Investigation of the genetic variation in ACE2 on the structural recognition by the novel coronavirus (SARS-CoV-2)

The outbreak of the coronavirus disease 2019 (COVID-19), caused by a novel (new) coronavirus (SARS-CoV-2), has been characterized as a global pandemic [1‐5]. COVID-19 is rapidly spreading across the world and affecting all populations. It has been documented that the S-protein of SARS-CoV-2 plays a key role in the recognition to the peptidase domain (PD) of the Angiotensin converting enzyme (ACE2) in humans [6, 7]. The three-dimensional protein structures of SARS-CoV-2 have recently been determined, which provide important insight into the treatment of the disease, such as vaccine development, antibody design and drug discovery [7]. The first X-ray crystallization structure of 3CLpro was resolved by Liu et al. at 2.16 Å resolution (Protein data bank (PDB) id 6lu7). The virus S-protein structure was first observed by Wrapp et al. at 3.46 Å resolution by electron microscopy (PDB id 6vsb) [8]. The first full-length S-protein in complex with human ACE2 Cryo-EM structure was observed by the institute of Xihu University [7], and at almost the same time, the X-ray structure of the S-protein RBD domain in complex with ACE2 was solved by Tsinghua University at 2.45 Å re solution [9]. Two X-ray structures of S2 subunit have been determined by Zhu et al. (PDB id 6lxt and 6lvn). In addition to SARS-CoV-2, several three-dimensional structures of ACE2, especially in complex with SARS S-protein, have been solved. It shows that ACE2 structure is flexible to toggle between open and close states when it binds an inhibitor or virus S-protein. Genetic variation, especially deleterious missense variants in these flexible regions, may affect its function and structure, and consequently alter the recognition by SARS-CoV-2. Thus, it’s important to systematically characterize and evaluate potentially deleterious variants in ACE2, which may affect SARS-CoV-2 recognition and infection, and COVID-19 susceptibility and treatment.

To illustrate how genetic variation may affect the structure, we analyzed the structural interactions between ACE2 and SARS-Cov-2. As displayed in Fig. 1a, b and Additional file 2: Figure S1, we demonstrated that ACE2 has two states, i.e. open and closed, for its native and ligand-binding states through a large hinge-bending motion [13]. In open state, ACE2 opens wide from its active site to wait for a ligand to enter. When the ligand enters ACE2 active site, it triggers ACE2 to close the active slot. Most SARS binding structures (e.g. PDB ids 2ajf, 3d0g, 3kbh, 3scl) show that S-protein binds the open/native state of ACE2. However, as depicted in Fig. 1c, d and Additional file 2: Figure S2, the two monomers in PDB 3scl display that SARS spike proteins can bind either in an open state or in a closed state of ACE2, which implies that the conformational change of ACE2 can be triggered either by an inhibitor from an inner active site or S-protein from an outer PPI (protein–protein interface) site. The huge conformational change of the two states can be up to 14 Å distance shift between Lys341 and Thr129. The two N-terminal helices (Ser19–Asn53, Ile54–Met82) that contact SARS-Cov-2 S-proteins are among the most flexible regions. The hinge movement of the helices pivots on the loop region of Trp83–Asn90. We also observed that the protein–protein interface of the SARS-CoV-2 spike glycoprotein to ACE2 has more hydrophilic residues than hydrophobic ones. The residues in ACE2 within distance of 3 Å to S-protein are Gln24, His34, Asp38, Tyr41, Gln42, Tyr83 and Lys353 to Gly446, Tyr449, Tyr453, Asn487, Thr500 and Gly502. The PPI interface binds with six hydrogen bonds (Gln24–Asn487, Gln42–Gly446, Gln42–Gln498, Lys353–Gly502), a network of π–cation interactions (Tyr41–Gln498, Tyr41–Asn501, Gln42–Tyr449, Tyr83–Asn487, Gln493–His34), one π-stacking interaction (Tyr83–Phe486), and only one hydrophobic interaction pair (Met82–Phe486). In summary, the structural flexibility of ACE2 implies that its structure could be distorted by potentially deleterious missense variants with the altered amino acids in ACE2, which may consequently affect its binding efficiency to the S-protein in the virus.

We next analyzed germline coding variants in ACE2 from the gnomAD and performed functional predications using six existing bioinformatics tools (see “Methods”). We characterized a total of 12 ACE2 putative deleterious missense variants, whereas the top variants with functional disruptions predicted by all tools included p.Leu731Phe (rs147311723, AF = 0.01 in African), p.Arg219Cys (rs372272603, AF = 7 × 10⁻⁴ in European), p.Ser547Cys (rs373025684, AF = 4 × 10⁻⁴ in European) and p.His378Arg (rs142984500, AF = 2 × 10⁻⁴ in European) (Table 1). Of those, we observed that these variants showed low frequency or rare in all populations (Fig. 2a; Table 1). AF of the characterized variants varied in populations, whereas a majority of them showed population specificities (Fig. 2a). In particular, we observed that two variants with low frequency were present in African (rs147311723) and in East Asian (s191860450), respectively (Fig. 2b). The top AF of the other missense variants were present including, in African (rs73635825, rs138390800, and rs149039346), East Asian (rs191860450), South Asian (rs148771870 and rs751603885) and Europeans (rs148771870) (Fig. 2b).

We further analyzed the structural flexibility of ACE2 and its interaction with the RBD of S-protein of SARS-CoV-2 using the two or three-dimensional interaction diagrams (see “Methods”) for nine missense variants on eight residues, as displayed in Fig. 3a. We showed that p.His378Arg could directly weaken the binding of catalytic metal atoms to decrease ACE2 catalytic activity, and p.Ser19Pro (rs73635825, AF = 3 × 10⁻³ in African) could distort the most important helix to interact with the S-protein.

As shown in Fig. 3, ACE2 has flexibility in its structure when it binds an inhibitor or virus S-protein. Therefore, the conformational change could be triggered by altered amino acids as well. Although some of missense variants we analyzed are not directly located on the PPI surface, the altered amino acids could affect the binding of virus S-protein allosterically. Since the binding of the inhibitor inside the active site triggers ACE2 to enter a closed state from an open state through a huge conformational change, the altered amino acids of active site residue could cause a structural change of ACE2 more easily. Thus, the His378Arg amino acid change may not only reduce ACE2 peptidase activity, but also change the structure of a PPI area to affect S-protein binding. When S19 mutates to the helix “killer” proline, it may destabilize the most important helix to contact with S-protein. For SARS, the 24QAK to 24KAE mutant of ACE2 slightly inhibits interaction with spike glycoprotein [15]. Gly211Arg, Asp206Gly, Arg219Cys/His, Lys341Arg, and Ile468Val may affect the interactions across secondary structures. Therefore, their mutations may destabilize the local structure significantly. Ser547Cys may only affect the stability of one secondary structure, which may have a minor effect on the S-protein binding. As listed in Table 1, Procko [16] studied the virus binding abilities of 2340 human ACE2 mutants by using deep mutagenesis experiments. It shows that the Ser19Pro mutant is a strong booster for viral binding, and His378Arg is a weak booster. Thus, the local structural change from the Ser19Pro or His378Arg mutation may enhance the S-protein interaction allosterically based on his experimental results. It should be noted that our predicted deleterious variants in protein structure lack of experimental validation. Further exploration would be required to further confirm their potential effects on ACE2 function. It should also be addressed that, for theoretical predictions, different researchers provided different conclusions on the mutation effects based on different criteria and methods. For example, for the Ser19Pro mutation, it is predicted to be an interaction-booster by some groups [17‐19], and an interaction-inhibitor by other groups [20, 21].

A recent genome-wide association study (GWAS) including 835 patients with COVID-19 and severe disease (defined as respiratory failure) and 1255 control participants from Italy, plus 775 patients and 950 control participants from Spain was conducted by the Severe COVID-19 GWAS Group [22]. They identified multiple genetic variants and genes associated with COVID-19 with respiratory failure. Although our study has characterized the putative functional and structurally related variants in ACE2 with top allele frequencies in various populations, the lack of phenotypes of COVID-19 prevents us from identifying susceptibility variants associated with a phenotype of COVID-19. However, our findings, together with other ACE2 genetic studies [17, 19], can prioritize the promising variants in ACE2 for further fast-track genotyping in blood samples from COVID-19 patients, which could provide a great opportunity to identify susceptibility variants in ACE2 related to symptoms of COVID-19 patients. On the other hand, our findings may also provide possible consideration of individuals carrying the identified variants in ACE2 for current vaccine development, especially those involved in ACE2 interaction with the S-protein of SARS-CoV-2.

In this study, we characterized a total of 12 putative deleterious missense variants in the gene ACE2. Of those, we further provided strong evidence of nine missense variants that may disrupt the flexible regions of ACE2 protein structure or its protein–protein interaction with the RBD of S-protein of SARS-CoV-2. Results from this study highlight an important role of deleterious missense variants in the gene ACE2 that are present in the specific populations, which may affect SARS-CoV-2 recognition and infection. These variants could be important for the development of appropriate strategies of COVID-19 prevention, control and treatment to distinguish individuals between carrying and non-carrying those deleterious variants. Our findings may also provide a clue to partially explain why there were substantial discrepancies about the morbidity and mortality in regional disparity and distinct populations.