COVA1-18 neutralizing antibody protects against SARS-CoV-2 in three preclinical models

Effective treatments against Severe Acute Respiratory Syndrome coronavirus 2 (SARS-CoV-2) are urgently needed. Monoclonal antibodies have shown promising results in patients. Here, we evaluate the in vivo prophylactic and therapeutic effect of COVA1-18, a neutralizing antibody highly potent against the B.1.1.7 isolate. In both prophylactic and therapeutic settings, SARS-CoV-2 remains undetectable in the lungs of treated hACE2 mice. Therapeutic treatment also causes a reduction in viral loads in the lungs of Syrian hamsters. When administered at 10 mg kg-1 one day prior to a high dose SARS-CoV-2 challenge in cynomolgus macaques, COVA1-18 shows very strong antiviral activity in the upper respiratory compartments. Using a mathematical model, we estimate that COVA1-18 reduces viral infectivity by more than 95% in these compartments, preventing lymphopenia and extensive lung lesions. Our findings demonstrate that COVA1-18 has a strong antiviral activity in three preclinical models and could be a valuable candidate for further clinical evaluation.

Across the world, the Coronavirus Disease 19 (COVID-19) pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continues to escalate1. Despite the progressive rollout of vaccines, there remains an urgent need for both curative and preventive measures, especially in individuals with high risk. Monoclonal neutralizing antibodies (NAbs), isolated from convalescent COVID-19 patients, are one of the most promising approaches and two NAb-based products have already received emergency use authorizations by regulatory agencies in both the US2,3 and Europe4,5. Although their clinical efficacy in hospitalized patients remains to be fully assessed, their capability to reduce viral loads and hospitalization in high risk individuals shows that NAbs constitute an effective treatment when administered early enough after symptom onset6,7,8.

We and others have previously isolated and characterized several highly potent monoclonal NAbs with half-maximum inhibitory concentration (IC50) values in the picomolar range9,10,11,12, with the majority of these targeting the receptor binding domain (RBD) on the S1 subunit of the S protein. We previously identified COVA1-18, an RBD-specific monoclonal Ab, as one of the most potent NAb in vitro9.

In this work, we use three different experimental models as well as mathematical modeling to demonstrate that COVA1-18 rapid and extensive biodistribution is associated with a very potent antiviral effect, and make it a promising candidate for clinical evaluation, both as a prophylactic or therapeutic treatment of COVID-19.

COVA1-18 in vitro potency is dependent on avidity

To advance our earlier in vitro results9 on COVA1-18 and allow for better comparability with other studies, we used two pseudovirus assays, one using lentiviral pseudotypes with an ACE2-expressing 293 T cell line13, and one using VSV-pseudotypes with Vero E6 cells14, to confirm the potency of COVA1-18. With these assays, we found that COVA1-18 IgG inhibited lentiviral SARS-CoV-2 pseudovirus with an IC50 of 1.7 ng ml−1 (11.3 pM) and VSV-based pseudovirus with an IC50 of 9 ng ml−1 (60 pM), confirming the remarkable potency previously observed against authentic virus9 (Supplementary Fig. 1a, Table 1). These results were corroborated in multiple independent labs and COVA1-18 was also equipotent against the D614G variant (Table 1) that now dominates worldwide15,16,17,18,19 as well as the recently emerged B.1.1.7 variant that includes the N501Y mutation in the RBD20,21 (Table 2).

COVA1-18 bound strongly to SARS-CoV-2 S protein and showed no cross-reactivity with S proteins of SARS-CoV, MERS-CoV and common cold coronaviruses HKU1-CoV, 229E-CoV and NL63-CoV (Supplementary Fig. 1b)9. Biolayer interferometry experiments showed that COVA1-18 IgG bound to soluble SARS-CoV-2 S protein with an apparent dissociation constant (KD) of 5 nM, and its affinity for RBD was similar at 7 nM (Fig. 1a, Supplementary Fig. 1c, d, Table 1). Its Fab displayed a 12-fold weaker binding to RBD compared to IgG (84 nM), with the difference mainly caused by a faster Fab off-rate (Fig. 1a, Table 1), as also observed in a different assay setting (Supplementary Fig. 1d). With an IC50 of 199 ng ml−1, the COVA1-18 Fab was 237-fold less potent at neutralizing SARS-CoV-2 pseudovirus, showing that the IgG avidity effect is important for COVA1-18 neutralization potency (Supplementary Fig. 1a, Table 1).

a Biolayer interferometry sensorgrams comparing COVA1-18 IgG and Fab binding to RBD. KDs are indicated. Representative of 3 independent experiments. b Study design with n = 5 per group, except mouse control group (n = 3). Hamsters were infected with 105 PFU on day 0 and treated on day 1. Mice received COVA1-18 24 h prior to or after exposure to 104 PFU. Lung viral titers at 3 days post-infection are shown for mice (c) and hamsters (d). Bars indicate medians. Mann-Whitney unpaired two-tailed t-test, p values: *:0.0179, **:0.0079. Ctl. control group (black), KD dissociation constant, PFU Plaque forming unit, PrEP pre-exposure prophylaxis (dark blue), Ther. therapeutic (light blue).

Full size image

COVA1-18 inhibits viral replication in rodents

We sought to evaluate whether COVA1-18 could control SARS-CoV-2 viral infection in a previously described Ad5-hACE2 mouse model22,23 using a 10 mg kg−1 dose. COVA1-18 administered intraperitoneally 24 h either prior to or after a SARS-CoV-2 challenge with 104plaque forming units (PFU) (n = 5 for treated groups, n = 3 for control group) was fully protective with no detectable viral replication in the lungs (Fig. 1b, c). We then tested the efficacy of COVA1-18 in the golden Syrian hamster model (n = 5 per group), which is naturally susceptible to SARS-CoV-2 and develops severe pneumonia upon infection24. We evaluated the effect on lung viral loads of 10 mg kg−1 of COVA1-18 given 24 h after a 105 PFU intranasal challenge (Fig. 1b, d). At 3 days post-infection (d.p.i.), 3/5 animals had high serum neutralization, while for 2/5 animals, low neutralization activity was observed (Supplementary Fig. 1e). On day 3, the COVA1-18 treated group had significantly lower median lung viral titers compared to the control group (3.5 vs 6.7 log10 PFU g−1, respectively, p < 0.01) with lowest viral titers corresponding to the higher neutralizing serum activity (Fig. 1d). The time of treatment (24 h post-infection) and 3-day study period did not allow for prevention of lung damage and recovery monitoring in this model (Supplementary Fig. 1f, gand Supplementary Table 1).

COVA1-18 PrEP prevents infection in NHP

We evaluated the potential of COVA1-18 to prevent SARS-CoV-2 infection in cynomolgus macaques in a pre-exposure prophylaxis (PrEP) study. The animals were treated intravenously 24 h prior to viral challenge with a dose of 10 mg kg−1 of COVA1-18 (Fig. 2a). Treated and control animals (n = 5 per group) were challenged on day 0 with 106 PFU of SARS-CoV-2 via combined intranasal and intratracheal routes using an experimental protocol developed previously25,26. On the day of challenge, the mean COVA1-18 serum concentration was 109 ± 2.7 μg ml−1 (Fig. 2b, Supplementary Fig. 2a). COVA-18 was also detected in all respiratory tract samples and rectal samples (Fig. 2c–e, Supplementary Fig. 2b–d), and represented on average 1.5% and 1.2% of the total IgG in heat-inactivated content in the nasopharyngeal and tracheal mucosae, respectively. These levels remained constant throughout the study period and similar levels were detected at 3 d.p.i. in bronchoalveolar lavages (BAL) and saliva (Fig. 2e, f). As SARS-CoV-2 can cause damage to non-respiratory organs, we performed a pharmacokinetic study on two additional macaques to characterize the COVA1-18 distribution within the first 24 h using non heat-inactivated samples (Fig. 2gand Supplementary Fig. 2e, f). COVA1-18 was found in all organs studied, including the lungs, at concentrations of 4 to 22 ng mg−1 of tissue, except for the brain where concentrations were substantially lower (250 pg mg−1 of tissue) (Fig. 2g). Altogether, these data showed that COVA1-18 administered intravenously was rapidly and efficiently distributed to the natural sites of infection as well as to organs affected by COVID-19 pathology.

a Study design. Two groups of n = 5 were exposed to 106 PFU of SARS-CoV-2 (intranasal and intratracheal routes). Treated animals received 10 mg kg−1of COVA1-18 1 day before challenge. bCOVA1-18 serum concentration (mean with range). COVA1-18 concentration reported as percent of total cynomolgus IgG in heat-inactivated (c) nasopharyngeal fluid, d tracheal fluid (means with range), e bronchoalveolar lavage (BAL) and f saliva (means ± SEMs) with n = 5, except for the control group in (f) where n = 1. g The two macaques from the pharmacokinetic study were euthanized at 24 h post-treatment and their organs analyzed to assess the biodistribution of COVA1-18. The concentration of COVA1-18 was normalized to the weight of each sample for every organ. Bars represent means. The red dashed line indicates challenge day. L left, PFU Plaque forming unit, PK pharmacokinetic, R right.

Full size image

Following viral challenge, control animals showed similar genomic (g)RNA and subgenomic (sg)RNA levels and kinetics as previously described25,26 with median peak viral loads (VL) of 6.4 and 6.2 log10 copies per ml at 1-2 d.p.i. in the nasopharyngeal and tracheal swabs, respectively (Fig. 3a). Active viral replication, as assessed by sgRNA levels, peaked at 1-2 d.p.i. in nasopharyngeal and tracheal swabs with median values of 4.6 and 4.0 log10 copies per ml, respectively (Fig. 3b). At 3 d.p.i., viral loads were detected in the BAL with a median value of 4.9 log10 copies per ml of gRNA and 3.2 log10 copies per ml of sgRNA, including 3 animals with no detectable sgRNA.

In comparison, treated animals had a reduction of 2.2 and 3.4 log10 median gRNA VL in tracheal swabs on days 1 and 2 (both p < 0.01 to controls), and had undetectable VL after day 4 (Fig. 3a). The difference was also evident in nasopharyngeal swabs, with treated animals having a reduction of 1.5 and 2.2 log10 gRNA VL on days 1 and 2 (both p < 0.01 to controls). By day 4, 4/5 treated animals had undetectable gRNA in the nasopharyngeal swabs while one animal (MF7) remained positive with a low residual gRNA signal up to 7 d.p.i. COVA1-18 treatment dramatically hindered viral replication in the upper respiratory tract as evidenced by the absence of detectable sgRNA in the nasopharyngeal and tracheal swabs for all treated animals with the exception of animal (MF9) that showed a low signal at 1 d.p.i. only in the tracheal swabs (Fig. 3b). Therefore, in the treated group, most upper respiratory tract gRNA VL likely represents the progressive elimination of the challenge inoculum, and does not result from active replication. The gRNA and sgRNA loads in BAL were also lower in COVA1-18 recipients compared to controls but the difference did not reach statistical significance (Fig. 3a, b). Cynomolgus anti-S IgM was detected as early as 6 d.p.i. in control animals, while no IgM was detected in treated animals at early timepoints (Supplementary Fig. 2g). Some IgM was detected at 28 d.p.i. in 3 treated animals (MF6, MF7, MF9), although levels remained lower than controls at 6 d.p.i. No anti-S specific cynomolgus IgG was detected up to the day of euthanasia in control animals (7 d.p.i.) or in treated animals up to 28 d.p.i. (Supplementary Fig. 2h). Overall, these results demonstrate that a 10 mg kg−1 dose of COVA1-18 PrEP dramatically reduced the acquisition and/or early spread of SARS-CoV-2 in the different respiratory compartments.

Analysis of lung lesions by chest computed tomography (CT) showed that all treated animals had few and small lung lesions as recorded by low CT scores at 3 d.p.i. while 2/5 controls showed mild pulmonary lesions characterized by non-extended ground-glass opacities (GGOs) with scores superior to 5, consistent with what was observed in historic controls25 and mirroring the heterogeneity of COVID-19 infection in humans27 (Fig. 3c). In addition, we observed that all control animals were lymphopenic at 2 d.p.i., consistent with previous studies25,26, while all treated animals had normal lymphocyte counts throughout the study (p < 0.01 for the comparison) (Fig. 3d and Supplementary Fig. 2i).

One concern about SARS-CoV-2 vaccines and NAb treatments is the possible generation of suboptimal concentrations of NAb in individuals, which could foster viral escape28. Sequencing analysis of nasopharyngeal, tracheal and BAL samples at 3 d.p.i. showed that COVA1-18 treatment resulted in enrichment of subclonal variations in N and ORF1ab. One mutation (E725G) was detected in the S gene in the MF7 BAL sample when applying standard quality filters, but this mutation has not been previously implicated in immune escape and located outside the epitope of COVA1-18 (Supplementary Fig. 3 and Supplementary Information). The high efficacy of COVA1-18 treatment prevented recovery of viral genetic information past 3 d.p.i.

Prediction models refine COVA1-18 dosage

Next, we used a viral dynamic model previously developed in the same SARS-CoV-2 NHP experimental model29 to evaluate the level of protection conferred by COVA1-18, and guide potential subsequent studies on SARS-CoV-2 MAbs. The model considers a target cell limited infection in both nasopharyngeal and tracheal compartments. In addition to the previously developed model, we assumed that sgRNA was a proxy for the total number of non-productively and productively infected cells (see supplementary methods) and we further assumed that COVA1-18 plasma drug concentrations over time, noted C(t), was the driver of drug efficacy. We modeled the changes in C(t) using a standard first order absorption and elimination model, and we estimated the half-life of COVA1-18 in plasma to be 12.6 days (Supplementary Fig. 4a). We assumed that COVA1-18 reduces infectivity rate in both tracheal and nasopharyngeal compartments with an efficacy, noted η(t), determined by the following model η(t)=C(t)C(t)+EC50$\eta (t)=\frac{\mathrm{C}(\mathrm{t})}{\mathrm{C}(\mathrm{t})+{\mathrm{E}\mathrm{C}}_{50}}$, where EC50 is the plasma COVA1-18 concentrations corresponding to a 50% reduction of viral infectivity. The model fitted the viral kinetics well in all animals (Fig. 4a, Supplementary Fig. 4b, Supplementary Table 2). The EC50 was estimated to be 2.2 and 0.053 µg ml−1 in the nasopharynx and trachea, respectively, which is roughly 50 and 2000 times lower than the plasma drug concentrations of 109 µg ml−1 observed on the day of infection (see above). Thus, these results confirm that the efficacy of COVA1-18 was very high, with efficacies above 95% and 99.9% in nasopharyngeal and tracheal compartments on the day of infection, respectively (Fig. 4a, Supplementary Fig. 4b). Given the long half-life of the drug, this efficacy was maintained over time, and we estimated that the mean individual efficacy of the COVA1-18 in the first 10 days following infection ranged between 96.67% and 97.50% in the nasopharynx and between 99.91% and 99.94% in the trachea (Supplementary Fig. 4c).

Next, we used our model to investigate changes in experimental conditions, such as COVA1-18 dose being administered at a lower dose and/or after the viral challenge (see methods). In all scenarios considered, a dose of 5 mg kg−1 was determined to provide nearly similar results than 10 mg kg−1 (Fig. 4b, c, Supplementary Fig. 5a, b). A dose of 1 mg kg−1 could be sufficient to prevent active viral replication as long as treatment is given prior to infection, but might be insufficient in a therapeutic setting. However, this dose could be relevant if lower doses of virus were used for infection, such as 104 or 105 PFU (Supplementary Fig. 4d–g).

COVA1-18/1-16 cocktail neutralizes B.1.351

Many highly potent RBD-targeting mAbs are affected by mutations in emerging variants-of-concern (VOC), in particular E484K30,31. We evaluated the ability of COVA1-18 to neutralize VOCs B.1.1.7 and B.1.351 as well as a E484K single mutant virus. While COVA1-18 retains its high potency against the B.1.1.7 strain, it lost its capacity to neutralize the B.1.351 strain due primarily to the RBD E484K mutation in the spike (Supplementary Fig. 6). Therefore, we also evaluated the in vitro potency of COVA1-18 in a cocktail with COVA1-16, an antibody that neutralizes B.1.351 as well as SARS-CoV-1, but is less potent than COVA1-189 (Supplementary Fig. 6 and Table 2). This mAb cocktail retained the high potency of COVA1-18 against wild-type, D614G and B.1.1.7 and also efficiently neutralized B.1.351, providing an avenue for broad mAb prophylaxis and treatment against VOCs.

Despite the recent approval of several SARS-CoV-2 vaccines by health authorities, the slow roll-out of vaccination campaigns will not result in resolution of the pandemic in the immediate future. Furthermore, the emergence of viral escape mutants may lead to reduced vaccine efficacy, and some individuals, such as immunocompromised patients or the elderly, may not mount adequate protective immune responses to vaccination. Thus, there is an urgent need to develop effective therapeutics, in particular for individuals with high risk of severe disease.

In hACE2-expressing mice and golden Syrian hamsters, COVA1-18 showed remarkable control of SARS-CoV-2 infection. These promising results were confirmed in NHPs, with COVA1-18 given one day prior to infection achieving nearly complete protection in the upper respiratory tract in cynomolgus macaques. Using a viral dynamic model, we estimated that COVA1-18 reduced viral infectivity by >95% and 99.9% in nasopharyngeal and tracheal compartments, respectively. The robustness of these results are reinforced by the high challenge dose that we used, which was 10 to 100-fold higher than in other NHP studies evaluating NAbs for PrEP against SARS-CoV-232,33,34,35,36,37,38. In fact, the model allowed us to predict, without using additional animals, that a high level of protection could be achieved with lower doses of 5 mg kg−1 and 1 mg kg−1 with lower inoculum doses of 105 or 104PFU, both in prophylactic and therapeutic settings (Supplementary Fig. 4, Supplementary Fig. 5).

How do these levels of efficacy greater than 95% translate into clinical efficacy? In previous work, we estimated that achieving 90% efficacy would be sufficient to confer a high level of protection against infection acquisition if treatment can be administered prophylactically or just after a high-risk contact38. In hospitalized patients, where viral load kinetics after admission are associated with the risk of death, we estimated that administration of treatment with an efficacy higher than 90% could reduce the time to viral clearance by more than 3 days in patients over 65 years of age, which could translate into significantly lower rates of mortality in this population39.

Several NAbs are being developed and some have achieved clinical endpoints, such as the reduction of the risk of hospitalization in patients that initiate treatment within 5 days of symptom onset6,7,8, leading to their approval for emergency use22,32,33,34,35,36,37,40. However, the narrow efficacy range of FDA-approved NAbs41,42,43, together with rapidly spreading new variants complicate treatment strategies30,31,44,45, highlighting the need for additional treatment options, including potent NAbs, such as COVA1-18, that could be used in combination with other NAbs. The plasma half-life was 12.6 days, albeit lower to what is found typically for human NAbs in humans37, ranging from 15 to 25 days, and consistent with values reported for other human NAbs in the macaque model (Supplementary Table 3). The efficacy in this model was high, despite the high challenge dose (106 PFU) used here. We estimated that with lower inoculum doses of 104 or 105 PFU, as used in other studies32,33,37(Supplementary Table 3), a dose of 10 mg kg−1 COVA1-18 could reduce the viral load even more dramatically (Supplementary Fig. 4d and Supplementary Fig. 5a, b). Although it is difficult to compare results obtained with different experimental and virological models, this model shows that the in vivo efficacy of COVA1-18 is comparable with what has been obtained for other advanced NAbs in clinical development.

An optimal cocktail should not only be based on intrinsic efficacy against wild-type virus of each NAbs, but rather whether synergy could be achieved in terms of binding domain and/or spectrum of efficacy. Indeed, the increasing prevalence of mutant strains has reduced the sensitivity to pre-existing NAbs, including those given in combination30. Escape mutations can arise following single NAb treatment as recently demonstrated37,46 and the one S mutation found in a unique sample from one animal treated with COVA1-18 is not in the epitope of COVA-18. Importantly, we and others have determined that COVA1-18 retains high potency against the B.1.1.7 variant, which includes the N501Y mutation20,21. However, COVA1-18 lost its potency against B.1.351 which harbors the E484K mutation that is also found in the B.1.1.28 lineage, similar to what has been found with first wave convalescent plasma and many NAbs30,31. This finding highlights the necessity of using NAbs cocktails targeting distinct epitopes and we propose the use of the SARS-CoV-1 cross-neutralizing antibody COVA1-16, which can effectively neutralize B.1.351, in a 1:1 cocktail with COVA1-18. In addition, the half-life of COVA1-18 can be extended by incorporating the LS or YTE47mutations, which can further reduce the protective dose required and reduce the cost of treatment.

While approved SARS-CoV-2 mAbs are given intravenously, other therapeutic mAbs are given intramuscularly or by subcutaneous injection48. SARS-CoV-2 mAbs could potentially also be administered intranasally or delivered via gene therapy to the airways49, to provide protection where it is most needed, i.e. the respiratory tract. The biodistribution of COVA1-18 by different routes of administration would also have to be investigated. In addition, we note that COVA1-18 and numerous potent neutralizing Abs isolated to date against SARS-CoV-2 have very low levels of somatic hypermutation. Thus, these antibodies are very close to the germline precursor and unlikely to trigger anti-idiotypic response in patients.

In conclusion, our COVA1-18 in vitro data translated into a powerful protective drug in three preclinical models to prevent SARS-CoV-2 replication. Together with our prediction model, these data showed that COVA1-18 could be used in patients at low doses either to prevent infection or to reduce viral loads in a therapeutic setting, with a potential greater impact in high-risk patients. The high in vivo efficacy of COVA1-18 and its demonstrated potency against the B.1.1.7. isolate also suggests that it is a promising candidate for a NAb cocktail.