Introduction Legionella pneumophila is a motile gram-negative bacterium that is the cause of a severe form of pneumonia called Legionnaires' disease [1]. Virulence of Legionella depends on its capacity to replicate inside host macrophages. Legionella utilizes a type IV secretion system (encoded by the dot/icm genes) to inject effector proteins into the cytosol of host cells [2,3]. The secreted effector proteins (more than 30 of which have now been identified [4,5]) direct the host cell to create a unique intracellular vacuole in which Legionella can thrive and multiply [6,7]. Early studies found that macrophages from A/J mice support more than three logs of Legionella growth [8], whereas C57BL/6J (B6) macrophages are almost entirely restrictive. Remarkably, this large phenotypic difference is controlled by a single gene on mouse Chromosome 13 called Naip5 (Neuronal apoptosis inhibitory protein 5, also called Birc1e [baculoviral IAP repeat-containing 1e]) [9,10]. Transgenic complementation experiments indicate that B6 mice encode a functional Naip5 allele that confers resistance to Legionella [9,10]. A/J mice also express a Naip5 gene; however, the A/J allele appears to be expressed at lower levels and encodes 14 amino acid polymorphisms as compared to the B6 allele. The precise polymorphism(s) within Naip5 that is responsible for differential permissiveness to Legionella has yet to be identified. Naip5 is a member of a superfamily of proteins that contain C-terminal leucine rich repeats (LRR) and a central nucleotide-binding domain (NBD) [11,12]. Evidence exists that many “NBD–LRR” proteins can function as microbe-detector proteins. For example, Nod1 and Nod2 confer recognition of bacterial cell wall–derived molecules [13–16] and Nod2 is required for resistance to orally delivered Listeria monocytogenes [17]. Human NALP1 is also reported to recognize muramyl dipeptide [18], and mouse Nalp1 is required for susceptibility to anthrax lethal toxin [19]. Ipaf, the closest homolog of Naip, has been shown to be required for caspase-1 activation in response to Salmonella [20]. Although many NBD–LRR proteins appear to function in pathogen detection, there is no evidence that they bind directly to microbial products. We hypothesized that B6 macrophages might restrict Legionella growth upon cytosolic detection of a Legionella-derived molecule. Here we identify flagellin in a genetic selection to identify essential elicitors of macrophage innate immunity. Our results suggest that flagellin initiates a caspase-1-dependent cell-death pathway that functions to restrict bacterial growth. Results/Discussion Legionella Flagellin Mutants Evade Macrophage Immunity We adopted a genetic approach to identify Legionella gene products that might be recognized by host macrophages. We reasoned that Legionella strains lacking such a gene product might be able to evade macrophage immunity and replicate in normally restrictive B6 macrophages. A mariner transposon, which inserts with low bias into frequently occurring TA dinucleotides, was used to generate 30 independent pools of L. pneumophila transposon mutants. Each pool, containing approximately 2,500 mutants, was used to infect an independent well of B6 macrophages. After 6 d, we observed numerous bacteria in all the wells infected with transposon mutants, whereas virtually no bacteria were observed in wells infected with unmutagenized Legionella. Sixty-four transposon mutants obtained from one round of selection in B6 macrophages were retested and, of these, 60 mutants (93.8%) were found to be capable of robust growth in B6 macrophages. Thus, growth in B6 macrophages represents a powerful and highly efficient selection procedure. The gene disrupted by the transposon was determined for 29 independent mutants. Twenty-four mutants were found to have insertions within the flaA open reading frame, four mutants were found to harbor insertions within the flaA promoter, and one mutant (with a weaker growth phenotype, unpublished data) was recovered with an insertion in fliA, a transcriptional activator of flaA [21]. The flaA gene encodes flagellin, the main structural subunit of flagella. Our screen was highly saturating, with an average of approximately 25 transposon insertions per Legionella gene, representing an average of one insertion per 45 bp. We conclude that flaA, and not genes encoding other components of the flagellar regulon, is the primary locus into which a transposon insertion can permit evasion of innate macrophage defenses that normally restrict bacterial growth. The precise transposon insertion site was determined for nine mutants (Figure 1A), of which four were selected for further analysis. LP02 flaA promoter::Tn harbors a transposon insertion 47 bp upstream of the flaA start codon. LP02 flaA 891::Tn, flaA 1365::Tn, and flaA 1418::Tn harbor insertions at nucleotides 891, 1,365, and 1,418 of the open reading frame, respectively. The latter strain is of particular interest since the transposon inserted only 10 bp from the stop codon. Translation read-through into the transposon is predicted to result in the last two amino acids of FlaA being replaced with a 37 amino-acid peptide. All Mutants Evading Macrophage Defenses Are Nonflagellated and Nonmotile In order to confirm the results of the transposon mutagenesis, an unmarked deletion of flaA was generated and then complemented by a chromosomal copy of flaA. For comparison, we also obtained a previously characterized fliI mutant [22]. FliI encodes an ATPase required for flagellar secretion and assembly. The flaA promoter::Tn, flaA 891::Tn, and the ΔflaA strains did not detectably express FlaA protein (Figure 1B), whereas the fliI, flaA 1365::Tn, and flaA 1418::Tn strains expressed flaA but were defective in the secretion of FlaA to the culture supernatant (Figure 1B). Owing to its extended open reading frame, the FlaA protein expressed by LP02 flaA 1418::Tn exhibited a higher molecular weight (Figure 1B). As expected, LP02 and the ΔflaA ahpC::flaA (complemented) strains were motile (unpublished data). The four transposon mutants examined were all nonmotile, as were the unmarked ΔflaA mutant and the fliI mutant (unpublished data). Each mutant was also examined by electron microscopy (Figure 1C). LP02 and the ΔflaA ahpC::flaA (complemented) strains were found to exhibit normal flagella, whereas the fliI mutant and all the flaA mutants were completely nonflagellated. Flagellin Itself Is Required for Restriction of Bacterial Growth We tested the ability of various mutants to grow in bone marrow–derived macrophages. In addition to using B6 macrophages, which express a fully functional allele of Naip5 and therefore normally restrict Legionella growth, we also used a consomic mouse strain, B6.A-Chr13, which is B6 at all loci (~95%), except for those that lie on Chromosome 13, which is derived from the A/J strain of mouse. Since Chromosome 13 harbors the Naip5 locus, both A/J and B6.A-Chr13 mice carry the A/J allele of Naip5 and are permissive for Legionella growth. As expected, the ΔflaA strain recapitulated the phenotype of the transposon mutants and grew robustly in B6 macrophages, indicating that it is indeed loss of flaA, and not polar or unlinked mutations, which results in the growth phenotype (Figure 2). This observation confirms identical results obtained independently by the Swanson group at the University of Michigan (Ari Molofsky and Michele Swanson, personal communication). In contrast, the ΔflaA ahpC::flaA (complemented) strain was restricted in growth in B6 macrophages, but grew well in permissive B6.A-Chr13 macrophages. The fliI mutant (which expresses but does not secrete or assemble flagellin) failed to grow in B6 macrophages (Figure 2A), but grew in B6.A-Chr13 macrophages (Figure 2B), suggesting that innate immunity to Legionella is triggered by FlaA itself, and not by motility or flagella. Interestingly, although the fliI, flaA 1365::Tn, and flaA 1418::Tn strains expressed similar intracellular levels of FlaA protein (Figure 1), only the fliI mutant was growth-restricted. We conclude that transposon insertions at the very C-terminus of flaA must affect secretion of flagellin into host cells or subsequent flagellin recognition by host cells. Macrophage Immunity Is Independent of Toll-Like Receptor 5 Flagellin stimulates immune signaling through toll-like receptor 5 (TLR5) [23], and there is evidence that TLR5 plays an important role in controlling Legionella infections in vivo [24]. However, mouse macrophages do not express TLR5 [25]. Thus, it is unlikely that TLR5 plays a role in mouse-macrophage restriction of Legionella growth in our in vitro system. This is difficult to prove directly since TLR5 knockouts have not yet been reported. Nevertheless, we sought to confirm that macrophage recognition of flagellin does not require TLR5 by examining the phenotype of macrophages deficient in MyD88, a signaling adaptor required for all TLR5 signaling [23]. After 48 h of growth, B6 MyD88−/− macrophages restricted Legionella growth as well as wild-type macrophages (Figure 2A), implying that TLR5 is not required for the initial restriction of flagellin+ bacteria. Interestingly, by 96 h, MyD88−/− macrophages exhibited more growth of wild-type Legionella than did B6 macrophages, and this is likely a consequence of reduced MyD88-dependent signaling (e.g., NF-κB activation), downstream of other toll-like receptors that may play a role in the later phases of infections. Flagellin Mediates Caspase-1-Dependent Cytotoxicity of B6 Macrophages The mechanism by which flagellin-mediated signaling restricts Legionella growth in B6 macrophages is unknown. However, a recent study demonstrated that wild-type Legionella activates caspase-1, and that caspase-1 is required for restriction of Legionella growth [26]. Caspase-1 does not mediate classical apoptosis but is instead required for processing of pro-IL-1 and pro-IL-18 into their mature (secreted) forms [27,28]. Since we found that addition of exogenous IL-1 and/or IL-18 did not render macrophages resistant to Legionella (unpublished data), we concluded that caspase-1 is required for host defense via a mechanism that is independent of IL-1 and/or IL-18. Interestingly, in response to several stimuli, including Salmonella [29,30], Shigella [31], and anthrax lethal toxin [19], caspase-1 has also been shown to be required for an unusual form of macrophage cell death that is distinguishable from apoptosis by its extremely rapid time-course ( 4.0) were diluted 10-fold, placed on a hemocytometer, and observed under high power by a neutral observer blinded with respect to bacterial genotype. Bacteria were considered motile (as opposed to merely subject to Brownian motion) if numerous bacteria could be observed crossing the gridlines on the hemocytometer. For E. coli, bacteria were inoculated at a single point in soft (motility) agar (0.35%), incubated at 37 °C, and observed after ~16 h. Electron microscopy. A stationary-phase culture (5 μl) of Legionella was adsorbed onto a carbon-coated grid that had been made hydrophilic by a 30-s exposure to a glow discharge in an Edwards Auto 306 vacuum evaporator (http://www.bocedwards.com). Excess liquid was removed with a filter paper (Whatman number 1), and the samples were stained with 1% uranyl acetate for 1 min. The grids were examined using a Tecnai 12 Bio Twin (FEI, Hillsboro, Oregon, United States) transmission electron microscope. Western blotting. Stationary-phase bacteria (25 × 106, or trichloroacetic acid [TCA] precipitates of supernatant from 0.5 ml of stationary-phase culture) were pelleted and resuspended in SDS-containing loading buffer. Proteins were separated on a 10% precast SDS-PAGE gel (Bio Rad, Hercules, California, United States), and transferred to PVDF membrane (Amersham Biosciences, Little Chalfont, United Kingdom). Blots were probed with a primary mouse monoclonal anti-flagellin antibody (hybridoma supernatant, kind gift of M. Swanson, University of Michigan, Ann Arbor, Michigan, United States) and a secondary anti-mouse HRP-conjugated antibody (Amersham Biosciences), and flagellin was visualized using enhanced chemiluminescence (PerkinElmer, Wellesley, California, United States). Legionella growth curves. Legionella growth in bone marrow–derived macrophages was assayed as described previously [10]. Cytotoxicity assays. The Cytotox assay (Promega, Madison, Wisconsin, United States) was used to measure LDH release. In this assay, 1 × 105 macrophages were plated per well of a 96-well plate. Two-fold dilutions of Legionella bacteria were grown overnight in liquid buffered-yeast-extract culture and, at the time of infection, samples were matched as closely as possible for optical density (600 nm). In experiments using Salmonella or E. coli, bacteria were grown to midlog phase in Luria-Bertani media before being used to infect macrophages. Bacteria (Legionella, Salmonella, or E. coli) were added to wells at the indicated MOI and the plate was then spun at 400 g (1,400 rpm) for 10 min, except where indicated. After 30 min of incubation at 37 °C, the media was removed from the infected cells and replaced with fresh media containing 50 μg/ml of gentamicin. After an additional 3.5 h of incubation at 37 °C, LDH release was calculated as a percentage of detergent-lysed macrophages for each strain of mouse. Alternatively, neutral red (Sigma, St. Louis, Missouri, United States) was added to macrophages 2 h after infection at a final concentration of 15 μg/ml. Two hours later, viable macrophages (positive for neutral red) and dead macrophages (unstained) were enumerated by light microscopy. Permeability to ethidium bromide-2 homodimer, indicative of cell death, was assessed using the Live/Dead kit (Invitrogen, Carlsbad, California, United States) according to the manufacturer's instructions. Supporting Information Figure S1 Salmonella Induction of Cell Death in Infected Macrophages The indicated strains were used to infect B6 or B6.A-Chr13 macrophages. (Infection and LDH release of macrophages was assayed as shown in Figure 3 and as described in Materials and Methods). The fljB fliC mutant strain does not express either fljB or fliC (the two Salmonella flagellin genes). The invA mutant is defective in SPI-1 type III secretion. The flhD gene is required for expression of the entire flagellar regulon. (A) MOI of 2. (B) MOI of 10. (52 KB PPT) Click here for additional data file. Accession Numbers The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) GI numbers for the genes and gene products discussed in this paper are Birc1e (26245344), caspase-1 (6753282), E. coli MG1655 fliC (B1923), Ipaf (74180700), L. pneumophila ahpC (19880966), L. pneumophila flaA (52627363), L. pneumophila fliI (1938359), MyD88 (31543276), S. flexneri fliC (30063365), S. typhimurium LT2 fljB (1254294), S. typhimurium LT2 fliC (1253480), S. typhimurium LT2 invA (16421444), and TLR5 (7648686).
