Structure Revision of Spiroleucettadine, a Sponge Alkaloid with a Bicyclic Core Meager in H-Atoms

Abstract

Our 2004 disclosure of the amino hemiketal-containing spiroleucettadine was met with keen interest by the natural products and synthetic communities. As repeated efforts to synthesize spiroleucettadine failed and questions regarding the original structure elucidation process arose, evidence mounted against the validity of the proposed structure. The low ratio of H/C in the core of spiroleucattadine complicated the original structure elucidation process. Speculation prompted a re-isolation of spiroleucettadine from an untouched portion of the original Luecetta collection and a thorough analysis of analytical data. In addition, a systematic analysis of candidate structures was performed via density functional theory (DFT) calculations; a favored high scoring structure 1b was ultimately confirmed to be spiroleucettadine via X-ray analysis of crystalline spiroleucettadine and reinforced the validity of DFT calculations in structure elucidation. We present the revised structure of spiroleucettadine, a bicyclic sponge alkaloid with a scarcity of H-atoms in its core.

Introduction

It has been somewhat surprising to observe the global reaction to our 2004 disclosure of a 2-aminoimidazole containing alkaloid, spiroleucettadine (1a).¹ This heterocycle, concluded to possess an uncommon trans-fused imidazolidine-oxalane, must now be revised to structure 1b, shown in Figure 1, based on the new findings reported herein. The compound was originally isolated from a Fijian Leucetta calcareous sponge and we outlined its structural features as being unprecedented. Our report tantalized many laboratories to further explore the properties associated with such a remarkable scaffold. Three nearly simultaneous disclosures by groups in the US, Canada, and Australia each concluded it was not possible to use a ring closure reaction in synthesis of structure 1a.^2-4 As an important counter point, the X-ray crystal structure of ergogaline⁵ firmly shows that an amino hemiketal embedded in a polycyclic ring system represents a stable moiety.

Figure 1.

Original and revised structure of spiroleucettadine.

Prompted by these synthetic failures, Watson et al. proposed structural alternatives 2 and 3, shown in Figure 2, and deftly used density functional theory (DFT) calculations to validate their postulate.⁴ We were inspired by this accomplishment but were unsettled by the result. While the calculation analysis for structure 1a provided a poor fit to the experimental data, structure 2 was also far from a match. Even though the DFT results for candidate 3 reasonably agreed with the empirical data, in comparison to the large family of Leucetta class compounds,^6-8 a C-methyl group did not seem biosynthetically viable. We eventually concluded that our structure elucidation was equivocal and complicated by the circumstance of an unfavorable H/C ratio⁹ associated with its bicyclic core, which also contained a plethora of heteroatoms. The results of our further investigation of several alternative structures for spiroleucettadine are outlined below.

Figure 2.

Watson’s suggested structures 2 and 3.

Results and Discussion

The biogenetic-based assumption that the three nitrogen atoms of spiroleucettadine (C₂₀H₂₄N₃O₄) were associated with a 2-aminoimidazole¹⁰ was an attractive and logical assignment but ultimately proved to be incorrect. We embarked on two parallel approaches to obtain further clarifying data. The first involved a successful re-isolation of spiroleucettadine to provide additional authentic compound. Especially fruitful was obtaining a crystalline sample for X-ray analysis, providing the structure 1b, shown in Figure 3. The second involved the drafting, prior to obtaining the crystal, 15 additional structural and stereoisomers as candidates (see Figure S7). These were all evaluated by DFT calculations and the results favored structure 1b.

Figure 3.

X-ray crystal structure of spiroleucettadine.

In view of the previous errors, more convincing data was sought to further affirm the revised structure 1b by obtaining a new sample. Our repository contained two distinct morphological types of Leucetta sponges that were considered as starting points for re-isolation efforts. One set possessed a lemon yellow color, oval shape and regular surface (Figure S1a) versus the other with an atypical highly corrugated surface (Figure S1b). No collections of the smooth surface type have ever yielded spiroleucettadine, so we focused on an untouched portion (170 g dry weight) of the original sample (coll no. 00111 obtained from Fiji). Extraction with MeOH followed by partitioning of the concentrated extract between dichloromethane and water afforded the crude organic phase (692 mg) used in the next step. Repetitive reversed-phase chromatography afforded 3.6 mg of spiroleucettadine that exhibited ¹H and ¹³C NMR properties in accord with the published values.¹

The re-isolated compound was the starting point for the next step that involved obtaining crystalline material. This began by dissolving 0.9 mg of spiroleucettadine in 300 μL of CDCl₃ for a vapor diffusion experiment employing hexanes as the co-solvent. After two weeks at -7 °C several very small colorless plate-like crystals of spiroleucettadine formed that were suitable for X-ray analysis, but this required using an advanced light source facility. The result of this examination (shown in Figure 3) was that the diffraction pattern obtained was consistent with structure 1b. Highlights of the crystallographic study are as follows. The compound crystallizes as a network of four hydrogen-bonded molecules, in the centric monoclinic space group P2₁/c (Figure S6). Surprisingly, the centrosymmetric nature of the space group indicates the presence of a racemic mixture of molecules. The optical rotation data (in CH₃OH) are relevant and include [α]_D = -27.1 of our original sample¹ and [α]_D = -5.1 of the new sample. It appears that scalemic mixtures of spiroleucettadine are being isolated. This phenomenon is not unique and bears remarkable similarity to the racemic crystal structure observed for pericosine E.¹¹ The biosynthetic basis for this observation is unclear yet a number of racemic and scalemic mixtures of natural products exist in the literature.^12-14

The new 2D NMR data obtained on spiroleucettadine verified a critical inconsistency noted by Watson.⁴ The correlation we reported¹ from H-8 to C-6, in Figure 4 for 1a in methanol-d₄, was actually due to a solvent J_CHpeak. This embarrassing error cast serious doubt on the validity of the amino hemiketal functional group shown in structure 1a. Another unsettling element of the original 2D NMR data set was the lack of parity displayed between the two N-methyl groups; the expected correlation from the protons of 3′-N-CH₃ to C-2 was observed but none were present from the protons of 1′-N-CH₃ to C-2. Finally, the 5,5-trans geometry was suspect as it was based on a very weak ¹H-¹H ROESY correlation. At this point it is important to summarize several important insights of Figure 5, based on new NMR data in CDCl₃. making it possible to favor structure 1b. Only one N-CH₃ group exhibited coupling to C-2, so the inequality of the two N-methyl groups persisted. An important new strong correlation (Figure S5) from the same hetero proton (either an NH or an OH) to C-2, C-4, and C-5 unambiguously meant that structure 1a could be ruled out. This new correlation also made Watson’s alternate structures 2 and 3 unlikely because it would require 4-bond couplings. In summary, only 1b was in harmony with the hetero attached proton HMBC data.

Figure 4.

Original structure (1a) with significant HMBC correlations (CD₃OD).¹

Figure 5.

Correlation data (CDCl₃) from spiroleucettadine on original structure 1a, final structure of spiroleucettadine 1b, and alternate structures 2⁴ and 3.⁴

During the course of this re-investigation it became clear that structure elucidation can be backbreaking or inaccurate when the ratio of H/C is less than one for a core substructure or an entire molecule. Under such a circumstance the process of assembling partial or total structures becomes tenuous as the NMR data sets will be less useful and other approaches must be used. Faced with this dilemma, we turned to DFT calculations to evaluate original structure 1a alongside Watson’s two alternatives (2 and 3), and an additional thirteen possibilities (Figure S7). We found it useful to evaluate the degree of fit using two parameters designated as the “mean absolute error (MAE)” and “% Score.” The MAE is calculated in a standard fashion¹⁵ while the % Score is defined by the following equation: % Score = [Σ of carbons - Σ error points / Σ of carbons]100. Based on extensive DFT evaluations of model compounds, we define acceptable agreement as follows: MAE < 2.2 and % Score > 85. The results for the top three new alternate structures, 1b, 4, and 5, plus that of 1a, 2 and 3 are summarized in Figure 6. The very poor match of the rating parameters for 1a [% Score = 65 and MAE = 4.3] and 2 [% Score = 75 and MAE = 3.6] are in accord with the observations of Watson.⁴ The calculation results for 4 are not within our defined acceptable range. While structure 3 provides an acceptable fit, it can be ruled out based on the discussion presented above. The two remaining structures 1b, and 5 provide good matches, but only the former is in harmony with the HMBC data discussed above. Thus, the final list could be pruned to a single candidate structure, 1b.

Figure 6.

DFT calculation results for spiroleucettadine (1b) and other candidate structures.

The revised structure of spiroleucettadine as 1b is now secure. The new structural core, comprised of a methylamino-furoimidazolone array, contains an enticing functional group moiety for total synthesis. It is reminiscent of the hydroxy-furoimidazolone array present in the Agelas-derived slagenins.¹⁶ The overall pathway taken to amend the original structure of 1a to 1b provides some important lessons learned. Caution should be exercised during the structure elucidation or dereplication phase of metabolites having three or more nitrogens isolated from a calcareous sponge. Its compounds may not necessarily possess an amino imidazole residue. More importantly, the outcome of this study underscores the value of total synthesis as a powerful tool to interrogate the possibility that a densely heteroatom-functionalized structure, assigned based solely on NMR and mass spec data, could be equivocal. The value of conducting an X-ray analysis is obvious but such results are not completely error free. Finally, as shown by the analysis of the structures in Figure 6, a powerful approach involves overlaying the DFT results with those obtained by HMBC experiment.

Experimental Section

Biological Material, Collection, and Identification

Specimens of Leucetta (coll. no. 00111) (6.2 kg, wet weight) were collected using scuba in 2000 from Caesar’s Rock, Fiji. Taxonomic identification was based on comparison of the biological features to other samples in our repository. Voucher specimens and underwater photos are available.

Extraction and Re-Isolation of Spiroleucettadine (1b)

The sponge (170 g, dry weight) was extracted with MeOH followed by partitioning of the concentrated extract between dichloromethane and water. The crude organic phase (692 mg) was purified via reversed-phase HPLC to afford 3.6 mg of 1b (3.6 mg).

Spiroleucettadine (1b)

¹H and ¹³C NMR properties (see supporting information Figures S2-S5 and Table S1) were in accord with the published value.¹ The NMR data recorded in CDCl₃ are reported in Table S1. [α]²³_D= -5.1 (c = 0.56 in MeOH). HRESIMS m/z [M + H]⁺ 370.1758 (calculated for C₂₀H₂₄N₃O₄ 370.1761. Crystals were too small to allow for a melting point to be determined.

Density Functional Theory Calculations

Carbon chemical shifts predictions were performed by ChemNMR (version 8.0), ACD (version 8.0), Spartan 06 and Gaussian 03. The PC computer used for Spartan 06 possessed an Athlon 64 × 2 5600+ (2.80 GHz, 1MB × 2) processor and 4 GB DDR2 SDRAM at 667 MHz with 222GB disk space. Gaussian 03 was run on a Unix machine with 8 GB RAM with 72 GB disk space. The Spartan calculations were performed by the basis set, B3LYP/6-31G*//B3LYP/6-31G*. Other basis sets described in Table 1 were performed by Gaussian 03. Tables S2 – S17 provide detailed results for calculated ¹³C vs experimental shifts. The calculation results for 16 spiroleucettadine candidate structures (Figure S7 and Tables S2-S17) and the seven diastereomers of 4 (Figure S9) are in the Supporting Information.

Crystallization method

Spiroleucettadine (0.9 mg) was dissolved in 300 μL of CDCl₃ for a vapor diffusion experiment employing hexanes as the co-solvent. After two weeks at -7 °C several colorless plate-like crystals of spiroleucettadine formed that were suitable for synchrotron crystallographic analysis.