Sanctolide A, a 14-membered PK-NRP hybrid macrolide from the cultured cyanobacterium Oscillatoria sancta (SAG 74.79)

Abstract

Sanctolide A (1), a 14-membered polyketide-nonribosomal peptide (PK-NRP) hybrid macrolide, was isolated from the cultured cyanobacterium Oscillatoria sancta (SAG 74.79). The planar structure was determined using various spectroscopic techniques including HRESIMS, and 1D and 2D NMR analyses. The relative configuration was assigned by J-based configurational analysis in combination with NOE correlations. The absolute configuration was determined by Mosher ester and enantioselective HPLC analyses. The structure of sanctolide A (1) features a rare N-methyl enamide and a 2-hydroxyisovaleric acid, which are incorporated to form a 14-membered macrolide ring structure, comprising a new type of cyanobacterial macrolides derived from a PKS-NRPS hybrid biosynthetic pathway.

Cyanobacteria (blue-green algae) have been shown to be prolific producers of bioactive secondary metabolites.^1–3 Polyketide-nonribosomal peptide (PK-NRP) hybrids represent a major class of cyanobacterial secondary metabolites.⁴ Diverse cyanobacterial metabolites belonging to this class have been isolated from both freshwater and marine cyanobacteria. A major group of cyanobacterial PK-NRP hybrid metabolites is lipopeptides, where linear or cyclic peptides contain one lipophilic residue of polyketide origin such as β-amino acid and N-acyl residues.² Another subclass of cyanobacterial PK-NRP hybrid metabolites is comprised of macrolides whose building blocks are mainly acetates with one or two amino acids incorporated into their macrolide ring structures. These compounds are less commonly found in cyanobacteria. Examples include laingolide, laingolides A and B, madangolide and palmyrolide A, all of which have 15-membered macrolide rings containing a rare N-methyl enamide functionality. ^5–8 These metabolites were all obtained from marine cyanobacteria belonging to the order Oscillatoriales.

In our continuing search for biologically active secondary metabolites from cultured cyanobacteria, the cell extract of Oscillatoria sancta (SAG 74.79) was initially evaluated for its activity in a brine shrimp toxicity assay and found to be active. Herein, we report the isolation and structure determination of a 14-membered PK-NRP hybrid macrolide, named sanctolide A (1). The planar structure was determined using spectroscopic techniques including HRESIMS, and 1D and 2D NMR analyses. The relative configuration of 1 was solved by J-based configurational analysis along with NOE correlations, and Mosher ester and enantioselective HPLC analyses were carried out for the assignment of the absolute configuration.

Oscillatoria sancta was obtained from the SAG (Sammlung von Algenkulturen Göttingen) culture collection of algae (strain ID: 74.79), and grown in inorganic media (BG-12).⁹ The freeze-dried cells were extracted with a mixture of CH₂Cl₂-MeOH (1:1 v/v). This organic extract was fractionated using Diaion HP-20 resin with an increasing amount of iPrOH in water. The fractions eluting at 70% and 80% iPrOH displayed toxicity against brine shrimp. HPLC-based activity profiling of these fractions using reversed-phase column identified one minor peak as the active component (Figure S2 in Supplementary data). Dereplication by LC-MS and ¹H NMR indicated this peak to be a potentially new metabolite with a molecular weight of 439 Da. Scale-up culture (4 × 2L) and re-isolation was performed using the method described above followed by reversed-phase HPLC to yield sanctolide A (1, 2.6 mg, Figure 1).

Figure 1.

The structure of sanctolide A (1).

Sanctolide A (1)¹⁰ was obtained as colorless oil. The molecular formula of 1 was established as C₂₄H₄₁NO₆ by HRESIMS analysis. Analysis of the ¹H NMR spectrum of 1 in combination with the DEPT-Q and HSQC spectra indicated the presence of two olefinic protons (δ_H 6.70 and 5.12), three oxygenated methines (δ_H 5.11, 5.00 and 3.12), one O-methyl (δ_H 3.28), one N-methyl (δ_H 3.06), two methines (δ_H 2.56 and 2.31), five diastereotopic methylenes (δ_H 2.0 – 1.0), three homotopic methylenes (δ_H 1.28 and 1.24), as well as three dolublet (δ_H 1.13, 0.94 and 0.90) and one triplet methyl (δ_H 0.87) protons. Analysis of the COSY spectrum established three partial structures 1a, 1b and 1c (Figure 2). The first partial structure (1a) was assembled by sequential COSY correlations from H-2 (δ_H 2.56) to H₃-12 (δ_H 0.87) along with a COSY correlation between H-2 and H₃-13 methyl (δ_H 1.13). Down-field chemical shifts of H-5 (δ_H 5.00) and H-7 (δ_H 3.12) indicated that C-5 and C-7 were oxygenated. The second partial structure (1b) was established from COSY correlations between H-15 (δ_H 5.11) and H-16 (δ_H 2.31), and between H-16 and H₃-17/H₃-18 methyl protons (δ_H 0.94/0.90). A downfield chemical shift of H-15 (δ_H 5.11) and the absence of a NH signal indicated this partial structure to be a 2-hydroxyisovaleric acid. This was further supported by chemical shifts of C-15 (δ_C 77.0) and C-19 (δ_C 168.7).. COSY correlations between H-20 (δ_H 3.20 and 3.15)/H-21 (δ_H 5.12)/H-22 (δ_H 6.70) determined the last partial structure as a 1, 3-functionalized propylene (1c) possessing the trans geometry due to the large ³J_HH (14.0 Hz) observed between H-21 and H-22.

Figure 2.

Key 2D NMR correlations used to determine the planar structure of 1.

The three partial structures (1a, 1b and 1c) were assembled through HMBC correlations (Figure 2). An HMBC correlation from the O-methyl singlet (δ_H 3.28) to C-7 methine (δ_C 77.7) placed a methoxy group at C-7. The chemical shift of C-14 (δ_C 170.0) combined with HMBC correlations from H-5 (δ_H 5.00) and H-16 (δ_H 2.31) to C-14 allowed the assembly of partial structures 1a and 1b via an ester linkage. The connection between substructures 1b and 1c was established by HMBC correlations from H-15 (δ_H 5.11) and H-21 (δ_H 5.12) to C-19 (δ_C 168.7), positioning the 2-hydroxyisovaleric acid between 1a and 1c via two ester linkages. HMBC correlations from an N-methyl singlet (δ_H 3.06) to C-22 (δ_C 132.4) and C-1 (δ_C 174.5), and from H₃-13 (δ_H 1.13) to C-1 linked 1a and 1c through the N-methyl enamide bridge, completing the assembly of a 14-membered macrolide ring scaffold.

The relative configuration between C5 and C7 was determined by J-based configurational analysis¹¹ in combination with NOE correlations (Figure 3). Homonuclear (³J_HH) and heteronuclear (²J_CH and ³J_CH) coupling constants were obtained from the DQF-COSY and GBIRD-HSQMBC spectra, respectively.^{12, 13} A small ³J_H5H6a (3.8 Hz) and a large ³J_H5H6b (8.1 Hz) indicated the gauche conformation between H-5 and H-6a and the anti conformation between H-5 and H-6b, respectively, leaving A3 and A4 as two possible conformations (Figure S3 in Supplementary data). An NOE correlation observed between H₂-4 and H-6a indicated A4 to be the only possible conformation, allowing the stereospecific assignments of the diastereotopic protons H-6a/b (Figure 3). For the C6–C7 bond, only two conformations B1 and B5 were possible based on a large ³J_H6aH7 (9.0 Hz) and a small ³J_H6bH7 (2.9 Hz) (Figure S3 in Supplementary data). An NOE correlation observed between H-6b and H₂-8 and between H-5 and OMe left B1 as the only possible conformation, thus allowing the assignment of the anti configuration for the 1,3-methine system at C5–C7. Determination of the relative configuration for the 1,4-methine system between C-2 and C-5 using NMR methods was complicated as analysis of the NOE spectrum indicated the presence of multiple conformations for the macrolide ring in 1. The proton H-5 showed NOE correlations with all of four diastereotopic protons including H-3a/b and H-4a/b, and a NOESY experiment recorded with a short mixing time (200 ms) verified these cross-peaks to be real NOE correlations as opposed to artifacts arising from spin diffusion. A J-based configurational analysis was attempted but unsuccessful due to a number of medium J values observed for this 1,4 methine system. Attempts to crystallize 1 were also unsuccessful due to the labile nature of an enamide moiety and the presence of a 12 carbons-long fatty acid branch. Therefore, the configuration of a stereocenter at C-2 could not be assigned.

Figure 3.

Newman projections for A4: C5–C6 and B1: C6–C7. The DQF-COSY and GBIRD-HSQMBC spectra were used for the measurement of homo- and hetero-nuclear coupling constants. Labels in parentheses denote the predicted size of coupling constants: S-small, M-medium and L-large. No correlation observed in the HSQMBC spectrum was assigned as a coupling of 0 Hz. Observed nOe correlations are presented as arched arrows. A full list of rotamers can be found in Supplementary data.

The absolute configuration at C-5, C-7 and C-15 was established by Mosher ester and enantioselective HPLC analyses (Figure 4). ¹H NMR and LC-MS analyses indicated that acid hydrolysis of 1 resulted in the formation of 2 due to the labile nature of an enamide moiety under acidic conditions, similar as reported for palmyrolide A by Pereira et al.⁷ The isolation of 2 from the acid hydrolysate was necessary for Mosher ester analysis, but challenging due to the lack of a chromophore. Alternatively, solvolysis of 1 was carried out using NaOMe in MeOH, and the resulting product was dried in vacuo for 24 hrs. ¹H NMR and LC-MS analyses identified the presence of 2 as a sole product¹⁴ probably due to volatile nature of the other methyl ester products. The resulting solvolysis product was subjected to Mosher ester analysis. Two equal portions of 2 were derivatized with (R)- and (S)-MTPA chlorides at C-5 to yield (S)- and (R)-MTPA esters (2S and 2R), respectively.^{15, 16} Interpretation of the ¹H NMR chemical shift differences (Δδ_S-R) between 2S and 2R assigned the S configuration to C-5, thereby leading to the assignment of the absolute configuration as 5S,7S. The absolute configuration of the remaining 2-hydroxyisovaleric acid residue (3) was assigned by enantioselective HPLC analysis of the acid hydrolysate.¹⁷ Comparison of the retention times between the acid hydrolysate of 1 and two authentic standards L- and D-2-hydroxyisovaleric acids allowed the absolute configuration of 3 to be assigned as L.

Figure 4.

Methanolysis and acid hydrolysis for the absolute configuration of 1 at C-5 and C-15; Δδ_S-R values (Δδ_S-R = δ_S − δ_R) for the MTPA esters 2S and 2R.

As observed for other enamide-containing natural products,^5–8 compound 1 was highly labile and underwent rapid enamide hydrolysis in the presence of water or acid, resulting in an opening of the macrolide ring. Pereira et al. proposed the mechanism of enamide hydrolysis in acidic conditions.⁷ This eneamide hydrolysis can also occur in netural pH in a similar manner and involves the reversible conversion of the eneamide moiety into an iminium ion, which undergoes addition by water followed by proton transfer, leading to the formation of a secondary amide and an aldehyde (Figure 5). A degradation experiment indicated that the enamide of 1 was completely hydrolyzed in 48 hrs at rt upon addition of water at similar concentrations used in our biological assay systems, indicating that the macrolide ring structure would be degraded during biological evaluation (see Supplementary data). Sanctolide A (1) exhibited moderate toxicity to brine shrimp with an LD₅₀ value of 23.5 μM. Compound 1 was also evaluated for its cytotoxicity against the HT-29 and MDA-MB-435 cell lines as well as antibacterial activity against E. coli and S. aureus, but no activity was found in either assay system at the highest concentration tested (25 μg/mL). Based on the degradation observed above, it is suggested that compound 1 was present in the acyclic form during the biological evaluation, and this acyclic form of 1 was moderately toxic to brine shrimp.

Figure 5.

Proposed enamide hydrolysis of 1 in neutral pH that leads to ring-opening of the macrolide ring structure.

Sanctolide A (1) features a new class of cyanobacterial macrolides where two amino acid precursors Gly and Val are incorporated into a polyketide chain to form a 14-membered macrolide ring structure. The biosynthesis of 1 was proposed as follows: The first step of biosynthesis involves the formation of a hexaketide (C1 – C12) chain by six PKS modules with α-branched methyl (C-13), which is likely originated from S-adenosyl-L-methionine (SAM).¹⁸ This hexaketide chain is further expanded with N-methyl glycine by a NRPS module followed by ketide extension by one acetate unit, reduction, dehydration and double bond isomerization, resulting in the formation of the C21–C22 enamide functionality as proposed for the palmyrolide A.⁷ The 2-hydroxyisovaleric acid linked to the hydroxyl group at C-5 is likely formed by transamination of Val followed by reduction.¹⁸ The cyclization between C-15 hydroxyl and C-19 carboxyl group is likely to occur in a final step, thus completing the proposed biosynthesis of 1.

Table 1.

NMR spectroscopic data for sanctolide A (1) in CDCl₃

		sanctolide A (1)
		δCa,c	δHb,c	mult.(J in Hz)	COSYb	HMBCb	NOESYb,d
1		174.5
2		38.1	2.56	m	3, 13	1, 3, 4, 13	3b, 4a, 13, 22
3	a	29.2	1.21	m	2, 4	1, 2, 4, 5, 13	3b, 4b, 13
	b		1.68	m			2, 3a, 4a, 5
4	a	34.1	1.42	m	3, 5	2, 3	3b, 4b, 13
	b		1.86	m			3a, 4a, 5, 13
5		73.3	5.00	m	4, 6	3, 4, 6, 7, 14	3a, 3b, 4a, 4b, 6a, 6b, 7, OMe
6	a	40.3	1.59	ddd (14.4, 9.0, 4.2)	5, 7	4, 5, 7, 8	4a, 4b, 6b, 7, OMe
	b		1.74	ddd (14.4, 7.8, 3.0)			6a, 5, 7, 8a, 8b
7		77.7	3.12	m	6, 8	5, 6, O-Me	5, 6a, 6b, 7, 8a, 8b
8	a	33.2	1.37	m	7, 9	6, 7, 9, 10	6a, 6b, 8b
	b		1.50	m			6b, 7, 8a, OMe
9		32.0	1.24	m	8, 10	7, 8, 10, 11
10		24.3	1.24	m	9, 11	8, 9, 11, 12
11		22.6	1.28	m	10, 12	9, 10, 12
12		14.2	0.87	t (7.1)	11	10, 11
13		14.9	1.13	d (6.9)	2	1, 2, 3	2, 3a, 4a, 4b
14		170.0
15		77.0	5.11	d (6.4)	16	14, 16, 17, 18, 19	16, 17, 18
16		29.6	2.31	sd (6.8, 6.4)	15, 17, 18	14, 15, 17, 18	15, 17, 18
17		18.6	0.94	d (6.8)	16, 18	15, 16, 18
18		17.2	0.90	d (6.8)	16, 17	15, 16, 17
19		168.7
20		34.5	3.15	dd (6.5, 1.3)	21	19, 21, 22	21, 22
			3.20	d (6.9)
21		102.7	5.12	m	20	19, 20, 22	20, 22, NMe
22		132.4	6.70	d (14.0)	21	1, 20, N-Me	2, 3b, 20, 21, NMe
O-Me		56.6	3.28	s		7	5, 6a, 8a, 8b, 17
N-Me		30.6	3.06	s		1, 22	21, 22

^aassigned using the DEPT-Q spectrum recorded at 226 MHz,

^brecorded at 600 MHz,

^cchemical shifts were referenced to the CDCl₃ solvent signals (δ_H 7.24 and δ_C 77.2)

^dmixing time: 600 ms