Biosynthesis of Cell Envelope-Associated Phenolic Glycolipids in Mycobacterium marinum

Abstract

Phenolic glycolipids (PGLs) are polyketide synthase-derived glycolipids unique to pathogenic mycobacteria. PGLs are found in several clinically relevant species, including various Mycobacterium tuberculosis strains, Mycobacterium leprae, and several nontuberculous mycobacterial pathogens, such as M. marinum. Multiple lines of investigation implicate PGLs in virulence, thus underscoring the relevance of a deep understanding of PGL biosynthesis. We report mutational and biochemical studies that interrogate the mechanism by which PGL biosynthetic intermediates (p-hydroxyphenylalkanoates) synthesized by the iterative polyketide synthase Pks15/1 are transferred to the noniterative polyketide synthase PpsA for acyl chain extension in M. marinum. Our findings support a model in which the transfer of the intermediates is dependent on a p-hydroxyphenylalkanoyl-AMP ligase (FadD29) acting as an intermediary between the iterative and the noniterative synthase systems. Our results also establish the p-hydroxyphenylalkanoate extension ability of PpsA, the first-acting enzyme of a multisubunit noniterative polyketide synthase system. Notably, this noniterative system is also loaded with fatty acids by a specific fatty acyl-AMP ligase (FadD26) for biosynthesis of phthiocerol dimycocerosates (PDIMs), which are nonglycosylated lipids structurally related to PGLs. To our knowledge, the partially overlapping PGL and PDIM biosynthetic pathways provide the first example of two distinct, pathway-dedicated acyl-AMP ligases loading the same type I polyketide synthase system with two alternate starter units to produce two structurally different families of metabolites. The studies reported here advance our understanding of the biosynthesis of an important group of mycobacterial glycolipids.

INTRODUCTION

Mycobacterial infections are responsible for devastating morbidity and mortality worldwide (1–5). A critical player in the ability of the mycobacteria to produce disease is a formidable cell envelope believed to be responsible for the intrinsic resilience of the mycobacteria to inhospitable environments, antimicrobial agents, and host immune defenses (6–12). Among the unique components found in the cell envelope of several pathogenic mycobacteria are two structurally related families of glycosylated and nonglycosylated lipids commonly referred to as phenolic glycolipids (PGLs) and phthiocerol dimycocerosates (PDIMs), respectively (for a review, see reference 13). PGLs and PDIMs have unusual lipid scaffolds consisting of β-diol-containing, long-chain, aliphatic polyketides esterified with long-chain, multimethyl-branched fatty acids onto the diol functionality (Fig. 1). PGLs and PDIMs are found in several mycobacterial pathogens (e.g., Mycobacterium tuberculosis strains, Mycobacterium bovis, Mycobacterium leprae, and Mycobacterium marinum) and thought to be constituents of the characteristic mycobacterial outer membrane. Multiple lines of investigation have provided considerable support for the idea that PGLs and PDIMs are implicated in virulence via complex mechanisms of action that are not fully elucidated (14–32). These (glyco)lipids are also believed to strengthen the cell envelope permeability barrier (20, 33) and to increase the bacterium's intrinsic resistance to antimicrobial drugs (20, 22, 34).

FIG 1.

Representative structures of mycobacterial PGLs and PDIMs. The carbon chain variability and glycosyl unit represented are from main variants found in the opportunistic human pathogen M. marinum. m and m', 16 to 20; n, 16 to 22; p, 14 to 22.

The relevance of PGLs in mycobacterial biology underscores the importance of developing a comprehensive knowledge of the PGL biosynthetic pathway, which remains incompletely understood. Our previous studies of the PGL biosynthetic pathway have revealed a functional cooperation between the M. marinum proteins FadD22 (p-hydroxybenzoate-AMP ligase/initiation module) and Pks15/1 (iterative type I polyketide synthase [PKS]) for production of p-hydroxyphenylalkanoate (PHPA) intermediates required for PGL biosynthesis (35, 36). The PHPA intermediates synthesized by the M. marinum FadD22-Pks15/1 iterative system, which is conserved in PGL producers, are thought to be further extended to form the phenolphthiocerol moiety of PGLs (Fig. 1). The extension of the PHPA intermediates has been proposed to be carried out by the modular type I PKS system PpsABCDE, a noniterative synthase complex known to extend fatty acids to form the phthiocerol moiety of PDIMs in M. tuberculosis and conserved in PGL/PDIM producers (13, 16, 25, 37, 38) (Fig. 1). However, the mechanism by which the PHPA intermediates assembled by the FadD22-Pks15/1 system would be transferred to the PpsABCDE system for acyl chain extension remains to be experimentally investigated, and the ability of the PpsABCDE system to extend PHPA intermediates has yet to be validated.

Here, we report mutational and biochemical studies that interrogate the mechanism by which the PHPA intermediates are transferred to PpsA, the first-acting enzyme of the PpsABCDE system, and probe the ability of PpsA to extend these intermediates. The studies were conducted with M. marinum, a nontuberculous mycobacterial species that is the closest genetic relative of the M. tuberculosis complex, is often utilized to model aspects of M. tuberculosis biology, and offers greater experimental tractability than other PGL/PDIM producers (39–42). The findings of our studies support a mechanistic model in which the PHPA intermediates are activated and loaded onto the loading acyl carrier protein (ACP_L) domain of M. marinum PpsA by a dedicated PHPA-AMP ligase. Our results also demonstrate that M. marinum PpsA is capable of extending PHPA intermediates. The conservation of the PGL biosynthetic genes across species (13, 43) (Fig. 2) and studies of members of the M. tuberculosis complex (44) strongly suggest that the mechanistic insights into PGL biosynthesis gained here are applicable to other PGL producers. Overall, these studies advance our understanding of the biosynthesis of an important group of mycobacterial cell envelope-associated glycolipids.

FIG 2.

Conservation of M. marinum and M. tuberculosis chromosomal loci involved in PGL-PDIM production. The five M. marinum genes targeted for mutational analysis (Δ, deletion; *, amino acid substitution) in this study and their respective orthologs in M. tuberculosis are highlighted with pattern-filled arrows. †, pks15/1 is essential for PGL production. The gene pks15/1 is disrupted (split into pks1 and pks15) by natural mutations in M. tuberculosis H37Rv and other Euro-American lineage strains, thus leading to PGL deficiency (18, 64). Additional genes implicated in PGL and/or PDIM production located downstream of fadD29 are not depicted. Adapted from reference 13 with permission of the publisher.

MATERIALS AND METHODS

Culturing conditions, recombinant DNA manipulations, and reagents.

M. marinum strain M (ATCC BAA-535) and its derivatives were cultured under standard conditions in Middlebrook 7H9 medium (Difco) supplemented with 10% ADN (5% bovine serum albumin [BSA], 2% dextrose, 0.85% NaCl) and 0.05% Tween 80 (supplemented 7H9) or on Middlebrook 7H11 agar (Difco) with ADN (supplemented 7H11) (45). The strains used in this study are listed in Table S1 in the supplemental material. Escherichia coli strains were cultured under standard conditions in Luria-Bertani (LB) medium (46). When required, kanamycin (30 μg ml⁻¹), hygromycin (50 μg ml⁻¹), sucrose (2%), and/or 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal; 70 μg ml⁻¹) were added to the growth media. DNA manipulations were carried out by standard methods and using E. coli DH5α (Invitrogen) as the primary cloning host (46). All PCR-generated DNA fragments used in plasmid constructions were sequenced to verify fidelity. The plasmids used in this study are listed in Table S2 in the supplemental material. Genomic DNA isolation and plasmid electroporation into mycobacteria were carried out as reported previously (45). Molecular biology reagents were obtained from Sigma, Invitrogen, Novagen, or Qiagen. The oligonucleotides used in this study are listed in Tables S3 and S4 in the supplemental material, and they were purchased from Integrated DNA Technologies, Inc. Solvents and nonradiolabeled chemicals were purchased from Sigma, Acros Organics, or Fisher Scientific. [1-¹⁴C]propionate (specific activity, 54 mCi mmol⁻¹) and [carboxyl-¹⁴C]p-hydroxybenzoic acid (specific activity, 55 mCi mmol⁻¹) were acquired from American Radiolabeled Chemicals, Inc. The compound 21-(4-hydroxyphenyl)henicosanoic acid (s-PHPA) was synthesized as described in the supplemental material.

Construction of mycobacterial mutants.

The mutants were engineered using the p2NIL/pGOAL19-based flexible cassette method (47) as reported previously (34, 35, 48). A gene-specific mutagenesis cassette delivery vector (see below) was used to construct each mutant. Each vector was electroporated into M. marinum, and the transformants with a potential single crossover (blue colonies) were selected on supplemented 7H11 containing hygromycin, kanamycin, and X-Gal. Potential single-crossover-bearing clones were grown in antibiotic-free supplemented 7H9 and then plated for single colonies on supplemented 7H11 containing sucrose and X-Gal. White colonies that grew on sucrose plates were restreaked onto antibiotic-free and antibiotic-containing plates to identify drug-sensitive clones, a trait indicating a possible double-crossover event with consequent allelic replacement or reversion to wild type (wt). Gene deletions were confirmed by PCR. For PCR analysis, genomic DNA from mutant candidates was used as the template along with two independent mutant-specific primer pairs (see Table S3 in the supplemental material) to produce amplicons permitting differentiation between mutant and wt genotypes based on amplicon size (see Fig. S2 in the supplemental material). Nucleotide substitutions in the M. marinum ppsA_S-to-A mutant were confirmed by DNA sequencing. The mutated region was PCR amplified with specific primers (see Table S3 in the supplemental material), and the resulting amplicon was sequenced (see Fig. S2 in the supplemental material).

Construction of mutagenesis cassette delivery vectors.

The following mutagenesis cassette delivery vectors were constructed: p2NILGOALc-ΔfadD22c, carrying a fadD22 (MMAR_1761) deletion cassette (ΔfadD22c); p2NILGOALc-ΔfadD26c, carrying a fadD26 (MMAR_1777) deletion cassette (ΔfadD26c); p2NILGOALc-ΔfadD28c, carrying a fadD28 (MMAR_1765) deletion cassette (ΔfadD28c); p2NILGOALc-ΔfadD29c, carrying a fadD29 (MMAR_1759) deletion cassette (ΔfadD29c); and p2NILGOALc-ppsAc, carrying a ppsA (MMAR_1776) mutagenesis cassette (ppsAc) (see Fig. S1 and Table S2 in the supplemental material). Each cassette was constructed by joining a 5′ arm and a 3′ arm using splicing by overlap extension (SOE) PCR (49). The primers and amplicon sizes are shown in Table S4 in the supplemental material. Each PCR-generated cassette was first cloned into pCR2.1-TOPO (Invitrogen), verified for sequence fidelity, and then subcloned into p2NIL (47). The cassettes ΔfadD22c, ΔfadD26c, ΔfadD28c, ΔfadD29c, and ppsAc were cloned into p2NIL as SalI-NotI, HindIII-KpnI, BamHI-NotI, HindIII-PmlI, and HindIII-HpaII fragments, respectively. Each resulting p2NIL-mutagenesis cassette construct and the plasmid pGOAL19 (47) were digested with PacI, and then the PacI fragment with the marker cassette (GOALc) of pGOAL19 was ligated to the p2NIL construct backbones to generate the final delivery vectors. The configuration of each final cassette was as follows: ΔfadD22c consisted of fadD22's 1,000-bp upstream segment plus fadD22's first 2 codons plus fadD22's last 2 coding codons plus stop codon plus 983-bp downstream segment; ΔfadD26c consisted of fadD26's 962-bp upstream segment plus fadD26's first 5 codons plus fadD26's last 3 coding codons plus stop codon plus 929-bp downstream segment; ΔfadD28c consisted of fadD28's 947-bp upstream segment plus fadD28's first 5 codons plus fadD28's last 4 coding codons plus stop codon plus 935-bp downstream segment; ΔfadD29c consisted of fadD29's 947-bp upstream segment plus fadD29's first 4 codons plus fadD29's last 5 coding codons plus stop codon plus 976-bp downstream segment; and ppsAc consisted of a 1,832-bp segment with Ser-to-Ala substitution mutations in the center.

Construction of pCP0 derivatives.

Plasmids pCP0-fadD22, pCP0-fadD26, pCP0-fadD28, pCP0-fadD29, and pCP0-ppsA, constitutively expressing fadD22, fadD26, fadD28, fadD29, and ppsA, respectively, from the mycobacterial hsp60 promoter were constructed using the expression vector pCP0 (34). DNA fragments each encompassing an M. marinum gene and its predicted ribosome-binding site were PCR generated using the primer pairs shown in Table S4 in the supplemental material. The fragments were first cloned into pCR2.1-TOPO, verified for sequence fidelity, and then subcloned into pCP0. The fadD22, fadD26, fadD28, fadD29, and ppsA fragments were cloned into pCP0 as HindIII-HpaI, EcoRI-HindIII, NheI-HindIII, EcoRI-HindIII, and HpaI-NheI inserts, respectively.

Analysis of PGLs and PDIMs.

Four-day-old cultures were diluted to an optical density at 595 nm (OD₅₉₅) of 0.6 in supplemented 7H9 and loaded into 12-well plates (1 ml per well). [¹⁴C]propionate (which labels both PDIMs and PGLs) or [¹⁴C]p-hydroxybenzoic acid (which selectively labels PGLs) was added to each well (0.2 μCi ml⁻¹), and the plates were incubated for 24 h (30°C, 170 rpm). After incubation, the OD₅₉₅ of the cultures was measured in a plate reader (Beckman Coulter, Inc.) and the cells were harvested for apolar lipid fraction extraction with a biphasic mixture of methanolic saline and petroleum ether as reported previously (35, 50). Lipid extracts were subjected to radiometric thin-layer chromatography (radio-TLC) for analysis of ¹⁴C-labeled PGLs and ¹⁴C-labeled PDIMs as described earlier (35, 50). Developed TLC plates were exposed to phosphor screens, which were scanned using a Cyclone Plus Storage Phosphor System (PerkinElmer, Inc.).

Construction of pETDuet-PpsA.

M. marinum ppsA was assembled from three fragments: NT1 (5′-end fragment, 2,100 bp); NT2 (middle fragment, 2,000 bp); and NT3 (3′-end fragment, 809 bp). The fragments were PCR amplified from genomic DNA with fragment-specific primers (see Table S4 in the supplemental material) and independently cloned into pCR2.1-TOPO to create pCR2.1TOPO-PpsA-NT1, pCR2.1TOPO-PpsA-NT2, and pCR2.1TOPO-PpsA-NT3. The inserts were verified for sequence fidelity. Subsequently, the NT1 insert of pCR2.1TOPO-PpsA-NT1 was recovered as a BamHI-AscI fragment and subcloned into pETDuet-1 (Novagen) digested with the same enzymes to create pETDuet-PpsA-NT1. Then, the NT2 insert of pCR2.1TOPO-PpsA-NT2 was recovered as an SpeI-AscI fragment and subcloned into SpeI-AscI linearized pETDuet-PpsA-NT1, resulting in pETDuet-PpsA-NT1NT2. Finally, the NT3 insert of pCR2.1TOPO-PpsA-NT3 was recovered as an AscI-HindIII fragment and subcloned into pETDuet-PpsA-NT1NT2 digested with the same enzymes to generate pETDuet-PpsA. This final plasmid was introduced into E. coli BAP1 (51) for isopropyl β-d-1-thiogalactopyranoside (IPTG)-inducible production of N-terminally His₆-tagged PpsA.

Construction of pCOLADuet-FadD29.

M. marinum fadD29 was PCR amplified from genomic DNA with gene-specific primers (see Table S4 in the supplemental material) and cloned into pCR2.1-TOPO to create pCR2.1TOPO-FadD29. The insert of pCR2.1TOPO-FadD29 was verified for sequence fidelity, recovered as an EcoRI-NotI fragment, and subcloned into pCOLADuet-1 (Novagen) linearized with EcoRI and NotI to generate pCOLADuet-FadD29. This final plasmid was introduced into E. coli BL21(DE3) (Stratagene) for IPTG-inducible production of N-terminally His₆-tagged FadD29.

Overproduction and purification of PpsA.

E. coli BAP1 carrying pETDuet-PpsA was cultured in LB broth supplemented with ampicillin (100 μg ml⁻¹) at 37°C with orbital shaking (220 rpm) to an OD₆₀₀ of 0.6. When the culture reached the target OD, the incubation temperature was reduced to 18°C, and PpsA expression was induced with IPTG (0.1 mM). After 20 h of additional incubation (18°C, 220 rpm), cells were harvested by centrifugation (6,000 × g, 20 min). The cell pellet was resuspended in lysis buffer (75 mM sodium phosphate [pH 7.5], 500 mM NaCl, 10% glycerol, 10 mM imidazole), and cells were disrupted using a high-pressure homogenizer (Avestin, Inc.). Cellular debris was removed from the lysate by centrifugation (1 h, 12,000 rpm, FX6100 rotor; Beckman Coulter Inc.) followed by subsequent filtration of the supernatant (0.45-μm filter). PpsA was purified from the clarified supernatant by Ni²⁺ column chromatography using Ni-nitrilotriacetic acid (Ni-NTA) Superflow resin according to the manufacturer's instructions (Qiagen). Proteins were eluted from the column using an imidazole gradient in lysis buffer run with an ÄKTA UPC10 fast protein liquid chromatography (FPLC) system (GE Healthcare). PpsA eluted at 125 mM imidazole. Fractions with the best purity were identified by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), pooled, concentrated using Amicon Ultra-15 centrifugal filter devices (50-kDa cutoff), and buffer exchanged into 100 mM sodium phosphate buffer (pH 7.2) using PD-10 desalting columns (GE Healthcare). Protein samples were aliquoted with 25% glycerol final concentration, flash-frozen in liquid nitrogen, and stored at −80°C (see Fig. S3 in the supplemental material). A typical yield was 14 mg liter⁻¹.

Overproduction and purification of FadD29.

E. coli BL21(DE3) carrying pCOLADuet-FadD29 was cultured in LB broth containing kanamycin at 37°C with orbital shaking (220 rpm). When the culture reached an OD_{600 nm} of 0.6, the incubation temperature was reduced to 30°C, and FadD29 production was induced with IPTG (0.1 mM). After 20 h of additional incubation (30°C, 220 rpm), cells were harvested by centrifugation and resuspended in lysis buffer (50 mM Tris-HCl, 500 mM NaCl [pH 7.5], 10 mM imidazole). Cell disruption and removal of cellular debris from the lysate were carried out as noted above. FadD29 was purified from the clarified supernatant by Ni²⁺ column chromatography. Proteins were eluted from the column using an imidazole gradient in lysis buffer run with an ÄKTA UPC10 FPLC system. FadD29 eluted at 140 mM imidazole. Fractions with the best purity were identified by SDS-PAGE, pooled, concentrated using Amicon Ultra-15 centrifugal filter devices (30-kDa cutoff), and buffer exchanged into 50 mM Tris-HCl–300 mM NaCl (pH 7.8) using PD-10 desalting columns. Purified proteins were aliquoted with 25% glycerol (final concentration), flash-frozen in liquid nitrogen, and stored at −80°C (see Fig. S3 in the supplemental material). A typical yield was 6.3 mg liter⁻¹.

In vitro FadD29-PpsA reconstituted system.

The standard complete reaction mixture (250 μl) contained 0.5 μM FadD29, 5.0 μM PpsA, 2.5 μM s-PHPA, 100 mM sodium phosphate buffer (pH 7.2), 1 mM tris(2-carboxyethyl)phosphine (TCEP), 0.5 mM MgCl₂, 0.5 mM ATP, 10% glycerol, 0.1 mM malonyl coenzyme A (malonyl-CoA) thioester, and 1 mM NADPH. After incubation (30°C, 2.5 h), the reaction was quenched by the addition of 50 μl of 1 M NaOH, and then the mixture was incubated at 65°C for 20 min to release the covalently bound products from PpsA. Following the alkaline hydrolysis, the reaction mixture was acidified with 50 μl of 2 M HCl, and the reaction products were extracted with ethyl acetate (750 μl × 2). The recovered organic layer was evaporated to dryness. The dried residual material was dissolved in 50 μl of dichloromethane and analyzed by high-resolution liquid chromatography-mass spectrometry (LC-MS) as described below. Control reaction mixtures lacking selected components were also set up, treated, and analyzed in the same manner as the complete reaction mixture.

Liquid chromatography-mass spectrometry instrumentation and analysis.

Mass spectral data were collected on an Agilent Technologies G6550A iFunnel high-resolution quadrupole time of flight (Q-TOF) mass spectrometer attached to an Agilent Technologies 1290 ultrahigh-performance liquid chromatography (UHPLC) system. Samples were ionized using Agilent's dual-sprayer Jet Stream electrospray ionization source (Dual AJS ESI) with the analysis performed in negative mode. Chromatography was performed on an Agilent Technologies Poroshell 120 EC-C₈ column (2.1 mm by 75 mm, 2.7 μm) using water containing 0.1% formic acid (solvent A) and methanol containing 0.1% formic acid (solvent B) at a flow rate of 350 μl min⁻¹. The UHPLC gradient was 75% solvent B (0 min) to 100% solvent B (15 min), and the analysis was stopped after 20 min. The column was equilibrated under the starting conditions for 10 min and held at 45°C for the entire analysis. The UHPLC stream was diverted to waste for the first 1.5 min of the analysis. Mass spectrometer parameters for the MS were as follows: 225°C drying gas temperature; 17-liters min⁻¹ drying gas flow; 35-lb/in² nebulizer pressure; 200°C sheath gas temperature; 12-liters min⁻¹ sheath gas flow; 3,500-V capillary voltage; 2,000-V nozzle voltage; and 365-V fragmentor. Data were collected with the instrument set to low mass range (100 to 1,700 m/z) and extended dynamic range conditions (2 GHz mode) at 2 spectra per second. Both centroid and profile data were stored (1.5 to 20 min) with a threshold of 300 counts for MS mode. The reference masses 112.985587 m/z (trifluoroacetate ion) and 966.000725 m/z (HP 921 reference compound + formate ion) were infused into the spray chamber through the second sprayer using an Agilent binary pump. A minimum height of 5,000 counts and a 100 ppm window were used. The instrument was controlled with Agilent MassHunter Workstation Acquisition Software B.05.00. The data were analyzed using Agilent MassHunter Workstation Qualitative Analysis Software B.05.00.

RESULTS AND DISCUSSION

Possible pathways for Pks15/1-to-PpsA PHPA intermediate transfer in M. marinum.

Previous studies have shown that the conserved modular (noniterative) type I PKS system PpsABCDE encoded in the PGL/PDIM biosynthetic gene cluster of M. tuberculosis (Fig. 2) extends fatty acids to form the phthiocerol moiety of PDIMs (Fig. 1). The fatty acids are activated and loaded by the conserved fatty acyl-AMP ligase FadD26 (Fig. 2) onto the loading acyl carrier protein (ACP_L) domain of PpsA, the first-acting enzyme of the PpsABCDE system (37, 52). The PpsABCDE system is proposed to also extend pathway-dedicated PHPA intermediates synthesized by the conserved iterative PKS system FadD22-Pks15/1 (Fig. 2) to form the phenolphthiocerol moiety of PGLs (36, 38) (Fig. 1). The mechanism by which PHPA intermediates synthesized by the iterative FadD22-Pks15/1 system would be transferred to the noniterative PpsABCDE system for chain extension has not been experimentally interrogated.

We hypothesized two possible pathways by which this PHPA intermediate transfer could take place in M. marinum (Fig. 3). In one of these pathways (Fig. 3, pathway A), the intermediates would be released first from the phosphopantetheinyl (P-pant) group of the acyl carrier protein (ACP) domain of Pks15/1 (step A1) and subsequently activated (step A2) and loaded (step A3) by a dedicated PHPA-AMP ligase onto the P-pant group of the ACP_L domain of PpsA. The PHPA intermediates bound to the ACP_L would be subsequently captured by the ketosynthase (KS) domain (step A4) to generate the loaded PpsA (Fig. 3, boxed acyl-PpsA species) ready for KS domain-dependent decarboxylation/condensation. The Pks15/1-to-PpsA PHPA intermediate transfer mechanism represented in pathway A (Fig. 3) emerges from analogy to the fatty acyl-AMP ligase-dependent mechanism of fatty acid activation and loading onto the ACP_L domain of PpsA during PDIM biosynthesis in M. tuberculosis (52). Pathway A is further supported by recent studies in the M. tuberculosis complex, leading to the proposal that the conserved fatty acyl-AMP ligase FadD29 (Fig. 2) loads PHPAs onto PpsA (44), yet this idea remains to be experimentally validated.

FIG 3.

Two possible mechanisms for transfer of p-hydroxyphenylalkanoate (PHPA) intermediates to PpsA in M. marinum. (A) PHPA-AMP ligase-dependent model. The model includes the release of PHPAs thioesterified to Pks15/1's ACP domain (A1), activation of free PHPAs by a PHPA-AMP ligase (A2), PHPA-AMP ligase-dependent loading of PHPAs onto PpsA's ACP_L domain (A3), and capture of the ACP_L domain-bound PHPAs by PpsA's KS domain to yield the fully loaded PpsA (boxed) ready for KS domain-dependent decarboxylation/condensation. (B) Direct KS domain capture model. In this model, the fully loaded PpsA (boxed) is generated directly by the capture of Pks15/1-bound PHPAs by PpsA's KS domain (B1), thus bypassing the need for steps A1 through A4. Decarboxylation/condensation leads to extension of PHPAs by a 2-carbon unit via the first noniterative extension cycle in the formation of phenolphthiocerols. In the scheme, the depicted carbon chain variability in the PHPA-Pks15/1 thioester intermediate is that expected during synthesis of M. marinum PGLs. Adenylated PHPAs are shown bound to the PHPA-AMP ligase via noncovalent linkages (by analogy to other acyl adenylating enzymes). PpsA's C-terminal ACP domain is shown loaded (AT domain dependent) with the malonyl-CoA-derived extender unit. Thiol groups of the phosphopantetheinyl group in the ACP domains and the catalytic Cys in the KS domain are depicted. The hydroxyl group in the extended PHPAs generated from the keto group by action of the KR domain of PpsA is shown. Sections of the PHPA intermediates are color coded based on origin: blue, derived from a p-hydroxybenzoic acid starter unit; red, derived from malonyl extender units via iterative extension cycles; black, section derived from a malonyl extender unit via a noniterative extension cycle. Domain abbreviations: A, adenylation; ACP, acyl carrier protein; ACP_L, loading acyl carrier protein; AT, acyltransferase; DH dehydratase; ER, enoylreductase; KR, ketoreductase; KS, ketosynthase.

Notably, pathway A (Fig. 3) requires free PHPA intermediates (step A1), yet sequence analysis of Pks15/1 orthologs does not reveal the presence of a possible thioesterase domain that would conveniently catalyze the release of the PHPA intermediates thioesterified to the P-pant group of the ACP domain of the synthase (13, 36). This does not rule out, however, the possibility that the PHPA intermediates are released without the assistance of an external (self-standing) thioesterase or by the action of one. In the latter option, the thioesterase could be TesA, which is encoded in the PGL/PDIM biosynthetic gene cluster (Fig. 2) and was recently shown to be required for PGL and PDIM production in M. marinum (22, 34).

In the second possible pathway for Pks15/1-to-PpsA PHPA intermediate transfer in M. marinum (Fig. 3, pathway B), the PHPAs thioesterified to the P-pant group of the ACP domain of Pks15/1 would be directly captured by the KS domain of PpsA (step B1), thus bypassing the need for steps A1 to A4. This direct Pks15/1-to-PpsA chain translocation would “skip” the ACP_L domain of PpsA. Domain skipping has in fact been demonstrated in a few PKS systems (53–55). Moreover, this direct-capture pathway would have adaptive value because it would not require ATP for PHPA intermediate activation, an essential step in pathway A (step A2). This pathway would also obviate the need to off-load the PHPA intermediates from the ACP domain of Pks15/1 (pathway A, step A1).

A functional ACP_L domain in PpsA is required for production of both PGLs and PDIMs in M. marinum.

The ACP_L domain of PpsA requires phosphopantetheinylation of the Ser residue embedded in the P-pant group attachment site motif (NCBI Conserved Domains Database [CDD] no. pfam00550/smart00823) of the domain to become functional (56–58). Phosphopantetheinylation of the ACP_L domain introduces the P-pant group onto which the fatty acids are loaded with the assistance of the fatty acyl-AMP ligase FadD26 to form the fatty acyl-ACP_L domain thioester intermediate required for PDIM biosynthesis in M. tuberculosis (37, 52). Formation of the analogous PHPA-ACP_L domain thioester intermediate would be required for PGL production in M. marinum if the Pks15/1-to-PpsA PHPA intermediate transfer takes place by the PHPA-AMP ligase-dependent pathway outlined in Fig. 3 (pathway A). On the other hand, formation of the PHPA-ACP_L domain intermediate would not be needed for PGL production if the intermediate transfer proceeds via direct capture by the KS domain of PpsA as depicted in Fig. 3 (pathway B). With these considerations in mind, we probed the essentiality of the P-pant group attachment site of the ACP_L domain of PpsA for PGL production by mutational analysis in M. marinum. To this end, we engineered an unmarked, site-directed mutant (M. marinum ppsA_S-to-A) with a Ser-to-Ala substitution that eliminated the phosphopantetheinylation site (Ser43) in the ACP_L domain of the synthase. We identified Ser43 as the phosphopantetheinylation target in the P-pant group attachment site sequence motif of the ACP_L domain of PpsA by sequence analysis (not shown). The M. marinum ppsA_S-to-A mutant carried also a Ser42-to-Ala substitution. Ser42 (adjacent to Ser43) was replaced in case it could become a surrogate phosphopantetheinylation target in the absence of Ser43, a potentially confounding scenario.

Evaluation of PGL production in the M. marinum ppsA_S-to-A mutant by radiometric thin-layer chromatography (radio-TLC) analysis revealed that the strain was PGL deficient (Fig. 4A, cf. lanes 1 and 2). Radio-TLC analysis revealed that the mutant strain was also unable to produce PDIMs (Fig. 4B, cf. lanes 1 and 2), a result in line with previous biochemical studies on PDIM biosynthesis in M. tuberculosis (37, 52). Introduction of the plasmid pCP0-ppsA (expressing M. marinum ppsA) into M. marinum ppsA_S-to-A restored the capacity of the mutant to produce both PGLs (Fig. 4A, cf. lanes 2 and 3) and PDIMs (Fig. 4B, cf. lanes 2 and 3). Overall, the findings of the mutational analysis in M. marinum are in line with the idea that the Pks15/1-to-PpsA PHPA intermediate transfer takes place by the PHPA-AMP ligase-dependent pathway outlined in pathway A of Fig. 3. The results also suggest that direct capture of the PHPA intermediates thioesterified to the ACP domain of Pks15/1 by the KS domain of PpsA (Fig. 3, pathway B) is not a transfer mechanism of physiological relevance in M. marinum.

FIG 4.

Inactivation of PpsA's ACP_L domain leads to a PGL⁻ PDIM⁻ phenotype in M. marinum. Radio-TLC analysis of ¹⁴C-labeled PGLs (A) and ¹⁴C-PDIMs (B) from M. marinum wt + pCP0 (lane 1), M. marinum ppsA_S-A + pCP0 (lane 2), and M. marinum ppsA_S-A + pCP0-ppsA (lane 3). The wild-type (wt) and mutant M. marinum strains carried the vector pCP0 so they could be cultured in the same kanamycin-containing medium used for the complemented M. marinum ppsA_S-A + pCP0-ppsA strain. TLC solvent systems used are indicated. CHCl₃, chloroform; MeOH, methanol; PE, petroleum ether; Et₂O, diethyl ether.

Mutational analysis in M. marinum identifies FadD29 as a PHPA-AMP ligase candidate.

The mechanistic model proposed in pathway A (Fig. 3) includes an acyl-AMP ligase competent to activate and load the PHPA intermediates onto the ACP_L domain of M. marinum PpsA (steps A2 and A3, respectively). Aside from the p-hydroxybenzoic acid-specific adenylation domain of M. marinum FadD22 (35, 36), there are three conserved acyl-AMP ligases encoded in the PGL/PDIM biosynthetic gene cluster, i.e., FadD26, FadD28, and FadD29 (13, 16, 25, 37, 43, 52) (Fig. 2). Recent studies in the M. tuberculosis complex have led to the proposal that FadD29 adenylates and loads PHPAs onto PpsA in M. tuberculosis (44), but the idea has not been experimentally explored.

To seek further support for the mechanistic model proposed in pathway A (Fig. 3), we undertook a systematic mutational analysis to conclusively establish the involvement of fadD22, fadD26, fadD28, and fadD29 in PGL and PDIM production in M. marinum and inform the identification of a PHPA-AMP ligase candidate in a nontuberculous mycobacterial species. To our knowledge, the involvement of M. marinum fadD22 and M. marinum fadD29 in the production of PDIMs and PGLs has not been probed by mutational analysis. M. marinum mutants with a transposon insertion in the promoter region of fadD26 or in fadD28 were recently shown to have defects in PGL and PDIM production, but potentially confounding polar effects produced by the transposon upstream of fadD26 on other genes of the pathway (ppsA to papA5 genes [Fig. 2]) preclude unequivocal gene-to-function assignment for fadD26 (20). To conclusively probe the involvement of the four M. marinum acyl-AMP ligases encoded in the PGL/PDIM biosynthetic gene cluster in PGL and PDIM production, we engineered four mutants (M. marinum ΔfadD22, M. marinum ΔfadD26, M. marinum ΔfadD28, and M. marinum ΔfadD29), each with an unmarked, in-frame deletion in one of the four acyl-AMP ligase genes. We then examined the ability of the M. marinum mutants to produce PGLs and PDIMs by radio-TLC analysis.

The analysis of the PGL and PDIM production capacity of the fadD gene mutants revealed that deletion of fadD28 led to a PGL⁻ PDIM⁻ phenotype (Fig. 5, cf. lanes 5a and 6a and cf. lanes 5b and 6b), deletion of fadD26 produced selective loss of PDIMs (Fig. 5, cf. lanes 1a and 4a and cf. lanes 1b and 3b), and deletion of fadD22 and fadD29 led to selective loss of PGLs (Fig. 5, cf. lanes 1a and 2a and cf. lanes 1b and 2b for fadD22 and cf. lanes 5a and 8a and cf. lanes 5b and 8b for fadD29). We also constructed and analyzed four corresponding genetic complementation control strains (M. marinum ΔfadD22 + pCP0-fadD22, M. marinum ΔfadD26 + pCP0-fadD26, M. marinum ΔfadD28 + pCP0-fadD28, and M. marinum ΔfadD29 + pCP0-fadD29). Each of these control strains carried a pCP0-based plasmid expressing the specific fadD gene deleted from the genome of the host strain. Radio-TLC analysis demonstrated that episomal expression of the fadD gene reasonably restored (fully or partially) the PGL and/or PDIM production capacity of each of the M. marinum mutants (Fig. 5, lanes 3a, 7a, 9a, 4b, and 7b). The complementation controls indicate that none of the deletions exerted a confounding polar effect preventing the functional assignment of the fadD genes. Overall, the findings of the mutational analysis of fadD genes conclusively demonstrate the specific roles of fadD22, fadD26, fadD28, and fadD29 in PGL and/or PDIM production in M. marinum and, in conjunction with the available biochemical information establishing the function of the p-hydroxybenzoic acid-adenylating FadD22 protein (35, 36, 44), point at M. marinum FadD29 as the likely PHPA-specific AMP ligase involved in PHPA intermediate activation and loading onto PpsA in M. marinum. The conclusions supported by our mutational analysis in the nontuberculous mycobacterial species M. marinum parallel those derived from experiments in the M. tuberculosis complex (35, 44, 59).

FIG 5.

Mutational analysis points at M. marinum FadD29 as a PHPA-AMP ligase candidate. Radio-TLC analysis of ¹⁴C-labeled PGLs (A) and ¹⁴C-PDIMs (B) from M. marinum wt + pCP0 (lanes 1a, 5a, 1b, 5b), M. marinum ΔfadD22 + pCP0 (lanes 2a and 2b), M. marinum ΔfadD22 + pCP0-fadD22 (lane 3a), M. marinum ΔfadD26 + pCP0 (lanes 4a and 3b), M. marinum ΔfadD26 + pCP0-fadD26 (lane 4b), M. marinum ΔfadD28 + pCP0 (lanes 6a and 6b), M. marinum ΔfadD28 + pCP0-fadD28 (lanes 7a and 7b), M. marinum ΔfadD29 + pCP0 (lanes 8a and 8b), and M. marinum ΔfadD29 + pCP0-fadD29 (lane 9a). The wild-type (wt) and mutant M. marinum strains carried the vector pCP0 so they could be cultured in the same kanamycin-containing medium used for the complemented strains. The TLC solvent systems used are as described for Fig. 4.

An in vitro-reconstituted M. marinum FadD29-PpsA system displays PHPA intermediate activation, transfer, extension, and reduction capacity.

Our mutational analysis in M. marinum supports the view that the Pks15/1-to-PpsA PHPA intermediate transfer takes place via the PHPA-AMP ligase-dependent pathway outlined in Fig. 3 (pathway A) and that FadD29 activates and loads PHPAs onto PpsA in M. marinum. We sought to further probe this mechanistic model by evaluating the ability of M. marinum FadD29 and M. marinum PpsA to functionally cooperate in vitro to produce an expected PHPA extended product. In these experiments, we utilized purified M. marinum FadD29 and PpsA proteins recombinantly produced in E. coli. Since PpsA was needed in its phosphopantetheinylated form, the synthase was coexpressed with the phosphopantetheinyl transferase Sfp to increase posttranslational modification stoichiometry. Sfp is a robust enzyme from Bacillus subtilis (60) that we have previously used to phosphopantetheinylate recombinant M. marinum FadD22 and Pks15/1 proteins (35, 36).

We sought to use a predicted physiological PHPA substrate in these in vitro experiments. To this end, we used the synthetic PHPA substrate 21-(4-hydroxyphenyl)henicosanoic acid (s-PHPA; C₂₇H₄₆O₃), which was synthesized by a novel seven-step procedure (described in Fig. S4 in the supplemental material). The s-PHPA substrate corresponds to the released form of one of the most abundant PHPA intermediates produced by the M. marinum FadD22-Pks15/1 system in vitro and in vivo (36). By the canonical polyketide biosynthetic pathway predicted for PpsA based on its domain composition (13, 37) (Fig. 3), the s-PHPA substrate would be expected to be extended by two carbons via a KS domain-dependent decarboxylative Claisen condensation and to undergo ketoreductase (KR) domain-dependent β-keto group reduction to form the corresponding C29 acyl-PpsA thioester intermediate (Fig. 3). The competency of an in vitro M. marinum FadD29-PpsA system to catalyze the formation of this predicted s-PHPA-derived product was investigated by high-resolution LC-MS analysis of the extracted reaction product chemically released from the synthase by alkaline hydrolysis. Representative results of these studies are depicted in Fig. 6 and Table 1.

FIG 6.

Extension of a synthetic PHPA substrate by the M. marinum FadD29-PpsA in vitro system. Extracted ion chromatogram and isotopic distributions for the synthetic PHPA substrate (s-PHPA) and the extended s-PHPA product (e-s-PHPA) in the complete reaction mixture (A) and extracted ion chromatograms from two representative negative-control reaction mixtures, i.e., one containing no malonyl-CoA (B) and the other no s-PHPA (C). The retention time (RT) is indicated in the chromatograms. The dashed-line boxes in the observed isotopic distributions indicate the calculated relative species abundance, which is in agreement with the experimental data.

TABLE 1.

LC-MS analysis of e-s-PHPA formation

Compound	Formula	Calculated [(M-H+)]− m/z	Reaction mixture	Experimental [(M-H+)]− m/z	Retention time (min)
s-PHPA	C27H46O3	417.3374	Complete	417.3375	8.97
			No Mal-CoA	417.3374	8.97
			No NADPH	417.3373	9.01
			No ATP	417.3374	8.97
			No s-PHPA	NDa
			No FadD29	417.3373	9.06
			No PpsA	417.3373	9.07
e-s-PHPA	C29H50O4	461.3636	Complete	461.3630b	8.45
			No Mal-CoA	ND
			No NADPH	ND
			No ATP	ND
			No s-PHPA	ND
			No FadD29	ND
			No PpsA	ND

^aND, not detected.

^bppm = −1.37. Replicates rendered comparable experimental [(M-H⁺)]⁻ mass data, e.g., 461.3633 (ppm = −0.83), 461.3632 (ppm = −1.07).

Gratifyingly, in vitro incubation of FadD29 and PpsA in the presence of s-PHPA substrate, malonyl-CoA (extender unit donor), ATP (for s-PHPA activation), and NADPH (for β-keto group reduction) led to the formation of a distinct s-PHPA-derived extended product (e-s-PHPA). The experimental mass for the e-s-PHPA product matched that calculated for 3-hydroxy-23-(4-hydroxyphenyl)tricosanoic acid (C₂₉H₅₀O₄; calculated m/z [M-H⁺]⁻ of 461.3636; experimental m/z [M-H⁺]⁻ of 461.3630), which is the expected released product after s-PHPA extension and reduction by PpsA (Fig. 6A; Table 1). Conversely, the e-s-PHPA product was not detected in control reaction mixtures lacking a single enzyme (FadD29 or PpsA) or a single substrate (malonyl-CoA, NADPH, ATP, or s-PHPA) (Table 1). The unreacted s-PHPA substrate was also detected in the complete reaction mixture (C₂₇H₄₆O₃; calculated m/z [M-H⁺]⁻ of 417.3374; experimental m/z [M-H⁺]⁻ of 417.3375) (Fig. 6A; Table 1). As expected, the s-PHPA substrate was detected by LC-MS in all control reaction mixtures lacking one of the enzymes or a substrate other than s-PHPA (e.g., malonyl-CoA [Fig. 6B]) but not in the control reaction mixture in which s-PHPA was omitted (Fig. 6C). Formation of the e-s-PHPA product was not observed in reaction mixtures in which FadD29 was replaced by recombinant M. marinum FadD26, although FadD26 was able to load a model fatty acid substrate (dodecanoic acid) on PpsA in vitro (not shown). The competency of M. marinum FadD26 to load PpsA with a fatty acyl starter unit is in line with the in vitro activity previously demonstrated for M. tuberculosis FadD26 (37, 52). The inability of M. marinum FadD26 to replace M. marinum FadD29 in vitro is consistent with our finding that M. marinum ΔfadD29 is PGL deficient despite having fadD26 (Fig. 5A). Altogether, the results with the reconstituted M. marinum FadD29-PpsA system demonstrate that M. marinum FadD29 and M. marinum PpsA cooperate in vitro to produce an expected PHPA extended product and support the genetic studies identifying FadD29 as the PHPA-AMP ligase required for PGL biosynthesis in M. marinum. Overall, the M. marinum FadD29-PpsA in vitro system appears to be competent to catalyze PHPA activation, loading, extension, and reduction in line with the model proposed in Fig. 3. The M. marinum FadD29-PpsA in vitro system along with the M. marinum FadD22-Pks15/1 in vitro system for biosynthesis of PHPAs that we have developed previously (36) lays a valuable foundation for further mechanistic analysis of the enzymatic machinery involved in PGL biosynthesis in M. marinum.

The M. marinum FadD29-PpsA functional partnership established by our in vitro studies represents the first acyl-AMP ligase and type I PKS partnership for acyl starter unit activation and PKS loading established in nontuberculous mycobacteria. Three analogous partnerships have been demonstrated in M. tuberculosis. These are the FadD26-PpsA partnership noted above for PDIM production (37, 52), a FadD32-Pks13 partnership that takes place during mycolic acid biosynthesis (52, 61, 62), and a FadD30-Pks6 partnership believed to be required for production of novel polar lipids (52, 63). To our knowledge, however, the partially overlapping PGL/PDIM biosynthetic pathways provide the first example of two distinct acyl-AMP ligases (i.e., FadD29 and FadD26) loading the same type I PKS (i.e., PpsA) with two alternate starter units (i.e., PHPAs and fatty acids). This bimodal loading strategy allows the bacterium to use the PpsABCDE megasynthase system to generate two structurally different products. Interestingly, recent host-pathogen interaction studies suggest that M. marinum PGLs and M. marinum PDIMs have different roles within a complex immune evasion mechanism (15). It will be interesting to investigate whether the FadD29-PpsA versus FadD26-PpsA alternative partnership is utilized by the pathogen as a control point to modulate the relative abundance of PGLs and PDIMs in the cell.

This paper is dedicated to the memory of our wonderful colleague Clifford E. Soll, who recently passed away.