Synthesis-Driven Stereochemical Assignment of Marine Polycyclic Ether Natural Products

Marine polycyclic ether natural products have gained significant interest from the chemical community due to their impressively huge molecular architecture and diverse biological functions. The structure assignment of this class of extraordinarily complex natural products has mainly relied on NMR spectroscopic analysis. However, NMR spectroscopic analysis has its own limitations, including configurational assignment of stereogenic centers within conformationally flexible systems. Chemical shift deviation analysis of synthetic model compounds is a reliable means to assign the relative configuration of “difficult” stereogenic centers. The complete configurational assignment must be ultimately established through total synthesis. The aim of this review is to summarize the indispensable role of organic synthesis in stereochemical assignment of marine polycyclic ethers.


Introduction
Marine polycyclic ethers, produced as secondary metabolites by marine microalgae, mainly dinoflagellates, have gained significant interest from the chemical community because of their impressively huge molecular architecture and diverse biological functions ( Figure 1) [1][2][3]. The majority of this family of natural products were identified as marine toxins responsible for human intoxications and massive fish kills. At the molecular level, many of marine polycyclic ethers are known to interact specifically with ion channels either as agonists, antagonists or partial agonists. Thus, marine polycyclic ethers are pharmacologically useful probes for structural and functional analyses of ion channels [4][5][6][7][8].
In this context, it is obvious that the structure determination of marine polycyclic ethers is of utmost importance to precisely discuss their structure-activity relationship as well as mode of interaction with ion channels.
The first member of this family of natural products to be structurally elucidated was brevetoxin B (1). The structure of brevetoxin B was established by Nakanishi and coworkers through an X-ray crystallographic analysis [9]. Thereafter, a number of marine polycyclic ethers have been isolated and structurally characterized. The common structural motif shared among this family of natural products is the ladder-shaped polycyclic ether skeleton fused in a trans/syn/trans fashion, and the ring size of cyclic ethers ranges from five to nine membered. A plausible biosynthetic mechanism postulated by Nakanishi/Shimizu involves a cascade cyclization of polyepoxides consisting of a single carbon chain to "zip up" the fused polycyclic ether skeleton [10][11][12][13]. Vilotijevic and Jamison demonstrated a polyepoxide cyclization cascade in neutral H 2 O at 70 • C to produce trans-fused polytetrahydropyrans, thereby supporting the Nakanishi's biosynthetic hypothesis [14].
The extraordinarily complex structure of marine polycyclic ethers has been characterized mostly by NMR spectroscopic analysis due to their extremely limited availability from natural sources as well as their physicochemical properties unsuitable for X-ray crystallographic analysis. From the late 1980s, the structure determination of ciguatoxins [15][16][17], gambieric acids [18][19][20], gambierol [21,22], maitotoxin [23][24][25][26][27], and yessotoxins [28][29][30] was accomplished by the Yasumoto group. Yasumoto and coworkers relied heavily on 2D-NMR and in certain cases 3D-NMR analyses for determining the gross structure and the relative configuration of the polycyclic ether skeleton of these compounds. However, NMR spectroscopic analysis has its own limitations, including configurational assignment of stereogenic centers within conformationally flexible systems. Stereocontrolled synthesis enables preparation of a series of diastereomeric compounds of a natural product in question. Comparison of the NMR spectroscopic data of diastereomeric compounds with those of the parent natural product (chemical shift deviation analysis) is a powerful means to elucidate the configuration of difficult-to-assign stereogenic centers [31][32][33][34][35]. However, not all the possible diastereomers need to be prepared; the number of possible diastereomers can be reduced through conformational analysis of the parent natural product. Synthesis and NMR analysis of a single model compound may be sufficient to confirm the configuration of a most likely stereoisomer.
In any case, it should be emphasized that the conformation of a model compound with correct configuration must reproduce that of the corresponding moiety of the parent natural product, and that the complete configurational assignment must be ultimately established through stereocontrolled total synthesis. This review summarizes selected examples of synthesis-driven stereochemical assignment of marine polycyclic ethers to highlight the significance of organic synthesis in structure determination of natural products.

Maitotoxin
Maitotoxin (5) was first isolated from the surgeonfish Ctenochaetus striatus as one of the causative toxins of ciguatera seafood poisoning [36] and was subsequently identified to originate from the dinoflagellate Gambierdiscus toxicus [37]. This natural product is the largest non-biopolymer secondary metabolite known to date and exhibits diverse biological activities that are supposed to be triggered by intracellular Ca 2+ elevation [37][38][39][40][41][42]. The LD 50 value of maitotoxin against mice (ddY strain, 16-20 g body weight) was determined to be 50 ng/kg (intraperitoneal, i.p.) [25]. The gross structure of maitotoxin, which consists of 142 carbon atoms and 32 cyclic ethers ranging from six to eight membered, was determined by chemical degradation of the authentic sample and extensive 2D-and 3D-NMR and negative FAB MS/MS experiments due to extremely heavy overlapping of the 1 H and 13 C NMR signals [23][24][25]. The relative configuration of the fused polycyclic ether domains was assigned by Murata, Yasumoto and coworkers on the basis of NOESY correlations and 3 J H,H values, with the aid of molecular mechanics calculations. However, the relative configuration of the remaining acyclic portions and the absolute configuration of maitotoxin could not be assigned solely by NMR spectroscopic analysis.

Determination of the Absolute Configuration of Maitotoxin by the Tachibana Group and Yasumoto Group
The configurational assignment of the stereogenic centers included in the acyclic portions and the determination of the absolute configuration of maitotoxin were achieved by a collaborative work of the Tachibana group and the Yasumoto group through extensive NMR spectroscopic analysis on natural and 13 C-enriched maitotoxin as well as synthesis of model compounds and comparison with those of the natural product [26,27,[43][44][45][46].

Stereochemical Assignment of the C63-C68 Linkage
The C63-C68 linkage is the acyclic portion flanked with the LM-and NO-rings and contains two stereogenic centers. Detailed conformational analysis on the basis of NOEs and 3 J H,H values led the Tachibana group to synthesize four candidate diastereomers 7-10 ( Figure 2) for this partial structure [43,44]. The synthesis of model compounds 7 and 8 started with methyl glycoside 11, which was prepared from 1,2,5,6-diisopropylidene-α-D-glucose (12) (Figure 3). Allylation of 11 with allyltrimethylsilane in the presence of TMSOTf [47], followed by cleavage of the allyl ether, gave alcohol 13. After a five-step sequence of manipulations, Sharpless asymmetric epoxidation [48] of the resultant allylic alcohol 14 provided epoxy alcohol 15. Oxidation, Wittig methylenation, and desilylation gave rise to vinyl epoxide 16, whose treatment with CSA triggered 6-endo cyclization [49] to afford alcohol 17. This compound served as a common precursor for the LM-and NO-ring fragments 18 and 19. Aldol coupling of 18 and 19 using LDA as a base provided β-hydroxy ketone 20 in 37% yield as the major diastereomer. The configuration of the C64 stereogenic center of 20 was confirmed by derivatization. Evans 1,3-anti reduction [50] of 20 followed by a three-step global deprotection of the protecting groups furnished model compound 7. Meanwhile, Narasaka-Prasad 1,3-syn reduction [51] of 20 and subsequent three-step deprotection led to model compound 8. The stereochemical consequence of the Narasaka-Prasad 1,3-syn reduction of 20 was confirmed by derivatization of the product 1,3-diol to the corresponding acetonide. The remaining model compounds 9 and 10 were similarly prepared starting from ent-21 and 19. Detailed comparison of the 1 H and 13 C NMR spectroscopic data of model compounds 7-10, recorded in CD 3 CN/H 2 O (1:1), with those of the corresponding part of the authentic maitotoxin clearly indicated that 7 represents the relative configuration of the C63-C68 acyclic linkage.

Stereochemical Assignment of the C35-C39 Acyclic Linkage
The C35-C39 acyclic linkage connects the F-and G-rings and contains two stereogenic centers. Based on the NOESY and E.COSY spectra of maitotoxin, the Tachibana group deduced the relative configuration and the conformation of this acyclic moiety as shown in Figure 4. To confirm this assignment, model compound 22 that represents the EF/GH-ring system was synthesized from ester 23, which was prepared from 1,2,5,6-diisopropylideneα-D-glucose (12) in ten steps [45] (Figure 5).  After two-step protecting group manipulations on 23, DIBALH reduction and Wittig olefination gave α,β-unsaturated ester 24. DIBALH reduction of 24 followed by Sharpless asymmetric epoxidation provided epoxy alcohol 25. Oxidation and Wittig olefination led to α,β-unsaturated ester 26, which upon exposure to TBAF and then to Pd(PPh 3 )/Ph 3 P induced 6-endo cyclization [52], affording bicyclic ether 27, after silylation of the cyclization product. Subsequent four-step manipulations gave rise to the EF-ring alkyne 28. The requisite coupling partner, the GH-ring triflate 29, was synthesized from methyl α-Dglucopyranoside by routine chemistry [45]. Coupling of the EF-ring alkyne 28 and the GH-ring triflate 29 afforded coupling product 30 in 70% yield. After desilylation with TBAF, the resultant alkyne was stereospecifically reduced with Red-Al ® to give, after benzylation, (E)-olefin 31. Stereoselective dihydroxylation of 31 under standard catalytic conditions provided diol 32 in 94% yield with approximately 5:1 diastereoselectivity, in accordance with Kishi's empirical stereochemical model [53]. The stereochemical outcome of the dihydroxylation was later confirmed on the basis of J C,H and J H,H values [46,54]. Subsequent four-step manipulations including sulfonylation of the C40 hydroxy group furnished model compound 22. The 1 H and 13 C NMR spectroscopic data of the C35-C39 acyclic linkage of 22, collected in C 5 D 5 N/CD 3 OD (1:1), were in excellent agreement with those of the corresponding part of maitotoxin, thereby confirming the configurational assignment of the C35-C39 acyclic linkage.

Stereochemical Assignment of the C1-C14 Side Chain
The C1-C14 side chain of maitotoxin involves seven stereogenic centers, and 2 7 = 128 stereoisomers are possible for this acyclic portion ( Figure 6). Obviously, however, the synthesis of all the possible stereoisomers is unrealistic. To reduce the number of candidate diastereomers, the Tachibana/Yasumoto group assigned all but one relative configurations of the stereogenic centers within the C1-C14 side chain by means of extensive conformational analysis. Since NOE-based conformational analysis was unsuitable for the conformationally flexible C5-C9 moiety, the configuration of the C5, C7, C8, and C9 positions was deduced based on J-based configurational analysis (JBCA) [54]. The JBCA method is widely applicable to conformationally non-biased acyclic compounds and involves measurement of long-range carbon-proton ( 2,3 J C,H ) and proton-proton ( 3 J H,H ) coupling constants. The C9-C12 moiety, on the other hand, appeared relatively rigid in conformation and the relative configuration between C9 and C12 could be correlated through conformational analysis based on NOEs and 3 J H,H values. The C12-C14 moiety was suggested to exist as an equilibrating mixture of two conformers, and the relative configurations of C12/C13 and C14/C15 were assigned by exploiting the JBCA method. Because the relative configuration between C13 and C14 could not be assigned by NMR spectroscopic analysis, the Tachibana/Yasumoto group determined to synthesize two diastereomeric model compounds 33 and 34 to compare their NMR spectroscopic data with those of the corresponding part of maitotoxin.
The synthesis of the left half of compound 33 started with methyl (R)-3-hydroxybutyrate (35), as shown in Figure 7. A four-step sequence of standard manipulations gave allylic alcohol 36. Sharpless asymmetric epoxidation delivered epoxy alcohol 37, which upon treatment with AlMe 3 [55,56] underwent regioselective and stereospecific epoxide opening to afford 1,2-diol 38 with desired configuration at the C7 and C8 positions. Subsequent three-step protecting group manipulations gave alcohol 39, which was oxidized under Swern conditions [57] to provide aldehyde 40. The enantiomer of 40, i.e., ent-40, was prepared in a similar fashion starting from ent-35.  The synthesis of the right half of compounds 33 and 34 commenced with Swern oxidation of alcohol 41, readily prepared from tri-O-acetyl-D-glucal in 10 steps, followed by Wittig reaction of the derived aldehyde ( Figure 8). DIBALH reduction of the resultant α,β-unsaturated ester gave allylic alcohol 42. Sharpless asymmetric epoxidation of 42 delivered epoxy alcohol 43. Regio-and stereospecific methylation of 43 was followed by cleavage of the resultant 1,2-diol with NaIO 4 .
The subsequent Horner-Wadsworth-Emmons reaction and DIBALH reduction led to allylic alcohol 44 with correct configuration at the C14 stereogenic center. Sharpless asymmetric epoxidation of 44 using (−)-DET/Ti(Oi-Pr) 4 as a chiral catalyst afforded an epoxy alcohol, whose exposure to Me 2 CuLi opened the epoxide in a regioselective/stereospecific manner [58] to provide 1,3-diol 45 after periodate workup, and thereby successfully installed the requisite C12 and C13 stereogenic centers. After three-step protecting group manipulations, the derived alcohol was oxidized and then iodoolefinated [59] to furnish iodoolefin 46. The diastereomeric iodoolefin 48 was synthesized from allylic alcohol 44 in the same way as 46, except for the Sharpless asymmetric epoxidation, where (+)-DET/Ti(Oi-Pr) 4 was used as a chiral catalyst to create the C12 and C13 stereogenic centers. The Nozaki-Hiyama-Kishi reaction [60,61] of aldehyde 40 and iodoolefin 46 afforded allylic alcohol 49 in 52% yield ( Figure 9). After oxidation, the resultant α,β-unsaturated ketone was reduced with NaHTe [62] to give ketone 50. Chelate-controlled reduction of 50 with Zn(BH 4 ) 2 [63] provided an alcohol with desired configuration at C9. Sulfonylation and hydrogenolysis of the benzyl ethers afforded compound 33. Following a similar synthetic strategy, compound 34 was synthesized from ent-40 and iodoolefin 48. Comparison of the 1 H and 13 C NMR spectroscopic data of 33 and 34 with those of the corresponding moiety of natural maitotoxin showed that 33 was in good accordance with the natural product with respect to the 13 C NMR chemical shift data. The 3 J H,H values of the C7-C9 and C12-C15 portions of 33 were also in good agreement with those of natural maitotoxin. From these results, the relative configuration of the C1-C14 side chain was successfully determined.

Stereochemical Assignment of the C135-C142 Side Chain and Absolute Configuration of Maitotoxin
Among the four stereogenic centers included in the C135-C142 side chain of maitotoxin, the relative configuration between C136 and C138 positions could be unambiguously assigned through a J-based configurational analysis ( Figure 10). The relative configuration of the C134, C135, and C136 stereotriad was also assignable by means of a JBCA. However, the NOEs and 3 J H,H values observed for this portion suggested the presence of two inter-converting conformers. To exclude the ambiguity in the assignment made by the JBCA method, the Tachibana/Yasumoto group synthesized model compound 51. The synthesis of model compound 51 started from known alcohol 52, which was available in five steps from tri-O-acetyl-D-glucal [64] (Figure 11). Oxidation followed by Wittig olefination gave α,β-unsaturated ester 53. Sharpless asymmetric dihydroxylation [65] of 53 with AD-mix β led to diol 54 with 6:1 diastereomer ratio. After protection of 54 as its acetonide, the desired diastereomer 55 was isolated in a stereochemically pure form by column chromatography using silica gel. DIBALH reduction and Wittig reaction delivered α,β-unsaturated ester 56, which was converted to alcohol 57 via a three-step sequence of manipulations. Deoxygenation of 57 through tin hydride reduction of a thionocarbonate [66], followed by removal of the acetonide under acidic conditions, afforded model compound 51. The 2 J C,H and 3 J H,H values of the C134-C136 portion of 51 corresponded with those of natural maitotoxin, thereby confirming the configurational assignment by the JBCA method. The configurational assignment of the C139 stereogenic center was not possible by the JBCA method because the broadened signals of H-138 and H-139 precluded the measurement of 3 J C,H and 3 J H,H values [27]. The Tachibana/Yasumoto group synthesized four stereoisomeric model compounds 57, ent-57, 58, and ent-58 to establish the relative configuration between C138 and C139 positions and at the same time the absolute configuration by comparing these stereoisomeric model compounds with the authentic degradation material ( Figure 12).  The synthesis of model compound 58 started from diol 65, which was prepared in two steps from di-(+)-menthyl fumarate via a Diels-Alder reaction with 1,3-butadine and a subsequent reduction using LiAlH 4 ( Figure 14) [67,68]. Iodination of 65 followed by dihydroxylation gave 1,2-diol 66. Deiodination under tin hydride conditions, periodate cleavage, and NaBH 4 reduction delivered diol 67. A four-step sequence of desymmetrization furnished model compound 58. The independent synthesis of 58 confirmed the relative configuration of 57. The synthesis of ent-58 was carried out in a similar manner as that described for 58. Chiral GC/MS analysis of four stereoisomeric compounds 57, ent-57, 58, and ent-58 with authentic reference prepared via degradation of natural maitotoxin (NaIO 4 oxidation to cleave the C135/C136 vic-diol moiety, followed by NaBH 4 reduction of the derived aldehydes) clearly demonstrated that the retention time and mass spectrum of 57 matched those of the authentic degradation sample. Accordingly, the absolute configuration of maitotoxin was determined to be shown by structure 5 [27].

Determination of the Relative Configuration of Maitotoxin by the Kishi Group
The Kishi group also elucidated, independently from the Tachibana/Yasumoto group, the complete relative configuration of maitotoxin [69]. Kishi et al. assigned the four acyclic portions of maitotoxin, i.e., the C1-C15, C35-C39, C63-C68, and C134-C142 portions, on the basis of stereocontrolled synthesis and NMR spectroscopic analysis of model compounds.

Stereochemical Assignment of the C1-C15 Side Chain
As described above, 128 stereoisomers are possible for the C1-C15 side chain. To narrow down the number of candidate diastereomers, Kishi divided this portion into the C1-C11 and C11-C15 portions and assigned the relative configuration of these two portions individually ( Figure 15). This approach was based on their previous work on the stereochemical assignment of AAL toxin T A and fumonisin B 2 backbone [70][71][72], and subsequently extended to the concept of "universal NMR database" [73][74][75]. In the present case, two methylene groups (C10 and C11) separate two stereoclusters, i.e., C5-C9 and C12-C15, and the steric and/or stereoelectronic interactions between these two stereoclusters were assumed to be very small. The Kishi group synthesized eight possible diastereomers for the C5-C9 portion and found 68 to be in accordance with the corresponding portion of natural maitotoxin with respect to 1 H and 13 C NMR data. The relative configuration of the C11-C15 portion was similarly assigned as shown by structure 69 through the synthesis of eight possible diastereomers. Based on these results, Kishi et al. determined to synthesize two candidate diastereomers 70 and 71 for the assignment of the relative configuration of the C1-C15 side chain. The synthesis of 70 was achieved in a convergent manner through a coupling of aldehyde 72 and dibromoolefin 73 (Figures 16 and 17). The synthesis of 72 started with Roush asymmetric crotylation [76] of (p-methoxybenzyloxy)acetaldehyde (74) to give homoallylic alcohol 75, which was converted to aldehyde 76 via a three-step sequence of manipulations ( Figure 16). Addition of vinyllithium to 76 provided allylic alcohol 77 as an approximately 1:1 mixture of diastereomers, which were separable by column chromatography using silica gel. A subsequent five-step sequence of standard manipulations led to aldehyde 72. Meanwhile, the synthesis of 73 started from known aldehyde 78, which was available from D-ribose [77]. A five-step sequence of manipulations led to aldehyde 79, which was reacted with Roush's (Z)-(R,R)-crotylboronate to give homoallylic alcohol 80 exclusively. Swern oxidation, removal of the PMB group, and reductive etherification afforded tetrahydropyran 81. After cleavage of the double bond, the resultant aldehyde was crotylated using crotyl bromide and zinc dust to give homoallylic alcohol 82. This compound was elaborated to dibromoolefin 73 through a six-step sequence of manipulations.  Coupling of aldehyde 72 and dibromoolefin 73 was efficiently achieved by treatment of 73 with n-BuLi, followed by addition of the resultant lithium acetylide to 72, to provide propargylic alcohol 83 in 97% yield, albeit with moderate diastereoselectivity (Figure 17). The desired isomer was isolated by silica gel column chromatography in 44% yield. After four steps of manipulations including the reduction of the alkyne and selective liberation/oxidation of the primary alcohol, addition of ((E)-4-t-butyldimethylsilyloxy)-2-buten-2-yllithium to the derived aldehyde 84, followed by oxidation of the resultant allylic alcohol and deprotection of the PMB group, delivered ketone 85. After masking the hydroxy group of 85 as its EE ether, Wittig methylenation of the carbonyl group and ensuing two-step protecting group manipulations gave alcohol 86.

Stereochemical Assignment of the C63-C68 Acyclic Portion
The Kishi group synthesized all eight possible diastereomers for stereochemical assignment of the C63-C68 acyclic portion containing four stereogenic centers ( Figure 20).  (Figures 21-23). Olefin 111 was synthesized from C-glycoside 113, the latter being available from 1,6-anhydro-D-glucose in five steps ( Figure 21). Ozonolysis of the double bond followed by Nozaki-Hiyama-Kishi coupling with methyl (E)-βiodoacrylate delivered alcohols 114a and 114b as an approximately 1:1 mixture of diastereomers, which were separable by column chromatography. The undesired diastereomer could be transformed into the desired one via a Mitsunobu reaction [80]/methanolysis sequence. After acetylation and desilylation, the resultant triol was treated with DBU to induce intramolecular oxa-Michael addition [81,82], giving rise to bicyclic ether 115, after resilylation of the resulting two hydroxy groups. A nine-step sequence of manipulations led to olefin 111. The enantiomer of 111, i.e., ent-111, was prepared from 1,6-anhydro-L-glucose in the same manner.   Meanwhile, the synthesis of aldoxime 112 started from C-glycoside 116 ( Figure 22). After five steps of standard manipulations, the derived aldehyde 117 was reacted with methyl (E)-β-iodoacrylate under Nozaki-Hiyama-Kishi conditions to deliver alcohols 118a and 118b in 80% yield with 2.5:1 diastereoselectivity. These diastereomers were separated by column chromatography on silica gel, and the undesired minor diastereomer 118b was converted to the desired 118a via a Mitsunobu reaction/methanolysis sequence. Acetylation followed by desilylation using buffered TBAF gave a diol, which upon exposure to DBU underwent intramolecular oxa-Michael addition to provide bicyclic ether 119. This compound was manipulated over 13 steps to afford aldoxime 112.
Coupling of olefin 111 and aldoxime 112 provided 1,3-dipolar addition product 120a,b in 50-60% yield with 3:2 diastereoselectivity (Figure 23). These diastereomers were separated by silica gel column chromatography. Treatment of the major diastereomer 120a with Mo(CO) 6 in aqueous CH 3 CN [83] followed by reduction of the derived β-hydroxy ketone with NaBH(OAc) 3 in AcOH/CH 3 CN gave 1,3-diols as an approximately 4:1 mixture of diastereomers at C66. These diastereomers were separated by silica gel column chromatography, and individually subjected to debenzylation to afford model compounds 103 and 104. The minor diastereomer 120b of the 1,3-dipolar addition product was similarly processed to provide model compounds 105 and 106. The remainder of model compounds, i.e., 107-110, was synthesized via a coupling of olefin ent-111 and aldoxime 112.
Among the eight diastereomeric model compounds, the 1 H and 13 C NMR spectroscopic data of 105 matched excellently those of the corresponding part of natural maitotoxin [69]. The relative configuration of 105 was secured by derivatization and NMR spectroscopic analysis of late-stage intermediates derived from the 1,3-dipolar adduct 120b.

Stereochemical Assignment of the C134-C142 Side Chain
The relative configurational assignment of the C134-C142 side chain of maitotoxin required the synthesis of 16 diastereomeric model compounds 121-136 ( Figure 24). The Kishi group synthesized these diastereomers from four diastereomeric alkynes 137-140 by means of dihydroxylation of (E)-and (Z)-olefins. The alkynes 137-140 in turn would be obtainable through coupling of two racemic dibromoolefins rac-141/rac-142 with optically active lactone 143, followed by separation of diastereomers. The synthesis of dibromoolefin rac-141 started with LiAlH 4 reduction of Diels-Alder cycloaddition product rac-144, followed by tosylation/reduction to remove superfluous hydroxy groups, giving cis-4,5-dimethylcyclohexene ( Figure 25). Subsequent ozonolysis and NaBH 4 -workup delivered diol rac-145. Monosilylation followed by an oxidation/dibromoolefination sequence provided dibromoolefin rac-141. In a similar fashion, dibromoolefin rac-142 was prepared from Diels-Alder cycloaddition product rac-146.  Coupling of dibromoolefin rac-141 and lactone 143, followed by reductive etherification of the resultant product, delivered a diastereomeric mixture of alkynes 137 and 138 ( Figure 27). Without separation, this mixture was transformed to a mixture of (E)-olefins 150/151 over three steps. Dihydroxylation of 150/151 provided four diastereomeric threo diols 152-155, which were separable by silica gel column chromatography. A series of standard manipulations furnished model compounds 121-124. Meanwhile, semi-reduction of 137/138 gave a mixture of (Z)-olefins 156/157, whose dihydroxylation delivered four diastereomeric erythro diols 158-161 ( Figure 28). These diastereomers were chromatographically separable, and subjected individually to a five-step sequence of manipulations to afford model compounds 125-128. In a similar manner, model compounds 129-136 were synthesized from dibromoolefin rac-142 and lactone 143 ( Figure 29). The 1 H and 13 C NMR spectroscopic data of model compounds 121-136 were compared with those of the corresponding portion of natural maitotoxin to show that model compound 121 was the best match. The configuration of 121 was confirmed by its independent synthesis from (S)-citronellal as a source of the C139 stereogenic center.   Overall, the Kishi group has successfully established the relative configuration of maitotoxin [69]. Moreover, they proposed the absolute configuration of this natural product on the basis of biosynthetic considerations, which turned out to be same as that determined experimentally by the Tachibana/Yasumoto group [27]. In the event, the Kishi group determined the absolute configuration of maitotoxin independently (see the Note Added in Proof of [69]).

Proposal of the Alternative Configuration of the JK-Ring Juncture by Gallimore and Spencer, and Its Disproof by Confirmation of the Originally Assigned Configuration of the JK-Ring Juncture by the Nicolaou Group
The biosynthesis of marine polycyclic ether natural products has not been fully understood. Nakanishi and Shimizu independently proposed a biosynthetic mechanism of brevetoxin A [12,13], in which a single carbon-chain polyene is epoxidized to an octaepoxide precursor and then a cascade of polyepoxide cyclization takes place to "zip up" the polycyclic ether core ( Figure 30). The strict trans/syn/trans stereochemical regularity shared among the family of marine polycyclic ether natural products suggests that a common mechanism is operating for their biosynthesis and that a polyene should be epoxidized from the same face of the molecule by a single monooxygenase enzyme. In 2006, Gallimore and Spencer pointed out that the configuration of the JK-ring juncture (C51/C52) of maitotoxin is an exceptional example that deviates from the common stereochemical regularity of marine polycyclic ethers ( Figure 31) [84]. Assuming the polyepoxide cyclization cascade, the JK-ring juncture should arise from an (S,S)-transepoxide, whereas the other ring junctures are derived from (R,R)-trans-epoxides. It appears odd that a monooxygenase epoxidizes all but one double bond from the same face of a polyene and another monooxygenase epoxidizes specifically the double bond in question from the opposite face. To make the matter complicated, the configurational assignment of the JK-ring juncture had been noted as being highly challenging due to heavy overlap of NMR signals [85]. Accordingly, Gallimore and Spencer proposed that the configuration of the JK-ring juncture should be opposite to that determined in previous works. In 2007, Nicolaou and Frederick suggested on the basis of DFT chemical shift calculations that the originally proposed configuration of the JK-ring juncture is likely correct, and proposed an alternative biosynthetic route that does not violate the hypothesis of Gallimore and Spencer [86]. Nicolaou and coworkers reported the synthesis of a GHIJK ring model compound in 2007 [87] and the GHIJKLMNO ring domain in 2008 [88] to provide a strong evidence that supports the originally assigned configuration.
Luche reduction of 177 followed by acylation, and subsequent dihydroxylation provided diol 178. After a two-stage protecting group manipulations to differentiate the hydroxy groups, allylation with allyltrimethylsilane/BF 3 •OEt 2 afforded olefin 179. After replacement of the acetyl group with a TBS group, migration of the double bond followed by ozonolysis gave aldehyde 180, which was alkynylated with lithiated cyclohexylacetylene and then oxidized to deliver α,β-unsaturated ketone 181. Desilylation and subsequent activation of the alkyne with a catalytic amount of AgOTf (CH 2 Cl 2 , 40 • C) provided dihydropyranone 182. A three-step functionalization of the dihydropyranone ring, including Luche reduction, hydroboration, and bis-silylation, led to bis-silyl ether 183. After reducing the pivaloates, oxidative lactonization of the derived diol, followed by enolization/triflation, gave rise to the IJK-ring enol triflate 164.
Hydroboration of the G-ring exocyclic enol ether 163 with 9-BBN-H and coupling of the resultant alkylborane with the IJK-ring enol triflate 164 under the influence of the Pd(OAc) 2 /S-Phos [97] catalyst and KHCO 3 as a base (THF/H 2 O, room temperature) afforded coupling product 184 in 78% yield ( Figure 34). Hydroboration with BH 3 •THF, Dess-Martin oxidation, [98] and subsequent desilylation/methyl acetalization delivered methyl acetal 185, which was reduced with Et 3 SiH/TMSOTf and then debenzylated to furnish the GHIJK ring model compound 162. The 13 C NMR spectroscopic data of the C42-C53 moiety of 162 were in excellent agreement with those of the corresponding portion of natural maitotoxin [87].

Brevenal
Bourdelais, Baden and coworkers isolated brevenal from a laboratory culture of the Florida red tide-forming dinoflagellate Karenia brevis [99]. This natural product competitively inhibits the binding of tritiated dihydrobrevetoxin B ([ 3 H]PbTx-3) to site 5 of voltage-gated sodium ion channels in a dose-dependent fashion without neurotoxicity and acts as a naturally occurring brevetoxin antagonist in vivo. Brevenal is also a potent inhibitor of catecholamine secretion induced by ciguatoxin without affecting other secretagogue activities, including nicotine-or barium-induced catecholamine secretion [100,101]. Ciguatoxin is the principal causative toxin of ciguatera sea food poisoning and exhibits potent neurotoxicity by binding to site 5 of voltage-gated sodium ion channels. Accordingly, it has been suggested that brevenal could be potentially useful for the treatment of ciguatera. Moreover, brevenal improves tracheal mucus clearance activity in an animal model of asthma, suggesting that it might be useful as a candidate in the drug development toward cystic fibrosis therapy [102]. The Baden group determined the gross structure of brevenal through extensive 2D-NMR analyses, and assigned the relative configuration on the basis of NOESY correlations. However, the absolute configuration of this natural product remained unassigned. The structure of brevenal is reminiscent of that of hemibrevetoxin B, a tetracyclic ether metabolite of K. brevis [103]. According to the Nakanishi hypothesis [10], these compounds should be biosynthetically generated from the respective polyepoxides via a cascade ring opening, as shown in Figure 35. The configuration of the C26 stereogenic center of the proposed structure 186 of brevenal appeared unusual on the basis of the biosynthetic consideration. Actually, the Sasaki group later revised the configuration of the C26 stereogenic center through total synthesis [104][105][106][107]. The first total synthesis of the proposed and correct structures of brevenal was accomplished by Sasaki and coworkers [104][105][106][107]. As shown in Figure 36, the Sasaki group built up the pentacyclic polyether core of 186 through a Suzuki-Miyaura coupling [89][90][91][92] of the AB-ring enol phosphate 188 and an alkylborane derived from the DE-ring exocyclic enol ether 189, followed by a stereoselective methylation of thioacetal 190. The unsaturated side chains at both ends of the polycyclic ether skeleton were introduced by means of a Stille reaction [108] and a Wittig olefination. The synthesis of the AB-ring enol phosphate 188 started with Evans syn-aldol reaction [109] of aldehyde 191 and N-propionyl (R)-4-benzyl-2-oxazolidinone (192), and subsequent reductive removal of the chiral auxiliary [110] provided 1,3-diol 193 (Figure 37). The diol 193 was elaborated to allylic alcohol 194 in seven steps. Sharpless asymmetric epoxidation of 194 using (+)-DET as a chiral ligand delivered epoxy alcohol 195. After oxidation and Wittig methylenation, the resultant vinyl epoxide 196 was exposed to DDQ to induce cleavage of the PMB ether and concomitant 6-endo cyclization, leading to tetrahydropyran 197, after protection of the resultant alcohol as its TES ether (89%, two steps). Additional six steps, including Yamaguchi lactonization [111] to construct the seven-membered ring, delivered the AB-ring enol phosphate 188. The synthesis of the DE-ring exocyclic enol ether 189 started from known alcohol 198 [112] that corresponds to the D-ring ( Figure 38). The alcohol 198 was manipulated over 11 steps to aldehyde 199. Exposure of 199 to SmI 2 (MeOH, THF, room temperature) under Nakata conditions [113,114] resulted in reductive cyclization to form the E-ring seven-membered ether, affording lactone 200 after acid treatment. The lactone 200 was transformed to ketone 201 via a three-step sequence of manipulations, at which point the relative configuration of the C22/C27 stereogenic centers was confirmed by an NOE. Introduction of the C26 methyl group was most reliably achieved by treatment of 201 with MeLi (THF, −78 • C to room temperature), giving rise to tertiary alcohol 202 in 97% yield with 10:1 diastereoselectivity. The configuration of the newly generated C26 stereogenic center was determined by observing an NOE enhancement between the C26 methyl group and the C27 oxymethine proton. Subsequent eight-step manipulations afforded the DE-ring exocyclic enol ether 189. Suzuki-Miyaura coupling of an alkylborane, derived from the DE-ring exocyclic enol ether 189 via hydroboration with 9-BBN-H, and the AB-ring enol phosphate 188 proceeded uneventfully under the influence of aqueous Cs 2 CO 3 and Pd(PPh 3 ) 4 in THF/DMF at 50 • C to deliver endocyclic enol ether 203 (Figure 39). Hydroboration using BH 3 •SMe 2 and alkaline oxidative workup provided alcohol 204 in 84% yield (two steps) as a single diastereomer (dr > 20:1). Oxidation of 204 under Ley's conditions [115] led to ketone 205 in 98% yield, at which point the configuration of the C16 and C18 stereogenic centers was confirmed by NOE correlations as shown. The C14 hydroxy group was introduced at this stage via a diastereoselecive dihydroxylation of an enol silyl ether derived from 205, giving rise to α-hydroxy ketone 206 (87%, two steps) with greater than 20:1 diastereoselectivity. Stereoselective reduction of 206 with DIBALH (THF, −78 • C) delivered cis-diol 207 in 76% yield, along with its C15 epimer (not shown) in 7% yield (starting material was recovered in 12% yield). The configuration of the C14 and C15 stereogenic centers was confirmed by NOE correlations observed for the corresponding cyclopentylidene derivative. After a three-step sequence of manipulations, the derived ketone 208 was exposed to EtSH/Zn(OTf) 2 [116] to promote cleavage of the TES ethers and spontaneous thioacetal formation, affording thioacetal 190 in 79% yield. The construction of the C-ring was completed by silylation of the hydroxy group of 190 and subsequent mCPBA oxidation and in situ methylation with AlMe 3 (CH 2 Cl 2 , −78 to 0 • C) [117] to furnish the pentacyclic ether 209 in 92% yield as a single stereoisomer (dr > 20:1). At this point, the configuration of 209 was further corroborated by NOE experiments and a 3 J H,H value. Finally, the side chains at both ends of the pentacyclic ether framework were attached to complete the total synthesis ( Figure 40). Thus, the pentacyclic ether 209 was converted to alkyne 210 in eight steps. Stannylcupuration of 210 with (Me 2 PhSi) 2 Cu(CN)Li 2 (THF, −78 to 0 • C) [118] delivered vinylsilane 211 with 9:1 regioselectivity, and ensuing iododesilylation with NIS [119] provided vinyl iodide 212 in 99% yield with E/Z 6:1 selectivity. Stille reaction of 212 with vinylstannane 213 under the influence of Pd 2 (dba) 3 /Ph 3 As and CuTC [120] in DMSO/THF at room temperature afforded conjugated diene 214 in 63% yield after removal of minor isomers by flash column chromatography using silica gel. The configuration of the conjugated diene moiety was confirmed by NOE experiments. Subsequent seven-step manipulations, including a Wittig olefination/selenoxide elimination to install the right-hand side chain according to the procedure of Nicolaou and coworkers [121], furnished the proposed structure 186 of brevenal [104]. However, the 1 H and 13 C NMR spectra of synthetic 186 do not match those of the authentic material. The COSY, HSQC, and HMBC correlations of synthetic 186 reproduced those of the natural product, indicating that the gross structure of 186 was correctly assigned. Upon chemical shift deviation analysis, significant inconsistencies were found in around the C26 tertiary alcohol ( Figure 41). Moreover, intense NOESY correlations were observed between C26 methyl/C27 methine and C26 methyl/C28 methylene protons of synthetic 186, whereas no such NOESY correlations were observed for natural brevenal. Thus, it was suggested that the configuration of the C26 stereogenic center might have been erroneously assigned. Importantly, this assumption was in accordance with the biosynthetic consideration described above. Eventually, the Sasaki group embarked on the total synthesis of the revised structure 187 of brevenal, as shown in Figure 42. The DE-ring exocyclic enol ether 215 (26-epi-189) was synthesized from olefin 216, the latter being available from 198. A three-step sequence of manipulations gave ketone 217. SmI 2 -mediated reductive cyclization of 217 under Nakata conditions provided a mixture of lactone 218 (57%) and hydroxy ester 219 (37%), which were separable by flash column chromatography using silica gel. These products were individually reduced with LiAlH 4 to give diol 220. The configuration of the newly formed E-ring was confirmed by NOE experiments on lactone 218 as shown. Subsequent 11-step manipulations afforded the DE-ring exocyclic enol ether 215. Assembly of the pentacyclic ether skeleton and subsequent introduction of the unsaturated side chains were achieved in the same manner as those described for the proposed structure 186 to complete the total synthesis of the revised structure 187. As anticipated, the 1 H and 13 C NMR spectra of synthetic 187 were fully consistent with those of the natural product.

Gambieric Acids
Nagai, Yasumoto and coworkers isolated gambieric acid A and its congeners from a cultured medium of the ciguatera causative dinoflagellate Gambierdiscus toxicus as highly potent antifungal substances [18,19]. The structure of gambieric acids consists of a nonacyclic polyether skeleton arranged with an isolated tetrahydrofuran ring. Although the structural characteristics of gambieric acids are similar to those of polycyclic ether neurotoxins produced by G. toxicus, gambieric acid A shows no toxicity in mice at a dose of 1 mg/kg (i.p.), and only weakly inhibits binding of tritiated dihydrobrevetoxin ([ 3 H]PbTx-3) to site 5 of voltage-gated sodium ion channels [122]. More interestingly, gambieric acid A exhibits potent antifungal activity against Aspergillus niger (10 ng/disk, >2000-fold potent than amphotericin B), and it is also suggested to be an endogenous growth regulator of G. toxicus [123].
The gross structure and relative configuration of the polycyclic ether skeleton of gambieric acids were determined on the basis of extensive 2D-NMR analyses. The complete configurational assignment of gambieric acid B (structure 222, Figure 43) was later reported by Satake, Yasumoto and coworkers [20], which entailed degradation experiments, application of chiral anisotropic reagents, chiral HPLC analysis, and conformational analysis based on J values and NOE and HMBC correlations. The structures of other members of gambieric acids were assigned accordingly, as shown in Figure 43. During the course of synthetic studies toward gambieric acids, Fuwa, Sasaki and coworkers found that the 1 H and 13 C NMR chemical shifts of an A/B-ring model compound 225 did not match those of the corresponding moiety of natural gambieric acid A ( Figure 44) [124][125][126]. Specifically, the C8-C11 moiety of 225 showed significantly deviated chemical shift values, bringing the relative configuration of C7/C9 and C9/C11 into question. Because the C7, C9, and C11 stereogenic centers were configurationally correlated with each other on the basis of conformational analysis on natural gambieric acid B, Fuwa et al. embarked on the synthesis of its four diastereomeric model compounds 226-229 through a Suzuki-Miyaura coupling of iodoolefins 230/231 with alkylborates 232/ent-232, and a diastereoselective bromoetherification to craft the isolated A-ring tetrahydrofuran ( Figure 45). The modular synthetic approach was crucial for efficient and stereocontrolled synthesis of all the requisite diastereomers.  The synthesis of iodoolefins 230/231 started with known alcohol 233 [127] (Figure 46). A five-step sequence of standard manipulations led to methyl ketone 234, which was exposed to SmI 2 under Nakata conditions to deliver lactone 235 in 74% yield as a single diastereomer (dr > 20:1). The configuration of 235 was confirmed by NOE enhancements as shown. After a four-step sequence of manipulations, Sharpless asymmetric epoxidation of the derived allylic alcohol 236 using (+)-DET as a chiral ligand provided epoxy alcohol 237 in 83% yield with greater than 20:1 diastereoselectivity. Chlorination of 237 followed by treatment with excess LDA [128] gave propargylic alcohol 238. Iodination of the terminal alkyne and diimide reduction [129] afforded iodoolefin 230. The configuration of the C9 stereogenic center created through the Sharpless asymmetric epoxidation was confirmed by a modified Mosher analysis [130] on a hydrogenated derivative of 238. The C9 epimer of 230, i.e., 231, was easily available from allylic alcohol 236 by Sharpless asymmetric epoxidation using (−)-DET as a chiral ligand and following the same four-step sequence as described for 230. The precursor of alkylborate 232, i.e., iodide 240, was prepared from known aldehyde 241 [131] over five steps including Evans syn-aldol reaction ( Figure 47). According to the Marshall procedure [132], iodide 240 was lithiated with t-BuLi and trapped with B-MeO-9-BBN to generate alkylborate 232. Without isolation, 232 was coupled with iodoolefin 230 under the influence of aqueous Cs 2 CO 3 and PdCl 2 (dppf)•CH 2 Cl 2 /Ph 3 As in THF/DMF at 50 • C to deliver allylic alcohol 243 in 75% yield. After two-step protecting group manipulations, diastereoselective bromoetherification of the resultant alcohol 244 (NBS, CH 3 CN, room temperature) and subsequent tin hydride reduction (Bu 3 SnH, AIBN, toluene, 110 • C) afforded 2,5-trans-substituted tetrahydrofuran 245 in 69% yield (two steps) as a single diastereomer (dr > 20:1). The configuration of the C4, C5, and C7 stereogenic centers was confirmed by NOE experiments as shown. This compound was elaborated to model compound 226 over seven steps of standard manipulations. Additional three diastereomers 227-229 were synthesized from 230/231 and 240/ent-240 in much the same way as described for 226.
Chemical shift deviation analysis of four diastereomeric model compounds 226-229 compared with natural gambieric acid B clearly indicated that the 1 H and 13 C NMR chemical shifts of 228 were in close agreement with those of the corresponding moiety of the natural product ( Figure 48). Moreover, the conformational analysis of 228 on the basis of J values and NOESY and HMBC correlations in pyridine-d 5 suggested that the conformation of 228 reproduced faithfully that of the A/B-ring moiety of natural gambieric acid B. Because the configuration of the C9 stereogenic center of 228 is opposite to that of the natural product, it was concluded that the C9/C11 relative configuration had been misassigned in the proposed structure 222. This resulted in a configurational reassignment of all the stereogenic centers embedded in the nonacyclic polyether skeleton, as shown in Figure 49. This conclusion was further supported by the synthesis/NMR analysis of an A/B-ring model compound of gambieric acid A and an A/BC-ring model compound of gambieric acid B [126]. Moreover, the revised structure of gambieric acid A was ultimately established through the first total synthesis by Fuwa, Sasaki and coworkers [133,134].

Conclusions
This review summarized the configurational assignment of marine polycyclic ether natural products, maitotoxin, brevenal, and gambieric acids. The maitotoxin case is an illustrative example which demonstrates the power of synthesis-driven configurational assignment of stereochemically complex natural products. At the same time, this case indicates that detailed conformational analysis by means of advanced NMR spectroscopic techniques may be helpful in narrowing down the number of possible diastereomers to be synthesized. The gambieric acid case underlines the significance of the modularity [135] of synthetic planning, which enables expedient access to a set of requisite stereoisomers for determination of the relative configuration in question. These are important points that should be carefully considered in order to save time, costs, and effort in synthesis.
Remarkable advances in NMR spectroscopy in recent decades enable the determination of the gross structure and also the relative configuration of conformationally rigid skeleton of super-carbon-chain polycyclic ethers. Nonetheless, configurational assignment of acyclic portions and remotely isolated stereogenic centers is still a daunting task. The JBCA method is effective for the configurational assignment of non-biased acyclic systems, but its application to more or less biased acyclic systems appears to need special care [136][137][138]. Genome sequence information of the producer organisms may be useful for assigning the configuration of natural products on the basis of biosynthetic predictions, although the biosynthetic genes are difficult to obtain from marine dinoflagellates. Molecular mechanics/DFT calculations may help in assigning those difficult stereogenic centers through the prediction of chemical shift values [139,140]. A recent study on the structure elucidation of gambierone, a novel polycyclic ether metabolite from the dinoflagellate Gambierdiscus belizeanus, took advantage of NMR chemical shift calculations at the B3LYP/6-31G(d)//B3LYP/STO-3G level of theory, and assigned, on the basis of DP4 probability analysis [141], the relative configuration of three stereogenic centers within acyclic portions [142]. It should be emphasized, however, that the configurational assignment made by means of computational and/or bioinformatic approaches still needs to be confirmed through synthesis/NMR analysis of suitably designed model compounds and finally be established in an unambiguous manner by total synthesis. Significant challenges remaining in this area are the complete stereochemical assignments of extremely large polycyclic ethers, brevisulcenal-F [143] and prymnesins [144,145].

Conflicts of Interest:
The author declares no conflict of interest.