The SARS-CoV-2 spike S375F mutation characterizes the Omicron BA.1 variant

Summary Recent studies have revealed the unique virological characteristics of Omicron, particularly those of its spike protein, such as less cleavage efficacy in cells, reduced ACE2 binding affinity, and poor fusogenicity. However, it remains unclear which mutation(s) determine these three virological characteristics of Omicron spike. Here, we show that these characteristics of the Omicron spike protein are determined by its receptor-binding domain. Of interest, molecular phylogenetic analysis revealed that acquisition of the spike S375F mutation was closely associated with the explosive spread of Omicron in the human population. We further elucidated that the F375 residue forms an interprotomer pi-pi interaction with the H505 residue of another protomer in the spike trimer, conferring the attenuated cleavage efficiency and fusogenicity of Omicron spike. Our data shed light on the evolutionary events underlying the emergence of Omicron at the molecular level.


INTRODUCTION
Since the emergence of SARS-CoV-2 at the end of 2019, this virus has become spectacularly diverse. In April 2022, the WHO defined two variants of concern, Delta (B.1.617.2 and AY lineages) and Omicron (originally the B.1.1.529 lineage, then reclassified into BA lineages) 1 ; currently, Omicron is the predominant variant spreading worldwide.
Even before detection of the Omicron B.1.1.529 lineage at the end of November 2021 in South Africa, 2 SARS-CoV-2 had become highly diversified from the original lineage, the B lineage, which was isolated in Wuhan, China, on December 24, 2019 (strain Wuhan-Hu-1, GISAID ID: EPI_ISL_402123). 3 Regarding the evolutionary scenario leading to the emergence of Omicron, the B.1 lineage, which had acquired the D614G mutation in the spike (S) protein, [4][5][6][7][8] was first reported on January 24, 2020 (GISAID ID: EPI_ISL_451345). Thereafter, the B.1.1 lineage was first reported in England on February 16, 2020 (GISAID ID: EPI_ISL_466615). The B.1.1 lineage is the common ancestor of both Alpha (B.1.1.7 lineage), a prior variant of concern by March 2022, and Omicron (B.1.1.529 lineage), and the Alpha variant caused a large surge of infection worldwide beginning in the fall of 2020. 9 Omicron was first reported in South Africa on September 30, 2021 (GISAID ID: EPI_ISL_7971523). 2 Soon after the press briefing on Omicron emergence on November 25, 2021, 2 the virological characteristics of Omicron, currently designated BA.1 (i.e., B.1.1.529.1 lineage, hereafter referred to as Omicron in this study), were intensively investigated. For example, Omicron exhibits profound resistance to the humoral immunity induced by vaccination and natural SARS-CoV-2 infection. [10][11][12][13][14][15][16][17][18][19][20][21] In addition, we demonstrated that the Omicron spike (S) protein is less prone to cleavage by furin, a cellular protease, and exhibits poor fusogenicity. 18,22 Moreover, we showed that the binding affinity of the receptor-binding domain (RBD) of Omicron S to human ACE2 is significantly lower than that of ancestral B.1 S. 14,23 However, it remains unclear why Omicron has spread so rapidly worldwide. In particular, although the explosive infectious spread of Omicron in the human population can be mainly characterized by the virological properties of Omicron S, the mutation(s) in Omicron S that are responsible for its virological characteristics, such  Figure 1A) were prepared. The pseudoviruses were inoculated into HOS-ACE2/TMPRSS2 cells at 1 ng HIV-1 p24 antigen, and the percentages of infectivity compared to that of the virus pseudotyped with B.1 S (spike 1) are shown.  Table S1). Consistent with recent studies, 10- 21 Omicron S (spike 2) was highly resistant to the vaccine sera [BNT162b2 ( Figure 2A) and mRNA-1273 ( Figure 2B)] as well as convalescent sera from individuals infected with earlypandemic virus (collected before May 2020) ( Figure 2C) or the Delta variant ( Figure 2D). Pseudoviruses with the Omicron S/B.1 S_RBD (spike 4) and Omicron S/B.1 S_NTD+RBD (spike 5) were significantly more sensitive to vaccine sera (Figures 2A and 2B) and convalescent sera obtained from early-pandemic virus-infected patients than was Omicron S (spike 2) ( Figure 2C). These results suggest that the RBD of Omicron S is closely associated with its pronounced resistance to the antiviral humoral immunity elicited by vaccination or previous SARS-CoV-2 infection. Moreover, we used convalescent sera from hamsters infected with B.1.1 (note that the S gene sequences of B.1 and B.1.1 are identical) and Omicron, as collected in our previous study, 22 for the assay. As shown in Figure 2E, Omicron S (spike 2) was completely resistant to the B.1.1 convalescent sera, whereas it was sensitive to the Omicron convalescent sera. Notably, chimeric Omicron S bearing the B.1 RBD [Omicron S/B.1 S_RBD (spike 4) and Omicron S/B.1 S_NTD+RBD (spike 5)] exhibited the opposite results: these chimeric pseudoviruses were sensitive to the B.1.1 convalescent sera ( Figure 2E) but completely resistant to the Omicron convalescent sera ( Figure 2F). These results further suggest that the Omicron RBD determines its immune resistance and is an immunodominant epitope for inducing humoral immunity. However, we found that Omicron S/B.1 S_NTD+RBD (spike 5) is significantly more sensitive to antisera than is Omicron S/B.1 S_RBD (spike 4) (Figures 2A-2C and 2E). These findings suggest that mutations in the NTD of Omicron S are also partly associated with the immune resistance of Omicron S.

The S S375F mutation increases binding affinity to human ACE2
Twelve substitutions are uniquely present in the RBD (residues 319-541) of Omicron S; another 3 substitutions (K417N, T478K and N501Y) are common among the other variants ( Figure 3A). 18 To determine the residue(s) responsible for the virological phenotype of Omicron, particularly the reduced binding affinity of the Omicron S RBD to human ACE2, 14,23 we prepared a series of B.1 S RBD point mutants that bear the respective mutations of Omicron and conducted screening experiments based on a yeast surface display assay. 14,23,24,26,27 As shown in Figure 3B (left panel), compared to the RBD of parental (i.e., B lineage-based) S, the K D values of the G339D, N440K and S477N mutants were significantly decreased, whereas those of the S375F, S371L/S373P/S375F, G496S and Y505H mutants were significantly increased.  (C) and quantified band intensity (the ratio of S2 to the full-length S plus S2 proteins for ''cell''; the ratio of S2 to HIV-1 p24 for ''supernatant'') (D) are shown. M, mock (empty vector-transfected). Uncropped blots are shown in Figure S4. Omicron emergence is closely associated with acquisition of the S S375F mutation The S375F, G496S and Y505H mutations in the S protein are almost exclusive to Omicron variants (Table S2).
To infer the evolutionary sequence of the emergence of these mutations in the Omicron lineage, we generated a time tree of 44 Omicron genomes detected in 2021 (for more detail, see STAR Methods) ( Figures 3C  and S2). The G496S and Y505H mutations were detected in all sequences used in this analysis, suggesting that these two mutations were present in the common ancestor of all Omicron variants reported thus far. In contrast, the S371L, S373P and S375F mutations are not present in the older Omicron sequences (shown in black in Figures 3C and S2). Although the emergence times of S371L and S373P cannot be estimated independently, our analysis assumed that the S371L and S373P mutations were first acquired between Node 1 [95% highest posterior density (HPD): August 26, 2021 to October 10, 2021] and Node 2 (95% HPD: September 24, 2021 to October 21, 2021) in Figure 3C, as based on the estimated time to the most recent common ancestor (tMRCA). The S375F mutation emerged thereafter, between Node 2 and Node 3 (95% HPD: October 5, 2021 to October 23, 2021) ( Figure 3C). Of interest, the Bayesian skyline plot of the 44 Omicron genomes suggested that the effective population size of Omicron increased around the time of S375F substitution acquisition ( Figure 3D). These data suggest that the emergence of the S375F mutation might have been a crucial event triggering the massive spread of Omicron variants in the human population.
To verify the possibility that the S375F mutation is crucial for the phenotype of Omicron, we performed yeast binding assays using the RBD of Omicron S. As depicted in Figure 3B (right panel), the F375S and L371S/P373S/F375S mutations in the RBD of Omicron S significantly increased binding affinity to human ACE2. Overall, these observations suggest that the three substitutions at positions 371, 373 and 375, particularly the S375F substitution, determine the reduced binding affinity of the Omicron S RBD to human ACE2.
The S S375F mutation determines the S cleavage efficacy, fusogenicity, and ACE2 binding affinity of the Omicron variant To investigate the impact of the S375F mutation, we prepared pseudoviruses with a series of Omicron S-based mutations ( Figure 4A). In the yeast surface display assay ( Figure 3B), the assay based on Omicron S showed that pseudovirus infectivity was clearly increased by the Omicron S F375S mutation (spikes 9 and 11-13 in Figure 4A) ( Figure 4B, top). Western blot analysis showed that the S1/S2 cleavage efficacy and level of S2 in virions were rescued by the F375S mutation ( Figures 4C and 4D, top). Similar to the results illustrated in Figures 1C and 1D, the mutated S proteins that were efficiently cleaved in cells (e.g., spikes 9 and 11-13) were also efficiently incorporated into the viral particles released ( Figures 4C and  4D). These results indicate that the level of virion-incorporated S2 is modulated by the S cleavage efficacy in producer cells and that pseudovirus infectivity can be an indicator of the level of S protein cleavage in Figure 2. Immune resistance conferred by the Omicron RBD Neutralization assays were performed with pseudoviruses harboring a series of S protein sequences (summarized in Figure 1A). The numbers are identical to those in Figure 1A.  Table S1. Each serum sample was analyzed in triplicate to determine the 50% neutralization titer (NT50). Each dot represents one NT50 value, and the geometric mean and 95% CI are shown. The numbers indicate the fold changes of resistance versus each antigenic variant. Horizontal gray lines indicate the detection limit of each assay (120 for A and C-F; 40 for B). Statistically significant differences between spikes 4 and 5 were determined by a two-sided Wilcoxon signed-rank test. See also Table S1.   Figure 4F, bottom). These results suggest that the S375F mutation in Omicron S is responsible for the decreased S cleavage efficacy in producer cells and the attenuated fusogenicity observed. However, the S371L/S373P/S375F mutations did not affect sensitivity to the antiviral humoral immunity elicited by vaccination and infection ( Figure S3), suggesting that the S375F mutation is not associated with the immune resistant phenotype of Omicron.  To further assess the impact of the S375F mutation, we generated two additional recombinant chimeric SARS-CoV-2 strains, B.1 S S375F-GFP (virus VI) and Omicron S F375S-GFP (virus VII) ( Figure 5A). Although the mutation at position 375 of the S protein did not affect the viral RNA load in the culture supernatant of infected VeroE6/TMPRSS2 cells ( Figure 5B), the GFP intensity in infected VeroE6/TMPRSS2 cells was significantly altered by this mutation: the S375F mutation in the B.1 S backbone decreased the GFP intensity, whereas the F375S mutation in the Omicron S backbone increased the intensity (Figures 5C and  S1). In addition, quantitative fluorescence microscopy showed that the GFP-positive area of B.1 S S375F-GFP (virus VI) was significantly lower than that of parental B.1 S-GFP (virus I); however, that of Omicron S F375S-GFP (virus VII) was significantly higher than that of parental Omicron S-GFP (virus II) ( Figure 5D). Moreover, plaque assays showed that the plaques formed by infection with B.1 S S375F-GFP (virus VI) were significantly smaller than those formed by B.1 S-GFP (virus I); conversely, plaque size was increased by the presence of the F375S mutation in Omicron S ( Figure 5E). Altogether, these results suggest that the S375F mutation in the Omicron S protein determines the major virological characteristics (i.e., decreased S1/S2 cleavage efficacy, decreased fusogenicity, and decreased ACE2 binding affinity) of Omicron. Here, we experimentally demonstrate that the S375F mutation attenuates the cleavage efficacy and fusogenicity of Omicron S (Figures 4 and 5). In addition, molecular phylogenetic analysis suggested that the emergence of this mutation was closely associated with the explosive growth of Omicron in the human population ( Figures 3C and 3D). Nevertheless, it remains unclear how the S375F mutation contributes to the decrease in cleavage efficacy and fusogenicity of Omicron S at the molecular level. We addressed this question using a structural biology approach. As shown in Figure 6A (top), we predicted that the F375 residue in a fully closed Omicron S trimer could form a pi-pi interaction, a sort of dispersion via van der Waals forces between aromatic residues, 30 with the H505 residue in another S protein of the same trimer. Importantly, the cryo-EM structure of the Omicron BA.1 S protein has been iScience Article determined. 31 The result demonstrated that the interprotomer interaction mediated by the F375 and H505 residues of Omicron S causes the S trimer conformation to be more rigid and leads to less cleavage efficacy, supporting our prediction. Because residue 375 in the B.1.1 S protein is a serine, the pi-pi interaction cannot be formed ( Figure 6A, bottom). To address the hypothesis that the F375-H505-mediated interprotomer pi-pi interaction contributes to the decreased cleavage efficacy and fusogenicity of Omicron S, we prepared the Omicron S H505A mutant, in which an aromatic side chain at position 505 is disrupted. Western blot analysis showed that the cleavage efficacy of Omicron S was increased by the insertion of the H505A mutation ( Figure 6B). To further test this possibility, the residues at position 375 of B.1 S were substituted with amino acids bearing aromatic side chains (i.e., F, Y and H). Similar to the S375F mutant, the B.1 S mutants bearing the S375Y or S375H mutation showed decreased S protein cleavage efficacy ( Figure 6C). These results further suggest that the interprotomer pi-pi interaction is formed between Y505 and S375F/Y/H. Moreover, insertion of the Y505A mutation in B.1 S bearing the S375F/Y/H mutation (i.e., disruption of the aromatic residue at position 505) rescued the S cleavage efficacy ( Figure 6C).
Finally, we verified the impact of the interprotein pi-pi interaction on S-mediated fusogenicity. The Omicron S F375S mutant exhibited decreased surface expression, but the H505A mutation did not (Figure 6D). In the case of the B.1 S-based mutants, the Y505A mutation decreased surface expression levels when the S375F/Y mutations were also present ( Figure 6E). Corresponding to western blot results (Figure 6B), disruption of the pi-pi interaction by F375S and H505A in Omicron S significantly increased fusion activity ( Figure 6F). Moreover, in the case of the B.1 S-based mutant, substitution of residue 375 with an aromatic residue (F, Y or H) significantly reduced fusion activity ( Figure 6G). However, when the Y505A substitution was present in the S375F/Y/H mutants, disrupting the aromatic residue at position 505, fusion activity was significantly increased ( Figure 6G). Altogether, our results suggest that the interprotomer pi-pi interaction mediated by the aromatic residues at positions 375 and 505 of the S protein contributes to the decreased cleavage efficacy and fusogenicity of Omicron S.

DISCUSSION
In the present study, we performed multiscale investigations to unveil the virological characteristics of the S protein of the SARS-CoV-2 Omicron variant, including (1) profound immune resistance, (2) decreased cleavage efficacy in cells, (3) poor fusogenicity, and (4) reduced ACE2 binding affinity. By using pseudoviruses, a yeast surface display system and the chimeric recombinant SARS-CoV-2 generated by reverse genetics, we showed that the RBD of Omicron S is responsible for these four virological features of this variant. In particular, the S375F mutation in the RBD of Omicron S is one of the most critical mutations that determine three of the four major virological properties of Omicron: decreased affinity to ACE2, attenuated efficacy of S cleavage, and reduced fusogenicity. Moreover, molecular phylogenetic analysis provided evidence suggesting that the acquisition of the S375F mutation was closely related to the onset of the explosive spread of Omicron in the human population ( Figure 3C). Furthermore, experiments based on structural biology revealed that the pi-pi interaction mediated by residues F375 and H505 is responsible for the observed decreased cleavage efficacy in cells and fusogenicity.
We and others demonstrate that the Omicron S RBD shows reduced binding affinity to human ACE2. 14,23 In this study, our mutagenesis experiment revealed that the S375F, G496S and Y505H substitutions are responsible for this reduced binding affinity of the Omicron S RBD to human ACE2 ( Figure 3B). Considering the importance of ACE2 binding in viral replication, it is intriguing how the Omicron variant acquired high transmissibility with decreased ACE2 binding. It may be reasonable to speculate that evasion from the preexisting immunity induced by previous infection or vaccination was the priority for the evolution of Omicron.
We revealed that the nascent pi-pi interaction of the Omicron S trimer is established by the F375 and H505 residues and characterizes Omicron S. After the initial submission of this study, structural analysis by cryoelectron microscopy (cryo-EM) showed that the interprotomer interaction mediated by the F375 and H505 residues of Omicron S firms the S trimer conformation and leads to reduced cleavage efficacy, 31 supporting our experimental results in this study. Because the Y505H mutation was already present in the ancestral Omicron sequences, our results suggest that acquisition of the S375F mutation during the evolution of Omicron resulted in attenuated S cleavage efficacy and fusogenicity in SARS-CoV-2 S protein, which led to the explosive spread of Omicron in the human population. The S375F mutation is highly conserved in the Omicron lineage and has not been detected in the other SARS-CoV-2 variants. However, our data ll OPEN ACCESS iScience 25, 105720, December 22, 2022 iScience Article suggest that substitution of residues possessing an aromatic ring, such as phenylalanine, tyrosine and histidine, at residue 375 may confer Omicron-like properties. Therefore, the emergence of SARS-CoV-2 variants bearing such substitutions at residue 375 should be considered a potential risk for health of the global population.
Our previous studies suggested a close association between viral fusogenicity and pathogenicity. 22,23,32 For example, Omicron S is less susceptible to cleavage than parental B.1.1 S harboring the D614G mutation. 18,22 This decreased S1/S2 cleavage is associated with a reduction in the fusogenicity of Omicron S and attenuates the pathogenicity of Omicron variant. 22 Here, we demonstrate that S cleavage efficacy and fusogenicity are determined by S375F mutation in the RBD of Omicron S (Figures 1C, 1F, 4C and 4F). Therefore, it is likely that acquisition of the S375F mutation in the S protein may, at least partially, contribute to the attenuated pathogenicity of the Omicron variant. Further investigation will be required to determine whether the S375F mutation is critical for viral pathogenicity because this mutation is present among more than 30 changes.
Here, we show the importance of the S375F mutation to the major virological properties of Omicron S, particularly its decreased cleavage efficacy, poor fusogenicity, and reduced ACE2 binding affinity. However, the following issues remain to be fully elucidated. First, although we showed that the S375F mutation determines a part of the virological features of Omicron S, it remains unclear which mutations in Omicron S determine its pronounced immune resistance. We showed that the RBD of Omicron S is closely associated with its resistance to the humoral immunity induced by vaccination and natural SARS-CoV-2 infection (Figure 2), yet there are dozen substitutions in the Omicron S RBD ( Figure 3B). Therefore, it would be reasonable to assume that multiple substitutions in the RBD cooperatively contribute to the profound immune resistance of Omicron S. Second, in addition to the Omicron BA.1 variant that we focused on this study, a variety of Omicron subvariants, such as BA.2 and BA.5, have emerged, and these subvariants also bear the S375F mutation. However, we have recently shown that the fusogenicity of BA.2 S is significantly higher than that of BA.1 S. 23 Together with the results of this study, these observations suggest that BA.2 S has acquired certain compensatory mutation(s) that increase fusion efficacy. Further investigations will be needed to unveil the full evolutionary history of the Omicron lineage. Furthermore, the question of why acquisition of the S375F mutation caused explosive spread despite reduced infectivity in tissue culture, S cleavage efficacy and fusogenicity also needs to be elucidated in detail by further studies.
In summary, our multiscale investigations reveal that the major virological characteristics of Omicron S, namely, attenuated S cleavage efficacy, attenuated fusogenicity, and reduced ACE2 binding affinity, are determined by one specific mutation, S375F, in the RBD. Assays based on structural biology revealed that the pi-pi interaction mediated by residues F375 and H505 is responsible for the observed attenuated S cleavage efficacy and fusogenicity. Furthermore, the molecular phylogenetic analysis suggested that acquisition of the S375F mutation was closely associated with the massive spread of Omicron in the human population. Altogether, our results suggest that acquisition of the S375F mutation was a crucial event for the emergence of a highly transmissible SARS-CoV-2 variant, Omicron.

Limitations of the study
In this study, we used lentivirus-based pseudovirus to examine impact mutations on S packaging; thus, the phenotype observed in this study may differ from the S2 incorporation occurring in authentic SARS-CoV-2 virions. However, our recent finding with authentic SARS-CoV-2 virions demonstrated that the S2 incorporation pattern of BA.1 was significantly lower than that of the Delta variant, 18 reproducing the low S2 incorporation of BA.1 S observed in this study. Therefore, these data suggest that lentivirus-based pseudovirus can be used to examine the S2 incorporation pattern of authentic SARS-CoV-2 virions. Further investigations, such as the identification of S protein incorporation in lentivirus-based pseudoviruses, are required to solve this issue.
In addition, it is reported that the Omicron variants deposited early contain artifactual reversions possibly derived from contamination of non-Omicron (mainly Delta) variants due to the low affinity of primers for sequencing. 33,34 Although such low-quality Omicron genomes were removed in this analysis (see STAR Methods), we cannot exclude the possibility that some genomes may contain artifactual reversions, which would affect the phylogenetic results.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:   GISAID IDs used in this study was available at the following https://doi.org/10.55876/gis8.221004su. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Ethics statement

Molecular phylogenetic analyses
The SARS-CoV-2 genomes and annotation information used in this study were downloaded from the GISAID EpiCoV database (https://www.gisaid.org/) on January 8, 2022 ( For each sequence, we counted the number of undetermined nucleotides (such as N, Y, W) for whole genomes as well as S genes and obtained 40,739 sequences with fewer than 1,000 undetermined nucleotides in the genome and fewer than 10 undetermined nucleotides in the S-coding region. We then obtained BA.1 variant genomes that met the following criteria: 1) genomes were isolated from humans; 2) genomes did not contain any undetermined nucleotides in genomic regions corresponding to amino acid positions 371-375 in the S protein; 3) genomes were sampled from September 2021 to November 2021; and 4) genomes did not contain any of the 3 amino acid replacements in the S protein.
We then selected 12 genomes and randomly selected 100 genomes that met criteria 1 and 2. We then removed Omicron genomes containing recombination sites using RDP4 v4.101 44 because such genomes may contain artifactual reversions possibly derived from contamination of non-Omicron (mainly Delta) variants due to the low affinity of primers. 33,34 We also checked the sequences manually, and 44 Omicron genomes were obtained.
The 44 Omicron genomes with two outgroup genomes EPI_ISL_402125 (strain Wuhan-Hu-1, B lineage) and EPI_ISL_406862 (B.1 lineage; one of the earliest sequences carrying the S D614G mutation) were aligned using FFT-NS-1 in MAFFT suite v7.407. 45 We then deleted gapped regions in the 5' and -3 0 regions. BEAST v1.10.4 46 was used to construct a timetree under an exponential growth coalescent model using a strict molecular clock. The GTR model with the four categories of discrete gamma rate variation was used as a nucleotide substitution model. 48 iScience Article iScience Article A linker fragment encoding hepatitis delta virus ribozyme, bovine growth hormone poly A signal and cytomegalovirus promoter was also prepared by PCR. The corresponding SARS-CoV-2 genomic region and the PCR templates and primers used for this procedure are summarized in Table S3. The 10 obtained DNA fragments were mixed and used for CPER. 25 To prepare GFP-expressing replication-competent recombinant SARS-CoV-2, we used fragment 9, in which the GFP gene was inserted in the ORF7a frame, instead of the authentic F9 fragment (Table S3). 25 To generate chimeric recombinant SARS-CoV-2 ( Figures 1G and 5A), mutations were inserted in fragment 8 by site-directed overlap extension PCR or the GENEART site-directed mutagenesis system (Thermo Fisher Scientific, Cat# A13312) according to the manufacturer's protocol with the primers listed in Table S3. To produce chimeric recombinant SARS-CoV-2, the CPER products were transfected into HEK293-C34 cells using TransIT-LT1 (Takara, Cat# MIR2300) according to the manufacturer's protocol. At 1 d posttransfection, the culture medium was replaced with Dulbecco's modified Eagle's medium (high glucose) containing 2% FCS, 1% PS and doxycycline (1 mg/mL; Takara, Cat# 1311N). At 7 d posttransfection, the culture medium was harvested and centrifuged, and the supernatants were collected as the seed virus. To remove the CPER products (i.e., SARS-CoV-2-related DNA), 1 mL of the seed virus was treated with 2 mL TURBO DNase (Thermo Fisher Scientific, Cat# AM2238) and incubated at 37 C for 1 h. Complete removal of the CPER products (i.e., SARS-CoV-2-related DNA) from the seed virus was verified by PCR. The working virus stock was prepared from the seed virus as described below (see ''SARS-CoV-2 preparation and titration'' section).

SARS-CoV-2 preparation and titration
To prepare the working virus stocks of chimeric recombinant SARS-CoV-2, 25,24 20 mL of the seed virus was inoculated into VeroE6/TMPRSS2 cells (5,000,000 cells in a T-75 flask). One hour post infection (h.p.i.), the culture medium was replaced with DMEM (low glucose) (Wako, Cat# 041-29775) containing 2% FBS and 1% PS. At 3 d.p.i., the culture medium was harvested and centrifuged, and the supernatants were collected as the working virus stock.
The titer of the prepared working virus was measured as the 50% tissue culture infectious dose (TCID 50 ). Briefly, one day before infection, VeroE6/TMPRSS2 cells (10,000 cells) were seeded into a 96-well plate. Serially diluted virus stocks were inoculated into the cells and incubated at 37 C for 4 days. The cells were observed under microscopy to judge the CPE appearance. The value of TCID 50 /mL was calculated with the Reed-Muench method. 53 To verify the sequence of chimeric recombinant SARS-CoV-2, viral RNA was extracted from the working viruses using a QIAamp viral RNA mini kit (Qiagen, Cat# 52906) and viral genome sequence was analyzed as described above (see "viral genome sequencing" section above). In brief, the viral sequences of GFP-encoding recombinant SARS-CoV-2 (strain WK-521; GISIAD ID: EPI_ISL_408667) 25,37 that harbor the S genes of respective variants were used for the reference. Information on the unexpected mutations detected is summarized in Table S4, and the raw data are deposited in DDBJ Sequence Read Archive (accession number: PRJDB13805).

SARS-CoV-2 infection
One day before infection, VeroE6/TMPRSS2 cells (10,000 cells) were seeded into a 96-well plate. SARS-CoV-2 (100 TCID 50 , m.o.i. 0.01) was inoculated and incubated at 37 C for 1 hour. The infected cells were washed, and 180 mL of culture medium was added. The culture supernatant (10 mL) was harvested at the indicated timepoints and used for RT-qPCR to quantify the viral RNA copy number (see ''RT-qPCR'' section below).
The binding affinities of SARS-CoV-2 S RBDs to human ACE2 were determined by flow cytometry titration experiments. The CF640R-labelled ACE2 at 12-14 different concentrations (200 nM to 13 pM in PBS supplemented with bovine serum albumin at 1 mg/mL) per measurement were incubated with expressed yeast aliquots and 10 nM bilirubin (Sigma-Aldrich, Cat# 14370-1G) and analyzed by using FACS S3e Cell Sorter device (Bio-Rad). The background binding subtracted fluorescent signal was fitted to a standard noncooperative Hill equation by nonlinear least-squares regression using Python v3.7 (https://www. python.org) as previously described. 26
In the time-course experiments ( Figures 1F, 1H, 1I, 4F, 5B, 5C, 6F, and 6G), a multiple regression analysis including experimental conditions (i.e., the types of infected viruses) as explanatory variables and timepoints as qualitative control variables was performed to evaluate the difference between experimental conditions thorough all timepoints. p value was calculated by a two-sided Wald test. Subsequently, familywise error rates (FWERs) were calculated by the Holm method. These analyses were performed in R v4.1.2 (https://www.r-project.org/).
In Figures 1J, 1I, 5C, 5D and S1B, assays were performed in triplicate. Photographs shown are the representatives of >18 fields of view taken for each sample.

ll
OPEN ACCESS