Total Synthesis of (+)-Arborisidine

Abstract

The first total synthesis of arborisidine, a unique Kopsia indole alkaloid possessing a fully substituted cyclohexanone ring system with two quaternary carbons, has been achieved in seven steps in racemic format from tryptamine and in nine steps in asymmetric format from D-tryptophan methyl ester. Key elements of the design include a carefully orchestrated decyanation protocol to finalize the asymmetric formation of an aza-quaternary center that is challenging to access in optically active format via direct Pictet–Spengler cyclizations with tryptamine, a metal-promoted 6-endo-dig cyclization of an enyne to establish the second core quaternary center, and regiospecific functionalizations of the resultant complex diene to finalize the target structure. The distinct and efficient nature of the developed solution is highlighted by several unsuccessful approaches and unexpected rearrangements.

Over the past several years, efforts to achieve the laboratory synthesis of an array of caged indolenine alkaloids drawn from several plant species have served as a powerful vehicle for driving innovation. Indeed, targets such as scholarisine A (1, Scheme 1),¹ strictamine (2),² and arboridinine (3),³ among others,⁴ have proven to be strong tests of available chemical tools, with their constrained frameworks typically necessitating the development of new reactions and synthetic strategies if they are to be fashioned effectively. Arborisidine (4)⁵ is a recent and valuable addition to this collection, having been isolated and characterized in 2016 in relatively minute quantities (0.2 mg/kg bark) from Malayan Kopsia arborea trees⁵ and shown in a patent disclosure to inhibit gastric cancer in vivo when used in combination with pimelautide.⁶ While 4 shares some general structural features with other caged alkaloids, such as 1–3, its most striking element is its fully substituted cyclohexanone core that includes two quaternary centers, one of which is an aza-quaternary carbon that has not yet been observed in related molecules. As highlighted by the only published study toward this target,⁷ that core system provides significant synthetic challenges in ways that might not be obvious upon initial inspection. Herein, we describe the first racemic and asymmetric syntheses of this target in seven and nine steps, respectively, through a plan featuring several regiospecific and chemoselective events to appropriately fashion its distinctive structural elements.

Scheme 1.

Structures of Several Alkaloids Containing a Shared Tetracyclic Indolenine Framework and a Generalized Retrosynthetic Plan for Arborisidine (4)

As shown in Scheme 1, our design for arborisidine (4) was predicated on forming its pyrrolidine ring system last. For that event, we anticipated the need to regiospecifically functionalize the exocyclic alkene of diene 5, either intermolecularly as controlled by stereoelectronic effects or intramolecularly as directed by the adjacent secondary amine. Our desire to use this substrate was based on lessons learned in failed attempts to effect late-stage C–C bond formation from compounds of type 7 (similarly described by Qin)⁷ as well as allylic functionalization of 6, likely due to the considerable conformational strain and steric bulk surrounding the starred methylenes. As a result, we envisioned that the exocyclic olefin of 5 could serve to circumvent these issues by pushing the reactive center away from the steric bulk of the aza-quaternary carbon. To fashion that diene, we projected that a 6-endo-dig cyclization^2c,e using enyne 8 could rise to the occasion, potentially obtained from aminoketone 9 through nucleophilic addition and alcohol dehydration. As a result, the total synthesis of (+)-4 would then require an asymmetric synthesis of the lone quaternary center within 9, an event of high challenge given the paucity of effective enantioselective Pictet–Spengler reactions using ketones.^8,9 However, as we observed in our own efforts with targets such as strictamine (2),^2e the challenges found in forging chiral centers at these positions (as in 10 → 11) and then effecting cyclization chemistries (as in 11 → 12) can also afford powerful opportunities for reaction discovery and development.

Our initial preparation of aminoketone 9 in racemic format proceeded without incident simply by stirring tryptamine (13) and 2,3-butadione in acidic MeOH at 60 °C,¹⁰ affording 9 in moderate yield (50%) in the gram-scale quantities needed to probe the remaining elements of the sequence (Scheme 2). By contrast, the means to achieve an asymmetric synthesis of the same compound was non-obvious based on existing Pictet–Spengler precedent, since no substrates of this type in a chiral format have been prepared directly; the closest analogies are a few examples that incorporated α-ketoamides⁸ with others utilizing cyclic N-acyliminium intermediates.⁹ Following initial failures to identify a de novo catalytic solution using varied Lewis and phosphoric acid promoters, we wondered if the use of D-tryptophan methyl ester (14) could provide an efficient diastereoselective alternative. Strong precedent to that general effect in Pictet–Spengler reactions between both D-/L-tryptophan methyl esters with aldehydes was known,¹¹ but not with diketones.¹² Pleasingly, moderate diastereoselectivity could in fact be achieved, favoring the formation of 15 versus its separable diastereomer in a 1.9:1 ratio, leading to an isolated yield of 62% for 15. Its structure and absolute configuration were confirmed by X-ray analysis of ent-15 formed from initial experiments using L-tryptophan methyl ester.

Scheme 2. Racemic and Enantioselective Syntheses of Aminoketone 9^a.

^aReagents and conditions: (a) 2,3-butadione (1.1 equiv), TFA (1.0 equiv), MeOH, 60 °C, 20 h, 50%; (b) 2,3-butadione (2.5 equiv), MeOH, 65 °C, 20 h, 95%, 1.9:1 dr, 62% desired diastereomer; (c) NH₃ in MeOH, 23 °C, 15 h; then concentrate; then Et₃N (2.0 equiv), TFAA (1.0 equiv), THF, 23 °C, 1 h; Et₃N (1.0 equiv), TFAA (0.5 equiv), 23 °C, 1.5 h, 96%; (d) 4-trifluoromethylbenzaldehyde (2.6 equiv), NaBH₃CN (2.0 equiv), MeOH/THF, 65 °C, 8 h, 71%.

Despite that success, the added steric bulk and electron-withdrawing effects of the ester and ketone groups α to the amine made subsequent chemistry quite difficult to achieve. First, efforts to protect the amine were universally non-successful, likely leading to the failures observed when common decarboxylation methods¹³ were screened directly with 15. Thus, we elected to transform that ester into the nitrile of 17 through a one-pot aminolysis/dehydration process,^11b hoping to effect a seemingly standard decyanation reaction,¹⁴ through reduction of the imine formed via cyanide expulsion, without impacting the ketone. Again, such chemoselectivity proved challenging but was necessary since if the ketone was reduced, its re-oxidation could not be achieved smoothly in the presence of the free amine. Extensive screening (see Supporting Information (SI) section for details) ultimately illustrated that NaBH₃CN¹⁵ could rise to the occasion, yet the desired product (9) was formed only in low yield, as the majority of the material converted instead into the aromatized natural product harmalane (18, Table 1, entry 1). That aromatization could not be suppressed even with degassed solution and the addition of BHT as a radical scavenger. Such results led us to believe that the expelled cyanide ion might be involved in the aromatization process in a redox-neutral pathway similar to a retro-benzoin condensation as shown in Scheme 3 via intermediate 20.

Table 1.

Screening of Conditions To Effect Enhanced Preparations of 9 and Minimize the Formation of Aromatic Side Product 18

entrya	additive	reaction time (h)	yield of 9 (%)b
1	none	8	<20c
2	C6H5CHO	3	50
3	C6H5CHO	8	56
4	4-MeOC6H4CHO	8	42
5	4-ClC6H4CHO	8	46
6	4-CF3C6H4CHO	8	69
7	4-CF3C6H4CHOd,e	8	71
8	BnOH	8	13c

^aConditions: substrate (0.2–0.3 mmol), NaBH₃CN (2.0 equiv, 1.0 M in THF), additive (3.0 equiv), MeOH (0.1 M), 65 °C.

^bIsolated yield.

^cThe major product is 18.

^dPerformed on 5 mmol scale.

^e2.6 equiv of aldehyde was used.

Scheme 3.

Proposed Mechanism for the Reductive Decyanation in the Presence of Aromatic Aldehydes

To test this hypothesis, we added benzaldehyde as a cyanide scavenger and, gratifyingly, the yield of aminoketone 9 increased to 50% (Table 1, entry 2). In support of our overall mechanistic hypothesis, the cyanohydrin of benzaldehyde was observed in the ¹H NMR spectrum of the crude mixture as was the deuterated cyanohydrin of the expelled methyl ketone when no reductant was present. As shown in the remainder of Table 1, further optimization of the scavenger revealed that more electron-deficient aldehydes were superior, with 4-trifluoromethyl benzaldehyde working best in 69% and 71% yield on test and gram-scale runs, respectively (entries 6 and 7). Finally, although full or partial reduction of the added aldehyde derivatives was observed in all of the experiments listed in entries 2–7, those alcohols had no impact in preventing aromatization as revealed by the control experiment using only added benzyl alcohol (entry 8).

With both racemic and asymmetric routes to 9 secured, we focused next on forming cyclic dienes of type 5 (cf. Scheme 1). As shown in Scheme 4, that goal could be achieved in three steps through propyne addition, dehydration of the resultant alcohol as promoted by trifluoroacetic anhydride (TFAA) (a step that also protected the amine), and a 6-endo-dig cyclization^2c,e promoted by a mixture of catalytic amounts of both Au(I) and Ag(I) salts in MeOH at 40 °C. While such endo selectivity has been commonly observed in other explorations using similar enynes,¹⁶ most systems studied to date underwent further rearrangements post-cyclization to afford products devoid of an sp³-hybridized quaternary center. These steps proceeded here in 39% overall yield on gram scale, noting that the yield values shown refer to reactions performed with racemic material; chiral material was advanced as well in commensurate yields, albeit on smaller scales (see SI).

Scheme 4. Completion of the Total Synthesis of Arborisidine (4)^a.

^aReagents and conditions: (a) 1-propynyllithium (4.0 equiv), THF, −78 °C, 2 h; (b) pyridine (3.5 equiv), TFAA (2.2 equiv), CH₂Cl₂, −78 to 23 °C, 2 h, 53% for two steps; (c) Ph₃PAuCl (0.1 equiv), AgBF₄ (0.1 equiv), MeOH, 40 °C, 17 h, 74%; (d) Br(coll)₂PF₆ (1.02 equiv), CH₂Cl₂, −78 to 23 °C, 1 h; then concentrate; then Pd(OAc)₂ (0.02 equiv), Xantphos (0.04 equiv), dioxane/MeOH/Et₃N, CO (balloon), 70 °C, 13 h, 56%; (e) Mn(dpm)₃ (0.5 equiv), PhSiH₃ (2.0 equiv), i-PrOH/(CH₂Cl)₂, 23 °C, 10 h; PhSiH₃ (1.0 equiv), 23 °C, 7 h; PhSiH₃ (1.0 equiv), 23 °C, 4 h; then concentrate; then NaBH₄ (10 equiv), MeOH, 23 °C, 1 h; 100 °C, 9 h, 50%, imine/amine = 1:2; (f) BH₃·THF (5.0 equiv), 23 °C, 12 h; then H₂O (2.0 equiv), Me₃NO· 2H₂O (1.1 equiv), THF, 65 °C, 2 h; then concentrate; then PhIO (1.5 equiv), CH₂Cl₂, 23 °C, 2.5 h; then NaHCO₃ (5.0 equiv), Dess–Martin periodinane (1.5 equiv), 23 °C, 1 h, 34% 4, 40% 29; (g) NaHCO₃ (5.0 equiv), Dess–Martin periodinane (1.0 equiv), CH₂Cl₂, 23 °C, 30 min, 74%; (h) Raney Ni, THF, H₂ (balloon), 23 °C, 86%, 25:26= 10:1. Br(coll)₂PF₆ = bis(2,4,6-trimethylpyridine)bromine(I) hexafluorophosphate.

Differentiation of the diene system was now needed. Initial efforts using polarity-inverse radical additions¹⁷ led to either recovery of starting material or decomposition as a result of the generally harsh conditions used for these processes. Fortunately, a regiospecific bromination of the diene motif at its exocyclic terminus could be achieved under mild conditions using bis(2,4,6-trimethylpyridine)bromine(I) hexafluorophosphate;¹⁸ following solvent removal and re-dissolution in 1,4-dioxane, that intermediate could be converted directly into methyl ester 24 in 56% yield through a Pd-catalyzed carbonylation using Xantphos.¹⁹ Of note, the isolated yield of vinyl bromide 23 was lower than this one-pot bromination/carbonylation sequence. We attribute this result to the possible formation of an allylic bromide which was unstable to column purification, but which under the carbonylation conditions could funnel to the desired dienoate through isomerization.

From here, we next sought to achieve a regioselective 1,4-reduction to generate 26; such an operation was viewed as feasible based on three-dimensional models showing that the β-position within this framework should be more accessible than the β-position, with the hydride source adding from the less-hindered Si-face. However, conventional dienoate hydrogenation^20a,20b,20c and conjugate reduction^20d methods either provided no conversion, no selectivity, or, in the case of Raney Nickel,^20e predominately 1,6-reduction to generate 25 (structure confirmed by X-ray crystallographic analysis) with a 10:1 preference over 26. Inspired by the pioneering work of Magnus^21a,21b and Shenvi^21c in manganese-catalyzed hydrogen atom transfer (HAT) processes, a reaction that was previously of high use to us in the synthesis of another complex alkaloid,^21d we were pleased to find that such conditions could deliver the desired 1,4-reduction product exclusively. Extensive optimization (see SI for more details) showed that the use of Mn(dpm)₃ in a 50 mol% loading in combination with PhSiH₃ gave the best conversion, using trace air as the activator^21b of the catalyst. When conducted on larger scales (>50 mg), the portion-wise addition of PhSiH₃ in excessive amounts was necessary to achieve full conversion, but at the price of the extra silane causing challenges in purification. As a result, the crude mixture of 26 was treated directly, following initial solvent evaporation, with methanolic NaBH₄. This operation excised the trifluoroacetamide protecting group, effected lactam formation, and afforded partial reduction of the indolenine system. The resultant mixture of imine and amine products (separable for purposes of characterization but also readily purified as a mixture) was then exposed to a carefully executed sequence in hope of affording arborisidine (4) directly. Those operations included initial treatment with excess BH₃ in THF to effect alkene hydroboration and full imine and lactam reduction. After quenching any remaining BH₃ with H₂O, the resultant alkylborane was oxidized into a secondary alcohol using Me₃NO·2H₂O,²² with PhIO²³ then added to oxidize the dihydroindole domain, and Dess–Martin periodinane (DMP)²⁴ added last to oxidize the secondary alcohol. Upon quenching, this one-pot operation afforded arborisidine (4) in 34% yield along with partially oxidized 29 in 40% yield. As such, a seven-step racemic synthesis (and nine-step asymmetric synthesis) of the target was achieved, with material supplies of the natural product enhanced by the separate oxidation of 29 using DMP. Pleasingly, all spectral and optical rotation data of 4 matched that of the natural sample as reported by Kam and co-workers⁵ with its structure further confirmed by X-ray analysis. To date, close to 50 mg of 4 has been accumulated (~40 mg racemic, ~5 mg enantioenriched) from all runs of the sequence.

As one example of the uniqueness of the developed solution in terms of advancing forward from diene 22 through intermolecular chemo- and regioselective transformations, significant efforts were also made to reach the target through intramolecular cyclizations as tethered by the adjoining N-atom of the aza-quaternary center. Scheme 5 documents one such attempt, where an intramolecular nitrile-oxide [3+2]-cycloaddition²⁵ with partial N-acylation could smoothly deliver a mixture of 32 and 33. Unfortunately, subsequent N–O bond cleavage of 32 using freshly prepared Mo(CO)₃(MeCN)₃,²⁶ while successful, also initiated an unexpected Wagner–Meerwein-type rearrangement affording 37 as the sole identifiable product (confirmed by X-ray). Attempts to ring-contract this material were not successful.

Scheme 5. Attempted Intramolecular [3+2]-Cycloaddition and an Unexpected Rearrangement^a.

^aReagents and conditions: (a) K₂CO₃ (4.0 equiv), MeOH, 60 °C, 4 h; (b) Et₃N (5.0 equiv), AcOCH₂CH₂NO₂(1.1 equiv), THF, 23 °C, 2 h; then 4-DMAP (0.2 equiv), Boc₂O (2.0 equiv), 23 °C, 20 h; 4-DMAP (0.2 equiv), Boc₂O (2.6 equiv), 23 °C, 7 h, 77% for two steps, 32:33 = 2.3:1; (c) Mo(CO)₃(MeCN)₃ (4.0 equiv), MeCN/CH₂Cl₂/H₂O, 23 °C, 1 h, 59% from 32.

In conclusion, this work highlights a number of unique operations leading to the first total synthesis of the caged alkaloid arborisidine (4). Most critical among those steps were (1) a chemoselective reductive decyanation strategy which affords access to chiral 1,1-disubstituted 3,4-dihydro-β-carbolines that are currently inaccessible via direct asymmetric Pictet–Spengler reactions with diketones, (2) an effective 6-endo-dig cyclization of an enyne leading to a cyclic diene, (3) regiospecific differentiation and functionalization of that resultant diene system, and (4) a carefully orchestrated oxidation state adjustment achieved in one pot. Overall, the brevity of the developed solution should inform strategies to access other alkaloids with hopefully similar expediency.