Epimeric Face-Selective Oxidations and Diastereodivergent Transannular Oxonium Ion Formation Fragmentations: Computational Modeling and Total Syntheses of 12-Epoxyobtusallene IV, 12-Epoxyobtusallene II, Obtusallene X, Marilzabicycloallene C, and Marilzabicycloallene D.

The total syntheses of 12-epoxyobtusallene IV, 12-epoxyobtusallene II, obtusallene X, marilzabicycloallene C, and marilzabicycloallene D as halogenated C15-acetogenin 12-membered bicyclic and tricyclic ether bromoallene-containing marine metabolites from Laurencia species are described. Two enantiomerically pure C4-epimeric dioxabicyclo[8.2.1]tridecenes were synthesized by E-selective ring-closing metathesis where their absolute stereochemistry was previously set via catalytic asymmetric homoallylic epoxidation and elaborated via regioselective epoxide-ring opening and diastereoselective bromoetherification. Epimeric face-selective oxidation of their Δ12,13 olefins followed by bromoallene installation allowed access to the oppositely configured 12,13-epoxides of 12-epoxyobtusallene II and 12-epoxyobtusallene IV. Subsequent exploration of their putative biomimetic oxonium ion formation-fragmentations reactions revealed diastereodivergent pathways giving marilzabicycloallene C and obtusallene X, respectively. The original configurations of the substrates evidently control oxonium ion formation and their subsequent preferred mode of fragmentation by nucleophilic attack at C9 or C12. Quantum modeling of this stereoselectivity at the ωB97X-D/Def2-TZVPPD/SCRF = methanol level revealed that in addition to direction resulting from hydrogen bonding, the dipole moment of the ion-pair transition state is an important factor. Marilzabicycloallene D as a pentahalogenated 12-membered bicyclic ether bromoallene was synthesized by a face-selective chloronium ion initiated oxonium ion formation-fragmentation process followed by subsequent bromoallene installation.


■ INTRODUCTION
Since the first report in the 1960s, 1 red algae of the family Rhodomelaceae, in particular of the genus Laurencia, have been found to give rise to fascinating structurally diverse nonterpenoid C 15 -acetogenin (ACG) metabolites as halogenated monocyclic, bicyclic, and tricyclic ring ethers 2,3 where these metabolites can be usefully classified on the basis of the largest ether ring size present. 4 These complex structures have attracted much attention as synthetic target molecules, 5,6 and recent further efforts have also been directed at further elucidating 7 and unifying 8 their biosynthetic origins. The largest reported ether ring sizes in these C 15 -ACGs are those compounds with 12-membered ether rings: the obtusallenes 9−15 and the more recently discovered marilzabicycloallenes. 16 Despite much synthetic effort in the wider family, the total synthesis of any of these 12-membered cyclic ethers remains unreported. Obtusallene II (1) ( Figure 1A) is considered to be the biogenetic precursor to the other obtusallenes and is, therefore, an interesting synthetic target. 17 What is more, obtusallenes II and IV (2), related as C 4 -epimers (C 15 -ACG numbering) and as enantiomeric R and S bromoallenes, respectively, have been hypothesized as the biogenetic precursors to marilzabicycloallenes A−D (3−6) ( Figure 1A) via 12S,13S-configured onium ions A (X = OH, Cl; R = −CHCCHBr) and, hence, 12R-configured oxonium ions B in transannular oxonium ion formation−fragmentations with attack of the nucleophile at C 9 ( Figure 1B). 16,18 Recently reported 15 12R,13R-configured coisolates 12-epoxyobtusallene IV (7) and obtusallene X (8) are therefore evidently related metabolites ( Figure 1C). Epoxide 7 is clearly related to olefin 2 by epoxidation of the Re face of the macrocyclic olefin. 19 On the basis of their absolute configurations at C 12 and C 13 , we propose that obtusallene X (8) arises biogenetically from 12epoxyobtusallene IV (7) via diastereomeric 12S-oxonium ion B′ in a transannular oxonium ion formation−fragmentation with a diastereodivergent nucleophilic attack at C 12 . The signature overall double-stereochemical inversion at this position implicates the intermediacy of the oxonium ion. It is interesting to speculate whether these different oxonium ion formation− fragmentation metabolites are formed with inherent selectivity because they arise from different starting diastereomeric forms or whether the compounds are simply representative isolates of all possible fragmentations of such oxonium ions. Herein, in an experimental exploration of the above, we report on an asymmetric strategy for the synthesis of two bicyclic 12membered ring ethers as C 4 -epimeric nitrile epimers 9 and 10 ( Figure 1, D) as putative synthetic precursors of obtusallene II and obtusallene IV. We demonstrate that the latter can serve as an advanced precursor for the synthesis of 12-epoxyobtusallene IV (7), thereby achieving the first total synthesis of a C 15 -ACG with a 12-membered ether ring from Laurencia species. Moreover, we demonstrate face-selective oxidations of nitrile epimers 9 and 10, thereby enabling remarkably selective and high-yielding diastereodivergent transannular oxonium ion formation−fragmentations for the total synthesis of obtusallene X (8) via 12S-oxonium ions of the type B′ and marilzabicycloallenes C (5) and D (6) via 12S,13S-onium ions A (X = OH, Cl, respectively) and thus 12R-oxonium ions of type B.

■ RESULTS AND DISCUSSION
Single enantiomer nitriles 9 and 10 were envisaged to be formed from epimeric acyclic dienes 11 and 12 via ring-closing metathesis, which in turn were expected to be accessible from bromochlorotetrahydrofuran 13 as a common intermediate (Scheme 1). We have previously reported the synthesis of (±)-13 18 which we have now adapted to an asymmetric method utilizing Yamamoto's catalytic enantioselective homoallylic epoxidation method. 20 Accordingly, known enediyne 14, prepared as previously described, 18 underwent transfer hydrogenation to (Z,Z)-doubly skipped triene 15 using a zinc− copper couple in a mixed solvent system of 2-propanol and water, avoiding over-reduction of the terminal alkene under these conditions (Scheme 2). Directed, catalytic asymmetric epoxidation of homoallylic alcohol 15 using Yamamoto's conditions and ligand 17 gave the desired epoxide 16. The er of epoxide 16 was determined by conversion to its O-trityl derivative 18 followed by chiral HPLC analysis revealing an er of 91:9 (see the Supporting Information) and where the absolute configuration of the major enantiomer was assigned by anology to Yamamoto's work. The subsequent steps of regioselective epoxide ring opening (giving 19), silyl protection of the primary alcohol (giving 20), diastereoselective bromoetherification (giving 21), and deprotection to bromochlorotetrahydrofuran 13 employed the previously reported conditions for the preparation of (±)-13 18 with minor modifications. With alcohol 13 in hand, it was oxidized to the corresponding aldehyde followed by immediate acetalization with the single enantiomer (S)-but-3-yn-2-ol under acidic catalysis to give acetal 22. In this reaction, the addition of

Scheme 1. Retrosynthesis of Nitriles 9 and 10
The Journal of Organic Chemistry Featured Article sodium borohydride as part of the workup procedure allowed recycling of unreacted material by recovery of alcohol 13. Subsequent cyanation of the acetal with trimethylsilyl cyanide as catalyzed by boron trifluoride etherate 21 gave the separable cyanoethers 23 and 24 in excellent yield with essentially perfect stereodivergence (1:1 ratio), presumably via the intermediacy of a planar oxonium ion. Hydrogenation of each individual epimer gave acyclic dienes 11 and 12 in high yield. With these acyclic dienes in hand, we explored the proposed ring-closing metatheses to form E-macrocyclic epimers 9 and 10. After much experimentation, and much to our delight, ring-closing metathesis using Hoveyda−Grubbs ruthenium benzylidene precatalyst 22a for diene 11 and second-generation Hoveyda− Grubbs precatalyst 22b for diene 12 in rigorously dry toluene at high dilution provided each of the desired E-macrocyclic nitrile epimers 9 and 10, respectively, as the major product in good yields. In this chemistry, the incorporation of enantiomerically pure (S)-but-3-yn-2-ol into enantiomerically enriched (91:9) aldehyde 13 resulted in the formation of a minor diastereoisomer. The minor diastereoisomer was carried through in each subsequent step to acyclic dienes 11 and 12 as an inseparable entity. After RCM, both macrocycles 9 and 10 were purified as single diastereoisomers, meaning that these compounds are enantiomerically pure.
Single crystal X-ray crystallography of each epimer 9 and 10 unambiguously established their structures confirming all relative and absolute stereochemistries ( Figure 2). These Xray crystal structures were compared with the previously obtained X-ray structures of obtusallenes II (1) 11 and IV (2), 13 respectively. This comparison established three important details. First, the epimeric macrocycles of 9 and 10 map perfectly onto the macrocyclic solid-state structures of 1 and 2, respectively (see the Supporting Information), showing that these compounds are excellent model compounds of the natural products. Second, in the solid state, each compound exposes its Re face of the C 12 −C 13 alkene where the Si face is blocked by the tetrahydrofuran. Thus, it is to be expected that 12R,13R-configured 12-epoxyobtusallene IV (7) could arise biogenetically by epoxidation of the exposed Re face of obtusallene IV (2). However, the proposed obtusallene-to-Scheme 2. Asymmetric Synthesis of Nitriles 9 and 10 Figure 2. Crystal structure of 9 (50% probability ellipsoids) (left). Structure of one of the two independent molecules present in the crystal of 10 (50% probability ellipsoids) (right).

The Journal of Organic Chemistry
Featured Article marilzabicycloallene interconversions require oxidation of the Si face of the olefins via 12S,13S-configured onium ions A (cf. Figure 1B). This apparent incongruity can be rationalized by noting that in solution obtusallenes II and IV have been shown to exist as interconverting alkene conformers, thereby exposing both Re and Si faces of the alkene. 12,14 This conformational interconversion manifests itself by broadened NMR signals for these compounds at room temperature. Epimeric nitriles 9 and 10 also display broadened NMR signals (see the Supporting Information) indicating that they behave in the same manner and are expected to expose both Re and Si faces of their alkenes in solution. As a third detail, we note that the X-ray crystal structures reveal that the two epimeric nitriles 9 and 10 have dif ferent local conformations around the C 5 −C 4 −O−C 14 torsion angle such that the nitrile group bisects the hydrogen atoms on C 5 in each case. Herein must lie the origin of their epimeric face-selective oxidative behavior (vide infra).
With epimeric nitriles 9 and 10 in hand, we planned to reduce each one to the corresponding aldehyde and then use well-established procedures 23 to install bromoallene functionalities leading to obtusallene II (1) and IV (2), respectively. Remarkably, there are no examples in the literature of the partial reduction of (allyloxy)acetonitriles, and DIBAL-H reduction of either nitrile proved to be unexpectedly troublesome and only minor quantities (ca. 10%) of the expected aldehydes could be obtained. 24 The use of model substrates 25 established that the allylic ether functional group in each is problematic, where the corresponding saturated or epoxidized model was converted to its aldehyde using DIBAL-H without incident. Accordingly, the attempted direct reduction of nitrile allylic ethers 9 and 10 with DIBAL-H was abandoned.
Epoxidation of nitrile 10 was then explored with a view toward accessing 12R,13R-configured 12-epoxyobtusallene IV (7) as a known natural product by subsequent selective nitrile reduction and installation of the requisite bromoallene. In the event, epoxidation of unsaturated nitrile 10 provided epoxides 25 and 26 in quantitative isolated yield in a 1:3 ratio, where the major product is the 12R,13R-configured epoxide (Scheme 3). As per the discussion above, this demonstrates that both Re and Si faces of the alkene in epimer 10 are accessible in solution, and for this epimer, the Re face is evidently subject to faster oxidation. Pleasingly, subsequent DIBAL-H reduction of epoxy nitrile 26 now proceeded smoothly, corroborating our findings from the earlier model studies. Bromoallene installation 23 was subsequently achieved via magnesium acetylide addition to the newly formed aldehyde to provide separable epimeric alcohols. The required alcohol 27 was converted to trisylate 28, and copper-mediated S N 2′ bromide incorporation provided 12epoxyobtusallene IV (7). 15,26 To test the proposed relationship of 12-epoxyobtusallene IV (7) to obtusallene X (8) (cf. Figure  1C) it was treated with HBr in dichloromethane solution. Much to our delight, obtusallene X (8) 15 was produced in essentially quantitative yield. 27 This experiment thereby supports its probable biogenesis via transannular formation of 12S-oxonium ion B′ and reinversion of configuration by attack of the nucleophile at C 12 . We recognized also that deoxygenation of 12-epoxyobtusallene IV (7) would provide synthetic access to obtusallene IV (2). However, despite successful deoxygenation in model studies with a representative chlorobromoepoxide, attempted deoxygenation of 12-epoxyobtusallene IV (7) under the same conditions was unsuccessful. 28 Finally, the formation of 12S,13S-epoxide 25 in the epoxidation of alkene 10 also has biogenetic significance. Marilzabicycloal-lene B (4) is proposed to arise from the 12S,13S-epoxide of obtusallene IV via oxonium ion B (cf. Figure 1B, X = OH, R = 4R-(S)-CHCCHBr). 16 This is the first experimental evidence that such an 12S,13S-epoxide can be accessed in the obtusallene IV skeleta. 29 However, we elected to explore the obtusallene-to-marilzabicycloallene rearrangements in the obtusallene II manifold instead (vide infra).
In contrast to the behavior of nitrile 10, epoxidation of epimeric nitrile 9 under the same conditions gave 12S,13Sepoxide 29 as effectively the only component in essentially quantitative yield (Scheme 4). Evidently, for this epimer, Si face oxidation is now favored. We suggest that for this epimer 12S,13S-epoxide formation results in a compound with minimal transannular strain. With a 12S,13S-epoxide of the obtusallene II framework in hand, we elected to explore the proposed obtusallene-to-marilzabicycloallene rearrangements (cf. Figure  1B). Much to our delight, on treatment with catalytic acid in methanolic solvent, 12S,13S-epoxide 29 was found to rearrange smoothly to bicyclo[5.5.1]tridecane nitrile 30 in essentially quantitative yield, thus validating the proposed transannular oxonium ion formation−fragmentation as mediated by a protonated epoxide where 12R-oxonium ion B (cf. Figure 1B, X = OH, R = 4S-CN) undergoes preferential nucleophilic attack (NuH = MeOH) at C 9 . Alternatively, DIBAL-H reduction of epoxy nitrile 29 followed by magnesium acetylide addition gave alcohol 31 along with its inseparable minor epimer, and bromoallene installation to give epoxide 32 was subsequently completed using an adaption of the established methods. 23 To our further delight, fully elaborated bromoallene epoxide 32 was also found to rearrange cleanly with catalytic acid in methanolic solvent to provide marilzabicycloallene C (5) 16,26 in essentially quantitative yield. Not only does this demonstrate the further validity of the proposed obtusallene-tomarilzabicycloallene biogenetic pathway (cf. Figure 1B   32, which we name 12-epoxyobtusallene II, as a yet to be discovered natural product from Laurencia species. 30 The experimentally demonstrated diastereodivergent selectivity observed for the position of nucleophilic attack in the above studies is intriguing and requires comment. There is obviously the initial question of epimeric face selectivity in the epoxidation of alkenes 9 and 10, but once in place the configuration of any 12S,13S or 12R,13R epoxide necessarily control the configurations of the resulting respective oxonium ion 12R-B versus 12S-B′ by stereospecific transannular epoxide ring opening. 31 For each trisubstituted oxonium ion 32 there are actually three possible positions of nucleophilic attack, C 6 , 33 C 9 , and C 12 ,where the nucleophile must approach with the normal backside stereoelectronic constraints of S N 2-type substitution. For both oxonium ions B and B′ inspection of the structures reveals that each one of these carbons is classified as secondary. Each one is also flanked by one methylene unit (C 5 , C 8 , and C 11 , respectively) and one secondary carbon each bearing a heteroatom (C 7 -Cl, C 10 -Br, and C 13 -OH). We chose to focus on oxonium ion 12R-B for a more detailed computational analysis of the possible factors controlling the regiochemical outcome.
Three different types of model were constructed, initially for X = OH, R = 4S-Me. The first involved inspecting the wave function of the reactant oxonium cation 12R-B itself. The conformational space of the larger 8-ring is complex; a partial exploration of this space showed the conformation of the reasonably related oxonium cation 34 for which a crystal structure is known, which coincided with the lowest energy conformation computed for 12R-B at the ωB97X-D/Def2-TZVPPD level using a self-consistent reaction field solvent model (cpcm, solvent = methanol). Both this conformation and computational method were used for the subsequent studies. 35 We also included in the study a reactant-based model as both a positively charged oxonium cation and with a model noninteracting counterion BF 4 − as a neutral ion pair. NBO (natural bond orbital) localization of the wave function for both models allowed the relative energies of the three C−O σ* accepting orbitals to be compared ( Figure 3). For the ion pair, the NBO energies increased in the order C 9 0.230 > C 12 0.241 > C 6 0.245 hartree, indicating the optimal position for nucleophilic attack is predicted by this approach to be at the best electron-accepting position, C 9 . The corresponding energies for 12R-B as just a cation were C 9 0.198 > C 6 0.207 > C 12 0.225 hartree, again predicting nucleophilic attack at the C 9 −O bond and coinciding with the actual outcome.
The next evolution of our reaction model was to compute the properties of transition states for reactions of a variety of nucleophiles interacting directly with the oxonium cation 12R-B (R = 4S-Me). With X = OH, we used clusters of both one MeOH and two MeOH, the latter interacting via hydrogen bonding and also anionic MeO − and Br − nucleophiles as ion pairs (Table 1). 36 In every case, the transition state of lowest free energy corresponded to attack at C 12 , promoted by the directing influence of the adjacent hydroxyl group at C 13 . When X = Cl, this effect was attenuated, and now the lowest free energy transition state emerged at C 6 . Finally, we added a noninteracting counterion BF 4 − to the transition-state model, locating the counterion in the same pocket for all three transition states. Now, C 9 and C 6 emerged as both lower than C 12 . The promotion of the C 9 position was because the dipole moment of this transition state was significantly lower than the This was also true for the other two ion-pair models using anionic MeO − and Br − as nucleophiles, where the transition state with the lowest dipole moment/charge separation was also the lowest in free energy.
Two principle conclusions can be drawn from these results. First, this model is an unusually complex one due to factors such as the conformational flexibility of the larger 8-membered ring, the hydrogen-bonding interactions possible with the incoming nucleophile, and the possibility of positional diversity of the counterion associated with the oxonium cation. A complete stochastic nondynamic exploration of each of these variables is not possible for a system of this size, and we cannot claim to have reduced each of these to the global lowest energy structures. Nevertheless, one interesting conclusion that can be drawn is that the reaction of an ion pair with a nucleophile may be strongly influenced by the charge-separation/dipole moment of the resulting highly ionic transition states. This is in addition to the more obvious structural features such as steric interactions or local hydrogen bonding. Thus, the regiochemical outcome of such reactions may well be determined by a complex blend of these various effects, with perhaps no one effect dominating. Certainly, the simpler analysis based purely on just the properties of the reactant oxonium cation should be considered as far too simplistic, even though in this specific case it predicts the "correct" outcome, for probably the wrong reasons.
Having demonstrated that a group VI onium ion from a preformed epoxide can drive these transannular rearrangements, we undertook to attempt the use of a group VII onium ion generated directly f rom olefin 9 to do so. With marilzabicycloallene D (6) as the intended target, we were delighted to find that the combination of catalytic quantities of TMG 37 and stoichiometric quantities of NCS and TMSCl 38 in dichloromethane effected the transformation of olefin 9 into trichlorobromide bicyclo[5.5.1]tridecane 33 in excellent isolated yield (Scheme 5). Evidently, the Si face of epimer 9 is again subject to kinetically controlled oxidation, now as the 12S,13S-chloronium ion A (cf. Figure 1B, X = Cl, R = 4S-CN), followed by stereospecific transannular oxonium ion 12R-B formation by chloronium ion ring-opening. As per the previously observed fragmentations of 12R-B oxonium ions (vide supra), the same remarkable selectivity for the C 9position is observed, presumably for the same reasons, but now with chloride anion functioning as the nucleophile. Subsequent DIBAL-H reduction of the nitrile, which proceeded without complication, and installation of the bromoallene 23 via alcohol 34 provided marilzabicycloallene D (6). 16  Thus, we have accomplished the total synthesis of 12epoxyobtusallene IV (7) and 12-epoxyobtusallene II (32) (as a yet to be discovered natural product from Laurencia species) as the first C 15 -ACGs with 12-membered ether rings. To the best of our knowledge, these also constitute the first total syntheses of any tricyclic ethers of this class of C 15 -ACGs. We also report the total synthesis of obtusallene X (8), marilzabicycloallene C (5), and marilzabicyclocallene D (6) via consideration, proposition, and exemplification of their biogeneses via oxonium ion formation−fragmentation reactions. These studies show that these metabolites are not simply representative isolates of all possible formation−fragmentations of such oxonium ions but rather are produced by inherently selective pathways. A density functional mechanistic exploration of one of these pathways involving ring opening of an intermediate ion-pair complex suggests that a major factor in the selectivity may be the dipole moment magnitude at the transition state.
■ EXPERIMENTAL SECTION General Information. Quinoline was dried over Na 2 SO 4 and distilled from and stored over Zn dust. Triethylamine was dried over CaSO 4 prior to distillation under nitrogen and was subsequently stored over 4 Å molecular sieves. TBCO was prepared according to the method of Matveeva. 39 Asymmetric epoxidation ligand 17 was prepared according to the method of Yamamoto. 20b All other reagents were obtained from commercial sources and used as received. All reactions were performed in anhydrous solvents unless used in combination with H 2 O. CH 2 Cl 2 , THF, and Et 2 O were dried by passing through a column of alumina beads. Toluene was distilled from sodium and benzophenone immediately before use. MeOH, EtOH, MeCN, and glacial AcOH were used as received. Extraction solvents and chromatography eluents were used as received. MeOH and CH 2 Cl 2 were HiPerSolv grade, EtOH was AnalaR grade, and n-hexane, Et 2 O, EtOAc, and petroleum spirit 40−60°C were GPR grade. Benzene was purchased and used as received. Reactions were carried out in oven-dried glassware under an inert atmosphere of nitrogen unless otherwise stated. Air-and moisture-sensitive reagents were transferred by syringe or cannula. Molecular sieves (4 Å) were dried by repeatedly heating under vacuum and flushing with nitrogen. Reaction temperatures other than room temperature were recorded as aluminum heating block or bath temperatures. Temperatures below room temperature were achieved by an ice/NaCl bath or acetone/dry ice bath. Brine refers to a saturated aqueous solution of NaCl. Column chromatography was performed on silica gel, particle size 33−70 μm or 40−63 μm. Analytical TLC was performed on Kieselgel 60 F254 precoated aluminum-backed plates which were visualized by ultraviolet light (254 and 350 nm) and/or chemical staining using potassium permanganate or an acidified solution of vanillin. Fourier transform IR spectra were recorded as neat samples using an ATR-IR spectrometer. 1 H NMR spectra were recorded at 400 or 500 MHz. 13 C{ 1 H} NMR spectra were recorded at 101 or 126 MHz. Chemical shifts (δ) are quoted in parts per million (ppm) and are referenced to the residual solvent peak. Coupling constants (J) are quoted in hertz (Hz). All NMR spectra were acquired at room temperature unless otherwise stated. Low-resolution MS were performed using ESI, EI, or CI methods and ToF or magnetic sector analysis. Chiral analytical HPLC was performed on a 25 cm × 4.6 mm ChiralPak AD or ODH column. All solvents for HPLC were HiPerSolv grade and used as received.