A Novel C3-like ADP-ribosyltransferase fromStaphylococcus aureus Modifying RhoE and Rnd3*

Clostridium botulinum C3 is the prototype of the family of the C3-like transferases that ADP-ribosylate exclusively RhoA, -B and -C. The ADP-ribose at Asn-41 results in functional inactivation of Rho reflected by disaggregation of the actin cytoskeleton. We report on a new C3-like transferase produced by a pathogenic Staphylococcus aureus strain. The transferase designated C3Stau was cloned from the genomic DNA. At the amino acid level, C3Stau revealed an identity of 35% to C3 from C. botulinum and Clostridium limosumexoenzyme, respectively, and of 78% to EDIN from S. aureus. In addition to RhoA, which is the target of the other C3-like transferases, C3Stau modified RhoE and Rnd3. RhoE was ADP-ribosylated at Asn-44, which is equivalent to Asn-41 of RhoA. RhoE and Rnd3 are members of the Rho subfamily, which are deficient in intrinsic GTPase activity and possess a RhoA antagonistic cell function. The protein substrate specificity found with recombinant Rho proteins was corroborated by expression of RhoE in Xenopus laevis oocytes showing that RhoE was also modified in vivo by C3Stau but not by C3 from C. botulinum. The poor cell accessibility of C3Stau was overcome by generation of a chimeric toxin recruiting the cell entry machinery of C. botulinum C2 toxin. The chimeric C3Stau caused the same morphological and cytoskeletal changes as the chimeric C. botulinum C3. C3Stauis a new member of the family of the C3-like transferases but is also the prototype of a subfamily of RhoE/Rnd modifying transferases.

Various bacterial protein toxins interfere with eukaryotic cell functions by catalyzing a posttranslational modification of essential cellular regulator proteins such as ADP-ribosylation of the Rho proteins by C3-like transferases. Clostridium botulinum exoenzyme C3 is the prototype of a family encompassing exoenzymes from Clostridium limosum and Bacillus cereus and from Staphylococcus aureus (EDIN) 1 (1)(2)(3)(4). The members of this family are similar in structure and homologous to each other. They are single-chained ADP-ribosyltransferases with a molecular mass of ϳ25 kDa. The C3-like transferases are in fact mere exoenzymes devoid of the cell entry apparatus harbored by other toxins, and they are thought to enter cells by nonspecific pinocytosis. C3 catalyzes ADP-ribosylation of the RhoA, -B and -C subtypes but not of other members of Rho and Ras subfamilies (3)(4)(5). Only in the presence of the detergent sodium dodecyl sulfate, the Rac protein is a poor substrate (4). The ADP-ribose moiety is transferred from NAD ϩ to the acceptor amino acid Asn-41 and is linked N-glycosidically to the amide group of the carboxylate side chain of Asn-41 (6).
The Rho proteins belong to the Ras superfamily of low molecular mass GTPases, which are the major regulators of the actin cytoskeleton, but they are also involved in cell cycle progression, transcriptional activity, and in cooperation with Ras in cell transformation (7)(8)(9). Because ADP-ribosylation of Rho in intact cells results in disaggregation of the actin cytoskeleton, ADP-ribosylation has been classified as an inactivating modification. The molecular basis of the inactivation is the decreased interaction of Rho with the exchange factors, which promote activation (10), and the sequestration by the negative regulator guanine nucleotide exchange factor. 2 Although C3 induces the breakdown of the actin cytoskeleton, an effect that can be easily monitored, it is now clear that C3catalyzed ADP-ribosylation inactivates all Rho functions (for reviews see Refs. [12][13][14]. Because of its confined protein substrate specificity, C3 has been advanced to a widely used tool in cell biology to selectively turn off cellular Rho functions. The poor cell accessibility of C3 has been overcome by the creation of chimeric C3 toxins using the cell entry apparatus of other toxins (15,16). We report here on a novel C3-like transferase produced by a pathogenic S. aureus strain that ADP-ribosylates RhoE/Rnd3 subtype proteins in addition to RhoA, -B and -C.

EXPERIMENTAL PROCEDURES
Materials and Chemicals-Culture supernatant was subjected to purification using the FPLC System, an anion exchange MonoQ column, and a cation exchange MonoS column from Amersham Pharmacia Biotech. Oligonucleotides were obtained from Mannfred Weichselgartner (Ebersberg, Germany), PCR was carried out using the Gene Amp 2400 System from PerkinElmer Life Sciences, and DNA sequencing was carried out with the Big Dye Terminator Cycle Sequencing Ready Reaction Kit from PerkinElmer Life Sciences. The Topo-TA-vector system was from Invitrogen (Groningen, The Netherlands), the pGEX2T vector system from Amersham Pharmacia Biotech, and the Quickchange Kit was from Stratagene (Heidelberg, Germany). Restriction * This work was supported by Deutsche Forschungsgemeinschaft Grant SFB 388 and Project Ak6/10. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Purification-S. aureus strain HMI6 was a clinical isolate from a patient with postoperative infection. The identity of this isolate was confirmed in the laboratory by routine tests. For purification of the ADP-ribosyltransferase, bacteria were grown in LB medium at 37°C overnight. After centrifugation at 4°C, ammonium sulfate was added to a concentration of 70% to precipitate proteins from the culture supernatant. The pellet was resuspended in 50 mM HEPES, pH 7.0, and dialyzed against the same buffer overnight to remove the salt. The solution was added onto the MonoQ column using a 10-ml super loop, and the flow through was collected, dialyzed against a buffer containing 50 mM HEPES, pH 6.0, and loaded onto the MonoS column. The transferase was eluted with a linear gradient at 0.3 M NaCl in 50 mM HEPES, pH 6.0.
ADP-ribosylation Reaction-Recombinant RhoA, RhoE/Rnd3, and other GTPases (2 M) were ADP-ribosylated in a buffer containing 50 mM HEPES, pH 7.3, 2 mM MgCl 2, 20 M [adenylate-32 P]NAD, and 100 g/ml bovine serum albumin for up to 4 h at 37°C. The total volume was 25 l. The recombinant toxin was applied in a concentration of 1 nM or as indicated.
SDS-PAGE-SDS-polyacrylamide gel electrophoresis was performed according to the methods of Laemmli (19). Gels were stained with Coomassie Brilliant Blue R-250, dried, and further analyzed by the PhosphorImager SI from Molecular Dynamics.
Amino Acid Sequencing Analysis-The eluted fraction from MonoS column containing the transferase was separated by SDS-PAGE, transferred onto polyvinylidene difluoride membrane (Amersham Pharmacia Biotech), and visualized with Amido Black. N-terminal amino acid sequencing was performed on an excised band using an Applied Biosystems 447A pulse-liquid protein sequencer.
Amplification from Genomic DNA-Genomic DNA from strain S. aureus HMI6 was prepared by standard methods. Primers for amplification of the HMI6 gene were designed according to the determined N-terminal amino acid sequence and in consideration of the staphylococci codon usage. For the 3Ј end, the primer was designed according to a region 100 base pairs downstream of the in frame stop codon in the published EDIN sequence (2). PCR was performed under the following conditions: 500 ng of template, 2.5 mM of each dideoxynucleotide, 5 l of 10-fold concentrated Mg 2ϩ -free buffer, different concentrations of MgCl 2 , and variant units of Taq polymerase (New England Biolabs). The primers were: HMI6-N, 5Ј AGA TCT GCC GAG ACT AAA AAT TTT ACA G 3Ј; and HMI6-C, 5Ј GGA TCC TAA GTT TAA AGC GTA TTT TTA G 3Ј. PCR products were separated by gel electrophoresis, purified, ligated into the TOPO-TA vector, and sequenced. For mobilization of the C3 Stau gene, the HMI6 -5Ј end primer and a primer corresponding to the 3Ј end of C3 Stau were used. The primers were flanked by a 5Ј BglI and a 3Ј BamHI site, which were used for the ligation of the C3 Stau gene into the pGEX vector. The amino sequence of C3 Stau is available in the GenBank TM /EBI data base under accession number AJ277173.
Expression of Recombinant C3 Stau Transferase-For expression of the transferase, bacteria were grown overnight in LB media in the presence of 100 g/ml of ampicillin, followed by inoculation in fresh media. At an A 600 of 0.8, isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.2 mM. After incubation for 6 h at 37°C, bacteria were pelleted, resuspended in lysis buffer containing 50 mM HEPES, pH 7.3, 2 mM MgCl 2 and 1 mM phenylmethylsulfonyl fluoride, and broken with a French Press (SLM Aminco). Debris was removed by centrifugation for 15 min at 15,000 ϫ g at 4°C, and GST fusion protein was purified from supernatant with glutathione-Sepharose beads. The GST carrier was cleaved with thrombin followed by its removal with benzamidine beads.
Preparation of Oocyte Lysates Overexpressing Human RhoE-Human RhoE cRNA was transcribed in vitro using the RhoE-pNK2 as template and injected into defolliculated Xenopus laevis oocytes. After 24 h at 19°C, lysates were prepared as described from cRNA-infected cells and noninjected cells (20).
Construction of the C2IN-C3 Stau Fusion Toxin-The fusion toxin containing the N-terminal 225 amino acids from C. botulinum C2 toxin and the full-length C3 Stau was constructed using the 5Ј BglI site and the 3Ј BamHI site flanking the toxin gene. Construction, expression, and purification of GST fusion protein were carried out as described (16).
Cytotoxic Assay-NIH3T3 and KB cells were grown to subconfluency at 37°C in Dulbecco's minimum essential medium containing 10% fetal calf serum, 2 mM L-glutamate, 100 units penicillin/ml, and 100 g/ml streptomycin at 5% CO 2 . For cytotoxic assays, subconfluent cells in 24-well plates containing coverslips were treated for 3 h with 200 ng/ml activated C2II toxin, C2IN-C3 Stau alone (100 ng/ml), or with 200 ng/ml activated C2II and C2IN-C3 Stau . As a control, cells were treated with 200 ng/ml activated C2II and 100 ng/ml C2IN-C3 (limosum) for the same time. Cells growing on coverslips were washed twice with phosphate-buffered saline and fixed with 4% paraformaldehyde and 0.1% Triton X-100 in phosphate-buffered saline for 30 min. Actin was stained with phalloidin-rhodamine (600 ng/ml) and the coverslips were mounted in Moviol.

RESULTS
Several clinical isolates of S. aureus were screened for C3like activity. An isolate designated HMI6 was identified to possess ADP-ribosyltransferase activity, which modified recombinant RhoA. The transferase was purified from the culture supernatant by applying ammonium sulfate precipitation and ion exchange chromatography. The partially purified protein was excised from SDS-polyacrylamide gel and N-terminally sequenced. Twenty-three amino acids of the N terminus were identified revealing an identity of 78% (18 of 23 amino acids) with the EDIN transferase from S. aureus (Fig. 1). The ADPribosyltransferase from S. aureus was designated C3 Stau to make a distinction from C3 bot (C. botulinum), C3 lim (C. limosum), and C3 cer (B. cereus).
The gene coding for C3 Stau was amplified from the genomic DNA by PCR. The 5Ј primers were deduced from the identified N-terminal amino acid sequence, and the basis for the 3Ј primers was the noncoding region EDIN, the sequence of which is known. Indeed, one major PCR product was obtained. From this product, the C3 Stau gene was further cloned. The determined sequence is presented in Fig. 2, and the alignment with the sequences of EDIN, C3 bot , and C3 lim is given in Fig. 3. The molecular mass of C3 Stau was calculated as 23,640 Da, and the theoretical isoelectric point was 9.4. The sequence comparison revealed an identity of 78% with EDIN and 35% with C3 bot and C3 lim . The C3 Stau transferase gene was cloned into the pGEX-2T expression vector. GST-C3 Stau was nicely expressed in E. coli, and the removal of the GST part by thrombin treatment resulted in a stable protein with a molecular mass of 24 kDa (Fig. 4). The recombinant C3 Stau was used to compare its properties with that of C3 bot exoenzyme, the prototype of C3like transferases.
To test the protein substrate specificity, recombinant Rho subfamily proteins (RhoA, Rac1, Cdc42, RhoD, RhoE, TC10, and RhoN) were incubated with C3 Stau in the presence of [ 32 P]NAD. As shown in Fig. 5A, RhoA was [ 32 P]ADP-ribosylated by C3 bot and C3 Stau . Surprisingly, RhoE was modified by C3 Stau but not by C3 bot . The new substrate specificity of C3 Stau was corroborated by showing that Rnd3 was also a substrate The time course revealed that RhoE/Rnd3 was ADP-ribosylated more slowly by C3 Stau compared with RhoA, and the linear phase of ADP-ribosylation lasted for hours. C3 bot , in contrast, did not catalyze any incorporation of ADP-ribose into RhoE/Rnd3 (Fig. 5B). RhoE and Rnd3 have been identified recently as GTPase-deficient low molecular mass GTP-binding protein, which has preferentially bound GTP (18,21,22). Because RhoA in the GDP-bound but not in the GTP-bound form is the preferred substrate for C3 bot (23), RhoE was artificially loaded with GDP and tested for ADP-ribosylation. However, the nucleotide occupancy did not change the kinetics of modification (data not shown). Thus, for unknown reasons, RhoE/ Rnd3 in contrast to RhoA was ADP-ribosylated with slow velocity.
Sequential ADP-ribosylation of recombinant RhoA, i.e. first with C3 bot followed by C3 Stau (and vice versa), indicated that both transferases linked the ADP-ribose to Asn-41 of RhoA (data not shown). This finding was confirmed by using the mutant Rho Asn41Ile , which was not a substrate for C3 Stau and C3 bot (Fig. 5C). Asn-41 is equivalent to Asn-44 in RhoE. Its exchange to Ile (RhoE Asn44Ile ) completely prevented ADP-ribosylation by C3 Stau , indicating that Asn-44 is the acceptor amino acid in RhoE (Fig. 5C).
To test whether cellular RhoE/Rnd3 was a substrate for C3 Stau , cell lysates and cellular subfractions were [ 32 P]ADPribosylated with C3 bot and C3 Stau . However, no radioactive band in addition to RhoA was detected (data not shown). It is conceivable that RhoE/Rnd3 is poorly expressed in the cell lines tested compared with RhoA and that radioactively labeled RhoA masked the traces of labeled RhoE/Rnd3. Therefore, a sequential ADP-ribosylation was performed, i.e. nonradioactive ADP-ribosylation by C3 bot was followed by C3 Stau -catalyzed [ 32 P]ADP-ribosylation. Preincubation with C3 bot followed by C3 Stau -catalyzed [ 32 P]ADP-ribosylation resulted in a radioactive labeling, whereas the opposite way showed no labeling, indicating a protein substrate that is modified by C3 Stau but not by C3 bot (Fig. 6A).
Another approach to check whether RhoE was an in vivo substrate for C3 Stau was the expression of RhoE in oocytes from X. laevis. In lysates from noninjected control oocyte, C3 Stau and C3 bot [ 32 P]ADP-ribosylated only RhoA (one band) (Fig. 6B). However, in lysates from oocytes expressing the RhoE protein, C3 Stau ADP-ribosylated two polypetides (double band), whereas C3 bot modified only one single band (Fig. 6B). Thus, cellular RhoE was a substrate for C3 Stau , corroborating the findings with recombinant Rho GTPases.
Exoenzymes C3 bot and C3 lim are cytotoxic to cultured cell lines to induce disaggregation of actin filaments but only when applied at micromolar concentrations (13). The same property was true for C3 Stau (data not shown). To overcome this limita-  tion, a chimeric C3 Stau was constructed analogous to that construct of C3 lim with C. botulinum C2 toxin (16). C3 Stau was fused to the enzymatically deficient C2I component, and the cell entry was mediated by the receptor binding component C2II from C. botulinum. C2IN-C3 Stau was as nontoxic to cells as C3 Stau was, but in the presence of C2II C2IN-C3 Stau induced the typical C3-like morphology and cytoskeletal changes (Fig.  7A). The differential ADP-ribosylation of lysates from intoxicated cells clearly demonstrated the in vivo ADP-ribosylation of cellular RhoA (Fig. 7B). DISCUSSION The first exoenzyme that was identified to ADP-ribosylate the Rho protein was exoenzyme C3 from C. botulinum. Because several isoforms of C3, all produced by C. botulinum types C and D, were identified, it was initially thought that C3 was associated with or even part of the botulinum neurotoxins. However, the later identification of homologues of C3 produced by C. limosum, B. cereus, and S. aureus, which definitely do not harbor genes for neurotoxins, proved that C3 and C3-like exoenzymes are independent from neurotoxins (24). All the C3 homologues are single chain proteins that are released from the bacteria and are, therefore, to be classified as exoenzymes. Compared with bacterial protein toxins such as pertussis toxin, cholera toxin, or Pseudomonas exotoxin A, the C3 homologues are devoid of a cell entry apparatus and seem to enter cells by nonspecific pinocytosis (25). Their function as virulence factors is unclear, but there is a report FIG. 6. RhoE is the intracellular substrate of C3 Stau . A, sequential ADP-ribosylation of rat brain membranes with C3 Stau and C3 bot . The membrane fractions from rat brain were ADP-ribosylated by C3 bot in the presence of unlabeled NAD. After washing, the membranes were divided and [ 32 P]ADP-ribosylated by C3 Stau and C3 bot , respectively. Samples were run on SDS-PAGE and analyzed by a PhosphorImager. B, the lysates of X. laevis oocytes were injected with RhoE cRNA, and noninjected controls were [ 32 P]ADP-ribosylated by C3 Stau or C3 bot . The PhosphorImager data of the SDS-PAGE are shown. that the C3 homologue EDIN from S. aureus inhibits differentiation and induces hyperplasia of epidermis (26). The C3 homologue C3 Stau , presented in this study, is also produced by a pathogenic S. aureus strain. C3 Stau shows 78% identity and 90% homology to EDIN but only 35% identity to C3 isoforms and the exoenzyme from C. limosum.
The identification of C3 exoenzyme is tightly linked to the elucidation of cellular Rho functions (13,24). The reason for this is the remarkable substrate specificity of C3 bot , which ADP-ribosylates only the subtypes RhoA, -B, and -C but not other target proteins. Surprisingly, C3 Stau modifies, in addition to RhoA, -B, and -C, the recently identified Rho subfamily members, RhoE/Rnd3. The acceptor amino acid in RhoE is Asn-44, the equivalent to the acceptor residue in RhoA, Asn-41. However, compared with RhoA, RhoE/Rnd3 is slowly modified; the linear phase lasts about 4 hours but eventually results in complete modification. The reason for this different property is unknown but may be based on its lack of GTP hydrolyzing activity. The basis for this is the exchange of three conserved amino acids in RhoE/Rnd3, which is canonically involved in the GTPase activity. GTPase activity of RhoE can be restored by replacement with the conserved amino acids (18). It is, therefore, conceivable that RhoE/Rnd3 possesses a conformation that resembles the active GTP-bound form independent of nucleotide occupancy. This notion explains the finding that loading of RhoE with GDP does not change the kinetics of ADPribosylation, whereas RhoA-GDP is modified about five times faster than RhoA-GTP (23). However, it is also conceivable that RhoE/Rnd3 needs interaction with other proteins or binding to membranes to be rapidly modified.
RhoE/Rnd3 belongs to a new branch of the Rho subfamily, the Rnd proteins (22). The members Rnd1, Rnd2, and Rnd3, the last one being nearly identical with RhoE, are about 50% identical with RhoA, but they show a remarkable functional difference to RhoA. They are deficient in intrinsic and GTPase acticating protein-stimulated GTPase activity (18,21,22). Furthermore, RhoE/Rnd3 has the opposite function of RhoA and can be classified as a functional RhoA antagonist (21,22). How RhoE/Rnd3 is regulated is so far unclear, possibly by expression or by subcellular translocation.
The initial step of RhoA activation, the interaction with guanine nucleotide exchange factor Lbc, is blocked by ADPribosylation (10). Based on this finding it is conceivable that RhoE/Rnd3 acts by sequestering exchange factors, thereby preventing activation of RhoA. ADP-ribosylation might inhibit binding of RhoE/Rnd3 to the exchange factors, thereby allowing the normal activation cascade of RhoA.
In the case of C3 bot and C3 lim , inactivation of RhoA by ADPribosylation allows RhoE/Rnd3 to act, thereby inducing disaggregation of the actin cytoskeleton. C3 Stau , however, inactivates both RhoA and its antagonist RhoE/Rnd3. Thus, one expects fewer morphological effects for C3 Stau than for C3 bot . However, the experimental data do not support this notion. The major reason for this might be the low expression of RhoE/Rnd3 in all tissues tested, which does not allow a direct functional antagonism to RhoA (22).
C3 Stau belongs to the family of C3-like transferases because of its homology and its ability to ADP-ribosylate RhoA. However, it is also the prototype of a novel subfamily of the C3-like transferases because of its extended protein substrate specificity, modifying RhoE and Rnd3 in addition to RhoA. C3 Stau and EDIN are produced by pathogenic S. aureus strains. S. aureus bacteria are reported to escape endosomes after phagocytosis by endothelial cells and to exist freely in the cytoplasm (11,27). Based on this scenario, C3 Stau as well as EDIN are released inside the cytoplasm and can immediately reach their targets RhoA, RhoE, and Rnd3. Under these conditions they in fact do not need any cell entry machinery.