A short total synthesis of (±)-epimeloscine and (±)-meloscine enabled by a cascade radical annulation of a divinylcyclopropane

Abstract

The first stereoselective synthesis of epimeloscine is accomplished in a longest linear sequence of 10 steps and with 13 total steps. The core of the synthesis takes only five steps, the key ones being acylation, stereoselective tandem radical cyclization of a divinylcyclopropane to make two rings, and group selective RCM of the resulting divinylcyclopentane to make the last ring.

Meloscine 1 is the parent of a small but important group of Melodinus alkaloids (Figure 1).¹ In nature, 1 and its less stable epimer epimeloscine 2 are thought to arise from scandine 3 by hydrolysis and decarboxylation. In turn, scandine arises from 18,19-dehydrotabersonine 4 by expansion of the B-ring and contraction of the C-ring.²

Figure 1.

Structures of meloscine alkaloids and a key synthetic intermediate of Bach and Mukai

The highly functionalized C-ring of meloscine with its four stereocenters (two of which are quaternary) presents a significant synthetic challenge. Overman met this challenge in 1989 with a 22-step synthesis that features a classic example of the aza-Cope Mannich reaction.³ Appealing syntheses of meloscine have been reported by Bach in 2008⁴ and very recently by Mukai.⁵ Bach made (+)-meloscine through key intermediate 5a, which was made by [2+2]-cycloaddition and ring expansion to construct rings B and C. Mukai made intermediate 5b by a Pauson-Khand cyclization. Both Bach and Mukai ultimately made the E-ring of meloscine by a ring-closing metathesis (RCM) reaction, but the sequences from 5a,b to the natural product took 10–11 steps.

The elegant syntheses of Bach and Mukai illustrate the challenge of late introduction of the E-ring with its C5 quaternary stereocenter. Herein we report an exceptionally short synthesis of the meloscines that constructs the B and C rings of an ABCD-ring product in a single step by a cascade radical annulation of a divinylcyclopropane. The subsequent synthesis of the E-ring is then expedited by the presence of the two vinyl groups that are essential for the radical cascade. As a bonus, the sequence produces exclusively (±)-epimeloscine, which is readily epimerized to (±)-meloscine.1a

Figure 2 shows our retrosynthetic analysis. Meloscine 1 or epimeloscine 2 should be readily available from 6 by an RCM end game à la Bach and Mukai but in just a few short steps (removal of the protecting group, N-allylation, RCM). Divinylcyclopentane 6 is the direct product of the cascade radical annulation of divinylcyclopropane 7. In turn, 7 is formed by acylation of aniline 8 by acid 9. Our recent work on radical cyclizations and ortho-alkenyl anilides⁶ and the early studies by several groups of radical annulations of monovinylcyclopropanes^7,8 supported the feasibility of this retrosynthetic analysis.

Figure 2.

Retrosynthetic analysis of meloscines based on RCM, a radical cascade and acylation

Divinylcyclopropane carboxylate 9 was readily prepared in five steps, as summarized in Scheme 1. Rhodium (II) catalyzed cyclopropanation of bis-benzyl ether 10 provided trisubstituted cyclopropane 11 in 66% yield. Hydrogenation of 11 with Pearlman's catalyst, followed by TEMPO oxidation of the resulting crude diol, afforded a dialdehyde 12 in 86% yield. Double Wittig reaction of 12 followed by hydrolysis of the crude product produced acid 9 in 60% yield over 2 steps.

Scheme 1.

Five-step synthesis of divinylcyclopropane carboxylate 9

The completion of the synthesis is shown in Scheme 2. Initially we pursued a traditional strategy with a protecting group on the anilide nitrogen. Aniline 13b (R = H) is a known compound that was prepared in three steps⁹ (see Supporting Information), then benzylated under standard conditions to provide 13a (R = Bn) in 73% yield. The acid chloride 14 was prepared in situ from 9 with Ghosez reagent (Me₂C=C(Cl)NMe₂),¹⁰ then aniline 13a was added to provide 15a in 83% yield. Importantly, this key precursor is expected to exist predominately as shown in the E rotamer,¹¹ and is therefore predisposed to undergo the first radical cyclization.¹²

Scheme 2.

Five- and six-step synthesis of (±)-epimeloscine and (±) meloscine from acid 9 and aniline 13b

Syringe pump addition of tributyltin hydride (2 equiv) and AIBN to a refluxing solution of 15a in toluene provided 16a in 55% yield after purification to remove the tin residues. This sole stereoisomer has the epimeloscine configuration at C3.

Construction of the E-ring followed on cue by removal of the N-Boc group with TFA and N-allylation to provide 17a, then RCM with the second-generation Grubbs-Hoveyda catalyst. As expected based on ring strain,⁵ only one of the two diastereotopic vinyl groups was engaged to provide pentacycle 18a in 94% yield. The synthesis then hit a minor roadblock when several pilot reactions to make epimeloscine by debenzylation of 18a were not successful.

Faced with the apparent choice of scaling up to make more 18a for renewed tries of debenzylation or with changing the N-protecting group to something easier to remove, we decided to do neither. Instead, we attempted to remove the N-protecting group entirely. This cuts two steps from the synthesis, but there is uncertainty because anilide 15b has a Z ground state geometry¹¹ so its derived radical is not predisposed for cyclization.¹² We elected this option since it was easy to prepare 15b.

Indeed, acylation of acid 9 and aniline 13b with Ghosez's reagent as before provided 15b in 77% yield. Now, syringe pump addition of tin hydride to 15b under the conditions optimized for 15a provided ABCD tetracycle 16b in 38% yield after careful purification. Removal of the Boc group and N-allylation provided 17b in 73% yield. Then RCM as above directly provided the natural product (±)-epimeloscine 2 in 89% yield.1a,c Overman produced epimeloscine in his classic synthesis,³ but from a minor stereoisomer (<10%) of a mixture on the way to meloscine. Thus, this is the first stereoselective synthesis of epimeloscine. Epimerization of 2 with KO^tBu provided (±)-meloscine 1 in 83% yield.

In summary, we have achieved the first stereoselective synthesis of (±)-epimeloscine 2 with a longest linear sequence of just 10 steps in about 6% overall yield. (±)-Meloscine 1 is readily produced from 2 by epimerization. The core part of the synthesis (Scheme 2)—coupling of two simple precursors (13b and 9) and rapid formation of rings C and D (in tandem) then ring E—takes just five steps and proceeds in almost 20% overall yield. There is room for improvement because yield of the radical cyclization (38%) is unoptimized. This core sequence features no oxidations, no reductions, no functional group transformations and only one deprotection (removal of the N-Boc group). The use of a divinylcyclopropane in the cascade radical annulation to make the C and D rings paves the way for immediate construction of ring E to complete the synthesis.