The final assembly of trehalose polyphleates takes place within the outer layer of the mycobacterial cell envelope

Laurie Thouvenel, Gautier Prevot, Laura Chiaradia, Julien Parra , Emmanuelle Mouton-Barbosa, Marie Locard-Paulet , Julien Marcoux , Maryelle Tropis, Odile Burlet-Schiltz, Mamadou Daffé, Christophe Guilhot , Gilles Etienne, and Christian Chalut* From the Institut de Pharmacologie et de Biologie Structurale, Université de Toulouse, CNRS, UPS, Toulouse, France and the Novo Nordisk Foundation Center for Protein Research, University of Copenhagen, Copenhagen, Denmark

Mycobacteria are endowed with an unusually thick lipid-rich cell envelope. This structure contains several families of trehalose-containing glycolipids that interact with mycolic acids attached to the arabinogalactan to form an atypical outer membrane called the mycomembrane (1). Among these, the ubiquitous trehalose monomycolates and trehalose dimycolates are essential for viability (2). This group also includes species-specific glycolipids, such as di-and polyacyltrehaloses (DAT and PAT) and sulfolipids (SL), which are restricted to the human pathogen Mycobacterium tuberculosis, or lipooligosacharides that have been isolated from diverse fast-and slow-growing mycobacteria (1). Recently, we reported that trehalose polyphleates (TPP), a family of surface-exposed glycolipids originally described in Mycobacterium phlei, are widely distributed across mycobacterial species, including Mycobacterium smegmatis and the opportunistic pathogens Mycobacterium abscessus and Mycobacterium avium (3)(4)(5). TPP consist of octoacylated trehalose bearing seven C36:5 and C40:6 polyunsaturated fatty acids called phleic acids and a C14-C19 fatty acid residue. They share a common basic structure with SL and PAT from M. tuberculosis (6). The biological function of TPP in mycobacteria remains enigmatic. A recent study showed that TPP production in M. abscessus correlates with clump and cord formation, suggesting a potential role for these molecules in the virulence of this opportunistic human pathogen (7).
Formation and export of TPP involve at least four biosynthetic enzymes and an MmpL transporter, namely MmpL10, that are encoded by genes clustered at the TPP locus (5). These proteins display sequence similarities with enzymes required for SL and PAT production, indicating that the TPP pathway can serve as an informative model to describe glycolipid production in M. tuberculosis. Biosynthesis of TPP begins in the cytoplasm with the formation of a 2,3-diacyltrehalose intermediate bearing a C14-C19 fatty acyl group and a phleic acyl substituent. The transacylase enzyme PE encoded by the MSMEG_0412 gene (previously referred to as MSMEI_0402), subsequently catalyzes transesterification of phleic acids between diacyltrehalose precursors to generate TPP. Finally, MmpL10 is involved in the translocation of TPP or/and of TPP precursors across the plasma membrane (5,8).
Despite this knowledge, the late stages of the TPP biosynthetic pathway and the relationship between biosynthesis and transport of these compounds to the cell envelope remain elusive. The PE protein is predicted to be a membrane-anchored protein harboring a putative N-terminal signal peptide and a Cterminal a/b-serine hydrolase domain, consistent with its proposed role as an acyltransferase in TPP assembly ( Fig. 1) (9,10). Interestingly, Chp1 and Chp2, two membrane-bound acyltransferases, respectively involved in SL and PAT production in M. tuberculosis, share overall domain organization with PE ( Fig. 1). It has been proposed that these enzymes catalyze the last step of SL and PAT assembly by a mechanism that is tightly coupled to lipid transport across the plasma membrane, but the molecular details of this process are not fully understood, partly because the in vivo subcellular localization of their catalytic domains remains controversial. Seeliger et al. (11) indeed demonstrated that the catalytic domain of Chp1 is located in the cytoplasm, but two studies reported contradictory results on the orientation of Chp2, the catalytic domain being either cytoplasmic or periplasmic (12,13). Likewise, the orientation of PE in the plasma membrane and thereby the subcellular location of its catalytic domain, is unknown, raising questions on the cell compartment in which the transacylation step takes place and on the identity of the compounds that are transported by MmpL10.
The goal of this study was to shed light on the last step of TPP biosynthesis by investigating the topological organization and subcellular location of the PE protein. Using a combination of genetic, biochemical, and proteomic approaches, we established that PE has a C-terminal periplasmic domain and found that proteolytic cleavage of PE is required for efficient formation of TPP. Based on these data, we present here a refined model of the TPP biosynthesis pathway. This work provides clues for a better understanding of the molecular mechanisms involved in DAT/PAT production in M. tuberculosis.

Results
The C terminus of PE is located outside the plasma membrane In silico analysis using the SignalP software predicts that PE possesses a putative signal sequence with a AXA signal peptidase I cleavage motif (PE 1-28 , cleavage site between positions 28 and 29: AAA-DD, Fig. 1) (14); it is thus possible that this protein is associated with the plasma membrane, facing either the periplasm or the cytoplasm, or exported to the bacterial cell wall after cleavage of the signal sequence. To discriminate between these possibilities, we first investigated the topological organization of PE in M. smegmatis. Plasmids expressing PE fused at the C terminus with either the GFP or the alkaline phosphatase (PhoA) (Table S1) were introduced in the WT strain of M. smegmatis, and the GFP and PhoA activities were monitored to determine the subcellular localization of the reporter proteins. GFP exhibits fluorescence only when localized to the cytosolic compartment, whereas PhoA is only active inside the periplasmic space (15,16). Expression of PE-GFP or PE-PhoA in an M. smegmatis pE knockout mutant (PMM229) restored TPP production, indicating that the fusion proteins are correctly folded and display proper subcellular localization in bacteria (Fig. S1). Additional M. smegmatis strains producing KatG1, a cytosolic protein, MmpL10, a transmembrane protein with a cytoplasmic C-terminal end, or MmpS4, a transmembrane protein with the C-terminal end facing the periplasm, fused to the same topology reporter proteins were generated to serve as controls in the experiments (16,17). Bacteria producing PE-GFP displayed no increase in average fluorescence when compared with WT or control bacteria expressing MmpS4-GFP ( Fig. 2A). Conversely, as expected, high fluorescence signals were detected for strains expressing KatG1-GFP or MmpL10-GFP. M. smegmatis strains producing the PhoA fusions were first assayed for alkaline phosphatase activity on LB agar plates containing the chromogenic substrate 5-bromo-4-chloro-3-indolyl phosphate (BCIP). Production of PE-PhoA resulted in blue colonies after incubation at 37°C, indicative of PhoA activity (Fig. 2B). A similar pattern was observed for the control strain expressing MmpS4-PhoA, whereas no color change was observed for strains expressing either KatG1-PhoA or MmpL10-PhoA. To confirm these data, the PhoA activity of fusion proteins was assessed on cells grown in LB using the chromogenic substrate p-nitrophenylphosphate (pNPP). Results obtained were consistent with those obtained on BCIP agar plates: bacteria that express PE-PhoA, as well as those producing MmpS4-PhoA, exhibited high levels of PhoA activity compared with the WT strain and control strains expressing KatG1-PhoA or MmpL10-PhoA (Fig. 2B). Collectively, these data indicated that the C-terminal domain of PE is exposed on the outside of the plasma membrane. In these studies, we also sought to confirm the topology of Chp1 and to resolve the contradictory research findings regarding the topological organization of Chp2, using M. smegmatis strains expressing each of these proteins fused to either the GFP or PhoA. Attempts to detect GFP fluorescence or PhoA activity in bacteria producing the Chp1 fusion proteins failed (data not shown). For Chp2, we observed a significant PhoA activity in cells expressing Chp2-PhoA on both solid and liquid media, but no increase in fluorescence in bacteria producing Chp2-GFP, compared with the WT or the negative control MmpS4-GFP (Fig. 2, A and B), thus establishing that like PE, the Chp2 protein has its C-terminal domain exposed to the periplasmic space.

PE is present in the cell wall and secreted into the culture filtrate
Our topology studies revealed that the C-terminal domain of PE is located outside of the cytoplasm. To examine whether PE remains attached to the plasma membrane or is released into the bacterial cell wall, we performed a comprehensive In silico analyses predict that PE (UniProtKB entry Q2M5K2, also referenced as "PE family protein" (UniProtKB entry I7FDB0)) and Chp2 (UniProtKB entry O50440) possess a putative N-terminal signal peptide (dark blue boxes) with an AXA signal peptidase cleavage motif (boldface type) (14), whereas Chp1 (UniProtKB entry O07801) contains a putative transmembrane helix (dark blue box) with no signal peptide cleavage site. Predicted cleavage sites in PE and Chp2 are indicated (black arrows). The three proteins harbor a conserved C-terminal domain (green boxes). Three-dimensional structure modeling of these domains revealed an a/b-hydrolase fold with an SDH catalytic triad (shown in red) and a conserved GXSX(G/S) pentapeptide motif sequence surrounding the serine residue, characteristic of serine hydrolase proteins (10). Predictions of transmembrane topology and signal peptides were performed using Phobius (RRID:SCR_015643) and SignalP 4.1 (RRID:SCR_015644).
proteomic analysis of the subcellular fractions of M. smegmatis. The proteomes of four independent cultures of M. smegmatis were split in four subcellular fractions: plasma membrane, culture filtrate (secretome), mycomembrane-containing cell-wall (MMCW), and soluble proteins from cytosol and periplasm. Plasma membrane and MMCW were isolated from the cell lysates by differential ultracentrifugation on a sucrose gradient, as described (18). The homogeneity of both membrane fractions has been demonstrated previously, using biochemical markers: arabinose and galactose (arabinogalactan), glucosamine, muramic acid, and diaminopimelic acid (peptidoglycan) as markers of MMCW and NADH oxidase and ATP synthase as markers of the plasma membrane. Only traces of MMCW markers were detected in the plasma membrane fraction, and conversely, very little NADH oxidase activity (3.6% of that measured in the plasma membrane) and no ATP synthase was detected in the MMCW fraction (18).
After concentration on SDS-PAGE, proteins from each fraction were digested in gel using trypsin. Peptides were then extracted from the gel and analyzed in quadruplicate injections by nano-LC-MS/MS using an Orbitrap Fusion TM Tribrid TM mass spectrometer. MS data were searched against the Uniprot M. smegmatis mc 2 155 database for protein identification, and the quantitative comparison of relative protein abundances was performed using MS1-based label-free quantification. This data set contains 3,510 proteins and their relative quantities in the different fractions. We first confirmed the presence of proteins of known subcellular localization in the different fractions and their clustering based on MS signal (see Fig. S2 and Table  S2 for the list of marker proteins). The PE protein was not detected in the cytosol or in the plasma membrane (Fig. 3A). It was detected with a very good sequence coverage in the MMCW (Fig. 3B). Although its signal was less intense, two specific peptides of PE were identified in the culture filtrate (see Fig. S3 for annotated spectra). Of note, the quantitative data relative to the subcellular specific proteomes generated in this study are available in Table S3. These data constitute an informative resource on the subcellular distribution of M. smegmatis proteins.

PE has a cleavable N-terminal signal sequence
As mentioned, PE possesses a putative signal sequence with a cleavage site. Therefore, we next examined whether the signal sequence is cleaved and whether the mature protein is exported to the MMCW using biochemical approaches. An HA-tagged PE protein (PE-HA) was expressed in the PMM229 mutant (DpE) of M. smegmatis for estimating its relative molecular weight in cells, based on SDS-PAGE migration and Western blotting analysis. Control strains producing either a mutated form of PE (PE DCS -HA), in which a stretch of amino acids (Ala 26 -Asp 29 ) surrounding the putative cleavage site has been deleted to interfere with signal peptide removal, or the predicted mature form of PE (mPE-HA, residues 29-383) were used to distinguish between cleaved and uncleaved forms of PE (Fig. 4A). Two protein bands that migrate at positions corresponding to mPE-HA and to the noncleavable PE DCS -HA variant, respectively, were detected in the cell extract of M. smegmatis expressing PE-HA, indicating that the PE protein underwent partial cleavage in this strain (Fig. 4A).
To further establish that PE contains a cleavable signal sequence, its N terminus region with or without the AAA motif and its local peptide environment was fused to the mature form of the Escherichia coli b-lactamase BlaTEM-1 (mBlaT, residues 24-286), generating fusion proteins PE 1-25 -mBlaT and PE 1-30 -mBlaT (Fig. 4B). BlaTEM-1 has been used before as a reporter for protein export with Sec and Tat substrates (19). PE 1-30 -mBlaT and PE 1-25 -mBlaT were expressed in PMM299, a M. smegmatis mutant lacking the major b-lactamase BlaS (see Fig.  S4), and bacteria were tested for susceptibility toward ampicillin using a disk diffusion assay and the broth dilution method for MIC determination. Production of PE 1-30 -mBlaT in PMM299 restored resistance to ampicillin beyond the WT level (more than 64-fold and 2-fold increases in MIC relative to the PMM299 mutant and the parent strain, respectively). In contrast, production of PE 1-25 -mBlaT conferred only moderate resistance to ampicillin on both solid and liquid media (2-4-fold increase in MIC compared with the PMM299 mutant) (Fig. 4B).
To confirm that the difference we observed between the two proteins reflected a real difference in protein cleavage, we performed Western blotting analysis with an anti-BlaTEM-1 antibody on cell lysates prepared from PMM299 expressing either PE 1-30 -mBlaT or PE 1-25 -mBlaT. We found that the two proteins were produced in similar quantities in bacteria (Fig. 4C). In accordance with their calculated molecular weights (CMW), PE 1-25 -mBlaT (CMW 31.8 kDa) migrated as a single band pro-tein with an apparent molecular mass slightly higher than that of the full-length b-lactamase BlaTEM-1 (CMW 31.6 kDa), confirming that the protein was not processed in PMM299 (Fig. 4C). By contrast, two bands of similar intensity were visible for PE 1-30 -mBlaT. Their migration on SDS-PAGE was consistent with the CMW of the full-length unprocessed form of the protein (32.3 kDa) and of the mature form of BlaTEM-1 with an additional DDKL amino acid sequence (DDKL-mBlaT, 29.4 kDa) (Fig. 4C). It can thus be concluded that PE 1-25 -mBlaT remains attached to the plasma membrane due to the lack of signal sequence cleavage, causing a moderate increase in antibiotic resistance, whereas PE 1-30 -mBlaT undergoes proteolytic processing, leading to mBlaT excretion in the cell envelope and high-level ampicillin resistance. Consistently, expression of BlaTEM-1 proteins with an altered cleavage signal sequence in E. coli confers intermediate-level resistance to ampicillin compared with the WT form of BlaTEM-1 (20). Altogether, data obtained with the BlaTEM-1 reporter protein, in addition to those obtained with the HA-tagged fusion proteins, confirmed that PE is excreted in the periplasmic compartment or the MMCW of M. smegmatis following signal peptide cleavage.

Cleavage of PE is required for efficient production of TPP
We finally explored whether formation of TPP in M. smegmatis depends on proteolytic cleavage of PE. Strain PMM229 (DpE) expressing either PE-HA or PE DCS -HA (Fig. 4A), as well as the WT and the PMM229 mutant strains, were grown to mid-exponential phase. The cellular and surface-exposed lipid fractions were extracted from each strain and separated by TLC. As reported previously, disruption of pE impaired TPP production and resulted in accumulation of a relatively polar compound that corresponds to the diacyl trehalose precursor of TPP, in both the cellular and surface compartments (5)   5A). Production of PE-HA in PMM229 restored the WT phenotype. In this complemented strain, TPP were found in the cellular and surface-exposed lipid fractions, and the intermediate product was undetectable in these lipid fractions. By contrast, PMM229 cells expressing the PE DCS -HA protein were defective in TPP production and accumulated the precursor in the cellular and surface compartments. These data demonstrate that the PE DCS -HA protein is unable to catalyze TPP formation, likely because the enzyme is anchored to the plasma membrane and has limited access to its substrate that is mainly located in the outer layers of the cell envelope. We cannot totally rule out that the small deletion present in PE DCS -HA may affect protein stability. However, amino acid residues surrounding the signal peptidase cleavage motif are located away from the predicted serine hydrolase core domain of PE (residues 90-324) and are probably not required for the correct folding of this domain. Supporting this, it has been shown that the Ala 30 -Gly 359 domain of Chp2, which starts 5 residues after the putative AXA cleavage motif (Fig. 1), displays acyltransferase activity in vitro (12).

Discussion
In a previous study, we proposed a model for the biosynthesis of TPP, but some steps of this biosynthetic pathway remain poorly characterized, including the one involving the acyltransferase PE (5). Here, we shed new light on the molecular mechanisms that underlie the latter stages of TPP production in M. smegmatis. Based on a series of genetic, biochemical, and proteomic approaches, we established that (i) the C-terminal domain of PE, and therefore its catalytic domain, resides outside the plasma membrane, (ii) PE is cleaved and excreted in the MMCW, and (iii) proteolytic processing of PE is required for efficient production of TPP in bacteria. These data, in  combination with our previous findings, allow us to propose an updated version of the TPP biosynthetic pathway (Fig. 6): the 2,3-diacyltrehalose precursor is synthetized through a series of cytoplasmic steps catalyzed by enzymes encoded by the TPP locus (FadD23, Pks, and PapA3) (5) and is translocated across the plasma membrane by the MmpL10 transporter. The intermediate molecule is then exported to the cell surface, where PE transacylates phleic acids between precursors to yield TPP. Several lines of evidence support this scenario. First, a mmpL10 mutant of M. smegmatis is completely devoid of TPP, indicating that the transacylase reaction depends on the translocation process and likely occurs after this step (5). Second, accumulation of the intermediate product in the surface-exposed lipid fraction of the PMM229 (DpE) mutant indicates that this compound is translocated through the plasma membrane by MmpL10 and transferred to the cell surface. Third, we established that only the mature form of PE catalyzes TPP formation, supporting the notion that transacylation takes place in the cell envelope. Presumably, the full-length unprocessed PE protein is unable to perform this reaction because it has a restricted access to its substrate in the neighborhood of the plasma membrane.
In our Western blotting analyses, we observed partial cleavage of PE, likely due to a relative overexpression of the protein in the PMM229 mutant compared with the WT expression level. To visualize PE, we indeed used a PMM229 strain in which expression of pE was under the control of the strong promoter pBlaF* (21) because we failed to detect protein production in cells carrying the pE gene under the control of its native promoter. Note that cleavage efficiency observed in our Western blotting experiments might be underestimated, due to a possible partial secretion of the processed form of PE, as evidenced by our proteomic studies. These studies, which identified PE in the MMCW but not in the plasma membrane of M. smegmatis, lend support to the idea that the protein exists primarily in a mature active form in the bacterium, when produced at physiological levels. BLAST searches in the Gen-Bank TM database revealed that at least 100 mycobacterial species harbor the TPP locus and can potentially synthetize this family of glycolipids (Table S4). Interestingly, the presence of a predicted N-terminal signal sequence with a conserved AXA motif in the PE orthologs putatively produced by species containing the TPP locus indicates that secretion of these enzymes is a general process in TPP producers (Table S4).
The basic mechanisms underlying synthesis and export of trehalose-containing lipids are evolutionarily conserved in mycobacteria (8). Notably, it has been shown that two transmembrane acyltransferases, namely Chp1 and Chp2, sharing similar domain organization with PE, participate in SL and PAT assembly in M. tuberculosis by catalyzing sequential transesterification between SL precursors and DAT molecules, respectively. Despite similar functions in the formation of trehalose glycolipids, these acyltransferases exhibit strict substrate specificity because Chp1 and Chp2 are not functionally interchangeable for PAT and SL formation (13) and neither of the two enzymes can restore TPP production in the PMM229 mutant of M. smegmatis (Fig. 5B). The C-terminal end of Chp1 was shown to reside in the cytosol, indicating that SL are built up intracellularly and exported to the cell envelope (11). The orientation of Chp2 in the plasma membrane is controversial because two studies demonstrated opposite orientation (12,13). We were unable to confirm the topological organization of Chp1 using the reporter proteins GFP and PhoA in M. smegmatis because both Chp1-GFP and Chp1-PhoA were found to lack activity (data not shown). To demonstrate the cytosolic localization of the C-terminal domain of Chp1, Seeliger et al. (11) used the b-galactosidase reporter instead of the GFP. The absence of fluorescence with the Chp1-GFP fusion in our studies may result from protein misfolding. This hypothesis is nevertheless difficult to test experimentally in M. smegmatis because this bacterium does not produce SL and, as mentioned above, Chp1 cannot restore TPP production in a mutant strain lacking PE. Here, we provide evidence that the catalytic domain of Chp2 has a periplasmic localization when expressed in M. smegmatis. Although we failed to provide clear evidence that Chp2 is cleaved in this bacterium, the presence of an AXA motif at the end of the signal sequence strongly suggests that the protein might be converted to a mature form that is excreted in the periplasmic space in M. tuberculosis. In that situation, the scenario for DAT/PAT production would be similar to that of TPP in M. smegmatis: DAT would be synthesized in the cytosol and transferred to the bacterial cell surface to serve as a substrate for Chp2.
Despite significant progress made in understanding the mechanisms of glycolipid formation in mycobacteria, the molecular machinery involved in their transport across the cell envelope remains to be identified. For instance, it is not known whether MmpL10 cooperates with other proteins to promote TPP export. Using a GFP-trap pulldown approach, Fay et al. (22) recently identified two transmembrane proteins interacting with the MmpL3 transporter during mycolic acid transport in both M. smegmatis and M. tuberculosis, but no protein was found to bind MmpL10 in a parallel control experiment. Recent studies pointed out the role of lipoproteins in the export of several families of lipids in M. tuberculosis, but, to date, none of them has been shown to be involved in the transport of trehalose-containing glycolipids (23). The location of PE in the cell envelope makes this protein a suitable candidate as a bait for identifying interacting partners putatively involved in TPP transport. Delineation of the basic mechanisms that drive TPP transport in the mycobacterial model organism M. smegmatis may serve as a guide to describe export of glycolipids in M. tuberculosis.

Experimental procedures
Bacterial strains, growth media, and culture conditions Plasmids were propagated at 37°C in E. coli DH5a in LB broth or LB agar (Invitrogen) supplemented with either kanamycin (Km) (40 mg/ml) or hygromycin (Hyg) (200 mg/ml). M. smegmatis mc 2 155 WT and derivatives (Table S5) were grown at 37°C in LB broth or in Middlebrook 7H9 broth (DB Difco) containing ADC (0.2% dextrose, 0.5% BSA fraction V, 0.0003% beef catalase) and 0.05% Tween 80 when necessary and on LB agar. When required, Km and Hyg were used at a concentration of 40 and 50 mg/ml, respectively. For subcellular fractionation and MS analysis, M. smegmatis mc 2 155 was grown in 7H9 Middlebrook broth medium supplemented with 0.2% glycerol (Sigma) at 37°C with shaking at 180 rpm.

Construction of the M. smegmatis blaS-disrupted mutant
The M. smegmatis blaS mutant (PMM299) was constructed by homologous recombination using pWM290, a derivative of the mycobacterial suicide plasmid pJQ200 harboring the counterselectable marker sacB (24). Briefly, two DNA fragments encompassing the regions located upstream and downstream of the blaS gene (MSMEG_2658) (25) were amplified by PCR from M. smegmatis mc 2 155 total DNA and inserted, flanking a res-km-res resistance cassette, into pWM290. The resulting plasmid (pWMbla) was transferred by electroporation into M. smegmatis mc 2 155 for allelic exchange, and transformants were selected on plates containing Km and sucrose. Several Km-and sucrose-resistant colonies were screened by PCR with different primer combinations. For further details, see Fig. S4.

Construction of protein expression vectors
Plasmids used in this study are indicated in Table S1. Details of plasmid constructions are described in the supporting information.
Fluorescence and alkaline phosphatase-based assays for topology determination of PE The subcellular location of the C-terminal region of PE was investigated using the topology reporter proteins PhoA and GFP (15,16). To measure the fluorescence intensities of strains expressing the GFP fusion proteins, bacteria were grown in LB containing 0.05% Tween 80 to logarithmic phase. Bacterial concentrations were determined by OD 600 measurements, and 500 ml of each culture were centrifuged for 1 min at 12,000 rpm. Pellets were resuspended in 500 ml of PBS containing 0.05% Tween 80, and 100 ml of each bacterial suspension were transferred to 96-well plates in triplicate. Fluorescence intensities were measured using a CLARIOstar microplate reader (BMG Labtech) at excitation and emission wavelengths of 470 and 515 nm, respectively. PhoA activities of M. smegmatis strains producing PhoA fusions were assessed by streaking bacteria on LB agar plates supplemented with 60 mg/ml BCIP. Plates were incubated at 37°C for 4 days. PhoA activities were also measured on cells grown in liquid medium using the chromogenic substrate pNPP (New England Biolabs). Bacteria were cultivated in LB at 37°C, and growth was followed by OD 600 measurements. Bacteria were then washed twice with 1 M Tris, pH 8.0, containing 0.05% Tween 80, and the OD 600 was adjusted to 2 with this buffer. 200 ml of cell suspension were mixed with 1 ml of buffer containing 10 mM pNPP and 10 mM MgCl 2 and incubated for 90 min at 37°C in the dark. The reaction was stopped by adding 0.2 ml of 1 M K 2 HPO 4 solution (Sigma). Tubes were centrifuged for 10 min at 13,000 rpm to remove aggregates, and 1 ml of supernatant was used to measure absorbance at 405 nm.

Antimicrobial susceptibility testing
Ampicillin susceptibilities of M. smegmatis strains expressing the BlaTEM fusion proteins were assessed using the disk diffusion method and by MIC measurements. For the disk diffusion assay, cultures were grown to mid-exponential phase (OD 600 = 0.3-0.5) in 7H9 containing 0.05% Tween 80. Cells were washed once with fresh medium and resuspended in an equal volume of 7H9. 200 ml of cell suspension were mixed with 5 ml of molten top agar (0.6% agar, 0.2% glycerol (v/v)) and poured onto 7H11 plates supplemented with oleic acid/albumin/dextrose/catalase. Homemade discs containing 100 or 200 mg of ampicillin were placed on top, and plates were incubated for 3 days at 37°C. MIC were determined by the broth microdilution method. Cultures were harvested at mid-exponential phase in 7H9 containing 0.05% Tween 80, and the OD 600 values were adjusted to 0.3 with fresh medium. 10 ml of each culture were used to inoculate 1 ml of 7H9 supplemented with 0.05% Tween 80 containing 2-fold serially diluted ampicillin (from 512 to 1 mg/ml) in a 24-well plate (Nunc). Plates were incubated for 3 days at 37°C with shaking at 150 rpm. The MIC corresponds to the lowest concentration of antibiotic inhibiting visible growth of bacteria.

Detection of proteins in bacteria by Western blotting analyses
To prepare whole-cell lysates of M. smegmatis, bacteria were grown in 50 ml of 7H9, collected by centrifugation, and resuspended in 500 ml of lysis buffer (20 mM Tris-HCl, pH 7.4). 500 ml of 0.1-mm glass beads (Biospec) were added to the cell suspensions, and bacteria were subjected to bead beating (Retsch Mixer Mill) for two cycles of 1 min with 2 min on ice between cycles. Tubes were centrifuged briefly to pellet beads and collect the supernatant fractions. Alternatively, 100-400 ml of bacterial cultures were collected and resuspended in 10 ml of lysis buffer. Cells were lysed by two passages through a French press, at 1,500 bars. For both methods, cell lysates were subjected to centrifugation at 3,000 3 g for 10 min to remove nonlysed cells. To perform Western blotting analyses, proteins from whole-cell lysates were separated by SDS-PAGE and transferred onto nitrocellulose membranes (Pall Corp.). Membranes were incubated in 1 3 PBS containing 0.1% Tween 20 and 10% nonfat dry milk and then incubated for 1 h with either a mouse monoclonal anti-HA antibody (clone HA-7, Sigma; 1:10,000) or a mouse monoclonal anti-BlaTEM-1 antibody (clone MA1-20370, Invitrogen; 1:400) in PBS containing 0.1% Tween 20 and 1% dry milk. Membranes were washed three times for 5 min with PBS containing 0.1% Tween 20 and incubated for 1 h with a horseradish peroxidase-conjugated goat anti-mouse IgG antibody (Sigma) at a dilution of 1:3,000 in PBS containing 0.1% Tween 20 and 1% nonfat dry milk. Protein bands were visualized using the EMD Millipore Immobilon enhanced chemiluminescence kit and a ChemiDoc Touch imaging system (Bio-Rad).

Extraction of mycobacterial lipids and TLC analyses
Bacteria were cultivated in LB medium at 37°C and recovered by centrifugation (3,000 3 g, 10 min). Total lipid fractions were extracted from bacterial pellets with chloroform/methanol (1:2) and chloroform/methanol (2:1), washed with water, and dried before analysis. Cellular and surface-exposed lipids were prepared as described previously (5,26). Mycobacterial cell pellets were shaken for 1 min with 10 g of glass beads (4mm diameter) per 2 g (wet weight) of cells and resuspended in distilled water. Extracellular materials were separated from cells by centrifugation (3,000 3 g, 10 min). Cellular lipid fractions were then extracted from pellets with chloroform/methanol (1:2) and chloroform/methanol (2:1), washed with water, and dried before analysis. Surface-exposed lipids were extracted from supernatants with chloroform and methanol according to the Bligh and Dyer procedure (27). After drying, the organic phases were washed. For TLC analyses, equivalent amounts of lipids from each strain were spotted on silica gel G60 plates (20 3 20 cm; Merck) run in CHCl 3 /CH 3 OH (90:10, v/v). The spots were visualized by spraying the plates with a 0.2% anthrone solution (w/v) in concentrated H 2 SO 4 , followed by heating.
Bacteria lysis and fractionation on sucrose density gradient for proteomics by MS M. smegmatis cells were fractionated as described previously (18). Briefly, exponentially growing bacteria were harvested by centrifugation at 3,000 3 g for 15 min at 4°C. The supernatant was filtrated through a 0.2-mm sterile Nalgene filter to yield the culture filtrate, and proteins were concentrated using a 3,000-Da molecular weight cutoff Vivaspin. The pellet was suspended in lysis buffer (20 mM Tris-HCl, pH 7.4, containing 5 ml of benzonase nuclease .250 units/ml (Sigma-Aldrich), 1 mM DTT, 2 mM 4-(2-aminoethyl)benzenesulfonyl fluoride (Euromedex), and 1 mM EDTA), and bacterial cells were broken by two passages through a French press cell (1,500 bars). Unbroken bacteria were removed by centrifugation (3,000 3 g for 10 min, twice). The bacterial lysate was then submitted to centrifugation at 10,000 3 g for 40 min to yield a crude MMCW fraction (pellet P10) and S10 supernatant. P10 was suspended in 20 mM Tris-HCl, pH 7.4, containing EDTA (1 mM), layered on a sucrose step gradient (from 10% (w/w) to 60% (w/w)), and then centrifuged for at least 2 h at 100,000 3 g. The MMCW fraction was recovered at the 30-36% sucrose cushion interface and washed twice with Tris-HCl buffer (10,000 3 g, 1 h). The S10 supernatant was centrifuged at 27,000 3 g for 30 min, and the resulting supernatant (S27) was further separated into supernatant (S100 fraction) and pellet (P100 fraction) by centrifugation at 100,000 3 g for 40 min. S100 yielded the cytosolic fraction. The P100 was suspended in 20 mM Tris-HCl, pH 7.4, containing 1 mM EDTA before being layered on a sucrose step gradient (from 10% (w/w) to 60% (w/w)) and centrifuged for at least 2 h at 100,000 3 g. Plasma membranes were recovered at the 20-30% sucrose cushion interface and washed twice with Tris-HCl buffer (100,000 3 g, 1 h).

Protein digestion and nano-LC-MS/MS analysis
Subcellular fractions were analyzed on an Orbitrap Fusion TM Tribrid TM mass spectrometer (Thermo Fisher Scientific), as described previously (18) but with slight modifications (see supporting information).

Bioinformatic MS data analysis
The Mascot (Mascot server version 2.6.2) database search engine was used for peptide and protein identification using automatic decoy database search to calculate a false discovery rate. MS/MS spectra were compared with the UniProt M. smegmatis strain ATCC700084/mc 2 155 database (release June 2017, 12,683 sequences). See supporting information for more details about the data analysis.

Data availability
All the R scripts associated to this study are freely available with the Proline export tables on Zenodo (3686899). The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (28) partner repository with the data set identifier PXD017602.