Neutralizing antibody responses elicited by SARS-CoV-2 mRNA vaccination wane over time and are boosted by breakthrough infection

The waning efficacy of SARS-CoV-2 vaccines, combined with the continued emergence of variants resistant to vaccine-induced immunity, has reignited debate over the need for booster vaccine doses. To address this, we examined the neutralizing antibody response against the spike protein of five major SARS-CoV-2 variants, D614G, Alpha (B.1.1.7), Beta (B.1.351), Delta (B.1.617.2), and Omicron (B.1.1.529), in health care workers (HCWs) vaccinated with SARS-CoV-2 mRNA vaccines. Serum samples were collected pre-vaccination, three weeks post-first vaccination, one month post-second vaccination, and six months post-second vaccination. Minimal neutralizing antibody titers were detected against Omicron pseudovirus at all four time points, including for a majority of patients who had SARS-CoV-2 breakthrough infections. Neutralizing antibody titers against all other variant spike protein-bearing pseudoviruses declined dramatically from one to six months after the second mRNA vaccine dose, although SARS-CoV-2 infection boosted vaccine responses. Additionally, mRNA-1273-vaccinated HCWs exhibited about two-fold higher neutralizing antibody titers than BNT162b2-vaccinated HCWs. Together these results demonstrate possible waning of antibody-mediated protection against SARS-CoV-2 variants that is dependent on prior infection status and the mRNA vaccine received. They also show that the Omicron variant spike protein can almost completely escape from neutralizing antibodies elicited in recipients of only two mRNA vaccine doses.


INTRODUCTION
Since its emergence in late 2019, the coronavirus disease 2019 (COVID- 19) pandemic has led to more than 323 million confirmed cases and more than 5 million deaths as of 18 January 2022 (1). In response, several vaccines have been developed against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of COVID-19, including two mRNA vaccines, Moderna mRNA-1273 and Pfizer/BioNTech BNT162b2. These highly effective vaccines have helped to stem COVID-19 hospitalizations and deaths. However, the rapid evolution of SARS-CoV-2, combined with waning vaccine efficacy, remains a threat to public health.
After its introduction into the human population, several SARS-CoV-2 variants of concern (VOCs) have emerged. Soon after initial identification, SARS-CoV-2 acquired a predominant D614G mutation in its spike protein. This mutation leads to enhanced transmissibility, likely due to increased stability of the spike protein, as well as increased viral titers in the nasopharynx and increased infectivity (2). As a result, nearly all currently circulating SARS-CoV-2 strains bear the D614G mutation (3). However, as greater proportions of the world population acquired immunity against SARS-CoV-2 through infection or vaccination, new VOCs emerged that had reduced susceptibility to antibody-mediated immune responses and continued to become more transmissible (4,5). One VOC, Alpha (B.1.1.7), is characterized by N-terminal domain (NTD) deletions and a key N501Y mutation in its receptor binding domain (RBD). The Alpha variant exhibited enhanced transmissibility and rapidly spread from Europe to other parts of the world (6). Another VOC to emerge at about the same time was Beta (B.1.351), which is characterized by other NTD mutations and deletions, as well as crucial RBD mutations, including K417N, E484K, and N501Y. Although the Beta variant did not disseminate as widely as the Alpha variant, it harbored strong resistance to vaccine-induced immunity (7). The recently emerged Delta (B.1.617.2) variant is characterized by additional NTD alterations together with crucial RBD mutations (L452R and T478K). The Delta variant led to an alarming number of vaccine breakthrough infections worldwide and initially prompted debate about the need for vaccine booster doses. Last, the Omicron (B.1.1.529) variant, responsible for the ongoing wave of the pandemic, is characterized by an alarming number of mutations, including 16 in the RBD alone. The Omicron variant has produced an unprecedented peak in COVID-19 cases worldwide and appears to continue to exhibit enhanced escape from vaccine-induced immunity (8)(9)(10).
The extent to which the rise in breakthrough infections is caused by increased resistance to vaccine-induced immunity in these variants or to waning durability of immunity conferred by vaccines remains unclear. For the Delta variant, reports from India, where the population was still pursuing mass vaccination efforts, show minor differences in breakthrough infection rates between Alpha and Delta variants. Specifically, BNT162b2 efficacy against symptomatic infection was reported to drop from 93.7% against Alpha to 88.0% against Delta (11). However, reports from the United States indicate that vaccine efficacy of BNT162b2 against infection with the Delta variant declined from 93% 1 month after vaccination to 53% at 4 months (12), consistent with an overall waning of vaccine efficacy over time (13). A critical goal of this study is to understand how the durability of vaccine efficacy might contribute to rates of breakthrough infections, especially in the context of evolving SARS-CoV-2 variants. Such insights will aid in decision-making for allocating booster doses, including updating recommendations for immunocompromised patients. These data may also help guide the reformulation of future SARS-CoV-2 booster doses. In this study, we examined neutralizing antibody (nAb) titers in 48 vaccinated health care workers (HCWs) against the major SARS-CoV-2 VOCs, as prior studies have shown that nAb titers are a major correlate for protection from SARS-CoV-2 infection (14). Serum samples were collected before vaccination, 1 month after the first dose of BNT162b2 or mRNA-1273, and 1 and 6 months after the second dose of vaccine.

SARS-CoV-2-specific immunity wanes over time after two mRNA vaccine doses
We produced pseudotyped lentiviruses expressing a Gaussia luciferase reporter gene and bearing SARS-CoV-2 spike proteins derived from D614G, Alpha, Beta, Delta, or Omicron (Fig. 1A). D614G was selected to serve as the control virus because this mutation emerged very early in the pandemic, is present in nearly all circulating SARS-CoV-2 variants, and substantially affects neutralization sensitivity. In light of this, comparisons of neutralization sensitivity to true wild-type SARS-CoV-2 would be confounded by the presence of the D614G mutation in all major SARS-CoV-2 variants. Pseudotyped virus infectivity was then determined by infection of human embryonic kidney (HEK) 293T cells expressing angiotensin-converting enzyme 2 (HEK293T-ACE2 cells). Gaussia luciferase secreted into the media of infected cells was assayed to determine the infectivity of these pseudotyped lentiviruses. We found that pseudotyped virus infectivity was reasonably comparable for all variants, with modest reductions seen for Alpha and Delta compared to D614G (Fig. 1B), despite reports of drastically increased transmission and spread for some VOCs, especially the Delta variant (15). However, the high transmission rate for the Delta and Omicron variants appears to be related to faster replication kinetics and improved replication in the nasopharynx for these viruses (16,17), which would not be reflected in a pseudotyped virus system.
We used our previously reported (18,19) (IQR = 190 to 197.75) after second vaccine dose. According to the predetermined titer of pseudotyped viruses, we adjusted the volumes of each so that equivalent amounts of infectious virus were used in neutralization assays. HCW serum samples were fourfold serially diluted, and then pseudotyped virus was added for 1 hour at final dilutions of 1:80, 1:320, 1:1280, 1:5120, and 1:20480 and no serum control. For a valid comparison, samples from the same individual collected at different time points were tested side by side on the same plate. HEK293T-ACE2 cells were infected with the virusserum mixture, and Gaussia luciferase activity was assayed 48 and 72 hours after infection. Neutralizing titer 50% (NT 50 ) values were determined by a nonlinear least squares regression model.
We compared the strength of the nAb titers over time against all five variants tested. After the first dose of mRNA vaccine, a strong nAb response was induced among HCWs compared to prevaccination across all variants except for Omicron; this was observed despite extensive variation in nAb titers of these individuals, including against D614G (mean = 1140, 95% confidence interval = 317 to 1963, and range = 100 to 15954) (Fig. 1C). A fraction of HCWs exhibited NT 50 values below the detection limit (NT 50 less than 100) against D614G (7 of 48), Alpha (17 of 48), Beta (22 of 48), and Delta (14 of 48) after the first dose of vaccine ( Fig. 1, C to F). In contrast, a notable 93.8% (45 of 48) fell below the detection limit for the Omicron variant (Fig. 1G). The below-detection rates fell to 0.0% (0 of 48) to 4.2% (2 of 48) for all variants except Omicron after a second vaccine dose, with a two-to threefold increase in mean nAb titers compared to the first dose (P < 0.001 for D614G, Alpha, Beta, and Delta) ( Fig. 1, C to F). Once again, 25.0% (12 of 48) of samples remained below the detection limit for Omicron for the post-second dose, and the overall increase in NT 50 was not significant compared to the first dose (P > 0.05; Fig. 1G). Four HCWs with higher nAb titers after the first vaccine dose did not show an increase, but rather, a plateau or slight decline in nAb titers after the second dose ( Fig. 1, C to F). These four individuals included one who had antibodies against the SARS-CoV-2 nucleocapsid (N) protein before vaccination and three who had anti-N protein antibodies after their first vaccine dose, which indicated infection either before or shortly after their first vaccine dose. The single patient exhibiting strong neutralization of Omicron had antibodies against N protein after the first vaccine dose (Fig. 1G), suggesting breakthrough infection. Overall, we found that, after two vaccine doses, the Alpha, Beta, Delta, and Omicron VOCs exhibited a 1.3-fold (P < 0.01), 3.2-fold (P < 0.001), 2.2-fold (P < 0.001), and 28.6-fold (P < 0.001) lower NT 50 values compared to D614G, respectively ( Fig. 1, C to G). Critically, 6 months after vaccination, there was a 3.5-to 10.7-fold reduction in nAb titers against the Alpha, Beta, and Delta variants, with 37.5% (18 of 48) to 56.3% (27 of 48) of HCWs exhibiting NT 50 values below the limit of detection ( Fig. 1, C to F). Consistent with earlier time points, minimal responses against Omicron were detected in samples collected 6 months after second vaccination, with 89.6% (43 of 48) of HCWs falling below the limit of detection (Fig. 1G). The mean NT 50 values for Alpha, Beta, Delta, and Omicron variants at 6 months were 1.3-, 1.7-, 3.6-, and 10.2-fold lower than that of D614G, respectively ( Fig. 1, C to G).
We also examined the correlation between time after second dose and log 10 -transformed NT 50 values. We found a statistically significant association (P < 0.001) between these values for all variants except Omicron ( Fig. 1, H to L). This corresponded to an about 10-fold decline in NT 50 for D614G, Alpha, and Delta (R 2 = 0.0452 to 0.594 and P < 0.0001) every 22 weeks compared with Beta (R 2 = 0.286 and P < 0.001) every 37 weeks (Fig. 1, H to K). For Omicron, only a weak, but significant, correlation was observed (R 2 = 0.064 and P = 0.0128) (Fig. 1L), likely due to the overall weak responsiveness against Omicron at all four time points. The schematics highlight the location of the S1 and S2 subunits of the spike protein as well as the N-terminal domain (NTD), receptor binding domain (RBD), fusion peptide (FP), transmembrane (TM) region, and the indicated mutations. (B) Pseudotyped lentiviruses were used to infect HEK293T-ACE2 cells. Media was harvested from infected cells 48 hours after infection and assayed for Gaussia luciferase activity to determine the relative infectivity of each variant pseudotyped virus; infectivity was measured as luminescence signal relative to D614G signal, relative units (ru). Significance relative to D614G was determined by one-way ANOVA with Bonferroni's correction (n ≥ 3 biological replicates); error bars represent means ± SE. (C to G) Lentivirus pseudotyped with SARS-CoV-2 spike protein from D614G (C), Alpha (D), Beta (E), Delta (F), and Omicron (G) was neutralized with serum samples from health care workers (HCWs) (n = 48 biological replicates) collected before vaccination (Pre), after vaccination with a first mRNA vaccine dose (Post 1st), after vaccination with a second mRNA vaccine dose (Post 2nd), and 6 months after second dose (6 months). Neutralization titers 50% (NT 50 ) values were determined by least squares fit nonlinear regression. Mean NT 50 values are shown at the top of the plots, and NT 50 values below 100 were considered background, as indicated by a dashed line; significance was determined by one-way repeated-measures ANOVA with Bonferroni's correction. (H to L) Log 10 -transformed NT 50 values against D614G (H), Alpha (I), Beta (J), Delta (K), and Omicron (L) variants were plotted against days after second vaccine dose of sample collection (n = 96 biological replicates). The goodness of fit (R 2 ) and P values are displayed on each plot as determined by least squares fit linear regression. The dashed lines correspond to the background (NT 50 < 100). In all cases, *P < 0.05, **P < 0.01, and ***P < 0.001; ns, not significant.

Breakthrough SARS-CoV-2 infection boosts nAb responses
Prior COVID-19 status is a critical parameter for the nAb response to vaccination (20). Of the 48 HCWs examined, 1 HCW was positive for antibodies against the SARS-CoV-2 N protein (anti-N protein positive) by enzyme-linked immunosorbent assay (ELISA) before vaccination, indicating a SARS-CoV-2 infection before vaccination. Four HCWs were anti-N protein positive at their post-first vaccine dose sample, indicating a SARS-CoV-2 infection before or within 3 weeks after their first vaccine dose. Three HCWs were anti-N protein positive at their post-second vaccine dose sample, indicating a SARS-CoV-2 infection more than 3 weeks after their first dose but within 4 weeks of their second dose. Last, samples from four HCWs collected 6 months after their second vaccine were anti-N protein positive, indicating a SARS-CoV-2 infection more than 4 weeks after their second dose. In total, 12 of 48 individuals were infected by SARS-CoV-2 at different phases of vaccination ( Fig. 2A). At the time of prevaccination sample collection, D614G was the major circulating SARS-CoV-2 variant, whereas at the time of post-first dose and post-second dose sample collections, D614G and Alpha were the major circulating variants. The Delta variant was the dominant circulating strain at the 6-month time point. Not all patients remained anti-N protein positive but were still considered to have been infected for the purpose of analysis.
After the first vaccine dose, anti-N protein-positive HCWs (n = 5) exhibited a 22.6-fold higher mean NT 50 value (P < 0.001) against all five viruses compared to the anti-N protein-negative HCWs (n = 43) (Fig. 2B). This difference fell to 2.6-fold after a second vaccine dose (P < 0.001) (n = 8 anti-N protein-positive HCWs and n = 40 uninfected anti-N protein-negative HCWs) (Fig. 2B). However, at 6 months after vaccination, anti-N protein-positive HCWs (n = 12) exhibited 5.9-fold higher NT 50 values than anti-N protein-negative HCWs (n = 36) for all variants (P = 0.033) (Fig. 2B). We found that the differences in NT 50 between anti-N protein-positive and anti-N protein-negative HCWs were greater for D614G and Alpha compared with the Beta, Delta, and Omicron variants, possibly because of the stronger neutralization resistance of the latter VOCs (Fig. 2C). For anti-N protein-negative HCWs, between 41.7% (15 of 36) and 66.7% (24 of 36) of individuals exhibited nAb titers against all variants except Omicron that were below the detection limit at 6 months, in contrast to anti-N protein-positive individuals, who were between 8.3% (1 of 12) and 25.0% (3 of 12) (Fig. 2C). The nAb titers for Omicron were too low at all time points to compare the anti-N proteinpositive and anti-N protein-negative HCWs (Fig. 2C).

Vaccine type and sex, but not age, affect nAb responses elicited by SARS-CoV-2 vaccination
We further examined the difference in nAb durability between HCWs who received either the Moderna mRNA-1273 vaccine or the Pfizer/BioNTech BNT162b2 vaccine. Across all variants over the full-time course, we observed that mRNA-1273 elicited an overall 2.2-fold higher nAb response than BNT162b2 (P < 0.001) (Fig. 3A). In particular, after two vaccine doses, mRNA-1273-vaccinated HCWs exhibited significantly higher nAb titers than BNT162b2vaccinated HCWs for the D614G (P < 0.001) and Alpha (P < 0.01) variants, respectively (Fig. 3B). However, by 6 months after second vaccine dose, no significant differences were observed between mRNA-1273-and BNT162b2-vaccinated HCWs (P > 0.05; Fig. 3B). Again, because most of the HCW samples exhibited an NT 50 value below the limit of detection against Omicron, a comparison between mRNA-1273 and BNT162b2 was not informative (Fig. 3B).
We examined additional factors that may also contribute to the strength and duration of the nAb response to vaccination, including age and sex. We found that male HCWs exhibited significantly higher NT 50 titers (P < 0.001) compared to females against all five variants over the postvaccination time points (Fig. 3C). However, this difference was not significant when examined for individual viruses at individual time points (P > 0.05; fig. S1). Studies on sex-based differences in response to COVID-19 mRNA vaccination have yielded conflicting results (21,22), although females seem to induce more robust humoral immune responses after vaccination (23). We observed no significant correlation for age and NT 50 against D614G at any time point (P > 0.05; fig. S2), which could be due to the relatively young pool of individuals in this cohort (median age of 37 and maximum age of 61) compared to older adults in a prior mRNA vaccine efficacy study (24).

DISCUSSION
Here, we report a decline of SARS-CoV-2 nAb at 6 months after mRNA vaccination and present several key factors that may contribute to these kinetics. We observed a drop in nAb titers from 1 to 6 months after the second vaccine dose, with 56.3 and 89.6% of HCWs exhibiting NT 50 values below the limit of detection against Delta and Omicron at 6 months, respectively. We found that the NT 50 values of anti-N protein-positive HCWs were about six times higher than that of anti-N protein-negative HCWs at 6 months after vaccination, with about 30% below the background of detection for the former compared to about 60% for the latter. This suggests that breakthrough infection can improve the durability of the SARS-CoV-2 nAb response, which is consistent with data collected after administration of mRNA vaccine booster doses (25). We observed a profound escape of the Omicron variant from mRNA vaccine-induced immunity, even at 3 to 4 weeks after the second dose of mRNA vaccination. Further, this escape was not rescued in most HCWs by breakthrough infection. Together, these results, along with recent reports of enhanced neutralization of Omicron after booster dose administration (9,26), support the need for boosters and alternative vaccination strategies to achieve long-term protection from infection with SARS-CoV-2.
In addition, we observed that individuals vaccinated with BNT162b2 exhibited lower nAb titers than individuals vaccinated with mRNA-1273. However, the trend for declining nAb titers was consistent for both vaccines. Thus, both mRNA vaccines require booster doses to maintain protective nAb titers, although the waning of nAb responses likely occurs over a relatively (C) NT 50 values against all variants at post-first dose, post-second dose, and 6 months after second dose combined were compared for male (n = 390) and female (n = 330) participants; mean NT 50 values are displayed at the top of the plot, with significance determined by unpaired, two-tailed t test with Welch's correction. For all panels, error bars indicate means ± SEs, and the dashed horizontal line indicates the limit of detection (NT 50 < 100). In all cases, **P < 0.01 and ***P < 0.001. long period for mRNA-1273. Further examination of the durability of cellular immunity after mRNA vaccination is warranted, because this more persistent immune response may limit the rates of hospitalization and death, which remain low for mRNA-vaccinated individuals (27).
In this study, we found that all four VOCs consistently had reduced NT 50 values compared to D614G at all time points, with the Omicron variant showing the most pronounced nAb resistance, followed by the Beta and Delta variants. These results are consistent with preliminary reports from our group and others (9,16,19,28,29). However, we found that the Delta variant exhibited comparable or even higher resistance to nAbs than the Beta variant for samples collected at 6 months after vaccination. The more modest drop in NT 50 values at 6 months for the Beta variant was unclear but could be the result of this variant's preexisting resistance to neutralization after the second dose of vaccination. Further, the more marked decline in nAb titers against the Delta variant could be attributed to a lower frequency and durability of nAb-producing plasma cells. As reported by others, the rampant spread of the Delta variant in vaccinated and unvaccinated populations is likely related to other factors, such as its high replication kinetics and transmissibility (30), coupled with its comparable neutralization resistance.
The major limitation of this study is its reliance on nAb data. Although nAb titers remain a crucial predictor of protection from viral infection, no clear cutoff is available to define an NT 50 threshold that indicates protective immunity. Further, other immune factors, including T cell immunity, may have a greater impact on protection from severe disease. In addition, this study uses a pseudotyped virus system rather than authentic SARS-CoV-2. Although our pseudotyped virus system has been shown to correlate well with authentic virus systems for determining neutralization titers (18), it is not a perfect recapitulation of the authentic virus system.
Overall, our data highlight the need for booster vaccinations because the nAb response elicited by a two-dose vaccine regimen wanes over time and is largely ineffective against the Delta and Omicron variants by 6 months. In addition, although breakthrough infection can boost nAb responses, it appears largely ineffective for providing protection from Omicron, at least for individuals infected before the Omicron wave. Last, further study is warranted on the durability of the nAb response in booster vaccine recipients. As SARS-CoV-2 variants continue to emerge, understanding the interplay between vaccine durability and virus evolution will be critical for revising vaccination strategies and formulations as we seek to curb the ongoing pandemic. vaccine dose between January and February of 2021. No other demographic information was collected on these HCWs. HCW COVID-19 status was determined by anti-N protein ELISA. Of the 48 HCWs examined, one was anti-SARS-CoV-2 N protein positive by ELISA for their prevaccination sample. In addition, four HCWs became anti-N protein positive for their post-first vaccine dose sample, three became positive for their post-second vaccine dose sample, and four became positive for their 6-month vaccine sample. This indicated that these 12 individuals were infected by SARS-CoV-2 at different phases after vaccination. HCW serum samples were then assessed for nAb titers against SARS-CoV-2 variant spike proteinpseudotyped lentiviruses. Individual-level NT 50 values are presented in data file S1.

Constructs for pseudotyped lentivirus production
The pseudotyped lentiviruses were produced using a previously reported protocol using the pNL4-3-HIV-1-inGluc vector (18,19,(31)(32)(33). This vector is a pNL4-3-HIV-1 Env construct and contains a Gaussia luciferase reporter gene with a cytomegalovirus promoter, both oriented in an antisense orientation relative to the HIV-1 genome. This Gaussia luciferase reporter gene contains a sense orientation intron, which prevents the expression of Gaussia luciferase in the virus-producing cells. However, after the intron is spliced from full-length virus genomes upon integrated into target cells, target cells can produce Gaussia luciferase, which is secreted in mammalian cell culture (34). Constructs encoding N-and C-terminal flag-tagged SARS-CoV-2 spike protein for each variant-D614G, Alpha (B.

Virus neutralization assays
Virus neutralization assays were performed as previously reported (18,19,33). In a 96-well format, HCW serum was fourfold serially diluted, and 100 l of pseudotyped virus was added to each well (final dilutions of 1:80, 1:320, 1:1280, 1:5120, and 1:20480 and no serum). To ensure comparable results between SARS-CoV-2 variants, equivalent amounts of infectious virus were used on the basis of the predetermined virus titers. Virus was incubated with HCW serum for 1 hour at 37°C, followed by infection of HEK293T-ACE2 cells seeded on a 96-well polystyrene tissue culture plate. Gaussia luciferase activity in cell culture media was then assayed 48 and 72 hours after infection, as described above. NT 50 for each serum sample was determined by nonlinear regression with least squares fit in GraphPad Prism 5 (GraphPad Software).

Anti-N protein ELISA
Anti-N protein ELISA was performed as previously reported (18). ELISA was performed by using the Epitope Diagnostics Inc. (EDI) Novel Coronavirus COVID-19 N protein IgG ELISA Kit (KT-1032, EDI) following the manufacturer's protocol. Briefly, 100 l of a 1:100 dilution of HCW serum was added to microplates coated with SARS-CoV-2 N antigen and incubated for 30 min. Plates were then washed and treated with 100 l of horseradish peroxidase (HRP)labeled anti-human immunoglobulin G antibody (31220, EDI) for 30 min. Then, plates were washed, and 100 l of ELISA HRP substrate (10020, EDI) was added and incubated for 20 min before 100 l of stop solution (10030, EDI) was added. Absorbance at 450 nm was read by a Biotek Synergy/HTX multimode plate reader using Gen 5 software.

Statistical analyses
Statistical analysis was done with GraphPad Prism 5. Comparisons between multiple groups were made using one-way analysis of variance (ANOVA) with Bonferroni's multiple testing correction (data in Figs. 1B and 2B) or one-way repeated-measures ANOVA with Bonferroni's multiple testing correction (data in Fig. 1, C to G). For comparisons between two treatments across multiple groups, a two-way ANOVA with Bonferroni's multiple testing correction was used (data in Figs. 2C and 3B and fig. S1). For comparisons between two groups, an unpaired, two-tailed Student's t test with Welch's correction was used (data in Fig. 3, A and C). For correlative analyses between two continuous variables, a linear regression model with least squares fit was used with log 10 -transformed NT 50 values to better approximate linearity (data in Fig. 1, H to L, and fig. S2). An  value of 0.05 was considered significant throughout.