Conformation-Specific Design: Engineering Extracellular Signal-Regulated Kinase 2 Variants with Bias toward Active or...

Area of Science:

• Protein engineering
• Computational biology
• Biochemistry

Background:

• Controlling protein conformational states is crucial for enzyme function but remains challenging.
• Enzymes regulated by structural switches require precise conformational control for targeted design.
• Existing methods often rely on phosphomimetic mutations or extensive experimental screening.

Purpose of the Study:

• To present a machine learning framework, Conformation-Specific Design (CSDesign), for conformationally biased protein design.
• To generate protein sequences predicted to favor specific conformations while disfavoring others.
• To demonstrate ML-driven control over protein conformational ensembles.

Main Methods:

• Utilized high-resolution structural data and probabilistic sequence-structure models.
• Developed a machine learning framework (CSDesign) for conformationally biased protein design.
• Applied CSDesign to extracellular signal-regulated kinase 2 (ERK2) to generate active and inactive state variants.

Main Results:

• CSDesign successfully generated ERK2 variants biased towards active or inactive states.
• An active-biased variant (CSD104) showed robust kinase activity without phosphorylation.
• An inactive-biased variant (CSD101) remained inactivated, demonstrating conformational control.
• Engineered interactions were observed to stabilize active-like features.

Conclusions:

• Machine learning can effectively control protein conformational ensembles.
• CSDesign enables the design of conformationally regulated proteins, including enzymes.
• This approach offers an alternative to phosphomimetic mutations and reduces experimental screening needs.