The Flagellar Regulon of Legionella—A Review

The Legionella genus comprises more than 60 species. In particular, Legionella pneumophila is known to cause severe illnesses in humans. Legionellaceae are ubiquitous inhabitants of aquatic environments. Some Legionellaceae are motile and their motility is important to move around in habitats. Motility can be considered as a potential virulence factor as already shown for various human pathogens. The genes of the flagellar system, regulator and structural genes, are structured in hierarchical levels described as the flagellar regulon. Their expression is modulated by various environmental factors. For L. pneumophila it was shown that the expression of genes of the flagellar regulon is modulated by the actual growth phase and temperature. Especially, flagellated Legionella are known to express genes during the transmissive phase of growth that are involved in the expression of virulence traits. It has been demonstrated that the alternative sigma-28 factor is part of the link between virulence expression and motility. In the following review, the structure of the flagellar regulon of L. pneumophila is discussed and compared to other flagellar systems of different Legionella species. Recently, it has been described that Legionella micdadei and Legionella fallonii contain a second putative partial flagellar system. Hence, the report will focus on flagellated and non-flagellated Legionella strains, phylogenetic relationships, the role and function of the alternative sigma factor (FliA) and its anti-sigma-28 factor (FlgM).


INTRODUCTION AND OVERVIEW
The Legionellaceae family consists of a single genus: Legionella that comprises more than 60 species so far (Gomez-Valero et al., 2009;Bajrai et al., 2016;Khodr et al., 2016). New species are identified continuously (i.e., Legionella drancourtii, Legionella gresilensis, and Legionella beliardensis), extending the list of known Legionella species (Lo Presti et al., 2001;La Scola et al., 2004;Gomez-Valero et al., 2009;Rizzardi et al., 2015;Bajrai et al., 2016;Khodr et al., 2016). More than 20 pathogenic Legionella species are known today that differ in their ability to infect hosts and to cause severe to mild diseases in humans (Rizzardi et al., 2015). Human pathogens known to cause the Legionnaires' disease-an atypical pneumonia-are for instance Legionella pneumophila, Legionella micdadei, and Legionella longbeachae (Yu et al., 2002;Whiley and Bentham, 2011). Legionella known to cause the Pontiac fever-a mild flu-like disease-are for instance Legionella feelei, L. micdadei and Legionella anisa, but also L. pneumophila. Often Legionella strains of the same species and same serogroup cause one of the mentioned diseases (Swanson and Hammer, 2000;Fields et al., 2002;Wang et al., 2015).
The assumption that the expression of flagella and virulence are linked was already made at an early stage (Rowbotham, 1986) and later on confirmed. It was shown that there is a regulatory link between the expression of a virulent phenotype and the flagellum (Pruckler et al., 1995;Byrne and Swanson, 1998;Hammer et al., 2002;Gal-Mor and Segal, 2003;Molofsky et al., 2005;Heuner and Albert-Weissenberger, 2008;Albert-Weissenberger et al., 2010;Schulz et al., 2012). The expression of flagellar genes is regulated on the flagellar regulon, extensively investigated in L. pneumophila due to its biphasic intracellular life cycle during which the bacterium undergoes a shape change. Within the host, after replication inside of Legionellacontaining vacuoles (LCVs), when nutrients become limited, L. pneumophila differentiates into a flagellated, non-replicating form. The flagellated, transmissible, mature form is stressresistant, virulent and metabolically resting as well as infectious (abbr. MIF) (Rowbotham, 1986;Byrne and Swanson, 1998;Heuner et al., 1999;Swanson and Hammer, 2000;Faulkner and Garduno, 2002;Garduno et al., 2002;Hammer et al., 2002;Molofsky and Swanson, 2004;Fonseca and Swanson, 2014;Eisenreich and Heuner, 2016). The actual release process of mature forms is still under discussion: either the bacteria are released from the LCV into the environment by lysis of the host or the bacteria are released first into the cytosol of the host and then after an additional putative round of replication into the environment (Rowbotham, 1986;Molmeret et al., 2004). The latter hypothesis implies that the flagellum is produced inside the cytosol of the host and not in LCVs as proposed earlier. Furthermore, there is also a possibility that the bacteria are released by the host via a non-lytic mechanism (Chen et al., 2004;Bouyer et al., 2007;Berk et al., 2008). However, the released form is well-prepared to reinfect new hosts or to differentiate into a viable-but-nonculturable form (VBNC) meant to enable a longterm survival of the bacteria (Rowbotham, 1986;Steinert et al., 1997;Ohno et al., 2003;Molmeret et al., 2010;Al-Bana et al., 2014). VBNC forms can be resuscitated when they are taken up by amoebae (Steinert et al., 1997;Ohno et al., 2003;Al-Bana et al., 2014). Further, different morphological forms of L. pneumophila have been recently discussed (Robertson et al., 2014). Next to the biphasic intracellular life cycle it was shown that L. pneumophila exhibits also a life stage-specific bipartite metabolism, an area for further investigations (Schunder et al., 2014;Eisenreich and Heuner, 2016;Gillmaier et al., 2016;Hauslein et al., 2016).

MOTILITY OF LEGIONELLA
Different forms of bacterial motility are known including swarming, twitching and sliding. Notably the flagellum-next to pili-allows bacteria to move. Bacterial motility is often related to chemotactic behavior that enables a bacterium to locate special environmental conditions and to get closer to higher concentrations of attractants (Szurmant and Ordal, 2004;Hazelbauer et al., 2008;Micali and Endres, 2016). Some Legionella have a chemotaxis system (L. longbeachae, Legionella parisiensis, and Legionella bozemanii) but most Legionella do not have the corresponding genes (e.g., L. pneumophila, L. micdadei, and L. oakridgensis). Moreover, the ability of Legionella to swarm and to show a chemotaxis behavior has not yet been reported.
Twitching motility is based on a functional type IV pilus. The ability to move forward by twitching has been reported for L. pneumophila (Coil and Anne, 2009;Hoppe et al., 2017). In addition, sliding motility, a surfactant-mediated motility, has been described for L. pneumophila (Stewart et al., 2009).

THE FLAGELLUM AND THE FLAGELLAR REGULON
The Structure of the Flagellum and Flagellar Systems Most Legionella species are motile due to a single polar flagellum (Figure 1) (Chandler et al., 1980;Elliott and Johnson, 1982;Heuner et al., 1995). More than 50 genes are involved in the expression of functional flagella, and due to high metabolic costs, a tight regulation is essential (Chilcott and Hughes, 2000;McCarter, 2006;Osterman et al., 2015). The flagellum of Legionella consists of a basal body, a hook structure and a filament (Figure 2; Heuner and Steinert, 2003;Heuner and Albert-Weissenberger, 2008). For the assembly of the flagellum, needed proteins (hook, rod and the filament forming proteins) are exported out of the cell by a flagellum-specific export apparatus, a type III-like secretion system (T3SS) (Heuner and Albert-Weissenberger, 2008;Altegoer and Bange, 2015).
The basal body consists of a rod, three rings ["membrane/supramembrane" (MS), "peptidoglycan" (P), and "lipopolysaccharide" (L)] and a motor switch complex (MotAB). The MotAB is providing the energy for the rotation of the flagellum (Minamino and Imada, 2015). For the formation of the slightly curved hook structure, FlgE and FlgD are essential. The hook cap protein FlgD assists when FlgE is incorporated into the hook structure (Altegoer and Bange, 2015).
Interestingly, an uncommonly straight hook has been reported for L. pneumophila mutant strains ( fliA, fliD, and flaA, i.e., a flagellin mutant) (Figure 1) (Schulz et al., 2012). The findings give credit to the assumption that L. pneumophila might have a straight hook that is hard to detect in wildtype strains. The filament consists mainly out of a single protein, the flagellin (FlaA or FliC) (Figure 2). The cap protein FliD is essential for the assembly of flagellin subunits into the filament. To assemble the filament, flagellin is exported through the filament structure by a flagellum-specific export apparatus (T3SS) and assembled at the tip of the filament (Heuner and Albert-Weissenberger, 2008;Altegoer and Bange, 2015).
Next to the regular flagellar system, a second putative flagellar system was suspected for two Legionella species: Legionella fallonii and L. micdadei (Gomez-Valero et al., 2014). Comparative genome analysis led to the suspicion that the strains do have homologs to flagellar genes of L. pneumophila. The identified genetic region is comprised of genes that encode a putative basal body, a secretion system, as well as a putative hook structure. No homologs to flaA or fliD were found in the predicted genomic region. Yet, further investigations are needed to find out if a T3SS or a putative second flagellum is encoded. Additionally, in silico investigation, performed on the draft genome sequence of Legionella israelensis, identified a similar operon. A BLAST search using the operon (10,041 bp, ctg_064, L. israeliensis draft-genome; Burstein et al., 2016) as query identified similar genes in L. drancourtii, L. fallonii, L. worsleiensis, L. quateirensis, L. birminghamensis, and L. drozanskii (Heuner, unpublished results). Initial findings show that the operon is present in two out of three Legionella clades (Figure 3), leaving a margin for additional studies.
In short, intracellular alarmone accumulation-ppGpp, a signal molecule, produced when environmental conditions are unfavorable (e.g., limited nutrient supply)-is regulated by RelA and SpoT. RelA senses the amount of available intracellular amino acids and SpoT senses the amount of intracellular fatty acids (Hammer and Swanson, 1999;Dalebroux et al., 2009Dalebroux et al., , 2010. Alarmone accumulation triggers the activation of an alternative stationary-phase sigma factor (RpoS) and of the two component system LetA/S (Byrne and Swanson, 1998;Hammer et al., 2002;Zusman et al., 2002;Molofsky and Swanson, 2004;Dalebroux et al., 2009Dalebroux et al., , 2010Edwards et al., 2009;Rasis and Segal, 2009;Sahr et al., 2009) (Figure 4). RpoS and LetA/S promote the transcription of small regulatory RNAs (RsmX, RsmY, RsmZ). RsmX plays a role in the virulence of L. pneumophila (Sahr et al., 2012). The transcription of rsmZ/rmsY is also influenced by a quorum sensing system regulator, LqsR (Tiaden et al., 2007;Schell et al., 2016). The two regulatory RNAs are able to bind a number of carbon storage regulator molecules (CsrA) at once (Sahr et al., 2017). CsrA is a negative regulator and through the binding on RsmY or RsmZ, other targets of the regulatory RNAs can be expressed. The expression of transmissive traits starts and main activator proteins (e.g., FleQ) are produced and flagellar genes are expressed (Zusman et al., 2002;Molofsky and Swanson, 2003;Rasis and Segal, 2009;Sahr et al., 2009Sahr et al., , 2017Albert-Weissenberger et al., 2010). Notably, some parts of the function of the negative regulator CsrA (flaA expression and motility of L. pneumophila) can be "complemented" by ectopically expressed csrT, a CsrA-like regulatory gene associated with integrative conjugative elements (Abbott et al., 2015).
CrsA is also controlling a major regulator involved in the expression of flagella, FleQ (Sahr et al., 2017). FleQ is responsible for the expression of early flagellar genes belonging to class II and III genes in an RpoN-dependent and RpoN-independent pathway (Jacobi et al., 2004;Albert-Weissenberger et al., 2010;Schulz et al., 2012). RpoN is an enhancer-binding protein encoding an alternative sigma factor that initiates transcription when activator proteins like FleQ, FleR, and PilR (Jacobi et al., 2004) are present. When class II and III genes are expressed, the activity of FliA-the alternative sigma factor-leads to the expression of class IV genes and the assembly of the flagellum (Figure 4). RpoN and FleR seem to be responsible for a negative feedback loop on flagellar genes (Albert-Weissenberger et al., 2010). It was found that RpoS and FlaR (transcriptional regulator FlaR, LysR family member) are also involved in the expression of the flagellin gene (Heuner et al., 2000;Swanson, 2001, 2004;Rasis and Segal, 2009;Sahr et al., 2009). The production of FlaA is also regulated by cyclic di-GMP, shown by the analysis of a gene (cdgS13) coding for a protein with diguanylate cyclase activity (Levi et al., 2011). The influence of cyclic di-GMP on flagellum-based motility has been shown for other bacteria than Legionella species (Wolfe and Visick, 2008).

THE FLAGELLUM AND VIRULENCE
Already early on, it has been hypothesized that virulence and flagellum expression are genetically linked with each other. It has been shown that motile, transmissive Legionella were more infectious for amoebae than non-motile replicative phase Legionella (Rowbotham, 1986;Pruckler et al., 1995;Bosshardt et al., 1997;Byrne and Swanson, 1998;Hammer et al., 2002;Heuner et al., 2002;Molofsky et al., 2005;Heuner and Albert-Weissenberger, 2008). Different experiments could show that the motility but not the flagellin promotes the contact with host cells. Motility increases the infectivity and the fitness. Furthermore, it turned out that the flagellum is not necessary for intracellular replication (Pruckler et al., 1995;Dietrich et al., 2001;Polesky et al., 2001;Heuner et al., 2002;Jacobi et al., 2004;Molofsky et al., 2005;Schulz et al., 2012).
The four major regulators of the flagellar regulon (RpoN, FleQ, FleSR, FliA) seems to be involved in the invasion process of L. pneumophila into hosts. These findings point out the proposed link between virulence traits and flagellum expression (Dietrich et al., 2001;Hammer et al., 2002;Molofsky et al., 2005;Heuner and Albert-Weissenberger, 2008;Albert-Weissenberger et al., 2010;Schulz et al., 2012). Especially the FliA regulon plays an important role (please, see the section: FliA and its implication in virulence below). FliA, but not the flagellin (flaA), is involved into the ability of L. pneumophila to form biofilm that allow bacteria to survive whenever environmental conditions are not favorable. (Mampel et al., 2006). Notably unwanted biofilms are a health issue causing a significant amount of nosocomial infections (Bryers, 2008). Legionella are known to survive within biofilm of other bacteria (e.g., Klebsiella pneumophila and Pseudomonas aeruginosa) (Molofsky et al., 2005;Stewart et al., 2012). More recently, findings about L. pneumophila's ability to form biofilms by itself in natural environments and on medical devices have attracted attention (Lau and Ashbolt, 2009;Abu Khweek et al., 2013). Biofilm formation is regulated by temperature, surface material and intracellular growth (Konishi et al., 2006;Piao et al., 2006;Bigot et al., 2013) and biofilm-derived L. pneumophila do FIGURE 2 | Overview of the structure of a flagellum. Schematic drawing of a flagellum consisting of basal body, hook structure and filament. Cap, FliD protein that is essential for the assembly of the filament, Motor, MotAB proteins that provide energy for the rotation; T3SS, type III-like secretion system; FlgD (not shown), the hook cap protein needed to incorporate FlgE; IM, inner membrane; OM, outer membrane; PG, peptidoglycan layer; (Heuner and Albert-Weissenberger (2008), modified). not express flagellin (Abu Khweek et al., 2013). More information about biofilms and L. pneumophila can be found in a recent review (Abdel-Nour et al., 2013).
The flagellum also affects the resistance of hosts to Legionnaires' disease and when Legionella do not produce flagellin they can evade the innate immune response in macrophages (Hawn et al., 2003;Molofsky et al., 2006;Ren et al., 2006;Abu Khweek et al., 2013). Resistance is mediated by the Naip5/Ipaf-dependent recognition of flagellin, which induces a protective immunity in non-A/J mouse models (Ricci et al., 2005). Detailed information about addressed points can be found in dedicated reviews (Fontana and Vance, 2011;Schell et al., 2016;Mascarenhas and Zamboni, 2017).

FliA Expression
One of the major regulators involved in the expression of the flagellum is FliA and an increased alarmone level leads to accumulation of functional FliA (Bruggemann et al., 2006;Heuner et al., 2006;Dalebroux et al., 2010). The alternative sigma factor (σ 28 ) is directly involved in the regulation and expression of the flagellin gene (flaA) and others (Figure 4). A fliA mutant of L. pneumophila does not produce flagellin and is consequently non-flagellated. Moreover, a fliA mutant of Escherichia coli can be completed with a fliA gene of L. pneumophila (Heuner et al., 1995(Heuner et al., , 1997(Heuner et al., , 2002Bruggemann et al., 2006;Albert-Weissenberger et al., 2010;Schulz et al., 2012).
The expression of flagellar class III and IV genes is induced in a FleQ-dependent manner. The FliA-regulated class IV genes are involved in the assembly of the filament and flagella motility (flgL, fliD, flaA, motY). Both lead to the complete synthesis of the flagellum (Jacobi et al., 2004;Albert-Weissenberger et al., 2010). The fliA gene itself is expressed in a FleQ-dependent but RpoN-independent manner (Albert-Weissenberger et al., 2010). Nevertheless, FleQ and RpoN are not necessary for a basal expression of fliA. For a basal expression, fliA is transcribed from a putative sigma-70 promoter element and later, during the exponential phase, the expression of fliA is induced in a FleQ-dependent manner (Schulz et al., 2012). Accordingly, it was hypothesized that during the exponential phase the basal fliA promotor activity may be mediated by DksA independent of the ppGpp concentration, whereas during the post-exponential phase DksA cooperates with ppGpp to activate fliA (Dalebroux et al., 2010). The identification of the transcription start point of fliA corroborates the presence of a putative DksA binding site, an A/T rich discriminator site (Schulz et al., 2012).

FliA and Its Implication in Virulence
FliA is a regulator that is also involved in the expression of putative virulence genes (Bruggemann et al., 2006;Albert-Weissenberger et al., 2010;Tlapak et al., 2017).
Several investigations performed on a fliA mutant strain of L. pneumophila pointed out that the mutant (at low MOI) is not replicating in host cells anymore (Dictyostelium discoideum). The fliA mutant seems to be less infectious for macrophages and non-cytotoxic to bone marrow-derived macrophages. Moreover, the mutant has a reduced fitness FIGURE 3 | Evolutionary relationships of Legionella species and distribution of flagellar regulon and chemotaxis genes. Shown is a phylogenetic tree of Legionella spp. based on the fliA gene (left side) and also the distribution of flagellar regulon and chemotaxis genes (right side). The phylogenetic tree was reconstructed on the basis of the fliA gene using the Neighbor-Joining method (Saitou and Nei, 1987) and a bootstrap of 1,000 (Felsenstein, 1985). The bootstrap >50 is reported next to the branches. The phylogenetic distances were computed using the Tamura-Kumar method (Tamura and Kumar, 2002)  potential in amoebae (Dietrich et al., 2001;Hammer et al., 2002;Heuner et al., 2002;Jacobi et al., 2004;Molofsky et al., 2005;Heuner and Albert-Weissenberger, 2008;Schulz et al., 2012). Likewise, a fliA mutant strain of L. oakridgensis showed a reduced fitness in its host (Acanthamoeba lenticulata) (Tlapak et al., 2017). In addition, another L. pneumophila fliA mutant strain exhibited a reduced ability to form biofilms (Mampel et al., 2006).
FliA is obviously a virulence factor, and target genes of fliA were investigated to understand its implication for virulence (Bruggemann et al., 2006;Albert-Weissenberger et al., 2010;Tlapak et al., 2017). Target genes of fliA in L. pneumophila strains include genes of the flagellar regulon (e.g., flaA and flgM) and others (e.g., enhA and lvrA), illustrated in Figure 4 and listed in Table 1 (Bruggemann et al., 2006;Albert-Weissenberger et al., 2010;Schulz et al., 2012). Other identified FliA-dependent genes encode for putative virulence factors corroborating the involvement of FliA in the establishment of Legionella infections. Identified putative virulence factors are: lpp0952, lpp1290, and lpp0972. The first one, lpp0952, is coding for a GGDEF/EAL and PAS/PAC domain protein (Bruggemann et al., 2006;Albert-Weissenberger et al., 2010). The two remaining genes are homologs of the enhanced entry proteins EnhA. Respective homologs were also found in L. longbeachae which is nonflagellated, putatively associated with the flagellar system (Kozak et al., 2010).
Recently, especially L. oakridgensis simplified the identification of fliA targets potentially involved in virulence (Tlapak et al., 2017). L. oakridgensis strains are non-flagellated, the entire flagellar regulon is missing and only homologs of FliA and FlgM are present (Figure 5). L. oakridgensis is less infectious then L. pneumophila, but still causes Legionnaires' disease. In addition, L. oakridgensis replicates in guinea pigs, in human cell lines, in Acanthamoeba lenticulata and for growth in media no additional cysteine is needed. (Orrison et al., 1983;Fields et al., 1986;O'Connell et al., 1996;Neumeister et al., 1997;Lo Presti et al., 2001;Brzuszkiewicz et al., 2013). Nevertheless, L. oakridgensis exhibits a functional T4SS, homologs of known virulence factors, as well as newly identified virulence factors (Brzuszkiewicz et al., 2013). L. oakridgensis is used to investigate FIGURE 4 | The flagellar regulon cascade. The regulatory cascade which leads to the expression of proteins that are needed to construct a functional flagellum is shown (Heuner and Albert-Weissenberger, 2008;Sahr et al., 2017, modified). Specific environmental factors lead to the activation of the pathway through alarmone (ppGpp) accumulation. Some main players of the flagellar regulon cascade are: the activator protein FleQ, the alternative sigma factor 28 (FliA), the two-component system (LetA/S) and the carbon storage regulator (CsrA) as examples. FleQ is the master regulator of the flagellar genes (Class II and class III). FliA is the regulator of flagellar genes of class IV. Black lined, continuous arrows refer to positive regulation events. The broken lines in black refer to negative regulatory effects of respective factors. Question marks (?) indicate pathway points with putative mode of actions that need to be specified by further investigations. RsmXYZ, regulatory RNAs; FlaA, flagellin; FlaR, transcriptional regulator (LysR family); FleQ and FleR, sigma factor 54 activator protein; FlgM, anti-sigma 28 factor; RpoS, alternative sigma 38 factor.
FliA since a fliA L. oakridgensis knockout will not cause the inactivation of the entire flagellar regulon and target genes of FliA can still be identified as well as genes involved in the expression of virulence traits. However, mutant strain analyses aimed at identifying target genes of FliA in L. oakridgensis yielded no results for putative FliA-dependent virulence genes yet ( Table 1; Tlapak et al., 2017). Additional investigations are needed including phenotypic characterizations and deletion analyses of further target genes of FliA.

FliA-FlgM Interaction in Legionella oakridgensis
Flagellated bacteria regulate the FliA activity often posttranscriptionally. For Salmonella, Escherichia, and Vibrio species it is known that an anti-sigma-28 factor (FlgM) binds FliA, preventing the binding of FliA to FliA-dependent promoter sites, and FliA-dependent genes are consequently not translated. After assembling of the hook-basal body structure, FlgM is exported and FliA is not repressed anymore (Gillen and Hughes, 1991;Ohnishi et al., 1992;Chilcott and Hughes, 2000;Aldridge et al., 2006). In Helicobacter pylori, the FlgM protein is inactivated instead of being exported out of the cells (Rust et al., 2009). FliA-FlgM interactions in Legionella are still unknown. Recent findings suggest that FliA-FlgM interaction might be different than in other flagellated bacteria, at least for L. oakridgensis. Respective species do not have a flagellum (Figure 1C), a flagellar regulon ( Figure 5) and a basal body although flgM and fliA homologs are present which encode for functional FlgM and FliA proteins (Brzuszkiewicz et al., 2013;Tlapak et al., 2017); consequently, the mechanism controlling FliA-FlgM interactions must be different.
For L. pneumophila as well as for L. oakridgensis it was found that the expression of FlgM or homologs is sigma-28dependent (Albert-Weissenberger et al., 2010;Tlapak et al., 2017). Although L. oakridgensis has no flagellar system, the expression of fliA-dependent genes is growth phase-and temperature-dependent (Heuner et al., 1999;Tlapak et al., 2017). Moreover, for L. oakridgensis it was demonstrated that FlgM is a negative regulator of FliA-dependent genes and the protein itself seems to be degraded in a growth phase-and temperaturedependent manner (Tlapak et al., 2017). Thus, it seems likely that, as described for H. pylori, FlgM in L. oakridgensis is degraded by protease activity instead of being secreted. However, investigations are needed to show if FlgM in L. pneumophila is effectively secreted in a basal body-dependent manner.

DISTRIBUTION OF THE FLAGELLAR SYSTEM AMONG LEGIONELLA SPECIES
Phylogenetic reconstruction-based on concatenated amino acid alignment of 78 orthologous ORFs-divided the Legionella species into three major clades (clade I and clade II, clade III) (Burstein et al., 2016). Clade I is comprised of most Legionella species including L. pneumophila, L. parisiensis, L. bozemanii, and L. longbeachae (Burstein et al., 2016). Clade II comprises among others L. feelei and L. micdadei and clade III, a deepbranching clade, includes three members: Legionella adelaidensis, L. oakridgensis and Legionella londiniensis (Burstein et al., 2016). Phylogenetic reconstructions performed herein, yielded similae results that are given in Figure 3. As opposed to former investigations, the phylogenetic tree was reconstructed on the basis of the fliA gene.
The flagellar system can be found in Legionella species classified as clade I or II (Cazalet et al., 2004(Cazalet et al., , 2010Chen et al., 2004;Bruggemann et al., 2006;Glockner et al., 2008;Kozak et al., 2010;Brzuszkiewicz et al., 2013;Gomez-Valero et al., 2014;Burstein et al., 2016), but not in clade III Legionella. Clade III Legionella do not have a functional flagellar system and do not have most of the flagellar regulon genes, except for fliA and its anti-sigma factor flgM (L. oakridgensis, L. adelaidensis, and L. londiniensis) (Cazalet et al., 2010;Brzuszkiewicz et al., 2013;Tlapak et al., 2017) and two additional genes: fleQ and fleN (L. londiniensis) (Figures 3, 5). Also, some clade I Legionella species do not have a functional flagellar system (clade I: L. longbeachae, Figure 5, Legionella gratiana, Legionella cincinnatiensis, Legionella sainthelensi, and L. israelensis). It has been hypothesized that the loss of flagellar genes has not happened recently (Kozak et al., 2010). This is corroborated by the finding that L. longbeachae and all subclade members are negative for the flagellar regulon but positive for genes coding for the sigma factor FliA, the regulator FleN and the two component system comprising of FleR and FleS, as well as FlgD (Figure 3) (Cazalet et al., 2010;Kozak et al., 2010). Also L. israelensis is negative for flaA (Heuner et al., 1995) and most flagellar regulon genes, except: fleQ, fliA, fleN, and flgM ( Figure 5). The finding allows to assume that the flagellar system may have been lost at different time points during the evolution of Legionella species (Figure 3). In addition, some genes that have regulatory functions outside of the flagellar system are still present (Albert-Weissenberger et al., 2010;Cazalet et al., 2010;Kozak et al., 2010;Tlapak et al., 2017). The same applies to flgD which is involved in the assembly of the hook structure of the flagellum with unassigned hypothetical alternative functions.
The investigated Legionella genomes were also screened for the presence/absence of genes of major regulators of the flagellar system as well as of the chemotaxis operon (Figure 3). It was found that the genes of the chemotaxis operon are only found in a subclade of the clade I Legionella. L. longbeachae is the first Legionella species described to exhibit chemotaxis genes (Cazalet et al., 2010;Kozak et al., 2010) that do not have flagellar genes. It seemed paradoxical that L. pneumophila is flagella positive but chemotaxis negative and L. longbeachae is flagella negative but chemotaxis positive. The distribution of the chemotaxis operon may indicate that the chemotaxis operon was acquired by a common 'ancestor' of this sub-tree clade (Figure 3).

CONCLUSION
The review aimed to summarize knowledge gained about flagella and the flagellar regulon of different Legionella species. The majority of Legionella species exhibit genes encoding for a functional flagellum and they are flagellated (Elliott and Johnson, 1981;Bornstein et al., 1991;Bangsborg et al., 1995;Heuner et al., 1995). Motility increases infectivity and fitness, helping the bacteria to reach new hosts after successful replication within protozoan host cells and release into aquatic environments.
Some Legionella-including some pathogenic species (e.g., L. longbeachae, and L. oakridgensis)-are not flagellated and most flagellar regulon genes are absent (Orrison et al., 1983;Heuner et al., 1995;Cazalet et al., 2010;Kozak et al., 2010;Brzuszkiewicz et al., 2013). With the advance in molecular techniques and the ability to produce and to process metagenomics datasets, it was found that some of the non-flagellated Legionella have still parts of the flagellar regulon, mainly genes with regulator functions.
However, in Legionella flagellum synthesis is associated with the expression of a virulence phenotype; and motility can be seen as a virulence and a fitness factor in Legionella and other bacteria. The alternative sigma factor FliA is also involved in the expression of virulence traits. FliA-dependent putative virulence genes were already identified by initial investigations that need to be extended. Also, additional investigations are needed to determine the role of FliA and molecular mechanisms of FliA-FlgM interactions in Legionellae. FlgM and FliA, two main players involved in the expression of the flagellum genes, are still present in non-flagellated Legionella, a promising takeoff for future investigations. Nevertheless, the flagellum is not necessarily needed for an intracellular replication within host cells. Moreover, in some hosts the Naip5/Ipaf-dependent recognition of flagellin can cause an innate immune response leading to resistance against Legionella infections (Molofsky et al., Genes are indicated by arrows and gene names are given above. Flagellar and chemotaxis genes are indicated in red and blue, respectively; other genes in black. Genes not present in a strain are indicated by thin arrows (Brzuszkiewicz et al., 2013, modified). 2006; Ren et al., 2006). For example, it was reported that biofilmderived L. pneumophila without flagellin expression evade the innate immune response in macrophages (Abu Khweek et al., 2013), as it was suggested for the non-flagellated L. longbeachae (Cazalet et al., 2010;Kozak et al., 2010). It seems that under certain conditions, the loss of the flagellum may increase the fitness of bacteria. For instance, L. pneumophila which can be found mainly in aquatic environments, is still flagellated whereas L. longbeachae which can be found predominantly in soil, is non-flagellated (Kozak et al., 2010). Nevertheless, if the loss of the flagellar system from Legionella species depends on the habitat or environmental conditions remains unanswered.
As outlined, flagellated and non-flagellated Legionella are positive for genes belonging to the chemotaxis operon. The ability of Legionella to swarm and to show off a chemotaxis behavior has not yet been reported. Interestingly, some chemotaxis-positive and flagellar operon-negative Legionella (e.g., L. longbeachae) give credit to the assumption that chemotaxis genes may not be involved in flagellum-mediated motility. Recent investigations could even show that chemotaxis sensory systems-different from those found in E. coli-in distinct bacteria (e.g., Myxococcus spp., Geobacter spp.) are not necessarily involved in bacterial flagellum-mediated motility (Kirby, 2009;Kozak et al., 2010). Chemotaxis-like systems seem to be involved among other things in type IV pilus-based motility and cell to cell interaction and/or social motility (Kirby, 2009;Kozak et al., 2010). Accordingly, additional experimentations are needed to investigate the role of the chemotaxis operon in flagellated and non-flagellated, chemotaxis-positive Legionella.

AUTHOR CONTRIBUTIONS
SA and KH contributed substantial to the conception and design of the work. SA and KH wrote the paper.