Enhanced fusogenicity and pathogenicity of SARS-CoV-2 Delta P681R mutation

During the current coronavirus disease 2019 (COVID-19) pandemic, a variety of mutations have accumulated in the viral genome of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and, at the time of writing, four variants of concern are considered to be potentially hazardous to human society1. The recently emerged B.1.617.2/Delta variant of concern is closely associated with the COVID-19 surge that occurred in India in the spring of 2021 (ref. 2). However, the virological properties of B.1.617.2/Delta remain unclear. Here we show that the B.1.617.2/Delta variant is highly fusogenic and notably more pathogenic than prototypic SARS-CoV-2 in infected hamsters. The P681R mutation in the spike protein, which is highly conserved in this lineage, facilitates cleavage of the spike protein and enhances viral fusogenicity. Moreover, we demonstrate that the P681R-bearing virus exhibits higher pathogenicity compared with its parental virus. Our data suggest that the P681R mutation is a hallmark of the virological phenotype of the B.1.617.2/Delta variant and is associated with enhanced pathogenicity.

During the current coronavirus disease 2019 (COVID-19) pandemic, a variety of mutations have accumulated in the viral genome of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and, at the time of writing, four variants of concern are considered to be potentially hazardous to human society 1 . The recently emerged B.1.617.2/Delta variant of concern is closely associated with the COVID-19 surge that occurred in India in the spring of 2021 (ref. 2 ). However, the virological properties of B.1.617.2/Delta remain unclear. Here we show that the B.1.617.2/Delta variant is highly fusogenic and notably more pathogenic than prototypic SARS-CoV-2 in infected hamsters. The P681R mutation in the spike protein, which is highly conserved in this lineage, facilitates cleavage of the spike protein and enhances viral fusogenicity. Moreover, we demonstrate that the P681R-bearing virus exhibits higher pathogenicity compared with its parental virus. Our data suggest that the P681R mutation is a hallmark of the virological phenotype of the B.1.617.2/Delta variant and is associated with enhanced pathogenicity.
The B.1.617 lineage emerged in India at the end of 2020 and is thought to have been a major driver of the massive COVID- 19 11,12 . Importantly, early evidence has suggested that infection with B.1.617.2/Delta may carry an increased risk of hospitalization compared with infection with B.1. 1.7 (refs. 13-15 ). However, the virological features of this newly emerging VOC, particularly its infectivity and pathogenicity, remain unclear. In this study, we demonstrate that B.1.617.2/Delta is more pathogenic than the prototypic SARS-CoV-2 in a Syrian hamster model. We also show that the P681R mutation in the S protein is a hallmark mutation of this lineage. The P681R mutation enhances the

The pathogenicity of the Delta variant
To investigate the pathogenicity of the B.1.617.2/Delta variant, we conducted hamster infection experiments using the B.1.617.2/Delta isolate and the B.1.1 isolate. The viral RNA loads in the oral swabs of B.1.617.2/ Delta-infected hamsters were comparable with those of B.1.1-infected hamsters across timepoints on average (P = 0.057, multiple regression) (Fig. 2c). Infected hamsters of both groups lost significant body weight beginning at 2 days post-infection (d.p.i.), and the weight loss of B.1.617.2/Delta-infected hamsters was significantly greater than that of B.1.1-infected hamsters across timepoints on average (P = 0.0082, multiple regression) (Fig. 2d). The peak weight loss was 16% after infection   (Fig. 2d).
In the lungs of infected hamsters of both groups, bronchitis with focal inflammatory cell infiltration around bronchi/bronchioles was observed at 1 d.p.i. followed by haemorrhage or congestion at 3 d.p.i. (Fig. 2e, f and Extended Data Fig. 5a). Crushed nuclear debris, suggesting the damage of the alveolar pneumocytes with macrophage infiltration, was observed from 3 to 5 d.p.i., and the area of inflammatory cell infiltration was expanded with time ( Fig. 2e, f and Extended Data Fig. 5a). In both cases, type II pneumocytes with an increased nuclearcytoplasmic ratio appeared at 5 d.p.i. Notably, in the lungs of B.1.617.2/ Delta-infected hamsters, prominently enlarged cells with large nuclei (greater than 10 μm in diameter) were recognized, suggesting large type II pneumocytes that reflect the severity of pneumonia at 5 d.p.i. (Fig. 2f, g and Extended Data Fig. 5b). Immunohistochemistry analysis of viral nucleocapsid (N) protein demonstrated that N proteins were detected in the bronchial epithelial cells with a small fraction of alveolar staining in both infection cases at 1 d.p.i. (Extended Data Fig. 5c

The effect of the P681R mutation on viral fusion
The P681R mutation in the S protein is a unique feature of the B.1.617 lineage, including the B.1.617.2/Delta variant (Fig. 1d). As the P681R mutation is located in proximity to the furin cleavage site (FCS; residues RRAR positioned at 682-685) of the SARS-CoV-2 S protein 16 , we hypothesized that the P681R mutation is responsible for the promotion of cell-cell fusion, leading to the formation of larger syncytia. To address this possibility, we generated a P681R-bearing artificial virus by reverse genetics (Extended Data Fig. 2b) and performed further virological experiments. Although the amounts of viral RNA in the culture supernatants of the D614G/P681R-infected Vero and VeroE6/TMPRSS2 cells were significantly lower compared with those of the D614G-infected cells at some timepoints, the growth of these two viruses was relatively comparable (Fig. 3a). However, the floating syncytia ( To clearly observe syncytium formation, we further generated GFP-expressing replication-competent D614G and D614G/P681R viruses. The levels of viral RNA in the supernatant and proportion of GFP-positive cells were similar in Vero, VeroE6/TMPRSS2 and Calu-3 cells (Extended Data Fig. 6). However, at 24 h.p.i., significantly larger GFP-positive adherent syncytia were observed in VeroE6/TMPRSS2 cells infected with the GFP-expressing D614G/P681R virus (Fig. 3c). Moreover, the GFP-positive floating syncytia at 72 h.p.i. in VeroE6/ TMPRSS2 cells infected with GFP-expressing D614G/P681R virus were significantly larger (2.4-fold) in size compared with those of VeroE6/ TMPRSS2 cells infected with GFP-expressing D614G virus (Extended Data Fig. 7a). Moreover, GFP-positive syncytia were observed in D614G/ P681R-infected Calu-3 cells but not in D614G-infected Calu-3 cells at 72 h.p.i. (Extended Data Fig. 6c). These results suggest that the features of the B.1.617.2/Delta virus observed in in vitro cell culture experiments, particularly the formation of larger syncytia, are well reproduced by the insertion of the P681R mutation. To further investigate the effect of the P681R mutation, the GFP-expressing viruses were inoculated into human primary nasal epithelial culture. Notably, the viral RNA levels of D614G/P681R virus on the apical side of culture at 2 and 3 d.p.i. were 12.3-fold and 7.0-fold higher, respectively, than those of parental D614G virus with statistical significance, and the rapid growth of D614G/P681R virus was supported by the observation of GFP expression (Extended Data Fig. 7b). Although the viral RNA levels of D614G/P681R virus gradually decreased after 5 d.p.i., plaque-like spots were observed after 7 d.p.i., and the sizes of these plaque-like spots in the culture of D614G/P681R infection were significantly larger than the plaque-like spots in the culture of parental D614G virus infection (Extended Data  Fig. 7b). These data suggest that the P681R mutation accelerates viral replication in human primary nasal epithelial culture and produces large plaque-like spots, which could be formed by cell-to-cell infection as the case of plaque formation.
To directly investigate the effect of the P681R mutation on the cleavage of the SARS-CoV-2 S protein, we prepared an HIV-1-based pseudovirus carrying the P681R mutation. Western blot analysis of the prepared pseudoviruses showed that the level of the cleaved S2 subunit was significantly increased in the presence of the P681R mutation (Extended Data Fig. 8a), suggesting that the P681R mutation Left, representative blots of SARS-CoV-2 full-length S and cleaved S2 proteins as well as ACTB as an internal control. Assays were performed in triplicate. Data are mean ± s.d. Right, the ratio of S2 to the full-length S plus S2 proteins in the S-expressing cells. e, SARS-CoV-2 S-based fusion assay. Effector cells (S-expressing cells) and target cells (ACE2-expressing cells or ACE2/ TMPRSS2-expressing cells) were prepared, and the fusion activity was measured as described in the Methods. Assays were performed in quadruplicate, and fusion activity (arbitrary units) is shown. Data are mean ± s.d. Statistically significant differences versus D614G (*P < 0.05) and uninfected culture ( # P < 0.05) were determined using two-sided unpaired Student's t-tests (a, d) or Mann-Whitney U-tests (b, c).

Article
facilitates furin-mediated cleavage of the SARS-CoV-2 S protein. We next performed a single-round pseudovirus infection assay using target HOS-ACE2 cells with or without TMPRSS2 expression. The infectivity of both the D614G and D614G/P681R pseudoviruses was increased approximately tenfold by the expression of TMPRSS2 in the target cells (Extended Data Fig. 8b). However, the relative infectivity of the D614G and D614G/P681R pseudoviruses was not altered by TMPRSS2 expression (Extended Data Fig. 8b). These data suggest that the P681R mutation does not affect the infectivity of the viral particles. We next addressed the effect of the P681R mutation on viral fusogenicity by a cell-based fusion assay. In the effector cells (that is, S-expressing cells), although the expression level of the D614G/P681R S protein was comparable to that of the D614G S protein, the level of the cleaved S2 subunit was significantly higher for the D614G/P681R mutant than for the D614G mutant (Fig. 3d). Consistent with the results of the pseudovirus assay (Extended Data Fig. 8a), these results suggest that the P681R mutation facilitates S cleavage. Flow cytometry analysis showed that the surface expression level of D614G/P681R S was significantly lower than that of D614G S (Extended Data Fig. 8c). Nevertheless, the cell-based fusion assay using the target cells without TMPRSS2 demonstrated that D614G/P681R S is 2.1-fold more fusogenic than D614G S-a statistically significant difference (P = 0.0002, Welch's t-test) (Fig. 3e). Moreover, a mathematical modelling analysis of the fusion assay data showed that the initial fusion velocity of D614G/ P681R S (0.83 ± 0.03 per hour) was significantly faster (2.8-fold) than that of D614G S (0.30 ± 0.03 per hour; P = 4.0 × 10 −6 , Welch's t-test) (Extended Data Fig. 8d, e). These data suggest that the P681R mutation enhances and accelerates SARS-CoV-2 S-mediated fusion. Furthermore, when we used targeted cells expressing TMPRSS2, both the fusion efficacy (about 1.2-fold) and initial fusion velocity (about 2.0-fold) were increased in both the D614G and D614G/P681R S proteins (Extended Data Fig. 8d, e). These results suggest that TMPRSS2 facilitates the fusion mediated by SARS-CoV-2 S and human ACE2 and that this TMPRSS2-dependent acceleration and promotion of viral fusion is not specific for the P681R mutant.

Neutralization of the P681R mutant
Resistance to neutralizing antibodies in the sera of COVID-19 convalescent individuals and vaccinated individuals is a hallmark of VOCs 17,18 , and it has recently been shown that the B.1.617.2/Delta variant is relatively resistant to vaccine-induced neutralization 19,20 . To determine whether the P681R mutation contributes to this virological phenotype, we performed a neutralization assay. The D614G/P681R pseudovirus was partially resistant (1.2-1.5-fold) to three monoclonal antibodies targeting the receptor-binding domain of the SARS-CoV-2 S protein (Extended Data Fig. 9a). Furthermore, neutralization experiments using 19 serum samples collected after two rounds of BNT162b2 vaccination showed that the D614G/P681R pseudovirus was significantly more resistant than the D614G pseudovirus to vaccine-induced neutralizing antibodies (P < 0.0001, Wilcoxon matched-pairs signed-rank test) (Extended Data Fig. 9b, c). These results suggest that the P681R-bearing pseudovirus is relatively resistant to neutralizing antibodies.

Pathogenicity of the P681R mutant
To assess the effect of the P681R mutation on viral replication and the pathogenicity of SARS-CoV-2, we intranasally infected Syrian hamsters with the D614G and D614G/P681R viruses. The D614G-infected hamsters exhibited no weight loss, although a slight decrease in body weight by 7 d.p.i. was observed for one of the hamsters (5.0%) (Fig. 4a). By contrast, all of the hamsters infected with the D614G/P681R virus experienced gradual body weight loss, and the hamsters showed a weight loss of 4.7-6.9% at 7 d.p.i., significantly greater compared with the weight loss of hamsters that were infected with the D614G virus (P = 0.011) (Fig. 4a). The weight loss of D614G/P681R-infected hamsters was significantly greater compared with that of D614G-infected hamsters on average across all timepoints (P = 0.00015, multiple regression) (Fig. 4a). We also assessed pulmonary function in infected hamsters by using a whole-body plethysmography system to measure enhanced pause (PenH), which is a surrogate marker of bronchoconstriction or airway obstruction. Infected hamsters of both groups showed increases in the lung PenH value, but the PenH values of D614G/P681R-infected hamsters were significantly higher than those of the D614G-infected hamsters on average across all timepoints (P = 0.038, multiple regression) (Fig. 4b) and nasal turbinates of the infected hamsters, and no significant difference in viral replication was observed between the two groups ( Fig. 4c). At 7 d.p.i., no differences in viral titres in the nasal turbinates were found between the two groups; however, the lung titres in the D614G/P681R-infected group were significantly higher than those in the D614G-infected group (P = 0.0013) (Fig. 4c).
Histopathological examination revealed cell infiltration in and around the bronchi/bronchioles at 3 d.p.i. in both groups, but solid bronchioloalveolar epithelial hyperplasia including type II pneumocytes was prominent at 7 d.p.i. in the D614G/P681R-infected hamsters (Fig. 4d). Microcomputed tomography (microCT) analysis revealed lung abnormalities in all of the infected hamsters on 7 d.p.i. that were consistent with commonly reported imaging features of COVID-19 pneumonia 21 (Extended Data Fig. 10a). Lung abnormalities included multifocal nodular ground glass opacity with a peripheral, bilateral, multilobar, peribronchial distribution with regions of lung consolidation. The CT severity scores of the D614G-infected and D614G/ P681R-infected hamsters ranged from 8 to 14, with an overall average CT severity score of 10.5 (median 9.5) (Extended Data Fig. 10b). The D614G/ P681R-infected hamsters had a higher CT severity score (mean 11 (range 9-14, median 10.5)), compared with the D614G-infected hamsters (mean 10 (range 8-13, median 9.5)). Two of the D614G/P681R-infected hamsters developed a small-volume pneumomediastinum, probably secondary to severe pulmonary damage, micropulmonary rupture and gas tracking into the mediastinum.

Discussion
Previous studies have demonstrated the close association of FCS in the SARS-CoV-2 S protein with the viral replication mode and its dependence on TMPRSS2. Johnson et al. 23 and Peacock et al. 22 showed that the loss of FCS results in an increase in viral replication efficacy in Vero cells and attenuates viral growth in Vero cells expressing TMPRSS2. By contrast, here we showed that the replication efficacy of the B.1.617.2/Delta variant was severely decreased in Vero cells compared with VeroE6/ TMPRSS2 cells. Importantly, although FCS-deleted SARS-CoV-2 is less pathogenic compared with its parental virus 23 , we revealed that the B.1.617.2/Delta variant and the P681R-harbouring virus exhibit higher pathogenicity. These findings suggest that enhanced viral fusogenicity, which is triggered by the P681R mutation, is closely associated with viral pathogenicity.
Although the P681R mutant is highly fusogenic, the virus containing the P681R mutation did not necessarily show stronger growth than the parental virus in in vitro cell cultures. HIV-1 variants with higher fusogenicity have been isolated from patients with AIDS, but the enhanced fusogenicity does not promote viral replication in in vitro cell cultures 24 . Similarly, a measles virus (Paramyxoviridae) containing mutations in viral matrix proteins 25 and substitution mutations in viral fusion proteins 26,27 is highly fusogenic and expands efficiently through cell-cell fusion. However, in in vitro cell cultures, the growth kinetics of these mutated measles viruses with higher fusogenicity are less efficient compared with those of the parental virus 25 . Thus, the discrepancy between the efficacy of viral growth in in vitro cell cultures and viral fusogenicity is not unique to SARS-CoV-2. However, higher fusogenicity is associated with the severity of viral pathogenicity, such as in HIV-1 encephalitis 28 and the fatal subacute sclerosing panencephalitis that is caused by measles virus infection in the brain 26,27 . Consistently, we showed that both the B.1.617.2/Delta variant and the P681R mutant exhibited higher fusogenicity in vitro and enhanced pathogenicity in vivo. Our data suggest that the greater COVID-19 severity and unusual symptoms caused by the B.1.617.2/Delta variant 13-15 are due in part to the higher fusogenicity caused by the P681R mutation.
After launching this research in May 2021, the B.1.617.2/Delta variant has rapidly surpassed the other VOCs and is a major driver of the current COVID-19 pandemic worldwide in only a few months after the emergence. Revealing the rationale of higher transmissibility of this variant is one of the most urgent and crucial issues in the current COVID-19 pandemic. However, transmission experiments using the B.1.617.2/Delta variant or P681R-bearing virus in animal models were not performed in this study, and it remains unaddressed why the B.1.617.2/ Delta variant has become more predominant than the other VOCs. By contrast, the greater severity and unusual COVID-19 symptoms caused by the B.1.617.2/Delta variant 13-15 should be another important issue; we therefore addressed the virological properties and virulence of this variant and showed evidence suggesting that this pandemic variant has enhanced fusogenicity and pathogenicity. We revealed the association of the P681R mutation with these virological features. An assumption from our observations is that the higher viral fusogenicity driven by the P681R mutation may be associated with the increased transmissibility of the B.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-021-04266-9.  Fig. 4, four hamsters per group were intranasally inoculated with the D614G or the D614G/P681R viruses (10 4 TCID 50 in 30 μl) under isoflurane anaesthesia. Body weight was monitored daily for 7 days. For virological examinations, four hamsters per group were intranasally infected with the D614G or the D614G/P681R viruses (10 4 TCID 50 in 30 μl); at 3 and 7 d.p.i., the hamsters were euthanized, and nasal turbinates and lungs were collected. The virus titres in the nasal turbinates and lungs were determined by plaque assays in VeroE6/ TMPRSS2 cells.

Histopathological analysis
Excised animal tissues were fixed with 4% paraformaldehyde in PBS, and processed for paraffin embedding. The paraffin blocks were sectioned with at a thicknes of 3 μm and then mounted on silane-coated glass slides (MAS-GP, Matsunami). H&E staining was performed according to a standard protocol. For immunohistochemistry analysis (Extended Data Fig. 5c), an Autostainer Link 48 (Dako) was used. The deparaffinized sections were exposed to EnVision FLEX target retrieval solution high pH (Agilent, K8004) for 20 min at 97 °C to activate, and a mouse anti-SARS-CoV-2 N monoclonal antibody (1:400 dilution; R & D systems, 1035111, MAB10474-SP) was used. The sections were sensitized using EnVision FLEX (Agilent) for 15 min and visualized by peroxidase-based enzymatic reaction with 3,3′-diaminobenzidine tetrahydrochloride as the substrate for 5 min.

Histopathological scoring of lung lesions
Pathological features, including bronchitis or bronchiolitis, haemorrhage or congestion, alveolar damage with epithelial apoptosis and macrophage infiltration, the presence of type II pneumocytes and the area of the presence of large type II pneumocytes ( Fig. 2f and Extended Data Fig. 5b), were evaluated by certified pathologists and the degree of these pathological findings were arbitrarily scored using four-tiered system as 0 (negative), 1 (weak), 2 (moderate) and 3 (severe). Especially, for the evaluation of the area of the large type II pneumocytes at 5 d.p.i., the presence of more than 5 large type II pneumocytes with a nuclear diameter more than 10 μm per 0.04 mm 2 were delineated and the areas were measured using Fiji software v.2.2.0 implemented in ImageJ v.2.2.0.

Lung function
Respiratory parameters were measured using a whole-body plethysmography system (PrimeBioscience) according to the manufacturer's instructions. In brief, hamsters were placed in unrestrained plethysmography chambers and allowed to acclimatize for 1 min, then data were acquired over a 3 min period using FinePointe v.2.8.0.12146 (Data Sciences International).

MicroCT imaging
Respiratory organs of the infected hamsters were imaged by using an in vivo microCT scanner (CosmoScan GXII; Rigaku) at 7 d.p.i. Under ketamine-xylazine and isoflurane for the induction and maintenance of anaesthesia, the hamsters were placed in the imaging chamber and were scanned for 4 min at 90 kV, 88 μA, FOV 45 mm and a pixel size 90.0 μm. After scanning, the lung images were reconstructed and analysed using the CosmoScan Database software v.3.3.27.100 (Rigaku). Qualitative and semiquantitative visual image analysis of the lungs was performed in three uninfected Syrian hamsters and the hamsters infected with D614G (n = 4) or D614G/P681R (n = 4) viruses at 7 d.p.i. A CT severity score (Extended Data Fig. 10b), which was adapted from a human scoring system, was used to grade the severity of the lung abnormalities 33 . Each lung lobe was analysed for degree of involvement and was scored from 0-4 as follows depending on the severity: 0 (none, 0%), 1 (minimal, 1%-25%), 2 (mild, 26%-50%), 3 (moderate, 51%-75%) or 4 (severe, 76%-100%). The scores for the five lung lobes were summed to obtain a total severity score of 0-20, reflecting the severity of abnormalities across the two infected groups. Images were anonymized and randomized; the scorer was blinded to the group allocation.

Viral genomes
All SARS-CoV-2 genome sequences and annotation information used in this study were downloaded from GISAID (https://www.gisaid.org) Article on 31 May 2021 (1,761,037 sequences). We first excluded genomes of viruses collected from non-human hosts. We obtained SARS-CoV-2 variants belonging to the B.1.617 lineage based on the PANGO annotation (that is, sublineages B. 1.617.1, B.1.617.2/Delta or B.1.617.3) for each sequence in the GISAID metadata. One variant annotated as belonging to the B.1.617 lineage (GISAID ID: EPI_ISL_1544002, isolated in India on 25 February 2021) was not used in the analysis because the variant was not assigned to any of the three sublineages, possibly due to the 212 undetermined nucleotides in the genome. To infer the epidemiology of the B.1.617 lineage (Fig. 1b-d), we excluded genomes for which sampling date information was not available. We analysed 2,855, 13,821 and 83 sequences belonging to the B. 1.617.1, B.1.617.2/Delta and B.1.617.3 sublineages, respectively.
A SARS-CoV-2 variant (GISAID ID: EPI_ISL_2220643) isolated in Texas, USA, on 10 August 2020, was also recorded to belong to B.1.617.1. However, the S protein of this viral sequence (GISAID ID: EPI_ISL_2220643) possesses neither L452R nor P681R mutations, both of which are features of the B.1.617 lineage. Thus, the EPI_ISL_2220643 sequence isolated in the USA is probably not the ancestor of the current B.1.617.1 lineage, and the EPI_ISL_1372093 sequence obtained in India can be considered to be the oldest example of the B.1.617 lineage.

Phylogenetic analyses
To infer the phylogeny of the B.1.617 sublineages, we screened SARS-CoV-2 genomes by removing genomes containing undetermined nucleotides at coding regions. As the numbers of genomes belonging to sublineages B.  34 . All sites with gaps in the alignment were removed, and the total length of the alignment was 29,085 nucleotides. A maximum likelihood tree was generated using IQ-TREE 2 v.2.1.3 with 1,000 bootstraps 35 . The GTR+G substitution model was used based on the BIC criterion.

SARS-CoV-2 preparation and titration
A B.1.617.2/Delta isolate (strain TKYTK1734; GISAID ID: EPI_ISL_2378732) and a D614G-bearing B.1.1 isolate (strain TKYE610670; GISAID ID: EPI_ISL_479681) were isolated from SARS-CoV-2-positive individuals in Japan. In brief, 100 μl of nasopharyngeal swabs obtained from SARS-CoV-2-positive individuals were inoculated into VeroE6/TMPRSS2 cells in a biosafety level 3 laboratory. After incubation at 37 °C for 15 min, maintenance medium supplemented with Eagle's minimum essential medium (FUJIFILM Wako Pure Chemical, 056-08385) containing 2% FBS and 1% PS was added, and the cells were cultured at 37 °C under 5% CO 2 . The cytopathic effect (CPE) was confirmed by observation under an inverted microscope (Nikon), and the viral load of the culture supernatant in which CPE was observed was confirmed by RT-qPCR. To determine the viral genome sequences, RNA was extracted from the culture supernatant using the QIAamp viral RNA mini kit (Qiagen, 52906). A cDNA library was prepared using the NEB Next Ultra RNA Library Prep Kit for Illumina (New England Biolab, E7530) and whole-genome sequencing was performed by MiSeq (Illumina).
To prepare the working virus stock, 100 μl of the seed virus was inoculated into VeroE6/TMPRSS2 cells (5 × 10 6 cells in a T-75 flask). Then, 1 h after infection, the culture medium was replaced with Dulbecco's modified Eagle's medium (low glucose) (Wako, 041-29775) containing 2% FBS and 1% PS. At 2-3 d.p.i., the culture medium was collected and centrifuged, and the supernatants were collected as the working virus stock.
The titre of the prepared working virus was measured as the 50% tissue culture infectious dose (TCID 50 ). In brief, 1 day before infection, VeroE6/TMPRSS2 cells (10,000 cells per well) were seeded into a 96-well plate. Serially diluted virus stocks were inoculated into the cells and incubated at 37 °C for 3 days. The cells were observed under microscopy to judge the CPE appearance. The value of TCID 50 per ml was calculated using the Reed-Muench method 36 .

SARS-CoV-2 infection
One day before infection, Vero cells (10,000 cells), VeroE6/TMPRSS2 cells (10,000 cells) and Calu-3 cells (10,000 cells) were seeded into a 96-well plate. SARS-CoV-2 was inoculated and incubated at 37 °C for 1 h. The infected cells were washed, and 180 μl of culture medium was added. The culture supernatant (10 μl) was collected at the indicated timepoints, and RT-qPCR was used to quantify the viral RNA copy number (see below). To monitor the syncytium formation in infected cell culture, bright-field photos were obtained using ECLIPSE Ts2 (Nikon). The sizes of floating syncytia were measured using the 'quick selection tool' in Photoshop 2020 v.21.0.2 (Adobe) as pixels, and the areas of floating syncytia were calculated from the pixel value. For the GFP-expressing recombinant viruses (Extended Data Fig. 6c), bright-field and green fluorescence images were obtained at the indicated timepoints using an All-in-One Fluorescence microscope BZ-X800 (Keyence), and the GFP fluorescence intensity was analysed using the BZ-X800 Analyzer v.1.

(Keyence).
For the infection experiment primary human nasal epithelial cells (Extended Data Fig. 7b), the working viruses were diluted with Opti-MEM (Thermo Fisher Scientific, 11058021). The diluted viruses (1,000 TCID 50 in 100 μl) were inoculated onto the apical side of the culture and incubated at 37 °C for 1 h. The inoculated viruses were removed and washed twice with Opti-MEM. To collect the viruses on the apical side of the culture, 100 μl Opti-MEM was applied onto the apical side of the culture and incubated at 37 °C for 10 min. Bright-field and green fluorescence images were obtained using ECLIPSE Ts2 (Nikon). The Opti-MEM applied was collected and we used RT-qPCR to quantify the viral RNA copy number (see below).

Plaque assay
Plaque assay (Extended Data Fig. 2a, c) was performed as previously described 37 . In brief, one day before infection, 200,000 VeroE6/ TMPRSS2 cells were seeded into a 12-well plate. The virus was diluted with serum-free Dulbecco's modified Eagle's medium (low glucose) (Wako, 041-29775) containing 1% PS and 20 mM HEPES. After removing the culture medium, the cells were infected with 500 μl of the diluted virus at 37 °C. At 2 h.p.i., 1 ml of mounting solution containing 3% FCS and 1.5% carboxymethyl cellulose (Sigma-Aldrich, C9481-500G) was overlaid, followed by incubation at 37 °C. At 3 d.p.i., the culture medium was removed, and the cells were washed three times with PBS containing 0.9 mM calcium chloride and 0.5 mM magnesium chloride and fixed with 10% formaldehyde neutral buffer solution (Nacalai Tesque, 37152-51). The fixed cells were washed with tap water, dried and stained with staining solution (2% crystal violet (Nacalai Tesque, 09804-52) in water) for 30 min. The stained cells were washed with tap water and dried, and the size of plaques was measured using ImageJ.

SARS-CoV-2 reverse genetics
Recombinant SARS-CoV-2 was generated by circular polymerase extension reaction (CPER) as previously described 32,37 . In brief, nine DNA fragments encoding the partial genome of SARS-CoV-2 (strain WK-521, PANGO lineage A; GISAID ID: EPI_ISL_408667) 29 were prepared by PCR using PrimeSTAR GXL DNA polymerase (Takara, R050A). A linker fragment encoding hepatitis delta virus ribozyme, bovine growth hormone polyA signal and cytomegalovirus promoter was also prepared by PCR. A summary of the corresponding SARS-CoV-2 genomic region and the PCR templates and primers used for this procedure is provided in Supplementary Table 3. The ten obtained DNA fragments were mixed and used for CPER 32 . To prepare GFP-expressing replication-competent recombinant SARS-CoV-2, we used fragment 9, in which the GFP gene was inserted into the ORF7a frame, instead of the authentic F9 fragment 32 (Supplementary Table 3).
To produce recombinant SARS-CoV-2, the CPER products were transfected into HEK293-C34 cells using TransIT-LT1 (Takara, MIR2300) according to the manufacturer's protocol. At 1 day after transfection, the culture medium was replaced with Dulbecco's modified Eagle's medium (high glucose) (Sigma-Aldrich, R8758-500ML) containing 2% FCS, 1% PS and doxycycline (1 μg ml −1 ; Takara, 1311N). At 6 days after transfection, the culture medium was collected and centrifuged, and the supernatants were collected as the seed virus. To remove the CPER products (that is, SARS-CoV-2-related DNA), 1 ml of the seed virus was treated with 2 μl TURBO DNase (Thermo Fisher Scientific, AM2238) and incubated at 37 °C for 1 h. Complete removal of the CPER products (that is, 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 above.

Pseudovirus assay
Pseudovirus assays were performed as previously described 31,32,37 . In brief, lentivirus (HIV-1)-based luciferase-expressing reporter viruses pseudotyped with the SARS-CoV-2 S protein and its derivatives, HEK293T cells (1 × 10 6 cells), were cotransfected with 1 μg of psPAX2-IN/ HiBiT 46 , 1 μg of pWPI-Luc2 46 and 500 ng of plasmids expressing parental S or its derivatives using Lipofectamine 3000 (Thermo Fisher Scientific, L3000015) or PEI Max (Polysciences, 24765-1) according to the manufacturer's protocol. At 2 days after transfection, the culture supernatants were collected and centrifuged. The amount of pseudovirus prepared was quantified using the HiBiT assay as previously described 31, 46 . The prepared pseudoviruses were stored at −80 °C until use. For the experiment, HOS-ACE2 cells and HOS-ACE2/TMPRSS2 cells (10,000 cells per 50 μl) were seeded in 96-well plates and infected with Article 100 μl of pseudoviruses prepared at four different doses. At 2 d.p.i., the infected cells were lysed with a One-Glo luciferase assay system (Promega, E6130), and the luminescent signal was measured using a CentroXS3 plate reader (Berthhold Technologies) or GloMax explorer multimode microplate reader 3500 (Promega).

SARS-CoV-2 S-based fusion assay
The SARS-CoV-2 S-based fusion assay was performed as previously described 37 . This assay uses a dual split protein (DSP) encoding Renilla luciferase (RL) and GFP genes; the respective split proteins, DSP 1-7 and DSP [8][9][10][11] , are expressed in effector and target cells by transfection 48,50 . In brief, on day 1, effector cells (that is, S-expressing cells) and target cells (that is, ACE2-expressing cells) were prepared at a density of 0.6-0.8 × 10 6 cells in a six-well plate. To prepare effector cells, HEK293 cells were cotransfected with 400 ng of the S expression plasmids and 400 ng pDSP 1-7 using TransIT-LT1 (Takara, MIR2300). To prepare the target cells, HEK293 cells were cotransfected with pC-ACE2 (0 ng, 200 ng or 1,000 ng) and pDSP [8][9][10][11] (400 ng). Target cells in selected wells were cotransfected with pC-TMPRSS2 (40 ng) in addition to the abovementioned plasmids. On day 3 (24 h after transfection), 16,000 effector cells were detached and reseeded into 96-well black plates (PerkinElmer, 6005225), and target cells were reseeded at a density of 1,000,000 cells per 2 ml per well in six-well plates. On day 4 (48 h after transfection), the target cells were incubated with EnduRen live cell substrate (Promega, E6481) for 3 h and then detached, and 32,000 target cells were added to a 96-well plate with effector cells. RL activity was measured at the indicated timepoints using a Centro XS3 LB960 (Berthhold Technologies). The S proteins expressed on the surfaces of effector cells were stained with rabbit anti-SARS-CoV-2 S monoclonal antibody (HL6, GeneTex, GTX635654) or rabbit anti-SARS-CoV-2 S S1/S2 polyclonal antibody (Thermo Fisher Scientific, PA5-112048). Normal rabbit IgG (Southern-Biotech, 0111-01) was used as a negative control, and APC-conjugated goat anti-rabbit IgG polyclonal antibody ( Jackson ImmunoResearch, 111-136-144) was used as a secondary antibody. Expression levels of surface S proteins were analysed using FACS Canto II (BD Biosciences) and FlowJo v.10.7.1 (BD Biosciences). RL activity was normalized to the mean fluorescence intensity of surface S proteins, and the normalized values are shown as fusion activity.

Mathematical modelling for fusion velocity quantification
The following cubic polynomial regression model was fitted to each of the time-series datasets (Fig. 3e): The initial velocity of cell fusion was estimated from the derivative of the fitted cubic curve.

Statistics and reproducibility
In the time-course experiments using hamsters (Figs. 2c, d, f, 4a, b), two types of statistical tests were performed. First, to evaluate the difference between experimental conditions through all timepoints, a multiple regression analysis including experimental conditions as explanatory variables and timepoints as qualitative control variables was performed. P values were calculated using two-sided Wald tests. In Fig. 2f, family-wise error rates were calculated using the Holm method. Second, to evaluate the difference between two conditions at each timepoint, two-sided Student's t-tests were performed. The data were analysed using Excel v. 16.16.8 (Microsoft) or Prism 9 v.9.1.1 (GraphPad Software).
In Fig. 4d and Extended Data Fig. 5, the photographs shown are the representative areas of two independent experiments using 3 hamsters (6 lungs) at each timepoint. In Extended Data Fig. 3, assays were performed in quadruplicate. Photographs shown are the representative of 40 fields of view taken for each sample.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.