Investigation of carbonyl amidation and O-methylation during biosynthesis of the pharmacophore pyridyl of antitumor piericidins

Piericidins are a large family of bacterial α-pyridone antibiotics with antitumor activities such as their anti-renal carcinoma activity exhibited recently in nude mice. The backbones of piericidins are derived from β, δ-diketo carboxylic acids, which are offloaded from a modular polyketide synthase (PKS) and putatively undergo a carbonyl amidation before α-pyridone ring formation. The tailoring modifications to the α-pyridone structure mainly include the verified hydroxylation and O-methylation of the C-4′ position and an unidentified C-5′ O-methylation. Here, we describe a piericidin producer, terrestrial Streptomyces conglobatus, which contains a piericidin biosynthetic gene cluster in two different loci. Deletion of the amidotransferase gene pieD resulted in the accumulation of two fatty acids that should be degraded from the nascent carboxylic acid released by the PKS, supporting the carbonyl amidation function of PieD during α-pyridone ring formation. Deletion of the O-methyltransferase gene pieB1 led to the production of three piericidin analogues lacking C-5′ O-methylation, therefore confirming that PieB1 specifically catalyses the tailoring modification. Moreover, bioactivity analysis of the mutant-derived products provided clues regarding the structure-function relationship for antitumor activity. The work addresses two previously unidentified steps involved in pyridyl pharmacophore formation during piericidin biosynthesis, facilitating the rational bioengineering of the biosynthetic pathway towards valuable antitumor agents.


Introduction
Piericidins are a large family of bacterial natural products featuring with a pyridone core linked to a linear polyene chain, representing some of the most commonly encountered compounds in natural product studies of Actinomycetes species collected from soil, insect symbiosis, and marine samples [1][2][3][4][5]. The structural diversity of natural piericidins is mainly derived from the variable polyene side chain and substitutions in the pyridyl core [6] (Fig. 1). Due to their bioactive potential against microorganisms, insects, and tumor cells, eighty-nine biosynthetic and eleven chemosynthetic piericidin analogues have been discovered thus far [7]. The activity of piericidins was highlighted again recently by piericidin A1, the firstly discovered member of this family. The compound was verified to induce apoptosis in renal carcinoma cells by reducing the ROS level caused by the upregulated peroxiredoxin 1 (PRDX1) and also demonstrated potent antitumor efficacy in renal carcinoma xenograft mice [8,9].
The piericidin biosynthetic gene cluster (pieBGC) was successively investigated in two Streptomyces species [2,10,11], and the biosynthesis of piericidin backbone was elucidated as follows: 1) the modular polyketide synthase (PKS) machinery of PieA1-A6 releases a polyene chain with a β, δ-diketo carboxylic acid terminal group after nine rounds of 2-carbon assembly; 2) the nascent polyene chain is amidated at the terminal carboxylic acid by the ATP-dependent amidotransferase PieD; 3) the amidated product undergoes cyclization spontaneously or driven monooxygenase PieE and C-4 ′ O-methylation by the O-methyltransferase PieB2 [10,11], whereas the C-5 ′ O-methylation putatively conducted by the O-methyltransferase PieB1 remains unverified (Fig. 2B). Moreover, there are additional unconventional modifications such as the obscure epoxidation of the polyene side chain and O-glycosylation of either the α-pyridone core or polyene side chain [12] (Fig. 1). Despite extensive efforts, at least two points remain to be identified in the piericidins biosynthetic pathway; the putative amidation function of PieD and the pyridyl C-5 ′ O-methylation thought to be conducted by PieB1 due to the lack of the accumulation of new intermediates after individual inactivation of the pieD and pieB1 genes [2,10].
We recently discovered that terrestrial Streptomyces conglobatus can produce piericidins, as this species contains a pieBGC with the genes found in two different loci of the genome ( Fig. 2A). The construction of gene deletion mutants in this new piericidin producer and structural elucidation of the intermediates accumulated by mutant strains allowed us to address the carbonyl amidation step that is essential for the formation of the α-pyridone ring pharmacophore and the tailoring Omethylation modification at the C-5 ′ of pyridyl core. Moreover, we investigated a structure-activity relationship by analyzing the antitumor activities of the five compounds isolated from mutant strains. This work further clarifies the piericidin biosynthetic pathway to facilitate rational bioengineering towards accumulating valuable antitumor agents.

DNA manipulation and chemicals
The oligonucleotides used in this work were ordered from Shanghai Generay Biotech Co., Ltd. Restriction endonucleases and T4 DNA ligase were purchased from New England Biolabs. Chemicals were purchased from Sigma-Aldrich. The Plasmid DNA Extraction Kit was purchased from Shanghai Generay Biotech. DNA fragments were assembled by using the U-Clone master mix kit (Evomic Science, Sunnyvale, CA, USA). Genomic DNA was prepared by using 10% Chelex 100 resin (Bio-Rad) solution. PCR amplifications were carried out by using Phusion High-Fidelity PCR Master Mix (New England Biolabs) for DNA cloning and 2 × Flash PCR MasterMix (CWBIO) for colony screening.

Strains, media, and culture conditions
Strain RJ8 is a mutant of terrestrial S. conglobatus [14] and was used as the starting strain/wild type strain in this work. E. coli DH10B was used as a host for plasmid construction. E. coli ET12567 containing the helper plasmid pUZ8002 was used as a transitional host for introducing the E. coli− Streptomyces shuttle plasmid into RJ8. TSBY medium (3% tryptone soy broth, 0.5% yeast extract, 10% sucrose, 0.1% antifoam) was used to culture mycelium of RJ8 or derived strains. SFM agar medium (2% soya flour, 2% D-mannitol, 2% agar) was used for manipulation of conjugation and mutant screening from the RJ8 strain. SGC medium consisting of 3% soybean flour, 5% glucose, 0.5% CaCO 3 , and 0.1% (v/v) antifoam was used as the medium for secondary metabolite production. Streptomyces fermentation was conducted in a 250 mL conical flask fitted with a metal spring, with 50 mL of SGC medium inoculated with 1% (v/v) of 3 d TSBY culture, and then incubated at 30 • C, 220 rpm for 5 d. E. coli strains were grown in Luria-Bertani (LB) broth (1% tryptone, 0.5% yeast extract, 0.5% NaCl) or an LB agar (1.5% agar) plate at 37 • C supplemented with the corresponding antibiotics.

Analytical methods
For routine HPLC-MS analysis, 7 mL of five-day fermentation broth was extracted with an equal volume of ethyl acetate. The extract was dried via vacuum evaporation and redissolved into 500 μL of methanol. Then, 20 μL of the final sample was injected into the HPLC-MS for analysis.
HPLC-MS analysis was conducted on an HPLC coupled with a Waters Acquity QDa detector. The HPLC was fitted with a Waters Xbridge C18 column (250 × 4.6 mm, 5 μm). Samples were eluted with the mobile phases of acetonitrile and H 2 O (0.1% formic acid, v/v) at a flow rate of 0.7 mL/min with a gradient elution of 30-95% acetonitrile over 30 min. The mass spectrometer was run in positive ionization mode, scanning from m/z 200 to 1250.
HRMS spectra were acquired with a UPLC Waters XeVO G2-XS Q TOF mass spectrometer in positive ionization mode, scanning from m/z 100 to 1200. MS/MS analysis was conducted with a collision energy ramp of 30-40 eV, scan time (sec) of 0.200, and interscan time (sec) of 0.014. The UPLC was fitted with a Waters Acquity UPLC BEH C18 column (2.1 mm × 50 mm, 1.7 μm). A mobile phase of acetonitrile and H 2 O (0.1% formic acid, v/v) was used for isocratic sample elution with 60% acetonitrile over 5 min at a flow rate of 0.4 mL/min. NMR spectra were recorded on a Bruker AV-600 MHz NMR spectrometer in CDCl 3 or DMSO-d 6 with TMS as an internal standard.

Deletion of pieB1CD to generate the mutant RJ10
To generate the platform mutant RJ10, the whole ORFs of the pieB1CD genes were selected for deletion from the RJ8 strain, which lost the ability to produce the dominant secondary metabolites of conglobatin and neoantimycins after inactivation of congE and natE from S. conglobatus [14,15]. To construct a homologous recombination cassette, primers Lpie-S and Lpie-A and primers Rpie-S and Rpie-A (Table S1) were used to amplify a 1578 bp left arm and a 1630 bp right arm, respectively. By using the U-Clone master mix kit, the two PCR fragments were assembled with the E. coli− Streptomyces shuttle plasmid pRJ2 linearized with XbaI and EcoRI [14,15]. The resulting plasmid pRJ32 was transformed into the RJ8 strain via conjugation from E. coli ET12567 with the helper plasmid pUZ8002. The target mutant was screened by using colony PCR with the primers Tpie-SS and Tpie-AA (Table S1) from the hygromycin-sensitive colonies prepared after two rounds of propagation on SFM plates without antibiotics added. The expected PCR products were 443 bp for the target mutation and 3560 bp for the wild type strain RJ8. The PCR product of the mutant was sequenced for final confirmation.

Complementation of RJ10 with different compositions of the pieB1, pieC, and pieD genes
To identify the functions of the pieB1, pieC, and pieD genes in RJ10, the ORFs of pieB1CD, pieCD, pieB1C, and pieD were constructed under the constitutive promoter KasO*p [16], respectively, based on the integrative vector pRJ5 [17]. In detail, the KasO*p promoter containing fragments (167-170 bp) was amplified from pRJ255 [17] (see primers in Table S1). The gene transcription frames consisting of 3172 bp (pieB1CD), 2441 bp (pieCD), and 1906 bp (pieD) were amplified from the genomic DNA of RJ8 (see primers in Table S1). By using U-Clone master mix kit, the KasO*p promoter fragment and the corresponding ORF fragment were assembled with the linearized plasmid pRJ5 (NsiI, EcoRV) to yield the plasmids pRJ261 for pieB1CD, pRJ262 for pieCD, and pRJ263 for pieD. Moreover, the 1447 bp KasO*p-pieB1C cassette was amplified from pRJ261 and introduced into the NsiI and EcoRV sites of pRJ5 to produce the plasmid pRJ267 (see primers in Table S1).
The resultant plasmids pRJ261, pRJ262, pRJ263, and pRJ267 were then introduced into RJ10 via conjugation. The exconjugants were confirmed via the phenotype of hygromycin resistance on SFM plates and subsequent colony PCR (see primers in Table S1).

Purification of compounds 1-7
Compound production was conducted by using RJ8 for 1 and 2, RJ10 for 3 and 4, and pieCD/RJ10 for 5-7. The 12 L of fermentation broth for each strain was extracted three times with an equal volume of ethyl acetate after adjusting the pH to 6.0 with formic acid. The combined organic phases were concentrated under vacuum, and subjected to vacuum liquid chromatography with silica gel (200-300 mesh) for fractionation by using stepwise elution with dichloromethane-methanol (from 50:1 to 0:1, v/v). Guided by HPLC-MS analysis, the fractions containing the expected compounds were combined and delivered for further separation through an ODS chromatography column (YMC-Pack Pro C18 RS, 20 × 250 mm, 5 μm) by using medium-pressure preparative liquid chromatography at a flow rate of 15 mL/min with a gradient of 30-95% acetonitrile/H 2 O (0.1% formic acid) for 4 h, detection at a wavelength of 254 nm. Finally, the target fractions were purified by using HPLC fitted with a semipreparative column (Waters Xbridge C18, 10 × 250 mm, 5 μm) with 80% methanol/H 2 O (0.1% formic acid) as the eluent and a flow rate of 3 mL/min.

Protein expression and purification of PieB2
The pieB2 gene was amplified from the genomic DNA of RJ8 (see primers in Table S1). The PCR fragments of pieB2 were introduced into the NdeI and XhoI sites of pET29a by using the U-Clone master mix kit, yielding the plasmid pRJ284. The resultant plasmid was transformed into E. coli BL21 (DE3) for protein expression. The overnight culture (1 mL) of the target transformant was inoculated into 500 mL of LB medium supplemented with kanamycin (50 μg/mL) (37 • C, 220 rpm).
Protein expression was induced by adding isopropyl-β-D-thiogalactopyranoside (IPTG) (0.2 mM) when cell turbidity reached to an OD 600 of 0.6-0.8, followed by 15 h of cultivation at 22 • C. Cells were harvested by centrifugation (11,325×g, 5 min) and then resuspended in buffer A (50 mM Tris HCl, 300 mM NaCl, pH 7.2) for sonication. The supernatant of the cell lysate (34,925×g, 25 min of centrifugation) was passed through a 0.22 μm filter before being loaded onto a His-Bind affinity column (1 mL bed volume). Target proteins were eluted via stepwise increases in the concentration of imidazole (10-500 mM). Concentration and buffer exchange of the target protein was performed by using Amicon Ultra-4 concentrators (Millipore, 10 kDa cut-off). Protein-containing fractions were analyzed by Bis-Tris Gel SDS-PAGE (4-12%). The protein concentration was measured with a NanoDrop 1000 spectrophotometer.

PieB2 enzyme assays
The enzymatic assay was carried out in a 100 μL reaction system with 100 μM compound 5, 10 μM PieB2, and 2 mM S-adenosylmethionine in 50 mM Tris-HCl buffer (pH 8.0) and incubated at 28 • C for 5 h PieB2 preboiled at 100 • C for 10 min was used as a negative control. The reactions were stopped by adding 400 μL of acetonitrile, and the solutions were centrifuged at maximum speed for 5 min before taking 20 μL of the supernatant for HPLC-MS analysis.

Bioactivity assays of compounds 1-7
For the antimicrobial activity assay, the minimum inhibition concentration (MIC) values of 1-7 were evaluated using the 96-well microtiter plate method [18]. The tested compounds were dissolved in DMSO at a concentration of 3200 μg/mL. The compounds used as positive controls were vancomycin for MRSA (CICC10201) and B. mycoides, rifampicin for M. smegmatis mc2155, and amphotericin B for C. albicans. DMSO was used as a negative control in the assay.
A cytotoxicity assay was conducted by using HL-60, A549, SMMC-7721, MDA-MB-231, and SW480 human cancer cells. The cells were incubated in RPMI-1640 medium for 12-24 h at 37 • C before the tested compounds were added and then incubated for an additional 48 h. For the first evaluation, each compound was used at a concentration of 40 μM, and for determining the IC 50 values (half-inhibitory concentrations), a gradient of compound concentrations was used. Cell viability was determined by adding MTS solution and incubating the cells at 37 • C for 4 h. The amount of the formazan crystal products dissolved in DMSO was detected with a 570 nm in spectrophotometer.

Discovery of piericidins and bioinformatics analysis of pieBGC from S. conglobatus
Terrestrial S. conglobatus was previously shown to produce the antitumor macrodiolide conglobatin and cyclic depsipeptide neoantimycins [14,19]. During a screening of neoantimycin analogues from the culture of S. conglobatus, we found two components that were structurally identical to piericidins A1 (1) and C1 (2) according to comparative analysis of the 13 C and 1 H NMR data (Tables S2 and S3). Interestingly, genome mining revealed that the expected pieBGC should be composed of the two clusters of genes located at a distance of 322 kb in the genome of S. conglobatus (GenBank OM674914) (Fig. 2A). The left part of the pieBGC contains five PKS genes, pieA1-A5, which are responsible for the loading starter unit and six two-carbon elongations. The right portion includes the pieA6 gene encoding the last two modules of the piericidin PKS, and the pieB1, C, D, B2, and E genes putatively involved in the α-pyridine ring formation and tailoring modifications ( Fig. 2A). Except for the different physical locations, the newly found pieBGC displays the same gene organization and a high level of protein homology compared to the reported pieBGC from Streptomyces piomogeues and Streptomyces sp. SCSIO 03032 [2,10] (Table 1), therefore suggesting a correlation to the production of 1 and 2.

Deletion of pieB1CD resulted in the accumulation of three new products
The putative ATP-dependent amidotransferase encoded by pieD in pieBGC was assumed to transfer a nitrogen from ammonia or glutamine to the terminal β, δ-diketo carboxylic acid group of the nascent C 18 polyene chain offloaded from the PKS assembly line of PieA1-A6, and the amidated product could undergo α-pyridine ring cyclization either spontaneously or driven by a putative cyclase encoded by pieC [2] (Fig. 2B). In a previous experiment, the pieD gene was inactivated by gene replacement with an apramycin resistance cassette. However, the resultant ΔpieD strain did not accumulate new products such as linear polyene carboxylic acids, although the production of 1 was destroyed as expected [2]. To probe the function of PieD, we designed a mutant with the open reading frame (ORF) of the pieB1CD gene cassette deleted from strain RJ8 (Fig. 3A), which was created from S. conglobatus by removing the dominant products conglobatin and neoantimycins [14,20].
According to HPLC-MS analysis of the fermentation extract, the resultant ΔpieB1CD mutant strain (RJ10) lost the ability to produce 1 and 2 as expected. Surprisingly, three new products, 3, 4, and 10, were accumulated in RJ10 with a considerable yield of 3 (Fig. 3B, iv). When pieB1CD cassette was delivered back to RJ10 via an integrative vector, the production of 1 and 2 was recovered to the same level as that of the parent strain RJ8, and the three new compounds disappeared in the fermentation culture (Fig. 3B, v and vi). When the pieB1 and pieC genes were introduced into RJ10, the resultant strain pieB1C/RJ10 (ΔpieD) gave the same metabolite profile as RJ10 (Fig. 3B, iii), therefore confirming that the accumulation of 3, 4, and 10 was directly correlated with the absent function of PieD.

Structural elucidation of 3 and 4 accumulated by the ΔpieD strain
To characterize the structures of the three new products accumulated by the ΔpieD strain, we performed a 24-L liquid fermentation to purify enough of each compound for spectral analysis. Finally, 27 mg of compound 3 and 9 mg of compound 4 were obtained as yellow oils; however, the instability of compound 10 during the isolation procedure prevented its quantitative preparation.
According to analysis of the 1 H and 13 C NMR data, 3 should be structurally identical to a reported fatty acid polyketide isolated from marine-derived S. youssoufiensis OUC6819 [20] (Fig. 4, Table S4). The molecular formula of 4 was determined to be C 20    C-10 and C-11 were proposed to be trans based on the coupling constant values of 7.1 Hz (Fig. 4, Table S5). The molecular formula of 10 was determined to be C 25  We speculate that 3 and 4 could be oxidatively degraded from the unstable β,δ-diketo carboxylic acids offloaded from the piericidin PKS, in which the starter units of acetate and isobutyrate were incorporated to yield 3 and 4, respectively (Fig. 2B), similar to the piericidin analogues derived from diverse starter precursors in nature (Fig. 1) [6]. Interestingly, the structure of 3 is identical to a compound discovered from the piericidin producer S. youssoufiensis OUC6819 [13,20], implying that 3 could be naturally produced with the inefficient activity of PieD. We hypothesized that compound 10 could be an α-pyrone product (Fig. S1) that was formed spontaneously from the instable β,δ-diketo carboxylic acid nascently offloaded from the PKS, in terms of the evidence that the α-pyrone ring was produced efficiently in vitro by the last module of the piericidin PKS with an inactivated C-terminal TE [2]. In conclusion, the accumulation of 3, 4 and 10 in the ΔpieD mutant provides indications for the previously speculated PieD-catalyzed carbonyl amidation of the carboxylic acid nascently released by the piericidin PKS before α-pyridone ring cyclization.

Deletion of pieB1 or pieB1C resulted in the accumulation of 5-7
In a previous study, based on gene disruption and enzymatic assay, the O-methyltransferase gene pieB2 was verified to conduct O-methylation at the C-4 ′ hydroxyl group generated by FAD-dependent monooxygenase PieE in the piericidin pyridyl core [10] (Fig. 2B). However, due to the lack of new intermediates detected in the ΔpieB1 mutant, it remains to be confirmed whether another O-methyltransferase gene, pieB1, performs O-methylation of the C-5 ′ hydroxyl group [10]. To probe the function of pieB1 in our piericidin producer, we created a ΔpieB1 mutant through complementation of RJ10 with the pieC and pieD genes based on an integrative vector. Interestingly, the ΔpieB1 strain (pieCD/RJ10) produced three new products, 5-7, with considerable yield for 6 according to HPLC-MS analysis of the fermentation extract (Fig. 3B, i). We speculate that the failed accumulation of 5-7 in the reported ΔpieB1 strain may be attributed to an interruption in the transcript cassette pieB1CDB2E, in which the pieB1 gene was destroyed by a fragment consisting of the apramycin resistance gene aac(3)IV and the conjugation gene oriT [10].
We also confirmed the function of PieC by creating a ΔpieB1C mutant, which was generated by complementation of RJ10 with the pieD gene. According to HPLC-MS analysis, the production of 5-7 was remarkably reduced in ΔpieB1C compared to the production levels in ΔpieB1 strain (Fig. 3B, i and ii), therefore supporting the proposal that PieC acted as a polyketide cyclase to increase the rate of α-pyridine ring cyclization during piericidin biosynthesis [2]. Moreover, compounds 3 and 4 were not detected in pieCD/RJ10 or pieD/RJ10 strain according to LC-MS analysis, indicating that PieD could convert int-I into int-II quite specifically and efficiently so that even trace amount of int-I was not remained to be degraded into 3 or 4 (Fig. 2B).

Structural elucidation of 5-7 to identify the function of PieB1
To prepare adequate samples of 5-7 for structure characterization, the liquid fermentation broth of the ΔpieB1 mutant strain (pieCD/RJ10) was scaled up to 12-L. Subsequent chromatographic separations yielded yellow oil compounds including 3 mg of 5, 13 mg of 6, and 30 mg of 7. Based on HRESIMS and NMR analyses, the planar structures of 5-7 were identified as new piericidin analogues without hydroxylation at C-4 ′ and O-methylation of the C-5 ′ hydroxyl group in the α-pyridone core of 1 or 2 (Fig. 4).
The molecular formula of 5 was determined to be C 23 (Table S6), 5 contains the same skeleton as the reported piericidin analogue Mer-A2026B [21,22], which possesses an additional methoxy group at the C-5 ′ hydroxyl compared to 5. The structure of 5 was further confirmed by 2D NMR, in particular with the use of the HMBC spectrum (Fig. 4).
The molecular formula of 6 was determined to be C 24 13 C NMR data revealed that 5 and 6 had the same skeleton (Table S6). The only difference between the two was that the C-3 ′ hydroxyl in 5 was substituted by a methoxy in 6. This difference was validated by the HMBC correlations from H-7 ′ (δ H 3.78) to C-3 ′ (δ C 169.40) in 6 (Fig. 4, Table S6, Fig. S9E). The geometries of olefins in 6 were deduced as 2E, 5E, 7E, and 11E based on the NOESY correlations of H-1 to H-14, H-8 to H-6, and H-10 to H-12, and the large 1 H-1 H coupling constant of 15.5 Hz between H-5 and H-6 (Table S6, Fig. S9F). A large coupling constant (9.0 Hz) suggested an anti-relationship between H-9 (δ H 2.67) and H-10 (δ H 3.62) in 6, and the relative configuration of 9R* and 10R* was supported by the NOESY correlations of H-8/H-10/H-16/H-17/H-9 (Fig. S9F).

Investigation of the activity of PieB2 for C-3 ′ O-methylation
The C-3 ′ O-methylation in 6 or 7 suggests an unidentified Omethylation step in the tailoring modifications of piericidin biosynthesis. Similarly, the C-3 ′ O-methylated version of 1 was recently discovered from a marine-derived Streptomyces (Peng et al., 2021). We first suspected that PieB2 possessed promiscuous activity for C-3 ′ Omethylation to convert 5 to 6. To probe this activity, we obtained the soluble proteins of C-His 6 -fused PieB2 from E. coli BL21 (DE3) harboring the pET29a-derived expression plasmid pRJ284. The proteins were purified to near homogeneity via Ni-NTA chromatography (Fig. S2). However, no conversion of 5 to 6 was observed in the assay containing 100 μM 5, 10 μM PieB2, and 2 mM S-adenosylmethionine (SAM) in Tris− HCl buffer (50 mM, pH 8.0) incubated at 28 • C for 5 h. Considering the widespread of O-methyltransferase in bacterial secondary metabolites biosynthesis [23], we suppose that an O-methyltransferase elsewhere could generate the C-3 ′ O-methylation of 6 and 7, being similar to the cryptic occurrence of C-10 O-methylation in piericidins (Fig. 1). Moreover, the C-11/C-12 epoxy ring in 7 could be formed by a monooxygenase, e.g. epoxidase and P450 monooxygenase, since that the epoxidations of alkenes in natural products are conducted by various monooxygenases [24].

The bioactivities of 3-7
The antitumor activities of 3-7 were evaluated in five human cancer cell lines including leukemia HL-60, lung cancer A549, liver cancer SMMC-7721, breast cancer MDA-MB-231, and colon cancer SW480 cells. Except for the weak inhibitory activity against SMMC-7721 cells (IC 50 26.93 ± 2.48) detected for 6, the cell inhibition rates of 3-7 at a concentration of 40 μM were all less than 50%. Moreover, only weak antimicrobial activities (MICs, above 100 μg/mL) of 3-7 were determined against pathogenic microorganisms such as methicillin-resistant Staphylococcus aureus (MRSA), Bacillus mycoides, Mycobacterium smegmatis, and Candida albicans. Of note, excellent antiproliferative activity (IC 50 , 80 nM) in the HL-60 cell line was detected with an analogue of 6, piericidin N produced by mangrove sediment-derived S. psammoticus SCSIO NS126 [25], which contains a C-5 ′ methoxyl instead of the C-3 ′ methoxyl group in 6 (Fig. S3), implying that the position of O-methylation in the pyridyl core should play an important role in the antitumor activities of piericidins.

Conclusion
The work has addressed two key steps in the biosynthetic pathway of piericidins with antitumor activity, including carbonyl amidation before pharmacophore α-pyridone ring formation and the tailoring C-5 ′ Omethylation modification in the α-pyridone core, both of which were not identified in previous investigations because no new products were detected with the same mutations of other piericidin producers. Moreover, we elucidated a structure-activity relationship by analyzing the antitumor activities of the five compounds isolated from the mutant strains generated in this work. Clarification of the piericidin biosynthetic pathway would facilitate the creation of diverse piericidin analogues for antitumor agent screening and the targeted accumulation of valuable components via rational bioengineering.

Declaration of competing interest
Authors declare that they have no conflict of interest.