Rapid homogeneous assay for detecting antibodies against SARS-CoV-2

LFRET incubation time, cutoff values and performance

LFRET assays for SARS-CoV-2 S and N were set up using Eu-labeled in-house antigens and AF-labeled protein L. First, the assay conditions were optimized separately for S and N using three known anti-S/-N ELISA-positive and three known anti-S/-N ELISA-negative samples (included in the full 129-sample panel). Thereafter the remaining 123 samples were tested in the optimized conditions. For detection of both anti-S and anti-N antibodies measurement at 7 minutes was found optimal.

Cutoffs for both anti-S and anti-N LFRET were set by measuring LFRET signals relative to buffer in 48 samples tested negative by anti-S and anti-N ELISA. The average plus four standard deviations (SD) was set as cutoff: 228.37 + 4 × 27.59 = 338.76 counts for anti-S and 220.94 + 4 × 27.73 = 331.86 counts for anti-N LFRET.

Performances of the anti-S and anti-N LFRET assays were then determined with the 129 samples including 77 sera or heparin/EDTA plasmas from 44 individuals with a previous RT-PCR-confirmed SARS-CoV-2 infection, four samples from four individuals negative for SARS-CoV-2 by both RT-PCR and serology, and 48 samples from individuals with a comprehensively negative SARS-CoV-2 serology. The sensitivities and specificities of SARS-CoV-2 anti-S and anti-N LFRET in detection of PCR-positive individuals were 87% and 100% & 78% and 100%, respectively. The combined anti-S/-N LFRET sensitivity and specificity were 90% and 100%: if either anti-S or -N LFRET was positive, the composite result was considered positive (Table 1). The development of LFRET signals over time among patients with follow-up samples available is shown in Figure S1.

ELISAs and microneutralization

In order to compare the performance of LFRET with classical serology, we tested the set of samples described above with SARS-CoV-2 anti-S and anti-N IgA, IgM and IgG ELISAs as well as SARS-CoV-2 microneutralization. The panel of seronegatives was excluded from anti-N IgA and IgM as well as anti-S IgM ELISAs. Altogether 107 samples underwent microneutralization, including 64 samples from RT-PCR positive patients and 43 seronegative samples. Microneutralization titers of ≥20 were considered positive.

The ELISA cutoffs were set at average plus four standard deviations of absorbances measured from 14 serum samples from SARS-CoV-2 seronegative Department staff members.

The sensitivities and specificities of ELISA for anti-S IgA, IgG and IgM in samples from SARS-CoV-2 RT-PCR-positive individuals were 91% and 98%, 90% and 100% & 66% and 100%, respectively. The corresponding sensitivities and specificities of anti-N ELISA were 75% and 100% (IgA), 92% and 94% (IgG) & 16% and 100% (IgM), respectively (Figure S2). Pearson correlation between anti-S and anti-N IgG ELISAs was 0.90, 0.79 between IgM ELISAs and 0.31 between IgA ELISAs.

Comparison of LFRET, ELISA and microneutralization

Comparison between the LFRET signals and ELISA absorbances is presented in Figure 2. For anti-N antibodies, the correlation between LFRET and IgA or IgM ELISA results was low (R=0.25 for IgA and R=0.13 for IgM). With IgG ELISA a stronger correlation of R=0.62 was seen, apparently hampered by saturation of the ELISA signal. For anti-S-antibodies, correlations between IgA, IgG and IgM ELISAs were R=0.52, R=0.62 and R=0.56, respectively. Higher LFRET signals were seen in samples from patients with severe disease, especially in anti-S LFRET and when samples taken less than two weeks from onset were excluded (Figure S3). The agreement between anti-N ELISA and LFRET was 88-89%, and that between anti-S ELISA and LFRET 96-98% (Table S1). The samples representing discordance between PCR, LFRET and/or ELISA are detailed in Table S2.

• Download figure
• Open in new tab

Figure 1. Simplified protocol for SARS-CoV-2 N and S LFRET assay. Eu-N/-S = Europium-labeled nucleocapsid protein/spike glycoprotein. AF-L = Alexa Fluor™ 647 - labeled protein L. TR-FRET = time-resolved Förster resonance energy transfer. RT = room temperature. TBS+BSA (50mM Tris-HCl, 150mM NaCl, pH 7.4, 0.2% BSA) was used for all dilutions. On-plate dilutions were 5 nM Eu-N/500 nM AF-L/serum 1/25 for anti-N and 5 nM Eu-S/250 nM AF-L/serum 1/100 for anti-S LFRET. For further details see the prior publication⁵.

• Download figure
• Open in new tab

Figure 2. ELISA (x-axis) vs. LFRET (y-axis) results by disease severity. a) Anti-N IgA ELISA vs. anti-N LFRET (N=81, R=0.25). b) anti-N IgG ELISA vs. anti-N LFRET (N=129, R=0.62). c) anti-N IgM ELISA vs. anti-N LFRET (N=81, R=0.13). d) anti-S IgA ELISA vs. anti-S LFRET (N=129, R=0.53). e) anti-S-IgG ELISA vs. anti-S LFRET (N=129, R=0.62). f) anti-S IgM ELISA vs. anti-S LFRET (N=81, R=0.56). Colour of the dot indicates SARS-CoV-2 PCR result and disease severity: cyan = PCR negative; yellow = non-hospitalized, PCR-positive; red = non-ICU hospitalized, PCR positive; black = hospitalized in ICU, PCR positive. Horizontal and vertical black lines indicate LFRET and ELISA cutoffs. On the X-axis, ELISA absorbance on a logarithmic scale and on the y-axis, LFRET signal on a logarithmic scale. S = spike glycoprotein. N = nucleocapsid protein. LFRET = protein L–based time-resolved Förster resonance energy transfer immunoassay. ELISA = enzyme immunoassay. R = Pearson’s correlation coefficient.

The LFRET and ELISA results are compared with microneutralization titers in Figure 3. Spearman’s rank correlation coefficient (rho) between anti-S LFRET and MNT was 0.87, whereas for anti-S IgG, IgA and IgM ELISAs it was 0.68, 0.86 and 0.81, respectively. Likewise, between anti-N LFRET and MNT Spearman’s rho was 0.83, while 0.81, 0.69 and 0.61 for anti-N IgG, IgA and IgM ELISAs. Higher neutralization titers were observed in samples from hospitalized individuals. Sensitivities and specificities of LFRET and ELISA in correctly identifying microneutralization-positive and -negative samples are shown in Table S3.

• Download figure
• Open in new tab

Figure 3. Microneutralization vs. LFRET and ELISA. Microneutralization titers are on the x-axis and LFRET signal or ELISA absorbance on the y-axis. Logarithmic scale is used on both axes. a) Microneutralization titer vs. anti-S LFRET signal (N=107, ρ=0.87). b-d) Microneutralization titer vs. anti-S IgG, IgA and IgM ELISA (N=107, 107 and 67, ρ=0.68, 0.86 and 0.81). e) Microneutralization titer vs. anti-N LFRET signal (N=107, ρ=0.83). f-h) Microneutralization titer vs. anti-N IgG, IgA and IgM ELISA (N=107, 67 and 67, ρ=0.81, 0.69 and 0.61). Colour of the dots indicate SARS-CoV-2 PCR result and disease severity: cyan = PCR negative; yellow = non-hospitalized, PCR-positive; red = non-ICU hospitalized, PCR positive; black = hospitalized in ICU, PCR positive. Horizontal black lines indicate LFRET/ELISA cutoffs. S = spike glycoprotein. N = nucleocapsid protein. LFRET = protein L–based time-resolved Förster resonance energy transfer immunoassay. ELISA = enzyme immunoassay. ρ = Spearman’s rank correlation coefficient.

Receiver operating characteristic analysis

We also rated the performances of both LFRET assays using receiver operating characteristic (ROC) curves (Figure S4). The curves were plotted in RStudio (version 1.3.1073) with the ROCR library (version 1.0-11). With the assumption that a positive result in either IgM, IgG or IgA ELISA for a given sample signifies ‘true’ positivity, the respective areas under the curve (AUCs) for both anti-N and anti-S LFRET assays were very high, 0.94 and 0.97.

Samples

This study included 77 serum/plasma samples from 40 individuals tested positive and four samples from four individuals tested negative for SARS-CoV-2 by RT-PCR from nasopharyngeal swab samples. The positive samples were taken at 8 to 81 days after onset of symptoms. Additionally, 48 serum samples from asymptomatic individuals with a comprehensively negative SARS-CoV-2 serology (Euroimmun IgG, IFA IgG virus, IFA IgG spike, microneutralization negative) were included in the study. The data and samples were collected under research permit HUS/211/2020 and ethics committee approval HUS/853/2020 (Helsinki University Hospital, Finland).

Nucleic acid testing

Nucleic acid testing for SARS-CoV-2 was done from nasopharyngeal swab samples with either the Cobas® SARS-CoV-2 test using the Cobas® 6800 system (Roche Diagnostics, Basel, Switzerland), a protocol based on one previously described by Corman et al.²³, or the Amplidiag® COVID-19 test (Mobidiag, Espoo, Finland).

Molecular cloning

For protein expression, we acquired the ZeoCassette Vector (pCMV/Zeo) from ThermoFisher Scientific, and excised the Zeocin resistance gene from the vector using FastDigest EcoRI and XhoI (ThermoFisher Scientific) according to manufacturer’s protocol. The excised gene was agarose gel purified, blunted using T4 DNA polymerase (ThermoFisher Scientific), and purified using Ampure XP beads (Beckman Coulter) both following the manufacturer’s protocol. The selection gene was inserted into pCAGGS/MCS and to the pCAGGS vector bearing SARS-CoV-2 S protein (described in ^24,25) gene by treating the plasmids with FastDigest SapI/LguI (ThermoFisher Scientific) according to manufacturer’s protocol, followed by blunting and purifications as above. The insert was ligated to the plasmids using T4 DNA ligase (ThermoFisher Scientific) according to manufacturer’s protocol, the ligation products transformed into Escherichia coli (DH5a strain), followed by plating the bacteria onto LB plates with 100 µg/ml of ampicillin and 50 µg/ml Zeocin (ThermoFisher Scientific). After overnight incubation at 37 °C, single colonies were picked and grown in 5 ml of 2xYT medium supplemented with 100 µg/ml of ampicillin and 50 µg/ml Zeocin overnight at 37 °C. The plasmids were purified using GeneJET Plasmid Miniprep Kit (ThermoFisher Scientific), and plasmids bearing the insert identified by restriction digestion (FastDigest EcoRI, ThermoFisher Scientific) and agarose gel electrophoresis. For both constructs, clones with the insert in reverse and forward direction were selected for ZymoPURE II Plasmid Maxiprep Kit (ZymoResearch) preparations done following manufacturer’s guidelines. A synthetic SARS-CoV-2 NP gene under Kozak sequence and a signal sequence MMRPIVLVLLFATSALA flanked by KpnI and SgsI restriction sites was obtained from ThermoFisher Scientific. The SARS-CoV-2 NP cassette was subcloned into pCAGGS/MCS-Zeo-fwd vector and plasmid maxipreps prepared as described above.

Protein expression and purification

We initially attempted producing SARS-CoV-2 S protein in Expi293F cells utilizing the Expi293 Expression System (ThermoFisher Scientific). Briefly, the pCAGGS plasmid bearing codon-optimized SARS-CoV-2 spike²⁴ was transiently transfected into Expi293F cells as advised^24,25, except that we used spinner flasks (disposable 125 ml spinner flask, Corning) for the culture. Protein purification from supernatants collected at five days post transfection followed the protocol described²⁵ and yielded 0.2-0.3 mg per 100 ml culture, in line with earlier reports^24,25.

Next, we transfected adherent HEK293T cells with pCAGGS-SARS-CoV-S-Zeo plasmids using Fugene HD at 3.5:1 ratio, in suspension as described²⁶. The transfected cells were plated onto six-well plates, and at 48 h post transfection subjected to Zeocin selection, 150 µg/ml Zeocin in high glucose DMEM (Dulbecco’s Modified Eagle’s Medium, Sigma) supplemented with 5% fetal bovine serum (Gibco) and 4 mM L-glutamine. Two days after initiating the selection, the cells were trypsinized and transferred into fresh wells, with fresh media and antibiotics provided at 2 to 3 day intervals. Once confluent, the cells were trypsinized, counted (TC20 cell counter, Bio-Rad), diluted to ∼30 cells/ml, and dispensed onto 96-well plates, 100 µl per well. After confluency, we switched to serum-free FreeStyle 293 Expression Medium (ThermoFisher Scientific) with 100 µg/ml Zeocin, and incubated the cells at 37 °C 5% CO2. At 48 h, we analyzed the medium for the presence of SARS-CoV-2 S protein by dot blotting, briefly 2.5 µl of the supernatant was dried onto a nitrocellulose membrane, the membrane blocked (3% skim milk in Tris-buffered saline with 0.05% Tween-20), washed, probed with rabbit anti-RBD (Sino Biological, 40592-T62), washed, probed with anti-rabbit IRDye800 (LI-COR Biosciences), washed, and read using Odyssey Infrared Imaging System (LI-COR Biosciences). The clone with the highest amount of SARS-CoV-2 S protein in the cell culture supernatant, HEK293T-spike-D5, was then expanded in DMEM with 5% FBS, 4 mM L-glutamine and 100 µg/ml Zeocin, and ampouled for storage in liquid nitrogen. We adapted the HEK293T-spike-D5 cells for suspension culture by placing trypsinized cells into a spinner flask with Expi293 Expression Medium (ThermoFisher Scientific) with 100 µg/ml of Zeocin. We stored an aliquot of the adapted cells in liquid nitrogen and tested their ability to produce SARS-CoV-2 S protein in both Expi293 and FreeStyle 293 Expression Medium (both ThermoFisher Scientific). Having been cultured in the spinner flask for 5 to 8 days, the cells reached a density of >3 × 10⁶ cells/ml. The protein from the supernatants was purified using the protocol described²⁵, with yields of 0.8-1.2 mg per 100 ml culture.

For production of SARS-CoV-2 N protein, we transfected Expi293F cells with pCAGGS-SARS-CoV-2-NP-Zeo using Fugene HD, briefly, 100 µg of the plasmid diluted into 10 ml of OptiMEM (Sigma), 350 µl of Fugene HD added followed by mixing and 15 min incubation at room temperature, after which the plasmid mix was added onto Expi293F cells in Expi293 Expression Medium at 2.5 × 10⁶ cells/ml. The incubation of the transfected cells was continued for four days, after which the supernatant was collected, and the protein purified (yield approximately 1 mg per 100 ml culture) as described for S protein²⁵. The remaining cells were treated with 0.25% Trypsin-EDTA (Sigma) to remove dead cells, and after two washes were put back into the spinner flask with fresh Expi293 Expression Medium supplemented with 100 µg/ml of Zeocin. Eventually a population of cells started to proliferate (Expi-NP-zeo cells), and were aliquoted in liquid nitrogen. The protein production and purification occurred as described above with yields of ∼1 mg per 100 ml culture.

Protein labeling

We labeled the SARS-CoV-2 S and N proteins with the donor fluorophore europium (Eu) using QuickAllAssay Eu-chelated protein labeling kit (BN Products and Services Oy) according to the manufacturer’s instructions to generate Eu-labeled S (Eu-S) and N (Eu-N). We also labeled recombinant protein L (Thermo Scientific) with the acceptor fluorophore Alexa Fluor 647 to generate AF647-labeled protein L (AF-L), as reported⁵. IgG-free bovine serum albumin (BSA) was from Jackson ImmunoResearch Inc.

TR-FRET assays

The LFRET assay was done as described⁵ and as briefly illustrated by the flowchart in Figure 1. For calculating the relative TR-FRET signal increase, we here replace the pool of negative sera with TBS-BSA (50mM Tris-HCl, 150mM NaCl, pH 7.4, 0.2% BSA). To establish LFRET assays for S and N, we optimized the component concentrations by cross-titration using three positive and three negative serum samples. For detection of anti-N antibodies, we found the optimal on-plate dilution for serum to be 1/25, and the optimal on-plate concentrations for AF-L and Eu-N to be 500 nM and 5 nM, respectively. For detection of anti-S antibodies, we found an on-plate dilution of 1/100 for serum, AF-L concentration of 250 nM and Eu-S concentration of 5 nM optimal. We measured TR-FRET at 0, 7, 15, 22, 30, 45 and 60 minutes after combining the reagents. TR-FRET counts were measured with Wallac Victor² fluorometer (PerkinElmer) and normalized as described³.

Enzyme linked immunosorbent assays (ELISAs)

We set up the SARS-CoV-2 S protein ELISA as described²⁵ with the following amendments. We coated the plates (ThermoScientific NUNC-immuno 446442 polysorp lockwell C8) with 50 ul/well of antigens diluted 1 µg/ml into 50 mM carbonate-bicarbonate buffer pH 9.6 (Medicago AB) and used 1-Step Ultra TMB-ELISA Substrate Solution (ThermoFisher Scientific). As secondary antibodies we made use of polyclonal rabbit anti-human IgA-horeseradish peroxidase (HRP), -IgM-HRP, and -IgG-HRP (all from Dako) at respective dilutions of 1:5000, 1:1500, and 1:6000. The colorimetric reaction was terminated by addition to 0.5 M sulphuric acid (Fluka), and the absorbances recorded (HIDEX Sense) at 450 nm. The N protein ELISA followed the same protocol.

Microneutralization

For the SARS-CoV-2 microneutralization assay we first cultured Vero E6 cells on 96-well plates (Thermo Scientific) overnight at +37°C in 2% MEM (Eagle Minimum Essential Media [Sigma-Aldrich] supplemented with 2% inactivated fetal bovine serum [Thermo Scientific], 2 mM L-glutamine [Thermo Scientific], 100 units penicillin, and 100 µg/ml streptomycin [Sigma-Aldrich]). The following day we made a two-fold dilution series (1:20 to 1:1280) of the serum samples in 2% MEM and combined 50 µl of each dilution with 50 µl of virus (1000 plaque forming units (pfu)/ml in 2% MEM). The serum-virus mixes were kept for 1h at +37°C. The cells were inoculated with the serum-virus mixes and grown in incubators at +37°C. After 4 days, the cultures were fixed with formalin, stained with crystal violet and the neutralization titers recorded.