Asymmetric synthesis of batrachotoxin: Enantiomeric toxins show functional divergence against NaV

See allHide authors and affiliations

Science  18 Nov 2016:
Vol. 354, Issue 6314, pp. 865-869
DOI: 10.1126/science.aag2981

Pluses and minuses of BTX behavior

Batrachotoxin is a potent neurotoxin produced by the endangered Colombian poison dart frog and is an agonist of voltage-gated sodium ion channels (NaVs). Logan et al. developed a chemical synthesis of this molecule, denoted (−)-BTX, by taking advantage of a tin hydride–mediated radical cyclization to stitch together the polycyclic framework. Using an analogous route, they also prepared the non-natural mirror image, (+)-BTX. Conversely to the natural product, (+)-BTX antagonized NaVs.

Science, this issue p. 865


The steroidal neurotoxin (−)-batrachotoxin functions as a potent agonist of voltage-gated sodium ion channels (NaVs). Here we report concise asymmetric syntheses of the natural (−) and non-natural (+) antipodes of batrachotoxin, as well both enantiomers of a C-20 benzoate–modified derivative. Electrophysiological characterization of these molecules against NaV subtypes establishes the non-natural toxin enantiomer as a reversible antagonist of channel function, markedly different in activity from (−)-batrachotoxin. Protein mutagenesis experiments implicate a shared binding side for the enantiomers in the inner pore cavity of NaV. These findings motivate and enable subsequent studies aimed at revealing how small molecules that target the channel inner pore modulate NaV dynamics.

The phenotypic effects of acute poisons found among the rich pharmacopeia of terrestrial and marine life have been documented from antiquity. Isolation and characterization of toxic compounds have made available important chemical reagents for studying complex biochemical circuits (1). Studies of this type have revealed a large number of peptide and small-molecule agents that target voltage-gated sodium ion channels (NaVs), an obligatory class of membrane proteins for bioelectrical signaling (14). Among the collection of known NaV modulators are three structurally related agents, (−)-batrachotoxin [(−)-BTX], veratridine, and aconitine (Fig. 1A)—sterically large, lipophilic amine derivatives believed to share a common binding locus in the inner pore region of NaV (3) (site 2, Fig. 1B). The influence of these toxins on ion gating, however, differs distinctly. On one extreme, (−)-BTX, the primary toxic constituent of Colombian poison dart frogs (genus Phyllobates), is a full NaV agonist, causing the channel to open more readily at hyperpolarized membrane potentials and blocking fast inactivation (among other characteristic effects) (35). Conversely, the activities of veratridine and aconitine are best described as partial agonism and inhibition of channel function, respectively (5). Despite recent insights from structural biology into the three-dimensional architecture of prokaryotic NaVs (69), a molecular understanding of the influence of the site 2 toxins on ion conduction and ion gating kinetics is lacking. Toxin structure-activity studies, in combination with protein mutagenesis experiments, can address questions related to the dynamical nature of channel function and may guide the rational design of small-molecule modulators of NaV activity (1). The potency of (–)-BTX (10), its storied history as the archetypal small-molecule site 2 probe (4), and its unparalleled effects on channel gating render it an optimal “lead” compound for such investigations.

Fig. 1 Background and synthetic plan.

(A) The structures of lipophilic site 2 toxins (−)-batrachotoxin (BTX), aconitine, and veratridine. (B) A pore model of NaV with (−)-BTX (depicted as spheres) docked at site 2. The structure is based on Magnetococcus marinus NaV crystallographic data (Protein Data Bank accession code 4F4L) (9, 19). Domain I, orange; domain II, red; domain III, gray; domain IV, teal. (C) Retrosynthetic analysis of BTX and BTX C-20 ester analogs. LD50, half-maximal lethal dose; Me, methyl; tBu, tert-butyl; Et, ethyl; TBS, tert-butyldimethylsilyl.

(−)-BTX binding to NaVs alters every aspect of channel function, resulting in a hyperpolarized shift in the voltage dependence of activation, inhibition of both fast and slow inactivation, a decrease in single-channel conductance, and reduction of ion selectivity (3, 4). The utility of this natural product as a NaV activator has led to a substantial depletion in the world supply, which once exceeded 1 g but was less than 170 mg as of 2009 (11, 12). Since the toxin was first isolated in 1963 by Märki and Witkop from poisonous frogs collected in the northern rain forest of Colombia (13), Phyllobates has been placed on the endangered species list, and thus collection of natural (−)-BTX from this source is restricted. (−)-BTX has also been identified in select species of birds (genus Pitohui and Ifrita) (14) and beetles (genus Choresine) (15), but only in small quantities (e.g., ~1.8 μg of (−)-BTX per beetle). Although semi- (16) and racemic syntheses (17) of BTX-A (Fig. 1C), a compound lacking the C-20 pyrrole ester, have been disclosed, the length of each of these works (>45 linear steps) precludes the facile production of (−)-BTX or select analogs. Accordingly, our desire to use BTX and modified forms thereof for examining channel dynamics and ion gating mechanisms has motivated our efforts to obtain the natural product through de novo synthesis.

Retrosynthetic analysis of (–)-BTX led us to outline a plan that would enable late-stage assembly of the homomorpholine E ring and elaboration of the C-20 allylic ester (Fig. 1C), thereby facilitating access to modified forms of the toxin. Previous structure-activity relationship studies using a small number of semisynthetic BTX derivatives (10, 18) and C/D/E-ring BTX analogs (19) revealed the importance of the C-20 ester, tertiary amine, and tetracyclic skeleton for NaV agonist activity. Unraveling BTX-A exposes a steroid-like frame 1, the assembly of which is confounded by two angular groups at C/D-ring junction, the C-11 exo-methylene and the C-8/C-9 alkene. To maximize convergence in our synthetic plan, we conceived a disconnection strategy for 1 across the C ring. This idea would reduce the problem of constructing 1 into two fragments, one expressing the A/B-ring system (3, 20) and a second comprising the D-ring cyclopentane (4, 21). The successful execution of this scheme could produce the desired toxin through a sequence of linear steps totaling no more than 20 to 25.

Our synthesis of (−)-BTX commenced with the coupling of methylenecyclopentanone 4 (21) (fig. S1A) and vinyl bromide 3, accessed from (S)-(+)-Hajos-Parrish ketone through a modified sequence of steps originally outlined by Parsons and co-workers (20) (fig. S1B). Conjoining fragments 3 and 4 to generate the linked A/B/D-tricycle 5 presented the first in a series of process development challenges. An initial attempt to effect this transformation involved Li-Br exchange of 3 with n-BuLi (Bu, butyl) and sequential addition of enone 4. Although 5 was delivered under these conditions, product yields never exceeded 30%. Deuterium quenching experiments with D2O validated our hypothesis that α-deprotonation of 4 was competitive with the desired ketone addition pathway. Transmetalation reactions of the vinyl-lithium species with ZnCl2, ZnBr2, MgBr2•OEt2 (Et, ethyl), CeCl3, Yb(OTf)3 (Tf, trifluoromethansulfonate), CeCl3•2LiCl, and LaCl3•2LiCl were examined, but none of these measures proved effective (22, 23). The addition of one equivalent of anhydrous LiBr to the reaction media of 3 improved the coupling efficiency by >20% (24). Following this lead, an optimized protocol using 2.1 equivalents of t-BuLi, which presumably generates one equivalent of LiBr in situ, reproducibly afforded 5 as a single diastereomer in 65% yield on a multigram scale. The ease of synthesis of this material and its desilylated form 6 enabled subsequent efforts to identify conditions for tandem annulation of the C ring and installation of the quaternary C-13 center.

An evaluation of available methods for ring closure of 1,6-enynes led us to consider radical-initiated processes (25). Under such conditions, an incipient C-13 3° radical could be intercepted to forge the angular aminomethylene unit (or a suitable surrogate). Efforts to first examine C-ring formation on 6, however, revealed the potential fallacy of this plan. Using n-Bu3SnH and triethylborane (Et3B) to promote the cyclization event resulted in the generation of two isomers, 7 and 8, in a 1:5 ratio favoring the undesired product (Fig. 2A). Studies by Stork and Beckwith and co-workers have demonstrated that substrate concentration and reaction temperature can influence the mode of cyclization (i.e., 5-exo-trig versus 6-endo-trig) in radical-mediated enyne reactions (26, 27). At elevated temperature (130°C) and with fivefold dilution of 6, a reversal in selectivity was observed, affording a slight excess of the desired tetracycle (1.3:1 ratio of 7 to 8; Fig. 2A). The combined product yield of this transformation exceeded 90%, thus encouraging further exploration of this chemistry, despite the modest selectivity results.

Fig. 2 Enyne radical cyclization to furnish the steroidal core of BTX.

Reagents, conditions, and product yields for steps a to p are as follows: (A) a, t-BuLi, THF, −90°C, then 4 (see Fig. 1) (65%); b, K2CO3, MeOH (94%); c, Et3B, air, n-Bu3SnH. (B) d, Me3SiC≡CSiEt2Cl, imidazole, CH2Cl2 (93%); e, O2, n-Bu3SnH, Et3B, Ph2O, 150°C (75%); f, n-Bu4NF, THF, 60°C (94%); g, 2-iodoxybenzoic acid, t-BuOH, 65°C, then OsO4 [7 mol %], NaIO4, pyridine, H2O (57%); h, MeNH2, CH2Cl2; NaB(O2CCF3)3H, CH2Cl2, −78°C, then ClCH2COCl, 2,6-lutidine, −78 to 0°C (52%); i, NaOEt, EtOH, 1:1 THF/C6H6 (92%); j, KN(SiMe3)2, PhNTf2, THF, −78 to 0°C (94%); k, CuCl2, O2, 1,4-dioxane, 73°C (85%); l, NaClO2, NaH2PO4, dimethyl sulfoxide/H2O; m, SOCl2, pyridine, CH2Cl2; n, NaN3, acetone/H2O; o, aqueous AcOH, 1,4-dioxane, 90°C (57% over four steps); p, p-TsOH, 4-Å molecular sieves, p-methoxyphenethyl alcohol (PMBCH2OH), C6H6 (89%). THF, tetrahydrofuran; Ph, phenyl; Tf, trifluoromethansulfonate; Ts, p-toluenesulfonate; Ac, acetate.

Repeated attempts to capture the intermediate C-13 radical with oxime and hydrazone derivatives generated from formaldehyde failed to deliver the expected aminomethylation product (28). Forced to consider alternative solutions, we recognized that a modified silyl ether group appended from the neighboring C-14 alcohol would be aptly positioned to intercept the 3° radical (29). Based on available precedent, an alkynylsilyl chloride, Me3SiC≡CSiEt2Cl (Me, methyl), was selected for modification of the C-14 alcohol in 6 (30, Fig. 2A). Treatment of the resulting silylether (9) with n-Bu3SnH and Et3B at 150°C resulted in a cyclization cascade to give pentacycle 10 as the exclusive product (31). Within the limits of proton nuclear magnetic resonance (1H NMR) detection, none of the corresponding five-membered C-ring isomer was generated in this process. Our preliminary efforts to understand the role of C-14 substituent groups on reaction selectivity suggest that silyl protection of the alcohol (along with the elevated reaction temperatures) favors 6-endo-trig ring closure. Although additional studies are warranted to appreciate these structure-selectivity data, our enyne cyclization cascade offers a convergent approach for synthesizing substituted steroid scaffolds and should facilitate access to a wide range of such compounds.

Close inspection of the radical cyclization products derived from either 6 or 9 revealed an unexpected outcome pertaining to the structure of the resulting organostannane moiety (Fig. 2, A and B). Carbostannylation of the alkyne group should afford a vinyl-tin product, as noted in the reaction of 6. Unexpectedly, when 9 was subjected to the reaction conditions, allylstannane 10 was the sole product, a result confirmed by both NMR and x-ray crystallography. Formation of allylstannane 10 can be rationalized through a mechanism involving 1,4-H-atom transfer of an intermediate vinyl radical (32) (Fig. 2B), a proposal supported by a deuterium labeling experiment (fig. S5). Although this result was unplanned, the efficiency and selectivity of the cyclization reaction compelled our decision to advance this material. Looking forward, the versatility of the allylstannane group should serve future efforts to prepare C ring–modified BTXs.

The availability of 10 in nine steps from the Hajos-Parrish ketone enabled the production of substantial quantities of material to complete the target synthesis. Excision of the bridging silyl ether in 10 was accomplished with excess n-Bu4NF, revealing a diol intermediate that was subsequently advanced to 11 through 2-iodoxybenzoic acid–mediated alcohol oxidation and chemoselective vinylsilane cleavage (57%; Fig. 2B). Conversion of aldehyde 11 to chloroacetamide 12 was performed by following a three-step, single-flask sequence (16, 33). From 12, efficient closure of the homomorpholinamide ring with NaOEt (92%) (17) provided a versatile intermediate for modification of both the C- and D-ring units. The latter could be achieved through the D-ring enol triflate, prepared using KN(SiMe3)2 and PhNTf2.

Introduction of the C-11α alcohol from the C-ring allylstannane 13 presented one of the more difficult challenges in our approach to BTX (Fig. 2B). Although protodestannylation of 13 to generate the corresponding C-11 exo-methylenecyclohexane had limited success (34), all subsequent attempts to oxidize this compound to ketone 15 (i.e., O3, OsO4, and RuO4) failed to give product. Inspired by a report from Kim and Fuchs, we attempted to convert 13 to the corresponding allyl chloride by using CuCl2 (35). Fortuitously, conducting this reaction in dioxane under aerobic conditions delivered enal 14 in 85% yield with only a minor amount of the chlorinated product (~10%). Although the mechanistic details of this transformation remain unclear, we are aware of only one other documented example of such an oxidation reaction, which uses a vanadium catalyst and O2 (36). Enal 14 is suitably disposed for conversion to C-11 ketone 15 by following a series of functional group interconversion steps highlighted by a Curtius rearrangement (37). The absence of a viable chromophore on batrachotoxinin A (BTX-A) makes purification of this material difficult; accordingly, in the sequence leading to 15, the C-3 methoxy acetal was exchanged with p-methoxyphenethyl alcohol.

Completion of the carbon skeleton of (–)-BTX was accomplished through a palladium-catalyzed cross-coupling of tributyl(1-ethoxyvinyl)tin to vinyl triflate 15 (Fig. 3A) (38). In situ hydrolysis of the incipient enol ether with 1 M oxalic acid supplied enone 16 (77%). Following an extensive screen of reducing agents, successful stereoselective global reduction of enone 16 was accomplished in 33% yield by treatment with freshly prepared AlH3 (39). We hypothesize that the Lewis-basic lactam (or a reduced form) acts as a pivotal stereocontrolling element, as treatment of enone 15 with alternative hydride reducing agents [e.g., AlH3•NMe2Et, NaBH4, NH3•BH3, (S)-Me-CBS-oxazaborolidine/BH3, or l-selectride] delivered the undesired C-11β alcohol exclusively. The use of AlH3 also favored generation of the correct C-20 allylic alcohol epimer, a sterochemical outcome that can be rationalized through a model invoking chelation control (38). Deprotection of the product from AlH3 reduction under acidic conditions afforded (–)-BTX-A in 83% yield (17). Finally, by employing a modification of Tokuyama, Daly, and Witkop’s (–)-BTX-A acylation protocol with the mixed anhydride prepared from ethyl chloroformate and 2,4-dimethyl-pyrrole-3-carboxylic acid (10), the synthesis of 2 mg of (–)-BTX was completed (79%, 0.25% overall yield) in 24 steps from (S)-(+)-Hajos-Parish ketone. The product was identical in all respects [as assessed by high-resolution mass spectrometry, thin-layer chromatography, and high-performance liquid chromatography (HPLC) coinjection] with a sample of the natural material and with previously recorded spectroscopic data (40, 41). Our synthetic plan also enabled milligram-scale preparation of the unnatural toxin antipode, (+)-BTX, the known benzoate ester of (–)-BTX-A (BTX-B; Fig. 3B) (42, 43), and the enantiomer of this compound (ent-BTX-B).

Fig. 3 Completion of the synthesis.

Reagents, conditions, and product yields for the preparation of (A) (–)-BTX (steps q to t) and (B) BTX-B and its enantiomer (steps q to s, then u) are as follows: q, LiCl, CuCl, Pd(PPh3)4, tributyl(1-ethoxyvinyl)tin, THF, 60°C, then 1 M oxalic acid, 0°C (77%); r, AlH3, THF, −78 to 0°C (33%); s, p-TsOH, 3:2 acetone/H2O (83%); t, (ethyl carbonic)-2,4-dimethyl-1H-pyrrole-3-carboxylic anhydride, Et3N, C6H6, 45°C (79%); u, benzoic (ethyl carbonic) anhydride, Et3N, C6H6, 45°C (70%).

Electrophysiological characterization of synthetic (–)-BTX and BTX-B against rat NaV1.4 (rNaV1.4) confirmed that the latter also functions as an agonist and is similar in potency to the natural product (fig. S6 and table S12). Previous reports and our own studies indicate that the ester group of BTX-B is more stable than the oxidatively sensitive acylpyrrole of BTX; thus, additional experiments were performed with the former compound (42, 43). Synthetic BTX-B was tested against a subset of representative NaV isoforms including rNaV1.4, human NaV1.5, and human NaV1.7. Application of 10 μM BTX-B to Chinese hamster ovary cells expressing a single NaV subtype resulted in sustained sodium current in all cases (Fig. 4A and figs. S7 and S8). Use-dependent agonism of NaV isoforms by BTX-B prevented steady-state inactivation of >80% of the sodium channel population (Fig. 4A and fig. S8). BTX-B also induced a characteristic hyperpolarizing shift (−44.9 to −51.5 mV) in the half-maximal voltage (V0.5) of activation of wild-type NaV isoforms (Fig. 4B and table S13). The similarity of these data is consistent with the high protein sequence conservation between NaV subtypes in the inner pore–lining S6 helices that form the putative toxin binding site (fig. S9).

Fig. 4 Effects of synthetic BTX-B and ent-BTX-B on wild-type rat NaV1.4 function.

(A) Representative trace for rat NaV1.4 (rNaV1.4) current before (black) and after (red) steady-state binding of 10 μM BTX-B. Current was evoked by a 150-ms test pulse from –120 to 0 mV after establishment of steady-state inhibition by repetitive depolarizing pulses to 0 mV. (B) Voltage dependence of activation for rNaV1.4 in the presence of 10 μM BTX-B (open circles) compared with control conditions (filled circles) for n ≥ 3 cells (mean ± SD). (C) Representative trace for rNaV1.4 current before (black) and after (red) steady-state binding of 5 μM ent-BTX-B. Current was evoked by a 24-ms test pulse from –120 to 0 mV after establishment of steady-state inhibition by repetitive depolarizing pulses to 0 mV. (D) Voltage dependence of activation for rNaV1.4 in the presence of 10 μM ent-BTX-B (open circles) compared with control conditions (filled circles) for n ≥ 3 cells (mean ± SD). (E) rNaV1.4 homology model highlighting residues that have previously been shown to abolish (–)-BTX activity. (F) Percent current inhibition of rNaV1.4 mutants by 5 μM ent-BTX-B (mean ± SD). WT, wild type; F, phenylalanine; K, lysine; L, leucine; N, asparagine.

Following earlier work from our laboratory (19) and others (44, 45), we questioned whether the enantiomer form of BTX would bind with high affinity to NaV with analogous functional effects. Such a question can only be answered with the availability of a de novo synthesis of the toxin. Accordingly, electrophysiological recordings with ent-BTX-B were performed against rNaV1.4. These data revealed ent-BTX-B to be a use- and state-dependent channel antagonist, with a measured half-maximal inhibitory concentration of 5.3 ± 0.6 μM [Fig. 4C and fig. S10; (+)-BTX also displays antagonistic activity (fig. S11)]. The concentration for half-maximal inhibition of NaV by ent-BTX-B is similar in magnitude to the half-maximal effective concentration for BTX-B agonism (1.0 ± 0.1 μM; fig. S10) measured under identical conditions. Notably, unlike the natural antipode, ent-BTX-B binding caused only a minimal shift in the V0.5 of activation and the V0.5 of steady-state inactivation (table S14). In addition, channel block was fully reversible by this inhibitor.

To determine whether BTX-B and ent-BTX-B share an overlapping binding site within the inner pore region of NaV, ent-BTX-B was tested against five rNaV1.4 single-point mutants that have been shown previously to destabilize BTX binding (Fig. 4, E and F, and fig. S12) (19). Mutation of N434 (46), L1280 (47), F1579 (48), and N1584 (48) to lysine resulted in a ~3- to 30-fold decrease in current block by 5 μM ent-BTX-B. Against F1236K (49), however, ent-BTX-B retained significant activity (33.6 ± 2.1% current inhibition). The evident difference between ent-BTX-B and BTX-B indicates an overlapping, but nonidentical, binding region within the inner pore cavity. It would seem that the open channel is sufficiently large to accommodate lipophilic tertiary amine ligands (19, 45, 50). Subtle alterations in the binding pose of these ligands appear to dramatically alter the functional response of the protein. Future investigations will be directed at delineating the precise channel-toxin interactions that distinguish activation from inhibition by BTX derivatives and related lipophilic toxins.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Tables S1 to S14

References (5158)

References and Notes

  1. A 10-μg sample of authentic (−)-BTX was purchased from Santa Cruz Biotech.
Acknowledgments: We are grateful to M. Maduke (Stanford University) for generous use of her laboratory space and equipment. We thank S. Lynch (Stanford University) for assistance with NMR experiments and analysis, Y. Kishi (Harvard University) for graciously providing NMR spectra of synthetic BTX-A, J. K. Maclaren (Stanford Nano Shared Facilities) for solving the crystal structure of 11 (supported by the NSF under award ECCS-1542152), G. Dick (Stanford University) for assistance with HPLC coinjection of natural and synthetic BTX, and the Vincent Coates Foundation Mass Spectrometry Laboratory, Stanford University Mass Spectrometry ( Metrical parameters for the structure of compound 11 are available free of charge from the Cambridge Crystallographic Data Centre under reference number CCDC-1509206. This work was supported in part by the NIH (R01NS045684) and by gifts from Pfizer and Amgen. T.T. was sponsored as a Japan Society for the Promotion of Science Fellow for research abroad. R.T.-T. is a NSF predoctoral fellow. M.M.L. and T.T. contributed to the synthesis of BTX, and R.T.-T. was responsible for electrophysiology experiments. The manuscript was prepared by M.M.L., R.T.-T., and J.D.B. J.D.B is a cofounder of and owns equity shares in SiteOne Therapeutics, a pharmaceutical startup company aimed at developing sodium channel subtype–selective inhibitors as antinociceptive agents.

Stay Connected to Science

Navigate This Article