Bordetella filamentous hemagglutinin plays a critical role in immunomodulation, suggesting a mechanism for host specificity

Materials and Methods

Bacterial Strains and Growth Media. Wild-type B. bronchiseptica RB50 (41) and mutant derivatives were grown at 37°C on Bordet-Gengou (BG) agar (Becton Dickinson Microbiology Systems) supplemented with 7.5% defibrinated sheep blood (Mission Laboratories, Los Angeles) or in Stainer-Scholte (SS) broth (42). Gentamicin-resistant (RB50G) and ΔfhaB (RBX9) derivatives of RB50 have been described (28, 43). RBFS4 is an RB50 derivative in which the entire fhaB gene, from 38 bp 5′ to the ATG start codon to 56 bp 3′ to the stop codon, has been replaced with the corresponding sequences from B. pertussis Tohama 1. The construction is described in Supporting Text, which is published as supporting information on the PNAS web site. Wild-type B. pertussis strain 536 (44) and Bpe138, a mutant derivative in which the fhaB coding sequence has been replaced by a chloramphenicol-resistance gene (a gift from David Relman, Stanford University, Stanford, CA) were grown on BG agar supplemented with 13% defibrinated sheep blood or in SS broth. Western blots with anti-FHA antibody confirmed that Bpe138 did not express FHA (data not shown).

Immunoblots. Immunoblots were done as described and probed with a polyclonal antibody raised against B. bronchiseptica FHA (45, 46). For whole-cell extracts, protein extracted from the equivalent of 4.2 × 10⁸ colony-forming units (CFU) from cultures grown overnight was loaded per lane. For supernatant fractions, total protein from the supernatant of a 1-ml culture grown to an OD₆₀₀ of 2.5 for B. bronchiseptica strains and 0.75 for B. pertussis was precipitated with trichloroacetic acid, resuspended in sample buffer, and loaded per lane.

In Vitro Adherence Assays. Adherence to rat-lung epithelial (L2) cells and human bronchial (BEAS-2B) cells was performed as described in ref. 28, except that bacteria were diluted in Ham's F12K nutrient mixture (for L2 assays) or Eagle's minimum essential medium (for BEAS-2B assays) and were added at a multiplicity of infection (MOI) of 250. Adherence was quantified by counting the total number of bacteria and eukaryotic cells in at least four microscopic fields from two separate experiments. Adherence to J774A.1 cells was performed as described in ref. 28, except that bacteria were added at a MOI of 100 and were stained and visualized by using immunofluorescence. Briefly, adherent bacteria and cultured cells were washed twice with 0.1 M cacodylate buffer and fixed in 4% formaldehyde. After a blocking step using 2 mg/ml BSA, primary antibody (serum from a rat infected with RB50) was added for 1 h, wells were washed, and a goat anti-rat Cy3-conjugated IgG secondary antibody (Jackson ImmunoResearch) was added for 30 min. After several washes, coverslips were mounted on glass slides by using 5% n-propylgallate in glycerol. Adherence was quantified by counting the number of bacteria associated with 25 individual macrophages from at least two independent experiments.

Rat-Colonization and Tracheal-Adherence Experiments. Female Wistar rats (3-4 wk old) were inoculated with 1,000 CFU of Bordetella delivered in 10 μl to the external nares. Animals were killed 14 and 28 d postinoculation, and colonization levels in the nasal cavity and trachea were determined as described in ref. 27. To deliver bacteria directly to the trachea, rats were inoculated intranasally with 5 × 10⁵ CFU delivered in 50 μl. At 1 and 24 h postinoculation, nasal septa and tracheas were recovered, and the levels of colonization were determined (27).

Mouse Lung Inflammation Experiments. Groups of 6-8 BALB/c mice (3-4 wk old; Charles River Laboratories) were inoculated intranasally with 5 × 10⁵ CFU of Bordetella in a 50-μl volume. At 1 h, 3 d, 11 d, and 28 d postinoculation, three left lung lobes were recovered and bacterial CFU enumerated (27), and two right lobes were isolated, fixed by inflating with 10% formalin, embedded in paraffin, sectioned, and stained with hematoxylin and eosin (Central Histology Facility, Diagnostic Pathology Medical Group, Sacramento, CA). Data shown are the results of two independent experiments. For coinfection experiments, groups of at least three mice were inoculated with a total of 5 × 10⁵ CFU, consisting of an equal mixture of RB50G and RBX9 or RBFS4, administered intranasally and the number of CFU of RB50G and RBX9 or RBFS4 recovered at 1 h, 3 d, and 11 d postinoculation was determined by plating lung homogenates on media with and without gentamicin. Two right lobes were isolated and prepared as described above for histological examination. Data shown are the results of two independent experiments.

Measurement of Bordetella-Specific Antibody Titers. IgG levels in mouse sera were determined by using ELISA as described in ref. 27. Reactions were analyzed by using a VICTOR3 1420 microplate reader (PerkinElmer Life Sciences). Absorbance at 405 nm was plotted against dilution, and antibody titers were reported as the reciprocal of the serum dilution at the x intercept, as extrapolated from the linear part of the curve.

Infection of Severe Combined Immunodeficient (SCID)-Beige Mice. SCID-beige mice (3-4 wk old, Charles River Laboratories) were inoculated with 500 CFU of Bordetella delivered in a 5-μl volume to the external nares. Animals were monitored daily for signs of disease and killed when moribund.

Statistical Analyses. Student's unpaired t test was used for all statistical analyses.

Results

B. bronchiseptica Strain RBFS4 Expresses fhaB from B. pertussis. We constructed strain RBFS4, a B. bronchiseptica RB50 derivative in which the entire fhaB gene was replaced with the B. pertussis fhaB gene by allelic exchange (see Materials and Methods and Supporting Information). Immunoblotting demonstrated that FHA_Bp was synthesized, processed, and secreted in RBFS4 in amounts similar to that of FHA_Bb in RB50 (Fig. 1).

FHA_Bp Can Mediate Adherence of B. bronchiseptica to Epithelial Cells and Macrophages in Vitro. Wild-type B. bronchiseptica RB50 bound efficiently to epithelial and macrophage-like cell lines, and this adherence almost entirely depended on FHA (Fig. 2). RBFS4 also adhered to these cells, demonstrating that FHA_Bp is functional in B. bronchiseptica. The somewhat decreased adherence of RBFS4 compared with RB50 probably reflects an inherent difference in the binding affinities of FHA_Bp and FHA_Bb, because the contribution of FHA_Bp to adherence of B. pertussis (BP536 versus Bpe138) is less than the contribution of FHA_Bb to adherence of B. bronchiseptica (RB50 versus RBX9) and similar to the contribution of FHA_Bp to adherence of RBFS4 (RBFS4 versus RBX9).

FHA_Bp Does Not Allow B. bronchiseptica to Colonize the Rat Trachea. At 14 and 28 d after intranasal inoculation of Wistar rats, RB50, RBX9, and RBFS4 were recovered in similarly high numbers from the nasal cavity, but only RB50 was consistently recovered in high numbers from the trachea (Fig. 3A). FHA_Bp is, therefore, unable to substitute for FHA_Bb in allowing B. bronchiseptica to colonize the rat lower respiratory tract.

Lack of tracheal colonization by RBFS4 in this model could have resulted from failure to move from the nasal cavity to the trachea, to adhere to tracheal epithelia, or to resist the innate and adaptive immune mechanisms that function to keep the trachea sterile. To investigate the ability of RBFS4 to adhere to tracheal epithelium and resist initial innate immune functions, we inoculated rats intranasally with 5 × 10⁵ CFU in a volume sufficient to deliver bacteria directly to the trachea, as evidenced by the number of CFU recovered at 1 h postinoculation (Fig. 3B). Twenty-four hours later, all strains had increased to similarly high levels in the nasal cavity, but only RB50 increased significantly in the trachea (Fig. 3B). We conclude that FHA_Bp is defective in adhering tightly to tracheal epithelial cells in a manner that prevents mucociliary clearance, or in contributing to resistance to initial innate immune responses elicited in the trachea, or both.

FHA_Bb, but Not FHA_Bp, Mediates Modulation of the Inflammatory Response and Protection of B. bronchiseptica from Inflammation-Mediated Clearance. To focus on the ability of FHA to influence and/or resist innate immune responses in vivo, we used a mouse lung inflammation model. At 1-3 hours postinoculation, similar numbers of CFU of each strain were recovered from the lungs, indicating equivalent delivery of the inocula (Fig. 4A). RB50 increased in the lungs by ≈0.5 log by day 4, then decreased gradually over the next 4 wk (Fig. 4A). Slightly more neutrophils, macrophages, and lymphocytes were evident in hematoxylin and eosin (H&E)-stained lung sections from mice killed 3 d postinoculation with RB50 than in lung sections from control PBS-inoculated mice (Fig. 4B a and b). By contrast, mice inoculated with RBX9 consistently displayed a bimodal response to infection: Half of the animals remained healthy, and half became moribund by day 4. Mice that remained healthy had numbers of CFU in their lungs at days 3-4 similar to those recovered at day 0, and the infection was cleared by day 11 postinoculation. Lung sections from these mice at day 3 were similar to those of RB50-infected mice (Fig. 4Bc). By contrast, moribund RBX9-inoculated mice had extremely high numbers of bacteria in their lungs at day 3 (Fig. 4A), and H&E-stained lung sections revealed severe inflammation, with almost complete tissue consolidation composed primarily of neutrophils and lymphocytes (Fig. 4Bd). We interpret these data as indicating that FHA allows B. bronchiseptica to modulate the innate immune response, resulting in decreased inflammation and increased bacterial persistence. In the absence of FHA, we hypothesize, a robust inflammatory response is induced that either clears the bacteria from the lungs rapidly without overt signs of disease or causes inflammation-mediated lung damage so severe that it prevents clearance of the bacteria, resulting in increased bacterial growth, pneumonia, and, ultimately, death.

Mice inoculated with RBFS4 displayed the same bimodal response as mice inoculated with RBX9 (Fig. 4A), and lung sections of healthy and moribund RBFS4-inoculated mice resembled those of healthy and moribund RBX9-inoculated mice, respectively (Fig. 4B). FHA_Bp, therefore, lacks FHA_Bb's ability to allow B. bronchiseptica to modulate the inflammatory response. RBFS4 was recovered in statistically significantly higher numbers than RBX9 at day 11, however, suggesting that FHA_Bp may contribute somewhat to the ability of B. bronchiseptica to defend against inflammation-mediated clearance.

These data suggest that B. bronchiseptica, in a FHA_Bb-dependent manner, plays an active role in modulating the inflammatory response to Bordetella infection. We hypothesized that this ability should function in trans (i.e., that it should affect the overall inflammatory response in the lungs) and that it should be dominant. When coadministered in equal numbers with the wild-type strain, both RBX9 and RBFS4 exhibited significantly increased persistence compared with single-infection experiments (compare days 11 in Fig. 4 A and C). Additionally, none of the coinoculated mice displayed any signs of disease at any time; in no case was the number of CFU of RBX9 or RBFS4 recovered from the lungs >8 × 10⁵, and lung sections from all coinfected mice resembled those of RB50-infected mice (or healthy RBX9- or RBFS4-infected mice) in single-infection experiments (see Fig. 6, which is published as supporting information on the PNAS web site). The presence of FHA_Bb-expressing bacteria in trans, therefore, prevented the severe inflammation caused by infection with RBX9 or RBFS4 alone and also significantly enhanced the survival of the mutant bacteria. These results implicate FHA as playing a critical role in determining the robustness of the inflammatory response and, subsequently, the outcome of infection.

FHA_Bb Is Required for B. bronchiseptica to Induce a Strong Anti-Bordetella Humoral Response. Using ELISAs with the corresponding bacteria as antigen, we detected high titers of anti-Bordetella IgG in sera from RB50-infected mice and significantly lower titers in sera from RBX9- and RBFS4-infected mice (see Fig. 7, which is published as supporting information on the PNAS web site). These data indicate either that strains lacking FHA_Bb are unable to induce a significant humoral immune response or that FHA_Bb (but not FHA_Bp) is the predominant antigen against which the humoral response is directed. To distinguish between these possibilities, we measured anti-Bordetella IgG titers using each of the different bacteria as antigen. IgG titers in sera from RB50-infected mice were low only when RBX9 cells were used as antigen (Fig. 7), indicating that FHA_Bb is a major immunogen and that antibodies to FHA_Bb also recognize FHA_Bp. IgG titers were consistently low in sera from both RBX9- and RBFS4-inoculated mice, regardless of which bacteria were used as antigen (Fig. 7), demonstrating that a significant antibody response was not generated in mice infected with bacteria that lack FHA_Bb.

FHA_Bb Is Required for B. bronchiseptica to Overcome Innate Immunity. Our results indicate that B. bronchiseptica modulates the inflammatory response and induces a strong humoral response in an FHA_Bb-dependent manner. To test the hypothesis that innate immunity may be sufficient to control infections established by bacteria that lack FHA_Bb, we used SCID-beige mice, which lack B and T cells and have decreased natural-killer-cell function (47, 48). Consistent with data reported in refs. 49 and 50, RB50 caused a systemic infection in these mice, resulting in 100% lethality by day 88 (Fig. 5). By contrast, all RBX9-inoculated mice remained healthy for the duration of the experiment, and bacteria were limited primarily to the upper respiratory tract. Although most RBFS4-inoculated mice also remained healthy for the duration of the experiment, and the difference in time to death was not significantly different from that for RBX9-infected mice, one mouse did succumb at day 70 and another at day 93 (Fig. 5). FHA_Bb is, therefore, required for B. bronchiseptica to overcome innate immunity, and, although severely defective compared with FHA_Bb, FHA_Bp is capable of contributing somewhat to resisting inflammation-mediated clearance.

Discussion

By constructing and characterizing RBFS4, a B. bronchiseptica strain that is isogenic with RB50, except for the precise replacement of fhaB_Bb with fhaB_Bp, we found that, although FHA_Bb and FHA_Bp are functionally interchangeable with regard to their ability to mediate adherence to epithelial and macrophage cell lines in vitro, they are not functionally interchangeable in vivo. Strikingly, RBFS4 was similar, if not identical, to the ΔfhaB strain RBX9 in its inability to colonize the lower respiratory tract of rats, to modulate inflammation and persist in the lungs of mice, to induce a humoral immune response, and to overcome innate immunity and cause lethality in immunodeficient mice. These results suggest that FHA contributes to host specificity and acute-versus-chronic disease characteristics.

FHA is proposed to represent the major adhesin expressed by Bordetella (51), and, thus, the simplest explanation for why FHA_Bp is unable to substitute for FHA_Bb in vivo is that FHA_Bp fails to mediate attachment to epithelial cells in the rodent lower respiratory tract. Although our data are consistent with this hypothesis, proof will require identification of a specific receptor(s) that can distinguish FHA_Bb from FHA_Bp. If such a receptor exists, the fact that both FHA_Bb and FHA_Bp can mediate adherence to a variety of epithelial and macrophage-like cell lines in vitro suggests that this receptor may be expressed only on cilia, to which Bordetella has been shown to display a distinct predilection (52, 53), or nonciliated cells within the context of the live respiratory mucosa. Furthermore, because B. bronchiseptica displays a broad host range that includes humans, it is possible that, whereas FHA_Bp may be able to bind a receptor or a form of a receptor that is present only in humans, FHA_Bb may display less stringent binding properties and be capable of binding both human-specific and non-human-specific receptors.

FHA has been shown to activate NF-κB in epithelial cells (19, 31) and PI3-K in monocytes (54) and to induce IL-10 production and suppress IL-12 secretion in J774A.1 macrophage-like cells and bone-marrow-derived dendritic cells (34, 35), suggesting that FHA may influence the initial inflammatory response induced during Bordetella infection. Using B. bronchiseptica and natural-host animal models, we have shown that (i) RB50, but not the ΔfhaB strain RBX9, increased in number 500-fold in the tracheas of rats within the first 24 h after delivery of bacteria directly to this immune-privileged site; (ii) RB50 induced only mild inflammation and persisted for >28 d in the lungs of immunocompetent mice after a high-dose/large-volume inoculation protocol, whereas RBX9 either induced massive inflammation and grew to extremely high numbers, killing the mice within 4 d or was controlled immediately and eliminated from the lungs within ≈11 d; (iii) RB50, but not RBX9, induced a strong anti-Bordetella antibody response in immunocompetent mice; (iv) RB50, but not RBX9, caused systemic lethal disease in SCID/Bg mice; and (v) when coinoculated with RBX9, RB50 prevented an overly robust inflammatory response, resulting in limited, yet persistent, lung colonization by both RB50 and RBX9. Together, these data indicate that FHA allows B. bronchiseptica to down-regulate the innate immune response, resulting in decreased inflammation and increased bacterial persistence.

Instead of differing in adherence properties, FHA_Bp may have failed to substitute for FHA_Bb in vivo because it was unable to modulate inflammation in the rodent respiratory tract. One possibility is that FHA_Bp is inherently immunostimulatory (or at least not immunomodulatory) in vivo. Such a scenario would be consistent with the fact that B. pertussis typically causes acute infections with signs and symptoms of disease consistent with respiratory inflammation, whereas B. bronchiseptica usually colonizes its hosts chronically and asymptomatically. Also consistent with this possibility is the fact that B. pertussis appears to be unable to survive for long periods outside the mammalian respiratory tract and, therefore, depends on efficient host-to-host transmission (and, thus, FHA_Bp has evolved to promote acute respiratory symptoms), whereas B. bronchiseptica is capable of surviving conditions of extreme nutrient limitation and, therefore, may depend less on host-to-host transmission. Alternatively, FHA_Bp may be immunomodulatory in humans but not rodents, possibly because of differences in the receptors that control signal-transduction pathways involved in inflammatory cytokine expression in the different mammalian species. The ability to either stimulate or suppress inflammation in different hosts could determine not only disease characteristics but also the eventual outcome of the infection, and, therefore, such differences could contribute significantly to host specificity as well. A third possibility is that FHA_Bp is inherently immunomodulatory (even in rodents), but this ability requires that the bacteria first be able to adhere specifically to host cells, i.e., that adherence and immunomodulation are mediated by separate domains within FHA and that each domain may display different abilities to distinguish receptors from different hosts. This hypothesis is supported by the fact that multiple binding domains within FHA_Bp, with apparently different specificity, have been identified (20, 29, 30, 55-57).

The fact that none of the coinfected mice displayed any outward or histological signs of respiratory illness or inflammation indicates that RB50 was able to function in trans to prevent the robust inflammatory response that occurred during single infection with either RBX9 or RBFS4. The immunomodulatory effect of FHA_Bb, therefore, functioned in a global way, presumably by controlling the repertoire and/or amount of cytokines induced throughout the lungs. This effect was further evidenced by the fact that RBX9 and RBFS4 persisted in the lungs significantly longer in coinfection experiments than in single-infection experiments. Whether this increased persistence resulted from diminished inflammatory cell recruitment or diminished antimicrobial activity of the recruited cells awaits further studies.

FHA may also play a role in allowing Bordetella to resist clearance by phagocytic cells by inhibiting the microbicidal activity of these cells. Occupancy of CR3 and binding of LRI/IAP in a way that prevents clustering can inhibit respiratory burst in neutrophils and monocytes (58, 59), and, thus, FHA may act directly in this regard. Alternatively, or additionally, FHA may perform this function indirectly by allowing the efficient delivery of toxins, such as adenylate cyclase or proteins delivered by the Bordetella type III secretion system, both of which have been implicated in phagocyte inhibition and/or immunomodulation (50, 60-62). Although neither RBX9 nor RBFS4 persisted as well as RB50 in the lungs of coinfected mice, RBFS4 persisted better than RBX9 in both single- and coinfection experiments. These data support the hypothesis that FHA plays a role in inhibiting phagocyte-mediated clearance and suggest that FHA_Bp is capable of performing this activity, to some extent, in mice. Our characterization of strains expressing chimeric FHA proteins should allow us to determine the specific FHA domains responsible for this ability.

Our results raise two concerns with regard to studying bacteria-host interactions in general. First, the fact that FHA_Bp was able to substitute for FHA_Bb with regard to its ability to adhere to various cell lines in vitro but was unable to substitute for FHA_Bb in vivo suggests that results obtained from in vitro experiments may not necessarily reflect events that actually occur during infection. Second, our results showed that without FHA_Bb, and even with FHA_Bp, B. bronchiseptica was unable to establish a productive infection that resembled that which B. bronchiseptica causes in its natural hosts. Notably, adaptive immunity was neither induced nor required to control the infection. It is well documented that B. pertussis induces a strong anti-Bordetella-antibody response in humans. Like RBX9, RBFS4 did not induce a strong anti-Bordetella-antibody response, and adaptive immunity was not required to control infection by RBFS4. Similarly, inoculation of immunocompetent mice with B. pertussis results in only a very weak humoral immune response, and B. pertussis is unable to cause a lethal infection in SCID-beige mice (49). Together, these data (i) suggest that FHA_Bb is required for Bordetella to infect nonhuman mammals, (ii) implicate FHA as a critical factor in determining host specificity, (iii) suggest that “infections” established by B. pertussis in mice may not reflect those which occur in humans, and (iv) demonstrate the importance of studying pathogens in the context of their natural hosts.

Finally, our results have important implications with regard to vaccine development, because they suggest that FHA plays an important role in the induction or suppression of inflammation. FHA is a primary component in acellular pertussis vaccines, and it will be important to determine whether FHA is acting directly to influence inflammation and the induction of adaptive immunity, because both the local effects resulting from immunization and the development of long-lasting immunity could be affected.

Supplementary Material