Speeding up new enzyme design

Enzymes can provide a green alternative to many chemical reactions. But designing the right enzymes for the chemical industry is a complex and time-consuming job. University of Groningen scientists Hein Wijma and Robert Floor have developed a computer-based approach to enzyme design that can speed up the process enormously by screening ‘in silico’ rather than in the laboratory.

Chemical synthesis can be a dirty business, requiring high energy input and toxic compounds. Enzyme-based synthesis is much milder, usually taking place in water at room temperature, and without the need for toxic additives. But nature doesn’t provide all the enzymes that industry needs. Enzyme design is therefore an important area of research.

‘The basic technique used to produce new enzymes is directed evolution’, says Wijma. Scientists make a large library of mutants from a known enzyme, and test them for the required activity. ‘Testing thousands of mutants is very time consuming, which is why we use computational enzyme design.’ One way to select the best mutants is to use high-throughput molecular dynamic simulations to test how each mutant interacts with the substrate.

‘But these simulations take a lot of computer time’, says Wijma. He wondered whether shorter simulations would render useful results. ‘A normal simulation spans 22 nanoseconds. I tried a very short simulation of just 10 picoseconds [a picosecond is one thousandth of a nanosecond] which I meant as a negative control, because I wasn’t expecting any results.’

As it turned out, the 10-picosecond simulations were able to predict which mutants would interact in the required fashion with the substrate. The results were published in the journal Angewandte Chemie on 2 February. ‘We found that most of the action between enzyme and substrate occurs in time scales as small as that. By running several short simulations of the same enzyme/substrate combination, each with a slightly different starting point, we get a good idea of their interaction. That really took us by surprise.’

In the paper, Wijma applies this technique to an epoxide hydrolase enzyme (see below). He screened some 3,000 mutants on the computer. ‘This took a week-and-a-half on 100 cores of the University’s computer cluster. The full 22 nanosecond simulations could have taken up to three years.’

The screening revealed 37 promising mutants, which were constructed and tested in the lab. ‘Other researchers have tried to find the same sort of mutant. They relied on directed and tested 4,700 mutants in the lab. That probably took them a whole year.’ Wijma’s 37 mutants amount to less than 1 percent of this, making his laboratory testing process some 100 times faster. ‘And we got very similar results.’

The technique appears to work well. An earlier paper by Wijma describing the basic theory has already generated a lot of interest from colleagues and industry. ‘Now we can show that it is not just theory, but actually works in the lab.’

Reference: Wijma HJ, Floor RF, Bjelic S, Marrink SJ, Baker D, Janssen DB Enantioselective Enzymes by Computational Design and In Silico Screening Angew Chem Int Ed Engl, Advanced Online Publication.

On epoxide hydrolase enzymes

These enzymes react with epoxide groups, which are an environmentally-friendly alternative to similar reactive groups containing chlorine, bromine or iodine, all of which are toxic. The enzyme can change the epoxide into a diol in two ways, resulting in two products with different chirality.

Chirality is a form of asymmetry in molecules, comparable to the difference between a left hand (often marked S, for sinister, Latin for left) and a right hand (R). In pharmaceutical products, either the Sz or the R form is the active compound, and the mirror image does not contribute or can even cause serious side effects.

The aim of the experiments Wijma and his colleagues describe in the paper was to design enzymes that produce either the S or the R diol, starting with an enzyme that produces both.

(Illustration: Wikipedia)