Optimization of a sarbecovirus llama nanobody–antigen binding interface via a combined computational and phage display protein engineering approach

Single-domain antibodies, such as variable heavy domain of heavy chain (VHH) domains from camelids, are an attractive platform for therapeutic purposes. VHHs have a smaller size than traditional antibodies, which harbors certain advantages such as tissue penetration or accessing epitopes that may be shielded by protein domains. However, the smaller size of VHHs typically involves a smaller binding site, which can limit the scope of related antigens they are able to bind and, in the specific case of antiviral VHHs, render them susceptible to escape by a single point mutation. Here, we present a combined computational and experimental protein engineering approach to broaden the reactivity of SARS-CoV-1 receptorbinding domain (RBD)-specific VHH-72 for SARS-CoV-2 and Delta. Our strategy focuses on increasing the size of the binding site by imparting "second site" interactions toward heterologous antigens. We utilized the residue-based pharmacophore modeling approach, Protein-ligandinterface design (ProtLID), to identify potential productive side chain interactions in this second site and then encoded the ProtLID-predicted restricted diversity into a VHH-72-based phage library, which we then screened for cross-reactive SARS-CoV-1/2 RBD clones. Based on sequence analysis from the resulting population of functional VHH clones, we identified a VHH-72 triple mutant (T60P.D61P.D100eY), which maintained high affinity for SARS-CoV-1 RBD but was enhanced by 18- and 20-fold for RBDs from SARS-CoV-2 and Delta relative to WT VHH-72. These results highlight the potential of computationally customizing phage display diversity for single-domain binding proteins and provide a strategy for designing extension of protein binding interfaces.