Biosynthesis of thiomarinol A and related metabolites of Pseudoalteromonas sp . SANK 73390 †

School of Chemistry, University of Bristol, E-mail: chris.willis@bristol.ac.uk; tom.simp 1295; Tel: +44 (0)117 928 7660 School of Biosiences, University of Birming UK Discovery Science and Technology Departme 16-13, Kita-Kasai, Edogawa-ku, Tokyo, 134† Electronic supplementary information details, feeding experiments, synthesis data and NMR spectra of labelled thioma ‡ These authors contributed equally to th


Introduction
Thiomarinol A 1 (Fig. 1), is the major metabolite produced by the marine bacterium Pseudoalteromonas sp.2][3] They display antibiotic activity against both Gram-negative and Gram-positive bacteria, and have particularly potent activity against methicillin-resistant Staphylococcus aureus (MRSA).They are unusual in being hybrid antibiotics, effectively two separate antibiotic classes in one molecule, and biosynthetically can be regarded as comprising three main elements: a highly functionalised polyketide-derived acid, which is esteried by a hydroxy fatty acid, which itself forms an amide with the bicyclic amino acidderived pyrrothine.The main carbon skeleton of the thiomarinols is very closely related to that of mupirocin, an important antibiotic used clinically against MRSA. 4 Mupirocin consists of a mixture of pseudomonic acids (PA A-C, 8 to 10), [5][6][7] with PA-A comprising ca.95% of the mixture.The major differences in thiomarinol A are the replacement of the 9-hydroxynonanoic acid moiety in the PAs with 8-hydroxyoctanoate in the thiomarinols, the presence of an extra 4-hydroxyl group, the lack of the 10,11-epoxide and the pyrrothine.Many pyrrothine natural products themselves display antibiotic activity, e.g.holomycin 11, N-propionylholothin 12, thiolutin 13, and aureothricin 14. [8][9][10] Stable isotope labelling studies were used to conrm the origins of all of the atoms in pseudomonic acid A and were consistent with a polyketide/fatty acid origin. 11,12his was subsequently conrmed by isolation of the biosynthetic gene cluster 13 which revealed that mupirocin was one of the rst members of what is now known to comprise an increasingly large number of biologically active compounds.These are most oen from diverse and unusual bacterial sources and are produced via trans-AT modular PKSs that have evolved independently from other modular systems. 14The main identifying feature of these systems is the lack of an integral acyl transferase (AT) domain in each chain extension module, this function being supplied "in trans" to each module by separately encoded mono-(or di-) functional ATs.In addition they show many other unusual features including extensive involvement of other trans acting enzymes during polyketide assembly.In particular b-branches are introduced onto b-keto-thiol ester intermediates by a cassette of enzymes that include an HMG CoA synthase (HMGS) analogue using chemistry similar to that involved in mevalonate formation in terpene biosynthesis. 15The biosynthesis of the pyrrothine class of natural products has not been studied in detail, although cysteine has been implicated as a precursor via radioisotope labelling studies. 16e have recently identied the thiomarinol (tml) biosynthetic gene cluster via full genome sequencing of SANK 73390.It is contained on a 97 kb plasmid consisting almost entirely of the thiomarinol biosynthetic genes. 17These consist of trans-AT PKSs (tmpA, C and D), a putative FAS (tmpB) and associated tailoring and resistance genes (tmlA-Z) with high homology to the mupirocin (mup) cluster (Fig. S1, ESI †).A non-ribosomal peptide synthetase (NRPS) linked to a set of tailoring enzymes (holA-H) is also present similar to that recently shown to control holomycin biosynthesis in Streptomyces clavuligerus. 18,19Analysis of the WT culture and mutant strains in which the PKS and NRPS parts of the cluster had been deactivated by in-frame deletions, resulted in isolation of a number of new thiomarinol related compounds. 20Analysis of extracts of the WT strain by reversed-phase HPLC-ESIMS led to the isolation of two additional polar metabolites which lacked the pyrrothine moiety.These were shown to be marinolic acid A 15 and the corresponding amide, marinolic amide 16.Similar analysis of the DNRPS mutant strain showed, as expected, that no pyrrothine containing compounds were produced, the major thiomarinol metabolite being marinolic acid A, but no amide 16.Two minor metabolites were isolated: marinolic acid A 6 17 and A 4 18 in which the octanoate moiety on marinolic acid A has been replaced by hexanoate and butanoate respectively.Re-examination of the WT extracts revealed trace amounts of 17.A number of new acyl-pyrrothine containing compounds were also isolated.On the basis of their similarity to metabolites previously isolated [21][22][23] from Xenorhabdus spp, they were designated as xenorhabdins 8-13 (19-25) respectively (Fig. 2).
We now report stable isotope labelling and other feeding experiments with WT and mutant strains of Pseudoalteromonas which conrm the origins of all of the atoms in thiomarinol, provide a rationalisation for the occurrence of the minor related metabolites, and for the timing of pyrrothine formation and its linkage to marinolic acid.

Results and discussion
The 13 C NMR spectrum of thiomarinol A was rigorously assigned.This necessitated reassignment of two side-chain carbons (C4 0 and C5 0 ) compared to that made previously. 1 This was achieved using a high resolution band-selective HMBC experiment with correlations observed via H3 0 (to C4 0 and C5 0 ) and H7 0 (to C6 0 and C5 0 ) allowing these two carbons, which are separated by 0.1 ppm, to be unambiguously assigned.In d 6 -DMSO, the 15-and 17-methyls are coincident, but they are resolved in d 4 -methanol.Prior to stable-isotope labelled precursor feeding, the production of thiomarinol A was monitored.This showed that thiomarinol production commences at around 7 hours aer the start of fermentation, reaches a maximum around 24 hours and levels then slowly decrease thereaer.The labelled precursors were therefore fed to the fermentation media ca.7 hours aer inoculation, and the fermentation extracted aer 24 hours.The crude extract obtained was puried to give thiomarinol A, which was then analysed by 13 C NMR to determine the extent and positions of 13 C and 18 O incorporations.The results of feeding labelled acetates and [methyl- 13 C]-methionine to wild-type Pseudoalteromonas SANK 73390 are summarised in Scheme 1.
The incorporation pattern conrms the similarity of the pseudomonic acid A 11,12 and thiomarinol A biosynthetic  pathways.A regular acetate incorporation pattern is observed in the C1-C14 fragment of thiomarinol.A high level of 13 C incorporation (5-6%) is observed at all seven expected positions in the polyketide-derived moiety with high levels of 18 O retention (ca.80%) at carbons 1, 5, 7 and 13 (Table 1).The 4-and 6hydroxyl groups are not labelled from 18 O-acetate and are at positions consistent with their introduction being via tailoring enzymes using molecular oxygen.The 16-methylene (55% enriched from [methyl- 13 C]methionine) and 17-methyl carbons (40% enriched) are methionine derived consistent with the presence of C-MeT domains in the rst and third extension modules, and the 15-methyl is derived from the methyl of a cleaved acetate unit consistent with it being derived via the HMGS b-branch pathway as in pseudomonic acid A.

Biosynthetic origin of the side-chain
Interestingly the labelling pattern in the octanoate side chain did not follow a standard polyketide labelling pattern.Whilst expected head-to-tail incorporation of intact acetate units from [1,2-13 C 2 ]acetate was observed for C1 0 to C4 0 (2% incorporation), a lower level of incorporation was observed in C5 0 and C6 0 (1%) and much lower into C7 0 and C8 0 (0.3%).[2-13 C]-Acetate also gave a similar pattern with much lower incorporation of 13 C into both C7 0 and C8 0 ($2% compared to 7% in other acetate-derived positions), and an intermediate level of enrichment (4%) at C6 0 (Table 1).[1-13 C, 18 O 2 ]-Acetate gave incorporation levels of 13 C ($6%) at carbons 1 0 and 3 0 as was observed elsewhere, but incorporation into C5 0 was half this level and at C7 0 was negligible.A low incorporation, ca.1%, into C8 0 was observed.These results suggest that a C 4 precursor is likely to be the starter unit for two rounds of FAS catalysed extension to give the octanoate side-chain.The observed scrambling of acetate incorporation at C7 0 and C8 0 suggested the possible involvement of a citric acid (Krebs) cycle intermediate.
It is known that 4-hydroxybutyrate can be formed via reduction of succinyl CoA. 24The overall results of acetate incorporation are also consistent with this: acetate is incorporated intact into only carbons 3 and 4 of succinyl CoA (Scheme S2, ESI †) which explains the signicant but lower incorporation into C5 0 /C6 0 of thiomarinol, presumably due to dilution by a pool of endogenously derived Krebs Cycle intermediate.Further cycling of labelled succinate would give single labels at carbons 1 and 2 of succinyl CoA which explains the lower levels of labelling from [2- 13 C]acetate into C7 0 and C8 0 .To explain the low level of intact acetate labelling of C7 0 /C8 0 , the conversion of succinyl CoA to symmetrical succinate must be partially reversible so that ca.30% of the succinate is incorporated via this route.The low level of labelling at C8 0 from [1-13 C]acetate is completely consistent with this hypothesis.
To further investigate the proposed C 4 side-chain precursor, [2,3-13 C 2 ]-4-hydroxybutyrate was prepared from [2,3-13 C 2 ]succinate via cyclisation to succinic anhydride, reduction to the lactone with LiAlH 4 followed by hydrolytic ring opening with aqueous NaOH (Scheme S1 †).Feeding [2,3-13 C 2 ]-4-hydroxybutyrate to cultures of the wild-type Pseudoalteromonas SANK 73390 showed a site-specic labelling (0.2%) of thiomarinol A at C6 0 , C7 0 conrming intact incorporation (Scheme 2, see insert expansion in Fig. S12 †).Thus the biosynthetic origin of C5 0 to C8 0 is somewhat similar to the situation presumed for 9-hydroxynonanoic in mupirocin biosynthesis, where labelling studies are consistent with a 3-hydroxypropionate starter unit, which then undergoes three reductive chain elongations.3-Hydroxypropionate can be formed in bacteria by reduction of the thiolester of malonyl CoA 25 and both the mup and tml clusters show the presence of genes encoding a malonyl/succinyl CoA dehydrogenase clustered with an acyl CoA synthetase and an ACP.
Interestingly, the biosynthetic gene cluster for difficidin, 26 another trans-AT metabolite of Bacillus amyloliquefaciens, also contains an oxo-acyl CoA reductase encoded by difE which as in mupirocin/thiomarinol is grouped with an acyl CoA synthetase and an ACP (difD and difC).In addition there is an adjacent kinase (difB) which would be suitable for catalysing dehydration of 3-hydroxypropionate to produce the acryl thiol ester required to prime difficidin chain elongation.
The retention of 18 O label at C1 0 of thiomarinol A is slightly less than half that at the other acetate-oxygen bearing positions.This is consistent with C1 0 existing as the free carboxylate, resulting in randomisation the label between both carboxylate oxygens, before ligation via an amide bond to the pyrrothine.Analysis of the structure of the pyrrothine moiety of thiomarinol suggest that it is formed via two molecules of cysteine.To gain evidence for this proposal, [2-13 C]cystine was prepared from ethyl [2-13 C]pyruvate, (Scheme S2 †) and fed to WT Pseudoalteromonas.A high level of incorporation of 13 C into the C2 00 and C4 00 positions (16 and 20% respectively) conrmed that the pyrrothine is indeed cysteine, and therefore NRPS, derived (Scheme 2).
The biosynthetic gene cluster for holomycin biosynthesis in Streptomyces clavuligerus has been isolated and a pathway from cysteine proposed.This pathway involves initial dipeptide formation followed by a series of oxidative cyclisations, the nal step being acetylation of the free pyrrothine. 18To prove that the pyrrothine moiety in thiomarinol is assembled and then incorporated as an intact unit rather than being assembled on marinolic acid, pyrrothine was synthesised as previously described 27 and fed to the DNRPS mutant strain of SANK 73390.The mutant was prepared by a deletion in the holA gene encoding the amino acid activation domain which, from sequence analysis of the active site, is predicted to selectively activate cysteine.As indicated above, the DNRPS mutant normally produces mainly marinolic acid.When cultures were supplemented with pyrrothine, the normal WT phenotype was restored indicating that Pseudoalteromonas takes the intact pyrrothine as substrate (Scheme 3).Interestingly, the concentration of the feeds was critical, with maximum restoration of thiomarinol A production being observed at a pyrrothine concentration of 20 mg ml À1 .At concentrations of >80 mg ml À1 , no thiomarinol A was observed and production of marinolic acid was also completely inhibited (Fig. S3, ESI †).

Novel acyl pyrrothines via mutasynthesis
Previous studies have shown that when TmlU, a putative amide ligase is knocked out, the mutant strain produces no thiomarinol but does produce marinolic acid and the full set of xenorhabdins. 20The HolE analogue, ORF3483, from the holomycin cluster in S. clavuligerus has been expressed and puri-ed and it has been shown 18 to act as an acyl transferase capable of transferring acetyl and longer acyl groups to pyrrothine.Thus it appears that the thiomarinol cluster has two ligases, TmlU which forms the amide in thiomarinol, and HolE which links pyrrothine to a wide range of acyl CoAs available in Pseudoalteromonas to produce the xenorhabdins.
To test if HolE can also transfer a range of non-endogenous substrates, various short, medium and long chain fatty acids and others with u-functionalized side chains, e.g.hydroxyl, amino, carboxyl and phenyl were fed to the DPKS mutant (Scheme 4).Some of these were incorporated with varying efficiencies, presumably aer conversion to their respective CoA esters, sometimes aer catabolic chain shortening to give 19 and the novel acyl-pyrrothines 26-30.

The biosynthesis of thiomarinol A and related metabolites
The isolation of the minor metabolites 15-18 from both WT and mutant strains and the incorporation of isotopically labelled substrates can be rationalised as shown in Scheme 5. Marinolic acid A 15 is produced by the DNRPS and DTmlU mutants, and is present in trace amounts in the WT which suggests that it is both a genuine biosynthetic product and intermediate in its own right and is not produced as an artefact, e.g. as a degradation product of thiomarinol A. The intermediacy of marinolic acid 15 was conrmed by feeding it to the DPKS mutant whereupon, thiomarinol production was restored (Fig. S19 †).The amide 16 on the other hand was only seen in the WT cultures which suggests that it is formed by degradation of thiomarinol A, and is not formed by, e.g.amidation of marinolic acid.This was further evidenced by feeding thiomarinol A to the DPKS mutant when the presence of both marinolic acid and marinolic amide in the crude extract could be observed by LCMS.This suggests that the amide is indeed a degradation product of thiomarinol.
The isolation of truncated metabolites marinolic acids A 4 and A 6 , is consistent with the isotope labelling studies which suggest that a C 4 intermediate is added to the product of the PKS, e.g. by trans-esterication.This would then be elongated to give an enzyme bound hexanoate analogue which would be further elongated to give the completed octanoate side chain.In the absence of the pyrrothine in the DNRPS mutant they would be released either by enzymatic or spontaneous hydrolysis as marinolic acids.We have previously shown that release of assembly intermediates occurs when ux of intermediates along the normal mupirocin pathway is impaired by a range of targeted mutations of trans-acting and other tailoring genes. 28imilar mechanisms would allow leakage of marinolic acids A 4 and A 6 into the medium.It is signicant that this appears to be more pronounced in the mutant than the WT.Truncated forms of mupirocin and related metabolites with C 7 and C 5 hydroxy acids have also been observed in both mutant strains and when the quorum sensing regulation mechanism has been manipulated to increase mupirocin production. 29

Conclusions
Using a combination of genetic and isotopic labelling studies, evidence has been presented for a mixed PKS-NRPS biosynthetic pathway to the antibiotic thiomarinol A in Pseudoalteromonas sp.SANK 73390.Particularly interesting features of the pathway include assembly of the 8-hydroxyoctanoic acid side-chain, which is generated via chain extension of a C 4 precursor (4-hydroxybutyrate), and both marinolic acids A 4 18 and A 6 17 have been isolated as minor metabolites from both wild-type and the DNRPS mutant.There was a good incorporation of [ 13 C]cystine into thiomarinol A and feeding studies with synthetic pyrrothine to the DNRPS mutant are in accord with heterocycle assembly (HolA-D, F-H) prior to intact incorporation into thiomarinol A (TmlU).A further minor metabolite, marinolic amide 16, present in wildtype but not the DNRPS mutant, is a degradation product of thiomarinol.Isotopic labelling studies indicate that the biosynthesis of the C1-C17 fragment is analogous to that of pseudomonic acid, a trans-AT modular PKS pathway with many fascinating and, as yet, unexplained features and these studies lay the foundation for a further understanding of the pathways to these compounds.

Scheme 3
Scheme 3 Feeding pyrrothine to the NRPS mutant gives thiomarinol A.