|Home | About | Journals | Submit | Contact Us | Français|
The myrmicarins are a family of air and temperature sensitive alkaloids that possess unique structural features. Our concise enantioselective synthesis of the tricyclic myrmicarins enabled evaluation of a potentially biomimetic assembly of the complex members via direct dimerization of simpler structures. These studies revealed that myrmicarin 215B undergoes efficient and highly diastereoselective Brønsted acid-induced dimerization to generate a new heptacyclic structure, isomyrmicarin 430A. Mechanistic analysis demonstrated that heterodimerization between myrmicarin 215B and a conformationally restricted azafulvenium ion precursor afforded a functionalized isomyrmicarin 430A structure in a manner that was consistent with a highly efficient, non-concerted ionic process. Recent advancement in heterodimerization between tricyclic derivatives has enabled the preparation of strategically functionalized hexacyclic structures. The design and synthesis of structurally versatile dimeric compounds has greatly facilitated manipulation of these structures en route to more complex myrmicarin derivatives.
The myrmicarins are a family of structurally fascinating alkaloids found in the poison gland secretion of the African ant species Myrmicaria opaciventris (Figure 1).1,2 Of the alkaloids produced by ants of the myrmicinae group, the higher molecular weight myrmicarins display an unprecedented level of complexity.3 In addition to challenges presented by their elaborate molecular architectures, the isolation and structure elucidation of the complex myrmicarins has been complicated by their air sensitivity.4 While the bicyclic and tricyclic myrmicarins can be purified chromatographically, myrmicarin 663 (M663) is the only complex member that has been isolated and characterized as a single component.5 By contrast, the extreme sensitivity of myrmicarin 430A (M430A) necessitated its characterization as a mixture with the tricyclic myrmicarins in a sample taken directly from the poison gland. Structural assignments for both M430A and M663 have been achieved using a combination of phase-sensitive two-dimensional NMR techniques. Although submilligram samples of myrmicarin 645 (M645) have been obtained, the scarcity and oxidative sensitivity of this compound have prevented assignment of its relative stereochemistry.1,6 Interestingly, while another alkaloid with a molecular weight of 430 was identified during mass spectrometric analysis of the poison gland secretion, no structural information on this compound was obtained.
The fascinating structures of these alkaloids, combined with the opportunities and challenges associated with their sensitivity and chemistry, inspired us to develop methods for their total synthesis. Specifically, we envisioned a concise assembly of the highly sensitive complex members through potentially biomimetic dimerization or trimerization of the tricyclic members. Herein we report our studies in this area, encompassing a succinct enantioselective synthesis of the tricyclic myrmicarins and the discovery of an unprecedented, highly efficient vinyl pyrroloindolizine dimerization to yield heptacyclic products. Mechanistic analysis of this dimerization using functionalized pyrroloindolizine derivatives revealed that the corresponding heptacyclic products are formed in a highly efficient, non-concerted ionic process. This unique reactivity has enabled us to employ strategic heterodimerization of functionalized pyrroloindolizines to prepare versatile dimeric structures for elaboration to more complex myrmicarin derivatives.
Shortly after their isolation of the tricyclic myrmicarins, Schröder and Francke reported the racemic synthesis of myrmicarin 217 (M217, Figure 1) through dehydrative cyclization of an unsaturated derivative of myrmicarin 237B (M237B, Figure 1).7 This result was consistent with a proposal that the complex myrmicarins may arise through oligomerization of bicyclic structures (Figure 2).8
As an alternative to this biosynthetic hypothesis, we proposed that the complex myrmicarins may originate via dimerization or trimerization of the tricyclic members. Specifically, M430A may arise directly from dimerization of M215B (Figure 3).9 According to this proposal, we envisioned three potentially biomimetic dimerization strategies for the formation of M430A (Figure 4).10 In the first we envisioned an ionic mode of dimerization involving the generation of the highly electrophilic azafulvenium ion 1 (Figure 4A). Nucleophilic addition of neutral M215B to this azafulvenium ion would furnish the hexacyclic azafulvenium ion 2, which could undergo intramolecular alkylation by the neutral pyrrole to establish the M430A heptacycle.
Alternatively, we considered that dimerization may occur in a single-step cyclopentannulation event involving a concerted [6πa+2πs] cycloaddition between M215B and the Z-azafulvenium ion 1 (Figure 4B).11 In this event, both bonds of the cyclopentane ring would be formed simultaneously via an antarafacial interaction between the 6π unit of the azafulvenium ion and one face of the M215B 2π alkene.12 Significantly, molecular orbital analysis establishes that the resulting stereochemistry at all of the contiguous stereocenters of the cyclopentane ring would match that of M430A. Finally, we considered that the dimerization may proceed through a pathway involving generation of the stabilized radical 4 (Figure 4C).13 Addition of 4 to the alkene of M215B and subsequent 5-exo trig radical cyclization of a C1 radical in 5 onto the pyrrole would provide the stabilized heptacyclic radical 6, which would be subject to oxidation. As dimerization of vinyl pyrroloindolizines via any of these approaches was without precedent at the onset of our studies, each of these manifolds represented an attractive means of accessing M430A. To identify conditions to achieve the desired dimerization, we began by probing the reactivity of the tricyclic myrmicarin structures.
Motivated by the prominent role of the poison gland secretion in the ecology of the Myrmicaria ants and the potent toxicity of the previously unidentified alkaloid constituents, Francke and coworkers embarked on elegant studies to identify the alkaloid poisons produced by Myrmicaria eumenoides. In 1995, they reported the first isolation, structure elucidation, and enanioselective synthesis of M237A and M237B (Figure 1).1b Mass spectrometric assisted analysis of the poison gland reservoir of this species revealed that the secretion consisted predominantly of (+)-limonene and two alkaloids with masses of 237. Although pure samples of each alkaloid could be obtained by chromatographic separation on neutral alumina gel, attempted storage at ambient temperature resulted in their rapid interconversion. Analysis of the mixture using a combination of two-dimensional NMR techniques revealed that the alkaloids were diastereomeric 3-butyl-5-(1-oxopropyl)indolizidines.
As the relative stereochemistry of each alkaloid could not be determined by NMR analysis of the mixture, Francke performed a series of structure correlation studies (Schemes 1 and and2).2). In all of these 5-(1-oxopropyl)-indolizine structures the C5 stereocenter could be readily epimerized under basic conditions to provide separable C5 stereoisomers, thus first considerations focused on controlling the relative stereochemistry of C3 and C9. Synthesis of the C3,C9-cis indolizine diastereomers commenced with conjugate addition of nitroester 8 to enone 7 to provide nitroketone 9 (Scheme 1). Stereoselective cis-pyrrolodine formation in the presence of palladium in ethanol provided the C3,C9-cis diastereomer (±)-10. One-pot α-iodination of the ester and intramolecular displacement by the pyrrolidine nitrogen provided the indolizine bicycle. Saponification of the ester and addition of ethyllithium provided a separable mixture of diastereomers (±)-11 and (±)-12. Comparison of the NMR spectra of these compounds to those of the poison gland alkaloids showed that neither matched M237A or M237B.
To confirm the trans disposition of the C3 and C9 methines in M237A and M237B, Francke synthesized the requisite trans pyrrolidine diastereomer (Scheme 2). Sequential alkylation of N-tert-butoxycarbonylpyrroline furnished the intermediate 15, whereupon employment of the previous synthetic sequence provided C3,C9-trans indolizine diastereomers. Comparison of the 1H NMR spectra of these compounds to those of the natural isolates revealed that the spectrum of the C5-(S) isomer matched that of M237A, while the spectrum of the C5-(R) isomer matched that of M237B.
After confirming the relative stereochemistry of the C3 and C9 stereocenters through synthesis of racemic samples of each diastereomer, Francke achieved the enantioselective syntheses of both M237A and M237B (Scheme 3). Lithiation of pyridine 18 and alkylation with enantioenriched epoxide 17 provided the advanced intermediate 19. Hydrogenation of the pyridine ring provided the desired piperidine 20 with 3:2 selectivity. Tosylation of the C3 alcohol and invertive displacement by the piperidine nitrogen in this mixture furnished the corresponding diastereomeric indolizidines, which were separated to provide the desired bicycle as a single stereoisomer. Acid-induced ketone deprotection occurred without epimerization at C5, exclusively providing (−)-M237A. To access (+)-M237B, silica gel-induced epimerization at C5 provided a 1:1 mixture of (−)-M237A and (+)-M237B, which were separated by alumina gel chromatography to provide (+)-M237B. This synthesis established the C9-(R) configuration in the M237A and M237B structures.14
In 1999, Lhommet et al reported an enantioselective synthesis of M237A (Scheme 4).15 To obtain the enantioenriched trans-pyrrolidine 22 they employed a highly diastereoselective reduction of β-enaminoester 21 derived from (S)-pyroglutamic acid. Horner-Wadsworth-Emmons reaction with a functionalized phosphonate provided the enone 23. After careful optimization of the reduction sequence, nitrogen deprotection and concomitant alkene reduction under hydrogenation conditions provided the corresponding free amino ketone 24. Trifluoroacetic acid-induced condensation of the pyrrolidine amine and the pendant ketone produced enone 25. Stereoselective iminium ion reduction using sodium cyanoborohydride in the presence of hydrochloric acid provided a 92:8 mixture of (−)-M237A and (+)-M237B, which were separated chromatographically to provide (−)-M237A as a single stereoisomer.
The first total synthesis of racemic M217 was reported by Schröder and Francke in 1998, two years after its isolation (Scheme 5).7 In line with a biomimetic proposal presented in the isolation paper, the key step of this very interesting synthesis involved dehydrative cyclization of an unsaturated derivative of M237A to provide M217. Commencing with piperidine 26, available in four steps from their reported intermediate 19 (Scheme 3), gentle heating in benzene under nitrogen atmosphere resulted in efficient conversion to M217. In situ 1H NMR monitoring revealed that condensation between the amine and the C3 ketone initially produced the C5,C9-cis indolizidine 27, the proposed unsaturated derivative of M237. Rapid epimerization at C5 provided a mixture of stereoisomers. In a slower process, intramolecular addition of the C3-C1" enamine to the C1' ketone and dehydration of the resulting alcohol afforded the pyrroloindolizine structure of M217.
Schröder and Francke note that formation of M237A, M237B and M217 from related monocyclic precursors provides support for an analogous biosynthetic relationship. They postulate that M215A and M215B might likewise arise via cyclization of a doubly unsaturated analogue of M237, citing the presence of trace amounts of alkaloids with mass 233 in the poison gland secretion as potential evidence for the existence of this intermediate. The corresponding bicylic and tricyclic substructures in the complex myrmicarins prompt them to propose a doubly unsaturated derivative of M237 as a common biogenetic precursor to these alkaloids.
The first enantioselective synthesis of (+)-(R)-M217 was disclosed by Vallée et al in 2000 (Scheme 6).16 Condensation between the D-glutamic acid diethyl ester and tetrahydro-2,5-dimethoxyfuran furnished the enantioenriched N-alkylpyrrole 29. Lewis acid-induced intramolecular Friedel-Crafts cyclization of diester 29 yielded bicycle 30.17 Exhaustive reduction of the C7 ketone with sodium cyanoborohydride in the presence of zinc diiodide followed by ester reduction with lithium aluminum hydride generated alcohol 31. Elaboration to the mixed anhydride 32 and boron trifluoride-induced cyclization yielded the tricylic ketone 33. Under the directing influence of the C3 carbonyl substituent, completely regioselective acylation at C1 provided diketone 34. Reduction of both carbonyl groups using lithium aluminum hydride and Vilsmeier-Haak acylation at the remaining unsubstituted pyrrole position generated the fully substituted M217 core. Finally, lithium aluminum hydride reduction of the C8 carbonyl substituent furnished (+)-(R)-M217.
Intercepting Vallée’s synthesis at intermediate 31 (Scheme 6), in 2000 Lazzaroni et al reported a formal synthesis of (−)-(S)-ent-M217 (Scheme 7).18 Cyclodehydration of the aldehyde 3719 provided bicycle 38, whereupon hydrogenation of the alkene and ester reduction provided alcohol ent-31.
In 2001 Vallée reported the synthesis of M215A and M215B as a mixture of geometric isomers (Scheme 8).20 For the purpose of investigating the inherent selectivity for the position of acylation in 39, they employed lithium aluminum hydride reduction to remove the electron withdrawing C3 ketone in 33. Careful optimization of acylating agent, reaction solvent, and reaction temperature revealed that formylation of 39 under Vilsmeier conditions in toluene at 83 °C afforded 40 as the exclusive monoformylation product in 53% yield. Elaboration to 35 followed by a second Vilsmeier-Haack reaction and Wittig homologation of the resulting C8 aldehyde provided M215A and M215B as a 4:1 mixture.
In order to explore our proposed dimerization strategies we required a succinct, convergent, and enantioselective route to the pyrroloindolizine core of the tricyclic myrmicarins.21 As a strategic element in designing our route we noted that use of a catalytic enantioselective method for introducing the C4a stereocenter (Figure 5) would enable us to efficiently access either antipode of the tricycles, and thus of the complex myrmicarins. Additionally, identification of air- and acid-stable intermediates would allow preparation of a range of derivatives for our proposed dimerization pathways. With these considerations, our strategy entailed rapid assembly of the advanced enantiomerically enriched intermediate 45 through N-vinylation of the substituted pyrrole 43 with a vinyl halide or the vinyl triflate 42 followed by an enantioselective conjugate reduction (Figure 5). Successive Friedel-Crafts cyclizations would then yield the pyrroloindolizine substructure. Critically, we envisioned that the acyl substituent of the pyrrole would direct the successive cyclizations and attenuate the electron rich nature of the pyrrole nucleus, conferring stability to these intermediates.
Inspiration for our N-vinylation/asymmetric conjugate reduction sequence arose from a report by Buchwald et al in 2004, in which they disclosed an efficient method for enantioselective copper-catalyzed reduction of β-azaheterocyclic enoates to the corresponding β-azaheterocyclic esters.22 In their report, the necessary β-azaheterocyclic enoates were prepared by copper-catalyzed N-vinylation of the corresponding β-iodoenoates. We initially envisioned the use of a vinyl iodide as our N-vinylation substrate, however reported conditions for formation of the Z vinyl iodide were incompatible with the acid-sensitive dimethoxyacetal functional group of our desired substrate.23 Furthermore, few methods were available for the direct stereoselective synthesis of E-β-iodoenoates. By contrast, several mild methods existed for the stereoselective formation of E or Z vinyl triflates from the corresponding β-ketoesters. Thus we sought to develop conditions for stereoselective N-vinylation of configurationally defined vinyl triflates that would be applicable to a range of nitrogen heterocycles, providing the resulting N-azaheterocyclic enoates as versatile compounds.24
While several methods were available for the copper-catalyzed N-arylation of amines, amides, azoles, and carbamates, far fewer systems had been developed for the corresponding N-vinylation reactions. Additionally, few reported methods for the N-vinylation of dialkylamines and azoles existed relative to those involving N-vinylation of amides and carbamates. At the time of our study, a single report of palladium-catalyzed N-vinylation of lithiated azoles with vinyl bromides had been described.25 While our initial studies showed that copper-based catalyst systems were ineffective, we discovered that the use of palladium catalysts in the presence of a phosphine ligand and an inorganic base efficiently yielded the desired N-vinylation products under mild conditions. Optimization of the catalyst system revealed that palladium dibenzylideneacetone (Pd2(dba)3), 2-dicyclohexylphosphino-2',4',6'-triisopropyl-1,1'-biphenyl (XPhos),26 and rigorously dry potassium phosphate tribasic in toluene or dioxane at 60 to 100 °C successfully effected N-vinylation of a range of heterocycles with both E- and Z-vinyl triflates, in each case with complete retention of the vinyl triflate geometry (Table 1). Importantly, use of rigorously dry base significantly reduced the formation of alkyne and ketone byproducts. Best results were obtained in N-vinylation of electron deficient azoles with β-ketoester derived vinyl triflates, however these conditions were also successfully applied to N-vinylation of unsubstituted pyrrole (Table 1, 44a and 44d) and indole (Table 1, 44b), and could be employed for simple ketone-derived vinyl triflates (Table 1, 44l).
This method was applied on multigram scale to N-vinylation of the readily accessible Z-vinyl triflate 42 with pyrrole 43 to provide the β-pyrrolyl enoate 44 in 95% yield with complete retention of vinyl triflate geometry (Scheme 9). Copper-catalyzed asymmetric conjugate reduction employing copper acetate dihydrate and S-bis(diphenylphosphino)-1,1’-binaphthyl (S-BINAP) in the presence of polymethylhydrosiloxane (PMHS) according to Buchwald’s protocol smoothly afforded the (R)-β-pyrrolyl ester 45,27 establishing the absolute stereochemistry at the C4a stereocenter. Subsequent acid-induced Friedel-Crafts cyclization occurred with greater than 10:1 selectivity in favor of the desired regioisomer under the directing influence of the C8 carbonyl. Hydrogenation of the resulting alkene, one-pot reduction of the tert-butyl ester through transient in situ protection of the C8 ketone, and conversion of the resulting C3 alcohol to the primary iodide 47 provided the precursor to the pyrroloindolizine core. Although radical-mediated cyclization of the primary iodide proceeded in the presence of tri-n-butyltin hydride and 2,2’-azobisisobutyronitrile (AIBN) in refluxing toluene, optimal yields were obtained by silver(I)-promoted cyclization under strictly anhydrous conditions. This sequence readily affords multigram quantities of the tricyclic ketone (R)-36 as an air stable solid.
This synthesis of the enantiomerically enriched ketone 36 allowed expedient access to each antipode of all three of the tricyclic myrmicarin alkaloids, providing the first geometrically pure samples of myrmicarin 215A (M215A, Scheme 1) and M215B. Commencing with (R)-36, exhaustive lithium aluminum hydride reduction of the C8 carbonyl group provided (+)-(R)-M217 in quantitative yield (Scheme 10). By contrast, M215B was prepared by low temperature lithium aluminum hydride reduction of the ketone 36 to yield an inseparable 1:1 mixture of C8 alcohol epimers, which underwent facile elimination under mildly acidic conditions to provide (+)-(R)-M215B as a single isomer. Stereoselective synthesis of the exceedingly air and acid sensitive (−)-(R)-M215A was achieved in a two-step sequence through 2-chloro-3-ethylbenzoxazolium tetrafluoroborate-induced dehydration of the C8 carbonyl followed by careful reduction of the intermediate alkyne under Lindlar conditions.28
Notably, exposure of M215A to mildly acidic conditions, including brief exposure to silica gel, effects partial conversion to M215B, requiring its manipulation under strictly neutral or basic conditions. Consistent with their isolation report, all three of the tricyclic myrmicarins undergo gradual conversion to their corresponding C6-C7 oxidized derivatives, accompanied by decomposition. Bubbling oxygen through a sample of pure M217 in benzene-d6 and monitoring of the mixture by 1H NMR shows almost complete conversion to myrmicarin 215C within 24 hours, accompanied by a complex mixture of minor decomposition products
(Equation 1). Importantly, ketone 36 does not undergo this mode of oxidation or decomposition, highlighting the advantage of the electron withdrawing carbonyl substituent in the manipulation and long term storage of these tricyclic derivatives.
Investigation into our proposed ionic mode of dimerization via formation of azafulvenium ion intermediates (Figure 4A) commenced by gathering evidence for the existence of tricyclic azafulvenium species. As noted, completely stereoselective formation of M215B could be achieved by exposure of a diastereomeric mixture of the alcohols 48 to acidic aqueous work up. Likewise, treatment of a benzene solution of 48 with acetic acid resulted in gradual conversion of the epimers to M215B exclusively (Scheme 11). 1H NMR monitoring revealed that the elimination proceeded via an approximately 1:1 mixture of the C8 acetate adducts 49, which were both slowly consumed to provide M215B. Although the intermediacy of the C8 acetate adduct and the transformation of both acetate epimers to M215B was consistent with the existence of an azafulvenium ion intermediate, this species was not directly observed by 1H NMR, which may be anticipated as a result of its high reactivity and short lifespan.
By contrast, in situ 1H NMR monitoring revealed that treatment of a benzene-d6 solution of M215B with trifluoroacetic acid (TFA, 1.10 equiv) resulted in quantitative and highly stereoselective dimerization to the heptacyclic iminium ion 50a within 45 minutes (Scheme 11).29 Likewise, in situ 1H NMR analysis showed that upon treatment with TFA (1.10 equiv) the epimeric alcohols 48 immediately generated M215B, which proceeded to exclusively furnish 50a. When M215B was subjected to substoichiometric quantities of TFA, 50a was produced to approximately the extent of the added acid, whereas addition of a large excess of TFA instead generated a mixture of ring-protonated M215B salts, which did not undergo dimerization. Compellingly, exposure of M215A to TFA (1.10 equiv) resulted in formation of the same heptacyclic structure. Under the latter conditions, while trace amounts of M215B were observed in the reaction mixture throughout the course of dimerization, the predominant tricyclic species was M215A, suggesting that the rate of isomerization of M215A was slower than the rate of dimerization of M215B. Cumulatively, these observations suggested an acid-promoted formation of a tricyclic azafulvenium ion and subsequent trapping by available M215B, consistent with our proposed ionic mode of dimerization. Although subsequent studies have established the reversibility of this dimerization, reversion of the heptacycle 50a to tricyclic intermediates was not observed under these conditions.
Initial efforts to isolate this heptacyclic iminium ion failed due to its extreme sensitivity. The oxidative instability of these alkaloids compelled us to consider chemical modifications that would provide a derivative more stable toward a robust means of characterization, such as X-ray crystallography or two-dimensional NMR analysis of a pure sample.30 In order to fully characterize the compounds en route to these isolable derivatives and to garner information about the reactivity of this dimeric species, we developed techniques that were compatible with in situ 1H NMR analysis of these reactions. In particular, use of resin-bound reagents enabled manipulation of the observed intermediates without obscuring the 1H or 13C NMR spectral regions of interest. Gratifyingly, direct addition of the resin-bound base 2-tert-butylimino-2-diethylamino-1,3-dimethyl-perhydro-1,3,2-diazaphosphorine (BEMP) to a benzene solution of 50a under inert atmosphere cleanly effected conversion to a heptacyclic diene, which we named isomyrmicarin 430A (isoM430A, Scheme 12). While this compound was stable as an argon-purged solution in benzene-d6 for approximately 12 hours, attempts at isolation resulted in decomposition, in line with the high oxidative sensitivity of an electron rich diene.
To enhance the stability of this compound we speculated that mild hydrogenation of this diene may provide a more stable derivative. Indeed, addition of catalytic palladium on carbon and saturation of the mixture with dihydrogen at atmospheric pressure provided the heptacyclic enamine 51 as a single diastereomer (Scheme 12). This compound was sufficiently stable to be purified through a short plug of triethylamine-pretreated silica gel and could be stored for 24 hours as an argon-purged solution in benzene-d6 without significant decomposition. Assignments based on a full complement of two-dimensional NMR data corroborated the structure of the heptacyclic enamine 51 and were in agreement with the assignments for the structures of 50a and isoM430A based on the corresponding two-dimensional NMR data.
The structure of the heptacyclic diene isoM430A differed from the M430A structure in the connectivity of the C1-C3b bond and the stereochemistry of the C3 substituent on the cyclopentane ring (Figure 6). The stereochemistry of the substituents on the cyclopentane ring were consistent with an ionic dimerization mechanism involving association of M215B and azafulvenium ion 1 from the convex face of each tricycle (Figure 7). Nucleophilic addition of M215B to 1 would provide the hexacyclic azafulvenium 2, which could undergo alkylation by the pyrrole in the same relative orientation. This sequence would generate the observed configuration at each of the cyclopentane stereocenters. The approach of M215B and the azafulvenium ion may be facilitated by π-stacking between the electron-rich and electron-deficient π systems, which is maintained throughout the proposed mechanism. Indeed, the generation of isoM430A as a single diastereomer was compelling evidence for our proposed ionic dimerization of M215B, representing an unprecedented mode of reactivity for vinyl pyrroloindolizines.
The charged and highly electrophilic nature of the tricyclic and heptacyclic azafulvenium ions in this ionic mechanism suggested that varying the reaction medium may have a pronounced effect on the rate or reversibility of each step. As such, we reasoned that use of solvents of different polarity, dielectric constant, and nucleophilicity, or introduction of additives may alter the observed regio- or stereoselectivity of the dimerization. To assess the influence of the reaction conditions on the rate and selectivity of the dimerization process we examined the acid-promoted behavior of the alcohol 48 and M215B in the presence of different solvents, acid promoters, inorganic salts, and nucleophilic additives. Due to the known sensitivity of isoM430A and the reported sensitivity of M430A, these reactions were monitored in situ by 1H NMR spectroscopy and the identity of the products was confirmed by conversion to previously prepared and characterized derivatives.
The nature of the solvent had a marked effect on the rate of dimerization. In hydrocarbon solvents such as benzene-d6 and cyclohexane-d12, trifluoroacetic acid (TFA)-induced dimerization of M215B was complete within 30 minutes, while in tetrahydrofuran-d8 dimerization of M215B was only 80% complete within 24 hours (Table 2, entries 1 and 2). In contrast, treatment of an acetonitrile-d3 solution of the alcohol 48 with TFA effected conversion to 50a within 5 minutes without visible accumulation of M215B (Table 2, entry 3). Addition of TFA to a methanol-d4 solution of the alcohol 48 resulted in methanolysis to afford a corresponding 1:1 mixture of C8 methanol adducts (52), which were slowly consumed to form 50a without observable persistence of M215B (Table 2, entry 4). In all of these experiments, the rate of addition and number of equivalents of acid used were crucial, as rapid addition or the use of large excess of acid resulted in protonation of the pyrrole ring (i.e. 54, Table 2), which prevented dimerization.
Introduction of salt additives also had a pronounced effect on the dimerization. Addition of stoichiometric strong acid (1.10 equiv) to tetrahydrofuran-d8 or methanol-d4 solutions of the alcohol 48 or M215B saturated with lithium perchlorate (LiClO4) led to protonation of the pyrrole ring, inhibiting dimerization (Table 2, entries 6 and 7). While dimerization proceeded in the presence of saturating lithium chloride (LiCl) in tetrahydrofuran-d8, the rate was reduced six-fold relative to that in tetrahydrofuran-d8 alone. Treatment of a methanol-d4 solution of the alcohol 48 with TFA in the presence of the nucleophilic additive p-methylbenzenethiol led to formation of C8 thiol adducts of the tricycle (53), which did not undergo dimerization (Table 2, entry 8). Significantly, none of these conditions yielded dimeric structures arising from trapping of hexacyclic azafulvenium or heptacyclic iminium intermediates.
While the rate of dimerization of M215B varied significantly in solvents of different polarity and in the presence of additives, the dimeric structure produced was exclusively the heptacycle 50. Dimerization proceeded rapidly in non-nucleophilic and hydrocarbon solvents, while polar and protic solvents reduced the rate of dimerization. Nucleophilic additives likewise retarded or inhibited dimerization by intercepting tricyclic azafulvenium ions to form C8 adducts. The inability to directly identify dimeric species other than 50 or to observe dimeric trapping products suggested that hexacyclic azafulvenium intermediates such as 2 (Figure 4A), if produced, were exceedingly fleeting species that underwent facile intramolecular capture to provide the heptacyclic structure 50.
The insensitivity of the structure of the heptacyclic product to the dimerization conditions prompted us to examine an alternative mechanism for its formation. Specifically, we considered that the exclusive formation of 50 as a single diastereomer may be the consequence of a concerted [6πa+2πs] cycloaddition that generated all of the new stereocenters and both bonds in a single event (Figure 8A).10,11 In this process, interaction between the 6π component of the azafulvenium ion 1 and the 2π component of M215B in a manner that was antarafacial with respect to the 6π component and suprafacial with respect to the 2π component would generate the heptacycle 50. This analysis requires the reactive azafulvenium ion E-1 to adopt the E alkene geometry, which molecular orbital calculations indicate to be 1.3 kcal/mol higher in energy than the Z-azafulvenium ion Z-1.31 A [6πa+2πs] cycloaddition would proceed through the higher energy azafulvenium ion E-1 via approach of the M215B alkene to the sterically accessible side of the azafulvenium π system (Figure 8A). Critically, an analogous [6πa+2πs] cycloaddition involving the sterically accessible sites of the isomeric Z-azafulveniun ion Z-1 and the M215B alkene would provide a heptacycle possessing the M430A connectivity and stereochemistry (Figure 8B).
As a means to investigate this mechanistic hypothesis we designed the conformationally restricted azafulvenium precursor 60 that would yield a Z-azafulvenium ion upon activation. To achieve this, we employed a sterically encumbered [1,3,2]-dioxasilocine tether between C3 and C9 of the tricycle to lock the azafulvenium ion Z-61 in the Z geometry (Figure 9). In this case, formation of the undesired E-azafulvenium ion E-61 would require introduction of a prohibitively strained trans alkene. As a further refinement to this design, we noted that the significant steric interaction between the C10 methyl group and the C11 methylene protons in the C9-(R) azafulvenium ion E-61, which were absent in the C9-(S) epimer, would more strongly favor formation of the Z- azafulvenium ion.
The tethered azafulvenium precursor 60 was readily accessed from the alcohol 55, an intermediate available from our synthesis of the tricyclic myrmicarins (Scheme 13).10 After examining a number of conditions to avoid oxidation of the electron rich pyrrole nucleus, conversion of the primary alcohol to the corresponding aldehyde was achieved using 2-iodoxybenzoic acid in dimethylsulfoxide. Triethylsilyl trifluoromethanesulfonate-promoted cyclization of 56 provided the protected alcohol 57 as a 4:1 mixture of diastereomers, which were separated chromatographically after desilylation. Reprotection of the oxidatively sensitive alcohol and hydroxylation using Davis oxaziridine provided the diol 58 as a 4:1 mixture of separable C9 epimers. Mitsunobu inversion of the C9 alcohol with p-nitrobenzoic acid and hydrolysis to the free alcohol provided the desired C9 epimer. Deprotection of the C3 alcohol, installation of the siloxy tether,32 and reduction of the C8 ketone afforded the conformationally restricted azafulvenium precursor 60.
To investigate our [6πa+2πs] cycloaddition mechanism we required selective ionization of the C8 alcohol in 60 under conditions that would avoid Brønsted acid-induced dimerization of M215B. On the basis of this consideration we explored conditions for Lewis acid activation of the C8 alcohol in 60 that would be compatible with M215B. After optimization we found that treatment of an acetonitrile solution of the alcohol 48 with scandium trifluoromethanesulfonate (Sc(OTf)3) in the presence of the E-thiolketene acetal 62 provided the thiolester 63 without visible Brønsted acid-promoted reactivity (Equation 2). A control experiment demonstrated that addition of Sc(OTf)3 to a solution of M215B in acetonitrile did not induce homodimerization.
With these results, we employed our optimal conditions to effect selective formation of the azafulvenium Z-61 in the presence of M215B as an electron rich π nucleophile (Figure 9).33 In situ monitoring by 1H NMR during portion-wise addition of 0.40 equivalents of Sc(OTf)3 to a 1:1 mixture of azafulvenium precursor 60 and M215B in acetonitrile-d3 showed complete and selective formation of the heptacycle 64, with no visible formation of other tricyclic or dimeric derivatives. This air and acid sensitive compound was sufficiently stable to be isolated and purified by flash column chromatography using triethylamine-pretreated silica gel. Exhaustive structural analysis using a combination of two-dimensional NMR techniques revealed that the obtained product possessed a connectivity and stereochemistry analogous to that of isoM430A.
The exclusive formation of heptacycle 64 provided strong evidence for a stepwise mechanism for heterodimerization involving M215B and the conformationally restricted azafulvenium precursor 60. Importantly, participation of the severely strained E-azafulvenium ion E-61 in a [6πa+2πs] cycloaddition manifold is unlikely, while a [6πa+2πs] reaction involving the corresponding Z-azafulvenium ion Z-61 would engender a C3 stereochemistry in the resulting heptacycle opposite to that of isoM430A. While these observations did not preclude a concerted [6πa+2πs] pathway for homodimerization of M215B to produce M430A or isoM430A, they did implicate two discrete bond forming events in the formation of the heterodimer 64.
The propensity of vinyl pyrroloindolizine derivatives to undergo ionic modes of dimerization inspired us to exploit this behavior to prepare functionalized dimeric derivatives. Data supporting the stepwise nature of these heterodimerizations prompted us to consider tricyclic structures that may enable us to capitalize on the efficiency of the first C2-C3 bond formation (Figure 10). In this vein, we envisioned introduction of a functional group at C8 of the nucleophilic partner in the dimerization that would allow us to secure the C2-C3 bond, then act as a ‘blocking’ group at C1 in the dimer. Ideally, this functional group would prevent formation of the isoM430A heptacycle by either inhibiting C3b alkylation or making it reversible. In this manner we could use a highly selective and efficient ionic dimerization to provide hexacyclic derivatives possessing a handle at C1 to investigate alternative means of accessing M430A heptacycle.
We identified the silyl enol ether 73 (Scheme 14) as a suitable tricyclic nucleophile for this proposed heterodimerization manifold. Nucleophilic addition to a tricyclic azafulvenium ion would provide the oxonium intermediate 69 (Figure 10, B=O), which could either desilylate directly or undergo alkylation to yield a labile C3b aminal that would be subject to fragmentation. In an initial attempt, we drew analogy to our previous mechanistic studies and employed the conformationally restricted azafulvenium precursor 72 as an electrophile (Scheme 14A). Significantly, it was necessary to establish conditions that would effect ionization of the C8 alcohol of this azafulvenium precursor without promoting desilylation of the electron-rich silyl enol ether nucleophile. Gratifyingly, in situ monitoring by 1H NMR showed that subjection of a mixture of the triisopropylsilyl enol ether 73 and the tethered azafulvenium precursor 72 (1.56:1) to catalytic Sc(OTf)3 in acetonitrile-d3 smoothly yielded the tethered hexacycle 74 as a 5:1 mixture of diastereomers in 73% yield (Scheme 14A). After rapid filtration through triethylamine-pretreated silica gel, a combination of NMR studies on the major diastereomer in the mixture enabled full structural assignment. One-dimensional nOe analysis established that the major diastereomer possessed the isoM430A configuration at both C2 and C3. While this tethered heterodimer was sufficiently stable to be stored for brief periods, the enhanced sensitivity of the C4 alcohol toward ionization in the absence of an electron-withdrawing pyrrole substituent hampered further manipulations.
Having established this heterodimerization with the conformationally restricted azafulvenium precursor 72, we began to examine the scope of tricyclic structures that could function as electrophiles in this manifold. Activation of the simple alcohol 48 by catalytic Sc(OTf)3 in the presence of 73 provided the hexacyclic ketone 75 in 77% yield, which showed a moderately enhanced stability relative to the tethered heterodimer (Scheme 14B). To assign the stereochemistry at C2 and C3 in this hexacyclic dimer we relied on derivatization to a more rigid heptacyclic structure. Low temperature reduction of the C1 ketone using lithium aluminum hydride and treatment of the sensitive hexacyclic alcohol with acetic acid resulted in cyclization to the isoM430A heptacycle with greater than 80% stereoselectivity (Scheme 14B).34 While the success of these heterodimerization reactions confirmed our ability to access functionalized hexacyclic structures, the persistent formation of the isoM430A stereochemistry at C3 prompted refinement of our approach.
To introduce additional flexibility in our heterodimeric structures, we sought conditions that would enable us to employ azafulvenium precursors at the C8 ketone oxidation state (Figure 10). Control over the C3 stereochemistry in the resulting dimeric structures would be facilitated by the corresponding increase in oxidation state at C3. Existing precedent for the addition of π nucleophiles to amides upon activation by trifluoromethanesulfonic anhydride developed in our laboratory35 suggested that the electron rich ketone in these pyrroloindolizine derivatives may likewise be subject to activation under these conditions. Activation of the C8 carbonyl in 36 with trifluoromethanesulfonic anhydride would generate a substituted azafulvenium ion that may be sufficiently electrophilic to undergo nucleophilic addition by the silyl enol ether 73 (Figure 11). We anticipated that loss of silyl triflate and elimination of trifluoromethanesulfonic acid would yield an enone structure, providing a handle for introduction of the desired stereochemistry at C3. In an initial attempt at this transformation, addition of 1.00 equivalent of trifluoromethanesulfonic anhydride (Tf2O) to a dichloromethane solution of 36 and 73 in the presence of 5.00 equivalents of 2-chloropyridine at −78 °C provided the desired heterodimer 76, albeit in approximately 10% yield. In this sequence, elimination of trifluoromethanesulfonic acid occurred via deprotonation at C9 to provide the β,γ-enone 76 as a mixture of alkene isomers. After careful investigation of the base additive, order of addition, and reaction temperature, an optimized procedure involving portion-wise addition of 1.50 equivalents Tf2O to a dichloromethane solution of 36 and 73 (1:1) at −78 °C in the presence of 5.00 equivalents of 2,6-di- tert-butyl-4-methylpyridine afforded the desired dimer 76 in 84% yield as a 3:2 mixture of olefin isomers.
As the C2 stereocenter in all of the previously obtained heterodimers possessed the M430A configuration, we focused on preparation of a structure that would maintain this selectivity and allow for alternative stereocontrol at C3.36 In this vein, we reasoned that use of an α-methoxy ketone as an electrophile in our trifluoromethanesulfonic anhydride dimerization protocol would afford a C9 methyl vinyl ether, which after hydrolysis to the C9 ketone would provide a handle for epimerization to the thermodynamic C2,C3-trans diastereomer after closure of the cyclopentane ring (Scheme 15). According to our optimized conditions, Tf2O-mediated heterodimerization of 73 and 77 smoothly provided the methyl vinyl ether 78 as a 3:2:1 mixture of stereoisomers in 84% yield. Upon treatment with mild aqueous acid the C3-C9 vinyl ether reacted rapidly to produce the desired diketone 79.37 Notably, as the lower pyrrole unit in this structure lacked an electron withdrawing substituent, it exhibited sensitivity comparable to that of M217, and as a result this compound and its derivatives had to be manipulated in the strict absence of oxygen. Investigation of a variety of proton sources revealed that optimal yields were obtained employing an aqueous solution of pyridinium p-toluenesulfonate in degassed acetonitrile-benzene. Differentiation between the C9 and C1 carbonyl groups was readily achieved via selective deprotonation at the C10 methyl group and trapping of the potassium enolate with triisopropylsilyl chloride to give 80. After extensive optimization to avoid over reduction at C1 and desilylation of the silyl enol ether, careful reduction of the C1 carbonyl using diisobutylaluminum hydride provided the hexacyclic alcohol 81 as a precursor to C1-functionalized structures.
Preparation of corresponding vinyl halide, triflate, and related derivatives provides substrates for alternative cyclization reactions (Figure 12). Cyclization of a stabilized C1 radical in a 5-exo-trig fashion followed by oxidation of the resulting α-amino radical could provide the heptacyclic structure 86. These derivatives may also allow us to investigate transition metal-mediated cyclization.38 Oxidative addition into the C1-X bond in 83 and β-migratory insertion would generate the desired C1-C8b bond, whereupon β-hydride elimination from C8 would likewise furnish the heptacycle 86 (Figure 12). Generation of a strained 4-membered ring from potential β-migratory insertion involving the C3a-C3b bond suggests that this process may proceed with regioselectivity for formation of the desired C1-C8b bond. In addition to providing functional groups in the C1-C3 linkage relevant to radical- and transition metal-mediated cyclization, preparation of hexacyclic structures possessing electronically distinct pyrrole units also provides opportunities for alternative electrophilic activation.
The broad scope, efficiency, and high level of diastereoselection in these heterodimerization reactions enable preparation of a range of hexacyclic cyclization substrates. Refinement of this strategy has led to the design of dimers that display enhanced stability and propensity for elaboration to advanced intermediates. Current efforts are focused on identifying the most concise sequence for derivatization and cyclization of the versatile dimer 83 and related structures. The advanced intermediates accessed via the proposed chemistry are expected to be subject to mild conversion to the highly sensitive naturally occurring complex myrmicarins.
The myrmicarins are a family of alkaloids that exhibit unique structural features and chemical reactivity. Substructure analysis of the complex members suggests that they may be assembled biosynthetically through succinct dimerization and trimerization of the simpler myrmicarin structures. Guided by this hypothesis, we proposed a potentially biosynthetic approach to M430A involving dimerization of M215B. Our convergent enantioselective route to a stable ketone intermediate possessing the pyrroloindolizine core of the tricyclic myrmicarins enabled us to access enantiomerically enriched samples of the highly sensitive myrmicarins M217, M215A, and M215B. Early investigations on formation of M430A through an ionic dimerization pathway established that M215B undergoes efficient and highly stereoselective homodimerization in the presence of Brønsted acid to produce the heptacyclic structure of isoM430A. Analysis of a potential [6πa+2πs] cycloaddition mechanism employing a conformationally restricted azafulvenium precursor indicated that the corresponding functionalized isoM430A heptacycle was formed in an ionic process involving two sequential bond forming events.
Studies on the reactivity of these heterodimeric structures have guided the strategic design of the tricyclic substrates to furnish functionalized dimeric derivatives. Hexacyclic structures possessing a functional group at C1 provide derivatives relevant to diverse modes of cyclization. This approach, based on a proposed biomimetic assembly of tricyclic myrmicarin structures, enables concise synthesis of complex structures and may provide key insight regarding the biosynthesis of this structurally fascinating family of alkaloids.
M.M. is an Alfred P. Sloan Research Fellow, a Beckman Young Investigator, and a Camille Dreyfus Teacher-Scholar. A.E.O. acknowledges a Novartis Graduate Fellowship. We are grateful for financial support by NIH-NIGMS (GM074825). We thank Dr. H. Ümit Kaniskan and Dr. Bin Chen for helpful discussions. We thank Professor Robert G. Griffin and Dr. Tony Bielecki for use of a high-field instrument at the MIT-Harvard Center for Magnetic Resonance (EB002026).