Antibody repertoire induced by SARS-CoV-2 spike protein immunogens

Multiple vaccine candidates against SARS-CoV-2 based on viral spike protein are under development. However, there is limited information on the quality of antibody response generated following vaccination by these vaccine modalities. To better understand antibody response induced by spike protein-based vaccines, we immunized rabbits with various SARS-CoV-2 spike protein antigens: S-ectodomain (S1+S2) (aa 16-1213), which lacks the cytoplasmic and transmembrane domains (CT-TM), the S1 domain (aa 16-685), the receptor-binding domain (RBD) (aa 319-541), and the S2 domain (aa 686-1213 as control). Antibody response was analyzed by ELISA, Surface Plasmon Resonance (SPR) against different Spike proteins in native conformation, and a pseudovirion neutralization assay to measure the quality and function of the antibodies elicited by the different Spike antigens. All three antigens (S1+S2 ectodomain, S1 domain, and RBD) generated strong neutralizing antibodies against SARS-CoV-2. Vaccination induced antibody repertoire was analyzed by SARS-CoV-2 spike Genome Fragment Phage Display Libraries (SARS-CoV-2 GFPDL), which identified immunodominant epitopes in the S1, S1-RBD and S2 domains. Furthermore, these analyses demonstrated that surprisingly the RBD immunogen elicited a higher antibody titer with 5-fold higher affinity antibodies to native spike antigens compared with other spike antigens. These findings may help guide rational vaccine design and facilitate development and evaluation of effective therapeutics and vaccines against COVID-19 disease. One Sentence Summary SARS-CoV-2 Spike induced immune response


80
Female New Zealand white rabbits were immunized twice intra-muscularly at a 14-day 81 interval with 50 g of the purified proteins mixed with Emulsigen Adjuvant. Sera were collected 82 before (pre-vaccination) and after the first and second vaccination and analyzed for binding 83 antibodies in ELISA and SPR, in a pseudovirion neutralization assay, and by GFPDL analysis. IgG to various spike proteins and domains in ELISA (S1+S2; black, S1; blue, RBD; red, and S2; 88 green) (Fig. 1C). Representative titration curves to spike ectodomain (S1+S2) and to the RBD in 89 IgG-ELISA are shown in Suppl. Fig. 2. End-point titers of the serum IgG were determined as the 90 reciprocal of the highest dilution providing an optical density (OD) twice that of the negative 91 control (Fig. 1C). All four immunogens elicited strong IgG binding to the spike ectodomain 92 (S1+S2). Binding to the individual domains (S1, S2, and RBD) was specific, in that sera generated 93 by S2 vaccination bound to S2, but not to S1 or RBD, and vice-versa (Fig. 1C).

94
SPR allows following antibody binding to captured antigens in real-time kinetics, including 95 total antibody binding in resonance units (Max RU) and affinity kinetics (Suppl. Fig. 3). In ELISA, 96 the antigens directly coated in the wells can be partially denatured increasing the likelihood of 97 presenting epitopes that are not seen on the native form of the protein by the polyclonal serum IgG.

98
On the other hand, in our SPR, the purified recombinant spike proteins were captured to a Ni-NTA 99 sensor chip to maintain the native conformation (as determined by ACE2 binding) to allow 100 comparisons of binding to and dissociation from the four proteins. Importantly, the protein density 101 captured on the chip surface is low (200 RU) and was optimized to measure primarily monovalent 102 interactions, so as to measure the average affinity of antibody binding in the polyclonal serum (8, 103 13). Additionally, while ELISA measured only IgG binding, in SPR, all antibody isotypes 104 contributed to antibody binding to the captured spike antigen. In the current study, all rabbit sera 105 contained anti spike antibodies that were at least 86% IgG (data not shown). Serial dilutions of 106 post-vaccination serum were analyzed for binding kinetics with different spike proteins (Suppl. 107 Fig. 3). The spike ectodomain (S1+S2) generated antibodies that predominantly bound to S1+S2 108 6 (black bar), followed by the S1 protein (blue bar), and 3-fold lower antibody binding to the RBD 109 and the S2 domain (red and green bars, respectively) (Fig. 1D). The S1 domain antigen induced 110 antibodies that bound with similar titers (Max RU values) to the S1+S2, S1 and RBD proteins 111 (black, blue and red bars, respectively), and did not show reactivity to the S2 domain (green bar).

112
However, the antibody reactivity of rabbit anti-S1 serum to S1+S2 domain was 3-fold lower than 113 the antibodies in the rabbit anti-S1+S2 serum. RBD immunization generated similar high-titer 114 antibody binding to S1+S2, S1 and RBD (black, blue and red bars, respectively), (Fig. 1D). In 115 contrast, the S2 domain induced antibodies that primarily bound to homologous S2 antigen (green 116 bars) and only weakly binding to the S1+S2 ectodomain (black bars), and no binding to either S1 117 or RBD (Fig. 1D).

118
Antibody off-rate constants, which describe the fraction of antigen-antibody complexes 119 that decay per second, were determined directly from the serum sample interaction with SARS-120 CoV-2 spike ectodomain (S1+S2), S1, S2, and RBD using SPR in the dissociation phase only for 121 sensorgrams with Max RU in the range of 20-100 RU (Suppl. Fig. 3) and calculated using the 122 BioRad ProteOn manager software for the heterogeneous sample model as described before(11).

123
These off rates provide additional important information on the affinity of the antibodies following 124 vaccination with the different spike proteins that are likely to have an impact on the antibody 125 function in vivo, as was observed previously in studies with influenza virus, RSV and Ebola virus 126 (13-15). Surprisingly, we observed significant differences in the affinities of antibodies elicited by 127 the four spike antigens (Fig. 1E). Specifically, the RBD induced 5-fold higher affinity antibodies 128 (slower dissociation rates) against S1+S2 (black), S1 (blue) and RBD (red) proteins, compared 129 with the post-vaccination antibodies generated by other three immunogens (Fig. 1E).   This region may not be highly exposed on the virions or infected cells but is clearly immunogenic 150 in the soluble recombinant spike ectodomain. In addition, the rabbit anti-S1+S2 antibodies bound 151 diverse epitopes spanning the RBD and to a lesser degree to the N-terminal domain (NTD) and the 152 C-terminal region of S1, and the N-terminus of S2, including the fusion peptide ( Fig. 2B and Suppl.
153 Table 1). The S1 domain elicited very strong response against the C-terminal region of S1 protein 154 and a diverse antibody repertoire recognizing the NTD and RBD/RBM regions ( Fig. 2C and Suppl. 155 Table 1). The recombinant RBD induced high-titer antibodies that were highly focused to the 156 RBD/RBM (Fig. 2E, and Suppl. Table 1). In contrast, the recombinant S2 immunogen after two 157 immunizations in rabbits elicited antibodies primarily targeting the C-terminus of the S2 protein 158 (CD-HR2).  . Table 1). Structural depiction of these antigenic sites on the SARS-CoV-2 spike (Suppl.   Table 1). The other epitopes identified in our study cover less conserved sequences 182 between the two SARS-CoV viruses that are unique to the SARS-CoV-2 spike and were not 183 identified in the in-silico approach by Grifoni et al. 184 Surprisingly, the S2 domain doesn't appear to elicit as many neutralizing antibodies as 185 RBD or S1. Although S2 contains the fusion peptide, it does not appear to be as immunogenic, 186 compared with S1 or RBD, in generating binding antibodies to the intact spike (S1+S2) 187 ectodomain, as observed in both IgG ELISA and SPR. Even though we characterized the purified 188 proteins in various assays, there is a possibility that the structure of the antigens used in the study 189 is different from the corresponding authentic spike protein on the surface of SARS-CoV-2 virion 190 particle.

303
Recombinant purified proteins used in the study were either produced in HEK 293 mammalian 304 cells (S1 and RBD) or insect cells (S1+S2 ectodomain and S2 domain).

307
Female New Zealand white rabbits (Charles River labs) were immunized twice intra-308 muscularly at 14-days interval with 50 g of purified proteins mixed with Emulsigen Adjuvant.

309
Sera were collected before (pre-vaccination) and after 1 st and 2 nd vaccination and analyzed for 310 binding antibodies in ELISA, SPR, neutralization assay and GFPDL analysis. 1hr, plates were washed as before and OPD was added for 10min. Absorbance was measured at 320 492 nm. End titer was determined as 2-fold above the average of the absorbance values of the 321 naïve serum samples. The end titer is reported as the last serum dilution that was above this cutoff.  ProteOn manager software (version 3.1). All SPR experiments were performed twice and the 339 researchers performing the assay were blinded to sample identity. In these optimized SPR 340 conditions, the variation for each sample in duplicate SPR runs was <5%. The maximum resonance 341 units (Max RU) data shown in the figures was the RU signal for the 10-fold diluted serum sample.

342
Antibody off-rate constants, which describe the fraction of antigen-antibody complexes that decay 343 per second, are determined directly from the serum/ sample interaction with SARS CoV-2 spike 344 ectodomain (S1+S2), S1, S2, and RBD using SPR in the dissociation phase only for the 345 sensorgrams with Max RU in the range of 20-100 RU and calculated using the BioRad ProteOn 346 manager software for the heterogeneous sample model as described before(11). Off-rate constants 347 were determined from two independent SPR runs.    The datasets generated during and/or analyzed during the current study are available from the 388 corresponding author on reasonable request.