Viruses use surface proteins on host cells as entry receptors. These molecules are specifically recognized by proteins on the viral particles, and sequence variations among receptors from different species could have the potential to limit the virus's host range. Mutations in the viral proteins can overcome this barrier, and one recent example of this occurring is the SARS (severe acute respiratory syndrome) coronavirus, which seems to have adapted from an animal virus infecting palm civets to include humans as its host. The human strains caused the 2002–2003 epidemic that killed ∼10% of those that were infected.

Angiotensin convertin enzyme 2 (ACE2) is the receptor for SARS coronavirus and interacts with the viral spike protein. Previous studies have identified a ∼200-residue receptor-binding domain (RBD) in the spike protein. Furthermore, these studies have shown that the RBDs derived from the palm civet and the human epidemic strains are very similar, differing at only four positions; yet, the affinity of the respective spike protein for human ACE2 varies by more than 1,000-fold. The crystal structure of the spike protein RBD, derived from a human SARS coronavirus infection, in complex with human ACE2 (Li et al. Science 309, 1864–1868, 2005) now provides a structural explanation for this difference in affinity.

The spike protein RBD has two subdomains, but only one of these two (∼70 residues) contacts the ACE2 receptor. This subdomain forms a shallow cleft (red) that curves snugly around the N-terminal tip of ACE2 (yellow green). A total of 32 residues contribute to the interaction interface (14 residues from the RBD and 18 residues from the ACE2). Most of the interactions between the RBD and ACE2 are mediated by side chains and are hydrophilic in nature.

Of the four residues that are different in the RBDs from palm civet and human epidemic SARS coronaviruses, only two—positions 479 and 487—are at the interaction interface. A previous study showed that having a lysine at position 479 (found in many civet-derived viral sequences) reduces affinity for human ACE2, whereas having an asparagine at this position (found in human epidemic-derived viral sequences) does not seem to have a significant effect on binding to the civet receptor. The structure shows that the ACE2 residues surrounding position 479 are not completely conserved.

In contrast, a previous study showed that changing position 487 from threonine (human epidemic) to serine (civet) reduces affinity of RBD for human ACE2 by more than 20-fold. The structure shows that the ACE2 residues important for making contacts with position 487 of RBD are essentially identical between humans and civets. These residues support the formation of an inflexible hydrophobic pocket filled by the methyl group of Thr487 from the RBD of the human epidemic strains. A serine residue found in the civet-derived viral sequences cannot fill this pocket and may affect the affinity between the two proteins. Other factors, such as residues not at the interface or glycosylation, could also attenuate the affinity between RBD and ACE2. Nevertheless, together with additional sequence analysis of sporadic SARS cases, the structure suggests that the γ-methyl group on the side chain of position 487 may be a key determinant for the severity of the disease.