SARS-CoV-2 correlates of protection from infection against variants of concern

Abstract

Serum neutralizing antibodies (nAbs) induced by vaccination have been linked to protection against symptomatic and severe coronavirus disease 2019. However, much less is known about the efficacy of nAbs in preventing the acquisition of infection, especially in the context of natural immunity and against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) immune-escape variants. Here we conducted mediation analysis to assess serum nAbs induced by prior SARS-CoV-2 infections as potential correlates of protection against Delta and Omicron infections, in rural and urban household cohorts in South Africa. We find that, in the Delta wave, D614G nAbs mediate 37% (95% confidence interval: 34-40%) of the total protection against infection conferred by prior exposure to SARS-CoV-2, and that protection decreases with waning immunity. In contrast, Omicron BA.1 nAbs mediate 11% (95% confidence interval: 9-12%) of the total protection against Omicron BA.1 or BA.2 infections, due to Omicron's neutralization escape. These findings underscore that correlates of protection mediated through nAbs are variant specific, and that boosting of nAbs against circulating variants might restore or confer immune protection lost due to nAb waning and/or immune escape. However, the majority of immune protection against SARS-CoV-2 conferred by natural infection cannot be fully explained by serum nAbs alone. Measuring these and other immune markers including T cell responses, both in the serum and in other compartments such as the nasal mucosa, may be required to comprehensively understand and predict immune protection against SARS-CoV-2.

Extended Data Fig. 1 |. Flowchart of participants included in the Delta-wave subgroup analysis.

Grey boxes represent participants excluded from the Delta-wave subgroup analysis. *Based on a previously published study. ^†Household with more than 6 infected individuals would be computationally intractable to track all possible transmission chain configurations (Methods Section 3).

Extended Data Fig. 2 |. Flowchart of participants included in the Omicron-wave subgroup analysis.

Grey boxes represent participants excluded from the Omicron-wave subgroup analysis. *Based on a previously published study. ^†Household with more than 6 infected individuals would be computationally intractable to track all possible transmission chain configurations (Methods Section 3).

Extended Data Fig. 3 |. D614G spike binding antibody (bAb) level for the Delta wave and the Omicron wave analysis.

a, for Delta wave subgroup, the distribution of the peak bAb level to BD5 (light blue dots) and the D614G spike bAb level at BD5 (dark blue dots), among individuals who had one prior SARS-CoV-2 infection before blood draw 5. Each dot represents one individual, with two measurements of the same individual connected through a gray line. OD: absorbance at 450 nm, measured in optical density; $\stackrel{-}{\mathrm{O}\mathrm{D}}$ the average of OD; $\stackrel{-}{\mathrm{O}\mathrm{D}}$ the average drop of OD. b, for Delta wave subgroup, the distribution of the peak D614G spike bAb up to BD5, stratified by individuals who were infected during the Delta wave (solid bar) vs those who were not infected (dashed bar). Independent samples t-test (two-sided) is used to determine the statistical significance (anti reported on the legend) of difference between the $\stackrel{-}{\mathrm{O}\mathrm{D}}$ of the two groups. c, same as b but for D614G spike bAb level at BD5. d, same as b but for ${\mathrm{\Delta }bAb}^W$. e, for Omicron wave subgroup, the distribution of the peak bAb level to BD8 (light red dots) and the D614G spike bAb level at BD8 (dark red dots), among individuals who had one prior SARS-CoV-2 infection before BD8. Each dot represents one individual, with two measurements of the same individual connected through a gray line. f, for the Omicron wave subgroup, the distribution of the D614G spike bAb level at BD8, stratified by individuals who were infected during the Omicron wave (solid bar) vs those who were not infected (dashed bar). Independent samples t-test (two-sided) is used to determine the statistical significance (p-value reported on the legend) of difference between the $\stackrel{-}{\mathrm{O}\mathrm{D}}$ s of the two groups. g, same as f but for D614G spike bAb level at BD8. h, same as f but for ${\mathrm{\Delta }bAb}^W$.

Fig. 1 |. Timing of cohort sample collections with respect to SARS-CoV-2 variants’ circulations in the two study sites.

a, Timing of the blood draws with respect to the SARS-CoV-2 epidemic waves in the rural site (Agincourt) of the PHIRST-C cohort. The bar plot represents the weekly incidence (per 100,000 population) of SARS-CoV-2 cases from routine surveillance data collected from the Ehlanzeni district in the Mpumalanga province (where rural participants reside). The shaded areas represent the timing of the serum sample collections for the ten blood draws. Each curve within the shaded area indicates the cumulative proportion of participants’ serum samples collected over time. The Delta wave subgroup analysis focuses on nAb titers among serum samples collected during BD5 (blue); the Omicron wave analysis focuses on nAb titers among serum samples collected during BD8 (red). b, Same as a, but for the urban site (Klerksdorp). The routine surveillance data (bar plot) were collected from the Dr. Kenneth Kaunda district in the North West province (where urban participants reside).

Fig. 2 |. D614G and BA.1 nAb titers for the Delta wave and the Omicron wave analysis.

a, For the Delta wave subgroup, the distribution of the peak D614G nAb titer up to BD5 (light blue dots) and the D614G nAb titer at BD5 (dark blue dots), among individuals who had one prior SARS-CoV-2 infection before BD5. Each dot represents one individual, with two measurements of the same individual connected through a gray line. GMFC, geometric mean fold change from peak D614G titer to that at BD5; r, Pearson correlation coefficient. b, For the Delta wave subgroup, the distribution of the peak D614G nAb titer up to BD5, stratified by individuals who were infected during the Delta wave (solid bar) versus those who were not infected (dashed bar). The independent-samples t-test (two-sided) was used to determine the statistical significance (P value reported in the legend) of the difference between the GMTs of the two groups. c, Same as b but for D614G nAb titers at BD5. d, Same as b but for ΔnAb^W (defined as the difference between D614G titers at peak and at BD5). e, For the Omicron wave subgroup, the distribution of D614G nAb titers (light red dots) and BA.1 titers at BD8 (dark red dots), among individuals who had one prior SARS-CoV-2 infection before BD8. Each dot represents one individual, with two measurements of the same individual connected through a gray line. f, For the Omicron wave subgroup, the distribution of the D614G nAb titer at BD8, stratified by individuals who were infected during the Omicron wave (solid bar) versus those who were not infected (dashed bar). The independent-samples t-test (two-sided) was used to determine the statistical significance (P value reported in the legend) of the difference between the GMTs of the two groups. g, Same as f but for BA.1 nAb titers at BD8. h, Same as f but for ΔnAb^E (defined as the difference between BA.1 and D614G titers at BD8).

Fig. 3 |. Causal diagrams for the mediation analyses.

a, Causal diagram of the Delta wave mediation analysis showing the hypothesized relationship between prior immunity (induced by prior SARS-CoV-2 infection) and SARS-CoV-2 infection (outcome of interest) during the Delta wave. The mediators of interest are D614G nAbs at BD5 and ΔnAb^W (the quantity of D614G nAbs waned from peak level to that at BD5). The direct effect represents protection operating through immune mechanisms other than the mediators of interest. We hypothesized that the direct effect could wane over time since the initial immune exposure. For the prospective cohort data, both mediator–outcome confounding and exposure–outcome confounding factors need to be adjusted for the mediation analysis, as the immune exposure (prior SARS-CoV-2 infection) was not randomly assigned (unlike SARS-CoV-2 randomized control vaccine trials where vaccination was randomly assigned to the participants). Furthermore, cohort participants may experience heterogeneous levels of SARS-CoV-2 exposure due to different intensity SARS-CoV-2 transmission in their household settings. We adjusted this by embedding the mediation analysis in a mechanistic household transmission model (Methods). We also look at the impact of prior immunity on the reduction of onward transmission, conditional on the failure of preventing reinfection. The estimates of the Delta wave mediation analysis are presented in Table 2. b, Same as a but for the Omicron wave analysis. The mediators of interest are BA.1 nAbs at BD8 and ΔnAb^E (the quantity of antibodies that can neutralize D614G but fail to neutralize Omicron BA.1 at BD8 due to Omicron’s immune escape). The estimates of the Omicron wave mediation analysis are presented in Table 2.