Ras proteins associate with cellular membranes by virtue of a series of post-translational modifications of their C-terminal CAAX sequences. The discovery that two of the three enzymes that modify CAAX proteins are restricted to the endoplasmic reticulum led to the recognition that all nascent Ras proteins transit endomembranes en route to the PM (plasma membrane) and that at steady-state N-Ras and H-Ras are highly expressed on the Golgi apparatus. To test the hypothesis that Ras proteins on internal membranes can signal, we developed a fluorescent probe that reports when and where in living cells Ras becomes active. We found that growth factors stimulated rapid and transient activation of Ras on the PM followed by delayed and sustained activation on the Golgi. We mapped one pathway responsible for this activity as involving PLCγ (phospholipase Cγ)/DAG (diacylglycerol)+Ca2+/RasGRP1. Using mammalian cells and fission yeast, we have shown that differential localization of activated Ras preferentially activates distinct signalling pathways. In very recent work, we have found that (i) the subcellular localization of K-Ras can be acutely modulated by phosphorylation of its C-terminal hypervariable region by PKC, (ii) among the membranes upon which phosphorylated K-Ras accumulates is the outer mitochondrial membrane and (iii) phosphorylated, internalized K-Ras promotes apoptosis. Thus the signalling output of Ras depends on its subcellular localization.
Association with cellular membranes is required for the biological activity of Ras family proteins. These proteins are synthesized in the cytosol on free polysomes and secondarily associate with membranes as a consequence of post-translational modifications . This processing is directed by a conserved CAAX motif at the C-terminus that contains an invariant cysteine as the fourth to last residue. CAAX sequences direct prenylation by one of two cytosolic prenyltransferases, farnesyltransferase or geranylgeranyltransferase I. These enzymes catalyse the addition of a 15-carbon farnesyl or a 20-carbon geranylgeranyl lipid through a stable thioether linkage to the CAAX cysteine . After prenylation, Ras proteins become substrates for a protease designated Ras converting enzyme 1 (Rce1) that removes the AAX amino acids of the CAAX sequence [3,4]. The new C-terminal prenyl cysteine then becomes a substrate for the third and final CAAX processing enzyme, Icmt (isoprenylcysteine carboxyl methyltransferase) that methyl esterifies the α carboxyl group . The end result of these modifications is to remodel a hydrophilic C-terminus into one that is hydrophobic.
Our discovery 7 years ago that Icmt is an integral membrane protein that is restricted to the ER (endoplasmic reticulum)  and the subsequent localization of Rce1 to the same compartment  suggested that nascent Ras proteins must traffic to the PM (plasma membrane) via the endomembrane. Subsequent studies using GFP (green fluorescent protein)-tagged Ras proteins revealed that this was indeed the case . The hypervariable regions immediately upstream of the CAAX sequence contain either cysteines that can be modified by palmitoylation or a highly basic polylysine sequence. Hancock et al.  showed 15 years ago that one or the other of these so-called ‘second signals’ was required for Ras trafficking to the PM. Once endomembrane trafficking was recognized, it was appreciated that the second signals are required for trafficking from the endomembrane to the PM [8,10]. At steady-state, the palmitoylated forms of Ras, N-Ras and H-Ras, are expressed on both the PM and Golgi apparatus. This is a function of not only anterograde transport on the cytoplasmic face of elements of the secretory system but also of a recently demonstrated retrograde transport of N-Ras and H-Ras that have reached the PM but are subsequently depalmitoylated and returned to the Golgi apparatus .
Ras is active on the Golgi
Because Ras proteins transduce signals from surface receptors that traverse the PM, it was generally considered that they functioned only at the PM. The discovery that all Ras proteins visit the endomembrane en route to the PM, and that at steady-state N-Ras and H-Ras are highly expressed on the Golgi, led us to hypothesize that Ras signalling may also occur on internal membranes. To test this hypothesis, we developed fluorescent probes that are capable of reporting when and where in the cell Ras is activated. Our probe consists of a GFP fusion with the Ras binding domain of Raf-1 that binds GTP-bound Ras with 104-fold higher affinity than GDP-bound Ras (Figure 1). When expressed in a serum-starved, quiescent cell GFP–RBD (Ras-binding domain) is distributed homogeneously throughout the cytoplasm and nucleoplasm. When such a cell expressing GFP–RBD is stimulated with a growth factor, such as EGF (epidermal growth factor), the probe rapidly (3–10 min) redistributes to the PM and clearing of the cytoplasm and nucleoplasm can be observed (Figure 2) . Recruitment of GFP–RBD to the PM reverses by 10–20 min after stimulation. Thus GFP–RBD recruitment to the PM recapitulates the fundamentals of Ras/MAPK (mitogen-activated protein kinase) signalling elucidated with biochemical measurements. The more unexpected finding using GFP–RBD technology was that subsequent to Ras activation on the PM there appeared a delayed but sustained activation on the Golgi apparatus (Figure 2) . Even more unexpected was the observation that, in Jurkat T-cells stimulated through the T-cell receptor, activation of Ras was detected only on the Golgi and within 3 min of stimulation .
A genetically encoded in vivo probe for activated Ras
Spatiotemporal analysis of Ras activation in a living cell
Role of calcium in Ras activation
Growth factor-stimulated Ras activation on the Golgi apparatus that is topologically removed from PTKRs (protein tyrosine kinase receptors) at the cell surface posed a cell biological conundrum. What was the basis for the translocation of the signal across the cytosolic void? The description of the so-called signalling endosomes  that incorporated in their membranes PTKRs that remained activated and carried with them adaptor and signalling molecules such as Grb2 and SOS provided one plausible explanation for signal propagation. However, we found that blocking endocytosis in particular, with dominant negative epsin, or vesicular transport in general, with a temperature block, had no effect on Ras activation at the Golgi . These observations suggested that rather than vesicular transport, a diffusible mediator might be responsible for propagating the signal. This hypothesis drew our attention to calcium, the best studied of the diffusible second messengers. Indeed, we found that whereas chelating calcium or blocking calcium channels did not block Ras activation at the PM, these manipulations completely inhibited Ras activation at the Golgi. In support of this observation we found that PLCγ (phospholipase Cγ), an enzyme responsible for elaborating the twin second messengers DAG (diacylglycerol) and inositol trisphosphate, the latter causing the release of stored calcium, was required for Ras activation on the Golgi .
The involvement of cytosolic calcium drew our attention to a family of Ras GEFs (guanine nucleotide-exchange factors), Ras GRPs (guanine nucleotide releasing proteins) that are regulated by calcium and DAG. Like PKCs (protein kinase Cs), RasGRPs possess both C1 and C2 domains that bind DAG and calcium respectively . As such, Ras GRPs were expected to translocate, like PKCs, to the PM. However, we found that the primary membrane compartment upon which RasGRP1 accumulated upon cellular activation was the Golgi apparatus. The basis for this membrane specificity is unknown but is likely to reside in specific sequences in the C1 domain and in the microenvironment of DAG in Golgi membranes . Whatever the reason for the affinity of activated RasGRP1 for the Golgi apparatus, it was clearly a strong candidate for a diffusible mediator that could activate Ras in situ on the Golgi. Moreover, its abundant expression in lymphocytes might explain the rapid activation of Ras on the Golgi observed by us in Jurkat cells. We demonstrated a requirement for RasGRP1 by knocking down this GEF with siRNA (small interfering RNA) and showing that growth factor-stimulated Ras activation was lost on the Golgi but not on the PM . Thus Ras can be activated on the Golgi apparatus through a pathway downstream of PTKRs that involves PLCγ followed by calcium- and DAG-dependent activation of RasGRP1 (Figure 3).
In situ activation of Golgi-associated Ras
As described above, calcium chelators and PLCγ inhibitors blocked Ras activation at the Golgi but not on the PM. However, Ras activation at the PM was not unperturbed. Under these conditions we observed markedly prolonged activation on this compartment. This observation drew our attention to a family of Ras GTPase activating proteins (GAPs) that are regulated by calcium. We studied one of these, CAPRI, that was known to translocate to the PM in response to elevations in cytosolic calcium . Knockdown of CAPRI prolonged Ras signalling. In Jurkat lymphocytes, we were surprised to observe Ras activation only at the Golgi after the engagement of the T-cell receptor . However, when CAPRI was knocked down, we observed Ras activation both on the Golgi and at the PM. Thus CAPRI is largely responsible for the deactivation cycle of Ras at the PM. This observation leads to the fascinating conclusion that a single second messenger, cytosolic calcium, regulates simultaneously Ras activity in opposite directions on two different subcellular compartments. This is a stark example of the power and utility of compartmentalized signalling.
Retrograde transport of Ras to the Golgi
Because Ras is a peripheral membrane protein that, in principle, can be released from membranes, we and others have also considered the possibility that, like N-Ras proteins, under certain circumstances, mature Ras can move from membrane compartment to compartment. Using fluorescence recovery after photobleaching and fluorescence correlation spectroscopy, we have recently derived evidence to support a model whereby mature N-Ras and H-Ras at the PM can be depalmitoylated and then traffic, in a retrograde fashion, to the Golgi apparatus . If Ras that enters this retrograde pathway starts out and remains GTP-bound, this could represent an additional pathway for Ras signalling at the Golgi. Recently, Rocks et al.  have reported data that support just such a pathway. Many questions remain with regard to elucidating this pathway fully. For example, is depalmitoylation regulated and, if so, is this pathway somehow coupled to one that protects activated Ras from the effects of PM-associated GAPs? What is the mechanism for transport of a farnesylated protein through the cytosol? Is there a chaperone involved analogous to RhoGDI (guanine nucleotide dissociation inhibitor) that solublizes prenylated Rho family proteins? Can a pool of soluble, GTP-bound Ras be detected in the cytosol? If retrograde traffic of depalmitoylated N-Ras and H-Ras prove to be a physiologically important mechanism of accumulating active Ras at the Golgi it need not be to the exclusion of the RasGRP1 pathway that activated Ras on this compartment in situ.
Role of Ras signalling on the Golgi
Having established that the Ras can be activated on the Golgi apparatus the next logical question becomes for what purpose? Are the signalling characteristics, and in particular the signal output, different from distinct subcellular compartments? As discussed above, the nature of peripheral membrane proteins like Ras that are secondarily targeted to membranes in a fashion that may be reversible makes difficult a study of their location-specific function in an unambiguous way. Therefore, to approach this question, we studied the signal output of Ras expressed constitutively on specific compartments using transmembrane tethers. We targeted activated H-Ras61L to the ER using the first transmembrane segment of the avian bronchitis virus M1 protein that affords strong ER restriction. To deliver H-Ras61L to the Golgi, we used the KDEL receptor (KDELR) that traffics between the ER and Golgi but at steady-state is highly expressed on Golgi. Importantly, neither of these constructs allowed any expression whatsoever on the PM. We measured three outputs of Ras signalling, two direct, Erk and Akt phosphorylation, and one indirect, Jnk activation. We found that H-Ras61L tethered to the Golgi through the KDELR showed robust activation of both the Erk and Akt pathways to a level comparable with that achieved with expression of the same level of natively targeted oncogenic H-Ras61L. However, KDELR tethered H-Ras61L only weakly activated Jnk. In contrast, H-Ras61L restricted to the ER with the M1 transmembrane segment activated Erk and Akt relatively weakly but activated Jnk to a higher level than natively targeted H-Ras61L . Thus subcellular localization exerts a considerable influence over signal output from Ras.
Metazoa have multiple Ras isoforms and numerous closely related family members as well as several modes of membrane targeting. Moreover, the pathways regulated by Ras proteins are numerous and multiplex. Therefore to demonstrate unambiguously that Ras signals in a compartmentalized fashion, we have turned to a lower eukaryote as a model system. We have used the fission yeast Schizosaccharomyces pombe because it has but one Ras protein designated Ras1. Ras1 of S. pombe controls two distinct pathways . One controls mating, is initiated by the engagement of mating pheromone by its cognate G-protein-coupled receptor, and proceeds through a classical MAPK pathway where Byr2 interacts with Ras1 and serves as the MAPKKK (MAPK kinase kinase). The other is the morphology pathway that maintains the characteristically elongated shape of S. pombe. In this pathway, Ras1 activates Scd1 that is a GEF for Cdc42, a Rho family GTPase that controls morphology through the actin cytoskeleton. Two GEFs for Ras1 have been described in S. pombe. These include Ecf25 and Ste6. Interestingly, genetic ablation of Ste6 leads to sterility but does not affect morphology, and removal of Ecf25 leads to short rounded cells but does not affect mating. Thus upstream elements in the Ras1 pathway seen to selectively engage downstream pathways. We tested the hypothesis that compartmentalized signalling could account for this bifurcation in Ras1 signalling. We showed that a palmitoylation mutant of Ras1 becomes restricted to the endomembrane and can rescue the morphology pathway but not the mating pathway. Conversely, Ras1 targeted directly and strictly to the PM with the C-terminal segment of the small GTPase RIT could rescue mating but not the morphological defect of Ras1 null cells. Thus Ras1 signalling through its two well-characterized pathways is spatially segregated in fission yeast. These observations represent the most unambiguous results to date that argue for compartment-specific Ras signalling.
Ras and mitochondria
In recent studies (T.G. Birona, S.E. Quatela and M.R. Philips, unpublished work), we have discovered another example of compartment-specific Ras signalling. As described above, the most prevalent splice variant of K-Ras is unique among Ras isoforms in being targeted to the PM by a polylysine sequence that operates as the ‘second signal’ in conjunction with a CAAX motif. The polybasic region is flanked by two serines that are putative PKC sites . We have now shown that PKC agonists stimulate phosphorylation of these residues (Ser181>Ser171) and cause a rapid (<3 min) translocation of K-Ras from the PM to internal membranes in a Ser181-dependent fashion. Unexpectedly, among the endomembranes upon which phosphorylated K-Ras accumulates is the outer mitochondrial membrane. A K-Ras mutant with a phosphomimetic Glu181 was constitutively internalized. Surprisingly, this K-Ras mutant induced apoptosis. Because K-Ras is the isoform most often associated with human cancer this observation was very exciting. It suggested that agents that promote K-Ras phosphorylation on Ser181 might induce apoptosis of K-Ras transformed cells and thereby serve as novel anti-cancer drugs. Although Ras proteins are generally known for their ability to drive cell proliferation and to promote cell survival, their ability to promote apoptosis in certain contexts has long been recognized . Our new observations add apoptosis as an outcome of Ras signalling that may be controlled in a compartment-specific fashion.
In summary, we have shown that the PM is not unique as a platform from which Ras proteins regulate signalling pathways. The cytoplasmic face of ER, Golgi and mitochondrial membranes are also signalling platforms for Ras. Until recently, the study of signal transduction involved only ‘what’ and ‘when’. It is now clear that ‘where’ must also be factored into the equation. Cell biologists have been at a loss to explain the complex and multiple outputs of single signalling molecules such as Ras. Spatial compartmentalization affords another level of regulation that allows for greater specificity and segregation of signal outputs.
Localization and Activation of Ras-like GTPases: Focused Meeting held at the Royal Agricultural College, Cirencester, U.K., 21–23 March 2005. Organized and Edited by A. Ridley (Ludwig Institute of Cancer Research, London, U.K.) and M. Seabra (Imperial College London, U.K.).
green fluorescent protein
isoprenylcysteine carboxyl methyltransferase MAPK, mitogen-activated protein kinase
protein tyrosine kinase receptor
guanine nucleotide-exchange factor
guanine nucleotide releasing protein
protein kinase C
Ras converting enzyme 1