Heterologous Production of the Marine Myxobacterial Antibiotic Haliangicin and Its Unnatural Analogues Generated by Engineering of the Biochemical Pathway

Despite their fastidious nature, marine myxobacteria have considerable genetic potential to produce novel secondary metabolites. The marine myxobacterium Haliangium ochraceum SMP-2 produces the antifungal polyketide haliangicin (1), but its productivity is unsatisfactory. The biosynthetic gene cluster hli (47.8 kbp) associated with 1 was identified and heterologously expressed in Myxococcus xanthus to permit the production of 1 with high efficiency (tenfold greater amount and threefold faster in growth speed compared with the original producer), as well as the generation of bioactive unnatural analogues of 1 through gene manipulation. A unique acyl-CoA dehydrogenase was found to catalyse an unusual γ,δ-dehydrogenation of the diketide starter unit, leading to the formation of the terminal alkene moiety of 1. Biological evaluation of the analogues obtained through this study revealed that their bioactivities (anti-oomycete and cytotoxic activities) can be modified by manipulating the vinyl epoxide at the terminus opposite the β-methoxyacrylate pharmacophore.


Identification and heterologous expression of the haliangicin biosynthetic gene cluster (hli).
The polyene structure of haliangicin (1) suggests that its biosynthetic machinery is a pure PKS system that is accompanied by various enzymes involved in tailoring modifications. To clone the biosynthetic gene cluster responsible for formation of haliangicin, we first constructed a cosmid library of the H. ochraceum genome by using the E. coli/Streptomyces shuttle vector pDW103 18 . The resulting library, consisting of approximately 1700 cosmid clones, was then screened by colony hybridization using probes homologous to the ketosynthase (KS) genes of terrestrial myxobacteria ( Supplementary Fig. S1). The end sequences of several KS-positive cosmids indicated that they might be responsible for the biosynthesis of 1 ( Supplementary Fig. S2); these included HMG-CoA synthase for installing a β -branch methyl (9-Me) group and a 3-hydroxyacyl-CoA dehydrogenase for biosynthesizing methoxymalonyl-ACP, a precursor of the 1,2-dioxy unit at the C-3-C-4 position. Of the cosmids that were obtained, two partially overlapping cosmids, c7-6E and c10-11C, were sequenced, revealing a continuous 63-kbp segment that putatively harbours the entire haliangicin biosynthetic gene cluster (hli). Sequence annotation suggested that the hli gene cluster (GC content 67.2%) spans 47.8 kbp and contains 21 open reading frames (Fig. 2a, Table 1 and Supplementary Tables S1-S3). We assigned these genes to hliA-U.
Five Type-I PKS genes (hliF, G, P, S and T) were predicted to be responsible for the construction of the haliangicin backbone. Other genes could be assigned to a β -methyl branching cassette (hliL, M, N, O, and C), a methoxymalonyl-ACP cassette (hliQ, H, I, J, and K), and modification genes encoding an O-methyltransferase (HliD), a metallo-β -lactamase-type thioesterase (HliE), an acyl-CoA dehydrogenase (HliR), and an epoxidase (HliU). As discussed below, there were particular features in the gene cluster, for example, a rather disorganized gene arrangement (scattered PKS genes and segmented gene cassettes) and the presence of insufficient modules to construct the haliangicin backbone.
Demonstration that this gene cluster is responsible for the biosynthesis of haliangicin was hampered by the slow growth and a severe tendency for cell aggregation of the native producer. We therefore set out to express the biosynthetic machinery for 1 heterologously in a more readily cultivated and more tractable host. Our initial attempt to express the hli genes in a Streptomyces sp. resulted in transformants that showed no titre of haliangicin (1). Finally, we found that Myxococcus xanthus, a fast-growing terrestrial myxobacterium, [19][20][21] is a suitable host. The aforementioned cosmids c7-6E and c10-11C, corresponding to the upstream and downstream parts of hli, respectively, were each modified through λ -Red recombineering 22 (Supplementary Fig. S3 and S4) and integrated into the chromosome of M. xanthus ATCC25232 in a stepwise manner to reconstitute the haliangicin biosynthetic pathway ( Supplementary Fig. S5-S7). The constructed strain, which we named M. xanthus c10-11C/c7-6E, was found to produce haliangicin (1) successfully, as revealed by HPLC (Fig. 3), verifying that the hli gene cluster is responsible for haliangicin biosynthesis. It is intriguing and worth mentioned that the heterologous host did not produce significant quantities of the 12,13-cis-epoxy isomer (cis-haliangicin), which usually accompanies 1 in the original producer (Fig. S8).
Next, we took advantage of the heterologous producer of 1, M. xanthus c10-11C/c7-6E, to settle several issues that were difficult to resolve with the native haliangicin producer.
Optimization of the heterologous production of haliangicin based on the identified biosynthetic precursors. The biosynthetic units for haliangicin (1) were first examined. The producer was fed various 13 C-labelled precursors, and the labelled haliangicins produced were analysed by NMR spectroscopy (Supplementary Fig. S9). This revealed that two acetate units, one acetate-derived methyl branch (9-Me), four propionate units, and three S-adenosylmethionine-derived O-methyl carbon atoms had been incorporated into 1 (Fig. 4a). The remaining two-carbon unit (C3-C4) bearing vicinal oxygen atoms was found to originate from glycerol, as supported by the incorporation of two carbon atoms from [U-13 C 3 ]glycerol ( Supplementary Fig. S9e). Such a unit is known as a glycolate extender in some metabolites. [23][24][25] Secondly, we evaluated the effects of these biosynthetic precursors on the productivity of haliangicin (1) (Fig. 4b). The haliangicin titre increased significantly when sodium acetate was added at a concentration higher than 50 mM and finally reached 11.0 ± 2.1 mg L -1 at 200 mM of sodium acetate after 5 days. The efficiency of haliangicin production was eleven times higher in quantity and three times shorter in culture period than that achieved with the native producer H. ochraceum.
Proposed mechanism for the biosynthesis of haliangicin. The use of the heterologous host for 1 also made gene manipulation much easier, allowing us not only to acquire a complete picture of the somewhat irregular biosynthetic machinery for 1, but also to generate several new analogues of haliangicin. On the basis of the deduced architecture of the hli gene cluster, we propose the haliangicin biosynthetic route shown in Fig. 2b. From the viewpoint of conventional PKS biosynthesis, the construction of the heptaketide backbone of 1 is likely to require six elongation steps following a C3 starter-loading step, but the hli machinery possesses only five elongation modules (modules 1-5 in Fig. 2b). An iterative function of any particular PKS is unlikely, because all the KS domains in hli are classified as modular not iterative ( Supplementary Fig. S10). One possible solution to this problem of a missing module might be the direct use of a diketide (C 6 ) starter, such as 2-methylpent-2-enoyl-CoA (2) ( Fig. 2b and Supplementary Fig. S11), which is suggested by a putative diketide synthase found in the genomes of H. ochraceum and M. xanthus ( Supplementary Fig. S12).
The diketide precursor 2 might undergo γ ,δ -dehydrogenation by HliR, a particular acyl-CoA dehydrogenase, to form 2-methylpent-2,4-dienoyl CoA (3), which might serve as a starter for the formation of the unique 14,15-terminal olefin of 1. A detailed functional analysis of HliR will be described later. The diene starter 3, prepared by HliR, is next loaded into the typical starter-loading module encoded by hliS, and the five PKS-mediated elongation steps take place sequentially to generate the haliangicin backbone (Fig. 2b). Sequence analysis showed that the domain arrangements of the PKSs corresponded well with the chemical structure of haliangicin (1). The module 2-mediated elongation step is followed by the formation of β -methyl branch at the 9-position by the methyl-branching cassette 26,27 (HliC and HliL-O) ( Supplementary Fig. S13). Module 4 incorporates a glycolate extender, biosynthesized by the methoxymalonyl-ACP cassette 24,28 (HliH-K and HliQ) ( Supplementary Fig. S14), to construct the vicinal oxygen functionality at C3-C4. Finally, the nascent carboxylic acid released by HliE, which is homologous to a metallo-β -lactamase-type thioesterase, 29,30 undergoes two post-assembly modification steps: O-methylation at the carboxyl terminus by HliD and epoxidation at the diene terminus by HliU (Fig. 2b).

Verification of some biosynthetic genes and generation of unnatural analogues of haliangicin.
By using the genetically amenable host producer, we manipulated three genes involved in the formation of haliangicin to validate the proposed biosynthetic pathway discussed above and to produce several unnatural analogues of 1 (Fig. 5). First, we targeted the acyl-CoA dehydrogenase gene hliR, which plays an important role in the early stages of haliangicin biosynthesis. Disruption of hliR successfully generated the first unnatural analogue 14,15-dihydrohaliangicin (4), clearly demonstrating that hliR is responsible for the formation of the terminal olefin moiety ( Supplementary Fig. S15). Subsequently, we inactivated two post-assembly tailoring enzymes: HliD (O-methyltransferase) and HliU (epoxidase). Inactivation of hliD led to the production of two free-acid intermediates, 1-O-demethyl-12,13-deoxyhaliangicin (5) and 1-O-demethylhaliangicin (6) as confirmed by  Fig. S16). However, because of their labile nature, these acids could not be purified for NMR analyses; consequently, their structures were chemically confirmed by methylation with trimethylsilyldiazomethane to give 12,13-deoxyhaliangicin (7) and haliangicin (1), respectively ( Supplementary Fig. S17).
Functional analysis of dehydrogenase HliR responsible for terminal alkene formation. Terminal alkene structures are found in a few natural products, for example, curacin A, 31,32 tautomycetin, 33 and FK506 34 The unique terminal vinyl epoxide alkene moiety in haliangicin (1) prompted us to elucidate the biosynthetic mechanism responsible for its formation. Because hliR is homologous to acyl-CoA dehydrogenases, the enzyme encoded by this gene might also be a good candidate for introducing the double bond between C-14 and C-15. Actually, the aforementioned gene disruption of hliR prevented the formation of 1 and resulted in the production of analogue 4 with a saturated terminus, thereby providing unequivocal evidence of the direct involvement of hliR. A detailed sequence and phylogenetic analysis indicated that HliR is homologous to short-chain acyl-CoA dehydrogenases and is closest to the propylmalonyl-ACP desaturase TcsD, which catalyses the desaturation of a propyl group to an allyl group in the side-chain biosynthesis of FK506 [35][36][37] (Fig. 6). Because functional analyses for such terminal dehydrogenases have been limited exclusively to FK506 and its related metabolites 35,38,39 , we next expressed HliR in Escherichia coli as a polyhistidine-tagged protein to analyse its biochemical features ( Supplementary Fig. S19). With the full-length dihydro analogue 4 in hand, we first examined the possibility of HliR-mediated dehydrogenation being the ultimate step in the biosynthesis of haliangicin. The lack of conversion of 4 into 1 by HliR in vitro precluded this hypothesis ( Supplementary Fig. S20). Next, we examined the possibility that formation of the terminal alkene might occur in the starter diketide 2. To prove this, a thio ester mimic of 2, S-(2-methylpent-2-enoyl)-N-acetylcysteamine (2a), was synthesized as the proposed substrate. The recombinant HliR converted 2a entirely into the corresponding dehydro product 3a (Fig. 7a). Preparation of 3a by feeding 2a to the HliR-expressing E. coli yielded a sufficient amount of 3a to permit comprehensive spectroscopic analysis (Supplementary Methods). The substrate specificity was then evaluated by using 14 additional short thio ester mimics, revealing that HliR is highly specific to α ,β -unsaturated short acyl compounds, but does not consume saturated or γ ,δ -unsaturated compounds ( Fig. 7b and Supplementary Fig. S20). Among the substrates tested, S-(pent-2-enoyl)-N-acetylcysteamine (2b), like 2a, was also completely converted into 3b, whereas S-(hex-2-enoyl)-N-acetylcysteamine (2c) and S-(4-methylpent-2-enoyl)-N-acetylcysteamine (2d) underwent partial dehydrogenation. These results suggested that the presence of substituents at the γ -or δ -position might  hinder the dehydrogenation reaction from the enzyme. In addition, unlike TcsD, which demands acyl-carrier protein-tethered substrates, HliR can utilize free small thio esters that mimic acyl-CoA substrates, offering a new strategy for preparing terminal alkenes. (1) is known to show potent inhibitory activity against the plant pathogenic oomycete Phytophthora capsici; in this activity, the terminal β -methoxyacrylate moiety serves as the pharmacophore 12 Because the unnatural analogues 4 and 7 obtained in this study are modified at the terminus opposite the β -methoxyacrylate, additional biological information might be obtained. In a test using P. capsici, analogue 4, which is saturated at the 14,15-olefin, showed the same activity as 1 and the versatile fungicide metalaxyl, whereas analogue 7, lacking the 12,13-epoxide moiety, was 30-fold less active than the other analogues ( Table 2), suggesting that the epoxide plays a key role in this activity. It is noteworthy that these  compounds showed potent cytotoxicity against a tumour cell line and that analogue 7 was more active than the natural compound 1. These compounds showed no inhibitory effects on other microorganisms tested.

Discussion
Since the isolation in 2002 of Haliangium ochraceum as the first marine myxobacterium, there have been few additional examples of this group. Although such a rare group of microorganisms might appear to be inconsequential, their secondary metabolites, once isolated, were found to be chemically and biologically novel. In addition, a report on Type I PKS genes of marine myxobacterial isolates suggested that the genes in marine myxobacteria were much more unusual than those in terrestrial strains. One of the greatest difficulties in studying marine myxobacteria is their fastidious nature (slow growth, low secondary metabolism, cell aggregation, etc.), which also makes it difficult to apply recently developed techniques for genetic manipulation. Therefore, the present work provides the first insights into biosynthesis in marine myxobacteria. The successful heterologous expression of the biosynthetic machinery for the PKS-type metabolite haliangicin (1) provided several benefits that would be difficult to achieve with the native producer, especially the improved productivity of the fungicide and the production of unnatural analogues, one of which was a more potent cytotoxin than the natural 1. The easier access to the biosynthetic machinery of 1 in a different organism also led to the functional characterization of a unique dehydrogenase that introduces the terminal alkene moiety of 1, thereby widening our understanding of the biosynthetic mechanism underlying the formation of terminal alkene groups. Such γ ,δ -dehydrogenation reactions hold promise for generating novel structures in vitro in the future. Genomic information on the native antibiotic producer H. ochraceum, which suggests the presence of six other latent biosynthetic gene clusters, prompts us to analyse them by applying our present expertise to produce novel metabolites that are not obtained by cultivation of the native marine myxobacterium.

Methods
Bacterial strains, plasmids, primers and culture conditions. All bacterial strains, vectors, and oligonucleotides (STAR Oligo; Rikaken Co., Ltd., Nagoya) used in this study are listed in Supplementary Tables  S4-S6. The marine myxobacterium Haliangium ochraceum SMP-2 was cultivated as previously described 11 . CTT medium 40 was used for the preculture of Myxococcus xanthus ATCC25232 (wild type) and all its mutants at 30 °C. Production medium 41 supplemented with 2-4% (w/v) Sepabeads SP207 resin (Mitsubishi Chemical Co., Tokyo) and a 0.1% (v/v) trace-elements solution (TES) 42 was used for the heterologous production of haliangicin (1) and its related metabolites. A PCR amplification was carried out with the degenerate primers KS1UP/KSD1 44 by using the H. ochraceum genomic DNA as a template to obtain two KS fragments HKS1 and HKS2 ( Supplementary Fig. S1). The genomic cosmid library containing about 1700 cosmids was then screened for PKS genes by colony hybridization with the above-mentioned KS probes HKS1 and HKS2. Five cosmids, c3-5F, c4-6B, c7-6E, c9-4F, and c10-11C, containing KS domain(s) were obtained (Supplementary Fig. S2). Each cosmid was end-sequenced (Supplementary  Table S7) by forward-primer DWF1 and reverse-primer DWR. Of these five cosmids, c7-6E and c10-11C were subjected to shotgun sequencing (Takara Bio Inc., Tokyo). Two discrete contigs of 33,197 bp and 22,829 bp were obtained and the remaining 7,420 bp were further sequenced by the genome-walking approach. One continuous region of 63,446 bp was finally assembled and subjected to sequence analysis.

Analysis of the haliangicin biosynthetic gene cluster (hli).
The obtained continuous contig of 63.4 kbp was annotated by various sequence-analysis tools, including BLAST, CDD, 45 and antiSMASH analysis 46 . The multiple alignments of amino acid sequences were generated by using the Clustal Omega program (http:// www.ebi.ac.uk/Tools/msa/clustalo/) provided by the European Molecular Biology Institute (Hinxton, UK) and neighbour-joining trees of proteins were generated by Geneious Tree Builder. KS domains in hli were extracted and analysed by using the NaPDoS Web tool (http://napdos.ucsd.edu/) 47 .
Heterologous expression of the hli gene cluster in M. xanthus. Cosmids c7-6E and c10-11C were modified through λ -Red recombineering. All techniques pertaining to the λ -Red recombineering are described in the published protocol 22 . For detailed modification procedures, refer to the Supplementary Methods and Figs S3 and S4. The resulting cosmids c7-6E TA-Kan R and c10-11C Tet R were introduced into the chromosome of M. xanthus ATCC25232 in a stepwise manner ( Supplementary Fig. S5). Namely, the first cosmid c7-6E TA-Kan R was integrated into the genome of M. xanthus through the TA fragment that is a 1.8 kbp internal fragment homologous to ta-1 responsible for the myxovirescin A biosynthesis 48 . The second cosmid c10-11C Tet R was then introduced through the 14 kbp region shared with the first cosmid. Electrocompetent cells of M. xanthus were prepared according to the reported methods 21,49 . In the first stage, 1.5 μ g of c7-6E TA-Kan R was added to 100 μ L of a chilled cell suspension of M. xanthus ATCC25232 and electroporated (Gene Pulser TM , BIO-RAD) at 25 μ F, 1.8 kV, and 200 Ω in 0.2 cm electroporation cuvettes. The kanamycin-resistant colonies (named as M. xanthus c7-6E TA-Kan R ) appeared after seven days and were verified by colony PCR using specific primers to detect the integration of c7-6E TA-Kan R into the chromosome of M. xanthus ATCC25232 (Supplementary Fig. S6a). The expected homologous recombination of the TA fragment was also confirmed by PCR (Supplementary Fig. S6b). Subsequently, 1.7 μ g of the second modified cosmid c10-11C Tet R was electroporated into M. xanthus c7-6E TA-Kan R under the same conditions as those for c7-6E TA-Kan R through the 14 kbp region shared with cosmid c7-6E. Mutants were selected on CTT agar plate (oxytetracycline 12.5 μ g/mL) for one week and checked by colony PCR using specific primers (Supplementary Fig. S7). The obtained construct carrying the entire length of hli was named M. xanthus c10-11C/c7-6E. The confirmation of the heterologous production of haliangicin (1)   72 hours, respectively, to give a final concentration of 4 mM. After cultivation for six or seven days in total at 30 °C and 180 rpm, labelled haliangicin was extracted, purified, and analysed by 13 C NMR spectroscopy ( Supplementary  Fig. S9).
Optimization of haliangicin production in the M. xanthus heterologous host. The heterologous host M. xanthus c10-11C/c7-6E was cultured in the production medium. 1 M Aqueous solutions of sodium acetate, sodium propionate, and glycerol were prepared, filter-sterilized, and added to the autoclaved culture medium. l-Methionine was added before autoclaving. For higher concentrations of sodium acetate (100 mM or 200 mM), an autoclaved 5 M stock solution was added in two portions to the medium on the second and third day, respectively, to avoid inhibition of the growth of M. xanthus in its early phase. After cultivation at 30 °C and 180 rpm for the scheduled time, the resins and cell pellets were separated from the medium by centrifugation (6000 rpm, 10 min) and extracted twice with acetone. The crude extracts were filtered, evaporated to dryness, and subjected to HPLC analysis [Develosil ODS-UG-5 (φ 4.6 × 250 mm), λ = 290 nm; 75% MeOH-H 2 O, flow rate: 1 mL/min]. The results are summarized in Supplementary Fig. 4b.