Total Synthesis of Rameswaralide Utilizing a Pharmacophore-Directed Retrosynthetic Strategy

Abstract

A pharmacophore-directed retrosynthetic (PDR) strategy was applied to the first total synthesis of the cembranoid, rameswaralide, in order to simultaneously achieve a total synthesis while also developing a structure-activity relationship profile throughout the synthetic effort. The synthesis utilized a Diels-Alder-Lactonization, including a rare kinetic resolution to demonstrate the potential of this strategy for an enantioselective synthesis providing both 5,5,6- and, through a ring expansion, the 5,5,7-tricyclic ring systems present in several Sinularia soft coral cembranoids. A pivotal synthetic intermediate, a tricyclic epoxy α-bromo cycloheptenone, displayed high cytotoxicity with interesting selectivity toward the HCT-116 colon cancer cell line. This intermediate enabled the pursuit of three unique D-ring annulation strategies including a photocatalyzed intramolecular Giese-type radical cyclization and a diastereoselective, intramolecular enamine-mediated Michael addition with the latter annulation constructing the final D-ring to deliver rameswaralide. The discovery of a serendipitous oxidation state transposition of the tricyclic epoxy cycloheptenone proceeding through a presumed doubly vinylogous, E1-type elimination, enabled facile introduction of the required α-methylene butyrolactone. Preliminary biological tests of rameswaralide and precursors demonstrated weak cytotoxicity however the comparable cytotoxicity of a simple 6,7-bicyclic β-keto ester, corresponding to the CD-ring system of rameswaralide, to the natural product itself suggests that such bicyclic β-ketoesters may constitute an interesting pharmacophore that warrants further exploration.

Graphical Abstract

Introduction

First isolated from the soft coral Sinularia dissecta in 1998,¹ rameswaralide (1) is a tetracyclic diterpene belonging to the broader class of marine cembranoid secondary metabolites.² Its structure, including both relative and absolute stereochemistry, was later verified by the same group in 2016 by X-ray crystallography,³ revealing a complex, highly oxygenated, caged 5,5,7,6 all cis-fused ring system containing seven stereogenic centers. Rameswaralide was reported to display moderate cytotoxicity³ and possible anti-inflammatory activity⁴ piquing the interest of various synthetic groups.⁵ Despite this broad interest,^{2, 5} a total synthesis of rameswaralide has not been reported nor a study of structure-activity relationships for this family. While several synthetic studies have been described toward cembranoid family members,^{2, 5-6} only two members possessing related tetracyclic frameworks have been reported, namely scabrolide A (5) by both the Stoltz^6d and Fürstner⁶ⁱ groups with the latter report also including the synthesis of nominal scabrolide B.⁷ The structure of scabrolide B was recently reassigned based on crystallography⁷ as a 5,5,7,6-core with similar connectivity to ineleganolide (2).

Our interest in rameswaralide (1) stems from its complex architecture coupled with an interest in applying pharmacophore-directed retrosynthesis (PDR) to this and related cembranoids to gain a greater understanding of their structure-activity relationships (SAR). We chose the common, densely functionalized 5,5,7 core ring system (Figure 1, blue) as the minimal structure required for bioactivity since family members which maintained this core generally displayed greater bioactivity (ineleganolide: 11.6 μM ED₅₀ P388 cells, ⁸ sinugyrosanolide: 11.8 μM ED₅₀ P388 cells⁹) compared to 5,5,6-tricyclic congeners (Figure 1, red) which to the best of our knowledge have weak to no activity.^{3-4, 8-10} Thus, we hypothesized that the 5,5,7-tricyclic core present in rameswaralide (1), ineleganolide (2), and sinugyrosanolide A (3) is critical for their bioactivity, while variations in the D-ring may impact bioactivity to a lesser extent. These considerations guided application of PDR to rameswaralide leading to the synthesis of several targeted derivatives en route to the first total synthesis of this cembranoid.

Figure 1.

Sinluaria cembranoid family members informing our proposed pharmacophore for rameswaralide.

To enable collection of SAR information en route to rameswaralide (1), we applied our recently described pharmacophore-directed retrosynthetic strategy (PDR).¹¹ Briefly, PDR begins by developing a working hypothesis for the pharmacophore of a natural product, or the minimal sub-structure required for bioactivity, premised on the bioactivity of congeners and/or chemical intuition. In the case of rameswaralide this sub-structure i is selected as a waypoint en route to the natural product along with other simplified derivatives e.g. ii and iii (i.e. synthetic intermediates) which importantly also contain the hypothetical pharmacophore along with varied functionality FG¹ and FG³ (Scheme 1a). Ultimately, this strategy delivers the entire natural product framework iv with potential for diverse functionality (FG² and FG³) for biological testing. In the case of rameswaralide, application of PDR led to the 5,5,6-core 11, based on a Diels-Alder lactonization strategy¹² of dienol 13 and acryloyl chloride, and the 5,5,7-core 11, the hypothesized pharmacophore, through a ring expansion strategy as initial waypoints in the synthetic plan (Scheme 1b). The overall retrosynthetic strategy and key disconnections (‘itinerary’) when applying PDR are dictated by the targeted simplified intermediates (‘waypoints’). These gross disconnections are set at the beginning given that a particular disconnection (e.g. D-ring annulation) must indeed be accomplished in order to investigate the desired structure-activity relationships. Therefore, we envisioned implementing various D-ring annulation strategies to deliver the CD ring system of rameswaralide with varied functionality at C1 (e.g. 7, 8). Despite well-known limitations, model systems can be useful tools to investigate key steps in a total synthesis on simpler substrates. When applying PDR, model studies gain additional utility by enabling access to additional simplified derivatives that can be bioassayed. We planned various D-ring annulation model studies with cycloheptenone, leading to the bicyclic target 9 and derivatives, corresponding to the CD ring system of rameswaralide. These proposed model studies took on greater importance given the challenges described in the literature for related transformations. Specifically, Mehta^5a and Trost^5f who independently synthesized a 5,5,7-tricyclic core in their efforts toward rameswaralide, were either unable or have not yet reported successful D-ring annulations onto a 5,5,7-core.

Scheme 1.

Pharmacophore-Directed Retrosynthesis (PDR): (a) application to rameswaralide showing proposed ‘pharmacophore’ i and gradual, increasing structural complexity leading to general targets ii-iv that will inform SAR (b) generalized retrosynthetic strategy highlighting required disconnections for targeted derivatives (e.g. ii, iii) serving as waypoints in a projected total synthesis including a model study for D-ring annulation to also access the CD ring system.

Previously, we described a racemic synthesis of the tricyclic 5,5,7-core of rameswaralide and an account of the challenges overcome to provide a viable ring expansion process.^5h These preliminary studies led to the synthesis of epoxy α-bromo enone 20 (cf. Scheme 2) which became a versatile intermediate for final D-ring annulation strategies described below (Scheme 2). Furthermore, enone 20 demonstrates the power of PDR, since while it was not specifically an initially envisioned target, implementation of this synthetic strategy leads to the synthesis of intermediates related to the natural product. In this instance, PDR led to the identification of the simplified 5,5,7-tricyclic enone 20 which surprisingly exhibited sub-micromolar activity with ten-fold selectivity toward the colon cancer cell line HCT116 (0.39 μM) compared to two other cancer cell lines studied (MDA-MB-231 and A549, 3.3 and 10.5 μM, respectively). The high potency of this epoxy α-bromo enone is likely due to the presence of a Michael acceptor and/or epoxide moiety.

Scheme 2.

Improved synthesis of the key, versatile epoxy α-bromo enone 20. Demonstration of a kinetic resolution leading to optically active enol ether (−)-15 and the transposed α-bromo enol ether (−)-16 (absolute stereochemistry determined through anomalous dispersion by X-ray analysis, inset).

Results and Discussion

Since our initial disclosure,^5h we optimized several early steps toward epoxy α-bromo enone (±)-20 during scale-up towards rameswaralide (Scheme 2). The addition of MeLi, which had to be prepared due to supply chain issues (see SI for details), and subsequent direct deprotection, performed without purification, led to greater overall yield and thus material throughput following a single recrystallization. Mother liquors from several runs were also combined and purified for greater material throughput. Furthermore, we demonstrated the first example of a kinetic resolution employing the Diels-Alder lactonization with racemic diene diol (±)-13 and acryloyl chloride albeit with higher loadings of (−)-BTM (25 mol%) for improved yields and enantioselectivities (46% yield, 97:3 er, >19:1 dr). The subsequent enol ether transposition mediated by NBS gave a single diastereomer of the crystalline bromide (−)-16 whose relative and absolute stereochemistry was verified by X-ray analysis (anomalous dispersion) which matched the absolute configuration of rameswaralide. However, in efforts to maximize material throughput, we utilized racemic enol ether (±)-15 obtained in 77% yield through the Diels-Alder lactonization using racemic BTM which, following the same alkene transposition utilizing NBS gave bromide (±)-16. For biological assays, we anticipated that racemic material would be assayed initially and then chiral HPLC separation could subsequently provide each enantiomer when warranted by biological studies of the racemate. The subsequent steps proceeded as previously described with one exception. Namely, an improvement in the ring expansion/elimination sequence was achieved by avoiding a problematic purification of the tribromide intermediate (±)-19. Optimal yields were obtained by taking the crude tribromide directly into the elimination leading to an improved 52% yield (vs 21%) of epoxy enone (±)-20 over 2 steps.

Initial D-Ring annulation strategy: An intramolecular Diels-Alder strategy.

Our initial strategy for D-ring annulation, involved the use of an intramolecular Diels-Alder (IMDA) cycloaddition with a tethered diene to overcome the low reactivity of the cycloheptenone previously observed by Mehta^5a (Scheme 3). In addition, α-bromo dienophiles, such as that present in epoxy enone 20, are known to increase the reactivity of dienophiles in Diels-Alder cycloadditions.¹³ The epoxide, if sufficiently stable, could serve as a protecting group for the labile tertiary alcohol. In addition, based on molecular models and computational studies, the epoxide leads to a more planar enone sub-structure increasing conjugation, which may increase the reactivity of the cycloheptenone toward Diels-Alder reactions. We presumed late stage SmI₂ reduction would result in simultaneous α-cleavage of the epoxide and reduction of the α-bromo substituent. β-Elimination of the neopentyl ether, dehydration of the resulting tertiary alcohol, and α-acylation of the cyclohexanone with Mander’s reagent¹⁴ would deliver rameswaralide (1).

Scheme 3.

D-Ring annulation strategy #1: (a) Retrosynthetic strategy for the intramolecular Diels-Alder (IMDA) cycloaddition. (b) Serendipitous C-ring oxidation state transposition discovered during attempted synthesis of the IMDA substrate.

Toward studying the proposed IMDA, we targeted neopentyl ether 22 by γ-oxidation of α-bromo enone 20 followed by etherification of the resulting alcohol with diene acid 23 by the method of Baran.¹⁵ Initial attempts to effect allylic oxidation of enone 20 with SeO₂ at C13 were unsuccessful likely due to the electron deficient nature of the enone. However, this reaction began to reveal the unexpected but welcome stability of α-bromo enone 20 (no degradation after heating 100 °C for up to 3 d) suggesting it would be an ideal substrate for various D-ring annulation strategies. We next studied a vinylogous Rubottom oxidation¹⁶ which has been used previously for γ-selective hydroxylation of enones.¹⁷ When α-bromo enone 20 was treated with TESOTf and Et₃N, rather than formation of the expected enol ether required for γ-oxidation, we noted two new products by TLC, a mono-silylated and a bis-silylated intermediate by MS analysis. Analysis of the crude ¹H NMR revealed a mixture of silylated and non-silylated tertiary alcohols 25 and 28, respectively. In this critical transformation, the epoxide was cleaved, effectively transposing the epoxide oxidation to the unsaturated butyrolactone found in rameswaralide. We quickly found that by adjusting the equivalents of TESOTf (< 2.5 equiv), the mono-TES enol ether 25 could be obtained as the exclusive product. Subsequent treatment with TBAF/AcOH gave tertiary alcohol 26 mixed with the resulting cross-conjugated dienone 27 derived from β-elimination of the tertiary alcohol (~9:1, ¹H NMR of crude mixture). Due to the inherent instability of tertiary alcohol 26 on silica gel (untreated and base treated) these products could not be separated. One possible mechanism for this transformation invokes the silyl ketene acetal intermediate 24 which leads to epoxide ring opening, through a presumed doubly vinylogous elimination followed by silylation of the tertiary alcohol if excess TESOTf is present. This serendipitous finding demonstrated that the epoxide, through an oxidation state transposition, is an efficient strategy to introduce the α-methylene butyrolactone present in rameswaralide. While we were unable to introduce the desired allylic alcohol to study the IMDA, these studies provided a way to introduce the α-methylene butyrolactone and demonstrated the versatility of the α-bromo enone 20.

Second D-Ring annulation strategy: A photocatalyzed, Giese-type 6-endo-trig cyclization.

We next considered an intramolecular Giese-type¹⁸ cyclization initiated by catalytic photoredox conditions building on the collaborative work of the Overman and MacMillan groups.¹⁹ The required allylic radical for this proposed 6-endo-trig cyclization would be generated from alcohol 29 which in turn would be derived from a Suzuki-Miyura cross coupling²⁰ between α-bromo enone 20 and pinacol boranate 31 (Scheme 4). We also planned to study the serendipitous oxidation state transposition on the more complex epoxide 30 as we believed this would avoid the inherent stability issues previously observed with dieneone 26 (cf. Scheme 3).

Scheme 4.

D-Ring annulation strategy #2: Intramolecular Giese-type, 6-endo trig cyclization of an allylic radical generated from alcohol 29.

The synthesis of TBS-protected pinacol borane 31 began from known iodide 33 through iodination of commercially available acetonide 32.²¹ Attempted direct aldol reaction of the dienolate derived from deprotonation of dioxinone 33 with methacrolein led to only 1,4-addition. However, a Mukaiyama aldol²² delivered the desired allylic alcohol 34 in 37% yield without isolation of the intermediate silyl ketene acetal. Silylation followed by a lithium-halogen exchange/borylation sequence gave the targeted pinacol borinate 31. We also prepared the TES-protected pinacol borinate 37 since subsequent studies revealed that a more labile silyl group was required (vide infra).

We chose to study the proposed Suzuki-Miyaura cross coupling in a model system as a means to not only preserve advanced intermediates, but also importantly provide access to the CD ring system of rameswaralide for further SAR information. Coupling of known bromocycloheptenone 38²³ and racemic pinacol borinate 31, after some optimization, delivered the dienone 39 in 60% yield. The stereocenter in borinate 31 is inconsequential since it becomes the radical center. Initial desilylation attempts with TBAF and HF led to elimination to the fully conjugated tetraene (not shown), however, prolonged stirring at ambient temperature (23 °C) with AcOH-buffered TBAF²⁴ gave clean conversion to allylic alcohol 40. We next turned our attention to formation of the cesium oxalate required for oxidation by the iridium photocatalyst.^19c Overman and MacMillan commonly employed methyl oxalates as intermediates for Cs-oxalate salt generation enabled by selective hydrolysis of the more accessible ester.^19c While methyl, benzyl, and p-nitrobenzyl oxalates²⁵ were prepared from alcohol 40, none of the oxalate esters lead to sufficiently selective hydrolysis returning quantities of starting alcohol 40 (not shown, see SI for details, pp. 16-18). While this work was in progress, Overman reported the use of TMS ethyl oxalates which, upon treatment with CsF, extrude ethylene gas providing the desired cesium oxalate.²⁶ However, during required heating for deprotection of the TMS-ethyl oxalate, the labile allylic alcohol underwent elimination again leading to a fully conjugated tetraene. We next studied direct treatment of allylic alcohol 40 with oxalyl chloride to avoid a deprotection step. Gratifyingly, exposure of allylic alcohol 40 to oxalyl chloride at -78 °C followed by addition of anhydrous Cs₂CO₃ exclusively provided the desired cesium oxalate 41. While our studies were in progress, Wang reported a similar process to access cesium oxalates as radical precursors.²⁷

With a reliable route to the cesium oxalate 41 in hand, we were in a position to study the photocatalytic radical cyclization for synthesis of the CD bicycle of rameswaralide. Irradiation of cesium oxalate 41 with blue LED light in the presence of the photoredox iridium catalyst 43 indeed gave the desired tricyclic ketone 42 as a mixture of three diastereomers (~1:1:1 dr) in 47% yield over 3 steps. We next studied the same sequence developed with α-bromo cycloheptenone 38 with the more elaborate epoxy α-bromo enone 20. However, we quickly learned that the TBS-protected side chain 31 gave low yields during attempted deprotection (cf. 47 → 29, Scheme 7). Therefore, we began this sequence with the TES-protected side chain 37 via a synthesis that followed the corresponding TBS-protected side chain described previously (see Scheme 5). Low reactivity observed with the pinacol borinate during attempted Suzuki-Miyaura coupling to α-bromo enone 20, led us to study the more reactive boronic acid 44 prepared by hydrolysis of the pinacol borinate 38 with buffered NaIO₄ (Scheme 7).²⁸ Subsequent coupling gave dienone 45 in 50% yield (1:1, dr) for the two-step sequence. We were pleased to find that the oxidation state transposition indeed worked well on this more complex epoxy enone 45 to give a mixture of the silylated and non-silylated silyl ethers 47 again proceeding through the presumed extended silyl enol ether 46. Following desilylation with TBAF buffered with AcOH,²⁴ the keto alcohol 29 was obtained in 30% overall yield for the two-step sequence. Once again, direct formation of the Cs-oxalate was achieved by use of oxalyl chloride at −78 °C followed by Cs₂CO₃. Unfortunately, after multiple attempts, we determined that the Giese-type radical cyclization did not deliver the desired pentacycle but only complex product mixtures. This could be attributed to the difference in reactivity between enone 41 and the more extensively conjugated enone 48 which could lead to undesired excitation of the enone under the photocatalytic conditions.

Scheme 7.

Synthesis of the required Cs-oxalate 48 for the intramolecular Giese-type cyclization for D-ring annulation which failed to cyclize.

Scheme 5.

Synthesis of required coupling partners 31, 37 for generation of the allylic radical for proposed intramolecular Giese-type cyclization.

Third D-Ring annulation strategy: An intramolecular enamine-mediated Michael cyclization.

We next investigated a ring annulation strategy involving a key Stork enamine-based²⁹ intramolecular Michael addition. Accordingly, we envisioned ultimately arriving at rameswaralide though a final acetonide deprotection, revealing the β-keto ester moiety (Scheme 8a). Several olefination methods could be employed to install the isopropenyl group from methyl ketone 50. To avoid issues with the oxidation state transposition on late-stage intermediates, we elected to perform the transposition first and then install the methyl ketone, through a Wacker oxidation.³⁰ Despite successful deployment of a Suzuki-Miyura coupling in the previous strategy, we chose to study a Stille coupling in efforts to improve yields of this critical lage-stage coupling with epoxy α-bromo enone 20 now with stannane 53.

Scheme 8.

D-ring annulation strategy #2: (a) Retrosynthetic analysis of rameswaralide for D-ring annulation strategy employing an enamine-mediated Michael cyclization. (b) Model study of the enamine-Michael cyclization leading to synthesis of the CD-ring system of rameswaralide (inset: single crystal X-ray analysis of alkene 42a).

The synthesis of stannane 53 proved to be more scalable than pinacol boraninate 31, avoiding both regioselectivity issues and a protection step (Scheme 8b). While alkylation of the extended enolate generated from vinyl iodide 33 with allyl iodide at -40 °C or higher led to mixtures of α- and γ-alkylation, maintaining a temperature of -78 °C provided only γ-alkylated vinyl iodide 54. Treatment of the resulting vinyl iodide with n-BuLi followed by tributyl tin chloride cleanly gave stannane 53. Stille coupling of α-bromo cycloheptenone 38 with stannane 53 proceeded in 71% yield to give the dienone 55. Wacker oxidation proceeded smoothly to deliver methyl ketone 56, the substrate for the key intramolecular Michael addition, along with traces of the corresponding aldehyde (~19:1 ratio). Simple treatment with super stoichiometric quantities of pyrrolidine³¹ provided low yields of diastereomeric tricyclic methyl ketones 57a,b. A drastic improvement in both reaction rate and yield of the cyclization was noted by inclusion of stoichiometric quantities of pyrrolidine•HCl or pyrrolidine•TFA salts in addition to pyrrolidine (1.0 equiv). While both salts resulted in comparable results, the lower hygroscopic nature of pyrrolidine•HCl ultimately proved ideal to presumably improve the rate of critical proton transfers in these transformations.³² Importantly, we observed the formation of only two of the possible four diastereomers in 42% and 15% yield (~2.5:1 dr), with both diastereomers displaying ring fusion coupling constants (J_5,14 < 8 Hz) indicative of a cis-fused 7,6 ring system compared to trans-fused bicyclic diastereomer (J_5,14 > 10 Hz) obtained through the photochemical route). However, the C1 relative stereochemistry was not readily determined at this stage. Nevertheless, we proceeded with the major diastereomer to study the olefination step. Wittig olefination stalled after ~10% conversion to the desired isopropenyl-substituted bicycle 42a, accompanied by what appeared to be diastereomeric by-products from epimerization under the basic Wittig reaction conditions. To circumvent this issue, we attempted Peterson³³ and Takai/Lombardo³⁴ olefinations with no success. We next turned to a Tebbe olefination³⁵ which provided 30% yield of the desired isopropenyl group along with 33% recovered starting material with no observed epimerization under these conditions. Interestingly, similar methylenation challenges were faced in the formal total synthesis of agelastatin A by Scheinmann,³⁶ studies toward lindene by Baldwin,³⁷ and several other cases.³⁸ In these instances, Julia-Kocienski olefination was successful when other olefination conditions failed. While Barbier-type conditions are typically employed for Julia-Kocienski olefinations to avoid self-condensation of the sulfenolate,^{33, 37-38} we had concerns that use of strong bases may lead to epimerization of several acidic sites in diketone 57a. Therefore, we utilized Kocienski’s phenyl tetrazole reagent 58³⁹ to pre-generate the corresponding sulfenolate which smoothly delivered olefin 42a (68%) with only trace undesired bis-olefinated tricycle 59 (4%).

The NMR spectra of alkene 42a correlated with one of the diastereomers obtained through the photoredox-initiated cyclization (Scheme 6). Ultimately, X-ray analysis of olefin 42a confirmed that the major diastereomer obtained possessed the stereochemistry corresponding to that found in rameswaralide at all three newly formed stereogenic centers. Deprotection of the acetonide through formation of ketene by retrocycloaddition⁴⁰ at elevated temperatures (140 °C)⁴¹ in a 4:1 mixture of mesityelene:MeOH gave the desired β-keto methyl ester 9 in 83% yield, corresponding to the CD ring system of rameswaralide.

Scheme 6.

Model study of the intramolecular Giese-type, 6-endo trig cyclization leading ot the synthesis of the CD-ring system of rameswaralide.

Total Synthesis of Rameswaralide via an Intramolecular Michael Addition Strategy.

Toward applying this successful enamine-mediated cyclization to the real system, we first joined epoxy α-bromo enone 20 and stannane 53 through a Stille coupling to deliver dienone 52 in 70% yield (~20% higher than the corresponding Suzuki coupling, Scheme 9). Exposure of this adduct to TESOTf in the presence of Et₃N resulted in the desired oxidation transposition following desilylation with TBAF/AcOH to deliver the α-methelyene butyrolactone 61 in 44% yield for the two steps. Wacker oxidation of the terminal alkene in tricycle 61 gave the desired methyl ketone 51, however, it was noted that the Wacker oxidation was slower on this substrate relative to the model system. Furthermore, the regioselectivity of this oxidation to provide ketone/aldehyde products was reduced from ~19:1 (model system) to ~7:1 with similar catalyst loadings. Increasing the Pd(OAc)₂ and CuCl led to slight improvements in both reaction rate and ketone selectivity (~10:1). When this ketone was subjected to Stork enamine cyclization conditions optimized through the model study, we were pleased to once again isolate only two diastereomeric ketones however, instead of both cis-fused products as obtained in the model study, we isolated cis-fused and trans-fused pentacyclic ketones (50a and 50b respectively). Based on nOe analysis of the two isolated diastereomers, and comparison to the natural product itself, we proposed that the major diastereomer once again correlated to rameswaralide at all three newly formed stereogenic centers by extensive NMR analysis which was subsequently confirmed by X-ray analysis. Interestingly, the cycloheptenone adopts a conformation in which the plane of the C6-ketone is ~80° off plane of the adjacent alkene which may explain the low reactivity toward Diels-Alder reactions noted by Mehta.^5a This conformation also has a profound impact on the chemical shifts of protons positioned in the deshielding (C4, C7, and C14 in 50a) and shielding (C14 in 50b) cone of the ketone due to anisotropic effects. Application of the Julia-Kocienski olefination, developed in the model system, with the major diastereomeric ketone 50a led to the desired pentacyclic alkene 49 in 75% yield. Subjecting acetonide 49 to the same deprotection conditions successful in the model system, resulted in extensive decomposition. However, simply lowering the temperature by 20 °C (from 140 °C to 120°C), allowed cleaner conversion of acetonide 49 to (±)-rameswaralide (1) with minimal decomposition after heating for >34 h leading to the first total synthesis of (±)-rameswaralide. All spectral data was consistent with that previously reported for rameswaralide in two separate reports (see p. 38, SI).^{1, 3}

Scheme 9.

D-ring annulation strategy #3: Synthesis of enamine-Michael cyclization substrate 61 and cyclization, olefination, and deprotection leading to (±)-rameswaralide (1).

The high diastereoselectivity (~6.5:1) observed in the enamine-based ring annulation leading to three new stereogenic centers, albeit proceeding in modest yield, warrants some discussion. The stereochemical setting step is the initial Michael addition leading to formation of the C1─C14 bond (Figure 2). As judged by molecular models and DFT calculations, due to geometrical constraints of the tethered dioxinone, the C1-vinyl proton of the reactive enamine conformer approaches the tricyclic core either from the top (indicated as concave addition) or from the bottom (convex addition) of the dieneone regardless of enamine geometry (shown as (Z)-geometry). To avoid lone pair repulsion of the carbonyl oxygens between the ketone and dioxinone that would occur upon addition from the convex face while maintaining conjugation, approach from the concave face appears favored. Developing torsional strain also appears to play a role in the observed diastereoselectivity. The strained cycloheptadienone has the C13-proton positioned slightly above the C14-proton, as shown in Fig. 2,with an ~30° dihedral angle (angles obtained from DFT geometry optimization of enamines i and ii at the B3LYP 6-31Gd,p level using Gaussian 16). Thus, in the transition state arrangement leading to the major diastereomer 50a, approach from the concave face avoids an unfavorable developing torsional strain between the C13/C14 protons which would transpire on approach from the convex face leading to 50b. Following this initial carbon-carbon bond formation, we anticipate that the C5-stereogenic center is derived from a kinetically-controlled protonation of the incipient enol from the more accessible convex face leading to the less strained C5,C14-cis-ring fusion. Finally, a thermodynamically-driven equilibration through enamine/iminium intermediates may lead to the methyl ketone being in a more favorable pseudo-equatorial position for both diastereomers 50a,b.

Figure 2.

Proposed transition state arrangements for the enamine-mediated Michael D-ring annulation process leading to ketones 50a,b.

Bioactivity of rameswaralide, synthetic intermediates, and truncated model systems accessed through a pharmacophore-directed retrosynthetic strategy.

Rameswaralide (1) and synthetic intermediates were assayed for cytotoxicity against the A549 lung cancer cell line along with the model system bicyclic structures which overlay with the CD ring system. Our synthetic (±)-rameswaralide showed some inhibition of A549 cells, albeit with low potency (apparent EC₅₀ ~200 μM) (Figure 3). This activity is consistent with the IC₅₀ value of 67 ± 3.7 μM reported for (−)-rameswaralide in the same cancer cell line which was the most potent activity reported for the natural product.³ The acetonide protected precursor of rameswaralide 49 and other acetonide protected compounds 50a, and 57a, were devoid of significant activity at the highest concentrations tested (>200 μM). . Interestingly, the simplified bicyclic β-keto ester 9 corresponding to only the CD ring system of rameswaralide displayed greater cytotoxicity (apparent EC₅₀ ~100-200 μM) compared to racemic rameswaralide toward A549 cells. Moreover, the acetonide protected version of this simplified bicycle 42a also showed greater cytotoxicity to A549 cells than rameswaralide despite showing reduced solubility at concentrations ≥ 100 μM. Surprisingly, the bicyclic methyl ketone 60 did not exhibit significant bioactivity (apparent EC₅₀ >200 μM) indicating that the more polar methyl ketone significantly impacts bioactivity. The small number of simplified derivatives does not allow a systematic structure-activity relationship profile to be ascertained, however our studies revealed antiproliferative activity for simplified versions of rameswaralide (e.g. 42a, 9).

Figure 3.

Cytotoxicity of rameswaralide and synthetic intermediates toward the A549 lung cancer cell line.

Summary.

In conclusion, we report the first total synthesis of the caged cembranoid (±)-rameswaralide (1) employing a pharmacophore-directed retrosynthetic strategy. Key reactions in the total synthesis include an organocatalytic Diels-Alder lactonization organocascade which provides rapid access to the common 5,5,6- tricycle core of several Sinularia natural product family members. In addition, this core can be accessed in optically active form through the first kinetic resolution involving our Diels-Alder lactonization organocascade which constitutes one of the few kinetic resolutions involving a Diels-Alder cycloaddition.⁴² Finally, a serendipitous, doubly vinylogous, E1-type elimination of an epoxy enone led to a desired net oxidation state transposition and paved the way to study three distinct D-ring annulation strategies. These strategies included a photoredox initiated 6-endo-trig Giese-type cyclization process and the ultimately successful Stork enamine-based intramolecular Michael addition. Application of PDR necessitated a more linear synthetic strategy to enable access to intermediates possessing our proposed 5,5,7-tricyclic pharmacophore which ultimately led to the discovery of a previously disclosed, selective cytotoxic compound, (±)-epoxy α-bromo enone 20 (cf. Scheme 2),^5h with high potency (EC₅₀ 0.39 ± 0.03 μM) toward A549 cells and selectivity (8-27 fold) over other cancer cell lines studied. In addition, we surprisingly found that the simplified CD ring system of rameswaralide possessing a β-keto ester displayed increased antiproliferative activity compared to (±)-rameswaralide.^5h We are currently pursuing identification of the cellular target(s) of epoxy α-bromo enone 20 and also a resolution to differentiate the activity of each enantiomer. Our studies revealed an interesting trend toward the increased bioactivity of the 5,5,7 core compared directly to the corresponding 5,5,6 core.^5h Overall, in addition to the completion of the preliminary biological tests of rameswaralide and simplified derivatives described herein, our results suggest that further biological studies of bicyclic β-ketoesters may be warranted.

ABBREVIATIONS

PDR

pharmacophore-directed retrosynthesis

BTM

benzotetramisole

IMDA

intramolecular Diels-Alder

MS

mass spectrometry

TLC

thin layer chromatography