Characterization of pertussis-like toxin from Salmonella spp. that catalyzes ADP-ribosylation of G proteins

Salmonella Typhimurium definitive phage type (DT) 104 produces a pertussis-like toxin (ArtAB-DT104), which catalyzes ADP-ribosylation of pertussis toxin sensitive G proteins. However, the prevalence of ArtAB and its toxicity have not been established. We report here that, in addition to DT104, S. Worthington, and S. bongori, produce ArtAB homologs, designated ArtAB-SW and ArtAB-Sb, respectively. We purified and characterized these ArtAB toxins, which comprise a 27-kDa A subunit (ArtA) and 13.8-kDa pentameric B subunits (ArtB). While the sequence of the A subunit, which is ADP-ribosyltransferase, is similar to the A subunit sequences of other ArtABs, the B subunit of ArtAB-Sb is divergent compared to the B subunit sequences of other ArtABs. Intraperitoneal injection of purified ArtABs was fatal in mice; the 50% lethal doses of ArtAB-DT104 and ArtAB-SW were lower than that of ArtAB-Sb, suggesting that ArtB plays an influential role in the toxicity of ArtABs. ArtABs catalyzed ADP-ribosylation of G proteins in RAW 264.7 murine macrophage-like cells, and increased intracellular cyclic AMP levels. ArtAB-DT104 and ArtAB-SW, but not ArtAB-Sb, stimulated insulin secretion in mice; however, unlike Ptx, ArtABs did not induce leukocytosis. This disparity in biological activity may be explained by differences in ADP-ribosylation of target G proteins.

of 2, 8, and 0.5 μg/well were required for the hemagglutination (HA) activity of ArtAB-DT104, ArtAB-SW, and Ptx, respectively, and no activity was observed for ArtAB-Sb (Fig. S2A). Furthermore, 100 ng of ArtAB-DT104 and ArtAB-SW induced CHO-K1 cell clustering as Ptx did, which did not occur for 1 μg of ArtAB-Sb (see Supplementary Fig. S2,B). ArtABs enhance isoproterenol-induced cyclic (c)AMP activation in macrophage-like cells. Ptx increases basal and humoral adenylate cyclase activity, which results in cAMP accumulation in the host cells. We next examined whether ArtABs could increase cAMP formation, and we observed intracellular cAMP levels within 15 min of treatment with isoproterenol (ISO) in the presence of the phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine (IBMX) in RAW 264.7 murine macrophage-like cells cultured with 0-500 ng of ArtABs. Cells treated with ArtAB without ISO stimulation served as a negative control. Incubation with ISO increased cAMP levels relative to those in unstimulated cells (Fig. 4A,C), and 500 ng/ mL ArtAB-DT104 increased cAMP levels relative to the levels in cells treated with control buffer (P < 0.05; Fig. 4C). Lysophosphatidic acid (LPA) did not induce cAMP accumulation (Fig. 4B); however, dose-dependent, ISO-specific ArtAB-DT104-induced cAMP activation was clearly observed in the presence of LPA (Fig. 4D). The cAMP levels observed in the presence of 100 and 500 ng/ml ArtAB-DT104 differed from those observed without ArtAB-DT104 ( Fig. 4A and D). Similar results were obtained in the presence of ArtAB-SW and ArtAB-Sb, but not in the presence of Ptx (Fig. 4E-G).

ADP-ribosylation of RAW 264.7 cell membrane proteins by ArtABs in vitro and in vivo.
ADP-ribosylation of G i protein by Ptx results in cAMP accumulation in cells 16 . Therefore, we examined the ability of the ArtA and ArtAB proteins to promote ADP-ribosylation of G i proteins in RAW 264.7 cell membranes, since their cAMP accumulation was observed by treatment with ArtABs. In the membranes of RAW 264.7 cells, ArtA catalysed the ADP-ribosylation of a 41-kDa protein that was the same size as the α subunit of the Ptx-sensitive heterotrimeric G proteins G i , G o , and G t from the bovine brain (Fig. 5A). Western blot analysis of the membrane fraction of RAW 264.7 cells and bovine brain G proteins using an antiserum that recognizes the α subunits of heterotrimeric G i proteins (both G αi1 and G αi2 ) showed that ADP-ribosylated G α , which migrated more slowly than non-ADP-ribosylated G α due to the modification contained G αi (Fig. 5A). These results indicate that the G αi protein is a substrate for ADP-ribosylation by ArtAB in vitro. We investigated whether ADP-ribosylation of G proteins occurs in intact RAW 264.7 cells by incubating the cells with various concentrations of ArtAB-DT104, and measuring the decrease in the amount of G protein available for biotin-nicotinamide adenine dinucleotide (NAD) in vitro. The membrane fraction containing G proteins was used to detect in vitro ADP-ribosylation using ArtA expressed in vitro, since ADP-ribosylation is more active using this as compared to purified ArtAB. While untreated cells contained G proteins that can be ADP-ribosylated by ArtA-DT104, incubation with increasing concentrations of ArtAB progressively reduced the amount of available ArtA-DT104-modifiable G protein (Fig. 5B). These results demonstrate that ArtAB-DT104 enters and modifies the G proteins of intact RAW 264.7 cells. Similar results were obtained with ArtAB-SW, ArtAB-Sb, and Ptx (see Supplementary Fig. S3). Along with Ptx, all the ArtA homologs catalysed ADP-ribosylation of the 41-kDa protein in the RAW 264.7 cell membranes in vitro (Fig. 5C); as noted above, one of these proteins is likely G αi . To determine whether the target proteins of ArtAB-DT104 are the same as those of ArtAB-Sw, ArtAB-Sb, and Ptx in RAW 264.7 cells, a cellular ADP-ribosylation assay was carried out. RAW 264.7 cells were treated with 500 ng/ml ArtAB-DT104 or Ptx, and their membrane fractions were used for an in vitro ADP-ribosylation assay with in vitro-translated ArtAs or Ptx. ArtAB-DT104 modified almost all of the available target proteins of ArtA-SW, ArtA-Sb, and Ptx, as evidenced by the absence of a protein band (Fig. 5D). However, preincubation of RAW 264.7 cells with Ptx did not fully block modification of the 41-kDa protein by ArtAs, indicating that ArtA might have ADP-ribosylated additional G i -proteins that was not ribosylated by Ptx (Fig. 5E).

Discussion
We examined the prevalence of Salmonella pertussis-like toxins (ArtABs), and found that artAB was present in 97.5% (237/243) of Salmonella Typhimurium DT104 isolates, but only 3.6% (11/303) of non-DT104 isolates. We also found that two other serotypes, S. Worthington and S. Agoueve, and an additional species, S. bongori, that harboured artAB homologs. artAB is located on a prophage in S. Typhimurium DT104, and a number of other toxin genes are encoded by phages; such genes are typically replicated and transcriptionally activated after prophage induction 17,18 . Shiga toxin (Stx) genes in certain Escherichia coli strains and cholera toxin (Ctx) genes in Vibrio cholera are located within prophages that are induced by agents that increase toxin production, such as MMC 18 . Although we did not examine the location of artAB in non-DT104 S. Typhimurium, S. Worthington, or S. bongori, similar to ArtAB-DT104, ArtAB production in these strains was induced by MMC, suggesting that these genes are also encoded by prophages. The pertussis-like toxins ArtAB-DT104, ArtAB-SW, and ArtAB-Sb of S. Typhimurium, S. Worthington, and S. bongori, respectively, were purified to homogeneity from MMC-treated culture supernatants, and all three showed potent ADP-ribosyltransferase activity for Ptx-sensitive G proteins. ADP-ribosylation by all three ArtABs was decreased in the absence of DTT, suggesting that ArtA-associated ADP-ribosyltransferase activity depends on thiol reduction, as in the cases of Ctx and Ptx [19][20][21] .
Ptx belongs to the AB 5 family of toxins with enzymatic A subunits that disrupt essential host functions and pentameric B subunits that mediate toxin uptake into target cells. AB 5 toxins also include Ctx, E. coli heat-labile enterotoxins (LT), Stx, and subtilase cytotoxin (SubAB); the latter is produced by certain Shiga-toxigenic E. coli strains 11,22 . The B subunit of Ptx is a heteropentamer comprising four different monomers (S2, S3, two copies of S4, and S5) 21,23 . ArtAB also belongs to the AB 5 family; however, ArtAB has a homopentameric B subunit, similar to LT, Ctx, Stx, and SubAB.
We demonstrated that the purified ArtAB is toxic to mice; LD 50 values were lower for ArtAB-DT104 and ArtAB-SW than ArtAB-Sb. The A subunits of ArtAB-SW and ArtAB-Sb had high sequence identity to that of ArtAB-DT104; all three proteins catalysed ADP-ribosylation of the 41-kDa α-subunits of Ptx-sensitive heterotrimeric G proteins. The B subunits of ArtAB-DT104 and ArtAB-SW were 83% identical, in contrast to the 26% identity between ArtAB-DT104 and ArtAB-Sb. Receptor binding mediated by the B subunit at the cell surface is an essential first step in the intoxication of target cells by AB 5 toxins. This triggers internalization and intracellular trafficking, allowing the catalytic A subunit to access its substrate. Although the mechanisms underlying the differences in potency among these ArtAB toxins are unclear, B subunit activities, such as receptor recognition or toxin internalization, may play a role. ArtAB-DT104 and ArtAB-SW, but not ArtAB-Sb, induced HA of chicken erythrocytes in vitro; in Ptx, this activity is associated with the B oligomer 24 . Furthermore, CHO cells treated with ArtAB-DT104 and ArtAB-SW exhibited clustered growth, similar to cells treated with Ptx. ArtAB-Sb did not have this effect, suggesting that its B subunit may not be recognized by the receptor, and is therefore not internalized by CHO cells.
Ptx catalyses the ADP-ribosylation of the α subunits of the heterotrimeric G i/o protein family, resulting in their inactivation and signal transduction involved in intracellular processes, such as cAMP formation. Ptx exerts pathological effects via cAMP accumulation in host cells 16,21,24,25 . In this study, preincubation of intact RAW 264.7 cells with ArtAB blocked ArtAB-induced ADP-ribosylation of the 41-kDa α subunits of heterotrimeric G i proteins in vitro, suggesting that these proteins are substrates for ArtAB toxins in vivo. Additionally, incubation of ArtAB-treated RAW 264.7 cells with the β-adrenoreceptor agonist isoproterenol -a specific ligand for G s protein-coupled receptors -leads to dose-dependent stimulation of cAMP production. This effect was enhanced by LPA, an agonist of G i protein-coupled receptors, and inhibits adenylyl cyclase via G i 26 . Taken together, although all the targets of ArtA are not known, we showed that ArtAB treatment of cells leads to cAMP accumulation at least via inhibition of G i signalling, similar to Ptx. In a previous study, treatment of G proteins modified by ArtA with HgCl 2 , which cleaves cysteine ADP-ribose bonds, released labelled ADP-ribose, suggesting that ArtA ADP-ribosylates G proteins at a cysteine residue 9 .
The biological effects of Ptx-mediated ADP-ribosylation of heterotrimeric G protein α subunits (G αi/o ) include histamine sensitization, stimulation of insulin secretion, and lymphocytosis 24 . All three ArtAB types contain the following three motifs that are conserved in bacterial ADP-ribosylating toxins [27][28][29][30] : (i) a putative catalytic glutamate at position 115, (ii) the β/α region with a motif (at positions 50-52) needed for the structural integrity of the NAD-binding site, and (iii) a conserved arginine residue at position 6, which is required for NAD binding in many ARTs (see Supplementary  Fig. S1,C). All three ArtAB toxin types catalysed ADP-ribosylation of Ptx-sensitive G proteins. ArtAB-DT104 also induced the activation of insulin secretion; however, this toxin did not exhibit leukocytosis-inducing activity. Two Ptx-sensitive target proteins, G αi2 and G αi3 , have been reported in CHO cells, and our previous report revealed differences between ArtA and Ptx with respect to reactivity against these proteins. Additionally, preincubation of RAW 264.7 cells with Ptx did not fully block subsequent ADP-ribosylation of the 41-kDa protein by ArtAB, suggesting that multiple G proteins of this size are substrates for ArtA, but not Ptx. The 41-kDa Ptx substrate in RAW 264.7 cells is a combination of G αi and G αo

31
. G αi includes different isoforms of G αi1 , G αi2 , and G αi3 that have similar molecular weights 32 . Although it is unclear which isoform in RAW 264.7 cells is a substrate for ArtAB, differences in the biological activity of ArtAB and Ptx may be explained by their variable reactivities to target proteins. Homologs of Ptx have also been identified in S. Typhi, and are referred to as pertussis-like toxins A and B (PltA and PltB, respectively) [12][13][14] . ArtA exhibits 61% identity to PltA for a 224-a.a. segment, while ArtB is 30% identical for a 114-a.a. region. PltA and PltB are required for the delivery of cytolethal distending toxin (Cdt)B from an intracellular compartment into target cells via autocrine and paracrine pathways 12 . CdtB forms a complex with PltA and PltB, which are designated as typhoid toxins. pltA, pltB, and cdtB have also been reported in non-typhoidal Salmonella serotypes, such as Montevideo, Schwarzengrund, Bredeney, and Javiana 33,34 . Therefore, pertussis-like toxins in Salmonella can be divided into ArtAB and PltAB types. Although ArtA is homologous to PltA in S. Typhi, the target of the ADP-ribosylating activity of PltA is a 100-kDa protein present in the postnuclear supernatant fraction of Henle-470 cells, which is distinct in size from the targets of ArtA and Ptx 12 . Therefore, to our knowledge, ArtA is the only known Salmonella enzyme that catalyses ADP-ribosylation of mammalian G proteins.
We previously found that not only MMC but also hydrogen peroxide, an oxidative stressor, induces ArtAB production 9 . Reactive oxygen species generated and released by host cells, such as macrophages or leukocytes, may induce artAB expression by intracellular bacteria. Since we do not yet have evidence for ArtAB expression in vivo, further studies on the regulation and expression of the toxin are required to clarify the importance of ArtAB in the virulence of Salmonella, especially S. Typhimurium DT104.

Methods
Bacterial strains and animals. The panel of 243 isolates of S. Typhimurium DT104, 302 isolates of non-DT104 strains from our laboratory strain culture collection, and NCTC 73 (see Supplementary Table S1) were previously described by Tamamura et al. 35 . Apart from serotype Typhimurium, 83 serotypes, including Worthington strain 182 and Agoueve strain 213, as well as S. bongori strain ATCC43975, are listed in Supplementary Table S2.
Japanese albino rabbits, weighing 1 kg, and 3-week-old female BALB/c mice were purchased from Hokudo Co. (Sapporo, Japan). Animals were maintained and used in accordance with the Guidelines for the Care and Use of Laboratory Animals of the National Institute of Animal Health.

Cells. CHO cells and the RAW 264.7 murine macrophage cell line were obtained from the American Type
Culture Collection (Manassas, VA, USA). CHO cells were cultured in Ham's F12 medium, and RAW 264.7 cells were cultured in Dulbecco's Minimal Essential Medium supplemented with 10% foetal calf serum, 0.1 mg/ml streptomycin, and 100 U/ml penicillin G at 37 °C in a humidified atmosphere of 5% CO 2 .
Purification of ArtAB. Salmonella cultures were grown overnight in syncase broth 37 supplemented with FeCl 3 to a final concentration of 10 μg/ml. Cultures were diluted 1:40 in 20 ml of syncase broth and grown for 2.5 h at 37 °C on a shaker at 120 rpm; mitomycin C (Sigma, St. Louis, MO, USA) was next added to a final concentration of 0.5 μg/ml, and cultivated for 16 h. Cells were concentrated by centrifugation; the supernatant was passed through 0.22-μm pore filters, and analysed by chromatography. ArtABs (except for ArtAB-Sb) were purified by consecutive elution from Affi-Gel Blue (Bio-Rad, Hercules, CA, USA) and hydroxyapatite columns followed by hydrophobic interaction chromatography (HIC), as previously described for Ptx purification 38 . Desalting or buffer exchange of protein solutions was performed using a PD-10 Column (GE Healthcare).
Culture supernatant filtrate (400 ml) was acidified to pH 6.0 with concentrated HCl before adding 6 ml of Affi-Gel Blue equilibrated with 0.25 M phosphate buffer (pH 6.0). The resultant slurry was stirred at 4 °C for 18 h. The resin was allowed to settle for 1 h and after siphoning off the supernatant solution it was packed into an empty 10-ml column with a Luer lock connection (MoBiTec, Gottingen, Germany) and washed with 0.25 M sodium phosphate (pH 6.0) and then 0.05 M Tris-HCl (pH 7.4), before elution with 0.05 M Tris-HCl (pH 7.4) containing 0.75 M MgCl 2 . Fractions were collected and assayed by western blotting using a rabbit antibody against a 14-a.a. peptide corresponding to the Arg 10 -His 23 sequence of S. Typhimurium DT104 ArtA. Fractions containing ArtAB were pooled, and buffer exchange was performed with 10 mM phosphate buffer (pH 6.0). Proteins were loaded on a 5-ml CHT-I hydroxyapatite column (Bio-Rad) equilibrated with 10 mM phosphate buffer (pH 6.0), washed with 10 mM phosphate buffer (pH 6.0) followed by 0.1 M phosphate buffer (pH 7.0), and eluted with 0.1 M phosphate buffer (pH 7.0) containing 0.5 M NaCl. The buffer in the eluted fraction containing ArtAB was exchanged with 1.5 M ammonium sulphate in 50 mM phosphate buffer (pH 7.0). Then, the proteins were loaded on to a Resource PHE Hydrophobic Interaction Column (GE Healthcare), and eluted with a linear gradient of 1.5 to 0 M ammonium sulphate in 50 mM phosphate buffer solution (pH 7.0) using a fast protein liquid chromatography system (GE Healthcare). The ArtAB fraction was desalted, concentrated in phosphate-buffered saline (PBS), and stored at −80 °C. Since ArtAB from S. bongori did not bind to the hydroxyapatite gel under these conditions, purification was performed using a Superpose 6 Column (10/300GL; GE Healthcare) with 50 mM phosphate buffer (pH 7.0) and 0.15 M NaCl, after Affi-Gel Blue chromatography, without HIC. For the preparative separation of ArtAB from DT104, the MonoQ Column (GE Healthcare) was loaded with purified ArtAB in a buffer containing 30 mM Tris-HCl (pH 8.8). Proteins were eluted with 20 ml of a linear gradient (0-500 mM) of NaCl in the same sample buffer. Protein concentrations were determined using a Bio-Rad protein assay, with bovine serum albumin as a standard.
Antibody production. Rabbit antibody against the 14-a.a. peptide corresponding to the Arg 10 -His 23 sequence of S. Typhimurium DT104 ArtA was produced by Sigma Genosys (Ishikari, Japan) as described by Uchida et al. 9 . Antibodies against ArtAB, ArtA, and ArtB were obtained from rabbits and mice injected subcutaneously with a 1:1 solution of purified antigen and Titer-Max-Gold (CytRx, Los Angeles, CA, USA) adjuvant three times every 2 weeks. Animals were bled 2 weeks after the last injection. IgG in sera from immunized animals was purified using a HiTrap Protein G HP Column (GE Healthcare) following the manufacturer's instructions. Anti-G αi rabbit serum was purchased from Calbiochem (San Diego, CA, USA).
Sandwich enzyme-linked immunosorbent assay (ELISA). ArtAB in the culture filtrate was measured by sandwich (capture) ELISA. The plate was coated with rabbit anti-ArtAB IgG (10 μg/ml in PBS containing 137 mM NaCl, 2.7 mM KCl, and 10 mM phosphate buffer at 100 μl/well) at 4 °C for 18 h. The plate was blocked with 3% bovine serum albumin in PBS containing 0.02% NaN 3 (blocking buffer; 400 μl/well) at room temperature for at least 2 h and washed twice with 400 μl of PBS. Culture supernatant was passed through a 0.22-μm membrane filter, and standards were prepared with the blocking buffer; these preparations were added at 50 μl/well, incubated at room temperature for 18 h, and washed four times with PBS (400 μl). Mouse anti-ArtAB IgG (diluted 1:200 in blocking buffer; 100 μl/well) was added and incubated at room temperature for 2 h, then removed by washing three times (400 μl per wash) with PBS. Goat anti-mouse IgG conjugated with horseradish peroxidase (diluted 1:500 in blocking buffer; 100 μl/well) was added and incubated at room temperature for 2 h, followed by three washes (400 μl per wash) with PBS. Tetramethylbenzidine-H 2 O 2 (TMB Peroxidase EIA Substrate Kit; Bio-Rad) was added, and the reaction was terminated by adding 100 μl of 1 N H 2 SO 4 . The optical density at 450 nm was recorded.

SDS-PAGE and immunoblotting.
Proteins were separated by 12.5% or 15% SDS-PAGE on a 0.1% SDS-Tris-glycine running buffer system and stained with Coomassie blue or silver using a silver staining kit (Daiichi Pure Chemical, Tokyo, Japan). For western blotting, SDS-PAGE gels were transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad) in a buffer containing 20% methanol, 25 mM Tris (pH 8.3), and 192 mM glycine. Membranes were incubated for 30 min in blocking buffer consisting of 1% Western Blocking Reagent (Roche Molecular Diagnostics, Indianapolis, IN, USA) in maleic acid buffer (100 mM maleic acid and 150 mM NaCl, adjusted to pH 7.5 with NaOH). The membranes were incubated with rabbit antiserum (1:1000 in maleic acid buffer), and then alkaline phosphatase-conjugated anti-rabbit IgG (Bio-Rad) diluted 1:10,000 in maleic acid buffer. Bands were visualized using a chemiluminescent substrate kit (Bio-Rad).
In vitro translation. In vitro transcription/translation of ArtA was performed using Pure System S-S (Post Genome Institute, Tokyo, Japan) as described by Uchida et al. 9 . Template DNA was generated by two-step PCR according to the manufacturer's instructions. Mature ArtA was amplified from chromosomal DNA, and the T7 promoter sequence was amplified. In vitro transcription/translation was initiated by adding PCR products from the second step to the Pure System reaction mixture, and incubating at 37 °C for 1 h.

Cell membrane preparation.
To prepare cell membranes, cells grown in 75-cm 2 flasks were washed twice with PBS, scraped from the flasks, and centrifuged for 5 min at 600 × g. Pellets were resuspended in a solution of 0.25 M sucrose, 25 mM Tris-HCl (pH 7.5), and 5 mM MgCl 2 , and homogenized using a sample-grinding kit (GE Healthcare). Homogenates were centrifuged at 600 × g for 10 min to remove nuclei and unbroken cells, and the supernatant was centrifuged at 40,000 × g for 20 min; the sediment that separated from the supernatant fraction was resuspended in a solution of 20 mM Tris-HCl (pH 7.5) and 5 mM MgCl 2 and stored at −80 °C. ADP-ribosylation assay. ADP-ribosylation of pertussis toxin-sensitive G proteins from the bovine brain (Calbiochem-Novabiochem, San Diego, CA, USA) or cell membrane proteins was determined using biotin-NAD (Trevigen, Gaithersburg, MD, USA), as described by Uchida et al. 9 . The in vitro ADP-ribosyltransferase reaction mixture (20 μl) contained 0.1 M Tris-HCl (pH 7.6), 0.1 mM ATP, 20 mM DTT, 5 mM thymidine, 10 μM β-NAD, and 0.1 μg of G protein from the bovine brain (Calbiochem-Novabiochem) or 5.6 μg of membrane proteins isolated from RAW 264.7 cells, and either the test substance or Ptx (Biomol, Hamburg, Germany). Ptx was preactivated by incubation in 50 mM Tris-HCl (pH 7.5) containing 50 mM DTT at 37 °C for 20 min. The reaction proceeded for 1 h at 37 °C and was terminated by adding an equal volume of 2× SDS-PAGE sample buffer. Samples were resolved by 12.5% SDS-PAGE. Biotin-ADP-ribosylated proteins were transferred to PVDF membranes, incubated in blocking buffer consisting of 1% Western Blocking Reagent in maleic acid buffer for 30 min, followed by peroxidase-conjugated streptavidin (Vector Laboratories, Burlingame, CA, USA) diluted 1:20,000 in maleic acid buffer for 1 h. Bands were visualized using an enhanced chemiluminescence kit (GE Healthcare) according to the manufacturer's instructions. ImageJ v.1.6 (National Institutes of Health, Bethesda, MD, USA) was used for the densitometric analysis of band intensity. To assess the efficacy of ADP-ribosylation of G proteins by the toxins in vivo, a 'back ADP-ribosylation' experiment was performed by pretreating intact cells with toxin, preparing cell membranes from these pretreated cells, and using them in an in vitro ADP-ribosylation experiment as described above 39 . Mouse lethality. Female BALB/c mice (4 weeks old; n = 5-8) were administered 0.2 ml of sample in graded doses by i.p. injection. Deaths were recorded daily. LD 50 was calculated as described by Reed 40 . ArtAB neutralization in BALB/c mice was assayed by i.p. injection of 0.2-ml mixtures of 2 μg of ArtAB and 100 μg of rabbit anti-ArtAB IgG preincubated at 37 °C for 18 h. As a control, mice were injected with the same amount of ArtAB incubated with IgG from non-immunized rabbit serum.
Cell-clustering activity. Clustering of CHO-K1 cells was evaluated as described by Hewlett et al. 41 . Briefly, CHO cells grown to confluence in flasks were trypsinized and diluted in F-12 medium with 5% foetal calf serum to a concentration of approximately 2 × 10 4 cells/ml. A 150-μl volume aliquot of the suspension was added to 8 wells of a flat-bottomed microtiter plate. After allowing 4 h for attachment and stabilization, the test substance (150 μl) was added.
Hemagglutination (HA) test. The HA test was performed as described elsewhere 15 . Briefly, 50 μl of 0.7% chicken erythrocytes in PBS (v/v) was added to 50 μl of the test substance or Ptx serially diluted with PBS in a microplate. The preparation was incubated at room temperature for 60 min. The minimum amount of sample causing complete agglutination of the red blood cells was recorded.

Leukocytosis-promoting and islet-activating activities. Assays for both leukocytosis and
islet-activating activity have been described elsewhere 15 . Briefly, 5-week-old female BALB/c mice (n = 4-10) were administered the test substance by i.p. injection. On day 3, tail vein blood samples were drawn and leukocytes were counted. To measure islet-activating activity, serum insulin in day 3 blood samples was measured 15 min after injection of 50% glucose (0.5 ml) using an ELISA kit (Morinaga Institute of Biological Science, Yokohama, Japan) according to the manufacturer's instructions.
Toxin treatment and cAMP assessment. RAW 264.7 cells were harvested and subcultured in a 24-well plate at a density of 1.0 × 10 5 cells/well. After 18 h, the medium was replaced with 0.5 ml of fresh medium containing toxins. Cultures were incubated for an additional 24 h and washed twice with 1 ml of buffered saline solution containing 137 mM NaCl, 5 mM KCl, 5.6 mM glucose, 1 mM EGTA, and 5 mM HEPES (pH 7.4). After washing, 0.4 ml of buffered saline supplemented with 1 mM MgCl 2 and 1 mM IBMX was added to the cultures, which were incubated at 37 °C for 30 min; this incubation was followed by a 15-min incubation at 37 °C with 10 μM isoproterenol or 10 μM forskolin and/or 50 μM LPA. The medium was next removed by aspiration and cells were washed with PBS. We next added 0.4 ml of lysis reagent from the cAMP Enzyme Immunoassay System Kit (GE Healthcare). The cAMP assay was performed according to the manufacturer's instructions.
Statistical analysis. Differences were evaluated using one-way ANOVA with Tukey's test or Dunnett's test.
All P values were calculated using GraphPad Prism version 6.0 and were interpreted as significant at values less than 0.05. Study approval. All animal procedures were carried out in strict accordance with local guidelines and with ethical approval from the National Institute of Animal Health.