Main

The SARS-CoV-2 Delta variant (B.1.617.2) has been detected in many countries. It has become predominant in the Indian state of Maharashtra and probably other Indian regions4, and represented 77% of the sequenced viruses circulating in the UK between 2 June and 9 June 20216. It has been classified as a variant of concern (VOC) and is believed to be 60% more transmissible than the Alpha variant (B.1.1.7). Little is known about the sensitivity of the Delta variant to the humoral immune response. Recent reports have indicated that members of the B.1.617 lineage exhibit a reduced sensitivity to certain monoclonal and polyclonal antibodies compared to the Alpha variant1,2,3,4,5,7,8,9.

Isolation and characterization of the Delta variant

We isolated the Delta variant from a nasopharyngeal swab of a symptomatic individual a few days after his return to France from India. The virus was amplified by two passages on Vero E6 cells. Sequences of the swab and the outgrown virus were identical, and identified the Delta variant (GISAID accession ID: EPI_ISL_2029113) (Extended Data Fig. 1). When compared to the D614G strain (belonging to the basal B.1 lineage), which was used here as a reference, the spike protein contained eight mutations, including four mutations in the NTD (T19R, G142D, Δ156–157 and R158G), two in the RBD (L452R and T478K), one mutation close to the furin-cleavage site (P681R) and one in the S2 region (D950N) (Extended Data Fig. 1). This set of mutations was different from that observed in other members of the B.1.617 lineage and other VOCs (Extended Data Fig. 1). Viral stocks were titrated using S-Fuse reporter cells and Vero cells10,11. The viral titres were similar in the two target cells and reached 105–106 infectious units per ml. Large syncytia expressing the spike protein were observed in cells that were infected with the Delta variant (Extended Data Fig. 2). Future work will help to determine whether Delta is more fusogenic than other variants, as suggested here by the large size of Delta-induced syncytia.

Phylogenetic analysis of the B.1.617 lineage

To contextualize the isolate of the Delta variant reported here, we inferred a global phylogeny by subsampling the diversity of SARS-CoV-2 sequences available on the GISAID EpiCoV database (Extended Data Fig. 3). The B.1.617 lineage, which is divided into three sublineages according to the PANGO classification12, derives from the B.1 lineage (D614G). The three sublineages show multiple changes in the spike protein, including the L452R substitution in the RBD, which has already been observed in other variants such as B.1.429, and P681R. This substitution is located in the furin-cleavage site and may enhance the fusogenic activity of the spike protein13. The E484Q substitution—which may be functionally similar to the antibody-escape mutation E484K that is found in the Beta and Gamma variants (B.1.351 and P.1, respectively)—is present in the B.1.617.1 and B.1.617.3 subtypes, and is likely to have reverted in the Delta sublineage, as it was present in a sequence (B.1.617) ancestral to the three sublineages14 (Extended Data Fig. 1). Whether the absence of E484Q, the presence of T478K or other changes in the spike protein or elsewhere may facilitate viral replication and transmissibility remains unknown. Notably, the B.1.617 lineage is not homogeneous, with multiple mutations that are fixed in a sublineage (for example, the spike protein mutations T19R, G142D or D950N) also being detected at lower frequencies in other sublineages. This may reflect founder effects or similar selective pressures acting on these emerging variants.

Mutational changes in the Delta variant

The locations of the spike protein mutations in the Delta variant showed a similar overall distribution to those that appeared in other VOCs. In particular, in addition to D614G, the D950N mutation mapped to the trimer interface (Extended Data Fig. 4a), suggesting that this mutation may contribute to the regulation of spike protein dynamics, as shown for the D614G mutation13. As with other VOCs, some mutations in the Delta variant cluster in the NTD (Extended Data Fig. 4b). The 156–157 deletion and G158R mutation in the Delta variant map to the same surface as the 144 and 241–243 deletions in the Alpha and Beta (B.1.351) variants, respectively. The T19R mutation maps to a surface patch that has several mutations in the Alpha variant. These altered residues are found in the NTD ‘supersite’ that is targeted by most anti-NTD neutralizing antibodies15. In the RBD, mutations appearing in VOCs map to the periphery of the ACE2-binding surface (Extended Data Fig. 4c), suggesting that the virus accumulates mutations there to reduce or avoid recognition by antibodies while maintaining binding to ACE2. For example, the L452R mutation found in the Delta variant impairs neutralization by antibodies16 and is located at this periphery. The only mutation within the ACE2 patch is at location 501, which increases the affinity of the RBD for ACE2 and is also involved in antibody escape13. The T478K mutation in the RBD is unique to the Delta variant and falls within the epitope region of potent neutralizing monoclonal antibodies categorized as ‘Class 1’17 (Extended Data Fig. 4c). This mutation is close to the E484K mutation that facilitates antibody escape13. These observations prompted us to analyse the neutralization potential, against the Delta variant, of monoclonal antibodies and sera from convalescent individuals who had recovered from COVID-19 and from individuals who had been vaccinated against COVID-19.

Neutralization of Delta by antibodies

We assessed the sensitivity of the Delta variant to a panel of human monoclonal antibodies using the S-Fuse assay. We tested four clinically approved monoclonal antibodies that target the RBD18,19 (bamlanivimab (LY-CoV555), etesevimab (LY-CoV016), casirivimab (REGN10933) and imdevimab (REGN10987)), as well as eight anti-RBD (RBD-48, RBD-85, RBD-98 and RBD-109) and anti-NTD (NTD-18, NTD-20, NTD-69 and NTD-71) monoclonal antibodies derived from convalescent individuals (C.P. et al., manuscript in preparation). Neutralizing anti-RBD monoclonal antibodies can be classified into four main categories17,20. RBD-48 and RBD-85 belong to the first category (Class 1) and act by blocking the binding of the ‘up’ conformation of RBD to ACE217. The precise epitopes of RBD-98 and RBD-109 are not yet defined but overlap with those of RBD-48 and RBD-85. The anti-NTD antibodies bind to uncharacterized epitopes.

We measured the potency of the four therapeutic antibodies against the Delta variant and included as a comparison D614G (B.1) and the Alpha and Beta variants. The antibodies neutralized D614G with a half-maximal inhibitory concentration (IC50) that ranged from 1.2 × 10−3 to 6.5 × 10−2 μg ml−1 (Fig. 1). Etesivimab exhibited a 200-fold increase in IC50 against the Alpha variant. As previously reported, bamlanivimab and etesivimab did not neutralize the Beta variant21. Bamlanivimab lost antiviral activity against the Delta variant, in line with previous results that showed that L452R is an escape mutation for this antibody16. Etesivimab, casirivimab and imdevimab remained active against the Delta variant (Fig. 1).

Fig. 1: Neutralization of the SARS-CoV-2 variants D614G, Alpha, Beta and Delta by therapeutic monoclonal antibodies.
figure 1

Neutralization curves of monoclonal antibodies. Dose–response analysis of neutralization of the D614G strain and the Alpha, Beta and Delta variants by four therapeutic monoclonal antibodies (bamlanivimab, etesivimab, casirivimab and imdevimab). Data are mean ± s.d. of four independent experiments.

Source data

Full size image

The four other anti-RBD monoclonal antibodies neutralized D614G. The IC50 values of RBD-48 and RBD-98 were about 15–100-fold higher for the Alpha variant than for D614G, whereas RBD-85 exhibited increased activity against the Alpha variant. Three monoclonal antibodies inhibited the Delta variant, whereas RBD-85 was inactive (Extended Data Fig. 5).

The four anti-NTD monoclonal antibodies were globally less efficient than anti-RBD monoclonal antibodies. They inhibited D614G with a high IC50 value (1–60 μg ml−1) (Extended Data Fig. 5). Three anti-NTD antibodies lost activity against the Alpha and Delta variants, whereas the fourth (NTD-18) inhibited the two variants to some extent. Thus, the Delta variant escapes neutralization by some antibodies that target the RBD or NTD.

We examined by flow cytometry the binding of each monoclonal antibody to Vero cells infected with the different variants, and used radar plots to show the binding of all antibodies tested (Extended Data Fig. 6). D614G was recognized by the 12 monoclonal antibodies tested. The Alpha and Delta variants were recognized by nine and by seven monoclonal antibodies, respectively. Bamlanivimab no longer bound to the Delta variant. We also analysed the binding of the 12 monoclonal antibodies to the Beta variant, which is more resistant to neutralization. Bamlanivimab and etesivimab lost their binding to the Beta variant, and only five of the antibodies bound to this variant (Extended Data Fig. 6). Thus, the escape of the Delta variant and other variants from neutralization is due to a reduction or loss of binding of the antibodies.

Sensitivity of Delta to convalescent sera

We examined the neutralization ability of sera from convalescent individuals. We first selected samples from 56 donors in a cohort of individuals from the French city of Orléans. All individuals had tested positive for SARS-CoV-2 infection by quantitative PCR with reverse transcription (qRT–PCR) or serology and included critical, severe, mild-to-moderate and asymptomatic cases of COVID-19 (Extended Data Table 1). The individuals were not vaccinated at the time at which samples were taken. We have previously characterized the potency of these sera against D614G, Alpha and Beta isolates11. We analysed samples taken from individuals a median of 188 days after the onset of symptoms (referred to as month-6 samples), and calculated the median half-maximal effective dilution (ED50) for each combination of serum and virus (Extended Data Fig. 7a). With the Alpha variant, we obtained similar ED50 values in this series of experiments to those in our previous analysis11 (Extended Data Fig. 7b). We thus included our published data for D614G and Beta in the comparison. With the Delta variant, neutralization titres were significantly decreased by four- to sixfold when compared to the Alpha and D614G strains, respectively (Extended Data Fig. 7a). This reduction in neutralizing titres was similar against the Delta and Beta variants (Extended Data Fig. 7a).

We asked whether this neutralization profile was maintained for longer periods of time. We analysed sera from 47 individuals from another cohort consisting of healthcare workers from Strasbourg University Hospitals who had a qRT–PCR-confirmed diagnosis of COVID-19 and who experienced mild disease22,23. Twenty-six individuals were unvaccinated, and 21 had received a single dose of vaccine 7–81 days before sampling. The samples were collected at a later time point (month 12), with a median of 330 and 359 days for unvaccinated and vaccinated individuals, respectively23 (Extended Data Table 1). As observed previously23, the neutralization activity was globally low at month 12 in unvaccinated individuals (Fig. 2a). There was a fourfold decrease in ED50 against the Beta and Delta variants relative to the Alpha variant (Fig. 2a). The 21 individuals within the month-12 cohort who had had a single dose of vaccine included 9 individuals who had received the AstraZeneca vaccine, 9 the Pfizer vaccine and 3 the Moderna vaccine. Sera from these vaccinated participants showed a marked increase in neutralizing antibody titres against the Alpha, Beta and Delta variants, as compared to convalescent individuals who had not been vaccinated (Fig. 2a). Therefore, as shown with other variants23,24, a single dose of vaccine boosts cross-neutralizing antibody responses to the Delta variant.

Fig. 2: Sensitivity of the SARS-CoV-2 variants D614G, Alpha, Beta and Delta to sera from convalescent individuals and vaccinated individuals.
figure 2

Neutralization titres of the sera against the indicated viral isolates are expressed as ED50 values. a, Neutralizing activity of sera from the Strasbourg cohort of convalescent individuals (n = 26; left) and convalescent individuals who had been vaccinated (n = 21; right). Samples were collected at month 12 (M12) after the onset of symptoms. b, Neutralizing activity of sera from recipients of the Pfizer vaccine, sampled at week 3 (W3) after vaccination (n = 16; left) and week 8 (W8) after vaccination (week 5 after the second dose) (n = 16; right). c, Neutralizing activity of sera from recipients of the AstraZeneca vaccine, sampled at week 10 (W10) after vaccination (n = 23; left) and week 16 (W16) after vaccination (week 4 after the second dose) (n = 20; right). The dotted line indicates the limit of detection (ED50 = 30). Data are mean from two independent experiments. A two-sided Friedman test with Dunn’s multiple comparison was performed between each of the viral strains. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Strasbourg cohort, unvaccinated (M12): D614G versus Beta, P = 0.0052; D614G versus Delta, P = 0.0052; Alpha versus Beta, P < 0.0001; Alpha versus Delta, P < 0.0001. Strasbourg cohort, vaccinated (M12): D614G versus Beta, P < 0.0001; Alpha versus Beta, P < 0.0001; Alpha versus Delta, P < 0.0001. Pfizer (W3): D614G versus Beta, P = 0.0001; D614G versus Delta, P = 0.0013. Pfizer (W8): D614G versus Beta, P = 0.0002; Alpha versus Beta, P < 0.0001; Alpha versus Delta, P = 0.0098. AstraZeneca (W10): D614G versus Beta, P < 0.0001; D614G versus Delta, P < 0.0001; Alpha versus Beta, P = 0.0006; Alpha versus Delta, P = 0.0056. AstraZeneca (W16): D614G versus Beta, P < 0.0001; D614G versus Delta, P = 0.0005, Alpha versus Beta, P < 0.0001; Alpha versus Delta, P < 0.0001.

Source data

Full size image

We then classified the individuals as neutralizers (individuals whose serum contained neutralizing antibodies that were detectable at the first serum dilution of 1/30) and non-neutralizers, for the viral variants and the two cohorts (Extended Data Fig. 7c). Sera from between 76% and 92% of individuals neutralized the four strains at month 6. The fraction of neutralizers was lower in the second cohort at month 12, and this effect was particularly marked for the Beta and Delta variants. Sera from 88% of individuals neutralized the Alpha variant, whereas sera from only 47% of individuals neutralized the Delta variant. After vaccination, sera from 100% of the convalescent individuals neutralized the four strains (Extended Data Fig. 7c).

Thus, the Delta variant exhibits enhanced resistance to neutralization by sera from convalescent individuals who have not been vaccinated, particularly by one year after the infection.

Sensitivity of Delta to sera from vaccinated individuals

We next asked whether vaccine-elicited antibodies neutralized the Delta variant in individuals who had not previously been infected with SARS-CoV-2. We randomly selected 59 individuals from a cohort of vaccinated individuals in Orléans. The characteristics of these individuals are outlined in Extended Data Table 2. Sixteen individuals received the Pfizer vaccine. They were sampled at week 3 after the first dose and week 8 (corresponding to week 5 after the second dose). Thirteen individuals were also sampled at week 16. Forty-three individuals received the AstraZeneca vaccine. Sera from 23 individuals were sampled after 1 dose (week 10) and from 20 other individuals after 2 doses (week 16, corresponding to week 4 after the second dose). We measured the potency of the sera against D614G and the Alpha, Beta and Delta variants (Fig. 2b, c).

With the Pfizer vaccine, after a single dose (at week 3) the levels of neutralizing antibodies were low against D614G, and almost undetectable against the Alpha, Beta and Delta variants (Fig. 2b). Titres significantly increased after the second dose. We observed a threefold and a sixteenfold reduction in the neutralization titres against the Delta and the Beta variants, respectively, when compared to the Alpha variant (Fig. 2b). Similar differences between strains were observed at a later time point (week 16), although titres were globally slightly lower (Extended Data Fig. 7b).

A similar pattern was observed with the AstraZeneca vaccine. A single dose (week 10) induced low levels of antibodies neutralizing the Delta and Beta variants, when compared to the D614G and Alpha strains (Fig. 2c). Four weeks after the second dose (week 16), neutralizing titres were strongly increased. There was, however, a fivefold and a ninefold reduction in neutralization titres against the Delta and the Beta variants, respectively, relative to the Alpha variant (Fig. 2c).

We then classified the vaccine recipients as neutralizers and non-neutralizers, for the four viral strains (Extended Data Fig. 7d, e). For the Pfizer vaccine, sera from 13% of individuals neutralized the Delta variant after a single dose. Sera from 81 to 100% of individuals neutralized any of the four stains after the second dose, at week 8. This fraction remained stable at week 16, with the exception of the Beta variant, which was neutralized by sera from only 46% of the individuals. Sera from 74% and 61% of individuals that received a single dose of the AstraZeneca vaccine neutralized the D614G and the Alpha strains, respectively. This fraction sharply dropped with the Beta and the Delta variants, which were inhibited by only 4 and 9% of the sera, respectively. Four weeks after the second dose of the AstraZeneca vaccine, sera from 95–100% of individuals neutralized the four strains.

Therefore, a single dose of Pfizer or AstraZeneca either showed low or no efficiency against the Beta and the Delta variants. Both vaccines generated a neutralizing response that efficiently targeted the Delta variant only after the second dose.

Discussion

We studied the cross-reactivity of monoclonal antibodies to pre-existing SARS-CoV-2 strains, sera from convalescent individuals six or twelve months after the onset of COVID-19 symptoms and sera from recent recipients of a COVID-19 vaccine against an infectious Delta isolate. Some monoclonal antibodies, including bamlanivimab, lost their ability to bind to the spike protein and no longer neutralized the Delta variant. We also showed that the Delta variant is less sensitive to sera from naturally immunized individuals. Vaccination of convalescent individuals boosted the humoral immune response to well above the threshold of neutralization. These results strongly suggest that vaccination of previously infected individuals is likely to be protective against a large array of circulating viral strains, including the Delta variant.

In individuals who had not previously been infected with SARS-CoV-2, a single dose of either the Pfizer or the AstraZeneca vaccine induced a barely detectable level of neutralizing antibodies against the Delta variant. About 10% of the sera neutralized this variant. However, a two-dose regimen generated high sero-neutralization levels against the Alpha, Beta and Delta variants in individuals sampled at week 8 to week 16 after vaccination. Levels of neutralizing antibodies are highly predictive of immune protection from symptomatic SARS-CoV-2 infection25. A recent report analysing all sequenced symptomatic cases of COVID-19 in England was used to estimate the effect of vaccination on infection26. Effectiveness was notably lower for the Delta variant than for the Alpha variant after one dose of the AstraZeneca or the Pfizer vaccine. The two-dose effectiveness against the Delta variant was estimated to be 60% and 88% for the AstraZeneca and the Pfizer vaccine, respectively26. Our neutralization experiments indicate that antibodies elicited by the Pfizer and AstraZeneca vaccines are efficacious against the Delta variant, but about three- to fivefold less potent than they are against the Alpha variant. There was no major difference in the levels of antibodies elicited by the Pfizer or the AstraZeneca vaccines.

Potential limitations of our work include the low number of vaccinated individuals analysed and the lack of characterization of cellular immunity, which may be more cross-reactive than the humoral response. Future work with more individuals and longer survey periods will help to characterize the role of humoral responses in the efficacy of vaccines against circulating variants.

Our results demonstrate that the emerging Delta variant partially—but notably—escapes neutralizing monoclonal antibodies and polyclonal antibodies elicited by previous infection with SARS-CoV-2 or by vaccination.

Methods

Data reporting

No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment. Our research complies with all relevant ethical regulations.

Orléans cohort of convalescent and vaccinated individuals

Since 27 August 2020, a prospective, monocentric, longitudinal, interventional cohort clinical study enrolling 170 individuals with SARS-CoV-2 infection (with different disease severities) and 59 healthy control individuals has been ongoing, and aims to describe the persistence of specific and neutralizing antibodies over a 24-month period. This study was approved by the Ile de France IV ethical committee. At enrolment, written informed consent was collected and participants completed a questionnaire that covered sociodemographic characteristics, virological findings (SARS-CoV-2 qRT–PCR results, including date of testing), clinical data (date of symptom onset, type of symptoms, hospitalization) and data related to anti-SARS-CoV-2 vaccination if ever (brand product, date of first and second doses). The serological status of participants was assessed every three months. Those who underwent anti-SARS-CoV-2 vaccination had regular blood sampling after the first dose of vaccine (ClinicalTrials.gov Identifier: NCT04750720). The primary outcome was the presence of antibodies to SARS-CoV-2 spike protein as measured with the S-Flow assay. The secondary outcome was the presence of neutralizing antibodies as measured with the S-Fuse assay. For the present study, we selected 56 convalescent individuals and 59 vaccinated individuals (16 with Pfizer and 43 with AstraZeneca). Study participants did not receive any compensation.

Strasbourg cohort of convalescent individuals

Since April 2020, a prospective, monocentric, longitudinal, interventional cohort clinical study enrolling 308 hospital staff from the Strasbourg University Hospitals who had a qRT–PCR-confirmed diagnosis of SARS-CoV-2 infection has been ongoing (ClinicalTrials.gov Identifier: NCT04441684). At enrolment (from 17 April 2020), written informed consent was collected and participants completed a questionnaire that covered sociodemographic characteristics, virological findings (SARS-CoV-2 qRT–PCR results, including date of testing) and clinical data (date of symptom onset, type of symptoms, hospitalization). This study was approved by the Institutional Review Board of Strasbourg University Hospital. The serological status of the participants has been described at months 3 and 6 after the onset of symptoms22,23. Laboratory identification of SARS-CoV-2 was performed at least 10 days before inclusion by qRT–PCR testing on nasopharyngeal swab specimens according to current guidelines (Institut Pasteur; WHO technical guidance). The assay targets two regions of the viral RNA-dependent RNA polymerase (RdRp) gene with a threshold of detection of 10 copies per reaction. The primary outcome was the presence of antibodies to the SARS-CoV-2 spike protein as measured with the S-Flow assay. The secondary outcome was the presence of neutralizing antibodies as measured with the S-Fuse assay. For the present study, we randomly selected 47 convalescent individuals at month 12 (26 unvaccinated and 21 vaccinated). Study participants did not receive any compensation.

Phylogenetic analysis

All SARS-CoV-2 sequences available on the GISAID EpiCov database as of 21 May 2021 were retrieved. A subset of complete and high-coverage sequences, as indicated in GISAID, assigned to lineages B.1.617.1, B.1.617.2 and B.1.617.3 were randomly subsampled to contain up to five sequences per country and epidemiological week in R with the packages tidyverse and lubridate. Together with a single B.1.617 sequence this subset was included in the global SARS-CoV-2 phylogeny reconstructed with augur and visualized with auspice as implemented in the Nextstrain pipeline (https://github.com/nextstrain/ncov; version from 21 May 2021)27. Within Nextstrain, a random subsampling approach capping a maximum number of sequences per global region was used for the contextual non-B.1.617 sequences. Acknowledgement of the contributing and originating laboratories for all sequences used in the analysis is provided in Supplementary Table 1.

Three-dimensional mapping of mutations on B1.617.2 and other variants to the surface of the spike protein

Panels in Extended Data Fig. 4 were prepared with the PyMOL Molecular Graphics System, v.2.1 (Schrödinger). The atomic model used (Protein Data Bank code: 6XR8) has been previously described28.

S-Fuse neutralization assay

U2OS-ACE2 GFP1-10 or GFP 11 cells, also termed S-Fuse cells, become GFP+ when they are productively infected by SARS-CoV-210,11. Cells tested negative for mycoplasma. Cells were mixed (ratio 1:1) and plated at 8 × 103 per well in a μClear 96-well plate (Greiner Bio-One). The indicated SARS-CoV-2 strains were incubated with serially diluted monoclonal antibodies or sera for 15 min at room temperature and added to S-Fuse cells. The sera were heat-inactivated 30 min at 56 °C before use. Eighteen hours later, cells were fixed with 2% paraformaldehyde (PFA), washed and stained with Hoechst (dilution 1:1,000, Invitrogen). Images were acquired with an Opera Phenix high-content confocal microscope (PerkinElmer). The GFP area and the number of nuclei were quantified using Harmony software (PerkinElmer). The percentage of neutralization was calculated using the number of syncytia as value with the following formula: 100 × (1 – (value with serum − value in ‘non-infected’)/(value in ‘no serum’ − value in ‘non-infected’)). The neutralizing activity of each serum was expressed as the ED50 value. ED50 values (in μg ml−1 for monoclonal antibodies and in dilution values for sera) were calculated with a reconstructed curve using the percentage of the neutralization at the different concentrations.

Clinical history of the patient infected with B.1.617.2

A 54-year-old man was admitted on 27 April 2021 to the emergency department of the Hôpital Européen Georges Pompidou hospital in Paris, France, for an acute respiratory distress syndrome with fever. He had no medical background and came back from a trip to India (West Bengal and a few days spent in Delhi) 10 days before (17 April 2021), where he stayed 15 days for his work. The onset of symptoms (abdominal pain and fever) was approximately 18 April 2021. The nasopharyngeal swab tested positive for SARS-CoV-2 at his date of admission. Lung tomo-densitometry showed a mild (10–25%) COVID-19 pneumonia without pulmonary embolism. He initially received oxygen therapy (2 l min−1), dexamethasone (6 mg per day) and enoxaparin (0.4 ml twice a day). His respiratory state worsened on day 3 (30 April 2021). He was transferred to an intensive care unit, in which he received high-flow oxygen therapy (maximum 12 l min−1). His respiratory condition improved, and he was transferred back to a conventional unit on day 8 (5 May 2021). He was discharged from hospital on day 15 (May 10 2021).

Virus strains

The reference D614G strain (hCoV-19/France/GE1973/2020) was supplied by the National Reference Centre for Respiratory Viruses hosted by the Institut Pasteur and headed by S. van der Werf. This viral strain was supplied through the European Virus Archive – Global (EVAg) platform, a project that has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement no. 653316. The variant strains were isolated from nasal swabs using Vero E6 cells and amplified by one or two passages. B.1.1.7 originated from an individual in Tours (France) who had returned from the UK. B.1.351 (hCoV-19/France/IDF-IPP00078/2021) originated from an individual in Creteil (France). B.1.617.2 was isolated from a nasopharyngeal swab of a hospitalized patient who had returned from India, as described above. The swab was provided and sequenced by the Laboratoire de Virologie of the Hopital Européen Georges Pompidou (Assistance Publique des Hôpitaux de Paris). All individuals provided informed consent for the use of the biological materials. Titration of viral stocks was performed on Vero E6, with a limiting dilution technique allowing a calculation of TCID50, or on S-Fuse cells. Viruses were sequenced directly on nasal swabs, and after one or two passages on Vero cells. Sequences were deposited in the GISAID database immediately after their generation, with the following IDs: D614G: EPI_ISL_414631; B.1.1.7: EPI_ISL_735391; B.1.1.351: EPI_ISL_964916; B.1.617.2: ID: EPI_ISL_2029113.

Flow cytometry

Vero cells were infected with the indicated viral strains at a multiplicity of infection (MOI) of 0.1. Two days after, cells were detached using PBS-EDTA and transferred into U-bottom 96-well plates (50,000 cells per well). Cells were fixed in 4% PFA for 15–30 min at room temperature. Cells were then incubated for 15–30 min at room temperature with the indicated monoclonal antibodies (1 μg ml−1) in PBS, 1% BSA, 0.05% sodium azide and 0.05% saponin. Cells were washed with PBS and stained using anti-IgG AF647 (1:600 dilution) (Thermo Fisher Scientific). Stainings were also performed on control uninfected cells. Data were acquired on an Attune Nxt instrument using Attune Nxt Software v.3.2.1 (Life Technologies) and analysed with FlowJo v.10.7.1 (Becton Dickinson).

Antibodies

The four therapeutic antibodies were provided by CHR Orleans. Human anti-SARS-CoV2 monoclonal antibodies were cloned from S-specific blood memory B cells of individuals who were convalescing from COVID-19 (C.P. et al., manuscript in preparation). Recombinant human IgG1 monoclonal antibodies were produced by co-transfection of Freestyle 293‐F suspension cells (Thermo Fisher Scientific) as previously described29, purified by affinity chromatography using protein G sepharose 4 fast flow beads (GE Healthcare) and validated by enzyme-linked immunosorbent assay (ELISA) against the trimeric S, RBD, S2 and NTD proteins (C.P. et al., manuscript in preparation).

Statistical analysis

Flow cytometry data were analysed with FlowJo v.10 software (TriStar). Calculations were performed using Excel 365 (Microsoft). Figures were drawn using GraphPad Prism 9. Statistical analysis was conducted using GraphPad Prism 9. Statistical significance between different groups was calculated using the tests indicated in each figure legend.

Reporting summary

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