Pseudomonas aeruginosa Exoenzyme S, a Double ADP-ribosyltransferase, Resembles Vertebrate Mono-ADP-ribosyltransferases*

Previous data indicated that Pseudomonas aeruginosa exoenzyme S (ExoS) ADP-ribosylated Ras at multiple sites. One site appeared to be Arg41, but the second site could not be localized. In this study, the sites of ADP-ribosylation of c-Ha-Ras by ExoS were directly determined. Under saturating conditions, ExoS ADP-ribosylated Ras to a stoichiometry of 2 mol of ADP-ribose incorporated per mol of Ras. Nucleotide occupancy did not influence the stoichiometry or velocity of ADP-ribosylation of Ras by ExoS. Edman degradation and mass spectrometry of V8 protease generated peptides of ADP-ribosylated Ras identified the sites of ADP-ribosylation to be Arg41 and Arg128. ExoS ADP-ribosylated the double mutant, RasR41K,R128K, to a stoichiometry of 1 mol of ADP-ribose incorporated per mol of Ras, which indicated that Ras possessed an alternative site of ADP-ribosylation. The alternative site of ADP-ribosylation on Ras was identified as Arg135, which was on the same α-helix as Arg128. Arg41 and Arg128 are located within two different secondary structure motifs, β-sheet and α-helix, respectively, and are spatially separated within the three-dimensional structure of Ras. The fact that ExoS could ADP-ribosylate a target protein at multiple sites, along with earlier observations that ExoS could ADP-ribosylate numerous target proteins, were properties that have been attributed to several vertebrate ADP-ribosyltransferases. This prompted a detailed alignment study which showed that the catalytic domain of ExoS possessed considerably more primary amino acid homology with the vertebrate mono-ADP-ribosyltransferases than the bacterial ADP-ribosyltransferases. These data are consistent with the hypothesis that ExoS may represent an evolutionary link between bacterial and vertebrate mono-ADP-ribosyltransferases.

Several vertebrate ADP-ribosyltransferases have been identified to date. The vertebrate ADP-ribosyltransferases possess specific properties that are distinct from the bacterial ADPribosyltransferases. Several of the vertebrate ADP-ribosyltransferases have the capacity to ADP-ribosylate multiple target proteins. For example, a murine lymphocyte transferase ADP-ribosylates a number of cell surface proteins (11), whereas rabbit skeletal muscle ADP-ribosyltransferase ADP-ribosylates several proteins in skeletal muscle T-tubules (12). In addition, several vertebrate ADP-ribosyltransferases modify target proteins at multiple sites. Both turkey erythrocyte type A ADP-ribosyltransferase (13) and chicken ADP-ribosyltransferase (14) ADP-ribosylate actin at two arginine residues. These sites of ADP-ribosylation are located in different regions of actin, which suggests that each ADP-ribosylation event is independent. Although the role of endogenous ADP-ribosylation in the regulation of eukaryotic cell physiology has not been defined in detail, the RT6 transferase appears to be involved in the regulation of T-lymphocyte function (15), and endogenous ADP-ribosylation may contribute to hippocampal long-term potentiation (16).
In this study, the ADP-ribosylation of Ras by Exoenzyme S was characterized. Using both mass spectrometry and rpHPLC coupled with Edman degradation, it was determined that ExoS ADP-ribosylates Ras at arginines 41 and 128. These residues are located in a ␤-sheet and ␣-helix, respectively, which are spatially separated in Ras. Thus, ExoS has functional similarities to the eukaryotic ADP-ribosyltransferases because ExoS ADP-ribosylates a number of target proteins in vitro and ADPribosylate Ras at two nonadjacent arginine residues. In addition, homology studies have identified more primary sequence homology between ExoS and the vertebrate ADP-ribosyltransferases than the bacterial ADP-ribosyltransferases. tagenesis kit from Amersham Pharmacia Biotech, bovine serum albumin from Pierce Biochemicals, DNA oligonucleotides from Operon, and GTP␥S from Sigma. Recombinant factor activating exoenzymes (FAS) and a c-Ha-Ras vector were gifts from H. Fu (Emory University).
Purification of His-tagged Ras Proteins-Recombinant His-tagged Ras proteins were expressed in Escherichia coli and purified by Ni 2ϩ affinity chromatography as described previously (10) with several modifications. His-tagged Ras proteins were eluted from the Ni 2ϩ affinity resin in elution buffer containing 3 M GTP and 10 mM MgCl 2. Eluted proteins were dialyzed into buffer (20 mM Tris-HCl (pH 7.6), 10 mM EDTA, 5 mM MgCl 2 , 1 mM dithiothreitol, and 10% glycerol) and stored at Ϫ70°C. These conditions were optimal for exchange of bound nucleotide from Ras proteins.
Nucleotide Loading of Ras Proteins-20 M Ras was incubated alone or with 1 mM of either GTP␥S or GDP for 30 min at 30°C. Reactions were stopped by the addition of 20 mM MgCl 2 . The stoichiometry of nucleotide loading of Ras was monitored radioanalytically, using ADP-ribosylation of Ras Proteins by ExoS-Reaction mixtures contained (25 l): 0.2 M sodium acetate (pH 6.0), 50 M [adenylate phosphate-32 P]NAD (specific activity 0.25 Ci per 1.25 nmol NAD), 5 M Ras or Ras⌬CAAX, FAS, and ⌬N222 (a catalytic, deletion peptide of ExoS). FAS and ⌬N222 were added in equivalent amounts. Reactions were stopped at the indicated times by spotting an aliquot of the reaction mixture onto trichloroacetic acid-saturated Whatman 3-mm paper. The paper was washed three times for 10 min each in 7.5% trichloroacetic acid, and radioactivity was quantitated by scintillation counting. Stoichiometry of ADP-ribosylation was determined as the moles of ADPribose incorporated/mole of Ras. Velocity of ADP-ribosylation was determined by linear regression analysis. Ras⌬CAAX is a deletion peptide of Ras where the four C-terminal amino acids have been deleted, which eliminates its capacity to be acylated, making it more amenable to biochemical manipulation.
Determination of the Sites of ADP-ribosylation within Ras-Ras⌬CAAX or Ras⌬CAAX-R41K,R128K (20 M) was incubated with 0.4 M ⌬N222, 400 M NAD, and 1.2 M FAS for 2 h at room temperature. Stoichiometry of ADP-ribosylated Ras was determined by subjecting an aliquot of the reaction mixture to SDS-polyacrylamide gel electrophoresis and subjecting the protein band corresponding to Ras to scintillation counting. Following ADP-ribosylation, the reaction mixture was dialyzed overnight into 10 mM Tris-HCl (pH 7.6), containing 20 mM NaCl, to remove unreacted NAD. Dialyzed samples were distributed into aliquots and lyophilized. Lyophilized samples were resuspended in 50 mM NH 4 Ac (pH 4.0), and Staphylococcus aureus V8 protease was added at a w:w ratio of 1:10 (protease:Ras). Samples were digested overnight, lyophilized, and suspended in 0.1% trifluoroacetic acid. The digested material was subjected to rpHPLC (C18), using a 5-40% acetonitrile gradient in 0.1% trifluoroacetic acid, and 1-min fractions were collected. Absorbance of eluted material was measured at 215 nm. Fractions were subjected to scintillation counting to identify peptides that were radiolabeled. Peptides in fractions containing radioactivity were subjected N-terminal amino acid sequencing. Protease-digested samples of ADPribosylated Ras and non-ADP-ribosylated Ras were also subjected to mass spectrometry as described previously (13).
Alignment of Protein Sequences-Protein sequences were aligned using the PILEUP program of GCG. Using an algorithm similar to that described in Ref. 17, this program creates a multiple sequence alignment.

RESULTS
ExoS ADP-ribosylates Ras at Arg-41 and Arg-128 -Previous studies indicated that ExoS ADP-ribosylated Ras at multiple sites. Although indirect, these studies suggested that ExoS ADP-ribosylated Ras at Arg 41 and at a second site. However, no specific second site of ADP-ribosylation could be resolved (10). In the present study, a direct approach was used to determine the sites of ADP-ribosylation within Ras by ExoS. Recombinant histidine-tagged Ras⌬CAAX was ADP-ribosylated by ExoS to a stoichiometry approaching 2 mol of ADP-ribose incorporated per mol of Ras. ADP-ribosylated Ras was digested with S. aureus V8 protease, and peptides were fractionated by rpHPLC. The absorbance of eluted peptides was measured at 215 nm, and elution fractions were assayed for radioactivity (Fig. 1A). Analysis of the V8 peptides of ADP-ribosylated Ras revealed three radioactive fractions. The first radioactive frac-tion corresponded to the void volume of the column (6 min), where NAD and free ADP-ribose eluted (data not shown). The two additional fractions that contained radioactivity eluted at 24 and 48 min, within the resolving region of the chromatogram.
The 24-and 48-min fractions were subjected to N-terminal amino acid sequencing. The first ten amino acids of the 24-min fraction contained a peptide corresponding to amino acids 38 -47 of Ras, which included Arg 41 (Table I). Amino acid yields at each cycle were similar (within about 2-fold), with the exception of cycle 4 where there was a reduction in the yield of the predicted amino acid, Arg 41 . Others have demonstrated that ADP-ribosylarginine is hydrolyzed by treatment with strong base (5). During the coupling of the free N terminus with phenyl isothiocyanate, the peptide was subjected to pH 9.0. When the products of cycle 4 of the Edman degradation were analyzed, no aberrant peak corresponding to ADP-ribosylarginine or hydrolyzed ADP-ribosylarginine was detected, and only minute amounts of free arginine were detected. The absence of hydrolytic products of ADP-ribosylarginine indicated that the ADP-riboseϪarginine bond was not cleaved during the Edman degradation procedure. Hydrophilic phenyl isothiocyanatemodified amino acids will not be extracted with butyl acetate from the glass fiber filter during Edman degradation as shown in the case of cysteine. 2 The fact that radioactivity was recovered on the glass fiber filter but not in the eluate of the Edman degradation is consistent with the idea that ADP-ribosylated arginine was not extracted from the glass fiber filter during cycle 4, and Arg 41 is one site of ADP-ribosylation. The first ten amino acids of the 48-min fraction contained a peptide corresponding to amino acids 127-136 of Ras, which included Arg 128 and Arg 135 (Table I). Amino acid yields at each cycle were similar (within about 2-fold), with the exception of cycle 2 where there was a reduction in the yield of the predicted amino acid, Arg 128 . This was consistent with Arg 128 having been ADPribosylated. The high yield of Arg at cycle 9 indicated that Arg 135 had not been ADP-ribosylated and indicated that the reduction in yield seen in cycle 2 of the 48-min fraction and cycle 4 of the 24-min fraction was not because of a generalized decrease in yield of arginine after Edman degradation.
The sites that ExoS ADP-ribosylated on wild-type Ras were also determined by mass spectrometry. Ras alone or following ADP-ribosylation by ExoS to a stoichiometry approaching 2 mol of ADP-ribose per mol of Ras was digested with trypsin and subjected to mass spectrometry. Mass spectra of peptides from ADP-ribosylated Ras possessed two unique peptides relative to peptides from non-ADP-ribosylated Ras. One peptide was identified which possessed a molecular mass of 3636 daltons, which corresponded to residues 17-42 of Ras plus 541 daltons, while a second peptide was identified which possessed a molecular mass of 1914 daltons, corresponding to residues 124 -135 of Ras plus 541 daltons. Addition of 541 daltons to each peptide was consistent with the addition of 1 ADP-ribose moiety to the peptide. These data supported the determination that ExoS ADP-ribosylates Ras at Arg 41 and Arg 128 .
Identification of Arg-135 as an Alternative Site of ADP-ribosylation of Ras by ExoS-Earlier data indicated that ExoS ADP-ribosylated Ras at multiple sites and that Ras possessed undefined alternative sites of ADP-ribosylation (10). To identify a possible alternative site of ADP-ribosylation within Ras, a double Ras mutant, Ras⌬CAAXR41K,R128K, was engineered by site-directed mutagenesis. Under saturation conditions, ExoS ADP-ribosylated Ras⌬CAAXR41K,R128K to a stoichiometry of approximately 1 mol of ADP-ribose/mol of Ras (Table II). This was consistent with the presence of an alternative site of ADP-ribosylation within the double mutant. Under linear velocity conditions, ExoS ADP-ribosylated wild type Ras⌬CAAX at a 4-fold greater velocity than Ras⌬CAAXR41K,R128K (Fig.  2). These data suggested the alternative site of ADP-ribosylation was not accessible for ADP-ribosylation upon the ADPribosylation of Arg 128 , possibly because of steric limitations. The identity of the alternative site of ADP-ribosylation within Ras⌬CAAXR41K,R128K was determined as described for wild type Ras⌬CAAX. rpHPLC analysis of peptides from V8 protease-digested ADP-ribosylated Ras⌬CAAXR41K,R128K showed the elution of two major fractions of radioactivity (Fig.  1B). The first fraction containing radioactivity eluted with the void volume of the column (6 min), corresponding to NAD or hydrolyzed ADP-ribose, whereas the second fraction containing radioactivity eluted at 46 min. N-terminal amino acid sequencing of the first eleven amino acids of the 46-min fraction iden-  A) or Ras⌬CAAX-R41K,R128K (panel B) was ADP-ribosylated by ⌬N222 ExoS to near completion. Following ADP-ribosylation, the protein was dialyzed to remove unreacted NAD and digested with S. aureus V8 protease. The resulting peptides were chromatographed on reversed-phase HPLC, and the absorbance of eluted material was measured at 215 nm (upper). Additionally, the absorbance of eluted material was measured at 259 nm to detect the presence of ADP-ribose (data not shown). Eluted fractions were assayed to detect radiolabeled peptides (lower). tified a peptide which corresponded to amino acids 127 to 137 of Ras, which included Arg 135 (note the presence of the R128K mutation at cycle 2) (Table I). Amino acid yields at each cycle were similar (within about 2-fold), with the exception of cycle 9 where there was a reduction of the yield of the predicted amino acid, Arg 135 . This was consistent with Arg 135 having been ADPribosylated. It should be noted that Ser is often recovered at lower yield during automated Edman degradation of peptides. 2 These data were consistent with Arg 135 being the alternative site for ADP-ribosylation of Ras by ExoS.
Influence of Nucleotide Occupancy on the Ability of ExoS to ADP-ribosylate Ras-Exoenzyme S ADP-ribosylates Ras at two distinct locations within the Ras molecule. One site of ADPribosylation, Arg 41 , lies in a ␤-sheet which is adjacent to the switch 1 domain of Ras, whereas the second and alternative sites of ADP-ribosylation, Arg 128 and Arg 135 , respectively, lie within an ␣-helix located toward the C terminus of Ras. The crystal structure of Ras (18) predicts that Arg 128 and Arg 135 lie on the external face of the ␣-helix and are located on a different surface of the Ras molecule, relative to Arg 41 (Fig. 3). Because both Arg 41 and Arg 128 are removed from the nucleotide binding site of Ras, nucleotide occupancy might not affect the ability of ExoS to ADP-ribosylate Ras. This was tested by loading Ras with GTP or GDP and measuring the stoichiometry and velocity of ADP-ribosylation relative to nucleotide-free Ras. Neither the stoichiometry nor the velocity of ADP-ribosylation of Ras was affected by the nucleotide occupancy (Table II). ExoS ADPribosylated either GTP-loaded, GDP-loaded, or nucleotide-free Ras to approximately 2 mol of ADP-ribose incorporated per mol of Ras. These results indicate that ExoS did not preferentially ADP-ribosylate either GTP-or GDP-bound Ras. Control experiments using [ 35 S]GTP␥S showed efficient nucleotide exchange conditions with the loading of nucleotide to a stoichiometry of approximately 1 mol of nucleotide per mol of Ras (data not shown). It should also be noted that Ras and Ras⌬CAAX showed the same velocities and stoichiometry as targets for ADP-ribosylation by ExoS. Thus, the presence of an acylation site does not influence the ADP-ribosylation properties of Ras.

Identification of Primary Amino Acid Homology between the Catalytic Domain of ExoS and Vertebrate ADP-ribosyltrans-
ferases-To date, several vertebrate ADP-ribosyltransferases have been identified, including rabbit skeletal muscle ADPribosyltransferase (RNART) (19), rat RT6 (20), a human ecto-ADP-ribosyltransferases (21), and chicken ADP-ribosyltransferase types I and II (22). Several of the vertebrate ADPribosyltransferases have the capacity to ADP-ribosylate several eukaryotic target proteins (11,12) and to ADP-ribosylate at two independent sites (13,14), which identified functional relationships with ExoS. Using the tFASTA algorithm, the vertebrate ADP-ribosyltransferases were observed to possess considerable primary amino acid homology with the catalytic portion of ExoS. ExoS possessed homologies of 26.6% with RNART, 28.3% with HNART, 31.9% with RT6, 35.4% with   CHAT 2b, and 25.7% with CHAT 1a. This alignment included the Ser-Thr-Ser motif and extended through the catalytic glutamic acids. Fig. 4 shows a PILEUP of the catalytic portion of ExoS and the vertebrate ADP-ribosyltransferases. These data suggest the possibility that the catalytic domain of ExoS shares both structural and functional properties with the vertebrate ADP-ribosyltransferases. In contrast, the alignments of the catalytic domain of ExoS with prokaryotic ADP-ribosyltransferases showed a similar degree of homology: 26.9% with cholera toxin, 28.9% with heat-labile enterotoxin of E. coli, and 23.8% with pertussis toxin. However, the homology centered on the Ser-Thr-Ser motif, but did not extend to the catalytic glutamic acids (data not shown). DISCUSSION Previous data indicated that exoenzyme S ADP-ribosylated Ras at multiple sites, including Arg 41 and a second site which could not be localized. The current study utilized direct biochemical and biophysical approaches to analyze how ExoS modified Ras. ADP-ribosylated Ras was digested with S. aureus V8 protease, and radiolabeled peptides were subjected to Edman degradation. During the Edman degradation reaction, radiolabeled ADP-ribosylated arginine was not recovered although a decrease in the yield of arginine at residues corresponding to Arg 41 and Arg 128 of Ras was detected. Others have demonstrated that radiolabeled ADP-ribosylarginine was not recovered when peptides ADP-ribosylated at arginine are subjected to Edman degradation although the site of ADP-ribosylation can be identified indirectly by a decrease in amino acid yield at a candidate arginine residue (14). The fact that arginines 41 and 128 were identified as the sites of ADP-ribosylation by two different methods, Edman degradation of ADPribosylated peptides and mass spectrometry, after digestion with two different proteases, V8 and trypsin, respectively, indicate that ExoS ADP-ribosylates Ras at arginines 41 and 128.
Analysis of the double mutant, RasR41K,R128K, revealed an alternative site for ADP-ribosylation by ExoS, Arg 135 . Because ExoS did not ADP-ribosylate Arg 135 in Ras, it appears that the ADP-ribosylation of Arg 128 sterically blocks the ADP-ribosylation of Arg 135 . Consistent with a steric mechanism of blockage, rather than an affinity effect, was the determination that ExoS ADP-ribosylated RasR41K,R128K at a rate that was within 4-fold of Ras. This indicates that ExoS can bind Ras in several orientations to facilitate ADP-ribosylation along that ␣-helix containing Arg 128 and Arg 135 and that ExoS can ADP-ribosylate Ras at Args on two distinct surfaces of Ras, the ␤-sheet containing Arg 41 and the ␣-helix containing Arg 128 and Arg 135 .
The plasticity of the ExoS-Ras interactions may explain the observed ability of ExoS to ADP-ribosylate numerous small molecular weight GTP binding proteins in vitro (7).
Coburn et al. (7) showed that ExoS ADP-ribosylates several members of the Ras superfamily in vitro. Only a limited subset of the family, Ras, Ral, and Rap, contained the Arg 41 homologue. Alignment of the ␣-helix, which contains the second site ADP-ribosylated by ExoS, Arg 128 , identified numerous members of the Ras superfamily that contained an arginine residue as a potential site of ADP-ribosylation. However, no distinct recognition motif is apparent from examination of the primary amino acid sequences of members of the Ras superfamily (Fig.  5). Consistent with this prediction, we have observed that ExoS ADP-ribosylates RhoA in vitro. 3 RhoA does not contain an Arg 41 homolog but does contain several Args within the secondary ␣-helix. The potential of ADP-ribosylating Ras superfamily members at their Arg 41 or Arg 128 homologues may have different functional implications. Arginine 41 is a contact residue in the Ras-Rap co-crystal structure (23) and is in a region of close contact in the Ras-SOS crystal structure (24). Modification of Ras at Arg 41 may inhibit Ras activation or Ras-Raf interactions. The secondary site helix has no ascribed function, although it is located adjacent to the Ras farnesylation site. Therefore, it is unclear how ADP-ribosylation of Ras at Arg 128 or 135 would affect Ras function, or alternatively ADP-ribosylation could affect Ras posttranslational modification. While ExoS has been shown to ADP-ribosylate Ras in vivo (9), a more detailed analysis of the in vivo targets of ExoS is needed.
Neither nucleotide occupancy nor Mg 2ϩ binding altered the 3 A. K. Ganesan and J. T. Barbieri, unpublished data.
FIG. 5. Alignment of exoenzyme S target proteins. Small molecular weight GTP binding proteins of the Ras superfamily that have been identified as targets for Exoenzyme S ADP-ribosylation in vitro were aligned using Pileup from GCG. The switch 1 domain of Ras (residues 32-37), the ␤-strand of Ras containing arginine 41, and the ␣-helix of Ras containing arginines 128 and 135 are depicted. Arginines within these regions in Ras and homologous proteins are shown in bold.

FIG. 4. Alignment of P. aeruginosa exoenzyme S with eukaryotic ADPribosyltransferases.
A portion of the catalytic domain of Exoenzyme S was aligned with the eukaryotic ADP-ribosyltransferases, using Pileup from GCG. RNART is the rabbit skeletal muscle NAD ADP-ribosyltransferase, HNART is the human NAD ADP-ribosyltransferase, CHAT 1 is the chicken ADP-ribosyltransferase type 1, CHAT 2 is the chicken ADPribosyltransferase type 2, and RT6 is the rat RT6 ADP-ribosyltransferase. Residues conserved between the eukaryotic transferases and Exoenzyme S are shown in bold. Region 2 of the ADP-ribosyltransferases corresponds to the STS sequence, whereas region 3 of the transferases corresponds to the catalytic glutamic acids. ability of ExoS to ADP-ribosylate Ras, as the velocity and stoichiometry of the ADP-ribosylation of GTP-loaded, GDPloaded, and nucleotide-free Ras were essentially identical. This indicated that the sites of ADP-ribosylation were not involved in nucleotide binding or chelation of Mg 2ϩ and was consistent with structural data which predicted that Arg 41 and Arg 128 were distant from these binding sites. In contrast to ExoS, several bacterial toxins that modify Ras superfamily members preferentially target GDP bound forms of those proteins; Clostridium sordellii LT preferentially glucosylates GDP bound Ras (25) and Clostridium botulinum C3 ADP-ribosyltransferase preferentially ADP-ribosylates GDP-bound Rho (26). The fact that ExoS modifies both the GTP-and GDP-bound forms of Ras whereas other toxins described to date modify only the GDP-bound form indicates that ExoS modifies Ras signal transduction by a mechanism that is distinct from the Clostridial toxins.
The endogenous vertebrate ADP-ribosyltransferases share several properties with ExoS, including the ability to ADPribosylate multiple target proteins and the ability to ADPribosylate target proteins at multiple sites. Sequence alignment showed that the catalytic domain of ExoS aligns more extensively with the eukaryotic ADP-ribosyltransferases than bacterial ADP-ribosyltransferases. The primary amino acid alignments of mono-ADP-ribosyltransferases predicts the conservation of a basic amino acid, a Ser-Thr-Ser sequence, and a catalytic glutamic acid (27). Alignment between the catalytic domain of ExoS and the vertebrate ADP-ribosyltransferases extended from the Ser-Thr-Ser sequence through the active site glutamic acid. In contrast, tFASTA alignment between the catalytic domain of ExoS and the bacterial ADP-ribosyltransferases, CT, LT, and PT, identified alignments only within the Ser-Thr-Ser sequence. This suggests that with respect to both functional and sequence alignments, ExoS is more similar to the vertebrate ADP-ribosyltransferases than the prokaryotic ADP-ribosyltransferases. Thus, ExoS may have an evolutionary link with the vertebrate ADP-ribosyltransferases. A better understanding of the mechanism by which ExoS modifies eukaryotic physiology may provide insight into the mechanisms of action of the vertebrate ADP-ribosyltransferases.