Involvement of a phospholipase C in the hemolytic activity of a clinical strain of Pseudomonas fluorescens

Background Pseudomonas fluorescens is a ubiquitous Gram-negative bacterium frequently encountered in hospitals as a contaminant of injectable material and surfaces. This psychrotrophic bacterium, commonly described as unable to grow at temperatures above 32°C, is now considered non pathogenic. We studied a recently identified clinical strain of P. fluorescens biovar I, MFN1032, which is considered to cause human lung infection and can grow at 37°C in laboratory conditions. Results We found that MFN1032 secreted extracellular factors with a lytic potential at least as high as that of MF37, a psychrotrophic strain of P. fluorescens or the mesophilic opportunistic pathogen, Pseudomonas aeruginosa PAO1. We demonstrated the direct, and indirect – through increases in biosurfactant release – involvement of a phospholipase C in the hemolytic activity of this bacterium. Sequence analysis assigned this phospholipase C to a new group of phospholipases C different from those produced by P. aeruginosa. We show that changes in PlcC production have pleiotropic effects and that plcC overexpression and plcC extinction increase MFN1032 toxicity and colonization, respectively. Conclusion This study provides the first demonstration that a PLC is involved in the secreted hemolytic activity of a clinical strain of Pseudomonas fluorescens. Moreover, this phospholipase C seems to belong to a complex biological network associated with the biosurfactant production.


Background
Pseudomonas fluorescens is a ubiquitous Gram-negative bacterium frequently encountered in the environment. Naturally resistant to a wide range of antibiotics and disinfectants, this bacterium is commonly found in hospitals as a contaminant of cleaning solutions and even injectable material [1]. This bacterium is characterized by its very large genome, containing a large number of regulatory genes conferring a very high potential to adapt to environmental variation [2]. P. fluorescens is generally considered psychrotrophic species, but the presence of this bacterium in hospitals has long been demonstrated [3]. The broad clinical spectrum of this bacterium, which may colonize the airways [4], urinary tract [5] and blood [6] of immunocompromised patients and previously healthy patients, demonstrates that human body temperature is not neces-sarily a barrier to the development of this microorganism. These findings strongly suggest that different strains have evolved to deal with this specific environment [7]. Previous investigations of the infectious potential of P. fluorescens have demonstrated that this bacterium can bind specifically to the cytoplasmic membrane of neurons and glial cells [8]. Attachment to the cytoplasmic membrane of the host cell is associated with the induction of apoptosis and necrosis [9]. LPS has clearly been implicated in cytotoxicity, but other factors released along with LPS during cytoadhesion also seem to be essential for the virulence of this bacterium [10]. The virulence factors of Pseudomonas aeruginosa and Burkholderia sp. have been studied in detail, but little is known about those of P. fluorescens [11,12]. P. fluorescens bacteria synthesize and release various extracellular enzymes, including a protease [13] and a lipase [14], which may be involved in virulence. They also produce phospholipase C (PLC) [15][16][17][18][19] -an enzyme produced by many bacterial pathogens and associated with various levels of virulence [20]. The pathophysiological role of secreted PLCs depends on bacterial species, extending from that of a major virulence factor to a minor metabolic factor involved in bacterial survival or dissemination only, without cytotoxic effects [21,22]. However, the effects of PLCs may be more subtle. PLC interferes with eukaryotic cellular signaling cascades and seems to be able to modulate the host immune response [21]. Several members of the Pseudomonas family produce PLCs [21], and the PLCs synthesized by P. aeruginosa have been studied in detail [23]. The virulence potential of the PLCs of this bacterium has often been associated with hemolytic activity [24]. Conversely, our knowledge about hemolysin production by P. fluorescens remains very limited. The PLCs identified in this species to date have essentially been studied biochemically, without considering their infectious potential.
We studied a recently identified strain of P. fluorescens (MFN1032) [7] with mesophilic behavior and hemolytic activity, comparing the cytotoxicity of the factors secreted by this clinical strain with those secreted by the environmental P. fluorescens strain MF37. We also included the opportunistic pathogen P. aeruginosa in the study, as a reference. We identified a phospholipase C (PlcC) from MFN1032 and compared its amino-acid sequence with those of the PLCs produced by other P. fluorescens species and other Gram-negative and Gram-positive bacteria. We then evaluated the involvement of PlcC in the hemolytic activity of MFN1032, by constructing plcC mutants.

Results
Characterization of the MFN1032 strain MFN1032 has been isolated from the sputum of a patient suffering from pneumonia. This bacterium has been characterized as P. fluorescens biovar I by polyphasic taxonomy, partial 16sRNA sequencing and siderotyping [7]. Culturability studies demonstrated that this strain was able to grow ( Figure 1A) and formed biofilms on polystyrene at 37°C. As PAO1, this bacterium shown strong beta-hemolytic activity on sheep, horse or rabbit blood-agar plates ( Figure 1B), whereas strain MF37 was non-hemolytic. MFN1032 also displayed lecithinase activity on egg-yolk agar plates ( Figure 1C).

Cytotoxic effects of bacterial culture supernatants
We assessed the cytotoxicity of supernatants from P. fluorescens MFN1032, P. fluorescens MF37 and P. aeruginosa PAO1 in two experimental models: LDH release from rat Growth and extracellular activities of MFN1032 Figure 1 Growth and extracellular activities of MFN1032. A-Growth of MFN1032 on LB agar plates after 24 hours at 37°C. B-Secreted hemolytic activity of MFN1032 on sheep red blood agar plate after 48 hours of incubation at 28°C (1), and comparison with PAO1 (2) and MF37 (3). C-MFN1032 lecithinase activity on egg-yolk agar plate after 48 hours of incubation at 28°C. We first investigated cytotoxicity in rat glial cells in primary culture, a model previously used to assess the cytotoxicity of P. aeruginosa and P. fluorescens [10]. Cell-free supernatants from stationary growth phase cultures were incubated overnight with primary cultures of rat glial cells. MFN1032 and MF37 supernatants were much more cytotoxic than those from the opportunistic pathogen PAO1 at all growth temperatures studied ( Figure 2A). MFN1032 supernatants from bacteria grown at 17°C and 28°C were equally toxic (55% lysis), whereas those from bacteria grown at 8°C were loss toxic (25% lysis). MF37 also displayed significantly higher levels of secreted cytotoxic activity against glial cells than PAO1 grown at 17°C or 28°C. This activity was clearly temperature-dependent, and was maximal at 17°C.
We monitored the levels of secreted hemolytic activity displayed by MFN1032 throughout bacterial growth. MFN1032 grown at 17°C displayed hemolytic activity only at the start of the stationary phase and similar profiles were observed at 8°C and 28°C. The kinetics of hemolysin production by PAO1 was investigated only at 37°C. Hemolytic activity appeared to be stable after 15 bacterial generations in culture, in all the cases. These conditions were therefore adopted in all subsequent studies. Major differences were observed in the hemolytic activities of supernatants from cultures of MFN1032, MF37 and PAO1 grown at 8, 17, 28 and 37°C ( Figure 2B). Culture supernatants from MF37 grown at 8, 17 and 28°C were non hemolytic (no test was carried out at 37°C because this strain cannot grow at this temperature). Hemolysis levels for the clinical strain MFN1032 were maximal at growth temperatures of 8 and 17°C (100% and 83% lysis, respectively), and lower at 28°C (26% lysis). MFN1032 supernatants were non hemolytic if the bacteria were cultured at 37°C. MFN1032 and PAO1 supernatants had similar hemolytical potentials if the bacteria were grown at 17°C (90% hemolysis) or 28°C (25% hemolysis) (Figure 2B). The hemolytic activity of PAO1 supernatants was maximal at growth temperatures of 37°C (optimal growth rate) and 17°C. By contrast, the secreted hemolytic activity of the MFN1032 strain was not maximal at the optimal growth rate for this strain (30°C).
As the P. fluorescens clinical strain MFN1032 displayed hemolysis, we investigated the factors potentially involved in this hemolytic activity.
Effect of growth temperature on the cytotoxicity of MFN1032 (white square), MF37 (grey square), and PAO1 (black square)

Measurement of protease activity, LPS and biosurfactant release and lecithinase activity in the supernatant
No protease activity or LPS release (KDO level) into the supernatants of MF37, MFN1032 and PAO1 cultured in LB medium at 28°C was detected (Table 1). We used egg yolk as a substrate for estimating PLC activity and biosurfactant release was assessed with the drop-collapse test. Both lecithinase activity and biosurfactant were detected in these three culture supernatants, although higher levels were observed for MFN1032 and PAO1 than for MF37 ( Table 1).

Effect of D609 on the secreted hemolytic activity of MFN1032
We investigated the possible involvement of a secreted PLC in the hemolytic activity of MFN1032 using D609, which specifically inhibits certain PLCs [25]. Supernatants from MFN1032 were incubated at room temperature for 1 h with D609 at a final concentration of 1 mM, and their hemolytic activity was then assessed. Incubation with D609 decreased hemolytic activity by 68% (Table 2). Thus, a phospholipase C sensitive to D609 is clearly involved in the hemolytic activity of MFN1032. By contrast, D609 treatment did not decrease the hemolytic activity of PAO1 supernatants, consistent with previous results [26].

Structural characterization of the PLC from MFN1032
SDS-PAGE and zymograms were used to determine the apparent molecular mass of an enzyme with lecithinase activity. A single band with lecithinase activity was recovered from a silver-stained SDS-PAGE gel placed on an eggyolk agar plate ( Figure 3). This protein had a molecular mass of approximately 42 kDa. We designed PCR primers based on the gene encoding a PLC of similar molecular mass described by Preuss et al. and produced by an uncharacterized P. fluorescens strain [15]. A single fragment of 1.2 kb was obtained and cloned with the pMOS kit. This fragment was sequenced by Genome Express (France) and the sequence registered in the Genbank database (accession number: DQ462712) named plcC. This sequence encodes a predicted PlcC protein of 385 residues, with a molecular mass of 42 kDa. These findings are consistent with SDS-PAGE and suggest that the protein is secreted without cleavage. In silico analysis indicated the absence of any known secretion signal sequence. No gene amplification was obtained if the same primers were used with MF37 and PAO1 (data not shown).
A search of the NCBI nucleotide and protein database showed that the PLC from MFN1032 (PlcC) was very similar to putative PLCs from other strains of P. fluorescens ( Table 3). The molecular mass of this group of PLCs ranges from 42 kDa to 59 kDa. PlcC is clearly related to the PLC described by Preuss et al. (98% identity), which was produced by a P. fluorescens strain isolated as a contaminant of placental extract of unknown origin. These two PLCs are also very similar to a putative PLC encoded by the P. fluorescens SBW25 genome (92% identity), but are more distantly related to the putative PLCs identified for the Pf-5 strain -a rhizosphere strain of P. fluorescens (76% identity) -and Pf0-1 (65% identity). Sequence identity was also observed with putative PLCs from the entomopathogen Pseudomonas entomophila (68%) and from Legionella pneumophila, although the level of identity was lowers (43%). No significant similarity was observed with the PLCs produced by P. aeruginosa (PlcH, PlcN and PlcB) or with the PC-PLC from B. cereus (Table 3).

Construction of MFN1032 plcC mutants
The MFN1036 strain, a clone of MFN1032 that overexpresses plcC, was obtained by electroporating MFN1032 with pUCP20, which carries the plcC gene and its promoter (1.5 kb). The expression of plcC from this plasmid was under the control of the constitutive plac promoter. We investigated the involvement of PlcC in hemolytic activity by culturing MFN1036 at 17°C and 28°C and assessing its hemolytic activity in liquid assays. MFN1036 was significantly more cytotoxic (74% lysis at 28°C) than the parental strain MFN1032 (26% lysis at 28°C) ( Table  4). We used a plcC-deficient mutant, MFN1037, to determine whether PlcC contributed to the hemolytic activity of MFN1032. We detected no hemolysis with MFN1037 culture supernatants, regardless of growth temperature ( Table 4). Complementation of the plcC mutation with pMF36 (strain MFN1038) did not restore the hemolytic phenotype of MFN1032 (Table 4), whereas it did result in levels of lecithin degradation on egg-yolk agar plates similar to those for the wild-type strain.  SDS-PAGE analysis of total extracellular proteins followed by PlcC zymogram detection on egg-yolk agar plates showed that there was no PlcC in the supernatant of cultures of the plcC-deficient mutant MFN1037, as no lecithin degradation was observed ( Figure 4, lane 2 and 5, arrow a). The mutant released smaller amounts of an approximately 50 kDa protein than did MFN1032 (arrow b). Analysis of the amino-acid sequence of the N-terminus of this protein (ALTVNTNIAS) suggested that it was a homolog of the flagellin protein FliC produced by P. aeruginosa and P. putida. The plcC-complemented strain MFN1038 displayed wild-type levels of lecithin degradation in zymogram analysis and a partial restoration of flagellin levels ( Figure 4, lanes 3 and 4, bands a and b respectively). MFN1036 had an electrophoretic profile identical to that of MFN1032, except for the 42 kDa pro-tein previously identified as PlcC, which was produced in larger amounts by MFN1036 (data not shown).

Pleiotropic effects of plcC gene mutation
Flagellin levels in the supernatant of the plcC-deficient strain MFN1037 were low. We therefore investigated the motility of the MFN1032 and the plcC mutants. In swimming conditions (i.e 0.3% agar, final concentration), MFN1036 completely invaded the plate over the course of a 16 h incubation at 28°C and displayed swarming motility, whereas MFN1032 displayed only swimming motility and diffused less ( Figure 5B and 5A, respectively). This may be due to higher levels of biosurfactant release or production, than observed for the wild-type strain, as suggested by the translucent appearance of the halo on the plates. This hypothesis was confirmed by the drop-col-Zymogram on egg-yolk agar of a silver-stained SDS-PAGE gel of MFN1032 supernatant Figure 3 Zymogram on egg-yolk agar of a silver-stained SDS-PAGE gel of MFN1032 supernatant. PLC activity was detected as an opaque band on the plate (arrow), the molecular mass of which was deduced. Concentrated MFN1032 supernatant was obtained following growth at 17°C in LB medium. 1: Front of a silver-stained gel placed on an egg-yolk agar plate (zymogram); 2: corresponding silver-stained gel; 3: other side of the zymogram; M: molecular size markers.

M 2 3
lapse test ( Table 5). The plcC-deficient mutant MFN1037 and the complemented strain MFN1038 (not shown) had swimming motility patterns similar to those of MFN1032 ( Figure 5C), but did not produce biosurfactant on LB agar plates, as shown by the drop-collapse test (Table 5). In motility assays on LB plates containing 0.6% agar, MFN1037 and MFN1038 had no swarming motility and no movement was observed even after 40 h of growth at 28°C, whereas MFN1032 continued to be motile ( Figure  5D and 5C, respectively).
The loss of biosurfactant production may be enough, in itself, to account for the loss of swarming motility of the plcC-deficient strain, but cannot account for the low levels of flagellin release. Electron microscopy analyses of MFN1032 and MFN1037 cells showed that the wild-type strain had a single polar flagellum whereas the plcC-deficient strain MFN1037 displayed a hyperflagellated phenotype, with most of the cells presenting three polar flagella ( Figure 6A and 6B, respectively). No difference in flagel-lum size was observed, but these differences in flagellation would probably affect the properties of the two strains. MFN1037 cells formed aggregates on the grid, whereas MFN1032 cells did not, despite having been prepared in the same conditions. Adhesion was evaluated by analyzing biofilm formation on polystyrene microarrays at 37°C. We found that MFN1037 had a greater capacity to form biofilms than the wild-type strain (255 ± 32% more adhesion than MFN1032 at 37°C; Figure 6C).

Analysis of the flanking regions of plcC
We investigated the genomic organization of the genes encoding PLCs in the genomes of P. fluorescens Pf0-1 and Pf5. We identified two genes of unknown function in the operon containing the plc gene in Pf0-1, and one in Pf5 ( Figure 7). We found no other gene in an operon with plcC in MFN1032, confirming that the pleiotropic effects of plcC disruption could not be due to a polar mutation. No gene encoding a potential flagellin or belonging to the flagellar regulon were found in the surrounding region. A gene encoding a putative dihydropycolinate synthase was found in the 5' flanking region of plcC, in an antisense orientation. In the 3' region, a similar phenomenon was observed, except that the gene encoded a putative transcriptional regulator of the GntR family [27]. This regulator is probably involved in plcC expression because the consensus GntR binding site (T.GT-N(0-3)-AC.A), as identified by Rigali et al. [27], is present in the sequence immediately upstream from the plcC gene.

Discussion
P. fluorescens is generally considered to be non pathogenic, but its infectious potential has nonetheless been demonstrated. In vitro studies of environmental strains, including P. fluorescens MF37, have shown that this psychrotrophic bacterium has most of the features of an opportunistic pathogen [8]. P. fluorescens is a highly heterogeneous species, ranging from avirulent strains that can be used in crop protection [28] to clinical strains involved in lung, urinary tract and blood infections. However, whereas the low virulence of industrial strains has been studied in detail, the factors involved in the virulence of clinical forms of P. fluorescens, including MFN1032, have not previously been investigated. Virulence results from the combined effects of direct contact between the bacterium and the target cell and the production of many soluble or SDS-PAGE of total extracellular proteins of MFN032 and the PlcC mutants MFN1037 and MFN1038 secreted molecules and exotoxins acting at some distance from the microorganism. MFN1032 has been shown to be highly cytotoxic in vitro to eukaryotic cells, to which it binds [7]. We investigated the role of factors secreted by this bacterium in cytotoxicity. We therefore carried out experiments with supernatants and culture medium extracts only.
In addition to its ability to multiply at 37°C, which is itself unusual for a strain of P. fluorescens, MFN1032 generated molecules with high levels of hemolytic and cytotoxic activity, as observed in in vitro models. The hemolytic activity of MFN1032 supernatants was similar to that of P. aeruginosa, which is an opportunistic pathogen. By contrast, a typical psychrotrophic strain, MF37, displayed no hemolytic activity under the same conditions. More surprisingly, MFN1032 supernatants were significantly more cytotoxic to rat glial cells than PAO1 supernatants, for bacteria grown at 17°C and 28°C. The secreted hemolytic and cytotoxic activities of P. fluorescens MFN1032 and P. aeruginosa PAO1 appeared to be temperature-dependent, as was P. fluorescens MF37 cytotoxicity. Surprisingly, the supernatant of MFN1032 grown at 37°C displayed no hemolysis. At first glance, this finding appears to contradict the hypothesis that hemolysis is a virulence factor. Low temperature induction likely reflects the evolutionary history of the factors involved in this activity and the function of those elements in non-infective conditions. In fact, it has been described that Pseudomonas aeruginosa from clinical and non-clinical environments are genetically and functionally similar [29,30]. So evolution of virulence determinants in opportunistic pathogens is not necessary linked to their interaction with the human host. Finally, invasin and the heatstable enterotoxin Yst from Y. enterocolitica are also virulence factors produced preferentially at temperatures below 37°C in vitro. However, in conditions of slight acidity or high osmolarity, these virulence genes are expressed at 37°C in vitro. High-temperature repression may therefore be overcome by other conditions stimulating expression in the host.  We found that neither protease nor LPS was responsible for the extracellular hemolytic activity of MFN1032. The absence of protease activity in these conditions was not surprising, as protease activity has been recovered from P. aeruginosa PAO1 only after at least 18 h of growth in LB medium and has never been observed with P. fluorescens MF0 in this medium [13]. Many pathogens display secreted hemolytic activity, which may be due to toxins, biosurfactants, and/or enzymes (essentially PLCs) [31]. Various species of Pseudomonas have been reported to produce hemolytic and/or non hemolytic PLCs [16,21,32]. The observed degradation of lecithin by MF37 on plates or in supernatants is consistent with this bacterium producing a non hemolytic PLC other than PlcC, as no gene amplification was observed with plcC primers. By contrast, the PLC produced by MFN1032 is involved in the hemolytic activity of this strain, as shown by the inhibitory effect of D609.
A protein with lecithinase activity was identified in MFN1032 supernatants by an egg-yolk agar plate zymogram method. This enzyme has a molecular mass of 42 kDa, as determined by SDS-PAGE. This apparent molecular mass is similar to that of previously described P. fluorescens PLC enzymes [15,16,18] and those of the putative PlcB  fied in various strains of P. fluorescens including, in particular, the PLC produced by a P. fluorescens isolated from placental extract [15] and the putative PLC from SBW25. This plcC gene was shorter than those predicted for Pf0-1 and Pf5, and the main difference between the protein encoded by this gene and Pf0-1 and Pf5 concerned the N-terminus of the protein. It is unclear whether plc genes are expressed in the other strains and it is possible that PLC production is not specific to adaptation to human infection but was acquired to survive in other environments.
No significant sequence similarity was found between the group of PLCs described here and the well characterized zinc-metallophospholipases C of Gram-positive bacteria. This group of PLCs is also only very distantly related to the enzymes described by Stonehouse et al., including the PlcH and PlcN of P. aeruginosa [26]. The differences between these enzymes concern not only their sequences, but also their catalytic sites, as D609 has no effect on PlcH activity [26]. We constructed a plcC-overexpressing mutant of MFN1032, MFN1036, to investigate the involvement of PlcC in hemolytic activity and we found that a higher level of PlcC production by MFN1032 was associated with higher levels of hemolytic activity. The loss of hemolytic activity in the plcC-deficient strain MFN1037 confirmed the involvement of this enzyme in hemolysis. Hemolytic activity was not restored in the MFN1038 strain, which overexpressed plcC, and extinction of the plcC gene had also pleiotropic effects, particularly as concerns the release of biosurfactant and flagellin. This was not due to a polar mutation involving plcC gene interruption, as the plcC gene was not associated with any other gene in an operon in the MFN1032 genome. Linares et al. described that the presence of low antibiotics concentrations in the culture media may induce changes in bacterial physiology (biofilm, motility and cytotoxicity) [42]. Their data could explain the phenotype change of the plcC mutant MFN1037, but we were not in conditions used by these authors (i.e subinhibitory concentrations of antibiotics). We probably disturb a complex regulatory network in MFN1037, and this hypothesis is also supported by plcC overexpression in MFN1036 resulting in the overproduction of biosurfactants, increasing swarming mobility.
Such complex regulatory systems often involve positive and/or negative feedback loops [43]. GntR regulators have been reported to maintain their own expression [44], so a positive feedback loop regulating GntR levels may exist. As previously reported, a simple change affecting a key element of this kind of system might lead to epigenetic modification [45]. Epigenetic switches, corresponding to phenotype modifications, arise and may be transmitted from a cell to its progeny in the absence of genetic modifications. This type of regulation has been reported for the cytotoxicity associated with the T3SS (type-three secretion system) of P. aeruginosa. Transient expression of the ExsA transcriptional regulator in non inducible strains leads to the acquisition of a cytotoxic phenotype [46]. The artificial extinction of plcC in MFN1037 may have triggered such an epigenetic switch.

Conclusion
These findings demonstrate that some P. fluorescens strains have some of the key characteristics of opportunistic pathogens. We provide the first demonstration, to our knowledge, of the involvement of a PLC in the secreted hemolytic activity of a clinical strain of P. fluorescens (MFN1032). We found that MFN1032 secretes a phospholipase C homologous to a PLC from an uncharacterized P. fluorescens strain previously studied biochemically by Preuss. This enzyme belongs to a new group unrelated to the PLCs produced by P. aeruginosa and seems to be produced by a wide range of P. fluorescens strains, although no homolog of the plcC gene was found in our model strain, MF37. Further studies including strains of different origins presenting hemolytic activity would clarify these observations. However, although PlcC is not specific to clinical isolates, this enzyme is a potential virulence factor as our data show that this enzyme is directly involved in the secreted hemolytic activity of MFN1032, as demonstrated by the inhibitory effect of D609. The direct involvement of PlcC in MFN1032 virulence could be further demonstrated or excluded by studying in vivo models.
Results obtained with the plcC mutants also suggest that this enzyme interferes with biosurfactant production, which might also account for the higher levels of hemolysis observed when plcC was overexpressed. The pleiotropic phenotype resulting from plcC mutation or overexpression suggests that PlcC is involved in a regulatory network. We are now investigating a possible role for the 3' flanking region of plcC, corresponding to the putative transcriptional regulator GntR, with the aim of determining the link between PlcC, GntR and biosurfactant production.

Bacterial strains and culture conditions
The MFN1032 strain was collected from a patient suffering from pulmonary tract infection (expectoration) at a hospital in Haute Normandie. This strain was the only bacterial contaminant in a normally sterile compartment and was considered to be the cause of the infection. MFN1032 was identified as a P. fluorescens biovar I strain [7]. PAO1 is a P. aeruginosa strain widely used in laboratory studies [47]. MF37 is a spontaneously rifampicin-resistant mutant of the MF0 strain, a psychrotrophic strain of P. fluorescens isolated from unpasteurized milk extensively studied in our laboratory [48]. These bacteria were cultured in Luria Bertani medium (LB), at various temperatures between 8 and 37°C, with shaking at 180 rpm. When necessary, 500 μg/mL mezlocillin or 40 μg/mL tetracycline was added. Bacterial density was determined by measuring optical density at 580 nm (Spectronic 601 spectrophotometer).

Glial cell cytotoxicity assays
Cytotoxicity was assessed by quantifying lactate dehydrogenase (LDH) release into the medium by cells, as this release reflects levels of necrosis. Concentrated supernatants (10 μL/mL culture medium) from bacterial cultures at various temperatures were incubated overnight with rat glial cells (8 × 10 6 cells/mL) that had been cultured in vitro for 12 days. Controls included LB concentrated with Ami-con Ultra-15 centrifugal filter units and incubated for the same period of time in culture medium for glial cells. The Cytotox 96 ® enzymatic assay (Promega, France) was used to quantify necrosis by measuring LDH release into the culture medium. The percentage total lysis was calculated as follows: % = [(X-B)/(T-B)] × 100, where B (baseline) is a control corresponding to LDH spontaneously produced by glial cells incubated with concentrated LB (10 μL/mL culture medium), T is a positive control (100%) corresponding to the amount of LDH detected in the culture medium after total lysis of the cell population by Triton X-100 (9% (v/v) in water) treatment and X is the amount of LDH detected in the culture medium of the sample tested.
The assay was sensitive enough to measure LDH concentrations equivalent to the lysis of 1% of the cell population. For hemolysis assays, we combined 600 μL of a 2% suspension of red blood cells (RBCs) with 600 μL of filtered supernatant and incubated this mixture for 2 h at 37°C. The suspension was centrifuged at 10,000 g for 8 min at 4°C, and hemoglobin release was assessed by determining absorbance at 540 nm. The percentage (%) of cells lysed was calculated as follows: % = [(X-B)/(T-B)] × 100. B (baseline) is a negative control corresponding to RBCs incubated with 600 μL sterile LB and T is a positive control corresponding to the total lysis obtained by incubating RBCs in LB supplemented with 0.1% SDS (final concentration). X is the absorbance value for the sample analyzed. All experiments were performed at least three times in triplicate. D609 was obtained from Sigma and used at a final concentration of 1 mM.

Protease and LPS assays
Protease assays were carried out with supernatants from bacteria cultured for 15 generations, as described by Hellio et al. [13]. LPS levels were quantified by determining 2keto-3-deoxyoctulonic acid (KDO) concentration, as described by Karkhanis et al. [50].

Biosurfactant production
Biosurfactant production was assessed by the drop-collapse test mainly as described previously [51]. Drops of Volvic water were dispensed into a Petri dish with a polystyrene platform. Drops of culture supernatant were added to the drops of water. If the culture broth contained biosurfactant, the droplets of water collapsed.

Gel electrophoresis conditions, zymogram methods and amino-acid sequencing
Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was carried out as described by Laemmli [52]. For a zymogram on egg-yolk agar, a silver-stained SDS-PAGE gel was washed three times in water and placed directly on an egg-yolk agar plate. After overnight incubation at 37°C, lecithinase activity was detected as an opaque band corresponding to lecithin hydrolysis. For Nterminal amino-acid sequencing, the supernatant proteins were transferred to PVDF membranes and subjected to Edman degradation in an Applied Biosytems 492 automated protein sequencer.

Motility assays
Motility assays were performed as described [53], with light modifications. Each strain was incubated on LB agar plates for 24 h at 28°C. Plates of LB medium solidified with 0.3% agar (for the assessment of swimming motility) were inoculated by stabbing colonies with a toothpick and inserting the end of the toothpick just below the surface of the agar. Three colonies were picked from three plates and incubated at 28°C until a migration halo appeared. We then spotted 5 μL of 3 independent suspensions of each strain onto LB medium plus 0.6% agar (swarming motility) and the plates were incubated until a migration halo appeared.

Sequence determination and analysis
The 1.2 kb or 1.5 kb PCR fragment was isolated with a DNA gel extraction kit (Genomics/Millipore) and cloned with the pMOSBlue Blunt-ended Cloning Kit (Amersham/Biosciences). MOS cells were transformed and, after blue/white colony screening, clones were picked and plasmid DNA was isolated with the QIAprep Spin Miniprep Kit (Qiagen). Plasmids were checked by digestion with HindIII/AvaI and sequenced by Genome Express (France). The predicted protein sequence of MFN1032 PlcC was for BLAST queries http://www.ncbi.nlm.nih.gov/BLAST/. . MFN1032 cells were conjugated with the S17.1 pMF1034 strain and selected for resistance to tetracycline and ampicillin. Clones were tested by PCR with PLC0 and PLC2 probes to confirm disruption of the plcC gene. MFN1037 (plcC-deficient MFN1032) had no1.5 kb fragment corresponding to the plcC gene under the control of its own promoter. This strain also had an attenuated egg-yolk degradation phenotype. It was cultured in LB with a final concentration of 40 μg/mL tetracycline. Complementation of the mutation was obtained by electroporating MFN1037 with pMF36 (strain MFN1038). Clones were selected for resistance to mezlocillin and tetracycline and egg-yolk degradation phenotype, and were checked by plasmid DNA isolation with the QIAprep Spin miniprep Kit (Qiagen), followed by enzymatic digestion.

Authors' contributions
GR wrote the manuscript and analyzed most of the data. AM participated in the molecular genetic studies, carried out the sequence comparison and participated in the design of the study and the manuscript. JG was involved in the swimming motility assays and flagellin identification. WV and OL carried out the glial cell cytotoxicity assays. MF helped to design and write the manuscript. NO was involved in designing the study. All authors read and approved the final manuscript.