Chemoenzymatic API synthesis

A chiral metal complex separates two mirror-image transition states by a few kilojoules per mole; it makes few contacts with the substrate and each contributes little. An enzyme makes dozens of contacts at once, and the productive orientation ends up being the only geometrically permitted one. The result is an enantiomeric excess that biocatalysis usually does not optimise so much as inherit: selectivity here is a property of the pocket, not of ligand tuning.

Transaminases and the sitagliptin case

Aminotransferases move an amino group through the cofactor pyridoxal-5’-phosphate. The substrate forms an external aldimine with PLP; deprotonation at the alpha carbon gives a resonance-stabilised quinonoid intermediate; tautomerisation releases the ketone and leaves the pyridoxamine form, which then aminates the acceptor. Stereochemistry is fixed by which face of the quinonoid plane the catalytic lysine can reach.

The canonical example is the sitagliptin route reported by Savile and colleagues in Science in 2010. The natural (R)-selective transaminase would not accept the bulky prochiral ketone at all; directed evolution produced a variant carrying 27 substitutions that runs at high substrate loading in the presence of an organic co-solvent. The step replaced an asymmetric hydrogenation over a rhodium catalyst under hydrogen pressure — and removing the metal removed the downstream step that had existed only to get the metal back out of the product.

Thermodynamics is the real constraint, not a detail

Transamination is close to thermodynamically neutral: the equilibrium constant is around unity, and left alone the reaction stalls at low conversion. An industrial process therefore always carries a mechanism for displacing that equilibrium — a large excess of isopropylamine with the acetone by-product stripped off, or coupling to lactate dehydrogenase, which removes pyruvate irreversibly.

Ketoreductases face the same logic. Reducing a ketone to a secondary alcohol consumes NADPH, and stoichiometric nicotinamide cofactor is economically impossible. The process works only with a recycling system: glucose dehydrogenase oxidising glucose to gluconolactone, whose irreversible hydrolysis pulls the cycle forward, or an alcohol dehydrogenase oxidising isopropanol to acetone that is then removed. The cofactor turnover number is what decides the cost of the step.

Where the ceiling sits

The first limit is solubility and medium. API substrates are typically lipophilic, and workable concentrations demand organic co-solvents that destabilise the protein. Either the enzyme is evolved for solvent tolerance, or it is immobilised on a support, where restricted chain mobility raises thermal stability and allows continuous-flow operation with the catalyst reused many times.

The second, and the decisive one, is narrow substrate scope. The same pocket that delivers selectivity fails when a substituent changes. There is no general-purpose “amine enzyme”: each new molecule needs its own screening and directed-evolution campaign. The cost of that campaign — weeks to months and tens of thousands of variants — is the actual barrier. Biocatalysis pays where volumes are large or the chemical alternative is unusually dirty, and does not pay in early development, where the route is still moving.

Regulatory treatment is ordinary: ICH Q11 requires the route to be justified and its critical parameters controlled, and the enzyme and its residues in the product become part of the specification.