Structural and Mechanistic Insights into the Regulation of the Fundamental Rho Regulator RhoGDIα by Lysine Acetylation*

Rho proteins are small GTP/GDP-binding proteins primarily involved in cytoskeleton regulation. Their GTP/GDP cycle is often tightly connected to a membrane/cytosol cycle regulated by the Rho guanine nucleotide dissociation inhibitor α (RhoGDIα). RhoGDIα has been regarded as a housekeeping regulator essential to control homeostasis of Rho proteins. Recent proteomic screens showed that RhoGDIα is extensively lysine-acetylated. Here, we present the first comprehensive structural and mechanistic study to show how RhoGDIα function is regulated by lysine acetylation. We discover that lysine acetylation impairs Rho protein binding and increases guanine nucleotide exchange factor-catalyzed nucleotide exchange on RhoA, these two functions being prerequisites to constitute a bona fide GDI displacement factor. RhoGDIα acetylation interferes with Rho signaling, resulting in alteration of cellular filamentous actin. Finally, we discover that RhoGDIα is endogenously acetylated in mammalian cells, and we identify CBP, p300, and pCAF as RhoGDIα-acetyltransferases and Sirt2 and HDAC6 as specific deacetylases, showing the biological significance of this post-translational modification.

Rho proteins are small GTP/GDP-binding proteins primarily involved in cytoskeleton regulation. Their GTP/GDP cycle is often tightly connected to a membrane/cytosol cycle regulated by the Rho guanine nucleotide dissociation inhibitor ␣ (RhoGDI␣). RhoGDI␣ has been regarded as a housekeeping regulator essential to control homeostasis of Rho proteins. Recent proteomic screens showed that RhoGDI␣ is extensively lysine-acetylated. Here, we present the first comprehensive structural and mechanistic study to show how RhoGDI␣ function is regulated by lysine acetylation. We discover that lysine acetylation impairs Rho protein binding and increases guanine nucleotide exchange factor-catalyzed nucleotide exchange on RhoA, these two functions being prerequisites to constitute a bona fide GDI displacement factor. RhoGDI␣ acetylation interferes with Rho signaling, resulting in alteration of cellular filamentous actin. Finally, we discover that RhoGDI␣ is endogenously acetylated in mammalian cells, and we identify CBP, p300, and pCAF as RhoGDI␣-acetyltransferases and Sirt2 and HDAC6 as specific deacetylases, showing the biological significance of this post-translational modification.
Rho proteins are guanine nucleotide-binding proteins (GNBPs) 3 predominantly regulating the actin and microtubule cytoskeleton (1,2). They are molecular switches and cycle between a GDP-bound inactive and a GTP-bound active conformation. In the GTP-bound state, Rho proteins bind to effector proteins regulating essential cellular processes: maintenance of cell architecture, intracellular transport, cell migration, cell movement, cytokinesis, and signal transduction. Rho protein dysfunction results in severe cellular disorders, such as neurodegenerative diseases, metastasis, and tumor invasion (3).
Rho proteins show a low intrinsic GTP hydrolysis and nucleotide exchange rate, which is strongly accelerated by Rho-GTPase-activating proteins and Rho guanine nucleotide exchange factors, respectively (4). In the GTP-bound state, they are mostly bound to the plasma membrane via a polybasic region and a prenyl group (farnesyl or geranylgeranyl) forming a thioether with the C-terminal CaaX box cysteine side chain. About 80 different RhoGTPase-activating proteins and 80 Rho guanine nucleotide exchange factors have been described in humans to date (5,6).
Another key regulator of Rho function is the Rho guanine nucleotide dissociation inhibitor (RhoGDI) that couples the GTP/GDP cycle to a membrane/cytosol cycle. Only three RhoGDIs have been found in mammals. RhoGDI␣ is ubiquitously expressed, RhoGDI␤ is mainly expressed in hematopoietic cells, and RhoGDI␥ is present in the brain, lung, kidney, and testis (7). This led to the hypothesis that RhoGDIs are housekeeping regulators of Rho proteins. However, recently, it has been found that RhoGDIs play more complex roles than originally expected. They are highly regulated by phosphorylation, can bind cytosolic GDP-and GTP-loaded Rho guanine nucleotide-binding protein (RhoGNBPs), are capable of transporting Rho proteins specifically to different cellular membranes, and regulate their turnover (8 -10).
The interaction of Rho proteins and GDIs has been studied functionally and structurally. The crystal structures of full-length RhoGDI␣ alone and in complex with RhoA, Cdc42, and Rac1 have been solved by NMR and by x-ray crystallography (11)(12)(13)(14). These studies revealed a modular structure of RhoGDI␣, a C-terminal immunoglobulin (Ig) domain forming a hydrophobic pocket accommodating the prenyl group of the RhoGNBPs and an N-terminal intrinsically unfolded domain. This domain adopts a helix-turn-helix con-formation upon binding the lipidated RhoGNBP contacting the switch I and II regions essential for effector binding.
For membrane extraction and membrane relocation of RhoGNBPs by RhoGDI␣, a two-step reaction mechanism has been postulated, supported by its modular structure (7). In the first step of delivery, positively charged patches in the Ig domain of RhoGDI␣ are electrostatically attracted to the negatively charged membrane phospholipids. In the second step, the RhoGNBP inserts its lipid moiety into the membrane. An electrostatic network encompassing the negatively charged RhoGDI␣ N terminus competes with the membrane phospholipids for binding to the positively charged C terminus of the RhoGNBP (polybasic region) (15). It is still unclear how the tight Rho⅐GDP⅐RhoGDI␣ complexes are dissociated for RhoGNBPs to be reactivated by GEF-catalyzed GTP loading (14,16).
Recently, RhoGDI␣ has been found to be SUMOylated at Lys-138, leading to a stabilization of RhoA⅐RhoGDI␣, resulting in decreased cancer cell motility (25). Several RhoGDI␣ lysine acetylation sites have been found in various quantitative proteomic screens performed in diverse cell and tissue types (20, 21, 26 -28). For one site, RhoGDI␣ Lys-141, it has been shown by site-directed mutagenesis (K141Q as an acetylation mimic) that it leads to formation of thickened actin stress fibers and filopodia in HeLa cells (21).
Functionally, the acetylation sites in RhoGDI␣ identified by quantitative mass spectrometry have only marginally been characterized so far. Here we present the first comprehensive study using a combined synthetic biological, biophysical, and cell biological approach to unravel how lysine acetylation regulates RhoGDI␣ function. Our results reveal general mechanisms of how lysine acetylation regulates protein function and might open up new therapeutic strategies.

Experimental Procedures
Expression and Purification of Proteins-All proteins were expressed as GST fusions using the pGEX-4T5/Tev vector derived from pGEX-4T1 (GE Healthcare) or as His 6 -tagged fusion proteins (pRSF-Duet; Merck) in Escherichia coli BL21 (DE3) cells. For protein expressions, cells were grown to an A 600 of 0.6 (37°C; 160 rpm). The expression was induced by the addition of 100 -300 M isopropyl-␤-D-thiogalactopyranoside and done overnight (18 -20°C; 160 rpm). Afterward, the cells were harvested (4000 ϫ g, 20 min) and resuspended in buffer A (50 mM Tris/HCl, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 , 2 mM ␤-mercaptoethanol) containing 200 M Pefabloc protease inhibitor mixture. After cell lysis by sonication, the soluble fraction (20,000 ϫ g, 45 min) was applied to the equilibrated affinity chromatography column. Washing was done with buffer B for the GSH column (50 mM Tris/HCl, pH 7.4, 300 mM NaCl, 5 mM MgCl 2 , 2 mM ␤-mercaptoethanol) and buffer C (buffer B plus 10 mM imidazole) for the Ni 2ϩ -NTA column. Tev protease cleavage was performed when suitable on the column or in batch overnight at 4°C. After Tev cleavage, the protein was concentrated by ultrafiltration and applied to a size exclusion chromatography column (GE Healthcare). The concentrated fractions were shock-frozen in liquid nitrogen and stored at Ϫ80°C. Protein concentrations were determined measuring the absorption at 280 nm using the protein's extinction coefficient. For nucleotide-bound RhoGNBPs, the concentration was determined by a Bradford assay (Expedeon).
Incorporation of N-(⑀)-Acetyl-L-lysine-The site-specific incorporation of N-(⑀)-acetyl-L-lysine was done by the addition of 10 mM N-(⑀)-acetyl-L-lysine (Bachem) and 20 mM nicotinamide to inhibit the E. coli CobB deacetylase to the E. coli BL21 (DE3) culture at an A 600 of 0.6 (37°C). Cells were grown for an additional 30 min before protein expression was induced by the addition of 100 -300 M isopropyl-␤-D-thiogalactopyranoside. Acetylated RhoGDI␣ was expressed from a pRSF-Duet-vector also encoding the synthetically evolved Methanosarcina barkeri tRNA CUA and the acetyl-L-lysyl-tRNA-synthetase as described previously (29). The incorporation of acetyl-L-lysine in E. coli is done cotranslationally as a response to an amber stop codon.
In Vitro Farnesylation-Geranylgeranylated proteins are prone to aggregation and are poorly soluble at micromolar concentrations needed for biophysical studies. Therefore, we used an in vitro farnesylation approach. Purified RhoA L193A/ Cdc42 L191A was enzymatically farnesylated by recombinantly expressed and purified human farnesyltransferase using farnesylpyrophosphate as substrate. The in vitro farnesylation was done in 1 ml of buffer containing 100 mM NaCl, 50 mM Tris/ HCl, pH 7.4, 5 mM MgCl 2 , 2 mM tris(2-carboxyethyl)phosphine, and 10 M ZnCl 2 by incubating 200 M protein with a 1.5-fold molar excess of farnesylpyrophosphate (Jena Bioscience) and 6 M farnesyltransferase (1 h at 30°C, 1 h on ice). Finally, farnesylated proteins were purified by size exclusion chromatography (Superdex 75 10/300, GE Healthcare).
Fluorescence Measurements of GEF-catalyzed Nucleotide Dissociation-For nucleotide exchange reactions, RhoA-F was loaded with mantGDP by incubating the protein with a 10-fold excess of fluorescently labeled nucleotide in the presence of 10 mM EDTA. Redundant nucleotide was removed by size exclusion chromatography, and loading of RhoA-F was checked by HPLC. Nucleotide exchange reactions were done at 25°C using a PerkinElmer Life Sciences LS55 spectrofluorimeter. All measurements were performed in standard buffer A containing a 50-fold molecular excess (final concentration 50 M) of unlabeled GDP. After 1:1 complexes of RhoGDI␣ and RhoA-F⅐mantGDP (final concentration 1 M) had been formed, the reaction was started by adding 500 nM mouse Dbs-GEF (PH (pleckstrin homology) domain-DH (dibble homology) domain; aa 624 -960). Nucleotide exchange reactions were followed by fluorescence quenching as a function of time.
In Vitro SUMOylation Assay-For the in vitro SUMOylation assay, recombinantly expressed and purified proteins/enzymes were used. The reactions were performed in a buffer containing 50 mM Tris/HCl, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 , 2 mM DTT, 1 mM PMSF, and 5 mM ATP. 100 ng/l RhoGDI␣ was mixed with 3 ng/l E1 (human Aos1/Uba2; both full-length), 3 ng/l E2 (full-length human Ubc9), and 300 ng/l human SUMO1. The reactions were incubated overnight at 30°C and terminated by adding SDS sample buffer. Proteins were analyzed by SDS-PAGE and IB.
Immunoprecipitation, Pull-down, and Immunoblotting-For immunoprecipitation of acetylated proteins, cells were sonicated in lysis buffer (10 mM Tris/HCl, pH 7.4, 150 mM NaCl, 2 mM EDTA, 1% (v/v) Triton X-100, and protease inhibitor mixture from Sigma). Lysates were incubated with anti-acetyl-Llysine-agarose beads (ImmuneChem) at 4°C overnight. The beads were washed three times in lysis buffer, and acetylated proteins were eluted by incubating the beads in elution buffer (50 mM Tris/HCl, pH 7.4, 150 mM NaCl, 0.1% (w/v) SDS, 1% (v/v) Triton X-100, 6 M urea) for 20 min at room temperature. For analysis of His 6 -tagged proteins, 12 h after transfection, cells were harvested and lysed in binding buffer (10 mM Tris/ HCl, pH 8.0, 100 mM NaH 2 PO 4 , 300 mM NaCl, 2 mM ␤-mercaptoethanol, 0.05% (v/v) Tween 20, 8 M urea, and 10 mM imidazole). Lysates were incubated with Ni 2ϩ -NTA magnetic beads (5 Prime) for 2 h at 4°C with rotation. Subsequently, the beads were washed three times in binding buffer supplemented with 20 mM imidazole, and His 6 -tagged proteins were eluted with binding buffer containing 250 mM imidazole for 20 min at room temperature. Eluates were then resolved by SDS-PAGE and analyzed by immunoblotting performed using a standard protocol. Bound antibodies were visualized by using enhanced chemiluminescence (Roth). Immunoblotting analysis was done by measuring mean gray intensities using ImageJ software. For statistical analyses, a two-tailed Student's t test was applied.
Membrane Extraction Assay-For separating membrane and cytoplasmic fractions, HEK293T cells were lysed in HEPES buffer (25 mM HEPES/NaOH, pH 8.0, 100 mM NaCl, 5 mM MgCl 2 , and protease inhibitor mixture). After preclearing (1000 ϫ g, 10 min, 4°C) lysates were centrifuged at 120,000 ϫ g for 1 h at 4°C. Subsequently, the membrane pellet was washed once and resupended in HEPES buffer. Membrane fractions were then incubated with 250 ng of recombinant RhoGDI␣ with rotation for 1 h at 4°C and again centrifuged at 120,000 ϫ g for 1 h at 4°C. The supernatant was analyzed by immunoblotting.
Cell Culture and Transfection-HEK293T and HeLaB cells were grown in DMEM or minimum essential medium (Life Technologies) supplemented with 10% fetal bovine serum (PAN-Biotech) in the presence of penicillin, streptomycin, and L-glutamine (Life Technologies). The stable cell line HeLa T-REx (pcDNA4/TO/MRGS-His 6 -SUMO1) for inducible His 6 -SUMO1 expression was grown in minimum essential medium containing the same supplements plus 150 g/ml Zeocin and 5 g/ml blasticidin (Life Technologies) (30). Transfections were performed with LipofectamineLTX with PLUS TM reagent (Life Technologies). For acetylation studies, HEK293T cells were grown in DMEM containing 2 M SAHA, 1 M trichostatin A, 10 mM nicotinamide, 10 M sirtinol, and 1 mM sodium butyrate for 6 h.
Preparation of Liposomes and Cosedimentation Assay-The liposomes were prepared in glass vials rinsed with chloroform. Brain extract from bovine brain type I (Folch Fraction I, Sigma B1502) was mixed with chloroform to a final concentration of 1 mg/ml as described. The chloroform was evaporated using a slow argon flow to produce a thin lipid film on the vial. The vial was placed in a dessicator for at least 60 min to dry the film completely. 1 ml of buffer A was added to the vial and incubated for 5 min at room temperature for rehydration while gently agitating. The liposomes were formed during incubation at 40°C for 5 min. The suspension was sonicated (10 min, 40°C), vortexed, and dropped into liquid nitrogen for 2 min. The thaw and freeze steps were repeated five times. Uniformly sized liposomes were formed by extrusion through 0.2-m polycarbonate filter membranes (Whatman catalog no. 800281) at least 21 times. The cosedimentation assay was done as described (31). In a total volume of 50 l, 40 l of freshly prepared liposomes (1 mg/ml) were mixed with the respective concentration of proteins and incubated on ice for 10 min in buffer A. The cosedimentation was done by ultracentrifugation at Ͼ100,000 ϫ g (20 min, 4°C). The obtained pellet and super-natant fractions were analyzed by SDS-PAGE and Coomassie staining or immunoblotting.
Immunofluorescence and Microscopy-HeLaB cells were grown on coverslips and fixed in 3% (w/v) paraformaldehyde for 20 min at room temperature. After permeabilization with 0.5% (v/v) Triton X-100, cells were blocked in PBS containing 5% (w/v) BSA and incubated with fluorescently labeled antibodies for 30 min at room temperature. Coverslips were embedded in ProLong Gold Antifade (Life Technologies) and examined by using an UltraView Vox Spinning Disc confocal microscope (PerkinElmer Life Sciences) collecting Z-stack images. Actin quantifications were performed using ImageJ software. After Z-stack projection to maximum phalloidin intensity, each transfected cell was outlined, and mean intensities of phalloidin fluorescence were measured. Data were normalized by dividing the fluorescence intensity of every single cell by the average of those of non-transfected cells. Student's t test was employed for statistical analysis.
Isothermal Titration Calorimetry (ITC) Measurements-The interactions of purified proteins were thermodynamically characterized by isothermal titration calorimetry on an ITC 200 instrument (GE Healthcare) based on Ref. 61. The measurements were done in buffer A at 25°C at a concentration of 40 -60 M RhoGDI␣ (syringe) and 4 -6 M farnesylated RhoA/ Cdc42 (cell), otherwise as indicated. The heating power per injection was plotted as a function of time until binding was saturated. A one-site binding model was fitted to the data using the MicroCal software. This resulted in the stoichiometry of binding (N), the enthalpy change (⌬H), and the equilibrium association constant (K A ) as direct readout. ⌬G, ⌬S, and the equilibrium dissociation constant (K D ) are derived. We used the standard EDTA-CaCl 2 sample tests as described by Micro-Cal to assess the statistical significance of individual observations. These gave values within the manufacturer's tolerances of Ϯ20% for K A values and Ϯ10% in ⌬H.
Crystallization-The RhoA-F⅐RhoGDI␣ Ac-Lys-178 crystals were obtained by the sitting drop vapor diffusion method in 0.2 M lithium sulfate, 0.1 M BisTris/HCl, pH 5.5, 20% (w/v) PEG3350 overnight at 20°C. 30% (w/v) D-glucose was used as a cryoprotectant. The crystals belonged to the space group P6 1 with one heterodimer per asymmetric unit. The native data set was collected at the Swiss Light Source in Villingen at 100 K on beamline X06DA at a wavelength of 1 Å using a Pilatus 2M detector. The oscillation range was 0.1°C, and 2700 frames were collected. The program XDS was used for indexing and integration (32). Scaling was done with Aimless (33). Initial phases were determined using the program Phaser within the program suite Phenix-dev-1893 and the Cdc42-G⅐RhoGDI␣ (PDB entry 1DOA) structure as a search model (34,35). Refinement was done using the program phenix.refine (36). Restraints for the cysteine-farnesyl thioether were obtained by the program eLBOW in Phenix (37). The program Coot version 0.7.1 was used to build the model into the 2F o Ϫ F c and F o Ϫ F c electron density maps in iterative rounds of refinement with phenix.refine for RhoA-F⅐RhoGDI␣ Ac-Lys-178 (38,39). In the final model, 100% of all residues are in the allowed regions of the Ramachandran plot as judged by the program Molprobity (40,41). All structure figures presented here were made with PyMOL version 1.7.2.0 (42). Data collection and refinement statistics are given in Table 2. R work is calculated as follows: and F c are the observed and calculated structure factor amplitudes, respectively. R free is calculated as R work using the test set reflections only.
Liquid Chromatography and Mass Spectrometry-Peptides derived from in-gel digestion were desalted using the stopand-go extraction technique prior to mass spectrometry analysis (43). Liquid chromatography and tandem mass spectrometry instrumentation consisted of an nLC 1000 nano liquid chromatography coupled via a nano-electrospray ionization source to a quadrupole-based mass spectrometer, QExactive. Peptide separation was done by linear increase of buffer B within a binary buffer system: 1) 0.1% formic acid in water and 2) 0.1% formic acid in acetonitrile. Within 40 min, the content of buffer B was raised to 27% followed by a washing (95%) and re-equilibration (5%) step within 10 min. The mass spectrometer acquired spectra in a data-dependent mode. After MS1 acquisition at 70,000 resolution at 200 m/z (automatic gain control target, 3e 6 , maximum injection time, 20 ms), the 10 most intense peaks were selected for HCD fragmentation and measurement in the orbitrap mass analyzer (35,000 resolution at 200 m/z, AGC target: 5e 5 , maximum IT 120 ms). Dynamic exclusion was set to 25 s, limiting extensive refragmentation.
Proteomics Data Analysis-MaxQuant and the implemented Andromeda search engine were used to analyze raw data (44,45). MS/MS spectra were correlated against the Uniprot E. coli database containing the target protein. Default mass tolerance settings, a peptide length of 7 amino acids, and a minimal score for modified and unmodified peptides of 0 were used. The false discovery rate was controlled at the peptide-spectrum match, and the protein level was controlled using the decoy approach by a revert algorithm to 1%. Acetylation at lysine residues, oxidation of methionine, and acetylation at the protein N terminus was set as a variable modification, whereas carbamidomethylation of cysteines was defined as a fixed modification. For data analysis, we used the raw intensities of specific acetylation sites for analysis and normalized these by label-free quantification of used protein. Ratios were calculated comparing these values with control experiments without any KAT. Hierarchical clustering was performed for visualization using the gplots package from R Bioconductor.

RhoGDI␣ Is Endogenously Lysine Acetylated in Human
Cells-By quantitative mass spectrometry, RhoGDI␣ has been found to be lysine-acetylated at least eight distinct sites: Lys-43 and Lys-52 in the N-terminal domain as well as Lys-99, Lys-105, Lys-127, Lys-138, Lys-141, and Lys-178 in the Ig domain (Fig.  1A). We tested whether RhoGDI␣ is lysine-acetylated endogenously in human HeLaB and HEK293T cells. To this end, we performed an immunoprecipitation of acetylated proteins using an agarose bead-coupled anti-acetyl-L-lysine antibody followed by immunoblotting using an anti-RhoGDI␣ antibody. In HEK293T and HeLaB cells, we identified endogenously lysine-acetylated RhoGDI␣. Moreover, the signal was increased upon treatment with several classical and sirtuin deacetylase inhibitors, indicating that endogenous RhoGDI␣ acetylation is regulated by KDACs ( Fig. 1B; data not shown for HeLaB cells). As expected, the amount of acetylated tubulin, immunoprecipitated using the anti-acetyl-L-lysine antibody-coupled beads, increased upon KDAC inhibitor treatment (Fig. 1B, bottom,  lane 4). Notably, the differences observed in the acetylation levels are not due to a variation in the expression levels because the amount of tubulin was similar in the samples treated with/ without KDAC inhibitors (Fig. 1B, Input). Additionally, we transiently expressed His 6 -RhoGDI␣ in HEK293T cells. Again, we found acetylation for His 6 -RhoGDI␣, and the acetylation level was increased upon the addition of a KDAC inhibitor mixture as shown by Ni 2ϩ -NTA pull-down (PD) and subsequent immunoblotting using an anti-acetyl-L-lysine antibody (Fig. 1C).
RhoGDI␣ Acetylation Affects Actin Filament Formation in Mammalian Cells-To test whether RhoGDI␣ acetylation interferes with Rho signaling, we analyzed how RhoGDI␣ acetylation affects cellular actin filament (F-actin) formation. To this end, we transiently overexpressed Gln and Arg mutants of RhoGDI␣-EGFP, mimicking the acetylated and non-acetylated state, in HeLaB cells and examined phalloidin-stained filamentous actin by immunofluorescence and subsequent fluorescence intensity quantification (Fig. 2, A and B). Notably, we observed by immunoblotting experiments that the expression levels for RhoGDI␣ and the mutants were nearly the same (data not shown). Overexpression of K178Q G (superscript "G" for RhoGDI␣) resulted in markedly thickened actin stress fibers and significantly increased the cellular F-actin content compared with cells expressing RhoGDI␣ WT or K178R G . In fact, the expression of K178R G reduced the actin filament level even below the value observed for the mock control, showing that neutralization of the charge and steric effects are phenotypically distinguishable. For K141Q G , we also observed an increase in F-actin; however, in contrast to the reported data, it was not statistically significant compared with RhoGDI␣ WT (Fig. 2B,  bottom) (21). The mutants K99Q/R G both led to a significant increase in total actin filament content. In contrast, from all of the acetylation mimetics analyzed, the RhoGDI␣ K52Q/R are the only mutants significantly reducing the cellular F-actin compared with RhoGDI␣ WT to levels observed for MOCK control (Fig. 2B, top). This suggests that K52Q/R G both switch off RhoGDI␣ function and that lysine acetylation at Lys-52 G might exert similar effects, mechanistically by electrostatic as well as steric components.
Preparation of Site-specifically Lysine-acetylated RhoGDI␣ and Farnesylated RhoA-To explain the phenotypes observed in cells, we mechanistically analyzed the impact of lysine acetylation on RhoGDI␣ function. To this end, we used a synthetically evolved, orthogonal acetyl-L-lysyl-tRNA-synthe- tase/tRNA CUA pair from M. barkeri in E. coli to site-specifically incorporate acetyl-L-lysine recombinantly into RhoGDI␣ (genetic code expansion concept) (29,46,47). We successfully produced homogeneously acetylated RhoGDI␣ in yields suitable for biophysical studies as judged from immunoblotting and electrospray ionization (ESI)-mass spectrometry (Fig. 3, A and  B). Notably, the anti-acetyl-L-lysine antibody recognizes the acetylated proteins differentially, suggesting that residues in vicinity are important for antigen recognition (Fig. 3A).
It was shown earlier that prenylation of RhoGNBPs is essential for their biological function; however, it is independent of a specific isoprenoid modification (48). Importantly, in vivo or in vitro geranylgeranylated RhoA (RhoA-G) was prone to aggregation at micromolar concentrations, and the reported picomolar affinities of RhoA-G toward RhoGDI␣ are above the detection limit of ITC (data not shown) (14).
To study the impact of RhoGDI␣ acetylation on RhoA function, we therefore used in vitro farnesylated RhoA as a model system. To this end, we used recombinantly expressed, purified human farnesyltransferase and farnesylpyrophosphate as a substrate to quantitatively farnesylate RhoA (RhoA-F) and Cdc42 (Cdc42-F) in vitro, as shown by ESI-mass spectrometry (Fig. 3C). RhoA-F was able to form a stable, low nanomolar 1:1 complex with RhoGDI␣ as shown by analytical size exclusion chromatography and ITC (Figs. 3D and 4A). Furthermore, it was able to bind to liposomes dependent on the presence of the farnesyl moiety, and it could be extracted by RhoGDI␣ in a concentration-dependent manner (Fig. 3E).
Functional Consequences of RhoGDI␣ Acetylation for the Interaction toward Prenylated RhoGNBPs-We examined whether acetylation of RhoGDI␣ impacts binding toward RhoA-F by ITC. RhoGDI␣ Ac-Lys-52 and the N-terminally truncated ⌬66 G mutant both strongly interfered with RhoA-F binding ( Fig. 4A and Table 1). However, whereas ⌬66 G completely abolishes RhoA-F binding, Ac-Lys-52 G showed at least residual, albeit 4 orders of magnitude reduced, binding toward RhoA-F as shown by ITC (K D ϭ 4.1 M; Table 1). Notably, Ac-Lys-52 G does not interfere with the structural integrity of RhoGDI␣ because the protein behaved as RhoGDI␣ WT in size exclusion chromatography analyses (data not shown). These results strongly suggest a pivotal role of the RhoGDI␣ N terminus for the proper insertion of the prenyl moiety into the hydrophobic cavity of the Ig domain. The acetylation at Lys-138 G leads to a 50-fold reduction in the affinity toward RhoA-F ( Fig.  4A and Table 1). The importance of this residue is in agreement  with recent findings showing that SUMOylation increases the affinity toward RhoA (25). Acetylation at Lys-138 G would block the SUMOylation, strongly indicating a PTM cross-talk. Structurally, Ac-Lys-138 G on ␤6 might favor interactions with His-131 G and Tyr-133 G on ␤5 (Fig. 4C, right). However, to finally judge how Ac-Lys-138 G mechanistically exerts this effect, we would need further structural data. RhoGDI␣ Ac-Lys-178 showed a strongly reduced affinity toward both RhoA-F (141 nM) and Cdc42-F (98 nM) ( Fig. 4A and Table 1). This supports previous findings showing that this region is of importance for Rho-binding because mutation of I177N G , directly adjacent to Lys-178 G , leads to a 20-fold reduction in Cdc42 affinity (49,50). Ile-177 G is oriented toward the hydrophobic pocket of the Ig domain, and Lys-178 G is surface-exposed. Acetylation at Lys-178 G , by affecting the polarity and hydrophobicity of the Ig domain, might affect the Ig domain's conformation, thereby affecting binding toward prenylated RhoA. As a support, whereas the RhoGDI␣ Ac-Lys-178 RhoA-F interaction is enthalpically and entropically favored, all of the other interactions are entropically unfavored, indicating that acetylation of RhoGDI␣ at Lys-178 alters the binding mechanism toward RhoA-F (Table 1).
Crystal Structure of RhoGDI␣ Ac-Lys-178 in Complex with Farnesylated RhoA⅐GDP-To show how Lys-178 acetylation of RhoGDI␣ alters the binding mechanism toward RhoA-F, we solved the structure of a RhoA-F⅐RhoGDI␣ Ac-Lys-178 complex at 3.35 Å resolution by x-ray crystallography (Fig. 4B and Table 2). The complex crystallized in the space group P6 2 with one heterodimer per asymmetric unit. We observed reliable electron density for the Ac-Lys-178 on ␤9 and also for the farnesyl moiety as judged by omit maps showing that Ac-Lys-178 and the farnesyl are correctly positioned (Fig. 4E). The overall conformation of the complex was almost unaltered compared with a previously solved structure of RhoA-G⅐ RhoGDI␣ (PDB entry 4F38, overall root mean square deviation 0.501 Å; Table 3). However, the hydrophobic pocket of the RhoGDI␣ Ac-Lys-178 Ig domain has a slightly bigger volume, and the farnesyl group is positioned differently, penetrating deeper into the pocket compared with the geranylgeranyl group in the RhoA-G⅐RhoGDI␣ structure (PDB entry 4F38; Fig. 4D). Compared with the non-prenylated RhoA⅐RhoGDI␣ complex (PDB entry 1CC0), it showed conformational differences mostly to RhoGDI␣ (overall root mean square deviation 0.804 Å; Table 3). Ac-Lys-178 on RhoGDI␣ ␤9 structurally couples three ␤-strands of the Ig domain ␤5, ␤9, and ␤10) by forming a stacking interaction with Tyr-133 G on ␤5 and a hydrogen bond with Asn-176 G on ␤9. Asn-176 G in turn forms a hydrogen bond to Glu-193 G on ␤10 (Fig. 4C, left). These interactions are lacking in the non-acetylated structure, as shown by superposition of the structure presented here with the non-acetylated structure (Protein Data Bank entry 4F38; Fig. 4C, left). The Ig domain is able to switch between an open, prenyl-bound, and closed, prenyl-free, state by undergoing a rigid body reorientation on two invariant glycine residues (Gly-125 G and Gly-147 G ). Glycine-based hinges are reported to be important for conformational flexibility also in other systems (51). By structurally coupling three ␤ strands in the Ig domain, Ac-Lys-178 G might interfere with the inherent Ig domain's conformational flexibility essential for the function of the Ig domain, resulting in the observed decrease in RhoA-F affinity.
To this end, HeLa T-REx cells stably expressing His 6 -SUMO1 were transiently transfected with RhoGDI␣-EGFP constructs, and His 6 -SUMO1-modified proteins were pulled down by Ni 2ϩ -NTA beads. Using an anti-GFP antibody, we showed that RhoGDI␣ WT is conjugated to SUMO1 in cells in contrast to RhoGDI␣ K141Q and K138A, which were not found to be SUMOylated (Fig. 5B).
To confirm these results and to quantify the SUMOylation efficiency, we additionally applied an in vitro SUMOylation assay. We used non-acetylated/lysine-acetylated RhoGDI␣ as substrate proteins and purified SUMO1, the E1 SUMO-activating enzyme Aos1/Uba2, and the E2 SUMO-conjugating enzyme Ubc9 to catalyze the reaction. Using an anti-SUMO1 antibody, non-acetylated RhoGDI␣ WT protein showed a clear SUMOylation signal (Fig. 5C). The SUMOylation of the K138A/D G mutants was completely abolished, confirming that Lys-138 is the major SUMOylation site in RhoGDI␣ directly competing with acetylation (25). RhoGDI␣ Ac-Lys-141 and RhoGDI␣ Ac-Lys-127,141 showed almost no SUMOylation, reflecting that acetylation at Lys-141 G blocks RhoGDI␣ SUMOylation at Lys-138 (Fig. 5C). Thus, these results provide evidence for both a direct and an indirect cross-talk of RhoGDI␣ acetylation with its SUMOylation in vitro and in vivo.
Acetylation of RhoGDI␣ at Lys-52 Impairs RhoA Extraction from Cellular Membranes-To test whether acetylation of RhoGDI␣ interferes with its capability to extract endogenous RhoA from cellular membranes, we prepared membrane fractions from HEK293T cells and incubated those with site-specifically acetylated RhoGDI␣ proteins. Solubilized endogenous geranylgeranylated RhoA was separated from residual membrane-attached RhoA via ultracentrifugation and detected by immunoblotting of the supernatant (Fig. 6A, left). Most strikingly, RhoGDI␣ Ac-Lys-52 was not able to significantly extract RhoA compared with RhoGDI␣ WT (Fig. 6A, right). This supports our previous results showing that acetylation at Lys-52 G abolishes RhoA binding. All of the other acetylated RhoGDI␣ proteins were capable of extracting modified RhoA to an extent comparable with RhoGDI␣ WT at least under the assay conditions used.
RhoGDI␣ Acetylation Interferes with GEF-catalyzed Guanine Nucleotide Exchange-To investigate how RhoGDI␣ acetylation interferes with GEF-catalyzed nucleotide exchange on RhoA, we performed a fluorescence-based nucleotide exchange assay. We used farnesylated RhoA loaded with fluorescently labeled mantGDP and preformed complexes with acetylated/ non-acetylated RhoGDI␣ (concentration 1 M). We followed the fluorescence quenching of the nucleotide exchange reactions catalyzed by 0.5 M Dbs-GEF in presence of a 50-fold molar excess of unlabeled GDP (concentration 50 M) (Fig. 6B). Surprisingly, we observed two phases in the exchange assays for all reactions with acetylated RhoGDI␣ but not for Ac-Lys-52 G and also not for the reactions without RhoGDI␣. This might reflect that the Dbs-GEF itself is able to act as a GDI displacement factor. As expected, the presence of RhoGDI␣ WT substantially decreased the Dbs-GEF-catalyzed nucleotide exchange compared with the reaction without RhoGDI␣ (Fig.  6B). All complexes with acetylated RhoGDI␣, except for Ac-Lys-52 G , showed exchange kinetics similar to non-acetylated RhoGDI␣ WT . However, for Ac-Lys-52 G , the nucleotide exchange rates approached values similar to the reactions in the absence of RhoGDI␣. The fact, that they are not exactly the same can be explained by the observed residual binding of Ac-Lys-52 G to RhoA-F (Table 1). Structurally, Lys-52 G forms a hydrogen bond with Tyr-63 R (superscript "R" for RhoA) and is in intramolecular hydrogen bond distance to the main chain carbonyl oxygen of Leu-41 G (Fig. 6C). Acetylation of Lys-52 G might sterically interfere with the formation of the N-terminal helix-loop-helix conformation contacting RhoA switch I and switch II, thereby abolishing RhoGDI␣-binding. These data suggest that acetylation at Lys-52 G could act as a displacement factor for the high affinity RhoA⅐GDP⅐RhoGDI␣ complex.
CBP, p300, and pCAF Lysine-acetylate RhoGDI␣-As shown above, RhoGDI␣ was acetylated at a basal level in HEK293T cells also in the non-KDAC inhibitor-treated samples, indicating KAT-catalyzed acetylation of RhoGDI␣ (Fig. 1, B and C). To identify RhoGDI␣-specific KATs, we first used an in vitro approach. We selected six major KATs (CBP, p300, pCAF, Gcn5, Tip60, and ␣-TAT1) and used these active enzymes and purified RhoGDI␣ as a substrate for in vitro acetylation. By immunoblotting, we identified p300 and pCAF as KATs for RhoGDI␣ (Fig. 6A). The acetylation was dependent on the presence of acetyl-CoA, KAT enzyme, and RhoGDI␣. We did not obtain a signal with acetyl-CoA alone or in the absence of RhoGDI␣, excluding the possibility that the signal derives from non-enzymatic acetylation or KAT-autoacetylation activity, respectively (Fig. 7A, left and right (control)).
To confirm our results, we also used an in vivo approach by transiently coexpressing His 6 -RhoGDI␣ together with Myctagged (CBP, p300, pCAF, Gcn5, and ␣-TAT1) or His 6 -tagged (Tip60) KATs in HEK293T cells. We pulled down His 6 - For comparison, the structural similarity between the complexes of RhoA⅐G⅐RhoGDI␣ (PDB entry 4F38) and RhoA⅐GDP⅐RhoGDI␣ (PDB entry 1CC0) is shown in the right column.

RhoA-F⅐GDP⅐RhoGDI␣ AcK178 (PDB entry 5FR2)
RhoA-G⅐RhoGDI␣ (PDB entry 4F38)  RhoGDI␣ and analyzed the amount of acetylated RhoGDI␣ by immunoblotting. As shown in Fig. 7C, overexpression of both p300 and CBP resulted in a significant increase of acetylated RhoGDI␣, supporting our in vitro data. Notably, the overexpression of pCAF did not increase the acetylation level of RhoGDI␣, indicating that different mechanisms operate to acetylate RhoGDI␣ in vitro and in vivo. Sirt2 and HDAC6 Act as RhoGDI␣ Deacetylases-Treatment of cells with KDAC inhibitors resulted in an increase in the acetylation level of endogenous and transiently expressed RhoGDI␣ (Fig. 1, B and C). To identify the KDACs responsible for RhoGDI␣ deacetylation, we performed an in vitro dot blot screen with all mammalian deacetylases (data not shown). We discovered the cytosolic enzymes Sirt2 and HDAC6 to act as RhoGDI␣ deacetylases. To verify these results, we determined the extent of RhoGDI␣ deacetylation by immunoblotting (Fig.  7D). Apparently, compared with the non-enzyme control, RhoGDI␣ Ac-Lys-43, Ac-Lys-52, Ac-Lys-138, and Ac-Lys-178 were significantly deacetylated by Sirt2 (Fig. 7D, bottom left).
For HDAC6, Ac-Lys-52 G again is the preferred substrate becoming significantly deacetylated (Fig. 7D, bottom right). The other acetylation sites were not significantly deacetylated by HDAC6. Time course experiments with Ac-Lys-52 G revealed a nearly complete deacetylation by Sirt2 after 30 min, whereas even using double the amount of HDAC6 did not show complete deacetylation after 6 h (Fig. 7E).

Discussion
RhoGDI␣ is a major regulator of RhoGNBP function modulating its subcellular distribution and turnover, thereby creating a cytosolic pool of inactive Rho easily activatable for a fast cellular response. Thus, RhoGDI␣ is essential for the precise spatial and temporal regulation of RhoGNBPs and affects their downstream signaling. How RhoGDI␣ is regulated to create a high specificity of action is only marginally understood. Here, we present the first functional and structural study describing how RhoGDI␣ function is controlled by post-translational lysine acetylation (Fig. 7F). Using the powerful genetic code  Fig. 5B). Lys-52 G is an integral part of the N-terminal domain within hydrogen bond distance to Tyr-63 R in switch II and the main chain carbonyl oxygen of Leu-41 G . The side chain of Lys-43 G is surface-exposed. expansion concept to produce site-specifically lysine-acetylated proteins allows us to study its real impact on protein function.
As we show here, RhoGDI␣ acetylation influences the membrane/cytosol and GTP/GDP cycle of Rho proteins, phenotypically observable by affecting the cellular actin cytoskeleton (Lys-52 G , Lys-99 G , Lys-141 G , and Lys-178 G ). It furthermore interferes with binding to prenylated (Ac-Lys-52 G , Ac-Lys-138 G , and Ac-Lys-178 G ) RhoGNBPs. Lysine acetylation per se leads to an electrostatic quenching of the lysine side chain's positive charge and an increase of the hydrophobicity. Additionally, we observe that Ac-Lys-178 G structurally couples ␤5, ␤9, and ␤10 of the Ig domain, most likely interfering with the inherent Ig flexibility (structural coupling). Moreover, we show structurally that the volume of the hydrophobic cavity of the Ig domain is increased, adopting the farnesyl group differently compared with the geranylgeranyl group (PDB entry 4F38). These mechanisms might contribute to the observed reduced RhoA-F binding. Ac-Lys-52 G is a loss-of-function modification completely abolishing RhoGDI␣ function, most likely by interfering with the transition of the intrinsically unfolded N-terminal domain to adopt the helix-loop-helix conformation upon binding to the Rho protein (structural destruction). Thereby, the cell has a system to quickly switch on/off RhoGDI␣ function. We observed that acetylation at the solvent-exposed Lys-138 G also reduced the binding toward RhoA-F. The exact molecular mechanisms need further examination but might also include hydrophobic shielding from the polar solvent as well as structural coupling. RhoGDI␣ acetylation interferes directly (Ac-Lys-138 G ) and indirectly (Ac-Lys-141 G ) with Lys-138 G SUMOylation, reported to increase the RhoA affinity (PTM cross-talk). By impairing SUMOylation, Lys-141 G acetylation might therefore indirectly affect the observed increase of RhoA at the membrane and cellular F-actin content. Using these mechanisms, RhoGDI␣ lysine acetylation is a powerful cellular system to tightly regulate the RhoGNBP turnover and residence time at the membrane, finally resulting in a modulation of RhoGNBP signaling.
We have discovered that acetylation at Lys-178 G lowers the affinity toward RhoA-F, resulting in more RhoA located at the plasma membrane. Consistent with an increased Rho signaling, we found more filamentous actin in cells expressing the respective acetylation mimicking mutant.
RhoGDI␣ Lys-52 acetylation completely blocks the interaction with RhoA, drastically reduces RhoA-F binding (4 orders of magnitude) and the membrane extraction capability. This is supported by previous results showing that the N-terminal domain of RhoGDI␣ is essential for membrane extraction as well as delivery of Rho proteins. Consistently, the expression of K52Q G in the N-terminal domain resulted in a cellular F-actin content comparable with mock control, indicating that acetylation at Lys-52 G leads to a complete loss of function. Notably, Lys-52 G acetylation restores the GEF-catalyzed nucleotide exchange rates, nearly approaching values observed for uncomplexed RhoA. This shows that it might act as a bona fide RhoGDI␣ displacement factor.
An open question for many proteins identified by quantitative mass spectrometry to be lysine-acetylated is the biological significance of the modification. Many acetylation sites are of low stoichiometry, for which only a gain of function might confer an additional property. Technical progress to determine absolute quantities of acetylation on a whole proteome scale will help to identify biologically important sites (52,53). Notably, in contrast to many reported studies, for RhoGDI␣, we were able to find endogenously acetylated protein in several human cell lines, showing that a significant amount is lysineacetylated and suggesting that acetylation is of biological importance to control RhoGDI␣ function. However, the exact absolute quantities could vary and might depend on the cellular metabolic, physiologic, and cell cycle state. Additionally, the expression level and activity of KDACs and KATs might play an important role. Here, we show that RhoGDI␣ is targeted by the KATs CBP, p300, and pCAF. The KATs p300 and pCAF acetylate RhoGDI␣ at Lys-52, -138, -141, and -178, the sites being important to control RhoGDI␣ function as shown here. Moreover, RhoGDI␣ is differentially deacetylated by Sirt2 and HDAC6 at these sites.
Sirt2 and HDAC6 are the main cytosolic deacetylases known to physically interact, to work synergistically, to colocalize with microtubules, and to deacetylate many proteins involved in cytoskeleton regulation, such as cortactin, mDia2, and ␣-tubulin (54 -56). Down-regulation or inhibition of Sirt2 and HDAC6 were both found to inhibit cell motility as well as cell migration (57). Sirt2 deacetylates and thereby inactivates p300, which in turn is able to inactivate both Sirt2 and HDAC6, constituting a feedback loop (58,59). The balance between acetylation/deacetylation of RhoGDI␣ might interfere with cell adhesion, cell migration, and cell motility by directly affecting Rho signal transduction pathways (60).
In summary, the presence of endogenously acetylated RhoGDI␣ in mammalian cells, site-specific KATs, and KDACs suggests that acetylation is a physiologically important posttranslational modification to control RhoGDI␣ function. Depending on when and where RhoGDI␣ acetylation takes place in vivo, the cell is able to directly control the Rho protein's lifetime and turnover, its residence time on the membrane, and its subcellular distribution. All of these processes strongly contribute to the initiation and termination of Rho signal transduction pathways, ultimately balancing cell resting and cell motility. A dysfunction in its regulation might therefore lead to severe cellular defects supporting tumorigenesis and neurodegeneration. Because there are just three RhoGDIs in mammalian cells, modification of RhoGDI␣ by post-translational lysine acetylation can create a sophisticated system to spatially and temporally balance RhoGNBP signal transduction pathways. Tackling the RhoGDI␣ lysine acetylation machinery might thus be a promising yet underestimated therapeutic approach.
Author Contributions-N. K., S. W. L. C. J., J. W. C., M. K., H. N., A. Z., G. P., and M. L. designed the experiments, discussed data, generated the data, and wrote the manuscript. L. S. and L. B. provided technical assistance with the project. S. W., U. B., and M. L. performed structural analysis. M. S., S. W., and M. L. took x-ray data sets. S. d. B., P. K., and J. B. performed experiments and discussed data.