Comparison of bivalent and monovalent SARS-CoV-2 variant vaccines: the phase 2 randomized open-label COVAIL trial

Vaccine protection against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection wanes over time, requiring updated boosters. In a phase 2, open-label, randomized clinical trial with sequentially enrolled stages at 22 US sites, we assessed safety and immunogenicity of a second boost with monovalent or bivalent variant vaccines from mRNA and protein-based platforms targeting wild-type, Beta, Delta and Omicron BA.1 spike antigens. The primary outcome was pseudovirus neutralization titers at 50% inhibitory dilution (ID50 titers) with 95% confidence intervals against different SARS-CoV-2 strains. The secondary outcome assessed safety by solicited local and systemic adverse events (AEs), unsolicited AEs, serious AEs and AEs of special interest. Boosting with prototype/wild-type vaccines produced numerically lower ID50 titers than any variant-containing vaccine against all variants. Conversely, boosting with a variant vaccine excluding prototype was not associated with decreased neutralization against D614G. Omicron BA.1 or Beta monovalent vaccines were nearly equivalent to Omicron BA.1 + prototype or Beta + prototype bivalent vaccines for neutralization of Beta, Omicron BA.1 and Omicron BA.4/5, although they were lower for contemporaneous Omicron subvariants. Safety was similar across arms and stages and comparable to previous reports. Our study shows that updated vaccines targeting Beta or Omicron BA.1 provide broadly crossprotective neutralizing antibody responses against diverse SARS-CoV-2 variants without sacrificing immunity to the ancestral strain. ClinicalTrials.gov registration: NCT05289037.


Study population
From 30 March to 6 May 2022, 602 participants were randomized and 597 received a mRNA-1273 vaccine (Moderna) in stage 1 (Table 1 and Fig. 1) 21 .From 12 to 27 May 2022, 313 participants were randomized and 312 received a BNT162b2 vaccine (Pfizer-BioNTech) in stage 2. From 8 to 17 June 2022, 153 participants were randomized and 152 received a pre-S DTM AS03 protein vaccine (Sanofi) in stage 3.This study was conducted whereby each stage was designed independently and sequentially based on the availability of new products and platforms.Vaccine selection in previous stages also informed the design of subsequent stages.

Safety
The frequency and severity of solicited local and systemic adverse effects (AEs) after vaccination were similar to other booster trials 22 and did not differ between arms in each stage (Extended Data Figs. 1 and 2).Multiple AEs could occur in a single participant.The most frequently reported solicited local AE was injection-site pain (83% of participants for stage 1, 77% for stage 2 and 74% for stage 3).The most common solicited systemic AEs were fatigue (50-67%) and myalgia (39-57%).Most solicited AEs were mild to moderate, with only 0-1% severe local AEs and 0.7-4% severe systemic AEs.A summary of all AEs is presented in Extended Data Figs.1-3.As of the data cut-off, 13 participants in stage 1, 4 participants in stage 2 and 1 participant in stage 3 had a serious AE; all were deemed unrelated to study product.There was one related AE of special interest in stage 1 of a young man who reported chest pain 1 day after vaccination that was initially evaluated as possible myocarditis, which was ultimately excluded due to a normal troponin I level and normal cardiac magnetic resonance imaging.There was one death unrelated to study product due to cardiac arrest from advanced coronary artery disease.
For stage 1, PsVN Ab responses peaked at day 15 after vaccination, remained relatively stable at day 29, were similar between older (≥65 years) and younger adults, and were 2-3 times higher in participants who were previously infected compared with those who were uninfected (Supplementary Tables 6 and 7).PsVN Ab GMTs against all variants declined from day 29 to day 91 by a factor of 1.74 (95% CI, 1.69, 1.80) in participants who were previously uninfected and by a factor of 1.34 (95% CI, 1.25, 1.44) in those who were previously infected (Fig. 2 and Extended Data Figs.4-6).
The antibody responses with Omicron BA.1-containing vaccines demonstrated a trend of greater durability, with a smaller estimate of geometric mean fold decline (GMFD) from day 29 to day 91 for B.1.351(GMFD B.1.351= 1.4 to 1.7) and Omicron subvariants (GMFD BA.1 = 2.0 to 2.2 and GMFD BA.4/5 = 1.8 to 2.0) when compared with the prototype vaccine (GMFD B.1.351= 1.8, GMFD BA.1 = 2.3 and GMFD BA.4/5 = 2.1).Within each study arm, the ratio in geometric mean neutralization titer against variant pseudoviruses compared with the ancestral D614G pseudovirus (geometric mean ratio against D614G (GMR D614G )) was used as a measure of boosting effect, where lower values correspond to stronger responses of variant vaccines to variants other than D614G.GMR D614G values also reflect the extent of neutralization escape, where higher values correspond to greater escape.In stage 1, less reduction in neutralization titers against Omicron variants was observed for Omicron BA.1-containing vaccines (GMR D614G = 7.13 to 8.72 for BA.1 and 13.40 to 16.13 for BA.4/5) than with the prototype vaccine (GMR D614G = 12.0 for BA.1 and 20.6 for BA.4/5) at day 91 (Extended Data Figure 5 and Supplementary Table 6).

Neutralizing antibody responses for stage 2
Stage 2 participants were boosted with either the Pfizer BNT162b2 wild-type vaccine or one of five different variant-targeting versions of Pfizer BNT162b2 COVID-19 vaccine, including monovalent BA.1, monovalent B.1.351, a bivalent BA.1 + wild-type vaccine, and two additional bivalent vaccines comprising B.1.351and either BA.1 or wild-type spike (Table 1).Neutralizing antibodies were assessed with the same assay used for the main dataset in stage 1.
Consistent with stage 1 results involving a similar mRNA vaccine, PsVN Ab GMT estimates peaked on day 15, remained relatively stable on day 29, were similar between older (≥65 years) and younger adults, and were 2-4 times higher in participants who were previously infected compared with those who were uninfected (Fig. 2, Supplementary Tables 8 and 9 and Extended Data Figs.4-6).
B.1.351,BA.1 and BA.4/5 share a common set of mutations in the RBD (K417N, E484K/A and N501Y), which might account for the modestly improved neutralizing antibody responses against Omicron seen with the Beta and Beta + wild-type vaccines compared with the wild-type monovalent vaccine.However, although monovalent Omicron BA.1 and monovalent Beta vaccines similarly boosted titers to the B.1.351variant (GMT B.1.351 of 6,253 and 6,247, respectively), they numerically differed in their ability to neutralize Omicron BA.1 (GMT BA.1 of 2,480 and 1,411, respectively).

Neutralizing antibody responses for stage 3
Stage 3 participants were boosted with one of three pre-S DTM AS03 protein vaccine products, including the prototype vaccine, a monovalent Beta vaccine and a bivalent Beta + prototype vaccine (Table 1).Neutralizing antibodies were assessed on days 1, 29 and 91 in the same assay used for the main datasets in stages 1 and 2. Day 15 samples were not tested for stage 3.
PsVN Ab GMT estimates at day 29 after vaccination with Sanofi variant vaccines were similar between older and younger adults and approximately 2-5 times higher in participants who were previously infected compared with those who were uninfected (Fig. 2, Supplementary Tables 10 and 11 and Extended Data Figs.4-6).For participants who were uninfected, all Beta-containing vaccines boosted D614G antibody titers (day 29 GMT D614G between 9,384 and 11,726) better than the prototype vaccine (day 29 GMT D614G = 6,942) (Extended Data Figure 5 and Supplementary Table 10).The prototype vaccine was less effective in boosting against most variants (day 29  Similar or a trend of modestly more durable antibody responses were seen in PsVN Ab titers from day 29 to day 91 with Beta-containing vaccines against Omicron subvariants (GMFD BA.1 = 1.5 to 2.1 and GMFD BA.4/5 = 1.5 to 1.7) when compared with prototype vaccine (GMFD BA.1 = 1.5 and GMFD BA.4/5 = 2.0).In addition, compared with responses against D614G, less reduction in neutralization titers for Omicron variants was observed for the Beta + prototype vaccine (GMR D614G = 9.1 for BA.1 and 11.6 for BA.4/5) than with the prototype vaccine (GMR D614G = 13.1 for BA.1 and 21.5 for BA.4/5) at day 91, based on point estimates, although CIs were overlapped (Fig. 2, Supplementary Table 10 and Extended Data Figure 5).

Analysis of covariance modeling of variant vaccines to prototype/wild type
In analysis of covariance (ANCOVA) models (adjusted for baseline titers, age and baseline infection status) for each stage, the day 91 GMR comparing neutralization titers with variant-containing vaccines to first generation prototype/wild-type vaccines against the ancestral D614G variant ranged from 1.01 to 1.40 for each variant vaccine within the 3 stages.
In stage 1, all Omicron BA.1-containing Moderna vaccines led to a day 91 GMR BA.1 ≥ 1.88, GMR BA.4.5 ≥ 1.70 and GMR B1.351 ≥ 1.50 compared with the prototype vaccine, with unadjusted lower-bound CIs >1 (Extended Data Table 2).In stage 2, all Omicron BA.1-or Beta-containing Pfizer vaccines led to a day 91 GMR BA.1 ≥ 1.99, GMR BA.4.5 ≥ 1.8 and GMR B.1.351≥ 1.78 compared with the wild-type vaccine (Extended Data Table 3).The day 91 GMRs in stages 1 and 2 were similar or higher to those observed for     day 29.In stage 3, all Beta-containing Sanofi vaccines led to a day 91 GMR of greater than 1 relative to the prototype vaccine, although the unadjusted lower-bound CI failed to exclude 1 (Extended Data Table 4).

Antigenic cartography and antibody landscapes
Antigenic cartography is a method to visualize antigenic relationships of virus variants in a two-dimensional map, where the distance in the map corresponds to neutralization properties of the variants 23 .We constructed antibody landscapes 24 , where neutralization titers are plotted in a third dimension above the variants in an antigenic map, to visualize how immunity in the different study arms distributes across antigenic space.The base map we used here was derived from a map by ref. 25 (Fig. 4a).Figure 4b shows the GMT antibody landscapes for each vaccine arm in the three stages stratified by prior infection, with the corresponding neutralizing antibody titers above the variant's map position.Lower landscapes correspond to day 1 and upper landscapes to day 91 immunity.To interpret landscapes, a day 91 response where the upper landscape is flat indicates the titers to all the variants were equivalent, whereas skewing up or down indicates titer differences across variants.The surface colors represent individual study arms.All vaccine arms for each of their respective stages in participants who were uninfected had similar prevaccination antibody landscapes, with the apex over D614G, as expected (Fig. 4b).After vaccination, all arms, in all 3 stages, had antibody titers that raised and flattened the landscape.In uninfected cohorts for all three stages, variant-containing vaccines lifted titers against BA.1 and BA.4/5 and produced flatter landscapes in the antigenic space surrounding these variants than did the prototype or wild-type vaccines.A second booster dose raised antibody titers in participants who were uninfected to the titers observed in participants who were previously infected at baseline (Fig. 4b).

SARS-CoV-2 infections
There were 267 self-reported COVID-19 illnesses occurring after randomization among 973 participants in single dose arms by data cut-off, 1 of which resulted in a brief hospitalization, lasting less than 24 hours, due to hypoxemia.The incidence of infections in this trial reflect the community transmission, with the majority occurring during the Omicron BA.5 wave in the United States.At any point in time, participants from different stages were in different points in follow-up, thereby preventing assessment of incidence across stages.Kaplan-Meier (KM) estimates of infections at the end of the follow-up period were similar among arms within a stage (Supplementary Tables 4-6).A higher percentage of infections, across all stages, was noted in participants with no history of prior infection (KM estimate, 37.8%; 95% CI, 31.8%,44.6%) compared with those with a history of prior infection (KM estimate, 12.1%; 95% CI, 8.4%, 17.2%).There were also fewer infections in adults ≥65 years (KM estimate, 19.3%; 95% CI, 15.1%, 24.5%) compared with their younger counterparts (KM estimate, 36.2%;95% CI, 29.2%, 44.4%) across all stages.

Discussion
The continued emergence of SARS-CoV-2 VOCs led to a recommendation to update COVID-19 vaccines 26 .The strains selected in 2022 for modified vaccines covered circulating strains at the time of vaccine development, not necessarily variants that would drift antigenically from Omicron BA.1 and BA.4/5 or evolve from other distinct locations on the phylogenetic tree.Therefore, it is important to investigate not only immune responses to known variants but also the antigenic relationships among different SARS-CoV-2 VOCs 25 and how variant vaccines may alter immunologic landscapes to cover antigenic areas where new strains may emerge.Here we described the magnitude, breadth and landscapes of the neutralizing antibody response following a second booster with investigational monovalent and bivalent variant-specific vaccines reflective of the diverse SARS-CoV-2 immunologic background seen in the general population.Our randomized study, using different vaccine platforms, offers the most comprehensive assessment of how vaccination with variants antigenically distinct from the ancestral strain compare in the ability to produce a broadly crossneutralizing antibody response and provides several insights to inform future SARS-CoV-2 vaccine policy.
First, our findings support that mRNA and adjuvanted protein variant vaccines elicit substantial crossreactive neutralizing antibodies to D614G and to B.1.351,1.617.2,Omicron BA.1, Omicron BA.4/5 and other Omicron subvariants, regardless of prior SARS-CoV-2 infection history and age.This is probably due to ongoing antibody somatic mutation, memory B cell clonal turnover and development of antibodies that are resistant to SARS-CoV-2 spike protein RBD mutations 27,28 .
Second, our ANCOVA modeling demonstrated that the mRNA variant vaccines offered a clear serologic advantage over the wild-type/ prototype vaccines against B.1.351,BA.1 and BA.4/5 that persisted up to 3 months after vaccination.Moreover, vaccine candidates without Omicron BA.1 variant, such as the Pfizer mRNA Beta vaccine, still provided superior heterologous coverage to Omicron BA.1 and BA.4/5 when compared with the wild-type vaccine, which was probably due to the common mutations in the spike RBD (K417N, E484K/A and N501Y) between B.1.351and these Omicron variants.Although a serologic advantage to BA.1 was not seen in ANCOVA modeling with the Sanofi Beta or Beta + prototype protein vaccine candidates, perhaps due to small sample size or undetected prior infection, a similar serologic benefit of boosting with the Beta monovalent vaccine 29,30 , and superior clinical efficacy against Omicron BA.1 and BA.2, was seen in the manufacturer's phase 3 clinical trial 31 .
The antibody landscapes visualizing the neutralization profile after vaccination further support inclusion of variants in booster vaccines.After vaccination, the antibody landscape rises with variant vaccine candidates, especially against more recent variants, and flattens the antibody landscape more than the prototype vaccine, suggesting there may be higher titers of neutralizing antibodies with variant-containing vaccines against future VOCs, especially if they emerge near B.1.351,Omicron BA.1 and BA.4/5 (ref.32).
Although specific correlates of protection for infection with recent Omicron subvariants are not well understood, neutralizing antibody titers have been used to infer protection during the D614G wave of the pandemic, when the circulating virus closely matched the vaccine strain 33 , and the resulting immunologic data have served as the basis for emergency-use authorization for booster vaccines by regulatory agencies 34,35 .The improved serologic response with either Omicron BA.1 or Beta variant-containing vaccines over prototype/wild-type vaccines in our study and others 36-38 provides evidence that broad crossprotection may be conferred without a variant-chasing approach and warrants further mechanistic exploration.
For all vaccine candidates, including vaccine products not containing prototype, the antibody titers were higher against D614G compared with the VOCs, supporting the hypothesis of back-boosting to the ancestral strain seen in previous studies 24,37,38 .This suggests that future generations of SARS-CoV-2 vaccines may be able to omit prototype or wild-type sequences without losing the ability to neutralize D614G, or other variants within close antigenic distance, in people who previously received the prototype vaccines.Furthermore, Omicron BA.1 or Beta monovalent vaccines were nearly equivalent to Omicron BA.1 + prototype or Beta + prototype bivalent vaccines for neutralization of B.1.351and both Omicron subvariants (BA.1 and BA.4/5), further supporting the premise that monovalent variant vaccines could replace bivalent vaccines as the updated boost in the future 31 .
Notably, although variant vaccines improved neutralizing activity against Omicron subvariants, these titers decreased for more recent Omicron subvariants.Although the serum inhibitory dilution required for 50% neutralization (ID 50 ) against BA.1 and BA.4/5 remained high, the neutralization titers for subvariants BQ.1.1 and XBB.1 were much lower.In addition, we noted a high rate of infections that occurred during the BA.4/5 wave and subsequent waves with XBB.1 and BQ.1.1.These infections occurred more frequently in individuals who were previously uninfected compared with those who were previously infected, highlighting the importance of hybrid immunity in protection against disease 32 .In addition, infections occurred in younger rather than older adults, probably reflecting behavioral differences affecting risk of exposure.Our study was not designed to assess VE.Although recent data suggest possibly higher VE against Omicron subvariants with bivalent vaccine boosts (prototype + Omicron BA.4/5 and prototype + Omicron BA.1) compared with the prototype vaccine 18,39 , our findings highlight concerns that variant vaccines are unlikely to keep pace with virus evolution and that other immune correlates of protection beyond antibody responses need to be explored.
Our study has several limitations.First, the sample size is small for certain subgroups of interest, such as prior infection (27%) and adults older than 65 years (31%).Second, T cell responses and antibody effector functions, which may be critical to preventing severe disease 40 ,  have not yet been evaluated.In addition, clonal and kinetic analyses of the memory B cell response, although underway, are not available to further differentiate the durability of the antibody response elicited by variant-containing vaccines.Finally, participants were only randomized to different arms within each stage and not between stages that enrolled sequentially at different calendar times, leading to different exposures to circulating variants before and after enrollment.This precludes head-to-head comparisons of rates of infections or neutralization titers across stages.These results may also not extend to adenovirus vector or inactivated vaccines licensed and used more frequently in other parts of the world or future next-generation vaccines.
In conclusion, these data demonstrate that updating vaccines to target recent variants provides modestly improved and broadly crossprotective neutralizing antibody responses against diverse SARS-CoV-2 variants without sacrificing boosting immunity to the ancestral strain.The precise degree to which the enhanced antibody response elicited by updated vaccines will restore protection against disease after infection with heterologous or homologous strains needs further confirmation by real-world effectiveness studies.Our study incorporating both antigenic distances and serologic landscapes serve as a framework for objectively guiding decisions for future vaccine updates.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41591-023-02503-4.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.https://doi.org/10.1038/s41591-023-02503-4

Fig. 3 |
Fig. 3 | Pseudovirus neutralization ID 50 titers by time point and variant in a subset (n = 22-23) of participants who were uninfected.Time points were days 1, 15 and 91.Variants were D614G and Omicron BA.1, BA.2.12.1, BA.4/BA.5,BA.2.75, BA.4.6, BF.7, BA.2.75.2, BQ.1.1 and XBB.1.a, Stage 1 mRNA-1273 prototype monovalent vaccine.b, Stage 1 mRNA-1273 Omicron BA.1 + prototype bivalent vaccine.In a and b, boxes and horizontal bars denote interquartile range and median ID 50 , respectively; whiskers denote 95% CI; and n represents the number of samples tested.c,d, Radar plots of the pseudovirus neutralization GMTs at day 15 (c) and day 91 (d) for the two vaccine arms in stage 1 mRNA-1273 prototype monovalent vaccine (red) and mRNA-1273 Omicron BA.1 + prototype bivalent vaccine (blue).Circles are GMT estimates for each variant.In the radar plots, each variant is represented by its own vertical line or spoke, and the spokes are evenly distributed around the circle.Each horizontal line along a vertical spoke represents the GMT at a ten-fold dilution, with the value closest to the center being 1 and farthest from the center being 10,000 or 10 4 .A line is drawn connecting the GMT data values for vaccine arm at the individual variants represented by its vertical spoke. Articlehttps://doi.org/10.1038/s41591-023-02503-4 /doi.org/10.1038/s41591-023-02503-4

Fig. 4 |
Fig. 4 | Antigenic cartography.a, An antigenic map by ref. 25 served as the base map for all antibody landscapes.Virus variants are shown as color-filled circles.Variants with additional substitutions from their root variant are shown as smaller circles.Variants associated with significant outbreaks or pandemic waves are secondarily encircled in red.Individual sera from individuals who were infected are displayed as open squares in the color of their root variant or gray for mRNA-1273 vaccinated sera; small dark squares represent clinical trial participants.One grid unit in the map corresponds to a twofold dilution in the neutralization assay.Within the x and y axes, the map orientation is free as antigenic distances are relative.Small triangles point to sera outside the shown map area.b, Day 1 and day 91 GMT antibody landscapes for individuals who Articlehttps://doi.org/10.1038/s41591-023-02503-4