Cellular Signalling Vol. 3, No. 3, pp. 179-187, 1991. Printed in Great Britain.
0898-6568/91 $3.00+.00 © 1991 Pergamon Prer,s pie
MINI REVIEW THE RAS S U P E R F A M I L Y OF M O L E C U L A R SWITCHES IAN G. MACARA
Environmental Health Sciences Center, University of Rochester Medical Center, Rochester, NY 14642, U.S.A. (Received 19 January 1991; and accepted 6 February 1991) Key words: GTPase, guanine nucleotide-binding proteins, signal transduction, oncogenes.
INTRODUCTION CRITICS have sometimes scorned the 'clone and sequence' approach to biology as providing data without insight. But important concepts have emerged from inspection of the large data set now available on protein sequences. One such concept is that of divergent functions arising from conserved mechanisms and structure, and a classic example is the universality of the GTPase timed switch to control cell processes. All switches cycle between at least two distinct, metastable states, 'on' and 'off', which in GTPases comprise the GTP-bound and GDP-bound conformations of the protein, respectively. There are other types of molecular switch, including ion channels and transcriptional factors, but the GTPases are unique in their possession of an intrinsic, regulatable timer, which is the enzymatic hydrolysis of bound GTP. The ratio of GTPase molecules in the 'on' and 'off' states at any particular time actually depends on both the rate of hydrolysis of bound GTP and on the rate of dissociation of GDP. (Fig. 1) The ratio of states provides a transient memory of the signal that altered it and the switching function arises from the differential interaction of the two states with other proteins ('targets' or 'effectors'). The utility of the GTPases as switches lies in the fact that, while the GTPase domain is highly conserved, the state-sensitive regions of the switch are variable, so that they can be designed
G'rp
GDP
I
ON
Pl
FlO. 1. Small GTPases function as timed switches. The timer is regulated by the rate constants, kotrand kc.t. This comprises a conformational change in discrete regions of the protein, induced by the presence or absence of the gamma phosphate on the guanine nucleotide. to interact with different effectors. The control function of the GTPases is enhanced by the potential to regulate the timing of the processes that determine the GTP/GDP state ratio. The GTPase superfamily is very large and comprises at least four branches: (1) the signaltransducing heterotrimeric GTPases such as Gs which possess guanine nucleotide-binding subunits (0t) of about 40,000 molecular weight [1]; (2) the protein synthesis-related GTPases such as EF-Tu, elF-2 and SRP of about 50,000-60,000 molecular weight [2-4]; (3) microtubule-associated GTPases such as DI00 and Mx of about 100,000 molecular weight [5, 6]; and (4) the small GTPases (20,000-25,000 molecular weight) such as ras [7-9]. Numerous excellent reviews have recently appeared on these various types of timed switch and their functions [1, 2, 5-9], but the pace of discovery, which has been exhilarating, continues to increase and reduces the sense of 179
180
I.G. MACARA
redundancy in adding to the list. However, so much information is already available that the present review necessarily confines itself to the group of ras-like GTPases, and will concentrate on their regulation and function. STRUCTURE OF THE SMALL GTPases A wealth of information has been amassed about the primary sequences of the small GTPases, all of which conform to the ras archetype in possessing four highly conserved regions that are involved in guanine nucleotide binding and hydrolysis [10]. These regions (GXXXXGK, DTAGXE, LXGNKXD and EXSAX) occur in the same order and with similar spacing in all known small GTPases. Mutations in the conserved amino acid residues within these sequences alter guanine nucleotide binding and hydrolysis [7, 10], and can convert ras into an oncogene by increasing the probability that p21 "~ remains in the GTP-bound 'on' state. The three-dimensional structures of some of these mutants have been determined at high resolution [11], in addition to the structures of the wild-type Ha-ras protein in its GDP- and GTP-bound states [12, 13]. In keeping with the concept of a conserved switch mechanism, the folds of the guanine nucleotide binding domains of p21 ras and the elongation factor EF-Tu are very similar, and will likely be almost identical in all the small GTPases. What regions then determine their very different functions? Mutational analysis of ras has located a region between residues 32-40, that interacts with regulatory and/or target proteins [7, 14]. In the three-dimensional structure
this region forms an external loop that undergoes a marked conformational shift between the GDP- and GTP-bound states. In particular, a Thr residue (T35 in ras), which is conserved among all the small GTPases, forms a hydrogen bond with the gamma phosphate of GTP, but is rotated out, away from the binding site, when GDP is bound [12, 13]. A second region, between residues 60-76, also exhibits different conformations in the two states, and might interact with other regulatory proteins. Both regions are variable between subclasses of the small GTPases, and likely interact with factors specific for each switch (or class of switch). Another highly variable region occurs near the C-termini of these GTPases, and may be involved in subcellular localization and membrane attachment [7, 10], and hence in specifying function. The final C-terminal sequences fall into a number of different classes which govern the type of post-translational modifications that occur, and these modifications also may regulate localization and membrane attachment. The C-terminal motifs, and the covalent modifications known at this time, are shown in Table 1. The processing steps have been worked out in most detail for the ras proteins, which possess a Cys-Ali-Ali-X consensus sequence (or CAAX box) like that of yeast a mating factor, rho and rac small GTPases, nuclear lamin B and the gamma subunits of the heterotrimeric G-proteins [14]. A farnesyl transferase that recognizes the ras CAAX sequence first attaches a C~5 prenyl group to the Cys via a thioether bond. A peptidase then cleaves (or 'axes') the final three residues (AAX), and the
TABLE1. POST-TRANSLATIONALMODIFICATIONSOF SMALLGTPASES Consensus sequence
Examples
Prenylation
Carboxymethylation Localization
-Cys-AIi-AIi-X -Cys-Ali-Ile-Phe/Leu -Asp-Cys-Ala-Cys -Gly-Gly-Gly-Cys-Cys
rho, ras raplB, G25K tab3, rab4 rabl, rab2
Cls(ras)
yes (ras) yes (raplB) yes (tab3) no (rab2)
C20 C2o(rab3)
C20
pro, cytoskel cytoskel? vesicles, cyt. Goigi/ER, cyt.
Ali, aliphatic residues; X, any residue; C~5, farnesyl; C20, geranylgeranyl; pm, plasma membrane; cyt, cytosol; ER, endoplasmic reticulum.
The ras superfamilyof molecular switches
new Cys-COO- is carboxymethylated by a methyltransferase that recognizes the prenylated Cys residue [14-16]. The Ha-ras protein is further modified by palmitoylation at a more N-terminal Cys residue [15], a step necessary to stabilize membrane attachment. The Ki-rasB protein does not possess this second Cys residue, and instead is localized to the plasma membrane by a sequence of six Lys residues just upstream from the CAAX box [17]. While this sequence of processing steps is fairly complex, it is now becoming apparent that the cell can play many baroque variations upon it, dictated by differences in the C-terminal consensus sequences. For instance, the smg p21B protein (or raplB), G gamma subunit and G25K, which possess a -C-AIi-I(F/L) consensus sequence, are carboxymethylated and prenylated with a C20 unit (geranylgeranyl) rather than with farnesyl [18]; while tab2 (-GGGCC) is geranylgeranylated but not carboxymethylated and rab3 (-DCAC) is geranylgeranylated and carboxymethylated in the cytosol but is not prenylated in the membrane [19, 20]. Different, specific prenyltransferase enzymes are probably responsible for the modification of these different C-termini, and it will be of great interest to determine how they are regulated. The loss of the prenyl group of tab3 following or during movement to the membrane is not understood but may involve a proteolytic cleavage reminiscent of that which truncates prelamin A [21]. It is also not known what other modifications may occur to the C-termini of these proteins to stabilize membrane attachment. Finally it remains to be established what other signals determine subcellular localization. R E G U L A T I O N OF S M A L L GTPases As was mentioned above, the state of the GTPase timer is controlled by two independent rate constants, kofr for GDP, and kca t for the hydrolysis of GTP. Because there is a large excess of GTP over G D P in the cell, the rate of release of GDP defines the rate at which GTP can bind to and switch the protein 'on', and kea t
181
eTP~_.I~GOP
-'-vF
i
>OE'
pl
GIP
GAP
FIG. 2. A multiplicity of factors could interact with the GTPase timed switches, including guanine nucleotide dissociation inhibitors (GDIs) and guanine nucleotide releasing factors (GRFs) that regulate ko~ and GTPase activating proteins (GAPs) and inhibiting proteins (GIPs) that regulate kca t. In addition, the 'off' and 'on' states of the GTPase switch must presumably interact with different proteins, such as receptors, or docking proteins (for control of vesicle movement), and downstream targets or effectors. controls the rate at which it switches 'off'. In principle, both of these rates could be independently increased or decreased by other cellular factors (Fig. 2), and examples of such factors are now known for several of the small GTPases. G A P S - - G T P a s e activating proteins
The first regulatory factor to be discovered was one which specifically increases the kcat for p21 r" [22]. This GAP has been cloned and sequenced [23], and exists in two forms, one of which--produced by alternative splicing--is specific to the placenta, while the other is expressed ubiquitously as a 120,000 molecular weight cytosolic protein [24]. The protein interacts with the 'effector domain' of p21 '~, which was described above and lies between residues 32-40. Because this region is necessary for transformation by v-ras, a debate has arisen as to whether GAP functions solely to switch p21 .... off' or whether it is also the downstream target for p21 r~ [8, 24]. This controversy will be discussed below. Two yeast proteins with both functional and sequence homologies to ras-GAP have also been identified (Iral and 2) [24, 25] and a new
182
I.G. M^CAR^
twist to the story has emerged with the exciting discovery that the neurofibramatosis type 1 gene (NF1) encodes a large protein with homology to the catalytic domains of Ira and r a s - G A P [26]. Moreover, this region of NF1 possesses a GAP-like activity with specificity for ras proteins [27-29]. Inheritance of a mutant form of the NF1 gene leads to the appearance of areas of dark pigmentation of the skin (caf~ au lait spots) and multiple benign tumours that emerge as outgrowths of the Schwann cells of peripheral nerves. The obvious explanation is that the mutant NF1 protein cannot catalyse the inactivation of p21'% which then remains predominantly in the 'on' state. However, the story must clearly be more complex than this, because micro-injection of the oncogenic p21 rQs (V12), which possesses a low GTPase activity unresponsive to r a s - G A P , actually stops the growth of Schwann cells [30]. Moreover, both NF1 and r a s - G A P appear to be expressed ubiquitously [26]. If they both perform the same function on the same protein, why are they both required? One possible answer to this dilemma is that the activities of the two proteins are differentially regulated. Another is that they perform different functions in addition to their effect on p2 lr~. Certainly the total sizes of these proteins ( > 2500 residues for NFI) are much larger than that required to catalyse ras GTPase activity (about 360 residues). The regulation of r a s - G A P remains mysterious. It is complicated by the discovery that GAP can associate with growth factor receptors such as those for PDGF [31], presumably through the SH2 domains, which specifically recognize tyrosine phosphate residues [32]. GAP also associates with two other proteins of unknown function, p190 and p62, and both r a s - G A P and the associated proteins are substrates for both tyrosine and serine-specific protein kinases [33, 34]. The role of phosphorylation in the regulation of G A P - - i f any--is not known. While there is evidence that the activation of growth factor receptors (tyrosine kinases) and of protein kinase C can rapidly increase the proportion of ras in the GTP-
bound 'on' state [35, 36], apparently through the inactivation of GAP, there is no evidence that this effect occurs through direct phosphorylation. A further complication is the observation that ras-GAP [37] and NFI [38] are inhibited by polyunsaturated fatty acids such as arachidonate, by lipoxygenase products of arachidonate and by certain acidic phospholipids such as phosphatidate, all of which are rapidly generated by cells in response to serum, and to many growth factors and hormones [39]. Moreover, we have recently discovered that ras-GAP (but not the NF 1 catalytic domain) is allosterically activated by certain prostaglandins [38]. Prostaglandins are generated from arachidonate by the action of cyclooxygenase. It is therefore possible that GAP action is modulated at least in part by the relative rates of conversion of arachidonate to its metabolites by lipoxygenase and cyclooxygenase, following the stimulation of phospholipase A 2 by growth factors. The function of r a s - G A P remains controversial. Its role as a negative regulator has received overwhelming evidential support, e.g. overexpression of GAP suppresses transformation by c - H A - r a s [40], and expression of human GAP in yeast decreases the stimulatory effect of ras on adenylyl cyclase [25]. However, recent observations on the modulation of atrial K ÷ channels [41] also favour a role as an effector for ras. Addition of either p21r~S.GTP complex or r a s - G A P to patches of atrial cell plasma membrane inhibits the coupling of the G k heterotrimeric G protein to muscarinic receptors, and anti-GAP antibodies block the effect, indicating that r a s - G A P is essential for p21 '~ function in this system. The situation is reminiscent of that in the control of protein synthesis, in which ribosomes act as both effector and GAP for EF-Tu [42]. How many GAPs are there? When the proteins from 3T3 cell plasma membranes are separated by high resolution anion exchange chromatography, at least a dozen distinct small GTPases can be resolved [43] and GTP hydrolysis is stimulated by addition of cell extracts for all but one of them (Wolfman and Macara,
The ras superfamilyof molecularswitches
unpublished observations). It therefore seems likely that GAPs exist for most members of the family. To date, a 29,000 rho-specific GAP has been identified by Garrett et al. [44], and GAPs with specificities for smg p21 (rap1, or Krev) [45] and for rab3 [46] have also been detected. The stringency of these specificities remains to be determined, however. The puzzle of GAP function is increased by the fact that ras-GAP will catalyse the GTPase of another member of the family, R-ras, which is not a protooncogene and possesses only about 55% homology to H-, N- and K-ras [47], and will bind with high affinity to rapl. In the latter case, no GTPase catalysis occurs, and the proteins remain associated in what is presumably an inactive complex [48]. This complex may be responsible for the antagonistic effect of rap1 on oncogenic transformation by v-ras [49].
GIPs--GTPase inhibiting protein
In addition to GAPs, factors could exist that decrease kca,. A cytoplasmic protein has now been discovered that performs this function for p21 "~ [50]. The protein (ras-GIP) has a Mr of about 60,000 and is distinct from ras-GAP. Remarkably, it is inhibited by the same phospholipids and fatty acids that activate ras-GAP. It is also potently inhibited by diacylglycerol, which is generated in response to many hormones and growth factors as a consequence of accelerated phosphatidylinositol and phosphatidylcholine turnover. Therefore these lipids and their metabolites may play a key role in the control of the ras switch through modulation of the activities of GAP and GIP. It seems likely that GIPs exist for other members of the small GTPase family, although none have yet been described.
GRFs--Guanine nucleotide releasing factors
The kofr for GDP from p21 ra' and most of the other small GTPases is extremely low (
0898-6568/91 $3.00+.00 © 1991 Pergamon Prer,s pie
MINI REVIEW THE RAS S U P E R F A M I L Y OF M O L E C U L A R SWITCHES IAN G. MACARA
Environmental Health Sciences Center, University of Rochester Medical Center, Rochester, NY 14642, U.S.A. (Received 19 January 1991; and accepted 6 February 1991) Key words: GTPase, guanine nucleotide-binding proteins, signal transduction, oncogenes.
INTRODUCTION CRITICS have sometimes scorned the 'clone and sequence' approach to biology as providing data without insight. But important concepts have emerged from inspection of the large data set now available on protein sequences. One such concept is that of divergent functions arising from conserved mechanisms and structure, and a classic example is the universality of the GTPase timed switch to control cell processes. All switches cycle between at least two distinct, metastable states, 'on' and 'off', which in GTPases comprise the GTP-bound and GDP-bound conformations of the protein, respectively. There are other types of molecular switch, including ion channels and transcriptional factors, but the GTPases are unique in their possession of an intrinsic, regulatable timer, which is the enzymatic hydrolysis of bound GTP. The ratio of GTPase molecules in the 'on' and 'off' states at any particular time actually depends on both the rate of hydrolysis of bound GTP and on the rate of dissociation of GDP. (Fig. 1) The ratio of states provides a transient memory of the signal that altered it and the switching function arises from the differential interaction of the two states with other proteins ('targets' or 'effectors'). The utility of the GTPases as switches lies in the fact that, while the GTPase domain is highly conserved, the state-sensitive regions of the switch are variable, so that they can be designed
G'rp
GDP
I
ON
Pl
FlO. 1. Small GTPases function as timed switches. The timer is regulated by the rate constants, kotrand kc.t. This comprises a conformational change in discrete regions of the protein, induced by the presence or absence of the gamma phosphate on the guanine nucleotide. to interact with different effectors. The control function of the GTPases is enhanced by the potential to regulate the timing of the processes that determine the GTP/GDP state ratio. The GTPase superfamily is very large and comprises at least four branches: (1) the signaltransducing heterotrimeric GTPases such as Gs which possess guanine nucleotide-binding subunits (0t) of about 40,000 molecular weight [1]; (2) the protein synthesis-related GTPases such as EF-Tu, elF-2 and SRP of about 50,000-60,000 molecular weight [2-4]; (3) microtubule-associated GTPases such as DI00 and Mx of about 100,000 molecular weight [5, 6]; and (4) the small GTPases (20,000-25,000 molecular weight) such as ras [7-9]. Numerous excellent reviews have recently appeared on these various types of timed switch and their functions [1, 2, 5-9], but the pace of discovery, which has been exhilarating, continues to increase and reduces the sense of 179
180
I.G. MACARA
redundancy in adding to the list. However, so much information is already available that the present review necessarily confines itself to the group of ras-like GTPases, and will concentrate on their regulation and function. STRUCTURE OF THE SMALL GTPases A wealth of information has been amassed about the primary sequences of the small GTPases, all of which conform to the ras archetype in possessing four highly conserved regions that are involved in guanine nucleotide binding and hydrolysis [10]. These regions (GXXXXGK, DTAGXE, LXGNKXD and EXSAX) occur in the same order and with similar spacing in all known small GTPases. Mutations in the conserved amino acid residues within these sequences alter guanine nucleotide binding and hydrolysis [7, 10], and can convert ras into an oncogene by increasing the probability that p21 "~ remains in the GTP-bound 'on' state. The three-dimensional structures of some of these mutants have been determined at high resolution [11], in addition to the structures of the wild-type Ha-ras protein in its GDP- and GTP-bound states [12, 13]. In keeping with the concept of a conserved switch mechanism, the folds of the guanine nucleotide binding domains of p21 ras and the elongation factor EF-Tu are very similar, and will likely be almost identical in all the small GTPases. What regions then determine their very different functions? Mutational analysis of ras has located a region between residues 32-40, that interacts with regulatory and/or target proteins [7, 14]. In the three-dimensional structure
this region forms an external loop that undergoes a marked conformational shift between the GDP- and GTP-bound states. In particular, a Thr residue (T35 in ras), which is conserved among all the small GTPases, forms a hydrogen bond with the gamma phosphate of GTP, but is rotated out, away from the binding site, when GDP is bound [12, 13]. A second region, between residues 60-76, also exhibits different conformations in the two states, and might interact with other regulatory proteins. Both regions are variable between subclasses of the small GTPases, and likely interact with factors specific for each switch (or class of switch). Another highly variable region occurs near the C-termini of these GTPases, and may be involved in subcellular localization and membrane attachment [7, 10], and hence in specifying function. The final C-terminal sequences fall into a number of different classes which govern the type of post-translational modifications that occur, and these modifications also may regulate localization and membrane attachment. The C-terminal motifs, and the covalent modifications known at this time, are shown in Table 1. The processing steps have been worked out in most detail for the ras proteins, which possess a Cys-Ali-Ali-X consensus sequence (or CAAX box) like that of yeast a mating factor, rho and rac small GTPases, nuclear lamin B and the gamma subunits of the heterotrimeric G-proteins [14]. A farnesyl transferase that recognizes the ras CAAX sequence first attaches a C~5 prenyl group to the Cys via a thioether bond. A peptidase then cleaves (or 'axes') the final three residues (AAX), and the
TABLE1. POST-TRANSLATIONALMODIFICATIONSOF SMALLGTPASES Consensus sequence
Examples
Prenylation
Carboxymethylation Localization
-Cys-AIi-AIi-X -Cys-Ali-Ile-Phe/Leu -Asp-Cys-Ala-Cys -Gly-Gly-Gly-Cys-Cys
rho, ras raplB, G25K tab3, rab4 rabl, rab2
Cls(ras)
yes (ras) yes (raplB) yes (tab3) no (rab2)
C20 C2o(rab3)
C20
pro, cytoskel cytoskel? vesicles, cyt. Goigi/ER, cyt.
Ali, aliphatic residues; X, any residue; C~5, farnesyl; C20, geranylgeranyl; pm, plasma membrane; cyt, cytosol; ER, endoplasmic reticulum.
The ras superfamilyof molecular switches
new Cys-COO- is carboxymethylated by a methyltransferase that recognizes the prenylated Cys residue [14-16]. The Ha-ras protein is further modified by palmitoylation at a more N-terminal Cys residue [15], a step necessary to stabilize membrane attachment. The Ki-rasB protein does not possess this second Cys residue, and instead is localized to the plasma membrane by a sequence of six Lys residues just upstream from the CAAX box [17]. While this sequence of processing steps is fairly complex, it is now becoming apparent that the cell can play many baroque variations upon it, dictated by differences in the C-terminal consensus sequences. For instance, the smg p21B protein (or raplB), G gamma subunit and G25K, which possess a -C-AIi-I(F/L) consensus sequence, are carboxymethylated and prenylated with a C20 unit (geranylgeranyl) rather than with farnesyl [18]; while tab2 (-GGGCC) is geranylgeranylated but not carboxymethylated and rab3 (-DCAC) is geranylgeranylated and carboxymethylated in the cytosol but is not prenylated in the membrane [19, 20]. Different, specific prenyltransferase enzymes are probably responsible for the modification of these different C-termini, and it will be of great interest to determine how they are regulated. The loss of the prenyl group of tab3 following or during movement to the membrane is not understood but may involve a proteolytic cleavage reminiscent of that which truncates prelamin A [21]. It is also not known what other modifications may occur to the C-termini of these proteins to stabilize membrane attachment. Finally it remains to be established what other signals determine subcellular localization. R E G U L A T I O N OF S M A L L GTPases As was mentioned above, the state of the GTPase timer is controlled by two independent rate constants, kofr for GDP, and kca t for the hydrolysis of GTP. Because there is a large excess of GTP over G D P in the cell, the rate of release of GDP defines the rate at which GTP can bind to and switch the protein 'on', and kea t
181
eTP~_.I~GOP
-'-vF
i
>OE'
pl
GIP
GAP
FIG. 2. A multiplicity of factors could interact with the GTPase timed switches, including guanine nucleotide dissociation inhibitors (GDIs) and guanine nucleotide releasing factors (GRFs) that regulate ko~ and GTPase activating proteins (GAPs) and inhibiting proteins (GIPs) that regulate kca t. In addition, the 'off' and 'on' states of the GTPase switch must presumably interact with different proteins, such as receptors, or docking proteins (for control of vesicle movement), and downstream targets or effectors. controls the rate at which it switches 'off'. In principle, both of these rates could be independently increased or decreased by other cellular factors (Fig. 2), and examples of such factors are now known for several of the small GTPases. G A P S - - G T P a s e activating proteins
The first regulatory factor to be discovered was one which specifically increases the kcat for p21 r" [22]. This GAP has been cloned and sequenced [23], and exists in two forms, one of which--produced by alternative splicing--is specific to the placenta, while the other is expressed ubiquitously as a 120,000 molecular weight cytosolic protein [24]. The protein interacts with the 'effector domain' of p21 '~, which was described above and lies between residues 32-40. Because this region is necessary for transformation by v-ras, a debate has arisen as to whether GAP functions solely to switch p21 .... off' or whether it is also the downstream target for p21 r~ [8, 24]. This controversy will be discussed below. Two yeast proteins with both functional and sequence homologies to ras-GAP have also been identified (Iral and 2) [24, 25] and a new
182
I.G. M^CAR^
twist to the story has emerged with the exciting discovery that the neurofibramatosis type 1 gene (NF1) encodes a large protein with homology to the catalytic domains of Ira and r a s - G A P [26]. Moreover, this region of NF1 possesses a GAP-like activity with specificity for ras proteins [27-29]. Inheritance of a mutant form of the NF1 gene leads to the appearance of areas of dark pigmentation of the skin (caf~ au lait spots) and multiple benign tumours that emerge as outgrowths of the Schwann cells of peripheral nerves. The obvious explanation is that the mutant NF1 protein cannot catalyse the inactivation of p21'% which then remains predominantly in the 'on' state. However, the story must clearly be more complex than this, because micro-injection of the oncogenic p21 rQs (V12), which possesses a low GTPase activity unresponsive to r a s - G A P , actually stops the growth of Schwann cells [30]. Moreover, both NF1 and r a s - G A P appear to be expressed ubiquitously [26]. If they both perform the same function on the same protein, why are they both required? One possible answer to this dilemma is that the activities of the two proteins are differentially regulated. Another is that they perform different functions in addition to their effect on p2 lr~. Certainly the total sizes of these proteins ( > 2500 residues for NFI) are much larger than that required to catalyse ras GTPase activity (about 360 residues). The regulation of r a s - G A P remains mysterious. It is complicated by the discovery that GAP can associate with growth factor receptors such as those for PDGF [31], presumably through the SH2 domains, which specifically recognize tyrosine phosphate residues [32]. GAP also associates with two other proteins of unknown function, p190 and p62, and both r a s - G A P and the associated proteins are substrates for both tyrosine and serine-specific protein kinases [33, 34]. The role of phosphorylation in the regulation of G A P - - i f any--is not known. While there is evidence that the activation of growth factor receptors (tyrosine kinases) and of protein kinase C can rapidly increase the proportion of ras in the GTP-
bound 'on' state [35, 36], apparently through the inactivation of GAP, there is no evidence that this effect occurs through direct phosphorylation. A further complication is the observation that ras-GAP [37] and NFI [38] are inhibited by polyunsaturated fatty acids such as arachidonate, by lipoxygenase products of arachidonate and by certain acidic phospholipids such as phosphatidate, all of which are rapidly generated by cells in response to serum, and to many growth factors and hormones [39]. Moreover, we have recently discovered that ras-GAP (but not the NF 1 catalytic domain) is allosterically activated by certain prostaglandins [38]. Prostaglandins are generated from arachidonate by the action of cyclooxygenase. It is therefore possible that GAP action is modulated at least in part by the relative rates of conversion of arachidonate to its metabolites by lipoxygenase and cyclooxygenase, following the stimulation of phospholipase A 2 by growth factors. The function of r a s - G A P remains controversial. Its role as a negative regulator has received overwhelming evidential support, e.g. overexpression of GAP suppresses transformation by c - H A - r a s [40], and expression of human GAP in yeast decreases the stimulatory effect of ras on adenylyl cyclase [25]. However, recent observations on the modulation of atrial K ÷ channels [41] also favour a role as an effector for ras. Addition of either p21r~S.GTP complex or r a s - G A P to patches of atrial cell plasma membrane inhibits the coupling of the G k heterotrimeric G protein to muscarinic receptors, and anti-GAP antibodies block the effect, indicating that r a s - G A P is essential for p21 '~ function in this system. The situation is reminiscent of that in the control of protein synthesis, in which ribosomes act as both effector and GAP for EF-Tu [42]. How many GAPs are there? When the proteins from 3T3 cell plasma membranes are separated by high resolution anion exchange chromatography, at least a dozen distinct small GTPases can be resolved [43] and GTP hydrolysis is stimulated by addition of cell extracts for all but one of them (Wolfman and Macara,
The ras superfamilyof molecularswitches
unpublished observations). It therefore seems likely that GAPs exist for most members of the family. To date, a 29,000 rho-specific GAP has been identified by Garrett et al. [44], and GAPs with specificities for smg p21 (rap1, or Krev) [45] and for rab3 [46] have also been detected. The stringency of these specificities remains to be determined, however. The puzzle of GAP function is increased by the fact that ras-GAP will catalyse the GTPase of another member of the family, R-ras, which is not a protooncogene and possesses only about 55% homology to H-, N- and K-ras [47], and will bind with high affinity to rapl. In the latter case, no GTPase catalysis occurs, and the proteins remain associated in what is presumably an inactive complex [48]. This complex may be responsible for the antagonistic effect of rap1 on oncogenic transformation by v-ras [49].
GIPs--GTPase inhibiting protein
In addition to GAPs, factors could exist that decrease kca,. A cytoplasmic protein has now been discovered that performs this function for p21 "~ [50]. The protein (ras-GIP) has a Mr of about 60,000 and is distinct from ras-GAP. Remarkably, it is inhibited by the same phospholipids and fatty acids that activate ras-GAP. It is also potently inhibited by diacylglycerol, which is generated in response to many hormones and growth factors as a consequence of accelerated phosphatidylinositol and phosphatidylcholine turnover. Therefore these lipids and their metabolites may play a key role in the control of the ras switch through modulation of the activities of GAP and GIP. It seems likely that GIPs exist for other members of the small GTPase family, although none have yet been described.
GRFs--Guanine nucleotide releasing factors
The kofr for GDP from p21 ra' and most of the other small GTPases is extremely low (
