Functional Type 1 Secretion System Involved in Legionella pneumophila Virulence

Abstract

Legionella pneumophila is a Gram-negative pathogen found mainly in water, either in a free-living form or within infected protozoans, where it replicates. This bacterium can also infect humans by inhalation of contaminated aerosols, causing a severe form of pneumonia called legionellosis or Legionnaires' disease. The involvement of type II and IV secretion systems in the virulence of L. pneumophila is now well documented. Despite bioinformatic studies showing that a type I secretion system (T1SS) could be present in this pathogen, the functionality of this system based on the LssB, LssD, and TolC proteins has never been established. Here, we report the demonstration of the functionality of the T1SS, as well as its role in the infectious cycle of L. pneumophila. Using deletion mutants and fusion proteins, we demonstrated that the repeats-in-toxin protein RtxA is secreted through an LssB-LssD-TolC-dependent mechanism. Moreover, fluorescence monitoring and confocal microscopy showed that this T1SS is required for entry into the host cell, although it seems dispensable to the intracellular cycle. Together, these results underline the active participation of L. pneumophila, via its T1SS, in its internalization into host cells.

INTRODUCTION

Legionella pneumophila is a Gram-negative pathogen that colonizes aquatic environments, where it survives by infecting water protozoans, especially amoebas. Pathogenic strains of L. pneumophila emerge from the environment after replication in such protozoans, are disseminated through aerosols of contaminated water, and reach human alveolar macrophages, where they begin a new infectious cycle, causing a severe form of pneumonia called Legionnaires' disease or legionellosis. The development of air-conditioning systems, cooling towers, and other aerosol-generating systems is frequently reported as the reason for the spread of this pathogen (1). Indeed, L. pneumophila is the second etiological agent of pneumonia requiring hospitalization in intensive care units, after Streptococcus pneumoniae (2). Moreover, the mortality rate of Legionnaires' disease, even with appropriate antibiotic treatment, ranges from 7 to 25%, thus making legionellosis a public health concern (3). Moreover, L. pneumophila is a scientifically relevant model to study intracellular pathogens (4).

Among pathogenic bacteria, secretion systems play a crucial role in virulence, whether by damaging the host or by being essential for bacterial replication, for example, by hijacking cellular pathways or promoting escape from the immune system (5). Bacterial protein secretion systems have been extensively studied, and to date, seven of them have been identified within two categories: those that address their substrates to the extracellular environment, such as type I secretion systems (T1SSs), T2SSs, T5SSs, and T7SSs, and those whose substrates are injected through the host membrane into its cytoplasm, such as T3SSs, T4SSs, and T5SSs (6). In L. pneumophila, a T2SS and a T4SS were identified several years ago (7–9) and their implication in the virulence of this bacterium has been extensively studied (10, 11). T4SS Dot/Icm has been particularly well investigated, as it is responsible for the translocation of more than 275 effector proteins into the cytoplasm of infected cells (12, 13). Those effectors are required for the entire intracellular cycle of L. pneumophila, as they are involved in the creation of a replicative niche inside the host cell, called a Legionella-containing vacuole (LCV), that is suitable for bacterial replication. A decade ago, bioinformatic studies predicted a set of genes that code for a putative T1SS in Legionella, but the functionality of such a system has never been demonstrated (14). The implication of T1SSs in the virulence of pathogenic bacteria, such as uropathogenic Escherichia coli (UPEC), Bordetella pertussis, or even the entomopathogen Photorhabdus, was demonstrated many years ago (15–17). To date, the most extensively studied examples of T1SSs are the Hly system in UPEC (18), the Apr system in Pseudomonas aeruginosa (19), and the Prt system in Dickeya dadantii (20). For example, in the case of UPEC, the system responsible for the secretion of the hemolytic toxin HlyA is based on HlyB, an inner membrane ABC transporter involved in substrate recognition and translocation, the periplasmic membrane fusion protein HlyD, and the outer membrane-spanning porin TolC (21). Together, these three components form a channel that spans both membranes to export substrates into the extracellular environment. The most commonly found substrates are toxins such as lipases and proteases, adhesins, and hemophores (22). Of the T1SS-secreted toxins, the repeats-in-toxin (RTX) family is the most extensively studied. RTX proteins are large, multidomain proteins that have diverse functions, depending on the embedded functional domains. HlyA exhibits a pore-forming activity, CyaA of Bordetella pertussis has an adenylate cyclase activity, and LapA and LapF of Pseudomonas putida are adhesins involved in cell surface or cell-cell interactions, respectively (23). They also contain several repeats of the glycine-rich nonapeptide motif GGXGXDXXX, which is the signature of RTX proteins (24). The RtxA protein of L. pneumophila also contains a high number of tandem repeats, whose exact number and nature vary among strains (25). In this pathogen, RtxA has been linked to the abilities of the bacterium to invade and replicate within amoebas and macrophages (26, 27). However, the mechanism of the potential secretion of this protein has never been investigated.

To date, there is no available study of the functionality of the T1SS in L. pneumophila or its role in virulence. In this paper, we report that LssD-LssD-TolC is a functional T1SS and that RtxA is a substrate of this T1SS. Moreover, we show that it is involved in virulence for amoeba and macrophages, precisely in early steps of infection, such as entry into the host cell.

MATERIALS AND METHODS

Strains and growth conditions.

All of the strains used in these study are listed in Table S1 in the supplemental material. L. pneumophila strains Lens and Paris were grown on buffered charcoal yeast extract (BCYE) agar or in liquid BYE medium. Unless otherwise mentioned, all cultures were grown at 30°C since we observed in our lab that virulence for amoebas was optimal at this temperature. Kanamycin (10 μg ml⁻¹), chloramphenicol (5 μg ml⁻¹), or gentamicin (5 μg ml⁻¹) was added when appropriate. E. coli strains were grown at 37°C in LB medium supplemented with chloramphenicol (5 μg ml⁻¹) or ampicillin (100 μg ml⁻¹) when appropriate. Axenic Acanthamoeba castellanii cells were grown on proteose-yeast extract-glucose medium at 30°C and split once a week. Dictyostelium discoideum Dd04 expressing calnexin-green fluorescent protein (GFP) (DBS0236184) was obtained from the Dicty Stock Center (http://dictybase.org/StockCenter/StockCenter.html; depositor, A. Muller-Taubenberger). D. discoideum cells were grown axenically at 22°C in HL5 medium supplemented with G418 at 20 μg ml⁻¹ when necessary. Human acute monocytic leukemia (THP1) cells and U937 cells were maintained at 37°C and 5% CO₂ in RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum. Differentiation into macrophages was triggered by the addition of phorbol 12-myristate 13-acetate at a final concentration of 100 ng ml⁻¹.

Plasmid constructions.

All DNA constructions were made by using E. coli XL1 Blue or DH5α as the host. L. pneumophila Lens and Paris genomic DNAs were used as the templates for PCRs with the primers listed in Table S2 in the supplemental material. Plasmids pXDC50 (28) and pXDC61 (29) were obtained from Xavier Charpentier (see Table S1 in the supplemental material). All recombinant plasmids were systematically verified by enzymatic digestion before electroporation into L. pneumophila (2,400 V, 200 Ω, 25 μF).

Gene inactivation in L. pneumophila.

Gene-specific knockout strains were constructed by the homologous-recombination method. Briefly, 2-kb flanking regions located upstream and downstream of the gene to be inactivated were amplified by PCR. A kanamycin resistance cassette was inserted between the two regions by a double-joint PCR. To obtain mutant strains derived from wild-type (WT) Lens, the final PCR fragment was cloned into the pLaw344 plasmid (30), which can be selected on chloramphenicol and counterselected on sucrose-supplemented medium. After electroporation and selection, Cm^r clones containing the plasmid were grown for 24 h in liquid medium without antibiotic and spread onto BCYE agar plates containing kanamycin (10 μg ml⁻¹) and sucrose (5% final concentration). Recombinant clones were then analyzed to confirm the deletion of the target gene(s) and the loss of pLaw344. To knock out genes in strain Paris, bacteria were grown overnight at 37°C to an optical density at 600 nm (OD₆₀₀) of 1 to 1.2. One to two micrograms of a purified PCR DNA fragment containing the resistance cassette flanked by the two 2-kb homologous regions were added to the bacteria. The tubes were incubated overnight at 30°C without shaking. Bacteria were then plated on BCYE agar containing 20 μg ml⁻¹ kanamycin. Recombinant clones were verified by PCR.

Intracellular growth.

A. castellanii cells were seeded into a 96-well microplate at 1 × 10⁵/well and allowed to adhere for at least 2 h at 30°C. After washing with peptone-yeast extract medium, amoebas were infected at a multiplicity of infection (MOI) of 10 with bacterial suspensions made by dilution of late-stationary-phase cultures of L. pneumophila strains expressing mCherry (pXDC50). Plates were then centrifuged for 10 min at 600 × g to phase bacterium-cell contact and incubated for 1 h at 30°C. Amoeba were washed three times to remove nonadherent bacteria, and plates were incubated at 30°C for 48 h.

D. discoideum cells were seeded at 1 × 10⁵/well and allowed to adhere for 16 h at 22°C. Infections were performed in the same way by using MB medium (yeast extract, 7.15 g liter⁻¹; peptone, 14.3 g liter⁻¹; MES [morpholineethanesulfonic acid], 3.9 g liter⁻¹; pH 6.9) at 25°C.

Intracellular growth was monitored by measuring the intensity of fluorescence emitted by mCherry (excitation wavelength, 580 nm; emission wavelength, 620 nm) with a TECAN InfinitePro plate reader. The plate reader was thermostat regulated at 25 or 30°C for D. discoideum or A. castellanii, respectively.

Cytotoxicity assays.

Cell viability, hence, bacterium-induced cell death or cytotoxicity, was measured by using the alamarBlue viability indicator according to the manufacturer's protocol (Life Technologies). Briefly, after 24 or 48 h of incubation, infected cells were washed three times in the appropriate medium and 100 μl of 10% alamarBlue diluted in medium was added to each well. Plates were incubated at the temperature suitable for the cells considered for 4 to 16 h, and absorbances at 570 and 600 nm were recorded with a BioTek Instruments μQuant plate reader. Cell viability was estimated by comparing the 570-nm/600-nm absorbance ratio with the ratio obtained with noninfected cells.

Recruitment of the endoplasmic reticulum (ER) to the LCV in D. discoideum.

D. discoideum cells producing calnexin-GFP were seeded into sterile glass coverslips on six-well plates at 5 × 10⁶/well in HL5 medium and allowed to adhere overnight. Monolayers were infected at an MOI of 100 with mCherry-producing L. pneumophila grown for 4 days at 30°C. The plates were centrifuged at 880 × g for 10 min and incubated for 1 h at 25°C. The medium was then carefully removed, and the monolayers were fixed with 3.7% paraformaldehyde (30 min, 4°C). Coverslips were examined with an inverted confocal microscope (Axiovert 200M; Zeiss, Thornwood, NJ).

Pore-forming activity assays.

The pore-forming activity assay used was previously described by Kirby et al. (31). Briefly, A. castellanii cells were used to seed coverslips placed on a six-well plate and infected at an MOI of 500 for 1 h at 30°C. The coverslips were then inverted onto a 5-μl drop of phosphate-buffered saline containing 25 μg ml⁻¹ ethidium bromide and 5 μg ml⁻¹ acridine orange placed on a glass side. Coverslips were immediately observed by using appropriate filters with a Zeiss fluorescence microscope.

β-Lactamase fusion construction and enzymatic-activity assay.

The last 162 or 731 amino acids of the coding sequence of the L. pneumophila Paris rtxA gene were PCR amplified with the primers listed in Table S2 in the supplemental material. The resulting PCR fragments were digested and cloned into pXDC61 in frame with the 3′ terminus of the blaM gene (with the sequence encoding the signal peptide deleted), creating translational fusions Bla-RtxA₁₆₂ and Bla-RtxA₇₃₁, respectively.

L. pneumophila strains carrying pblaM-rtxA₁₆₂, pblaM-rtxA₇₃₁, or pblaM-fabI (no-secretion control) were grown in BYE liquid medium to stationary phase. Cells were then pelleted by centrifugation, adjusted to an OD₆₀₀ of 20 in BYE medium supplemented with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) to induce the production of the hybrid proteins, and incubated for 3 h at 30°C without shaking. Bacterial suspensions were then added to a sterile paper disk placed onto an E. coli MG1655 layer made on an ampicillin-supplemented LB agar plate, and the growth of E. coli around the disk was monitored after a 16-h incubation at 37°C. Supernatants were also collected after centrifugation and dropped onto an E. coli MG1655 layer made on an ampicillin-supplemented LB agar plate. Growth of E. coli was monitored as previously mentioned.

mCherry fusion construction and secretion assay.

The blaM-containing fragments of plasmids pblaM-rtxA₁₆₂ and pblaM-rtxA₇₃₁ were excised (NdeI/KpnI), and the PCR-amplified mCherry-encoding gene was cloned into the same sites, generating pmCherry-rtxA₁₆₂ and pmCherry-rtxA₇₃₁, respectively. Plasmids plssBD_h, ptolC_h, and plssBD-tolC_h were constructed by cloning the lssBD, tolC, or lssBD and tolC genes into the p15A-derived plasmid pACYC184kan under the control of the constitutive pKan promoter. For the secretion assay, E. coli MG1655 strains carrying one fusion-expressing plasmid and one T1SS gene-expressing plasmid were grown to late exponential phase, centrifuged, and resuspended in fresh LB medium supplemented with 1 mM IPTG. Induction of hybrid protein expression was carried out at 37°C for 30 min, the supernatant was collected, and mCherry fluorescence was monitored in a 96-well microplate for 16 h at 30°C in a TECAN InfinitePro plate reader.

RESULTS

L. pneumophila encodes two putative members of a T1SS.

Jacobi and Heuner identified two genes, lssB and lssD, of L. pneumophila as part of a putative T1SS, based on similarity with toxin transporter systems of Vibrio cholerae, Salmonella typhi, and E. coli (14). Moreover, we showed in a previous study that a tolC gene is present in the L. pneumophila genome and that it is involved in early steps of host invasion, as a ΔtolC mutant exhibits impaired virulence for amoebas and macrophages (32). A similar phenotype has also been observed in a ΔrtxA mutant strain (26, 27). Carefully analyzing the LssB protein sequence, and also the classical ABC transporter domains (a transmembrane domain and a nucleotide binding domain), we noticed the presence of a C39 peptidase-like motif (from position 16 to position 137) that has been recently reported as a signature of T1SSs that export RTX proteins (33). In the case of a functional C39 peptidase, the T1SS is involved in the export of bacteriocins, and the corresponding activity (C39 peptidase) is essential for the maturation of the secreted bacteriocins (cleavage downstream of a GG pattern of a prodomain). At the difference, the C39 peptidase-like motif corresponds to a degenerated catalytic site of a C39 peptidase and the exported proteins all belong to the RTX family and are not matured during export.

We therefore hypothesized that LssB-LssD-TolC acts as a functional T1SS and has a role in the virulence of this bacterium, presumably by allowing L. pneumophila to secrete the RtxA protein. Therefore, we inactivated the lssB, lssD, and rtxA genes in the epidemic strain Lens, as well as in the endemic strain Paris, by homologous recombination and studied the roles of these genes in the physiology and virulence of the bacteria. It should be noted that lssB and lssD have been found to be the last genes of a six-gene operon by studying mRNA synthesis (14), therefore reducing a possible polar effect of inactivation. The roles of the other genes products within this operon have yet to be documented. The rtxA gene is annotated on genomes with its own transcript unit.

The Legionella lssB and lssD genes are required for intracellular replication.

As a preliminary test, the growth of each strain was followed by measuring absorbance at 600 nm. No growth difference between the WT and mutant strains could be observed, indicating that lssB and lssD are not essential for optimal growth in the laboratory (see Fig. S1 in the supplemental material). The ability to replicate within different hosts was then investigated by using the strains carrying pXDC50. The intensity of mCherry fluorescence was monitored for 2 to 6 days after contact with two models of environmental hosts, A. castellanii and D. discoideum (Fig. 1A and B). As expected, a ΔdotA mutant cannot replicate within amoebas. This mutant is unable to assemble a functional T4SS, which is necessary for the translocation of many effectors in the cytosol of infected cells and for intracellular growth (34). WT strain Lens showed detectable replication after 8 h in A. castellanii and 40 h in D. discoideum (Fig. 1A and B), the difference in time being attributed to the difference in the temperatures used in these infection experiments. For A. castellanii, the optimal temperature is 30°C, while D. discoideum infections were carried out at 25°C (closer to the optimum temperature of growth for these amoebas, 22°C). A delay in the appearance of ΔlssBD mutant strain replication in both hosts was clearly observed (38 and 105 h, respectively). A ΔlssBD mutant strain carrying a plasmid with the WT copy of the native lssB and lssD genes exhibits an intracellular replication profile in A. castellanii that is similar to the WT strain profile. This complementation is only partial in D. discoideum, but the difference from the ΔlssBD mutant strain profile is still significant.

FIG 1.

Intracellular replication of L. pneumophila Lens and mutant derivatives in eucaryotic host cells. Intracellular replication of the ΔlssBD mutant in A. castellanii (A) and D. discoideum (B) was monitored by measuring fluorescence intensity with a TECAN plate reader (excitation wavelength, 580 nm; emission wavelength, 620 nm). All bacteria expressed the mCherry fluorescent protein under the control of an IPTG-inducible promoter (plasmid pXDC50). The results are representative of at least four infection experiments (MOI = 10) performed in triplicate (errors bars are indicated). RLU, relative fluorescence units.

lssB and lssD mediate bacterium-induced cytotoxicity.

Cell death induction is known to be correlated with the virulence of L. pneumophila. We investigated the cytotoxicity of the ΔlssBD mutant strain for the amoeba A. castellanii and found that its ability to cause cell death was dramatically lower (<30%) than that of the WT strain (70%) (Fig. 2A). This cytotoxicity was at least partially restored by complementation with a WT copy of these genes. Observations by optical microscopy confirmed this lack of cytotoxic effect of the ΔlssBD mutant (see Fig. S2 in the supplemental material). Moreover, this phenotype was also observed in THP1 macrophages (10% in the case of the ΔlssBD mutant strain versus 65% in the case of the WT) and plasmid complementation totally restored Legionella cytotoxicity (Fig. 2B). Interestingly, these results suggest that the mechanism underlying this cytotoxicity may be effective against these phylogenetically distant hosts.

FIG 2.

Cytotoxicity of L. pneumophila Lens and mutant derivatives for eucaryotic host cells. Infections of A. castellanii (A) and THP1 macrophages (B) were carried out as described in Materials and Methods, and host cell death was quantified by using the alamarBlue dye 48 h after the initial contact. Standard deviations are represented as error bars; the results are mean values of three independent experiments (MOI = 10) performed in triplicate (n.s., not significantly different; **, P < 0.01; ****, P < 0.0001 [Student's t test]).

LssB and LssD are involved in early steps of the infectious cycle.

By using confocal microscopy, we analyzed the ability of the ΔlssBD mutant strain to create a functional replicative niche within the infected cell (Fig. 3A). The availability of numerous genetic tools and a library of D. discoideum mutants made it a good model for such studies. Indeed, we used an ER marker (calnexin) coupled to GFP expressed in D. discoideum to show that the average number of bacteria per infected cell was around nine for the WT strain (Fig. 3B), which is similar to the number observed when using the ΔdotA mutant strain, indicating that a functional T4SS is not required for efficient entry into the host cell. Interestingly, the ΔlssBD mutant strain exhibited a significantly decreased internalization level, as only five bacteria were found inside infected host cells, on average. The WT phenotype could be restored by plasmid complementation (Fig. 3B).

FIG 3.

L. pneumophila entry into host cells and ER recruitment to the LCV. D. discoideum cells expressing calnexin-GFP were used to seed glass coverslips and infected with L. pneumophila WT Lens and mutant derivatives expressing mCherry (MOI = 100). After 1 h, the coverslips were fixed, mounted. and observed by confocal microscopy (A). Statistical analysis was performed by observing at least 200 host cells in three independent experiments, and the average number of bacteria in each cell (B) and the proportion of internalized bacteria actually recruiting ER to their LCV (C) are shown (n.s., not significantly different; ****, P < 0.0001 [Student's t test]). The symbol ‡ indicates that no statistical analysis could be performed because the ΔdotA mutant is, by definition, unable to recruit ER.

A statistical analysis of the observation of more than 200 host cells from three independent experiments was performed to estimate the level of ER recruitment to LCV within infected cells. One hour after contact, approximately 65% of the internalized WT bacterial strain constituted an ER-surrounded LCV (Fig. 3C). As expected, a ΔdotA mutant strain is unable to form a mature LCV and no ER recruitment could be detected (Fig. 3A and C). The ability of the ΔlssBD mutant to recruit ER-derived vesicles around the LCV was found to be statistically comparable to that of the WT strain, as well as to that of the complemented strain. These results showed that the lssB and lssD genes are required in an early step of the infectious cycle, presumably during entry into the host cell, but do not seem essential to the constitution of the ER-surrounded LCV within the first hour of infection.

Secretion of the protein RtxA is abolished in ΔlssBD and ΔtolC mutant strains.

RTXs are known substrates of T1SSs. On the basis of the characteristics shared by proteins of the RTX family, Cirillo et al. identified an rtx locus in the L. pneumophila AA100 genome, containing the rtxA gene, that is required for virulence to amoebas and human macrophages (27). Indeed, the ability of a ΔrtxA mutant to adhere to and enter host cells was reduced to 40% of the ability of the WT strain. We obtained similar results when using an L. pneumophila ΔrtxA mutant strain derived from WT strain Paris (see Fig. S3 in the supplemental material). Moreover, the cytotoxicity of the ΔrtxA mutant strain for A. castellanii amoebas was lower (40%) than that of the WT Paris strain (60%). This phenotype was similar to that of the ΔlssBD strain (see Fig. S4 in the supplemental material).

Although the secretion signal for T1SS substrates is not conserved, it has been shown to be uncleaved and located in the last 50 to 60 C-terminal residues of the substrate proteins (21). To assess whether LssB, LssD, and TolC are able to export Legionella RtxA protein, we constructed two different translational fusions of the gene encoding the β-lactamase (blaM) with its signal sequence deleted and 3′ portions of the rtxA gene respectively encoding 162 and 731 amino acids of the RtxA C-terminal region. A blaM-fabI fusion was used as a negative control for secretion, as FabI is an L. pneumophila cytoplasmic protein (29). The production of all three hybrid proteins was assessed by Western blotting with anti-BlaM antibodies (see Fig. 5). Comparable amounts of protein were detected with BlaM, BlaM-RtxA₇₃₁, and BlaM-FabI, but it seems that smaller amounts were produced in the case of BlaM-RtxA₁₆₂. Moreover, two bands can be seen for hybrid proteins, certainly because of the instability of the fusion products. The secretion of these hybrid proteins was then studied by monitoring the growth of ampicillin-sensitive E. coli in the presence of L. pneumophila Paris and mutant derivatives producing these hybrid proteins. This test and the time of study were designed to limit the background, resulting in the detection of hybrid proteins released after cell lysis, which was followed with BlaM-FabI fusion protein. Strain Paris and its ΔlssBD derivative mutant were used for these experiments, and the results presented in Fig. 4A showed that the secretion of the two BlaM-RtxA₁₆₂ and BlaM-RtxA₇₃₁ hybrid proteins was abolished in a ΔlssBD genetic background (no growth of E. coli around the central disk containing the Legionella strain) compared to the active secretion in a WT genetic background. The BlaM-FabI hybrid did not allow good growth of the E. coli ampicillin-sensitive strain, indicating that there was no secretion of the hybrid protein and no lysis (WT) or very little lysis (ΔlssBD) of the L. pneumophila strain during the time of the experiment. A similar experiment was performed with cell-free culture supernatant of hybrid-producing L. pneumophila and showed that the TolC protein is also involved in the secretion of the BlaM-RtxA₇₃₁ hybrid protein as ampicillin was degraded and allowed the ampicillin-sensitive E. coli strain to grow around the supernatant drop (Fig. 4B).

FIG 5.

Heterologous mCherry-RtxA secretion assay. The fluorescence intensity in cell-free culture supernatants of E. coli producing mCherry-RtxA hybrid proteins was monitored 16 h after bacterium removal by centrifugation (folding time). The fluorescence was normalized by the amount of fluorescence in bacteria (production of the protein) and by the fluorescence of the native mCherry-producing strain for each pACYC construct. Results are expressed as the ratio of fluorescence of the plssBD_h-, ptolC_h-, and plssBD-tolC_h-carrying strains to the observed fluorescence of the empty vector (pACYC184kan). Tests were performed for both the mCherry-RtxA₁₆₂ and mCherry-RtxA₇₃₁ hybrid proteins (n.s., nonsignificantly different; **, P < 0.01 [Student's t test]).

FIG 4.

Secretion assay. The secretion of BlaM-RtxA hybrid proteins in the L. pneumophila Paris WT or ΔlssBD or ΔtolC mutant background was assayed by monitoring the growth of an ampicillin-sensitive E. coli strain plated on ampicillin-containing medium. Growth can be restored if the ampicillin on the agar plate is degraded by the secreted β-lactamase activity. L. pneumophila strains expressing hybrid proteins were dropped onto a sterile paper disk, and production of hybrid proteins by bacteria prior to the test was verified by Western blotting of whole-cells extracts (anti-BlaM) (A), or cell-free culture supernatants were dropped directly onto the E. coli monolayer (B).

RtxA is secreted in a LssB-LssD-TolC-dependent manner.

To ensure that the secretion of the hybrid proteins observed in an L. pneumophila WT background is specific to the presence of LssB, LssD, and TolC, the corresponding genes, as well as the genes encoding new hybrid proteins mCherry-RtxA₁₆₂ and mCherry-RtxA₇₃₁, were expressed in an E. coli background (WT strain MG1655). The mCherry fluorescent protein was chosen because of its low folding rate (35), as it was shown that rapid folding of the substrate prior to secretion by the T1SS could block its export (36). The secretion of mCherry-RtxA hybrid proteins can be evaluated by monitoring the fluorescence intensity at 620 nm in the culture supernatant (excitation wavelength, 580 nm). The results showed that the three proteins LssB, LssD, and TolC are required for efficient secretion of the two hybrid proteins tested in this heterologous background (Fig. 5). We also noticed a higher level of secretion (more than twice as high) when using the shortest RtxA end (162 amino acid residues). Since those two hybrid proteins were produced at similar levels (data not shown), we assumed that this is due to inappropriate folding of the larger hybrid protein. The expression of the LssB and LssD proteins in the absence of the Legionella TolC protein led to a low level of fusion protein secretion (increase by a factor of up to five compared to the control strain) that might result from E. coli TolC protein involvement but does not appear statistically significant. It is worth noting that expression of the Legionella TolC protein alone did not result in fusion protein secretion by E. coli.

The lssB, lssD, and tolC genes are required for the RtxA-mediated pore-forming activity of L. pneumophila.

In addition to adherence and virulence, L. pneumophila RtxA was shown to be involved in pore formation in the membranes of infected cells (20). We thus compared the pore-forming activity of the ΔrtxA mutant to those of the ΔlssBD and ΔtolC mutants strains (Fig. 6) and observed a strong decrease in this activity (10-fold decrease) compared to that of WT strain Paris (75% of the visualized amoeba cells lost their membrane integrity) when using all three mutant strains. It is important to note that this decrease was similar when all of the mutant (ΔlssBD, ΔtolC, and ΔrtxA) strains were used, suggesting a shared implication in this activity. Moreover, complementation of the ΔlssBD mutant strain restored the pore-forming activity. Interestingly, the ΔdotA mutant strain exhibited pore-forming activity similar to that of the WT strain, which showed that although it is necessary for intracellular replication and thus for global virulence to host cells, the T4SS is not required for the pore-forming ability of L. pneumophila. These results demonstrate that RtxA pore-forming activity is nearly absent from both T1SS-deficient ΔlssBD and ΔtolC mutants, indicating that the secretion of hybrid proteins observed in the previous experiments was not an artifact and actually occurs in an in vivo infection situation.

FIG 6.

Pore-forming activities of L. pneumophila Paris and mutant derivatives in amoeba cells. A. castellanii cells used to seed coverslips were infected with L. pneumophila at an MOI of 500 for 1 h. Coverslips were inverted on a drop of acridine orange-ethidium bromide staining solution and analyzed by fluorescence microscopy. Cells stained green were considered intact, whereas the membrane of cells stained red was considered damaged. Standard deviations are represented as error bars; the results are mean values of three independent experiments (n.s., not significantly different; ****, P < 0.001 [Student's t test]).

DISCUSSION

The role of secretion systems in bacterial pathogenicity is well established. In L. pneumophila, two secretion systems have been clearly identified (T4ASS Icm/Dot and T2SS Lsp), and their role in the virulence of this bacterium has been extensively studied over the years. Recently, the possibility that the T4BSS Lvh VirD4 protein complements the defect in virulence of a T4ASS-invalidated ΔdotA mutant was reported (11, 37). Genome sequencing data also predicted the existence of a putative autotransporter (T5aSS) in strain Paris (38). Bioinformatic predictions also suggested that a T1SS is present in all L. pneumophila strains (14, 39–42). Moreover, we pointed out the presence of a C39 peptidase-like motif in the L. pneumophila LssB protein, which strongly suggested its classification as a T1SS inner membrane partner involved in the recognition and secretion of RTX proteins (33). In this paper, we demonstrated that the corresponding protein RtxA in Legionella is a substrate of an LssB-LssD-TolC-based T1SS. The structure of L. pneumophila RtxA, however, raised several interesting perspectives concerning the role of the tandem repeats that shape more than half of the protein. Using both homologous and heterologous expression of fusion proteins, we showed that the last 162 amino acids of the RtxA protein were sufficient to promote its secretion, hence eliminating the necessity of tandem repeats for recognition and/or translocation through the T1SS. We are currently investigating the localization of RtxA during the infection of host cells by L. pneumophila, as it was shown in Pseudomonas fluorescens that the RtxA homolog LapA was a surface-associated adhesin (43, 44). Although the roles of the different regions of such large RTX proteins remain to be experimentally assessed, the variability in the number and nature of the tandem repeats between the RtxA proteins of different strains of L. pneumophila may suggest a modulation of the virulence based on these repeats.

Additionally, we documented the phenotype of an L. pneumophila strain defective for this T1SS. The ΔlssBD mutant strain displayed strongly attenuated virulence for A. castellanii and D. discoideum. Moreover, preliminary results showed moderately attenuated virulence for U937 macrophages (data not shown). We were able to study the early steps of the infectious cycle, demonstrating that a ΔlssBD mutant strain is defective for entry into host cells, which is consistent with the observations of Cirillo et al., who used a ΔrtxA mutant strain (26). The absence of a T1SS did not seem to cause a defect in the creation of the replicative LCV, as the recruitment of ER-derived vacuoles to the surface of the phagosome is not altered after 1 h of infection. This recruitment is considered a good marker of the constitution of the LCV and of successful hijacking of host vesicular trafficking (28). Therefore, we hypothesized that the initial entry into the host cell was responsible for the general intracellular replication defect observed. However, the attenuation of entry may not to be sufficient to explain the quasiabolished virulence for amoebas, suggesting that the increased phagosome-lysosome fusion observed in the ΔrtxA mutant by Cirillo and coworkers might also be involved in this phenotype.

Altogether, these results demonstrate for the first time the functionality of a T1SS in L. pneumophila and underline its importance in the virulence of this pathogen. Indeed, the lss locus seems to be present in the majority of the 217 L. pneumophila isolates analyzed by Cazalet and coworkers (25). The rtxA gene encoding the T1SS substrate is also present in all of the isolates tested, though its structure is different between several strain groups. These results underline the active participation of L. pneumophila in its internalization into host cells via its T1SS and its cognate substrate RtxA.