The Eukaryotic Host Factor 14-3-3 Inactivates Adenylate Cyclase Toxins of Bordetella bronchiseptica and B. parapertussis, but Not B. pertussis

ABSTRACT Bordetella pertussis, Bordetella bronchiseptica, and Bordetella parapertussis share highly homologous virulence factors and commonly cause respiratory infections in mammals; however, their host specificities and disease severities differ, and the reasons for this remain largely unknown. Adenylate cyclase toxin (CyaA) is a homologous virulence factor that is thought to play crucial roles in Bordetella infections. We herein demonstrate that CyaAs function as virulence factors differently between B. bronchiseptica/B. parapertussis and B. pertussis. B. bronchiseptica CyaA bound to target cells, and its enzyme domain was translocated into the cytosol similarly to B. pertussis CyaA. The hemolytic activity of B. bronchiseptica CyaA on sheep erythrocytes was also preserved. However, in nucleated target cells, B. bronchiseptica CyaA was phosphorylated at Ser375, which constitutes a motif (RSXpSXP [pS is phosphoserine]) recognized by the host factor 14-3-3, resulting in the abrogation of adenylate cyclase activity. Consequently, the cytotoxic effects of B. bronchiseptica CyaA based on its enzyme activity were markedly attenuated. B. parapertussis CyaA carries the 14-3-3 motif, indicating that its intracellular enzyme activity is abrogated similarly to B. bronchiseptica CyaA; however, B. pertussis CyaA has Phe375 instead of Ser, and thus, was not affected by 14-3-3. In addition, B. pertussis CyaA impaired the barrier function of epithelial cells, whereas B. bronchiseptica CyaA did not. Rat infection experiments suggested that functional differences in CyaA are related to differences in pathogenicity between B. bronchiseptica/B. parapertussis and B. pertussis.

and Bordetella bronchiseptica, which also cause respiratory infections in various host animals with severe to mild coughing. A genomic analysis indicated that B. pertussis and B. parapertussis independently evolved from a B. bronchiseptica-like ancestor (3). Classical Bordetella is known to share many virulence factors, including adenylate cyclase toxin (CyaA), dermonecrotic toxin (DNT), and filamentous hemagglutinin, but not pertussis toxin, which is specific to B. pertussis. The molecular actions of these major virulence factors have been extensively examined. However, the pathogenesis of Bordetella infections has yet to be elucidated in detail (4). Furthermore, disease-causing propensities and host specificities apparently differ among classical Bordetella (5). B. pertussis is a strictly human pathogen that causes acute and severe respiratory infections with paroxysmal coughing and vomiting. B. parapertussis infects sheep and humans and is associated with a milder pertussis-like disease. B. bronchiseptica exhibits the broadest host range of mammals and leads to chronic and often asymptomatic infections. Limited information is currently available on how these three species exhibit different characteristics as pathogenic bacteria despite sharing many homologous virulence factors.
CyaA, one of the virulence factors commonly produced by classical Bordetella, is considered to play a key role in the establishment of infection (5). The toxin consists of two functional modules: the N-terminal adenylate cyclase domain (ACD), which is activated by eukaryotic calmodulin and causes supraphysiological cyclic AMP (cAMP) accumulation after being translocated into the cytosol of target cells, and the Cterminal RTX (repeats-in-toxin) domain, which is responsible for binding to target cells and organizing cation-selective toxin pores on the cell membrane (6). The toxin affects myeloid cells expressing CD11b/CD18 and subverts host immunity by inhibiting phagocytosis, chemotaxis, and superoxide production and modulating dendritic cell maturation and inflammatory cytokine/chemokine production (7)(8)(9)(10)(11)(12)(13). CyaA also lyses erythrocytes and induces apoptosis in macrophages (14)(15)(16). Moreover, previous studies showed that epithelial cells may be targets of CyaA (17,18).
Since CyaA is highly conserved among classical Bordetella, its function and role have been considered to be similar in bacterial infections. However, we herein demonstrated that B. bronchiseptica CyaA, in contrast to B. pertussis CyaA, exhibited only weak toxicity in nucleated cells. After being translocated into target cells, B. bronchiseptica CyaA was phosphorylated in the cytosol and associated with the eukaryotic host factor 14-3-3, which resulted in the abrogation of adenylate cyclase activity. To the best of our knowledge, this is the first study to demonstrate that a virulence factor of pathogens is inactivated by 14-3-3 in target cells. Our results also suggest that a difference in cytotoxic effects between B. pertussis CyaA and B. bronchiseptica CyaA influences the pathogenicity of each Bordetella species in a rat infection model.

RESULTS
CyaA of B. pertussis, but not CyaA of B. bronchiseptica or B. parapertussis, induces morphological changes and cAMP accumulation in L2 cells. We previously reported that B. pertussis CyaA caused cell rounding in L2 cells through its adenylate cyclase activity (19): anti-CyaA serum neutralized cell-rounding activity of the culture supernatant of B. pertussis, indicating that CyaA (B. pertussis CyaA) is the sole cause of cell rounding. In contrast, we found that the culture supernatants of B. bronchiseptica and B. parapertussis, which contain CyaA, did not alter cell morphology ( Fig. 1A and B; see Fig. S1A in the supplemental material). Ten strains each of B. pertussis and B. bronchiseptica that were maintained in the laboratory were similarly examined for their cell-rounding activities. Although the culture supernatants of B. pertussis strains caused cell rounding, a result that is consistent with previous findings (19), the culture supernatants of all the B. bronchiseptica strains tested did not (data not shown). These results indicate that a discrepancy in CyaA toxicity was commonly observed between B. pertussis and B. bronchiseptica but was not due to some unusual strains. When L2 cells were treated with purified preparations of recombinant CyaAs, similar results were obtained: B. pertussis CyaA induced cell rounding in a dose-dependent manner, while B. bronchiseptica CyaA did not at any of the concentrations tested (Fig. 1C). As previously reported (19), B. pertussis CyaA K58Q, which is defective in adenylate cyclase activity, did not cause cell rounding. We then examined the enzyme activities of B. pertussis CyaA and B. bronchiseptica CyaA to generate cAMP from ATP in vitro. Both CyaAs similarly produced cAMP, indicating that the enzyme activity of B. bronchiseptica CyaA is intact (Fig. 2). In contrast, when cultured cells were treated with the toxins, a marked increase in intracellular cAMP was observed only in B. pertussis CyaA-treated cells (Fig. 2). These results indicate that B. bronchiseptica CyaA is enzymatically active in vitro, but it does not appear to exhibit cytotoxicity that is mediated by the accumulation of intracellular cAMP.
Phe 375 is the key residue for the cytotoxicity of CyaA. B. bronchiseptica CyaA and B. pertussis CyaA show 98% homology with 36 amino acid replacements in the overall sequence. In order to identify the crucial structural differences responsible for the different cytotoxic activities of B. bronchiseptica CyaA and B. pertussis CyaA, we initially attempted to localize the region of B. pertussis CyaA necessary for cell-rounding activity. We examined chimeric CyaAs, in which the RTX domains were mutually exchanged between B. pertussis CyaA and B. bronchiseptica CyaA, for their cell-rounding activities (Fig. 3A)  Inactivation of CyaA by 14-3-3 ® we experimentally confirmed that this was not the case and that the translation start site of B. bronchiseptica CyaA was identical to that of B. pertussis CyaA (Fig. S2B). In order to identify which amino acid residue is crucial for the cytotoxicity of B. pertussis CyaA, we generated recombinant CyaAs by replacing the amino acid residue of the chimeric B. pertussis-B. bronchiseptica CyaA with that of B. bronchiseptica CyaA at each position and then examined their cell-rounding activities ( Fig. 3C and S2A). The recombinant CyaAs F375S and V800A, the amino acids of which were exchanged with those of B. bronchiseptica CyaA at amino acid positions 375 and 800, respectively, failed to cause cell rounding, whereas the activities of other recombinant CyaAs were intact. We and Ser 375 , respectively. In contrast, Val 800 was conserved in B. pertussis, whereas Ala 800 was not conserved in B. bronchiseptica: 43 out of 87 deposited data sets of B. bronchiseptica CyaA showed Ala 800 , whereas the others showed Val 800 . These results indicate that the 375th amino acid replacement is phylogenetically conserved, and Val 800 appears to be involved in the cytotoxic activity of B. pertussis CyaA for unknown reasons. Thus, we focused on the amino acid residue at position 375.
B. bronchiseptica CyaA is not defective in each intoxication step on target cells. In order to elucidate the reason why B. bronchiseptica CyaA does not exert cytotoxic activity, we examined purified preparations of B. bronchiseptica CyaA, B. pertussis CyaA, B. bronchiseptica CyaA S375F , and B. pertussis CyaA F375S for actions in each intoxication step ( Fig. 4 and Fig. S1B). Following their addition to a culture of L2 cells, B. pertussis CyaA and B. bronchiseptica CyaA S375F increased intracellular cAMP levels, whereas B. bronchiseptica CyaA and B. pertussis CyaA F375S did not, which is consistent with the results of the cell-rounding assay ( Fig. 3D and 4A). However, the in vitro enzyme activities of these CyaA preparations were intact ( Fig. 2 and 4A). The actions of CyaA on target cells are at least divided into the following steps (6)  Inactivation of CyaA by 14-3-3 ® fore, we examined CyaA preparations for each intoxication step. A flow cytometric analysis showed that all CyaA preparations similarly bound to L2 cells (Fig. 4B). The formation of toxin pores and translocation of the enzyme domain have been reported to occur in an unrelated manner after toxin binding (20). All CyaA preparations showed equivalent hemolytic activities on sheep erythrocytes (Fig. 4C). Translocation of the enzyme domain, as previously estimated by increases in cAMP levels in sheep erythrocytes (21), was confirmed for all CyaA preparations, although B. pertussis CyaA F375S and B. bronchiseptica CyaA were less efficient than B. pertussis CyaA and B. bronchiseptica CyaA S375F (Fig. 4D). B. pertussis CyaA has been reported to undergo proteolytic cleavage in the cytosol and liberate fragments of~45 kDa, including the enzyme bronchiseptica CyaA S375F , we detected 46-kDa fragments in the cytosol using the 3D1 antibody, which recognizes the enzyme domain ( Fig. 5A and B CyaA with Ser 375 interacts with 14-3-3 after phosphorylation. In the abovedescribed results of immunoblotting (Fig. 5B for B. bronchiseptica CyaA and B. bronchiseptica CyaA S375F ), we noted that the fragment containing ACD of B. bronchiseptica CyaA and B. pertussis CyaA F375S showed slightly slower mobility on SDS-polyacrylamide gels than B. pertussis CyaA and B. bronchiseptica CyaA S375F , implying that S 375 in the fragment underwent some covalent modification in the cytosol of target cells. Mass spectrometric analyses revealed that 46-kDa fragments of B. bronchiseptica CyaA and B. pertussis CyaA F375S , which were composed of at least~399 (~R 399 ) amino acid residues covering the enzyme domain, were phosphorylated at Ser 375 (Fig. 5C).
In order to examine whether the phosphorylation of Ser 375 influences the toxic action of CyaA on target cells, we prepared recombinant ACD peptides of B. pertussis CyaA (B. pertussis ACD) and B. pertussis CyaA F375S (B. pertussis ACD F375S ) phosphorylated by pretreatment with protein kinase A (PKA) in vitro and examined them for adenylate cyclase activity. B. pertussis ACD and B. pertussis ACD F375S , which consist of the region from amino acids at positions 2 to 400, were both phosphorylated with PKA, as judged by mobility shifts in SDS-polyacrylamide gels with Phos-tag and mass spectrometry (Fig. S4). The in vitro adenylate cyclase activities of these ACDs were unchanged before and after phosphorylation, indicating that the phosphorylation of Ser 375 per se does not influence the enzyme activity of CyaA (Fig. 6A).
We then searched a phosphorylation-related motif in ACD carrying Ser 375 with Scansite 3 (http://scansite3.mit.edu/) and found that the region surrounding phosphorylated Ser 375 corresponded to a 14-3-3 binding motif (mode 1, RSXpS/pTXP, in which pS and pT represent phosphoserine and phosphothreonine, respectively, and X is any type of residue [ Fig. 5C]) (23,24). The 14-3-3 protein family comprises seven isoforms that are expressed ubiquitously in a wide range of eukaryotic organisms and are known to regulate phosphoserine/phosphothreonine-mediated intracellular signal transduction by interacting with a number of phosphorylated proteins (23,25). Therefore, we examined the influence of 14-3-3 on the enzyme activity of B. pertussis ACD F375S phosphorylated with PKA and found that each isoform of the 14-3-3 protein, except for the isoform, significantly inhibited the in vitro adenylate cyclase activity of phosphorylated B. pertussis ACD F375S to various extents (Fig. 6B). In contrast to phosphorylated B. pertussis ACD F375S , the activities of B. pertussis ACD and nonphosphorylated B. pertussis ACD F375S were not influenced by 14-3-3 (Fig. 6C). These results indicate that the inhibition of adenylate cyclase activity was dependent on the phosphorylation of Ser 375 . The immunoprecipitation assay revealed that 14-3-3␥ interacted with phosphorylated, but not nonphosphorylated, B. pertussis ACD F375S . The apparent interaction between B. pertussis ACD and 14-3-3␥ was not detected regardless of the phosphorylation state (Fig. 6D). Similar results were obtained when the cytosolic fraction of J774A.1 cells, which contained 14-3-3, was used instead of Escherichia coli-produced recombinant 14-3-3␥ (Fig. S5). These results indicate that the activity of CyaA with Ser 375 was inhibited by 14-3-3 of the target cells after the phosphorylation of Ser 375 .
B. pertussis CyaA, but not B. bronchiseptica CyaA, disrupts epithelial barrier function. CyaA targets CD11b-expressing myeloid cells, such as macrophages, neutrophils, and dendritic cells, and subverts the immune responses of hosts by inhibiting bactericidal activities and affecting cytokine secretion (8,9,(11)(12)(13). The cAMP-elevating activity of CyaA was previously reported to up-and downregulate the induction of interleukin 6 (IL-6) and tumor necrosis factor alpha (TNF-␣), respectively (13,17,26). Therefore, we compared the effects of B. pertussis CyaA and B. bronchiseptica CyaA on the expression of IL-6 and TNF-␣ in L2 and NR8383 (rat alveolar macrophage) cells using real-time quantitative reverse transcription-PCR but observed no significant differences between B. pertussis CyaA and B. bronchiseptica CyaA, which is consistent with previous findings (13,17,26). This may have been because a slight increase in intracellular cAMP by residual active B. bronchiseptica CyaA was sufficient to affect cytokine production (note that B. bronchiseptica CyaA shows residual activity at higher concentrations, as shown in Fig. 2). We examined the effects of the toxin on epithelial cells, which were reportedly sensitive to CyaA (17,18 (Fig. 7A). Immunofluorescence microscopy revealed that the structure of tight junctions visualized by anti-ZO-1 antibody was destroyed by treatment with B. pertussis CyaA and B. bronchiseptica CyaA S375F , but not by treatment with B. bronchiseptica CyaA and B. pertussis CyaA F375S (Fig. 7B).  3). The statistical significance of differences was analyzed by an unpaired t test and indicated by asterisks as follows: *, P Ͻ 0.05; **, P Ͻ 0.001. (D) Immunoprecipitation assay for the interaction between ACDs and 14-3-3␥. The reaction solutions in panel C were immunoprecipitated (IP) using an anti-ACD antibody (Ab), followed by immunoblotting (IB) with an anti-14-3-3 antibody and anti-CyaA antibody (3D1).
Inactivation of CyaA by 14-3-3 ® In addition, B. pertussis CyaA, but not B. bronchiseptica CyaA, partly caused cell death under these experimental conditions (Fig. 7C). Consistent results were also obtained in rat infection experiments, in which rats were inoculated intranasally with a low dose (500 to 700 CFU) of B. bronchiseptica producing B. bronchiseptica CyaA (wild-type B. bronchiseptica [Bb WT]) or B. bronchiseptica CyaA S375F (Bb S375F) (Fig. 7D). On day 3 postinoculation, the tracheas, the epithelia of which were labeled with N-hydroxysuccinimide (NHS)-linked biotin that was introduced into the tracheal lumen, were excised and subjected to microscopy (Fig. 7D). The infection with wild-type (WT) B. bronchiseptica and B. bronchiseptica ΔcyaA did not disrupt the barrier function of the tracheal epithelium, as judged by the intact apical area labeled with biotin. In contrast, in rats infected with B. bronchiseptica S375F, the introduced biotin permeated the submucosal layers through the epithelium, indicating that B. bronchiseptica CyaA S375F impaired the barrier function of the tracheal epithelium (Fig. 7D). As shown in Fig. 7E, all bacteria tested colonized the rat trachea to similar extents.

DISCUSSION
The CyaAs of classical Bordetella are highly homologous: B. pertussis CyaA shows 97.8 and 97.7% amino acid identities to B. bronchiseptica CyaA and B. parapertussis CyaA, respectively, while B. bronchiseptica CyaA shows 99.8% amino acid identity to B. parapertussis CyaA. Although most amino acid replacements between the CyaAs of classical Bordetella were found in the C-terminal region, which binds to the cell receptor and elicits protective immunity (27,28), differences in the toxic activities of CyaAs were not reported (29,30). In the present study, we demonstrated that the strength of the intracellular enzyme activity of B. bronchiseptica CyaA was markedly less than that of B. pertussis CyaA. B. bronchiseptica CyaA was enzymatically active in vitro but was inactivated by phosphorylation and a subsequent association with 14-3-3 in target cells. These results are not in conflict with the previous study showing an intracellular increase in cAMP levels after the treatment of cells with B. bronchiseptica CyaA (31), because our results also demonstrated that the intracellular enzyme activity of B. bronchiseptica CyaA was not completely abrogated (e.g., in J774A.1, THP-1, and EGV-4T cells in Fig. 2 and in J774A.1 cells in Fig. S3 in the supplemental material]). In addition, we found that B. bronchiseptica CyaA increased intracellular cAMP in erythrocytes (Fig. 4D), and this may have been because of the weak expression of 14-3-3 proteins (32) or unfavorable conditions for phosphorylation in erythrocytes. The amount of CyaA that translocated into the cytosol may also influence the efficiency of inactivation by 14-3-3. Furthermore, differences in responses to B. bronchiseptica CyaA among various cell types may occur due to different cell conditions as described above.
Ser 375 of B. bronchiseptica CyaA was the crucial phosphorylation site responsible for the 14-3-3 association. NetPhos 3.1 (http://www.cbs.dtu.dk/services/NetPhos-3.1/) (33,34) predicted PKA and CDC2 as possible kinases to phosphorylate Ser 375 . However, the types of kinases actually involved in the intracellular phosphorylation of Ser 375 have not yet been identified, although PKA phosphorylated Ser 375 of ACD in vitro. B. pertussis CyaA carries Phe 375 instead of Ser 375 and therefore is not phosphorylated or inactivated by 14-3-3. B. parapertussis CyaA, which has Ser 375 , did not cause cell rounding in L2 cells, indicating that it is also inactivated in an identical manner to B. bronchiseptica CyaA. A previous study showed that B. pertussis CyaA was cytotoxic against macrophages, whereas B. parapertussis CyaA was not (35), which is consistent with the present results.
14-3-3 proteins bind to the phosphoserine/threonine-containing sequence motifs of various target proteins and modulate their functions in diverse manners as follows: alterations in the intracellular localization or ability of target proteins to interact with other partners, the direct augmentation or inhibition of target protein activity, the protection of target proteins from proteolysis or dephosphorylation, and serving as a scaffold to bridge two distinct target molecules (24,36,37). The region of CyaA corresponding to the 14-3-3 motif (Arg 372 -Pro 377 ) is not directly involved in enzyme activity or the association with calmodulin (38). However, the binding of 14-3-3 may interrupt the interaction between ACD and calmodulin because this region presumably faces the calmodulin binding space according to the crystal structures of the ACDcalmodulin complex (38). Alternatively, 14-3-3 may interfere with the conversion of ACD from the inactive state to the active state in response to calmodulin binding because the switch region of CyaA (Val 343 -Ala 364 ) responsible for catalytic activation is sterically in the vicinity of the Arg 372 -Pro 377 region (38). 14-3-3, which is known to regulate various biological processes, including mitogenic signal transduction, apoptotic cell death, and cell cycle control, has been reported to interact with hundreds of proteins, including Raf-1, tyrosine and tryptophan hydroxylases, Bcr, and Bad (25,36,37,(39)(40)(41). Regarding the virulence factors of pathogens, a previous study reported that Pseudomonas aeruginosa ExoS required 14-3-3 to exert its ADP-ribosylating effects (42 and Bacillus anthracis edema factor (EF) were also recently revealed to disrupt host cell barrier function by inhibiting Rab11-mediated endocytic recycling, which is essential for transporting junctional proteins at tight/adherens junctions (43,44). These effects were mediated by increases in intracellular cAMP induced by CT and EF. The mechanism by which CyaA with Phe 375 disrupted the epithelial cell barrier may be similar to that by CT and EF. We also found that some cells in the confluent cell layer were killed by the treatment with CyaA with Phe 375 (Fig. 7C), indicating that a long stimulation of active CyaA may induce epithelial cell death, which may exacerbate the disruption of the epithelial barrier. Similar epithelial damage was observed in rat tracheas infected with B. bronchiseptica expressing B. bronchiseptica CyaA S375F , but not B. bronchiseptica CyaA (Fig. 7D). These results imply that B. pertussis CyaA causes more extensive damage to the respiratory epithelium than B. bronchiseptica/B. parapertussis CyaA. This may explain the less severe clinical manifestations in B. bronchiseptica and B. parapertussis infections than in B. pertussis infections. However, it is important to note that CyaA plays a diverse role in Bordetella infections through several properties, such as adenylate cyclase activity, pore-forming activity, and potent immunogenicity (45)(46)(47)(48)(49)(50), and only adenylate cyclase activity was disturbed by 14-3-3 in CyaAs of B. bronchiseptica and B. parapertussis. Furthermore, B. bronchiseptica CyaA modulated the production of IL-6 and TNF-␣, similar to B. pertussis CyaA (data not shown), indicating that slight increases in cAMP are sufficient to induce biological effects in some cases. Thus, our results do not deny the roles of B. bronchiseptica CyaA as a virulence factor in B. bronchiseptica infection. Although classical Bordetella share major virulence factors, they vary in disease severity and host specificity. The exact nature of this difference in pathogenesis remains unknown; however, previous studies pointed out distinct virulence factors produced by each Bordetella species, including pertussis toxin specific for B. pertussis, the type III secretion system apparently expressed in B. bronchiseptica, and the different lipopolysaccharide structures of each Bordetella species (45,(51)(52)(53). In addition to specific virulence factors, homologous ones may also play different roles in each Bordetella infection. We previously demonstrated that the expression level of DNT varied among classical Bordetella because of polymorphisms in the promoter region for the dnt gene (54). Most pig isolates of B. bronchiseptica carried the promoter types with increasing transcription activities, whereas human isolates of B. pertussis and B. parapertussis carried the least active promoter. These results appear to reflect the important role of DNT in the pathogenesis of pig infections with B. bronchiseptica. In the present study, we showed that CyaA may function differently as a virulence factor between B. bronchiseptica/B. parapertussis and B. pertussis infections. Studies on functional differences in homologous virulence factors, in addition to specific factors, may provide insights into the underlying reasons for the different pathogenicities and host specificities of classical Bordetella.
Note. After the manuscript had been completed, Hasan et al. reported the disruption of paracellular barrier function in human bronchial epithelial VA10 cells by B. pertussis CyaA (55). Further studies are warranted in order to evaluate the importance of epithelial barrier disruption in the pathogenesis of B. pertussis.

MATERIALS AND METHODS
Cultured cell lines, bacterial strains, gene constructions, and other commonly utilized methods are given in Text S1 in the supplemental material.
Isolation of cytosolic and membrane fractions from CyaA-treated cells. J774A.1 cells grown to confluence were treated with 0.5 g/ml of each recombinant CyaA at 37°C for 1 h. After the cells were washed with cold Dulbecco-modified phosphate-buffered saline (D-PBS), the cells were scraped and