key: cord-0775199-49nhfymh
authors: Wang, Kang; Jia, Zijing; Bao, Linlin; Wang, Lei; Cao, Lei; Chi, Hang; Hu, Yaling; Li, Qianqian; Jiang, Yinan; Zhu, Qianhui; Deng, Yongqiang; Liu, Pan; Wang, Nan; Wang, Lin; Liu, Min; Li, Yurong; Zhu, Boling; Fan, Kaiyue; Fu, Wangjun; Yang, Peng; Pei, Xinran; Cui, Zhen; Qin, Lili; Ge, Pingju; Wu, Jiajing; Liu, Shuo; Chen, Yiding; Huang, Weijin; Qin, Cheng-Feng; Wang, Youchun; Qin, Chuan; Wang, Xiangxi
title: A subset of Memory B-derived antibody repertoire from 3-dose vaccinees is ultrapotent against diverse and highly transmissible SARS-CoV-2 variants, including Omicron
date: 2022-01-03
journal: bioRxiv
DOI: 10.1101/2021.12.24.474084
sha: 54d6e4cd01fbeaa2b094f6287d1d3f3c29c47b33
doc_id: 775199
cord_uid: 49nhfymh

Omicron, the most heavily mutated SARS-CoV-2 variant so far, is highly resistant to neutralizing antibodies, raising unprecedented concerns about the effectiveness of antibody therapies and vaccines. We examined whether sera from individuals who received two or three doses of inactivated vaccine, could neutralize authentic Omicron. The seroconversion rates of neutralizing antibodies were 3.3% (2/60) and 95% (57/60) for 2- and 3-dose vaccinees, respectively. For three-dose recipients, the geometric mean neutralization antibody titer (GMT) of Omicron was 15, 16.5-fold lower than that of the ancestral virus (254). We isolated 323 human monoclonal antibodies derived from memory B cells in 3-dose vaccinees, half of which recognize the receptor binding domain (RBD) and show that a subset of them (24/163) neutralize all SARS-CoV-2 variants of concern (VOCs), including Omicron, potently. Therapeutic treatments with representative broadly neutralizing mAbs individually or antibody cocktails were highly protective against SARS-CoV-2 Beta infection in mice. Atomic structures of the Omicron S in complex with three types of all five VOC-reactive antibodies defined the binding and neutralizing determinants and revealed a key antibody escape site, G446S, that confers greater resistance to one major class of antibodies bound at the right shoulder of RBD through altering local conformation at the binding interface. Our results rationalize the use of 3-dose immunization regimens and suggest that the fundamental epitopes revealed by these broadly ultrapotent antibodies are a rational target for a universal sarbecovirus vaccine. One sentence summary A sub-set of antibodies derived from memory B cells of volunteers vaccinated with 3 doses of an inactivated SARS-CoV-2 vaccine work individually as well as synergistically to keep variants, including Omicron, at bay.

more concerning about effectiveness for two-dose regime against Omicron 115 infection. Among two doses of CoronaVac recipients, NT50 titer against Delta was 116 6.3 with a 5-fold reduction when compared to WT, but none of the serum 117 specimens had an NT50 titer of >8 against Omicron (Fig. 1c ). Compared to 2-dose 118 vaccinees, sera of the 3-dose vaccinees displayed lower reduction in neutralization 119 titers against Delta, which is consistent with previous observations that 3-dose 120 administration of inactivated vaccine leads to enhanced neutralizing breadth to 121 SARS-CoV-2 variants 5 .

Three doses of vaccine-elicited monoclonal antibodies 124 We previously sorted immunoglobulin (IgG+) memory B cells from peripheral blood mononuclear cells (PBMCs) of four 3-dose CoronaVac vaccinees using 126 prefusion SARS-CoV-2 S as a bait 5, 15 . In total, we sorted 1800 SARS-CoV-2 S-127 specific memory B cells, obtained 422 paired heavy-and light-chain antibody 128 sequences, and selected 323 antibodies for expression. Characterization by ELISA 129 showed that 163, 100 and 51 recognized the RBD, NTD and S2, respectively and 9 130 failed to bind S (Fig. 2a) . Biolayer interferometry affinities (BLI) measurements 131 showed that nearly all RBD-directed antibodies bound to WT SARS-CoV-2 at sub-132 nM levels and 127 of them showed neutralization activities against both authentic 133 and pseudotyped WT SARS-CoV-2 were selected for further investigation. Of these 134 antibodies, over 93% of these antibodies exhibited broad binding activities to most 135 VOCs and VOIs. Notably, 85% of these antibodies cross-reacted with the Omicron 136 RBD. Contrarily, ~80% of NTD antibodies lost their associations with Omicron. 137 Additionally, NTD antibodies also showed relatively poor cross-reactivity to other 138 four VOCs due to the greater diversity of the NTD (Fig. 1, a, b) . respectively. These neutralizations are as potent as those exhibited by best-in-class 150 antibodies against WT (Fig. 2b and 2d) . We obtained IC50 values of 0.24 and 0.28 151 ng/μl for well-studied therapeutic antibodies like S309 and DXP-604, respectively.

These values are 10~40-fold higher than those of the subset antibodies. 153 Concerningly, some antibody drugs, such as REGN10933, REGN10987, CoV555, LY-CoV016, AZD1061 and AZD8895, almost lost their neutralization 155 activities against Omicron (Fig. 2b) 18 . Meanwhile, specific VOC-resistant antibodies with high neutralizing potency against WT and some other VOCs (IC50 157 <0.2 ng/μl) were identified and these comprise ~30% of the antibody repertoire, 158 indicative of the evolution of a wide range of antibodies after 3-dose vaccination. 159 Experiments repeated using authentic virus, including WT and all five VOCs, 160 showed similar neutralization patterns by all these antibodies, further verifying the 161 neutralizing potency and breadth for this subset of antibody repertoire elicited by 3-162 dose vaccination (Fig. 2e ). with >500-fold decreased neutralization against Omicron, but potent neutralization 180 against other four VOCs were selected for structural investigations (Fig. 2b) . We 181 determined cryo-EM reconstructions of these complexes at 3.2 -3.6 Å, and 182 performed local refinement to further improve the densities around the binding 183 interface between RBD and antibodies, enabling reliable analysis of the interaction 184 details (Fig. 3, Extended Data Fig. 3, 4 and 5, Extended Data Fabs bound to S trimer with one down and one up RBD, although the XGv265-bound 193 up RBD conformation was weakly resolved and therefore not modeled (Fig. 3a). 194 Antibody XGv347 binds to an epitope at the tip of RBD, largely overlapping with the 195 patch targeted by ACE2 (Fig. 2d, 3b, Extended Data Fig. 1 the RBD-down conformation in the WT S, but become accessible for either up or 204 down RBDs in the Omicron S due to a markedly outward expansion and a ~10º 205 clockwise rotation of three RBDs, leading to an approximately 9 Å conformational 206 movement for RBM (Fig. 3b and Extended Data Fig. 6 ). The XGv347 paratope 207 constituted five complementarity determining regions (CDRs) with heavy chain and 208 light chain contributing 70% and 30% of the binding surface area, respectively (Fig. 209 3b and Extended Data Table. 2). Overall XGv289, XGv282 and XGv265 bind 210 patches surrounding the right shoulder of RBD with various orientations, but in a 211 manner similar to those observed for DH1047, BD-812 and REGN10987; antibodies 212 known to generally neutralize most VOCs with high potency 24-26 , but showing 213 declined, to varying degrees, binding and neutralizing activities against Omicron due 214 to the presence of new N440K and G446S mutations (Fig. 2b, Extended Data Fig. 7 215 and Extended Data Table. 3). Notably, XGv265 and REGN10987 recognize almost 216 same epitopes, both nearly losing their neutralizing activities against Omicron, despite 217 retaining weak binding (Extended Data Fig. 7) . Structural superimpositions reveal that XGv347 and either XGv289 or XGv265 can simultaneously bind to S, informing 219 strategies to rationally design two-antibody cocktails (Extended Data Fig. 8 ). ~10,000-fold in the lungs compared to the control group (Fig. 4b ). Remarkably, a single dose of XGv289, XGv265, XGv347, XGv052 or antibody cocktails of XGv282 281 + XGv347, XGv052 + XGv289 resulted in a complete clearance of viral particles in 282 the lungs ( Fig. 4b and 4c ). A potential synergistic effect was observed for combined 283 therapies of XGv282 + XGv347 at 2.5 mg/kg for each ( Fig. 4b and 4c ). In addition, The ongoing pandemic has witnessed frequent occurrences of SARS-CoV-2 variants 296 that increase transmissibility and reduce potency of vaccine-induced and therapeutic 297 antibodies 2,11 . More recently, there has been unprecedented concern that the Omicron 298 variant has significantly increased antibody escape breadth due to newly occurred and 299 accumulated mutations in the key epitopes of most neutralizing antibodies.

Alarmingly, Omicron nearly ablates the neutralization activity of most FDA approved 301 antibody drugs, including LY-CoV555, LY-CoV016, REGN10933, REGN10987, 302 AZD8895 and AZD1061 18 . These issues raise an urgent need to develop next-303 generation antibody-based therapeutics that can broadly neutralize these variants, as 304 well as future variants of concern. Our previous study revealed that the regimen of 3-305 dose vaccination (0, 1, 7 months) of inactivated vaccine leads to an improved 306 immunity response with significantly enhanced neutralizing breadth via ongoing 307 antibody somatic mutation and memory B cell clonal turnover 5,28 . Correlated with 308 this, one subset of highly potent neutralizing antibodies with broad activities (IC50 < The selected 323 antibodies were subjected to gene codon optimization, 491 construction and expression as described previously 5 . Then the clones were 492 transiently transfected into mammalian HEK293F cells and incubated for 5 days in 493 a 5% CO2 rotating incubator at 37°C for antibody expression, which were further 494 purified using protein A. The purified mAbs XGv265, XGv282, XGv289 and 495 XGv347 were then processed to obtain their Fab fragments using the Pierce FAB 496 preparation kit (Thermo Scientific) as described previously 30 . Briefly, the samples 497 were first applied to desalination columns to remove the salt and the flow-throughs 498 were collected and incubated with papain that was attached with beads to cleave Fab 499 fragments from the whole antibodies for 5 hours at 37°C. After that, the mixtures 500 were transferred into Protein A columns and the flow-throughs, i.e., the Fab fragments 501 were collected and dialyzed into Phosphate Buffered Saline (PBS) (ThermoFisher, 502 catalog #10010023). ResMap. All dataset processing is shown in Extended Data Fig. 3 and also 574 summarized in Extended Data Table 1 .

Model fitting and refinement 577 The atomic models of the complexes were generated by first fitting the chains of the 578 native apo SARS-CoV-2 S trimer (PDB number of 6VYB) and Fabs (PDB number of 579 7LSS and 7CZW for XGv265, 5MES and 5VAG for XGv282, 6UDA and 7MEG for 580 XGv289 as well as 7E3K for XGv347) into the cyo-EM densities of the final S-Fab- Briefly, we used OPLS force field with TIP3P water model to prepare the dynamic 594 system and add Na+ and Clions to make the system electrically neutralized. Then, 595 the system was subjected to energy minimization using the steepest descent algorithm until the maximum force of 1,000 kJ mol-1 has been achieved. NVT ensemb1e via the 597 Nose-Hoover method at 300 K and NPT ensemble at 1 bar with the Parinello-Rahman 598 algorithm were employed successively to make the temperature and the pressure 599 equilibrated, respectively. Finally, a MD production runs of 100 ns were performed 600 starting from random initial velocities and applying periodic boundary conditions.

The non-bonded interactions were treated using Verlet cut-off scheme, while the long-602 range electrostatic interactions were treated using particle mesh Ewald (PME) 0.020 ng/μL; green: 0.020-1.000 ng/μL; red: 1.000-10.000 ng/μL). Data marked with 716 '*' represents the data referred from the available publication in which the whole 717 experiment system and condition of pseudovirus neutralization assay remains the samewith this study 18 ; Data marked with '-' means no related datasets here. c, 719 Heatmap with values shown in the form of AUC represents the competition ability 720 between the selected mAbs and hACE2. Color gradient ranging from white (1) Defocus range (μm) -1.5--2.5 -1.5--2.5 -1.5--2.5 -1.5--2.5 -1.5--2.5 -1.5--2.5 -1.5--2.5 -1.5--2.5 -1.5--2.5

Resolution ( 

4| Protection against SARS-CoV-2 Beta variant strain challenge in mice

Groups of BALB/c mice were infected intranasally with SARS-753

CoV-2 Beta variant strain, followed by a single dose of an antibody, or an antibody 754 cocktail, or PBS as control one hour after infection as indicated in a

Examination of lung tissues of mice collected at 5 dpi for b, virus titer, c

Virus RNA loads in the lungs at 5 dpi were measured 757 by RT-qPCR and are expressed as RNA copies per gram. Data are represented as 758 mean ± SD. Dashed line represents limit of detection. c, SARS-CoV-2 genome RNA 759 ISH was performed with a SARS-CoV-2 specific probe. Brown-colored staining 760 indicates positive results

Extended Data Fig. 2 | Data Sheets of ELISA assay of representative Mabs 776 against Omicron RBD. Different Classes of Nabs (Class I-VI) are colored by yellow, 777 green, red, blue, brown and magenta, respectively. Values are filled with black (>75)

Extended Data Fig. 7 | Binding modes of XGv289, 282 and 265. Binding modes of 810

XGv282 and XGv 265. RBD is colored in light cyan and color scheme of 811

XGv282 and XGv265 is the same as in Fig. 3a. DH1047, BD-812 and 812 REGN10987 are colored in orange