TIBS 23 – JULY 1998 REVIEWS

The use of activation of nucleoside triphosphate hydrolysis as a regulatory mechanism is not confined to ; GTPase-activating proteins: ATP-converting are also regulated in this way. Actin stimulates helping hands to complement an ATP turnover by , thereby acting as an ATPase-activating protein (‘AAP’). Similarly, the ATPase activity of the bacterial chaperone DnaK is stimulated by DnaJ. GAPs that are specific for the Ras, Rho, , and Arf subfamilies Klaus Scheffzek, Mohammad Reza of Ras-related GTP-binding proteins have been described (see Box 1, p. 260). Ahmadian and Alfred Wittinghofer Although members of any one subfamily share sequence homology, GAPs from 4 Stimulation of the intrinsic GTPase activity of GTP-binding proteins by different subfamilies do not . Accordingly, GTPase-activating proteins (GAPs) is a basic principle of GTP-binding- they are termed RasGAPs, RhoGAPs, protein downregulation. Recently, the molecular mechanism behind this etc. Modular architecture is commonly used to combine the downregulatory reaction has been elucidated by studies on Ras and Rho, and their activity of GAPs with various other func- respective GAPs. The basic features involve stabilizing the existing catalytic tions, including signalling. Originally machinery and supplementing it by an external arginine residue. This detected genetically in yeast, as negative represents a novel mechanism for active-site formation. regulators of G-protein signalling (RGSs), a large number of GAPs that target GTP HYDROLYSIS IS a key process in in- respective GTP-binding proteins4. GAPs heterotrimeric-G-protein a subunits (Gas) tracellular . are primarily downregulators of the GTP- have been described recently10. Numerous vital processes, including bound form, but some are also active The variability of GAP function is protein synthesis, visual perception, signal transduction molecules. For exam- demonstrated by Tu vesicular and nucleocytoplasmic trans- ple, the Ras-specific p120GAP contains (EF-Tu). EF-Tu possesses an almost un- port, protein targeting, growth control signalling domains that have a dramatic measurable, intrinsic GTPase activity, and differentiation, are controlled enzy- impact on the reorganization of the which is stimulated dramatically by the matically by the conversion of GTP to cytoskeleton5. The importance of GTPase large subunit of the mRNA-primed ribo- 1 GDP and inorganic (Pi) . GTP- regulation is evident from diseases associ- some. It has been proposed that the binding proteins are the molecular ated with mutations in either GTP-binding L7/L12 protein C-terminal domain acts machines that catalyse this reaction. As proteins themselves or GAPs: certain as an EF-TuGAP (Ref. 11). FtsY and Ffh, essential factors in protein biosynthe- GTP-binding-protein mutants are onco- the homologues of the sis, heterotrimeric G proteins or small proteins6; and loss of GAP function (as a signal-recognition particle (SRP) and its Ras-related GTP-binding proteins func- consequence of disruption or mutation of receptor, respectively, catalyse the tion as molecular switches that cycle be- the presumed tumour-suppressor gene) is cotranslational targeting of proteins into tween GTP-bound ‘ON’ and GDP-bound responsible for the disease phenotype in membranes. By a mechanism that ‘OFF’ states. Exchange of the bound GDP type 1 neurofibromatosis patients7. remains to be defined, they stimulate is facilitated by guanine-nucleotide- each other’s GTPase activity12. exchange factors (GEFs), which increase GTPase activation – ten years after Most of the concepts underlying the the dissociation rate of nucleotides. During microinjection studies of Ras structure, function and biochemical This promotes binding of GTP, which (originally termed p21ras) function, mechanism of small GTP-binding pro- allows the GTP-binding proteins to inter- Trahey and McCormick noticed that, teins have been derived from studies act with effector molecules. Hydrolysis contrary to observations in vitro, in vivo on Ras1,2,6. During the past two years, of bound GTP is the timing mechanism ‘Gly12p21 was predominantly guanosine Ras has again come into focus: studies that returns these proteins to their GDP- diphosphate (GDP)-bound because of a on the Ras–RasGAP system have eluci- bound OFF state and thereby completes dramatic stimulation of Gly12p21-associ- dated the mechanism behind GAP what is called the GTPase cycle1,2. ated guanosine triphosphatase (GTPase) catalysis, both biochemically13,14 and GTP-hydrolysis by GTP-binding pro- activity’8. The cytosolic protein respon- in terms of structural biology15,16. teins is intrinsically very slow but can be sible for this increased activity, now This mechanism has been confirmed, accelerated by orders of magnitude upon known as p120GAP, stimulated GTP- independently, by biochemical and interaction with GTPase-activating pro- hydrolysis by normal Ras in vitro, but structural studies of the Rho–RhoGAP teins (GAPs)3, which are specific for their had no effect on oncogenic Ras mutants8,9. system17–21. These two systems are Since the discovery of this first GTPase- therefore the focus of this review. activating protein, many studies have K. Scheffzek, M. R. Ahmadian and shown GTPase activation to be a general Hypotheses on the GTPase-activating A. Wittinghofer are at the Max-Planck-Institut für Molekulare Physiologie, Rheinlanddamm regulatory principle within systems that protein mechanism 201, 44139 Dortmund, Germany. involve GTP-binding proteins and have Investigation of GTPase acceleration Email: alfred.wittinghofer@mpi-dortmund. provided considerable insight into the focused on two questions: which steps mpg.de mechanism of GAP action. of the GTPase reaction are controlled by Copyright © 1998, Elsevier Science Ltd. All rights reserved. 0968 – 0004/98/$19.00 PII: S0968-0004(98)01224-9 257 REVIEWS TIBS 23 – JULY 1998

GAP; and does a universal mechanism aspects of the GTP-bound form, was and RhoGAP can form in the presence of Ϫ exist? In the acto-myosin system, confirmed by the crystal structures of AlF4 , and experiments with Ran and 17 release of Pi is rate limiting and is stimu- Ga–AlFx complexes. In addition, these Rap have yielded similar results . Ϫ Ϫ lated by actin. By contrast, in both structures showed AlF4 to be a square- Ga proteins bind AlF4 , but are also intrinsic and GAP-stimulated reactions, planar entity; together with biochemical targets of specific GAPs, the RGS pro- the release of Pi by GTP-binding proteins data, this observation supported the teins. Some of these (such as RGS4) bind 22,23 Ϫ is not rate limiting . Two models for idea that GDP–AlF4 does not simply act much more tightly to the transition- the mechanism of GAP action have been as a GTP analogue but, rather, mimics state complex (as represented by Ϫ discussed. The first postulates that the the transition state in the GTPase reac- Gai–GDP–AIF4 ) than to the ground- GTP-binding protein is itself an efficient tion29,30. On the basis of numerous stud- state complex (as represented by 10 Ϫ GTPase and that GAP acts catalytically ies on GTP- and ATP-converting enzymes, Gai–GTPgS) . Thus, stabilizing AlF4 to drive the GTP-binding protein into an aluminium fluoride is now considered to binding is not the only indication of enzymatically competent conformation. be a general mimic of the phosphoryl GTPase activation. Experiments using fluorescently labelled group transferred in GTP/ATP hydrolysis GTP analogues designed to test such a and other phosphotransfer reactions31. Structures of GTPase-activating proteins model yielded conflicting results24–26. The observations that Ras–GDP does Numerous GAPs that are specific for

The second hypothesis proposes that not bind AlFx (Ref. 32) and that the helical Rho/Rac/Cdc42 have been identified, GAP participates actively in the process domain in the effector region (as defined including p190 and p50. The RhoGAP- of GTP hydrolysis, possibly by con- in Ras) of Ga subunits acts as an internal like domain of the p85 subunit of phos- tributing a catalytic residue to the active GAP (Ref. 2) prompted studies of the phoinositide 3-kinase comprises a 200- Ϫ 18 site; in this case, stoichiometric effect of RasGAPs on the AlF4 –Ras interac- residue helical protein that is highly amounts of GAP would be needed tion. Indeed, Ras–GDP forms a stoichio- similar in structure to the correspond- 3 for catalysis . The latter model is based metric complex with AlFx in the presence ing domain of p50RhoGAP (Ref. 19) but on the structure of very efficient of NF1-333 or GAP-334, the catalytic has no GAP activity33. Its core contains phosphoryl-transfer enzymes, such as domains of neurofibromin and p120GAP, a four-helix bundle, one face of which adenylate kinase, where a number of respectively13. Complex formation does contains most of the conserved positively charged residues are involved not occur if the invariant arginine residues and has been proposed to be in catalysis27. residue, Arg1391 (from the RasGAP iden- the G-protein- (Fig. 1a)18,19.

tifier motif FLRX3PAX 3P) in NF1-333, is Of the five mammalian RasGAPs de- Experiments with aluminium fluoride mutated to methionine, or if an onco- scribed to date, p120GAP and neurofi- A major breakthrough in the eluci- genic mutant of Ras is used. The experi- bromin are the best studied34. The dation of the nature of GTPase acceler- ments using transition-state analogues catalytic fragment of p120GAP, GAP-334, ation came from studies using fluor- favoured the hypothesis that GAPs ac- is a helical, elongated protein (Fig 1b)15. escently labelled guanine nucleotides tively participate in the process of Ras- The structure defined a central domain and aluminium fluoride (AlFx). AlFx was mediated GTP hydrolysis. The general of 218 amino acid residues that contains originally found to activate heterotrimeric implications of the AlFx experiments, for all the residues conserved among G proteins in their inactive GDP-bound the interaction between GAPs and their RasGAPs and corresponds to a minimal 28 state . The hypothesis that AlFx was respective GTP-binding proteins, are catalytic domain of neurofibromin trapped in the ␥-phosphate-binding site, obvious from studies of the Rho system. that retains full GAP activity (Fig. 1b)35. thereby mimicking at least some Stable complexes between Cdc42–GDP On the basis of a large number of

Figure 1 Structures of GTP-binding proteins and their GTPase-activating proteins (GAPs; drawn using MOLSCRIPT48 and Raster3D49) shown in the ori- entations found for their respective complexes. In each case the nucleotide has been omitted for clarity. The G domains are shown in yellow; additional elements not present in Ras and not associated with GAP activity are shown in pale yellow; the helical domain of G␣ is shown in orange; the common switch I/II regions and the P-loop of the GTP-binding proteins are shown in green. GAPs are shown in red; the positions of the catalytic arginines and the critical glutamines are indicated by cyan and white dots, respectively. Helices belonging to the proposed evo- lutionary module in RasGAP and RhoGAP are shown as solid, pink cylinders. Gly12 in Ras is shown in yellow. (a) Structures of Rac1 (PDB ac- cession code 1MH1), representative of Rho proteins50, and p50RhoGAP (PDB accession code 1RGP)19. (b) H-Ras (PDB accession code 5P21)44 and GAP-334 (PDB accession code 1WER)15. (c) Gai and regulator of G-protein signalling 4 (RGS4; PDB accession code 1AGR)39. 258 TIBS 23 – JULY 1998 REVIEWS biochemical studies of Ras–RasGAP in- teraction9,34, a docking model has been 2+ Ras/Rho proposed, in which two invariant arginine Mg residues (Arg789, Arg903) – candidates for residues involved in GAP catalysis – are brought within reach of the nucleotide15. The catalytic domains of p120GAP P W GMP b and p50RhoGAP have been reported to Pg share no detectable tertiary structural similarity20,21. However, alignment of the d Ð models (on the basis of the way in d Ð which the GAPs communicate with their + GTP-binding protein partners16,21) reveals at least distant structural Gln relationships (Fig. 1a,b). An additional region (present in GAP-334) that includes the C-terminal part of the central domain is missing in RhoGAP, and an a-helical hairpin corresponding to a3c and a4c in GAP-334 is consider- Arg Arg/Lys ably shorter and adopts an orientation that is different from that of the equiva- Secondary lent region in GAP-334. A structural overlay suggests that the helical core Primary 18,19 described for the RhoGAP domain (finger arginine) GAP is a possible evolutionary module (Fig. 1a,b). No apparent similarity between the Figure 2 Rho/RasGAPs and RGS4 (Fig. 1c) has yet Complementation of the active site of the small GTP-binding proteins Ras and Rho by their been detected. From the structural stud- respective GTPase-activating proteins (GAPs). A ‘primary’, finger-arginine residue, together ies, it appears that the GTP-binding pro- with the finger loop, crosses the ‘gap’ between the proteins in order to neutralize develop- teins represent a tema con variazioni ing charges in the transition state of the reaction and stabilize the critical glutamine (variations on a theme). In contrast, residue. A ‘secondary’, positively charged residue, (Arg in RasGAP and Lys in RhoGAP) sta- their GAPs share far less structural simi- bilizes the finger loop. The transition state is shown as having a pentacoordinate phos- phate group, in which the degree of bond making and bond breaking between the trans- larity, although they are not completely ferred phosphate, and the leaving group and nucleophilic oxygen (broken green lines), unrelated (Fig. 1a–c). respectively, determines its associative/dissociative character. GMP, guanosine monophosphate. GTPase-activating-protein communication in three dimensions chain–main-chain interactions. This situa- The situation observed in the

Within the Ras–RasGAP complex tion is depicted schematically in Fig. 2. active site of the Ras–GDP–AlF3–GAP-334 formed by Ras–GDP and GAP-334 in the Mutations of Gly12 and Gln61 in Ras complex was confirmed by the structure presence of AlF3, GAP-334 interacts that are commonly found in human of the corresponding complex between predominantly with the switch regions tumours lock the GTP-binding protein in the catalytic domain of p50RhoGAP and and the P-loop of Ras. This interaction its active conformation, thereby activat- RhoA, which revealed the details of is similar to that proposed in the dock- ing its oncogenic potential36,37; homolo- communication between the two pro- ing model15, and the complex is stabi- gous mutations in other GTP-binding teins21. In this structure, the invariant lized by hydrophobic and hydrophilic proteins show a similar, constitutively arginine (Arg85, which corresponds to contacts (Fig. 1b)16. An exposed loop activated, phenotype. The structure Arg282 of the full-length protein) con- in RasGAP is placed close to the provides a simple explanation for why tacts the nucleotide and a fluoride Ϫ nucleotide, the guanidinium group of these mutants are insensitive to GAP: ligand of the square-planar AlF4 . As in Arg789 interacting with the b phosphate Gly12 lies sufficiently close to the finger the Ras–RasGAP complex, the carbonyl of GDP and AlF3. In addition, the main- loop that even the smallest possible oxygen of this arginine forms a hydro- chain carbonyl oxygen of Arg789 forms amino acid change (to alanine) would gen bond with the amide group of the a hydrogen bond with the side-chain sterically interfere with the geometry of critical glutamine (Gln63), which also Ϫ amide group of the catalytically impor- the transition state. Because Gly12 mu- contacts AlF4 and the nucleophilic tant Gln61 in Ras. Because Arg789 and tants bind to GAP with almost wild-type water molecule. The loop carrying Arg85 the loop point into the active site, they affinity, it appears that larger side is stabilized by an invariant have been called the ‘arginine finger’ chains at position 12 can be tolerated in (Lys122), which appears to play a similar and the ‘finger loop’, respectively16. the Ras–RasGAP ground-state complex role to that of Arg903 of RasGAP (Fig. 2). Ϫ Gln61 also contacts AlF3 and a water but not in the transition state. The appar- Comparison of the AlF4 -bound com- molecule that corresponds to the ent involvement of Gln61 in stabilization plex with the ground-state complex, as attacking nucleophile. As in unligated of the transition state, together with bio- represented by p50RhoGAP(242-residue GAP-334 (Ref. 15), Arg903 of the FLR chemical data13, confirms the notion fragment)–Cdc42–GppNHp (Ref. 20), motif stabilizes the finger loop by side- that Gln61 has a vital role in catalysis. revealed that major structural changes 259 REVIEWS TIBS 23 – JULY 1998

RapGAPs Acc. No. Motif 1 Motif 2 Motif 3 Box 1. Are all GTP- half of the protein’s catalytic domain hsRapGAP M64788 KGFRGGL LQRKRHI EERTRAA (motif 1). Interestingly, the motif-1- dmRapGAP AF023478 KGYRGGL LQRKRHI EQRTRTS binding proteins arginine residues are preceded by an ceRap1GAP P91315 QKYRGGL LQRKRHI AERTRSS aromatic amino acid residue mmSpa1 P46062 EQYRAQL LLRKRHI ATRTRQQ (phenylalanine or tyrosine) in RasGAPs, mmRap1GAP P70204 ESYRAQL LLRKRHI ATRTRQQ switched off by hsRap1GAP D1023058 ESYRAQL LLRKRHI ATRTRQQ RhoGAPs, RapGAPs and YPTGAPs. In rnSpa1 G2555183 EKYRAQL LLRKRHI ATRTRQE arginine fingers? RasGAPs and RhoGAPs, this aromatic Consensus cxaRsxL LxRKRHI xpRTRxx amino acid residue stabilizes the adjacent RasGAPs hydrophobic core and balances the hsNF1 M89914 TLFRGNS MFLRFIN The increasing number of new GTPase orientation of the arginine finger. This rnNF1 M89914 TLFRGNS MFLRFIN hydrophobic stabilization, however, is not ip4BP activating proteins (GAPs) identified for ssGAP Q14644 TIFRGNS IFLRFFG members of subfamilies of GTP-binding realized in all GAPs. In RanGAP and btR-GAP Q28013 TIFRGNS IFLRFFG ArfGAP, and in the presumed GAP for the hsRasGAP Q15283 TIFRGNS VFLRFFA proteins such as Rap, Ran, Arf, Ypt and m elongation factors L7/L12, the invariant rnGAP1 Q29594 AIFRGNS VFLRFFA elongation factors prompted us to mmGAP3 D30734 TIFRGNS IFLRFFA analyse the sequence relationships arginine residues are not preceded by an dmGAP1 U202238 TIFRGNT IFLRFFA between the various GAP subfamilies in aromatic residue. Furthermore, ARD1 is ceGAP M86655 LMFRGNT IFLRFLC order to identify possible catalytic an Arf protein that has an ArfGAP domain btp120 P09851 TLFRATT VFLRLIC in the N-terminal regioni. The invariant rnp120 P50904 TLFRATT VFLRLIC arginine residues. GAPs within a sub- family share high sequence similarity, arginine residue (Arg164) in the ArfGAP hsp120 M23379 TLFRATT VFLRLIC domain of ARD1 has recently been shown scIRA1 M24378 DILRRNS VFLRFIG but the degree of sequence similarity scIRA2 M33779 DILRRNS VFLRFFC between subfamilies is lowa–g. to be critical for the ARD1 GAP activity. Its spSar1 S37449 SLLRANT FFLRFVN replacement by a glycine residue almost Nevertheless, we were able to identify i Consensus xhhRsNo hFLRFhx sequence motifs that contain invariant completely abolishes GTP-hydrolysis , RhoGAPs arginine residues within the GAPs de- which supports the proposal that it is in hsp50GAP Z23024 GIFRRSA scribed for the Rap, Ypt, Ran and Arf fact an arginine finger. mmRip1 Q62172 GVYRVSG proteins [motifs 1, 2 and 3 (see figure)]. Although the evidence is far from GIYRVSG rnCytocentrin G2697022 The ribosomal L7/L12-proteins, which conclusive, and the structural data for the hsRlip76 Q15311 GIYRVSG other GAPs are not available, our analysis rnRalBP1 Q62796 GIYRVSG are proposed to be GAPs for elongation h of GAPs has identified a limited number of hsABR Q13693 GIYRISG factors , contain one invariant arginine hsGRP Q12979 GIYRISG conserved in 38 protein sequences arginine residues that are good hsBcr P11274 GIYRVSG analysed from different bacteria and candidates for arginine fingers. Arginine hsN-Chimerin P15882 GLYRVSG chloroplasts. fingers must meet the following hs§-Chimerin P52757 GLYRVSG The arginine-finger hypothesis is requirements: (1) they are invariant mmp190-B P97393 GLYRVSG within a subfamily of GAPs; (2) they hsRGC1 P98171 GIFRVSG strongly supported by biochemical and structural data on RasGAPs and cannot be replaced, not even by a lysine rnp190GAP P81128 GIYRVSG residue; (3) a mutation in the residue scRga1/Dbm1 P39083 GIYRKSG RhoGAPs showing that the arginine mm3BP-1 X87671 GLFRLAA finger (motif 1) is the primary element drastically impairs GAP activity without consensus GhaRhSG required for GTPase-rate enhancement. changing binding affinity. In other words, The second invariant arginine residue the critical arginine finger should show up YptGAPs if it is broken by site-directed mutagenesis. scGyp1 Ref. 4 FAFRWMN LLMREFQ in RasGAPs (motif 2) stabilizes the ceGyp6 P32806 WLIRWTR LFLRELP finger loop, a function that appears to scGyp7 P48365 FCFRMLL WFHREFE be dependent on a conserved lysine Acknowledgements Gyp7-like P09379 FFFRMLL LFRRELS residue at the equivalent position in We thank Jörg Becker and Dan ylGyp7 E339717 FFFRMLL WFHRELL Consensus FhFRhhx xFhREhx RhoGAPs (Fig. 2 in review). The Rho Cassel for helpful discussions. and Ras systems reveal characteristics RanGAPs that might be common to other xlRanGAP1 G2062659 FTGRLRP AGRNRLE References arginine fingers. The catalytic residues stpRanGAP1 G2623618 FTGRLRS AGRNRLE a Ahmadian, M. R. et al. (1996) J. Biol. Chem. mmRGP1 P46061 FTGRLRS AGRNRLE in RasGAPs and RhoGAPs are located in 271, 16409–16415 hsRanGAP1 P46060 FTGRLRT AGRNRLE the N-terminal portion of the catalytic b Lancaster, C. A. et al. (1994) J. Biol. Chem. mmRNA1 Q60801 FTGRLRS AGRNRLE domains (motif 1). Although RasGAPs 269, 1137–1142 spRNA1 P41391 FTGRVKD CGRNRLE and RhoGAPs do not share obvious c Chen, F. et al. (1997) Proc. Natl. Acad. Sci. scRNA1 P11745 YTSRLVD CGRNRLE sequence homology, their structures U. S. A. 94, 12485–12490 Consensus FTGRh+x AGRNRLE are related. It is therefore likely that d Du, L-L., Collins, R. N. and Novick, P. J. (1998) ArfGAPs the putative arginine fingers in J. Biol. Chem. 273, 3253–3256 hsARD1 P36406 AKHRRVP members of other GAP subfamilies e Bischoff, F. R et al. (1995) Proc. Natl. Acad. Sci. U. S. A. 92, 1749–1753 rnArfGAP Q62848 GRHRGLG are also localized within the N-terminal dmArfGAP G2286211 GKHRSLG f Cukierman, E., Huber, I., Rotman, M. and scGcs1 P35197 GIHRGLG Cassel, D. (1995) Science 270, 1999–2002 scSps18 P32572 NLLRGMG g Bachrach, G., Bar-Nir, D., Banai, M. and scGlo3 P38682 AVHRNMG The invariant arginine residues (highlighted in yellow) Bercovier, H. (1994) FEMS Microbiol. Lett. btIn3PBP O02753 GIHRNIP are aligned. Invariant residues are shown in red; con- 120, 237–240 rnCentaurin Q63629 GIHRNIP served residues are green; homologous residues are h Liljas, A. and Al-Karadaghi, S. (1997) Nat. ssp42IP4 O02780 GIHRNIP shown in blue. Acc. No., accession number; a, aro- Struct. Biol. 4, 767–771 Consensus shHRxhx matic; c, charged; h, hydrophobic; o, Ser or Thr; s, Gly i Vitale, N., Moss, J. and Vaughan, M. (1998) L7/L12 or Ala; ϩ, positive; x, any residue. at, Arabidopsis J. Biol. Chem. 273, 2553–2560 E. coli V00339 KAVRGAT thaliana; bt, Bos taurus; sec, Serinus canaria; ce, S. typhimur. P18081 KAVRGAT Caenorhabditis elegans; dm, Drosophila P. putida P31855 KAVRELT MOHAMMAD REZA AHMADIAN, melanogaster; hs, Homo sapiens; mm, Mus muscu- KLAUS SCHEFFZEK AND M. luteus P02395 KVVREIT lus; rn, Rattus norvegicus; sc, Saccharomyces cere- B. stearoth. P05392 KVVREIT ALFRED WITTINGHOFER Th. maritima P29396 KVVREIT visiae; sp, Schizosaccharomyces pombe; ss, Sus H. pylori P55834 KVVREIT scrofa; stp, Strongylocentrotus purpuratus; xl, Consensus KxVRxhT Xenopus laevis; yl, Yarrowia lipolytica. Email: [email protected]

260 TIBS 23 – JULY 1998 REVIEWS occur upon formation of the transition Њ state. These involve a 20 , rigid, body ro- (a)G domain: Ga (b) G domain: Ras/Rho tation of the two proteins relative to each other21. In the ground-state com- G G helical plex, Arg85 contacts the P-loop of Cdc42 P Arg P and is not in a position that would sup- domain: P P Ga port catalysis. Phosphorus-NMR experi- P P ments using Ras have shown that the Arg Gln Gln presence of RasGAP does not induce Sw I Sw II Sw I Sw II a chemical-shift change in any of the Ras–GppNHp-complex phosphate reso- RGS GAP nances, which would be expected if they were contacted by a positively charged 38 arginine side chain . This, together with Figure 3 biochemical studies, suggests that, as Common principles in the requirements for efficient GTP-hydrolysis. The universal G domain with RhoGAP, the RasGAP finger (shown in yellow) contains a number of functionally important residues, especially a critical arginine is not in a position that would glutamine, that provide the scaffold for the active site. (a) In heterotrimeric-G-protein ␣ sub- accelerate GTP-hydrolysis, in the ground units (Gas; shown in yellow), a catalytically essential arginine is supplied in cis and pos- state of the RasGAP–Ras complex. itioned by an inserted helical domain (shown in orange), which is supplemented by a regu- 16 lator of G-protein signalling (RGS; shown in pink) that stabilizes the switch regions. (b) In As in the Ras–RasGAP complex , the case of the small GTP-binding proteins Ras and Rho (shown in yellow), the arginine larger side chains at the critical Gly12 residue is supplied in trans, by GAPs (shown in pink). The GAP also provides components position in Cdc42/RhoA can be accom- that stabilize the switch regions. Sw, Switch. modated in the ground state but would cause steric hindrance upon transition- state formation (K. Rittinger and S. In RhoGAPs and RasGAPs, invariant themes in order to realize efficient GTP Smerdon, pers. commun.). This explains arginines are critical for interaction with hydrolysis: (1) use of arginine residues why these mutants bind, but are not the GTP-binding protein15,19. In GAP-334, for stabilizing the transition state; (2) sensitive to, GAP, although the confor- Arg789 (Arg1276 in neurofibromin) is ex- stabilization of the switch regions in mational changes observed in these tremely important for GTPase acceler- order to optimize the orientation of the studies did not involve complexes con- ation; even conservative mutation of this catalytic machinery in the GTP-binding taining identical GTP-binding species. residue, to lysine, has dramatic effects. protein, the most important element of Ga proteins differ from small GTP- Arg903 (Arg1301) is less critical, but which is a glutamine residue. In het- binding proteins in that they contain an double mutants such as Arg789®Lys erotrimeric G proteins the critical argi- additional helical domain that is appar- Arg903®Ala are unable to accelerate the nine residue is part of the GTP-binding ently important for positioning a cata- GTPase activity beyond the intrinsic rate protein itself and is supplied in cis with lytically important arginine residue of Ras-mediated GTP hydrolysis14. These an extra domain necessary for orien- such that it contacts the nucleotide and observations are in very good agreement tation and an extra protein (RGS) required a fluoride ligand in the complex with with the structure of the Ras–GAP-334 for proper alignment of the entire Ϫ GDP–AlF4 (Refs 29, 30). The crystal complex. Arg789 points into the active machinery (Fig. 3a). For Ras-related structure of the complex with RGS4 re- site and neutralizes negative charges, GTP-binding proteins, this residue is vealed that RGS4 has a helical architec- and, by means of the finger loop, anchors supplied as an arginine finger (i.e. ture (Fig. 1c)39 and that it predominantly Gln61. This stabilizes the transition state in trans) by the respective GAP (Fig. 3b). contacts the switch regions of Ga. RGS4 (as represented by Ras–GDP–AlF3). As Stabilization of the switch region is is therefore believed to stabilize the one would expect, mutation of the invari- best documented in Ras, where the transition state by reducing the mobility ant ‘finger’ arginine in p190 has a detri- Gln61 region is highly mobile in the of these regions, and thus acts similarly mental effect on catalysis in Rho isolated protein44–46. to RasGAPs and RhoGAPs. proteins41. The finger loop is stabilized by Although the basic features of GAP- Arg903. Interestingly, an invariant lysine catalysed GTPase reactions have been The mechanism of GTPase activation residue (Lys122) seems to play the role of worked out, many questions remain. For Arginine and lysine residues play this residue in the RhoGAP system (Fig. example, why do most known phos- critical roles in phosphoryl-transfer reac- 2). Gln61 apparently positions the water phoryl-transfer enzymes prefer arginine tions. Positively charged under physio- molecule for nucleophilic attack and sta- residues for catalysis? Why do we find dif- Ϫ logical conditions, these residues are bilizes the transferred phosphoryl group. ferently coordinated AlFx (AlF4 or AlF3) able to neutralize negative charges that The structures of Rho and Ras in com- in the active sites? We also do not know develop on the transferred phosphoryl plexes with their respective GAPs did how the transition-state mimic containing group or the leaving group oxygen, not reveal a general base for the activation AlFx, which is kinetically and thermody- depending on whether the mechanism of the nucleophilic water molecule, which namically very stable, is related to the is associative or dissociative40. In addi- is consistent with the notion that the high-energy state of the real transition tion, their side chains are comparatively proposed mechanism of - state. In addition, do AAPs such as DnaJ long, which allows them to bridge larger assisted catalysis42 also applies to the work in a similar manner? What is the distances at the protein–protein-complex GAP-catalysed reaction43. mechanism of phosphoryl transfer in interfaces. In nucleoside monophosphate By comparing the roles of GAPs with myosin, which has a non-actin-stimulated kinases, arginine residues are essential those of RGSs, we can conclude that single-turnover ATPase of about 100 sϪ1, for catalysis27. nature has developed at least two but does not have a unique positively 261 REVIEWS TIBS 23 – JULY 1998 charged amino acid residue in the active Another reason for the physical separ- 15 Scheffzek, K. et al. (1996) Nature 384, site47? Clever studies have to be designed ation of the GTPase active centre could 591–596 16 Scheffzek, K. et al. (1997) Science 277, 333–338 in order to answer these questions. be a requirement for efficient and ineffi- 17 Ahmadian, M. R., Mittal, R., Hall, A. and cient GTPase machineries under differ- Wittinghofer, A. (1997) FEBS Lett. 408, A heterodimeric enzyme ent physiological conditions. In order 315–318 18 Musacchio, A., Cantley, L. C. and Harrison, S. C. Transition-state stabilization is the to avoid unnecessary GTP turnover (1996) Proc. Natl. Acad. Sci. U. S. A. 93, basic principle of . because of nucleotide exchange and 14373–14378 In GTP-binding proteins, a substrate- GTP-hydrolysis, both reactions are very 19 Barrett, T. et al. (1997) Nature 385, 458–461 20 Rittinger, K. et al. (1997) Nature 388, 693–697 binding site is formed by amino acid slow in the absence of GEFs and GAPs, 21 Rittinger, K. et al. (1997) Nature 389, 758–762 residues derived from fingerprint se- respectively. Only when the signalling 22 John, J., Frech, M. and Wittinghofer, A. (1988) quence motifs, and the catalytic machin- mechanisms of the cell require fast Ras J. Biol. Chem. 263, 11792–11799 ery is in principle able to perform GTP activation or deactivation do external 23 Nixon, A. E. et al. (1995) Biochemistry 34, 15592–15598 cleavage at a rate significantly greater factors increase the rates of these reac- 24 Neal, S. E., Eccleston, J. F. and Webb, M. R. than that of spontaneous hydrolysis in tions. It appears that the newly discov- (1990) Proc. Natl. Acad. Sci. U. S. A. 87, water. However, this rate is increased ered principle that an inefficient GTPase 3652–3565 25 Rensland, H., Lautwein, A., Wittinghofer, A. and even more upon interaction with centre is complemented, under certain Goody, R. S. (1991) Biochemistry 30, GAPs, in a way that represents a novel physiological conditions, by a GAP is a 11181–11185 biological principle. universal principle in the regulation of 26 Moore, K. J. M., Webb, M. R. and Eccleston, J. F. (1993) Biochemistry 32, 7451–7459 What is special about this enzyme? many if not all GTP-binding proteins. 27 Müller, C. W. and Schulz, G. E. (1992) J. Mol. Nature has developed numerous strat- Biol. 224, 159–177 egies for optimizing metabolism and for Acknowledgements 28 Chabre, M. (1990) Trends Biochem. Sci. 15, regulating enzyme activities, such as 6–10 We thank all our colleagues for sharing 29 Coleman, D. E. et al. (1994) Science 265, allosteric control, proteolytic activation, skill, knowledge and discussions through- 1405–1412 reversible covalent modification and acti- out our efforts to understand GTP hy- 30 Sondek, J. et al. (1994) Nature 372, 276–279 vation by control proteins. Enzymes are 31 Wittinghofer, A. (1997) Curr. Biol. 7, drolysis and apologize for not being able R682–R685 also commonly composed of two or more to reference comprehensively their con- 32 Kahn, A. (1991) J. Biol. Chem. 266, subunits, and the active site can be tributions because of space limitations. 15595–15597 shared between subunits – a strategy that 33 Lamarche, N. and Hall, A. (1994) Trends Genet. We thank Frank Schmitz for help with 10, 436–440 optimizes regulation and formation of visualization of the structural models and 34 Wittinghofer, A., Scheffzek, K. and Ahmadian, the induced-fit conformation required figure preparation, Heiner Schirmer for M. R. (1997) FEBS Lett. 410, 63–67 35 Ahmadian, M. R. et al. (1996) J. Biol. Chem. for catalysis. In the Ras–RasGAP/ comments on the manuscript, Rita Rho–RhoGAP system, the active site is 271, 16409–16415 Schebaum for secretarial assistance, Greg 36 Seeburg, P. H. et al. (1984) Nature 312, 71–75 shared by two completely different pro- Petsko for discussions on the mechanis- 37 Der, C. J. et al. (1986) Cell 44, 167–176 teins. They come together as a transient tic and evolutionary aspects of GAPs, 38 Geyer, M. et al. (1996) Biochemistry 35, heterodimeric enzyme, in order to cata- 10308–10320 and Ken Holmes for continuous 39 Tesmer, J. J. G., Berman, D. M., Gilman, A. G. lyse GTP-hydrolysis, and separate after support. K. S. thanks the Peter and and Sprang, S. R. (1997) Cell 89, 251–261 the job is done. Traudl Engelhorn Stiftung and the 40 Maegley, K. A., Admiraal, S. J. and Herschlag, D. Separation of components of the enzy- (1996) Proc. Natl. Acad. Sci. U. S. A. 93, National Neurofibromatosis Foundation 8160–8166 matic machinery is of major physiologi- for support. 41 Li, R., Zhang, B. and Zheng, Y. (1997) J. Biol. cal importance. p120GAP is a cytosolic Chem. 272, 32830–32835 protein and becomes localized to 42 Schweins, T. et al. (1995) Nat. Struct. Biol. 2, the plasma membrane by binding References 36–44 1 Bourne, H. R., Sanders, D. A. and McCormick, F. 43 Schweins, T. et al. (1996) Biochemistry 35, (through its SH2 domain) to activated re- (1990) Nature 348, 125–132 14225–14231 ceptor tyrosine kinases, such as the 2 Bourne, H. R., Sanders, D. A. and McCormick, F. 44 Pai, E. F. et al. (1990) EMBO J. 9, 2351–2359 platelet-derived-growth-factor receptor. (1991) Nature 349, 117–127 45 Milburn, M. V. et al. (1990) Science 247, 3 Gideon, P. et al. (1992) Mol. Cell. Biol. 12, 939–945 Considering the low affinity of p120GAP 2050–2056 46 Kraulis, P. J. et al. (1994) Biochemistry 33, for Ras–GTP, which is in the micromolar 4 Boguski, M. S. and McCormick, F. (1993) 3515–3531 range, and the low concentrations of the Nature 366, 643–654 47 Smith, C. A. and Rayment, I. (1996) 5 McGlade, J. et al. (1993) EMBO J. 12, Biochemistry 35, 5404–5417 reaction partners, it seems reasonable to 3073–3081 48 Kraulis, P. J. (1991) J. Appl.Crystallogr. 24, assume that GAP acts on Ras only when 6 Lowy, D. R. and Willumsen, B. M. (1993) Annu. 946–950 both are localized to the plasma mem- Rev. Biochem. 62, 851–891 49 Merrit, E. A. and Bacon, D. J. (1997) Methods brane (where the local concentration in 7 Gutmann, D. H. and Collins, F. S. (1993) Neuron Enzymol. 277, 505–524 10, 335–343 50 Hirshberg, M., Stockley, R. W., Dodson, G. and the two-dimensional space is very high). 8 Trahey, M. and McCormick, F. (1987) Science Webb, M. R. (1997) Nat. Struct. Biol. 4, 147–152 Neurofibromin is predominantly cyto- 238, 542–545 plasmic and has been found attached to 9 Polakis, P. and McCormick, F. (1993) J. Biol. Chem. 268, 9157–9160 microtubules, where it might not be 10 Dohlman, H. G. and Thorner, J. (1997) J. Biol. STOP PRESS available for Ras downregulation under Chem. 272, 3871–3874 normal conditions. The affinity of neuro- 11 Liljas, A. and Al-Karadaghi, S. (1997) Nat. Following acceptance of this manuscript, three fibromin for Ras–GTP is about 50-fold Struct. Biol. 4, 767–771 other reports of structural similarities between 12 Powers, T. and Walter, P. (1995) Science 269, RhoGAPs and RasGAPs have appeared. higher than that of p120GAP, which indi- 1422–1424 Bax, B. (1998) Nature 392, 447–448 cates a differential requirement for the 13 Mittal, R., Ahmadian, M. R., Goody, R. S. and Rittinger, K. et al. (1998) Nature 392, 448–449 Wittinghofer, A. (1996) Science 273, 115–117 GAP reaction. Other GAP isoforms might Calmels, T. G. et al. (1998) FEBS Lett. 426, 14 Ahmadian, M. R., Stege, P., Scheffzek, K. and also exhibit differential behaviour; how- Wittinghofer, A. (1997) Nat. Struct. Biol. 4, 205–211 ever, detailed knowledge is limited. 686–689 262