Standardization of enzyme-linked immunosorbent assays for serosurveys of the SARS-CoV-2 pandemic using clinical and at-home blood sampling

Abstract

The extent of SARS-CoV-2 infection throughout the United States population is currently unknown. High quality serology is a key tool to understanding the spread of infection, immunity against the virus, and correlates of protection. Limited validation and testing of serology assays used for serosurveys can lead to unreliable or misleading data, and clinical testing using such unvalidated assays can lead to medically costly diagnostic errors and improperly informed public health decisions. Estimating prevalence and clinical decision making is highly dependent on specificity. Here, we present an optimized ELISA-based serology protocol from antigen production to data analysis. This protocol defines thresholds for IgG and IgM for determination of seropositivity with estimated specificity well above 99%. Validation was performed using both traditionally collected serum and dried blood on mail-in blood sampling kits, using archival (pre-2019) negative controls and known PCR-diagnosed positive patient controls. Minimal cross-reactivity was observed for the spike proteins of MERS, SARS1, OC43 and HKU1 viruses and no cross reactivity was observed with anti-influenza A H1N1 HAI titer during validation. This strategy is highly specific and is designed to provide good estimates of seroprevalence of SARS-CoV-2 seropositivity in a population, providing specific and reliable data from serosurveys and clinical testing which can be used to better evaluate and understand SARS-CoV-2 immunity and correlates of protection.

Figure 1:. Serology testing protocol for evaluation of SARS-CoV-2 seropositivity in a large-scale population.

Utilizing both venipuncture-derived fresh blood and dried blood spots, we have standardized a dual-antigen ELISA platform for highly specific (IgG = 100%, 95% confidence interval = 98.5 – 100) detection of SARS-CoV-2 antibodies for application in precise, large-scale serosurvey efforts.

Figure 2:. Production and sensitivity of two full spike ectodomain (Spike) and two receptor binding domain (RBD) constructs used as antigens in ELISA.

(a) Schematic of the spike and RBD constructs used to generate recombinant proteins. Abbreviations are 3C, rhinovirus 3C protease cleavage site; Strep2×2, dual Strep2 epitope tag; T7, bacteriophage T7 fibritin trimerization domain; SBP, streptavidin-binding peptide. (b) SDS-PAGE Coomassie Blue staining of purified (1) VRC spike, (2) Mt Sinai spike, (3) Mt Sinai RBD, and (4) Ragon RBD proteins. (c) Analytical size exclusion chromatography of purified VRC and Mt Sinai spike proteins. Peak elution volumes of sizing standards are noted (670 kDa – thyroglobulin, 158 kDa – gamma-globulin, 44 kDa - ovalbumin, 17 kDa – myoglobin). Inset: transmission electron microscopy of VRC and Mt Sinai spike trimers. (d) Left: Full spike ectodomain at 3 different concentrations of protein coating density for both VRC (NIAID Vaccine Research Center, blue) and Sinai (Mount Sinai, orange) constructs. Right: RBD constructs at 3 different concentrations of protein coating for both Ragon (Ragon Institute, blue) and Sinai (Mount Sinai, orange) constructs. Anti-spike or anti-RBD monoclonal recombinant antibody spiked into negative serum at 1:100, 1:1000, and 1:10000 dilution. Data are means ± SD. n = 4.

Figure 3:. Specificity of antigens against a preliminary panel of pre-2019 archival sample controls.

(ac) Spike (VRC) and RBD (Ragon) antigens tested against 100 archival controls collected from 2014 – 2018 (blue) compared with 14 SARS-CoV-2 PCR+ controls (red) for (a) IgG, (b) IgM, and (c) IgA. (d) Microsampler controls of PCR+ samples against healthy controls. (e-f) Seropositivity thresholding for (e) IgG, (f) IgM) and (g) IgA calculated from average signal intensity of archival controls. n = 100 archival controls, n = 14 nasal swab SARS-CoV-2 PCR+ patients. Manual ELISA, titer = 1:400 (serum) or 1:10 (microsampler eluate).

Figure 4:. Small-scale testing of SARS-CoV-2 seropositivity from a hard-hit community.

(a) Signal intensity (absorbance) sorted by IgG (b) Thresholding of seropositivity in small-scale test cohort for SARSCoV-2 IgG, black = archival negative controls, red = known PCR-diagnosed positive controls, green = test cohort. (c) Signal intensity (absorbance) sorted by IgM. (d) Thresholding of seropositivity in small-scale test cohort for SARS-CoV-2 IgM (e) Relation between IgM expression and IgG expression of spike and RBD antigens. (f) Seropositivity in non-symptomatic individuals that have not tested (PCR) positive for SARS-CoV-2 infection show robust IgG expression in absence of symptoms. n = 68 symptomatic donors, n = 6 non-symptomatic donors.

Figure 5:. Cross-evaluation of high seroprevalence community samples with spike antigens from other coronaviruses.

(a) IgG absorbance of spike antigens from SARS-CoV-2, previous epidemic (MERS and SARS1), and seasonal (OC43 and HKU1) coronaviruses in archival pre-2019 samples (left) and high-incidence community (right). (b) Scatter plots (with linear models and R-squared values) between IgG signal from SARS-CoV-2 spike and other coronavirus spike proteins. (c) Signal intensity for recombinant antibody control (rAB1 = spike monoclonal; rAB2 = RBD monoclonal), known SARS-CoV-2 nasal swab positive patient control (PCR+), and pre-2019 archival control (ARCH) samples for OC43, HKU1, MERS and SARS1.

Figure 6:. Simulation results showing confidence intervals for serosurvey prevalence calculations.

Each graph displays 95% confidence intervals (CI) from 1000 replications of each condition. The CIs are sorted by the lower bound. For graphs a-c the true sensitivity is 0.90 and it is estimated with 100 samples. The true specificity is 0.99 and it is estimated with sample sizes of 100 (black), 300 (red), 1000 (green). Each graph shows the different underlying true prevalence values of either 0.001, 0.01, or 0.1. The true prevalence for each simulation is given by a vertical line.