Introduction

Screening for bioactive microbial secondary metabolites has been performed for more than 70 years. Although the discovery of bioactive natural products has resulted in great benefits to humankind, the rate of discovery of skeletally-novel compounds from microorganisms has decreased significantly over time [1, 2]. To address this situation, researchers investigating natural products have made great effort to exploit the synthetic potential of microorganisms using a variety of media [3] and co-culture systems [4], and by inducing mutations in candidate microbes [5]. In contrast to these more conventional methods, recent advances in sequencing technology have revealed the presence in microorganisms of many cryptic gene clusters thought to be capable of the biosynthesis of secondary metabolites [6, 7]. This observation reminds us that we have not fully applied the abilities of microorganisms to produce various bioactive compounds. We hypothesized that heterologous expression of these cryptic biosynthetic genes will serve as a process for inducing the production of novel compounds because it has already been reported that heterologous expression can initiate even a cryptic biosynthetic gene cluster and confirm its metabolite [8]. While many studies on heterologous expression have been performed, there exist (to our knowledge) only a limited number of reports of the heterologous production of middle-molecular-weight compounds such as macrolides, which are biosynthesized by type I polyketide synthases (PKSs). This fact presumably reflects the extremely large sizes of the corresponding genes and gene clusters, which would have led to technological challenges in cloning, transformation, and expression. To overcome these technical difficulties, we have been employing bacterial artificial chromosomes (BACs), which permit the cloning of DNA fragments up to 300 kb in size [9]. Using this technology, we have succeeded in the heterologous production of mediomycin and neomediomycin [10], JBIR-156 [11], quinolidomicin [9], and desertomycin [12], molecules generated by the products of biosynthetic gene clusters (BGCs) that are 161, 183, 137, 213, and 127 kb in length, respectively.

Concanamycin A (folimycin) (Fig. 1) [13, 14], originally isolated as an antifungal metabolite of Streptomyces neyagawaensis nov. sp. [15], is a potent inhibitor of the vacuolar-type proton-ATPase (V-ATPase) [16]. The concanamycin BGC, an approximately 100-kb locus consisting of six genes encoding type I PKSs, was identified by assembling the DNA sequences of three cosmid clones generated from the genome of S. neyagawaensis ATCC 27449 [17]. However, the heterologous expression of this BGC has not been achieved. Therefore, we targeted concanamycin(s) for the further development of our heterologous expression system for the biosynthesis of polyketide compounds derived from modular polyketide synthases. Heterologous expression of the BAC clone pKU503ccn, which harbors the concanamycin BGC, resulted in the production of concanamycins at high yield, compared to the wild-type strain, S. neyagawaensis IFO 13477. Interestingly, we found that a host strain carrying this BAC also produced two other aromatic polyketides (Fig. 1) when cultured in a production medium distinct from that used to obtain concanamycin. Therefore, we also report herein a newly identified BGC for the production of ent-gephyromycin and a novel compound that we designate JBIR-157.

Fig. 1
figure 1

Compounds identified in this study. The structure of concanamycin A (presented here as a representative concanamycin) and ent-gephyromycin (1) are shown. The proposed substructure of JBIR-157 (2) is shown with key correlations obtained via COSY and HMBC experiments

Full size image

Materials and methods

Bacterial strains and growth conditions

Streptomyces neyagawaensis IFO 13477 was obtained from Institute for Fermentation, Osaka, Japan (currently, the microorganism can be distributed as NBRC 13477 from National Institute for Technology and Evaluation Biologocal Resource Center, Chiba, Japan). The growth of Escherichia coli strains DH5α, NEB10β (New England Biolabs, Ipswich, MA), and GM2929 hsdS::Tn10 used Luria broth (LB), which contained (per L of deionized water) 10 g of tryptone, 5 g of yeast extract, and 5 g of NaCl, and was adjusted to pH 7.5 following formulation. The corresponding solid medium was prepared by supplementing LB with agar at 15 g L−1. For the production of secondary metabolites, S. neyagawaensis IFO 13477, S. avermitilis SUKA32 [11], or SUKA54 [18] was inoculated into a 50-mL test tube containing 15 mL of seed medium, containing (per L of deionized water) 5 g of glucose, 15 g of soya flour, and 5 g of yeast extract; the cells were cultured with reciprocal shaking at 320 rpm and 27 °C for 2 days. A 375-μL aliquot of the resulting vegetative culture was used to inoculate a 125-mL flask containing 15 mL of either of two production media. Medium 1, used for the production of concanamycins, contained (per L of deionized water) 40 g of β-cyclodextrin, 20 g of Pharmamedia, 5 g of glycerol, 5 mg of ZnSO4·7H2O, 5 mg of CuSO4·5H2O, and 5 mg of MnCl2·4H2O; the pH was adjusted to 7.0 following formulation. Medium 2 (synthetic medium [19]), used for the production of ent-gephyromycin and JBIR-157. The fermentation cultures were grown on a rotary shaker at 180 rpm and 27 °C for 7 days. For large-scale preparations of the products, cells were grown under the same conditions in baffled 500-mL flasks, each containing 100 mL of the respective production medium.

Isolation of a concanamycin BGC-containing BAC

Genome sequencing of the S. neyagawaensis IFO 13477 strain was performed using PacBio RS II (Pacific Biosciences, Menlo Park, CA) and Roche 454 GS FLX Titanium chemistry (Roche, Basel, Switzerland). The sequence data were assembled using HGAP2 (Pacific Biosciences) and Newbler software package (Roche). A BAC library of S. neyagawaensis IFO 13477 was constructed with pKU503 [20] according to a previously reported protocol [8]. BAC clones obtained were transferred into 384-well plates containing Plusgrow II medium (Nacalai Tesque, Inc., Kyoto, Japan) supplemented with 100 µg mL-1 ampicillin and 20 v/v% glycerol and their plates were stored at -80 °C. A clone carrying the concanamycin and ent-gephyromycin/JBIR-157 BGCs was identified by screening via PCR amplification using two pairs of primers, as follows. For the concanamycin BGC, the primer pairs con_1F (5′- AGACGCTGTACTTCCCCGACCTCA -3′) + con_1 R (5′- GTTCTGATCGGCCTGCTCCTTCTG -3′) and con_2F (5′- TCTTCTGGATGTGGCCGTTCTTCA -3′) + con_2R (5′- CAGCCGGACTTCACCATCTCCTTC -3′) were used to amplify the upstream and downstream regions (respectively) of that cluster; for the ent-gephyromycin/JBIR-157, the primer pairs gpn_1F (5′- GTACCGAGTGCCGCAGATACTCCT -3′) + gpn-1R (5′- GAATCGTTACCCCACAATGGTGTC -3′) and gpn_2F (5′- CATCGAAGAAGATGTGTCGGAAGA -3′) + gpn_2R (5′- GACGCTTTCGAGAGATGTGTTGAG -3′) were used to amplify the upstream and downstream regions (respectively) of that cluster. The identities and spans of the clones that were positive for the respective BGCs were confirmed by end-sequencing and aligning with the genome sequence obtained as described above. The inserted sequence of pKU503ccn (which includes both the concanamycin and ent-gephyromycin/JBIR-157 BGCs) has been deposited in the DDBJ under Accession Number LC780117.

Introduction of BACs into S. avermitilis SUKA strains

First, each clone was introduced into S. lividans TK24 ΔattBϕK38-1::aadA ΔattBϕBT1 ΔattBϕC31 ΔattBTG1 harboring the SAP1 vector, a linear plasmid containing a synthetic sequence corresponding to the lysogenic bacteriophage attachment sites attBϕK38-1, attBR4, attBϕBT1, attBϕC31, and attBTG1 of S. avermitilis MA-4680 [21]. After the introduction of each BAC clone, the clone was integrated into the attBϕC31 site of the SAP1 vector by site-specific recombination (but not into the chromosome because it lacks attB sites). The SAP1 vector carrying BAC clone in S. lividans was transferred to S. avermitilis SUKA SAP1− strains by simple conjugation as described previously [21]. The resulting exoconjugants were selected based on their antibiotic-resistance phenotype; the presence of a linear plasmid of the expected size was confirmed by contour-clamped homogeneous electric field (CHEF) electrophoresis [22].

Liquid chromatography/mass spectrometry (LC/MS) analysis of culture extracts

The metabolites in exoconjugants were extracted from the cultured broth with an equal volume of n-butanol. The organic phase was collected and evaporated to dryness under the reduced pressure. The resulting residue was dissolved in an appropriate volume of dimethyl sulfoxide (DMSO) and analyzed by LC/MS. Specifically, analytical ultra-high performance liquid chromatography (UHPLC) and high-resolution electrospray ionization MS (HR-ESI-MS; positive mode) were performed using an ACQUITY UPLC System (Waters, Framingham, MA) in conjunction with an ethylene-bridged hybrid octadecyl-silica (BEH ODS) column (2.1 i.d. × 100 mm, 55 °C; Waters), an ACQUITY UPLC photodiode array eλ detector (Waters), and a Xevo G2 time-of-flight (TOF) system (Waters). Mobile Phase A was water + 0.1% formic acid; Mobile Phase B was acetonitrile + 0.1% formic acid. The elution program consisted of 5 − 100% B over 5 min, followed by 100% B for 1 min; the flow rate was 0.8 mL min−1. A standard curve for quantification of concanamycin A was prepared by triplicate injection of 2 µL of concanamycin A samples at concentrations of 50, 25, 12.5, 6.25, 3.13, and 1.56 mg L−1.

Nuclear magnetic resonance (NMR) spectroscopy

NMR spectra were recorded on an NMR 600 NB CL spectrometer (Varian, Palo Alto, CA). The coupling constants (J) are given in hertz units. Measurements were carried out at room temperature. Chemical shifts (δ) are reported in ppm with the residual solvent signals as the internal standards (methanol-d4 at δH 7.26 and δC 77.0 ppm). The data are reported as (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, or unresolved; coupling constant(s); integration). 13C NMR spectra were recorded with complete 1H decoupling.

Isolation of Compounds 1 and 2 from the fermentation culture of S. avermitilis SUKA32/pKU503ccn

One liter of a fermentation culture of S. avermitilis SUKA32/pKU503ccn was centrifuged to clear the cell suspension; the resulting supernatant was extracted with ethyl acetate (1 L × 3). The resulting ethyl acetate fraction was concentrated in vacuo, to afford 111.0 mg of crude extract. The crude extract was subjected to silica gel middle pressure-liquid chromatography (SNAP Ultra 25 g; Biotage, Uppsala, Sweden; 1 column volume (CV) = 45 mL); elution was performed using a gradient of n-hexane–ethyl acetate (0% ethyl acetate for 4 CV followed by a linear gradient of 0–25% ethyl acetate over 6 CV; 75 mL min−1 flow) followed by a stepwise gradient of chloroform–methanol (0% methanol for 2 CV followed by a stepwise gradient of 1, 3, 5, 10, 50, and 90% methanol each for 3 CV; 75 mL min−1 flow). The 5% methanol fraction (17.9 mg) was further purified by preparative reverse-phase HPLC using a CAPCELL PAK MG-II C18 column (5.0 μm, 20 i.d. × 150 mm; Shiseido, Tokyo, Japan), and eluted with a 10 mL min−1 flow of 54% aqueous methanol supplemented with 0.1% formic acid, yielding 2.1 mg of 1 (retention time 10.5 min). The 50% methanol fraction (10.5 mg) was further purified by preparative reverse-phase HPLC using a CAPCELL PAK MG-II C18 column (5.0 μm, 20 i.d. × 150 mm; Shiseido), and eluted with 54% aqueous methanol supplemented with 0.1% formic acid, yielding 2.8 mg of 2 (17.5 min).

ent-gephyromycin (1):

[α]24D + 67.3 (c 0.2, MeOH);

NMR: see Supplementary Table 1 and Supplementary Data 1–5;

HR-ESI-MS m/z ([M + Na]+): Calcd. for C19H18O8Na: 397.0899, Found: 397.0906

JBIR-157 (2):

UV λmax (MeOH)nm (ε): 236 (15,890), 270 (14,710), 388 (2,830), 489 (2,210);

NMR: see Supplementary Table 2 and Supplementary Data 6–10;

HR-ESI-MS m/z ([M + Na]+): Calcd. for C19H18O8Na: 397.0899, Found: 397.0902

Bioinformatics

Protein database (nr) was obtained from the National Center for Biotechnology Information (ftp://ftp.ncbi.nlm.nih.gov/blast/db/FASTA/) and similarity searches were conducted using Basic Local Alignment Search Tool (BLAST) binary program. The binary program was made by the compilation of a source package (ftp://ftp.ncbi.nlm.nih.gov/blast//executables/blast+/2.14.0/). The Pfam database version 36.0 (September 2023, 20,795 entries, 659 clans) from the European Bioinformatics Institute (http://ftp.ebi.ac.uk/pub/databases/Pfam/releases/Pfam36.0/) and pfam search binary program was made by the compilation of a source code (http://hmmer.org/). The alignment of sequences was analyzed using Clustal Omega version 1.2.4 (http://www.clustal.org/omega/) [23].

Results and discussion

Isolation of a BAC clone containing the concanamycin BGC

As a first step, we sequenced the genome of S. neyagawaensis IFO 13477 using a PacBio RSII long-read sequencing system and a Roche 454 GS FLX short-read system. The final assembly consisted of three contigs (including a circular plasmid molecule) with a total length of 10.1 Mb. We used the results of this assembly for all sequence analyses in the present study.

Since the BGC size of concanamycin is around 100 kb, we first prepared a BAC library with 150-kb inserts, a size that was expected to encompass the complete BGC. Screening of approximately 1500 colonies by PCR failed to yield a clone containing the entire concanamycin BGC. Therefore, we constructed a second BAC library with inserts of over 180 kb, permitting the PCR screening of an additional 1500 colonies. Among these BAC clones, we obtained two positive clones that harbored identical 211-kb fragments, which were located within the largest contig by end-sequencing the inserts of BACs. One such clone was designated pKU503ccn (Accession No. LC780117). A library of 1500 clones with fragments averaging 150 kbp in size from 10 Mbp Streptomyces genome could theoretically contain at least one fragment of the desired 100 kbp. In practice, however, a BAC library with much larger inserts (>180 kb) was required, and only one pair of identical clones spanning the entire concanamycin BGC was detected. The biosynthetic gene cluster for cephamycin C of S. clavuligerus is approximately 35 kbp, but no entire gene cluster could be obtained from partially digested fragments with restriction enzyme of S. clavuligerus chromosomal DNA using the cosmid vector. However, full-length clones could be obtained using fragments cut by shearing forces and cosmid vector [8]. Thus, because of the recognition sequence bias in partial digestion with restriction enzyme, it may be necessary to use larger fragments for genomic library preparation.

Analysis of secondary metabolites produced by the heterologous expression of pKU503ccn

The pKU503ccn plasmid was introduced successfully into S. avermitilis SUKA32, a “clean” host (lacking production of major endogenous products) that has been used previously for the heterologous production of secondary metabolites of interest [8]. Exoconjugants were cultured in several distinct media, and the resulting metabolite profiles were analyzed by UHPLC/TOF-MS, confirming the production of concanamycin A along with several other minor analogues. These metabolites were identified by comparing the retention time, UV spectra, and HR-ESI-MS profiles to those of S. neyagawaensis IFO 13477 and the authentic sample of concanamycin A (Fig. 2). The production yield of total concanamycins was 24.3 ± 1.6 mg L−1, a value twice that of the parent strain S. neyagawaensis IFO 13477 (12.3 ± 1.0 mg L−1) grown in the same medium. Thus, as previously reported, heterologous expression provides increased, stable production of secondary metabolites [9,10,11,12].

Fig. 2
figure 2

Heterologous production of concanamycins. The culture extract of S. avermitilis SUKA32/pKU503ccn was analyzed by UHPLC/TOF-MS. a UV chromatograms (240 nm) of culture extracts. S. avermitilis SUKA32/pKU503ccn produced larger amounts of concanamycins than did S. neyagawaensis IFO 13477 (indicated by the trace labeled “IFO 13477”). Concanamycin A was used as the standard (indicated by the trace labeled “Std.”). Each culture experiment was carried out in triplicate. Representative data are shown. b The HR-ESI-MS of concanamycin A produced by S. avermitilis SUKA32/pKU503ccn. Calculated m/z for [M+Na]+: 888.5080; observed m/z: 888.5095; MS error +1.7 ppm

Full size image

Interestingly, a chromatogram of n-butanol-extract of S. avermitilis SUKA32/pKU503ccn cultured in Medium 2 showed two sharp, novel (compared to Medium 1) peaks that exhibited UV spectra distinct from those of concanamycins (Fig. 3a–c). Since these peaks (designated Compounds 1 and 2) were not observed in the culture extract of S. neyagawaensis IFO 13477 under any of the tested culture conditions, it was expected that the cryptic gene cluster would be activated by the heterologous expression. We subsequently purified these two compounds using silica gel chromatography followed by reversed-phase chromatography. They were identified to be ent-gephyromycin (1) and a novel compound JBIR-157 (2), respectively. HR-ESI-MS analysis indicated that 1 and 2 had the same molecular formula, C19H18O8 (Fig. 3d–e). Extensive analyses of 1D (1H and 13C) and 2D (DQF-COSY, HSQC, and HMBC) NMR spectra revealed that 1 possessed the same planar skeleton as gephyromycin, an angucycline with a rare ether bridge [24] (Supplementary Table 1, Supplementary Data 1–5). Based on this result and the optical rotation observed with this molecule, we concluded that 1 is ent-gephyromycin [25]. On the other hand, for Compound 2, the UV and visible spectra, together with the analyses of a series of NMR spectra, suggested the existence of a 2,5-dihydroxy-1,4-naphthoquinone moiety (Fig. 1), but the 2D NMR spectra obtained in the present work was insufficient to permit determination of the remaining substructure (Supplementary Table 2, Supplementary Data 6–10). The detailed structural determination of 2 will be reported elsewhere.

Fig. 3
figure 3

Heterologous production of ent-gephyromycin (1) and JBIR-157 (2) by S. avermitilis SUKA32/pKU503ccn. The culture extracts of S. avermitilis SUKA32/pKU503ccn (following growth in two different media) were analyzed using UHPLC/TOF-MS. a UV chromatograms (240 nm) of culture extracts. S. avermitilis SUKA32/pKU503ccn produced 1 and 2 when grown in Medium 2 (indicated by trace “i”) but not when grown in Medium 1 (trace “ii”). S. neyagawaensis IFO 13477 did not produce detectable levels of 1 or 2 even when the strain was cultured in Medium 2 (trace “iii”). b The UV absorption spectrum of 1. c The UV absorption spectrum of 2. d The HR-ESI-MS of 1. Calculated m/z for [M + Na]+, 397.0899; observed m/z, 397.0906; MS error +1.8 ppm. e The HR-ESI-MS of 2. Calculated m/z for [M + Na]+, 397.0899; observed m/z, 397.0902; MS error +0.8 ppm

Full size image

Sequence analysis of the concanamycin BGC

The sequence of the concanamycin BGC determined in this study (ccn cluster) was compared to the previously reported sequence (con cluster; Genbank Accession No. DQ149987 [17]). The set of biosynthetic genes (six genes encoding PKSs and 19 flanking open reading frames (ORFs) and their order and orientation were consistent between the two BGC sequences, although several sequence differences (representing substitutions and insertion/deletions) were observed (Supplementary Fig. 1). All mismatch regions were re-confirmed by Sanger sequencing of pKU503ccn to exclude possible errors in genome sequencing or assembly. The most notable difference was a 168-bp insertion in the dehydratase-encoding region of ccnF in pKU503ccn compared to the previously reported sequence (conF). This inserted sequence corresponds to an unusual 56-amino-acid insertion between β4 and β5 of EryDH4 (PDB Accession No. 3EL6 [26]); to our knowledge, this insertion has not previously been discussed in the literature (Supplementary Fig. 2). Although the specific effect of this additional sequence remains undefined, the dehydratase (DH) domain of CcnF seems to function normally to give an olefin moiety at the C2 position of the growing chain. Notably, the sequence of the concanamycin BGC within the independently deposited genome sequence (NCBI Accession No. GCF_001418645.1) of S. neyagawaensis NRRL B-3092 (which should be identical to ATCC 27449 and IFO 13477) is consistent with that obtained in the present work. On the other hand, the corresponding PKS gene, bfmA5, in the BGC responsible for the biosynthesis of bafilomycin (a natural product that is structurally related to concanamycin) lacks such an insertion [27].

BGC for Compounds 1 and 2

The BGC for ent-gephyromycin (1) has not previously been elucidated. Since both 1 and 2 are aromatic polyketides, these compounds are presumed to be generated via a type II or III PKS system. Gephyromycin (enantiomer of ent-gephyromycin), produced by Gephyromycinifex aptenodytis isolated from gut of Antarctic emperor penguin, was generated from tetraketide and geranyl pyrophosphate by type III PKS (stilbene synthase) and prenyltransferase [28]. But type III PKS system(s) and prenyltransferase were not present in the pKU503ccn and in the genome of heterologous host S. avermitilis. A BLAST search of sequences within the pKU503ccn insert (but outside the concanamycin BGC) revealed that these regions contained two gene clusters consisting of type II PKS system; each of these clusters included genes encoding two 3-oxoacyl-ACP synthase subunits (KSα and KSβ) and an acyl carrier protein (ACP) (Fig. 4a). ORFs in one of these clusters showed strong sequence similarity to ORFs of conserved biosynthetic gene cluster for Streptomyces spore pigment (whiE cluster; Fig. 4b and Supplementary Table 3 [29]). Therefore, we inferred that the other set of type II PKS system-encoding ORFs and their flanking genes (hereafter referred to as the egp cluster) is responsible for the biosynthesis of both 1 and 2 (Fig. 4c). To prove this hypothesis, we re-screened the BAC library to obtain a BAC containing the egp cluster entirely lacking the concanamycin BGC or the whiE-like BGC. One such BAC clone (which we designated pKU503egp; Fig. 4a) was identified and introduced into S. avermitilis SUKA54. As expected, the exoconjugants harboring pKU503egp produced both 1 and 2, confirming the identity of the egp cluster (Supplementary Fig. 3). Simultaneously, these results indicate that the type I PKS cluster (ccn cluster) and the type II PKS BGC (egp cluster) are located next to each other, and each BGC is functional for producing secondary metabolites. Also, it is rare that three BGCs (ccn, egp, and whiE-like cluster) are densely encoded within about 200 kb in actinomycetal genome (Fig. 4a).

Fig. 4
figure 4

Location and organization of biosynthetic gene clusters (BGCs) characterized in the present study. Schematic diagrams of BGC-containing regions. a The chromosomal locations of the inserts of the BACs (pKU503ccn and pKU503egp) that were isolated in the present study and of the BGCs that were characterized. A segment of the genome (bold line) is depicted. Scale bar, 10 kb. b Gene organization of the whiE-like BGCs from IFO 13477 and S. avermitilis. Each pentagon represents a gene, with the pointed end indicating the orientation of the open reading frame (ORF). Vertical lines are used to indicate ORFs that are similar between the two sources (see also Supplementary Table 3). The names of ORFs similar to genes of the whiE cluster of S. coelicolor A3(2) are indicated directly below the S. avermilis sequence. c Gene organization of the egp BGC

Full size image

We further annotated the 26 ORFs in the region of overlap between the pKU503ccn and pKU503egp inserts in an effort to predict the biosynthetic pathway for 1 and 2 (Fig. 5). Three of these ORFs (ccn1, ccn2, and ccn3) were identical to con1, con2, and con3, respectively, that had been proposed to belong to the con cluster, although the functions of the respective gene products remain unknown (Although unpublished data, Con3 positively regulates the expression of con cluster). While the actual boundary between the ccn and egp clusters still is unclear, we tentatively have assigned 21 genes as components of the egp cluster (Table 1). It is noteworthy that each egp gene has at least one similar gene associated with a previously reported type II PKS-encoding gene cluster that has been deposited in the MIBiG (Minimum Information about a Biosynthetic Gene cluster) database [30]. In addition, the egp cluster was identified in the previously reported S. neyagawaensis genome, and also found in the S. cyaneochromogenes genome with the extremely high sequence similarity (Supplementary Table 4).

Fig. 5
figure 5

Proposed biosynthetic pathway for ent-gephyromycin (1) and JBIR-157 (2). Based on the annotation analyses of the egp cluster, 1 and 2 likely are biosynthesized through a shared intermediate, UWM6 (3). Compound 4 is a hypothetical intermediate whose production from 3 would require multiple oxygenation/reduction steps. Definition of the detailed structure of 2 is expected to further improve the accuracy of the proposed biosynthetic pathway

Full size image
Table 1 Deduced functions of ORFs involving ent-gephyromycin in egp cluster
Full size table

Proposed biosynthetic pathway for 1 and 2

EgpA, EgpB, and EgpC show significant similarity to JadA (85% amino acid identity), JadB (78%), and JadC (73%), respectively; JadABC constitutes a minimal PKS [31]. In addition, EgpD, EgpL, and EgpF are close homologues of JadE (84% amino acid identity), JadD (75%), and JadI (81%), respectively. The jadABCDEI gene set is conserved across several type II PKS-encoding clusters whose products are known to be responsible for the biosynthesis of UWM6 (3) [32]; notably, 3 is a common biosynthetic intermediate of well-studied angucyclines such as landomycins [33], urdamycins [34], and jadomycins [35]. Accordingly, we presume that the biosynthesis of 1 and 2 employs 3 as an intermediate (Fig. 5).

The conversion from 3 to 1 requires several oxygenation/reduction steps. The hydroxylation at C-12b of angucyclines has been investigated in the context of the biosynthesis of urdamycin [36] and gaudimycin [37]. In both cases, a NADPH-dependent flavoprotein hydroxylase (UrdE or PgaE, respectively) was assigned as the responsible enzyme. In the egp cluster, EgpO4 has 78.1% amino acid identity to PgaE. Additionally, the C-terminal domain of EgpO5 has 78.5% amino acid identity to the respective region of PgaM, a short-chain dehydrogenase/reductase domain that has been proposed to catalyze the formation of the C-6 hydroxy group [36]. An ethylene moiety is present on the B-ring of several angucyclines, including urdamycins I and J [38], gaudimycin A [39], and SF2315A and B [40]. However, the responsible enzyme(s) remain unidentified. The candidate participants encoded by the egp cluster include EgpO1 and EgpO2. EgpO1 has a 48.0% amino acid identity to ActVI-orf2, an enoyl reductase involved in actinorhodin biosynthesis [41]. On the other hand, domain analysis indicates that EgpO2 is a didomain protein. Specifically, the N-terminal domain (EgpO2-N) has a Rossmann-fold motif, while the C-terminal domain (EgpO2-C) resembles a methyltransferase. In the MIBiG database, the protein with the strongest similarity to EgpO2-N is BexH (64.0% amino acid identity), a putative sugar epimerase in the BE-7585A biosynthetic pathway [42]. However, EgpO2-N also shows some similarity (39.4% amino acid identity) to SimC7, a ketoreductase involved in simocyclinone biosynthesis [43]. The oxygenation at the C-ring to form a quinone is presumably performed by EgpO3, a protein with intermediate similarity (34.7% amino acid identity) to ActVA-orf6. ActVA-orf6 is a cofactor-independent monooxygenase that oxygenates the phenolic intermediate, 6-deoxydihydrocarafungin, into the quinone, dihydrokalafungin, during actinorhodin biosynthesis [44]. Based on these speculations, we tentatively hypothesize that 3 is converted to 1 via an intermediate consisting of 4 (Fig. 5), an enantiomer of one of two predicted precursors for gephyromycin.

Among the remaining genes, egpR1-R2 and egpT1-T2 may encode transcriptional regulators and transporters, respectively. Additionally, two transcriptional regulators, ORF−1 and ORF−2, are tandemly encoded in opposite direction to the putative egpJ-I operon. Each of them shows 52% and 47% amino acid identity, respectively, to KiqA, the positive regulator for the kinanthraquinone BGC [45]. The particular functions of ORF−1 and ORF−2 are still elusive. EgpK and EgpE show ~90% amino acid identity to the α and β subunits (respectively) of the S. avermitilis acetyl-CoA carboxylase. Although the genes within egp cluster do not appear to encode an acetyl-CoA carboxylase ε subunit, the products of the egpK and egpE genes may enhance polyketide production by supplying malonyl-CoA. EgpO2-C has 49.2% amino acid identity to BexP, a methyltransferase predicted to be involved in self-resistance to the angucycline BE-7585A [42]. However, the function of EgpO2-C remains unclear, given that 1 and 2 did not show antibacterial activity (data not shown). EgpJ (which includes a DsbA-like thioredoxin domain), EgpG (a putative NADPH-dependent FMN reductase), and EgpI (a putative member of the α/β-hydrolase family) also have strong similarities to proteins encoded by type II PKS BGCs. These proteins may be responsible for protein folding, flavin mononucleotide (FMN) regeneration, and the off-loading step of polyketide synthesis, respectively, although these suggestions will need to be confirmed experimentally. Lastly, EgpH constitutes a 66-amino-acid protein without any apparent motifs; the role of this gene product in the biosynthesis of 1 and 2 remains unknown.

Concluding remarks

This work represents (to the best of our knowledge) the first report of heterologous production of concanamycins. The heterologous expression system with a suitable production titer is expected to facilitate the generation of novel concanamycins via engineering the BGC. We additionally note that our sequence analysis, performed using a long-read sequencer, refined the BGC at the nucleotide level. The data presented here should serve as a superior reference for studies requiring precise sequences, such as those involving PKS engineering.

The capacity of the BAC cloning technique to capture a large insert, along with the use of a host strain that lacks intrinsic relevant secondary metabolites, enabled us to identify not only the target compound (concanamycin) but also two type II PKS products. In combination with the results of our heterologous expression system, bioinformatic analyses provided a milestone for our understanding of the biosynthesis of 1, a unique bridged angucycline. Further biosynthetic analyses are in progress.