Asymmetric Total Synthesis of (–)-Verrucarol

Abstract

We disclose a concise, asymmetric synthesis of (−)-verrucarol, the natural product precursor of bioactive trichothecene macrocycles including verrucarins A and J. Historically, enantioselective construction of the 6/6/5 (A/B/C) ring system of the 12,13-epoxytrichothec-9-ene (EPT) core of trichothecene natural products has been a significant challenge. This architecture was constructed beginning with a chiral Lewis acid-catalyzed, enantioselective [4+2] cycloaddition to afford a cis-fused A/B ring fragment. A mechanistic hypothesis for the pivotal [4+2] cycloaddition was developed through experimentation and density functional theory (DFT). En route to completing the 6/6/5 core, we demonstrate diastereoselective modifications of the [4+2] cycloadduct including silylative generation and addition to a vinyl oxocarbenium. Final C-ring closure was achieved by a diastereoselective 5-endo-trig cyclization, thereby completing formal [3+2] annulation over two steps. Late-stage alkene transposition and hydroxyl-directed epoxidation furnished the target compound, (−)-verrucarol.

Graphical Abstract

A concise, asymmetric synthesis of (−)-verrucarol has been achieved in 17 steps from a commercially available starting material. Upon construction of a functionalized 4-pyrone substrate, the key transformations to accomplish the synthesis include chiral Lewis acid-catalyzed, enantioselective [4+2] cycloaddition, selective methyl addition, and diastereoselective formal [3+2] annulation.

Introduction

Since the isolation of trichothecin from T. Roseum,^[1] trichothecene sesquiterpenes have garnered significant interest by exhibiting antibiotic, antiviral, antifungal, antimalarial, and anticancer activities (Scheme 1a).^[2–6] Verrucarin A (1) indirectly inhibits steroid receptor coactivator-3 (SRC-3), an oncoprotein overexpressed in breast, pancreatic, gastric, prostate, and colorectal cancers, among others (e.g. MCF-7 cells, IC₅₀=4.86 nM).^[7,8] Moreover, 1 exhibited activity in vivo against neuroendocrine tumors (NETs), glioblastoma (GBM), and triple-negative breast cancers (TNBCs) via monoclonal antibody-directed extracellular vesicle delivery.^[9–11] A congener of 1, verrucarin J (2), showed nanomolar activity against ovarian and lung cancer cell lines and targeted cancer stem cells (CSC’s).^[12–14] Several biological modes of action for trichothecenes have been elucidated,^[3,4] with a primary mechanism of cytotoxicity involving inhibition of protein synthesis.^[15–17] Key insights into this mechanism came from a co-crystallographic study by Yusupov and coworkers who reported non-covalent coordination of several trichothecenes within the 60S tRNA A-site of the eukaryotic peptidyl transferase center.^[17] In light of the therapeutic potential of trichothecene macrocycles, developing an efficient route to the core natural product scaffold (−)-verrucarol (3) is crucial.^[18] (−)-Verrucarol (3) is a non-macrocyclic trichothecene first obtained via base-mediated hydrolysis of macrocyclic congeners and through isolation from S. atra, F. sporotrichioides, and S. chatarum.^[19,20] While 3 exhibits poor cytotoxicity relative to macrocyclic congeners,^[21] it features the 12,13-epoxytrichothec-9-ene (EPT) core, making it a critical synthetic entry point towards bioactive trichothecenes (Scheme 1a). The EPT core is a 6/6/5 ring system comprised of six total stereocenters, two contiguous quaternary centers, a bridged spiro-epoxide, and C9–C10unsaturation. Given these structural complexities, the synthesis of 3 represents a formidable synthetic challenge that unsurprisingly garnered interest from the synthesis community throughout the late 20^th century. This challenge was overcome using racemic routes employing both biomimetic and non-biomimetic approaches.^[22–27] The first asymmetric total synthesis was reported by Tadano and coworkers (Scheme 1b).^[28] After substantial elaboration of D-glucose,^[29] precursor 4 was subjected to fluoride-mediated 1,2-migration (as previously demonstrated by the Trost group) to afford the 6/6/5 trichothecene core in 42 steps. Further functional group interconversions afforded 3 in 49 total steps.

Scheme 1.

a) Cytotoxic trichothecene macrocycles, (−)-verrucarol, and 12,13-epoxytrichothec-9-ene (EPT) with A/B/C ring assignments and carbon numbers. b) The first asymmetric synthesis of 3 started from D-glucose and utilized fluoride-induced 1,2-migration. c) Our retrosynthetic analysis of 3.

The racemic and/or lengthy approaches described are restrictive towards trichothecene medicinal chemistry efforts; therefore, further development of synthetic methods towards 3 were warranted. Methods of extracting trichothecenes from natural sources and the semi-synthesis of 3 also remain uneconomical or may not be amenable to structure-activity relationship (SAR) studies.^[30–32] Accordingly, we sought to achieve a concise asymmetric synthesis of 3 using the strategy outlined in Scheme 1c. By utilizing late-stage hydroxyl-directed epoxidation and alkene transposition, we envisioned rapid construction of the 6/6/5 (A/B/C) ring system of 3 through use of sequential [4+2] and formal [3+2] cycloaddition methodologies. Disconnections at C5 and C2 to install the C-ring may be enabled by diastereoselective, formal [3+2] annulation of vinyl oxocarbenium (5). The requisite cis-fused dihydrochromene 6 (A/B ring fragment) could be accessed from the achiral 4-pyrone 7 through a [4+2] cycloaddition rendered enantioselective through asymmetric catalysis.

Results and Discussion

Our early-stage synthesis approach drew inspiration from the formal synthesis of (±)-verrucarol by White and coworkers who reported a thermal, dearomative [4+2] cycloaddition of a 4-pyrone ester with 2-ethoxy-1,3-butadiene.^[26] We aimed to differentiate from and improve upon this approach by developing a [4+2] cycloaddition between benzylated 4-pyrone-ester 10 and isoprene, thereby enabling a new asymmetric strategy (Scheme 1c) utilizing a simpler diene to minimize late-stage functional group interconversions. We began with the preparation of 10 starting from commercially available ethyl 4-(benzyloxy)acetoacetate 8 on a decagram scale (Scheme 2a).^[33] Compound 8 was subjected to N,N-dimethylformamide dimethylacetal (DMF-DMA) to afford enamine 9, which underwent subsequent aromatization to 10 in the presence of ethyl formate and potassium tert-butoxide (KOtBu). Efforts to induce [4+2] cycloaddition between 10 and isoprene using both thermal and Lewis-acidic conditions were unsuccessful; hence, 10 was modified via a debenzylation/tosylation sequence to afford 11a – a more electron-deficient dienophile (Please see Section III of the Supporting Information for further details).^[34] Conversion to cycloadduct (±)-14a was then achieved by implementing stoichiometric Lewis acids including zinc(II) chloride (ZnCl₂) or scandium triflate (Sc(OTf)₃). These findings prompted exploration of chiral bis(oxazoline), pyridine-2,6-bis(oxazoline), and N,N’-dioxide ligands as Zn(II), Sc(III), and Cu(II) complexes, which did not induce enantioselectivity (Tables S1–S4).^[34–37] Chiral phosphoric acids were also investigated as organocatalysts, but these efforts were similarly fruitless (Table S5).^[38]

Scheme 2.

Please see the Supporting Information for detailed reaction conditions. a) Construction of the A/B ring system leveraging preferential enrichment of (R,R)-14a. b) Table detailing an investigation of additives on the [4+2] cycloaddition of 11a to (R,R)-14a. c) [4+2] cycloaddition of modified4-pyrone sulfonates. d) Density functional theory (DFT) geometry optimizations of catalyst/substrate 11a complexes.

Ultimately, modest success was achieved by employing Jacobsen’s Cr(III)–salen catalyst 12, which afforded partial conversion of 11a to cycloadduct (R,R)-14a in 28% ee (Scheme 2b, Entry 1).^[39–41] The absolute configuration was determined by X-ray crystallography.^[42] Poor conversion was attributed to the insolubility of catalyst 12; hence, we examined hexafluoroisopropanol (HFIP) as a polar protic additive.^[43] This modification resulted in quantitative conversion to (R,R)-14a and a remarkable improvement in selectivity (66% ee, Scheme 2b, Entry 2).

As we recognized that HFIP could operate through several modes including coordination to metal complexes and substrate activation, we chose to investigate conditions further.^[44] Use of catalyst 13 featuring an SbF₆⁻ counterion without additives (Scheme 2b, Entry 3) gave comparable enantioselectivity to prior entries and was consistent with precedent detailing improvements through weakly coordinating anions.^[41,45] No further improvement was observed employing HFIP with catalyst 13 (Scheme 2b, Entry 4). Molecular sieves, 4Å (4Å MS) are a Brønsted basic zeolite known to facilitate alkoxide ligand exchange to metal centers.^[46,47] When omitting 4Å MS, no change in selectivity was observed (Scheme 2b, Entry 5). Collectively, results summarized in Scheme 2b support that the primary role of HFIP is to solvate the chloride anion of catalyst 12.^[48,49] We also note the substantial impact of 4Å MS on catalytic turnover. Pilot reactions formerly incorporating 4Å MS to minimize the deleterious effect of water afforded poor conversion of substrate 11a to (R,R)-14a at low catalyst loadings (e.g. below 15 mol%); however, omitting 4Å MS enabled a 10-fold decrease in catalyst loading from 30 mol% to 3 mol% (Table S8).^[34] This improvement suggests that the reactions using 4Å MS are likely impeded by competitive chelation and/or catalyst degradation by an in situ-generated HFIP alkoxide.

We next probed the impact of 4-pyrone sulfonate modifications on [4+2] enantioselectivity (Scheme 2c). Subjecting mesylated 4-pyrone 11b to conditions using the catalyst 12/HFIP system resulted in complete loss of enantioselectivity towards 14b (0% ee). Next, mesityl 11c and tripsyl 11d sulfonate derivatives were tested, with the greatest enantioselectivity observed for 11d →14d (92% ee).

Our experimental and computational results support the relevancy of DFT-optimized Complex I (Scheme 2d) as a plausible substrate 11a chelation mode for selectivity towards cycloadduct (R,R)-14a (Please see Section VI of the Supporting Information for further details). Complex I depicts bidentate chelation of the β-keto ester moiety of 11a to form a cis-beta complex, a rare binding mode for metal–salen complexes.^[39,50,51] Examination of Complex I suggests that modest enantioinduction may be imparted by steric encumbrance of the 3’-tert-butyl salen substituent, which blocks the Re face of 11a from diene approach. The geometry of a second higher energy 11a/catalyst conformer (Scheme 2d, Complex II, +10.45 kcal/mol) with a sterically encumbered si face was also optimized and features chelation of the β-keto sulfonate moiety to Cr(III). This higher energy complex may result from disfavored sulfonate chelation due to steric encumbrance imparted by the 3-tert-butyl salen substituent and is consistent with results reported in Scheme 2c. In addition to steric effects, the conjugated π systems of 4-pyrones 11a, 11c, and 11d may decrease the Lewis basicity of their respective sulfonates to disfavor chelation analogous to Complex II. This may also explain the significant increase in % ee between mesityl 14c and tripsyl 14d cycloadducts. Based on literature precedent, tripsyl sulfonates also exhibit C–H hydrogen bonds between ortho-isopropyl groups and their respective sulfonate oxygens which may further decrease Lewis basicity of the 11d sulfonate.^[52,53]

While we demonstrated proof-of-concept for our route and gram-scale synthesis using tripsyl sulfonate 11d, we proceeded in the synthesis towards 3 using cycloadduct (R,R)-14a due to late-stage route development using this compound. Furthermore, recrystallization efforts revealed that batches of (R,R)-14a (68% ee) could be enriched to 98% ee through preferential enrichment.^[54] This process was optimized to access gram-quantities of (R,R)-14a using commercially available catalyst 12. Our synthesis efforts next proceeded with an investigation of conditions to install the C14 methyl group of 3, a prerequisite for the C5 quaternary center, via selective 1,2-addition to (R,R)-14a (Scheme 3). Challenges encountered during reaction optimization towards 15a included suppression of 1,4-conjugate addition byproduct 15b and deconstructed byproduct 15c arising from a retro-aldol/retro-Michael elimination pathway (Tables S10–S12).^[34] These undesired outcomes were avoided by employing stoichiometric excesses of methyl magnesium bromide (MeMgBr) and anhydrous cerium trichloride (CeCl₃) to afford allylic alcohol 15a as a single diastereomer through convex addition. Important optimizations to achieve full conversion and selectivity to 15a was the order of reagent addition (Imamoto’s Method B),^[55] the rate of MeMgBr addition, and temperature. Fast addition of MeMgBr (as opposed to dropwise) was required to observe full conversion of starting material. Elevated temperatures resulted in higher conversion to 15c. Inverse substrate addition (Imamoto’s Method A), resulted in poor conversion to 15a and increased formation of 15b. These findings imply that chelation of (R,R)-14a to CeCl₃ is necessary for selective 1,2-addition, however, the gradually forming CeCl₃-Grignard complex is a softer nucleophile and less reactive in this context.^[56] Tertiary alcohol 15a was then reduced to 16 using lithium aluminum hydride (LiAlH₄); however, 16 was prone to decomposition and formation of bridged acetal 17 if left on the bench or on silica gel for extended periods. Accordingly, 16 was converted to the bench-stable bridged acetal 17 using pyridinium triflate (Pyr∙TfOH) as a mild acid. Conversion to bridged acetal 17 was also important for efficiency in the forthcoming transformation in contrast to a stepwise oxocarbenium addition/alcohol protection sequence. A modified, telescoped sequence to 17 was developed to minimize column chromatography and reaction times on multigram scale.^[34]

Scheme 3.

Completing of the total synthesis of (−)-verrucarol (3). Please see the Supporting Information for specific reaction conditions.

At this stage, installation of a C2_-pendant aldehyde was required to serve as a homologated precursor for C-ring construction and completion of the 6/6/5 (A/B/C) ring system of (−)-verrucarol 3. We envisioned that this intermediate could then serve as either an electrophile in an intramolecular aldol cyclization or a ketyl radical precursor to engage in a ketyl-olefin coupling. Pendant aldehyde intermediate 18 was obtained as a single diastereomer by inventing a one-pot silylative addition of ethyl vinyl ether (EVE).^[57,58] This reaction likely proceeds by initial acetal ring opening by triethylsilyl triflate (TESOTf) to furnish a protected primary alcohol and vinyl oxocarbenium (III), which is then intercepted by EVE followed by deprotonation of an oxonium species and extrusion of ethylene to afford 18.^[34]

With aldehyde 18 in hand, we investigated methods to form the C4–C5 connection and establish the key quaternary center. Three factors made this ring closure challenging: 1) 5-endo-trig annulations are classically disfavored,^[59] 2) the C5 quaternary center is highly congested, and 3) multiple enolizable protons could complicate anionic cyclization attempts. Colvin and coworkers demonstrated limited success in establishing this connectivity through a lactone reduction/intramolecular aldol cascade in their synthesis of (±)-trichodermin.^[60] Our initial efforts examined a detosylation/intramolecular aldol cascade. Under both nucleophilic and reductive detosylation conditions, only trace conversion to the desired intermediate ketone was observed, and further efforts to induce enolate cyclization were unsuccessful.^[61–63] Alternatively, we investigated an umpolung strategy via generation of a nucleophilic ketyl radical under single electron transfer (SET) conditions that would leverage the innate stability of the resultant sulfinate radical anion.^[64,65] Focusing efforts on this methodology, classic ketyl-generating conditions employing tributyltin hydride (Bu₃SnH) with radical initiators did not generate reactivity.^[66,67] The analogous tris(trimethylsilyl)silane (TTMSS) system afforded reduction of the aldehyde and protection as the silyl ether without cyclization.^[68] Photochemical and electrochemical ketyl-generating conditions were also non-productive.^[69,70] We ultimately found success utilizing excess samarium(II) iodide (SmI₂) to afford bridged ketone 19 as a single diastereomer (Tables S13 and S15).^[34] Overall, we demonstrate that the trichothecene core can be accessed in 11 steps, a significant advance from the 42-step route previously reported,^[28,29] while also providing an opportunity for expedited SAR development using a scaffold analogous to natural trichothecene congeners.

Several mechanisms by which SmI₂ could induce coupling reactions are possible including “carbonyl first,” “olefin first,” parallel single electron transfer (SET), and samarium-enolate pathways.^[71–73] We initially believed that a ”carbonyl first” pathway was operable for conversion of 18 to 19, as suggested by the precedent of Sarpong and coworkers.^[74] To investigate further, bridged acetal 17 was subjected to the optimized SmI₂-mediated coupling conditions which resulted in conversion to α-keto acetal 27 (Scheme 4a). Cyclic voltammetry (CV) of 17 and 18 with and without SmI₃ included as a stoichiometric additive was also performed (Scheme 4b). In the absence of SmI₃, 17 and 18 exhibit near identical onset reduction potentials (Figures S8 and S9, V ≈ −2.5 vs. Fc^+/0 in THF).^[34] Upon incorporation of SmI₃ as a stoichiometric additive, cathodic shifts (ΔV ≈ 0.125) for Sm(III) reduction were observed in the presence of both substrates and suggested chelation of Sm(III) as expected.^[75] We next investigated the effect of incorporating stoichiometric SmI₃ in the reductive cyclization of 18 to 19 based on the possibility of Sm autocatalysis and surface catalysis.^[76] This additive did not have a noticeable impact on reaction rate or isolated yield (Table S14).^[34] Furthermore, we propose that the pivotal 5-endo-trig annulation of 18 to 19 could proceed by either a “carbonyl first” mechanism (Scheme 4c, pathway 1) or by parallel SET (Scheme 4c, pathway 2); future kinetic studies may offer further insight.^[77]

Scheme 4.

a) Enol-tosylate reductive deprotection using optimized annulation conditions b) Cyclic voltammetry of compounds 17 and 18 in tetrahydrofuran (THF) c) Possible mechanisms towards compound 19.

Final hurdles to complete the synthesis included alkene transposition from C8–C9 to C9–C10 and C12 bridgehead epoxidation, which required protection of 19 via benzoylation to 20 (Scheme 3).^[78] Energy comparison of the trichothecene core olefin regioisomers (20 and 23) highlighted the challenge of direct alkene isomerization from the C8–C9 to C9–C10 position, as many isomerization methods favor thermodynamic products.^[79,80] Furthermore, previous trichothecene olefin isomerization efforts by Ziegler and coworkers using Ir(I) and Rh(III) catalysis resulted in the undesired C8–C9 olefin.^[81] Using substrate (20), we observed similar difficulties. Alkene isomerization under kinetic control unfortunately did not afford the desired isomerization.^[82,83] Accordingly, we investigated a hydration/elimination approach.^[84] Subjecting 20 to Mukaiyama hydration conditions afforded 21 and 22 as a 1:1 mixture of diastereomers.^[85,86] The subsequent elimination products using Martin’s sulfurane were dependent on the alcohol stereochemistry, in which the axial R-alcohol 21 furnished the requisite C9–C10 olefin 23 in excess (7.5:1, 23/24). Implementing identical conditions for dehydration of the equatorial S-alcohol 22 resulted in an excess of the C16 exo-olefin 24 (10:1, 24/20). These dehydration outcomes are consistent with a predominant E2 elimination mode and agree with results observed in Still and Tsai’s synthesis of (±)-trichodermol.^[87,88] The favorable elimination product distribution (21 → 23) can also be rationalized by hyperconjugative effects wherein the C10 axial proton of 21 has higher acidity relative to protons of the neighboring methyl group due to C10–H σ to C11–O σ* and C9–O σ* donation.^[89,90] Attempts to induce diastereoselective hydration using nitroarene conditions reported by Studer and coworkers afforded 21 and 22 as a 1:1 mixture.^[91] Syn-elimination via Burgess’ reagent on 22 also led to a trisubstituted alkene distribution in favor of 20.

For the final steps of the synthesis, 23 was subjected to a Wittig olefination/reductive deprotection sequence to furnish 26, which was primed for hydroxyl-directed epoxidation. We adapted biphasic methyltrioxorhenium (MeReO₃) conditions based on the Seiple group’s recent report, which afforded the desired epoxide as the major diastereomer.^[27,92] In addition to secondary alcohol chelation, other factors contributing to regio/diastereoselective epoxidation could be molecule concavity and the electron-poor nature of the C9–10 olefin. Following epoxidation, subjection of the silyl ether to tetrabutylammonium fluoride (TBAF) furnished the target compound (−)-verrucarol (3).

Conclusion

In summary, we have completed a concise asymmetric total synthesis of (−)-verrucarol (3) in 17 linear steps. The total synthesis was enabled by enantioselective [4+2] cycloaddition utilizing Jacobsen’s Cr(III)–salen catalyst 12. Initial studies have provided information on modulation of enantioselectivity by modification of the sulfonate moiety of a 4-pyrone ester. Cycloadduct (R,R)-14a was expediently enriched to a single enantiomer via preferential enrichment. Selective methylation of β-keto ester (R,R)-14a to 15a was enabled by modifications to Imamoto’s protocol in which order of reagent addition, rate of MeMgBr addition, and temperature were key optimization parameters. We also developed a one-pot procedure to synthesize the trichothecene core precursor 18 using silylative conditions to generate a vinyl oxocarbenium followed by addition of EVE as nucleophile. The 6/6/5 core was constructed diastereoselectively through a novel samarium (II) iodide-mediated 5-endo-trig annulation to afford 19 wherein both “carbonyl first” and parallel SET mechanisms are possible. Our synthesis of 19 demonstrates rapid, enantioselective access to the trichothecene core in 11 steps and provides a unique opportunity to develop trichothecene SAR from a scaffold containing a valuable C12 ketone handle. The total synthesis was completed using a selective hydration/elimination sequence to effect the required alkene transposition followed by a regio- and stereoselective epoxidation using MeReO₃. Ongoing studies will leverage this route to synthesize targeted trichothecene analogues and develop structure-activity relationships (SAR) for known biological targets of this natural product class.

Supplementary Material

The authors have cited additional references within the Supporting Information.^[93–126]

Data Availability Statement

The data that support the findings of this study, including reaction optimization tables, are available in the Supporting Information of this article.