Unified Total Synthesis of Pyrroloazocine Indole Alkaloids Sheds Light on Their Biosynthetic Relationship.

The total synthesis of seven members of the lapidilectine and grandilodine family of alkaloids has been accomplished in racemic and enantiopure form without protection/deprotection of functional groups. The two key steps, an 8- endo-dig hydroarylation and a 6- exo-trig photoredox cyclization, were catalyzed using gold. A rationale for the formation of the cyclopropane ring of the lundurines is also provided.

Introduction

Lapidilectines and grandilodines are indole alkaloids, isolated from peninsular Malaysia species Kopsia grandifolia D. J. Middleton, that feature lactone (1 and 2) or diester (3–7) motifs (Figure ).[1,2] Together with tenuisines and the cyclopropane-containing lundurines isolated from Kopsia tenuis species,[3] they represent a family of 16 indole alkaloids containing the common pyrroloazocine core. Preliminary studies on the biological activity of lapidilectines and grandilodines demonstrated their ability to reverse multidrug resistance in vincristine-resistant cancer cells.[2]

Figure 1

Pyrroloazocine indole alkaloids.

The first total synthesis of (±)-lapidilectine B (1) was achieved by Pearson et al. in 25 steps.[4] Only very recently, the total syntheses of (+)-lapidilectine B (1)/(+)-grandilodine C (2)[5] and racemic grandilodine B (7)[6] have been developed. However, a synthetic approach to lapidilectine A-type natural compounds 3–5 has not yet been disclosed. Our interest in Kopsia indole alkaloids launched with the discovery of a gold-catalyzed hydroarylation of alkynes that allows a fast access to the pyrroloazocine indole core of these natural compounds.[7] On the basis of this strategy, we developed a concise total synthesis of lundurines A–C.[8]

Besides the synthetic challenge the pyrroloazocine indole alkaloids pose,[9,10] we were intrigued by the biosynthetic relationships among the main classes of this family of alkaloids and, in particular, fascinated by the origin of the cyclopropane ring of the lundurines. Here we report a unified total synthesis of seven members 1–7 of the lapidilectine/grandilodine family streamlined by the use of two gold-catalyzed reactions for the key cyclization steps. With ready access to the main members of this natural product family, we have examined their chemical interconversions, uncovering that the cyclopropane ring can arise from the photochemical decarboxylation of a lactone. This photochemical origin differs from an early biosynthetic hypothesis[11] and is unprecedented among the many pathways that have been elucidated for the biosynthesis of cyclopropane-containing natural products.[12]

Results and Discussion

We considered that compounds 1/2 and 3/4 might be biosynthetically connected through an oxidative decarboxylation[13] via carbocation A, which could become a key intermediate for the synthesis of lactone 8, or be treated with a carbon nucleophile to construct the quaternary stereocenter in 9 en route to diesters 3–7 (Scheme ). The diastereoselectivity of the latter transformation would be controlled by the CO2Me moiety, which shields one face of carbocation A from nucleophilic attack. Intermediate A could be readily accessed from alkene 10, whose synthesis was envisioned through a radical photoredox-catalyzed cyclization of 11. This process was expected to favor the formation of the product with an endo-methoxycarbonyl group. Precursor 11 could ultimately be obtained from aldehyde 12, whose 5-OMe-analogue was an intermediate in our synthesis of the lundurines.[8]

Scheme 1

Bioinspired Unified Retrosynthesis of Lapidilectine/Grandilodine Family of Alkaloids

Unified Total Synthesis

The synthesis of advanced pyrroloazocine compound 12 was realized in both enantiopure and racemic form in 5 steps, on gram scale, from commercially available tryptamine (Scheme ). The first step in our synthetic sequence, a condensation/lactamization/Claisen rearrangement cascade of oxoesters 13a,b with tryptamine occurred with high yield (83%) and for (+)-13b with a good level of chirality transfer from the allylic fragment to the C20 stereocenter (>99% ee for (+)-13b to 70% ee for (−)-14b).[8,14] Aldehydes 14a,b underwent a Seyferth–Gilbert homologation to give alkynes 15a,b.

Scheme 2

Synthesis of Enantiopure Aldehyde Intermediate 12

See Supporting Information for the synthesis of (+)-13b.

Initial 5 mol % of AuCl catalyst employed, then additional 2 mol % added after 0.5–1 h to reach full conversion.[14]

The Au-catalyzed cyclization of 15 was found to be more challenging than that of its 5-OMe derivative.[8] In the presence of the standard AuCl-based catalytic system, the reaction stopped at an unsatisfactory ca. 60% conversion, due to catalyst decomposition. With ligand-stabilized cationic gold complex [IPrAu(NCCH3)]SbF6 (2 mol %), (±)-15a undergoes the gold-catalyzed cyclization with a satisfactory ca. 20:1 8-endo/7-exo selectivity. After subsequent reaction with methyl chloroformate and crystallization (removing traces of 7-exo product), (±)-17a was obtained in 65% yield over two steps.[14] Alternatively, in the presence of acetic acid, the AuCl catalyst was found to be more stable,[15] and full conversion could be achieved with 5 + 2 mol % catalyst loading (Scheme ). The reaction proceeds with excellent 8-endo selectivity (>50:1 endo/exo) and with high yield (16a,b, 76–82%). After introducing the methylcarbamate, the exocyclic double bond of 17a,b was cleaved by a one-pot OsO4/NaIO4 protocol to give aldehyde 12 in racemic and enantioenriched form. For the latter, the enantiomeric excess was enhanced from 70% to >99% by crystallization from acetone, and its absolute configuration was confirmed by high-resolution single-crystal X-ray diffraction.[16]

At this point the synthetic route diverges from the one previously developed for the lundurines.[8] Hydrogenation of the double bond in 12 required carefully controlled conditions[14] in order to prevent over-reduction of the indole into indoline. This was achieved with 10 wt % Pd/C catalyst in EtOAc/CH2Cl2 to give 18 in 65–77% and 92% brsm yields (Scheme ). With aldehyde 18 in hand, we focused on the development of an approach to an ester-containing precursor for the radical cyclization. Current methods to access α-halo esters by aldehyde homologation are essentially limited to cyanohydrin synthesis, requiring several steps and harsh conditions. Thus, seeking an alternative, our attention turned to the work of Denmark et al., where α-hydroxy methyl esters were synthesized from aldehydes and t-BuNC, in a Passerini-type reaction, with an exceptional functional group tolerance.[17] To apply this transformation to compound 18, we had to account for the presence of several Lewis-basic centers in the molecule, increasing the amount of SiCl4 (4.35 equiv instead of 1.1 equiv) and raising the reaction temperature (from −74 to −40 °C).[14] The α-hydroxy ester 19 was obtained in an excellent 88% yield, while the relatively labile methylcarbamate remained intact (Scheme ). Finally, bromide 11 was obtained in ca. 90% yield through an Appel-type reaction using a modified protocol.[14]

Scheme 3

Synthesis of α-Bromo Ester 11

The photoredox process to build the rigid azabicyclic [4.2.2] skeleton was expected to be challenging, as it consists of a rare 6-exo-trig radical spirocyclization onto an indole through a transition state that cannot adopt a chairlike conformation.[18,19] Nevertheless, our initial study employing [Ru(bpy)3]Cl2[20] showed that 11 indeed underwent cyclization to the key synthetic intermediate 10 with excellent endo-diastereoselectivity, which most likely originates from the CO2Me group adopting an equatorial position in a twist-boat-like transition state (Scheme ). However, the reaction did not reach full conversion because of catalyst decomposition, and several other evaluated catalytic systems did not provide a significant improvement.[14] Remarkably, the digold photoredox catalyst [(dppmAuCl)2] introduced by Barriault and co-workers[21] showed an outstanding efficiency, leading to 10 in 91% yield (Scheme ).

Scheme 4

Photoredox Cyclization of 11 into 10

Values for exo-CO2Me isomer (R2b).

The main challenge of the photoredox transformation presumably arises from the lack of driving force for the cyclization of the relatively stable α-CO2Me radical R1 into strained benzyl radicals R2a (endo-CO2Me) and R2b (exo-CO2Me). Density functional theory (DFT) calculations[22] provided ΔG‡ values of 20.2 and 23.2 kcal/mol for the cyclization of R1 into R2a and R2b, respectively, which is in accord with the observed high endo-selectivity (Scheme ). In addition, these barriers indicate that the radical cyclization is a relatively slow process, 4–6 orders of magnitude slower than a standard 6-exo-trig cyclization of 6-hepten-1-yl radical.[14] Furthermore, from a thermodynamic point of view, the open form of radical R1 was found to be even more stable than the cyclized radical intermediates (ΔG0 = 1.7 kcal/mol for R2a and 7.5 kcal/mol for R2b). This suggests that the oxidation of benzylic radicals R2 is the main driving force of the transformation that shifts the equilibrium between R1 and R2. Both kinetic and thermodynamic data of the radical cyclization imply that R1 may accumulate in the reaction medium, causing side-reactions and catalyst decomposition. For example, the undesired reduction of R1 leads to 11H, which was isolated and identified as the main byproduct in both Au- and Ru-catalyzed photoredox processes.[14] This reduction into 11H naturally results in the oxidation of the photoredox catalyst and ultimately leads to the deactivation of the catalytic system.[23] For [Ru(bpy)3]Cl2 catalyst, such a deactivation product, deep-purple trans-[Ru(bpy)2Br2]Br, was isolated and structurally characterized.[14]

Alkene 10 was used as a precursor of benzylic carbocation A, a common intermediate in the synthesis of both lactones 1 and 2 and diesters 3–7 (Schemes and 5). Under strong acidic conditions (50% aqueous H2SO4), the styrene moiety underwent protonation. The subsequent hydrolysis of the ester, presumably involving cyclized cation A′ (see below), led to the desired lactone 8 (Scheme ). Forging the quaternary carbon center bearing the benzylic CO2Me group proved to be challenging. Besides the obvious increase in molecular strain, this transformation goes in the opposite direction to that of the proposed natural biosynthetic scheme, in which the benzylic C–C bond is cleaved, not constructed. Our scouting experiments suggested that carbocation A undergoes proton elimination, providing alkene 10 or lactonization into 8 faster than it reacts with carbon nucleophiles (t-BuNC, trimethylsilyl cyanide (TMSCN), anisole, or 1,3-dimethoxybenzene).[14]

Scheme 5

Lactonization of 10 and Benzylic Allylation of 20

We found that benzylic alcohol 20, accessible from 10 via Mukaiyama hydration,[24] can partake in a Hosomi–Sakurai-type allylation with allylTMS,[25] providing the desired product 9. Although promising, this result was not suitable for application in the total synthesis because 9 was formed together with alkene 10 as an inseparable mixture in ca. 1:2 ratio. Other allylic nucleophiles were tested (Scheme ), and allylSnBu3[26] demonstrated outstanding efficiency (ca. 30:1 ratio of 9/10). Allylated product 9 was obtained in 72% yield over two steps as a single diastereomer (Scheme ), confirming that the ester moiety provides sufficient steric hindrance to favor the exclusive attack of the nucleophile on one face of carbocation A. Our attempts to perform a one-pot radical cyclization/benzylic C–C bond construction were unsuccessful.[14,27] The [(dppmAuCl)2]-catalyzed photoredox cyclization of 11 in the presence of nucleophiles such as TMSCN and allylTMS led to elimination product 10. Employing allylSnBu3 (10 equiv) as nucleophile, this one-pot procedure provided a mixture of 9 and 10 in an unsatisfactory 1:2 ratio.[28]

Scheme 6

Introduction of Benzylic CO2Me Group; X-ray Structures of (±)-20 and (±)-24

Two-carbon degradations of allyl moieties to aldehydes are typically performed through double-bond migration/ozonolysis sequence.[29] However, low conversion and difficult separation of isomeric alkenes made this strategy impractical in our case. To circumvent this issue, we envisioned to first cleave the double bond in 9 with OsO4/NaIO4 to aldehyde 21, which could be further converted to an enamine that could undergo a second oxidative cleavage (Scheme ).[30] Treating aldehyde 21 with pyrrolidine afforded enamine 22 quantitatively (1H NMR), which was then fragmented to aldehyde 23, under similar oxidative conditions. A base (Na2CO3) was utilized in the second C=C cleavage to efficiently generate the enamine intermediate and suppress the retro-Stork enamine alkylation reaction, which results in the formation of alkene 10 and/or alcohol 20 byproducts.[14] As both events involved OsO4/NaIO4, we combined them, thereby developing a novel one-pot degradation of an allyl group to the two-carbon lower aldehyde homologue. With this strategy, 9 was converted into 23 in 72% yield.[31] The oxidation of aldehyde 23 to carboxylic acid 24 was hampered by the steric hindrance of the substrate and the oxidative decarboxylation of the carboxylic acid product (see below). The oxidation under Pinnick conditions[32] was found to be less selective than that with KMnO4. Excess of oxidants (KMnO4 and NaIO4) was employed to minimize the side decarboxylation, presumably mediated by reduced manganese species. Finally, carboxylic acid 24 was converted into methyl ester 25 by treatment with TMS-diazomethane (80% yield over two steps).

Seven pyrroloazocine indole alkaloids were accessed from lactone 8 and diester 25 utilizing a unified end-game strategy (Scheme ). The introduction of a double bond in (+)-8 and (−)-25 following Magnus’ thioamide sequence[8,33] provided (+)-grandilodine C (2) and (−)-lapidilectam (4), respectively. These results allowed us to revise the sign of the optical rotation for lapidilectam (4) from the previously reported[1b] (+) to (−).[14] Treatment of diester 4 under basic conditions led to a ca. 3.5:1 mixture of 4 and 7, which could be separated, affording (+)-grandilodine B (7). The final reduction of amides 2, 4, and 7 with Me3OBF4/NaBH4 led to (+)-lapidilectine B (1), (−)-lapidilectine A (3), and (+)-isolapidilectine A (6),[14] while (−)-grandilodine A (5) and unnatural (+)-dihydrolapidilectine B (26) were synthesized by direct reduction of (−)-25 and (+)-8 with borane.

Scheme 7

Endgame: Diversification of Intermediates (+)-8 and (−)-25 into Seven Natural Products of the Lapidilectine/Grandilodine Family; X-ray Structures of (±)-1, (±)-3, (±)-4, (±)-5, and (±)-7

On the Biosynthesis of Pyrroloazocine Indole Alkaloids

With the natural products and synthetic intermediates in hand, we studied the biosynthetic relationships of pyrroloazocine indole alkaloids. The initial proposal contains a rare oxidative decarboxylation of a methyl ester in lapidilectine A-type diesters, followed by a heterolytic decarboxylation of the lactone, furnishing the cyclopropane of the lundurines (Scheme ).[11,14] However, to date, there is still no experimental support for this hypothesis. To probe the first decarboxylation event, we attempted to transform diester 25 into alcohol 20, alkene 10, or directly into lactone 8. However, 25 was unreactive to SET oxidation agents (ceric ammonium nitrate (CAN)), and no oxidation peak was observed (up to +2.0 V) by cyclic voltammetry. In contrast, the sodium salt of carboxylic acid 24 (oxidative potential peak of +0.9 V) underwent the desired oxidative decarboxylation with CAN in the presence of MeOH.[34] Under these conditions, a ca. 1:1 mixture of alkene 10 and the methyl ether of 20 (20Me) was formed. Treatment of this mixture with 50% aqueous sulfuric acid afforded lactone 8 in 62% NMR yield (Scheme ).

Scheme 8

Cyclopropane Formation by Stepwise Decarboxylation of Acid 24 and Lactone 8; Original and New Proposal for the Origin of the Cyclopropane of Lundurines

Simplified version. For the full scheme, see Supporting Information and ref (11).

Next, the hypothesis of heterolytic cleavage of lactones to cyclopropanes was tested. When lactone 8 was treated with KCl in wet dimethylsulfoxide (DMSO) at 85 °C for 17 h,[35] no cyclopropane product was observed. This led us to consider an alternative mechanism for the generation of the cyclopropane via homolytic photochemical decarboxylation of the γ-lactone.[36] To our delight, irradiation of lactone 8 with UVB light (300 nm) or UVC light (254 nm) gave the corresponding cyclopropane 27 in moderate NMR yield, with UVB wavelength being more selective and efficient (Scheme ).

Finally, it was noticed that, despite their lower stability in comparison with the corresponding pyrrolones,[14] 3-pyrrolines (lundurine B, lapidilectines A and B, isolapidilectine A, and tenuisine A) were isolated in significantly greater amounts,[1a,2] suggesting that the pyrrolones (lundurine A, grandilodines B and C, lapidilectam, and tenuisine C) could arise from an auto-oxidation process. Indeed, synthetic lundurine B[8] spontaneously converted (>50%) into lundurine A upon storage under air for 16 months.

We suggest an alternative biosynthetic scheme that features carboxylic acids 28 as key intermediates and a homolytic mechanism for decarboxylation of lactone into cyclopropane. The new proposal is summarized in Scheme . It is in line with the experimentally observed reactivity and additional DFT studies (see below).

The new biosynthetic hypothesis starts with the hydrolysis of kopsijasminilam-type precursors to pyrroline-containing carboxylic acids 28 (Scheme ). The close proximity between C20 and amide N atom in known kopsijasminilam-type structures[37] suggests that lactam hydrolysis might involve N-acyl ammonium cations 29 as key intermediates.[38,39] Indeed, the DFT optimization[22] of kopsijasminilam-type structures with planar amide and the carbocation at C20 led to more stable cations 29 that have short C20–N distances (ca. 1.62 Å), nonplanar nitrogen atoms, and elongated amide C21–N bonds (ca. 1.52 Å).[14] This geometry resembles the one previously reported for N-alkylated twisted amides, which are prone to hydrolysis of N–CO bond.[40]

Carboxylic acids 28 can be further converted into methyl esters lapidilectine A and isolapidilectine A. These compounds were previously proposed as precursors to lactones. Our experimental results suggest that this scenario is unlikely, but carboxylic acids 28 might undergo an oxidative decarboxylation/lactonization, leading to the formation of lactones lapidilectine B and tenuisine A. The proximity of the lactone carbonyl group and the pyrroline nitrogen atom in lapidilectine B (2.5 Å, X-ray structure, Scheme ) suggests a possible intramolecular assistance of the latter in the lactonization process. While we were not able to obtain amino acid 28 or its 14,15-dihydro analogue to test this hypothesis experimentally, our DFT calculations[22] support this possibility. On one hand, open benzylic carbocation A is more stable than closed carbocation A′ by 5.3 kcal/mol (Scheme ). This suggests that the lactonization of lactam derivatives is hampered by this unfavorable equilibrium. Experimentally, we indeed observed that this transformation only takes place efficiently in a strongly acidic medium, and that under milder conditions the major pathway observed is toward elimination product 10. On the other hand, pyrroline-containing cation B′ benefits from additional stabilization by the nitrogen and, in this scenario, is more stable than open benzylic cation B by 5.6 kcal/mol.[41] This, in turn, suggests that the lactonization of pyrroline derivatives should be a favorable natural process.

Scheme 9

Relative Stability of Cation Intermediates A and A′, and B and B′

The subsequent decarboxylation into cyclopropane may be a light-induced process, as was demonstrated experimentally for 8. Tenuisine A, after a photochemical decarboxylation, would give lundurine B, featuring a 3-pyrroline fragment. This transformation presumably proceeds with the initial excitation of the arene system (absorption band at 290 nm in 1, 2, and 8), which results in the homolytic cleavage of the benzylic C–O bond, which is perpendicular to the aromatic system. Irradiation of a thin film of neat 8 between two quartz plates with sunlight over 10 days also showed formation of 27, suggesting the relevance of the photochemical decarboxylation in the biosynthetic scheme and the sufficiency of UVb light to trigger this transformation.[42]

Finally, pyrrolines might undergo allylic oxidation, producing pyrrolones. As the decarboxylation of lactone and allylic oxidation do not require any specific enzyme, the corresponding compounds could have been formed after collection of the plant material and even can be artifacts of the isolation process.

Conclusions

In summary, we have developed concise total syntheses (11–19 steps) of enantiomerically pure (+)-lapidilectine B (1), (+)-grandilodine C (2), (−)-lapidilectine A (3), (−)-lapidilectam (4), (−)-grandilodine A (5), (+)-isolapidilectine A (6), (+)-grandilodine B (7), and unnatural (+)-dihydrolapidilectine B (26) by means of two highly efficient gold-catalyzed cyclization processes. The skeleton of grandilodines/lapidilectines (10) was assembled in only 9 steps and 16% overall yield from tryptamine. We also propose a new hypothesis of biosynthetic relationship among Kopsia pyrroloazocine indole alkaloids by means of two decarboxylation events: the elimination of a carboxylic acid to form a lactone and a photoinduced conversion into the cyclopropane present in the lundurines.