Ectopic Expression of O Antigen in Bordetella pertussis by a Novel Genomic Integration System

Some bacterial phenotypes emerge through the cooperative functions of a number of genes residing within a large genetic locus. To transfer the phenotype of one bacterium to another, a means to introduce the large genetic locus into the recipient bacterium is needed. Therefore, we developed a novel system by combining the advantages of a bacterial artificial chromosome vector and phage-derived gene integration machinery. In this study, we succeeded for the first time in introducing a gene locus involved in O antigen biosynthesis of Bordetella bronchiseptica into the chromosome of B. pertussis, which intrinsically lacks O antigen, and using this system we analyzed phenotypic alterations in the resultant mutant strain of B. pertussis. The present results demonstrate that this system successfully accomplished the above-described purpose. We consider this system to be applicable to a number of bacteria other than Bordetella.

genic mechanisms for the different host specificities among Bordetella species, which share many virulence factors with each other, have yet to be elucidated.
Recent comparative genome analyses on closely related pathogens revealed a genetic background that involves the adaptation of pathogens to the host environment. In Yersinia pestis, the acquisition of the plasmid pPCP1 containing the gene pla, which encodes a protease, is an important step for adaptation to the pulmonary environment in host animals (4). A loss-of-function mutation in a gene for urease also provides a new niche for Y. pestis (5). The host tropism of Staphylococcus aureus and Salmonella enterica serovar Typhimurium are reportedly influenced by a single mutation in a gene that encodes a membrane protein and an allelic variation in an adhesin, respectively (6,7).
In the case of Bordetella, previous studies indicated that B. pertussis and B. parapertussis independently evolved from B. bronchiseptica-like ancestors through large-scale gene losses and translocations and, as a result, the genome sizes of B. pertussis and B. parapertussis became approximately 1.3 Mbp and 0.6 Mbp smaller, respectively, than that of B. bronchiseptica (8)(9)(10). Limited evidence is currently available for the large acquisition of genes through horizontal gene transfer during evolution (8)(9)(10). Differences in host tropism among Bordetella species have been suggested to result from gene decay. If this is the case, the narrow host specificity of B. pertussis may be altered by supplying a number of genes of B. bronchiseptica, which has a wide range of host animals. We decided to examine this possibility. To achieve this, we needed an efficient means to enable the introduction of the large genomic DNA (gDNA) fragments of B. bronchiseptica into the chromosome of B. pertussis.
In the present study, we propose a novel method to achieve this purpose; we named it the BPI system. This method is based on the combination of a bacterial artificial chromosome (BAC) vector and phage-derived integration machinery. To verify the usability of the BPI system, we attempted to generate O antigen-expressing B. pertussis, the parental strain of which does not intrinsically express O antigen, by introducing the B. bronchiseptica wbm gene locus, which is approximately 32 kbp (11)(12)(13). The resultant B. pertussis strain expressed the O antigen of B. bronchiseptica and became more resistant to complement-or polymyxin B-mediated bactericidal effects than the parental B. pertussis. In addition, an in vivo competitive infection assay showed that the expression of O antigen contributed to the efficient colonization of B. pertussis in the mouse respiratory tract. These results demonstrated that the BPI system has potential as a powerful tool to transform phenotypes through the introduction of large DNA fragments into the chromosomes of recipient bacteria.

RESULTS
Construction of the BPI system in Bordetella species. We constructed the vector for the BPI system by combining the advantages of the BAC vector to carry large DNA fragments (14) and the phiC31 phage to integrate these fragments into a bacterial chromosome in a site-specific manner (15) (Fig. 1A and B). We initially introduced a kanamycin (Km) resistance gene into pBeloBAC11 in place of the chloramphenicol resistance gene, because resistance to chloramphenicol is often unreproducible in Bordetella grown on Bordet-Gengou (BG) agar plates. The resultant plasmid was designated pBeloBAC11-Km. An origin of transfer (oriT), which is necessary for the conjugative transfer of the vector from Escherichia coli to B. pertussis, was subsequently introduced. To utilize the integration machinery of the phiC31 phage, we also placed the attachment sites of the phage (attP) and bacterium (attB) into the chromosome of B. pertussis and pBeloBAC11-Km, respectively. The strain and plasmid obtained were designated Bp attP and pBPI. The ColE1 origin of replication (ori) was also introduced into pBPI from high-copy-number pBlueScript to obtain the plasmid from E. coli at a high yield. The resulting plasmid was designated pBPIori. The ColE1 ori was removed from pBPIori via a treatment with BamHI before use. The plasmid after the removal of the ColE1 ori was expected to be identical to pBPI.
We initially attempted to introduce pBPI itself into Bp attP carrying pBBR1MCS5-Int, which expresses the phiC31 integrase, to test whether this system functions properly in B. pertussis. After the introduction of pBPI, we successfully isolated the Km-resistant colonies of Bp attP (Bp attP -pBPI) only when the integrase-expressing plasmid was introduced (Fig. 1D). PCR with primers specific to the junction sequences attL and attR, which are generated after integration, revealed integration that was mediated by attB and attP (Fig. 1C). Therefore, the BPI system appeared to function as expected.
Introduction of large B. bronchiseptica gDNAs into the B. pertussis chromosome. The BAC system is capable of maintaining high-molecular-weight DNA in E. coli (14). To examine the cloning capacity of pBPI, we ligated large DNA fragments of B. bronchiseptica with pBPI and introduced plasmids into E. coli. We recovered the plasmids from 21 Km-resistant clones that were randomly picked, we roughly evaluated the sizes of the inserted fragments from the results of electrophoresis after BamHI digestion, and we found that the inserted fragments ranged between 8 and 50 kbp. Construction of pBPI. pBeloBAC11 was modified for the BPI system. oriT for conjugative transfer and attB for site-specific integration were introduced into pBeloBAC11-Km. High-copy-number-yielding ColE1 ori was cloned into the BamHI site of pBPI in order to obtain a large amount of the plasmid (pBPIori). The ColE1 ori was removed by subsequent BamHI treatment, and the resulting plasmid was identical to pBPI. (B) Schematic diagram of the BPI system. pBPI was integrated into the chromosome of Bp attP via attB on pBPI and attP on the chromosome by an ectopically expressed integrase. attL and attR regions are junction regions generated after integration. (C) The integration of pBPI into the Bp attP chromosome was verified with PCR. PCR was performed using the genomes of Bp attP and Bp attP -pBPI as the templates. As judged by electrophoresis, the newly generated attL and attR sites after integration were only amplified when the latter template was used. (D) Colonies of B. pertussis grown on agar plates containing kanamycin. Only Bp attP harboring the integrase-expressing plasmid (pBBR1MCS5-Int) grew after the introduction of pBPI.
Sequencing of the largest fragment revealed that it covered an approximately 49.6-kbp region, including the genes BB3602 to BB3665. The plasmid carrying this fragment was named pBPI-L1.
Following the conjugative transfer of pBPI-L1 into Bp attP , we examined the integration of the vector via PCR by using pBPI-L1-integrated B. pertussis gDNA as a template. Although the attL and attR regions were amplified, the B. bronchiseptica gDNA region was not (data not shown), indicating undesirable homologous recombination, possibly because of large homologous sequences between the 49.6-kbp fragment and the chromosome of B. pertussis. To overcome this issue, we generated a recA deletion mutant of Bp attP (Bp attP ⌬recA::gfp), because RecA is mainly involved in recombination in bacteria (16,17). Bp attP ⌬recA::gfp did not show any growth defect during in vitro culture (see Fig. S1 in the supplemental material). We then transferred pBPI-L1 into Bp attP ⌬recA::gfp in the presence of pBBR1MCS5-Int and checked the integration of pBPI-L1 by PCR performed with pBPI-L1 itself or with the gDNA of Bp attP ⌬recA::gfp after transformation as the template. In this PCR, we used pairs of primers to amplify short fragments, named I to VIII. Fragments II to VII, covering the region of B. bronchiseptica gDNA, were amplified by PCR with each template ( Fig. 2A and B). In contrast, fragments I and VIII, containing attL and attR, were only amplified when the gDNA of Bp attP ⌬recA:: gfp after transformation (Bp attP ⌬recA::gfp-L1) was used as the template. These results indicated that full-length pBPI-L1 containing the 49.6-kbp fragment of B. bronchiseptica gDNA was successfully integrated into the chromosome of B. pertussis. Thus, the BPI system is expected to be capable of integrating gDNA fragments as large as 50 kbp into the chromosome of B. pertussis.
Introduction of the wbm locus of B. bronchiseptica into B. pertussis via the BPI system. To assess the usability of the BPI system, we attempted to introduce a functional genetic locus into the chromosome of B. pertussis. We selected the wbm locus, which comprises 24 genes and has a length of approximately 32 kbp in B. bronchiseptica RB50 (12,18) and is one of the gene clusters lost during evolution of B. pertussis (10,13). The wbm locus is also known to be involved in the biosynthesis of O antigen, which is the outermost component of lipopolysaccharide (LPS). B. pertussis does not have O antigen because of the lack of the wbm locus. Therefore, we attempted to introduce the wbm locus into the chromosome of B. pertussis by using the BPI system (Fig. 3A). If the genes inserted by the BPI system are functional, O antigen should be expressed in B. pertussis recipients. We constructed pBPI carrying the wbm locus (pBPI-wbm) and introduced it into Bp attP ⌬recA::gfp. The incorporation of pBPI-wbm into the chromosome was verified (Fig. 3A) in a series of PCRs in a manner similar to those shown in Fig. 2, and LPS of the tested strains was examined for the expression of O antigen by using Tricine-SDS-PAGE followed by silver staining. B. bronchiseptica and B. pertussis share similar lipid A and core structures, called band A (19-21) (Fig. 3B, lower panel). While all of the tested strains showed the band A unit at the position of approximately 9 kDa, as reported previously (20,21), O antigens of approximately 14 kDa were detected in B. bronchiseptica wild-type and wbm-integrated B. pertussis (Bp attP ⌬recA::gfp-wbm) strains, but not in the B. pertussis wild-type or pBPI-integrated B. pertussis (Bp attP ⌬recA::gfp-pBPI) strain (Fig. 3B, upper panel). In addition, Western blot analysis revealed that the 14-kDa band in LPS extracted from Bp attP ⌬recA::gfp-wbm was recognized by anti-B. bronchiseptica serum but not by anti-B. pertussis serum (Fig. 3C), suggesting that Bp attP ⌬recA::gfp-wbm expressed O antigen corresponding to that of B. bronchiseptica.
Characterization of O antigen-expressing B. pertussis. Previous studies suggested that O antigen plays a role in the protection of Gram-negative bacteria from host immune responses, such as complement-mediated killing (22)(23)(24). Therefore, we examined whether the expressed O antigen conferred some protective ability on B. pertussis against serum killing. Approximately 40% of Bp attP ⌬recA::gfp-pBPI was killed by treatment with 90% rat serum, whereas Bp attP ⌬recA::gfp-wbm not only survived but also multiplied, even in the presence of serum (Fig. 4A). These two B. pertussis derivatives grew similarly in the presence of 90% heat-inactivated serum (Fig. 4A). Polymyxin B, a cationic antimicrobial peptide, interacts with the lipid A moiety of LPS and exhibits potent antibacterial activity (25,26). We also found that approximately 40% of Bp attP ⌬recA::gfp-wbm cells survived in the presence of 0.08 or 0.16 g/ml of polymyxin B, whereas approximately 10% of Bp attP ⌬recA::gfp-pBPI cells survived under the same conditions (Fig. 4B).

Contribution of O antigen to respiratory infections in mice.
Since O antigenexpressing B. pertussis was more resistant to the bactericidal activity of serum than vector-integrated B. pertussis, the former may colonize host animals more efficiently than the latter. To examine this possibility, we performed a competitive infection assay in mice. After the inoculation of equal amounts of Bp attP ⌬recA::gfp-pBPI-Gm and Bp attP ⌬recA::gfp-wbm, competitive index (CI) values were calculated from CFU isolated from the nasal septum, trachea, and lungs 8 and 14 days postinfection. Bp attP ⌬recA:: gfp-pBPI-Gm and Bp attP ⌬recA::gfp-wbm equally colonized the nasal cavity, as judged by CI values of~1 on days 8 and 14 (Fig. 4C). In contrast, Bp attP ⌬recA::gfp-wbm outcompeted Bp attP ⌬recA::gfp-pBPI-Gm in the trachea and lungs, showing mean CI values of 3.33 (day 8) and 3.88 (day 14) and of 2.41 (day 8) and 4.99 (day 14), respectively (Fig. 4C). Bp attP ⌬recA::gfp-pBPI and Bp attP ⌬recA::gfp-pBPI-Gm, which carry kanamycin and gentamicin resistance genes, respectively, in their genomes, were subjected to a similar competitive assay to assess the influence of drug resistance genes on the infectivities of the bacteria, and CI values of~1 on days 8 and 14 were obtained, indicating that these genes do not affect the ability of these bacteria to colonize mouse respiratory organs (Fig. S2). Collectively, these results indicated that the ability of O antigenexpressing B. pertussis to colonize the lower respiratory tract was superior to that of the parental strain.

DISCUSSION
In the present study, we developed a novel genome integration system, named the BPI system. To accomplish this, we incorporated the features of various genes related to DNA maintenance and transfer, as follows.
We utilized phiC31 integrase, which mediates unidirectional and site-specific recombination between the attachment sites attP and attB (15). Since no host factor is required for this reaction, this recombination system is applicable to genetic engineering in bacteria and human cells (27)(28)(29). In the present study, we constructed the novel cloning vector pBPI by introducing attB into pBeloBAC11-Km and attP into the chromosome of B. pertussis (Fig. 1A and B), based on a previous study showing that the integration efficiency of attB-containing DNA into an attP-containing chromosome was higher than that for attP-containing DNA integration into an attB-containing chromosome (30).
BAC vectors, including pBeloBAC11, are derived from the E. coli F factor and are maintained at a single or few copies in an E. coli cell. This feature makes BACs useful for the cloning and stable maintenance of large DNA fragments in E. coli, and they are widely used for the construction of various genomic libraries (31)(32)(33)(34). On the other hand, low-copy-number plasmids are difficult to obtain in large numbers and are unsuitable for subsequent cloning procedures (35). To overcome this disadvantage, we introduced the ColE1 ori of high-copy-number pBlueScript into the BamHI site of pBPI (Fig. 1A). The final yield of the resultant plasmid, named pBPIori, was approximately 50-fold higher than that of pBPI. Since the whole region of ColE1 ori is easily removed by BamHI digestion and is not carried over into subsequent cloning procedures, we were able to use pBPIori without adverse effects on vector preparation.
In the first attempt to apply the BPI system, we succeeded in introducing pBPI into the chromosome of B. pertussis, as shown in Fig. 1C and D. However, when pBPI carrying the gDNA fragment of B. bronchiseptica was used, genomic integration was unsuccessful (data not shown), suggesting unexpected recombination between homo-  (8)(9)(10). To prevent this unwanted recombination, we generated a recA deletion mutant of B. pertussis. RecA is known to mediate DNA recombination between two homologous regions (16,17,36). By using the recA deletion mutant, we successfully introduced an approximately 50-kbp fragment of B. bronchiseptica DNA into the chromosome of B. pertussis ( Fig. 2A and B).
To examine whether the BPI system is of practical use, we attempted to introduce the wbm locus of B. bronchiseptica into B. pertussis. This locus is involved in the biosynthesis of the O antigen of LPS and is 32 kbp, a size that is applicable to the BPI system (12). LPS is one of the major components of the outer membrane of Gramnegative bacteria and generally consists of three representative components: lipid A, a core oligosaccharide, and O antigen (37). Although B. bronchiseptica and B. pertussis share the basic structure of lipid A, a core oligosaccharide, and trisaccharide, the latter does not express O antigen due to the loss of the entire wbm locus (12,18). After the introduction of the wbm locus into the chromosome via the BPI system, B. pertussis expressed O antigen corresponding to that of B. bronchiseptica ( Fig. 3B and C). These results indicated that the BPI system introduces a large DNA fragment into the chromosome of the recipient bacterium without genetic damage.
The ectopic expression of O antigen conferred resistance to serum and polymyxin B on B. pertussis ( Fig. 4A and B). Complement activation is known to occur on the surface of pathogens, and polymyxin B, which is a cationic antimicrobial peptide, exhibits bactericidal activity through interactions with the negatively charged lipid A moiety of LPS (26). Therefore, our results suggest that the expressed O antigen reduces the accessibility of complement and polymyxin B; however, further studies are needed to clarify the mechanisms underlying the protective role of O antigen. O antigenexpressing B. pertussis colonized the mouse trachea and lungs better than the isogenic parental strain, suggesting that the expression of O antigen is specifically advantageous for B. pertussis during lower respiratory tract infections in mice. On the other hand, the amount of the O antigen-expressing strain recovered from the nasal septum of mice was similar to that of the isogenic parental strain, indicating that the presence of O antigen has a negligible impact on the ability of bordetellae to colonize the nasal cavity. These results are consistent with previous studies in which wbm-deleted mutants of B. bronchiseptica and B. parapertussis were characterized (22). In the present study, we succeeded, for the first time, in characterizing O antigen-expressing B. pertussis by using the BPI system.
Collectively, the present results indicate that the BPI system, which alters the phenotype of B. pertussis by introducing the large wbm locus to generate O antigen, provides us with a means to generate a genomic library carrying the large gDNA fragments of B. bronchiseptica into B. pertussis. Using the BPI system, we are currently attempting to screen the genetic loci that influence the host tropism of Bordetella species.

MATERIALS AND METHODS
Bacterial strains and plasmids. All bacterial strains and plasmids used in this study are shown in Table S1 and S2. B. bronchiseptica RB50 was provided by Peggy A. Cotter (University of North Carolina), and B. pertussis Tohama I was maintained in our laboratory. B. pertussis and B. bronchiseptica strains were grown at 37°C in Stainer-Scholte (SS) broth or BG agar (Becton, Dickinson) containing 1% glycerol and 15% defibrinated horse blood. E. coli strains were grown at 37°C in Luria-Bertani (LB) broth or LB agar. Gentamicin, kanamycin, and ceftibuten were added at a final concentration of 10 g/ml when necessary. The bacterial cell numbers of B. pertussis were estimated based on optical density values at 650 nm (OD 650 ) of fresh cultures, according to the following equation: OD 650 of 1.0 ϭ 3.3 ϫ 10 9 cells/ml. The experimental procedures for construction of mutant strains and plasmids are described in Text S1. BPI reaction. All plasmids were transferred to B. pertussis strains by triparental conjugation with the E. coli helper strain HB101 carrying pRK2013, as described elsewhere (38). For example, E. coli DH5␣ carrying pBPI and HB101 carrying pRK2013 were inoculated into LB with kanamycin. Bp attP ⌬recA::gfp carrying pBBR1MCS5-Int was inoculated into SS broth containing gentamicin. After overnight incubation, the OD values of these bacterial cultures were adjusted to 1.0. Eight hundred microliters of B. pertussis and 100 l of each E. coli strain were then mixed. The bacterial mixture was centrifuged, and the pellet was suspended in 10 mM MgSO 4 and spread on LB plates. After a 6-h incubation at 37°C, the colonies were collected and suspended in SS broth. The appropriately diluted suspension was plated on BG agar plates containing ceftibuten and kanamycin for the selection of recombinants that had integrated pBPI into the chromosome. The integration of DNA fragments into the chromosome was confirmed by PCR with PrimeSTAR GXL DNA polymerase (TaKaRa) and the primers listed in Table S2. Briefly, a fresh colony was suspended in 10 l of deionized and autoclaved water with a toothpick and heated at 98°C for 3 min. A 1-l aliquot from the boiled supernatant was used for the PCR template.
LPS purification and Tricine-SDS-PAGE. LPS was extracted from B. bronchiseptica and B. pertussis by using a modified hot aqueous phenol extraction method as previously described (39). B. bronchiseptica and B. pertussis were grown at 37°C in SS broth for 6 and 12 h, respectively. Twenty milliliters of the bacterial culture was centrifuged, and the resulting pellets were suspended in 1 ml of 10 mM Tris-HCl, pH 8.0, containing 2% SDS. The suspension was mixed with 2 l of recombinant DNase I (5 U/l; TaKaRa) and 2 l of PureLink RNase A (40 mg/ml; Invitrogen) and incubated at 37°C for 1 h. After the addition of 1 l of proteinase K (20 mg/ml; Sigma), the mixture was then incubated at 37°C overnight. Samples were mixed with an equal volume of 90% phenol preheated to 68°C, and mixtures were incubated at the same temperature for 20 min. Following centrifugation, the aqueous layer was dialyzed against deionized and autoclaved water. The dialyzed solutions were centrifuged at 100,000 ϫ g for 3 h, and the precipitate was dissolved in 30 l SDS sample buffer (62.5 mM Tris-HCl [pH 6.8] containing 6% glycerol, 2% SDS, 12.5 mM dithiothreitol, and 0.01% bromophenol blue). Collected LPS was subjected to Tricine-SDS-PAGE by using a 16% separating gel and 4% stacking gel (40). LPS in the gel was visualized by silver staining with the Sil-Best Stain One kit (Nacalai Tesque) according to the manufacturer's instructions. Samples on the other gels were transferred to a polyvinylidene difluoride membrane (Millipore) and visualized by immunoblotting with rabbit antiserum against B. bronchiseptica or B. pertussis at dilution ratios of 1:3,000 and 1:5,000, respectively. For immunoblotting, a goat anti-rabbit IgG-horseradish peroxidase (HRP)-conjugated antibody (1:10,000; Jackson) was used as a secondary antibody. Signals were detected with an ECL Western blotting detection system (GE Healthcare). The rabbit antiserum against B. bronchiseptica or B. pertussis was prepared according to the standard protocol, by injecting formalin-fixed B. bronchiseptica or B. pertussis into female Kbl:JW rabbits (Kitayama Labs, Japan).
Assays for sensitivity of bacteria to rat serum and polymyxin B. The sensitivities of bacteria to serum and polymyxin B were evaluated by previously reported methods, with slight modifications (22,41). Normal rat serum was purchased from Japan SLC. Polymyxin B sulfate (Wako) was dissolved in deionized and autoclaved water to obtain a stock solution (1 mg/ml). The bacteria for the assays were grown for 12 h in SS broth to the mid-log phase. Approximately 500 CFU of bacteria were incubated in the presence or absence of 90% serum in Hanks' balanced salt solution at 37°C for 6 h. For the polymyxin B sensitivity assay, approximately 2 ϫ 10 5 CFU of bacteria were incubated in phosphate-buffered saline (PBS) with the indicated concentrations of polymyxin B at 37°C for 2 h. After each treatment, bacteria were cultivated on BG agar and CFU were enumerated. Survival rates after the serum treatment were expressed in relation to the input bacterial number (as 100%). Survival rates after the polymyxin B treatment were calculated in relation to the bacterial number when incubated without polymyxin B (as 100%).
Mouse intranasal infection. All animal experiments were approved by the Animal Care and Use Committee of the Research Institute for Microbial Diseases, Osaka University, and performed in accordance with the Regulations on Animal Experiments at Osaka University. Bp attP ⌬recA::gfp-pBPI-Gm and Bp attP ⌬recA::gfp-wbm strains were grown for 12 h in SS broth to the mid-log phase and combined in approximately equal bacterial numbers. Seven-week-old BALB/c mice were anesthetized with an intraperitoneal injection of midazolam (2.0 mg/kg), medetomidine (0.3 mg/kg), and butorphanol (5.0 mg/kg). Groups of four mice were intranasally inoculated with 5 ϫ 10 6 CFU of the bacterial mixture, prepared as described above, in 50 l of PBS. Mice were sacrificed 8 or 14 days postinfection, and the nasal septum, trachea, and lungs were excised, homogenized in PBS, and plated on BG agar containing ceftibuten and kanamycin or gentamicin to assess CFU. The CI for each organ was calculated as follows: CI ϭ [(output CFU of Bp attP ⌬recA::gfp-wbm)/(output CFU of Bp attP ⌬recA::gfp-pBPI-Gm)]/[(input CFU of Bp attP ⌬recA::gfpwbm)/(input CFU of Bp attP ⌬recA::gfp-pBPI-Gm)]. A CI of Ͼ1 indicated that Bp attP ⌬recA::gfp-wbm outcompeted Bp attP ⌬recA::gfp-pBPI-Gm.