Ras, an Actor on Many Stages

Abstract

Among the wealth of information that we have gathered about Ras in the past decade, the introduction of the concept of space in the field has constituted a major revolution that has enabled many pieces of the Ras puzzle to fall into place. In the early days, it was believed that Ras functioned exclusively at the plasma membrane. Today, we know that within the plasma membrane, the 3 Ras isoforms—H-Ras, K-Ras, and N-Ras—occupy different microdomains and that these isoforms are also present and active in endomembranes. We have also discovered that Ras proteins are not statically associated with these localizations; instead, they traffic dynamically between compartments. And we have learned that at these localizations, Ras is under site-specific regulatory mechanisms, distinctively engaging effector pathways and switching on diverse genetic programs to generate different biological responses. All of these processes are possible in great part due to the posttranslational modifications whereby Ras proteins bind to membranes and to regulatory events such as phosphorylation and ubiquitination that Ras is subject to. As such, space and these control mechanisms act in conjunction to endow Ras signals with an enormous signal variability that makes possible its multiple biological roles. These data have established the concept that the Ras signal, instead of being one single, homogeneous entity, results from the integration of multiple, site-specified subsignals, and Ras has become a paradigm of how space can differentially shape signaling.

Introduction

Ras family GTPases are included in a large superfamily of small GTP-binding proteins that act as molecular switches, regulating the flow of intracellular signals by cycling between an inactive, GDP-bound and an active, GTP-bound state. The “classical” Ras proteins (H-Ras, N-Ras, K-Ras4A, and K-Ras4B) are a key control node for multiple biological functions, including proliferation, differentiation, apoptosis, survival, motility, and adhesion, among many others.¹ The importance of Ras proteins in cell physiology is highlighted by the dramatic results of their deregulation in some pathological conditions: Activating mutations in Ras are detected in about 30% of human cancers (Catalogue Of Somatic Mutations In Cancer [COSMIC]; available online at http://www.sanger.ac.uk/genetics/CGP/cosmic/).

Despite sharing a common set of regulators, both guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs), and interacting with and activating identical effector molecules, a wealth of animal and cell biology studies, complemented by clinical observations, indicates that the 4 Ras isoforms may not be completely redundant in their functions. This can be clearly envisioned from their distinctive implication in human carcinogenesis: Activated mutations in K-ras are primarily associated with tumors of the lung, colon, pancreas, and biliary track; H-ras mutations are predominant in oral and urinary track tumours, whereas N-ras mutations are related to leukemias and skin neoplasias.² Such differences are evident even at the very early stages of carcinogenetic processes, where the main Ras isoforms (H-Ras, N-Ras, and K-Ras4B) seem to play different roles. In transformed fibroblasts, N-Ras regulates adhesion, whereas K-Ras4B (K-Ras hereafter) coordinates motility,³ and in the colonic epithelium, K-Ras stimulates hyperproliferation, whereas N-Ras prevents apoptosis,⁴ and in pluripotent cells, only K-Ras can trigger stem cell expansion to initiate endodermal tumors.⁵

From the regulatory point of view, there also exist relevant differences among Ras isoforms. In vivo studies demonstrate that the GEF RasGRF1 is mostly active on H-Ras,^6,7 whereas Smg-21GDS preferentially catalyzes nucleotide exchange on K-Ras.⁸ Similarly, the GAP NF1 is 4 times more efficient in down-regulating H-Ras than N-Ras.⁹ Substantial differences are also found in their effector functions: K-Ras is the most potent activator of Raf-1, followed by N-Ras and H-Ras.^10,11 Conversely, H-Ras is the most potent activator of PI3K.¹⁰ RASSF2 is specifically activated by K-Ras,¹² and Rac-1 is differentially activated by H-Ras and K-Ras.¹³

It would be hard to find an explanation for these dissimilar behaviors based on structural divergences, as all Ras isoforms are identical within their G-domains, in particular the Switch I and II regions, which dictate the interactions with regulators and downstream targets.¹⁴ Some of these functional differences among Ras isoforms began to be understood when the concept of space was introduced into the equation. Today, we know that within the cell, the 3 Ras isoforms occupy different microlocalizations, as a consequence, in great part, of particularities in their posttransductional processing and trafficking. Thus, space can, distinctively and ultimately, shape Ras isoforms’ functions by introducing variability in the availability, accessibility, and functionality of their common sets of effectors and regulators and by bringing into play novel localization-specific control mechanisms. This review focuses on the recent advances in our understanding of the spatial regulation and functions of Ras signals.

Many Places for a Ras

Seminal studies identified Ras as a peripheral membrane protein bound to the inner leaflet of plasma membranes (PMs), an essential requisite for its functionality.¹⁵ More recently, resulting from Mark Philips’s pioneer studies, it was found that significant pools of Ras proteins were also present and active in endomembranes, including the endoplasmic reticulum (ER), the Golgi complex (GC),^16,17 early and late endosomes,^18-20 and lysosomes²¹ (for a review, see Fehrenbacher et al.²²). The 3 Ras isoforms could also be detected in cytoplasmic organelles such as mitochondria.^7,23,24 An additional level of complexity was introduced following the obsolescence of the concept of the PM as a uniform, homogeneous structure, to be substituted by the notion of the PM as a mosaic of domains of different composition and physicochemical properties. These include the following: Lipid rafts are tightly packed domains enriched in cholesterol and glycosphingolipids. Caveolae are a subtype of lipid rafts containing caveolin family proteins. Disordered membrane domains correspond with high-fluidity membrane regions rich in unsaturated phospholipids. These PM features have been thoroughly reviewed elsewhere.²⁵ Following this new line of thought, it was detected that, within the PM, Ras isoforms segregated differently at these distinct microlocalizations: H-Ras was present at lipid rafts, both of caveolar and noncaveolar nature, and was also detected in disordered membrane domains. Likewise, N-Ras was also found in caveolin-positive and caveolin-negative raft domains.^7,26 By contrast, K-Ras was preferentially located in nonraft, disordered PM,^27-29 although, probably depending on the cell type, small amounts could also be found in caveolae.²⁶ Hitherto, the spatial distribution of K-Ras4A remains largely unstudied.

Ras segregation appears to extend further down to the “nano” scale, as within microdomains, such as the disordered membrane, H-Ras and K-Ras4B are not intermixed but cluster into separate, spatially distinct microenvironments, as small as 6 nm, referred to as “nanoclusters.”^28,29 The formation of these nanostructures requires the interaction with galectins, a family of pleiotropic carbohydrate-binding proteins that, in this case, serve a scaffold role. Galectin-1 is recruited to the PM in response to H-Ras activation,³⁰ inducing the formation of activated H-Ras/galectin-1 complexes³¹ that become enriched in cholesterol-independent microdomains.³² Antagonizing galectin-1 results in the displacement of H-Ras from the PM and in the inhibition of Ras biological activity.³³ Galectin-3 functions in a similar way, although specifically for K-Ras.³⁴ The formation of K-Ras but not H-Ras nanoclusters is also dependent on actin, and the disassembly of the actin cytoskeleton prevents K-Ras signals.³⁵ It is very likely that the concept of the PM being formed of distinct microdomains will extend to endomembranes in the future, so it is not unlikely that Ras proteins in endomembranes also partition further in different microlocalizations. Indeed, the GC harbors two distinct microenvironments with different capabilities for sustaining C-Raf-induced differentiation of PC-12 cells.³⁶ What makes Ras isoforms segregate differently in these membrane environments? To a great extent, such dispersion is dictated by the posttranslational modifications suffered by the Ras isoforms in the process to become active proteins and other regulatory modifications such as phosphorylation and ubiquitination.

Ras Processing

Ras proteins are synthesized as hydrophilic proteins in free ribosomes in the cytoplasm and have a half-life of 24 hours.³⁷ To be active, these proteins have to associate with the cytosolic leaflet of membranes.¹⁵ Since they are not intrinsic membrane proteins, for this purpose, they need to undergo several posttranslational modifications within their C-terminus, including prenylation, proteolysis, and carboxyl methylation (for a review, see Wright and Philips³⁸). Seminal studies revealed that Ras proteins were palmytoylated.³⁹ This notion was clarified further when it was later demonstrated that all Ras proteins were isoprenylated, but only some isoforms were palmitoylated.⁴⁰ The initial step in Ras maturation is a farnesylation reaction, a process in which a farnesyl-transferase enzyme transfers a 15-carbon isoprenyl group from a farnesyl phosphate to cysteine 186, conserved in all Ras isoforms, within the C-terminal CAAX box (where C is cysteine, A is an aliphatic amino acid, and X is any amino acid).⁴¹ This modification forces the nascent Ras proteins to transiently associate with the ER. At this organelle, the endopeptidase Ras-converting enzyme-1 (Rce1) cleaves the C-terminal -AAX residues.^42,43 Finally, the alpha-carboxyl group of the newly C-terminal prenyl-cysteine becomes methylated by the isoprenylcysteine carboxyl methyl-transferase Icmt,⁴⁴ also at the ER.⁴⁵ K-Ras and N-Ras can also be modified by geranylgeranylation, a process that becomes particularly active when farnesylation is inhibited (for a review, see Konstantinopoulos et al.⁴⁶). These modifications increase Ras hydrophobicity and therefore its capability to associate to phospholipid bilayers, but this is not enough to stably attach Ras to cellular membranes. To this end, a second modification is required.

H-, N-, and K-Ras4A are further modified by palmitoylation in cysteine residues within the hypervariable (HVR) region, immediately adjacent to the CAAX box.⁴⁷ Palmitoylation is the reaction by which a palmitoyl group is added to a cysteine residue through a thioester bond. Ras palmitoylation takes place in the GC. In Saccharomyces cerevisiae, it is undertaken by the enzyme Erf2/4.⁴⁸ In mammalians, the farnesyl-transferase activity on Ras is mediated by its ortholog DHHC9/GCP16, resident at the GC.^49,50 Whereas prenylation alone serves as a relatively weak membrane anchor, palmitoylation markedly increases Ras affinity for membranes and enables their tethering to the peripheral PM.¹⁶ Palmitoylation is essential for H- and N-Ras association to the PM; proteins rendered unpalmitoylable by mutations in the modified residues, cysteine 181 in N-Ras and cysteines 181 and 184 in H-Ras, cannot be transported to the PM and are therefore retained in endomembranes.^16,51

In the case of K-Ras4B, the second membrane-binding signal is provided by the presence in its HVR of a polybasic domain formed by 6 lysine residues. This poly-lysine motif provides a net positive charge in K-Ras4B C-terminus, enabling an electrostatic interaction with the negatively charged PM phospholipids.^47,52 However, such an interaction is not permanent as it has been shown that PM K-Ras4B is in a dynamic equilibrium with a cytoplasmic pool.⁵³

Ras Trafficking to the Plasma Membrane

Once fully processed, H- and N-Ras travel to the plasma membrane through the classical secretory pathway, in vesicles generated at the GC. Indeed, blockade of vesicle formation in this organelle, either by low temperature (16°C) or by treatment with brefeldin A, results in the accumulation of H- and N-Ras in this endomembrane.¹⁶ Initial studies by John Hancock’s laboratory⁵⁴ demonstrated that N-Ras monopalmitoylation on cys181 was sufficient for its translocation from the GC to the PM. In the case of H-Ras, the palmitoylations on cysteines 181 and 184 affected differently its trafficking. Monopalmitoylation on cys184 was not sufficient for localizing H-Ras to the PM, and this form remained at the GC. On the other hand, palmitoylation of H-Ras on cys181 was required and sufficient for efficient trafficking of H-Ras to the PM. As such, monopalmitoylated H-Ras cys181 behaved like N-Ras on its trafficking to the PM.⁵⁴

A new twist has emerged recently with the identification of recycling endosomes as a midway stop en route from the GC to the PM. After exiting the GC, H- and N-Ras are transported to recycling endosomes before reaching the PM (Figure 1). Interestingly, the different palmitoyl groups also play different roles in this process. Only bipalmitoylated H-Ras can associate to recycling endosomes. In the absence of a palmitoyl group on cys184, cys181 H-Ras is directly transported to the PM, in agreement with Hancock’s results. Counterintuitively, N-Ras, also bearing a single palmitoyl group at cys181, can have access to recycling endosomes.⁵⁵ This can be explained on the basis that in N-Ras, leucine 184 provides an additional membrane anchor, supported by hydrophobic interactions between the leucine side chains and the membrane phospholipids.^56,57 Indeed, a mutant H-Ras C184L could efficiently localize to recycling endosomes.⁵⁵ These results suggest that, in addition to Ras posttranslational modifications, also the nature of the recipient membrane could have a significant impact on Ras localization. K-Ras4B does not follow the same path toward the PM; its polybasic domain directs K-Ras through a different route, possibly mediated by microtubules. Indeed, prenylated K-Ras exhibits high affinity for binding to microtubules.⁵⁸ When vesicle transport is inhibited, K-Ras transport is not affected; on the other hand, disorganizing the microtubule network (e.g., by using placlitaxol) disrupts K-Ras access to the PM.^16,35

Figure 1.

Ras trafficking routes among subcellular compartments. Ras proteins, synthesized in cytoplasmic polysomes, are farnesylated and bind to the endoplasmic reticulum (ER), where they are processed (dark blue arrows). Therein, H-Ras and N-Ras are palmitoylated and subsequently translocated to the Golgi complex (GC) and to recycling endosomes from where they are transported, via vesicular transport, to their final destination in plasma membrane (PM) lipid rafts and disordered membrane microdomains (light blue arrows). K-Ras4B is transported by an unknown mechanism, probably mediated by microtubules to disordered membrane microdomains at the PM (brown arrow). At the PM, H-Ras can diffuse between lipid rafts and disordered membrane microdomains depending on its activation state (red arrow). At the PM, H-Ras and N-Ras are depalmitoylated and sent back to the GC, closing the acylation cycle (light blue arrows). The trafficking mechanism is unclear and could involve RhoGDI-like chaperones. H-Ras and N-Ras can also be internalized through the endocytic pathway via endosomes (dashed orange arrow). At the PM, K-Ras4B can be phosphorylated by protein kinase C (PKC), dislodging it from the PM to endomembranes, particularly mitochondria (green arrow). Also at the PM, H-Ras and N-Ras can be ubiquitinated, promoting their association to endosomes (purple arrow). Calmodulin binding to K-Ras4B at the PM brings about its recruitment to endosomes and the GC (orange arrow).

Ras Acylation Cycle

Originally, mature Ras proteins were thought to remain bound to the PM until degraded. However, it was also known that the half-lives of H- and N-Ras were longer than those of their palmitoyl groups, 24 hours³⁷ versus 20 minutes and 2 hours, respectively.⁵⁹ Unlike farnesylation, which is stable, palmitoyl lipids are linked through a labile thioester bond, making it a reversible process. Thus, the idea of a palmitoylation/depalmitoylation cycle taking place gained foot. FRAP was used to demonstrate that palmitoylated Ras isoforms at the PM were de-palmitoylated therein and trafficked from the PM back to the GC via a nonvesicular route. When de-palmitoylated, H- and N-Ras accumulated in the GC, a new palmitoylation process needed to take place to enable them to regain access to the PM.⁶⁰ Such a cycle was later verified using Ras proteins tagged with photoactivatable fluorescent probes, showing that PM-bound H- and N-Ras trafficked back to the GC, getting enriched therein after they had been depalmitoylated. Since N-Ras required a single depalmitoylation event, it cycled faster and was more abundant in the GC than in the H-Ras, which must be depalmitoylated on 2 cysteines.⁶¹ Such a cycle provides the cell with a highly accurate protein sorting system and a safeguard against spillage to undesired localizations as the de/reacylation cycle will continuously reset the system. The removal of palmitoyl groups is mediated by acyl-thioesterases.⁵⁰ Two acyl-thioesterases have been described thus far, PPT-1 and APT-1, the latter being responsible for the elimination of the palmitoyl group of proteins located at the inner leaflet of the PM, including, at least, the H-Ras isoform.^62,63 Treatment with palmostatin B, an inhibitor of APT-1, reduces H- and N-Ras levels at the GC, fostering their enrichment at the cellular periphery.⁶⁴

An aspect still unclear about Ras retrograde transport is the mechanism whereby depalmitoylated, but still farnesylated, Ras proteins traffic “in solution” back to the GC.⁶⁰ This is consistent with initial studies demonstrating that large amounts of Ras were present in the cytosol,¹⁶ including farnesylated forms.⁴⁴ One possibility is that H- and N-Ras traffic back to the GC, escorted by chaperones analogous to Rho-GDIs, enclosing the lipid moiety and shielding it from the aqueous phase. Several of these farnesyl Ras binding proteins have been described, including PRA1,⁵³ galectin-1,⁶⁵ and PDE δ.⁶⁶ Conversely, other in vitro studies have demonstrated that monolipidated peptides can transit between bilayers within seconds without requiring shuttle factors.⁶⁷ Thus, the issue is still open and awaits clarification.

Similar to H- and N-Ras, K-Ras has also been shown to traffic in a retrograde fashion. In hippocampal neurons following stimulation with glutamate, K-Ras was released from the PM and accumulated in the GC and endosomes by a mechanism dependent on calcium/calmodulin.⁶⁸ Recent studies using FRAP have shown that the continuous cycle between the periphery of the cell and inner membrane systems has profound implications on the nature of Ras signals, by propagating Ras activity between subcellular compartments as in a “radio transmission,” thereby generating a “carrier wave” of Ras signaling.⁶⁹

Palmitoylation regulates not only H-Ras transport between PM and endomembranes but also its lateral segregation and distribution between different PM microdomains. Palmitoylation at cys181 is sufficient to target H-Ras to PM lipid rafts, but once there, monopalmitoylated cys181 H-Ras is locked in this compartment, with palmitoylation at cys184 being necessary for H-Ras segregation to nonraft microdomains following GTP loading.^54,61 Intriguingly, at the PM, monopalmitoylated cys181 H-Ras emulates N-Ras in its interaction with this type of membrane,⁵⁴ unlike recycling endosomes that permit the association of N-Ras but not of monopalmitoylated cys181 H-Ras.⁵⁵ This once again strongly suggests that, in addition to the lipid anchors, the composition of the targeted membrane must play some role in determining Ras distribution. In this respect, using model membranes, it has been demonstrated that prenylated peptides have very little affinity for liquid-ordered domains such as lipid rafts, whereas multi-palmitoylated peptides show significant partitioning to such domains.⁷⁰ However, spacing between palmitates or between a palmitate and the prenyl group is also a factor to be taken into account when evaluating membrane affinity. This is exemplified by the fact that H-Ras monopalmitoylated at cys181 exhibits affinity for lipid rafts domains, whereas H-Ras monopalmitoylated at cys184, probably too close to prenylated cys186, does not.⁵⁴ It is likely that the proximity of these two bulky lipid anchors causes an intolerable distortion on the structure of highly ordered PM domains such as lipid rafts, and its insertion therein is prevented.

Ras lateral displacement between different microdomains is also dependent on its activation status, as demonstrated by Hancock and collaborators. When inactive, GDP-loaded H-Ras resides mainly at lipid rafts, but once activated, GTP-bound H-Ras segregates to disordered membrane.²⁷ This result implies that Ras Switch I and/or Switch II domains, the only regions that alter their conformation depending on the bound nucleotide, must have some role in membrane binding. Another possibility is that GDP/GTP binding somehow also affects the conformation of the C-terminus in such a way that membrane association is affected. Indeed, nuclear magnetic resonance (NMR) studies of full-length, farnesylated H-Ras show that the farnesyl moiety, the HVR, and residues 23-30 within the G-domain contribute to effector interactions,⁷¹ suggesting that membrane binding could also be affected by the Ras active/inactive state. Since we still lack structural information on the lipidated Ras C-terminus, this is an open possibility. Interestingly, the rate of Ras depalmitoylation has been shown to be dependent on Ras activation state: GTP-bound H-Ras is depalmitoylated at a significantly faster rate than inactive GDP-loaded H-Ras.⁷² As such, it is tempting to speculate that disordered membrane microdomains could represent locations where Ras depalmitoylation is actively taking place. Unlike palmitoylation that, thus far, has been detectable only in GC, depalmitoylation seems to occur everywhere in the cell,⁷³ but this does not discard that significant differences could exist in depalmitoylation activity at different sublocalizations, in different cell types. In this respect, Hancock’s model of Ras activation-driven diffusion between PM microdomains, postulated in BHK cells, awaits confirmation as a widespread mechanism among different cellular contexts.

Even though lipid modifications are essential for Ras binding to and trafficking between membranes, some regions within Ras itself also play critical roles in these processes. Such is the case for the HVR in H-Ras; two regions in the linker domain within its HVR are required for H-Ras segregation from lipid rafts to disordered membrane following activation. Deletion of amino acids 166-172 or 173-179 confines H-Ras to cholesterol-rich domains and prevents its downstream signaling.⁷⁴ Apparently, for all palmitoylated Ras isoforms, acidic residues within the HVR are required to stabilize palmitoylation, and basic residues are likely to interact electrostatically with PM phospholipids, thereby contributing to stabilize Ras localization to membranes.⁷⁵ Interestingly, operating against the attractive interactions of the palmitate anchors and the HVR toward lipid rafts, there is a repulsive force generated by the N-terminal 165 amino acids, comprising the Ras catalytic domain, that increases when Ras is GTP loaded, pushing H-Ras toward disordered membrane domains.⁷⁶ In this respect, the basic residues in the helix α4, within the HVR, play a critical role in stabilizing GTP- or GDP-H-Ras interactions with the membrane by a conformational switch that reorients the G-domain in order to enhance association to effector molecules when in the GTP-bound form.⁷⁷

Ras Trafficking by Phosphorylation

Phosphorylation is another posttranslational modification that can have a significant impact on Ras localization and traffic and therefore on its functions, although contrary to the acylation cycle that affects H and N-Ras, in this case, the regulated isoform is K-Ras. In the early days, protein kinase C (PKC) was shown to phosphorylate K-Ras at serine 181, in its polybasic region, in response to phorbol ester stimulation.⁷⁸ More recently, it was demonstrated that PKC-mediated phosphorylation triggers a significant translocation of oncogenic K-Ras from the PM to internal localizations such as the ER, GC, and mitochondria. Apparently, phosphorylation serves to partially neutralize the positive charge of the K-Ras polylysine region, reducing its electrostatic attraction with phospholipids and promoting its release from the PM. At mitochondria, K-Ras induces apoptosis by some mechanism mediated by the proapoptotic protein BCL-X_L.⁷⁹ However, a large pool of K-Ras still remains at the PM.^80,81 Therein, phosphorylation inhibits the formation of GTP-bound K-Ras nanoclusters, preventing its downstream signaling,⁸¹ an effect that could also contribute to apoptosis. Interestingly, PKC-mediated effects on K-Ras can be antagonized by calmodulin. Calmodulin binds to the polybasic region of K-Ras^68,82 and prevents phosphorylation of ser181 within this region by PKC.⁸³ Noticeably, unphosphorylated K-Ras is more susceptible to p120GAP-mediated inactivation.⁸⁴ Interestingly, calcium/calmodulin binding also redirects K-Ras to internal membranes such as endosomes and the GC.⁶⁸ It is also remarkable that PKC-mediated phosphorylation of ser181 prevents calmodulin binding.⁸⁰ Thus, there exists a marked antagonism between PKC and calmodulin, the result of which can modulate K-Ras4B localization and functions. In a similar fashion to other scaffolds, calmodulin could be promoting the formation of K-Ras4B microdomains where K-Ras4B phosphorylation would be prevented, making it more sensitive to p120GAP, leading to an attenuation in K-Ras4B signals.

Ras Trafficking by Ubiquitination

Lipidic additions and phosphorylations are not the only means by which Ras intracellular trafficking is regulated. It has been recently shown that, in mammalian cells, H-Ras and N-Ras but not K-Ras are subject to mono- and lys63-linked diubiquitination, a process that promotes their association with endosomes.⁸⁵ Ubiquitination is determined by the Ras C-terminus, suggesting that it could be site specified, by restricting the enzymatic machinery that undertakes this task to defined cellular sublocalizations. Interestingly, Ras ubiquitination is regulated by the interaction between the Rab family GEF Rabex-5 and the Ras effector Rin,⁸⁶ suggesting some kind of feedback mechanism whereby Ras activation can regulate its own ubiquitination. Although indirectly, ubiquitination can also influence Ras localization by interfering with its processing. The deubiquitinizing enzyme USP17 down-regulates the activity of RCE1, a process that takes place in the ER. As a consequence, Ras cleavage is blocked and its subsequent membrane localization and activation impeded.⁸⁷ Intriguingly, USP17 blocks H-Ras and N-Ras but not K-Ras membrane trafficking.⁸⁸

Although it is unquestionable that the aforementioned posttranslational modifications are essential factors for Ras localization and trafficking, it must be envisioned that the Ras final destination will not be defined solely by these, but rather by the intimate relationship established between Ras intrinsic localization determinants and the substrate they are interacting with: membranes. Thus, the composition and physicochemical characteristics of the membranes on which Ras proteins will anchor will significantly affect Ras distribution. In this respect, in model membranes, it has been shown that the proportion of acidic phospholipids markedly influences the binding affinity of prenylated and polybasic peptides.⁸⁹ Also in model membranes, K-Ras binds poorly to vesicles composed of neutral lipids, and its affinity increases exponentially as increasing amounts of anionic lipids are incorporated.⁹⁰ Moreover, it has been show that the electrical properties of PM can modulate the distribution of K-Ras: inhibition of polyphosphoinositide synthesis causes a redistribution of K-Ras to endosomes, whereas treatment with ionophores that modify transmembrane potential induce K-Ras translocation to the cytoplasm and endomembranes.⁹¹ Interestingly, alterations in the PM outer leaflet can also affect Ras localization: Depletion of sialic acid results in K-Ras redistribution to recycling endosomes.⁹¹ Likewise, the clustering of outer leaflet and transmembrane lipid raft-associated proteins retards the lateral diffusion of H-Ras but not K-Ras.⁹² PM viscosity, as determined by cholesterol content, can also profoundly affect Ras mobility: In cells loaded with excess cholesterol, the diffusion of all Ras isoforms is reduced.⁹³ Contrarily, cholesterol depletion increases the lateral diffusion of H-Ras but has little effect on the other Ras isoforms⁹⁴; in consequence, H-Ras but not K-Ras signaling is attenuated.⁹⁵ Despite these data, quite surprisingly, little is known on how membrane composition and structure affect Ras localization, regulation, and functions. It is hoped that more information will be provided on this crucial aspect.

Site-Specified Ras Regulation

Expectedly, being present at multiple subcellular localizations implies functioning in different microenvironments where Ras could be subject to distinct control mechanisms depending on the presence, availability, and functionality of the different regulatory proteins that orchestrate Ras activation and also on site-specific processes that could somehow impinge on Ras regulation. Initially, GEFs were thought to activate Ras exclusively at the PM.⁹⁶ In parallel to the findings demonstrating that Ras was also activated at endomembrane systems, some GEFs were found to act specifically at distinct internal sublocalizations. Accordingly, in T cells, Ras-GRP1 was shown to activate Ras at the GC in a phospholipase Cγ and calcium-dependent fashion.^97,98 However, in T cells, RasGRP1 can also activate Ras at the PM in response to the activation of phospholipase D and DAG generation,⁹⁹ illustrating how the same regulator under different stimuli can function at distinct locations. Similarly, Ras-GRF family GEFs are responsible for switching on Ras at the ER, but they can also activate Ras at the PM depending on the stimulus: In epithelial cells, LPA promotes Ras activation at the ER, whereas calcium ionophores trigger Ras activation at the PM.¹⁰⁰ Other GEFs (e.g., RasGRP2) seem to act exclusively at the PM¹⁰¹ (for a review, see Omerovic et al.¹⁰²). GEF functions can be affected by scaffold proteins such as galectin-3, which inhibits Ras-GRP4, thereby diverting incoming signals away from H-Ras and N-Ras nanoclusters toward K-Ras nanoclusters, promoting the activation of this last isoform.¹⁰³ Also, small molecules seem to exert a regulatory effect in a site-specific fashion; such is the case for nitric oxide, derived from endothelial nitric oxide synthase action, which specifically activates N-Ras at the GC of T cells upon TCR engagement.¹⁰⁴

Site-specific regulatory effects have also been observed for Ras inactivation by GAPs. Initial studies suggested that GAP activity was more pronounced in the interior of the cell than in the peripheral PM,^105,106 which would agree with reports indicating that the PM is the predominant site for Ras function.¹⁰⁷ However, it is likely that such notions are not universal and that the situation varies significantly depending on the cellular context. In epithelial cells, calcium activates RasGRF to stimulate Ras activation at the PM,¹⁰⁰ whereas in lymphocytes, it triggers Ras activation at the GC by Ras-GRP1 and promotes Ras inactivation at the PM by switching on the calcium-responsive GAP CAPRI.⁹⁸ Indeed, CAPRI¹⁰⁸ and other GAPs such as GAP1^IP4BP are predominantly found at the PM,¹⁰⁹ whereas others such as NF1 are ubiquitous.^110,111 As in the case of GEFs, scaffolds and ancillary proteins can also affect GAP functions: In Hela cells, annexin A6 promotes the recruitment of p120GAP to PM nonraft microdomains by forming a complex with p120GAP, whereby Ras activity is modulated.¹¹²

As illustrated above, depending on its localization, Ras is subject to the action of different GEFs and GAPs and probably other regulatory proteins, which in turn are subject to site-specific regulatory events depending on, for example, the acting stimulus, as mentioned above. In consequence, compartmentalization can result in a remarkable variability of Ras signal outputs. In this respect, it has been shown that at the PM, Ras activation is fast and transient, whereas at endomembranes such as the ER, GC,¹¹³ and endosomes,¹¹⁴ it is slow and prolonged. Scaffold proteins can affect Ras signal output in a site-specific fashion: For example, galectin-1 stabilizes H-Ras nanoclusters, resulting in an enhanced recruitment of effectors and greater signal intensity.³¹ This is achieved on the basis that galectin-1 binding to H-Ras is sensitive to the orientation of the G-domain relative to the surface of the membrane, in such a way that it stabilizes the active G-domain orientation. This stabilization allows the formation of H-Ras nanoclusters and enhances the recruitment of effector molecules,¹¹⁵ an effect that may be related to the potentiation of cellular transformation by galectin-1.^33,65 Furthermore, the formation of Ras nanoclusters seems to be important for the transformation of analog signal inputs into digital outputs.¹¹⁶ In some cases, effector molecules can serve the role of scaffold proteins by retaining Ras at specific sublocalizations. This is the case for PLC-ε that, in response to calcium stimulation, can sequester Ras on the GC, decreasing Ras-GTP levels in the PM and turning Ras transient signals at the GC into sustained ones.¹¹⁷

Ras Effector Usage Can Be Spatially Defined

Importantly, Ras sublocalization also has a notorious impact on the determination of which effector pathways are to be activated and how intensively. Experiments demonstrating that the subcellular localization markedly influences ERK signal output¹¹⁸ clearly highlight this concept. Probably, this has a lot to do with how abundant and how available effector molecules are at different sites. For instance, C-Raf and B-Raf are ubiquitous proteins,¹¹⁹ whereas A-Raf is particularly enriched in mitochondria.¹²⁰ IMP is mostly present in PM cholesterol–rich microdomains,¹²¹ and Rain/RasIP1 is exclusively activated at the GC¹²² (for a review, see Omerovic et al.¹⁰²). Of course, it is very likely that extreme variability will be encountered in the subcellular distribution of the different Ras effector molecules depending on the cellular context. Not only the availability of effector molecules will shape Ras signals but also how functional they are at different microenvironments. For example, B-Raf and C-Raf are known to interact with membrane phospholipids. Some of these, such as phosphatidylethanolamine and phosphatidylinositol, inhibit their kinase activity.¹²³ Thus, it is possible that the degree of Raf activation could be regulated to some extent by the phospholipid composition of the membranes where Ras is acting. In a similar fashion, the activity of regulatory proteins such as RKIP, a Raf inhibitory protein known to bind phosphatidylethanolamine,¹²⁴ could also be dependent on the composition of the different sublocalizations where it is present.

A great part of our current knowledge on Ras compartmentalized functions derives from the use of artificially tethered constructs. This was the strategy used by Mark Philips for the first time to show that in COS1 cells, H-RasV12 tethered to the ER activated ERK1/2 MAP kinases (ERKs), PI3K, and JNK. Engagement of the pathway leading to ERK activation was also observed when H-RasV12 was tethered to the GC.¹¹³ A later study from our laboratory used the same approach in NIH3T3 cells to study the differences in Ras effector usage between endomembranes and PM and among PM microdomains as well: It was found that H-RasV12 efficiently activated ERKs, PI3K, and Ral-GDS at lipid rafts, and at the ER, at the bulk membrane, it behaved similarly, although with lower PI3K activation. However, at the GC, H-Ras profusely activated Ral-GDS but was unable to signal to ERKs.¹²⁵ Even though there may be some discrepancies between these two studies, probably due to experimental details, the concept of different subcellular localizations conferring variability in Ras effector usage was firmly established. This notion has also held at the “nano” scale as it has been found that in response to EGF stimulation, Raf is recruited to K-Ras but not to H-Ras nanoclusters.⁸¹ Furthermore, recent findings from our laboratory have gone a step further by demonstrating that the microenvironment in which Ras signals unfold can also affect events further downstream, by regulating the effectors’ substrate specificity. Ras sublocalization determines which substrates will be preferentially phosphorylated by activated ERKs, with this specificity being defined by the participation of distinct scaffold proteins. As such, ERKs activated by Ras at the ER would use Sef-1 to phosphorylate cPLA₂. This substrate was also activated by ERKs in response to Ras signals coming from lipid rafts, but in this case, the mediating scaffold was KSR1. Ras signals emanating from lipid rafts also regulated the phosphorylation of other ERK substrates such as EGFr; in this process, the participating scaffold was IQGAP1.¹²⁶

This brings into the equation the concept that space can affect Ras signals not only by orchestrating the functions of its regulatory molecules and by controlling its interactions with effector proteins but also by setting up the conditions whereby its signals are going to flow through key nodes further downstream (for a review, see Kholodenko et al.¹²⁷). One such node is ERK. The central role played by ERKs in Ras biological functions is unquestionable, and recent advances have made clear that ERKs are subject to strict spatial regulation (for a review, see Yao and Seger¹²⁸). As mentioned above, one mechanism whereby space regulates ERK signals is by the participation of scaffold proteins. These proteins regulate the amplitude, intensity, and spatial specificity of its signals. Scaffold proteins play no role in the enzymatic reactions taking place along the MAPK cascade and serve as structural backbones where the different members of the signaling cascade are brought together, forming a complex whereby optimization of the signal is achieved. At the same time, they exclude other related components that operate in parallel cascades, insulating a MAPK module from undesired external interferences.^129,130 In addition, scaffolds play important roles in the spatial selectivity of ERK signals. Evidence gathered in the past few years points to different scaffold proteins acting at distinct subcellular localizations. For example, KSR1 acts preferentially in response to signals emanating from PM cholesterol-rich domains¹²¹; MP-1 regulates ERK activity in endosomes¹³¹; Sef functions at endomembranes, particularly in the GC^126,132; and paxillin at focal adhesions¹³³ and β-arrestins are abundant in clathrin-coated pits,¹³⁴ where they enhance ERK activation by assembling the different tiers of its cascade.¹³⁵ Importantly, β-arrestins prevent ERK nuclear translocation, favoring ERK cytosolic functions.¹³⁵ A similar effect was observed for Sef, which sequesters activated ERKs at the GC, preventing their nuclear translocation¹³² and inhibiting PC12 differentiation.¹³⁶ Noticeably, this phenomenon seems to be widespread among scaffold proteins as down-regulating their levels using siRNA interference results in an enhancement of ERK nuclear functions and a drop in ERK cytoplasmic activity.¹³⁷ Indeed, scaffold proteins serve as dimerization platforms where ERK dimers are assembled; these scaffold-dimer complexes ensue the interaction of ERKs with their cognate cytoplasmic substrates.¹³⁷ Thus, space through the participation of specific scaffold proteins can orchestrate the spectrum of ERK substrates to be phosphorylated in response to Ras signals.

The selection of the effector pathways to be activated by Ras can be influenced by other types of site-specific regulatory proteins. This is clearly exemplified by galectins: In response to EGF stimulation, galectin-1 could direct H-Ras signals toward the activation of C-Raf at the expense of the PI3K pathway.⁶⁵ Likewise, galectin-3 enhanced EGF-induced K-Ras activation with the subsequent increment in C-Raf and PI3K activity but with an attenuated ERK response.³⁴

Ras trafficking between different cellular compartments can also impact profoundly on which effector pathways are going to be switched on and how intensively. As previously mentioned, in CHOK1 cells, ubiquitination of H- and N-Ras promotes their association to endosomes, where reduced availability of C-Raf in comparison to the PM results in diminished activation of the ERK pathway.⁸⁵ Consistently, in Drosophila, a threshold of Ras ubiquitination, and therefore a pool of endosomal Ras, is required to prevent excessive Ras-ERK signaling.¹³⁸ These results suggest that ubiquitination could be confining Ras in a subcellular compartment defective for coupling Ras signals to ERK activation, thereby down-regulating this pathway. However, endosomal signaling appears to be quite complex, and it is likely that different types of endosomes affect signaling in different ways. For instance, H-Ras but not K-Ras signaling requires endocytosis.¹⁹ However, some endosomes appear to serve as sites for K-Ras degradation.²¹ Likewise, active H-Ras is internalized together with EGF receptors after stimulation with EGF, insulin, or NGF to endosomes where phosphorylated ERKs can be detected,^139,140 but Ras-ERK signals can be inhibited by impeding clathrin-mediated endocytosis.¹⁴¹ This may reflect that clathrin-dependent and clathrin-independent endosomes affect Ras signaling in different ways. Overall, these results clearly point to endosomes as independent signaling environments as opposed to the notion that they constitute a sequential continuation of the signaling events taking place at the PM once the PM has been internalized. Indeed, the fact that, as mentioned before, scaffold proteins such as MP-1 specifically regulate ERK signaling at endosomes¹³¹ illustrates the unfolding of specific signaling events taking place at this microenvironment.

Ultimately, the variability in the use of effector pathways depending on Ras sublocalization results in marked differences in the transcription programs switched on by Ras from its different sites. In NIH3T3 fibroblasts, H-RasV12 signals regulate the expression of 1074 genes in total, of which 329 genes, nearly a third, are commonly regulated by Ras from all of its compartments. One hundred twenty-one genes are uniquely regulated by Ras signals emanating from plasma membrane microdomains, all of them from disordered membrane microdomains. Interestingly, not a single gene is specifically controlled by Ras signals derived from lipid rafts—something that could explain the fact that H-Ras/N-Ras double knockout mice, devoid of Ras isoforms at lipid rafts, are viable and exhibit a normal phenotype.¹⁴² Noticeably, only 9 genes are exclusively regulated by Ras signals originating from endomembranes.¹⁴³

Different Responses in Different Environments

It is nothing but intuitive that the aforementioned differences in timing, intensity, and selection of effector pathways that define each site-specific Ras subsignal must ultimately lead to differences in the biological outputs emanating from the different sublocalizations. This notion has been demonstrated to be correct right from the lower end of the eukaryotic evolutionary tree. In Schizosaccharomyces pombe Ras1p, the only Ras protein in this organism controls cell morphology from the ER, whereas signals emanating from the PM regulate mating.¹⁴⁴ In mammalian cells such as PC12, it has been long known that transient Ras-ERK signals induce proliferation, whereas sustained activation causes differentiation.¹⁴⁵ This effect must require Ras lateral diffusion between PM microlocalizations, as confining H-Ras to lipid rafts significantly inhibits sustained ERK activation and prevents PC12 differentiation.⁷⁴ Recently, it has been possible to assign some of the Ras biological outputs to specific localizations, thanks to the use of artificially tethered constructs. In murine fibroblasts, H-Ras could effectively support proliferation and cellular transformation from the ER, disordered membrane, and lipid rafts but not from the GC when directed thereto using the KDEL receptor as a GC-targeting signal.¹²⁵ Likewise, H-Ras tethered to the GC by another tether, the E1 ABV protein, was also unable to transform NIH3T3 fibroblasts.¹⁴⁶ Furthermore, RKTG, a 7-transmembrane protein that specifically localizes to the GC, is capable of unproductively sequestering C-Raf therein, leading to the inhibition of ERK activation and preventing PC12 differentiation.¹⁴⁷ The Ras family GTPase Rap1, capable of antagonizing Ras functions, is also enriched in this compartment,¹⁴⁸ further suggesting that the GC platform negatively regulates Ras-ERK signals. In agreement, in thymocytes, positive selection leads to proliferation, whereas negative selection induces apoptosis. Both responses require Ras-ERK activation, but whereas negative selection is triggered by a strong PM-derived activation of the Ras-ERK pathway, positive selection is induced by weaker antigens, generating a low-amplitude Ras-ERK signal at the GC.¹⁴⁹ It must be noticed that some discrepancy exists with respect to Ras transforming potential at the GC, as another study finds H-Ras targeted to this organelle capable of transforming murine fibroblasts.¹¹³ It is likely that such differences reside in the GC tethers used, although it is worth mentioning that in both studies, transformation correlates with the degree of ERK activation. Furthermore, in T lymphocytes, N-Ras signal at the GC is required for T cell activation in response to TCR signaling.¹⁵⁰ Thus, Ras at the GC appears to be required at least for a subset of cellular processes in some cell types.

In the case of the ER, in murine fibroblasts, the only site-specific H-RasV12 signal capable of sustaining survival in response to growth factor starvation emanated from this organelle.¹²⁵ Indeed, H-RasV12 tethered to the ER switches on a transcriptional program remarkably enriched in genes functionally related in survival and antiapoptotic roles.¹⁴³ In agreement, N-Ras, which generates a potent antiapoptotic signal,²⁴ is particularly enriched at the ER, pointing to this organelle as an important site for the regulation of cellular survival. However, the ER is not likely to be the only localization for such role. As previously mentioned, PKC-mediated phosphorylation of K-Ras triggers its rapid dissociation from the PM, where it generates proliferative signals, and its translocation to mitochondria, where it induces apoptosis by some mechanism mediated by Bcl-X_L.⁷⁹ This process is a clear example of how a Ras isoform—in this case, K-Ras—can generate diametrically opposed biological effects, proliferation versus cell death, when functioning at two different subcellular localizations, the PM versus mitochondria. N-Ras is also present in the outer and inner mitochondrial membrane compartments, but contrary to K-Ras, its functions therein have been related to antiapoptotic effects.^24,151,152

Noticeably, the presence of specific Ras isoforms at endomembranes is associated with the homeostasis of these cellular systems. Cells lacking N-Ras or K-Ras exhibit an abnormal mitochondrial morphology.¹⁵² Likewise, the overexpression of oncogenic H-Ras, but not K- or N-Ras, induces the vacuolization and expansion of the ER, typical of an ER-associated stress response.¹⁵³ Similarly, oncogenic N-Ras at the GC induces alterations in its architecture, evidenced by a collapsed morphology, accompanied by an increase in protein transport from the trans-Golgi network to the cell surface.¹⁵⁴

Conclusion

The findings described above clearly demonstrate the dynamic nature of Ras localization at different subcellular compartments, where its activation and its signals are subject to distinct site-specific control mechanisms. These phenomena clearly endow Ras signals with a broad variability in amplitude, duration, and effector usage that, depending on the settings, can ultimately lead to a vast spectrum of biological outputs. It is very likely that signal compartmentalization accounts to a great extent for the differences in biochemical and biological functions encountered among the three Ras isoforms and for the functional variability that each isoform displays in different cellular contexts.

Evidently, what holds under physiological conditions may very well be pertinent under pathological conditions. In this respect, it will be interesting to see how space affects Ras functions particularly in tumor cells, where aberrant Ras signals play a major role, in comparison to their normal counterparts. Much to our distress, we cannot provide an answer yet. To begin with, to date, there are no data available comparing Ras distribution in tumor with normal cells, probably due, to a great extent, to technical difficulties, such as the absence of good isoform-specific antibodies for immunofluorescence studies. However, the data gathered thus far can suffice for making the educated anticipation that it is likely to be quite different. For example, as mentioned before, H-Ras localization in different PM microdomains depends on its activation state.²⁷ Therefore, whereas in normal cells, wild-type H-Ras, cycling between GTP and GTP bound states, would traffic between lipid rafts and disordered membrane microdomains, in tumor cells, oncogenic H-RasV12, locked in the GTP-bound form, would be confined to the disordered membrane, thereby altering significantly the nature of the signals emitted. The differences in subcellular distribution of wild-type versus oncogenic H/N-Ras may be even greater, considering that Ras-GTP is depalmitoylated much faster than Ras-GDP,⁷² which could result in significant variations in the acylation cycle and therefore in the proportion of Ras molecules that signal from the PM versus endomembrane pools. Moreover, some tumor cells exhibit overexpression of proteins involved in the Ras acylation cycle, as is the case for the palmitoyl transferase DHHC9.¹⁵⁵ Overall, these alterations in the mechanisms whereby Ras subcellular distribution is orchestrated may lead to gross and grievous changes in Ras sublocalization, ultimately leading to substantial differences in Ras functions in tumor versus normal cells.

At this point, it seems clear that the Ras signal is not a single, unique, and homogeneous entity but rather the sum of multiple, site-specified, distinct subsignals.¹⁷ Conceptually, in our quest for drugs that prevent aberrant Ras signaling to fight cancer, a compound that selectively blocks those Ras subsignals essential for tumor progression, but not those regulating other functions, should yield drugs with reduced toxicity and undesired side effects, compared to compounds that block Ras signaling completely (for reviews, see Calvo et al.¹⁵⁶ and Matallanas and Crespo¹⁵⁷). As a proof of principle, there are already data demonstrating that cellular transformation can be prevented or reverted just by inhibiting Ras at specific sublocalizations. For example, fibroblast transformation induced by the oncogenes v-Src and Sis can be inhibited by interfering with Ras signals coming from lipid rafts or disordered membrane using site-specific dominant inhibitory mutants.¹²⁵ Ras-induced transformation can be suppressed by the overexpression of annexin A6 that recruits p120GAP specifically to nonraft PM microdomains.¹¹² Also, inhibiting galectin-1 with the use of an inhibitory mutant destabilizes H-Ras “nanoclusters” at the PM and prevents fibroblast transformation.³³ Although the experimental approaches of these experiments may be somewhat rudimentary, this set of data serves to illustrate that it may not be necessary to fully suppress Ras signals to achieve antitumor responses. Thus, although conceptual at this moment, strategies directed at interfering with Ras site-specific signals could be a valid therapeutic option for fighting tumors sometime in the future. In the meantime, the notion of space as a key aspect of Ras biology is already deeply rooted, and the path is paved for a better understanding of this novel aspect of these unique proteins.