Nucleotide-binding oligomerization domain containing-like receptor family, caspase recruitment domain (CARD) containing 4 (NLRC4) regulates intrapulmonary replication of aerosolized Legionella pneumophila

Background Legionella pneumophila (Lp) flagellin activates signaling pathways in murine macrophages that control Lp replication. Nucleotide-binding oligomerization domain (NOD) containing-like receptor (NLR) family, caspase recruitment domain (CARD) containing 4 (NLRC4) and Toll-like Receptor (TLR5) both recognize Lp flagellin in vitro, but whether these two receptors play redundant or separate functional roles in vivo is unknown. Methods The immune response of Nlrc4−/−, Nlrc4−/−/Tlr5−/−, and wild type C57Bl/6 mice was analyzed after in vivo infection with aerosolized Lp. Results Lp clearance from the lungs was delayed in Nlrc4−/− mice over seven days in comparison to wild type controls. Nlrc4−/−/Tlr5−/− mice had no additional defect. In contrast to TLR5, NLRC4 did not regulate recruitment of neutrophils to the lung. Although there were no differences among the mouse strains in the lung transcriptome at 4 hours, Nlrc4−/− and Nlrc4−/−Tlr5−/− mice had increased lung inflammation at 72 hours in comparison to WT. Nlrc4−/−/Tlr5−/− mice also had altered cytokine production at both 4 and 24 hours post infection when compared to wild-type (WT) and Nlrc4−/− mice. Lp replication in murine alveolar macrophages was NLRC4-dependent and TLR5-independent. Conclusion These studies reveal that NLRC4 and TLR5 mediate different roles in the inflammatory response to Lp flagellin in an aerosolized infection model and NLRC4 regulates replication in both lungs and alveolar macrophages.


Background
Legionella pneumophila (Lp), an intracellular gramnegative rod, is a frequent cause of severe pneumonia [1]. Lp replicates and survives in the environment by infecting and persisting in waterborne amoeba and human infection occurs following exposure to aerosols of contaminated water [2]. Despite the discovery of Lp over three decades ago, we are just beginning to understand the role that innate recognition of molecules such as flagellin plays in host defense to the pathogen.
Lp is recognized by several classes of innate immune receptors that regulate different steps in the immune response. Extracellular and endosomal receptors (Toll-like receptor (TLR)-4, TLR2, TLR9 and TLR5) and cytoplasmic receptors, NLRC4, Neuronal apoptosis inhibitory protein-5 (NAIP5), Retinoic acid inducible gene -I, Melanoma differentiation associated protein-5, NLR family, pyrin domain containing-7 (NLRP7), NOD1, and NOD2) detect Lp in the mouse and regulate the immune response to the pathogen [3][4][5][6][7][8][9][10][11][12][13]. Two receptor complexes, TLR5 and NLRC4/NAIP, detect Lp flagellin (FlaA) [4,[14][15][16]. Through a series of in vitro studies of Lp replication in murine bone marrow derived macrophages (BMDM), NAIP5 was identified as a key regulator of bacterial growth [17,18]. Later studies showed that NAIP5, NLRC4, and apoptosis-associated speck protein containing CARD (ASC) oligomerize into an inflammasome complex after recognition of cytoplasmic flagellin leading to enhanced caspase-1 dependent IL-1β secretion and pyroptotic cell death [10,11,14,16]. In addition in the mouse Lp flagellin is detected by multiple NAIP orthologs (NAIP5 and NAIP6), causing oligomerization of the NLRC4 inflammasome and proteolytic activation of caspase-1 [19,20]. While replication of Lp in cultured murine BMDM is controlled through activation of the NLRC4/NAIP inflammasome, the role of the inflammasome in mediating resistance to Lp in alveolar macrophages and in the lungs of whole animals is only partially understood. Respiratory challenge studies with wild type and flagellin-deficient (FlaA-) Lp indicated that flagellin recognition is important for control of Lp replication in the lungs [4,15,21,22]. The A/J mouse, which has impaired signaling through the NLRC4/NAIP5 inflammasome, permits intrapulmonary replication of Lp, in contrast to mice with intact NLRC4 inflammasome activation [23]. In addition, pulmonary clearance of Lp is impaired in Nlrc4−/− mice, a finding that is reproduced by flagellin deficient organisms [21,22]. Therefore we know that the NLRC4/NAIP5 inflammasome helps to clear Lp from the lung, but the mechanism of clearance is largely unknown, and whether other flagellin detectors such as TLR5 play a redundant role also is unknown.
FlaA expression in Lp is a major virulence factor and important in the early host response to the organism [15,24]. Lp tightly controls the expression of flagellin to enhance its infectivity and suppression of flagellin expression is important in intracellular replication in the macrophage [25,26]. Most experiments identifying flagellin as an important determinant in macrophages, however, have been performed in bone marrow derived macrophages differentiated ex vivo, and the role that tissue macrophages such as the alveolar macrophage play in control of Lp replication in vivo remains largely unknown.
In this study we examine the in vivo immune response to aerosolized Lp infection in Nlrc4−/− and Nlrc4 −/−/Tlr5−/− mice to determine whether impaired signaling through two separate flagellin detection pathways influences host immune responses and Lp survival in vivo. In addition we identify that alveolar macrophages restrict Lp growth in an NLRC4-dependent, but not TLR5-dependent manner.

Animal model
Tlr5−/− and Nlrc4−/− C57Bl/6 mice strains were obtained from Drs. Shizuo Akira and Vishva Dixit, respectively [27,28], and were backcrossed at least six generations onto C57BL/6. Wild type C57BL/6 mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Mice were housed in laminar flow cages with ad lib access to sterile food and water. Male and female mice were used in approximately equal numbers, and were 8-10 weeks of age at the time of experimentation. The University of Washington Institutional Animal Care and Use Committee approved all experiments.

Airborne infection model
Mice were exposed to aerosolized bacteria in a whole animal exposure chamber as described previously [4,29]. Lp was cultured on buffered yeast charcoal extract (BYCE) agar plates for 4 days at 35°C. Colonies were harvested by rinsing plates with PBS. The pooled suspension was pelleted by centrifugation and resuspended in PBS to approximately 3×10 10 cells/ml, as estimated by optical density. The slurry was transferred to twin jet nebulizers from Salter Laboratories (Arvin, CA USA) and mice were exposed to aerosolized bacteria for 30 minutes. To determine actual bacterial deposition in the lungs in each experiment, four sentinel animals were euthanized with intraperitonal pentobarbital and exsanguinated immediately after aerosol exposure. Left lungs were homogenized in PBS and serial dilutions in Mueller Hinton broth were plated onto BYCE agar. At subsequent time points, left lungs were similarly homogenized and cultured. Unused lung homogenates were mixed in lysis buffer containing protease inhibitors (Roche, NJ) on ice for 30 minutes, centrifuged at 2,500 rpm, and stored at −80°C for further use. Right lungs were used for histology and alveolar cell recruitment, as described [4,30]. Following cannulation of the trachea, right lungs were lavaged four times with 0.5 ml volumes of 0.9% sodium chloride containing 0.6 mM EDTA. Cell counts in bronchoalveolar lavage samples were determined with a hemocytometer, and differentials were determined by examining cytocentrifuge specimens stained with a modified Wright-Giemsa technique (Diff Quick, Dade Behring, Dudingen, Switzerland). After lavage, right lungs were fixed by instillation of 4% paraformaldehyde at 20 cm of pressure. Sections of paraffin-embedded tissue were stained with hematoxylin and eosin. 20 separate high power fields of lung sections were examined by a pathologist and scored to determine the percentage of airspaces with inflammation that involved the alveolar airspaces.

Protein detection
Lung homogenate cytokine levels were determined using a multiplex fluorescent bead array system (Luminex 100). Briefly dilutions of homogenates were captured onto antibody-coated fluorochrome embedded microspheres and read using a flow based sorting and detection platform. Cytokine levels were measured using a sandwich ELISA technique (R&D systems) as previously described [4,30].

Alveolar macrophage isolation and infection
Resident alveolar macrophages (AMs), were harvested from uninfected mice as previously described [4]. Cells pooled from 8-10 mice were suspended in RPMI 1640 supplemented with 10% heat-inactivated FCS, 100 U/ml penicillin, 100 ug/ml streptomycin, and 2 mM L-glutamine, counted in a hemocytometer, and the viability was determined by the exclusion of trypan blue. The cells were cultured in poly-L-lysine-treated 96 well plates at a density of 1 × 10 6 viable AM/ml (0.85 ml/well). After a 2 h incubation at 37°C in humid air with 5% CO 2 , the adherent monolayers were washed five times with prewarmed Hank's balanced salt solution (HBSS) and re-fed with antibiotic-free medium. Lp was added at an MOI of 1 and the plates were centrifuged at 400 × g for 7 minutes, then incubated for 1 hour. The monolayers were washed 4 times with warm HBSS and re-fed with antibiotic-free medium. Immediately and after 24, 48, and 72 h incubation at 37°C in humid air with 5% CO 2 triplicate monolayers were lysed by the addition of saponin to a concentration of 0.1%. The lysate was serially diluted in Mueller Hinton broth and quantitatively cultured on buffered charcoal yeast extract (BCYE) agar. Colonies were counted after 3 days of incubation at 37°C in 5% CO 2 .
Lung RNA isolation Mice exposed to aerosolized Lp 4 h previously and uninfected controls were euthanized with pentobarbital and exsanguinated by cardiac puncture. The pulmonary vascular bed was perfused with 10 ml cold PBS and both lungs were harvested into RNALater (Qiagen, Hamburg, Germany). Total lung RNA was isolated with an RNAeasy column (Qiagen, Hamburg, Germany). Reverse transcriptase was performed with Superscript III reverse transcriptase (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. Total lung cDNA was applied to murine exon arrays (Affymetrix 430 2.0) (Affymetrix, Santa Clara, CA) containing 45,100 probes for 22,006 genes. Each probe was analyzed separately. Induced genes in lung mRNA were determined by comparing the expression values of the probes (which were analyzed separately) in infected and uninfected mice. Genes induced in infected wild type C57Bl/6 mice (n = 2) were compared to genes from infected in Nlrc4−/− mice (N = 2 each) as well as uninfected C57Bl/6 (n = 2) and Nlrc4−/− (n = 2) mice. Raw intensity values were background corrected, log2 transformed and quantile normalized and adjusted by robust multichip average (RMA) prior to analysis [31].

Statistics
Data are expressed as mean ± standard error of mean (SEM). For comparisons of continuous variables among multiple groups, significance was determined using oneway ANOVA with bonferroni post-hoc test analysis. Comparisons between two groups was analyzed with two-tailed T tests. For mRNA expression arrays, a generalized linear model for microarray by limma (Bioconductor [32]) was used. Unadjusted significance was determined for expression arrays by an empirical Bayes moderated t-statistic as described [33] using the equations: . For adjusted p values, a false discovery rate of 5% was used for the analysis in the manner described by Benjamini and Hochberg [34]. The cutoff of significance for genes included in Additional file 1: Table S1 and Additional file 2: Table S2 was an adjusted P value of 0.01, no cutoff for log 2 fold change was used. Gene set enrichment was done using publicly available software (Broad institute [35,36]), using a false discovery rate of 25%.

Results
NLRC4 regulates pulmonary clearance of Lp, but not cellular recruitment to the lungs We infected WT, Nlrc4−/−, and Tlr5−/−Nlrc4−/− mice with aerosolized Lp to determine which immune responses are regulated by NLRC4 and TLR5 during pneumonic infection. Nlrc4−/− mice exhibited markedly delayed clearance of Lp from the lungs compared to WT mice; a 25-fold difference in lung CFUs was evident at the 72 hour time point and a 10-fold difference persisted to 10 days after infection ( Figure 1A and data not shown (for day 10)). Nlrc4−/−/Tlr5−/− mice had no additional impairment in Lp clearance compared to the Nlrc4−/− animals ( Figure 1A). Interestingly, Lp-infected Nlrc4−/− animals showed no difference in lung recruitment of neutrophils at 4, 24, or 72 hours in comparison to WT mice. In contrast, TLR5 and NLRC4 double knockout showed impaired neutrophil recruitment at 4 hours, similar to previously reported findings in Tlr5−/− animals ( Figure 1B) [4]. These data demonstrate that NLRC4 is required for optimal clearance of Lp after airborne infection, but does not influence inflammatory cell recruitment to the lungs. We also saw that during the 72 hour timepoint, there were increased cytokines (IL-1β, MIP2, CXCL-1, MCP-1, and IL-6) observed in Nlrc4-/-/Tlr5-/-mice (data not shown). Together, these data suggest that animals deficient in both NLRC4 and TLR5 pathways produce higher levels of lung pro-inflammatory cytokines than WT or NLRC4 single knockout animals.

NLRC4 does not regulate early lung transcriptional responses
The NLRC4 inflammasome activates caspase-1 and also influences macrophage survival. However, little is known if genes downstream of NLRC4 control replication in the lungs during in vivo infection. To explore this, we isolated mRNA from the lungs of Lp-infected WT and Nlrc4−/− mice 4 hours after infection (n = 2) and measured expression levels with an Affymetrix 430 2.0 exon array. 1701 probes from 1275 genes were differentially regulated (972 probes were induced and 729 repressed) in the WT mice and 1315 probes from 978 genes (980 probes were induced and 335 repressed) in Nlrc4−/− animals compared to the uninfected controls (using a linear model fit from Bioconductor, p < 0.01 [33,37]) ( Figure 3, Table 1, Additional file 1: Table S1, Additional file 2: Table S2, Additional file 3: Table S3, and Additional file 4: Table S4). When comparing infected animals, only 4 probes from 3 genes (D site albumin promoter binding protein, RIKEN cDNA G530011O06 gene, testis expressed gene 11) were different between WT and Nlrc4−/− mice (Table 1, using the formula [(WT stim -WT unstim ) -(Nlrc4−/− stim -Nlrc4−/− unstim )] with a generalized linear model with a false discovery rate of <5%) ( Table 1). In addition gene set enrichment analysis (GSEA) was performed using a 25% false discovery rate cutoff and no gene sets were noted to be significantly different in the lungs of Nlrc4−/− mice compared to WT after infection with LP. Together, these data suggest that the whole lung transcriptional response to Lp is very similar between WT and Nlrc4−/− mice.

NLRC4 regulates lung inflammation
We next examined lung histology in WT, Nlrc4−/−, and Nlrc4−/−/Tlr5−/− mice at 1 and 3 days after infection and determined the percentage of inflamed airspace ( Figure 4A). No difference was seen between WT and Nlrc4−/−, and Nlrc4−/−/Tlr5−/− mice at 24 hours following aerosol inoculation ( Figure 4A and B). At 3 days following inoculation, WT lungs showed significantly less inflammation then the Nlrc4−/− and Nlrc4−/−/Tlr5 −/− animals ( Figure 4B). Compared to WT mice, the Nlrc4−/− and Nlrc4−/−/Tlr5−/− mice had increased patchy infiltration of the alveolar space by neutrophils with some macrophages, as is characteristic for Lp pneumonia in humans. No significant difference was seen between Nlrc4−/− and Nlrc4−/−/Tlr5−/− mice. These data suggest that increased replication in Nlrc4−/− mice is associated with an increased inflammatory response in the lungs 3 days following Lp exposure.   Table S1 and Additional file 2: Table S2 respectively. 3 genes were significantly increased or decreased in WT versus Nlrc4−/− lungs (Table 1). Significant difference in genes was determined by general linear model using a false discovery rate of 0.05 (see methods).  ted alveolar macrophages from NLRC4-null mice ( Figure 5A-C). These data indicate that NLRC4, but not TLR5, restricts the growth of Lp in alveolar macrophages, and suggest that flagellin induces NLRC4mediated resistance to Lp in murine alveolar macrophages.

Discussion
Our primary finding is that TLR5 and NLRC4 mediate different roles in the inflammatory response to Lp flagellin in an aerosolized infection model. In earlier studies, TLR5 detection of Lp flagellin was shown to regulate early recruitment of PMNs to the alveolar space that results in organizing pneumonia at later stages without a difference in bacterial replication [4]. In this study, we again observed that TLR5 regulates early recruitment of PMNs to the alveolar space but not bacterial replication in Nlrc4−/−/Tlr5−/− mice compared to Nlrc4−/− and WT mice. In contrast to TLR5, we show here that NLRC4 regulates bacterial replication in vitro in alveolar macrophages and in vivo by accelerating Lp clearance from the lungs. NLRC4 control of in vivo Lp growth could be mediated via several mechanisms including pyroptosis or via  IL-1β or IL-18-dependent mechanisms. Previous in vitro studies in murine BMDM demonstrated that control of Lp replication is NLRC4/NAIP5-dependent and likely mediated through regulation of pyroptosis [14]. Therefore increased growth of Lp in the lungs of Nlrc4−/− and Nlrc4−/−/Tlr5−/− mice may be due to enhanced replication or failure of alveolar macrophage to undergo pyroptosis in NLRC4-deficient lungs. Alternatively elevated CFUs in Nlrc4−/− and Nlrc4−/−/Tlr5−/− mice could also be explained by enhanced levels of IL-1β or IL-18 in WT animals. Although we found no difference in IL-1β production in lungs of Lp-infected mice WT mice compared to Nlrc4−/− and Nlrc4−/−/Tlr5−/− [38]. Transcriptional analysis of the array data reveal that IL-1β is transcriptionally increased at similar levels in both animals, our methods, which involve measurement of cytokines in lung homogenates, may not adequately distinguish cleaved bioactive IL-1β from pro-IL-1β Nlrc4 −/− and WT animals infected by Lp, suggesting that early induction of pro-IL-1β is found in both WT and Nlrc4−/− animals. In support of a possible IL-18-dependent mechanism, a previous study demonstrated that IL-18 blockade with monoclonal antibodies contributes to control of Lp in the lungs when IL-12 is concurrently blocked, likely through an IFN-γ dependent mechanism [35]. However, others have found that Il1β−/−/Il18−/− mice are not more susceptible to Lp intraperitoneal infection when compared to WT mice [39]. Such a mechanism could contribute to NLRC4-dependent regulation of Lp growth during pulmonary infection.
In addition our study reveals elevated TNF and CXCL2 at 24 hours post-infection in Nlrc4−/−/Tlr5−/− mice compared to Nlrc4−/− and WT mice. The mechanisms underlying cytokine suppression seen in Nlrc4−/− animals compared to Nlrc4−/−/Tlr5−/− animals are currently only speculative. Perhaps signaling through TLR5 may evoke an anti-inflammatory response by causing refractory signaling in TLR signaling pathways at later timepoints (24 h). Another explanation may be that early TLR5mediated neutrophil recruitment seen in WT mice may help to contain the inflammatory response either by eliciting an anti-inflammatory response (for example, by eferocytosis of apoptotic PMNs [40]) or by clearing a pro-inflammatory stimulus (presumably by phagoctosis).
We also observed that NLRC4-deficient animals have increased alveolar space inflammation at 72 h postinfection. One possible explanation for these findings is that failure to control Lp replication in alveolar macrophages of Nlrc4−/− animals contributes to delayed pulmonary clearance of Lp. The higher bacterial burden would therefore enhance later recruitment of lung inflammation by CFU-dependent increases in signaling through other pathways (such as Myd88-dependent pathways) known to be activated by Lp [5]. At 72 hours, the increase in inflammatory cells seen histologically in the airways of Nlrc4−/− mice is different from the bronchoalveolar lavage data that has no difference. The likely explanation for this difference is that the BAL fluid samples cells that are easily recovered from the conducting airways and the alveolar lumina, while the histology measures the aggregates of inflammatory cells that occupy the alveoli and may not be readily dislodged and sampled during lavage. Our findings are consistent with a delayed Lp clearance model with increased bacterial loads even at early time points in Nlrc4−/− mice compared to WT.
In our study we show that growth of Lp in alveolar macrophages is NLRC4-dependent and TLR5-independent, and provides a plausible cellular mechanism for the increased in vivo susceptibility of Nlrc4−/− mice to Lp. Murine alveolar macrophages differ from other primary and BMDMs in their ability to sense flagellin in a TLR5dependent manner [4]. Despite these unique features of TLR5 in alveolar macrophages, our data show that control of Lp replication is not influenced by the presence of TLR5. Furthermore, our observations suggest that flagellin induces NLRC4-mediated resistance to Lp in murine alveolar macrophages, as has been described for BMDMs [10,11,15,16,19,20]. The implications of these data for human biology are not clear. Previous work has shown that human alveolar macrophages are more permissive to Lp replication than blood-derived monocytes [41]. Moreover, human alveolar macrophages differ from monocytes in the requirements for activation of the NLRC4 inflammasome. Alveolar macrophages lack constitutively activated caspase-1 and require a second stimulus for IL-1β secretion in response to TLR ligands, in contrast to monocytes [42][43][44]. Lp growth in human alveolar macrophages is at least partially NLRC4 and NAIP-dependent as determined by enhancement of Lp growth by siRNA knockdown of NLRC4 or NAIP in vitro [45]. Furthermore Lp has been shown to enhance the transcription of β-defensin-3 in a TLR5-dependent manner, which may influence Lp replication in humans differently from the mouse [46,47]. Whether TLR5 knockdown regulates replication of Lp in human AMs is unknown. The precise molecular mechanism of NLRC4 and NAIP control in the both murine and human alveolar macrophage is the focus of current research efforts.
Messenger RNA arrays performed on genes expressed in whole lung showed no difference between Nlrc4−/− and WT animals suggesting that NLRC4 inflammasome signaling at 4 hours is independent of transcription and may occur by post-translational mechanisms. After recognition of flagellin, the NLRC4/NAIP5 inflammasome oligomerizes and activates a proteolytic cascade beginning with activation of caspase-1. The lack of transcriptional regulation of NAIP5/NLRC4 inflammasome accompanied by the activation of proteolytic cascades by Lp has been observed previously at 4 hours in vitro in BMDMs isolated from C57Bl/6 mice compared to the A/J strain [48]. A separate study demonstrated that functionally deficient NAIP derived from A/J mice may regulate transcriptional signaling through IRF1 and IRF8 [49]. Due to experimental differences, it is difficult to compare our data with these previous findings (bone marrow derived macrophages infected ex vivo with Lp versus whole lung mRNA after in vivo aerosol infection). Our study is also potentially limited by power, and it is possible that significant differences in the transcriptome might be obscured by the small sample size. The absence of mRNA differences seen in vivo suggest that the observed phenotype is from posttranslational, caspase-1 dependent events such as pyroptosis. Induction of NLRC4-mediated pyroptosis by Lp may enhance bacterial release by susceptible macrophages and permit phagocytosis by newly recruited neutrophils. This may promote bacterial clearance in the alveolar space, as seen in other models of intracellular infection [39]. Whether pyroptosis contributes to resistance to pneumonic legionellosis is currently unknown.

Conclusions
These studies reveal that NLRC4 and TLR5 control different responses to Lp flagellin in an aerosolized murine infection model and NLRC4 regulates replication in both lungs and alveolar macrophages.

Additional files
Additional file 1: Table S1. Genes differentially expressed in WT lungs after aerosol infection with Lp. Genes in this list had differentially expressed mRNA in WT Lp-infected lungs compared to uninfected controls. Log 2 fold change was determined by the averaged difference between the Log 2 mRNA values in lungs infected with Lp (WTstim) and uninfected lungs (WTunstim). Average expression in WTstim and WTunstim is the average of Log 2 transformed mRNA expression values in two mice. Genes named "-" had no identifiers in probeset. Probe identifiers are from Affymetrix 2.0 array designations. Unadjusted p values were determined as described in methods. Adjusted p values were determined using a false discovery rate of 5% in the method of Benjamini and Hochberg (see methods).
Additional file 2: Table S2. Genes differentially expressed in Nlrc4−/− lungs after aerosol infection with Lp. Genes in this list had differentially expressed mRNA in Nlrc4−/− Lp-infected lungs compared to uninfected Nlrc4−/− controls. Log 2 fold change was determined by the averaged difference between the Log 2 mRNA values in lungs infected with Lp (Nlrc4−/− stim ) and uninfected lungs (Nlrc4−/− unstim ). Average expression in Nlrc4−/− stim and Nlrc4−/− unstim is the average of Log 2 transformed mRNA expression values in two mice. Genes named "-" had no naming identfication. Probe identifiers are from Affymetrix 2.0 array designations. Unadjusted p values were determined as described in methods. Adjusted p values were determined using a false discovery rate of 5% in the method of Benjamini and Hochberg (see methods).
Additional file 3: Table S3. Genes signficantly decreased in WT Lp-infected lungs but not significant in Nlrc4−/− mice.
Additional file 4: Table S4. Genes signficantly increased in WT Lp-infected lungs but not significant in Nlrc4−/− mice.

Competing interests
The authors declare that they have no competing interests.
Authors' contributions WRB conceived and designed the study, carried out the molecular array studies, performed the statistical analysis, and drafted the manuscript. SJS conceived and designed the study and performed the in vivo infection, studies, and edited the manuscript. KDS examined the lung histology and edited the manuscript. TRH conceived and designed the study and helped with drafting the manuscript and statistical analysis. All authors read and approved the final manuscript.