1 INTRODUCTION

Early efficacy trials and real-world data on the BNT162b2 messenger RNA (mRNA) vaccine confirmed its high effectiveness in reducing laboratory-confirmed infection, COVID-19 hospitalization and death.^1-4 Nevertheless, a gradual decline in vaccine efficacy (VE) over time has been observed within the first months after the initial two-dose regimen.^4-7 This waned efficacy was consistent with the decrease of neutralizing antibodies observed by multiple independent studies^8-10 supporting neutralizing antibodies as a strong correlate of COVID-19 protection.^11-14

Moreover, since the beginning of the pandemic, several mutations occurred in the SARS-CoV-2 genome leading the different lineages of the virus.¹⁵ Five of these lineages have been designated as a variant of concern (VOC) by the World Health Organization, namely the Alpha, Beta, Gamma, Delta and Omicron variants.¹⁶ Discovered in November 2021, the Omicron lineage is to date the leading variant over the world.¹⁵ This variant is characterized by 32 dominant mutations in the spike (S) protein, 15 of which are located in the receptor-binding domain (RBD) conferring an increased transmissibility and a considerable immune escape from acquired protection through SARS-CoV-2 vaccination or a previous infection.^17-24 Currently, Omicron is largely dominant and several subvariants have emerged including BA.2, BA.2.12.1, BA.4, and BA.5.¹⁵ All these subvariants have also demonstrated a considerable escape to acquired immunity.^{20, 25, 26}

The current BNT162b2 vaccine, which has been elaborated on the sequence of the wild-type SARS-CoV-2,²⁷ has been shown to be less effective against Omicron compared to other VOCs^28-31 and the VE also waned over time to reach zero to 22.3% 6 months after the second BNT162b2 dose.^{5, 7, 30, 32, 33} With the decreased efficacy of vaccines over time and the emergence of highly transmissible SARS-CoV-2 variants that escape neutralization, many countries have deployed third doses of COVID-19 vaccines.

The administration of a homologous BNT162b2 booster dose increased the VE to 58.9% (interquartile range  [IQR] = 52.7%–63.3%) within 2–4 weeks.^{7, 30, 32-34} This increase was consistent with the rise of binding (median fold increase = 26.8; IQR = 13.6–51.7)^35-37 and neutralizing antibodies (median fold increase = 27.3; IQR = 10.2–52.7).^{18, 19, 38-41} However, a waning of protection against symptomatic diseases was rapidly observed 8–14 weeks after the booster (median VE = 37.9%; IQR = 24.6%–45.1%).^{7, 30, 33} Although protection against severe COVID-19 remains higher, the Center for Disease Control and Prevention reported that after receiving both 2 and 3 doses, the VE was lower during the Omicron-predominant than during the Delta-predominant period at all time points evaluated.⁴²

Evidence about the long-term persistence of the booster-mediated immunity against Omicron is crucial knowledge for pandemic response. The aim of this study was to evaluate 6-month humoral response in a cohort of healthcare workers (HCW) who received the homologous BNT162b2 booster.

2 MATERIALS AND METHODS

3 RESULTS

3.2 Neutralizing antibodies against the Omicron BA.1 variant

In participants with no history of SARS-CoV-2 infection, the highest measured neutralizing capacity was reached at day 28 with a GMT of 221 (95% CI = 175–277), representing a 15.5-fold increase from baseline (i.e., 14.3; 95% CI = 12.1–16.8). A continuous decrease was then observed up to day 180 with an observed GMT of 19.3 (95% CI = 15.1–24.6), which represents a 11.5-fold decrease. At 6 months, the mean neutralizing titer was not significantly different from baseline (Table 1 and Figure 1). In participants with history of SARS-CoV-2 infection, the highest neutralizing capacity was also reached at day 28 with a GMT of 264 (95% CI = 186–373), corresponding to a 15.4-fold increase from baseline (i.e., 17.1; 95% CI = 12.8–22.7). As observed in COVID-19 naive individuals, a continuous decline was observed up to day 180 with a GMT of 26.0 (95% CI = 18.3–36.8), which represents a 10.2-fold decrease. The mean titer at 6 months was comparable to baseline (Table 1 and Figure 1). For each time point, no significant differences were observed in individuals with or without history of SARS-CoV-2 infection (p > 0.05) (Table 1). The proportion of detectable Omicron-specific neutralizing antibodies was low at baseline (16.0% and 29.4% for participants without or with history of SARS-CoV-2 infection, respectively) and progressively increased to achieve 100% at day 28 for COVID-19 naive individuals and 100% between days 14 and 56 for past-COVID-19 subjects. Afterward, the seroprevalence progressively decreased to achieve 37.2% and 63.2% after 6 months in individuals without or with previous history of SARS-CoV-2 infection (Figure 1). The estimated T_1/2 of neutralizing antibodies for COVID-19 naive participants was 36 days (95% CI = 25–65 days). The T_max was estimated at 18 days (95% CI = 14–22 days). In previously infected subjects, the estimated T_1/2 of neutralizing antibodies was 42 days (95% CI = 25–137 days) and the T_max was reached at 24 days (95% CI = 15–32 days). Estimations for T_1/2 and T_max were not significantly different between groups. According to the model, a mean time of 182 days (95% CI = 118–234) in COVID-19 naive participants and 214 days (95% CI = 110–297) in previously infected subjects would be needed to cross the dilution titer threshold of 1/20 (Figure 3A).

Table 1. Fifty percent relative inhibition pseudovirus-neutralization titers and binding antibodies titers of sera from vaccine recipients, collected before and after the homologous BNT162b2 booster

	Never infected (n = 104)		History of infection (n = 51)
	GMT (95% CI)	% pos. samples	GMT (95% CI)	% pos. samples	p value
pVNT50 titer (dilution−1)
Before booster	14.3 (12.1–16.8)	16.0	17.1 (12.8–22.7)	29.4	>0.99
7 days	42.4 (29.6–60.8)	59.6	47.5 (26.7–84.3)	74.0	>0.99
14 days	177 (122–266)	92.3	168 (104–269)	100	0.91
28 days	221 (175–277)	100	264 (186–373)	100	>0.99
56 days	125 (94.0–165)	92.7	170 (105–275)	100	0.99
90 days	33.3 (25.8–42.9)	71.4	53.9 (34.4–84.4)	80.0	>0.99
180 days	19.3 (15.1–24.6)	37.2	26.0 (18.3–36.8)	63.2	>0.99
Binding antibodies (BAU/ml)
Before booster	480 (407–566)	0.0	1999 (1590–2512)	6.1	<0.0001
7 days	14 879 (12 056–18 364)	86.6	15 842 (12 618–19 891)	91.9	>0.99
14 days	18 834 (17 295–20 509)	92.0	17 461 (15 028–20 288)	97.1	0.99
28 days	17 386 (15 834–19 090)	93.4	15 271 (13 241–17 613)	90.2	0.85
56 days	14 463 (13 002–16 088)	81.0	12 123 (9724–15 113)	68.8	>0.99
90 days	11 505 (9915–13 351)	73.4	9610 (7017–13 160)	62.5	>0.99
180 days	6508 (5080–8338)	38.6	6868 (4461–10 573)	52.6	>0.99

• Note: The percentage of positive sera according to the assay considered are also represented. GMT stand for geometric mean titers. Positive cut-offs were >20 dilution titer-1 and >8434 BAU/ml for neutralizing and binding antibodies, respectively. The p value expresses the statistical difference between GMT of seronegative and seropositive persons.
• Abbreviations: CI, confidence interval; pVNT, pseudovirus-neutralization test.

3.3 Binding antibodies

In participants with no history of SARS-CoV-2 infection, the highest measured binding antibody response was reached at day 14 with a GMT of 18 834 BAU/ml (95% CI = 17 295–20 509), representing a 39.2-fold rise from baseline (i.e., 480 BAU/ml; 95% CI = 407–566). A continuous decrease was then observed up to day 180 with an observed GMT of 6508 BAU/ml (95% CI = 5080–8338), which represents a 2.9-fold decrease compared to day 14. Levels of binding antibodies at 6 months were higher compared to baseline (Table 1 and Figure 2). In participants that were previously infected, the highest binding antibody response was reached at day 14 with a GMT of 17 461 BAU/ml (95% CI = 15 028–20 288), corresponding to a 8.7-fold increase from baseline (i.e., 1999 BAU/ml; 95% CI = 1590–2512). A continuous decline was observed up to day 180 with a GMT of 6868 BAU/ml (95% CI = 4461–10 573), which represents a 2.5-fold decrease in binding antibody titers at 6 months. Six-month titers were higher compared to baseline titers (Table 1 and Figure 2). Except at baseline (i.e., just before the administration of the booster), no significant differences were observed in individuals with or without history of SARS-CoV-2 infection (p > 0.05) (Table 1). All participants still had detectable positive binding antibodies 6 months after the booster (i.e., >0.8 BAU/ml). The estimated T_1/2 of binding antibodies for COVID-19 naive participants was 139 days (95% CI = 113–180 days) and the T_max was reached at 11 days (95% CI = 9–13 days). In previously infected subjects, the estimated T_1/2 of binding antibodies was 168 days (95% CI = 116–303 days) and the T_max was reached at 9 days (95% CI = 0–19 days). Estimations for T_1/2 and T_max were not significantly different between groups. According to the model, a mean time of 186 days (95% CI = 155–223) in COVID-19 naive participants and 194 days (95% CI = 142–283) in previously infected subjects would be needed to cross the threshold of 8434 BAU/ml (Figure 3B). This threshold represents the binding antibody titer needed to ensure a neutralizing activity of 1/20 (Figure 4B).

3.4 Binding antibodies versus neutralizing antibodies and correlation to VE

A significant correlation between binding antibodies and neutralizing titers was found (r = 0.51, 95% CI = 0.46–0.56, p < 0.0001) but the strength of agreement was null using the manufacturer's cut-off of 0.8 BAU/ml since all results for binding antibodies were positive. Furthermore, there was a proportional and significant increase in binding antibodies according to categories of neutralizing titers. GMT for binding antibodies corresponding to neutralizing titer categories <20, 20–80, 81–160, 161–320, and >320 were 3286, 10 351, 12 481, 16 588, and 20 036 BAU/ml (Figure 4A). Based on the ROC curve analyses, an alternative cut-off of 8434 BAU/ml for binding antibodies was identified to predict the neutralization of the Omicron BA.1 variant with a calculated sensitivity and specificity of 83.1% and 63.4%, respectively (area under the curve = 0.77, p < 0.0001) (Figure 4B). Therefore, there was a high proportion of sera (i.e., 36.6%) without neutralizing antibodies that had high titers of binding antibodies. Using this adapted cut-off induced a Cohen's kappa of 0.45 (95% CI = 0.38–0.51) that corresponds to a moderate agreement. The GMT of binding and neutralizing antibodies obtained in our study correlated strongly with the VE (%) from symptomatic infection (r = 0.83 [95% CI = 0.63–0.93], p < 0.0001 and r = 0.89 [95% CI = 0.72–0.95], p < 0.001, for binding and neutralizing antibodies, respectively), with the respective equations: “y = 0.01890*x + 3.251” and “y = 0.02106*x + 1.085” (Figure 5).

4 DISCUSSION

Although homologous boosting with BNT162b2 vaccine elicited high titers of binding and neutralizing antibodies to Omicron BA.1 in the first weeks following vaccine administration, the response waned substantially within 6 months. The rapid increase in the humoral response was observed regardless of previous SARS-CoV-2 history. After 6 months, the decay of neutralizing antibodies was higher compared to binding antibodies (11.5- and 10.2-fold decrease vs. 2.9- and 2.5-fold-decrease in COVID-19 naive and in previously infected subjects, respectively). Accordingly, binding antibodies presented a significantly higher T_1/2 (139–168 days) as compared to neutralizing antibodies (36–42 days). Nevertheless, the mean time to reach the neutralization cut-off was similar, i.e., 186–194 days versus 182–214 days, for binding and neutralizing antibodies, respectively. The proportion of participants that was considered negative for binding and neutralizing antibodies was also related.

Interestingly, the global humoral response was not different in participants with a history of SARS-CoV-2 infection compared to COVID-19 naive participants. This feature was not seen after the second dose for binding antibodies.^{9, 49} This is somewhat in contradiction with the study of Alatarwneh et al.³² who found that vaccination enhanced protection among persons who had had a previous SARS-CoV-2 infection. Hybrid immunity resulting from previous infection and recent booster vaccination conferred the strongest protection in this study.

Very few studies documented the long-term kinetics of antibodies following the booster administration. After a follow-up period of 3 months, Lyke et al.²⁰ found a 3.5-fold decrease in neutralizing antibodies against Omicron in a population of 50 participants having received the BNT162b2 booster. The observe decay was quite similar as in our study at 3 months (i.e., 5.5-fold decrease). Munro et al.⁵⁰ found a considerable decay of immunoglobulin G (IgG) titers 7 months after the homologous BNT162b2 booster in a small population of 31 COVID-19 naive subjects (26 982 at 28 days and 3761 U/ml at 7 months: 7.2-fold decrease). Neutralizing antibodies were not evaluated. Regev-Yochay et al.⁵¹ also identified a decrease of IgG titers 4–5 months after the homologous booster administration (2102 at 28 days and 383 BAU/ml at 4–5 months resulting in a 5.5-fold decrease). A parallel decay in neutralizing antibodies was also observed (2629 at 28 days and 480 BAU/ml at 4–5 months resulting in a 5.5-fold decrease). Only COVID-19 naive participants were included (n = 154). The waning of the humoral response observed in our study and in the literature for the postbooster period^{50, 51} was consistent with the one observed after the second dose of BNT162b2.^{8, 9} The distinct kinetics observed for IgG and total antibodies may be explained by the additional response of non-IgG antibody isotypes, which may persist several months after vaccination.

The waning of antibodies over time after the BNT162 booster was proportional to the decrease of VE identified in the literature^{7, 29-31, 33, 34, 42, 48} and we found a stronger correlation between GMT of neutralizing antibodies (r = 0.89) compared to GMT of binding antibodies (r = 0.83), yet the difference was not significantly different. Previous studies focusing on primary vaccine schemes and mostly on the wild-type virus identified that SARS-CoV-2 antibodies are a strong correlate of protection.^11-14 Lower levels of binding and neutralizing antibodies during the peri-infection period were described in breakthrough patients in comparison to control patients (i.e., patients that did not develop infection), supporting the role of antibodies in protecting against infection.^{52, 53} Our results therefore reinforce the conclusions of these preliminary studies and show that these also applied to Omicron after the booster administration.

Although a waning of binding and neutralizing antibodies was observed as well as similar correlations between mean titers and VE, the two methods used to measure these antibodies were not commutable. Indeed, neutralizing antibodies, that represent a first layer of adaptive immunity against COVID-19, were only modestly correlated (r = 0.51) against the commercial assay used. This latter was therefore not adapted to predict the presence of neutralizing antibodies. The refining of the cut-off for binding antibodies at 8434 BAU/ml allowed us to improve the prediction of neutralizing antibodies, but the performance remained moderate since there is still a significant proportion of samples with high binding antibody titers that do not correspond to neutralizing activity against Omicron (Figure 3). Therefore, diagnostic companies should need to rethink their current commercial assays (i.e., modification of antigen and epitopes) to design assays capable of predicting neutralizing activity against emerging and highly mutated SARS-CoV-2 variants.⁵⁴ This would also avoid any misinterpretation (i.e., high protection in case of high binding antibody titers).⁵⁴ Methods used to measure neutralizing antibodies present a low throughput, are time-consuming, need skillful operators, and require high levels of biosafety (especially for live virus neutralization assay).⁵⁴ It would therefore be easier to use commercial assays that can be a surrogate of these reference methods.

In our study, we evaluated the vaccine-induced neutralizing activity against Omicron BA.1. We were not able at that time to evaluate the neutralizing activity against Omicron sublineages BA.2, BA.2.12.1, BA.3, or BA.4/5. Lyke et al.²⁰ identified similar neutralizing titers 29 days after the Moderna mRNA-1273 booster between BA.1, BA.2, and BA.3, while a modest decline was observed for BA.2.12.1 (1.5-fold) and BA.4/5 (2.5-fold) in a total of 16 subjects. Accordingly, Hachmann et al.²⁶ found no difference between BA.1 and BA.2, but significant lower titers for BA.2.12.1 (2.2-fold) and BA.4/5 (3.3-fold), 14 days after the BNT162b2 homologous booster in 27 participants. Cao et al.²⁵ confirmed the same mean titers of neutralizing antibodies in 50 participants between BA.1 and BA.2 28 days after the CoronaVac homologous booster, and lower neutralizing activity compared to BA.1 for BA.2.12.1 (1.2-fold) and BA.4/5 (1.6-fold). However, Bowen et al.⁵⁵ found similar neutralizing titers between BA.1, BA.2, BA.2.12.1 and B4/5, ±30 days after mRNA-1273/BNT162b2 booster in 13 participants. A strong increase after the booster was also observed for Omicron sublineages.^{20, 26, 55} Our results 6 months after the booster might therefore be overestimated compared to BA.2.12.1 and B4/5 sublineages.

The administration of a fourth dose is currently under discussion⁵⁶ and some interim recommendations have been formulated.^{57, 58} Considering the waning of VE over time and the considerable immune escape of new emerging variants, the fourth dose seems unavoidable to restore a sufficient level of neutralizing antibodies. The efficacy of a fourth dose (or second booster) against symptomatic disease, hospitalization and severe COVID-19 has already been proved.^59-61 The protection against confirmed infection, however, started to wane from 4 weeks since the fourth dose.^{59, 61} Magen et al.⁵⁹ estimated a VE of 61% (95% CI = 58%–64%) against symptomatic disease 14–30 days after the fourth dose, which is consistent with the VE found 14–30 days after the third dose (57%; 95% CI = 51%–64%).^{7, 30, 32-34} The vaccine-induced antibody titers after the fourth dose have been evaluated in few studies.^{50, 51} Two studies have enrolled 31 and 154 individuals and found a 7.8 and 11.4-fold rise in IgG 14 days after the fourth BNT162b2 dose.^{50, 51} A consistent 10.7-fold increase in neutralizing antibodies against Omicron was observed.⁵¹ This increase in neutralizing antibodies was similar to the one reported in our study 14 days after the third BNT162b2 dose (9.8- to 12.4-fold increase). It was concluded that the maximal immunogenicity of mRNA vaccines that was achieved after three doses was similar compared to antibody levels generated after the fourth dose.^{50, 51} Taken all together, VE and antibody levels after the second booster were consistent with the ones after the first booster if considering the same time intervals since injection. Our linear model that could predict the level of VE according to neutralizing titers might therefore also be applicable for the fourth dose, but this deserves further validation.

The efficacy of the current formulation of the vaccine is at most around 65% against symptomatic disease 4 weeks after the administration of the first booster. On 25 June 2022, Pfizer and BioNTech announced that they are working on a Omicron-adapted mRNA vaccine.⁶² On 31 August 2022, the U.S. Food and Drug Administration amended the emergency use authorization of the Moderna COVID-19 Vaccine and the Pfizer-BioNTech COVID-19 Vaccine to authorize bivalent formulations of the vaccines. This bivalent vaccine, also called “updated booster,” contains two mRNA components of the virus (i.e., one original strain and one in common between the BA.4 and BA.5 lineages).⁶³ This adapted version is expected to boost the protection against Omicron.

5 CONCLUSION

A rapid and significant increase in booster-induced antibodies was observed from 7 days after the homologous BNT162b2 booster. Thereafter, a considerable antibody waning was noticed within 6 months, which was strongly correlated to VE data available in the literature. The impact of previous SARS-CoV-2 infection on the humoral response was nonsignificant after a first complete cycle of vaccination. Binding and neutralizing antibodies against Omicron followed a similar kinetics of decay but were only modestly correlated. A substantial increase of cut-off for binding antibodies was needed to increase the prediction of a neutralizing activity. Nevertheless, there was still a considerable number of participants with high binding antibody titers that did not present any neutralizing capacity. Commercial assays available in clinical laboratories might therefore require adaptation to better predict neutralizing antibodies, which represent the best correlate of protection. Our kinetic models might also be useful to determine the timing of fourth dose administration.

AUTHOR CONTRIBUTIONS

Conceptualization: Julien Favresse and Jonathan Douxfils. Methodology: Julien Favresse, Constant Gillot, Jean-Louis Bayart, Clara David, Germain Simon, Loris Wauthier and Mélanie Closset. Software: Julien Favresse, Constant Gillot, and Jonathan Douxfils; Validation: Julien Favresse and Jonathan Douxfils. Formal analysis: Julien Favresse, Constant Gillot, and Jonathan Douxfils. Investigation: Julien Favresse, Constant Gillot, Jean-Louis Bayart and Jonathan Douxfils. Resources: Jean-Michel Dogné and Jonathan Douxfils. Data curation: Julien Favresse, Constant Gillot, Jean-Louis Bayart, Clara David, Germain Simon, Loris Wauthier, and Mélanie Closset. Writing—original draft preparation: Julien Favresse. Writing—review and editing: Julien Favresse, Constant Gillot, Jean-Louis Bayart, Loris Wauthier, Jean-Michel Dogné, Jonathan Douxfils. Supervision: Jonathan Douxfils. Project administration: Julien Favresse, Jean-Louis Bayart, Mélanie Closset, and Jonathan Douxfils. Funding acquisition: Julien Favresse, Jean-Louis Bayart, Mélanie Closset, Jean-Michel Dogné, and Jonathan Douxfils. All authors have read and agreed to the published version of the manuscript.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.