Regulating the Regulator: Post-Translational Modification of Ras

CAAX processing: prenylation, proteolysis and methylation

Ras is a member of a large class of proteins known as CAAX proteins. The CAAX sequence is modified by three enzymes that work sequentially⁴³. First, unmodified CAAX sequences serve as substrates for prenylation by one of two cytosolic prenyltransferases^{44, 45}. If the amino acid in the X position is leucine⁴⁶, as is the case for most Rho family GTPases and γ subunits of heterotrimeric G proteins, then geranylgeranyltransferase type I (GGTase I) adds a 20-carbon polyisoprene lipid to the CAAX cysteine via a stable thioether bond. Geranylgeranylation of Rho proteins not only allows them to associate with membranes, upon which they exert their biological effects, but it also promotes their association with a cytosolic chaperone, RhoGDI. Because RhoGDI possesses a hydrophobic pocket into which the geranylgeranyl lipid is sequestered it can regulate the trafficking of Rho proteins on and off membranes⁴⁷. If the amino acid in the X position of CAAX is not leucine⁴⁶, as is the case for all Ras proteins, then farnesyltransferase (FTase) modifies the CAAX cysteine with a 15-carbon farnesyl lipid (FIG. 2b). Farnesylation affords Ras proteins relatively weak affinity for cellular membranes⁴⁸. Several farnesyl binding proteins analogous to RhoGDI, such as PDE^™49, have been described that may facilitate the trafficking of farnesylated proteins.

Proteins such as Ras that are modified only with a farnesyl lipid accumulate on the cytoplasmic face of the endoplasmic reticulum¹¹, where they encounter the next CAAX processing enzyme, Ras converting enzyme type 1 (Rce1). Ras prenylation is a prerequisite for the action of Rce1, not only because it is required for the co-localization of Ras and Rce1 on the ER, but also because of the substrate specificity of this protease. Rce1 is an endoprotease that removes the AAX amino acids so that the farnesylcysteine is the new C-terminus⁵⁰. Ras is then modifed by another ER resident enzyme, known as isoprenylcysteine carboxylmethyltransferase (Icmt), which catalyzes the methyl esterification of the α-carboxyl group of the farnesylcysteine⁵¹. Of these three modifications, only the one catalyzed by Icmt is reversible (FIG. 2b). The end result of these three modifications is the remodeling of the C-terminus of Ras proteins from a hydrophilic region to a hydrophobic one that is capable of insertion into cellular membranes, which is a requirement for Ras biological activity.

There is no evidence that CAAX processing is regulated with respect to cell activation, metabolic status or cell cycle, suggesting that it is a housekeeping process. One caveat is the recent description of the binding of a splice variant of smgGDS, a GEF that is active against a wide range of small GTPases, to unprenylated Ras and thereby retarding its prenylation, suggesting a mode of regulation⁵². Two classes of drugs can block prenylation of Ras and other proteins: statins that inhibit HMG-CoA and therefore limit the availability of farnesyl pyrophoshate, which is an intermediate in the cholesterol biosynthetic pathway initiated by HMG-CoA reductase, and farnesyltransferase inhibitors (FTIs) that directly inhibit farnesylation. A pool of unprocessed, endogenous Ras in the absence of prenylation-inhibiting drugs has not been described. Ras processed by the three sequential PTMs of the CAAX sequence can be distinguished from unprocessed Ras by SDS-PAGE because the processed form has a slightly faster electrophoretic mobility. The unprocessed form of endogenous Ras can be observed in SDS-PAGE only after treatment with statins or FTIs^{53, 54}. Concordant with these results, whereas endogenous Ras is readily detected in the cytosolic fraction of cells treated with statin or FTI, the vast majority of Ras is in the membrane fraction of untreated cells (M. Zhou and M.R.P, unpublished observation). These observations argue strongly against a pool of unprocessed Ras that is awaiting processing, and suggest that CAAX processing follows translation immediately in an efficient and constitutive manner. Nevertheless, the CAAX processing pathway can be saturated since, when Ras is overexpressed, a significant pool is unprocessed based on the results of electrophoretic mobility and subcellular fractionation (M. Zhou and M.R.P, unpublished observation).

Although CAAX processing is necessary for the delivery of Ras proteins to, and stable association with, the plasma membrane, it is not sufficient for these events. Elements within the HVR that are upstream of the CAAX sequence⁵⁵ are also required. These elements have been called “second signals” for plasma membrane targeting. There are two types of second signal: one consists of cysteines that serve as acceptor sites for palmitoylation and the other consists of polybasic regions rich in lysines. N-Ras, H-Ras and K-Ras4A have the former second signal and K-Ras4B has the latter (FIG 2a).

Palmitoylation as a second signal

The covalent attachment of the acyl chain of a fatty acid to a protein is known as protein acylation⁵⁶. Although acyl chains of a variety of lengths have been shown to be incorporated into proteins, the 14-carbon myristoyl chain and the 16-carbon palmitoyl chain are the most commonly utilized. Whereas myristoylation is largely restricted to the N-termini of proteins, palmitoylation of cysteine residues occurs throughout the polypeptide chain. The palmitoylation of Ras was first described 25 years ago⁵⁷, and it was later appreciated that this PTM is isoform specific⁵⁸. Whereas H-Ras has two cysteines that can accept palmitate (residues Cys181 and Cys184), N-Ras only has one (residue Cys181). Although the best studied splice variant of K-Ras, K-Ras4B, does not undergo palmitoylation, the other splice form, K-Ras4A, is palmitoylated at Cys180⁵⁹. Palmitate is linked to Ras via a labile thioester bond and the modification is readily reversible⁶⁰.

A palmitoyl acyltransferase (PAT) capable of modifying Ras was recently identified as the DHHC9–GPC16 complex⁶¹, which is an ortholog of the yeast Erf4–Erf2 complex⁶². DHHC9–GPC16 is a member of a family of 25 PATs that all contain a DHHC motif but differ in their subcellular localization and substrate specificity⁶³. Whereas all members of this family have a DHHC motif, most do not have binding partners analogous to GPC16. Knockdown studies suggest that DHHC9–GPC16 is not the only PAT with activity toward Ras⁶⁴. DHHC9 and other DHHC motif-containing members of this family of enzymes are transmembrane proteins. Thus, CAAX processing of Ras, which affords the protein some affinity for membranes, would appear to be a prerequisite for palmitoylation. However, Ras mutants in S. cerevisiae that lack the CAAX cysteine can still be palmitoylated, demonstrating that prenylation is not absolutely required for palmitoylation, at least in yeast⁶⁵.

Palmitoylation is required for the trafficking of N-Ras and H-Ras from the endomembrane system to the plasma membrane^{11, 66} (FIG 3). Palmitoylation of Ras takes place on the cytosolic face of the Golgi apparatus where DHHC9–GPC16 resides. Whereas farnesylated Ras has modest affinity for membranes, Ras that is both farnesylated and palmitoylated has more than 100 fold higher affinity^{67, 68} and therefore palmitoylation of Ras at the Golgi serves as an affinity trap for the protein. The high affinity binding to membranes of dually lipidated proteins promotes their subcellular trafficking via vesicular transport.

The polybasic region as a second signal for K-Ras4B

K-Ras4B is unique among Ras proteins in that it cannot be palmitoylated. Nevertheless, as is the case for all CAAX proteins, it still requires a second signal for trafficking to the plasma membrane⁵⁵. The K-Ras4B second signal consists of a lysine-rich domain in the HVR that has a net positive charge of 8. Thus, although no PTM is required to create this second signal, the motif is considered here because it complements farnesylation and can itself be modified by a PTM (see below). The polybasic region forms an electrostatic interaction with the negatively charged headgroups of the inner leaflet of the plasma membrane. Neither this electrostatic interaction alone nor the insertion of the farnesyl group on Ras into the phospholipid bilayer provides sufficient affinity for stable membrane association, but the two interactions together provide sufficient affinity for relatively stable membrane association. Although the steady-state distribution of K-Ras4B appears to be predominantly at the plasma membrane, molecular trapping studies in live cells have revealed that the interaction of K-Ras4B with the plasma membrane is dynamic⁶⁹.

Palmitoylation–depalmitoylation cycling

Unlike farnesylation, palmitoylation is readily reversible under physiologic conditions. Indeed, it was shown more than 25 years ago that the half-life of palmitate on Ras was considerably shorter than the half-life of the protein⁶⁰. More recently it was found that N-Ras and H-Ras undergo an acylation–deacylation cycle that is linked to Ras trafficking to and from the Golgi apparatus^{70, 71}(FIG 3). Anterograde trafficking from the Golgi to the plasma membrane requires palmitoylation and proceeds via vesicular transport^{11, 66}. The localization of DHHC9–GPC16 on the Golgi and the fact that dually lipidated proteins are affinity trapped by membranes supports this model. Photobleaching^{70, 71} and photoactivation⁷⁰ studies have revealed that retrograde trafficking of Ras from the plasma membrane to the Golgi requires depalmitoylation and is too rapid to occur via vesicular trafficking. These observations support the current model in which N-Ras and H-Ras are palmitoylated on the Golgi apparatus and thereby affinity trapped in a membrane compartment, transported to the plasma membrane on vesicles and, after a certain period of time, depalmitoylated there and released back into the cytosol (FIG 4). From the cytosol the Ras proteins can diffuse back to the Golgi for another round of palmitoylation and a return to the plasma membrane. Evidence that Ras signaling from the Golgi apparatus differs from Ras signaling from the plasma membrane in terms of relative downstream pathway utilization^{12, 13, 72–74} and, in the case of T lymphocytes, in biological outcome³¹, supports the idea that the acylation/deacylation cycle is a way to modulate signaling.

There is no evidence that palmitoylation of Ras by DHHC9–GPC16 or any other PAT is regulated, suggesting that, like farnesylation, palmitoylation is an immediately post-translational “housekeeping” modification, although this is an under-investigated area. In contrast, there is evidence that depalmitoylation is regulated by GTP-loading of H-Ras⁷⁵. Thus, elucidation of the mode of regulation of depalmitoylation rests, to some extent, on identifying the mechanism of palmitate removal. The thioester linkage between palmitate and its substrates is quite labile, and one school of thought holds that depalmitoylation may not be enzymatic. Another school holds that there are one or more Ras specific palmitoyl thioesterases that catalyze the hydrolysis reaction. One candidate is acyl protein thioesterase 1 (APT1)⁷⁶. However, this protein has many other substrates and is found in the cytosol, raising the question of how it could have access to the palmitate modifications of membrane-bound Ras⁷⁷. A recent, highly innovative study used semisynthetic Ras proteins with cleavable or non-cleavable acyl modifications and came to the conclusion that, although Ras could only be palmitoylated on the Golgi, depalmitoylation can occur anywhere in the cell⁶⁴. This suggests a model whereby the entropy that would otherwise distribute Ras over all intracellular membranes is overcome by the restricted localization of palmitoylation on the Golgi and unidirectional vectorial vesicular transport ⁶⁴.

Peptidyl-prolyl isomerization

A recent study has implicated FKBP12, a cis-trans prolyl isomerase, in the regulation of Ras depalmitoylation⁷⁸ (FIG 4). The extent of H-Ras palmitoylation was shown to depend on the presence or absence of a proline at position 179 in the HVR. FK506 and other chemical inhibitors of the prolyl isomerase activity of FKBP12, including cycloheximide and rapamycin, inhibited H-Ras depalmitoylation and this effect was recapitulated by silencing FKBP12. In addition, H-Ras bound to FKBP12 in a palmitoylation-dependent fashion. Together these observations suggest that isomerization of the Gly–Pro peptidyl-prolyl bond at position 178–179 of H-Ras regulates depalmitoylation and thereby constitutes a molecular timer for acylation. Cis-trans isomerization about this bond, which accelerates Ras depalmitoylation, is catalyzed by FKBP12. Thus, peptidyl-prolyl isomerization is the most recently recognized Ras PTM. Although originally thought to only play a role in the proper folding of nascent proteins, cis-trans isomerization of peptidyl-prolyl bonds is increasingly recognized as a mechanism for signaling, particularly when a molecular timer is required⁷⁹.

Phosphorylation and the farnesyl-electrostatic switch

K-Ras4B was first shown to be phosphorylated by protein kinase C (PKC) at its C-terminus in 1987, although no biochemical effect of this modification was reported⁸⁰. Recently, the PKC phosphorylation site was mapped to serine 181, which is positioned within the polybasic region⁸¹. Phosphorylation is stimulated by calcium ionophore, implicating a conventional PKC as the relevant kinase, although which member(s) of this class of kinase are involved in regulating K-Ras4B has not been determined. The physiologic stimulus that induces K-Ras4B phosphorylation also remains to be elucidated. Phosphorylation of serine 181 reduces the net charge of the polybasic region, causing K-Ras4B to lose affinity for the plasma membrane and to accumulate on endomembranes. This mechanism is reminiscent of the myristoyl-electrostatic switch that regulates the membrane association of the myristoylated alanine-rich C-kinase substrate (MARCKS) protein⁸², and therefore it has been called the farnesyl-electrostatic switch (FIG 5).

K-Ras4B is not the only small GTPase that has a farnesyl-electrostatic switch; the Rho family GTPase Rnd3 is also regulated in this fashion⁸³. Furthermore, RalA, a Ras family GTPase, is phosphorylated on serine 194 in its HVR by Aurora A kinase and this causes it to translocate to mitochondria where it regulates mitochondrial fission ⁸⁴. Interestingly, when K-Ras4B has an activating mutation, relocation to endomembranes as a consequence of the farnesyl-electrostatic switch is associated with cell death⁸¹. The mechanism for the cell toxicity is unclear, but, paradoxically, the anti-apoptotic protein Bcl-XL is required⁸¹. The farnesyl-electrostatic switch of oncogenic K-Ras4B might be exploited in the development of anti-cancer drugs. Dissociation of K-Ras4B from the plasma membrane was also observed in neurons stimulated with glutamate, and a mechanism involving Ca⁺⁺/calmodulin binding to the K-Ras polybasic region was proposed⁸⁵. Because calmodulin binds to the polybasic region of K-Ras4B and other proteins through a calcium-regulated electrostatic interaction, phosphorylation of serine 181 would be expected to diminish calmodulin binding. Thus, under conditions in which the farnesyl-electrostatic switch is engaged by the phosphorylation of K-Ras4B on serine 181, calmodulin is unlikely to play a role in causing the dissociation of K-Ras4B from the plasma membrane.

Ubiquitination

H-Ras,N-Ras and K-Ras4B were recently shown to be substrates for mono- and diubiquitination^{86, 87}. The conditions under which Ras becomes ubiquitinated have not been determined. Nor have all of the ubiquitin acceptor sites been mapped, although, based on mutagenesis studies, several surface exposed lysines are candidates⁸⁶. Mass spectrometry of affinity purified, dual epitope-tagged H-Ras from cells expressing tagged ubiquitin revealed mono- and di-ubiquitination on lysines 117, 147 and 170⁸⁷. The E3 ligase responsible for these modifications was identified as Rabex-5⁸⁸. Furthermore, the Ras effector RIN1 is required for Rabex-5-dependent Ras ubiquitination⁸⁸. RIN1, through its function as a Rab5 GEF, is thought to promote the Rab5-GTP-dependent recruitment of Rabex-5 to endosomal sites where H-Ras ubiquitination takes place. Whereas the presence of H-Ras that could not be ubiquitinated was rare on endosomes, H-Ras that was artificially and constitutively ubiquitinated was enriched on endosomes. Thus, ubiquitination regulates the trafficking of H-Ras to and from endosomes. Importantly, the ubiquitin-mediated endosomal restriction of H-Ras was also associated with a reduction in MAPK signaling, demonstrating that ubiquitination is yet another PTM that regulates Ras compartmentalization and the spatial control of its signaling output. K-Ras4B was recently found to be monoubiquitinated on lysines 104 and 147⁸⁷. This modification led to enhanced GTP loading and, in the setting of an oncogenic V12 mutation, increased the affinity of K-Ras4B for the downstream effectors Raf-1 and PI3K.

S-Nitrosylation

Cysteine 118 is highly conserved among Ras isoforms and orthologs and it is the most surface exposed cysteine on these GTPases. This residue was shown to be a redox switch more than fifteen years ago⁸⁹, and subsequent studies revealed that cysteine 118 could be nitrosylated when exposed to nitric oxide (NO)⁹⁰. The mechanism of nitrosylation was later shown to be a direct interaction of the thiol of cysteine 118 with either a nitrogen dioxide radical (•NO₂) formed when NO interacts with O₂, or with a glutathionyl radical (GS•)⁹¹. S-nitrosylation does not alter the structure of Ras but leads to enhanced guanine nucleotide exchange^{92, 93}, which promotes more efficient Ras activation. This has led to the speculation that Ras signaling can be modulated by redox reactions, although the physiologic relevance of Ras S-nitrosylation remains to be determined.

Bacterial toxins and exoenzymes

The Ras superfamily of small GTPases is a favorite target of bacterial virulence factors that evolved as enzymes that modify eukaryotic signaling molecules. Most such toxins and exoenzymes target the Rho family of small GTPases that regulate the actin cytoskeleton⁹⁴. However, there are also bacterial enzymes that are active against Ras. Exoenzyme S (ExoS) of Pseudomonas aeruginosa is an ADP-ribosyl transferase that modifies Ras. ExoS adds ADP-ribose to arginines 41 and 128 of Ras, leading to inefficient guanine nucleotide exchange⁹⁵. Lethal toxin (LT) of Clostridium sordelli is a monoglucosyltranferase that uses UDP-glucose as a substrate and glucosylates Ras on threonine 35. Toxin B of Clostridium difficile has a similar activity towards Rho proteins⁹⁶. Glucosylation of Ras inhibits signaling to MAPK⁹⁷.

Whereas nascent Ras requires stoichiometric CAAX processing for maturation as a membrane protein, it is now clear that the mature protein can be modified in ways that modulate subcellular trafficking and signaling (FIG 6). Although these modifications are conditional in the sense that pools of Ras exist in both the modified and unmodified forms, the conditions in which these PTMs are stimulated, the stoichiometry of the modifications and their biological significance remain to be elucidated.

Ras trafficking

CAAX processing and palmitoylation of Ras were originally thought to be a way to convert a globular hydrophilic protein into one with a hydrophobic C-terminus that has generalized affinity for cell membranes. The discovery that two of the CAAX processing enzymes, Icmt and Rce1, are polytopic membrane proteins restricted to the ER revealed that Ras trafficking is more complex than originally believed^{51, 98, 99}.

Our current understanding of Ras trafficking begins in the cytosol where nascent Ras is synthesized on free polysomes (FIG. 3). FTase, the first and rate limiting, of the CAAX processing enzymes, is cytosolic ¹⁰⁰. Once farnesylated, Ras gains modest affinity for membranes, particularly those of the ER. How farnesylated Ras is transferred from the cytosolic FTase to membranes is a matter of speculation. Are there chaperones involved? If so, does a single chaperone or family of chaperones both deliver farnesylated Ras to Icmt and Rce1 at the ER and also retrieve the fully CAAX processed product? The existence of and physiologic role of RhoGDI, a geranylgeranyl-binding chaperone for Rho family GTPases⁴⁷, has led Ras biologists to speculate that an analogous chaperone may exist for farnesylated proteins. Several proteins have been identified in yeast-two hybrid screens as farnesyl-binding proteins, including PDE6δ¹⁰¹, PRA1¹⁰² and Galectin¹⁰³. Among these proteins, PDE6δ is particularly intriguing because its farnesyl binding domain bears structural similarity to RhoGDI and because it was recently reported that its capacity to bind Ras and other farnesylated proteins is regulated by its interaction with the non-prenylated small GTPases Arl1 and Arl2 ⁴⁹. Moreover, a recent study reported that PDE6δ is required for proper trafficking of Ras and other farnesylated proteins as well as for efficient Ras signaling. Another intriguing possibility is that FTase itself serves as the chaperone, since it was found to have an unusual mode of catalysis; it has the ability to retain its product in a secondary exit surface near the active site¹⁰⁰.

Whether or not chaperones plays a role in Ras trafficking, the first membrane compartment visited by nascent Ras is the ER. The basis for the particular affinity of Ras for ER membranes is not universally accepted nor understood, but it may relate to the biophysical properties of this compartment and the branched, unsaturated nature of the polyisoprene lipid that is added by farnesylation. Some investigators have argued that GFP-tagged prenylated proteins appear on the ER simply because the ER constitutes by far the greatest reservoir of membrane in the cell⁶⁴. However, without a second signal, GFP-tagged Ras and other CAAX proteins appear neither on the plasma membrane nor on other abundant membrane-bound organelles, such as mitochondria¹¹, calling into question the notion of non-specific membrane association. A particular affinity of Ras for the ER makes teleological sense since it is here that the next two enzymes in the CAAX processing cascade (Rce1 and Icmt) reside. Indeed, it is conceivable that Rce1 itself helps attract proteins with farnesylated CAAX sequences to the ER and thereby accounts for the observed membrane localization.

Once CAAX processing is complete, Ras isoforms diverge in their subsequent trafficking⁴³. Palmitoylated isoforms visit the cytoplasmic face of the Golgi where they are acylated and thereby affinity trapped in Golgi membranes from where they can be incorporated into transport vesicles and enter the secretory pathway. K-Ras4B cannot be trapped on the Golgi and proceeds to the plasma membrane either by passive diffusion or by an as yet uncharacterized delivery system that could involve cytosolic chaperones. Recent observations have revealed that the plasma membrane is not the end of the road for palmitoylated Ras or K-Ras4B. The former undergoes retrograde trafficking, which is mediated by the acylation cycle described above^{70, 71}, and the latter undergoes retrograde trafficking from the plasma membrane to endomembranes upon phosphorylation and engagement of the farnesyl-electrostatic switch⁸¹.

Ras is also trafficked to endosomes by two routes, one from the plasma membrane and one from the Golgi. Ras that is associated with clathrin-coated regions of the plasma membrane is internalized into primary endosomes during clathrin-mediated endocytosis. As clathrin-coated membrane domains also contain the activated protein tyrosine kinase receptors that signal through Ras and Raf-1, all of these elements are present on these so-called ‘signaling endosomes’ and signaling can persist¹⁰⁴. Indeed, in some contexts, endocytosis is absolutely required for Ras–MAPK signaling¹⁰⁵. Recently, signaling endosomes were differentiated from their non-signaling counterparts by expression of the APPL1 adaptor protein¹⁰⁶. APPL1 directly binds to Rab5, protein tyrosine kinase receptors and the phospholipid bilayer, thus serving as a scaffold for signaling complexes that include Ras. The association of APPL1 with endosome membranes was dependent on low PtdIns(3,4,5)P₃ levels, suggesting that phosphoinositide remodeling interconverts signaling and non-signaling endosomes¹⁰⁶. A second route to endosomes, which is taken by palmitoylated Ras, was recently described, in which N-Ras and H-Ras traffic from the Golgi to recycling endosomes and then on to the plasma membrane¹⁰⁷.

There remains some disagreement as to whether K-Ras4B associates with, and signals from, endosomes like its palmitoylated cousins. Several groups reported that, whereas N-Ras and H-Ras were associated with endosomes, K-Ras4B was not^{86, 108, 109}. These data are consistent with the observation that diubiquitination promotes the association of Ras with endosomes but, whereas all three Ras isoforms examined can be mono-ubiquitinated, only N-Ras and H-Ras are di-ubiquitinated⁸⁶. More recently, K-Ras4B was reported to associate with early endosomes in a clathrin-dependent fashion and then to traffic to late endosomes, leaving H-Ras behind¹¹⁰. Moreover, K-Ras4B was found to reside on, signal from and undergo degradation in late endosomes, lysosomes and multivesicular bodies¹¹⁰. Clearly we have more to learn about Ras trafficking through the endosomal compartment.

Microdomain affinity and nanoclusters

Ras trafficking does not end at the plasma membrane because this organelle is not homogeneous but rather is composed of several types of microdomains¹¹¹. Like its trafficking between organelles, the partitioning of Ras into membrane microdomains is regulated by PTMs. The best-characterized microdomain is a cholesterol-dependent, liquid-ordered domain often referred to as a “lipid raft.” Once thought to encompass relatively large patches of membrane analogous to the liquid-ordered domains observed in artificial bilayers¹¹², cholesterol-dependent domains are now known to exist on a nanoscale, to be highly dynamic and to incorporate on the order of 10 or fewer signaling molecules such as Ras^{113, 114}. In addition to lipid rafts, at least two cholesterol-independent microdomains have been detected, and these are also thought to be highly dynamic and on a nanoscale level¹¹¹. Microdomains serve to bring together multiple signaling components. In addition, nanoclusters have been shown to convert analog Ras–MAPK signaling into a digital signal that improves fidelity¹¹⁵.

Palmitoylation appears to be one of the strongest determinants of protein association with lipid rafts. Therefore, it is not surprising that H-Ras, N-Ras, and K-Ras4B occupy distinct plasma membrane microdomains¹¹⁴. Inactive, GDP-bound H-Ras has been found to group in nanoclusters of ~6 molecules within ordered, cholesterol-rich microdomains. Interestingly, exchange of GDP for GTP causes H-Ras to lose affinity for the lipid raft domains and to migrate laterally into adjacent regions of disordered plasma membrane^{116, 117}. Coincident with this migration is its interaction with a scaffolding protein, Galectin-1, which acts to enrich GTP-bound H-Ras in non-ordered microdomains^{118, 119}. Surprisingly, although the final 10 amino acids of H-Ras, which encompasses the CAAX sequence and both palmitoylation sites, is sufficient to direct plasma membrane localization, it cannot fully recapitulate the dynamic behavior of full-length H-Ras with regard to plasma membrane nanoclusters and microdomains^{114, 120}. Instead, additional elements within the HVR and even the G-domain help stabilize the membrane orientation and microdomain preference of full length H-Ras at the plasma membrane and thereby contribute to differential effector engagement^120–122.

Interestingly, the GTP binding status of N-Ras also affects microdomain partitioning, but for this isoform it is the GTP-bound form that favors partition into cholesterol-dependent nanoclusters¹²³, which in turn controls signal output. Thus, with regard to plasma membrane microdomains, mono- versus dipalmitoylation appears to differentially control localization. Moreover, whereas H-Ras that is mono-palmitoylated on cysteine 181 behaves like N-Ras, H-Ras that is mono-palmitoylated at cysteine 184 does not traffic efficiently to the plasma membrane¹²³. Thus, it appears that the spacing between the prenyl and acyl modifications is also critical for proper trafficking and microdomain partitioning. The linker region comprising amino acids 170–179 within the H-Ras HVR has also been shown to specifically contribute to the association of H-Ras within non-raft microdomains¹²⁰.

K-Ras4B partitions into cholesterol-independent microdomains that do not overlap with those into which GTP-bound H-Ras partitions^{114, 124}. Interaction with the Galectin-3 scaffold enhances K-Ras4B nanocluster formation¹²⁵. Phosphorylation of GTP-bound K-Ras4B on serine 181 reduces nanoclustering¹²⁶. Thus, like palmitoylation, phosphorylation plays a role in nanocluster formation.

Generating a comprehensive model for how the C-terminus of Ras proteins interacts with phospholipid bilayers is confounded by the inability to produce crystals of Ras that include its C-terminus. All Ras structures have been solved from crystals of Ras proteins that lacked a C-terminus, which is thought to be disordered¹²⁷. However, recent molecular dynamic simulations and NMR data collected from full-length farnesylated H-Ras in a 1,2-dimyristoylglycero-3-phosphocholine bilayer have provided computational models of membrane insertion, and revealed putative contributions of both the HVR and G-domains in this process^128–130. In these models the GDP-bound conformation of H-Ras favors electrostatic interactions between basic residues within its HVR and the inner leaflet of the plasma membrane. In contrast, residues within the α-4 helix of the G-domain facilitate the membrane association of GTP-bound, active, H-Ras, control its orientation with respect to the membrane and thereby contribute directly to effector binding specificity^{131, 132}. Moreover, in these models the depth of membrane insertion of palmitates was greater when H-Ras was GTP-loaded, perhaps explaining the activation-specific association of H-Ras with Galectin-1¹²⁹.