Characterization of the biosynthetic gene cluster of the polyene macrolide antibiotic reedsmycins from a marine-derived Streptomyces strain

Background Polyene antibiotics are important as antifungal medicines albeit with serious side effects such as nephrotoxicity. Reedsmycin (RDM) A (1), produced by marine-derived Streptomyces youssoufiensis OUC6819, is a non-glycosylated polyene macrolide antibiotic with antifungal activity comparable to that of clinically used nystatin. To elucidate its biosynthetic machinery, herein, the rdm biosynthetic gene cluster was cloned and characterized. Results The rdm cluster is located within a 104 kb DNA region harboring 21 open reading frames (ORFs), among which 15 ORFs were designated as rdm genes. The assembly line for RDM A is proposed on the basis of module and domain analysis of the polyketide synthetases (PKSs) RdmGHIJ, which catalyze 16 rounds of decarboxylative condensation using malonyl-CoA as the starter unit (loading module), two methylmalonyl-CoA (module 1 and 2), and fourteen malonyl-CoA (module 3–16) as extender units successively. However, the predicted substrate specificity of AT0 in the loading module is methylmalonyl-CoA instead of malonyl-CoA. Interestingly, the rdm cluster contains a five-gene regulation system RdmACDEF, which is different from other reported polyene gene clusters. In vivo experiments demonstrated the XRE family regulator RdmA and the PAS/LuxR family regulator RdmF function in negative and positive manner, respectively. Notably, inactivation of rdmA and overexpression of rdmF led to increased production of RDM A by ~ 2.0-fold and ~ 2.5-fold, reaching yields of 155.3 ± 1.89 and 184.8 ± 9.93 mg/L, respectively. Conclusions Biosynthesis of RDM A is accomplished on a linear assembly line catalyzed by Rdm PKSs harboring a unique AT0 under the control of a complex regulatory system. These findings enable generation of new biologically active RDM derivatives at high yield and with improved properties by engineered biosynthesis. Electronic supplementary material The online version of this article (10.1186/s12934-018-0943-6) contains supplementary material, which is available to authorized users.


Background
Pathogenic fungi are a leading cause of human mortality, particularly among an ever-increasing population of immunocompromised individuals, and have become critical threats to global health [1]. Polyenes were the first antifungal drugs for clinical use. Compared to any other antifungal agents, they have the broadest spectrum of activity [1]. Up to date, polyene macrolide antibiotics have been proved to be one of the most effective antifungal agents, especially when dealing with life-threatening systemic fungal infections [2]. However, the emergence of rare or unidentified species of drug-resistant fungal pathogens and the clinical need for non-toxic antifungal drugs demand us to develop new polyene derivatives and formulations, given the apparent toxicity and serious side effects of existing polyene macrolides [2,3].
The biosynthetic mechanisms for several polyene macrolide antibiotics, including candicidin (FR-008) [4,5], amphotericin [6], pimaricin [7], nystatin [8], tetramycin [9], and NPP [10] have been widely studied. The formation of the macrolactone ring is typically catalyzed by a modular type I polyketide synthetase (PKS), after which two tailoring steps usually occur to furnish the exocyclic carboxyl group and the mycosamine sugar(s) [11]. Engineering of the polyene PKSs and tailoring genes have generated several polyene analogues. For example, inactivation of KR16 in amphotericin PKS resulted in 7-oxoamphotericin B, which had good antifungal activity and was less hemolytic than amphotericin B [12]. By inactivation of the corresponding P450 gene in each cluster, the exocyclic carboxyl group was replaced with a methyl group in amphotericin B [13], nystatin [14], candicidin [15], and rimocidin [16], leading to decreased haemolytic activity but not antifungal activity. Recently, Kim et al. [17] carried out cross genetic complementation of the Pseudonocardia autotrophica ∆nppY mutant strain with nypY from Pseudonocardia sp. P1, leading to isolation of a mannosylated NPP analogue with reduced antifungal activity while higher nephrotoxical activity against human hepatocytes. Albeit many potentially valuable polyene analogues have been developed, none of them has advanced into clinical medicine so far [11].
Genome sequencing has uncovered an increasing number of polyene biosynthetic gene clusters, facilitating genome-directed novel polyene discovery [11]. Previously, we did genome scanning of the reeds rhizosphere soil-derived Streptomyces youssoufiensis OUC6819 (previously Streptomyces sp. CHQ-64) and found a 28-kb DNA fragment encoding two partial type I PKSs [18]. Bioinformatic analysis indicated it is comprised of 6 modules putatively encoding a polyene-polyol compound. Inspired by this fact, we identified a series of reedsmycins (RDMs 1-5, Fig. 1), featuring a polyene-polyol macrolide devoid of sugar moiety. Notably, the major component reedsmycin A (1) (RDM A) exhibited comparable inhibition against Candida albicans as that of the positive control nystatin [18]. The purpose of this study was to verify the involvement of this 28-kb DNA fragment in the biosynthesis of RDMs, and characterize the entire rdm gene cluster to set the stage to genetically engineer new bioactive RDM analogues at high yield. Herein we report (i) the cloning and sequencing of the rdm gene cluster; (ii) bioinformatics analysis of the rdm cluster and a proposed assembly line of RDM A; (iii) gene inactivation to support function of the rdm cluster; (iv) genetic characterization of the unique regulatory system for RDM biosynthesis.

Cloning and sequencing of the entire rdm cluster
The cosmid library of S. youssoufiensis OUC6819 was firstly constructed and screened by PCR, resulting in four positive cosmids pWLI511-514 (Fig. 2a). The sequence gaps within the PKS genes were filled by chromosomal walking using the pWLI512 or pWLI513 as template. Combined with our previous genome scanning results, a 104-kb DNA sequence containing the entire rdm cluster was obtained and deposited in GenBank database under the accession number MG947597. Twenty-one open reading frames (ORFs) were identified within the sequence (Fig. 2b), and their proposed functions were summarized in Table 1, among which fifteen ORFs were designated as rdm genes. The deduced encoding products of the rdm genes include four PKSs (RdmG, RdmH, RdmI, and RdmJ), five regulatory proteins (RdmA, RdmC, RdmD, RdmE and RdmF), one resistance protein RdmK, one PaaI family thioesterase (TE) RdmM, one acyl carrier protein (ACP) RdmN, one acyl-CoA synthetase RdmO, and two proteins with predicted functions of methyltransferase (RdmB) and ferritin (RdmL) ( Table 1).
The functions of all the domains were deduced based on sequence homology. Except the loading module, which contains a mutated ketosynthase (KSq) with Cys replaced by Gln, the other KS domains contain the CHH/N catalytic triad required for the decarboxylative condensation [19]. All the ACP domains display the conserved L/IG(x)DS motif with Ser being essential for 4′-phopspho-panthehenylation [20]. The substrate specificity of the AT domains was predicted by sequence multiple alignments and comparison of key residues to published data [21]. AT3-8 and AT15-16 were predicted to recognize malonyl-CoA as substrate, which is the same as that of AT9-14 [18], conversely AT0-2 were predicted to be specific for methylmalonyl-CoA (Fig. 4).
All the modules harbor KR domains with the conserved NADP(H)-binding motif and the catalytic triad KSY [22], among which KR9-11 and KR12-14 were previously predicted to be A-type and B-type, respectively [18]. We analyzed all the other KRs and found that: (i) KR1-3 and KR15-16 belong to B-type with the conserved LDD-motif in KR1-3, or the replaced LED-motif in KR15-16, thus putatively generating R-configured alcohols; (ii) KR6-7 possess a Trp residue and are A-type, catalyzing the formation of S-configured alcohols; (iii) conversely, no signature motif was seen in KR4-5 and KR8, and thus it is difficult to predict the corresponding products (Additional file 1: Table S4). In addition to modules 12-14, functional DH domains featuring the conserved consensus sequence HxxxGxxxP [23] were found for modules 1-3 and 15-16 which catalyze elimination of water from R-configured hydroxyl group to form E-configured double bond. Only one ER domain was identified in module 1 with the conserved motif LxHxxxGGVGxx-AxxxA [24].
Thus the entire assembly line for RDMs was proposed as shown in Fig. 3. RdmG-J catalyze the RDMs biosynthesis by carrying out 16 rounds of decarboxylative condensation using malonyl-CoA as the starter unit (loading module), two methylmalonyl-CoA (module 1 and 2),  Table 1 and fourteen malonyl-CoA (module 3-16) as extender units. A full-length polyketide intermediate is offloaded and cyclized by the dedicated RdmJ-TE domain, generating RDM A. The present analysis of RdmG-J well supported the proposed pathway but with the following two inconsistencies: (i) phylogenetic analysis supported AT0 to be specific for methylmalonyl-CoA (Fig. 4), however, the skeleton of RDMs indicates AT0 is supposed to utilize malonyl-CoA instead (Fig. 3); (ii) albeit the presence of an active DH domain in module 2, it is most likely "skipped" due to unknown reason (Fig. 3).

Genes encoding regulators
There are five ORFs (rdmA, rdmC, rdmD, rdmE and rdmF) located upstream of the PKS genes. As shown in Table 1, RdmA is a putatively XRE family transcriptional regulator with 57% identity/65% similarity to ASD26_RS06115 (WP_056788122.1) from Streptomyces sp. Root1319. RdmC-E are putative LuxR family twocomponent system (TCS) regulators homologous to KasWX from the biosynthetic gene cluster of kasugamycin in Streptomyces kasugaensis [25]; RdmC is the homolog of the response regulator KasW, containing a signal receiver domain REC and a helix-turn-helix (HTH) DNA binding domain; interestingly, RdmD and RdmE display homology to different regions of the same protein, the sensor kinase KasX (BAF79687.1). RdmF is located right adjacent to the PKSs and harbors a PAS fold motif (pfam08448) at N-terminus and a LuxRtype HTH DNA binding domain at C-terminus, with 83% identity/89% similarity to a hypothetical protein AOM46_RS26735 (WP_062215992.1) from Streptomyces sp. NBRC 109706, and 38% identity/51% similarity to NysRIV (AAF71781.1) from the nystatin gene cluster. In comparison with the regulators in the other polyene biosynthetic gene clusters, PAS-LuxR superfamily regulators are well-conserved and function in positive manner [26]; conversely, no homologs of RdmA and RdmC-E were found in these clusters.

Functional confirmation and determination of cluster boundaries of the rdm gene cluster
To detect if this locus is indeed involved in RDM biosynthesis, gene inactivation was performed. The PKS genes rdmG, rdmH and rdmJ were replaced with the apramycin resistance cassette, respectively (Additional file 1: Figures S1-S3). Fermentations of the resulting mutants were carried out followed by HPLC and LC-MS analysis (Additional file 1: Figure S4). The results showed that ∆rdmG, ∆rdmH and ∆rdmJ all abolished production of RDMs as expected (Fig. 5iii-v), confirming this locus is indeed involved in RDMs biosynthesis. Based on the sequence analysis result, orf(−1) and orf(−2), which are located upstream of the regulatory gene rdmA and putatively encode a methyltransferase and a FAD-binding monooxygenase (Table 1), respectively, were inactivated to determine the upstream boundary (Additional file 1: Figures S5, S6). Orf1 is homologous to the hypothetical protein AMJ94_18775 (KPK85730.1) from Deltaproteobacteria bacterium SM23_61 and is interestingly split into two proteins AOK24_RS06255 and AOK24_ RS06250 in S. niveiscabiei (Table 1). Orf1 was inactivated to determine the downstream boundary (Additional file 1: Figure S7). HPLC analysis results showed that no impact was observed for all the three mutants ( Fig. 5viviii), suggesting they are probably located outside of the gene cluster.

A complex regulatory system featuring a negative regulator RdmA and a positive regulator RdmF
To probe the function of the regulatory genes, rdmA and rdmC-F were each inactivated as described in the "Methods" section. Fermentations and HPLC analysis were carried out for the wild type strain and the confirmed mutants. The production of RDM A was quantified in each strain as described in "Methods" section, and the results showed: (i) inactivation of rdmA led to increased production of RDM A by ~ 2.0-fold, reaching a considerable yield of 155.3 ± 1.89 mg/L, as compared to 73.8 ± 0.76 mg/L in the wild type strain (Fig. 6aiii, d), suggesting RdmA functions as a negative regulator; (ii) in contrast, inactivation of rdmF completely abolished the production of RDMs (Fig. 6biii, d), supporting it to be a positive regulator; (iii) while no obvious influences were observed for the ∆rdmC (Fig. 6ciii) and ∆rdmD (Fig. 6civ) strains, inactivation of rdmE resulted in moderately decreased production of RDM A to 50.5 ± 0.49 mg/L (Fig. 6cv, d). To further confirm the function of rdmA and rdmF, genetic complementation and gene overexpression were performed. HPLC analysis results indicated that the genetically complemented strains of ∆rdmA (Fig. 6aiv, d) and ∆rdmF almost restore the production of RDMs (Fig. 6biv, d) to levels comparable to those in the wild type strain (Fig. 6a, bii). Overexpression of rdmA decreased the production of RDM A by ~ 1.4-fold to a yield of 52.5 ± 0.82 mg/L (Fig. 6av, d), and conversely, overexpression of rdmF enhanced the titer of RDM A remarkably by ~ 2.5-fold, reaching 184.8 ± 9.93 mg/L (Fig. 6bv, d), supporting their contrary action of manners.

Other genes in the rdm cluster involved in RDM biosynthesis
As indicated in Table 1, in addition to the five regulatory genes, a putative methyltransferase gene rdmB is found upstream of the PKS genes. To detect if rdmB is related to RDM biosynthesis, rdmB was inactivated, resulting in no influence at all on the production of RDMs (Fig. 7iii), indicating it might be unnecessary for RDM biosynthesis. The downstream genes adjacent to the PKS genes exhibit considerably high homology to genes from Streptomyces niveiscabiei, with RdmK-Orf3 being homologous to AOK24_RS06280-AOK24_RS06240 (Table 1), respectively. We further searched the neighboring region of this locus in the genome of S. niveiscabiei (NZ_ LIRL00000000.1), and found the presence of a hypothetical protein AOK24_RS06285 (WP_055719057.1) containing a conserved family A glycosyltransferase motif instead of PKSs. rdmK encodes a putative MFS transporter, suggesting it is probably a resistance gene; RdmL harbors a conserved ferritin-like motif, which participate in a range of functions such as iron regulation, mono-oxygenation, and reactive radical production; interestingly, rdmM encodes a PaaI family TE belonging to the TE13 family, instead of a type II TE belonging to the TE18 family normally found in PKS gene clusters [27]; the putative acyl-CoA synthetase RdmO is possibly involved in substrate formation for the PKSs. To detect if these genes are involved in RDMs biosynthesis, rdmL, rdmM and rdmO were inactivated, respectively. As shown in Fig. 7, no impact was observed when rdmL was inactivated (iv), but the production of RDMs in ∆rdmM (v) and ∆rdmO (vi) was obviously decreased by ~ 1.4-fold as compared with that in the wild type strain (ii), indicating they are probably involved in the biosynthesis of RDMs. The decreased instead of abolished production of RDMs might be due to presence of homologous genes of rdmM and rdmO in other places of the genome.

Discussion
High throughput sequencing has greatly promoted genome-directed natural product discovery as well as understanding of underlying biosynthetic machinery. Herein, we identified and characterized the entire biosynthetic gene cluster of the polyene macrolide compound RDMs, and proposed its biosynthetic pathway based on the genetic organization of the rdm gene cluster and in vivo gene manipulation, setting the stage for generating novel RDM analogs with improved bioactivities.
The module and domain organization of RdmG-J is colinear with the RDM structure excluding the predicted superfluous DH domain in module 2 (Fig. 3). In general, KSq-type loading ATs exhibit a strict substrate specificity with selection of either malonyl-CoA or methylmalony-CoA, which is decarboxylated in situ to provide acetyl or propionate starter units for the polyketide initiation [28]. Interestingly, the KSq-type loading AT0 of RdmG is presumably to recognize malonyl-CoA based on the structure of RDM A, however, it is clustered with ATs recognizing methylmalonyl-CoA instead (Fig. 4). Further biochemical experiments would be performed to elucidate the substrate selectivity of RdmAT0.
Stereochemistry determination is always very challenging for the polyol system of polyene macrolides antibiotics. During the PKS assemble line, the hydroxyl groups are formed by KR domains in a stereospecific manner, thereby bioinformatics analysis would contribute to configuration assignment. In the case of RDM A, S-configured hydroxyl groups at C-13, C-15, C-17, C-21 and C-23 of RDMs would be expected, resulting respectively from A-type KRs 11, 10, 9, 7 and 6; conversely configurations of hydroxyl groups at C-19, C-25 and C-27 are not predictable being lack of conserved residues. Nevertheless, X-ray crystallography would be necessary to determine their stereochemistry. Being the dominant product, RDM A (1) harbors six trans (E)-double bonds (Δ 2 -Δ 10 and Δ 28 ) (Fig. 1). This is supported by presence of DH domains in module 3, and 12-16, which catalyze syn elimination operating on R-configured β-hydroxy intermediates (B-type ketoreduction). The occurrence of cis-double bonds in RDM B-E (2-5) might be due to spontaneous trans-cis isomerization [11]. In comparison to other polyene macrolide gene clusters, such as pimaricin [7], amphotericin [6], candicidin [4,5], nystatin [8], NPP [10] and tetramycin [9] biosynthetic gene clusters, the rdm gene cluster is devoid of mycosamine biosynthetic genes and cytochrome P450 genes, which are involved in tailoring steps after macrocyclization. Many studies have demonstrated that mycosaminyltransferases tolerate structural changes in their aglycones and show moderately strict specificity for NDP-sugar donors [11,29]. For example, mycosaminyltransferases AmphDI (in the amphotericin gene cluster) and NysDI (in the nystatin gene cluster) are capable to mycosaminylate candicidin and pimaricin aglycones in vivo albeit with less efficiency compared to the cognate enzymes [29]. The broad substrate promiscuity displayed by mycosaminyltransferases could enable addition of sugar moiety onto RDM A, and thus optimize its pharmacological properties, as glycosylation can usually improve water solubility and reduce toxicity [30,31]. In general, hydroxylation or epoxidation further happens after glycosylation, which is accomplished by cytochrome P450 monooxygenases. These enzymes were demonstrated to be able to recognize substrates with altered structures [32], allowing for possibility of combinatorial engineering of them with the rdm biosynthetic gene cluster to generate RDM analogues.
The rdm gene cluster seems to have a regulation system different from those of the other reported polyene gene clusters. In addition to the well conserved PAS/ LuxR family regulatory gene rdmF, the rdm cluster also harbors an XRE family transcriptional regulator RdmA and a set of LuxR family two-component system RdmC-E. Interestingly, the sensor kinase gene was split into two genes rdmD and rdmE, which are homologous to different regions of kasX from the kasugamycin gene cluster. Inactivation of rdmD had no obvious impact on RDM production, while inactivation of rdmE led to decreased production of RDM A by ~ 1.5-fold (Fig. 6c, d), different from KasX acting in a negative manner during kasugamycin biosynthesis [33]. The hierarchical relationships between the rdmACDEF genes are under investigation in our lab. Elucidation of the regulation mechanism would contribute to boosting the production of bioactive RDM compounds in a rational way.

Conclusions
The biosynthetic gene cluster of non-glycosylated polyene antibiotic RDMs was cloned and characterized in vivo. The proposed model for RDM A assembly line agrees well with its chemical structure, and is supported by gene inactivation. A complex regulatory system consisting of 5 genes was characterized by gene inactivation, genetic complementation as well as gene overexpression, leading to yield improvement of RDM A by ~ 2.0-to 2.5-fold. These findings set the stage for further generation of new biologically active RDM derivatives with improved properties and for yield enhancement by genetic engineering.

Bacterial strains, plasmids, and culture conditions
Bacterial strains and plasmids used in this study are listed in Additional file 1: Table S1 and the PCR primers in Additional file 1: Tables S2, S3. Escherichia coli strains including DH5α, BW25113/pIJ790 and ET12567/ pUZ8002 were cultivated at 37 °C in Luria-Bertani (LB) liquid medium or on LB agar. S. youssoufiensis OUC6819 and its derivatives were cultured at 30 °C in TSBY (yeast extract 5 g/L, tryptic soy broth 30 g/L, sucrose 103 g/L) for DNA extraction and on International Streptomyces Project Synthetic Salts-Starch Medium (ISP4) agar with 0.5% glycine for genetic manipulation and on R2YE agar for sporulation. For fermentation, spores of the strain S. youssoufiensis OUC6819 and its derivatives were inoculated in fermentation medium (soluble starch 10 g/L, glucose 20 g/L, corn syrup 4 g/L, yeast extract 10 g/L, beef extract 3 g/L, MgSO 4 ·7H 2 O 0.5 g/L, KH 2 PO 4 0.5 g/L, CaCO 3 2 g/L, sea salt 30 g/L, pH = 7.0) and incubated at 30 °C, 220 rpm for 7 days in a 250 mL flask. When necessary, the medium was supplemented with apramycin 50 μg/mL, chloramphenicol 25 μg/mL, kanamycin 100 μg/mL, thiostrepton 25 μg/mL, or ampicillin 100 μg/ mL. Common biochemicals and chemicals were purchased from commercial sources.

Genomic library construction and screening
Genomic DNA of S. youssoufiensis OUC6819 was partially digested with Sau3AI, and fragments with the size of 40-50 kb were recovered and dephosphorylated with FastAP (Thermo Scientific, Pittsburgh, USA), and then ligated into SuperCos1 that was pretreated with XbaI, dephosphorylated, and digested with BamHI. The ligation product was packaged into lambda particles with the MaxPlax Lambda Packaging Extract (Epicenter, Madison, WI, USA) as per the manufacture's instruction and plated on E. coli Top10. The titer of the primary library was about 5 × 10 6 cfu per μg of DNA. The primer pairs used for cosmid library screening are listed in Additional file 1: Table S2.

DNA manipulation and bioinformatic analysis
All DNA manipulations were performed according to standard procedures [34] or manufacturer's instruction. Plasmid extractions and DNA purification were carried out using commercial kits (OMEGA, BIO-TEK). Both primer synthesis and DNA sequencing were performed at Tsingke Biotech Co. Ltd. (Qingdao, China). ORF assignments and their proposed function were accomplished by using the FramePlot4.0beta (http://nocar dia. nih.go.jp/fp4) [35]. Sequence comparisons and database searches were carried out by BLAST algorithm [36] (http://blast .ncbi.nlm.nih.gov/Blast .cgi). Domain analysis were accomplished at SBSPKS [37]. Additional sequence alignments were conducted by ClustalX [38].

Construction of gene inactivation mutants
Gene inactivation in S. youssoufiensis OUC6819 was performed using the REDIRECT Technology according to the literature protocol [39]. For the construction of gene mutants, the aac(3)IV-oriT resistance cassette from pIJ773 or pMT3 was amplified to replace an internal region of the target gene. The target gene disruption plasmids were transformed into E. coli ET12567/pUZ8002, and the conjugation between E. coli ET12567/pUZ8002 and S. youssoufiensis OUC6819 was performed using ultrasonic fragmented mycelia as acceptors. The ultrasonic program is set to cycle 20 times with 3 s ON and 4 s OFF. Desired mutants were selected by apramycin-resistant and kanamycin-sensitive phenotype, followed by PCR confirmation (Additional file 1: Table S3).

Genetic complementation and overexpression
For genetic complementation and overexpression, rdmA and rdmF was each put under the control of the constitutive promoter P gapdh from S. youssoufiensis OUC6819. Construction of rdmA expression plasmid was performed as follows: P gapdh was amplified using primer pair of pGFP/3′-OH phosphorylated pGRP (Additional file 1: Table S3) and was digested with EcoRI; the rdmA fragment was amplified with the primer pair of rdmAEF/ rdmAER (Additional file 1: Table S3) and digested with XbaI; and then these two digested fragments were ligated and cloned into the EcoRI and XbaI sites of pSET152C to give pWLI515 (Additional file 1: Table S1). Similarly, the fragments of rdmF were cloned into the same sites of pSET152C to obtain pWLI516 (Additional file 1: Table S1). After confirmation by sequencing, the resulting plasmids were passed through E. coli ET12567/ pUZ8002 and introduced into the rdmA and rdmF mutants for genetic complementation or into the wild type S. youssoufiensis OUC6819 for overexpression via conjugation.

Production and analyses of Reedsmycin A
Spores of Streptomyces strains were inoculated into 50 mL medium in a 250 mL flask, and were incubated on a rotatory shaker at 30 °C, 220 rpm for 7 days. The fermentation cultures were harvested by centrifugation and the supernatant was extracted twice with an equal volume of ethyl acetate. The precipitate was extracted by acetone and dried in vacuo then merged to ethyl acetate. After the residue was dried in vacuo, the crude extract was dissolved in 1.5 mL methanol, filtered through a 0.2 μm filter, and subjected to HPLC analysis with injection volume of 25 μL for each sample. The HPLC system consisted of Agilent 1260 Infinity Quaternary pumps and a 1260 Infinity diode-array detector. Analytical HPLC was performed on an Eclipse C18 column (5 μ, 4.6 × 150 mm) developed with a linear gradient from 40 to 65% ACN/H 2 O in 15 min followed by an additional 15 min at 100% ACN at flow rate of 1 mL/ min and UV detection at 365 nm. The identity of reedsmycin A (RDM A) was unambiguously confirmed by comparison with authentic standard as well as HR-ESI-MS analysis. HR-ESI-MS was carried out on Thermo LTQ-XL mass spectrometer. RDM A was quantified on the basis of peak area at 365 nm using standard curve obtained with RDM A standard. All experiments were repeated at least three times.

Nucleotide sequence accession number
The sequence of the rdm biosynthetic gene cluster has been deposited in GenBank under the accession number MG947597.