Deep learning–enabled scaffolding of spatial arrays of PfCSP epitopes

Malaria is a leading cause of disease in developing countries. The licensed malaria vac-cine RTS,S/AS01 confers partial protection in part due to the elicitation of circum-sporozoite protein (CSP) antibodies, of which those to the CSP repeat and junctionalregions offer the most potent protection. Anti-repeat region antibodies, including theprotective antibody L9, frequently develop mutations that promote inter-Fab contactswhen bound to CSP in “spiral” quaternary structures. As a first step toward the designof immunogens that elicit L9-like antibodies, we utilized generative deep learning mod-els to design epitope scaffolds that incorporated up to three junctional repeat epitopeswith structural conformations and relative spatial orientations matching those of themultivalent complex of CSP bound to three copies of L9. Affinity and structural studiesdemonstrated accurate scaffolding of two epitopes with the intended relative orientation,and displacement of the third epitope, while maintaining inter- Fab contacts betweenL9 antibodies. In a mouse model of malaria liver invasion, immunization with nano-particles displaying these scaffold immunogens inhibited liver invasion as potently asmatched nanoparticles displaying a short junctional peptide but less potently than thesame nanoparticles displaying longer junctional peptides. This study demonstrates asubstantial advance for design of multiepitope scaffolds with predetermined relativeepitope spatial positioning. The study also represents an initial step toward developmentof multiepitope immunogens to elicit antibodies that utilize homotypic interactions tobind pathogens in multivalent clusters