Aeromonas Exoenzyme T of Aeromonas salmonicida Is a Bifunctional Protein That Targets the Host Cytoskeleton^*

Aeromonas salmonicida subsp. salmonicida (A. salmonicida) is the causative agent of furunculosis, a systemic disease that affects salmonid fish (salmon, trout, and char). A. salmonicida expresses a variety of extracellular toxins, many of which have been implicated in virulence. Several of these factors, including the pore-forming toxin aerolysin, serine protease, and the phospholipase GCAT, are secreted into the environment via the well characterized type II or general secretory pathway (1, 2). However, like many Gram-negative pathogens, A. salmonicida also possesses a type III protein secretion system (3-5). Such secretion systems translocate toxins, or type III effector proteins, directly into the cytosol of eukaryotic cells. Here the effector proteins are able to modulate cell signaling pathways or, alternatively, disrupt the dynamics of the cytoskeleton (6, 7).

To date, the following genes encoding four type III effector proteins have been identified in A. salmonicida: aexT, aopP, aopO, and aopH (8-10). Although functional studies of AopO and AopH have not yet been carried out, sequence homology with type III effector proteins YopO and YopH of Yersinia sp. suggests that AopO and AopH modify the actin cytoskeleton of target eukaryotic cells. In contrast, AopP has been shown to interfere with the NF-κB signaling pathway by inhibiting translocation of NF-κB into the nucleus of target cells (10).

AexT, the first type III effector of A. salmonicida to be characterized, has been shown to function as an ADP-ribosylating toxin (8). However, its substrate has never before been identified. AexT also displays significant sequence homology to the bifunctional type III effector proteins ExoS and ExoT, expressed by Pseudomonas aeruginosa. Both ExoS and ExoT are known ADP-ribosylating toxins (11) but also function as GTPase-activating proteins (GAP)² (12, 13).

Here we report that AexT is also a bifunctional protein possessing both ADP-ribosylating and GAP activities. We identify actin as the only specific target of the APD-ribosylating activity of AexT, and we show that AexT also exhibits GAP activity against the monomeric GTPases Rho, Rac, and Cdc42. Both enzymatic activities are involved in AexT-dependent actin filament depolymerization in eukaryotic cells.

EXPERIMENTAL PROCEDURES

Bacterial Strains—Cloning was routinely carried out in Escherichia coli strain XL1-Blue (14). E. coli strain BL21(DE3) (15) was used for the expression of recombinant AexT proteins, and strain S17-1 (16) was used as the donor strain in bacterial conjugation. The wild type (WT) isolate A. salmonicida strain JF2267 (8) used in this study was routinely cultured on Luria-Bertani agar plates at 18 °C.

Site-directed Mutagenesis—The WT aexT gene was amplified from A. salmonicida strain JF2267 by PCR using primers aexTHIS-forward and aexTHIS-reverse (Table 1) or aexTGFP-forward and aexTGFP-reverse (Table 1) before being cloned into the vector pGEM-T Easy (Promega). Site-directed mutagenesis via the overlap extension-PCR method (17) was then carried out using mutagenesis primers as indicated in Table 1. Resulting PCR products were digested with DpnI to eliminate the methylated (parental) DNA (18) and subsequently used to transform E. coli XL1-Blue. Mutations were verified by DNA sequencing before cloning into the expression vectors pETHIS-1 (Novagen) and pEGFP-N2 (BD Biosciences).

Overexpression of AexT and AexT Mutants—Vector pETHIS-1, containing either WT or mutated aexT genes, was transformed into E. coli BL21(DE3) cells for expression. Induction and subsequent purification of recombinant proteins were performed as described previously (19).

ADP-ribosylation Assay—In vitro ADP-ribosylation assays were carried out in 50 mm triethanolamine buffer (pH 7.5) containing 5 mm MgCl₂, 10 mm dithiothreitol, 10 mm thymidine, and 5 × 10⁵ cpm [³²P]NAD (specific activity 800 Ci mmol^-1; PerkinElmer Life Sciences). Lysates of EPC or Vero cells (20 μg), muscular or nonmuscular actin (1 or 4.5 μg), Rho-GST or Rac-GST fusion proteins (1 μg), and recombinant WT AexT (10 μm) were added to a total volume of 20 μl. After incubation at 37 °C for 30 min, samples were subjected to SDS-PAGE and processed by autoradiography.

Kinetic analysis was carried out as described previously (20) using 0.5 μm AexT. NAD concentrations were 5, 3.3, 1, 0.33, 0.1, and 0.03 μm. ADP-ribosylating activity constants (V_max and K_m) were determined using Lineweaver-Burk plots. Determination of ADP-ribosylating activity of iota toxin from Clostridium perfringens was included for comparison.

Peptide Analysis by HPLC, Sequencing, and Mass Spectrometry—Nonmuscular actin (6 μg) was incubated with AexT (10 μm) and [³²P]NAD (1 μm) or cold NAD (10 mm) for 1 h at 37 °C. The actin was then subjected to SDS-PAGE and stained with Amido Black. Actin bands were cut from the gel, washed with distilled H₂O, dried, and suspended in 150 μl of 50 mm Tris-HCl (pH 8.6), containing 0.01% Tween 20 and 0.2 μg of trypsin (Promega). Following 18 h of incubation at 30 °C, the preparations were centrifuged, and supernatants were recovered. Samples were injected onto a DEAE-HPLC column (Interchim Hema Bio 1000 DEAE, 33 × 1 mm) linked to a C18 reverse phase HPLC column (Interchim Uptisphere ODBD, 150 × 1 mm) and eluted with a 2-70% acetonitrile gradient in 0.1% trifluoroacetic acid. Radioactive material in fractions from actin ADP-ribosylated with [³²P]NAD was screened by blotting 3 μl of each fraction onto a polyvinylidene difluoride membrane and analyzing with a PhosphorImager. Protein sequencing was performed on an Applied Biosystem 494 sequencer.

Separated peptides (1-2 μl on a gold ProteinChip or H4 ProteinChip array) were identified by their mass using SELDI-TOF-MS (PCS 4000) from Bio-Rad (focus mass = 4000 daltons; laser intensity = 1100 nJ). The matrix (0.8 μl) was α-cyano-4-hydroxycinnamic acid saturated in 50% acetonitrile, 0.5% trifluoroacetic acid.

GAP Activity Assay—GAP activity was determined as described previously (21). The assays were performed three times, each time in duplicate.

Immunofluorescent Staining of the Actin Cytoskeleton—Epithelioma papulosum cyprinid (EPC) cells were transiently transfected with 2 μg of plasmid DNA using the PolyFect reagent (Qiagen) according to the manufacturer's instructions. Cells were incubated overnight at 18 °C. Eighteen hours post-transfection, cells were fixed with 4% formaldehyde and then incubated in 50 mm NH₄Cl for 15 min to block free aldehydes remaining after fixation. Cells were permeabilized with 0.1% Triton X-100 for 10 min, followed by incubation for 1 h at room temperature with TRITC-phalloidin (Sigma). Stained cells were observed using a Nikon Eclipse 80i microscope, and images were produced by digital confocal imaging using Open-lab software (Improvision). All transfections were carried out twice, and a minimum of 20 transfected cells was examined for each mutant.

Marker Replacement Mutagenesis—The aexT gene and flanking regions were amplified from A. salmonicida strain JF2267 using primers aexTREcoRI and aexTLSacI (Table 1) and cloned into vector pUC19 (Invitrogen). The kanamycin cassette from plasmid pSSVI186 (22) was excised, and blunt ends were created using S1 nuclease (Promega) before ligation into the ApaI and StuI cut sites (blunt-ended with S1 nuclease) within aexT. The inactivated aexT gene was then cloned into the mobilizable suicide vector pSUP202sac (23), and the resulting plasmid, pSUP202sac-aexT::K_m was transformed into E. coli S17-1. Conjugation into A. salmonicida strain JF2267 was carried out by filter mating (16). Double-crossover mutants were selected for directly by growth on tryptic soy agar containing 15% (w/v) sucrose, 40 μg of kanamycin ml^-1, and 20 μg of chloramphenicol ml^-1 at 15 °C for 7 days. Chloramphenicol was used to select against the E. coli donor as A. salmonicida strain JF2267 is resistant to this antibiotic. PCR was then used to ensure the correct mutation was present in A. salmonicida.

Fish Cell Infections with A. salmonicida—Infection of EPC cells was performed as described previously (24) using a multiplicity of infection of 20:1 (bacteria to fish cells). Two hours post-infection, the cells were photographed under phase-contrast microscopy (Axiovert 100; Zeiss).

RESULTS

AexT ADP-ribosylates Cellular Actin—To identify potential targets of the ADP-ribosylating activity for AexT, an in vitro ADP-ribosylating assay was performed using lysates of both EPC cells, a fish cell line derived from carp epithelium, and Vero cells. In each cell line, WT AexT ADP ribosylated only one protein of ∼45 kDa (Fig. 1A, lanes 2 and 3), corresponding to the size of cellular actin (Fig. 1A, lane 1). When the assay was performed using AexT alone, no ADP-ribosylation was detected indicating AexT does not undergo auto-ribosylation (Fig. 1A, lane 5). As a further control, we also carried out this assay using purified actin and iota toxin, a well characterized ADP-ribosylating toxin of C. perfringens that targets all actin isoforms (25, 26, 28). The result with this toxin was the expected one (Fig. 1A, lane 4), thereby adding strength to our finding that actin is the substrate of the ADP-ribosylating activity for AexT.

To further verify our results, a second assay was performed using both muscular (bovine) and nonmuscular (human platelet) actin. The GTPases Rho and Rac were also included as potential targets as these proteins are ADP-ribosylated by the AexT homologue ExoS of P. aeruginosa and are also the substrate of various other toxins such as exoenyzme C3 of Clostridium botulinum and large clostridial toxins (33). The results indicate that AexT exhibits specific ADP-ribosylating activity against actin, with the reaction against nonmuscular actin being significantly stronger than that against muscular actin (Fig. 1B, lanes 1 and 2). No AexT-dependent ADP-ribosyltransferase activity could be detected against Rho and Rac (Fig. 1B, lanes 3 and 4), although ADP-ribosylation of Rho by C3, which served as a positive control, was detected (Fig. 1B, lane 5).

To quantitate the difference seen between ADP-ribosylation of muscular and nonmuscular actin, a kinetic assay was carried out using both targets. The results, shown in Table 2, confirmed our finding that ADP-ribosylating activity of AexT is stronger against nonmuscular actin. A further comparison between AexT and iota toxin indicated that ADP-ribosylating activity of AexT is lower than that of iota toxin for both isoforms of actin tested (Table 2).

In an effort to identify the actin residue or residues that are ADP-ribosylated by AexT, nonmuscular actin was incubated with AexT in the presence of [³²P]NAD. Trypsin digestion was then carried out, and the resulting peptides were separated by HPLC. Radioactivity was primarily recovered in fraction 51 with lesser amounts recovered in three fractions immediately adjacent (Fig. 2A). This indicated that a specific peptide contained the ADP-ribosylated residue.

To obtain a higher yield of actin ADP-ribosylated with AexT, the reaction was then carried out with 10 mm nonradioactive NAD. The peptide profile of actin modified by AexT and non-radioactive NAD was similar to that of unmodified actin; however, peak 51 was slightly enlarged in the actin ADP-ribosylated by AexT (data not shown). Peaks 51 from both ADP-ribosylated and unmodified actin were analyzed by N-terminal sequencing and mass spectrometry. Several peptides were identified in both fractions beginning with the sequences FRCPEA, LCY-VAL, and CPEALF. An additional peptide beginning with the sequence TTGIVM was identified only in AexT-treated actin. Mass spectrometry by two different arrays showed that a 4391-Da peptide was present in peak 51 of AexT-treated actin but not in the corresponding peak of the control actin (Fig. 2, B and C). The size of this peptide corresponds to the sequence TTGIVMDSGDGVTHTVPIYEGYALPHAILRLDLAGR (where the underlined R residue is found at position 177) with an oxidized Met and an ADP-ribose and Na⁺ molecule (4389.33 Da). A small amount of the same peptide with the nonoxidized Met and an ADP-ribose and Na⁺ molecule (4373.34 Da) was also recovered. Modification of Arg-177 by ADP-ribosylation thereby preventing trypsin cleavage at this residue explains the presence of this peptide in peak 51 from ADP-ribosylated actin but not in the equivalent peak from control actin. Taken together, these results indicate that Arg-177 is mono-ADP-ribosylated by AexT.

Glutamic Acid Residues Glu-398 and Glu-401 and Arg-303 Are Involved in ADP-ribosylating Activity of AexT—ADP-ribosylating toxins retain a conserved structure that forms the enzymatic site. This region accommodates a molecule of NAD and possesses a catalytic glutamic acid residue essential for the enzymatic activity (27). The active site of certain toxins, including actin-ADP-ribosylating toxins such as iota toxin, contain not one but two glutamic acid residues whereby both residues are required for the ADP-ribosylation activity (27, 28). The amino acid sequence of AexT possesses such a biglutamic acid motif within its C-terminal domain (Fig. 3A) (8). According to sequence alignment, the two glutamic acid residues within the motif of AexT are found at positions 401 and 403. To determine the influence of these residues on the ADP-ribosylating activity of AexT, mutants containing glutamic acid to alanine substitutions were generated. The ADP-ribosylating activity of the recombinant mutants was then assayed in vitro using nonmuscular actin as a substrate. A single mutant, whereby the second of the two glutamic acid residues in the ADP-ribosylating domain of AexT was mutated to alanine, AexTE403A, displayed only a slight decrease in ADP-ribosyltransferase activity when compared with the WT protein (Fig. 3B, lane 2). A similar result was obtained with a double mutant whereby both glutamic acid residues were substituted, AexTE401A/E403A (Fig. 3B, lane 3). Examination of the amino acid sequence of AexT surrounding the biglutamic acid site then identified a third glutamic acid upstream of Glu-401 at position 398 (Fig. 3A). Two more AexT mutants, AexTE398A/E403A and AexTE398A/E401A, were therefore constructed. Although AexTE398A/E403A still displayed significant ADP-ribosylating activity (Fig. 3B, lane 4), the activity of mutant AexTE398A/E401A was virtually abolished (Fig. 3B, lane 5). A triple mutant, AexTE398A/E401A/E403A, containing glutamic acid substitutions in all three positions, was also found to possess virtually no ADP-ribosylating activity (Fig. 3B, lane 6). The results indicate that the two catalytic glutamic acids of AexT are Glu-398 and Glu-401. It is interesting to note that these residues are separated by two amino acids, instead of only one as in other ADP-ribosylating toxins possessing a biglutamic motif (Fig. 3A) (27).

ADP-ribosylating toxins can be divided into two groups according to conserved residues in the catalytic site. The cholera toxin group of ADP-ribosylating toxins, which includes iota toxin and other ADP-ribosylating toxins that target actin, contains a conserved arginine residue that is required to maintain the stability of the active site pocket. In the diphtheria toxin group, a histidine residue rather than an arginine carries out this function (27). Sequence alignment of AexT with cholera toxin suggests that Arg-303 or Arg-306 in AexT could be the conserved, functional arginine residue of the cholera toxin group (Fig. 4A). Two AexT mutants, whereby these arginine residues were mutated to alanine, were constructed, and their ADP-ribosylating activity was compared with that of WT AexT. Although the mutant R306A retained full enzymatic activity, AexT carrying the mutation R303A displayed an almost complete loss of ADP-ribosylating activity (Fig. 4B). This result indicates that AexT retains the main features of the cholera-like toxins.

AexT Alters the Actin Cytoskeleton and Mutation of the Catalytic ADP-ribosylating Residues Does Not Prevent AexT-dependent Actin Depolymerization—Eukaryotic cell transfections, carried out using aexT constructs cloned in-frame with enhanced green fluorescent protein (EGFP), and immunofluorescence microscopy were performed to assay the effect of AexT expression on EPC cells. Fluorescence microscopy showed that cells transfected with the pEGFP-N2 vector only displayed a typical morphology and a well defined actin cytoskeleton (Fig. 5A). In contrast, all cells transfected with WT pAexT-EGFP had a rounded appearance and no longer displayed actin stress fibers (Fig. 5B). The AexT-dependent changes in the actin cytoskeleton are reminiscent of the effects induced by actin-ADP-ribosylating toxins. When the mutants AexTE398A/E401A and AexTE398A/E401A/E403A, both of which have significantly reduced ADP-ribosylating activity, were expressed in EPC cells, significant actin depolymerization was still observed in at least 18 of 20 transfected cells examined (90%), resulting in the presence of actin clouds and a rounded cell morphology (Fig. 5, C and D). This indicates that WT AexT induces disruption of actin filaments and cell rounding and that mutations within the ADP-ribosylating site do not prevent this effect. Therefore, an additional activity is likely involved in the actin filament disruption caused by AexT.

AexT Exhibits GAP Activity toward GTPases Rho, Rac, and Cdc42—Sequence alignment of AexT with type III effector proteins that are known to act as GAP proteins identified a potential RhoGAP consensus motif (29) in the N-terminal domain of the protein (Fig. 6A). The RhoGAP motif contains an arginine residue that is known to be essential for efficient catalysis (30), and in AexT, this residue is found at position 143 (Fig. 6A).

The results of an in vitro assay of GTP hydrolysis performed with recombinant WT AexT indicated that almost 100% of the [³²P]GTP bound to Rac, Cdc42, and Rho was hydrolyzed following 5 min of incubation at room temperature (Fig. 6, B-D). In each case, the GAP activity of WT AexT was similar to that of RhoGAP, which served as a positive control. In contrast, an AexT mutant whereby the arginine residue at position 143 was substituted with a lysine residue, AexTR143K, showed a marked decrease in GTP hydrolysis. Five min following the addition of this mutant, ∼60% of bound [³²P]GTP was hydrolyzed by Rac and Cdc42 (Fig. 6, B and C), and ∼30% was hydrolyzed by Rho (Fig. 6D). Similar results were obtained with the negative control where no GAP protein was present. These results demonstrate that AexT possesses GAP activity toward all three GTPases tested and indicate that the arginine residue found within the RhoGAP consensus motif is required for this activity.

Aext-dependent Actin Cytoskeleton Alteration Is Mediated by Both Actin-ADP-ribosylation and GAP Activity toward Rho-GTPases—The effect of the GAP domain of AexT on actin depolymerization in EPC cells was also assayed by cell transfection, using plasmid pEGFP-N2 expressing AexTR143K. After 18 h of incubation, at least 60% of fish cells transfected with this construct displayed an altered actin cytoskeleton and a rounded morphology (Fig. 7A) similar to that seen in cells transfected with WT AexT (Fig. 5B). Following 22 h of incubation, at least 80% of transfected cells displayed these morphological changes. This indicates that, like the loss of ADP-ribosylating activity, loss of GAP activity alone does not prevent actin depolymerization by AexT.

Finally, transfection with a quadruple mutant, AexTR143K/E398A/E401A/E403A, whereby the essential residues within the GAP and the ADP-ribosylating domains were mutated, was performed. At least 90% of examined cells transfected with this mutant displayed a well defined cytoskeleton, and no cell rounding was observed (Fig. 7B). This suggests that both the GAP and ADP-ribosylating domains of AexT have the ability to affect the actin cytoskeleton independently of one another. Similar results were obtained when the mammalian cell line, COS-7, was used in the transfection assay (data not shown).

AexT Contributes to the Cytopathology of A. salmonicida—The cytopathic effect of AexT was assessed by infecting cultured EPC cells with WT A. salmonicida strain JF2267 and an isogenic ΔaexT mutant containing a deletion of both the GAP and ADP-ribosylating domains (Fig. 8A). Although infection with WT A. salmonicida strain JF2267 induced cell rounding and detachment within 2 h post-infection (Fig. 8B), EPC cells infected with the isogenic ΔaexT mutant displayed no marked morphological changes (Fig. 8C). These cells were similar in appearance to cells incubated with phosphate-buffered saline only (Fig. 8D). After 3 h of incubation, EPC cells infected with the ΔaexT mutant began to display cell rounding, and after 4 h, these cells were similar in appearance to those infected with the WT strain (data not shown). The results indicate that although AexT induces a cytopathic effect on target fish cells, other proteins expressed by A. salmonicida also contribute to the cytopathic effect of the bacterium.

DISCUSSION

AexT was found to be a bifunctional protein with two distinct enzymatic activities, ADP-ribosylating activity and GAP activity. We have shown that, in vitro, AexT ADP-ribosylates both muscular and nonmuscular actin with the reaction against non-muscular actin being stronger. AexT is therefore a member of the large group of ADP-ribosylating toxins that target actin. Other members of this group include the clostridial binary toxins (31-33), vegetative insecticidal proteins from Bacillus cereus and Bacillus thuringiensis (32, 34), the Streptococcus pyogenes toxin SpyA (35), as well as the potential type III effector molecule SpvB of Salmonella enterica (36, 37). Like the ADP-ribosylating effector SpvB (37) and toxins such as the C2 toxin of C. botulinum (32), AexT preferentially modifies nonmuscular actin versus muscular actin.

Another marked similarity found between AexT, SpvB, and C2 toxin is the site of modification on the actin molecule. AexT was found to mono-ADP-ribosylate actin at residue Arg-177. This residue is located at the contact site between actin monomers. ADP-ribosylation of monomeric actin prevents contact with other actin monomers thereby preventing polymerization (27). Arg-177 has also been shown to the modification site of SpvB and C2 toxin as well as of iota toxin (26, 38, 39).

Despite a low level of sequence homology, ADP-ribosylating toxins possess a conserved structure that forms the site of ADP-ribosylation. This active site is formed by a β-strand followed by an α-helix bend together creating a cavity in which the nicotinamide ring of NAD is anchored during catalysis (27). In all ADP-ribosylating toxins, the catalytic residue is a glutamic acid. This residue forms a hydrogen bond with the 2′-hydroxyl group on the nicotinamide ribose of NAD, which mediates the transferase reaction. Toxins of the cholera toxin group have another glutamic acid or glutamine residue located two amino acids upstream of the catalytic glutamic acid. C3 has the QXE sequence, whereas toxins from the iota and C2 families, cholera toxin, as well as other ADP-ribosyltransferases possess an EXE motif (27). Type III-secreted ADP-ribosylating toxins such as ExoS, ExoT, and SpvB also retain an EXE motif (37, 40). Detailed functional studies of iota toxin have shown that both of the glutamic acid residues in the EXE motif are required for ADP-ribosyltransferase activity (28, 41). An EXE motif has been identified previously in AexT and, by sequence alignment, was found to encompass residues Glu-401 and Glu-403 (Fig. 3A) (8). Surprisingly, however, this study has found that recombinant AexT mutants containing glutamic acid to alanine substitutions of these residues, AexTE403A and AexTE401A/E403A, still retained significant ADP-ribosylating activity (Fig. 3B). This is in marked contrast to clostridial binary toxins and the type III secreted ADP-ribosyltransferases ExoS, ExoT, and SpvB whereby mutation of the glutamic acid residues within the EXE motif abolishes ADP-ribosylation activity and subsequent cell effects (32, 42, 43). For example, mutation of the second glutamic acid within the EXE motif of ExoS and within the S. pyogenes toxin SpyA is sufficient to eliminate ADP-ribosylating activity (35, 44). In the case of the C. perfringens iota toxin, mutation of either of the two glutamic acid residues within the EXE site is sufficient to abolish the activity of the protein (28). Still another study has found that a double mutant of SpvB, whereby both glutamic acid residues had been replaced by aspartates, was no longer active (37).

In AexT, we have found an additional glutamic acid residue located directly upstream of the EXE motif, at position 398 (Fig. 3A). In contrast to the AexT mutants AexTE401A/E403A and AexTE398A/E403A, which retain significant ADP-ribosyl-transferase activity, the mutant AexTE398A/E401A displayed almost no ADP-ribosylating activity (Fig. 3B). This finding suggests that the key glutamic acid residues in AexT are Glu-398 and Glu-401, although this is not consistent with the position of the EXE motif. Interestingly, ExoT also contains a glutamic acid two amino acids upstream of the EXE motif at a position equivalent to Glu-398 in AexT (Fig. 3A). Yet substitution of the EXE motif of ExoT by AAA results in loss of cell rounding (42). Another study has found that mutation of the EXE motif to DXD results in a specific loss of ADP-ribosylating activity (45). Therefore, despite the sequence homology of the catalytic site, AexT and ExoT do not require equivalent glutamic acid residues for their enzymatic activity. Structural analysis of these proteins, in particular of the catalytic site, would help to solve why AexT does not appear to conform to the functional EXE consensus sequence.

Unlike secreted toxins, including actin-ADP-ribosylating toxins, many type III effector proteins have a bifunctional activity (33). Among them, the AexT homologues ExoS and ExoT display ADP-ribosylation and GAP activities. Like ExoS and ExoT, AexT was shown, in vitro, to possesses GAP activity for monomeric GTPases of the Rho family, RhoA, Rac1, and Cdc42 (Fig. 6). During the preparation of this study, this result was confirmed by an independent study, which expressed only the N-terminal region of AexT (amino acids 93-255) (46). In contrast to ExoS, which ADP-ribosylates numerous cell proteins (47-49), and to ExoT, which targets Crk proteins (45), AexT specifically ADP-ribosylates cellular actin. Indeed, the bifunctional activity of AexT is specifically directed against the actin cytoskeleton, including modification of actin monomers and down-regulation of Rho proteins leading to actin filament depolymerization.

Fish cells transfected with a plasmid expressing WT AexT displayed a distinct, rounded morphology, and the actin filaments were clearly disrupted. This observation is consistent with both the ADP-ribosylating and GAP activities of AexT. It is therefore not surprising that despite the strong GAP activity of AexT, transfection of EPC cells with a plasmid expressing AexTR143K, the mutant that no longer possesses GAP activity (Fig. 6), did not prevent the cytopathic effect of AexT (Fig. 7A). This finding is in contrast to studies of the type III effector YopE where expression of a corresponding mutant (R144A) completely inhibited the disruption of the microfilament structure of HeLa cells (50). However, YopE, which is shorter in length than AexT, does not possess an ADP-ribosylating domain. In contrast, a single mutation of R149K in the bifunctional protein ExoT did not abolish the cytopathic effect of this toxin. Disruption of the actin cytoskeleton was still visible when HeLa cells were transfected with ExoTR149K indicating that the ADP-ribosylating domain of ExoT also contributes to the biological functions of this protein (42). Our observation that expression of AexTR143K within transfected EPC cells causes an altered cytoskeleton indicates that the ADP-ribosylating domain of AexT also contributes to the activity of this protein.

Transfection of EPC cells with plasmids expressing mutants that no longer possess ADP-ribosylating activity, the double mutant AexTE398A/E401A and triple mutant AexTE398A/E401A/E403A, also resulted in significant alteration of the actin cytoskeleton and a rounded appearance similar to that seen when cells were transfected with the WT protein (Fig. 5, B-D). Together with the transfection results of the GAP mutant, these results indicate that both the GAP and ADP-ribosylating activities of AexT are able to affect the cytoskeleton independently of one another. Accordingly, no disruption of the actin cytoskeleton was seen when EPC cells were transfected with a quadruple mutant, AexTR143K/E398A/E401A/E403A, which is devoid of both GAP and ADP-ribosylating activities (Fig. 7B).

When epithelial fish cells were infected with the WT A. salmonicida isolate, strain JF2267, the cells displayed extensive cell rounding and retraction 2 h post-infection (Fig. 8B). In contrast, cells inoculated with an isogenic aexT mutant did not display any morphological changes (Fig. 8C). This is consistent with our findings that AexT affects the eukaryotic cytoskeleton. However, after a longer incubation time, cells infected with the ΔaexT mutant also began to display a rounded appearance. We should note that these findings are inconsistent with a previous study by Braun et al. (8), which indicated a ΔaexT mutant had no toxic effect on RTG-2 cells even after prolonged incubation. It has since become clear that the type III secretion genes of A. salmonicida strain JF2267, the strain used in both the previous study and this study, are easily lost during laboratory cultivation (51). It is therefore possible that the long term effect reported by Braun et al. (8) was because of a loss of the type III secretion system itself and not by the mutation in aexT. The ΔaexT mutant created in this study was analyzed by PCR to ensure the strain still maintained the type III secretion genes that are responsible for the translocation of AexT into the eukaryotic cytosol.

The observation that the ΔaexT mutant has a cytopathic effect after prolonged incubation indicates that other virulence factors expressed by A. salmonicida also contribute to the cytopathogenicity of the bacterium. This is supported by a recent report whereby an aexT mutant displayed only a subtle reduction in virulence, in vivo, when compared with the parent strain (9). As previous studies have shown that the cytopathic effect of A. salmonicida strain JF2267 is dependent upon a functional type III secretion system (3, 24), these other factors must be type III effector proteins. This conclusion is also supported by two independent in vivo studies carried out in rainbow trout (5) and Atlantic salmon (9), both of which have shown that inactivation of the type III secretion system attenuates A. salmonicida virulence.

AexT homologues ExoS, ExoT, and YopE are primarily involved in antiphagocytosis permitting Pseudomonas and Yersinia, respectively, to escape host defense (7, 40). The functional role of the double activity on the actin cytoskeleton of AexT in the pathogenesis of Aeromonas is not yet known. However, recent results have shown that a functional type III secretion system, responsible for the translocation of AexT into target fish cells, prevents uptake of A. salmonicida by peripheral blood leukocytes (4). We speculate that AexT-dependent actin depolymerization plays a role in this process.

In conclusion, the AexT produced by A. salmonicida displays bifunctional enzymatic activity, ADP-ribosylation of cellular actin, and GAP activity toward Rho-GTPases. This is the first bacterial toxin known to disrupt actin filaments in target cells via these two specific targets.