|Home | About | Journals | Submit | Contact Us | Français|
RHO GTPase-activating proteins (RHOGAPs) are one of the major classes of regulators of the RHO-related protein family that are crucial in many cellular processes, motility, contractility, growth, differentiation, and development. Using database searches, we extracted 66 distinct human RHOGAPs, from which 57 have a common catalytic domain capable of terminating RHO protein signaling by stimulating the slow intrinsic GTP hydrolysis (GTPase) reaction. The specificity of the majority of the members of RHOGAP family is largely uncharacterized. Here, we comprehensively investigated the sequence-structure-function relationship between RHOGAPs and RHO proteins by combining our in vitro data with in silico data. The activity of 14 representatives of the RHOGAP family toward 12 RHO family proteins was determined in real time. We identified and structurally verified hot spots in the interface between RHOGAPs and RHO proteins as critical determinants for binding and catalysis. We have found that the RHOGAP domain itself is nonselective and in some cases rather inefficient under cell-free conditions. Thus, we propose that other domains of RHOGAPs confer substrate specificity and fine-tune their catalytic efficiency in cells.
Hydrolysis of the bound GTP to GDP and inorganic phosphate is the timing mechanism that terminates signal transduction of the majority of RHO family proteins returning them to their inactive GDP-bound state (1, 2). The intrinsic GTP hydrolysis (GTPase) reaction is usually very slow (2) but can be stimulated by several orders of magnitude through the interaction of the RHO proteins with RHO GTPase-activating proteins (RHOGAPs)4 (3,–8). RHOGAPs are defined by the presence of a conserved catalytic domain of ~190 amino acids, which supplies a conserved arginine residue termed the “arginine finger.” This complements an inefficient active site by stabilizing the transition state of the GTPase reaction of the RHO proteins (4, 9,–14). Most remarkably, the same mechanistic strategy has been shown for bacterial GAPs, such as the Salmonella typhimurium virulence factor SptP, the Pseudomonas aeruginosa cytotoxin ExoS, and the Yersinia pestis YopE, even though they do not share any sequence or structural similarities with eukaryotic RHOGAP domains (15,–17).
Mining in the UniProt database led to the identification of 66 distinct human proteins containing a common RHOGAP domain (Fig. 1; Table 1), a number that is slightly different from previous reports (18,–25). Among them p50RHOGAP (26), also known as CDC42GAP (27), p190 (28), and BCR (29) were the first identified and also the best characterized family members. Apart from conserved RHOGAP domains, RHOGAP family proteins possess several sequence motifs and structural domains, which play a role in autoregulation (30), lipid membrane association, subcellular localization, and connection to upstream signals (8, 18, 21, 31, 32). The majority of RHOGAP family members are largely uncharacterized. To date, the selectivity of about one-third of RHOGAPs has been experimentally determined, mainly for their activities toward CDC42, RAC1, and RHOA using diverse methods (Table 1) (8, 18, 21, 25, 31, 33, 34). Despite their significance, the data reported so far do not allow general conclusions about their selectivity (bimolecular recognition and interaction), efficiency (the capability of GAPs to accelerate GTP hydrolysis), and specificity (multimodal interaction at a specific subcellular site influenced by additional domains, motifs and scaffold/adapter proteins) toward RHO proteins. This is mainly due to a large variation of methods and experimental designs.
To revise this status quo, we performed a meta-analysis aiming to evaluate the sequence-structure-function relationship of a variety of RHOGAPs and RHO proteins under cell-free conditions. Therefore, we first measured the GAP activities of 14 RHOGAPs toward 12 GAP-competent RHO proteins, related them to the intrinsic GTP hydrolysis properties of RHO proteins (35, 36) and calculated the fold activation by RHOGAP domain. Second, we combined obtained data with sequence alignments, evolutionary analysis, and accessible structural and functional data from previous studies. All information was then systematically assessed in an ensemble approach focusing on various biochemical aspects of the RHO-RHOGAP interactions. Extracted data at the final stage enabled us to predict activity and selectivity of 66 RHOGAPs and to conclude that the specificity of the non-selective GAP domain in cells is most likely determined by other functional motif(s) and domain(s) in the respective polypeptide chain.
Our original intention was to inspect the GAP activity of 14 representative RHOGAPs (Table 1; supplemental Fig. S1) toward 14 RHO proteins with GTPase activity (supplemental Fig. S2) (35, 36). GTPase-deficient and GAP-insensitive RHO family members, such as RND proteins and RHOH/TTF (35), were excluded. As purified WRCH1 and CHP/WRCH2 proteins were not stable in our hands, the following 12 RHO proteins were included in our study: RHOA, RHOB, RHOC, RAC1, RAC2, RAC3, RHOG, CDC42, TC10, TCL, RHOD, and RIF.
Stimulated GTP hydrolysis reaction was measured by real time fluorescence spectroscopic methods using purified recombinant proteins. In this method, rapid hydrolysis of tamraGTP by the respective RHO proteins was monitored in the presence of an excess amount of the respective RHOGAPs using a stopped-flow instrument (Fig. 2). Fluorescent tamraGTP has been previously described as a GTP hydrolysis sensor for most RHO proteins (6). To be able to detect activity of RHOGAPs with even very low efficiency, we used a 50-fold higher molar concentration of the RHOGAP domains above the respective RHO protein, in all GAP-stimulated reactions. However, even such excess did not lead to a measurable GAP activity of some proteins, such as OCRL1 and p85α (data not shown), which were therefore excluded. Ultimately, p50GAP (hereafter called p50), oligophrenin 1 (hereafter called OPHN1), GRAF1, RICH1 (also called Nadrin), p190A (hereafter called p190), ABR, MGCRACGAP (also called RACGAP1; hereafter MGC), DLC1, DLC2, and DLC3 were used in this study.
The measurement of the GAP-stimulated tamraGTP hydrolysis was suboptimal for RHOA, RHOB, and RHOC (Fig. 2). Therefore, we turned to other fluorescent nucleotides that enable reliable monitoring of real time kinetics of the hydrolysis reaction by RHO isoforms. We examined the following GTP analogs: ATTO550-GTP, ATTO495-GTP, ATTO488-GTP, FAM5-GTP, and cy3GTP. As the p50-stimulated cy3GTP hydrolysis by RHOA provided a substantially distinct decrease in fluorescence (Fig. 2), we measured and evaluated the RHOGAP activities toward RHOA, RHOB, and RHOC using cy3GTP.
In total, we have measured hydrolysis of fluorescently labeled GTP by 12 RHO proteins in all mutual combinations with 10 RHOGAPs. Evaluated observed rate constants (kobs) are shown in Fig. 3 as a bar diagram and summarized in Table 2 in fold activation. Intrinsic GTP hydrolysis was used as control experiments.
In general, the investigated GAPs do not show any selectivity toward particular RHO proteins or their isoforms. p50 appears to be the most universal GAP as it stimulated the GTP hydrolysis of all RHO proteins more than 200-fold as compared with the intrinsic GTPase reaction (Fig. 3). However, there is a large difference between the highest activity toward CDC42 with a kobs value of 14 s−1 and the lowest activity toward RHOD with kobs value of 0.12 s−1 (Fig. 3). p190 was also highly active on all RHO proteins with a relatively low stimulation of the RIF GTPase reaction. Noteworthy, p190 revealed a high activity toward RHOD that was as efficient as p190 activity toward RHOA. OPHN1, a Bin/Amphiphysin/Rvs (BAR) domain-containing protein (see Fig. 1), exhibited overall the highest activities reaching a stimulation of 5 orders of magnitude in the case of RHOA and RHOB. GRAF1, an OPHN1-homolgous protein, was found with an absolute kobs value of 90 s−1 for CDC42 as a fastest stimulated GTP hydrolysis reaction among the GAPs investigated in this study. Differences between the fastest and slowest stimulation for OPHN1 or GRAF1 remarkably exceed 3 orders of magnitude, pointing to an extreme span of measured activities for a single GAP.
Intermediate activities were measured for MGC, RICH1, and ABR, which are still able to stimulate the GTP hydrolysis of measured RHO proteins but to a significantly lower extent than the previously mentioned RHOGAPs, especially for TCL and RHO isoforms. Accordingly, all three proteins can be classified as preferential to CDC42 and RAC isoforms, with an addition of MGC acting on RHOD and RICH1 acting on TC10. Rather inefficient GAP activities were detected for the DLC isoforms, except for the DLC1 activity toward the RHO isoforms and marginally toward CDC42, RAC1, and TC10. Overall disability of DLC2 and DLC3 proteins to operate on analyzed RHO proteins raises a question about proper conditions under which these proteins may exert their GAP functions.
A remarkable finding of our analysis is a broad spectrum of catalytic efficiencies and substrate-selective properties of investigated proteins ranging from a 1-fold to a 120,000-fold stimulation of the intrinsic GTP hydrolysis (Table 2). To illustrate this explicitly, we plotted all 120 pairs of RHOGAP and RHO proteins (x axis) against fold activation (y axis) in numeric order starting with OPHN1-RHOB with the highest efficiency and ending with DLC2-RHOG with no activity (Fig. 4; Table 2). Overall, the RHOGAP-RHO protein pairs were subdivided into six groups based on their catalytic efficiency to stimulate the intrinsic GTP hydrolysis of the RHO proteins (Fig. 4). OPHN1 and its homolog GRAF1 form the first group as they have emerged as highly efficient GAPs not only for the RHO isoforms but also for CDC42 and TC10 (Fig. 4). p50 and p190 also belong to this group of GAPs with the highest catalytic efficiency particularly for their selectivity toward RHOB and RHOD, respectively. MGC and RICH1 rank as the second highest efficiency group because of their activities toward RAC1, RHOD, and TC10. We also indexed p190 to this group as it clearly revealed significantly higher activities toward RAC1 and RAC3. The third group with intermediate efficiency is interestingly populated by RHO isoform-specific DLC1 and RAC-specific ABR. The p50-RIF pair is the most active in the fourth group, which overall displays low efficiency. However, as p50 showed a poor selectivity (Table 2), it is rather doubtful that p50 may be a physiological RIFGAP. Caution should be applied when looking at the data of group five (with 45 pairs in the largest group) in Fig. 4 (pairs between 10- and 150-fold activations). To this group belong protein pairs with the lowest activity, for instance the DLC isoforms on the one hand and RIF and RHOD on the other hand. We scored this group despite their obvious but low GAP activities as inefficient pairs. A very small group of only four pairs with an output of less than 10-fold activation was categorized as the sixth group and graded as “inactive” due to their extremely low catalytic efficiency.
Two critical steps that may control the catalytic efficiency of the GAPs under the conditions used in this study are as follows: (i) association of the RHOGAP with the GTP-bound RHO protein and (ii) the stimulation of the GTP hydrolysis reaction itself. To examine whether an association-controlled mechanism is a reason for the extreme differences in the catalytic efficiency, we loaded CDC42 with tamraGppNHp, a non-hydrolysable fluorescent GTP analog (6), and measured in real time its association with the RHOGAPs. As shown in Fig. 5, there is a clear correlation between the reaction rates of association and GTP hydrolysis. Correspondingly, DLC isoforms, for example, revealed a 1500-fold lower association rate when compared with OPHN1 (Fig. 5, left panel). These data strongly suggest that the catalytic efficiency of the RHOGAPs is directly proportional to the rate of their association with the RHO proteins.
To inspect molecular details of interaction between RHO proteins and RHOGAPs, we have analyzed 42 structures of RHOGAP alone and in the complex with RHO proteins available in the PDB (supplemental Table S1). Residues involved in intermolecular interaction were defined to have at least one inter-atomic distance shorter than 4.0 Å. They constitute interacting interface highlighted on the crystal structure of RHOA-p50GAP (Fig. 6A). We have extracted information about interacting amino acids from different complex structures and combined them with sequence alignments of all investigated proteins in the form of an interaction matrix (Fig. 6B). Each element of the matrix, which we call “hot spot,” relates one homologous residue from RHO proteins to one homologous residue from RHOGAPs (see also supplemental Figs. S1 and S2). The number value of this element represents the number of complex structures in which these residues interact. Thus, a zero value of the element means that these two residues do not face each other in any structure, whereas value 8 means that this particular interaction is to be found in all known structures. We have sorted the residues at both sides of the matrix according to the conservation versus variability. As can be seen (Fig. 6B), more than a half of the residues (17 out of 24) on the side of RHO proteins are identical or highly conserved (Gly/Ala-15, Ser/Thr-37, Asp/Glu-64, and Asp/Glu-65). These residues comprise mostly switch I and switch II and create a continuous patch on the surface (Fig. 6A, left panel). On the other side, only six amino acids are identical in RHOGAPs; five are homologous, and the majority of the GTPase interacting residues are variable (Fig. 6A, right panel). Strikingly, identical and conserved residues form a patch on the GAP domain. We postulate that the interaction between conserved patches of RHO proteins and RHOGAPs is responsible for both the recognition of the two proteins and the catalysis of GTP hydrolysis. This hypothesis is supported by the fact that an identical arginine 282 (p50 numbering), known as the arginine finger, essential for the catalysis (12, 13, 37), is a central residue of the conserved patch on GAP and contacts only identical residues on RHO proteins (Fig. 6). Interactions on this region are not expected to contribute to the differences in activities (or selectivities) and therefore were excluded from further analysis. However, the number of remaining variable pairs is still high and indicates that the relationship between observed diversity of stimulation and molecular interactions is not simple but is quite complex and multifaceted.
The four most effective GAPs, including OPHN1, GRAF1, p190, and p50, share an asparagine (glutamine in p190) at position 417 (p50 numbering), which is not present in other investigated GAP domains (Fig. 6B). Their predominant counterpart residue in RHO proteins is Tyr-66. Its particular interaction with the amide group may contribute to higher activity. Moreover, OPHN1, the most efficient RHOGAP (Fig. 4; Table 2), has in this region two unique residues containing the hydroxyl group, Thr-283 and Ser-286 (supplemental Fig. S1). Thr-283 undergoes a favorable contact with serine and asparagine at position 88 as well as variable residues at position 90 of RHO proteins (Fig. 6B). RAC isoforms have at the latter position hydrophobic residues that are disfavored by Thr-283 in OPHN1, which contributes to the lower OPNH1-stimulated GTP hydrolysis by RAC isoforms. Going beyond the variable regions of interacting interface, three of four most active GAPs have unique leucine at the position 386, which is otherwise replaced by a lysine or a glutamine (Fig. 6B). Its counterpart residue on the side of RHO proteins is the invariant Tyr-34. The nature of the interactions, in which they are involved, is contributing to observed differences. Tyrosine can be involved either in hydrophobic interactions with leucines by utilizing its phenyl moiety or in electrostatic interactions with lysines employing its hydroxyl group.
Similarly, DLC isoforms that were found to be least efficient in stimulating the GTP hydrolysis of RHO proteins contain within the interacting interface unique positively charged residues, arginine and lysine at positions 409 and 413, respectively (Fig. 6B). Such amino acids at these positions are rather unfavorable because the presence of prevailingly hydrophobic residues is required in this region of the GAP surface as it contacts the hydrophobic patch on RHO proteins formed by invariant Val-36, Phe-37, Leu-67, and Leu-70 (RHOA numbering; Fig. 6). However, considering only DLC GAPs, there are no differences in their interacting residues that could explain partial selectivity of DLC1 for the RHO isoforms. Although it is also not directly possible to interpret an overall low activity of RHOGAPs on RHOG, RHOD, and especially RIF, the interaction matrix enables us to determine the regions that are very likely responsible for these low activities. They include variable positions on RHO proteins, e.g. 90, 97, and 134, and on RHOGAPs, e.g. 283, 286, 287, 288, and 309 (Fig. 6B). A similar situation exists for a considerable effect of p190 on RHOD. RHOD has a unique threonine at position 35, but it interacts with identical residues of GAPs (Fig. 6B). On the other side, p190 also has unique amino acids at positions 323, 390, 408, and 413, but they interact reciprocally with identical residues of RHO proteins. Both proteins interact further through variable regions; thus, a full elucidation of a broad spectrum of GAP catalytic activities on RHO proteins would require global evaluation of synergic effect of multifaceted interaction between varying amino acids.
Taking into account that RHO proteins and RHOGAPs are highly homologous, it is legitimate to assume that yet unknown complex structures will share the same structural architecture. Consequently, corresponding residues according to sequence alignments are expected to interact in the manner of known complex structures. Thus, in the absence of structural information for some RHO-RHOGAP complexes, the interaction matrix enables us to deduce which amino acids could be involved in the interaction between these two proteins.
To prove the validity of such an hypothesis, we solved the crystal structure of RHOA in complex with the GAP domain of p190 at high resolution (supplemental Table S3; Fig. 7A), and we used this structure as a benchmark for the verification of our assumptions.
As expected, the overall structure arrangement of the RHOA-p190-GAP complex is similar to complexes of RHO proteins with other GAPs (similarity of p190 to p50 and GAP20 is 25.1 and 17.9%, respectively; supplemental Table S2). The RHOA structure corresponds to an active GTP-bound conformation and clearly differs from its GDP-bound form (Fig. 7A). Conformation of p190 differs in some regions from the structures of p50 and ARHGAP20 (Fig. 7B). However, most relevant for our study were both its high structural similarity in the conserved region of the interacting interface and its conformational variability within the loop between residues 1406 and 1419 (Fig. 7C). Position and orientation of catalytic Arg-1284 are very similar to those found for the arginine fingers in other complexes (supplemental Table S1). We note that the weak electron density around the residues 26–31 of RHOA did not allow us to fully build its switch I region.
Relating the complex structure to interaction matrix, nonzero matrix elements can predict possible interactions in the RHOA-p190 complex. A higher number of the element indicates higher probability of contact occurrence in complex (Fig. 6B). We have thus calculated interaction matrix only for RHOA-p190 complex structure and compared it with the original interaction matrix. The vast majority of conserved residues, which were predicted to be in the contacts, are indeed presented in the structure (Fig. 7). One exception is a conserved arginine at position 323 (Fig. 6B, p50 numbering), which is exclusively a serine in p190A (Ser-1326). These residues are supposed to interact with Glu-65 of RHOA. Structure of the RHOA-p190 complex revealed that Ser-1326 at this site is in the vicinity of Glu-65 but is simply not long enough to form the contact with it (data not shown). Largest discrepancies between predicted and observed contacts comprise the interaction of “conserved” patches of RHO proteins and “variable 1” of RHOGAPs. The reason is that sequence alignment of all RHOGAPs (supplemental Fig. S1) contains in this region many gaps and shifts that preclude reliable prediction of its structure and proper assignment of similar residues. However, it has to be borne in mind that the interaction matrix was constructed on the basis of only four crystal structures. Including more structures for its calculation would certainly increase its reliability and enable more precise prediction of unknown complexes between RHO proteins and RHOGAPs.
An obviously demanding question is as follows. To what extent does the substrate selectivity of the RHOGAP domain determined in this study under cell-free conditions reflect the cell-based specificity of RHOGAPs and how relevant is that to multicomponent and multidomain cellular machineries? Addressing this question is of ultimate importance due to the fact that RHOGAPs did not in general reveal strong selectivity for some particular RHO proteins (Fig. 3; Table 2). An interesting observation in this regard is the high activity of p190 toward RHOD that is comparable with its activity toward RHOA (Fig. 4; Table 2). To answer the question of whether the p190 is also a GAP for RHOD in cells, we used GST fusion of DIA-RBD and RTKN-RBD to pull down GTP-bound RHOD and RHOA, respectively (38). Strikingly, the obtained data revealed that the amount of pulled down RHOD-GTP remained unchanged in cells overexpressing either full-length (FL) or the GAP domain of p190 (Fig. 8, A and B). Control experiments showed that p190-FL indeed acts as a GAP for RHOA, but overexpression of only the GAP domain failed to stimulate GTPase activity of RHOA (Fig. 8, C and D). These results provide a clear indication that other domains and motifs of p190-FL determine its specificity for RHOA in cells and not, for example, for RHOD. Extrapolating it on the activities of the GAP domains measured under cell-free conditions for members of the RHO family, we postulate that they are not directly proportional to specificities of GAP proteins within the cell.
Results described above brought us the question about the activity of GAP domains in the context of the full-length proteins and their niche within the cell. Another question in the same context is whether and to what extent GAPs can regulate multiple signaling pathways, which in turn seem to be dominated by the composition of their domains as illustrated with our data. Therefore, we conducted a phylogenetic analysis of 66 human RHOGAP proteins (Table 1) based only on the sequence of respective domains of the RHOGAP family members. The phylogenetic order correlated with the arrangement of proteins according to their domain and motif compositions (Fig. 1). We assigned so far 33 different domains with different properties (supplemental Table S4). The majority of them can be classified into the following three major groups: (i) lipid and membrane binding domains; (ii) peptide and protein interacting domains; and (iii) catalytic domains with enzyme activities (Fig. 1). Most widespread domains are pleckstrin homology (30), CC (25), P (16), Src homology 3 (15), and BAR/F-BAR (14). Most RHOGAPs have 3–4 additional domains, whereas CNT-D1 has another 10 and MYO9B even 11 domains (Fig. 1; supplemental Tables S4 and S5). Thirteen GAPs lack any additional putative domains but contain highly variable regions at their N and C termini. It is possible that these regions consist of not yet identified motifs, which may contribute to their specific function in the cell. ARHGAP11B is the smallest RHOGAP protein belonging to this group (Fig. 1; Table 1). A blast search of the terminal 63 and 17 amino acids of ARHGAP11B revealed a consensus motif, KLL(X5)RED, at its C-terminal region, which exist in many proteins (data not shown). Both KLL and RED motifs have been reported to be involved in protein-protein interactions (39,–41).
The most efficient activator of the GTPase reaction among investigated GAP domains is OPHN1 that stimulated GTP hydrolysis of RHOA and RHOB up to 5 orders of magnitude as compared with the other investigated RHOGAPs. Its second striking feature is that it can efficiently deactivate a broad spectrum of investigated RHO proteins, including the RHO and RAC isoforms, CDC42, TC10, and TCL (Table 2). The ability to stimulate efficiently GTP hydrolysis of various RHO proteins was also observed for GRAF1, p190, and p50. These four GAPs are in general active on the same RHO proteins. Least susceptible are RHOG and RIF that could be in fact deactivated only by p50.
Although the GTP hydrolysis of RHOG was to a limited extent stimulated by OPHN1 and p190, there is no GAP in our set that could actually act on RIF. Remarkably, GTP hydrolysis of RHOD was only markedly stimulated by p190 and MGC. The spectrum of activity for MGC on the different RHO proteins is not as broad as the spectrum of above mentioned GAPs. Its activity on RHO isoforms is about 10-fold lower compared with RHOD, CDC42, and RAC isoforms, and it is practically inactive on TC10 and TCL. RICH1 was found to be even less effective, as it significantly stimulated GTP hydrolysis reaction of only TC10 and to a lesser extent of CDC42, RAC1, and RAC2. Interestingly, TCL, RHOG, and RHO isoforms are less sensitive to hydrolysis by RICH1, but in contrast, RHO isoforms appeared to be an exclusive substrate for DLC1. In fact, a unique selectivity was observed for the DLC1 GAP domain, which acts specifically on RHOB, RHOA, and RHOC (Table 2). ABR was the least effective, but it effectively deactivated RAC1 and RAC3. However, we do not designate it as selective for these RAC isoforms because its hydrolytic activity toward other GTPases was not so distinct as the activity of DLC1 for RHO proteins. DLC2 and DLC3 did not show any considerable GAP activity, an observation that raises the question whether these proteins still can be considered as RHOGAPs.
Taken together, there is no certain selectivity between investigated RHOGAP domains and particular RHO proteins or their isoforms. Our observation regarding the selectivity of GAP domains is in contrast to the guanine exchange factors (GEFs) of diffuse B-cell lymphoma (DBL) family that activate RHO proteins by accelerating the GDP/GTP exchange reaction. We showed in previous study that their isolated DBL homology (DH) domain, which is actually responsible for nucleotide exchange, showed both selectivity and specificity for their substrate RHO proteins (36). This finding challenges the fundamental principles of cell signaling exemplifying that there are two principally different manners of interplay between the RHO proteins and their regulatory proteins. In the first case of DBL family GEFs, the catalytic DH domain directly interacting with the substrate RHO protein is itself able to selectively discriminate among the RHO proteins, and other additional domains and motifs of the full-length RHOGEF protein provide an additional degree of regulation in the cell. In the second case, when the catalytic domains directly interacting with the substrate RHO proteins do not show any distinct selectivity, as we have found for the RHOGAP domains, secondary domains and motifs of the full-length RHOGAP inevitably determine, beyond other features, the specificity for the substrate RHO proteins. This is nicely demonstrated in this study by an example of p190 protein. Its GAP domain was equally and highly active on RHOA and RHOD under cell-free conditions, but its full-length version in the cells was able to specifically inactivate only RHOA and not RHOD. The existence of multiple determinants in full-length RHOGAPs, which dictate their localized recruitment, activation, “specific” function in cells by including distinct protein and lipid interaction domains and motifs, as well as post-translational modification, has been suggested by several previous articles (18, 42,–47). It has been shown that the spatial distribution of RHOGAPs and their specificity toward individual RHO proteins are controlled by their interactions with various proteins within signaling complexes (48, 49). Our results thus elegantly complement the scenario for the function of GAP proteins in which a concerted action of the whole protein is required. Accordingly, we conclude that p190 cannot be recruited to RHOD because it is RHOA-specific or that p190 and RHOD do not find each other under the used experimental conditions. We can, however, not exclude the possibility that p190 might specifically operate on RHOD in a specific cell type.
To shed light on the molecular interactions of RHOGAPs with RHO proteins, we have analyzed available crystal structures of their complexes and combined the data about interacting residues with two multiple sequence alignments of investigated RHOGAPs and RHO proteins in the form of a structure-based interaction matrix (Fig. 6A). Such interaction matrix allows us to predict which residues of two sets of homologous proteins are likely to interact in their binary complexes. In addition, it provides a complete overview of the occurrence of particular contacts in analyzed structures as well as the conservation or variability of respective amino acids utilized by both GAP domain of RHOGAP proteins and G domain of RHO proteins upon interaction.
In terms of conservation, GAP side residues are largely variable (supplemental Fig. S2) in contrast to the RHO side residues, which are mostly identical. The variability of the latter originates almost exclusively from the helix 3 and the insert helix (Figs. 6 and supplemental Fig. S2). Reordering of residues in the matrix according to their conservation enabled us to divide hot spots into three distinct regions that also correspond to three exclusive regions on the interacting interface. Each of these regions also includes distinctive interacting pairs of amino acids (Fig. 6; supplemental Fig. S1–S3, color-coded regions). They can be classified into three different groups as follows: interaction pairs of conserved residues, pairs of variable residues on both sides, and the interactions of conserved residues on RHO side and variable residues on GAP side.
We have hoped that our analysis would reveal special distinctiveness in the interactions between RHO and RHOGAP proteins that could at least semi-quantitatively explain differences in observed activities. What we have found instead was an abundance of combinatorial possibilities and the complexity incorporated in the formation of binary protein complexes. To relate observed stimulations of GTP hydrolysis with sequence differences among investigated proteins and to describe them quantitatively, the contributions from all matrix points have to be considered. Each element in the matrix represents in principle the combination of 12 RHO protein and 10 RHOGAPs, e.g. 120 possibilities. To assess the contributions of all such combinations requires the evaluation of the impact of all different amino acids at each particular spot. For example, in the case of invariant Tyr-34 of RHO proteins, all types of its interactions with leucine, lysine, or glutamine of GAPs at position 386 have to be considered (Figs. 6B and supplemental Fig. S3). The interaction properties of these amino acids are as diverse as their chemical properties. The situation is even more complex for the second discussed invariant Tyr-66 in RHO proteins because it contacts nine different amino acids of RHOGAPs (Fig. 6B). Spots that have variable residues on both sides of the matrix would require an even more thorough evaluation. Finally, an overall contribution of all individual elements from the interaction matrix would have to be correlated with observed differences in activities to obtain fully qualitative description.
Interaction matrix also allows to predict which residues of one RHO protein would interact with one of RHOGAPs. To validate this approach, we have solved the structure of RHOA-p190GAP complex (Fig. 7A), calculated the interaction matrix exclusively for this structure, and compared it with the original interaction matrix (Fig. 7C). As can be seen, a majority of residues interacting between conserved patches could be successfully anticipated. Most of deviated contacts pertain to the interactions between conserved RHO residues and variable 1 of RHOGAPs. The reason is the gaps in sequence alignment of GAP domains, namely in hypervariable regions (supplemental Fig. S1). They enable shifts of corresponding amino acids so their space positions in the presumed complex structures might differ from positions found in known structures. A comparison of three distinct RHOGAPs from complex structures (supplemental Table S1), i.e. p50, RAGAP20, and p190, nicely demonstrates that the conformation of this variable region is indeed very diverse. The whole loop in p190 comprising residues 1406–1419 is folded completely differently when compared with the corresponding loop of p50 (Fig. 7B). Interaction of this region of RHOGAPs with conserved residues of RHO proteins seems to be responsible for observed differences in their activities. However, interaction of conserved patches on both sides of complexes is preserved and exclusively determines the formation of the complex between RHO proteins and RHOGAPs.
Several cell-based studies have shown that there is specificity between RHOGAP and RHO proteins as follows: ARHGAP15, BCR, β-chimaerin, 3BP1, p68RACGAP, and FILGAP are specific for RAC1 (29, 50,–54); RALBP1 and MGCRACGAP1 are specific for both CDC42 and RAC1 (55,–57); RICH1 and CDGAP are specific for CDC42 (58, 59); ARHGAP6, DLC1, DLC3, myosin IXb, OPHN1, p190A, and RA-RHOGAP are specific for RHOA (34, 47, 60,–68); ARHGAP18, ARHGAP21, and DLC1 are specific for RHOC (64, 66, 69); TCGAP is specific for TC10 (70); PARG1 is specific for RHOA (71), and ARHGAP30 is specific for WRCH1 (72).
Our observations, however, show that RHOGAP domains are not able to selectively deactivate particular RHO protein or its isoforms. Such discrepancy raises the question of what other factors, processes, or circumstances may determine the specificity between RHOGAPs and RHO proteins. Cellular context is certainly crucial as, for example, studies on p190 (73, 74) and myosin IXb (75) showed that these proteins have different specificity in vitro and in vivo. There are several possibilities for the regulation of the GAP activity in cells ranging from intermolecular autoinhibition (β-chimerin, p50, OPHN1, and DLC1) (8, 30, 76, 77) to post-translational modifications (78). For instance, “activating” phosphorylation generates in p190 a new contact site for p120RASGAP (79), releases DLC1 from its autoinhibited state (76), or converts MGCGAP to a RHO-specific GAP (80). Furthermore, SUMOylation of ARHGAP21 may represent a way of guiding its function (81), and non-proteolytic ubiquitination of p250GAP controls axon growth (82).
The fact that almost all GAP proteins consist of several diverse domains and motifs strongly indicates that the regions accompanying the GAP domain are crucial for their function. Sixty-six GAPs identified in the human genome contain 33 different domains (Fig. 1; supplemental Tables S4 and S5). Some of them possess up to five different domains, and there are some proteins that contain four or even five copies of the same domain (Fig. 1; supplemental Table S5). Domain composition of GAPs together with the nature of individual domains demarcates on a higher level their subcellular localization and function. For example, the BAR domain of OPHN1 and GRAF1 is simultaneously involved in membrane tubulation and GAP inhibitory functions (8). The SEC14-like domain of p50 homology appears not only to regulate the GAP activity (77) but also to localize p50 in the endosomal membrane as a link between RHO and RAB proteins (83). Similarly, phospholipid binding to the C1 domain both recruits β-chimerin to the plasma membrane and activates its RACGAP activity (30). The RHOGAP activity of DLC1 has been proposed to be inhibited by an intramolecular interaction between the SAM and RHOGAP domains (84). Phosphorylation by CDK5 and association with both phospholipids and the scaffold proteins tensin and talin has been shown to release DLC1 from its inhibited state and to significantly promote its RHOGAP activity (67, 76, 84, 85). C2 or pleckstrin homology domains of GRAF, ABR, OPHN1, and MGC are the modules mediating association with the membrane according to the calcium-dependent phospholipid binding or phosphatidylinositol concentration (8, 31, 80, 86,–90).
Being the smallest human RHOGAP, ARHGAP11B is not decorated with any known domain or motif (Fig. 1). The involvement of ARHGAP11B in neuronal development by promoting basal progenitor amplification and neocortex expansion has been reported recently (91). This study has shown that ARHGAP11B does not exhibit RHOGAP activity as compared with AHRGAP11A and its variants. However, its GAP activity was measured indirectly by monitoring more downstream RHOA/ROCK activity. ARHGAP11A and -11B share 90% identical sequences in their RHOGAP domain and mainly differ at the very C-terminal end (91), which is highly variable in all RHOGAPs (supplemental Fig. S1). Other residues that are essential for the RHOGAP activity are highly conserved in ARHGAP11B indicating that this GAP, although very small, may act as GAP for RHO proteins. In this context, the C-terminal KLL and RED motifs that were detected in a blast search in this study (see above) may play a role in protein-protein interactions (39,–41).
Functionalization of the RHOGAPs with various modular building blocks, especially the membrane-associating domains, is a prerequisite for successful orchestration of a series of spatiotemporal events, including recruitment, subcellular localization, assembly of proactive protein complexes, and ultimately association with and inactivation of the substrate RHO protein. Reduced dimensionality on distinct regions of the cell membrane does not only achieve high specificity of the RHOGAPs but also tremendously enhances their overall catalytic activity.
The efficiency of a RHOGAP depends largely on the cellular processes, in which they are involved. There are very fast processes, e.g. calcium fluxes, exocytosis, or muscle contraction, and very slow processes, e.g. differentiation, apoptosis, or metabolism, which are also very much dependent on cell types. The GAP protein that is inefficient under cell-free conditions may efficiently operate through the function of its other domains in an appropriate cellular niche. An example is provided by DLC proteins that were mostly found as inefficient or even inactive in this study (Fig. 4). DLC1 has been thought to play a major role as a tumor suppressor probably in a GAP domain-independent manner (92). However, the DLC1 activity is, as compared with DLC2 and DLC3 activities, relatively high toward the RHO isoforms, RAC1, CDC42, and TC10 (Fig. 4). Thus, it is conceivable that additional mechanisms contribute to further enhancement of its GAP activity, comprising CDK5 phosphorylation, association with scaffold proteins, such as tensin and talin, and/or association with lipid membranes (67, 76, 84, 85). However, there are also mechanisms to inhibit the DLC1 RHOGAP activity, including phosphorylation by protein kinases C and D, and subsequent association with 14-3-3 proteins (62) or direct association of the Src homology 3 domain of p120RASGAP with its RHOGAP domain (32, 93). Similar regulatory mechanisms have been proposed for DLC2 and DLC3 (67) suggesting that inefficient RHOGAPs under cell-free conditions can be highly efficient in proper cellular context and appropriate protein network.
An inefficient GAP can otherwise be employed in the control of a slow cellular process, including actin dynamics. A group of nonconventional RHO proteins, such as RHOD, RIF, and RAC1b, mainly persists in their active state under resting conditions (35, 94). They accumulate in their GTP-bound state and thus are essentially dependent on a specific GAP to be switched off (35). Both RHOD and RIF are involved in the integration of cytoskeletal reorganization and membrane trafficking (95, 96); however, specific RHOGAPs for these atypical members of the RHO family remain to be discovered.
According to the mechanism of the GAP-stimulated GTPase reactions, the RHOGAP domain supplies an arginine finger directly into the active site of the substrate RHO proteins to stabilize the transition state (13, 97). A first inspection of the sequence alignment of the 66 RHOGAP domains revealed that ARHGAP36, CNT-D1, DEP1, DEP2, FAM13B, INPP5P, and OCRL1 lack an arginine finger at the corresponding position (supplemental Fig. S1). These proteins have serine, threonine, or glutamine instead and thus cannot substitute for the arginine function. ARHGAP36 is poorly investigated. It has been shown to be involved in Gli transcription factor activation but independent on its GAP domain (98). The ARFGAP and RHOGAP domain-containing CNT-D1 (also called ARAP2; Table 1) lacks RHOGAP activity and acts as an ARF6 GAP (99). DEP1 and DEP2 coordinate cell cycle progression and interfere with RHOA and signaling despite lacking RHOGAP activity (100). OCRL1 has been shown to interact with GTP-bound RAC1 without the stimulation of its hydrolysis (101). p85α and p85β (85-kDa regulatory subunits of the phosphoinositide 3-kinases) can also be included on the list of RHOGAP-like proteins (Table 1; supplemental Fig. S1), as they do not show any detectable GAP activity toward different RHO proteins (102). An essential prerequisite of the GAP function is that the GAP domain, in order to position its catalytic residue Arg-282 (p50 numbering), must employ a number of amino acids that are responsible for binding and stabilizing the protein complex (Fig. 6A). Both p85 isoforms lack most of these binding determinants, e.g. Arg-323, Asn-391, Val-394, and Pro-398, along with the conserved amino acids around the arginine finger (p50 numbering; supplemental Fig. S3) (4).
Unlike the RHOGEF domains (so-called DBL homology or DH domains), which exhibit high selectivity for the RHO-, CDC42-, and RAC-like proteins (36), we have found that the RHOGAP domain itself is nonselective and in some cases rather inefficient under cell-free conditions. Thus, we propose that other domains of RHOGAPs confer substrate specificity and fine-tune their catalytic efficiency in cells. They dictate the specificity of the respective RHOGAPs most likely through different successive steps as follows: (i) recruitment to a specific subcellular structure at a given time; (ii) release of its (auto)inhibited and most likely membrane-associated state; (iii) recognition and association with the substrate RHO protein; (iv) complementation of an inefficient active site with a catalytic arginine; (v) stimulation of GTPases reaction by orders of magnitudes; and (vi) finally dissociation from the inactivated GDP-bound RHO proteins. One approach of verifying this hypothesis is conducting RHOGAP domain-swapping experiments in cells using two RHOGAPs with verified specificities. Results may show that the specificity of these RHOGAPs remains unchanged irrespective of recombined RHOGAP domain.
Formation of binary complexes between two classes of proteins, such as RHO proteins and RHOGAPs, is a straightforward biochemical process. However, we have shown that its detailed description requires a sophisticated approach capable of covering a huge number of combinatorial possibilities incorporated in such molecular system. Our structural analysis based on the interaction matrix aspires to be such an approach. Its application led to the division of interacting interface into two parts. The first part determines the formation of complexes and supports the catalytic mechanism, and the second part is responsible for the diversity in catalytic activities. Although it remains to be proven whether such an approach is also applicable to other protein systems, we believe that its further elaboration will enable a precise prediction of interacting residues in the unknown structure of complexes between RHO proteins and RHOGAPs.
A critical issue regarding experimental determination of the specificity of RHOGAPs in cells is that arginine finger mutants, mostly to alanine (RA mutant), are often used to compromise the RHOGAP function. This approach is in principle very useful under cell-free conditions but is not really optimal in the cells because an RA mutant may provide a similar readout as the wild type; it interferes with downstream signaling by competing with the effector(s) for binding to the RHO proteins. RHOGAP mutants at this site are able to persistently bind to and sequester the target RHO protein. This most likely displays a similar readout as the activity of wild type RHOGAP. Instead of the catalytic arginine, we recommend mutating critical “binding determinants,” particularly Lys-319 and Arg-323 (p50 numbering; Figs. 6 and and77B). Charge reversal of these residues most likely leads to loss of RHOGAP association with its substrate RHO proteins and consequently the activity of the GAP domain. This is not only a tool for determining the specificity of RHOGAPs but also for investigating GAP domain-independent function(s) of the RHOGAPs.
Constructs containing the GAP domain of human p50 (amino acids or aa 198–439), GRAF1 (aa 383–583), RICH1 (aa 245–499), p190A (aa 1250–1531), OPHN1 (aa 375–583), ABR (aa 559–822), MGCRACGAP (aa 343–620), DLC1 (aa 609–878), DLC2 (aa 644–916), and DLC3 (aa 620–890) were amplified by standard PCR and cloned in either pGEX-4T1 or pGEX-4T1-Ntev vector. All RHO protein constructs have been reported before (35). Human RHOA, its Gln-63 variant to leucine (Q63L mutant), and RHOD were cloned in pRK5-Myc. Human p190-FL and its GAP domain and their Arg-1284 variant to alanine (RA mutant) were cloned in HA-pKH3 and pRK5, respectively. Rat p190 GAP domain (1242–1439) was cloned into pGST-parallel vector (103).
All RHO proteins and GAP domains of RHOGAPs were purified as glutathione S-transferase (GST) fusion proteins from Escherichia coli BL21(DE3) pLysS or CodonPlusRIL as described previously (2, 104). All RHO proteins and their nucleotide-free forms were prepared as described (2, 105).
Various fluorescence reporter groups, including mant, tamra, and cy3, have been coupled to 2′(3′)-hydroxyl group of the ribose moiety of the guanine nucleotide GTP via ethylenediamine to obtain fluorescent GTP variants (Jena Bioscience, Germany) for the analysis of the GAP-stimulated GTP hydrolysis reactions of the RHO proteins.
All GAP-stimulated GTP hydrolysis fluorescence measurements of RHO proteins were performed at 25 °C. Fluorescent GTP-bound RHO proteins (pre-mixing 0.3 μm nucleotide-free RHO and 0.2 μm tamra-/cy3-labeled GTP) and the catalytic domain of RHOGAPs (10 μm) were rapidly mixed in a buffer containing 30 mm Tris-HCl, pH 7.5, 10 mm KH2PO4·K2HPO4, 10 mm MgCl2, and 3 mm dithiothreitol using a Hi-Tech Scientific (SF-61) stopped-flow spectrophotometer instrument with mercury xenon light source as described (32). For excitation, wavelengths of 546 and 550 nm were used for tamra and cy3 fluorophores, respectively, and a 570 nm (tamra and cy3) cutoff filter (Schott glass) was used to collect emitted light.
Sequence alignments were performed with BioEdit program using ClustalW algorithm (106). The intermolecular contacts were determined (≤4.0 Å) between RHOGAP and RHO proteins using available RHO-RHOGAP complex structures in the Protein Data Bank (supplemental Table S1). A python code has been written including BioPython modules (pairwise2 and SubsMat.MatrixInfo) (107) to get PDB and alignment files and returns corresponding interaction pairs in a matrix form. RHOGAP domains discussed in the matrix have sequence similarities between 20 and 80% (supplemental Table S2) (108) and are assumed to have identical fold and form molecular complexes with similar arrangement. All structural representations were generated using PyMOL viewer (109).
A mixture of RHOA-GDP and p190-GAP with a small molar excess of RHOA was dialyzed overnight in a buffer, containing 20 mm Hepes, 100 mm NaCl, 5 mm NaF, 5 mm MgCl2, 5 mm β-mercaptoethanol. The sample was loaded on a Superdex 200 gel filtration column. Fractions containing RHOA-GDP-MgF3−-p190-GAP complex were pooled and concentrated to 8 mg/ml for crystallization trials. The vapor diffusion method was used for crystallization with sitting drops of 1:1 ratio of protein and crystallization reagent. Best crystals grew from JSCG+ screen (Molecular Dimensions) reagent 82. The crystallization conditions were further optimized, and a buffer, containing 30% PEG2000 MME, 0.15 m KCSN in 0.1 m MES, pH 6.5, produced crystals of diffraction quality. For data collection the crystals were frozen in a cryo-solution containing mother liquor with an addition of 0.2 m ascorbic acid and 12.5% glycerol. X-ray data were collected at Argonne National Laboratory, South-Eastern Region Collaborative Access Team (SER-CAT) beamline of Advanced Photon Source (supplemental Table S3). The structure was solved by molecular replacement method using program Balbes (110) and refined using Phenix (111). Manual rebuilding of the model during refinement was performed using Coot (112). Final refinement statistics can be found in Table S3. Structure was deposited with PDB accession number 5IRC.
RHOD and RHOA were pulled down in their activated states as described previously (38), and HEK293T cells were seeded in 6-cm dishes. Next day, cells were transfected with 2 μg of DNA of pRK5-Myc-RHOA or pRK5-Myc-RHOD together with 1 μg of DNA of HA-pKH3-p190-FL (WT) or pRK5-FLAG-p190-GAP domain using the PolyPlus JetPEI transfection reagent. Cells were incubated 24 h post-transfection, followed by lysis in ice-cold buffer, including 50 mm Tris-HCl, pH 7.5, 1% Triton X-100, 150 mm NaCl, 10 mm MgCl2, and protease inhibitor mixture (cOmpleteTM, EDTA-free, Roche Applied Science). Cell lysates were transferred to pre-chilled tubes and centrifuged at 13,000 rpm for 5 min at 4 °C. 1/25th of each cell lysate was transferred to a new tube, and 3× SDS-PAGE sample buffer was added. The rest of the sample lysates were transferred to new tubes, and GST fusion proteins on glutathione beads were added. GST-RTKN was added to the pRK5-Myc-RHOA samples and GST-DIA1 to the pRK5-Myc-RHOD samples. GST-RTKN and GST-DIA1 were overexpressed in E. coli and isolated from the lysate using glutathione beads as described previously (38). Samples were carefully rotated at 4 °C for 10 min and then centrifuged and washed four times with 0.5 ml of ice-cold buffer, including 50 mm Tris-HCl, pH 7.5, 1% Triton X-100, 150 mm NaCl, 10 mm MgCl2. SDS-PAGE sample buffer (3×) was added to each sample. Samples were separated on 10% SDS-polyacrylamide gels followed by transfer to nitrocellulose membrane. Proteins were detected with 9E10 mouse monoclonal anti-c-Myc antibody (Covance), 12CA5 mouse monoclonal anti-HA antibody (Roche Applied Science), and M2 mouse monoclonal anti-FLAG antibody (Sigma).
M. R. A. conceived and coordinated the study; E. A., M. J., R. D., K. T. K., and K. N. designed, performed, and analyzed the experiments. K. R. and P. A. performed pulldown assays; U. D. and A. V. S. coordinated the structure determination of the RHOA-p190 complex; E. A., R. D., and M. R. A. designed the study and wrote the paper. All authors reviewed the results and approved the final version of the manuscript.
We thank P. Billuart, N. Nassar, M. A. Olayioye, K. Rittinger, I. Whitehead, and L. van Aelst for sharing reagents, cDNA, and plasmids; M. A. Olayioye and B. Vanhaesebroeck for discussions, and Ilse Meyer and Natalya Olekhnovich for technical assistance. The University of Chicago Argonne LLC operates Argonne National Laboratory for the United States Department of Energy, Office of Biological and Environmental Research under Contract DE-AC02-06CH11357. Use of the Advanced Photon Source, an Office of Science User Facility operated for the United States Department of Energy Office of Science by Argonne National Laboratory, was supported by United States Department of Energy under Contract DE-AC02-06CH11357.
*This work was supported by the International Research Training Group 1902 Intra- and Interorgan Communication of the Cardiovascular System Grant IRTG 1902, Deutsche Forschungsgemeinschaft Grant AH 92/5-1, Research Committee of the Heinrich-Heine Düsseldorf Grant 26/2015, and National Institutes of Health Grant R01GM086457. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
This article contains supplemental Tables S1–S5 and Figs. S1–S3.
4The abbreviations used are: