While sometimes chemists get lucky and find the perfect enzyme straight away, more commonly a huge raft of them must be screened to find one that will carry out a reaction. And even when a ‘hit’ is found, it may not be ideal: perhaps the reaction is too slow, or needs to be run at very high dilution. However, in recent years chemists have become very skilled at giving nature a helping hand, by engineering enzymes to be even more selective and efficient.

The simplest way of doing this is by site-directed mutagenesis, where a specific base in a gene is altered, but the diversity this can introduce is limited. Much more diversity is available by using an error-prone version of the polymerase chain reaction that is routinely used to make strands of DNA. This is not ideal, either, as it is uncontrolled, and the diversity it can create is still limited.

Much more powerful is gene shuffling, where pieces of DNA are cut up and rearranged to make new genes. One way of doing this is the MolecularBreeding process developed by Codexis. This has been applied to great effect in the synthesis of the hydroxynitrile (R)-4-cyano-3-hydroxybutyrate, used to introduce the chiral sidechain of the world’s biggest selling drug, Pfizer’s Lipitor (atorvastatin). Demand for this intermediate is about 200 tonnes a year, so the cost savings of developing a faster, cheaper, more efficient process would be substantial.

The new enzymatic process, developed on behalf of Lonza, has been so successful that in June it was awarded the US Environmental Protection Agency’s Greener Reaction Conditions award. The improvements over traditional processes are dramatic: it saves a resolution step where the ‘wrong’ half of a racemic product is separated out giving a maximum yield of just 50%, and the need for hydrogen bromide and a strongly alkaline cyanation reaction are eliminated, along with a difficult purification involving high-vacuum fractional distillation.

Codexis’s Chris Davis explained how it was done at Scientific Update’s recent Industrial Biotransformations conference in Barcelona. The company’s MolecularBreeding process uses its proprietary DNAShuffling technology to create new enzymes. The gene that codes for an enzyme with some activity is split into fragments that recombine in a variety of different ways to create new genes. This leads to a library of novel genes retaining many of the characteristic of the parent. The new enzymes they code for are expressed, and those with better activity and specificity identified. The process can be repeated several times to give enzymes with ever better properties.

Codexis developed three separate enzymes for the hydroxynitrile process. In the first step, two of these catalyse the enantioselective reduction of the ketone ethyl 4-chloroacetate by glucose to form an enantiopure chlorohydrin. The third enzyme then catalyses a novel cyanation of this chlorohydrin under neutral conditions. The improvements over the old process are dramatic: the volumetric productivity of the reduction is improved 100-fold, and that of the cyanation is about 4,000 times better, with a 14-fold reduction in reaction time, a 7-fold increase in substrate loading, a 25-fold reduction in enzyme use and a 50% improvement in isolated yield.

The starting point for the cyanation enzyme was halohydrin dehydrogenase, explained Davis. Its activity was first described nearly 40 years ago, and while it had been applied with some success to the cyanation of epichlorohydrin, its usefulness was limited by its low volumetric productivity. By applying MolecularBreeding, interesting mutations could be identified and recombined to give libraries. ‘Which of the potential 300 mutations to the enzyme are good?’ he said. Combinatorial libraries were used to determine the beneficial mutations, and these are kept and the process started again.

‘Screening identifies genes with novel combinations of beneficial mutations, and fewer deleterious mutations,’ Davis said. ‘This way, rapid, large improvements are obtained.’ The enzyme finally used in the process has 37 mutations from the wild-type enzyme, amounting to 14.5% of the amino acid residues. Computer modelling indicates that each one of these mutations contributes to the improvement in the enzyme.

Many other companies and groups have developed directed evolution methods for improving enzyme activity. DSM, for example, has applied such a strategy to the enzyme 2-deoxyribose-5-phosphate aldolase, or DERA, developed by Chi-Huey Wong at the Scripps Institute in California for aldol reactions. However, its activity is not particularly high and it is easily inactivated by aldehyde substrates, so large quantities of the enzyme are needed to drive the reaction. It has great potential in the synthesis of intermediates such as statin sidechains, like that of Lipitor, so a more efficient variant would be very useful

DSM’s Marcel Wubbolts told the Barcelona conference error-prone PCR was used to create random genetic diversity in the enzyme to counteract the sensitivity to aldehyde substrates. The DERA variant library was screened for more productive enzymes, and directed evolution was then applied. Around 10,000 randomly chosen DERA variant clones were screened, and 63 hits found. After two rounds of recombination – screening 3,000 clones each time – 10 good hits were pinpointed. This allowed mutation ‘hotspots’ to be identified which had a positive effect on stability. A similar process was carried out to find out which parts of the enzyme would improve productivity, and nine hits were identified.

By combining the beneficial mutations from both screenings, Wubbolts said, a much more effective strain was discovered. They also found the mutations from both screenings are synergistic. ‘DERA variants with improved stability towards the reagents and with more than 10-fold increased productivity for the synthesis of statin intermediates were obtained,’ he said. And they are potentially of much more widespread use than just statin intermediates. ‘Applications of aldolases are emerging as these enzymes enable the asymmetric synthesis of multiple chiral centres starting from simple and low-cost starting materials.’

Daicel Chemical Industry has also had success at directed evolution, as Brian Freer of its European subsidiary Chiral Technologies Europe told the conference. While naturally occurring L-amino acids are readily – and cheaply – prepared by bioprocesses, unsurprisingly this is not the case for the unnatural D-amino acids. A number are important starting points for pharmaceuticals, but as they are much more expensive than their natural analogues, enzymatic processes to make them directly would be advantageous.

The aim was to make D-tryptophan using L-tryptophan as the starting point, because it is cheap. The process involves first a racemising acetylation, Freer explained, followed by a bioselective deacetylation to separate out the desired D-isomer, and then re-racemising the remaining acetylated L-isomer.

‘A wild-type enzyme was found which is capable of deacetylating several acetyl amino acids, including acetyl tryptophan,’ Freer said. ‘However, the wild-type enzyme is inhibited by the two isomers of tryptophan. We therefore had to improve it.’ Error-prone PCR was first used to determine which amino acid would be effective if changed. PCR was then used to carry out site-specific mutagenesis at this point. The third generation of mutation involved optimisation by the combination of mutation sites. The result of this process was an enzyme mutant that is not inhibited by any of the substrates or products, Freer said. ‘The process has been scaled, and is being prepared for piloting.’