Marine actinomycetes are chemically rich sources of structurally diverse secondary metabolites. The vast majority of these secondary metabolites is mainly derived from members of the genus Streptomyces [1,2]. Approximately 289 secondary metabolites from the marine-derived genus of Streptomyces are reported in the Marinlit database, covering a wide variety of chemical structures, including peptides, macrolides, lactones, indoles, terpenes, and quinones. These compounds show an extensive range of activities, such as cytotoxic, antibacterial, antifungal and antimalarial [3]. The bioactive secondary metabolites from marine-derived Streptomyces have thus attracted increasing interest during the last decades [4,5,6].

In our endeavor to search for novel antibacterial secondary metabolites from marine Streptomyces using bioassay and LC-UV/Vis-MS guided methods, we isolated two new antimycin A analogues. Antimycin A is characterized by the presence of a nine-membered dilactone ring, a carboxyl phenol amido unit and two alkyl side chains of varying lengths, with the only exception being antimycin A9, which has an aromatic 8-acyl residue [7]. The first group of antimycin A compounds was isolated as fungicides from Streptomyces sp. in 1949 [8]. Molecular studies showed that these compounds could inhibit the mitochondrial electron transport chain between cytochromes b and c [9]. To date, 20 antimycin A compounds have been reported [10]. A recent report analyzing the biosynthetic gene clusters of antimycin A from Streptomyces sp. S4 revealed that these compounds were synthesized by a hybrid NRPS/PKS [11]. Here, we report on the discovery of the first naturally occurring ring-opened antimycin A analogues which were named as antimycin B1 and B2 (Figure 1) as well as their antibacterial activity.

Compound 1 was obtained as a light-brown amorphous solid with a molecular formula of C₂₄H₃₄N₂O₁₀, which was deduced from NMR and HRESITOFMS data (Obsd [M + H]⁺ at m/z 511.2290), which suggested nine degrees of unsaturation. The UV spectrum indicated absorption at λ_max (ε) 220 (22,900) and 317 (5800) nm. The ¹H NMR data for 1 (Table 1) indicated the presence of five exchangeable protons at δ_H 13.00 (s), 12.60 (s), 9.83 (brs), 8.81 (d, J = 8.0 Hz), and 5.09 (s), a formyl proton at δ_H 8.32 (d, J = 1.5 Hz), three aromatic protons at δ_H 8.27 (d, J = 8.0 Hz),7.80 (d, J = 8.0 Hz), and 6.92 (t, J = 8.0 Hz), four methine protons directly connected to heteroatoms δ_H 5.23 (dd, J = 10.0, 5.5 Hz), 4.98 (dq, J = 6.5, 5.5 Hz), 4.45 (dd, J = 8.0, 4.5 Hz), and 4.20 (m), one methine signal at δ_H 1.50 (m), and two methylene signals at δ_H 2.26 (m) and 1.37 (m) together with five methyl groups at δ_H 1.19 (d, J = 6.5 Hz), 1.16 (d, J = 6.5 Hz), 1.00 (d, J = 7.0 Hz), 0.84 (d, J = 7.0 Hz), and 0.82 (d, J = 6.5 Hz). These data suggested that 1 is an antimycin A analogue [8].

Two substructures I and II were assigned by analyzing the 1D and 2D NMR spectra (Figure 2). Substructure I was assembled starting from the amino acid threonine moiety that was inferred from the ¹H–¹H COSY correlations of δ_H 4.20 (m, 2-H) with δ_H 4.45 (dd, J = 8.0, 4.5 Hz, 3-H) and δ_H 1.16 (d, J = 6.5, 1-H), of δ_H 4.45 (3-H) with δ_H 8.81 (d, J = 8.0 Hz, 1′-CONH) and HMBC correlations of 3-H with δ_C 169.3 (C-4). The ¹H–¹H COSY correlations from δ_H 6.92 (t, J = 8.0 Hz, 5-H′) to δ_H 8.27 (d, J = 8.0, 4-H′) and 7.80 (d, J = 8.0 Hz, 6-H′) indicated the presence of an 1,2,3-trisubstituted benzene. The HMBC correlations from phenolic hydroxyl proton observed at δ_H 13.00 (s, 2′-OH) to δ_C 113.9 (C-1′), 150.7 (C-2′) and 127.0 (C-3′), from the formamide proton observed at δ_H 9.83 (brs, 3′-NHCHO) to C-3′, from H-6′ to δ_C 170.0 1′-CONH, suggested that 2′-OH, 3′-NHCHO, and 1′-CONH groups were attached on C-2′, C-3′ and C-1′, respectively. HMBC correlations from H-6′, 3-NH and H-3 to the carbonyl carbon δ_C 170.0 (1′-CONH), allowed the trisubstituted aromatic ring moiety and the threonine moiety to be connected.

Substructure II was assembled from the following ¹H–¹H COSY and HMBC spectral data. ¹H–¹H COSY correlations from δ_H 4.98 (dq, J = 6.5, 5.5 Hz, 6-H) to δ_H 1.19 (d, J = 6.5 Hz, 6-Me) and δ_H 5.23 (dd, J = 10.0, 5.5 Hz, 7-H), from δ_H 2.69 (m, 8-H) to 7-H and δ_H 1.00 (d, J = 7.0, 8-Me) suggested the presence of a 2,3,4-trisubstituted pentanoic acid group. HMBC correlations from 7-H, 8-H and 8-Me to δ_C 174.5 (C-9) indicated a carbonyl group attached on C-8. ¹H-¹H COSY correlations from δ_H 1.50 (m, 4′′-H) to 0.84 (d, J = 7.0 Hz, 5′′-H), δ_H 0.82 (d, J = 6.5, 6′′-H) and δ_H 1.37 (m, 3′′-H), and from 3′′-H to δ_H 2.26 (m, 2′′-H), together with HMBC correlations from 3′′-H and 2′′-H to δ_C 172.2 (C-1′′) indicated the presence of a 4-methylvaleryl group. The long-range coupling between 7-H and C-1′′ revealed that the 4-methylvaleryl residue was connected to C-7 via an ester bond. The HMBC correlation of 6-H with C-4 allowed the substructures I and II to be connected.

There are five stereocenters in compound 1. To assign the absolute configurations, two reactions were considered. The strategy was to establish the absolute stereochemistry of the threonine group using Marfey’s reagent, followed by lactonization to form the nine-membered dilactone ring found in cyclized analogs. If lactonization was successful, then NOEs could be obtained via NOESY or ROESY NMR experiments to complete the assignments. The threonine moiety of 1 was defined by acid hydrolysis and Marfey’s method [12,13], using standard amino acids, L-Threonine, L-allo-Threonine, D-Threonine and D-allo-Threonine. The derivatized threonine residue from compound 1 gave the same retention time as that prepared from standard L-Threonine and L-allo-Threonine (RT = 29.3 min). Thus, the absolute configuration of the C-3 was determined as 3S, but C-2 remains unresolved. The relative configurations at C-6, C-7 and C-8 were assigned as 6S*, 7R*, 8R* based on comparison of ¹H and ¹³C chemical shifts and ¹H-¹H coupling constant data with literature values [14]. To determine the absolute configurations of C-2, C-6, C-7 and C-8, lactonization was attempted using a modification of Shiina’s method [15]. The mixture products were checked using LC-MS. However, no expected dehydrated molecule mass data was observed. Due to the limited amount of the compound, it was not possible to perform additional experiments to determine the absolute configurations of C-2, C-6, C-7 and C-8. Thus, the structure of 1 was determined as depicted in Figure 1; it is named antimycin B1.

Compound 2 had a molecular formula of C₂₆H₃₁N₂O₁₀ deduced from its HRESITOFMS data (Obsd [M + H]⁺ at m/z 531.1985) and NMR spectra. Comparison of the ¹H and ¹³C NMR spectral data (Table 1) revealed close similarities between 2 and 1. The difference between them was the absence of the 4-methylvaleryl residue in Substructure II of 1 and the appearance of a phenyl-acetyl group in 2, which is verified by HMBC correlations from δ_H 3.66 (s, 2′′-H) to carbonyl carbon δ_C 170.5 (C-1′′) and carbons on the second aromatic ring δ_C 133.8 (C-3′′), 129.1 (C-4′′, 8′′). The correlation between δ_H 5.22 (dd, J = 7.0, 6.5 Hz, 7-H) and C-1′′ revealed that the phenyl-acetyl residue was connected to C-7 by an ester bond. The absolute configuration of C-3 in 2 was also determined as 3S by applying Marfey’s method. The relative configurations at C-6, C-7 and C-8 were assigned as 6S*, 7R*, 8R* as in compound 1. Thus, the structure of compound 2, named antimycin B2, was determined and is also shown in Figure 1.

Compounds 1 and 2 were evaluated for their antibacterial activities against Staphylococcus aureus, Bacillus subtilis and Loktanella hongkongensis. Compound 1 did not show any activity against any of the three strains. Compound 2 showed selective activity against two strains, S. aureus and L. hongkongensis. See Table 2.

Two new antimycin A analogues, antimycin B1 (1) and antimycin B2 (2), were isolated from a spent broth of marine-derived actinomycete Streptomyces lusitanus. The key structural features of compounds 1 and 2 were characterized by spectroscopic analyses and chemical methods. The compounds were suspected to be artificial products originated during the treatment process. A previous degradation study of antimycin A compounds revealed that mild alkaline hydrolysis of antimycin A compounds yielded blastmycic acid and antimycin lactone, while blastmycic acid would be further degraded into antimycic acid under more severe conditions [16]. However, due to intramolecular reaction [17], no ring-opened dilactone products were observed. Thus, the antimycin B compounds described here are the first naturally occurring antimycin A dilactone opened products to be identified. Antibacterial activities were evaluated using MIC assays. Compound 2 showed moderate antibacterial activities against S. aureus, L. hongkongensis and B. subtilis, with the activity against L. hongkongensis being stronger than that of streptomycin. Compound 1 did not show any activities against these three bacterial strains.