Bordetella pertussis is the bacterial microbe responsible for the disease "Pertussis" (commonly referred to as 'Whooping Cough'). In this site, we will explore the physical attributes of the microbe, as well as the various toxins produced which facilitate both the infection and proliferation of the disease. Specifically, we will analyze the history of disease in Canada, the steps taken to eradicate disease, and potential treatments for the prevention of future outbreaks.
Description of the microbe
The bacterium Bordetella Pertussis is the causative agent of Whooping Cough. This small bacterium (about .8um in length and .4um by width) cannot survive in the open environment, and therefore must reside in a host, and is considered pathogenic. This host is almost always a human, where its “natural” environment is said to be the mucus in the human respiratory tract. The optimum temperature of growth for B.pertussis is about 35-37C, which is the temperature inside a living human host.
B.pertussis is an aerobic gram negative bacterium that does not produce spores. Due to its lack of flagella, it is also immotile. It resides in the phylum Proteobacteria, class Betaproteobacteria, order Burkholderiales, and family Alcaligenaceae. Because it is an aerobe, it utilizes aerobic respiration. Therefore, the bacterium consists of an electron transport chain on its membrane, and is considered a chemoheterotroph. Like other gram negative bacteria, it possesses an inner and outer membrane, with a thin peptidoglycan cell wall in between. This cell wall is attached to the outer membrane via lipoproteins. Like other gram negatives, the outer membrane of B.pertussis is coated with LPS (lipopolysachharides), which are long sugar linked and lipid-anchored polysaccharides coating the outer membrane. LPS are endotoxins, meaning that they are toxic to host (or potentially other bacteria) cells when detached from the bacteria(2). However, B.pertussis consists of very unusual LPS compared to other gram negatives. Namely, it contains two forms of LPS that have a different phosphate composition than standard Lipid A form LPS.
Bordetella pertussis does not necessarily cause whooping cough by itself, but rather produces toxins that do so. Among producing pertussis toxin, filamentous hemagglutinin, and hemolysin, B.pertussis produces the adenylate cyclase toxin (CyaA) – a direct causative agent of Whooping Cough (1). B.pertussis attaches to host cells via the virulence factor P.69 pertactin, which is an extracellularly exposed domain of a membrane protein expressed by B.pertussis – and targets the adhesion to mammalian cells. In addition to this, the filamentous hemagglutinin (FHA) also helps mediate cell adhesion, and contains similar motifs as P.69 pertactin (2). Once attached, B.pertussis is able to transfer CyaA to the target host cell.
The pertussis toxin (PT), also produced by B.pertussis, is known to also be an important factor in whooping cough respiratory infections. The resultant coughing is painful, and also is what spreads the highly contagious B.pertussis to others (3).
The Tahoma 1 strain of B.pertussis has its entire genome sequenced – it is 1 circular chromosome 4,086,189 nucleotides long (3867 genes), of which about 67% of the entire genome is GC rich (4).
Link to MicrobeWiki Page: Bordetella pertussis http://microbewiki.kenyon.edu/index.php/Bordetella_pertussis
Description of the disease
Bortadella Pertussis secretes numerous toxins known to affect cellular homeostasis. The primary disease caused by the microbe is known as “Pertussis” (commonly referred to as 'Whooping Cough').
General Disease Information:
Whooping cough is characterized by uncontrollable, heavy coughing fits which persist for many weeks. Initially the symptoms may mimic that of the common cold, later developing into severe coughing and the potential development of further illness i.e. pneumonia, hypoxia (oxygen deprivation to tissues), apnea (suspension of breathing), seizures, etc. The disease progression can be broken up into several stages, each characterized by different symptoms. Initially the patient enters the “Catarrhal Phase” in which they display symptoms like that of the common cold. This usually develops within the first two weeks of disease. The second major stage is referred to as the “Paroxysmal Stage” and is characterized by uncontrollable fits of coughing followed by high pitched ‘whoop’ sounds upon air uptake. This stage is the most severe and is often the time when a patient will exhibit other disease symptoms. If the patient has the immune strength to combat the illness, they enter the final stage called the “Convalescent Stage”. Here the patient begins recovery from disease and uncontrollable coughing slowly diminishes (6).
Molecular Mechanisms of Disease:
Pertussis can act through numerous virulent mechanisms to affect the host organism. Two of the more notable toxins secreted by Bortadella pertussis are Pertussis Toxin (PT) and Adenylyl Cyclase Toxin (CyaA).
Pertussis Toxin (PT)
Pertussis toxin directs its activity intracellularly by affecting the inhibitory subunit of G-protein coupled receptors. PT catalyzes the ADP-Ribosylation of the Gi subunit and prevents inhibition of the protein Adenylyl Cyclase (AC). Adeneylyl Cyclase is responsible for the production of intracellular cAMP, a common downstream messenger in many cell signaling pathways.
In a normal cell, Gi subunits work concurrently with Gs proteins. Gs proteins are responsible for the activation of AC while Gi is responsible for inhibiting AC. When functioning properly, the two proteins regulate intracellular cAMP levels to maintain cellular homeostasis (3, 4, 6). In a cell affected by PT, Gi is inhibited and cannot act to “shut off” AC. As a result, the AC is not regulated and will constitutively produce cAMP until the Gs protein is removed. These high levels of cAMP have extreme downstream effects on gene expression and the viability of the cell. This destroys the ciliated epithelial cells that normally would sweep away mucus, and without this ciliated action, vigorous coughing is needed to clear the airway (3, 4, 6).
Adenylate Cyclase Toxin (CyaA)
CyaA is the primary toxin that contributes the development of 'Whooping cough' symptoms. CyaA has a series of G and D rich polypeptide residues, which is a characteristic repeat for many toxins. It predominantly controls the initial localization of microbe to the lung and airway tissues for colonization. B. pertussis adheres to the host cell via P.69 pertactin virulence factor. It consists of an RGD tripeptide motif which attaches to various sites on mammalian adhesion proteins (some include fibronectin, vitronectin, and fibrinogen) which facilitates interaction between the mammalian host cell and the bacterial microbe (2). CyaA acts on cells in both calcium and temperature-dependent manners where the CyaA catalytic domain is directly moved across the target cells plasma membrane. Specifically, it binds and invades cells by forming transmembrane cation selective channels. These channels can further affect the affected cell by offsetting cellular homeostasis, occasionally resulting in cellular lysis. CyaA may also contribute to the increase in intracellular cAMP levels of the host cell (1,5,6,7).
Transmission of disease
Bordetella pertussis cannot live in the outside environment. As a result, it has evolved to become a highly contagious microbe which is most commonly spread airborne. B. pertussis has negative effects on the airway tissues resulting in the loss of cilia mucus sweeping functions. The affected individual must now cough rigorously to clear the airway of excess mucus. The strong, uncontrollable coughing provides a mechanism for B. pertussis to leave the internal host environment and infect other individuals.
How/why is pertussis a problem in Canada
In Canada, the number of reported infected individuals of pertussis cases has fluctuated throughout the years, due to outbreaks and despite new advances in vaccination technology; reported infected individuals has ranged from 2,165 to 10,151 within this period of time. These numbers are predicted to be under-representations of the true values due to incomplete diagnosis or misreporting (8). The number of individuals affected per 100,000 of population has been increasing for adults through the time period of 1987-2004. However, this disease has been reported to be more common in infants than in older children and adults. The percentage of reported infected adolescents (defined to be less than 15 years of age) and adults has been “increasing from 9.6% in 1995 to 16.4%, 21.2%, and 31.3% in 1998, 2001 and 2004 respectively” (8).
Use of whole cell pertussis vaccine
In the 1940s, a whole cell pertussis fluid vaccine was introduced in Canada and was then replaced by an absorbed whole cell prtussis fluid vaccine in the 1980s. The absorbed whole cell pertussis fluid vaccine is made by suspending killed B. pertussis cells with an addition of other agents. These agents included in the vaccine are either absorbed diphtheria and tetanus toxoids (DPT), absorbed diphtheria and tetanus toxoids plus inactivated polio vaccine (DPT-Polio), or DPT or DPT-Polio combined with Haemophilus influenzae type b (Hib) conjugate vaccine (DPT-Hib, DPT-Polio-Hib) (9). The combined vaccines trigger an increase antibody production to B. pertussis, as well as diphtheria and tetanus antigens (9).
There have been adverse reactions towards the whole-cell pertussis vaccines. There have been reports of minor local reactions, which result in redness, swelling, as well as pain, in 50% to 75% of vaccine patients. It was proposed that a new type of vaccine was needed since the whole cell pertussis vaccine did not seem protected, and the polymorphism mutations of B. pertussis rendered the vaccine less effective (10). This low efficacy resulted in high numbers of reported cases of B. pertussis, and thus became a outbreak.
The efficacy of the whole-cell pertussis vaccines was 70%-90% in 1940s and 1950s. However, its efficacy has dimished through the last four decades, with efficacy percentages ranging from 40-100%, due to the vaccination increase and thus increasing the immunity against the vaccine (11).
Use of polymerase chain reaction as detection
One of the culprits behind few of the pertussis outbreaks could be the use of nucleic acid amplification tests, such as polymerase chain reaction (PCR), as diagnosis of pertussis.
The PCR method, for testing for pertussis, is carried out by culturing the sample and adding florescent antibody conjugates and PCR. Then, after the sample is cutured, swabs are elided and lysed. The PCR test are carried out on lysate using primers and then the product is detected through gel electrophoresis methods and stained with ethium bromide (14).
PCR is a very sensitive method of the detection of B. pertussis and the test can remain positive for a longer period of time compared to other methods. The sensitivity also allows for contamination to occur, which may often lead to false positives (15). This could have led to the documented outbreaks of pertussis, if many tests were conducted in the same laboratory. Also, until recently, the Canadian policy of diagnosis of pertussis stated that a positive test on a PCR was enough to diagnose an individual with pertussis, even without a clinical confirmation (15).
On the other hand, PCR could be a helpful tool in detecting certain cases of pertussis that would have been undetectable by standard culture, which could include milder or partially treated cases (11).
The cohort effect of pertussis
In the 1990s, Canada experienced a significant increase of infected, infant pertussis individuals. The cause of this outbreak was not due to unvaccination, as 95% of the infected infants have had three or more vaccinations (3). Once vaccinated, an individual may be protected from B. pertussis, but that protection may decrease through evolution of the bacteria, and thus there was a waning of the vaccination. As that protection decreases, the infect population that was once protected may be vulnerable. Therefore, that population can be infected again and the average age group that is infected increases as it ages (3).
The low efficacy of the whole cell pertussis vaccination may have led to a whole infant population that was vulnerable to B. pertussis. Despite the vaccination advances, there have still be a large group of individuals that have stemmed from the 1985 outbreak that continues to be effected (3).
The assumption that pertussis is more common in infants is a trend that has been noticed throughout the years of research on B. pertussis. This has led to more attention given to infants and are increasingly tested for pertussis. However, this results in misdiagnosis of pertussis in older individuals (3). Also, it is generally known that sensitivity of diagnosis is higher in infants than in older individuals, which makes testing more convenient and easier for infants. There may be many cases of undocumented pertussis cases due to the lack of attention given to pertussis symptoms in the older population (2).
What is being done to address this problem
Use of acellular pertussis vaccine
In 1996, PMC Canada released the "Five Component Pertussis" vaccine as part of the DTaP (Diptheria, Tetanus, acellular Pertussis) vaccine. This new vaccine updated the original that was developed in 1930's, replacing the old whole-cell design with new acellular components comprising of FHA (filamentous hemaglutinin), 69K (69KDA) protein, or FIM (fimbriae) (17). By not using the whole cell, the DTaP vaccine proves to be a safe alternative. The acellular pertussis vaccine produces a longer lasting immunity, reduced side effects, reduced number or required injections. However, due to the nature of the vaccine, it cannot be combined for injection with Haemophilus Influenzae type B or polio vaccinations (18). The vaccination is given on an immunization schedule of 2 months, 4 months, 6 months. Additional booster doses are administered at 18 months of age, and the final shot is given around 4-6 years of age (19).
Use of Antibiotics
A variety of antibiotics may be used to combat pertussis, including erythromycin estolate, erythromycin ethylsuccinate, azithromycin, clarithromycin. They may be applied before the paroxysmal phase, or given as a preventative to individuals in close contact with a person affected by pertussis. The standard antibiotic treatment for pertussis is a 14 day course of 15-20 mg/kg of erythromycin per day. Alternative antibiotics are offered in substitution of erythromycin for patients unable to handle erythromycin. If applied before the paroxysmal stage, antibiotics often reduce the severity and duration of the symptoms. If applied after the paroxysmal stage, the antibiotics are effective at eradicating the disease, but clinical studies offer varying opinions of its efficacy against the symptoms. Some studies report a reduction in symptoms while others state the antibiotics have no discernible effect. Erythromycin cannot be used for children under the age of 14 days. Clinical studies have shown that the use of erythromycin in infants can be associated with pyloric stenosis (21).
What could be done to improve the situation?
With the advent of the acellular pertussis vaccine, pertussis can be prevented in individuals that have been immunized with the vaccine. Pertussis still poses a problem for infants as the vaccine cannot be given until 2 months of age. That 2 month window can be slightly ameliorated with the use of antibiotics as a preventative, but antibiotics cannot be used before the age of 2 weeks. This phase before the vaccine can be used is the window of opportunity for pertussis. Future advances in pertussis prevention should be directed towards counteracting the vulnerability of infants.
Detection of pertussis
As testing with PCR has led to possibilities of misreporting of the number of cases of pertussis in Canada, a more conclusive way of determining the true number of reported cases of pertussis must be applied. Until recently, the governmental policy for determining the effect of pertussis was the PCR testing, even without clinical testing (8). Now that clinical testing is required to report a case of pertussis, it would be even more ideal to have other methods of testing to guarantee that the individual is suffering from pertussis to accurately describe its virulence upon the population.
As the disease “whooping cough” could be attributed to the bacterium Bordetella pertussis, it is actually the toxins, rather than the presence of the bacteria, that accounts for the various symptoms associated with whooping cough. The effect that it has had on Canadians has been ongoing, despite vaccination development.
The research done on B. pertussis can be quite complicated, as the problem of polymorphism has been associated with all bacteria. However, the research for better vaccinations or other methods of resistance must not ever halt. At times, the disease may seem eradicated, but it may sporadically become a nuance once again. Research on methods of detection, vaccination, or antibiotics can always be improved and should never be ignored.
1. R Benz, E Maier, D Ladant, A Ullmann and P Sebo; Adenylate cyclase toxin (CyaA) of Bordetella pertussis. Evidence for the formation of small ion-permeable channels and comparison with HlyA of Escherichia coli. http://www.jbc.org/cgi/content/abstract/269/44/27231
2. Char, Ian; Chatfield, Steven; Douce, Gillian; Everest, Paul; Li, Jingli; “Role of the Bordetella pertussis P.69lpertactin protein and the P.69/pertactin RGD motif in the adherence to and invasion of mammalian cells”; http://mic.sgmjournals.org/cgi/reprint/142/11/326.
3. Cox, Michael M.; Nelson, David L.; Principles of Biochemistry. W.H. Freeman and Company, New York, NY.
4. Fricks IP, Carter RL, Lazarowski ER, Harden TK. Gi-dependent cell signaling responses of the human P2Y14 receptor in model cell systems. http://www.ncbi.nlm.nih.gov/pubmed/19339661?ordinalpos=9&itool=EntrezSystem2.PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_DefaultReportPanel.Pubmed_RVDocSum
5. Guermonprez P,et al. The Adenylate Cyclase Toxin of Bordetella pertussis Binds to Target Cells via the aMb2Integrin (CD11b/CD18); http://jem.rupress.org/cgi/reprint/193/9/1035
6. N/A; Pertussis; http://www.cdc.gov/ncidod/dbmd/diseaseinfo/pertussis_t.htm
7. Tang WJ, Guo Q; The adenylyl cyclase activity of anthrax edema factor. http://www.ncbi.nlm.nih.gov/pubmed/19560485?ordinalpos=11&itool=EntrezSystem2.PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_DefaultReportPanel.Pubmed_RVDocSum
8. "Statement on Pertussis Vaccine." Public Health Agency of Canada: Canada Communicable Disease Report Vol. 23 (ACS-3). 15 July 1997. 19 Aug 2009 < http://www.phac-aspc.gc.ca/publicat/ccdr-rmtc/97vol23/23sup/acs3.html>.
9. Ewanowich C, Chui L, Paranchych M et al. Major Outbreak of Pertussis in Northern ALberts, Canada: Analysis of Discrepant Direct Fluorescent-Antibody and Cuture Results by Using Polymerase Chain Reaction Methology. Journal of Clinical Microbiology. July 1993.
10.Ntzeayabo B, De Serres G, Duval B et al. Pertussis resurgence in Canada largely caused by a cohort effect. Social and Community Medicine. Social and Community Medicine.
11.Lambert HJ. Epidemiology of a small pertussis outbreak in Kent County, Michigan. Public Health Rep 1965
12. Decker MD, Edwards KM, Steinhoff MC et al. Comparison of 13 acellular pertussis vaccines: adverse reactions. Pediatrics 1995.
13.Gustafsson L, Hallander HO, Olin P et al. A controlled trial of a two-component acellular, a five-component acellular, and a whole-cell pertussis vaccine. N Engl J Med 1996.
14.Skowronsky D, De Serres G, MacDonald D et al. The Changing Age and Seasonal Profile of PErtussis in Canada. J Infectious Disease 2002.
15.Waters V, Jamieson F, Richardson S et al. Outbreak of Astypical Pertussis Detected by Polymerase Chain Reaction in Immunized Preschool-Aged Children. J Infectious Disease.
16.“Canadian Immunization Guide. Seventh Edition – 2006.” Public Healthy Agency of Canada. 19 Aug 2009. <http://www.phac-aspc.gc.ca/publicat/cig-gci/index-eng.ph> .
17. "Connaught and The History of Modern Pertussis Vaccines" Article originally published in Conntact, Vol. 9 #5 (November 1996), p. 4 http://www.healthheritageresearch.com/Pertussis/Pertussis-CONN-9611.pdf
18. Halperin SA: Canadian Journal Infectious Diseases 7: 6; 359-60, Dec. 1996. http://www.utoronto.ca/kids/acelular.html
19. Canada Communicable Disease Report Volume 29, ACS-5 1 September 2003 “Prevention of Pertussis in Adolescents and Adults”
21. The Pediatric Infectious Disease Journal Vol 24 Number 5 may 2005 Use of Antibiotics in the Prevention and Treatment of Pertussis Carl-Heinz Wirsing von Koenig, MD
22. "Pertussis vaccination." Online image. Department of Health and Human Services: Center for Disease Control and Prevention. 27 Aug 2009. <http://www.cdc.gov/vaccines/vpd-vac/pertussis/default.htm>.
23. "Bordetella pertussis." Online image. DPT - Difteria/Tetano/Coquelunche. 27 Aug 2009. <http://www.virtual.epm.br/material/tis/curr-bio/trab2001/grupo5/triplice.htm>.
Edited by Bryan Dieffenbach, Gorjan Hrustanovic, Patricia Lee and Andrew Chen, students of Rachel Larsen
Figure 1: Bordetella pertussis (23)
Figure 3: Vaccinations are often performed on adolescents (22)
Citation: Diavatopoulos DA, Cummings CA, Schouls LM, Brinig MM, Relman DA, Mooi FR (2005) Bordetella pertussis, the Causative Agent of Whooping Cough, Evolved from a Distinct, Human-Associated Lineage of B. bronchiseptica. PLoS Pathog 1(4): e45. https://doi.org/10.1371/journal.ppat.0010045
Editor: Scott Hultgren, Washington University in St. Louis, United States of America
Received: August 23, 2005; Accepted: November 28, 2005; Published: December 30, 2005
Copyright: © 2005 Diavatopoulos et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: CGH, comparative genomic hybridization; ISE, insertion sequence element; LCA, last common ancestor; LPS, lipopolysaccharide; mAb, monoclonal antibody; Mya, million years ago; MLST, multilocus sequence typing; MST, minimum spanning tree; ST, sequence type
The genus Bordetella is composed of several species, of which three are exclusively respiratory pathogens of mammalian hosts: B. bronchiseptica, B. pertussis, and B. parapertussis (henceforth referred to as the mammalian bordetellae). B. bronchiseptica causes chronic and often asymptomatic respiratory tract infections in a wide variety of mammals. It is only sporadically isolated from humans [1,2], particularly from immunocompromised individuals, and human infections have been considered to be zoonotic . B. parapertussis consists of two distinct lineages: one found in humans and the other found in sheep (B. parapertussishu and B. parapertussisov respectively) . B. pertussis and B. parapertussishu have been isolated only from humans and cause acute, transient infections and disease, designated whooping cough or pertussis. Pertussis is especially severe in young, unvaccinated children and has reemerged in recent years in vaccinated populations [5–7]. Previous research indicated that B. pertussis and B. parapertussishu independently evolved from a B. bronchiseptica–like ancestor [8.9], and comparison of the genomes of the three isolates chosen for sequencing suggested that the time to the last common ancestor (LCA) for B. pertussis and B. parapertussishu and for B. bronchiseptica was 0.7 to 3.5 and 0.27 to 1.4 million years, respectively . Despite their different host tropisms, the mammalian bordetellae are very closely related [8,9]. Analysis of their genome sequences revealed that the adaptation of B. pertussis and B. parapertussishu to the human host was accompanied by extensive genome decay .
Their differences in host tropism in contrast to their close genetic relationships make the mammalian bordetellae attractive candidates for the study of host-adaptation. Such studies are facilitated by the availability of genome sequences of B. bronchiseptica, B. pertussis, and B. parapertussishu . So far, only a single representative of each species has been sequenced, and it is important to determine their relationships to the Bordetella population as a whole. To that end and to identify genetic events that may be associated with host adaptation, we used a combination of multilocus sequence typing (MLST) , comparative genomic hybridization (CGH) with whole-genome microarrays , and the distribution of several insertion sequence elements (ISEs) to characterize 132 mammalian Bordetella strains with diverse host associations. This work identified two B. bronchiseptica lineages, the first of which is composed of mainly strains of animal origin and includes the B. bronchiseptica strain from which the genome sequence has been determined. The second lineage, comprising strains mainly of human origin, is more closely related to B. pertussis than the first lineage. Comparison of the two B. bronchiseptica lineages to B. pertussis revealed genetic differences that may be associated with adaptation to the human host.
Population Structure of the Mammalian Bordetellae, Based on Multilocus Sequence Typing
To determine the relationships between the mammalian bordetellae, we determined the partial sequences of seven housekeeping genes from 132 strains (Table S1 and http://pubmlst.org/bordetella). We observed 32 sequence types (STs) among the 132 Bordetella isolates. Allele segments were divided into five equally sized subloci, and a minimum spanning tree (MST) algorithm was used to cluster the subloci . Complexes were defined as groups of strains differing at fewer than five of 35 subloci with a minimum of two STs per complex. Using this criterion, strains could be assigned to one of four complexes, designated complexes I through IV (Figure 1).
Figure 1. Minimum Spanning Tree of B. bronchiseptica, B. pertussis, and B. parapertussis
The tree was based on the sequence of seven housekeeping genes. Individual genes were split into five subloci, and a categorical clustering was performed. In the minimum spanning tree, sequence types sharing the highest number of single locus variants were connected first. Each circle represents a sequence type (ST) the size of which is related to the number of isolates within that particular ST. Colors within circles indicate host distribution. The numbers between connected STs represent the number of different subloci between those STs. The clonal complexes (I, II, III, and IV) are indicated by colored strips between connected STs. ST16 (B. bronchiseptica complex I) harbors the B. parapertussisov strains. STs containing strains of which the genome has been sequenced (B. pertussis Tohama, B. parapertussis 12822 or B. bronchiseptica RB50) are indicated by a thickset, dashed line. The distribution of the insertion sequence elements IS481, IS1001, IS1002, and IS1663 is shown in boxes (see also Table S1); numbers between parentheses indicate the percentage of strains that contained the ISE as determined by PCR amplification. The divergence times between B. bronchiseptica complexes I and IV and B. pertussis complex II are shown.
Complexes II and III contained the B. pertussis and B. parapertussishu isolates, respectively. Both of these complexes showed very limited genetic diversity (H = 0.65 and 0.35, respectively), as described previously [8,9]. B. bronchiseptica was divided into two distinct populations, designated complexes I and IV, respectively. The genetic diversity of these two complexes (H = 2.16 and 2.45, respectively) was much higher than that of complexes II and III. Complex I contained the majority of the B. bronchiseptica strains (76 of 91 strains), including the sequenced RB50 strain. In addition, it contained the B. parapertussisov isolates in the study population (ST16). B. bronchiseptica complex IV was more closely related to B. pertussis than was B. bronchiseptica complex I. Furthermore, the host species associations of the two complexes were quite distinct. Of the B. bronchiseptica complex IV isolates, 80% were isolated from humans, while this was the case for only 32% of the complex I isolates. It should be noted that human B. bronchiseptica isolates were overrepresented in our strain collection, in comparison with their occurrence in natural populations. However, the human complex IV isolates originated from different continents, comprising North America, South America, and Europe. Previously, phylogenetic analysis based on CGH suggested the existence of a distinct B. bronchiseptica lineage that was closely related to B. pertussis .
The relationships among the mammalian bordetellae inferred from housekeeping gene sequences were confirmed by an analysis based on the pertactin gene (prn), which codes for a surface-associated virulence factor involved in adherence [14,15]. A UPGMA tree was constructed from the aligned prn sequence data, and the topology of this tree was very similar to the MLST tree (Figure 2). B. bronchiseptica strains are grouped into two lineages, corresponding to complex I and IV in the MLST tree. Also, B. bronchiseptica complex IV and B. pertussis strains clustered together in one branch, which was supported by bootstrapping. B. parapertussishu comprised a separate branch within a larger cluster that also contained the B. bronchiseptica complex I strains. The B. parapertussisov strains were indistinguishable from B. bronchiseptica complex I strains, as was observed in the MLST tree.
Figure 2. UPGMA Tree Based on the Analysis of the Pertactin Gene of Bordetella Isolates Used in the MLST Analysis
The DNA segment coding for the extracellular domain of pertactin (P.69) was used for analysis, with the exclusion of the repeat regions 1 and 2. Bootstrap values are shown for the nodes separating the complexes and are based on 500 bootstrap replicates. The scale indicates the genetic distance along the branches. Colors of the branches indicate the four complexes defined by MLST. The number of strains of each branch is shown in boxes, as well as the host distribution.
Distribution of Insertion Sequence Elements
The distribution of ISEs has been used to reveal evolutionary relationships among the Bordetella population . Toward this end, we screened our strain collection for the presence of IS481, IS1001, IS1002, and IS1663 using PCR (Table S1). The distribution of the ISEs was mapped onto the MST (see Figure 1). IS481 was detected in all B. pertussis strains but not in any other species, with the exception of two B. bronchiseptica isolates, both from a horse (B1975, B0230, ST6), consistent with previous observations [9,16]. IS1001 was detected in all B. parapertussishu and B. parapertussisov strains. Additionally, IS1001 was detected in most (21 of 25) B. bronchiseptica strains belonging to ST7 in complex I, but not in other STs, including STs in complex II and IV. IS1002 was detected only in B. pertussis and in B. parapertussishu strains, confirming previous observations [4,9]. IS1663  was detected in all B. pertussis isolates but also in ten of 13 B. bronchiseptica complex IV strains. The three complex IV strains in which IS1663 was not detected belonged to STs 18 and 21.
Divergence Times of Complexes
Under the assumption that the mutation rate in prokaryotes is relatively constant, the time since descent from the LCA can be estimated using pairwise mean allele distances (KS) [17,18]. The clock rates described by Whittam  and by Guttman and Dykhuizen  were used to estimate a range of divergence times between complexes. Calculations indicated that B. pertussis and B. bronchiseptica complex IV separated approximately 0.3 to 2.5 million years ago (Mya), which suggests a more recent divergence time than B. pertussis and B. bronchiseptica complex I, estimated at 1.1 to 5.6 Mya. B. parapertussishu and B. bronchiseptica complex I diverged between 0.7 and 3.5 Mya according to our calculations. The divergence times of combinations of complexes is shown in Figure 1.
Gene Content of Strains from B. bronchiseptica Complexes I and IV
The high percentage of complex IV strains of human origin compared to the percentage of those in complex I suggested a preference for the human host in complex IV strains. To identify genetic events that may have played a role in host adaptation or host restriction, we used CGH with Bordetella DNA microarrays [17,18].
Genomic DNA from 26 B. bronchiseptica complex I and 13 complex IV strains was hybridized to microarrays (CGH data files have been deposited in ArrayExpress, accession E-TABM-32). Significance Analysis of Microarrays (SAM) was used to identify probes with statistically significant log intensity ratio differences between the two complexes (Table S2). This approach detected sequences that have been deleted more often in one of the two complexes, as well as DNA sequences diverging from the reference sequences in one of the two complexes.
Thirty-one probes, representing 29 genes and an ISE, hybridized more strongly or more frequently to the genomes of complex IV strains than to those of complex I. Two of these probes represented the B. pertussis homologs of the virulence genes prn and tcfA, encoding pertactin and tracheal colonization factor, respectively [15,21] (Figure 3). Sequence analysis confirmed that the B. bronchiseptica complex IV prn sequences were more similar to those of B. pertussis than to those of the B. bronchiseptica complex I (see Figure 2).
Figure 3. Gene Content of the Differentially Hybridizing Virulence Loci between B. bronchiseptica Complex I and IV, as Determined by CGH
Each column represents one strain. Strain numbers and STs are indicated above the columns. Each row represents one ORF (in B. bronchiseptica RB50 gene order), ORF designations are shown to the right of the rows. In the case of tcfA and prn, the origins of the probes are indicated between parentheses. The BP probe of tcfA was 100% similar to B. pertussis Tohama and 85.1% similar to B. bronchiseptica RB50. The BP prn probe was 100% similar to B. pertussis Tohama and 86% similar to B. parapertussis 12822 and B. bronchiseptica RB50. The BB/BPP prn probes were both 100% similar to B. parapertussis 12822 (BPP) and B. bronchiseptica RB50 (BB) and 86% similar to B. pertussis Tohama. The yellow-black-blue color scale indicates the hybridization value relative to the reference; references are B. bronchiseptica RB50, B. parapertussis 12822, and B. pertussis Tohama. For B. bronchiseptica RB50 and B. pertussis Tohama, the data in the figure are based on the genomic sequences. Yellow indicates decreased hybridization, black indicates hybridization values comparative to the references, and blue indicates gene duplications. Intermediate values indicate partial deletions or sequence divergence. Missing data are represented in gray.
Sixteen of these 31 probes had been identified previously as B. pertussis–specific . In most cases, these probes hybridized to nine or more complex IV strains but not to any complex I or III strain. The genes represented by these probes encode diverse functions involved in metabolism, transport, regulation, and transposition (Table S2). With the possible exception of BP0703, which encodes a TonB-dependent iron receptor, no obvious virulence genes were observed among them. Most of the 16 genes were located in small clusters along the B. pertussis Tohama chromosome. The presence of these genes in B. bronchiseptica complex IV and B. pertussis but not in B. bronchiseptica complex I or B. parapertussishu suggests that they were acquired by the common ancestor of B. pertussis and B. bronchiseptica complex IV.
We also identified 248 probes, representing 237 genes, that exhibited significantly stronger hybridization to complex I genomes than to complex IV genomes (Table S2). Sixty-eight (27%) corresponded to genes associated with mobile elements such as prophages, while many of the other probes represented genes involved in metabolic, transport, and regulatory functions. Surprisingly, several virulence-associated genes were found to be missing or divergent in the complex IV strains, as compared to complex I strains. These included the B. bronchiseptica homologs of tcfA and prn, Bvg-regulated intermediate phase gene A (bipA), the alcaligin biosynthesis locus (alcA/E), the pertussis toxin synthesis and transport locus (ptx/ptl), the dermonecrotic toxin gene (dnt), and the lipopolysaccharide (LPS) biosynthetic locus (Figures 3 and 4).
Figure 4. Expression of LPS by B. bronchiseptica Complex I and Complex IV Strains and Gene Content Variation at the LPS Biosynthesis Locus
(A) Top: Electrophoretic LPS profiles obtained by tricine-SDS-PAGE and silver staining. Middle: Western blot of the same samples with mAb 36G3, which detects band A. Bottom: Western blot of the same samples with mAb BL8, which detects band B.
(B) Gene content of the LPS biosynthesis locus as determined by CGH. See Figure 3 for details. For B. bronchiseptica RB50 and B. pertussis Tohama, the data in the figure are based on the genomic sequences. The genes wbmPQRSTU represent an alternative LPS O-antigen biosynthesis sublocus that is orthologous to the genes found in B. parapertussis 12822  and B. bronchiseptica C7635E . LPS genetic profiles as described in the text are indicated at the top of the columns. Color scale as in Figure 3. Missing data are represented in gray.
Interestingly, ten of 13 complex IV strains harbored deletions in the ptx and ptl loci, which encode pertussis toxin (Ptx) and its secretion machinery, respectively (Figure 3) [22,23]. While conditions under which Ptx is expressed by either B. parapertussis or B. bronchiseptica have not been found, the structural genes are generally conserved among these species , suggesting selective pressure to retain the ability to produce functional Ptx under certain circumstances. The complex IV strains in which ptx/ptl was still present (ST3/17/29) were tested for expression of Ptx by immunoblotting, and these strains were found not to express Ptx under the growth conditions used (unpublished data). Another distinguishing characteristic of the complex IV strains was the deletion of the dermonecrotic toxin gene (dnt) in 8 of 13 strains. In contrast, this gene was detected in all B. pertussis, B. parapertussishu, B. parapertussisov, and B. bronchiseptica complex I strains.
LPS Genetic and Structural Diversity in B. bronchiseptica
The genetic structure of the LPS biosynthesis locus differed between complex I and IV strains. The LPS molecules of Gram-negative bacteria usually consist of three, covalently linked, major domains: the lipid A, the branched chain oligosaccharide core, and the hydrophilic O-antigen. A number of genetic loci have been implicated in the synthesis of these domains in Bordetella, such as the lpx locus (lipid A), the waa locus (inner core), the wlb locus (outer core), and the wbm locus (O-antigen) [25–29].
B. pertussis LPS usually consists of lipid A and an inner core, to which the outer core (a trisaccharide) is attached; this is also referred to as band A. Certain B. pertussis strains produce only band B, which is identical to band A except that it lacks the trisaccharide . The O-antigen, which is added to the trisaccharide, is found only in B. bronchiseptica and B. parapertussis. This structure is missing in B. pertussis due to the deletion of the genes wbmA-U . In Figure 4, the gene content of the LPS locus is shown for 13 complex I and 13 complex IV strains and for B. pertussis Tohama and B. bronchiseptica RB50. Four LPS gene content profiles, designated LPS 1–4, could be distinguished among the B. bronchiseptica strains. Strains with the LPS 1 profile had an LPS gene composition similar to RB50, characterized by the absence of wbmPQRSTU. The LPS 2 profile was characterized by the absence of the genes wbmORS and BB0124–BB0127 and the presence of an alternative O-antigen locus, comprising wbmPQRSTU, orthologous to the B. parapertussis 12822 genes . Both of these genotypes appear competent for the production of a full length LPS. Strains with the LPS 3 profile lacked wbmD-U and BB0124-BB0127, suggesting that they may not produce O-antigen. The LPS 4 profile was similar to the LPS 3 profile but additionally lacked wbmABC and wlbD-L, suggesting that strains of this genotype may be deficient for the production of trisaccharide as well as O-antigen. The deletion in the O-antigen genes of LPS 4 strains was similar to that observed in the O-antigen genes of B. pertussis Tohama. All complex I strains displayed either an LPS 1 or an LPS 2 profile. Nine of 13 complex IV strains had either an LPS 1 or an LPS 2 profile, while four complex IV strains showed more extensive deletions, resulting in LPS 3 and LPS 4 profiles.
To study the effect of these deletions on LPS production, proteinase K–treated cell lysates were analyzed by Tricine-SDS-PAGE, followed by silver staining or immunoblotting with monoclonal antibodies (mAbs) directed against either band A or band B (mAbs 36G3 and BL-8, respectively [31,32]) (Figure 4). Silver-staining showed that all complex I strains produced band A, except for B1985, which produced band B, and B2112, which produced a band migrating at a position between bands A and B. These strains showed no obvious deletions at their wlb or wbm loci, and the fact that they did not produce band A and O-antigen may be attributed to point mutations, e.g., in their wlb locus, or to regulatory effects. These results were confirmed by immunoblotting with mAb 36G3. The epitope in the trisaccharide that is recognized by this mAb was also present in the O-antigenic repeats, as was described previously . In general, most strains that produced band A also produced an additional band just above band A. This extra band was also recognized by 36G3 and therefore is likely derived from band A.
The nine complex IV strains with LPS 1 or 2 profiles all produced band A. The two strains with the LPS 4 profile, B2490 and B2506, produced a band smaller than band A, which failed to be recognized by mAb 36G3. Of the two strains with an LPS 3 profile, one strain (B0243) produced only band B and no O-antigen. Unexpectedly, the other strain, B2494, produced band A and O-antigen as detected by immunoblotting, indicating that this strain contains as yet uncharacterized O-antigen biosynthesis genes. Silver staining also suggested that the LPS 4 strains may produce O-antigen, although the O-antigen failed to be recognized by mAb 36G3.
Although it has long been speculated that B. pertussis evolved from a B. bronchiseptica strain [8,9], a specific lineage has not been identified. Here we identify and characterize such a B. bronchiseptica lineage. Analysis of MLST data from the mammalian bordetellae identified four distinct complexes. Complex I and IV comprised B. bronchiseptica strains, while complex II and III comprised the human pathogens B. pertussis and B. parapertussishu, respectively. Our results suggest that B. pertussis and B. parapertussishu evolved from complexes I and IV, respectively, indicating that adaptation to humans occurred as two independent events, consistent with previous data [9,10].
The population structure of the mammalian bordetellae inferred from MLST data largely corresponded with a maximum parsimony phylogeny derived from a previous CGH study , with the exception of the relationship of B. parapertussisov and B. parapertussishu. In the current study, B. parapertussishu and B. parapertussisov are clearly derived from different STs in complex I. Further, in contrast to B. parapertussishu, B. parapertussisov is actually part of B. bronchiseptica complex I. The closer relationship between the sheep- and human-derived B. parapertussis lineages that was inferred from CGH data may be an artifact of long-branch attraction in the maximum parsimony tree .
Consistent with previous studies [8,9,35], B. pertussis and B. parapertussishu showed a relatively low degree of genetic diversity, suggesting that they evolved recently or encountered a recent evolutionary bottleneck. Of the three B. pertussis STs observed, two were found exclusively before 1960, whereas all modern strains belong to ST2. The temporal shift in B. pertussis STs is consistent with our previous studies on antigenic shifts that show major changes in the B. pertussis population after the introduction of mass vaccination against pertussis in the 1950s and 1960s [35,36].
Most human disease is by far caused by B. pertussis, and we therefore focused on the relationship of B. bronchiseptica complex IV with B. pertussis. B. bronchiseptica complex IV strains were found to be more closely related to B. pertussis than to the complex I B. bronchiseptica strains. A tree based on prn nucleotide sequences also suggested a closer relationship of complex IV strains to B. pertussis than to complex I strains. A number of other features of complex IV strains were consistent with their close relationship with B. pertussis. Most complex IV strains were isolated from humans (80%), while the majority of complex I strains were of animal origin (68%). Almost all B. bronchiseptica complex IV strains were isolated from patients with whooping cough symptoms. Further, complex IV strains and B. pertussis shared an IS element, IS1663, that was not found outside these two lineages. The sharing of an IS element may be explained by either vertical or horizontal transfer. The former suggests a common ancestry, while the latter would point to niche sharing of B. pertussis and B. bronchiseptica complex IV. It seems unlikely that the association of complex IV strains with humans is due to a sampling artifact, as the strains analyzed were from widely separated geographic regions, including North America, South America, and Europe. Thus, these strains were not epidemiologically related.
Three STs (ST12, ST23, and ST27) found in complex I also contained a high percentage of human strains (55%, 43%, and 77%, respectively). All other, nonhuman isolates of these STs were collected from domesticated animals. Thus, both complex I and IV contain B. bronchiseptica strains that are well adapted to the human host. However, the particular relevance of the human-associated lineage in complex IV appears to be its evolutionary relationship with B. pertussis. The high frequency of human isolates observed in complex IV may be due to the close interaction of humans with the animal hosts in which these strains reside or to the fact that complex IV strains are better adapted to a human environment than B. bronchiseptica complex I strains. In either case, the B. bronchiseptica complex IV infections of humans would be zoonotic. Another intriguing possibility is that B. bronchiseptica complex IV strains are to a large extent adapted to the human host and primarily transmitted between humans.
Microarray-based CGH revealed 29 genes and IS1663 to be more frequently present in, or more similar to, B. pertussis orthologs in B. bronchiseptica complex IV than to complex I strains. Of these genes, 16 were unique to B. pertussis and B. bronchiseptica complex IV, suggesting they were acquired after the common ancestor of complex IV and B. pertussis diverged from complex I. With the exception of prn and tcfA, which hybridized more strongly to the B. pertussis–derived probe, no known virulence genes were identified among these 30 genes. Conversely, 237 genes were absent or divergent in complex IV compared to complex I strains, suggesting that the B. bronchiseptica complex IV genome is decaying, as has been assumed for B. pertussis and B. parapertussishu. Genome decay has been associated with host restriction or niche change in a number of pathogens such as Yersinia pestis [37,38] and Burkholderia mallei  and has been suggested to be a driving force of host restriction for B. pertussis and B. parapertussis as well . Likewise, the apparent preference of complex IV strains for humans may also be associated with genome decay. Because putative complex IV–specific sequences were not represented on the microarray used here, we were unable to address the possibility that complex IV strains have acquired, through lateral transfer, genetic loci that may have promoted host preference. However, gene acquisition appears to have been a rare event in the evolution of B. pertussis and B. parapertussis from B. bronchiseptica complex I .
Differences between complex IV strains and B. pertussis were observed with respect to three major virulence factors, Ptx, Dnt, and LPS. All B. bronchiseptica complex I strains examined contained intact Ptx genes. Although conditions under which the Ptx genes are expressed in B. bronchiseptica have not been identified, their conservation suggests that they may confer a selective advantage in the ecology of complex I strains. In contrast to complex I strains, the genes encoding Ptx and its secretion apparatus were deleted from most complex IV strains (10 of 13). In the strains that did retain the Ptx locus, no in vitro expression of Ptx was observed, even though these strains were closely related to B. pertussis.
Another characteristic that sets complex IV strains apart from all other mammalian bordetellae is that in eight of 13 strains, dnt was deleted. Dnt is an intracellular toxin that activates the small GTPase Rho through deamidation or polyamination . It has been shown that Dnt is important for turbinate atrophy and the colonization of the upper respiratory tract by B. bronchiseptica in pigs .
Based on CGH, four LPS genetic profiles were distinguished. The LPS genetic locus was generally more polymorphic in complex IV strains than in complex I strains, and deletions were observed in the O-antigen and trisaccharide biosynthesis genes in some complex IV strains. In complex IV strains with the LPS 4 profile, the extent of deletion in the O-antigen genes was very similar to that seen in B. pertussis Tohama. B. pertussis does not produce repetitive O-antigen as it lacks the wbm genes but makes a lipo-oligosaccharide that consists of lipid A to which a single trisaccharide is attached . Like B. pertussis, four complex IV strains lacked the O-antigen genes known to be present in the sequenced genomes of B. bronchiseptica and B. parapertussishu. Despite the absence of these genes, in at least one of these strains O-antigen was detected by immunoblotting, suggesting that this strain carries LPS genes distinct from those in RB50 or 12822. The two LPS 4 strains, both isolated from humans, also lacked the genes required for biosynthesis of the trisaccharide and failed to produce trisaccharide detectable by immunoblotting. The absence of the trisaccharide is intriguing in view of the fact that it was found to be otherwise conserved in all Bordetella strains analyzed.
It seems likely that in addition to gene loss and acquisition, differences in gene regulation have significantly contributed to host adaptation . The differences observed between complex IV strains and B. pertussis, particularly with respect to Ptx, Dnt, and LPS, may be due to differences in niches occupied. Another possibility is that these differences have arisen in response to immune competition between B. bronchiseptica complex IV strains and B. pertussis. Gupta and co-workers  provided evidence that immunodominant surface antigens are organized into nonoverlapping combinations as a result of selection by the host immune system. This process could also have driven the inactivation and conservation of virulence factors in mammalian bordetellae infecting the same host. In a similar vein, Bjørnstad and Harvill  hypothesized that, since B. pertussis and B. parapertussishu both infect humans, they may have evolved to evade cross-immunity by the other pathogen. The authors propose that immune competition provides an explanation for differences observed between B. pertussis and B. parapertussishu. For example, B. pertussis but not B. parapertussishu expresses Ptx, although both contain the required genes. Conversely, B. parapertussishu expresses O-antigen, while the corresponding genes have been deleted from B. pertussis. Similarly, the deletion of the genes for Ptx, Dnt, and genes involved in trisaccharide syntheses by complex IV strains may have been driven by immune competition with B. pertussis and possibly also with B. parapertussishu.
The origin of the disease whooping cough is still a mystery. Although the disease has very typical symptoms in children and was one of the major causes of child mortality before the introduction of vaccination, the first written reference to the disease in Europe is found in 1540 . The first description of an epidemic, which occurred in Paris, was given by Baillon in 1578 . Particularly interesting are the observations made by Nils Rosen von Rosenstein in 1766, who wrote , “The hooping cough never appeared in Europe originally, but was transported thither from other parts of the world by means of merchandise, seamen and animals. Its first appearance in Sweden cannot be determined with any certainty; but in France it began in the year 1414.” In contrast, 16th- and 17th-century descriptions of the disease and epidemics in Europe are documented frequently in the literature . The absence of references to pertussis-like symptoms in the ancient literature has been taken as evidence that the association of B. pertussis with humans is of recent origin.
We propose that the association of B. pertussis with humans is, in fact, ancient but that the introduction of B. pertussis into Europe may be more recent. Complex IV strains showed a degree of diversity that was comparable to complex I strains (2.16 and 2.45, respectively), and thus, assuming that complex IV strains are primarily adapted to the human host, this association must be ancient. Parkhill et al.  previously estimated the time to the LCA of a B. bronchiseptica complex I strain (RB50) and B. pertussis to be 0.7 to 3.5 Mya, based on the mean number of synonymous substitutions per synonymous site of orthologous gene pairs. Our data indicate that current B. pertussis strains expanded clonally from the B. pertussis–B. bronchiseptica complex IV LCA 0.32 to 2.53 Mya, further supporting an ancient association of B. pertussis with humans. However, we cannot rule out the possibility that more recent human-associated ancestors of B. pertussis are extinct or undiscovered. Such recent ancestors would indicate a more recent origin of B. pertussis.
Although it is tempting to speculate that the LCA of B. pertussis and B. bronchiseptica complex IV was associated with humans, the possibility remains that this association emerged after the split with B. pertussis. A possible evolutionary scenario (Figure 5) represents the adaptation of an ancestral B. bronchiseptica complex I strain to humans or their hominid ancestors. From this lineage, the LCA of B. bronchiseptica complex IV and B. pertussis evolved, subsequently giving rise to B. bronchiseptica complex IV and B. pertussis.
Figure 5. Model of the Evolution of the Mammalian Bordetellae
The bar on the left indicates increasing degrees of adaptation to the human host. Arrows indicate descent; double arrows between complexes indicate possible within-host immune competition. In boxes, genetic events are shown that may have played a role in speciation and niche adaptation. Numbers between parentheses refer to references. See text for details.
Recent emergence of a pathogenic clone from a more ancient human-associated progenitor species has been proposed as the mechanism for the origin of Mycobacterium tuberculosis . Although previous genetic analysis had suggested that M. tuberculosis emerged as little as 20,000 years ago, phylogenetic analysis of M. tuberculosis and a closely related but more diverse group of smooth tubercle bacilli indicated that this more broadly defined species has been associated with hominids for up to 3 million years.
Yersinia pestis, the causative agent of plague, is a clone that evolved from Y. pseudotuberculosis 1,500 to 20,000 years ago, shortly before the first known pandemics of human plague, and its recent origin is further suggested by the complete lack of polymorphism in housekeeping genes . Similarly, B. pertussis also shows limited diversity. However, in contrast to Y. pestis, which reveals absolutely no polymorphisms in housekeeping genes, we observed three STs in B. pertussis. This may suggest an older origin of B. pertussis compared to Y. pestis, although other factors, such as population size and bottlenecks, could also explain these differences. The most plausible explanation from our data is that the association of B. pertussis with humans originated in the LCA of B. pertussis and B. bronchiseptica complex IV. Based on that assumption, the apparent emergence of pertussis in Europe within the last 500 years may be attributable to import via travel or migration or to the recent acquisition by B. pertussis of the ability to cause more severe, whooping cough–like symptoms. Although most of the B. bronchiseptica complex IV strains in our collection were isolated from patients suspected to have pertussis, we know little of the severity of the symptoms caused by these strains. It is conceivable that B. bronchiseptica preceded B. pertussis in Europe and that its disease was not documented because of its relatively mild and nonspecific course.
The work presented here places the three sequenced mammalian Bordetella strains within a phylogenetic context, thereby facilitating rational selection of strains for further genomic sequencing. In particular, sequencing of one or more members of complex IV may shed more light on processes involved in host adaptation and immune competition. Further, the identification of a B. bronchiseptica lineage which circulates in human populations may be important for public health. In recent years, whole cell vaccines have been replaced by acellular vaccines comprised of one to five antigens derived from B. pertussis . The acellular vaccines induce a less cross-reactive immune response compared to whole cell vaccines  and may therefore result in an increase in B. parapertussis and B. bronchiseptica infections in vaccinated human populations.
Materials and Methods
A total of 132 Bordetella isolates were used in this study: 91 B. bronchiseptica, 9 B. parapertussishu, 3 B. parapertussisov, and 29 B. pertussis isolates (see Table S1). The three strains from which the genome sequence has been determined, B. bronchiseptica RB50, B. pertussis Tohama and B. parapertussis 12822 , were included. The collection included clinical isolates from humans and a broad range of animal species. Strains were grown on Bordet Gengou (BD, Franklin Lakes, New Jersey, United States) agar supplemented with 15% sheep blood at 37 °C for 2 to 5 days. Chromosomal DNA was isolated using the Wizard Genomic DNA Purification Kit (Promega, Madison, Wisconsin, United States), according to the manufacturers' protocol for Gram-negative bacteria.
The nucleotide sequences were determined for internal regions of seven housekeeping genes for all strains (http://pubmlst.org/bordetella ). The nucleotide sequence of the prn region encoding the extracellular domain of the surface-associated autotransporter pertactin, P.69, was determined for 116 strains, with the exclusion of the repeat regions 1 and 2 . These regions are comprised of amino acids repeats and are highly polymorphic due to insertion or deletion of the repeat unit. Primer information is listed in Table S3.
Detection of ISEs.
The distribution of IS481, IS1001, IS1002, and IS1663 was determined for all strains using PCR amplification (see Table S1). For PCR amplification of IS481, IS1001, and IS1002, primers were used that have been described previously [4,52]. Primer characteristics are listed in Table S3.
LPS SDS-PAGE and Western blotting.
BG-agar grown bacteria were harvested, boiled in 1× sample buffer (7.5% glycerol, 0.125 M Tris-HCl [pH 6.8], 1.5% SDS), and treated with proteinase K . Tricine-SDS-PAGE was then performed in 4% stacking and 16% separating gels, as previously described by Lesse et al. . Silver staining was performed as described by Tsai and Frasch . LPS was transferred to PVDF membranes (Amersham Biosciences, Buckinghamshire, United Kingdom) and blocked with 0.5% (w/v) Protifar nonfat dried milk, 0.5% bovine serum albumin (w/v), and 0.1% Tween 20 in PBS. Immunoblotting was performed with monoclonal antibodies 36G3 and BL-8, directed against band A and band B LPS, respectively [31,32].
Sequence data analysis.
Analysis of nucleotide sequence data was performed using Bionumerics software package version 4.0 beta 4 (Applied Maths, Sint-Martens-Latem, Belgium). The Bordetella MLST database can be accessed at http://pubmlst.org/bordetella .
For each locus in the MLST analysis, the allele sequences for all strains were trimmed to a uniform length, and an allele number was assigned to each unique allele sequence. The combination of the allele numbers at the seven loci defines the ST or allelic profile of each strain. Construction of trees based on allelic profiles may not accurately reflect the true genetic distance because both single and multiple nucleotide polymorphisms are given equal weight. Consequently, the degree of sequence difference between two alleles is not quantitatively reflected in the MLST profile. Conversely, tree construction based on concatenated allele sequences does not take into account the introduction of clustered multiple base substitutions due to a single recombinational event. As a result, trees based on MLST sequences often contain long branches, incorrectly suggesting a large genetic distance. Therefore, we used a method designated as split-MLST, in which each locus is split into a user-defined number of equally sized subloci (D. A. Diavatopoulos, P. Vauterin, L. Vauterin, F. R. Mooi, and L. M. Schouls, unpublished data). Using this method, the sensitivity of categorical clustering could be increased, without the perturbing effect of recombination. The topology of the tree appeared to vary if the number of subloci per MLST locus was lower than five. However, above the value four, increasing the number of subloci had no significant effect on the topology of the tree, and we therefore selected the lowest possible split value, five, resulting in a total of 35 subloci.
The genetic diversity for each complex was calculated using the Shannon-Weiner index of diversity (H) using the following formula:
where Pi is the frequency of the ith type .
For estimation of divergence times between complexes, we calculated the pairwise mean distance (Ks) between alleles using DNASP 4.00 . The divergence time was calculated using the following formula:
where Ks is the number of synonymous substitutions per synonymous site and r is the molecular clock rate of Escherichia coli as determined by Whittam  or by Guttman and Dykhuizen . We used these two rates to calculate a range of divergence times. The divergence time was first calculated for each combination of STs between complexes, and from these the averaged age between complexes was calculated.
Comparative genomic hybridization.
The preparation of PCR product-based microarrays and the comparative genomic hybridization was performed essentially as described by Cummings et al. (12). This study employed a new array design that contained all of the probes from the first array plus 1,417 additional probes that brought the theoretical ORF coverage of these arrays up to 97.4% for B. pertussis Tohama, 98.5% for B. bronchiseptica RB50, and 97.9% for B. parapertussis 12822. Like the previously used array probes, these additional probes were PCR products with a size of less than 300 base pairs and amplified from the sequenced reference genomes with ORF-specific oligonucleotides (Illumina, San Diego, California, United States) designed with Microarray Architect (C. A. Cummings and D. A. Relman, unpublished data).
A total of 26 complex I and 13 complex IV strains were hybridized to the arrays. The genomic DNA of B. bronchiseptica was labeled with Cy5 and hybridized to the array in conjunction with a Cy3-labeled genomic DNA reference comprising the three sequenced mammalian Bordetella genomes (B. pertussis Tohama, B. parapertussis 12822, and B. bronchiseptica RB50). For the list of strains analyzed by CGH, see Table S1. Labeled probes were purified using the Cyscribe GFX Purification Kit (Amersham Biosciences, Freiburg, Germany) following the manufacturer's protocol for probes produced by the CyScribe First-Strand cDNA Labelling Kit. After purification, the test and reference-labeled DNA samples were concentrated to less than 8.5 μl using a Savant SpeedVac SVC-100H. The test and reference samples were combined and 150 μg of yeast tRNA (Invitrogen Life Technologies, San Diego, California, United States) was added to block nonspecific binding. The probe volume was adjusted to 24 μl with water and then 5.1 μl of 20× SSC (1× SSC = 0.15 M NaCl plus 0.015 M sodium citrate) and 0.9 μl of 10% sodium dodecyl sulfate (SDS) were added. Thirty microliters of the probe was added to the array and covered with a 25 × 40 mm No. 1 glass coverslip. Hybridization was performed in GeneMachines Hybchambers (Genomic Solutions, Ann Arbor, Michigan, United States) with 2× 30 μl of 3× SSC to maintain humidity and incubated at 65 °C overnight.
Arrays were washed in 0.5× SSC, 0.03% SDS for 30 s, 0.1× SSC, 0.01% SDS for 30 s, 0.05× SSC, 0.005% SDS for 1 min, and 0.025× SSC for 1 min. The first wash was performed at 65 °C, and the remaining washes were performed at room temperature. Slides were dried using a Quick-Dry Filtered Air Gun (Matrix Technologies Corporation, Hudson, New Hampshire, United States). Images were acquired on a PerkinElmer ScanArray 4000XL scanner using Scanarray Express software (PerkinElmer Life and Analytical Sciences, Inc., Boston, Massachusetts, United States). Images were analyzed with GenePix Pro software (Axon Instruments, Union City, California, United States).
Processed two-color array image data were submitted to an in-house microarray database. Data were extracted using filters to eliminate automatically and manually flagged spots and spots with very low background subtracted signal intensity (<150) in the reference channel. B. bronchiseptica complex I and complex IV enriched sequences were identified using the Significance Analysis for Microarrays software (SAM) . The probes were analyzed using 26 complex I and 13 complex IV strains that were hybridized to the arrays. SAM analysis was run using the two-class option with KNN missing value imputation. In addition to a statistically significant difference, a 2-fold difference in mean signal intensity ratio for each probe was also required.
We are grateful to Dr. Geoffrey Foster (SAC Veterinary Science Division, Inverness) and to Dr. Gary Sanders (Centers for Disease Control and Prevention, Atlanta, Georgia, United States) for providing B. bronchiseptica strains. We thank Dr. Eric Harvill (Penn State University, Pennsylvania, United States) for sharing unpublished data and discussions and Ing. Marjolein van Gent, Ing. Betsy Kuipers, and Ing. Hendrik-Jan Hamstra for assistance and introduction to LPS work. We also thank Dr. Martin Maiden and Dr. Keith Jolley for assistance with setting up the Bordetella MLST database. This work was supported by a travel grant from Netherlands Organization for Scientific Research (NWO). CAC was supported by an American Lung Association Research Training Fellowship. DAR received grant support from the National Institutes of Health (grants AI54970 and AI057188).
DAD and FRM conceived and designed the experiments. DAD performed the experiments. DAD, CAC, LMS, and MMB analyzed the data. CAC and LMS contributed reagents/materials/analysis tools. DAD, CAC, LMS, DAR, and FRM wrote the paper.
- 1. Gueirard P, Weber C, Le Coustumier A, Guiso N (1995) Human Bordetella bronchiseptica infection related to contact with infected animals: Persistence of bacteria in host. J Clin Microbiol 33: 2002–2006.P. GueirardC. WeberA. Le CoustumierN. Guiso1995Human Bordetella bronchiseptica infection related to contact with infected animals: Persistence of bacteria in host.J Clin Microbiol3320022006
- 2. Bauwens JE, Spach DH, Schacker TW, Mustafa MM, Bowden RA (1992) Bordetella bronchiseptica