Serological studies are critical for understanding pathogen−specific immune responses and informing public health measures. A robust antibody response indicates a high adaptive immune response, with the building of immunological memory and predicting rapid recovery from repeated infection.
However, it isn't easy to arrive at such results using non-standardized tests. A new Swedish study published on the preprint server medRxiv* in July 2020 describes the results of using highly sensitive and specific anti-spike antibody tests in a cohort of patients with COVID-19 over a range of disease severity.
Study: Disease-associated antibody phenotypes and probabilistic seroprevalence estimates during the emergence of SARS-CoV-2. Image Credit: Kateryna Kon / Shutterstock
The Issues with Current Testing
The currently available data on seroprevalence suffers from the use of commercial antigens based on the viral spike protein, which can detect cross-reacting antibodies, may miss low levels of antibody, or lack the epitopes essential for specific antibody detection. The present study, therefore, is based upon an ELISA assay developed by the researchers based on a spike (S) trimer closely identical to the native protein and stabilized in the prefusion state, as well as of the receptor-binding domain (RBD) and the nucleocapsid (N) protein.
The Study: Antibodies in Patients and Community Seroprevalence
The study aimed to assess the level of antiviral antibodies in patients and key community groups. In order to arrive at a more accurate seroprevalence estimate, the researchers chose a large number of historical controls, at almost 300, consisting of blood samples collected pre-pandemic, and 20 individuals positive for endemic coronaviruses (ECV). They trained probabilistic algorithms to process the data from the patient population.
The investigators make use of the fact that Sweden allowed the pandemic to spread with relative freedom, given that no lockdown was imposed. This allows a unique understanding of how immunity develops in this population.
They analyzed IgM, IgG, and IgA levels in addition to in vitro neutralization assays, mapping them to the clinical severity. The population assayed consists of a COVID-19 patient population of 105, as well as of 900 pregnant women between weeks 14 and 25, and 1,000 blood donors.
The first represents a high-risk population as well as one that has an exceptional immune response. The second is generally active and working-age individuals, with a good grasp of health principles.
Validating the Assay
In the validation assay, all patients were positive for anti-S IgG and almost all for anti-RBD IgG at lower titers. Among the negative controls, two patients with ECV positivity were repeatedly IgM positive to the N and S protein. Two pre-pandemic (2019) donors were weakly positive for IgM to the S protein.
This showed that IgG levels differed by up to 1,000 times between positive individuals, with similar titers for anti-S and anti-N IgG. Anti-RBD titers were lower. About a tenth had no detectable IgG to the N protein, as shown in earlier studies as well.
Patient Antibody Responses
The next step was to screen all the patients, finding strong anti-S and anti-RBD IgG responses in 97% of them. In the donor and pregnant groups, antibodies were present in a subset only, and with marked variation between individuals. Some showed antibody titers comparable to those of COVID-19 patients, and some only levels comparable to the high end of the 2019 control range. These were termed, weak responders. This is the category that creates confusion in any assay that relies on a pre-determined cut-off value.
IgM and IgA titers were not as strong or as consistent as the IgG responses. However, antibody titers were significantly correlated with clinical severity within three categories, namely, non-hospitalized or category 1, hospitalized or category 2, and intensive care on mechanical ventilation, category 3.
First, anti-S titers related to anti-RBD titers across all antibody isotypes. Secondly, higher levels of all isotypes correlated to disease severity. The most significant correlation was with titers of specific IgA antibodies, with higher levels indicating increased mucosal infection.
IL-6 titers also increased with clinical severity, since this cytokine promotes antibody production and is seen to be dysregulated in metabolic diseases as well as in acute respiratory distress syndrome (ARDS), both of which increase the risk of death in COVID-19. The specific anti-RBD IgA levels were also lower in females overall.
The researchers conclude: "Severity showed the most consistent relationship with any measure and was the primary driver of Ig levels." Advancing age increased the risk of severity. IL-6 was not correlated with IgM titers, indicating that this cytokine indicates a more prolonged disease course. IgA titers and IL-6 levels were threefold and tenfold higher in severe disease compared to non-hospitalized patients. Low levels of IgA were found in a tenth of healthy donors from 2020.
IgG remained positive two months after the disease onset or dating from the first positive PCR, but IgM and IgA decreased as expected. Testing along a timeline showed that the highest levels of all three isotypes occurred along the same trajectory, though all were not always seen to develop. The isotype most often absent in category 1/2 was IgA.
Neutralizing antibody responses were observed in all patients and all antibody-positive donors over a range of titers. Antibody binding was highly correlated with neutralization titer. The highest neutralizing antibody levels were in category 3. Anti-RBD IgG titers were most strongly related to neutralizing activity.
The researchers set their cut-off value after taking into account the large number of samples with weak reactivity. They took random 20-strong negative control samples from the ~300 2019 donors and calculated the seroprevalence in both donors and pregnant women at pandemic weeks 17 to 19. They found that 5.7% to 8.7% of the samples were positive, which is a difference of over a third.
Weak responders are probably the result of the interactions of genes, previous health, total serum Ig and protein levels, and the assay used – as well as the level of antibody in the test sample. Based on the current testing, they identified seroprevalence of 7.7% in all the healthy donors tested two months after the highest mortality peak in Sweden.
The Course of the Curve
The probabilistic model used to assess the true positivity of a given sample is based on Bayesian analysis. They found an initial sharp rise in positivity, which then slowed over weeks 17 to 25, to reach about 7.2% at the last time point measured.
They compared this against machine learning approaches and found the same pattern. This was especially so when seroprevalence was predicted using the same Bayesian model, which followed the trajectory of deaths in Stockholm county, but 2.5 weeks behind.
The researchers comment, "ICU occupancy and deaths are a better proxy for viral spread than PCR+ diagnoses, which are highly dependent on the number of tests carried out."
Implications and Applications
The finding that seroprevalence can predict the course of the mortality trajectory will allow researchers to calculate the case fatality rate based on seroprevalence. The study may also help to identify individuals who should be re-tested or investigated further by identifying the probability of positivity.
The researchers conclude, "Humoral immunity to the virus develops slowly in these populations despite considerable virus spread in the community." This shows that even without lockdowns, antibody production within the population to herd immunity levels will take a long time and require much more massive levels of infection, and of course, a correspondingly higher number of severe cases and deaths.
This will also help to decide how to manage future epidemics, and motivate the effort to uncover the genetic and environmental factors that determine the individual antibody responses to a pathogen.
medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.