Structure-selected RBM immunogens prime polyclonal memory responses that neutralize SARS-CoV-2 variants of concern

Successful control of the COVID-19 pandemic depends on vaccines that prevent transmission. The full-length Spike protein is highly immunogenic but the majority of antibodies do not target the virus: ACE2 interface. In an effort to concentrate the antibody response to the receptor-binding motif (RBM) we generated a series of conformationally-constrained immunogens by inserting solvent-exposed RBM amino acid residues into hypervariable loops of an immunoglobulin molecule. Priming C57BL/6 mice with plasmid (p)DNA encoding these constructs yielded a rapid memory response to booster immunization with recombinant Spike protein. Immune sera antibodies bound strongly to the purified receptor-binding domain (RBD) and Spike proteins. pDNA primed for a consistent response with antibodies efficient at neutralizing authentic WA1 virus and two variants of concern (VOC), B.1.351 and B.1.617.2. These findings demonstrate that immunogens built on structure selection can focus the response to conserved sites of vulnerability shared between wildtype virus and VOCs and induce neutralizing antibodies across variants.


Introduction 57
The current SARS-CoV-2 pandemic has been globally disruptive (Nicola et al, 2020). 58 Nonpharmaceutical interventions (NPI) have reduced viral transmission but are difficult to 59 (Benton et al, 2020). A primary antibody response was evident in group 1 and 2 suggesting 156 a more favorable surface exposure and spatial orientation of the FNCY patch and 157 surrounding residues in the immunogenic proteins coded by pDNA1 and pDNA2. 158 The antibody response post-booster immunization was analyzed at an early (day 30) 159 and late (day 45) time point (i.e., 9 and 24 days after booster), to capture the evolution of 160 the memory response. Nine days post-boost antibodies to RBD were markedly increased 161 relative to day 21 both in group 1 and 2. The response was also greater than in group 3 and 162 4. Since group 4 controls for magnitude and speed of the antibody response after booster 163 with intact Spike protein, the data show that pDNA priming accelerated the recall antibody 164 response by the Spike protein. Not surprisingly, sera from group 1 and 2 also had a 165 markedly stronger response to the Spike protein compared to group 3 and 4 (Fig. 2). 166 Twenty-four days post-boost the responses against the RBD and Spike proteins were 167 considerably stronger than on day 30 and substantially similar among groups (Fig. 2). 168 Collectively, the three pDNA immunogens differed in their ability to prime a specific B cell 169 response and generate an anamnestic response in a prime-boost regimen. A greater 170 response against the RBM B cell epitope was associated with expression of the RBM epitope 171 in the CDR2 loop of the VH, suggesting context-dependent immunogenicity perhaps owing 172 to better folding and more favorable recognition by B cells relative to expression in CDR3. 173 The B cell epitope selected for these studies is contained in a narrow region of the 174 RBM. To monitor the reactivity of serum antibodies with greater precision we synthesized 175 peptide 475 AGSTPCNGVEGFNCYFPLQSYGFQPT 500 . Reactivity against the RBM peptide 176 occurred after booster immunization with binding profiles on day 9 post-booster mimicking 177 those on RBD and Spike proteins in group 1 and 2, which displayed overall stronger binding 178 relative to group 3 and 4 (Fig. S2A). Intra-spleen pDNA priming induces predominantly IgM 179 antibodies (Gerloni et al, 1998), which may account for a weak binding to the RBM peptide 180 at 1:50 dilution, suggesting low antibody concentration in sera, low avidity, or both. 181 Antibody binding increased by day 45 (Fig. S2A). Thus, priming with pDNA1 and pDNA2 182 generated a pool of memory B cells specific for an RBM epitope comprised within residues 183 475 AGSTPCNGVEGFNCYFPLQSYGFQPT 500 that were expanded during the anamnestic 184 response. 185 Immune sera antibodies cross-compete RBD binding of neutralizing human 186 monoclonal antibodies 187 The RBM amino acids grafted into CDR loops comprise contact residues shown to be targets 188 of potently neutralizing antibodies derived from COVID-19 patients. Among those are 189 antibodies B38  and CC12.1 (Rogers et al., 2020), which map an 190 overlapping RBM epitope including contact residues Phe -Asn and Tyr of the 486 FNCY 489 191 patch in both cases ( Fig 3A). 192 We reasoned that the paratope of these two antibodies could be used to determine 193 shared epitope specificity between immune serum polyclonal antibodies and human 194 neutralizing antibodies. To this end, we designed a competitive ELISA assay where the 195 binding of horseradish peroxidase (HRP)-labeled B38 and CC12.1 to the RBD protein was 196 competed by individual immune sera. HRP-labeled B38 and CC12.1 bound RBD with similar 197 characteristics ( Fig. 3B and D) so that both antibodies could be used at the same (~1.5 198 ng/well) effective ~50% binding concentration. Unlabeled antibodies B38 and CC12.1 199 inhibited homologous RBD binding at similar concentration (50% inhibition at ~ 60 200 ng/well) ( Fig. 3B and D, inset). 201 The majority of immune sera tested individually at a 1:50 dilution inhibited >50% 202 the binding to RBD by both HRP-B38 and HRP-CC12.1. Inhibition in group 2 was overall the 203 strongest (range 25-85% for B38; and 30-84% for CC12.1) (Fig 3C and E). Surprisingly, 204 group 3 immune sera also inhibited (range 62-80% on B38; and 38-52% on CC12.1). Group 205 4 had overall the weakest inhibitory activity (<30%) with the exception of one mouse. The 206 fact that serum antibodies from immune mice cross-competed the RBD binding of the two 207 human neutralizing antibodies indicates that a component of serum antibodies in the 208 immune serum share the paratope with antibodies B38 and CC12.1. We estimated that the 209 upper limit serum concentration of such antibodies as high as ~12 µg/ml. Collectively, the 210 results show that intra-spleen immunization with pDNA coding for the selected RBM 211 epitope activates, and subsequently facilitates, the expansion of B cell clonotypes producing 212 RBM antibodies found in COVID-19 patients. Oddly, neither HRP-B38 nor HRP-CC12.1 213 bound to 475 AGSTPCNGVEGFNCYFPLQSYGFQPT 500 peptide in ELISA ( Fig. S2B and C), 214 suggesting perhaps that the synthetic peptide lacks the conformation/structure the 215 paratope of these antibodies needs for binding. Thus, immune sera have a wider spectrum 216 of paratopes than those defined by the two human monoclonal antibodies.  In conclusion, we demonstrate that immune mice sera neutralize not only authentic WA1 255 virus but also two VOCs responsible for rapid spreading of infection and disease. 256

Structure-function considerations on B cell epitope expression 257
We used computer modeling techniques to model the RBM B cell epitope CNGVEGFNCYFP particularly by group 2 sera, was not. Compellingly, this shows that few critical solvent-306 exposed amino acids in the RBM can expand B cells with a paratope specific for the RBM. 307 One may argue that neutralization of the two VOCs by polyclonal antibodies in immune sera 308 is owed to recognition of a discrete epitope within the RBM that is conserved among the 309 study. Collectively, this shows that when this B cell epitope is conformationally-constrained 315 in the CDR2 is not only immunogenic but also captures a spectrum of paratopes efficient at 316 neutralizing across the WA1 virus, and the B.1.351 and B.1.617.2 variants of concern. This 317 stresses the importance of defining immunogenic sites at the structure/function level for 318 efficient precision immunization. 319 Our study has limitations. First, the sample size was deliberately low since the study 320 was designed to identify (a) pDNA with ability to prime RBM-specific B cells that could be 321 re-expanded during a recall response, and (b) yield consistent virus neutralization. In spite 322 of this limitation we managed to identify the immunogen yielding the overall more 323 consistent response. The second is that the dose of Spike protein used in the booster 324 immunization was probably excessive as this protein is highly immunogenic. This may have 325 masked the true potential of the memory response. Future experiments will need to assess 326 memory responses against booster immunization with lower doses of antigen. Finally, 327 although the study was successful in identifying an optimal prime-boost combination The experimental design is illustrated in Fig. 1. The overall objective of the study was to 381 identify an immunogen able to prime a polyclonal response expandable as a memory 382 response upon boost immunization while enriching for antibodies recognizing a specific 383 site in the RBM ridgeline. 384

417
Engineering methods were a modification of those described in (Xiong et al, 1997). Briefly, 418 the three VH genes matching the descriptions of Model 1-3 as shown in Fig. 1 were 419 synthesized with unique ends for cloning into the ZUC1.1 (9.7 Kb) target vector using the 420 cloning site as shown in Fig. S4. The ZUC1.1 plasmid lacks Amp resistance gene and SV40 421 sequences, and is optimized for human use. The amino acid sequences of the Variable 422 domain of Model 1-3 are shown in Fig. S1. pDNAs were prepared from transformed DH5a 423 Escherichia coli according to standard procedures and were analyzed for purity using the 424 following equation: %N=(11.1R-6.32)/(2.16-R) where R=260nm/ 280nm, %N=% of nucleic 425 acid 42 . pDNAs were resuspended in sterile saline solution and stored at -20 °C until use. 426

Mice and immunizations 427
Twelve-week-old female C57Bl/6 (H-2 b ) were bred at the University of California, San Diego 428 animal facility where they were kept throughout the performance of the experiment. RBD protein (4 µg/ml). Binding was revealed as described above. 454

SARS-CoV-2 viruses 455
All work with SARS-CoV-2 was conducted in Biosafety Level-3 conditions at the University 456 of California San Diego following the guidelines approved by the Institutional Biosafety 457 Committee. SARS-CoV-2 isolates WA1(USA-WA1/2020, NR-52281) and B. Computational models utilized the coordinates of rearranged murine VH62 (Sollazzo et al, 499 1989). Model predictions have made using SWISS-MODEL, an automated protein structure 500 homology-modelling server and the coordinates of the crystal structure of an antibody: 501 antigen complex (SMTL ID: 5mhe.1) where the antibody is specific for L-Thyroxine, which is 502 also the epitope specificity of monoclonal antibody 62 from which VH62 was originally 503 cloned (Sollazzo et al., 1989).