The marine environment is a plentiful source of bioactive natural products with potential for development as new pharmaceutical agents. To date, the success rate of drug discovery from the marine world is seven approved drugs out of 22,000 discovered molecular entities, which is approximately 1.7 to 3.3-fold higher than the industry average [1]. Among the diverse spectrum of marine organisms, culturable marine bacteria are prime candidates due to their production of bioactive natural products for possible use in chemical and clinical research [2].

Methicillin-Resistant Staphylococcus aureus (MRSA), a traditional nosocomial opportunistic pathogen, has spread from hospitals to the community in the last few decades [3]. Compared with methicillin-susceptible S. aureus bacteremia, MRSA bacteremia cases are associated with greater lengths of hospital stay, significantly higher mortality, and increased costs, posing a heavy burden on healthcare systems [4,5,6]. In the United States, the standardized mortality rate from MRSA infection in 2005 was 6.3 per 100,000 [7]. MRSA contains a mutant gene that manufactures a protein that protects the bacterium from the binding of penicillin [8]. Due to the current prevalence of multidrug-resistant bacteria such as MRSA in many countries, the discovery of new antibiotics with novel modes of action is crucial to tackle the threat of multidrug-resistant bacteria [9].

Citreamicins are polycyclic xanthones isolated from Streptomyces, and have been reported to show antibacterial activity against a variety of Gram-positive bacteria, including MRSA and vancomycin-resistant Enterococcus faecalis (VRE). To date, 10 citreamicins have been reported by several research groups [10,11,12,13,14]. Citreamicins share a common 5/6/6/6/6/6/6 seven-ring system, and identifying their respective structures is usually difficult due to the existence of multiple quaternary carbon atoms.

With the aim of discovering novel antibacterial natural products from marine bacteria, we examined a rarely explored area of rich biodiversity—the Red Sea. We performed a systematic biological and chemical analysis of the secondary metabolites produced by bacteria isolated from the Red Sea based on various bioassays, ultra performance liquid chromatography-mass spectrometry (UPLC-MS) analysis, and UV profile analysis. One of the hits resulting from our efforts was identified by its 16S rRNA gene as the strain Streptomyces caelestis. The genus Streptomyces belongs to the actinomycete family, which is responsible for producing about half of all of the discovered secondary metabolites known to have broad biological activity, such as antibacterial, anticancer, and anti-inflammatory activity [15,16,17,18,19]. Bioassay-Guided fractionation of the S. caelestis extract led to the isolation of four new compounds (1–4), all of which showed antibacterial activity against Staphylococcus haemolyticus, Staphylococcus aureus, and Bacillus subtillis. Citreamicin θ A (1), citreamicin θ B (2), and citreaglycon A (3) also exhibited low minimum inhibitory concentration (MIC) values of 0.25, 0.25, and 8.0 μg/mL, respectively, against methicillin-resistant strain Staphylococcus aureus (MRSA) ATCC 43300. Compounds 1 and 2 also displayed cytotoxic activity against HeLa cells. Here, we report the fermentation, isolation, structural elucidation, and bioactivity evaluation of these isolated compounds from S. caelestis.

Compound 1 was isolated as a red crystal with the molecular formula C₃₀H₂₅NO₁₁, based on a high resolution electrospray ionization mass spectroscopy (HRESIMS) of 576.1508 for the [M + H]⁺ value (calcd 576.1506), indicating 19 degrees of unsaturation. The UV spectrum showed the existence of conjugation based on absorption at λ_max 237, 278, 323, and 441 nm. Analysis of the ¹³C NMR spectra indicated the presence of three carbonyls (δ_C 183.1, 173.7, 166.3), including a ketone at δ_C 183.1; 18 sp² carbons in the down field region (three methines (δ_C 122.6, 118.2, 107.7) and fifteen quaternary carbons); and two sp³ quaternary carbons (δ_C 93.6, 65.5), four methylene groups (δ_C 61.6, 40.4, 29.4, 23.6), two methyl groups (δ_C 26.5, 19.2) and one methoxy (δ_C 57.3) in the high field region. Together, these signals reveal a skeleton of xanthones very similar to that identified in citreamicin ε (5) [13] (Figure 1). A fragment ion at m/z 475 in ion source collision-induced dissociation electrospray ionization mass spectrometry (ISCID ESIMS) corresponding to citeamicin ε (5) without the G ring ion was also observed (Figure 2).

Further comparison of the 1D and 2D NMR data of compound 1 and citreamicin ε [13] indicated that they are structural isomers with the same molecular weight and similar chemical shifts. Heteronuclear multiple bond correlation (HMBC) correlations from the methyl group at δ_H 1.61 (H-21) to carbons at δ_C 173.7 (C-2), 65.5 (C-3), and 61.6 (C-20), together with HMBC correlations from δ_H 1.67 (H-22) to δ_C 40.4 (C-18) and 93.6 (C-19), reveal the G ring. However, as rings B and C are constituted by quaternary carbons, it was difficult to obtain an unambiguous structure through 2D NMR analysis. To determine the exact difference in the constituent parts of compound 1 and citreamicin ε, we attempted to culture a single crystal of compound 1. From the crystal data [20], we were able to identify two ortho-phenolic hydroxyls at positions 13 and 14 and to determine the planar structure of compound 1 as that shown in Figure 1.

Compound 2 was also isolated as a red crystal with an identical UV profile to that of compound 1 with absorption at λ_max 237, 278, 323, and 441 nm. Its molecular formula of C₃₀H₂₅NO₁₁ ([M + H]⁺ 576.1500 calcd 576.1506) obtained from HRESIMS and the ¹H and ¹³C NMR data (Table 1) are very similar to those of 1, with the significant differences being the methylene signals at position 18 (δ_H 3.37 in 1 and δ_H 3.45 in 2) and at position 20 (δ_H 3.55/4.31, δ_C 61.6 in 1 and δ_H 3.69/4.10, δ_C 63.4 in 2) in the 1D NMR spectra. Further analysis of the 2D NMR spectra of compound 2 showed that a proton signal at δ_H 7.54 (H-9) had a H¹-H¹ correlation (H¹-H¹ COSY) to δ_H 6.82 (H-8) and HMBC correlations to δ_C 145.5 (C-7), 149.1 (C-10), 143.3 (C-11). Together, these data elucidate the construction of the A and B rings. HMBC correlations from the methyl group at δ_H 1.61 (H-21) to carbons at δ_C 173.4 (C-2), 65.6 (C-3) and 63.4 (C-20), together with HMBC correlations from δ_H 1.69 (H-22) to δ_C 40.0 (C-18) and 93.7 (C-19) reveal the G ring. The similar HMBC correlations in 2 to those observed in 1 suggest that they are a pair of diastereomers.

As both compounds 1 and 2 possess two stereogenic centers, there were two possibilities regarding the directions of the two methyl groups at C-3 and C-19. The relative configuration of 2 was determined from the nuclear overhauser effect spectroscopy (NOESY). The structures of 1 and 2 were optimized by the molecular modeling (MM2) force field and the calculated distance between the H₃-21 (δ_H 1.61) and H-18 (δ_H 3.45) are shown in Figure 3. The distances between protons beyond 4 Å would give no or only very weak nuclear overhauser effect (NOE) signals [21]. Although, the two protons at position 18 were not distinguished, the distance between H₃-21 and one of the H-18 is 4.658 Å that is beyond the cutoff of 4 Å [21]. The NOE correlation between H₃-21 (δ_H 1.61) and H-18 (δ_H 3.45) revealed that they are on the same side of the molecule, which indicates that the methyl groups at positions C-3 and C-19 have opposite orientations (Figure 3).

The circular dichroism (CD) spectrum of compound 1 displayed a negative Cotton effect with weak minima at short wavelengths and a maximum at higher wavelengths [λ_max 220 nm (Δε −0.41), λ_max 248 nm (Δε +1.14)] (Figure 4). However, the CD spectrum of compound 2 showed an almost opposite Cotton effect to that of compound 1. The CD spectra of compounds 1 and 2 confirmed that these two compounds were diastereomers [22,23].

Compound 3 has the molecular formula of C₂₇H₂₀O₁₁, as established by the HRESIMS of the pseudomolecular ion peak at m/z 521.1079 [M + H]⁺ (calcd for 521.1084). The UV spectrum displayed absorption at λ_max 238, 298, 412 nm, which is similar to the findings for compound 1. The ¹H NMR spectrum revealed the presence of four phenolic protons (δ_H 9.29, 10.3, 11.7, 12.6), three aromatic protons (δ_H 6.82, 7.49, 7.63), one methoxy group (δ_H 3.31), one methyl (δ_H 1.66), and three methylene groups (Table 2). The ¹³C NMR spectrum displayed 27 carbon signals, which were classified as three carbonyls (δ_C 167.5, 169.0, 180.9); nineteen carbons between δ_C 160 and 100, indicating multiple aromatic rings; three methylene (δ_C 23.1, 29.6, 36.8), one methoxy (δ_C 50.1); and one methyl group (δ_C 21.7). Compound 3 had the same xanthone skeleton as compound 1, but without the G ring.

The HMBC spectrum of 3 revealed correlations from δ_H 3.31 and δ_H 1.66 to δ_C 106.1 (C-16), indicating the attachment of the methyl and methoxy group at C-16. The downfield chemical shift of C-16 suggests that this quaternary carbon attaches to an additional oxygen. The construction of the F ring was verified by an HMBC correlation from δ_H 1.66 to carbonyl carbon δ_C 169.0 (C-2) and δ_C 36.8 (C-15). HMBC correlation from δ_H 7.49 (H-8) to δ_C 167.5 (C-18) provided evidence for the presence of a carboxylic acid group at C-7 (Figure 5), and the planar structure of compound 3 was thus determined to be that shown in Figure 1. However, the relative structure of 3 at C-16 remains unconfirmed due to the limited signals in the NOESY spectrum. Furthermore, since the single crystal of compound 3 was not obtainable, and no CD signal was observed, the absolute configuration of compound 3 is still undetermined.

Compound 4 has the molecular formula C₂₇H₂₀O₁₀, as deduced from its HRESIMS data ([M + H]⁺ at m/z 505.1130, calcd for 505.1135), which is 16 Dalton less than the value for compound 3. Comparison of the ¹H and ¹³C NMR spectral data (Table 2) indicates close similarities between 4 and 3, with the only difference occurring at C-11. The phenolic hydroxyl group at C-11 in compound 3 is replaced by an aromatic proton in compound 4, as confirmed by an HMBC correlation in compound 4 from δ_H 6.90 (H-11) to δ_C 22.4 (C-12). These findings revealed the structure of 4 to be that shown in Figure 1. The relative configuration at C-16 was again undetermined due to insufficient NOESY key correlations. The absolute configuration of 4 was still unconfirmed due to the absent of single crystal data and the CD signal.

The antibacterial and cytotoxic activity of compounds 1–4 was evaluated. The diastereomers—compounds 1 and 2—showed similar antibacterial activity against four strains: B. subtillis 769, S. haemolyticus UST950701-004, S. aureus UST950701-005, and S. aureus ATCC 43300, with MIC values of ≤1.0 μg/mL. Compounds 3 and 4 also displayed similar biological activity against these bacterial strains, except that compound 4 revealed no activity against S. aureus ATCC 43300. The cytotoxic activity of these four compounds was also assessed against HeLa cells by the standard 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay [24]. Compounds 1 and 2 displayed significant cytotoxic activity at nanomolar concentrations, whereas compounds 3 and 4 showed no activity (Table 3). Compounds 1 and 2 displayed much stronger antibacterial activity than compounds 3 and 4, suggesting that the five-member nitrogen heterocycle, which is absent from compounds 3 and 4, is essential for activity. The five-member nitrogen heterocycle is quite similar to that in oxazolidinones, which are an approved class of antibiotics [25,26]. The mode of action of compounds 1 and 2 with the five-member nitrogen heterocycle is currently being studied.

Four new xanthones were isolated from a marine-derived Streptomyces caelestis. Compounds 1 and 2, named citreamicin θ A and B, are a pair of diasteromers that possess a 5/6/6/6/6/6/6 seven-ring system. It is the first report of the citreamicins with such a unique ring system that is bending with an acute angle. The similar MIC values of citreamicin θ A (1) and B (2) to those of B. subtillis, S. haemolyticus, and S. aureus, and the MRSA strain indicate that these two diasteromers possess equal antibacterial activity. Citreaglycon A and dehydrocitreaglycon A showed weaker antibacterial activity than citreamicin θ A (1) and citreamicin θ B (2). The greater antibacterial activity of citreamicin θ A (1) and B (2) may be due to the five-member nitrogen heterocycle. The role of this nitrogen heterocycle in citreamicins merits further study.