A Cell-permeable Fusion Toxin as a Tool to Study the Consequences of Actin-ADP-ribosylation Caused by the Salmonella enterica Virulence Factor SpvB in Intact Cells*

The virulence factor SpvB is a crucial component for the intracellular growth and infection process of Salmonella enterica. The SpvB protein mediates the ADP-ribosylation of actin in infected cells and is assumed to be delivered directly from the engulfed bacteria into the host cell cytosol. Here we used the binary Clostridium botulinum C2 toxin as a transport system for the catalytic domain of SpvB (C/SpvB) into the host cell cytosol. A recombinant fusion toxin composed of the enzymatically inactive N-terminal domain of C. botulinum C2 toxin (C2IN) and C/SpvB was cloned, expressed, and characterized in vitro and in intact cells. When added together with C2II, the C2IN-C/SpvB fusion toxin was efficiently delivered into the host cell cytosol and ADP-ribosylated actin in various cell lines. The cellular uptake of the fusion toxin requires translocation from acidic endosomes into the cytosol and is facilitated by Hsp90. The N- and C-terminal domains of SpvB are linked by 7 proline residues. To elucidate the function of this proline region, fusion toxins containing none, 5, 7, and 9 proline residues were constructed and analyzed. The existence of the proline residues was essential for the translocation of the fusion toxins into host cell cytosol and thereby determined their cytopathic efficiency. No differences concerning the mode of action of the C2IN-C/SpvB fusion toxin and the C2 toxin were obvious as both toxins induced depolymerization of actin filaments, resulting in cell rounding. The acute cellular responses following ADP-ribosylation of actin did not immediately induce cell death of J774.A1 macrophage-like cells.

Several pathogenic bacteria produce toxins and effector proteins, which attack the cytoskeleton of eukaryotic cells by mono-ADP-ribosylation of actin. In past years, we focused on the mode of action of the Clostridium botulinum C2 toxin as the prototype of binary actin-ADP-ribosylating toxins (1). The enzyme component of the C2 toxin (C2I) 3 ADP-ribosylates G-actin at arginine 177 (2). This leads to depolymerization of actin filaments and finally to cell rounding. The proteolytically activated binding/translocation component (C2IIa) forms heptamers, which assemble with C2I and bind to the cellular receptor (3). Following receptor-mediated endocytosis, C2IIa forms pores in the membrane of acidic endosomes. For translocation of the C2I protein through the lumen of these pores, a partial unfolding of C2I is required (4). The subsequent refolding of C2I in the cytosol is facilitated by the host cell chaperone Hsp90 (5). The interaction of C2I with C2IIa is mediated by the N-terminal domain of the C2I protein (C2IN, amino acid residues 1-225). C2IN, which is enzymatically inactive and does not induce cell rounding when applied in combination with C2IIa to cells, was successfully used as an adaptor for the C2IIa-mediated delivery of different proteins into the cytosol of eukaryotic cells (6).
The SpvB protein from Salmonella enterica was identified as a new member of bacterial actin-ADP-ribosylating enzymes (7). We identified arginine 177 as the modification site for C/SpvB within the actin homolog Act88F (Drosophila indirect flight muscle actin) (8). Our finding was confirmed by a recent publication (9) reporting that arginine 177 is the modification side in mammalian actin for SpvB. S. enterica is a Gram-negative facultative intracellular pathogen, which causes diseases ranging from mild gastroenteritis to severe systemic infections in humans (10,11). During infection Salmonella grows and replicates inside macrophages in a special membrane compartment, called the Salmonella-containing vacuole (SCV) (12).
As the final step in the infection process, Salmonella induces cell death of infected macrophages. Intracellular pathogenesis depends on the presence of the Salmonella pathogenicity island 2 (SPI2), which encodes for several effectors and for a type III secretion system (SPI2-encoded TTSS, for review see 13). A plasmid, which contains the Salmonella plasmid virulence (spv) gene cluster, is also essential for intracellular growth of Salmonella and therefore for the virulence (14,15). This plasmid comprises the four-gene operon spvABCD (16 -18), in which spvB encodes for the ADP-ribosyltransferase (7,19). It is supposed that SpvB is delivered into the cytoplasm via the SPI2-encoded TTSS (13,20). However, the mechanism underlying the translocation of the SpvB protein across the membrane of the Salmonella-containing vacuole is not known. The N-terminal domain of SpvB might be involved in translocation of SpvB, and the C-terminal catalytic domain (C/SpvB, amino acid residues 375-591) contains the ADP-ribosyltransferase activity (7,8). Between both domains a region of seven proline residues with a yet unknown function is present.
Biochemical studies on the mode of action of SpvB in intact cells are limited by the fact that SpvB is not taken up when applied to cultured cells. To characterize the cellular consequences of actin-ADP-ribosylation by C/SpvB, we used the C. botulinum C2 toxin as a delivery system for the catalytic domain of SpvB into the cytosol. We constructed a recombinant fusion toxin, consisting of the N-terminal adaptor domain of the C2I component (C2IN) and the catalytic domain of SpvB (C/SpvB), and analyzed the cellular effects caused by the fusion toxin. Finally, we identified the functional role of the polyproline stretch, located between the domains C2IN and C/SpvB in the fusion toxin.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotides were obtained from Hermann GbR (Freiburg, Germany). PCRs were performed with a T1 thermocycler from Biometra (Göttingen, Germany). DNA sequencing was done with an ABI PRISM 310 genetic analyzer from PerkinElmer Life Sciences. Taq polymerase was purchased from Roche Diagnostics, and Turbo Pfu polymerase was purchased from Stratagene (La Jolla, CA). Glutathione-Sepharose 4B was from Amersham Biosciences; cell culture medium was from Biochrom (Berlin, Germany); fetal calf serum was from PAN Systems (Aidenbach, Germany); and thrombin was from Sigma. ECL, Complete protease inhibitor mixture, annexin-V-FLUOS, trypsin, and trypsin inhibitor were from Roche Diagnostics. Alexa594/phalloidin, 4,6-diamidino-2-phenylindole, and ProLong Gold antifade reagent were obtained from Molecular Probes (Eugene, OR); ␤/␥-actin was from Cytoskeleton (Denver, CO). The 6-biotin-17-NAD was purchased from R&D Systems. The horseradish peroxidase-coupled anti-rabbit antibody was from Santa Cruz Biotechnology. The antiserum raised against C2IN was from a rabbit. The Cell-Titer 96 AQ ueous Non-radioactive Cell Proliferation Assay was from Promega (Mannheim, Germany).
Expression and Purification of Recombinant Proteins-The components of C. botulinum C2 toxin, C2I and C2II, were expressed as GST fusion proteins in E. coli BL21 cells. Proteins were purified as described previously (3,6,21,22) and incubated with thrombin (3.25 NIH units/ml of bead suspension) for cleavage of the GST domain. C2II was activated with trypsin for 30 min at 37°C (3). The C/SpvB protein was purified following the protocol of Otto et al. (7). However, after isopropyl 1-thio-␤-D-galactopyranoside induction the E. coli cells grew at 37°C, and no sterile filtration was done. The proteins C2IN-C/ SpvB(pro5), C2IN-C/SpvB(pro7), C2IN-C/SpvB(pro9), and C2IN-C/SpvB(⌬pro) were expressed as recombinant glutathione S-transferase fusion proteins in E. coli. Bacteria were grown at 37°C in 2 liters of LB medium, containing 100 g/ml ampicillin to an absorbance (600 nm) of 0.8. After addition of isopropyl 1-thio-␤-D-galactopyranoside the cultures were grown for 18 h at 29°C. The bacteria were sedimented at 5.000 ϫ g (10 min, 4°C) and resuspended in ice-cold lysis buffer (10 mM NaCl, 20 mM Tris-HCl (pH 7.4), 1% Triton X-100, and 1 mM phenylmethylsulfonyl fluoride). Following sonification, the bacterial debris was sedimented at 12.000 ϫ g (10 min, 4°C), resuspended, and sedimented again. The supernatant was added to a 50% slurry of glutathione-Sepharose 4B in PBS (400 l/1 liter) and incubated for 60 min at room temperature. After sedimentation at 3350 ϫ g (3 min at 4°C), the pellet was washed twice with ice-cold wash buffer (150 mM NaCl, 20 mM Tris-HCl (pH 7.4)) and two times with ice-cold PBS (8000 ϫ g, 3 min at 4°C). Thereafter, the beads were incubated with 500 l of PBS, containing thrombin (3.25 NIH units/ml) for 60 min at room temperature. For the elimination of thrombin, the beads were sedimented (8000 ϫ g, 1 min at 4°C), and the supernatant was incubated with a benzamidine bead suspension for 10 min at 4°C and sedimented at 8000 ϫ g for 1 min at 4°C. Afterward, an aliquot of the supernatant was subjected to a 12.5% SDS-PAGE, and the protein concentration was determined by densitometry by using Photoshop 7.0 software.
Cell Culture and Cytotoxicity Assays-Vero cells were cultivated in DMEM (Invitrogen) containing 10% heat-inactivated fetal calf serum, 1 mM sodium pyruvate, 2 mM L-glutamate, and 0.1 mM nonessential amino acids. HeLa cells were cultivated in minimum Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum and 2 mM L-glutamate. NIH 3T3 cells and J774.A1 macrophage-like cells were cultivated in DMEM (Invitrogen) containing 10% fetal calf serum, 4 mM L-glutamate, 1.5 g/liter sodium pyruvate, and 4.5 g/liter glucose. All cell lines were cultivated at 37°C and 5% CO 2 . Cells were trypsinized and reseeded for at most 15-20 times. For cytotoxicity experiments, cells were seeded in culture dishes and incubated with the respective toxin in complete medium.
Immunocytochemistry and Microinjection-Cells were seeded on coverslips and cultivated for at least 16 h. For actin staining, cells were washed with prewarmed PBS and fixed with 4% of paraformaldehyde for 20 min, washed twice with PBS, permeabilized with 0.2% (v/v) Triton X-100 in PBS, and washed again twice with PBS. Subsequently, cells were incubated with Alexa594-conjugated phalloidin (optional with 4,6-diamidino-2-phenylindole) for 30 min, washed with PBS, and embedded on glass slips with ProLong© Gold antifade reagent according to the manufacturer's manual. Cells were visualized by using a Zeiss Axiovert 200 (Oberkochen, Germany) with Cairn Optoscan monochromator system (Faversham, UK) and Photometrics CoolSNAP HQ CCD camera (Tucson, AZ). Images were processed by using Metamorph 6.1 software from Visitron (Munich, Germany). For microinjection, cells were seeded at a subconfluent density on glass coverslips (Cellocate; Eppendorf, Germany) and cultured overnight. Microinjection buffer (20 mM Tris-HCl (pH 7.5)), C2IN-C/SpvB(pro7) (50 nM in microinjection buffer), or C/SpvB (50 nM in microinjection buffer) was microinjected into Vero cells with an Eppendorf 5242 microinjector.
ADP-ribosylation Assay-The ADP-ribosylation reaction was performed for 30 min at 37°C in a buffer containing 50 mM Hepes buffer (pH 7.4), 2 mM MgCl 2 , 150 M [adenylate-32 P]NAD, 6 M of ␤/␥-actin, and 1 M of the enzymes C2I, C/SpvB, and C2IN-C/SpvB. The reaction was stopped by addition of SDS sample buffer and heating of the samples for 5 min at 95°C. Radiolabeled proteins were detected by SDS-PAGE and analyzed by PhosphorImaging. For ADP-ribosylation of cell lysates, 10 g of whole-cell lysate were used, and ADP-ribosylation was performed as described above with different amounts of toxins. The samples were run on SDS-PAGE, and [ 32 P]ADP-ribosylated proteins were detected by autoradiography with a PhosphorImager from Amersham Biosciences. For actin-ADP-ribosylation of cell lysates, cultured cells were lysed with 20 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1 mM dithiothreitol, 5 mM MgCl 2 , and Complete protease inhibitor (according to the manufacturer's manual). The complete cell protein samples (20 -50 g of protein) were incubated together with 10 M biotinylated NAD and 300 ng of C2I at 30°C for 30 min. The reaction was stopped by addition of 5ϫ SDS sample buffer (625 mM Tris-HCl (pH 6.8), 20% SDS, 8.5% glycerol, 0.2% bromphenol blue, 100 mM dithiothreitol) and subsequent heating of the samples for 5 min at 95°C. Biotin-ADP-ribosylated proteins were detected by immunoblot analysis with peroxidase-coupled streptavidin.
SDS-PAGE and Immunoblot Analysis-SDS-PAGE was performed according to the method of Laemmli (23). The proteins were stained with Coomassie Brilliant Blue R-250. For immunoblot analysis, the proteins were transferred onto a nitrocellulose membrane. The membrane was blocked for 30 min with 5% nonfat dry milk in PBS containing 0.05% Tween 20 (PBS-T), and the proteins were probed with anti-C2IN antiserum (rabbit, 1:5000 in PBS-T). After washing with PBS-T, the blot was incubated for 1 h with an anti-rabbit antibody coupled to horseradish peroxidase (1:2000 in PBS-T). The membrane was washed again, and proteins were visualized using the ECL system according to the manufacturer's instructions.
Assay for Toxin Binding to the Cell Surface-Confluently grown Vero cells were incubated in serum-free DMEM at 4°C with the toxins (C2IIa ϩ C2I; C2IIa ϩ fusion toxins) to allow the binding of the proteins to the cell surface. Before adding to the cells, the toxin components were preincubated on ice for 15 min. As a control, the single components of the binary toxins (C2IIa, C2I, fusion toxins) were taken. Following a washing step with serum-free DMEM, cells were lysed in RIPA buffer, and equal amounts of protein (50 g of each sample) were subjected to SDS-PAGE. The proteins were transferred from the gel onto a nitrocellulose membrane using the semi-dry system. The membranes were blocked overnight with 5% nonfat dry milk in PBS containing 0.05% Tween 20 (PBS-T) followed by a 1.5-h incubation with the antiserum (anti-C2IN, rabbit, 1:5000). After washing with PBS-T, the membrane was incubated with a donkey anti-rabbit antibody coupled to horseradish peroxidase (1:2.000 in PBS-T) for 1 h and washed, and the C2IN-containing proteins were detected using the ECL system.
Assay for Translocation of Toxins across the Cell Membrane-Subconfluently grown HeLa and Vero cells were treated in serum-free DMEM for 30 min at 37°C with 100 nM bafilomycin A1 to inhibit the physiological uptake pathway for C2 toxin. Subsequently, the cells were incubated for 20 min at 4°C with the toxins (C2I ϩ C2IIa; C2IN/C/SpvB ϩ C2IIa; preincubated for 15 min at 4°C to allow complex formation) and for 15 min at 4°C in the presence of bafilomycin to allow the binding of the toxins to the cell surface. The medium was removed, and warm acidic medium (37°C, pH 4.6, containing bafilomycin A1) was added to the cells for 5 min at 37°C. This acidic pulse mimics the endosomal conditions. Finally, the acidic medium was replaced by neutral medium (pH 7.5) containing bafilomycin A1, and cells were incubated at 37°C. The cytopathic effect of the toxin, i.e. cell rounding, was detected by taking pictures of the cells after various incubation periods.
Reproducibility of the Experiments and Statistics-All experiments were performed independently at least three times.

Cloning, Expression, and Biochemical Characterization of the C2IN-C/SpvB Fusion Toxin-
To deliver the catalytic domain of the ADP-ribosyltransferase SpvB (C/SpvB) into the cytosol of eukaryotic cells, we used a C2 toxin-based transport mechanism. To this end, a fusion toxin was cloned, consisting of amino acid residues 1-225 of the C2I enzyme component (C2IN, adaptor for the interaction with the C2IIa transport component) and the C/SpvB domain, encompassing residues 374 -591 of the SpvB protein (Fig. 1A). The wild type S. enterica SpvB protein harbors a region of seven-proline residues between the N-and C-terminal domains, according to the amino acid residues 367-373 in the SpvB protein. We included this seven-proline region into the C2IN-C/SpvB fusion toxin between the two domains C2IN and C/SpvB (C2IN-C/SpvB-(pro7) in the following referred to as C2IN-C/SpvB). To compare the fusion toxin with the C/SpvB domain and the C. botulinum C2I ADP-ribosyltransferase, the proteins C/SpvB, C2I, and C2IN-C/SpvB were expressed as recombinant glutathione S-transferase fusion proteins in E. coli cells and purified by affinity chromatography and subsequent cleavage of the GST domain with thrombin. The resulting proteins were analyzed by SDS-PAGE and by an immunoblot analysis with an antibody raised against the C2IN protein. Fig. 1B shows the Coomassiestained proteins C/SpvB (ϳ25 kDa), C2I (ϳ50 kDa), and the C2IN-C/SpvB fusion protein (ϳ51 kDa). The specific antibody against C2IN recognized C2I and the C2IN-C/SpvB fusion toxin but not C/SpvB (Fig. 1C). In preparations of the C2IN-C/ SpvB protein, a degradation product of ϳ26 kDa (Fig. 1, B and C) was also recognized by the anti-C2I antiserum, i.e. this protein was C2IN or contained a part of C2IN. Matrix-assisted laser desorption ionization time-of-flight analysis of the two degradation products confirmed that one protein contained the C2IN peptide and the other contained the C/SpvB peptide.
Next, we tested whether the C2IN-C/SpvB fusion toxin was enzymatically active. We performed an ADP-ribosylation assay with non-muscle (␤/␥) actin as a substrate, [ 32 P]NAD as a cosubstrate, and the mono-ADP-ribosyltransferases C/SpvB, C2I, and C2IN-C/SpvB as enzymes. The autoradiography of radio-labeled, i.e. ADP-ribosylated actin, is shown in Fig. 1D. This result demonstrates that the C2IN-C/SpvB fusion toxin was active in vitro. To test the activity of C2IN-C/SpvB in intact cells, the protein was microinjected into Vero cells. Directly and after the indicated incubation periods, pictures from identical cells were taken (Fig. 1E). Injection of control buffer as a control for the microinjection procedure did not influence the morphology of cells (Fig. 1E, left panel). When the C/SpvB protein as a positive control was injected, cells rounded up and finally detached from the coverslip (Fig. 1E, middle panel). Similarly, cells microinjected with C2IN-C/SpvB rounded up within a period of 60 min (Fig. 1E, right panel), indicating that the C2IN-C/SpvB fusion toxin destroyed the structure of the actin cytoskeleton in infected cells.
Cellular Uptake Mechanism of the C2IN-C/SpvB Fusion Toxin-Next, we tested whether the fusion toxin C2IN-C/SpvB was delivered into cells by the transport component C2IIa. To this end, subconfluently growing Vero cells were incubated with the C2IN-C/SpvB fusion toxin (1 g/ml) together with C2IIa (2 g/ml). As a positive control, the C. botulinum C2 toxin (C2I ϩ C2IIa) was used. As further controls, cells were incubated with the single components C2I (200 ng/ml), C2IN-C/SpvB (1 g/ml), and C2IIa (1 g/ml) and without any protein. Cells were incubated for 3 h at 37°C; pictures were taken, and the morphology of the cells was determined as an end point for the cytopathic activity of the individual protein combinations ( Fig. 2A). Only such cells, which have been treated with the combination of C2IIa plus C2IN-C/SpvB or C2IIa plus C2I, showed a round morphology. As expected, treatment of the cells with the single components C2IIa, C2I, or C2IN-C/SpvB did not induce any morphological alterations, indicating that for transport of the C2IN-C/SpvB fusion toxin into cells C2IIa is essentially needed. After a 3-h incubation period, the majority of the cells were rounded up ( Fig. 2A). This change in morphology was comparable with the rounding of cells, which have been treated with C2I and C2IIa.
To confirm that the morphological effects caused by the toxins were because of the ADP-ribosylation of actin within the cytosol of intact cells, we analyzed the state of actin-ADP-ribosylation from cells, which have been incubated with C2IIa plus C2IN-C/SpvB. Again, the combination C2IIa plus C2I was used as a control. Following incubation with the toxins, cells were lysed, and the cellular protein was subsequently incubated with biotin-labeled NAD ϩ and C2I for ADP-ribosylation of actin. Proteins were separated by SDS-PAGE, and ADP-ribosylated actin was detected by the use of streptavidin coupled to peroxidase and a subsequent chemiluminescence detection of the labeled proteins (Fig. 2B). The more actin already ADP-ribosylated in the intact cells by the action of the intracellular toxin, the less was labeled during the subsequent in vitro reaction, i.e. no actin signals were detectable from intoxicated cells. The ADP-ribosylation of actin from cell lysates confirmed that actin was already ADP-ribosylated in the intact cells by the C2IN-C/ SpvB protein during incubation of the cells with C2IN-C/SpvB plus C2IIa (Fig. 2B). Incubation of cells with the single components C2IIa and C2IN-C/SpvB resulted in a strong signal of biotin-labeled actin, indicating that actin was not ADP-ribosylated in intact cells. This result clearly demonstrates that cell rounding was because of the activity of the fusion toxin within the cytosol of treated cells.
The C2IIa-mediated translocation of C2I from acidic endosomes into the cytosol was blocked by the drug bafilomycin A1, which prevents acidification of the endosomal lumen. Cells, which have been pretreated with bafilomycin A1, were protected from the cytopathic effects and did not round up. As shown in Fig. 3A, bafilomycin A1 protected the cells from cytopathic effects of the C2IN-C/SpvB fusion toxin, indicating that the fusion toxin takes the same route into the cytosol as C2 toxin does.
For C2 toxin and C2 toxin-derived fusion toxins, it is known that inhibition of the host cell chaperone Hsp90 prevents the C2IIa-facilitated translocation of C2I from acidic endosomes into the cytosol. Again, the cytopathic effect of the C2IN-C/SpvB fusion toxin, i.e. the number of rounded cells, was reduced when cells were pretreated with the Hsp90 inhibitor radicicol prior to application of C2IIa plus C2IN-C/SpvB (Fig. 3B). This observation corroborates that the C2IN-C/SpvB fusion toxin is taken up into the cytosol of A, morphologic effects of the C2IN-C/SpvB fusion toxin. Vero cells were incubated with C2IN-C/SpvB (1 g/ml) plus C2IIa (2 g/ml) or with C2I (200 ng/ml) plus C2IIa (400 ng/ml). For control, cells were treated with the single toxin components in the same concentrations or without any protein (control). Following a 3-h incubation period at 37°C, pictures were taken. B, ADP-ribosylation of actin in lysates from toxin-treated cells. Control cells that have not been treated with any toxin, as well as cells treated with the single component C2IN-C/SpvB/SpvB (1 g/ml), with the combination of C2IN-C/SpvB (1 g/ml) plus C2IIa (2 g/ml), and with the combination of C2I (200 ng/ml) plus C2IIa (400 ng/ml), respectively, were lysed, and the samples (50 g of lysate protein each) were analyzed for ADP-ribosylation of actin by incubation with 10 M biotinylated-NAD and 300 ng of C2I protein. Following SDS-PAGE and transfer of the proteins onto a nitrocellulose membrane, the ADP-ribosylated actin was detected with peroxidase-labeled streptavidin and a subsequent detection via enhanced chemiluminescence.
target cells via the same Hsp90-dependent mechanism as the C2 toxin. Most likely, the translocation of the fusion toxin through the C2IIa pore requires at least a partial unfolding of the protein and a subsequent Hsp90-dependent refolding of the enzyme in the cytosol.
To demonstrate that the effect of C2IN-C/SpvB was not restricted to one specific cell type, several eukaryotic cell types were tested for their sensitivity toward the new fusion toxin. The cell lines HeLa, Vero, NIH 3T3, and Caco-2 were incubated at 37°C in the presence of C2IN-C/SpvB plus C2IIa, and following a 3-h incubation period, the cytopathic effect of the toxin was analyzed by monitoring alterations in the cell morphology. All the tested cell lines, which have been treated with C2IN-C/SpvB (1 g/ml) plus C2IIa (2 g/ml), showed cell rounding (Fig. 4A). In contrast, an incubation of cells with the single components C2I, C2IN-C/SpvB, and C2IIa did not change the morphology of the cells (data not shown). The cells were lysed, and a subsequent in vitro ADP-ribosylation of actin was performed. A strong signal of ADP-ribosylated actin was observed for the control cells, whereas all samples from toxintreated cells showed only a weak signal (Fig. 4A). In toxintreated cells, the actin was already modified in the intact cells by the toxin, clearly indicating that the ADP-ribosylation status of  the cellular actin correlated with the altered morphology of the cells because of the activity of the toxin in the cytosol. The time-dependent effects of the toxin combination C2IN-C/SpvB (1 g/ml) plus C2IIa (2 g/ml) on Vero cells and HeLa cells were analyzed by determination of round cells after toxin treatment (Fig. 4B).
Consequences of ADP-ribosylation of Actin for the Filament Structure-To study the effects of the toxin-induced ADP-ribosylation of G-actin on the morphology of the actin cytoskeleton in intact cells, we used transiently GFP-actin-transfected Ptk2 cells, because they are large cells with well defined actin filaments. Cells were treated either with C2I (200 ng/ml) plus C2IIa (400 ng/ml) or with C2IN-C/SpvB (1 g/ml) plus C2IIa (2 g/ml). Real time microscopy was performed to follow the effects on filamentous actin.
Interestingly, following treatment of cells with either the C2 toxin (C2I ϩ C2IIa) or with the fusion toxin C2IN-C/SpvB plus C2IIa, the depolymerization of stress fibers started in an unexpected way, beginning in the middle of the filament (Fig. 5, arrows) and continuing to both ends. However, both toxins showed no obvious difference in the mode of action in respect of the actin filaments. The corresponding movies are in the supplemental material.
Consequences of the ADP-ribosylation of Actin on the Viability of J774.A1 Macrophage-like Cells-We tested whether ADPribosylation of actin by C/SpvB and by the C2 toxin induced any cell death of the murine macrophage-like cell line J774.A1. These cells were sensitive toward C2IIa-mediated delivery of C2I and C2IN-C/SpvB fusion toxin into the cytosol, as the actin from toxin-treated cells was completely ADP-ribosylated (data not shown). Cells were incubated with the C2IN-C/SpvB fusion toxin plus C2IIa or with C2I plus C2IIa. As a positive control, we used the Clostridium difficile toxin B (TcdB), which induces apoptosis of J774.A1 cells by inactivation of Rho GTPases. 4 Following a 7-h incubation period with the toxins at 37°C, cell viability was determined with the CellTiter 96 AQ ueous Nonradioactive Cell Proliferation Assay. To this end cells were further incubated at 37°C in the presence of the toxins and the MTS reagent. Although TcdB reduced the number of viable cells to about 20%, no significant decrease of viable cells was detected after treatment of cells with each of the actin-ADP-ribosylating toxins (data not shown). This result indicates that within the time frame, in which the total amount of cellular actin was ADP-ribosylated by the toxins, cell death is not essentially and directly a consequence of actin-ADP-ribosylation in J774.A1 cells.
Role of the Proline Linker Region between the Domains C2IN and C/SpvB in the C2IN-C/SpvB Fusion Toxin-Finally, we studied the role of the seven proline residues located between the domains C2IN and C/SpvB in the C2IN-C/SpvB fusion toxin (see Fig. 1A). The catalytic domain of SpvB (C/SpvB), containing no polyproline region, is capable to ADP-ribosylate actin in vitro (see Fig. 1D). Moreover, the C/SpvB protein induced morphological alterations, i.e. cell rounding of Vero cells following microinjection (Fig. 1E, middle panel). Therefore, the presence of the polyproline sequence is not absolutely essential for the ADP-ribosyltransferase activity in vitro and in the cytosol of intact cells.
We constructed the following C2IN-C/SpvB fusion toxins with variations in the length of the proline region: C2IN-C/ SpvB(⌬pro) without any proline region; C2IN-C/SpvB(pro5) containing five proline residues; the C2IN-C/SpvB(pro7), which corresponds to the above characterized C2IN-C/SpvB and reflects the seven proline residues in the S. enterica SpvB protein; and C2IN-C/SpvB(pro9), which contains nine proline residues. The constructs were confirmed by DNA sequence analysis. The recombinant GST fusion proteins were purified as described above, and the resulting proteins were analyzed by SDS-PAGE and subsequent Coomassie staining (Fig. 6A). Moreover, all fusion toxins were detected by the specific antiserum raised against C2IN (Fig. 6B). The enzyme activities of the proteins were tested by ADP-ribosylation of non-muscle (␤/␥) actin with [ 32 P]NAD as a co-substrate. All tested fusion toxins showed a detectable enzyme activity (Fig. 6C). However, in presence of 5 nM of each fusion toxin, the kinetics of actin-ADP-ribosylation revealed that depending on the length of the proline linker the enzyme activity of the toxins decreased because C2IN-C/SpvB(pro9) exhibited the highest enzyme activity followed by C2IN-C/SpvB(pro7), C2IN-C/SpvB(pro5), and C2IN-C/SpvB(⌬pro). If high concentrations of each fusion toxin (i.e. 1 M) were tested in the actin-ADP-ribosylation assay, no differences in the in vitro enzyme activities were detected in our assay.
Next, the various fusion toxins were tested for their cytopathic activity toward intact cells. To this end, the toxins (400 ng/ml and 1.5 g/ml of each protein) were applied with C2IIa (800 ng/ml and 2 g/ml) to Vero cells. Pictures from the cells were taken after the indicated incubation periods, and the percentages of rounded cells were determined. In the presence of C2IIa, the effect of the fusion toxins in cells correlated with the existence and the length of the proline linker region between 4 K. Heine, unpublished observations. FIGURE 5. Effect of the C2IN-C/SpvB fusion toxin and the C2 toxin on the actin filaments of PtK2 cells. PtK2 cells were transfected with GFP-actin and incubated with C2IN-C/SpvB(pro7) (1 g/ml) ϩ C2IIa (2 g/ml) or alternatively with C2I (200 ng/ml) ϩ C2IIa (400 ng/ml). The dashed boxes mark the region of interest, where the destruction of actin filaments was followed. The initial effects were located in the middle of the filament structure (marked with arrows), and the depolymerization continued from the middle to the ends of the filaments. The cytopathic effect of the toxin was monitored over a period of 5 min. The corresponding movies are shown as supplemental data.
the C2IN and the C/SpvB domain of the protein (Fig. 6D). Within 6 h of incubation, the fusion toxin C2IN-C/SpvB(⌬pro) plus C2IIa induced only a marginal number of rounded cells, comparable with untreated cells. Interestingly, the fusion toxin C2IN-C/SpvB(pro9) showed an advanced cytopathic effect compared with the C2IN-C/SpvB(pro7) protein and with the C2IN-C/SpvB(pro5) protein, when the proteins were applied to cells together with C2IIa.
We tested whether the proline linker region had any influence on the transport of the fusion toxin by C2IIa, and we compared the two proteins C2IN-C/SpvB(⌬pro) and C2IN-C/ SpvB(pro7) for their ability to interact with C2IIa and to translocate into the host cell cytosol. The first step in the cellular uptake mechanism was the C2IIa-mediated binding of the fusion toxins to the cell surface receptor. This step was analyzed by detecting cell-bound toxin with a specific antiserum against the C2I protein. To prevent any endocytosis, cells were incubated for 20 min at 4°C with C2IIa (1 g/ml) together with 2 g/ml of the respective fusion toxin. To exclude nonspecific binding of the fusion toxins C2IN-C/SpvB(⌬pro) and C2IN-C/ SpvB(pro7) to the cell surface, cells were incubated with the fusion toxins in the absence of C2IIa. Cells were washed and lysed, and following SDS-PAGE with equal amounts of lysate protein, the cell-bound fusion toxins were detected by a subsequent immunoblot analysis with an antibody raised against C2IN. The result is shown in Fig. 7A. The fusion toxins C2IN-C/SpvB(⌬pro) and C2IN-C/SpvB(pro7) exclusively bound to the cells in the presence of C2IIa (Fig. 7A, left panel). Moreover, no difference in the amount of the cell-bound proteins C2IN-C/SpvB(⌬pro) and C2IN-C/SpvB(pro7) was observed. Even when cells were incubated with the fusion toxins plus C2IIa for 3 min, the amount of the cell-bound proteins C2IN-C/ SpvB(⌬pro) and C2IN-C/SpvB(pro7) was absolutely comparable (Fig. 7A, right panel), clearly indicating that the proline linker region did not influence the specific interaction of the fusion toxins with C2IIa.
To study the C2IIa-mediated translocation of the fusion toxins C2IN-C/SpvB(⌬pro) and C2IN-C/SpvB(pro7) into the cytosol, we performed an assay that mimics the situation of acidic endosomes on the surface of intact cells (3). Cells were pretreated with bafilomycin A1 to block endosomal acidification and thereby the physiological uptake of C2 toxin and the C2 toxin-derived fusion toxins (see Fig. 3A). Subsequently, cells were incubated at 4°C with C2IIa plus C2IN-C/SpvB(⌬pro) and C2IN-C/SpvB(pro7), respectively. As a positive control, C2I plus C2IIa was used. As further controls, the single components or no protein was applied to the cells. Following binding of the toxins to the cell surface, cells were subjected for 5 min to pre-warmed medium (37°C), containing bafilomycin A1 at pH 4.6 versus pH 7.5. The acidic conditions trigger membrane insertion and pore formation of C2IIa, which enables C2I or the respective fusion toxin to translocate through the C2IIa pore into the cytosol. After the short acidification, cells were incubated at 37°C at pH 7.5 in the presence of bafilomycin A1, and the numbers of rounded cells were determined. As shown in Fig. 7B, only cells that have been treated with either C2IIa plus C2I or C2IIa plus C2IN-C/SpvB(pro7) rounded up, even in the presence of bafilomycin A1 and after exposition to the acidic shift, although no cell rounding was observed for cells that have not been incubated in acidic medium or for cells treated with the single components C2I, C2IIa, C2IN-C/SpvB(pro7), or C2IN-C/SpvB(⌬pro), respectively (data not shown). Only few cells were rounded when C2IIa plus C2IN-C/SpvB(⌬pro) was applied. This finding indicates that the proline linker region is essential for the C2IIa-facilitated translocation of the fusion toxins across cell membranes under acidic conditions.

DISCUSSION
The pathogenesis of systemic Salmonella infections is characterized by survival and proliferation of bacteria inside macrophages, leading to cell death and to intercellular spread of the bacteria. To promote cellular uptake and intracellular survival, Salmonella has evolved sophisticated strategies to directly modify the host cell cytoskeleton. To this end, a variety of virulence factors is delivered into the cytosol of infected cells in a precisely spatiotemporal coordinated manner by two different type III secretion systems encoded on the pathogenicity islands SPI1 and SPI2. The intracellular survival and the replication of Salmonella in macrophages are promoted by the SPI2-mediated translocation of bacterial effectors into the host cell cytosol. The SPI2 effectors induce local actin polymerization, resulting in a focal actin meshwork around the SCV (24), which is important for intracellular replication of bacteria in the SCV. Some strains of Salmonella harbor the spv virulence cluster, which encodes for the ADP-ribosyltransferase SpvB and enhances the ability of Salmonella to induce systemic disease (15). SpvB ADP-ribosylates G-actin at arginine 177, resulting in depolymerization of actin filaments in host cells. In Salmonella-infected cells, SpvB modulates the vesicle-associated actin polymerization to achieve release of bacteria from SCV, and it was reported that SpvB contributes to cell death of infected macrophages (25). However, the precise consequence of actin-ADP-ribosylation by SpvB in intact cells and its contribution to cell death are not completely understood.
To study the effects of actin-ADP-ribosylation by SpvB in intact cells apart from other SPI2 effector proteins, the cellpermeable fusion toxin C2IN-C/SpvB was used. In microscopic time lapse experiments with GFP-actin-transfected cells, the depolymerization of actin filaments because of C2 toxin-and C2IN-C/SpvB-mediated ADP-ribosylation was analyzed. Interestingly, many filaments started to disrupt from their central region. Based on previous reports, it was expected that depolymerization starts at the end of the filaments because of the capping function of ADP-ribosylated actin (26). The disruption of actin filaments following treatment with actin-ADPribosylating toxins might be because of interference of ADPribosylated actin monomers in the treadmilling process of actin polymerization and depolymerization (for review see Ref. 27). A deranged actin polymerization cycle might be leading to an increased tension in the actin filaments resulting in their disruption and subsequent depolymerization. Recently, Stebbins and co-workers (9) reported on a steric antagonism of actin polymerization by SpvB. By crystallizing ADP-ribosylated and thereby polymerization-deficient actin, no dramatic conformational changes in actin were found. Based on their study, they draw the conclusion that polymerization of ADP-ribosylated actin is inhibited most likely through a steric disruption of intrafilament contact sites. It requires further investigation to FIGURE 7. Role of the proline residues in the fusion toxin for the C2IIamediated receptor binding and translocation into the cytosol. A, detection of the cell-bound fusion toxins C2IN-C/SpvB(⌬pro) and C2IN-C/SpvB-(pro7). Vero cells were incubated at 4°C for 20 min with C2IN-C/SpvB(⌬pro) or with C2IN-C/SpvB(pro7) (each 1 g/ml) together with C2IIa (2 g/ml). For control, cells were left untreated. To demonstrate the specific C2IIa-mediated binding of the fusion toxins to cells, cells were incubated with C2IN-C/ SpvB(⌬pro) and C2IN-C/SpvB(pro7) in the absence of C2IIa. Cells were washed and lysed, and following SDS-PAGE, the proteins were blotted onto nitrocellulose. The fusion toxins were detected with an anti-C2IN antibody and a secondary anti-rabbit antibody coupled to horseradish peroxidase. Note that binding of the fusion toxins to cells is specific and dependent on C2IIa (left panel). Right panel, detection of the amount of cell-bound fusion toxins C2IN-C/SpvB(⌬pro) and C2IN-C/SpvB(pro7) after 3 and 20 min. The conditions of the experiment were precisely as described above. The fusion toxins were applied to cells only in combination with C2IIa. B, C2IIa-mediated translocation of the toxins into the cytosol. HeLa cells were pretreated for 1 h at 37°C with bafilomycin A1 (100 nM final concentration) to block endosomal acidification. Subsequently, cells were incubated at 4°C with 2 g/ml C2IN-C/SpvB(⌬pro) plus 1 g/ml C2IIa or with 2 g/ml C2IN-C/SpvB(pro7) plus 1 g/ml C2IIa, respectively. As a positive control, cells were treated with C2IIa (1 g/ml) plus C2I (0.5 g/ml). As a negative control, cells were treated without any toxin. The cells were subjected for 5 min to an acidic pulse (medium at 37°C; pH 4.6, containing bafilomycin A1). The cells were further incubated at 37°C at pH 7.5 in the presence of bafilomycin A1, and pictures were taken after 90 min. test whether this model is the underlying mechanism for the toxin-induced effects on actin filaments observed in intact cells.
We used the C2IN-C/SpvB fusion toxin and the C. botulinum C2 toxin to separate the consequences of actin-ADP-ribosylation, e.g. the depolymerization of actin filaments from the induction of cell death. According to the postulated model of Salmonella infection, the intracellular secretion of effectors by the SPI2-TTSS induces a delayed type of apoptosis in infected macrophages (25), and SpvB was proposed to contribute to this phenomenon. Interestingly, it was reported earlier that C2 toxin did not immediately induce cell death following ADP-ribosylation of actin in various cell types (28 -32). When macrophages were challenged with the C2IN-C/SpvB fusion toxin, ADP-ribosylation of actin did not trigger cell death, even when the total amount of cellular actin was ADP-ribosylated by the toxin. 4 However, this observation does not exclude that actin-ADP-ribosylation might result either directly or indirectly in delayed macrophage cell death. In our experiments the fusion toxin harbors only the catalytic domain of SpvB; therefore, we cannot exclude that the N-terminal domain of SpvB may contribute either directly or indirectly to cell death by modulating C/SpvB to exhibit any cytopathic activity. To further investigate the role of SpvB in cell death of macrophages, we will construct fusion toxins containing C2IN and either the N-terminal domain of SpvB or the full-length SpvB protein.
As an advantage of the fusion toxin, the consequences of actin-ADP-ribosylation can be studied in intact cells without overlapping effects from intracellular growing Salmonella. On the other hand, the artificial delivery of an isolated domain of a virulence factor into the cytosol most likely does not reflect the spatiotemporal distribution of the ADP-ribosyltransferase in Salmonella-infected cells. In infected cells, SpvB is delivered from SCV-located Salmonella into the cytosol, most likely via the SPI2-TTSS. Depending on the amount of translocated SpvB, a more local distribution of SpvB around the SCV might be possible, initially resulting in an actin-ADP-ribosylation around the vacuoles, accompanied by a vacuole-associated actin depolymerization. On the other hand, it was shown that the complete amount of actin was ADP-ribosylated in Salmonella-infected cells.
The translocation mechanism of proteins through the C2IIa pore and through a Salmonella type III needle complex might be comparable. For translocation through the C2IIa pore, C2I and C2I-derived fusion toxins require at least a partial unfolding (4). Refolding of the proteins in the cytosol is facilitated by the host cell chaperone Hsp90 (5). Here we report that the cytopathic mode of action of the C2IN-C/SpvB fusion toxin also requires Hsp90 activity, suggesting that this protein becomes unfolded during C2IIa-mediated translocation and is refolded into its active conformation in the cytosol. If SpvB translocates through a type III needle complex from intracellularly growing Salmonella into the host cell cytosol and as the inner diameter of the C2IIa pore (2.7 nm) and of the type III needle complex are similar (33)(34)(35), unfolding for passage through the needle and subsequent refolding might be also important during the physiological translocation of SpvB.
Finally, we analyzed the role of the polyproline region located in the fusion toxin for its functional role. The full-length SpvB protein also harbors a seven-proline residue region between the N-and C-terminal domains. We demonstrated recently that the proline region has no significant effects on the ADP-ribosyltransferase activity of the catalytic domain of SpvB (8). Moreover, we can exclude that the proline region is compulsive for the ADP-ribosylation of actin by C/SpvB in intact cells, because microinjection of C/SpvB without the proline residues also led to depolymerization of actin filaments. When the conditions of C2IIa-mediated toxin translocation from acidic endosomes were experimentally mimicked on the surface of intact cells, the cytopathic effects of the C2IN-C/SpvB fusion toxins revealed a crucial influence of the presence and the length of the proline region for translocation of the fusion toxins through the C2IIa pore. The proline stretch between the two domains may enhance the flexibility of the fusion toxin and facilitate its unfolding and subsequent refolding, as the presence of proline residues results in special type helices (36). If the SpvB protein actually translocates via the SPI2-TTSS in Salmonella-infected macrophages, it would be interesting to study whether the seven proline residues between the two domains of SpvB are essentially involved in the intracellular secretion of SpvB into the host cell cytosol.