Inactivation of tesA Reduces Cell Wall Lipid Production and Increases Drug Susceptibility in Mycobacteria

Abstract

Phthiocerol dimycocerosates (PDIMs) and phenolic glycolipids (PGLs) are structurally related lipids noncovalently bound to the outer cell wall layer of Mycobacterium tuberculosis, Mycobacterium leprae, and several opportunistic mycobacterial human pathogens. PDIMs and PGLs are important effectors of virulence. Elucidation of the biosynthesis of these complex lipids will not only expand our understanding of mycobacterial cell wall biosynthesis, but it may also illuminate potential routes to novel therapeutics against mycobacterial infections. We report the construction of an in-frame deletion mutant of tesA (encoding a type II thioesterase) in the opportunistic human pathogen Mycobacterium marinum and the characterization of this mutant and its corresponding complemented strain control in terms of PDIM and PGL production. The growth and antibiotic susceptibility of these strains were also probed and compared with the parental wild-type strain. We show that deletion of tesA leads to a mutant that produces only traces of PDIMs and PGLs, has a slight growth yield increase and displays a substantial hypersusceptibility to several antibiotics. We also provide a robust model for the three-dimensional structure of M. marinum TesA (TesAmm) and demonstrate that a Ser-to-Ala substitution in the predicted catalytic Ser of TesAmm renders a mutant that recapitulates the phenotype of the tesA deletion mutant. Overall, our studies demonstrate a critical role for tesA in mycobacterial biology, advance our understanding of the biosynthesis of an important group of polyketide synthase-derived mycobacterial lipids, and suggest that drugs aimed at blocking PDIM and/or PGL production might synergize with antibiotic therapy in the control of mycobacterial infections.

Introduction

Mycobacterium tuberculosis (Mtb),³ Mycobacterium leprae, and several opportunistic mycobacterial human pathogens (e.g. M. marinum (Mm)) produce two related groups of diesters of β-glycol-containing aliphatic polyketides (e.g. phenolphthiocerols and phthiocerols) and polyketide synthase-derived multimethyl-branched fatty acids (e.g. mycocerosic acids) (Fig. 1). One of these groups is represented by phthiocerol dimycocerosates (PDIMs). The other group is represented by phenolphthiocerol dimycocerosates, which are glycosylated compounds generally known as phenolic glycolipids (PGLs). These complex lipids, which are believed to be noncovalently bound constituents of the outer leaflet of the unique mycobacterial outer membrane, are known important effectors of virulence (for review, see Ref. 1).

FIGURE 1.

Mycobacterial PDIMs and PGLs. Structural variability is strain- and species-specific. The variability shown is representative of that found in M. marinum (Ref. 1 and references therein). m and m′, 16–20; n, 16–22; p, 14–22.

Cox et al. (3) and Camacho et al. (2) independently demonstrated in 1999 that loss of PDIMs in PGL-deficient Mtb strains correlates with attenuation in animal models of tuberculosis. It has also been documented that production of PGLs confers a hyperlethality phenotype to PDIM-producing Mtb in murine disease models (4). Since these seminal reports, an overwhelming body of evidence has accumulated demonstrating that PDIMs and PGLs play key roles in virulence and host-pathogen interaction via multiple mechanisms in various mycobacterial pathogens (4–15). Moreover, PGL production has been suggested as a trait predisposing Mtb strains of the W-Beijing family to their characteristic epidemic spread and increased likelihood of developing drug resistance (16). Thus, elucidation of PDIM and PGL biosynthesis will not only expand our understanding of cell wall biosynthesis in clinically relevant mycobacteria, but it may also illuminate potential routes to alternative therapeutics against mycobacterial infections (17, 18). These considerations have guided our previous studies on PDIM and PGL biosynthesis (17, 19–22), which include the development of the first PGL biosynthesis inhibitor (17). The inhibitor works in a manner analogous to that of the first reported inhibitor of siderophore (microbial iron chelator) biosynthesis (18, 23), and it blocks PGL production in Mtb and other mycobacterial pathogens (17).

Despite significant progress made toward understanding PDIM and PGL biosynthesis (Ref. 1 and references therein), many aspects of this pathway remain unclear. One such aspect is whether tesA (encoding a putative type II thioesterase (24)) is required for PDIM and/or PGL production. An ortholog of tesA is clustered with genes with confirmed or suspected involvement in PDIM and/or PGL production in each PDIM/PGL producer for which the genome has been analyzed (1). Moreover, Mtb TesA (TesAmt) has been shown (by two-hybrid system and pulldown analyses) to interact with Mtb PpsE (24), the terminal polyketide synthase in the synthesis of the phthiocerol and phenolphthiocerol constituents of PDIMs and PGLs, respectively (1). These observations have led to the hypothesis that TesA is a thioesterase required for release of polyketide products from PpsE (24).

Two Mtb tesA mutants, each with a transposon-based disruption of tesA, have been reported to be PDIM-deficient (25, 26). Unfortunately, the possibility of polar effect on genes located downstream of tesA and needed for PDIM production has not been ruled out in either of these insertional mutants, thus precluding a conclusive determination as to the requirement of tesA for PDIM production. Moreover, Mtb is known to spontaneously and irreversibly lose its PDIM production capacity with high frequency during propagation in vitro (27, 28), and the possibility that the PDIM deficiency of the transposon mutants is due to spontaneous loss cannot be conclusively dismissed either. Last, because the parental Mtb H37Rv and Erdman strains from which the transposon mutants were derived are naturally PGL-deficient (29), neither mutant provides information regarding PGL production either. The PGL deficiency of these strains arises from a natural mutation in a gene encoding a polyketide synthase that assembles biosynthetic intermediates for PGL production (22). Thus, the relevance of tesA for PDIM production has not been conclusively established, and the relevance of tesA for PGL production remains unexplored. Our analysis reported herein conclusively demonstrates that deletion of tesA leads to a drastic reduction in PDIM and PGL production and suggests that drugs aimed at blocking PDIM and/or PGL production might synergize with antibiotic therapy in the control of mycobacterial infections.

EXPERIMENTAL PROCEDURES

Routine Culturing and Recombinant DNA Manipulations

Mm (strain M; ATCC BAA-535) was cultured at 30 °C in Middlebrook 7H9 (Difco) supplemented with 10% ADN (5% BSA, 2% dextrose, 0.85% NaCl) (Difco) and 0.05% Tween 80 (7H9) or Middlebrook 7H11 (Difco) with ADN (7H11) (30). Escherichia coli DH5α (Invitrogen) was grown in Luria-Bertani media under standard conditions (31). When required, kanamycin (30 μg/ml), hygromycin (50 μg/ml), sucrose (2%), and/or X-gal (70 μg/ml) was added to the media. General recombinant DNA manipulations were carried out by standard methods and using E. coli as the primary cloning host (31). PCR-generated DNA fragments used in plasmid constructions were sequenced to verify fidelity. Genomic DNA isolation and plasmid electroporation into Mm were carried out as reported (30). Supplemental Table S1 shows the plasmids and oligonucleotides used in this study.

Construction of M. marinum ΔtesA

Mm ΔtesA was engineered using the p2NIL/pGOAL19-based flexible cassette method (32). A suicide delivery vector (p2NIL-GOALc-ΔtesAc, see below) carrying a tesA deletion cassette (ΔtesAc) was used to generate Mm ΔtesA. The vector was electroporated into Mm, and transformants with potential single-crossovers (blue colonies) were selected on 7H11 containing hygromycin, kanamycin, and X-gal. Potential single-crossovers were grown in antibiotic-free 7H9 and then plated for single colonies on 7H11 containing sucrose and X-gal. White colonies that grew on the sucrose plates were re-streaked onto antibiotic-free and antibiotic-containing plates to identify clones that lost drug resistance, thus indicating a possible double-crossover event with consequent loss of tesA or wild-type (WT) generation. The tesA deletion in antibiotic sensitive clones was screened for and confirmed by PCR. To this end, genomic DNA isolated from mutant candidates was used as the template along with two independent primer pairs (tes-of and tes-or, tes-f and tes-r), each of which produced diagnostic amplicons permitting differentiation between WT and mutant.

Construction of p2NIL-GOALc-ΔtesAc

Supplemental Fig. S1 outlines the construction of the plasmid. ΔtesAc was generated using splicing by overlap extension (SOE) PCR (33). ΔtesAc contained a 5′-arm (984 bp = 954-bp region upstream of the tesA + tesA first 10 codons) and a 3′ arm (993 bp = the tesA last 15 codons + stop codon + 945-bp downstream segment). Each arm was PCR-generated from genomic DNA. Primer pair tes-of and tes-ir and primer pair tes-or and tes-if were used to generate the 5′ and 3′ arms, respectively. The arms were then used as templates for PCR with primers tes-of and tes-or to fuse the arms and generate ΔtesAc (1977 bp). The PCR-generated ΔtesAc was first cloned into pCR2.1Topo (Invitrogen). ΔtesAc was subsequently excised from the pCR2.1Topo construct using HindIII and SalI, and the excerpt was ligated to p2NIL (32) linearized by HindIII-SalI digestion. The resulting p2NIL-ΔtesAc plasmid and pGOAL19 (32) were digested with PacI, and the PacI cassette (GOALc, 7939 bp) of pGOAL19 was ligated to the linearized p2NIL-ΔtesAc to create p2NIL-GOALc-ΔtesAc.

Construction of pCP0-tesAmm and pCP0-tesAmmS92A

Plasmid pCP0-tesAmm was constructed as follows. Mm tesA (tesAmm, 795 bp) was PCR-amplified from genomic DNA using primer pair tes-f and tes-r. The PCR product was cloned into pCR2.1Topo. Insert tesAmm was recovered from the pCR2.1Topo construct as a HindIII-NheI fragment, and the excerpt was subcloned into pCP0 (19) linearized by HindIII-NheI digestion. To construct pCP0-tesAmmS92A, the 300-bp 5′-end fragment of the tesAmm insert of pCR2.1Topo-tesAmm was amplified with primers tes-f and tes-s2a (S92A mutation). The mutated fragment and pCR2.1TOPO-tesAmm were both digested with HindIII and BsmI, and the mutated fragment was then ligated to the plasmid backbone to replace the WT 5′ end of tesAmm and generate pCR2.1Topo-tesAmmS92A. The tesAmmS92A insert of pCR2.1Topo-tesAmmS92A was recovered from the plasmid as a HindIII-NheI fragment and subcloned into pCP0 linearized with HindIII and NheI to create pCP0-tesAmmS92A. The tesA genes were placed under the control of the hsp60 promoter of pCP0 for expression in mycobacteria.

PDIM and PGL Analysis

Five-day-old cultures were diluted to an A₅₉₅ of 0.6 in 7H9 and loaded into 24-well plates (1 ml/well). The plates were incubated for 24 h (170 rpm, 30 °C). After incubation, the A₅₉₅ of the cultures was measured in a DTX 880 Plate Reader (Beckman Coulter, Inc.), and the cells were harvested for PDIM and PGL extraction using a biphasic mixture of methanolic saline and petroleum ether as reported (17, 19, 20). PDIM and PGL were analyzed by MS. Mass spectral data were collected at the CUNY Mass Spectrometry Facility at Hunter College on an Agilent Technologies G6520A high resolution quadrupole-TOF mass spectrometer attached to an Agilent Technologies 1200 Capillary HPLC system. Samples were ionized by electrospray ionization in positive mode. Chromatography was performed on an Agilent Technologies Poroshell 120 EC-C8 2.1-mm × 75-mm, 2.7-μm column (Part no. 697775-906) using 2-propanol (Solvent A) and methanol containing 0.1% formic acid and 50 μm ammonium formate (Solvent B) at a flow rate of 400 μl/min. The HPLC gradient was 95% B for 0–6 min and 95–80% B for 6–45 min. Total analysis time was 45 min. The HPLC flow was diverted to waste for the first 6 min. The temperature of the column was held at 45 °C for the entire analysis. Mass spectrometer parameters for the MS (non-MS/MS) data presented were as follows: fragmentor = 120 V, drying gas temperature = 200 °C, drying gas flow = 11 liters/min, nebulizer pressure = 30 p.s.i., capillary voltage = 3500 V. Data were collected with the instrument set to low mass range (100–1700 m/z)-extended dynamic range conditions (2 GHz mode), and data were stored (6–45 min) as both centroid and profile with a threshold of 10 counts for MS mode. MS spectra were collected over a range of 400–1700 m/z at 2 spectra/s. The reference mass used, HP-922 with M+H⁺ ion at 922.009798 m/z, was infused into the spray chamber using the Agilent calibrant delivery system. The instrument was controlled with Agilent MassHunter Work station Acquisition Software B.03.01, and data were analyzed using Agilent MassHunter Work station Qualitative Analysis Software B.04.00.

Growth Curves

Cultures grown to exponential phase (A₅₉₅ = 0.6) in 7H9 were diluted in fresh medium to an A₅₉₅ of 0.005 and loaded into 96-well plates (200 μl/well). The growth medium used for strains carrying pCP0 or its derivatives contained kanamycin. Plates were incubated with shaking (170 rpm) at 30 °C, and growth was assessed as A₅₉₅ using a plate reader at time points indicated under “Results.”

Antibiotic Susceptibility Assays

Dose-response experiments using standard 2/3-fold-microdilution assays in a 96-well plate platform were performed as reported (34). Cultures started at an A₅₉₅ of 0.005 were grown in 7H9 containing the antibiotic being tested and kanamycin (for plasmid maintenance). Test antibiotics were added from DMSO solution (rifampicin) or aqueous solution (all others). In the former case, DMSO was kept at 0.5% in antibiotic-treated cultures and DMSO controls. Plates were incubated with shaking (170 rpm) at 30 °C for 8 days. After incubation, growth was assessed as A₅₉₅ using a plate reader. IC₅₀ values were calculated from sigmoidal curves (23) fitted to triplicate dose-response data sets using KaleidaGraph (Synergy Software). MIC₉₀ values were determined as the lowest antibiotic concentration tested that inhibited growth by ≥90% relative to controls. Standard disk-diffusion assays were carried out as reported (35). Exponentially growing cultures (A₅₉₅ = 0.6) in 7H9 were diluted in fresh medium to an A₅₉₅ of 0.2, and 100 μl of diluted culture were used to seed 7H11 plates (25 ml agar/plate). Antibiotic disks were placed onto the inoculated agar, and the plates were incubated at 30 °C for 15 days before analysis.

Structure Modeling

Three-dimensional models for TesAmm and TesAmmS92A were constructed based on homology modeling techniques using several software tools. The majority of fold recognition and sequence alignment algorithms (supplemental Table S2) identified the structure of the type II thioesterase RifR (Protein Data Bank code 3FLA) (36) as a high confidence structural representation of TesAmm sequence and was, therefore, used as a structural template in modeling. Alternative models based on other identified close structural templates (Protein Data Bank code 1JMK (37) and 2RON (38)) were also constructed and analyzed to confirm for robustness of structural fold and consistency of biophysical properties. The sequence alignments between TesAmm or TesAmmS92A and the structural template, RifR, were manually edited and constructed by combining (a) the results of alignment algorithms (supplemental Table S2), (b) threading analysis (supplemental Table S2), and (c) alignment of predicted (for the target sequence (supplemental Table S2)) and known (for the template sequence (36)) secondary structure elements. Homology models were constructed by overlaying the target sequence on the template structure according to the optimized sequence alignment using the program Modeller (39). The best representative model was chosen based on the fitness scores, z-scores, and knowledge-based energy plots obtained in the structure evaluation programs Verify 3D (40) and Prosa-web (41). The TesAmm model was superimposed with the structure of RifR and TesAmmS92A using the program CE (42) for comparing their structural fold. The three-dimensional model of TesAmm was visualized and analyzed using PyMol (43).

RESULTS

The TesA Orthologs Are Highly Conserved

We previously reported an analysis of the organization of the genes in the chromosomal locus involved in PDIM and PGL production of several species (1), yet the sequence similarity of TesA orthologs, each encoded at the 5′-end of its cognate PDIM/PGL biosynthetic gene cluster, has not been highlighted. We examined herein the sequence relatedness of the TesA orthologs from several PDIM/PGL producers. The orthologs have 243–269 amino acids and remarkable sequence identity (69–100%) (supplemental Fig. S2). Two paralogs of TesAmm were included in this analysis (supplemental Fig. S2). One of these paralogs, MMAR_3693 (240 amino acids, 31% identity with TesAmm), is located in the gene cluster involved in the synthesis of mycobactin siderophores (44). The second paralog, MMAR_0367 (252 amino acids, 28% identity with TesAmm), forms a predicted operon with MMAR_0368, a nonribosomal peptide synthetase gene of unknown function. The sequence identity of all the aligned homologs extends to a predicted Ser-Asp-His catalytic triad (supplemental Fig. S2). The triad is characteristic of the α/β-hydrolase superfamily (45), and we identified it in TesA proteins by sequence alignment with characterized type II thioesterases (not shown) and structural modeling of TesAmm (see below) based on the structure of the type II thioesterase RifR (36) (259 amino acids, 30% identity with TesAmm, supplemental Fig. S2).

Model of the Three-dimensional Structure of TesAmm

We have constructed a model structure of TesAmm by means of homology modeling to gain better insight into its catalytic properties and function (Fig. 2). Fold recognition and sequence alignment algorithms identified the structures of several thioesterases (36–38, 46) as potential templates to construct a three-dimensional structural model for the sequence of TesAmm. However, the top-ranked structural template in nearly all programs was the type II thioesterase RifR (Protein Data Bank code 3FLA (36)). Models based on RifR also evaluated significantly better than alternative potential structural templates identified by the programs (supplemental Fig. S3). The refined three-dimensional model of TesAmm predicts that the α/β-hydrolase fold found in various hydrolytic enzymes (45) is also maintained in the protein, and it closely mimics the structural details of RifR, showing a root mean square deviation of only 0.5 Å. Structural superposition of TesAmm and RifR (Figs. 2, A and B) and sequence alignment of TesAmm with other known thioesterases (Fig. 2D, supplemental Fig. S2) predict the conservation of the classic catalytic triad required of a functional enzyme at both the primary and tertiary level of protein structure. A three-dimensional model of the TesAmmS92A mutant (Fig. 2C) predicts that the S92A substitution in this protein (see below) does not disrupt the structural fold of the protein and should only affect the biochemistry of the enzyme.

FIGURE 2.

Molecular models of TesAmm and TesAmmS92A. A, structural superposition of TesAmm (marine) and RifR (gray) rendered in a schematic using the program Pymol (43). B, shown is the active site of TesAmm and RifR. The catalytic triad (Ser, Asp, and His) residues are shown as sticks in the same color as the parent protein. The inset shows a rotated and zoomed-in version of the residues for clarity. The numbering of residues is based on the sequence of TesAmm (corresponding residues in RifR are Ser-94, Asp-D200, and His228). C, shown is the active site of TesAmm and TesAmmS92A. The catalytic triad (Ser, Asp, and His) residues are shown as sticks in the same color as the parent protein (TesAmm, marine; TesAmmS92A, deep salmon). The inset shows a rotated and zoomed-in version of the residues for clarity. D, shown is the sequence alignment of TesAmm and RifR using ESPRIPT (55). Secondary structure elements are indicated above the amino acid sequence: α-helices with squiggles, β-strands with arrows, and turns with the letters TT. Conserved residues are boxed. Blue triangles indicate the position of the catalytic residues. The last six residues in TesAmm were not modeled because the structure of RifR is missing the coordinates of the corresponding residues and also are omitted from the sequence alignment in panel D.

Deletion of tesA in M. marinum

The requirement of tesA for PDIM production has been hypothesized, but not demonstrated, whereas the requirement of this gene for PGL production remains to be investigated. In this study we sought to conclusively establish whether tesA was required for PDIM and/or PGL production. Toward this end, we utilized the opportunistic human pathogen Mm as a prototype representative of mycobacteria that produce both PDIMs and PGLs. We engineered strain Mm ΔtesA, an unmarked in-frame ΔtesA mutant of Mm, and investigated the capacity of the strain to produce PDIMs and PGLs as described below. The deletion was engineered using the tesA deletion cassette-delivery suicide vector p2NIL-GOALc-ΔtesAc (∼15 kb) in a homologous recombination- and counterselection-based approach that replaced tesA by a 25-codon remnant engineered into the tesA deletion cassette of the vector (supplemental Fig. S1). The deletion in Mm ΔtesA encompassed 224 central codons of tesA and was verified by PCR using two independent primer pairs, each producing diagnostic amplicons of different sizes depending on whether the genomic DNA used as template was WT or carried the tesA deletion (supplemental Fig. S4). The engineering of Mm ΔtesA is the first reported example of the construction of an unmarked site-directed deletion mutant of this pathogen.

The Gene tesA Is Required for PDIM and PGL Production

We investigated the effect of the tesA deletion in Mm ΔtesA on PDIM and PGL production using LC-MS analysis. The presence of PDIMs and PGLs was probed for in lipid samples from Mm WT + pCP0 (expression vector), Mm ΔtesA + pCP0, and Mm ΔtesA + pCP0-tesAmm (a plasmid expressing Mm tesA) by LC-MS. Mm WT + pCP0 and Mm ΔtesA + pCP0, rather than the respective plasmid-free strains, were used in the experiments so that the WT, the mutant, and the complemented strain could all be cultured under identical conditions (i.e., kanamycin-containing growth medium) for comparative analysis. Representative results from this analysis are shown in Table 1 and supplemental Figs. S5 and S6. MS analysis of the samples from Mm WT + pCP0 revealed the expected ion series consistent with the characteristic PDIM and PGL mass patterns from acyl chain length heterogeneity (1). Conversely, the PDIM and PGL ion series was not detected by LC-MS analysis of the samples from Mm ΔtesA + pCP0. The analysis indicated that the mutant strain produces only traces of PDIMs and PGLs. The PDIM and PGL production defect was corrected by transformation of the mutant with pCP0-tesAmm. The complemented strain displayed PDIM and PGL ion series with ion signals comparable with those from Mm WT + pCP0. The complementation control rules out the possibility that the deletion exerted a meaningful polar effect on the neighboring genes known to be required for PDIM and PGL production (1). Overall, the results of the LC-MS conclusively demonstrate that tesA plays a critical role in the production of both PDIMs and PGLs in Mm.

TABLE 1.

Mass spectrometry analysis of PDIMs and PGLs

^a Variants with either a methoxy or a keto group at C3 exist. Only methoxy variants are shown in Fig. 1.

^b Calculated neutral exact masses and calculated exact masses for [(M+NH₄⁺)]⁺ were determined using Agilent's MassHunter Qualitative Analysis Software. The experimental exact masses for [(M+NH₄⁺)]⁺ were determined by processing the data using Agilent's MassHunter Qualitative Analysis Software's “Find Compounds by Formula” feature. The software searches the data for ions based on a given molecular formula and, in this case, extracts the ions for the [(M+NH₄⁺)]⁺ ion for the molecular formulas of interest.

^c ppm values were determined by Agilent's MassHunter Qualitative Analysis Software from comparison of the experimental exact mass for [(M+NH₄⁺)]⁺ and the calculated exact mass for [(M+NH₄⁺)]⁺ for a given molecular formula.

^d Abundance corresponds to the integrated area for all ion adducts.

^e 1, Mm WT + pCP0; 2, Mm ΔtesA + pCP0; 3, Mm ΔtesA + pCP0-tesAmm; 4, Mm ΔtesA pCP0-tesAmmS92A. nd, not detected.

A S92A Replacement in TesAmm Eliminates the Protein Ability to Support PDIM and PGL Production

The predicted Ser-92—Asp-196—His-224 catalytic triad in TesAmm is well positioned for catalysis, as judged by structural modeling (Fig. 2). Should TesAmm be indeed a thioesterase required for product release from PpsE as earlier hypothesized (24), a S92A replacement would be expected to abrogate thioesterase activity and lead to a drastic reduction in PDIM and PGL production. To probe the biological relevance of Ser-92, we engineered a site-directed S92A mutant allele (tesAmmS92A) and investigated its ability to complement Mm ΔtesA. As noted above, the three-dimensional model of TesAmmS92A (Fig. 2C) predicts that the S92A substitution does not disrupt the structural fold of the protein. LC-MS analysis demonstrated that introduction of pCP0-tesAmmS92A (expressing tesAmmS92A) in Mm ΔtesA failed to complement the mutant (Table 1 and supplemental Figs. S5 and S6). These results reveal a critical role for Ser-92 and are consistent with the proposed catalytic function of Ser-92.

Loss of TesAmm Correlates with Increased Growth Fitness

At the onset of this study, we noticed a small, yet consistent, growth yield increase for Mm ΔtesA over the WT. Interestingly, a slight in vitro growth advantage has been reported for the PDIM-deficient Mtb mutant over the WT (27). In view of these precedents, we investigated whether deletion of tesA correlated with a growth advantage ex vivo. We compared the growth of Mm WT and Mm ΔtesA in liquid culture using a 96-well plate-based assay that facilitated execution of replicates and collection of abundant data sets. Representative results from these experiments are shown in Fig. 3A. Results from additional experiments are shown in supplemental Fig. S7. Comparison of growth curves derived from this assay revealed no apparent difference in growth rate between Mm ΔtesA and WT. However, a small, yet reproducible, increase in growth yield for Mm ΔtesA relative to Mm WT was observed (Fig. 3A and supplemental Fig. S7). We also compared the growth of Mm WT + pCP0, Mm ΔtesA + pCP0, and Mm ΔtesA + pCP0-tesAmm. The plasmid-bearing Mm ΔtesA + pCP0 and Mm WT + pCP0 strains, rather than the respective plasmid-free strains, were compared with the complemented mutant in this experiment so that all compared strains could be cultured under identical conditions. These experiments consistently revealed an increase in growth yield for Mm ΔtesA + pCP0 relative to Mm WT + pCP0 (Fig. 3B and supplemental Fig. S7). The growth yield gained by the loss of tesA was considerably reduced in the complemented strain, Mm ΔtesA + pCP0-tesAmm (Fig. 3B and supplemental Fig. S7). No apparent difference in growth rate between the strains was observed. These results indicate that deletion of tesA leads to an increase in growth yield and suggest that wild-type production of PDIMs and/or PGLs is associated with a reduction of fitness ex vivo.

FIGURE 3.

Deletion of tesA correlates with an increase in growth yield. A, growth curves of the plasmid-free Mm WT (○) and Mm ΔtesA (△) strains in antibiotic-free growth medium are shown. B, growth curves of the plasmid-bearing Mm WT + pCP0 (○), Mm ΔtesA + pCP0 (△), and Mm ΔtesA + pCP0-tesAmm (◊) strains in kanamycin-containing growth medium are shown. Data were generated in a multiwell plate-based growth assay, and data points shown represent the means of 30 wells (±S.E.). Each plot is representative of three independent growth comparison experiments. Two-sample unpaired t test at 15-day time point: A, p < 0.01 (**); B, p < 0.0001 (****).

Deletion of tesA Leads to Pan-antibiotic Hypersensitivity

It is conceivable that a drastic reduction of PDIMs and PGLs in the outer leaflet of the mycobacterial outer membrane (47) might produce structural or fluidity changes that lead to an increase in outer-membrane permeability in Mm ΔtesA. With this idea in mind and considering that increased outer-membrane permeability might increase antibiotic susceptibility by facilitating compound penetration, we probed the antibiotic susceptibility of the tesA mutant. We compared the susceptibility of Mm WT + pCP0, Mm ΔtesA + pCP0, and Mm ΔtesA + pCP0-tesAmm to six clinical antibiotics of different classes. The group tested included rifampicin, doxycycline, ciprofloxacin, and streptomycin, all of which have reported use in the treatment of Mm infections (48). Remarkably, Mm ΔtesA + pCP0 displayed increased susceptibility to five of these antibiotics compared with Mm WT + pCP0, as judged by IC₅₀ and MIC values (Table 2). The pan-antibiotic hypersensitivity of the tesA mutant was also exposed by disk-diffusion assays (Table 2). Mm ΔtesA + pCP0-tesAmm showed a susceptibility pattern similar to that of Mm WT + pCP0, indicating that the hypersensitive phenotype of the mutant was complemented by episomal expression of tesA (Table 2). Representative dose-response curves and results from disk-diffusion assays are shown in Fig. 4. Overall, these results conclusively link loss of tesA to pan-antibiotic hypersensitivity.

TABLE 2.

Antibiotic susceptibility

MIC, minimal inhibitory concentration; ND, not determined.

Antibiotic name (antibiotic class)	Mm WT + pCP0		Mm ΔtesA + pCP0		Mm ΔtesA + pCP0-tesAmm		IC50-fold changec	MIC-fold change	Disk assay
	IC50a	MICb	IC50	MIC	IC50	MIC
	μg/ml	μg/ml	μg/ml	μg/ml	μg/ml	μg/ml
Cefuroxime (2nd generation cephalosporin)	2.73 ± 0.70	10.42 ± 2.60	0.26 ± 0.07	1.06 ± 0.24	2.32 ± 0.05	9.11 ± 1.30	10.4	9.8	+
Doxycycline (tetracycline antibiotic)	0.84 ± 0.40	3.67 ± 1.67	0.10 ± 0.01	0.33 ± 0.00	0.64 ± 0.09	3.00 ± 1.00	8.2	11.0	+
Rifampicin (rifamycin antibiotic)	0.18 ± 0.04	1.17 ± 0.39	0.04 ± 0.02	0.15 ± 0.05	0.15 ± 0.06	1.04 ± 0.26	4.7	8.0	+
Ciprofloxacin (2nd generation fluoroquinolone)	0.22 ± 0.05	1.42 ± 0.58	0.10 ± 0.04	0.42 ± 0.08	0.17 ± 0.02	0.71 ± 0.29	2.3	3.4	+
Ampicillin (β-lactam antibiotic)	99.68 ± 6.83	ND	11.74 ± 1.38	ND	114.08 ± 8.92	ND	8.5	ND	+
Streptomycin (aminoglycoside antibiotic)	0.55 ± 0.14	2.34 ± 0.78	0.56 ± 0.03	2.34 ± 0.78	0.41 ± 0.06	2.15 ± 0.98	1.0	1.0	−

^a IC₅₀ (±S.E.) values were calculated from sigmoidal curves fitted to duplicate sets of dose-response data.

^b MIC (±S.E.) values are the means of duplicates.

^c IC₅₀-fold change = IC₅₀ Mm WT + pCP0/IC₅₀ Mm ΔtesA + pCP0 ratio, MIC-fold change = MIC Mm WT + pCP0/MIC Mm ΔtesA + pCP0 ratio. +, increased susceptibility of Mm ΔtesA + pCP0 relative to Mm WT + pCP0 and Mm ΔtesA + pCP0-tesAmm in disk-diffusion assays. −, no change.

FIGURE 4.

Deletion of tesA leads to increased antibiotic susceptibility. A, dose-response plots for growth inhibition of the plasmid-bearing Mm WT + pCP0 (○), Mm ΔtesA + pCP0 (△), and Mm ΔtesA + pCP0-tesAmm (◊) strains as a function of antibiotic concentration are shown. Each data point represents the mean of triplicates (±S.E.). Each plot is representative of two independent experiments. B, representative standard disk-diffusion assays show growth inhibition zones at the antibiotic concentrations indicated.

DISCUSSION

The results of our LC-MS analyses conclusively demonstrate that tesA plays a critical role in the production of both PDIMs and PGLs. Deletion of tesA leads to a very drastic reduction in PDIM and PGL production. This mutant phenotype is fully complemented by episomal expression of WT tesAmm. Conversely, a tesAmmS92A allele with a Ser-to-Ala substitution in the predicted Ser-Asp-His catalytic triad of TesAmm does not complement Mm ΔtesA, a result that reveals a critical role for Ser-92 in TesAmm function and is consistent with the proposed involvement of Ser-92 in catalysis.

The mechanistic basis underlying the PDIM and PGL production defect in Mm ΔtesA remains to be investigated. It is likely, however, that this reduction is due to pathway stalling resulting from loss of TesAmm thioesterase-dependent release of the polyketide products thioesterified to PpsE. Alternatively, or in conjunction with a product-release function, TesAmm and its orthologs might be involved in the removal of inappropriate acyl units and/or aberrant acyl intermediates thioesterified to acyl carrier protein domains of the polyketide synthases involved in PDIM and PGL synthesis. Such a “housekeeping function” has been attributed to RifR (36) and other type II thioesterases and is needed for efficient metabolite production (49–51). Notably, the thioesterases with this housekeeping function have been found in pathways where a type I thioesterase or another enzyme is involved in product release. The fact that TesA is the only protein with predicted competence to catalyze polyketide product release encoded in proximity to the PDIM/PGL biosynthetic genes in conjunction with the documented TesAmt-PpsE interaction (24) suggests that TesAmm and its orthologs are indeed required for efficient product release.

The production of the PDIM and PGL remnants observed in Mm ΔtesA could be explained by enzyme-independent hydrolytic product release and/or by nonspecific action of other Mm thioesterases. We identified two tesA paralogs in Mm. One of these paralogs, MMAR_3693, is located in the gene cluster encoding the nonribosomal peptide synthetases and polyketide synthases involved in the synthesis of mycobactin siderophores (44). The second paralog, MMAR_0367, forms a bicistronic operon with MMAR_0368, a large nonribosomal peptide synthetase gene of unknown function. Based on genetic context, we predict that MMAR_3693 is involved in mycobactin production, whereas MMAR_0367 is likely to function in the production of a nonribosomal peptide synthesized by MMAR_0368. Nonspecific action of either or both of these TesAmm paralogs could perhaps support the production of the PDIM and PGL remnants observed in the absence of TesAmm.

The TesA orthologs have a predicted Ser-Asp-His catalytic triad characteristic of the α/β-hydrolase superfamily (45), and the Ser residue of this triad is critical for function in TesAmm. To further evaluate the relevance of the Ser-Asp-His triad of TesAmm and generate a first model of the three-dimensional structure of the proteins, we modeled the structure of TesAmm based on the crystal structure of RifR (36). The refined three-dimensional model strongly suggests that the TesAmm folds in three dimensions to present the conserved catalytic residues correctly within a structural context for it to function as a thioesterase. If the biochemistry is disturbed without affecting the structural fold, as predicted by modeling the mutant TesAmmS92A, the catalytic function is lost as expected. Interestingly, the modeled TesAmm seems to best fit the subtle details of tertiary structure of RifR within the range of available structures for the structural fold of the α/β-hydrolase superfamily, including the lid subdomain that covers the active site. The lid flexibility of RifR suggested by the variation among RifR crystal structures (36) may well be present in TesAmm based on the observation that alternative homology models of TesAmm show the most variation in this region compared with the rest of the modeled structure (data not shown). This suggested motion of the lid could impact the size and shape of the substrate chamber and, therefore, the substrate accommodated and also access to the catalytic triad from the presumed 4′-phosphopantetheine prosthetic group entrance (36).

In addition to the drastic reduction in PDIM and PGL production, two other mutant traits are associated with the deletion of tesA in Mm ΔtesA. The tesA mutant has increased in vitro growth and pan-antibiotic hypersusceptibility compared with the parental WT. Both these phenotypic characteristics are complemented by episomal expression of tesA, and thus the possibility that they originate from a polar effect created by the tesA deletion can be ruled out. Interestingly, a slight in vitro growth increase has been documented for the PDIM-deficient Mtb mutant over the parental WT (27). These results suggest that wild-type PDIM and/or PGL production is associated with a reduction of fitness ex vivo. This scenario contrasts with the increased fitness conferred by these virulence effectors in vivo (2, 3).

The molecular mechanisms behind the increased growth ex vivo and the pan-antibiotic hypersensitivity arising from the lack of tesA remain obscure. It is tempting to hypothesize, however, that these two traits have a common mechanistic origin, a weakening of the permeability barrier of the mycobacterial cell envelope in the tesA mutant. It is generally believed that the permeability barrier imposed by the unique mycobacterial outer membrane limits the growth of these microbes by restricting influx of selected nutrients and reduces antibiotic susceptibility by decreasing compound penetration (47, 52, 53). It is indeed possible that the drastic reduction of PDIMs and/or PGLs in the outer leaflet of the outer membrane of Mm ΔtesA produces structural or fluidity changes that lead to an increase in nutrient and antibiotic permeability. In connection with this possibility, others have reported that a PDIM-deficient mutant of Mtb strain Mt103 exhibits increased cell wall permeability compared with the parental PDIM producer (54). It is also possible that the increased growth yield of Mm ΔtesA suggests that the large amount of carbon normally funneled to PDIMs and PGLs is being used for the construction of other cell structures, thus allowing Mm ΔtesA to produce a higher yield of cells on the same amount of carbon.

Overall, our studies demonstrate a relevant role for tesA in mycobacterial biology, advance our understanding of the biosynthesis of an important group of mycobacterial virulence factors, and suggest that antivirulence therapy aimed at blocking PDIM and/or PGL production might synergize with antibiotic therapy in the control of mycobacterial infections. This previously unrecognized possibility for therapeutic synergy warrants further investigation, as does the pan-antibiotic hypersensitivity of Mm ΔtesA. Understanding the molecular basis of the pan-antibiotic hypersensitivity in Mm ΔtesA will provide insights into the mechanisms of antibiotic penetration in the mycobacteria, a process with clinical significance that remains poorly understood.