Ras GTPases are important regulators of pathways controlling proliferation, differentiation and transformation. Three ubiquitously expressed almost identical Ras genes are not functionally redundant; this has been attributed to their distinctive trafficking and localization profiles. A palmitoylation cycle controls the correct compartmentalization of H-Ras and N-Ras. We review recent data that reveal how this cycle can be regulated by membrane organization to influence the spatiotemporal signalling of Ras.
Ras proteins are low-molecular-mass GTPases that mediate signalling emanating from cell-surface receptors. Oncogenic mutations of Ras at codons 12, 13 and 61 are found in 16% of human cancers and the majority of all cancers display aberrant activation of Ras signalling pathways . Key effector cascades are the PI3K (phosphoinositide 3-kinase)–Akt pro-survival and the Raf–MEK [mitogen-activated protein kinase/ERK (extracellular-signal-regulated kinase) kinase]–ERK proliferative pathways. Although understanding Ras function is important from a human health perspective it has also emerged as a model for studying fundamental questions about membrane biology and the context-dependence of isoform-specific signalling. The reason for this is the presence in all cells of three closely related isoforms: H-Ras, K-Ras and N-Ras, which despite a high degree of identity are not functionally redundant [2,3]. Differences include the observation that only K-Ras is essential for normal mouse development, the preponderance of K-Ras mutations compared with the other isoforms in a wide range of cancer types and distinct signalling profiles associated with each isoform in overexpression studies [4–7].
Examination of the amino acid sequences of the Ras isoforms reveals that all of the regions required for GDP/GTP binding and effector interactions are essentially identical (Figure 1). This means that the key interactions which regulate Ras activity and determine downstream signalling should be indistinguishable. The only region of significant difference is the C-terminal HVR (hypervariable region) that contains isoform-specific membrane binding and trafficking determinants. Consequently, each Ras isoform shares overlapping but distinct localizations at the plasma membrane and on intracellular organelles. This may enable access to different pools and concentrations of activators and effectors. Notably, the dynamic nature of Ras compartmentalization was proposed to generate tunable isoform-specific signalling.
Ras HVRs contain membrane binding and trafficking determinants
Trafficking of Ras isoforms
All Ras proteins are synthesized on free ribosomes in the cytosol before undergoing a series of post-translational modifications that enable them to stably interact with membranes. The cysteine residue in the C-terminal CAAX box of all Ras isoforms is farnesylated to allow weak membrane binding. ER (endoplasmic reticulum)-localized Rce1 cleaves off the AAX motif before the farnesylated cysteine residue is methylated by ICMT (isoprenyl-cysteine-carboxymethyltransferase) . A second signal is required to stabilize Ras membrane interactions, and also promote, in some cases, trafficking to the plasma membrane . For the K-Ras4B splice variant of K-Ras this second signal is a polybasic domain comprising six lysine residues that electrostatically interact with membrane phospholipids. A minimal 15-amino-acid sequence of the K-Ras tKs (targeting domains) is sufficient to promote plasma membrane trafficking of GFP (green fluorescent protein) via a Golgi-independent route [10,11]. Recent work has suggested that PDE6δ (phosphodiesterase 6δ) acts as a cytosolic chaperone shielding the hydrophobic farnesyl group of K-Ras during transport . Membrane binding is reversible and K-Ras is constantly sampling membranes with a half-life of a couple of minutes .
The other Ras isoforms are all palmitoylated either once or twice on cysteine residues in the HVR by the yeast palmitoyltransferase Erf2 or mammalian DHHC9 [14,15] (Figure 1). Based on observations with H-Ras, palmitoylation in conjunction with the C-terminal farnesylated signal was thought to be sufficient for promoting stable plasma membrane localization. A minimal ten-amino-acid sequence comprising the C-terminal tH (targeting region of the H-Ras HVR) attached to GFP traffics normally to the plasma membrane via the Golgi apparatus . However, the monopalmitoylated Ras isoforms, N-Ras and the K-Ras4A splice variant, both require additional basic/hydrophobic motifs within the HVR to traffic to the plasma membrane . In the absence of these motifs a minimal monopalmitoylated sequence comprising the tN (targeting region of the N-Ras HVR) stably associates only with the Golgi apparatus. As no change was observed in the palmitoylation half-life upon removal of the basic/hydrophobic motif, it was speculated that this motif regulates the promotion of more stable insertion of the palmitoyl group into the membrane to facilitate membrane trafficking .
Importantly, palmitoylation is reversible; APT1 (acyl protein thioesterase 1) cleaves the thioester bond thus solubilizing the Ras proteins . Radioactive palmitate kinetic analysis indicated a half-life of N-Ras palmitoylation of 20 min . This means that over the life-time of the N-Ras proteins there will be >100 cycles of palmitoylation/depalmitoylation. Re-palmitoylation of Ras occurs in the Golgi apparatus; therefore, a unidirectional cycle between the cell surface and Golgi is established by this dynamic modification [19–21]. This flux is critical for ensuring correct localization of the palmitoylated Ras proteins; when the APT1 inhibitor palmostatin B or constructs impervious to depalmitoylation were used, palmitoylated Ras localized to all membranes within the cell [17,20]. Therefore all Ras isoforms display dynamic, regulated and relatively short-lived interactions with membranes that modulate their correct localization to the cell surface and on endomembranes.
Membrane nanoclustering and regulated membrane interactions
Upon reaching the cell surface Ras isoforms partition into distinct signalling nanoclusters. These are sub-20-nm protein and lipid assemblies consisting of five to ten copies of Ras with life-times of 0.1–0.5 s [22–24]. The clustered distributions of each Ras isoform display specific combinations of actin-, cholesterol- and galectin-dependence . Importantly, these associations are dynamic and are regulated by the activation state of the Ras protein. H-Ras localizes to both cholesterol-dependent and cholesterol-independent nanoclusters; however, on activation its cell-surface distribution is shifted to purely cholesterol-independent clusters that rely on galectin-1 for their integrity [23,26]. In contrast, N-Ras moves into cholesterol-dependent nanoclusters upon activation [27,28]. K-Ras clustering is independent of cholesterol but relies on galectin-3 when GTP-bound . Therefore all three Ras isoforms when activated occupy non-overlapping signalling domains.
The two palmitoyl groups of H-Ras play different roles in nanoclustering and localization, with the palmitoyl at Cys181 (equivalent to the N-Ras palmitoyl group) favouring localization in cholesterol-dependent clusters [27,30]. The potential mechanisms underpinning the dynamic association of palmitoylated Ras proteins with different nanoclusters has been provided by molecular dynamics simulations [31,32]. The HVR linker domain between the G-domains (residues 1–166) and the C-terminal targeting domain contains basic lysine and arginine residues that interact with the membrane when H-Ras is in the inactive GDP-bound conformation. Upon GTP binding, Ras undergoes a conformational change that brings arginine residues in the α4 helix of the G-domain into contact with the membrane. This alters the orientation of the G-domain and HVR with respect to the membrane, and is predicted to influence the depth of insertion of the palmitoyl moieties so that they are more exposed in the GTP-bound conformation . This change in membrane interactions facilitates galectin-1 binding to stabilize localization outside of cholesterol-dependent nanoclusters . It is tempting to speculate that it will also facilitate interaction with APT1 resulting in depalmitoylation and solubilization of the H-Ras protein. A further function of the altered orientation is to specify isoform-specific signalling, since K-Ras adopts a different orientation than H-Ras. These differences are recognized by Raf and PI3K, which favour the K-Ras and H-Ras orientations respectively .
We previously investigated the role of nanocluster localization in the regulated depalmitoylation of N-Ras using a combination of live-cell fluorescence microscopy and high-resolution electron microscopy imaging [27,28]. FRAP (fluorescence recovery after photobleaching) using two different beam sizes for bleaching membrane-associated proteins allows discrimination between fluorescence recovery by lateral diffusion in the membrane or direct exchange with molecules in the cytoplasm [26,33,34]. Metabolic depletion of cholesterol resulted in a decrease in clustering and an increase in the lateral mobility of N-Ras. This was more pronounced for the GFP–N-Ras (G13V) constitutively active mutant than the wild-type protein, corroborating previous observations that N-Ras association with cholesterol-dependent raft domains increases upon its activation [27,28]. Intriguingly, antibody-mediated clustering (patching) of an extracellular raft protein, the HA–GPI (glycosylphosphatidylinositol-anchored influenza haemagglutinin), shifted N-Ras recovery to exchange from the cytosol, but only when N-Ras was also activated. Clustering of fibronectin receptors by fibronectin, a natural ligand for endogenous raft-resident integrins, reproduced this effect. These effects were inhibited by cholesterol depletion or palmostatin B. Notably, the activated N-Ras molecules that dissociated from the plasma membrane accumulated in the Golgi apparatus, where re-palmitoylation occurs . Taken together, these results indicate that clustering of raft domains facilitates depalmitoylation of raft-resident active N-Ras, increasing its exchange rate and enhancing its accumulation in and signalling from the Golgi compartment.
The FRAP beam-size analysis studies demonstrated that clustering of raft-associated proteins (e.g. HA–GPI) shifts the FRAP recovery mode of constitutively active GFP–N-Ras (G13V) from lateral diffusion to exchange . To further explore the underlying molecular mechanism, a novel quantitative full-fitting FRAP analysis, which measures simultaneously the lateral diffusion and the plasma membrane binding/unbinding rate constants (Kon and Koff) of non-integral membrane proteins , was employed. This analysis (Figure 2) shows that both Kon and Koff of N-Ras (G13V) increase dramatically (2–3 orders of magnitude) following HA–GPI clustering. Comparison of the characteristic times for recovery by lateral diffusion in the plasma membrane (τD=ω2/4DM, where ω is the Gaussian radius of the laser beam and DM is the lateral diffusion coefficient in the membrane) and by exchange [(τex=1/(Kon+Koff)] yields values of τex/τD=380 and τex/τD=0.22 before and after HA–GPI clustering respectively (Figure 2; for derivation of equations see ). These results demonstrate that whereas patching of HA–GPI (raft clustering) has only a minor effect on the lateral diffusion (DM) of N-Ras (G13V), it dramatically alters the membrane-interaction kinetics of the activated N-Ras. They change from conditions where τD is much smaller than τex (resulting in recovery by diffusion, with a negligible contribution by the much slower exchange) to conditions where τex is significantly smaller than τD (i.e. recovery mainly by exchange).
Full-fitting FRAP kinetics analysis shows that the plasma membrane exchange rate of GFP–N-Ras (G13V) is dramatically enhanced by HA–GPI clustering
The plasma membrane is the main location for Ras function. Occupancy of signalling nanoclusters enables high-fidelity signal transmission across the membrane by operating as an analogue–digital–analogue circuit where each nanocluster can be considered as a switch-like digital unit. The graded stimulation provided by extracellular ligands results in activation of all of the Ras proteins within a nanocluster, which then generates a proportional analogue activation of intracellular cytosolic cascades . Although not formally investigated, a second function of nanoclusters may be to determine access to specific pools of effectors and regulators that favour association with lipid and protein environments unique to each type of isoform nanocluster. In this way isoform-specific coupling to downstream effectors may be achieved.
Activation-dependent changes in membrane insertion of the lipid moieties result in dynamic exchange between nanoclusters for H-Ras and N-Ras that is essential for observing a full effector activation profile . The consequent depalmitoylation that we observed in the context of active N-Ras and clustered raft domains results in solubilization and re-localization of activated N-Ras to the Golgi complex . Although in situ activation of Ras in the Golgi complex has been observed in other systems , we found that inhibition of depalmitoylation resulted in loss of the resident Golgi pool of active Ras. Furthermore, use of inhibitors of the in situ PLC (phospholipase C) and calcium-dependent pathway for activation of Ras in the Golgi failed to inhibit the raft-clustering-induced accumulation of GTP-loaded N-Ras in the Golgi . These findings indicate direct cytosolic transit of the GTP-loaded N-Ras to the Golgi in response to clustering of raft-associated proteins. This alternative mode of generating a pool of active Ras in the Golgi compartment is analogous to that seen previously in EGF (epidermal growth factor)-stimulated MDCK (Madin–Darby canine kidney) cells .
The functional consequences of Ras signalling from the Golgi apparatus are contentious. However, in lymphocytes, endogenous Golgi Ras signalling has been shown to be important for positive thymocyte selection . Under conditions where N-Ras redistribution to the Golgi complex was promoted, we observed a significant decrease in short-term plasma membrane ERK signalling but more potent ERK activation at longer (60 min) time points, typical of signalling from the Golgi complex .
The Ras isoforms exhibit dynamic localizations modulated by post-translational modifications that have important signalling consequences. Our recent data have revealed a novel co-stimulatory effect between Ras activation state and plasma membrane organization that regulates depalmitoylation and redistribution of N-Ras, depending on its activation state . The functional consequences of the switch from short-term signalling at the plasma membrane to long-term signalling from the Golgi remain to be determined.
Regulation of Protein Trafficking and Function by Palmitoylation: A Biochemical Society Focused Meeting held at St Anne's College, Oxford, U.K., 23–25 August 2012. Organized and Edited by Luke Chamberlain (University of Strathclyde, U.K.) and Tony Magee (Imperial College London, U.K.).
We gratefully acknowledge funding from the Royal Society, North West Cancer Research Fund, the Wellcome Trust (to I.A.P.) and from the DGF-DIP [grant numbers KL 1948/1-1 and GA 309/10-1] and the Ministry of Science & Technology, Israel (to Y.I.H.). Y.I.H. is an incumbent of the Zalman Weinberg Chair in Cell Biology.