Trimbody with rigid AI-designed scaffolds enables atomic-resolution cryo-EM structure determination of small proteins.

Abstract

Cryo-electron microscopy (cryo-EM) single-particle analysis faces significant challenges in resolving the structures of small proteins due to low signal-to-noise ratios and insufficient structural features. Here, we present Trimbody, a simple yet robust method that leverages rigid AI-designed scaffolds to overcome these limitations. Trimbody comprises two components: a trimeric scaffold (H3-PrAC-5350A) and a reformatted nanobody fusion (Nb-TAIL). The de novo 3-helix bundle (H3) and TAIL domains, engineered via RFdiffusion and ProteinMPNN, stabilize the interactions between Nbs and the scaffold and enhance the overall rigidity of the system, thereby enabling high-resolution cryo-EM analysis. Using Trimbody, we resolve the atomic structures of four sub-50 kDa test proteins: human Gal10 (2.62 Å), Aequorea coerulescens GFP (2.29 Å), the IgV domain of human Nectin4 (2.43 Å), and membrane protein Escherichia coli LacY (2.50 Å), demonstrating the method's versatility. Trimbody's simplicity, cost-effectiveness, and compatibility with standard cryo-EM workflows position it as a universal tool for structural studies of small proteins, further advancing nanobody-based drug development and life science research.