Synthesis of resveratrol tetramers via a stereoconvergent radical equilibrium

See allHide authors and affiliations

Science  09 Dec 2016:
Vol. 354, Issue 6317, pp. 1260-1265
DOI: 10.1126/science.aaj1597

Catching a break in polyphenol synthesis

Chemical synthesis is usually rather different from playing with a modeling kit. If two large fragments of a molecule are not properly oriented, it is not typically possible to break them apart, rotate one, and then paste them back together. Yet that is precisely the trick that Keylor et al. used to synthesize two plant-derived polyphenols. Resveratrol forms a variety of dimers, trimers, and tetramers. When one central carbon-carbon bond links the fragments, it is weak enough to break spontaneously and reversibly at room temperature. The authors leveraged this equilibrium to generate an efficient route to two of the tetramers, nepalensinol B and vateriaphenol C.

Science, this issue p. 1260


Persistent free radicals have become indispensable in the synthesis of organic materials through living radical polymerization. However, examples of their use in the synthesis of small molecules are rare. Here, we report the application of persistent radical and quinone methide intermediates to the synthesis of the resveratrol tetramers nepalensinol B and vateriaphenol C. The spontaneous cleavage and reconstitution of exceptionally weak carbon-carbon bonds has enabled a stereoconvergent oxidative dimerization of racemic materials in a transformation that likely coincides with the biogenesis of these natural products. The efficient synthesis of higher-order oligomers of resveratrol will facilitate the biological studies necessary to elucidate their mechanism(s) of action.

Resveratrol (1), a naturally occurring and biologically important polyphenol (1), is widespread within the plant kingdom. It it serves as the progenitor to an arsenal of phytoalexins—antimicrobial defense compounds that accumulate rapidly at sites of pathogenesis to neutralize invading microorganisms and promote plant survival (2). Several resveratrol oligomers (dimers, trimers, and tetramers) have been shown, primarily via in vitro studies, to exert biological effects that transcend this natural role, including but not limited to anti-inflammatory, immunomodulatory, and cytotoxic activities (3). Unfortunately, the requirement for laborious extraction and purification of resveratrol oligomers from plant matter has imposed severe limitations on the extent to which their mechanism(s) of action can be elucidated. It is therefore imperative that synthetic advances be made in order to confirm or refute the biological activities ascribed to the isolated natural products; these efforts will identify chemical frameworks that hold potential as small-molecule chemopreventives and/or chemotherapeutics and enable structural modification for both structure-activity relationship (SAR) studies and the development of congeners with improved potency, efficacy, and bioavailability. Although the synthetic community has presented several innovative approaches to the resveratrol dimers (410), access to higher-order oligomers remains a serious challenge. In 2011, Snyder and coworkers reported a de novo synthetic approach to address this problem, accessing several higher-order oligomers through homologation of dimeric core structures (11). This strategy has since been used in two additional trimer syntheses (12, 13) and represents the only successful strategy to date for the preparation of these compounds.

Organisms that produce resveratrol (1) are able to harness the reactivity of delocalized phenoxyl radicals (such as 1•) (Fig. 1) generated upon its oxidation; the resultant oligomers are typically isolated as optically active materials (3). Although biomimetic approaches have been reported (1417), such remarkable levels of regio-, chemo-, and stereoselectivity have proven challenging to replicate in the laboratory because of the transient nature of the putative radical and quinone methide intermediates (Fig. 1, inset). We reasoned that if these intermediates could be rendered more persistent (18), then it would be possible to gain the advantages of efficiency offered by biomimicry without sacrificing the modularity offered by de novo synthetic approaches. Recently, we were able to recapitulate one mode of resveratrol oligomerization for the synthesis and antioxidant evaluation of two dimeric natural products, quadrangularin A (2) and pallidol (3) (Fig. 1) (19). The synthesis featured a remarkably persistent bis(p-quinone methide) intermediate 4a, similar to those (such as 5) invoked by Niwa, Pan, and coworkers in their studies on the structural elucidation and biogenesis of resveratrol trimers and tetramers from ε-viniferin (6) (14, 17). This biosynthetic logic can similarly be applied to higher-order oligomers, in principle providing access to the gamut of 8–8ʹ linked resveratrol tetramers (such as 7 and 8) through a convergent oxidative coupling of ε-viniferin (6) followed by regio- and/or stereodivergent cyclizations of the resultant bis(p-quinone methide) 5 (Fig. 1) (3).

Fig. 1 Synthetic design.

Merging de novo and biomimetic strategies allows for selective, scalable, and efficient preparation of resveratrol oligomers through the use of persistent radical and bis(p-quinone methide) intermediates. Bn, benzyl.

The proposed oxidative coupling presented several challenges. First, dimerization of a racemic form of 6 could, in principle, provide products derived from both homo- and heterodimerization between the (+)- and (−)-enantiomers; the (+)/(–) product is not represented in any known resveratrol oligomer. Furthermore, the dimerization reaction would generate two additional stereocenters, suggesting that the formation of six stereoisomers of product 5 is possible (fig. S9) (20). Last, it was unclear whether the intended double intramolecular cyclization of 5 would exhibit inherent preferences for specific regio- and stereochemical outcomes. Each hemisphere of tetramer 5 possesses two resorcinol moieties that are both capable of engaging each prochiral p-quinone methide in Friedel–Crafts cyclizations. The possibility for both symmetrical (2×5-exo-trig, 2×7-exo-trig, and 2×8-exo-trig) and unsymmetrical (5-exo-trig/7-exo-trig and 5-exo-trig/8-exo-trig) cyclization modes, each capable of producing several diastereoisomers (fig. S10) (20), posed a daunting challenge.

Here, we describe the execution of this strategy for the efficient biomimetic total synthesis of the resveratrol tetramers nepalensinol B (7) (21, 22) and vateriaphenol C (8) (23). Critical to the success of these efforts was the identification and rigorous characterization of an unconventional equilibrium between isolable dimeric (4a/b) and tetrameric (5a/b) bis(p-quinone methide) intermediates and their phenoxyl radicals (Fig. 1) (24), a physical property initially explored as a mechanistic curiosity but which we have found to have remarkable—and potentially biogenically relevant—implications for dynamic stereocontrol in the context of resveratrol oligomer synthesis and biosynthesis. Synthetic access to these natural products and their derivatives will enable further explorations of their already promising biological activities. For instance, nepalensinol B (7) is a potent inhibitor of topoisomerase II [median inhibitory concentration (IC50) = 0.02 μg/mL] (22); this is 3000 times more potent than etoposide (VP-16, IC50 = 70 μg/mL) (25), a clinically approved chemotherapeutic on the World Health Organization (WHO) model list of essential cytotoxic and adjuvant medicines (26).

During our studies toward resveratrol dimers (19), we discovered that intermediate 4a, which was isolated as a 4:3 mixture of meso:dl diastereomers, could be quantitatively isomerized to trans,trans-indane 9 (Fig. 2), a product that can only derive from meso-4a. Although it was tempting to conclude that epimerization of dl-4a was proceeding via tautomerization followed by stereorandom vinylogous protonation, independent preparation of the presumptive intermediates and their subjection to these reaction conditions did not lead to any detectable formation of 9 (fig. S7) (27). Thus, an alternate mechanism had to be responsible. In 1969, Becker reported that bis(p-quinone methides) similar to 4a equilibrate in chloroform solution at room temperature with the corresponding phenoxyl radicals through a homolytic C–C bond scission process analogous to that of Gomberg’s historic triphenylmethyl radical (24, 28). To probe whether such a mechanism could be operative for this transformation, we performed a thermal crossover experiment using differentially protected derivatives of 4, and our observations were consistent with the formation of a statistical mixture of homo- and cross-coupled products (fig. S8) (27). Intrigued by this unusual reactivity, we undertook an extensive analysis of the homolytic dissociation equilibrium of 4a and related derivatives.

Fig. 2 Diastereoconvergent cyclization of 4a.

TFA, trifluoroacetic acid.

Solutions of the bis(p-quinone methide) 4a in benzene yielded prominent electron paramagnetic resonance (EPR) spectra at room temperature (Fig. 3A). The spectrum is fully consistent with what is expected for the phenoxyl radical derived from 1a (hereafter 1a•) (Fig. 3B); the hyperfine coupling constants derived from the simulated spectrum are in good agreement with values from related compounds in the literature (29), as well as those predicted from the spin density distribution in 1a• calculated by using density functional theory (DFT) at the B3LYP/TZVP level of theory (Fig. 3C) (30, 31). Integration of the signals afforded Keq (1a•/4a) = 1.8 × 10–10 M (32). We recorded spectra at several temperatures between 10° and 50°C and used the corresponding equilibrium constants to provide an estimate of the thermodynamics of the homolysis-recombination process (Fig. 3D). Corresponding experiments were carried out by means of ultraviolet/visual (UV/vis) spectroscopy—with an expanded temperature range up to 85°C (33)—by following the increase in intensity of the low-energy absorption maximum at 414 nm (which was attributed to 1a• with ε = 79,000 M–1 cm–1, vide infra) as a function of temperature (Fig. 3E) (34). The measurements agree that the central C−C bond dissociation enthalpy (BDE) in 4a is 17.0 ± 0.7 kcal/mol. Although this is not the weakest C–C bond reported to date (the C–C BDE in the 4,4′–dimer of 2,6-di-tBu-4-methoxyphenoxyl is reported to be a mere 6.1 ± 0.5 kcal/mol) (35), it does afford a meaningful equilibrium at room temperature.

Fig. 3 Characterization of homolytic dissociation equilibrium of representative bis(p-quinone methide) 4a.

(A) Experimental (black) and fitted (red) EPR spectrum of 1a• recorded at 295 K. (B) Experimental (black) and calculated (red) hyperfine coupling constants for 1a•. (C) Spin-density distribution in 1a• predicted with DFT (B3LYP/TZVP). (D) Temperature dependence of equilibrium constants determined with EPR (black) and UV/vis (red) spectroscopy and corresponding calculated thermochemical parameters. The error bars represent standard deviations for three or more combined measurements. (E) Temperature dependence of absorbance corresponding to 1a•. (Inset) Full spectrum of 4a at 323 K. (F) Second-order decay of absorbance at 414 nm after nanosecond-pulsed irradiation of 4a at 308 nm under an atmosphere of either nitrogen (black) or oxygen (red).

To provide insight into the kinetics associated with this process, 4a was subjected to laser flash photolysis (LFP) with the 308-nm emission of a nanosecond-pulsed XeCl excimer laser, and the rates of recombination of the radicals were recorded. The transient species generated by photolysis exhibited the same low-energy absorption centered at 414 nm as in the spectrum of 4a (27). The decay of this absorbance could be fit to a second-order function by using the extinction coefficients determined from the UV/vis–EPR equilibrium experiments to afford the radical recombination rate constants, kr (Fig. 3F). Although 1a• features substantial spin density at C8 (resveratrol numbering) (Fig. 1)—enabling its dimerization at that position to produce 4a—the recombination rates were insensitive to the presence of O2. Rate constants in the absence and presence of oxygen were found to be within error (6.7 ± 1.5 × 107 M–1 s–1 and 7.4 ± 1.0 × 107 M–1 s–1, respectively), with the combined data set affording kr = 6.9 ± 1.4 × 107 M–1 s–1. The homolysis rate constant could be estimated from the equilibrium constant and recombination rate constant to be kf = 1.2 × 10−2 s–1.

The homolytic dissociation equilibrium held tremendous potential for the biomimetic preparation of higher-order resveratrol oligomers. Drawing on the similarity between the putative biogenesis of 8–8ʹ resveratrol dimers and tetramers (Fig. 1), we sought to realize a selective dimerization of ε-viniferin (6) (or a suitably substituted derivative). In their total synthesis of the resveratrol trimer caraphenol A, Snyder and Wright reported a highly effective eight-step preparation of aldehyde 10 (12), which we have leveraged for the present synthesis. This intermediate was converted into tBu-ε-viniferin derivative 6a almost exclusively as the (E)-isomer via Wittig olefination with phosphonium salt 11a (27) in 85% yield (Fig. 4A). With 6a in hand, we were poised to explore the key oxidative coupling reaction.

Fig. 4 13-Step total synthesis of resveratrol tetramers.

(A) Preparation of ε-viniferin derivatives, reagents, and conditions are as follows: iPr2NH, n-BuLi, THF, then 11a/b, then 10, –78°C to room temperature, then 0°C, TBAF, 0.5 hours, 85% for 6a, 80% for 6b. (B) Synthesis of nepalensinol B and vateriaphenol C, reagents and conditions are as follows: (a) THF, 0°C, then KHMDS, then FeCp2PF6, 97% for 5a, 63% (14% RSM) for 5b. (b) For 5a, BF3∙OEt2, CH2Cl2, –60° to –30°C (44% 12a + 9% 13/14a). For 5b, BF3∙OEt2, CH2Cl2, –78°C (59% 12b + 15% 13/14b). (c) 1. Pd/C (30 weight %), MeOH/EtOAc (1:1 v/v), H2 (1 atm), filter, concentrate, then 2. TFA (0.05 M) in CH2Cl2/MeNO2 (1:1 v/v), 75% for 7, 60% for 8/15. THF, tetrahydrofuran; TBAF, tetra-n-butylammonium fluoride; KHMDS, potassium hexamethyldisilazide; dr, diastereomeric ratio; SM, starting material; RSM, recovered starting material.

Despite concerns about stereo- and regioselectivity in the proposed transformations, our observations in the diastereoconvergent cyclization of 4a (vide supra) (Fig. 2) suggested that thermodynamic differentiation of the various diastereoisomers of 5—interconvertible via C–C homolysis-recombination—may afford some level of selectivity upon oxidative coupling, whereas the conformational requirements for productive orbital overlap may favor selected cyclization modes in the ensuing Friedel–Crafts reactions. Remarkably, subjection of 6a to our ferrocenium-mediated oxidative dimerization conditions afforded the desired tetrameric bis(p-quinone methide) 5a as nearly a single diastereoisomer (~19:1 major isomer:all other isomers)—derived from the coupling of two monomers of the same absolute configuration (Fig. 4B). Although stereoselectivity has been observed previously in biomimetic dimerizations of racemic precursors (3638), examples are rare and typically proceed via polar, irreversible mechanisms. Although it is possible that the stereochemical outcome of oxidative coupling of 5a is kinetically determined during the dimerization event, it is far more likely that initial coupling produces a mixture of diastereomers that rapidly equilibrate in solution via bond homolysis-recombination.

To support this hypothesis, we subsequently carried out analogous characterization of the 5a/6a• equilibrium as described above for 4a/1a•. Once again, EPR spectra consistent with 6a• were obtained from room-temperature samples of 5a (fig. S1, A to C) (27), and the temperature dependence of the equilibrium constants (fig. S1D; also determined via UV/vis spectroscopy, fig. S1E) (27) once again enabled the determination of the key C–C BDE in 5a to be 17.1 ± 0.4 kcal/mol—which is within error of 4a, as were the kinetics: kr = 2.0 ± 1.1 × 107 M–1s–1 (fig. S1F) and kf = 3.6 × 10−3 s–1. Given the overall similarity of the kinetics and thermodynamics of homolysis-recombination of tetramer 5a when compared with dimer 4a, the equilibration of 5a to nearly a single diastereomer was likely. Gratifyingly, exposure of 5a to BF3·OEt2 at –60°C followed by warming to –30°C furnished a mixture of just two regioisomeric cyclization products: 12a (44%, single diastereomer) and 13/14a (9%, 9:1 dr) (Fig. 4B). Each of these compounds derive from the trans,cisoid (S)/(S) [or (R)/(R)] diastereomer of 5a. This does not unequivocally demonstrate that this is the lowest-energy diastereomer of 5a, only that the cyclization of this diastereomer is favored over that from the (R)/(S) configuration. Structures 12a, 13a, and 14a represent the carbon skeletons of nepalensinol B (7) (21, 22), vateriaphenol C (8) (23), and hopeaphenol (15) (39), respectively. The stereochemical outcome of the Friedel–Crafts cyclization leading to major product 12a is complementary to that achieved by Snyder and coworkers through iterative homologation of the pallidol (3) core, which is capable of producing the stereoisomer ampelopsin H (11).

Global debenzylation of 12a and 13/14a via Pd/C–mediated hydrogenolysis proceeded in 45 and 60% yields, respectively (27). However, attempts at removal of the four remaining tert-butyl groups were unsuccessful under a variety of reaction conditions, resulting in decomposition to an intractable mixture. Although extensive investigation of this transformation may have eventually revealed less destructive conditions for dealkylation of the penultimate intermediates, the use of an isosteric functional group with increased lability that could nevertheless enforce regio- and diastereoselectivity upon oxidative coupling of 6 seemed to be a more attractive solution. Our initial thoughts focused on the use of trimethylsilyl (TMS) groups in place of the tBu moieties. To our surprise, given the vast literature on the chemistry of hindered phenols and phenoxyl radicals, the persistence of 2,6-di-TMS-phenoxyl radicals had yet to be investigated. High accuracy CBS–QB3 quantum chemical calculations (40, 41) predicted that the O–H BDEs in 2,6-di-TMS-4-methylphenol (S2) and 2,6-di-tBu-4-methylphenol (BHT) were 80.7 and 78.6 kcal/mol, respectively [the O–H BDE in BHT has been determined experimentally to be 81.0 kcal/mol (42)]. These calculations suggest that the electronic effects of TMS and tBu groups on the thermodynamic stability of phenoxyl radicals are similar. Moreover, the reactivity of S2 toward peroxyl radicals in inhibited autoxidations of cumene was essentially indistinguishable from that of BHT (k = 1.4 × 104 versus 2.1 × 104 M–1 s–1, respectively) (fig. S6) (27), suggesting that the kinetics of the reactions of the silylated phenol and phenoxyl radicals would be similar to that of their t-butylated counterparts.

Accordingly, 1b (Fig. 1) could be readily oxidatively dimerized to the corresponding bis(p-quinone methide) dimer 4b. Crystals of the meso isomer of 4b suitable for x-ray analysis were obtained (Fig. 2) (27), confirming its identity as a bis(p-quinone methide) and seemingly revealing a conformational preference for antiparallel alignment of the carbonyl moieties, likely resulting from the combined influence of sterics and dipole minimization. Like its tert-butylated predecessor, 4b yielded a prominent EPR spectrum at room temperature that was consistent with the phenoxyl radical derived from 1b (fig. S2, A to C) (27). Efforts to determine the temperature dependence of the 4b/1b• equilibrium were limited by the fact that 4b was not sufficiently persistent to obtain reproducible spectra above 50°C, instead undergoing an unusual rearrangement to a derivative of the natural product δ-viniferin (27). Nevertheless, a Van’t Hoff plot of the available data between 10° and 50°C afforded an estimate of the C–C BDE of 16.4 ± 0.5 kcal/mol (fig. S2, D and E) (27), which is only slightly lower (and within error) of that in 4a. The recombination rate constant was determined with laser flash photolysis of 4b to be kr = 5.3 ± 3.0 × 108 M–1s–1 (fig. S2F), which is almost one order of magnitude faster than that obtained for 4a and explains the slightly less favorable equilibrium, given the similar values of kf (2.1 × 10−2 and 1.2 × 10−2 s–1 for 4b and 4a, respectively).

Likewise, the TMS derivative of ε-viniferin, 6b, proved to be a competent substrate for FeCp2PF6-mediated oxidative dimerization (Fig. 4B), and the resultant bis(p-quinone methide) 5b was found to undergo the same homolysis-decomposition as that of 4b (fig. S3) (27). Cyclization with BF3∙OEt2 at –78°C cleanly afforded 12b (59%, single diastereomer) and 13/14b (15%, 9:1 dr), respectively. The increased reactivity of the silylated derivative 5b was evident because it was less stable to silica gel chromatography and could be cyclized at a lower temperature than could the tBu analog. Cyclization at slightly elevated temperatures (–60°C) afforded 12b and 13/14b in a 1.3:1 regioisomeric ratio, suggesting that the current level of product selectivity is sensitive to conformational effects. Furthermore, allylic strain minimization before 7–exotrig cyclization would predict the hopeaphenol isomer 14b to predominate, and therefore the observed preference for formation of the vateriaphenol C core 13b suggests a likely contribution of transannular strain and/or conformational restriction in controlling facial selectivity during attack of the prochiral p-quinone methides. Hydrogenolysis of 13/14b followed by protodesilylation afforded vateriaphenol C (8) and hopeaphenol (15) in 60% yield over two steps as a 9:1 mixture of diastereoisomers. Subjecting 12b to the same reaction sequence provided nepalensinol B (7) in 75% yield over two steps.

The total syntheses of resveratrol tetramers nepalensinol B (7, 5.1% overall yield) and vateriaphenol C (8, 1.1% overall yield) described here required only 13 linear synthetic steps. Critical to our strategy was the application of thermodynamic stereocontrol in the dimerization of persistent free radicals, a process that we have extensively characterized. The efficiency of the route has enabled the preparation of sufficient quantities of material that the biological activities of these natural products can now be more thoroughly evaluated.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S10

Tables S1 to S8

References (4350)

References and Notes

  1. A side by side comparison of these structures is provided in the supplementary materials.
  2. Materials and methods are available as supplementary materials on Science Online.
  3. The reduction by three orders of magnitude relative to Becker’s equilibrium constant likely results from inductive withdrawal of electron-density in 1a• by the two meta-benzyloxy ether substituents on the resorcinol ring relative to the simple phenyl ring in Becker’s example.
  4. Considerable decomposition was observed for experiments conducted at or above 90°C.
  5. The experimental extinction coefficient is in good agreement with Becker’s unsubstituted radical (24) at 75,000 M–1 cm–1.
Acknowledgments: Financial support is gratefully acknowledged from the NIH–National Institute of General Medical Sciences (grant R01-GM096129), the Camille Dreyfus Teacher Scholar Award Program, and the University of Michigan to C.R.J.S., and the National Sciences and Research Council of Canada, the Canada Research Chairs program, and the University of Ottawa to D.A.P. J-P.R.C. acknowledges the support of Ontario Graduate Scholarships. We thank J. W. Kampf for conducting x-ray diffraction experiments and solving the structures of compounds 4b and 9. Atomic coordinates and structure factors for the crystal structures reported are available free of charge from the Cambridge Crystallographic Database under accession nos. CCDC 1499423 and CCDC 1485815, respectively. Additional characterization data are available in the supplementary materials. Reprints and permissions information is available at The authors declare no competing financial interests. Readers are welcome to comment on the online version of the paper. Correspondence and requests for materials should be addressed to C.R.J.S. ( or D.A.P. (
View Abstract

Navigate This Article