Bordetella pertussis pathogenesis: current and future challenges

Key Points

• Bordetella pertussis, which is the causative agent of whooping cough, has re-emerged as a public health threat despite broad vaccine coverage.

• Re-emergence of this disease correlates with a transition from the use of whole-cell pertussis vaccines to acellular component vaccines, and the epidemiology of pertussis has shifted while the overall number of pertussis cases has increased.

• Bordetella spp. are paradigms for studying global virulence-control systems, bacterial toxins, type V secretion proteins and the evolution of pathogens.

• Bordetella spp. produce many virulence factors that contribute to pathogenesis, including toxins, adhesion factors, iron-acquisition systems and surface structures.

• Considering the fact that Bordetella pertussis is a strict human pathogen, the development of appropriate animal models has been challenging. However, a novel baboon infection and transmission model provides a promising new research avenue.

• Future goals include improving the efficacy of vaccines, protecting unvaccinated infants from infection and developing better treatment strategies for infants who become infected with B. pertussis. Achieving these goals requires a more thorough understanding of the mechanisms that are used by B. pertussis to establish infection and cause disease.

Abstract

Pertussis, also known as whooping cough, has recently re-emerged as a major public health threat despite high levels of vaccination against the aetiological agent Bordetella pertussis. In this Review, we describe the pathogenesis of this disease, with a focus on recent mechanistic insights into B. pertussis virulence-factor function. We also discuss the changing epidemiology of pertussis and the challenges facing vaccine development. Despite decades of research, many aspects of B. pertussis physiology and pathogenesis remain poorly understood. We highlight knowledge gaps that must be addressed to develop improved vaccines and therapeutic strategies.

Main

Pertussis is a highly contagious respiratory disease that is transmitted directly from human to human¹, probably via aerosolized respiratory droplets. The primary causative agent, Bordetella pertussis, is a Gram-negative bacterium that was first described by Bordet and Gengou in 1906 (Ref. 2). The closely related bacterium Bordetella parapertussis _Hu is responsible for a minority of cases (approximately 14%) and is less capable of causing severe disease³. Both B. pertussis and B. parapertussis_Hu are human-specific, and phylogenetic analyses indicate that they evolved from Bordetella bronchiseptica or a B. bronchiseptica-like ancestor^4,5 (Box 1). B. bronchiseptica infects a broad range of mammals, including humans, and although it can cause overt disease such as kennel cough in dogs and atrophic rhinitis in pigs, it typically colonizes its hosts chronically and asymptomatically⁶. Despite differences in host range and disease-causing propensity, B. pertussis, B. parapertussis_Hu and B. bronchiseptica are so closely related that they are now considered to be subspecies. Together, these organisms provide a paradigm for understanding bacterial adaptation to humans and the dichotomy between acute disease and chronic asymptomatic infection^4,5. Although other Bordetella species have been isolated from humans, they seem to be primarily opportunistic human pathogens.

In the pre-vaccine era, pertussis was widespread and mainly affected young children (1–9 years old)⁷. The classical manifestation of the disease in this age group, is characterized by three phases: the catarrhal phase, the paroxysmal phase and the convalescent phase⁸. Clinical observations, combined with results from studies using animal models (Box 2), suggest that classic pertussis is initiated by the adherence of bacteria to the ciliated respiratory epithelium in the nasopharynx and trachea^9,10. Adherent bacteria survive innate host defences, such as mucociliary clearance and the action of antimicrobial peptides, multiply locally and resist elimination by inflammatory cells. Symptoms during this catarrhal phase are similar to those of many upper respiratory infections, such as the common cold. After 1–2 weeks, the disease progresses to the paroxysmal phase, which can persist for 1–10 weeks and is characterized by periods of normal airway function that are interspersed with multiple severe spasmodic coughing fits, followed by characteristic inspiratory whoops, and often, emesis. The onset of adaptive immunity coincides with bacterial clearance but not with the cessation of symptoms, which typically decline gradually over another month but can persist for much longer (this is known as the convalescent phase)⁸. In infants (<1 year old), pertussis can take a more serious course, in which bacteria disseminate into the lungs and cause necrotizing bronchiolitis, intra-alveolar haemorrhage and fibrinous oedema¹⁰. In severe cases, extreme lymphocytosis occurs, which positively correlates with intractable pulmonary hypertension, respiratory failure and death¹⁰.

The introduction of whole-cell pertussis (wP) vaccines in the late 1940s resulted in a rapid reduction in both the incidence of pertussis and the number of deaths caused by the infection. However, the success of these vaccines was undermined by concerns about their safety (Box 3); thus, they were replaced with acellular pertussis (aP) vaccines in the late 1990s in many developed countries¹¹. Since then, the number of pertussis cases has increased and dramatic epidemic cycles have returned. In 2012, 48,277 cases of pertussis and 18 deaths associated with the infection were reported to the US Centers for Disease Control and Prevention (CDC); this represents the greatest burden of pertussis in the United States in 60 years, and similar outbreaks are occurring in other countries^12,13,14. However, the epidemiology of contemporary pertussis does not replicate that of the pre-vaccine era. Disease is now more common in infants and older children (aged 9–19) and, strikingly, older children who develop pertussis are often fully vaccinated according to current recommendations^15,16. Ominously, studies that have analysed the incidence of pertussis among children who were born and vaccinated during the transition to aP vaccines have found that the rate of infection is significantly higher among children who have been vaccinated with only aP vaccines compared with those who have been vaccinated with even a single dose of the wP vaccine¹⁷. To combat the rise of infections in this group, regulatory agencies have called for boosters to be administered earlier¹⁸. However, the benefit of boosting with aP vaccines is unclear as it is unknown whether the re-emergence of pertussis is due simply to waning immunity or whether it is due to fundamental differences in the nature of the immune response that is induced by aP vaccines compared with wP vaccines or with natural infection.

The increased incidence of disease among older children and adults is especially worrying because of the corresponding risk of transmission to incompletely or non-immunized infants¹. Compounding the problem, antibiotic treatment has minimal efficacy by the time most diagnoses are made, and severe cases can be unresponsive to standard therapies for respiratory distress¹⁰. Therefore, the re-emergence of pertussis as a global public health problem presents two challenges: first, the development of vaccines that have an acceptable safety profile, provide long-lasting immunity, reduce the burden of infection and prevent transmission; and second, the development of therapeutic agents and treatment strategies that reduce morbidity and mortality in vulnerable populations. Both goals require a better understanding of the aetiological agents of pertussis and the mechanisms by which they cause disease.

In this Review, we discuss our current understanding of the mechanisms that are used by Bordetella spp. to cause respiratory disease, focusing on the roles and functions of virulence factors in pathogenesis. For the interested reader, more specialized recent reviews on pertussis toxin biology^19,20, virulence gene regulation²¹, immunity^22,23 and vaccines²⁴, as well as an earlier comprehensive review on Bordetella spp. pathogenesis²⁵ are available.

Box 1: Bordetella phylogenomics

The Bordetella genus includes nine species, and the 'classic' or 'mammalian' bordetellae (which are Bordetella pertussis, Bordetella parapertussis and Bordetella bronchiseptica) are the most important species in the context of mammalian infection. Comparative analyses of their genomes have been informative on at least two levels. The first level involves comparisons between subspecies, which has revealed intriguing clues regarding the mechanisms of host adaptation and virulence evolution, and the second level involves studies of the population dynamics of B. pertussis in the context of recent outbreaks and the possibility of vaccine-driven evolution.

Early studies revealed limited genetic diversity between isolates, arguing that the classic bordetellae should be regarded as closely related subspecies and that B. pertussis and B. parapertussis_Hu evolved independently from different Bordetella bronchiseptica-like ancestors^172,173. This hypothesis suggested that B. bronchiseptica, which is a zoonotic generalist, has the propensity to give rise to host-restricted specialists that cause acute disease. Whole-genome sequence comparisons support this idea and show that the transition by B. pertussis and B. parapertussis_Hu to the human-restricted niche was accompanied by large-scale gene loss, an accumulation of pseudogenes and an expansion of insertion sequence elements^174,175. Based on this framework, a seminal study characterized a set of 132 Bordetella isolates from diverse mammalian hosts using a combination of multilocus sequence typing, comparative whole-genome microarray analysis and insertion sequence typing⁴. Four distinct complexes comprised of related strains were resolved, representing B. pertussis (complex II), B. parapertussis_Hu (complex III) and two distinct B. bronchiseptica complexes (complex I and complex IV; see the figure). Surprisingly, although B. bronchiseptica complex I isolates were primarily of animal origin (68%), 80% of B. bronchiseptica complex IV strains were from humans who had whooping cough-like symptoms. Notwithstanding the fact that human isolates were over-represented in the collection, the clustering of B. bronchiseptica strains that had been isolated from humans into a genetically related group was intriguing. A follow-up study that used whole-genome sequencing suggested that B. pertussis evolved from a B. bronchiseptica complex IV-like ancestor⁵. These observations illustrate the evolutionary dynamics of Bordetella subspecies and raise the question of whether complex IV B. bronchiseptica are on a path towards human adaptation.

At the other end of the evolutionary spectrum, the resurgence of pertussis has prompted many efforts to characterize polymorphisms in B. pertussis populations in a search for correlations with the introduction of aP vaccines. Although several shifts in allelic frequencies have been documented in different countries since the vaccines were introduced^176,177, a comprehensive analysis of isolates from the United States shows that most of these polymorphisms predate the wP-to-aP transition and are not the result of aP-driven selection¹⁷⁸. Nonetheless, currently circulating strains in the Americas, Asia, Australia and Europe have a single-nucleotide substitution in the pertussis toxin (PT) promoter (the ptxP3 allele) that confers a slight (∼1.6-fold) increase in PT production compared with the previously dominant allele¹⁷⁹. The ptxP3 mutant strains were present before aP vaccines, but, given the roles of PT in immunosuppression, it has been suggested that their expansion and apparent fixation reflects a selective advantage for establishing infection in partially immune hosts, regardless of the source of immunity¹⁸⁰. Cause and effect are unclear, as is the probability that such a small difference in PT production could have such a major effect. One of the difficulties in interpreting these studies is their reliance on allelic variability in small subsets of genes, which raises the possibility that 'hitchhiker' mutations located elsewhere in the genome are responsible for apparent clonal shifts¹⁷⁷. It is essential that these analyses continue, but they need to be based on whole-genome sequences and appropriate sampling of circulating and reference strains.

Box 2: Animal models

One of the greatest challenges in studying the pathogenesis of pertussis has been the development of animal models that accurately reflect human disease, as the most commonly used laboratory animals do not cough and are not natural hosts for Bordetella pertussis. Nonetheless, the availability of wild-type, knockout and transgenic mouse strains, as well as a plethora of mouse-specific reagents, have made mice attractive and commonly used model organisms. To establish infection in mice, high numbers of B. pertussis must be delivered directly to the lungs, where they multiply for the first week or so post-inoculation and are eventually cleared. Although mouse models do not mimic classical whooping cough, they have provided insights into the importance of several virulence factors, the roles of various host immune responses in controlling infection and the potential efficacy of vaccines.

Suckling pigs have been used to model infant pertussis, and intrapulmonary inoculation with B. pertussis results in low-grade fever, mild cough, hypoglycaemia, lymphocytosis, weight loss and pneumonia¹⁸¹. In this model, passively transferred immunity provides protection for newborn piglets upon challenge¹⁸², which supports the recently instituted policy of maternal immunization during pregnancy^183,184. Studies with infant and adult pigs have also shown that B. pertussis is susceptible to host antimicrobial peptides, whereas Bordetella bronchiseptica and Bordetella parapertussis_Hu are resistant to antimicrobial peptides¹⁸⁵.

The most recent and exciting advance in modelling human pertussis is the development of a baboon (Papio anubis) model¹⁸⁶. Delivery of B. pertussis to the nasopharynx of weanling baboons results in low-grade fever, paroxysmal coughing, lymphocytosis, robust production of pertussis toxin-specific antibodies and protection from subsequent challenge, all of which are manifestations of human pertussis¹⁸⁶. A unique advantage of this model is the ability to study transmission, which was shown to occur via contact and aerosols¹⁸⁷. Importantly, a recent study showed that aP vaccination provided protection against the development of disease symptoms but not against colonization or transmission¹⁶⁵, which suggests that non-symptomatic individuals might be capable of transmitting the infection to unprotected infants. Use of this model to investigate mechanisms of transmission and disease will be crucial for the development of new vaccines and therapeutic drugs.

Rodents, rabbits and swine are not natural hosts for B. pertussis or B. parapertussis_Hu; however, they are commonly infected with B. bronchiseptica in nature. This fact, in combination with the close phylogenetic relationships between these subspecies and the conservation of many known virulence factors, has prompted several groups to use B. bronchiseptica infection of natural hosts to study features of pathogenesis that are common to B. pertussis and B. parapertussis_Hu. Studying B. bronchiseptica pathogenesis is also of veterinary and clinical importance, as B. bronchiseptica is a common pathogen of domestic animals^188,189 and occasionally causes disease in humans¹⁹⁰. These models have revealed roles for several conserved virulence determinants and regulatory factors and have also shown that several factors are functionally interchangeable between B. pertussis and B. bronchiseptica^111,125,191, thereby validating the use of B. bronchiseptica as a model for studying B. pertussis virulence and regulatory function. Thus far, the results of these studies suggest that host specificity is determined by differences other than sequence polymorphisms in highly conserved virulence genes.

Box 3: Vaccination against pertussis

The introduction of whole cell pertussis (wP) vaccines in the 1940s resulted in dramatic decreases in the morbidity and mortality caused by pertussis, and by the early 1970s, pertussis was nearly eradicated in the developed world⁷. Attention then turned to the side effects that are associated with immunization. The National Childhood Encephalopathy Study (NCES), which was a prospective case-controlled study carried out in the U.K. in the late 1970s to evaluate acute neurological illnesses in children aged between 2 and 26 months, concluded that pertussis vaccination was associated with brain damage¹⁹²; however, this correlation was later shown to be unfounded. Lack of circulating disease was misinterpreted as being equivalent to a lack of risk of contracting the disease, and the NCES report (among others) led to a precipitous decline in vaccine coverage in many countries worldwide. The reported association with brain damage also effected a change in manufacturing policies: decreased vaccination compliance, combined with an increase in legal action associated with previous vaccinations, led many pharmaceutical companies to abandon the production of wP vaccines. However, the decrease in vaccine coverage almost immediately sparked the largest pertussis epidemics since the pre-vaccine era in many countries, which resulted in a substantial increase in infant mortality — a much more severe outcome than the alleged dangers of the wP vaccine. Eventually, public concern over the incidence of pertussis disease and the number of infant deaths increased, and vaccine coverage returned to high levels. Efforts to develop improved pertussis vaccines were rekindled, with the subsequent development, testing and deployment of acellular pertussis (aP) vaccines. Although supposedly safer and undoubtedly less reactogenic than the wP vaccine, we now know that aP vaccines are also less effective. The experience with vaccination against pertussis clearly highlights the effect that public perception and misconception can have on the implementation of public vaccination programs.

Bordetella virulence regulation

Several Bordetella virulence factors were identified and biochemically characterized before genetic tools became available, including pertussis toxin (PT), adenylate cyclase toxin (ACT), dermonecrotic toxin (DNT), filamentous haemagglutinin (FHA) and fimbriae (Fim). The first transposon mutagenesis screen of B. pertussis identified the genes that encode these factors, as well as a locus — now known as bvgAS — which encodes a two-component regulatory system that is required for their expression²⁶. Reasoning that BvgAS also activates the expression of genes that encode additional unknown Bordetella virulence factors, mutagenesis screens using Tn5lac and Tn5phoA were conducted^27,28. These, together with subsequent genome-wide analyses, showed that BvgAS controls hundreds of genes in response to changing environmental conditions, including genes that encode surface structures and secreted proteins involved in pathogenesis, factors required for survival outside the mammalian host, enzymes involved in cellular metabolism and physiology and additional regulatory systems^29,30.

The BvgAS phosphorelay. BvgA is a typical response-regulator protein with a receiver domain at its amino terminus and a DNA-binding helix–turn–helix domain at its carboxyl terminus²¹ (Fig. 1a). BvgS is a polydomain sensor kinase that contains two N-terminal venus flytrap (VFT) domains, which are located in the periplasm³¹. C-terminal to the VFT domains is a membrane-spanning region, followed by a cytoplasmically located PAS domain, a histidine kinase (HK) domain, a receiver domain and a histidine phosphotransferase (Hpt) domain. During growth in standard medium at 37 °C, BvgAS is active and uses ATP to phosphorylate a conserved histidine in the HK domain³². The phosphoryl group is subsequently relayed to an aspartate in the receiver domain, then to a histidine in the Hpt domain and, finally, to an aspartate in the receiver domain of the response regulator BvgA³². Phosphorylated BvgA is competent for dimerization and binds to specific DNA sequences to either activate or repress transcription^33,34. Although the signal (or signals) to which BvgS responds in nature is unknown, growth at a low temperature (∼25 °C) or in the presence of magnesium sulphate (MgSO₄) or nicotinic acid (which are 'chemical modulators' of BvgS) inactivates BvgS; thus, BvgA remains unphosphorylated and is unable to regulate transcription.

Figure 1: The BvgAS master regulatory system.

a | BvgS is a polydomain histidine sensor kinase that contains (from the amino to the carboxyl terminus) two periplasmically located venus flytrap domains (VFT1 and VFT2), a transmembrane domain, a PAS domain, a histidine kinase domain (HK), a receiver domain (Rec) and a histidine phosphoryl transfer domain (Hpt). BvgA is a response regulator protein that has an N-terminal Rec and a C-terminal helix–turn–helix domain (HTH). BvgS is active at 37 °C and becomes autophosphorylated at a conserved histidine (H) in the HK domain. The phosphoryl group is then transferred to the Rec domain, followed by the Hpt and finally to the Rec domain of BvgA. Phosphorylated BvgA (BvgA-P) dimerizes and activates the expression of virulence-associated genes (vag loci; which are subdivided into class 1 and class 2 genes) and represses the expression of virulence-repressed genes (vrg loci; which are class 4 genes). BvgS is inactive and remains unphosphorylated when bacteria are grown at a low temperature (∼25 °C) or at 37 °C in the presence of chemical modulators (such as magnesium sulphate (MgSO₄) or nicotinic acid). b | BvgAS controls four classes of genes and three distinct phenotypic phases. The effect of BvgAS activation on the various classes of genes is shown in the upper panel. The Bvg⁺ phase occurs when BvgAS is fully active and is characterized by maximal expression of genes that encode adhesins (class 2 genes, such as fhaB and fim; expression levels are indicated by the green line) and toxins (class 1 genes, such as cyaA–E, ptx–ptl and bsc genes; expression levels are indicated by the red line) and minimal expression of class 3 and class 4 genes (expression levels are indicated by purple and blue lines, respectively). The Bvg⁺ phase is necessary and sufficient to cause respiratory infection in vivo. The Bvg⁻ phase occurs when BvgAS is inactive and is characterized by maximal expression of class 4 genes and minimal expression of class 1, class 2 and class 3 genes. Notably, the regulation of some vrg loci is indirect; when BvgAS is inactive, it does not repress the expression of frlAB, which is a positive regulator at the top of the motility regulon, and it does not activate the expression of bvgR, which is an indirect negative regulator of vrg loci. The Bvg⁻ phase is required for growth under nutrient-limiting conditions, such as those that may be encountered in the ex vivo environment. The Bvgⁱ phase occurs when BvgAS is partially active and is characterized by the maximal expression of class 2 and class 3 genes and minimal expression of class 1 and class 4 genes. The only class 3 gene that has been characterized so far is bipA, which is activated by BvgA under Bvgⁱ phase conditions and repressed by BvgA under Bvg⁺ phase conditions. The Bvgⁱ phase may be important for transmission between hosts, but this has not been fully elucidated.

PowerPoint slide

BvgAS controls multiple phenotypic phases. The genes that are regulated by the BvgAS phosphorelay belong to four classes and their differential regulation results in at least three distinct phenotypic phases (Fig. 1b). Class 1 genes include the ptx–ptl operon (which encodes PT and its transport system), cyaA–E (which encodes ACT) and the bsc operon (which encodes a type III Secretion System (T3SS)). These genes are maximally expressed when BvgAS is fully active (known as the Bvg⁺ phase). Class 2 genes are maximally expressed in both the Bvg-intermediate (Bvgⁱ) and Bvg⁺ phases. The Bvgⁱ phase occurs when bacteria are grown in the presence of low concentrations of chemical modulators or within the first few hours following a switch from Bvg⁻ phase conditions to Bvg⁺ phase conditions. Class 2 genes include fhaB (which encodes filamentous haemagglutinin (FHA)), fim genes (which encode fimbriae) and bvgAS itself; thus, bvgAS is positively autoregulated. Class 3 genes, of which only one (bipA, which encodes an outer membrane protein of unknown function) has been characterized so far^35,36, are maximally expressed in the Bvgⁱ phase. Class 4 genes, which are also known as vrg (virulence-repressed genes) loci, are maximally expressed in the Bvg⁻ phase and include genes that are required for flagella synthesis and motility in B. bronchiseptica.

Role of BvgAS-mediated gene regulation. The conservation of BvgAS among Bordetella spp. and its ability to control multiple phenotypic phases in response to environmental cues suggests that it has an important and conserved role in the infectious cycle. As B. pertussis and B. parapertussis_Hu strains are unable to survive for extended periods of time outside the human host (unpublished observations from various research groups), it was hypothesized that BvgAS-mediated gene regulation must occur in the mammalian respiratory tract. Experiments with mutants that were locked in either the Bvg⁺ phase or the Bvg⁻ phase, or that ectopically expressed Bvg⁻ phase factors in the Bvg⁺ phase, showed that the Bvg⁺ phase is necessary and sufficient for respiratory infection, that cells in the Bvg⁻ phase are unable to survive in vivo and that failure to repress Bvg⁻ phase factors (such as flagella) is detrimental to the development of infection^37,38,39,40. Moreover, recent studies using sensitive reporter systems have provided strong evidence that switching to the Bvg⁻ phase does not occur in vivo^41,42. In B. bronchiseptica, the Bvg⁻ phase is required for survival under nutrient-limiting conditions, such as those that might be encountered in an external environment⁴³. It has been hypothesized that the Bvgⁱ phase is important for transmission, and with the development of the baboon model (Box 2), this hypothesis is now testable. Although additional regulatory systems are undoubtedly important during the infectious cycle of Bordetella spp., their precise roles have not yet been determined.

Toxins

Pertussis toxin. PT, which is sometimes referred to as lymphocytosis-promoting factor owing to its ability to induce lymphocytosis in mammals, was one of the first identified, and is one of the most extensively characterized, B. pertussis virulence factors⁴⁴. The presumed requirement of PT for the development of infection and the observed positive correlation between PT-specific immunity and bacterial clearance led to the hypothesis that pertussis, like cholera and diphtheria, is a toxin-mediated disease⁴⁵. However, although PT is important for pathogenesis, it is now clear that pertussis results from the coordinated function of many different virulence factors⁴⁶.

PT is an ADP-ribosylating AB₅-type toxin⁴⁷ (Fig. 2a). The holotoxin is composed of one catalytic subunit (the A subunit) and five membrane-binding or transport subunits (B subunits; which form the B pentamer), which are assembled in the periplasm and then exported by the type IV secretion system that is encoded by the ptl locus⁴⁸. PT holotoxin can bind to almost any sialic acid-containing glycoprotein⁴⁹, and thus, multiple receptors have been identified and characterized in a broad range of cell types in vitro²⁰; however, the specific cell types that are targeted by PT in vivo are unknown. After binding, PT enters the host cell by receptor-mediated endocytosis and follows a retrograde transport pathway to the Golgi apparatus and then the endoplasmic reticulum (ER)⁵⁰ (Fig. 2b). The A subunit exits the ER, possibly by 'hijacking' the ER-associated degradation pathway that normally expels misfolded proteins⁵¹. In the cytoplasm, the A subunit catalyses the transfer of ADP–ribose from NAD⁺ to a cysteine residue near the C terminus of the α-subunit of heterotrimeric G proteins, some of which are inhibitory G proteins. Among other downstream effects, this modification eliminates the ability of these inhibitory G proteins to inhibit adenylate cyclase activity (resulting in increased cyclic AMP (cAMP) levels in the cell) and blocks other G protein-regulated enzymes and pathways^20,52, leading to dysregulation of the immune response.

Figure 2: Toxin-mediated virulence of Bordetella spp.

a | Pertussis toxin (PT; Protein Data Bank (PDB) accession 1PRT) is an AB₅-type toxin that is composed of one catalytic subunit (the A subunit) and five membrane-binding or transport subunits (the B pentamer)⁴⁷. PT is assembled in the bacterial periplasm and exported by a type IV secretion system. b | Following binding to a sialoglycoprotein host cell receptor, PT is endocytosed and trafficked through the Golgi apparatus to the endoplasmic reticulum (ER). In the ER, the B pentamer binds to ATP and dissociates from the A subunit. The A subunit is then transported into the cytoplasm and traffics on exosomes to the cytoplasmic membrane, where it ADP-ribosylates the α-subunit of heterotrimeric G proteins. This modification alters the ability of G proteins to regulate multiple enzymes and pathways, including their ability to inhibit cyclic AMP (cAMP) formation. The overall result of these modifications is an initial suppression of inflammatory cytokine production and an inhibition of immune cell recruitment to the site of infection. c | Bordetella adenylate cyclase toxin (ACT) is composed of two primary domains: a calmodulin-responsive adenylate cyclase enzymatic domain and an RTX (repeats in toxin) domain, which are connected by hydrophobic segments. d | The RTX domain of ACT interacts with complement receptor 3 (CR3), which is expressed on host cell membranes from a wide range of cell types. The hydrophobic segments of the linker region form pores in the membrane, enabling the passage of cations and the adenylate cyclase domain is translocated into the cytoplasm. These two activities are mediated by distinct conformations of ACT. Adenylate cyclase activity is stimulated by binding to calmodulin in the host cell, leading to an increase in cAMP production. The combined effects of ACT intoxication and pore formation result in inhibition of complement-dependent phagocytosis, induction of anti-inflammatory cytokines, suppression of pro-inflammatory cytokines and inhibition of immune cell recruitment.

PowerPoint slide

PT has an extraordinarily broad range of pharmacological effects in cell culture and animal models, which has confounded efforts aimed at identifying its precise role (or roles) during human infection. PT inhibits the migration of cells that express G protein-coupled chemokine receptors in vitro, such as neutrophils, monocytes and lymphocytes⁵³. In mouse models, the production of PT by B. pertussis correlates with decreased pro-inflammatory chemokine and cytokine production, decreased recruitment of neutrophils to the lungs and increased bacterial burdens early in infection^54,55. Experiments in mice, in which alveolar macrophages are depleted using clodronate, suggest that PT initially targets these cells⁵⁶. PT production at the peak of infection correlates with exacerbated inflammation and pathology in the airways⁵⁷. Although these and other observations in animal models suggest that PT contributes to the establishment of infection by suppressing early inflammation and inhibiting the microbicidal action of inflammatory cells, in addition to contributing to inflammatory pathology at the peak of infection, it is not known whether PT induces these effects during human infection. However, it has been shown that the production of PT positively correlates with the extreme lymphocytosis that occurs in primary human pertussis cases⁵⁸, and antibodies against PT protect against severe disease⁵⁹.

Adenylate cyclase toxin. ACT (Fig. 2c, d), which is a member of the RTX (repeats in toxin) toxin family, is encoded by cyaA and is produced by all Bordetella subspecies that infect mammals¹⁹. ACT is secreted by the cyaBDE-encoded type I secretion system and is palmitoylated by the product of cyaC (Refs 60, 61). The toxin contains two distinct functional modules: the C-terminal domain, which contains the RTX repeats, mediates binding to target cells and forms cation-selective pores in plasma membranes^62,63; and the N-terminal domain, which is a calmodulin-dependent adenylate cyclase that converts ATP to cyclic AMP (cAMP)^64,65. Recent studies indicate that ACT can adopt multiple conformations and that these forms are distinct in their ability to effect pore formation or the translocation of adenylate cyclase into the host cell⁶⁶. Thus, the observed effects of ACT on different cell types are the result of a combination of ion permeability, increased levels of cAMP (which leads to the perturbation of downstream signalling events) and possibly, the depletion of intracellular ATP.

Although ACT can intoxicate many cell types, it binds with high affinity to complement receptor 3 (CR3; also known as αMβ2 integrin or CD11b/CD18), which is present on neutrophils, macrophages and dendritic cells⁶⁷, and early work correlated ACT-dependent cAMP production in human neutrophils with the inhibition of phagocytosis and oxidative burst⁶⁸. More recent studies have shown that ACT blocks complement-dependent phagocytosis by macrophages⁶⁹. In addition, this toxin also suppresses the activation and chemotaxis of T cells⁷⁰. The importance of these in vitro observations is unclear; however, a recent study using the baboon model and clinical samples from humans showed that the concentrations of ACT in B. pertussis-infected respiratory tissues are considerably lower than the concentrations of purified protein that are used in most in vitro studies⁷¹. In mouse models, ACT-deficient bacteria are cleared faster than wild-type bacteria, and studies using immunodeficient and neutropenic mice suggest that ACT has a crucial role in enabling bacteria to resist neutrophil-mediated clearance^72,73. These data, in addition to the fact that ACT is one of the few virulence factors that is conserved and produced by all pathogenic Bordetella species⁵, suggest that ACT has the potential to be an effective antigen in future vaccine formulations⁷⁴.

Type III secretion. For reasons of experimental tractability, the Bordetella Bsc type III secretion system (T3SS) is most extensively studied in B. bronchiseptica and induces caspase-independent necrotic death in a diverse range of cell types in vitro⁷⁵. Mutations that eliminate T3SS activity decrease bacterial persistence in the lower respiratory tracts of rats and mice, following intranasal inoculation^76,77. Infection of mice with T3SS-defective B. bronchiseptica mutants also results in a more robust antibody response, and re-stimulated splenocytes from animals that have been infected with these mutants show increased production of pro-inflammatory interferon-γ (IFNγ) and decreased production of anti-inflammatory interleukin-10 (IL-10)⁷⁸. Consistent with these data, IFNγ has been shown to facilitate clearance of B. bronchiseptica from the lower respiratory tract, whereas IL-10 delays clearance⁷⁸. Together, these observations suggest that the Bsc T3SS has an immunomodulatory role that promotes persistence in the lower respiratory tract; however, the mechanistic basis for this phenomenon remains to be determined.

Remarkably, and despite concerted efforts by several research teams, only a single effector protein, BteA, has been definitively identified as a translocated substrate of the Bsc T3SS^79,80. BopN — a homologue of outer membrane protein YopN (which regulates type III secretion in pathogenic Yersinia spp.) — has been proposed to be a second effector⁸¹, but thus far, evidence that BopN is translocated by the Bsc system is lacking. BteA is both necessary and sufficient for cytotoxicity in vitro, and mutations in bteA recapitulate the phenotypes that are associated with eliminating T3SS activity in vitro and in vivo^77,79. Following translocation into host cells, the N-terminal targeting domain results in BteA localization to the ezrin-rich lipid rafts that underlie sites of bacterial attachment⁸². However, the mechanisms responsible for the potent cytotoxicity of BteA remain unclear.

Type III secretion is tightly regulated in Bordetella spp. The bteA and bsc genes are transcriptionally activated by the alternative sigma factor BtrS, which is activated by BvgAS⁸³. Expression of the bcs genes is also upregulated by iron starvation⁸⁴. In addition to these regulatory mechanisms, the partner-switching proteins BtrU, BtrV and BtrW mediate a cycle of serine phosphorylation and dephosphorylation events, which regulate secretion activity^83,85.

Perhaps the most pressing question regarding the Bsc T3SS relates to its potential role during human infection. A requirement for T3SS activity in B. pertussis cytotoxicity has not been documented, despite the fact that T3SS genes are intact and are highly conserved, transcribed and regulated, in addition to the observation that bteA alleles are functionally interchangeable between subspecies^82,83. Fortunately, recent studies are beginning to shed light on this paradox. Although Bsc activity is not generally observed in laboratory-adapted B. pertussis strains, the tip complex of the T3SS, Bsp22, is secreted by clinical isolates in vitro, and mutations in the ATPase gene, bscN, result in elevated production of pro-inflammatory cytokines and accelerated clearance of B. pertussis from the lungs of aerosol-infected mice⁸⁶. Furthermore, T3SS activity seems to be lost following laboratory passage of B. pertussis and regained after passage in mice^86,87.

Tracheal cytotoxin. Tracheal cytotoxin (TCT) is a disaccharide–tetrapeptide monomer of peptidoglycan that is produced during cell wall remodelling⁸⁸. Although most Gram-negative bacteria recycle this molecule^89,90, B. pertussis does so inefficiently and releases a large amount of TCT into the extracellular environment. TCT is the only known B. pertussis virulence factor that is not regulated by BvgAS. In hamster tracheal rings, TCT functions synergistically with lipo-oligosaccharide to stimulate the production of pro-inflammatory cytokines (such as tumour necrosis factor α (TNFα), IL-1α, IL-1β and IL-6) and inducible nitric oxide synthase (iNOS), resulting in the destruction and extrusion of ciliated cells from the epithelial surface^91,92. The biological activity of TCT depends on NOD1 (nucleotide-binding oligomerization domain-containing protein 1), which is a cytosolic pattern recognition receptor that senses bacterial peptidoglycan and induces the production of pro-inflammatory mediators⁹³. NOD1-dependent detection of TCT seems to be host specific, as human NOD1 poorly detects TCT, whereas mouse NOD1 detects TCT efficiently⁹³. Although it has been postulated that TCT-mediated cytopathology contributes to the characteristic cough in pertussis, the lack of appropriate animal models has prevented this hypothesis from being tested. Thus, the contribution of TCT to pertussis pathogenesis in humans remains unclear.

Dermonecrotic toxin. Subcutaneous injection of B. pertussis or B. bronchiseptica cells into mice results in the formation of necrotic lesions, owing to the activity of dermonecrotic toxin (DNT)⁹⁴. Consistent with a role in infection, the production of DNT is positively regulated by BvgAS^26,29 and there is evidence that DNT contributes to the ability of B. bronchiseptica to induce turbinate atrophy and lung pathology in swine⁹⁵. DNT has transglutaminase activity, can activate Rho GTPases^96,97 and inhibits osteogenic cell differentiation in vitro, which suggests that the toxin acts directly on host cells^98,99. However, as DNT lacks a signal sequence for export and is not secreted from bacterial cells grown in culture^94,100, it may actually function in the bacterial cytoplasm during infection — possibly by facilitating bacterial survival within a specific host niche and hence indirectly functioning in pathogenesis.

Surface Structures

Filamentous haemagglutinin. Filamentous haemagglutinin (FHA) (Fig. 3a) is a large rod-shaped protein and, together with FhaC (filamentous haemagglutinin transporter protein), it functions as a prototypical member of the two-partner secretion pathway (TPS pathway)¹⁰¹. It is initially synthesized as a ∼370 kDa pre-pro-protein (FhaB) that undergoes processing to produce the mature ∼250 kDa FHA as it is translocated across the cytoplasmic membrane by the Sec translocation system and across the outer membrane by FhaC¹⁰². The N-terminal signal peptide is probably removed by leader peptidase and the C-terminal prodomain is processed by SphB1 (autotransporter subtilisin-like protease) and other, as yet unidentified, factors^103,104. Mature FHA is oriented with its mature C terminus (the MCD) distal to the bacterial surface, and a substantial amount of FHA is also released into culture supernatants when the bacteria are grown in vitro¹⁰⁴.

Figure 3: Presentation of filamentous haemagglutinin, fimbriae and pertactin on the Bordetella cell surface.

a | Filamentous haemagglutinin (FHA) is a TpsA exoprotein that is translocated across the outer membrane through its cognate TpsB pore protein, FhaC. This translocation occurs via the two-partner secretion pathway. Processing during translocation removes the carboxy-terminal prodomain (yellow) from the full-length FhaB protein to produce the mature ∼250 kDa FHA protein. The C-terminal domain of mature FHA (the mature C-terminal domain, MCD) is required for functions that include adherence to ciliated epithelial cells and for persistence during infection, possibly by directly or indirectly modulating the host immune system. b | Bordetella fimbriae are type I pili. FimB is similar to chaperone proteins that traffic major fimbrial subunits (Fim2 and Fim3, in this case) to the outer membrane usher protein. FimC is probably the usher protein and, by analogy with other systems, it probably functions as a dimer, although its activity has not been experimentally demonstrated. Fim2 and Fim3 are the major pilin subunits and FimD is likely to be the fimbrial tip protein. Fimbriae are required for persistence during infection, possibly by functioning similarly to FHA and directly or indirectly modulating the immune response. Furthermore, studies have suggested that fimbriae are necessary for adherence to ciliated epithelial cells. c | Pertactin is a classical autotransporter. The C-terminal ∼30 kDa domain (orange) forms a channel in the outer membrane, which is required for the translocation of the ∼70 kDa β-helical passenger domain (blue) to the cell surface. Although the precise role of pertactin is unclear, data suggest that pertactin may contribute to virulence by resisting neutrophil-mediated clearance.

PowerPoint slide

FHA is both necessary and sufficient to mediate bacterial adherence to several eukaryotic cell types in vitro^105,106. However, FHA is only one of several factors that contribute to the adherence of bacteria to tracheal explants^9,107, which suggests that additional adhesins are important for adherence in vivo. Studies using cultured, non-ciliated cells have reported that FHA binds to CR3, very late antigen V (VLA-5) and leukocyte response integrin–integrin-associated protein (LRI–IAP) complexes, and an RGD (Arg-Gly-Asp) motif that is located in the centre of the FHA molecule is implicated in this process^108,109,110. More recent studies that have examined B. bronchiseptica infection using animal and cell culture models have shown that the FHA molecules that are produced by B. pertussis and B. bronchiseptica are functionally interchangeable. These studies have also shown that the production of an FHA protein that contains an RAE (Arg-Ala-Asp) motif instead of an RGD motif results in no observable differences and that the MCD is required for function¹¹¹. Whether FHA interacts with CR3, VLA-5, LRI–IAP or other mammalian receptors during infection has yet to be determined.

Experiments in which B. bronchiseptica is delivered in a small volume to the nasal cavities of rats and pigs have shown that FHA is essential for progression of the infection from the upper to the lower respiratory tract^111,112. In mouse models, in which large numbers of bacteria are directly delivered into the lungs, FHA-deficient B. bronchiseptica strains induce a more robust inflammatory response than wild-type bacteria^73,111. This response is characterized by increased production of pro-inflammatory cytokines and chemokines (such as TNFα, keratinocyte-derived chemokine (KC), CC-motif ligand 2 (CCL2; also known as MCP-1) and IL-17) in lung tissue and increased recruitment of neutrophils to the lungs during the first 4 days post-inoculation⁷³. Animals that do not succumb to inflammation-mediated pulmonary damage clear the FHA-deficient bacteria from their lungs much faster than animals that are inoculated with wild-type bacteria^73,111. These data suggest that FHA enables B. bronchiseptica to modulate inflammation during the establishment of infection, thereby facilitating bacterial persistence. It is currently not known whether FHA exerts these effects by binding directly to host receptors while it is attached to the bacterial cell surface or whether these effects occur after FHA release from the bacterial cell. Furthermore, it has been suggested that FHA functions as a scaffold to direct the delivery of other virulence factors (such as ACT¹¹³); however, the in vivo relevance of this activity has not been determined.

Fimbriae. Bordetella spp. produce type I pili, which are also known as fimbriae (Fig. 3b). The putative chaperone (FimB), usher (FimC) and tip adhesin (FimD) are encoded by the fimBCD operon, which is located between the fhaB and fhaC genes¹¹⁴. The fim2 and fim3 genes, which encode the two major fimbrial subunits, are located elsewhere on the chromosome and can undergo phase variation¹¹⁵. Alternative major fimbrial subunit genes (fimA, fimN and fimX) have also been identified^116,117,118. Although in vitro adherence assays using cultured cells have yielded variable results^119,120, studies with tracheal explants indicate a role for fimbriae in mediating adherence to ciliated respiratory epithelium^9,107. Studies with both B. pertussis and B. bronchiseptica have shown a requirement for fimbriae during colonization of the lower respiratory tract in rodents^120,121, and mice that have been inoculated with Fim-deficient B. pertussis show a more robust inflammatory response than mice that have been inoculated with wild-type bacteria¹²². Similarly to FHA, fimbriae seem to be involved in adherence and/or suppression of the initial inflammatory response to infection, potentially contributing to persistence.

Pertactin. Pertactin (PRN) is a member of the classical autotransporter family of outer membrane proteins¹²³ (Fig. 3c). The surface-localized 'passenger' domain forms a β-helix in which β-strands are connected by short turns or, in a few cases, large extrahelical loops¹²⁴. Similarly to fimbriae, studies using non-ciliated mammalian cells to investigate a role for PRN in adherence or invasion have yielded equivocal results^112,125. Studies using ciliated rabbit tracheal explant cultures suggest that PRN contributes to the adherence of B. pertussis to ciliated respiratory epithelium⁹, although experiments in mice failed to identify a role for PRN in vivo¹²⁶. However, in the case of B. bronchiseptica, studies indicate that PRN is involved in resistance to neutrophil-mediated clearance and promoting persistence in the lower respiratory tract^112,125. In recent years, B. pertussis strains that do not produce PRN have been isolated from patients¹²⁷, raising the concern that such strains have been selected owing to the presence of PRN-specific antibodies, which are generated in response to immunization with PRN-containing aP vaccines. Whether vaccine-driven evolution of B. pertussis strains is actually occurring is currently under investigation, as it has decisive implications for the development of new and improved vaccines.

Lipopolysaccharide. B. pertussis, B. parapertussis_Hu and B. bronchiseptica produce different forms of lipopolysaccharide (LPS). B. pertussis produces a penta-acylated lipid A that is linked to a complex core trisaccharide, B. bronchiseptica produces a hexa-acylated lipid A that is linked to a similar, if not identical, complex core trisaccharide and O-antigen repeats, and B. parapertussis_Hu produces a hexa-acylated lipid A that is linked to an altered core structure and O-antigen repeats^128,129,130. Because it lacks O-antigen, B. pertussis LPS is often referred to as lipooligosaccharide (LOS)¹³¹. The genes that are required for the synthesis of O-antigen in B. bronchiseptica and B. parapertussis_Hu are repressed by BvgAS¹³²; however, some O-antigen is produced under Bvg⁺ phase conditions, and mutants that are unable to produce O-antigen show defective virulence in mouse models^132,133.

In mice, B. bronchiseptica LPS is sensed by Toll-like receptor 4 (TLR4), resulting in an early TNFα response and recruitment of neutrophils to the lungs^134,135. Although B. parapertussis_Hu LPS and B. pertussis LOS can stimulate mouse TLR4, they do so less efficiently, and Tlr4^−/− mice are only modestly impaired in their ability to control infection by these organisms^136,137,138. In addition, it has been reported that B. pertussis LOS-mediated stimulation of mouse dendritic cells results in the development of anti-inflammatory regulatory T cells¹³⁶. On the basis of these observations, it has been suggested that B. pertussis and B. parapertussis_Hu have evolved to be less inflammatory than B. bronchiseptica and that diminished inflammation might facilitate persistence during human infection^136,137. However, subsequent studies have shown that human and mouse TLR4–MD-2–CD14 complexes differ in their ability to recognize different forms of lipid A. Although mouse TLR4–MD-2–CD14 responds similarly to both penta- and hexa-acylated lipid A, human TLR4–MD-2–CD14 responds robustly to hexa-acylated lipid A but only weakly to penta-acylated lipid A¹³⁹. Furthermore, in contrast to mouse TLR4–MD-2–CD14, which responds to B. pertussis lipid A regardless of whether the phosphate groups are modified or not, human TLR4–MD-2–CD14 responds more robustly to lipid A that contains glucosamine (GlcN)-modified phosphate groups than to lipid A that contains unmodified phosphate groups¹⁴⁰. Although it seems that most B. pertussis LOS contains GlcN-modified phosphate groups¹⁴¹, the fact that it is penta-acylated suggests that its ability to stimulate TLR4 in humans is even weaker than its ability to stimulate TLR4 in mice. These data provide additional support for the hypothesis that B. pertussis and B. parapertussis_Hu strains have evolved to be relatively non-inflammatory in humans. However, these data also raise concerns about extrapolating conclusions that have been drawn from mouse studies to humans, as the TLR4–MD-2–CD14-dependent immune responses clearly differ in these hosts.

Additional surface proteins. Many additional BvgAS-activated genes encode known or predicted surface-localized or secreted proteins, which are suspected to have roles in pathogenesis^29,142. BrkA, TcfA, BapC, BatB, Vag8, SphB1 and Phg are BvgAS-activated classical autotransporter proteins, and their putative roles in pathogenesis include mediating adherence, serum resistance, the evasion of antibody-mediated clearance and the proteolytic processing of other surface proteins^{103,143,144,145,146,147,148}. BipA and BcfA are BvgAS-regulated members of the intimin–invasin family and, although their roles in pathogenesis are unknown^35,149, immunization of mice with BcfA can accelerate clearance of B. bronchiseptica following intranasal challenge¹⁵⁰. These data suggest that these poorly characterized surface molecules should be considered for the development of new vaccines containing different or alternative antigens.

Metabolic proteins

Many BvgAS-regulated genes encode proteins that are probably involved in metabolism, respiration and other physiological processes^29,142, which presumably reflects the diversity of environmental conditions encountered by Bordetella spp. as they travel within and outside the mammalian respiratory tract. Among these factors, those that are involved in the acquisition and use of iron have been the focus of most studies. In addition to producing and using the siderophore alcaligin¹⁵¹, B. pertussis and B. bronchiseptica can use various xenosiderophores (including enterobactin¹⁵²) and haem–iron sources, such as haemoglobin¹⁵³. Most, if not all, of these iron-acquisition mechanisms are required during respiratory infection in mice^154,155, which demonstrates the necessity of iron for bacterial survival, the variety of mechanisms that are used by the host to sequester iron and the reciprocal array of mechanisms that are used by the bacterium to acquire this essential element.

In addition, accumulating evidence suggests that biofilm production by pathogenic Bordetella spp. in vitro and during infection may contribute to colonization of the respiratory tract. Biofilm formation is regulated by a complex programme of both Bvg-dependent and Bvg-independent gene expression^{156,157,158,159}; genes that promote this process are maximally expressed in the Bvgⁱ phase¹⁵⁶. Bvg-independent production of an exopolysaccharide via expression of the bps locus and the presence of extracellular DNA are also required for biofilm formation^157,160,161. Recent evidence suggests that the second messenger cyclic-di-GMP is also crucial for the regulation of biofilm formation¹⁶².

Current and future challenges

Despite high rates of immunization with aP vaccines, epidemics of pertussis have recently occurred in the United States, Europe, Australia and Japan (US Centers for Disease Control and Prevention (CDC), Australian Government Department of Health and Ageing and Japanese National Institute of Infectious Diseases; see further information)^163,164 and similar outbreaks seem to be imminent in developed countries throughout the world. Moreover, irrespective of socioeconomic status, the highest rates of mortality are in infants, who are also the most difficult population to treat and protect. In considering these challenges and looking ahead, we suggest three priorities for future studies.

The first priority is to improve the robustness and duration of protection that is conferred by vaccination, which will require further study of the immunological responses to infection and vaccination (Box 4). The deficiencies of current aP vaccines are well documented, including the striking observation that aP vaccination of baboons protects against disease symptoms but not against colonization or transmission¹⁶⁵. Many efforts are in progress to overcome these deficiencies²⁴, such as the inclusion of additional antigens in aP vaccines, reformulation with adjuvants that favour T helper 1 (T_H1) cell and T_H17 cell responses, as opposed to the T_H2 cell-type immunity that is generated by alum-adjuvanted vaccines, as well as the development of live attenuated B. pertussis vaccines¹⁶⁶. The development of live attenuated vaccines has considerable advantages, including the ability to generate mucosal immunity, but the issue of public acceptance looms large. Similarly, it is interesting to note that outside North America, Europe and parts of Asia, wP vaccines remain in widespread use, and approaches to decrease their reactogenicity while retaining their immunogenicity should be considered¹⁶⁷. The known efficacy of wP vaccines, combined with the cost-effectiveness of this approach, might be of more benefit to people than the development of improved but more costly vaccines that are composed of purified proteins. It is important to remember that the development and approval of novel vaccines will be a prolonged process. In light of recent findings concerning the lack of protection against colonization or transmission by aP vaccination¹⁶⁵, maximizing the efficacy of current vaccines by prenatal vaccination, additional boosting and alternative strategies is imperative.

A second priority is to mitigate infant mortality. Nearly 90% of all deaths from pertussis occur in infants who are less than 4 months old¹⁶⁸, and the most frequent cause is intractable pulmonary hypertension associated with marked lymphocytosis and bronchopneumonia. Currently, the only efficacious therapy for severe cases is rapid leukodepletion, which is only available at advanced critical care centres^169,170. Respiratory samples that have been obtained during autopsies show luminal aggregates of leukocytes occluding small pulmonary arteries, along with an abundance of B. pertussis^10,171. The pathology of fatal pertussis pneumonia seems to be mostly caused by PT. Thus, in addition to protecting susceptible infants by maternal vaccination or by vaccination at birth, it is also imperative to pursue approaches for limiting PT activity during infection. Potential therapeutic modalities include the development of humanized monoclonal antibodies and small molecules that inhibit PT interactions with host cell receptors or the enzymatic activity of PT, as well as therapeutic compounds designed to target regulatory factors such as the BvgAS system.

Finally, although animal models have proven to be useful, we need to improve our understanding of human disease. Decades of research on B. pertussis virulence determinants have primarily been based on tissue culture models and mouse infections. These studies have shown what adhesins, toxins and other virulence factors can do under laboratory conditions, but very little, if anything, is known about what they really do during human disease. Specificity is the rule for human-adapted pathogens and it can manifest at several levels, including gene expression, virulence-factor delivery, binding-specificity and activity. Perhaps the most vivid illustration of our lack of understanding of B. pertussis is that we still don't know why infection makes people cough!

Box 4: Immunity to pertussis

Clinical studies suggest that both humoral and cell-mediated immunity are important for controlling human pertussis. Measurements of Bordetella pertussis-specific T cell proliferation, cytokine production and titres of different IgG subclasses from human samples suggest that naive infection primarily causes a T helper 1 (T_H1) cell response, resulting in the production of pro-inflammatory cytokines and opsonizing antibodies, combined with the stimulation of antigen presenting cells^193,194. Whole cell pertussis (wP) vaccines also stimulate a T_H1 cell response^195,196, whereas acellular pertussis (aP) vaccines seem to produce a mixed T_H1 cell and T_H2 cell response^196,197,198. Many studies have shown that vaccination with aP vaccines produces antibody responses that are equal to, or that exceed, those produced by vaccination with wP vaccines^194,196,198. However, the immunity that is induced by aP vaccines is inefficient, highlighting the importance of investigating and evaluating the cell-mediated immune responses that are induced by vaccination and infection. Studies that have evaluated human infection or vaccination have not comprehensively addressed the role of T_H17 cell responses, which result in pro-inflammatory cytokine production and the stimulation of professional phagocytes. However, studies in the baboon infection model show that infection causes a mixed T_H1 cell and T_H17 cell response that is long-lived and protective against colonization following subsequent challenge^186,199. Immunological data from animal models mostly agree with these findings^23,165,199. Regardless of the natural immune response to infection, efficacious vaccines need to be long-lasting, prevent transmission and reduce disease burden. As B. pertussis is primarily an extracellular respiratory pathogen, it is likely that an effective immune response requires the induction of a mixed T_H1 cell and T_H17 cell response that stimulates the production of opsonizing, toxin-neutralizing and mucosal antibodies, along with memory T cells that produce cytokines to recruit and activate professional phagocytes at the site of infection. As such, there is an urgent need to re-evaluate antigens, adjuvants and immunization routes to achieve these goals.