Automated Western immunoblotting detection of anti-SARS-CoV-2 serum antibodies

A first set of 27 serum samples from 27 different patients with RT-PCR-documented COVID-19 [12], collected between March and April 2020, at least 2 weeks after the onset of symptoms were incorporated as a positive control group. All of them presented with an IgG titer ≥ 1:100 using in-house indirect immunofluorescence assay (IFA) [13]. Of these, 16 serum samples were used for conventional immunoblotting including 3 samples exhibiting low (1:200), moderate (1:800), and high (1:3200) IgG titers using IFA that were used to fix optimal conditions to be used for AWB (antigen, serum, and secondary antibodies concentrations). One serum collected in 2018, before the onset of COVID-19 (having a negative RT-PCR for HCoVs on homologous respiratory specimen), was included as negative control.

As for AWB, 223 serum samples (including the 27 serum samples described above) collected from 223 different RT-PCR-confirmed COVID-19 patients between March and September 2020 were incorporated as a positive control group. Twenty-seven of these sera were tested for antibodies to the recombinant S1 protein by EUROIMMUN® SARS-CoV-2 IgG ELISA (Euroimmun, Bussy Saint-Martin, France) performed using the Elispeed DUO system (Euroimmun) according to the manufacturer’s recommendations. The ratio (AUC sample/AUC calibrator) was interpreted as follows: <0.8 negative; ≥0.8 to <1.0 undetermined; ≥1.1 positive. We considered undetermined results as negative for statistical analyses. A negative control group (37 serum samples) consisted of (i) 10 serum samples obtained less than 5 days after the onset of symptoms in patients presenting high viral loads of SARS-CoV-2 (Ct values < 20) collected in March and April 2020; (ii) 14 sera from asymptomatic healthcare workers largely exposed to the virus but exhibiting negative results for RT-PCR and serology by IFA for SARS-CoV-2 during follow-up collected in March and April 2020; and (iii) 13 sera from patients collected between April and December 2019 before the pandemic and harboring negative RT-PCR results for the 4 HCoVs in their nasopharyngeal specimens. These 37 serum samples were also all tested by ELISA. A third group of 76 serum samples was retrieved from patients diagnosed with seasonal coronavirus infections (HCoV-NL63 (n=19), HCoV-OC43 (n=21), HCoV-229E (n=8), and HCoV-HKU1 (n=28)) between April and December 2019 an and were collected at least 2 weeks after the diagnosis, of which 45 were also tested by ELISA. A fourth group of 266 sera was collected between January 2018 and February 2020 from children and adults admitted in surgery departments (n=145) and other medical units (=121) before the pandemic, of which 88 serum samples were also tested by ELISA; their HCoVs status was unknown. Altogether, 197 sera tested by ELISA included 27 sera from COVID-19-positive patients and 170 from COVID-19-negative patients. Serum samples were collected from patients visiting four university hospitals in Marseille, France. They were stored frozen at −30°C and were tested retrospectively. No sample was collected specifically for this study which was approved by our institution’s ethics committee under No.2020-024.

The SARS-CoV-2 IHUMI2 strain (lineage 20a) was used as antigen as previously described [13]. One liter of infected cells was collected and clarified by centrifugation at 700g for 10 min and by filtering the supernatant through a 0.45-μm pore-sized filter (VWR, Radnor, PA, USA) and further a 0.2-μm pore-sized filter (Thermo Fisher Scientific, Waltham, MA, USA ). Virions were then aggregated by overnight precipitation at 4°C with 10% polyethylene glycol 8000 white flake type (PEG-8000, BioUltra, SIGMA-ALDRICH, USA) and 2.2% crystalline NaCl, with gentle swirling. Precipitated virus particles were then centrifuged at 10,000g for 30 min using a SORVALL Evolution centrifuge with SLA-3000 Recent 1 fixed angle rotor pre-cooled at 4°C (Kendro Laboratory Products, Newtown, CT, USA). The pellet was resuspended with HEPES-saline (0.9% NaCl, 10 mL of 1 M HEPES, 990 mL purified water) previously vacuum-sterilized through a 0.2-μm pore size membrane, swirled in the cold HEPES-saline until dissolution to avoid using pipetting as it may hurt viral spikes at this step. The resuspended pellet was then treated with a 30% sucrose cushion in 25 × 89 mm centrifuge tubes (Ultra-Clear, BECKMAN COULTER, CA, USA). Final purification was achieved by ultracentrifugation at 100,000g for 90 min at 4°C followed by two 30-min washes with Hanks Balanced Salt Solution using SORVALL Discovery 90SE with Surespin 630 rotor (Kendro Laboratory Products). The final pellet was resuspended in 400 μL of HEPES-buffered saline and heat-inactivated at 65°C for 1h.

SARS-CoV-2 antigens diluted to 0.5 mg/mL were mixed (v/v) with 2X Laemmli Sample Buffer (Bio-Rad, Hercules, CA, USA) before a heating step of 5 min at 95°C. This preparation and a ladder were dispensed in wells shaped in a 5 % polyacrylamide stacking gel. The protein separation was then performed in a 10% polyacrylamide separating gel with a Mini Trans-blot cell device (Bio-Rad) at 160 V for 90 min. After transferring proteins from the gel to a 0.45-μm-pore size nitrocellulose membrane (Bio-Rad) at 100 V and 15°C for 90 min, the membrane was left at 4°C overnight with 5% non-fat milk powder in Tris buffered saline (TBS) with 0.5% Tween 20 (Euromedex, Souffelweyersheim, France). Blocked strips were incubated with sera diluted at 1:50 for 60 min. Three washes of 10 min were performed before a 90-min incubation of the strips with goat peroxidase-conjugated anti-human IgG/IgM/IgA (Jackson ImmunoResearch, Ely, UK) diluted 1:1000. Three washes of 10 min with TBS buffer with 0.5% Tween 20 were made. Strips were put in contact with ECL Western Blotting Substrate (Promega, Madison, WI, USA) and the reaction with secondary antibody peroxydases was revealed with a Fusion Fx chemiluminescence imaging system and analyzed with the Fusion software (Vilber, Marne-la-Vallée, France).

The Jess^TM Simple Western system (ProteinSimple, San Jose CA, USA,) is an automated capillary-based size separation and nano-immunoassay system. To quantify the absolute serological response to viral antigens, we followed the manufacturer’s standard method for 12-230-kDa Jess separation module (SM-W004). The SARS-CoV-2 antigen (1 μg/μL) was mixed with 0.1X Sample buffer and Fluorescent 5X Master mix (ProteinSimple) to achieve a final concentration of 0.25 μg/μL in the presence of fluorescent molecular weight markers and 400 mM dithiothreitol (ProteinSimple). This preparation was denatured at 95°C for 5 min. Ladder (12-230-kDa PS-ST01EZ) and SARS-CoV-2 proteins were separated in capillaries as they migrated through a separation matrix at 375 volts. A ProteinSimple proprietary photoactivated capture chemistry was used to immobilize separated viral proteins on the capillaries. Patients’ sera diluted at a 1:2 were added and incubated for 60 min. After a wash step, goat HRP-conjugated anti-human IgG/IgM/IgA antibodies (Jackson ImmunoResearch) diluted 1:500 were added for 30 min. The chemiluminescent revelation was established with peroxyde/luminol-S (ProteinSimple). Digital image of chemiluminescence of the capillary was captured with Compass Simple Western software (version 4.1.0, Protein Simple) that calculated automatically heights (chemiluminescence intensity), area, and signal/noise ratio. Results could be visualized as electropherograms representing peak of chemiluminescence intensity and as lane view from signal of chemiluminescence detected in the capillary. An internal system control was included in each run.

Protein profiles of the purified SARS-CoV-2 antigen and uninfected Vero cells were verified on silver-stained 2-D gels. As expected, the viral-specific and major-dominant proteins were N, S, S1, and S2 proteins at 42, 170, 110, and 90 kDa, respectively. All 16 serum samples collected from 16 different COVID-19 patients exhibited reactivity against the nucleocapsid and spike proteins. Parameter optimization to translate these data on AWB included an antigen concentration of 0.25 μg/μL, a serum dilution at 1:2, and a secondary antibody dilution of 1:500 (data not shown). In these conditions, AWB of positive serum samples yielded a prominent 56-kDa band interpreted as the nucleocapsid and a 217-kDa band interpreted as the spike protein (Fig. S1). Higher molecular weight values observed with AWB than with SDS-PAGE were due to the different composition of gels in the capillaries. In total, the 16 sera from COVID-19 patients tested with conventional and AWB gave similar results except that AWB failed to detect the spike protein in one sample. Furthermore, AWB yielded significantly higher S/N ratios, peak height, and area under curve for the nucleocapsid (p < 0.0001) and spike proteins (p < 0.0001) in the 27 serum samples from COVID-19 patients than in 37 serum samples collected in negative control patients (Table 1). The S/N ratio presented higher Youden Index for nucleocapsid detection, being therefore interpreted as the most pertinent parameter to interpret AWB results. Optimal threshold for the S/N ratio of 110.4 conferred a 96.3% sensitivity and 94.6% specificity for the nucleocapsid detection. Determination of a cut-off to interpret results of spike protein was not useful and could be based only on presence/absence of signal with sensitivity to 66.7% and 100% specificity (Table 2). Therefore, we further used the presence of antibodies to the nucleocapsid with S/N ratio ≥ 110.4 and/or to the spike protein, as criteria to define a positive AWB.

AWB yielded 395/602 (66%) negative and 207/602 (34%) positive serum samples (Table 3), giving an 81% sensitivity as 181/223 COVID-19 patients were positive (nucleocapsid detected in 180/223 (76%) and spike in 67/223 (30%), respectively), and a 93% specificity as 26/379 (7%) non-COVID-19 patients were positive; applying above-reported cut-off criteria (Figs. 1 and 2a). Accordingly, positive (PPV) and negative predictive values (NPV) were of 87% and 89%, respectively. Sera from COVID-19 patients were collected with a median of 13 days (range 0 to 165) after the onset of symptoms. Sensitivity was 54% among sera collected less than 10 days after the onset of symptoms and increased to 94% among sera collected more than 10 days after the onset of symptoms (Fig. 3). AWB had a 90% agreement with the herein compared ELISA assay (Cohen’s Kappa=0.64) as the latter was positive in 22/27 (81.5%) COVID-19 patients and 6/170 (3.5%) non-COVID-19 patients, yielding a sensitivity of 81.5% and specificity of 97% (Table 4).

Detailing false-positive AWB, antibodies to the nucleocapsid were detected in 3/37 (8%) negative control serum samples. Also, 9/76 (14%) of serum sampled from patients diagnosed with seasonal HCoVs reacted with the nucleocapsid which was detected in 5/28 (18%) of patients with HCoV-HKU1, 2/19 (10.5%) with HCoV-NL63, 2/21 (9.5%) with HCoV-OC43, but in none of HCoV-229E patients (Fig. 2b). In addition, one HCoV-HKU1 serum and one HCoV-NL63 serum reacted against the spike protein, increasing the number of total cross-reactions to 11/76 (14.5%) for this group. Among 266 serums sampled before the COVID-19 epidemic in France, albeit of unknown status for HCoVs, 12/266 (4.5%) reacted against the nucleocapsid but none against the spike protein.

Most cross-reactivities were detected in 46–65-year-old patients (7/63) and more than 65-year-old (4/43) patients (Fig. 2c). Cross-reactivity was more prevalent in subjects > 21 years (15/173) than in children ≤ 15 years (6/126) but this difference was not significant (p = 0.25, Fisher’s exact test).