Coibacins A-D, Anti-leishmanial Marine Cyanobacterial Polyketides with Intriguing Biosynthetic Origins

Abstract

Four unsaturated polyketide lactone derivatives, coibacins A-D, were isolated from a Panamanian marine cyanobacterium, cf. Oscillatoria sp. The two different types of termini observed in these co-occurring metabolites, either a methyl cyclopropyl ring as seen in curacin A or a methyl vinyl chloride similar to that observed in the jamaicamides, suggest an intriguing flexibility in the “beta branch” forming biosynthetic process. The coibacins possess selective anti-leishmanial activity as well as potent anti-inflammatory activity.

The secondary metabolites of filamentous marine cyanobacteria show a surprising degree of structural diversity and biological activity.¹ For example, a methyl cyclopropyl ring is found in the curacin series of anti-cancer metabolites isolated from Moorea producens (formerly Lyngbya majuscula²) (Figure 1, compound 5)³ whereas vinyl chloride moieties are present in the jamaicamides, isolated from a Jamaican collection of M. producens,⁴ as well as in most of the more than two dozen malyngamide natural products.⁵ Study of the biosynthetic mechanism of formation of these two moieties revealed they are variant products of closely related manifolds, differing only in the site of proton delivery following enoyl CoA hydratase mediated decarboxylation, and subsequently, enoyl reductase catalyzed hydride addition in the case of curacin A to form the cyclopropyl ring.⁶ Previously, these two distinctive structural features have never been reported from a single organism, and their co-occurrence here suggests an intriguing flexibility in the biosynthetic manifold.

Figure 1.

Structures of coibacins A-D (1-4) with two other marine cyanobacterial metabolites of related biosynthetic origin, curacin A (5) and jamaicamide A (6).

Prefractionation of the CH₂Cl₂:MeOH extract of a cf. Oscillatoria species collected in the Coiba National Park as part of the Panama International Cooperative Biodiversity Group (ICBG) resulted in several fractions with potent activity to axenic amastigotes of Leishmania donovani.⁷ Purification using RP-SPE followed by RP HPLC (C₁₈) gave coibacins A-D (1-4) as the active compounds (Figure 1). ¹H NMR analysis indicated that all four compounds contained an α,β-unsaturated-δ-lactone. However, metabolites 1 and 2 exhibited signals typical of a methyl substituted cyclopropyl ring, whereas compounds 3 and 4 showed signals diagnostic for methyl vinyl chloride moieties (Table 1).

Tablel.

NMR data for coibacins A-D (1-4) in CDCl₃

	Coibacin A (1)		Coibacin B (2)		Coibacin C (3)		Coibacin D (4)
no.	δc a,mult.	δH b,mult., J (Hz)	δC a,mult.	δH b,mult., J (Hz)	δC a,mult.	δh b,mult., J (Hz)	δC a,mult.	δH b,mult., J (Hz)
1	164.0 (C)		164.0 (C)		164.2 (C)		164.5 (C)
2	121.7 (CH)	6.05,c m	121.7 (CH)	6.05,c m	121.5 (CH)	6.03, m	121.5 (CH)	6.02, dt, 1.9, 9.7
3	144.5 (CH)	6.88, dt, 4.4, 9.8	144.5 (CH)	6.88, dt, 4.9, 9.8	144.8 (CH)	6.88, dt, 4.9, 9.7	144.9 (CH)	6.88, ddd, 3.6, 5.1, 9.8
4	29.8 (CH2)	2.45, m	29.8 (CH2)	2.45, m	29.4 (CH2)	2.35, m	29.4 (CH2)	2.33, m
5	77.9 (CH)	4.94, dt, 7.6, 7.8	77.9 (CH)	4.95, m	77.9 (CH)	4.43, m	77.9 (CH)	4.42, m
6	126.7 (CH)	5.64, dd, 6.7, 15.5	126.6 (CH)	5.64, dd, 6.7, 15.3	34.4 (CH2)	1.73, m 1.92, m	34.7 (CH2)	1.64, m 1.79, m
7	133.7 (CH)	6.30, dd, 10.7, 15.7	133.8 (CH)	6.31, dd, 10.5, 15.4	27.8 (CH2)	1.25, c m	24.3 (CH2)	1.39, m 1.51, m
8	129.1 (CH)	6.05,c m	129.0 (CH)	6.05,c m	131.0 (CH)	6.05, m	29.1 (CH2)	1.39, m
9	136.8 (CH)	5.77, dt, 6.6, 15.2	137.1 (CH)	5.78, dt, 6.6, 15.3	131.6 (CH)	5.57, m	32.2 (CH2)	2.03, dt, 6.7, 7.0
10	32.6 (CH2)	2.16, m	32.8 (CH2)	2.15, m	128.5 (CH)	5.51, m	126.6 (CH)	5.47, dt, 7.3, 15.1
11	32.1 (CH2)	2.16, m	32.0 (CH2)	2.07, m	132.3 (CH)	6.05, m	132.8 (CH)	5.34, dt, 6.9, 15.2
12	129.9 (CH)	5.50, dt, 7.3, 14.5	126.4 (CH)	5.43, dt, 6.9, 15.3	40.1 (CH2)	2.79, d, 7.1	40.2 (CH2)	2.72, d, 6.8
13	130.8 (CH)	5.97, m	134.3 (CH)	5.00, m	137.4 (C)		137.9 (C)
14	127.4 (CH)	6.05,c m	22.4 (CH)	1.03, m	16.7 (CH3)	1.78, s	16.6 (CH3)	1.75, d, 1.2
15	136.3 (CH)	5.15, dd, 9.4, 14.9	14.8 (CH2)	0.41, ddd, 4.7, 5.6, 8.2 0.48, ddd, 4.5, 4.5, 8.2	113.1 (CH)	5.82, br. s	112.6 (CH)	5.81, q, 1.5, 2.7
16	22.8 (CH)	1.07, m	14.8 (CH)	0.70, m
17	15.7 (CH2)	0.49, ddd, 4.7, 5.3, 8.2 0.55, ddd, 4.7, 4.7, 8.4	18.5 (CH3)	1.05, d, 6.2
18	15.7 (CH)	0.76, m
19	18.5 (CH3)	1.05, d, 6.0

^aMeasured at 100 MHz.

^bMeasured at 400 MHz.

^cObscured due to overlapping ¹H signals

High resolution mass measurement of the sodiated molecular ion for coibacin A (obs. m/z [M+Na]⁺ 307.1670; calc. m/z 307.1669) yielded a molecular formula of C₁₉H₂₄O₂ with 8 degrees of unsaturation. By COSY and HMBC, 1 was determined to possess two conjugated dienes separated by two methylene groups, and by coupling constant analysis (all J values > 12 Hz), the double bonds were entirely trans (Table 1). At one end, this linear chain was linked to the unsaturated lactone ring, and at the other to the methyl cyclopropyl ring, thus completing the planar structure. The absolute configuration of the α,β-unsaturated lactone ring was determined using circular dichroism. The CD of 1 exhibited a positive Cotton effect at λ 259 nm, which matches well with the S configuration of model α,β-unsaturated lactone rings.⁸ However, because coibacin A has other UV chromophores adjacent to chiral centers, and these could potentially contribute to the CD, we also meaured the CD of compound 4 (see below) as it lacks any chromophore other than the unsaturated lactone. The CD curves of 1 and 4 were nearly superimposible, thus substantiating our assignment of the configuration of C-5.

The relative configuration of the methyl substituted cyclopropyl ring of 1 was determined using NOESY correlations and J coupling constant analysis (Table 1). Correlations were observed between the adjacent vinyl proton (H-15, δ 5.15), one of the diastereotopic H₂-17 (δ0.55), and methine H-18 (δ 0.76). Additional correlations were found between methine H-16 (δ 1.07), the other H-17 (δ 0.49), and H₃-19 (δ 1.05). Taken together with the coupling constants between these protons (Table 1), these data indicated that coibacin A (1) possessed a trans configured methyl cyclopropyl ring (S*,S*). Unfortunately, due to the low yield of natural product, the absolute configurations at C-16 and C-18 were not assigned.

The unsaturated chain of coibacin B (2) was found to be two carbons shorter than in 1 by HRMS and NMR analysis. This deletion was localized to one less olefin in the diene motif adjacent to the methyl cyclopropyl ring. In all other respects, including optical rotation, the spectroscopic data for 2 were essentially identical to those recorded for coibacin A (1).

While coibacins C (3) and D (4) had spectroscopic features similar to coibacins A and B, they were clearly of a different structural series due to their molecular formulas, which both contained one chlorine atom. Coibacin C (3), of molecular formula C₁₅H₁₉ClO₂ (m/z [M+Na]⁺ 289.0968), possessed the same α,β-unsaturated-δ-lactone as 1 and 2, but oxymethine H-5 was approximately 0.5 ppm upfield, and by COSY was located adjacent to a methylene rather than an olefin.

COSY correlations sequentially extended this spin system to a second methylene (H₂-7), a conjugated diene (E,E by J values, see Table 1), and a bis-allylic methylene (δ 2.79, H₂-12). HMBC correlations connected the latter doublet to quaternary C-13, methyl C-14 and olefinic C-15 of an adjacent terminal olefin with β-chloro and α-methyl substituents. Comparison of the ¹H and ¹³C NMR shifts for 3 and for model compounds from the jamaicamide and malyngamide families of metabolites confirmed the identity of this moiety,^4,5 and nOe (H-15 to H₂-12) was used to define the E-geometry of this trisubstituted olefin. Coibacin D (4) was of two mass units greater than 3 (m/z [M+Na]⁺ 291.1124), and consistent with this was found to lack one of the two olefins forming the conjugated diene in 3; by COSY this saturation was localized to the C-8/C-9 position. The geometry of the Δ¹⁰ olefin was trans based on a 15.1 Hz vicinal J value, and the Δ¹³ olefin was assigned as E given very similar ¹H and ¹³C NMR chemical shifts to those of coibacin C (3). Coibacins C and D had optical rotations similar to that of coibacins A and B, and as noted above, coibacin D (4) had a CD absorption curve consistent with the S-configuration of an α,β-unsaturated-δ-lactone (see Supporting Information).

Purified coibacins A-D (1-4) were tested for activity in the Panama ICBG suite of tropical disease assays (Figure S36, Supporting Information). Coibacin A (1) presented potent activity against Leishmania donovani axenic amastigotes with an IC₅₀ value of 2.4 μM, while the other coibacins exhibited slightly less activity in this assay (Table 2). The anti-leishmanial activity of the coibacins was confirmed using L. mexicana axenic amastigotes (Figure S37, Supporting Information); however, using aL. mexicana macrophage assay they were inactive, perhaps indicating a failure to cross the cell membrane of these latter cell types.

Table 2.

Biological assay data for coibacins A-D (1-4) in anti-leishmanial, cytotoxicity, and nitric oxide assays (IC₅₀ values in μM).

	Leishmania donovani axenic amastigotes	cytotoxicity (NCI-H460 cells)	nitric oxide
Coibacin A (1)	2.4	31.5	20
Coibacin B (2)	7.2	17.0	5
Coibacin C (3)	18.7	21.3	11
Coibacin D (4)	7.8	11.4	21

The coibacins were also tested for activity against malaria and Chagas’ disease and were inactive (IC₅₀ values > 25 μM). In testing against NCI-H460 human lung cancer cells, coibacin D (4) was the most cytotoxic whereas coibacin A (1) was least cytotoxic (Table 2). From these data, tentative therapeutic indices against Leishmania parasites for 1-4 were calculated to be 13.1,2.4, 1.1, and 1.5, respectively; thus coibacin A possesses the greatest selectivity, and may be a lead for development of an antileishmanial drug candidate.

The coibacins were also tested for anti-inflammatory activity in a cell-based nitric oxide (NO) inhibition assay.⁹ In this assay, coibacin B (2) was determined to be the most active (Table 2); however, because coibacin A was more abundant, it was further evaluated in gene transcription (TNF-α, IL-6, IL-1b and iNOS) and protein expression studies. Coibacin A (1, at 10 μg/mL) was found to significantly reduce gene transcription of all four cytokines tested, with especially notable effects on IL-1β and iNOS (Figure 2). Using ELISA, changes in protein expression for some of these inflammatory cytokines were measured in murine RAW264.7 cells stimulated with lipopolysaccharide (LPS) in the absence or presence of the coibacins. Coibacin A (1, at 10 μg/mL) was found to significantly reduce TNF-α and IL-6 secretion (Figure S38 and S40, Supporting Information). Coibacins B-D (2-4) also affected protein expression of TNF-α and IL-6 albeit with slightly less potency (Figure S39 and S41).

Figure 2.

Transcription of pro-inflammatory genes after treatment with coibacin A (1) at 10 μg/mL (bars represent the mean ± standard deviation; N = 3. *P-value < 0.05 compared to LPS treatment alone. **P-value < 0.01 compared to LPS treatment alone).

The isolation of both cyclopropyl ring and methyl vinyl chloride moieties from the same organism suggests intriguing biosynthetic possibilities when compared to the metabolic pathways of curacin A (5) and jamaicamide A (6). The latter two pathways, originally located in two different source organisms, were found to have high sequence identity for the first three enzymatic steps involved in beta branch formation [HMG Co-A synthase (HCS), halogenase (HAL), and dehydratase (ECH₁)], and these steps are partially responsible for formation of the two different functional groups.⁶ There is considerably less sequence identity for the decarboxylase (ECH₂) and the enoyl reductase (ER) enzymes. The isolation of co-occurring metabolites containing these two unique structural moieties from a single organism may indicate variability in the site of proton delivery (α vs. γ) following ECH₂-catalyzed decarboxylation (Figure S42 Supporting Information). Subsequently, in the latter case of γ-proton delivery, ER-catalyzed hydride delivery to the β-carbon leads to cyclopropyl ring formation; no further reaction occurs in the former case of α-proton addition. Another structural feature distinguishing the coibacin A/B series from the C/D series is chain length following the β-branch manifold; thus, it is possible that the variable presence of either the cyclopropyl ring or vinyl chloride moieties determines subsequent PKS extension steps.

It should be noted that the cyclopropyl ring of the coibacins A and B are of trans stereochemistry, which contrasts with the cis cyclopropyl ring of curacin A. Thus, there must be additional, as yet unstudied points of divergence between these two biosynthetic pathways.

In summary, we report here the isolation and structure elucidation of coibacins A-D (1-4), polyketide metabolites possessing novel structures with provocative biosynthetic motifs. These cyanobacterial metabolites possess significant anti-leishmanial and anti-inflammatory activities, and thus, may be useful lead structures for drug development. Additionally, their co-occurrence gives insight into how Nature has developed alternative processing to produce cyclopropyl rings versus vinyl chloride moieties.