Stereodivergent Total Syntheses of (+)‐Mycaperoxides C, D, G Methyl Ester and (−)‐Mycaperoxide B

Abstract Mycaperoxides are natural endoperoxides isolated from different Mycale genus sponges, showing significant antiviral or antibacterial activities. We report herein the first total syntheses of representative congeners of this family from sclareol using a stereodivergent approach. Thus, an innovative oxidative ring expansion of cyclobutanol was used to bring the 1,2‐dioxane subunit, and a Mukaiyama aldol reaction on peroxycarbenium species was utilized to install the propionic acid subunit. During the study toward (+)‐mycaperoxide D methyl ester (2), the isolation of the eight possible diastereomers under their ethyl thioester form allowed to build a pertinent database for further NMR assignment studies. Thus, we completed the total syntheses of (+)‐mycaperoxides D, C, G methyl ester, and (−)‐mycaperoxide B in 11 to 15 steps, confirming their original assignment.


Introduction
The endoperoxide motif is an uncommon chemical function in natural products. However, a relatively large group of marine norsesterterpene metabolites represent a significant class of natural endoperoxides. [1][2][3] Indeed, more than 70 members have been isolated from various sponges of the class Demospongiae, including the genera Prianos, Sigmosceptrella, Lacuntrulia, Diacarnus or Mycale. Structurally, these endoperoxides have a common backbone consisting of a 1,2-dioxane ring substituted by a propionic acid subunit on one side and a sesquiterpene subunit connected via a methylene bridge on the other side. The main difference between all members lies in the nature of the sesquiterpene side chain and, in particular, the degree of cyclization of the chain, the intervention of methyl transpositions, and the enantiomeric series of this chain. The absolute configuration of the stereogenic centers of this subclass of natural products is not identical from one product to another, depending mainly on their origin. Despite many members, this subclass has been barely studied at the synthetic level. Indeed, Harwood reported several synthetic studies towards the total synthesis of mycaperoxide B, leading, with difficulty, to the obtention of four inseparable diastereomers, in a total of 16 steps, sharing the gross structure of the target.
However, none of them corresponded to the natural product. [4][5][6][7] Later, Seifert reported the total synthesis of diacarnoxide C in about 20 steps but obtained as an inseparable mixture of cis and trans-1,2-dioxanes with epimerization at C 6 ( Figure 1). [8] Lately, Wu and co-workers reported the enantioselective total synthesis of (+)-Muqubilin from seudenol, by adding hydrogen peroxide to a chiral epoxide. Still, a long synthetic sequence was necessary to reach this objective (29 linear steps). [9] Recently, we have developed a synthetic method to construct 1,2-dioxanes from the oxidative expansion of cyclobutanols, [10] followed by the alkylation of the resulting peroxyacetals. [10][11][12] Thus, we felt it would be particularly well suited to the total synthesis of several members of this family of norsesterterpene endoperoxides. Mycaperoxides were targeted as some members have been reported to have significant antiviral activities (IC 50 = 0.25-1.0 μg/mL against vesicular stomatitis virus and herpes simplex virus type-1) and antibacterial properties (inhibition of the growth of gram-positive bacteria Bacillus subtilis and Staphylococcus aureus). [13]

Results and Discussion
In terms of synthetic access, these molecules, mainly mycaperoxides B, C, D, and G, would be strategically accessible from sclareol (Salvia sclarea) used as starting synthon. Concerning the chirality control strategy, mycaperoxide B (4) would be directly accessible from sclareol as the latter corresponds to the proper enantiomeric series of the decalin moiety. For the mycaperoxides C, D, and G, their antipodes (ent-mycaperoxides C, D, and G) would be accessible by total synthesis from sclareol. For these three compounds, methyl esters (1), (2), and (3) were targeted as the natural products were isolated and characterized in this form after derivatization with diazomethane. [14,15] Thus, a stereodivergent total synthesis was collectively designed for all these products. They could be obtained by a Mukaiyama aldol reaction performed on an acetoxy-endoperoxyketal. [10][11][12] The 1,2-dioxane ring of the latter could be constructed from a cyclobutanol by ring expansion with molecular oxygen mediated by Co(acac) 2 . [10] The cyclobutanol could be easily obtained from a methyl ketone derived from sclareol. The four required decalins would thus be prepared from the same starting synthon via divergent functional arrangements. (Scheme 1) The synthesis of the different targeted mycaperoxides consisted first of the functionalization of sclareol (5). The latter underwent an oxidative cleavage with KMnO 4 leading to methyl ketone 6. [16] A Wittig olefination then led to the formation of cyclopropylidene 7, [17] the precursor of mycaperoxide D. To obtain the other congeners, a hydroxyl elimination gave rise to a mixture of isomers 8 and 8' in an 85 : 15 ratio and thus gave access to the other required decalins. Epoxidation of the 8/8' mixture followed by Wittig olefination provided the three separable isomers 9 a, 9 b, and 9 c in a 61 : 30 : 9 ratio. Compounds 9 a and 9 b were subjected to reduction with LiAlH 4 . From 9 a, hydride attack led exclusively to 10 a, the precursor of mycaperoxide B and ent-mycaperoxide G. The same reaction performed on 9 b gave only 10 b, the precursor of ent-mycaperoxide C. The regioselectivity of hydride attack is in agreement with the Fürst-Plattner rule. [18][19][20] The synthesis of ent-mycaperoxide D was firstly studied for its better accessibility. Thus, after the protection of the hydroxyl in the form of a silyl ether (TES), the cyclopropylidene was epoxidized. In situ acid-catalyzed rearrangement led then directly to cyclobutanone 11. After reduction with NaBH 4 , the resulting cyclobutanol underwent a ring enlargement reaction in the presence of Co(acac) 2, affording 1,2-dioxanol 12 in good yield. [10] Although any initial stereochemistry at C 3 and C 6 was lost during the process due to hydroperoxyacetal equilibrium and radical mechanism of the reaction, we observed that the decalin slightly induced stereoselectivity during the cyclobutanol oxidative ring opening (see below for further discussion).
The hydroxyl was acetylated in the presence of pyridine, [21] and the resulting acetoxy-endoperoxyacetal 13 reacted with silylketenethioketal 14 [22] as well as a catalytic amount of Sc(OTf) 3 . [10] At this step, the first resolution of diastereomers [2,3-syn/2,3anti and cis/trans-1,2-dioxanes] was thus performed by silica gel chromatography. Therefore, three fractions (F1, F2, and F3) were obtained in 8 %, 32 %, and 31 % yield, respectively; each containing a mixture of defined diastereomers among the eight possible (Scheme 3, A). The stereoselectivity of the reaction was estimated to be 4 : 1 in favor of cis-1,2-dioxanes, but no selectivity was observed for the syn/anti relationship at C 2 -C 3 . After cleavage of the silyl ether by acid methanolysis performed independently on each fraction (F1, F2 and F3), a further separation of diastereomers was again done to deliver thioesters 16 a-h. All cis-1,2-dioxanes could be isolated as single stereoisomers but, trans-1,2-dioxanes, do not allow a separation of their 2,3,6-epimer and were thus obtained as a mixture of two diastereomers. For the trans-1,2-dioxanes mixtures, a selectivity about 55 : 45 in favor of one particular 2,3,6-isomer, allowed the differentiation of carbon for attribution.
The structures of the different diastereomers were attributed as follow by using Capon's empirical rules [23] based on the 13 C NMR chemical shift. Firstly, cis/trans-1,2-dioxanes were assigned thanks to 13  ] were the most difficult to discriminate due to the slight chemical shift differences due to the spacer between decalin and 1,2-dioxane units. Structure of cis-dioxanes 16 g and 16 h were deduced from the similarities observed on 13 C NMR with natural (À )-mycaperoxide D. Although tight chemical shift differences at C 7 and C 8 , these positions are characteristic (� 0.2-0.8 ppm difference for each 2,3,6-epimer pairs) and cis-dioxanes 16 c and 16 d were next assigned. A more significant difference can be observed at C 11 for cis-dioxanes 16 c and 16 g versus trans-dioxanes 16 d and 16 h (+ 1.6 ppm), which is indicative of a hydrogen bond of the hydroxyl at C10. Indeed, DFT calculation at the B3LYP/6-31G(d,p) showed that among the lowest energy conformers, an evident hydrogen bond exists between this hydroxyl group and carbonyl group at C1 with a distance of 2.18 Å (Figure 2). This particularity also explains the ability of cis-dioxanes 6 c-d and 6 g-h to be isolatable as lone diastereomers compared to protected compounds 15 a-h and trans-dioxanes where this hydrogen bond is not possible. Thus, collectively taking into account the 13 C NMR chemical shifts at positions C 7 , C 8 , and C 11 , 2,3,6-epimers 16 a, 16 b, 16 e, and 16 f were assigned.
Among the eight diastereomers 16 a-h, thioester 16 h was spectroscopically the best match with mycaperoxide D methyl ester. Thus, in two steps, 16 h was transformed into methyl ester 2 by transesterification. Spectral data of synthetic 2 were in agreement with the description of the natural product and the antipode sign of the optical rotation compared to those of natural mycaperoxides D methyl ester [+ 68 (c 0.25, CHCl 3 ); lit: [13] À 52 (c 0.3, CHCl 3 )] confirmed the original assignment of mycaperoxide D (Scheme 3A).
Afterward, the synthesis of (+)-mycaperoxide C methyl ester (1) was investigated. Thus, endoperoxyacetal 19 was obtained in five steps from 10 b, using a sequence of ring-expansion of cyclobutanone 17 producing endoperoxide 18, similar to those described for the total synthesis of 2. A notable difference between the two synthetic routes lies in using TMS as the protecting group for the hydroxy group at C10. The conformation of 10 b did not allow the introduction of TES; consequently, only the use of TMS-triflate with Et 3 N succeeded in the protection step. However, the use of TMS led to an unexpected consequence. Upon Mukaiyama-type aldol reaction, in the presence of Sc(OTf) 3 on 19, the hydroxyl at C10 was eliminated. The use of less active Y(OTf) 3 [11] allowed the aldolization to be carried out without significant side-reaction (Scheme 3B). The selectivities were comparable for the stereocenters at C 6 and C 2 to that was obtained in the total synthesis of 2. Again, during this synthesis, the formation of the cis-dioxane rings (C 3 ) was favored over the trans one, with a 2 : 1 cis:trans ratio. As described in the total synthesis of 2, we obtained a mixture of diastereomers distributed over three fractions. The fraction F3 was also containing the isomer corresponding to the natural product according to 13 C NMR data. After deprotection of the alcohol function, the compound of interest 21 h was separated from its isomer 21 g. The final transesterification performed on 21 h led to the synthesis of the (À )-mycaperoxide methyl ester C (1). Spectral data matched to the described natural product and, again, the optical rotation [+ 50 (c 0.198, CHCl 3 ); lit: [13] À 71 (c 1.1, CHCl 3 )] confirmed the synthesis of ent-mycaperoxide C methyl ester and the original assignment of mycaperoxide C (Scheme 3B).
Regarding (+)-mycaperoxide methyl ester G (3) and (À )mycaperoxide B (4), their total synthesis was planned from compound 10 a. The same reaction sequence used to prepare intermediates 13 and 19 (Scheme 3A and B) led to the formation of 24 in five steps. However, here again, a particular reactivity of substrate 10 a led us to some adaptations. Hydroxyl in position C 9 was reluctant to protection and systematically underwent an elimination reaction during all attempts to form a silyl ether. It was thus chosen to leave it free for the rest of the synthesis, its tertiary position inducing a steric hindrance making it less accessible and therefore less reactive. Indeed, the acetylation of 23 was carried out chemoselectively on the 1,2dioxane hydroxyl. To avoid further elimination reactions of the hydroxyl at the C 9 position during the Mukaiyama aldol reaction, Y(OTf) 3 was used again, which led to the obtention of two fractions, F1 and F2 in 87 % yield (1,2-dioxane cis:trans = 3 : 2; 2,3-syn:2,3-anti = 1 : 1). The fraction of interest, F2, contained a mixture of three products, including isomers 25 e and 25 f, which corresponded to the relative stereochemistry of mycaperoxide B (4) and (+)-mycaperoxide G methyl ester (3) respectively. The utilization of preparative HPLC for fraction F2 led to the separation of isomer 25 h [24]  From the fraction F2 containing isomers 25 e-g, elimination of hydroxyl group induced by the action of SOCl 2 led selectively to the tetrasubstituted olefin. Compound 27 g and a mixture of 27 e and 27 f were thus obtained. The combination of 27 e/27 f was transesterified, finalizing the synthesis of the (+)-mycaperoxide methyl ester G (3) as a mixture with 28 e. As for 26 e and 4, the 13 C NMR data for products 3 and 28 e were very close. The most significant differences were observed for the ethylenic carbons at C 9 and C 10 [δ = 0.2 ppm]. The synthetic product with NMR signals matching the most with the natural product was attributed to structure 3 (Scheme 3C).

Conclusion
In summary, we were able to construct for the first time several members of the mycaperoxides, about thirty years after their discovery, employing an oxidative expansion of cyclobutanol to 1,2-dioxane followed by a Mukaiyama aldol reaction on a peroxycarbenium species as key steps. Thus, the methyl esters of (+)-mycaperoxides C (1) and D (2) were obtained in 15 and 11 steps, respectively. The methyl ester of (+)-mycaperoxide G (3) and (À )-mycaperoxide B (4) were obtained in 14 and 13 steps but combined with their respective 2,3,6-epimer. Original assignments were thus confirmed. We highlighted the utility of yttrium compared to scandium in the Lewis acid-catalyzed Mukaiyama aldol reaction, which allowed smoother conditions reducing elimination side reaction with tertiary alkoxides. Although our approach did not allow us to control all the stereogenic centers of the synthetic targets, the different mycaperoxides studied were obtained efficiently accordingly to literature precedents. [7,8] In addition, obtaining several wellidentified diastereomers provides a valuable NMR database for the assignment of new natural endoperoxides of the same family. 2,3-syn/2,3-anti and cis/trans-1,2-dioxanes relationships are easily determined by this method. However, the assignment of the 2,3,6-epimers is more challenging, leading to small differences in NMR spectral data. Moreover, the affected positions are also a function of the nature of the decalin unit, which makes a general rule impossible for these pairs of diastereomers. This project also underlines the lack of enantiomeric methods to build organic peroxides, with few reported examples. [25][26][27] Therefore, significant work must be undertaken to overcome the actual limitations in the field. [28] Such effort will be reported from our laboratory in due course.

Experimental Section
General information, detailed experimental procedures, characterization data, cartesian coordinates of 16 h, tables of NMR data for each mycaperoxides and copies of 1 H and 13 C NMR spectra of all compounds are in the Supporting Information.