Enantioselective Total Syntheses of (+)-Hippolachnin A, (+)-Gracilioether A, (-)-Gracilioether E and (-)-Gracilioether F

The Plakortin polyketides represent a structurally and biologically fascinating class of marine natural products. Herein, we report a unified strategy that enables the divergent syntheses of various Plakortin polyketides with high step-economy and overall efficiency. As proof-of-concept cases, the enantioselective total syntheses of (+)-hippolachnin A, (+)-gracilioether A, (-)-gracilioether E and (-)-gracilioether F have been accomplished based on a series of bio-inspired, rationally designed or serendipitously discovered transformations, which include 1) an organocatalytic asymmetric 1,4-conjugate addition to assemble the common chiral -butenolide intermediate en route to all of the aforementioned targets, 2) a challenging biomimetic [2+2]-photocycloaddition to forge the oxacyclobutapentalene core of (+)-hippolachnin A, 3) a [2+2]-photocycloaddition followed by one-pot oxidative cleavage of methyl ether/Baeyer-Villiger rearrangement to access (-)-gracilioether F, and 4) an unprecedented hydrogen-atom-transfer (HAT)-triggered oxygenation of vinylcyclobutane to afford (+)-gracilioether A and (-)-gracilioether E in one pot.


INTRODUCTION
The Plakortin polyketides constitute a growing family of marine natural products that display remarkable structural and biological diversity. 1 As one of the most prominent family members, hippolachnin A (1, Figure 1) was identified by Lin and coworkers from the South China Sea sponge Hippospongia lachne in 2013. 2 Preliminary biological evaluation revealed that hippolachnin A exhibited potent antifungal activity against three pathogenic fungi including Cryptococcus neoformans, Trichophyton rubrum, and Microsporum gypseum, with a MIC value of 0.41 μM for each species. 2 Therefore, it represents a new chemotype of anti-fungal agents or leads. Further studies suggested that hippolachnin A could also potentially function as a therapeutic agent to treat various diseases such as renal fibrosis, chronic heart failure and rhinitis. 3 Structurally, hippolachnin A bears an unprecedented molecular architecture that features a highly strained, bowl-shaped 5/5/4 tricyclic core. Moreover, it contains six consecutive stereogenic centers, four of which bear an ethyl substituent projected toward the convex orientation. Besides hippolachnin A, a series of closely relevant congeners, namely gracilioethers, have also been discovered in nature, as exemplified by the structures 2-8. 4 In analogy to hippolachnin A (1), most of gracilioethers possess a typical tricyclic system, in which the A and B rings are relatively conserved, but the C rings vary greatly in regard to the ring sizes (4-, 5-and 6-membered rings) and oxidation patterns (lactone, furan and 1,2-dioxane). Of note, the gracilioether family is also endowed with diverse biological profiles. For example, gracilioether A (2) shows promising antimalarial activities against Plasmodium falciparum (IC50 = 10 µg/mL) and gracilioether H (6) displays significant antiplasmodial activity against chloroquine resistant CR FC29 strain (IC50 = 3.26 M). 4a,4b In adidtion, several members of gracilioethers such as 3, 5, 7 and 8 exhibit notable pregnane-X-receptor (PXR) agonistic activity, which renders them the promising leads for the development of antiinflammatory drugs. 4c Not surprisingly, the fascinating molecular architectures and promising biological profiles of hippolachnin A and gracilioethers stimulated extensive interests from the synthesis community, [5][6] which has culminated in a number of elegant total syntheses of these targets. In 2015, Carreira and coworkers completed the first total synthesis of (±)-hippolachnin A (1), with an intermolecular [2+2]-photocycloaddition and an intramolecular ene-cyclization employed as key steps. 5a Meanwhile, an asymmetric synthesis of the key intermediate en route to (+)-1 was also described in this work. Subsequently, the Wood and Brown groups collaboratively accomplished a landmark seven-step total synthesis of (±)-1, which hinges on an intriguing [2π + 2σ + 2σ] cycloaddition of quadricyclane and a late-stage intramolecular allylic C-H oxidation. 5b More recently, another impressive total synthesis of (±)-1 was achieved by the Trauner group, the key elements of which include the photoisomerization of tropolone and an unusual thiocarbonyl ylide-mediated [3+2] cycloaddition. 5c Besides above-mentioned total syntheses, a formal synthesis (±)-1 and an asymmetric synthesis of the tricyclic core of (+)-1 were reported by the Wu 5d and Ghosh 5e,5f groups respectively, both of which employed the bio-inspired [2+2]-photocycloadditions as the key steps. As to the gracilioethers, the Brown group completed the first total synthesis of (±)-gracilioether F (4) in 2014, featuring a ketene-alkene [2+2] cycloaddition and a latestage carboxylic acid directed C-H oxidation. 6a In 2016, another two total syntheses of (±)-gracilioether E (3) and (or) F (4) were successively reported by the Carreira 6b and Wong groups. 6c In the early of 2017, Wu and co-workers disclosed a total synthesis of 4 as well as a formal synthesis of 3. 5d More recently, an asymmetric synthesis of (−)-gracilioether E (3) was achieved by the Ghosh group based on a chiral pool strategy. 6d In spite of great advances, there exist some unmet challenges for the total syntheses of hippolachnin A and gracilioethers. First of all, most of the previous syntheses were achieved in a racemic manner. So far, only one asymmetric total synthesis of 3 6d and one asymmetric formal synthesis of 1 5a have been documented. In this context, the asymmetric syntheses of these targets remain relatively underdeveloped. Secondly, only limited members (1, 3 and 4) of this family of natural products have been conquered by synthetic chemists, and some more challenging targets such as 2 and 5- 8 have not yet succumbed to total synthesis. Thus, it has remained an unfulfilled task to develop a flexible synthetic approach enabling the access of diverse polycyclic Plakortin polyketides, particularly those ones bearing different frameworks. Last but not least, while the biosynthetic origins of hippolachnin A and gracilioethers seem to be inspiring, their real biomimetic syntheses have not been realized yet. Taking hippolachnin A as an example, although the bio-inspired [2+2]-photocycloadditions have been successfully realized with some truncated substrates, 5d-f the attempts to effect the [2+2]-photocycloaddition with the biosynthetic precursor to 1 turned out to be problematic, 5d,6e which casted a shadow on such a biomimetic approach. Herein, we report our own contribution on this topic, which led to the development of a unified synthetic route that can address all of the above-mentioned challenges encountered for the chemical syntheses of hippolachnin A and gracilioethers. Hinging on a series of bio-inspired, rationally designed or serendipitously discovered transformations, the presented chemistry enables the divergent syntheses of (+)-hippolachnin A, (+)-gracilioether A, (-)-gracilioether E and (-)-gracilioether F from a common intermediate with high step-economy and overall efficiency. Our synthetic strategy for hippolachnin A and gracilioethers was largely built on their plausible biosynthetic origins that are rationalized in Figure 2A. We assumed that both hippolachnin A and gracilioethers could be traced back to three monocyclic precursors 9a-c. 4a,7 In particular, hippolachnin A (1) could arise from 9a through an intramolecular [2+2]-photocycloaddition, 8 whereas gracilioether A (2) might be derived from 9b through an aerobic [2+2+2]-photocycloaddition. 9 In analogy, 9a and 9c could also undergo [2+2+2]-photocloaddition to give the corresponding products 2a and 2c, which can further advance to the other gracilioether congeners through late-stage structural diversification. For example, 2a could be converted to gracilioether H (6) through oxidative cleavage of the C2=C3 double bond, and 2c might undergo sequential Kornblum-DeLaMare rearrangement (KDR), 10 hemiketalization and oxa-Michael addition to afford gracilioether K (8). In addition, the hemiketal intermediate 11 could also divert to gracilioethers E (3) and F (4) through oxidative cleavage of the C10-C11 single bond or (and) C2=C3 double bond.

RESULTS AND DISCUSSION
According to above rationalization, our retrosynthetic analysis is then traced back to the proposed biosynthetic precursors 9a-c. Further bond-disconnection of 9a-c reveals the chiral -butenolide 12 as a common precursor. Notably, although various methods have been developed for the synthesis of chiral butenolides, 11 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 Scheme 1. Enantioselective Biomimetic Total Synthesis of (+)-Hippolachnin A approach to access the highly functionalized chiral butenolides like 12 remained a considerable challenge at the outset of this work. Inspired by the pioneer work of MacMillan 11b and Pihko, 11c,11d we envisioned that an organocatalytic asymmetric Mukaiyama-Michael addition might meet this challenge, which necessitates the enal 13 and silyloxyfuran 14 as precursors. Coincidently, a synthesis of racemic 12 was reported previously by Wu and co-worker. 5d However, their attempt to achieve the enantioselective synthesis of 12 through the asymmetric Mukaiyama-Michael reaction turned out to be unsuccessful. We commenced our study by exploring the organocatalytic asymmetric Mukaiyama-Michael addition. For this end, the known compound 3,5-diethylfuran-2(5H)-one 15 12 was first converted to the silyloxyfuran 14a in a quantitative yield under the standard conditions (TBSOTf/Et3N). In view of the fragile nature of 14a, it was directly used in the next step without purification. Initially, the readily accessible Jørgensen-Hayashi catalyst A was employed to effect the Mukaiyama-Michael addition. To our delight, the reaction did work, affording the desired product 12 and its diastereoisomer 16 in 61% combined yield albeit with low enantiomeric excess (entry 1, Table 1). Simply replacing the silyloxyfuran 14a with 14b increased the ee value slightly (entry 2). Striving for higher efficiency and asymmetric induction, a range of other organocatalysts were evaluated, among which the diphenylpyrrolidine catalyst F was proved to be the optimal choice by affording the notably improved ee and dr values (entry 7). Furthermore, the effect of reaction concentration and temperature were also examined. It was found that both the yield and ee value were improved when the reaction was conducted at the higher concentration and lower temperature (-10 o C) (entries 8 and 9). However, decreasing the temperature to -40 o C only resulted in a trace amount of the expected product (entry 10). Eventually, a satisfactory result was obtained when an excess amount of enal 13 was used, which delivered 12 in 68% yield and 91% ee (entry 11). The absolute configuration of 12 was assigned as (6R, 8R) by X-ray crystallographic analysis of its derivative 18. Of note, the practical utility of this asymmetric Mukaiyama-Michael reaction was demonstrated by a gram-scale synthesis of 12 with no loss of enantiopurity (Scheme 1).
Having 12 in hands, we then installed the C10-C12 side chain through a Julia-Kocienski olefination. Thus, deprotonation of 19 with KHMDS at -78 o C followed by addition of 12 led to 20 as a single (E)-isomer. Upon irradiation, 20 underwent [2+2]photocycloaddition to give the tricycle 21 as a single diastereoisomer in 94% yield. It should be noted that an identical transformation was reported by Wu and co-workers. 5d Since 21 has been employed as an advanced intermediate in Wood and Brown's synthesis of hippolachnin A, 5b the above work represents a formal synthesis of 1. Nevertheless, given that the biomimetic synthesis of 1 from its precursor 9a had not been achieved yet, we decided to undertake this challenge. For this end, reduction of the lactone 20 was effected with DIBAL-H, which led to the corresponding hemiketal as a mixture of diastereoisomers. Upon treatment with the ylide 22 in refluxing toluene, the hemiketal intermediate underwent a tandem Wittig reaction/oxa-Michael addition to give 23 in 91% yield over two steps. 13 Finally, 23 was converted to 9a through sequential selenylation and oxidative elimination. With 9a in hands, we focused our attention on exploring the key biomimetic [2+2]-photocycloaddition. At first, we attempted to effect the reaction under the conditions employed for the synthesis of 21. Unfortunately, no desired product could be obtained (entry 1, Table 2), which was consistent with the results reported by the Wu' and Perkins' groups. 5d,6e To identify the suitable conditions to effect the transformation, we conducted a comprehensive condition screening. Initially, different light sources were evaluated. It turned out that the reactions irradiated with an UV-lamp (254 or 365nm) or tungsten-lamp failed to give satisfactory outcomes (entries 1-3). Comparably, a promising result was obtained using the highpressure Hg-lamp (500 W), which delivered the desired cycloadduct 1 in a moderate yield (26%). Interestingly, two other products, namely 10-epi-hippolachnin A (24) (7%) and (2E)-hippolachnin A (25) (10%), were also identified in the reaction. Besides the light sources, we also examined various solvents (toluene, MeOH, CH3CN and acetone), among which acetone was proved to be the optimal choice by providing the best yields (1: 39%; 24: 16%; 25: 15%). Furthermore, it was found that the reaction temperature also exerted a notable impact on the distribution of products. As shown, increasing the temperature to 45 o C resulted in an improved overall efficiency, favoring the formation of 1 as predominant product (50%) (entry 6). Comparably, the lower reaction temperature gave an inferior outcome (entry 7), wherein 24 was obtained as major isomer. Finally, we found that 1 and 25 could interchange from each other upon irradiation, generally favoring 1 as major product (dr = 2:1). On the basis of above observations and some related precedents, 5e,8e we assume that the [2+2]-photocycloaddition of 9a proceeds through a stepwise mechanism (Scheme 2). Following the so-called rule of five, the 1,4-diradical intermediate I-1 would be generated first, which may adopt two conformers, as represented by I-1a (path a) and I-1b (path b). Apparently, I-1a is thermodynamically more stable than I-1b, since its bulky ethyl substituent on the C-10 radical center is projected toward

Scheme 2. Mechanistic Rationalization of the [2+2]-Photocycloaddition of 9a
the less hindered convex orientation. As a result, I-1a would be preferentially adopted in the following cyclization step, thus leading to 1 as major product. As to another product 25, it may be generated from 1 directly through the photo-induced C2=C3 double bond isomerization. Alternatively, it could also arise from 9a through sequential C2=C3 double bond isomerization and [2+2]-photocycloaddition (path c). In both cases, the formation of 25 is unfavorable because of the visible steric effect between the carboxylate and ethyl substituent at C-4. Of note, although another isomer 26 could also be generated theoretically (path d), we did not identify any trace amount of this product in practice, presumably attributed to its kinetically and thermodynamically unfavorable nature.

Scheme 3. Total Synthesis of (-)-Gracilioether F
With the total synthesis of 1 secured, we went on to synthesize gracilioether F (4), another popular target in the gracilioether family (Scheme 3). 6 Thus, treatment of aldehyde 12 with the in situ generated ylide MeOCH=PPh3 provided methyl vinyl ether 27 as a mixture of Z/E isomers (1:5) in 70% yield. Without separation, 27 was directly irradiated with highpressure Hg-lamp (500 W) in DCM in the presence of Et3N, which led to the [2+2] adduct 28 as a single diastereoisomer in 83% yield. Of note, the usage of Et3N as additive was crucial  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 for securing the high reaction yield. Otherwise, the severe hydrolysis of the methyl vinyl ether 27 was observed in the reaction. Subsequently, oxidative cleavage of the methyl ether of 28 was effected with RuCl3/NaIO4 under ultrasonic irradiation, 14 which led to the cyclobutanone 29 in 80% yield. Upon treatment with m-CPBA in the presence of NaHCO3, 29 was smoothly converted to gracilioether F (4) and its regioisomer 30 via Baeyer−Villiger oxidation in an excellent combined yield (90%) but with no chemoselectivity (4:30 = 1:1). 15 It should be noted that the Baeyer−Villiger oxidation has also been employed as the key step in Brown's and Carreira's studies, both of which displayed excellent chemoselectivity. 6a, 6b We assumed that the different stereoelectronic effect associated with these substrates may account for the observed results. For the case of 29, the more substituted C4-C10 bond is adjacent to an electron-withdrawing carbonyl group, which exerts a detrimental effect on the desired reaction leading to 4. In order to improve the regioselectivity, we examined various conditions by using different oxidants and additives, but failed to get satisfactory results. Serendipitously, we found that if the transformation from 28 to 29 was conducted with extending reaction time (16 h), a small amount of gracilioether F (4) could also be detected. Importantly, the regioisomer 30 was not identified in the scenario. Inspired by this interesting discovery, an operationally one-pot protocol was developed to effect the oxidative cleavage of the methyl ether and Baeyer-Villiger rearrangement, which delivered the desired regioisomer 4 in 60% overall yield.

Scheme 4. Failed Attempts to Synthesize (+)-Gracilioethers A and H
To further prove the flexibility of our synthetic strategy, the total syntheses of gracilioether A (2) and H (6), two more challenging targets that have not yet achieved in the previous studies, 6 were undertaken by us. Initially, we sought to assemble gracilioether H (6) through the proposed biomimetic [2+2+2]photocycloaddition (Scheme 4A). Both 20 and 9a were evaluated as potential substrates in the reaction. However, we failed to get promising result after extensive tries. In most cases, the severe decomposition of substrates occurred. Thus, we had to search for a new strategy to achieve this goal. Among the many known approaches to access cyclic peroxides, 16 the radical-mediated formal [3+2] cycloaddition of 1-vinylcyclopropane with molecular oxygen caught our attention. 17 We envisioned that the 1,2-dioxane ring of 2 and 6 could be assembled from the vinylcyclobutane precursor 31 through an analogous formal [4+2]-cycloaddition. To test this idea, we first obtained the requisite precursor 31 from the aldehyde 12 through seven steps (for the details, see Scheme 5). However, to our disappointment, when we submitted 31 to the standard conditions (Ph2Se2, AIBN, CH3CN, 0 o C) reported by Feldman, 17b-d we failed to obtain the desired product 32, with most of the starting material recovered (Scheme 4B).

Scheme 5. Total Syntheses of (+)-Gracilioether A and (-)-Gracilioether E
The above failure led us to reconsider our synthetic strategy towards 2 and 6. Recently, a landmark total synthesis of (+)cardamom peroxide, a 1,2-dioxane-containing natural product, was reported by Maimone and co-workers, 18 which employed an interesting hydrogen-atom-transfer (HAT)-triggered double peroxidation reaction to install the two requisite oxygenated functionalities (one peroxide and one hydroxyl) in a single step. 19 Inspired by this seminal work, a seemingly daring design came into our minds: the characteristic 1,2-dioxane ring and the hydroxyl group at C-11 of gracilioether A (2) could also be assembled from the vinylcyclobutane derivative 31 through a HAT-triggered double peroxidation reaction. To validate this idea, we had to obtain the requisite precursor 31 first. For this end, the aldehyde 12 was converted to the 1,3-diene 34 through Horner-Wadsworth-Emmons olefination. Upon irradiation, 34 underwent [2+2]-photocycloaddition smoothly, giving rise to the vinylcyclobutane 35 as a single isomer. Subsequently, a three-step sequence (reduction, acetylation and addition of silyl ketene acetal 36) was used to introduce the carboxylate side chain, 5b which afforded 37 in 87% overall yield. After selenylation and oxidative elimination, 37 was converted to the corresponding -unsaturated ester as a mixture of Z/E isomers (31/38 = 1:1) in 79% combined yield. Of note, the (2E)-isomer (38) could convert to the desired (2Z)-isomer (31) upon treatment with HCl/MeOH, thus improving the overall yield of 31 to 68%.
Having 31 in hands, we then attempted the designed HATtriggered double peroxidation reaction using the identical conditions reported by the Maimone's group (entry 1, Table 3). Encouragingly, although the reaction products appeared complicated, we did identify a small amount of the expected product gracilioether A (2) (5%)! Unexpectedly, another natural product, gracilioether E (3), was also detected in the reaction (7%). Although the efficiency of the transformation was far from ideal, the preliminary result indicated that our new strategy was feasible. To improve the reaction, we further evaluated several  reaction parameters including the metal species, solvent, temperature. Gratifyingly, we found that the Mukaiyama/Isayama hydrosilylperoxidation reaction system [Co(acac)2, PhSiH3, t-BuOOH, iPrOH, DCE) 19b-d displayed superior reactivity, affording 2 and 3 in 25% and 24% yields, respectively (entry 2). In addition, substantial amounts of another product was also isolated in this case, which was determined to be 2c on the basis of extensive spectroscopic study (for details, see Supporting Information). Furthermore, it was found that decreasing the reaction time (6 h) enabled the access of 2 as a major product (39%) (entry 3). Comparably, the longer reaction time (24 h) mainly resulted in 3 in 30% yield (entry 4). The above transformation was noteworthy, since it enables the simultaneous generation of the natural products 2 and 3 in one pot with high step-economy. Mechanistically, we assumed that the reaction should proceed through a hydrogen-atom-transfer (HAT)-triggered cascade process illustrated in Scheme 6. Thus, the transfer of a hydrogen atom from HCo(III) complex to the terminal alkene of 31 affords the radical species I-3, which readily undergoes cyclobutane ring-opening to form the radical intermediate I-4. Subsequently, I-4 can be trapped by dioxygen to give the peroxy radical I-5, which then undergoes 6-exo-trig cyclization to yield the radical species I-6. Next, the second dioxygen insertion takes place to afford the peroxy radical I-7, which can further advance to 2 and 2c through hydrogen abstraction followed by peroxide reduction. As to gracilioether E (3), we assumed that it could be generated from 2 and 2c through a series of intriguing transformations. As shown, both 2 and 2c could undergo Co(II)catalyzed ring-opening reaction to afford the oxygen-centered radical I-8. Subsequently, 1,5-hydrogen atom abstraction takes place to form the carbon radical I-9, which then advances to the ketone 10 with the regeneration of Co(II) species. Once formed, 10 readily undergo a hemiketalization to give the lactol 11, which can further advance to the final product 3 through the oxidative cleavage of the 1,2-diol with the action of Co(II)catalyst and molecular oxygen. Interestingly, both the Co(II)catalyzed Kornblum-DeLaMare-type endoperoxide rearrangement 20 and the Co(II)-catalyzed aerobic oxidative cleavage of the 1,2-diol have been well documented, 21 which supports our mechanistic rationalization. More convincingly, we proved that Scheme 6. Mechanistic Rationalization of the HAT-Induced Oxygenation of Vinylcyclobutane both 2 and 2c could convert to 3 smoothly upon treatment with Co(acac)2 in the presence of molecular oxygen (Scheme 6).