Keywords

13.1 Introduction

In 2012, the structure of (+)-aquatolide, a humulane-derived sesquiterpenoid lactone, was revised from 1a to 1b (Fig. 13.1) by Shaw et al. [1]. This terpenoid was originally isolated by San Feliciano et al. from Asteriscus aquaticus [2]. Previously, structurally related sesquiterpenoids asteriscunolides A–D (25), called humulanolides, were isolated from the same source [3,4,5]. The proposed structure of aquatolide 1a consisted of a tetracyclic 5/4/4/7-ring system with a characteristic bicyclo[2.2.0]hexane motif. However, the Shaw and Tantillo group found that the calculated NMR data for structure 1a were inconsistent with those reported by San Feliciano et al. Revised structure 1b was assigned by computational chemistry and confirmed by X-ray crystallography. We were interested in structure 1b, which has an unusual, intricate bridged 5/5/4/8-ring system.

Fig. 13.1
A chart presents the following chemical structures. Aquatolide, negative asteriscunolide A, negative asteriscunolide B, negative asteriscunolide C, and negative asteriscunolide D. Each molecule has a complex arrangement of carbon, hydrogen, and oxygen atoms with hydroxyl, carbonyl, and methoxy groups.

Structures of (+)-aquatolide and (–)-asteriscunolides A–D

The biosynthesis of aquatolide has been proposed to involve a transannular [2 + 2] cycloaddition of (–)-asteriscunolide C (4) (Fig. 13.2). Parallel addition forms the C2–C10 and C3–C9 bonds in proposed structure 1a, whereas crossed addition forms the alternate bonds (C2–C9 and C3–C10) to afford real structure 1b. Although the racemic synthesis of 1b has been completed by two groups [6, 7], the total synthesis via a biomimetic [2 + 2] cycloaddition has not been reported. We decided to attempt the biomimetic approach, and then complete the total synthesis of the natural enantiomer (+)-aquatolide (1b). After much effort, we have achieved the total synthesis of 1b [8]. In this chapter, we described our endeavors toward the synthesis of 1b, including some unsuccessful approaches.

Fig. 13.2
Chemical reactions present the biosynthesis of aquatolide. It is a double step process that begins with asteriscunolide that undergoes parallel and crossed cycloaddition to form the originally proposed structure of aquatolide and plus aquatolide.

Proposed biosynthesis of aquatolide

13.2 Unsuccessful Route: ROM/RCM/RCM Approach

Our retrosynthetic analysis of (+)-aquatolide (1b) is shown in Scheme 13.1. The advanced intermediate was a putative biosynthetic precursor of 1b, (–)-asteriscunolide C (4), which has a disubstituted γ-butenolide skeleton. For concise access to γ-butenolides, we have reported the ring-opening/ring-closing metathesis (ROM/RCM) reaction of cyclobutenecarboxylates in the total synthesis of (+)-clavilactone A (9) [9,10,11]. In the present work, we expected to construct the asteriscunolide skeleton through a combination of the ROM/RCM approach and a ring-closing metathesis (RCM) reaction. Namely, compound 4 could be obtained by RCM expelling ethylene from γ-butenolide 6, which would be derived from cyclobutenecarboxylate 7 by the ROM/RCM reaction. Sequential metathesis reactions can be performed as one-pot reactions. Substrate 7 would be synthesized by our acylation method from alcohol 8.

Scheme 13.1
2 reaction schematics. Top. Plus aquatolide on transannular cycloaddition forms minus asteriscunolide to form compound 6 through R C M, compound 7 to R O M and R C M and compound 8 through acylation. Bottom. Cyclobutenecarboxylate on treatment with R O M, R C M forms plus clavilactone A.

First-generation retrosynthetic analysis of (+)-aquatolide (1b)

Preliminary experiments were performed as racemates (Scheme 13.2). Known racemic secondary alcohol 10 [12] was treated with DDQ under anhydrous conditions to provide acetal 11, which was reduced with DIBAL-H to primary alcohol 12. Parikh–Doering oxidation of 12 afforded aldehyde 13. The lithium enolate generated from ketone 14 reacted with aldehyde 13 to give aldol adduct 15 as a diastereomeric mixture (d.r. = 2:1). β-Elimination of 15, followed by removal of the MPM group in resultant dienone 16, provided alcohol rac-8. According to our previous procedure [9,10,11], the acylation of rac-8 was achieved by using acid anhydride 17. Cyclobutenecarboxylate rac-7 was obtained, and a substrate for the planned sequential metathesis was synthesized.

Scheme 13.2
A synthesis of the molecule rac-7, which is a cyclo butene carboxylate. It starts with M P M O-protected cyclohexenone and T B S O through a 7-step process. Some of the reagents used in this phase include hydrogen cyanide, ether or dimethoxy ethane, lithium tetramethyl piperide, methyl iodide, T B S O C l, and isopropoxide triethylamine complex.

Synthesis of cyclobutenecarboxylate rac-7

With substrate rac-7 in hand, we attempted the ROM/RCM/RCM reaction (Scheme 13.3). Cyclobutenecarboxylate rac-7 was treated with the Grubbs catalyst in hot toluene. Unfortunately, desired (±)-asteriscunolide C (rac-4) was not obtained, and only dimerized product 18 was obtained in approximately 30% yield. The ROM/RCM reaction proceeded to form a γ-butenolide skeleton-like compound 6 (Scheme 13.1), but the final RCM reaction failed. In anticipation of different reactivity, several additional substrates (1921) were synthesized. However, the asteriscunolide skeleton could not be constructed by any of the metathesis reactions.

Scheme 13.3
An R O M, R C M, and R C M reaction for the synthesis of the asteriscunolide skeleton with 6 attempted reactions, each involving a starting material like rac-7 or rac-4 and Grubbs catalyst. The attempted reactions failed to produce the desired asteriscunolide skeleton.

Attempted ROM/RCM/RCM reactions

The conformational inflexibility of the substrates was thought to be responsible for the failure of the RCM. It was presumed that the E-olefin in rac-7 and 19 or the oxygen substituents in 20 and 21 could have an adverse effect, preventing the alkene partners from getting close enough for metathesis. Therefore, we next turned our attention to preparing more flexible substrates for the sequential metathesis reaction (Scheme 13.4). Chemoselective addition of a vinyl group to known cyanoaldehyde 22 [13] afforded alcohol 23. After silylation of 23, resulting nitrile 24 was reduced to aldehyde 25 by DIBAL-H reduction followed by hydrolytic work-up. The isopropenyl Grignard reagent reacted with aldehyde 25 to provide adduct 26 as a mixture of diastereomers (d.r. = 1:1). Via a simple sequence of reactions, TBS ether 26 was converted to MPM ether 27, which was acylated by the same method as in the synthesis of rac-7 to give cyclobutenecarboxylate 28. As further substrates for metathesis, alcohol 29 and ketone 30 were synthesized from 28 by removal of the MPM group and Dess–Martin oxidation.

Scheme 13.4
A synthesis of cyclobutene carboxylates 28, 29, and 30 starts with vinyl magnesium bromide and toluene, followed by reactions with isocyanides to form intermediates. Then, DIBAL-H affords compounds 25 and 26. A two-step process using M P M C I and T B A F yields the final products.

Synthesis of cyclobutenecarboxylates 28, 29, and 30

Using new substrates 2830, the ROM/RCM/RCM reaction was attempted again (Scheme 13.5). When the second-generation Hoveyda–Grubbs catalyst was applied to substrates 28 and 29, the reaction proceeded and moderate yields of the 11-membered products were obtained (15% and 27%, respectively). However, NOE experiments on the products showed that the geometry of the ring-closing site olefin was the undesired E-configuration, and compounds 32 and 34 were produced by the RCM. Furthermore, no cyclized product was obtained from the reaction of ketone 30. At this stage, we decided to abandon this approach because the metathesis yields were lower than expected and we believed that the Z-configuration was required for preparing the precursor of (+)-aquatolide (1b). Later, we realized that the E-configuration would work. This allowed us to develop a concise, high-yielding synthetic route for 1b. Independently, Li et al. have reported the total synthesis of (–)-asteriscunolide D (5) by the ROM/RCM/RCM approach [14, 15].

Scheme 13.5
An R O M reaction followed by 2 R C M reactions to synthesize cyclobutene carboxylate 29 from 28 which is the starting material. Hoveyda-Grubbs 2 catalyst initiates the reaction in toluene under reflux conditions. The sequence results in the formation of cyclobutene carboxylate 29 in 34% yield.

ROM/RCM/RCM reaction of cyclobutenecarboxylates 28 and 29

13.3 Successful Synthetic Strategy Toward (+)-Aquatolide

Next, we combined cross-metathesis (CM) with the ROM/RCM approach instead of RCM. This approach required a reaction to construct the 11-membered ring. We planned to rely on an intramolecular Nozaki–Hiyama–Takai–Kishi (NHTK) reaction [16,17,18,19,20], which we have used extensively [21,22,23,24]. A second-generation retrosynthetic analysis of (+)-aquatolide (1b) was conducted (Scheme 13.6). In this analysis, all-trans-humulene lactone 36 was set as a key intermediate. By tuning the structure of 36, we expected to find asteriscunolide-type compound 35 that could undergo the transannular [2 + 2] cycloaddition. Key intermediate 36 could be cleaved to iodoalkene–aldehyde 37 by the intramolecular NHTK reaction, and then 37 would be synthesized by the ROM/RCM/CM reaction of cyclobutenecarboxylate 38 with methacrolein (39).

Scheme 13.6
A retrosynthetic analysis of plus aquatolide disconnects multiple bonds in the molecule, breaking it down into a 5-membered lactone ring, a vinyl iodide containing a carbon-iodine bond, and a carbonyl compound with a C O double bond.

Second-generation retrosynthetic analysis of (+)-aquatolide (1b)

13.4 Construction of the Asteriscunolide Skeleton

The second-generation synthesis was performed asymmetrically. Readily available D-(–)-pantolactone (40) was chosen as the starting material (Scheme 13.7). After protection of 40, DIBAL-H reduction of MPM ether 41 followed by Wittig reaction of resulting lactol 42 provided known alcohol S-12 [25] in 98% ee, confirmed by chiral HPLC analysis. Dess–Martin oxidation of S-12 afforded aldehyde S-13, and a large-scale Takai–Utimoto olefination [26] was investigated, for which economic conditions were found. Chromium(II) chloride (CrCl2) reduced from less expensive CrCl3 with LiAlH4 [27] was also effective in this reaction, yielding E-iodoalkene 43. Cyclobutenecarboxylate 38 was obtained by deprotection of 43 and acylation of alcohol 44 with acid anhydride 17, and the stage was set to develop the ROM/RCM/CM reaction.

Scheme 13.7
A synthesis of cyclobutene carboxylate 38 starts with D-pantolactone and undergoes several steps to reach the final product. Reagents used in the pathway include triflic acid, chlorinated imine, diethyl ether, T H F, an oxidizing agent, iodoform, B U O K, and dichloromethane.

Synthesis of cyclobutenecarboxylate 38

First, we examined the ROM/RCM and CM reactions in a stepwise manner (Scheme 13.8). In our previous work [9,10,11], we used the first-generation Grubbs catalyst for the ROM/RCM, but the same conditions were not suitable for the reaction of 38. Fortunately, the second-generation Grubbs catalyst showed good activity and corresponding γ-butenolide 45 and its dimerized product were obtained. Both products were reacted with methacrolein (39) in the presence of the second-generation Grubbs catalyst to afford α,β-unsaturated aldehyde 37 as a single E-isomer in 30% overall yield. Next, a cascade method was examined. A mixture of 38, the catalyst, and methacrolein (39) in toluene was heated, and desired product 37 was produced directly in an improved yield (60%). As expected, the iodoalkene moiety was not involved in the metathesis. The cascade ROM/RCM/CM reaction was established and became a central part of this work.

Scheme 13.8
An R O M, followed by a R C M, and a C M to synthesize molecule 45 from cyclobutene carboxylate 38. The reaction is initiated by Grubbs 2 catalyst in toluene at a reflux temperature of 60 degrees Celsius to form molecule 45 in 60% yield.

ROM/RCM/CM reaction of cyclobutenecarboxylate 38

The next task was forming the 11-membered ring (Scheme 13.9). The intramolecular NHTK reaction of 37 formed the asteriscunolide skeleton to provide 36 in a remarkable yield (96%). Once again, the NHTK reaction exhibited tremendous power. Cyclized product 36 was obtained as a single diastereomer with a pseudoequatorial hydroxy group. This is a common trend in NHTK reactions [20,21,22,23,24]. Alcohol 36 was oxidized to (–)-asteriscunolide D (5). Thus, the total synthesis of 5 was achieved in 10 steps from 40 with an overall yield of 32%, which was approximately four or seven times higher than the yields of previously reported syntheses [14, 15, 28].

Scheme 13.9
A total synthesis reaction of negative asteriscunolide D 5 starts with D-pantolactone 40 and uses a thionium ion intermediate for a key ring formation step to create the 11-membered ring structure. The stereochemistry is controlled throughout the synthesis.

Total synthesis of (–)-asteriscunolide D (5)

13.5 Synthesis of the Proposed Structure of Aquatolide (1a)

Because asteriscunolide and asteriscunolide-type compounds were obtained, the [2 + 2] cycloaddition was investigated (Scheme 13.10). First, we irradiated (–)-asteriscunolide D (5) with a high-pressure Hg lamp (100 W). However, only olefin isomerization was observed and (–)-asteriscunolide A (2) was obtained as the major product. Although a small amount of (–)-asteriscunolide C (4) was produced, [2 + 2] cycloadducts were not detected. The Li group reported that irradiation of 5 with a UV lamp (10 W, 254 nm) gave similar results [14, 15]. We concluded that compound 5 was an unsuitable substrate for the direct synthesis of (+)-aquatolide (1b) via a photochemical reaction.

Scheme 13.10
A photoisomerization reaction of negative asteriscunolide D 5 to yield a new isomer exhibits irradiation of negative asteriscunolide D 5 with light yielding a new isomer which is isoausteriscunolide D in 88% yield.

Irradiation of (–)-asteriscunolide D (5)

Next, we irradiated dienol 36 instead of dienone 5 (Scheme 13.11). In this reaction, chemoselective isomerization of the trisubstituted olefin gave compound 46, but the desired [2 + 2] cycloaddition did not occur. We thought that the more reactive trisubstituted olefin needed to be masked. Therefore, 46 was epoxidized with m-CPBA. The reagent approached from the outside of the 11-membered ring, yielding only isomer 47. As expected, the photoreaction of epoxide 47 gave a [2 + 2] cycloadduct. In fact, the cycloaddition proceeded in parallel mode and ladder-like adduct 48 was formed, although at this stage, we did not notice the undesired result. Oxidation of 48 afforded keto-epoxide 49, which was reduced to the final compound. Unfortunately, the compound was not (+)-aquatolide (1b). The 1H and 13C NMR data for the synthetic sample did not match those for 1b, but were consistent with those calculated by Shaw and Tantillo’s group for the proposed structure of aquatolide (1a) [1]. Consequently, we realized that it was non-natural product 1a.

Scheme 13.11
A multi-step synthesis of the proposed structure of aquatolide from acetone through reactions with a high-pressure mercury lamp, m-C P B A, Dess-Martin periodinane, and a final ring-forming step.

Synthesis of the proposed structure of aquatolide (1a)

13.6 Completion of the Total Synthesis of (+)-Aquatolide (1b)

We used an oxy-Michael reaction as an alternative way to mask the olefin and expected the corresponding adduct to give different results from epoxide 47. Therefore, we investigated the conditions for regioselective 1,4-addition of methanol to (–)-asteriscunolide D (5) (Scheme 13.12). Under basic conditions (NaOMe/MeOH), low regioselectivity was observed, giving a complex mixture of 1,4-adducts to di- and/or tri-substituted olefins. Fortunately, the reaction with acids such as BF3·OEt2 in methanol provided 1,4-adduct 50 with high regio and stereoselectivity. The crossed [2 + 2] cycloaddition of 50 proceeded to provide desired product 51 with the bicyclo[2.1.1]hexane core. Thus, we achieved the first biomimetic transannular [2 + 2] cycloaddition for the synthesis of aquatolide. Treatment of 51 with BF3·OEt2 afforded the eliminated product and completed the total synthesis of (+)-aquatolide (1b). Compared with previous racemic syntheses [6, 7], our route was shorter (13 steps from 40) and resulted in a high yield (5.7% overall yield).

Scheme 13.12
A total synthesis of plus aquatolide undergoes two main steps. The first step removes a protecting group from negative asteriscunolide D using T F A and dichloromethane at 0 degrees Celsius. The deprotected intermediate undergoes a light-induced cycloaddition to form plus aquatolide in 74% yield in the next step.

Total synthesis of (+)-aquatolide (1b)

Although the transannular [2 + 2] photocycloaddition of compound 47 provided parallel product 48 (Scheme 13.11), the reaction of compound 50 gave crossed adduct 51 (Scheme 13.12). We analyzed the extremely high regioselectivity in these reactions by conformational searches of the substrates (Fig. 13.3). H-2 and H-10 were cis to each other in the most stable conformer of epoxide 47. In contrast, these hydrogens were trans in 50. Therefore, the parallel cycloaddition of 50 would not proceed because a trans-fused 4/4-ring system is impossible. These results suggested that the parallel or crossed modes of cycloaddition are controlled by the conformation of the substrates.

Fig. 13.3
A regioselectivity reaction of photocycloadditions of molecules 47 and 50. It involves cycloaddition between double bonds to form a new cyclobutane ring.

Regioselectivity of [2 + 2] photocycloadditions of 47 and 50

13.7 Total Syntheses of Related Humulanolides

In addition, we investigated the syntheses of other related humulanolides (Scheme 13.13). In the photoreaction of (–)-asteriscunolide D (5), the total synthesis of (–)-asteriscunolide A (2) was achieved (11 steps from 40 with an overall yield of 14%) (Scheme 13.10), but pure (–)-asteriscunolide C (4) was not isolated. In contrast, the oxidation of alcohol 46 (Scheme 13.11) afforded pure 4 (11 steps from 40 with an overall yield of 13%). Chemo and stereoselective epoxidation of alcohol 36 with m-CPBA achieved the first total synthesis of (–)-asteriscunolide I (52) [29], a recently isolated humulanolide (10 steps from 40 with an overall yield of 26%). Our next goal was to construct the asteriscanolide skeleton, a tricyclic 5/5/8-ring system. A clue to its construction was found by chance in a study of the 1,4-addition of methanol to 5 (Scheme 13.12). When 5 was treated with n-Bu3P in methanol, an intramolecular Rauhut–Currier (vinylogous Morita–Baylis–Hillman) reaction [30, 31] occurred to afford (+)-tetradehydroasteriscanolide (53) [32]. After optimization, the yield was improved and the efficient total synthesis of 53 was also completed (11 steps from 40 with an overall yield of 32%).

Scheme 13.13
A total synthesis reaction of several related humulanolides. The top section exhibits a 2-step process involving deprotection and a biomimetic transannular photocycloaddition reaction with a 74% yield. It involves several steps including Dess-Martin periodinane oxidation and a stereoselective reduction with n-B u 3 P dot H.

Total syntheses of related humulanolides

13.8 Biological Activity of Natural Humulanolides and Analogs

Although several humulanolides show anti-tumor activity [33, 34], their target molecule has not been identified. We conducted a structure–activity relationship study using synthetic samples of natural humulanolides and their analogs, expecting to elucidate the mode of action. First, two additional compounds, 54 and 55, were prepared by oxidation of the corresponding alcohols (Scheme 13.14) [35].

Scheme 13.14
A retrosynthetic analysis for the synthesis of humulanolide analogs with a seven-membered lactone ring and a cyclobutane unit as the target molecule. It breaks down the molecule into a 6-membered ring ketone and an allylic alcohol containing a double bond next to an alcohol group.

Synthesis of humulanolide analogs

Twelve compounds were selected, and anti-proliferative activity was examined against eight human cancer cell lines. Whereas most compounds, including aquatolide, were inactive (1a, 1b, 36, 47, 49, 52, 53, and 55), asteriscunolide A (2), asteriscunolide C (4), and asteriscunolide D (5) showed some activity. 54 was the most potent compound, exhibiting anti-proliferative activity against all tested cell lines (e.g., the IC50 values against human gingival carcinoma cell line Ca9-22: 9.9 μM for 2, 7.5 μM for 4, 4.8 μM for 5, and 2.9 μM for 54). The results suggested that the unsaturated carbonyl moiety on the 11-membered ring is essential for the anti-cancer activity and that the stereochemistry around the epoxide moiety is also important.

A further biological study was conducted using 54. 54 caused morphological changes in cells similar to those observed during geldanamycin treatment. Thus, heat-shock protein 90 (HSP90), which is a chaperone that helps proper protein folding, was probably the target protein. Actually, we confirmed that 54 increased the expression of HSP70 and decreased that of HSP90 client proteins, such as AKT and CDK4.

13.9 Conclusion

We were inspired by the biosynthesis of (+)-aquatolide (1b) to achieve the efficient total synthesis of 1b. In the early stage of the synthesis, the cascade ROM/RCM/CM reaction of cyclobutenecarboxylate was developed to construct the γ-butenolide with an unsaturated aldehyde side chain. The intramolecular NHTK reaction efficiently formed an all-trans-humulene lactone skeleton. Finally, the transannular [2 + 2] photocycloaddition of an asteriscunolide-like compound was realized in a crossed mode. In addition, the [2 + 2] cycloaddition proceeding in a parallel mode was also found. Thus, we established a concise, high-yielding synthetic route to 1b. Related humulanolides (2, 4, 5, 52, and 53) were also synthesized with our strategy and a structure–activity relationship study was performed using the synthesized samples. We hope that our findings will contribute to the development of natural product synthesis.