Liang, Lalonde, Borup, Mitchell, Mundorff, Trinh, Kochrekar, Cherat, Pai. Development of a Biocatalytic Process as an Alternative to the (−)-DIP-Cl-Mediated Asymmetric Reduction of a Key Intermediate of Montelukast. Org. Process Res. Dev. 2010, 14, 193-198. DOI: 10.1021/op900272d
This article from researchers at Codexis describes the development of a biocatalytic (i.e. enzyme-catalyzed) method for creating the lone stereocenter in the synthesis of montelukast sodium, aka Merck’s asthma drug Singulair. The original Merck process route includes an enantioselective ketone reduction using a boron reagent derived from alpha-pinene called (-)-DIP-Cl. The reaction works well: high yield, high enantioselectivity (although still requiring a recrystallization step to upgrade from ~95% to 99% ee), and (-)-DIP-Cl is made in one step from cheap starting materials. The downside is that at least 1.5 equivalents of (-)-DIP-Cl must be used, and the reagent is moisture sensitive and corrosive. Codexis, being in the enzyme business, decided to find an enzyme that would catalyze this same reaction.
The ketoreductase class of enzymes will catalyze this type of transformation, using the cofactor NADPH as a hydride source. Using a stoichiometric quantity of NADPH as a reductant is undesirable from an atom-economy standpoint in addition to being unreasonably expensive, so regeneration of NADPH is necessary. The authors envisioned a process in which the oxidation of excess isopropanol regenerates NADPH, catalyzed by the same enzyme that reduces their substrate. The overall transformation would be similar to a MPV reduction, or a transfer hydrogenation.
The authors screened ketoreductases from their in-house collection, as well as commercially available enzymes, and remarkably they found several of their enzymes that produced the desired enantiomer of the reduced product in >99.9% ee. Despite the high enantioselectivity, these initial hits had low catalytic activity and thermal stability, and were inactivated by the organic cosolvents needed to dissolve the starting material. In their directed-evolution studies, enzyme-catalyzed reactions were conducted in a 1:5:3 THF/IPA/water solvent system and variants were selected for enhanced yield and activity at higher temperatures. Since a co-catalytic amount of NADPH was added to these reactions, this automatically selected for variants that could use IPA for cofactor regeneration. After three rounds of mutations and screening, they had an enzyme whose product output was increased 400-fold over the starting enzyme. (I won’t try to do justice to all the interesting science behind directed evolution, but the first few slides of this presentation gives a nice overview, and lots of references). Swapping THF for toluene increased product output by 2.5 times to reach their target of 100g/L over 24 hours, and two additional rounds of evolution allowed them to halve either the reaction time or the catalyst loading. Fortunately, the 19 mutations accumulated over the course of the directed evolution did not erode their enantioselectivity at all – their final biocatalyst still yields products with >99.9 %ee.
Over the course of the reaction, the monohydrate of the reduced product crystallizes out of solution, which drives the reaction to completion (interestingly, the anhydrous material is not a solid). It also simplifies product isolation – the reaction mixture is filtered, and the crude product is obtained in 95% yield, 98.5% purity and perfect ee. The process also scales well, up to 230 kg substrate (!) without any drop in yield or ee.
The authors go on to compare their biocatalytic process to the original (-)-DIP-Cl reduction. The biocatalytic process gives a slightly higher yield (90-98% vs 85-90%), requires a lower catalyst/reagent loading (3-5% vs 150% by weight), and does not require recrystallization to achieve >99% ee (DIP-Cl reaction does). They also compared the two reactions using a metric called process mass intensity(PMI), which is the the mass of all material inputs divided by the mass of the product (similar, but not the same as the E-factor, which divides the mass of waste by the mass of product): The Codexis reaction is the clear winner with a PMI of 18 kg input per kg product, compared to the PMI of 52 for the original Merck process. If the product of the biocatalytic reaction is extracted and then recrystallized (which they do for the reaction run on 200kg scale for some reason), the PMI is 34. This really goes to show that the method of purification matters almost as much as the reaction conditions – and no purification is the best of all!
As great as this all sounds, there is the question of whether similar gains could be made switching from stoichiometric DIP-Cl to a transition metal-catalyzed hydrogenation reaction. The authors mention the results of three such reactions in the footnotes, which all seem somewhat promising. The first two reactions, developed by the Noyori group and the Lonza company, would require product recrystallization in order to upgrade the enantioselectivity – a process that added 16 to the PMI value of the Codexis reaction. For the final reaction (developed by the Pliva company), if they could work out conditions where they precipitate the pure product straight from the reaction mixture, it might be an attractive alternative. There’s also the issue of whether it’s better to use a biocatalyst (whose preparation involves some amount of material inputs for expression and purification) or a ruthenium complex (obtaining the metal itself sounds pretty resource intensive, not to mention ligand synthesis etc). In any case, this biocatalytic method is certainly more attractive than the existing protocol, and looks like it could find use soon on an industrial scale.