Coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has become a global pandemic with continued transmission [1, 2]. Since there are currently no effective treatments for COVID-19, considerable efforts are focused on developing vaccines and therapeutic drugs [3]. However, the dynamics of humoral immune responses of COVID-19 patients using different serological assay platforms are largely unclear.

A wide range of SARS-CoV-2 binding antibody assays have been developed with different antigen targets and assay formats [4]. These assays can detect either isotype-specific antibodies (IgG, IgA, IgM) or total antibodies based on different assay principles such as chemiluminescent immunoassay (CLIA), enzyme immunoassay (EIA), lateral flow immunoassay (LFIA), and microsphere-based antibody assay. Most of these assays mainly target either spike (S) protein (the most exposed viral protein) or its receptor-binding domain (RBD), or nucleocapsid (N) protein, which is abundantly expressed. Although the performance and clinical utility of different binding antibody assays continue to be identified, the currently available assays show variable performance, and data on the early immune response and seroconversion are insufficient [4–6]. Many questions have also been raised about the index value of antibody assays for COVID-19 monitoring.

There is great interest in identifying SARS-CoV-2 neutralizing antibodies for measuring immune status and assessing vaccine responses. Neutralizing antibodies against both the viral S and N proteins have been found in COVID-19 patients, pointing to the potential value of SARS-CoV-2 binding antibody assays as a surrogate for neutralizing titers [7–9]. A SARS-CoV-2 surrogate virus neutralization test (sVNT) (GenScript Inc., Leiden, the Netherlands) is available for detecting neutralizing antibodies that block the interaction between angiotensin-converting enzyme 2 (ACE2) receptor protein and the RBD. However, limited data are available correlating the results of commercial assays with the presence of neutralizing antibodies detected by the sVNT.

We compared the serological characteristics and seroconversion ability in serial serum samples from COVID-19 patients using 12 commercial antibody assays: three automated, high-throughput CLIAs [SARS-CoV-2 IgG assay (Abbott, Chicago, IL, USA), Elecsys Anti-SARS-CoV-2 assay (Roche, Basel, Switzerland), and SARS-CoV-2 Total assay (Siemens, Munich, Germany)]; three EIAs [COVID-19 ELISA IgM+IgA and COVID-19 ELISA IgG (Vircell Microbiologists, Granada, Spain), anti-SARS-COV-2 ELISA (IgA) and anti-SARS-COV-2 ELISA (IgG) (Euroimmun AG, Lübeck, Germany), and Platelia SARS-CoV-2 Total Ab (BioRad Laboratories, Hercules, CA, USA)]; five LFIAs [ichroma COVID-19 Ab (Boditech Med Inc., Gangwon-do, Korea), STANDARD Q COVID-19 IgM/IgG Combo assay (SD Biosensor Inc., Gyeonggi-do, Korea), PCL COVID-19 IgG/IgM Rapid Gold (PCL Inc., Seoul, Korea), SGTi-flex COVID-19 IgM/IgG (Sugentech Inc., Daejeon, Korea), and Biocredit COVID-19 IgG+IgM Duo (Rapigen Inc., Gyeonggi-do, Korea)]; and one SARS-CoV-2 sVNT (GenScript Inc., Piscataway, NJ, USA). To our knowledge, this is the first study to compare 12 SARS-CoV-2 antibody assays using various assay platforms for assessing the early antibody response, seroconversion, neutralizing capacity, and association with disease severity during the early infection period in COVID-19 patients.

Expanding the testing capacity with accurate, validated, and reliable assays is critical in the response to the ongoing COVID-19 pandemic and vaccine trials [12–14]. We compared 12 commercially available assays with various platforms and method principles. More than 100 LFIAs are now commercially available. Although these LFIAs are convenient for testing and are useful in small or emergency laboratories with limited resources, their clinical performance remains limited with various sensitivities and specificities [15]. We evaluated five LFIAs authorized by the Korean Food and Drug Administration for export using serum samples instead of fingerstick whole blood.

For samples obtained more than 14 days after symptom onset, the overall PPA ranged from 84.0% to 98.5%. The three automated CLIAs showed similar sensitivities. Our PPA results were generally higher than those reported previously [16] and were comparable for all 12 assays. This might be related to the studied population, as we used serum samples only from symptomatic patients. Approximately 40% of the samples were from patients with severe disease courses, highlighting the importance of determining whether the performance of an antibody assay is affected by disease stage or severity. Overall, IgM or IgA isotype assays detected seroconversion earlier than IgG assays. This result is consistent with previous data showing early detection of acute patients using IgM-based assays [17, 18]. Patient 3 was an 83-year-old man who was suspected of having delayed seroconversion due to an age-related decrease in immunity. Patient 10, who had a mild disease course, showed early seroconversion with IgM assays (9–11 days), although this patient showed late seroconversion with IgG assays (14–25 days). Interestingly, the three automated CLIAs did not detect antibodies until day 25 after symptom onset in patient 10. This might be because CLIAs have defined thresholds to improve the negative likelihood ratio [19]. We determined the index values of eight assays and their correlation with disease severity. Assays targeting the S protein or RBD showed a tendency of early detection of seroconversion with higher index values, which supports previous reports demonstrating earlier seroconversion and higher titers in patients with severe disease using EIAs targeting the S protein or RBD [20–22]. However, the Roche and Abbott assays targeting the N protein antibody showed a different tendency, demonstrating later seroconversion (>14 days) and lower antibody titers in patients with severe disease (Fig. 1). This finding is similar to a previous report demonstrating delayed detection of nucleocapsid antibody in severely ill patients [20]. Although the number of patients was small, these findings suggested that the antibody response might be different according to the targeting antibodies and assay method. Since ACE2 expression varies in different ethnic populations, further studies are needed to understand the factors contributing to SARS-CoV-2 antibody responses, including genetic variability, age-related variation, and comorbidities [15].

The antibody-mediated humoral immune response is critical to prevent viral infections. The most useful information is the correlation between antibody values and a metric of protective immunity such as the neutralizing capacity [23]. The current gold standard is the conventional VNT, which shows a good correlation with the neutralizing antibody titer [7, 24]. We observed good agreement rates (>83.0%) between the 11 IgG or total binding antibody assays and the sVNT and found moderate to strong correlations (r=0.517–0.874) between the index values of binding assays and neutralizing antibody results. Neutralizing antibodies are primarily against the S1 domain, S2 domain, and RBD of the SARS-CoV-2 S protein [25, 26], suggesting that antibody assays targeting these regions might be better at predicting neutralizing capacities. As expected, the Euroimmun (targeting S1) and Siemens (targeting S1 and the RBD) assays tended to show better correlations with the sVNT (r=0.874 and r=0.839, respectively) compared with the Roche and Abbott assays (r=

0.705 and r=0.650, respectively). These results are in contrast to previous reports showing similar performances across antibody assays targeting the N and S proteins [25]. We also found that patients with severe infection showed better correlations between antibody levels and sVNT results than those with mild infection. This result confirmed previous reports showing a wide range of SARS-CoV-2 neutralizing antibody titers depending on disease severity [22, 27, 28]. Due to the limitation of using the sVNT instead of the conventional VNT, further investigation is needed to verify the immune response dynamics. Antibody detection differences can be associated with patient and assay characteristics [4–6].

This study has several limitations. First, because of the emergency isolation of patients with a positive molecular testing result and mild COVID-19 in Korea, serum samples of asymptotic patients were not included, and only a small number of patient samples were tested. We were also not able to evaluate the waning antibody response in each assay because of the lack of follow-up samples of discharged or transferred patients.

In summary, different SARS-CoV-2 antibody assays showed reliable performance, demonstrating a PPA of 84.0% or greater for samples tested more than 14 days after symptom onset. All assays detected seroconversion within less than two weeks for most patients without immune complications. However, their positivity rates and seroconversion of SARS-CoV-2 antibodies varied depending on the assay kits, disease severity, and antigen target. Commercial antibody assays should be further evaluated using serial samples over time. This study contributes to gaining a better understanding of the antibody response using currently available assays in symptomatic COVID-19 patients.