Sensitive and specific serological ELISA for the detection of SARS-CoV-2 infections

Reagents were obtained from following manufacturers: Advansta Corporation (San Jose, USA): WesternBright Sirius®; Carl Roth GmbH & Co. KG (Karlsruhe, Germany): Carbenicillin disodium salt, lysozyme (≥ 45 000 FIP U/mg), ROTI®Fair Carbonate Bicarbonate Buffer pH 9.6, ROTI®Stock 10 × PBS, ROTI®Stock 10 × PBS-T, sodium chloride (≥ 99.5%), sodium dodecyl sulfate (SDS, ≥ 99.5%), sulfuric acid, Terrific-Broth- (TB-) Medium and urea (> 99.5%); GenScript Biotech BV (Leiden, Netherlands): SARS-CoV-2 Nucleocapsid protein (His-tag), SARS-CoV-2 Nucleocapsid Protein mAbs; MORPHISTO GmbH: Acetate buffer pH 5.0; Promega GmbH (Mannheim, Germany): Peroxidase-conjugated anti-human IgG antibody; Roche Deutschland Holding GmbH (Mannheim, Germany): cOmplete™ Mini EDTA-free protease inhibitor cocktail (from bovine pancreas); Seramun Diagnostika GmbH (Heidesee, Germany): TMB substrate solution; SERVA Electrophoresis GmbH (Heidelberg, Germany): Acrylamide/bis(acrylamide) (30% T, 2.67% C), BlueBlock PF 10x, Coomassie Brilliant Blue G-250, TEMED and trypsin (sequencing grade, MS approved); Sigma Aldrich Chemie GmbH (Taufkirchen, Germany): 2-mercaptoethanol (BioUltra), Antifoam Y-30 emulsion and imidazole (≥ 99.5%); Surmodics IVD, Inc. (Eden Prairie, USA): StabilZyme™ SELECT; Thermo Fisher Scientific (Waltham, Massachusetts, USA): AcroMetrix™ Inhibition Panel, SuperBlock^® (PBS).

The coding sequence of SARS-CoV-2 N-protein (NCBI accession # YP_009724397.2, downloaded from https://​www.​ncbi.​nlm.​nih.​gov/​protein) containing a N-terminal His-tag was codon-optimized for Escherichia coli, synthesized, cloned into a pET45b(+) vector (GenScript Biotech BV), and expressed in E. coli BL21(DE3). Briefly, an overnight culture from TB-Medium containing Carbenicillin (0.5 g/L) and Antifoam Y-30 emulsion (0.0075%, v/v) was inoculated with a single colony of E. coli BL21(DE3) pET45b(+)-N-protein and incubated on a horizontal shaker (180 rpm, Certomat® BS-T, B. Braun Biotech GmbH, Melsungen, Germany) at 30 °C for 17 h. Cells were harvested, resuspended in lysis buffer (PBS (337 mmol/L NaCl, 2.7 mmol/L potassium phosphate, 10 mmol/L disodium hydrogen phosphate, 2 mmol/L potassium dihydrogen phosphate) containing 50 µmol/L lysozyme and 1 tablet/10 mL protease-inhibitor-mix), and disrupted on a French® Press (Thermo Fisher Scientific, Waltham USA). For purification by immobilized metal affinity chromatography the protein concentration determined by NanoPhotometer NP80® (Implen GmbH, München, Germany) was adjusted with purification buffer (PBS, pH 7.4) to 10 g/L and imidazole added to obtain a final concentration of 25 mmol/L. The sample was loaded and proteins were eluted using a linear 25-min gradient from 25 mmol/L–0.5 mol/L imidazole in purification buffer. Fractions were collected in 1.5-min intervals and analyzed by SDS-PAGE. Fractions containing mostly the N-protein were combined, dialyzed against storage buffer (PBS), and stored at -20 °C. The purity of the N-protein was verified by SDS-PAGE. The most intense band expected to contain the N-protein was excised from the gel and the protein digested with trypsin (in-gel digest). The extracted tryptic peptides were analyzed by LC–MS. Additionally, proteins in the gel were semi-dry electroblotted onto a PVDF membrane (Bio-Rad Laboratories GmbH, Feldkirchen, Germany), blocked with BlueBlock solution at room temperature for 1 h and incubated with an anti-SARS-CoV-2 Nucleocapsid protein antibody (GenScript Biotech BV) diluted to 10,000-fold in BlueBlock solution at room temperature. After 1 h, peroxidase-conjugated anti-Human IgG-HRP (Promega GmbH, 1: 30,000 in BlueBlock) was added at room temperature. After 1 h, the bands were visualized with WesternBright™ Sirius substrate solution and the chemiluminescence was recorded on ChemiDoc MP (Bio-Rad Laboratories GmbH).

Serum samples from PCR-confirmed SARS-CoV-2 positive patients were obtained from two individual donors and hospitalized patients (Krankenhaus Nordwest, Frankfurt, Germany, Klinikum Chemnitz gGmbH, Chemnitz, Germany, Institut für Transfusionsmedizin, Universitätsklinkum Leipzig, and Hospital St. Georg gGmbH, Leipzig, Germany) (Additional file 1: Tables S1–S4). These investigations represent parts of the analyzes in the COVID genetics cohort Leipzig-Chemnitz, which was approved by the Institutional Review Board of Leipzig University (reference numbers 195/20-ek and EK-allg-37/10–1). Control serum samples collected from 2009 to 2014 were enriched for older people, higher BMI, and a history of chronic obstructive pulmonary disease (COPD) to specifically test the assay specificity in populations at higher risk for severe COVID-19 conditions and with presumably a higher incidence of previous viral infections including other coronaviruses, although this information was not available (Additional file 1: Table S5). These 1500 serum samples obtained from the population-based LIFE-Adult study of the Leipzig Research Center for Civilization Disease (LIFE) [10]. All samples have been processed and stored by the team of the Leipzig Medical Biobank. Probands were collected all over the year and thus should well represent different antibody titers in response to seasonal bacterial and viral infections, e.g., influenza, coronaviruses, and rhinoviruses, as well as allergies to achieve a robust performance in epidemiological studies. Furthermore, serum samples from persons with confirmed antibody titers against human immunodeficiency viruses (HIV) 1/2, parvovirus B19, hepatitis A/B virus, cytomegalovirus, Epstein Barr virus, and herpes simplex virus were used for cross-reactivity studies (INSTAND, Düsseldorf, Germany).

Medium binding microplates (Greiner Bio-One, Frickhausen, Germany; 12xF8, PS, F-bottom) were coated with 150 ng SARS-CoV-2 N-protein per well in PBS at 4 °C overnight. All following steps were performed at room temperature. Wells were washed three times with PBS-T (300 µL) using a Hydro Flex ELISA washer (Tecan Group AG, Männedorf, Switzerland) and blocked with superblock (200 µL) for 60 min. Human serum (off-the-clot, sterile filtered; PAN-Biotech GmbH, Aidenbach, Germany) was used as negative control. The positive control consisted of an anti-SARS-CoV-2 N-protein antibody (1 g/L, GenScript Biotech BV) 100-fold diluted in the negative control human serum. Both controls and the serum samples were diluted 100-fold in sample buffer (Indical Bioscience, Leipzig, Germany) and incubated for 45 min. Wells were washed with PBS-T (300 µL) using the Hydro Flex ELISA washer before conjugate solution (100 µL/well) was added (anti-Human IgG-HRP, 1:30,000 in Stabilzyme Select). After 30 min, wells were washed three times as described above and TMB substrate solution added (100 µL/well). The reaction was stopped after 10 min by the addition of sulfuric acid (0.3 mol/L; 100 µL/well) and the absorbance recorded at 450 nm using a SUNRISE microplate reader (Tecan Group AG, Männedorf, Switzerland).

Additionally, serum samples were tested with a commercial anti-SARS-CoV-2 ELISA (IgG) kit (SARS-CoV-2 N-protein based ELISA, EUROIMMUN Medizinische Labordiagnostika AG, Lübeck, Germany) according to the manufacturer’s instruction.

To test potential interference, a positive serum sample was 128-fold diluted with plasma containing different interfering substances: hemoglobin (up to 20 g/L), bilirubin (up to 0.3 g/L), and triglycerides (up to 15 g/L).

The absorbance values of serum samples were converted to sample-to-positive (S/P-) ratios using the absorbances recorded at 450 nm (OD₄₅₀) of the positive (PC) and negative controls (NC), using the following equation:

The cut-off value was determined by Receiver Operating Characteristics (ROC) to obtain the best sensitivity and specificity. ROC analysis was performed by GraphPad Prism 9.0.2 (Graph Pad Software, La Jolla, CA, USA).

Control sera collected before 2015 but unexpectedly tested positive in the N-protein ELISA, were probed by an immunoblot. Commercial and in-house expressed N-proteins were separated by SDS-PAGE (0.5 µg per lane) using 12% gels. Gels were stained with brilliant Coomassie Blue G250 or semi-dry electroblotted onto a PVDF membrane (Bio-Rad Laboratories GmbH) [7]. Briefly, the membrane was blocked with BlueBlock solution at 4 °C overnight and incubated with SARS-CoV-2 positive sera, negative sera or potentially false-positive sera (1:10,000) at room temperature for one hour. After three washing steps with PBS-T for 5 min, the membrane was incubated with peroxidase-conjugated anti-human IgG-HRP (Promega GmbH, 1: 20,000 in BlueBlock) at room temperature for one hour. After washing, bands were detected with WesternBright™ Sirius substrate solution and the chemiluminescence was recorded on ChemiDoc MP (Bio-Rad Laboratories GmbH).

The SARS-CoV-2 N-protein was successfully expressed at high levels with N-terminal His-tag at a basal level without IPTG induction. After affinity chromatography, a single band was detected in SDS-PAGE at the expected apparent molecular weight at ~ 47 kDa (Fig. 1), which was confirmed by an immunoblot using an anti-N-protein antibody and after in-gel digestion by LC–MS. When larger quantities were loaded, a few weak bands appeared at apparent molecular weights of ~ 30 kDa, which were identified as C-terminally truncated N-proteins by LC–MS. At -20 °C the purified protein was stable in PBS containing 0.2 mol/L NaCl and 0%, 10% or 25% glycerol for the tested period of 360 days (Additional file 1: Fig. S1). When stored in the same buffers at 4 °C for 180 days, several bands appeared in SDS-PAGE below an apparent molecular weight of 47 kDa and at around 100 and 150 kDa (Additional file 1: Fig. S1).

The indirect ELISA for anti-SARS-CoV-2 IgG antibodies was initially established using a commercial N-protein with “research grade” purity to accelerate the assay development, while the N-protein was expressed and purified in-house. First, coating of the N-protein was evaluated for carbonate buffer (pH 9.6), PBS (pH 7.4), and acetate buffer (pH 5.0) using an anti-N-protein IgG, a SARS-CoV-2 positive serum, and a negative control serum (Additional file 1: Fig. S2a). While carbonate buffer provided the highest absorbance for positive samples, the best ratio between positive and negative samples was obtained for PBS. Considering the OD₄₅₀-values obtained for coating different protein quantities (Additional file 1: Fig. S2b) and the efforts and costs to produce pure N-protein, 150 ng of N-protein were coated per well. Using this condition, the best performance among the tested microtiter plates was observed for Maxisorp (Nunc, Roskilde, Denmark) and medium binding (Greiner) based on the ratios of the OD₄₅₀-values measured for the anti-N-protein IgG antibody and the negative serum pool (Additional file 1: Fig. S2c). All following experiments used medium binding plates and a 30,000-fold dilution of the secondary anti-human IgG-HRP antibody in Stabilzyme Select (Additional file 1: Fig. S2d). The robustness and reproducibility of the ELISA was warranted by including positive and negative control samples to normalize the OD₄₅₀-values percentage-wise.

The optimized ELISA using the purified in-house N-protein was probed with 68 sera collected at least 14 days after symptom onset and confirmation of a SARS-CoV-2 infection by PCR along with 1500 sera collected in Germany from 2009 to 2014 considered to be SARS-CoV-2 and SARS-CoV-1 negative. A ROC curve analysis provided a sensitivity of 89.7% and a specificity of 99.3%. The cut-off value was 30% (Table 1, Additional file 1: Fig. S3) with the uncertain region (grey zone) ranging from 20 to 30%, which was considered as negative. Among the 1500 control sera were ten sera tested positive in the ELISA. When seven of the ten samples that were available in sufficient quantities were probed in an immunoblot against the N-protein, no bands were detected. As neither the N-protein nor any contamination, such as an E. coli protein, were recognized (Additional file 1: Fig. S4), we considered these samples as false positive in ELISA.

As the N-protein was coated in ELISA using non-denaturing conditions but probed in the immunoblot after being denatured by SDS-PAGE, the seven false-positive samples were tested again in ELISA, but this time after coating the plate with denatured N-protein (Additional file 1: Fig. S5). Compared to the N-protein ELISA using native coating conditions, the absorbance decreased on average for positive samples by ~ 7% and ~ 65% for the false-positive and borderline samples (up to 90% for one false-positive sample). Thus, denaturing conditions identified 50% of the false-positive and 78% of borderline samples improving the specificity to 99.7%. An RNase/DNase digestion of purified N-protein prior to denaturation did not provide further improvement (data not shown).

Serum samples collected from persons infected with SARS-CoV-2 were assorted based on the time period passed since confirmation of the disease by a positive PCR result, i.e., sera collected within the first six days (group 1, n = 23), from day 7 to day 13 (group 2, n = 22), and on day 14–55 (group 3, n = 68). Our ELISA correctly identified nine sera of group 1 (39.1%), 19 sera of group 2 (86.4%), and 61 sera of group 3 (89.7%) as positive (Table 1, Fig. 2). The specificity of 99.3% was maintained with only 10 of 1500 negative samples being repeatedly recognized as positive (Table 1, Fig. 2). Patients with a history of COPD and smoking as well as gender and obesity had no significant effect on the false-positive rates (P > 0.05; Table 2).

The seven serum samples from group 3 tested negative for IgG antibodies against the SARS-CoV-2 N-protein with the in-house ELISA were also negative when tested with the CE-labelled (European Economic Area) anti-SARS-CoV-2 NCP IgG ELISA (Euroimmun) (Fig. 3a). This confirms previous reports indicating that anti-N-protein IgG antibodies cannot be detected by ELISA in sera from some patients despite a PCR confirmed SARS-CoV-2 infection [9, 14].

When the ten serum samples tested false positive in our ELISA were tested with the CE-labelled anti-SARS-CoV-2 NCP IgG ELISA, four samples remained positive and six were negative (Fig. 3b). It should be noted that the ten samples tested here were selected from initially 1500 sera based on false positive results of our in-house assay. It was not our intention to determine the specificity of the commercial assay for all 1500 control samples, but we randomly selected 30 positive sera from group 3 tested with high, middle, and low OD₄₅₀-values (10 sera per OD-range) in our ELISA and 33 of the 1490 remaining negative control serum samples. While the commercial ELISA confirmed all positive samples, one control sample negative in our in-house ELISA was detected as positive (Fig. 4). It should be noted that the resulting specificity of 96.9% for the commercial ELISA, which is well below the specificity of 99.8% reported by Euroimmun (Additional file 1: Table S6), should not be over-interpreted due to the low case numbers of the random sample set. Interestingly, both assays showed a similar trend in the calculated normalized absorbance and ratios from high to low readouts.

As the humoral immune system responds to a SARS-CoV-2 infection with IgA production before IgG is secreted [25], we tested also the 23 sera from group 1 and 29 randomly selected negative sera for IgA titers using the conditions of the optimized anti-N-protein ELISA but a secondary anti-IgA antibody. Only six sera were positive for IgA (26.1%) including four that were also positive for IgG (Additional file 1: Fig. S6). Considering that only two additional serum samples were correctly identified as positive, we did not further investigate an IgA-based ELISA.

As hemoglobin, bilirubin, and triglycerides in serum and plasma samples may interfere with the ELISA, plasma samples were spiked with these substances at different concentrations. The OD₄₅₀-values between the collected and the corresponding spiked plasma samples varied by less than 20% indicating that the in-house ELISA was not affected by these substances in the concentration range to be expected in human sera (Additional file 1: Fig. S7).

Finally, 26 serum samples with antibody titers confirmed for various other diseases, such as HIV 1/2, parvovirus B19, hepatitis A/B virus, cytomegalovirus, Epstein-Barr virus, and herpes simplex virus, and collected before 2019 were tested negative by the in-house N-protein based ELISA (Additional file 1: Fig. S8).

For nine patients, follow-up samples were tested showing the seroconversion, increasing antibody titers during the second week after PCR confirmation, and relatively stable antibody titers for at least 54 days (Additional file 1: Fig. S9). Only the antibody titer of one patient decreased from day 9 but it remained positive for 65 days (Additional file 1: Fig. S9, light blue line).