Molecular insights into receptor binding of recent emerging SARS-CoV-2 variants

Multiple SARS-CoV-2 variants of concern (VOCs) have been emerging and some have been linked to an increase in case numbers globally. However, there is yet a lack of understanding of the

molecular basis for the interactions between the human ACE2 (hACE2) receptor and these VOCs. Here we examined several VOCs including Alpha, Beta, and Gamma, and demonstrate that five

variants receptor-binding domain (RBD) increased binding affinity for hACE2, and four variants pseudoviruses increased entry into susceptible cells. Crystal structures of hACE2-RBD complexes

help identify the key residues facilitating changes in hACE2 binding affinity. Additionally, soluble hACE2 protein efficiently prevent most of the variants pseudoviruses. Our findings

provide important molecular information and may help the development of novel therapeutic and prophylactic agents targeting these emerging mutants.

Coronavirus disease 2019 (COVID-19) which is caused by SARS-CoV-2 has rapidly been declared as pandemic since its identification in late December 20191,2. Up-to date, although tremendous

efforts have led to the development of vaccine, combine with the preventive measures prescribed by the government around the world, the transmission of the virus among human are still going

and numerous SARS-CoV-2 mutations with uncertain consequences for viral replication and transmission are being increasingly identified. The mutation of the spike (S) protein in SARS-CoV-2

has drawn wide concerns because S proteins mediate viral entry via their interaction with the angiotensin-converting enzyme 2 (ACE2) receptor3,4 and are the major target for vaccine

development. SARS-CoV-2 variant D614G in the S protein rapidly become dominant around the world5. Recently, several novel SARS-CoV-2 variants of concern (VOCs) carrying D614G mutation have

been linked to an increased number of infections at a global scale (Supplementary Fig. 1).

With the addition of variants circulating in humans, mutants have also been identified in susceptible farmed animals for fur. In Denmark, a SARS-CoV-2 variant referred to as the “Cluster 5”

variant12, and designated as Mink-Y453F in this study, was identified in farmed minks and subsequently found to be transmissible between minks and humans13. Besides, both F486L and N501T

mutations have been found in SARS-CoV-2 isolates from minks, as well as humans14, and these strains were designated as Mink-F486L and Mink-N501T in this study, respectively (Supplementary

Fig. 1). Given that residues at positions 417, 453, 486, and 501 are involved in the interactions between the RBD and human ACE2 (hACE2)4,15, it is important to understand the detailed

binding of these novel variant RBDs to hACE2.

Here, we examined six different SARS-CoV-2 RBD variants, including Alpha, Beta, Gamma, Mink-Y453F, Mink-F486L, and Mink-N501T. We found that these dominant mutations result in enhanced

binding affinity to hACE2 receptor except for Mink-F486L, and that four variants pseudovirus particles except for Gamma and Mink-F486L show enhancement of viral entry to human cells.

Analysis of molecular features of crystal structures revealed that replacements of key amino acids change bonding forces in the interaction interface. In addition, we demonstrate that the

soluble hACE2 protein efficiently inhibit most of SARS-CoV-2 variants infections. Taken together, these data highlight the entry of new emerging variants and will provide a useful

information to develop antiviral drugs against SARS-CoV-2.

The binding of RBD to the receptor hACE2 is an essential step for virus to entry cell, thus we first explored if the binding of six SARS-CoV-2 variant (including Alpha, Beta, Gamma,

Mink-Y453F, Mink-F486L and Mink-N501T) RBDs to hACE2 were changed in comparison with wild-type (WT) SARS-CoV-2 (the first strain of SARS-CoV-2 which was isolated from a clinical patient on

Jan 6, 2020, GISAID: EPI_ISL_402119) RBD. The SARS-CoV-2 WT RBD and variant RBDs were purified and then evaluated for hACE2 binding using flow cytometry (FACS) (Supplementary Fig. 2).

SARS-CoV-2 Alpha, Beta, and Gamma, as well as Mink-Y453F and Mink-N501T RBDs, bound to the hACE2-expressing cells with higher affinity than WT RBD, with 33.4% positive cells (for WT RBD)

versus 82.7% (for Alpha RBD), 48.1% (for Beta RBD), 57.7% (for Gamma RBD), 84.7% (for Mink-Y453F RBD), and 76.8% (for Mink-N501T RBD), respectively (Fig. 1). While Mink-F486L RBD bound to

fewer hACE2-expressing cells with a 1.65% positivity rate (Fig. 1).

BHK cells stably expressing GFP and hACE2 were incubated with His-tagged MERS-CoV RBD, SARS-CoV-2 WT RBD, Alpha RBD, Beta RBD, Gamma RBD, Mink-Y453F RBD, Mink-N501T RBD, and Mink-F486L RBD,

respectively. APC anti-His antibodies were used to detect the His-tagged protein binding to the cells. Representative results from three experiments are shown. The mean ± SD percentages of

RBD-binding cells in the three experiments are shown in the right lower bar chart. Statistical significance was analyzed using one-way ANOVA with a Tukey’s multiple comparison test for

multiple groups.

To better understand the interactions between SARS-CoV-2 variant RBDs and hACE2, we measured their binding affinity using surface plasmon resonance (SPR). Mouse Fc (mFc) tagged hACE2 was

captured in a CM5 chip that pre-immobilized with anti-mFc antibody, and the serially diluted variant RBDs were flowed through the chip. As shown in Fig. 2, SARS-CoV-2 WT RBD protein bound to

hACE2 with an equilibrium dissociation constant (KD) of ~26.34 nM (Fig. 2a), which is similar to the previous results16,17. Alpha RBD, Beta RBD, Gamma RBD, Mink-Y453F RBD, and Mink-N501T

RBD displayed higher affinities to hACE2 than WT RBD, with ~7, ~3, ~5, ~8, and ~4-fold increase in binding strength, respectively (Fig. 2b–f). Notably, both Beta RBD and Gamma RBD contain

two more mutated residues than Alpha RBD, but displayed a little lower binding affinities for hACE2 than Alpha RBD. In order to understand which residues in Beta RBD and Gamma RBD contribute

to the lower affinities, three single mutations (including K417N, K417T, and E484K), and three double mutations (containing N501Y/E484K, N501Y/K417N, and N501Y/K417T) were prepared and used

to measure the binding affinities for hACE2. Both K417N and K417T mutations in RBD decreased ~2 fold affinity for hACE2, while E484K mutation exerted little effect (Supplementary Fig. 3).

Similarly, both N501Y/K417N RBD and N501Y/K417T RBD also displayed ~2 fold lower affinity than N501Y RBD, but N501Y/E484K did not (Supplementary Fig. 3). Although both N501Y and N501T

strengthened the interactions between RBD and hACE2, N501Y exhibits higher affinity than N501T. In contrast, Mink-F486L RBD bound to hACE2 with ~4-fold lower binding affinity than WT RBD

(Fig. 2g). These results were consistent with the FACS assays. In a word, five variant (Alpha, Beta, Gamma, Mink-Y453F, and Mink-N501T) RBDs but not Mink-F486L RBD, increased association

with hACE2.

a–k Mouse Fc (mFc)-fused hACE2 or miACE2 in the supernatant was captured in the CM5 chip via its interaction with the pre-immobilized anti-mFc antibody. Various concentrations of SARS-CoV-2

WT RBD (a), Alpha RBD (b), Beta RBD (c), Gamma RBD (d), Mink-Y453F RBD (e), Mink-N501T RBD (f), and Mink-F486L RBD (g) protein were used to evaluate their binding affinity for hACE2.

Serially diluted WT RBD (h), Mink-Y453F RBD (i), Mink-N501T RBD (j), and Mink-F486L RBD (k) protein were measured the binding to miACE2. KD, ka, and kd values are all recorded and the

representative results from three experiments are shown. The data are presented as the mean ± SEM of three independent replicates (n = 3).

Current data indicate that Mink-Y453F, Mink-N501T, and Mink-F486L emerged in farmed minks, when we evaluated their RBDs binding affinity for mink ACE2 (miACE2, Neovison vison13,18), we noted

that WT RBD bound to miACE2 with ~8.16 μM affinity, which was ~310-fold lower than the affinity of WT RBD to hACE2 (Fig. 2h). In addition, Mink-Y453F RBD, Mink-N501T RBD, and Mink-F486L RBD

exhibited much higher binding affinity for miACE2 than WT RBD with ~21, ~9, and ~15 fold, respectively (Fig. 2i–k). This provides some molecular evidence to explain why the mutant strains

efficiently transmitted among minks.

To further elucidate the molecular mechanism underlying the binding of these SARS-CoV-2 variant RBDs to hACE2, we prepared for RBD-hACE2 complexes and obtained five crystal structures,

namely Alpha RBD-hACE2, Beta RBD-hACE2, Gamma RBD-hACE2, Mink-Y453F RBD-hACE2, and Mink-F486L RBD-hACE2 at a resolution of 2.9 Å, 2.6 Å, 2.8 Å, 2.4 Å, and 2.7 Å, respectively (Table S1).

Each complex structure was comprised of one copy of the RBD-hACE2 molecule in one asymmetric unit. The overall structure of Alpha RBD-hACE2, Beta RBD-hACE2, Gamma RBD-hACE2, Mink-Y453F

RBD-hACE2, and Mink-F486L RBD-hACE2 were similar to the WT RBD-hACE2 structure, with a root-mean-square deviation (RMSD) of 0.196 Å (for 736 Cα atoms), 0.109 Å (for 725 Cα atoms), 0.198 Å

(for 743 Cα atoms), 0.158 Å (for 712 Cα atoms) and 0.150 Å (for 763 Cα atoms), respectively (Supplementary Fig. 4), when compared with the WT RBD-hACE2 (PDB: 6LZG) structure.

For the detailed mutated positions (Supplementary Fig. 5), in Alpha RBD-hACE2, Beta RBD-hACE2, and Gamma RBD-hACE2 structures, N501 is located in a loop structure and could be replaced by a

Y without creating folding problems, it seems that the N501Y does not induce a large conformational change. Some very weak hydrogen bonds are possible between N501 and ACE2, but the phenyl

of the Y501 side chain could make many new favorable nonbonded interactions with hACE2, i.e., a cation-π interaction with hACE2 K353 and a π-π stacking interaction with hACE2 Y41 (Fig. 

3a–c). These noncovalent interactions in Y501 variant are stronger as compared to the WT type. Thus, Y501 significantly increased the interaction between RBD and hACE2 compared to N501, so

the N501Y replacement in this region of the interface should be favorable for the interaction with hACE2. These results were consistent with cryo-electron microscopy structures of the N501Y

SARS-CoV-2 spike protein in complex with ACE219. In addition, the E484K substitutions would seem neutral or even unfavorable because it is far away from the RBD-interacting residues of hACE2

(Fig. 3a, b). This can also be confirmed by the results of the binding affinity of hACE2 with E484K single or double mutation constructions (Supplementary Fig. 3). Both K417N mutation in

Beta RBD and K417T mutation in Gamma RBD destroyed the salt bridge formed by K417 and hACE2 D30 (Fig. 3a, b). Therefore, Beta RBD and Gamma RBD exhibited a higher binding affinity for hACE2

than WT RBD, but a little lower than Alpha RBD, consistent with the SPR results. In the Mink-F486L RBD-hACE2 structure, the mutation at F486L impairs the π-π stacking interaction formed by

RBD F486 and hACE2 Y83 (Fig. 3d), resulting in the decreased interaction between Mink-F486L RBD and hACE2. We also predicted the structure of miACE2 and compared it with both Mink-F486L

RBD-hACE2 and WT RBD-hACE2 structures. The residue T82 in miACE2 was clashed with the phenyl group of RBD F486 but not L486 (Supplementary Fig. 6). Thus, F486L mutation may be helpful for

the interaction of SARS-CoV-2 RBD with miACE2.

a–e The WT RBD-hACE2 structure (PDB: 6LZG) is shown in the center. Superimposition of WT RBD-hACE2 and each variant RBD-hACE2 (including Beta RBD-hACE2 (a), Gamma RBD-hACE2 (b), Alpha

RBD-hACE2 (c), Mink-F486L RBD-hACE2 (d), Mink-Y453F RBD-hACE2 (e)) are shown in each surrounding panel. In each structure the hACE2 is colored in light pink. SARS-CoV-2 WT RBD, Beta RBD,

Gamma RBD, Alpha RBD, Mink-F486L RBD, and Mink-Y453F RBD are colored in gray, cyan, orange, yellow, green, and magenta, respectively. The key contact residues are shown as stick structures

and labeled appropriately. The cation-π interaction, π-π stacking interaction, salt bridge, and hydrogen bonds are colored in magenta, blue, orange, and yellow, respectively. Hydrogen bond

interactions were analyzed at a cutoff of 3.5 Å. f The detailed hydrogen bonds between the Mink-Y453F RBD and hACE2 are shown.

Considering that the binding affinity of SARS-CoV-2 variants RBD to hACE2 are changed, we further tested the potential influence of the SARS-CoV-2 variants on cellular infection using

pseudovirus transduction. The same amount of pseudovirus that incorporate into the various SARS-CoV-2 variants S protein were infected hACE2-positive Huh7 cells and the GFP of pseudovirus

was quantified by FACS for the transduction efficiency. As we saw in the binding affinity assays, pseudovirus particles of Alpha, Beta, Mink-N501T, and Mink-Y453F, but not Mink-F486L, showed

increased transduction efficiency when compared to the D614G strain in Huh7 cells (Fig. 4a–f, and h). However, inconsistent with binding affinity, Gamma pseudovirus displayed the similar

transduction efficiency with the D614G pseudovirus (Fig. 4g, h). In order to exclude the influence of binding of RBD to hACE2 on the different transduction efficiency between Beta and Gamma,

N417T mutation was introduced to Beta S protein, which was designated as Beta-N417T that contains the same RBD sequence as Gamma. The transduction efficiency of Beta-N417T pseudovirus was

similar to Beta pseudovirus and higher than Gamma pseudovirus (Supplementary Fig. 8). It suggested that other substitutions outside of the RBD also contribute to the change in transduction

efficiency. Taken together, four out of six variants exhibited increased transduction efficiency.

a–g The SARS-CoV-2 D614G (a), Alpha (b), Beta (c), Mink-N501T (d), Mink-Y453F (e), Mink-F486L (f), and Gamma (g) pseudoviruses’ entry into Huh7 cells as evidenced by GFP expression in

transduced cells. The representative results from three experiments are shown. h The GFP-positive cells were quantified using FACS and representative results from three experiments are

shown. The values indicate the mean of the three experiments and the bar suggest the SD. Relative infectivity was normalized against that of the D614G pseudovirus. Statistical significance

was analyzed using a one-way ANOVA with Tukey’s multiple comparison test for multiple groups.

The EC50 values and folds change values are shown on the right panel. The folds change was normalized to WT, and the EC50 values are presented as the mean ± SEM of three independent

replicates (n = 3). The representative results from three experiments are shown.

The deep mutational scanning suggests that single residue mutation K417N or K417T is likely to have minimal effect on the binding to hACE2, and that the E484K mutation may predictably

enhance the binding21, while in this study the structures of Beta RBD-hACE2 and Gamma RBD-hACE2 indicated the mutations, K417N and K417T, decreased the RBD-hACE2 interactions while the

mutation E484K displayed little impact. The structural information was confirmed by the SPR results that the single mutation (K417N or K417T) or double mutation (N501Y/K417N or N501Y/K417T)

RBD displayed a lower binding affinity for hACE2 than WT RBD or RBD N501Y, but E484K or N501Y/E484K did not. Interestingly, the decreased impact caused by K417N or K417T could be

quantitatively counteracted by the increased binding effect associated with the N501Y mutation.

SARS-CoV-2 may acquire adaptive mutations that ensure efficient viral replication and transmission in other species, for example, by optimizing the interaction with host ACE2. The RBDs of

three mink-origin variants: Mink-Y453F, Mink-F486L, and Mink-N501T, displayed a higher binding capacity to mink ACE2. These mutations may be adapted for the efficient use of mink ACE2 for

entry. Thus, the number of both Mink-Y453F and Mink-F486L S sequences grew rapidly before the beginning of November 2020 (Supplementary Fig. 9c, d). However, Mink-N501T, whose binding

affinity for hACE2, transduction efficiency, and frequency of S sequences (Supplementary Fig. 9e) were similar with both Alpha and Beta, was likely to adapt to transmission among humans, but

Mink-F486L not. This may partly explain why Mink-F486L did not efficiently transmit to humans and suddenly disappeared following the implementation of the mink cull policy in the

Netherlands and Denmark (Supplementary Fig. 9d). While, increasing human samples were detected to carry N501T (Supplementary Fig. 9e). However, Mink-N501T has not drawn enough attention.

Considering SARS-CoV-2 has been detected in farmed minks in ten countries in Europe and North America28, our results suggest it should be evaluated in more detail in the future, especially

in people who live or work on mink farms and are close to mink.

As more SARS-CoV-2 variants continue to emerge and the major SARS-CoV-2 variants continue to spread, characterization of the hACE2-binding affinity and transduction efficiency of SARS-CoV-2

variants will help us understand SARS-CoV-2 transmission. The molecular features of variant RBDs binding to hACE2 provides valuable information helping us understand the entry mechanism of

SARS-CoV-2 variants and aiding in the development of novel vaccines and specific drugs that target the SARS-CoV-2 entry process.

Expi293F cells (Gibco) were cultured at 37 °C in SMM 293-TII Expression Medium with 5% CO2 in a shaking incubator (140 rpm). Sf9 cells (Invitrogen) and High Five cells (Invitrogen) were

cultured at 27 °C in Insect-XPRESS medium (LONZA) in a shaking incubator (120∼130 rpm).

HEK293T (ATCC), BHK21 (ATCC) and Huh7 cells (3111C0001CCC000679) were cultured at 37 °C in Dulbecco’s Modified Eagle medium (DMEM) supplemented 10% fetal bovine serum (FBS) at 5% CO2.

The coding sequence of hACE2-mFc (residues 1-740, GenBank: NP_001358344) or miACE2-mFc (residues 1-740, GenBank: QPL12211) were cloned into pCAGGS vector (MiaoLingPlasmid). The plasmids were

transiently transfected into HEK293T cells using PEI and then, 48 h later, the cell supernatants were collected, concentrated and used in the SPR assays.

The DNA sequence encoding hACE2 (residues 19-615, GenBank: NP_001358344) was inserted into the Baculovirus transfection vector pFastBac1 (Invitrogen) using the EcoRI and XhoI restriction

sites. The gp67 signal peptide sequence was added to the N-terminus of the hACE2 gene for protein secretion, and the Hexa-His tag sequence was added to the C-terminus of the hACE2 sequence

for protein purification. The hACE2 protein was expressed using the Bac-to-Bac Baculovirus expression system and used for crystallization. The pFastBac1-hACE2 plasmids were transformed into

DH10Bac E. coli to produce recombinant bacmids. Transfection of the bacmids using FuGENE 6 Transfection Reagent (Promega) and virus amplification were carried out in Sf9 cells, and the

proteins were expressed in High Five cells. The supernatants were collected 48 h post-infection.

The DNA sequences encoding hACE2 (residues 1-740, GenBank: NP_001358344) were cloned into the pCAGGS vector with Hexa-His tag at the C-terminus. The DNA sequences encoding SARS-CoV-2 WT RBD

(spike residues 319-541, GISAID: EPI_ISL_402119) or MERS-CoV RBD (spike residues 367-606, GenBank: JX869050) were inserted into the pCAGGS vector with IL10 signal peptide sequence at the

N-terminus and the Hexa-His tag at the C-terminus. The SARS-CoV-2 variant RBD plasmids (including Alpha RBD, Beta RBD, Gamma RBD, Mink-Y453F RBD, Mink-F486L RBD, Mink-N501T RBD, RBD

N501Y/E484K, RBD N501Y/K417N, RBD N501Y/K417T, RBD K417N, RBD K417T, and RBD E484K) were constructed via site-directed mutagenesis using the Mut Express II Fast Mutagenesis Kit V2 (Vazyme).

The recombinant RBD and hACE2 (used for pseudovirus neutralization assays) proteins were expressed in Expi293F cells after plasmid transfection using Sinofection Transfection Reagent (Sino

Biological). The supernatants were collected 5 days post-transfection.

The supernatants containing hACE2 or RBD proteins were purified via affinity chromatography using a HisTrap HP 5 mL column (GE healthcare) and the target proteins were eluted in an elution

buffer composed of 20 mM Tris (pH 8.0), 150 mM NaCl, and 300 mM imidazole. The samples were then purified using gel-filtration chromatography on a HiLoad 16/600 Superdex 200PG column (GE

healthcare) in a buffer containing 20 mM Tris (pH 8.0) and 150 mM NaCl.

Purified hACE2 and each SARS-CoV-2 variant RBD protein (including Alpha RBD, Beta RBD, Gamma RBD, Mink-Y453F RBD, and Mink-F486L RBD) were mixed and incubated on ice for 2 h. The mixture was

then purified on HiLoad 16/600 Superdex 200PG column in a buffer containing 20 mM Tris (pH 8.0) and 50 mM NaCl. The SARS-CoV-2 variant RBD-hACE2 complex proteins were then concentrated to

15 mg/mL for crystallization. All crystallizations were performed using a vapor-diffusion sitting-drop method with 0.8 μL protein mixing with 0.8 μL reservoir solution at 18 °C. High-quality

crystals for both the Beta RBD-hACE2 and Gamma RBD-hACE2 complexes were obtained when using 0.1 M MES (pH 6.5), 12% w/v PEG 20000 at a concentration of 15 mg/mL at 18 °C. Complex crystals

of Alpha RBD-hACE2, Mink-Y453F RBD-hACE2, and Mink-F486L RBD-hACE2 were grown in 0.1 M MES (pH 6.5), 10% w/v PEG 5000 MME, 12% v/v1-Propanol at a concentration of 15 mg/mL at 18 °C.

Prior to collecting diffraction data, all crystals were cryo-protected by briefly soaking in reservoir solution supplemented with 20% (v/v) glycerol and then flash-cooled in liquid nitrogen.

All X-ray diffraction data were collected at Shanghai Synchrotron Radiation Facility (SSRF) BL17U. The datasets were indexed, integrated, and scaled using HKL200034. The structures of

variant RBD-hACE2 were determined via molecular replacement method using Phaser35 with the previously reported structures of SARS-CoV-2 RBD-hACE2 (PDB: 6LZG) as a search model. The atomic

models were built using Coot 0.8.236 and the refinements were completed using Phenix.refine37. MolProbity was used to assess the stereochemical quality of the final models38. The data

collection, processing, and refinement statistics were summarized in Supplementary Table 1. All structural figures were generated using the PyMOL 4.5 software (https://pymol.org/2/).

The SPR assays were performed to test the interactions between mFc-fused ACE2 (including hACE2 and miACE2) and SARS-CoV-2 variant RBDs using a BIAcore 8 K (GE Healthcare) with a CM5 chip (GE

Healthcare) at 25 °C in single-cycle mode. SARS-CoV-2 WT RBD was used as a positive control. The buffer system was PBST (10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4, 137 mM NaCl, 2.7 mM KCl, 0.005%

Tween 20) and the anti-mIgG antibody (Cytiva) was pre-immobilized on the CM5 chip using standard amine coupling chemistry with a 50 μg/mL concentration. Concentrated supernatant containing

hACE2-mFc or miACE2-mFc protein was captured on the chip using this immobilized antibody. Serially diluted WT RBD (12.5, 25, 50, 100, 200 nM), Alpha RBD and Mink-Y453F RBD (3.125, 6.25,

12.5, 25, 50 nM), Beta RBD, Gamma RBD, and Mink-N501T RBD (6.25, 12.5, 25, 50, 100 nM), Mink-F486L RBD (25, 50, 100, 200, 400 nM), RBD K417N and RBD K417T (20, 40, 80, 160, 320 nM), RBD

N501Y/E484K (0.5, 1, 2, 4, 8 nM), RBD N501Y/K417N (4, 8, 16, 32, 64 nM), RBD N501Y/K417T (2, 4, 8, 16, 32 nM), and RBD E484K (5, 10, 20, 40, 80 nM) were flowed over the chip to evaluate

hACE2 binding. Various concentrations of WT RBD (2.5, 5, 10, 20, 40 μM), Mink-Y453F RBD (40, 80, 160, 320, 640 nM), Mink-F486L RBD (160, 320, 640, 1280, 2560 nM), and Mink-N501T RBD (320,

640, 1280, 2560, 5120 nM) were also used to evaluate miACE2 binding. The chip was regenerated after each reaction using glycine (pH 1.7). The equilibrium dissociation constants (KD) of each

pair of RBD-hACE2 interaction and RBD-miACE2 interaction were analyzed using a 1:1 binding model and steady state affinity model in the Biacore Insight Evaluation 2.0.15.12933 software (GE

Healthcare), respectively. These results were then visualized using Graphpad Prism 8.

These assays used a stable BHK21 cell line expressing hACE2 which was first constructed. Briefly, lentiviruses were packaged in HEK293T cells co-transfected with pLV-C-GFPSpark-hACE2 (Sino

Biological) and helper plasmid pLP1, pLP2, and pLP/VSV-G (Invitrogen) at a ratio of 20:20:13:5 using Lipofectamine 2000. After 72 h, the supernatants containing the hACE2 lentiviruses were

collected and used to infect BHK21 cells. GFP-positive cells were selected using a BD FACSAriaIII (Becton, Dickinson and Company) and separated into 96-well cell culture plates. A single

positive colony was then picked for further culture. The cells were stored in liquid nitrogen following two times of cell sorting. The hACE2-expressing cells were grown for approximately 12

passages, and incubated with Phosphate Buffer Saline (PBS) or 2 μg/mL RBD protein (including MERS-CoV RBD, SARS-CoV-2 WT RBD, Alpha RBD, Beta RBD, Gamma RBD, Mink-Y453F RBD, Mink-N501T RBD,

and Mink-F486L RBD) at 37 °C for 30 min. MERS-CoV RBD was used as a negative control. After washing with PBS, the cells were stained with APC anti-His tag antibody (1:500; BioLegend) at 37 

°C for 30 min. These cells were then washed and resuspended in 200 μL PBS before being evaluated using a BD FACSCanto II (Becton, Dickinson and Company). The percentage of RBD-binding cells

can be described as the ratio of RBD-binding cells (Q2) to hACE2-positive cells (Q2 and Q3). Each group comprised at least three replicates and the FACS graphics were generated using FlowJo

V7.6 software. Statistical analysis was performed using Graphpad Prism 8.

Briefly, both the WT RBD-hACE2 (PDB: 6LZG) and Mink-Y453F RBD-hACE2 structures were stripped of their N-acetyl-β-glucosaminide glycans and crystal structural waters and then used in ten

parallel molecular dynamics (MD) simulations with different random seeds. All MD simulations were performed using GROMACS (version 2020.5)39 on GPU with the CHARMM36 protein force field and

TIP3P water model40,41. All calculations were applied using an atom-based truncation scheme and updated heuristically with a list cutoff of 12 Å, a non-bond cutoff of 12 Å, and a force

switching function initiated at 10 Å for Van der Waals interactions. Long-range electrostatic interactions were computed using Particle Mesh Ewald method with fourth-order cubic

interpolation and 1.6 Å grid spacing42. To equilibrate solvent molecules around the solute, each system was minimized using the steepest descent algorithm with a maximum force of