Revisiting the role of phospholipases C in virulence and the lifecycle of Mycobacterium tuberculosis

Mycobacterium tuberculosis, the agent of human tuberculosis has developed different virulence mechanisms and virulence-associated tools during its evolution to survive and multiply inside the host. Based on previous reports and by analogy with other bacteria, phospholipases C (PLC) of M. tuberculosis were thought to be among these tools. To get deeper insights into the function of PLCs, we investigated their putative involvement in the intracellular lifestyle of M. tuberculosis, with emphasis on phagosomal rupture and virulence, thereby re-visiting a research theme of longstanding interest. Through the construction and use of an M. tuberculosis H37Rv PLC-null mutant (ΔPLC) and control strains, we found that PLCs of M. tuberculosis were not required for induction of phagosomal rupture and only showed marginal, if any, impact on virulence of M. tuberculosis in the cellular and mouse infection models used in this study. In contrast, we found that PLC-encoding genes were strongly upregulated under phosphate starvation and that PLC-proficient M. tuberculosis strains survived better than ΔPLC mutants under conditions where phosphatidylcholine served as sole phosphate source, opening new perspectives for studies on the role of PLCs in the lifecycle of M. tuberculosis.


Results
Genome analysis and deletion of the plcABC operon in an M. tuberculosis H37Rv genetic background. Analysis of M. tuberculosis genome data from public databases shows that most M. tuberculosis strains harbour four PLC encoding genes. These genes, named plcA, plcB, plcC and plcD are located at two different genomic loci in M. tuberculosis, with plcA-B-C organised as an operon (rv2351c-rv2350c-rv2349c) at genome coordinates 2632-2627 kb (reverse strand) of strain H37Rv, and plcD, represented as a single gene (rv1755c), located about 640 kb upstream of plcA-C 25 . It is also known that PLC encoding genes are preferred integration sites (or hotspots) for the IS6110 insertion element, which may lead to the presence of two insertion elements in close proximity, favouring homologous recombination between the adjacent IS6110 elements and deletion of the intervening sequences [26][27][28] . The widely used reference strain M. tuberculosis H37Rv shows such IS6110-mediated truncation of the plcD gene 26 . However, despite plcD inactivation, M. tuberculosis H37Rv retains a fully virulent phenotype in mice 29 . We thus chose the H37Rv strain to construct a null PLC mutant, taking in consideration that truncation of plcD facilitated the construction of the PLC complete knock-out strain, as only the PlcA-B-C operon had to be deleted.
The M. tuberculosis H37Rv PLC null mutant (H37RvΔ PLC) was constructed by using a previously described recombineering-based approach 30 . The different construction steps included the generation by 3-step-PCR 31 of a linear DNA fragment containing an apramycin resistance cassette embedded in the flanking regions of the plcABC cluster (Fig. 1A), which was genetically transformed into the H37Rv strain. Selection of an appropriate clone that showed replacement of the plcABC cluster with the apramycin cassette was assessed by PCR and then confirmed by Southern blotting analysis (Fig. 1B,C). In addition, a H37RvΔ PLC::plcABC complemented strain was obtained by integrating the plcABC gene cluster into the genome of H37RvΔ PLC using the plasmid pPlcABC. This pYUB412-based vector 32 contains the plcABC operon expressed under the control of its natural promoter.
As controls for selected experiments, we also included the previously described Myc2509Δ PLC mutant strain 15 , here referred to as MT103Δ PLC, and the isogenic MT103 parental M. tuberculosis strain.
Evaluation of phospholipase C activity in mutant and WT M. tuberculosis strains. In a first step, we used a spectrophotometric assay to determine the PLC activity of whole-cell extracts from WT M. tuberculosis strains and the two PLC-deletion mutants. This assay is based on the detection of the hydrolysis of colourless p-nitrophenylphosphorylcholine (p-NPPC) to p-nitrophenol, which absorbs light at 410 nm and is yellow. As shown in Fig. 2 pPlcABC restored PLC activity to the level of the H37Rv WT M. tuberculosis strain. As expected, the plcABC-unrelated M. tuberculosis H37Rv mutant ∆ESX1, which is lacking a functional ESX-1 secretion system due to the deletion of the region of difference RD1 23,24,33 , showed a phospholipase C activity very similar to the H37Rv WT strain ( Fig. 2A,B). Taken together, these results confirmed the loss of PCL activity in the H37RvΔ PLC mutant.
Phospholipase C is not involved in M. tuberculosis-induced phagosomal rupture. According to host-pathogen interaction data reported from a range of bacterial pathogens, phospholipase C activity is often required for egress of bacteria from phagosomal containment and cytosolic access [20][21][22] . We thus investigated the ability of PLC mutants H37RvΔ PLC and MT103Δ PLC and WT strains to access the cytosol during infection of THP-1 human macrophage-like cells, by using a recent flow-cytometric phagosomal rupture screening method 23 . Briefly, this sensitive assay relies on the change in the emission spectrum of the cephalosporin-like FRET substrate CCF-4 upon cleavage by β -lactamases 34,35 , which serves as a readout for the detection of contact between β -lactamase-producing M. tuberculosis and the FRET substrate in different environments, including the host cytosol. As CCF-4 cannot enter an intact vacuole, the assay assesses whether cytosolic contact of M. tuberculosis occurs in the host cell during infection. Differentiated THP-1 cells were infected with the various M. tuberculosis strains at an MOI of 1:2, and the CCF-4 emission spectrum of cells was monitored over a three-day period. Both the WT and the Δ PLC-deletion M. tuberculosis strains were able to induce a switch in the emission spectrum from ~535 nm to ~450 nm, indicating that they were gaining access to the cytosol of the infected THP-1 cells (Fig. 3). In contrast, the attenuated Δ ESX-1 (Δ RD1) M. tuberculosis strain, which is impaired in inducing phagosomal rupture in host cells 23,24 and was included in the analysis as a negative control, was  Fig. 4, the H37RvΔ PLC mutant and the corresponding WT strain showed similar intracellular growth kinetics, resulting in a 1.5-Log increase in CFU number over a 7-day period. Consistent with previous observations from Raynaud and colleagues 15 , no differences were observed in the intracellular growth abilities of MT103Δ PLC and its isogenic parental strain. In contrast, the Δ ESX-1 (Δ RD1) M. tuberculosis strain, showed attenuated growth relative to WT and Δ PLC strains (Fig. 4). These results indicate that PLCs, in contrast to the ESX-1 proteins, are not essential for M. tuberculosis intracellular survival and optimal growth in host macrophages.

Virulence of M. tuberculosis in mouse infection models.
To further test whether PLC inactivation might result in a potential defect in in vivo growth ability that might not be detectable in macrophages cell lines, we evaluated the virulence properties of the H37RvΔ PLC and WT strains in different mouse infection models.
Given the previously established suitability of the SCID (severe combined immune deficient) mouse infection model for distinguishing attenuated Δ ESX-1 (Δ RD1) and virulent WT M. tuberculosis strains 33,37,38 , the potential impact of PLC-inactivation on virulence of M. tuberculosis was first assessed by testing the in vivo growth characteristics of Δ PLC and WT M. tuberculosis strains in SCID mice. Both the H37Rv Δ PLC mutant and the WT strain displayed an indistinguishable, high bacterial load in lungs and spleen of infected mice (Fig. 5A,B). This was also confirmed by visual inspection of the organs, which showed typical signs of massive infection (Supplementary Figure 2). Similarly, the MT103Δ PLC  and WT strains both showed comparable, intense in vivo growth in SCID mice, as indicated by the presence of ~10 8 CFU in the organs after 3 weeks of infection, although it should be mentioned that for this latter strain couple, the CFU numbers determined for day 1 were somewhat higher for the Δ PLC mutant in comparison with the WT strain (Fig. 5A,B).
To determine whether the findings obtained in the SCID mouse model were also relevant in immunocompetent mice, virulence studies with the H37Rv and MT103 Δ PLC and WT strain-pairs were also performed in an aerosol infection model of C57BL/6 mice, where the bacterial load in target organs was determined after 6 weeks of infection. As shown in Fig. 5C and Supplementary Figure 3, we did not observe a significant difference between WT and Δ PLC mutants in their in vivo growth properties. These findings, which are in overall agreement with results from the phagosomal rupture screen and the THP-1 infection assay, suggest that PLCs from M. tuberculosis might not represent very obvious virulence factors of M. tuberculosis.
Expression of plcABC genes seems to linked to phosphate concentration. Previous studies on PLCs of P. aeruginosa have shown that induction of PLC expression was phosphate regulated, suggesting  Figure 4). For monitoring promoter activities, a recombinant Δ PLC M. tuberculosis H37Rv strain expressing a translational 5′-plcA-egfp fusion under the natural plcABC promoter was constructed and named H37RvΔ PLC::plcA-egfp. Results obtained from growth experiments with this strain showed that during the first 9 days fluorescence remained low, while starting from day 10 post-inoculation a strong increase in fluorescence was noted (Fig. 6A). By this time-point the phosphate concentration in the medium was below 0.3 mmol.l -1 . The expression of the plcABC genes thus seems to be induced by low phosphate concentration, although an impact of other potential stress factors linked to the consumption and limitation of essential nutriments may not be excluded. To further investigate this point, an H37RvΔ PLC::plcA-egfp strain that also expressed DsRed under a constitutive promoter was constructed. With the help of this strain promoter activity was studied at different phosphate ion concentrations, simultaneously controlling for the impact of cell density on fluorescence levels. Monitoring of green fluorescence relative to red fluorescence and absorbance levels showed that under low phosphate conditions green fluorescence increased strongly relative to the constant level of red fluorescence, confirming that the plcABC promoter was more strongly induced under low phosphate conditions (Fig. 6B). Finally, we also evaluated the GFP-fluorescence normalized to the cell density measured in OD, and again observed that at low phosphate concentration the promoter activity of the plcABC operon was increased (Fig. 6C). Starvation of phosphate ions thus seems to represent a stress that the bacteria try to counterbalance by induction of PLC production.
Finally, we also conducted experiments wherein the WT, Δ PLC and Δ PLC::plcABC H37Rv M. tuberculosis strains were grown in liquid medium supplemented with phosphatidylcholine as the sole phosphate source. As shown in Fig. 6D, the WT M. tuberculosis H37Rv strain and the complemented strain survived better under these conditions compared to the Δ PLC mutant. It should be emphasized, however, that none of the strains was able to actively grow under these experimental settings.

Discussion
PLCs are widely distributed enzymes in living organisms. PLCs hydrolyze phospholipids such as phosphatidylcholine or sphingomyelin at the phosphodiester bond. In bacteria, these enzymes have been reported to function in a wide variety of cellular tasks during infection, including membrane lysis, intracellular signalling, lipid metabolism and/or pathogenicity-associated activity 40,41 . In our study, we focused on the PLCs of M. tuberculosis, which belong to the superfamily of haemolytic phosphocholine-specific PLCs for which PLC of P. aeruginosa is the paradigm member 42 . Our initial objective was to evaluate whether these enzymes were involved in the process of phagosomal rupture induced by M. tuberculosis during infection of macrophages. In L. monocytogenes, or C. perfringens, PLCs play important roles together with pore forming listeriolysin or perfringolysin, respectively, to lyse the phagosomal membrane and allow the bacteria to gain access to the cytosol and promote cell-to-cell spread 22,43 . Concerning the infection with M. tuberculosis, the scenario seems more complex. While it was long thought that M. tuberculosis resists degradation in the phagosome by inhibiting the fusion with lysosomes, favoring intra-phagosomal survival and multiplication 44 , more recent studies by van der Wel and colleagues, using cryo-electron microscopy, provided evidence of cytosolic presence of M. tuberculosis at later stages of infection 45 . Similarly, cytosolic access of virulent M. tuberculosis strains was recently also reported by using a FRET-based read-out, combined with automated confocal microscopy 24 or flow cytometry 23 . We here used the latter method to test the M. tuberculosis Δ PLC and WT strains for their ability to cause phagosomal rupture in comparison with a Δ ESX-1 (Δ RD1) negative control and found that the M. tuberculosis Δ PLC mutants and WT strains all showed very similar abilities to gain cytosolic access.
Given the result that PLCs of M. tuberculosis were not required for inducing phagosomal rupture and cytosolic contact of M. tuberculosis, which are attributes usually linked to mycobacterial pathogenicity 36 , we subjected the Δ PLC mutants and WT strains to virulence analyses in in vitro/ex vivo and in vivo models. In the obtained data, we could only detect minor, not significant virulence differences between the Δ PLC and WT M. tuberculosis strains of two genetic backgrounds, i.e. MT103 and H37Rv, in the 3 models used. These results, which were different from those of previous work reporting that PLCs were involved in virulence of M. tuberculosis 15 , remained puzzling.
Review of the available literature suggests that the number of functional PLC-encoding genes in different strains of the M. tuberculosis complex is highly variable and ranges from 0 to 4 copies. In many M. tuberculosis strains, including H37Rv, the plcD gene, which represents a hotspot for IS6110 insertion, is inactivated or deleted 27,28,46 . Similarly, extensive IS6110 insertion is also observed for the plcABC locus, but to a lesser extent 47 . Moreover, in a study on genetic polymorphisms affecting the four PLC encoding genes in M. tuberculosis isolates, Viana-Niero and coworkers found that 19 of 25 clinical isolates showed loss of parts of genes or complete genes from the plcABC and/or plcD loci, whereby five isolates retrieved from patients with active tuberculosis had all 4 plc genes interrupted 48 . PLC-encoding loci are also variable in different lineages of the M. tuberculosis complex; PlcA/B/C, which are also known as the "mtp40" mycobacterial protein(s), are missing from the M. bovis lineage due to the RD5 deletion, and also are absent from certain other tubercle bacilli 27,49 . In this respect it is also noteworthy that M. bovis strains with an IS6110 insertion in the remaining plcD gene were described. Interestingly, these strains without a functional PLC encoding gene were responsible for causing tuberculosis lesions in cattle for which no differences in the organ distribution relative to other M. bovis strains were noticed 50 . These findings are also in agreement with results from a high-density transposon screen, wherein PLC-encoding genes have not been identified as essential for in vivo growth of M. tuberculosis in the mouse model 51 . Taken together, these reports and our experimental findings with two different Δ PLC mutants of M. tuberculosis cast doubt on an essential role of PLC in virulence of tubercle bacilli. PLCs of M. tuberculosis might play a less important role in the infectious lifecycle of M. tuberculosis than previously thought.
However, it is intriguing that despite the apparently marginal role of PLCs in virulence of M. tuberculosis, most strains have conserved one or more copies of PLC-encoding genes in their genomes, similar to certain non-tuberculous (NTM) mycobacteria. There are only few mycobacterial species that harbour genes encoding PLCs. Database analyses show that for the group of slow-growing mycobacteria, PLC-encoding genes are present in the genomes of smooth tubercle bacilli 52 , members of the M. tuberculosis complex, members of the Mycobacterium kansasii-Mycobacterium gastri cluster, Mycobacterium asiaticum, and members of the Mycobacterium marinum-Mycobacterium ulcerans cluster. PLCs are absent from the genomes of Mycobacterium leprae, and members of the Mycobacterium avium-intercellulare complex. In the more distantly related rapid growing mycobacteria, only M. abscessus is known to carry a PLC, which shows 37% amino-acid identity with PLCs from M. tuberculosis and seems to be the result of a specific horizontal gene transfer (HGT) into M. abscessus 13,53 . In contrast, the PLCs in M. tuberculosis and other slow-growing mycobacteria seem to share a common origin with PLCs from different Gordonia species, with which they show about 60% amino acid identity. It seems thus likely that a common progenitor of the phylogenetic subgroup of slow-growing mycobacteria comprising M. kansasii, M. gastri, M. asiaticum, M. marinum and M. tuberculosis has acquired the PLC-encoding genes during evolution through HGT from more distantly related actinobacteria. This feature prompted us to search for potential alternative biological functions of PLCs in slow growing mycobacteria, not necessarily linked with virulence. As one hypothesis, PLCs of M. tuberculosis might help in the acquisition of phosphate. Cleavage of phospholipids by PLCs results in the generation of two molecular entities, a glycerol part and a residue containing a phosphate group, which might serve as a potential source of phosphate for the bacterium. The finding that expression of the plcA-egfp fusion was inversely correlated with the phosphate concentration in the medium suggests that the specific promoter activity might be downregulated under phosphate-sufficient environmental conditions. This assumption is in agreement with previous observations with PLCs from P. aeruginosa, for which an impact of phosphate concentration on plc gene regulation was noted 20,39 . Moreover, it has previously been reported that during infection of THP-1 cells by M. tuberculosis, the expression of the plcABC operon was upregulated for the first 24 h of infection 15 . Given the results obtained with our GFP-fusion assay, it is thus tempting to speculate that this upregulation might be related to a limited phosphate concentration inside the phagosome. Limitation of phosphate during phagosomal containment was also postulated by results from large-scale transcriptome studies, which found genes encoding phosphate transporters upregulated during infection 18,54 . It is plausible that M. tuberculosis can vary its supply in phosphate between inorganic phosphate, which is the preferred source of phosphorus for many bacteria 18 , and acquisition of organic phosphates through the action of phosphatases and/or phospolipases. A similar scenario was recently suggested for SpmT (Rv0888) of M. tuberculosis, which harbours a surface-exposed C-terminal sphingomyelinase domain and a putative N-terminal channel domain that mediates glucose and phosphocholine uptake across the outer membrane 55 . However, at present it remains unknown if the PLCs of M. tuberculosis may contribute to the phosphate supply of the bacterium in a similar way. Our results point to such a possibility, although more in depth studies are needed to clarify this question.
In conclusion, our study calls into question the impact of PLCs on virulence of M. tuberculosis, and provides new hints on putative alternative functions of PLCs in M. tuberculosis.

Methods
Bacterial strains and culture conditions. Escherichia coli DH10B and Top10 (Invitrogen) strains, used for cloning procedures, were grown on LB agar medium and/or LB broth. M. smegmatis mc 2 155 and M. tuberculosis strains were obtained from stock held at the Institut Pasteur. The M. tuberculosis MT103 strain and the corresponding Myc2509Δ PLC mutant strain 15 were a gift of Prof. Gicquel, Institut Pasteur.

Construction of a plcABC deletion mutant in M. tuberculosis H37Rv. The M. tuberculosis H37Rv
ΔplcABC mutant was constructed by allelic replacement using the recombineering method 30 . The allelic exchange substrate plcABC::Apra was obtained by a three step PCR approach 56 . Briefly, two 500-bp fragments corresponding to the plcABC upstream and downstream regions were amplified by PCR from the M. tuberculosis H37Rv genomic DNA and linked to a third PCR fragment encoding the apramycin resistance cassette, to generate the 2 kb-fragment plcABC::Apra. The plcABC::Apra fragment was thus used to transform a M. tuberculosis H37Rv recombinant strain containing the pJV53 vector. The pJV53 plasmid encodes the recombination proteins gp60 and gp61 57 , whose expression is induced by incubation with 30 0.2% acetamide for 24 h. The H37Rv-pJV53 acetamide-activated transformants were selected on solid medium for resistance to Kanamycin and Apramycin. The obtained Kanamycin and Apramycin resistant clones were thus tested for the plcABC deletion by PCR. One out of 116 tested clones revealed an amplification profile consistent with the replacement of the plcABC cluster with the apramycin cassette, and was thus subjected to Southern blot analyses. Genomic DNAs from M. tuberculosis strains were digested with AvrII, separated by gel electrophoresis and transferred onto Hybond-C-Extra nitrocellulose (GE). Hybridization was performed with [α -32 P] dCTP-labeled PCR-probe, specific for the plcABC downstream region, in 6x SSC, 0.5% SDS, 0.01 M EDTA, 5x Denhardt's solution, 100 μg.ml −1 salmon-sperm DNA, at 68 °C. After washing, membranes were exposed to phosphorimager screens, which were scanned in a STORM phosphorimager 31,57 . Construction of M. tuberculosis H37RvΔPLC complemented strains. Two different integrative pYUB412-based plasmids (pPlcABC and pYUB412-Pr_plcA-egfp) harbouring the plcABC operon and the plcA gene, respectively, were constructed. To obtain the pPlcABC plasmid, the plcABC operon and its natural promoter region, were amplified by PCR and cloned into the pYUB412 vector backbone. Similarly, to construct the pYUB412-Pr_plcA-egfp plasmid, the plcA gene and the plcABC promoter region were amplified by PCR using modified primers (Supplementary Table 1), which allow the introduction of additional HindIII and NheI recognition sequences in the amplified fragment obtained. The resulting PCR product was digested and ligated into the HindIII-NheI-digested pYUB412::egfp (a pYUB412 derivative cosmid that allows the expression of transcriptional eGFP fusion protein constructs 32 ).
Both pPlcABC and pYUB412-Pr_plcA-egfp constructs were used to transform the M. tuberculosis H37RvΔ PLC mutant strain. Transformed clones were selected on solid medium for resistance to hygromycin.
PCR amplification and DNA Sequencing. PCR reactions to obtain fragments used in cloning procedures or in screening of transformed clones were carried out with Pwo (Roche) or similar high fidelity DNA polymerases, respectively, as previously reported 58 . Sequences of primers used in amplification reactions are listed in Table S1. All amplified PCR products and plasmids were sequenced by using the Big Dye cycle sequencing Kit (Applied Biosystems) in an automated DNA sequencer (Applied Biosystems, 3130xl genetic analyser). Phospholipase C assay. PLC activities were measured using the p-nitrophenylphosphorylcholine (p-NPPC) substrate (Sigma-Aldrich). Activity is defined as the ability of an enzyme to catalyse p-NPPC into p-nitrophenyl (p-NP) of yellow chromogenic nature. Briefly, 0.5 mg of total protein was incubated in 3 ml of 10 mM Tris HCl (pH = 7.2), containing 5 mM of p-NPPC and 1.5% of sorbitol 16 . The reaction mix was incubated at 37 °C under shaking at 100 rpm. The reaction was stopped after 0, 1, 2, 3, 4, or 7 days by NaOH at 0.1 N (final concentration). The release of p-NP was measured at 410 nm. Buffer without proteins served as blank reference. For initial experimental setup, a control assay was performed with 40 U of purified PLC from B. cereus (Invitrogen).
Phosphate assay by colorimetric method. Phosphate ions in presence of L(+ ) ascorbic acid (Merck) and ammonium molybdate tetra-hydrate (Sigma) form a complex showing a blue/green colour. Briefly, comparator samples containing between a standard range of 0 and 6.10 −5 mol. L −1 of phosphate (KH 2 PO 4 ) were prepared in 15 mL glass tubes. After adding 1 mL of stock solution of ascorbic acid at 0.1 mol.L −1 and ammonium molybdate tetra-hydrate at 0.2 mol.L −1 , the react volume was adjusted to 10 ml final volume with Milli-Q (Millipore) purified water. All tubes were incubated at 80 °C in a water bath during 10 minutes and slowly cooled down to room temperature on the bench. Then all samples and comparator samples were diluted (d = 1/2) before measurement at 750 nm.
Infection of THP-1 derived macrophages with M. tuberculosis. THP-1 cells were grown at 37 °C with 5% CO 2 . Cells were maintained in RPMI 1640 + glutamax (Life technologies) and 10% of foetal bovine serum. THP-1 cells were seeded in 96 well plates at 7.5 × 10 4 cells per well, and differentiated by incubation with 10 ng.ml − 1 of PMA for 2 days. Before infection, the medium was removed and the wells were washed 3 times with PBS. Bacterial strains were added at a multiplicity of infection (MOI) = 1: 20 (1 bacterium: 20 macrophages). After 2 hours (day 0) or 3, 5 and 7 days post-infection, cells were lysed in PBS 0,01% of Triton X-100. The number of viable intracellular mycobacteria was determined/counted by plating serial dilutions of macrophage lysates on solid medium.

M. tuberculosis virulence studies in mice.
Six-week-old female CB17/Ico SCID mice (Charles River) were infected intravenously via the lateral tail vein with 200 μ l of bacterial suspension of 1 × 10 6 CFU.ml −1 . For aerosol infection, a customized apparatus was used following a previously established procedure 59 Six-week-old female C57BL/6B mice (Charles-River) were infected with a suspension containing 5 × 10 5 bacteria.ml −1 to obtain an inhaled dose of ca. 100 CFU. At selected time points after infection, mice were killed and organs homogenised using a tissue Lyser apparatus from Quiagen and 2.5 mm diameter glass beads to determine CFU numbers as previously reported 52,59 .
All animal studies were approved by the Institut Pasteur Safety Committee (Protocol 11.245; experimentation authorization number 75-1469), in accordance with European and French guidelines plcABC-promoter induction assay. The H37RvΔ PLC::Pr_plcA-egfp strain was complemented with a DsRed expressing plasmid. The resulting strain was named H37RvΔ PLC::Pr_plcA-egfp::Pr-hsp60-DsRed. In this construct DsRed is expressed via a constitutive promoter while GFP expression is dependent on plcA promoter activity. Briefly, for these experiments bacteria were grown in 7H9 + ADC + Hygromycin 50 μ g.ml −1 /Zeocin 25 μ g.ml −1 medium until 0.4-0.6 OD. Bacteria of this preculture were inoculated in fresh Sauton medium containing 0.05% Tween80 and Hygromycin 50 μ g.ml −1 /Zeocin 25 μ g.ml −1 at 0.05 OD. After 7 days of culture, bacteria were harvested and centrifuged during 5 min at 5000 g. After 3 washing steps, using 5 ml of fresh Sauton medium lacking PO 4 3-, bacteria were inoculated at 0.05 final OD in Sauton medium containing phosphate or not. In addition, the bacterial quantities were monitored by plating aliquots of each strain onto agar plates (in triplicate) for CFU determination. The GFP (Ex = 475 nm/Em = 504 nm) and DsRed fluorescence (Ex = 558/Em = 583) was monitored using a microplate reader (BGM Labtech) and analysed by Omega software. To avoid an increase of fluorescence due to the growth, we chose to normalize fluorescence values relative to absorbance. These ratio measurements were converted in percentage, with the day 0, as reference point.
Mycobacterial phosphatidylcholine survival assay. M. tuberculosis H37Rv WT, ΔplcABC and complement strains were inoculated into phosphate-free Sauton medium supplemented with phosphatidylcholine (3.6 mM) (Sigma) as the sole phosphate source, considering that 1 mole of phosphate was equal to 1 mole of phosphatidylcholine. Addition of phosphatidylcholine rendered the medium turbid, which precluded the use of OD measurement. CFU counting was used as an alternative for quantification of bacteria, as previously reported 60 .
Statistical analyses. Potential statistical differences in bacterial loads were evaluated by ANOVA test with Tukey correction, after conversion of CFU numbers in Log10 CFU values. Statistical significance was considered to be a P value ≤ 0.05.