Combinatorial mosaics dissect the residue level basis of ACE2 binding divergence between SARS-CoV-2 and RaTG13

The spike proteins of SARS-CoV-2 and the closely related bat coronavirus RaTG13 differ in their binding affinity of human ACE2. This is due to differences in the receptor binding domain (RBD), the region primarily responsible for ACE2 binding, which are only 89% identical on the amino acid level, compared to 96% for the remainder of their genomes. Here, we constructed a combinatorial library of RBD mosaics containing combinations of 18 amino acid differences between SARS-CoV-2 and RaTG13 and measured human ACE2 binding using yeast surface display and fluorescence-activated cell sorting. The mosaics revealed a broad range of binding phenotypes. The effects on ACE2 binding were concentrated at eight positions, including all six variable positions that directly contact ACE2. Although the SARS-CoV-2 spike binds human ACE2 stronger than RaTG13 spike, the SARS-CoV-2 residues Q493 and Q498 each reduced ACE2 binding relative to RaTG13’s tyrosines. By measuring the effects of these residues across many combinations of other sequence differences, we identified sequence context dependence for residues 493 and 498, finding that the deleterious effects of Q493 and Q498 were attenuated by the SARS-CoV-2 residues at positions 483, 484, and 494. Together, these results provide a residue-level account of the sequence differences underlying the difference in ACE2 binding by SARS-CoV-2 and RaTG13 RBDs, including which differences have the largest effect and which depend on sequence context. This work demonstrates that mosaic libraries provide a robust approach for resolving the contributions of sequence differences across diverse backgrounds.