TACKLE Study Design

TACKLE (NCT04723394) was a phase 3, randomised, double-blind, placebo-controlled study, conducted across 95 sites in Europe, Japan, Latin America, and the USA, designed to assess the safety and efficacy of a single dose of AZD7442 for the prevention of severe outcomes from COVID-19, including death. Full details of the study, including primary and key secondary outcomes, were reported previously [8]. In short, between January and July 2021, this study enrolled unvaccinated, non-hospitalised adults (≥ 18 years) with mild-to-moderate COVID-19 confirmed by laboratory reverse-transcription polymerase chain reaction (RT-PCR) or antigen test on a sample collected ≤ 3 days prior to enrolment. Participants were randomised and dosed ≤ 7 days from symptom onset with a single 600-mg dose of AZD7442 (two consecutive 3-mL intramuscular injections, one each of 300 mg tixagevimab and 300 mg cilgavimab) or matched saline placebo (two consecutive 3-mL intramuscular injections). Randomisation in a 1:1 ratio was stratified, using centralised blocked randomisation, by time from symptom onset (≤ 5 vs. > 5 days), and risk of progression to severe COVID-19 (high vs. low risk; high risk included those aged ≥ 65 years, immunocompromised individuals, and those with comorbidities, such as cancer and chronic diseases). Per protocol, cases of severe COVID-19 were diagnosed by the treating principal investigator and defined as pneumonia (fever, cough, tachypnoea or dyspnoea, and lung infiltrates) or hypoxaemia (saturation of arterial blood with oxygen < 90% in room air, severe respiratory distress, or both), and a World Health Organization (WHO) Clinical Progression Scale score of 5 or more [16]. Participants, investigators, and the sponsor were blind to treatment-group assignments.

Ethical Approval

The study was conducted in accordance with the Good Clinical Practice guidelines and the Declaration of Helsinki, Council for International Organizations of Medical Sciences International Ethical guidelines, applicable International Conference on Harmonization Good Clinical Practice guidelines, and all applicable laws and regulations. The protocol, protocol amendments, and all other relevant documentation were reviewed and approved by an institutional review board or ethics committee (Supplementary Material Table S1) and are available with the previous report [8]. An independent Data Safety Management Board provided oversight throughout the study to ensure the safe and ethical conduct of the study. All participants provided written informed consent.

Objective

We examined treatment-emergent viral variants following AZD7442 treatment as an exploratory objective of the TACKLE study. With longitudinal viral genotypic analysis, we explored the relationship between treatment-emergent viral variants and in vitro changes in viral susceptibility to AZD7442, viral load, and prevalence of severe COVID-19 outcomes, including death.

Sample Collection and Handling

Nasopharyngeal (NP) swabs were taken at study visits occurring at baseline, and 3, 6, and 15 days after administration of AZD7442 or placebo. Immediately after collection, all swabs were placed in transport medium for storage at − 70 °C. Samples then were shipped to the centralised testing laboratory on dry ice where they were stored at − 70 °C until viral nucleic acid extraction.

Genotypic AnalysisNucleic Acid Extraction, Amplification, and Sequencing

Viral nucleic acids were extracted from NP swabs on a Kingfisher Flex instrument using the MagMax™ Viral/Pathogen Nucleic Acid Isolation Kit (Thermo Fisher Scientific, Waltham, MA, USA). A validated protocol for next-generation sequencing (NGS) of the SARS-CoV-2 spike gene was employed on SARS-CoV-2 RT-PCR-positive swabs to identify viral mutations and assign SARS-CoV-2 lineages at Monogram Biosciences (South San Francisco, CA, USA). Briefly, the full-length S gene (amino acids 1–1274) was amplified and sequenced in a GenoSure SARS-CoV-2 spike 2 × 150 base pair MiSeq sequencing (v2 chemistry; Illumina, San Diego, CA, USA) NGS assay. Sequence files were analysed to determine frequency of amino acid polymorphisms (consensus; reported at allele fraction [AF] ≥ 25%) and minor variants (minimum coverage > 1000×; reported at AF 3% to 25%).

SARS-CoV-2 Lineage Designation

For each sample, a consensus SARS-CoV-2 spike nucleotide sequence was derived from NGS representing variants at AF ≥ 25% with International Union of Pure and Applied Chemistry-based codes. A spike-only version of Pangolin COVID-19 lineage assigner (Hedgehog; https://github.com/aineniamh/hedgehog, version 1.0.7) was used to classify SARS-CoV-2 spike sequences to current Pango lineages or sets of lineages [17]. Variant of concern (VOC) and variant of interest (VOI) classification followed nomenclature as defined by the WHO during the study period [18, 19].

Treatment-Emergent Substitution Analysis

NGS-derived consensus SARS-CoV-2 spike nucleotide sequences were aligned to Wuhan-Hu-1/2019 SARS-CoV-2 reference sequence (GenBank accession number NC_045512) using Geneious (version 2023.0.4; Dotmatics, Boston, MA, USA). Longitudinal sequence quality control was performed by computing intra-participant Hamming’s distances (R version 4.2.0 and DescTools package version 0.99.47), and participants with at least one pairwise Hamming’s distance greater than the mean + 3 standard deviations of the intra-participant distribution (i.e. 12.21 substitutions) were retained in the baseline analysis but were removed from the longitudinal analysis, as they may have represented sample mislabelling/misassignment.

Sanger sequencing, a common method of minor variant detection, has a traditionally accepted capacity to detect minor variants at > 25% AF [20,21,22]. The technique we used in this study, NGS, has been predicted to accurately detect minor sequence variants at a 1% AF [20]. However, it has also been suggested that variants in the 1–3% AF range can only be accurately determined if the obtained sequence data are of high quality and when the variants are confirmed by a secondary method [23]. Therefore, our treatment-emergent substitution analysis was performed by intra-participant comparison of baseline to post-baseline amino acid sequences, and by computing amino acid substitutions (i.e. amino acid replacements, and in-frame insertions and deletions) at AF ≥ 25% and AF 3% to 25% in the binding sites of tixagevimab (positions 455–456, 458, 475–480, 483–489, and 493) and cilgavimab (positions 345–346, 439–441, 443–447, 449–450, 452, 484, 490, 492–494, and 499) [14]. The notation employed for treatment-emergent substitutions is of the following format: amino acid residue(s) present at baseline plus residue position, followed by amino acid residue(s) observed at the post-baseline time point.

In Vitro Microneutralisation of SARS-CoV-2 Spike Protein Pseudotyped Lentivirus

Spike protein sequences representing the major SARS-CoV-2 lineages identified at baseline, and single substitutions in the spike protein identified at baseline and treatment-emergent analyses were engineered into SARS-CoV-2 Wuhan-Hu-1/2019 + D614G spike protein pseudotyped lentiviruses and assessed for in vitro susceptibility to AZD7442 and its component mAbs via a microneutralisation assay. Generation of spike protein pseudotyped lentivirus and pseudovirus microneutralisation assays were performed as previously reported, with several modifications [24, 25]. Inserts encoding for residues 1–1254 of the Wuhan-Hu-1/2019 + D614G SARS-CoV-2 spike protein with single substitutions or of major viral variants (Supplementary Material Table S2) were designed through codon optimisation and were incorporated in the pCAGG-Sdl19 plasmid. SARS-CoV-2 spike pseudotyped lentiviral particles were generated using a third-generation human immunodeficiency virus (HIV)-based lentiviral vector system expressing luciferase. Pseudotyped lentiviral particles were produced, purified, and titrated to assess AZD7442 susceptibility in the microneutralisation assay.

In vitro susceptibility of SARS-CoV-2 variants to AZD7442 and its component mAbs was assessed using one of three versions of the microneutralisation assay (i.e. research-grade, qualified version 1.0 [pVNTv1.0], and qualified version 2.0 [pVNTv2.0]). For all three versions of this assay, serial dilutions of mAbs were prepared in a 384-well microtiter plate and pre-incubated with pseudovirus for 30 min at 37 °C. Following incubation, AD293-ACE2-ARCB cells stably expressing ACE2 (for research-grade and pVNTv1.0 assay versions) or HEK-Blue-ACE2/TMPRSS2 cells stably expressing ACE2/TMPRSS2 (for pVNTv2.0 assay version; InVivogen, San Diego, CA, USA) were added to the wells and the plates were incubated for 48 h at 37 °C. Relative quantitation of infectivity was determined by measuring the luminescence of the expressed luciferase activity as relative luminescence units on an EnVision 2105 Multimode Plate Reader (Perkin Elmer, Waltham, MA, USA) using the Steady-Glo™ Luciferase Assay System (Promega, Madison, WI, USA) for research-grade and pVNTv1.0 assay version or Bright-Glo™ Luciferase Assay System for pVNTv2.0 assay version, according to the manufacturer’s recommendations. Percent inhibition was calculated by normalisation to the virus-only control for all assays. For the research-grade assay, half-maximal inhibitory concentration (IC50) values were determined by non-linear regression using Prism software (GraphPad, version 9.0.0). The average IC50 value for each antibody was determined from a minimum of two independent experiments. The corresponding fold change IC50 against each variant was determined for each mAb individually and together (tixagevimab and cilgavimab) compared with the reference SARS-CoV-2 Wuhan-Hu-1/2019 + D614G spike protein pseudovirus. Reference IC50 was measured within each run of the research-grade assay. The assay methodology underwent a qualification based on specificity, intermediate precision, inter-assay precision, and intra-assay precision to support regulatory submission [26]. For assays performed using the qualified pseudovirus microneutralisation assay (pVNTv1.0 and pVNTv2.0), IC50 values were subsequently determined by fitting a four-parameter logistic model to the replicate relative luminescence signal obtained across a dilution of mAb concentrations and back-calculating the concentrations that give 50% inhibition between the virus only and the no virus control (X50) using ordinary least squares in RStudio, version 4.0.2 (PBC, Boston, MA, USA). The average reference IC50, established during qualification, was utilised for all qualified assay runs for fold-change calculations.

Quantitative PCR to Measure SARS-CoV-2 Viral Load

NP swab samples acquired at baseline and study day 6 that were assessed as positive in the Cobas® SARS-CoV-2 RT-PCR assay (Roche Diagnostics International AG, Rotkreuz, Switzerland) were further evaluated by quantitative RT-PCR to measure viral load using the Centers for Disease Control and Prevention’s 2019-nCoV Real-Time RT-PCR Diagnostic Panel (Centers for Disease Control and Prevention, Atlanta GA, USA). The testing was performed at Covance Central Laboratory Services (Indianapolis, IN, USA). The validation included precision, sensitivity, accuracy, and linearity. The lower limit of quantitation for the viral load assay was 3.348 log10 copies/mL (2228 copies/mL).

Statistical Analysis

Statistical analysis of viral shedding was conducted by first calculating the log-transformation of the reduction in viral shedding at study day 6 versus baseline. The reduction was then fit to a robust linear model (R version 4.1.3). A one-sided t test was used to compare the mean reduction in viral shedding between participants who exhibited treatment-emergent substitutions and participants who had no detected substitutions.