Bacterial strains and growth conditions

Sample preparation, protein electrophoresis and immunoblotting

For SDS-polyacrylamide gel electrophoresis (SDS-PAGE) sample preparation, bacteria were cultured in SS media overnight and harvested by centrifugation at 10,000 x g at 4°C for 10 min. The resulting supernatant, containing secreted proteins, was filtered through a 0.2-μm membrane to remove contaminating bacterial cells. Protein from supernatants (equivalent of 3.75 O.D.₆₀₀) was precipitated with 15% trichloroacetic acid (TCA) for 1 h on ice and samples were centrifuged at 15,000 g for 15 min at 4°C. After centrifugation, TCA was removed and the pellet was resuspended in 1 X SDS-loading dye with 25 mM freshly prepared DTT. To neutralize the acidic pH of the samples, a few crystals of tris-base were added. Protein pellets were dissolved by shaking over a bench-top shaker for 30 min at room temperature prior to fractionation by various fixed percentage or gradient (as indicated) pre-cast SDS-polyacrylamide gels (Bio-Rad). The pellet samples after normalization to 12.5 O.D.₆₀₀/ml, were boiled for 10 min in 1 x SDS-loading dye as above. After the run, proteins were either Coomassie stained or transferred onto a polyvinylidene difluoride (PVDF) membrane (Immobilon P, Millipore) using a semi-dry blot. BvgS, a non-secreted protein control was detected using polyclonal mouse antiserum at a dilution of 1:1000 [21]. Pertactin (PRN), which is secreted by a non-T3SS dependent pathway, was identified using a monoclonal mouse antibody at a dilution of 1:1000 [22]. Bsp22, a T3SS substrate control, was detected using polyclonal mouse serum at a dilution of 1:10,000 [23]. Immunodetection was carried out by chemifluorescence using horseradish peroxidase-labeled goat anti-mouse IgG and the ECL plus® detection substrate (GE Healthcare). Chemifluorescent signals were visualized using a Typhoon scanner (GE Healthcare).

Genomic DNA extraction, PCR-based detection and genome sequencing

DNA was extracted from overnight cultures of various isolates using the PureLink genomic DNA kit as per manufacturer’s instructions (Invitrogen Corporation, USA). PCR was performed according to the manufacturer's instructions (0.5 U of iproof polymerase, 200 μM each of the four dNTPs and 1 μM each primer) and supplemented with 3% dimethyl sulphoxide (DMSO). Primers B77_QseC1F (5^′- ATGACTTTGCAGCGCAGGTT −3^′) and B77_QseC1R (5^′- AGAAACGCGATCAGCACGGG −3^′) or primers B77_QseC2F (5^′- GGAGATCTTGCCGTCGCCAT-3^′) and B77_QseC2R (5^′-ACTTCCCATTGCGCGCGTAG-3^′) were used to amplify qseC sequences, and primers B77_QseB1F (5^′- GAGAATTCTTATTGTCGAAG-3^′) and B77_QseB1R (5^′- GATTCCCAGTCATACAGCTT −3^′) were used to amplify qseB. Cycling parameters were: one cycle of 98°C for 1 min; 25 cycles of 98°C for 10 s, 55°C for 20 s and 72°C for 30 s; and a final incubation at 72°C for 5 min. The PCR products were fractionated on 1% agarose gel using 1X TBE buffer containing 5 μg/ml ethidium bromide. PCR products of the extracted DNA were then purified for sequencing using Qiagen's QIAquick purification kit (Qiagen, Valencia, USA). Bordetella genomes were sequenced by the Sequencing Group at the Sanger Center and can be obtained from ftp://ftp.sanger.ac.uk/pub/pathogens/bp.

Construction of bscN and bteAin-frame deletion mutants

To construct in-frame deletions of codons 171–261 in the bscN locus, allelic exchange was performed using pEGBR1005 suicide plasmid derivatives as previously described by Yuk et al. [15]. For construction of bteA in-frame deletions (codons 4–653), suicide plasmid pRE112-bteA was used as previously described by Panina et al. [11]. All mutants were verified by sequencing target open reading frames.

Cell lines

Cell lines used in this study were obtained from the American Tissue Culture Collection (ATCC). Human Cervical Adenocarcinoma, HeLa (ATCC CCL-2^TM) and Mouse Monocyte-macrophage, J774A.1 (ATCC TIB-67^TM) cell lines were maintained in Dulbecco's Modified Eagle Medium(DMEM), and Human Lung Carcinoma, A-549 cells (ATCC CCL-185^TM) were maintained in Ham’s F-12 K medium (F-12 K) supplemented with 10% Fetal bovine serum (FBS) at 37°C with 5% CO_2.

Cytotoxicity assays

Bacteria were cultured in SS media overnight and were then sub-cultured in SS media to an optical density of ~0.5 at 600 nm. For cytotoxicity assays, bacteria were added to previously seeded cell monolayers in 12- or 24-well tissue culture plates at the indicated MOIs. The plates were centrifuged for 5 min at 60 x g and incubated for up to 4 h at 37°C with 5% CO_2. To measure cell cytotoxicity, Lactate dehydrogenase (LDH) release was used as a surrogate marker for cell death. LDH release in the supernatant media was assayed using a CytoTox 96® non-radioactive cytotoxicity assay kit (Promega, Madison, WI), according to the manufacturer's instructions. The maximal LDH release was defined as 100% and was determined by adding lysis solution to uninfected monolayers, determining the absorbance at 490 nm, and then subtracting the background value. Each sample was measured in triplicate in at least three independent experiments.

Animal infection experiments

Wild-type female C57BL/6NCr (B6) mice, 4–6 weeks of age, were purchased from Charles River Breeding Laboratories (Wilmington, MA). The animals were lightly sedated with isoflurane (Novation Laboratories, TX) prior to intranasal infection with the indicated number of CFU of bacteria in a total volume of 40 μl of phosphate-buffered saline (PBS, Mediatech Inc, VA). Bacteria were cultured in SS media overnight and were then sub-cultured in SS media to an optical density of ~0.5 at 600 nm. Inocula were confirmed by plating serial dilutions. For survival curves, groups of four mice were inoculated with the indicated dose, and the percent survival was monitored over a 30-day period. Mice with lethal bordetellosis, indicated by ruffled fur, labored breathing, and diminished responsiveness, were euthanized to alleviate unnecessary suffering [24]. To enumerate the number of bacteria in respiratory organs, groups of three to four mice were sacrificed at the indicated time points and bacterial numbers in the lungs and tracheas were quantified by plating dilutions of tissue homogenates on BG plates with appropriate antibiotics, following incubation at 37°C for 2 days. The mean ± the standard error was determined for each group. The statistical significance between different groups was calculated by Student's two-tailed t-test. A significance level was set at P values of ≤0.05. All animal experiments were repeated at least three times with similar results. Murine survival percentage was analyzed with the Log-Rank (Mantel-Cox) test. All mice were maintained in UCLA animal research facilities according to National Institutes of Health and University of California Institutional Animal Care Committee guidelines. Animals were housed under specific pathogen-free conditions with free access to food and water. All experiments were approved by the UCLA Chancellor's Animal Research Committee.

Histopathological analysis

Lungs were inflated with 10% neutral buffered formalin at the time of necropsy. Following fixation, tissue samples were embedded in paraffin, sectioned at 5 μm, and stained with hematoxylin-eosin, Giemsa, and Warthin-Starry for light microscopic examination at the Translational Pathology Core Laboratory of UCLA. Sections were scored for pathology by a veterinarian with training and experience in rodent pathology who was blinded to experimental treatment. The degree of inflammation was assigned an arbitrary score of 0 (normal = no inflammation), 1 (minimal = perivascular, peribronchial, or patchy interstitial inflammation involving less than 10% of lung volume), 2 (mild = perivascular, peribronchial, or patchy interstitial inflammation involving 10-20% of lung volume), 3 (moderate = perivascular, peribronchial, patchy interstitial, or diffuse inflammation involving 20-50% of lung volume), and 4 (severe = diffuse inflammation involving more than 50% of lung volume).

In vitroadherence assays

Human lung epithelial (A549) cells and Human cervical epithelial (HeLa) cells were grown in F-12 K and DMEM medium, containing 10% fetal calf serum on cover slips in standard 12-well tissue culture plates, respectively. Bacteria in their mid-log phase were added to cell monolayers at a MOI of 200 as previously described [25]. The plates were spun at 200 × g for 5 min and then incubated for 15 min at 37°C. The cells were then washed six times with Hanks' balanced salts solution, fixed with methanol, stained with Giemsa stain (Polyscience, Warrington, PA) and visualized by light microscopy. Adherence was quantified by counting the total number of bacteria per eukaryotic cell in at least three microscopic fields from two separate experiments.

Trypsin digestion of polypeptides for mass spectrometry

For secretome analysis by mass spectrometry, bacteria were cultured in SS media overnight and were then sub-cultured in SS media to an optical density at 600 nm of ~1.0. A 5 ml aliquot was removed and centrifuged at 10,000 x g at 4°C for 10 min to remove bacterial cells. The resulting supernatant, containing proteins secreted into the culture medium, was filtered through a 0.2 μm membrane to remove contaminating bacterial cells. The filtered supernatants were then desalted and concentrated using a centrifugal filter device (Amicon Ultra-3 K, Millipore) into ~300 μl of 50 mM ammonium bicarbonate buffer. The samples were reduced by incubation in 10 mM dithiotreitol (DTT) in 50 mM ammonium bicarbonate at 37°C for 1 h. They were then alkylated by adding 55 mM iodoacetamide in 50 mM ammonium bicarbonate and incubated at 37°C in dark for 1 h. Finally, the samples were digested at 37°C overnight with addition of 75 ng trypsin (EC 3.4.21.4, Promega) in 50 mM ammonium bicarbonate. For in-gel trypsin digestion of polypeptides, a previously described method was used [26].

Nano-liquid chromatography with tandem mass spectrometry (nLC-MSMS)

nLC-MS/MS with Collision Induced Dissociation (CID) was performed on a linear trap quadrupole fourier transform (LTQ FT, Thermo Fisher, Waltham, MA) integrated with an Eksigent nano-LC. A prepacked reverse-phase column (Microtech Scientific C18 with a dimension of 100 μm x 3.5 cm) containing resin (Biobasic C18, 5-μm particle size, 300-Å pore size, Microtech Scientific, Fontana, CA) was used for peptide chromatography and subsequent CID analyses. ESI conditions using the nano-spray source (Thermo Fisher) for the LTQ-FT were set as follows: capillary temperature of 220°C, tube lens 110 V, and a spray voltage of 2.5 kV. The flow rate for reverse-phase chromatography was 5 μl/min for loading and 300 nl/min for the analytical separation (buffer A: 0.1% formic acid, 1% acetonitrile; buffer B: 0.1% formic acid, 100% acetonitrile). Peptides were resolved by the following gradient: 2–60% buffer B over 40 min, then increased to 80% buffer B over 10 min and then returned to 0% buffer B for equilibration of 10 min. The LTQ FT was operated in data-dependent mode with a full precursor scan at high-resolution (100000 at m/z 400) and six MSMS experiments at low resolution on the linear trap while the full scan was completed. For CID the intensity threshold was set to 5000, where mass range was 350–2000. Spectra were searched using Mascot software (Matrix Science, UK) in which results with p < 0.05 (95% confidence interval) were considered significant and indicating identity. The data was also analyzed through Sequest database search algorithm implemented in Discoverer software (Thermo Fisher, Waltham, MA).

Identification of the core, non-core, and pan-genome of Bordetella

"Core" regions were defined as genome sequences that were present in all 11 Bordetella genomes, while "non-core" regions were defined as genome sequences that are not present in all genomes. RB50 was used as the reference genome. For each of the other 10 sequences, genomes were mapped to the reference genome using Nucmer [27]. All 10 “.coords” output files from the Nucmer program were analyzed to identify overlap regions based on RB50 coordinates using a Perl script. Finally, “core” sequences were extracted based on the genome sequence of RB50 with the coordinates calculated above. Unshared regions were then added to the reference genome to make a “revised” reference genome, which contained the original sequence plus unshared sequences. This process was repeated until all of the genomes were compared to include all unshared sequences included in the pan-genome. The core region was subtracted from the pan-genome of all the 11 genomes, and the remaining regions were identified as non-core regions.

Hierarchical clustering using Cluster and Java Tree View

844 non-core fragments with more than 1000 bp were identified. An 844 row x 11 column matrix, in which 1 means "present" while 0 means "absent" for each non-core region, was entered to the Cluster program (http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm#ctv) [28]. Average linkage was used for clustering. The Java Tree View program [28] was used to show the clustering result.

Hypercytotoxicity of complex IV isolates in vitro

Cytotoxicity against a broad range of cell types is a hallmark of B. bronchiseptica infection in vitro[11, 12, 14, 16, 23]. To measure relative levels of cytotoxicity, human epithelial cells (HeLa), murine monocyte-macrophage derived cells (J774A.1), or human pneumocyte-derived cells (A549) were infected with an array of complex I or complex IV B. bronchiseptica isolates (Figure 1A-C). These strains represent different multilocus sequence types (STs), and they were isolated from both human and non-human hosts (Table 1). Lactate dehydrogenase (LDH) release was used as a surrogate marker for cell death, and RB50, an extensively characterized complex I rabbit isolate classified as ST12, was used as a positive control for cytotoxicity [20]. An isogenic RB50 derivative with a deletion in bscN, which encodes the ATPase required for T3SS activity [15], served as a negative control.

For HeLa (Figure 1A) and J774A.1 cells (Figure 1B), single time point assays showed a distinct trend, in which complex IV strains displayed higher levels of cytotoxicity than complex I isolates. For A549 cells the results were more dramatic (Figure 1C). Unlike other cell types previously examined [11, 16, 29], A549 cells are nearly resistant to cell death mediated by the RB50 T3SS (see RB50 vs. RB50ΔbscN; Figure 1C). Similarly, other complex I strains displayed little or no cytotoxicity against these cells. In striking contrast, incubation with complex IV isolates resulted in significant levels of cell death (p < 0.0001; Figure 1C). For A549 cells, strains D444 (ST15), D445 (ST17), D446 (ST3) and Bbr77 (ST18) were 10- to 15-fold more cytotoxic than RB50. Parallel assays measuring bacterial attachment to A549 cells did not detect significant differences between complex I and complex IV isolates, indicating that relative levels of adherence are not responsible for the observed differences in cytotoxicity (Additional file 1 Table S1).

Kinetic studies were performed next to increase the resolution of the analysis. We examined relative levels of cytotoxicity conferred by five complex IV strains towards HeLa, J774A.1 or A549 cells as a function of time, using RB50 as a complex I representative strain and RB50ΔbscN as a negative control. Following infection at an MOI of 50, cultures were sampled over a 4 h time period for measurements of LDH release. Using HeLa cells, which are exceptionally sensitive to T3SS-mediated killing, only minor differences were detected between strains (Figure 2A). Although two of the complex IV strains, Bbr77 and D444, displayed slightly elevated cytotoxicity at intermediate time points, all strains reached maximum lysis by the end of the 4 h time course. For J774 cells, differences between complex IV strains and RB50 were apparent throughout the experiment, with Bbr77 and D444 showing the highest levels of activity (Figure 2B). As expected, the most dramatic differences were seen with A549 cells (Figure 2C). Most complex IV strains displayed a marked hypercytotoxicity phenotype compared to RB50, with the exception of Bbr69 which had an intermediate phenotype. Interestingly, Bbr69 is a dog isolate whereas all of the other complex IV strains tested were cultured from human infections.

Roles of the bsc T3SS and the BteA effector in hypercytotoxicity by complex IV B. bronchisepticaisolates

To examine the hypercytotoxicity phenotype in detail, two representative highly toxic complex IV strains of human origin, D445 (ST17) and Bbr77 (ST18), were chosen for further analysis. To measure the contribution of the bsc T3SS, nonpolar in-frame deletions were introduced into the bscN loci of D445 and Bbr77. As shown in Figure 3AbscN mutations eliminated in vitro cytotoxicity against all three cell types, demonstrating an essential role for type III secretion. We next examined the involvement of the BteA effector in hypercytotoxicity. Previous studies have shown that BteA is essential for T3SS-mediated cell death induced by RB50, and it is sufficient for cytotoxicity when expressed in mammalian cells [11]. For both complex IV strains, bteA deletion mutations had a similar effect as ΔbscN mutations and abrogated cytotoxicity (Figure 3A).

The BteA proteins expressed by Bbr77 and D445 are identical except for the absence of two amino acids at the extreme carboxyl end of D445 BteA (Figure 3B). In contrast, when compared to RB50 BteA, the complex IV effectors from Bbr77 and D445 differ at 22 or 24 positions, respectively (Figure 3B). Interestingly, BteA sequences from complex IV strains were more closely related to BteA in B. parapertussis _hu Bpp12282 than to homologs in B. bronchiseptica RB50 or B. pertussis Tohama I. To determine whether BteA polymorphisms are responsible for differences in cytotoxicity phenotypes, bteA deletion derivatives of all three strains were complemented with the RB50 bteA allele on a medium copy vector (Figure 3A) [11]. In each case, complemented levels of cytotoxicity were similar to those of the wild type isolate. Most importantly, complemented ΔbteA derivatives of strains D445 and Bbr77 regained cytotoxicity against A549 cells, whereas RB50 ΔbteA/pbteA remained non-cytotoxic against this cell line. Taken together, these results demonstrate that the bsc T3SS and the BteA effector are essential for cytotoxicity by D445 and Bbr77. The hypercytotoxicity phenotypes of the complex IV isolates, however, are not due to polymorphisms in BteA. This is consistent with the conserved nature of this effector, both within and between Bordetella species [11]. Differential regulation of T3SS activity, the presence of novel secretion substrates, or alterations in accessory factors could account for phenotypic differences between strains (see Discussion).

T3SS secretome analysis

Virulence of complex IV strains during respiratory infections

To determine if relative levels of cytotoxicity measured in vitro correlate with virulence in vivo, we used a murine respiratory intranasal challenge model [24]. Groups of 4–6 week old female specific-pathogen-free C57BL/6NCr mice were intranasally infected with 5 x 10⁵ CFU. At this dose, RB50 establishes nonlethal respiratory infections that generally peak around day 10 post-inoculation and are gradually cleared from the lower respiratory tract, while persisting in the nasal cavity [33].As shown in Figure 4A, complex IV strains segregated into two groups. The first caused lethal infections in some (D444, Bbr77) or all (D445) of the infected animals. The second group (D446, Bbr69) caused nonlethal infections similar to RB50.

In the experiment shown in Figure 4B, animals were intranasally inoculated with 5 x 10⁵ CFU of RB50 or the two most virulent complex IV isolates, D445 and Bbr77, and sacrificed three days later. Both complex IV isolates were present in lungs at levels that were 10 to 30-fold higher than RB50 (p < 0.001). Histopathological examination of lung tissue from mice infected with D445 or Bbr77 showed severe and widespread inflammation, affecting nearly the entire volume of the lung for D445 and up to 40% of the tissue for Bbr77 (Figure 4C & D). Extensive migration of lymphocytes, macrophages, and neutrophils resulted in severe consolidation of large areas of lung parenchyma. Alveolar and interstitial edema as well as extensive perivascular and peribronchiolar inflammation were also observed. In contrast, lungs from animals infected with RB50 displayed only mild inflammation that covered less than 25% of the total lung volume.

We also examined the relative roles of the bsc T3SS and the BteA effector in the in vivo virulence phenotypes of D445 and Bbr77. As shown in Figure 4A, deletions in bscN or bteA abrogated lethality following infection by either strain. Consistent with these observations, ΔbscN and ΔbteA mutants also showed significantly decreased numbers of bacteria in the lungs at day 3 post infection (Figure 4B) and a corresponding decrease in histopathology (Figure 4C). These results demonstrate that in comparison to the prototype complex I strain RB50, D445 and Bbr77 are more virulent in mice following respiratory infection, and hypervirulence is dependent on type III secretion and BteA.

Comparative whole-genome analysis of complex I and complex IV B. bronchisepticastrains

To determine if hypervirulent complex IV B. bronchiseptica strains share common genomic regions that might be responsible for the phenotypes reported here, we obtained whole genomic sequences of D444 (MO149), Bbr77, and D445 using next-generation high throughput sequencing. We included in our analysis the genomic sequences of B. bronchiseptica strains BBE001 and 253 (complex I human isolates) [34, 35], BBF559 (complex IV human isolate) [34], and RB50 [20]; B. pertussis strains Tohama I and 18323 [36]; and B. parapertussis strains 12822 and Bpp5 (human and ovine isolates, respectively) [37, 38]. The B. bronchiseptica sequences were in various stages of assembly at the time of analysis (Table 3). Hierarchical clustering of virtual comparative genomic hybridization data supports previous MLST assignments of phylogenic relationships between Bordetella strains [10], as isolates from each complex are clustered together (Figure 5). Genome alignments reveal that these strains share approximately 2.5 Mb of "core" genome sequence.

Strain	Size (Mb)	ST (complex)	Contigs/Scaffold
RB50	5.4	12 (I)	1
253	5.3	27 (I)	4
D444	5.1	15 (IV)	1
D445	5.2	17 (IV)	11
Bbr77	5.2	8 (IV)	16
BBE001	5.1	11 (I)	175
BBF579	4.9 (+IS481)	novel (IV)	319

We next carried out a comparative analysis of the non-core genome to identify potential loci shared only by complex IV strains. Despite sequences that are shared by more than one complex IV isolate, we did not identify complex IV genomic sequence(s) that uniquely differentiate complex IV from complex I strains. Strains D445, Bbr77 and D444 do, however, contain clusters of shared genes that are not present in other Bordetella genomes (Figure 5B, yellow boxes). Although these loci are missing in BBF579, the virulence properties of this isolate has not been reported, raising the possibility that one or more of these loci may contribute to hypervirulence by a subset of complex IV strains. Blastn analysis of overlap regions revealed a diverse set of genes involved mainly in signal transduction, metabolism, adhesin/autotransporter expression and type IV secretion of unknown substrates (Figure 5B).

One locus of potential interest, found in two out of four sequenced complex IV isolates (Bbr77 and D444) but none of the other Bordetella genomes, is predicted to encode homologs of the QseBC two-component regulatory system found in numerous bacterial pathogens [39]. In enterohemorrhagic E. coli (EHEC) and Salmonella sp., QseBC has been shown to sense host stress hormones (epinephrine and norepinephrine) and regulate virulence gene expression [40–44]. The qseBC open reading frames from Bbr77 and D444 are identical, and their predicted products share 47% amino acid identity and 63% similarity with EHEC QseB, and 34% identity and 51% similarity with EHEC QseC, respectively. Using a PCR-based assay, we screened for the presence of qseBC in a larger collection of B. bronchiseptica isolates. As shown in Figure 5C, this locus is present in 7 out of 9 complex IV isolates, but only 1 out of 10 complex I isolates. Sequence analysis of PCR amplicons revealed high levels of nucleotide identity (> 97%) between B. bronchisepticaqseBC alleles. Although highly enriched in complex IV strains, qseBC is unlikely to represent a single, conserved pathway for hypervirulence since it is absent from strain D445. Nonetheless, the potential role of QseBC in Bordetella-host interactions warrants further study. In addition to examining gross genomic differences, we also analyzed polymorphisms in virulence loci. Nearly all of the virulence genes shared a high degree of homology (Additional file 3 Table S2). The bsc T3SS locus, the btr genes involved in T3SS regulation, as well as their upstream promoter regions had greater than 97% sequence conservation between RB50 and complex-IV strains. Additionally, our analysis confirms the absence of ptx/ptl loci and divergence in tcfA and prn genes in sequenced complex-IV isolates as previously described by Diavatopoulos et al. [10].