As of December 2022, the successive waves of the COVID-19 pandemic have been driven by five major SARS-CoV-2 variants, known as variants of concern (VOC): Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1), Delta (B.1.617.2 and AY lineages) and Omicron (BA lineages)13. Omicron is the most recently recognized VOC, and was first documented in South Africa, Botswana and in a traveller from South Africa in Hong Kong in November 2021 (GISAID ID: EPI_ISL_7605742)14,15. It quickly swept through the world, displacing the previously dominant Delta variant within weeks and accounting for most new SARS-CoV-2 infections by January 2022 (refs. 16,17,18). At least five lineages of Omicron have so far been identified: BA.1, BA.2, BA.3, BA.4 and BA.5. BA.1 (hereafter referred to as Omicron) exhibits a remarkable escape from infection- and vaccine-induced humoral immunity3,19. Furthermore, it is less virulent than other VOCs in humans and in vivo models of infection4,5,7,11,12,20. Omicron differs from the prototype SARS-CoV-2 isolate, Wuhan-Hu-1, by 59 amino acids; 37 of these changes are in the S protein, raising the possibility that S is at the heart of Omicron’s pathogenic and antigenic behaviour.

S mutations affect Omicron replication in cell culture

The Omicron S protein contains 30 amino acid substitutions, 6 deletions and one insertion of 3 amino acids in length, as compared to Wuhan-Hu-1 (Extended Data Fig. 1). Twenty-five of these changes are unique to Omicron relative to other VOCs, although some of them have been reported in waste water and minor SARS-CoV-2 variants21,22. To test the role of the S protein in Omicron phenotype, we generated a chimeric recombinant virus containing the S gene of Omicron (USA-lh01/2021) and all other genes of an ancestral SARS-CoV-2 (GISAID EPI_ISL_2732373)23 (Extended Data Fig. 2a). This chimeric virus, named Omi-S, was made by using a modified form of circular polymerase extension reaction (CPER)24 (Extended Data Fig. 2b) that yielded highly concentrated virus stocks, containing 0.5 × 106–5 × 106 plaque-forming units (PFU) per ml, from transfected cells within two days of transfection (Extended Data Fig. 2c,d), obviating the need for further viral amplification.

We first compared the infection efficiency of Omi-S with an ancestral D614G-containing virus (GISAID EPI_ISL_2732373; generated by CPER; hereafter referred to as wild type (WT)) and an Omicron isolate (USA-lh01/2021) in cell culture (Fig. 1a). For this, we infected ACE2- and TMPRSS2-expressing Caco-2 (hereafter, ACE2/TMPRSS2/Caco-2) and Vero E6 cells with Omi-S, WT and Omicron at a multiplicity of infection (MOI) of 0.01, and monitored viral propagation by flow cytometry and plaque formation assay. The WT virus and Omi-S spread rapidly in ACE2/TMPRSS2/Caco-2 cells, yielding 89% and 80% infected cells, respectively, at 24 hours post-infection (hpi) (Fig. 1b). By contrast, Omicron replicated more slowly, leading to 48% infected cells at 24 hpi. A similar pattern was seen in Vero E6 cells, in which 60% and 41% of cells were positive for WT and Omi-S, respectively, at 48 hpi, as opposed to 10% for Omicron (Fig. 1c). The plaque assay showed that although both Omi-S and Omicron produced lower levels of infectious virus particles compared with WT, the viral titre of Omi-S was significantly higher than that of Omicron. In ACE2/TMPRSS2/Caco-2 cells, Omi-S produced 5.1-fold (P = 0.0006) and 5.5-fold (P = 0.0312) more infectious particles than Omicron at 12 hpi and 24 hpi, respectively (Fig. 1d). Similarly, in Vero E6 cells, the infectious virus titres of Omi-S were 17-fold (P = 0.0080) and 11-fold (P = 0.0078) higher than that of Omicron at 24 hpi and 48 hpi, respectively (Fig. 1e). The difference between viruses became less obvious at later time points owing to higher cytotoxicity caused by Omi-S compared to Omicron (Extended Data Fig. 3a).

Fig. 1: Effect of S on the in vitro growth kinetics of Omicron.
figure 1

a, Schematic of viruses. S, spike; N, nucleocapsid. be, ACE2/TMPRSS2/Caco-2 cells (b,d) and Vero E6 cells (c,e) were infected at an MOI of 0.01, and the percentage of N-positive cells (n = 6 replicates) (b,c) and the release of infectious particles (n = 3 replicates) (d,e) were determined by flow cytometry and by plaque assay, respectively. f, ACE2/TMPRSS2/Caco-2 cells were infected with virus mixtures at a 1:1 ratio to obtain the final MOI of 0.005 for each virus. The cells were fixed at the indicated times and subjected to flow cytometry. Left, representative dot plot; right, percentage of uninfected, Omi-S/mCherry-infected, Omicron/mNeoGreen (Omicron/mNG)-infected and doubly infected cells. Singly infected cells were used for compensation. Error bars, mean ± s.d. (n = 3 replicates). g, Plaque sizes. Left, representative images of plaques on ACE2/TMPRSS2/Caco-2 cells. Right, the diameter of plaques is plotted as the mean ± s.d. of 20 plaques per virus. h, Human iAT2 epithelial cells were infected at an MOI of 2.5 for 48 h or 96 h. The apical side of cells was washed with 1× phosphate-buffered saline (PBS) and the levels of infectious virus particle were measured by plaque assay. Error bars, mean ± s.d. (n = 4 replicates). Experiments were repeated twice, with each experimental repeat containing 3 (bg) or 4 (h) replicates. P values were calculated by a two-tailed, unpaired t-test with Welch’s correction. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; NS, not significant. The gating strategy for flow cytometry is shown in Supplementary Fig. 1.

The increased replication efficiency of Omi-S relative to Omicron was preserved when tested at varying MOIs (Extended Data Fig. 3b). We further confirmed the fitness advantage of Omi-S over Omicron by a direct competition assay. For this, we first generated recombinant Omicron (rOmicron), which, in our cell culture assays, mimicked the replication kinetics of natural Omicron (Extended Data Fig. 4). Next, we created mCherry-containing Omi-S and mNeonGreen-containing Omicron, and inoculated ACE2/TMPRSS2/Caco-2 cells with these viruses mixed at a 1:1 ratio. Flow cytometric analysis of infected cells at various times of infection showed that Omi-S/mCherry was clearly superior to Omicron/mNeonGreen in terms of replication (Fig. 1f). Finally, the higher infection efficiency of Omi-S was also reflected in the plaque size; although WT virus produced the largest plaques (around 4.1 mm), the size of Omi-S plaques (around 2.2 mm) was twofold (P < 0.0001) larger than that of Omicron plaques (around 1.1 mm) (Fig. 1g). These results indicate that although mutations in the S protein influence the infection efficiency of Omicron, they do not fully explain the Omicron phenotype.

Several lines of evidence indicated that the S protein incorporated into Omi-S behaved the same way as in natural Omicron. For instance, as described previously20,25, Omicron S was poorly cleaved compared to WT S; whereas 71% of S in WT virions was in the cleaved form, only 45% and 47% was cleaved in Omi-S and Omicron, respectively (Extended Data Fig. 5a). The same pattern of S cleavage was evident in virus-infected cells (WT, 63% cleaved; Omi-S, 33% cleaved; Omicron, 42% cleaved) (Extended Data Fig. 5b). These experiments also revealed that Omicron S was inefficiently incorporated into virus particles compared to WT S (S to nucleocapsid (N) ratio: 3.40 for WT virus, 1.91 for Omi-S and 2.04 for Omicron) (Extended Data Fig. 5a). Similarly, both Omi-S and Omicron produced smaller syncytia compared to the WT virus, an observation that has previously been reported for Omicron20,26 (Extended Data Fig. 5c). Finally, consistent with the published literature25, Omi-S and Omicron showed a preference for cathepsin-mediated entry, as reflected by their higher sensitivity to the cathepsin inhibitor E64d (Extended Data Fig. 6).

We next compared the replication kinetics of WT, Omi-S and Omicron in lung epithelial cells, which form a major viral replication site in patients with COVID-19 (refs. 27,28). Accordingly, we used human induced pluripotent stem (iPS)-cell-derived lung alveolar type 2 epithelial (iAT2) cells. AT2 cells represent an essential cell population in the distal lung and constitute one of the primary targets of SARS-CoV-2 infection28,29. We infected iAT2 cells, grown as an air–liquid interface (ALI) culture, at an MOI of 2.5 and monitored the secretion of viral progeny on the apical side of cells at 48 hpi and 96 hpi. In congruence with the results obtained from cell lines, WT virus produced the highest levels of infectious virus particles (Fig. 1h). Among Omi-S and Omicron, the former yielded around fivefold (P = 0.0008) higher infectious viral titre at 48 hpi. The viral titres for WT and Omi-S decreased at 96 hpi compared to 48 hpi owing to the cytopathic effect (CPE) of infection. However, no CPE was seen for Omicron, leading to sustained production of infectious virions. Overall, these results corroborate the conclusion that mutations in S do not fully account for the attenuated replication capacity of Omicron in cell culture.

Minimal role of S in Omicron pathogenicity in mice

To examine whether Omi-S exhibits higher in vivo fitness compared with Omicron, we investigated the infection outcome of Omi-S relative to WT SARS-CoV-2 and Omicron in K18-hACE2 mice. In agreement with the published literature4,5, intranasal inoculation of mice (aged 12–20 weeks) with Omicron (104 PFU per mouse) caused no significant weight loss, whereas inoculation with WT virus triggered a rapid decrease in body weight, with all mice losing over 20% of their initial body weight by 8 days post-infection (dpi) (Fig. 2a). Notably, 80% of mice infected with Omi-S also lost over 20% of their body weight by 9 dpi (Fig. 2a and Extended Data Fig. 7a). The evaluation of clinical scores (a cumulative measure of weight loss, abnormal respiration, aberrant appearance, reduced responsiveness and altered behaviour) also revealed a similar pattern; whereas Omicron-infected mice exhibited few to no signs of clinical illness, the health of those infected with WT and Omi-S rapidly deteriorated, with the WT virus causing a more severe disease (P = 0.0102) (Fig. 2b and Extended Data Fig. 7b). Because SARS-CoV-2 causes fatal infection in K18-hACE2 mice4, we compared the survival of mice after viral infection. Agreeing with the results of body-weight loss and clinical score, WT and Omi-S caused mortality rates of 100% (6/6) and 80% (8/10), respectively. By contrast, all mice infected with Omicron survived (Fig. 2c). These findings, which are consistent with a recent publication30, indicate that the S protein is not the exclusive determinant of Omicron’s pathogenicity in K18-hACE2 mice.

Fig. 2: Role of S in Omicron pathogenicity.
figure 2

ac, Male and female K18-hACE2 mice (aged 12–20 weeks) were intranasally inoculated with 1 × 104 PFU of WT (n = 6 mice), Omi-S (n = 10 mice) or Omicron (n = 10 mice) virus. Two independently generated virus stocks were used in this experiment. Body weight (a), clinical score (b) and survival (c) were monitored daily for 14 days. Mice that lost 20% of their initial body weight were euthanized. d,e, K18-hACE2 mice were intranasally inoculated with 1 × 104 PFU of WT (n = 14 mice), Omi-S (n = 14 mice) and Omicron (n = 14 mice). Lung samples of the infected mice were collected at 2 or 4 dpi to determine the viral titre (n = 4 mice) (d) or for immunohistochemistry (IHC) detection of the N protein (n = 3 mice) (e). In e, representative IHC images showing SARS-CoV-2 N (brown colour) in alveoli (arrows) and bronchioles (arrowheads) in mice lungs at 2 dpi are presented. Scale bars, 100 µm. f, Percentage of N-positive bronchioles in the lungs of infected mice (n = 3 mice) at 2 dpi. Each dot represents an infected mouse. Statistical significance was determined using a two-tailed, unpaired t-test with Welch’s correction (a,b,d,f) and log-rank (Mantel-Cox) test (c). *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; NS, not significant.

Source data

Next, we compared the propagation of Omi-S with Omicron and WT SARS-CoV-2 in the lungs and nasal turbinates of K18-hACE2 mice. The mice (12–20 weeks old) were intranasally challenged with 104 PFU (seven mice per virus), and viral titres in mice lungs were measured at 2 and 4 dpi. Consistent with in vitro findings, the infectious virus titre in the lungs of WT-infected mice was higher than that detected in mice infected with the other two viruses (Fig. 2d). Notably, however, Omi-S-infected mice produced 30-fold (P = 0.0286) more infectious virus particles compared with Omicron-infected mice at 2 dpi. The titre decreased at 4 dpi for WT- and Omi-S-infected mice, but it showed an increasing trend for Omicron-infected mice, pointing to the possibility of mild but persistent infection by Omicron in K18-hACE2 mice. All three variants recovered from the lungs of mice maintained the same plaque size phenotype as the original inoculum, indicating that replication in mice had no detectable effect on genotypes of these viruses (data not shown).

To evaluate the viral pathogenicity in lungs and nasal turbinates of K18-hACE2 mice, we performed histopathological analysis of these tissues at 2 dpi. As previously reported4,31, an extensive near-diffused immunoreactivity of the SARS-CoV-2 N protein was detected in lung alveoli of mice infected with WT virus (Fig. 2e). By contrast, Omi-S and Omicron infection produced localized foci of alveolar staining with fewer foci for Omicron compared with Omi-S. The most marked phenotype was seen in bronchiolar epithelium, in which Omi-S caused pronounced, routinely circumferential infection, with around 10–15% of bronchioles being positive for viral N protein at 2 dpi, whereas only 3–5% of bronchioles were N-positive for Omicron (Fig. 2f). WT virus infected around 1% of bronchioles and in all cases only included a single isolated epithelial cell per bronchiole. Furthermore, bronchiolar infection was associated with epithelial necrosis in Omi-S-infected mice, as determined through serial haematoxylin and eosin (H&E) section analysis, whereas no histological evidence of airway injury was observed in Omicron- or WT-infected mice (Extended Data Fig. 8a,b). The nasal turbinates of mice inoculated with WT and Omi-S viruses both contained abundant SARS-CoV-2-positive cells, which were associated with overt cytopathic effects, whereas Omicron produced rare, sporadic positive cells, with no apparent signs of epithelial injury (Extended Data Fig. 8c). Overall, these findings suggest that replication of Omicron in the mice respiratory tract is substantially attenuated compared to Omi-S, supporting our conclusion that mutations in S are only partially responsible for the attenuated pathogenicity of Omicron.

Mutations in S and nsp6 define Omicron attenuation

In addition to the S protein, Omicron has amino acid changes in non-structural protein 3 (nsp3), nsp4, nsp5, nsp6, nsp14, envelope (E), membrane (M) and N proteins, when compared with WT virus (Extended Data Fig. 9a). To identify non-spike proteins that are involved in Omicron attenuation, we generated a large panel of fluorescently labelled chimeric viruses, each containing Omicron S in combination with one non-spike protein of Omicron, with the remaining proteins coming from WT virus (Extended Data Fig. 9b). When we combined Omicron S with Omicron nsp6 (Omi-S/nsp6), we observed a strong decrease in viral replication, with infection kinetics mimicking those of Omicron in cell culture (Fig. 3a–d); no such decrease was seen for other chimeric viruses. Poor replication efficiency of Omi-S/nsp6 was also corroborated by our finding that both Omi-S/nsp6 and Omicron took almost five to six days to recover by CPER, whereas all other variants were recovered in two days (data not shown). Finally, like Omicron, Omi-S/nsp6 was clearly outcompeted by Omi-S in a direct competition assay (Fig. 3e).

Fig. 3: Mutations in S and nsp6 drive Omicron pathogenicity.
figure 3

ad, Replication kinetics of indicated mNeonGreen reporter viruses in ACE2/TMPRSS2/Caco-2 cells (MOI = 0.01) determined by flow cytometry (n = 3 replicates) (a,c) and plaque assay (n = 3 replicates) (b,d). Experiments were repeated twice. e, ACE2/TMPRSS2/Caco-2 cells were infected with virus mixtures at a 1:1 ratio to obtain the final MOI of 0.005 for each virus. The cells were fixed at the indicated times and analysed by flow cytometry. Percentage of uninfected, singly infected and doubly infected cells is shown. Singly infected cells were used for compensation. Individual data points are plotted along with the mean ± s.d. (n = 3 replicates). The experiment was repeated twice. fh, K18-hACE2 mice were intranasally inoculated with 1 × 104 PFU of viruses. Lung samples of infected mice were collected at 2 dpi for IHC detection of the N protein (n = 3 mice) (f) or for determination of viral titres (n = 4 mice) (g). In f, representative images of H&E staining of N-positive bronchioles are shown in insets. Bronchiolar epithelial necrosis is indicated with arrows. No evidence of necrosis was seen in bronchioles of mice infected with Omicron or Omi-S/nsp6. Scale bar, 100 µm. The right graph in f shows the percentage of N-positive bronchioles in the lungs of infected mice. Each dot represents an infected mouse. h, Survival of infected mice monitored daily for 14 days. Mice that lost 20% of their initial body weight were euthanized. Statistical significance was determined using a two-tailed, unpaired t-test with Welch’s correction (ag) and log-rank (Mantel-Cox) test (h). *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; NS, not significant.

Source data

In lungs of K18-hACE2 mice, whereas Omi-S caused extensive bronchiolar infection and injury, both Omicron and Omi-S/nsp6 showed decreased infection with no evidence of epithelial damage (Fig. 3f). Consistent with these findings, lungs of Omi-S/nsp6-infected mice produced viral titres equivalent to those seen for rOmicron and Omicron isolate (Fig. 3g). Finally, 71% of mice infected with Omi-S/nsp6 survived (Fig. 3h)—in contrast with the survival rates of only 20% that were observed in mice infected with Omi-S (Fig. 2c). Overall, these results indicate that mutations in S and nsp6 are sufficient to define Omicron’s attenuated virulence. These observations support and further extend the findings of a previous study showing that mutations in the 5′-UTR–nsp12 region, in which nsp6 resides, contribute to Omicron’s attenuation in K18-hACE2 mice30.

The S protein RBM drives Omicron’s vaccine escape

A large body of literature has provided evidence of the extensive escape of Omicron from vaccine-induced humoral immunity14,19. To define S regions that are associated with the immune-escape phenotype of Omicron, we first compared the in vitro neutralization activity of sera from vaccinated individuals against WT SARS-CoV-2 (USA-WA1/2020), Omi-S and Omicron. Sera collected within two months of the second dose of mRNA-1273 (Moderna mRNA vaccine; n = 12) or BNT162b2 (Pfizer-BioNTech mRNA vaccine; n = 12) vaccine were included (Extended Data Table 1). We performed a multicycle neutralization assay using a setting in which the virus and neutralizing sera were present at all times, mimicking the situation in a seropositive individual. All sera poorly neutralized Omicron, with an 11.1-fold (range: 4.4-fold to 81.2-fold; P < 0.0001) lower half-maximal neutralizing dilution (ND50) for Omicron compared with WA1 (Fig. 4a,b). In fact, around 80% of samples did not completely neutralize Omicron at the highest tested concentration (Extended Data Fig. 10). Notably, Omi-S exhibited identical ND50 values to Omicron (11.5-fold lower than that of WA1; P < 0.0001) (Fig. 4a,b), suggesting that the Omicron S protein, when incorporated into a WT virus, behaves the same way as it does in Omicron.

Fig. 4: Role of S in the immune resistance of Omicron.
figure 4

a, ND50 values for WA1, Omi-S and Omicron in sera from individuals who received two shots of Moderna (donors 1–12) or Pfizer (donors 13–24) vaccine (further details of sera are provided in Extended Data Table 1; individual curves are shown in Extended Data Fig. 10). b, Trajectories of ND50 values against WA1, Omi-S and Omicron (the data from a are plotted). The fold change in ND50 values is indicated (n = 24 serum samples). cf, Schematic of the chimeric (left; c,d) and mutant (left; e,f) viruses. The amino acid numbering for WA1 mutants in e is based on the WA1 S sequence, whereas the numbering for Omicron mutants in f is based on the Omicron S sequence. Six of the 24 sera (3 from Moderna and 3 from Pfizer) were tested. Each serum sample is represented by a dot of specific colour. The data are plotted as the fold change of the parental virus. Statistical significance was determined using a two-tailed, unpaired t-test with Welch’s correction. *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001; NS, not significant.

The SARS-CoV-2 S protein comprises two domains: the S1 domain, which interacts with the ACE2 receptor, and the S2 domain, which is responsible for membrane fusion32. Within the S1 domain lie an N-terminal domain (NTD) and a receptor-binding domain (RBD), which contains the receptor-binding motif (RBM) that makes direct contact with the ACE2 receptor33. The NTD of Omicron S has 11 amino acid changes, including 6 deletions and one 3-amino-acid-long insertion, whereas the RBD contains 15 mutations, 10 of which are concentrated in the RBM (Extended Data Fig. 1). Both the NTD and the RBD host neutralizing epitopes34,35,36,37, but the RBD is immunodominant and represents the primary target of the neutralizing activity that is present in SARS-CoV-2 immune sera37,38. To determine whether the neutralization resistance phenotype of Omicron is caused by mutations in a particular domain of the S protein, we generated two groups of chimeric viruses. The first group comprised the WA1 virus carrying the NTD, RBD or RBM of Omicron (Fig. 4c), and the second group consisted of Omi-S virus bearing the NTD, RBD or RBM of WA1 (Fig. 4d). The neutralization assay showed that mutations in the RBM were the main cause of Omicron’s resistance to vaccine-induced humoral immunity: replacing the RBM of WA1 with that of Omicron decreased the ND50 by 5.4-fold (P < 0.0001), and, conversely, substituting the RBM of Omi-S with that of WA1 increased the ND50 by 5.6-fold (P = 0.0003) (Fig. 4c,d). The fact that none of the RBM-swap viruses achieved the difference of around 11-fold that was seen between WA1 and Omi-S suggests that mutations in other parts of S also contribute to vaccine resistance.

To investigate whether specific mutations in Omicron RBM drive vaccine escape, we generated two additional panels of recombinant viruses, one with WA1 S carrying Omicron RBM mutations, either singly or in combination (Fig. 4e), and the other with Omicron S lacking the same set of mutations (Fig. 4f). Two WA1 mutants—mutant 3 (with an E484A substitution) and mutant 4 (bearing a cluster of five substitutions Q493R, G496S, Q498R, N501Y and Y505H)—exhibited a moderate but statistically significant decrease of 1.4-fold (P = 0.0002) and 1.8-fold (P = 0.0003) in ND50 values, respectively, compared with WA1 (Fig. 4e). The opposite was observed when these mutations were removed from Omicron S; the Omicron mutant 3 (lacking the E484A substitution) and mutant 4 (lacking Q493R, G496S, Q498R, N501Y and Y505H) had 1.9-fold (P = 0.0082) and 3.1-fold (P = 0.0025) higher ND50 values compared with Omicron (Fig. 4f). As none of the mutants captured the overall phenotype of Omicron, we assume that the vaccine escape is a cumulative effect of mutations distributed along the length of the S protein. It is possible that mutations alter the conformation of Omicron S in such a manner that most of the immunodominant neutralizing epitopes are disrupted and become unavailable for neutralization.


This study provides key insights into viral proteins that contribute to SARS-CoV-2 pathogenicity. We show that S, the most mutated protein in Omicron, has an incomplete role in Omicron attenuation. In cell-based infection assays, the Omi-S virus exhibits an intermediate replication efficiency between the ancestral virus and Omicron. Similarly, in K18-hACE2 mice, Omi-S contrasts with non-fatal Omicron and leads to 80% mortality; the ancestral virus causes 100% mortality in these mice. Notably, when we combined S mutations with mutations in nsp6, the virus exhibited an attenuated phenotype largely resembling that of Omicron, indicating that these two proteins are major determinants of Omicron pathogenicity. Future studies will decipher the mechanism(s) by which nsp6 mutations affect viral replication.

One potential limitation of our study is the use of K18-hACE2 mice for pathogenesis studies instead of primate models that have more similarities with humans39. It should however be noted that K18-hACE2 mice are a well-established model for investigating the lethal phenotype of SARS-CoV-24,31. Although these mice develop lung pathology after SARS-CoV-2 infection, mortality has been associated with involvement of the central nervous system owing to viral neuroinvasion and dissemination31,40. The fact that infection of K18-hACE2 mice with Omi-S, but not with Omicron, elicits neurological signs (for example, hunched posture and a lack of responsiveness) suggests that the neuroinvasion property is preserved in Omi-S, probably as a result of its higher replication efficiency, and that the determinants of this property lie outside of the S protein. These findings are consistent with a previous study showing that hamsters infected with Omi-S shed significantly more virus and lost more weight than those infected with Omicron, suggesting that mutations outside of S contribute to the attenuated pathogenicity of Omicron41.

We found that although the ancestral virus mainly replicates in lung alveoli and causes only rare infection of bronchioles in K18-hACE2 mice, both Omi-S and Omicron exhibit an increased propensity to replicate in bronchiolar epithelium, indicating that the S protein is accountable for the changed tropism. The mechanism behind this switch is unknown, but it is possible that Omicron S is more efficient than WT S in using cathepsin B or cathepsin L (refs. 25,42,43), which form an active viral entry pathway in bronchioles and other airway cells41. By contrast, the entry of SARS-CoV-2 into alveolar epithelial cells is mainly driven by TMPRSS2 (refs. 28,44), which Omicron S is deficient in utilizing25,45, leading to poor infection of these cells4,5,25,43. These findings may explain the attenuated lung pathology caused by Omicron.

Omicron nsp6 has two altered sites relative to the prototype SARS-CoV-2 Wuhan-Hu-1 isolate: a three-amino-acid deletion (LSG, positions 105–107) and an I189V substitution (Extended Data Fig. 9). Several functions of nsp6 in coronavirus replication have been described; chief among them is the biogenesis of double-membrane vesicles (DMVs), which represent the site of viral RNA synthesis46,47,48,49,50. A previous study showed that SARS-CoV-2 DMVs are mainly generated by the concerted action of three viral proteins: nsp3, nsp4, and nsp6; although nsp3 and nsp4 are sufficient for the formation of DMVs, nsp6 connects these DMVs with the endoplasmic reticulum and channelizes the essential communication between these structures46. Whether the constellation of mutations in Omicron nsp6 affects the formation or functions of DMVs needs further investigation. Nsp6 also activates NLR3-dependent cytokine production and pyroptosis in the lungs of patients with COVID-19, serving as a key virulence factor47. Of note, an nsp6 variant that is associated with asymptomatic COVID-19 exhibited a reduced ability to induce pyroptosis47, prompting speculation that mutations in Omicron nsp6 may also influence pyroptosis. Detailed mechanistic studies will be required to dissect the effect of Omicron mutations on the functions of nsp6.

It is at present unknown whether mutations in S and nsp6 work in concert with each other to drive Omicron attenuation. Given that Omicron S showed a higher predilection for bronchioles, it is possible that S is responsible for the altered viral tropism, whereas non-spike mutations—including those in nsp6—are mere adaptations to the changed tissue environment. It is worth mentioning that although nsp6 seems to be the major non-spike protein behind Omicron attenuation, the contribution of other viral proteins cannot be completely ruled out. In vitro experiments examining the role of non-spike mutations were all carried out in ACE2/TMPRSS2/Caco-2 cells. Using other, more immune-competent cell types could reveal the effect of other non-spike mutations as well. In addition, our chimeric viruses contained Omicron S paired with only one non-spike protein at a time, which limited long-range epistatic interactions between mutations in multiple viral proteins.

Our study shows that mutations in the RBM of Omicron S are the main determinants of Omicron’s escape from neutralizing antibodies, although mutations in other regions of S also contribute. Within the RBM, we identify two hotspots of mutations, which give Omicron S the ability to resist neutralization: one bearing the E484A substitution and the other containing a cluster of five substitutions—Q493R, G496S, Q498R, N501Y and Y505H. The E484A substitution has been shown to escape neutralization by convalescent sera51. Moreover, structural modelling suggests that some therapeutic monoclonal antibodies establish highly stable salt bridges with the E484 residue, entirely losing their binding when this residue is changed to A or after Q493K and Y505H changes52. Similarly, mapping of RBM residues that directly interact with 49 known neutralizing antibodies revealed N440, G446, S477 and T478 as low-frequency interactors, N501, Y505 and Q498 as medium-frequency interactors and E484 and Q493 as high-frequency interactors53, in line with our neutralization assay results. Notably, although the antibody-binding potential of Omicron S is impaired54, its receptor-binding capacity is intact. In fact, the Omicron RBD has a higher affinity for ACE2 relative to the Wuhan-Hu-1 and Delta RBDs25. This indicates that mutations in Omicron S have evolved in such a manner that they hinder antibody binding but preserve the receptor engagement. This opens up the possibility of targeting the conserved and structurally constrained regions of S that are involved in ACE2 recognition for the design of broad-spectrum vaccines to control the COVID-19 pandemic.


Cells, antibodies and plasmids

The cell lines were incubated at 37 °C and 5% CO2 in a humidified incubator. Human embryonic kidney HEK293T cells (ATCC; CRL-3216), human lung adenocarcinoma A549 cells (ATCC; CCL-185), human colorectal adenocarcinoma Caco-2 cells (ATCC; HTB-37) and African green monkey kidney Vero E6 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco; 11995-065) containing 10% fetal bovine serum (FBS) and 1× non-essential amino acids. A lentiviral delivery system was used to generate cells stably expressing human ACE2 and TMPRSS2. The mycoplasma-negative status of all cell lines was confirmed.

Anti-SARS-CoV N protein antibody (Rockland; 200-401-A50; 1:2,000) was used to detect the SARS-CoV-2 N protein by immunofluorescence and western blot. Mouse anti-SARS-CoV-2 spike antibody (GeneTex; GTX632604; 1:1,000), directed against the S2 subunit, was used for western blot analysis of S-protein cleavage in virus particles and infected cells.

Plasmids encoding various fragments of the SARS-CoV-2 genome (Hu/DP/Kng/19-020 isolate) were a gift from Y. Matsuura55. We replaced the S gene in the plasmid pCSII-SARS-CoV-2 F8 (ref. 55) with the chemically synthesized Omicron S gene and named this plasmid pCSII-SARS-CoV-2 F8_Omicron. We replaced the open reading frame (ORF) 7 in the plasmid pcDNA3.1-SARS-CoV-2 F9+10 with mNeonGreen or mCherry to obtain plasmids pcDNA3.1-SARS-CoV-2 F9+10_mNG and pcDNA3.1-SARS-CoV-2 F9+10_mCherry, respectively. The plasmids pMW-CoV-2-UTRlinker (ref. 55) and pGL-CPERlinker (ref. 24), both containing a linker fragment comprising hepatitis delta virus ribozyme (HDVr), the bovine growth hormone polyadenylation signal sequence (BGH-polyA) and cytomegalovirus (CMV) promoter, have been previously reported. The lentiviral vectors, pLOC_hACE2_PuroR and pLOC_hTMPRSS2_BlastR, containing human ACE2 and TMPRSS2, respectively, have been described previously56.


All procedures were performed in a state-of-the-art biosafety level 3 (BSL3) facility at the National Emerging Infectious Diseases Laboratories (NEIDL) of Boston University using biosafety protocols approved by the Institutional Biosafety Committee (IBC). The experimental plans, including the generation of recombinant chimeric viruses, were reviewed and approved by the IBC, which comprises scientists, biosafety and compliance experts and members of the local community. Furthermore, the research was approved by the Boston Public Health Commission. All personnel received rigorous biosafety, biosecurity and BSL3 training before participating in experiments. Personal protective equipment, including scrubs, disposable overalls, shoe covers, double-layered gloves and powered air-purifying respirators, was used. Biosecurity measures are built in the environment through building and security systems and are reinforced through required training programmes, standing meetings and emergency exercises. The researchers involved in working with chimeric viruses received at least two booster shots of the SARS-CoV-2 mRNA vaccine before the study was started. Finally, all researchers were medically cleared by the Boston University Research Occupational Health Program.

Collection of serum samples

Sera from individuals who received two doses of mRNA-1273 (Moderna) or BNT162b2 (Pfizer) vaccine were collected at Boston Medical Center at least two weeks after the final dose. These individuals had no prior history of SARS-CoV-2 infection. Serum samples were collected using protocols reviewed and approved by the Institutional Review Board at Boston Medical Center. All methods were performed in accordance with relevant guidelines and regulations. The participants provided electronic informed consent. De-identified samples were used in this research. Additional information for serum samples is provided in Extended Data Table 1.

Omicron stock preparation and titration

The SARS-CoV-2 BA.1 Omicron virus stock was generated in ACE2/TMPRSS2/Caco-2 cells. In brief, 5 × 105 cells, grown overnight in DMEM, 10% FBS and 1× NEAA in one well of a six-well plate, were inoculated with the collection medium in which the nasal swab from a patient with COVID-19 was immersed. The swab material was obtained from the Department of Public Health, Massachusetts, and it contained the sequence-verified Omicron virus (NCBI accession number: OL719310). Twenty-four hours after infecting cells, the culture medium was replaced with 2 ml of DMEM, 2% FBS and 1× NEAA and the cells were incubated for another 72 h, at which point the CPE became visible. The culture medium was collected, passed through a 0.45 µm filter and kept at −80 °C as a P0 virus stock. To generate a P1 stock, we infected 1 × 107 ACE2/TMPRSS2/Caco-2 cells, seeded the day before in a T175 flask, with the P0 virus at an MOI of 0.01. The next day, the culture medium was changed to 25 ml of 2% FBS-containing medium. Three days later, when the cells exhibited excessive CPE, the culture medium was collected, passed through a 0.45 µm filter and stored at −80 °C as a P1 stock.

To titrate the virus stock, we seeded ACE2/TMPRSS2/Caco-2 cells into a 12-well plate at a density of 2 × 105 cells per well. The next day, the cells were incubated with serial 10-fold dilutions of the virus stock (250 µl volume per well) for 1 h at 37 °C, overlaid with 1 ml per well of medium containing a 1:1 mixture of 2× DMEM/4% FBS and 1.2% Avicel (DuPont; RC-581) and incubated at 37 °C for another three days. To visualize the plaques, the cell monolayer was fixed with 4% paraformaldehyde and stained with 0.1% crystal violet, with both fixation and staining performed at room temperature for 30 min each. The number of plaques was counted and the virus titre was calculated.

Generation of recombinant SARS-CoV-2 by CPER

SARS-CoV-2 recombinant viruses were generated using a previously described optimized CPER protocol24. Full-length SARS-CoV-2 cDNA cloned into a bacterial artificial chromosome (BAC)23 was used to generate WT and Omi-S viruses. In brief, the BAC was amplified into eight overlapping fragments (F1, F2, F3, F4, F5, F6, F7 and F9) covering the whole SARS-CoV-2 genome. The pCSII-SARS-CoV-2 F8 (containing a D614G substitution) and pCSII-SARS-CoV-2 F8_Omicron plasmids, which were used to generate S mutants, served as templates for the amplification of fragment 8 (F8). The UTR linker plasmids pMW-CoV-2-UTRlinker (ref. 55) or pGL-CPERlinker (ref. 24) were used as a template to amplify the linker sequence. The 5′ termini of all ten DNA fragments (F1–F9 and the linker) were phosphorylated by using T4 PNK (NEB; M0201). The CPER reaction containing equimolar amounts (0.05 pmol) of each fragment was performed with PrimeStar GXL DNA polymerase (Takara Bio; R050A) as previously described24. The nicks in the circular product were sealed by using HiFi Taq DNA ligase (NEB; M0647S).

To generate chimeric viruses containing a combination of Omicron S and non-spike proteins (Omi-S/nsp3, Omi-S/nsp4, Omi-S/nsp5, Omi-S/nsp6, Omi-S/nsp14, Omi-S/E, Omi-S/M and Omi-S/N), we used SARS-CoV-2 plasmids described previously55 as templates, provided by Y. Matsuura. These plasmids contained SARS-CoV-2 sequences derived from the SARS-CoV-2/Hu/DP/Kng/19-020 strain. We introduced mutations into these plasmids using the standard DNA recombination technology. Our chimeric viruses also contained a P323L substitution in nsp12. Plasmid sequences were confirmed by the Sanger method.

To transfect cells with the CPER product, we seeded ACE2/TMPRSS2/Caco-2 cells into a six-well plate at a density of 5 × 105 cells per well. The transfection mix was prepared by mixing 26 µl of the original 52-µl CPER reaction volume with 250 µl of Opti-MEM (Thermo Fisher Scientific; 31985070) and 6 µl of TransIT-X2 Dynamic Delivery System (Mirus Bio; MIR 6000). After incubation at room temperature for 25 min, the transfection mix was added to the cells. The next day, the culture medium was replaced with fresh DMEM containing 2% FBS. The CPE became visible in three to four days, at which point the culture medium was collected and stored as a P0 virus stock. The P0 stock was used for experiments described in this manuscript. The sequence of CPER-generated viruses was confirmed by next-generation sequencing.

SARS-CoV-2 whole-viral sequencing and genome assembly

cDNA synthesis was performed using Superscript IV reverse transcriptase (Invitrogen). Whole-viral amplification was performed with the NEB Varskip protocol using multiplexed primer pools designed with Primal Scheme, generating 400-bp tiling amplicons. PCR products from the Varskip protocol were pooled together and Illumina library construction was performed using the Nextera XT Library Prep Kit (Illumina). Deep-sequencing data analysis was performed using the Stanford Coronavirus Antiviral & Resistance Database (CoV-RDB) platform57. Input FASTQ sequence alignment with the Wuhan-Hu-1 reference was done using MiniMap2 v.2.22 in the CodFreq pipeline ( The output of MiniMap2, an aligned SAM file, was converted to a CodFreq file by a Python script written in-house, using a PySam library (v.0.18.0), and was further analysed with the CoV-RDB. SARS-COV-2 variant calling was done using CoV-RDB Scorpio call v.1.2.123 (ref. 58) ( and Nextclade v.1.13.2 (ref. 59) ( PCR and sequencing run were performed once with the appropriate positive and negative controls.

SARS-CoV-2 neutralization assay

For neutralization assays, initial 1:10 dilutions of plasma, obtained from individuals who received two shots of either the Moderna or the Pfizer mRNA-based SARS-CoV-2 vaccine, were fivefold serially diluted in Opti-MEM over seven or eight dilutions. These plasma dilutions were then mixed at a 1:1 ratio with 1.25 × 104 infectious units of SARS-CoV-2 and incubated for 1 h at 37 °C. Thereafter, 100 µl of this mixture was directly applied to ACE2/A549 cells seeded the previous day in poly-l-lysine-coated 96-well plates at a density of 2.5 × 104 cells per well in a volume of 100 µl. Thus, the final starting dilution of plasma was 1:20 and the final MOI was 0.5. The cells were incubated at 37 °C for 24 h, after which they were fixed and stained with an anti-N antibody. When PBS instead of plasma was used as a negative control, these infection conditions resulted in around 40–50% infected cells at 24 hpi.

Generation and infection of iAT2 cells

The detailed protocol for generation of human iPS-cell-derived alveolar epithelial type II cells (iAT2 cells) has been published in our previous papers28,60. The ALI cultures were established by preparing single-cell suspensions of iAT2 three-dimensional (3D) sphere cultures grown in Matrigel. In brief, Matrigel droplets containing iAT2 spheres were dissolved in 2 mg ml−1 dispase (Sigma) and the spheres were dissociated in 0.05% trypsin (Gibco) to generate a single-cell suspension. Transwell inserts (6.5 mm; Corning) were coated with dilute Matrigel (Corning) in accordance with the manufacturer’s protocol. Single-cell iAT2 cells were plated on Transwells at a density of 520,000 cells per cm2 in 100 µl of CK+DCI medium containing 10 µM of rho-associated kinase inhibitor (“Y”; Sigma Y-27632). Six hundred microlitres of this medium was added to the basolateral compartment. Twenty-four hours after plating, the basolateral medium was changed with fresh CK+DCI+Y medium. Forty-eight hours after plating, the apical medium was aspirated to initiate the ALI culture. Seventy-two hours after plating, the basolateral medium was replaced with CK+DCI medium to remove the rho-associated kinase inhibitor. Basolateral medium was changed every two days thereafter. The detailed composition of CK+DCI medium is provided in our previous publications28,60.

iAT2 cells in ALI cultures were infected with purified SARS-CoV-2 stock at an MOI of 2.5 based on the titration that was done on ACE2/TMPRSS2/Caco-2 cells. For infection, 100 µl of inoculum prepared in 1× PBS (or mock-infected with PBS only) was added to the apical chamber of each Transwell and incubated for 2 h at 37 °C, followed by the removal of the inoculum and washing of the apical side three times with 1× PBS (100 µl per wash). The cells were incubated for two or four days, after which the newly released virus particles on the apical side were collected by adding 100 µl of 1× PBS twice to the apical chamber and incubating at 37 °C for 15 min. The number of infectious virus particles in the apical washes was measured by the plaque assay on ACE2/TMPRSS2/Caco-2 cells. For flow cytometry, iAT2 cells were detached by adding 0.2 ml Accutase (Sigma; A6964) apically, and incubated at room temperature for 15 min. The detached cells were pelleted by low-speed centrifugation, fixed in 10% formalin and stained with anti-SARS-CoV-2 N antibody.

Detection of S-protein incorporation and cleavage in SARS-CoV-2 particles

The culture medium of ACE2/TMPRSS2/Caco-2 cells transfected with the CPER product was collected and passed through 0.22-µm filters. The SARS-CoV-2 particles were pelleted down by mixing an equal volume of the culture medium with 20% PEG6000 in PBS followed by overnight incubation at 4 °C. The mixture was centrifuged at 12,000g for 30 min at 4 °C, and the pellet was resuspended in 1× SDS sample buffer. The protein concentration was measured by the BCA assay using Pierce BCA Protein Assay kit (Thermo Fisher Scientific; 23225). Equal amounts of protein were resolved on 4–12% SDS page. The S protein was detected with mouse anti-SARS-CoV-2 spike antibody (GeneTex; GTX632604; 1:1,000) and IRDye 800CW donkey anti-mouse IgG secondary antibody (LI-COR Biosciences; 926-32212; 1:5,000). The bands were visualized by scanning the membrane with the LiCor CLx infrared scanner. The open-source package ImageJ (v.1.53a) was used to measure the intensity of protein bands.

Flow cytometry

For flow cytometry, fixed cells were permeabilized in 1× permeabilization buffer (Thermo Fisher Scientific; 00-5523-00) and stained with SARS-CoV-2 nucleocapsid antibody (Rockland; 200-401-A50, 1:1,000), followed by donkey anti-rabbit IgG-AF647 secondary antibody (Thermo Fisher Scientific; A-31573). Cells infected with fluorescent reporter viruses were fixed and analysed without staining. Gating was based on uninfected, stained control cells. The extent of staining was quantified using a BD LSR II flow cytometer (BD Biosciences), and the data were analysed with FlowJo v.10.6.2 (FlowJo, Tree Star). The gating strategy for flow cytometry is shown in Supplementary Fig. 1.


Immunofluorescence was performed as described in our previous publication56. In brief, virus-infected cells were fixed in 4% paraformaldehyde and permeabilized in a buffer containing 0.1% Triton X-100 prepared in PBS. After blocking in a buffer containing 0.1% Triton X-100, 10% goat serum and 1% BSA, the cells were incubated overnight at 4 °C with anti-SARS-CoV N antibody (1:2,000 dilution). The cells were then stained with Alexa Fluor 568-conjugated goat anti-rabbit secondary antibody (1:1,000 dilution) (Invitrogen; A11008) in the dark at room temperature for 1 h and counterstained with DAPI. Images were captured using the ImageXpress Micro Confocal (IXM-C) High-Content Imaging system (Molecular Devices) with a 4× S Fluor objective lens at a resolution of 1.7 μm per pixel in the DAPI (excitation, 400 nm/40 nm; emission, 447 nm/60 nm) and TexasRed (excitation, 570 nm/80 nm; emission, 624 nm/40 nm) channels. Both channels were used to establish their respective laser autofocus offsets. The images were analysed using MetaXpress High Content Image Acquisition and Analysis software (Molecular Devices). First, the images were segmented using the CellScoring module. The objects between 7 μm and 20 μm in diameter and greater than 1,800 grey level units in intensity were identified and classified as nuclei. Positive cells were taken as nuclei having a TexasRed signal of 1,500 grey level units or above within 10–20 μm of each nucleus. The remaining objects were set to negative cells. From these objects, the following readouts were measured and used for downstream analysis: total number of positive and negative cells; total area of positive cells; and integrated intensity in the TexasRed channel for positive cells. To calculate the ND50, we performed a non-linear regression curve fit analysis using Prism 9 software (GraphPad).

Mice maintenance and approvals

Mice were maintained in a facility accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care (AAALAC). Animal studies were performed following the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocols were approved by the Boston University Institutional Animal Care and Use Committee (IACUC). Heterozygous K18-hACE2 C57BL/6J mice (Strain 2B6.Cg-Tg(K18-ACE2)2Prlmn/J) were purchased from the Jackson Laboratory (JAX). Mice were housed in groups of four or five in ventilated cages (Tecniplast) and maintained on a 12:12 light cycle at 30–70% humidity, 20 °C temperature, ad libitum water and ad libitum standard chow diet (LabDiet).

Mice infection

Male and female K18-hACE2 mice (12–20 weeks old) were inoculated intranasally with 104 PFU of SARS-CoV-2 in 50 µl of sterile 1× PBS. The inoculations were performed under 1–3% isoflurane anaesthesia. In vivo experiments were not blinded, and mice were randomly assigned to infection groups. No a priori sample size calculation was performed. Instead, samples sizes were determined on the basis of our previous animal work. Twenty-six mice (6 for WT, 10 for Omi-S and 10 for Omicron) were enrolled in a 14-day survival study, and another 42 mice (14 for each of the WT, Omi-S and Omicron viruses) were used for virological and histological analysis of infected lungs. During the survival study, the mice were monitored for body weight, respiration, general appearance, responsiveness and neurological signs. An IACUC-approved clinical-scoring system was used to monitor disease progression and define humane end-points. The score of 1 was given for each of the following situations: body weight, 10–19% loss; respiration, rapid and shallow with increased effort; appearance, ruffled fur and/or hunched posture; responsiveness, low to moderate unresponsiveness; and neurological signs, tremors. The sum of these individual scores constituted the final clinical score. Mice were considered moribund and humanly euthanized in case of weight loss greater than or equal to 20%, or if they received a clinical score of 4 or above for two consecutive days. Body weight and clinical score were recorded once per day for the duration of the study. For the purpose of survival curves, mice euthanized on a given day were counted dead the day after. Mice that were found dead in the cage were counted dead on the same day. For euthanasia, an overdose of ketamine was administered followed by a secondary method of euthanasia.

For the quantification of SARS-CoV-2 infectious particles in lungs by plaque assay, lung tissues were collected in 600 µl of RNAlater stabilization solution (Thermo Fisher Scientific; AM7021) and stored at −80 °C until analysis. Twenty to forty milligrams of tissue was placed in a tube containing 600 µl of Opti-MEM and a 5-mm stainless steel bead (Qiagen; 69989) and homogenized in the Qiagen TissueLyser II by two dissociation cycles (1,800 oscillations per min for 2 min) with a 1-min interval between cycles. The homogenate was centrifuged at 15,000g for 10 min at room temperature and the supernatant was transferred to a new tube. Tenfold serial dilutions of this supernatant were used for the plaque assay on ACE2/TMPRSS2/Caco-2 cells, as described above.

For IHC and histological analysis, the insufflated whole lung tissues were inactivated in 10% neutral buffered formalin at a 20:1 fixative-to-tissue ratio for a minimum of 72 h before removal from BSL3 in accordance with an approved IBC protocol. Tissues were subsequently processed and embedded in paraffin, and 5-μm sections were stained with H&E following standard histological procedures. IHC was performed using a Ventana BenchMark Discovery Ultra autostainer (Roche Diagnostics). An anti-SARS-CoV-2 S antibody (Cell Signaling Technologies; clone E5S3V) or anti-SARS-CoV-2 N antibody (Cell Signaling Technologies; clone 1C7C7) that showed equivalent immunoreactivity against WT and Omicron proteins was used to identify virus-infected cells. For the SARS-CoV-2 N antibody, given its mouse origin, an additional rabbit anti-mouse anti-Ig1 + Ig2a + IgG3 antibody (Abcam; 133469) was used to prevent non-specific binding. A HRP-conjugated goat anti-rabbit IgG polymer detection was then used to detect the viral specific antibodies (Vector Laboratories; MP-7451) and finally developed using 3,3′-diaminobenzidine (DAB) chromogen and counterstained with haematoxylin. Negative and positive controls for IHC included blocks of uninfected and SARS-CoV-2-infected Vero E6 cells, respectively.

For the quantification of N protein in the nasal turbinate epithelium, digitalized whole slide scans were analysed using the image analysis software HALO (Indica Labs). The respiratory epithelium was manually annotated to create a layer for downstream analysis. Area quantification was performed to determine the percentages of SARS-CoV-2 N in the annotated layer, which generated the percentage of immunoreactivity output.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.