Ten emerging SARS-CoV-2 spike variants exhibit variable infectivity, animal tropism, and antibody neutralization

Emerging mutations in SARS-CoV-2 cause several waves of COVID-19 pandemic. Here we investigate the infectivity and antigenicity of ten emerging SARS-CoV-2 variants—B.1.1.298, B.1.1.7(Alpha), B.1.351(Beta), P.1(Gamma), P.2(Zeta), B.1.429(Epsilon), B.1.525(Eta), B.1.526-1(Iota), B.1.526-2(Iota), B.1.1.318—and seven corresponding single amino acid mutations in the receptor-binding domain using SARS-CoV-2 pseudovirus. The results indicate that the pseudovirus of most of the SARS-CoV-2 variants (except B.1.1.298) display slightly increased infectivity in human and monkey cell lines, especially B.1.351, B.1.525 and B.1.526 in Calu-3 cells. The K417N/T, N501Y, or E484K-carrying variants exhibit significantly increased abilities to infect mouse ACE2-overexpressing cells. The activities of furin, TMPRSS2, and cathepsin L are increased against most of the variants. RBD amino acid mutations comprising K417T/N, L452R, Y453F, S477N, E484K, and N501Y cause significant immune escape from 11 of 13 monoclonal antibodies. However, the resistance to neutralization by convalescent serum or vaccines elicited serum is mainly caused by the E484K mutation. The convalescent serum from B.1.1.7- and B.1.351-infected patients neutralized the variants themselves better than other SARS-CoV-2 variants. Our study provides insights regarding therapeutic antibodies and vaccines, and highlights the importance of E484K mutation.

F ollowing the discovery of SARS-CoV-2, the emergence of multiple variants has been reported 1,2 . Mutations of the virus may cause changes in its infectivity and pathogenicity, resulting in the emergence of highly infectious or lethal mutant strains, they may also change the antigenicity of the virus, leading to failures of existing antibody treatments or the vaccine 1,3,4 . Additionally, mutations may cause cross-species transmission and the virus may undergo further evolution in the new host, triggering a new wave of virus spread 4 . Therefore, the mutations of SARS-CoV-2 have received close attention from scientists worldwide. Beginning in March 2020, the D614G mutant strain became the dominant strain globally, and the current prevalence has exceeded 95% 3 . In November 2020, the mink strain B.1.1.298 (cluster 5) was reported to spread between humans and minks 4,5 . Since December, increasing numbers of SARS-CoV-2 variants have been reported worldwide, among which B.1.1.7, B.1.351, and P.1 have been listed as viruses of concern (VOCs) by the WHO 6 . As of March 2021, B.1.1.7(alpha, VOC 202012/01 or 501Y.V1), which first appeared in the United Kingdom, has spread to 125 countries; this variant exhibits increased transmissibility, risk of hospitalization, severity, and mortality [7][8][9] . B.1.351(beta or 501Y.V2) first appeared in South Africa and leads to immune escape of the spike protein because of mutation E484K in the RBD; this variant may influence the efficacies of vaccines and therapeutic monoclonal antibodies (mAbs) and sera 10,11 . P.1 (gamma, B.1.1.28.1, or 501Y.V3) and P.2 (Zeta), which first appeared in Brazil, led to the disappointment regarding Brazil's herd immunity dream and almost caused the collapse of Brazil's medical system 12,13 . Variants B.1.429 (Epsilon or B.1.427) in California and B.1.526 (Iota) in New York comprised the largest proportion of the new COVID-19 cases in those areas, eliciting widespread concern 1,14-17 . Moreover, the Nigerian variant B.1.525 (Eta), which contains subsets of mutations previously observed in variants B.1.1.7 and B.1.351, has spread rapidly in Nigerian and the United Kingdom 18,19 . Additionally, a new variant B.1.1.318 recently appeared in the United Kingdom; this variant requires close attention because of its E484K mutation 18,20 . In this study, we investigated the 10 currently prevalent variants (B.  (Fig. 1). We compared infectivity, host tropism, and neutralization characteristics with the D614G reference strain, with the aim of providing clues for the prevention and control of COVID-19, particularly with respect to designing mAbs and vaccines.

Results
Infectivities of 10 SARS-CoV-2 variants. The infectivities of the 10 variants and seven RBD-located single mutations were first tested in four SARS-CoV-2-susceptible cell lines, including two human cell lines (Huh-7, and Calu-3) and two non-human primate cell lines (LLC-MK2 and Vero). Notably, although most of the examined SARS-CoV-2 variants showed slightly increased infectivity, none of them had more than fourfold increased infectivity, compared with the D614G reference strain (Fig. 2a-d). The L452R single mutation and B.1.526-2 led to increased infectivity, whereas the B.1.1.298 variant exhibited significantly decreased infectivity in all the four cell lines (Fig. 2a-d). Moreover, the variants B.1.351 B.1.525 and B.1.526-2(E484K) showed significantly increased infectivity for Calu-3 cells (Fig. 2b). We further analyzed the reason for decreased infectivity of B.1.1.298. The single mutations M1229I may be the key mutation that caused the decreased infectivity of B.1.1.298 ( Supplementary  Fig. 1), while the expression level of B.1.1.298 spike protein was significantly decreased compared to D614G mutation, which may be responsible for the observed reduced infection ( Supplementary  Fig. 2).
The activity of furin/TMPRSS2/cathepsin L the ten SARS-CoV-2 variants. Because furin, TMPRSS2, and cathepsin L play important roles in SARS-CoV-2 infection 21-23 , mutation-related structural changes in the virus may influence the functions of these enzymes. We found that the ability of the D614G reference strain to infect 293T-ACE2 cells was significantly increased when furin, TMPRSS2, or cathepsin L was overexpressed (Fig. 2e). We subsequently investigated the infectivities of the variants in respective furin-, TMPRSS2-, or cathepsin L stable overexpressing cells. The results showed that the increased infectivity to the 293T-ACE2 cells in the presence of furin, TMPRSS2, or cathepsin L was further increased among most of the tested variants, excluding B.1.1.298 and B.1.1.318 in the TMPRSS2 group, as compared to the D614G reference strain (Fig. 2f-h).
Epitope analysis of the mAbs-RBD complex. To further analyze the reason behind the failure of mAbs to neutralize the tested variants, we performed a structural analysis of the mAbs-RBD complex based on the reported structural information [25][26][27] . mAb 7B8 binding to the RBD relies on five main hydrogen bonds. The interaction between S477 with E99 is particularly important. The S477N mutation weakens this interaction and causes escape (Fig. 4b). mAb X593 binding to the RBD relies on both salt bridges and hydrogen bonds, particularly the salt bridge E484-R102. This interaction is greatly weakened when the E484K mutation occurs. The contribution of L452 to the affinity between the two proteins is not as significant as that of E484K, causing only minor changes in local structures (Fig. 4b) Calu-3 **** **** *** ** **** **** **** in the process of mAb-CB6 binding to the RBD. K417N/T destroys the salt bridge, thus reducing the affinity significantly. Furthermore, N501Y causes minor changes in the local structure of the CB6-RBD complex, which also weakens the affinity between CDRL and the RBD (Fig. 4b). As for mAb 10D12, K417 and D104 located on CDRH3 form a salt bridge. The K417N/T mutation destroys the salt bridge, reducing affinity significantly. K417N/T and N501Y also cause minor changes in local structure, which further weakens the affinity between CDRL and RBD, especially Y503 on RBD (Fig. 4b) sites are located far from the interaction surface, the mutations causing local conformational changes in the RBD may be the reason behind binding failure.
Neutralization activities of immunized sera against 10 SARS-COV-2 variants. To determine whether the antigenicity change in the 10 variants could change their neutralizing sensitivity with respect to vaccine immunization, animals were immunized with different types of SARS-CoV-2 antigens: trimer spike protein (in mice), pseudotyped virus (in guinea pigs), recombinant DNA containing full-length spike gene (in guinea pigs) or purified RBD protein (in horses). The neutralization reactions elicited by immunized sera generated with these different antigens were compared among the 10 SARS-CoV-2 variants. Of the 10 variants, only B.  Fig. 5a-e). The reduced sensitivities were observed regardless of SARS-CoV-2 immunogen and source of animal serum (Fig. 5a-e). Furthermore, the neutralization activity against the K417T/N single-mutation strain was increased among all serum samples (Fig. 5e). The neutralization ID 50 of the various immunogens against the D614G reference strain was shown in Fig. 5f. Regarding the E484K-carrying variants, neutralization sensitivity of the RBD immunized sera reduced 2.6 to 6.2 folds, which is much obvious than other immunogens. On the other hand, the trimer protein immunized sera are much stable, the neutralization activity reduced only 1.1 to 1.8 folds. (Fig. 5a-e). Additionally, although the variants B.1.351 and P.1 carry nearly identical RBD mutations, variant B.1.351 exhibited a much greater reduction in neutralization sensitivity. Moreover, although P.2 only has one additional mutation (V1176F) in the S2 domain, its neutralization activity was less reduced in the pseudovirus-and RBD protein-induced sera compared with the E484K single-mutation strain (Fig. 5a, d).  (Fig. 6a). Convalescent sera from D614G infected patients showed a neutralization pattern similar to the pattern exhibited by SARS-CoV-2 immunized animal sera (Fig. 6b)  reference strain (Fig. 6c). B.1.1.7 variant-infected sera also showed the highest neutralization activity against the variant itself; it was comparatively resistant to B.1.351 and other E484K-carrying sera (Fig. 6d).
Neutralization activities of vaccine elicited sera. Neutralization ability of sera elicited by two vaccines approved in China was tested, including inactivated vaccine KCONVAC 28 (Shenzhen Kangtai Biological Products Co.) and adenovirus vaccine Ad5-nCoV 29 (CanSino Biologics Inc.). The results were similar to those observed in the immunized animal sera and D614G convalescent patient sera experiments. However, an obvious decrease in neutralization sensitivity of E484K-containing SARS-CoV-2 variants was apparent in the vaccine group, especially the inactivated vaccine group (Fig. 6e and f) compared to the convalescence sera group (Fig. 6b). The decrease in neutralization against P.1 was also less pronounced compared with B.1.351; however, the increase in neutralization against the K417T/N single-mutation strain was less than that of the immunized animal sera or convalescence sera group (Fig. 6e and f). Upon comparison of the two vaccines, the results were quite similar because they both targeted full-length spike protein. Overall, although the variants reduced the neutralization abilities of vaccines by approximately 0.9-5.0-fold for the inactive vaccineelicited sera and 0.9-3.1-fold for the adenovirus vaccine-elicited sera, both vaccines continued to exhibit a protective effect (Fig. 6e  and f).    been caused by community transmission or been detected in multiple countries. As of March 30, 2021, there were nine VOIs, of which three were variants of concern (VOCs; e.g., B.1.1.7, B.1.351 and P.1.) that showed increased transmissibility and potentially reduced neutralization by convalescent and postvaccination sera 6 . Furthermore, a previous study demonstrated increased binding of N501Y to mouse ACE2, which implies differences in host tropism 31 . Here, we evaluated the host tropism characteristics of 10 SARS-CoV-2 variants, which variants included most VOIs and additional potentially important variants. Pseudoviruses of multiple variants and single mutants at RBD sites presumably related to these variants were examined for their ability to infect four SARS-CoV-2 susceptible cells. Slightly enhancements of infectivity were observed among most of the tested variants, compared with the D614G reference strain, whereas the B.1.1.298 variants (mink cluster 5) displayed significantly decreased infectivity. Dr Fomsgaard's study using authentic virus showed that the mink cluster 5 was less pronounced and had an approximate 10-fold lower titer 24 h postinoculation compared to human SARS-CoV-2 isolates 32 , which was also consistent with our previous study focused on mink SARS-CoV-2 variants 33 . Further analysis showed that M1229I may be the key mutation that responsible for the reduced infectivity. Moreover, the increased infectivity of L452R mutation and B.

Adenovirus vaccine elicited sera
Ratio to D614G Ratio to D614G ** ** **** ** *** ** ** * Dashed lines indicate the threshold of fourfold difference. The experiments were repeated twice due to the sample limitation. e ID 50 ratios of sera from inactived-vaccine, compared with the D614G reference strain, are displayed using dot plots. f ID 50 ratios of sera from adenovirus-vaccine, compared with the D614G reference strain, are displayed using dot plots. b-f, data are presented as means ± SEMs. Asterisks indicate statistical significance.
virus and harvested the signal at around 10 4 -10 5 , it was not stable enough to yield reliable results. In addition, we also found slight enhancements of infectivity involving the B.1.351, B.1.525,  B.1.526, and B.1.1.318 variants, suggesting that these variants should receive close attention.
Furin, TMPRSS2 and Cathepsin L are important proteolytic enzymes, which are key regulators of SARS-CoV-2 infection 22,23 . Overexpression of these enzymes facilitates cellular infection. Surprisingly, almost all SARS-CoV-2 variants further enhanced the enzymes that promoted SARS-CoV-2 infection, especially when furin was overexpressed in 293T-ACE2 cells. When TMPRSS2 overexpressed infection was compared to mock 293T-ACE2 cells, the enhancement of infectivity by TMPRSS2 for the P.1 variant was almost fourfold that for the D614G variant. The increased affinity to ACE2 receptors caused by RBD mutations (e.g., E484K, N501Y) 31 or increased cleavage activity by mutations adjusted to the furin site (e.g., P681H) 35 may facilitate the function of enzymes. The underlying mechanisms of these increased infectivities require further analyses. The underlying mechanisms of these increased infectivities require further analyses.
Cross-species infections caused by viral mutations contribute to the extensive spreading of many animal-derived viruses in human populations. In this study, we evaluated the abilities of current SARS-CoV-2 variants to bind ACE2 proteins of 14 different animal species. We found that K417N/T, E484K, and N501Y mutations significantly increased the ability of SARS-CoV-2 to infect 293 T cells overexpressing mouse ACE2; variants carrying these mutations (i.e., B. Regarding neutralization, we found that mutations in the RBD enabled to escape from various mAbs; these escape findings were consistent with the activities of SARS-CoV-2 variants carrying the corresponding mutations. Six antibodies showed reduced neutralization against N501Y, including mAbs CB6, which was consistent with previous reports [36][37][38] . Chen et al. also examined a group of monoclonal antibodies, but did not find out significantly reduced neutralization, which may be due to the different groups of monoclonal antibodies that they tested 39 . Variants with more mutations in the RBD region (e.g., B.1.1.7, B.1.351, and P.1) more frequently escaped from mAbs. The results indicate that the neutralization activities of mAbs against epidemic variants should be examined during the development of new therapeutic mAbs; additionally, specific mAbs are presumably more effective against specific variants, implying that cocktail therapy might be appropriate in clinical practice.
Analysis of serum neutralization resistance revealed that most single mutations, including those in the RBD region, could not directly and substantially change the serum neutralization effect on SARS-CoV-2. However, variants carrying the E484K mutation had distinct reductions in neutralization susceptibility. Moreover, some studies have shown that selective pressure from therapeutic mAbs or convalescent serum could induce E484K or E484Q mutations 24,40 . These results indicate the importance of E484 in the viral epitope. Our previous study revealed that the K417N mutation in the RBD region of B.1.351 led to the enhanced convalescence sera neutralization activity 10 . The increased neutralization against K417N by antibodies against RBM was also observed in other group 39 . Here we found that the K417T mutation in the P.1 variant has a similar effect. The increase in K417N sensitivity to serum neutralization has been discussed in our previous paper 10 . K417 forms hydrogen bonds with N370 of S protein resulting in stabilized and closed conformation, which presents a reduced overall area accessible to antibodies or ACE2 41 . The K417N/T mutation increases the probability of conversion to the open conformation, thus exposing epitopes to neutralizing antibodies, which would increase the likelihood of virus neutralization by sera containing polyclonal antibodies. Although variants B.1.351 and P.1 have similar RBD mutation sites, the neutralization resistance of P.1 was less robust than the resistance of B.1.351. Mutations in other regions of the virus (e.g., NTD or S2) may also cause antigenicity changes 42 . Furthermore, the findings that neutralization by full-length trimer spikeimmunized sera was reduced to a lesser extent than RBDimmunized sera suggest that full-length spike protein may induce more complete antibodies. Our preliminary study found that mice immunized with the NTD fragment produced almost no neutralizing antibody, while mice immunized with S2 fragment produced neutralizing antibody. Because the S2 region is a key region that mediates virus fusion, mutation of this region may change viral structure and influence neutralization activity. These results emphasize the need to consider regions other than the RBD in vaccine design efforts. Notably, the RBD is a robust immunogen for neutralizing antibody production, and the corresponding antibody titer is also relatively high. However, modification of E484K in the RBD led to the most marked change in immune escape. Therefore, the RBD alone may not be an ideal immunogen for vaccine development.
Neutralization ability of sera elicited from two vaccines approved in China were tested. E484K was found to be the key mutation that caused the most obvious neutralization insensitivity, while B. In summary, we systematically analyzed the infectivity and host tropism of 10 SARS-CoV-2 variants. The infectivities of most tested SARS-CoV-2 variants were slightly increased compared with the reference strain, and also exhibited considerable enhancements of infectivity in the presence of furin, TMPRSS2, and cathepsin L. Our results demonstrate that the K417N/T, E484K, and N501Y mutations change host tropism, implying possible transmission of SARS-CoV-2 variants in mice. The neutralization activity of mAbs, immunized sera and convalescent sera from different variants infected patients were analyzed against the 10 variants, which would provide insights for the development of therapeutic antibodies and vaccine design.
Monoclonal antibodies. mAbs A261-262, A247 and 76 A (acquired from Professor Linqi Zhang of Tsinghua University, Beijing, China) were produced by RBD-specific single B cells isolated by FACS from SARS-CoV-2 Wuhan-1 strain-infected patients. mAb H00S022 (Sino Biological, Beijing, China) was screened from a phage display scFv library constructed from splenic mRNA of mice immunized with recombinant SARS-CoV-2 RBD of Wuhan-1 strain. mAbs 1F9, 7B8, 4E5, 2H10, 10D12, 10F9 and 9G11 (Biocytogen Pharmaceuticals [Beijing] Co., Bejing, China) were produced by humanized mouse hybridoma against the RBD of SARS-CoV-2 Wuhan-1 strain. X593 (acquired from Prof. Sunney Xie of Peking University, Beijing, China) was identified by high-throughput single-cell RNA and VDJ sequencing of antigenenriched B cells from 60 convalescent patients. mAb CB6(acquired from Prof. Jinghua Yan, Institute of Microbiology, Chinese Academy of Sciences, Beijing, China) was generated by FACS sorting of the membrane B cells from a patient convalescing from COVID-19 and clone the V H and V L genes to human IgG1.
Immunized sera. The study protocol was approved by the Animal Care and Use Committee at the National Institutes for Food and Drug Control (NIFDC). The Animals were handled in accordance with the protocol and guidelines for laboratory animal care and use. Mice were immunized with purified Trimer protein with aluminum adjuvant (20 µg per mouse, once per week for 3 weeks). Serum samples were collected at 4 weeks after the third immunization. Serum samples from 10 mice of each group were pooled to produce combined samples. Each two mice were combined to make one sample. Guinea pigs were immunized with SARS-CoV-2-Spike plasmid at 200 µg per guinea pig or peudotyped virus at 6×10 5 TCID 50 per guinea pig(once every 2 weeks for 6 weeks). Serum samples from five guinea pigs in each group were collected 28 days after the third immunization. Horses were immunized with SARS-CoV-2 RBD protein plus Freund's incomplete adjuvant (once every 10 days for 30 days) at doses of 3 mg, 6 mg and 12 mg. Serum samples were collected at 1 week after the third immunization.
Convalescence sera. Twenty convalescence serum samples were collected from patients with COVID-19, 2-3 months after SARS-CoV-2 infection. Of these 10 samples, 5 were from D614G reference strain-infected patients, four were from B.1.1.7-infected patients, and three were from B.1.351-infected patients. All patients provided written informed consent to participate in the study.
Sera from vaccinated participants. Ten serum samples from individuals immunized with the inactivated vaccine (KCONVAC, Shenzhen Kangtai Biological Products Co.) were collected at 14 days after completion of a three-dose immunization procedure (5 µg/dose; doses given at 0, 28, and 58 days). The study protocol was approved by the Ethics Committee of Jiangsu Provincial Center for Disease Control and Prevention. Written informed consent was signed by all volunteers prior to blood collection.
Ten serum samples were collected from individuals immunized with the adenovirus vaccine (Ad5-nCoV, CanSino Biologics Inc.) at 28 days after a standard immunization procedure (one dose at day 0; 300 µl/dose). The study protocol was approved by the Ethics Committee of Jiangsu Provincial Center for Disease Control and Prevention. Written informed consent was obtained from all volunteers prior to blood collection.
SARS-CoV-2 pseudotyped virus. In accordance with our published method 21 , the spike protein expression plasmid was constructed using GenBank sequence MN908947. The spike gene was codon-optimized and inserted into pcDNA3.1 plasmid between the enzyme digestion sites BamHI and XhoI. Replicationdefective SARS-CoV-2 viral particles were generated by transfection with pcDNA3.1-SARS-CoV-2 and concurrent infection with G*ΔG-VSV (Kerafast). The cell supernatant containing pseudotyped virus was harvest 24 and 48 h later, then aliquoted and stored at −80°C. Site-directed mutagenesis based on circular PCR was used to construct mutants of SARS-CoV-2 pseudotyped virus. DpnI (NEB) was used to digest template DNA. Primers used for mutation are listed in Supplementary Table 1. Pseudotyped SARS-CoV-2 variants were quantified via RT-PCR using VSV P protein as an internal control. Viruses of multiple variants were diluted to the same number of copies before use in experiments.
Infection and neutralization assays. As described in our previous paper 2 , pseudotyped viruses were purified through 25% sucrose cushion by ultracentrifugation. 7 mL pseudovirus was loaded on 3 mL sucrose cushion, and centrifuged at 10,0000 g for 4 hours at 4°C. The RNA of the sediment was extracted and quantified by real-time PCR. The same copy numbers of different pseudotyped SARS-CoV-2 variants were mixed with the indicated cells for infection assays. For neutralization assays, the pseudoviruses were diluted to make sure the control group (without antibody) infected Huh7 cells at the same level (same RLU after 24 h infection, Supplementary Figure 4). mAbs or sera were serially diluted and preincubated with the pseudotyped virus at 37°C for 1 h, then mixed with Huh-7 cells. The cells were then incubated for 20-28 hours at 37°C, in an environment containing 5% CO 2 . Chemiluminescence signals were collected by PerkinElmer Ensight using luciferase substrate (PerkinElmer). Duplicated wells were analyzed for each group. Each experiment was repeated for 2-5 times. Reed-Muench method was used to calculate the half-maximal inhibition dilution (ID 50 ) 43 .
Structural modelling. The spike protein was modelled based on the following Protein Data Bank coordinate sets: RBD-7B8 for 7B8, RBD-Ab5 for 9G11, RBD-Ab1 for 10D 12, 7chh for X593 and 7c01 for CB6. Mutant simulations were carried out in the Mutabind2 web server. MutaBind2 calculates changes in binding affinity upon single or multiple mutations and provides a structural model of the mutated complex. The MutaBind2 model uses molecular mechanics force fields, statistical potentials and fast side-chain optimization algorithms built via a random forest method. Protein-protein interactions were calculated in PDBePISA. Images demonstrating structures were generated in PyMOL.
FACS analysis. Cell surface expression of the spike protein was assessed by flow cytometry. 293 T cells were transfected using the same procedure as for packaging the pseudotyped virus (see above). The medium was removed after transfection for 36 h, following which the cells were digested to produce a single-cell suspension, washed once with PBS, and resuspended with PBS solution containing 1% BSA at 1 × 10 6 cells/tube. MW06 anti-spike antibody (Kohnoor Science & Technology Co., Beijing, China, no neutralization escape for all the tested variants) solution with a final concentration of 1 μg/mL was used as the primary antibody and a FITClabeled goat anti-human IgG (ZF-0308, Zhongshan Jinqiao, Beijing, China) with a final concentration of 6 μg/mL was added as the secondary antibody. The fluorescent signal was examined using a BD FACS CantoTM II Flow Cytometer (BD, Franklin Lakes, NJ).
Statistics and reproducibility. ID 50 was calculated using the Reed-Muench method. Graphical representations were generated using GraphPad Prism 8. Oneway ANOVA and Holm-Sidak's multiple comparisons test were used to identify differences relative to the D614G reference strain. Values are shown as means ± SEMs. For all figures, *P < 0.05, **P < 0.01, ***P < 0.005, and ****P < 0.001.