LipG a bifunctional phospholipase/thioesterase involved in mycobacterial envelope remodeling

Tuberculosis caused by Mycobacterium tuberculosis is currently one of the leading causes of death from an infectious agent. The main difficulties encountered in eradicating this bacteria are mainly related to (i) a very complex lipid composition of the bacillus cell wall, (ii) its ability to hide from the immune system inside the granulomas, and (iii) the increasing number of resistant strains. In this context, we were interested in the Rv0646c (lipGMTB) gene located upstream to the mmaA cluster which is described as being crucial for the production of cell wall components and required for the bacilli adaptation and survival in mouse macrophages. Using biochemical experiments combined with the construction of deletion and overexpression mutant strains in Mycobacterium smegmatis, we found that LipGMTB is a cytoplasmic membrane-associated enzyme that displays both phospholipase and thioesterase activities. Overproduction of LipGMTB decreases the glycopeptidolipids (GPL) level concomitantly to an increase in phosphatidylinositol (PI) which is the precursor of the PI mannoside (PIM), an essential lipid component of the bacterial cell wall. Conversely, deletion of the lipGMS gene in M. smegmatis leads to an overproduction of GPL, and subsequently decreases the strain susceptibility to various antibiotics. All these findings demonstrate that LipG is involved in cell envelope biosynthesis/remodeling, and consequently this enzyme may thus play an important role in mycobacterial physiology.


Introduction
With more than 10 million new cases and almost 1.6 million deaths in 2017, tuberculosis (TB) caused by the etiologic agent Mycobacterium tuberculosis (M. tuberculosis) still remains the deadliest infectious disease worldwide [1]. The current antibiotherapy, which is a combination of four antibiotics (i.e. isoniazid (INH), rifampicin (RIF), pyrazinamide, and ethambutol), allows to heal 95% of infected people with active disease when administered under directly observed therapy [2]. This treatment is, however, not effective enough to overcome the emergence of multidrug and extensively drug-resistant strains of M. tuberculosis [3,4]. More recently, some totally drug-resistant M. tuberculosis strains have also been detected, which are virtually incurable [5]. Most of the time this antibiotic resistance is directly related to the unique mycobacterial cell wall composition which is an essential factor contributing to virulence, pathogenesis, and survival of the bacilli. Therefore, identifying proteins and mechanisms involved in the biosynthesis and/or remodeling of this complex structure could lead to the development of new therapeutics to control and fight TB.
Indeed, amongst all bacterial phyla, Actinobacteria and more specifically M. tuberculosis, possess an unusual cell envelope with a specific architecture constituted by lipids which may represent up to 40-60% of the dry weight of the cell, as compared with only 5-10% in other bacteria [6,7]. The outer membrane is an extremely hydrophobic barrier mainly composed of long-chain mycolic acids (MAs) and several (LipG MTB ) was also generated, by using the I-TASSER server [35,36] and visualized with PyMOL software (The Py-MOL Molecular Graphics System, version 1.8 Schrödinger, LLC). Electrostatic surface potentials of LipG MTB model structure was built using freely available PDB2PQR web service [37,38]. The electrostatic potential data thus generated, were further visualized using the PyMOL APBS Tools 2 plugin [39].

Construction of recombinant plasmids, expression and purification of recombinant LipG MTB protein
Rv0646c ORF encoding LipG MTB protein, was amplified from M. tuberculosis H 37 Rv genomic DNA by polymerization chain reaction, using primers containing the attB1, Shine-Dalgarno, 6× His, and a TEV (tobacco etch virus) NIa site sequences at the 5 -end and the attB2 recombination site at the 3 -end (Supplementary Table S1, P1 MTB and P2 MTB ). The resulting PCR fragments were cloned using the Gateway R technology (Invitrogen, Carlsbad, CA, U.S.A.) into the pDonR221 entry vector, and then into the pDEST14 expression vector (pDEST14::lipG MTB ). A similar strategy was used to clone the MSMEG 1352 ORF encoding LipG MS protein. PCR was performed with P1 MS and P2 MS primers (Supplementary Table S1), thus giving the pDEST14::lipG MS construct. DNA sequences of recombinant clones were analyzed by GATC Biotech (Ebersberg, Germany). The newly generated pDEST14::lipG MTB was used as a template for generating pDEST14::lipG MTB S123A by PCR mutagenesis using specific primers (Supplementary Table S1, P3 and P4). The resulting plasmids pDEST14::lipG MTB , pDEST14::lipG MTB S123A , and pDEST14::lipG MS were used to transform E. coli C41 (DE3)/pLyS strain harboring the pRare plasmid (Novagen).
For protein expression, 1 L of Terrific Broth medium was inoculated with the recombinant strains at initial OD 600 of 0.05, and was further incubated at 37 • C and 200 rpm. When the OD 600 reached 0.6, cells were put on ice for 30 min. Then, 0.5 mM isopropyl-d-thiogalactopyranoside was added to the medium and the culture was further incubated at 17 • C under shaking. After 18 h of incubation, cells were pelleted, re-suspended in 30 ml of ice-cold lysis buffer (50 mM Tris/HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 0.1% (v/v) Triton X-100, 0.25 mg/ml lysozyme), and stored at −80 • C. To purify recombinant LipG MTB or LipG MS , the supernatant (S1) obtained after cell lysis and centrifugation at 17000×g for 30 min, was stored at 4 • C. Pellet was then re-suspended in lysis buffer, sonicated twice for 30 s with 30 s breaks between each cycle, solubilized by stirring at 4 • C for 1 h using a magnetic bar, and further centrifuged at 17000×g for 30 min (giving S2 supernatant). Both supernatants (S1 and S2) were pooled and loaded on to a Ni 2+ -NTA affinity column previously equilibrated with buffer A (10 mM Tris/HCl pH 8.0, 150 mM NaCl). The column was washed with buffer A before performing elution with buffer A and 125 and 250 mM of imidazole. Fraction purity was assessed on SDS/PAGE. The corresponding 35 kDa protein band was further analyzed by tryptic digestion and MS. Pure protein was concentrated to 0.5-1 mg/ml, flash frozen in liquid nitrogen, and stored at −80 • C.

Thioesterase activity assay
Hydrolysis of palmitoyl-Coenzyme A (palmitoyl-CoA) (Sigma-Aldrich) was used to measure thioesterase activity. After the hydrolysis of the thioester bond, the free sulphur on CoA is attacked by DTNB (5-5 -Dithio-bis 2-nitrobenzoic acid) (Sigma-Aldrich) which releases a measurable nitrophenyl group, TNB 2− (5-Thio-2-nitrobenzoate) ion detectable at 415 nm. Fresh stock solution of palmitoyl-CoA (1 mM) and DTNB (0.4 mM) were prepared in water. Palmitoyl-CoA and DTNB were added in each well of a 96-well plate at final concentration of 10 μM. The thioesterase activity of rLipG MTB or rLipG MS (30 μg) was compared with that of TesA (Rv2928) [42] and LipY [43] used as positive and negative controls, respectively. Triplicate assays were done at 37 • C in 10 mM Tris (pH 8.0) 150 mM NaCl buffer. The absorbance at 415 nm was continuously measured and the release of TNB 2− was quantitated using a calibration curve (0.5-30 nmoles).
Inhibition experiments were performed using the classical enzyme-inhibitor pre-incubation method, as described previously [30,47,48]. Stock solutions (10 mM) of each CyC and OX inhibitor, as well as Orlistat were prepared in DMSO. rLipG MTB was pre-incubated for 30 min at 25 • C with each compound at various inhibitor molar excess (x I ), related to 1 mol of enzyme, ranging from 1 to 200. Sample of the incubation medium (1.0 μg rLipG MTB ) was then collected for measuring the enzyme residual activity using the colorimetric assay with the substrate pNP-C6, as described above. The variation in the residual activity allowed to determine the inhibitor molar excess which reduced the enzyme activity to 50% of its initial value (x I50 ) [47,48]. Thereby, a x I50 value of 0.5 is corresponding to a 1:1 stoichiometric ratio between the enzyme and the inhibitor, and is therefore the highest level of inhibitory activity that can be achieved. In each case, control experiments were done in the absence of inhibitor but with the same volume of solvent. It is worth noting that DMSO at a final volume concentration of less than 10% has no effect on the enzyme activity. Inhibition dose-response curves were fitted in Kaleidagraph 4.2 Software (Synergy Software). Results are expressed as mean values + − S.D. of at least three independent assays.

Creation of M. smegmatis MSMEG 1352 disrupted strain
The MSMEG 1352 gene encoding LipG MS was amplified with 500-bp flanking regions by PCR using M. smegmatis mc 2 155 genomic DNA as template and specific primers (Supplementary Table S1, P9 and P10). PCR product of ∼2000 bp was cloned into pJET1.2/blunt vector following manufacturer's instructions giving pJET-lipG MS . Recombinant plasmid was then digested with NruI, and a hygromycin resistance cassette amplified from pUC-Hygro using primer (Supplementary Table S1, P17 and P18). The resulting blunt-ended PCR product was cloned in order to disrupt lipG MS gene. This new plasmid, pJET-lipG MS ::Hyg which contains an inactive version of lipG MS gene was used as template for amplifying an allelic exchange substrate by using specific primers (Supplementary Table S1, P9 and P10). Approximately 300 ng of purified PCR product was electroporated into electro-competent M. smegmatis mc 2 155 strain harboring pJV53 plasmid, and cells were plated on to Middlebrook 7H10 agar medium supplemented with 50 μg/ml of both kanamycin and hygromycin and incubated at 37 • C during 5 days as previously described [49]. Finally, recombinant clones were analyzed by PCR using P7-P8 and P11-P12 primers' pairs (Supplementary Table  S1).

Creation of overexpression/complementation vectors
ORFs Rv0646c and MSMEG 1352 were amplified by PCR with specific primers (Supplementary Table S1, P5-P6 and P7-P8) containing NdeI and BamHI restriction sites in 5 and 3 ends, respectively. Digested and purified products were cloned into the pVV16 mycobacterial expression vector in frame with the 6 His-tag located in C-terminal position giving the pVV16::lipG MTB and pVV16::lipG MS complementation plasmids allowing a constitutive production of recombinant LipG. Rv0646c ORFs was also cloned within a homemade pMyC Gateway R vector, allowing a strong overexpression following acetamide induction, thus giving pMyC::lipG MTB . Competent M. smegmatis mc 2 155, M. smegmatis groEL1ΔC, and lipG MS ::Hyg cells were prepared as previously described [50] and were electroporated with a single pulse of 2.5 kV, 25 μF, and 1000 by using a Gene Pulser Xcell TM Electroporation System (Bio-Rad). Selection was achieved by plating cells on complete Middlebrook agar medium supplemented with the proper antibiotic.

Subcellular fractionation and immunoblotting
Recombinant M. smegmatis groEL1ΔC harboring pVV16::lipG MTB and pVV16::lipG MS were cultured in 7H9-S medium supplemented with 50 μg/ml kanamycin until OD 600 was approximately 1-1.5. Cells were harvested for 10 min at 4000×g, then washed twice with PBS (pH 7.4) buffer containing 0.05% (v/v) Tween-20, and re-suspended in lysis buffer (PBS pH 7.4 containing EDTA and 1 mM PMSF). Cells were then disrupted by using a French pressure cell at 1100 psi. Subcellular fractionation was performed as previously described [51] with slight modifications. Briefly, the cell lysate was centrifuged two times at 1000×g at 4 • C in order to remove unbroken cells. Supernatant (total lysate) was further centrifuged two times at 27000×g at 4 • C to separate cell wall/membrane from the cytoplasmic fraction. The pellet containing insoluble material was washed twice with PBS (pH 7.4) to remove any cytoplasmic contaminant, and supernatant containing the cytoplasmic fraction was further ultracentrifuged at 100000×g for 90 min at 4 • C to remove any membrane contaminant. Each fraction was loaded and separated on to 12% SDS/PAGE and transferred on to a nitrocellulose membrane using a Trans-Blot Turbo Transfer System (Bio-Rad). Immunoblotting of 6× His-tagged proteins was performed using anti-HisProbe TM HRP conjugate (Thermo Fisher Scientific). MSMEG 0220 was used as the control for subcellular location, and immunoblotting was performed by using rabbit polyclonal antibodies [52,53] and horseradish peroxidase-conjugated anti-rabbit IgG (Sigma-Aldrich). Revelation was achieved using Pierce TM ECL Western Blotting substrate solution (Thermo Fisher Scientific) and visualized with a ChemiDoc TM MP Imaging System (Bio-Rad).

Genapol R extraction and immunoblotting
Approximately 10 uOD 600 of recombinant cells cultured in 7H9-S medium supplemented with 50 μg/ml kanamycin at OD 600 ∼1-1.5 were harvested for 10 min at 4000×g. Cells were washed twice in PBS (pH 7.4) buffer containing 0.05% (v/v) Tween-20 and surface-exposed proteins were isolated by incubating cells with PBS containing 0.5% (v/v) Genapol R -X080 (Sigma-Aldrich) at room temperature [54,55]. Control samples were treated with a PBS buffer without Genapol R -X080. After approximately 30 min, all supernatants were collected and precipitated with 12% (v/v) trichloroacetic acid. Finally, samples were separated on to 12% SDS/PAGE and immunoblotting was achieved as described above.

Dynamic subcellular location using fluorescence microscopy
Gene encoding superfolder GFP (sfGFP) was amplified from pKD4-sfGFP plasmid with specific primers (Supplementary Table S1, P13-P15 and P14-P15) which were cloned in pMV261 and pVV16 vectors at either EcoRI/HindIII or only HindIII restriction sites. Thus, creating pMV261::sfGFP and pVV16::sfGFP. Rv0646c and MSMEG 1352 ORFs were amplified by PCR with specific primers (Supplementary Table S1, P5-P6 and P7-P8) containing NdeI and BamHI restriction sites in 5 and 3 ends, respectively. Digested and purified products were cloned into the pVV16::sfGFP in frame with the sfGFP gene and thus generating a translational fusion. M. smegmatis was electroporated with 1.5 μg of pMV261::sfGFP and these latter constructs; recombinant clones were selected on to specific medium and overexpression of the two chimeric proteins were checked by immunoblotting as described above. Recombinant strains were cultured until OD 600 reaches ∼1-1.5, and then harvested by centrifugation at 4000×g for 10 min. Cells were washed once in PBS (pH 7.4) buffer, and bacterial suspension (5 μl) was spotted between a coverslip of 170 μm thickness and a 1.5% agarose-PBS pad. Bacteria were analyzed by snapshot imaging at room temperature using an Olympus FV1000 confocal microscope and 100× oil-objective. Exposure time was 800 ms for both phase-contrast and fluorescence images (λ exc \ λ em = 490/530 + − 10 nm). Images recorded were processed using the open source program ImageJ 1.51K (NIH, U.S.A.).

Lipid extraction and analysis
Mycobacterial strains were grown in 1 l of 7H9-S broth until an OD 600 of ∼1 was reached, except when using acetamide inducible strains where cells were collected at specific time points. Cultures were then centrifuged at 4000g for 15 min at 4 • C and washed twice with distilled water. Pellets were lyophilized overnight and weighed to determine the exact mass of bacterial dry extract for normalization calculations.
MAs methyl esters (MAMEs) and fatty acids methyl esters (FAMEs) were extracted following previously described protocol [56,57]. Approximately, 10 mg of cellular dry weight were re-suspended in 2 ml TBAH (40% v/v, Sigma-Aldrich) and incubated at 100 • C overnight. Then 4 ml of CH 2 Cl 2 , 300 μl CH 3 I, and 2 ml of water were added to the mixture and incubated for 1 h at room temperature under shaking. Sample was centrifuged and the lower organic phase was kept, washed three times with distilled water, dried over MgSO 4 , and then concentrated under nitrogen stream. Finally, 3 ml of DTT were added to the dried extract, and sonication was performed for 10 min in a water bath. Sample was centrifuged for 10 min at 4000×g to remove any precipitate. The DTT layer was evaporated and the residue was re-suspended in 100 μl of DTT. The same volume of extract containing MAMEs and FAMEs were separated on 1D TLC using hexane/ethyl acetate (19:1, v/v) as eluent, and revealed by using a 10% phosphomolybdic acid solution in absolute ethanol.
Total lipids were extracted as previously described [58] with slight modifications. Briefly, lipids from dry extract were incubated for 16 h with CHCl 3 -MeOH (1:2; v/v), at room temperature under shaking. Residual lipids were re-extracted for 16 h with the same solvents, but using alternative ratios (1:1 and 2:1 v/v). All the three organic phases were pooled and concentrated under reduced pressure. Samples were re-suspended in a CHCl 3 -MeOH solution (3:1, v/v), and washed with 0.3% (w/v) NaCl in water. Organic and aqueous phases were separated by centrifugation 4000×g for 10 min, and only the organic phases were conserved and dried over MgSO 4 . Samples were finally evaporated under a nitrogen stream, weighed, and re-suspended in a CHCl 3 -MeOH solution (3:1, v/v).
Equal volume of extract containing glycopeptidolipids (GPLs) were separated on TLC (Silica Gel 60, Merck) by using ; v/v/v/v) as solvent system 2. Phospholipids were visualized by vaporization of a copper acetate/orthophosphoric acid mixture prepared by mixing a copper acetate saturated solution with a 85% (1:1, v/v) aqueous orthophosphoric acid solution. Phospholipid standards were purchased from Sigma-Aldrich. Finally, each resolved plate was heated at 120 • C for 5-10 min, scanned using a ChemiDoc TM MP Imaging System (Bio-Rad), and densitometric analyses was performed using the ImageLab TM software version 5.0 (Bio-Rad) allowing to determine the relative content of each sample.

REMA for MIC determination
Susceptibility testing was performed using the Middlebrook 7H9 broth microdilution method. All assays for each strain were carried out at least in triplicate. MICs against the various M. smegmatis bacterial strains were determined in 96-well flat-bottom Nunclon Delta Surface microplates with lid (Thermo Fisher Scientific) using REMA method [59]. Briefly, log-phase bacteria were diluted to a cell density of 5 × 10 6 cells/ml in 7H9-S medium. Then 100 μl of the above inoculum (i.e. 5 × 10 5 cells per well) was added to each well containing 100 μl of 7H9-S medium, serial two-fold dilutions of the selected antibiotics (RIF or INH), CyC 17 [28,29], HPOX [27], or Orlistat to a final volume of 200 μl. Growth controls containing no inhibitor (i.e., bacteria only), inhibition controls containing 50 μg/ml kanamycin and sterility controls (i.e. medium only) without inoculation were also included. Plates were incubated at 37 • C in a humidity chamber [60] to prevent evaporation for 3-5 days. Then, 20 μl of a 0.02% (w/v) resazurin solution was added to each well, and the plates were incubated at 37 • C for color change from blue to pink or violet and for a reading of fluorescence units (FU). Fluorescence corresponding to the resazurin reduction to its metabolite resorufin was quantitated using a Tecan Spark 10M multimode microplate reader (Tecan Group Ltd, France) with excitation at 530 nm and emission at 590 nm. Fluorometric MICs were determined by fitting the RFU% sigmoidal dose-response curves [28,29] in Kaleidagraph 4.2 software (Synergy Software). MIC values determined as the lowest compound concentrations inhibiting 50% of growth were defined as the MIC 50 .

LipG is a highly conserved mycobacterial protein
Gene Rv0646c encoding LipG MTB protein was not only found within M. tuberculosis genome. Genomic context containing Rv0646c gene and its orthologs was indeed surprisingly highly conserved within many mycobacterial species. In M. tuberculosis, Rv0646c is surrounded by an essential uncharacterized gene (Rv0647c) and the mmaA cluster composed of four mmaA genes (mmaA1-4) encoding proteins involved in MAs maturation processes (Figure 1).

Figure 1. Genomic organization of lipG clusters in mycobacterial species
Schematic organization of the lipG genomic loci from M. tuberculosis, M. leprae, M. marinum, and M. smegmatis were designed based on the Mycobrowser database [32]. lipG genes are represented in yellow whereas genes involved in MAs maturation processes are in black. Furthermore, a conserved ORF encoding a putative ABC transporter was highlighted in red. Co-ordinates of each gene cluster are indicated in kb. Scale represents 500 bp.
Interestingly, the occurrence and genome organization of lipG clusters are also highly conserved in M. tuberculosis, M. leprae, M. marinum, or M. smegmatis genomes (Figure 1), with two copies of genes encoding for LipG proteins in M. marinum, M. abscessus, or M. ulcerans genomes. Multiple sequence alignment of these various LipG proteins and sequence comparison, revealed a high level of amino-acid conservation, also with some consensus amino-acids blocks between each mycobacterial species [21] (Supplementary Figure S1). The LipG characteristics (i.e. sizes of genes/proteins and sequence identities) from different mycobacterial strains are summarized in Table 1. Each of these proteins possess a molecular weight ∼32 kDa, with the number of amino acids ranging from 301 to 324. Taken together, all these findings suggest that the cluster including the lipG gene may be dedicated to a common mycobacterial feature either in strict pathogenic, opportunistic, or non-pathogenic strains.

Cloning, expression purification, and biochemical characterization of LipG MTB
In an attempt to characterize LipG biochemical activities more finely, to confirm its carboxylesterase properties [21], and also to decipher the physiological role of this protein, the Rv0646c and MSEG 1352 genes have been cloned. The respective recombinant rLipG MTB and rLipG MS were produced in E. coli C41 (DE3)/pLyS strain harboring the pRare plasmid. rLipG MTB and purification was achieved using Ni 2+ -NTA affinity chromatography, yielding ∼1 mg of pure recombinant protein per culture liter. The purity and the expected molecular weight (∼35 kDa) were confirmed by 12% SDS/PAGE ( Figure 2A) and MALDI-TOF analysis.
The lipase activity of rLipG MTB has also been tested using a large panel of triacylglycerol (TAG) substrates, such as trioctanoin and triolein using the pH-stat technique, and pomegranate oil-coated TAG method [48]. In all cases, rLipG MTB was not able to hydrolyze any of these TAGs, even when using highly sensitive fluorogenic TAG substrates [40]. It is well known that lipases differ from esterases by their ability to hydrolyze long-chain acylglycerols at the oil-water interface, whereas esterases can only hydrolyze substrates with short-or medium-fatty acid chains [61]. From these findings, rLipG MTB could be considered as a true carboxylesterase, as recently proposed by Rastogi et al. [21].
Due to its genomic localization in the same cluster as the mmaA genes encoding for methyltransferase proteins, it is tempting to suggest that rLipG MTB might also be able to hydrolyze the thioester bond of acyl-CoA or MA intermediates as previously shown for Cut6 (Rv3802c). Cut6 is an essential bifunctional phospholipase A and thioesterase enzyme belonging to the Cutinase-like family proteins from M. tuberculosis [62]. Thioesterases are a group of enzymes that catalyze the hydrolysis of a thioester bond between a carbonyl group and a sulphur atom. Here, the hydrolysis of palmitoyl-CoA was used to measure the thioesterase activity of rLipG MTB . As shown in Figure 2C, rLipG MTB also exhibited a thioesterase activity with an S.A value of 2.10 + − 0.59 mU/mg. It is noteworthy that this value is of the same order of magnitude as that of TesA, a thioesterase of M. tuberculosis [42] with a S.A of 5.95 + − 0.85 mU/mg.
In order to investigate further rLipG MTB enzyme activity, its phospholipase activity was also investigated. Interestingly, rLipG MTB was able to hydrolyze only fluorescent phospholipids present at the sn−1 stereospecific position with S.A value of 2.75 + − 0.25 mU/mg ( Figure 2D) suggesting that rLipG MTB is a strict PLA 1 . T. lanuginosus lipase PLA 1 (TL PLA 1 ) and Bee venom PLA 2 (Bv PLA 2 ) were used as positive controls for the PLA 1 and PLA 2 activities, respectively.
Finally, since Rastogi et al. [21] had proposed that the catalytic machinery was governed by Ser 123 , we thus generated by site-directed mutagenesis a rLipG MTB S123A variant, in which this Ser 123 residue was mutated into an Ala residue. As predicted, no enzyme activity was observed with fluorescent-phospholipids, pNP-substrates or palmitoyl-CoA with this inactive mutant ( Figure 2E and data not shown). Taken together, these data provide full evidence that Ser 123 contained in the consensus sequence G 121 ASMG 125 is the catalytic residue responsible for the enzyme activity, and consequently, confirms that LipG MTB is a true serine hydrolase. More specifically and given the level of S.A reached with the various substrates investigated, our results suggest that rLipG MTB is a phospholipase A 1 enzyme which also displays esterase and thioesterase activities.

rLipG MTB is inhibited by OX derivatives and CyC analogs
Two series of recently reported antitubercular inhibitors discovered in our laboratory (OX derivatives and CyC analogs), exhibit activity against enzymes by forming a covalent bond with the catalytic serine or cysteine residue in their active sites [27][28][29]. In order to further investigate the ability of these compounds to efficiently inhibit the lipolytic activity of rLipG MTB , six OX derivatives (i.e. MmPPOX, iBPOX, iBpPPOX, HPOX, HpPPOX, and BePOX) [27]; and five CyC analogs (i.e. the CyC 17 , CyC 7α , CyC 7β , CyC 8α , and CyC 8β ) [28,29] were then tested for inhibition of rLipG MTB . In our previous work, whereas iBpPPOX, and CyC 7β and CyC 8α exhibited strong activity against both extracellular and intramacrophagic M. tuberculosis; iBPOX and HpPPOX, and CyC 7α and CyC 8β were found active on infected macrophages only [28,42]. In contrast, HPOX and BePOX, and CyC 17 efficiently block M. tuberculosis growth in vitro with no activity against intracellular bacilli [28,42]. Orlistat which has been previously reported as an inhibitor of LipG was used as positive control [63]. The chemical structures of the 12 compounds used in this study are provided in Supplementary Figure S2. rLipG MTB was inactivated by all the six OX derivatives, with medium 60.6% to good 93.8% inhibition levels at x I = 100 ( Table 2). The best inhibitors were HPOX, MmPPOX, and HpPPOX, which displayed inhibitor molar excess leading to 50% enzyme inhibition, i.e. x I50 values of ∼3.0 (Table 2 and Figure 3A). Amongst the CyC analogs, CyC 7β , CyC 8α , and CyC 8β exhibited very weak inhibitory effects up to a high molar excess x I = 100. Whereas, a strong dose-dependent inhibition was observed with CyC 17 and CyC 7α (Table 2 and Figure 3B). In particular, CyC 17 was found to react almost stoichiometrically with rLipG MTB , as confirmed by its x I50 value of 0.98. It is noteworthy, that Orlistat, used as reference inhibitor, displayed only a moderate inhibitory activity against pure rLipG MTB with only 69.5% inhibition at x I = 100 and a x I50 value of 15.2, similar to those of iBpPPOX, iBPOX, and BePOX.  1 Inhibition data (% of initial activity), at inhibitor molar excess (x I ) of 4 and 100 related to 1 mol of enzyme. Results are expressed as mean values + − S.D. of at least three independent assays. The inhibitor molar excess leading to 50% lipase inhibition, x I50 , was determined as described in Experimental procedures section.

rLipG MTB and rLipG MS are peripheral membrane protein
Although LipG MTB has been identified in the membrane fraction of M. tuberculosis H37Rv by using proteomic profiling approach [64], in silico experiments performed using SignalP 4.1 server [65] revealed that neither a peptide signal for the general secretion pathway or for the twin-arginine translocation pathway, nor a Type VII consensus secretion motif YxxxD/E were present in the primary sequence of LipG proteins. In this context, recombinant M. smegmatis strains, overexpressing either rLipG MS (i.e. MSMEG 1352 gene) or rLipG MTB (i.e. Rv0646c gene) cloned under the control of the hsp60 promoter in frame with a C-terminal 6× His-tag, were subjected to subcellular fractionation followed by immunoblotting experiments. The well-characterized cell wall exported monoglyceride lipase, MSMEG 0220, was used as control for cell fractionation [52,53]. In each case, both proteins were found only within the cell wall fraction and not in the soluble cytoplasmic fraction suggesting that LipG is a membrane/cell wall-associated protein ( Figure 4A,B). Additional differential detergent extraction of surface-exposed mycobacterial proteins using Genapol R -X080 [54,55], followed by precipitation and immunoblotting (Figure 4C), demonstrated that LipG protein remains associated with the pellet fraction. This results suggest that this enzyme would not be accessible to the detergent and thus is not a surface-exposed protein. In contrast, MSMEG 0220 was found within the supernatant fraction, as expected ( Figure 4C). All these data strengthen the fact that mycobacterial LipG are peripheral proteins, interacting with the inner leaflet of the cytoplasmic membrane.
To go further, the dynamic location of LipG within living cells was investigated by using fluorescence microscopy [66]. Gene fusions between Rv0646c or MSMEG 1352 and the sfGFP gene encoding the sfGFP [67] were generated and used to transform M. smegmatis cells. Recombinant clones harboring stable fusions (Supplementary Figure S3) with a constant and important expression level were grown to mid-log exponential phase (i.e. OD 600 = 1) in classical 7H9-S medium and further analyzed by fluorescence microscopy. As shown in Figure 4D (right panel), the strains harboring LipG MTB -sfGFP chimeric construct emitted a strong fluorescence signal mainly localized at the periphery of the inner mycobacterial cell membrane. In contrast, the fluorescence signal of the recombinant strain expressing only a wild-type sfGFP was less intense and diffuse through the cytoplasmic compartment ( Figure 4D, left panel). Similar results were obtained when the bacteria was processed either at early exponential phase (i.e. OD 600 = 0.2) or at the stationary phase (i.e. OD 600 = 3). These findings strongly suggest that the resulting chimeric LipG proteins remain clustered into the bacterial cytoplasmic membrane whatever the phase of bacterial growth.
To get structural insights concerning LipG MTB and its potential interaction within the cytoplasmic membrane, a 3D structural model was built using the I-TASSER server [35,36]. The X-ray structure of Streptomyces lividans methylesterase RdmC (PDB id: 1Q0R), solved at 1.45Å resolution was used as a template [68]. This generated model revealed that LipG MTB is a serine hydrolase constituted by 11 α-helices and 8 β-strands that are forming a central β-sheet (Supplementary Figure S4) as previously described [21]. A putative catalytic triad composed by Ser 123 , Asp 251 , and His 279 is located within a highly accessible active site surrounded by hydrophobic residues forming the catalytic pocket (Supplementary Figure S4). In an effort to better characterize the molecular interactions that might govern LipG binding on to the inner cell membrane, the electrostatic surface potential of the previously built 3D model structure of LipG MTB (Supplementary Figure S4) was generated and is displayed in Figure 4E,F. Interestingly, a positive electrostatic patch located opposite to the active site, on top of the model structure, was found. Notably, this specific amino-acid patch is mainly constituted by six arginine residues highly conserved in several LipG orthologs ( Figure  4E,F and Supplementary Figure S5). Such arginine residues, which are indeed known to play important role in protein  binding to zwitterionic and anionic glycerolipid polar head groups [69,70], might thus contribute to the binding of LipG to the cytoplasmic membrane ( Figure 4E,F).

Disruption of MSMEG 1352 triggers cell wall modifications
In order to identify the physiological role of the LipG proteins within the mycobacterial lifecycle, a M. smegmatis MSMEG 1352 disrupted strain named lipG MS ::Hyg, in which the gene has been interrupted with a hygromycin resistance cassette, as well as a lipG MS ::Hyg complemented strain (i.e. lipG MS ::Hyg Comp) were produced (Supplementary Figure S6). When grown in 7H9-S medium for 48 h at 37 • C, no significant differences in growth rate were observed between the WT, lipG MS ::Hyg, and lipG MS ::Hyg Comp strains. Interestingly, when the strains were grown on to 7H11TG agar plates (i.e. 7H11 medium supplemented with 0.02% Tween 80 and 0.2% glycerol), no differences in colony morphologies were observed with naked eyes. In contrast, when agar plates were spot-inoculated on to the same medium devoid of detergent and observed under a stereomicroscope, the lipG MS ::Hyg mutant strain harbored a slightly smoother and rounder phenotype than the WT and lipG MS ::Hyg Comp strains ( Figure 5A). This finding would suggest that disruption of lipG MS may trigger some minor cell surface modifications ( Figure 5A). TLC analysis of total lipid extract reveals that disruption of lipG MS induces a slight but significant overproduction of GPLs by 1.35-fold compared with WT strain ( Figure 5B). In contrast, no detectable changes were observed within FAMEs and MAMEs levels as well as phospholipid content between each strain (Supplementary Figure S7A,B). Although bacterial cell wall remodeling (including misregulation of GPL production) may also affect biofilm/pellicle formation, neither global nor visible modifications where observed amongst the WT, lipG MS ::Hyg, and lipG MS ::Hyg Comp strains regarding pellicle formation (Supplementary Figure S7C) in 7H9-S medium containing either glycerol or acetate as carbon source.
Since the mycobacterial cell wall is the first effective barrier against bactericidal compounds, we hypothesized that increase in GPLs could also play an important role in cell wall integrity. To answer this question, the WT, lipG MS ::Hyg, and lipG MS ::Hyg Comp strains were subjected to antibiotic susceptibility testing toward standard antibiotics INH and RIF, using the REMA method [28,59] (Figure 5C and Table 3). The effects of Orlistat, HPOX as well as the CyC 17 inhibitor on the growth of each strains were also investigated.
It is noteworthy that the respective MIC 50 values obtained with INH (0.76 μg/ml), RIF (1.4 μg/ml), and CyC 17 (0.56 μg/ml) against the WT strain are in accordance with previously published studies ( Figure 5C and Table 3) [28,29]. Regarding Orlistat, this well-known serine hydrolase inhibitor [23,71] is able to impair M. smegmatis growth with an MIC 50 of 0.97 μg/ml. Surprisingly, deletion of lipG MS leads to an increased drug tolerance toward the four compounds. As depicted in Figure 5C and Table 3, MIC 50 of INH, RIF, Orlistat, and CyC 17 against the lipG MS ::Hyg strain amounted to 2.9, 3.3, 6.5, and 9.6 μg/ml, respectively, thus corresponding to a respective fold change compared with the WT strain of ×2.0, ×4.3, ×6.7, and ×15.7. The complementation of the mutant strain with the pVV16::lipG MS restored the drug susceptibility with mean MIC 50   to restore the drug susceptibility, suggesting that the two proteins may possess similar enzymatic activities and/or physiological functions (Table 3).
To further confirm this result, the recombinant rLipG MS was produced, purified, and biochemically characterized as described above for rLipG MTB (Supplementary Figure S8 and Table S2). Both pure rLipG MS and rLipG MTB proteins thus exhibit the same substrate specificity toward pNP ester substrates, the best S.A being also found with pNP caproate (C6). Moreover, rLipG MS was also able to hydrolyze acyl-CoA and phospholipids as substrates, confirming that the substrate specificity is well conserved between the two proteins. Despite their similar reactivity pattern with various esters, we noticed, however, that the rLipG MS was more active than its ortholog (Supplementary Table S2). Interestingly, these differences where only observed in vitro using pure proteins, but not during trans-complementation with the lipG MTB gene during in vivo experiments as described above.

Overexpression of lipG MTB impacts free fatty acids and phospholipids homeostasis
Despite the observed impact of lipG MS gene disruption on to the penetration/diffusion of the latter antitubercular molecules within the mycobacterial envelope, no clear detectable phenotype has been observed. To overcome such issue, a recombinant strain harboring the pMyC::lipG MTB plasmid which allows a strong overproduction of the LipG MTB protein upon acetamide induction, has been constructed. Approximately 6 and 24 h following induction by acetamide, total lipid extraction was performed on cell cultures either in exponential (OD 600 ∼1-1.5) or stationary (OD 600 ∼3-4) phase, respectively. The effect of LipG MTB overproduction was scrutinized by TLC analysis. As expected from our previous results which showed a slight increase in GPL production in absence of LipG MS ( Figure  5A Regarding phospholipid levels, surprisingly overexpression of lipG MTB slightly increases the phospholipid level of phosphatidylethanolamine (PE) and phosphatidylglycerol (PG) subspecies (1.75-and 1.07-fold in comparison with the control strain, respectively), but more importantly, this overproduction drastically impacts phosphatidylinositol (PI) production with a fold-change related to the control strain of ×2.3 ( Figure 6C). These results were further confirmed by using 2D-TLC analysis, where a nice and efficient separation of all species of phospholipids allowed to confirm a consequent increase in PI biosynthesis upon LipG MTB overproduction (Supplementary Figure S9).
Altogether, the increase in the levels of both FAMEs and specific subclass of phospholipids suggest that the LipG MTB acyl-CoA-thioesterase activity may be the predominant biochemical activity in vivo, and that the products of the reaction are directly influencing cell wall phospholipids composition.

Discussion
M. tuberculosis is one of the most successful infectious agents worldwide. Therefore, understanding the molecular mechanisms responsible for its pathogenesis remains one of the most important challenges in TB fundamental research. By using highly elaborate genetic screens, several independent teams have indeed tried to identify the genes required for in vitro or in vivo bacterial survival [26,72,73]. These approaches have highlighted the Rv0646c gene of M. tuberculosis which is required for adaptation and survival of the bacilli in mouse macrophages [26]. Interestingly, this gene has been annotated as encoding a putative lipase/esterase protein based on the presence of a G-x-S-x-G consensus pentapeptide motif. More importantly, the Rv0646c gene encoding LipG MTB is widely conserved through the Mycobacterium genus and is associated with a gene cluster involved in MAs maturation processes [19]. Strikingly, bioinformatics analysis demonstrated that this gene is highly conserved in both slow-and fast-growing mycobacteria, and even in the obligate intracellular parasite M. leprae. This latter unusual pathogen is harboring a genome with a large number of inactivated pseudogenes which are not required anymore in vivo [74]. However, amongst the 24 genes encoding lipolytic enzymes from the Lip family in M. tuberculosis, only lipE, lipU, and lipG are conserved in M. leprae, thus reinforcing an essential role played by these enzymes in intracellular survival and proliferation [74,75]. In this study based on biochemical characterization of rLipG MTB protein associated with microbiology experiments, we uncovered the potential role of this enzyme in mycobacterial lipid metabolism.
In accordance with a recent report, we confirmed that recombinant LipG MTB protein is an esterase acting preferentially on medium chain-length substrates [21,75]. More interestingly, we also found that the protein was able to hydrolyze phospholipids at the sn−1 position and cleave the thioester bond from palmitoyl-CoA, showing that rLipG MTB possesses esterase and phospholipase, as well as thioesterase activities (Figure 2). Both phospholipase and thioesterase activities have been already observed by Parker et al. [62], who described that M. tuberculosis Rv3802c Figure 6. LipG MTB impacts free fatty acids and phospholipids biosynthesis Mycobacterial cultures harboring pMyC and pMyC::lipG MTB were grown in 7H9-S medium until reaching OD 600 nm ∼0.5, then 0.2% of acetamide was added inside the culture medium. Cells were collected at 6 or 24 h post-induction corresponding to exponential or stationary phase, lyophilized and the same amount of dry cell weight was use for lipid extraction. After normalization, the same volume of each sample was loaded on to TLC. (A) Global GPL levels are decreasing in an M. smegmatis pMyC::lipG MTB overexpression strain. Lipid extracts of each samples were analyzed by TLC followed by densitometry analysis. (B) FAMEs levels are increasing in an M. smegmatis pMyC::lipG MTB overexpression strain. Lipid extracts of each samples were analyzed by TLC followed by densitometry analysis. In each case, results are expressed as mean + − S.D. of each respective total lipid species levels from at least two independent experiments, where control strain level at t 6 h was arbitrary adjusted to 1 relative unit. (C) Phospholipid biosynthesis is increased in an M. smegmatis pMyC::lipG MTB overexpression strain. Lipid extracts of each samples were analyzed by TLC followed by densitometry analysis. TLC was performed with CHCl 3 -MeOH-H 2 O (65:25:4; v/v/v) as eluent and PI overproduction is indicated with a red asterisk (left panel). Lipid sample from M. smegmatis pMyC::lipG MTB at t 6 h was also co-injected with pure PI standard as control lane, to clearly identify PI as an overproduced subspecies. Histograms correspond to densitometry analysis at t 6 h . Results are displayed as intensity ratio where phospholipids levels from the pMyC::lipG MTB strain were divided by the phospholipids levels from the pMyC strain (right panel). Results are expressed as mean + − S.D. from three independent experiments. Statistical analysis was performed by using the Student's t test where * correspond to a P-value <0.05. Abbreviation: PI, phosphatidylinositol. gene (cut6) encodes a functional phospholipase/thioesterase protein. Using radiolabeled phospholipids coupled with TLC analysis, the authors showed that Cut6 was not only able to act on PE, PC, and phosphatidylserine (PS) molecules [62], but that it was also able to hydrolyze both palmitoyl-and decanoyl-CoA, suggesting that the protein may act as an acyl donor in vivo.
The fact that LipG was required for adaptation and survival within macrophages, prompted us to investigate the inhibitory properties of two new series of serine hydrolase inhibitors active against M. tuberculosis growth [27][28][29]. During in vitro inhibition experiments on pure protein ( Figure 3 and Table 2), the best inhibitors from the two series of compounds were HPOX (for the OX derivatives) on the one hand, and CyC 17 (for the CyC analogs) on the other hand; two molecules that are only active against extracellularly growing M. tuberculosis strain [27,28]. In the case of CyC 17 , this result was a posteriori predictable. Following activity-based protein profiling experiment coupled with proteomics analyses [28], LipG was indeed found as a potential target of this compound from a lysate of M. tuberculosis (unpublished data). However, its low coverage (21% sequence identity) with only five distinct peptide sequences identified together with a peptide score ∼2.0-3.4 (given by SequestHT algorithm) made that this protein was below our threshold and was not selected. Conversely, LipG was not identified from similar ABPP experiments performed with HPOX directly on living M. tuberculosis cells [27]. We recently demonstrated that both the OX and CyC families of compound are multitarget inhibitors, which are probably displaying pleiotropic and cumulative effects toward mycobacterial Ser-and Cys-containing enzymes involved in lipid metabolism and/or cell wall maintenance, thus leading to M. tuberculosis growth arrest [27,28]. From these findings and despite their strong inhibitory effect in vitro on pure LipG enzyme, we thus proposed that in vivo HPOX and CyC 17 preferentially impair other previously identified target enzymes such as the Ag85 complex or the thioesterase TesA rather than LipG MTB .
Based on proteomic analysis of the M. tuberculosis membrane fraction, it has been recently proposed that LipG MTB may be localized to the cell surface and directly interact with the host immune system [21,64]. However, bioinformatics analysis did not reveal the presence of any secretion signals or specific motifs, which may indicate that the protein is surface exposed. Herein, using different cell-fractionation approaches, we demonstrated that LipG MTB is associated with the membrane fraction [54,55,66], but not translocated across the cytoplasmic membrane and remains a cytoplasmic protein which is mainly peripheral (Figure 4). It has been described that such peripheral proteins require either a lipophilic attachment (i.e. acylation) or the presence of basic residues positively charged [69,70] which induce electrostatic interaction with the acidic and negatively charged phospholipids head-groups, thus allowing anchoring of the protein within the cytoplasmic membrane [70]. Based on our 3D structure model, surface electrostatic potential analysis allows identification of a region on top of the active site mainly composed by positively charged residues ( Figure 4E,F). This area is essentially constituted by a C-terminal cluster of arginine residues fully conserved in numerous LipG proteins (Supplementary Figure S5). Since this type of electrostatic interaction has already been described for a mycobacterial protein involved in acyl-CoA metabolism [76], one can assume that this cluster of Arg residues would play an important role in LipG binding to the cell membrane, and thus for its physiological function.
In contrast with Cut6, which is known for being secreted and acting outside the mycobacterial cell [62,77], LipG MTB and LipG MS are presumably hydrolyzing cytoplasmic substrates. Two distinct strategies were employed: (i) the generation of a lipG MS disrupted strain which was trans-complemented with either lipG MS or lipG MTB and (ii) the generation of a lipG MTB heterologous overexpression strain which were useful in identifying the potential physiological role of LipG proteins and to define their substrates in vivo. On one hand, the observed increase in GPLs production due to lipG gene disruption ( Figure 5) resulted in a significant reduced susceptibility to RIF, INH, Orlistat, HPOX, and above all CyC 17 for which the highest fold change compared with WT strain in MIC 50 value has been reached ( Table 3). Since RIF and INH are inhibiting specifically cytoplasmic targets, we hypothesized that such increase in GPL will affect cell wall fluidity and integrity; which subsequently impacts the penetration of the main two first-line anti-TB drugs [78]. Such results are in agreement with Khoo et al. work [79] who reported that Mycobacterium avium GPLs positive serovar were less sensitive to ethambutol than GPLs negative serovar, and concluded that hydrophobic GPLs production may alter cell wall permeability and correlate with antibiotic tolerance [79]. In other hand, the overproduction of the LipG MTB protein leads to a significant increase in the FAMEs level (1.42-fold and 1.25-fold at 6 and 24 h post-induction, respectively), but more importantly a decrease in GPLs levels within M. smegmatis concomitant to an increase in level of phospholipid species and more particularly, the phosphatidyl inositol ( Figure 6). However this is not the first time that such modifications of the cell wall composition have been observed. Similar phenotypes with drastic alteration of FAMEs/MAMEs and phospholipid subspecies levels have already been described in 2009, by Meniche et al. [58]. The authors demonstrated that overexpression of the Corynebacterium glutamicum gene NCgl2775 encoding a functional thioesterase/phospholipase protein leads to an increase in MAs production concomitantly with a decrease in glycerophosphoslipids levels [58]. A similar result was also obtained when the MSMEG 6394 protein (i.e. NCgl2775 and Cut6 orthologs) was overproduced, which prompted the authors to propose that this kind of bifunctional enzyme is may be able to regulate membrane compositions under specific conditions [58].
Since not all the mycobacterial strain are GPLs producers, especially M. tuberculosis H 37 Rv, we postulated that this phenotype was more an indirect effect, than a direct catabolic process transposable to M. tuberculosis.
From all these findings, it is tempting to propose that LipG is mainly acting as a thioesterase dedicated to phospholipid metabolism in vivo. In a large number of bacteria, the contribution of thioesterase proteins to phospholipid biosynthetic process is however not well-defined. For example, in E. coli, glycerol starvation leads to an increase in acyl-ACP and a drastic diminution of phospholipid levels. Overexpression of TesA, a type I thioesterase, drastically reduced the accumulation of long-chain acyl-ACP species, probably by rerouting free fatty acids to phospholipids [80]. It has also been proposed that the ybgC gene encoding a functional acyl thioesterase is playing a role in phospholipid metabolism in Gram-negative bacteria [81][82][83]. Astonishingly, this latter enzyme is known for interacting with a large number of proteins dedicated to phospholipid homeostasis, such as the acyl-carrier protein, the glycerol 3-phosphate acyltransferase (PlsB), and the PS synthase (PssA) [82]. Such systems dedicated to phospholipids have not been described so far in mycobacteria, but a similar multicomponent complex is involved in MA biosynthesis [16,17]. Overall, it might be possible that some proteins work together in order to respond to specific stimuli and thus modulate biological membrane composition and fluidity. Such a physiological process has also been reported in eukaryotic cells, where upon inflammatory stimulus, the acyl-CoA thioesterase 7 (ACOT7) converts acyl-CoA molecules into free fatty acids, thus triggering a remodeling of phospholipid containing unsaturated long (≥C 20 )-acyl chains in macrophages [84].
Finally, PI is known for being an essential and abundant phospholipid species within mycobacterial cytoplasmic membrane [85], but also for being a major precursor for complex glycoconjugated lipids such as phosphatidyl mannosides, lipomannan, or lipoarabinomannan [86,87]. Observation of an important increase in PI subspecies during LipG MTB overproduction may also impact the formation of such glycolipids that are highly contributing to virulence and modulation of the host immune response. The potential role of the LipG MTB protein within these pathways is currently under investigation in our laboratory and may bring new insight concerning the biosynthesis of such complex lipids, but also the role of this highly conserved protein during host-pathogen cross-talk.