Serendipitous Discovery of an Immunoglobulin-Binding Autotransporter in Bordetella Species

Abstract

We describe the serendipitous discovery of BatB, a classical-type Bordetella autotransporter (AT) protein with an ∼180-kDa passenger domain that remains noncovalently associated with the outer membrane. Like genes encoding all characterized protein virulence factors in Bordetella species, batB transcription is positively regulated by the master virulence regulatory system BvgAS. BatB is predicted to share similarity with immunoglobulin A (IgA) proteases, and we showed that BatB binds Ig in vitro. In vivo, a Bordetella bronchiseptica ΔbatB mutant was unable to overcome innate immune defenses and was cleared from the lower respiratory tracts of mice more rapidly than wild-type B. bronchiseptica. This defect was abrogated in SCID mice, suggesting that BatB functions to resist clearance during the first week postinoculation in a manner dependent on B- and T-cell-mediated activities. Taken together with the previous demonstration that polymorphonuclear neutrophils (PMN) are critical for the control of B. bronchiseptica in mice, our data support the hypothesis that BatB prevents nonspecific antibodies from facilitating PMN-mediated clearance during the first few days postinoculation. Neither of the strictly human-adapted Bordetella subspecies produces a fully functional BatB protein; nucleotide differences within the putative promoter region prevent batB transcription in Bordetella pertussis, and although expressed, the batB gene of human-derived Bordetella parapertussis (B. parapertussis_hu) contains a large in-frame deletion relative to batB of B. bronchiseptica. Taken together, our data suggest that BatB played an important role in the evolution of virulence and host specificity among the mammalian-adapted bordetellae.

Bordetella pertussis, human-derived Bordetella parapertussis (B. parapertussis_hu), and Bordetella bronchiseptica are gram-negative bacteria that cause respiratory infections in mammals. Multilocus enzyme electrophoresis, insertion sequence polymorphism, and genome sequence analyses have revealed that these bacteria are so closely related that they should be considered the same species (32, 34, 39). Despite their remarkable similarity, however, they differ in host specificity and disease manifestations. While B. bronchiseptica infects nearly all mammals, typically causing long-term asymptomatic or subclinical infections (4), B. pertussis and B. parapertussis_hu infect only humans and cause the acute disease known as whooping cough (2, 18, 20). Although many Bordetella factors with proven or postulated roles in pathogenesis have been identified and characterized, including some that differ substantially in amino acid sequence or expression profile among these subspecies, the molecular basis for differences in host specificity and disease is unknown.

The autotransporter (AT) family comprises a large and diverse group of proteins that are exported across the outer membranes of gram-negative bacteria, where they either remain associated with the cell surface or are secreted into the extracellular milieu (see reference 21 for a review). All classical ATs consist of three domains: an N-terminal signal sequence, which directs the protein to the Sec secretion complex of the cytoplasmic membrane, a passenger domain, containing the effecter function, and a C-terminal β-domain, which facilitates the export of the passenger domain across the outer membrane. β-Domains are highly conserved and are the identifying feature of ATs. Passenger domains, however, are extremely diverse in both sequence and function. The majority of characterized passenger domains have roles in bacterial virulence through such actions as adherence to host cells, the proteolytic cleavage of host proteins, and the processing of other bacterial cell surface proteins (see reference 33 for a review). An analysis of the sequenced genomes of B. bronchiseptica, B. pertussis, and B. parapertussis_hu revealed the presence of 22 genes with the potential to encode classical ATs, of which the functions of only a few have been examined (34). These include pertactin (Prn), a putative adhesin that is included in B. pertussis acellular vaccines, BrkA, a protein that is involved in serum resistance and possibly adherence, and SphBI, a serine protease that is involved in the maturation of filamentous hemagglutinin (FHA), an adhesin of the two-partner secretion family (9, 14, 26, 29, 34, 36). Like all known protein virulence factors produced by Bordetella that have been identified so far, the expression of the genes encoding the characterized ATs is positively regulated by the BvgAS phosphorelay.

We describe here the serendipitous discovery of BatB, a putative autotransporter produced by B. bronchiseptica. The characterization of a polyclonal antibody generated against an internal fragment of FHA revealed the presence of an unknown ∼180-kDa protein in whole-cell lysates (WCLs) of Bvg⁺-phase B. bronchiseptica. Since the protein was associated with the outer membrane and positively regulated by BvgAS, we sought to determine its identity. We determined that the protein was a previously uncharacterized putative Bordetella AT, BatB. Our results show that BatB binds immunoglobulin (Ig) and is necessary for B. bronchiseptica to resist inflammatory clearance from the respiratory tract. We also characterized differences in the sequence and expression of batB between B. bronchiseptica, B. pertussis, and B. parapertussis_hu. Our results suggest that BatB contributes to differences in disease and/or host specificity between the human- and non-human-adapted bordetellae.

MATERIALS AND METHODS

Bacterial strains.

The bacterial strains used in this study are listed in Table 1. Bordetella spp. strains were cultured on Bordet-Gengou (BG) agar (BD Biosciences, San Jose, CA) supplemented with 7.5% defibrinated sheep blood (Mission Laboratories, Diamondbar, CA) for 48 to 72 h at 37°C. For β-galactosidase assays and total RNA isolations, bacteria were grown in Stainer-Scholte (SS) broth (38) at 37°C with shaking. Escherichia coli strains were cultured on Luria-Bertani (LB) agar or in LB broth. When appropriate, culture media were supplemented with gentamicin (20 μg ml⁻¹), streptomycin (25 μg ml⁻¹), or ampicillin (100 μg ml⁻¹).

TABLE 1.

Strains and plasmids used in this analysis

Strain or plasmid	Description	Reference or source
Strains
E. coli
DH5α	Used for molecular cloning	Gibco-BRL, Gaithersburg, MD
SM10λpir	Used for conjugation with Bordetella	31
B. bronchiseptica
RB50	WT	5
RB53	Bvg+-phase-locked derivative of RB50	5
RB53i	Bvgi-phase-locked derivative of RB50	6
RB54	Bvg−-phase-locked derivative of RB50	5
RBX9	ΔfhaB derivative of RB50	7
B. parapertussishu
12822	Clinical B. parapertussishu isolate	20
Plasmids
RB50::pRH4	RB50 with plasmid pRH4 integrated, resulting in the disruption of ORF452	This study
RB50::pCW85	RB50 with B. bronchiseptica batB promoter region-lacZYA fusion	This study
RB50::pCW86	RB50 with B. pertussis batB promoter region-lacZYA fusion	This study
RB50::pCW87	RB50 with B. parapertussishubatB promoter region-lacZYA fusion	This study
RBAT	ΔbatB derivative of RB50	This study
RBAT::pCW98	Complemented ΔbatB strain	This study
RBatpp	RB50 with 1,593-bp deletion within batB	This study
BPSM	Streptomycin-resistant derivative of clinical B. pertussis isolate Tohama I	30
BPSM::pCW85	BPSM with B. bronchiseptica batB promoter region-lacZYA fusion	This study
BPSM::pCW86	BPSM with B. pertussis batB promoter region-lacZYA fusion	This study
BPSM::pCW87	BPSM with B. parapertussishubatB promoter region-lacZYA fusion	This study
12822::pCW85	12822 with B. bronchiseptica batB promoter region-lacZYA fusion	This study
12822::pCW86	12822 with B. pertussis batB promoter region-lacZYA fusion	This study
12822::pCW87	12822 with B. parapertussishubatB promoter region-lacZYA fusion	This study

Molecular cloning and DNA sequence analysis.

Standard cloning techniques were used for all DNA manipulations (37). Restriction enzymes, T4 DNA ligase, Taq polymerase, and high-fidelity Phusion polymerase were purchased from Promega Corp. (Madison, WI), New England Biolabs (Beverly, MA), or MJ Research (Waltham, MA), respectively, and were used according to the manufacturers' instructions.

Construction of bacterial strains and plasmids.

B. bronchiseptica ORF0452 was disrupted by plasmid pRH4. This plasmid contains a 632-bp fragment corresponding to nucleotides 111 to 742 of BB0452 cloned into pEGZ (27). Allelic exchange plasmid pEG7S is a derivative of Bordetella suicide plasmid pEG7 (27), in which the sacB gene from pRE118 (13) has been cloned into the EcoRI site. We deleted the batB allele from B. bronchiseptica strain RB50 by allelic exchange using plasmid pRH5. Plasmid pRH5 was created by cloning a 737-bp fragment containing upstream sequences and the first three codons of the batB coding sequence and a 722-bp fragment that includes the last three codons of the batB coding sequence and downstream sequences into pEG7S. A sequence analysis of the cloned DNA (Laragen, Los Angeles, CA) was performed to confirm the fidelity of B. bronchiseptica sequences. pRH5 was transformed into E. coli SM10λpir, which was used strictly for its mobilization functions, and then delivered to RB50 by conjugation. Cointegrants in which pRH5 had recombined into the chromosome were selected on medium containing streptomycin (B. bronchiseptica strains are naturally streptomycin resistant) and gentamicin. Recombinants that had excised the plasmid from the chromosome were selected on LB agar containing 10% sucrose (the sacB gene codes for levansucrase, which is toxic to cells grown in the presence of high concentrations of sucrose) (12). PCR was used to screen for mutants that had undergone a recombination event that resulted in a chromosomal deletion of the batB gene, and Western blot analysis was used to confirm that BatB was not produced in these strains (see Fig. 7A and D). To construct a B. bronchiseptica strain containing a 1,593-bp deletion corresponding to the deletion naturally present in the batB allele of B. parapertussis_hu, we amplified a 643-bp region from B. parapertussis_hu batB surrounding the 1,593-bp deletion and cloned it into the allelic exchange vector pEG7S. The resulting plasmid (pCW81) was transformed into E. coli SM10λpir and then mobilized into RB50 by conjugation. Cointegrants and exconjugants were selected as described above. The resulting strain, RBat_pp, was confirmed by PCR and Western blot analysis (see Fig. 7C and E).

FIG. 7.

BatB is necessary for persistence in the lower respiratory tract of mice. (A) Construction of strains used. RBAT was created using allelic exchange to delete codons 4 to 2297 of batB from the chromosome of RB50. RBAT was complemented with plasmid pCW98, which, when integrated at the promoter region of batB in the RBAT chromosome, leaves the original deletion of batB and the genes downstream of the deletion unchanged. Suc, sucrose. (B) Three- to 4-week-old BALB/c mice were inoculated with 5 × 10⁵ CFU of RB50 (black circles with solid black line), RBAT (gray triangles with dashed gray line), RBAT::pCW98 (gray circles with solid gray line), or RBat_pp (black triangles with dashed black line). A dotted line indicates the lower limit of detection. An asterisk denotes a result different from that for RB50, with a P value of ≤ 0.05. Two asterisks denote a result that is significantly different from that for RB50, with a P value of ≤0.01. Two crosses indicate that the result is significantly different from that for RBAT, with a P value of ≤0.01. (C) RBat_pp was created by deleting the same 1,593 nucleotides that are missing within batB of B. parapertussis_hu from the RB50 chromosome. (D) Outer membrane preparations of RB50, RBAT::pCW98, and RBAT were probed with anti-BatB antibody (α-BatB). (E) Outer membrane preparations of RB50, RBAT, RBat_pp, and 12822 (Bpp) were probed with anti-BatB antibody.

The batB gene was cloned from RB50 using plasmid rescue. A 510-bp fragment immediately 3′ of the B. bronchiseptica batB coding sequence was amplified and cloned into pEG7 (1). The resulting plasmid, pCW89, was transformed into E. coli SM10λpir and then mobilized into RB50, and cointegrants were selected on the appropriate medium. Chromosomal DNA was purified from the resulting strain RB50::pCW89, digested with EcoRI, and then ligated. The resulting plasmid, pCW91, contained the entire batB allele. We originally tried to complement the mutation in strain RBAT by cloning batB and its promoter into pBBR1MCS (24), a medium-copy plasmid (∼30/cell) that replicates in Bordetella. When pBBR1MCS containing batB with its native promoter was transferred into RBAT, the resulting colonies grew very slowly or were bvg mutants. To overcome this problem, we created a suicide plasmid containing batB and its promoter. This plasmid, pCW98, was transformed into E. coli SM10λpir and then mobilized into RBAT by conjugation. Since pCW98 contained homology to the entire batB allele, it could integrate either 5′ or 3′ of the batB deletion in RBAT. Only the integration of pCW98 into the 5′ end of the deletion would result in a strain with the deletion and the region downstream of the deletion intact (see Fig. 7A), ensuring that batB was truly complemented in RBAT. PCR was used to confirm that the plasmid integrated into the 5′ region of the batB deletion.

To construct strains harboring lacZYA fusions with the batB promoter region, we amplified a 510-bp region upstream of the batB translational start sites of B. bronchiseptica, B. pertussis, and B. parapertussis_hu. We cloned each promoter region into pSS3110 (40). The resulting plasmids, pCW85 (B. bronchiseptica batB promoter region), pCW86 (B. pertussis batB promoter region), and pCW87 (B. parapertussis_hu batB promoter region), were sequenced to confirm that no PCR errors had been introduced (Laragen, Los Angeles, CA) and were transformed into E. coli SM10λpir and mobilized into RB50, BPSM, and 12822 by conjugation. The integration of these plasmids into the correct sites on the chromosome was confirmed by PCR.

Bacterial conjugations.

Matings between B. bronchiseptica strains and E. coli strain SM10λpir were achieved by mixing stationary-phase cultures of each strain on BG agar plus 7.5% sheep blood in a 10:1 (B. bronchiseptica/E. coli) ratio. The mating mixture was incubated at 37°C for 5 h and then plated onto BG agar plus 7.5% sheep blood containing gentamicin and streptomycin to select for cointegrants.

Bacterial cell fractionation and immunoblot analysis.

Stationary-phase cultures were used for all bacterial cell preparations. WCLs were prepared as described previously (27). Outer membrane preparations were isolated from bacterial cell sonicates as the Triton X-100-insoluble fraction. Briefly, clarified cell sonicates were centrifuged at 16,000 × g, the pellet was resuspended in 2% Triton X-100 and centrifuged again, and the pellet was resuspended in an appropriate volume of 10 mM HEPES (pH 7.5) and 1× sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer.

SDS-PAGE was performed by the method of Laemmli (25) using denaturing 8% gels (acrylamide-bisacrylamide, 29:1). Proteins from whole cells or outer membranes were run on SDS-PAGE and transferred to nitrocellulose membranes (Schleicher and Schuell BioScience, Dassel, Germany). Membranes were incubated with a 1:10,000 dilution of chicken polyclonal FHA antibody (23), a 1:1,000 dilution of preimmune IgY from chickens, a 1:10,000 dilution of rabbit polyclonal BatB antibody (see below), or a 1:1,000 dilution of preimmune sera collected from naïve rabbits, rats, mice, or guinea pigs. Antigen-antibody complexes were detected by incubation with a 1:20,000 dilution of goat anti-chicken IRdye 800-conjugated secondary antibody (Rockland, Gilbertsville, PA), a 1:20,000 dilution of goat anti-rabbit IRdye 800-conjugated secondary antibody (Li-Cor Biotechnologies, Lincoln, NE), a 1:20,000 dilution of goat anti-rat Alexa Flour 680-conjugated secondary antibody (Invitrogen Inc., Carlsbad, CA), a 1:20,000 dilution of goat anti-mouse Alexa Flour 680-conjugated secondary antibody (Invitrogen Inc., Carlsbad, CA), or a 1:20,000 dilution of goat anti-guinea pig IRdye 700-conjugated secondary antibody (Rockland, Gilbertsville, PA). The complexes then were visualized using the Odyssey infrared imaging system (Li-Cor Biosciences, Lincoln, NE). Purified horseradish peroxidase (HRP)-conjugated mouse IgG Fc and Fab fragments were purchased from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA), and used at a 1:5,000 dilution.

Generation of BatB-specific antibody.

To generate the BatB-specific antibody, we cloned the nucleotides corresponding to amino acids 590 to 802 of BatB into expression vector pET28 (Novagen, Darmstadt, Germany). The peptide was purified to a concentration of 12 μg/μl and injected along with Freund's adjuvant (Pacific Immunology, Ramona, CA) into specific-pathogen-free New Zealand White rabbits. Rabbits were boosted every 2 weeks, and sera were collected 2 months postimmunization.

Total RNA isolation and cDNA synthesis.

Total RNA was isolated from strains RB50, RB53, RB53i, RB54, BPSM, and 12822 grown overnight under Bvg⁺-phase conditions (SS broth, 37°C incubation on a roller) and/or Bvg⁻-phase conditions (SS broth supplemented with 12 mM MgSO₄, 37°C incubation on a roller) by following the protocol of the RNAqueous-4PCR kit from Ambion Inc. (Austin, TX). For the reverse transcription step, 10 ng of total RNA was transcribed using oligo(dT) and random priming by following the protocol supplied with Super Script II reverse transcriptase (Invitrogen Inc., Carlsbad, CA).

mRNA quantification.

Relative levels of batB and recA transcripts were determined using quantitative real-time PCR (qRT-PCR) using the following primers: RTrecAF (5′GCCAGGGCAAGGACAATGT′3), RTrecAR (5′CTTCGCTGGCGGGAAG′3), RTbatBF (5′TCGCAGCCACCAAGATACTCAGTT′3), and RTbatBR (5′TCAATGCCGGTACGCTGCAGATAT′3). qRT-PCRs were performed in SYBR green Super Mix (Bio-Rad Laboratories, Hercules, CA) using the Bio-Rad iCycler PCR machine and software (Bio-Rad Laboratories, Hercules, CA). All samples were run in triplicate, and batB transcription was normalized to recA transcription for each sample.

β-Galactosidase assays.

Stationary-phase cultures from strains harboring lacZ transcriptional fusions were used to measure β-galactosidase activity. Bacteria were grown overnight in SS broth and then permeabilized by the addition of SDS and CHCl₃, and β-galactosidase activity was determined essentially as described previously (27), except that measurements were taken using a Victor3 1420 microplate reader (PerkinElmer Life Sciences, Boston, MA).

Rat colonization experiments.

Female Wistar rats (3 to 4 weeks old; Charles River Laboratories) were sedated lightly with isoflurane and then inoculated with 1,000 CFU of bacteria in 10 μl phosphate-buffered saline (PBS) that was pipetted directly into the nares. Animals were euthanized by halothane inhalation at days 14 and 24 postinoculation. Tracheas and nasal septa were collected and homogenized, and the number of CFU was determined as described previously (28). All experiments were performed as described in the animal use protocols that have been approved by the UCSB IACUC (protocol no. 6-01-601).

Mouse lung inflammation experiments.

BALB/c and SCID mice (3 to 4 weeks old; Charles River Laboratories) were sedated with isoflurane and inoculated intranasally by pipetting 50 μl PBS containing 5 × 10⁵ CFU of Bordetella into the nasal passage. Animals were euthanized by halothane inhalation, and nasal septa, tracheas, and lungs were collected at 0, 4, 7, 11, and 21 (BALB/c only) days postinoculation. Tissues were homogenized, and the numbers of CFU were determined as previously described (28). All experiments were performed as described in the animal use protocols that have been approved by the UCSB IACUC (protocol no. 6-01-601).

RESULTS

Serendipitous discovery of batB in B. bronchiseptica.

The Western blot characterization of a chicken-derived polyclonal antibody generated against the central region of FHA revealed that both pre- and postimmune antibody preparations detected a protein of ∼180-kDa that was present in WCLs of wild-type and Bvg⁺-phase-locked strains of B. bronchiseptica (RB50 and RB53, respectively) but not the WCLs of Bvg-intermediate (Bvgⁱ)- or Bvg⁻-phase-locked strains (RB53i and RB54, respectively) (Fig. 1A and B). The protein was detected in WCLs of RBX9, a ΔfhaB derivative of RB50, demonstrating that it was not a degradation product of FHA. The protein appeared to be an integral outer membrane protein, as it was present in outer membrane-enriched fractions but not culture supernatants (Fig. 1C and data not shown). Because this 180-kDa protein was expressed only in Bvg⁺-phase bacteria and was present in the outer membrane, we hypothesized that it was involved in pathogenesis and therefore sought the identity of its structural gene. Given its outer membrane location, we hypothesized further that the protein was one of the ATs predicted from the B. bronchiseptica genome sequence. Two of the open reading frames (ORFs) predicted to encode ATs were large enough to encode proteins of at least 180 kDa. These were BB0452 and BB0450. We constructed a suicide plasmid (pRH4) to disrupt BB0452 but were unsuccessful in our attempts to construct a plasmid to disrupt BB0450. The integration of plasmid pRH4 into the chromosome of RB50 resulted in a strain (RB50::pRH4) that did not produce the 180-kDa protein (Fig. 1D). We then constructed a strain containing a large in-frame deletion of BB0452 by allelic exchange. This strain, called RBAT, is missing all but the first four codons and last three codons of BB0452 from the chromosome. The 180-kDa protein was not detected in WCLs of RBAT (Fig. 1D). These data indicate that BB0452 is required for the production of the 180-kDa outer membrane protein in B. bronchiseptica.

FIG. 1.

Preimmune chicken IgY recognized an unknown protein of ∼180 kDa in B. bronchiseptica. (A) WCLs of Bvg⁻-phase-locked strain RB54, Bvgⁱ-phase-locked strain RB53i, Bvg⁺-phase-locked strain RB53, and ΔfhaB strain RBX9 were probed with a chicken-derived polyclonal anti-FHA antibody (α-FHA). The arrowhead indicates FHA, and the asterisk denotes the 180-kDA protein. (B) WCLs of RB50 grown under Bvg⁺-phase conditions [RB50(+)]; RB50 grown under Bvg⁻-phase conditions [RB50(−)]; RB53, RB53i, RB54, and RBX9 grown under Bvg⁺-phase conditions [RBX9(+)]; and RBX9 grown under Bvg⁻-phase conditions [RBX9(−)] were probed with preimmune chicken IgY. (C) Outer membranes prepared from RB50 grown under Bvg⁺-phase conditions, RB50 grown under Bvg⁻-phase conditions, and RBX9 grown under Bvg⁺-phase conditions were probed with preimmune chicken IgY. (D) WCLs of RB50, RB50::pRH4, and RBAT grown under Bvg⁺-phase conditions were probed with preimmune chicken IgY.

BB0452 was previously named batB (for Bordetella autotransporter B) by Henderson et al., who compiled a list of all known and putative Bordetella ATs (21). Although identified in this review, there has been no reported experimental characterization of batB. batB homologs are present in the sequenced genomes of B. pertussis strain Tohama I (BP0529) and B. parapertussis_hu strain 12822 (BPP0452). The predicted translational start site of batB is annotated differently for each of the sequenced Bordetella strains (http://www.ncbi.nlm.nih.gov) (Fig. 2A). As all ATs characterized so far have either canonical N-terminal signal sequences or canonical signal sequences preceded by an N-terminal extension (21), it is likely that BatB also contains an N-terminal signal sequence. We entered the first 60 amino acids from each predicted BatB protein sequence into the signal sequence prediction program SignalP 3.0 (http://www.cbs.dtu.dk/services/SignalP/). Only the BatB sequence predicted for B. parapertussis_hu strain 12822 contained a probable signal sequence. It is most likely, therefore, that the translational start sites of batB in RB50 (B. bronchiseptica) and Tohama I (B. pertussis) are the same as the ATG codon annotated in the 12822 genome (Fig. 2A). Based on these ATG translational start sites, a comparison of the predicted amino acid sequences of BatB indicates that BatB of RB50 and Tohama I are 96% identical (Fig. 2B). The predicted BatB proteins of RB50 and 12822 are 98% identical, exclusive of amino acids 1092 to 1623 relative to BatB of RB50 (Fig. 2B), which are missing in BatB of 12822 due to a 1,593-bp deletion within the batB gene.

FIG. 2.

Comparison of B. bronchiseptica, B. pertussis, and B. parapertussis_hu. (A) Alignment of the intergenic region of batB from B. bronchiseptica strain RB50, B. pertussis strain Tohama I, and B. parapertussis_hu strain 12822. The stop codon of the upstream ORF is gray. Nucleotide differences from the B. bronchiseptica sequence are in boldface type. The batB translational start sites denoted by the Sanger Institute are different in B. bronchiseptica strain RB50 (red box), B. pertussis strain Tohama I (purple box), and B. parapertussis_hu strain 12822 (green box). The translational start site denoted for B. parapertussis_hu most likely is the correct translational start site, since it is predicted to encode a signal sequence. All BatB comparisons in this paper use this as the translational start (green lettering). (B) Comparison of BatB from strains RB50 (Bb), Tohama I (Bp), and 12822 (Bpp). Red lines represent single-amino-acid changes away from the B. bronchiseptica sequence, and blue lines represent amino acid changes that are the same in B. pertussis and B. parapertussis_hu. The BatB signal sequence is represented in black, the region homologous to IgA proteases is blue, the PL2 domain is light gray, and the β-domain is purple. The inverted arrows indicate the region of the protein that was disrupted by plasmid insertion. The solid line below the sequence indicates the region of the protein used to generate an anti-BatB antibody (α-BatB). (C) WCLs of B. bronchiseptica strain RB50, B. pertussis strain BPSM (a streptomycin-resistant derivative of Tohama I), and B. parapertussis_hu strain 12822 grown under Bvg⁺- and Bvg⁻-phase conditions were probed with anti-BatB antibody.

A BLAST search of the RB50 BatB sequence against other microbial protein sequences (http://www.ncbi.nlm.nih.gov/BLAST/) revealed several conserved domains, including a β-domain, which is found in all ATs, a PL2 Prn-like domain, and a region with similarity to IgA protease family proteins (Fig. 2B). The similarity with IgA proteases did not encompass the catalytic domain.

We produced a polyclonal antibody against amino acids 590 to 802 of BatB from B. bronchiseptica strain RB50 after expressing the corresponding gene fragment in E. coli and purifying the resulting polypeptide (Fig. 2B). This antibody detected a protein of ∼180 kDa in WCLs of Bvg⁺-phase B. bronchiseptica strain RB50, but no protein was detected in the batB deletion strain RBAT (Fig. 2C). These data confirm that the 180-kDa protein detected by preimmune chicken sera was BatB. The anti-BatB antibody detected a protein of ∼120 kDa in WCLs of Bvg⁺-phase B. parapertussis_hu strain 12822, which is consistent with its predicted size (Fig. 2B and C), and protein was not detected in B. parapertussis_hu harboring the batB disruption plasmid pRH4 (12822::pRH4) (Fig. 2C). Unexpectedly, the BatB antibody did not detect any protein in WCLs of Bvg⁺-phase B. pertussis strain BPSM (a streptomycin-resistant derivative of Tohama I) (Fig. 2C).

batB transcription is regulated by the BvgAS phosphorelay.

To measure batB transcription, we performed qRT-PCR on RB50 grown under Bvg⁺-phase conditions, Bvg⁺-phase-locked strain RB53, Bvgⁱ-phase-locked strain RB53i, and Bvg⁻-phase-locked strain RB54. Consistent with the Western blot results, batB transcript levels were high in both RB50 grown under Bvg⁺-phase conditions and RB53 (Fig. 3). Minimal levels of batB transcript were detected in the Bvgⁱ-phase-locked strain RB53i, and almost no transcript was detected in ΔbvgS strain RB54 (Fig. 3). These data demonstrate that batB transcription is dependent on both the presence and activation of the BvgAS phosphorelay.

FIG. 3.

Relative batB mRNA levels in B. bronchiseptica. qRT-PCR was used to determine the relative levels of batB transcript in wild-type RB50 (light gray bar), Bvg⁺-phase-locked mutant RB53 (white bar), Bvgⁱ-phase-locked mutant RB53i (dark gray bar), and ΔbvgS strain RB54 (black bar).

batB is not transcribed in B. pertussis due to nucleotide differences within the promoter region.

batB transcription was compared among B. bronchiseptica strain RB50, B. parapertussis_hu strain 12822, and B. pertussis strain BPSM grown under both Bvg⁺-phase and Bvg⁻-phase conditions by qRT-PCR (Fig. 4). Consistent with the Western blot results, batB was transcribed at a high level in both RB50 and 12822 grown under Bvg⁺-phase conditions but not when either strain was grown under Bvg⁻-phase conditions (Fig. 4). batB transcription was negligible in strain BPSM grown under both Bvg⁺- and Bvg⁻-phase conditions (Fig. 4). We hypothesized that the lack of batB transcription in BPSM was due either to nucleotide differences within its promoter region or the differential presence or activity of intermediate regulatory factors in BPSM compared to that of strains RB50 and 12822. An alignment of the 445-bp intergenic regions between the ORF 5′ of batB (BB0451, BPP0451, and BP0527) and the predicted translational start site of batB revealed six nucleotide differences between RB50 and BPSM and four nucleotide differences between RB50 and 12822 (Fig. 2A). We cloned each intergenic region upstream of the promoterless lacZ (′lacZ) gene in plasmid pSS3110 (40), a suicide plasmid that integrates into a nonessential region of the Bordetella chromosome. We then introduced each plasmid into RB50, BPSM, and 12822 and measured β-galactosidase activity in the cointegrant strains grown under Bvg⁺-phase conditions. All strains containing plasmids with either the RB50 or the 12822 intergenic region upstream of ′lacZ displayed high levels of β-galactosidase activity (Fig. 5). β-Galactosidase activity was minimal, however, in all strains containing the plasmid with the BPSM intergenic region upstream of ′lacZ (Fig. 5). Differences within the BPSM batB promoter region compared to the batB promoter regions of RB50 and 12822 therefore are responsible for the lack of batB transcription in BPSM.

FIG. 4.

Relative batB mRNA levels in B. bronchiseptica, B. pertussis, and B. parapertussis_hu. qRT-PCR was used to determine the relative levels of batB transcripts in B. bronchiseptica, B. pertussis, and B. parapertussis_hu grown in Bvg⁺-phase conditions (gray bars) or Bvg⁻-phase conditions (black bars).

FIG. 5.

batB is not transcribed in B. pertussis. β-Galactosidase activity from the batB promoter from B. bronchiseptica (Bb promoter), B. pertussis (Bp promoter), and the B. parapertussis_hu promoter (Bpp promoter) in B. bronchiseptica strain RB50 (black bars), B. pertussis strain BPSM (white bars), and B. parapertussis_hu strain 12822 (gray bars) grown under Bvg⁺-phase conditions.

BatB is not essential for respiratory tract colonization in a rat model of infection.

We used a rat model to determine the contribution of batB to the colonization of the upper and lower respiratory tract. Female Wistar rats were inoculated with 1,000 CFU of either wild-type B. bronchiseptica strain RB50 or ΔbatB strain RBAT delivered in a small volume (10 μl) to the nares. The colonization of tracheas and nasal septa was determined at days 14 and 28 postinoculation. RBAT was recovered at levels similar to those of RB50 at both time points in both tissues (Fig. 6), indicating that batB is not necessary for B. bronchiseptica to colonize the upper and lower respiratory tracts of rats.

FIG. 6.

BatB is not necessary for the colonization of the nasal septum and trachea of rats. Three- to 4-week-old female Wistar rats were inoculated with 1,000 CFU of either RB50 (black circles) or RBAT (white circles). Each circle represents the number of CFU recovered from a single rat. The horizontal line represents the average number of CFU. The dashed line represents the lower limit of detection.

BatB is required for B. bronchiseptica to resist inflammatory clearance.

Although it is a contrived model, the inoculation of mice with 5 × 10⁵ CFU of wild-type B. bronchiseptica delivered in a larger (50 μl) volume has proven to be a useful tool for investigating the ability of B. bronchiseptica to induce an inflammatory response and resist clearance by inflammatory cells (19, 22). We inoculated 3- to 4-week-old female BALB/c mice with wild-type B. bronchiseptica RB50 or RBAT (ΔbatB) and determined the number of CFU in nasal septa, tracheas, and lungs at 4, 7, 11, and 21 days postinoculation. Additional animals were sacrificed 1 to 3 h postinoculation to confirm that similar numbers of RB50 and RBAT were delivered to each site initially (Fig. 7B). RB50 and RBAT were recovered in similar numbers from nasal septa at all time points (Fig. 7B). The numbers of CFU of RBAT recovered from the lungs and tracheas, however, were dramatically lower than the numbers of CFU of RB50 recovered at all time points (Fig. 7B). To determine if the decreased recovery of RBAT in the lower respiratory tract was due to a lack of batB expression, we complemented the ΔbatB mutation in RBAT by integrating a plasmid containing wild-type batB and its promoter region into the RBAT chromosome (Fig. 7A and D; also see Materials and Methods). This plasmid was integrated such that the batB deletion mutation and sequences downstream of the deletion were the same as those in the RBAT strain (Fig. 7A and Materials and Methods). As described in Materials and Methods, this approach was used because the expression of batB from a multicopy plasmid was not tolerated. Similar numbers of CFU were recovered from mice inoculated with the complemented strain (RBAT::pCW98) and from mice inoculated with RB50 (Fig. 7B), confirming that the defect in the persistence of RBAT in the mouse lower respiratory tract was due to the absence of batB.

To determine if the 531 amino acids missing from BatB of B. parapertussis_hu strain 12822 that are present in RB50 contribute to BatB function in vivo, we deleted the same 1,593 nucleotides from RB50 batB that are missing in 12822 batB by allelic exchange (Fig. 7B and E). We then analyzed this mutant, called RBat_pp, in the mouse model. RBat_pp colonized the nasal septa of mice at levels similar to those of RB50 and RBAT (Fig. 7B). However, at days 4, 7, and 11 postinoculation, RBat_pp was recovered at significantly lower levels in the lungs compared to those for RB50 (Fig. 7B). This level was significantly higher, however, than the level of RBAT recovered from lungs at days 4 and 7. By day 11, RBat_pp and RBAT were recovered at similar levels in the lungs (Fig. 7B). The level of colonization of RBat_pp in the trachea was significantly lower than that of RB50 at days 4, 7, and 11 but was higher than that of RBAT at days 4 and 7. These data suggest that the BatB protein produced by B. parapertussis_hu is partially functional.

BatB binds Ig.

Although it does not appear to contain a catalytic domain, BatB shares significant sequence similarity with IgA proteases. We hypothesized, therefore, that the detection of BatB in Western blots using preimmune chicken IgY that we observed initially resulted from BatB binding to Ig rather than the antigen-specific binding of chicken IgY to BatB. To explore this possibility, outer membrane preparations of wild-type B. bronchiseptica strain RB50 and the ΔbatB strain RBAT were analyzed by Western blotting using preimmune sera collected from naïve rabbits, guinea pigs, mice, and rats. BatB was detected in all cases (Fig. 8A). To examine this hypothesis more rigorously, we probed the outer membrane fractions of both RB50 and RBAT with commercially available HRP-conjugated Fab and Fc fragments purified from IgG collected from naive mice. Both Fab and Fc fragments bound to BatB (Fig. 8B). Taken together, these data support the hypothesis that BatB binds Ig.

FIG. 8.

(A) BatB binds preimmune sera from rabbit, guinea pig, mouse, and rat. Outer membrane fractions prepared from wild-type B. bronchiseptica strain RB50 and ΔbatB strain RBAT were probed with preimmune sera collected from naïve rabbit, guinea pig, mouse, or rat. (B) BatB binds to both Fc and Fab regions of preimmune mouse IgG. Outer membranes prepared from RB50 and RBAT were probed with commercially available HRP-conjugated Fc fragments or HRP-conjugated Fab fragments purified from preimmune mouse IgG.

BatB function in vivo is dependent on B and T cells.

To explore the possibility that BatB's role in resisting clearance from the lower respiratory tract during the first week postinoculation is related to its ability to bind Ig, we inoculated 3- to 4-week-old SCID mice, which do not produce Ig due to the lack of B and T cells, with 5 × 10⁵ CFU of RB50, RBAT, and RBat_pp and determined the level of colonization in nasal septa, tracheas, and lungs at 4, 7, and 11 days postinoculation. (Although not specifically defective in the production of antibodies, SCID mice were chosen for these studies because they are available on the BALB/c background, which allows comparison with the results described above.) Two additional animals in each group were sacrificed 1 to 3 h postinoculation to confirm that similar amounts of RB50, RBAT, and RBat_pp were delivered initially to each site (Fig. 9). RB50, RBAT, and RBat_pp were recovered at similar levels in both nasal septa and tracheas at all time points postinoculation (Fig. 9). At day 4 postinoculation, the number of CFU of RB50 recovered from the lungs of SCID mice was slightly higher than the number of CFU of RBAT and RBat_pp recovered (Fig. 9). Although the small difference in the number of CFU of RB50, RBAT, and RBat_pp recovered from the lungs at day 4 postinoculation was statistically significant, the inability of RBAT and RBat_pp to resist clearance in the lower respiratory tract (tracheas and lungs) generally was eliminated in this mouse strain (compare Fig. 7B and 9). Unlike the results for BALB/c mice, the numbers of CFU of RB50, RBAT, and RBat_pp recovered from the lungs of SCID mice at days 7 and 11 postinoculation were similar. These data indicate that the increased clearance of RBAT and RBat_pp compared to that of RB50 from the tracheas and lungs of BALB/c mice is due to some function provided by B and T cells, possibly the production of nonspecific Ig.

FIG. 9.

BatB is not necessary for persistence in the lower respiratory tract of mice lacking B and T cells. Three- to 4-week-old SCID mice were inoculated with 5 × 10⁵ CFU of RB50 (black circles with solid black line), RBAT (gray triangles with dashed gray line), or RBat_pp (black triangles with dashed black line). A dotted line indicates the lower limit of detection. Two asterisks denote a result that is significantly different from that for RB50, with a P value ≤0.01.

BatB production in multiple B. bronchiseptica, B. pertussis, and B. parapertussis_hu isolates.

We have shown that the sequenced B. pertussis strain Tohama I does not produce BatB, and the sequenced B. parapertussis_hu strain 12822 produces a smaller BatB protein compared to that of the sequenced B. bronchiseptica strain RB50. To determine if BatB production by RB50, Tohama I, and 12822 is representative of each species, we prepared WCLs from several B. bronchiseptica, B. pertussis, and B. parapertussis_hu isolates within our strain collection and probed them with anti-BatB antibody. BatB of the same size as that of RB50 was present in WCLs of all B. bronchiseptica isolates tested. No B. pertussis isolate tested produced BatB (Table 2). All B. parapertussis_hu isolates produced a BatB protein similar in size to that of strain 12822 (Table 2). These data suggest that full-length (and fully functional) BatB is produced only in non-human-adapted Bordetella subspecies.

TABLE 2.

BatB expression among Bordetella isolates

Species	Isolate name	BatB expression
B. bronchiseptica	RB50	+
B. bronchiseptica	CB1A	+
B. bronchiseptica	MB-9	+
B. bronchiseptica	RBSB	+
B. bronchiseptica	KM22	+
B. bronchiseptica	JC100	+
B. bronchiseptica	545	+
B. bronchiseptica	548	+
B. bronchiseptica	590	+
B. bronchiseptica	680	+
B. bronchiseptica	705	+
B. bronchiseptica	723	+
B. bronchiseptica	782	+
B. pertussis	BPSM	−
B. pertussis	GMT1	−
B. pertussis	18323	−
B. pertussis	6068	−
B. pertussis	CS	−
B. parapertussishu	12822	+ (∼120 kDa)
B. parapertussishu	BppHU1	+ (∼120 kDa)
B. parapertussishu	BppHU3	+ (∼120 kDa)
B. parapertussishu	BppHU6	+ (∼120 kDa)

DISCUSSION

The sequenced genomes of the mammalian-adapted bordetellae are predicted to encode 22 different ATs, all of the classical type. All five that have been studied so far (Prn, BrkA, SphB1, Vag8, and TcfA) have been shown to contribute to virulence-related phenotypes (9, 14-16, 26, 29, 34, 36). We report here the serendipitous discovery of BatB and show that it, too, contributes to pathogenesis. In B. bronchiseptica and B. pertussis, BatB is predicted to be a 229-kDa AT protein with a typical signal sequence, an ∼186-kDa passenger domain, and an ∼30-kDa β-domain. The predicted β-domain of BatB is 27% identical and 43% similar to the β-domain of the well-characterized Bordetella AT Prn, and we have shown that a protein of ∼180 kDa that is recognized by antibodies generated against amino acids 590 to 802 is present in outer membrane fractions of B. bronchiseptica. These results suggest that BatB is a classical AT and that its passenger domain, like those of many other classical ATs, is cleaved from the β-domain but remains noncovalently associated with the cell surface. We also showed that, as predicted from the genome sequence, B. parapertussis_hu produces a BatB protein with an outer membrane-associated passenger domain of ∼120 kDa. Unexpectedly, B. pertussis did not produce BatB under any of the in vitro growth conditions tested, which included growth in SS broth at 37°C, a condition under which all known Bordetella virulence factors are expressed. The lack of BatB production under these conditions was found to be due to nucleotide differences in the intergenic region between batB and the ORF 5′ of batB in B. pertussis and those regions in B. bronchiseptica and B. parapertussis_hu. While we did not map the batB promoter precisely, the fact that qRT-PCR data correlate with lacZ fusion data obtained using plasmids containing the intergenic region between batB and the ORF 5′ of batB suggests strongly that the batB promoter lies within the intergenic region. Our data also show that the transcription of batB, like that of all known or putative Bordetella virulence factor-encoding genes identified so far, is positively regulated by the BvgAS phosphorelay.

Our in vivo studies showed that although a batB deletion mutant did not differ from wild-type B. bronchiseptica in its ability to colonize the nasal septa and tracheas of rats, it was recovered at dramatically reduced levels from the lower respiratory tracts of mice beginning at day 4 postinoculation. This apparent disparity in virulence between these two models has been observed for strains containing mutations in other virulence genes as well (e.g., cyaA, the gene encoding adenylate cyclase toxin [ACT], and sphB1, which encodes a serine protease AT) (8, 19, and our unpublished data). We hypothesize that although the rat colonization model is highly sensitive with regard to the establishment of infection (fewer than 20 CFU delivered in a 5-μl droplet to the nares are required) and likely reflects a natural course of persistent, asymptomatic colonization, it may be relatively insensitive with regard to revealing phenotypes for virulence factors that perform redundant or overlapping functions. The mouse lung inflammation model, on the other hand, while artificial and unnatural (it has even been used to study enteric pathogens such as Shigella flexneri and Vibrio cholerae [3, 17, 35]), is highly sensitive in its ability to identify bacterial virulence factors involved in the induction or suppression of inflammation and/or defense against inflammatory responses. Our data indicate that the batB mutant is defective at overcoming innate immunity and suggest that, like ACT, BatB plays a role in resisting the microbicidal action of inflammatory cells. In fact, the defect displayed by the batB mutant in the mouse model is more pronounced than that of cyaA mutants (19 and our unpublished data), suggesting that BatB plays an even more critical role than ACT in this regard.

Despite the fact that the batB phenotype manifests within the first few days postinoculation, which is before the induction of an antigen-specific adaptive immune response, we explored the possibility that the Ig-binding ability of BatB observed in vitro was related to its function in vivo. Wild-type B. bronchiseptica increased to similar numbers in the lungs of both wild-type and SCID mice at days 4 and 7 postinoculation. This result suggests that B and T cells are not contributing to the control of B. bronchiseptica during this time. The growth of the ΔbatB mutant, however, was severely restricted in wild-type mice but not in SCID mice, indicating that B- and T-cell functions can participate in the early clearance of B. bronchiseptica if B. bronchiseptica does not express batB; i.e., these results indicate that BatB functions to abrogate B- and T-cell-mediated activities during the first few days postinoculation. The control of B. bronchiseptica in the mouse model is absolutely dependent on polymorphonuclear neutrophils (PMN); PMN-deficient animals die within 4 days postinoculation (19). Taken together, these data are consistent with the hypothesis that BatB prevents nonspecific antibodies in the lungs from facilitating PMN-mediated clearance. (Although the similarity between BatB and IgA proteases does not encompass the IgA protease catalytic domain, the possibility that BatB functions as a protease, or inhibits antibody function in another way, has not been tested.) Wijburg et al. recently demonstrated the importance of innate secretory antibodies in the control of pathogenic bacteria in the gastrointestinal tract (41). While many more experiments will be required to draw definitive conclusions (the most important being those using mice that are deficient specifically in the production of antibodies, such as μMT mice), our results are consistent with the hypothesis that innate antibodies play a similarly important role in the lower respiratory tract. If so, it will be interesting to determine if normal flora (in either the gut or the nasopharynx) are required for the production of the protective innate (secretory) antibodies in the respiratory tract, as Wijburg et al. showed for the gastrointestinal tract (41).

The B. parapertussis_hu BatB protein and a BatB-LacZ fusion protein containing the N-terminal 210 amino acids of BatB bind antibody (data not shown). A B. bronchiseptica strain expressing a batB gene that is missing the same nucleotides that are missing in batB_Bpp but present in batB_Bb displayed a phenotype intermediate between those of wild-type B. bronchiseptica and the ΔbatB mutant in wild-type mice, and the defect (compared to the phenotype of wild-type B. bronchiseptica) was abrogated in SCID mice (Fig. 7 and 9). These data suggest that BatB performs at least one function in addition to antibody binding (which may or may not be dependent on antibody binding), and that antibody binding alone is sufficient for BatB to contribute, at least somewhat, to resisting inflammation-mediated clearance. One possibility is that BatB binds antibody molecules and then inactivates them, perhaps via proteolysis and in a catalytic manner like IgA proteases, and that simply binding antibody molecules renders them inactive, at least while bound, such that the B. parapertussis_hu BatB protein (or B. parapertussis_hu-like BatB protein expressed in our mutant B. bronchiseptica strain) inactivates only a small number of antibody molecules relative to the number inactivated by the B. bronchiseptica BatB protein. We are currently performing experiments to test these hypotheses.

None of the B. pertussis strains included in our study produced BatB, and all of the B. parapertussis_hu strains investigated produced BatB proteins that were similar in size to BatB of strain 12822. While it has been established that B. pertussis and B. parapertussis_hu diverged independently and at different times from B. bronchiseptica-like ancestors, the lineages from which each diverged is unknown (11, 39). All B. bronchiseptica strains included in our study, however, produced BatB proteins that were similar in size to BatB of strain RB50. The loss of the ability to produce a B. bronchiseptica-like BatB protein (under Bvg⁺-phase conditions, at least), therefore, appears to have occurred when B. pertussis and B. parapertussis_hu strains diverged from B. bronchiseptica, suggesting it represents an important step in the evolution of the strictly human-adapted bordetellae. While it is possible that the B. bronchiseptica-like BatB protein simply does not contribute to human infection or to infectious cycles that are limited to growth within the human respiratory tract, both B. pertussis and B. parapertussis_hu strains appear to have retained some genes that they apparently never express (10, 34). It seems more likely, therefore, that the B. bronchiseptica-like BatB protein is incompatible with strict human adaptation. Whether the production of a B. bronchiseptica-like BatB protein is detrimental to human infection cannot be tested experimentally, but we are hopeful that our investigations into how BatB functions mechanistically will provide insight into this issue and also into understanding if the B. parapertussis_hu-like BatB protein performs an important function in the B. parapertussis_hu infectious cycle. It does appear, however, that BatB is not sufficient to broaden the host range of B. pertussis, because the expression of batB in B. pertussis strain BPSM did not alter its ability to infect mice (our unpublished observations).

Our future experiments will focus on determining how BatB functions at the molecular level to allow B. bronchiseptica to resist clearance from the lower respiratory tract during the first few days postinoculation. Although we suspect that Ig binding is involved, the data supporting this hypothesis are only correlative at this point. If Ig binding is involved, it will indicate an important role for nonspecific antibodies in the protection of the lower respiratory tract from infection, which has not been demonstrated previously. We are hopeful, therefore, that in addition to shedding light on the evolution of virulence and host specificity among bordetellae, our studies will reveal new insights into the function of innate immunity in the respiratory tract in general.