Seroprevalence and correlates of SARS-CoV-2 neutralizing antibodies from a population-based study in Bonn, Germany

Abstract

To estimate the seroprevalence and temporal course of SARS-CoV-2 neutralizing antibodies, we embedded a multi-tiered seroprevalence survey within an ongoing community-based cohort study in Bonn, Germany. We first assessed anti-SARS-CoV-2 immunoglobulin G levels with an immunoassay, followed by confirmatory testing of borderline and positive test results with a recombinant spike-based immunofluorescence assay and a plaque reduction neutralization test (PRNT). Those with a borderline or positive immunoassay result were retested after 4 to 5 months. At baseline, 4771 persons participated (88% response rate). Between April 24^th and June 30^th, 2020, seroprevalence was 0.97% (95% CI: 0.72-1.30) by immunoassay and 0.36% (95% CI: 0.21-0.61) when considering only those with two additional positive confirmatory tests. Importantly, about 20% of PRNT+ individuals lost their neutralizing antibodies within five months. Here, we show that neutralizing antibodies are detectable in only one third of those with a positive immunoassay result, and wane relatively quickly.

Fig. 1. Flow chart.

Overview of the number of participants, their test results as well as reasons for non-participation or missingness. A minus or a plus sign indicates a negative or positive confirmatory test result, respectively. IFT immunofluorescence test, PRNT plaque reduction neutralization test.

Fig. 2. Factors associated with the presence of neutralizing antibodies.

a Living in the same household with a person with confirmed or suspected SARS-CoV-2 infection as well as a higher number of reported symptoms were significantly associated with the odds of having neutralizing antibodies. b A reduced sense of taste or smell, fever in the last month, pain in arm/legs, chills/hot flashes, pain while breathing as well as muscle weakness and pain were significantly more often reported by individuals with versus those without neutralizing antibodies; seizures and confusion were not reported in the seropositive group, and because of a very low background prevalence, the associated odds ratios could not be estimated reliably. The statistical comparisons were between 16 cases with mild-to-moderate symptoms who had SARS-CoV-2 neutralizing antibodies and 4754 individuals without SARS-CoV-2 (neutralizing) antibodies. All estimates are adjusted for age, sex and household clustering. The red dots represent the odds ratio point estimates, while the whiskers depict the corresponding 95% confidence intervals, on a logarithmic scale.

Fig. 3. Relation between IgG response and neutralizing antibodies.

a Individuals with neutralizing antibodies had a significantly higher IgG antibody response as represented by the ELISA IgG ratio (a minus or a plus sign indicates a negative or positive test result, respectively). Each individual measurement is represented by one dot. The red dotted line indicates the threshold for a positive ELISA result. The box-plots indicate the medians (bold horizontal lines) and interquartile ranges (box boundaries), while the whiskers represent 1.5× interquartile ranges. Sample sizes: 23 ELISA+/IFT−/PRNT−, 7 ELISA+/IFT+/PRNT−, and 17 ELISA+/IFT+/PRNT+. P-values were obtained by the two-sided non-parametric Wilcoxon test. b Only in the ELISA+/PRNT+ subgroup there was a significantly higher IgG response with increasing age. The shaded areas around the regression lines represent the 95% confidence intervals of the mean. The red dotted line indicates the threshold for a positive ELISA result. Please refer to the main text for further details. IFT immunofluorescence test, PRNT plaque reduction neutralization test.

Fig. 4. Factors differentiating between ELISA+ individuals with and without neutralizing antibodies.

a A higher number of reported symptoms was significantly associated with a higher odds of having neutralizing antibodies. b Loss of appetite, muscle weakness, chills or hot flashes as well as a reduced sense of taste were significantly more often reported by individuals with versus those without neutralizing antibodies. All estimates are adjusted for age, sex and household clustering. The statistical comparisons were between 16 ELISA+/PRNT+ cases and 30 ELISA+/PRNT− control subjects. The red dots represent the odds ratio point estimates, while the whiskers depict the corresponding 95% confidence intervals, on a logarithmic scale.

Fig. 5. Time course of neutralizing antibody response.

a Titers of neutralizing antibodies decreased in most individuals during follow-up, becoming undetectable in four (PRNT₅₀ titers; the four individuals in whom neutralizing antibodies became undetectable after follow-up are highlighted by red arrows). b Within the subgroup of individuals with the highest neutralizing antibody titers at baseline, there were four individuals in whom these titers continued to increase (PRNT₉₀ titers). Each two-points joined by a dashed line represent one participant. The box-plots indicate the medians (bold horizontal lines) and interquartile ranges (box boundaries), while the whiskers represent 1.5× interquartile ranges. Sample size: 22 individuals with SARS-CoV-2 neutralizing antibodies at baseline. Note that neutralizing antibody titers were measured on a semi-quantitative scale (i.e., <1:20 (undetectable), 1:20, 1:40, 1:80, or >1:80), therefore, random vertical jitter was added to the points to avoid masking of data points with similar values. PRNT plaque reduction neutralization test.

Fig. 6. Time course of immunoassay-based IgG levels in relation to neutralizing antibody response at baseline.

There was a strong association between the magnitude of the immunoassay-based IgG response and the probability of levels of neutralizing antibodies (NAb) decreasing below detection limit (odds ratio for time (in days) × IgG ratio interaction = 0.96, 95% CI: 0.94–0.98, two-sided P << 0.001, using generalized estimating equations with a cumulative logistic link function and an independent covariance structure, and deriving the confidence intervals from the robust Huber-White sandwich variance estimator). The four individuals, who became NAb negative during follow-up, had relatively modest IgG responses (right panel). Sample size: 22 individuals with SARS-CoV-2 neutralizing antibodies at baseline. The horizontal dotted lines represent the borders of the indeterminate range.