# AI-guided prefusion stabilization of the human coronavirus OC43 spike protein enables universal embecovirus antigen design

## TL;DR

Researchers used AI to stabilize the spike protein of a human coronavirus, enabling better vaccine design for a range of coronaviruses.

## Contribution

A novel AI-guided method for stabilizing embecovirus spike proteins in their prefusion state, applicable across species.

## Key findings

- Stabilized OC43 spike proteins showed enhanced expression and thermal stability.
- The stabilization strategy was successfully transferred to equine and HKU1 coronaviruses.
- Cryo-EM structures confirmed the prefusion conformation and revealed the stabilization mechanism.

## Abstract

The continued threat of zoonotic coronavirus spillovers underscores the need for cross-species applicable vaccine design strategies. The genus Embecovirus includes human coronaviruses OC43 and HKU1 as well as relevant veterinary pathogens. The coronavirus spike (S) fusion glycoprotein, key to viral entry and protective immunity, is inherently metastable, complicating vaccine development. Using the ReCaP AI tool, we stabilized the prefusion conformation of OC43 S through rationally combined amino acid substitutions, resulting in markedly enhanced expression and thermal stability. The substitutions were transferable to equine coronavirus (ECoV) S and HKU1. Cryo-EM structures of stabilized OC43 and ECoV S revealed that stabilization was achieved by arresting the release of the fusion peptide and keeping the S1B receptor binding domain in the ‘down’ state by improving the complex polar interactions of neighboring S1B domains and the bound free fatty acid at the interprotomer S1B interface. This work provides the first ECoV S structure and a broadly applicable framework for engineering stabilized Embecovirus S antigens.

Coronaviruses pose a constant threat to global health due to the potential transmission from animal to human hosts. Vaccine development has been limited by the instability of the coronavirus spike protein in the prefusion state and since future spillover events may involve previously unidentified coronaviruses there is a need for broadly applicable stabilization strategies. In this study, we describe an approach to stabilize embecovirus prefusion spike proteins using the human coronavirus OC43 spike as a prototype. By combining phylogenetic analysis of OC43 laboratory strains, structure-guided predictions using the machine-learning algorithm ReCaP, and rational structure-based design, we engineered spike proteins with improved prefusion stability. Importantly, these stabilizing features were transferable to other embecoviruses, including equine coronavirus (ECoV) and human coronavirus HKU1. We further determined high-resolution cryo–electron microscopy structures of the stabilized OC43 and ECoV spikes, confirming preservation of the prefusion conformation and providing insights into the mechanism by which the substitutions improve stability. Together, our results establish a generalizable framework for stabilizing embecovirus spike proteins and support pandemic preparedness efforts to mitigate future zoonotic spillover events.