Tradition of eating cyanobacteria is seen in the coastal area of Lake Chad, east and southeast Asian countries, and among Peruvian highlanders.1 In Japan, consumption of Nostoc commune, N. verrucosum and Aphanothece sacrum (‘suizenji-nori’ in Japanese) has been reported, only the last of which continues on a commercial basis.2 Because cyanobacteria are known as a prolific source of bioactive secondary metabolites,3 edible but less investigated species hold promise for new drug leads.

As part of our program to evaluate the biomedical potential of edible cyanobacteria, we examined the ethanolic extract of A. sacrum and discovered sacrolide A as an antimicrobial principle.2 It belongs to the oxylipin class metabolites, which are enzymatically or chemically derived fatty acid peroxidation products,4 and from cyanobacteria only four precede it which are as follows: malyngic acid,5 mueggelone,6 (9R, 10E, 12Z, 15Z)-octadecatrienoic acid7 and (9R, 10E, 12Z)-octadecadienoic acid.7 Although oxylipins are known to function as intra- and interspecific signaling molecules or as microbicides in plants,4 their role in prokaryotes is practically unknown. To gain further insights into the function and structural diversity of prokaryotic oxylipins, congeners of sacrolide A (Figure 1) were pursued in A. sacrum, which resulted in the isolation of two new congeners, 9-epi-sacrolide A (1) and 15,16-dihydrosacrolide A (2).

Figure 1
figure 1

(a) Structure of sacrolide A, 9-epi-sacrolide A (1) and 15,16-dihydrosacrolide A (2). (b) Selected COSY/TOCSY (bold lines) and HMBC (arrows) correlations for 1.

Full size image

The ethanolic extract of the algal colonies was partitioned between dichloromethane and 60% aqueous MeOH, and the former between n-hexane and 90% aqueous MeOH. The antimicrobial principles in the latter layer were purified by open column chromatography on silica gel, gel filtration on Sephadex LH-20, normal phase HPLC on silica gel to yield crude sacrolide A, as previously described. NMR-based examination of the peaks eluting before and after sacrolide A revealed the presence of its congeners, which prompted further purification of both fractions by reversed-phase HPLC to yield 1 and 2, respectively.

Compound 1 has the same molecular formula as sacrolide A (C18H30O4) as evidenced by a molecular ion peak at m/z 331.1864 ([M+Na]+, C18H28NaO4, Δ-1.6 mmu) in the HRESITOF-MS analysis.2 The 1H NMR spectrum of 2 (Supplementary Table S4) assorted all signals from sacrolide A,2 some of which though shifted upfield and changed shape, indicating that 1 is either a diastereomer or transesterified isomer of sacrolide A. Indeed, interpretation of a full set of NMR spectra supported this idea: the presence of spinsystems H22/H23/H24, H26/H27/H28/H9/H10/H11 and H13/H214/H15/H16/H217/H318 was evident from the sequence of COSY and TOCSY crosspeaks (Figure 1); the 10E and 15Z geometries from 3JH10-H11=15.8 Hz and 3JH15-H16=~10 Hz; placement of carboxylate at C1 from HMBC correlations H2/C1 (δH 2.56/δC 172.8 Figure 1); connectivity between C4 and C6 through C5 from HMBC correlations H4/C5 (δH 1.32/δC 27.0), H4/C6 (δH 1.32/δC 25.4) and H6/C5 (δH 1.36/δC 27.0); interruption of ketone between C11 and C13 from HMBC correlations H11/C12 (δH 6.47/δC 196.8) and H13/C12 (δH 5.13/δC 196.8). Thus, an acyloxy backbone with an α,β-unsaturated-α’,γ-ketodiol motif, present in sacrolide A, was assembled. Formation of an ester linkage between C1 and C9 was denied by the fact that 1 and sacrolide A both gave the same oxidation product (see below), leaving diastereomeric structure as the only possibility. Difference in a coupling constant between H9 and H10 (1, 7.6 Hz; sacrolide A,2 4.7 Hz) and an inverted intensity of a pair of NOESY correlations H9/H10 and H9/H11 (1, weak and strong; sacrolide A, strong and weak: see S14 and S15) endorsed this assignment.

To specify which of the two stereocenters (C9 or C13) is epimerized, 1 and sacrolide A were separately oxidized with MnO2 and CD spectra of the resulting diketone products 3 was compared, which retain the configuration of C13 from the mother compounds (Supplementary Figure S5). Because both 3 exhibited a weak but obvious negative CD Cotton effect, they were enantiomerically identical. Considering the (9R,13R) configuration of sacrolide A, 1 was concluded to have a (9S, 13R) configuration.

Compound 2 gave a molecular ion at m/z 333.2002 [M+Na]+ (C18H30O4, Δ−3.4 mmu) in an ESITOFMS measurement, which is larger by two hydrogens than sacrolide A. The 1H NMR spectrum of 2 was reminiscent of sacrolide A, as represented by the keto-conjugated E-olefin (H10: δH 6.95 dd, J=4.7 and 15.8 Hz, H11: δH 6.57 dd, J=1.4 and 15.9 Hz, Supplementary Table S6) and adjacent two oxymethine (H9: δH 4.50 m, H13: δH 5.13 dd, J=6.7 and 6.9 Hz) signals. However, signals for two olefinic methines (δH 5.30 and δH 5.54) and allylic methylenes (δH 2.61, 2.56, and 2.06) were missing, implying the saturation of the side chain in 2. This assignment was confirmed by HMBC correlations from a methyl proton (H18: δH 0.89) to two aliphatic carbons (C16: δC 31.4 and C17: δC 22.4). A closer similarity of the 1H NMR chemical shift values and coupling constants to those of sacrolide A, rather than of 1, verified the same conformation and relative configuration to sacrolide A. The positive sign of the optical rotation was also the same (2: [α]D24.1 +4.9, c 0.038: sacrolide A: [α]D24.1 +24.5, c 0.392, both in MeOH), supporting the (9R,13R)-absolute stereochemistry. Thus, the structure of 2 was concluded to be a 15,16-dihydro derivative of sacrolide A.

Compound 1 was inactive against four microorganisms sensitive to sacrolide A (Staphylococcus aureus, Streptomyces lividans, Saccharomyces cerevisiae and Penicillium chrysogenum) at the tested concentration range, while 2 was marginally active against eumycetes (Table 1). These results demonstrate that the 9S-hydroxy group is crucial for the antimicrobial activity of sacrolide A but unsaturation at the aliphatic tail is less so against eumycetes. A substantial number of oxylipins bear an α,β-unsaturated ketone motif and are known to disable a certain set of redox-controlled proteins by modifying the active site sulfhydryl groups.8, 9 Sacrolide A belongs to the same oxylipin class, but loss of the activity upon epimerization of C9 suggests that it is more of a target-specific inhibitor than promiscuous Michael acceptor.10

Table 1 Comparison of the antimicrobial activity among sacrolide A congeners (MIC μg ml−1)
Full size table

Conservation of an S-configuration at C13 among sacrolide A, 1 and 2 implies that hydroxylation of this carbon is an enzymatic process. In contrast, variation in the chirality of C9 carbinol and the degree of unsaturation at C15 is allowed, underpinning a non-enzymatic formation of α,β-unsaturated-γ-ketol from 12,13-allene oxide precursors (Supplementary Scheme S7).11