Total Synthesis of (±)- and (−)-Actinophyllic Acid

Abstract

Development of efficient sequences for the total syntheses of (±)-actinophyllic acid (rac-1) and (−)-actinophyllic acid (1) are described. The central step in these syntheses is the aza-Cope/Mannich reaction, which constructs the previously unknown hexacyclic ring system of actinophyllic acid in one step from much simpler tetracyclic precursors. The tetracyclic hexahydro-1,5-methano-1H-azocino[4,3-b]indole ketone rac-37 is assembled from o-nitrophenylacetic acid in four steps, with oxidative cyclization of a dienolate derivative of tricyclic precursor rac-35 being the central step. In the first-generation synthesis, this intermediate is transformed in two steps to homoallyl amine rac-43, whose formaldiminium derivative undergoes efficient aza-Cope/Mannich reaction to give pentacyclic ketone rac-44. In four additional steps, this intermediate is advanced to (±)-actinophyllic acid. The synthesis is streamlined by elaborating ketone rac-37 to β-hydroxyester intermediate rac-53, which is directly transformed to (±)-actinophyllic acid upon exposure to HCl and paraformaldehyde. This concise second-generation total synthesis of (±)-actinophyllic acid is realized in 22% overall yield from commercially available di-tert-butylmalonate and o-nitrophenylacetic acid by a sequence that proceeds by way of only six isolated intermediates. The first enantioselective total synthesis of (−)-actinophyllic acid (1) is accomplished by this direct sequence from tricyclic keto malonate (S)-35. Catalytic enantioselective reduction of α,β-unsaturated ketone 66 is the key step in the preparation of intermediate (S)-35 from the commercially available Boc-amino acid 65. Discussed also is the possibility that the aza-Cope/Mannich reaction might be involved in the biosynthesis of (−)-actinophyllic acid.

Introduction

Thrombotic diseases are a major cause of mortality and morbidity in the developed world. In healthy individuals, a complex network of enzymatic processes carefully regulates the balance between blood clotting and blood thinning.¹ Inhibition of activated thrombin-activatable fibrinolysis inhibitor (TAFIa), an unstable zinc-dependent carboxypeptidase, is a promising approach toward upregulating fibrinolysis, the process whereby small blood clots are removed from circulation.^2,3 In a screening program designed to discover natural product inhibitors of TAFIa, 40,000 extracts from Australian plants and marine organisms were screened by Carroll and co-workers, initially leading to the identification of promising extracts from the tree Alstonia actinophylla, growing on the Cape York Peninsula, Far North Queensland.⁴ Ultimately a new indole alkaloid, (−)-actinophyllic acid (1, Figure 1), was identified from this source as a potent inhibitor in the coupled enzyme assay TAFIa/hippuricase (IC₅₀ = 0.84 μM).

Figure 1.

Structure of (−)-actinophyllic acid and its three unique fragments.

The carbon connectivity and relative configuration of actinophyllic acid (1) were determined largely by detailed NMR analysis.⁴ The 2,3,6,7,9,13c-hexahydro-1H-1,7,8-(methanetriyloxymethano)pyrrolo[1′,2′:1,2]azocino[4,3-b]indole-8(5H)-carboxylic acid skeleton⁵ of actinophyllic acid is unique among natural products. Moreover, the simpler 1-azabicyclo[4.4.2]dodecane (2), 1-azabicyclo[4.2.1]nonane (3) and octahydropyrrolo[1,2-a]azocine (4) fragments that define its structure are found in no other indole alkaloids. The absolute configuration depicted in structure 1 for (−)-actinophyllic acid was advanced on the basis of its proposed biogenesis from precondylocarpine via a novel biogenetic pathway.⁴ Rigorous definition of the absolute configuration of (−)-actinophyllic acid (1) by spectroscopic and computational methods⁶ was realized only after this laboratory completed the first total synthesis of (±)-actinophyllic acid in 2008.⁷

We describe in this article the development of an efficient strategy for assembling the ring system of actinophyllic acid, which culminated in the first total synthesis of this unique alkaloid. A simplification of the later stages of this sequence leading to an improved second-generation total synthesis of (±)-actinophyllic acid is also reported. In addition, the first enantioselective total synthesis of (−)-actinophyllic acid (1), which confirms the spectroscopic assignment of its absolute configuration,⁸ is disclosed. The possibility that the aza-Cope/Mannich reaction is involved in the biosynthesis of natural products is considered and a potential biosynthetic route to actinophyllic acid is proposed.

Results and Discussion

Synthesis Plan

The retrosynthetic analysis that guided our efforts to prepare (−)-actinophyllic acid (1) is outlined in Scheme 1. Disconnecting the tetrahydrofuran ring at the hemiketal C–O bond reveals pentacyclic ketone 5. This intermediate contains a 3-acylpyrrolidine unit, which suggests its potential formation by aza-Cope/Mannich rearrangement of formaldiminium ions derived from hexahydro-1,5-methano-1H-azocino[4,3-b]indole precursors such as 6 or 7.⁹ Of these possibilities, the postulated transformation of intermediate 6 to 5 is particularly attractive as actinophyllic acid would result directly. If the relative configuration of the ester and hydroxymethyl side chains of precursor 6 could not be established in an efficient fashion, an alternate possibility would be to carry out the aza-Cope/Mannich transformation with precursor 7, and subsequently elaborate the product to intermediate 5 by reaction of an ester or acid enolate with formaldehyde. Disconnecting the allylic alcohol intermediates 6 and 7 identifies tetracyclic ketone 8 as an important subgoal of our synthesis plan.

Scheme 1.

Retrosynthetic Analysis of Actinophyllic Acid

The pivotal aza-Cope/Mannich rearrangement step of our projected synthesis plan is analyzed in more detail in Scheme 2. Although this reaction had not been employed previously to transform a 3-vinylpiperidine to a 1-azabicyclo[4.2.1]nonan-5-one (atoms highlighted in red in Scheme 2), the prospects for success appeared good. Molecular modeling of intermediates such as 11 showed that the overlap between the vinyl and iminium fragments, although far from ideal, was comparable to that of several other successful aza-Cope/Mannich processes.^9,10

Scheme 2.

The Pivotal Aza-Cope/Mannich Rearrangement Step

Moreover, there was evidence from early studies of Grob and coworkers that the proposed cationic aza-Cope rearrangement step, 11 → 12 (Scheme 2), would likely take place readily.¹¹ Specifically, they had shown that solvolytic Grob-fragmentation of tosylate 13 generated largely 4-azocine iminium ion 14, which was transformed rapidly to the 3-vinylpiperidine iminium ion 15 (eq 1). By the principle of microscopic reversibility, the reverse transformation, as postulated in the conversion of intermediate 11 to 12, should be possible. That the equilibrium of the proposed iminium ion isomers likely lies on the side of the 3-vinylpiperidine isomer should be of no concern, as a stereoelectronically favorable intramolecular Mannich reaction would be expected to capture iminium ion isomer 12 in the postulated aza-Cope/Mannich transformation.

(1)

The hexahydro-1,5-methano-1H-azocino[4,3-b]indole ring system of intermediate 8 is a structural feature of several indole alkaloid families and the ring system of the uleine alkaloids.¹² As a result, a number of methods for assembling this tetracyclic scaffold have been developed.^13,14 Particularly attractive to us was a new construction in which intermediate 8 would be assembled from two fragments of similar complexity: an indole-2-malonate (9) and a six-membered, azacyclic synthon having electrophilic sites for bond construction at C2 and C4.

Several aspects of the plan adumbrated in Scheme 1 warrant additional comment. The aza-Cope/Mannich disconnection is highly productive because this transformation, if successful in the synthetic direction, would construct the previously unknown hexacyclic ring system of actinophyllic acid in one step from a much simpler tetracyclic precursor. However, this strategy is not without significant risk. Besides deferring the pivotal aza-Cope/Mannich step to a late stage of the synthesis,¹⁵ intermediate 8, and later ones derived from this structure, contain a potentially labile gramine fragment that could result in unraveling of the piperidine ring. At the outset, we hoped that we could arrive at intermediate 8 by a sufficiently direct sequence that these key issues could be addressed relatively quickly in our experimental studies.

Total Synthesis of (±)-Actinophyllic Acid

Attempted Formation of the 2,5,6,7-Tetrahydro-1,5-methano-1H-azocino[4,3-b]indole Ring System by Sequential Pyridinium Ion Alkylation/Pictet–Spengler-Type Cyclization

One of our early attempts to assemble the hydro-1,5-methano-1H-azocino[4,3-b]indole ring system followed the general approach to this ring system developed by Bosch and coworkers.¹⁶ We envisioned constructing tetracyclic ketone intermediate 8 by the sequence enunciated in Scheme 3. Intermediate 17 would arise from addition of the conjugate base of indole malonate 9 to C4 of pyridinium salt 16.¹⁷ Oxidation of one of the prochiral double bonds of the dihydropyridine fragment of this adduct could potentially promote intramolecular attack by the pendant indole with introduction of an oxygen substituent on the resulting one-carbon bridge of the product.

Scheme 3.

Initial Plan for Preparing a Tetracyclic Precursor from an Indole-2-malonate and a Pyridinium Salt

To pursue this potential construction of tetracyclic ketone 8, dimethyl indole-2-malonate (19)^18,19 was deprotonated with 1.2 equiv of a variety of strong bases [LDA, NaHMDS, KHMDS or BrMgN(i-Pr)₂] in THF at temperatures between 0 and −78 °C,²⁰ and the resulting anion was allowed to react at −78 °C with the pyridinium salt generated in situ from the reaction of pyridine with 2,2,2-trichloroethyl chloroformate (Troc–Cl).²¹ Product 20 resulting from the addition of the malonate side chain to C4 of the pyridinium salt was never observed. The major product produced in these reactions, adduct 21, resulted from coupling at the 3-position of the indole malonate nucleophile. When the bromomagnesium salt of indole-2-malonate 19 was used, adduct 21 was formed in high yield.

As an alternative approach, we investigated the reaction of a less-basic anion generated from α-keto malonate 22¹⁸ with several pyridinium salts, with the goal of forming the indole following the contruction of the azabicyclo[3.2.1]octane ring system (Scheme 5). The initial condensation was most efficient with the in situ-generated N-triflylpyridinium triflate salt,²² giving product 23 in excellent yield. However, attempted epoxidation of the N-sulfonylenamine functionality of adduct 23 with a variety of oxidants (DMDO, m-chloroperbenzoic acid, Shi’s dioxirane²³) did not lead to the formation of tetracyclic product 24. The major mode of reactivity observed under most of the conditions examined was fragmentation of bond a of adduct 23 to regenerate keto malonate 22.

Scheme 5.

Addition of Tricarbonyl Intermediate 22 to a 1-(Trifluoromethanesulfonyl) pyridinium Salt

Formation of the Keto-Bridged Hexahydro-1,5-methano-1H-azocino[4,3-b]indole Ring System by Intramolecular Oxidative Dienolate Coupling

The observation that anions derived from indole malonate 19 reacted with electrophiles at C3 of the indole suggested that the order of bond formation in the construction hexahydro-1,5-methano-1H-azocino[4,3-b]indole ketone 8 be reversed (Scheme 6). In such a sequence, coupling of the indole malonate with a six-membered iminium electrophile, in the ideal case one derived from a precursor such as 25 that incorporates a carbonyl group at C3, would deliver adduct 26. The hexahydro-1,5-methano-1H-azocino[4,3-b]indole ring system would then be fashioned by bond formation between the starred carbons of intermediate 26. One possibility we envisioned for this bond construction was oxidative coupling of a dienolate intermediate such as 27.

Scheme 6.

Revised Plan for Preparing Ketone 8 by Intramolecular Oxidative Dienolate Coupling

The formation of C–C bonds by oxidative coupling of enolates generated from ester, ketone, and carboxylate precursors has a long history.^24,25 Throughout the 1970s and 1980s, intramolecular oxidative couplings of dienolates to form three-, four-, five- and six-membered rings were disclosed,²⁶ as was the intramolecular couplings of enolates derived from two different functional groups.²⁷ Nonetheless, this C–C bond-forming method has received only modest attention for the construction of more elaborate structures such as polyfunctional natural products.²⁸ Three impressive examples from the Paquette, Cohen, and Baran laboratories are summarized in Scheme 7.²⁹ Absent from existing precedent was the intramolecular coupling of malonate and ketone enolates, as well as a demonstration that an unprotected indole might survive such a sequence. Nonetheless, because of the potential brevity of the synthetic sequence postulated in Scheme 6, we were drawn to examine the prospect that tricarbonyl intermediates such as 26 could be transformed directly to 1,5-methanoazocino[4,3-b]indole ketone 8.

Scheme 7.

Selected Examples of Intramolecular Oxidative Enolate Couplings Used in Complex Molecule Syntheses

The short sequence for assembling 1,5-methanoazocino[4,3-b]indole ketones 36 and 37 that ultimately resulted from these studies is summarized in Scheme 8. The synthesis begins with acylation of the methoxymagnesium salts of dimethyl (28) or di-tert-butyl (29) malonate with acid chloride 30, to give keto malonates 22¹⁸ and 31 in good yields. On small scales, dimethyl intermediate 22 could be transformed to indole dimethyl malonate 19 in a yield of 62% by catalytic hydrogenation over Pd(OH)₂/C in methanol.¹⁸ However, over multiple runs we found the yields of indole malonates 19 and 32 to be irreproducible using this procedure. These reactions suffered from formation of variable amounts of N-hydroxyindole products, which underwent reduction of the N–O bond only slowly. Forcing conditions, such as elevated reaction temperatures or high catalyst loadings, did lead to reduction of the N–O bond; however, these conditions also promoted competitive reduction of the indole C2–C3 double bond. Difficulties in optimizing the Pd(OH)₂/C reduction prompted us to investigate alternative methods for reducing the nitro group of intermediates 22 and 31. In the dimethyl series, simply carrying out the reaction with excess zinc in acetic acid at 50 °C delivered indole malonate 19 reliably in 48–55% yield.³⁰ However, this procedure was problematic in the di-tert-butyl ester series, particularly in large-scale runs wherein the reaction exotherm was difficult to control. In these cases, zinc and acetic acid reduction gave product 32 contaminated with various amounts of tert-butyl 2-indoleacetate. After examining several alternative procedures, we finally found that transfer hydrogenation over Pd/C in a 2:1 mixture of formic acid and triethylamine in the presence of catalytic amounts of ammonium metavanadate at 50 °C promoted reproducible transformation of keto malonate 31 to indole di-tert-butyl malonate 32 in 74–79% yield.³¹ In the absence of NH₄VO₃, the reaction rapidly produced a mixture of the desired indole malonate 32 and the corresponding N-hydroxyindole, which was slow to undergo further reduction. Inclusion of a catalytic amount of NH₄VO₃ accelerated reduction of the hydroxyindole.³²

Scheme 8.

Synthesis of Hexahydro-1,5-methano-1H-azocino[4,3-b]indole ketones 36 and 37

We next examined methods for joining indole malonate and piperidone fragments. We soon found that the desired transformation could be accomplished by simply allowing the indole to react at room temperature in N,N-dimethylformamide (DMF) with crude bromopiperidone 33, a reactant readily generated by radical bromination of commercially available 1-tert-butoxycarbonyl-3-piperidone.³³ This condensation was carried out on multigram scale to provide indole piperidones rac-34 and rac-35 in 74% and 85% yield, respectively. Presumably the highly reactive N-acyloxy, C-acyl iminium cation generated by ionization of bromide 33 is an intermediate in this coupling step.³⁴

With convenient access to keto malonates rac-34 and rac-35 in hand, we turned to examine the intramolecular oxidative coupling of dienolates generated from these intermediates. In early studies carried out largely with dimethyl malonate precursor rac-34, we surveyed several bases including lithium diisopropylamide (LDA), lithium hexamethyldisilazide, and potassium hexamethyldisilazide, along with various metal oxidants including ferrocenium hexafluorophosphate, iron(III) chloride, iron(III) acetylacetonate, [Fe(DMF)₃Cl₂][FeCl₄], Cu(II) 2-ethylhexanoate, and Cu(II) chloride. The best results were obtained with a combination of LDA and [Fe(DMF)₃Cl₂][FeCl₄], a complex simply formed by combining FeCl₃ with DMF.³⁵ Deprotonation of indole piperidone rac-34 or rac-35 with 3.2 equiv of LDA in tetrahydrofuran (THF) at −78 °C, followed by adding a THF solution of 3.5 equiv of [Fe(DMF)₃Cl₂][FeCl₄], and allowing the reaction to warm to room temperature over 60–90 min provided crystalline tetracyclic ketones rac-36 (68–72% yield) or rac-37 (60–63% yield) on scales up to 10 g. Single-crystal X-ray analysis of rac-37 confirmed the 1,5-methanoazocino[3,4-b]indole structure of this product.^36a

Although mechanistic details of this oxidative cyclization have not been examined thoroughly, several aspects merit mention. To avoid destabilizing A^1,3 interactions between the indole and tert-butoxycarbonyl (Boc) groups,³⁷ the piperidine ring of the piperidone indole malonate precursors should exist in a conformation, 38, in which the indole moiety is axial, thus positioning the methine hydrogen adjacent to the indole orthogonal to the π-bond of the carbonyl group (eq 2). For this reason, we anticipated that regioselection in the deprotonation of the 3-piperidinone at the methylene carbon would be high. As the yields of the cyclized products rac-36 and rac-37 were reduced significantly if only 2.2 equiv of LDA were employed, it is possible that the indole of 38 is also converted to its conjugate lithium base.³⁸ However, it is also plausible that the third equivalent of LDA deprotonates the indole of the cyclized product as it forms.

(2)

A major byproduct produced under the reaction conditions of the oxidative dienolate cyclization of precursor rac-35 is a symmetrical homodimer resulting from coupling at the methylene carbon adjacent to the piperidone carbonyl group, rac-40. Although a solid, we have thus far been unable to obtain single crystals to allow its structure to be fully established. Nonetheless, its relative configuration can be assigned, because the same dimer is produced as a byproduct in the oxidative dienolate coupling of enantioenriched (S)-35 (see below).³⁹ In the enantiomerically enriched series, the only symmetrical dimers that could result would have C₂-symmetry, of which two are possible. The C₂-symmetric dimer assigned as rac-40, would result from dimerization of intermediate 39 from the face of the piperidone enolate opposite to the quasi-axial indole fragment. Fortunately, the yield of the dimer decreased as the oxidative coupling reaction was carried out at higher concentration. ⁴⁰

First-Generation Synthesis of (±)-Actinophyllic Acid (rac-1)

The next step in advancing tetracyclic ketones rac-36 and rac-37 to (±)-actinophyllic acid was introduction of a vinyl group from the Re face of the carbonyl group (Scheme 9). It was anticipated that the proximal ester substituents, particularly in the tert-butyl ester series, would shield the Si faces of the ketones during addition of a vinyl nucleophile (see X-ray model in Scheme 8). Vinyllithium and vinylmagnesium bromide did not add to ketones rac-36 or rac-37 at −78 °C in THF, and produced complex product mixtures at higher temperatures. However, premixing these ketones with anhydrous cerium(III) chloride in THF, followed by addition of vinylmagnesium bromide at −78 °C did bring about addition to the ketone carbonyl group.⁴¹ In the dimethyl series, this reaction resulted in formation of lactone rac-41 (IR 1790 cm⁻¹) in 29–44% yield, with 22–30% of the starting ketone recovered. Under the same conditions, di-tert-butyl ketone precursor rac-37 was converted solely to allylic alcohol, rac-42, in nearly quantitative yield. Both the lactone and allylic alcohol products were viewed as viable intermediates in route to (±)-actinophyllic acid (rac-1). We chose to investigate elaboration of allylic alcohol rac-42 first.

Scheme 9.

Vinylation of Ketones rac-36 and rac-37

The first sequence developed to elaborate intermediate rac-42 to (±)-actinophyllic acid (rac-1) is summarized in Scheme 10. Reaction of allylic alcohol rac-42 with trifluoroacetic acid (TFA) at 0 °C selectively cleaved the Boc group to deliver, after aqueous base workup, amino alcohol rac-43 in high yield. This intermediate was not purified, but immediately allowed to react with 1 equiv of paraformaldehyde and a catalytic amount of camphorsulfonic acid (CSA) in benzene at 70 °C to promote aza-Cope/Mannich transformation to yield pentacyclic keto diester rac-44. Exposure of this crude product to neat TFA at room temperature gave the amino acid trifluoroacetate salt rac-45a as a single stereoisomer in 76% overall yield from rac-42. Fischer esterification of this amino acid, followed by counter ion exchange delivered amino ester trifluoroacetate salt rac-46 as a 2:1 mixture of a and β ester epimers in 92% yield. For characterization purposes, these methyl ester epimers could be separated by HPLC.⁴²

Scheme 10.

Elaboration of Allylic Alcohol rac-42 to (±)-Actinophyllic Acid Hydrochloride (rac-1·HCl)

The total synthesis (±)-actinophyllic acid (rac-1) was then completed in two additional steps. Deprotonation of the 2:1 mixture of a and β ester epimers rac-46 with LDA at −78 °C, followed by addition of a THF solution of monomeric formaldehyde⁴³ gave largely one aldol adduct (diastereomer ratio, dr = 14–20:1), which was partially purified by reverse phase chromatography to give (±)-actinophyllic acid methyl ester trifluoroacetate salt (rac-47).^44,45 Hydrolysis of this product with 4 M HCl and purification of the product by reverse-phase HPLC provided pure (±)-actinophyllic acid hydrochloride (rac-1·HCl) in 48% yield from amino ester rac-46.

The predominant formation of (±)-actinophyllic acid methyl ester from the aldol reaction of the lithium enolate of rac-46 with formaldehyde is attributable to steric factors. As depicted in Figure 2, approach of formaldehyde from the Re (a) face of the enolate is hindered by the relatively bulky 2-carbon bridge (atoms b and c). In contrast, approach to the Si (β) face of the double bond is relatively free of obstruction, as the ketone bridge is small and tilted away from the π bond of the enolate.⁴⁶

Figure 2.

Rationale for stereoselection in the reaction between the ester enolate of rac-46 and formaldehyde; for clarity, water molecules rather than THF are shown as ligands on lithium.

Purification of synthetic (±)-actinophyllic acid hydrochloride (rac-1·HCl) by HPLC, as reported for the natural product,⁴ does not reproducibly give samples of (±)-actinophyllic acid (rac-1) that exhibit identical ¹H NMR spectra. Moreover, the ¹H NMR spectra in DMSO-d₆ of these samples does not precisely match those reported for natural 1 in this solvent. We ascribe these differences to samples of (±)-actinophyllic acid (rac-1) purified in this way containing variable small amounts of the conjugate acid. To confirm this supposition, incremental amounts of sodium methylsulfinylmethylide-d₅ were added to a sample of (±)-actinophyllic acid hydrochloride in DMSO-d₆, which resulted in substantial changes for several ¹H NMR resonances. When just less than 1 equiv of base was added, a ¹H NMR spectrum identical to that reported for natural 1 was obtained (see the Supporting Information). Unfortunately, a sample of natural actinophyllic acid is no longer available for direct comparison.^47a

This first-generation total synthesis of (±)-actinophyllic acid hydrochloride (rac-1·HCl) could be streamlined by combining the four acid-catalyzed steps in the conversion of rac-42 → rac-46 into a single operation (Scheme 11). Reaction of allylic alcohol rac-42 with neat trifluoroacetic acid at room temperature resulted in cleavage of the Boc group and the tert-butyl esters, promoting decarboxylation to provide amino acid salt rac-48 as a single stereoisomer. Removal of trifluoroacetic acid in vacuo, dissolution of the crude residue in acetonitrile, addition of 1 equiv of paraformaldehyde, and heating at 70 °C for 3 h promoted aza-Cope/Mannich reorganization to the carboxylic acid salt rac-45.⁴⁸ Removal of acetonitrile in vacuo, followed by dissolution of the resulting residue in a 0.5 M methanolic solution of HCl and heating at 50 °C provided amino ester trifluoroacetate salt rac-46 as a 1:1 mixture of a and β ester epimers in 62% overall yield. In this streamlined fashion, the first-generation total synthesis of actinophyllic acid was completed in 8% overall yield by a sequence that proceeds via only seven isolated intermediates.

Scheme 11.

Streamlined First-Generation Total Synthesis of (±)-Actinophyllic Acid

In an attempt to shorten the synthesis even further, we examined the aldol reaction between carboxylic acid rac-45 and formaldehyde in an attempt to generate (±)-actinophyllic acid (rac-1) directly from this precursor. In one such experiment, carboxylic acid trifluoroacetate salt rac-45, which is available in 73% yield from precursor rac-42, was deprotonated with 4.5 equiv of LDA at 0 °C in THF and after 30 min was cooled to −78 °C (eq 3). Addition of a THF solution of monomeric formaldehyde,⁴³ followed by quenching with trifluoroacetic acid before allowing the reaction to warm to room temperature, returned a mixture of the unreacted amino acid starting material, lactone rac-49, actinophyllic acid hydrotrifluoroacetate (rac-1·CF₃CO₂H), and N-hydroxymethylindole rac-50 after reverse-phase C18 column chromatography. The relative configuration of aldol adduct rac-49 was assigned on the basis of a diagnostic lactone carbonyl stretch at 1762 cm⁻¹ and 2D NMR analysis. The unexpected reversal in diastereoselection in the aldol reaction of the carboxylic acid dianion compared to that of corresponding methyl ester led us to abandon this shorter sequence.⁴⁹

(3)

Second-Generation Total Synthesis of (±)-Actinophyllic Acid (rac-1): An Improved Endgame

The total synthesis of (±)-actinophyllic acid (rac-1) summarized in Schemes 8–11 suffered from a low-yielding aldol–hydrolysis sequence (Scheme 10) used to transform aza-Cope/Mannich product rac-46 to (±)-actinophyllic acid. Moreover, in this inaugural route the all-carbon quaternary center present in allylic alcohol intermediate rac-42 is sacrificed by the decarboxylation–formaldehyde aldol sequence used to establish the relative configuration of the quaternary carbon stereocenter of (±)-actinophyllic acid. The formation of pentacyclic lactone rac-41 from vinyl cerium addition to the keto dimethyl ester rac-36 (Scheme 9) showed that it would be possible to differentiate the two ester substituents of the malonate fragment prior to the aza-Cope/Mannich step. Thus we turned to examine the possibility of optimizing the generation of such pentacyclic lactone intermediates and subsequently transforming them to actinophyllic acid.

Successful elaboration of pentacyclic ketone rac-36 via lactone intermediate rac-41 to (±)-actinophyllic acid is summarized in Scheme 12. The formation of lactone intermediate rac-41 was improved by employing 2.5 equiv of cerium(III) chloride and 3.5 equiv of vinylmagnesium bromide in the reaction with ketone dimethyl ester intermediate rac-36. Under these conditions, all of the starting ketone was consumed at −78 °C, with lactone rac-41 being formed reproducibly in 49–51% yield.⁵⁰ Selective reduction of the lactone carbonyl of this intermediate with excess sodium borohydride in methanol/THF at −20 °C delivered hydroxy ester rac-51 in 56% yield. Removal of the Boc group with TFA in dichloromethane at room temperature, followed by concentration in vacuo, dissolution of the residue in acetonitrile, and heating with 1 equiv of paraformaldehyde at 70 °C generated (±)-actinophyllic acid methyl ester hydrotrifluoroacetate (rac-47). This intermediate was not purified, but directly hydrolyzed with 4 M HCl to furnish (±)-actinophyllic acid hydrochloride (rac-1·HCl) in 69% yield.

Scheme 12.

Synthesis of (±)-Actinophyllic Acid Hydrochloride via Lactone Intermediate rac-41

The sequence summarized in Scheme 12 showed that the primary alcohol side chain generated by chemoselective reduction of lactone rac-41 presented no problem in the pivotal aza-Cope/Mannich transformation. This improved synthesis of (±)-actinophyllic acid would be further streamlined if a related sequence could be realized in the di-tert-butyl ester series, because a global deprotection–aza-Cope/Mannich sequence could potentially directly deliver actinophyllic acid.

The optimized second-generation total synthesis of (±)-actinophyllic acid that resulted from these considerations is summarized in Scheme 13. The first obstacle to overcome was transformation of keto tert-butyl diester rac-37 to lactone rac-52. After some experimentation, we found that this conversion could be realized in good yield by first allowing rac-37 to react with 2.5 equiv of both cerium(III) chloride and vinylmagnesium bromide at −78 → −70 °C in THF. After the ketone was consumed, as judged by thin layer chromatography, 1.5 equiv of acetic acid were added to quench the excess organometallic reagent; allowing the reaction to then warm to −20 °C promoted lactonization to give pentacyclic lactone rac-52 in 83% yield.⁵¹ The rate of vinylation was increased if 5 equiv of lithium chloride was added to the reaction mixture,⁵² however carbonate rac-55 was then a significant byproduct (rac-52:rac-55 = 6:1). This byproduct was not observed when lithium chloride was absent. Lithium chloride likely activated the ester carbonyl and promoted elimination of the tert-butyl ester enolate from the tetrahedral intermediate formed upon cyclization. The propensity of lanthanides to interact only weakly with esters likely explains why byproduct rac-55 was not formed in the absence of lithium chloride.⁵³

Scheme 13.

Improved Endgame of the Concise Second-Generation Total Synthesis of (±)-Actinophyllic Acid

In three additional steps pentacyclic lactone rac-52 was elaborated in high yield to (±)-actinophyllic acid. Chemoselective Luche reduction of lactone rac-52 delivered hydroxy ester rac-53 in 86% yield.⁵⁴ The use of cerium(III) chloride was crucial to obtain high yields in this transformation. Carrying out the reduction of the lactone with sodium borohydride in methanol/THF at 0 °C provided a 3.4:1.0 mixture of hydroxy ester rac-53 and formate ester rac-56 in 75% yield, whereas this formate byproduct was not formed under Luche conditions. Exposure of hydroxy ester rac-53 to 5 N aqueous HCl at 60 °C removed the two protecting groups. Concentration of this reaction mixture, dissolution of the residue of rac-54 in 5:1 acetonitrile/water, addition of 1.1 equiv of paraformaldehyde, and heating to 70 °C promoted aza-Cope/Mannich reaction to furnish (±)-actinophyllic acid hydrochloride (rac-1·HCl) in 93% yield.⁵⁵ This considerably improved second-generation total synthesis of (±)-actinophyllic acid was realized in 22% overall yield from commercially available di-tert-butylmalonate and o-nitrophenylacetic acid by a sequence that proceeds by way of only six isolated intermediates.

Enantioselective Total Synthesis of (−)-Actinophyllic Acid

As noted earlier, there was no experimental evidence concerning the absolute configuration of (−)-actinophyllic acid (1) at the time our investigations in this area began. Contemporaneously with our collaboration with the Nakanishi group that established the absolute configuration of (−)-actinophyllic acid (1) by chiroptical methods,⁶ we initiated efforts to ascertain its absolute configuration by enantioselective total synthesis. The first chiral intermediate in our synthetic route to actinophyllic acid is indole piperidone 35 (Scheme 8). Before beginning to prepare this intermediate enantioselectively, we wished to confirm that it would not be racemized under the basic conditions of the ensuing oxidative dienolate coupling step. To pursue this issue, the enantiomers of rac-35 were separated by enantioselective HPLC. As illustrated in eq 4, conversion of these enantiomers to hexahydro-1,5-methano-1H-azocino[4,3-b]indole ketone 37 was accompanied by little, if any, racemization.⁵⁶

(4)

Success has been registered recently in several laboratories in accomplishing some bimolecular nucleophilic additions to N-acyliminium ions in catalytic asymmetric fashion.⁵⁷ As a result, we examined briefly the possibility of preparing (S)-35 directly by coupling of 1-(tert-butoxycarbonyl)-2-methoxy-3-piperidone (the methoxy analog of 33) with indole malonate 32 in the presence of several Bronsted acid catalysts derived from (R)-1,1′-bi(2-naphthol).⁵⁸ Thus far, our efforts in this area have not led to a satisfactory enantioselective synthesis of intermediate (S)-35 (see the Supporting Information for details). As a result, we turned to develop a diasteroselective construction of this key intermediate.

The obvious approach was to replace piperidone intermediate 33 with an appropriate piperidine electrophile bearing an alcohol substituent (or alcohol derivative) at C3 that would direct bond formation at C2. Prospects for success in such an endeavor appeared promising, as Kobayashi and coworkers had reported high trans diasteroselectivity in Lewis acid-catalyzed reactions of 2-acetoxy-3-acyloxy(or 3-alkoxy)-1-(benzyloxycarbonyl)piperidine with β-substituted enoxysilanes and silyl ketene acetals.⁵⁹

Salient results of our initial examination of the coupling of indole malonate 32 with related precursors in the N-Boc piperidine series are summarized in Scheme 14. Kobayashi’s conditions were effective in promoting the desired transformation, as combining piperidine diol rac-57a⁶⁰ and indole 32 in chloroform with a catalytic amount of scandium(III) triflate at 0 °C provided indole piperidine rac-58 in 40% yield, as a 5:1 mixture of trans and cis stereoisomers. Byproduct 60 arising from the reaction of piperidine rac-57a with two equiv of the indole nucleophile was isolated in 9% yield. Similar reaction of the 2-methoxy derivative rac-57b led to a decreased yield of adduct rac-58 (trans:cis = 5:1) and an increased yield (35%) of byproduct 60. The piperidine ring-opening pathway was not significantly suppressed by employing an electron-withdrawing 2,2,2-trifluoroethoxy substituent at C2, as the reaction of piperidine rac-57c with indole 32 gave a similar product distribution to that of piperidine derivative rac-57b. In contrast, diacetoxypiperidine rac-57d^60c condensed with indole malonate 32 in good yield to deliver indole piperidine rac-59 as a 20:1 mixture of inseparable trans and cis stereoisomers.⁶¹ In this case, no products arising from ring opening of the piperidine were detected.

Scheme 14.

Diastereoselective Heteroarylation of Piperidine Diol Derivatives

With conditions for diastereoselective heteroarylation in hand, we sought to develop an efficient enantioselective synthesis of the 3R isomer of diacetoxypiperidine 57d.⁶² Initial work focused on sequential enantioselective epoxidation–hydrolysis of commercially available tetrahydropyridine 61 (eq 5).⁶³ Shi epoxidation was unselective and provided the diol product as a mixture of epimers, each in 7% ee.^64,65 Two conditions were examined for enantioselective epoxidation with Jacobsen’s catalyst:⁶⁶ use of aqueous sodium hypochlorite as the stoichiometric oxidant⁶⁷ and the low-temperature procedure that employs m-chloroperbenzoic acid as the stoichiometric oxidant.⁶⁸ The low-temperature procedure was more successful; however, enantioselectivity was still modest (59–60% ee for both diol diastereomers).⁶⁵

(5)

Proline-catalyzed α-oxidation⁶⁹ of Boc-protected amino aldehyde 62⁷⁰ proved highly enantioselective, providing alkoxyamine 63 as a single stereoisomer in 98% ee (eq 6). However, under the best conditions we identified the yield was low, likely reflecting the facile cyclization of aldehyde 62 to hydroxypiperidine 64.⁷¹

(6)

We eventually discovered that diacetoxypiperidine (3R)-57d could be prepared by the convenient sequence shown in Scheme 15. Commercially available amino acid 65 was initially converted to its Weinreb amide,⁷² which underwent cerium(III) chloride-mediated vinylation to furnish enone 66 in 88% yield over the two steps.⁷³ In a sequence carried out without purification of intermediates, α,β-unsaturated ketone 66 was hydrogenated using Noyori’s catalyst 68⁷⁴ in isopropanol to give allylic alcohol (R)-67 in 91% ee.^75,76 Concentration of this reaction mixture, dissolution of the residue with dichloromethane, ozonolysis at −78 °C, quenching with triphenylphosphine, and finally addition of acetic anhydride, triethylamine, and a catalytic amount of 4-dimethylaminopyridine (DMAP) at room temperature gave diacetoxypiperidine (3R)-57d in 73% yield from enone 66.

Scheme 15.

Enantioselective Synthesis of Diacetoxypiperidine (3R)-57d

The elaboration of diacetoxypiperidine (3R)-57d to (−)-actinophyllic acid is summarized in Scheme 16. Piperidine electrophile (3R)-57d was added to a stirring mixture of 1.3 equiv of indole malonate 32 and 5 mol % of scandium(III) triflate in dichloromethane at 0 °C to give adduct (2S,3R)-59 in 88% yield (dr = 17:1) on a multigram scale. Deacetylation of this product with diisobutylaluminum hydride (DIBAL) in toluene at −78 °C delivered, after column chromatography, the pure trans-alcohol (2S,3R)-58 in 82% yield.⁷⁷ Attempts to cleave the acetyl group of (2S,3R)-59 with sodium or potassium methoxide in methanol resulted in competitive transesterification of the tert-butyl esters. Swern oxidation of alcohol (2S,3R)-58 furnished indole piperidone (S)-35,⁷⁸ which underwent intramolecular oxidative dienolate coupling to provide ketone (1S,5R)-37 in 57–59% yield and 91% ee. Whereas the corresponding racemate could be recrystallized from toluene, this enantioenriched ketone could not be purified in this fashion. Instead, tetracyclic ketone (1S,5R)-37 was purified by column chromatography followed by trituration with diethyl ether; this change in the purification procedure likely accounts for the slightly lower yield realized in the enantioenriched series. Elaboration of (1S,5R)-37 as described in Scheme 13 furnished (−)-actinophyllic acid hydrochloride (1·HCl) in high yield. Reverse-phase HPLC of 1·HCl afforded the zwitterion, which was crystallized from methanol to provide single crystals of a methanol and water solvate, allowing the first X-ray analysis of (−)-actinophyllic acid (1) to be accomplished (Figure 3).^36b The optical rotation of an analytical specimen of synthetic 1 (>99% ee) showed [a]_D −199 (c 0.67, MeOH). A nearly identical rotation was observed for the hydrochloride salt 1·HCl. The optical rotation of 1 at the sodium D line did not compare well to the reported rotation of [a]_D −29 (c 0.001, MeOH).⁴,^47b The total synthesis of enantioenriched (−)-actinophyllic acid 1·HCl summarized in Scheme 15 proceeds by way of nine isolated intermediates and was accomplished in 18% overall yield (91% ee); enantiopure 1·HCl (>99% ee) was accessed in 8% yield.⁷⁹

Scheme 16.

Total Synthesis of (−)-Actinophyllic Acid Hydrochloride (1·HCl)

Figure 3.

X-ray model of (−)-actinophyllic acid. The asymmetric unit has two molecules each of actinophyllic acid, methanol, and water; for clarity, only one molecule of actinophyllic acid is shown.

Potential Biosynthetic Relevance of the Aza-Cope/Mannich Reaction

The aza-Cope/Mannich reaction has proven to be a remarkably robust reaction that has been employed to construct a wide variety of pyrrolidine-containing ring systems.⁹ This cascade reaction typically takes place in high yields under extremely mild reaction conditions, often at or near room temperature and at neutral pH. It is these features, along with the wide occurrence of pyrrolidine-containing natural products, that has led us over the years to wonder whether the aza-Cope/Mannich reaction is utilized in natural product biosynthesis. Our demonstration that (−)-actinophyllic acid (1) is formed in high yield in one step from a much simpler tetracyclic precursor (+)-54 by an aza-Cope/Mannich reaction (see Scheme 16) surely raises this question in the current context.

The possibility that the biogenesis of (−)-actinophyllic acid (1) involves an aza-Cope/Mannich reaction is hightened by the isolation from Alstonia plant species indigineous to Malaysia of the indole alkaloids (−)-undulifoline (69)⁸⁰ and (−)-alstilobanines C (70) and B (71)⁸¹ that contain a uleine alkaloid ring system and the complete carbon scaffold found in synthetic aza-Cope/Mannich precursor (+)-54. A biosynthetic sequence,^82–84 potentially beginning with (+)-stemmadenine (72),^85,86 that delivers alkaloids 69–71 could plausibly lead to an intermediate such as tetracyclic diol 73 (Scheme 17). Oxidative transformation of this intermediate to formaldiminium ion 74 would give rise to (−)-actinophyllic acid (1) by an aza-Cope/Mannich sequence.

Scheme 17.

Plausible Biosynthesis of (−)Actinophyllic Acid (1) from an Intermediate Having a Uleine Aklaloid Skeleton by an Aza-Cope/Mannich Reaction

Conclusion

The first total syntheses of (±)-actinophyllic acid (rac-1) and (−)-actinophyllic acid (1) have been accomplished by short and efficient synthetic routes. (±)-Actinophyllic acid was prepared in 22% overall yield from commercially available di-tert-butylmalonate and o-nitrophenylacetic acid by a sequence that proceeds by way of only six isolated intermediates. The enantioselective total synthesis of (−)-actinophyllic acid (1) proceeds by way of nine isolated intermediates to deliver enantioenriched (−)-actinophyllic acid 1 (91% ee) in 18% overall yield or enantiopure 1 (>99% ee) in 8% overall yield.⁷⁹ In these syntheses, no protecting groups are introduced, and in the notably concise synthesis of rac-1, nearly all steps form skeletal bonds of actinophyllic acid.

A number of steps in the synthetic sequence are noteworthy. The aza-Cope/Mannich reaction allows the previously unknown hexacyclic ring system of actinophyllic acid to be constructed in one step from much simpler tetracyclic precursors. These total syntheses entail the first use of this powerful cascade reaction for forming medium azacyclic rings and 1-azabicyclic ring systems. An oxidative intramolecular dienolate cyclization is the pivotal step in an efficient construction of the commonly occuring 2,3,4,5,6,7-hexahydro-1,5-methano-1H-azocino[4,3-b]indole ring system found in the uleine alkaloids. This step represents the first intramolecular coupling of malonate and ketone enolates, as well as the first demonstration that an unprotected indole can survive such a coupling reaction. Tetracyclic intermediates 36 and 37 produced in this way could well serve as precursors of other families of indole alkaloids.

In conclusion, the efficient construction of actinophyllic acid by an aza-Cope/Mannich reaction suggests the possibility that nature utilizes this powerful cascade reaction in natural product biosynthesis.

Scheme 4.

Addition of Conjugate Bases of Indole-2-malonate 19 to a 1-Acyloxypyridinium Salt