Humoral Immune Correlates Analysis of Four Vaccines Against SARS-CoV-2 in Rhesus Macaques

We assessed correlates of protection (COP) in a vaccine trial of 126 Rhesus macaques randomized to placebo or one of four vaccines (mRNA-1273 [Moderna], Ad26.COV2.S [Janssen], NVX-CoV2373 [Novavax], and CoV2 preS dTM+AS03 [Sanofi]) at varying doses and challenged with SARS-CoV-2 (USA-WA1/2020 strain). Four immune marker assay readouts (live virus microneutralization assay, Meso scale discovery electrochemiluminescence immunoassay, and pseudovirus neutralizing assay using 50% or 80% inhibitory dose) measured at two time points (2 weeks prior to challenge and day of challenge) were assessed as correlates of protection. Protection was defined as ability to predict reduction in one of 12 measures of viral load (two RT-qPCR assays [subgenomic (E protein) and genomic (N1)], two time periods [two days post challenge and area under the curve (AUC) from challenge to the last day of sample collection], and three locations [nasal swab, oropharyngeal swab, and bronchioalveolar lavage]). There was high correlation among all immune marker measurements (all correlations ≥0.86) and viral load measurements (all correlations≥0.87). In 28 of 96 models predicting viral load from an immune marker, adjusted R-squared was ≥0.4 (maximum=0.52). In most of those 28 models, viral load was measured using nasal swab and summarized by AUC. In 56% of cases, a simple model predicting viral load from an immune marker explained >80% of variability compared to models that additionally contained dose parameters for different vaccines. Further, the observed viral load for animals receiving each vaccine showed good agreement to their predicted values modelled only using data from animals receiving the other three vaccines (agreement coefficients between 0.25-0.59). These analyses suggest that neutralizing titers and antibody binding levels are correlates of protection that predict reduction in viral load across these four different vaccines.