Characterization of the Biosynthetic Gene Cluster and Shunt Products Yields Insights into the Biosynthesis of Balmoralmycin

ABSTRACT

Angucyclines are a family of structurally diverse, aromatic polyketides with some members that exhibit potent bioactivity. Angucyclines have also attracted considerable attention due to the intriguing biosynthetic origins that underlie their structural complexity and diversity. Balmoralmycin (compound 1) represents a unique group of angucyclines that contain an angular benz[α]anthracene tetracyclic system, a characteristic C-glycosidic bond-linked deoxy-sugar (d-olivose), and an unsaturated fatty acid chain. In this study, we identified a Streptomyces strain that produces balmoralmycin and seven biosynthetically related coproducts (compounds 2−8). Four of the coproducts (compounds 5−8) are novel compounds that feature a highly oxygenated or fragmented lactone ring, and three of them (compounds 3−5) exhibited cytotoxicity against the human pancreatic cancer cell line MIA PaCa-2 with IC₅₀ values ranging from 0.9 to 1.2 μg/mL. Genome sequencing and CRISPR/dCas9-assisted gene knockdown led to the identification of the ~43 kb balmoralmycin biosynthetic gene cluster (bal BGC). The bal BGC encodes a type II polyketide synthase (PKS) system for assembling the angucycline aglycone, six enzymes for generating the deoxysugar d-olivose, and a hybrid type II/III PKS system for synthesizing the 2,4-decadienoic acid chain. Based on the genetic and chemical information, we propose a mechanism for the biosynthesis of balmoralmycin and the shunt products. The chemical and genetic studies yielded insights into the biosynthetic origin of the structural diversity of angucyclines.

IMPORTANCE Angucyclines are structurally diverse aromatic polyketides that have attracted considerable attention due to their potent bioactivity and intriguing biosynthetic origin. Balmoralmycin is a representative of a small family of angucyclines with unique structural features and an unknown biosynthetic origin. We report a newly isolated Streptomyces strain that produces balmoralmycin in a high fermentation titer as well as several structurally related shunt products. Based on the chemical and genetic information, a biosynthetic pathway that involves a type II polyketide synthase (PKS) system, cyclases/aromatases, oxidoreductases, and other ancillary enzymes was established. The elucidation of the balmoralmycin pathway enriches our understanding of how structural diversity is generated in angucyclines and opens the door for the production of balmoralmycin derivatives via pathway engineering.

INTRODUCTION

Angucyclines are aromatic polyketides that contain or are derived from an angular tetracyclic decaketide ring system, with typical members containing the angular tetracyclic benz[α]anthracene structure (1–3). The number of naturally occurring angucyclines has increased steadily in the literature since the isolation and characterization of tetrangomycin from Streptomyces rimosus in 1965 (4). Angucyclines and other aromatic polyketides exhibit a wide range of bioactivities, and some of them have been developed into antibiotics and other therapeutic drugs (1, 3, 5). New aromatic polyketides with promising bioactivity are still emerging, as highlighted by the recent discovery of the antifungal polyketide turbinmicin from a bacterium residing in the gut of the sea squirt Ecteinascidia turbinata (6). In bacteria, many structurally complex, carbon-skeleton-containing, secondary metabolites are synthesized by PKSs that are broadly grouped into type I, type II, and type III PKSs (7, 8). The decaketide skeletons of angucyclines are assembled from the acetate units of acyl-CoA starters and extenders by type II PKSs. Type II PKSs are dissociative enzyme complexes that rely on a ketosynthase subunit, an acyl carrier protein (ACP), and a chain length factor to catalyze the condensation of acyl-CoA starter and extender units to synthesize the polyketide chains (1, 3, 9). The polyketide chains assembled by the type II PKSs usually undergo cyclase/aromatase-catalyzed cyclization and post-PKS modifications to generate mature aromatic products (1, 10–18). The structural diversity in angucyclines arises mainly from the different patterns of cyclase-controlled folding and aromatization of the decaketide chains as well as from the diverse, post-PKS modifications introduced by tailoring enzymes (1, 11, 19–22). The tailoring enzymes are able to not only decorate the angucycline skeleton (e.g., glycosylation) but also rearrange the angucycline carbon skeleton to generate atypical angucyclines in some cases (23–26).

Among the angucyclines, there is a family of urdamycin-type angucyclines (UTA) that feature a classical angucycline ring system and a unique C-glycosidic bond-linked deoxy-sugar d-olivose moiety (1–3, 27). While some urdamycin-type angucyclines (UTA-1 group) are further decorated with mono-, di-, and tri-saccharides (e.g., aquayamycin [28], grecocyclines [29], urdamycins [27]), there are a group of urdamycin-type angucyclines (UTA-2 group) that contain an unsaturated fatty acid chain attached to the d-olivose moiety via an ester linkage (30, 31). Simocyclinones, which are the best-studied UTA-2 members, feature a complex molecular architecture that is composed of an angucycline aglycone, a d-olivose, a tetraene fatty acid chain, and an aminocoumarin moiety (31). The other known members of the UTA-2 group, including capoamycin (32), dioxamycin (33), fradimycins (34), waldiomycin (34, 35), antibiotic C104 (36), and balmoralmycin (30), do not contain the aminocoumarin moiety. It should also be noted that the angucycline moieties of the UTA-2 members differ from each other slightly in their degree of unsaturation and level of oxygenation. The unsaturated fatty acid chains also vary among the members by the length of the carbon chain and by the degree of unsaturation. Some of the UTA-2 compounds display notable antimicrobial bioactivities, with simocyclinone D8 acting as a strong inhibitor of the type II DNA topoisomerase and as a potent antibacterial agent against Gram-positive bacteria (5, 37, 38). Waldiomycin inhibits the growth of methicillin-resistant Staphylococcus aureus via a novel mechanism of interacting with the signaling protein histidine kinase WalK (34, 35). Several UTA-2 members display anticancer activity, with fradimycin B exhibiting a notable tumor cell-growth inhibitory activity (IC₅₀ = 0.13 μM).

The biosynthetic pathways for the UTA-2 members remained unknown until the characterization of the biosynthetic gene cluster (BGC) of simocyclinones (39, 40). The simocyclinone pathway remains the only characterized UTA-2 pathway as of today. In this study, we report the isolation of balmoralmycin (compound 1), the structurally related antibiotic C104 (compound 2), C-olivosyltetrangulol (compound 3), and five putative shunt products (compounds 4−8) from a newly isolated Streptomyces strain. The identification of the balmoralmycin BGC (bal BGC) unveiled a biosynthetic pathway that explains the structural differences between balmoralmycin and simocyclinones. The three coproducts, namely, compounds 3−5, not only are informative about the biosynthetic mechanism but also exhibit significant cytotoxicity against human cancer cell lines. The genetic and chemical information allows us to propose a biosynthetic mechanism to explain how balmoralmycin and the shunt products are formed.

RESULTS AND DISCUSSION

Isolation and characterization of balmoralmycin and the shunt products from Streptomyces sp. P01.

We recently isolated Streptomyces sp. P01 and other actinobacterial strains from the sediment samples obtained from a freshwater lake in Singapore. Our metabolite profiling-based strain prioritization suggested Streptomyces sp. P01 as a biosynthetically talented strain that produced several novel compounds that display a distinct molecular mass and a UV-visible spectrum. When we cultivated Streptomyces sp. P01 in various liquid and solid culture media, the strain produced a series of compounds in abundance on the solid PM3 medium. Scaled-up fermentation, organic extraction, and normal-phase and reverse-phase chromatography yielded eight colored compounds (compounds 1−8) with distinct UV-visible absorbance (Scheme S1). Compounds 1−3 were the major fermentation products that were isolated in copious quantities, whereas compounds 4−8 were minor components that were produced in much lower titers. Based on the high-resolution mass spectrometry (HRMS) and nuclear magnetic resonance (NMR) spectral data, four compounds were found to be previously characterized compounds: balmoralmycin A (compound 1), antibiotic C104 (compound 2), C-olivosyltetrangulol (compound 3), and galtamycinone (compound 4). Interestingly, compounds 1−4 were previously separately isolated from four different actinobacterial strains, with compound 1 produced by Streptomyces sp. P6417 (30), compound 2 produced by an unnamed Streptomyces strain (36), compound 3 by Streptomyces cyanogenus S136 (15), and compound 4 by Micromonospora sp. Tű6368 (41). Compounds 5−8, which seem to be structurally related to compounds 1 and 2, are novel compounds that have not yet been reported. Below, we briefly describe how the structures of compound 5−8 were established, based on the HRMS and NMR data.

Compound 5 has a molecular formula of C₃₅H₃₆O₁₁ and is based on the (+)-HRESIMS ion at m/z 633.2347 ([M+H]⁺, calcd for C₃₅H₃₇O₁₁, 633.2330) and the ¹³C NMR data (Table S1). The ¹³C NMR spectrum supports the structural similarity with compounds 1 and 2 (30, 36), as supported by the presence of the signals for the olivose moiety (δ_C 78.7, 74.9, 71.8, 70.4, 40.9, and 18.2), the 2,4-decadienoic acid (δ_C 166.9, 145.7, 145.1, 128.6, 119.4, 32.8, 31.2, 28.3, 22.4, 13.9), and the anthraquinone-containing aglycone (carbonyls: δ_C 188.4, 180.0). Further inspection of the ¹H NMR data (Table S2) suggested that the olefinic protons of the A-ring in compounds 1 and 2 (Fig. 1A) are replaced by a sp³ methylene (δ_H 3.64, 3.58) in compound 5. The ¹³C-NMR signals for the methylene (δ_C 43.2), the additional carbonyl (δ_C 171.9), and the oxygenated carbon (δ_C 89.3) were reminiscent of a seven-membered lactone ring (A-ring) seen in urdamycin L, a shunt product in the angucycline urdamycin pathway (42). The seven-membered lactone in compound 5 possesses a higher oxidation degree than that of urdamycin L because of the keto group at C-4 (δ_C 202.1). The 2D NMR data (Fig. 1B), including the key heteronuclear multiple bond correlations (HMBC) from H-5 (δ_H 8.23) to C-4 (δ_C 202.1), from 3-Me (δ_H 1.63) to C-2 (δ_C 43.2)/C-3 (δ_C 89.3)/C-4, and from H₂-2 (δ_H 3.64, 3.58) to C-1 (δ_C 171.9)/3-Me (δ_C 22.4) further support the existence of the seven-membered lactone ring in compound 5. Given the similar NMR chemical shifts and the shared biogenetic origins (discussed later) between compounds 5 and 2, the olivose in compound 5 was likely to be d-olivose, as in compound 2, whose absolute configuration was unambiguously determined via total synthesis (36). This deduction is also supported by their similar proton-proton coupling constants in the sugar moieties, including the apparent large constants for axial protons H-1′ (br d, J = 11.0 Hz), Ha-2′ (ddd, J = 12.8, 11.3, 11.0 Hz), H-4 (t, J = 9.4 Hz), and the doublet methyl (d, J = 6.3 Hz) (Table S2). While the absolute configuration of C-3 remains to be determined, the biogenetic origin of C-3 that is known to arise from the early cyclization stage of angucycline biosynthesis suggests that it is likely to adopt the S-configuration (1). Together, the results established compound 5 as a new compound that is structurally related to balmoralmycin but contains a unique seven-membered lactone ring that also features additional hydroxyl and keto groups at the C-3 and C-4 positions, respectively (Fig. 1A).

FIG 1.

(A) Chemical structures of compounds 1−8 isolated from Streptomyces sp. P01. (B) Observed 2D NMR correlations for compounds 5−8.

With a molecular formula of C₃₅H₃₈O₁₀, compound 6 features ¹H and ¹³C NMR spectra (Tables S1 and S2) that share similarities with those of compound 5. The data suggest that although compounds 5 and 6 share a similar balmoralmycin-type carbon framework, the seven-membered lactone ring in compound 5 was replaced by a linear 3-methylbut-2-enoic acid in compound 6. The presence of the 3-methylbut-2-enoic acid fragment was reinforced by the HMBC cross-peaks (Fig. 1B) between H₂-4 (δ_H 3.69 and 3.65) and C-2 (δ_C 119.7)/3-Me (δ_C 19.0)/C-5 (δ_C 137.7) as well as those between H-2 (δ_H 6.16) and C-1 (δ_C 169.0)/C-3 (δ_C 155.2)/3-Me (δ_C 19.0). Given that Z/E (or trans/cis) geometrical isomers generally display different patterns of proton-proton correlations in the Nuclear Overhauser Enhancement SpectroscopY (NOESY) experiment, the clear NOESY correlation between H-2 and H₂-4 (Fig. 1B) supports that the C2=C3 double bond in compound 6 adopts an E/trans-configuration, which resembles that of angucycline (E/trans)-himalaquinone-E (43). The NMR data (Tables S1 and S2) suggest that compound 7 (C₃₆H₄₀O₁₀) is the methyl ester of compound 6 with an extra methyl group at C-1. In contrast to the E/trans-configuration of the C2=C3 double bond of compound 6, the C2=C3 double bond adopts a Z/cis-configuration in compound 7, as supported by the γ-gauche effect-induced upfield shift of C-4 (7: δ_C 32.4; 6: δ_C 40.1)⁴³ and the NOESY correlation (Fig. 1B) between 3-Me (δ_H 1.83) and H-2 (δ_H 6.04). Compound 8 has a molecular formula of C₂₅H₂₄O₉, and the NMR data suggest that compound 8 lacks the 10-carbon 2,4-decadienoic acid moiety seen in compounds 5 and 6. Like compound 6, the C2=C3 double bond in compound 8 is also likely to adopt the E/trans-configuration, as supported by the NOESY correlations (Fig. 1B).

Antibacterial activity and cytotoxicity in cancer cells.

As many aromatic polyketides display potent bioactivities, we set out to assess the in vitro cell-based antibacterial, antifungal, and cytotoxic activities of compounds 1−8 using the bacterial strains Staphylococcus aureus (ATCC 25923) and Klebsiella aerogenes (ATCC 13048), the fungal pathogenic strains Candida albicans (ATCC 10231) and Aspergillus fumigatus (ATCC 46645), and the human cancer cell lines PANC-1 and MIA PaCa-2 and A549 as test strains or cell lines. Gentamicin, amphotericin B, and puromycin were included as positive controls in the antibacterial, antifungal, and cytotoxicity assays, respectively. While compounds 1, 2, 6, 7, and 8 displayed negligible activity (IC₅₀ > 100 μg/mL), compounds 3, 4, and 5 showed significant cytotoxic activity against the three cancer cell lines, with the IC₅₀ values ranging from 0.9 to 7.8 μg/mL (Table 1). Compound 5 was also found to inhibit the growth of S. aureus, with a MIC₉₀ value of 1.8 μg/mL.

TABLE 1.

Antibacterial, antifungal, and cytotoxic activities (in human cancer cell lines) of compounds 1−8^a

Compound	Antibacterial (IC90, μg/mL)		Antifungal (IC50, μg/mL)		Cytotoxic (IC50, μg/mL)
	S. aureus	K. aerogenes	C. albicans	A. fumigatus	A549	MIA PaCa-2	PANC-1
1	>100	>100	>100	>100	>100	>100	>100
2	>100	>100	>100	>100	>100	>100	>100
3	71.4	>100	>100	>100	3.0	0.9	5.5
4	>100	>100	>100	>100	5.3	1.1	5.3
5	1.8	>100	>100	>100	2.8	1.2	7.8
6	>100	>100	18.8	>100	50.8	49.3	57.4
7	>100	>100	>100	>100	>100	>100	>100
8	6.7	>100	>100	>100	50.2	46.4	53.0
Gentamicin	0.15	0.10	/	/	/	/	/
Amphotericin B	/	/	0.28	0.12	/	/	/
Puromycin	/	/	/	/	1.04	0.68	1.13

^a/, denotes undetected bioactivity.

Balmoralmycin (compound 1) was reported previously to be an inhibitor of human protein kinase C-α (PKC-α) with an IC₅₀ of 50 μM, though the cell-based anticancer activity was not revealed in that report (30). Our cell-based assays showed that compound 1 did not inhibit the growth of cancer cell lines despite the reported inhibitory effect on PKC-α. While C-olivosyltetrangulol (compound 3) was previously found to exhibit modest antimicrobial activities (44), our assays showed that compound 3 inhibited S. aureus growth with a modest IC₉₀ of 71.4 μg/mL but exhibited notable cytotoxicity against the three human cancer cell lines (IC₅₀ = 0.9 to 5.5 μg/mL). The novel compound 5 also exhibited cytotoxicity against all three cancer cell lines (IC₅₀ = 1.2 to 7.8 μg/mL). As compounds 1, 2, 6, and 7 did not display noticeable cytotoxicity against the cancer cell lines, the lactone ring of compound 5 seems to be an important contributing factor to its bioactivity.

Identification of the balmoralmycin biosynthetic gene cluster (bal BGC).

Although compounds 1 and 2 were isolated over 3 decades ago, the biosynthetic pathways for these two compounds remained unknown. We decided to sequence the genome of Streptomyces sp. P01 to unveil the balmoralmycin BGC and other novel cryptic BGCs. The complete genome of Streptomyces sp. P01 was obtained using the PacBio single-molecule real-time (SMRT) sequencing technology to reveal a linear chromosome of 8,239,033 bp with an average guanine-cytosine (GC) content of 73.4%. Phylogenetic analysis based on the 16S rDNA sequences suggested Streptomyces cyslabdanicus K04-0144 as the closest relative, with a shared nucleotide identity of 99.03% (45). An antiSMASH 5.0 (46) analysis indicated that the genome contains at least 27 BGCs, including BGCs that encode type I and type II PKSs, nonribosomal peptide synthases (NRPS), terpene/terpenoid synthases, and other scaffold-forming biosynthetic enzymes (Table 2).

TABLE 2.

BGCs of Streptomyces sp. P01 predicted by antiSMASH 5.0

BGC	Position from	Position to	Product/type
Cluster 1	557,421	611,003	NRPS
Cluster 2	1,134,310	1,143,378	Melanin
Cluster 3	1,419,537	1,462,424	NRPS
Cluster 4	2,137,189	2,146,241	Ectoine
Cluster 5	2,361,202	2,386,397	Unknown
Cluster 6	2,942,195	2,998,132	NRPS
	2,925,940	2,941,812	T2PKS/spore pigment
Cluster 7	2,999,108	3,059,065	NRPS
Cluster 8	3,207,575	3,287,472	Bafilomycin
Cluster 9	3,636,637	3,889,360	T1PKS
Cluster 10	3,901,076	3,921,444	Isorenieratene
	3,923,121	3,958,038	NRPS-T1PKS
Cluster 11	4,055,622	4,064,129	ɣ-butyrolactone
Cluster 12	4,328,715	4,365,209	T2PKS/balmoralmycin
Cluster 13	4,409,376	4,442,191	Unknown
Cluster 14	4,697,177	4,707,309	Melanin
Cluster 15	5,133,774	5,171,858	T1PKS
	5,171,918	5,191,200	Lanthipeptide
Cluster 16	5,328,697	5,343,365	Hopene
Cluster 17	5,478,575	5,493,801	Unknown
Cluster 18	5,648,944	5,668,281	Unknown
Cluster 19	5,696,139	5,708,735	Siderophore
Cluster 20	5,856,754	5,874,955	Terpene
Cluster 21	5,902,770	5,913,380	RiPP
Cluster 22	6,026,541	6,054,413	Tomaymycin
Cluster 23	6,155,780	6,174,762	ɣ-butyrolactone
Cluster 24	6,235,374	6,245,156	Siderophore
Cluster 25	6,871,887	6,891,271	Albaflavenone
Cluster 26	7,001,270	7,022,559	Lanthipeptide
Cluster 27	7,590,043	7,628,408	NRPS

Within the genome, there are two gene clusters (BGC6 and BGC12) that encode type II PKS systems. Type II PKS systems, which depend on stand-alone acyl carrier proteins (ACP) for shuttling biosynthetic intermediates among polyketide chain-extending and processing enzymes, are well known to be responsible for the biosynthesis of bacterial aromatic polyketides (e.g., angucycline polyketides) (1, 3, 9). While the small type II PKS BGC6 (~5.5 kb) is commonly found in Streptomyces for pigment production, the large type II PKS BGC12 (~43 kb) was found to contain a type II PKS cassette (i.e., balA7-A9) that is typically involved in the biosynthesis of angucycline polyketides (Table 3). BGC12 also shares high similarities with the simocyclinone BGCs (i.e., sim and smc BGCs) as well as with the sugar moiety-synthesizing gene cassette (balB7-B3) and the fatty acid biosynthesis-related genes (Fig. 2), further suggesting that BGC12 could be the gene cluster responsible for the biosynthesis of balmoralmycin (39, 40). First isolated from Streptomyces antibioticus Tü6040 and later from Kitasatospora sp. and Streptomyces sp. NRRL B-24484, simocyclinones are structurally complex secondary metabolites with distinct structural components and potent antibacterial activity (37, 39). The best-studied simocyclinone D8 features a chlorinated aminocoumarin that is linked to the angucyclic aglycone via a tetraene linker and an acetylated d-olivose sugar (31). The four groups of genes (groups A to D) that are responsible for the biosynthesis of structural units A to D (i.e., angucycline aglycone, d-olivose, tetraene linker, and chlorinated aminocoumarin moieties) (Fig. 2), have been identified by Luzhetskyy and coworkers (39). BGC12 seems to share the group-A and group-B genes for the biosynthesis of the angucycline aglycone (A1 to A12) and d-olivose (B1 to B5, B7) with the sim and scm BGCs (Fig. 2) (39), although the sim/scmB6 gene that encodes an acyltransferase for the acylation of the deoxysugar in the simocyclinone pathway was not present in BGC12.The group-C genes in the sim/scm gene clusters are responsible for the biosynthesis of the tetraene linker. Two of the genes (sim/scmC4, C5) that are putatively involved in the carboxylation of the tetraene linker in simocyclinones are missing in BGC12. The group-D genes for aminocoumarin biosynthesis in the sim and smc BGCs are completely missing in BGC12, which is consistent with the absence of an aminocoumarin unit in balmoralmycin. Taken together, the bioinformatic information suggests that BGC12 is likely to produce an aromatic polyketide-derived product that contains an angucycline ring, a d-olivose, and a fatty acid chain that is likely to be different from that of the simocyclinones. This consideration led us to propose that BGC12 (hereafter referred to as bal BGC) is involved in the biosynthesis of compound 1 and the coproduced compounds 2−8.

TABLE 3.

The genes of bal BGC and their annotated functions^a

Gene	AA	Closest homolog	UniProt ID	% Identity/% similarity	Predicted function	Homologs in sim and other angucycline BGCsb
balA7	485	Hydroxylase/dehydrase	A0A1M5TK19	90/93	Flavin monooxygenase	simA7, lanE, sqnF, sprB, may8, bexE
balA4	109	Polyketide cyclase	A0A1M5TKA0	91/96	Polyketide cyclase	simA4, lanF, jadI, sprC, may17, bexF
balA1	420	PKS ketosynthase	A0A1M5TK21	97/98	Polyketide ketosynthase	simA1, lanA, sprD, sqnH, jadA, bexA
balA2	404	Beta-ketoacyl synthase	A0A101ULX5	92/95	Polyketide ketosynthase	simA2, lanB, sim3, sprE, bexB
balA3	89	ACP	A0A1M5TK95	81/86	ACP	simA3, lanC, sqnJ, sprF, bexC
balA6	261	Ketoacyl reductase	A0A117RWC5	97/98	Ketoreductase	simA6, lanD, sprG, sqnK, bexD
balA5	314	Polyketide cyclase	A0A117RWC7	92/95	Cyclase	simA5, lanL, sim6, spr H, sqnL, bexL
balA8	513	Oxygenase	A0A1M5TKE0	84/88	Flavin monooxygenase	simA8, lanM, sim7, sqnM, sprB, bexE
balA9	253	Short-chain dehydrogenase/reductase	A0A1M5TKH0	90/96	Oxygenase/reductase	simA9, lanV, sim8, sprI, sqnM, bexM
balEx	486	MFS domain protein	A0A117RWH8	75/80	Transporter	sqnN, sprJ, saqJ, BexJ
balA10	191	Flavin reductase	A0A1M5TL23	87/92	Flavin reductase	simA10, lanZ4, sim9, saqO, sqnC, urdO
balA11	164	Acyl-CoA carboxylase	A0A101UMQ2	66/70	Unknown
balB7	381	C-glycosyltransferase	A0A117RWK8	88/92	C-glycosyltransferase	simB7, lanGT2, sqnG3, saqGT5, sprGT3
balB1	355	Thymidylyltransferase	A0A1M5TM58	88/93	Thymidylyltransferase	lanG, sqnS2, saqG, sprL
balB4	359	Dehydrogenase	A0A1M5TM89	75/82	Oxidoreductase	simB4, lanT, saqT, sqnS8, sprS
balB2	327	dTDP-glucose 4,6-dehydratase	A0A101ULU7	88/93	NDP-hexose 4,6-dehydratase	lanH14, sqnS3, saqH, sprM
balB5	321	Epimerase	A0A1M5TLG1	79/84	NDP-hexose 4-ketoreductase	simB5, lanR, saqR, sqnS6, sprP
balB3	462	NDP-hexose 2,3-dehydratase	A0A101ULW8	84/88	NDP-hexose 2,3-dehydratase	simB3 lanS, sim20, sqnS7, sprR
balC3	338	3-oxo-ACP synthase	A0A1M5TLT0	85/92	3-oxo-ACP synthase	simX5
balC8	270	Stilbenecarboxylate synthase	A0A101ULZ3	78/84	Type III PKS synthase
balA12	544	Carboxyltransferase	A0A101UM08	91/95	Decarboxylase	simA12, lanP, sqnP, sprU, BexH
balC7	88	ACP	A0A1M5TLZ2	87/94	ACP	simorf1
balC2	418	Ketosynthase	A0A1M5TLW4	92/95	3-oxo-ACP synthase II	simorf3, lanA, sqnH, saqA, sprD
balC1	360	Ketosynthease	A0A1M5TLP8	85/92	3-oxo-ACP synthase II	simorf2
balC6	248	3-oxoacyl-ACP reductase	A0A1M5TLW7	86/93	3-oxo-ACP-reductase	simD4, simJ2, sqnM
balE	319	Histidine kinase	A0A101ULU1	75/82	Histidine kinase
balReg1	220	Response regulator	A0A1M5TLZ1	92/93	Transcriptional regulator	simReg1, sim1, sqnR, sprR3
balF	163	TIR domain-containing protein	A0A0M9YDF9	58/91	Hypothetical protein

^aGenes responsible for the biosynthesis of the angucycline aglycone, deoxysugar and unsaturated fatty acid moieties are shaded in red, brown and green, respectively.

^bSim, simocyclinone BGC; Urd, urdamycin BGC; Lan, landomycin BGC; Sqn, saquayamycin BGC; Spr, saprolmycin BGC; May, mayamycin BGC; Bex, BE-7585A BGC; Jad, jadomycin BGC.

FIG 2.

Comparison of BGC12 (bal BGC) and the two simocyclinone BGCs (sim, smc). The genes are colored based on their predicted functions in the biosynthesis of the angucycline aglycone (red), deoxysugar (brown), unsaturated fatty acid chain (green), and aminocoumarin (yellow).

To confirm the involvement of the bal BGC in the biosynthesis of balmoralmycin and compounds 2−8, we targeted the bal genes for deletion using the CRISPR/Cas9-mediated gene knockout methods that are routinely used in our lab for gene inactivation (47, 48). However, no exo-conjugants could be obtained when Streptomyces sp. P01 was transformed with several Cas9-expressing plasmids. The observation suggested that Cas9 could be toxic to Streptomyces sp. P01. The toxicity caused by Cas9 expression is a rather common phenomenon for Streptomyces strains, based on our own observations and on the observations of others (49, 50). As an alternative, we adopted the CRISPR/dCas9-mediated interference (CRISPRi) approach, considering that dCas9 is generally not known to be toxic to microbial cells because it does not cause double-strand DNA cleavage. Using the dCas9-expressing vector pSET152-dcas9 developed by Lu and coworkers (51), we designed the sgRNA sequences to target balA1, A7, and A8 individually (Tables S3 and S4). The three genes encode a ketosynthase (balA1) and two FAD-dependent monooxygenases (balA7 and A8) that are putatively needed for the formation of the angucycline core. The three knockdown (KD) mutants (balA1_KD, balA7_KD, and balA8_KD) were generated via E. coli-Streptomyces conjugation, with the exo-conjugants being readily observed on the agar plates. DNA sequencing confirmed the integration of the plasmids, including the dCas9 and sgRNA fragments, into the host chromosome via the PhiC31 attachment site. When we cultured the three KD mutants along with the WT strain on PM3 solid medium, we noticed the loss of the red or orange color for the three KD mutant strains compared to the WT and empty-plasmid-containing control strains (Fig. 3A). Consistent with the color change, we found that the production of the colored compounds 1−8 was almost abolished for the balA1_KD strain and was drastically reduced for the balA7_KD and balA8_KD strains (Fig. 3B). It should be noted that one of the limitations of the dCas9-based CRISPRi method is that dCas9 suppresses not only the expression of the targeted gene but also the downstream genes from the same operon. As the bal angucycline core genes (i.e., balA7-A9 cassette) are likely to be clustered in a single operon, as seen for the simA7-A9 cassette in the sim BGC (39), the binding of dCas9 would negatively impact the expression of all of the genes that are transcribed together. Hence, the observed effect on the production of compounds 1 − 8 is likely caused by the suppressed expression of multiple bal genes, rather than a single gene. Despite the limitation of the CRISPRi, the drastic decreases in the production of compounds 1−8 that were observed for the three KD strains and the similarity shared between the bal and sim BGCs support the conclusion that bal BGC is responsible for the biosynthesis of balmoralmycin.

FIG 3.

Gene knockdown supports the involvement of bal BGC in balmoralmycin production. (A) The balA1_KD, balA7_KD, and balA8_KD mutants generated using the CRISPR/dCas9 method exhibited different colony morphology on PM3 agar plates. Control, P01 with an empty CRISPR/dCas9 vector; WT, P01 wild-type. The dark red and orange colors exhibited by the WT and the control plates were not observed for the three mutant strains, likely due to the reduced production of the colored compounds 1 − 8. (B) High-performance liquid chromatography (HPLC) analysis of the organic extract of the strains suggested that the CRISPR/dCas9-assisted knockdown of balA1, A7, or A8 drastically decreased the production of balmoralmycin (compound 1) and the shunt products (compounds 2−8). The HPLC detector was set at 260 nm wavelength.

Biosynthesis of d-olivose and the unsaturated fatty acid.

The C-glycosidic-linked deoxysugar d-olivose is a hallmark of the UTA family of angucyclines. The biosynthesis of d-olivose has been established from the studies of urdamycin and other pathways (1, 39, 52, 53). The genes encoding the d-olivose-synthesizing enzymes are conserved in the bal BGC. Accordingly, a cascade of reactions catalyzed by five enzymes (BalB1 to B5) is likely to transform d-glucose-1-phosphate into the key intermediate NDP-d-olivose. BalB7 is a C-glycosyltransferase that putatively catalyzes the C-glycosylation of the angucycline core to yield the C-glycosylated intermediate compound 3 (Fig. 4) (15).

FIG 4.

The proposed biosynthetic mechanism for balmoralmycin (1) and the shunt products (compounds 2−8).

All UTA-2 type angucyclines contain an unsaturated fatty acid chain that differs in the length of the carbon-chain and the degree of unsaturation. The 10-carbon-long fatty acid chain of simocyclinones is synthesized by a group of standalone enzymes that include ketosynthases (SimKSI, KSII, X5), the acyl carrier protein (SimP), ketoreductase (SimC6), and oxidoreductases (SimC4, SimC5) (39). The balmoralmycin pathway seems to contain a similar group of type II PKS proteins that include BalC1 (=KSI), BalC2 (=KSII), BalC3 (=X5), BalC7 (=P or =ACP), and BalC6. However, the balmoralmycin pathway lacks the two enzymes (SimC4 and SimC5) that are required for the formation of the terminal carboxylic acid group in simocyclinones, but it does possess a type III acyl-CoA ketosynthase (BalC8) that is not seen in the simocyclinone pathway. The shared genes (balC1-C3, C6, C7) suggest that the 2,4-decadienoic acid chain of balmoralmycin could be synthesized in a comparable mechanism via the type II PKS machinery. Unlike simocyclinones, the five-terminal carbons of the 2,4-decadienoic acid chain in balmoralmycin are fully reduced. Because of the lack of an enoylreductase (ER) gene in the bal BGC, the three terminal acetate units of the fatty acid chain in bal are most likely derived from the starter hexanoyl-CoA (54). Based on these considerations, we propose that the biosynthesis of the fatty acid chain is initiated with hexanoyl-CoA synthesized by the type III ketosynthase BalC8 or is recruited from a primary fatty acid synthesis pathway (Fig. 4) (55–58). With hexanoyl-CoA as the starter unit, the ketosynthase BalC3 proceeds to catalyze the condensation between hexanoyl-CoA and another unit of malonyl-CoA. Next, the resulted acyl-CoA is loaded onto the acyl carrier protein BalC7, and BalC1 and BalC2 form a heterodimer to extend polyketide chains as is observed with other KS/CLF type II PKS systems. BalC6 is a homolog of 3-oxoacyl-ACP-reductases and probably reduces the β-keto group of the extending polyketide chain. Similar to the simocyclinone pathway, the balmoralmycin pathway also does not seem to contain a dehydratase (DH) protein that is needed to generate the C=C double bonds in the unsaturated fatty acid chains (39). We speculate that the dehydration reaction is either catalyzed by an uncharacterized or moonlighting enzyme from the Bal pathway or results from the crosstalk between the type II PKS system and a fatty acid synthetase system.

Biosynthesis of the tetracyclic angucycline structure.

Compounds 1 and 2 are likely the end products of the biosynthetic pathway encoded by the bal BGC, based on the observations that compounds 1 and 2 contain all three of the structural units and that the putative immediate precursor of compounds 1 and 2 (i.e., compound 3) were produced in much higher titers than were compounds 4−8, which are likely metabolic shunt products, as we will discuss in more detail below.

Previous studies on urdamycin, landomycin, simocyclinone, and jadomycin have shed light on the key roles played by type II PKS enzymes, cyclases, and oxidoreductases in the biosynthesis of the angucycline ring systems (1, 10, 12, 16, 59). Based on sequence homology, BalA1, BalA2, and BalA3 are the ketoacyl synthase, chain length factor, and ACP components of the “minimal PKS” complex that assembles the linear decaketide precursor (Fig. 4). The decaketide is putatively cyclized and aromatized by the ketoreductase BalA6, aromatase BalA5, and cyclase BalA4 to yield UWM6. UWM6 is a common intermediate in angucycline biosynthesis, and structurally diverse angucyclines can be generated from UWM6 via the actions of oxygenases, reductases, and other tailoring-enzymes (1, 11, 13, 16, 17, 20, 23, 60). Based on the functions of the homologs from the landomycin/gilvocarcin/jadomycin pathways (1, 10, 61), the FAD-dependent monooxygenase BalA8 could catalyze a 2,3-dehydration reaction to generate the C2=C3 double bond. The short-chain dehydrogenase/reductase BalA9, whose homologs can also be found in many angucycline pathways (17, 20), putatively reduces the C-6 keto group to a hydroxyl group. BalA7, another FAD-dependent monooxygenase that is also conserved in the landomycin/jadomycin pathways (17, 19, 60, 62), is predicted to catalyze an oxidation reaction to generate the anthraquinone-type intermediate dehydrorabelomycin (Fig. 4). The subsequent glycosylation of dehydrorabelomycin by BalB7 generates the stable intermediate compound 3 that accumulates in the culture broth. The final installation of the unsaturated fatty acid chain via an esterification reaction completes the biosynthesis of compound 1. Compound 2, which only differs from compound 1 in its lack of the C6-hydroxyl group, is likely to be generated via the intermediate tetrangulol instead of dehydrorabelomycin (Fig. 4).

An interesting finding from our study is the isolation of the shunt product (compound 5) that features a seven-membered lactone ring. A similar seven-membered lactone ring was seen in urdamycin L, a shunt product from the urdamycin pathway (42). The lactone ring of urdamycin L was proposed to be generated via a Baeyer-Villiger oxidation reaction that inserts an oxygen atom between C1 and C12b. Although the Baeyer-Villiger oxidation reaction was originally speculated to be catalyzed by the flavin-dependent monooxygenase UrdM, the direct experimental evidence for the proposal is still lacking (42). We envision that a similar Baeyer-Villiger oxidation is involved in the formation of 5, though we do not know whether the reaction proceeds nonenzymatically or enzymatically and is catalyzed by one of the flavin-dependent monooxygenases (BalA7 or A8) (Fig. 4). The low fermentation titers of compounds 5−8, relative to those of compounds 1−4, seem to favor the argument that the lactone ring of compounds 5−8 are generated via a sluggish nonenzymatic Baeyer-Villiger reaction. Further hydroxylation, oxidation, and dehydration steps led to the formation the anthraquinone-containing intermediate-I and the C-glycosylated intermediate-II. The shunt product compound 8 is likely derived from intermediate-II via nonenzymatic dehydration and lactone hydrolysis whereas compound 4 could also be derived from intermediate-II via a rearrangement of the carbon skeleton, based on studies on the biosynthesis of saquayamycin Z (28, 41). Attachment of the 2,4-decadienoic acid chain to intermediate-II is likely to yield intermediate-III, which undergoes oxidation at the C-4 position to yield compound 5 or dehydration and lactone hydrolysis to yield compound 6 (and compound 7 with further methylation). Considering that the seven-membered lactone is prone to hydrolysis (Fig. S6), the shunt products 6, 7, and 8 are likely artifacts generated during fermentation and compound isolation. Although it is debatable whether compound 5, instead of compound 1 or compound 2, is the real final product of the balmoralmycin pathway, we favor compound 1 or compound 2 as the final product, considering that the production of compound 5 involves spontaneous nonenzymatic steps and that only negligible amounts of compound 5 were produced under most fermentation conditions.

In summary, balmoralmycin (compound 1) and several biosynthetically related coproducts (compounds 2−7) were isolated from a Streptomyces strain in this study. Four of the isolated products (compounds 5−8) are novel compounds that feature a highly oxygenated or fragmented seven-membered lactone ring. The balmoralmycin bal BGC was found to encode a type II PKS system for constructing the angucycline ring system, a set of enzymes for synthesizing the deoxysugar d-olivose, and a hybrid type II/III PKS system for assembling the unsaturated 2,4-decadienoic acid chain. A biosynthetic mechanism was proposed to account for the formation of balmoralmycin and the coproducts. Together with the recently characterized simocyclinone pathway, the elucidated balmoralmycin pathway enriches our understanding of how structural diversity is generated in the UTA-2 angucycline family. The delineation of the balmoralmycin pathway also opens the door for the production of bioactive balmoralmycin derivatives via pathway engineering.

MATERIALS AND METHODS

General.

UV-visible spectra were measured on a Denovix spectrophotometer. NMR experiments were carried out on an Avance NEO 400 MHz or Avance III 600 MHz spectrometer at 298 K. Chemical shifts are expressed in δ (ppm) and referenced to the residual solvent signals. HRESIMS spectra were acquired on a Thermo LTQ XL spectrometer with an electrospray Ionization (ESI) source. Semipreparative high-performance liquid chromatography (HPLC) was conducted using an Agilent 1200 series HPLC-DAD system with an ODS column (Pursuit XRs: diphenyl, 250 mm × 10 mm, 5 μm; Cosmosil: cholester, 250 mm × 10 mm, 5 μm) or using a Shimadzu liquid chromatography system equipped with an ODS column (ACE: C18-HL, 250 mm × 10 mm, 5 μm). Flash column chromatography (CC) was performed using silica gel (230 to 400 mesh, Merck, Darmstadt, Germany). All reagents and materials were purchased from Sigma-Aldrich unless otherwise indicated.

Whole-genome sequencing.

Streptomyces sp. P01 was isolated from a sediment sample collected from Pulau Ubin Quarry Lake, Singapore, at a depth of approximately 25 m (50). The complete genome of P01 was obtained as a linear chromosome of 8,239,033 bp using single molecule real time (SMRT) (Pacific Biosciences, CA, USA) sequencing and the CLC Genomics Workbench (CLC bio, Denmark). The DNA sequence of the balmoralmycin biosynthetic gene clusters (bal BGC) has been deposited into NCBI GenBank (accession number: ON960037). The genes within the bal BGC and their predicted functions are tabulated in Table 3.

Construction of knockdown mutants.

pSET152-dcas9 is a pSET152 derivative harboring the dCas9 expression cassette that includes the ermEp* promoter, dCas9 gene, and fd terminator (51). The primers balA1-sgRNA-F and balA1-sgRNA-R were annealed, and this was followed by Gibson assembly with SpeI-digested pSET152-dcas9-actII-4-NT-S1 (Addgene plasmid number 110185) to replace the 20 nt protospacer (51). The Gibson mix was then transformed into chemically competent TOP10 E. coli. The correct plasmids (i.e., pSET152-dcas9-balA1) were selected via colony PCR and DNA sequencing and subsequently transformed into E. coli ET12567/pUZ8002 for conjugation with P01 spores. Briefly, the P01 spores were heat-shocked at 45°C for 10 min and incubated at 30°C in a shaking incubator for 6 h to allow for germination. The germinated spores were then mixed with E. coli ET12567/pUZ8002 harboring the constructed pSET-dcas9-balA1 at a ratio of 1:3 and spread on mannitol-soy agar supplemented with 20 mM MgCl₂ and 20 mM CaCl₂. After 20 h of incubation at 30°C, the plates were overlaid with 20 μg/mL apramycin and 25 μg/mL nalidixic acid. The correct exconjugants were then checked for the integration of the plasmid via colony PCR to obtain the knockdown mutant balA1_KD. The mutants balA7_KD and balA8_KD were generated following the same protocol.

Fermentation and metabolite profiling. Spores of P01 wild-type or the knockdown strain (balA1_KD, balA7_KD, balA8_KD) were plated on mannitol-soy agar (20 g/L mannitol, 20 g/L pollen soy powder, 20 g/L agar) and incubated for 4 days at 30°C. Starter cultures were prepared by inoculating the mycelia to 50 mL GYM broth (4 g/L glucose, 4 g/L yeast extract, 10 g/L malt extract) in a 250 mL conical flask and shaking for 5 to 7 days at 170 rpm until dispersed. 100 μL starter cultures were used for spreading on PM3 agar (20 g/L oatmeal, 2.5 g/L glycerol, 0.00004 g/L ZnCl₂, 0.0002 g/L FeCl₃·6H₂O, 0.00001 g/L CuCl₂·2H₂O, 0.00001 g/L MnCl₂·4H₂O, 0.00001 g/L Na₂B₄O₇·10H₂O, 0.00001 g/L (NH₄)₆Mo₇O₂₄·4H₂O, 10 g/L agarose) and incubated at 30°C. After 2 weeks of fermentation, the resulting agar plates were mashed and extracted with MeOH. The filtered MeOH extract was concentrated under a vacuum and was redissolved in 500 μL MeOH to obtain the crude sample. HPLC analysis was performed on an Agilent1200 HPLC-DAD system equipped with an ODS column (Pursuit XRs: 250 mm × 4.6 mm, 5 μm) using a linear gradient of CH₃CN in H₂O with 0.1% (vol/vol) formic acid (0 to 5 min, 10% to 20% CH₃CN, vol/vol; 5 to 35 min, 20% to 70% CH₃CN, vol/vol; 35 to 50 min, 70% to 90% CH₃CN, vol/vol; 50 to 60 min, 90% to 100% CH₃CN, vol/vol; 60 to 70 min, 100% CH₃CN) at a flow rate of 1.0 mL/min.

Isolation of compounds 1−8 from Streptomyces sp. P01. Compounds 1−8 were isolated from the P01 wild-type strain, which was cultivated on PM3 agar plates. 10 liters of PM3 agar were autoclaved and poured into 150 mm × 15 mm petri dishes. The P01 spores from glycerol stock at −80°C were diluted and streaked on mannitol-soy agar. After incubation at 30°C for 72 h, single colonies were inoculated into several 250 mL conical flasks containing 50 mL GYM broth. After shaking and incubation at 30°C for 96 h, 100 μL of the starter culture were used to spread on the 10 liters of PM3 agar and were incubated at 30°C for 14 days. The agars were mixed and ultrasonically extracted with MeOH three times. The MeOH extract was combined and concentrated via evaporation under a vacuum to yield the crude extract. The combined extract was then fractionated by silica gel column chromatography and eluted with a step gradient of hexane/EtOAc/MeOH (20/1/0, 5/1/0, 3/1/0, 1/1/0, 0/1/0, 0/20/1, 0/10/1, 0/5/1, 0/1/1, 0/0/1) to obtain 10 fractions (Fr. A–J). Further isolation was guided by HPLC analysis, which showed that compounds 1−8 were mainly distributed in Fr. D–G. Fr. D was purified by repeated semipreparative reversed-phase-HPLC (RP-HPLC) with a high-load ACE ODS column (CH₃CN in H₂O [containing 0.1% formic acid, vol/vol] 88:12, vol/vol; flow rate, 4.7 mL/min) to afford compound 7 (3.2 mg, t_R = 10.5 min), while Fr. E was subjected to Sephadex LH-20 (MeOH) column separation to yield C104 (2, 32.5 mg) and balmoralmycin (1, 85.2 mg). Flash column chromatography of Fr. F over an ODS column gave four subfractions: Fr. F1–F4. Fr. F2 was further purified via semipreparative HPLC using a Cosmosil cholester column (CH₃CN:H₂O [containing 0.1% formic acid, vol/vol] 65:35, vol/vol; flow rate, 3.0 mL/min) to afford compound 3 (42.5 mg, tR = 9.9 min). Accordingly, compound 4 (3.6 mg, t_R = 8.2 min) and compound 5 (5.7 mg, t_R = 11.9 min) were obtained from Fr. F3 and Fr. F4, respectively, by using a diphenyl column (for compound 4, CH₃CN:H₂O [containing 0.1% formic acid, vol/vol] 65:35, vol/vol; flow rate, 3.0 mL/min) or an ACE PFP column (for compound 5, CH₃CN:H₂O [containing 0.1% formic acid, vol/vol] 76:24, vol/vol; flow rate, 3.0 mL/min). Fr. G was purified via semipreparative HPLC with a high-load ACE ODS column (CH₃CN in H₂O [containing 0.1% formic acid, vol/vol] 80:20, vol/vol; flow rate, 4.7 mL/min) to furnish pure compound 6 (5.4 mg, tR = 11.3 min) and crude compound 8. The high-pure compound 8 (3.5 mg, tR = 9.7 min) was achieved via further PFP column separation (CH₃CN:H₂O [containing 0.1% formic acid, vol/vol] 80:20, vol/vol; flow rate, 3.0 mL/min).

Compound 5: yellow-brown solid; UV (MeOH) λ_max (log ε) 240 (4.04), 420 (3.38); ¹H and ¹³C NMR data, see Tables S1 and S2; HRESIMS m/z 633.2347 [M+H]⁺ (calculated for C₃₅H₃₇O₁₁, 633.2330).

Compound 6: yellow-brown solid; UV (MeOH) λ_max (log ε) 230 (3.97), 260 (4.05), 438 (3.40); ¹H and ¹³C NMR data, see Tables S1 and S2; HRESIMS m/z 619.2554 [M+H]⁺ (calculated for C₃₅H₃₉O₁₀, 620.2616).

Compound 7: yellow-brown solid; UV (MeOH) λ_max (log ε) 240 (3.98), 264 (3.95), 422 (3.38); ¹H and ¹³C NMR data, see Tables S1 and S2; HRESIMS m/z 633.2707 [M+H]⁺ (calculated for C₃₆H₄₁O₁₀, 633.2694).

Compound 8: yellow-brown solid; UV (MeOH) λ_max (log ε) 230 (4.03), 258 (3.91), 440 (3.39); ¹H and ¹³C NMR data, see Tables S1 and S2; HRESIMS m/z 469.1506 [M+H]⁺ (calculated for C₂₅H₂₅O₉, 469.1493).

Antimicrobial and cytotoxic assays of compounds 1−8. Antimicrobial activity testing was done on four test pathogens, including two bacteria Staphylococcus aureus (ATCC 25923) and Klebsiella aerogenes (ATCC 13048) as representatives of Gram-positive and Gram-negative bacteria, respectively, and two fungal pathogens Candida albicans (ATCC 10231) and Aspergillus fumigatus (ATCC 46645). Cytotoxic testing was performed on three cancer cell lines: the human pancreatic cancer cells PANC-1 and MIA PaCa-2 and the human lung cancer cells A549. Both antimicrobial and cytotoxic dose-response testing were performed in triplicate in an eight-point, two-fold serial dilution assay with a starting concentration of 100 μg/mL of each compound, following the procedures described in our earlier study (63).