Structures of Human Antibodies Bound to SARS-CoV-2 Spike Reveal Common Epitopes and Recurrent Features of Antibodies

Abstract

Neutralizing antibody responses to coronaviruses mainly target the receptor-binding domain (RBD) of the trimeric spike. Here, we characterized polyclonal immunoglobulin Gs (IgGs) and Fabs from COVID-19 convalescent individuals for recognition of coronavirus spikes. Plasma IgGs differed in their focus on RBD epitopes, recognition of alpha- and beta-coronaviruses, and contributions of avidity to increased binding/neutralization of IgGs over Fabs. Using electron microscopy, we examined specificities of polyclonal plasma Fabs, revealing recognition of both S1^A and RBD epitopes on SARS-CoV-2 spike. Moreover, a 3.4 Å cryo-electron microscopy (cryo-EM) structure of a neutralizing monoclonal Fab-spike complex revealed an epitope that blocks ACE2 receptor binding. Modeling based on these structures suggested different potentials for inter-spike crosslinking by IgGs on viruses, and characterized IgGs would not be affected by identified SARS-CoV-2 spike mutations. Overall, our studies structurally define a recurrent anti-SARS-CoV-2 antibody class derived from VH3-53/VH3-66 and similarity to a SARS-CoV VH3-30 antibody, providing criteria for evaluating vaccine-elicited antibodies.

Keywords: COVID-19; ELISA; Fab; IgG; MERS-CoV; SARS-CoV; SARS-CoV-2; convalescent plasma; coronavirus; electron microscopy.

Graphical abstract

Figure 1

Coronavirus S Proteins Show Localized Regions of Conservation and Variability (A) Schematic of SARS-CoV-2 S protein domain architecture. The S1 and S2 subunits are indicated, with scissors representing the locations of proteolytic cleavage sites required for S priming prior to fusion. UH, upstream helix; FP, fusion peptide; HR1, heptad repeat 1; CH, central helix; BH, β-hairpin; HR2, heptad repeat 2; TM, transmembrane region; CT, cytoplasmic tail. (B) Phylogenetic trees of selected coronaviruses based on protein sequences of S proteins and RBD/S1^B domains. (C) Sequence conservation of 7 human coronaviruses plotted as a surface. The sequence alignment was generated using SARS-CoV-2 (GenBank: MN985325.1), SARS-CoV (GenBank: AAP13441.1), MERS-CoV (GenBank: JX869059.2), HCoV-OC43 (GenBank: AAT84362.1), HCoV-229E (GenBank: AAK32191.1), HCoV-NL63 (GenBank: AAS58177.1), and HCoV-HKU1 (GenBank: Q0ZME7.1). Conservation was calculated by ConSurf Database (Landau et al., 2005) and displayed using a surface representation of the structure of the SARS-CoV-2 S protein (PDB: 6VXX).

Figure 2

Plasma Fabs Bind to SARS-CoV-2 S Protein (A) Schematic of polyclonal IgG and Fab purification from human plasma for nsEMPEM protocol. (B and C) SEC profile of Fabs (B) and SDS-PAGE of purified IgGs and Fabs (C) from COV21, COV57, and COV107 plasma samples. (D) SEC demonstration that plasma-derived Fabs from COV21 and COV57 shift the SARS-CoV-2 S protein trimer to a higher apparent molecular weight. No shift was observed when Fabs from COV107 were analyzed by SEC with S protein (data not shown). Fractions pooled and concentrated for nsEMPEM are boxed. See also Figure S1.

Figure S1

Characterization of Purified Proteins, Related to Figures 2 and 3 (A-C; G-I) SEC-MALS profiles of CoV S trimers. The absorbance at 280 nm (left y axis) is plotted against the SEC elution volume and overlaid with the molar mass determined for each peak (right y axis). The molar mass determined for each peak is indicated. (D-F; J-L) Corresponding representative nsEM images shown below the SEC-MALS profiles. Scale bars on micrographs represent 50 nm.

Figure 3

Convalescent Plasma IgG and Fab Binding Properties Demonstrate Recognition of Diverse Coronaviruses and Effects of Avidity (A–F) Results from ELISAs assessing binding of IgGs and Fabs purified from plasmas from 10 COVID-19 individuals (x axis) presented as area under the curve (AUC; shown as mean ± SEM of values derived from experiments conducted in triplicate). Binding was assessed against S and RBD proteins for SARS-CoV-2 (A), SARS-CoV (B), MERS-CoV (C), HCoV-NL63 (D), HCoV-OC43 (E), and HCoV-229E (F). Polyclonal IgGs or Fabs were evaluated at a top concentration of 50 μg/mL and 7 additional 4-fold serial dilutions. Binding of the IgG and Fab from IOMA, an antibody against HIV-1 (Gristick et al., 2016), was used as a control in each assay. (G) In vitro neutralization assays comparing the potencies of purified plasma IgGs and purified plasma Fabs. COV21, COV57, and COV107 plasma Fabs and IgGs are highlighted in the indicated colors; curves for 10 other plasmas (listed in H) are gray. (H) Molar IC₅₀ values for purified plasma IgGs and Fabs for the indicated plasmas are listed with the molar ratio for IC₅₀ (Fab) to IC₅₀ (IgG) shown in the right column. SEM was plotted versus SD. See also Figures S2 and S3.

Figure S2

SARS-CoV-2, SARS-CoV, MERS-CoV, and Common Cold Coronavirus ELISA Curves, Related to Figure 3 Anti-S IgG (left panels), Anti-S Fab (middle left panels), Anti-RBD/S1^B IgG (middle right panels), and Anti-RBD/S1^B Fab (right panels) ELISA binding data for (A) SARS-CoV-2, (B) SARS-CoV, (C) MERS-CoV, (D) HCoV-NL63, (E) HCoV-OC43, and (F) HCoV-229E. COV21: red curves; COV57: green curves; COV107: magenta curves. Curves for other plasmas are in gray. Each curve represents the average of three independent experiments. Binding of the IgG and Fab from IOMA, an antibody against HIV-1 (Gristick et al., 2016), was used as a control in each assay.

Figure S3

RBD Adsorption Experiments to Assess Degrees of Cross-Reactive RBD Recognition by Plasma IgGs, Related to Figure 3 (A-F) Purified IgGs from COVID-19 plasmas (indicated by numbers) and control plasmas (indicated as “con”) were adsorbed with one of two resins: a SARS-CoV-2 RBD resin (IgGs remaining after RBD adsorption; light gray bars) and a 2G12 mAb control resin (IgGs remaining after control adsorption; dark gray bars). IgGs remaining after adsorption were evaluated in ELISAs against the indicated RBD (or S1^B) domains. Binding of IgGs after adsorption to IOMA, an antibody against HIV-1 (Gristick et al., 2016), was used as a control in each assay. Results are presented as area under the curve (AUC; shown as mean and ± SEM of experiments conducted in duplicate). SEM was plotted versus SD.

Figure S4

Representative 2D Class-Averages and 3D Models from nsEMPEM of Human Convalescent Plasma, Related to Figure 4 (A,C,E) Representative reference-free 2D class-averages obtained from EM data collections of (A) SARS-CoV-2 S trimers alone, (C) SARS-CoV-2 S complexed with COV21 polyclonal Fabs, and (E) SARS-CoV-2 S complexed with COV57 polyclonal Fabs. For COV21 and COV57, class-averages demonstrating extra density beyond the S trimer core are highlighted (red boxes). For COV107, no extra density was observed in class averages or a 3D construction (data not shown). (B,D,F) Refined 3D models after iterative rounds of 2D and 3D classification. Features corresponding to Fabs are denoted.

Figure 4

EM Reveals Distinct Predominant Epitopes Targeted by Convalescent Plasma Antibodies (A) Side and top views for representative 3D reconstructions of four nsEMPEM datasets (S protein alone, S + COV21 Fabs, S + COV57 Fabs, S + COV107 Fabs). Bound Fabs observed in reconstructions from COV21 and COV57 plasmas are highlighted with false coloring as orange and green, respectively. No Fabs were observed in the reconstruction of COV107 Fabs plus S protein. Refined 3D models for SARS-CoV-2 S trimer-polyclonal Fab complexes from COV21 (B), and COV57 (C) were rigid-body fit with reference structures in Chimera (Goddard et al., 2007; Pettersen et al., 2004), displayed as cartoons (S1^A: blue, S1^B: red, S2: gray). (B) For COV21, the volume was best-fitted with PDB 6VYB (SARS-CoV-2, one “up” S1^B conformation, inset). Overlay of PDB 6NB6 showed similarities in S1^B epitope targeting of COV21 Fab (orange) and the human SARS-CoV neutralizing antibody, S230 (magenta, cartoon). (C) COV57 was fitted with PDB 6VXX (closed, prefusion conformation, inset). Fab density (green) was focused on the S1^A domain. See also Figure S4.

Figure 5

A Cryo-EM Structure of a Monoclonal Fab-S Protein Complex Resembles the COV21 Fab(s)-S Reconstruction (A) Reconstructed volumes for mAb C105 bound to SARS-CoV-2 S trimers in state 1 (two “up” RBDs, two bound Fabs) and state 2 (three “up” RBDs, three bound Fabs). (B) Left: cartoon representation of VH-VL domains of C105 bound to an RBD. Right: CDR loops of C105 overlaid on surface representation of the RBD (shown as a gray surface). (C) RBD surface showing contacts by C105 VH-VL (contacts defined as an RBD residue within 7 Å of a VH or VL residue Cα atom). (D) RBD surface fitted with volume representing the variable domains of the COV21 Fab(s) nsEMPEM reconstruction. (E) CDR loops of B38 mAb overlaid on surface representation of the RBD (from PDB: 7BZ5). (F) RBD surface showing contacts by ACE2 (contacts defined as an RBD residue within 7 Å of an ACE2 residue Cα atom) (from PDB: 6VW1). See also Figures S5 and S6.

Figure S5

Data Collection and Processing Pipeline for the Cryo-EM Structure of the C105-SARS-CoV-2 S Complex, Related to Figure 5 (A) Representative micrograph of C105-S complex in vitreous ice. Inset: Power spectrum of micrograph determined during CTF estimation showing Thon rings to 3.2 Å. (B) Reference-free 2D classification of extracted particles. (C) Workflow for classification and refinement of selected particles. Briefly, after selection of good 2D class averages, an ab initio model was generated, which was then homogeneously refined before further 3D classification. To improve features at the SARS-CoV-2 RBD-C105 Fab interface, particles from states 1 and 2 were combined and used for non-uniform, focused refinement to yield a state 1-like reconstruction to an FSC = 0.143 resolution of 3.4 Å.

Figure S7

CDRH3 Length Distributions, Related to Figure 5 (A) The CDRH3 lengths (IMGT definition) (Lefranc et al., 2015) of anti-SARS-CoV-2 RBD-binding mAbs (Robbiani et al., 2020) are shown in three groups: all 534 mAbs (dark gray), those derived from VH3-53 (red), and those derived from VH3-66 (green). For comparison, the CDRH3 length distribution from the human antibody repertoire (Briney et al., 2019) is also shown (normalized to the same total count as the set of 534). The CDRH3 length of mAb B38 is indicated with an arrow. (B) Length of CDRH3s in human antibodies versus predicted clashes with SARS-CoV-2 RBD if binding in the orientation observed for the mAbs B38 and C105. The VH domains of 1364 human antibody structures with resolutions of ≤ 3.5 Å downloaded from SAbDab (Dunbar et al., 2014) were aligned to the B38 VH domain in complex with SARS-CoV-2 RBD (PDB code 7BZ5) (Wu et al., 2020c). In cases in which there was more than one Fab in the crystallographic asymmetric unit, each VH was evaluated and enumerated separately. CDRH3 clashes were defined if any CDRH3 atom was within 2.0 Å of an atom in the RBD, a stringent criterion devised to account for not allowing CDR flexibility or different side chain rotamer conformations.

Figure 6

Identified S Mutations Are Unlikely to Affect Epitopes Revealed by nsEMPEM and Single-Particle Cryo-EM (A and B) Side (A) and top (B) views of refined 3D model of SARS-CoV-2 S trimer alone fitted with a reference structure (PDB: 6VYB; gray cartoon) to illustrate locations of mutations observed in circulating SARS-CoV-2 isolates (Table S3) (red spheres). Residues affected by mutations that are disordered in the SARS-CoV-2 S structure (V483A) or in regions that are not included in the S ectodomain (signal sequence or cytoplasmic tail) are not shown. Densities corresponding to Fabs were separated, colored, and displayed on the same 3D volume. (C) C105-RBD interaction from the cryo-EM structure of the C105-S complex (Figure 5) showing locations of RBD mutations. V483 is ordered in this structure. See also Tables S3 and S4.

Figure 7

S Protein Epitopes Offer Different Possibilities for Avidity Effects during IgG and Receptor Binding (A) Left: model of two adjacent S trimers separated by ∼15 nm, as seen on coronaviruses by cryo-electron tomography (Neuman et al., 2011), demonstrating that the orientation of COV21 Fab(s) on S could accommodate inter-spike crosslinking by a single IgG. The Fc portion of the IgG (PDB: 1IGT) was modeled assuming flexibility between the Fabs and the Fc (Sandin et al., 2004) and with the hinge region indicated by a dotted line since it is disordered in crystal structures of intact IgGs (Harris et al., 1992; Harris et al., 1998; Saphire et al., 2001). Right: Example of a model of two adjacent S trimers with bound Fab(s) in the orientation observed in the COV57 Fab(s)-S reconstruction demonstrating that inter-spike crosslinking is unlikely due to the “downward” orientation of the Fab(s), which does not permit linking by an Fc region, and predicted steric clashes between adjacent Fabs. Inter-spike crosslinking is also not possible for other orientations of two adjacent COV57 Fab(s)-S complexes (not shown). (B) Model of S trimers with two RBDs in an “up” position based on a cryo-EM structure of SARS-CoV S trimer (Kirchdoerfer et al., 2018) (PDB: 6CRX) interacting with full-length ACE2 receptors from the cryo-EM structure of soluble SARS-CoV-2 RBDs bound to the dimeric membrane form of ACE2 (Yan et al., 2020) (PDB: 6M17). Inter-spike crosslinking is possible if ACE2 dimers cluster in the membrane. (C) Model of intra-spike crosslinking between dimeric ACE2 and an S protein trimer with two RBDs in an “up” position. The RBDs were rotated by ∼180° about their long axes to allow binding of the ACE2 ectodomains. Rotation of the RBD is a possibility since its position is flexible with respect to the remaining part of the S trimer (Walls et al., 2019). In this model, RBDs from a single S trimer could bind the ACE2 dimer in the same configuration as seen in the B^oAT1-ACE2-SARS-CoV-2 RBD structure (PDB: 6M17).

Figure S6

Cryo-EM Structure Validation, Related to Figure 5 (A) Fourier shell correlation (FSC) plots calculated from half-maps of state 1 (black), state 1 after focused refinement (blue) and state 2 (red). Dotted lines for FSC values of 0.5 and 0.143 are shown. (B,C) 2D angular distribution plot for state 1 (panel B) and state 2 (panel C) reconstructions. (D) Local resolution estimations for states 1 and 2 and at the RBD-C105 Fab interface. (E) Representative density from S trimer and Fab regions of the state 1 reconstructed volume. Maps are contoured at 6σ.