An antibody from single human VH-rearranging mouse neutralizes all SARS-CoV-2 variants through BA.5 by inhibiting membrane fusion

Abstract

SARS-CoV-2 Omicron subvariants have generated a worldwide health crisis due to resistance to most approved SARS-CoV-2 neutralizing antibodies and evasion of vaccination-induced antibodies. To manage Omicron subvariants and prepare for new ones, additional means of isolating broad and potent humanized SARS-CoV-2 neutralizing antibodies are desirable. Here, we describe a mouse model in which the primary B cell receptor (BCR) repertoire is generated solely through V(D)J recombination of a human V_H1-2 heavy chain (HC) and, substantially, a human Vκ1-33 light chain (LC). Thus, primary humanized BCR repertoire diversity in these mice derives from immensely diverse HC and LC antigen-contact CDR3 sequences generated by nontemplated junctional modifications during V(D)J recombination. Immunizing this mouse model with SARS-CoV-2 (Wuhan-Hu-1) spike protein immunogens elicited several V_H1-2/Vκ1-33-based neutralizing antibodies that bound RBD in a different mode from each other and from those of many prior patient-derived V_H1-2-based neutralizing antibodies. Of these, SP1-77 potently and broadly neutralized all SARS-CoV-2 variants through BA.5. Cryo-EM studies revealed that SP1-77 bound RBD away from the receptor-binding motif via a CDR3-dominated recognition mode. Lattice light-sheet microscopy-based studies showed that SP1-77 did not block ACE2-mediated viral attachment or endocytosis but rather blocked viral-host membrane fusion. The broad and potent SP1-77 neutralization activity and nontraditional mechanism of action suggest that it might have therapeutic potential. Likewise, the SP1-77 binding epitope may inform vaccine strategies. Last, the type of humanized mouse models that we have described may contribute to identifying therapeutic antibodies against future SARS-CoV-2 variants and other pathogens.

Fig. 1.. A humanized mouse model with diverse BCR repertoire derived from single human V_H and Vκ recombination.

(A) Schematic representation of modified Igh and Igκ loci of V_H1-2/Vκ1-33-rearranging mice. The V_H1-2^IGCR1∆ allele was made previously (33). We deleted 2 MB region upstream of V_H1-2 that contained all mouse V_Hs to generate V_H1-2^{mVH∆/IGCR1∆} allele. We replaced mouse Vκ3-2 with human Vκ1-33 and deleted Cer/sis from the Vκ to Jκ interval to generate Vκ1-33^CS∆ allele. (B) HTGTS-rep-seq analysis of Vκ usage in 129/Sv wild-type (left panel) and Vκ1-33-rearranging mouse splenic B cells in presence (middle panel) or absence (right panel) of Cer/sis. The x axis listed all functional Vκs from distal to the Jκ-proximal ends. The histogram displayed percent usage of each Vκ among all productive VκJκ rearrangements. The junction number of each Vκ was shown in Table S1. (C-D) Length distribution of V_H1-2 HC CDR3 (left panel in (C)) and Vκ1-33 LC CDR3 (left panel in (D)) in splenic B cells. Data were mean ± SD of three libraries from different mice and shown in Table S2. Venn diagram showed the V_H1-2 HC CDR3 (right panel in (C)) and Vκ1-33 LC CDR3 (left panel in (D)) complexity. Unique reads derived from the same libraries were in the left panel. Little overlap of V_H1-2 HC CDR3 sequences among three independent mice indicated tremendous CDR3 complexity.

Fig. 2.. Immunizing the V_H1-2/Vκ1-33-rearranging mouse model with SARS-CoV-2 spike or RBD elicited multiple V_H1-2/Vκ1-33 antibodies.

(A) Immunization scheme. Prime plus boost immunizations were performed at an interval of four weeks. (B) Binding curves showing reactivity of sera to SARS-CoV-2 spike, RBD, NTD and SARS-CoV-1 spike protein. The upper panel showed the sera from the SARS-CoV-2 Spike immunized mice at week 0, 2 and 6. The bottom panel showed the sera from VHH7-RBD immunized mice. Data were mean ± SD of three mice. (C) Table showed the V_H1-2/Vκ1-33 antibodies isolated from SARS-CoV-2 spike-specific or RBD-specific IgG⁺ B cells. The antibody sequences and sequence features were shown in Table S3.

Fig. 3.. SP1-77 potently neutralized SARS-CoV-2 VOCs, including Omicron sub-variants.

(A) Table showed the neutralization activities (IC₅₀: top; IC₈₀: bottom) of three monoclonal antibodies against all variants of concern (VOCs) and some variants of interest (VOIs) in pseudovirus neutralization assays. Experiments were done in 293 T/ACE2 cells. The neutralization curves were shown in Fig. S4A. The mutations on the spike proteins of different variants were listed in Table S4. Data were representative of 2 biologically independent experiments for most VOCs and VOIs, but one experiment for BA.3. Each independent experiment contained 2 technical replicates. IC₅₀ and IC₈₀ values were color-coded based on the key shown at the right. (B) Table showed the neutralization activities of three antibodies against VOCs in PRNT live virus neutralization assays. The neutralization curves were shown in fig. S4C. Data were representative of one independent experiment with 2 technical replicates. IC₅₀ and IC₉₀ values were color-coded based on the key shown at the right. (C) Final 3D reconstruction of Fab-S complexes shown in top view and side view with the S in gray and the Fabs colored (SP1-77: green; VHH7-5-82: blue; VHH7-7-53: magenta).

Fig. 4.. cryo-EM structures of SP1-77 Fab in complex of a full-length SASR-CoV-2 S trimer.

(A) Cryo-EM structures of SP1-77 Fabs in complex with full-length S trimer in the one-RBD-up (2.9 Å) and three-RBD-down (2.7 Å) conformations. EM density was in gray and structures were in ribbon diagrams with RBD in blue, NTD in yellow and the rest in dark gray. SP1-77 HC was in green and LC was in cyan. Three SP1-77 Fabs bound one S trimer in both conformations. (B) Close-up view of interactions between SP1-77 Fab and the RBD and NTD of the S trimer in RBD-down conformation. Left, HC CDR3 of SP1-77 made main contact with RBD away from RBM in light blue, while HC CDR2 touched the N-linked glycan from the NTD. Right, zoom-in views of binding interface, showing 17-residue HC CDR3 wedging into a groove formed by two segments of residues 339-346 and residues 440-445. The Asn343 glycan of RBD also interacted with Tyr99 from HC CDR3. Two glycans at Asn122 and Asn165 in the NTD were in proximity to CDR2 (Asn53, Ser54, Gly56, Thr57 and Asn58) and HC FW3 (Thr73, Ser74 and Ile75). (C) Comparison of binding mode and footprint of SP1-77 with ACE2 and other neutralizing antibodies, including S309 and LY-CoV1404. Left, RBD was in surface representation in gray. ACE2 and the antibodies were in ribbon diagram with SP1-77 in green, S309 in orange, LY-CoV1404 in magenta, and ACE2 in pink. Right, footprints on RBD of various antibodies with SP1-77 outlined in green, S309 in orange and LY-CoV1404 in magenta. The interface residues of SP1-77 were indicated with major contacting residues in red.

Fig. 5.. Modeled SP1-77 binding site on various SARS-CoV-2 variants.

(A) The potential footprint of SP1-77 on the modeled RBDs from different SARS-CoV-2 variants in a top view. The RBD was shown in surface representation in gray with the SP1-77 footprint highlighted in green and the mutations in each variant in blue. The Fv region of SP1-77 was shown in ribbon diagram in green. Most mutations on spike variants were not located at the SP1-77 footprint. (B) Side view of a selected panel from A. (C) Structural comparison of the SP1-77 binding interface among the RBDs of wildtype G614, Mu and Omicron variants. The conservative mutation Arg346Lys in Mu preserved the salt bridge between the residue 346 in the RBD and the SP1-77 Asp95. The mutations in Omicron variant reconfigured the local conformation near the N343-glycan, which is on the edge of SP1-77 footprint. The RBD was colored in gray, the SP1-77 heavy chain was colored in green and the light chain was colored in light green. Mutations in the Mu and Omicron variants were shown in stick and ball model in blue.

Fig. 6.. LLSM single virus tracking revealed SP1-77 inhibited S-fragment shedding and membrane fusion.

(A-I) Trajectories of VSV-SARS-CoV-2-Atto 565 virus imaged every 4 seconds with volumetric LLSM beginning 5 minutes after inoculation of Vero TMPRSS2 cells with virus at MOI ~ 2 without (A-D) or with 500 ng/mL SP1-77 treatment (E-H). Trajectories were colored magenta when virus was localized to cell surface, light blue when virus internalized, and dark blue if the Atto 565 signal spreaded from a point spread function (PSF) to a larger distribution indicating viral envelope fusion with endosomal membrane. Single virus trajectories (B, F) with insets showing x-axis projection through 4 planes (top), 3D integrated intensity profiles (C, G, bottom) and corresponding heat maps of fluorescence intensities from 4 plane z- projections (C, G, top). Intensity line-profiles through center of virions (D, H) appeared as a single point spread function at the cell surface and in endosomes in the absence (D, t1-t4) or presence (H, t1-t6) of SP1-77. Atto 565 spreaded upon fusion in the absence of SP1-77 (D, t5-t6). (I) Summary of all single virus trajectories over the course of single experiments, in the absence and presence of SP1-77, each from 3 cells imaged consecutively for 10 min. Red dots indicated virus trajectories in which a TMPRSS2-dependent Atto 565 signal decrease was observed. (J) Quantification of 5 experiments for each condition for the number of viruses in the total cell volume of every cell imaged (left), the number of Atto 565 decreases observed in all the trajectories of virus when co-localizing to the cell surface (middle), and the number of instances of Atto 565 dye spreading within endosomes (right). Data were mean ± SD of 5 independent experiments. p values were accessed by unpaired two-tail t-test. Data were shown in Table S6.

Fig. 7.. SP1-77 inhibited ACE2-mediated S1 dissociation on trypsin-treated VSV-SARS-CoV-2 Atto 565 viruses.

(A) Histogram of the number of Atto 565 molecules on VSV-SARS-CoV-2-Atto 565 determined by single molecule counting. (B) Histogram of averages of peak distribution determined by a Gaussian fit of 3 independent experiments. p values were accessed by unpaired two-tail t-test. Data were shown in Table S7. Treatments prior to adsorption included none (control), incubation with 1 μg/mL trypsin for 30 min at 37°C, treatment without or with trypsin followed by an incubation with 0.5 μM of recombinant ACE2 for 10 min at 37°C, or treatment with trypsin then incubated with 100 ng/mL SP1-77 IgG or SP1-77 Fab for 1 hour at 37°C followed by incubation with ACE2. (C) Schematic representation of proposed mechanism of SP1-77 inhibition of SARS-CoV-2 infection. Left panel: Without antibody treatments, the spike protein on the viral surface bound to the ACE2 receptor on the infected cell surface. Membrane fusion was activated either by TMPRSS2 protease on the cell membrane or by cathepsin L protease following endocytosis. Cleavage at the S2’ site by these proteases led to dissociation of the S1 subunit, which exposed the fusion peptide on the S2, facilitating viral-host membrane fusion and viral entry into the infected cells. Middle panel: Pre-treatment of the virus with SP1-77, a non-ACE2-blocking antibody, did not appreciably impact binding of viruses to the cell surface and their endocytosis. However, SP1-77 greatly inhibited the dissociation of S1 subunit, thereby, blocking activation of the fusion peptide and membrane fusion. Right panel: Pre-treatment of the virus with VHH7-5-82, an ACE2-blocking antibody, prevented binding of the virus to the cell surface.