Introduction

Soon after the emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), there were prevailing concerns that prior vaccination might trigger immunopathological responses to the infection. This led to recommendations for monitoring the human phases of the clinical development of candidate vaccines1. Severe clinical outcomes following respiratory syncytial virus (RSV) infection in infants who had been vaccinated with a formalin-inactivated, whole-virus RSV vaccine motivated these concerns2. Later, non-clinical studies were able to reproduce the clinical phenotype in macaques in pursuit of the elucidation of relevant mechanisms3. Additional non-clinical observations of antibody-dependent enhancement of infection following vaccination with candidate SARS-CoV-14,5,6,7 or Middle East Respiratory Syndrome coronavirus (MERS)8 vaccines provided reasons for vigilance. Much of this preceding research focused on characteristics of the antigen design or vaccine construct and composition, and less on host factors.

Subsequent experience demonstrated that human infection with SARS-CoV-2 is associated with a spectrum of clinical outcomes, ranging from asymptomatic to fatal disease. This heterogeneity of clinical outcome following infection with SARS-CoV-2 is not fully accounted for by baseline health status or comorbid conditions9. More recently, emerging viral variants have been notable for increased efficiency of transmission with apparently lesser clinical severity of clinical infection10. Pre-existing immune responses have also been identified as a factor for differences in clinical outcomes11. By contrast, human populations have regular seasonal exposure to ‘common cold’ human endemic coronaviruses (HCoVs), which are notable for a less severe course of clinical illness12.

Pre-existing immunity to HCoVs could impact the response to SARS-CoV-2 vaccination in several ways. For example, boosting cross-reactive responses could augment viral clearance. Alternatively, boosting of pre-existing responses could focus immune responses on previously generated epitopes rather than novel epitopes that differ between the viruses13. Furthermore, antibody function that facilitates uptake by macrophages is hypothesized as a mechanism whereby pre-existing, non-neutralizing responses could lead to more severe manifestations of clinical illness following infection14. It was also not known whether pre-existing immune responses to endemic coronaviruses could impact immune responses to SARS-CoV-2 candidate vaccines.

Given these considerations, we included measurement of pre-existing binding responses to four human endemic coronaviruses in the first-in-human clinical evaluation of a SARS-CoV-2 spike protein vaccine candidate that was subsequently licensed as a booster vaccine15,16,17. We have previously evaluated the cellular immune response to this vaccine candidate, where we found little evidence of baseline cellular reactivity to SARS-CoV-2 spike protein or spike peptides among participants in this trial18. Here, we evaluate the impact of baseline HCoV antibody response magnitude and breadth on neutralizing and binding antibody responses to the candidate SARS-CoV-2 vaccine.

Results

Among 436 VAT00001 trial participants (99.3% of the safety analysis population) with available Binding Antibody Multiplex Assay (BAMA data), reactivity to four endemic coronaviruses was detected in all participants and did not differ by age group or dosing regimen [Fig. 1A, Supplementary Tables 2 and 3]. The distribution of the HCoV log10-transformed binding antibody responses at baseline (D01) shows considerable overlap across age range and dosing regimen [Fig. 1A]. Samples were tested at a 1:100 dilution factor, and the log10-transformed response magnitude (MFI) ranged from 2.3 to 4.5 (Supplementary Table 2), with a minimal percentage of sample/analyte combinations exhibiting responses above the linear range of the assay (123/3488 sample/analyte combinations (3.5%). There were modest levels of correlation of the binding antibody response magnitude between endemic coronaviruses, being generally strongest between the RBD and spike antibody response magnitude for a given endemic coronavirus, e.g. highest between OC43 RBD and OC43 spike than other HCoVs (Supplementary Fig. 2). Principal components analysis (PCA) revealed that there was no clustering of antibody responses among the endemic human coronaviruses (Fig. 1C).

Fig. 1: Prior exposure to endemic alpha and beta-coronaviruses among healthy adult volunteers.
figure 1

Distribution of A human endemic coronavirus log 10 transformed binding antibody multiplex assay (BAMA) response magnitudes in baseline samples (D01) among VAT00001 trial participants, B fold-rise in neutralization titer and SARS-CoV-2 binding antibody concentration at D36 from baseline, stratified by age stratum and dosing schedule. Treatment assignments are shown for both panels, and C principal components analysis (PCA) plot of human endemic coronavirus response magnitude and treatment assignment. Ab antibody, D01 day 1, D36 day 36, HCoV human coronavirus, MFI mean fluorescence intensity, PC principal component, RBD receptor binding domain.

Full size image

In linear regression models, we evaluated whether treatment assignment and the response magnitude of binding antibody responses at D01 to four endemic coronavirus RBD and full-length spike proteins were predictive of neutralizing antibody response to SARS-CoV-2, as assessed in the microneutralization assay. Among participants 18–49 years of age allocated to the two-dose cohort, treatment assignments were significant predictors of neutralizing antibody response at D36, but none of the binding antibody responses at baseline to endemic coronaviruses were significant predictors of neutralizing response (Table 1). The lack of predictive findings of the endemic coronaviruses held true for the two-dose cohort in the 50+ years of age stratum (Supplementary Table 4) and single-dose or two-dose cohorts evaluated at D22 (Supplementary Tables 5 and 6). The regression analyses for the age-dosing schedule—D22/D36 permutations (not shown)—were consistent with these findings. Combining antibody response magnitude across all spike assays or all RBD assays into a single measure of CoV spike or RBD breadth was also not predictive of neutralizing responses when accounting for treatment assignment (18–49 years two-dose cohort, p = 0.11 and p = 0.93, respectively). Scatter plots of D36 neutralizing response (log10 transformed fold rise of neutralizing response at D36 compared to baseline) versus the D0 BAMA response magnitude [log10 transformed breadth scores or mean fluorescence intensity (MFI)], stratified by age stratum and dosing schedule, further support the lack of meaningful effect of the prior endemic human coronavirus exposure on the neutralizing antibody response to vaccination (Fig. 1B).

Table 1 Linear regression analysis of treatment assignment and baseline BAMA HCoV antibody response magnitude as predictors of SARS-CoV-2 WA-1 microneutralization titer at D36 among 18–49 yr-old participants receiving a two-dose regimen (N = 173)
Full size table

There were no meaningful differences in baseline endemic coronavirus response magnitude among participants who achieved a four-fold rise in neutralizing titers by D36 compared to baseline. However, a slightly higher response magnitude to the betacoronavirus OC43 spike was observed among participants displaying a four-fold rise (4.06 vs. 4.12 log10 units, p = 0.018).

We further evaluated whether treatment assignment and D01 endemic coronavirus response magnitude were predictive of binding antibody response to SARS-CoV-2 in a validated GCN-4 enzyme-linked immunosorbent assay (ELISA). As for neutralizing responses, among participants 18–49 years of age allocated to the two-dose cohort, treatment assignments were significant predictors of binding antibody response at D36, but none of the endemic coronaviruses were significant predictors of binding antibody response (Table 2). The lack of predictive findings held true for the two-dose cohort in the 50+ years of age stratum and single-dose or two-dose cohorts evaluated at D22 (data not presented). In keeping with the neutralizing antibody analyses, breadth scores for cumulative HCoV spike or RBD response at baseline were not significant predictors of binding antibody response to SARS-CoV-2 post-baseline.

Table 2 Linear regression analysis of treatment assignment and baseline BAMA HCoV antibody response magnitude as predictors of SARS-CoV-2 GCN-4 ELISA binding antibody concentration at D36 among 18–49 yr-old participants receiving a two-dose regimen (N = 173)
Full size table

Discussion

We found that prior exposure to endemic alpha and beta-coronaviruses was universally detected among healthy adult volunteers in a Phase I/II trial conducted in eight clinical trial sites in the United States. There was no meaningful predictive association between pre-existing binding antibody responses to four endemic coronaviruses and either neutralizing or binding responses to SARS-CoV-2 following vaccination with stabilized prefusion SARS-CoV-2 spike trimer recombinant protein vaccine candidates. By contrast, the dose and schedule of these candidate formulations were meaningful predictors of the SARS-CoV-2 immune response in the trial volunteers, who were not previously exposed to SARS-CoV-2 and who did not exhibit antibody reactivity to SARS-CoV-2 at the baseline visit.

Our findings are similar to those observed in an occupational longitudinal cohort of individuals who received mRNA vaccinations19. In that study, while baseline endemic coronavirus antibody levels were correlated with SARS-CoV-2 antibody levels following infection, the authors did not find an association between baseline human common cold virus immunity and SARS-CoV-2 antibody levels post-vaccination. Crowley et al. also found that antibody responses to endemic HCoV were not significantly associated with vaccine responses20. The consistency of these findings despite near-universal prior exposure to HCoVs indicates that these highly immunogenic vaccines are not impacted by HCoVs-related immune imprinting as observed with infection. Our findings suggest that recent HCoV exposure, which strongly impacts vaccine response in murine models, may not be relevant to human vaccine response. However, it is important to note that the timing of exposure in the current trial volunteers is not known. One limitation of the current study is that we did not assess the HCoV antibody response magnitude following vaccination.

The theoretical possibility for suboptimal immune responses to vaccination, which could be associated with increased severity of respiratory disease upon infection with circulating SARS-CoV-2, was a significant concern in the initial stages of coronavirus disease 2019 (COVID-19) vaccine development. As exposure to human endemic coronaviruses is extremely common, the theoretical concern for immune imprinting to adversely affect the development of responses to SARS-CoV-2 vaccines was raised as a potential mechanism that could lead to vaccine-associated enhanced respiratory disease (VAERD)1,21,22. VAERD has not been observed in clinical development of multiple COVID-19 vaccines that achieved regulatory authorization, including the CoV2 dTM preS-AS03 vaccine, nor following widespread deployment in public vaccination campaigns23,24,25. In the present study, we found no meaningful evidence for an impact of prior HCoV exposure on measures of early immune response to vaccination with prefusion-stabilized trimer recombinant protein vaccine candidates. The current study provides an immunological view that prior exposure to endemic coronaviruses does not substantially impact the vaccine-elicited antibody responses.

Methods

Clinical trial

VAT00001 was a Phase I/II randomized, modified double-blind, placebo-controlled, first-in-human, dose-escalation trial. The trial design and interim results have been described in the prior publication26. In brief, participants were healthy volunteers without prior exposure to SARS-CoV-2, stratified by age (18–49 years of age, 50 and above). The vaccine formulations evaluated included one or two dose levels (1.3 or 2.6 μg) of a recombinant protein vaccine comprising the stabilized prefusion spike of the ancestral (WA-1) SARS-CoV-2 strain, delivered with AS03 (GSK) or AF03 (Sanofi) adjuvant or without adjuvant. The dosing schedules evaluated included a single administration (D01) or two doses separated by a 21-day interval (D01, D22).

Assay methods

The microneutralization assay (Global Clinical Immunology, Sanofi Vaccines, Swiftwater, PA) and GCN-4 binding antibody enzyme-linked immunosorbent assay (ELISA; Nexelis, Laval, QC) methods have been described elsewhere26,27. Blood samples obtained at baseline and post-vaccination (D01, D22, D36) were assessed in the above assays.

A qualified Binding Antibody Multiplex Assay (BAMA) was used to detect endemic coronavirus-specific IgG in serum from participants at baseline (D01) in the VAT00001 trial. The assay detects antibodies to the receptor binding domain (RBD) and to full-length spike of endemic human coronaviruses HKU1, 229E, NL63, and OC43. Both forms of CoV antigens were chosen since RBD exhibits lower homology across CoVs, whereas the full-length spike contains regions that are more conserved across CoVs20,28,29. Antigens included in the assay are detailed in Supplementary Table 1. The method for detecting HCoV-specific antibodies, described previously9, includes sample incubation with antigen-coated beads for 30 min (MagPlex beads, Luminex Corporation). NeutrAvidin-coated beads are bound to a biotinylated rabbit anti-6xhis antibody, which then captures His-tagged HCov proteins in BAMA diluent (PBS, 1% milk blotto (w/v), 5% normal goat serum, 0.05% Tween). Samples were tested at a single dilution of 1:100 in assay diluent, then added to the bound beads and then incubated for 30 min on an orbital plate shaker. Human IgG HCoV-specific antibodies are then detected using an anti-human IgG phycoerythrin (PE) labeled detection antibody (Southern Biotech). Plates containing the bound microspheres/antigen/antibody complex are read on a BioPlex 200 instrument to measure fluorescence intensity (FI). Duplicate wells are averaged to provide a mean fluorescence intensity (MFI) value. Background binding (no sample and no antigen controls) is subtracted from each sample well (Net MFI), which is then used for statistical analysis. The dynamic range of the Bioplex instrument is 2.0–4.5 (log10 transformed MFI). Assay controls included a titrated HCoV seropositive plasma sample (HCoV-PC55 plasma, kindly provided by Dr. Paul Goepfert (University of Alabama) and HCoV spike (AB711725 mAb, kindly provided by Dr. Kevin Saunders, Duke Human Vaccine Institute) as outlined in Supplementary Table 1. Additionally, an irrelevant human monoclonal antibody (7B2) was used as a negative control. Additional details regarding the specificity of the assay for endemic HCoV and the 7B2 negative control mAb binding are shown in Supplementary Fig. 1 (A and B), including the non-reactivity of a SARS-CoV-2-specific monoclonal antibody (AB712384 mAb, kindly provided by Drs. Kevin Saunders and Barton Haynes, Duke Human Vaccine Institute) to the endemic HCoV antigens.

The study was done in compliance with the International Conference on Harmonization guidelines for Good Clinical Practice and the principles of the Declaration of Helsinki. The protocol and amendments were approved by the applicable Independent Ethics Committees and Institutional Review Boards and the regulatory agency as per local regulations. Written informed consent was obtained from the participants before any study procedures were done26.

Statistical methods

We evaluated the RBD or spike HCoV antibody magnitudes at D01 and vaccine formulation as predictors of the D22 or D36 post-vaccination neutralizing antibody titer or binding antibody concentration using linear regression stratified by age (18–49, 50+) and dosing schedule (single dose, two doses). Additionally, we examined RBD and spike breadth scores as predictors by calculating the mean magnitude of the four HCoV RBDs or four HCoV spike proteins, respectively. Scores were calculated by first averaging the response magnitudes of the four individual antigens within each panel for each sample (i.e., mean of [HKU1, 229E, NL63, and OC43 RBDs] and mean of [HKU1, 229E, NL63, and OC43 spike proteins]). The resultant mean was then log10 transformed for analysis and plotting.

Statistical analyses were conducted using R v4.3.0. Principal component analysis (PCA) was conducted using the R function ‘prcomp,’ and a plot of PC1 and PC2 was generated to visualize potential outliers or clusters and the relationship between those RBD or spike HCoV antibodies plotted using the R package ggfortify. Correlations were evaluated using Spearman’s rank correlation. An alpha level of 0.05 was used to denote statistical significance. P-values were not adjusted for multiple comparisons.