Aminotransferases don’t work alone, though—they require a sacrificial partner, 2-ketoglutarate, to accept the amino acid’s lost amine. This usually creates a mixture of amino acids and ketoacids that settle into an unhelpful equilibrium.

Wu’s team relied on two more enzymes to shift this equilibrium, so that all of the starting amino acid was converted into its 2-ketoacid, and only catalytic amounts of 2-ketoglutarate were needed. “That’s quite an elegant feature,” O’Reilly says.

Once the reaction was complete, the researchers could then add more enzymes to the same pot to transform the 2-ketoacids into grams of high-value compounds, including N-methylamino acids and D-amino acids, which are being used to build “mirror” peptides and proteins as potential therapeutic drugs.

The researchers aim to improve the activity of their aminotransferases by tweaking their structures, in the hope that they could be used to manufacture 2-ketoacids on a larger scale. “We are also planning to use this enzyme cascade to convert L-amino acids into more diverse high-value compounds, such as N-heterocycles and vitamins,” Wu says.

The researchers aimed to exploit the growing availability of cheap amino acids, made by the fermentation of microorganisms. Global amino acid production hit 8.5 million metric tons in 2017, making them an increasingly useful raw material in chemical synthesis.

In contrast, 2-ketoacids are more difficult to produce by fermentation because they slow cell growth and tend to be unstable inside cells. Consequently, chemists primarily make these compounds by lab processes that may require expensive catalysts or costly starting materials.

Bian Wu and colleagues at the Chinese Academy of Sciences’ Institute of Microbiology hoped that aminotransferase enzymes could provide a more efficient route to compounds such as 2-ketoisocaproate, which can treat inflammatory kidney disease, and the animal feed additive 2-keto-3-methylvalerate.

To find an optimal enzyme that might aid large-scale 2-ketoacid synthesis, they set out on a computational fishing expedition in protein sequence and structure databases. That search identified more than 13,000 proteins that had features characteristic of aminotransferases. The researchers then mapped the enzymes into six broad families, based on shared traits such as the amino acid sequences around the active sites.

“Our goal was primarily to find an enzyme that could accept many different amino acids,” Wu says. The researchers reasoned that enzymes that didn’t fit neatly within a single family were more likely to offer this broader activity. So they picked 27 of these outliers, expressed them inEscherichia coli, and screened them for activity against 18 amino acids.

Although this didn’t identify a single universal aminotransferase, it did narrow the list down to a handy toolbox of just seven enzymes. By selecting a suitable enzyme from this group, the team could convert almost any L-amino acid into its 2-ketoacid with more than 99% yield, in some cases at gram scale. “The nice thing about this type of database mining is that you are pulling out catalysts that we don’t know anything about yet,” says Elaine O’Reilly of University College Dublin, who develops enzymes for use in synthesis.