IgG Seroconversion and Pathophysiology in Severe Acute Respiratory Syndrome Coronavirus 2 Infection

We investigated the dynamics of seroconversion in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. During March 29–May 22, 2020, we collected serum samples and associated clinical data from 177 persons in London, UK, who had SARS-CoV-2 infection. We measured IgG against SARS-CoV-2 and compared antibody levels with patient outcomes, demographic information, and laboratory characteristics. We found that 2.0%–8.5% of persons did not seroconvert 3–6 weeks after infection. Persons who seroconverted were older, were more likely to have concurrent conditions, and had higher levels of inflammatory markers. Non-White persons had higher antibody concentrations than those who identified as White; these concentrations did not decline during follow-up. Serologic assay results correlated with disease outcome, race, and other risk factors for severe SARS-CoV-2 infection. Serologic assays can be used in surveillance to clarify the duration and protective nature of humoral responses to SARS-CoV-2 infection.

contained controls: negative (diluent solution), negative cutoff (purified human IgG in 10 mM Tris buffered saline) and positive (purified human IgG in 10 mM Tris buffered saline), in duplicates. The plate was incubated at room temperature for 30 min before 3 washes with wash buffer (10 mM Tris-buffered saline with detergent, pH 7.2). Following the wash step, 100 µL of conjugate solution (IgG against humans, conjugated to horseradish peroxidase with protein stabilization and antimicrobial agents) was added to every well and the plate incubated for 30 min at room temperature before 4 washes with wash buffer. Indicator/substrate solution (3,3′,5,5′-tetramethylbenzidine, TMB, with H2O2; 100 µL) was added to every well and the plate incubated at room temperature for 10 min before addition of 100 µL stop solution (0.25 M H2SO4) per well. The plate was read with a spectrometer at 450 nm within 10 min of stop solution addition. Wells with an optical density of 10% greater than the negative cutoff value were regarded as positive for antibodies to SARS-CoV-2 antigen.

Normalization
For each plate the mean cutoff OD value plus 10% (lower cutoff) was subtracted from each mean patient sample OD value (thus samples with negative values were considered negative and samples with positive values were considered positive for SARS-CoV-2 antibodies). To normalize, the resulting value was divided by the mean positive control OD value. Any assayed duplicates (serum samples taken from the same patient on the same day) were removed from further analysis.

Resolving Inconsistencies
Inconsistent data points (n = 3) were identified (e.g., a data point that suggested a seropositive patient subsequently lost their antibody response). In these cases: samples were rerun, including sequential samples from the same patient from either side of the timepoint; alkaline phosphatase results from that sample were checked; and sample aliquots were resourced from stocks in South West London Pathology laboratories.

Normalized ELISA Values
Normalized OD values for ELISA results are presented in Appendix Figure 1.

Analysis
A 1-way analysis of variance was conducted to compare the effect of race (white/nonwhite/unknown) on demographics and severity indices in patients. Results showed that antibody levels (mean NOD) across the 3 groups were unequal, F (2,174) = 3.46, p = 0.03 and post hoc comparisons using Tukey's test showed that antibody levels in whites were significantly different from nonwhites. There was no observed significant effect of race on other measured severity indices and demographics (Appendix Table 3).

Multiple Linear Regression (Independent Variable: Mean NOD)
Multiple linear regression was performed to determine the relationship between mean NOD and age, sex, peak CRP levels, number of concurrent conditions, presence or absence of respiratory symptoms, and race. Upon examination of regression β coefficients and associated tstatistic p values, there was a significant association found between higher age, higher peak CRP levels, nonwhite race, and higher NOD values. The p value of the F-statistic is 8.6 × 10 -5 with an adjusted R-squared (R 2 ) of 0.10 (10% of the variance in mean NOD levels can be predicted by the 3 variables).

Logistic Regression Model (Independent Variable: Seroconversion)
Logistic regression was performed to determine the relationship between seroconversion and age, sex, peak CRP levels, number of concurrent conditions, presence or absence of respiratory symptoms, and race. A significant association was found between higher age, higher peak CRP levels and seroconversion (p = 0.001 and p = 0.035 respectively).

Our final model equation can be written as follows:
Log odds of seroconversion = −1.37 + 0.051 × age + 0.004 × peak CRP level

Logistic Regression Model (Independent Variable: Severity/outcome)
Logistic regression was performed to determine the relationship between poor outcome (death and or ICU admission) and age, gender, peak CRP levels, number of concurrent conditions, presence or absence of respiratory symptoms, and race. A significant association was found between only higher peak CRP levels and poor outcome (p = 1.1 × 10 -8 ).

Our final model equation can be written as follows:
Log odds of a poor outcome = −2.08 + 0.008 × peak CRP levels.  Table 3. Laboratory values at diagnosis from symptomatic and asymptomatic patients with severe acute respiratory syndrome coronavirus 2, United Kingdom, 2020.
Test ( respiratory symptoms. Symbols indicate the last available seronegative sample for a patient subsequently right censored (i.e, no further data was available). B) Patients >70 years of age (filled squares) or <70 years of age (filled circles). Symbols indicate the last available seronegative sample for a patient subsequently right censored. C) Men (filled squares) or women (filled circles). Symbols indicate the last available seronegative sample for a patient subsequently right censored.