Analysis of the Loading and Hydroxylation Steps in Lankamycin Biosynthesis in Streptomyces rochei

Abstract

The biosynthetic gene cluster of lankamycin (LM), a 14-member macrolide antibiotic, is encoded on the 210-kb linear plasmid pSLA2-L in Streptomyces rochei 7434AN4. LM contains a 3-hydroxy-2-butyl group at the C-13 position, which is different from an ethyl group in erythromycin. The following two possibilities could be considered for the origin of this starter moiety of LM biosynthesis: (i) an extra module exists in the biosynthetic gene cluster and loads an additional acetate molecule, or (ii) 3-hydroxy-2-butyrate or its equivalent is loaded and incorporated as a starter. The former possibility was eliminated by the complete sequencing of pSLA2-L, which showed no extra module. On the other hand, the latter was confirmed by incorporation of deuterium in [3-²H]dl-isoleucine into the C-14 position of LM. The timing of hydroxylation reactions at the C-15 and C-8 positions of LM was studied by constructing disruptants of two P450 hydroxylase genes, lkmF (orf26) and lkmK (orf37). The lkmF disruptant produced 8-deoxylankamycin, while the lkmK disruptant produced both 15-deoxylankamycin and 8,15-dideoxylankamycin. These results clearly showed that LkmF is a C-8 hydroxylase and LkmK is a C-15 hydroxylase in LM biosynthesis and in addition suggested the order of hydroxylation steps; namely, hydroxylation may occur at first at C-15 by LkmK and then at C-8 by LkmF.

Streptomyces, a genus of gram-positive soil-inhabiting bacteria, produces a great number of secondary metabolites, including antibiotics, herbicides, immunosuppressants, and other biologically active compounds. Among them, polyketides represent a pivotal class of natural products with structural and biological diversity (28). Bacterial modular type I polyketide synthases (PKSs) are multifunctional megaenzymes for polyketide biosynthesis. Usual modular PKSs are composed of several sets of modules, each of which contains acyltransferase (AT), ketosynthase, and acyl carrier protein domains for one condensation step. Additional catalytic domains such as ketoreductase, dehydratase, and enoylreductase domains are also present in modules to reduce growing polyketides at different levels after each condensation reaction. Thus, typical modular PKSs show a strict colinear relationship between the module organization and all of the biosynthetic steps. However, during the last few years, unusual iterative type I PKSs without colinearity were reported (1, 21, 32).

Lankamycin (LM; Fig. 1, compound 2) is a 14-member macrolide antibiotic (5, 11) that shows moderate antimicrobial activity against several gram-positive bacteria such as Staphylococcus aureus, Bacillus subtilis, and Micrococcus luteus (18). Streptomyces rochei 7434AN4 produces LM in addition to the 17-member macrocyclic polyketide lankacidin (Fig. 1, compound 1) and carries three large linear plasmids, pSLA2-L, -M, and -S. Correlation between the antibiotic-producing ability and the plasmid profiles of mutants derived from strain 7434AN4 suggested that the largest plasmid, pSLA2-L, is involved in the production of both antibiotics (14). This idea was further supported by identification on pSLA2-L of a region homologous to eryAI, a typical modular PKS gene for erythromycin (EM), and an experiment in which it was disrupted (15, 29). Finally, we determined the complete nucleotide sequence of pSLA2-L (210,614 bp) and identified 143 open reading frames on it (19). It was revealed that pSLA2-L contains two type I PKS gene clusters for LM (lkm) and lankacidin (lkc), a cryptic type II polyketide gene cluster (roc), and a carotenoid biosynthetic gene cluster (crt). This finding is interesting because only a few cases are known where the antibiotic biosynthetic gene cluster is located on a linear plasmid (3, 4, 6, 13).

FIG. 1.

Chemical structures of lankacidin C (compound 1), LM (compound 2), and EM A (compound 3). Me, methyl; Ac, acetyl.

The macrolide skeleton (lankanolide) of LM is quite similar to that of EM (compound 3 in Fig. 1); the differences are the positions of hydroxylation (C-8 and C-15 in LM and C-6 and C-12 in EM) and the attached group at C-13 (3-hydroxy-2-butyl in LM and ethyl in EM). The C-13 side chain corresponds to a starter moiety in macrolide biosynthesis; an ethyl group in EM is derived from propionate. The following two possibilities could be considered for the origin of the 3-hydroxyl-2-butyl group in LM: (i) an extra module exists in the lkm gene cluster and loads an additional molecule of acetate, or (ii) the loading module of lkm can recognize and load 3-hydroxy-2-butyrate or its equivalent as a starter. The former possibility was finally eliminated by the complete sequencing of pSLA2-L, which showed that lkmAI does not contain any extra module and that its loading module and module 1 are quite similar to those of eryAI.

Therefore, the second possibility has remained to be studied. In avermectin biosynthesis, l-isoleucine and l-valine were incorporated as a starter (22), and branched-chain amino acids were also used in other polyketide compounds (reference 20 and references cited therein). Branched-chain amino acids are converted via a branched-chain fatty acid metabolic pathway. Namely, l-isoleucine is first converted to 3-methyl-2-oxopentanoic acid, which in turn is decarboxylated to give 2-methylbutyrate. LM contains two hydroxyl groups at C-15 and C-8, both of which seem to be introduced by P450 hydroxylases encoded on pSLA2-L (orf26 and orf37). Therefore, another interesting question is when these hydroxyl groups are introduced into a macrolide skeleton. To answer these questions, we carried out feeding and gene disruption experiments, results of which are described in this paper.

MATERIALS AND METHODS

Bacterial strains and media.

S. rochei strain 51252 (14), which contains only a linear plasmid, pSLA2-L, was used to construct P450 hydroxylase gene disruptants. Escherichia coli strain XL1-Blue {recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [F′ proAB lacI^qZΔM15 Tn10 (Tet^r)] } was used for routine cloning and construction of targeting plasmids. Targeting plasmids were propagated in E. coli ET12567 (dam dcm hsdM) (16) to obtain unmodified DNAs in order to overcome a strong restriction barrier in Streptomyces. E. coli strains were grown in Luria-Bertani medium supplemented with ampicillin (100 μg/ml). YM medium (0.4% yeast extract, 1.0% malt extract, 0.4% glucose, pH 7.3) and TSB medium (tryptic soy broth, 30 g/liter) were used for antibiotic production and bioassay, respectively.

DNA manipulation.

DNA isolation and manipulation for Streptomyces (12) and E. coli (26) were performed according to standard procedures. Plasmids pRES18, a shuttle vector containing a thiostrepton resistance gene (9), and pUC4-KIXX, containing a kanamycin resistance gene cassette (2), were used for gene disruption. For protoplast preparation, S. rochei strains were grown in YEME liquid medium containing 34% sucrose (12). Southern hybridization was carried out with a digoxigenin DNA labeling and detection kit (Roche) according to the manufacturer's protocol.

Spectroscopic instruments.

Nuclear magnetic resonance (NMR) spectra were recorded on a JEOL LA-500 spectrometer equipped with a field gradient accessory. Deuteriochloroform and deuterium oxide were used as NMR solvents in ¹H and ¹³C NMR analyses. Chemical shifts were recorded in δ values based on the resident solvent signals (δ_C = 77.0 in CDCl₃, δ_H = 4.65 in HDO) or the internal standard signals of tetramethylsilane (δ_H = 0) or dioxane (δ_C = 66.5). For ²H NMR analysis, CHCl₃ or H₂O was used as the solvent. Correlations in the LM framework were determined by several two-dimensional NMR spectra (double quantum filtered-correlated spectroscopy, heteronuclear multiple-quantum correlation, and heteronuclear multiple-bond correlation [HMBC]). Mass spectra were monitored on a JEOL SX-102A mass spectrometer.

Synthesis of labeled isoleucine.

[3-²H]3-methyl-2-oxopentanoic acid (compound 4-d). To a solution of 3-methyl-2-oxopentanoic acid (compound 4) (1.0 g, 7.6 mmol) in pyridine (10 ml) was added deuterium oxide (5.0 ml), and the mixture was stirred at 40°C for 24 h. The solvent was removed in vacuo, and this cycle was repeated five times to afford deuterium-labeled acid compound 4-d (1.0 g). Compound 4-d (keto form obtained as a pyridinium salt): ¹H NMR (D₂O) δ 0.70 (3H, t, J = 7.5 Hz), 0.91 (3H, s), 1.27 (1H, m), 1.51 (1H, m); ¹³C NMR (D₂O) δ 11.4, 14.6, 25.3, 43.8 (t, J = 19.5 Hz), 128.3, 141.9, 148.1, 169.6, 208.5; ²H NMR (H₂O) δ 2.80.

[3-²H]dl-isoleucine (compound 5-d).

A mixture of 2-keto acid compound 4-d (1.4 g, 11 mmol), ammonium chloride (1.5 g), formic acid (1.0 ml), sodium acetate (1.5 g), and 10% Pd-C (1.5 g) in water (20 ml) was stirred at 40°C for 3 days. The mixture was passed through a pad of Celite, and the filtrate and washings were lyophilized. The resultant was subjected to ion-exchange chromatography (Dowex 50W-X2, H⁺ form) and eluted with 1 M aqueous ammonia to afford [3-²H]dl-isoleucine (compound 5-d; 0.33 g, 25%) as a mixture of four diastereoisomers. Compound 5-d: ¹H NMR (D₂O) δ 0.77 to 0.85 (6H, m), 1.07 to 1.35 (2H, m), 1.85 to 1.96* (0.5H, m), 3.54 (0.5H, H-2 in l isomer), 3.62 (0.5H, H-2 in d isomer); ¹³C NMR (D₂O) δ 11.8, 11.9, 14.1, 15.5, 25.2, 26.3, 35.9*, 36.2*, 59.4, 60.4, 175.0, 175.5; ²H NMR (H₂O) δ 1.77, 1.85. Signals with an asterisk were derived from nonlabeled dl-isoleucine.

Feeding experiments.

S. rochei strain 51252 was precultured at 28°C for 2 days in 10 ml of YM liquid medium in a test tube, and then a 1-ml culture was transferred to 100 ml of YM liquid medium in a 500-ml Sakaguchi flask. Labeled compounds, compound 4-d and compound 5-d (each 200 mg), were added at 10 h, and the culture was stopped and analyzed at 72 h.

Construction of a targeting plasmid for orf26 (lkmF).

The 3.0-kb BamHI fragment containing lkmF was cloned into pUC19 to give pKK2601. The 0.7-kb NruI-Eco47III fragment of pKK2601 was replaced with the 1.2-kb SmaI fragment of pUC4-KIXX to generate pKK2602, the vector of which was replaced with pRES18 to afford targeting plasmid pKK2603.

Construction of a targeting plasmid for orf37 (lkmK).

The 2.8-kb PvuII fragment containing lkmK was cloned into pBluescript SK-plus predigested with EcoRV to give pKK3704. This plasmid was digested with EheI and ScaI and self-ligated to generate pKK3705, which lost the central 903-bp EheI-ScaI fragment. The 1.9-kb EcoRI-HindIII fragment of pKK3705 containing a mutated lkmK gene was inserted into pRES18 to afford targeting plasmid pKK3706.

Gene inactivation.

Streptomyces protoplasts were transformed by targeting plasmids propagated in E. coli ET12567, regenerated on R1M plates (33), and overlaid with soft nutrient agar (Difco) containing thiostrepton (final concentration of 10 μg/ml). Thiostrepton-resistant colonies were picked up and subjected to successive liquid cultures in YEME medium containing 10 μg/ml kanamycin to facilitate a double crossover. Gene replacement in kanamycin-resistant and thiostrepton-sensitive colonies was confirmed by Southern hybridization analysis.

Isolation and analysis of metabolites.

S. rochei strains were reciprocally cultured in Sakaguchi flasks at 28°C for 3 days, and the supernatant was extracted twice with the same volume of ethyl acetate. The combined organic phase was dried with Na₂SO₄, filtered, and concentrated to dryness. The resulting crude extract was subjected to Sephadex LH-20 chromatography (1 by 40 cm; Amersham Pharmacia Biotech AB) with methanol. LM and its derivatives (molecular weight of ca. 800) were eluted in 20- to 25-ml fractions, while lankacidins (molecular weight of ca. 500) were eluted in 28- to 34-ml fractions. The former fractions were purified by successive silica gel chromatography with chloroform-methanol (100:1 to 50:1) and then toluene-ethyl acetate (1:1 to 1:2). The average yield of LM (compound 2) from parent strain 51252 was 2.0 mg/liter. The yields of 8-deoxylankamycin (compound 6) from strain KK01 (Δorf26) and 15-deoxylankamycin (compound 7) and 8,15-dideoxylankamycin (compound 8) from strain KA26 (Δorf37) were 1.2, 0.7, and 0.2 mg/liter, respectively.

Bioautography of the metabolites was carried out as described previously (1). M. luteus was used as the test microorganism. Purified LM and deoxylankamycins (each 0.20 mg) were spotted onto paper disks (diameter, 8 mm), and the relative activities of deoxylankamycins were estimated from the calibration curve of the inhibitory zone of LM.

Nucleotide accession number.

The nucleotide sequence reported in this paper has been deposited in the DDBJ database under accession number AB088224.

RESULTS

Feeding experiment with deuterium-labeled isoleucine.

Deuterium-labeled isoleucine was synthesized from 3-methyl-2-oxopentanoic acid (compound 4) as shown in Fig. 2A. The hydrogen at C-3 was replaced with deuterium in the presence of deuterium oxide and pyridine to afford [3-²H]3-methyl-2-oxopentanoic acid (compound 4-d). It was then converted to [3-²H]dl-isoleucine (compound 5-d) by reaction with palladium carbon and ammonium chloride (23), although the deuterium labeling at C-3 in compound 5-d was decreased to around 50%. The ²H NMR spectrum of LM obtained by feeding of compound 5-d gave a distinct signal at 1.8 ppm (Fig. 2B, part i), indicating that isoleucine was incorporated into the C-13 branched chain of LM. The deuterium atom in synthetic intermediate compound 4-d was also incorporated into C-14 of LM (Fig. 2B, part ii). These results suggest that l-isoleucine was oxidized to (S)-3-methyl-2-oxopentanoic acid, decarboxylated to (S)-2-methylbutyrate, and then incorporated into the macrolide (lankanolide) skeleton.

FIG. 2.

Synthesis of labeled isoleucine and NMR spectra of labeled LMs. (A) Synthetic scheme of deuterium-labeled compounds [3-²H]3-methyl-2-oxopentanoic acid (compound 4-d) and [3-²H]dl-isoleucine (compound 5-d). (B) ²H NMR of labeled LM obtained by feeding of compound 5-d (i) or compound 4-d (ii) and ¹H NMR of unlabeled LM (iii). Me, methyl; Py, pyridine.

Alignment of two P450 hydroxylases, LkmF and LkmK.

Incorporation of isoleucine into the starter moiety of LM raised the next question, i.e., when the hydroxyl group at C-15 is introduced. In addition, LM contains another hydroxyl group at the C-8 position. Hydroxylation reactions of the macrolide skeletons are usually done by cytochrome P450 hydroxylases encoded on their biosynthetic gene clusters, for example, EryF and EryK for EM (7, 27), OleP for oleandomycin (25), and ChmPI and ChmHI for chalcomycin (30). Two P450 hydroxylase genes (orf26 and orf37) were also identified in the lkm cluster on pSLA2-L (19). The gene products of orf26 (lkmF) and orf37 (lkmK) show significant similarity to P450 hydroxylases; LkmF shows 64% identity and 78% similarity to EryF, while LkmK shows 49% and 65%, respectively. Partial amino acid sequences of LkmF, LkmK, and other bacterial P450 hydroxylases are aligned and compared in Fig. 3A. All of them contain a cysteine residue for heme binding and a threonine residue for O₂ binding (24), except for the latter residue being replaced with alanine in EryF.

FIG. 3.

Functional analysis of P450 hydroxylases LkmF and LkmK in LM biosynthesis. (A) Partial alignment of P450 hydroxylases involved in antibiotic production. EryF, EM C-6 hydroxylase (accession number AAA26496); OleP, oleandomycin C-8 hydroxylase (accession number AAA92553); ChmPI, probable chalcomycin C-8 hydroxylase (accession number AAS79447); ChmHI, probable chalcomycin C-20 hydroxylase (accession number AAS79453); EryK, EM C-12 hydroxylase (accession number P48635); AveE, P450 hydroxylase for avermectin synthesis (accession number BAA84477). The O₂-binding and heme-binding residues are indicated. (B) TLC analysis of the metabolites of P450 hydroxylase gene disruptants. Lanes: i, lankacidin C; ii, LM; iii, strain 51252 (parent); iv, KK01 (ΔlkmF); v, KA26 (ΔlkmK). The TLC plate was developed with chloroform-methanol (20:1), sprayed with anisaldehyde in 10% sulfuric acid, and baked. Compounds 6 and 8 showed gray-blue spots, lankacidin C (compound 1) showed a blue-violet spot, and LM (compound 2) and compound 7 showed violet spots. (C) Chemical structures of deoxylankamycins isolated from mutants KK01 and KA26. (D) Antimicrobial activities of deoxylankamycins. i, LM (compound 2); ii, 8-deoxylankamycin (compound 6); iii, 15-deoxylankamycin (compound 7); iv, 8,15-dideoxylankamycin (compound 8). Me, methyl; Ac, acetyl.

Metabolite of the orf26 (lkmF) disruptant.

To determine the function of LkmF and LkmK in LM biosynthesis, we constructed disruptants of the genes for both proteins and analyzed their metabolites by thin-layer chromatography (TLC). The lkmF disruptant KK01 did not produce LM but produced the novel metabolite LM-KK01 (compound 6) (Fig. 3B, lane iv). This compound was isolated and subjected to mass spectrometry (MS) and NMR analysis, the data of which are summarized in Table 1. High-resolution fast atom bombardment-MS analysis determined the molecular formula of LM-KK01 to be C₄₂H₇₃O₁₅ (m/z 817.4937 for [M+H]⁺, calculated 817.4951). Since the [M+H]⁺ ion of LM was observed at m/z 833, LM-KK01 is one oxygen atom smaller than LM. In its ¹H NMR spectrum, an H-8 proton newly appeared at 2.78 ppm and a singlet C-8 methyl signal at 1.34 ppm in LM was shifted to 1.12 ppm because of the loss of a vicinal hydroxyl group. In the spectrum determined by distortionless enhancement by polarization transfer, a tertiary C-8 carbon (80.2 ppm) in LM was changed to a methine carbon (44.7 ppm) in LM-KK01. Other signals of LM-KK01 were almost identical to those of LM. From these data, LM-KK01 was determined to be 8-deoxylankamycin (Fig. 3C, compound 6).

TABLE 1.

¹H- and ¹³C-NMR data for LM and deoxylankamycins

No.	LM (compound 2)		8-Deoxylankamycin (compound 6)		15-Deoxylankamycin (compound 7)		8,15-Dideoxylankamycin (compound 8)
	δCa	δHb	δC	δH	δC	δH	δC	δH
1	176.7 (s)		175.9 (s)		175.2 (s)		175.1 (s)
2	44.8 (d)	2.80 (br)	44.7 (d)	2.84 (m)	44.9 (d)	2.81 (m)	44.9 (d)	2.86 (m)
3	77.8 (d)	3.95 (d, 4.9)e	78.4 (d)	3.86 (d, 6.3)e	78.3 (d)	3.85 (d, 6.4)e	79.5 (d)	3.75 (m)
4	44.1 (d)	1.83 (m)	43.6 (d)	1.86 (m)	44.0 (d)	1.85 (m)	41.2 (d)	1.84 (m)
5	84.5 (d)	3.51 (m)	84.5 (d)	3.55 (d, 6.7)	85.0 (d)	3.48 (t, 6.7)	85.0 (d)	3.50 (m)
6	33.9 (d)	2.21 (br)	33.3 (d)	1.80d	34.5 (d)	2.20 (br)	37.0 (d)c	2.31 (br)
7	39.4 (t)	1.93d	39.3 (t)	1.96d	38.7 (t)	1.98d	41.3 (t)	2.05d
8	80.2 (s)		44.7 (d)	2.78 (m)	80.3 (s)		41.2 (d)	2.86 (m)
9	214.4 (s)		216.3 (s)		214.6 (s)		215.1 (s)
10	38.2 (d)	3.15 (q, 6.7)f	40.5 (d)	3.03 (q, 6.7)f	37.9 (d)	3.12 (q, 6.8)f	43.0 (d)	3.02 (q, 6.7)f
11	71.0 (d)	4.87 (d, 8.9)f	72.4 (d)	4.88 (d, 9.8)f	70.4 (d)	4.73 (d, 11.0)f	72.7 (d)	4.85 (d, 9.8)f
12	39.4 (d)	1.93d	39.3 (d)	1.97d	35.9 (d)	1.98d	36.1 (d)c	2.03d
13	73.0 (d)	4.83 (d, 6.4)g	73.4 (d)	4.90 (d, 7.0)g	75.6 (d)	4.71 (d, 10.4)g	76.7 (d)	4.75 (d, 9.8)g
14	42.7 (d)	1.82 (m)	42.5 (d)	1.87 (m)	35.6 (d)	1.70 (m)	35.8 (d)	1.71 (m)
15	69.1 (d)	3.71 (m)	68.8 (d)	3.74 (m)	25.4 (t)	1.21, 1.45 (m)	25.5 (m)	1.11, 1.45 (m)
1′	102.5 (d)	4.33 (d, 7.7)	102.6 (d)	4.29d	102.0 (d)	4.30 (d, 7.3)	102.9 (d)	4.25 (d, 7.6)
2′	75.3 (d)	3.34 (dd, 7.7, 8.9)	75.5 (d)	3.34d	75.4 (d)	3.34 (t, 8.2)	75.5 (d)	3.32 (t, 8.0)
3′	80.2 (d)	3.25 (m)	80.1 (d)	3.26 (m)	80.1 (d)	3.24 (m)	80.1 (d)	3.24 (m)
4′	37.1 (t)	1.25, 2.07 (m)	37.1 (t)	1.25, 2.07 (m)	37.1 (t)	1.25, 2.05 (m)	37.1 (t)	1.25, 2.03 (m)
5′	67.3 (d)	3.51 (m)	67.4 (d)	3.49 (m)	67.2 (d)	3.50 (m)	67.4 (d)	3.93 (m)
1"	96.6 (d)	5.05 (d, 4.6)	97.4 (d)	4.99 (d, 4.3)	97.0 (d)	5.04 (d, 4.6)	97.9 (d)	4.97 (d, 4.6)
2"	30.6 (t)	1.67, 2.09 (m)	30.9 (t)	1.68, 2.10 (m)	30.9 (t)	1.68, 2.05 (m)	31.1 (t)	1.68, 2.08 (m)
3"	72.6 (s)		72.5 (s)		72.5 (s)		72.4 (s)
4"	73.9 (d)	4.67 (s)h	73.8 (d)	4.70 (s)h	73.9 (d)	4.67 (s)h	73.8 (d)	4.69 (s)h
5"	62.6 (d)	4.47 (q, 6.6)h	62.5 (d)	4.47 (q, 6.4)h	62.7 (d)	4.47 (q, 6.4)h	62.5 (d)	4.47 (q, 6.4)h
3′-OCH3	56.9 (q)	3.44 (s)	56.9 (q)	3.44 (s)	56.8 (q)	3.44 (s)	56.9 (q)	3.44 (s)
3"-OCH3	49.3 (q)	3.30 (s)	49.3 (q)	3.30 (s)	49.3 (q)	3.29 (s)	49.3 (q)	3.28 (s)
8-CH3	27.0 (q)	1.34 (s)	16.9 (q)	1.12d	27.8 (q)	1.32 (s)	16.9 (q)	1.05d
10-CH3	10.1 (q)	1.10d	10.1 (q)	1.01d	10.1 (q)c	1.09 (d, 6.6)	9.2 (q)	1.00 (d, 6.7)
15-CH3	19.7 (q)	1.15 (d, 6.4)	19.2 (q)	1.14d	10.2 (q)c	0.90 (t, 7.3)	10.2 (q)	0.90 (t, 6.7)
11-COCH3	20.9 (q)c	2.08 (s)	21.0 (q)	2.05 (s)	21.0 (q)	2.07 (s)	20.8 (q)c	2.03 (s)
11-COCH3	170.1 (s)		169.8 (s)		169.5 (s)		169.7 (s)
4"-COCH3	21.0 (q)c	2.13 (s)	21.0 (q)	2.15 (s)	21.0 (q)	2.13 (s)	20.9 (q)c	2.14 (s)
4"-COCH3	170.7 (s)		170.6 (s)		170.6 (s)		170.6 (s)

^aMultiplicity is shown in parentheses.

^bMultiplicity and J values in Hz are shown in parentheses.

^cAssignments are exchangeable.

^dObscured by overlapping.

^eVicinal proton coupling constant J_3,4 is ∼0 Hz.

^fVicinal proton coupling constant J_10,11 is ∼0 Hz.

^gVicinal proton coupling constant J_12,13 is ∼0 Hz.

^hVicinal proton coupling constant J_4",5" is ∼0 Hz.

Metabolites of the orf37 (lkmK) disruptant.

The lkmK disruptant KA26 produced two LM derivatives (LM-KA26A and LM-KA26B; R_fs, 0.5 and 0.6 in CHCl₃-methanol at 20:1) (Fig. 3B, lane v). The fast atom bombardment-MS spectrum of LM-KA26A (compound 7) showed a parent peak at m/z 817, indicating that it is also monodeoxylankamycin. The ¹H NMR spectrum of LM-KA26A gave distinct triplet methyl protons (CH₃-C₁₅) at 0.90 ppm. This proton signal showed an HMBC relationship with a methylene carbon (C-15) at 25.4 ppm, which in turn had another HMBC correlation with the C-14 methyl group. Since other signals are almost identical to those of LM, LM-KA26A was determined to be 15-deoxylankamycin (Fig. 3C, compound 7; C₄₂H₇₃O₁₅, m/z 817.4948 for [M+H]⁺, calculated 817.4951).

Another compound, LM-KA26B (compound 8), showed a parent ion at m/z 801, suggesting that it is dideoxylankamycin. In the ¹³C NMR spectra of LM-KA26B and LM, the C-8 quaternary carbon at 80.2 ppm in LM was changed to a methine carbon at 41.2 ppm and the C-15 methine carbon at 69.1 ppm was changed to a methylene carbon at 25.5 ppm. In addition, in the ¹H NMR spectrum of LM-KA26B, an H-8 signal newly appeared around 2.86 ppm and the H-15 signal was shifted to the hydrocarbon region around 1.1 to 1.5 ppm. These data revealed that LM-KA26B is 8,15-dideoxylankamycin, losing both hydroxyl groups at C-8 and C-15 (Fig. 3C, compound 8; C₄₂H₇₃O₁₄, m/z 801.4978 for [M+H]⁺, calculated 801.5002).

Possible order of two hydroxylation steps.

The structures of the metabolites produced by two P450 hydroxylase-deficient mutants revealed that LkmF is a C-8 hydroxylase and LkmK is a C-15 hydroxylase. Isolation of 15-deoxylankamycin from mutant KA26 (ΔlkmK) indicates that 3-methylbutyrate was loaded and incorporated as a starter and a hydroxylation event occurred at C-15 after completion of macrolide synthesis. In addition, isolation of 8,15-dideoxylankamycin from mutant KA26, not from mutant KK01 (ΔlkmF), suggested that hydroxylation reactions occurred first at C-15 and then at C-8. We speculate that mutant KA26 (ΔlkmK) could partially convert 8,15-dideoxylankamycin to 15-deoxylankamycin because of the flexible substrate specificity of the C-8 hydroxylase LkmF.

Antimicrobial activities of metabolites.

The three deoxylankamycins produced by mutants KK01 and KA26 were tested for antimicrobial activity against M. luteus. Compared with LM (compound 2, 100%), they showed decreased antimicrobial activity in the reverse order of the number of hydroxyl groups (Fig. 3D; compound 6, 50% activity; compound 7, 18% activity; compound 8, 2.5% activity). A similar relationship was also observed in EM A and its 6-deoxy derivative (31). These results suggest that the hydroxyl groups in the lankanolide skeleton contribute substantially to the compound's antimicrobial activity.

DISCUSSION

In this study, we have analyzed the origin of the starter moiety and the function of two P450 hydroxylase genes in LM biosynthesis. On the basis of the results, we propose the LM biosynthetic pathway depicted in Fig. 4. Namely, isoleucine is first oxidized to 3-methyl-2-oxopentanoic acid and then converted to (S)-2-methylbutyrate. These oxidation reactions may be carried out by aminotransferase and dehydrogenase in the isoleucine metabolic pathway (Fig. 4A). These genes are not located on pSLA2-L; however, the dehydrogenase gene bkdF (8) from S. avermitilis JCM5450 gave a hybridizing signal to S. rochei chromosomal fragments (datanot shown). Therefore, (S)-2-methylbutyryl coenzyme A (CoA) may be synthesized by primary metabolic enzymes encoded on the chromosome and used as a starter unit in LM biosynthesis. Incorporation of (S)-2-methylbutyryl-CoA and subsequent hydroxylation of the macrolide skeleton were supported by the isolation of 15-deoxylankamycin from the lkmK disruptant.

FIG. 4.

Proposed LM biosynthetic pathway. Panels: A, origin of the starter moiety; B, polyketide chain assembly. Me, methyl; PLP, pyridoxal-5′-phosphate; 2-OG, 2-oxoglutarate; CoA, coenzyme A; AT, acyltransferase; KS, ketosynthase; ACP, acyl carrier protein; KR, ketoreductase; DH, dehydratase; ER, enoylreductase; TE, thioesterase.

In many macrolide antibiotics, including EM, propionyl-CoA is used as a starter unit. Recognition of a starter molecule is carried out by an AT domain of the loading module of PKS. Thus, replacement of the loading module of eryAI with that of aveAI resulted in the incorporation of isobutyryl-CoA and (S)-2-methylbutyryl-CoA into the EM skeleton (17). Consequently, we compared the AT domains of the loading modules to seek a recognition motif, but no apparent differences have been found among LkmAI [recognizes (S)-2-methylbutyryl-CoA], AveAI [isobutyryl-CoA and (S)-2-methylbutyryl-CoA], EryAI, TylGI, and OleAI (propionyl-CoA).

Metabolite analysis of the gene disruptants revealed that LkmF is a C-8 hydroxylase and LkmK is a C-15 hydroxylase and also suggested that C-15 hydroxylation by LkmK occurred first and then C-8 hydroxylation by LkmF occurred. This speculation was supported by the isolation of 8-deoxylankanolide from the fermentation broth of Streptomyces spp. (10). In EM biosynthesis, C-6 hydroxylation of 6-deoxyerythronolide B occurs before subsequent glycosylation events (27). The resultant erythronolide B is combined with dTDP-l-mycarose and dTDP-d-desosamine, in that order, and then hydroxylated at C-12 to give EM A. All of the three LM derivatives we isolated from the hydroxylase mutants were glycosylated. However, we cannot speculate about the timing of the hydroxylation and glycosylation events because glycosyltransferases and P450 hydroxylases show wide substrate specificities. In vitro transformation of deoxylankamycins by the isolated hydroxylases, which is in progress in our laboratory, will answer this question.