludes the effector-binding interface.⁵⁶ Similarly, several myristoylated peptides have been proposed to target the effector-binding surface and lock KRAS in a signaling incompetent orientation.⁵⁷ Ubiquitination of KRAS may also be targeted using traditional approaches, but this requires a better understanding of the ubiquitin ligases and deubiquitinases that regulate the monoubiquitination of KRAS.

CONCLUSIONS

KRAS is a small GTPase that plays a key role in regulating cell growth and proliferation. Mutations in KRAS drive many cancers with low survival rates, including pancreatic, colorectal, and lung cancer. Studies to identify novel mechanisms regulating KRAS functions have begun to focus on its post-translational modifications, one of which is monoubiquitination. KRAS undergoes monoubiquitination at K104 and K147. At a functional level, mUbKRAS¹⁰⁴ has been previously shown to behave similar to WTKRAS despite small compensatory defects in GEF and GAP activity and loss of the critical role of K104 in stabilizing the helix-2/helix-3 interface. On the other hand, mUbKRAS¹⁴⁷ has impaired GAP activity and behaves like most oncogenic KRAS mutants. However, the structural details and mechanisms behind these observations were not fully understood. In this study, we used MD simulations and NMR spectroscopy to gain insights. We showed that while ubiquitin samples a range of conformers and large conformational space around KRAS, it samples certain conformations more frequently due to stabilization by weak nonspecific interactions with KRAS. We propose that regulatory and effector proteins or membrane dynamics can bias the mUbKRAS population toward certain conformations. This conformational selection and population shift, in turn, helps mUbKRAS to selectively overcome the potentially deleterious effects arising from ubiquitination at K104 and K147. Further, we showed that ubiquitin occludes the KRAS–GAP interface in nearly all of the simulated conformers of mUbKRAS147 and some conformers of mUbKRAS104, explaining previous experimental observations on GAP activity. Our simulations also predicted that KRAS organization at the membrane may be disrupted by monoubiquitination at K104 but not K147. Our findings thus provide novel insights into the structural interplay governing the impact of ubiquitination on KRAS dynamics and interactions, and its ability to overcome them in a selective manner.