The pseudoGTPase group of pseudoenzymes

Abstract

Pseudoenzymes are emerging as significant mediators and regulators of signal transduction. These proteins maintain enzyme folds and topologies, but are disrupted in the conserved motifs required for enzymatic activity. Among the pseudoenzymes, the pseudoGTPase group of atypical GTPases has recently expanded and includes the Rnd and RGK groups, RhoH and the RhoBTB proteins, Miro and centaurin-γ groups, CENP-M, dynein LIC, EhRabX3, LRRK2 and the p190RhoGAP proteins. The wide range of cellular functions associated with pseudoGTPases include cell migration and adhesion, membrane trafficking and cargo transport, mitosis, mitochondrial activity, transcriptional control and autophagy, placing the group in an expanding portfolio of signaling pathways. In this review we examine how the pseudoGTPases differ from canonical GTPases and consider their mechanistic and functional roles in signal transduction. We review the amino acid differences between the pseudoGTPases and discuss how these proteins can be classified based on their ability to bind nucleotide and their enzymatic activity. We discuss the molecular and structural consequences of amino acid divergence from canonical GTPases and use comparison with the well-studied pseudokinases to illustrate the classifications. PseudoGTPases are fast becoming recognized as important mechanistic components in a range of cellular roles and we provide a concise discussion of the currently identified members of this group.

Pseudoenzymes

Enzyme families are usually built around a conserved protein fold with enzymatically required amino acids located at spatially defined locations within the fold. Members of each enzyme family generally have a similar structural topology, and this allowed much of the early identification work for enzyme family members, for example, the presence of the G motifs helped identify and classify many small GTPases [1]. As enzyme identification progressed, members of these families were observed that maintained the enzyme fold, but had degenerate conserved motifs [2, 3]. These ‘pseudoenzymes’ occur across the kingdoms of life, and their extensive presence is becoming widely appreciated [4, 5]. The classic example is the protein kinase family, where approximately 10% of its ~500 members are altered in the conserved kinase enzymatic motifs [6, 7]. The functional roles of these proteins have become the focus of much recent study, and have been found to be important in a wide range of signaling pathways [4, 5].

The Ras-like small GTPases

Small GTPases are important signaling molecules that function widely as regulators of signal transduction and cellular function. The 150+ members of this family have been extensively studied and classified into five major groups: Ras, Rho, Rab, Ran and Arf [8]. These proteins are built around a Rossmann fold architecture to bind and hydrolyze guanine nucleotides, and usually comprise 160–200 amino acids [9]. The enzymatic activity of small GTPases is key to their cellular functions; they slowly hydrolyze guanosine-5’-triphosphate (GTP) to release the γ-phosphate and generate guanosine-5’-diphosphate (GDP). By doing so, the molecular conformation of the protein changes – the GTP-loaded and GDP-loaded states are significantly different. Because downstream effector proteins bind the GTP-loaded state, the differences between the GDP and GTP conformational states control the extensive repertoire of GTPase-mediated signal transduction. This process is termed “GTPase cycling” (Figure 1A).

Figure 1. GTPase cycling and conserved regions.

A) Schematic illustrating GTPase cycling. GTPase domain indicated as grey circle. GDP and GTP nucleotides indicated. Inorganic phosphate indicated as PO₄. GTPase activating protein indicated as GAP, guanine nucleotide exchange factor indicated as GEF. Insets are of crystal structures illustrating the conformational differences between the GDP and GTP loaded states of Ras (PDB codes: 5P21, 4Q21) [111, 112]. Major differences in Switch I and Switch II are indicated with red arrows. G motif colors are: G1 (P-loop) in purple, G2 (Switch I) in yellow, G3 (Switch II) in light green, G4 in dark green, and G5 in blue. Nucleotide shown in stick format. B) Structure of Ras in complex with GTP analogue GMP-PNP (PDB code: 5P21) with the G motifs shown in stick format and colored as per panel A. Structural images generated using CCP4mg [113].

As the small GTPases are usually slow enzymes but are responsive to upstream signals, their activity in signal transduction pathways must be regulated. To do this, conformational changes are accelerated by the actions of guanine nucleotide exchange factor (GEF) proteins which assist GDP release, and GTPase activating proteins (GAP) which increase the rate of hydrolysis [10]. Furthermore, control of the system can be achieved by a multitude of processes, including the actions of guanine dissociation inhibitors (GDI) on Rho- and Rab-family small GTPases [11, 12], post-translational modification [13], trafficking and sub-cellular localization [13], and oligomerization changes such as the recently discovered dimerization of Ras [14].

Catalysis, G motifs, and Switches

The small GTPase group has a number of common structural features that contribute to GTPase cycling, enzymatic activity, and conformational regulation of binding partner interactions. These features are the five conserved “G box” GTP/GDP-binding motif elements that mediate nucleotide binding and catalysis. Termed G1 through G5, the G motifs are short amino acid sequences that are highly conserved through the family and comprise the key amino acid residues that create the nucleotide binding site [15, 16]. The G1 motif (GxxxxGKS/T) is also known as the P-loop or Walker A motif and forms a cavity to bind the nucleotide phosphates. G2 is also known as ‘Switch I’ and contains a threonine residue that helps coordinate the GTP γ-phosphate and Mg²⁺ ion. G3 (DxxGQ/H/T) is also known as ‘Switch II’ or Walker B motif and allows a water-mediated coordination of the nucleotide-bound Mg²⁺ ion. G4 (T/NKxD) and G5 (C/SAK/L/T) create specificity for the guanine ring [8, 17, 18] (Figure 1B).

The “Switch regions”, Switch I (G2) and Switch II (G3), are so termed because they display extensive conformational changes in response to nucleotide binding state, i.e. they switch conformations. The conformation of the Switch regions is directed by the bound nucleotide (GDP or GTP), and protein binding partners recognize these specific conformations; GAP and effector proteins recognize the GTP-bound state, GEF proteins recognize the GDP-bound and nucleotide-free states, and GDI proteins the GDP-bound state (Figure 1A). Broadly speaking, the small GTPases therefore have defined regions that mediate nucleotide binding, catalysis, and protein interactions [9, 15, 16].

PseudoGTPases: a rapidly growing group

The small GTPase group includes members which are altered in their conserved catalytic and nucleotide-binding G motifs [5, 19], and the term “pseudoGTPase” seems to have been coined in 2014 to describe these atypical or non-canonical members [20, 21]. Identification of pseudoGTPases within the human genome is not fully complete, but the 150+ small GTPases suggests 15–20 pseudoGTPases in total (10–15%) [2, 4]. These currently include six Rho-family members, the Miro mitochondrial Rho-like GTPases, members of the Ras family RGK group, the Ras-like Roc domain of LRRK2, domains within the multidomain centaurin-γ and p190RhoGAP GAP proteins, the kinetochore protein CENP-M, and the dynein light intermediate chain. PseudoGTPases also exist in non-mammalian genomes, including the fungal dynein light intermediate chain and an extensive array of predicted and demonstrated pseudoGTPases in the parasitic protozoan, E. hystolytica. They are also structurally related to the bacterial response regulator proteins. This array of identified pseudoGTPases fits well into the classification scheme for nucleotide-binding pseudoenzymes (Figure 2), and it is notable that they have varying levels of conservation for their G motifs (Figure 3). The roles of these pseudoenzymes in signal transduction at the molecular and functional level are being elucidated and are extensive.

Figure 2. Illustration of pseudoGTPase classification.

Classifications of for pseudoGTPases are shown and compared to pseudokinases. Class i nucleotide non-binding catalytically incompetent pseudoenzymes are colored purple, class ii nucleotide binding catalytically incompetent pseudoenzymes in green, and class iii nucleotide binding catalytically competent pseudoenzymes in yellow [23, 24]. *Note that RhoBTB3, AGAPs, Miro1-C and Miro2-C bind and hydrolyze non-GTP nucleotides.

Figure 3. Alignment of G motifs for pseudoGTPase family members.

Red indicates residues that do not match the consensus. Sequences are human unless indicated. Alignments are structure-based where available and performed by the DALI server [114], PROMALS3D [115], and CLUSTAL Omega [116]. RGK family members are Gem, Rad, Rem1 and Rem2. hLIC is human dynein LIC1, fLIC is fungal LIC (C. thermophilum). AGAP1 is centaurin-γ2, AGAP2 is centaurin-γ1, and AGAP3 is centaurin-γ3. p190A-N, p190A-pG1 and p190A-pG2 are p190RhoGAP-A N-terminal, pG1 and pG2 pseudoGTPase domains. p190B-N, p190B-pG1 and p190B-pG2 are p190RhoGAP-B N-terminal, pG1 and pG2 pseudoGTPase domains. LRRK2 Roc is the Ras of complex domain of leucine rich repeat kinase 2. EhRabX3-N and EhRabX3-C indicate the N- and C-terminal pseudoGTPase domains of EhRabX3, respectively. All sequences are from UniProt; accession codes are as follows: p190RhoGAP-A: Q9NRY4; p190RhoGAP-B: Q13017; fLIC: G0S0R6; CENP-M: Q9NSP4; RhoBTB1: O94844; RhoBTB2: Q9BYZ6; EhRabX3: Q5NT25; Rnd1: Q92730; Rnd2: P52198; Rnd3: P61587; hLIC1: Q9Y6G9; RhoH: Q15669; Gem: P55040; Rad: P55042; Rem1: O75628; Rem2: Q8IYK8; AGAP1: Q9UPQ3; AGAP2: Q99490; AGAP3: Q96P47; Miro1: Q8IXI2; Miro2: Q8IXI1; RhoBTB3: O94955; LRRK2: Q5S007.

Classification of pseudoGTPases

The most well-studied group among the pseudoenzymes is the pseudokinases. These are degenerate in residues that mediate ATP nucleotide binding and catalysis, and accordingly the consequences of these differences have been extensively studied [7, 22]. Analysis of the pseudokinases allows their further classification into three classes: those with (class i) loss of nucleotide binding and consequent loss of catalytic competence, (class ii) retained nucleotide binding but loss of catalytic competence, and (class iii) retained catalytic activity [23, 24] (Figure 2). Importantly, this classification system may be thought of as a general scheme for other pseudoenzymes belonging to a nucleotide-binding protein fold, including the GTPases.

For class i pseudoGTPases, amino acid changes in the conserved G motifs substantially alter the nucleotide binding pocket. Loss of nucleotide binding can be accomplished by mutation in the G1 motif which accommodates the phosphates, and/or mutations in the G4 and G5 motifs which form the pocket for the guanosine. Consequently, the nucleotide non-binding/catalytically incompetent group of pseudoGTPases are often very degraded in the G1, G4 and G5 motifs compared to canonical GTPases. As these pseudoGTPases do not bind nucleotide, the Switch regions (G2 and G3 motifs) are also divergent among the sub-group (Figure 3). For class ii pseudoGTPases, loss of enzymatic activity results from alterations in catalytic residues, often attributed in part to the catalytically important glutamine of the G3 motif (DxxGQ). Despite these changes the class ii domains maintain the ability to bind nucleotide, and consequently the phosphate binding G1 motif is conserved with canonical GTPases. The class ii pseudoGTPases usually adopt a GTP-bound conformation with the differences in catalytically important residues degrading the ability to hydrolyze nucleotide (Figure 3). For class iii pseudoGTPases, both nucleotide binding and catalysis are maintained despite changes in the consensus motifs. Commonly these pseudoGTPases contain mutations in the G3, G4 and G5 motifs, and often the conserved threonine residue of the G2 (Switch I motif) is lost. For some of this class, specificity for guanosine nucleotides is lost because of the G motif substitutions, particularly in G4 and G5, for others, catalysis requires elements outside of the standard GTPase fold contributed by other regions of the protein, or by partner proteins (Figure 3). As is clear from these descriptions, sequence analysis can yield some degree of functional prediction for the pseudoGTPases, but sequence comparison is often not conclusive so experimental studies are usually required to define the nucleotide binding and catalytic competence of pseudoGTPases. As described below, many of the family members have yielded novel and interesting functionality upon biochemical, structural and functional analyses.

Members of the PseudoGTPase group

The pseudoGTPases of the Rho-family (the Rnd group, RhoH and RhoBTBs)

There are 20 Rho-family small GTPases, several of which are atypical GTPases and can be classified as pseudoGTPases. The Rnd proteins, Rnd1 (Rho6), Rnd2 (Rho7) and Rnd3 (Rho8/RhoE), are three of these and are important regulators of cytoskeletal organization, migration and adhesion [25–27]. Upon their discovery, non-canonical residues at key sites in the G motifs were noted [25, 28], and today these differences classify the Rnd proteins as pseudoGTPases. Conserved phosphate-binding residues including the G1 motif, as well as some conservation in nucleotide-binding residues in G4 and G5, suggested that GTP binding is maintained. This was confirmed biochemically, but divergent catalytic residues including the Gln to Ser change in G3/Switch II contribute to catalytic inactivity [25, 28, 29]. Therefore the Rnd proteins can be defined as class ii pseudoGTPases, belonging to the group that constitutively bind GTP but are catalytically incompetent (Figures 2 and 3). So how do the Rnd proteins function in signaling? Many studies have found that their roles as regulators of neuronal signaling pathways are regulated by mechanisms such as expression, localization and phosphorylation, rather than conformational switching as a readout of GTP/GDP-bound state [27, 30, 31]. These mechanisms commonly alter the interactions of Rnd proteins with binding partners and are important for regulation of signal transduction pathways. For example, Rnd binding to p190RhoGAP proteins targets these RhoGAPs (which are also pseudoGTPase proteins) to locations where activation of Rho’s enzymatic activity is required [30, 32], likewise Rnd proteins regulate RasGAP activity of Plexin receptors by direct interactions, although the molecular mechanisms of these are still to be revealed [30, 33]. Overall, the Rnd small GTPases often act as either binders or regulators for GAP proteins and/or as scaffolds, and probably represent the best studied group of pseudoGTPases.

RhoH is a hematopoiesis-specific pseudoGTPase that retains GTP binding but not catalytic activity, owing possibly to its altered G3/Switch II motif sequence (Figure 3), and is resistant to GAP-induced activation by p50RhoGAP [34]. Although it is not involved in actin cytoskeleton rearrangement like its active counterparts such as RhoA, it has inhibitory effects on transcriptional pathways [34, 35]. Specifically, RhoH has been observed to act as a negative regulator of p38 and NFkB activation via inhibiting the active Rho GTPases Rac1, RhoA and CDC42, but the mechanisms are currently not well understood.

RhoBTB1, RhoBTB2 and the distantly related RhoBTB3 are multidomain proteins that contain GTPase-like domains, in addition to BTB domains [36]. The GTPase domains of RhoBTB1 and RhoBTB2 have been described as “atypical”, due to divergent G3, G4 and G5 motifs [37]. The G4 and G5 divergence potentially indicates a deficiency in nucleotide binding, and indeed RhoBTB1 and RhoBTB2 have been shown not to bind nucleotide [38, 39]. RhoBTB3 has approximately 50% sequence identity with RhoBTB1 and RhoBTB2 [40] and is so divergent that it was not initially classified as Rho-like [41]. Its GTPase-like domain is predicted to assume the GTPase fold with atypical sequences across its G motifs plus a predicted insert in G1 [40] (Figure 3). The pseudoGTPase domain of RhoBTB3 does not bind GTP making it divergent from the other RhoBTB proteins [42] however, it can bind and hydrolyze ATP, a selectivity probably achieved because of its atypical G motifs [40]. Although RhoBTB3 has this unusual ATPase activity, the best described roles of the RhoBTB pseudoGTPase domains are as protein-protein interaction domains [36]. The RhoBTB pseudoGTPase domain is thought to act as the substrate binding domain for a Cullin-3, Roc1, RhoBTB ubiquitin ligase complex [36, 42, 43], and the RhoBTB pseudoGTPase domain may also directly interact with Rab9 as an effector important for in vesicle transport to the Golgi [40]. The pseudoGTPase domain therefore is thought to be critical to RhoBTB function.

The mitochondrial Rho-like small GTPases (Miro1 and Miro2)

The mitochondrial Rho-like small GTPases, or Miro proteins, Miro1 (RhoT1) and Miro2 (RhoT2), are transmembrane multidomain proteins that contain two GTPase-like folds which flank two EF-hand motifs in the cytoplasmic side of the mitochondrial membrane [44, 45]. These proteins help regulate many aspects of mitochondrial activity including transport, morphology, and apoptosis among others [45, 46]. The N-GTPase domain of both Miro1 and Miro2, which are 73% identical, possess conserved G motif sequences at the G1, G4 and G5 positions, and non-conserved G2 and G3 motifs. Despite the lack of conservation of the G2 and G3 motifs, the isolated N-GTPase domain of both proteins exhibit GTP hydrolysis activity in vitro [47]. The Miro1 and Miro2 C-GTPase domains (approximately 50% identical) are often referred to as “atypical” or “relic” GTPase domains, due to the highly divergent nature of their G2 and G3 regions [45, 48]. Interestingly, however, hydrolysis activity has been demonstrated for the isolated C-GTPase domains of both Miro1 and Miro2, and furthermore these domains are able to hydrolyze ATP and UTP [47]. The C-GTPase domains of Miro1 and Miro2 are thus broad-specificity NTPases. The potential roles of enzymatic activity in these catalytically-active pseudoGTPase domains is still in question, with the GTPase activity of N-GTPase thought to be required for proper Miro function in mitochondria, but the C-GTPase activity to be dispensable [46, 49, 50]. These GTPase-like domains have been suggested to affect mitochondrial processes via allosteric regulation of target proteins, but the specific targets and mechanisms have not yet been fully determined [51].

The Ras family RGK group (Rad, Rem1, Rem2, Gem/Kir)

The RGK small GTPases compose a sub-family within the Ras-like small GTPases. The four members of this group - Rad, Rem1, Rem2, Gem/Kir - contain a central GTPase domain with an N-terminal extension of variable length and a conserved C-terminal extension [52–55]. The best known function of the RGKs is to regulate voltage-gated calcium channels and is achieved by two major mechanisms, interaction with the Ca_vβ-subunit to directly influence channel open probability or regulation of surface expression [52–54]. The RGKs can bind both GDP and GTP despite amino acid substitutions in their G motifs (including in the Switch I and Switch II regions) (Figure 3), and although the nucleotide-binding affinities are lower than typical GTPases, some studies have shown the RGKs possess slow GTP hydrolysis activity [56–58]. Among the G motif substitutions, the Switch II (G3 motif) contains a bulky tryptophan residue which has been suggested to disturb bound magnesium and thus contribute to the lower nucleotide binding affinity (Figure 3) [59]. Furthermore, structural studies indicate that there are very few conformational differences between the GDP- and GTP-bound states, making it unlikely that these GTPases act like Ras as molecular switches [57]. Lastly, the unique N- and C-terminal extensions may affect hydrolysis; for example, deletion of these extensions has been observed to reduce the GTPase activity of Gem [56]. The RGK GTPases are therefore indicated to be members of the class iii pseudoGTPases.

The Roco protein kinase family (LRRK1, LRRK2, DAPK1, and MFHAS1)

Roco proteins are multidomain proteins that contain a tandem of Roc (Ras of complex) and COR (C-terminal of Roc) domains, in addition to other domains including a serine/threonine kinase [60]. There are four human Roco proteins, leucine rich repeat kinase (LRRK1), LRRK2, DAPK1 and MFHAS1, the best studied of which is LRRK2 because of its clinical relevance and frequent mutation in Parkinson’s disease [61–63]. It was originally noted that the Roc domain of LRRK2 shares homology with Ras-like GTPases and contains conserved G1, G2 and G3 motifs, but altered G4 and G5 [60] including a conserved histidine at the position of the typical G4 lysine (Figure 3). These differences classify the Roc domain as a pseudoGTPase domain, and nucleotide binding (albeit with low affinity) and hydrolysis activity have both been demonstrated [64] placing LRRK2 Roc domain in the class iii pseudoGTPases (Figure 2). Intriguingly, the crystal structure of Roc domains revealed a unique domain-swapped GTPase dimer that forms the active site, with the G1, G2 and G3 motifs from one monomer and G4 and G5 from the other [65]. However, whether this dimerization occurs in cells is still the subject of debate (discussed in [66]); nonetheless, dimerization has been proposed to be related to the nucleotide-bound status of the protein [67]. The pseudoGTPase domain seems to be intrinsic to regulation of LRRK2 function, as the kinase and GTPase activities are mutually regulated; kinase activity is dependent on the Roc domain, and loss of GTP binding and/or hydrolysis leads to a decrease in LRRK2 autophosphorylation [68–70]. Conversely, LRRK2 autophosphorylation of residues in the Roc domain can enhance GTPase activity [71]. The Roc/pseudoGTPase domain also interacts with tubulin and microtubules in a GTP-dependent manner, [72, 73], and some pathogenic mutants show enhanced colocalization of LRRK2 to microtubules [74, 75]. Additionally, many Parkinson’s disease mutations in LRRK2 result in increased kinase activity and/or decreased GTPase activity (reviewed in [76]). Thus, the pseudoGTPase Roc domain of LRRK2 functions as an important regulatory domain, with critical pathogenic roles.

Centaurin-γ ArfGAP proteins (AGAP1, AGAP2, AGAP3)

The centaurins are multidomain GAPs for Arf GTPases. The -γ subgroup contains three members, centaurin-γ2, -γ1, and -γ3 (also named AGAP1, 2 and 3, respectively) and in addition to the GAP, pleckstrin homology and ankyrin repeat domains, these proteins contain a pseudoGTPase domain (termed GTPase-like domain, GLD, in centaurin literature) [77, 78]. These proteins function as regulators of Arf signaling, with roles including in endosomal trafficking, the actin cytoskeleton, apoptosis and cancer cell invasion [77–79]. The centaurin-γ GTPase domains contain multiple sequence divergences in their G2, G3, G4 and G5 motifs [80], and biochemical studies have found that the centaurin-γ group harbor very low nucleotide selectivity. The non-canonical G4 motif is thought most likely to cause the reduced selectivity [81], nonetheless the centaurins can hydrolize nucleotides, so these proteins are potentially NTPases rather than GTPases per se and can thus be considered class iii pseudoGTPases (Figure 2). Mechanistically, the pseudoGTPase domain may play an intramolecular regulatory role for centaurin-γ GAP domains [82].

The p190RhoGAP proteins (ARHGAP35, ARHGAP5)

The p190RhoGAP proteins, p190RhoGAP-A (ARHGAP35) and p190RhoGAP-B (ARHGAP5), are multidomain GAPs for Rho GTPases. The p190RhoGAPs negatively regulate Rho signaling to control actin cytoskeleton rearrangement and affect such cellular processes as adhesion, spreading, and migration [83, 84]. Their localization to the membrane, where RhoA is associated, is thought to be driven in part by interaction with protein binding partners [30, 32, 85]. The p190RhoGAPs include a RhoGAP domain at the C-terminus and a GTPase-like domain at the N-terminus [86–88]. Recent structural and biochemical studies have demonstrated that the N-terminal GTPase domain is a nucleotide binding but catalytically incompetent class ii pseudoGTPase with substitutions at key catalytic positions in the Switch II motif [89] (Figure 2). Strikingly, the crystal structure of the N-terminal GTPase contains 6 unique inserts that create a contiguous sub-domain which seems to occlude canonical GAP/GEF and effector protein binding sites. This sub-domain is well conserved, and consequently it has been suggested that the N-terminal pseudoGTPase domain may act as a signal modifier or protein binding domain [89].

In addition to the N-terminal GTPase domain, p190RhoGAP proteins have also recently been shown to be multidomain-pseudoenzymes. Two additional pseudoGTPase domains were discovered within a previously untabulated region, revealing that the p190RhoGAP proteins contain an unprecedented three pseudoGTPase domains [90]. These new domains were termed pG1 and pG2 and are severely degenerate in their G motifs and do not bind nucleotide, so are considered class i nucleotide non-binding, catalytically incompetent pseudoGTPases [24] (Figure 2). Functional studies showed that deletion of the newly identified pseudoGTPase domains (pG1 and pG2) results in an increase in the levels of active (GTP-bound) RhoA, however, the individual importance of each of these domains and the mechanism for this increase are not currently understood [24]. These domains may act as a signal modifier or functional scaffolds, but there is also the possibility that similar to FERM domains (which comprise a tri-foil of three sub-domains: ubiquitin-like, acyl-CoA-binding protein-like, and phosphotyrosine binding/pleckstrin homology-like [91]), the pseudoGTPase domains of p190RhoGAP proteins could be component parts of a larger folded architecture.

Centromere protein M (CENP-M)

Centromere protein M (CENP-M) is part of the large kinetochore complex that links centromeric chromatin to spindle microtubules during mitosis. It is a required component for chromosomal alignment as well as mitosis; specifically, CENP-M is a member of a tetramer in the inner kinetochore that also includes CENP-I, CENP-H, CENP-K, and binds directly to CENP-I. The integrity of this CENP-H, -I, -K, -M complex is required for proper kinetochore assembly [20]. Bioinformatics including secondary structure analysis first predicted that CENP-M folds as a small GTPase but with key residue changes that suggested it lacked catalytic activity [92]. Accordingly, the crystal structure confirmed the GTPase fold, and is most similar to the Rab proteins; furthermore, biochemical studies supported the conclusion that CENP-M does not bind GTP, GDP, ATP or ADP [20]. Thus, CENP-M is a class i pseudoGTPase. Examination of its sequence shows it contains divergent G1 (P-loop), G4 and G5 motifs, is mutated in the G3 motif (Switch II), and contains a large deletion which removes the G2 motif (Switch I) and an adjacent β-strand [20]. It has been postulated that CENP-M evolved from active GTPases. This suggestion is based on distantly related orthologs which contain a conserved G1/P-loop motif but lack other consensus catalytic residues, suggesting that GTP binding was retained even after enzymatic activity was lost [20]. Phylogenetic analysis also shows that CENP-M clusters near the base of the Ras/Rab subfamily, indicating it may have originated at the same time as these active GTPases [20]. Interestingly, CENP-M is missing from fungi, which raises the possibility that the critical scaffolding role in fungi is fulfilled by a different GTPase protein [20, 93]

Dynein light intermediate chain (LIC1, LIC2 and LIC3)

The dynein light intermediate chain (LIC) proteins are components of the cytoplasmic dynein complex which acts as a motor to transport cargo throughout the cell. Dynein LIC1 (and the similar LIC2 and LIC3) contains a globular domain which packs against the dynein heavy chain [94], and was predicted to contain a P-loop nucleotide binding sequence with a Ras-like GTP binding fold [21, 95]. This was confirmed by the crystal structure of C. thermofilum LIC that revealed a GTPase-like fold, but with additional secondary structure features which create a unique topology. Specifically, fungal LIC is a class iii pseudoGTPase with atypical G1 (P-loop), G3 (Switch II), G4 and G5 motifs (Figure 2), and a closed nucleotide binding site unable to bind nucleotide [21]. In contrast, human LIC1 (23% identical to fungal LIC) has conserved G1, G4 and G5 motifs, and thus reportedly binds nucleotide, but is predicted to be a noncatalytic class ii pseudoGTPase based on degenerate G2 and G3 motifs and a preference for binding diphosphate nucleotides [21]. Similar to human LIC1, other metazoan LIC1, LIC2 and LIC3 proteins have conserved G1, G4 and G5 motifs and are also expected to bind nucleotide [96]. The evolutionary differences between fungal and metazoan LICs raises an interesting question about the potential role of nucleotide binding; similar to the N-terminal pseudoGTPase domain of p190RhoGAP proteins [89], nucleotide bindings has been proposed to help stabilize the metazoan LIC domain [21]. Notably, the LIC acts as a scaffolding pseudoGTPase, since it can bind the dynein heavy chain directly.

Non-mammalian pseudoGTPase, EhRabX3

The parasitic protozoan, E. hystolytica, causes amoebic dysentery and liver abscesses in humans. Its genome was recently found to contain over 90 predicted Rab-like GTPase proteins, and these seem to be important for vesicular transport and to play roles in virulence and survival [97]. Notably, 21 of the GTPase sequences lack one or more of the consensus G motifs [97], and the best studied of these, EhRabX3, contains two GTPase domains in tandem (the N-terminal and C-terminal domains), both of which diverge from the consensus sequence of standard GTPases (Figure 3). The C-terminal domain is divergent from consensus GTPases in its G3 and G4 motifs, and biochemical and structural studies show that this domain cannot bind nucleotide (class i) but instead plays a role in stabilization of the N-terminal domain [98, 99]. The N-terminal domain is also divergent in both its G3 (Switch II) and G2 (Switch I) motifs, and when it is studied in isolation cannot bind nucleotide, however in the context of a tandem module containing both N- and C-terminal domains it exhibits cation independent nucleotide binding and slow GTP hydrolysis, consequently it is a class iii pseudoGTPase [98]. Interestingly, the enzymatic activity of the N-terminal domain can be significantly increased by mutating Switch II to a consensus sequence (DIVGK to DIAGQ), suggesting that EhRabX3’s slow hydrolysis is due to its unusual G3 motif [98, 99]. The function of EhRabX3 is currently unknown, but the presence of intramolecular disulfide bonds important for activity and oligomerization suggests it may be a redox sensitive nucleotide binding protein [100]. These results highlight that multiple bona fide pseudoGTPases exist in the simple E. hystolytica genome, with potentially significant implications for signal transduction.

CheY and related bacterial response regulator proteins

Response regulator proteins are not found in metazoans, but are present in bacteria, yeast and plants, where their receiver (or regulatory) domain is well conserved [101]. Interestingly, early work on the E. coli response regulator protein, CheY, suggested a structural similarity to the Ras-like GTPase domain [102, 103]. This receiver (or regulatory) domain does not bind nucleotide, lacks the G2 motif (Switch I), contains a shortened G1 motif (P-loop) which is closed and sterically blocks phosphate binding, and is also non-conserved in the G4 and G5 motifs [103]. In most two-component signaling pathways, the receiver domain is phosphorylated on an Asp residue by the sensor histidine kinase in response to environmental chemical signals; in turn, phosphorylated regulator proteins affect cellular motility by directly binding to motor proteins [101]. Accordingly, the receiver domain has been referred to as a “switch” activated by phosphorylation, and functional comparisons between the phosphorylated receiver domain and GTP-bound GTPases have been made [101, 104, 105]. Nonetheless, the field has not maintained classification of the fold of the receiver domains as Ras-like, and convergent evolution has been posited as the reason for the structural and functional similarities with GTPases [106], but perhaps the recent advances in the pseudoenzyme field suggest that the CheY and similar response regulators could be related to the pseudoGTPases.

Concluding remarks and future perspectives

Recent work has highlighted the importance of “pseudo-” members of most enzyme classes, and that these proteins can have critical roles in regulation and transduction of cellular signals [4, 5, 19, 107, 108]. Among these, the atypical GTPases have been recognized since the 1990’s, and the well-studied Rnd and RGK groups illustrate the functional importance of these proteins [27, 80]. In recent years, the term ‘pseudoGTPase’ has come into use and there have been extensive discoveries of new members of this pseudoGTPase group, including in the p190RhoGAP proteins, dynein LIC, CENP-M and what seems to be a large number of pseudoGTPases in the parasite E. hystolytica. These studies have highlighted that pseudoGTPases are more prevalent than previously recognized but are perhaps less well annotated than other pseudoenzyme classes, suggesting that the study and understanding of this group will benefit from advances in pseudoenzyme classification [109, 110].

As discussed, the loss of conserved residues in the GTPase G motifs of the pseudoGTPases places these proteins into similar classes compared to those observed for other nucleotide-binding pseudoenzymes, such as the pseudokinases [23, 24]. These classes separate into, class i which do not bind nucleotide and are catalytically incompetent, class ii which bind nucleotide but are also catalytically incompetent, and class iii which are divergent in their G motifs but retain catalytic activity by unusual mechanisms. Some of the members of this family are easy to classify (e.g. the complete loss of G motifs in the Ras-like p190RhoGAP pG1 domain supported by extensive biochemistry strongly indicate a class i pseudoGTPase [24]), but due to the divergence of function and ability of unusual residues and factors to compensate for loss of conservation in the G motifs, the classification of pseudoGTPases remains an ongoing process. The continuing studies of pseudoGTPases, supported by biochemistry, structural biology and functional analyses, will improve our understanding of the function and roles of these very interesting proteins.

Abbreviations

ADP

adenosine-5′-diphosphate

ATP

adenosine-5’-triphosphate

BTB domain

broad-complex, tramtrack and bric a brac

CENP-M

centromere protein-M

COR

C-terminal of Roc

DAPK1

death-associated protein kinase 1

EhRabX3

Entamoeba histolytica RabX3

GAP

GTPase activating proteins

GDI

guanine dissociation inhibitor

GEF

guanine nucleotide exchange factor

GDP

guanosine-5’-diphosphate

GTP

guanosine-5’-triphosphate

LRRK2

leucine rich repeat kinase 2

LIC

light intermediate chain

MFHAS1

malignant fibrous histiocytoma-amplified sequence 1

Miro

mitochondrial RhoGTPase

RGK

Rad, Rem1, Rem2, Gem/Kir

Roc

Ras of complex