Nanobodies from camelid mice and llamas neutralize SARS-CoV-2 variants

Abstract

Since the start of the COVID-19 pandemic, SARS-CoV-2 has caused millions of deaths worldwide. Although a number of vaccines have been deployed, the continual evolution of the receptor-binding domain (RBD) of the virus has challenged their efficacy. In particular, the emerging variants B.1.1.7, B.1.351 and P.1 (first detected in the UK, South Africa and Brazil, respectively) have compromised the efficacy of sera from patients who have recovered from COVID-19 and immunotherapies that have received emergency use authorization^1-3. One potential alternative to avert viral escape is the use of camelid VHHs (variable heavy chain domains of heavy chain antibody (also known as nanobodies)), which can recognize epitopes that are often inaccessible to conventional antibodies⁴. Here, we isolate anti-RBD nanobodies from llamas and from mice that we engineered to produce VHHs cloned from alpacas, dromedaries and Bactrian camels. We identified two groups of highly neutralizing nanobodies. Group 1 circumvents antigenic drift by recognizing an RBD region that is highly conserved in coronaviruses but rarely targeted by human antibodies. Group 2 is almost exclusively focused to the RBD-ACE2 interface and does not neutralize SARS-CoV-2 variants that carry E484K or N501Y substitutions. However, nanobodies in group 2 retain full neutralization activity against these variants when expressed as homotrimers, and-to our knowledge-rival the most potent antibodies against SARS-CoV-2 that have been produced to date. These findings suggest that multivalent nanobodies overcome SARS-CoV-2 mutations through two separate mechanisms: enhanced avidity for the ACE2-binding domain and recognition of conserved epitopes that are largely inaccessible to human antibodies. Therefore, although new SARS-CoV-2 mutants will continue to emerge, nanobodies represent promising tools to prevent COVID-19 mortality when vaccines are compromised.

Fig. 1. Production of nanomice.

a, Thirty VHHs selected from alpaca, dromedary and Bactrian camel were inserted via CRISPR–Cas9 in lieu of the 2.5-Mb mouse VH locus. CH1 exons from Cμ and Cγ1 were also deleted to avoid misfolding of the antibody heavy chain. b, Flow cytometry analysis of splenic B220⁺ B cells from wild-type (WT) mice or heterozygous nanomice. IgM⁺Igκ⁺ cells express conventional heavy–light chain antibodies, whereas IgM⁺Igκ⁻ cells are mostly Igλ⁺ in wild-type mice (not shown) or single-chain-antibody B cells in nanomice. c, Flow cytometry analysis of splenic cells from unimmunized and immunized nanomice and controls stained with CD95 and IgG1. d, Pie charts showing VHH somatic hypermutation in unimmunized and immunized nanomice. Pie segments are proportional to the VHH sequences carrying the mutations indicated on the periphery of the chart. The middle circle shows the total number of sequences, and mutation frequency is given below.

Fig. 2. Isolation of nanobodies against SARS-CoV-2.

a, Immunization of llama and nanomice to obtain high-affinity nanobodies against SARS-CoV-2 RBD. b, Biolayer interferometry (BLI) analysis of difference concentrations of Nb17 monomer (left) and trimer (right) binding to immobilized RBD. Red trace represents the raw data; the kinetic fit is shown in grey underneath. Equilibrium (K_D) constants are provided. c, Table summarizing pseudovirus neutralization potency (IC₅₀) of selected nanobodies. Values are provided in molarity (left) or as ng ml⁻¹ (right). d, Diagrams showing nanobodies used in neutralization assays as monomers, bivalent or trimers (the last two fused to human IgG1 Fc via the human or llama IgG2a hinge domain). e, Neutralization of SARS-CoV-2 pseudovirus by the 20 nanobodies shown in c. Nb12 monomer (red), bivalent (cyan) and trimer (magenta), as well as Nb19 trimer (blue), are highlighted. Data are representative of two independent experiments and the error bars are mean ± s.d. of triplicates.

Fig. 3. Neutralization of wild-type and mutant SARS-CoV-2.

a, Neutralization assays (IC₅₀ values) of pseudoviruses carrying wild-type or mutant SARS-CoV-2 spike. Colour gradient indicates values ranging from 0 (blue) to 50,000 pM (red). Pseudotyped viruses containing E484K or K417N, E484K and N501Y (KEN) also contain the R683G substitution. b, Neutralization assay showing the sensitivity of SARS-CoV-2 B.1.351 to different concentrations of trivalent Nb15, Nb56 and Nb12, and bivalent Nb30. Data are representative of two independent experiments and the error bars are mean ± s.d. of triplicates. c, Schematics summarizing the BLI competition assay, in which nanobody–RBD immunocomplexes attached to a biosensor are incubated with different nanobodies to measure binding. d–g, Binding of nanobodies to Nb12–RBD (d), Nb30–RBD (e), Nb15–RBD (f) and Nb56–RBD (g) immunocomplexes.

Fig. 4. Structures of leading nanobodies in complex with SARS-CoV-2 spike.

a, Cryo-electron microscopy structure of Nb12 in complex with SARS-CoV-2 spike. b, As in a, for Nb30. c, Interface between nanobodies and spike. d, Surface properties of RBD, including sequence diversity (dark purple indicates diversity among sarbecoviruses), and prevalence of RBD-recognized regions by human antibodies (dark raspberry indicates high prevalence) and binding site for ACE2 (cyan).

Extended Data Fig. 1. VHH genes used in the array and gene unit assembly.

Alignment of the 30 VHH genes, highlighting the 100% amino acid conservation (in red) and the 4 hydrophilic amino acids in framework (FR) 2 (in blue). In VH proteins, these latter four amino acids are hydrophobic and mediate the interaction with light chains. Schematics below show the configuration of VHH gene units, composed of a mouse VH promoter (250 bp, containing the octamer and TATA box); mouse leader exons–intron (about 150 bp) encoding the signal peptide cleaved off during heavy chain processing in the endoplasmic reticulum; the camelid VHH open reading frame (about 300 bp); and mouse downstream sequences (100 bp) containing the recombination signal sequences (RSSs).

Extended Data Fig. 2. Igλ expression and recombination frequency of VHH genes.

a, Flow cytometry analysis of Igλ expression in B220⁺IgM⁺ splenic B cells from wild type and heterozygous nanomice. b, VHH–DJ recombination was monitored by genomic PCR in bone marrow and spleen samples using an FR1 VHH-specific primer and a second primer downstream of JH4. The expected PCR products for each recombination event between a given VHH and JH1, JH2, JH3 or JH4 are provided. Data are representative of two independent experiments. c, Bar graph showing VHH percentage use among splenic B cells in three nanomouse littermates.

Extended Data Fig. 3. B cell development in nanomice.

a, Flow cytometry analysis of bone marrow (left two columns), spleen (third and fourth columns) and peritoneal cavity B cells (last column) in wild-type controls and nanomice. First column shows the percentage of B220-gated CD43⁺IgM⁻ proB and CD43^-IgM⁺ immature B cells. Second column shows percentage of IgM and Igκ within the B220-gated population. Third column denotes the total number of B220⁺ B cells in the spleen. The y axis shows viability staining with eFluor506 (eBiosciences). The fourth column shows the percentage of B220-gated CD23^lowCD21^low immature, CD23^highCD21^low follicular, and CD23^lowCD21^high marginal-zone splenic, cells in the two strains. The last column shows the percentage of B1 (IgM^highB220^low) and B2 (IgM^lowB220^high) cells in the peritoneal cavity. Examples of gating for bone marrow and splenic B cells is provided in Extended Data Fig. 10d. b, Histograms depicting the percentage of Igκ (left), IgD (middle) and IgG1 (right row) in wild-type, nanomice, and Ighm-CH1^−/− mice. The latter measured in ex vivo cultures treated with LPS, IL-4 and anti-CD180. Population gates are represented with a line and the percentage of total cells is provided. c, Proliferation assay of nanomouse and control B cells cultured for 96 h with LPS, IL-4 and anti-CD180. d, Immunization regimen. Nanomice were immunized with 50 μg HIV-1 envelope trimer at the indicated dates. e, Per cent nucleotide substitutions (adjusted for base composition) observed in nanobodies isolated from immunized nanomice. Phage library was selected for binding to HIV-1 envelope trimer.

Extended Data Fig. 4. Nanomouse immune response to HIV-1 envelope trimer.

a, Top table shows nanobody and JH use for the 16 VHH clones isolated from immunized nanomice. Middle graph shows protein alignment for VHHs isolated from HIV-1 trimer immunized nanomice. CDRs are boxed. Bottom table shows hypermutation profiles for VHH, D and J domains of selected nanobodies. b, Left, BLI analysis of BG505 DS-SOSIP binding to immobilized VHH9-1. Red trace represents the raw data, and the kinetic fit is shown in grey underneath. Right, table showing the kinetic constants for association (k_on), dissociation (k_off) and equilibrium (K_D) for all four VHH9 nanobody variants.

Extended Data Fig. 5. Isolation of anti-SARS-CoV-2 RBD nanobodies.

a, Table indicating (1) the total number of unique nanobody genes identified from llama and three nanomice phage display libraries following selection for RBD binding; (2) the number of nanobodies enriched at least tenfold after selection; (3) the number of nanobodies with a unique CRD3; and (4) the different clusters of nanobodies that share similar CDR3 s (with no more than 2 amino acid differences). b, Table showing in vitro neutralization results for the six leading nanobodies using the sVNT kit of GenScript. c, Dot plot depicting the extent of enrichment (y axis) and frequency (x axis) of unique nanobodies after RBD selection of llama (left) or nanomouse 1 (right) libraries. Green circles represent nanobodies that block ACE2–RBD interactions in vitro, black circles are nanobodies that do not efficiently block ACE2–RBD interactions, and grey dots represent untested nanobodies. d, Top graph shows protein alignment of the six nanobodies isolated from llama and nanomice immunized with SARS-CoV-2 spike and RBD. Bottom table shows detailed information of the VHH, D and J domains of nanomouse nanobodies. e, Table depicting equilibrium (K_D), association (K_on) and dissociation (K_off) constants obtained for each nanobody as a monomer (black) or trimer (red) form.

Extended Data Fig. 6. Neutralization of pseudo and authentic SARS-CoV-2 viruses.

a, Comparison of neutralization activities of leading nanobodies in monovalent, bivalent or trivalent form (results for monovalent and trivalent reproduced from Fig. 3a). b, Neutralization assays for wild-type (WA1) and SARS-CoV-2 variants B.1.1.7 and P.1 for trivalent Nb56, Nb15, Nb12 and bivalent Nb30. Data are representative of two independent experiments. Data are mean ± s.d. of triplicates. c, IC₅₀ neutralization values for experiments shown in Fig. 3b and b. Top table values are in pM, lower table values are in ng ml⁻¹. d, Related to Fig. 3c–f. Immunocomplexes used were Nb17–RBD (left) and Nb19–RBD (right). e, Coomassie staining showing nanobody integrity following nebulization. With the exception of Nb30 (bivalent), all nanobodies were fused to Fcs as trimers. Data are representative of two independent experiments. f, Bar graph showing in vitro neutralization (percentage) of RBD–ACE2 interactions by the different trivalent nanobody (bivalent for Nb30) before and after nebulization at two different concentrations (0.1 μg ml⁻¹ (blue) and 0.02 μg ml⁻¹ (orange)). g, Coomassie staining showing integrity of nanobody monomers (left) or multimers (right) following heat treatment (98 °C for 10 min). Data are representative of two independent experiments.

Extended Data Fig. 7. Cryo-EM data processing and validation for Nb12–spike complex.

a, A representative cryo-EM micrograph showing Nb12–spike complex embedded in vitreous ice. b, A contrast-transfer function (CTF) fit of the micrograph. c, Representative 2D average classes. d, Overall resolution estimation (Fourier shell correlation (FSC) of 0.143). e, Local resolution estimation of the cryo-EM map. f, Cryo-EM density and models for an interface region between RBD and Nb12 after local refinement.

Extended Data Fig. 8. Cryo-EM data processing and validation for Nb30–spike complex.

a, A representative cryo-EM micrograph showing Nb30–spike complex embedded in vitreous ice. b, A CTF fit of the micrograph. c, Representative 2D average classes. d, Overall resolution estimation (FSC of 0.143). e, Local resolution estimation of the cryo-EM map. f, Cryo-EM density and models for an interface region between RBD and Nb30 after local refinement.

Extended Data Fig. 9. Structural analysis of nanomice and llama nanobody interface with the SARS-CoV-2 spike.

a, Structure of Nb12 and RBD region (inset, interface between Nb12 and RBD with contact resides shown in stick representation). b, Structure of Nb30 and RBD region (inset, interface between Nb30 and RBD with contact resides shown as stick representation). c, Cryo-EM defined structures of nanomouse nanobodies recognize regions on RBD distal from residues 417, 484 and 501 (affected by mutations in emerging variants). d, SARS-CoV-2 spike (HexaPro) structure in two perpendicular views. e, Spike–Nb15 (red) complex structure in two perpendicular views. f, Spike–Nb17 (light green) complex structure in two perpendicular views. g, Spike–Nb19 (dark green) complex structure in two perpendicular views. h, Spike–Nb56 (purple) complex structure in two perpendicular views.

Extended Data Fig. 10. Binding and neutralization of sarbecoviruses by nanomouse nanobodies and gating strategy for nanomouse and wild-type B cells.

a, Neutralization using trivalent Nb56 and Nb12, and bivalent Nb30, against pseudoviruses carrying SARS-CoV (left) or bat coronavirus WIV16 (right) spikes. Data are representative of two independent experiments and the error bars are mean ± s.d. of triplicates. b, BLI analysis of trivalent Nb56 and Nb12, and bivalent Nb30, binding to immobilized RBD from SARS-CoV-2 (left), SARS-CoV (middle) and bat coronavirus WIV16 (right). Equilibrium (K_D) constants are provided. c, IC₅₀ (pM) values for neutralization in culture assays, showing the sensitivity of HIV-1 and VSV pseudotyped viruses containing 13 sarbecoviral spike proteins. d, Analysis of bone marrow (top) or splenic (bottom) B cells was done by gating lymphocytes (first plot), avoiding aggregates (plots 2 and 3), B220⁺apoptotic⁻ gaiting (plot 4), and visualization with cell-surface makers as indicated (plots 5 and 6).