Adaptation of Bordetella pertussis to the Respiratory Tract

Abstract

There is a lack of insight into the basic mechanisms by which Bordetella pertussis adapts to the local host environment during infection. We analyzed B. pertussis gene expression in the upper and lower airways of mice and compared this to SO₄-induced in vitro Bvg-regulated gene transcription. Approximately 30% of all genes were differentially expressed between in vitro and in vivo conditions. This included several novel potential vaccine antigens that were exclusively expressed in vivo. Significant differences in expression profile and metabolic pathways were identified between the upper versus the lower airways, suggesting distinct antigenic profiles. We found high-level expression of several Bvg-repressed genes during infection, and mouse vaccination experiments using purified protein fractions from both Bvg^- and Bvg⁺ cultures demonstrated protection against intranasal B. pertussis challenge. This study provides novel insights into the in vivo adaptation of B. pertussis and may facilitate the improvement of pertussis vaccines.

Infection with Bordetella pertussis causes whooping cough (or pertussis) in susceptible humans. Although progression to the lungs can result in severe symptoms, infection is typically restricted to the upper respiratory tract. Severe morbidity and mortality due to pertussis has nearly been eradicated in children following the introduction of whole-cell pertussis vaccines. In most high-income countries, whole-cell pertussis vaccines have been replaced with acellular pertussis vaccines, which have a better safety profile than whole-cell pertussis vaccines but also provide less durable protection [1–3].

The resurgence of pertussis has stimulated research into B. pertussis vaccine antigens. Many known Bordetella virulence factors are controlled by the BvgASR system, which transduces environmental cues, such as temperature or concentrations of nicotinic acid and sulfate [4], into distinct gene expression and virulence phenotypes: Bvg⁺, Bvgⁱ, and Bvg^-. While B. pertussis locked in the Bvg⁺ and Bvgⁱ phases can infect mice, Bvg^- phase–locked mutants are attenuated in virulence [4–7].

Knowledge about B. pertussis gene expression and regulation is mainly derived from in vitro studies. All current acellular pertussis vaccine antigens (ie, pertactin, pertussis toxin, filamentous hemagglutinin, and fimbriae 2/3) are optimally expressed in the Bvg⁺ phase. Similarly, whole-cell pertussis vaccines are also produced from B. pertussis Bvg⁺ cultures. Although Bvg-associated genes are clearly important for infection, it remains largely unclear how local in vivo conditions involving changes in nutritional status, quorum sensing, temperature, pH, and immune factors influence global B. pertussis gene networks. To gain a better understanding of this process, this study provides a comprehensive analysis of B. pertussis gene expression in the mouse respiratory tract as compared to Bvg-regulated gene expression in vitro.

METHODS

Ethics Statement

Animal experiments were approved by the Radboud University Medical Center Committee for Animal Ethics and were conducted according to Dutch legislation.

Bacterial Strains and Culture Conditions

B. pertussis strains B1917 (pertussis toxin promoter 3 [ptxP3]) and B1920 (PtxP1) [5, 8] were grown under Bvg⁺, Bvgⁱ, and Bvg^- conditions in chemically defined THIJS medium as described previously [5]. Bacteria were harvested at mid-log phase (OD₆₂₀, 0.5–0.6) for RNA and protein isolation.

Protein Isolation

Membrane proteins were isolated using the ReadyPrep Protein Extraction Kit Membrane I (Bio-Rad Laboratories, Hercules, CA). Membrane and membrane-associated proteins were purified using the ReadyPrep 2-D clean-up Kit (Bio-Rad Laboratories) and dissolved in 8 M urea and 10 mM Tris-HCl (pH 8.0).

Vaccination

Membrane proteins were mixed 1:1 with 1.3% alum adjuvant (Alhydrogel; Sigma, Saint Louis, MO), and 1 µg of protein was used to immunize naive, female, 6–8-week-old BALB/c mice (Charles River, Wilmington, MA) subcutaneously on days 0 and 14. As controls, mice were immunized with Alhydrogel mixed with phosphate-buffered saline (PBS; 1:1) or 1/50th human dose of the acellular pertussis vaccine Infanrix (GSK, London, UK). On day 35, mice were anesthetized and infected intranasally with 2 × 10⁷ colony-forming units (CFU) of B. pertussis strain B1917 in 40 µL. On day 38 and 42, the bacterial loads in the nasopharynx and lungs were determined as described previously [9]. Serial dilutions of nasal lavage (NL) specimens and lung homogenates were plated onto Bordet-Gengou agar plates and incubated at 37°C for 3–4 days, followed by determination of the number of CFU.

In Vivo Transcriptional Analysis

Naive, female, 6–8-week-old BALB/c mice were infected intranasally with B. pertussis strain B1917 or B1920 as described above. Mice were euthanized on days 3 and 7, and bacteria were recovered by flushing the nasopharynx (ie, by NL) or lungs (ie, by bronchoalveolar lavage [BAL]) with PBS. The bacterial load was determined, and the remaining sample was mixed immediately with 2 volumes of RNA Protect Bacteria Reagent (Qiagen, Hilden, Germany). To control for false-positive signals due to potential cross-hybridization of mouse-derived complementary DNA (cDNA) to the microarray, 2 mice were challenged intranasally with 10 µg of Escherichia coli lipopolysaccharide (LPS) in 40 µL of PBS. LPS treatment recruits immune cells to the bronchoalveolar space, with the cell count peaking on day 2 [10]. At this time point, mice were euthanized, and NL and BAL samples were collected and processed as described above.

Total RNA was extracted using the RNeasy Mini kit (Qiagen), and contaminating genomic DNA was removed by DNase treatment (DNAfree; Ambion, Foster City, CA). Bacterial RNA was enriched and amplified using the MicrobEnrich (Ambion) and SensationPlus FFPE Amplification (Affymetrix, Santa Clara, CA) kits, respectively. RNA was reverse transcribed using the SuperScript One-Cycle cDNA Kit (Invitrogen, Carlsbad, CA), and 5 µg of cDNA was labeled and hybridized to custom-designed microarrays as described previously [5].

Microarray results were validated by reverse-transcription quantitative polymerase chain reaction (RT-qPCR) analysis, as described previously [11]. Relative expression levels were determined using the recA (BP2546) household gene. PCR results are expressed as follows: ΔC_t − ΔC_tBvg⁺.

Immunoblot Analysis

Bvg⁺ and Bvg^- bacteria were lysed by a microfluidizer and used for immunoblot analysis [12]. Pooled sera were then added from mice infected with B1917 (14 days after infection). As detection antibody, rabbit-anti-mouse immunoglobulin G horseradish peroxidase (Dako, Santa Clara, CA) was used with ECL Western blotting substrate (GE Healthcare, Chicago, IL).

Because the experimental vaccine formulations potentially contained B. pertussis lipooligosaccharide, immunoblot was used to evaluate whether vaccination had induced antibodies against B1917 lipooligosaccharide purified by the hot phenol/water method [13]. As a positive control, monoclonal antibody clone 88F3 was used [14].

Statistical Analysis

In vivo expression data were combined with previously published transcriptome data on SO₄ modulation [5]. Normalized expression data were analyzed by ArrayStar (DNAStar, Madison, WI), using the robust multiarray analysis algorithm for background correction and quantile normalization [15]. Based on investigation of the LPS-treated mice, 34 genes were excluded from further analysis. Gene expression values were log₂ transformed, and differentially expressed genes were identified on the basis of log₂ fold changes. P values were calculated with the limma package [16], followed by use of the Benjamini and Hochberg method to correct for multiple testing [17]. For all comparisons, samples from both strains were combined for the NL specimens, as well as for the BAL specimens. Differential gene expression between in vivo and in vitro conditions was defined as a log fold change in gene expression of >2 between the NL/BAL specimens versus any of the in vitro conditions and a corrected P value of < .05 from a 5-way comparison combining both time points. Gene expression between NL and BAL samples was compared for both time points in a 4-way comparison. Genes were considered differentially expressed if the log fold change in gene expression was >1 and the corrected P value was < .05. Enrichment for specific KEGG pathways were assessed using STRING [18]. Hierarchical clustering and principal component analysis of samples was performed on log-transformed gene expression values of differentially expressed genes, using k-means and Pearson correlations, respectively, yielded by Genesis [19].

RESULTS

Transcriptional Profiling of B. pertussis During Respiratory Infection

To examine how B. pertussis adapts to local conditions, we studied bacterial gene expression during infection. Naive adult BALB/c mice were infected intranasally with B. pertussis strain B1917 and B1920, 2 clinical isolates derived from the P3 and P1 lineages that dominated globally after and before the 1990s, respectively [5, 8, 20, 21]. Three and 7 days later, bacteria were recovered from the upper respiratory tract (via NL specimens) and the lower respiratory tract (via BAL specimens) for transcriptional analysis (Figure 1A). At least 3 independent biological replicates were obtained for each strain and anatomical location, except for B1920 recovered from the nasal cavity at day 3 (2 replicates) and B1917 recovered from the nasal cavity at day 7 (0 replicates), owing to low RNA yield. Previously obtained gene expression data on in vitro modulation by sulfate [5] was used to study gene expression differences between in vivo and in vitro conditions (Figure 1A). Microarray results were validated by qPCR analysis for a selection of genes (Supplementary Table 1).

Principal component analysis (PCA) was used to explore differences in the B. pertussis transcriptome, including all genes regulated by at least log 2-fold by in vitro sulfate modulation (defined as Bvg⁺, Bvgⁱ, and Bvg^-). Although we previously found differences in in vitro regulation between B1917 and B1920 [5], the in vivo gene signatures of these strains clustered closely together. Consequently, we pooled the expression data of both strains to increase the statistical power of our analysis. The PCA showed very distinct dimensional segregation between in vivo and in vitro conditions (Figure 1B), indicating large differences in Bvg-dependent gene expression. As expected, in vitro Bvg^- phase bacteria showed a distinct expression pattern as compared to Bvg⁺ and Bvgⁱ phase bacteria (Figure 1B), whereas Bvg⁺ and Bvgⁱ conditions clustered closely together.

To study gene expression changes during infection in an unbiased manner, we compared the replicas of both B. pertussis strains (P1 and P3), all conditions (Bvg⁺, Bvgⁱ, and Bvg^- phases and NL and BAL specimens), and both time points (days 3 and 7). This yielded 1622 differentially expressed genes (4-fold cutoff), visualized by hierarchical clustering. In line with PCA analysis, hierarchical clustering showed strong segregation between in vivo and in vitro samples (Figure 1C and Supplementary Tables 2–4). Fifteen expression clusters were distinguished on the basis of k-means clustering [19]. Figure 1D shows the gene expression pattern from 6 of these clusters (clusters 4, 5, 6, 13, 14, and 15). The clusters and genes of interest shown were selected on the basis of a clear segregation in mean log₂ signal intensity (SI) between ≥2 growth conditions and/or their potential or known relevance for pathogenesis.

Twenty-one strongly induced Bvg-repressed genes (average log₂ SI, >11), primarily encoding hypothetical, surface-exposed, and secreted proteins, were moderately expressed in vivo and weakly expressed during the Bvg⁺/Bvgⁱ phase (average log₂ SI, <10; cluster 4 in Figure 1D). In contrast, several Bvg-repressed genes (eg, the kps locus involved in polysaccharide capsule formation) were not expressed under all other tested conditions (average log₂ SI, >10 vs <8; cluster 5 in Figure 1D).

Several virulence-associated, Bvg-activated genes (average log₂ SI, >10) were relatively weakly expressed during infection (average log₂ SI, <9; cluster 6 in Figure 1D). This includes 23 genes of the type III secretion system (T3SS), fimbrial genes (fimBC), Bordetella intermediate protein A (bipA), autotransporter subtilisin-like protease (sphB1), virulence regulatory genes (bvgR and bvgS), adenylate cyclase (cyaA locus), and complement resistance protein (brkA). Other known virulence genes, such as pertactin (prn), filamentous hemagglutinin (fhaB), pertussis toxin subunits (ptxDE), tracheal colonization factor (tcfA), vag8, and the T3SS toxin-encoding bteA, were found to be expressed under all conditions except Bvg^- (average log₂ SI, >10 and <8, respectively; cluster 13 in Figure 1D).

Finally, 166 genes showed high expression during infection (average log₂ SI, >0), with only weak expression in vitro (average log₂ SI, <9). This includes genes involved in iron acquisition (alc locus) and flagellar biosynthesis (fli and flg loci), as well as components of the T3SS (bsc locus) and fimbrial protein X (fimX; cluster 14 in Figure 1D). A similar but less pronounced pattern (average log₂ SI, <7 in vitro vs >9 in vivo) was observed for several genes encoding proteins that directly interact with the host environment (cluster 15 in Figure 1D). Examples include iron acquisition/use systems (bhu, hur, bfe, and bfr loci) and the sulfate transport system (cys operon). Autotransporter sphB3, a 2-component sensor system (kdp), and flagellar biosynthesis genes also belonged to this cluster.

To study gene networks, Bvg⁺, Bvgⁱ, and Bvg^- phases were grouped together and compared to NL and BAL specimens together, combining both strains and time points. Using a 4-fold cutoff, 465 genes were upregulated and 605 genes downregulated during infection versus in vitro growth. Functional category analysis showed that infection was associated with significant downregulation of several central metabolic pathways, protein production, and LPS biosynthesis. Concomitantly, chemotaxis and flagellar assembly genes, ABC transport systems, and amino acid and fatty acid metabolic pathways were upregulated (Figure 2).

Distinct Gene Signatures in the Upper and Lower Respiratory Tracts

Although PCA of the expression values of all in vitro Bvg-regulated genes (2-fold cutoff) could not separate NL from BAL samples (Figure 1B), hierarchical clustering of all 1622 differentially expressed genes (4-fold cutoff) resulted in distinct clustering (Figure 1C). Gene expression between NL and BAL specimens was compared 3 and 7 days after infection, using a 4-way comparison in which the 2 strains were grouped for each condition.

In the NL specimens, only the phage-related gene BP3372 was differently expressed over time (Supplementary Table 5). Conversely, 28 and 53 genes were more highly expressed in the BAL specimens on days 3 and 7, respectively. These included 12 ribosomal genes (24.5%) on day 7 (P < .001), suggesting increased metabolic activity at later stages of infection. Of note, the complement evasion protein Vag8 was also more highly expressed on day 7 (Supplementary Table 5).

Spatial analysis identified 147 and 199 genes with increased expression on day 3 in BAL and NL specimens, respectively. Forty-seven genes upregulated in BAL specimens were significantly enriched for metabolic pathways (P < .001). In NL specimens, 33 genes were associated with ribosomal activity (P < .001), 14 were associated with oxidative phosphorylation (P < .001), 11 were associated with the bacterial secretion system (P < .001), and 5 were associated with protein export (P = .037; Supplementary Table 5). In addition, virulence-associated genes, such as virulence regulatory genes (bvgA and risA), vag8, fimD, dnt, T3SS genes, and pertussis toxin structural and transport genes (ptxC/E and ptlA/D), were upregulated in NL specimens (Table 1). On day 7, 229 and 186 genes showed significantly increased expression in the lower and upper respiratory tracts, respectively. Twenty-three genes upregulated in BAL specimens were significantly enriched in microbial metabolism in diverse environments (P = .038). No pathways were significantly enriched in NL specimens (Supplementary Table 5).

Protective Antigens Are Expressed Under Both Bvg⁺ and Bvg^- Conditions

Gene expression analysis suggested that several Bvg-repressed genes were expressed during infection. To evaluate whether Bvg-repressed antigens may be immunogenic during infection, Bvg^- and Bvg⁺ lysates were analyzed by immunoblot, using serum from mice previously infected with B. pertussis. Infection-induced antibodies recognized several antigens that were either exclusively expressed or upregulated during the Bvg- phase (Figure 3A). Next, we studied whether vaccination with Bvg^- phase membrane proteins could protect against infection. Purified Bvg⁺ and Bvg^- membrane proteins were adjuvanted with aluminum hydroxide and used to vaccinate BALB/c mice. Vaccinated mice were challenged intranasally with strain B1917 two weeks later, and bacterial loads in lung and nose specimens were determined 3 and 7 days after infection.

Mice that received acellular pertussis vaccine showed a strong reduction (>220-fold) in lung CFU number at both time points, compared with mock-treated mice (P = .001; Figure 3B). Three days after challenge, mice immunized with Bvg⁺ or Bvg^- membrane proteins showed a significant and comparable approximately 10-fold reduction in lung CFU number, compared with the adjuvant-treated control (P = .001; Figure 3B). Seven days after infection, more-pronounced 107-fold (P = .005) and 16-fold (P = .001) reductions in lung bacterial loads were observed after vaccination with Bvg⁺ and Bvg^- proteins, respectively (Figure 3B). The difference between Bvg⁺ and Bvg^- treated groups almost reached statistical significance (P = .067). None of the tested formulations reduced the bacterial load in the upper respiratory tract (Figure 3C).

DISCUSSION

Although genetic and phenotypic changes in circulating B. pertussis strains are thought to contribute to the resurgence of whooping cough [22], few studies have investigated bacterial gene regulation during infection [23–26]. By studying changes in B. pertussis gene expression during respiratory infection, this study provides novel insights into the in vivo antigen repertoire and gene regulation.

Unsupervised clustering by PCA and hierarchical clustering showed that the in vivo transcriptional profile of B. pertussis is predominantly Bvg⁺/Bvgⁱ-like (Figure 1B, C), in line with previous studies showing that most virulence factors are highly expressed in the Bvg⁺ phase [4, 27]. In contrast to a previous study, in which we found small but significant strain-specific differences in susceptibility to SO₄-induced modulation [5], in vivo expression profiles of strains B1917 and B1920 clustered together. A potential explanation for this may be higher variability in gene expression between biological replicates from the respiratory tract. Alternatively, environmental signals during infection might overrule strain-specific differences in susceptibility to SO₄ modulation. Our observations on in vitro Bvg-regulated gene expression are similar to findings reported by Hot et al about differential modulation of virulence genes by SO₄ [28]. Some of the minor differences could be due to factors such differences in bacterial strains or culture conditions.

In addition to Bvg⁺/Bvgⁱ-associated genes, several Bvg-repressed genes were found to be expressed in vivo. Further experiments showed that vaccination with Bvg^- membrane proteins reduced the bacterial load in lungs (Figure 3B). Similar results were reported previously following vaccination with outer-membrane vesicles derived from Bvg^- phase B. pertussis [29]. Western blot analysis suggested that several antigens that induce an antibody response following infection are exclusively expressed or significantly upregulated under Bvg^- conditions.

The expression of Bvg-repressed genes contributes to the survival of Bordetella bronchiseptica outside the host. Our findings suggest that Bvg-repressed genes may also play a role during infection of the respiratory tract. The upper and lower respiratory tracts differ with respect to characteristics such as epithelial cellular composition, nutrients, microbiota, and temperature [30, 31], and it is therefore conceivable that distinct gene networks, including Bvg^- genes, facilitate survival in these microniches [32]. This is in line with a previous study that demonstrated that the Bvg-repressed polysaccharide capsule transport genes play an important role during infection [33]. A recent study identified positive selection for Bvg^- phase bacteria toward the later stages of infection in macaques, resulting in up to 50% of the total bacterial population becoming avirulent, as defined by the presence of an IS481 insertion in the BvgAS regulon [34]. Understanding the exact role of Bvg-repressed genes during infection is important because these genes may provide novel targets that can complement the current vaccines, which are all based on Bvg-activated genes.

Approximately 30% of the 3500 analyzed genes were classified as Bvg-independent genes, characterized by undetectable or very low expression in vitro (ie, Bvg regulated) and moderate-to-high levels in vivo. Notably, this included genes involved in interaction with the host environment, encoding proteins for iron, heme, and sulfate uptake. It is likely that these genes are important for survival in the host [35, 36].

To our surprise, we also found that flagellar biosynthesis genes were expressed during infection, even though B. pertussis is not motile and has no functional flagella [37]. It is possible that flagellar genes still play a role during B. pertussis infection, such as during biofilm formation [38]. Similarly, genes involved in chemotaxis may have secondary functions, such as roles in sensory transduction. The role of these genes during infection is currently being investigated.

Gene classification analysis suggests that bacteria are metabolically less active during infection as compared to growth under laboratory conditions. Differences in LPS biosynthesis, aminoacyl–transfer RNA biosynthesis, ribosomal activity, purine metabolism, and oxidative phosphorylation may be explained by faster bacterial replication in vitro as compared to in vivo. This is perhaps not a surprise, owing to the abundance of nutrients in culture medium versus the limited nutrient availability during colonization. Reduced metabolic activity during infection, paired with increased expression of flagellar genes, may point toward a biofilm-like mode of growth [11], as has been described for B. bronchiseptica [39]. Similar to other respiratory pathogens, B. pertussis upregulates the expression of many ABC transporters during infection, possibly in response to nutrient stress [40, 41]. In addition, metabolism of fatty acids, including propanoate, was upregulated in vivo, which was previously shown to be (partially) Bvg regulated and is thought to play a role in innate immune evasion [42, 43].

Interestingly, several genes that typically function in concert or encode multiprotein complexes (eg, fimBC vs fimAD and fhaC vs fhaB) did not cluster together (Figure 1D). This could be due to differences in hybridization to the array probes or to differential gene modulation by environmental factors, such as temperature. Previous studies also reported variable virulence gene expression, using a recombinase-based in vivo technology approach [26], which was important for successful intranasal infection in mice [44]. However, whole bacterial population gene expression values, presented here, may not accurately reflect variable expression within bacterial subpopulations (eg, planktonic bacteria vs bacteria residing in biofilms [45, 46] or inside cells [47]). Although in vivo gene expression is normally distributed, establishing an exact cutoff that distinguishes high-level from low-level expression is difficult. Thus, in vivo data should primarily be used for analysis of relative gene expression.

Neither the acellular vaccine nor any of the other experimental aluminum-adjuvanted formulations protected against colonization of the upper respiratory tract, an observation that closely resembles findings with acellular pertussis vaccines in baboons and that we previously described in mice [1, 11, 48, 49]. This could reflect induction of suboptimal local mucosal immunity. Another, not mutually exclusive explanation may be the relatively high bacterial load used to challenge mice, which may overwhelm vaccine-induced immunity in the upper airways. Another reason could be the differential expression of 415 genes between the upper and lower airways, suggesting that these 2 anatomical sites represent distinct niches [50]. Interestingly, early during respiratory tract infection, bacteria expressed several important virulence factors, including Vag8, dermonecrotic toxin, pertussis toxin, and Bsp2, at a higher level than lung-derived bacteria. All of these factors are known to interfere with the host immune system [27]. We hypothesize that differential gene regulation in the upper versus the lower respiratory tract influences local immune response modulation and may translate to distinct antigenic profiles.

This study provides important new insights into the mechanisms by which B. pertussis adapts to local conditions during infection. Findings from this study may be used to guide more-rational design of laboratory growth conditions that yield antigenic cocktails that are more similar to in vivo conditions. Together with novel adjuvants and/or administration routes, these strategies may provide new opportunities for the design of more effective vaccines. Future comparisons between B. pertussis gene expression in mice and in baboons or humans may strengthen the relevance of animal models in B. pertussis research.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. We thank Fred van Opzeeland, Elles Simonetti, and Saskia van Selm (Section Pediatric Infectious Diseases, Laboratory of Medical Immunology, Radboud Institute for Molecular Life Sciences, Radboud University Medical Centre, Nijmegen, The Netherlands) for assistance regarding the animal experiment.

Disclaimer. The funders had no role in study design, data collection, and interpretation or the decision to submit the work for publication.

Financial support. This work was supported by the Netherlands Organization of Scientific Research (grant 125020001 to D. d. G.).

Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Present affiliations: Global Center of Expertise in Microbiology, MSD Oss, Oss (D. d. G.), and Department of Infectious Diseases and Immunology, Faculty of Veterinary Medicine, Utrecht University, Utrecht (A. Z.), the Netherlands.

Presented in part: 10th International Symposium on Bordetella, Dublin, Republic of Ireland, 8–11 September 2013.

References

Author notes