Stereoselective Synthesis of the I–L Fragment of the Pacific Ciguatoxins

The I–L ring system found in all the Pacific ciguatoxins has been prepared from a tricyclic precursor in a highly stereoselective manner. Subtle differences in the reactivity of the enones present in the seven- and eight-membered rings of the tricyclic ether starting material have been exploited to allow selective protection of the enone in the eight-membered ring. Subsequent distereoselective allylation of the seven-membered ring has been accomplished by a palladium-mediated Tsuji-Trost reaction. The K-ring methyl and hydroxyl groups have been installed in a highly stereoselective manner by sequential conjugate reduction and enolate oxidation reactions. Ring L has been constructed by a use of a novel relay ring-closing metathesis reaction to complete the tetracyclic framework, which possesses the functionality necessary for elaboration of rings I and L and the introduction of ring M.


Structures and Bioactivities of the Ciguatoxins
The ciguatoxins are a family of large and structurally complex fused polycyclic ether natural products of marine origin. Although the ciguatoxins had been implicated as the causative agents of ciguatera poisoning in humans since the 1960s, it was not until the late 1980s that the first structures of members of the ciguatoxin family-P-CTX-1 and P-CTX-4B-were fully elucidated by Yasumoto et al. (Figure 1) [1,2]. Subsequently, more than 30 other ciguatoxins have been isolated from marine dinoflagellates or from fish and marine organisms that accumulate the toxins by predation. The ciguatoxins can be grouped as Pacific, Caribbean or Indian according the ocean or sea where samples containing each toxin were first collected [3][4][5][6]. Pacific ciguatoxins are the most numerous and well-characterised of the ciguatoxins and this group accounts for more than three quarters of the ciguatoxins that are known [4][5][6]. Four Caribbean ciguatoxins have been characterized-C-CTX-1 and C-CTX-2 are equilibrating lactols and C-CTX-3 and C-CTX-4 are diastereomers that arise by reduction of C-CTX-1/C-CTX-2 [7,8]. A further eight Caribbean ciguatoxins have been detected by mass spectrometry but have yet to be characterised and the structures of the six ciguatoxins-I-CTX-1-6-isolated from samples collected in the Indian Ocean have yet to be elucidated [6,9,10]. The Pacific ciguatoxins can be divided into two distinct structural sub-groups. The first is exemplified by P-CTX-1 and P-CTX-4B, and the second by P-CTX-3C ( Figure 1) [3,5]. Natural products in the first group have a four-carbon chain attached to the A-ring (C-5) and a seven-membered E-ring. This group comprises more than a dozen congeners that have a variety of structural modifications such as a hydroxyl substituent at C-54, inverted configuration at C-52, a seco M-ring or oxidation at the C-4 and/or C-7 positions. The second and larger group is based on a P-CTX-3C skeleton. Compounds in this group have an expanded (eight-membered) E-ring and lack the A-ring side chain found in P-CTX-1 and related ciguatoxins. Variations on the P-CTX-3C structure include A-ring and M-ring seco compounds, a C-49-epi compound (P-CTX-3B) and oxidised analogues in which there is a hydroxyl group at the C-51 position, and/or the A-ring alkene is replaced by a C-2 carbonyl group, a C-2 hydroxyl group or a 1,2-diol.
The ciguatoxins are extremely potent neurotoxins and some of them have been shown to have LD 50 values of 4 µg/kg or lower in mice. Ingestion of fish or seafood that is contaminated by ciguatoxins causes a type of food poisoning known as ciguatera that can result in the victim experiencing a myriad of unpleasant and sometimes severe neurological (both peripheral and central nervous system), cardiovascular and gastrointestinal symptoms [11]. It has been estimated that tens or even hundreds of thousands of people suffer from ciguatera poisoning each year. Symptoms can persist for many months or even years and, in extreme cases, ciguatera food poisoning can be fatal. The pharmacological effects of the ciguatoxins result primarily from their ability to cause persistent activation of voltage-gated sodium channels in nerve membranes and thereby prevent the nerve repolarisation that is essential for normal nerve function [12,13]. There is also evidence that the ciguatoxins can block voltage-gated potassium channels [14].

Total Syntheses of Pacific Ciguatoxins
The ciguatoxins are daunting targets for total synthesis because of their size, their stereochemical complexity, the large number of rings they possess and the high proportion of synthetically challenging medium-sized cyclic ethers in their trans-fused polycyclic structures. The synthetic challenges presented by the ciguatoxins combined with their potent neurotoxic activities makes them highly alluring synthetic targets and many research groups have performed methodological studies in which substantial fragments or smaller sub-units of various ciguatoxins have been constructed. However, despite a continuous stream of highly innovative synthetic work spanning more than two decades, the ciguatoxins have proved to be very difficult to synthesise. The only successes in this area have been achieved by Hirama et al., who reported the first total synthesis of a ciguatoxin-P-CTX-3C-in 2001 followed by the total syntheses of 51-hydroxy-P-CTX-3C and P-CTX-1 in 2006 [15][16][17], and by Isobe et al., who completed a total synthesis of P-CTX-1 in 2009 [18]. These elegant and impressive syntheses are the only complete total syntheses of members of the ciguatoxin family of natural products to have been reported.

Proposed Synthetic Strategy and Previous Synthetic Work Concerning the Total Synthesis of P-CTX-3C and Related Ciguatoxins
We have been engaged in the development of new reactions, tactics and strategies for the rapid and efficient synthesis of fused polycyclic ether arrays for many years [19]. As part of this research program, we have invented and explored a variety of iterative and bidirectional strategies to assemble large fragments of fused polycyclic ether natural products of marine origin such as the brevetoxins, gambieric acids and ciguatoxins [20][21][22][23]. In our previous studies concerning the synthesis of P-CTX-3C, the A-E fragment of the natural product was assembled by a route in which ring-closing metathesis (RCM) reactions were used to construct the rings in an iterative manner [18]. We have also reported a novel bidirectional RCM approach to the synthesis of the tricyclic I-K fragment of P-CTX3-C from a simple monocyclic precursor and now present the results of studies concerning its elaboration to give an I-L fragment that can serve as an advanced intermediate for the preparation of the terminal fragment of any of the Pacific ciguatoxins [24].
Retrosynthesis of P-CTX-3C commences with disconnection through rings G and H to give two fragments-the hexacyclic A-F fragment i and the pentacyclic I-M fragment ii-that are of similar size and structural complexity ( Figure 2). The focus of the synthetic work described herein was the I-M fragment ii. Disconnection of this pentacyclic system by removal of the methyl substituents in rings I and scission of the M-ring spiroacetal leads to the tetracyclic fragment iii. This intermediate contains an α,β-unsaturated ketone in ring I and an α,β-unsaturated lactone in ring L to enable introduction of the methyl substituents by conjugate addition in the case of ring I and sequential conjugate addition and enolate alkylation in the case of ring L. In previous work concerning the synthesis of the I-M fragment of P-CTX-3C and related ciguatoxins, we have demonstrated that the tricyclic sub-unit 1 can be prepared from a simple tetrahydropyranyl precursor by double ring-closing metathesis (RCM) and that subsequent bidirectional functionalisation of the diketone 1 is possible by sequential double allyl enol carbonate formation and double palladium-mediated Tsuji-Trost allylation (Scheme 1) [24,25]. Although it was possible to direct the double Tsuji-Trost reaction of the substrate 2 to give just two of the four possible diastereomeric products by the choice of a suitable chiral ligand, we were unable to identify a catalyst that would deliver the required diketone 3 as the sole or even predominant isomer. Thus, it was necessary to revise the synthetic strategy in order to allow rings I and K to be differentiated and functionalised independently thereafter. The following discussion provides an account of construction of the I-L sub-unit by such an approach.

Results and Discussion
It was not possible to perform simultaneous and diastereoselective bidirectional functionalisation of the bis-enone 1 by double Tsuji-Trost allylation [24,25] and so it was necessary to elaborate rings I and K independently in order to extend the fused polyether framework (Scheme 2). The tricyclic bis-enone 1 is quasi-symmetric and so independent functionalisation of rings I and K by discrimination between reactive functional groups appears to be a challenging task at first glance. However, spectroscopic analysis of the bis-enone 1 reveals subtle structural differences between the enone functionality in rings I and K. In the IR spectrum of 1, the carbonyl group in ring I has a stretching frequency of 1669 cm −1 whereas the carbonyl group in ring K has a stretching frequency of 1653 cm −1 -the identities of the carbonyl peaks in the IR spectrum of 1 were established unambiguously by comparison to data for the hydroxy ketones 4 and 5, produced by partial reduction of the diketone 1 ( Figure 3). The IR data suggest that the degree of enone conjugation in the eight-membered I-ring is significantly lower than that in the seven-membered K-ring. This finding was further supported by 13 C NMR spectroscopy where the β-carbon of the enone in ring I has a chemical shift (137 ppm) that is more typical of an unconjugated alkene at this position in an eight-membered cyclic ether, as can be seen by comparing the chemical shifts for the alkene carbons in the simple fused bicyclic ether 6 [26], whereas the corresponding carbon in ring K has a chemical shift (154 ppm) for the enone β-carbon that is consistent with a fully conjugated enone. The significant differences in the degree of conjugation in the enone groups suggested that it might be possible to functionalise them selectively by exploiting differences in their chemical reactivity. Indeed, there are pertinent examples of highly selective acid-mediated acetal formation in polycyclic substrates that contain both an enone and unconjugated ketone and so this approach was used to protect the I-ring carbonyl group selectively [27].  Completely selective protection of the I-ring carbonyl group as a dioxolane was accomplished by reaction of the bis-enone 1 with bis(trimethylsilyl)ethylene glycol in the presence of trimethylsilyl trifluoromethanesulfonate (Scheme 2) [28]. The enone 7 was then converted into the allyl enol carbonate 8 by deprotonation with sodium bis(trimethylsilyl)amide and enolate trapping with allyl chloroformate. Tsuji-Trost allylation was then performed thereafter by treatment of the allyl enol carbonate 8 with the complex generated from tetrakis(triphenylphosphine) palladium(0) and the (R)-t-butyl-PHOX ligand (9) in situ [24,25,29]. The reaction was highly stereoselective (>95:5 dr; minor isomer not detectable by 1 H NMR analysis) and delivered the enone 10 in 95% yield (Scheme 2). The stereochemical outcome of this and subsequent diastereoselective reactions was confirmed by 1 H NMR NOE analysis (for details, see Supplementary Materials).
Ring K was elaborated further by conjugate reduction of the enone (Scheme 3). Reduction of the enone 10 with Stryker's reagent [30] resulted in a mixture of the diastereomeric ketones 11a and 11b (3:7 ratio) in which the required diastereomer 11a was the minor component. This reaction also suffered from the problem that it did not proceed to completion, even when an excess of the reducing agent was used. A conjugate reduction protocol developed by Yamamoto et al. was deployed with the expectation that it would solve these reactivity and selectivity problems [31]. Coordination of the ketone to aluminium tris(2,6-diphenylphenoxide) followed by treatment with the reductant generated by the reaction of diisobutylaluminium hydride with n-butyllithium afforded the ketone 11b in 39% yield and the diketone (13% yield) resulting from cleavage of the acetal protecting group of 11b; none of the required diastereomer 11a was obtained from the reaction. Stereocontrolled conjugate reduction of the enone 10 to give the required ketone was finally performed in good yield by reaction with the 'hot' Stryker's reagent generated from 1,2-bis(diphenylphosphino)benzene, copper(II) acetate hydrate and 1,1,3,3-tetramethyldisiloxane as described by Lipshutz and co-workers [32]. Reduction of the enone 10 with this reagent at 0 • C in toluene afforded a diastereomeric mixture of the ketones 11a and 11b (combined yield of 66%; 9:1 ratio) in which the required isomer was the major product. The diastereomeric ketones 11a and 11b could not be separated by standard silica gel chromatography. Stereoselective conjugate reduction of the enone 10 enabled the requisite K-ring hydroxyl group to be installed by enolate oxidation (Scheme 4). Regioselective deprotonation of a mixture of ketones 11a and 11b (4:1 ratio) with potassium bis(trimethylsilyl)amide and reaction of the resulting enolate with the N-sulfonyl oxaziridine 12 [33,34] produced the α-hydroxy ketone 13 in 43% yield (54% when adjusted to account for the isomer ratio of starting material 11) and with a very high level of diastereocontrol. The same transformation was performed by generation of the triethylsilyl enol ether and subsequent Rubottom oxidation with 3-chloroperoxybenzoic acid [35], but poor yields were obtained and so this sequence was not a viable alternative to direct enolate oxidation. The α-hydroxy ketone 13 was then subjected to stereoselective reduction with diisobutylaluminium hydride to give the diol 14 [36]. The stereochemical outcome of the reaction can be explained by a reaction mechanism in which complexation of the aluminium reagent to the hydroxyl and carbonyl groups produces a cyclic chelate and addition of diisobutylaluminium hydride occurs from the top face of the molecule (as drawn). At this stage, it was necessary to differentiate the hydroxyl groups of the syn diol 14 to enable construction of ring L. Selective protection of the hydroxyl group adjacent to the K-ring methyl substituent was accomplished by reaction of the diol 14 with dibutyltin oxide to give a cyclic dialkylstannylene acetal followed by regioselective fluoride-mediated alkylation with 2-bromomethylnaphthalene [37,38]. The Nap-protected alcohol 15 was obtained in reasonable yield and a small amount of unreacted diol 14 was recovered from the reaction. An alternative approach to functionalisation of ring K that did not involve sequential conjugate reduction and enolate oxidation was explored (Scheme 5). The tricyclic enone 10 was first subjected to Luche reduction which proceeded to give the allylic alcohol 16 in excellent yield and with a high level of diastereocontrol. Directed epoxidation of the allylic alcohol with t-butyl hydroperoxide mediated by vanadyl acetylacetonate produced the epoxide 17 in a highly stereoselective manner. Protection of the hydroxyl group as a triisopropylsilyl ether afforded the epoxide 18 in a yield of 66% from the enone 10. It was now possible to rearrange the epoxide to give an allylic alcohol by reaction of the epoxide 18 with lithium 2,2,6,6-tetramethylpiperidide in the presence of the Lewis acid diethylaluminium chloride [39]. The allylic alcohol 19, in which the acetal on ring I had been cleaved under the reaction conditions to reveal the enone, was obtained. Rearrangement of the epoxide 18 by treatment with p-toluenesulfonic acid was also successful; in this case ethylene glycol was added to the reaction mixture to ensure that the acetal protecting group was retained in product 20. In both cases, the yield of the rearranged product was modest. Unfortunately, subsequent chemoselective and stereoselective hydrogenation of the exocyclic alkene to produce the K-ring methyl substituent proved to be difficult to accomplish. It was expected that selective directed hydrogenation of the allylic alcohol 20 would be possible by use of Crabtree's catalyst [40]. However, the hydrogenation reaction did not take place when a catalyst loading of 1 mol % was used and decomposition of the substrate occurred when higher catalysts loadings (≥10 mol %) were employed. When Wilkinson's catalyst was used to hydrogenate the alkoxide generated by deprotonation of the allylic alcohol 20 with sodium hydride in the presence of 15-crown-5 [41,42], the terminal alkene in the side chain was reduced instead of the exocyclic 1,1-disubstituted alkene in ring K. Construction of ring L to complete the tetracyclic ciguatoxin I-L core framework was accomplished as shown in Scheme 6. The allyl side chain in the alcohol 15 was subjected to isomerisation by treatment with the ruthenium complex 21 in methanol at reflux to give the alcohol 22 as a mixture of alkene isomers (E/Z) along with some side-products [43]. At this juncture, construction of ring L as a pentenolide by conversion of the alcohol 22 into a simple acrylate ester followed by direct RCM was considered. However, there were concerns regarding the rate of initiation of the metathesis reaction because the substrate would contain an acrylate and an internal alkene branched at the α position; consequently, a relay RCM reaction was employed as a precaution [44]. Subjection of the alcohol 22 to Steglich esterification with the known carboxylic acid 23 [45] delivered the relay metathesis precursor 24 and subsequent treatment with the ruthenium complex 21 resulted in relay RCM to give the lactone 25 corresponding to the tetracyclic framework found in the I-L ring system of the Pacific ciguatoxins. The modest yield for the three-step sequence implies an average of 50% per step; it can be accounted for by the scale on which the reactions were performed and by-product formation during isomerisation (15 → 22). The unsaturation in ring L would enable subsequent introduction of the requisite methyl substituents by conjugate addition with a suitable methyl nucleophile and trapping of the resulting enolate with a methyl electrophile [46]. Scheme 6. Use of relay RCM to construct ring L and form the I-L ring system of the Pacific ciguatoxins.

Conclusions
The synthesis of the I-L ring system of the Pacific ciguatoxins has been accomplished in 10 steps starting from the tricyclic bis-enone 1. Elaboration of ring K by stereoselective Tsuji-Trost allylation was accomplished after selective protection of the I-ring enone by acetal formation. The configuration at the methyl-bearing stereogenic centre in ring K was controlled during conjugate reduction of the enone and the hydroxyl substituent was installed in a highly stereoselective manner by enolate oxidation according to the Davis procedure. Stereoselective ketone reduction followed by selective protection of the resulting 1,2-diol then allowed the metathesis trigger to be attached to the K-ring by esterification of the hydroxyl group. Relay RCM, mediated by the Grubbs second generation ruthenium complex, resulted in formation of ring L and completed the tetracyclic framework with the functionality necessary for subsequent elaboration of ring I, installation of the methyl substituents in ring L and construction of ring M.

Materials and Methods
Air-and moisture-sensitive reactions were performed under an atmosphere of argon in flame-dried apparatus. Tetrahydrofuran (THF), toluene, dichloromethane and diethyl ether were purified using a Pure-SolvTM 500 Solvent Purification System. Other dry solvents and starting materials were obtained from commercial sources and used as received unless stated otherwise. Petroleum ether (pet. ether) used for column chromatography was the 40-60 • C fraction. Triethylamine and 2,2,6,6-tetramethylpiperidine were distilled and stored under argon prior to use. n-Butyllithium solutions were titrated against diphenylacetic acid to obtain accurate molarity. 4 Å molecular sieves were oven dried prior to use.
Reactions were monitored by thin layer chromatography (TLC) using Merck silica gel 60 covered aluminium-backed plates F254. TLC plates were visualised under UV light and stained using potassium permanganate solution or acidic ethanolic anisaldehyde solution. Flash column chromatography was performed with silica gel (Geduran Si 60 35-70 µm) as solid support.
IR spectra were recorded using a Shimadzu FT IR-8400S ATR instrument (Shimadzu UK, Milton Keynes, UK). The IR spectrum of each compound (solid or liquid) was acquired directly on a thin layer at ambient temperature.