The Salmonella spvB virulence gene encodes an enzyme that ADP-ribosylates actin and destabilizes the cytoskeleton of eukaryotic cells

Authors

  • Marc L. Lesnick,

    1. Department of Medicine 0640, School of Medicine, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093, USA.
    Search for more papers by this author
  • Neil E. Reiner,

    1. Department of Medicine, Division of Infectious Diseases, and the Department of Microbiology and Immunology, University of British Columbia, Vancouver, British Columbia, Canada.
    Search for more papers by this author
  • Joshua Fierer,

    1. Department of Medicine 0640, School of Medicine, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093, USA.
    2. Department of Pathology, School of Medicine, University of California at San Diego, La Jolla, CA 92093, USA.
    3. Department of Veterans Affairs Medical Center,
      San Diego, CA 92161, USA.
    Search for more papers by this author
  • Donald G. Guiney

    Corresponding author
    1. Department of Medicine 0640, School of Medicine, University of California at San Diego, 9500 Gilman Drive, La Jolla, CA 92093, USA.
    • *For correspondence. E-mail dguiney@ucsd.edu; Tel. (+1) 858 534 6030; Fax (+1) 858 534 6020.

    Search for more papers by this author

Abstract

ADP-ribosylating enzymes, such as cholera and diphtheria toxins, are key virulence factors for a variety of extracellular bacterial pathogens but have not been implicated previously during intracellular pathogenesis. Salmonella strains are capable of invading epithelial cells and localizing in macrophages during infection. The spvB virulence gene of Salmonella is required for human macrophage cytotoxicity in vitro and for enhancing intracellular bacterial proliferation during infection. Here, we present evidence that spvB encodes an ADP-ribosylating enzyme that uses actin as a substrate and depolymerizes actin filaments when expressed in CHO cells. Furthermore, site-directed mutagenesis demonstrates that the ADP-ribosylating activity of SpvB is essential for Salmonella virulence in mice. As spvB is expressed by Salmonella strains after invasion of epithelial cells or phagocytosis by macrophages, these results suggest that SpvB functions as an intracellular ADP-ribosylating toxin critical for the pathogenesis of Salmonella infections.

Introduction

Salmonella strains produce a broad spectrum of infections ranging from asymptomatic colonization through enteritis to fatal systemic disease. Although host factors play an important role in the outcome of Salmonella infection, genetic variation between strains is a critical determinant of clinical manifestations (J. Fierer and D. G. Guiney, submitted). Salmonella infection is initiated in the intestine, with dissemination occurring through the lymphatics and bloodstream. Progressive infection of systemic organs by non-typhoid Salmonella serovars is facilitated by the spv genetic locus. During disseminated infection, virulent Salmonella localize within macrophages of the spleen and liver, proliferate and induce cytotoxicity (Richter-Dahlfors et al., 1997). Evidence obtained in vivo suggests that the spv genes act to enhance bacterial replication in macrophages (Fields et al., 1986; Gulig and Doyle, 1993). Recently, studies of Salmonella infection in bovine monocyte-derived macrophages have indicated that the spv genes are required for bacterial growth (Libby et al., 1997). Furthermore, expression of the spv genes is required for Salmonella strains to induce cytotoxicity in infected human monocyte-derived macrophages, characterized by cell detachment and eventual apoptosis (Libby et al., 2000).

The spv genes are located on large virulence plasmids found in non-typhoid Salmonella serovars of subspecies I that are commonly associated with severe systemic infection (Gulig et al., 1993; Guiney et al., 1995). Chromosomal spv loci have also been found in several other lineages of Salmonella subspecies found in cold-blooded vertebrates (Boyd and Hartl, 1998). The plasmid-encoded spv locus consists of a four-gene operon (spvABCD) that is positively controlled by the upstream transcriptional regulatory protein SpvR and requires the alternative sigma factor RpoS for efficient expression (Fang et al., 1991; 1992). Both rpoS and spv expression are dramatically upregulated after phagocytosis of Salmonella strains by macrophages (Fierer et al., 1993; Chen et al., 1996). Mutational analysis has demonstrated that the spvB gene, encoding a 65.6 kDa protein, is essential for the virulence phenotype (Roudier et al., 1992). Database searches reveal that SpvB is likely to have two functional domains based on homologies to two different classes of proteins (Bowen et al., 1998; Otto et al., 2000). The N-terminus of SpvB has 47% homology at the amino acid level to TcaC from Photorhabdus luminescens, part of a secreted multicomponent insecticidal toxin with an unknown mechanism of action. The C-terminus of SpvB has 25% homology at the amino acid level to Vip2 from Bacillus cereus, a member of the large family of bacterial ADP-ribosylating enzymes (bAREs) (reviewed by Pizza et al., 1999). These toxins are expressed by a broad range of extracellular toxigenic bacteria and act by covalently transferring an ADP-ribose moiety from NAD to an essential cell component or regulatory protein, irreversibly altering the activity of the target. Preliminary evidence indicates that SpvB also possesses ADP-ribosylating activity, as predicted by the sequence homology (Otto et al., 2000).

Results

The C-terminus of SpvB has homology to ADP-ribosylating enzymes

Figure 1 shows the proposed division of SpvB into two functional domains. The homology to the insect toxin TcaC ends immediately before a string of nine proline residues, which probably provides a flexible bridge between the two domains. The C-terminal domain contains critical conserved residues for the NAD binding site required for ADP-ribosylating activity: two active-site glutamates, an upstream serine–threonine–serine motif and a tyrosine–arginine pair further upstream (not shown). In other ADP-ribosylating enzymes, these residues are folded into a pocket that binds NAD and forms the catalytic site for ADP-ribosylation (Pizza et al., 1999).

Figure 1.

The SpvB protein has homology to two types of bacterial toxins.

A. The N-terminus of SpvB (hatched) has homology to a protein toxin from Photorhabdus luminescens. This region is followed by a string of 8–10 prolines (black) that separates the N-terminal region from the C-terminal region, which has homology (grey) to bacterial ADP-ribosylating enzymes (bAREs). The bars beneath the protein denote the regions of SpvB present in the N- and C-terminal polypeptides used in CHO transfection experiments.

B. A comparison of SpvB with the active site of four ADP-ribosylating enzymes: Vip2 from Bacillus cereus; Iota toxin from Clostridium perfringens; and C2 toxin component I and C3 toxin from Clostridium botulinum, using the clustalw program, where | denotes identity, * denotes amino acids in common among three of the proteins, : denotes homology, and ◊ denotes the critical glutamate residues mutated in the SpvBmut1 protein. The underline highlights a conserved region that is part of the putative catalytic site.

SpvB ADP-ribosylates actin

In order to confirm that SpvB has ADP-ribosylation activity, the full-length protein was overexpressed in Escherichia coli, purified and incubated in the presence of a protein extract from cultured CHO cells and radiolabelled NAD. Purified SpvB ADP-ribosylated a protein of ≈ 45 kDa (Fig. 2A) present in the CHO cell lysate. Similar results were obtained with HeLa cell extract (data not shown). Three of the bAREs that are most homologous to SpvB, Vip2, Iota toxin and C2 toxin have been shown to use actin (43 kDa) as a substrate (Aktories et al., 1986; Vandekerckhove et al., 1987; Han et al., 1999), suggesting that this might be the substrate for SpvB as well. To test this possibility, we incubated SpvB with purified actin and found that SpvB was capable of ADP-ribosylating actin (Fig. 2A). To demonstrate that native SpvB produced in Salmonella has a similar activity to the recombinant protein expressed in E. coli, we tested an extract of Salmonella made from bacteria grown under conditions that maximize spv gene induction (Fang et al., 1991). These extracts ADP-ribosylated actin as well (Fig. 2B). Further, extracts from Salmonella with an insertion in the spvB gene that abolished its expression (Roudier et al., 1992) were unable to ribosylate actin (Fig. 2B). These findings demonstrate that SpvB can ADP-ribosylate actin and that there are no other proteins expressed by Salmonella under the conditions examined that have significant actin ADP-ribosylating activity.

Figure 2.

The SpvB protein ADP-ribosylates actin.

A. Purified SpvB was mixed with a protein extract from CHO cells or with purified rabbit muscle actin and radiolabelled NAD, incubated for 60 min, subjected to PAGE and then exposed to film. The arrows denote a band dependent on the presence of the SpvB protein that appears at the molecular weight of actin. The higher molecular weight band appearing in both CHO extract lanes is apparently the substrate of a eukaryotic ADP-ribosylating enzyme and is not dependent on the presence of SpvB.

B. Protein extracts from three different strains of Salmonella dublin Lane were assayed as above with CHO cell extract. These strains include S. dublin pCR4, which contains the entire spv locus on a low-copy-number plasmid and overexpresses the SpvB protein, S. dublin Lane wild type and S. dublin D44, which has an insertion into the spvB gene in the context of the pCR4 plasmid (Roudier et al., 1992).

C. SpvB was incubated either without actin or with equal amounts of actin purified from muscle (rabbit), non-muscle (platelets) or cardiac muscle (porcine) obtained from Cytoskeleton Inc. and assayed as above.

To define SpvB's substrate specificity further, the ability of the protein to ribosylate actin from three different sources, cardiac muscle, striated muscle and non-muscle (platelet) cells, was examined. Actin from these different sources contains varying proportions of the three actin isotypes, alpha, gamma and beta, each of which varies chiefly in its N-terminal amino acid composition. bAREs that ADP-ribosylate actin can show distinct preferences for certain isotypes of actin (Pizza et al., 1999). This was found to be the case with SpvB as well, as non-muscle actin (80% beta-isotype, 20% gamma-isotype) was ribosylated to much greater levels than were the other actins tested (Fig. 2C). This is consistent with the observation that Salmonella generally infect epithelial cells or macrophages, both of which contain non-muscle actin isotypes.

SpvB disrupts actin microfilaments when transiently transfected into CHO cells

Ribosylation of actin by other bAREs has been shown to disrupt actin polymerization, causing a breakdown of filamentous actin throughout the cell (Reuner et al., 1987). As the addition of purified active SpvB had no effect on eukaryotic cells, we assayed the intracellular activity of SpvB using transient transfection. The spvB gene was cloned into the pTRACER vector, which is designed to express a cloned gene under the direction of the CMV promoter and co-express green fluorescent protein (GFP) using an independent constitutive promoter. When transiently transfected into eukaryotic cells, the presence of GFP identifies transfected cells expressing the protein of interest. CHO K1 cells transfected with pTRACER expressing either full-length SpvB or the C-terminus of SpvB contained no detectable F-actin (Fig. 3B and D), although they readily express GFP, suggesting that they remain viable (Pederson and Barbieri, 1998). In contrast, cells transfected with the empty vector control or the vector expressing the N-terminus of the SpvB protein showed a normal distribution of F-actin (Fig. 3A and C). Further, monolayers transfected with vectors to express the full-length SpvB or its C-terminus appeared to have a large number of rounded and detached cells (data not shown). We have also observed loss of F-actin staining in human monocyte-derived macrophages infected with wild-type Salmonella dublin Lane. Twenty-four hours after infection, > 37% of macrophages lack detectable F-actin, compared with < 1% in cells infected with an spvR mutant that abolishes spvB expression (S. Browne and D. Guiney, unpublished observation). Together, these findings show that the SpvB protein is capable of disrupting polymerized actin, thereby contributing to intracellular pathogenesis, and demonstrate that the C-terminus of the protein is necessary and sufficient for this activity.

Figure 3.

Expression of the spvB gene in CHO cells causes disruption of filamentous actin. CHO K1 cells were transfected using lipofectamine reagent (Gibco BRL) with the pTRACER vector that co-expresses GFP and the cloned gene to be tested. Transfected monolayers were stained to detect the presence of filamentous actin using rhodamine-conjugated phalloidin, and individual transfected cells were identified by GFP fluorescence. CHO cells were transfected with (A) the pTRACER vector alone; (B) pTRACER expressing the full-length SpvB protein; (C) pTRACER expressing the N-terminus of SpvB (amino acids 1–367); or (D) pTRACER expressing the C-terminus of SpvB (amino acids 376–594). A single field was photographed separately to capture GFP fluorescence (GFP column) or rhodamine–phalloidin fluorescence (phalloidin column), and the images were merged (merged column) to determine whether phalloidin staining co-localized with GFP fluorescence.

The ADP-ribosylating activity of SpvB is crucial for Salmonella virulence in mice

Salmonella strains lacking the spv locus are dramatically attenuated in their ability to cause disease in the mouse (Gulig et al., 1993; Guiney et al., 1995). To assess further the contribution of the actin ADP-ribosylating activity of SpvB to spv-mediated pathogenesis in vivo, we constructed a mutation in the putative active site of SpvB by substituting aspartates for the two conserved glutamate residues (E538D and E540D) that have been shown to be essential for NAD binding and enzyme activity in other bAREs (Pizza et al., 1999). This altered SpvB protein (SpvBmut1) was completely defective in its ability to ribosylate non-muscle actin in the in vitro assay system described above (Fig. 4), indicating that these glutamate residues are essential for catalytic activity.

Figure 4.

The SpvBmut1 protein is defective in its ability to ADP-ribosylate actin. Increasing amounts of purified SpvB or SpvBmut1 protein, containing a substitution of Glu-538 and Glu-540 to Asp, were assayed for their ability to ADP-ribosylate non-muscle actin as above (Fig. 2). The SpvB protein was added in the range of 0.13 µg to 0.5 µg, SpvBmut1 from 0.25 µg to 2 µg.

SpvBmut1 was recombined into the virulence plasmids of S. dublin Lane and S. typhimurium 14028s, replacing the wild-type spvB allele. Expression of the SpvBmut1 protein was confirmed by immunoblotting using antisera raised to the C-terminal domain of SpvB. Expression of spvC and spvD, two genes downstream of spvB, was found to be unaffected by the presence of the spvBmut1 allele, as determined by immunoblotting using antisera specific for their gene products.

We then compared the virulence of wild-type Salmonella with isogenic strains that express the SpvBmut1 protein. The active-site mutation in SpvB markedly attenuated Salmonella virulence in both susceptible BALB/c (Nramp1Asp169/Itys) and resistant congenic BALB/c. DBA/2 mice homozygous for the wild-type Nramp1Gly169/Ityr locus (Potter et al., 1983) (Fig. 5). The degree of attenuation was similar to that observed using a Salmonella strain defective for the expression of the entire spv locus (spvR, Fig. 5). As expected, Ityr mice were more resistant to Salmonella infection than Itys mice. Regardless of the functional status of the murine Nramp1 locus, the spvB mutation significantly decreased the virulence of Salmonella, as measured by cfu in liver and spleen. All bacteria recovered from mice infected with the spvBmut1 strain were shown to have retained the virulence plasmid. These findings suggest that the ADP-ribosylating activity of SpvB is essential for the virulence phenotype of the spv locus.

Figure 5.

Mutation of the active site of SpvB attenuates Salmonella virulence in mice. Groups of six BALB/c (itys) (A) or congenic BALB/c DBA/2 (ityr) (B) mice were inoculated intraperitoneally with either 100 or 500 cfu, respectively, of wild-type Salmonella dublin Lane, an isogenic strain containing the spvBmut1 gene, or Salmonella dublin with a mutation in spvR, which abolishes expression of the entire spv locus. The number of bacteria present in the spleen and liver of the infected animals was determined 5 days after infection. Error bars denote standard error. The P-values between wild-type and the SpvB mutant Salmonella were less than 0.006 for all data points. Similar results were obtained with Salmonella typhimurium 14028s (data not shown).

Discussion

The discoveries that the SpvB protein has actin ADP-ribosylating activity and that this activity is required for virulence place new emphasis on the importance of actin cytoskeletal modifications during Salmonella pathogenesis. Previous work has concentrated on effector proteins involved in the uptake of Salmonella by non-phagocytic cells (reviewed by Galán, 1999). In this process, the type III protein secretion system encoded by Salmonella pathogenicity island I (SPI-1) injects the SopE and SipA proteins into host epithelial cells, where SopE activates the GTPases Cdc42 and Rac-1, recruiting cytoskeletal components for the actin-bundling reaction promoted by SipA (Hardt et al., 1998; Zhou et al., 1999). The resulting membrane ruffling promotes uptake of Salmonella.

Although the depolymerization of actin mediated by the SpvB protein would potentially antagonize the events leading to cell invasion mediated by the SPI-1 locus, several lines of evidence suggest that the SpvB protein is expressed later in the cell infection cycle, after invasion or phagocytosis has occurred. The spv locus is regulated by an entirely different mechanism from SPI-1, involving a central role for the alternative sigma factor RpoS (Fang et al., 1992). In addition, SpvB expression increases steadily for 6 h after bacterial entry into epithelial cells and macrophages (Fang et al., 1992), and SpvB-dependent cytotoxicity does not become apparent in macrophages until 10–12 h after infection (Libby et al., 2000). This differential regulatory scheme ensures sequential modification of the actin cytoskeleton, first by SPI-1-mediated factors to promote bacterial uptake and, later, by SpvB to induce cytotoxicity. Thus, Salmonella has evolved a sequential and carefully orchestrated programme of actin cytoskeletal rearrangements during infection.

As SpvB is unable to cause actin disorganization in CHO cells when added to growth media, SpvB is probably actively transported by a specialized bacterial system into host cells to have its effect. A type III secretion system would be the most obvious possibility for a mechanism of transport, as these systems are able to transport a protein across both bacterial and host cell membranes. Supporting this hypothesis is the observation that SPI-2 mutant Salmonella are defective in a fashion similar to spv mutant bacteria, namely in proliferation and cytotoxicity (Ochman et al., 1996). Future work should help to discern whether SpvB-dependent pathogenesis requires a functional SPI-2 type III secretion system.

SpvB is the first ADP-ribosylating enzyme shown to be a virulence factor for an intracellular pathogen. The Clostridial and Bacillus actin-modifying toxins act extracellularly and have not been shown to have a role in pathogenesis (Thelestam et al., 1999). SpvB cytotoxicity is required for the intracellular proliferation of Salmonella in human monocyte-derived macrophages, and its expression correlates temporally with bacterial growth in these cells (Libby et al., 2000). As Salmonella grow within an intracellular vacuole, SpvB activity may cause local or global actin depolymerization that affects either vesicular trafficking or cellular physiology, or both, to promote Salmonella proliferation. Ultimately, this cytotoxicity leads to apoptosis (Libby et al., 2000), which could promote cell-to-cell spread of Salmonella by phagocytosis of infected, apoptotic cells.

Experimental procedures

Protein preparations and ADP-ribosylation assays

The spvB gene was cloned using standard molecular techniques into the pBAD/Myc-his plasmid (Invitrogen), induced in E. coli with 0.2% arabinose and purified using a nickel affinity column (Qiagen) according to the manufacturer's instructions. The fractions containing the purified protein were then concentrated using a Centricon-30 (Amicon) and frozen for later use. Salmonella extracts were prepared by growing the bacteria in LB for 12 h, centrifuging for 10 min at 4000 g and resuspending the bacteria in lysis buffer (50 mM NaH2PO4, 300 mM NaCl, 10% glycerol, 5 mM 2-mercaptoethanol and Complete protease inhibitor cocktail; Boehringer Mannheim). The cells were then sonicated three times for 30 s on ice, centrifuged for 30 min at 12 000 g to sediment particulates, filtered through a 0.45 µm filter and the supernatant concentrated using a Centricon-30 and frozen. The presence of SpvB in these extracts was confirmed by Western blots using a rabbit polyclonal antibody raised against the SpvB protein. CHO cell extracts were prepared by detaching confluent cultures with trypsin–EDTA, washing and resuspending in buffer (250 mM sucrose, 3 mM imidazole, 0.5 mM EDTA), passing the cells repeatedly through a 23-g needle and sonicating. The extract was then concentrated and frozen for later use.

ADP-ribosylation assays were performed essentially as described previously (Perelle et al., 1996). Either 0.12–2 µg of purified SpvB (wild type or SpvBmut1) or 15 µg of Salmonella extract was mixed with 6 µg of CHO extract or 1 µg of purified actin in the presence of 100 mM HEPES, pH 7.2, 5 mM ATP, 2.5 mM ADP-ribose and 1.25 µCi of 800 Ci mM−1[32P]-nicotinamide adenine dinucleotide (NEN), incubated at 37°C for 1 h and then run on a 12% polyacrylamide gel, dried and exposed to film.

Transfection of eukaryotic cells

CHO K1 cells were transfected using lipofectamine reagent (BRL) according to the manufacturer's instructions. Briefly, 2.5 × 104 CHO cells were added to 0.5 ml media well−1 in 8-well Lab Tek II chamber slides and allowed to grow overnight. The cells were then transfected with 0.5 µg of DNA in 2.5 µl of lipofectamine reagent in 250 µl of media lacking serum. After 4 h, 250 µl of media containing 10% fetal calf serum was added to each well, and the transfected cells were then incubated overnight to allow gene expression. Monolayers were fixed with 3.3% paraformaldehyde, stained with phalloidin–rhodamine (Molecular Probes) and examined using a Nikon E800 fluorescent microscope at 400× magnification. Transfections were repeated at least three times with each DNA tested, and representative fields from these transfections were photographed. GFP and rhodamine fluorescence were photographed separately, and the resulting images were merged using photoshop software (Adobe).

Construction and analysis of Salmonella mutations

A 991 bp HincII–StuI fragment of spvB (nucleotides 1088–2079) was subcloned into pBluescript II, and site-directed mutagenesis was carried out using Stratagene's QuickChange mutagenesis kit and primers designed to replace Glu-536 and Glu-538 with Asp at each position. In addition, a silent nucleotide change just upstream of the active site was made to create a new DraI site to facilitate screening. The new sequence of the spvBmut1 allele from nucleotides 1602 to 1622 is as follows: TTTAAAGGAGACGCAGACATG, in which the underlined nucleotides denote changes from wild type. Complete sequencing of the mutated 991 bp fragment verified that the mutation was generated correctly.

The mutated spvB gene fragment was introduced into S. dublin Lane and S. typhimurium 14028s using standard allellic exchange techniques (Chen et al., 1996). Briefly, the mutated fragment generated above was cloned into the suicide plasmid pEP185.2 (Pepe and Miller, 1993) and maintained in SM10λpir. This plasmid was then transferred into strains of S. typhimurium and S. dublin Lane containing a kanamycin cassette downstream of the spv locus and selected by plating on chloramphenicol and kanamycin. Growth on this media indicated plasmid integration via homologous recombination. Transformed Salmonella were then passaged in liquid media for 4 days without selection, at which point cycloserine enrichment was used to select against chloramphenicol resistance and find those clones that had deleted the plasmid sequences by a second recombination event (Cole et al., 1993). These strains were then checked by polymerase chain reaction (PCR) and Southern blotting to confirm that plasmid sequences were absent and that the mutated spvB allele was present. Finally, the mutated gene was transferred into a clean background strain using P22 transduction and the downstream kanamycin marker to reduce the possibility of background mutations. Proper expression of the mutant allele was confirmed by immunoblotting using an anti-SpvB polyclonal antibody raised in rabbits to the C-terminal domain. Proper expression of the downstream spv genes was confirmed by immunoblot using polyclonal antibodies raised in rabbits against the SpvC and SpvD proteins (El-Gedaily et al., 1997).

Acknowledgements

We thank S. Okamoto, C. Waters and P. Hasegawa for their expert technical assistance, and members of the Guiney laboratory for helpful discussions. This work was supported by NIH grants AI32178 and DK35108 to D.G.G. and J.F., and an Isaak Walton Killam Faculty Research Fellowship award to N.E.R.

Ancillary