In situ recycling of chiral ligand and surplus nucleophile for a noncatalytic reaction: Amplification of process throughput in the asymmetric addition step of efavirenz (DMP 266)

The synthesis of efavirenz (DMP 266) involves a highly enantioselective asymmetric reaction of ketone 2a with lithium cyclopropylacetylide 3a in the presence of (1R,2S)-pyrrolidinylnorephedrine (PNE) 4b as the chiral mediator to produce 6b, a key advance intermediate bearing a tetrasubstituted chiral carbon. The major drawback of this reaction is that it requires at least 2 mol of cyclopropylacetylene (CPA) 3b, 2 mol of the chiral mediator 4b, and 4 mol of an n-alkyllithimn base to generate a 1 mol of addition product 6b. In a program to improve the cost-effectiveness of the process, we have studied this asymmetric addition reaction to reduce the stoichiometry of the chiral moderator and CPA nucleophile. The initial experimental designs to improve the stoichiometry were based on the assumption that the second equivalent of lithium cyclopropylacetylide simply acted as a base to deprotonate the N-H of ketones 2a or 2b, leaving 1 equiv of unlithiated and presumably unreactive CPA. It was reasoned that if we lithiate this CPA in the postreaction mixture, it would complex with the already lithiated ligand in the reaction which could then be used to convert additional 2a or 2b into product thus improving the stoichiometry. After some trial experiments, a dramatic decrease in stoichiometry was achieved and it was found that by simply adding one more equivalent of an n-alkyllithium to the system after the first reaction cycle, it was possible to obtain at least a 14 % throughput increase with just a 5% dilution of the reaction volume in this simple and more cost-efficient process. The chiral addition reactions could be run with multiple cycles in the same pot with the sequential addition of n-alkyllithium 5, CPA 3b, and ketone 2a. However, after four cycles, there was some decrease in the enantioselectivities (90.8%) that ultimately places a practical limit on the number of recycles possible. The process throughput increase can be explained in the light of a recent mechanistic investigation by Collum and co-workers. We postulate that the introduction of an n-alkyllithium to the reaction mixture after the completion of the first cycle regenerates the reactive cubic