Enzymes to the rescue

Biocatalysis and enzyme-based racemic separation are finding increasing use in process chemistry. Sarah Houlton highlights how these technologies are finding application in the production of APIs

An enormous proportion of the molecules found in nature are chirally pure - ranging from simple amino acids and sugars through to the beautiful and hugely complex molecules that are made by bacteria, plants and other organisms.

It is hardly surprising, therefore, that nature is so much better at chiral synthesis than we are. However, in recent years, chemists have increasingly been taking a leaf out of nature's book, using bacteria, cells and isolated enzymes to catalyse chemical reactions and separations to give single enantiomer products. They can also be used to effect synthetic transformations that are difficult - or even impossible - to achieve by purely chemical means.

Biocatalysts have, essentially, been applied to small molecule synthesis in two areas. The first is as more traditional catalysts, giving single enantiomer products in high ees and good yields; the second is in the separation of racemates, where an enzyme is used to split one enantiomer from the other. Both are finding increasing use in process chemistry and are being carried out at production scale.

chirally pure

Because enzymes are designed by nature to give single isomers, if a suitable enzyme is available it is an appealing option for chemists looking to make chirally pure molecules. It's rarely as simple as picking an enzyme off a shelf and adding it to a reaction, though - the right strain has to be identified, not to mention the right solvent and other conditions to give high yields and ee, plus good turnover. If the enzyme can be recycled and used more than once, then so much the better for the economics of the process.

Valrubicin is a derivative of the anthracycline antibiotic doxorubicin. This latter compound is used as an antineoplastic agent in a variety of cancers, but it has a number of disadvantages, including dose-dependent cardiotoxicity and acute myelo-suppression. The derivative valrubicin has lower cardiotoxicity in animals. It acts as an inhibitor of topoisomerase II and RNA synthesis and is indicated as a treatment for refractory bladder cancer. Existing chemical syntheses of the drug give low yields - 24% and 35% - and chemists at Albany Molecular in Albany, NY, US have developed a synthetic route that gives an overall yield of 79%.¹

key step

The key step is a lipase catalysed esterification of N-trifluoroacetyl doxorubicin. More than 80 commercially available lipases and proteases were screened in a variety of solvents, using either valeric acid or vinyl butyrate as the acid donor. Ultimately, a combination of Chirazyme L2,C3 - an immobilised Candida antarctica lipase - with valeric acid was identified as the best combination in an automated enzyme screen with mass spectrometric analysis. Because the enzyme is immobilised it is easy to remove from the reaction mixture by filtration or centrifugation, and it can be recycled. It even allows the possibility for the process to be carried out continuously. No chromatographic separation is necessary. As an added bonus, they found that the technique was widely applicable and around 60 doxorubicin 14-ester derivatives were made for biological testing.

The drug clofibrate is in widespread use for the treatment of hyperlipidaemia and atherosclerosis, and for the prevention of heart failure. A team at the University of Bari in Italy has been investigating a range of clofibrate analogues and discovered that yeasts could be used to effect the stereoselective reduction of prochiral ketones in their synthesis.² They report that while baker's yeast is commonly used to reduce carbonyl groups, a number of other yeasts can be used too, and can give better results. They looked at a group of para-chlorophenol oxoesters and, despite the structural similarity of the substrates, found that the same yeast organism did not give best results for all of them. Yields were from 88% to quantitative, and diastereoselectivities of up to 99% were obtained (Figure 1).

yeast option

Yeast has also been applied to the reduction of 1-acetonaphthone by a team at the National Institute of Pharmaceutical Education and Research in Punjab, India.³ They identified a new yeast strain, Candida viswanathii MTC 5158, isolated from soil, that could effect the transformation of 1-acetonaphthone to S-(-)-1-(1"-naphthyl) ethanol, an intermediate in the synthetic sequence to make mevinic acid analogue, a potential inhibitor of 3-hydrogy methyl glutaryl coenzyme A reductase (Figure 2).

The strain is an oxidoreductase, and was isolated using an enrichment and isolation technique. It can be used in whole cell format, which has one big advantage over isolated NADH-dependent carbonyl reductase in reduction reactions: the cells regenerate the cofactor, removing the need to include a regeneration process in the reaction.

Other biocatalytic reductions had already been reported for 1-acetonaphthone, but in all cases the reaction times were long and conversions relatively low. Using the new strain reaction times were reduced to just 12 hours, the substrate tolerance of 2g/l was higher and the cell concentration was lower at 200g/l. Yields in excess of 97% were achieved, along with an ee of 99%.

(R)-Pantolactone is a precursor of vitamin B5 and is made on vast scale for uses in pharmaceuticals, as well as in food and cosmetics. However, current manufacturing processes involve complicated routes that include the resolution of a racemate or an oxidation, followed by an asymmetric hydrogenation. A group at Warsaw University of Technology, Poland, has developed a process that uses oxynitrilase catalysis in an asymmetric hydrocyanation.⁴ It shortens the synthetic sequence, removing both the resolution and redox reactions, although the yields and ees are not that great.

kernel of hope

Oxynitrilases are present in a number of plants, and in the past an isolated and purified (R)-oxynitrilase from almonds has been used, either in solution or immobilised on a support. The Polish team have been investigating the potential of using unpurified meals made from almonds, as well as apple and plum kernels. Almonds remained the best, with a yield of 50% and an ee of 30%.

Often, a good and robust process exists for making an API or intermediate, but the product it creates is racemic. It can often be much quicker to create a process that makes a racemate rather than a single isomer and then separate the two components, and in the early stages of drug development when time is of the essence, a route that involves a separation can often be the best method. However, the big drawback of separation is that the maximum achievable yield is 50% - the other 50% is the "wrong" isomer.

There are several methods that can be used to separate enantiomers, the simplest being to scale up the lab chromatographic separation. However, while this is fine in the early stages of the development process when speed is vital, it can require prodigious quantities of solvents, which is expensive and far from practical. If a suitable enzyme can be identified that carries out some form of chemical transformation selectively on one enantiomer, then this can be used to differentiate the two. It is also sometimes possible to recycle the "wrong" isomer into the "right" one by using some form of racemisation reaction.

A team at Pfizer in La Jolla, CA, US, has developed a method for the preparation of optically active secondary amines that uses a protease enzyme to carry out a kinetic resolution.⁵ General, efficient and practical routes to chirally pure secondary amines remain a synthetic challenge, despite the frequency with which the moiety appears in drug molecules, and dynamic resolution remains the method of choice for making them.

Enzyme-catalysed resolutions proved difficult, in all likelihood because of steric hindrance at the enzyme active site, but the Pfizer team found that oxalamic esters were amenable to enzyme action. The first step is the acylation of a racemic cyclic amine with ethyl chlorooxoacetate to give an oxalamic ester. After trying a number of different enzymes, they found a number that selectively hydrolyse the ester bond rather than the amide in the R-enantiomer, leaving the S-enantiomer unchanged.

The best results were achieved with Aspergillus protease. On separation, the ester was removed with HCl, giving the two enantiomers in 43% and 45% yields respectively, with 96% ee for the former and 99% for the latter. The conditions can be applied to a number of different amines, including substituted piperidines, pyrrolidines and alkyl-aryl amines.

The Pfizer scientists have applied this enzymatic separation to the investigational glycinamide ribonucleotide formyltransferase (GARFT) inhibitor pelitrexol, which has good antiangiogenesis activity.⁶ The process used to make the compound for early stage trials was a lengthy, inefficient 20 step linear synthesis and produced an overall yield of less than 2%.

Although a convergent synthesis was later developed, this gave a racemic product that was heavily reliant on a chromatographic separation of diastereomers in the penultimate step, which has a disastrous effect on the overall yield. This is quite apart from the enormous amounts of solvents required - neither diastereomer is particularly soluble in the common organic solvents, so hundreds of litres were required for every kilogram of material. Separation and recovery were both poor too, and less than 25% of the required product was obtained.

effect amplification

The problem they faced was the remoteness of the stereocentre in the molecule from the racemic carbon - it is six bonds away. However, by applying their oxalamic ester technique, the increased size amplified the difference in chirality and they found that the inexpensive lipase Candida antarctica lipase B selectively de-esterified the S-terminal ester. As an added bonus, the oxalamic ester greatly increased the solubility, having a dramatic effect on the overall throughput because of the higher concentration. (Figure 3).

ß-Amino alcohols are a common structural feature in a variety of drug classes, including antibiotics, and especially ß-adrenergic blocking agents such as propranolol and atenolol. These chiral aryloxy propanol amines bear a strong structural resemblance to the hormone noradrenaline, and the S-enantiomer is the active one. Historically they have been marketed as racemates because classical separations are rarely straightforward and the off isomer tends to cause few side-effects. Chiral syntheses exist, such as processes that use D-mannitol as a starting point, but resolution remains the preferred technique. Now enzymatic routes have been developed.

simple separation

An enzymatic resolution for an intermediate to these drugs has been developed by scientists at the Indian Institute of Chemical Technology in Hyderabad.⁷ They found that immobilised forms of the lipase from Pseudomonas cepacia can be used to separate the two enantiomers of the nitrile precursors of propranolol, alprenol and moprolol. The racemic ß-hydroxy nitriles were made by ring opening of the corresponding epoxides. These were then reacted with the lipase and vinyl acetate, which selectively forms the acetate ester of the R enantiomer, leaving the S enantiomer untouched. Separation is then simple to give the desired enantiomer in good ee. (Fig. 4).

1,2-O-Isopropylglycerol, or solketal, is a useful precursor for a variety of drug classes, including prostaglandins, leukotrienes and ß-adrenoceptor antagonists. It is cheap and easy to make racemically, and enzymatic processes exist for the production of the S isomer by resolution through the selective hydrolysis of the acetate ester with esterases from Bacillus species. A group in Italy has developed a method that gives the R enantiomer instead.⁸ They used whole cells of the yeast Kluveromyces marxianus and achieved good enantioselectivity. They also found it was possible to produce multigramme quantities of the desired enantiomer by carrying out the process in an ultrafiltration membrane reactor.

A team at Merck has developed an enzyme catalysed kinetic resolution to make an indole-ethyl derivative, an intermediate required for the synthesis of a prostaglandin D2 receptor antagonist being developed to treat allergic rhinitis.⁹ The initial chemical synthesis was acceptable up to kilogramme scale, but its resolution step suffered from a large loss of material. As this resolution is towards the end of the synthetic sequence, it means large amounts of material have to be carried through from the early stages of the synthesis to make up for the loss. If the wrong isomer could be recycled into the correct one, then much less material would have to be made earlier on in the route. Making the resolution step much earlier in the synthesis would also be of great benefit here.

The first step in the original route, a Fischer indole reaction, was already known to give fully racemised products, so the next step was investigated for the separation. The indole ester was screened against a library of enzymes and the lipase from Pseudomonas fluorescens was identified as the best option, along with a triphasic solvent system consisting of dimethyl formamide, hexane and an aqueous buffer to overcome the problem of the lack of solubility of the organic substrate in the aqueous medium required for the enzymatic reaction. The desired R-enantiomer was left as the ester in 98% ee, and the unwanted S-enantiomer was hydrolysed to the alcohol in 97% ee (Figure 5).

cost savings

The team found that the off isomer could be epimerised to the R-enantiomer in the same pot by the addition of a catalytic amount of sulphuric acid, giving complete racemisation and esterification at once. Costs could also be saved by recycling the enzyme - after a lyophilisation process, the recovered enzyme lost no catalytic potential after at least three sequential cycles of use. The process has been carried out on a multi-kilogramme scale.