Broad Host Range of SARS-CoV-2 Predicted by Comparative and Structural Analysis of ACE2 in Vertebrates

Joana Damas; Graham M. Hughes; Kathleen C. Keough; Corrie A. Painter; Nicole S. Persky; Marco Corbo; Michael Hiller; Klaus-Peter Koepfli; Andreas R. Pfenning; Huabin Zhao; Diane P. Genereux; Ross Swofford; Katherine S. Pollard; Oliver A. Ryder; Martin T. Nweeia; Kerstin Lindblad-Toh; Emma C. Teeling; Elinor K. Karlsson; Harris A. Lewin

Angiotensin I converting enzyme 2 (ACE2) coding and protein sequences . All human ACE2 orthologs for vertebrate species, and their respective coding sequences, were retrieved from NCBI Protein (March 20, 2020) (70) . ACE2 coding DNA sequences were extracted from available or recently sequenced unpublished genome assemblies for 123 other mammalian species, with the help of genome alignments and the human or within-family ACE2 orthologs. The protein sequences were predicted using AUGUSTUS v3.3.2 (71) or CESAR v2.0 (72) and the translated protein sequences were checked against the human ACE2 orthologue. ACE2 gene predictions were inspected and manually curated if necessary. Identification of ACE2 residues involved in binding to SARS-CoV-2 S protein . We identified 22 ACE2 protein residues that were previously reported to be critical for the effective binding of ACE2 RBD and 15 SARS-CoV-2 S (13, 21) . These residues include S19, Q24 , T27, F28, D30, K31, H34, E35, E37, D38, Y41, Q42, L45, L79, M82, Y83, N330, K353, G354 , D355, R357, and R393. All these residues were identified from the co-crystallization and structural determination of SARS-CoV-2 S and ACE2 RBD (13, 21) . The known human ACE2 RBD glycosylation sites N53, N90 and N322 were also included in the analyzed residue set (11) . Protein structure analysis . We applied an orthogonal approach to assess the likelihood of binding of a sampling of species that were predicted to bind SARS-CoV-2 across the categories of high , medium , low or very low likelihood of binding. ACE2 amino acid sequences from 28 species were extracted from the multiway alignment and loaded into SWISS-MODEL (25) in order to generate homology derived models.