Characterization of the biosynthetic gene cluster for cryptic phthoxazolin A in Streptomyces avermitilis

Abstract

Phthoxazolin A, an oxazole-containing polyketide, has a broad spectrum of anti-oomycete activity and herbicidal activity. We recently identified phthoxazolin A as a cryptic metabolite of Streptomyces avermitilis that produces the important anthelmintic agent avermectin. Even though genome data of S. avermitilis is publicly available, no plausible biosynthetic gene cluster for phthoxazolin A is apparent in the sequence data. Here, we identified and characterized the phthoxazolin A (ptx) biosynthetic gene cluster through genome sequencing, comparative genomic analysis, and gene disruption. Sequence analysis uncovered that the putative ptx biosynthetic genes are laid on an extra genomic region that is not found in the public database, and 8 open reading frames in the extra genomic region could be assigned roles in the biosynthesis of the oxazole ring, triene polyketide and carbamoyl moieties. Disruption of the ptxA gene encoding a discrete acyltransferase resulted in a complete loss of phthoxazolin A production, confirming that the trans-AT type I PKS system is responsible for the phthoxazolin A biosynthesis. Based on the predicted functional domains in the ptx assembly line, we propose the biosynthetic pathway of phthoxazolin A.

Introduction

Polyketides and non-ribosomal peptides are structurally diverse classes of natural products representing important therapeutic and agricultural chemicals [1,2]. In bacteria, polyketides are typically produced by polyketide synthase (PKS) assembly lines that are composed of functional modules. Each module is responsible for one step of chain extension of the growing products and consists of a set of domains which dictate the chemical functionality of the incorporated building block [3,4]. The minimal domains of a PKS module are an acyltransferase (AT), ketosynthase (KS) and acyl carrier protein (ACP) domain, where the AT domain loads a coenzyme A-activated dicarboxylic acid extender unit onto the ACP domain and the KS domain of the preceding domain catalyzes condensation of the nascent polyketide with the extender unit. Meanwhile, nonribosomal peptide synthetase (NRPS) also forms multienzyme assembly lines that are similar to those of PKS, and produces peptide-containing compounds [5]. In the NRPS module, an adenylation (A) domain activates an amino acid, and loads onto a peptidyl carrier protein (PCP). Subsequently, a condensation (C) domain generates an amide bond with the growing chain tethered to the upstream PCP domain. Many of the bacteria sequenced to date have been found to possess a “hybrid” PKS/NRPS enzyme [6] and show that both PKS and NRPS function simultaneously in the same assembly line [7], but the products synthesized by most of the hybrid PKS/NRPS enzymes are still unknown.

Oxazole-containing polyketides are typical PKS/NRPS hybrid products, and have various biological activities for medicinal and agricultural purposes; e.g., oxazolomycin is an oxazole-triene antibiotic of Streptomyces albus JA3453 showing antibacterial and antitumor activities [8, 9], rhizoxin is a methyl-oxazole macrolide from the bacterial endosymbiont Burkholderia rhizoxina [10], and conglobatin is an unusual symmetrical macrolide from Streptomyces conglobatus showing antitumor activity [11]. Within these oxazole-containing polyketides, oxazolomycin and rhizoxin are synthesized by the trans-AT type I PKS system together with NRPS machinery [10,12]. One distinct feature of the trans-AT type I PKS system is that the PKS modules lack an AT domain and one or a few free-standing ATs provide extender units for each elongation step [13,14]. This type of PKS modular system has contributed to the expansion of complexity in the polyketide biosynthetic machinery.

Phthoxazolin A (Fig 1A), one of the oxazole-containing polyketides, is an inhibitor of cellulose biosynthesis and exerts growth-inhibitory activity against plant pathogenic oomycetes [15–17]. The structure of phthoxazolin A includes a unique 5-substituted oxazole ring connected to a triene moiety, which corresponds to the substructure of oxazolomycin. We have previously demonstrated that phthoxazolin A is a cryptic metabolite of Streptomyces avermitilis, the producer of the important anthelmintic agent avermectin [18]. Although the in silico analysis predicted that the genome of S. avermitilis harbors at least 38 gene clusters for secondary metabolites [19], there is no assembly line similar to that for oxazolomycin biosynthesis. Thus, the biosynthetic gene cluster for phthoxazolin A would be hidden somewhere in the S. avermitilis genome.

Fig 1. Phthoxazolin A production in the S. avermitilis progeny.

(A) Chemical structure of phthoxazolin A. (B) HPLC chromatograms of MeOH extracts from S. avermitilis KA-320 (top), S. avermitilis SUKA22 (middle), and S. avermitilis K139 (bottom). mAU, milliabsorbance units at 275 nm. Phthoxazolin A was detected at a retention time of 33.9 min, and is indicated by an inverted triangle.

In this study, we performed genome sequencing, comparative genomic analysis, and mutagenesis to identify a biosynthetic gene cluster for phthoxazolin A, and demonstrated that the gene cluster is laid on the extra genomic region of the original avermectin producer, and phthoxazolin A is synthesized by the trans-AT type I PKS system. The proposed assembly line for phthoxazolin A biosynthesis also suggests a new cleavage system in the PKS/NRPS machinery.

Materials and methods

Bacterial strains, plasmids, and growth conditions

S. avermitilis KA-320 (isogenic to MA-4680, ATCC 31267 and NRRL 8165) [20], K139 (a progeny of KA-320) [21], SUKA22 (K139 as a genetic background) [22] strains were obtained from the culture collection of Kitasato Institute, and S. avermitilis KA-320 ΔavaR3 [20] was used in this study. All these strains were grown on YMS-MC medium for spore formation [20]. Escherichia coli DH5α was used for general DNA manipulation, and E. coli F^-dcm Δ(srl-recA) 306::Tn10 carrying pUB307-aph::Tn7 was used for E. coli/Streptomyces conjugation. The plasmids pKU451, pKU470, pKU479, pKU480, and pKU250 were used to construct a vector for gene deletion [23]. The media and general E. coli and Streptomyces manipulations were as described previously [20]. For analysis of phthoxazolin A production, spores (1.0 X 10⁸ CFU) of the S. avermitilis strains were inoculated into 70 mL APM medium in a 500-mL baffled flask, and mycelia were harvested after 48 h of cultivation. The mycelia were washed, re-suspended in fresh APM medium and stored at -80°C until use as a seed culture. All the primers are listed in S1 Table.

Analysis of phthoxazolin A production

The seed culture was inoculated on 2.5 mL YMD solid medium [18], followed by incubation at 28°C for 3 days. The agar culture was diced and extracted with an equal volume of methanol. The methanol extract was analyzed by using a HPLC system as described previously. [18].

Construction of Streptomyces avermitilis large-deletion (SALD) mutant strains

For the SALD-1 mutant, two regions (position 77,219–79,457 nt and 79,479–81,744 nt) were PCR-amplified by the primer pairs (sav68-up-Fw/sav68-up-Re and sav71-dw-Fw/sav71-dw-Re). These fragments were digested by HindIII and SpeI, and inserted to pKU451 resulting in pLT143. This plasmid was digested with SpeI and ligated together with a fragment containing a hygromycin B phosphotransferase gene (hph) with mutant loxP sequence (mut-loxP) (PCR-amplified with pKU480 and the mutloxP-SpeI-Fw/wo-mutloxP-SpeI-Re primer pair) to generate pLT144. A 6.6 kb HindIII fragment, recovered from pLT144, was cloned into pKU250 at the HindIII site to obtain pLT145. pLT145 was introduced by intergeneric conjugation to S. avermitilis KA-320 ΔavaR3 mutant to yield S. avermitilis ΔavaR3/sav71::mut-loxP-hph mutant. Another two regions (582,859–584,876 nt and 587,127–584,906 nt) were also PCR-amplified by the primer pairs (sav432-up-Fw/sav434-up-Re and sav434-dw-Fw/sav434-dw-Re). These fragments were treated with HindIII/SpeI, and ligated to pKU451 to get pLT146. A kanamycin-resistant gene (aphII) with mut-loxP at downstream end was PCR-amplified by the primer pair wo-mutloxP-SpeI-Fw/ mutloxP-SpeI-Re using pKU479 as a template, and introduced into the SpeI site of pLT146 to generate pLT147. A 6.0 kb HindIII fragment, recovered from the resultant plasmid, cloned into at the HindIII site of pKU250 to obtain pLT148. pLT148 was introduced to S. avermitilis ΔavaR3/sav71::mut-loxP-hph mutant, resulting in S. avermitilis ΔavaR3/sav71::mut-loxP-hph mutant/Δsav434::aphII-mut-loxP. The cre expression plasmid pKU470 was introduced into the strain for removal of a 0.51 Mb region covering from sav71 to sav434.

For the SALD-2 mutant, the aphII gene of pLT148 was replaced with a hph gene harboring mut-loxP by SpeI digestion and ligation, resulting in pLT149. pLT149 was introduced into S. avermitilis KA-320 ΔavaR3 mutant to yield S. avermitilis ΔavaR3/sav434::mut-loxP-hph mutant. Two regions (999,468–997,573 nt and 1,004,009–1,002,110 nt) were PCR-amplified by the primer pairs (sav845-up-Fw/sav845-up-Re and sav845-dw-Fw/sav845-dw-Re), and ligated with pKU451 to get pLT150. This plasmid was ligated with the aphII gene prepared previously to generate pLT151. A 5.4 kb HindIII fragment, recovered from pLT151, was cloned into pKU250 at the HindIII site to obtain pLT152. pLT152 was introduced to S. avermitilis ΔavaR3/sav434::mut-loxP-hph mutant, resulting in S. avermitilis ΔavaR3/sav434::mut-loxP-hph mutant/Δsav845::aphII-mut-loxP. pKU470 was introduced into the strain for removal of a 0.42 Mb region covering from sav434 to sav845.

For the SALD-3 mutant, the aphII gene of pLT152 was replaced with a hph gene harboring mut-loxP by SpeI digestion and ligation, resulting in pLT153. pLT153 was introduced into S. avermitilis KA-320 ΔavaR3 mutant to yield S. avermitilis ΔavaR3/sav845::mut-loxP-hph mutant. Two regions (1,271,761–1,273,916 nt and 1,273,956–1,276,187 nt) were PCR-amplified by the primer pairs (sav1007-up-Fw/sav1007-up-Re and sav1007-dw-Fw/sav1007-dw-Re), and ligated with pKU451 to get pLT154. This plasmid was ligated with the aphII gene prepared previously to generate pLT155. A 6.0 kb HindIII fragment, recovered from pLT155, was cloned into pKU250 at the HindIII site to obtain pLT156. pLT156 was introduced to S. avermitilis ΔavaR3/sav845::mut-loxP-hph mutant, resulting in S. avermitilis ΔavaR3/sav845::mut-loxP-hph mutant/Δsav1007::aphII-mut-loxP. pKU470 was introduced into the strain for removal of a 0.28 Mb region covering from sav845 to sav1007.

For the SALD-4 mutant, the aphII gene of pLT156 was replaced with a hph gene harboring mut-loxP by SpeI digestion and ligation, resulting in pLT157. pLT157 was introduced into S. avermitilis KA-320 ΔavaR3 mutant to yield S. avermitilis ΔavaR3/sav1007::mut-loxP-hph mutant. Two regions (1,593,533–1,595,311 nt and 1,595,580–1,597,661 nt) were PCR-amplified by the primer pairs (sav1286-up-Fw/sav1286-up-Re and sav1286-dw-Fw/sav1286-dw-Re), and ligated with pKU451 to get pLT158. This plasmid was ligated with the aphII gene prepared previously to generate pLT159. A 5.5 kb HindIII fragment, recovered from pLT159, was cloned into pKU250 at the HindIII site to obtain pLT160. pLT160 was introduced to S. avermitilis ΔavaR3/sav1007::mut-loxP-hph mutant, resulting in S. avermitilis ΔavaR3/sav1007::mut-loxP-hph mutant/Δsav1286::aphII-mut-loxP. pKU470 was introduced into the strain for removal of a 0.32 Mb region covering from sav1007 to sav1286.

The genotype of candidate strains for the desired large deletion mutation was confirmed by PCR analysis and DNA sequencing. The S. avermitilis KA-320 ΔavaR3 lacking each region (sav71-sav434, sav434-sav845, sav845-sav1007, and sav1007-sav1286) was designated as SALD-1, SALD-2, SALD-3, and SALD-4.

Genome sequencing and bioinformatics analyses

The genomic DNA of S. avermitilis KA-320 was prepared according to the previous procedures [24]. Sequence data of S. avermitilis KA-320 was obtained by assembling both data generated from the old-type DNA sequencer (MegaBACE 1000) and the next-generation DNA sequencer (Illumina). After assembling them, at least 21 contigs were generated. One contig (ca. 164 kb) contained the putative biosynthetic gene cluster for phthoxazolin A (a region around the gene cluster has been deposited in the DDBJ; accession number LC315614). Annotation of open reading frames (ORFs) and gene functions was performed manually by using the FramePlot 4.0beta program (http://nocardia.nih.go.jp/fp4), the 2ndFind program (http://biosyn.nih.go.jp/2ndfind/), the BLAST algorithm and the web-based PKS/NRPS analysis program (http://nrps.igs.umaryland.edu/nrps/).

In-frame deletion of the ptxA gene in the avaR3 mutant

Two 2.0 kb flanking regions of the ptxA gene were PCR-amplified by the primer pairs (ptxA-up-Fw/ptxA-up-Re and ptxA-dw-Re/ptxA-dw-Fw). These fragments were digested by HindIII and SpeI, and inserted into pKU451, resulting in pLT140. This plasmid was digested with SpeI and ligated together with a fragment containing a hph gene (PCR-amplified with pKU480 and the hph-Fw/hph-Re primer pair) to generate pLT141. A 6.2 kb HindIII fragment was recovered from pLT141, and cloned into the HindIII site of pKU250 to obtain pLT142. pLT142 was introduced into the S. avermitilis KA-320 ΔavaR3 mutant by intergeneric conjugation [20], and the DNA region including the ptxA gene was replaced with the disrupted allele by homologous recombination. The genotype of candidate strains for the ptxA mutation was confirmed by PCR analysis and DNA sequencing. The S. avermitilis KA-320 avaR3/ptxA double mutant was abbreviated as ΔavaR3 ΔptxA.

Complementation of the avaR3/ptxA double mutant

A 3.2 kb fragment containing the entire ptxA gene was PCR-amplified by using the primer pair ptxA-comp-Fw/ptxA-comp-Re, and then cloned to the BamHI site of pLT101 [25] using GeneArt Seamless Cloning and Assembly Kit (Life Technologies). The resultant plasmid was introduced into the avaR3/ptxA double mutant by intergeneric conjugation and integration. Integration of the plasmid was confirmed by apramycin resistance and PCR analysis.

Results

Phthoxazolin A production in the S. avermitilis progeny

To identify a biosynthetic gene cluster for phthoxazolin A, in silico screening was performed with the public genome sequence of S. avermitilis using oxazolomycin biosynthetic genes as probes. However, we could not find any assembly line in the genome that was similar to that of the oxazolomycin biosynthesis. Moreover, the genome has no orthologue of OzmO (an NRPS enzyme), which is necessary for the biosynthesis of an oxazole ring moiety. It thus appeared that the public genome sequence contained no clues for identifying the phthoxazolin A biosynthetic gene cluster.

S. avermitilis K139 SUKA22 is genetically constructed by deleting the 1.5 Mb left region of the parental strain [22], and was found to lack the ability to produce phthoxazolin A (Fig 1B), suggesting that the deleted DNA region might contain the phthoxazolin A biosynthetic gene cluster. Thus, we generated several series of large-deletion mutants (SALD mutants) from the S. avermitilis KA320 ΔavaR3 mutant [20], and evaluated phthoxazolin A production in these SALD mutants (S1 Fig). To our surprise, all the large-deletion mutants (SALD-1, SALD-2, SALD-3 and SALD-4 strains) still produced phthoxazolin A at a level comparable to the parental ΔavaR3 mutant strain.

S. avermitilis K139 is one progeny of our wild-type strain (S. avermitilis KA-320, which is isogenic to MA-4680, ATCC 31267 and NRRL 8165), and K139 has been used in the genome sequencing project [18]. The strain K139 also did not produce phthoxazolin A (Fig 1B), suggesting that genetic differences may exist between S. avermitilis KA-320 and K139 to confer ability for the phthoxazolin A biosynthesis.

Identification of the phthoxazolin A biosynthetic gene cluster

To reveal genetic differences between strains KA-320 and K139, we performed a contour-clamped homogeneous electrical field (CHEF) electrophoresis analysis of these chromosomes, and found that the genome size of strain KA-320 was about 800–1,000 kb larger than that of strain K139 and that the extra region of strain KA-320 was located at the right-hand region of the K139 genome. Based on these results, there was a strong possibility that the extra region of strain KA-320 encodes a biosynthetic gene cluster for phthoxazolin A. Thus, we sequenced the genome of strain KA-320 to find the extra region. By using the same approach as described above, we performed in silico screening with the genome sequencing of S. avermitilis KA-320, and found a possible phthoxazolin A biosynthetic gene cluster that spanned 99.9 kb (Fig 2). Annotation analysis of the sequence and comparison with genes in the public databases revealed 34 ORFs. The genetic organization and proposed functions are shown in Fig 2 and Table 1, respectively.

Fig 2. Genetic organization of the phthoxazolin A biosynthetic gene cluster.

Arrows indicate the direction of transcription and relative gene size. ORFs predicted to participate in phthoxazolin A biosynthesis are shaded. The proposed functions of individual ORFs are indicated here and summarized in Table 1.

Table 1. Deduced functions of ORFs in the phthoxazolin A biosynthetic gene cluster.

Gene	Sizea	Homologb and origin	Identity/ similarity (%)	Proposed function
orf 5	504	GlpK (WP_015654981), Streptomyces davawensis JCM 4913	91/95	Glycerol kinase
orf 6	255	ASC56_RS09905 (WP_055490882), Streptomyces sp.TP-A0356	92/96	IclR-family transcriptional regulator
orf 7	142	ADL25_RS11400 (WP_059127556), Streptomyces sp. NRRL F-5122	67/76	Histidine kinase
orf 8	80	IQ62_RS20385 (WP_037697207), Streptomyces scabiei NCPPB 4086	85/91	Hypothetical protein
orf 9	333	Ppk2 (WP_007385561), Streptomyces sviceus ATCC 29083	88/93	Polyphosphate kinase
orf 10	180	SAMN05216482_0059 (SEB58800), Streptomyces sp. PAN_FS17	78/87	Hypothetical protein
orf 11	481	G412_RS0110405 (WP_02881204), Streptomyces flavidovirens DSM 40150	96/98	Glyceraldehyde 3-phosphate dehydrogenase
orf 12	137	NF37_RS0107960 (WP_032755078), Streptomyces alboviridis NRRL B-1579	81/91	Integrase
orf 13	501	AWV61_RS50755(WP_060880896), Streptomyces scabiei 95–18	83/85	Transposase
orf 14	188	AVL59_RS26005 (WP_067308799), Sreptomyces griseochromogenes ATCC 14511	81/86	Two-component system sensor kinase
orf 15	113	CCN44_RS40620 (WP_086704188), Streptomyces tricolor NRRL B-16925	86/86	Two-component system response regulator
orf16	94	IG08_RS0113085 (WP_030600335), Streptomyces fulvoviolaceus NRRL B-2870	87/88	LuxR-family transcriptional regulator
orf 17	169	BIV24_RS13170 (WP_071366454), Streptomyces sp. MUSC 93	74/82	Polyketide cyclase
ptxA	1065	OzmM (ABS90474), Streptomyces albus JA3453	57/68	Acetyl transferase
ptxB	5939	OzmH (ABS90470), S. albus JA3453	53/62	Hybrid NRPS-PKS
ptxC	877	OzmQ (ABS90478), S. albus JA3453	67/76	Type I PKS
ptxD	362	OzmP (ABS90477), S. albus JA3453	77/88	Hypothetical protein
ptxE	1154	OzmO (ABS90476), S. albus JA3453	53/62	NRPS
ptxF	4885	OzmN (ABS90475), S. albus JA3453	51/60	Type I PKS
ptxG	3542	NRPS (OMI35273), Streptomyces sparsogenes DSM 40356	65/74	NRPS
ptxH	2860	PKS 1–1 (ADI03434), Streptomyces bingchenggensis BCW-1	63/72	Type I PKS
orf 26	155	SibV (ACN39745), Streptosporangium sibiricum DSM 44039	66/77	Dioxygenase
orf 27	306	BZL62_RS04865 (WP_086716558), Streptomyces angustmyceticus NRRL B-2347	68/78	Hypothetical protein
orf 28	347	AOK13_RS10670 (WP_055559528), Streptomyces luridiscabiei NRRL B-24455	81/87	IMP dehydrogenase
orf 29	517	BZL62_RS04825 (WP_086716551), Streptomyces angustmyceticus NRRL B-2347	79/86	Acetolactate synthase
orf 30	295	SAMN05444920_109123 (SEG95653), Nonomuraea solani CGMCC 4.7037	55/70	Hydroxyacid dehydrogenase
orf 31	410	BR98_RS37570 (WP_083976095), Kitasatospora azatica KCTC 9699	70/82	Cytochrome P450
orf 32	71	WT80_RS35315 (WP_081087741) Burkholderia stagnalis MSMB774WGS	43/56	unknown
orf 33	526	KCH_RS21250 (WP_084223811), Kitasatospora cheerisanensis KCTC 2395	66/77	Fatty acid Co A ligase
orf 34	409	SAMN05216533_5065 (SEF00159), Streptomyces sp. Ag109_O5-10	85/92	5-Aminolevulinate synthase
orf 35	509	Ann2 (AGY30678), Streptomyces calvus ATCC 13382	67/81	5-Aminolevulinate synthase
orf 36	225	ColR1 (AIL50186), Streptomyces aureus SOK1/5-04	50/62	LuxR-family transcriptional regulator
orf 37	248	IF01_RS0119920 (WP_051755722), Streptomyces purpeofuscus NRRL B-1817	75/83	TetR-family transcriptional regulator
orf 38	516	OO66_RS31780 (WP_051763676), Streptomyces virginiae NRRL B-8091	74/83	Multidrug MFS (major facilitator superfamily) transporter

^a Numbers refer to amino acid residues.

^b Parenthetical codes are National Center for Biotechnology Information accession numbers.

Three PKS genes (ptxC, ptxF, and ptxH), one hybrid NRPS/PKS gene (ptxB), and two NRPS genes (ptxE and ptxG) were identified in the cluster, together with ptxA, which encoded a putative discrete AT enzyme. To investigate the involvement of PtxA in the biosynthesis of phthoxazolin A, we disrupted the ptxA gene by deleting 1,032 amino acids in the ΔavaR3 genetic background. The double mutant (ΔavaR3 ΔptxA) strain was unable to produce phthoxazolin A (Fig 3) and the ptxA-complemented avaR3/ptxA double mutant (ΔavaR3 ΔptxA/ptxA) produced phthoxazolin A to the level in the parental strain (ΔavaR3) (Fig 3), indicating that the discrete AT (PtxA) is essential for the phthoxazolin A biosynthesis. This result also suggested that the PKS genes and the hybrid NRPS/PKS gene, which require enzymatic activity of PtxA, are involved in the biosynthesis of phthoxazolin A (see next section).

Fig 3. Phthoxazolin A production in the avaR3/ptxA double mutant.

(A) Schematic representation of the strategy for the ptxA gene disruption. ΔavaR3, avaR3 mutant; ΔavaR3 ΔptxA, avaR3/ptxA double mutant; ΔavaR3 ΔptxA/ptxA, ptxA-complemented avaR3/ptxA double mutant. (B) PCR analysis to confirm gene-disruption of the ptxA gene and its complementation. With the primer pair ptzA-tFw/ptzA-tRe, a fragment (4,193 bp) containing an intact ptxA gene or a fragment (3,067 bp) containing the mut-loxP-hph-mut-loxP was amplified with PCR. An intact ptxA gene (485 bp) was detected by using the primer pair ptzA-Fw/ptzA-Re. An internal region of aac(3)IV gene (974 bp) was amplified using the primer pair apr-Fw/apr-Re. (C) HPLC chromatograms of MeOH extracts from the avaR3/ptxA double mutant. mAU, milliabsorbance units at 275 nm. Phthoxazolin A is indicated by an inverted triangle.

A gene encoding a discrete AT

The PtxA protein contains two tandem AT domains together with an oxidoreductase domain, and should function as a discrete AT enzyme in the trans-AT type I PKS gene cluster. In the ptx gene cluster, three PKS genes (ptxC, ptxF, and ptxH) and a hybrid NRPS/PKS gene (ptxB) are embedded to encode 10 PKS modules (modules 1 to 5 and 7 to 11). The PKS modules encoded by ptxB, ptxC, ptxF, and ptxH all lack cognate AT domains, implying that the PtxA protein provides the missing AT activity by acting in trans for these PKS modules. The oxazolomycin pathway has also been established to have the trans-AT type I PKS system with a discrete AT enzyme (OzmM containing the tandem AT domains) [12]. Phylogenetic tree analysis of the AT domains from PtxA (PtxA-AT1 and PtxA-AT2) with other tandem-type discrete AT enzymes revealed that PtxA-AT1 and Ozm-AT1 are positioned in a different clade from PtxA-AT2, Ozm-AT2 and KirC1-AT2 (S2 Fig). Zhao et al. [12] established that OzmM-AT2 is involved in the oxazolomycin biosynthesis, whereas OzmM-AT1 is dispensable. Moreover, KirC1-AT2 loads malonyl-CoA extender units to the ACPs in the kirromycin biosynthesis [26]. With the observation that PtxA-AT1 seems inactive due to the replacement of important amino acid residues [27] (substitution of Glu⁶³ and His⁹¹ with His and Ser, respectively), much as in the cases of OzmM-AT1 and KirC1-AT1, it can be concluded that PtxA-AT2 supplies malonyl-CoA units to the Ptx PKS modules.

Genes involved in the polyketide assembly

Ten PKS modules were identified in the ptx gene cluster, encoded by ptxB, ptxC, ptxF, and ptxH. PtxC and PtxF closely resembled OzmQ (67% identity) and OzmN (51% identity), respectively. Both OzmQ and OzmN have PKS modules 2, 3, 4, and 5 in the oxazolomycin assembly line, and render a triene moiety in the structure of oxazolomycin [12]. PKS modules 4 and 5 of PtxB contain two methyltransferase (MT) domains for the addition of methyl groups at C-2 and C-4 of phthoxazolin A.

With respect to the KS domains in the ptx cluster, analysis of the conserved active site (catalytic triad of Cys-His-His) [28] revealed that two KS domains (KS9 and KS11) contain a mutation in the conserved motif (S3 Fig), suggesting that they should be inactive in the polyketide assembly line; these domains were called KS⁰ domains [14]. Regarding the ACP domains, the Ser residue in the signature motif plays a crucial role as the 4’-phosphopantetheine attachment site [29], but the first ACP domain (ACP9a) of module 9 lacks the Ser residue, while other ACP domains in the Ptx PKS modules retain the conserved Ser residue (S4 Fig), indicating that the PKS module 9 employs two types of ACPs (inactive-type and active-type). Along with the ketoreductase (KR) domains, seven KR domains are predicted in the PKS modules, all of which show a characteristic Rossmann fold for NADP(H)-binding [30] (S5 Fig). In addition, all the KR domains except KR4 and KR5 contain a typical catalytic triad (Ser-Tyr-Asn) [30], while KR4 and KR5 have different sequences (Ser-Tyr-Cys for KR4 and Ser-Tyr-Ser for KR5). Because these minor modifications in the catalytic triad have also been identified in other trans-AT type I PKSs, such as those in OocL and OocR in the oocydin A biosynthesis [31], all the ptx KR domains seem to be active in the PKS assembly line. Regarding the DH domains, all six DH domains possess the conserved His residue in the signature motif HXXXGXXXXP [32] (S6 Fig), indicating that the dehydratase (DH) domains are active, and are likely responsible for the formation of a double bond. Taking these results together, the first 5 PKS modules (modules 1 to 5) appear to be responsible for synthesizing the polyketide moiety of phthoxazolin A, whereas the remaining 5 PKS modules (modules 7, 8, 9, 10, and 11) might be capable of synthesizing an additional triene structure (as discussed later).

Genes involved in the formation of an oxazole ring

The ptxE gene encodes a single module of NRPS comprising a formylation (F) domain, an A domain, and a PCP domain, which probably serve as loading modules to incorporate an initial amino acid into the assembly line. Because the PtxE-A domain is predicted to be a glycine-specific A domain based on bioinformatic analysis using the AntiSMASH database [33], glycine can be considered to be activated by the PtxE-A domain and loaded onto the PtxE-P domain as a glycyl-S-PCP. The presence of an F domain in the loading module has sometimes been identified in other biosynthetic machineries, such as those for the biosynthesis of gramicidin [34], rhizopodin [35], oxazolomycin [12], and conglobatin [11], and the PtxE-F domain can be predicted to catalyze formylation of glycyl-S-PCP to generate formyl-glycyl-S-PCP, from the analogy of the F domain of LgrA1 in the gramicidin biosynthesis.

The downstream region of the ptxE gene includes the ptxD gene, which encodes a protein homologous to OzmP (in oxazolomycin biosynthesis) and CongE (in conglobatin biosynthesis). The gene arrangement (ptxD-ptxE) is identical to those in the biosynthetic gene clusters of oxazolomycin (ozmO-ozmP) and conglobatin (congA-congE). Moreover, these three proteins (PtxD, CongE, and OzmP) have an ATP-pyrophosphatase domain that includes the signature motif SGGKDS in the N-termini for ATP binding [36]. Because CongE has been proposed to activate the amide oxygen by adenyltransfer from ATP with the release of pyrophosphate for formation of the oxazole ring moiety [11], PtxD is likely to be responsible for conversion of the formyl-glycyl intermediate into the oxazole ring by cyclodehydration.

Genes putatively involved in the nonribosomal peptide assembly

The assembly line in the ptx gene cluster possesses four NRPS modules, including a loading module encoded by ptxE. NRPS module 6 is a part of the PKS/NRPS hybrid protein PtxB, and NRPS modules 12 and 13 are encompassed in the PtxG protein. The specificity-conferring codes of the A domains indicated that amino acids activated by PtxB-A6, PtxG-A12 and PtxG-A13 are glycine, serine and tyrosine, respectively. The glycine residue recognized by PtxB-A6 would be incorporated into the structure of phthoxazolin A as a part of a carbamoyl moiety. The terminal module 13 of the PtxG protein includes an N-MT domain and a cytochrome P450 domain. The MT domain probably modifies the Tyr residue incorporated by PtxG-A13. Because the NRPS module 6 is likely to be a final module in the phthoxazolin A assembly line, it is difficult to predict the involvement of the NRPS modules 12 and 13 in phthoxazolin A biosynthesis at present.

Proposed model for phthoxazolin A biosynthesis

The in silico analyses of the Ptx proteins described above allowed us to propose a model for the phthoxazolin A biosynthesis (Fig 4). Phthoxazolin A biosynthesis starts with activation of glycine and the tethering to the PCP domain of PtxE, followed by formylation by the F domain to generate a formyl-glycyl-S-PCP. PKS module 1 of PtxC, representing a minimal module of the trans-AT type I PKS systems, incorporates a malonyl-CoA into the formyl-glycyl intermediate. The cyclodehydration by PtxD on the intermediate would occur after the condensation between the formyl-glycyl-S-PCP and the malonyl-CoA, although the timing of cyclodehydration remains unclear. Subsequently, four malonyl-CoA units and three methyl groups are incorporated by four PKS modules (modules 2 to 5) of PtxF and PtxB. The KR-DH domain pair in the modules 2, 3 and 4 provides three conjugated double bonds (C8-C9, C6-C7 and C4-C5), and generates the trans, cis, and cis configuration of double bonds, respectively. KR5 of PKS module 5 could be classified as a KR domain found in the partially reducing PKS [37], and introduces a hydroxyl group in the R configuration at C-3 of phthoxazolin A. NRPS module 6 activates a glycine residue, and loads it onto the polyketide intermediate to generate a carbamoyl moiety. The remaining five PKS modules could produce a triene moiety, and two NRPS modules could incorporate two additional amino acids (Ser and N-Tyr) into the polyketide assembly line. However, considering the structure of phthoxazolin A, these additional substructures are unnecessary for the biosynthesis of phthoxazolin A. The thioesterase (TE) domain usually catalyzes the release of a completed chain of PKS/NRPS products from the assembly line [38], but no such TE domain was found in the ptx assembly line. Finally, the product generated by the ptx chain assembly should be processed by some enzymatic reactions to complete phthoxazolin A biosynthesis. To date, however, no clear candidate enzyme has been found in the ptx gene cluster or its adjacent regions.

Fig 4. Proposed model for phthoxazolin A biosynthesis.

A, adenylation; ACP, acyl carrier protein; AT, acyltransferase; C, condensation; Cyp, cytochrome P450; DH, dehydratase; F, formylation; KS, ketosynthase; KS⁰, KS lacking His in the HTGTG motif; KR, ketoreductase; MT, methyltransferase; PCP, peptidyl carrier protein. The presumed inactive ACP domain of module 9 is shaded in black.

Discussion

The oxazole-containing polyketides have the distinct feature of an oxazole-moiety linked with a polyketide structure or a hybrid peptide-polyketide structure, and bestow remarkable biological activities such as antibacterial, antitumor and herbicidal properties. Only a few biosynthetic gene clusters of oxazole-containing polyketides have been identified [10–12,35,39] and their biosynthetic mechanisms have provided us valuable genetic information of the complicated pathways involved in heterocycles formation and sequential extensions by PKSs or PKS/NRPSs. Recently, we identified phthoxazolin A, which is a polyketide compound containing a 5-substituted oxazole ring and a triene moiety, as a cryptic metabolite of the original avermectin producer S. avermitilis KA-320 [18]. In the present study, we have shown that strain KA-320 harbors an extra genomic region, which is absent in the publicly available genome of strain K139, and that a biosynthetic gene cluster for phthoxazolin A is laid on the extra region of strain KA-320. The biosynthetic gene cluster could be classified into a growing number of trans-AT type I PKS systems that utilize a discrete AT enzyme to supply a malonyl-CoA unit into all the PKS modules as opposed to the cis-acting integrated AT domains of canonical PKSs [13,14].

The structure of phthoxazolin A resembles the partial structure of oxazolomycin. Comparison of the functional domains of the biosynthetic gene clusters revealed a virtually complete architectural identity with the corresponding portion of the ptx and ozm PKS-NRPS clusters over the first eight modules, with the exception of a dispensable module (PKS module 7) for the phthoxazolin A biosynthesis. The Ptx proteins are 51%-71% identical to their orthologues in the ozm cluster. Similar to the oxazolomycin biosynthetic pathway, the phthoxazolin A biosynthesis is initiated by formylation of the glycine residue, followed by extension by a malonyl-CoA unit and cyclodehydration to generate an oxazole ring moiety. The biosynthetic maturation of phthoxazolin A should require additional extension by three malonyl-CoA units to yield a triene moiety and termination by glycine incorporation to generate a carbamoyl moiety. The geometries of the conjugated triene moiety of phthoxazolin A are assigned as 4Z,6Z,8E. A PKS module encompassing the pair of a B-type KR domain and a DH domain normally biosynthesize polyketide chain elongation intermediates with a trans (E) double bond [30]. Sequence analysis of KR domains in the ptx cluster demonstrated that KR2, KR3, and KR4 contain the stereochemistry signature “LDD” motif for B-type KR domains [40, 41], although their motifs vary among the individual KR domains. These analyses indicated that the polyketide intermediate of phthoxazolin A is expected to have three trans conjugated double bonds, which are different from those of phthoxazolin A. Whether the change of stereochemistry in the biosynthetic process requires an enzymatic reaction remains obscure. However, this contradictory feature for the formation of conjugated double bonds has also been observed in the biosynthetic pathways of oxazolomycin and chivosazol [12,39].

The ptx gene cluster encodes a series of modules indispensable (from the loading module to NRPS module 6) for the phthoxazolin A biosynthesis, but also includes additional functional modules (PKS module 7 to NRPS module 13) for the biosynthesis of an unidentified compound, which might be a precursor of phthoxazolin A. In particular, PKS modules 8 and 9 include the interesting feature of a non-elongating KS (KS⁰) domain. These modules are composed of KS-KR-ACP-KS⁰ domains followed by DH-ACP domains, and are grouped as type A bimodules for dehydration [14,42] In addition, the domain pair of KS⁰-ACP as a non-elongating module is also embedded in PKS module 11. Both the type A bimodule for dehydration and the domain pair of KS⁰-ACP are ubiquitously found in the trans-AT type I PKS system [14]. The remaining modules (PKS modules 10 and 11, and NRPS modules 12 and 13) also might be responsible for a polyketide elongation and an incorporation of amino acids, as well as those of the oxazolomycin biosynthesis, suggesting that the ptx gene cluster would produce a larger compound than that of phthoxazolin A; namely, phthoxazolin A is a cleaved compound of the ancestral compound. HPLC analysis of the avaR3/ptxA double mutant demonstrated that, in addition to phthoxazolin A, the strain lost production of several more compounds that are present in the parental strain (Fig 3C): any of the compounds can be the ancestral compounds synthesized by the ptx gene cluster. In the pederin biosynthetic pathway, a putative FAD-dependent monooxygenase (PedG) has been proposed to oxidatively cleave the final product generated by the whole assembly line and produce a pederin precursor [43]. On the other hand, the ptx gene cluster and its flanking regions have no genes encoding an FAD-dependent monooxygenase, suggesting that another cleavage mechanism might be applied to the synthesis of phthoxazolin A.

In conclusion, we have shown that an extra genomic region of the original avermectin producer has a unique trans-AT type I PKS system for the biosynthesis of phthoxazolin A. These findings could provide further insight into the diversity of trans-AT type I PKS systems, which are widely distributed in bacteria [6,14]. Moreover, our finding of additional PKS/NRPS modules in the ptx assembly line suggests that S. avermitilis is capable of producing a larger intermediate than phthoxazolin A, and that the PKS/NRPS machinery incorporates a new cleavage system. Further understanding of the phthoxazolin A biosynthetic pathway should provide interesting perspectives into the engineering of polyketide/non-ribosomal peptide biosynthetic pathways.

Supporting information

S1 Fig. Phthoxazolin A production in the SALD mutants.

(A) Schematic representation for the construction of Streptomyces avermitilis large-deletion (SALD) mutant strains. The details of the construction are described in the Materials and Methods. Gray solid line represents the extra genomic region of S. avermitilis KA-320, and gray dashed lines represent the deleted DNA region in the genome. KA-320, S. avermitilis KA-320; K139, S. avermitilis K139; SUKA22, S. avermitilis SUKA22 (K139 as a genetic background). (B) HPLC chromatograms of MeOH extracts from S. avermitilis SALD mutants. KA-320/ΔavaR3, S. avermitilis ΔavaR3 (KA-320 as genetic background). mAU, milliabsorbance units at 275 nm. Phthoxazolin A is indicated by an inverted triangle.

(PDF)

S2 Fig. Phylogenetic analysis of AT domains of the trans-AT PKS system.

BaeC (CAG23950), BaeD (CAG23951), and BaeE (CAG23952) from Bacillus amyloliquefaciens FZB42; BatH (ADD82949) and BatJ (ADD82951) from Pseudomonas fluorescens BCCM_ID9359; BryP (ABK51299) from Candidatus endobugula sertula; DifA (CAG23974) from Bacillus amyloliquefaciens FZB42; DszD (AAY32968) from Sorangium cellulosum So ce12; ElsA (WP_012792904) and ElsB (WP_012792903) from Chitinophaga pinensis DSM 2588; KirC1 (CAN89639) from Streptomyces collinus Tu 365; LnmG (AAN85520) from Streptomyces atroolivaceus; LkcD (BAC76473) from Streptomyces rochei; MmpC (AAM12912) from P. fluorescens NCIMB 10586; OzmM (ABS90474) from S. albus JA3453 and OzmM (ADI12766) from Streptomyces bingchenggensis BCW-1; PedC (AAS47559) and PedD (AAS47563) from symbiont bacterium of Paederus fuscipes; PsyH (ADA82589) from an unculturable symbiont of sponge Psammocinia aff. Bulbosa; RhiG (CAL69887) from Burkholderia rhizoxina; SorO (ADN68489) from S. cellulosum So ce12; VirI (BAF50719) from Streptomyces virginiae; FabD (CAA77658) form E. coli; and FabD SAV5788 (BAC73500) from S. avermitilis.

(PDF)

S3 Fig. Sequence alignment of the conserved motifs in the KS domain core regions from ptx PKSs.

The conserved catalytic triad of C-H-H is marked with an asterisk. The numbers indicate amino acid positions within each domain.

(PDF)

S4 Fig. Sequence alignment of the conserved motif in the ACP/PCP domain core regions from ptx PKSs and NRPSs.

The Ser residue functioning as the phosphopantetheine-binding site is marked with an asterisk. The numbers indicate amino acid positions within each domain.

(PDF)

S5 Fig. Sequence alignment of the conserved motifs in the KR domain core regions from ptx PKSs.

The conserved catalytic residues are marked with asterisks. The core region for the NADP(H)-binding motif is underlined. The numbers indicate amino acid positions within each domain. The “LDD” motif for B-type KR domains is shown in a box.

(PDF)

S6 Fig. Sequence alignment of the conserved motif in the DH domain core regions from ptx PKSs.

The proposed active catalytic His residue is marked with an asterisk. The numbers indicate amino acid positions within each domain.

(PDF)

S1 Table. Oligonucleotides used in this study.

(PDF)

Data Availability

All relevant data are within the paper and its Supporting Information files.

Funding Statement

This study was supported by the Japan Society for the Promotion of Science (JP15K07358, http://www.jsps.go.jp/index.html, SK); and the Japan Association for Chemical Innovation (New Chemical Technology Research Encouragement Award, http://jaci.or.jp/, SK). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.