INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

All of the principal Bordetella species, B. pertussis, B. parapertussis, B. bronchiseptica, and B. avium, can cause upper respiratory disease. In many instances, the pathogenesis involves the interaction of the bacteria with ciliated tracheal epithelial cells (2, 22, 30), resulting first in ciliastasis and eventually in death of the ciliated cells (26, 31, 36). Further, all species can cause diseases with similar tracheal lesions and outward signs and symptoms that involve ocular-nasal discharge and persistent severe coughing (5, 28, 32). However, each Bordetella species causes the hallmark signs of disease only in particular hosts. In the case of B. pertussis, children are susceptible and the disease is called whooping cough (reviewed in reference 40). B. parapertussis causes a milder form of whooping cough in humans and a chronic pneumonia in lambs (6). B. bronchiseptica infects the upper respiratory tract of a number of domestic, companion, and laboratory animals and can cause a variety of upper respiratory diseases in these hosts (e.g., kennel cough in dogs [41]), most of which are complicated when seen naturally (reviewed in reference 16). With B. avium, birds are the susceptible host and the disease produced is called avian bordetellosis or turkey coryza, the latter name reflecting the most economically important animal commonly afflicted (reviewed in reference 32).

B. parapertussis, B. bronchiseptica, and B. avium all have characters associated with virulence in the type species, B. pertussis (reviewed in reference 39). Of those characters, B. avium has the smallest subset: dermonecrotic toxin (DNT), tracheal cytotoxin, hemagglutination, and fimbriae (reviewed in reference 32). Also, B. avium is the furthest removed from the other Bordetella species by systematic measurements (8). These observations might lead one to suspect that B. avium is fundamentally different from the other Bordetella species in its method of disease production. However, it may be that generation of the basic tracheal lesion and production of the most pronounced clinical features of the disease in the natural host depend upon characteristics (e.g., tracheal cytotoxin [7]) common to all Bordetella species. This possibility remains open in part due to the lack of a practical experimental animal that gets the clinically pronounced symptoms and the histological features of bordetellosis caused by B. pertussis and B. bronchiseptica under well-controlled experimental conditions (16, 39).

Turkeys (Meleagris galapavo) are plentiful and readily show pronounced signs of bordetellosis under experimental conditions (27). Whereas birds are indeed evolutionarily distant from humans and other mammals, the insights gained from a systematic study of avian bordetellosis may be useful in understanding the pathogenesis of all Bordetella infections. Furthermore, avian bordetellosis is of significant economic concern to producers of turkeys worldwide (32). Turkeys grow faster and use feed more efficiently than chickens (10). Consequently, worldwide agriculture has an interest in their development as a food source.

As a first step in analyzing B. avium virulence, we have systematically examined several variables in the experimental infection. These include a statistical analysis of the effects of turkey age at infection, the time required for tracheal colonization, the effect of group size on 50% infectious dose (ID₅₀) measurements, and the communicability of B. avium within groups. In addition, since turkeys are not inbred animals (but can be divided into distinct strains or varieties, referred to here as lines to avoid confusion with bacterial strains), we examined the five major commercial lines for possible host-associated differences in susceptibility to the parental strain and three mutant strains of B. avium. Finally, we developed and employed an in vitro turkey tracheal ring assay that examined the adhesion of the B. avium parental and mutant strains.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacterial strains and growth conditions. B. avium strains used in this study are all derivatives of strain 197N, a spontaneous nalidixic acid-resistant mutant of strain 197 (14). Strain 197 was chosen from a laboratory collection of three B. avium strains that were tested for virulence. One strain, GOBL271 (15) was of reduced virulence compared to the other two. Strain 197 was the better described of the two remaining (14) and was chosen for our studies. All bacterial strains, plasmids, and one generalized transducing phage are described in Table 1. B. avium and Escherichia coli (when used for matings) were grown on brain heart infusion (BHI) medium (Difco) at 37°C. Broth cultures were shaken vigorously. E. coli was routinely maintained on medium composed of L broth or L agar (Difco). Lactose MacConkey agar (composed of 1% lactose and MacConkey agar base [Difco]) was used to enrich and isolate B. avium from infected turkey tracheas. B. avium minimal medium was prepared as described previously (21) except that dextrose was omitted. Stainer-Scholte agar (18) was prepared with modifications for B. avium (SSM agar [14]) and contained 10 mM MgSO₄ when used for mating. Bordet-Gengou agar was prepared as directed (Difco) with 15% sheep blood added.

Source of animals and housing conditions. The various lines of turkeys used in this study were obtained from three sources depending upon the type of experiment to be performed. (i) Turkey poults used for comparing line susceptibility were obtained as breeding stock from a commercial supplier, and the resulting poults were laid and hatched at the North Carolina State University Poultry Science facility. Birds for these experiments were obtained at hatching from this facility. (ii) For studies involving tracheal ring assays, fertilized eggs of the BUT-A line were obtained from the British United Turkeys of America breeding location in Lewisburg, W.Va. These embryonated eggs were incubated in a Kuhl (Flemington, N.J.) model AT-600-110 incubator for 26 days before sacrifice and use. (iii) Birds not specifically used in host susceptibility tests were obtained from Tarheel Turkey Hatchers in Raeford, N.C., and from the British United Turkeys of America breeding location in Lewisburg, W.Va., at 1 day after hatching. Hatched birds were kept at the AAALAC-accredited facility at North Carolina State University and housed in stainless steel brooders (a maximum of 10 birds were kept in a single 61- by 91- by 28-cm brooder) for the 3 weeks (average) needed to complete an experiment.

Standard infection protocol. The standard protocol for infecting turkeys was established after a number of parameters had been examined. Some of these parameters are described in Results. Features common to all experiments are mentioned here. Just prior to an experiment, approximately 5% of the population was removed (three to five birds) and tracheal cultures and serum samples were obtained. Serum samples were evaluated for antibody to B. avium in a slide agglutination test performed by Rollins Animal Diagnostic Laboratory, Raleigh, N.C. In all experiments, the tested birds were uniformly negative in both cultural and immunological tests. Birds were infected by B. avium with an inoculum obtained from overnight, 37°C, BHI plate-grown cultures that had been resuspended in phosphate-buffered saline (PBS) to give three concentrations necessary for an ID₅₀ determination. Parent and mutant microorganisms were always compared in a given experiment in order to detect any variation in the general health of received birds. Each experiment involved an analysis of the fraction of birds infected (infection rate) in three groups of 10 birds per group, each group given a 10-fold increasing dose of B. avium. Each experiment also contained a group of control (PBS sham-inoculated) birds of the same size as an inoculated group (10 birds).

Genetic methods. B. avium hemagglutination (Hag) and motility (Mot) mutants were generated by using pUT/mini-Tn5lacZ2 and -Tn5phoA plasmids constructed by DeLorenzo et al. (9). A triparental mating was performed by mixing overnight broth cultures of (i) E. coli MM294 containing mating plasmid pRK2013, (ii) E. coli CC118 (pir) containing the pUT/mini-Tn5 plasmid, and (iii) the parental strain B. avium 197N in a 1:1:10 ratio. The mating mixture was dropped onto the surface of dry SSM agar plates (without selective antibiotics). Mating was allowed to proceed for 4 to 8 h at 35°C, after which the mixture was removed with a cotton swab and plated on selective L agar (containing, per ml, 30 µg of nalidixic acid, 150 µg of kanamycin, and 40 µg of 5-bromo-6-chloro-3-indolyl--D-galactoside or 5-bromo-6-chloro-3-indolyl phosphate) and incubated for 2 days at 35°C. Blue B. avium exconjugants were patched onto selective medium plates and tested for hemagglutination and motility phenotypes. Prospective Hag and Mot mutants were colony purified, the phenotypes were rechecked, and mutant strains were stored in 50% glycerol-50% L broth at 80°C. The DNT mutant (Dnt) was isolated by insertion mutagenesis, except that in this case a cloned B. pertussis dnt gene was interrupted in vitro following cloning onto a mobilizable suicide plasmid (see Table 1 and Results).

DNT measurements. Cell-free lysates of bacteria were obtained by first growing B. avium from glycerol stock cultures on SSM agar plates at 35°C for 24 to 36 h. Bacterial cell suspensions (approximately 10⁸ bacteria/ml) were prepared in sterile saline and sonicated (Branson Sonifier model 350; 1-cm tip) on ice at 50% power for 2 min (or until the optical density at 600 nm of the suspension was reduced by at least 50%). The sonicate was then subjected to centrifugation for 1 h at 100,000 × g at 5°C, and the supernatant was sterilized by passage through a 0.2-µm filter. DNT activity of bacterial preparations was determined in 7- to 8-day-old albino outbred Swiss Webster mice. Sonicated cell suspensions (0.05 ml/mouse) were injected intradermally on one side of the back. The mice were marked to distinguish different treatments and kept with the mother. They were regularly observed for 2 days to note the time of lesion appearance. Dark purple to black lesions at the inoculation site were considered DNT positive.

Motility and flagellin assay. B. avium strains were tested for motility by touching an isolated colony with a sterile toothpick and stabbing it into SSM containing 0.4% agar. Plates were observed for expanding zones of bacterial growth, indicating motility, after 24 h at 35°C. The presence of flagella on nonmotile mutants was checked by transmission electron microscopy (TEM) and by colony immunoblotting using monoclonal antibody to E. coli flagellin as described by Feng et al. (11).

Hemagglutination assay. Suspensions of 10¹⁰ bacteria/ml were prepared from strains grown on L agar overnight at 35°C. These were isolated by centrifugation and resuspended to give 10¹¹ cells/ml in normal saline (0.15 N NaCl), and 0.1 ml of each suspension was placed in the first well of a round- or pointed-bottom 96-well plate (Costar). Normal saline (50 µl/well) was added to each subsequent well in the row, and serial twofold bacterial dilutions were made. Guinea pig erythrocytes (Cocalico, Reamstown, Pa.) that were packed at 500 × g for 5 min were used to prepare a 1% suspension in normal saline. The erythrocytes (50 µl) were then added to each well containing bacteria. After mixing, the plate was covered and incubated at 4°C for 4 to 12 h and observed. The lowest dilution in which no button was visible was recorded as the hemagglutination titer for that sample.

Tracheal attachment assay. Bacterial strains were grown overnight on Bordet-Gengou agar containing nalidixic acid (30 µg/ml) and 15% sheep blood at 35°C. The bacteria were harvested in magnesium-free Earle's balanced salt solution (EBSS) (Sigma Chemical Co., St. Louis, Mo.) containing 1.0 mM CaCl₂ and diluted to approximately 2 × 10⁷ bacteria/ml, and 0.5 ml of suspension was placed in each well of a 24-well plate (Costar). Twenty-six-day-old turkey embryos were decapitated, the tracheae were removed, and transverse 2-mm-long rings were cut. Three tracheal rings were placed into each well containing EBSS with or without bacteria and incubated at 42°C for 3 h with constant rocking. Microscopic observation revealed beating cilia for 4 to 6 h postculturing. Also, scanning electron microscopy (SEM) revealed an intact ciliated layer for the same time period. Upon completion of incubation, the rings were washed with rocking three times with 1.0 ml of EBSS for 2 min each at 42°C. The rings were removed, placed individually into separate tubes containing 1.0 ml of PBS with 1% Triton X-100, incubated at 4°C for 1 to 2 h, and then mixed for 1 min on a vortex mixer to distribute the bacteria. Triton X-100, at this concentration, had no effect on bacterial viability. Dilutions were plated out onto lactose MacConkey agar and incubated for 2 days at 35°C. Resulting colonies were counted, and the numbers of CFU/tracheal ring were calculated. Tracheal rings to be used for SEM were incubated with approximately 10⁹ bacteria for 3 h. Washed rings were fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.6, overnight and prepared for electron microscopy.

Recombinant DNA techniques. Conditions for restriction endonuclease digestion, chromosomal and plasmid DNA isolation, and agarose gel electrophoresis have been previously described (3). Southern blots (33) of the Hag, Mot, and Dnt strains were prepared from EcoRI-digested B. avium chromosomal DNA and a probe derived from the kanamycin resistance gene of Tn903 (bp 581 to 1501; GenBank accession no. X06404). Probes were prepared by random priming of the BamHI-XhoI fragment with biotinylated nucleotides by using a kit from Life Technologies (Gaithersburg, Md.) and detected by chemiluminescence (Tropix, Inc., Bedford, Mass.), as directed by the manufacturer.

Electron microscopy. TEM was performed with overnight cultures grown on agar at 31°C. A drop of water containing some of the overnight growth was placed on Formvar-coated grids, the grids were rinsed, and the cells were stained with 2% phosphotungstic acid. Negatively stained preparations were examined on a Philips 910 TEM. For SEM, the glutaraldehyde-fixed tracheal ring samples were dehydrated through an ethanol-acetone series, critically point dried with liquid CO₂, mounted on stubs, and sputter coated with gold-palladium (60:40), and images were acquired by the PGTIMIX image analysis system and a Topcon ABT-35 SEM at 20 kV.

Histological methods. Tracheae were obtained at necropsy from birds at 2 weeks postinfection. Longitudinal sections were fixed in 10% buffered formalin, dehydrated, and embedded in paraffin. Sections (4 µm) were stained with hematoxylin-eosin. All microscopic histological material was viewed by an avian pathologist. Samples were coded to assure objectivity.

Statistical methods. The statistical package developed by Microsoft, EXCEL 4.0, was used to calculate parameters (e.g., standard deviations [SD] and averages) and to determine statistical significance. Specific statistical tests (e.g., analysis of variance [ANOVA], t test) are noted, where applicable, in the text.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Identification and characterization of nonmotile mutants of B. avium. Screening of several hundred Tn5lacZ2 and Tn5phoA insertion mutants of strain 197N identified four isolates that showed no motility. The motility-negative mutant (Mot) chosen for this study (strain G146) had a Tn5lacZ2 translational fusion, did not have flagella when examined by TEM, and did not react to a monoclonal antibody (11) directed against a conserved flagellin epitope from E. coli (12). In contrast, the parental strain was positive for both features. Cotransduction experiments using Ba1 and Southern blotting with a probe made to the neoR gene both indicated that a single insertion was responsible for the mutant phenotype (data not shown). We designated the locus interrupted by this insertion mot146. No revertants to the wild type were detected under in vivo selective pressure.

Identification and characterization of hemagglutination mutants of B. avium. Over 20 independent insertion mutants of strain 197N deficient in hemagglutination were identified. The hemagglutination-negative mutant (Hag) chosen for this study (strain P206) had a Tn5phoA translational fusion and required at least 30 times more bacteria than the parent to show hemagglutinating activity. Cotransduction experiments using Ba1 and Southern blotting with a probe made to the neoR gene both indicated that a single insertion was responsible for the mutant phenotype (data not shown). We designated the locus interrupted by this insertion hag206. In contrast to the stable Mot mutant, the Hag character was more problematic due to strong in vivo selection. All Hag mutants inoculated were recovered as Kan^r Hag⁺ pseudorevertants. That is, isolates still had the original insertion but also a second (undefined) lesion that restored the hemagglutination-positive phenotype (data not shown).

Identification and characterization of a DNT-negative B. avium mutant. A DNT-negative mutant of strain 197N (strain WBA16) was isolated following a recombination event between the chromosome of strain 197N and the suicide plasmid pKEW16-7 (37). Plasmid pKEW16-7 carries the B. pertussis DNT gene insertionally inactivated by a neoR gene from Tn903 (Table 1). Strain WBA16 was negative for DNT production (Dnt) when assayed in the infant mouse model (see Materials and Methods). Southern blot hybridization using a 500-bp neoR gene fragment as a probe and the antibiotic resistance phenotype of WBA16 were consistent with a single, double-crossover insertion event (37). Further, in vivo selection (ostensibly for DNT production) revealed a 100% correlation between restoration of the DNT-positive phenotype and loss of the neoR insertion. (Coreversion data was gathered from 25 Kan^s clonal isolates of WBA16, each independently isolated from 25 turkeys.) We have designated the locus interrupted by the neoR insertion dnt1. In spite of the above, we have been unable to formally conclude that the dnt1 locus is actually the structural gene for DNT. In part, this is because of the low homology between the B. pertussis dnt gene and its B. avium counterpart.

Defining parameters for differentiating virulent and avirulent B. avium in vivo. The in vivo infection protocol (Materials and Methods) was based upon observations with the parental strain 197N and one of its derivatives, strain WBA16, the Dnt mutant. In pilot experiments, these two strains represented the extremes of virulence and avirulence, respectively. Several parameters were examined: (i) the age of the birds at infection, (ii) the time required to colonize, (iii) the number of birds per group used in ID₅₀ determinations, and (iv) communicability of B. avium within groups.

View this table: [in this window] [in a new window]   TABLE 2.   Susceptibilities of birdsa at different ages to the parental strain (197N) and an avirulent (Dnt) mutant (WBA16)

View this table: [in this window] [in a new window]   TABLE 3.   Proportion of birds colonized by the parental strain (197N) compared to an avirulent Dnt mutant strain (WBA16) as a function of time

View this table: [in this window] [in a new window]   TABLE 5.   Relative communicabilitiesa of the parental B. avium strain and a nonmotile (Mot) mutant

Host susceptibility to B. avium. Turkeys are not inbred animals. However, in the United States and Great Britain, there are five major commercial turkey lines whose individual characteristics are maintained by interbreeding within the line's restricted gene pool. These five lines constitute the majority of turkeys sold worldwide. Since each line's gene pool is different, uniform susceptibility to infectious agents, across lines, cannot be assumed. To test the relative resistance or susceptibility of the five major commercial lines, ID₅₀ values for the parental strain and three mutant derivatives of strain 197N were tested. Statistical analysis (ANOVA) of the results (Table 6) indicated that there were no host-associated differences in susceptibility to any of the B. avium strains tested. This is most apparent in comparing ID₅₀ values when lines were infected with the parental strain and the Mot mutant (Table 6). Whereas there was a wide fluctuation in the degree of resistance to our avirulent Dnt strain, all lines were at least 100-fold more resistant to the Dnt mutant (Table 6). This fluctuation between lines in susceptibility to the Dnt mutant likely stems from the extremely low infection rate of Dnt mutants, necessitating extrapolation (rather than interpolation) of ID₅₀ calculations (Table 6, footnote c). All lines were resistant to the Hag mutants in the sense that isolates recovered from infected birds never retained Hag character (all were Kan^r Hag⁺ pseudorevertants). We infer from the selection of the Hag⁺ character in birds that hemagglutination is important for colonizationsuch selection does not take place during in vitro growth. Taking all of the host susceptibility data together, we conclude that both the Dnt and Hag mutants were extremely avirulent, such that ID₅₀ values were difficult or impossible (respectively) to calculate, and that lack of motility had no effect on virulence.

In vivo tracheal pathology. Turkeys naturally infected with B. avium exhibit severe microscopic histopathological signs of bordetellosis (32). Such signs were apparent in our experimental system, in which 10 tracheae were compared from groups of 3-week-old birds (10 birds per group) that were either sham infected or infected with approximately 2 × 10¹⁰ CFU of parental strain or the Dnt mutant at 1 week of age. Birds were examined at 2 weeks postinfection (examples are shown in Fig. 1). Tracheae from the sham-inoculated birds had a well-defined ciliated border with a pseudostratified columnar layer with defined goblet cells and mucous secreting cells (Fig. 1A). Tracheae from birds infected with the parental strain showed a dramatic loss of ciliated epithelial cells and other changes associated with severe tracheitis (Fig. 1B). The Dnt mutant produced much milder tracheal pathology (Fig. 1C). With a pathological scoring system, with +5 being the most severe and 0 being normal, the sham-inoculated group received an average score of +0.1, the Dnt mutant +1.8, and the parental strain-infected birds +4. In this particular experiment, all of the Dnt mutant-infected birds were colonized by Kan^s Dnt⁺ revertants by the time of evaluation. Consequently, the individual scores of Dnt mutant-infected individuals were likely dictated by the time at which revertants arose in each bird rather than by any pathology caused specifically by the Dnt mutant. This interpretation is supported indirectly by the wide variation in the scores observed in this group relative to the sham- and parental strain-inoculated groups (data not shown). Hag and Mot strains were not scored in this histological assay because of the high numbers of Hag⁺ pseudorevertants that colonized the tracheae and the negligible loss in virulence of the Mot mutant (Table 6).

View larger version (68K): [in this window] [in a new window]   FIG. 1.   Hematoxylin-eosin-stained 4-µm sections of tracheae from 3-week-old turkeys sham inoculated or exposed 2 weeks previously to parent or Dnt mutants of B. avium, as described in the text. (A) Sham-inoculated control with normal trachea. (B) Parental strain-infected turkey. Trachea shows marked disruption of mucosal architecture; absence of ciliated epithelium and goblet cells; epithelium composed of immature cuboidal cells; glands depleted of mucus; and hyperemia, mild fibrosis, and mixed inflammatory cell infiltrate of lamina propria and submucosa. (C) Dnt mutant-infected turkey. Trachea shows normal mucosal architecture; normal ciliated epithelium, except for small focal areas of deciliation (arrowheads); partial depletion of mucus from glands; and increased mononuclear cells in lamina propria. Changes are intermediate between sham-inoculated and wild-type-exposed turkeys. Bars, 100 µm.

In vitro tracheal cell adherence assay. SEM revealed numerous B. avium 197N (parental strain) cells adhering to tracheal cilia but not to nonciliated cells (Fig. 2A). The Mot strain also showed dense numbers of bacteria attached to cilia (Fig. 2B). The Dnt mutant (Fig. 2C) showed fewer bacteria, and the Hag mutant (Fig. 2D) showed much lower levels of attachment. In order to quantitate the relative ciliated cell binding efficiency of B. avium, an in vitro assay was developed. The tracheal ring adherence assay (see Materials and Methods) indicated that Dnt and Hag mutants were statistically less able to adhere to tracheal rings than were the parental strain and the nonmotile mutant (Table 7). The difference in the adherence ability between the parent and the Mot mutant was insignificant. The difference was most dramatic with the Hag mutant (Table 7). These results were remarkably similar to those obtained in vivo.

View larger version (152K): [in this window] [in a new window]   FIG. 2.   SEM of tracheae incubated with B. avium. Tracheal rings were incubated with approximately 109 bacteria/ml for 3 h under the same conditions as used for the tracheal attachment assay. (A) Parental strain 197N. (B) Mot mutant, strain G146. (C) Dnt mutant, strain WBA16. (D) Hag mutant, strain P206. All images are shown at the same magnification. Bar, 1 µm.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The studies described herein examined some of the basic parameters for experimentally infecting turkey poults with B. avium and established host susceptibility to bordetellosis across five turkey lines. Additionally, we developed a tracheal ring adherence assay using embryonic turkeys that provided an organ, species, and line-matched in vitro corollary of colonization and tracheal pathogenesis in vivo. Interestingly, we found that examinations of several B. avium insertion mutants both in vitro and in vivo produced strikingly similar conclusions about the relative colonizing ability of each mutant.

A systematic study of the factors influencing the infection rate in experimental B. avium infections was carried out in order to firmly establish the effects of several variables that could influence the results of experimental infections. To establish the initial infection parameters (i.e., age, duration, and number of birds) and to assess the histopathological consequences of infection, we employed the parental strain 197N and an avirulent insertion mutant that did not produce DNT. Whereas the Dnt mutant was constructed by site-directed mutagenesis, we have been unable to conclude that the insertion is in the structural gene for DNT. Nevertheless, its use here as an avirulent control strain proved to be quite valuable.

One of the more interesting results we obtained in testing parameters that could influence experimental infections was that B. avium did not spread extensively within groups. This finding is somewhat paradoxical for a respiratory pathogen. Nevertheless, in order to see any spreading among birds, we had to give the index case a very large inoculum and to assay for spreading at 1 week after inoculation rather than the normal 2 weeks. We found that at 2 weeks, rather than getting increased spreading, the index case was often cleared of infection. In spite of the modest spreading of the parental strain, we found significantly less spreading with a nonmotile mutant. This finding probably reflects the role of motility in the environment rather than in the bird, since B. avium is nonmotile and flagella are not detectable in vitro at the internal temperature (42°C) of turkeys (12).

One of the more practical and useful findings that emerged from this study was that the five major lines of turkeys showed no differences in susceptibility to experimental infection. This uniformity applied not only to the parental B. avium strain but to three mutant strains tested. The Hag, Mot, and Dnt mutants provided a source of strain variation that we used here as a tool to try to detect and better understand any variation in host susceptibility to B. avium. The B. avium mutants created and employed in this study, while not fully characterized at the molecular level, did present distinctly different phenotypes to the hosts. With regard to virulence, the phenotypes ranged from no loss (Mot mutants) to complete loss (Dnt mutants). The only significant technical problem encountered in the use of the mutants was with the stability of the Hag mutant, which produced Hag⁺ pseudorevertants under in vivo selective pressure. The basis for this pseudoreversion is not understood, but it was a property of all 20 Hag mutants isolated and tested. In any case, all host lines reacted uniformly to the Hag mutant used in this study (i.e., pseudorevertants were isolated from all turkeys). Curiously, earlier work by Moore et al. (25), while indicating the importance of Hag in turkey virulence (and the lack of importance of motility), did not encounter instability of the Hag character. The uniform response of all turkey lines tested suggested that practical measures for controlling the disease in one line of turkeys (e.g., development of a live vaccine strain) may be universally applicable.

Along with an in vivo model, an in vitro corollary of infection is very helpful in understanding the pathogenesis of disease at the molecular level. To our knowledge, there are no avian ciliated tracheal epithelial cell lines available. However, tracheal rings in culture have been used for a number of years (17). We utilized embryonic tracheal rings in our studies (rather than rings from live poults) because embryonic rings were easily obtained aseptically and gave uniform, reproducible results. In addition to providing an organ that is matched to the colonization site of our in vivo experiments, the rings can additionally be line matched (as they were in our experiments) so that in vitro and in vivo experiments are more closely comparable. Whereas this matching is probably not as significant a factor as originally thought, it still seems prudent to have the in vitro and in vivo experiments be as closely matched as possible. In our studies, we found that the strains that were best able to adhere to tracheal cells in culture were best able to colonize turkey poults. Whereas this finding may seem reasonable or even expected, it is remarkable from the standpoint of the different time frames involved (3 h in vitro, 2 weeks in vivo) and the number of factors coming into play in the in vivo situation (e.g., the onset of the immune response at 1 to 2 weeks postinoculation [35]).

The combination of in vivo and in vitro tools for examining avian bordetellosis and the uniform susceptibility of all commercial turkeys should facilitate development of a well-defined and practical method of preventing disease and enable a more rigorous investigation of its pathogenesis at the host and cell biological level. Also, our in vitro and in vivo studies may form a basis and provide a rationale for a more sophisticated bacterial genetic analysis of B. avium on a level comparable to those B. pertussis and B. bronchiseptica (1, 20, 23, 24, 38).