Qualified serological assays are critical for epidemiological studies of SARS-CoV-2. Serological assays are necessary for not only understanding the nature of the immune response(s), but also allow a more accurate estimate of prevalence by including the rate of asymptomatic infections and more precise estimates of infection fatality rate, enhance contact tracing, aid evaluation of vaccine trials, and identify donors for the generation of convalescent serum/plasma therapeutics. Serology testing is debated due to the various test method performance characteristics, including choice of antigen, source of antigen, seroconversion time, and isotype switching. The highly immunogenic S protein, especially the RBD, is the target of neutralizing antibodies8,19,20. Recombinant S and RBD proteins have been produced in mammalian or insect cells, however mammalian cell produced antigens have superior performance in reactivity with SARS-CoV-2 sera21. N protein is known to be immunogenic, but N is highly homologous among coronaviruses, and therefore is expected to induce more cross-reactive antibodies. Long et al. found some COVID-19 patients with cross-reactivity to N of SARS-CoV but not S of SARS-CoV6. Seroconversion time must be considered when analyzing antibodies. Recent studies have revealed people rarely develop specific antibodies against SARS-CoV-2 within the first 7 days of symptoms. However, from day 7 to 14 IgM and IgG antibodies rapidly increase6,22,23. Therefore, to remove the timing limitation, we set our positive patient cohort for ELISA development, as RT-PCR plus antibody positive confirming all patients have seroconverted.
For our studies, we examined the same cohort of patients to understand the variation in serological responses to different SARS-CoV-2 antigens by studying IgG, IgM, and IgA to S, RBD, and N. All SARS-CoV-2 antigens were manufactured in mammalian cells and purified to a similar degree. ROC analysis revealed S and RBD having better sensitivity and specificity characteristics than N. Immunoglobulin to N gave the most varied response. IgG, IgA and IgM all demonstrated poor sensitivity and specificity and when combining IgG and IgM the combined positive predictive value (PPV) is only 29.2%, hence N is not the best antigen for SARS-CoV-2 diagnosis. In concordance with other studies, this suggests N has more cross-reactivity when compared with S and RBD and may not be the best antigen for clinical diagnosis.
The IgG, IgM or total antibodies are traditionally used for diagnosis having similar performance characteristics, however, a few studies have used SARS-CoV-2 specific IgA for diagnosis1,24. The detection of IgM antibodies may indicate a more recent infection and detection of IgG aids in examining the duration of immune response. Less is known about SARS-CoV-2 specific IgA, however, additional information gained about the IgA immune response is valuable for understanding SARS-CoV-2 infection. Overall, our analysis demonstrates that S may be the most robust detection antigen and that both—IgG and IgM are valuable immunoglobulins for diagnostic purposes. All SARS-CoV-2 infected volunteers developed IgG and IgM to S therefore no false negatives were found. However, IgG to S has a specificity of 96.6 leading to a few false positives. Analyzing IgG and IgM together yields a positive predictive value of 100%. Anti-S IgA gave a sensitivity of 94.7% therefore not all SARS-CoV-2 patients produce detectable IgA antibodies to S. IgG antibodies to RBD also gave a sensitivity of 94.7% showing some SARS-CoV-2 positive patients do not develop IgG to RBD at levels above our assay’s limit of detection.
In our study, the sensitivity and specificity of the RBD IgM and IgA ELISAs are not ideal, but this can be attributed to IgM being transiently produced in the early stages of infection only, and to the possibility that not all individuals produce detectable IgA. Other ELISA studies have shown differences in sensitivity and specificity between IgA and IgG based assays. A recent study found RBD IgA is more sensitive than RBD IgG 4–10 days after the onset of symptoms with the sensitivity of RBD IgG increasing 16 days after the onset of symptoms, thus the timing of blood sampling is important24. Thus, IgM ELISAs may be used as an additional supplemental analysis for RT-PCR to improve diagnosis of recent SARS-CoV-2 infection. IgA may be used to supplement diagnosis, and IgA response deserves further investigation.
The virus-neutralizing activity of serological samples confirm ELISA results due to neutralization specificity, identify immunoglobin levels need for “immunity”, and identify individuals for convalescent sera donation. Interestingly, we found patients that were negative or had low titers of RBD IgG but had high titers of S IgG, had neutralizing antibodies suggesting the possibility of another epitope besides RBD that produces neutralization. Furthermore, a significant correlation between S IgG and neutralizing antibodies was found in patients with low RBD titers. Taken together, these data indicate that S IgG antibodies exhibit strong neutralization capabilities against SARS-CoV-2. Although many studies have reported neutralizing antibodies against SARS-CoV-2 targeting the RBD of the S protein25,26,27,28,29, a recent publication in Science found neutralizing antibodies targeting an epitope on the N terminal domain (NTD) of the S protein that is independent of RBD in SARS-CoV-218. Collectively, this supports our findings of high titers of S IgG exhibiting neutralization against SARS-CoV-2.
We have demonstrated differences in neutralization that are dependent on higher levels of S IgG and the possibility that S IgA antibodies are neutralizing when low titers of S IgG and S RBD occur. Recent serological studies have provided further insight into IgA and IgG antibody neutralization against SARS-CoV-2. IgG and IgA are known to play important roles in protection against respiratory viral infections30,31,32. IgG is the major antibody type produced systemically, while IgA is the most prevalent type on mucosal surfaces and provides protection against respiratory pathogens30,31,32. Previous studies have demonstrated that IgA is more effective than IgG in preventing influenza infections in both mice and humans30,32,33,34,35. Furthermore, IgA may have a role in SARS-CoV infections and immunity. Intranasal delivery of SARS-CoV proteins in mice provided better protection against viral replication in the lungs than intramuscular SARS-CoV challenge30,36. Additionally, intranasal MERS-derived vaccine reported neutralizing IgA antibodies in the bronchoalveolar lavage fluid in mice30,37. A more recent study evaluated the effectiveness of a chimpanzee adenoviral vaccine encoding stabilized spike protein from SARS-CoV-2 in mice38. This study compared intranasal and intramuscular delivery of the vaccine and concluded both routes of delivery resulted in protection from SARS-CoV-2 lung infection and pneumonia in mice38. Intranasal delivery induced a robust IgA and neutralizing antibody response which uniquely protected mice from both upper and lower respiratory tract SARS-CoV-2 infection and resulted in sterilizing immunity38. On the other hand, intramuscular administration did not induce an IgA response or result in sterilizing immunity38. Comparatively our study found examples of patients with neutralizing antibodies without extremely high titers of S or RBD IgG. This indicates that virus neutralization may be occurring due to higher titers of S IgA. For example, patient 11, who has much higher serum IgA, has nearly a two-fold increase in virus neutralization when compared to patient 10 (Table 2). Furthermore, a recent study found that SARS-CoV-2 infections resulted in early elevated titers of IgA neutralizing antibodies in patient serum, and a rapid decline in serum IgA neutralizing antibodies over time30. With only two patients studied we have seen this same trend (data not shown), further studies are needed to elucidate this observation.
Healthcare workers are a subpopulation thought to be at high risk for SARS-CoV-2 infection. In May of 2020, we enrolled healthcare workers from three Louisville, KY healthcare systems for concurrent viral load and antibody detection. The worker’s self-administered a nasal swab for RT-PCR analysis and a finger stick for serum collection. We identified 2 healthcare workers positive for active viral infection and 19 positive for S or RBD IgG antibodies for a seropositive rate of 1.4%. When this study was performed, the incidence rate in Kentucky was low (4.020 per 100,000 inhabitants), hence the 1.4% is not surprising. Similar to what we found, published studies have found various positivity rates in healthcare workers ranging from 1.6 to 31.6%39,40,41,42,43,44,45,46,47,48,49,50,51. Furthermore, comparisons to overall community infections rates are not included in all studies, but range from below the community spread to several times over community reported values48,49,50,51 which is consistent with the higher rate of infection seen in the healthcare workers when compared to the community.
Virus neutralization analysis of the healthcare workers in our study revealed that 13 of the 19 positive sera contained neutralizing antibodies. Some healthcare workers did not neutralize SAR-CoV-2 in microneutralization analysis. These individuals had IgG antibodies to SARS-CoV-2 S, and conversely did not have IgA or IgM nor any immunoglobulins to RBD. Two reasons for this are the presence of cross-reactive antibodies, especially for healthcare worker 14 or low levels of IgG to S. Healthcare worker 5 is an extremely rare find because in this individual the S IgG levels do not suggest seropositivity, however, there was significant RBD IgG and the antibodies neutralized the virus very well. Further studies are needed to confirm these findings. Two workers tested positive during the study and we were able to detect presence of S IgG and IgM 6 and 7 days after viral diagnosis. One patient had a very robust immune response showing a high level of all S immunoglobulins, RBD immunoglobulins and virus neutralization. A limitation of this phase of the study is the serum sample volume due to self-administered finger sticks as 128 workers did not have sufficient sample for testing and in some positive patients there was not enough sample to complete analysis endpoint titers or repeat analysis.
In the second round of our healthcare worker study we tested 932 individuals for SARS-CoV-2 antibodies. We found that the positivity rate increased from 1.4 to 2.3% in round 2, corresponding with the increased incident rate (11.847 per 100,000 inhabitants) in July for Kentucky. Notably 78.5% (11 of 14) of the healthcare workers that returned for round 2 maintained high levels of SARS-CoV-2 antibodies. For example, healthcare worker 2 maintained high levels of S and RBD IgG, IgA, and IgM along with virus neutralization from May to July. During round 2, we had 3 healthcare workers test positive for active virus by RT-PCR. Antibodies were not detected in the finger stick of these individuals, which was taken on the same day as the nasal swab, however the individuals developed detectable antibodies on day 4, 9, and 14 after viral diagnosis. In the future we will follow up with previously positive antibody or CPR healthcare workers and track their antibody development.
A limitation of our study is that most seropositive individuals we examined to validate ELISAs had a robust immune response (ELISA AUs greater than 1.0). In dissecting Fig. 1, for all scenarios, except N IgM, there is a significant difference between negative and positive antibody response. For example, the S IgG data (Fig. 1) shows there are only three patients between 0.2 and 1.0 AU. Consequently, for our ROC analysis there is a sharp distinction between sensitivity and specificity. However, analysis of the healthcare worker study reveals that positive patients have values between 0.2 and 1.0 AU. For example, patient 18, who was PCR positive in our study, had values of 0.52 and 0.59 AU on days 6 and 7 days after infection. This gray area (area between 0.2 and 1.0 AUs) shows some individuals with neutralizing responses and some without, therefore an increase in the SARS-CoV-2 positive cohort will provide improvement to ROC analysis. In addition, it is useful to look at IgM and IgA levels to determine seropositive patients and predict neutralization activity. To determine the root cause of this gray area, further analysis is ongoing in our laboratory as we analyze more seropositive individuals. It is likely that disease severity and timing post infection impact the diagnostic performance. A recent study has shown a clear distinction between symptomatic and asymptomatic individuals, with symptomatic individuals having a statistically higher RBD IgG titer. Furthermore, the study found diagnostic sensitivity and specificity values are different when considering symptomatic and asymptomatic patients separately as well as the timing of when antibodies are measured after infection52.
Qualified serological assays will continue to be an important tool and are central to disclosing accurate information during the SARS-CoV-2 pandemic. The sample size for our positive and negative cohort are average, hence as we continue to collect patient samples we intend to update our ROC analysis for continuous improvement of ELISA threshold values and performance characteristics. A serological toolbox to measure and understand antibody responses to SARS-CoV-2 are continuously vital to understanding the current pandemic and monitoring immunity for vaccine development. This study provides valuable data on ELISA based serological testing, correlations with neutralization, understanding the broad immune response to SARS-CoV-2 and seropositive rates in healthcare workers.
