Operon for Biosynthesis of Lipstatin, the Beta-Lactone Inhibitor of Human Pancreatic Lipase

Abstract

Lipstatin, isolated from Streptomyces toxytricini as a potent and selective inhibitor of human pancreatic lipase, is a precursor for tetrahydrolipstatin (also known as orlistat, Xenical, and Alli), the only FDA-approved antiobesity medication for long-term use. Lipstatin features a 2-hexyl-3,5-dihydroxy-7,10-hexadecadienoic-β-lactone structure with an N-formyl-l-leucine group attached as an ester to the 5-hydroxy group. It has been suggested that the α-branched 3,5-dihydroxy fatty acid β-lactone moiety of lipstatin in S. toxytricini is derived from Claisen condensation between two fatty acid substrates, which are derived from incomplete oxidative degradation of linoleic acid based on feeding experiments. In this study, we identified a six-gene operon (lst) that was essential for the biosynthesis of lipstatin by large-deletion, complementation, and single-gene knockout experiments. lstA, lstB, and lstC, which encode two β-ketoacyl–acyl carrier protein synthase III homologues and an acyl coenzyme A (acyl-CoA) synthetase homologue, were indicated to be responsible for the generation of the α-branched 3,5-dihydroxy fatty acid backbone. Subsequently, the nonribosomal peptide synthetase (NRPS) gene lstE and the putative formyltransferase gene lstF were involved in decoration of the α-branched 3,5-dihydroxy fatty acid chain with an N-formylated leucine residue. Finally, the 3β-hydroxysteroid dehydrogenase-homologous gene lstD might be responsible for the reduction of the β-keto group of the biosynthetic intermediate, thereby facilitating the formation of the unique β-lactone ring.

INTRODUCTION

Lipstatin (Fig. 1, structure 1) was originally isolated from fermentation broth of Streptomyces toxytricini as a very potent, selective, irreversible inhibitor of human pancreatic lipase (1, 2). The saturated derivative of lipstatin, tetrahydrolipstatin (Fig. 1, structure 2), commonly known as orlistat, is currently the only available FDA-approved oral drug for long-term treatment of obesity because of its cardiovascular safety and its benefits for diabetes control in obese patients (3, 4). Orlistat blocks the activity of human pancreatic and gastric lipases, thereby reducing the absorption of fat from diets (5, 6). Orlistat was also found to exhibit antitumor activity, by virtue of its ability to inhibit the thioesterase domain of fatty acid synthase of tumor cells (7).

FIG 1.

Structures of lipstatin (1), tetrahydrolipstatin (2), and esterastin (3).

Lipstatin features a 2,3-trans-disubstituted β-propiolactone with two linear alkyl chains of C₆ at the α-site and C₁₃ at the β-site, respectively (i.e., an α-branched 3,5-dihydroxy fatty acid β-lactone backbone), and an N-formyl-l-leucine group attached as an ester to the 5-hydroxy of the long chain of the α-branched fatty acid backbone (8, 9) (Fig. 1, structure 1). Opening of the 4-membered ring results in an almost complete loss of the lipase-inhibitory activity of lipstatin, suggesting that the β-lactone moiety of the molecule is pivotal to its enzymatic inhibition (10). Other β-lactone natural products produced by microorganisms include esterastin (11, 12), panclicins (13), valilactone (14), and ebelactones (15, 16), which differ only in the structure of the side chains and the nature of the linked amino acids. For instance, esterastin, a potent inhibitor of lysosomal acid lipase and esterase, has an N-acetyl asparagine side chain instead of an N-formyl-leucine (Fig. 1, structure 3) (11, 12).

The biosynthesis of lipstatin in S. toxytricini has been intensively studied based on ¹⁵N, ¹³C, and deuterium labeling experiments by the Bacher group (17–21). The α-branched fatty acid moiety of lipstatin is derived from Claisen condensation between octanoyl coenzyme A (octanoyl-CoA) and 3-hydroxytetradeca-5,8-dienoyl-CoA, both of which are obtained from incomplete β-oxidation of linoleic acid (17–21). In addition, the H-2 atom of the octanoic acid precursor is displaced with the solvent proton in the final product lipstatin (20). In contrast, the two long-chain fatty acid substrates of the mycolic acids (complex α-branched β-hydroxy long-chain fatty acids) of mycobacteria and corynebacteria are synthesized by type I and type II fatty acid synthetase systems (FASI and FASII, respectively) (22). Subsequently, Pks13, a type I polyketide synthase (PKS) with multiple functional domains, takes the two substrates by attaching them to two phosphopantetheine-binding (PPB) domains and catalyzes Claisen condensation between two tethered substrates by the ketosynthase (KS) domain (23). Furthermore, when [¹³C-formyl,¹⁵N]-N-formyl-leucine was fed to S. toxytricini fermentation broth, only the ¹⁵N-leucine moiety of the double-labeled chemical was incorporated into lipstatin, suggesting that the incorporation of leucine takes place after hydrolysis of the formamide motif and that the formyl group is transferred to the l-leucine residue afterwards (21).

Here we report the identification of the lipstatin biosynthetic operon based on genome sequencing, bioinformatics analyses, and genetic manipulations. A six-gene operon, encoding homologues of two β-ketoacyl-acyl carrier protein synthases III (FabH), an acyl-CoA synthetase, a 3β-hydroxysteroid dehydrogenase, a nonribosomal peptide synthetase (NRPS), and a formyltransferase, was found to be sufficient for the production of lipstatin in Streptomyces. Furthermore, new metabolites were produced by three S. toxytricini gene deletion strains. The model of the lipstatin biosynthetic pathway is updated accordingly.

MATERIALS AND METHODS

Bacterial strains, plasmids, and culture conditions.

S. toxytricini NRRL15443 was provided by the ARS Culture Collection. Streptomyces lividans TK24 (24) was used as the host for heterologous expression of the lipstatin biosynthetic operon. Escherichia coli DH10B was used for routine cloning experiments. E. coli XL1-Blue MR (Stratagene) was used as the host for construction of the genomic cosmid library. E. coli BW25113/pIJ790 was used for λ-red-mediated recombination (25). E. coli ET12567 carrying RK2-derived conjugation helper plasmid pUZ8002 was used as the donor for E. coli-Streptomyces intergeneric conjugation (26).

pJTU2554 containing an origin of transfer (oriT) from RK2, a functional int-attP (integrase gene-attachment site of phage) region of actinomycete phage ϕC31, and two lambda phage cos sites was used as the vector for the genomic cosmid library (27). 27F11, a genomic cosmid clone containing 46.5 kb of S. toxytricini genomic DNA including the putative lipstatin biosynthetic operon, was used to construct gene deletion mutants. oriT-containing plasmid pOJ260 (28), which does not replicate in streptomycetes, was used as a suicide vector for gene replacement in S. toxytricini.

E. coli strains were grown in the Luria-Bertani (LB) broth (29) at 37°C, except for BW25113/pIJ790, which was grown at 30°C. S. lividans strains were grown on MS agar at 30°C for sporulation and conjugation (24). S. toxytricini strains were grown on International Streptomyces Project Medium 4 (ISP4; BD Biosciences, San Jose, CA, USA) for sporulation and conjugation. Fermentation seed medium (10 g of soya bean flour, 5 g of Bacto soytone, 5 ml of glycerol, 10 ml of soya oil, and 2 ml of Triton X-100 [pH 6.5] per liter) and fermentation medium (30 g of soya bean flour, 14 ml of glycerol, 1 g of Bacto soytone, 1 ml of Triton X-100, and 60 ml of soya oil [pH 7.0] per liter) (30) were used for shake flask fermentation of S. toxytricini and S. lividans strains.

Genomic sequencing and bioinformatics analyses.

S. toxytricini NRRL15443 genomic DNA was prepared as described previously (24). The genome sequence was determined by using the Roche 454 GS (FLX Titanium) sequencing platform. All reads, providing ∼25-fold genome coverage, were assembled by using Newbler 2.5.3 software (Roche Diagnostics, Branford, CT), and 212 contigs were obtained. The assembled genome sequence was annotated on a high-performance server (NF8560M2; Inspur) with the program Glimmer 3.0 for identification of protein-coding genes (31). The genes in the lipstatin biosynthetic locus were analyzed by a BLASTp search against the NCBI nonredundant protein sequence, UniProtKB/Swiss-Prot, and Conserved Domain Database (CDD) databases (32–34).

S. toxytricini NRRL15443 genomic cosmid library.

pJTU2554 vector DNA was digested by BamHI and HpaI, and the resulting cos-containing 6.8-kb and 2.7-kb fragments were gel purified. Genomic DNA of S. toxytricini NRRL15443 was partially digested with Sau3AI, and DNA fractions of 35 to 45 kb were obtained by agarose gel electrophoresis separation and gel extraction. Ligation, competent-cell preparation, library packaging, transfection, storage, and screening of the library were conducted according to standard molecular biology protocols (29). The genomic library was screened for cosmids that contained DNA of the putative lipstatin biosynthetic operon (lst) by Southern hybridization using a PCR-amplified 429-bp DNA fragment internal to the formyltransferase gene lstF. Oligonucleotides used for PCR amplification of the 429-bp DNA were lstF-F (5′-GTCGCTGGGGCTCCGCATCGT-3′) and lstF-R (5′-TCGGCGACTTCGGGTGCGTG-3′).

Construction of the large-deletion mutant strain SBT11.

SBT11, a large-fragment-deletion mutant of S. toxytricini, was generated by removing a 38.4-kb genomic DNA fragment, including the lst operon from the wild-type chromosome, via homologous recombination using a suicide construct, pHTL6. The gene displacement construct pHLTL6 was derived from pOJ260, harboring a 4.9-kb BamHI fragment and a 2.7-kb XbaI-BamHI fragment identical to the left and right ends of the S. toxytricini genomic DNA insert of cosmid 27F11, respectively. pHLTL6 was conjugated into S. toxytricini, and offspring single colonies of the exconjugants were screened for double-crossover events, indicated by a loss of apramycin resistance and PCR amplification band patterns. The resulting mutant strain, named SBT11, was verified by PCR experiments to carry the 38.4-kb large-fragment deletion (Table 1 and Fig. 2a; see also Fig. S1 in the supplemental material).

TABLE 1.

Primers used for verification of the deletion mutants of S. toxytricini NRRL15443

Mutant strain	Primer designed to verify mutant strain	Primer sequence (5′→3′)	Size of desired PCR fragment (bp)
			Mutant	Wild type
SBT11	rightArm-F	GGCTTCGGTGGTGTTCTCCC	No band	938
	rightArm-R	TGGCGTCACTCCTGGCTCCT
SBT11	Scar-F	CAGACGACGAAGCCGACC	891	No band
	Scar-R	CCTACGAGGCGATGACCC
SBT11	leftArm-F	CCGAGCAGGGTGAGGCAGAC	No band	785
	leftArm-R	GGCGACCAGAGCGTCAAGG
Δright	deRight-dgF	GGATGTGCTGCAAGGCGATTA	401	No band
	deRight-dgR	CCGAAGGCGGGTTCAAGGT
Δleft	deLeft-dgF	GGCCGTCGAAAGGTCAGTG	739	No band
	deLeft-dgR	GCCGTGCGGACATAGGAAG
ΔlstA	deA-dgF	GGATGCGGTGATGACGATGC	201	1,230
	deA-dgR	TGACGGCCCTCAGCCATTTAC
ΔlstB	deB-dgF	CGAAGGCGAGGGCGAACAGT	878	1,650
	deB-dgR	AGCGACCGCAGCCAGGGAAT
ΔlstC	deC-dgF1	GCTGGAGACGAACACGAACCG	912	3,410
	deC-dgR1	CGACTACCCCTACGCCGACC
ΔlstC	deC-dgF2	GAAGCGTTCCAGAGCGTCG	No band	602
	deC-dgR2	GCATCACCAAGGGCACCAAG
ΔlstD	deD-dgF	ACCTGCGGTGAGGTCGAAGC	709	2,314
	deD-dgR	GCAGGAGTTCCTGCACCACC
ΔlstE	deE-dgF1	GACCGACCAGACGATCCCG	772	No band
	deE-dgR1	TCCACCGAACAGCCCTTCC
ΔlstE	deE-dgF2	CGGAATACCTCCCGCACCCA	No band	652
	deE-dgR2	CGTACCTCGCGTCAGGCAACA
ΔlstF::aadA	deF-dgF	TCGTCTGGTCGGTCAACA	1,729	472
	deF-dgR	GGCGACTTCGGGTGCGTG
Δorf1	de1-dgF	TGCCCACCCGCTTCCACA	357	885
	de1-dgR	GCAGGCGAACTCCACATCCA
Δorf2	de2-dgF	CGGCTGGTGATGATGTGCG	882	1,407
	de2-dgR	AAGCAGCAGGACGGCAAGG
Δorf3	de3-dgF	TTTCTCCGGTCGTCGTTCGTC	537	1,362
	de3-dgR	CGCACCATCTGGGATTCCTGT
Δorf4::aadA	de4-dgF	AGCTGTTGGCCGTCCAGTACC	1,751	494
	de4-dgR	CCCAGGTGGCGAGGGATT
Δorf6	de6-dgF	TGCCATCAGTGCCTGTTGTCC	287	792
	de6-dgR	CGCCATCCATGCCTGTGAA

FIG 2.

Localization of the DNA sequence essential for lipstatin biosynthesis by large deletion, complementation, and heterologous expression. (a) Schematic representation of the 46.5-kb DNA region spanning the putative lst operon in the S. toxytricini wild-type (WT) strain, a large-deletion mutant (SBT11), and three integrative plasmids (27F11, pDelL, and pDelR). Thick lines represent genomic DNA. Dash lines indicate deleted regions. Wide arrows marked by letters or numbers on the top line are genes in the 19.4 kb-DNA fragment shared by three integrative plasmids, including 6 genes (lstA-F) of the putative lst operon and 8 downstream genes (orf1 to orf8). (b) Detection of lipstatin production in S. toxytricini large-deletion and complementation strains by HPLC performed on an Agilent 1100 HPLC system (210 nm) under standard conditions. Lipstatin (marked by 1) is eluted at 16 min. (c) Detection of lipstatin production in the S. lividans/27F11 heterologous expression strain. HPLC was performed on an Agilent 1260 Infinity HPLC system (210 nm). Lipstatin (marked by 1) was eluted at 13.5 min. pJTU2554 is the int-attP^ϕC31 integrative vector for construction of cosmid 27F11. ST, lipstatin standard; mAU, milli-absorbance units.

Construction of 27F11-derived integrative plasmids carrying gene deletions.

27F11-derived integrative plasmids carrying deletions in the lst operon and flanking sequences were constructed by λ-red-mediated recombination (PCR targeting technology) between 27F11 and PCR-amplified DNA fragments harboring the spectinomycin/streptomycin resistance gene aadA (24, 25). PmeI restriction sites (GAAATTTC) were set in some pairs of oligonucleotides for amplification of the aadA cassette. The aadA cassettes of some 27F11-derived plasmids were removed to generate in-frame deletion mutants, when necessary, by PmeI restriction digestion and religation after λ-red-mediated recombination (see below). Oligonucleotides for construction and verification of gene deletion mutants are listed in Tables 1 and 2, respectively.

TABLE 2.

Oligonucleotides used for PCR targeting to construct mutants

Oligonucleotide	Nucleotide sequence (5′→3′)a
deRight-vec-F	GGTAACGCCAGGGTTTTCCCAGTCACGACGTTGTAAAACGACGGCTCTAGATGACGATGCGTGGAGACC
deRight-R	CGAGCGCACCGCCAGTGGCTACCGCGTCTACACCCCGCTGCACTCTAGATGCGGATGTTGCGATTAC
deLeft-F	CCATGTTCACCTCGATGACGTACGGGAACGCCATGTCGATGCCACTAGTTGACGATGCGTGGAGACC
deLeft-vec-R	GCCAGTTCATCCATCGCTTTCTTGTCTGCTGCCATTTGCTTTGTACTAGTTGCGGATGTTGCGATTAC
delstA-F	ATGAGTACCACCGAGCGCCGCAGCCGAATAGAGGCCCTCGGCGCgtttaaacTGACGATGCGTGGAGACC
delstA-R	TCAGGCCCCTGCGTGGGTGACGGTTGCGGACAGCGCGCCGGTGAgtttaaacTGCGGATGTTGCGATTAC
delstB-F	ATGGGCATCGTCATCACCGCATCCGCGACCGCCACCCACACCGAgtttaaacTGACGATGCGTGGAGACC
delstB-R	TCACCATCCCTGCGGGCGGTAGGAGGCGACGGCTGCTCGCGGTCgtttaaacTGCGGATGTTGCGATTAC
delstC-F	GTGGCGACCACGACCGCCACCCCGGCGGCGGCCCGGCCCGCAGCgtttaaacTGACGATGCGTGGAGACC
delstC-R	TCACAGCTGCCGCCACCAGGTCGCCCCGTAGAGCATGGTCAGGCgtttaaacTGCGGATGTTGCGATTAC
delstD-F	GTGAAGATCCTGATCACCGGAGCCACCGGCTTCCTCGGCGGCCAgtttaaacTGACGATGCGTGGAGACC
delstD-R	TCATCGTGAGGTGTCCTCTCGGCCGGCAGGTCCGGCCGGATCGGgtttaaacTGCGGATGTTGCGATTAC
delstE-F	ATGAGCACCAGCACACCGAACCCGCCCGGCACCTCGGAACCACAgtttaaacTGACGATGCGTGGAGACC
delstE-R	TCATGCGGTGGTTCCTTCGCGTACGGCGGCGAGGCGGTCGCAGAgtttaaacTGCGGATGTTGCGATTAC
delstF-F1	CCGACGGGATCGTCTGGTCGGTCAACAACCGGCAGCTTTTCCGTGTGACGATGCGTGGAGACC
delstF-R1	CCGATCGGGAAGCGGTGCTCGGCGAGGACGGGGCCGGTGTCGATGTGCGGATGTTGCGATTAC
deorf1-F	GTGCGCCGACTCGTCTACTACATCGCCACCACGCTCGACGGCTTgtttaaacTGACGATGCGTGGAGACC
deorf1-R	TCAGGCGGTGTCGGTGTCGGTGTCGGTGTCGGTGGGGCGGGTGTgtttaaacTGCGGATGTTGCGATTAC
deorf2-F	GTGGGCAGGGACATGGCACGACCCCGCGGCGTCGAGGACGCGGTgtttaaacTGACGATGCGTGGAGACC
deorf2-R	TCATCGTTCCTCCGGGCGTGGGTGGTGGGGTTGGTGGGCGCGGAgtttaaacTGCGGATGTTGCGATTAC
deorf3-F	ATGACGTACGCCACCCCCGCCCGGCCCCTCGCCGGCAAGGTCGCgtttaaacTGACGATGCGTGGAGACC
deorf3-R	CTACAGCGGCTCGCCGAGGGGGCCGTGGACCGGTTCGAGGGCATgtttaaacTGCGGATGTTGCGATTAC
deorf4-F	GTCTTCGGCCACAGCATGGGCTCGCTCGTCGCGTACGAGACCGTCTGACGATGCGTGGAGACC
deorf4-R	AGGGCATCAGGAGCTCGCGGAGCTCGGGGATGTCGTACACCTCGGTGCGGATGTTGCGATTAC
deorf6-F	GTGAGGATCCTCCTGGTCGGAGCGGGCGGCACGCTCGGCGGCGCgtttaaacTGACGATGCGTGGAGACC
deorf6-R	TCAGTGGACGCGGTAGATCCGGCCCGTCTGGGCGCCCTCGATCGgtttaaacTGCGGATGTTGCGATTAC

^aUnderlined sequences are homologous to the ends of the aadA gene cassette. The PmeI restriction site (gtttaaac, shown in lowercase) in some oligonucleotides was used for removal of the inserted aadA gene cassette by restriction digestion after λ Red-mediated recombination.

Thirteen integrative plasmids carrying different gene deletions were generated from integrative cosmid 27F11 by employing these methods. These plasmids included (i) pDelL, in which a 15-kb left-terminal fragment of the genomic DNA insert of 27F11 was deleted; (ii) pDelR, in which a 9-kb right-terminal fragment of the genomic DNA insert of 27F11 was deleted; (iii) pDlstF and pDorf4, in which the open reading frames of lstF and orf4 were replaced by aadA, respectively; and (iv) pDlstA, pDlstB, pDlstC, pDlstD, pDlstE, pDorf1, pDorf2, pDorf3, and pDorf6, in which internal fragments of lstA, lstB, lstC, lstD, lstE, orf1, orf2, orf3, and orf6 were deleted from 27F11, respectively, affording in-frame deletion mutants of these genes.

Construction of S. toxytricini gene deletion mutants.

27F11-derived integrative plasmids carrying deletions in the lst operon and flanking sequences were transferred by E. coli-Streptomyces conjugation into large-deletion mutant strain SBT11 separately. Integration of these plasmids into the chromosome of S. toxytricini SBT11 resulted in the generation of gene disruption strains with deletions in the ectopic copy of the lst genes and adjacent sequences. For instance, integration of pDlstF and pDorf4 into the chromosome of S. toxytricini SBT11 resulted in the generation of the ΔlstF::aadA and Δorf4::aadA mutant strains, respectively. Similarly, S. toxytricini Δleft, Δright, ΔlstA, ΔlstB, ΔlstC, ΔlstD, ΔlstE, Δorf1, Δorf2, Δorf3, and Δorf6 mutant strains were generated by transferring pDelL, pDelR, pDlstA, pDlstB, pDlstC, pDlstD, pDlstE, pDorf1, pDorf2, pDorf3, and pDorf6, respectively, into SBT11. These mutants were verified by PCR experiments (Table 1; see also Fig. S2 to S4 in the supplemental material).

Analysis of lipstatin and new metabolites in fermentation broths.

Fermentation broths of wild-type or mutated S. toxytricini strains were treated as described previously (35). Briefly, 50 ml of fermentation broth of each strain was extracted with 150 ml of acetone-hexane (3:2, vol/vol) four times. The organic phase was combined, dried with sodium sulfate, and concentrated by evaporation to yield a yellow oil with a volume of ∼1 ml. One hundred microliters of the crude extract was dissolved in 1 ml methanol and centrifuged at 12,000 rpm for 5 min. The supernatant was filtered and subjected to high-performance liquid chromatography (HPLC) or HPLC-mass spectrometry (MS) analysis. HPLC analyses were performed on an Agilent 1260 Infinity HPLC system or an Agilent 1100 series HPLC system (with a diode array detector set at 210 nm). HPLC-MS analysis was performed on the same Agilent 1100 HPLC series instrument coupled with a mass-selective detector (MSD) ion trap mass spectrometry system with an electrospray ionization (ESI) source. HPLC and HPLC-MS were conducted with a Cosmosil Cholester reversed-phase HPLC column (4.6 by 150 mm) under the following conditions. The column was equilibrated with 20% solvent A (H₂O) and 80% solvent B (methanol). Water (solvent A)-methanol (solvent B) mobile phases were used (20% solvent A–80% solvent B from 0 to 10 min, from 20% solvent A–80% solvent B to 10% solvent A–90% solvent B from 10 to 11 min, 10% solvent A–90% solvent B from 11 to 20 min, from 10% solvent A–90% solvent B to 100% solvent B from 20 to 21 min, 100% solvent B from 21 to 25 min, from 100% solvent B to 20% solvent A–80% solvent B from 25 to 26 min, and 20% solvent A–80% solvent B from 26 to 35 min) at a flow rate of 200 μl/min. Under these conditions, lipstatin was eluted at 16 min when the Agilent 1100 HPLC system was used or at 13.5 min when the Agilent 1260 Infinity HPLC system was used.

High-resolution mass spectrometry and nuclear magnetic resonance analyses.

High-resolution mass spectrometry (HR-MS) was performed by using a 6530 Accurate-Mass quadrupole time of flight (QTOF) spectrometer coupled to an Agilent 1200 series HPLC system. Nuclear magnetic resonance (NMR) spectra were recorded on an Agilent 500-MHz spectrometer. Proton chemical shifts are reported in ppm (δ) relative to internal tetramethylsilane (TMS) (δ, 0.0 ppm) or with the solvent reference relative to chloroform (CHCl₃) (δ, 7.26 ppm). Data are reported as follows: chemical shift (multiplicity [singlet {s}, doublet {d}, triplet {t}, multiplet {m}, and broad singlet {brs}], coupling constants, and integration). Carbon chemical shifts are reported in ppm (δ) relative to TMS with CDCl₃ as the internal standard.

Isolation and structure elucidation of compound 4 from the S. toxytricini ΔlstF::aadA mutant.

The S. toxytricini ΔlstF::aadA strain was fermented in 4 liters of fermentation medium at 220 rpm at 30°C for 5 days. Mycelium harvested by centrifugation was soaked overnight in 400 ml acetone and extracted three times with 600 ml hexane. The organic phase was combined, dried with sodium sulfate, and concentrated to yield a yellow oil. The yellow crude extract was extracted with 2 volumes of acetonitrile and centrifuged at 12,000 rpm for 5 min. The resulting yellow layer was further purified twice by semipreparative HPLC on an Agilent 1260 Infinity HPLC system under the following elution conditions, where solvent A is H₂O plus 0.1% formic acid and solvent B is acetonitrile plus 0.1% formic acid: 20% solvent A–80% solvent B from 0 to 10 min, from 20% solvent A–80% solvent B to 10% solvent A–90% solvent B from 10 to 11 min, 10% solvent A–90% solvent B from 11 to 20 min, from 10% solvent A–90% solvent B to 100% solvent B from 20 to 21 min, and 100% solvent B from 21 to 30 min. Compound 4 was eluted at 16.9 min under these conditions. Finally, 22.7 mg of pure compound was obtained. The chemical structure was elucidated based on high-resolution ESI-QTOF-MS, ¹H NMR, ¹³C NMR, ¹H-¹H two-dimensional correlated spectroscopy (COSY), distortionless enhancement by polarization transfer at 135° (DEPT135), heteronuclear single quantum coherence (HSQC), and heteronuclear multiple-bond correlation (HMBC) spectra (see Table S1 and Fig. S5 and S6 in the supplemental material). For compound 4, HR-ESI-QTOF-MS m/z = 351.2891; ¹H NMR (500 MHz, CDCl₃) δ 5.54 (m, 1H), 5.42 (m, 2H), 5.31 (m, 1H), 4.76 (m, 1H), 4.29 (brs, 1H), 2.8 (t, J = 7 Hz, 2H), 2.52 (m, 1H), 2.44 (m, 1H), 2.3 (m, 1H), 2.12 (m, 1H), 2.09 (m, 1H), 2.04 (m, 2H), 1.75 (m, 1H), 1.57 (m, 1H), 1.43 (m, 1H), 1.36 (m, 1H), 1.3 (m, 12H), 0.89 (m, 6H); ¹³C NMR (500 MHz, CDCl₃) δ 173.0, 131.8, 130.8, 127.1, 123.1, 75.2, 64.8, 46.3, 35.6, 33.4, 31.7, 31.5, 29.3, 29.2, 27.3, 27.0, 26.5, 25.8, 22.6, 22.6, 14.0, 14.0 (see Table S1 in the supplemental material for more detailed NMR data).

Isolation and structure elucidation of compound 5 from the S. toxytricini ΔlstD strain.

Compound 5 was isolated from the S. toxytricini ΔlstD strain by employing the same method as that used for compound 4. Compound 5 was eluted at 25.3 min under the same semipreparative HPLC conditions. Finally, 38.2 mg of compound 5 was purified from 2 liters of fermentation broth. The chemical structure was elucidated based on high-resolution ESI-QTOF-MS, ¹H NMR, ¹H-¹H COSY, ¹³C NMR, and DEPT135 spectra (see Tables S2 and S3 and Fig. S7 and S8 in the supplemental material). For compound 5, HR-ESI-QTOF MS m/z = 486.3555; ¹H NMR (500 MHz, CDCl₃) δ 8.20 (s, 1H), 5.99 (brs, 1H), 5.52 (td, J = 11 Hz, 7 Hz, 1H), 5.41 (td, J = 11 Hz, 7 Hz, 1H), 5.35 (m, 3H), 4.66 (td, J = 9 Hz, 5 Hz, 1H), 2.78 (m, 2H), 2.75 (dd, J = 12 Hz, 8 Hz, 1H), 2.63 (dd, J = 12 Hz, 5 Hz, 1H), 2.40 (m, 4H), 2.04 (td, J = 7 Hz, 7 Hz, 2H), 1.58 (m, 5H), 1.32 (m, 14H), 0.94 (m, 6H), 0.89 (m, 6H); ¹³C NMR (500 MHz, CDCl₃) δ 207.4, 171.9, 160.5, 132.1, 130.8, 126.9, 123.1, 71.1, 49.4, 45.7, 43.4, 41.8, 31.6, 31.6, 31.5, 29.1, 29.05, 28.3, 27.3, 25.7, 24.8, 23.6, 22.8, 22.6, 22.6, 21.9, 14.1, 14.0 (see Tables S2 and S3 in the supplemental material for more detailed NMR data).

Nucleotide sequence accession number.

The DNA sequence of the 19.4-kb lipstatin biosynthetic operon and downstream region was deposited in the NCBI GenBank database (http://www.ncbi.nlm.nih.gov/GenBank/) under accession number KJ872771.

RESULTS

Gene organization of the putative lipstatin biosynthetic operon and its downstream region.

The genome of S. toxytricini NRRL15443 was sequenced for searching for genes likely to be associated with lipstatin biosynthesis. In the assembled draft genome sequence of S. toxytricini, a nonribosomal peptide synthetase (NRPS) homologue attracted our attention because it was situated next to a formyltransferase homologue. The NRPS homologue showed similarity to many NRPS genes, such as the mycosubtilin synthase subunit C gene (mycC) (36). The derived NRPS protein contained one unknown domain (amino acids [aa] 202 to 602; E value, 3.8 × 10⁻⁴), one adenylation domain (A domain) (aa 799 to 1324; E value, 2.4 × 10⁻⁵⁵), and one peptidyl carrier protein (PCP) domain (aa 1499 to 1543; E value, 3.4 × 10⁻¹⁴) according to prediction by PKS/NRPS analysis (37). The predicted Stachelhaus code of the A domain of the NRPS was DIFALGGVAK, which shared 70% identity to the signature of leucine-specified A domains (38, 39). The N-terminal unknown domain (aa 202 to 602) showed marginal similarity (E value, 1.7 ×10⁻²) to the condensation domain of the starting module (CDAPS1_C1_start) of the CDAS1 NRPS governing biosynthesis of the lipopeptide calcium-dependent antibiotic (CDA), with 39% identity over a 59-aa aligned segment (aa 296 to 351 of the query NRPS) when analyzed by using the Natural Product Domain seeker (40). However, it is worth noting that an HXXXDG catalytic motif, which was reported previously to be essential for the condensation reaction (41, 42), present in this short aligned region. We therefore named this unknown domain the condensation-like domain (C^# domain), on the basis of its low-level similarity, short aligned region, and predicted catalytic motif missing one of the two histidines of canonical condensation domains (HHXXXDG) (41, 42).

Furthermore, the NRPS gene and formyltransferase homologue were located at a six-gene operon, here named lst for putative lipstatin biosynthesis. The relative orientation of the six lst genes (lstA to lstF) and a few downstream genes (orf1 to orf8) are schematically shown in Fig. 2a (top). lstE is the NRPS homologue. lstF is the formyltransferase homologue. Other genes of the lst operon encode homologues of two β-ketoacyl–acyl carrier protein (ACP) synthetases III (FabH) (LstA and LstB), one acyl-CoA synthetase (LstC), and a 3β-hydroxysteroid dehydrogenase (LstD). The proposed functions, predicted active sites or conserved domains, and BLASTp search results for the six putative Lst proteins (LstA to LstF) and eight downstream gene products (Orf1 to Orf88) are summarized in Table 3.

TABLE 3.

Summary of bioinformatics analysis of gene products of the putative lipstatin biosynthetic operon and downstream sequence

Protein	No. of aa	Proposed function	Predicted active site(s) and/or domain(s) (reference[s])a	Top BLAST hit from the Swiss-Prot database
				Accession no.	Name	No. of aa	Identity/similarity/coverage (%)
LstA	374	β-Ketoacyl-ACP synthase III	Cys128, His271, Asn305 (43–45)	C5CSE2	FabH	325	30/44/83
LstB	288	β-Ketoacyl-ACP synthase III	Cys96 (43–45)	Q1IQ32	FabH	333	32/48/46
LstC	874	Acyl-CoA synthetase	FAC_like_1 (46)	O31826	YngI	549	24/38/57
LstD	563	Dehydrogenase/isomerase	Asn89, Ser113, Tyr142, Lys146, Rossmann fold (47–49)	P9WQP6	Rv1106c	370	35/48/60
LstE	1,581	Nonribosomal peptide synthetase	C#, A, PCP (37–42; this study)	Q9R9I9	MycC	2,609	34/50/35
LstF	230	Formyltransferase	Asn88, His90, Asp128 (50)	Q8KCG8	FMT	314	36/50/62
Orf1	206	Unknown	NA	P45862	YwjB	169	31/41/79
Orf2	202	TetR family regulator	NA	B4SHW1	BetI	196	29/41/53
Orf3	306	Putative dehydrogenase	NA	Q99L04	DhrS1	313	36/53/89
Orf4	268	Putative thioesterase	NA	P33586	PabT	361	35/48/96
Orf5	297	Transcriptional regulator	NA	P77559	YnfL	297	31/47/92
Orf6	199	Putative dehydrogenase	NA	P50161	AflM	262	32/47/75
Orf7	156	Unknown	NA	Q44115	CpcE	273	32/50/55
Orf8	396	Putative esterase	NA	P05789	EreB	419	23/39/89

^aNA, not applicable.

Localization of the DNA region essential for lipstatin biosynthesis by large deletion, complementation, and heterologous expression.

Genetic experiments, including large deletion and complementation, were carried out to localize the DNA region that was essential for lipstatin biosynthesis. First, a 429-bp fragment internal to lstF, the formyltransferase gene, was used to screen the genomic library by Southern hybridization, resulting in the isolation of 16 cosmids carrying the speculated lst operon. One of them, cosmid 27F11, containing a 46.5-kb insert of genomic DNA including the putative lst operon, was then used to construct a gene replacement vector (pHL6) to generate a large deletion mutant of the S. toxytricini strain via homologous recombination. In the resulting double-crossover strain, SBT11, a 38.4-kb DNA fragment covering the putative lst operon was deleted from the S. toxytricini genome (Fig. 2a; see also Fig. S1 in the supplemental material). To genetically complement the large deletion, the int-attP^ϕC31 integrative cosmid 27F11 was transferred into SBT11. The empty vector pJTU2554, as a negative control, was also transferred. S. toxytricini strains SBT11, SBT11/27F11, and SBT11/pJTU2554 and the wild type were fermented under standard conditions, and productions of lipstatin in these strains were analyzed by HPLC. No lipstatin was detected in the organic extract obtained from deletion mutant strain SBT11, and lipstatin production was restored in complementation strain SBT11/27F11 (Fig. 2b, peaks marked by 1). Furthermore, when 27F11 was transferred into the heterologous host Streptomyces lividans, the resulting exconjugant produced lipstatin although at a yield ∼30% of that of the original producer S. toxytricini (Fig. 2c, peaks marked by 1).

To further narrow down the DNA region needed for lipstatin biosynthesis, a 15-kb left-end fragment and a 9-kb right-end fragment were removed from 27F11 by PCR targeting technology to yield two int-attP^ϕC31 integrative plasmids, pDelL and pDelR, respectively (Fig. 2a). pDelL and pDelR were introduced into SBT11, and the resulting strains were fermented and analyzed by HPLC. Lipstatin was found to be produced in both SBT11/pDelL (Δleft) and SBT11/pDelR (Δright) (Fig. 2b). This indicated that the 19.4-kb DNA region shared by pDelL and pDelR, encoding the 6 lst genes and 8 downstream genes, carried the whole set of genes for lipstatin biosynthesis.

Probing the function of the putative lst genes by single-gene knockouts.

All 6 genes of the putative lst operon (lstA to lstF) and 5 of 8 genes immediately downstream of the lst operon (orf1, -2, -3, -4, and -6) were disrupted in cosmid 27F11 separately by using PCR targeting technology (Table 2). Among these genes, lstF and orf4 were disrupted by replacement with a marker gene, aadA. As gene disruption of lstA, lstB, lstC, lstD, lstE, orf1, orf2, orf3, and orf6 by marker insertion or marker replacement would likely cause a polar effect on their downstream genes that might be transcribed from shared promoters, disruptions of these genes were made by in-frame deletion to exclude the probable polar effect. The resulting modified plasmid constructs that carried single-gene disruption mutations were introduced into large-deletion mutant strain SBT11 separately, to afford single-gene-disrupted S. toxytricini strains (see Fig. S2 to S4 in the supplemental material). These strains were fermented and analyzed by HPLC under standard conditions. HPLC analyses showed that lipstatin production was abolished in the ΔlstA, ΔlstB, and ΔlstC single-gene mutants (Fig. 3a), suggesting that lstA, lstB, and lstC play an essential role in the lipstatin biosynthetic pathway. Meanwhile, the Δorf1, Δorf2, Δorf3, Δorf4::aadA, and Δorf6 single-gene mutants kept producing lipstatin (Fig. 3a, peaks marked by 1), suggesting that these genes are not involved in biosynthesis.

FIG 3.

Detection of lipstatin and new metabolites in S. toxytricini strains carrying single-gene disruptions by HPLC. (a) Detection of lipstatin production in ΔlstA, ΔlstB, ΔlstC, Δorf1, Δorf2, Δorf3, Δorf4::aadA, and Δorf6 single-gene mutants. An Agilent 1260 Infinity HPLC system (210 nm) was used. Lipstatin peaks are marked by 1. (b) Detection of lipstatin and a new metabolite in ΔlstE and ΔlstF::aadA single-gene mutants. An Agilent 1100 HPLC system (210 nm) was used. Lipstatin is marked by 1 (production) or a dashed line (abolishment). New peaks (asterisks) at 11 min were further analyzed by ESI-MS (insets) (see Fig. S10 in the supplemental material). +MS, positive-ion-mode MS. (c) Detection of lipstatin and a new metabolite in the ΔlstD mutant. An Agilent 1100 HPLC system (210 nm) was used. Lipstatin is marked by 1 (production) or a dashed line (abolishment). The new peak (filled circle) at 21 min was further analyzed by LC-ESI-MS (inset) (see Fig. S11b in the supplemental material).

In the organic extracts obtained from the ΔlstE and ΔlstF::aadA mutants, lipstatin production was not detected in the HPLC-UV and MS chromatograms (Fig. 3b; see also Fig. S9 in the supplemental material), suggesting that both lstE and lstF are involved in lipstatin production. Furthermore, a new peak (Fig. 3b, asterisks) emerged at 11 min in HPLC traces of both the ΔlstE and ΔlstF::aadA mutants. The new peaks of two mutants produced a quasimolecular ion peak at m/z 351.3 ([M + H]⁺) and an ammonium adduct peak at m/z 373.2 ([M + NH₄]⁺) by liquid chromatography (LC)-ESI-MS analyses (Fig. 3b, insets; see also Fig. S10 in the supplemental material), suggesting that these two mutants accumulate compounds with an identical m/z of 350.3.

In the organic extract obtained from the ΔlstD mutant, lipstatin production was not detected, indicating that lstD is essential for lipstatin production. In addition, a new peak was observed by HPLC analysis (at 21 min) (Fig. 3c, filled circle; see also Fig. S11a in the supplemental material). This new peak of the ΔlstD organic extract at 21 min produced a quasimolecular ion peak at m/z 464.4 ([M + H]⁺) and adduct peaks at m/z 486.3 ([M + Na]⁺) and m/z 481.3 ([M + NH₄]⁺) by LC-ESI-MS analyses (Fig. 3c, inset; see also Fig. S11b in the supplemental material), suggesting that the ΔlstD strain accumulates a new metabolite with an m/z of 463.4, whose m/z value was 27.99 lower than that of lipstatin.

Compound 4 accumulated by the S. toxytricini ΔlstF::aadA strain.

The ΔlstF::aadA strain was fermented (4 liters), and 22.7 mg of the compound was purified from the fermentation broth. The structure of this compound was assigned as (3S,4S,6S)-3-hexyl-4-hydroxy-6-((2Z,5Z)-undeca-2,5-dien-1-yl)tetrahydro-2H-pyran-2-one, on the basis of analysis of the ¹H NMR, ¹³C NMR, ¹H-¹H COSY, and HMBC NMR spectroscopic data (compound 4) (Fig. 4; see also Table S1 in the supplemental material). The configurations at C₂ and C₃ were determined to be 2S and 3S based on analysis of the NOESY NMR spectrum, assuming that the configuration at C₅ was 5S (the same as that of lipstatin) (see Fig. S6g in the supplemental material). The HR-ESI-QTOF-MS analysis of compound 4 afforded a quasimolecular ion peak at m/z 351.2891, consistent with the [M + H]⁺ ion of the molecular formula C₂₂H₃₈O₃ + H⁺ (calculated m/z 351.2899). Compound 4 was isolated previously from alkaline hydrolysis of esterastin (compound 3), a lipstatin congener (12). Most of the chemical shifts of 22 carbons and protons matched well with those from the literature, except that the chemical shifts of C₄ and C₆ were exchanged, as supported by two-dimensional NMR spectra (see Table S1 and Fig. S6 in the supplemental material).

FIG 4.

Plausible lipstatin biosynthetic pathway in S. toxytricini. Octanoic acid may be activated by LstC (an acyl-CoA synthetase homologue), the acyl-CoA carboxylase (ACCase) complex (33), and ACP borrowed from primary metabolism. Two FabH homologues, LstA and LstB, may conduct Claisen condensation between 3-hydroxytetradeca-5,8-dienoyl-CoA and hexylmalonyl-ACP, affording the 22-carbon α-branched fatty acid backbone (compound 6) (17–21). LstE (NRPS) and LstF (a formyltransferase homologue) are responsible for the attachments of leucine and formyl groups. LstD (a 3β-hydroxysteroid dehydrogenase/isomerase homologue) is involved in the reduction of the 3-keto group of compound 7. The β-lactone ring may be formed spontaneously by attack of the 3-hydroxy group on the carbonyl of the ACP-tether acyl intermediate, in analogy with the ebelactone biosynthesis pathway (53). SCoA, CoA thioester; C^#, condensation-like domain; A, adenylation domain; T, thiolation domain or PCP domain. Compound 4 is accumulated by both the ΔlstE and ΔlstF::aadA mutants. Compound 5 is accumulated by the ΔlstD mutant (see the text for more explanation).

Compound 5 accumulated by the S. toxytricini ΔlstD strain.

The new metabolite accumulated by the ΔlstD mutant was purified to yield a yellow oil (38.2 mg) readily dissolved in chloroform and acetonitrile. The chemical structure of this metabolite was elucidated as (S)-(S,12Z,15Z)-8-oxohenicosa-12,15-dien-10-yl 2-formamido-4-methylpentanoate (compound 5) (Fig. 4) on the basis of analysis of the ¹H NMR, ¹³C NMR, and ¹H-¹H COSY NMR spectroscopic data (see Tables S2 and S3 and Fig. S8 in the supplemental material). High-resolution ESI-QTOF-MS analysis of the purified compound produced a sodium adduct ion at m/z 486.3555 (see Fig. S7 in the supplemental material), consistent with the [M + Na]⁺ ion of the molecular formula C₂₈H₄₉NO₄ (calculated m/z 486.3559).

DISCUSSION

In this study, we have identified a six-gene operon that was essential for lipstatin biosynthesis by genetic manipulations. Deletions of lstA, lstB, lstC, lstD, lstE, and lstF all abolished lipstatin production. Meanwhile, both the ΔlstE and ΔlstF::aadA mutants produced a new metabolite with the same HPLC retention times, equal molecular masses, and probably identical chemical structures. The metabolite produced by the ΔlstF::aadA mutant was purified and elucidated as (3S,4S,6S)-3-hexyl-4-hydroxy-6-((2Z,5Z)-undeca-2,5-dien-1-yl)tetrahydro-2H-pyran-2-one (compound 4). In addition, the ΔlstD mutant produced another new metabolite, which was purified and elucidated as (S)-(S,12Z,15Z)-8-oxohenicosa-12,15-dien-10-yl 2-formamido-4-methylpentanoate (compound 5).

We therefore propose that the LstA, LstB, and LstC enzymes are involved in early steps of the lipstatin biosynthetic pathway, while the LstE, LstF, and LstD enzymes are responsible for the following steps. On the basis of the genetic data, bioinformatics analyses, biosynthetic logics of nonribosomal peptides and fatty acids, and previously reported lipstatin biosynthetic models, we update the lipstatin biosynthetic pathway as follows (Fig. 4).

Formation of the 22-carbon α-branched fatty acid backbone.

According to lipstatin biosynthesis models established on the basis of early feeding experiments with S. toxytricini, the first committed step is Claisen condensation between (3S,5Z,8Z)-3-hydroxytetradeca-5,8-dienoyl-CoA and activated octanoic acid to afford the 22-carbon α-branched fatty acid backbone (17–21). Both substrates of Claisen condensation are obtained from incomplete degradation of linoleic acid. One substrate, 3-hydroxytetradeca-5,8-dienoyl-CoA, is directly obtained from β-oxidation. However, in analogy to the initiation of fatty acid synthesis, activation and carboxylation of octanoic acid are needed to form hexylmalonyl-CoA to serve as the second substrate (17–21). LstA, LstB, LstC, and a few enzymes from primary metabolism are plausibly involved in these substrate activation and condensation steps. First, the acyl-CoA synthetase homologue LstC might catalyze the ATP-dependent activation of octanoic acid to produce an acyl-CoA thioester via an acyl-adenylate intermediate (46). The octanoyl-CoA ester is then carboxylated to give hexylmalonyl-CoA, probably by an acyl-CoA carboxylase (ACCase) complex, the inactivation of which caused a decrease of lipstatin production by ∼80% in S. toxytricini (30). An unknown route may be present to complement the ACCase pathway for the generation of hexylmalonyl-CoA, however. The hexylmalonyl-CoA thioester is then transferred to ACP borrowed from primary metabolism. The resulting hexylmalonyl-ACP serves as one of two substrates for Claisen condensation. The C₁₄ substrate of the condensation, as mentioned above, is directly obtained from the incomplete degradation of linoleic acid. Next, LstA, probably together with LstB, both of which are FabH homologues, catalyzes the decarboxylation of hexylmalonyl-ACP and successive condensation with 3-hydroxyl-5,8-tetradecadienoyl-CoA to afford a 3-keto-5-hydroxy-C₂₂-ACP intermediate (compound 6) (Fig. 4). Double-FabH homologues may be employed to incorporate unusual starter units into final products in some biosynthetic pathways, for instance, AsuC3/C4 in the asukamycin production pathway (51). Substrate selectivity and stereochemistry of LstAB and LstC enzymology will be subjects of further research. Nevertheless, the formation of α-branched β-hydroxylated fatty acid derivative in S. toxytricini is different from the synthesis of mycolic acids in corynebacteria and mycobacteria, which is carried out by a multifunctional type I polyketide synthase, Pks13 (23).

Incorporation of formyl and leucine groups into the 3-keto-5-hydroxy-C₂₂-ACP intermediate.

The subsequent steps are the incorporation of leucine and formyl groups into the 3-keto-5-hydroxy-C₂₂-ACP intermediate (21). Accumulation of compound 4 in the ΔlstE and ΔlstF::aadA mutants supports that the NRPS LstE and the formyltransferase LstF are responsible for the attachment of leucine and formyl groups and that the attachment occurs after Claisen condensation. First, LstE might activate leucine with ATP to form a leucinyl-AMP intermediate by the A domain and then load the activated leucinyl group onto the PCP domain at the serine-attached 4′-phospho-pantethine side chain as a thioester. LstF then transfers a formyl group to the α-amine group of the PCP-tethered leucine residue to form a PCP-tethered formyl-leucine. Finally, LstE might catalyze the nucleophilic attack of the 5-hydroxyl of intermediate compound 6 on the acyl carbon of the PCP-tethered formyl-leucine by its C^# domain to yield intermediate compound 7, probably via a mechanism analogous to that of acyltransferase reactions (Fig. 4, curved arrow) (41). Thus, because of the absence of either functional LstE or LstF proteins in the ΔlstE and ΔlstF::aadA mutant strains, the free 5-hydroxy of the 3-keto-5-hydroxy-C₂₂-ACP intermediate (compound 6) attacks on the C-1 carbonyl to give a δ-lactone derivative spontaneously (52). The 3-keto of the resulting δ-lactone derivative might be reduced by LstD or an unknown enzyme to give compound 4 (see below).

Reduction of the 3-keto group and formation of β-lactone.

The next steps are reduction of the 3-keto group and formation of the β-lactone to afford the final product lipstatin. The production of metabolite 5 in the ΔlstD mutant supports that the reduction of the 3-keto group and the formation of β-lactone are the final steps leading to lipstatin. In addition, the displacement of the H-2 proton with the solvent proton, which was shown by deuterium feeding experiments, may also occur before the formation of β-lactone (20). The LstD protein sequence is similar to those of many members of the 3β-hydroxysteroid dehydrogenase extended-short-chain dehydrogenase/reductase (3β-HSD-like eSDR) superfamily, such as Rv1106c (53). It is worth noting that the characterized 3β-HSD from animals is a bifunctional dehydrogenase/isomerase enzyme (54). Therefore, it is plausible that LstD is involved in the reduction of the 3-keto group to a 3-hydroxyl group, which is probably also involved in the exchange of the H-2 proton with the solvent proton by an unknown mechanism. After the 3-keto is reduced, the resulting 3-hydroxy group may attack spontaneously on the carbonyl moiety of the ACP-tether acyl intermediate to give the β-lactone ring, in analogy with the formation of the β-lactone ring in the ebelactone biosynthesis pathway (55). On the other hand, the production of compound 4 in the ΔlstE and ΔlstF::aadA mutants implies that the LstD dehydrogenase might also reduce the 3-keto group to a 3-hydroxyl in the absence of a 5-O-formyl-leucine group to afford compound 4, or another, unknown enzyme in S. toxytricini contributes to this reduction. The catalytic mechanism and substrate specificity of LstD will be the subjects of further research.