Calibrated comparison of SARS-CoV-2 neutralizing antibody levels in response to protein-, mRNA-, and vector-based COVID-19 vaccines

SARS-CoV-2 neutralizing antibodies have been suggested to reflect the efficacy of COVID-19 vaccines. This study reports the direct comparison of the SARS-CoV-2 neutralizing antibody response elicited by a protein- (NVX-CoV2373), an mRNA- (Comirnaty), and a vector-based (Vaxzevria) COVID-19 vaccine, calibrated against the WHO international SARS-CoV-2 antibody standard, and further supports the use of neutralizing antibody levels as a correlate of protection.


Full Text
Since late 2019, Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has rapidly spread around the globe, and the resulting coronavirus disease 2019 (COVID- 19) pandemic has had unprecedented impact on healthcare systems, economics, and social interactions. At similarly unprecedented speed, the development, clinical investigation, and regulatory assessment of COVID -19 vaccines have been pursued 1 , resulting in emergency use authorization (EUA) of several vaccines just about one year after the virus had initially been described. The rst two formulations licensed in Western countries -Comirnaty (BioNTech-P zer) and Spikevax (Moderna) -employed an mRNA vaccine technology, while shortly thereafter, two vaccines based on recombinant adenoviral vectors -Vaxzevria (AstraZeneca) and Janssen COVID-19 Vaccine (Janssen) -received regulatory approvals in the European Union (EU) or the United States (US); all of these vaccines were characterized by high e cacy across gender, age groups and ethnicities [2][3][4][5] .
Another large group of COVID-19 vaccine candidates employs the biotechnological production of immunogenic viral proteins or protein subunits, in case of SARS-CoV-2 the spike (S) protein or its receptor-binding domain 1 . One of these candidates is NVX-CoV2373 (TAK-019; Novavax), a recombinant nanoparticle vaccine for which safety and immunogenicity 6 and subsequently high e cacy 7 have been demonstrated. Compared to mRNA vaccines, long-term storage above freezing is a bene cial characteristic that is especially important for the supply of low-and middle-income countries.
Correspondingly, as the rst protein-based vaccine, NVX-CoV2373 has recently been granted EUA in Indonesia 8 and the Philippines 9 , as well as a conditional marketing authorization in the EU 10 .
Large vaccination campaigns have meanwhile substantiated the effectiveness of the aforementioned mRNA-and vector-based vaccines, especially with respect to severe COVID-19 and thus COVID-19-related death. With respect to further vaccine candidates currently in the pipeline, there is a demand for estimating such performance indicators upfront, as the ultimately expected global reduction in case numbers (due to the already licensed vaccines) and ethical considerations (avoiding placebo groups when vaccines are already standard care) argue against large phase 3 e cacy studies 11 . Not unexpectedly, the levels of SARS-CoV-2 neutralizing antibodies (nAbs), and antibodies binding to the S protein, have recently been identi ed as promising correlates of protection 12,13 , i.e., measuring SARS-CoV-2 antibodies in serum samples of vaccinees enables to predict the risk of developing COVID-19. However, while the clinical evaluation of vaccine candidates regularly includes such antibody readouts, the heterogenous design of the underlying virus neutralization and binding assays limits the quantitative comparison of the primary result (e.g., the neutralization titer, or arbitrary binding units) across distinct studies. These technical hurdles can be alleviated by the incorporation of an international standard, which has recently been made available to study SARS-CoV-2 antibody-containing samples 14 . Indeed, this international standard has recently been employed for analyses of vaccinee sera from the two mRNA-and two vector-based vaccines mentioned above to predict an overall protective level of S proteinbinding antibodies 15 . Further, using a pseudovirus neutralization assay, Gilbert et al. found that anti-SARS-CoV-2 potencies of 8 and 140 international units per milliliter (IU/mL) correspond to 70% and 90% e cacy of the Spikevax vaccine, respectively 16 . By additional comparison to a similar study on Vaxzevria recipients' sera 17 , the authors also showed an encouragingly congruent quantitative relationship (4 and 83 IU/mL corresponding to 70% and 90% e cacy, respectively), despite the different vaccine platforms and pseudoviral assays.
Previously, the explanatory power of SARS-CoV-2 nAbs with respect to vaccine e cacy has been investigated by an indirect approach, i.e., via normalization of neutralization titers to cohorts of convalescent (post-COVID-19) individuals analyzed in parallel 12,13 . However, such cohorts are subject to considerable variation (e.g., due to different parameters that de ne convalescence, due to diverging fractions of individuals that had suffered from severe versus mild disease, or sample size), as is the sequence of mathematical operations for normalization. Thus, while Khoury et al. found the average SARS-CoV-2 nAb levels induced by Vaxzevria to be lower than average post-COVID-19 levels 12 , Earle et al.
found slightly higher mean anti-SARS-CoV-2 potency of the vaccine 13 . The latter result is similar to our own data, as the mean vaccine-induced SARS-CoV-2 nAb level of 202 IU/ml (Table 1) is above the mean (140 IU/ml) of a post-COVID-19 group analyzed in one of our earlier studies 18 . The accuracy of the earlier used comparison to convalescent plasma for normalization must therefore be considered limited.
In contrast, the results of the present study are derived from the same assay and provide a rst direct comparison of the SARS-CoV-2 nAb response between three COVID-19 vaccines based on different immunogenic principles. Comparable anti-SARS-CoV-2 potency in sera of NVX-CoV2373 and Comirnaty recipients and the slightly lower levels in response to Vaxzevria are in line with previously published levels of vaccine e cacy 2,4,7 , directionally con rm the previous, more indirect approaches 12,13 and lend further support to the notion that neutralizing antibody responses represent a suitable correlate of protection.
Between the three vaccination groups, there is a somewhat diverging mean time between second vaccination and sample collection (Table 1), which is a limitation of our study. However, it should be noted that the peak immune response (and hence the achievement of a 'fully vaccinated' status) is generally considered to commence two weeks after the second shot (i.e., when serum samples from the NVX-CoV2373 group were obtained) and antibody waning occurs over several weeks to months rather than days (the latter being the difference in sampling between our three groups). Further, while distinct ethnicities might be a confounding factor of the present study, it should be noted that diverging COVID-19 vaccine e cacy between Asian and White vaccine recipients has not been reported 2,4 . Most importantly, the calibration of results against the rst international SARS-CoV-2 antibody standard 14 allows for an objective comparison beyond the present study population.
The neutralization assay included several validity criteria, i.e., con rmatory titration of input virus infectivity, cell viability, and neutralization testing of an internal reference standard, all of which had to comply with de ned ranges. Further, the internal standard was calibrated against the First WHO International Standard for anti-SARS-CoV-2 immunoglobulin (human; NIBSC code: 20/136) 14 to enable the quanti cation of anti-SARS-CoV-2 potency in international units per milliliter (IU/ml). Data from 28 subjects of the Comirnaty group has been included in a previous study 20 .

Ethics
The study was performed in accordance with applicable regulations, policies and procedures, and the authors' institutions required informed consent that was obtained from all study subjects. For Comirnaty and Vaxzevria groups, all subjects received two vaccinations according to EMA approval. For NVX-

DATA AVAILABILITY
The data that support the ndings of this study are available from the corresponding author upon reasonable request.
All authors declare no competing non-nancial interests, but the following competing nancial interests apply: MK, MRF and TRK are employees of Takeda Manufacturing Austria AG, Vienna, Austria, and have