Antibody-mediated protection against symptomatic COVID-19 can be achieved at low serum neutralizing titers

Multiple studies of vaccinated and convalescent cohorts have demonstrated that serum neutralizing antibody (nAb) titers correlate with protection against coronavirus disease 2019 (COVID-19). However, the induction of multiple layers of immunity after severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) exposure has complicated the establishment of nAbs as a mechanistic correlate of protection (CoP) and hindered the definition of a protective nAb threshold. Here, we show that a half-life–extended monoclonal antibody (adintrevimab) provides about 50% protection against symptomatic COVID-19 in SARS-CoV-2–naïve adults at serum nAb titers on the order of 1:30. Vaccine modeling results support a similar 50% protective nAb threshold, suggesting that low titers of serum nAbs protect in both passive antibody prophylaxis and vaccination settings. Extrapolation of adintrevimab pharmacokinetic data suggests that protection against susceptible variants could be maintained for about 3 years. The results provide a benchmark for the selection of next-generation vaccine candidates and support the use of broad, long-acting monoclonal antibodies as alternatives or supplements to vaccination in high-risk populations. Description A passively transferred monoclonal antibody protects against COVID-19 in humans, even at low serum neutralizing titers. Correlating protection Monoclonal antibodies have become an essential component of the SARS-CoV-2 treatment toolkit. However, the serum neutralizing titers necessary for these antibodies to confer protection against symptomatic infection have not been determined. Here, Schmidt et al. studied the protection conferred by a half-life extended monoclonal antibody, adintrevimab, in SARS-CoV-2–naive adults. Serum neutralizing antibody titers of 1:30 were sufficient to confer 50% protection against symptomatic infection in these individuals. These data, in combination with pharmacokinetics studies in humans, suggested that adintrevimab could confer protection against susceptible variants for up to 3 years. Together, these results support the use of half-life extended monoclonal antibodies for SARS-CoV-2 prophylaxis, especially in high-risk populations. —CM


INTRODUCTION
The establishment of correlates of protection (CoPs), defined as specific immunological markers associated with protection against infection or disease caused by a pathogen (1), is of critical importance for accelerating the licensure of new vaccines, evaluating susceptibility to disease, and determining the need for and optimal timing of booster vaccinations. There are two types of CoPs: mechanistic CoPs, which are directly responsible and statistically interrelated with protection, and nonmechanistic CoPs, which are correlated with the mechanistic factor without directly conferring protection (1).
Studies in both humans and animal models have universally demonstrated that the induction of neutralizing antibody (nAb) responses correlates with protection against coronavirus disease 2019 (COVID-19) (2)(3)(4)(5)(6)(7)(8). However, the human CoP studies performed to date have been confounded by the presence of other forms of immunity induced by prior severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and vaccination (such as memory B cells and T cells). These other immune responses may also contribute to, and independently correlate with, protection and therefore complicate the establishment of serum nAb as a mechanistic CoP. Furthermore, the lack of standardized serological assays has required the normalization of vaccine-induced nAb titers to those observed after convalescence, allowing only for the determination of relative, rather than absolute, protective titers (3,8). Last, although clinical studies have demonstrated that passively transferred monoclonal antibodies can protect against COVID-19 in the absence of other forms of immunity, the high neutralizing titers conferred by these therapies have precluded the definition of a protective neutralization threshold (9,10).
We conducted a phase 2/3 clinical study to evaluate the efficacy of a half-life-extended human monoclonal antibody (adintrevimab) in the prevention of symptomatic COVID-19 in SARS-CoV-2-naïve adults during the emergence and global spread of SARS-CoV-2 variants Delta and Omicron BA.1/BA1.1. Because of marked differences in adintrevimab potency against these two variants (11), combined with the natural waning of passively transferred nAb titers over time, we were afforded the unique opportunity to assess the relationship between serum nAb titers and clinical protection against symptomatic COVID-19 in the absence of preexisting SARS-CoV-2 immunity. The observed protective efficacy of adintrevimab in our trial, combined with vaccine modeling data, provide strong evidence that nAbs are mechanistic in mediating protection against symptomatic COVID-19 and suggest that clinically meaningful efficacy can be achieved at serum neutralizing titers on the order of 1:30.

Adintrevimab exhibits intact effector functions in vitro and extended half-life in vivo
We previously described a human immunoglobulin G1 (IgG1) broadly neutralizing antibody (bnAb), ADG2, directed to the SARS-CoV-2 receptor-binding domain (RBD) (12). The precursor to this bnAb was isolated from a 2003 SARS-CoV survivor and subsequently engineered in vitro to improve its neutralization breadth and potency against a wide range of human angiotensin converting enzyme 2-using sarbecoviruses (12,13). We introduced a twoamino acid modification (M428L/N434A; "LA") into the fragment crystallizable (Fc) region of ADG2 (ADG2-LA; hereafter referred to as adintrevimab) to enhance binding affinity to the neonatal Fc receptor (FcRn) under the acidic conditions of the lysosome (pH 6.0) and prolong serum half-life in vivo. Previous studies have shown that these two individual mutations improve binding to human FcRn at low pH, and the N484A variant has been shown to exhibit decreased clearance relative to wild type in rhesus macaques (14)(15)(16). In accordance with these studies, adintrevimab bound with sixfold and fivefold higher affinity to human and cynomolgus macaque FcRn, respectively, relative to ADG2 (Fig. 1A). and adintrevimab for human FcγRs were measured by surface plasmon resonance (SPR), and binding affinities for C1q were determined by biolayer interferometry. N.B., nonbinding. (D) In vitro Fc-mediated functional activities of ADG2 and adintrevimab. Primary human NK cells were analyzed for the production of tumor necrosis factor-α (TNF-α) (left) and interferon-γ (IFN-γ) (right) and for surface expression of CD107a (middle), indicating degranulation, after incubation with antibody-RBD immune complexes. (E) Antibody-mediated phagocytosis of RBD-coated beads by differentiated THP-1 monocytes (left) or HL-60 neutrophil-like cells (right) was measured by flow cytometry. (F) ADCD was measured by detection of complement component C3 on RBD-coated beads after incubation of guinea pig complement with immune complexes. MFI, mean fluorescence intensity. Data points and error bars represent means and SDs, respectively. All data are representative of two independent experiments. Statistical comparisons were determined by two-way analysis of variance (ANOVA) with Tukey's multiple comparison test. **P < 0.01, ***P < 0.001, and ****P < 0.0001. To determine whether the improvement in FcRn binding at low pH translated to extended serum half-life in vivo, we evaluated the pharmacokinetics (PK) of adintrevimab for up to 98 days in naïve cynomolgus macaques after a single intramuscular (IM) injection (10 mg/kg) or intravenous (IV) infusion. After IV infusion, we observed low clearance (mean = 0.126 ml/hour per kg) and a mean half-life (T 1/2 ) of 19.7 days compared with 8 to 10 days for non-Fc modified IgGs in nonhuman primates (NHPs) (17) (Fig. 1B).
Systemic exposure after a single IM dose at 10 mg/kg demonstrated high relative bioavailability (111%) with a similarly long half-life (T 1/2 mean = 22.2 days) ( Fig. 1B and table S1). Thus, the LA modification endowed adintrevimab with an extended half-life in NHPs.
Because mutations that enhance binding to FcRn can adversely affect binding to human Fcγ receptors (FcγRs) and the complement component C1q (18), we evaluated the impact of the LA modification on binding to recombinant FcγRI, FcγRIIa, FcγRIIb, FcγRIIIa, and C1q. We also assessed antibody-dependent natural killer cell activation and degranulation (ADNKA), antibody-dependent cellular phagocytosis mediated by monocytes and neutrophils (ADCP and ADNP), and antibody-mediated complement deposition (ADCD) using in vitro effector function assays. ADG2 and adintrevimab displayed comparable binding affinities to human FcγRs and C1q and induced similar degrees of ADNKA, ADCP, ADNP, and ADCD activity in vitro, suggesting that the LA modification enhances FcRn binding without substantially affecting Fc-mediated effector activities (Fig. 1, C to F). Last, we confirmed that the LA modification did not substantially affect the biophysical properties of the molecule, as assessed by in vitro polyreactivity, hydrophobicity, and thermal stability assays (fig. S1) (19).

Adintrevimab demonstrates clinical efficacy against symptomatic Delta and Omicron BA.1/BA1.1 infection
On the basis of its neutralization profile, biophysical properties, and extended half-life in NHPs, we advanced adintrevimab into a phase 1 study to assess safety and PK and a phase 2/3 prevention study with a preexposure prophylaxis (PrEP) cohort (EVADE; NCT04859517) to evaluate the ability of a single 300-mg IM dose to prevent the development of symptomatic COVID-19 in SARS-CoV-2-naïve adults. In the phase 2/3 study, the primary analysis population comprised all participants who were seronegative and reverse transcription polymerase chain reaction (RT-PCR) negative at baseline. The primary efficacy end point was symptomatic COVID-19 (SARS-CoV-2 infection confirmed by RT-PCR) occurring after administration of adintrevimab or placebo through day 90. The study enrolled between 27 April 2021 and 11 January 2022, which spanned the transition from Delta to Omicron BA.1/ BA1.1 as the dominant circulating variant ( fig. S2).
Given the substantial (>100-fold) loss of adintrevimab neutralizing activity against Omicron BA.1/BA1.1 relative to the Delta variant ( fig. S3) (11), which we hypothesized would translate into a substantial reduction in clinical efficacy, the trial population was divided into two subsets for the purposes of this analysis: the "pre-Omicron population," which consisted of participants enrolled on or before 30 November 2021 and with events on or before 15 December 2021 (the date Omicron BA.1 became the predominant circulating variant in study locations), and the "Omicron population," which consisted of participants enrolled between 1 December 2021 and 11 January 2022 and with events occurring before 11 April 2022. Whole-genome sequencing (WGS) on a subset of trial participants confirmed that the vast majority (97.7%) of infections in the pre-Omicron population were caused by the Delta variant, and 90.5% of infections in the Omicron population were caused by Omicron BA.1 or BA1.1 variants ( fig. S2 and table S2).
In the pre-Omicron population, adintrevimab demonstrated a relative risk reduction of 71% versus placebo in the development of RT-PCR-confirmed symptomatic COVID-19 through 3 months, the primary end point (P < 0.0001). A post hoc analysis . Bar heights and error bars represent geometric mean FRNT 50 titers ± SD, and the geometric mean 50% FRNT 50 titer is indicated above each bar. The dotted line represents the limit of detection for Omicron BA.1 titers based on a starting serum dilution of 1:24. Samples that did not display 50% neutralization at the lowest dilution tested (1:24) are shown at half of the lowest dilution tested and were imputed to 24 to calculate geometric mean titers. Data were excluded from phase 1 study participant samples after confirmed COVID-19 infection or vaccination. Individuals receiving mRNA-1273 had no history of prior SARS-CoV-2 infection or vaccination. of a subset of pre-Omicron EVADE participants randomized before 15 June 2021 revealed an 84% relative risk reduction versus placebo in the development of RT-PCR-confirmed symptomatic COVID-19 through 6 months (P = 0.041; table S3). The increased efficacy through day 180 was driven by a greater percentage of events in the placebo group than in the adintrevimab group during months 3 through 6 as compared with months 0 through 3. Efficacy waned more rapidly in the Omicron PrEP population, as expected given the lower potency of adintrevimab against this variant. Here, we observed a relative risk reduction of 60, 41, and 37% versus placebo in the development of RT-PCR-confirmed symptomatic COVID-19 through days 28, 60, and 90, respectively (P = 0.026, 0.02, and 0.032, respectively; table S3).

Adintrevimab mediates protection against symptomatic Omicron BA.1/BA1.1 infection at low serum neutralizing titers
The notable difference in serum neutralizing activity conferred by adintrevimab against Delta and Omicron BA.1/BA1.1 provided us with the opportunity to potentially determine a threshold degree of serum neutralization associated with protection against symptomatic COVID-19 in the absence of other forms of immunity. Because participants in the primary and exploratory efficacy analysis populations were primarily infected with Delta and Omicron BA.1/BA1.1, respectively, we normalized our population PK model-derived median serum concentrations of adintrevimab to authentic virus neutralization data for Delta and Omicron BA.1/ BA1.1 to project serum nAb titers against these variants over time, where serum neutralization titer = adintrevimab serum concentration/variant half-maximal inhibitory concentration (IC 50 ) (11) (table S4 and Fig. 2A). On the basis of this analysis, adintrevimab-conferred serum neutralization against the Delta variant peaked on day 8 with a median neutralizing titer of 1:6157 and declined to a median titer of 1:987 on day 360 ( Fig. 2A). Because adintrevimab neutralizes Omicron BA.1 and BA1.1 with about 180-fold lower potency than Delta (fig. S3) (11), the projected serum neutralizing titers had a proportionally lower peak (1:34 on day 8) against Omicron BA.1/BA1.1 and declined to below 1:20 titers by day 120 ( Fig. 2A). To validate the calculation used for converting serum concentrations to neutralizing titers, we also experimentally measured authentic virus serum neutralizing titers in all participants in our phase 1 study, which showed that the normalized and measured serum neutralizing titers were within threefold for both Delta and Omicron BA.1 ( fig. S4). However, many serum samples did not reach 50% neutralization against Omicron BA.1 at the lowest dilution tested (1:24). For comparison, we also measured vaccineinduced serum neutralizing titers against Delta and Omicron BA.1 in 12 healthy adult donors who had received a second dose of an mRNA vaccine (mRNA-1273) 7 to 30 days before sampling (tables S5 and S6). In this cohort, serum neutralizing titers ranged from 1:260 to 1:1024, with a geometric mean titer of 1:478 against the Delta variant (Fig. 2B). Consistent with prior studies (11), we observed lower vaccine-induced serum neutralizing titers against the Omicron BA.1 variant, which ranged from less than 1:24 to 1:53 with a geometric mean titer of 1:23 (Fig. 2B). We concluded that administration of a 300-mg dose of adintrevimab resulted in serum neutralizing titers of greater than 1:500 against the Delta variant for at least 1 year, with geometric mean titers at 6 months exceeding peak titers achieved after two doses of an mRNA vaccine, whereas the same dosing regimen resulted in low serum neutralizing titers against Omicron BA.1, even at peak concentrations.

Serum nAb titers predict protection in both passive monoclonal antibody prophylaxis and vaccination settings
Passive transfer experiments in animal models have demonstrated that certain antiviral antibodies are unusually potent in protection relative to their neutralization abilities (7,20). To determine whether adintrevimab protects at lower serum neutralizing titers than polyclonal nAbs induced by vaccination, we investigated the relationship between geometric mean serum neutralizing titers induced after ChAdOx1 or bNT162b2 vaccination, as measured in the same authentic virus focus reduction neutralization test (FRNT) assay used to calculate normalized adintrevimab serum neutralizing titers (11,21,22), and protection against symptomatic COVID-19 reported in phase 3 clinical trials or real-world vaccine effectiveness studies (table S7) (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37). Despite uncontrolled variables across the vaccine studies, we observed a strong correlation between geometric mean 50% FRNT (FRNT 50 ) titers and ]. However, as observed in prior vaccine CoP studies (3,8), the shape of the protection curve suggests that protection increases gradually with neutralization titer, with no absolute threshold above which protection is achieved.
To test the predictive utility of the model, we used normalized adintrevimab serum neutralizing titers to predict clinical efficacy against Delta and Omicron BA.1/BA1.1. The model predicted an efficacy of 92.8% for adintrevimab against Delta on day 180, which was similar to the observed efficacy in our prevention trial, where individuals in the adintrevimab arm experienced an 84.4% relative risk reduction in the development of symptomatic disease through 6 months ( fig. S7). Similarly, the model predicted an efficacy of 51.6% (95% CI, 44.4 to 58.8%), 46.7% (95% CI, 38.8 to 54.5%), and 41.5% (95% CI, 33.4 to 49.7%) against Omicron BA.1 on days 28, 60, and 90, respectively. These predicted efficacies were within the range of the observed efficacies of 60, 41, and 37% at a median follow-up of 28, 60, and 90 days, respectively ( fig. S7). Inclusion of the observed adintrevimab efficacy and normalized serum neutralizing titer data in the CoP model did not substantially change the 50% protective neutralizing titer calculated using only vaccine data ( Fig. 4 and fig. S6). Extrapolation of serum adintrevimab concentration over time demonstrated that adintrevimab-conferred serum neutralizing titers against the Delta variant would remain over this 50% protective neutralization titer for about 3 years ( Fig. 2A), suggesting that potent, half-life-extended monoclonal antibodies have the potential to provide prolonged protection against susceptible SARS-CoV-2 variants.

DISCUSSION
Our study demonstrates that serum nAb titers on the order of 1:30 are protective against symptomatic COVID-19 in both vaccination and monoclonal antibody prophylaxis settings. Although prior vaccine studies have consistently shown that spike protein-specific serum antibody concentrations are correlated with protection (2,3,8), the observed clinical efficacy conferred by adintrevimab against symptomatic Delta and Omicron BA.1/BA1.1 infection in SARS-CoV-2-naïve individuals provides strong evidence that nAbs are mechanistic in protection and that relatively high degrees of protection can be achieved at low serum nAb titers even in the absence of other forms of immunity such as T cell and memory B cell responses. However, the protective serum neutralization threshold defined here is not absolute, because breakthrough infections still occurred at time points associated with very high titers of serum nAb. Similar results have been observed in the context of respiratory syncytial virus and influenza virus, where the probability of infection decreased with increasing concentrations of antibody; however, breakthrough infections still occurred at high serum nAb titers, suggesting that an absolute threshold of protection does not exist (38,39). The reasons for this are not well understood but may be related to the mucosal nature of these infections. In contrast, for infections in which viremia is key to pathogenesis (such as smallpox, measles, and polio), protective thresholds can be identified because sufficient concentrations of serum antibody prevent dissemination of the pathogen through the bloodstream. Reported median vaccine-induced serum neutralizing titers and normalized monoclonal antibody neutralizing titers (adintrevimab), measured in an authentic virus neutralization assay (11,21,22), are plotted against reported efficacy in phase 2/3 clinical trials or real-word vaccine effectiveness studies at the time points indicated in the legend (23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37). The brown solid line indicates the best fit of the nonlinear regression, and the yellow shading indicates 95% confidence intervals. Data points and error bars represent means ± SD. The neutralizing titer associated with 50% protection against symptomatic COVID-19 is indicated by a vertical dotted red line. Omi, Omicron; Geo., geometric.
Because adintrevimab displays intact Fc effector functions, we cannot determine the relative contributions of neutralization versus Fc-mediated activities to the observed protection in our clinical trial. However, studies in animal models have shown that RBDdirected monoclonal antibodies do not require Fc-dependent effector functions for optimal protection in prophylaxis settings (40). In support of this notion, the monoclonal antibody combination AZD7442 (tixagevimab and cilgavimab) demonstrated similar clinical efficacy against symptomatic Delta infection as adintrevimab despite containing Fc modifications that abolish Fc-mediated effector function (9,41). Nevertheless, we acknowledge that Fc-dependent effector functions may play a role in the protection afforded by adintrevimab and may be especially important in protection by non-RBD antibodies, such as those targeting the S2 subunit (42).
The high degree of efficacy conferred by adintrevimab against symptomatic Delta infection through 6 months, combined with PK modeling showing maintenance of neutralizing titers associated with at least 50% protection against symptomatic disease for about 3 years, suggests that broad, highly potent, and half-life extended monoclonal antibodies have the potential to offer more durable protection than currently available vaccines. This advantage may be especially notable in the context of antigenically divergent variants, such as Omicron, given the limited durability of protection conferred by vaccination. However, the rapid evolution of SARS-CoV-2 has also presented challenges for monoclonal antibody therapies, with most of the previously approved SARS-CoV-2 antibodies demonstrating limited to undetectable neutralizing activity against current Omicron subvariants (43)(44)(45). Thus, the future challenge will be in the development of next-generation antibody therapeutics that recognize functionally constrained and antigenically invariant epitopes, such as the highly conserved S2 stem helix region (42,46,47), or in modifying the current regulatory paradigm such that potent but relatively narrow-spectrum RBD antibodies can be developed and deployed at a pace that keeps up with viral evolution.
Our report has several limitations. First, the protection afforded by systemically delivered monoclonal antibodies, such as adintrevimab, is likely mediated by IgG transudated from the serum into the mucosal surfaces of the upper respiratory tract. Thus, although we have defined a serum nAb titer associated with protection against symptomatic COVID-19, it is currently unclear how serum titers translate to protective mucosal antibody titers. Second, the prolonged duration of protection afforded by adintrevimab against the Delta variant is due, in part, to its unusually long serum halflife of 145 days. Other half-life extended monoclonal antibodies may have shorter serum half-lives, which may reduce the durability of protection. Furthermore, we found that the NHP model was unable to accurately predict the degree of half-life extension of adintrevimab in humans. The likely reason for this discordance is that FcRn-antibody binding constants do not scale with body weight using traditional allometry, as suggested previously (48). Third, we were unable to determine the relative contributions of neutralization versus Fc-mediated effector activities to the observed protection in our clinical trial, and it is possible that the relative role of Fcdependent effector functions in protection may vary across monoclonal antibodies. Last, because of the limited number of hospitalizations and deaths in the EVADE trial, we were unable to assess the relationship between serum neutralizing titer and protection against severe disease.
In conclusion, the results presented here provide strong evidence that neutralizing antibodies are a mechanistic CoP against symptomatic COVID-19 and that relatively high degrees of protection can be achieved at low serum neutralizing titers. The observed relationship between serum nAb titers and protection against symptomatic disease could be used to define surrogate clinical end points, allowing for the accelerated approval of variant-based vaccines and potentially next-generation monoclonal antibodies.

Study design
ADG-001-01 is a phase 1, randomized, double-blind, single-ascending-dose study to evaluate the safety, tolerability, PK, and immunogenicity of a single IM or IV dose of adintrevimab or placebo administered to healthy participants while confined to the clinical unit. A total of 60 participants were enrolled in the study across six cohorts. In each cohort, 10 participants were randomized to receive adintrevimab (eight participants) or placebo (two participants). A sentinel group of two participants (one adintrevimab and one placebo) was administered study medication first, before advancing to the other eight for higher-dose cohorts not previously evaluated. A Safety Review Committee reviewed short-term safety and tolerability data for the preceding cohorts before a dose increase was made. Participants were followed for up to 12 months after receipt of study medication. Ex vivo serum neutralizing activity of adintrevimab against SARS-CoV-2 was also assessed. EVADE (NCT04859517) is a phase 2/3, multicenter, doubleblind, placebo-controlled, randomized study of adintrevimab in the prevention of symptomatic COVID-19 in adults and adolescents with no known history of SARS-CoV-2 infection but whose circumstances place them at increased risk of acquiring SARS-CoV-2 infection and developing symptomatic COVID-19. This objective was independently evaluated in a cohort of participants with reported recent exposure to an individual diagnosed with an SARS-CoV-2 infection (cohort A; postexposure prophylaxis) and in a cohort of participants with no reported exposure to SARS-CoV-2 (cohort B; PrEP). These cohorts were enriched for participants whose advanced age (≥55 years old) or health status places them at risk for severe COVID-19 or COVID-19 complications.

Trial participants
The PrEP cohort consisted of adolescents (12 to 17 years of age) and adults more than 18 years of age who had an increased risk of exposure to SARS-CoV-2 owing to vocational or social circumstances. All participants were required to have a negative SARS-CoV-2 serologic test result at screening. Participants were excluded if they had a history of SARS-CoV-2 infection, a positive SARS-CoV-2 result at screening, or previous receipt of a vaccine or biologic agent indicated for the prevention of SARS-CoV-2 infection. All participants were randomized 1:1 to receive adintrevimab or placebo.

End point
The primary objective for the PrEP cohort was to evaluate the efficacy of adintrevimab compared with placebo in the prevention of RT-PCR-confirmed symptomatic COVID-19 through 6 months in participants with negative SARS-CoV-2 tests (RT-PCR and serology) at baseline as assessed by the proportion of participants with RT-PCR-confirmed symptomatic COVID-19 through 3 months. Secondary or exploratory end points included RT-PCR-confirmed symptomatic COVID-19 through 6 months, PK analyses, and safety analyses.

Viral testing, WGS, and variant analysis
Virological testing, WGS, and bioinformatics analysis of SARS-CoV-2 variants for the EVADE clinical study were performed at Eurofins Viracor BioPharma. A quantitative real-time RT-PCR assay was used to detect SARS-CoV-2 infection from nasopharyngeal and saliva specimens during illness visits (49). For WGS, nucleic acid extraction for respiratory specimens was performed using the KingFisher (Thermo Fisher Scientific) with the GSD No-vaPrime RNA Extraction (AE1) Kit (Eurofins). Extracted RNA was converted into cDNA using LunaScript RT Supermix (New England Biolabs). After reverse transcription, cDNA was amplified using ARTIC SARS CoV2 Primer Pools (Eurofins Genomic Laboratories). PCR reactions from the pools were combined and subjected to magnetic bead clean up. The concentration of amplicons in each sample was quantified using the Qubit FLEX Fluorometer (Thermo Fisher Scientific) and Qubit 1X dsDNA HS reagents (Thermo Fisher Scientific) and normalized. Preparation of libraries was performed using the NEBNext Ultra II FS library prep kit (New England Biolabs) in conjunction with the BRAVO liquid handling platform (Agilent). After normalization, amplicons underwent enzymatic fragmentation and end repair using the NEBNext Ultra II FS library prep kit (New England Biolabs). After end-repair fragmentation, AT-tailed uracil-linked hairpin adapters were ligated onto amplicons. After adapter ligation, USER enzyme was added to each reaction to cleave the uracil linker in the hairpin of the adapter molecules. After USER digestion, automated purification of the library reactions was performed using the BRAVO liquid handler (Agilent) and SPRIselect magnetic beads (Beckman Coulter). Index PCR unique dual-indexed primers (New England Biolabs) were used to barcode and amplify each individual library. After index PCR, automated purification of the library reactions was performed using the BRAVO liquid handler (Agilent) and SPRIselect magnetic beads (Beckman Coulter). The mass of each purified indexed library was quantified using the Qubit FLEX Fluorometer (Thermo Fisher Scientific) and Qubit 1X dsDNA HS reagents (Thermo Fisher Scientific) and normalized. Libraries were pooled in equal volumes. The fragment size distribution of the final pooled library was confirmed using the TapeStation 4200 (Agilent) before preparation for sequencing. Pooled libraries were denatured, diluted, and sequenced on the Illumina NextSeq 500/ 550 instrument using a NextSeq Mid Output flow cell and reagents running a 2 × 150 cycle paired-end sequencing protocol.
Raw sequencing data (bcl files) were demultiplexed and converted to fastq files with bcl2fastq from Illumina. The fastq files were analyzed with the Eurofins Genomics SARS-CoV-2 normal goat serum bioinformatic pipeline that trimmed primer sequences and filtered low-quality reads, mapped high-quality reads to the reference genome (MN908947), called single-nucleotide variants and small insertions and deletions, generated a consensus sequence, and assigned SARS-CoV-2 lineages using pangolin (cov-lineages. org) (50).

PK of adintrevimab in NHPs
A non-good laboratory practice NHP study was conducted at WuXi AppTec to determine the serum PK properties of adintrevimab after a single IV infusion or IM injection administration. Twelve naïve female cynomolgus macaques, aged 32 to 34 months, were divided into two groups with three animals per group (animals supplied by Laboratory Animal Co.). Animals in the IV group were administered adintrevimab by a single 60-min IV infusion administration at 10 mg/kg. Animals in the IM group were administrated adintrevimab by a single IM injection at 10 mg/kg. Blood samples were collected at predose (0) and at 1 (immediately after 1-hour infusion), 6,24,48,96,144,168,192,336,504,672,1008,1344,1680,2016, and 2352 hours after dose. Concentrations of adintrevimab in NHP serum samples were determined by a qualified enzyme-linked immunosorbent assay. The serum concentration-time profiles of adintrevimab in study animals were analyzed using a noncompartmental PK method [Phoenix WinNonlin software (version 6.3; Pharsight)]. The protocol was approved by Wuxi AppTec (Suzhou) Co. Ltd. Institutional Animal Care and Use Committee (reference number SZ20200529-Monkeys).

Bioanalytical assay to measure adintrevimab concentration in NHP sera
Quantification of adintrevimab in NHP sera was performed using a qualified bioanalytical immunoassay method at WuXi AppTec. SARS-CoV-2 spike protein (RBD from GenScript) was coated onto a microplate, and NHP serum samples, quality control samples (QCs), or standards were added to coated wells. After incubation, the serum was removed, wells were washed, and mouse anti-human IgG Fc-horseradish peroxidase detection antibody was added. 3,3′,5,5′-Tetramethylbenzidine substrate (SeraCare), including low pH stop, was used for detection, followed by absorbance reading at 450/630 nm. Concentrations of adintrevimab in samples and QCs were determined by back-calculation to a standard curve.

Bioanalytical assay to measure adintrevimab concentration in human sera
Quantification of adintrevimab in human sera was performed using a validated bioanalytical method at Q2 Solutions. Adintrevimab was isolated from serum-based calibration standards, QCs, and participant samples by immunoprecipitation in which samples were combined with biotinylated monoclonal anti-adintrevimab (antiidiotypic antibody), which specifically binds adintrevimab in samples. Streptavidin beads were used to capture the biotinylated antibody-adintrevimab complex. Adintrevimab was then digested with trypsin to yield signature peptides specific to adintrevimab. A stable isotope-labeled signature peptide internal standard was then added to the calibration standards, QCs, and participant samples. All samples were assayed using liquid chromatography/ tandem high-resolution mass spectrometry. All concentration calculations are based on the peak area ratio of adintrevimab signature peptide to the internal standard. Concentrations of the analyte in QC and participant samples were determined by back-calculation from the calibration curve.

Phase 1 clinical study serum neutralization assay
Clinical samples were evaluated for serum neutralization of SARS-CoV-2 using a microneutralization assay at Viroclinics Biosciences. The sources of SARS-CoV-2 isolates included Delta [hCoV-19/ USA/MD-HP05647/2021 (BEI Resources NR-55672)] and Omicron BA.1 [hCoV-19/Netherlands/NH-RIVM-72291/2021 (European Virus Archive)]. Briefly, virus was preincubated with serial dilutions of serum; then, the virus-serum mixtures were added to a confluent monolayer of Vero-E6 cells [American Type Culture Collection (ATCC)]. After 1 hour of incubation, cell medium was replaced, and cells were incubated for an additional 15 to 23 hours. Cells were then fixed with formalin, followed by immunostaining using a primary antibody targeting the viral nucleocapsid protein and a peroxidase-conjugated secondary reagent. Virus-positive cells were detected using TrueBlue substrate and quantified using an ImmunoSpot analyzer (Cellular Technology Limited). The 50% neutralization titers were calculated according to the method described by Zielinska et al. (51).
Population PK model development A population PK model was developed for adintrevimab using serum adintrevimab concentration-time data from the two clinical trials and a phase 1, first-in-human study (ADG20-1-001). The development of the population PK model involved three steps: (i) construction of the base structural and statistical model, (ii) covariate analysis to identify participant descriptors associated with the interindividual variability in PK, and (iii) evaluation and qualification of the final model. Base structural model development was initially conducted using the phase 1, intensive PK sampling data only and consisted of the fitting of one, two, and three compartmental population PK models with linear elimination and first-order drug absorption. The base structural model was then fit to the pooled phase 1 and phase 2/3 data and refined as necessary to assure a robust fit across all participants. The potential for targetmediated drug disposition was evaluated as necessary on the basis of comparisons of the observed data in SARS-CoV-2-infected participants relative to those who were not infected. After an appropriate base structural model was identified, population PK covariate model development was undertaken using forward selection followed by a backward elimination procedure. The resultant final population PK model was then evaluated for potential revisions, such as the removal of extraneous covariate relationships or modifications to the interindividual and residual variability models. The model was then qualified by performing a prediction-corrected visual predictive check, which graphically examines the agreement between the 5th, 50th, and 95th percentiles of the observed and the individual simulated concentrations across time intervals. The population PK analysis was conducted using NONMEM software version 7.4 (ICON Development Solutions) implementing the first-order conditional estimation method with interaction. Model-based simulations were conducted by converting the NONMEM model code to C++ code so that simulations could be conducted using mrgsolve, a package for R statistical software that facilitates simulations from differential equation-based models.

FcγR binding studies
Surface plasmon resonance analysis was conducted at 25°C in a HBS-EP + buffer system [10 mM Hepes (pH 7.4), 150 mM NaCl, 3 mM EDTA, and 0.05% Surfactant P20] using a Biacore 8K optical biosensor equipped with a CAP sensor chip (Global Life Sciences Solutions USA). The sample compartment was maintained at 10°C for the duration of the experiment. This assay orientation allows for reproducible capture of biotinylated samples to the sensor surface. Before each analysis, the sensor chip surface was first conditioned with three pulses (60 s at 10 ml/min) of regeneration solution (6 M guanidine-HCl in 0.25 M NaOH). Each experiment cycle began with an injection (300 s at 2 ml/min) over flow cells 1 and 2 of a 1:20 solution of Biotin CAPture Reagent (Global Life Sciences Solutions USA, catalog no. 29423383) in HBS-EP + buffer. This was followed by an injection (180 s at 10 ml/min) of the biotinylated RBD antigen over flow cell 2. Upon capture of the antigen to the sensor surface, ADG2 or adintrevimab was injected (180 s at 30 ml/min) over flow cells 1 and 2. The dissociation of the IgG was monitored for 120 s before injection (180 s) of FcγR. Dissociation of the FcγR from the sensor surface was monitored for 180 s. Last, an injection (120 s at 10 ml/min) of regeneration solution over flow cells 1 and 2 prepared the sensor surface for another cycle. The data were first cropped to include only the steps that involve the FcγR association and dissociation. These selected data were then aligned and double reference-subtracted, and then nonlinear least squares were fit to a 1:1 binding model using Biacore Insight Evaluation software version 3.0.11.15423.
FcRn binding studies SPR analysis was conducted at 25°C using a Biacore 8K optical biosensor equipped with either a CM3 or CM5 sensor chip (Global Life Sciences Solutions USA). The sample compartment was maintained at 10°C for the duration of the experiment. These studies were conducted in an HBS-EP + buffer system (10 mM Hepes, 150 mM NaCl, 3 mM EDTA, and 0.05% Surfactant P20) at pH 6.0 or pH 7.4. A pH scouting study helped determine the buffer pH and approximate concentration for direct immobilization of each IgG to the sensor surface. The sensor surface was prepared as follows: A 1:1 mixture of EDC and NHS was injected (420 s) over flow cells 1 and 2; the antibody was injected (120 s) over flow cell 2, and last, ethanolamine was injected (420 s) over flow cells 1 and 2. Each experiment cycle began with an injection (180 seconds at 30 ml/minute) of FcRn over flow cells 1 and 2. The dissociation of the FcRn was observed for 180 s before the sensor surface was regenerated through two injections (20 s at 30 ml/min) of HBS-EP + buffer (pH 7.4), which prepared the sensor surface for another cycle. The data were aligned and double reference-subtracted, and then, nonlinear least squares were fit to a 1:1 binding model using Biacore Insight Evaluation software version 3.0.11.15423.
C1q binding studies BLI analysis was conducted at 25°C in phosphate-buffered saline (pH 7.4) with 0.1% bovine serum albumin (PBSF) buffer system using a ForteBio Octet HTX (Sartorius Bioanalytical Instruments) equipped with streptavidin sensor tips. These assay conditions were adapted from the work of Zhou et al. (52). The sensor tips were soaked in PBSF buffer for 10 min and then exposed (60 s) to wells containing biotinylated RBD antigen. Each experiment cycle began with dipping (180 s) the sensor tip into PBSF to establish a stable baseline. This was followed by exposure (180 s) of the antigen-loaded sensor tip to wells containing the IgG. After a short dip (60 s) into fresh wells of PBSF, the sensor tip was dipped (180 s) into wells containing the C1q or blank buffer. The sensor tips were then immediately dipped (180 s) into fresh wells containing PBSF buffer to monitor (first 30 s) the dissociation of C1q from the sensor tip surface. The data were x and y axisaligned, and then, nonlinear least squares were fit to a 1:1 binding model using ForteBio Data Analysis software version 11.1.3.10.

ADCP with monocytes and neutrophils
For ADCP assays with neutrophils, HL-60 promyeloblast cells (ATCC, catalog no. CCL-240) were maintained in Iscove's modified Dulbecco's medium (ATCC, catalog no.  with 20% FBS and 1% Pen/Strep. HL-60 cells were differentiated into neutrophils by growth for 5 days in the presence of 1.3% dimethyl sulfoxide. Recombinant SARS-CoV-2 RBD protein was coupled to fluorescein isothiocyanate (FITC)-loaded fluorescent beads (Thermo Scientific, catalog no. F8819) by carbodiimide coupling. Antibodies were diluted in a fivefold dilution curve in HL-60 culture medium (1000 to 0.32 ng/ml) and incubated with RBD-coated fluorescent beads for 2 hours at 37°C. Cells (5 × 10 4 per well) were incubated for 18 hours at 37°C. Cells were then stained for CD11b (clone M1/ 70 APC-Fire750, BioLegend, catalog no. 101262; 0.2 μg per well) and CD16 (clone 3G8 Pacific Blue, BioLegend, catalog no. 302032; 0.125 μg per well), fixed with 4% paraformaldehyde, and analyzed by flow cytometry. CD11b + and CD16 + cells were analyzed for uptake of fluorescent beads. A phagocytic score was determined using the following formula: (percentage of FITC + cells) × [geometric mean fluorescent intensity (gMFI) of the FITC + cells]/100,000. Cells were analyzed on a Cytek Aurora spectral flow cytometer, and Cytek SpectroFlo software was used for data analysis.
For ADCP assays with monocytes, THP-1 monocytes were maintained in RPMI 1640 supplemented with 10% FBS, 1% Pen/ Strep, 1% L-glutamine, and β-mercaptoethanol. Recombinant SARS-CoV-2 RBD-coated beads were generated as described above. Antibodies were diluted in a fivefold dilution curve in THP-1 culture medium (5000 to 0.064 ng/ml) and incubated with RBD-coated beads for 2 hours at 37°C. Unbound antibodies were removed by centrifugation before the addition of THP-1 cells at 2.5 × 10 4 cells per well. Cells were fixed with 4% paraformaldehyde and analyzed on a Cytek Aurora spectral flow cytometer. A phagocytic score was determined as described above. Cytek SpectroFlo software was used for data analysis.

Antibody-mediated complement deposition
Recombinant SARS-CoV-2 RBD-coated beads were generated as described for ADCP assays. Antibodies were diluted in a fivefold dilution series in RPMI 1640 (5000 to 0.064 ng/ml) and incubated with RBD-coated beads for 2 hours at 37°C. Unbound antibodies were removed by centrifugation before the addition of reconstituted guinea pig complement (Cedarlane Labs, catalog no. CL4051) and diluted in veronal buffer supplemented with calcium and magnesium (Boston BioProducts, catalog no. IBB-300) for 20 min at 37°C. Beads were washed with PBS containing 15 mM EDTA and stained with a FITC-conjugated anti-guinea pig C3 antibody (MP Biomedicals, catalog no. 855385; 1:100 dilution). C3 deposition onto beads was analyzed on a Cytek Aurora spectral flow cytometer. The gMFI of FITC for all beads was measured. Cytek SpectroFlo software was used for data analysis.

Polyreactivity assay
Polyspecificity reagent binding of antibodies was performed as described previously (19). Briefly, soluble membrane protein (SMP) and soluble cytosolic protein (SCP) fractions were extracted from Chinese hamster ovary cells and biotinylated using NHS-LC-Biotin (Thermo Fisher Scientific) reagent. Yeast-presented IgGs were incubated with 1:10 diluted stock of biotinylated SMP and SCP for 20 min on ice, followed by two washes with PBSF, and stained with 50 μl of a secondary labeling mix containing ExtrAvidin-R-PE (Sigma-Aldrich; 1:50 dilution), anti-human LC-FITC (Southern Biotech; 1:100 dilution), and propidium iodide (Invitrogen; 1:500 dilution) for 15 min on ice. Cells were subsequently washed with PBSF and resuspended in PBSF for flow cytometric analysis on a BD FACSCanto II (BD Biosciences).

Fab thermal stability
Apparent melting temperatures (T m App ) of Fab fragments were obtained as previously described (19). Briefly, 20 μl of test antibody solution at 1 mg/ml were mixed with 10 μl of 20× SYPRO orange. The plate was scanned with the CFX96 Real-Time System (Bio-Rad) from 40°C to 95°C at a rate of 0.25°C/min. T m App was calculated from the primary derivative of the raw data using the Bio-Rad analysis software.

Hydrophobic interaction chromatography
Antibody hydrophobicity was evaluated using HIC as previously described (19). Briefly, test antibody samples were diluted in phase A solution [1.8 M ammonium sulfate and 0.1 M sodium phosphate (pH 6.5)] to a final concentration of 1.0 M ammonium sulfate. A linear gradient from phase A solution to phase B solution [0.1 M sodium phosphate (pH 6.5)] was run for 20 min at a flow rate of 1.0 ml/min using the Sepax Proteomix HIC butyl-NP5 column. Peak retention times were obtained by monitoring ultraviolet absorbance at 280 nm.

Statistical analysis
RT-PCR-confirmed symptomatic COVID-19 was determined by COVID-19 symptoms (defined by the protocol) occurring within 14 days from the sample collection date of a positive central or local (in the absence of central test) RT-PCR. Any COVID-19related hospitalization with a positive local SARS-CoV-2 test (within 14 days) or all-cause death is counted toward the end point. The date of event is the earliest date of COVID-19 symptom onset, COVID-19-related hospitalization, or allcause death.
The analysis of the primary estimand, incidence of RT-PCRconfirmed symptomatic COVID-19 for adintrevimab versus placebo, was analyzed using the methodology for determining a standardized estimator for a binary outcome with adjustment for the prognostic factors. The standardized risk difference-associated P value and 95% CI, as well as the standardized relative risk reduction with 95% CI, were provided for efficacy assessment.
A treatment policy strategy was used to handle the intercurrent events (ICEs) of interest in the primary analysis. The ICEs included the use of rescue medications (such as COVID-19 vaccines or COVID-19 mAb treatments for the purposes of prevention) or participant unblinding by the investigator before the primary end point outcome. Participants with missing primary end point outcome data were imputed as not having the primary end point outcome in the primary analysis.

Supplementary Materials
This PDF file includes: Fig. S1 to S7 Table S1 to S7 References (49)(50)(51)(52)(53) Other Supplementary Material for this manuscript includes the following: Data file S1 MDAR Reproducibility Checklist View/request a protocol for this paper from Bio-protocol.