Haemolytic and cytotoxic activities of the Tween 80-extracted putative haemolysin of Pasteurella multocida B:2.

The objective of this study was to investigate the haemolytic and cytotoxic activity of Pasteurella multocida B:2 strains, originally from cases of haemorrhagic septicaemia in cattle. All six P. multocida B:2 strains were non-haemolytic on sheep blood agar (SBA) and horse blood agar (HBA) when grown aerobically and on SBA anaerobically but they were haemolytic on HBA when grown anaerobically. No haemolytic activity against horse red blood cells was detected in culture supernates from aerobically or anaerobically grown cultures and only very weak haemolytic activity was obtained in supernates or pellet fractions from sonicated cells. However, after repeated extraction of sonicated cells with Tween 80, haemolytic activity was found in various cell fractions, both Tween-soluble and -insoluble. The Tween-extracted putative haemolysin and other bacterial fractions were also cytotoxic for mouse macrophage-like J774.2 cells. Further characterisation of the putative haemolysin revealed it to be a heat-labile, non-pore-forming protein of molecular weight >10 kDa whose activity was completely destroyed by trypsin and greatly reduced with protease and proteinase K treatment. Congo red also reduced the haemolytic activity. Non-denaturing gel-electrophoresis and RBC agar overlay revealed clear haemolytic zones but suggested that Tween was bound to some component of the P. multocida B:2 fractions and was responsible, to some extent, for the haemolytic activity observed. However, the effect of heat and other reagents on the Tween-extracted fractions and the lack of haemolytic activity in different Tween-extracted cell fractions of organisms other than P. multocida suggested that some proteinaceous component of the organism could indeed act as a haemolysin. This putative haemolysin may be one of the virulence attributes of P. multocida, but its characterisation and role in pathogenesis require further study.

Virulence spectra of typed strains of Campylobacter jejuni from different sources: a blinded in vivo study.

Campylobacter jejuni is a major cause of human diarrhoeal disease, but specific virulence mechanisms have not been well defined. The aims of the present blinded study were to measure and compare the in vivo properties of 40 serotyped, biotyped and genotyped C. jejuni isolates from different sources and genetic makeup. An 11-day-old chick embryo lethality assay, which measured embryo deaths and total viable bacteria over 72 h following inoculation of bacteria into the chorioallantoic membrane, revealed a spectrum of activity within the C. jejuni strains. Human and chicken isolates showed similar high virulence values for embryo deaths while the virulence of the bovine isolates was less pronounced. A one-way ANOVA comparison between the capacity of the strains to kill the chick embryos after 24 h with cytotoxicity towards cultured CaCo-2 cells was significant (P=0.025). After inoculation with a Campylobacter strain, mouse ligated ileal loops were examined histologically and revealed degrees of villous atrophy, abnormal mucosa, dilation of the lumen, congestion and blood in lumen, depending on the isolate examined. A 'total pathology score', derived for each C. jejuni strain after grading the pathology features for degree of severity, showed no apparent relationship with the source of isolation. Some relationship was found between amplified fragment length polymorphism groups and total ileal loop pathology scores, and a one-way ANOVA comparison of the mouse pathology scores against total chick embryo deaths after 72 h was significant (P=0.049).

Comparison of virulence-associated in vitro properties of typed strains of Campylobacter jejuni from different sources.

Campylobacter jejuni is a major cause of human diarrhoeal disease, but specific virulence mechanisms have not been well defined. This blinded study was undertaken with 40 C. jejuni isolates from different sources to determine their haemolytic, cytotoxic and adhesion and invasion activities towards mammalian cells. The results were correlated with source of isolation and genetic makeup by amplified fragment length polymorphism (AFLP) typing. The isolates had variable degrees of haemolytic activity against rabbit erythrocytes and cytotoxicity towards CaCo-2, HeLa and Vero cells. The data indicated that the haemolytic and cytotoxic activities were due to separate factors. A range of cytotoxicity was exhibited, whereby some strains had no activity against the target cells and others had activity against all three cell lines. Certain strains had activity against CaCo-2 cells but little or no activity against the other cells, while others exhibited the opposite phenotype. The data suggested that the cytotoxicity assay with the different cell lines may have detected more than one cytotoxin. A wide variation between isolates was observed for both adherence and invasion with all three cell lines, yet, overall, the strains showed a significantly greater invasion capacity for CaCo-2. There was no clear relationship between source of isolation or disease manifestation and possession of statistically significantly higher levels of particular virulence-associated factors although, in some cases, a correlation between cytotoxicity and cell invasion was evident. Five AFLP clusters, each representing two to eleven isolates with similar profiles, were observed at the 90 % similarity level. Some AFLP groups contained isolates with a common serotype, but each group had C. jejuni isolates from more than one source with the exception of group IV, which contained only human isolates. Isolates with high cytotoxic activity against CaCo-2 cells were confined to groups I, III and IV and a group of unrelated strains (U). Group II isolates had uniformly low cytotoxicity. Isolates in groups I, V and U were more invasive for CaCo-2 cells than isolates in groups II, III and IV. The strain differences in cytotoxicity or invasion did not correlate with source of isolation.

ADP-ribosylation activity in pertussis vaccines and its relationship to the in vivo histamine-sensitisation test.

Pertussis toxin (PTx) is a major virulence factor produced by Bordetella pertussis. In its detoxified form (PTd), it is an important component of acellular pertussis vaccines although some residual PTx activity may likely be present because of the limitations of the detoxification processes used. Furthermore, different detoxification procedures have been shown to result in different amino acid side-chain modifications for the resulting PTd. The histamine-sensitisation test (HIST) in mice is currently used for the safety testing of these vaccines. However, an alternative test is needed because of large assay variability and ethical concerns. The ADP-ribosylation enzyme activity of PTx is thought to be the major factor responsible for the histamine-sensitising activity detected in vivo. In the present study, the ADP-ribosylation activity in different acellular pertussis-based combination vaccine formulations was measured and compared with reactivity in the HIST. The results indicated that different products showed differences in ADP-ribosylation activity and a level which would be significant in relation to the reactivity seen in the HIST could not be defined, except for vaccines that contain genetically detoxified PTx, which do not have enzymatic activity nor in vivo toxicity. Different detoxification procedures as well as formulation factors could contribute to this variation. Relying solely on the residual enzyme activity of PTx in vaccines containing chemically detoxified PTd may not fully reflect the in vivo reactivity observed by the HIST. Refinement of the in vitro test to include a step which monitors the B-subunit activity of PTx may provide a better correlation with the in vivo HIST.

Environmental sensing mechanisms in Bordetella.

The success of a bacterial pathogen may depend on its ability to sense and respond to different environments. This is particularly true of those pathogens whose survival depends on adaptation to different niches both within and outside the host. Members of the genus Bordetella cause infections in humans, other animals and birds. Two closely related species, B. pertussis and B. bronchiseptica, cause respiratory disease and express a similar range of virulence factors during infection, but exhibit different host ranges and responses to environmental change. B. pertussis has no known reservoir other than humans and is assumed to be transmitted directly via aerosol droplets between hosts. B. bronchiseptica, on the other hand, has the potential to survive and grow in the natural environment. Comparison of the manner in which these two organisms respond to external signals has provided important insights into the co-ordinate regulation of gene expression as a response to a changing environment. During infection, both species produce a range of virulence factors whose expression is co-ordinated by two members of the two-component family of signal transduction proteins, the bvg (bordetella virulence gene) and ris (regulator of intracellular stress response) loci. When active, the bvg locus directs the activity of a number of virulence determinants in both species whose products, such as adhesins and toxins, establish colonization of the host by the bacteria, although each organism has evolved a slightly different strategy during pathogenesis. B. pertussis, the causative agent of whooping cough, promotes an acute disease and tends to be more virulent than B. bronchiseptica which generally causes chronic and persistent asymptomatic colonization of the respiratory tract. The recently identified ris locus appears to control the expression of factors important for intracellular survival of B. bronchiseptica, but a role for this regulatory locus in B. pertussis infection has not been established. Expression of the virulence determinants controlled by the bvg and ris loci is subject to modulation by different environmental signals, such as low temperature, which act through these two-component systems. Evidence indicates that, for B. bronchiseptica, bvg-controlled determinants expressed under modulating conditions, such as motility, facilitate adaptation and survival in environments outside the host. With B. pertussis, however, there is no apparent requirement for prolonged survival outside the host and this difference is reflected in the expression of different, as yet uncharacterized, determinants as a response to modulating signals. The nature of the gene products involved and their assumed role in the life cycle of B. pertussis remains to be determined. Thus, comparative analysis of these species provides an excellent model for understanding the genetic requirements for pathogenesis of respiratory infection and adaptation to changing environments, both within and outside the host.

PCR methods and plasmid profile analysis for characterisation of Histophilus ovis strains.

The value of polymerase chain reaction (PCR)-based DNA fingerprinting and plasmid profile analysis for differentiation of Histophilus ovis isolates was assessed. Nineteen isolates of H. ovis were typed by PCR-ribotyping, repetitive extragenic palindromic element (REP)-PCR and enterobacterial repetitive intergenic consensus (ERIC)-PCR. These methods distinguished five types by PCR-ribotyping, 11 types by REP-PCR and seven types by ERIC-PCR. The ribotyping method produced a relatively simple pattern and a small number of distinct types and was useful for differentiation of H. ovis from the phenotypically similar organism, Haemophilus somnus. REP- and ERIC-PCR both produced complex banding patterns, but increased the discrimination between strains. Plasmids were found in 12 of the 19 isolates and there were four different plasmid profiles. A combination of the PCR methods and plasmid profile analysis provided a high resolution typing method for H. ovis.

Bioluminescence as a reporter of intracellular survival of Bordetella bronchiseptica in murine phagocytes.

The uptake and persistence of Bordetella bronchiseptica was characterized in murine phagocytes by using a novel bioluminescence-based reporter system. A mini-Tn5 promoter probe carrying the intact lux operon from the terrestrial bacterium Photorhabdus luminescens which allowed measurement of light output without the addition of exogenous substrate was constructed. It was used to create a pool of bioluminescent fusion strains of B. bronchiseptica. The internalization and persistence in murine macrophages of a constitutive bioluminescent strain of B. bronchiseptica was monitored by luminometry and by fluorescence and electron microscopy. The number of bacteria internalized, in a microfilament-dependent process, by a mouse macrophage-like cell line after 2 h was approximately 1% of the inoculum for several different multiplicities of infection (MOI). At an MOI of <500:1 (bacteria to macrophages), viable numbers of intracellular bacteria declined over a 4-day period. However, at an MOI of >/=500:1, long-term survival was enhanced, with viable bacteria recovered up to 4 days postinfection with little decline in numbers, indicating that a critical population size may have been essential for intracellular persistence. No evidence of macrophage killing by intracellular bacteria was detected over the 4-day period. Intracellular bioluminescent B. bronchiseptica organisms in mouse peritoneal cells were detected at 24 and 48 h after intraperitoneal injection of mice. Bioluminescence is shown to act as a convenient real-time technique for monitoring of intracellular survival of B. bronchiseptica in vitro and may provide a suitable means for examining the role of long-term intracellular survival of the bacterium in the host.

PCR methods for rapid identification and characterization of Actinobacillus seminis strains.

Twenty-four isolates of Actinobacillus seminis were typed by PCR ribotyping, repetitive extragenic palindromic element (REP)-based PCR, and enterobacterial repetitive intergenic consensus (ERIC)-based PCR. Five types were distinguished by REP-PCR, and nine types were distinguished by ERIC-PCR. PCR ribotyping produced the simplest pattern and could be useful for identification of A. seminis and for its differentiation from related species. REP- and ERIC-PCR could be used for strain differentiation in epidemiological studies of A. seminis.

Specific PCR primers from the 16S-23S rRNA spacer region for the rapid detection and identification of Actinobacillus seminis.

Actinobacillus seminis is a common cause of ovine epididymitis and ram infertility. The ability to detect and identify this organism promptly is important commercially for the quality control of ram semen samples. Actinobacillus seminis is a fastidious and slow-growing bacterium and primary isolation and presumptive identification can be difficult and time-consuming. In this study, two ribosomal operons, termed rrnA and rrnB, have been characterized in the A. seminis genome, and these contain one and two tRNAs, respectively, in the spacer region between the 16S and 23S rRNA genes. Species-specific primers for A. seminis were developed from the sequence of the spacer region of rrnB for the identification and detection of A. seminis by PCR. The PCR assay was specific for A. seminis and gave no amplification products with phenotypically similar organisms such as Histophilus ovis. Storage solution used to preserve semen for long-term storage was found to inhibit the PCR. Therefore, for diagnostic purposes, the assay would best be performed after primary isolation or perhaps on fresh semen prior to storage if obvious contamination is indicated.

Characterisation of the leukotoxin produced by different strains of Pasteurella haemolytica.

Pasteurella haemolytica isolates from cattle and sheep, including representatives of different serotypes and untypable strains, were examined for leukotoxin (Lkt) production at the end of the log phase of growth in brain heart infusion broth. There were marked differences in leukotoxic activity in culture supernate samples, as measured by chemiluminescence-inhibition assays with bovine and ovine neutrophils, even between strains of the same serotype. There was also some variation in the amount and mol. wt of the Lkt protein produced by different strains, as judged by SDS-PAGE, immunoblotting and ELISA. Some strains produced normal amounts of Lkt protein which had only low leukotoxic activity. Most strains produced Lkt of 105 kDa whereas four strains produced a higher mol. wt form of c. 108 kDa, including two of the five serotype A2 strains examined. Thus, the P. haemolytica isolates showed considerable heterogeneity in terms of leukotoxin production, mol. wt and activity, even within a given serotype.

Genomic fingerprinting of Haemophilus somnus by a combination of PCR methods.

Twenty-three isolates of Haemophilus somnus were typed by repetitive extragenic palindromic (REP) element-based PCR, enterobacterial repetitive intergenic consensus (ERIC)-based PCR, and PCR ribotyping. A total of 11 types were distinguished by REP-PCR, 13 types were distinguished by ERIC-PCR, and 5 types were distinguished by PCR ribotyping. PCR ribotyping produced a relatively simple pattern and a small number of distinct types, whereas REP- and ERIC-PCR both produced more complex banding patterns but increased the discrimination between strains. Clearly distinguishable profiles were obtained for respiratory and genital isolates of H. somnus by all three typing methods. The results suggest that a combination of all three primer sets provides a high-resolution fingerprinting method for epidemiological studies of H. somnus and for its differentiation from related species.

Bordetella pertussis adenylate cyclase toxin: proCyaA and CyaC proteins synthesised separately in Escherichia coli produce active toxin in vitro.

Bordetella pertussis produces a cell-invasive adenylate cyclase toxin (CyaA) which is related to the RTX family of pore-forming toxins. Like all RTX toxins, CyaA is synthesised as a protoxin (proCyaA), encoded by the cyaA gene. Activation to the mature cell-invasive toxin involves palmitoylation of lysine 983 and is dependent on co-expression of cyaC. The role of the cyaC gene product in the acylation reaction has not been determined. We have developed an efficient T7 RNA polymerase system for over-expression of cyaA and cyaC separately in Escherichia coli. Each protein accumulated intracellularly in an insoluble form and could be collected by centrifugation of lysed cells. A single-step purification was achieved by extraction of the aggregated material with 8 M urea. Active cell-invasive CyaA was produced in vitro when the proCyaA and CyaC proteins were mixed with a cytosolic extract of either E. coli or B. pertussis. Activation was assumed to occur by an acylation reaction requiring acyl carrier protein (ACP) as cofactor, as the cytosolic factor required for toxin activation was lost if the S100 extract was dialysed before use and the cytosolic factor could be replaced in the in vitro reaction by ACP charged separately in vitro with palmitic acid, as reported previously for activation of the homologous E. coli haemolysin (HlyA). The in vitro activation system may be used to investigate the mechanism of the CyaC-dependent acylation of proCyaA and the effect of variation of the modifying fatty acyl group on target cell specificity and toxic activity of CyaA.

A rapid, colourimetric assay for cytotoxin activity in Campylobacter jejuni.

Cell extracts and culture supernates of Campylobacter jejuni NCTC 11168 and three isolates from faecal samples from patients with enteritis were tested for cytotoxic activity on HeLa and Vero cells using a sensitive and rapid dye reduction assay which represents a simple assay for cytotoxin activity that can be assessed visually or spectrophotometrically in the wells of microplates. The assay was as sensitive as trypan blue exclusion and did not require the use of radioisotopes. A low level of cytotoxin activity, compared to that produced by a control verotoxin 2-producing Escherichia coli strain, was detected in cell extracts of all four strains, but no activity was detected in culture supernates. Production of an enterotoxin was evaluated by reverse passive latex agglutination with anti-cholera toxin antibody, a procedure which also represents a rapid and simple assay for this toxin. No enterotoxin activity was detected in cell extracts or culture supernates from any of the isolates.

Comparison of outer-membrane proteins of Pasteurella haemolytica expressed in vitro and in vivo in cattle.

Outer-membrane protein (OMP) profiles of two serotype A1 isolates of Pasteurella haemolytica were compared by SDS-PAGE and Western blotting with bovine convalescent serum after growth (a) in vitro under iron-sufficient and -deficient conditions, (b) in vivo in the lungs of experimentally infected calves and (c) in vivo in diffusion chambers implanted into the peritoneal cavities of calves. Lung-grown bacteria differed from iron-sufficient in vitro-grown bacteria in having enhanced expression of the previously recognized 71, 77 and 100 kDa iron-regulated proteins, reduced expression of 18, 31, 39.5 and 50 kDa proteins, and expression of a 19 kDa protein. Differences were also apparent in the Western blot profiles of OMPs of in vitro- and lung-grown bacteria. These included the apparent lack of recognition of the 100 kDa protein in the lung-grown bacteria, but not in the in vitro-grown bacteria, and more intense staining of a 47 kDa protein in in vitro-grown bacteria, but not in lung-grown bacteria. The OMP profiles of the chamber-grown bacteria resembled those of the lung-grown bacteria in that expression of the 18, 19, 31 and 39.5 kDa proteins was similar. These similarities demonstrated that the chamber-grown bacteria had adapted to the in vivo environment, and that growth conditions within the chambers resembled, but not perfectly, those within the lungs. For example, expression of the three iron-regulated OMPs was very low in the chamber-grown bacteria compared to the lung-grown bacteria.(ABSTRACT TRUNCATED AT 250 WORDS)

Construction of conjugative shuttle and suicide vectors for Pasteurella haemolytica and P. multocida.

A shuttle cloning vector, pAKA16, and suicide derivatives pAKA19 and pAKA22 have been developed for gene transfer to Pasteurella haemolytica and P. multocida. pAKA16 was constructed by insertion of the lacZ alpha-peptide-encoding region and a multiple cloning site into a plasmid which was originally isolated from P. haemolytica serotype A1. The vector encodes ampicillin resistance, and contains at least 14 unique restriction sites and the property of phenotypic identification of recombinant clones in Escherichia coli by insertional inactivation of beta-galactosidase activity. It can be transferred by conjugation to P. haemolytica or P. multocida and is stably maintained in both species. The type-II chloramphenicol acetyltransferase-encoding gene (cat), cloned into pAKA16, was stably expressed in both P. haemolytica and P. multocida. Plasmids pAKA19 and pAKA22 were constructed by replacement of the origin of DNA replication (ori) of pAKA16 with a ColE1-type ori from pBR322 or an ori of plasmid R6K (oriR6K) from pJM703.1, respectively. These derivatives replicate in E. coli, but not in either P. haemolytica or P. multocida, and are suitable for use as suicide vectors for these Pasteurella species.

Development of an intraperitoneal implant chamber for the study of in vivo-grown Pasteurella haemolytica in cattle.

An intraperitoneal implant chamber was developed for the study of the in vivo growth of Pasteurella haemolytica in calves. The chamber had a volume of approximately 100 ml and featured an external sampling port which allowed multiple and sequential sampling of the chamber contents. A single polycarbonate diffusion membrane with a pore size of 0.22 micron allowed host peritoneal fluid to diffuse into the chamber and maintained the bacterial population free of white blood cells. Chambers were implanted into the peritoneal cavities of four five-month-old dairy-cross calves, demonstrated to be sero-negative by indirect haemagglutination assay. Three days later, four different P. haemolytica isolates, of serotypes A1 or A2, were inoculated into the chambers. In all cases, there was a slow decline in the viable bacterial numbers within the chambers. Western blot analysis of the antibody content of the chamber fluids revealed IgG antibodies to P. haemolytica OMPs in the fluid prior to inoculation and both 9 and 15 days after inoculation. Furthermore, there was no significant change in the IgG antibody content of the chamber fluid, either quantitatively or qualitatively, during the course of the experiment. Analysis of the bactericidal activity of pre-inoculation chamber fluid against the corresponding bacterial isolate suggested that an antibody-dependent complement-mediated process was not responsible for the decline in bacterial numbers. Overall, the chamber design was demonstrated to be extremely effective for in vivo studies of P. haemolytica in calves, allowing easy and regular sampling of the chamber contents and maintaining bacteria free of white blood cells. Although there was a slow decline in bacterial numbers over time, sufficient numbers of cells could be obtained for analysis of cell-surface antigens.

A new assay for the invasive adenylate cyclase toxin of Bordetella pertussis based on its morphological effects on the fibronectin-stimulated spreading of BHK21 cells.

When baby hamster kidney (BHK) cells are allowed to spread on fibronectin-coated substrata in the absence of serum and the presence of agents which elevate intracellular 3':5'-cyclic AMP (cAMP) levels they adopt an abnormal, stellated morphology. To determine whether the invasive adenylate cyclase (AC) toxin of Bordetella pertussis induced the same response, cell extracts were prepared from several B. pertussis strains. They were characterized for AC toxin production by enzymic assay and by immunoblotting with an AC-toxin-specific monoclonal antibody. Extracts of strains producing AC toxin induced elevated levels of intracellular cAMP in BHK cells and promoted a stellation response during cell spreading. Extracts prepared from strains defective in AC toxin production showed no effect. Using image analysis to quantify the morphological change, we have demonstrated that the effect of AC toxin on cell spreading is dose dependent. This technique is a rapid and sensitive assay for the invasive AC toxin.

Evaluation of different methods for the detection of outer membrane proteins and lipopolysaccharides of Pasteurella haemolytica by immunoblotting.

The optimal conditions for the detection of outer membrane proteins (OMPs) and lipopolysaccharide (LPS) of Pasteurella haemolytica by immunoblotting were evaluated. The variables examined included the equilibration time of the gels before transfer, composition of the transfer buffer, type of blotting membrane, blocking agent, effect of the zwitterionic detergent Empigen BB on protein renaturation, and the development reagent. The composition of the transfer buffer and time of gel equilibration significantly affected the efficiency of transfer of both OMPs and LPS. However, the optimal conditions for the transfer of OMPs were not the same as those for LPS. Thus, optimal transfer of OMPs occurred in Tris-glycine buffer, with prior equilibration of the gels to allow for expansion, whereas optimal transfer of LPS was achieved in Tris-glycine-methanol buffer with no equilibration of the gels. In Tris-glycine-methanol buffer, gel equilibration resulted in a significantly reduced transfer of both OMPs and LPS, probably due to the removal of SDS from these components. The use of Zeta-Probe blotting membrane which, unlike nitrocellulose, does not require methanol for optimal protein binding, did not result in improved binding of OMPs or LPS in the absence of methanol and, even after prolonged blocking (> 2 h), gave higher background staining than did nitrocellulose. Effective blocking of nitrocellulose was achieved with 3% (w/v) gelatin, 2.5% (w/v) skimmed milk or 0.3% (v/v) Tween 20, whereas increased background staining occurred with 1% (w/v) bovine serum albumin or 1% (w/v) ovalbumin. The incorporation of Empigen BB in the primary antibody buffer did not improve antibody recognition of proteins as a result of their renaturation. For the horseradish-peroxidase enzyme development system, the substrate 3,3'-diaminobenzidine tetrahydrochloride was more sensitive, and developed more quickly, than 4-chloro-1-naphthol, but faded more rapidly after drying of the membrane. 4-chloro-1-naphthol was more suitable for identifying OMPs because less background staining occurred, whereas 3,3'-diaminobenzidine tetrahydrochloride was more suitable for the detection of LPS due to its greater sensitivity.

Chromosomal DNA from both flagellate and non-flagellate Bordetella species contains sequences homologous to the Salmonella H1 flagellin gene.

The genus Bordetella contains four species: two are non-motile, the human pathogens B. pertussis and B. parapertussis; and two are motile, the broad host-range mammalian pathogen B. bronchiseptica, and the avian pathogen B. avium. The motility of the latter two species is due to peritrichous flagella. Here we show that strains of all four species contain DNA sequences homologous to flagellin genes. Two types of gene probe were hybridised to Bordetella chromosomal DNa in Southern blots: the structural gene for H1 flagellin of Salmonella typhimurium and an oligonucleotide derived from the conserved N-terminal amino acid sequences of various flagellin proteins. ClaI-digested DNa from all four Bordetella species hybridised with both probes in Southern blots, although each species gave a characteristic pattern of hybridisation. This indicates that the non-motile B. pertussis and B. parapertussis species contain non-expressed flagellin genes.