Phosphoinositide lipids recruit proteins to the plasma membrane involved in the regulation of cytoskeleton organization and in signalling pathways that control cell polarity and growth. Among those, Rgd1p is a yeast GTPase-activating protein (GAP) specific for Rho3p and Rho4p GTPases, which control actin polymerization and stress signalling pathways. Phosphoinositides not only bind Rgd1p, but also stimulate its GAP activity on the membrane-anchored form of Rho4p. Both F-BAR (F-BAR FCH, and BAR) and RhoGAP domains of Rgd1p are involved in lipid interactions. In the Rgd1p–F-BAR domain, a phosphoinositide-binding site has been recently characterized. We report here the X-ray structure of the Rgd1p–RhoGAP domain, identify by NMR spectroscopy and confirm by docking simulations, a new but cryptic phosphoinositide-binding site, comprising contiguous A1, A1′ and B helices. The addition of helix A1′, unusual among RhoGAP domains, seems to be crucial for lipid interactions. Such a site was totally unexpected inside a RhoGAP domain, as it was not predicted from either the protein sequence or its three-dimensional structure. Phosphoinositide-binding sites in RhoGAP domains have been reported to correspond to polybasic regions, which are located at the unstructured flexible termini of proteins. Solid-state NMR spectroscopy experiments confirm the membrane interaction of the Rgd1p–RhoGAP domain upon the addition of PtdIns(4,5)P2 and indicate a slight membrane destabilization in the presence of the two partners.
For a cell, polarity is the ability to asymmetrically organize itself in a non-random manner. This is characterized by a particular and organized intracellular morphology, and the precise localization of structures and various cell constituents, for example organelles, cytoskeleton, proteins and lipids. According to their role, cells optimize their organization by activating specific cellular signalling programmes, leading to this cell asymmetry. Cell polarity is thus an essential and common phenomenon . All these processes must be tightly regulated in a co-ordinated manner. These are controlled by signalling pathways which happen to be conserved from yeast to human. The CDK (cyclin-dependent kinase) proteins and small G proteins (guanine nucleotide-binding proteins) are key players in this control . Among the small G proteins, members of the Rho (Ras homology) family are involved in actin cytoskeleton polarization, intracellular trafficking, cytokinesis and cell cycle control . Rho GTPases usually exist in two states: an active GTP-bound form anchored in the membrane and an inactive GDP-bound form. GTPases cycle between these two states in response to various stimuli, a property that has led to the notion of Rho GTPases as ‘molecular switches’. Switching between the active and inactive forms is regulated by two protein families: the Rho guanine nucleotide exchange factors, which facilitate the exchange between GDP and GTP and thus activate the GTPase, and Rho GTPase-activating proteins (RhoGAPs), which negatively regulate the Rho proteins by stimulating the intrinsic activity of GTP hydrolysis by Rho GTPases . A challenge to integrate molecular functions to cellular process consists in determining how the regulators of Rho family are controlled. The GDI, GEF and GAP activities should be co-ordinated leading to an increase in the GTP-bound form through concomitant activation of GEFs and inhibition of GAPs to act in signal transduction controlling actin dynamics [5,6].
In Saccharomyces cerevisiae, six Rho GTPases have been described that are mainly involved with cell polarity and are regulated by RhoGAPs . Previous work demonstrated that Rgd1p is a RhoGAP shown to increase GTP hydrolysis by Rho3p and Rho4p in S. cerevisiae . These two proteins are involved in the establishment of cell polarity at the bud tip and neck, respectively, in yeast [9,10]. The Rgd1 protein contains a RhoGAP domain at its C-terminal part (aa 486–666) and an F-BAR (F-BAR FCH, and BAR) domain at its N-terminal extremity (aa 1–300). In a previous work, we demonstrated that phosphoinositides bind Rgd1p and specifically stimulate its RhoGAP activity on Rho4p . By gel filtration and circular dichroism, we provided the first evidence for a specific interaction between the Rgd1-RhoGAP (RhoGAP domain from related GAP domain 1 protein) and PtdIns(4,5)P2 (phosphatidylinositol-4,5-bisphosphate) , a phospholipid involved in key signalling pathways related to cell polarity [13,14,15].
In this work, we used X-ray diffraction, NMR and docking approaches to characterize the structure of Rgd1-RhoGAP and its interaction with PtdInsP. For both crystal and solution states, we observed all the structural characteristics of the RhoGAP family and the elements necessary for activation of Rho proteins (Figure 1). Between helices A and B, the domain displays an additional non-conserved 310 helix, termed A1′. In a second step, we performed NMR titration experiments with several phosphoinositides. Our results show the presence of a common phosphoinositide-binding site in the protein comprising helices A1, A1′ and B (Figure 1); a novel and unexpected observation for a RhoGAP domain.
Sequence alignment between Rgd1-RhoGAP and close structural homologues.
Protein expression and purification
Rgd1-RhoGAP was prepared as previously described . Briefly, the RGD1 fragment coding for a RhoGAP domain (aa 450–666) was cloned into pET21a vector and expressed at 37° C using the Escherichia coli BL21 (DE3) strain as a fusion protein with a six histidine tag at the C-terminus. The unlabelled protein was produced in DYT-rich medium for crystallogenesis. The uniformly 15N,13C- and 15N-labelled proteins were prepared for NMR experiments by growing E. coli bacteria in M9 minimal medium containing 15N-ammonium chloride (1 g/l) and 13C-labelled and -unlabelled glucose (2 g/l), respectively. The protein was purified by Ni2+ affinity chromatography using Ni-NTA (nickel-nitrilotriacetic acid) resin (Qiagen) followed by size-exclusion chromatography on a Superdex 75 10/300 (GE Healthcare Life Sciences).
Single crystals suitable for diffraction experiments were obtained by the sitting drop vapour diffusion method, starting from condition # 44 of the Index screen (Hampton Research): protein solution was mixed with equal volume of reservoir solution containing HEPES 100 mM (pH 7.5) and PEG 3350 25%. The protein solution was 14 mg/ml in Tris–HCl 20 mM (pH 7.3), DTT 5 mM and NaN3 1 mM. Optimal conditions were obtained by vapour diffusion in hanging drops at 293 K, using a reservoir containing HEPES 100 mM (pH 7.56), PEG 3350 27% and NaN3 3 mM. Co-crystallization was carried out using protein, di-C4-PtdIns(4)P (phosphatidylinositol-4-phosphate) or di-C4-PtdIns(4,5)P2 mixtures, with ratios of 1/1.2 up to 1/6, and after 2–29 days of incubation time. Crystals appeared in the same conditions as for the protein alone.
Data collection, structure solution and refinement
Low-temperature X-ray diffraction data extending to 2.19 Å were collected at a wavelength of 0.954 Å, on beamline ID23-2 at the ESRF (Grenoble, France). The diffraction data were processed and scaled using XDS . The crystal belongs to the hexagonal system with space group P6n. The unit cell volume indicates that the asymmetric unit is composed of one protein chain, with a solvent content of 46% and a Matthews coefficient of 2.28 Å3/Da. The structure was solved by molecular replacement using Phaser . Protein p50-RhoGAP (human Ras-related homology GTPase-activating protein; pdb entry 1rgp) was used as a search model. Its amino acid sequence exhibits 22.8% identity with that of yeast Rgd1-RhoGAP. Side chain atoms and external loops of the p50-RhoGAP model were discarded. The search model contained only 116 residues. A satisfying solution was found in the space group P65. Further construction and refinement of the model were carried out using Coot , Refmac5  and PDB-REDO , with the stereochemical restraints as defined in Refmac5. Structure was validated using Molprobity . Final statistics of data collection and refinement are given in Supplementary Table S1. The refined model contains 206 residues out of 227 of the expressed protein and 52 water molecules. Parts of the protein chain could not be modelled and correspond to segment 1–18. Intermolecular contacts exclusively involve hydrogen bonds between neighbouring residues or via water molecules.
Co-ordinates have been deposited to the RCSB Protein Data Bank with the accessing code 5my3. The PyMOL molecular-graphics system  was used to prepare the figures presented herein.
Liquid-state NMR spectroscopy
HSQC (heteronuclear single quantum coherence transfer) spectra were carried out at 313 K on a Bruker Avance III 800 MHz spectrometer equipped with a 5 mm TCI 1H/13C/15N/2H cryoprobe. The NMR data were processed using the TOPSPIN 3.2 software and analyzed using the CCPNMR Analysis program package . The backbone assignment data were obtained by a standard combination of three-dimensional NMR experiments, such as HNCO/HNcaCO, HNCA/HNcoCA, and HNCACB/CBCAcoNH. Chemical shift values have been deposited to the Biological Magnetic Resonance Bank with the accession code 19589.
Titration experiments were performed at 313 K using 15N-labelled Rgd1-RhoGAP domain in NMR buffer [20 mM Tris–HCl and 5 mM DTT (pH 7.3)] incubated with different amounts of di-C8 derivatives of the d-myo-inositol forms of PtdIns(4)P, PtdIns(4,5)P2 and PtdIns(3,4,5)P3 (phosphatidylinositol-3,4,5-trisphosphate; Avanti Polar Lipids, Inc.). Protein concentration was measured by UV absorbance at 280 nm and fixed at 150 μM. The PtdInsP/protein ratio varied from 0 to 3.3. Chemical shift perturbations were determined using the normalized chemical shift formula Δδ = [(ΔδH)2 + (0.17 × ΔδN)2]1/2 .
Docking was carried out with AutoDock Tool 1.5.6  in accordance with D'Avanzo et al.  whereby all PtdInsP ligands were restricted to only the head groups to best mimic the PtdInsPs. PDB files for the ligands PtdIns(4)P, PtdIns(4,5)P2 and PtdIns(3,4,5)P3 were generated by modifying files taken from the PDB using ligand codes PIF, 3PT and IP9, respectively, and incorporating partial charges as defined by Lupyan et al. . Ligands had 9, 10 and 11 rotatable bonds, respectively. The 5MY3 PDB file was used as the macromolecule for analysis. Residues with missing side chain atoms were completed using Coot  and its rotamer library. Docking was initialized from a grid box of 126 × 126 × 126 with 0.275 Å spacing, centred on the core NMR-CSP (nuclear magnetic resonance chemical shift perturbation) region using AutoGrid 4.2. Docking calculations were carried out for each ligand with 100 individual simulations using a Lamarckian genetic algorithm  with default parameters and 25 000 000 maximum evaluations with AutoDock 4.2 . Initial blind-docking was also carried out in a similar fashion, using a grid spacing of ∼0.4 Å to calculate 100 poses with 2 500 000 maximum evaluations; however, similar binding was observed. All calculations were carried out on a 2012 MacBook Pro with a 2.9 GHz core i7 processor and 8 GB RAM.
Sample preparation for solid-State NMR spectroscopy
To prepare multilamellar vesicles (MLVs), POPC (1-palmitoyl-2-oleoylphosphatidylcholine), POPS (1-palmitoyl-2-oleoylphosphatidylserine), POPE (1-palmitoyl-2-oleoylphosphatidylethanolamine), SoyPI and ergosterol powders were mixed in organic solvent (chloroform : methanol at 2 : 1 ratio). Solvent was evaporated under a N2 airflow to obtain a thin film. Lipids were rehydrated with ultrapure water before lyophilization. The powder obtained was hydrated with an appropriate amount of deuterium-depleted water (92% hydration ratio) and homogenized by three cycles of shaking in a vortex mixer, freezing (liquid nitrogen, 1 min) and thawing (40°C in a water bath, 10 min). Protein was suspended in a buffer composed of Tris–HCl 20 mM at pH 7.4, to a lipid/protein molar ratio of 100. Final MLV composition in moles was as follows: POPC-d31 (13%), ergosterol (37%), POPE (24%), POPS (8%) and SoyPI (18%) with or without 2% PtdIns(4,5)P2.
Solid-state NMR spectroscopy
2H NMR experiments were carried out using a Bruker Avance III 800 MHz spectrometer equipped with a Dual Resonance 1H/15N-31P Broad Band 4 mm DVT CPMAS probe. 2H NMR experiments on POPC-d31 were performed with a phase-cycled quadrupolar echo pulse sequence (90°x–τ–90°y–τ–acq). Acquisition parameters were as follows: spectral window of 500 kHz, π/2 pulse width set to 4.5 μs, interpulse delays τ of 40 μs and a recycle delay of 2 s. 4k scans were used for 2H NMR data. The spectra were processed using a Lorentzian line broadening of 400 Hz for 2H spectra before the Fourier transform from the top of the echo. Samples were equilibrated 30 min at a given temperature before data acquisition. All the spectra were processed and analyzed using the Bruker TOPSIN 3.2 software.
De-Pake-ing and simulation of deuterium spectra
A mathematical deconvolution using the NMR dePaker 1.0rc1 software was performed on experimental deuterium data to obtain oriented-like spectra. Experimental quadrupolar splittings were extracted from these deconvoluted spectra. In-house simulation software was used to calculate the quadrupolar splitting values (ΔíBQ) all along the lipid acyl chain. They are directly proportional to the order parameter of the C-D bond (SCD) using the following equation: Δí, where KB is a coupling constant of 125.25 kHz for bilayer geometry.
The structure of Rgd1-RhoGAP was successfully solved by molecular replacement using the atomic co-ordinates of the human p50-RhoGAP structure as a search model (PDB code 1rgp) , despite a rather low sequence identity (22.8%) between RhoGAP domains. This is most likely due to the high conservation of secondary and tertiary structure elements in this protein family. The final refined model consists of 206 out of 227 residues. Secondary structure matching (SSM) superimposition  between the RhoGAP model and the p50 structure yields an average rms deviation of 1.56 Å over the 177 superimposed Cα positions.
The Rgd1-RhoGAP domain adopts a typical RhoGAP fold (see Figure 2A and topology in Supplementary Figure S1) [30,32,33] composed exclusively of helices and unstructured connecting loops: a four α-helix bundle (A, B, E and F) makes up a highly conserved hydrophobic domain core. The root mean square deviation (r.m.s.d.) of Cα between this domain and the corresponding one from p50 yields a value of 0.82 Å. As shown in Figure 2B, when superimposing the present structure with that of p50, the largest deviations occur in the periphery of the central four helix bundle. This bundle is capped on one end by a short α-helix (A0) and by connecting loops on the other end. Five other helices are positioned against two of the bundle sides: helix G lies perpendicular to the bundle axis and is stabilized across its E-F side by two peripheral α-helices C and D. On the opposite side, two helices A1 and A1′ stack against side A-B. Whereas the α-helix A1 is conserved in all of the known structures, an 8 aa long helix, labelled A1′, is unique to the yeast Rgd1-RhoGAP and singularly exhibits a 310 turn. This secondary structure element inserts into the longest connecting loop observed between helices A1 and B among six structures (Figure 1). In the present structure, the α-helices have similar lengths compared with those observed in the other structures, with the exception of helix A0 that is one of the shortest of the six structures and helix B that exhibits a supplementary half turn on its N-terminus. Supplementary Figure S2 presents the B factors of the crystal structure. We notice high values for loops between helices, and more specifically between helices A1′ and B.
X-ray structure of the Rgd1-RhoGAP domain.
As a general trend, the ternary structure is almost exclusively stabilized via interhelical hydrophobic interactions, with the exception of hydrogen bonds between side chains of polar residues present on the protein surface, at the helix ends and along the connecting loops.
As for Myo9b-RhoGAP (Rho GTPase-activating protein Myo9b), the protein contains two arginine fingers at its catalytic sites. The first arginine finger (R60) resembles the one within the canonical RhoGAP domain. It is critical for the GAP activity in trans to the catalytic core of Rho proteins in order to participate in the GTP hydrolysis reaction [34,35]. The second arginine finger (R172) anchors the switch I loop of Rho proteins and interacts with the nucleotide, stabilizing the transition state of GTP hydrolysis .
In the present structure, the first arginine finger loop (residues G52–N64) adopts a conformation similar to those observed in p50-RhoGAP and Myo9b-RhoGAP, and is stabilized by comparable interactions. One may notice that, after SSM superimposition of the yeast Rgd1-RhoGAP and human p50-RhoGAP structures (see Supplementary Figure S3A for details), the r.m.s.d. between Cα positions of the first arginine finger loop is 0.77 Å, a value very close to the one mentioned for the embedded four α-helix bundle (A, B, E and F). Similarly, a value of 0.68 Å is obtained when superimposing the Myo9b-RhoGAP structure onto that of Rgd1-RhoGAP. Thus, anchoring of the first arginine finger loop leads to a rather well-conserved conformation among the various structures. As far as the R60 side chain conformation is concerned, no conclusion could be drawn since it happens that this side chain is involved in crystal packing contacts. In the structures of human p50-RhoGAP, Myo9b-RhoGAP and RICS (RhoGAP involved in the catenine-N-cadherin and NMDA receptor signalling), the corresponding arginine side chains are also involved in crystal contacts. Thus, the conformation observed here appears not to be significant with respect to the active role of this residue, adding the fact that this latter adopts a different conformation in the structure of the p50–RhoGAP–RhoA complex (pdb code 1tx4) .
Concerning the second arginine finger (R172), a superimposition of Myo9b-RhoGAP onto Rgd1-RhoGAP (see Supplementary Figure S3B) shows that the corresponding loops adopt a similar conformation with a comparable r.m.s.d. (0.91 Å) between Cα positions. Here, arginine side chains adopt an extended conformation pointing towards the solvent.
So far, co-crystallization of the Rgd1-RhoGAP domain with either PtdIns(4)P or PtdIns(4,5)P2 leads to crystals which diffracted to comparable resolution, but no electron density corresponding to PtdInsP molecules was observed. Moreover, the corresponding refined structures lead to very close protein models, with r.m.s.d. on Cα lower than 0.34 Å, i.e. no protein conformational change could be observed.
Assignment of backbone resonances and secondary structure of Rgd1-RhoGAP domain
Analysis of the experiments allowed the sequence-specific assignment for 187 out of the 204 (226 less 15 Pro, the first Met residue and the 6xHis tag) backbone 15N and amide proton resonances (92%). The assignment is nearly completed with the exception of H4, Q34, H91, H160, L161, K162, R176, E216 and 15N chemical shifts of 15 proline residues. The catalytic loop («arginine finger», residues 58–65) has not been assigned because peaks were missing, even if the temperature was lowered to 278 K. Cα and Cβ chemical shift values allowed us to assess the secondary structure of the protein in solution. We identified 10 helices in agreement with the structure solved by X-ray diffraction.
Identification of a phosphoinositide-binding site on the Rgd1-RhoGAP domain
Using size-exclusion chromatography and circular dichroism, we previously showed that the Rgd1-RhoGAP domain could bind lipids with a moderate affinity: half maximal effective concentration of PtdIns(4,5)P2 (EC50) is estimated to 30 μM . We provided the first evidence that this domain selectively bound PtdInsP, whereas it bound neither PC (phosphatidylcholine) nor PS (phosphatidylserine). Phosphoinositides were previously shown to enhance Rho4p GTPase activity in the presence of the full Rgd1 protein or simply the isolated domain . Since we were not able to obtain crystals of Rgd1p–RhoGAP/PtdInsP in complex, we decided to explore the interaction with NMR spectroscopy in order to determine an eventual phosphoinositide-binding site. A series of 1H15N-HSQC spectra of the protein were acquired in the presence of increasing PtdInsP concentration. Experiments were carried out with short chain di-C8 lipids (Supplementary Figure S4), as di-C16 lipids used for size-exclusion chromatography led to a massive line broadening of the protein peaks. NMR titration experiments were performed with three different phosphoinositides, chosen for their physico-chemical properties and their biological relevance. PtdIns(4)P and PtdIns(4,5)P2 were chosen to mimic the RhoGAP interaction within secretory vesicles and plasma membrane, respectively. PtdIns(3,4,5)P3, though absent in the yeast S. cerevisiae, was used to check the potential link between PtdIns phosphorylation state versus binding. Figure 3 shows the chemical shift perturbations for the RhoGAP domain with the three different PtdInsPs at 500 μM. The chemical shift perturbations are displayed along the protein sequence (Figure 3B) and mapped on the X-ray structure (Figure 3C). The conservation of the global identity of the HSQCs between 0 and 500 μM PtdInsP means that the protein structure remains conserved through the titration. A limited number of peaks are significantly affected by phosphoinositide binding (Figure 3A). For example, several residues experience a significant chemical shift perturbation upon PtdIns(4,5)P2 binding: G14, K17 (side chain), F19, Q43 (side chain), K70, L71, K72, S83, L86, S88, I96 and Y97 (Figure 3B—2). All these residues could be mapped on the Rgd1-RhoGAP surface (Figure 3C), except for G14 and K17 absent in the crystal structure. The affected residues belong to the same structural region, constituted by the flexible N-terminal loop, the helices A1, A1′, B and their interconnecting loops. The comparison of data between the three different phosphoinositides allowed us to identify a common binding site (Figure 4), composed of the side chains of K17 and Q43 residues, as well as the backbone of L71, S83, L86 and S88 residues. The main observed difference between the three phosphoinositides corresponds to an extent of perturbations related to the number of phosphate groups. One may notice that, for most of these residues, the greater the number of phosphates, the more significant the chemical shift differences.
Chemical shift perturbation and mapping of the phosphoinositide-binding site on the Rgd1-RhoGAP structure.
Determination of a common PtdInsP-binding site for Rgd1-RhoGAP.
The side chains of residues K17 and Q43, as well as the backbone of L86 at the end of helix A1′, experience the most significant chemical shift perturbations. It is interesting to note that these side chains are involved in the interaction with the lipid headgroup IP3 (inositol tri-phosphate).
The majority of the residues involved in the lipid binding are localized in two non-conserved regions of the RhoGAP domain family: the A1′ helix (77–88) and the N-terminal loop (Figure 1). Extended sites are also observed on the two neighbouring helices A1 and B, depending on the nature of the lipid head groups. Using Blast and clustal Omega softwares, we showed that the presence of an additional A1′ helix, compared with the canonical model , is restricted to the fungi kingdom (data not shown). This extra helix was never detected in srGAPs (Slit-Robo GAP) from mammals which share strong similarities in domain organization with the yeast Rgd1p [37,38,39].
In silico docking
In silico docking was performed with the program AutoDock for the three PtdInsP ligands in the presence of the Rdg1-RhoGAP protein, namely PtdIns(4)P, PtdIns(4,5)P2 and PtdIns(3,4,5)P3. All PtdInsP ligands were restricted to their respective head groups (truncated at the C1 position), as previously published , to best incorporate ligand flexibility into the calculation. Supplementary Figure S5 corresponds to a conformational ensemble of 100 complexes of the Rgd1-RhoGAP protein bound to the PtdIns(4,5)P2. Of the 100 docked poses, 87 were located around the K89 side chain positioned in the core lipid-binding site as observed by NMR-CSPs (Figure 4). The remaining conformations show potential interactions with the side chain of lysine's 20, 70 or 77; however, these may not be as significant due to minor perturbations observed by NMR. Similar results were obtained with PtdIns(4)P and PtdIns(3,4,5)P3. In general, protein–lipid binding was shown to be stabilized by a network of hydrogen bonds between the side chain amine protons of a lysine and the phosphate moieties at positions 4 and 5 of the lipid headgroup. In particular, the core lipid-binding site around K89 also showed a close spatial proximity of the lipid phosphate moiety with residues S83, L86 and S88 of helix A1′ and residue Q43 of helix A (Figure 5). This conformation would then suggest stabilization by Van der Waals interactions between the lipid acyl chains and other Rgd1-RhoGAP side chains; a potential explanation for the more minor NMR-CSPs observed for adjacent residues. It is interesting to note that the amide protons of the K89 side chain were beyond detection using 1H15N-HSQC NMR spectra acquired at 313 K, probably due to the fast exchange with water at high temperature.
Docking of PtdIns(4,5)P2 on Rgd1-RhoGAP surface.
Effect of Rgd1-RhoGAP domain on liposome dynamics
We used solid-state NMR spectroscopy to explore the effect of the Rgd1-RhoGAP domain on fluid-phase MLV lipid dynamics in the presence or absence of phosphoinositides. NMR experiments were carried out with deuterium-labelled POPC-d31 (13%) in the presence of PE/PI/PS/ergosterol (24/18/8/37). This composition was chosen to mimic the yeast plasma membrane, as it has been previously shown [40,41]. We chose to explore the selectivity of the protein–lipid interaction by experimentation in the presence of 2% PtdIns(4,5)P2. We first followed the spectral moment (M1) at different temperatures (Figure 6A). M1 reflects lipid dynamics as it is proportional to the average quadrupolar splitting and gives a measure of the width of all spectrum components. When the temperature is increased, the M1 value decreases as the lipids become more mobile. Within the accuracy of the measurement, we could not observe any differences in membrane dynamics either in the presence or absence of the Rgd1-RhoGAP domain or PtdIns(4,5)P2. However, when PtdIns(4,5)P2 was added at a cellular concentration (2%), a decrease in the spectral moment was observed upon the addition of protein (Figure 6A). This therefore suggests that the RhoGAP domain has a disordering effect on the membrane in the presence of PtdIns(4,5)P2. Simulations and De-Pake-ing of experimental spectra were carried out to accurately measure individual quadrupolar splittings for POPC-d31 in fluid phase. Order parameters |SCD| are determined for each C-D bond along the lipid acyl chain using the simulated quadrupolar splitting values (Figure 6B). The order parameter profile of POPC-d31 is characterized by a plateau region corresponding to the rigid positions close to the glycerol backbone (2–8). It is then followed by a rapid decrease in the SCD to the terminal methyl moiety that exhibits rapid motions in the middle of the membrane. We determined the order parameter profile for four different systems (Figure 6B): MLV alone, MLV + RhoGAP, MLV + PtdInsP and MLV + PtdInsP + RhoGAP. No differences in the order parameter profile were detected between the first three systems. As previously described, the RhoGAP domain did not induce any perturbation on the membrane dynamics. There was also no effect when 2% of PtdIns(4,5)P2 was added to the membrane composition. However, in the presence of both phosphoinositide and the Rgd1-RhoGAP domain, we observed a disordering effect on the first positions of the lipid acyl chains. This result shows that there is a specific interaction between membrane-containing PtdInsP and the protein. Moreover, the RhoGAP domain was not deeply inserted in the membrane as the interaction affects only the first carbon positions of the acyl chain.
Lipid dynamics of phosphoinositide-enriched liposomes in the presence of RhoGAP.
Rgd1p is a GAP protein specific for the Rho3 and Rho4 small GTPases which control actin cytoskeleton organization and stress signalling pathways [8,42,43]. This protein shows a specific domain organization (F-BAR and RhoGAP), which is also found among the members of the mammalian srGAP family involved in axonal growth . The present paper identifies the fold and characterizes the lipid-binding properties of the RhoGAP domain (aa 486–666) from Rgd1p. It first presents the three-dimensional structure of the domain characterized by a double approach using X-ray crystallography and NMR spectroscopy. The protein displays both in crystal and in solution the typical RhoGAP fold, composed of helices A0, A, A1, B, C, D, E, F and G. As with Myo9b-RhoGAP, the protein also possesses two arginine fingers at its catalytic site, critical for the GAP activity on Rho proteins for GTP hydrolysis stimulation (Figure 1). Interestingly, an 8 aa long helix, labelled A1′, between helices A1 and B, is shown to be unique to Rgd1p–RhoGAP. Interestingly, the Myo9b-RhoGAP displays a single helical turn in this region . NMR titration studies show that the region A1–A1′–B constitutes a new phosphoinositide-binding site, a result unexpected from the protein sequence alone. The lipid-binding site involves both polar and non-polar residues (K17, Q43, L71, S83, L86 and S88), suggesting the presence of both electrostatic and hydrophobic interactions. In silico docking with PtdInsPs is in full agreement with the experimental NMR results and suggests the presence of a network of hydrogen bonding between the phosphate moieties of the lipid head group and mainly the side chain amine protons of K89. Our results explain at the residue level the specific binding between phosphoinositides and the Rgd1-RhoGAP domain previously observed by gel filtration and circular dichroism . The presence of a lipid-binding site in the yeast RhoGAP domain could explain the modulation of its GAP activity by phosphoinositides observed on Rho4p, but not on Rho3p . A similar situation was previously described for the human p190A. A PBR adjacent to its RhoGAP domain has been shown to bind negatively charged phospholipids. Such interaction is crucial for the RhoGAP activity of p190A with RhoA [45,46]. The presence of PBR is observed in flexible regions of some other RhoGAP domains, such as cdGAP and human DLC1 (deleted in liver cancer 1). This last protein, linked to breast cancer metastatic potential, is known to selectively bind PtdIns(4,5)P2 . Other lipid-binding motifs can be found at the N-terminal side of RhoGAP domains, such as PH, C1, C2 or Sec14 domains [7,47,48]. In the case of the Rgd1p-RhoGAP protein, the phosphoinositide-binding site is not in a flexible region, but is included within the domain itself. Other regions of Rgd1p bind phosphoinositides and increase RhoGAP activity: the F-BAR domain (aa 1–300) and the internal region (aa 301–350) . Previous works showed that the F-BAR domain from Rgd1p binds PtdIns(4,5)P2 and is essential for its in cell distribution. Moreover, a phosphoinositide-binding site for F-BAR was revealed by a crystallographic approach . Some previous papers indicate that regulation of some RhoGAP proteins is mediated by either phosphorylation (p190 RhoGAP) or lipid binding (n-chimaerin) . Phosphoinositides as PtdIns(4,5)P2 regulate the specificity of p190A RhoGAP in switching its GTPase substrate: in the presence of PtdIns(4,5)P2, the GAP activity was decreased for RhoA, and in the contrary, the GAP activity was increased for Rac1 . Similarly, in our case, phosphoinositides stimulate RhoGAP only for Rho4p GTPase and not for Rho3p . Rho3p and Rho4p are involved during bud growth and cytokinesis establishment, respectively. We hypothesize that phosphoinositides could act as a regulator of Rgd1-RhoGAP activity, according to the cell cycle.
In summary, Rgd1p and Rho proteins bind the membrane through multiple phosphoinositide-binding sites and a geranylgeranyl anchor, respectively (Figure 7). Weak but specific lipid–protein and protein–protein interactions may both be necessary to keep the protein partners close to the plasma membrane during cell growth and division for the acquisition of cell polarity.
Model of Rgd1p/Rho4p complex bound to the plasma membrane.
Rho GTPase-activating protein 11A
F-BAR FCH, and BAR
Fes, CIP4 homology
- G protein
guanine nucleotide-binding proteins
heteronuclear single quantum coherence transfer
Rho GTPase-activating protein Myo9b
nuclear magnetic resonance chemical shift perturbation
human Ras-related homology GTPase-activating protein
RhoGAP domain from related GAP domain 1 protein
Rho GTPase-activating proteins
secondary structure matching
B.O. developed and led the project. D.M. performed and analyzed NMR experiments. B.L.E. crystallized the protein and collected X-ray diffraction data. D.M., T.G. and B.G. solved the X-ray structure. J.T. performed the in silico docking of PtdIns on the crystal structure of RhoGAP domain. Solid-state NMR experiments were acquired by B.O. and C.C. B.O. and F.D. wrote the paper with input from D.M., T.G., B.L.E. and M.H..
Financial support for the TGIR-RMN-THC Fr3050 CNRS is gratefully acknowledged. D.M. was supported by a French PhD fellowship afforded to the University of Bordeaux by the Ministère de la Recherche (MNERT).
We thank the ESRF for provision of synchrotron radiation facilities, and the staff of beamline ID23-2 for their kind assistance. We also thank the structural biology platform at the Institut Européen de Chimie et Biologie [UMS 3033] for access to NMR spectrometers and technical assistance.
The Authors declare that there are no competing interests associated with the manuscript.
Deceased; this article is dedicated to the memory of Bernard Gallois.