|Home | About | Journals | Submit | Contact Us | Français|
Widespread utilization of small GTPases as major regulatory hubs in many different biological systems derives from a conserved conformational switch mechanism that facilitates cycling between GTP-bound active and GDP-bound inactive states under control of guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs), which accelerate slow intrinsic rates of activation by nucleotide exchange and deactivation by GTP hydrolysis, respectively. Here we review developments leading to current understanding of intrinsic and GAP catalyzed GTP hydrolytic reactions in small GTPases from structural, molecular and chemical mechanistic perspectives. Despite the apparent simplicity of the GTPase cycle, the structural bases underlying the hallmark hydrolytic reaction and catalytic acceleration by GAPs are considerably more diverse than originally anticipated. Even the most fundamental aspects of the reaction mechanism have been challenging to decipher. Through a combination of experimental and in silico approaches, the outlines of a consensus view have begun to emerge for the best studied paradigms. Nevertheless, recent observations indicate that there is still much to be learned.
Regulated hydrolysis of GTP by small GTPases (also known as small G proteins) is a biologically important example of the more general class of phosphoryl transfer reactions in aqueous solution and structured protein environments1-3. The deceptively simple reaction chemistry involves nucleophilic attack by a water molecule on the terminal γ phosphate of a GTP substrate and cleavage of the phospho-monoester bond to produce GDP (the leaving group) and inorganic phosphate. Unlike classical enzymes that evolved to rapidly convert weakly bound substrates into products with catalytic efficiencies (kcat/KM) in some cases approaching diffusion controlled limits, the majority of GTPases, including nearly all small GTPases, evolved under very different selective pressures to bind GDP/GTP with very low dissociation constants (KD typically in the nanomolar-picomolar range) and slowly hydrolyze GTP with turnover numbers rarely exceeding 10-1 s-1 and more generally in the range of 10-3-10-6 s-1 or even lower4-7. With co-evolution of biosynthetic pathways to maintain free GTP at steady state concentrations in excess of free GDP by at least 10 fold8, the result is a superfamily of enzymes that slowly exchange GDP for GTP and slowly hydrolyze GTP to GDP in an effectively unidirectional cycle between metastable GDP- and GTP-bound states characterized by structurally and dynamically distinct conformations (Fig. 1A). Indeed, the timescale for interconversion between nucleotide-bound states is generally much slower than that of the relevant biological processes such that two major classes of regulatory proteins co-evolved to control exchange and hydrolysis. Guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) accelerate GDP/GTP exchange and GTP hydrolysis by orders of magnitude, with catalytic turnover numbers exceeding the intrinsic rates by as much as 105 fold9 or even more in cases where the intrinsic rates are particularly low.
The general functional significance of this remarkable switch mechanism becomes apparent with the co-evolution of another major class of proteins known as ‘effectors’, which selectively bind the GTP-bound conformation to couple the GTPase cycle with the downstream molecular machinery involved in signaling, cytoskeletal dynamics, trafficking, nuclear transport and many other cellular processes10. Thus, the GTPase cycle emerged as one of the most highly conserved switch mechanisms, in which cell autonomous or external signals coupled through GEFs and GAPs are the primary inputs controlling transitions between GDP-bound ‘inactive’ and GTP-bound ‘active’ conformations while selective effector binding to the GTP-bound conformation is the primary output mediating physiological responses. The remarkable versatility of the GTPase cycle as a widespread regulatory mechanism for numerous biological processes reflects the structural diversity of the GTPases, GEFs, GAPs and effectors, which contributes to high functional specificity at the level of molecular interactions. Finally, the co-evolution of additional ancillary proteins, including GDP-dissociation inhibitors (GDIs) and related factors that selectively bind the GDP-bound conformation, is crucial for post-translational modification of GTPases as well as delivery to sub-cellular sites of action.
This review focuses on current understanding of the molecular mechanisms, structural underpinnings and chemical energetics/dynamics related to the tightly regulated GTP hydrolytic reaction in small GTPases. Selected examples are used to illustrate general conceptual frameworks as well as variations and exceptions that have extended, shifted or redefined established paradigms. Given this emphasis, other very important aspects of the GTPase cycle including relevant biology and associated diseases are largely neglected. The interested reader can find further information on related topics in many excellent reviews in this issue and elsewhere1,2,10-21.
As illustrated in Fig. 1B, small GTPases of the Ras superfamily share a common architecture with a core ‘GTPase domain’ consisting of five helices (α1-α5) surrounding a six-stranded beta sheet (β1-β6)22. At the primary structural level, five highly conserved ‘G motifs’ (G1-G5) encode invariant residues that mediate critical contacts with the nucleotide as well as an essential Mg2+ ion required for high affinity nucleotide binding and hydrolytic activity4. The G1 motif (GxxxxGKS/T) encodes the phosphate binding loop (P-loop) characteristic of nucleotide binding proteins. The cryptic G2 motif (xTx) contains an invariant threonine residue that mediates hydrogen bonding interactions with γ phosphate and is also coordinated by the Mg2+ ion in the GTP-bound state. In addition to invariant aspartic acid and glycine residues involved in stabilization of Mg2+ binding and hydrogen bonding with the γ phosphate, respectively, the G3 motif (DxxG) is typically followed by a glutamine residue that is conserved in many families (including Ras, Rho, Rab, Arf, Arl, and Gα subunits) and plays a particularly important role in the intrinsic and, with notable exceptions discussed below, GAP-catalyzed hydrolytic reactions. Finally, the G4 (N/TKxD) and G5 (xAx) motifs encode residues important for base contacts and guanine nucleotide binding specificity.
N- and C-terminal extensions to the GTPase core vary considerably between and even within families. Although typically too short to form independent domains, the variable extensions include distinctive sequence motifs or other structural features that confer additional functionality, most notably related to subcellular localization or binding specificity for effectors, GEFs, GAPs and GDIs. Ras, Rho and Rab GTPases posses C-terminal cysteine motifs that specify post-translational modification of the cysteine thiol group to form a stable thioether linkage to either one (CAAX motifs) or two (CC, CXC motifs) non-polar farnesyl (15 carbon) or geranyl geranyl (20 carbon) moieties consisting of three or four isoprenoid units, respectively23-27. The prenyl groups anchor these GTPases at membrane surfaces but confer little if any targeting specificity, which depends on other elements in the variable regions and/or molecular interactions with GDIs, GEFs, GAPs, and effectors12,16,28. Moreover, since many small GTPases are involved in or otherwise influenced by membrane dynamic processes (e.g. vesicle budding/transport/fusion) their steady state distributions can differ from their initial sites of membrane targeting.
Membrane targeting of all Arf GTPases depends on partitioning of an N-terminally myristoylated (14 carbon) amphipathic helix that inserts into the hydrocarbon core 29-31. Some Arl GTPases are also known or predicted to be N-terminally myristoylated (e.g. Arl1, Arl4, Arl5, Arl11, and Arl14)32-34 whereas others either lack the essential glycine at position 2 or are otherwise predicted to be poor substrates for N-myristoyl transferase (Arl2, Arl3, Arl8, Arl9, Arl10, Arl13, Arl15, and Arl16)31,35. Arl2 and Arl3 have the required glycine at position 2; however, there is no evidence of myristoylation even though the N-terminal amphipathic helix contributes to membrane association36. A few non-myristoylated Arl GTPases are instead N-terminally acetylated, a modification that is functionally important and (e.g. yeast Arl3 and fly Arl8)37-39. Whether myristoylated or not most Arl GTPases have an N-terminal amphipathic helix that is known or expected to contribute to membrane partitioning31,36,40,41. As with prenylated GTPases, membrane targeting and steady state localizations of Arfs and Arls are dependent on multiple molecular interactions.
Comparisons of GDP- vs. GTP-bound crystal structures of Ras42-44, and subsequently others from every major family45-49, revealed prominent conformational differences within two localized ‘switch regions’, referred to as the switch I (α1/β2 loop) and switch II (β3/α2 loop, α2, and α2/β4 loop). To a first approximation, the switch regions are flexible in the GDP-bound state but adopt relatively well ordered conformations in the GTP-bound state due to hydrogen bonding interactions between the γ phosphate and: i) the main chain NH as well as side chain hydroxyl of the invariant threonine in the switch I G2 motif; and ii) the main chain NH of the invariant glycine in the switch II DxxG motif. In addition, waters displaced from two sites in the coordination sphere of the Mg2+ ion in the GDP-bound state are replaced by a non-bridging oxygen from the γ phosphate and the side chain hydroxyl oxygen of the invariant switch I threonine in the GTP-bound state. The nucleotide-dependent conformational changes in Arf GTPases include a distinctive ‘myristoyl switch’ involving a two residue shift of the β2 strand known as the ‘interswitch toggle’ that displaces the myristoylated N-terminal amphipathic helix to facilitate membrane targeting46.
The effects of oncogenic mutations in Ras provided insights into the GTPase mechanism and implicated the conserved DxxGQ glutamine as a critical catalytic residue50. The Q61L mutation, for example, abolishes intrinsic as well as GAP-accelerated hydrolysis and is constitutively activating. The analogous QL substitutions have been widely used to generate constitutively active forms of small GTPases. Another common oncogenic mutation in the P-loop, G12V, has a similar effect and in fact every Gly 12 substitution except proline is oncogenic51. The lack of oncogenicity for G12P Ras is likely related to the observation that its intrinsic hydrolytic rate is slightly higher than that of wild type Ras despite being GAP insensitive whereas other mutations typically have lower intrinsic rates52,53. A second important insight came from comparison of Ras with the Gα subunits of heterotrimeric G proteins, which have considerably higher intrinsic hydrolytic rates on the seconds rather than minutes timescale. In addition to conserving the catalytic glutamine equivalent to Gln 61 in Ras, Gα subunits also posses a catalytic switch I arginine that is absent in Ras and other small GTPase families4. It was hypothesized that GAPs for small GTPases including Ras might supply a catalytic arginine in trans.
Metal fluoride coordination complexes (e.g. AlF3, AlF4-, BeF3- and MgF3-) have been extensively used in combination with nucleotide diphosphates to mimic the ground or transition states for phosphoryl transfer reactions54. BeF3- favors tetrahedral coordination geometry with a β phosphate oxygen of GDP as the fourth ligand and should in principle approximate the GTP ground state55. Interestingly, the stability of RasGAP complexes with Ras-GDP-BeF3- are less sensitive to oncogenic substitutions than complexes with Ras-GppNHp, suggesting that the two ground state GAP complexes though similar are not identical56. Aluminum fluoride forms mixed coordination species and trigonal bipyramidal (AlF3) as well as octahedral (AlF4-) geometries have been reported in crystal structures of GDP-bound GTPases alone (Gα subunits57,58) and in complexes with GAPs59-66. In both geometries, the equatorial sites are occupied by fluoride ions while the apical sites are occupied by a β phosphate oxygen of GDP and a water. The possibility that aluminum fluoride complexes might be transition state mimetic was suggested on the basis of biochemical evidence67,68 and strongly supported by crystal structures of GDP-bound Gα subunits activated with aluminum fluoride57,58. MgF3- favors trigonal bipyramidal geometry and has a net charge approximating that expected for the transition state and thus may be a better transition state mimetic than octahedral AlF4- or charge neutral AlF3 complexes69. Here it is worth noting that the electron density in a number of medium resolution crystal structures has been interpreted as AlF3; however, the atomic scattering factors and expected geometry for AlF3 are nearly indistinguishable from MgF3- and it is possible that MgF3- is the actual chemical species since Mg2+ is also present in the crystallization buffers69,70.
Although metal fluorides were originally shown to activate heterotrimeric G proteins and stabilize complexes of small GTPases with GAPs55,71,72, Cdc42-GDP binds effectors in the presence of aluminum or beryllium fluorides73 and a GDP-bound Ras variant, in which the switch I Tyr 32 was mutated to tryptophan to monitor activation by intrinsic tryptophan fluorescence, binds BeF3- with a high Kd in the millimolar range74. It is not known whether the Y32F mutation influences BeF3 binding. However, a Rab5 variant capable of binding aluminum fluoride in the absence of a GAP was generated by substituting proline for alanine at a P-loop position proximal to the hydrolytic site 75. BeF3- also binds to the Ran-GDP complex with Ran binding protein in the absence of RanGAP76. Thus, the affinity of metal fluorides for GDP-bound GTPases in the presence and absence of GAPs depends on multiple determinants and not just the presence of catalytic arginine/glutamine residues.
The remarkable stability of phosphomonoester bonds in aqueous solution derives from the net negative charge of the ionizable phosphoryl group at physiological pH, which impedes nucleophilic attack3. Reduction of negative charge on the terminal phosphate through electrostatic interactions with the divalent Mg2+ cation, the P-loop lysine and often a cis or trans arginine is a general feature exploited by nucleotide triphosphatases to reduce this barrier; however, other factors are also important, including differences in the stereochemistry of polar interactions, Mg2+ coordination and steric strain in the ground vs. transition states1,18. From a theoretical perspective, GTP hydrolytic mechanisms are generally considered with respect to the lowest energy pathway over a potential surface defined by reaction coordinates for phosphomonoester bond breakage and bond formation with the nucleophilic water. Three extreme cases (Fig. 2) illustrate the range of possible trajectories over the high energy transition state/intermediate region: i) a fully dissociative transition state intermediate for which bond breakage precedes bond formation; ii) a fully associative transition state intermediate for which bond formation precedes bond breakage; and iii) a SN2-like transition state for which bond breakage and formation are concerted. In fact, the reaction pathways can be mixed and the possibilities are even more complex when the low lying d orbitals of phosphorous or pre/post-bond breakage/formation steps are considered1,77. The problem has been intensively investigated through experimental as well as theoretical approaches, with considerable debate reflecting the inherent complexity of the underlying chemical energetics and the dependence on the detailed structural and chemical environment of the hydrolytic site.
Structural and other experimental observations have been interpreted to support mainly dissociative or associative mechanisms as well as concerted mechanisms with substantial dissociative or associative character in GTPases with or without GAPs14,64,75,78-80. Computational studies on intrinsic and GAP catalyzed hydrolytic reactions in small GTPases have focused primarily on Ras, for which multiple constitutively-activating oncogenic substitutions have well established effects on hydrolytic activity52,81. Detailed thermodynamic analysis of linear free energy relationships and ab initio quantum mechanical(QM)/molecular mechanical(MM) free energy calculations suggested that experimental observations do not distinguish between associative, dissociative or concerted mechanisms82-84. The calculations further suggested that the lowest energy reaction pathways for both associative and dissociative mechanisms traverse flat transition regions such that dissociative (metaphosphate) or associative intermediates would be weakly populated and therefore difficult to detect. Consequently, mechanistic details can differ significantly between solution, intrinsic GTPase and GAP-catalyzed reactions, and need to be worked out for each case using theoretical calculations subject to experimental observations.
A related and also much debated mechanistic problem concerns transfer of the proton from the nucleophilic water to an acceptor, which can in principle be an amino acid residue, the γ phosphate or a second water14. Based on the crystal structure of active Ras-GppNHp, it was noted that the DxxGQ glutamine (Gln 61) was sufficiently close to potentially activate the nucleophilic water43. However, the glutamine carboxamide group is a very weak base and free energy perturbation calculations combined with electrostatic considerations suggested that Gln 61 is unlikely to function as a general base due to the high barrier for protonation85. An alternative proposal based on analysis of linear free energy relationships suggested that the γ phosphate, which has a pKa near 6.5 in solution, might function as a general base in a substrate-assisted mechanism86. Subsequent QM/MM calculations suggested that proton transfer to the γ phosphate may be indirectly mediated by Gln 61 in the GAP reaction and involve a second water assisted by Gln 61 in the absence of the GAP70,87. Recent QM/MM calculations including intermediates after bond breakage revealed a seemingly counterintuitive possibility that proton transfer to Gln 61 results in the imide tautomer form of the carboxamide group and that regeneration of the amide form may be the highest energy chemical barrier along the GAP-accelerated hydrolytic pathway77.
Regardless of the proton transfer pathway, analyses of QM/MM free energy surfaces suggest that the acceleration provided by the GAP is likely due to a combination of electrostatic stabilization and protein conformational changes influencing the transition region preceding proton transfer83,84. FTIR studies combined with QM/MM simulations indicate that the non-bridging phosphate oxygens of GTP adopt a low energy staggered configuration in solution compared to a high energy eclipsed configuration for the non-bridging oxygens of the β-γ phosphates when bound to Ras and eclipsed configurations for the non-bridging oxygens of all three phosphates in the GAP complex88. These staggered-to-eclipsed configurational changes correlate with increases in the length of the scissile (Pγ–Oβγ) bond and are predicted to contribute to ground state destabilization.
Finally, another issue relates to the protonation state of the γ phosphate in Ras-GTP, which is thought to be fully deprotonated based on the pH dependence of 31P NMR chemical shifts52,86,89. In a recent neutron scattering structure of ground state Ras-GppNHp exchanged with D2O, however, a deuteron is observed on one of the γ phosphate oxygens at a calculated pD of 8.490. This apparent discrepancy may be due to differences in the pKa of the γ phosphate, which is higher for GppNHp than GTP in solution52. Given the potential mechanistic implications, the question of whether the γ phosphate is protonated in Ras-GTP with or without RasGAP would benefit from clarification. It might be interesting, for example, to compare the protonation of state of constitutively active Ras mutants loaded with GTP vs. GppNHp.
Beyond the chemical steps, there are pre- and post-transition state barriers related to the conformational transitions in the GTPase and GTPase/GAP complex on the pathway from the ground to transition state and from the transition state to products. The ground to transition state conformational changes could in principle be rate limiting, although it is difficult to experimentally distinguish chemical steps from conformational changes if the latter are rate-limiting. In some ground state crystal structures of active GTPases, the switch I region adopts a ‘closed’ conformation incompatible with the insertion of catalytic residues into the hydrolytic site, suggesting that transition to a open conformation either precedes or accompanies formation of the initial GAP complex. For example, a switch I tyrosine in Rab1 (equivalent to Ras Tyr 32) interacts with the γ phosphate and occludes the hydrolytic site, except when trapped by crystal contacts in an open conformation91. The equivalent switch I tyrosine in Ran also blocks the active site and detailed studies combining X-ray crystallography, FTIR and MD simulations indicate that the closed conformation contributes substantially to the slow rate of GTP hydrolysis compared to Ras76. One role of the GAPs in these instances may be to stabilize an open switch I conformation compatible with insertion of catalytic residues. In addition, allosteric modulation at a remote site by small molecules including calcium acetate in various crystal forms of Ras-GppNHp has been observed to shift the equilibrium between catalytically competent and resistant conformations, suggesting that global conformational changes in the ground state, potentially modulated by physiological interactions, may also play an important role92,93.
On the product side, there are at least three barriers, in particular relaxation to the GDP-bound conformation, release of phosphate after hydrolysis and, for GAP accelerated reactions, dissociation of the GTPase-GAP complex94. Using time resolved FTIR to monitor vibrational modes for a number of GTPase-GAP complexes18, it has been possible in certain cases to distinguish the hydrolytic event from subsequent phosphate release, with concomitant accumulation of a GDP + Pi intermediate state. In the case of Ras/RasGAP, for example, phosphate release is roughly 8 fold slower than hydrolysis95,96. Phosphate release is also rate-limiting for Rap1b with either Rap1GAP or the dual specificity Ras/Rap1GAP97 whereas the hydrolytic step is evidently rate limiting for RhoA/RhoGAP, Ran/RanGAP, and Rab1b/TBC1D2060,98. Relaxation of the switch regions to the GDP-bound conformation is presumably the major post-hydrolysis barrier preceding phosphate release. For GAP-catalyzed reactions, dissociation of the GAP-GTPase complex may limit the rate of conformational relaxation and consequently the rate of phosphate release. Whether conformational relaxation and/or phosphate release can precede GAP dissociation is unclear; however, the crystal structure of ArfGAP in complex with Arf1-GDP99 suggests that this possibility warrants further investigation.
A key insight facilitating atomic resolution crystallographic studies of GAP complexes with small GTPases involved the discovery that aluminum fluoride stabilizes complexes between the GDP-bound forms of several small GTPases including Ras and their cognate GAPs72. The first crystal structure of a small GTPase-GAP complex was determined for Ras-GDP-AlF3 in complex with the catalytic domain of RasGAP64. The structure revealed polar and electrostatic interactions involving Gln 61 and an ‘arginine finger’ from the GAP that have similar stereochemistry to interactions mediated by the catalytic glutamine and arginine residues in Gα subunits, despite different aluminum fluoride coordination geometries and different arginine rotomer configurations (Fig. 3A and B). Consistent with an important role in catalysis, mutation of the arginine finger reduces GAP activity by three orders of magnitude9. The RasGAP-Ras-GDP-AlF3 structure further clarified how oncogenic substitutions interfere with formation of the transition state. The frequently occurring G12V oncogenic mutation, for example, introduces van der Waals conflicts with both the arginine finger and cis-glutamine (Fig. 3C). The van der Waals conflict with the cis-glutamine involves the Cβ and may help to explain why all Gly12 mutations except proline are oncogenic51. Indeed, as noted above, the intrinsic activity of G12P Ras is slightly higher than that for wild type Ras52,53.
The crystal structure of RhoA-GDP-AlF3 in complex with the catalytic domain of RhoGAP revealed an arginine finger/cis-glutamine arrangement similar to Ras63 (Fig. 3D) and, unexpectedly, a related overall architecture for the RhoGAP and RasGAP catalytic domains despite limited sequence homology100. The crystal structure of RhoGAP in complex with Cdc42-GppNHp suggests that the arginine finger and DxxGQ glutamine are not engaged with the γ phosphate in the ground state complex101. From these and subsequent crystallographic studies of RhoGAP in complex with RhoA-GDP-MgF3-69, which more closely approximates the expected transition state geometry than the AlF4- complex, as well as a catalytically compromised Cdc42GAP (arginine finger mutated to alanine) in complex with Cdc42-GDP-AlF3102, a coherent paradigm emerged for transition state stabilization involving a cis-glutamine and trans-arginine finger that mediate equivalent polar/electrostatic interactions even though the arginine finger inserts from different directions and adopts different rotomer configurations in RhoGAP vs. RasGAP complexes.
Since the majority of small GTPase families conserve the DxxGQ glutamine, it was anticipated that their GAP complexes would conform to the cis-glutamine/trans-arginine paradigm. Although this is indeed the case for several GTPase families, there are nevertheless differences in structural detail that have implications for the catalytic mechanism. For example, the retinitis pigmentosa protein 2 (RP2) GAP in complex with Arl3-GDP-AlF3 revealed an arginine finger/cis-glutamine configuration similar to that observed for Ras/Rho GAPs65 (Fig. 3E). However, the cis-glutamine is not optimally positioned to donate a hydrogen bond to an equatorial oxygen in the transition state due to a van der Waals conflict with an aspartic acid residue at the P-loop position equivalent to Ras Gly 12. An arginine finger/cis-glutamine configuration was also observed in the crystal structure of the zinc finger GAP domain-ankyrin repeats of the ArfGAP ASAP3 in a complex with Arf6-GDP-AlF4- that was stabilized by generating a fusion construct103 (Fig. 3F). Although the resolution of the latter structure is too low for detailed stereochemical analysis, the same van der Waals conflict is predicted. Notably, however, both GAPs supply the arginine finger from an angle that circumvents potential steric conflict with the P-loop aspartic acid.
The hydrolytic reactions in some small GTPase/GAP systems proceed in the absence of a DxxGQ glutamine and/or arginine finger. When crystal structures were determined for RanGAP in ternary complexes with Ran binding protein (RanBP) and ground state Ran-GppNHp or transition state mimetic Ran-GDP-AlF3, it became clear that while the DxxGQ glutamine mediates polar interactions similar to its counterpart in the Ras and Rho GAP complexes, RanGAP does not supply an arginine finger104 (Fig. 4A). Since RanBP binds to a distal site, the catalytic mechanism likely involves allosteric effects, including stabilization of a catalytically competent orientation of the cis-glutamine similar to RGS GAPs for Gα subunits105,106, in addition to proposed solvent shielding of the hydrolytic site by the switch I tyrosine equivalent to Ras Tyr 32. Indeed, the switch I tyrosine in Ran adopts a closed conformation that contrasts with the open conformation required for insertion of the arginine finger in most GTPase/GAP complexes. Interestingly, a conserved arginine from RanGAP packs against the switch I tyrosine and appears to be sufficiently close to potentially insert into the hydrolytic site if the switch I tyrosine was in an open conformation. Alanine substitution of the conserved arginine, or a nearby conserved glutamic acid with which it forms an ion pair, has little affect on GAP activity whereas alanine substitution of another conserved arginine proposed to be a catalytic arginine finger reduces kcat by two orders of magnitude107,108. The proposed arginine finger, however, is located roughly 15 Å from the hydrolytic site in the transition state mimetic complex104. Time resolved FTIR studies reveal a small conformational change in the hydrolytic site preceding rate limiting bond breakage, and further suggest that the switch I tyrosine inhibits intrinsic hydrolysis since alanine substitution increases the intrinsic rate by roughly two orders of magnitude but has little affect on GAP-accelerated hydrolysis98.
Although closely related to Ras, the small GTPase Rap has a non-catalytic threonine rather than a glutamine following the DxxG motif. Rap is deactivated by two distinct GAPs, Rap-specific RapGAPs and dual specificity Ras/Rap GAPs. Unlike the dual specificity Ras/Rap GAPs, RapGAPs lack an arginine finger109. A mutational analysis based on the structure of the RapGAP1 catalytic domain indicated that RapGAPs supply an ‘asparagine thumb’ in lieu of the cis-glutamine110. The crystal structure of RapGAP1 in complex with Rap-GDP and the ground state mimetic BeF3- confirmed that the asparagine thumb is appropriately oriented towards BeF3- to function in catalysis111 (Fig. 4B). In addition, the switch I tyrosine equivalent to Ras Tyr 32 adopts a closed conformation similar to that in the Ran/RanGAP complex. The tyrosine phenyl moiety rather than its hydroxyl group appears to play an important role in the GAP reaction, since mutation to alanine reduces catalytic activity 25 fold compared to less than two fold reduction for mutation to phenylalanine. Time resolved-resolved FTIR experiments combined with mutational analyses of wild type and mutated dual specificity Ras/Rap GAPs (GAP1 and RASAL) provide compelling evidence that the GAP domain reorients switch II such that a glutamine three residues after the DxxG motif (i.e. not the typical cis-glutamine) assumes a catalytic role analogous to the DxxGQ glutamine in Rho, Ras and Gα subunits112. As these examples illustrate, different GAPs can exploit structurally distinct catalytic mechanisms to deactivate common GTPase substrates.
In the Sar1 GTPase involved in vesicle budding from endoplasmic reticulum exit sites, and also in evolutionarily conserved prokaryotic and eukaryotic GTPases involved in translation elongation (EF-Tu, EF1A, EF-G and IF2), the DxxGQ glutamine is replaced by a catalytic histidine. Unlike glutamine, the relatively low pKa for histidine (typically near 6) would appear to make it a reasonable candidate for general base catalysis through activation of the nucleophilic water. However, simulations for EF-Tu in the context of the ribosome suggest that the pKa for the catalytic histidine is several units higher due to nearby counter ions such that the histidine is protonated in the ground state and instead functions in electrostatic compensation of the hydroxide ion generated during transfer of the proton from the nucleophilic water to the γ phosphate113. Although it is not clear if the histidine in Sar1 functions in an analogous substrate-assisted mechanism, the Sar1 GAP (i.e. the Sec23/24 complex) supplies a putative arginine finger that is directed into the hydrolytic site in the crystal structure of the Sar1-GppNHp complex with Sec23/24114.
Finally, it is interesting that several GTPases substitute the DxxGQ glutamine with a non-catalytic hydrophobic amino acid residue and have been termed ‘HAS-GTPases’115,116. These include the cation-dependent MnmE and dynamin family GTPases, which utilize either K+ (MnmE) or Na+/K+ in lieu of a catalytic arginine. In the crystal structure of dimeric MnmE with the transition state mimetic GDP-AlF4-, a conserved glutamate residue from switch II compensates for the absence of a cis-glutamine through an indirect water-mediated polar interaction with the nucleophilic water117. These and related examples have been reviewed in the context of potentially unifying principles, including steric hindrance arising from P-loop substitutions at the position equivalent to Ras Gly12, as a selective pressure for evolution of diverse GAP-GTPase mechanisms118.
Rab proteins comprise the largest GTPase family, with 11 members in yeast, 20-30 in lower metazoan organisms, and 60 members in humans119,120. Multi-genome phylogenetic analyses predict that the last eukaryotic common ancestor (LECA) had approximately 23 different Rab GTPases, indicating that the Rab genes diverged early in eukaryotic evolution to regulate trafficking between different organelles of the endomembrane system121,122. TBC domain proteins also emerged early during eukaryotic evolution with slightly lower numbers than Rab GTPases in diverse genomes, suggestive of a co-evolutionary relationship123. From the initial characterization of TBC proteins in yeast, which are known as GAPs for Ypt (Rab) GTPases or Gyps, it was clear that most yeast TBC domain proteins were GAPs for different though overlapping subsets of Rab GTPases124-127. It is generally thought that most TBC proteins in other organisms are Rab GAPs and that most Rab GTPases have TBC domain GAPs, although the Rab substrate specificity profiles remain to be established for a number of mammalian TBC domain proteins11.
For the prototypical yeast Rab GAP Gyp1, a putative arginine finger was identified by mutational analysis based on the crystal structure of the TBC domain128. Given that Rab GTPases conserve the DxxGQ glutamine (except Rab25, which has a natural leucine substitution), the standard cis-glutamine/trans-arginine configuration was expected. Rab-GDP-aluminum fluoride complexes with Rab GAPs are marginally stable and thus difficult to crystallize; however, the Gyp1 complex with mammalian Rab33 was a fortuitous exception62. High resolution phylogenetic analyses indicate that Rab33 first appeared in multicellular metazoan organisms and is absent from the fungi lineage122. Nevertheless, Gyp1 has 5 fold higher catalytic efficiency for Rab33 than the best characterized yeast substrate Ypt1, and forms sufficiently stable complexes with Rab33-GDP and aluminum fluoride to support crystallization62. Although the crystal structure of Gyp1 in complex with Rab33-GDP-AlF3 confirmed the predicted arginine finger, it also a revealed a major surprise regarding the role of the DxxGQ glutamine (Fig. 4C). Rather than adopting the expected transition state stereochemistry, the DxxGQ glutamine instead mediates bivalent polar interactions with the main chain of the TBC domain, consistent with a role in binding rather than catalysis. Substituting for the DxxGQ glutamine, is a ‘glutamine finger’ from the TBC domain, which mediates canonical bivalent polar interactions with the transition state mimetic. Notably, the glutamine and arginine fingers are both conserved in nearly all TBC domains, with the exception of the small TBC1D3 subfamily in which both are replaced. The dual finger configuration has also been observed in the crystal structure of the TBC1D20-Rab1-GDP-AlF3 transition state complex60 (Fig. 4D), and mutational analyses of these and other TBC domain/Rab systems suggest that the dual finger modality is a fundamental feature of TBC domain GAPs62,129.
Some pathogenic bacteria have evolved GAP proteins that deactivate one or more small GTPases in host cells during infection. Several examples and the underlying evolutionary significance are discussed below.
The structurally similar bacterial proteins VirA from Shigella flexneri and EspG from enteropathogenic E. coli have high GAP activity for Rab1 and crystal structures of VirA and EspG in transition state mimetic complexes with Rab1-GDP-AlF3 revealed a dual arginine/glutamine finger configuration analogous to that of TBC domains59 (Fig. 4E). Considering that the overall tertiary structures of VirA/EspG and TBC domains are unrelated, the common dual finger mechanism is evidently an example of convergent evolution.
Another Rab1 GAP LepB is injected into lung macrophages during infection by the intracellular pathogen Legionella pneumophila130. The catalytic efficiency of LepB for Rab1 (kcat/KM ≈ 106 M-1 s-1) is an order of magnitude higher than that of the host Rab1 GAP TBC1D20, a difference which likely reflects a very high turnover number for LepB (kcat ≈ 102 s-1) approaching the range of traditional enzymes131,132. Crystal structures of LepB in complex with either the ground state mimetic BeF3- or the transition state mimetic AlF3 revealed a cis-glutamine/trans-arginine configuration rather than the dual trans-finger configuration of TBC domains and VirA/EspG66,131,133 (Fig. 4F). Although the low resolution of the transition state mimetic structures limits the certainty of conclusions regarding specific polar interactions, there is nevertheless an elaborate network of polar residues, four from Rab1 and two from LepB, that likely contribute directly to transition state stabilization as suggested by their disposition in the hydrolytic site of the ground/transition state mimetic structures and by the relative magnitude of changes in catalytic efficiency for systematic mutation of polar residues throughout the binding interface131. In addition to the DxxGQ glutamine and arginine finger, the other putative catalytic residues include a P-loop serine corresponding to Ras Gly 12, a switch I serine preceding the invariant threonine, and a switch II arginine that forms an ion pair with a ‘glutamic acid finger’ from LepB. The P-loop and switch I serines are appropriately disposed to mediate polar interactions with the γ phosphate in the transition state. The switch I serine may also interact with the cis-glutamine and nucleophilic water. The ionic interaction between the switch II arginine and LepB glutamic acid finger secures the cis-glutamine in a catalytically competent orientation. Alanine substitution of either the DxxGQ glutamine or switch II arginine reduces catalytic efficiency by four orders of magnitude, while individual alanine substitutions of the P-loop/switch I serines or arginine/glutamic acid fingers reduces catalytic efficiency by two orders of magnitude. Alanine substitutions of other polar residues in the binding interface reduce catalytic efficiency by 10 fold or less. Finally, LepB facilitates insertion of its arginine finger by clamping the switch I tyrosine equivalent to Ras Tyr 32 in an open conformation.
The convergent evolution of dual finger mechanisms in TBC domains and VirA/EspG along with independent evolution of LepB to exploit an atypical polar catalytic network, implies a selective pressure that is likely due to the inability of the cis-glutamine to adopt an optimal stereochemistry for stabilization of the transition state in GTPases with residues other than glycine at the P-loop position equivalent to Ras Gly 12118. Indeed, all but a few Rab GTPases including Rab1 have residues other than glycine at the equivalent P-loop position. One solution to this problem is for the GAP to supply a glutamine finger that inserts from a different orientation to avoid van der Waals conflicts as in VirA/EspG and TBC1D20. An alternative solution is to exploit a polar network of catalytic residues to compensate for the sub-optimal stereochemistry of the cis-glutamine as in LepB.
Given that most small GTPases are membrane localized it is not surprising that a major paradigm for regulation of GAPs involves membrane recruitment, which can be mediated directly by lipid interactions and/or indirectly by interactions with membrane-bound proteins. Another important regulatory paradigm involves post-translation modification of either the GAP or substrate GTPases, which can alter membrane recruitment, interactions with the GTPase, and potentially the hydrolytic activity. Allosteric and autoregulatory mechanisms have also been identified for some GAPs, although the extent of autoregulation and the underlying structural mechanisms remain poorly characterized in most cases. Cross-talk between small GTPases and non-cognate GAPs has also been reported to influence the GAP activity. Finally, expression levels of GAPs can be regulated, either during development or during progression of physiological or pathological processes. Several selected examples illustrating different mechanisms and functional consequences are highlighted below. GAP regulatory mechanisms and their implications for feedback amplification loops have been recently reviewed15.
The RacGAP β2-chimaerin is activated by diacylglycerol (DAG) binding to its C1 domain. The crystal structure of the inactive form of β2-chimaerin revealed an autoinhibitory mechanism whereby the N-terminus of the protein intrudes into the Rac binding site in the GAP domain134 (Fig. 5A). The autoinhibited conformation is stabilized by multiple intramolecular interactions involving the N-terminus, SH2 domain, RacGAP domain and the SH2-C1 linker, which occlude the DAG/phospholipid membrane binding site in the C1 domain. DAG/phospholipid membrane binding to the C1 domain likely relieves autoinhibition by destabilizing these interactions. Another Rho family GAP, p50RhoGAP, is proposed to be autoregulated, although the structural basis remains unknown135. Biochemical and yeast two hybrid analyses of truncation constructs suggest that the GAP activity may be autoinhibited by a mechanism involving the N-terminal region of the protein as well as the Sec14 domain, which has also been implicated in endosome targeting136, and relieved by binding of the N-terminal region to the prenylated C-terminus of Rac. The activities of p190B RhoGAP and RA-RhoGAP are modulated by interaction with Rac and Rap1, respectively137,138; however, it is not known if activation involves relief of autoinhibitory constraints or another mechanism.
The Arf GAP ASAP1 has several modular domains including a PH domain. The GAP activity of ASAP1 exhibits cooperative dependence on phosphatidylinositol 4,5-bisphosphate (PIP2) density in liposome membranes139. Although the cooperative effect requires PIP2 binding to the PH domain, the underlying structural mechanism was enigmatic140. The recent crystal structure of the PH domain in complex with soluble dibutyl PIP2 offered an unexpected insight141 (Fig. 5B). In contrast to the phosphoinositide binding modalities observed in other PH domains, where the phosphoinositide typically binds to either a ‘canonical site’ or a proximal ‘non-canonical’ site, both sites are simultaneously occupied in the ASAP1 PH domain. The cooperative mechanism, which is also thought to involve dynamic interactions between the PH and GAP domains detected by NMR140, is proposed to confer switch-like sensitivity of ASAP1 GAP activity to PIP2 density. The GAP activity of ArfGAP1, which is required for COPI vesicle budding from Golgi membranes, increases by two orders of magnitude as the membrane curvature of liposomes is increased to approximate that of a budding vesicle142. Membrane curvature is sensed by the folding of an ArfGAP1 Lipid Packing Sensor (ALPS) motif proximal to the GAP domain143. ALPS motifs form amphipathic helices stabilized by partitioning with high curvature membranes. Unlike typical amphipathic helices, however, the hydrophilic surfaces of ALPS motif helices are rich in serine/threonine rather than basic residues. How curvature sensitive folding/partitioning of the ALPS motif increases GAP activity remains to be established.
The membrane recruitment mechanism for the Rab GAP Gyp1, which as noted above deactivates the yeast Rab1 ortholog Ypt1, illustrates a functional ‘Rab cascade’ in which the GAP for Ypt1, a Rab required for fusion of ER-derived vesicles with the cis-Golgi, is recruited to the Golgi through an effector binding interaction with active Ypt32, a downstream Rab involved in intra-Golgi transport125,126,144. This Rab cascade enhances segregation between Ypt1 and Ypt32 through a Gyp1-dependent temporal mechanism in which Ypt1 compartments are converted to Ypt32 compartments.
The Akt kinase substrate AS160 (TBC1D4) has GAP activity for several Rab substrates including Rab10145. Under basal conditions in adipocytes, AS160 is recruited to vesicles containing the GLUT4 glucose transporter, where it deactivates Rab10 and possibly other Rab GTPases146,147. In response to insulin stimulation, Akt phosphorylation of AS160 results in its release into cytoplasm and concomitant accumulation of active Rab10, which is important for GLUT4 translocation to the plasma membrane.
During infection of lung macrophages, the intracellular bacterial pathogen Legionella pneumophila injects at least 270 distinct ‘effector proteins’ (i.e. bacterial effectors as opposed to GTPase effectors), including a GEF (DrrA/SidM), GAP (LepB) and binding protein (LidA) for Rab1130,148,149. In addition to a P4M domain that selectively binds PI4P to facilitate targeting to the Legionella-containing vacuole150 and a GEF domain that activates Rab1151-153, DrrA/SidM contains an AMP transferase (ATase) domain that covalently attaches AMP to a conserved tyrosine in the switch II region of Rab191. AMPylated Rab1 is inert to deactivation by LepB or the host Rab1 GAP TBC1D20 and must be de-AMPylated by another Legionella effector SidD prior to de-activation by GAPs132,154. Notably the AMPylated tyrosine is located at the binding interface in the crystal structures of the LepB (Fig. 5C) and TBC1D20 complexes with Rab1. In contrast, AMPylation does not impair binding to LidA, which has picomolar affinity for Rab1 independent of nucleotide state132,155,156. Thus, AMPylation repurposes Rab1 for selective interactions with LidA, possibly to facilitate tethering of ER-derived vesicles for fusion with the LCV157.
The crystal structure of Arl3-GDP-AlF3 in complex with the ASAP3 GAP domain revealed a Ca2+ ion at the GTPase-GAP binding interface that is coordinated by residues from both proteins103 (Fig. 5D). Ca2+ was found to simulate the GAP activity of ASAP1 and ASAP3 (and likely ASAP2) but not ArfGAPs from other subfamilies. Ca2+ activation may couple the GAP activity of ASAP isoforms to release of Ca2+ from internal stores and thereby generate crosstalk between Ca2+ and Arf signaling pathways.
Given that small GTPase regulatory networks control many fundamental biological processes, it is not surprising that they are increasingly implicated in disease initiation, progression, severity and/or therapeutic resistance. The examples below illustrate some of the different roles for GAPs in pathological processes as well as some potential targets and strategies for therapeutic intervention. More detailed discussion of disease relationships for small GTPases and their regulatory factors can be found in other in depth reviews13,15,20,50,158-161.
Although constitutively activating Ras mutations (including the aforementioned Gly 12 and Gln 61 substitutions) occur in roughly 30% of human tumors, they are curiously absent in some cases where down stream signaling pathways are nevertheless hyperactivated160. In this context, the 14 human RasGAPs are increasingly appreciated as bona fide or potential tumor suppressors. The RasGAP neurofibromin encoded by the NF1 gene is the earliest and best characterized example162. Deletion or loss of function mutations in NF1 are responsible for peripheral nervous system tumors (neurofibromas) that are the hallmark of neurofibromatosis type 1 and studies in mice provide evidence for a more general role of NF1 in familial as well as sporadic cancers160.
The tuberous sclerosis complex (TSC), consisting of the hamartin and tuberin proteins encoded by the TSC1 and TSC2 tumor suppressor genes respectively, functions as a GAP for the small GTPase RheB in the mechanistic target of Rapamycin (mTOR) signaling pathway163. Interestingly, missense mutations that map to the Rap1GAP-like domain of tuberin and disrupt GAP activity for RheB have been identified in patients with tuberous sclerosis. The unconventional TBC domain protein TBC1D7 was shown to be a third component of the TSC164. Although TBC1D7 knockdown results in increased mTOR Complex 1 (mTORC1) signaling, delayed initiation of autophagy and resistance to glucose or growth factor withdrawal, it does not appear to be associated with tuber sclerosis mutations. Homology modeling based on the Rap1GAP-Rap1 structure described above suggests that at least one patient mutation (R388P) in the TSC2 GAP domain disrupts binding to RheB (or possibly Rap1) rather than hydrolytic activity per se110.
In the Arl3 GAP RP2 described above, loss of function mutations in the GAP domain including the arginine finger interfere with trafficking between the Golgi and sensory cilium membrane, leading to retinal degeneration, and account for ~15% of X-linked retinitis pigmentosa cases165,166. Several missense mutations involve residues in the GAP domain including the arginine finger65.
Mutations in TBC1D24 are associated with DOORS syndrome167,168, a rare familial disorder characterized by multiple clinical presentations including infantile epileptic seizures and developmental delay, as well as rare autosomal-dominant hearing loss169-171. TBC1D24 interacts with Arf6, and evidently functions as a negative regulator172. Notably, TBC1D24 lacks the characteristic arginine and glutamine fingers, suggesting that it is does not function as a Rab GAP or otherwise deviates from the dual finger paradigm. Whether the TBC domain of TBC1D24 has GAP activity for Arf6 or Rab substrates remains to be determined.
Rab3GAP is a heterodimer consisting of catalytic and regulatory subunits, and deactivates Rab3 isoforms in neurons to regulate synaptic vesicle fusion173,174. Mutations in both subunits have been linked to the autosomal recessive neurological disorders Warburg Micro Syndrome (catalytic subunit mutations) and Martsolf Syndrome (regulatory subunit mutations)175,176. In the case of Warburg Micro Syndrome, the majority of mutations occur in the C-terminal region of the catalytic subunit required for catalytic activity and are proposed to disrupt GAP activity. Since the catalytic subunit lacks homology with other GAP domains and the tertiary structure remains to be determined, the underlying structural bases and etiology are poorly understood.
Several pathogenic prokaryotes supply GAPs that manipulate small GTPases during infection15. These include VirA/EspG and LepB described above as well as the Salmonella typhimurium RhoGAP SptP, which functions in conjunction with the Salmonella typhimurium RhoGEF SopE to control Rac-1 and Cdc42 activation and actin dynamics177,178. SptP supplies an arginine finger and utilizes the DxxGQ glutamine but otherwise does not resemble the tertiary structure of host Rho family GAPs, indicative of a convergent evolutionary process that preserves underlying stereochemical functionality179. Several other pathogenic bacteria have proteins with structurally related RhoGAP domains including Yersinia pestis YopE180, Pseudomonas aeruginosa ExoS181 and ExoT182, and Aeromonas hydrophila AexT183. As with bacterial RabGAPs, the catalytic activities of the bacterial RhoGAP domains are considerably higher than those of host RhoGAPs15.
At least some GTPase regulatory factors, in particular those required for maintenance of pathologic conditions that can also be validated as in vivo targets in animal models for disease, are potential candidates for development of new mechanisms-based therapeutic interventions. Inspired in part by natural inhibitors such as the fungal metabolite Brefeldin A, which inhibits Arf GTPase activation by forming an interfacial ternary complex with Arf1-GDP and Golgi Arf GEFs184, efforts to identify chemical inhibitors of small GTPase GEFs have led the way185,186. However, there is no fundamental reason to exclude GAPs or mutated GTPases as potential targets. Indeed, it has been proposed that chemical compounds might be found that would counter the effects of constitutively activating Ras mutations, possibly by mimicking acceleration of GTP hydrolysis by GAPs, trapping ternary complexes with Ras GAPs, or modulating protein expression or degradation13. A fundamentally different approach that has recently generated considerable excitement as well as concern involves correction of mutations or knockout/inactivation of mutated genes through gene-editing187. High efficacy gene-editing approaches would require improved specificity to minimize off-target effects, possibly in conjunction with better vehicles for delivery to affected cells in order to increase potency, penetrance and selectivity. Short of therapeutic applications, gene editing tools are already playing an important role in target validation and assessment of potential therapeutic strategies in cell and animal models188.
Since the discovery of small GTPases several decades ago, considerable progress has been made with respect to delineating the structural bases and mechanisms underlying the intrinsic vs. GAP-catalyzed GTP hydrolytic reactions for all of the major GTPase families. Most GAPs provide a catalytic arginine finger, although some GAP/GTPase systems utilize a K+ or Na+ ion to achieve analogous electrostatic compensation whereas others lack a catalytic arginine and do not depend on monovalent ions. Simple generalizations regarding the role of the largely conserved DxxGQ glutamine are complicated by structural variations and exceptions. In the known cases where the DxxGQ glutamine is either non-catalytic or replaced by a non-catalytic residue, a distinct though functionally equivalent glutamine/asparagine may be supplied in trans or even in cis. In retrospect, it appears that some of the structural variations and trans vs. cis alternatives may be related to oncogenic mutations in Ras. For example, the trans glutamine alternative avoids van der Waals conflicts imposed by natural substitutions at the P-loop position corresponding to Ras Gly12 that prevent the DxxGQ glutamine from adopting optimal stereochemistry in the transition state. Thus, the chemical functionality is better preserved than individual residues or their functional roles.
The nature of the transition state/intermediates and specific chemical functions of catalytic residues in the hydrolytic site has been the subject of considerable debate, with multiple alternatives proposed on the basis of experimental observations and computational analyses. It is generally accepted that the arginine finger contributes to electrostatic reduction of developing charge in the transition state. The cis/trans-glutamine or asparagine equivalent is thought to orient the nucleophilic water and stabilize the transition state geometry, with some differences in stereochemistry forced by P-loop substitutions as noted above. For Ras/RasGAP, which is by far the best characterized GTPase/GAP system, there appears to be an emerging though still developing consensus supporting a transition state/intermediate reaction pathway with substantial associative character but without a general base for activation of the nucleophilic water. Proton transfer from the nucleophilic water to the γ phosphate may not be direct, but instead mediated either by a second water in the intrinsic reaction or the DxxGQ glutamine in the GAP reaction. Regardless of the proton transfer pathway, the catalytic effect of the GAP appears to derive primarily from electrostatic compensation by the arginine finger in combination with allosteric modulation. The allosteric contribution includes a staggered-to-eclipsed configurational change involving the non-bridging oxygens of the α-β phosphates, which adds to the steric strain in the phosphate-bonding network incurred by an analogous configurational change for the non-bridging oxygens of the β-γ phosphates accompanying GTP binding to Ras.
Beyond well characterized instances, the regulatory mechanisms for most GAPs are either unknown or partially characterized. The dearth of detailed information is due in part to the challenges associated with expressing and purifying full length GAPs proteins and/or complexes in quantities required for structural and biophysical studies. Nevertheless, it is already clear that the mechanisms for regulation of GAP localization/activity are even more diverse than the catalytic mechanisms. Likewise, many diseases including a substantial fraction of cancers are genetically-linked to disruption or impairment of hydrolytic activity as a consequence of mutations in small GTPases as well as their GAPs. Pathogens also supply GAPs to manipulate the hydrolytic activity of host small GTPases.
Despite considerable progress, there are still major questions that require clarification and additional paradigm-shifting observations can be expected. With respect to intrinsic and GAP catalyzed reaction mechanisms, it is not entirely clear whether or how the results from extensive studies of Ras/RasGAP can be extended to other GTPases/GAP systems, especially those with significant structural divergence in the hydrolytic site. Further biophysical and computational investigation of the intrinsic and GAP reactions for other GTPase/GAP families would be useful. Meaningful comparisons of computational analyses with experimental observations will require more detailed kinetic and time-resolved spectroscopic studies in order to distinguish chemical steps from other potentially rate limiting pre/post-transition state barriers including phosphate release. There is also a clear need for exploration of GAP regulatory mechanisms in the context of the full length proteins, complexes and membranes as well as the application of quantitative experimental approaches to extend in vitro/in silico observations to cellular environments. Finally, there is a growing awareness of the important role of small GTPases and GAPs with respect to disease onset, progression and severity. The surface has hardly been scratched with respect to development of new mechanism-based therapeutics targeting components of GTPase regulatory networks. A strong case can be made for further exploration of GTPases and GAPs in disease, including the underlying molecular/structural mechanisms, target validation, and testing of therapeutic strategies in disease models.
We thank Sanchaita Das for comments on the manuscript. This work was supported by NIH Grant GM056324 to D.G.L.