The use of enzymes is having a major impact on the synthesis of pharmaceutical ingredients in terms of yield. Dr Sarah Houlton highlights some recent successes.
The ability to harness the power of nature in the form of enzymes has had a real impact on the synthesis of pharmaceutical ingredients in recent years. These white biotechnology applications have made it possible to improve some syntheses – often dramatically – and even make some molecules that would otherwise be either impossible or impractical to synthesise.
Many standard chemical reactions can now be achieved using enzyme catalysts. One recent example from the labs of CLEA Technologies in Delft, Netherlands, is the use of Pseudomonas stutzeri lipase, or PSL, to carry out aminolysis reactions.1 It proved particularly effective on sterically hindered substrates. The current common techniques for carrying out aminolysis reactions predominantly use acid chlorides or coupling reagents. These tend to be non-catalytic and not very atom efficient, with toxic or corrosive reagents or by-products. Using an enzyme catalyst would improve the process’s green credentials.
The researchers took a range of bulky methyl esters and amines as substrates – the acyl donors were activated as methyl esters to avoid salts forming between the amine and the carboxylic acid. For comparison, the same substrates were also treated with Novozym 435, an immobilised Candida Antarctica lipase B. While the CAL-B enzyme was more efficient in those aminolysis reactions where methyl 2-phenylpropionate was the acyl donor, for secondary amines PSL worked better, as it did with piperidine.
Neither worked with t-butylamine, nor diisopropylamine, nor esters that were bulky at the alpha position to the carbonyl group. The team speculates that the relatively broad substrate scope shown by the lipase in terms of both the acyl and amine partners might enable this to become a green and general method for the synthesis of amines under mild conditions.
Molecular biology tools such as directed evolution could also be used to improve the enzyme’s productivity and substrate specificity above those seen with the wild type enzyme.
Transaminases are another important class of enzyme because of their ability to make the enantiopure amines that are common intermediates in the pharmaceutical sector. Most transaminases found in nature are selective for S-stereochemistry, and attempts to genetically engineer synthetically useful R-selective ω-transaminases have met with only limited success. Thus, biocatalytic approaches to R-amines have largely relied on the kinetic resolution of the racemic mixture, using S-selective transaminases to remove the ‘wrong’ isomer. But of course this leaves the maximum possible yield at 50%.
Most transaminases found in nature are selective for S-stereochemistry, and attempts to genetically engineer synthetically useful R-selective ω-transaminases have met with only limited success
A group at the University of Graz in Austria has been looking for further R-selective transaminases to add to the one that is commercially available, isolated from a species of arthrobacter.2 They started with three wild type R-selective Ω-transaminases, for which there was little data on stereoselectivity or substrate range, isolated from Hyphomonas neptunium, Aspergillus terreus and Arthrobacter sp. The test reaction involved the amination of prochiral ketones using D-alanine as the amine donor, but the pyruvate that is formed must be removed from the reaction mixture by a second enzymatic process, either by using an alanine dehydrogenase enzyme to give ketoreduction, or by alanine dehydrogenase mediated reductive amination.
A range of prochiral ketone substrates were tested. The Arthrobacter-derived enzyme worked for all those tested except a-tetralone, in each case giving the desired amine as the R-enantiomer with an ee in excess of 99%. The enzyme from Aspergillus terreus converted three of the test substrates.
Meanwhile, by starting with the previous commercially available enzyme, after 11 rounds of mutations a transaminase that accepts bulky ketones was developed. However, this does not work with alanine as the amine donor. By using a large excess of 2-propylamine instead, the reaction was successful, with acetone produced as the co-product. For all four of these new enzymes, DMSO could be added as a cosolvent to improve the solubility of the substrates, in some cases giving a better conversion than without it.
A great example of how biocatalysis can be used to improve a process comes out of the Pfizer labs at Sandwich, UK.3 The original synthesis of a potent β-3 agonist designed to treat overactive bladder involved a Noyori asymmetric reduction that gave only a modest ee of 77%; this meant a preparative reversed-phase chiral HPLC was necessary to purify the chlorohydrin made in this reaction and, as a result, the overall yield was just 46%. Various other reduction conditions, such as homogenous asymmetric hydrogenation and DIP-Cl asymmetric reductions proved no better.
A second problem was that of the synthesis of an alcohol intermediate used in the tin chemistry, which is far from ideal from the standpoints of handling, cleaning and waste disposal.
To get around these problems, the researchers replaced the Noyori reduction with a ketoreductase catalysed asymmetric ketone reduction, and used a lipase-catalysed route involving desymmetrisation instead of the tin chemistry. For the first, a panel of 260 ketoreductases was screened. An enzyme from Julich looked promising, but on scaling up the reaction stopped at 60% conversion. The next attempt involved KRED-130 from Codexis. Although the conversion had been better in the screen, the ee it produced was not that good. However, when it was tried on a 1g scale, complete conversion was seen in five hours, and the R-chlorohydrin product isolated in 97% yield, and an ee of 95%, which was much better than they had expected. Finally, they tried their own aldehyde reductase mutant from Sporobolomyces salmonicolor, again at a 1g scale, giving 85% yield of product in 97% ee.
For the tin replacement reaction, they looked to take a commercially available nitrile, which was straightforward to hydrolyse to the diacid, and then esterify to the diethylester in the same pot, which could then be desymmetrised to the monoester, and then elaborated to the desired alcohol intermediate. There was literature precedent for the desymmetrisation, but it involved hydrochloric acid at 140°C, and a biocatalytic process would be milder and more environmentally friendly. This time, they identified 15 commercially available hydrolases that could selectively hydrolyse the diester to the monoester, some of which also made small quantities of diacid.
The first ‘hit’ was Meito lipase MY from Candida cylindracea, and it was tested on a preparative scale. Although after 21 hours HPLC showed 94% conversion to the monoacid, only 49% could be isolated, and they thought the ‘missing’ product might be left behind in the biocatalyst residues. Another promising enzyme, Lipozyme TL 100L from Novozymes, which is a Thermomyces lanuginosus lipase, was tried, and this time it was possible to avoid product entrainment during the workup process. It was scaled up to 200g, with the monoacid being isolated in 94% yield. With no chromatographic separations or highly toxic reagents, this route is a significant improvement on the original tin mediated process.
Sometimes an enzymatic resolution remains the only sensible way to prepare a chiral intermediate, perhaps because there is no obvious starting point for developing an enzyme to carry out the reaction in an enantioselective fashion. In these cases, the best strategy can be to use a dynamic kinetic resolution, where the ‘wrong’ enantiomer is recycled, and optimise the enzyme’s properties so that the yield is maximised, while giving a product with an extremely good ee. This is the strategy adopted by a group at Merck’s process research department in New Jersey, US.
The synthesis of the cathepsin K inhibitor odancatib, being developed for the treatment of osteoporosis, involves a biocatalytic dynamic kinetic resolution to make a chiral fluoroleucine intermediate.4 Initially they used the immobilised CAL-B lipase Novozym 435, but while this gave reasonable activity, its stability was not good enough to facilitate a commercially viable process, and none of the other commercially available immobilised CAL-B enzymes were good enough under their process conditions, either.
Sometimes an enzymatic resolution remains the only sensible way to prepare a chiral intermediate, perhaps because there is no obvious starting point for developing an enzyme to carry out the reaction in an enantioselective fashion
Thus, they looked to create their own immobilised CAL-B that worked better in their system. The starting point was five different Sepabead resins from Mitsubishi, which had a variety of different compositions and functional groups, allowing both covalent and hydrophobic immobilisation binding techniques. They incubated liquid CAL-B from Novozymes with each of these resins at room temperature for 24 hours, before filtering off the resin, washing with a buffer, and then drying under vacuum.
Each of the five resins was then tested for activity, with the best results being seen with the enzyme immobilised by hydrophobic binding on the polymethacrylate resin EXE-120. Compared with Novozym 435, it had a 50% higher specific activity. In addition, after 48 hours the activity of the commercial enzyme had dropped to just 6% of its initial value under process conditions. In contrast, the EXE-120 immobilised enzyme retained 94% of its activity after this time. It was also significantly more stable in a continuous packed bed plug flow reactor, and gave better product yield and ee – 95% and 88%, compared with 90% and 86%. As it is more stable, enzyme to substrate loading could be reduced from 1:20 for the commercial enzyme to below 1:100.
They also looked at alternative substrates, and for each of the seven aromatic secondary alcohols and amines they tried, an ee in excess of 99% was obtained after an 18-hour run. Furthermore, the resin could be recharged with further enzyme after it had deactivated in the process, further reducing costs.
As well as immobilising enzymes onto beads, it is also possible to immobilise them by encapsulation. For example, biomimetic silica mineralisation can be used to trap them within a silica support. A group at the University of Toulouse in France has taken enzymes and encapsulated them in silica particles, which altered both the activity and enantioselectivity.5 They took two enzymes – Pseudomonas fluorescens esterase I (PFE-I) and CAL-B lipase – and fused them directly with the silica-precipitating R5 peptide from Cylindrotheca fusiformis, which promotes silica condensation. At first, they had tried this peptide along with the polycationic polymer polyethyleneimine to catalyse the silica deposition, but the encapsulation yields were less than 20%, with the majority of the enzyme protein remaining in solution. Instead, they created a recombinant fusion protein between the R5 peptide and the protein they were looking to immobilise, expressing PFE-I from E. coli and CAL-B from the oleaginous yeast Yarrowia lipolytica.
The purified recombinant proteins were then mixed with the silicate precursor tetramethyl-orthosilicate, hydrolysed in HCl. The resulting silica precipitates were separated by centrifugation before being washed with water and freeze dried. They found by examining the remaining supernatant that in both cases more than 95% of protein had been immobilised within the silica particles.
They then looked at the efficiency of the encapsulated enzymes at carrying out an enantioselective process. The test reaction was the esterification of racemic 3-phenylbutanoic acid – a common chiral synthon – with ethanol. They were surprised to find no conversion with CAL-B-R5, despite its activity in previous test reactions. However, PFE-I-R5 not only gave better conversion and a higher enantioselectivity than the free enzyme, it also gave the opposite enantiomer. They believe this is the first time such an inversion after encapsulation has been reported, and they postulate that the silica matrix encourages entrapment in conformations that give the opposite enantiomer to the free enzyme.
Flow reactors are increasingly being used in chemical synthesis because of the advantages they offer in areas such as improved mixing and reducing explosion risk from exothermic reactions. However, they have found much less use in the area of biocatalysis because of the challenges presented by the heterogeneous slurry nature of the reaction mixture, and problems with gas and liquid handling.
Flow reactors have found much less use in biocatalysis because of the challenges presented by the heterogeneous slurry nature of the reaction mixture, and problems with gas and liquid handling
Residence times generally need to be longer for a bioreaction, and the multiphase nature of the process makes it extremely tricky to use a traditional tubular reactor. To address this, a team from the Biochemist project co-funded by the UK Technology Strategy Board has been looking to develop continuous manufacturing methods for biocatalytic reactions.6 The partners in the project are Ingenza, CTech Innovation and AM Technology.
Most large-scale flow reactors have one of two designs – either they use static mixing, or dynamic. A reactor using static mixing relies on turbulent flow or baffles on the side of the reactor tubes; these are typically narrow and thus with a heterogeneous biocatalytic reaction there is a tendency for the reactor to block, the phases to separate, or the mixing to be suboptimal. The alternative dynamically mixed reactors use mechanical stirrers, giving efficient mixing with no reliance on the nature of the flow through the reactor. The team used a Coflore agitated tube reactor, which has loose reactor elements combined with shaking to give mixing, which also overcomes some of the problems encountered with a traditional rotating mixer within the tube.
The test reaction was the dynamic kinetic resolution of racemic amino acids. Selective oxidation of the D-amino acid within the racemate gives a mixture of L-amino acid plus α-ketoacid, with the catalyst for the oxidation being an oxidase enzyme expressed by Pichia pastoris, and flavin adenine dinucleotide as a redox cofactor. Oxygen is also needed as a co-substrate, and is introduced into the reactor via a sparged gas inlet.
As long as there is sufficient enzyme in the reaction mixture, the main hindrance to a good reaction rate is the presence of oxygen, so efficient mixing is important
As long as there is sufficient enzyme in the reaction mixture, the main hindrance to a good reaction rate is the presence of oxygen, so efficient mixing is important. Indeed, as the agitator speed was increased in a batch vessel, the reaction rate increased. However, as the batches are scaled up, the reaction rate falls, and the limited enzyme life means there is a significant productivity drop.
In the Coflore reactor, a 1-litre scale reaction had a three times higher reaction rate than that carried out in a 1-litre batch reactor. When scaled up to 10L, the reaction rate remains largely unchanged – altering the tube length has little impact on the mixing and gas distribution, compared with a batch reactor where such parameters change dramatically as the vessel increases in size. Importantly, the solid material such as organic material and dead cells that can block the narrow tubes of a statically mixed flow reactor caused no problem, and no blockage problems were seen, which the team put down to the high mixing efficiency and large tube diameters within the Coflore reactor.
1. S. van Pelt et al. Green Chem. 2011, 13, 1791
2. F.G. Mutti et al. Adv. Synth. Catal. 2011, 353, 3227
3. M. Badland et al. Green Chem. 201, 13, 2888
4. M.D. Truppo and G. Hughes, Org. Proc. Res. Dev. 2011, 15, 1033
5. S. Emond et al. Chem. Commun. 2012, 48, 1314
6. G. Gasparini et al. Org. Proc. Res. Dev. in press doi 10.1021/OP2003612I