In-vitro cleavage of RtxA by LapG

Previous studies report that some RTX family proteins are susceptible to cleavage by periplasmic proteases in a controlled regulatory process [18, 21]. And since homologs of both substrate and enzyme were identified in L. pneumophila [27], we hypothesized that the periplasmic protease LapG can indeed cleave RtxA. Given that the size of the entire RtxA protein in Legionella pneumophila Paris strain is 6765 amino acids [22], a truncated COOH-His-tagged fragment of RtxANterm (492 amino acids ~ 51.6 kDa; Fig. 1), and a NH2-His-tagged LapG protein minus its secretion signal (188 amino acids ~ 22 kDa) were cloned in appropriate plasmids and overproduced in E. coli. Both expression in E. coli and subsequent purification were conducted independently.

Following co-incubation of the two purified proteins in a suitable buffer for 2 h, reaction mixture components were visualized using SDS-PAGE (Fig. 2; supporting information Figure S8).

Fig. 2

Gel electrophoresis of RtxANterm incubated with LapG protease. Purified RtxANterm (shown in the second lane ~ 55 kDa) was incubated with LapG (~ 23 kDa) in an elution buffer from TAKARA purification kit (50 mM sodium phosphate, 300 mM NaCl, 150 mM imidazole, pH 7.4) containing 40 mM MgCl2 and 80 mM CaCl2 for 2 h at 37 °C. The third lane represents the outcome of the incubation showing two protein bands at 42 and 12 kDa corresponding to the COOH and NH2 fragments of the cleaved RtxANterm in addition to residual uncleaved amount of the latter protein and LapG. Proteins were run on a 12% polyacrylamide gel, PageRuler™ pre-stained protein ladder is used

In the control lane corresponding to RtxANterm incubated alone in the assay buffer, no degradation was observed during the 2 h incubation period. On the contrary, the lane containing the co-incubated proteins show two protein bands (42 and 12 kDa) in addition to LapG and RtxANterm which indicates that part of RtxANterm was cleaved in vitro. Furthermore, the sum of the molecular weights of the 2 RtxANterm post-cleavage fragments is consistent with that of the full-length recombinant protein (54 kDa).

These fragments also coincided with previous reports stating the most probable cleavage site to be around 108th or 109th residue. Therefore, to further pinpoint the exact in vitro cleavage site, the C-fragment (42 kDa) of the cleaved RtxANterm was sequenced by Edman degradation and the identity of its 7 N-terminal amino acids was found to be AGAEAVG, indicating that the cleavage occurs between residues 108 and 109. This result experimentally confirmed that the L. pneumophila RtxA cleavage site is consistent with the putative cleavage site proposed for other RTX proteins, like the one of P. fluorescens [18] (Fig. 3).

Fig. 3

Primary sequence alignment of LapG cleavage site region of various potential predicted RtxA proteins within Legionella species with Pseudomonas LapA proteins. Cleavage site between two alanine residues is indicated by an arrow. The dashes correspond to gaps in aminoacid sequences. The red box surrounds a highly conserved region close to the cleavage site recognized by the LapG protease. The green box surrounded sequence is the outcome of Edman degradation analysis of the cleaved RtxA C-fragment, with cleavage occurring between residues at positions 108 and 109, shown at the top of the figure. This position is well conserved in most of RtxA and LapA proteins with few variations (from 106 to 112 for the first alanine). Purple shading gradation highlights amino acid conservation from an intense purple color (indicating highly conserved residues) to a light purple color (indicating poor conservation). Alignment of RtxA cutting site regions was performed with Jalview software using Clustal Omega software to align the protein sequences (version 2.10.5; [29])

Interestingly, using protein BLAST search, RtxA N-terminus homologues were identified among many Legionella species and the alignment of LapG cutting site region revealed few differences. Most notably the threonine residue in position 107 (before the double A recognized cutting site), present in most cleavage sites among LapA family proteins and Legionella species (except L. waltersii and L. birmighamensis), is substituted by an alanine residue in all L. pneumophila RtxA proteins (Fig. 3). Together, these data emphasize that L. pneumophila RTX proteins defines a well conserved sub-clade within the closely related RTX proteins in the Legionella clade but distinct from Pseudomonas LapA. The phylogeny of the first N-terminal 210 amino acids from predicted proteins supports this observation (supporting information Figure S1).

Phylogeny of LapD/LapG system among Legionella species

Recently, T1SS was reported to be restricted to L. pneumophila species among the genus Legionella [10]. However, the authors searched for the entire lss operon in Legionella, but the whole genetic organization does not seem to be conserved among the genus which may have led the authors to such conclusion. Conversely, our BLAST search found RtxA homologs in many Legionella species, which corroborates the results of another recent study reporting the presence of C39-like T1SS exporters involved in “adhesin-like” export (T1SS-RtxA system) in 19 Legionella species among 45 Legionella species studied [30].

Thus, we performed protein BLAST searches on published Legionella genomes looking for LapG and LapD proteins and components of RtxA/LapA cleavage system, in order to construct phylogenetic trees (LapG phylogeny shown on Fig. 4A).

Fig. 4

LapG family proteins phylogenetic tree inferred using maximum-likelihood (PhyML 3.0 software). (A) The proteins reported in the tree were chosen for their high similarities with Pseudomonas LapG. Only one representative protein is conserved in each bacterial species except in Legionella pneumophila. All bacterial species are Gammaproteobacteria except the two marked with an asterisk which are in Betaproteobacteria class. The branch length is proportional to the number of substitutions per site (scale at the bottom). L: Legionella. Phylogenetic trees were inferred using maximum-likelihood with PhyML 3.0 software online pipeline (http://phylogeny.lirmm.fr; [31]). (B) Distribution of the genes lssB, lssD, tolC, lapD and lapG (with gene copy number) from Legionella in 647 different Legionella strains including L. pneumophila and non-pneumophila strains. The protein sequences of interest of L. pneumophila strain were blasted against our genome database of family-clustered proteins to determine the presence or absence of proteins in the corresponding genomes [32]

Interestingly, we were able to identify LapD and LapG in many species of the Legionella genus that form a clade within L. pneumophila strains. A similar tree was obtained using LapD homologous proteins (supporting information Figure S2).

To further analyze the presence of T1SS genes and lapD/lapG genes in Legionella genomes, we studied their distribution in 647 Legionella genomes (including 540 Legionella pneumophila strains; Fig. 4B). As the rtxA gene is a very long gene harboring many repeat sequences, it is often not annotated in genomes and sequencing data might correspond to uncertain sequence area for these genomes. Thus, we did not include rtxA genes in the analysis at this stage. In a few cases, lssB or lssD genes were present in more than one copy, but only within non-pneumophila genomes. Interestingly, studying the co-occurrence of T1SS genes, lapD/lapG and all the five genes together revealed high conservation of the global machinery (secretion and release of RtxA) among Legionella pneumophila species (over 95% co-occurrence; Fig. 4B). It is worth to note that these genes are located in two different loci in Legionella pneumophila Paris genome: a first one with lssB and lssD, a second one with tolC, lapD and lapG, whereas rtxA is present in a third independent locus.

RtxA release from cell surface is controlled by LapD/LapG system in vivo

We previously demonstrated that L. pneumophila RtxA, as other RTX proteins is transported by a T1SS, namely the LssB/LssD /TolC T1SS [12]. However, the possible fates of the protein after its passage through the secretion machinery were not clear. Following evidence from P. fluorescens [21] and building on the results of our previous in vitro cleavage assays, we hypothesized that LapG cleavage of RtxA results in its release from the cell surface of Legionella.

Specific rabbit polyclonal antibodies against the C terminal region of RtxA were produced. L. pneumophila ΔrtxA, ΔlapG and ΔlapD deletion mutants were designed and constructed to abolish the RtxA production or to disrupt the hypothetical RtxA release process. The ΔrtxA mutant strain was used as a negative control in immunofluorescence microscopy.

As expected, wild type cells (WT, Fig. 5A) show clear fluorescence dots on the contrary to ΔrtxA mutants which are defective for RtxA production and hence do not exhibit any fluorescence.

Fig. 5

Immunofluorescence microscopy of four L. pneumophila strains using anti RtxACterm antibodies. (A) Three days old cultures of L. pneumophila Paris wild type, ΔlapG mutant, ΔlapD mutant and ΔrtxA mutant were used in an indirect immunofluorescence assay using antibodies targeting the C-terminus of RtxA protein at 0.374 µg ml-1. The secondary antibody used was Alexa Fluor® 568 goat anti-rabbit (Invitrogen Inc. USA). Scale bar shown is 100 μm. Aggregated bacteria are pointed out with white arrows. (B) Confocal microscopy image of L. pneumophila Paris and ∆lapG strains stained with DAPI and immunolabelled using antibodies targeting the C-terminus of RtxA protein at 0.374 µg ml-1. The secondary antibody used was Alexa Fluor® 594 goat anti-rabbit (Invitrogen Inc. USA)

Few red dots are present on the image but that might result from antibody aggregates with no link to bacteria cells. Therefore, our negative control is efficient and no cross-reaction of our primary antibodies with other Legionella proteins is detected. In addition, as almost all WT bacteria were marked with red fluorescent dots, our experimental growth conditions seemed to maintain a significant amount of L. pneumophila RtxA on cell surface. Concerning the ΔlapD strain, the absence of fluorescence associated with bacteria cells allowed us to conclude that no RtxA was present at the surface of the cells (Fig. 5A). This result is consistent with the proposed role of LapD. Indeed, our results show that LapD controls the activity of the periplasmic protease LapG, possibly by physically sequestering it close to the outer-face of the inner membrane (similarly to P. fluorescens LapA [27]) and preventing RtxA N-terminus periplasmic cleavage. Therefore, the absence of LapD in the ∆lapD mutant strain may result in continuously free LapG in the periplasm that is able to cleave and release RtxA from the cell surface. Finally, the strain lacking lapG exhibits lots of red dots at the surface of the bacterial cells corresponding to RtxA protein embedded into the outer membrane as visualized in Fig. 5B. The confocal microscopy images clearly showed the presence of RtxA antibodies labelled in red close to the blue stained bacterial cells. Moreover, it is worth to note certain phenotypic characteristics such as aggregation, possibly due to the interaction between the RtxA proteins and neighboring cells, which may in turn cause clumping of the cells (Fig. 5A). The localization of RtxA at the surface of the cells was also revealed on many WT bacterial cells, dependent in that case on the activity/inactivity of LapG at the time of the experiment (Fig. 5B). Therefore, all these results confirmed that the LapD/LapG system controls RtxA localization (embedded or released) in L. pneumophila by blocking or promoting its N-terminus cleavage in the periplasmic space. Interestingly, amoebae and macrophage infection experiments using L. pneumophila ∆lapG strain that retains RtxA on its cell surface allowed us to visualize RtxA immunolabeled bacteria at the surface of eukaryotic cells as soon as 20 min post-infection, suggesting the efficiency of L. pneumophila in targeting its host cells (supporting information Figure S3).

RtxA, a key role in the initial stage of virulence

The rtxA gene has been previously correlated to Legionella virulence, mainly adherence and entry into various host cells [12, 25, 26]. A mutant strain, ΔdotA, defective for Dot/Icm T4SS and hence intracellular replication was used as a negative control. All constructed mutants were tested for growth in AYE medium and no difference in growth capacity/fitness compared to wild-type strain was noticed (supporting information Figure S4 and Fuche et al., 2015). We then followed the progress of amoeba infection for the different strains using light microscopy. It is worth mentioning that all infection experiments made during this work followed a special protocol; L. pneumophila cells were added to host cells (amoebae or macrophages) without any centrifugation, hence avoiding forced contact and adhesion to host cells. Severity of the infection was based on amoebae mortality as well as morphology since they display a more round shape when infected [33], compared to the elongated morphology with possible pseudopods. We sought to recapitulate the insights gained above, regarding the regulated transport of RtxA to the bacterial cell surface, with the impact of RtxA on L. pneumophila virulence. ΔrtxA mutant, ΔlssBD and ΔtolC for the T1SS mutants, and ΔlapD or ΔlapG for the regulatory component of the RtxA localization were used to infect Acanthamoeba castellanii, as described above. Figure 6 images were captured three days after infection at MOI (multiplicity of infection) 1.

Fig. 6

Importance of RtxA secretion and release systems in L. pneumophila virulence. A. castellanii were infected with L. pneumophila strains (WT, ΔlapG, ΔlapD. ΔlssBD, ΔtolC and ΔrtxA) at MOI 1 in a growth inhibitory medium. Light microscopy images were taken 3 days post infection. The round morphology of A. castellanii corresponds to stressed infected cells

Two patterns can be observed regarding the severity of A. castellanii infection. Mutant strains for the T1SS, ΔlssBD and ΔtolC display lower virulence toward A. castellanii, a statement in agreement with our previous work with L. pneumophila Lens [25, 26]. Interestingly, ΔlapD infected cells were stressed in a manner similar to the WT infected cells whereas ΔlapG infected amoebae seemed to have a slight delay of infection (Fig. 6). However, monitoring the bacterial growth within amoebae during the infection showed no difference between WT, ∆lapD and ∆lapG strains (supporting information Figure S4). We can say that disrupting the T1SS responsible for RtxA secretion results in the lower virulence of these strains when compared to WT. It seems that manipulating the presence of RtxA on the cell surface (ΔlapG) versus being constantly released (ΔlapD) didn’t attenuate the virulence of L. pneumophila towards A. castellanii even if RtxA release seemed to be more favorable. Overexpressing lapG in WT strains also resulted in a small increase of fluorescence during amoebae infection which reinforces this hypothesis (supporting information Figure S4C). Complemented strains for the mutants were also constructed, e.g., L. pneumophila ΔlapD/pXDC50-lapD and ΔlapG/pXDC50-lapG. However, the axenic growth was altered in the case of ΔlapD/pXDC50-lapD strain with a start-up delay in AYE medium (supporting information Figure S4E). Moreover, the overexpression of lapD gene in the WT strain was deleterious in liquid media (supporting information Figure S4B). To better understand this phenotypic characteristic, we analyzed the transcriptional profile of the genes involved in RtxA secretion (T1SS), and release (lapD/lapG) at different physiological states: exponential, post-exponential (corresponding to the high infectious potential of Legionella with motility acquisition) and stationary phase (supporting information Figure S5). Interestingly, lapD/lapG genes were expressed at low level (~ 40 counts for lapG and ~ 100 counts for lapD) from the beginning of Legionella growth (exponential phase) and this expression decreased slowly during the post-exponential and stationary phase, corresponding to an early implementation of the complex in inner membrane and periplasm. A similar expression pattern was observed for tolC, mainly expressed during exponential phase (~ 2000 counts), in agreement with its pleiotropic role as a partner of efflux pumps complexes as well as in T1SS machinery. On the contrary, the inner membrane and periplasmic component of T1SS, lssB and lssD, were highly expressed during the post-exponential phase (from 3000 to 7000 counts) and corresponded to the beginning of RtxA expression which was still increased at stationary phase at a very high level (170,000 counts; supporting information Figure S5). Such a high production of RtxA is compatible with the release of RtxA that can be achieved when LapG cuts the N-terminal domain (108 AA). Thus, TolC now being free can interact with other substrate-engaged LssB/lssD and form a new T1SS complex. This RtxA hydrolysis is only possible if LapG is not sequestered by LapD in low c-di-GMP conditions. In the case of LapD overexpression, the constant “trapping” of LapG may result in a saturated T1SS linked to RtxA with the newly synthesized RtxA proteins retained in cytoplasm of bacteria cell with a toxic effect. However, the fact that ∆lapG mutants seemed viable without any toxic effect may reflect more complex organization and interactions between T1SS, RtxA and LapD/lapG components that still have to be elucidated.

In conclusion, transporting RtxA by the T1SS to the cell surface appears to be a crucial factor in L. pneumophila virulence efficiency, most likely in the early steps of infection such as host cell adhesion and/or recognition. We can also hypothesize that the presence of RtxA in the infection medium whether on the cell or released can play this role, independently of its location. However, this protein is not a limiting factor in L. pneumophila virulence as the infection can proceed without it albeit less efficiently.

Protection against L. pneumophila infection of A. Castellani by anti-RtxA antibodies

To further demonstrate the role of RtxA in L. pneumophila virulence, we monitored the effect of anti-RtxA antibodies on A. castellani infection. A. castellanii were infected by L. pneumophila in the presence of the purified L. pneumophila anti-RtxACOOH antibodies described above. Since our antibodies were preserved in glycerol that might affect L. pneumophila virulence, appropriate controls were taken. Figure 7 shows that glycerol on its own is able to reduce the infection potential of L. pneumophila by approximately 36% even at 1% glycerol concentration.

Fig. 7

Anti-RtxACOOH antibodies attenuate L. pneumophila infection of A. castellanii. L. pneumophila Paris wild type (WT) was used to infect A. castellanii cells at MOI 1. The curves represent variation in mCherry fluorescence versus time reflecting the number of bacteria. Results are displayed as mean fluorescence (3 replicates) ± one standard error of the mean. **** P < 0.0001 indicates that the means of our results are significantly different among each other. The significance was calculated by using a multiple comparisons test (Tukey’s test) to compare the mean of all time points in one infection experiment versus that of other infections at alpha 0.05

However, the curve representing an infection in presence of 37.4 µg ml− 1 Anti-RtxACOOH (containing an equivalent of 1% glycerol) was able to further decrease the peak fluorescence and consequently infection efficiency by approximately 62%. Taking into consideration that significance analysis confirmed the significance of our results at P < 0.0001, this suggests that anti-RtxA antibodies are able to hinder the infection process by binding and possibly disrupting RtxA role in amoeba infection.

RtxA does not play a key role in the targeting of amoeba by L. Pneumophila

In order to place the insights acquired here in an ecological context of interaction between L. pneumophila and its protozoan host, we assessed the role of RtxA in L. pneumophila virulence when the bacteria are in competition with other bacterial prey of amoebae. Indeed, phagotrophic protists including amoebae can be very selective consumers that recognize prey organisms [34]. Given the role of RtxA in the early steps of amoeba infection, we addressed the question whether RtxA could participate in the initial interaction/recognition/targeting between L. pneumophila and its environmental host. We thus investigated the effects of the presence of alternative prey cells of amoebae on the infection potential of L. pneumophila. We carried out A. castellanii infection with L. pneumophila WT, ΔdotA and ΔrtxA at MOI 1 in the presence of 100-fold excess cells of E. coli MG1655 (known prey of amoeba) in the infection medium. Figure 8A shows that after three days, A. castellanii cells infected with WT Legionella display significant stress, similar to a normal infection without E. coli, as observed above in Fig. 6.

Fig. 8

L. pneumophila may target A. castellanii during infection. (A) Infection of A. castellanii cells with different L. pneumophila strains (WT, ΔrtxA, ΔdotA) at MOI 1 in a growth inhibitory medium in the presence of 100-fold E. coli cells (MG1655) as a source of nutrition for amoebae. Observations were made after 3 and 7 days. Round morphology of A. castellanii corresponds to stressed infected cells. (B)L. pneumophila Paris wild type (WT) and ∆rtxA mutant harboring pXDC50 plasmid were used to infect A. castellanii cells at MOI 1 in absence or presence of E. coli cells (100x more). The curves represent variation in mCherry fluorescence versus time reflecting the number of bacteria. Results are displayed as mean fluorescence (3 replicates) ± one standard error of the mean. **** P < 0.0001 indicates that the means of our results are significantly different among each other. The significance was calculated by using a multiple comparisons test (Tukey’s test) to compare the mean of all time points in one infection experiment versus that of other infections at alpha 0.05

Moreover, as expected, the ∆rtxA mutants displayed lower virulence than WT after three and seven days compared to WT (Fig. 8A and supporting information Figure S6). However, a quantitative assay using L. pneumophila strains expressing mCherry has shown that the infection cycle was modified in the presence of competitor E. coli cells (Fig. 8B). Although there is no clear delay in the onset of L. pneumophila WT growth within the amoebae, the fluorescence level is lowered at each point in the kinetics. That might be the result of a lower number of amoebae cells infected, thus lowering the fluorescence level as the total Legionella cells number replicating in the whole population of amoebae was decreased. This attenuation of infection potential was expected to be more emphasized in the presence of an alternative nutrient source for amoebae. In other words, the addition of an alternative nutrient source did not drastically alter the infection process. The attenuating effect of this competitor on infection was greater for ∆rtxA strain and might be the consequence of the slower capacity of this defective mutant to enter amoebae, thus giving advantage to phagocytosis of E. coli which might compensate Legionella entry. Keeping in mind that we did not force any contact between amoebae and L. pneumophila or E. coli, we can assume that L. pneumophila may actively target its host cell rather than simply being engulfed and consumed by phagocytic cells, even if the presence of a competitor in high excess (100x) may reduce the total of amoebae cells infected.