Pertactin contributes to shedding and transmission of Bordetella bronchiseptica

Whooping cough is resurging in the United States despite high vaccine coverage. The rapid rise of Bordetella pertussis isolates lacking pertactin (PRN), a key vaccine antigen, has led to concerns about vaccine-driven evolution. Previous studies showed that pertactin can mediate binding to mammalian cells in vitro and act as an immunomodulatory factor in resisting neutrophil-mediated clearance. To further investigate the role of PRN in vivo, we examined the functions of pertactin in the context of a more naturally low dose inoculation experimental system using C3H/HeJ mice that is more sensitive to effects on colonization, growth and spread within the respiratory tract, as well as an experimental approach to measure shedding and transmission between hosts. A B. bronchiseptica pertactin deletion mutant was found to behave similarly to its wild-type (WT) parental strain in colonization of the nasal cavity, trachea, and lungs of mice. However, the pertactin-deficient strain was shed from the nares of mice in much lower numbers, resulting in a significantly lower rate of transmission between hosts. Histological examination of respiratory epithelia revealed that pertactin-deficient bacteria induced substantially less inflammation and mucus accumulation than the WT strain and in vitro assays verified the effect of PRN on the induction of TNF-α by murine macrophages. Interestingly, only WT B. bronchiseptica could be recovered from the spleen of infected mice and were further observed to be intracellular among isolated splenocytes, indicating that pertactin contributes to systemic dissemination involving intracellular survival. These results suggest that pertactin can mediate interactions with immune cells and augments inflammation that contributes to bacterial shedding and transmission between hosts. Understanding the relative contributions of various factors to inflammation, mucus production, shedding and transmission will guide novel strategies to interfere with the reemergence of pertussis.

Introduction Pertussis, or whooping cough, is an acute respiratory disease caused by the gram-negative bacterial pathogen Bordetella pertussis. Historically, pertussis had a high morbidity and mortality rate and was the predominant childhood killer before the introduction of whole cell vaccines (WCV) that greatly reduced its prevalence [1]. In highly vaccinated populations, the reduced incidence of severe disease led to attention being focused on the moderate side effects associated with WCV, including common local reactions (swelling and pain), uncommon systemic reactions (fever, irritability, drowsiness, loss of appetite and vomiting), and very rare neurological reactions (acute encephalopathy in newborns) [2]. To address these concerns, less reactive pertussis acellular vaccines (ACV) containing up to five purified, detoxified B. pertussis proteins, including pertussis toxoid, pertactin, type 2 and 3 fimbriae, and filamentous hemagglutinin, were increasingly used in the 1980s and 1990s, and eventually replaced the WCV in most developed countries [3]. Coincident with the switch to ACV, the incidence of whooping cough has been increasing in all age groups, from infants to adults [4]. Moreover, in recent years several countries have seen an increased percentage of clinical isolates that fail to express pertactin (PRN), a prominent outer membrane protein [5,6]. The loss of PRN, a prominent ACV component, has led to concern that ACV vaccines are driving the evolution of B. pertussis to evade vaccine-induced immunity [7]. The long-term consequences of this possibility are difficult to predict as we have limited knowledge of the biological functions of PRN.
PRN is an autotransporter protein that appears on the outer membrane of B. pertussis [8]. This protein has been included in acellular vaccines in several countries, based on evidence that pertactin specific antibodies contribute to protection against pertussis. Previous studies revealed that PRN exerts cell binding through its RGD motif and contributes to adherence and invasion of mammalian cells [9][10][11][12]. However, the adhesion functions of pertactin were not detected in some other studies [13][14][15][16][17]. In vivo studies have shown an immunomodulatory role of PRN [17,18], in the context of severe pneumonic infection initiated by super-natural challenge with roughly a million CFU of B. pertussis or B. bronchiseptica delivered deep into the lungs of mice where they caused severe lung pathology. This experimental system was established to study the most extreme form of disease and has enabled the development and testing of vaccines and therapeutics to prevent and treat severe disease in the lower respiratory tract. But B. pertussis is highly infectious, suggesting very small initial inocula, and natural infections begin with the weeks-long catarrhal stage, which is limited to the nasopharyngeal region and is a highly contagious period. Thus, colonization with low initial numbers followed by weeks of growth within and shedding from the upper respiratory tract are critical aspects of the infectious cycle of B. pertussis. However, these aspects are not simulated in the standard inoculation approach that delivers as many as a million bacteria deep into the lungs and measures the consequences of the subsequent extreme pneumonic pathology [19][20][21].
We have focused on simulating the progression of natural infections in order to study the contributions of individual bacterial factors and host immune functions [22][23][24][25]. We use B. bronchiseptica because it naturally and highly efficiently colonizes, grows, causes pathology, is shed, and transmits between mice. Importantly, it can efficiently colonize mice when delivered to the external nares in very small numbers (<5 CFU), allowing the entire natural progression of the infection to be studied, including many aspects that are obviated by the conventional approach of washing millions of bacteria deep into the lungs [22][23][24][25]. B. pertussis and B. bronchiseptica are very closely related and shared genes are~98% identical at the nucleotide level, but differ in some notable regards. While most of the best studied factors are shared, each expresses a somewhat different subset of factors [5,[26][27][28][29][30][31], that are likely to explain their differences in host range and/or the diseases they cause: while B. pertussis cause an acute and severe disease in humans, B. bronchiseptica causes infection that can range in severity and often persists in the upper respiratory tract of infected animals. However, similarities far outbalance differences. Both species infect mammals and, with the notable exception of expression of the pertussis toxin, share nearly all genes implicated in interacting with their hosts, including the vaccine antigens fimbriae (FIM), filamentous hemagglutinin (FHA) and both species express PRN orthologs that are 92% similarity of PRN at the amino acid level and are likely to perform similar functions [31].
The study of the true in vivo functions of any of Bordetella spp. factors requires experimental systems in which their effects can be measured. Unfortunately, B. pertussis poorly colonizes mice, making studies of the details of natural infections challenging. Unlike B. pertussis, B. bronchiseptica efficiently colonizes mice with an infectious dose less than 5 CFU with infections progressing efficiently, allowing the study of the mechanistic basis for their complex interactions in the context of naturally progressing infection in the natural host [22][23][24][25]. We also previously described B. bronchiseptica efficiently transmitting among TLR4-deficient C3H/HeJ mice. B. bronchiseptica grows and is shed in higher numbers in these mice, allowing more efficient transmission [22][23][24][25]. This experimental model allows all the bacterial components necessary for efficient transmission to be experimentally manipulated to better understand bacterial mechanisms involved in the transmission process.
In this study, to investigate the biological role(s) of PRN in the infectious process and pathogenicity of Bordetella species, we used a mutant of B. bronchiseptica with an in-frame deletion in the prn gene which was generated by allelic exchange as described by Inatsuka et al. [17]. In vivo experiments showed that prn was not necessary for efficient colonization or early growth in the nasopharynx, but was required for efficient bacterial shedding and transmission between mice. Furthermore, wild-type B. bronchiseptica induced higher levels of inflammation and more mucus secretion than the Δprn mutant, revealing a role for PRN in promoting inflammation and mucus secretion, which was supported by in vitro assays showing higher secretion levels of the pro-inflammatory cytokine TNF-α from WT infected compared to Δprn mutant infected murine macrophages. Moreover, the WT strain, but not the prn deletion strain, was recovered from splenocytes of infected mice, indicating that PRN contributes to to systemic dissemination that involves intracellular survival. Together these data suggest that PRN promotes the induction of inflammation and mucus production, mediates shedding and transmission to new hosts and is involved in the intracellular survival and systemic dissemination of the pathogen.

PRN is not necessary for B. bronchiseptica to efficiently colonize the host and grow within the respiratory tract
To evaluate the contribution of pertactin to various aspects of the biology of B. bronchiseptica, we used an isogenic prn gene deficient mutant (BbΔprn) of B. bronchiseptica [17]. The parental strain RB50 (Bb WT) is well-established as being highly efficient in colonizing, persisting, and transmitting among mice [22][23][24][25]. To confirm the deletion was clean and not complicated by other changes, we re-sequenced the whole genome and confirmed the in-frame deletion of prn (S1 Fig). The prn mutant strain showed no defect in in vitro growth, adherence to human alveolar epithelial cells or cytotoxicity to murine RAW 264.7 macrophages (S2-S4 Figs). Using the conventional pneumonic infection model, C57BL/6 mice were inoculated intranasally with 50 μL PBS buffer containing 5x10 5 CFU of either WT or BbΔprn bacteria. Bacterial numbers in nasal cavities, trachea and lungs harvested at 7-, 14-and 28-days post-inoculation (dpi) did not differ significantly between mice infected with Bb WT or BbΔprn, (S5 Fig). Interestingly, BbΔprn showed higher colonization levels in the nasal cavity and lungs at 14 dpi, but these differences were not observed in later timepoints. Histopathological analysis of nasal cavities at 7 dpi and 14 dpi detected mild suppurative inflammation in both groups with similar incidence and severity. Thus, the standard high dose pneumonic infection experimental system did not reveal a significant role for PRN in the pathogenesis of B. bronchiseptica. Rather than extending the use of this conventional experimental system that generates extreme lung pathology but poorly simulates natural infection, we considered newer assays and approaches we have developed.

PRN contributes to transmission
We recently developed an experimental system to study various aspects of the process of transmission between hosts using C3H/HeJ mice [22][23][24][25]32]. To relate shedding and transmission to other aspects of the infectious process often assumed to affect shedding, such as bacterial load, we first assessed the time course of infection in these mice. C3H/HeJ mice were inoculated with 150 CFU of either Bb WT or BbΔprn in 5 μL of PBS, a volume that deposits the bacterial inoculum only into the nasal cavity. Respiratory organs were harvested at 7-, 14-and 28-dpi. Both bacteria grew over time, with similar numbers observed in nasal cavity, trachea and lung (S6 Fig), suggesting that PRN has no critical role in colonization, growth, and spread within the respiratory tracts of these mice.
To assess the contribution of PRN to transmission, pairs of C3H/HeJ mice were inoculated with 150 CFU of either Bb WT or BbΔprn strain (in index mice) and then co-housed with two naïve C3H/HeJ mice (uninfected recipient mice) in cages of four. To evaluate bacterial transmission, the nasal cavities of 12 such co-housed naïve mice per strain (from 6 cages) were assessed for bacterial colonization after 21 days of co-housing with infected mice. In the Bb WT group, all 12 co-housed naïve mice became colonized, while BbΔprn bacteria were only transmitted to 5 out of 12 co-housed naïve mice (Fig 1A), suggesting that PRN contributes to bacterial transmission between hosts.

PRN contributes to bacterial shedding
The deficiency of BbΔprn in transmission could be due to either a defect in shedding from infected hosts or colonization of exposed hosts. To test the efficiency of colonization, we inoculated C3H/HeJ mice with decreasing doses of bacteria. Both Bb WT and BbΔprn efficiently colonized the respiratory tracts of all mice with a calculated inoculation dose of~5 CFU (S7 Fig) indicating that PRN is not required for efficient colonization of the respiratory tract.
Since PRN contributes to transmission but is not required for efficient colonization, we investigated its effect on shedding from the noses of challenged mice. C3H/HeJ mice (n = 12) were inoculated with 150 CFU of either Bb WT or BbΔprn, and bacteria shed from the nose were collected by gently swabbing the external nares with Dacron-polyester tipped swab every two-or three-days post inoculation. Throughout the experiment, WT bacteria were shed from the nares at high numbers, > 1000 CFU at multiple consecutive timepoints, indicating prolonged, high level shedding. In contrast, the BbΔprn strain was shed at 10 to 100-fold lower numbers from the first day of shedding to the end of the experiment at day 21 ( Fig 1B), revealing a critical role for PRN in efficient shedding from the nose to the environment.

PRN contributes to inflammation in the nasal cavity
The striking difference in shedding, without a substantially different load of bacteria in the respiratory tract, led us to speculate that PRN might induce shedding by altering the inflammatory state and/or mucus production. To test this hypothesis, C3H/HeJ mice infected with Bb WT or BbΔprn were collected at 7 and 14 dpi for histopathological analysis. Histopathology showed that in the WT group, all 10 analyzed mice had inflammation in the nasal cavity ranging from mild to severe (Fig 2G-I). In contrast, of the BbΔprn infection group, 2 out of 10 had no apparent inflammation and the other 8 had severity scores which varied from minimal to mild (Fig 2D-F). These data indicate that PRN is involved in induction of inflammation in the nasopharynx (Fig 3A and S1 Table). To probe the effect of PRN on immune cell recruitment, flow cytometry was used to analyze immune cell populations in nasal cavities 14 days after inoculation with Bb WT or BbΔprn (Fig 4). Bb WT induced a significant increase in numbers of neutrophils, NK cells, macrophages, and B cells in the nasal cavities (Fig 4C-F). In contrast, BbΔprn induced an increase in macrophages, but no significant increase of other cell types compared to the uninfected control (Fig 4). These results indicate that PRN plays a role(s) in activation and recruitment of neutrophils, NK cells, and B cells, leading to higher inflammation levels in the nasal cavity.

PRN increases mucus secretion in acute inflammation
The dramatically higher number of bacteria shed from mice infected with Bb WT than those given BbΔprn led us to examine whether increased mucus production might be involved. To compare mucus secretion in the nasal cavity of Bb WT infected and BbΔprn infected mice, Alcian blue-Periodic acid Schiff (PAS) staining was performed on tissue sections. At 7 dpi all 5 Bb WT infected mice displayed moderate mucus secretion. In contrast, only 2 out of 5 BbΔprn infected mice displayed some mucus accumulation and only in a small area, while the other 3 showed no mucus accumulation at all (Fig 3B). The substantially higher mucus secretion in Bb WT infected mice compared to BbΔprn infected mice indicates that PRN plays a role in the induction of mucus production during the acute phase of infection.

PRN induces secretion of pro-inflammatory cytokines
Because of the very different numbers of neutrophils we observed in the noses of the two groups of mice, we hypothesized that PRN may affect the induction of two pro-inflammatory cytokines that recruit and activate neutrophils, IL-1 and TNF-α [33][34][35]. We harvested the nasal cavities of mice inoculated with Bb WT or BbΔprn at 24-and 48-hpi and observed somewhat higher levels of IL-1β and TNF-α in the former, albeit without statistical significance (S8 Fig). To test the effect of PRN on cytokine secretion from macrophages, we exposed macrophages to Bb WT or BbΔprn and observed higher levels of TNFα in the former group, suggesting that PRN promotes secretion of TNF-α, but did not affect IL-1β (Fig 5 and S9 Fig).

PRN contributes to intracellular survival and systemic dissemination
Bb has been shown to be able to invade and persist in immune cells, which can have profound effects on the local and systemic immune responses [36][37][38]. Since PRN has been implicated in interactions with immune cells, we tested its effects on access to and invasion of immune organs [17,39]. We harvested spleens from mice inoculated with 150 CFU of either Bb WT or BbΔprn at 7-, 14-and 28-dpi. Bb WT was recovered from 3 of 4 spleens and 5 of 8 spleens of infected mice on days 7 and 14, respectively, while no bacteria were found in the spleens of any BbΔprn inoculated mice (Fig 6A). To examine whether Bb WT in the spleen were surviving intracellularly, single cell suspension of splenocytes were treated with gentamicin treatment to kill the extracellular bacteria. Gentamicin killed 100% of control bacteria from culture (without host cells present), but did not kill all of those recovered from spleen, indicating that some were within host cells (Fig 6B), indicating that Bb WT expressing PRN can survive within the splenocytes. Additionally, we detected a higher IgG titer in Bb WT infected mice compared to that in BbΔprn infected mice (S10 Fig), suggesting that a strong immune response was induced to against this systemically disseminated infection.

Discussion
Previous studies have used different and unrelated clinical isolates of B. pertussis with or without PRN, demonstrating, for example, that an isolate lacking PRN sustained a longer infection in mice immunized with an acellular pertussis vaccine [40]. Such differences cannot be specifically attributed to the loss of PRN, since the genomes of these two separate isolates differ in many other genes. Moreover, because B. pertussis poorly colonizes, sheds, and transmits between mice, it has been difficult to assess the role of PRN in these critical aspects of the infectious cycle. To address these problems, we and Inatsuka et al. used an in-frame deletion of prn in B. bronchiseptica, a close evolutionary progenitor of B. pertussis that naturally infects mice, to test the roles of PRN in infections of a natural host. In the conventional Bordetella spp. experimental system in which a very high inoculation dose (5 x 10 5 CFU) delivered in a large volume (50 μL) to deposit bacteria deep in the lungs, PRN had a modest effect in BALB/c mice in the prior study [17]. In our high dose pneumonic challenge experiment, using different mice (C57BL/6) assessed at different time points, we did not observe measurable defects of the prn mutant. Since BALB/c and C57BL/6 mice have different immune responses to B. pertussis infection and vaccination [41][42][43], the different genetic backgrounds of two mouse strains and/or the different time points may explain the different observations in these two studies. Rather those extensive comparisons between experimental conditions using the conventional experimental system, which is an extreme model of severe pneumonic disease, we directed our efforts toward assays that are more sensitive to various aspects of the infection and transmission processes. To better simulate a more natural, gradually progressing Bordetella spp. infection, we used a low dose inoculation model in which an experimental dose as low as 5 CFU is delivered into the external nares of mice in a 5μl droplet that deposits bacteria only in the nose. In this more natural experimental system, PRN was not required for efficient colonization and spread within inoculated hosts. When delivered only into the nasal cavity, rather than deep into the lungs, Bb WT-inoculated mice shed significantly more bacteria and transmitted between mice much more efficiently than BbΔprn-inoculated mice (Fig 1). Bb WT also induced significantly more inflammation and mucus production than RB50Δprn. Inflammation and mucus production contribute to the rhinorrhea that facilitates shedding and transmission, as noted previously with Streptococcus pneumonia transmission [44]. The significantly higher number of neutrophils were observed in the nares of Bb WT infected mice, suggesting that PRN may play critical roles in recruitment and activation of neutrophils and/ or in resistance to neutrophils-mediated clearance, as was observed by Inatsuka CS et al [17]. PRN induced secretion of TNF-α from RAW 264.7 macrophages in in-vitro, which may explain the higher inflammation levels observed in mice infected with Bb WT. The role of TNF-α in the transmission of B. bronchiseptica deserves more attention in future studies as this factor may work as a potential therapeutic target for blocking transmission of Bordetella spp. and other pathogens.
PRN, as an outer surface protein with a demonstrated ability to affect adhesion to mammalian cells in vitro [9][10][11], is generally considered an "adhesin". However, it is not clear that PRN contributes to adhesion to either bronchial or laryngeal cells in vivo. In this study, deleting PRN did not affect the ability of B. bronchiseptica to efficiently colonize mice with a remarkably small inoculum, indicating that PRN is not required for the efficient adherence involved in initial colonization. However, Bb WT showed significantly higher recovery from spleens of infected hosts, indicating that PRN affects the ability to get to and/or survive in the spleen. The observation of substantial numbers of Bb WT, but not BbΔprn, within splenocytes (Fig 6), suggests PRN might affect invasion of immune cells and/or phagocytic killing within splenocytes. Upon encountering invading bacteria, immune cells trigger cascades of inflammatory responses by secreting cytokines, chemokines, small lipid mediators (SLM), and antimicrobial peptides (AMPs) that can contribute to increased phagocytic capacity and bacterial clearance [45], amongst other potential effects. After phagocytosis, peripheral immune cells can carry engulfed bacteria to deeper tissues including draining lymph nodes for T cell priming [46]. Thus, systemic dissemination of Bb WT, but not BbΔprn, may result in a stronger immune response against the former, consistent with the histopathology analysis and IgG antibody titers observed (Fig 2 and S10 Fig). In addition, the role of PRN in inducing TNF-α secretion may also contribute to the higher inflammation observed in the nasal cavity of Bb WT infected mice (Fig 5).
The increasing prevalence of PRN-deficient B. pertussis strains raises questions about both positive and negative (purifying) selection. It appears increasingly likely that PRN-containing ACVs select for PRN-deficient B. pertussis. It is also likely that the epidemiology of pertussis has changed as host behavior, population density, and worldwide travel affect the network of connected hosts. Even though PRN-deficient B. pertussis isolates may be shed less, a dense and well-connected population of vaccinated hosts in countries like the USA may be sufficient for a successful chain of transmission. Alternatively, the loss of PRN may have less cost in B. pertussis due to compensation by other genes. There are 15 other autotransporter genes in the genome of B. pertussis [5,31], one or more of which may compensate for the loss of PRN. Since PRN plays roles in systemic dissemination and induction of inflammation in murine transmission models (Figs 2-6), other factors facilitating systemic dissemination and promoting inflammation may compensate for its loss. Alternatively, C3H/HeJ mice may fail to mimic immunocompetent hosts in every aspect of pathogenesis or B. pertussis may be different from B. bronchiseptica in various aspects, and it is possible that the functions of PRN in these two species may differ to some extent. However, to test this, an efficient B. pertussis transmission model is needed. In addition, further studies are needed to identify factors that might compensate for the loss of PRN in current circulating strains. This might lead to the identification of alternative antigens that could be included in next generation vaccines against B. pertussis.

Ethics statement
Mouse experiments used in this study were performed in strict accordance to recommendations outlined in the Guide for Care and Use of Laboratory Animals of the National Institute of Health. Protocols were approved by the Institutional Animal Care and Use Committee at University of Georgia (A2016 02-010-Y2-A3, Bordetella-Host Interactions). Mice were closely monitored during experiments and any mouse found moribund was euthanized using CO 2 inhalation to prevent unnecessary suffering.

Adherence assay
Adherence assays were conducted following protocols described earlier [12,47]. In brief, human epithelial lung A459 cells were seeded in triplicate in 24-well plates at a density of 2.5 × 10 5 cell/well in Dulbecco's modified Eagle's medium (DMEM) (supplemented with 10% fetal bovine serum, 10 mM glutamine, 25 mM sodium pyruvate, 10 mM HEPES). Log-phase bacteria were suspended in warm DMEM medium and added to each well at a multiplicity of infection of 10:1 (bacteria: eukaryotic cells). The plate was centrifuged at 300Xg for 10 minutes to synchronize infection and the assay plates were incubated for 5 minutes at 37˚C. Unattached bacteria were then removed by washing the cells 4 times with 1 mL phosphate-buffered saline (PBS). A459 cells were lysed with 100 μL of 0.1% sodium deoxycholate for 5 minutes and released bacteria suspended in 900 μL of PBS. The bacteria were enumerated by dilution plating on BG agar plates. RB50 ΔfhaB [12], a mutant strain deleted of the gene encoding the filamentous hemagglutinin and known to be defective in adherence, was used as the negative control.

Cytotoxicity assay
Cytotoxicity assays were conducted on RAW 264.7 cells, using the CytoTox 96 Nonradioactive Cytotoxicity Assay Kit (Promega) following manufacturer's protocols. In brief, 100 μL of 2.5 × 10 4 macrophages were seeded in triplicate in a 96-well plate followed by adding bacteria at a multiplicity of infection of 10:1. The assay plate was centrifuged at 300Xg for 10 minutes. Bacteria were incubated with the macrophages for 4 hours, following which the plate was centrifuged for 5 minutes (300Xg). 50 μL of the supernatant was placed into a fresh flat-bottomed 96-well plate and was calorimetrically assayed for lactate dehydrogenase. A noninfected group was used as a negative control.

Cytokine test
To assess secretion of TNF-α and IL-1β from RAW 264.7 macrophages, the supernatant of infected cells (MOI = 10 or MOI = 100) was collected at 1-or 4 hpi. To assess the levels of TNF-α and IL-1β in the noses of mice infected with Bb WT or BbΔprn, noses were collected in PBS and homogenized using a bead tissue disruptor. Cytokine levels were determined using the commercially available ELISA kits R&D Systems Mouse IL-1 beta DuoSet ELISA and TNF-alpha DuoSet ELISA following the manufacturer's instructions.

Mouse infections
All work with mice was conducted following institutional guidelines. Six-week-old female C57 BL/6 or C3H/HeJ mice (Jackson Laboratories, Bar Harbor, Maine) were used for assessing the colonization and the progress of infection of the respiratory tract. As required, 5-150 CFU of bacteria was delivered in 5 μL of PBS to the nares of mice anesthetized with isoflurane/oxygen. For the colonization profiles, groups of 4 mice were inoculated with WT or mutant bacteria, and at the indicated days 4 mice of each group were euthanized with carbon dioxide (CO 2 ) and the nasal cavity, trachea, and lungs were collected in PBS and homogenized using a bead tissue disruptor. Bacterial load was enumerated by dilution plating.

Transmission and shedding assay
Transmission assays were conducted using the transmission permissive C3H/HeJ (TLR4 deficient) strain of mouse (Jackson Laboratory) whereby infected (index) mice were co-housed with uninfected (naive) mice [25]. In brief, mice were lightly anaesthetized with isoflurane/ oxygen and inoculated with 150 CFU of bacteria delivered in 5 μL of PBS onto the nares. Inoculated (index) mice were then placed in cages with 2 uninfected (naive) mice. Transmission of B. bronchiseptica was assessed after 3 weeks of co-housing by enumerating the bacterial load in the nasal cavities of the naive mice. To monitor shedding, the external nares of the index mice were swabbed (32 swipes) with a dry Dacron polyester tipped swab at the indicated times. The swab was vortexed vigorously in 1 mL PBS for 30 seconds and bacteria enumerated on BG agar plates.

Histopathology
Following fixation in neutral-buffered, 10% formalin solution and subsequent decalcification in Kristensen's solution, coronal sections were made through the nose. Tissues were subsequently processed, embedded in paraffin, sectioned at approximately 5μm, and stained with hematoxylin and eosin. Histopathological examination consisted of evaluation of the nose for the incidence (presence or absence), severity, and distribution of inflammation. Histopathologic severity scores were assigned as grades 0 (no significant histopathological alterations); 1 (minimal); 2 (mild); 3 (moderate); or 4 (severe) based on an increasing extent and/or complexity of change, unless otherwise specified. Lesion distribution was recorded as focal, multifocal, or diffuse, with distribution scores of 1, 2, or 3, respectively [48].

Flow cytometry
Five mice per experimental group were euthanized by CO 2 inhalation. Nasal cavities were harvested, placed in 1mL of RPMI 1640 medium and homogenized via a syringe plunger against a 40μm cell strainer. Cell suspensions were centrifuged at 1,500rpm for 10 minutes and remaining red blood cells were lysed with ACK lysing buffer. After washing with PBS, the cells were incubated with 1μl Zombie aqua (Biolegend) for 20 minutes, washed again, and incubated with 1μl Fc Block (Biolegend) for 30 minutes. Surface marker staining was added to each sample from a master mix of antibodies. Cells were fixed, washed, and resuspended in 250μl FACS buffer. Flow cytometry (Acea Novocyte Quanteon) was performed was used to sort neutrophils (CD11b-Pe-Cy7, CD115-APC, Ly6G-AF488) and macrophages (CD11b-PE-Cy7

Cell isolation for intracellular survival test
The spleen organs harvested from WT or mutant infected Hej mice was cut into 3-4 pieces using sterile scissors. Spleen pieces were mashed with the rough surfaces of two sterile frosted microscope slides and slide surfaces were rinsed with 1 mL (FBS-free) DMEM medium. The samples were then passed through a 70 μm cell strainer to obtain a single-cell suspension. A 100μl sample was plated on BG agar containing 20 μg/ml streptomycin to estimate the number of total bacteria. Cells were washed twice with PBS and then incubated for 1 h with 300 μg/ml gentamicin to eliminate extracellular bacteria. Cells were washed twice with PBS and then lysed with 0.1% triton X-100 for 15 minutes at room temperature. CFU counting was performed on cell lysates by plating 10-fold serial dilutions onto BG agar plates containing 20 μg/ ml streptomycin to estimate the number of intracellular bacteria. Control experiments to assess the efficacy of gentamicin were performed in parallel [49]. Briefly, samples of 10 3 , 10 4 or 10 5 bacteria were incubated with 300 μg/ml gentamicin for 1 hour at 37˚C and plated on BG agar. There were 3 replicates for each group. The results showed that more than 99.9% bacteria were killed.

Statistical analysis
Statistical analysis of differences between the WT and mutant groups was performed using the Unpaired Student 2-tailed t test, One-way ANOVA and Two-way ANOVA test, as appropriate. P value less than 0.05 shown as � ; P value less than 0.01 shown as �� ; P value less than 0.001 shown as ��� ; P value less than 0.0001 shown as ���� . Adherence to A549 lung epithelial cells of Bb WT (blue), BbΔprn (orange) and BbΔfha (yellow) that was previously shown to be impaired in its ability to adhere to epithelial cells. There were 3 replicates in each group. Error bar shows the standard error of mean. Statistical significance was calculated using One-way ANOVA. ��� p < 0.001. (TIF) A) The IL-1β detected in the supernatant when RAW macrophages were exposed with Bb WT (blue) or BbΔprn (orange) with a MOI at 10 for 1 hour. B) The IL-1β detected in the supernatant when RAW macrophages were exposed with Bb WT (blue) or BbΔprn (orange) with a MOI at 100 for 1 hour. Error bar shows the standard error of mean. Statistical significance was calculated using unpaired T test. (TIF) S10 Fig. PRN contributed to the generation of anti-B. bronchiseptica IgG antibodies. IgG antibody titers against B. bronchiseptica were determined in sera of C3H/HeJ mice infected with either Bb WT (blue) or BbΔprn (orange) at 28 dpi. There were 4 mice per group. Error bars show the standard error of mean. Statistical significance was calculated by using One-way ANOVA. � p < 0.05, �� p < 0.01, ��� p < 0.001. (TIF) S11 Fig. Gating strategy for flow cytometry. A) Gating strategy for myeloid cell types. B) Gating strategy for lymphoid cell types. (TIF) S1 Table. Scoring standard for histopathology analysis.