Enantioselective Synthesis of the Predominant AB Ring System of the Schisandra Nortriterpenoid Natural Products

: An enantioselective synthesis of the AB ring system common to the majority of the Schisandra nortriterpenoid natural products is reported. Key steps include a stereospeci ﬁ c ring opening of a trisubstituted epoxide and the use of a β lactone to enable installation of the gem -dimethyl functionality of the B ring. An acetalization strategy played a key role in a late-stage biomimetic AB ring bicyclization.

E xtracts from the Schisandra genus of herbal plants have long been used in Chinese herbal medicines. Studies by Sun et al. over the past decade have resulted in the isolation of more than 120 bioactive nortriterpenoids from these plants, many of which have anti-HIV or other bioactivities. 1 While these properties make them attractive targets for synthetic endeavors, 2 it is the challenge posed by their beautiful and complex skeletons which has driven organic chemistry research in this field−a challenge that is underlined by the solitary total synthesis to date of just one family member, (±)-schindilactone A (1, Scheme 1). 3 Common to almost all of these natural products is a highly oxygenated [3.3.0] bicyclic lactone AB ring system, epitomized by lancifodilactone I and rubriflordilactones A and B (2−4). In this respect, schindilactone A in fact represents a structural outlier in featuring an additional alcohol substituent at the AB ring junction, an attribute that played a key role in the formation of its A ring via aldol cyclization in the work of Yang et al. 3 The rubriflordilactones and lancifodilactones are longstanding synthetic targets within our group and have inspired the development of various metal-catalyzed methods for the formation of their CDE ring systems. 4 In planning for an assault on the full framework of these natural products, we ideally required a preformed AB ring system, onto which appropriate functionality could be attached in readiness for CDE ring cyclization. It is notable that all previous approaches to the ABC rings have used a preformed C ring (or equivalent) as a template, 2 a strategy that we were keen to avoid due to the skeletal diversity that might be introduced from CDE cyclization subsequent to AB ring formation. Here we report an enantioselective route to the AB rings that permits such future explorations, which is characterized by biomimetic lactonization/oxy-Michael transformations.
A retrosynthetic analysis that depicts these considerations in the context of rubriflordilactone A is illustrated in Scheme 1, the CDE rings of which could arise from a generalized bromoendiyne precursor 5. This disconnection reveals the 'unified' nature of our synthetic plan, as addition of a variety of alkynes into aldehyde 6 could lead to different cyclization substrates (including 5) and therefore different natural products. Aldehyde 6 in turn was envisaged to arise from possible biomimetic processes such as an oxy-Michael addition and lactonization from diol-enoate 7. The two stereocenters in this enoate could be installed from a regioselective ring opening of an appropriate trisubstituted epoxide with an allyl organometallic reagent. The synthesis of aldehyde 6, which is key to all our future synthetic work, forms the focus of this paper.
Our work commenced with an exploration of this epoxide ring opening. An ideal precursor to 7 would be a diol such as 8 (Scheme 2), which we might approach either from a hydroxyl-(or steric-) directed ring opening 5 of hydroxymethyl epoxide 9, followed by oxidation and dimethylation, or from a direct but challenging opening of a more hindered epoxide such as 10. The route to 9 began with a Pd-catalyzed hydrostannylation 6 of but-2-yne-1,4-diol, followed by regioselective monosilylation. Subsequent π-allyl Stille coupling 7 of stannane 11 with 2,3dibromopropene proved surprisingly challenging. Despite the use of an excess of 2,3-dibromopropene (3 equiv), a significant amount of byproduct 12 was produced (21%) which arises from a second Stille coupling of desired product 13, implying that the vinyl bromide in 13 is at least as reactive toward Pd(0) as the allylic bromide in 2,3-dibromopropene. Fortunately, these products were easily separable, and 13 could be isolated in a respectable 56% yield on multigram scale. In order to reveal the hydroxyl group required to direct epoxide opening and also mediate an asymmetric epoxidation, a protecting group switch afforded 14; epoxidation with m-CPBA, or a Sharpless AE (97%, 93% ee), delivered epoxide 9 in racemic and enantioenriched form, respectively. 8 For comparison in the ring-opening reactions, disubstituted epoxide 15 was prepared in two steps from cis-butene-1,4-diol, and disubstituted epoxide 16, which would enable us to evaluate the tolerance of steric effects around the 'directing' alcohol adjacent to the site of nucleophilic attack, was prepared in two steps from acetone adduct 17. 9 Having assembled a selection of epoxides, we moved to the challenge of epoxide ring opening (Table 1). We first confirmed that typical Cu-catalyzed hydroxyl-directed epoxide opening could be achieved on disubstituted epoxide 15 (entry 1) using conditions employed by Paquette et al. for the opening of a trisubstituted epoxide with a butenyl Grignard reagent (cat. CuCN, allylMgCl, THF). 2i This ring opening proceeded uneventfully, with product 18 isolated in good yield. However, application of these conditions to the more hindered epoxide 16 failed, resulting in no observable reaction even in the presence of excess nucleophile and stoichiometric Cu(I) salt (entries 2, 3). Hindered ring openings of this nature are unprecedented and suggest that neopentylic steric considerations prohibit nucleophile approach. This result thus dictated that we would need to install the B ring gem-dimethyl group into tertiary alcohol 8 af ter allylation.
With this in mind, we turned to trisubstituted epoxide 9. Again, the use of the previously successful copper(I) cyanide promoted conditions failed; these, and indeed any other Cucatalyzed methods, 10 led only to the halohydrin product 19 or 20 in good-to-excellent yields (entries 4, 5), where the halide derives from the counterion of the Grignard reagent. Halohydrins have been reported by Hatakeyama et al. in the opening of a related epoxide with an allyl Grignard reagent; their solution simply involved exclusion of the copper salt from the reaction. 11 To our delight this also proved effective for us, delivering the targeted epoxide 21 in excellent yield (94%, entry 6) and in a short time scale.
With an enantioselective route to diol 21 in place, installation of the gem-dimethyl group was next addressed. We first attempted a stepwise oxidation/methylation sequence via intermediate ketone 22 (Scheme 3). However, the second addition of MeMgBr to 22 proved surprisingly low yielding (24%), 12  ketone from this reaction had undergone partial epimerization at the allyl-bearing C3 stereocenter, indicating preferential enolization over nucleophilic addition. 13 An alternative strategy of oxidation to the methyl ester 23, followed by treatment with methylmagnesium bromide, proved even less fruitful; 23 was resistant to the Grignard reagent even at or above rt. We ascribe the epimerization of 22 and the surprisingly low reactivity of 23 to a chelated reactive conformation 24 (box, Scheme 3), which on stereoelectronic grounds requires addition of a methyl nucleophile to the hindered concave face of this bicyclic complex; meanwhile, the C−H bond α to the ketone is ideally aligned for enolization. A solution to this problem presented itself first by serendipity 14 and then by design in the form of βlactone 25, which could be produced in high yield from acid 26 using BOPCl. This strained lactone does not suffer from the immediate problem of forming chelate complex 24 and, to our delight, indeed underwent successful double addition of methylmagnesium bromide to give the targeted tertiary alcohol 8 (69%).
With 8 in hand, the route to the AB rings now appeared straightforward (Scheme 4). First, acidic deprotection of the PMB ether (which proved more efficient than oxidative methods) and a two-step oxidation sequence afforded the lactols 27. Notably, the choice of oxidant in this step (SO 3 ·py) was crucial to avoid the formation of ketone 28 via oxidative fragmentation of the intermediate triol under other conditions (e.g., Dess-Martin); precedent for this side reaction is found in Dess and Martin's own work 15 and in a mechanistic study by Santagostino et al. 16 We presumed (on the basis of literature precedent of a related C ring-tethered lactol) 2l that 27 would be well-suited to a Wittig or HWE olefination/in situ cascade cyclization to the AB rings. Indeed, this exciting transformation led, on a single occasion, to a product that was tentatively assigned as the AB rings 29. However, this success proved irreproducible, and despite many attempts, only decomposition pathways were observed. It would appear that without the imposed conformational influence of a template C ring, Thorpe−Ingold effects (from the two quaternary carbons in lactol 27) strongly disfavor the equilibrium of the lactol anion with the open chain aldehyde and thus prevent 27 from participating in productive olefination.
This seemingly insurmountable obstacle necessitated a change of strategy. Given that the allyl side chain in 8 was eventually destined for oxidative cleavage to reveal an aldehyde, we realized that this aldehyde could also provide a transient means to prevent the formation of the troublesome B ring lactol by sequestering the tertiary alcohol as an acetal. Thus, oxidative cleavage of the alkene in 8 gave an epimeric mixture of γ-lactols. After acidic cleavage of the PMB ether, treatment with acidic methanol led to the methyl acetals 30 as the exclusive regioisomers (despite the presence of two other hydroxyl groups which might form cyclic acetals), tying up the B ring oxygen and, we hoped, allowing A ring construction. This required the oxidation of the primary alcohol in 30 to form aldehyde 31, which again required Parikh−Doering conditions to avoid similar oxidative fragmentation as observed previously (see 28) and gave aldehyde 31 in quantitative yield.
Although an (E)-selective Horner−Wadsworth−Emmons reaction and subsequent cascade cyclization on 31 should afford the AB rings, we recognized that a (Z)-selective olefination and in situ A ring lactonization would render the subsequent B ring oxy-Michael cyclization stereospecific (a strategy related to the work of other researchers where the C ring template is present), 2b,g rather than relying on an equilibrium situation to control B ring formation. We tested both the Still−Gennari 17 and Ando phosphonates 18 and found that while the Still−Gennari olefination was moderately stereoselective, affording butenolide 32 as the major product along with some of the noncyclized (E)-alkene, the Ando variant gave exclusively the desired (Z)-olefination product 32. An excess of phosphonate reagent (>1.5 equiv) was necessary to drive the reaction to completion, as otherwise a significant amount of byproduct 33 (30−55%), formed by a basepromoted rearrangement of the aldehyde 31, was observed. Separation of the excess phosphonate proved difficult, but after deprotection of the methyl acetal 32 with TFA, the lactols 34 were isolated cleanly. Completion of the AB ring fragment 6 required only ring opening of the lactol in α,β-unsaturated lactone 34 and subsequent oxy-Michael addition of the tertiary alcohol onto the α,β-unsaturated lactone. To our delight, this reaction was triggered under mild basic conditions to afford 6 in 99% yield. The structure and stereochemistry of this product was confirmed by NMR spectroscopy, including 1 H NMR NOE experiments. 19 In summary, we have established an enantioselective synthesis of the AB rings common to many of the Schisandra nortriterpenoid natural products, using an oxy-Michael/ lactonization strategy. The final aldehyde product is well suited to a subsequent diversity-oriented approach to members of this family, the details of which will be reported in due course.
■ ASSOCIATED CONTENT * S Supporting Information Experimental procedures, characterization data for novel compounds, and copies of 1 H and 13 C NMR spectra. This material is available free of charge via the Internet at http:// pubs.acs.org.

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
We thank the EPSRC for an Advanced Research Fellowship (EAA) (EP/E055273/1), A*STAR for a National Science Scholarship (SSG), and postdoctoral support through a fellowship from the German Academic Exchange Service (DAAD) (B.G.).