Structural diversity in nudibranch chemistry: elucidation of norditerpenes with a dendrillane scaffold from the Australian nudibranch Goniobranchus coi

Introduction

Goniobranchus coi (Risbec, 1956) is a nudibranch species with distinctive purple markings that is found in tropical Western Pacific locations. In common with other Goniobranchus nudibranchs, these colourful marine animals have a distinctive crawling motion in which the mantle edge flaps rhythmically up and down. According to Rudman and Bergquist,¹ ‘Goniobranchus coi is a sponge predator known to feed upon Chelonaplysilla violacea of the Darwinellidae family. It separates and concentrates the noxious distasteful antifeedant chemicals produced by the sponge and contained in its tissues, into the previously mentioned mantle glands for its own defensive purposes.’² This statement is backed up by photographs on the website SeaSlug Forum that show the mollusc feeding on Chelonaplysilla violacea.³ Although the chemistry of C. violacea has been well documented by Bobzin and Faulkner,^4,5 there is limited published data on the secondary metabolite composition of G. coi. Dendrillolide A (1) is the major metabolite of C. violacea from Palau,⁵ and a minor component in New Zealand collections.⁶ Dendrillolide A (1) was not reported from C. violacea specimens collected from New South Wales, Australia.^7,8

Our research group previously analysed the chemistry of a single specimen of G. coi collected from Mackay, north-east Queensland, and identified the known norditerpenes dendrillolide A (1), norrisolide, macfarlandin E, cheloviolene C and chromolactol.⁹ We then examined the chemistry of two specimens of G. coi collected from the Percy Isles and Coolum in Queensland and identified 11 known oxygenated terpenes. These two specimens contained a suite of oxygenated metabolites, including a 5,9-epoxy derivative 2 and a (nor)ketone 3, both of which exhibited extensive line broadening in their ¹H NMR spectra, and whose structures and stereochemistry were therefore solved by application of variable temperature NMR together with density functional theory (DFT) calculations.¹⁰ The current report extends the chemical description of these two specimens of G. coi through structural and stereochemical investigation of additional oxygenated metabolites (Fig. 1). In addition to 10,20-epoxide-functionalised structures 4 and 5, we identified aldehyde rearrangement products 6 and 7 that could be linked to both the epoxide metabolite(s) and the previously reported (nor)ketone 3.¹¹ Both specimens were also found to contain lactol analogues 8 and 9 related to dendrillolide A. Our findings provide an insight into the chemical diversity of G. coi and its dietary relationship with C. violacea.

The nudibranch specimens were collected under permit, with specimen number 1426 collected by Dr A Roberts-Thomson (Salty Pets Pty Ltd, Qld) and specimen number 1647 collected by Gary Cobb (permit number 183990, General Fisheries, Qld).

Fig. 1.

Structures of diterpenes 1–9.

Results and discussion

Two specimens of G. coi were collected from Percy Isles (2015) and Coolum (2017), Australia, and chemical extracts of the mantle and gut tissue prepared and analysed as described previously.¹¹ Metabolite 4 was isolated from both the mantle and viscera as a colourless oil and displayed a sodiated molecular ion peak in the high-resolution electrospray ionisation mass spectrum (HRESIMS) at m/z 415.2083 [M + Na]⁺, corresponding to a molecular formula of C₂₂H₃₂O₆. Comparison of the ¹H NMR and ¹³C NMR spectroscopic data (Table 1) of 4 with those of dendrillolide A (1)^12,13 supported a dendrillane-derived skeleton. Gradient correlation spectroscopy (gCOSY) correlations confirmed the H₂-1/H₂-2/H₂-3 and H-9/H-5/H₂-6/H₂-7 spin systems of a perhydroazulene fragment shown to be linked to the 2,8-dioxabicyclo[3.3.0]octane ring system through heteronuclear multiple bond correlations (HMBC) from H-14 to C-7 (δ_C 38.4), C-8 (δ_C 47.7) and C-9 (δ_C 53.4) (Fig. 2). In the 2,8-dioxabicyclo[3.3.0]octane ring system, gCOSY connectivities linked H-14 (δ_H 2.70) to both H-13 (δ_H 3.05) and H-15 (δ_H 6.38), as well as H-13 to H-16 (δ_H 5.98). There were HMBC correlations from H-15 to C-14 at δ_C 56.2 and the acetate carbonyl at δ_C 169.5, as well as correlations from H-16 to C-15 (δ_C 96.9), C-13 (δ_C 42.4), C-12 (δ_C 29.1) and the lactone at δ_C 175.7 (C-11). The characteristic exomethylene signals of dendrillolide A at δ_H 4.83 (d) and 4.57 (d) were replaced with signals typical of an epoxide at δ_H 2.64 (d, J = 5.2 Hz) and 2.40 (dd, J = 5.2, 1.5 Hz). HMBC correlations from the epoxide protons (H₂-20) to signals at δ_C 37.2 (C-1), 53.4 (C-9) and 61.2 (C-10) established the spiroepoxide at the C-10 position. Accordingly, 4 was named 10,20-epoxydendrillolide A.

Fig. 2.

Selected COSY and HMBC correlations for 10,20-epoxydendrillolide A (4).

Nuclear Overhauser Enhancement spectroscopy correlations (NOESY) between H-5/H-9, and between H-13/H-16, H-14/H-16, and H-13/14 were fully consistent with the relative configuration of dendrillolide A (1). Although NOESY correlations were observed from the epoxide proton at δ_H 2.40 (H-20b) to Me-19 (δ_H 0.99) and H-9 (δ_H 1.73), inspection of molecular models revealed that the configuration at C-10 could not be conclusively assigned from the NOESY data. Two candidate stereostructures, 4 or its C-10 epimer 5, were considered. A Monte Carlo Multiple Minimum (MCMM) conformational search was undertaken using MacroModel (ver. 2018-1, Schrödinger Inc., New York, NY, USA) providing eight conformers for 4 and 12 conformers for 5. The lowest energy conformers (<3 kcal mol⁻¹ of the global minimum) were next optimised by DFT calculations using the B3LYP/6-31G(d,p) method with the integral equation formalism polarisable continuum model (IEF-PCM) for chloroform solvent (Gaussian, ver. 16W, revision B.01, Gaussian, Inc., Wallingford, CT, USA), converging to six and five conformers for 4 and 5 respectively. The two lowest energy conformers of 4 were within 0.07 kcal mol⁻¹ of the global minimum and thus represented >75.6% of the conformational population. These two conformers both had a chair conformation for the cycloheptane ring and differed only in the orientation of the 2,8-dioxabicyclo[3.3.0]octane moiety relative to the perhydroazulene fragment. The two conformers of 4 showed a C(13)–C(14)–C(8)–C(9) dihedral angle of 68.5 and 173.2° respectively. The lowest energy conformer for 4 represented 40.3% of the conformational population. The lowest energy conformer of 5, which represented >84.6% of the conformational population, had a chair conformation for the cycloheptane ring and a C(13)–C(14)–C(8)–C(9) dihedral angle of 175.7° (Fig. 3). The free energies of 4 and 5 were used to calculate the Boltzmann-weighted ¹H and ¹³C NMR chemical shifts in chloroform solvent (IEF-PCM) (see Tables S1–S4 in the Supplementary material). The calculated chemical shifts were then examined using the modified DP4 computational approach developed by Grimblat et al. to assign the most probable diastereomer.¹⁴ Using both the ¹H and ¹³C NMR data, the probability was 99.99% that the C-10 configuration of 4 was R.

Fig. 3.

Lowest energy conformer of (a) (10R)-spiroepoxide 4 and (b) (10S)-spiroepoxide 5.

An aldehyde product 6 was purified as a colourless oil from both the mantle and the viscera and displayed a sodiated molecular ion peak in the HRESIMS at m/z 415.2092 [M + Na]⁺. The corresponding molecular formula of C₂₂H₃₂O₆ was identical to that of spiroepoxide 4. Comparison of the ¹H NMR spectroscopic data (Table 1) of 6 with those of dendrillolide A (1)^12,13 again supported a dendrillane-derived skeleton, except that the exomethylene signals of dendrillolide A (1) were replaced by a methine signal at δ_H 2.55 (m) and an aldehyde methine at δ_H 9.57 (d, J = 4.4 Hz, H-20). HMBC correlations from the aldehyde proton (H-20) to the signals at δ_C 30.3 (C-1), 49.5 (C-9) and 52.8 (C-10) established the position of an aldehyde at the C-10 position. By inspection of NOESY and coupling constant data, the relative configurations of the perhydroazulene and 2,8-dioxabicyclo[3.3.0]octane moieties were found to be consistent with a dendrillolide A-like scaffold. Signals for aldehyde 6 were not observed in the ¹H NMR spectrum of the crude extract and were only observed after the sample was purified by silica chromatography, implying that 6 was a rearrangement product formed during the purification process.

Epoxides isomerise into carbonyl compounds under mild acid conditions (House–Meinwald rearrangement) by hydride migration across the unhindered face of the epoxide, antiperiplanar to the breaking C–O bond.^15,16 As a result of this preference, an epoxide with 10R configuration was expected to convert into an aldehyde with 10R configuration, whereas the diastereomeric S-epoxide would convert into the 10S aldehyde.

To establish the stereochemical relationship between epoxide 4 and aldehyde 6, a sample of dendrillolide A (1) was reacted with m-chloroperbenzoic acid (mCPBA) in a biphasic solution with phosphate buffer and DCM for 3 days, yielding a mixture of the 10R and 10S spiroepoxides (Scheme 1). From the ¹H NMR data of the crude reaction mixture, the signals for the major component matched those of the 10R-epoxide 4 whereas the signals of the minor epoxide were in agreement with the computationally derived data for 10S-epoxide 5. During acquisition of 2-D NMR data, the signals for the epoxide diastereomers disappeared, and aldehyde signals became apparent. Fortunately, the degradation process was relatively slow, so the initially collected heteronuclear single quantum correlation (HSQC) and HMBC data provided sufficient information for full characterisation of the minor epoxide 5 from the reaction mixture (Table 1). Re-inspection of the ¹H NMR data then revealed evidence of traces of epoxide 5 in the G. coi crude extract.

Scheme 1.

Chemical correlation of spiroepoxides 4 and 5 and aldehyde products 6 and 7 following epoxidation of dendrillolide A (1) with mCPBA.

Subsequent normal-phase high-performance liquid chromatography (NP-HPLC) (25% EtOAc in hexanes) of the mCPBA reaction mixture afforded directly two aldehyde products, one of which had ¹H NMR data that matched those for the isolated aldehyde from G. coi. Inspection of a model of 4 revealed the epoxide protons are orientated towards the centre of the cycloheptane ring, placing them in proximity (2.65 Å) to one of the geminal dimethyl groups (Me-19). This steric hindrance shields the epoxide from undergoing rapid rearrangement to form the 10R-aldehyde 7. By contrast, in 5 the epoxide protons are orientated outward from the cycloheptane ring, placing an epoxide proton in proximity to Me-17 (2.54 Å), and the epoxide can undergo isomerisation to the 10S-aldehyde product 6. Consequently, it was inferred that both epoxides were originally present in the crude extracts of G. coi, however, when the extract was exposed to silica chromatography, the 10S-epoxide 5 isomerised to form the 10S-aldehyde product 6. On inspection, traces of aldehyde 7 were also detectable in the ¹H NMR spectrum of the crude extract. The specific optical rotation measurement of the major aldehyde reaction product gave a value of [α]²¹_D +11 (c 0.042, CHCl₃) that was comparable to that of the isolated aldehyde artefact 6 [α]²¹_D +16 (c 0.055, CHCl₃). The sample of the 10R-aldehyde reaction product 7 degraded on storage to the (nor)ketone, 10-oxonordendrillolide A (3), which we previously reported from G. coi¹¹; the specific optical rotation of 7 could not therefore be measured.

In a recent study on plant sesquiterpenes,¹⁷ Bombarda et al. reported that oxidation of an alloaromadendrene using mCPBA produced two diastereoisomeric epoxides, which differed in configuration at the C-10 position, as well as two aldehyde diastereomers as rearrangement products. Likewise, they found that the minor 10R epoxide was more labile than the major 10S epoxide, rapidly isomerising to an aldehyde during column chromatography. Detection of the epoxides 4 and 5, together with their aldehyde and ketone rearrangement products, brings into question their natural product status, given that the extract had been stored at −4°C for 6 months prior to detailed investigation. 1,1-Disubstituted alkenes are known to be unreactive substrates for chemical epoxidation relative to other alkenes.¹⁸

Lactols 8 and 9 were isolated from both the mantle and the viscera as an inseparable mixture by NP-HPLC, but the key individual signals could be identified in the ¹H NMR spectrum (Table 2); integration of the two sets of signals indicated a 1:2 ratio of 8 and 9 respectively. The colourless oil displayed a sodiated molecular ion peak in the HRESIMS at m/z 357.2045, corresponding to a molecular formula of C₂₀H₃₀O₄. Each terpene showed exomethylene signals (8: δ_H 4.82 (d) and 4.59 (d); 9: δ_H 4.81 (d) and 4.53 (d)), three quaternary methyl signals (8: δ_H 0.94 (s, Me-18), 0.96 (s, Me-17) and 0.97 (s, Me-19); 9: δ_H 0.95 (s, Me-18), 0.97 (s, Me-19) and 1.04 (s, Me-17)) and the diagnostic bridging methine proton (8: δ_H 2.88 (d, J = 8.9 Hz, H-9); 9: δ_H 2.70 (d, J = 9.3 Hz, H-9)), indicative of a perhydroazulene ring system. For each terpene, gCOSY connectivities and HMBC correlations were fully consistent with perhydroazulene and 2,8-dioxabicyclo[3.3.0]octane ring systems and with a lactol group at C-15, as observed in the metabolites cheloviolene A and cheloviolene B.¹⁹ The relative and absolute configuration of cheloviolene A was recently confirmed by single-crystal X-ray analysis, following an 11-step enantioselective total synthesis carried out by the Overman group. This synthetic work also revised the relative configuration at C-13, C-14 and C-16 of cheloviolene B.^20,21

The ¹H NMR data for 8 and 9 did not match that of the cheloviolenes, however, in lactol 8, the coupling constants of H-12a (dd, J = 18.1, 9.3 Hz), H-12b (dd, J = 18.1, 9.8 Hz), H-15 (d, J = 7.0 Hz) and H-16 (d, J = 4.8 Hz) were in accordance with those observed for dendrillolide A (1), suggesting an identical configuration of the 2,8-dioxabicyclo[3.3.0]octane ring system, whereas for 9, the coupling constant of H-15 (d, J = 3.7 Hz) suggested a change in configuration at C-15. Metabolites 8 and 9 were named dendrillactol and 15-epidendrillactol respectively. Following treatment of the lactol mixture 8 and 9 with acetic anhydride and pyridine, the single acetylated product formed was identified as dendrillolide A (1). The specific optical rotation measurement of the reaction product gave a value of [α]²¹_D +26 (c 0.02, CHCl₃); for comparison, dendrillolide A (1) isolated from G. coi has [α]²¹_D +61 (c 0.88, CHCl₃) compared to a literature value [α]²¹_D +87 (c 0.31, CHCl₃).^12,13 A detailed molecular modelling and computational study involving DP4/DP+ probability calculations, comparison of experimental vs calculated chemical shifts and of coupling constant values further supported the proposed stereochemistry (see Fig. S20 and Tables S5–S8 Supplementary material for details).

The crude extracts of the two specimens investigated were compared by their ¹H NMR spectra. A number of metabolites were found in both the mantle and viscera tissues, including epoxides 2, 4 and 5, the ketone 3, the lactols 8 and 9 plus cheloviolenes A and B, as well as the major metabolite dendrillolide A (1). The only differences between the two tissue types were that the small amounts of dendrillolide C detected were found in the mantle tissue alone, whereas norrisolide, cheloviolene C, macfarlandin D and polyrhaphin C were only detected in the viscera. This general pattern of similar metabolite distribution in both tissue types mirrors that detected in G. hunterae and G. verrieri, and in a previous report on a specimen of G. coi collected from Mackay,⁹ for all of which there was similar metabolite content in both mantle and visceral tissues. By contrast, in six other species (G. tinctorius, G. tasmaniensis, G. collingwoodi, G. splendidus, G. aureopurpureus and Goniobranchus sp. 1),²² the terpene metabolites detected were predominantly localised in mantle tissues, where they may collectively function as an anti-predator defense. Notably three of these six species contain spongian-16-one metabolites.

Conclusion

Chemical correlation experiments coupled with computational studies led to the determination of the relative configuration of spiroepoxide 4 and of aldehyde 6. The investigation into characterising the mixture of lactols 8 and 9 isolated from G. coi, successfully assigned the relative configuration at C-15 as 15S and 15R respectively, by spectroscopic analysis and chemical conversion into dendrillolide A (1), as well as by modelling and computational analysis. These studies illustrate how combining experimental NMR-derived and computational data can support the configurational analysis of conformationally flexible molecules.

Experimental

Supplementary material

An image of Goniobranchus coi (Fig. S2); copies of 1-D and 2-D NMR spectra of compounds 4–9 (Fig. S5–S19); details of the molecular modelling and computational calculations (Tables S1–S8) are available as Supplementary material. The Supplementary material (Fig. S1–S20, Tables S1–S8) is available online.