Vibrio vulnificus Biotype 3 Multifunctional Autoprocessing RTX Toxin Is an Adenylate Cyclase Toxin Essential for Virulence in Mice

Abstract

Vibrio vulnificus is an environmental organism that causes both food-borne and wound infections with high morbidity and mortality in humans. The annual incidence and global distribution of infections associated with this pathogen are increasing with climate change. In the late 1990s, an outbreak of tilapia-associated wound infections in Israel was linked to a previously unrecognized variant of V. vulnificus designated biotype 3. The sudden emergence and clonality of the outbreak suggest that this strain may be a true newly emergent pathogen with novel virulence properties compared to those of other V. vulnificus strains. In a subcutaneous infection model to mimic wound infection, the multifunctional autoprocessing RTX (MARTX) toxin of biotype 3 strains was shown to be an essential virulence factor contributing to highly inflammatory skin wounds with severe damage affecting every tissue layer. We conducted a sequencing-based analysis of the MARTX toxin and found that biotype 3 MARTX toxin has an effector domain structure distinct from that of either biotype 1 or biotype 2. Of the two new domains identified, a domain similar to Pseudomonas aeruginosa ExoY was shown to confer adenylate cyclase activity on the MARTX toxin. This is the first demonstration that the biotype 3 MARTX toxin is essential for virulence and that the ExoY-like MARTX effector domain is a catalytically active adenylate cyclase.

INTRODUCTION

Vibrio vulnificus is a Gram-negative pathogen that inhabits coastal waters throughout the world (1). This pathogen causes severe food-borne and wound infections in humans that can progress to tissue necrosis, septicemia, organ failure, and death, often within 48 h of pathogen exposure (1–3), particularly among individuals who suffer from cirrhosis of the liver and other underlying predisposing conditions (4, 5). Mortality rates can reach as high as 50%, with extreme morbidity among survivors, including amputations, surgical tissue debridement, and long hospital stays (1–3). Although the disease remains rare, the incidence of V. vulnificus is rising because of global warming (6, 7) and the geographical area impacted by V. vulnificus in the water and food supply is expanding (7–11).

Biotype 1 (BT1) strains are most commonly associated with clinical infections, while BT2 strains cause infections in eels (1, 12, 13). From 1996 to 1999, changes in marketing practices to allow vendors to sell live fish in Israel led to an outbreak of a new lineage of V. vulnificus, resulting in a large number of wound infections associated with the handling and preparation of tilapia and carp (14). These BT3 strains continue to be present in fish ponds and the Sea of Galilee and have caused nearly 200 infections with 10% mortality from 1996 to 2008 (15).

The primary virulence factor of V. vulnificus BT1 and BT2 strains is the multifunctional autoprocessing RTX (MARTX) toxin (16–18). The secreted MARTX toxins are very large single-polypeptide toxins composed of repeat sequences that form a pore in eukaryotic cell plasma membranes for the translocation of centrally located effector domains (19). Within the cell, the autoproteolytic cysteine protease domain (CPD) is activated by the binding of inositol hexakisphosphate to release individual effector domains into the cytosol (20). Among the BT1 and BT2 V. vulnificus isolates studied, at least four different MARTX variants have been annotated that carry distinct repertoires of effector domains and these have been shown to be exchanged frequently in this naturally competent bacterium by horizontal acquisition of DNA, followed by recombination (21–23).

In this study, it is shown that BT3 strains have a unique variant of the V. vulnificus MARTX toxin that has been converted to an adenylate cyclase toxin. This toxin is essential for the virulence of BT3 strains in mice infected by the subcutaneous (s.c.) route.

MATERIALS AND METHODS

Bacterial strains, cell lines, and reagents.

All of the bacterial strains used in this study are listed in Table 1. Ten-locus multilocus sequence typing (MLST) was used to verify strains as BT3 according to the methods of Bisharat et al. (24) by using the pubmlst.org website developed by Keith Jolley at the University of Oxford (25). A spontaneously rifampin-resistant isolate of BAA87 was selected by serial dilution into Luria-Bertani (LB) broth containing increasing concentrations of rifampin. Bacteria were grown in LB broth and on LB agar supplemented with the antibiotics ampicillin (100 μg/ml), chloramphenicol (10 μg/ml), kanamycin (50 μg/ml), and rifampin (50 μg/ml) as needed.

TABLE 1.

V. vulnificus strains used in this study

Strain	Relevant description	Source or reference
CMCP6	BT1, clinical isolate, South Korea, Rifr	16
BAA86 (CDC 9505-95)	BT3, human blood, Haifa, Israel, 1995	ATCCa
BAA87 (CDC 9530-96)	BT3, human wound, Afula, Israel, 1996	ATCC
11028	BT3, human clinical, Israel, 1996	E. F. Boyd, University of Delaware
2095/00	BT3, environmental isolate, Israel, 2000	A. Depaola, FDA
JK57 (BAA87rif)	Spontaneously rifampin-resistant BAA87	This study
KZ1 (BAA87rtxA1)	BAA87 rtxA1::kan, Kanr	This study
KZ4 (BAA87rtxA1/c)	BAA87rtxA1 restored to Kans by recombination with wild-type sequence	This study

^aATCC, American Type Culture Collection.

Chinese hamster ovary (CHO) cells obtained from Rich Longnecker (Northwestern University) were cultured in Life Technologies Ham's F12 nutrient mixture with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin obtained from Gemini Bio-Products (West Sacramento, CA).

All other common chemical reagents were obtained from Fisher or Sigma. Molecular biology enzymes were from New England BioLabs (Ipswich, MA), and DNA primers were obtained from Integrated DNA Technologies (Coralville, IA).

rtxA1 gene sequencing.

The effector domain regions of rtxA1 genes were amplified, and their nucleotide sequences were determined as previously described (21). The full-length sequence of rtxA1 from BAA87 was extended from the effector domain region to the 5′ and 3′ ends of the gene by using primers designed to highly conserved regions of the rtxA1 genes of BT1 and BT2 strains for amplification into regions of known BAA87 sequence, followed by sequencing of products by primer walking. Coverage was 1× to 6× by this strategy. The accuracy of the sequence across zones of poor coverage was later verified by whole-genome sequencing of BAA87 with the Life Technologies Ion Torrent PGM technology with a 318 chip paired with a 400-bp library. The whole-genome sequence of BAA87 (GenBank accession no. JDSE00000000.1) has been published separately (55).

Generation of BAA87 rtxA1 mutant and complemented strain.

The region of BAA87 rtxA1 from bp 341 to bp 5446 was amplified by PCR and captured in the plasmid pCR-XL-TOPO (Life Technologies) by TOPO cloning according to the manufacturer's protocol to generate plasmid pJK.206. The fragment corresponding to bp 341 to 4974 was transferred into sacB-counterselectable plasmid pDS132 (26) with SacI and SalI. The resulting plasmid, pKZ38, was digested at the unique FspI restriction site at bp 2042 into which the 1,170-bp HincII kanamycin resistance cassette from pKanπ (27) was inserted so as to create a null mutation in rtxA1. The resulting plasmid, pKZ37, has a kanamycin resistance gene under the control of its own promoter with flanking rtxA1 DNA for recombination. pKZ37 was mated from S17-1λpir to rifampin-resistant BAA87, and merodiploids were selected by gain of chloramphenicol and kanamycin resistance. The colonies were purified and then passaged on LB agar with kanamycin in the absence of chloramphenicol. Recombination events that excised the plasmid from the merodiploid were selected by growth on no-salt LB agar with 6% sucrose and kanamycin. The gain of the kanamycin resistance gene in the correct location on the chromosome of the selected KZ1 isolate (BAA87rtxA1) was confirmed by PCR and sequencing of the PCR product through the rtxA1::kan junctions. The strain was then reverted to the wild-type sequence by mating KZ1 with S17-1λpir(pKZ38) and sacB counterselection as described above with screening for loss of kanamycin resistance to create strain KZ4 (BAA87rtxA1/c). PCR amplification, followed by sequencing, was done to confirm that the strain was restored to the wild-type sequence. The strains were confirmed to produce a MARTX toxin by gain or loss of rtxA1-dependent pore-forming cell lysis.

Cloning of LF_NVvExoY expression vector and site-directed mutagenesis.

The region of BAA87 rtxA1 from bp 9702 to bp 10791 was amplified by PCR with primers modified for ligation-independent cloning (LIC) (28). LIC was used to insert the PCR product into plasmid pRT24, a vector modified from the anthrax toxin lethal factor N terminus (LF_N) expression vector pABII (29) by addition of the LIC site (28). The resulting plasmid, pKZ5, was also used as the template for site-directed mutagenesis with Agilent (Santa Clara, CA) cloned Pfu polymerase following the manufacturer's protocol. Amplified plasmids were transformed into Agilent XL10-Gold Ultracompetent cells and plated on LB agar with ampicillin. Recovered plasmids were screened for gain of the mutation by sequencing and then transformed to BL21(DE3) for expression of the modified protein.

Expression and purification LF_NVvExoY.

BL21(DE3) carrying either plasmid pKZ5 or mutagenized variants was grown in LB with ampicillin (37°C, 250 rpm) to an A₆₀₀ of ∼0.6, when expression was induced with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 3 h at 37°C with shaking. Bacteria were collected by centrifugation, resuspended in lysis buffer (500 mM NaCl, 20 mM Tris [pH 8.0], 5 mM imidazole, 1 mg/ml lysozyme, 1% Triton X-100), and lysed on ice by sonication with a Branson Digital Sonifier with a disruptor horn attachment by eight 30-s pulses at 40% amplitude. The soluble fraction containing LF_NVvExoY was clarified by centrifugation at 27,000 × g, filtered across a 0.45-μm syringe filter, and loaded onto a GE Healthcare 1-ml HisTrap HP column with a GE Healthcare ÄKTApurifier fast protein liquid chromatography system. The column was washed with 10 ml of loading buffer (500 mM NaCl, 20 mM Tris [pH 8.0], 5 mM imidazole) and then with a 20-ml buffer gradient of 5 to 250 mM imidazole, wherein the protein eluted at 55 mM. The protein was further purified by size exclusion in 20 mM Tris (pH 8.0)–500 mM NaCl with a Superdex 200 10/30 GL column. The protein was concentrated with Amicon Ultracel 30K and then stored at −20°C in 10% glycerol. The concentration of the protein was determined with a NanoDrop 100, and the purity of 10 μg was assessed by SDS-PAGE, followed by staining with Coomassie blue.

In vitro adenylate cyclase assay.

A 100-μl reaction mixture contained 10 μl of 10× reaction buffer (100 mM Tris [pH 8.0], 60 mM MgCl₂, 2 mM CaCl₂, 20 mM dithiothreitol), 30 μg bovine serum albumin (BSA) or a postnuclear CHO cell extract (CE, prepared as described by Xu and Barbieri [30], except with a Sonicel microdisruptor to dismembranate the cells), and 1 μmol of LF_NVvExoY, except where noted otherwise. A 0.2 μM concentration of ATP was always added last to prevent ATP rolling, and the reaction was carried out for 30 min at 30°C. The reaction was stopped by adding 186 μl of ethanol and centrifugation at 16,000 × g for 5 min. The supernatant was then lyophilized with a Savant SpeedVac DNA 120 until dry. The presence of cyclic AMP (cAMP) in the lyophilized sample was initially assessed by separating the rehydrated reaction mixture on a C₁₈ high-performance liquid chromatography (HPLC) column in 100 mM potassium phosphate, pH 6.0, across an 8-ml gradient of 0 to 10% methanol on the ÄKTApurifier configured for HPLC. The identity of the cAMP peak was confirmed by treatment with phosphodiesterase (PDE; Sigma P9529) for 1 h at 30°C. For quantification, the concentration of cAMP in rehydrated samples was determined by enzyme-linked immunosorbent assay (ELISA) as described below.

cAMP production in CHO cells cocultured with V. vulnificus.

CHO cells (5 × 10⁵) were seeded into 12-well tissue culture dishes. Two hours prior to bacterial infection, cells were washed three times with phosphate-buffered saline (PBS) and the medium was changed to Ham's F12 plus 10% FBS without antibiotics. V. vulnificus grown to mid-log phase in LB was washed in PBS, and the A₆₀₀ of all samples was equalized with a Beckman Coulter DU530 spectrophotometer. Bacteria were added to medium over cells at a multiplicity of infection (MOI) of 20, and the plates were centrifuged at 500 × g for 5 min. The plates were then incubated at 37°C with 5% CO₂ for 90 min. Cells were washed in PBS and then lysed in the wells with 600 μl of PBS with 1% Triton X-100 at 37°C for 10 min, and the lysate was stored at −20°C. The concentration of cAMP in 100 μl of lysate was determined as described below. Cell viability throughout the assay was verified by phase microscopy for the presence of cells and in some experiments by measuring the release of lactate dehydrogenase (LDH) into the medium as an indicator of cell lysis with the Promega CytoTox96 kit according to the manufacturer's instructions. To generate a positive control for all cAMP assays, 10 μM forskolin was added to an extra well 30 min prior to the end of the incubation period and assayed for maximum cAMP production in parallel with the other samples.

cAMP ELISA.

cAMP concentrations from in vitro assays or cell lysates were measured with the Enzo Life Sciences cAMP Complete ELISA by following the manufacturer's protocol, except for the use of the alternative lysis buffer indicated above for lysate assays. For some in vitro assays, lysates were diluted in the lysis buffer to reduce the cAMP concentration to the detectable range. Plates were read on a Molecular Devices SpectraMax M5 plate reader, and concentrations were determined on the basis of a standard curve prepared on the same assay plate.

Mouse survival assay.

All mouse infections were conducted in accordance with protocols approved by the Northwestern University Institutional Animal Care and Use Committee. Five- to 6-week-old female C57BL/6 mice for assays were obtained from Harlan Laboratories. V. vulnificus was grown to mid-log phase and washed in PBS, and the A₆₀₀ was equalized. A 50-μl volume of PBS containing V. vulnificus was injected s.c. under the dorsal skin of mice anesthetized with 100 μl of a cocktail of 10 μg/ml ketamine and 2 μg/ml xylazine in PBS. After 8 h of infection, mouse survival was observed every 2 h. Severely moribund mice were humanely euthanized by cervical dislocation under anesthesia and counted as nonsurvivors. Surviving mice showing no symptoms of infection were euthanized at the termination of the experiment.

Wound pathology.

Five mice per group were inoculated s.c. with 10⁶ CFU of V. vulnificus and euthanized after 12 h. Skin wounds were excised with the fur intact, and the tissue was fixed in 10% neutral buffered formalin overnight and then embedded in paraffin. Sections were stained with hematoxylin and eosin. Slides were viewed by a trained skin pathologist (R.M.L.) on a Zeiss Axioplan2 microscope for pathology, which was consistent across all five animals, and representative images were taken at magnifications of ×400 and ×630.

Nucleotide sequence and MLST data accession numbers.

MLST data from this study have been deposited with pubmlst.org/vvulnificus as follows: BAA87, id285; 11028, id288; 2095/00, id289. All of the rtxA1 nucleotide sequences obtained in this study have been deposited in GenBank as follows: full-length BAA87 rtxA1, accession number KJ131555; partial rtxA1 sequences of BAA87, 11028, 2095/00, and BAA86, accession numbers KJ143989 to KJ143992, respectively.

RESULTS

Sequencing of the rtxA1 genes of various BT3 V. vulnificus strains.

BT3 V. vulnificus BAA87 isolated from a patient wound during the original epidemic in 1996 was obtained. The sequence of the rtxA1 gene, which encodes the MARTX toxin, was determined by sequencing of PCR products generated with primers designed to rtxA1 genes from other strains. The gene was determined to be 15,627 nucleotides (nt) long and to encode a deduced protein of 5,208 amino acids (aa). The accuracy of our BAA87 sequence was later verified, as it matched the rtxA1 gene from whole-genome sequencing of BT3 isolate VVyb1, which was from tilapia fish (31), except that BAA87 has a confirmed 15-nt deletion of a sequence that is precisely duplicated in VVyb1 (Table 2).

TABLE 2.

Nucleotide polymorphisms in BT3 rtxA1 genes compared to BAA87

Strain	Nucleotide polymorphism	Amino acid change in MARTX
ATCC BAA86a	G10916C, T12332C	S →T, V →A
BT3 clinical isolates
11028a	C6214T, C9273T, C11062G, T11841C	P →S, SYN, L →V, SYN
2095/00a	C2937T, 18 other mutations	N → stop (CAG →TAG)
VVyb1b	15-nt duplication inserted at nt 3114	LSVTA inserted

^aOnly the effector domain region (rtxA1 nt 5678 to 12746) was sequenced.

^bCompared to rtxA1 gene from published whole-genome sequence (31).

Surprisingly, neither the BAA87 nor the VVyb1 rtxA1 sequence matched a previously annotated rtxA1 sequence from BT3 clinical isolate 11028 (GenInfo identifier 283558386), a sequence of 14,112 nt with a deduced product of 4,703 aa (22). Thus, we obtained three additional BT3 isolates to determine which sequence is representative of the BT3 rtxA1 gene. To confirm the identities of the strains, we first conducted a 10-locus MLST analysis based on the methods of Bisharat et al. (24). The sequence typing (ST) profile of clinical isolate ATCC BAA86 was previously published and deposited at www.pubmlst.org/vvulnificus (25), and this published ST profile, number 8, was used as the reference. The ST profile of BAA87 was a 100% match for that of BAA86, as was the profile of isolate 11028, which was obtained from E. Fidelma Boyd (University of Delaware), and that of 2095/00 (32), which was obtained from the FDA. Thus, all of the strains in this study were confirmed by MLST to be BT3 strains.

We next sequenced the 7,068-nt region of the rtxA1 gene corresponding to the effector domain region of each strain. The nucleotide sequences in the effector domain regions of BAA86 and 11028 were identical in length to that of BAA87 but with two and four single nucleotide polymorphisms, respectively (Table 2). Within the 7,068 nt sequenced from strain 2095/00, there were 19 single nucleotide polymorphisms. Most notably, one of these changes introduced a stop codon, indicating that this is a natural rtxA1 null strain.

Overall, it is concluded that, with the exception of spontaneously rtxA1 null strain 2095/00, all BT3 strains, including the 11028 isolate we obtained, have rtxA1 genes that are nearly identical in sequence and would produce a new variant of the MARTX toxin that has the same domain structure.

The BT3 MARTX toxin is a novel variant.

Annotation of the deduced amino acid sequence of the BT3 MARTX toxin showed that the predicted arrangement of effector domains differs from that of previously annotated BT1 and BT2 strain variants (Fig. 1A and B). Similar to the variants produced by commonly studied BT1 strains CMCP6 and MO6-24/O (also known as type I and II MARTX toxins), the BT3 strain variants have a domain of unknown function in position 1 (DUF1), a Rho inactivation domain (RID) in position 2, and a domain with homology to the alpha-beta hydrolase (ABH) family of enzymes in position 3 (Fig. 1A) (21, 22).

FIG 1.

The BT3 MARTX toxin is unique among MARTX toxins. Linear schematics of MARTX toxins are depicted to scale with the conserved regions shown with repeats indicated as hatched gray boxes and the autoprocessing CPD in light blue. The effector domain arrangement is different for each MARTX toxin. (A) Deduced structures of all BT3 strains, represented by BAA87, compared to those of BT1 CMCP6 and MO6-24/O, which are the closest in overall structure. (B) BT2 strains have a distinct structure that bears some similarity to BT1 E64MW, which was found here to also have acquired an ACD and an ExoY-like domain. (C) Other species noted as having genes for MARTX toxins with an ExoY-like domain are shown; some also have DmX. Effector domain designations are DUF1, RID, ABH family protein, makes caterpillars floppy homology (Mcf), DUF5, ExoY-like adenylate cyclase (ExoY), DmX, and the ACD. More detailed descriptions of the effector domains can be found in a recent review article (19).

Two new effector domains not previously detailed as being present in the V. vulnificus MARTX toxins are in the next two positions. The domain in the fourth position is most similar to the type III secretion effector ExoY adenylate cyclase from Pseudomonas aeruginosa. Sequence comparison studies have also identified this domain in putative MARTX toxins from environmental V. cholerae isolate RC385 and coral V. caribbeanicus isolate BAA-2122 and from the plasmid-borne MARTX toxin produced by the shrimp pathogen V. nigripulchritudo (Fig. 1C) (19, 33, 34), although no functional studies have established that this domain confers adenylate cyclase activity on these putative toxins. Gene annotation, done for this study, of the rtxA1 gene from recently sequenced BT1 V. vulnificus environmental isolate E64MW (23) also revealed that this strain carries an ExoY-like adenylate cyclase effector domain, as well as an actin cross-linking domain (ACD), a domain shared with the BT2 MARTX toxin (Fig. 1B), although again no functional studies have shown that this putative MARTX toxin has adenylate cyclase activity.

The effector domain in the fifth position is a DUF. As “DUF in the fifth position” (DUF5) is already used to designate the Pasteurella-like effector domain of the CMCP6 MARTX toxin (35), we refer to this domain as domain X (DmX). Sequence analysis revealed that this effector domain is present also in the putative MARTX toxin from V. cholerae environmental isolate RC385, where it is duplicated and present on either side of the ExoY-like domain and in the plasmid-borne MARTX toxin of V. nigripulchritudo in the first position (Fig. 1C) (19, 33, 34). No studies of the functional activity of this domain have been conducted.

BT3 V. vulnificus coculture increased intracellular cAMP concentrations in CHO cells.

A notable feature of the unique sequence of the BT3 MARTX toxin is the gain of a domain with 26% amino acid sequence identity and 42% similarity to P. aeruginosa ExoY (PaExoY) (30) and 19 to 20% identify and 35 to 36% similarity to the adenylate cyclase domains of the Bordetella pertussis adenylate cyclase toxin (BpACT) and the Bacillus anthracis edema factor (EF) (36). A domain analysis revealed strong conservation in the regions corresponding to the active site defined by the crystal structures of the adenylate cyclase domains of BpACT and EF (Fig. 2).

FIG 2.

Alignment of VvExoY with adenylate cyclase toxins and effectors. (A) Schematic diagrams of the V. vulnificus MARTX toxin VvExoY effector domain (corresponding to its putative cleavage sites) and full-length secreted bacterial proteins with adenylate cyclase activity P. aeruginosa ExoY (PaExoY), B. pertussis adenylate cyclase toxin (BpACT), and B. anthracis EF. On the basis of comparative analysis of the crystal structures of BpACT and EF bound to CaM (36), the catalytic domain is composed of C_B and the C_A1 and C_A2 segments, which fold to form the C_A subdomain. The binding site for ATP is formed at the interface between the two subdomains and contacting switch B (SwB). The boxed residues are known to be essential for cyclase activity. Poorly conserved switch A (SwA) contacts the C terminus of CaM in both EF and BpACT. (B) ClustalW alignment generated in MacVector shows conservation in the catalytic regions of VvExoY aa 3255 to 3547, PaExoY aa 34 to 307, BpACT aa 9 to 314, and EF aa 299 to 594. The C-terminal tail portion of EF, which is known to contact the N terminus of CaM (N-CaM), is shown in diagrams, but this part is not conserved in ExoY proteins and is poorly conserved in BpACT. GD, Gly-Asp-rich repeats; PA, protective antigen.

When CHO cells were cocultured with BT3 strains BAA86, 11028, and BAA87rif at an MOI of 20 for 90 min, their intracellular concentration of cAMP increased by >30-fold (Fig. 3A and B). In contrast, the spontaneous rtxA1 null strain 2095/00 induced only a 3- to 10-fold increase in cAMP production in CHO cells during coculture, suggesting a link to the rtxA1 gene (Fig. 3A and B). The loss of cAMP was not due to a growth defect in 2095/00, as all of our BT3 isolates grew at similar rates in Luria broth, although all grew more slowly than the representative BT1 strain CMCP6 (Fig. 3C), consistent with previous studies that BT3 strains have a lower growth rate in all media (37). Thus, it is most likely that the lack of induction of cAMP in CHO cells by 2095/00 is due to its naturally occurring null mutation in rtxA1.

FIG 3.

Coculture of V. vulnificus BT3 increases the intracellular cAMP concentration in CHO cells. (A, B) The strains indicated were cocultured with CHO cells at an MOI of 20 for 90 min. The intracellular concentration of cAMP in the clarified lysate was determined by ELISA. (C) Growth kinetics of the BT3 strains indicated and BT1 CMCP6 measured in Luria broth with shaking at 37°C. Experiments with three biological replicates are representative of two independent experiments performed with similar results. Statistical significance was determined first by analysis of variance as P < 0.05, and the specific P values indicated were obtained for compared samples with the Student t test. OD, optical density.

To affirmatively link the adenylate cyclase activity to the rtxA1 gene, an isogenic null mutation in the BAA87rif rtxA1 gene was generated by recombining the kanamycin resistance cassette nptII into the 5′ end of the rtxA1 gene in a way that completely disrupted the transcription and translation of the downstream rtxA1 gene. The KZ1 mutant strain (BAA87rtxA1) was unable to induce cAMP on coculture with CHO cells (Fig. 3B). In order to show that the loss of function was not due to spontaneous mutation at another locus during genetic manipulation, we used genetic recombination to remove the kanamycin resistance cassette from KZ1 and restore the wild-type sequence. The restored KZ4 strain regained the ability to stimulate cAMP production (Fig. 3B). Overall, these data demonstrate that the BT3 rtxA1-encoded MARTX toxin induces cAMP production in cells.

The BT3 MARTX toxin also causes CHO cell lysis.

A phenotype closely associated with V. vulnificus BT1 MARTX toxin is the ability to induce lysis of cells in culture. Upon coculture at an MOI of 20 for 90 min as conducted for the cAMP assays, lysis of cells was either undetectable or minimal (<5%) and cells generally appeared normal, although some cell rounding and distention was observed (Fig. 4A and B). This low level of lysis increased over the next 90 min such that after 3 h, 90 to 100% of the CHO cells incubated with BT3 strain BAA87rif were lysed and the cells were visually necrosed. The disruption of rtxA1 resulted in an 87% decrease in LDH release at 180 min, indicating that lysis at an MOI of 20 was due primarily to the MARTX toxin. However, the detection of 6 to 15% LDH release and the appearance of necrotic cells demonstrate that other secreted cytolysins also contribute. At a higher MOI of 100, the amount of cell lysis at 90 min varied up to ∼25%, with 80 to 100% lysis occurring by 180 min with all of the strains, including the rtxA1 mutant. These data demonstrate that BT3 MARTX toxin can induce cell lysis, although cytolysis is not due solely to the MARTX toxin, as has been shown also for BT1 and BT2 (16–18, 38, 39).

FIG 4.

BT3 rtxA1-dependent and -independent CHO cell lysis. (A) Supernatant fluids of CHO cells incubated with V. vulnificus as indicated for the times indicated were assayed for LDH release as an indicator of cell lysis. (B) Prior to collection of supernatant fluids, digital phase-contrast images of cells were taken with a Nikon TS100 at a magnification of ×200. The experiment shown was conducted with three biological replicates and is representative of at least two independent experiments performed with similar results. The differences between the rtxA1 mutant and (i) BAA87rif and (ii) the complemented mutant at an MOI of 100 were determined for each time point and found by analysis of variance to be significant (P < 0.0001) at 90 min (dagger) but not significant (ns) at 180 min. All other differences were not significant or were within the range of variability observed across repeated assays.

Purified recombinant V. vulnificus ExoY (VvExoY) has adenylate cyclase activity.

As the BT3 MARTX toxin is associated with perturbation of the cell at an MOI of 20, one mechanism for the increase in cAMP caused by the BT3 MARTX toxin could be induction of endogenous host adenylate cyclases activated by pore formation. To further demonstrate that the ExoY-like effector domain is the adenylate cyclase, the sequence of the gene for the ExoY-like effector domain was amplified and cloned fused to the anthrax toxin lethal factor N terminus (LF_N). MARTX toxin effectors are highly unstable in solution because of a natural tendency to unfold and difficulties identifying the termini of the domains within the larger holotoxins. However, we have found through studies of the ACD (40) and RID (41) that fusion proteins can be sufficiently stable for limited biochemical and cell biological analyses.

In an in vitro reaction, LF_NVvExoY catalyzed the conversion of ATP to cAMP. When the terminated reaction was separated on a C₁₈ reverse-phase column, a new peak corresponding to the expected retention time of cAMP was observed. This peak was eliminated when the reaction was conducted in the presence of PDE to degrade the phosphodiester bond in cAMP (Fig. 5A).

FIG 5.

The ExoY-like domain of BT3 MARTX toxin has in vitro adenylate cyclase activity. Purified recombinant VvExoY fused to anthrax toxin lethal factor to enhance stability (LF_NVvExoY) or purified P. aeruginosa ExoY (PaExoY) as a positive control (gift from W. J. Tang, University of Chicago) was used in an in vitro assay of ATP conversion to cAMP. (A) The product of the completed reaction of 1 μmol of LF_NVvExoY in the presence (solid line) or absence (dashed line) of PDE was separated by C₁₈ reverse-phase chromatography. The expected peaks are marked. (B) Final reactions conducted in the presence of either 30 μg BSA or CE were diluted in buffer for quantification by ELISA. Values were corrected against a mock control containing either BSA of CE to negate potential effects due to cellular adenylate cyclases present in the extract. (C) Reaction mixtures with 1 μmol of LF_NVvExoY enzyme in the presence of BSA (no CE) were not diluted to allow direct comparison with low-activity mutants. Statistical significance was determined first by analysis of variance as P < 0.05, and the specific P values indicated for the samples compared were obtained with the Student t test. Daggers in panel B indicate statistically significant differences from BSA controls and in panel C from the LF_NVvExoY control set at a maximum value of 200 for all samples. All results shown are means and standard deviations of three biological replicates in experiments conducted at least twice to ensure reproducibility.

The amount of cAMP produced during the in vitro reaction was quantified by ELISA, which showed that cAMP production occurred in a dose-dependent manner. The overall activity was, however, about 100-fold less efficient than the basal level of activity of the nearest protein homolog, PaExoY (Fig. 5B). Yet, purified LF_NVvExoY containing single amino acid substitutions in four conserved residues required for catalysis by PaExoY (42) (Fig. 2) did not catalyze cAMP production, confirming that the ExoY-like effector domain is a class II adenylate cyclase (Fig. 5C).

The adenylate cyclase BpACT and EF are activated by calmodulin (CaM) through contact with the catalytic domain switch A region and a helical domain to the C-terminal side of the catalytic domain (36, 43, 44) (Fig. 2). However, the closest homolog of VvExoY is PaExoY, and this protein is activated not by CaM but by an unknown eukaryotic protein (30). However, in a parallel experiment, CE, which activated PaExoY, did not stimulate the activity of LF_NVvExoY (Fig. 5B) and, indeed, partially inhibited its activity. It is unknown if the VvExoY effector domain is constitutively active or if it requires an activator that is distinct from the unknown activator used by PaExoY.

BT3 MARTX toxin is essential for virulence and tissue damage at the wound site.

Both CMCP6- and MO6-24/O-type MARTX toxins of BT1 strains and the actin-cross-linking MARTX toxins of BT2 strains have been shown to be important for virulence with mice as a model system (16–18, 38, 39, 45, 46). As the BT3 MARTX toxin is unique, it was necessary to determine if the toxin is still an important virulence factor. As BT3 strains in the outbreak in Israel are all linked to wounds, we used s.c. infection of 5- to 6-week-old inbred C57BL/6 mice as a model for wound-induced disease. Mice injected with iron dextran complex to increase serum iron levels rapidly succumbed to infection with as few as 100 CFU (Fig. 6A). However, BT3 infections are not strongly correlated with underlying infection or high iron content, as 45% of patients have no predisposing conditions (15). In mice not treated with iron dextran, mice also succumbed to infection with BAA87rif in a dose-dependent manner (Fig. 6B). By comparison, at a dose that was lethal to 10/10 mice, the BAA87rtxA1 mutant was significantly less virulent, resulting in the deaths of only 2/10 mice. Genetic exchange of the kanamycin resistance cassette for the wild-type sequence restored virulence to BAA87rtxA1, demonstrating that the virulence defect was directly associated with the rtxA1 gene disruption and not due to unlinked mutations acquired during genetic manipulation (Fig. 6C).

FIG 6.

BAA87 is virulent in mice inoculated s.c. dependent upon rtxA1. Survival curves after s.c. infection from iron dextran-treated (A) and untreated (B) mice (five per group) infected with increasing numbers of CFU of BAA87. (C) Pooled data from two experiments with 5 mice per group infected with 10⁵ CFU of BAA87 and the mutant strains indicated. The rtxA1 mutant in panel C is significantly different from both BAA87 and the complemented strain, as determined by the Mantel-Cox log rank test.

Histological evaluation of skin revealed a thin epidermis overlying a dermis with a marked inflammatory infiltrate (Fig. 7B and C). The fabric of the dermis appeared fractured throughout and lacked the customary fibrillar organization seen in mock-treated animals and in mice treated with the BAA87rtxA1 mutant (Fig. 7A and D). The dermis of BAA87-infected mice contained greater numbers of inflammatory cells than the dermis of mutant-infected mice. Most striking in the BAA87-infected mice was the appearance of the s.c. fat and the panniculus carnosus of the skin. The s.c. fat layer was markedly thicker and filled primarily with polymorphonuclear lymphocytes (PMNs). In addition, small, rod-like bacteria were noted within the s.c. fat. A brisk, dense infiltrate was observed at the junction of the s.c. fat and the panniculus carnosus. At high magnification, this infiltrate consisted mainly of PMNs. In contrast, the s.c. fat layer of mice infected with BAA87rtxA1 contained fewer infiltrating cells, although more than noted in the mock-treated animal. Also, by 12 h postinoculation, no bacteria were noted in the mutant-infected animal, consistent with resolution of the skin wound infection. Overall, the severity of the pathology of wounds of BAA87-infected mice was dependent on rtxA1 and the wounds were highly inflamed, with severe damage that affected every layer of the skin tissue.

FIG 7.

Pathology of wounds in BAA87- and BAA87rtxA1-infected mice. Mice used for pathology determination were either mock inoculated by injection with PBS (A) or inoculated s.c. with 10⁶ CFU of the V. vulnificus strain indicated (B to D). Mice were euthanized after 12 h, and wounds were prepared for histopathological examination. Photographs are from representative mice imaged at a magnification of ×400 (large panels) or ×630 (insets and right portion of panel B). Epidermis (E), s.c. fat (S), and muscle (M) tissues are indicated, as are V. vulnificus BAA87 rod-shaped bacteria (Vv) and infiltrating PMNs.

DISCUSSION

BT3 V. vulnificus is associated with an outbreak of infections in Israel, and these infections are apparently currently limited to Israel. These strains have been considered hybrids of two lineages of BT1 strains, the E strains enriched for environmental isolates (also known as lineage 2 and clade 1) and the C strains enriched for clinical isolates (also known as lineage 1 and clade 2) (24, 47, 48). In this study, we considered if these strains might be unique in their pathogenesis, as well as their genetics. As the V. vulnificus MARTX toxins are known to be variable in effector domain composition (21, 22), it was considered that any difference in pathogenesis determined could be linked to a unique toxin composition.

Sequencing revealed that all BT3 strains have nearly identical rtxA1 gene sequences. All of the BT3 strains examined are predicted to have the same unique arrangement of five effector domains. We cannot account for the variation compared to the previously published sequence of 11028 (22), as the 11028 isolate we obtained matched the other BT3 rtxA1 sequences examined. Our sequencing in the 5′ arm of BAA87 rtxA1 also differed from the rtxA1 gene of VVyb1 derived by whole-genome sequencing (31), as the latter had a 15-nt exact duplication increasing the length of the deduced protein by 5 aa. The lack of this duplication in BAA87 was confirmed, although it is unknown if this difference is due to an error in the automated assembly of VVyb1 or if this represents a true difference between the two sequences.

The effector domain repertoire of the BT3 MARTX toxin was distinct from all other previously annotated MARTX toxins, carrying two effector domains not previously known to be present in V. vulnificus MARTX toxins. DmX in the fifth position remains a DUF. This domain joins DUF1, RID, and ABH, which are shared with the BT1 strains and whose biological function also are not well characterized.

In this study, we show that the new domain in the fourth position is an adenylate cyclase—the first demonstration that this effector domain is a catalytically active cyclase enzyme. The acquisition of this domain resulted in V. vulnificus BT3-dependent cAMP production upon coculture with CHO cells. This gain of function was due to the MARTX toxin, since loss of the rtxA1 gene either by a naturally occurring null mutation in 2095/00 or by insertional mutagenesis in BAA87 eliminated the ability of V. vulnificus to induce cAMP production.

This activity was linked directly to the ExoY-like domain, which was able to convert ATP to cAMP in vitro. The activity depended on residues critical for catalysis by other adenylate cyclase toxins, confirming that this enzyme is a class II adenylate cyclase. Compared to the activity of purified PaExoY, it was less efficient and was not induced by the unknown activator present in CHO cell extract that was able to induce the activity of PaExoY. Variable activity of class II adenylate cyclases has been previously noted, as the adenylate cyclase activity of EF is ∼10- to 100-fold lower than that of BpACT, which also functions optimally at lower Ca²⁺ and CaM concentrations (49). It is unclear if the observed activity of VvExoY indicates a weak constitutively active enzyme or if the in vitro conditions identified for this initial report are simply not yet optimized. A problem with studying MARTX effectors is that the “ends” of the structurally stable domain can be difficult to predict because of long unstructured regions found between effector domains. This can lead to the inclusion of either too much or too little of the unstructured linker, affecting the stability of recombinant proteins. In addition, purified MARTX effectors tend, in general, to behave poorly in solution, as they seem to be evolutionarily selected to be unstable, likely because of a need to unfold easily for type I secretion out of the bacterium and again for translocation into the eukaryotic cell cytosol. ACD and CPD are both known to require cofactors (50, 51), and for the CPD, it has been demonstrated that the recombinant protein does not become stably folded in vitro until it associates with its activator (52). Thus, it is possible that VvExoY does require a cofactor that would increase both its in vitro stability and its activity. This activator will likely be different from that for PaExoY, as the same extract that stimulated PaExoY did not stimulate VvExoY and indeed seemed to inhibit it. Another distinct difference between the two proteins is that PaExoY has been shown to also be efficient in the conversion of GTP to cGMP (53). We found that under the reaction conditions employed here, VvExoY did not function as a guanylate cyclase either during coculture with CHO cells or in vitro (data not shown). However, as these are negative data, we cannot exclude the possibility that, under optimized conditions or with an appropriate activator, improved production of both cAMP and cGMP would be observed. Under the conditions detailed here, the enzyme functioned solely as an adenylate cyclase.

This gain of adenylate cyclase activity by the MARTX toxin will likely be shared with other species, as effector domains with homology to VvExoY are found in Vibrio MARTX toxins from V. cholerae, V. nigripulchritudo, and V. caribbeanicus (19). As the MARTX toxin effector domains seem to be freely exchangeable in the environment among the natural competent bacteria, it seems reasonable that this domain will appear via whole-genome sequencing in the MARTX toxins of even more species.

Since the effector domain repertoire of the BT3 MARTX toxin was distinct, it was determined if the toxin would be a virulence factor, as has been shown for BT1 and BT2 strains. It is shown here that the BT3 rtxA1 gene is essential for wound-induced infection in mice inoculated s.c. The wounds showed significant epithelial damage and inflammation. In the absence of rtxA1, 2/10 mice still succumbed to infection and showed some inflammation, although the overall pathology was much less severe. These results are consistent with other studies of BT1 V. vulnificus wound infection in mice, where the cytolysin encoded by the gene vvhA is also linked to virulence, and the potential remains that other secreted proteases and elastase contribute to these infections as well (17).

In a final set of experiments, we attempted to compare the pathology of the wounds induced by BT1 strain CMCP6 and the pathology of the wounds induced by BT3 strain BAA87 to determine if the conversion from a BT1-type MARTX toxin to the adenylate cyclase toxin would affect qualitative aspects of the wound. However, the observed differences in inflammatory infiltration and the degree of damage could not be assigned specifically to the BT3 rtxA1 gene (our unpublished results). These findings indicated that it is the sum of all of the variations in the overall genome context, in addition to changes in the MARTX content, that caused the difference in disease progression between BT1 CMPC6 and BT3 BAA87. This important finding demonstrates that V. vulnificus infection is highly variable and strain specific and can involve changes in many genetic components that dictate virulence.

This vast heterogeneity of strains may pose a problem, as seen in this outbreak in Israel, for there are potentially regional differences in the nature of the infections, and thus, pathogenic findings based on studies of a single strain may not be globally applicable. This may also pose challenges for the development of therapeutic agents. For example, it has recently been determined that an antibody raised against the carboxyl-terminal fragment of the MARTX toxin from BT1 strain MO6-24/O can be used in mice as a passive treatment for infections with that same strain of V. vulnificus (54). However, if a patient has a strain with MARTX toxin epitopes different from those of the strain used to generate the antibody, the therapy may be limited in its application. These regional differences in strain and the resulting therapeutic barriers have the potential to then be exported as food markets expand globally, just as the incidence of V. vulnificus is also expanding globally because of climate change.