According to a study published in the journal Nature, researchers using artificial intelligence-based protein sequence redesign to stabilize enzyme starting points achieved superior outcomes in automated directed evolution, mitigating stability-activity trade-offs that traditionally constrain enzyme engineering. By combining the deep-learning model ProteinMPNN with the computational stabilization tool PROSS on botulinum neurotoxin (BoNT) protease models, scientists unlocked mutational spaces that wild-type enzymes could not access.
AI Sequence Redesign Overcomes Stability Trade-Offs in BoNT Proteases
Traditional evolutionary campaigns for novel enzymes often stall because mutations conferring new catalytic functions can destabilize the protein’s structural integrity. To counter this, investigators applied ProteinMPNN and PROSS to redesign the catalytic domains of BoNT/E, BoNT/F, and BoNT/X. According to the study findings, 78% of the initial 74 ProteinMPNN BoNT/E designs retained catalytic activity. The top redesigned variants, designated D1 through D3, displayed catalytic efficiencies 1.7 to 2.8 times higher than the wild-type enzyme. Specifically, variant D2 reached a catalytic efficiency of up to 310 mM−1s−1, compared to 110 mM−1s−1 for the wild-type BoNT/E.
Did you know? Thermal stability also improved significantly in the redesigned enzymes, with melting temperatures reaching up to 59.5°C, which helped drive a 24-fold increase in HEK293T cell expression for variants D2 and D3 when combined with prior mutations.
Evolutionary Campaigns Reveal Superior Outcomes Against Complex Substrates
To test evolutionary potential, researchers ran 44 parallel continuous evolution campaigns on an automated eVOLVER platform. Wild-type and redesigned BoNT/E proteases faced a panel of altered SNAP25 substrates of increasing difficulty, specifically labeled substrates 415, 413, and 412. When challenged against the most difficult substrate, 412, wild-type evolutions failed in 50% of lagoons, whereas all redesigned lagoons succeeded. Furthermore, the redesigned proteases successfully tolerated destabilizing mutations like K225E that drove high catalytic function but caused no detectable activity when placed in a wild-type background.
Reprogramming Protease Specificity Toward Human Ataxin-2
The research team applied the workflow to reprogram BoNT/E specificity toward human ataxin-2, a protein linked to neurodegeneration and amyotrophic lateral sclerosis (ALS). According to the study, an AI-redesigned protease variant achieved more than 79-fold greater specificity for the selected ataxin-2 substrate compared to the top wild-type-evolved enzyme. This variant showed a 16% sequence divergence from the natural protein framework. FRET assays confirmed that even the top D3-evolved protease at a concentration of 50 µM displayed no detectable cleavage of the native substrate SNAP25, demonstrating high target specificity.
Frequently Asked Questions
What AI tools were used in the study?
Researchers used ProteinMPNN, a deep-learning protein sequence-design model, alongside PROSS, a computational protein-stabilization method, to redesign the starting points of the enzymes.
How did redesigned enzymes perform against wild-type proteins?
According to the Nature study, redesigned starting points adapted faster, accessed functional mutational spaces unavailable to wild-type enzymes, and achieved significantly higher catalytic efficiencies and thermal stability.
What disease target was tested using the redesigned proteases?
The workflow successfully reprogrammed BoNT/E specificity to cleave human ataxin-2, a protein implicated in neurodegeneration and ALS, yielding markedly higher target specificity than wild-type-evolved enzymes.
Interested in more breakthroughs in biotechnology and computational biology? Subscribe to our newsletter for regular updates, or explore our latest research archives to read more in-depth analyses.
Worth a look