Cyclic Allene Approach to the Manzamine Alkaloid Keramaphidin B

Abstract

We report an approach to the core of the manzamine alkaloid keramaphidin B that relies on the strain-promoted cycloaddition of an azacyclic allene with a pyrone trapping partner. The cycloaddition is tolerant of nitrile and primary amide functional groups and can be complemented with a subsequent retro-Diels–Alder step. These efforts demonstrate that strained cyclic allenes can be used to build significant structural complexity and should encourage further studies of these fleeting intermediates.

Graphical Abstract

Manzamine alkaloids serve as attractive synthetic targets due to their impressive structural complexity. Some members of this class of natural products display important biological activities, including but not limited to anticancer, antibacterial, and anti-inflammatory properties.^1–3 The present study pertains to the manzamine alkaloid keramaphidin B (1, Figure 1), which was first isolated in 1994 as a racemate and features a pentacyclic framework with an azadecalin core.^4,5 The eastern portion of the azadecalin core bears a [2.2.2]-bridged bicycle containing four stereogenic centers, one of which is quaternary (C8a). Additionally, the natural product possesses 11- and 13-membered macrocyclic rings. Two prior syntheses of keramaphidin B (1) have been achieved: Baldwin’s biomimetic synthesis of (±)-1 in 1998^6–8 and Fürstner’s synthesis of (+)-1 via an intermolecular Michael/Michael addition cascade to generate the central core in 2021.⁹

Figure 1.

(−)-Keramaphidin B (1), a complex manzamine alkaloid, and azacyclic allene 2, an underutilized synthetic building block.

We viewed keramaphidin B (1) as a suitable target to probe the scope and limitations of the strained cyclic allene methodology in a complex setting. Specifically, we questioned whether azacyclic allenes (2) could be used to assemble the azadecalin core and generate the C8a quaternary stereocenter of keramaphidin B (1, Figure 1). The use of strained cyclic allenes in the context of total synthesis has remained underexplored, especially in comparison to related aryne chemistry, which has been used in >100 total syntheses to date.^10–12 Encouraged by our group’s recent success in using an azacyclic allene (2) in the total synthesis of (−)-lissodendoric acid A,¹³ we sought to further expand cyclic allene chemistry by utilizing these fleeting intermediates to build the highly complex bridged azadecalin core of keramaphidin B (1).

Strained cyclic allenes provide new tools for the rapid generation of highly complex sp³-rich scaffolds under mild reaction conditions. Although they were experimentally discovered over 50 years ago,^14–17 they have recently emerged as valuable synthetic building blocks.^13,18–41 Most relevant to the studies described herein, it has been demonstrated that azacyclic allenes (2) serve as competent dienophiles in strain-promoted Diels–Alder cycloadditions due to their appreciable strain energy of ~27 kcal/mol.³⁷ Generally, studies have focused on the use of electron-rich dienes, but in the context of our total synthesis studies,¹³ we found that regioselectivity of the cycloaddition with respect to the olefins of the cyclic allene can be modulated by using relatively electron-deficient pyrones. This approach has also proven useful in assembling the azadecalin core of keramaphidin B (1) in the presence of potentially reactive functional groups, as we describe in this report.

Figure 2 shows our retrosynthetic analysis of keramaphidin B (1). The natural product 1 was expected to arise from [2.2.2]-azabicycle 3 through late-stage functional group manipulations and the installation of the 13-membered macrocycle. Next, we envisioned the assembly of [2.2.2]-azabicycle 3 via a series of cycloaddition processes. [2.2.2]-Azabicycle 3 would be accessed from intermediate 4, which would be generated in situ from precursor 5, via an intramolecular iminium or N-acyl iminium Diels–Alder (ImDA) cycloaddition.^42–44 In turn, intermediate 5 would arise from cycloadduct 6 via a retro-Diels–Alder reaction that proceeds with the loss of carbon dioxide. Lastly, cycloadduct 6 would be accessed from strained cyclic allene 7 and pyrone 8 via the key strain-promoted Diels–Alder cycloaddition. This step would simultaneously introduce a quaternary center and construct the azadecalin core of the natural product. As will be discussed herein, several variations of this overall strategy were considered,⁴⁵ including the tether length (n) for the ImDA step, the oxidation state for the ImDA precursor (i.e., iminium vs N-acyl iminium),⁴⁶ and whether cycloadduct 6 was isolated in the Diels–Alder/retro-Diels–Alder sequence.

Figure 2.

Retrosynthetic analysis of (−)-keramaphidin B (1).

We first attempted to prepare a cycloadduct of type 6 with a pendant primary amine (see Figure 2; n = 7), as this could plausibly allow for direct assembly of the core of keramaphidin B (1) and install all the carbons necessary for the 11-membered macrocycle (Figure 3). The model system⁴⁷ shown is proposed to provide valuable insight into the fundamental reactivity of the strained cyclic allene for constructing the bicyclic core of the natural product. The necessary cyclic allene precursor 11 was derived from silyl ketone 9⁴⁸ using a scalable three-step cross-metathesis, hydrogenation, and triflation sequence. This route, which mimics a general strategy developed by our laboratory for prior studies, gave access to 11 in 26% yield over three steps. To evaluate the key step, 11 was subjected to pyrone 12 in the presence of CsF at 23 °C. Notably, pyrone 12 was chosen as the trapping partner for this cycloaddition, as it has been demonstrated to be a competent diene for the strain-promoted Diels–Alder cycloaddition with azacyclic allenes.¹³ We were delighted to obtain cycloadduct 14 in 56% yield. Of note, this key step forms two new C–C bonds and sets three new stereocenters, including a quaternary carbon at C8a, which highlights the utility of cyclic allene methodology to generate structural complexity in a single synthetic step.

Figure 3.

Model system for assembly of the azadecalin scaffold (14).

The strain-promoted Diels–Alder cycloaddition of the cyclic allene proceeds with high regio- and diastereoselectivity, presumably via an approach shown in transition structure 13 (Figure 3). With regard to regioselectivity,⁴⁹ cycloaddition could occur at either olefin of the cyclic allene; however, the more substituted, relatively electron-rich olefin preferentially undergoes the reaction with the relatively electron-deficient pyrone, presumably via an inverse electron demand Diels–Alder cycloaddition.¹³ Considering the diastereoselectivity, the pyrone is thought to approach the allene in an endo fashion, with favorable orbital overlap between the diene and the nonreactive olefin of the cyclic allene.³⁴ In this case, these stereoelectronic factors led to the formation of 14 with excellent regio- and diastereoselectivity.

With cycloadduct 14 in hand, we aimed to construct the desired [2.2.2]-azabicycle via an ImDA reaction (Figure 4). We first sought to reduce the C4a–C5 alkene in the piperidinyl ring to install the requisite stereochemistry at C4a.⁵⁰ This was achieved by subjecting cycloadduct 14 to allylic oxidation conditions, followed by a diastereoselective 1,4-reduction using modified Stryker’s reagent.¹³ Product 15 was obtained in 48% yield over two steps with >20:1 dr. Next, heating 15 to 80 °C in acetonitrile facilitated a retro-Diels–Alder reaction. Treatment of the product with TFA led to the removal of the Boc protecting groups, thus affording the desired substrate 16. Considerable effort was put forth to enable the desired intramolecular ImDA⁵¹ and generate the [2.2.2]-azabicyclic core. Although the formation of the iminium intermediate was validated,⁵² exhaustive attempts to access the desired product 17 were not successful. We hypothesize that there is a significant entropic penalty to achieve the necessary reactive conformation for the ImDA reaction, resulting from the lengthy tether.⁵³ Consequently, we opted to pursue a modified substrate bearing a shorter and more geometrically constrained tether.

Figure 4.

Synthesis of ImDA precursor 16 and evaluation of the intramolecular ImDA reaction.

Our revised approach to synthesize the [2.2.2]-azabicycle is shown in Figure 5. We envisioned accessing cyclic allenes 18, which bear a shorter tether with either pendant nitrile or primary amide functional groups for the introduction of N-substituents. Trapping with pyrone 8 and a subsequent retro-Diels–Alder reaction furnished cycloadducts 19. If the primary amide and nitrile functional groups are tolerated in this transformation, this step would expand upon known azacyclic allene cycloaddition chemistry. Subsequently, the s-cis diene present in 19 could engage in the intramolecular N-acyl ImDA or ImDA cycloaddition,⁴⁶ depending on the oxidation state at C16, to furnish 20. Cleavage of the C16–N bond⁵⁴ could enable introduction of the 11-membered macrocycle.

Figure 5.

Revised strategy to construct the [2.2.2]-azabicycle.

The requisite cyclic allene precursors were synthesized, as shown in Figure 6. Precursor 21, accessible in one step from ketone 9 (see SI for details), was elaborated to nitrile-containing cyclic allene precursor 22a through a hydroboration/oxidation, oxidation, and nitrile formation sequence.⁵⁵ To access the other desired cyclic allene precursor, nitrile 22a was treated with the Ghaffar–Parkins catalyst to furnish primary amide 22b in 84% yield.⁵⁶ It is notable that the silyl group and triflate motif withstand the various reaction conditions shown in Figure 6, thus demonstrating the utility of these cyclic allene precursors in multistep synthesis.

Figure 6.

Synthesis of cyclic allene precursors 22a and 22b.

With cyclic allene precursors 22a and 22b in hand, we evaluated each compound in a strain-promoted Diels–Alder cycloaddition/retro-Diels–Alder cascade (Figure 7). Each substrate was independently subjected to pyrone 12 and CsF in acetonitrile. At ambient temperatures, as described previously, cyclic allene generation and Diels–Alder trapping occurred, leading to the formation of two new C–C bonds and installation of the C8a quaternary carbon. By directly heating the reactions to 80 °C, the subsequent CO₂ extrusion was achieved in one pot to generate trienes 23a and 23b in 51 and 27% yield,⁵⁷ respectively. The presumed mechanism is depicted for the reaction of amide 22b (22b → 25 → 26 →23b), with regio- and diastereoselectivity in the key cyclic allene trapping step, paralleling observations described earlier (see Figure 3). Despite the low yield of 23b, it is notable that the primary amide functional group is tolerated in this complexity-generating step given the known electrophilicity of strained cyclic allenes.⁵⁸ Although extensive efforts to access 24 using ImDA and N-acyl ImDA approaches were ultimately curtailed due to the exploration of an alternative approach, our current results underscore the synthetic utility of strained cyclic allene intermediates for the generation of complex scaffolds in the context of a total synthesis study.

Figure 7.

One-pot cycloaddition and CO₂ extrusion to generate trienes 23a and 23b.

Strained cyclic allenes have seen sparse use in total synthesis, especially in comparison to their close relatives, arynes, and cyclic alkynes. In this study, we have validated that these relatively understudied fleeting intermediates can be used to generate significant structural complexity in the context of a synthetic approach to the manzamine alkaloid keramaphidin B (1). Our key step involves the regio- and diastereoselective trapping of a strained azacyclic allene with a pyrone, leading to the formation of two new C–C bonds and three stereocenters, including a quaternary carbon center. These efforts should inform future synthetic design plans, including our ongoing efforts to assemble the bridged azadecalin core of keramaphidin B (1). Furthermore, we expect that these studies will help fuel the exploration of strained azacyclic allenes for the assembly of complex structural architectures.

Data Availability Statement

The data underlying this study are available in the published article and its Supporting Information.