The SARS-CoV-2 Lambda variant exhibits enhanced infectivity and immune resistance

SARS-CoV-2 Lambda, a variant of interest, has spread in some South American countries; however, its virological features and evolutionary traits remain unclear. In this study, we use pseudoviruses and reveal that the spike protein of the Lambda variant is more infectious than that of other variants due to the T76I and L452Q mutations. The RSYLTPGD246-253N mutation, a unique 7-amino acid deletion in the N-terminal domain of the Lambda spike protein, is responsible for evasion from neutralizing antibodies and further augments antibody-mediated enhancement of infection. Although this mutation generates a nascent N-linked glycosylation site, the additional N-linked glycan is dispensable for the virological property conferred by this mutation. Since the Lambda variant has dominantly spread according to the increasing frequency of the isolates harboring the RSYLTPGD246-253N mutation, our data suggest that the RSYLTPGD246-253N mutation is closely associated with the substantial spread of the Lambda variant in South America.


INTRODUCTION
During the current pandemic, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of coronavirus disease 2019 (COVID-19), has diversified. As of December 2021, there are 5 variants of concern (VOCs)-Alpha VOCs and VOIs harbor multiple mutations in their spike (S) protein and are relatively resistant to neutralizing antibodies (NAbs) that are elicited in convalescent and vaccinated individuals Collier et al., 2021;Garcia-Beltran et al., 2021;Hoffmann et al., 2021;Liu et al., 2021aLiu et al., , 2021bPlanas et al., 2021;Wall et al., 2021;Wang et al., 2021aWang et al., , 2021b. Because the receptor-binding domain (RBD) of the SARS-CoV-2 S protein is the major site that NAbs recognize, mutations in this domain can lead to immune evasion (Piccoli et al., 2020). In addition, mutations in the N-terminal domain (NTD) of the SARS-CoV-2 S protein are associated with neutralization escape (McCallum et al., 2021). Moreover, antibodies that enhance viral infectivity (enhancing antibodies [EAbs]) were detected in severe COVID-19 patients, and these EAbs target the NTD Liu et al., 2021c). Because natural mutations in the S NTD crucially influence the sensitivity to antibodies (Gobeil et al., 2021), the accumulation of mutations in this domain is closely associated with the infectious spread of VOCs and VOIs.
The Lambda variant (also known as the C.37 lineage) was classified as a VOI on June 14, (WHO, 2021a and has spread in South American countries such as Peru, Chile, Argentina, and Ecuador (WHO, 2021a). Based on the data from the Global Initiative on Sharing All Influenza Data (GISAID) database (https:// www.gisaid.org; as of June 29, 2021), the Lambda variant has been isolated in 26 countries. Notably, the vaccination rate in Chile is relatively high; the percentage of people who received at least one dose of COVID-19 vaccine was $60% on June 1, 2021 (https://ourworldindata.org/covid-vaccinations). A recent article also suggested that vaccines have effectively prevented COVID-19 in Chile (Jara et al., 2021). Nevertheless, a large COVID-19 surge occurred in Chile in spring 2021 (WHO, 2021b), suggesting that the Lambda variant can effectively escape from the antiviral immunity elicited by vaccination. In this study, we reveal the evolutionary traits of the Lambda variant by molecular phylogenetic analysis. We further demonstrate that the RSYLTPGD246-253N mutation, a unique mutation in the NTD of the Lambda S protein, is responsible for the virological phenotype of the Lambda variant that is associated with the major infectious spread mainly in South American countries.

Epidemic dynamics of the Lambda variant in South American countries
As of June 29, 2021, 1,908 genome sequences of the Lambda variant belonging to the PANGO C.37 lineage have been isolated from 26 countries and deposited in GISAID. Although the Lambda variant was first reported in Peru in December 2020 (WHO, 2021a), our in-depth analysis revealed that this variant was first detected in Argentina on November 8, 2020 (GISAID: EPI_ISL_2158693) ( Figure 1A; see STAR Methods for details). The prevalence of the Lambda variant is increasing in South American countries, including Peru, Chile, and Argentina (Figure 1A and Table S1), suggesting that this variant is spreading predominantly in these countries (Table S2). We then generated a maximum likelihood-based phylogenetic tree of the Lambda variant ( Figure S1A). Although there were some isolates that were misclassified as Lambda, which could be the ''sister group of Lambda variants,'' our phylogenetic tree indicated the monophyly of the ''genuine Lambda variants'' isolates ( Figure S1A).
Association of the Lambda variant spread with the increasing frequency of isolates harboring the RSYLTPGD246-253N mutation In the 1,908 sequences of the Lambda variant (C.37 lineage), the majority of the S protein sequence of the genuine Lambda variants contains 6 substitution mutations (G75V, T76I, L452Q, F490S, D614G, and T859N) and a 7-amino acid deletion in the NTD (RSYLTPGD246-253N) ( Figure S1A). The Lambda variant first isolated in Argentina contains the 6 amino acid mutations as well as the deletion in the S protein. However, the sister group of Lambda variants does not contain any of the 7 mutations, excluding D614G in their S proteins ( Figure S1A).
We then analyzed the mutations in the S protein using the 1,908 sequences of the Lambda variant (C.37 lineage) and showed that the 6 amino acid mutations are highly (>90%) conserved ( Figure 1B). Although a large deletion in the NTD, the RSYLTPGD246-253N mutation, is also highly conserved, and only 287 of the 1,908 sequences (15.0%) of the Lambda variant (C.37 lineage) genomes do not harbor this mutation (Figures 1B and S1A). To determine whether the epidemic dynamic of the Lambda variant is associated with the emergence of the RSYLTPGD246-253N mutation, we examined all of the amino acid changes in the S protein of the SARS-CoV-2 genomes deposited in the GISAID database (2,084,604 sequences; as of June 29, 2021). Our analysis showed that the RSYLTPGD246-253N mutation was first found in Argentina on November 8, 2020, which is the first isolate of the Lambda variant ( Figure 1A), suggesting that this deletion event uniquely occurred in the ancestral lineage of the genuine Lambda variants. We then analyzed the molecular evolutionary dynamics of the Lambda variants by performing Bayesian tip-dating analysis. We showed that the estimated time to the most recent common ancestor of the genuine Lambda variants was approximately July 12, 2020 (95% highest posterior density [HPD], January 5, 2020-October 22, 2020) (Figures 1C and S1B). To infer the population dynamics of the lineage, we performed a Bayesian skyline plot analysis. This analysis showed that the effective population size of the Lambda variant increased at the beginning of 2021 ( Figure 1D). Intriguingly, when we plotted the proportion of the Lambda variant with the RSYLTPGD246-253N mutation, we found that it increased after the emergence of the Lambda variant and closely associated with the increase in effective population size ( Figure 1D). These results suggest that the acquisition of the RSYLTPGD246-253N mutation is associated with the outbreak of the Lambda variant in South America.

Enhanced infectivity and resistance to NAbs of the Lambda variant
To address the virological phenotype of the Lambda variant (hereafter, only genuine Lambda variant is referred to as ''Lambda variant''), we prepared viruses pseudotyped with the S proteins of the Lambda variant as well as 4 VOCs, Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1) and Delta (B.1.617.2) (Figure S1C). We also prepared pseudoviruses with the S protein of the Epsilon (B.1.427/429) variant ( Figure S1C), which was used as a VOC/VOI and was reclassified on July 6, 2021 (WHO, 2021a). As the control of this experiment, the pseudoviruses with the S protein of the D614G-bearing isolate (B.1 lineage), which spread in the early pandemic (i.e., during early 2020 and before the emergence of VOCs/VOIs) , were prepared, and this is referred to as ''parental virus'' in this study. As shown in Figure 2A, the infectivities of the Alpha and Beta variants were significantly lower than that of the parental virus (p = 0.0012, Alpha versus parental; p = 0.0012, Beta versus parental by Student's t test), and the infectivities of the Gamma variant and the parental virus were comparable. Since the epidemic of SARS-CoV-2 in the early pandemic, which includes the B.1 lineage (parental virus in this study), was outcompeted by VOCs including the Alpha and Beta variants during late 2020 (WHO, 2021a), it was unexpected that the infectivities of Alpha and Beta variants were lower than that of the parental virus (Figure 2A). However, the lower infectivity of the Alpha variant compared to the D614G-bearing early pandemic virus, analyzed by pseudovirus assay (Figure 2A), is consistent with a previous report . In contrast, the infectivities of the Delta, Epsilon, and Lambda variants were significantly higher than that of the parental virus (  Wuhan- 532,199), the oldest SARS-CoV-2 (isolated on December 26, 2019), is indicated in red. Bars on the internal nodes correspond to the 95% HPD. The tree noted with the GISAID ID and sampling date at each terminal node is shown in Figure S1B. (D) Transition of the effective population size of the Lambda variant and the proportion of the Lambda variant harboring the RSYLTPGD246-253N mutation. The effective population size of the Lambda variant harboring the RSYLTPGD246-253N mutation (left y axis) was analyzed by a Bayesian skyline plot. The initial date was when the first Lambda variant was sampled (November 8, 2020). The 95% HPD is shaded in brown. In the same panel, the proportion of the Lambda variants harboring the RSYLTPGD246-253N mutation for each month (right y axis) is also plotted. The number at each time point indicates the number of Lambda variants harboring the RSYLTPGD246-253N mutation. The number in parentheses indicates the number of Lambda variants deposited in the GISAID database. See also Figure S1 and Tables S1, S2, and S3.
Cell Reports 38, 110218, January 11, 2022 3 Report ll OPEN ACCESS  Figure S1C. The pseudoviruses were inoculated in HOS-ACE2/TMPRSS2 cells at 1,000 ng HIV-1 p24 antigen, and the percentages of infectivity compared to that of the virus pseudotyped with parental S are shown. (B) Neutralization assay. A neutralization assay was performed using the pseudoviruses with the S proteins of the parental, Lambda, and Lambda + N246-253RSYLTPGD and 18 BNT162b2-vaccinated sera as described in STAR Methods. The raw data are shown in Figure S2B. The number in the panel indicates the fold change in neutralization resistance of Lambda S to the parental S or the Lambda + N246-253RSYLTPGD derivative. (C and D) Structural insights of the mutations in Lambda S. (C) Overlaid overviews of the crystal structure of SARS-CoV-2 S (PDB: 6ZGE, white) (Wrobel et al., 2020) and a homology model of Lambda S (brown) are shown. The mutated residues in the Lambda S and the regions in the NTD and RBD are indicated in red and blue. The squared regions are enlarged in (D). Mutated residues in the NTD (left) and RBD (right) of Lambda S. The residues in parental S and the Lambda S are indicated in red and black. (E) Pseudovirus assay. HIV-1-based reporter viruses pseudotyped with SARS-CoV-2 S proteins bearing respective mutations of the Lambda variant in parental S (G75V, T76I, GT75-76VI, RSYLTPGD246-253N, L452Q, F490S, L452Q/F490S, or T859N) as well as the parental S were prepared. The pseudoviruses were inoculated into HOS-ACE2/TMPRSS2 cells at 1,000 ng HIV-1 p24 antigen, and the percentages of infectivity compared to that of the virus pseudotyped with parental S are shown. (F) Binding affinity of the SARS-CoV-2 S RBD (residues 336-528) to ACE2 by yeast surface display. The K D values of the binding of the SARS-CoV-2 S RBD expressed on yeast to soluble ACE2 are shown. (G) CTL assay. HLA-A24 + CTL lines established from 6 BNT162b2-vaccinated donors were stimulated with 1 nM NF9 peptide or its derivatives, NF9-L452Q (NYNYQYRLF) or NF9-L452R (NYNYRYRLF). The percentage of IFN-g + cells in CD8 + T cells (bottom, n = 6) is shown. Representative fluorescence-activated cell sorting (FACS) plots showing intracellular expression of IFN-g in the CD8 + T cell subset of a vaccinated donor (top, donor GV36-2) are shown in Figure S2H. (H) Neutralization assay. A neutralization assay was performed using the pseudoviruses used in Figure 2B and 18 BNT162b2-vaccinated sera, as described in STAR Methods. The raw data are shown in Figure S2B. The number in the panel indicates the fold change in neutralization resistance to the parental S.
(legend continued on next page) 4 Cell Reports 38, 110218, January 11, 2022 Report ll OPEN ACCESS parental; p = 0.0031, Epsilon versus parental; p = 0.0004, Lambda versus parental by Student's t test). This pattern was independent of the input dose of pseudoviruses used ( Figure S2B).
To assess the effect of a characteristic mutation of the Lambda variant, the RSYLTPGD246-253N mutation (Figure 1), on viral infectivity, we prepared a pseudovirus with the Lambda S derivative with the recovery of this deletion (Lambda + N246-253RSYLTPGD) ( Figure S2A). The infectivity of this mutant was comparable to that of the Lambda S (Figures 2A and S2B). To further verify the effect of the N246-253RSYLTPGD mutation on the infectivity of the Lambda variant, we prepared the fluorescent protein-expressing lentivirus vectors pseudotyped with Lambda S (expressing Crimson) and Lambda + N246-253RSYLTPGD S (expressing ZsGreen1). The competition assay using these fluorescent protein-expressing pseudoviruses showed that both viruses exhibit similar infectivities ( Figure S2C), suggesting that the 7-amino acid deletion in the NTD does not affect viral infection.
Because NAb resistance is a notable phenotype of most VOCs (reviewed in , we next analyzed the sensitivity of Lambda S to NAbs induced by BNT162b2 vaccination (Table  S4). As shown in Figure 2B, Lambda S was, on average, 1.5-fold (2.63-fold at a maximum) more resistant to BNT162b2-induced antisera than the parental S (p = 0.0077 by the Wilcoxon matched-pairs signed rank test). However, the neutralization level of Lambda + N246-253RSYLTPGD was similar to that of the parental S pseudovirus (P = 0.21 by the Wilcoxon matched-pairs signed rank test), and this recovered variant was 1.37-fold more sensitive to vaccine-induced neutralization than the Lambda variant (P = 0.016 by the Wilcoxon matchedpairs signed rank test) ( Figures 2B and S2D). These results suggest that Lambda S is highly infectious and resistant to vaccine-induced humoral immunity, and the robust resistance of Lambda S to vaccine-induced neutralization is determined by a 7-amino acid deletion in the NTD.

Effect of consensus mutations in Lambda S on viral infectivity
We next examined 6 substitution mutations (G75V, T76I, L452Q, F490S, D614G, and T859N) and a deletion mutation (RSYLTPGD246-253N) of the Lambda variant in the structure of the SARS-CoV-2 S protein. Structural analysis showed that the 3 mutations, G75V, T76I, and RSYLTPGD246-253N, are in the NTD ( Figure 2C), and the RSYLTPGD246-253N mutation is located in a loop structure, which was designated as loop 5 (residues 246-260) in a previous study (Chi et al., 2020) (Figure 2D). The G75V and T76I mutations are located at the flexible loop adjacent to loop 5 ( Figure 2C). The L452Q and F490S mutations are situated in the receptor-binding motif in the RBD ( Figures 2C and 2D), but neither residue is in direct contact with the angiotensin-converting enzyme 2 (ACE2) receptor (Figure S2E). The T859N mutation is in the heptad repeat 1 of the S2 subunit ( Figure 2C).
To investigate the effects of these seven consensus mutations in Lambda S on viral infectivity and NAb sensitivity, we prepared viruses pseudotyped with parental S-based derivatives that have mutations of the Lambda variant. Figure 2E shows that the G75V mutation significantly reduces viral infectivity (p = 0.0004, by Student's t test), while the T76I and GT75-76VI mutations significantly increase viral infectivity (p = 0.0011, T75I versus parental; p = 0.012, GT75-76TI versus parental by Student's t test). In addition, 91.5% (1,746/1,908) of the Lambda variant sequences had these 2 mutations, and the phylogenetic tree of the Lambda variant indicated that the variant harboring either G75V or T76I sporadically emerged during the epidemic of the Lambda variant ( Figure S1A). However, there were no viral sequences harboring the G75V mutation without the T76I mutation ( Figure S1A). These findings suggest that T76I is a compensatory mutation to recover the decreased infectivity by the G75V mutation.
Similar to the results from the experiment using pseudoviruses with Lambda S and Lambda + N246-253RSYLTPGD (Figure 2A), the insertion of the RSYLTPGD246-253N mutation did not affect viral infectivity ( Figure 2E). The infectivity of the T859N mutation was also similar to that of the parental pseudovirus ( Figure 2E). When we focused on the effect of the mutations in the RBD, the L452Q and L452Q/F490S mutations significantly increased viral infectivity (p = 0.0011, L452Q versus parental; p = 0.0003, L452Q/F490S versus parental by Student's t test), while the F490S mutation did not ( Figure 2E). In addition, the binding affinity of the SARS-CoV-2 RBD to soluble ACE2 was significantly increased by the insertion of the L452Q mutation, but not the F490S mutation ( Figure 2F: p = 0.0004, L452Q versus parental; p = 0.92, F490S versus parental; p = 0.0002, L452Q/F490S versus parental by Student's t test). These results suggest that the T76I and L452Q mutations are responsible for the higher infectivity of Lambda S ( Figure 2E). The effect of each mutation was similar with different amounts of pseudoviruses used and in the target cells without TMPRSS2 expression ( Figure S2F). In (G), statistically significant differences (*p < 0.05) versus the NF9 peptide were determined by the Wilcoxon matched-pairs signed rank test. In (I) and (K), vertical dashed lines indicate the MFIs of mock-transfected cells. In (J), statistically significant differences (*p < 0.05) versus the value without the antibody were determined by Student's t test. NS, no statistical significance. In (L), statistically significant differences versus the value without the antibody (*p < 0.05) and the parental S (#p < 0.05) were determined by Student's t test. See also Figure S2 and Tables S4 and S5. Cell Reports 38, 110218, January 11, 2022 5 Report ll OPEN ACCESS

Evasion of vaccine-induced cellular immunity induced by the L452Q mutation
Previous studies showed that the 9-mer peptide spanning the 448-456 residues of the SARS-CoV-2 S protein (designated NF9) is an immunodominant epitope presented by human leukocyte antigen (HLA)-A24 in convalescent individuals with SARS-CoV-2 infection Hu et al., 2021;Kared et al., 2021). Interestingly, we recently demonstrated that the L452R mutation, a hallmark mutation in the Delta and Epsilon variants, could induce evasion of the antiviral effects triggered by HLA-A24-restricted cytotoxic T lymphocytes (CTLs) . To determine whether the L452Q mutation in the Lambda variant can contribute to evasion of HLA-A24restricted anti-SARS-CoV-2 cellular immunity, as in the case of the L452R mutation , we obtained peripheral blood mononuclear cells (PBMCs) from HLA-A*24:02 + individuals vaccinated with BNT162b2 and stimulated these cells with the NF9 peptide (Table S5). As shown in Figure S2G, the NF9 peptide activated the PBMCs from HLA-A*24:02 + vaccinated individuals, but not those from HLA-A*24:02 À vaccinated individuals or HLA-A*24:02 + unvaccinated individuals, suggesting that the NF9 peptide is an immunodominant epitope for PBMCs from HLA-A*24:02 + BNT162b2-vaccinated individuals. We then assessed whether the L452Q mutation evades the HLA-A24-restricted cellular immunity induced by BNT162b2 vaccination. Consistent with our previous study , the NF9 peptide significantly induced interferon-g (IFN-g) production ( Figures  2G and S2H: p = 0.031 by the Wilcoxon matched-pairs signed rank test). In sharp contrast, 2 substitutions of the NF9 peptide, NF9-L452R and NF9-L452Q, faintly activated the NF9specific CTLs (Figures 2G and S2H). These results suggest that the NF9 peptide is an immunodominant epitope in HLA-A*24:02 + vaccinated individuals and that the L452Q mutation potentially results in evasion of HLA-A24-restricted cellular immunity.

Evasion of vaccine-induced humoral immunity
We next assessed the sensitivity of these pseudoviruses with the mutated S proteins to BNT162b2-induced antisera. As shown in Figure 2H, the G75V, T76I, GT75-76VI, and T859N mutations did not affect vaccine-induced neutralization. In contrast, the RSYLTPGD246-253N mutant exhibited significant resistance to the vaccine-induced neutralization ( Figure 2H: p = 0.027 by the Wilcoxon matched-pairs signed rank test), which is relevant to the experiment with the S proteins of the Lambda variant and the Lambda + N246-253RSYLTPGD derivative ( Figure 2B). In addition, we found that the L452Q and F490S mutations confer resistance to vaccine-induced antisera ( Figure 2H). The finding that the F490S mutation does not affect viral infectivity ( Figure 2E) but confers the resistance to the vaccine-induced antisera (Figure 2H) suggests that this mutant has acquired resistance to antiviral humoral immunity. In contrast, the L452Q mutation not only increases viral infectivity ( Figure 2E) and affinity to ACE2 (Figure 2F) but also augments resistance to the vaccine-induced antiviral cellular ( Figure 2G) and humoral immunities ( Figure 2H), suggesting that this mutation is critical for the viral dissemination in the human population.

Effects of NTD-targeting monoclonal antibodies on Lambda variant infection
To further assess the effect of the mutations in Lambda S, particularly those in the NTD, we used 2 monoclonal antibodies that target the NTD: an NTD-targeting NAb, clone 4A8 , and an EAb, clone COV2-2490, that recognizes the NTD and enhances viral infectivity . As shown in Figures 2I and 2J, the 4A8 antibody detected the cell surface expression of S proteins of the parental virus, Lambda + N246-253RSYLTPGD derivative, G75V, T76I, and GT75-76VI and inhibited the infections of these pseudoviruses in a dosedependent manner. Intriguingly, the 4A8 antibody failed to detect the surface expression of the S proteins of the Lambda and the RSYLTPGD246-253N mutant ( Figure 2I), and the pseudoviruses with the S proteins of the Lambda and the RSYLTPGD246-253N mutant were resistant to the antiviral effect mediated by the 4A8 antibody ( Figure 2J). Because a polyclonal antibody targeting the SARS-CoV-2 S protein detected these mutants, similar to the parental S ( Figure S2I), our results suggest that the RSYLTPGD246-253N mutation critically affects the sensitivity to certain NAbs targeting the NTD, such as the 4A8 antibody.
For the COV2-2490 antibody, flow cytometry analysis showed that this antibody more efficiently detects the Lambda S than the parental S ( Figure 2K). Although this antibody enhanced the infectivities of all of the pseudoviruses tested ( Figure 2L), it is important to note that the infectivity of the Lambda S was more significantly enhanced than that of the parental S (Figure 2L: p = 0.032 at 0.010 mg/mL; p = 0.013 at 0.1 mg/mL; p = 0.024 at 1.0 mg/mL by Student's t test). These data suggest that Lambda S is more susceptible to EAb-mediated enhancement of viral infection. Because an apparent enhancement of viral infectivity compared to the parental S pseudovirus was not observed in the substituted mutants (G75V, T76I, GT75-76VI, and RSYLTPGD246-253N) ( Figure 2L), our data suggest that multiple mutations in the NTD of Lambda S are crucial for the augmented enhancement of viral infectivity. Although the surface detection level of the Lambda variant was comparable to those of the Lambda + N246-253RSYLTPGD derivative and other NTD mutants tested (T76I, GT75-76VI, and RSYLTPGD246-253N) ( Figure 2K), the augmented enhancement of viral infectivity was specific for the Lambda S ( Figure 2L). These observations suggest that the detection level of the S protein on the cell surface by flow cytometry does not necessarily reflect the augmentation level of viral infection by the COV2-2490 antibody, and further actions of the S protein following the binding of the COV2-2490 antibody, such as the Lambda S-specific conformational change, may be needed to lead to the augmented enhancement of viral infectivity.

Association of a nascent N-linked glycosylation site (NLGS) by the RSYLTPGD246-253N mutation with sensitivity to NTD-targeting antibodies
We revealed that Lambda S is more resistant to vaccineinduced antiviral humoral immunity ( Figures 2B and 2H) as well as an NTD-targeting NAb ( Figure 2J) and that the RSYLTPGD246-253N mutation is responsible for NAb resistance. In addition, the sensitivity of Lambda S to EAb was higher than that of the parental S and was also attributed to the multiple mutations in the NTD, including the RSYLTP-GD246-253N mutation ( Figure 2L). These data suggest the importance of the RSYLTPGD246-253N mutation in determining the virological phenotype of Lambda S; however, it remains unclear how this large mutation affects sensitivity to neutralizing/enhancing antibodies. When we further assessed the amino acid sequences of the S proteins of the parental and Lambda variant, we found that the RSYLTPGD246-253N mutation generates an NLGS (N-X-S/T consensus sequence) de novo ( Figure 3A). Because emergent N-linked glycans in viral spike proteins, such as HIV-1 envelope protein (Li et al., 2008;Quinones-Kochs et al., 2002;Reitter et al., 1998;Wagh et al., 2020), and influenza A virus hemagglutinin protein (Abe et al., 2004;Kobayashi and Suzuki, 2012;Skehel et al., 1984;Wang et al., 2009), can shield the viral particles from NAbs, we hypothesized that this property of Lambda S is attributed to the nascent N-glycan conferred by the RSYLTPGD246-253N mutation. To address this possibility, we prepared two Lambda S derivatives, N246A and N246Q, which abolished the NLGS ( Figure 3A). We also prepared two parental S RSYLTPGD246-253N mutant-based derivatives, RSYLTPGD246-253A and RSYLTPGD246-253Q (Figure 3A). As shown in Figure 3B, the infectivities of the parental S RSYLTPGD246-253N mutant-based pseudoviruses (p < 0.0001, RSYLTPGD246-253A versus parental; p < 0.0001, RSYLTPGD246-253Q versus parental by Student's t test) but not that of the Lambda S-based pseudoviruses (p = 0.081, RSYLTPGD246-253A versus Lambda; p = 0.26, RSYLTPGD246-253Q versus Lambda by Student's t test) were significantly increased by the NLGS disruption. However, NLGS disruption did not affect the sensitivity to vaccineinduced NAbs ( Figure 3C). As in the cases of the Lambda S  Figure 2I), flow cytometry analysis showed that the 4A8 antibody faintly detected the NLGS-disrupted mutants on the cell surface ( Figure 3D). In addition, a neutralization assay revealed that the NLGS disruption did not affect the sensitivity to the 4A8 antibody ( Figure 3E). For an NTD-targeting EAb, the COV2-2490 antibody, the detection on the cell surface by flow cytometry ( Figure 3F), and the enhancement of the pseudovirus infectivity ( Figure 3G) were comparable between the NLGSdisrupted mutants and the Lambda S and the parental S RSYLTPGD246-253N mutants. These results suggest that the addition of an NLGS does not affect the sensitivity of Lambda S to the antibodies targeting the NTD.

DISCUSSION
In this study, we demonstrated that 3 mutations, the RSYLTP-GD246-253N, L452Q, and F490S mutations, confer resistance to vaccine-induced antiviral humoral immunity. In addition, the L452Q mutation allows evasion of HLA-A24-restricted cellular immunity. Moreover, the T76I and L452Q mutations contribute to enhanced viral infectivity. Our investigations using pseudoviruses suggest at least 2 virological features of the S protein of the Lambda variant: increasing viral infectivity (by the T76I and L452Q mutations) and exhibiting resistance to antiviral immunity (by the RSYLTPGD246-253N, L452Q, and F490S mutations).
Molecular phylogenetic analyses showed that the acquisition of a 7-amino acid deletion, the RSYLTPGD246-253N mutation, in the Lambda variant is closely associated with the increase in the effective population size of this variant. Here, we demonstrated that the RSYLTPGD246-253N mutation does not affect viral infectivity, but is responsible for resistance to vaccineinduced neutralization as well as an NTD-targeting NAb (4A8). Although the RSYLTPGD246-253N mutation generates a nascent NLGS, the addition of N-linked glycan did not affect NAb resistance in our assays. Notably, the RSYLTPGD246-253N mutation overlaps with a component of the NTD ''supersite'' . Our results suggest that the 7-amino acid deletion rather than the addition of the asparagine residue leading to the formation of a new NLGS confers NAb resistance and further supports the possibility that the NTD supersite is immunodominant and the mutations in this site are closely associated with the efficacy of vaccine-induced neutralization.
It has been reported that the SARS-CoV-2 S forms some conformational patterns, and the opened RBD state exposes the ACE2 binding interface Plante et al., 2021;Yurkovetskiy et al., 2020). Also, Liu et al. (2021c) revealed that the binding of EAbs, including COV2-2490, to the SARS-CoV-2 S NTD induces the conformation of the opened RBD state leading to enhanced binding to ACE2. Here, we demonstrated that multiple mutations in the NTD, including the RSYLTP-GD246-253N mutation, contribute to the augmented enhancement of viral infection by an EAb (COV2-2490). Because the epitope of COV2-2490 does not overlap the mutations in the NTD of Lambda S (G75V, T76I, and RSYLTPGD246-253N) , these mutations would not directly affect the binding affinity of this antibody to S. Rather, these mutations may indirectly increase the affinity of this antibody to S through the conformational change of NTD overall. Our results suggest that the emergence of the RSYLTPGD246-253N mutation was a driving force of the spread of this variant in the human population, possibly due to the acquisition of 2 antibody-associated properties, resistance to NAbs and enhanced benefits from EAbs.
We showed that the infectivity of the viruses pseudotyped with the Lambda S is significantly higher than that of the parental S. In addition, consistent with our previous reports (Mlcochova et al., 2021;Motozono et al., 2021), the infectivity of the viruses pseudotyped with the S proteins of the Delta and Epsilon variants was significantly higher than that of the parental S. A common feature of the Lambda, Delta, and Epsilon variants is a substitution in L452 of the SARS-CoV-2 S protein: the Lambda variant harbors the L452Q mutation, while the Delta and Epsilon variants possess the L452R mutation (Mlcochova et al., 2021;Motozono et al., 2021). Here, we demonstrated that the L452Q mutation increases viral infectivity. Together with our previous observation that the L452R mutation enhances viral infectivity , these results strongly suggest that the relatively higher infectivity of the Lambda, Delta, and Epsilon variants is attributed to the L452Q/R mutation. The fact that the Lambda and Delta variants are currently a VOI and a VOC, respectively, suggests that their increasing spread in the human population is partly attributed to their higher infectivity than that of the parental SARS-CoV-2. In contrast, despite its higher infectivity, the Epsilon variant, an ex-VOC, was excluded from the VOC/ VOI classification on July 6, 2021 because this variant was stamped out (WHO, 2021a). The transient and unsuccessful (compared to the other VOCs) spread of the Epsilon variant in the human population implies that increasing viral infectivity is insufficient to maintain efficient spread in the human population. In addition to increased viral infectivity, the Delta variant exhibits higher resistance to vaccine-induced neutralization (Mlcochova et al., 2021;Wall et al., 2021). Similarly, here, we showed that the Lambda variant shows increased infectivity and resistance against antiviral immunity. These observations suggest that acquiring at least two virological features, increased viral infectivity and evasion from antiviral immunity, is pivotal for efficient spread and transmission in the human population.
We showed that the Lambda S is more resistant to the vaccine sera than the parental S ( Figure 2B). By the loss-of-function experiment, we also showed that this immune resistance was canceled by the reversion of the NTD deletion (RSYLTPGD246-253N mutation) in the context of the Lambda variant ( Figure 2B), suggesting that the immune resistance of Lambda variant is mostly dependent primarily on the RSYLTPGD246-253N mutation located in NTD. However, in the gain-of-function experiment based on the parental S ( Figure 2H), the 3 mutations detected in the Lambda variant, RSYLTPGD246-253N, L452Q, and F490S, can lead to the resistance to vaccine sera. The discrepancy between these 2 experiments can be explained by the context of S proteins used as the backbone. The gain-of-function experiment (Figure 2H) suggests the potential of the RSYLTPGD246-253N, L452Q, and F490S mutations to confer resistance to vaccine sera, when the backbone is parental S protein. However, the loss-of-function experiment ( Figure 2B) suggests that the RSYLTPGD246-253N mutation determines the resistance to 8 Cell Reports 38, 110218, January 11, 2022 Report ll OPEN ACCESS vaccine sera, when the backbone is the Lambda S. Because the Lambda S contains the other mutations compared to the parental S, our data suggest that the resistance to the vaccine sera (i.e., polyclonal antibodies elicited by vaccination) is dependent on the context of S protein sequence and/or structure. Similarly, the context-dependent effect by the mutation insertion was also observed in the pseudovirus infection assay ( Figure 3B); the RSYLTPGD246-253A/Q mutation increased the infectivity of the parental RSYLTPGD246-253N-based S but did not increase that of the Lambda S. Our findings suggest that the effects of mutations on viral infectivity and immune resistance should be carefully evaluated based on the context of S protein used as the backbone.
In addition to the aforementioned virological property of the Lambda S via the mutations in its NTD, we revealed that a mutation in its RBD, L452Q, confers 2 virological properties: increase in viral infectivity and evasion from the HLA-A24-restricted cellular immunity. These are reminiscent of our recent study showing that the L452R mutation confers similar virological phenotypes . In particular, regarding the latter phenotype, it is possible that the immune pressure elicited by the HLA-A24 + individuals living in South American countries, where the Lambda variant mainly spreads, was the driving force in the emergence of this mutation. In fact, the allele frequencies of HLA-A24 in the populations of South American countries such as Peru, Chile, and Argentina, where the Lambda variant has been widespread ( Figure 1A), were 9.1%, 11.1%, and 10.7% for people in Peru (n = 251), Chile (n = 920), and Argentina (n = 2,516), respectively (Gonzalez-Galarza et al., 2020). Assuming that the Hardy-Weinberg law is valid, the frequencies of the people who have at least 1 HLA-A24 allele are 17.5% (Peru), 21.0% (Chile), and 20.2% (Argentina), respectively, which are relatively higher numbers considering the diversity of HLA alleles. Therefore, it would be reasonable to assume that the L452Q mutation that emerged via the HLA-A24-mediated immune pressure and the acquisition of this mutation to evade HLA-A24-mediated cellular immunity may further contribute to the relatively higher prevalence of the Lambda variant in these countries. Gobeil et al., 2021 suggested that mutations in the S NTD drive viral transmission and escape from antiviral immunity. In fact, three of the current four VOCs harbor deletions in the S NTD (reviewed in Harvey et al. [2021]). The Alpha variant has a 2-amino acid deletion, the HV69-70 deletion, in the NTD. Meng et al. (2021) showed that the HV69-70 deletion does not affect NAb sensitivity but does increase viral infectivity, suggesting that the virological importance of the deletion of a portion of the NTD between the Alpha (the HV69-70 deletion) and the Lambda variants (the RSYLTPGD246-253N mutation) is different. A cluster of the Beta variant also possesses a 3-amino acid deletion, the LAL242-244 deletion. This deletion does not critically affect sensitivity to vaccine-induced neutralization but results in resistance to some NTD-targeting NAbs, such as 4A8 Wang et al., 2021b). Interestingly, the Delta variant (B.1.617.2 lineage), a VOC, harbors a 2-amino acid deletion, the EFR156-158G mutation, while its relative VOI, the Kappa variant (B.1.617.1 lineage), does not (WHO, 2021a). Although the virological importance of the EFR156-158G mutation re-mains unclear, this mutation in the S NTD may be associated with the spread of the Delta variant worldwide. Together with our findings that RSYLTPGD246-253N confers resistance to vaccine-induced antisera and an NTD-targeting NAb, 4A8, the accumulating mutations in the S NTD may be closely associated with the virological features of the variants that can explain their spreading efficacy in the human population. In particular, 4A8 targets the NTD supersite Lok, 2021) that includes RSYLTPGD246-253 . Because the RSYLTPGD246-253N mutation is responsible for the resistance to the vaccine-induced antiviral effect of the Lambda variant, our data indicate that the NTD supersite is immunodominant and that acquiring immune escape mutations in this region is associated with the efficacy of viral dissemination in the human population, which has been proposed in previous reports McCallum et al., 2021). Moreover, we revealed that an EAb, COV2-2490, preferentially enhances Lambda Smediated infection. Although this enhancing effect is not specifically attributed to the RSYLTPGD246-253N mutation, such enhancement may be associated with the potential spread of the Lambda variant.
Through molecular phylogenetic analyses and virological experiments, we elucidated how the Lambda variant originated and acquired its virological properties. Because the Lambda variant is a VOI, it may be believed that this variant is not an ongoing threat compared to pandemic VOCs. However, because the Lambda variant is relatively resistant to vaccineinduced antisera, this variant may cause breakthrough infections (Hacisuleyman et al., 2021;Jacobson et al., 2021;Nixon and Ndhlovu, 2021;Rana et al., 2021). Moreover, elucidating the evolutionary traits of threatening SARS-CoV-2 variants can help predict the possibility of a wider epidemic, and revealing the virological features of the respective mutations in VOCs and VOIs is essential to assess the risk of newly emerging SARS-CoV-2 variants in the future.

Limitations of the study
The experiments using pseudoviruses showed that the Lambda variant is more infectious than the parental virus ( Figure 2A). However, it remains unaddressed whether the live SARS-CoV-2 Lambda variant is more replicative than the live parental SARS-CoV-2 (B.1 lineage). Although our data showed that the Alpha pseudovirus is less infectious than the parental (B.1 lineage) pseudovirus, which is consistent with a previous study , Touret et al. (2021) showed that the growth capacity of the live Alpha variant is comparable to that of the B.1 lineage virus. Because the cells used for the infection experiments are different, the discrepancy between the pseudovirus assay and the experiment using live virus may be due to the differences in the experimental setup. Further investigation will be needed to evaluate the replication capacity of SARS-CoV-2 variants using live viruses and multiple cell types. However, the use of live SARS-CoV-2 variants for in-depth experiments is still limited by experimental, processing, and safety limitations.
Consistent with a previous report , we showed that an EAb, COV2-2490, enhanced viral infectivity in in vitro cell culture experiments (Figures 2L and 3G). In addition, COV2-2490 confers greater benefit to the Lambda pseudovirus Cell Reports 38, 110218, January 11, 2022 9 Report ll OPEN ACCESS ( Figures 2L and 3G). However, Li et al. (2021) reported that the EAbs do not necessarily promote viral expansion in an in vivo animal model. Because the effect of EAbs on viral replication in vitro and in vivo may be inconsistent, the possibility of the contribution of EAbs for the expansion of Lambda variant as well as other VOCs/VOCs in vivo should be addressed in the future.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Kei Sato (KeiSato@g.ecc.u-tokyo.ac.jp).

Materials availability
All unique reagents generated in this study are listed in the key resources table and available from the lead contact with a completed Materials Transfer Agreement.
Data and code availability d The data reported in this paper will be shared by the lead contact upon request. d This paper does not report original code. d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Ethics statement
For the use of human specimens, all protocols involving human subjects recruited at Kyoto University were reviewed and approved by the Institutional Review Boards of Kyoto University (approval number G0697) and Kumamoto University (approval numbers 2074 and 477) and The Institute of Medical Science, The University of Tokyo (approval number 2021-1-0416). All human subjects provided written informed consent.
Collection of sera from BNT162b2-Vaccinated individuals Peripheral blood was collected four weeks after the second vaccination with BNT162b2 (Pfizer-BioNTech), and the sera of 18 vaccinees (average age: 39.0 years, range: 28-59 years, 22% male. Average time between doses, 21.3 days. Average time of sampling after second vaccination, 26.3 days) were isolated from peripheral blood (Table S4). Sera were inactivated at 56 C for 30 min and stored at À80 C until use.