Posttranslational Modification of RAS Proteins

Introduction

Mutations of RAS genes drive cancer more frequently than any other oncogene. The recent explosion of cancer genomic analysis has revealed the paramount role of the mitogen activated protein kinase (MAPK) pathway in RAS-mediated oncogenesis. RAS proteins integrate signals from a wide array of receptors and initiate MAPK signaling, making them an exceedingly attractive target for cancer drug discovery. RAS proteins are fundamentally binary molecular switches in which the off/on state is determined by the binding of GDP or GTP, respectively. As such, the intrinsic and regulated nucleotide binding and hydrolytic properties of the RAS GTPase were historically believed to account for the entirety of the regulation of RAS signaling. However, it is increasingly clear that RAS proteins are also regulated by a vast array of posttranslational modifications (PTMs) (Fig. 1). The first PTMs recognized were those of the C-terminus of RAS that convert a globular, hydrophilic protein into one that is lipidated and has affinity for cellular membranes. Subsequently a large assortment of PTMs of the GTP-binding (G) domain (aa 1–165) have been reported. The current challenge is to understand what are the functional consequences of these modifications and which are physiologically relevant. Because direct inhibitors of RAS have been difficult to develop and because PTMs are catalyzed by enzymes that may offer targets for drug discovery, the study of RAS PTMs has been a high priority for RAS biologists.

Figure 1. Posttranslational modifications of RAS proteins.

All reported and validated modifications of the four mammalian RAS proteins are indicated. Prenylation and palmitoylation of the C-terminal portion of the hypervariable region (HVR) create affinity for phospholipid bilayers whereas phosphorylation and lysine methylation of KRAS4B modulate that affinity. Methylation, phosphorylation, sumoylation, nitrosylation and mono-, di- and polyubiquitination of the G domains regulate trafficking, GDP/GTP exchange and degradation.

CaaX Processing

The three mammalian RAS genes give rise to four proteins because the KRAS locus harbors alternative 4^th exons [1]. Whereas RAS proteins are 97% identical in their G domains (aa1–165), the 3’ half of the fourth exons encode a hypervariable region (HVR) that mediates subcellular trafficking and membrane association. This is accomplished by posttranslational modification (PTM) of this region that converts nascent RAS from a globular, hydrophilic protein to a peripheral membrane protein with a hydrophobic C-terminus that affords affinity for phospholipid bilayers [2].

The last four amino acids of each HVR consist of a CaaX sequence (cysteine, aliphatic x 2, variable) found in >200 peripheral membrane proteins [3]. RAS and other CaaX proteins are synthesized in the cytosol on free polysomes. Upon termination of translation, CaaX sequences are recognized by a prenyltransferase that attaches a C15 farnesyl or C20 geranylgeranyl polyisoprene lipid to the CaaX cysteine. These enzymes utilize as substrates farnesyl or geranylgeranyl pyrophosphates that are intermediates or derivatives, respectively, in the cholesterol biosynthetic pathway [4]. Whereas protein acylation is accomplished on cysteines via a labile thioester linkage, prenyl lipids are attached via a stable thioether linkage that is irreversible and persists throughout the life of the protein. Prenyltransferases are cytosolic, heterodimeric enzymes composed of an α and β subunit. Four α subunits combine with three β subunits to generate four holoenzymes, farnesyltransferase (FTase) and geranylgeranyltransferase (GGTase) types 1–3 [5]. GGTase 2 modifies RAB proteins that lack CaaX sequences and adds two geranylgeranyl lipids to C-terminal CXC or CC sequences. The other three enzymes modify CaaX proteins. The X position of the CaaX sequence determines which prenyltransferase modifies the protein. RAS proteins end with CaaM or CaaS and are substrates for FTase [6]. Many RAS-related small GTPases, including most RHO family proteins, end in a CaaL sequence, which directs modification by GGTase I [7].

Once CaaX processing of RAS was elucidated more than three decades ago, FTase became a logical target for anti-RAS drug discovery and numerous FTIs were developed [8]. Although these drugs were highly potent, orally bioavailable, hit their target in vivo, and inhibited HRAS-driven tumors, they failed in the clinic as anti-cancer drugs. This unfortunate result was shown to be a consequence of the fact that the vast majority of RAS mutant human cancer is driven by KRAS. Although KRAS is normally a substrate only for FTase, in the presence of an FTI it can be modified by GGTase 1 such that its membrane association and signaling function are preserved [9,10].

Prenylation is the first of three sequential modifications of the CaaX sequence. It is followed immediately by endoproteolytic removal of the aaX amino acids by RAS converting enzyme 1 (RCE1) and methylesterification of the newly C-terminal prenylcysteine by isoprenylcysteine carboxylmethyltransferase (ICMT). These latter modifications position the lipid moiety at the extreme C-terminus of the protein and neutralize the negative charge on the C-terminal carboxylate adding hydrophobicity and allowing the C-terminus to intercalate into phospholipid bilayers. Interestingly, before proteolysis and carboxyl methylation, prenylated CaaX sequences appear to be targeted specifically to the membranes of the endoplasmic reticulum (ER). The biophysical properties of the ER membrane that supports this affinity are unknown and some have argued that it is simply the vast surface area of the ER relative to that of the plasma membrane (PM) that accounts for the observed distribution [11]. However, at all expression levels, green fluorescent protein (GFP) extended with CaaX sequences lacking other membrane targeting motifs decorate the ER but are not observed on endosomes, mitochondria or the PM [12]. The mechanism of targeting to the ER notwithstanding, it is on this compartment that both RCE1 [13] and ICMT [14], both polytopic membrane proteins, are restricted such that we can be certain that fully processed CaaX proteins must visit the cytoplasmic face of the ER before trafficking to the PM. Like FTase, RCE1 and ICMT have been considered targets for anti-RAS drug discovery but no inhibitors have reached clinical testing.

CaaX processing (prenylation, proteolysis and carboxyl methylation) occurs immediately after translation and is highly efficient. Several observations support this view. First, top down proteomic analysis of endogenous KRAS4B in tumor cells revealed only farnesylated species, although the degree of carboxyl methylation varied with or without an activating mutation [15]. Second, unprocessed RAS proteins (e.g. those that include a C>S substitution in the CaaX sequence or those in cells treated with high concentrations of FTIs) have a slower electrophoretic mobility allowing quantification of processed and unprocessed species. Whereas markedly overexpressed RAS proteins migrate with both mobilities, endogenous RAS migrates only as fully processed protein [16–18]. Thus, CaaX processing appears to be a constitutive feature of the proteome. Nevertheless, some evidence suggests that CaaX processing can be modulated. Carol Williams’ group has shown that splice variants of small GTP-binding protein GDP-dissociation stimulator (SmgGDS), a guanine nucleotide exchange factor active against a wide array of small GTPases, can regulate RAS prenylation [19,20]. Noncoding RNAs have also been reported to modulate access of KRAS to FTase [21]. Interestingly, Drosophila Ras1 is inefficiently prenylated as a consequence of a lysine in the a1 position of its CaaX sequence, but the unprocessed form can nevertheless support eye development [22]. Although FTase activity may be constitutive, recently it was shown that its expression is regulated by scaffold association factor B (SAFB), a nuclear DNA and RNA binding protein required for the transcription of FNTA, the α subunit of FTase [23]. Silencing SAFB sensitized KRAS-driven tumor cells to growth inhibition by FTIs [23].

Aside from insertion into phospholipid bilayers, CaaX processing also establishes affinity for prenyl-binding proteins that serve as cytosolic shuttling factors with the ability to shield the C-terminus of prenylated proteins from the aqueous environment [24]. The best studied of these is RhoGDI that sequesters RHO, RAC, CDC42 and other RHO family proteins in the cytosol where they are poised for redistribution to cellular membranes as part and parcel of their activation cycle [25]. RabGDI serves a similar function for RAB family small GTPases [26]. The δ subunit of phosphodiesterase type 6 (PDE6δ) serves this function for RAS and many other RAS-related GTPases [27,28]. In photoreceptors PDE6δ binds PDE6α and PDE6β, both prenylated CaaX proteins, and allows rapid transport to the outer segment where the enzyme is integral to visual signal transduction [29]. Interestingly, although there is little homology between RhoGDI and PDE6δ at the amino acid sequence level, there is marked structural homology whereby the prenyl group is inserted into a hydrophobic cleft bounded on both sides by β sheets suggesting convergent evolution [30,31]. Somewhat surprisingly, PDE6δ can accommodate both a C15 farnesyl or C20 geranylgeranyl polyisoprene lipid in its binding pocket [32]. RhoGDI is postulated to be regulated by RHO GTPase releasing factors (GRFs) that promote discharge of the GTPase from its carrier protein to allow insertion into membranes [25]. Whereas no GRF has been reported for RhoGDI, GTP-bound ARL2/3, a small GTPase lacking a CaaX sequence, serves as the GRF for PDE6δ [31] suggesting that the localization of ARL proteins marks a subcellular compartment as an acceptor membrane for RAS. PDE6δ is not the only protein that functions as a prenyl binding protein for RAS. Others include prenylated RAB acceptor protein 1 [33], SmgGDS [34], galectin-1 [35], and VPS35 [18].

Palmitoylation

CaaX processing is necessary but not sufficient for delivery of RAS proteins to the PM. Also required is the so called “second signal” immediately upstream of the CaaX sequence in the HVR. The second signal imparts both higher affinity for the phospholipid bilayer as well as the trafficking information required to transport RAS from the cytosolic face of the ER to the PM (Fig. 2). There are two types of second signal. One requires no further PTM and consists of a stretch of basic amino acids that allows for an electrostatic interaction with the negatively charged phospholipid headgroups of the inner leaflet of the PM. The second is additional PTM with one or two palmitoyl acyl chains that modify cysteines. Whereas a farnesyl only modification imparts relatively low affinity to membranes, each of these second signals adds the affinity required for trafficking and stable association with membranes [36].

Figure 2. Postprenylation modification and trafficking of RAS.

All RAS proteins are prenylated in the cytosol and then delivered to the cytosolic face of the endoplasmic reticulum (ER) where CaaX processing is completed through the actions of RAS converting enzyme 1 (RCE1) and isoprenylcysteine carboxylmethyltransferase (ICMT). CaaX processed NRAS and HRAS then traffic to the cytosolic face of the Golgi apparatus where they are palmitoylated and thereby gain enough affinity for membranes to engage in vesicular traffic to the plasma membrane (PM). NRAS also traffics through the cytosol in complex with chaperones such as PDE6δ and VPS35. Depalmitoylation occurs at the PM allowing NRAS and HRAS to cycle back to the Golgi for another round of palmitoylation. KRAS4A is palmitoylated, although the location for this modification has not been determined. KRAS4A is also depalmitoylated at the PM and moves to endomembranes that include the outer mitochondrial membrane. KRAS4B has a strong polybasic region that substitutes for palmitoylation, the prenylated protein traffics in complex with chaperones like PDE6δ and loses affinity for the PM upon phosphorylation of S181 in its HVR.

S-acylation of cysteines via a thioester linkage is an exceedingly common PTM and may affect up to 10% of the proteome [37]. Although acyl chains of various lengths have been documented, by far the most common is C16 palmitate. Importantly, no consensus sequence for cysteine S-acylation has been defined. Early studies of acylated proteins employed metabolic labeling with [³H]palmitate but more recently, acyl exchange chemistry and click-chemistry labeling have afforded both quantification of acylation and affinity capture of acylated proteins [38–41]. Whereas NRAS and KRAS4A have one HVR cysteine modified by palmitate, HRAS has two. HRAS was among the first proteins characterized with this modification [42]. The enzymes that modify proteins with palmitate are known as palmitoyl acyltransferases (PATs). These enzymes all possess a DHHC sequence in their catalytic sites. Twenty-three PATs are encoded in the human genome with a variety of subcellular localizations [43]. PATs are typically promiscuous in that they modify a large spectrum of proteins. DHHC9 in complex with GCP16 was identified as a Golgi-localized PAT that modifies NRAS and HRAS [44]. However, silencing of DHHC9 diminished but did not eliminate NRAS palmitoylation demonstrating that RAS proteins can be modified by more than one PAT [45]. The PAT responsible for the PTM of KRAS4A has not been identified.

Unlike prenylation, S-palmitoylation is readily reversible and the lability of this PTM is integral to its biological function allowing for the establishment of palmitoylation/deplamitoylation cycles that direct protein trafficking. Although some depalmitoylation may occur non-enzymatically, a variety of esterases can accelerate this process. Acylprotein thioesterase 1 and 2 (APT1/2) were the first such esterases shown to have RAS depalmitoylating activity [46]. More recently the ABHD17 family of serine hydrolases have been shown to be physiologically relevant RAS thioesterases [47,48]. In addition to S-acylation, KRAS4A has been reported to be N-acylated on lysines in the HVR [49]. Although N-acylation is not readily reversible, SIRT2 can remove the acyl modification and alter the signaling potential of oncogenic KRAS4A [49].

Upon palmitoylation RAS proteins acquire a 100-fold greater affinity for membranes than that of the prenylated only protein [50,51]. This increased affinity creates a kinetic trap that enriches NRAS and HRAS at the cytoplasmic face of Golgi membranes from where these RAS proteins can engage vesicular transport and gain access to other membranes, including the PM [52]. Hancock and colleagues have shown that another consequence of RAS palmitoylation is to promote partition into PM microdomains such that RAS forms nanoclusters that promote signaling [53–55]. Bastiaens and colleagues have developed models wherein the continuous cycle of palmitoylation at the Golgi and depalmitoylation at the PM counteracts the entropy that would otherwise distribute RAS evenly among membranes and insures a sufficient concentration of RAS proteins at the PM to allow for signaling [56–58]. Since the dwell time on the PM of palmitoylated RAS proteins is critical for signaling, the regulation of depalmitoylation may play an important role, but is poorly understood. Ahearn found that prolyl isomerization in the HVR of HRAS catalyzed by FKBP12 promotes depalmitoylation and therefore operates as an internal timer for PM association [59]. Compartmentalized signaling of RAS proteins from distinct subcellular localizations is a well-established paradigm in RAS biology [60–63] and the palmitoylation/depalmitoylation cycle is among the most important regulators of this system. The best characterized functional consequence of the palmitoylation/depalmitoylation cycles of RAS proteins is the regulation of hexokinase 1 on the outer mitochondrial membrane (OMM) by KRAS4A [64]. Whereas palmitoylated KRAS4A is expressed primarily at the PM, depalmitoylated KRAS4A traffics to endomembranes, including the OMM, where it encounters hexokinase 1 and reverses its allosteric regulation by glucose-6-phosphate [64]. Other differential biological outcomes controlled by palmitoylation-driven RAS localization include positive and negative selection of thymocytes [65]. Recently Shannon et al. demonstrated in vivo that palmitoylation is required for Nras-driven myeloid transformation adding encentive to target this pathway in cancer [66].

Phosphorylation

Reversible phosphorylation of signaling molecules is an exceedingly common mode of regulation. HRAS, NRAS and KRAS4B have been shown to be phosphoproteins although the functional consequences of this PTM remain poorly defined. Ora Rosen and colleagues reported in 1987 that KRAS4B is a substrate for protein kinase C (PKC) [67]. Around the same time Arimura et al. reported that HRAS is a substrate for protein kinase A (PKA), although the stoichiometry of phosphate incorporation was too low to be physiologically meaningful [68]. More recently Kim et al. found that phosphorylation of HRAS by glycogen synthase kinase 3 β (GSK3β) on threonine residues 144 and 148 led to polyubiquitination by β-transducin repeat-containing protein (β-TrCP) and degradation [69]. Yin et al. recently reported that NRAS is phosphorylated on S89 by STK19. This PTM activates NRAS signaling and inhibitors of STK19 are effective in a mouse model of NRAS-driven melanoma [70].

Phosphorylation of KRAS4B in its HVR is the best characterized modification of this type. Bivona et al. found that PKC phosphorylated KRAS4B on serine 181, which weakened the electrostatic interaction of the polybasic HVR with the PM (Fig. 2) and allowed the GTPase to translocate to endomembranes in a process designated the farnesyl-electrostatic switch [71] that is analogous to the myristoyl electrostatic switch of the MARKS protein [72]. Importantly, phosphorylation of oncogenic KRAS4B on serine 181 was associated with decreased cell survival as a consequence of interference with the ability of Bcl-XL to modulate calcium flux between ER and mitochondria [73]. Wang et al. showed that phosphorylation of KRAS4B on serine 181 inhibits tumor initiation by blocking interaction with calmodulin and thereby abrogating suppression of non-canonical Wnt signaling [74]. Thus, phosphorylation of KRAS4B appears to be a negative regulatory event. One contrary result was reported by Alvarez-Moya who found that activated KRAS4B that could not be phosphorylated was impaired as an oncogene [75]. However, we have found that KRAS4B-12V,181A is as potent in vivo as KRAS4B-12V,181S in producing tumors (unpublished results) strongly arguing against a requirement for phosphorylation. Phosphorylation of KRAS4B on serine 181 also affects its ability to partition into membrane microdomains [76,77]. In addition to PKC, phosphorylation of serine 181 of KRAS4B was shown to catalyzed by cyclic GMP-dependent protein kinases 2 (PKG2) [78]. Recently, Iijima has made the remarkable discovery that KRAS4B is also phosphorylated on serine 181 by PKA downstream of insulin signaling and that phospho-KRAS4B combines with RHOA to regulate the mTORC2 complex (unpublished results).

In addition to serine/threonine phosphorylation, RAS proteins have also been reported to be phosphorylated on tyrosine 32 and 64 [79]. The non-receptor tyrosine kinase SRC has long been known to cooperate with oncogenic RAS in driving cell growth and tumorigenesis. In 2014, Michael Ohh and colleagues reported that SRC directly phosphorylates GTP-bound RAS proteins on tyrosine 32 and thereby reciprocally regulates binding to RAF1 and RasGAP to inhibit RAS signaling [80]. These authors went on to show that the ubiquitous tyrosine phosphatase SHP2 dephosphorylates RAS at position 32 and thereby potentiates signaling [81]. Most recently, this group found that tyrosine 64 was also phosphorylated by SRC such that the kinase modifies residues in both the switch I and II regions and that SRC and SHP2 cooperate to take RAS proteins in and out of a “dark state.” [82]. The availability of clinically useful SHP2 inhibitors makes this a very exciting model. However, because numerous investigators have interrogated the RAS/MAPK pathway for decades using anti-phosphotyrosine antibodies without observing phospho-RAS, these observations await confirmation.

RAS Cysteine Oxidation

The cellular redox state plays an important role in numerous signaling pathways that regulate cellular growth. When this balance is disrupted, aberrant proliferation can occur [83]. Indeed, protein modification by reactive oxygen species constitutes another class of PTMs that affect RAS. Redox homeostasis requires a balance of cellular oxidation and reduction. The reactive intermediates involved in these processes include reactive oxygen species (ROS), reactive nitrogen species (RNS), and reactive thiols. While cellular redox reactions are diverse and can generate many products, in this review, we will focus on regulation of RAS proteins by reversible thiol oxidation. The earliest studies on RAS modulation by ROS and RNS date back to 1995. Studies both in vitro and in T cells [84,85] support enhanced RAS activation by reactive intermediates. In vitro analyses indicated that incubation of ROS and RNS with recombinant RAS stimulated exchange of GDP for GTP, suggesting a mechanism for activation. The reaction site was a later shown to be a solvent accessible cysteine (Cys118) [84,86] within the NKCD nucleotide binding motif. Intriguingly, RAS activation by ROS and RNS occurred via a cysteine thiol radical intermediate rather than stable cysteine oxidation or nitrosation [86–90]. Reactive species that promote RAS Cys118 thiol radical intermediate led to oxidation of the guanine nucleotide and destabilization of nucleotide binding, which in turn stimulated guanine nucleotide exchange and RAS activation. These studies also led to identification of a variant of RAS, RAS C118S that retains RAS structure yet is redox insensitive [91]. The RAS C118S redox insensitive variant proved useful for a variety of cell-based studies aimed at discriminating direct or indirect effects of ROS/RNS on RAS activity. While RNS and ROS have been shown in many studies to regulate RAS activation and downstream signaling [92,93], a land mark study by the Counter lab, linked endothelial nitric oxide synthase (eNOS) to NO-mediated RAS activation and tumorigenesis in mice [94]. Oncogenic RAS activates the phosphoinositide 3-kinases (PI3K) pathway, which results in AKT-mediated eNOS phosphorylation. Lim et al. identified a feedback mechanism whereby eNOS-generated nitric oxide (NO) activates wild type RAS and stimulates downstream signaling. Consistent with their findings, introducing the RAS C118S mutation in HRAS and NRAS circumvented PI3K pathway activation and reduced tumor growth. In addition to RAS C118, three cysteines within the C-terminal hypervariable region RAS can undergo modification by RNS and ROS both in vitro and in mouse fibroblasts [93]. While these C-terminal cysteines may be protected from modification due to lipid PTM, redox conditions in which cysteine oxidation occurs at these sites could prevent lipid modification that facilitates membrane association and RAS activity. Hence, modification of these residues may downregulate RAS function by preventing proper membrane localization, in contrast to the upregulation of RAS by RNS/ROS modification of C118.

Whereas RAS is regulated by redox-driven PTMs, the reciprocal is also true in that RAS signaling modulates oxidant and anti-oxidant production and consequently redox balance. Oncogenic RAS can drive redox mediated tumorigenesis [95].

Lysine modifications (ubiquitylation, SUMOlyation, acetylation, methylation)

Ubiquitylation

RAS proteins are substrates for ubiquitination. Distinct ubiquitin linkages (mono-, di- and polyubiquitination) have been identified at multiple lysine residues in RAS [96]. The type and site of ubiquitin modification can alter RAS function by at least three different mechanisms; RAS subcellular localization, protein–protein interactions, and degradation (Fig. 3).

Figure 3. Three modes of RAS regulation by ubiquitination and SUMOylation.

A. RAS activity is regulated by ubiquitination. Primary sites of monoubiquitination occur at residues 147 in KRAS, whereas in HRAS it is at 117. Monoubiquitination of KRAS at 147 upregulates RAS activity through a GAP defect leading to enhanced MAPK activation. In contrast, monoubiquitination of HRAS at 117 induces fast exchange and activates RAS in a GEF-independent manner. All RAS isoforms undergo SUMOylation at Lys 42 which upregulates downstream signaling by an unknown mechanism. B. RAS localization is regulated by ubiquitination. Rabex-5 promotes mono- and diubiquitination of HRAS and NRAS resulting in endosome localization and reduced MAPK signaling. The deubiquitinase, OTUB1, removes ubiquitin from RAS and promotes plasma membrane localization and MAPK signaling. C. Ubiquitination by β-TrCP1 and SMURF2 promote RAS degradation through proteasome and autolysosomes resulting in reduced MAPK signaling.

The earliest studies examining the effects of RAS ubiquitination identified the RAB5 exchange factor, Rabex-5, as a key regulator of RAS function. Studies performed in both mammalian and Drosophila cells showed that Rabex-5, which contains both ubiquitin binding and E3 ligase domains, catalyzes mono- and diubiquitination of NRAS and HRAS but not KRAS [97,98]. Ubiquitination of either HRAS or KRAS protein promoted endosomal membrane localization and reduced MAPK signaling. It is unclear whether the ability of Rabex-5 to selectively ubiquitinate HRAS and NRAS but not KRAS in CHO-K cells is due to specific colocalization of Rabex-5 complexes with these RAS isoforms or through another mechanism. These findings indicate that Rabex-5 downregulates RAS function through directing ubiquitin-mediated relocalization of RAS to endosomes. However, the sites of ubiquitination that drive endosomal localization have yet to be established. It is also unclear whether mono- versus diubiquitination have distinct effects of HRAS and NRAS signaling and whether pathways other than MAPK are affected. On the flip-side, the deubiquitinase OTUB1 functions a negative regulator of RAS mono- and diubiquitination by sequestering RAS on the plasma membrane. OTUB1 is commonly overexpressed in non-small cell lung carcinoma and can promote RAS activation and tumorigenesis in wild- type RAS cells [99].

While these studies revealed a role of ubiquitination in localization and suppression of HRAS and NRAS function, KRAS monoubiquitination was found to alter RAS interactions with regulatory proteins and effectors and up-regulate its activity. In one study comparing KRAS to HRAS, sites of ubiquitination in RAS were identified by mass spectrometry following isolation of ubiquitinated KRAS and HRAS from HEK293T cells [100]. Lysine 147 was identified as the most frequent site of KRAS ubiquitination whereas lysine 117 was the most populated site in HRAS. Compared with unmodified KRAS, the ubiquitinated subpopulation was predominantly in the activated GTP-bound state and showed increased association with the downstream effectors RAF, PI3K, and RalGEF. Mutation of lysine 147 to prevent KRAS ubiquitination impaired tumor growth in a mouse xenograft model system. In vitro biochemical and pulldown studies suggested that KRAS monoubiquitination at 147 impairs GAP-mediated down-regulation and enhances nucleotide independent RAF binding [101]. In contrast, ubiquitin modification at lysine 117 in HRAS promotes RAS activation due to faster nucleotide cycling, rather than by disrupting GAP interactions [102]. Intriguingly, the covalently bound ubiquitin does not appear to make specific and stable intramolecular contacts with RAS, but rather transient interactions that modulate RAS structure and dynamics and consequently recognition by regulatory factors and downstream effectors [103,104]. This appears to be a novel mechanism of RAS activation. Taken together, these findings indicate that ubiquitination can activate RAS-mediated signaling and tumorigenesis.

The above studies support a role of mono- and diubiquitination in the regulation of RAS localization and protein–protein interactions. However, a more classically recognized mechanism by which ubiquitin mediates protein function is through degradation. Proteasomal degradation of HRAS has been reported to occur in response to activation by the Wnt/α-catenin signaling pathway [69,105,106] by recruitment of the E3 ligase β-TrCP after HRAS phosphorylation by glycogen synthase kinase 3β. Conversely, inhibition of this pathway enhances RAS expression levels and RAS-induced colorectal tumorigenesis. In an effort to develop anticancer drugs targeting RAS, small molecules have been identified that enhance formation of the α-catenin destruction complex to induce degradation of α-catenin and/or RAS [105,107–109]. In addition, evidence for KRAS isoform specific degradation has been observed [110]. The Smad ubiquitination regulatory factor 2 (SMURF2) can promote monoubiquitination of E2 ubiquitin-conjugating enzyme 5 (UBCH5) to form an active E3/E2 complex. This complex polyubiquitinates and degrades, in an activation-dependent manner, the β-TrCP1 E3 ligase that modulates KRAS protein stability [110]. Here, degradation appears dependent on lysosomal proteolysis, as a lysosome inhibitor protected KRAS from SMURF2, whereas proteasomal inhibition was ineffective, consistent with previous observations that KRAS can undergo lysosomal degradation [111]. Thus, both proteasome- and lysosome-mediated degradation mechanisms modulate KRAS levels and do so in a stimulus-dependent manner.

Recently, the leucine zipper-like transcription regulator 1 protein (LZTR1), which associates with the cullin 3 E3 ubiquitin ligase, has been postulated to regulate RAS. Although ubiquitination by LZTR1 was initially proposed to modulate RAS function by a nondegradative mechanism [*112,*113], a more recent study suggested that LZTR1-dependent proteolysis of RAS was prevented by treatment with a proteasome inhibitor, suggesting that LZTR1 facilitates polyubiquitination and degradation of RAS via the ubiquitin–proteasome pathway [*114] to modulate RAS protein levels. LZTR1 appears to downregulate RAS signaling, as LZTR1 haploinsufficiency in mice results in Noonan syndrome, a developmental disorder that arises from enhanced RAS and MAPK signaling. Moreover, loss of LZTR1 in Schwann cells promotes differentiation and proliferation [*113]. However, these findings have been challenged by Castel et al. who found that LZTR1 co-immunoprecipitates with RIT1 and MRAS but not HRAS, NRAS or KRAS [115]. Additional studies are needed to define the substrate specificity of LZTR1 and to determine whether LZTR1 directly regulates RAS function or affects RAS signaling indirectly.

RAS lysine SUMOylation

In addition to ubiquitination, RAS proteins are targets of SUMOylation [116,117]. While this small ubiquitin-like modifier (SUMO) is similar in size to ubiquitin, SUMOylation does not appear to initiate degradation. Rather, one well-known role of SUMOylation lies in regulation of cell growth, migration and tumorigenesis [118]. SUMOylation of RAS proteins occurs at a conserved lysine, Lys42, which is proximal to the effector region (residues 30–40). RAS modification by SUMO3 appears to be associated with activation as inhibition of SUMOylation or mutation of Lys42 results in downregulation of RAS signaling, reduced cell migration and invasion in multiple cell lines and tumor development in vivo. However, the mechanism of activation is unknown.

RAS lysine acetylation

Acetylation occurs through transfer of an acetyl CoA acetyl group by a cognate lysine acetyltransferase. Lysine acetylation modifies the lysine side chain, reduces the positive charge and can alter protein function and recognition. The best-studied example of this is the lysine modifications of histones that regulate chromatin. Acetylation of WT and G12V KRAS has been detected in the core GTPase domain at multiple lysine residues (K101, K104, K128 and K147) [119]. Two of these identified acetylation sites in RAS (K104 and K147) have been studied. Yang et al. generated a K104Q mutant to mimic acetylation and found that this substitution impaired SOS-mediated exchange [120]. This variant in the context of an oncogenic RAS G12V mutation showed reduced proliferation and clonogenic survival of NIH3T3 cells in a focus-formation assay relative to unmodified KRAS G12V. These findings indicated that KRAS K104 acetylation negatively regulates RAS function. The investigators [121] also found that knockdown of HDAC6 and SIRT2 reduced viability in NIH3T3 cells expressing K-RASG12V, but not in cells expressing KRASG12V/K104A, suggesting that these deacetylases modulate RAS acetylation. Using molecular dynamic (MD) simulations, they predicted that mutations at K104 disrupt the structure of helix 2 in switch 2 and consequently inhibited recognition by GEFs. However, subsequent NMR studies using the same K104Q mutant found only partial disruption of helix 2 [122], which explained both the GEF and GAP defects that were detected in biochemical assays. In contrast to previous findings, the KRAS G12V/K104Q variant did not significantly alter steady-state GTP levels, cellular growth or proliferation in NIH 3T3 cells, leading to the conclusion that the GEF and GAP defects were compensatory in nature, and the K104Q mutant likely did not impact overall Ras activity [122]. Moreover, K104-acetylated KRAS4B WT and G12V mutation did not alter SOS-catalyzed nucleotide exchange in vitro [119,122], indicating that glutamine substitution at K104 in RAS does not mimic acetylation. While both CBP and p300 lysine-acetyltransferases were able to acetylate both wildtype and G12V KRAS4B in vitro [119], Sirt2 and HDAC6 did not stimulate KRAS 104 deacetylation. Taken together, these somewhat conflicting studies indicate that the RAS K104Q variant does not mimic acetylation, and RAS acetylation may not alter RAS downstream function as originally proposed. RAS K147 acetylation has also been observed in KRAS4B, however, this modification does not alter intrinsic and the SOS-catalyzed nucleotide exchange, and the cellular consequences of acetylation at this site are unknown [119]. Given that intrinsic RAS function is not significantly altered upon acetylation K104 and K147, acetylation may serve as a docking site to promote new interactions. Additional studies are clearly needed to better understand the role of RAS acetylation in RAS-mediated signaling.

Lysine Methylation.

While lysine methylation has been primarily studied in the context of histone regulation, it is a critical PTM that can modulate the function of non-histone proteins [123]. Research in our lab and others have identified novel methylation sites (Lys5, Lys147) within the core RAS G-domain, but the functional role of these modifications is unclear. In contrast to acetylation, methylation does not alter the side chain charge and it is challenging to deduce its impact on protein structure by conventional or unnatural amino acid substitutions. Methylation of conserved lysine residues (Lys5, Lys16 and Lys117) in the core GTPase domain of RAS superfamily GTPases have been identified through an algorithm developed by the Sasaki lab called GoMADScan [124], suggesting that methylation at these sites may play a conserved role in GTPase function. To follow up on these observations, endogenous RAS was immunoprecipitated from HEK293T cells, and mass spectrometric analysis conducted. RAS residues, Lys5 and Lys147 were found to be di- and monomethylated. Lysine 147 methylation was unique to HRAS and the GoMADScan algorithm did not detect methylation at the equivalent position to Lys147 in other small GTPases [124]. Given that substitutions at Lys147 to alanine, cysteine, or leucine do not significantly alter RAS activity [102], it is tempting to speculate that Lys147 methylation may function to create a docking site or block other PTMs (acetylation, ubiquitination), rather than alter RAS structure and intrinsic function. Mutations at Lys5 have been identified in cancers and RASopathies [125–131], however, it is unclear how these mutations upregulate RAS function. To assess the impact of Lys5 dimethylation on RAS structure and dynamics, molecular dynamics (MD) simulations were conducted, which predicted that dimethylation of Lys-5 does not significantly alter RAS conformation. This suggests that Lys-5 methylation may alter existing protein interactions, create a docking site to foster new interactions or other PTMs at this position to modulate RAS function.

Conclusion

RAS is a highly regulated signaling node that plays a key role in controlling cellular growth. Dysregulation of accessory proteins (e.g. GEFS and GAPs) and point mutations in RAS can alter RAS activity and cause aberrant signaling and pathologies, resulting in RASopathies and cancer. PTMs represent yet another layer of RAS regulation. In comparison to GEFs and GAPs, modulation of RAS function by PTMs is poorly understood. Nevertheless, it is clear that a diverse array of often reversible PTMs can regulate RAS function by controlling membrane association and subcellular trafficking, that active state of RAS, its protein-protein interactions, and its expression levels. To better understand the role of various PTMs in RAS function, it will be important to identify the sites of modification in a site-specific manner and the enzymes responsible for both the PTMs and their reversal. While advances in mass spectrometry have identified a subset of PTMs and the sites of modification, PTMs can be transient and present at low levels thereby evading MS detection. Conversely, MS is sensitive enough to report PTMs that lack physiological significance. Indeed, one of the great challenges in the study of RAS PTMs has been to establish the stoichiometry of the modifications that must be relatively high to affect processes such as protein-protein interaction is a way to have significant impact. Another challenge has been how to study the effects of PTMs when they cannot be genetically encoded. PTMs are often probed by mutating the site of modification to block or mimic a modification. However, it has become increasingly apparent that blocking mutations may alter the structure of the protein or adjacent modification sites and that mimetic substitutions may not truly mimic the PTM of interest. While synthetic biology can unambiguously replicate some PTMs in vitro, the ability to express or modify proteins in cells remains limited. Extension of genetic, biochemical, and synthetic biological approaches are needed to better understand RAS PTM.

RAS has proven difficult to target for drug discovery due to the lack of druggable pockets and exceedingly high affinity of guanine nucleotides. Nevertheless, despite over 30 years of being dubbed an ‘undruggable’ target, progress has recently been made on this front, as covalent inhibitors of an oncogenic RAS cysteine containing mutant are showing efficacy in clinical trials. However, because the KRAS G12C mutant is only one of over 100 oncogenic RAS mutations found in various RAS-driven cancers, additional therapeutic intervention approaches are needed. Given that PTMs can regulate RAS activity by a variety of mechanisms, enzymes that control RAS PTMs may be viable targets. Indeed, early efforts to antagonize oncogenic RAS signaling relied on inhibition of farnesyltransferases that target RAS to the membrane, an approach that has recently been revived for HRAS-driven head and neck cancer [132].

Enzymes in the native RAS ubiquitination pathways represent another class of targets. In addition, engaging E3 ligases that do not normally target RAS to create neomorphic RAS degraders is another approach that co-opts PTMs. These include both proteolysis targeting chimeric small molecules (PROTACS) [**133,134] and genetically encoded degraders [*135–137]. Palmitoyl acyltransferases, palmitoyl thioesterases, carboxyl methyltranferases, aaX endoproteases, kinases, phosphatases, acetyltransferases, deacetylases and methyltransferases may represent a rich tapestry of additional targets.