Characterization of the Amicetin Biosynthesis Gene Cluster from Streptomyces vinaceusdrappus NRRL 2363 Implicates Two Alternative Strategies for Amide Bond Formation

Abstract

Amicetin, an antibacterial and antiviral agent, belongs to a group of disaccharide nucleoside antibiotics featuring an α-(1→4)-glycoside bond in the disaccharide moiety. In this study, the amicetin biosynthesis gene cluster was cloned from Streptomyces vinaceusdrappus NRRL 2363 and localized on a 37-kb contiguous DNA region. Heterologous expression of the amicetin biosynthesis gene cluster in Streptomyces lividans TK64 resulted in the production of amicetin and its analogues, thereby confirming the identity of the ami gene cluster. In silico sequence analysis revealed that 21 genes were putatively involved in amicetin biosynthesis, including 3 for regulation and transportation, 10 for disaccharide biosynthesis, and 8 for the formation of the amicetin skeleton by the linkage of cytosine, p-aminobenzoic acid (PABA), and the terminal (+)-α-methylserine moieties. The inactivation of the benzoate coenzyme A (benzoate-CoA) ligase gene amiL and the N-acetyltransferase gene amiF led to two mutants that accumulated the same two compounds, cytosamine and 4-acetamido-3-hydroxybenzoic acid. These data indicated that AmiF functioned as an amide synthethase to link cytosine and PABA. The inactivation of amiR, encoding an acyl-CoA-acyl carrier protein transacylase, resulted in the production of plicacetin and norplicacetin, indicating AmiR to be responsible for attachment of the terminal methylserine moiety to form another amide bond. These findings implicated two alternative strategies for amide bond formation in amicetin biosynthesis.

INTRODUCTION

Nucleoside antibiotics are a large family of microbial secondary metabolites exhibiting potent and diverse bioactive properties (36). Amicetin (compound 1 in Fig. 1) is a disaccharide pyrimidine nucleoside antibiotic produced by Streptomyces vinaceusdrappus and Streptomyces fasciculatis (22). Amicetin has been known for its activity against a number of both Gram-negative and Gram-positive bacteria, particularly Mycobacterium tuberculosis (16, 33), and was also found to have effects against herpesvirus 1 and poliovirus (5). Amicetin has been described as a universal antibiotic, exerting effects against microorganisms of different evolutionary origins, including archaea, bacteria, and eukarya (44), by functioning as a peptidyl transferase inhibitor to block protein biosynthesis (41). Although the binding of amicetin to a conserved structural motif of 23S rRNAs has been investigated by using site-directed mutations, chemical footprinting, and nuclear magnetic resonance (NMR) structure and molecular modeling studies (18, 44), the exact working mechanism for amicetin remains elusive.

Fig 1.

Chemical structures of amicetin (compound 1), bamicetin (compound 2), plicacetin (compound 3), and norplicacetin (compound 4). Me, methyl.

The chemical structure of amicetin was established in 1962 after extensive studies of its hydrolytic products (22, 56) and was finally confirmed by crystallization studies (55). From various actinobacterial strains, a group of 10 amicetin analogues had been isolated, including bamicetin (compound 2) and plicacetin (compound 3) (31), norplicacetin (compound 4) (21), oxamicetin (42), SF2457 (37), oxyplicacetin (11), and cytosaminomycins A to D (30). The presence of an α-(1→4)-glycoside bond between amosamine and amicetose is the most characteristic feature of the amicetin group antibiotics. A p-aminobenzoic acid (PABA) moiety is present in amicetin. PABA is an important precursor of folates in primary metabolism and is also incorporated into many secondary metabolites (61). PABA has been shown to serve as the starter unit initiating the biosynthesis of aureothin (32) and FR-008/candicidin (7) and contributes to the skeleton of chloramphenicol (3). PABA moieties in these secondary metabolites are derived from chorismate via two enzymatic steps catalyzed by a PABA synthase and a 4-amino-4-deoxychorismate (ADC) lyase (6, 32, 61). The primary amino group of cytosine is linked by PABA, to which a (+)-α-methylserine residue is attached, forming two amide bonds in the amicetin structure.

The biosynthesis gene clusters for a number of pyrimidine nucleoside antibiotics have been elucidated, including nikkomycin (4, 46), polyoxin (8), blasticidin S (13), A-500359s and A-503803s (25), caprazamycin (38), lipisidomycin (39) and its analogue A-90289 (24), pacidamycins (50, 60), muraymycin (12), and tunicamycin (9). However, biosynthesis studies for the amicetin group of disaccharide nucleoside antibiotics are not yet available. Herein we report the cloning and characterization of the amicetin biosynthesis gene cluster from S. vinaceusdrappus NRRL 2363. Heterologous expression of the ami gene cluster in Streptomyces lividans TK64 resulted in the production of amicetin and its analogues. The inactivation of 4 biosynthesis genes (amiI, amiF, amiL, and amiR) in the ami gene cluster led to mutants lacking amicetin production or accumulating biosynthesis intermediates, which provided experimental evidence for the functions of these genes in amicetin biosynthesis. This work paves the way to further study the unusual enzymology of amicetin biosynthesis, including the unusual retaining of glycosylation and the biochemical logic of forming two amide bonds.

MATERIALS AND METHODS

Bacterial strains, plasmids, and reagents.

Bacterial strains and plasmids used and constructed in this study are listed in Table S1 in the supplemental material. The strain S. vinaceusdrappus NRRL 2363 was obtained from the National Center for Agricultural Utilization Research culture collection. Streptomyces strains were grown at 28°C, either in ISP2 medium (40) for growth and sporulation or in production medium [2.5% glucose, 0.7% soybean powder, 0.25% yeast extract, 0.5% (NH₄)₂SO₄, 0.4% NaCl, 0.04% KH₂PO₄, 0.8% CaCO₃] for fermentation. Chemicals, enzymes, and other molecular biological reagents were purchased from standard commercial sources and used according to the manufacturers' recommendations.

DNA isolation, manipulation, and sequencing.

DNA isolation from and manipulation in Escherichia coli and Streptomyces were carried out according to standard procedures (40, 53). Primers were synthesized at the Shanghai Invitrogen Biotech Co., Ltd. PCR amplifications were carried out on an authorized thermal cycler (Eppendorf AG). DNA sequencing was performed at the Invitrogen Biotech Co., Ltd. (Guangzhou, People's Republic of China), and the Chinese National Genome Center (Shanghai).

Genomic library construction and screening.

A pOJ446-based cosmid library of S. vinaceusdrappus NRRL 2363 genomic DNA was constructed using a Gigapack III XL packaging kit according to the manufacturer's specifications. For screening of the library, a ca. 0.55-kb conserved nucleoside diphosphate (NDP)-glucose 4,6-dehydratase gene fragment was obtained via a PCR approach using the following pair of degenerate primers: NGDH-F (5′-CSGGSGSSGCSGGSTTCATCGG-3′) and NGDH-R (5′-GGGWRCTGGYRSGGSCCGTAGTTG-3′) (17). Alternatively, a ca. 1.1-kb DNA fragment of a PABA synthase gene homologue was amplified using the degenerated primer pair PAPB-F (5′-GGGVGTSCAGTTCCMMCCSGAGTC-3′) and PABA-R (5′-CAGGTCGACGATCATSAGGTTCTCGGC-3′). The two PCR products were cloned, sequenced, and subsequently used as digoxigenin (DIG)-labeled probes for Southern blot colony hybridization (52).

Sequence analysis.

The sequence of the ami gene cluster was determined, and the open reading frames (ORFs) were deduced from the sequence by using the FramePlot 4.0 beta program. The corresponding deduced proteins were compared with known proteins by BLAST methods.

Construction of targeting vectors.

The subcloning strategy for pCSG3104 is depicted in Fig. S1 and the resulting subclones are summarized in Table S1 in the supplemental material. The gene inactivation experiments with S. vinaceusdrappus NRRL 2363 were carried out using lambda Red-mediated gene replacements according to standard procedures (29). Detailed procedures for disrupting and complementing individual ami genes are described in the supplemental material (see Fig. S1 to S8).

Heterologous production of amicetin and its analogues in S. lividans TK64.

The plasmid pCSG3104 was introduced into S. lividans TK64 by protoplast transformation (40). A single transformant was inoculated into 25 ml production medium [2.5% glucose, 0.7% soybean powder, 0.25% yeast extract, 0.5% (NH₄)₂SO₄, 0.4% NaCl, 0.04% KH₂PO₄, 0.8% CaCO₃] for 5 to 7 days, and production was monitored by high-performance liquid chromatography (HPLC). S. lividans TK64 transformed with the cosmid vector pOJ446 was used as a control.

Isolation and structural elucidation of accumulated novel analogues.

For the isolation of amicetin and accumulated analogues, each 400-ml culture was centrifuged at 4°C and 4,000 rpm for 10 min. After removal of the cell pellets, the resulting supernatant was extracted four to five times with 1-butanol at about one-third of the volume of the supernatant. The extracts were concentrated under vacuum conditions and dissolved in appropriate volumes of methanol for HPLC or HPLC-mass spectrometry (MS) analysis. HPLC analysis was carried out on a Luna C₁₈ reversed-phase column (150 by 4.6 mm; particle size, 5 μm [Phenomenex]) with UV detection at 254 nm. The solvent system consisted of solvent A, 10% CH₃CN in water supplemented with 0.08% trifluoroacetic acid (TFA), and solvent B, 90% CH₃CN in water. The program was as follows: 0% to 40% solvent B (0 to 15 min), 40% to 80% solvent B (15 to 18 min), 80% to 0% solvent B (18 to 19 min), and 0% solvent B (19 to 25 min) at a flow rate of 1 ml/min. The targeted compounds from large-scale fermentation of corresponding mutants were isolated and purified by various chromatographic methods, including silica gel chromatography, thin-layer chromatography, medium-pressure column chromatography, reversed-phase silica gel column chromatography, and Sephadex LH-20 gel column chromatography (see the supplemental material for details). The purified compounds were analyzed by ¹H and ¹³C NMR spectroscopy.

Nucleotide sequence accession number.

The sequence of the ami gene cluster was deposited in GenBank under accession number HM748814.

RESULTS AND DISCUSSION

Cloning, identification, and heterologous expression of the amicetin biosynthesis gene cluster.

Structurally, amicetin contains two deoxysugar units and a PABA moiety. On the basis of these structural features, two complementary strategies were employed to screen for a putative amicetin biosynthesis gene cluster in a pOJ446-based genomic library from S. vinaceusdrappus NRRL 2363. Using a pair of previously reported degenerate primers targeting the conserved NDP-glucose 4,6-dehydratase gene, we were able to amplify a 550-bp PCR fragment, which was proven to be related to a dTDP-glucose 4,6-dehydratase gene (accession number YP_480774) from Frankia sp. strain CcI3. Alternatively, based on the universally conserved PABA synthase motifs, we designed a pair of degenerate primers (PABA-F and PABA-R) capable of amplifying a 1.1-kb fragment, which was sequenced and determined to be a homologue of aurG, a PABA synthase gene from the aureothin pathway in Streptomyces thioluteus (32). Among 1,920 clones in the genomic library from S. vinaceusdrappus NRRL 2363, 9 were positive with the NDP-glucose 4,6-dehydratase gene probe and 6 were positive with the PABA probe. A common positive clone, designated pCSG3104, was identified. Subsequently, pCSG3104 and pOJ446 were introduced into S. lividans TK64 by protoplast transformation and the resulting transformants were cultivated for 5 days. Interestingly, 3 metabolites were detected in the butanol extract of S. lividans TK64::pCSG3104 (Fig. 2A, trace II); however, none were present in the control, S. lividans TK64::pOJ446 (Fig. 2A, trace I). Upon HPLC-MS analyses, these metabolites were identified as amicetin (m/z 617.2 [M − H]⁻), norplicacetin (m/z 501.2 [M − H]⁻), and plicacetin (m/z 515.2 [M − H]⁻) by comparisons of their UV/visible spectra, retention times, and molecular masses with authentic standards (Fig. 2A, traces III and IV).

Fig 2.

Heterologous expression of the ami gene cluster in S. lividans TK64 and preliminary determination of the boundary. (A) HPLC analyses of the heterologous production of amicetin and its analogues in S. lividans TK64::pOJ446 (I) and TK64::pCSG3104 (II) in comparison to an amicetin (compound 1) standard (III) and plicacetin (compound 3) and norplicacetin (compound 4) standards (IV). (B) HPLC analyses of the metabolite profiles of wild type S. vinaceusdrappus NRRL 2363 (V), the Δorf(−1) mutant AM1001 (VI), and the Δorf1 mutant AM1011 (VII). mAU, milli-absorbance units.

The insert in pCSG3104 was sequenced by a shotgun approach, yielding a 37,337-bp contiguous DNA sequence with an overall G+C content of 68.5% (GenBank accession number HM748814). Bioinformatic analysis of the sequence revealed the presence of 26 ORFs (Fig. 3), and the deduced functions of individual ORFs are summarized in Table 1. The four ORFs, orf(−4) to orf(−1), located upstream of amiA encode proteins whose functions are not relevant to amicetin biosynthesis (Table 1) and are probably outside the ami gene cluster. Indeed, the inactivation of orf(−1) (encoding an Mg- or Mn-dependent protein phosphatase) by replacement of an internal fragment with the aac(3)IV gene cassette (see Fig. S2 in the supplemental material) led to a mutant, AM1001, that produced amicetin at a yield comparable to that from the wild-type strain (Fig. 2B, traces V and VI). This indicated that orf(−1) was not involved in amicetin biosynthesis. The inactivation of the orf1 gene (see Fig. S3 in the supplemental material), encoding a GCN5-related N-acetyltransferase (GNAT) family protein, led to a mutant, AM1011, displaying no effects on amicetin production (Fig. 2B, trace VII), excluding a functional role for orf1 in amicetin biosynthesis. Based on these experimental results, we preliminarily outlined the ami gene cluster in S. vinaceusdrappus NRRL 2363 to contain 21 genes, from amiA to amiU, spanning a DNA region of approximate 27 kb. Interestingly, these 21 genes are transcribed in the same direction (Fig. 3).

Fig 3.

Restriction map of the DNA insert in pCSG3104 and organization of the ami gene cluster for amicetin biosynthesis. Restriction sites: E, EcoRI site; B, BamHI site; X*, XbaI site originating from the cosmid vector pOJ446; S*, SpeI site originating from the cosmid vector pOJ446. Proposed functions of individual ORFs are identified and are summarized in Table 1.

Table 1.

Deduced ORFs and their predicted functions in the ami gene cluster

Gene	Product size (aa)a	Proposed function of gene productb	Closest protein homologue (origin)	Accession no.	% similarity/% identity
orf(−4)	137	Putative NAD(P)H-dependent FMN reductase	SSQG_07138 (Streptomyces viridochromogenes DSM 4073)	EFL36620	90/84
orf(−3)	346	Putative transcription antitermination regulator	DSMT0073 (Streptomyces ambofaciens)	CAK50916	90/84
orf(−2)	173	Putative regulator	SAMT0056 (Streptomyces ambofaciens ATCC 23877)	CAI78259	78/70
orf(−1)	396	Mg- or Mn-dependent protein phosphatase	SSLG_00983 (Streptomyces sp. strain SPB78)	EFK98924	83/75
amiA	264	Putative 4-amino-4-deoxychorismate lyase	PabC (Streptomyces sp. FR-008)	AAQ82550	61/49
amiB	372	NDP-sugar aminotransferase	SULAZ_0355 (Sulfurihydrogenibium azorense)	YP_002728349	53/35
amiC	458	NDP-hexose 2,3-dehydratase	SchS2 (Streptomyces sp. strain SCC 2136)	CAF31368	51/36
amiD	322	NDP-hexose 3-ketoreductase	SpnN (Saccharopolyspora spinosa)	AAG23275	44/29
amiE	257	Glucose-1-phosphate thymidylyltransferase	Thal_1465 (Thermocrinis albus DSM 14484)	YP_003474221	55/37
amiF	468	GCN5-related N-acetyltransferase	Dbac_1649 (Desulfomicrobium baculatum DSM 4028)	YP_003158157	50/32
amiG	495	Glycosyltransferase, group 1	DAMO_2983 (NC10 bacterium)	CBE70056	46/30
amiH	249	NDP-aminohexose N-dimethyltransferase	Veis_4849 (Verminephrobacter eiseniae)	YP_999558	51/34
amiI	182	(Deoxy)cytidine deoxyribosyltransferase	MilB (“Streptomyces rimofaciens”)	ACA14349	46/27
amiJ	403	Cytosylglucuronic acid synthase/glycosyltransferase	BlsD (Streptomyces griseochromogenes)	AAP03118	51/39
amiK	315	NDP-hexose 4-ketoreductase	PokS6 (Streptomyces diastatochromogenes)	ACN64824	44/35
amiL	498	Benzoate-CoA synthase	SBI_01640 (Streptomyces bingchenggensis)	ADI04761	50/38
amiM	677	4-Amino-4-deoxychorismate synthase	PapA (Streptomyces venezuelae)	BAD21140	61/49
amiN	439	NDP-4-keto-6-deoxyglucose-3-dehydratase	Gra-orf23 (Streptomyces violaceoruber)	CAA09644	86/75
amiO	559	ABC transporter	Kfla_6094 (Kribbella flavida DSM 17836)	YP_003383896	83/74
amiP	212	TetR family transcriptional regulator	HMPREF9621_00706 (Propionibacterium acnes)	EFS75074	75/59
amiQ	397	Putative major facilitator superfamily transporter	Svir_26710 (Saccharomonospora viridis DSM 43017)	YP_003134482	76/65
amiR	299	Malonyl CoA-acyl carrier protein transacylase	ECA2695 (Pectobacterium atrosepticum)	YP_050786	48/29
amiS	452	Putative glycine/serine hydroxymethyltransferase	GlyA (Azospirillum sp. strain B510)	YP_003450671	64/50
amiT	821	Nonribosomal peptide synthetase (A-PCP)	McnA (“Microcystis” sp. strain NIVA-CYA 172/5)	AAZ03550	54/36
amiU	354	dTDP-glucose 4,6-dehydratase	AaLAA1DRAFT_2011 (Alicyclobacillus acidocaldarius LAA1)	EED06826	65/49
orf1	179	GCN5-related N-acetyltransferase (GNAT) family protein	Sros_7636 (Streptosporangium roseum DSM 43021)	ACZ90310	75/65

^aaa, amino acids.

^bFMN, flavin mononucleotide; A-PCP, adenylation domain and peptidyl carrier protein.

Genes involved in transport and regulation.

Two putative transporter-encoding genes, amiO and amiQ, are identified in the ami gene cluster. AmiO shows 73% sequence identity to a putative ABC transporter, Kfla_6094, from Kribbella flavida DSM 17836. AmiO has two typical ABC-cobalt-CbiO domains, one at the N terminus and one at the C terminus. Each domain contains motifs characteristic of an ATP binding site, Walker A and Walker B motifs, and Q-loop, D-loop, and H-loop motifs (27). AmiQ shares 64% sequence identity with an arabinose efflux permease family protein (ACU97655), belonging to the major facilitator superfamily (MFS), from Saccharomonospora viridis DSM 43017. MFS proteins constitute a large and diverse group of secondary transporters that facilitate the transport of a variety of substrates across cytoplasmic or internal membranes (43). AmiP exhibits 59% sequence identity to a putative TetR family transcriptional regulator (accession no. EFS75074) from Propionibacterium acnes. Functions regulated by TetR family members have been clustered into 10 groups, of which the most frequent function is the regulation of efflux pumps and transporters involved in antibiotic resistance and tolerance to toxic chemicals (51). Thus, AmiP may play a regulatory role in amicetin biosynthesis.

Genes involved in deoxysugar biosynthesis.

Eight genes, amiB to amiE, amiH, amiK, amiN, and amiU, are putatively involved in the biosynthesis of two deoxysugar moieties, d-amosamine and d-amicetose. The sequence alignment of the products of these genes with known deoxysugar biosynthesis enzymes leads to a proposal of the biosynthesis pathway for both sugar units (Table 1; Fig. 4A). The d-glucose-1-phosphate (compound 5) thymidylyltransferase AmiE may catalyze the formation of TDP-d-glucose (compound 6), which is further converted to TDP-4,6-dideoxy-4-keto-d-glucose (compound 7) by the TDP-glucose 4,6-dehydratase AmiU. The biosynthesis of amosamine and amicetose may branch at a common intermediate, compound 7, which is universally found in many deoxysugar pathways (57). The amiB gene, encoding a pyridoxal phosphate (PLP)-dependent aminotransferase, may be responsible for the introduction of an amino group at the C-4 position of compound 7 to form TDP-d-viosamine (compound 8). AmiH carries a type 12 methyltransferase domain (pfam08242) and is probably involved in the dimethylation of compound 8 to form TDP-d-amosamine (compound 9).

Fig 4.

Proposed amicetin biosynthesis pathway. (A) Proposed biosynthesis of two deoxysugars. (B) Two proposed alternative strategies for amide bond formation in amicetin biosynthesis. The dashed-line arrows denote biosynthesis steps predicted solely from bioinformatic analyses, and the solid-line arrows represent biosynthesis steps based on data from in vivo gene inactivation experiments. The orders for sugar attachments are deduced from the accumulation of compound 17 in the ΔamiF mutant AM1003 and the ΔamiL mutant AM1007 (Fig. 5). Sugar attachments can also occur after the formation of the two amide bonds. X, domain with unknown function; A, adenylation domain; PCP, peptidyl carrier protein.

Four enzymes, AmiC, AmiD, AmiN, and AmiK, are expected to catalyze the conversion of compound 7 to TDP-d-amicetose (compound 13) (Fig. 4A). The C-2 deoxygenation converting compound 7 to TDP-3,4-diketo-2,6-dideoxy-d-glucose (compound 10) is putatively catalyzed by AmiC, which shows the highest homology to SchS2 (36% identity), a predicted NDP-hexose 2,3-dehydratase in the biosynthesis of amicetose in angucyclines Sch 47554 and Sch 47555 (2). AmiD displays 29% identity to SpnN, a characterized TDP-hexose-3-ketoreductase which catalyzes the formation of TDP-4-keto-2,6-dideoxy-d-glucose in the TDP-d-forosamine biosynthesis pathway for spinosyn in Saccharopolyspora spinosa (35). Thus, AmiD may reduce the C-3 keto in compound 10 to form the equatorial hydroxyl group in TDP-4-keto-2,6-dideoxy-d-glucose (compound 11). The subsequent C-3 deoxygenation of compound 11, leading to compound 12, is probably performed by AmiN in a way similar to that of its homologue SpnQ (72% identity), a mechanistically characterized TDP-4-keto-6-deoxyglucose-3-dehydratase in the spinosyn pathway (34). Finally, the C-4 reduction on compound 12 to form TDP-d-amicetose (compound 13) is putatively catalyzed by AmiK, which is homologous to a predicted NDP-hexose 4-ketoreductase, PokS6 (35% identity), from the polyketomycin pathway (15).

Genes involved in glycosylation.

Genes for two putative glycosyltransferases (GTs), AmiJ and AmiG, are found in the ami gene cluster. By using a BLAST search of the GenBank database, AmiJ is found to exhibit 39% sequence identity to the cytosylglucuronic acid synthase BlsD from the blasticidin S pathway, which catalyzes the attachment of glucuronic acid to cytosine to form cytosylglucuronic acid (13). Thus, AmiJ is thought to catalyze the coupling of amicetose and cytosine to form an N-glycosidic bond in the amicetin pathway. AmiG is closely related to the GT1 family of GTs, with the highest similarity (30% identity) to a putative GT from an anaerobic bacterium (20). Distinct from most natural-product GTs in secondary metabolic pathways, which often catalyze glycosyl group transfer with inversion of the anomeric stereochemistry with respect to the donor sugar (14), AmiG may be responsible for the formation of the α-(1→4)-glycosidic bond between amosamine and amicetose, proceeding with retention of the anomeric configuration of amosamine. Consistent with this unique property, AmiG displays little similarity to a number of natural-product inverting GTs that have been well studied by in vivo inactivation or by in vitro biochemical characterization (57, 58).

Genes involved in cytosine and PABA biosynthesis.

Similar to the biosynthesis of blasticidin S and mildiomycin (13, 45), the biosynthesis of amicetin may originate from CMP (compound 14) (Fig. 4B). AmiI shows sequence similarity to various nucleoside 2-deoxyribosyltransferases that catalyze the cleavage of the glycosidic bond of 2-deoxyribonucleosides, among which BlsM (34% identity) from the blasticidin S pathway and MilB (27% identity) from the mildiomycin pathway have been functionally characterized (28, 45). The amiI gene was inactivated to probe its functional role in amicetin biosynthesis, resulting in the ΔamiI mutant AM1006. AM1006 lost the ability to produce amicetin or detectable intermediates (Fig. 5A, trace I). The production of amicetin could be partially restored by in trans complementation (see Fig. S8 in the supplemental material). These data supported the possibility that AmiI might participate in an early step of amicetin biosynthesis, providing free cytosine (compound 15) from hydrolysis of CMP (Fig. 4B), in analogy to the well-studied enzymes BlsM and MilB.

Fig 5.

Production analyses of ami gene inactivation mutants. (A) HPLC analyses of the metabolite profiles of the ΔamiI mutant AM1006 (I), the ΔamiL mutant AM1007 (II), the ΔamiF mutant AM1003 (III), the ΔamiR mutant AM1009 (IV), and the wild type S. vinaceusdrappus NRRL 2363 (V). (B) Chemical structures of compounds isolated from the mutant AM1007.

Four genes in the ami gene cluster, amiM, amiA, amiL, and amiF, may encode enzymes responsible for the incorporation of the compound 20 moiety into amicetin. AmiM contains both type I glutamine amidotransferase (cd01653) and chorismate binding (pfam00425) motifs and exhibits high levels of sequence identity to various putative PABA synthases, such as AurG in aureothin biosynthesis (32) and PabAB for chloramphenicol in S. venezuelae (47% identity), which was functionally characterized (3). AmiA encodes an ADC lyase with 48% identity to PabC-1 for FR-008/candicidin biosynthesis in Streptomyces sp. strain FR-008 (61). Thus, AmiM may act to transfer the amide nitrogen of glutamine to chorismate (compound 18), forming ADC (compound 19), and subsequently, AmiA carries out the elimination of pyruvate from ADC to produce PABA (compound 20), which is further modified by AmiL, a benzoate coenzyme A (benzoate-CoA) ligase, to form PABA-CoA (compound 21) (Fig. 4B). Bioinformatic analysis shows that the amiF gene encodes a protein of 468 amino acids. Interestingly, the N terminus of AmiF contains a conserved NAT_SF (cd04301) domain for N-acetyltransferase superfamily enzymes. However, the N terminus of AmiF exhibits similarities to only two GCN5-related N-acetyltransferases, from Segniliparus rotundus DSM 44985 (ACU8974; 32% identity) and Desulfomicrobium baculatum DSM 4028 (ADG97057; 34% identity). Given the capability of N-acetyltransferases to catalyze the transfer of the acetyl group from acetyl-CoA to a primary amine (19), we hypothesized that AmiF may be responsible for the formation of an amide bond between cytosine and PABA (Fig. 4B).

To probe the specific functions of amiF and amiL in amicetin biosynthesis, both genes were inactivated, yielding two mutants, AM1003 (ΔamiF) and AM1007 (ΔamiL). HPLC analyses of the butanol extracts from both mutants showed that they lost amicetin-producing ability and accumulated the same two products (Fig. 5A, traces II and III), and in trans complementation of AM1003 (ΔamiF) and AM1007 (ΔamiL) restored amicetin production (see Fig. S8 in the supplemental material). Upon large-scale (9-liter) fermentation of AM1003 and subsequent product isolation efforts, one product was revealed to be cytosamine (compound 17) by comparison of its ¹H NMR data (see Table S3 in the supplemental material) with those reported previously (42). Another product was elucidated as 4-acetamido-3-hydroxybenzoic acid (compound 25) (Fig. 5B). Careful analyses of the ¹H ¹³C distortionless enhancement by polarization transfer (DEPT)-NMR spectra of compound 25 (see Table S3 in the supplemental material) indicated that there was a 4-amino-3-hydroxybenzoic acid moiety [δ_H 8.72 (d, H-5), 8.30 (s, H-2), 8.08 (d, J = 8.5 Hz, H-6); δ_C 169.3 (s, C-7), 148.2 (s, C-3), 132.5 (s, C-4), 128.2 (s, C-1), 122.2 (d, C-6), 121.1 (d, C-5), 117.4 (d, C-2)]. Additionally, an acetyl moiety [δ_H 2.34 (s, Me-2′); δ_C 170.2 (s, C-1′), 24.5 (q, Me-2′)] was present in compound 25. The heteronuclear multiple-bond correlation (HMBC) data from the proton [δ_H 10.02 (s)] of the amino group to C-3, C-5, and C-1′ of the acetyl moiety indicated that the C-4 amino group was acetylated. Based on these NMR data, compound 25 was elucidated as 4-acetamido-3-hydroxybenzoic acid (Fig. 5B). However, it is hard to explain the role of compound 25 in the amicetin biosynthesis pathway. We hypothesized that PABA was accumulated as a biosynthesis intermediate in both AM1003 (ΔamiF) and AM1007 (ΔamiL). PABA was converted to compound 25 by a hydroxylase and an N-acetyltransferase encoded outside the ami gene cluster. This hypothesis could be inferred from the biotransformation of exogenously supplemented PABA (compound 20) into compound 25 by the ΔamiI mutant AM1006 (see Fig. S9 in the supplemental material). Given the predicted role of AmiL as a PABA-CoA synthase, production of cytosamine and compound 25 in both AM1003 and AM1007 indicated that AmiF probably functioned to coordinate amide bond formation, linking PABA-CoA (compound 21) and cytosamine (compound 17) (Fig. 4B). Additionally, these data indicated that the glycosyltransferase AmiG was capable of utilizing amicetosyl-cytosine (compound 16) as a substrate (Fig. 4B). However, the timing for AmiG in the compound 1 biosynthesis pathway was still not clear; AmiG glycosylation could also occur as a tailoring step after amide bond formation.

Genes involved in the biosynthesis and incorporation of (+)-α-d-methylserine.

Three genes, amiS, amiT, and amiR, were predicted to be involved in the biosynthesis and incorporation of (+)-α-d-methylserine into amicetin. AmiS shows high similarity to glycine/serine hydroxymethyltransferase (SHMT), belonging to the PLP-dependent aspartate aminotransferase superfamily (fold I), which catalyzes the transfer of the hydroxymethyl group of N⁵,N¹⁰-methylene tetrahydrofolate to glycine, resulting in the formation of serine and tetrahydrofolate (54). Therefore, AmiS may catalyze the conversion of d-alanine (compound 23) to α-d-methylserine (compound 24) in the amicetin pathway (Fig. 4B). Subsequently, α-d-methylserine may be activated by AmiT, a protein displaying significant similarity to nonribosomal peptide synthetases (Fig. 4B). Bioinformatic analysis shows that AmiT contains an adenylation domain (A domain), with predicted substrate specificity for serine, and a putative thiolation domain (T domain). Interestingly, no domains are predicted in the first 200 amino acid residues at the N terminus of AmiT, and a required condensation domain (C domain) is missing in AmiT. AmiR exhibits weak similarity (less than 30% identity) to putative acyl-CoA-acyl carrier protein transacylases, which transfer an acyl group to an acyl carrier protein, resulting in the formation of a carbon-sulfur bond. Recently, this kind of enzyme has also been shown to be capable of forming a carbon-nitrogen bond (1).

To determine the function of AmiR in amicetin biosynthesis, the amiR gene was inactivated to give the double-crossover mutant AM1009. The mutant was devoid of amicetin production (Fig. 5A, trace IV), which could be restored by in trans complemention (see Fig. S8 in the supplemental material). Two compounds distinct from amicetin were produced by the ΔamiR mutant AM1009 (Fig. 5A, trace IV). The two compounds were then isolated and determined to be plicacetin (compound 3) and norplicacetin (compound 4) by ¹H ¹³C correlation spectroscopy (COSY), heteronuclear single-quantum correlation (HSQC), and HMBC NMR characterizations (see Tables S3 and S4 in the supplemental material). Comparing the ¹H and ¹³C NMR spectra of compounds 3 and 4 with those of amicetin revealed that the signals of the (+)-α-methyl serine moiety were absent in the spectra of both compounds. The only difference between compounds 3 and 4 was that the double N-methyl signals [δ_H 2.50 (6H, s, 7″-Me), δ_H 42.4 (q, 7″-Me)] in compound 3 were replaced by a single N-methyl signal [δ_H 2.51 (3H, s, 7″-Me), δ_C 33.0 (q, 7″-Me)] in compound 4. Thus, compounds 3 and 4 were confirmed to be plicacetin and norplicacetin. Interestingly, compound 17 was also found to be a minor product in the ΔamiR mutant AM1009. These data indicated that AmiR played an essential role in amide bond formation between PABA and the terminal (+)-α-d-methylserine (Fig. 4B). However, its exact function remains unclear.

Implications for amide bond formation in amicetin biosynthesis.

Commonly occurring amide bonds in natural products are most often formed by the condensation domain of nonribosomal peptide synthetases. Nature also develops diverse ways to form amide bonds. For instance, amide bonds in aminocoumarins and fredericamycins are introduced by amide synthethases (10, 47), and thioesterase domains at the C termini of polyketide synthases (PKS) are proposed to catalyze amide bond formation in macrolactams (48). Transglutaminase homologue AdmF-catalyzed amide bond formation in andrimid represents a novel condensation strategy (23). Unusual strategies for amide bond formation in the biological assembly of peptidyl nucleoside antibiotics have been demonstrated recently. For instance, an ATP-independent strategy for amide bond formation in the A-503803 pathway was revealed by Funabashi et al. (26), and tRNA-dependent peptide bond formation in pacidamycin biosynthesis was validated by Zhang et al. (59). In amicetin biosynthesis, the amide linkage between PABA and (+)-α-d-methylserine moieties is likely to be formed by a common strategy through a nonribosomal peptide synthetase, AmiT. However, bioinformatic analysis shows that AmiT lacks a crucial condensation domain. We propose that AmiR, an acyltransferase-like protein, may act as a free-standing C domain in compensation for AmiT, to furnish the amide bond formation between PABA and (+)-α-d-methylserine. This hypothesis is supported by the fact that the amiR inactivation mutant accumulated two biosynthesis intermediates, plicacetin (compound 3) and norplicacetin (compound 4), both lacking the terminal (+)-α-d-methylserine moiety. Interestingly, the inactivation of amiF, an N-acetyltransferase-encoding gene, led to a mutant that produced the same two compounds (17 and 25) as the mutant with the inactivation of amiL, a PABA-CoA synthase gene, indicating that AmiF was essential for amide bond formation between the cytosine and PABA moieties. In rifamycin biosynthesis, an arylamine N-acetyltransferase, RifF, was also proposed as an amide synthase (49). Taken together, the data indicate that two alternative strategies, involving a nonribosomal peptide synthetase complex (comprising AmiT and AmiR) and an N-acetyltransferase (AmiF), were utilized to form the two amide bonds present in the amicetin structure.

Conclusion.

In this study, we cloned and characterized the amicetin biosynthesis gene cluster from S. vinaceusdrappus NRRL 2363. Additionally, the ami gene cluster was expressed in a heterologous host, S. lividans TK64, to produce amicetin and its analogues plicacetin and norplicacetin. The identity of the ami gene cluster was further confirmed by in vivo inactivation experiments with 4 putative biosynthesis genes (amiI, amiF, amiL, and amiR) and subsequent isolation of metabolites from the gene knockout mutants, thereby providing hints of two alternative strategies for amide bond formation in amicetin biosynthesis.