Highly functionalized donor–acceptor cyclopropanes applied toward the synthesis of the Melodinus alkaloids

Abstract

A series of highly substituted vinylcyclopropanes were prepared and examined as reaction partners in a palladium-catalyzed (3 + 2) cycloaddition with nitrostyrenes. Described herein are our efforts to synthesize an elusive 1,1-divinylcyclopropane by several distinct approaches, and to apply surrogates of this fragment toward the synthesis of the Melodinus alkaloids.

The Melodinus alkaloids are a class of dihydroquinolinone natural products related to the Aspidosperma alkaloids through an oxidative rearrangement of dehydrotabersonine (1, Scheme 1).^1,2 Despite their lack of known biological activity,^3,4 the structural complexity of the Melodinus alkaloids and the prospects of preparing non-natural derivatives for biological evaluation were both extremely appealing to our lab.

Scheme 1.

Proposed biosynthesis of the Melodinus alkaloids.

In the case of (+)-scandine (3),¹ (+)-meloscandonine (4),⁵ and others,⁶ three of the four contiguous stereocenters on the characteristic central cyclopentane ring are quaternary. To date, the only members of the family to have been synthesized are meloscine (5) and epimeloscine (6), both of which possess only two quaternary stereocenters on the central C ring.^7–9 It is hypothesized that (+)-scandine (3) is the biosynthetic precursor to the other Melodinus alkaloids.² Thus, we began to pursue the synthesis of scandine (3), which could allow access to the related dihydroquinolinone natural products.

In planning a concise synthesis, we chose to exploit elements of symmetry found within the target natural product. In particular, the quaternary stereocenter at C(20) bears two olefinic substituents, and C(16) bears two carbon substituents in the carboxylic acid oxidation state. Accordingly, after disconnection of the E ring via benzylic C–H insertion, we envisioned that the D and B rings of 7 could be formed by substrate-controlled diastereoselective ring-closing metathesis and lactamization steps of divinylcyclopentane 8 (Scheme 2). This intermediate could arise, in turn, from nitrocyclopentane 9, the product of a transition metal catalyzed, intermolecular formal (3 + 2) cycloaddition between a trans-β-nitrostyrene (10) and divinylcyclopropane 11.¹⁰

Scheme 2.

Retrosynthetic analysis of scandine (3).

At the outset of our synthetic efforts, we examined several possible approaches toward the synthesis of the desired divinylcyclopropane (11, Scheme 3). The geminal vinyl groups could potentially be installed through substitution of 1,1-dihalocyclopropane 12,¹¹ itself generated from a dihalocarbene 13 and methylidene dimethylmalonate (14).¹² Alternatively, the two vinyl groups could be formed by elimination from cyclopropane 15, derived from the reaction of olefin 17 with a malonate-derived carbenoid (16). Finally, we envisioned utilizing an S_N2′ displacement of alkylidene cyclopropane 18 with a vinyl nucleophile. This cyclopropane could be synthesized from allene 19.

Scheme 3.

Retrosynthetic analyses of cyclopropane 11.

We first examined the use of a 1,1-dihalocyclopropane (e.g. 12) toward divinylcyclopropane 11 (Pathway A, Scheme 3). The synthesis and reactions of these building blocks have been extensively researched.¹² 1,1-Dihalocyclopropanes are known to react with dialkyl cuprates,¹³ trialkyl zincates,¹⁴ manganates,¹⁵ or magnesates¹⁶ to yield alkylated cyclopropylmetals, which can react with an electrophile to deliver products with geminal substitution. Furthermore, the cyclopropylmetal intermediates can be used in metal-catalyzed cross-coupling reactions with vinyl halides to deliver vinylcyclopropanes.¹⁵

Due to the highly reactive nature of methylidene dimethylmalonate (14),¹⁷ we sought to first examine the vinylation of gem-dihalocyclopropanes using a reduced substrate. Accordingly, acrylate derivative 20 was prepared by a known procedure and protected as a silyl ether (21, Scheme 4).¹⁸ Olefin 21 was then cyclopropanated using phase-transfer catalysis to afford gem-dibromocyclopropane 22.

Scheme 4.

Synthesis of reduced gem-dihalocyclopropane 22.

Unfortunately, efforts to directly vinylate cyclopropane 22 failed (Scheme 5). A Stille coupling with tetravinyltin was unsuccessful, as was the palladium-catalyzed cross coupling of the in situ-generated organomanganate with vinyl bromide.^15b An attempt at a bis-alkynylation using Sonogashira coupling was also unfruitful. Since no desired substitution products were observed with this substrate, we did not pursue this route further and we shifted our focus to an alternative approach.

Scheme 5.

Efforts to substitute dibromocyclopropane 22.

We turned our attention toward the formation of the desired vinyl groups by elimination of two leaving groups (Pathway B, Scheme 3). In this vein, we set out to prepare dimesylate 30 as a divinylcyclopropane precursor (Scheme 6). Baylis–Hillman reaction of methyl vinyl ketone (25) with acetaldehyde by a known procedure furnished adduct 26 which was then reduced to afford diol 27 as a mixture of diastereomers.¹⁹ Although this substrate underwent mesylation cleanly, the product (28) was unstable as a neat oil, and underwent spontaneous, rapid decomposition.²⁰ Furthermore, when a solution of the dimesylate in dichloromethane was subjected directly to cyclopropanation with diazodimethylmalonate, a complex reaction mixture was observed and no desired cyclopropane product (30) could be isolated.

Scheme 6.

Synthetic approach to elimination substrate 30.

To avoid problems of substrate stability, we opted to protect diol 27 as a disilyl ether (31, Scheme 7). This substrate was cyclopropanated efficiently using Du Bois’ catalyst to give cyclopropane 32, which was immediately subjected to alcohol deprotection under acidic conditions,²¹ however, one hydroxyl group underwent an undesired lactonization to give bicyclic lactone 33. This product was mesylated and eliminated to yield vinylcyclopropane 35. Although an interesting structure, we were not able to advance lactone 35 to divinylcyclopropane diester 11.

Scheme 7.

Vinylcyclopropane synthesis via diol 27.

Finally, we examined a route to the divinylcyclopropane through S_N2′ displacement of a substituted alkylidenecyclopropane (Pathway C, Scheme 3). De Meijere and coworkers have demonstrated that vinylcyclopropanes and methylenecyclopropanes with allylic leaving groups will react under palladium catalysis to form a common palladium allyl intermediate, which can then be alkylated.²²

We sought to prepare an analogous alkylidenecyclopropane bearing the necessary methyl ester functionalities. Beginning with the known homoallenyl acetate 36,²³ we screened cyclopropanation conditions using diazodimethylmalonate (29), examining several catalysts, carbenoid precursor equivalents, and addition times (Scheme 8). On our first attempt (entry 1), we were able to isolate the desired alkylidenecyclopropane (37) in 42% yield, although an excess of allene 36 was required. While using an excess of the diazo compound lowered the yield (entry 2), increasing the catalyst loading and the equivalents of the diazo improved the yield to 58% (entry 3). Increasing or decreasing the slow addition rate of the diazo reagent had a detrimental effect on the yield (entries 4 and 5). Changing the catalyst to the electron-poor trifluoroacetate complex resulted in a mixture of products (entry 6), and use of the electron-rich caprolactamate complex gave low conversion of the starting material (entry 7). Microwave heating of a neat mixture of the reaction components (entry 8) afforded considerably shortened reaction times, however, the yield was not improved. Finally, the use of Du Bois’ catalyst (Rh₂(esp)₂) gave the highest isolated yield (80% yield, entry 9), with a short reaction time, low catalyst loading, and no need for syringe-pump addition of the diazodimethylmalonate.²⁴

Scheme 8.

Cyclopropanation of homoallenyl acetate 36.

With the desired alkylidenecyclopropane 37 in hand, we examined an array of allylic substitution conditions with vinyl nucleophiles, including those reported by de Meijere,²² as well as other catalytic systems with vinyl alanes and cuprates (Scheme 9). Unfortunately, in all cases, none of the desired divinylcyclopropane 11 was observed, and only ring-opened products were obtained.^25,26 It is possible that the diester functionality serves to weaken the distal bond of the methylenecyclopropane, favoring ring-opening rather than substitution. We did find, however, that we could smoothly remove the acetate protecting group through a two-step procedure from homoallenyl acetate 36 to furnish primary allylic alcohol 38 in 94% yield (Scheme 10).

Scheme 9.

Attempted vinylation of diester cyclopropane 37.

Scheme 10.

Synthesis of primary allylic alcohol 38.

At this stage, we considered that the use of a Claisen rearrangement might offer an alternative pathway to install the desired quaternary carbon on the cyclopropane (Scheme 11a).²⁷ The use of Claisen rearrangements to install vicinal quaternary centers is well precedented.²⁸ Furthermore, the relief of ringstrain (i.e., from alkylidenecyclopropane to cyclopropane) was predicted to aid the efficiency of the C–C bond formation. However, we envisioned potential chemoselectivity and side-reactivity problems in the conversion of Claisen product 40 to the desired divinylcyclopropane (11). Particularly, conditions would be necessary that could reduce the product carbonyl in the presence of the methyl esters and prevent concomitant lactonization.

Scheme 11.

Proposed Claisen rearrangement routes.

Accordingly, we turned to the Eschenmoser–Claisen reaction, since numerous examples exist in the literature for chemoselective reduction of amides in the presence of esters,²⁹ and the resulting tertiary amines (42) would not be expected to react with the pendent ester functionalities and can be converted to olefins by means of the Cope³⁰ or Hofmann³¹ elimination (Scheme 11b).

We therefore treated alcohol 38 under typical reaction conditions with dimethylacetamide dimethyl acetal, and observed the formation of amide 43 in moderate yield (Scheme 12).³² The main side product of the reaction was conjugated amide 44, likely formed by base-promoted ring opening of the desired product, and extensive screening of reaction temperatures and times could not improve the yield of the desired vinylcyclopropane 43. Amide 43 was reduced with alane to dimethylamine 45 in 36% yield. Efforts to eliminate the amine (45) to form the desired divinylcyclopropane (11) have been unsuccessful to date. Fortunately, our efforts to this point provided three unique vinylcyclopropanes (35, 43, and 45) which we could examine in the palladium-catalyzed (3 + 2) reaction.

Scheme 12.

Eschenmoser-Claisen rearrangement of 38.

With three highly functionalized vinylcyclopropanes in hand, we set out to determine their compatibility with palladium-catalyzed (3 + 2) cycloaddition conditions originally developed by Tsuji.¹⁰ Under an array of conditions, no cyclopentane products could be isolated (Scheme 13). In the case of dimethylamide substituted cyclopropane 43, the starting material was isomerized in high yield to conjugated amide 44 as a mixture of olefin isomers. Dimethylamine analogue 45 and bicyclic vinylcyclopropane 35 showed no reactivity, even at elevated temperatures.

Scheme 13.

Palladium-catalyzed (3 + 2) cycloaddition attempts.

The isomerization of dimethylamide 43 is attributed to the presence of acidic protons on the substrate: upon formation of the palladium(II) allyl species (48), the pendant malonate acts as a base, eliminating Pd(0) via deprotonation to give conjugated amide 44 (Scheme 14).

Scheme 14.

Mechanistic rationale for the formation of amide 44.

As for vinylcyclopropanes 35 and 45, we propose that the lack of reactivity results from a demanding allylation step of the catalytic cycle (Scheme 15). Whereas hard nucleophiles such as Grignard reagents typically add to the more highly substituted terminus of the allyl fragment under palladium catalysis via an inner-sphere mechanism, soft nucleophiles often attack at the least-substituted position through an outer-sphere mechanism. In the case of an unsubstituted vinylcyclopropane (3 + 2) cycloaddition (i.e. 49, R = H), conformational effects in the ring closure presumably override this innate selectivity, resulting in addition to the more highly substituted internal position of the allyl fragment. However, in the case of our substituted vinylcyclopropanes (i.e. R ≠ H), the steric demand is possibly too high to form the desired cyclopentane product (50) under these conditions.

Scheme 15.

Rationale for the lack of desired reactivity of highly substituted vinylcyclopropanes.

In the course of our studies, Curran and Zhang completed the total syntheses of (±)-meloscine (5), (±)-epimeloscine (6), and several unnatural analogs by a route similar to our own original strategy (Scheme 16).^7d,7i They were able to construct necessary divinylcyclopropane 55 through a tandem oxidation-Wittig methylenation sequence from cyclopropane 53. After coupling of acid 55 with aniline 57, the core tetracycle 59 was formed via an intramolecular radical-mediated cycloaddition and quickly advanced to epimeloscine (6) and meloscine (5). Scandine (3), the parent of the natural product family, was not accessed via this route but the similarity of their approach to our own original pathway, as well as the challenges we faced in effecting a transition metal catalyzed intermolecular (3 + 2) cycloaddition encouraged us to modify our synthetic plan.

Scheme 16.

Total syntheses of epimeloscine and meloscine by Curran and Zhang (ref. 7d).

The primary revision to our retrosynthesis involves using a monovinylcyclopropane (63) in the palladium-catalyzed (3 + 2) cycloaddition, and appending the second vinyl group at a later stage by C–H functionalization (Scheme 17).

Scheme 17.

Revised retrosynthetic analysis.

In 2011, we disclosed our progress toward scandine, using the palladium-catalyzed intermolecular (3 + 2) cycloaddition strategy as planned in our revised retrosynthesis (Scheme 18).³³ We were able to synthesize monovinylcyclopropane 63 from dimethylmalonate (64) and dibromide 65 via a known procedure.³⁴ The subseuqent palladium-catalyzed (3 + 2) cycloaddition of cyclopropane 63 and nitrostyrene 47 proceeded smoothly. Tandem reduction and lactamization provided tricycle 67 as a 2:1 mixture of diastereomers at C(20) in favor of the undesired stereoisomer. Nevertheless, after reductive amination, acetylation, and ring-closing metathesis, we were able to access the tetracyclic ABCD ring system of the Melodinus alkaloids (72) in only six steps from commercial sources.

Scheme 18.

Assembly of the ABCD ring system.

In summary, efforts to synthesize and apply a 1,1-divinylcyclopropane toward the total synthesis of scandine are described. Furthermore, we have applied a monovinylcyclopropane toward the preparation of a tetracyclic precursor to scandine via a palladium-catalyzed (3 + 2) cycloaddition. The remaining challenges to overcome in the synthesis include E ring closure by benzylic C–H insertion and installation of the C(20) vinyl group. Finally, the derivatization of scandine to other members of the natural product family will be examined.