Minimization of the Thiolactomycin Biosynthetic Pathway Reveals that the Cytochrome P450 Enzyme TlmF is Required for Five-membered Thiolactone Ring Formation

Abstract

Thiolactomycin (TLM) belongs to a class of rare and unique thiotetronate antibiotics inhibiting bacterial fatty acid synthesis. Although this group of natural product antibiotics was first discovered over 30 years, the study of TLM biosynthesis remains in its infancy. We recently discovered the biosynthetic gene cluster (BGC) for TLM from the marine bacterium Salinispora pacifica CNS-863. Here, we report the investigation of TLM biosynthetic logic through mutagenesis and comparative metabolic analyses. Our results reveal that only four genes (tlmF, tlmG, tlmH, and tlmI) are required for the construction of the characteristic γ-thiolactone skeleton of this class of antibiotics. We further showed that the cytochrome P450 TlmF does not directly participate in the sulfur insertion and C-S bond formation chemistry but rather in the construction of the 5-membered thiolactone ring, as upon its deletion, we observed the alternative production of the 6-membered δ-thiolactomycin. Our findings pave the way for future biochemical investigation of the biosynthesis of this structurally unique group of thiotetronic acid natural products.

Graphical abstract

The minimal biosynthetic gene cluster required for the production of the antibiotic thiolactomycin has been identified through heterologous expression and mutagenesis. A cytochrome P450 was confirmed to play an important role in forming the structurally unique 5-membered thiolactone ring, as upon its deletion, a 6-membered δ-thiolactomycin was instead synthesized.

Polyketides (PKs) and nonribosomal peptides (NRPs) represent two of the most important classes of natural products (NPs) with life-saving pharmaceutical properties.^[1] The corresponding biosynthetic machineries, polyketide synthase (PKS) and non-ribosomal peptide synthetase (NRPS), share striking structural and catalytic similarities that support the hybridization of PKS and NRPS proteins for the assembly of structurally unusual NPs in nature. Many important NPs, such as the immunosuppressant rapamycin,^[2] the genotoxin colibactin,^[3] and the anticancer agent salinosporamide A,^[4] are produced by PKS-NRPS hybrid biosynthetic gene clusters (BGCs). Typically, PKS-NRPS hybrid NPs are often heavily modified during and/or after chain elongation on the assembly lines, yielding bioactive NP scaffolds with a variety of heterocyclic moieties.^[5]

Thiolactomycin (TLM, 1) (Figure 1), a structurally unique antibiotic, was originally isolated from a soil actinomycete identified as Nocardia sp. ATCC 31319.^[6] It consists of an unprecedented five-membered thiolactone ring system and a methyl-branched butadienyl side chain. Since its original discovery in 1982, 1 and analogues have also been identified from species of Streptomyces and Salinispora.^[7] This group of antibiotics has a wide-range antimicrobial activity against both Gram-positive and Gram-negative bacteria.^[8] The mechanism of action of 1 has been extensively investigated, revealing that it selectively targets the bacterial type II fatty-acid synthase (FAS-II) through the reversible inhibition of the β-ketoacyl-ACP synthases (KAS).^[9] In a mouse model, 1 was more effective than carbenicillin in treating experimental acute urinary tracts infected with Serratia marcescens.^[10] Furthermore, 1 has emerged as an attractive lead for developing new anti-tuberculosis,^[11] anti-malarial^[12] and anti-trypanosomal drugs.^[13]

Figure 1.

Structures of thiolactomycins 1-4, the tlm biosynthetic gene cluster (top) and the engineered minimal tlm BGC under control of the ermE promoter (bottom), and HPLC profiles of S. coelicolor M1152 mutant extracts at 239 nm.

Although this group of antibiotics, especially 1, has been extensively studied in many aspects, such as mode of action,^[14] pharmacological activity improvement,^[15] and total synthesis^[16] over the last 30 years, the study of TLM biosynthesis remains in its infancy. Isotope feeding studies performed in Nocardia sp. by Reynolds and colleagues^[17] suggested that 1 is derived from a PK pathway involving one acetate-derived starter unit, three propionate extender units, and a sulfur atom originating from cysteine. Recently, we discovered the BGCs for 1-4 (Figure 1) and the related thiotetroamide (TTM) pathway from the marine bacterium Salinispora pacifica CNS-863 and terrestrial bacterium Streptomyces afghaniensis NRRL5621,^[7e] respectively. Simultaneously, others identified similar BGCs in four different strains.^[18] These findings revealed that 1 and its analogues are constructed via an unconventional PKS-NRPS hybrid assembly line, indicating an unusual biosynthetic logic at play. Here we report the exploration of the biosynthesis of the tlm BGC from S. pacifica CNS-863 through genetic and analytical strategies. We show that four genes, tlmF-I, are the minimal set of genes needed for producing 1-4 (Figure 1). Furthermore, we identified a 6-membered shunt product upon genetic disruption of the annotated cytochrome P450 (CYP450) encoded by tlmF, indicating that TlmF is required for the formation of TLM's distinctive 5-membered thiolactone ring.

The tlm BGC from S. pacifica strains consists of nine open reading frames (tlmA-I), which are assumed to code for biosynthesis, resistance, and regulation (Figure 1). We previously reported that the heterologous expression of the tlm BGC in the antibiotic production super-host Streptomyces coelicolor M1152 led us to identify four compounds, including 1 and three other analogues (2-4) (Figure 1 and Figure S1 in the Supporting Information).^[7e] To investigate which genes are essential for production of TLM's distinctive heterocycle, we interrogated the functions of most of the genes in the cluster. We first deleted tlmA, encoding a putative type II thioesterase (TE), on the plasmid pMXT13^[7e] containing the tlm BGC via lambda Red–mediated recombination.^[19] We then heterologously expressed the mutated plasmid pMXT13ΔtlmA in S. coelicolor M1152 and examined the extract of mutant culture by HPLC and high resolution LCMS analyses. The result showed that the tlmA deletion mutant still had the ability to produce 1-4, however, at a greatly reduced level (Figure 1 and Figure S1). Type II TEs are hydrolytic enzymes often encoded within PKS and NRPS gene clusters where they often perform important roles in maintaining the efficiency of the assembly line through exclusion of aberrant residues, selection of accurate substrates, and release of terminal products.^[20] Our data thus strongly suggests that the product of tlmA is a type II TE that is responsible for maintaining the productivity of the modular PKS-NRPS of tlm. Next, we concentrated on genes located between tlmB and tlmD of the tlm BGC. This region encodes a crotonyl-CoA reductase/carboxylase (TlmB), a 3-hydroxybutyryl-CoA dehydrogenase (TlmC), and a LuxR family transcriptional regulator (TlmD), which are not likely to be critical for the assembly of 1-4. The first two enzymes are expected to participate in the generation of the precursor ethylmalonyl-CoA for 2-4 assembly,^[21] while LuxR family transcriptional regulators often act as transcriptional activators.^[22] To verify our assumption as well as further minimize the tlm BGC, we deleted tlmA-D on pMXT13 to generate the mutant plasmid pMXT13ΔtlmA-D. The resulting plasmid was integrated in S. coelicolor M1152, and the culture extract of the mutant was analyzed by HPLC and high resolution LCMS. We only observed a small amount of 1 by HPLC, while 2-4 were only detected by LCMS (Figure 1 and Figure S1), thereby clearly supporting the nonessential roles of TlmB, TlmC, and TlmD in the assembly of 1-4.

We previously reported that the FabB/F homolog TlmE functions as the self-resistance element for 1-4 biosynthesis.^[7e] In order to verify that tlmE is not involved in the formation of the products, we decided to remove the region tlmA to tlmE. However, we realized that direct removal of tlmE may cause polar mutation. Therefore, we cloned the aac(3)IV-ermE cassette from the plasmid pMXT19^[7e] to replace the designated region generating an operon from tlmF to tlmI (pMXT13tlmF-I), which is exclusively under control of the constitutive ermE promoter (Figure 1). Remarkably, expression of this minimized BGC, tlmF-I, still produced normal levels of 1 but very small quantities of 2-4 (Figure 1 and Figure S1). The dramatically reduced production of 2-4 can be explained by the absence of the two dedicated ethylmalonyl-CoA generating enzymes, TlmB and TlmC. We further deleted the gene tlmF, encoding a CYP450, to generate a three-gene (tlmG-I) BGC promoted by the ermE promoter (Figure 2). As expected, we did not observe any production of 1-4 in the mutant strain (Figure 2). This result strongly supported the hypothesis that the four genes tlmF-I compose the minimal BGC for 1-4 production.

Figure 2.

Identification and structure of δ-thiolactomycin (5). (a) Organization of the three-gene gene cluster for 5 production. (b) HPLC profiles of extracts from S. coelicolor M1152 and S. coelicolor M1152/pMXT13tlmG-I. (c) Key HMBC and COSY correlations of compound 5.

Surprisingly, we detected a new product from this mutant by HPLC with a distinct UV chromatogram (Figure 2). Upon purification from a 2 L culture of the mutant strain, the HRESIMS of this compound (5) exhibited a protonated molecular ion peak at m/z 213.0940 (Figure S2), consistent with a molecular formula of C₁₁H₁₆O₂S (calcd 213.0944 [M + H]⁺, 1.9 ppm). The analysis of extensive NMR spectra (Supplementary Figure S3-S6) indicated that although 5 contained a thiolactone moiety, neither the characteristic Δ^5,7-conjugated double bond of the side chain at C-4 nor the enol unit at C-3 were present as in other thiolactomycin analogues with a γ-thiolactone ring. Instead, only one olefin quartet at δ_H 5.44 (H-7) was observed, along with three resonance signals at δ_H 3.03 (H-4), 3.92 (H-2), and 4.54 (H-5) attributed to methines on an oxygenated aliphatic ring (Table S3 and Figure S3). These chemical shifts suggested a significantly different cyclization pattern of the thiolactone ring in 5. A key HMBC correlation between H-5 and C-1 indicated the linkage of C-5 to the sulfur atom (Table S3 and Figure S6), and thus established the core skeleton of 5 as a six-membered thiolactone ring in comparison with the five-membered thiolactone core present in other thiolactomycin antibiotics. A COSY correlation between the H-7 olefin quartet and the tertiary H-8 methyl doublet (δ_H 1.67) suggested the formation of a Δ⁶-double bond in the side chain at C-5 of compound 5 instead of the Δ^5,7-conjugated double bond (Figure S4). This assignment was confirmed by the HMBC correlations of H-7 to C-5 and H-8 to C-6 (Figure S6). Therefore, the planar structure of compound 5 was established. Interestingly, an enol form at C-3 of compound 5, thiolactomycin δ, was previously reported from a Nocardia sp. strain, and the structure was established using crystallography without any spectroscopic data being reported.^[7c]

The remaining biosynthetic genes embedded in the tlm BGC are co-transcribed as indicated by the overlap of start and stop codons. This subcluster encodes two modules of PKSs (TlmG and TlmH) and one module of a NRPS (TlmI). In-frame deletion of tlmG completely abolished the production of 1-5 (Figure 1 and Figure S1), suggesting tlmG-I are the minimal set of genes for thiolactone ring production. We therefore propose that TlmG, a trans-AT KS-ACP di-domain, initiates polyketide chain priming with an acetate-derived starter unit that is extended by the iterative PKS module TlmH (KS-AT-DH-KR-ACP-C) with two or three branched chain malonates to form the carbon backbone of 1 (Figure 3). Leadlay and co-workers recently presented chemical and biochemical evidence confirming that the second module of the tlm assembly line indeed encodes an unusual iterative PKS.^[24] Finally, the tri-domain NRPS TlmI (A-PCP-TE) appears to be involved in the incorporation of a sulfur atom from a carrier protein-bound cysteine residue, since the adenylation domain has been predicted to activate a cysteine residue (Figure 3). This proposal is consistent with the previous isotope study that revealed cysteine as the source of sulfur.^[17]

Figure 3.

Model for the biosynthesis of five- and six-membered thiolactomycin via TlmF P450-dependent (route a) and independent (route b), respectively.

Together, these findings suggest a new mechanism for TLM biosynthesis and the intriguing role of TlmF in the unprecedented formation of the thiolactone ring. CYP450s are among the most common and versatile tailoring enzymes in biosynthesis where they have been shown to catalyze various reactions, including hydroxylation, epoxidation, decarboxylation, ring-coupling, C-C bond cleavage, and carbocation formation.^[23] Leadlay and co-workers previously reported the gene inactivation of stuD1 and tlmD1, two homologs of the CYP450-encoding gene tlmF, in two separate TLM-producing strains, that led to the elimination of TLM production without detecting any additional products.^[18] Thus, our findings not only support the essential role of the CYP450 TlmF in the formation of the 5-membered thiolactone ring, but the identification of compound 5 further implicates that the insertion of sulfur into TLM is independent of TlmF.

We propose that the TlmH-bound polyketide intermediate condenses with the cysteinyl-S-TlmI to eventually form the polyketide thiocarboxylic acid intermediate 6 (Figure 3). While the mechanism for the desulfuration reaction has not yet been established, Leadlay and co-workers suggest that the sulfur atom is incorporated by the action of a cysteine desulfurase along with a sulfur transferase via primary metabolic reactions.^[18] Following the generation of 6, we envision that the competing routes, a or b, proceed divergently in the presence or absence of the CYP450 TlmF (Figure 3). In the presence of the TlmF P450, formation of the 5-membered TLM 1 may proceed via a number of oxidative routes, including an epoxide intermediate as shown via route a. Alternatively, in the absence of the TlmF P450, 6 is structurally configured to form the 6-membered thiolactone ring in 5 by attack at C-5 via route b. The exact mechanism involved in the formation of the thiolactone ring is actively under investigation.

In summary, we report the identification of the minimal set of genes involved in thiolactomycin biosynthesis and established that the CYP450 TlmF is involved in the construction of the five-membered thiolactone ring. This work paves the way for future in vitro studies with recombinant proteins to explore the mechanistic details of this unprecedented ring formation enzymology to an important group of antibiotic polyketides.

Experimental Section

General experimental procedures

All chemicals were purchased from Fisher Scientific and Sigma-Aldrich and used as such unless stated otherwise. PCR was performed with the PrimeSTAR HS kit with GC Buffer (Takara Bio USA, Inc.). All solvents used were HPLC grade solvents or higher. Analytical HPLC analyses were conducted with an Agilent 1200 HPLC system with diode array detection connected to a Phenomenex Luna C18 reversed-phase HPLC column (5 μm, 250 mm × 4.6 mm). High resolution LC-MS analyses were performed with Agilent 6530 Accurate-Mass Q-TOF MS coupled to an Agilent 1260 LC system. Preparative HPLC was carried out by using an Agilent 218 purification system equipped with a Pro-Star 410 automatic injector, an Agilent ProStar UV-Vis dual wavelength detector, and a 440-LC fraction collector connected to Phenomenex Luna C18(2), 10.0×250 mm, 5 mm column. NMR data were acquired at the UCSD Skaggs School of Pharmacy and Pharmaceutical Sciences NMR Facility on a 600 MHz Varian NMR spectrometer (Topspin 2.1.6 software, Bruker) with a 1.7 mm cryoprobe.

Generation of tlm mutants in S. coelicolor M1152

An apramycin resistance (aac(3)IV) cassette was amplified from plasmid pMXT19 (Table S1)^[7e] using primer pairs tlmA-KO-F/tlmA-KO-R, tlmA-D-KO-F/tlmA-D-KO-R, tlmA-E-KO-F/tlmA-E-KO-R, tlmA-F-KO-F/tlmA-F-KO-R and tlmG-KO-F/tlmG-KO-R (Table S2). The genes were replaced in E. coli BW25113/pIJ790/pMXT13 by using the PCR targeting system (Table S1).^[19] Resulting plasmid were confirmed by restriction analysis. If necessary, excision of the cassette was performed in E. coli BT340 (Table S1) taking advantage of the FLP/FRT recognition sites adjacent to the resistance cassette. Positive plasmids were screened for their apramycin sensitivity and verified by restriction analysis. Plasmid pMXT13ΔtlmA, pMXT13ΔtlmG, pMXT13ΔtlmA-D, pMXT13tlmF-I (ΔtlmA-E), and pMXT13tlmG-I (ΔtlmA-F) were transferred into E. coli ET12567 (Table S1) and introduced into S. coelicolor M1152 (Table S1) by triparental intergeneric conjugation with the help of E. coli ET12567/pUB307. ^[25] Kanamycin resistance clones were selected, confirmed by PCR, and designated as S. coelicolor M1152/pMXT13ΔtlmA, S. coelicolor M1152/pMXT13ΔtlmG, S. coelicolor M1152/pMXT13ΔtlmA-D, S. coelicolor M1152/pMXT13tlmF-I, and S. coelicolor M1152/pMXT13tlmG-I.

Fermentation, extraction and detection of 1-5

TSB broth (3 mL) with 5 μL spore suspension of S. coelicolor M1152 or a derivative thereof was incubated for two days at 30 °C and 220 rpm. One mL of the preculture was inoculated into 50 mL of the Streptomycete Production Medium (SPM) (1 L: 10 g soytone, 10 g soluble starch, and 20 g D-maltose) with components purchased from BD Biosciences, US. After incubation of 7 days at 30 °C and 220 rpm, the culture supernatant was adjusted to pH 4.0 and subsequently extracted with an equal volume of ethyl acetate (EtOAc). The organic phase was evaporated, dissolved in MeOH (1 mL), and filtered through Acrodisc MS PTFE Syringe filters (Pall Inc., Ann Arbor, MI, USA) prior to LC-MS analysis. For LC-MS analysis, a 10 μL aliquot of the EtOAc-soluble extract was injected onto a Phenomenex Luna C18 reversed-phase HPLC column (5 μm, 250 mm × 4.6 mm i.d.) and analyzed with an Agilent 6530 Accurate-Mass LC-MS coupled to an Agilent 1260 LC system by a gradient elution (A: CH₃CN with 0.1% formic acid; B: H₂O with 0.1% formic acid: 35–70% A over 23 min, 70–100% A from 23 to 28 min, and 100% A from 28 to 33 min; 0.7 mL/min). Q-TOF MS settings during the LC gradient were as follows: Acquisition—mass range m/z 100–1700, MS scan rate 1s^-1, MS/MS scan rate 2s^-1, fixed collision energy 20 eV; Source—gas temperature 300 °C, gas flow 11 Lmin^-1; Nebulizer 45 psig, ion polarity positive; Scan source parameters-VCap 3000, Fragmentor 100, Skimmer 65, OctopoleRFPeak 750. The MS was autotuned using Agilent tuning solution in positive mode before each measurement. LC (DAD) data were analyzed with ChemStation software (Agilent), and MS data were analyzed with MassHunter software (Agilent).

Isolation of δ-thiolactomycin (5)

An EtOAc-soluble extract (ca. 50 mg) of the cell culture (2 L) of S. coelicolor M1152/pMXT13tlmG-I was obtained using the same fermentation and extraction methods mentioned above, followed by passage over a reversed-phase C18 silica gel column. The polar fraction was removed by elution with 5% MeOH–95% H₂O, and then the column was eluted using 95% MeOH-5% H₂O, with the latter fraction being evaporated and resuspended in MeOH (1 mL) prior to preparative HPLC separation on a Phenomenex Synergi hydro-RP column (10 μm, 250 mm × 21.2 mm i.d.), along with an Agilent Technologies system composed of a PrepStar pump, a ProStar 410 autosampler, and a ProStar 325 UV detector (Agilent Technologies, Inc. Santa Clara, USA). The elution condition was MeCN–H₂O (45:65, each containing 0.1% TFA) at a flow rate of 9.0 mL/min. The fraction collected at t_R 26.2 min were subsequently subjected to a semi-preparative HPLC purification, using a Phenomenex Luna C18 reversed-phase HPLC column (5 μm, 250 mm × 10 mm i.d.) with isocratic elution (40% MeCN–60% H₂O, each solvent contains 0.1% TFA; flow rate 2.5 mL/min), to yield 5 (2.0 mg, t_R = 27.8 min).

δ-thiolactomycin (5)

colorless resin; [α]²⁰_D +86 (c 0.1, MeOH); UV (MeOH) λ_max (log ε) 247 (3.83), 285 (3.54) nm; ¹H and deduced ¹³C NMR data shown in Supplementary Table S3; HRESIMS obsd m/z 213.0940 [M+H]⁺ (calcd for C₁₁H₁₇O₂S⁺, 213.0944).

Abbreviations

KS

Ketosynthase

AT

Acyltransferase

DH

Dehydratase

KR

Ketoreductase

ACP

Acyl Carrier Protein

C

Condensation Domain

A

Adenylation Domain

PCP

Peptide Carrier Protein

TE

Thioesterase