Featuring ACE2 binding SARS-CoV and SARS-CoV-2 through a conserved evolutionary pattern of amino acid residues

Abstract

Spike (S) glycoproteins mediate the coronavirus entry into the host cell. The S1 subunit of S-proteins contains the receptor-binding domain (RBD) that is able to recognize different host receptors, highlighting its remarkable capacity to adapt to their hosts along the viral evolution. While RBD in spike proteins is determinant for the virus–receptor interaction, the active residues lie at the receptor-binding motif (RBM), a region located in RBD that plays a fundamental role binding the outer surface of their receptors. Here, we address the hypothesis that SARS-CoV and SARS-CoV-2 strains able to use angiotensin-converting enzyme 2 (ACE2) proteins have adapted their RBM along the viral evolution to explore specific conformational topology driven by the residues YGF to infect host cells. We also speculate that this YGF-based mechanism can act as a protein signature located at the RBM to distinguish coronaviruses able to use ACE2 as a cell entry receptor.

Communicated by Ramaswamy H. Sarma

1. Introduction

Viruses are the most numerous type of biological entity on Earth and the identification of novel viruses continues to enlarge the known viral biosphere (Pan et al., 2017; Shi et al., 2018). This collection of all viruses presents enormous morphological and genomic diversity as a result of continuous exchange of genetic material with the host cells (Koonin et al., 2020; Nasir et al., 2014). Moreover, this well succeeded long-term virus–host interaction indicates that viruses are more than simple genomic parasites in all cellular life forms (Claverie, 2006). A number of evidences has led to the proposal that viruses play an astonishing role as agents of evolution because of their capacity in propagating between biomes (Sano et al., 2004) and in gene transfer between species (Enard et al., 2016; Filée et al., 2003; Koonin & Dolja, 2013; Van Blerkom, 2003). For this purpose, viruses have developed large number of genome replication and protein expression strategies to benefit from the host translational machinery over time (Baranowski et al., 2003).

Despite all of such enormous diversity in gene sequence, it is not possible to achieve huge number of highly distinct protein structures mainly because of stereochemical constraints on the possible protein folds (Ng et al., 2020). In fact, it has been observed common secondary structures throughout different virus families while the sequences are not fully conserved (Ahola & Karlin, 2015; Ng et al., 2020). This may result in evolutionary efficiency once viruses can exploit already well-designed motifs from similar cellular functions (Baranowski et al., 2003).

Currently, the world population is confronting a new coronavirus disease (COVID-19), a highly infectious disease to humans. This disease is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and is affecting human health worldwide. Coronaviruses (CoVs) belong to the large and diverse family Coronaviridae, within the order Nidovirales and suborder Cornidovirineae (Gorbalenya et al., 2020). Their subfamily Orthocoronavirinae contains four genera based on phylogeny and termed as α, β, γ, and δ-coronavirus.

SARS-CoV-2 belongs to the β-coronavirus genus as well as SARS-CoV, middle east respiratory syndrome coronavirus (MERS-CoV), and hCoV-HKU1, to cite a few (Jaimes et al., 2020). Other important representative human viruses as hCoV-NL63 and hCoV-229E belong to α-coronavirus. Phylogenetic relationships among the known members of this subfamily indicate that α and β-coronavirus infect mammals, while γ and δ-coronavirus infect both mammalians and avians.

Members of Coronaviridae family are enveloped, positive single-stranded RNA (+ssRNA) viruses and render the largest genomes among all known RNA viruses (Cheng et al., 2007; Li, 2016; Masters, 2006; Su et al., 2016). The + ssRNA genomes undergo rapid mutational changes (Sanjuán & Domingo-Calap, 2016), leading to faster adaptation to new hosts, though they also contain conserved sequence motifs as observed, for example, in multiple alignments do CoV strains (Ahola & Karlin, 2015; Lau et al., 2007; Woo et al., 2012).

Coronaviruses attach to host cell surface receptors via their spike (S) glycoproteins, located on the viral envelope, to mediate the entry into the host cell. Each monomer of trimeric S-protein comprises two subunits S1 and S2, responsible for the viral attachment and for the membrane fusion, respectively (Beniac et al., 2006; Li, 2012; Song et al., 2018). The S1 coronavirus subunit contains the receptor-binding domain (RBD) that is able to recognize different host receptors, highlighting its remarkable capacity to adapt to their hosts along the viral evolution. Thus, it is not unexpected to observe in this domain high sequence divergence even for the same coronavirus identified in different host species. In contrast, the S2 subunit presents the most conserved region in the S-protein.

The binding of RBD spike proteins to the receptor on the host cell is the first step in virus infection. This initial step is followed by an entry mechanism of enveloped viruses into target cells. Usually, most viruses enter cells through endocytotic pathways with the fusion occurring in the endosomes, although a direct entry into cells can occur by fusion of their envelopes with the cell membrane (Belouzard et al., 2012).

A number of CoVs utilizes angiotensin-converting enzyme 2 (ACE2) as the entry receptor into cells, exemplified by β-genus human respiratory SARS-CoV, SARS-CoV-2, and α-genus hCoV-NL63 (Hoffmann et al., 2020; Jaimes et al., 2020; Letko et al., 2020; Walls et al., 2020). In particular, SARS-CoV, as well as SARS-CoV-2, enter the cell via endocytosis induced by RBD complexed with human ACE2 (hACE2) receptor (Milewska et al., 2018; Ou et al., 2020; Wang et al., 2008, 2008; Yuan et al., 2017). In contrast, the β-genus MERS-CoV and its genetically related bat CoV-HKU4 utilize dipeptidyl peptidase 4 (DPP4) as the viral receptor (Yang et al., 2014). Other viral receptor is aminopeptidase N (APN), recognized for example by the α-genus hCoV-229E (Li, 2015).

The human coronaviruses hCoV-HKU1, hCoV-229E, hCoV-NL63, and hCoV-OC43, cause mild to moderate upper respiratory tract infections (Weiss & Navas-Martin, 2005), while SARS-CoV and SARS-CoV-2 cause severe respiratory diseases, with SARS-CoV-2 being far more lethal than SARS-CoV. SARS-CoV strains vary enormously in infectivity, which can be connected to their binding affinities to hACE2 (Cui et al., 2019). This binding affinity, in turn, can be correlated with disease severity in humans (Li, Zhang, et al., 2005).

While RBD in spike proteins is determinant for the virus–receptor interaction, the active residues lie at the receptor-binding motif (RBM), which is part of RBD and plays a fundamental role binding the outer surface of their receptors (Cui et al., 2019; Hoffmann et al., 2020; Letko et al., 2020; Li, Farzan, et al., 2005; Wan et al., 2020). The importance of the RBM is further explored here in relation to its structural topology. Thus, instead of only analyzing specific residues that make contacts with ACE2 after binding, we go a step further and track the molecular origin that drives the viral attachment to this cell receptor. This investigation has revealed a highly conserved amino acid residue sequence Tyr-Gly-Phe (YGF) in coronavirus variants that employ this receptor. Consequently, we hypothesize that the short sequence YGF is vital for RBD–ACE2 interaction because of the formation of a hydrophobic pocket proper to the receptor specificity (He et al., 2020; Lan et al., 2020; Wan et al., 2020; Wu et al., 2012). Moreover, we recognize that a similar binding mechanism is characteristic of the interaction between ubiquitin-associated (UBA) domain proteins and ubiquitin. In this vein, we conclude that is plausible that SARS-CoV and SARS-CoV-2 strains able to use ACE2 proteins have adapted their RBM along the viral evolution to explore such a mechanism to infect host cells.

1.1. The conserved XGF loop in UBA–ubiquitin interaction

Amino acid sequences of type XGF, where the residue X is frequently the residue Met, form a highly conserved loop characteristic of ubiquitin-associated (UBA) domain that occurs in a variety of proteins. The UBA domain is a conserved motif through eukaryotic evolution and is found in many proteins related to the ubiquitin metabolism and in particular, associated with ubiquitin-mediated proteolysis (Finley, 2009; Pickart & Cohen, 2004). The MGF loop in the UBA domain is typical of a hydrophobic pocket that is critical for recognition and binding affinity to ubiquitin through a hydrophobic surface patch located in the vicinity of this loop (Bertolaet et al., 2001; Cabe et al., 2018; Dieckmann et al., 1998; Long et al., 2008; Madura, 2002; Raasi et al., 2005; Tse et al., 2011; Wilkinson et al., 2001). UBA domains are ubiquitin receptors whose binding is a fundamental step for diverse regulatory functions.

NMR analysis of UBA–ubiquitin interactions identify hydrophobic surface patches formed by the conserved MGF sequence as the main determinants for the protein–protein interaction. A number of alignments of UBA domains has revealed mutations in the MGF sequence, mainly M → L, M → Q, and F → Y, but these mutations still maintain the overall hydrophobic characteristic for the main set of residues located at the protein-interacting surface in these UBA domains (Geetha & Wooten, 2002; Mueller & Feigon, 2002).

An illustrative example of this protein–protein interaction mechanism is presented in Figure S1, supplemental material. This figure displays the UBA domain of Dsk2p protein complexed with ubiquitin (PDB ID: 1WR1) where the residues M342, G343, and F344 form the core of the hydrophobic surface patch for interacting with ubiquitin. The importance of this hydrophobic surface patch has been strengthened by mutagenesis experiments (Ohno et al., 2005) of single substitutions G343A, F344A, and L369A in Dsk2p–UBA domain, producing a large decrease of its binding affinity for ubiquitin.

Figure 1.

Detailed surface view of SARS-CoV and SARS-CoV-2 RBD. (a) Residues YGFY at the interface of SARS-CoV complexed with hACE2 (PDB ID: 2AJF). These residues are in yellow color and form a hydrophobic pocket located in the RBM (magenta color). (b) Residues YGF and EGF (yellow color) at the interface of SARS-CoV-2 complexed with hACE2 (PDB ID: 6LZG). The first sequence is located in a hydrophobic pocket, while the second sequence EGF is on the RBM surface (magenta color). Ribbon representation of ACE2 is in blue color.

2. Results and discussion

2.1. Spike receptor-binding motifs in human CoVs

Here, we investigate the occurrence and importance of the specific amino acid residue sequence YGF for SARS-CoV and SARS-CoV-2 strains able to use ACE2 proteins as receptors. It is displayed in Figure 1a the interface of SARS-CoV RBD spike-protein (magenta and green color) complexed with hACE2 (blue color) to gain insight about the importance of this type of conformational mechanism in creating a shape complementarity between receptor and ligand. The RBM is in magenta color, with the yellow color displaying the YGFY sequence in that pocket, which establishes the proper relative position for favorable binding to surface-exposed hACE2 residues. The YGFY sequence seems strongly conserved in SARS-CoV spike RBD, more precisely located at residues 481–484 in the receptor binding motif. Noteworthy, this sequence seems to be unique because even the shorter YGF sequence does not occur in this region, neither in the RBD. As a consequence of this hydrophobic pocket, amino acid residues responsible for binding interaction are located close to this conformational structure as, for example, the residues N479 and T487 (Figure 2a). These residues have been identified to be essential for SARS-CoV spike RBD/ACE2 binding (Li et al., 2005, 2005; Qu et al., 2005). The residue N479 in SARS-CoV is located near K31 of hACE2 which in turn makes a salt bridge with E35, a residue buried in that hydrophobic environment. The residue T487 is located close to K353 on hACE2, and in turn makes a salt bridge with D38, also buried in that pocket. Other important residues for this attachment are Y442, L472, and D480 (Wan et al., 2020).

Figure 2.

SARS-CoV and SARS-CoV-2/hACE2 RBD interfaces. Ribbon diagrams of SARS-CoV RBD (a) and SARS-CoV-2 RBD (b) complexed with hACE2 (blue color), where the RBM is highlighted in magenta color. The main residues responsible for the structural binding are displayed in the stick representation.

Figure 3a displays the residues of SARS-CoV RBM in direct contact with hACE2 as determined by the hydrophobic–hydrophilic properties of the interface residues as predicted by the CSU program (Sobolev et al., 1999). This bipartite network of contacts highlights the importance of residues that are located near the YGF sequence and contributes to the stabilization of SARS-CoV complexed with hACE2. For example, Y475 makes hydrogen bond (H–B) contacts with Q24, F28, and Y83; N479 with K31, and H34; Y486 with Y41, N330, and R357; and T487 with Y41.

Figure 3.

Contact networks between (a) SARS-CoV residues, and (b) SARS-CoV-2 residues located in the RBM regions with ACE2. SARS-CoV and SARS-CoV-2 residues are in magenta color while human ACE2 residues are in blue color.

Figure 1b displays the interface of SARS-CoV-2 RBD spike protein complexed with hACE2 (blue color). Now, the sequence YGFY observed in SARS-CoV is replaced by YGFQ as a result of sequence alignments shown in Figure 4. The single-point mutation Y484 → Q498 replaces a hydrophobic residue in SARS-CoV by a hydrophilic one in SARS-CoV-2.

Figure 4.

Sequence alignments of human CoVs restricted to RBM residues. The medium purple color highlights the YGFY pattern followed by the mutation Y498Q in the RBM of SARS-CoV-2 strains.

Figure 4 compares residue sequences of human SARS-CoV and SARS-CoV-2 strains aligned with RBM of SARS-CoV Tor2, an epidemic strain isolated from humans during the SARS epidemic in 2002–2003. The human Tor2 strain has high affinity for hACE2 (Cui et al., 2019). We highlight in this figure in medium purple color the hydrophobic sequence YGFY typical of SARS-CoV, occurring at positions 481–484 in the spike protein. The corresponding mutated sequence occurs now at positions 495–498 in SARS-CoV-2 spike protein.

The important residues for the interface interaction found in SARS-CoV are mutated in SARS-CoV-2. The sequence alignments show the mapping, Y442 → L455, L472 → F486, N479 → Q493, D480 → S494, and T487 → N501. These mutations do not present a drastic change in their hydrophobic character (Wimley & White, 1996), thus preserving the overall receptor-binding topological structure for these viruses. In particular, residues L455 and Q493 in SARS-CoV-2 preserve the noted favorable interactions with the residues E35 and K31 in hACE2 (Yi et al., 2020) (Figure 2b). Interestingly, a new GF sequence appears in the RBM of SARS-CoV-2 strains as a consequence of the mutation L472 → F486, producing a small hydrophobic surface, but does not seem to disrupt the proposed topological formation mechanism for ACE2 binding. No other GF sequence appears in their RBD.

Figure 3b displays the SARS-CoV-2 RBM residues in direct contact with hACE2 as predicted by the CSU program, showing again the importance of residues close to the hydrophobic pocket. Details of protein–protein binding interfaces can be quite different among strains, likely related to their infectivity degree. It has been noted that mutations in RBM residue T487 in SARS-CoV have an important role in the human-to-human and animal-to-human transmission of SARS-CoV (Cui et al., 2019; Li, 2015; Qu et al., 2005).

Now, we investigate the relevance of the hydrophobic pocket driven by the YGFY sequence in promoting the stability of SARS-CoV spike receptor binding domain complexed with hACE2. To this end, we conducted a series of mutations to estimate the change in binding affinity $\Delta \Delta G$ using the MutaBind2 method (Zhang et al., 2020).

Initially, we investigate the influence of N479 mutation by the residues E, K, Q, R, and S on the complexation. The mutations N479E, N479K, N479Q, N479R, and N479S have been observed, respectively in pangolin strains (see Figure 5), bat and palm civet strains (see Figures 5 and 6), human SARS-CoV-2 (see Figure 4), bat and palm civet strains (see Figures 5 and 6), and bat strains (see Figure 6). The calculation of $\Delta \Delta G$ for these mutations does not indicate any appreciable effect on SARS-CoV spike RBD/hACE2 binding affinity due to its small variation, as displayed in Table 1. Therefore, we may conclude that N479 does not enhance the binding affinity of spike RBD to hACE2, and could as well be replaced by any of the above residues, preserving the hydrophilic character (Wimley & White, 1996) and conformational stability.

Figure 5.

Sequence alignment of bat CoVs restricted to RBM residues of SARS-CoV Tor2. The residues of YGFY pattern are in medium purple color. Last three alignments are placed together for direct amino acid sequence comparison.

Figure 6.

Sequence alignment of palm civet CoVs and pangolin PCoVs restricted to RBM residues of SARS-CoV Tor2. The residues of YGFY pattern are in medium purple color. Last line includes the SARS-CoV-2 sequence for comparison.

Table 1.

Changes in binding affinity of human SARS-CoV and human SARS-CoV-2 spike RBD complexed with hACE2 upon mutation as predicted by MutaBind2 method. Prediction effects are classified as low-confidence prediction: (1), or high-confidence prediction: (2). Here we adopt the classification presented in MutaBind method (Minghui et al., 2016) because of the high similarity between the ROC (receiver operating characteristic) curves in both methods.

	$\Delta \Delta G$	Effect on the
SARS-CoV	(kcal/mol)	complexation
N479E	0.76	neutral ${}^{(2)}$
N479K	0.66	neutral ${}^{(2)}$
N479Q	0.40	neutral ${}^{(2)}$
N479R	0.08	neutral ${}^{(2)}$
N479S	0.63	neutral ${}^{(2)}$
G482A	1.07	neutral ${}^{(2)}$
G482D	1.85	highly destabilizing ${}^{(1)}$
G482V	1.78	highly destabilizing ${}^{(1)}$
G482D, Y484F	2.40	highly destabilizing ${}^{(2)}$
G482D, Y484H	2.68	highly destabilizing ${}^{(2)}$
G482D, Y484N	3.24	highly destabilizing ${}^{(2)}$
G482D, Y484Q	2.47	highly destabilizing ${}^{(2)}$
G482D, Y484T	3.26	highly destabilizing ${}^{(2)}$
Y484F	0.64	neutral ${}^{(**)}$
Y484H	1.57	highly destabilizing ${}^{(1)}$
Y484N	2.19	highly destabilizing ${}^{(1)}$
Y484Q	1.44	neutral ${}^{(*)}$
Y484T	1.92	highly destabilizing ${}^{(1)}$
N479E, G482D, Y484F	4.58	highly destabilizing ${}^{(2)}$
N479E, G482D, Y484H	4.90	highly destabilizing ${}^{(2)}$
N479E, G482D, Y484N	4.69	highly destabilizing ${}^{(2)}$
N479E, G482D, Y484Q	4.58	highly destabilizing ${}^{(2)}$
N479E, G482D, Y484T	4.67	highly destabilizing ${}^{(2)}$
SARS-CoV-2
G496D	0.96	neutral ${}^{(2)}$
G496A	1.10	neutral ${}^{(2)}$
G496V	1.69	highly destabilizing ${}^{(1)}$
Q498Y	0.16	neutral ${}^{(2)}$
G496D, Q498F	2.75	highly destabilizing ${}^{(2)}$
G496D, Q498H	2.52	highly destabilizing ${}^{(2)}$
G496D, Q498N	1.72	highly destabilizing ${}^{(1)}$
G496D, Q498T	1.68	highly destabilizing ${}^{(1)}$
Q493S	0.89	neutral ${}^{(*)}$
Q493S, G496D, Q498F	3.63	highly destabilizing ${}^{(2)}$
Q493S, G496D, Q498H	4.33	highly destabilizing ${}^{(2)}$
Q493S, G496D, Q498N	2.96	highly destabilizing ${}^{(2)}$
Q493S, G496D, Q498T	2.85	highly destabilizing ${}^{(2)}$

Specific mutations were also performed to investigate the importance of the residues forming the hydrophobic surface patch for the complexation stability. Residues Y481 and F483 are conserved through all SARS-CoV strains we have analyzed in this work, while G482 and Y484 are mutated in some strains (bats and pangolins). Therefore, we mutated G482 and Y484 by plausible residues, i.e. the ones that occur in other strains to evaluated the changes in the binding affinity. To perform this double mutation, we fixed G482D, a mutation observed in bat strains, followed by Y484F, Y484N, and Y484T, observed in bats; Y484H, observed in pangolins; and Y484Q, observed in human SARS-CoV-2. The free-energy changes for these double mutations strongly indicate the desestabilization of human SARS-CoV complexed with hACE2 (Table 1). It is interesting to note that G482D together with Y484Q decreases the binding affinity in human SARS-CoV because they decrease the hydrophobicity of the initial YGFY pocket. This conclusion is supported by the mutant G482D-Y484Q structure obtained by the bioinformatic tool Mutabind2, after a 100 step energy minimization, which does not display steric hindrances between these residues and the ones at the interface with ACE2. Table 1 also displays the changes in binding affinity for single mutations of Y484, a highly connected residue with hACE2 (see Figure 3).

We have also applied the predicting tool Mutabind2 to the single mutations G482A, and G482V to evaluate under which conditions we may obtain disruptive S1–protein–virion interaction. This tool yielded 1.07, and 1.78 kcal/mol, respectively for G482A, and G482V. V842 produces a single steric hindrance at Y436:CZ from spike protein, highlighting the importance of hydrophobicity against steric effects.

Next, we mutated N479 followed by mutations at sites 482 and 484 to analyze the consequences on the binding affinity of this triple mutation by disrupting the hydrophobic surface patch. Again, we fixed, for example, the mutations N479E and G482D. The impacts of this set of mutations can be seen in Table 1. What was considered to be a neutral mutation, N479E shows high destabilizing effect in the new conformational environment. Similar destabilizing effects on the human SARS-CoV spike RBD complexed with hACE2 are obtained for the important residue T487 when one mutates the residues forming the hydrophobic pocket (data not shown).

Now, we repeat the above procedure to investigate the role of residues YGFQ for human SARS-CoV-2 RBD/hACE2 binding affinity. The mutation Q498Y does not alter the binding affinity for the complexation because the predicted $\Delta \Delta G$ is about 0.16 kcal/mol. We also analyzed the impact of double mutations in the YGFQ sequence on the stability of the complexation, see Table 1. To this end, we fixed, for example G496D and replaced Q498 by the residues F, H, N, and T that appear in strains of other species. The predicted changes in binding affinity by mutations indicate destabilization of new complexations.

2.2. Spike receptor-binding motifs in bats

It is known that not all SARS-CoV strains isolated from bat hosts have exploited ACE2 as a cellular attachment. Therefore, the set of amino acid sequences displayed in Figure 5 may exemplify the successful relation between virus evolution and the binding mechanism. This set highlights in medium purple color the preserved amino acid residues in the sequence YGFY, characteristics of human SARS-CoV. For comparison, we also display CoV strains with mutations in that SARS-CoV pattern to explore the relation between the hypothesized mechanism and the cell receptor recognition.

It has been demonstrated that LYRa11 (Letko et al., 2020), Rs3367 (Ge et al., 2013), Rs4874 (Hu et al., 2017), WIV1, and WIV16 (Letko et al., 2020; Yang et al., 2015), have the capacity to use ACE2 for cell entry as well RaTG13, in line with our hypothesis. Also, the near single-point mutation Y → F in the next six strains Rs7327, Rs9401, YN2018B, RsSHC014, Rs4084, and Rs4231, does not interfere, as expected, in the attachment mechanism. This conclusion is supported by cell entry studies for Rs7327 (Hu et al., 2017; Letko et al., 2020), Rs9401, RsSHC014, Rs4084, and Rs4231 (Hu et al., 2017), because they are in a group that is likely to use the ACE2 receptor. This mutation replaces a hydrophobic residue by another one with higher hydrophobicity, reinforcing the conformational topology for binding with the receptor. This single-point mutation Y→F in human SARS-CoV produced a neutral effect in the binding hACE2 (Table 1).

We identified in the next group constituted by BtKY72, BB9904/BGR/2008, and BM48-31/BGR/2008, respectively the mutations Y487T, Y488T, and Y485T, decreasing the initial hydrophobicity of the expected pocket in these strains. It seems unlikely that this mutation and amino acid residue deletions associated to Tor2 RBM sequence affect the YGF-based attachment mechanism for BtKY72 and BB9904/BGR/2008. Unfortunately, there is no available experimental data concerning their receptors. It is important to remark that the residue F492 in BM48-31/BGR/2008 produces another hydrophobic sequence IGF at residues 490–492 (Figure 5). We speculate that this double occurrence may disrupt the aforementioned mechanism because of indications that BM48-31/BGR/2008 does not interact, at least with human ACE2 (Letko et al., 2020). No other GF sequence occurs in the RBD of these strains.

Next CoV strains in Figure 5 do not contain such specific YGF sequences of residues in the RBM neither in their RBD. We find the two-letter sequence GF in Rf1/2004, but it is located in RBD and with GF surrounded by hydrophilic residues. Although we have considered only part of the sequences that better align with RBM of Tor2, it has been demonstrated that the spikes of HuB2013, HKU3, CoVZC45, CoVZXC21, Rf1, Rf4092, and Shaanxi2011 do not use hACE2, a result that is not just a consequence of deletions at the RBD (Letko et al., 2020). Further support has been presented against HKU3 in using hACE2 (Gralinski & Menachery, 2020). It seems unlikely that Rm1/2004 infects hACE2 because its unfavorable binding free energy (Armijos-Jaramillo et al., 2020). Another result concludes that Rp3 is unable of infect hACE2 or even bat ACE2 (Hoffmann et al., 2013).

We have placed together the alignments involving Tor2, RaTG13, and SARS-CoV-2 at the end of Figure 5 for further comparison. The whole genome of RaTG13 shares 96% amino acid sequence identity with SARS-CoV-2, and it is considered the most closely related genome to this CoV (Zhang et al., 2020). Considering its spike protein, and RBM, RaTG13 shares respectively 97% and 76% amino acid identity with SARS-CoV-2. For comparison, RaTG13 shares 79%, 77%, and 53% identity, respectively, for the whole genome, spike protein, and RBM with SARS-CoV Tor2. Therefore, SARS-CoV-2 is mostly similar to RaTG13 than SARS-CoV strains in all regions.

2.3. Spike receptor-binding motifs in palm civets and pangolins

To explore further the role of YGF-based attachment mechanism, we exhibit comparative residue sequences for civet and pangolins, again aligned with RBM of SARS-CoV Tor2 (Figure 6). This figure shows that the pattern YGFY characteristic of human SARS-CoV is maintained for the collected data, but with a single-point mutation Y → H for pangolin hosts PCoV. It is worth to observe that even the shorter two-letter GF sequence is not found in the RBD of these strains, which could promote another hydrophobic pocket.

We have included SARS-CoV-2 on the last line of Figure 6 for a direct comparison. PCoV GX-P2V shares 79%, 77%, and 50% amino acid identity with Tor2, respectively for whole genome, spike protein, and RBM aligned with Tor2. In relation to SARS-CoV-2, PCoV GX-P2V shares 85%, 92%, and 75% amino acid identity, respectively for whole genome, spike protein, and RBM. It is believed that human SARS-CoV passed from palm civets to humans in the 2002–2003 epidemic because their genome sequences are highly similar (Cui et al., 2019; Li, 2008; Qu et al., 2005). The amino acid alignments show an almost identical RBM between human SARS-CoV, represented by Tor2 strain, and collected data from palm civet strains. This identification also includes the YGF-based mechanism able to use ACE2 proteins. Nevertheless, these alignments display high similarity between pangolins and SARS-CoV-2, which also support previous conclusions on pangolins being the probable origin of SARS-CoV-2 (Lam et al., 2020; Zhang et al., 2020). However, based on our data related to host receptor binding and their RBM and S-protein alignments, we cannot discard bat RaTG13-like strain as also the possible origin of SARS-CoV-2.

2.4. SARS-CoV and hCoV-NL63: only functionally related

Although there are no many available experimental data identifying the viral receptor-binding protein for CoVs, it is well established that human SARS-CoV and hCoV-NL63 both employ ACE2 as the cell receptor to infect host cells (Hofmann et al., 2005, 2006). Interestingly, SARS-CoV and hCoV-NL63 domains do not present high sequence similarity. For example, their spike-S1 subunities share only 10% in similarity. Other features separate SARS-CoV and hCoV-NL63 (Milewska et al., 2014). SARS-CoVs are classified as β- coronavirus with subgenus sarbecovirus, while hCoV-NL63 is in genus α-coronavirus and subgenus setracovirus. Although hCoV-NL63 also enters the cell via endocytosis, its functional receptor requires heparan sulfate proteoglycans for the initial attachment, representing an important extra factor for ACE2 to act as a functional receptor (Milewska et al., 2014, 2018). Moreover, the spike-S1 glycoprotein of SARS-CoV binds more efficiently ACE2 than the corresponding spike-S1 of NL63 (NL63-S) (Glowacka et al., 2010). This may be linked to the fact that SARS-CoV and NL63-S contact ACE2 differently, a conclusion based upon the experimental results that NL63-S does not bind to ACE2 through a single and large domain (Hofmann et al., 2006; Wu et al., 2009). Actually, different RBD regions have been identified within NL63-S. One of these regions was positioned at residues 476–616 and comprising three discontinuous RBM regions, RBM1 (residues 497–501), RBM2 (residues 530–540), and RBM3 (residues 575–594) (Li et al., 2007; Lin et al., 2008, 2011). A slightly different RBD has been identified for this CoV (Wu et al., 2009). It would be located at residues 482–602, also with three discontinuous RBM regions, which surround a shallow cavity at hCoV–NL63–ACE2 binding interface. Curiously, its spike protein alignment with Tor2 does not show the expected residue pattern in the corresponding RBM of Tor2 nor in the aforementioned RBD regions of NL63-S. This may help to explain the unusual pathway of binding to ACE2 for this CoV.

3. Methods

The analysis of inter-molecular contacts in the experimentally determined high resolution crystal structures complexed with hACE2 (PDB ID: 2AJF (Li, Farzan, et al., 2005) and PDB ID: 6LZG (Wang et al., 2020)) was performed with the bioinformatic tool Contacts of Structural Units (CSU) (Sobolev et al., 1999).

Single and multiple residue mutations were introduced at positions located on the hydrophobic surface patch to evaluate the importance of these residues in establishing specific interactions with hACE2. Mutations may affect the spike receptor-binding complexed with hACE2 either leading to higher, lower or even neutral binding affinity. We apply the fast and accurate MutaBind2 method (Zhang et al., 2020) to estimate the binding free-energy change $\Delta \Delta G=\Delta G^{\text{mut}}-\Delta G^{\text{wt}}$ upon mutation to predict its functional effects. This method compares free-energy changes between mutated and wild-type three-dimensional conformations. The binding free-energy change upon single mutation was also evaluated with the predictor BeAtMuSiC (Dehouck et al., 2013) based on a set of statistical potentials extracted from experimental mutational data. This computational method predicted very similar effects on the complexation (data not shown) as described in Table 1.

3.1. Sequence alignment

We have also used the bioinformatic tools BLAST and ClustalW for sequence alignment and analysis of CoV strains, and Jalview to examine and edit various sequence alignments. Figures showing the conformational complexations were prepared using PyMol. The list of GenBank accession codes for the spike proteins analyzed in this work is available in Supplementary Table S1.

4. Conclusion

We have analyzed a number of CoV strains to support the hypothesis that SARS-CoV and SARS-CoV-2 strains share a common evolutionary mechanism for the initial attachment to ACE2. Moreover, we speculate that the YGF-based mechanism can act as a protein signature to distinguish CoVs able to use ACE2 as a cell entry receptor whenever this residue sequence is located at the CoV RBM region. For example, bat-SL-CoV ZC45 and ZXC21 are closely related sequences to SARS-CoV-2 with overall genome identity of ∼89% and can be promptly put under suspicious in their ACE2 binding affinity because the lack of such signature. Of course, as exemplified by hCoV-NL63, we cannot discard that another mechanism can act helping such ACE2 binding. It must be accentuated that the occurrence of other XGF sequences, mainly with X being a hydrophobic residue, in the RBM, or even in the RBD region, can disrupt the proposed topological mechanism for ACE2 binding. This because it might introduce hydrophobic loops promoting a new ligand–substrate recognition.

Funding Statement

P.P.D.C. and N.A.A. gratefully acknowledge financial supports from the Brazilian agencies CAPES, and FAPESP, process 2015/16116-3, respectively.

Disclosure statement

The authors declare no competing financial interests.