Molecular Pathogenesis of Infections Caused by Legionella pneumophila

Legionella bacteria are ubiquitous organisms within freshwater environments (128). L. pneumophila has been recovered from a wide range of both human-made and natural aquatic habitats, from lakes and streams to air-conditioning cooling towers, fountains, and spa baths (130, 198, 235, 300, 314). The exception to this ecological niche is L. longbeachae, which resides primarily in soil, and infection is often associated with exposure to commercial potting soil (317, 318). Legionella bacteria are not free-living aquatic bacteria; rather, they parasitize or form a commensal relationship with free-living, freshwater, and soil amoebae (288, 307, 319). Legionella species multiply intracellularly in many types of protozoa, and this relationship is central to the ecology of the organism in both aquatic and soil environments. While providing a niche for Legionella replication, amoebae also protect Legionella from harsh environmental conditions. This relationship increases the resistance of L. pneumophila to biocides, antibiotics, acid, and osmotic and thermal stress (15, 31, 84, 85, 191). Furthermore, some amoebal species expel biocide-resistant vesicles containing large numbers of L. pneumophila bacteria, which may act as airborne agents for the transmission of the bacteria (40).

Within human-made water systems, Legionella bacteria are found almost exclusively within complex biofilms (282). The characterization of Legionella within these ecosystems is difficult, but model biofilm systems have demonstrated that the replication of L. pneumophila within this niche depends on the presence of a protozoan host (239). Legionnaires' disease is most strongly associated with human-made aquatic environments that contain water at elevated temperatures. In particular, many disease outbreaks are linked to air-conditioning cooling towers and evaporative condensers, which can produce contaminated water droplets that are inhaled by passers by. The increased presence of these large-scale, human-made aquatic reservoirs has likely led to the increased human exposure to Legionella and subsequently an increased incidence of Legionella infection in the latter half of the 20th century. It should be emphasized that accidental human infection is a dead end for Legionella replication, and person-to-person transmission has never been reported. Therefore, the evolution of virulence traits in L. pneumophila has resulted largely from the organism's need to replicate in an intracellular niche and also avoid predation by environmental protozoa.

Phagosomes traditionally mature into digestive vacuoles via the endocytic pathway. This involves a progressive interaction with the endosomal network, leading to the acidification of the vacuole and the degradation of most microbes (141). The endocytic pathway is characterized by the acquisition of early endosome markers, including Rab5, a small GTPase predicted to confer specificity to membrane fusions, and early endosome antigen 1 (EEA1) (306, 365). Following the recruitment of Rab5 and EEA1, endosomes acquire Rab7 and other late endosomal proteins, including lysosome-associated membrane glycoproteins (LAMPs). The accumulation of LAMPs, cathepsin D, and other acid hydrolases ultimately leads to phagolysosome formation (210). The unusual biogenesis of the Legionella vacuole arises from the simultaneous delay of endosome fusion and the recruitment of vesicles and membrane from the host cell secretory pathway (186, 187, 241). An initial examination of L. pneumophila-containing vacuoles (LCVs) reported the formation of a novel ribosome-studded vacuole that did not undergo acidification (173, 174). Indeed, within 5 min of phagocytosis, LCVs begin to recruit ER vesicles that can be detected by the presence of the resident ER protein BiP and analysis of membrane width by electron microscopy (337). After 15 min, the phagosomal membrane becomes thinner, resembling that of the intimately attached ER vesicles rather than the plasma membrane (328, 337). Most endocytic markers, including LAMP-1, cathepsin D, and Rab5, are absent until 18 to 24 h after LCV formation (88, 171, 325, 328). Interestingly, Rab7, a small GTPase that regulates vesicular traffic from early to late endosomal stages of the endocytic pathway, is also found abundantly on LCVs (89, 339). Therefore, although the LCV does not mature along the classical endocytic pathway, there are nevertheless interactions with vesicles trafficking in both the secretory and endocytic pathways (338).

Soon after phagocytosis, the LCV rapidly acquires proteins from secretory vesicles that cycle between the ER and Golgi apparatus. The LCV specifically recruits Rab1, but not Rab2 or Rab6, to the LCV within minutes of uptake, preceding any remodeling of the vacuole (187). Rab1, Rab2, and Rab6 are small GTPases that recruit factors necessary for the fusion of ER-derived vesicles with the Golgi apparatus (7, 236). The inhibition of Rab1 activity impedes the intracellular replication of L. pneumophila, providing further evidence that its recruitment to the LCV is important for the biogenesis of the L. pneumophila replicative organelle (187). Aside from GTPase activity, vesicle tethering and fusion also involve interactions between soluble N-ethylmaleimide-sensitive factor attachment protein receptors (SNAREs) on both the vesicle and target membranes (312). In mammalian cells, the SNARE Sec22b is found on ER-derived vesicles and normally binds with a SNARE complex formed by membrin, syntaxin-5, and Bet1 on the pre-Golgi intermediate compartments (160, 360). Sec22b is delivered to the LCV membrane, providing clear evidence of fusion between LCVs and ER-derived vesicles (187). While most research examining the biogenesis of the LCV has utilized mammalian tissue culture models of infection, the formation of the replicative niche within protozoan hosts also involves a substantial diversion from the default endosomal pathway and the recruitment of rough ER (1, 49). This finding suggests that the ability of L. pneumophila to replicate in protozoa is closely linked to its ability to replicate in alveolar macrophages and that the fundamental processes are conserved between the two hosts.

Apart from Rab1 and Sec22b, double-stranded RNA interference techniques using Drosophilamelanogaster Kc167 tissue culture cells have shown that the knockdown of the small GTPases Sar1 and ADP-ribosylating factor 1 (Arf1) decreases the intracellular replication of L. pneumophila (108). Reduced levels of expression of many other transport proteins had no effect on replication despite resulting in a severe disruption of normal host organelle biogenesis. The disruption of multiple pathways in this system, particularly Sec22 and the Golgi transport protein particle (TRAPP), did hinder L. pneumophila replication, suggesting that multiple sources of vesicles supply the LCV (108). Interestingly, the depletion of Cdc48/p97, which functions as an ATPase in a complex that recognizes ubiquitinated proteins targeted for ER-associated degradation, resulted in diminished L. pneumophila replication in both Drosophila and mammalian tissue culture cells (108). This complex was subsequently shown to associate with the LCV. Although the role of Cdc48/p97 in L. pneumophila infection remains unclear, it may assist in the translocation of bacterial effector proteins into the host cell. Overall, the ability of L. pneumophila to establish and maintain this intracellular niche indicates that the bacteria actively and continually manipulate host cell trafficking from within the replicative vacuole.

Within 1 to 2 h of the recruitment of ER-derived vesicles to the LCV, resident ER proteins such as calnexin also become associated with the LCV, and within 4 to 6 h, the LCV membrane appears to be completely comprised of rough ER (186, 328). Replication of L. pneumophila bacteria then commences between 4 and 10 h after phagocytosis, and the bacteria continue to replicate within a vacuole that maintains a neutral pH (356). This finding concurs with the fact that the optimal growth of L. pneumophila bacteria in vitro occurs between pH 6.3 and 7.2 (2, 278). The nutritional environment of the LCV is poorly defined but presumably supplies the bacteria with amino acids for replication. Indeed, amino acid transporters are important for bacterial viability during intracellular replication (294). The oxidation of amino acids may also contribute to pH balance in the LCV.

Other evidence suggests that the manipulation of autophagy by L. pneumophila is important for bacterial replication. Autophagy is an important process for cellular homeostasis in which double-membrane vesicles (autophagosomes) derived from the engulfment of cytoplasmic components and organelles traffic to lysosomes for degradation (114). This commonly occurs in response to nutrient starvation, as the breakdown of macromolecules provides substrates for essential biosynthetic pathways. Within permissive bone marrow-derived macrophages, L. pneumophila appears to activate the autophagy process upon infection (14). The autophagy enzymes Atg7 and Atg8 show a transient association with the LCV that eventually acquires lysosomal markers in a manner similar to that of traditional autophagosomes (14). It is possible that interactions with the autophagic pathway provide a mechanism for supplying replicating L. pneumophila bacteria with a ready source of nutrients (109). However, investigations of L. pneumophila replication within D. discoideum have shown that autophagy is not linked to L. pneumophila replication. Mutations within five different genes involved in autophagy produced defects in the autophagy process but had no impact on LCV development or the replication of L. pneumophila (257). Therefore, the contribution of autophagy to L. pneumophila replication may be a peculiarity of cell type rather than a core element of LCV formation.

Other species of Legionella replicate within various protozoan species and human-derived cell lines but with differing levels of efficiency, and very little is known about the intracellular vacuoles of these opportunistic pathogens. The properties of the replicative vacuole have been examined only for L. micdadei and L. longbeachae. Within mammalian macrophages, alveolar epithelial cells, and H. vermiformis, L. micdadei multiplies within a spacious vacuole that shows no association with rough ER (138, 353). However, in MM6 human monocytic cells, the L.micdadei-containing vacuole colocalizes with calnexin, a resident protein of the ER (145). These differences suggest that different strains of L. micdadei may evoke different mechanisms to establish an intracellular niche. Virulent strains of L. longbeachae show the same intracellular phenotype as L. pneumophila, with the replicative vacuole associated with rough ER and studded with ribosomes (145). The L.longbeachae-containing vacuole does, however, acquire LAMP-2, a marker of late endosomes (22). Interestingly, cytoplasmic replication has also been reported for some Legionella species, including L. dumoffii (181, 219), suggesting that there is great diversity in the mechanisms used by Legionella species to multiply in eukaryotic cells.

The four L. pneumophila genome sequences available are all human clinical isolates with worldwide distributions and include Philadelphia-1, derived from the original Philadelphia outbreak (223); the Paris strain, an endemic strain responsible for around 12% of cases of Legionnaires' disease in France (23); the Lens strain, the causative agent of a large outbreak in France (252); and the Corby strain (184). The general features of the genomes are conserved in size (3.3 to 3.5 Mb), G+C content (∼38%), number of predicted genes (3,001 to 3,259), percentage of coding regions (88 to 90.2%), and average length of the coding regions (68, 75, 147, 150). In general, the genetic order of the chromosome is conserved; however, synteny between the Lens strain and Paris, Philadelphia-1, and Corby is disrupted by a 260-kb inversion. Overall, the strains share around 80% of genes that constitute the core genome, while around 10% of the genome is strain specific (150). Much of the strain-specific genetic content is present in genomic regions associated with different G+C contents that may represent pathogenicity islands (321). For example, a 65-kb Philadelphia-1-specific genomic island harbors a plasmid-like element comprising a new trb/tra region and several efflux pumps and putative membrane transporters (51). Both Paris and Lens harbor plasmids that in the Paris strain include virulence determinants also found on a large L. longbeachae plasmid (111). The very high level of conservation between the sequenced section of the L. longbeachae plasmid and the 132-kb L. pneumophila Paris plasmid suggests recent horizontal transfer between these species (68). Another 60-kb plasmid present in the Lens strain encodes several proteins with homology to the transfer region of the F plasmid of Escherichiacoli. Interestingly, the region encoding the Lvh type IVa secretion system of L. pneumophila Paris may be present as a multicopy plasmid or integrated in the chromosome (68, 107). This kind of genetic mobility has undoubtedly contributed to the high overall level of plasticity in the L. pneumophila genome and diversity among different strains (67, 68, 75, 147).

The core L. pneumophila genome contains many of the factors associated with the ability of the bacteria to infect eukaryotic cells and replicate (67). In fact, a DNA array screen of 217 L. pneumophila strains showed that there is a high degree of conservation among virulence-associated genes (67). Although no specific hybridization pattern was associated with environmental or clinical isolates in that study, those authors identified L. pneumophila Paris as a globally distributed epidemic strain, suggesting that genomic background does contribute to environmental persistence and clinical disease. That work also identified four phylogenetic lineages (lineages I, II, III, and IV) of L. pneumophila, but these lineages were not associated with geographic region or epidemiology (67, 150). In short, it appears that L. pneumophila is a highly diverse species, and the capacity of any one strain to cause human infection cannot be predicted from its gene content.

Perhaps the most intriguing aspect of the L. pneumophila genome is the unusual number of genes predicted to encode products with amino acid sequence similarity to eukaryotic proteins and/or containing eukaryotic domains (68, 150, 207). Although genes of apparent eukaryotic origin have also been identified for other intracellular pathogens, including Coxiella, Wolbachia, Mycobacterium, and Rickettsia species, L. pneumophila possesses the widest variety. Upon subsequent investigation, many of the gene products were found to play a role in host-pathogen interactions, and many of them are translocated into the eukaryotic host cell, where they presumably interfere with pathways through functional mimicry (207).

One interesting set of eukaryote-type enzymes are the three serine/threonine protein kinases (STPKs) identified in all the L. pneumophila genomes. Mycobacteriumtuberculosis produces 11 eukaryote-like STPKs, with 1 clearly involved in the inhibition of phagosome-lysosome fusion and the promotion of intracellular survival (91). L. pneumophila may utilize eukaryotic STPKs for a similar purpose. L. pneumophila also secretes two proteins exhibiting a significant level of similarity to the mammalian CD39 family of ectonucleoside triphosphate diphosphohydrolases (NTPDases) (134, 291). The presence of two NTPDase genes in L. pneumophila is surprising since these enzymes are associated almost exclusively with eukaryotes. In fact, the recently determined crystal structure of Lpg1905 shows that the bacterial protein is a conserved structural mimic of mammalian NTPDases (344). Lpg1905 is a functional NTPDase that hydrolyzes both ATP and GTP with similar efficiencies (291, 292). The inactivation of lpg1905 results in the defective replication of L. pneumophila bacteria in amoebae, epithelial cells, and macrophages (291). Substantial levels of enzyme activity are also required for the full virulence of L. pneumophila in a mouse model of infection, although it is not yet clear if the ATPase or GTPase activity or both activities contribute to L. pneumophila infection (292).

Another eukaryote-type enzyme and a further addition to the family of secreted metalloproteases predicted for the L. pneumophila genome is an M12 zinc metalloprotease (68, 95). The eukaryotic M12 family of metalloproteases has diverse biological roles and includes members such as adamalysin, which is a component of rattlesnake venom (129), as well as ADAM17, a tumor necrosis factor alpha (TNF-α)-converting enzyme, and ADAMTS13, which cleaves plasma von Willebrand factor (259). Other members of this family target components of the extracellular matrix such as collagen (259). The M12 metalloprotease of L. pneumophila is secreted in vitro and translocated into the host cell, but to date, its substrate specificity and function are unknown (95, 99). Other gene products with eukaryotic similarity include enzymes involved in sphingomyelin degradation, a pathway that is not present in prokaryotic cells. L. pneumophila possesses a sphingomyelinase, sphingosine kinase, and sphingosine-1-phosphate lyase, strongly suggesting that the bacteria interfere with sphingomyelin degradation, although the consequence of this for bacterial replication and LCV formation is not yet known (150). The sphingosine-1-phosphate lyase is at least translocated into eukaryotic cells, where it is targeted to mitochondria (100).

Eukaryotic domains found within predicted L. pneumophila proteins include at least 5 proteins with multiple Sel1 repeats and 20 proteins with ankyrin repeat domains (Ank). Sel1 repeat domains are linked to a range of protein-protein interactions in both prokaryotic and eukaryotic cells (149). Three of the Sel1 repeat proteins, EnhC, LidL, and LpnE, are linked to L. pneumophila entry and/or endosome evasion, although it is not yet known if they play a direct role in host-pathogen interactions (87, 92, 205, 249, 251). For example, EnhC function has been linked to the maintenance of cell wall integrity (205). The Ank proteins are the most abundant group of eukaryote-type proteins encoded by the L. pneumophila genome and exhibit significant diversity among different L. pneumophila strains (67). In eukaryotes, ankyrin repeats mediate protein-protein interactions and are commonly found together with other domains that determine protein function. Therefore, the Ank proteins of L. pneumophila are predicted to have varied activities. The carriage of a large number of Ank proteins is a property shared by the closely related bacterial pathogen Coxiella burnetii (346). In the case of both pathogens, the Ank proteins are translocated into the host cell, where they may interfere with various cell functions, including vesicular transport (258, 346).

Other L. pneumophila proteins carry domains found in eukaryotic guanine exchange factors (GEFs). The Sec7 domain protein RalF is translocated into the host cell and functions as a GEF for the host cell GTPase Arf1, which regulates COPI vesicle trafficking (241). Another protein with a predicted Ras GEF domain (Lpg0276/Lpp0350/Lpl0328/LPC_0353) has not been characterized. Some L. pneumophila proteins appear to target host ubiquitination pathways though U-box and F-box domains, and these too are translocated into the host cell (6, 195; M. Lomma, D. Dervins-Ravault, H. J. Newton, F. M. Sansom, M. Jules, T. Sahr, T. Nora, M. Bonazzi, E. L. Hartland, and C. Buchrieser, submitted for publication). The carriage of such diverse proteins with eukaryotic similarity and/or eukaryotic domains suggests that a major survival strategy of L. pneumophila is the functional mimicry of multiple eukaryotic pathways.

The origin of the eukaryotic proteins of L. pneumophila is an intriguing issue. The proteins seem unlikely to have emerged through convergent evolution, and hence, the current hypothesis is that genes were acquired by horizontal gene transfer from a range of eukaryotic hosts (150). L. pneumophila is capable of exogenous DNA uptake through natural transformation and may have acquired new genetic information through its long evolutionary association with protozoa. In support of this hypothesis, there is increasing phylogenetic evidence of interkingdom DNA transfer between the intra-amoebal bacterium Legionella drancourtii and its amoebal host (225, 226). However, the integration of unrelated DNA would present a significant challenge for the bacterium, and until more protozoan genome sequences become available, the evolution and origin of the eukaryotic genes in L. pneumophila will remain unexplained (150).

In addition to the Dot/Icm T4SS, L. pneumophila possesses a type II protein secretion system (T2SS) termed Lsp, for Legionella secretion pathway, which is required for full virulence and environmental persistence (153, 310). T2SSs are highly conserved protein secretion machines of Gram-negative bacteria that play an important role in the disease progression of many bacterial pathogens through the export and directed release of toxins, proteases, and other enzymes (80). Substrates of T2SSs possess an N-terminal signal sequence that mediates their translocation across the inner membrane via Sec or Tat machinery (80). Once in the periplasm, the signal peptide is removed and the folded protein is transported across the outer membrane by a multiprotein complex comprising the T2SS secretin. The L. pneumophila T2SS has 12 components located in five loci throughout the chromosome. These include the prepilin peptidase pilD, which processes the pilin of type IV pili and pseudopilins of the T2SS (201); the outer membrane secretin and ATPase lspDE; the pseudopilins lspFGHIJK; and lspC and lspLM, which are predicted to promote secretion (287). Mutant strains of L. pneumophila lacking lspDE, lspG, or lspK demonstrate drastically impaired replication within A. castellanii and H. vermiformis cells and a more moderate replication defect in macrophages (263, 284). In addition, lsp mutants show a reduced capacity to replicate in the lungs of mice compared to wild-type L. pneumophila (287).

The ability of L. pneumophila to remain viable under a wide range of conditions is fundamental to its environmental persistence. A disruption of the Lsp secretion system results in an inability of L. pneumophila to survive at low temperatures, both in tap water at temperatures ranging from 4°C to 17°C and in the presence of aquatic amoebae at temperatures of 22°C to 25°C (310, 311). The Lsp system presumably secretes factors that aid L. pneumophila persistence at low temperatures, as the growth of lsp mutants was stimulated by the addition of culture supernatant from the wild-type strain (310). One of these factors is a secreted peptidyl-prolyl cis-trans isomerase present in the culture supernatant (308). Additional factors were recently identified from a transposon mutant screen for growth defects at low temperatures. The screen identified 11 mutants with a compromised ability to replicate at 17°C, and these included mutants with insertions into components of the T2SS. Other major factors included a homologue of the LpxP lipid A acyltransferase; an RNA helicase, CsdA, which was particularly important for bacterial survival; and RNase R, which was implicated previously in replication at low temperatures (70, 309).

Recently, L. pneumophila was observed to undergo sliding motility on agar plates, and this phenotype was linked to the function of the Lsp secretion system (323). Sliding motility was independent of flagella and pili but required a functional Lsp secretion system. Sliding motility appears to be linked to the production of an undefined amphipathic, diffusible surfactant, which may be directly secreted by the T2SS or processed by an Lsp effector protein (323). Surfactant production still occurs in mutants lacking characterized Lsp substrates and in a tatB mutant; therefore, the molecular basis of surfactant production is yet to be identified. Other species of Legionella also demonstrate sliding motility, albeit with various efficiencies, except L. micdadei. This finding is consistent with data from previous reports showing that while L. micdadei harbors the lsp genes, this Legionella species does not express Lsp-dependent phenotypes, including phosphatase, hemolytic and lipolytic activity, and growth at low temperatures (26, 125, 287, 310). It is unclear whether these phenotypic differences are due to a lack of expression of the components of the Lsp secretion system or a specific lack of the Lsp substrates that mediate these phenotypes.

The Lsp secretome has been characterized by proteomics and comprises a cohort of at least 27 proteins (95). Curiously, the Dot/Icm component IcmX is among the secreted proteins (95). A chitinase enzyme, ChiA, is also secreted by the Lsp T2SS, and further investigation demonstrated that while this protein is not required for in vitro intracellular growth, it does promote persistence within the mouse model of infection. Infected mice also produce antibodies to ChiA (95). A range of proteins with enzymatic activities are secreted by the Lsp system, including the most abundant secreted protein, the zinc metalloprotease ProA or Msp, as well as acid phosphatases, multiple lipases, phospholipases C and A, lysophospholipase A (LPLA), an RNase, both tartrate-resistant and tartrate-sensitive acid phosphatases, and an endoglucanase (17-19, 153, 201, 260, 284, 287). An analysis of mutants lacking the tartrate-sensitive acid phosphatase, LipA and LipB lipase, LPLA, phospholipase C, or two aminopeptidases, LapA and LapB, found no impairment in intracellular replication (17, 19, 126, 224, 286, 329). Additionally, these single mutants showed no growth impairment at low temperatures (311). Recently, ProA/Msp and the RNase SrnA were shown to be required for the optimal replication of L. pneumophila bacteria in H. vermiformis (285, 286). This is an interesting finding, as these enzymes are dispensable for growth in all other host cell models tested, which may implicate substrates of the Lsp secretion system in host-specific responses. Indeed, not all Legionella strains have the same Lsp substrate repertoire. For example, the T2SS-dependent secreted endoglucanase is not found in all L. pneumophila strains (260).

As with the substrates of the Dot/Icm secretion system, the vast majority of Lsp substrates confer no phenotype that can be measured by comparing the replications of wild-type and mutant strains. With examples such as ProA and SrnA, this can perhaps be attributed to the substrate playing a specific role in a subset of environmental hosts. However, functional redundancy is also prevalent among Lsp substrates. L. pneumophila encodes at least 15 proteins with phospholipase A (PLA) and LPLA activities, and at least 2 of these proteins, PlaA, the major secreted phospholipase, and PlaC, are secreted via the Lsp system (29, 126; for a review, see reference 28). Phospholipases cleave fatty acids from lipids having a direct impact on host cells through membrane destruction. In the case of a lung pathogen, phospholipases may also act to break down lung surfactant. In addition, by-products of phospholipase activity can act as signaling molecules, influencing important host functions such as inflammation and apoptosis (28). PlaA and PlaC are therefore likely to be important for L. pneumophila virulence, yet single mutants demonstrate no discernible phenotype (29, 126). Further functional redundancy may be expected, as L. pneumophila also has phospholipase activity from a third GDSL-like phospholipase, PlaD (28), and a cell-associated phospholipase, PlaB (127). A cohort of patatin-like phospholipases are also encoded by the L. pneumophila genome, at least two of which are translocated into host cells by the Dot/Icm T4SS (304, 340).

Future studies on the Lsp system will continue to unravel the contribution of Lsp substrates to Legionella biology and may also uncover other unexpected roles of the T2SS, such as its possible function in Rab1 recruitment to the LCV and inhibition of interleukin-8 (IL-8) secretion (N. Cianciotto, personal communication). The putative Lsp/Rab1 connection was made when an lspE mutant was identified among a small group of transposon mutants that recruited Rab1 poorly to the LCV (238). In addition, there is likely to be cross talk between the Lsp and Dot/Icm secretion systems. For example, the M12 zinc metalloprotease is secreted by the Lsp system and is also translocated into host cells by the Dot/Icm system (95, 99). How these two secretion systems may function together at a molecular level remains to be determined.

In addition to the Dot/Icm and Lsp protein secretion systems, L. pneumophila harbors a type I secretion system (Lss), a twin-arginine translocation system (Tat), and a second type IV secretion system (Lvh). The lssXYZABD locus includes the typical components of a type I secretion system, including an ABC transporter (LssB) and a membrane fusion protein (LssD) (182). Lss may play a role in the biology of L. pneumophila, although the secretion system appears to be largely dispensable for host-pathogen interactions (182). In contrast, the Lvh system may play a role in certain aspects of L. pneumophila intracellular replication by complementing Dot/Icm function (27). Following the incubation of L. pneumophila bacteria in water or the encystment of L. pneumophila bacteria in amoebae, the Lvh region was required to rescue the host cell uptake of Dot/Icm mutants (27). This finding suggests that both the Dot/Icm and Lvh systems are conditionally required for certain virulence attributes such as host cell invasion. Another curious feature of the Lvh system is that the locus may exist as an extrachromosomal plasmid or as an integrated genomic island (68). The excision and integration of the lvh region into a chromosomal tRNA gene depend on the growth phase-dependent expression of genes carried by the plasmid-like element (107). For both Lss and Lvh, the proteins or substrates exported by these secretion systems have not yet been identified.

The Tat system complements the general secretory pathway (Sec) by transporting folded proteins across the inner membrane. In the case of L. pneumophila, the Tat pathway contributes to biofilm formation and intracellular replication in macrophages and amoebae and aids growth under low-iron conditions (98, 283). Several substrates of the Tat pathway have been confirmed, including phospholipase C, LvrE, and cytochrome c oxidase, although the list of likely Tat substrates is much longer and also includes components of the Lss and flagellar biosynthesis systems (96-98, 283). There appears to be little crossover between the Lsp and Tat secretion systems. Very few of the Lsp substrates seem to utilize the Tat pathway to reach the periplasm, and the supernatants of tatB mutants exhibit normal levels of Lsp-dependent enzymatic activity (283). In addition, a mutant lacking both Tat and Lsp systems demonstrates a defect in replication greater than that of either of the single mutants (283), which supports the idea that the Tat and Lsp systems work independently.

The gene encoding the peptidylprolyl cis-trans isomerase (PPIase) Mip was one of the first genes associated with the ability of L. pneumophila to replicate in eukaryotic cells, and the mip nucleotide sequence has been used extensively as a target for Legionella molecular diagnostics and typing for 2 decades (69, 115, 118, 185, 203, 269). Mip stands for macrophage infectivity potentiator, and mip mutants of L. pneumophila are defective for replication in macrophages, epithelial cells, and amoebae and are attenuated in both guinea pig and mouse models of infection (81, 83, 358).

The mip product is a 24-kDa protein with PPIase activity that shares amino acid sequence similarity with the mainly eukaryotic family of FK506 binding proteins, a class of immunophilins (123, 152). Mip is exported to the cell surface, where the protein exists as a stable homodimer (82, 275, 297). The catalytic PPIase domain of the enzyme is located in the C terminus of the protein, which is connected to the N-terminal dimerization domain by a long alpha-helix that acts as a hinge between the two domains (168, 275). The crystal structure of Mip revealed that the fold of the catalytic domain resembles the family of human protein FK506 binding proteins; therefore, Mip is another structural and functional mimic of a eukaryotic protein (275). Recently, the PPIase activity of Mip was associated with the ability of L. pneumophila to penetrate an epithelial barrier and to disseminate from the lung in a guinea pig model of infection (347). Mip binds to collagen and may sensitize the extracellular matrix to proteolysis by converting protease-insensitive cis prolyl bonds to protease-sensitive trans prolyl bonds (347). Mip then acts in concert with a serine protease of either bacterial or host origin to degrade the extracellular matrix. Mip may also act synergistically with other secreted proteins of L. pneumophila, such as a p-nitrophenol phosphorylcholine hydrolase (94).

Many studies have highlighted the importance of iron for L. pneumophila virulence and replication by exposing L. pneumophila bacteria to iron-limiting conditions in vitro (183, 272). Iron is also important during intracellular replication, as monocytes and macrophages that have been treated with an iron chelator do not support L. pneumophila replication (143, 342). In addition, a low level of expression of the transferrin receptor in monocytes and macrophages correlates with a nonpermissive state (62).

Given that a source of iron is essential for L. pneumophila replication, it is not surprising that the pathogen has developed mechanisms to regulate iron utilization and obtain iron from the environment and from host cells (79). A key development in our understanding of the acquisition of iron by L. pneumophila was the discovery that the bacteria secrete a siderophore although only at particular phases of growth under iron-limiting conditions (202). Representative strains from all L. pneumophila serogroups and other Legionella species also have siderophore activity under defined conditions (316). The production and export of the L. pneumophila siderophore legiobactin have been linked to the genetic locus lbtAB. LbtA is the predicted siderophore synthetase, and LbtB exhibits homology to multidrug efflux pumps and is likely responsible for the export of legiobactin (9). Most recently, lbtA or lbtB was confirmed as being essential for the production of legiobactin, and lbtA was identified as being important for infection of mice. Coinfection studies were able to show that legiobactin secreted from wild-type L. pneumophila bacteria can promote the replication of the lbtA mutant in trans (8).

Once the iron-siderophore complex is internalized across the outer membrane, it is likely acted upon by a periplasmic ferric reductase to yield ferrous iron that is transported across the inner membrane by FeoB (79). L. pneumophila also secretes pyomelanin, which has ferric reductase activity and may supply ferrous iron substrate to FeoB (72). Mutational analysis has shown that feoB is important for extracellular growth and promotes replication in H. vermiformis and macrophage cell lines as well as infection of mice (279). L. pneumophila also produces a multicopper oxidase, McoL, that is essential for aerobic extracellular growth under iron-limiting conditions or where ferrous iron is the only iron source (176). Rather than aiding iron acquisition, McoL may help prevent the toxic effects of utilizing ferrous iron during aerobic growth. In addition to these fundamental mechanisms, several other factors appear to be important for iron acquisition and assimilation, most notably the cytochrome c maturation (ccm) locus, which is essential for intracellular multiplication, and iraAB (for iron acquisition/assimilation), where iraA is important for intracellular replication and iraB contributes to extracellular growth under iron-limiting conditions (244, 264, 342, 343).

While iron is a critical limiting nutrient for many pathogens, other nutrient acquisitions by L. pneumophila must also occur intracellularly during the replicative phase of L. pneumophila growth. In fact, the nutritional environment of the LCV is quite poorly defined. It is clear that in vitro, amino acids are preferentially utilized by L. pneumophila as a source of carbon, nitrogen, and energy in broth culture. Early reports demonstrated that L. pneumophila requires the presence of arginine, isoleucine, leucine, methionine, serine, threonine, valine, cysteine, and glutamic acid to replicate (144, 332, 333). Interestingly, genome sequencing revealed that L. pneumophila does possess genes for the synthesis of cysteine and methionine, suggesting that under certain conditions, the bacterium may not be auxotrophic for these amino acids (68, 75).

Clear evidence for the importance of amino acids for L. pneumophila replication is demonstrated by the discovery of a family of amino acid transporters that are crucial for the ability of L. pneumophila bacteria to differentiate and replicate within host cells (73, 294). These transporters have been termed phagosomal transporters (Pht) and belong to a subfamily of the major facilitator superfamily transporters. The first member of the family to be characterized was PhtA, which was shown to be a threonine transporter (294). L. pneumophila phtA mutants have an inability to replicate in macrophages, as the transmissive bacteria that invade and establish the LCV do not differentiate into a replicative phase (294). This demonstrates that the phagosomal transporters not only act to provide essential nutrients but in doing so can also assess the nutrient supply in a given environment, thereby providing a cue for replication. In addition, host amino acid transporters likely contribute to the vacuolar conditions that support L. pneumophila intracellular replication (357).

Including PhtA, all sequenced L. pneumophila strains possess 11 Pht paralogues, with strain Lens also having a strain-specific Pht (73). PhtJ, which transports valine, is at this stage the only additional Pht with a defined function (73); however, recent evidence suggests that other Pht family members, namely, PhtC and PhtD, may act as thymidine nucleoside transporters (M. Swanson, personal communication). Interestingly, several other intracellular human pathogens, including Chlamydia and Coxiella, possess Pht genes, indicating that this may be a common mechanism for nutrient sensing and acquisition by vacuolar pathogens (73).

The environmental and intracellular niches of L. pneumophila expose the organism to a range of nutrient and temperature conditions that require the bacteria to be highly adaptable. Similar to other intracellular pathogens such as Chlamydia and Coxiella, L. pneumophila has a biphasic growth cycle where, under nutrient-rich conditions, the bacteria are in a replicative phase and express few virulence traits, but under nutrient-limited conditions, the bacteria enter a transmissive phase and become highly motile and resistant to various stresses (63). The transmissive phase allows bacteria to egress from spent host cells and commence the next round of infection, and this coincides with the expression of many virulence determinants.

A comprehensive microarray analysis examining and comparing the transcription profiles of L. pneumophila bacteria in replicative and transmissive phases both during infection of A. castellanii and in vitro highlighted the dramatic changes that differentiate these phases (59). That study found an increased transcription of metabolic genes involved in carbohydrate utilization and amino acid metabolism during the replicative phase (59). In contrast, during the transmissive phase, L. pneumophila upregulates the expression of many Dot/Icm substrates, genes involved in flagellar biosynthesis, and many other known virulence determinants. A large class of GGDEF and EAL genes, known to modulate the concentration of the second messenger cyclic di-GMP, are also upregulated during the transmissive phase, suggesting that cyclic di-GMP signaling is important for the regulation of this biphasic cycle (59).

A range of regulatory genes interact in a complex manner to control the transition of L. pneumophila between replicative and transmissive forms. Initial studies examining regulation during broth culture growth of L. pneumophila bacteria found that amino acid depletion, occurring following exponential growth, triggered the bacteria to release the alarmone guanosine 3′,5′-bispyrophosphate (ppGpp) (154). This response has been reported for a variety of microbes and involves uncharged tRNAs binding to ribosomes and activating ppGpp synthetases such as RelA and SpoT in L. pneumophila (93, 154). The accumulation of ppGpp stimulates the transition of L. pneumophila from its replicative form to its transmissive form via the sigma factors RpoS and FliA and the two-component response regulator LetA/LetS (for Legionella transmission activator and sensor) as well as the positive regulator LetE (25, 155, 367).

L. pneumophila RpoS regulates motility, sodium sensitivity, and evasion of the endocytic pathway (24, 153a). As such, it is not surprising that an isogenic mutant of rpoS is unable to replicate within A. castellanii (153a). Among other factors, RpoS regulates the expression of fliA, which itself is important for regulating contact-dependent cytotoxicity, infectivity, lysosomal avoidance, and biofilm formation (155, 215, 232). In concert or in parallel with RpoS and FliA, other two-component response regulators contribute to the regulation of transmissive traits. Both PmrAB and CpxRA directly regulate the transmissive-phase expression of Dot/Icm components and substrates (5, 11, 135, 366).

Perhaps the most important regulatory switch controlling the replicative- to transmissive-phase cycle is the two-component system LetAS (136, 155, 211). LetAS initiates a global regulatory mechanism that acts through the derepression of CsrA (233). Only recently have the small noncoding RNAs that mediate this process been identified (268, 290). A signal is relayed from LetAS to increase the expression levels of the small RNAs RsmY and RsmZ. These noncoding RNAs possess multiple CsrA binding motifs that act to sequester the transcriptional repressor CsrA and allow the transcription of genes that mediate the transmissive phenotype (268, 290).

An additional layer of complexity is incorporated into this regulatory network by quorum sensing. The response regulator LqsR contributes to virulence processes such as phagocytosis and the formation of the LCV by relaying signals from the adjacent sensor kinase LqsS (334, 335). Neighboring this system is an autoinducer synthase, LqsA, which was recently shown to synthesize 3-hydroxypentadecan-4-one, the small Legionella autoinducer 1 (LAI1) (313). Although an lqsA mutant of L. pneumophila showed only a modest decrease in phagocytosis compared to the wild-type strain, the overexpression of LqsA in an lqsR or lqsS mutant background restored the more severe uptake defects of these mutants, suggesting that LqsA does have a role in modulating virulence, possibly through regulating the expression of a novel genomic island (336).

Despite the significant shortcomings of the mouse infection model for Legionnaires' disease, the fact that mice are by and large resistant to L. pneumophila infection has allowed investigators to identify elements of the immune response that are critical to the control of L. pneumophila replication in macrophages and in the lung. Crosses between the permissive A strain and highly resistant C57BL/6 mice allowed the unique susceptibility of A mice to be mapped to the lgn1 locus on chromosome 13 (34, 104). Within this region are a series of highly polymorphic repeats of neuronal apoptosis-inhibitory protein (Naip), also known as baculoviral IAP (inhibitor of apoptosis) repeat-containing 1 (Birc1) (151, 361). Variations, either by missense polymorphisms or altered expression levels, in Naip5/Birc1e alleles have been defined as the basis for the susceptibility of the A mouse strain to L. pneumophila (105, 359). Naip5 is an intracellular flagellin recognition molecule that shows homology to plant resistance proteins involved in innate immune responses to secreted bacterial virulence proteins (178). In macrophages, Naip5 activates caspase-1 upon the phagocytosis of L. pneumophila, leading to IL-1β production and an increased fusion of the LCV to endosomes, followed by bacterial degradation (363). Although a Naip5 equivalent is not found in humans, a study of Ipaf, another intracellular flagellin recognition molecule, produced similar findings, where Ipaf was shown to restrict L. pneumophila replication (12). Both Naip5 and Ipaf detect cytosolic flagellin, and either the deletion of flagellin from L. pneumophila or the deletion of Naip5 or Ipaf from nonpermissive macrophages reverses the restriction of L. pneumophila intracellular replication (12, 231, 273).

Since the publication of the first in vivo mouse model of L. pneumophila infection (162), various studies have identified key components of the innate immune response as being important for limiting and eliminating infection. These components include cytokines such as gamma interferon (IFN-γ) (54, 303), tumor necrosis factor alpha (TNF-α) (56, 133), IL-12 (57), and IL-18 (55) as well as the cell types that produce these cytokines, namely, neutrophils (101) and natural killer (NK) cells (20, 101).

More recently, investigations of the innate immune response to L. pneumophila have focused on Toll-like receptors (TLRs), a group of receptors that recognize various microbial products that trigger an innate immune response upon binding a specific ligand. Considerable study has also been directed toward myeloid differentiation primary response gene 88 (MyD88), an important adaptor molecule for most TLR signaling. MyD88-deficient mice infected with L. pneumophila have an increased bacterial burden in the lung and decreased survival rates, develop more severe lung pathology, and suffer disseminated bacterial infection in the spleen compared to wild-type mice (20, 157). However, investigations of individual TLRs such as TLR2, TLR4, TLR5, or TLR9 have revealed that their contribution to L. pneumophila resistance is negligible or, at best, very modest compared to that of MyD88 (21, 41, 156, 157, 197, 248, 315). For example, in an attempt to determine if multiple TLRs were responsible for the MyD88-dependent immune response to L. pneumophila, TLR2/9-deficient mice were infected with a flagellin-null (ΔflaA) mutant strain of L. pneumophila. By using this strategy, the combined contribution of TLR2, TLR5, and TLR9 to the immune response to L. pneumophila could be assessed. The ΔflaA bacterial burden in TLR2/9-deficient mice was similar to that of control mice, suggesting that the MyD88-dependent immune response was not restricted to the actions of TLR signaling (20). MyD88 also functions as an adaptor molecule for IL-1 and IL-18 receptor signaling, both of which are induced upon L. pneumophila infection (12, 216, 293, 363). The fact that L. pneumophila infection of caspase-1-deficient mice results in significantly higher bacterial loads in the lung and significantly less IL-18 production than that in control mice indicates that caspase-1- and Ipaf-deficient mice are more susceptible to L. pneumophila infection due to reduced IL-18 (or IL-1β) signaling that is MyD88 dependent (66).

Apart from the activation of the inflammasome through Ipaf and Naip5, L. pneumophila has diverse effects on cytokine signaling pathways. In macrophages, L. pneumophila stimulates cytokine activity in a TLR- and RIP2-independent and Dot/Icm-dependent manner (302). In fact, robust mitogen-activated protein kinase (MAPK) signaling (p38 and Jun N-terminal protein kinase [JNK]) is induced in response to the Dot/Icm system in infected macrophages, and the fact that icmS and icmW mutants do not induce MAPK activation suggests that one or more Dot/Icm effectors are either directly or indirectly responsible (302). The consequences of MAPK activation for L. pneumophila replication are unknown, but this activity may constitute an important host response to infection in vivo. Certainly, in amoebae, L. pneumophila inhibits the equivalent of the MAPK pathway by inducing the expression of DupA, a tyrosine kinase/dual-specificity phosphatase, which is likely to be a negative regulator of MAPK signaling (200). L. pneumophila upregulates MAPK phosphatases in macrophages, which supports the idea that this pathway is manipulated during infection (208).

The induction of a robust type I interferon response was also observed during infection of macrophages and epithelial cells with L. pneumophila (204, 256, 295, 302). Exogenous type I interferon restricts L. pneumophila replication in permissive macrophages, while endogenous type I interferon restricts L. pneumophila replication in nonpermissive macrophages (234, 262, 295). As evidence for the latter, macrophages from type I interferon receptor-deficient mice support L. pneumophila replication (90, 234, 262, 295). The type I interferon response is independent of flagellin and results from the transfer of nucleic acid to the cell cytosol through a Dot/Icm-dependent process (90, 234, 322). Until recently, it was thought that bacterial DNA was recognized by a cytosolic DNA sensor (322). However, it has now been shown that RNA stimulates the type I interferon response through the host helicases RIG-I and MDA5 and the adaptor Ips-I, a pathway previously known to respond only to viral infections (76, 234). The source of RNA may originate from the pathogen or from RNA generated by host RNA polymerase III acting on pathogen-derived DNA (76, 234). The latter possibility seems unlikely given that cytosolic L. pneumophila DNA alone does not stimulate an Ips-I-dependent type I interferon response in macrophages, even in the presence of RNA polymerase III (234). Therefore, L. pneumophila may translocate RNA into the host cytosol, or alternatively, as-yet-unknown Dot/Icm effectors may stimulate the production of RNA intermediates in the host cell that then stimulate RIG-I/MDA5 signaling (234). Given the currently accepted view that L. pneumophila has not evolved to avoid the mammalian immune system, it is surprising that the Dot/Icm effector SdhA is a potent inhibitor of the type I interferon response (234). Ectopically expressed SdhA inhibits the RIG-I/MDA5 pathway, although it is not yet clear if this occurs through a direct interaction or whether the inhibition results from an indirect effect of SdhA on mitochondria. Since SdhA is also implicated in inhibiting host cell death (196), the two effects of SdhA on macrophages could be linked. Despite the effort that has gone into characterizing the mechanism of type I interferon induction, Ips-I- and type I interferon receptor-deficient mice are no more susceptible to L. pneumophila infection, indicating that the type I interferon response is dispensable for limiting L. pneumophila infection in vivo, at least in mice (234; D. K. Y. Ang, E. L. Hartland, and I. R. van Driel, unpublished data).

Few studies have examined other Legionella species in animal infection models; however, L. micdadei does not infect the lungs of A mice as readily as L. pneumophila, highlighting differences in pathogenicity between L. pneumophila and L. micdadei in mice (138). A study of the ability of L. longbeachae isolates to infect guinea pigs also demonstrated that there was a substantial variation in the virulences of different L. longbeachae strains in this infection model (110). Strains of L. longbeachae serogroup 1 are highly virulent in mice, with the intratracheal inoculation of 10³ bacteria capable of causing disease (148). Unlike L. pneumophila, mouse infection with L. longbeachae is not restricted to A mice, and both BALB/c and C57BL/6 mice are also susceptible to infection with serogroup 1 strains (148). This difference in virulence may in part be related to the fact that L. longbeachae serogroup 1 strains are nonmotile, as the deletion of flagellin from L. pneumophila also increases the virulence and host range of the pathogen for mice (148, 273).

Although the characterization of innate immune responses to L. pneumophila infection of mice is well investigated, far less is known about the innate immune response during legionellosis in humans. Certainly, the great majority of people exposed to L. pneumophila remain asymptomatic or suffer only a mild self-limiting infection. Risk factors associated with the development of severe disease include age, smoking status, male gender, chronic obstructive pulmonary disease, alcohol intake, and immune suppression (324). A retrospective analysis of Legionnaires' disease patients showed that these individuals released less IFN-γ in response to bacterial LPS than non-Legionnaires' disease patients, suggesting that an impairment in the IFN-γ response may also increase susceptibility to the disease (199). In general, patients with acute Legionnaires' disease have elevated serum levels of IFN-γ and IL-12, which are typical of a Th1-type response (331). This finding is supported by in vitro studies showing that primary human macrophages or cell lines derived from human macrophages or epithelial cells produce a Th1-type cytokine response following L. pneumophila infection that ultimately restricts bacterial replication (42, 220, 243, 348). Thus, an early and robust inflammatory response appears to be critical to limit infection.

Attempts to examine possible genetic susceptibilities to Legionnaires' disease are frustrated by the small numbers of patients in disease cohorts. Nevertheless, the examination of patient genotypes in a large outbreak of Legionnaires' disease at a flower show in the Netherlands correlated some human TLR polymorphisms with the development of disease independently of other risk factors. A common TLR5 polymorphism that introduces a premature stop codon (TLR5^392STOP) occurs in around 10% of the population and is associated with a small but significantly increased risk of Legionnaires' disease (159). In heterozygotic individuals, the TLR5^392STOP polymorphism is dominant and impairs the production of proinflammatory cytokines. As the recognition of bacterial flagellin by TLR5 on alveolar epithelial cells is a major driver of an IL-8 and IL-6 response, the TLR5^392STOP mutation likely increases an individual's risk of Legionnaires' disease by weakening the cytokine response (159). These results support findings from the mouse model of L. pneumophila infection where deficiencies in the recognition of flagellin by the innate immune system also increase susceptibility to infection. In contrast, a polymorphism in the TLR4 receptor TLR4^A896G was associated with an increased resistance to Legionnaires' disease despite the fact that L. pneumophila LPS does not strongly activate inflammatory signaling through TLR4 (158). Another innate immune factor that has been associated with resistance to L. pneumophila infection is mannose binding lectin (MBL). Patients with Legionnaires' disease show impaired MBL-mediated complement activation (116, 164). However, this deficiency does not reflect an underlying genetic predisposition to Legionnaires' disease but rather appears to be a physiological consequence of severe L. pneumophila infection (117, 164).

Although a vigorous inflammatory response is critical for limiting bacterial infection of the lung during Legionnaires' disease, T and B cells are ultimately required for the clearance of the infection. Evidence for the role of T cells comes from the depletion of CD4 or CD8 T cells using monoclonal antibodies, which resulted in decreased survival rates for mice infected with high doses of L. pneumophila compared to nontreated animals (326). For mice infected with lower doses of L. pneumophila, CD4 or CD8 T-cell-depleted animals were unable to clear the bacteria from their lungs for up to 11 days after inoculation (326). In all cases, animals depleted of both CD4 and CD8 T cells were more severely affected than mice depleted of only one subset (326). Another study showed that mice immunized with L. pneumophila-pulsed, fractalkine-expressing DCs were protected from a lethal dose of L. pneumophila. This protection was completely abolished in CD4 T-cell-, CD8 T-cell-, and B-cell-deficient animals (190). These results demonstrate that T and B cells are essential for protecting mice against L. pneumophila infection. However, the exact mechanisms by which these cells become activated and confer protection are still largely unexplored.

T cells become activated after recognizing their antigen via the presentation of peptide-major histocompatibility complexes (MHCs) by antigen-presenting cells (APCs). The primary APC is the dendritic cell, and several studies performed with mice have implicated this cell type as the APC responsible for initiating the adaptive immune response to L. pneumophila. For example, murine bone marrow-derived dendritic cells (BMDCs) infected with L. pneumophila are able to stimulate IFN-γ production by CD4 T cells in vitro (245). This activation of CD4 T cells was dependent on the successful formation of the LCV and the prevention of lysosome fusion by the bacteria (245). Interestingly, this scenario also occurred with murine bone marrow-derived macrophages, although BMDCs stimulated more IFN-γ production in CD4 T cells than did macrophages (245, 246). These data indicate that although L. pneumophila is sequestered in the LCV, this is insufficient to prevent the presentation of bacterial antigens to CD4 T cells. The exact mechanisms by which the bacterial antigens are processed and presented by MHC class II molecules are still unknown. BMDCs that come into contact with L. pneumophila also upregulate fractalkine (CX3CL1), a strong chemoattractant for T cells (190). In contrast to macrophages, murine BMDCs restrict the intracellular growth of L. pneumophila (245). This and the fact that BMDCs undergo rapid apoptosis upon L. pneumophila infection may be a mechanism that not only prevents bacterial dissemination but also delays antigen presentation and the activation of L. pneumophila-specific T cells (255).

Early studies showed that a cell-mediated immune response to L. pneumophila also occurs in humans (43, 169). Although there is a paucity of information on the nature of this response, patients with immune deficiencies resulting from corticosteroid therapy and some types of leukemia are at an increased risk of severe and persistent Legionnaires' disease (165, 296). L. pneumophila infections are uncommon in patients with AIDS (122), although some reports suggested that the disease is more severe (261). The nature of the effector cells controlling L. pneumophila infection in humans is not clear, but in vitro studies show that Th1-type CD4 T cells respond to L. pneumophila by secreting IFN-γ (193).

Immune regulation has emerged in recent times as an important concept describing the delicate balance between protective immune responses to pathogens and damaging immune-mediated pathology. During L. pneumophila infection of mice, it appears that this balance may be achieved via the production of prostaglandins by infected macrophages, which in turn inhibit IFN-γ production in T cells (246). It will be interesting to see if other aspects of immune regulation such as CD4⁺ Foxp3⁺ T-regulatory cells also play a role in regulating the cell-mediated immune response to L. pneumophila.

Both patients and laboratory animals produce antibodies in response to L. pneumophila infection. Major characterized antigens include lipopolysaccharide, the Msp/ProA protease, flagellin, Hsp60, and OmpS (45, 48, 132, 276, 352). Early studies investigated the potentials of several of these antigens in the development of vaccines to L. pneumophila, and although some antigens showed significant promise in animal infection models, there has been little activity in this area in recent years (44-48, 276). Nevertheless, vaccine studies have informed our understanding of the immune control of L. pneumophila infection and support a role for both antibody- and cell-mediated immunity in the clearance of the bacteria from the lung (46, 48, 276).