Building an enzyme means arranging atoms so that a reaction can happen repeatedly. A promising-looking protein shape is only the beginning. Research by Anna Lauko, Samuel J. Pellock, Kiera H. Sumida and colleagues tackles this challenge by designing serine hydrolases around their catalytic machinery.
When EnzymeNews first reported the work in September 2024, it was a preprint. The peer-reviewed study appeared online in Science on 13 February 2025. Its published results provide a fuller account of what the designs achieved and why some worked better than others.
Designing serine hydrolases from the active site outward
The team’s publicly available workflow starts with small arrangements of catalytic residues. RFdiffusion generates protein backbones around these arrangements, while LigandMPNN helps assign amino acid sequences. Structure-prediction checks then assess whether a proposed sequence is compatible with the intended fold.
Another component, PLACER, examines possible arrangements of the active site and a small molecule. Applying it to different reaction states helps identify designs whose catalytic groups remain suitably positioned throughout the process. This is the name used in the published study; the earlier report referred to ChemNet.
The distinction between a fold check and a reaction-state check matters. A protein can adopt the intended overall shape while placing one crucial side chain badly. Likewise, a useful arrangement at the start of a reaction may become unsuitable after the substrate changes chemically.
Why completing the cycle matters
Serine hydrolases use a reactive serine residue to attack a substrate. During ester hydrolysis, part of that substrate becomes temporarily attached to the enzyme. Water must then help release it so the enzyme can act again.
The researchers tested designs with progressively more elaborate active sites. Simple designs could show chemical reactivity without sustained turnover. More complete catalytic arrangements, including a serine–histidine–aspartate triad and an oxyanion hole that stabilizes charge during the reaction, supported repeated catalysis.
The published study reported catalytic efficiencies up to 2.2 × 10⁵ M⁻¹ s⁻¹. That highest value concerned the model ester 4-methylumbelliferyl phenylacetate. It is not a measurement of plastic-bottle breakdown, nor a universal performance rating for the designs.
Experimental structures put the designs to the test
Crystal structures closely matched selected computational models. One deposited example is win31, recorded as PDB 9DEH, with an X-ray structure determined at 2.21 Å resolution. Its coordinates support the molecular illustration accompanying this updated article.
A solved structure and a kinetic assay answer different questions: one tests the arrangement of the protein, while the other measures its performance under specified conditions. Together they offer stronger evidence than a computer-generated model alone.
A useful advance, with application work still ahead
The work shows a route to constructing and evaluating catalytic machinery beyond an existing natural enzyme scaffold. It does not establish a ready-to-use industrial process or medical treatment. Those applications would require their own substrate, stability, production and operating-condition tests.
For enzyme engineers, the practical lesson is to assess the entire catalytic cycle. Successful binding or a brief burst of product formation does not by itself demonstrate a reusable catalyst. Related questions about interpreting measured activity also arise in comparisons of enzymes for plastic breakdown.
Sources
- Lauko, A., Pellock, S. J., Sumida, K. H., et al. (2025). Computational design of serine hydrolases. Science, 388, eadu2454. DOI: 10.1126/science.adu2454.
- Lauko, A., et al. (2024). Computational design of serine hydrolases. bioRxiv, version 1, the preprint cited in the original report. DOI: 10.1101/2024.08.29.610411.
- Authors’ serine-hydrolase design workflow and win31 structural record.
Update note: This revision adds the 2025 peer-reviewed publication, identifies PLACER and the model substrate, and clarifies the difference between reactivity, turnover and application readiness. The original publication date and URL are retained.
