Site-specific monoubiquitination activates Ras by impeding GTPase-activating protein function

Abstract

KRas has recently been shown to be activated by monoubiquitination (mUb). Similar to oncogenic mutations, mUb of Ras at position 147 activates Ras by causing a defect in GTPase activating protein (GAP) function. To characterize the mechanism by which mUb impairs GAP-mediated downregulation of Ras, we made various modifications at position 147 of Ras and examined the impact on Ras sensitivity to GAP function. Whereas small modifications (iodoacetamide and glutathione) at position 147 of Ras do not affect GAP-mediated hydrolysis, ligation of Ras to Ub^G76C (native linker), Ub^X77C (one residue longer), and PDZ2 (with a native ubiquitin linker) was defective in GAP-mediated GTP hydrolysis. However, restoration of GAP activity was observed for Ras modified with the PDZ2 domain containing a shorter and stiffer linker region than ubiquitin. Therefore, the properties of the linker region dictate whether modification affects GAP-mediated hydrolysis, and our data indicate that the GAP defect requires a minimum linker length of 7 to 8 residues.

Ras GTPases regulate a variety of cellular functions, including gene expression, cell growth, and differentiation.¹ Although Ras proteins are part of a larger superfamily of small guanine nucleotide binding proteins, Ras is the only family member that is frequently mutated in human cancer. Approximately 30% of all tumors contain Ras mutations. There are three isoforms of Ras (HRas, NRas, and KRas), which differ by their subcellular location. KRas is the most commonly mutated isoform of Ras in tumors.² Ras GTPases cycle between the GDP-bound ‘off’ state and GTP-bound ‘on’ state to regulate a plethora of signaling pathways that control cellular growth. As the intrinsic rates of nucleotide dissociation and GTP hydrolysis are too slow to function on the time scale of most cellular events, Ras activity is dictated by regulatory proteins. These regulatory proteins include guanine exchange factors (GEFs), which activate Ras by facilitating nucleotide exchange, and GTPase-activating proteins (GAPs), which downregulate Ras activity by increasing the rate of GTP hydrolysis. Oncogenic mutations lead to constitutive activation of Ras most commonly by causing a GAP defect.

Recent work by Sasaki et al.³ demonstrated that post-translational modification of KRas at position 147 by monoubiquitination (mUb) is another, yet novel, mechanism of Ras activation. While mUb has been shown to regulate a number of cell functions, including DNA repair, gene expression, endocytosis, and nuclear export,⁴ the regulation of protein activity by mUb is emerging as another functional role.⁵ Furthermore, mUb is highly regulated and can be reversed by deubiquitinating enzymes.⁶ KRas is predominately monoubiquitinated at Lys¹⁴⁷. Sasaki et al. showed that mUb of KRas (mUbRas) resulted in an increased fraction of GTP-bound Ras, altered effector specificity, and increased tumorigenicity. Lysine¹⁴⁷ is part of a highly conserved motif (SAx) in Ras superfamily GTPases and forms critical interactions with the guanine nucleotide base. Within the SAx motif in the Ras subfamily, Lys¹⁴⁷ is less conserved (67%) relative to Ser¹⁴⁵ (97%) and Ala¹⁴⁶ (89%).⁷ Mutation of Ala¹⁴⁶ to valine results in oncogenic transformation of Ras due to increased nucleotide exchange, which populates Ras in the GTP-bound form.⁸ Thus, Sasaki et al. hypothesized that mUb of Ras at Lys¹⁴⁷ activates Ras by enhancing nucleotide exchange and populating Ras in its active GTP-bound state.

While mutation of the site of KRas mUb (K147L) reduced tumorigenesis in 3T3 cells injected into mice,³ it is unclear whether altered levels of Ras mUb occur in primary tumors as most studies probe for Ras mutations and not upregulation of Ras activity. Moreover, the levels of mUb Ras are controlled by enzymes responsible for ubiquitination and deubiquitination, which can become altered in many cancers.⁹ Aside from elucidating how ubiquitin alters Ras properties, determining the enzymes necessary for Ras-mediated ubiquitination is a critical future direction in understanding the role of this modification in healthy and diseased states. Sasaki et al. found that mUbRas^G12V was more oncogenic than Ras^G12V.³ This is an interesting result as KRas mUb may contribute to tumorigenicity in the context of a GAP-defective oncogenic mutation (G12V), possibly because of enhanced association with select effectors.³ Thus, more studies probing Ras oncogenicity should be performed in cell lines with and without Ras mutation-driven tumors. In addition, the role of mUbRas in regulating effector specificity should be investigated. Understanding the functional consequences underlying this post-translational modification will be critical to understanding the role of mUbRas under normal and oncogenic conditions.

Toward this goal, Baker et al. recently elucidated how mUb of KRas at position 147 upregulates Ras activity in vitro. Moreover, they were able to verify the GAP-deficient defect of mUb KRas immunoprecipitated from HEK293T cells.¹⁰ First, it was shown that mutating Lys¹⁴⁷ to less conservative substitutions, such as K147A and K147L, had only a modest effect on intrinsic and regulator (GAP/GEF)-mediated GTP hydrolysis and nucleotide exchange. This finding indicates that Lys¹⁴⁷ does not play a critical role in nucleotide binding, but rather, Lys¹⁴⁷ may be conserved in Ras GTPases for regulation by ubiquitin modification. Baker et al. developed a method to chemically modify Ras in vitro with ubiquitin using variants containing a cysteine substituted at the site of ubiquitination (Ras^K147C and ubiquitin^G76C).¹⁰ The studies were conducted in the context of the Ras^C118S substitution, which does not alter Ras structure or biochemical properties,^11,12 so that position 147 was specifically modified. The intrinsic and regulator-mediated rates of nucleotide dissociation and GTP hydrolysis on Ras^C118S/K147C (hereafter called Ras^CSKC) were measured. While mUb of Ras did not alter the intrinsic rates of nucleotide dissociation or GTP hydrolysis, mUbRas impaired GAP-mediated GTP hydrolysis. This finding explains the observed increase in active GTP-loaded Ras observed for the mUbRas pool in vivo.³ Baker et al. verified these in vitro findings in HEK293T cells as immunoprecipitated mUbRas, but not unmodified KRas, was GAP defective. Thus, increased Ras activation upon mUb results from the loss of GAP activity and not an increase in intrinsic nucleotide exchange.¹⁰ Monoubiquitination of Ras may be activated in a similar manner to common GAP-defective oncogenic Ras mutations at positions 12, 13, and 61.²

It is unclear how mUb of Ras impairs GAP-mediated hydrolysis while retaining the intrinsic rate of GTP hydrolysis. One possibility is that mUbRas is not responsive to GAP-mediated hydrolysis because GAPs cannot bind to the transition state of mUbRas, similar to many oncogenic mutations at position 12 and 13.¹³ Using the transition state analog that consists of AlF₃ and GDP, Baker et al. determined that mUbRas could still bind GAPs while in the transition state, albeit with decreased binding.¹⁰ However, binding to the transition state is not necessarily indicative of GAP sensitivity because the Ras^G12A mutant is unresponsive to GAP-mediated hydrolysis but can still bind to GAPs in the presence of the transition-state analogs.¹³ It is currently postulated that GAPs enhance the rate of GTP hydrolysis by stabilizing the switch regions around position Gln⁶¹ for in-line nucleophilic attack on the γ phosphate as well as to move Tyr³² to an open position, which provides space for the GAP arginine ‘finger’ to insert into the binding pocket and catalyze phosphate hydrolysis.¹⁴ The data from Baker et al. indicate that ubiquitin ligation to Ras impairs GAP-mediated hydrolysis without forming a high-affinity interaction with a particular surface of Ras. Using nuclear magnetic resonance (NMR) spectroscopy and Rosetta-based computer simulations, Baker et al. demonstrated that ubiquitin does not specifically interact with Ras but that ubiquitin ligation to Ras affects the dynamics of the switch regions and phosphoryl-binding loop (p-loop).¹⁰ This finding suggests that mUbRas may alter the conformational dynamics of regions critical for GAP-mediated GTP hydrolysis.

While Ras has not been shown to directly interact with ubiquitin, it is unclear if there are transient or low affinity interactions between Ras and ubiquitin that alter the ability of GAP proteins to facilitate the rate of GTP hydrolysis. To simplify this post-translational modification, mUb may be viewed as two components: a ball, which consists of the core ubiquitin domain, and a chain (tether), which consists of the 7-residue linker from the core ubiquitin domain to the carboxy-terminus. To confirm that the deficiency in GAP-mediated GTP hydrolysis was not due to specific interactions formed between Ras and ubiquitin, we ligated Ras^CSKC to a modified PDZ2 domain (PDZ2^UL; with the same C-terminal linker as ubiquitin), which is a protein with a similar size and fold as ubiquitin. We found that Ras ligated to PDZ2^UL was also defective in GAP-mediated GTP hydrolysis, which supports earlier NMR and modeling studies¹⁰ that indicate that ubiquitin does not specifically interact with Ras but samples a range of conformations on the surface of Ras to render Ras insensitive to GAP-mediated GTP hydrolysis. To address whether the size of the modification (ball) or length of the tether (chain) at position 147 significantly contributes to the loss of GAP-mediated hydrolysis, we expanded on our earlier work by modifying Ras with various modifications. These modifications include iodoacetamide (IAA), glutathione, PDZ2 with a shortened linker region (PDZ2^SL; the chain was truncated to 6 residues), and ubiquitin with a slightly longer linker Ub^X77C (chain length was increased to 8 amino acids) and measured the intrinsic and GAP-mediated rates of GTP hydrolysis.

Although mutation of Ras at Lys¹⁴⁷ does not significantly alter nucleotide hydrolysis,¹⁰ it is possible that small modifications of Ras at Lys¹⁴⁷ could affect GAP-mediated GTP hydrolysis. Thus, we first modified purified KRas^CSKC with IAA. This modification was selected because IAA is small, uncharged, and the chemistry of cysteine oxidation has been sufficiently documented.^15,16 Using a fluorescence-based GTP hydrolysis assay,^10,11 we found that iodoacetylation of KRas^CSKC did not affect either the intrinsic or GAP-mediated rates of GTP hydrolysis when using the p120 RasGAP catalytic domain (GAP³³⁴) compared with unmodified KRas^CSKC (Fig. 1A). We also modified KRas^CSKC with glutathione. Glutathione was selected because it is larger than IAA (305 Daltons compared with 58, respectively) and has a negative charge. Furthermore, the cysteine in glutathione allows for disulfide bond formation between KRas^CSKC and glutathione, which is similar to the ubiquitin ligation technique presented in Baker et al.¹⁰ We verified that KRas^CSKC was 100% glutathiolated at position 147 using mass spectrometry (Fig. 1B, C), and similar to our findings with iodoacetylated Ras^CSKC, we find that glutathiolated Ras^CSKC possesses similar rates of intrinsic and GAP-mediated GTP hydrolysis compared with unmodified KRas^CSKC (Fig. 1A). These in vitro results indicate that perturbations at position 147 of Ras, by mutation or small molecule modifications (up to ~300 Da), do not affect GAP-mediated hydrolysis.

Figure 1. (A) Rates of intrinsic and GAP-mediated hydrolysis of KRas with various modifications. Using the minimal catalytic domain of p120 RasGAP, residues 1–334, GTP hydrolysis was measured with and without RasGAP. The relative rates of hydrolysis in the absence and presence of RasGAP are presented using the phosphate detection protein FlipPi, as previously reported.^10,11 All reactions were performed in triplicate and were fit to a single exponential dissociation curve using GraphPad Prism. The error is reported as the standard deviation of the replicates. (B) MS3 product ion showing c- and z-type product ions resulting from the dissociation of the N-C (α) bonds due to electron transfer dissociation (ETD) of a 6⁺ charge of collision-induced dissociation (CID)-derived [Y30+G] product ions. The [Y30+G] product ions were generated by CID-MS/MS, which generated b- and y-type fragment ions from the dissociation of the backbone amide bonds of a 20⁺ product ion from intact KRas^CSKC modified with a single glutathione modification. (C) MS4 product ion spectrum derived from CID-MS3 of the singly charged glutathione product ion (G⁺). Note that the glutathione moiety was confidently localized to cysteine 147 of KRas^CSKC by tandem mass spectrometry.

We next asked whether the linker length associated with either PDZ2 or ubiquitin ligation to Ras was important for GAP-mediated GTP hydrolysis. We have previously shown that a ~2 kDa larger ball (PDZ2^UL instead of ubiquitin; Fig. 2D) ligated to position 147 of Ras retained the GAP defect observed with mUb¹⁰; however, it was unclear how different chain lengths associated with the ligation affect GAP-mediated GTP hydrolysis. Toward this end, we altered the properties of the linker between Ras to either ubiquitin or PDZ2. We first increased the linker length by 1 residue by ligating Ras to a ubiquitin variant (Ub^X77C; Fig. 2C) and observed a similar GAP defect to that of mUbRas. However, when we shortened the chain length by 1 residue, by modifying Ras with PDZ2 with the native PDZ2 carboxy-terminus (PDZ2^SL; 6 residue linker; Fig. 2E), we observed that ligation of PDZ2^SL to Ras restored sensitivity to GAP-mediated GTP hydrolysis. While the linker length for this ligation is 1 residue shorter, the composition of the linker differs from that found in ubiquitin and is predicted to be less flexible by FlexServ.¹⁷ Taken together, these data indicate that the linker length and/or flexibility are critical for the GAP defect upon ligation to position 147 of Ras and suggest that conformational sampling or accessibility to a particular surface(s) of Ras may play an important role in the ability of mUb at position 147 of Ras to impede GAP-mediated GTP hydrolysis.

Figure 2. Depiction of the protein:protein ligations in this study. Ras is indicated in red and the attached modifications are in black. The single letter amino acid code associated with the tether of ubiquitin and PDZ2 are highlighted in bold font. (A) The native ubiquitin modification of Lys¹⁴⁷ of Ras. (B) The chemical ubiquitination by disulfide formation between Ras^K147C and Ub^G76C. (C) The chemical ubiquitination of Ras^K147C by Ub^X77C. Gly⁷⁶ is highlighted by a black box to show the increased size of the Ub^X77C modification compared with the native modification. (D) The chemical ubiquitination of Ras^K147C to PDZ2^UL. (E) The chemical ubiquitination of Ras^K147C to PDZ2^SL. The linker region of PDZ2^SL is outlined to highlight the differences within the linker region. The core domain of each ligated protein is also shown. Figure generated using ChemBioDraw Ultra, version 13.0.

By comparing the findings in Baker et al. to the present results, the composition and length of the linker (chain) associated with either PDZ2 or ubiquitin ligation to position 147 of Ras appear critical for inhibition of GAP-mediated GTP hydrolysis. Although we find that a ball size between 8–10 kDa attached to a ubiquitin linker of at least 7 residues can impede GAP-mediated hydrolysis, we have not determined whether there is a minimum ball or maximum chain size associated with the inhibition of GAP function. However, we have shown that there is a minimal chain length for effective inhibition. We modified Ras at position 147 with Ub^X77C (Fig. 2C), which is 1 C-C bond length longer than the native ubiquitin modification as well as Ub^G76C (Fig. 2B), which is 2 C-C bond lengths shorter than the native modification, and both of these modifications fully inhibit GAP-mediated GTP hydrolysis in our assays. Further, when using the PDZ2^SL modification (Fig. 2E), which is 5 C-C bond lengths shorter than the native modification, Ras was responsive to GAP-mediated GTP hydrolysis (see Fig. 3 for a structural representation of the modifications). These results suggest that there is a minimum chain length for the ubiquitin modification to inhibit GAP-mediated GTP hydrolysis of Ras. However, as ubiquitin does not appear to specifically interact with Ras, we reviewed proposed mechanisms of GAP-mediated hydrolysis to generate a hypothesis as to why mUbRas displays a GAP defect.

Figure 3. Structural representation of KRas^K147C modifications. Ribbon diagram of: (A) KRas (PDB 3GFT) with the K147C variant modified with iodoacetamide; (B) KRas (PDB 3GFT) with the K147C variant modified with glutathione; and (C) HRas (PDB 1CRR) with the K147C variant ligated to ubiquitin. Ubiquitin is ligated to Ras by a disulfide bond between K147C and a C-terminal cysteine in ubiquitin (G76C), and (D) HRas (1CRR) with the K147C variant ligated to PDZ2^SL through a disulfide bond with the PDZ2^SL C-terminal cysteine. Structural depictions in (A) and (B) were generated in PyMol.²³ Depictions in (C) and (D) were generated using Rosetta²⁴ and represent low-energy models of the complex. Ras is shown in teal in all images, switch I (residues 25–40) is in red, and switch II (residues 57–75) is in green. Each thiol modification is in magenta. The nucleotide has been removed from (B), (C), and (D) for clarity.

GAP proteins interact with Ras and greatly facilitate GTP hydrolysis rates via three primary steps, according to a report by Kötting et al.^18,19 Using Fourier-transformed infrared (FTIR) spectroscopy and photo-caged GTP-bound Ras complexed with p120 RasGAP, the authors first observed movement of switch I from an ordered ‘off’ state to an ordered ‘on’ state. The second step of GAP-mediated hydrolysis involves the movement of the arginine finger of the GAP into the catalytically active position, and the third step is cleavage of the γ phosphate. Recently, Rudack et al. used FTIR and biomolecular simulations to study the mechanism of GAP-mediated GTP hydrolysis in Ras in more detail.²⁰ In this study, the authors suggest that GTP hydrolysis is a result of the α, β, and γ phosphates being pulled into an eclipsed conformation by GAP binding, which results in the γ phosphate becoming more electro-positive; this allows the catalytic water to hydrogen bond with the γ phosphate and accelerate hydrolysis by nucleophilic attack. Further, Rudack et al. used FTIR to show that the GAP-arginine finger forms hydrogen bonds with the α and γ phosphates of GTP.^20,21 Molecular dynamics simulations performed by Kötting et al.^18,19 indicate that when the arginine finger inserts into the active site of Ras, it displaces 5 water molecules and increases the negative charge on the β phosphate and imparts partial positive charge to the γ phosphate of GTP. The removal of the water molecules has been theorized to provide the thermodynamic (entropic) energy necessary for GAP-mediated hydrolysis.¹⁸ Furthermore, Grigorenko et al.²² suggested that the placement of the arginine finger locks a conformation of the GTP binding pocket that decreases the free energy necessary for γ phosphate cleavage. In the GAP-bound state, Grigorenko et al. determined that Gln⁶¹ of Ras and Arg⁷⁸⁹ of p120 RasGAP, as well as Thr³⁵ (switch I), Mg²⁺, Gly⁶⁰ (switch II) and Lys¹⁶ (p-loop), provide further orientation of the phosphates of GTP and facilitates hydrolysis. However, in intrinsic Ras-mediated GTP hydrolysis, only Thr³⁵, Mg²⁺, Gly⁶⁰, Lys¹⁶, and water molecules are directly involved in GTP hydrolysis, which results in an increased activation energy.²² The primary factors proposed for driving increased nucleotide hydrolysis of Ras by GAPs include reorientation of the γ phosphate, Gln⁶¹, and catalytic water of Ras as well as the placement of the arginine finger from the GAP.

These mechanistic studies of GAP-mediated hydrolysis can be used to generate a hypothesis for the results observed by Baker et al.¹⁰ Although Baker et al. showed that GAPs still bind to the transition state of mUbRas, the binding affinity and whether the GAP favors the proper conformation of Ras for GTPase activity was not determined. The data by Kötting et al. imply a structural rearrangement by insertion of the arginine finger by the GAP after binding to Ras, and it is unclear whether this structural rearrangement by the GAP is impeded in mUbRas. NMR and Rosetta simulations indicate that when Ras¹⁴⁷ is ligated to ubiquitin, ubiquitin samples a large surface area of Ras, including parts of the p-loop and switch regions. As GAP binding has been shown to restrict the motions of the phosphate moieties of GTP and switch regions,²⁰ these regions in mUbRas may not be conducive for GAP-mediated hydrolysis.¹⁰ Specifically, the NMR data suggest that conformational dynamics of the p-loop and switch regions are altered by ubiquitin ligation at position 147, which could change either the rates or population of the conformers associated with these regions. However, it will be important to better characterize the conformational dynamics associated with GTP-bound mUbRas as only the GDP-bound form of mUbRas was studied by NMR.¹⁰ The inability to populate the correct orientation of the active site, specifically the region near the p-loop, could prevent stabilization of the switch regions in Ras as well as the eclipsed conformation of GTP, which would likely prevent GAP-mediated hydrolysis. This hypothesis is consistent with the simulations performed by Rudack et al., who proposed that the eclipsed phosphate conformation is necessary for hydrolysis. Thus, modification alone at position 147 is not enough to prevent hydrolysis-competent conformations; however, the ability of tethered ubiquitin to sample regions associated with the catalytic site (p-loop and switch regions) is necessary to interfere with GAP function. In this case, other modifications of Ras, such as PDZ2^SL, that possess a shorter tether may not alter the conformational dynamics of the Ras switch regions and p-loop, and thus retain the ability to be downregulated by GAPs, which is likely why the smaller modifications (IAA and glutathione) and PDZ2^SL do not affect GAP-mediated hydrolysis.

In summary, we present additional data that provide further insight into how ligation of ubiquitin to Ras¹⁴⁷ results in a GAP-deficient phenotype. Using the KRas^CSKC variant, we were able to build on the recent publication by Baker et al. and show that smaller modifications of Ras at position 147 (IAA and glutathione) do not affect GAP-mediated hydrolysis. However, as both the PDZ and ubiquitin modifications result in a GAP defect, the minimum size of the modification necessary to generate the GAP defect is not known. Rather, by changing the tether length of the PDZ2 domain, we were able to show that there is a critical tether length needed to produce a GAP defect. Thus, with the PDZ2^SL modification, no loss of GAP function was observed; however, with the longer tether (PDZ2^UL), a complete loss of GAP function was observed. The ability of ubiquitin ligation to Ras at position 147 to impede GAP-mediated GTP hydrolysis appears to be dependent on the length and composition of the linker, which we speculate to be important for the conformational sampling of a surface(s) on Ras that prevents GAPs from catalyzing GTP hydrolysis and downregulation of Ras activity. As mUbRas appears to mediate Ras-induced tumorigenesis in HEK293T cells, understanding how this modification alters Ras function could aid in providing a new direction for inhibiting Ras-driven tumors.

10.4161/sgtp.26270

Baker R, Lewis SM, Sasaki AT, Wilkerson EM, Locasale JW, Cantley LC, Kuhlman B, Dohlman HG, Campbell SL. Site-specific monoubiquitination activates Ras by impeding GTPase-activating protein function. Nat Struct Mol Biol. 2013;20:46–52. doi: 10.1038/nsmb.2430.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.