The natural alternative: protozoa as cellular models for Legionella infection

Authors

  • Christine Hoffmann,

    1. Max von Pettenkofer Institute, Department of Medicine, Ludwig-Maximilians University Munich, Munich, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Christopher F. Harrison,

    1. Max von Pettenkofer Institute, Department of Medicine, Ludwig-Maximilians University Munich, Munich, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Hubert Hilbi

    Corresponding author
    1. Max von Pettenkofer Institute, Department of Medicine, Ludwig-Maximilians University Munich, Munich, Germany
    • For correspondence. E-mail hilbi@mvp.uni-muenchen.de; Tel. (+49) (0)89 2180 72964; Fax (+49) (0)89 2180 99 72964.

    Search for more papers by this author

Summary

The severe pneumonia known as Legionnaires' disease occurs following infection by the Gram-negative bacterium Legionella pneumophila. Normally resident in fresh-water sources, Legionella are subject to predation by eukaryotic phagocytes such as amoeba and ciliates. To counter this, L. pneumophila has evolved a complex system of effector proteins which allow the bacteria to hijack the phagocytic vacuole, hiding and replicating within their erstwhile killers. These same mechanisms allow L. pneumophila to hijack another phagocyte, lung-based macrophages, which thus avoids a vital part of the immune system and leads to infection. The course of infection can be divided into five main categories: pathogen uptake, formation of the replication-permissive vacuole, intracellular replication, host cell response, and bacterial exit. L. pneumophila effector proteins target every stage of this process, interacting with secretory, endosomal, lysosomal, retrograde and autophagy pathways, as well as with mitochondria. Each of these steps can be studied in protozoa or mammalian cells, and the knowledge gained can be readily applied to human pathogenicity. Here we describe the manner whereby L. pneumophila infects host protozoa, the various techniques which are available to analyse these processes and the implications of this model for Legionella virulence and the pathogenesis of Legionnaires' disease.

Introduction

Legionella pneumophila came to the attention of the world in 1976 when it caused a fatal outbreak of a pneumonia which was to become known as Legionnaires' disease (Fraser et al., 1977). L. pneumophila is a ubiquitous fresh-water bacterium, found in numerous natural and man-made water sources around the world, commonly in association with biofilm communities (Newton et al., 2010; Hilbi et al., 2011a). Biofilms are ubiquitous in the environment and contain a range of bacterial species, often in communication via quorum sensing molecules (Hall-Stoodley et al., 2004). They can also be ‘grazed’ by protozoan predators such as amoeba or ciliates, which have numerous methods available to track, engulf, and kill their bacterial meals. There is thus a significant selective pressure on bacteria to avoid this fate, and this is achieved in a variety of ways (Greub and Raoult, 2004; Matz and Kjelleberg, 2005; Hilbi et al., 2007). Amoeba resistance constitutes a mechanistic basis of Legionella virulence, since it underlies macrophage-resistance, which in turn is a prerequisite for Legionnaires' disease (Newton et al., 2010; Hilbi et al., 2011a).

Legionella pneumophila utilizes a molecular-scale needle known as the Icm/Dot type IV secretion system (T4SS) to avoid destruction (Berger et al., 1994; Brand et al., 1994). Upon phagocytosis by amoeba or macrophages, this T4SS injects a plethora of proteins into the host cell, hijacking the ‘default’ phagocyte-lysosome pathway (Isberg et al., 2009; Hubber and Roy, 2010; Zhu et al., 2011). Thus instead of being shuttled to the lysosome and degraded, L. pneumophila can hide and replicate within a dedicated vacuole known as the Legionella-containing vacuole (LCV). After replication the bacteria lyse and exit their erstwhile predator, ready to infect a new target (Horwitz and Silverstein, 1980) (Fig. 1).

Figure 1.

Overview of L. pneumophila infection. The course of L. pneumophila infection can be divided into five main stages: (1) pathogen uptake, (2) formation of the replication-permissive vacuole, (3) intracellular replication, (4) host cell response, and (5) bacterial exit. The host cell processes and components implicated at each step are indicated, as well as the Icm/Dot-translocated effectors and other bacterial factors, which target them.

In addition to the Icm/Dot T4SS, L. pneumophila produces the Lsp type II secretion system (T2SS), which is also implicated in pathogen–host interactions (Hales and Shuman, 1999a; Liles et al., 1999). The T2SS is required for growth within amoeba and macrophages, and it secretes a number of hydrolytic enzymes including proteases, aminopeptidases and phospholipases (Rossier and Cianciotto, 2001; Rossier et al., 2008), some of which determine the host protozoa specificity of L. pneumophila among Acanthamoeba, Hartmanella and Naegleria spp. (Tyson et al., 2013).

As protozoa are the environmental hosts of Legionella species, they are extremely useful in examining the processes by which infection and replication occur in the ‘natural’ environment. Several protozoan genera have been particularly valuable to researchers, namely Acanthamoeba, Hartmanella, Tetrahymena and – although likely not a natural host – Dictyostelium. Acanthamoeba species comprise one of the most commonly found environmental protozoa, having been identified in soils, rivers, oceans, the air and even the Antarctic (Brown et al., 1982; Sawyer, 1989; Rodriguez-Zaragoza et al., 1993; John and Howard, 1995). By contrast the social amoeba Dictyostelium discoideum is only found in soil, while Tetrahymena ciliates and Hartmanella amoeba are found in freshwater ponds (Unal and Steinert, 2006; Berk and Garduno, 2013).

Although these protozoa are known to be potential hosts for L. pneumophila, they possess distinct properties suitable for different experimental approaches, which answer diverse biological questions. At present, genome sequences of A. castellanii (Clarke et al., 2013), D. discoideum (Eichinger et al., 2005) and the macronucleus of Tetrahymena thermophila (Eisen et al., 2006) are available. In contrast to Acanthamoeba and Tetrahymena spp., D. discoideum is readily genetically tractable (Hägele et al., 2000; Solomon and Isberg, 2000; Kessin, 2001). As such, gene knockouts can be generated to examine the role of host cell factors on Legionella replication, and GFP-fusions with host proteins of interest can be generated and examined by fluorescence microscopy. Moreover, time-lapse life cell microscopy can reveal insights into the dynamics of an L. pneumophila infection (Lu and Clarke, 2005). One disadvantage of D. discoideum is that these amoeba are unable to survive at higher temperatures, and so infections with L. pneumophila must be performed at 25°C, which requires increased replication times when compared with A. castellanii (30°C/37°C), Tetrahymena spp. (30°C) or mammalian cell lines (37°C) (Moffat and Tompkins, 1992; Faulkner et al., 2008).

In contrast to protozoa macrophages are not ‘natural’ host cells although, unfortunately, the mechanisms evolved to avoid predation by protozoa are also extremely well suited to avoid degradation by macrophages of the human immune system (Gao et al., 1997). Macrophages and the human lung are a dead-end from the bacterial point of view, as no human-to-human transmission has been reported. However, inhalation of Legionella-contaminated aerosols can lead to bacterial replication within the alveolar macrophages and epithelial cells, eventually causing the severe pneumonia Legionnaires' disease (Newton et al., 2010).

Legionella pneumophila is an intracellular pathogen, and thus much research has focused on pathogen–host cell interactions. While the natural environment of L. pneumophila is highly diverse, researchers have gravitated towards a few model cell lines, using the strengths and weaknesses of these to determine different features of the bacteria. Here we summarize recent findings on the most important processes of cellular virulence of Legionella, as well as the most useful cell-based models and techniques presently available for Legionella research (Table 1).

Table 1. Techniques employed to study Legionella infection of host cells
 A. castellaniiD. discoideumMacrophagesEpithelial cells
  1. Notes: Uptake of L. pneumophila is much less efficient in epithelial cells versus protozoa or macrophages (primary cells and cell lines). RNA interference is feasible with epithelial cells but not with protozoa or macrophages. For details see text.
UptakeFluorescence microscopy, cfu, flow cytometryFluorescence microscopy, cfu, flow cytometryFluorescence microscopy, cfu, flow cytometryFluorescence microscopy, cfu, flow cytometry
LCV formationFluorescence microscopyMicroscopy (GFP fusions), proteomicsFluorescence microscopy, proteomicsFluorescence microscopy, RNAi
ReplicationMicroscopy, cfu, GFP fluorescence, competition, plate testMicroscopy, cfu, GFP fluorescence, transcriptomicsMicroscopy, cfu, GFP fluorescence, transcriptomicsMicroscopy, cfu, GFP fluorescence, RNAi
Host responseMicroscopy, flow cytometryMicroscopy, flow cytometry, transcriptomicsMicroscopy, flow cytometryMicroscopy, flow cytometry, RNAi
ExitMicroscopy, cfu, flow cytometryMicroscopy, cfu, flow cytometryMicroscopy, cfu, flow cytometryMicroscopy, cfu

Uptake of L. pneumophila

Attachment and uptake are the primary steps of infection. L. pneumophila expresses several adhesins to facilitate specific binding to different host cells (Fig. 1). Although the molecular mechanisms underlying attachment and uptake are still not fully understood, some adhesins and the corresponding host receptors have already been determined. RtxA, PilEL, EnhC and Hsp60 mediate attachment to A. castellanii as well as to epithelial cells (Garduno et al., 1998; Cirillo et al., 2001; Rossier and Cianciotto, 2001), whereas the integrin analogue LaiA mediates binding to epithelial cells exclusively (Chang et al., 2005).

The galactose/N-acetylgalactosamine (Gal/GalNAc)-inhibitable lectin facilitates binding of L. pneumophila to the amoeba Hartmanella vermiformis, which results in invasion associated with tyrosine dephosphorylation of other host cell proteins (Venkataraman et al., 1997; 1998). Proteins such as Lcl act by binding to sulfated glycosaminoglycans of the host extracellular matrix (Duncan et al., 2011). The L. pneumophila major outer membrane proteins (MOMPs) are targeted by complement factors, facilitating binding to macrophages via the host cell complement receptors (Bellinger-Kawahara and Horwitz, 1990). For other adhesins which target macrophages, such as LadC, the receptor is still unknown (Newton et al., 2008). After attachment to the host cells, bacteria are either taken up via coiling phagocytosis (Horwitz, 1984), or by macropinocytosis as shown for D. discoideum and macrophages (Watarai et al., 2001; Peracino et al., 2010). In general, the uptake of L. pneumophila into their hosts is dependent on both actin and actin-binding coronin (Lu and Clarke, 2005).

To analyse these primary uptake steps experimentally, different technical approaches are available. Flow cytometry allows quantitative analysis of a large number of individual cells. Using GFP-producing L. pneumophila, the rate and degree of phagocytosis, e.g. by wild-type or mutant D. discoideum can be assessed at very early time-points by determining the uptake index, calculated from the signal-strength and percentage of infected cells (Harf et al., 1997; Kessler et al., 2013; Tiaden et al., 2013).

Another common method for measuring bacterial uptake is fluorescence microscopy, which can be used to determine uptake and attachment of L. pneumophila to amoeba or macrophages. To this end an anti-Legionella antibody is employed with and without permeabilization to differentiate between intra- and extracellular bacteria (Hilbi et al., 2001; Newton et al., 2008). In the gentamicin-protection-assay, extracellular bacteria are killed after infection by the antibiotic gentamicin, the cells are lysed by a detergent, plated on appropriate agar, and colony-forming units (cfu) are counted (Hilbi et al., 2001). The uptake of L. pneumophila by amoeba and macrophages is regulated by the Icm/Dot T4SS, yet the effectors regulating this process have not been identified (Hilbi et al., 2001; Watarai et al., 2001).

Formation of the Legionella-containing vacuole

After uptake, L. pneumophila remodels the phagosome into a membrane-bound compartment termed LCV (Fig. 1). LCVs acquire components of many host cell organelles, such as early and late endosomes, mitochondria, ribosomes and the endoplasmic reticulum (ER), however fusion with lysosomes is strictly avoided (Isberg et al., 2009; Hilbi and Haas, 2012). In order to determine on a global scale, which host cell proteins are recruited to the LCV membrane, proteomics studies using D. discoideum amoeba were performed (Shevchuk et al., 2009; Urwyler et al., 2009b).

An established protocol for LCV isolation from D. discoideum uses an antibody against the L. pneumophila effector protein SidC (see below), which specifically localizes to the pathogen vacuole membrane, and a secondary antibody coupled to magnetic beads. This allows the isolation of intact LCVs by immuno-magnetic separation, followed by classical density gradient centrifugation (Urwyler et al., 2010). In principle, this purification protocol is applicable to any L. pneumophila-infected host cell, including D. discoideum, RAW 264.7 macrophage-like cells (Hoffmann et al., 2013), primary macrophages (J. Naujoks, C. Hoffmann, B. Opitz and H. Hilbi, unpublished) and likely also A. castellanii. Recent proteomics analysis identified more than 670 or 1150 host cell proteins on LCVs from D. discoideum or RAW 264.7 macrophages respectively (C. Hoffmann, A. Otto, D. Becher and H. Hilbi, unpublished). Proteomics data can be validated by fluorescence microscopy using D. discoideum producing GFP fusion proteins of interest (Urwyler et al., 2009b; Rothmeier et al., 2013), and by RNA interference (RNAi) in epithelial cells (Finsel et al., 2013; Rothmeier et al., 2013) or Drosophila Kc167 phagocytes (Dorer et al., 2006; Brombacher et al., 2009).

The proteomics data generated implicate that LCVs communicate with numerous host cell compartments and vesicle trafficking pathways (Urwyler et al., 2009a). Indeed, L. pneumophila produces the amazing number of ∼ 300 different effector proteins that are translocated into host cells via the Icm/Dot T4SS and are thought to subvert host processes (Hubber and Roy, 2010; Zhu et al., 2011; Lifshitz et al., 2013). Although the functions of many effector proteins are still unknown, several of them target small GTPases regulating the secretory or endosomal trafficking pathways.

The small GTPase Arf1 is activated by a single effector, RalF, which functions as a guanine exchange factor (GEF) (Nagai et al., 2002). In contrast, Rab1 is a targeted by at least six L. pneumophila effector proteins, which function as a GEF (SidM), AMPylase (SidM), de-AMPylase (SidD), phosphocholinase (AnkX), de-phosphocholinase (Lem3), activator/stabilizer (LidA) or a GTPase activating protein (GAP; LepB) (Itzen and Goody, 2011; Hilbi and Haas, 2012). Furthermore, the L. pneumophila effector VipD tightly binds to the small GTPases Rab5 and Rab22, thereby blocking endosomal trafficking (Ku et al., 2012). Thus, small host GTPases, essential for secretory or endocytic vesicle trafficking, are modified in a number of ways by L. pneumophila effectors.

Recently, the small GTPase Ran has been shown to be the target of the L. pneumophila Icm/Dot substrate LegG1 (Rothmeier et al., 2013). Ran and its effector RanBP1 localize to LCVs and play an important role for intracellular growth of L. pneumophila. The effector LegG1 promotes intracellular growth, accumulates on LCVs and activates Ran, which results in a profound stabilization of the microtubules required for pathogen vacuole motility. D. discoideum amoeba producing GFP-α-tubulin have been instrumental to analyse microtubule stabilization by LegG1 and the motility of LCVs harbouring wild-type or legG1 mutant L. pneumophila (Rothmeier et al., 2013).

Phosphoinositide (PI) lipids represent another class of eukaryotic factors, which play key roles in signal transduction and membrane dynamics and which are also exploited by L. pneumophila. Several Icm/Dot-translocated effector proteins anchor to the LCV membrane via PI lipids (Weber et al., 2009b; 2013; Hilbi et al., 2011b). The effectors SidC and SidM specifically bind to phosphatidylinositol-4-phosphate (PtdIns(4)P) and interfere with ER fusion and the secretory pathway respectively (Weber et al., 2006; Brombacher et al., 2009). The 20 kDa PtdIns(4)P-binding domain of SidC, P4C, can be ectopically produced in D. discoideum and used to localize PtdIns(4)P in uninfected or Legionella-infected amoeba (Ragaz et al., 2008). PtdIns(4)P might be produced on the LCV membrane by the L. pneumophila PI polyphosphate 3-phosphatase SidF (Hsu et al., 2012). Furthermore, the Icm/Dot-translocated glucosyl-transferase SetA and another effector, RidL, bind PtdIns(3)P (Jank et al., 2012; Finsel et al., 2013).

Defined D. discoideum mutant strains are valuable tools to identify and characterize host factors implicated in L. pneumophila-phagocyte interactions, which are also relevant for mammalian cells. The D. discoideum PI 5-phosphatase Dd5P4 localizes to LCVs in the amoeba, as does its mammalian homologue OCRL1 (oculocerebrorenal syndrome of Lowe) in macrophages (Weber et al., 2009a). L. pneumophila grows much more efficiently in D. discoideum lacking Dd5P4 (Weber et al., 2009a), and correspondingly, in epithelial cells upon depletion of OCRL1 or other components implicated in the retrograde endosome to trans-Golgi network (TGN) vesicle trafficking pathway (Finsel et al., 2013).

A recent study addressed the role of retrograde trafficking for intracellular replication of L. pneumophila (Finsel et al., 2013). The Icm/Dot substrate RidL supports intracellular growth of L. pneumophila, localizes to LCVs and selectively binds to the Vps29 subunit of the eukaryotic retromer cargo recognition subunit, which by promoting retrograde trafficking restricts intracellular bacterial growth. Using GFP fusion proteins or antibodies the retromer cargo recognition subunits were found to localize to LCVs in D. discoideum or macrophages respectively. The ectopic production of RidL inhibited retrograde trafficking of Cholera toxin in epithelial cells, and dependent on ridL, L. pneumophila blocked retrograde transport of Shiga toxin at endosome exit sites in infected macrophages, indicating that RidL inhibits retromer function to promote bacterial replication (Finsel et al., 2013).

L. pneumophila effectors not only interfere with endocytic, secretory and retrograde vesicle trafficking, but also with lysosomes, autophagosomes and mitochondria. The Icm/Dot substrate SidK interacts with the VatA subunit of the late endosomal/lysosomal V-ATPase, thereby impeding acidification of the LCV (Xu et al., 2010). The effector RavZ hydrolytically removes the phosphatidylethanolamine (PE) moiety from the autophagy component Atg8, thus irreversibly releasing this protein from autophagosome membranes and inhibiting autophagy (Choy et al., 2012). Finally, a homologue of eukaryotic sphingosine-1-phosphate lyase, LegS2, is targeted to mitochondria; yet the function of this Icm/Dot substrate during infection is unknown (Degtyar et al., 2009).

The genetic deletion of distinct L. pneumophila effectors usually does not significantly impair intracellular bacterial replication, likely due to the robustness of the system and redundancy of effector functions. However, most L. pneumophila mutant strains lacking single effectors or regulatory factors are outcompeted by wild-type bacteria in A. castellanii amoeba competition assays (Finsel et al., 2013; Kessler et al., 2013; Rothmeier et al., 2013). In contrast, L. pneumophila strains lacking single or multiple gene clusters are severely defective for intracellular growth (Tiaden et al., 2008; O'Connor et al., 2011). The chromosomal deletion of around 13% of all protein-encoding genes and 31% of all known Icm/Dot substrates yielded an L. pneumophila mutant strain impaired for growth in amoeba (Dictyostelium, Hartmanella and Acanthamoeba) but not in macrophages, suggesting that the effectors (co-)determine host cell specificity (O'Connor et al., 2011).

Intracellular replication of L. pneumophila

Following formation of the LCV, L. pneumophila switches to a replicative phase and begins to multiply (Fig. 1). Studies of the L. pneumophila transcriptome indicated that expression of almost half the genome is altered during the shift between replicative and transmissive phases, as observed for intracellular growth in A. castellanii or macrophages, as well as in broth cultures (Brüggemann et al., 2006; Faucher et al., 2011). Upon intracellular replication in the human monocyte cell line THP-1, L. pneumophila induced genes involved in amino acid biosynthetic pathways as well as amino acid and iron uptake transport systems (Faucher et al., 2011). Moreover, genes involved in catabolism of glycerol were also upregulated, suggesting that glycerol is used as a carbon source for intracellular growth. At the end of the replicative cycle, genes encoding several Icm/Dot-translocated effectors were strongly induced. L. pneumophila also contains a number of regulatory non-coding RNAs (Weissenmayer et al., 2011; Sahr et al., 2012). The expression of ncRNAs changes during the biphasic life cycle or following infection of A. castellanii, underscoring a role for these RNAs in virulence.

The intracellular growth rate of L. pneumophila varies according to the host cell and the temperature encountered. In general, L. pneumophila replicates faster at higher temperatures (Moffat and Tompkins, 1992; Ohno et al., 2008), with a steadily decreasing viability observed above 40°C (Kusnetsov et al., 1996). This observation also holds true for outbreaks in buildings: lower temperatures such as those found in natural watercourses are less conductive to bacterial survival and replication within A. castellanii than the higher temperatures usually found within manmade water systems (Ohno et al., 2008). Growth assays in A. castellanii indicated that the number of viable Legionella bacteria drops significantly in the first few hours following uptake, indicating that only some Legionella bacteria are successful in creating a replication-permissive niche (Moffat and Tompkins, 1992; Harrison et al., 2013).

Using amoeba or macrophages as host cells, intracellular replication of L. pneumophila (i.e. increasing numbers of bacteria) can relatively easily be determined. A number of techniques are available, each of which has advantages and disadvantages. The most traditional assay comprises plating out diluted samples on appropriate agar plates and then counting cfu formed within 3 days. Intracellular replication can be followed by lysis of the host cells using shear forces or mild detergent treatment (e.g. 0.8% saponin). This approach has been used to follow the infection and replication of many Legionella species in both A. castellanii and D. discoideum (Moffat and Tompkins, 1992; Hägele et al., 2000). The requirements for replication vary to some extent between phagocytic cells; deletions of some L. pneumophila genes caused severe growth defects in amoeba, while less strongly affecting replication within macrophage-like cells (Hales and Shuman, 1999b; Segal and Shuman, 1999).

Minor growth phenotypes can be observed in the sensitive amoeba plate test, where serial dilutions of bacterial cultures are spotted onto an A. castellanii lawn. This approach allows a semi-quantitative comparison of L. pneumophila mutants to determine small defects in intracellular replication (Albers et al., 2005). A modification of this method, in which mutagenized bacteria are used to infect A. castellanii prior to plating, has been employed to screen clonal libraries of L. pneumophila mutants for diminished virulence phenotypes (Aurass et al., 2009). Small differences in bacterial fitness also become evident in the amoeba competition test, where A. castellanii is co-infected with L. pneumophila wild-type and a mutant strain (Kessler et al., 2013). Even if initially present at a ratio of 10:1, most L. pneumophila effector or other single gene deletion mutants are out-competed by wild-type bacteria within days (Rothmeier et al., 2013). This sensitive assay yielded a replication phenotype for L. pneumophila mutants (lacking, e.g. quorum sensing genes) that otherwise do not show strong or any defects (Spirig et al., 2008; Tiaden et al., 2010; Kessler et al., 2013).

These quantitative cfu-based methods are end-point and single sample assays, not allowing continuously measuring bacterial numbers at consecutive time points. To overcome these limitations, a non-invasive fluorescence-based continuous replication assay has recently been developed (Harrison et al., 2013). The assay performs most robustly when A. castellanii is used as host cells (2 days, 30°C), but also operates with D. discoideum, macrophages and epithelial cells. Moreover, the assay is medium- or high-throughput compatible, is suitable for screening of chemical and siRNA libraries, as well as L. pneumophila deletion mutants (Harrison et al., 2013; Kessler et al., 2013; Rothmeier et al., 2013).

Host cell response to Legionella

During replication L. pneumophila employs and manipulates many host pathways in order to avoid vacuole acidification and degradation. However, to counter a productive bacterial infection host cells have developed antimicrobial mechanisms targeting various intracellular processes such as iron availability, autophagy and apoptosis or pyroptosis (Fig. 1).

The analysis of transcriptional changes in D. discoideum 24 h after L. pneumophila infection showed that a significant number of genes were upregulated, which are involved in metabolic activities generating products required for pathogen proliferation (Farbrother et al., 2006). In contrast, genes encoding components of the autophagy machinery, enzymatic activities involved in bacterial degradation and various cellular activities (e.g. protein biosynthesis and fatty acid modification) were downregulated.

Recent studies demonstrated that L. pneumophila replication is strictly dependent on the intracellular iron pool. Nramp1 (natural resistance-associated macrophage protein 1) is a proton/divalent cation antiporter that protects host cells from Legionella infection by regulating the iron flux into the phagosomes (Fortier et al., 2005). Members of the Nramp superfamily identified in the genome of D. discoideum (Nramp1 and Nramp2) were shown to be required for resistance to L. pneumophila infections (Peracino et al., 2010; 2013). Nramp1 homologues were also identified in A. castellanii (Clarke et al., 2013) and Tetrahymena (Eisen et al., 2006), suggesting that this antimicrobial mechanism is evolutionarily conserved.

Another important innate immune response is autophagy, which is considered a protective mechanism against bacterial infections. In agreement with this notion, L. pneumophila grows more efficiently in some defined D. discoideum mutant strains lacking specific components of the autophagy machinery, e.g. Atg9 (Otto et al., 2004; Tung et al., 2010). The component of the autophagy machinery, Atg8, is coupled to the lipid PE on early autophagosomes, thus marking this compartment for degradation in lysosomes (Ichimura et al., 2000). L. pneumophila produces the effector RavZ, which directly uncouples Atg8 from PE and thus inhibits autophagy (Choy et al., 2012).

Protozoa such as Dictyostelium, Acanthamoeba or Tetrahymena do not harbour caspase cysteine proteases or the transcription factor NF-κB (Eichinger et al., 2005; Eisen et al., 2006; Clarke et al., 2013). Therefore, caspase-mediated apoptosis or NF-κB-dependent immune regulation obviously does not play a role in these phagocytes. In contrast, upon infection with L. pneumophila macrophages recognize flagellin through the NAIP5 (Birc1e)/NLRC4 (Ipaf) inflammasome, which readily triggers caspase-1 activation, pore formation and pyroptosis (Amer et al., 2006; Molofsky et al., 2006; Ren et al., 2006; Zamboni et al., 2006; Fortier et al., 2007; Massis and Zamboni, 2011). Recognition of flagellin through the NAIP5/NLRC4 inflammasome represents an important mechanism of L. pneumophila restriction.

Apoptotic processes are modulated through the Icm/Dot T4SS, suggesting that by inducing anti-apoptotic factors L. pneumophila delays cell death to sustain intracellular replication (Amer, 2010). A recent study demonstrated that caspase-3 activation and apoptosis in macrophages is regulated by at least five Icm/Dot substrates (SidF, LegS2, VipD, Ceg18 and Lem12) (Zhu et al., 2013). Moreover, several Icm/Dot substrates, including LegK1, LnaB, SdbA and LubX activate NF-κB, thereby triggering the expression of at least 12 anti-apoptotic genes (Abu-Zant et al., 2007; Ge et al., 2009). Thus, L. pneumophila modulates host cell death pathways in an intricate manner.

Exit of L. pneumophila from phagocytes

Once replication is complete, L. pneumophila shifts into the infectious, ‘transmissive’ phase, which is characterized by the upregulation of invasion and virulence genes, such as the flagellar machinery and substrates of the Icm/Dot T4SS (Brüggemann et al., 2006). Completion of the growth cycle requires exit of the host cell and return to the environment. Several processes have been described, which may allow this to occur (Fig. 1).

Legionella pneumophila might lyse host cells by secreted proteins. Non-apoptotic lysis of the LCV has been reported in Acanthamoeba polyphaga, allowing release of the bacteria into the cytoplasm, accompanied by further cell necrosis (Molmeret et al., 2002; 2004; 2010). L. pneumophila exhibits cytolytic activity against a variety of cells, which is thought to be mediated by haemolytic proteins such as legiolysin (Wintermeyer et al., 1994) and Msp (Keen and Hoffman, 1989; Szeto and Shuman, 1990), as well as bacteria-associated and type II- or type IV-secreted phospholipases (Bender et al., 2009; Lang et al., 2012; Aurass et al., 2013). However, individual deletion of these genes does not affect toxicity towards amoeba or macrophages.

Legionella pneumophila also induces apoptosis in host macrophages (Müller et al., 1996), a process which appears to utilize the caspase pathway (Gao and Abu Kwaik, 1999), which might represent another exit strategy. Several effector proteins affect caspase-3-dependent apoptosis, balancing pro- and anti-apoptotic signalling (Zhu et al., 2013). However, apoptosis does not appear to occur following infection of natural hosts such as A. castellanii (Hägele et al., 1998), indicating that this mechanism is not relevant to leave natural hosts. Yet, amoebae infected with L. pneumophila still lose their membrane integrity and are killed, as measured by propidium iodide uptake and flow cytometry (Tiaden et al., 2007; 2008).

Legionella pneumophila might also exit host cells by non-lytic release. Hijacking of the exocytosis pathway by effector proteins such as LepB has been described (Chen et al., 2004). Beyond this, Acanthamoeba spp. also undergo encystment during times of stress, a process which includes the expulsion of food vacuoles into the surrounding environment (Stewart and Weisman, 1972). Legionella bacteria within these expelled vesicles are able to survive for up to 6 months in nutrient-poor media, and are partially protected from antibiotic treatment (Bouyer et al., 2007). Similar processes have been observed for Tetrahymena spp., as concentrated pellets of live, highly virulent L. pneumophila in transmissive phase are expelled from the ciliates (Berk et al., 2008).

Legionella grown within and exiting from eukaryotic host cells show significantly different phenotypes when compared with bacteria grown in broth. For example, L. pneumophila released from A. polyphaga is more resistant to antibiotic treatment (Barker et al., 1995), while bacteria grown within A. castellanii are more virulent than the same strains grown on agar plates (Cirillo et al., 1994). Moreover, the specific amoeba host also seems to affect the physiology of L. pneumophila, as bacteria released from H. vermiformis are more resistant to chlorination compared with A. castellanii (Chang et al., 2009). Thus, the protozoan environment wherein Legionella spp. replicate determines bacterial physiology and egress.

Conclusions

The progression of Legionella infection is complex, involving a multitude of bacterial effector proteins targeting molecular processes of host cells. Protozoa represent invaluable model organisms to study these interactions, both as natural hosts and as a paradigm for infection of human macrophages. The various stages of the infection process have been successfully studied using the different methods outlined and as such provide a strong basis for further understanding the pathogenesis of L. pneumophila.

Acknowledgements

Work in the group of H. H. was funded by the Max von Pettenkofer Institute, Ludwig-Maximilians University Munich, the German Research Foundation (DFG; HI 1511/1-1, HI 1511/2-1, SPP1580), the ‘Bundesministerium für Bildung und Forschung’ (BMBF; ‘Medical Infection Genomics’ initiative, 0315834C) and the Swiss National Science Foundation (31003A-125369).

Ancillary