Broad neutralization of SARS-related viruses by human monoclonal antibodies

Seeking broad protection As scientists develop therapeutic antibodies and vaccines against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the risk of emergent coronaviruses makes it important to also identify broadly protective antibodies. Wec et al. isolated and characterized hundreds of antibodies against the viral spike protein of SARS-CoV-2 from the memory B cells of a survivor of the 2003 outbreak caused by the related coronavirus, SARS-CoV. In both of these viruses, the spike protein facilitated viral entry by binding to the angiotensin-converting enzyme 2 (ACE2) receptor on human cells. The antibodies targeted multiple sites on the spike protein, but of nine antibodies that showed strong cross-neutralization, eight targeted the domain that binds to ACE2. These eight antibodies also neutralized a bat SARS-related virus. Illuminating the epitopes on the viral spike protein that bind cross-neutralizing antibodies could guide the design of broadly protective vaccines. Science, this issue p. 731

Broadly protective vaccines against known and preemergent human coronaviruses (HCoVs) are urgently needed. To gain a deeper understanding of cross-neutralizing antibody responses, we mined the memory B cell repertoire of a convalescent severe acute respiratory syndrome (SARS) donor and identified 200 SARS coronavirus 2 (SARS-CoV-2) binding antibodies that target multiple conserved sites on the spike (S) protein. A large proportion of the non-neutralizing antibodies display high levels of somatic hypermutation and cross-react with circulating HCoVs, suggesting recall of preexisting memory B cells elicited by prior HCoV infections. Several antibodies potently cross-neutralize SARS-CoV, SARS-CoV-2, and the bat SARS-like virus WIV1 by blocking receptor attachment and inducing S1 shedding. These antibodies represent promising candidates for therapeutic intervention and reveal a target for the rational design of pan-sarbecovirus vaccines. I n December 2019, a novel pathogen emerged in the city of Wuhan in China's Hubei province, causing an outbreak of atypical pneumonia [a disease known as coronavirus disease 2019 (COVID- 19)]. The infectious agent was characterized as a lineage B betacoronavirus, named severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and shown to be closely related to SARS-CoV and several SARS-like bat CoVs (1). There are currently no approved vaccines or therapeutics available for the prevention or treatment of COVID-19.
CoV entry into host cells is mediated by the viral S glycoprotein, which forms trimeric spikes on the viral surface (2). Each monomer in the trimeric S assembly is a heterodimer of S1 and S2 subunits. The S1 subunit is composed of four domains: an N-terminal domain (NTD), a C-terminal domain (CTD), and subdomains I and II (3)(4)(5). The CTD of both SARS-CoV and SARS-CoV-2 functions as the receptor-binding domain (RBD) for the shared entry receptor, human angiotensinconverting enzyme 2 (hACE2) (6)(7)(8)(9)(10). The S2 subunit contains the fusion peptide, heptad repeats 1 and 2, and a transmembrane domain, all of which are required for fusion of the viral and host cell membranes.
The S glycoprotein of HCoVs is the primary target for neutralizing antibodies (nAbs) (11). SARS-CoV and SARS-CoV-2 share 76% amino acid identity in their S proteins, raising the possibility of conserved immunogenic surfaces on these antigens. Studies of convalescent sera and a limited number of monoclonal antibodies (mAbs) have revealed limited to no cross-neutralizing activity, demonstrating that conserved antigenic sites are rarely targeted by nAbs (5,9,12,13). However, the frequencies, specificities, and functional activities of crossreactive antibodies induced by natural SARS-CoV and SARS-CoV-2 infection remain poorly defined.
We aimed to comprehensively profile the cross-reactive B cell response induced by SARS-CoV infection by cloning an extensive panel of SARS-CoV-2 S-reactive mAbs from the peripheral B cells of a convalescent donor (donor 84) who survived the 2003 SARS outbreak. To isolate cross-reactive antibodies, we obtained a blood sample from this donor about 3 years after infection and stained purified B cells with a panel of memory B cell (MBC) markers and a fluorescently labeled SARS-CoV-2 S protein. Flow cytometric analysis revealed that 0.14% of class-switched MBCs were SARS-CoV-2 S-reactive, which was about threefold greater than background staining observed with a SARS-CoV-naïve donor sample (Fig. 1A). Cognate antibody heavy-and light-chain pairs were amplified from 315 individual SARS-CoV-2-reactive B cells by single-cell reverse transcription polymerase chain reaction (RT-PCR) and subsequently cloned and expressed as fulllength immunoglobulin Gs (IgGs) in an engineered strain of Saccharomyces cerevisiae (14). Of the 315 cloned antibodies, 200 bound to SARS-CoV-2 S in preliminary binding screens (Fig. 1B). Sequence analysis revealed that about half of the clones were members of expanded clonal lineages, whereas the other half were unique (Fig. 1C). Moreover, about 30% of isolated antibodies displayed convergent VH1-69/VK2-30 germline gene pairing (Fig. 1C). As expected, almost all the antibodies were somatically mutated, with members of clonally expanded lineages showing significantly higher levels of somatic hypermutation (SHM) compared with unique clones (Fig. 1D). Finally, consistent with the respiratory nature of SARS-CoV infection, index sorting analysis revealed that 33% of binding antibodies originated from IgA + MBCs and the remaining 66% from IgG + MBCs (Fig. 1E). We conclude that SARS-CoV infection elicited a high frequency of long-lived, cross-reactive MBCs in this donor. We next measured the apparent binding affinities (K D App s) of the antibodies to prefusionstabilized SARS-CoV and SARS-CoV-2 S proteins (5). Although most antibodies (153 out of 200) showed binding to both S proteins, a subset appeared to be SARS-CoV-2 S-specific ( Fig. 2A). This result was unexpected given that the antibodies were isolated from a SARS-CoVexperienced donor and may relate to differences between the infecting SARS-CoV strain and the recombinant SARS-CoV S protein (Tor2) used for the binding studies. Alternatively, this result may be due to inherent differences in the stability or antigenicity of recombinant prefusion-stabilized SARS-CoV and SARS-CoV-2 S proteins. Indeed, about 30% of antibodies that failed to bind recombinant SARS-CoV S displayed reactivity with SARS-CoV S expressed on the surface of transfected cells, providing some evidence for differences in the antigenicity of recombinant and cell-expressed forms of S ( fig. S1).
Paradoxically, most of the highly mutated and clonally expanded antibodies bound weakly (K D Apps > 10 nM) to both SARS-CoV and SARS-CoV-2 S (Fig. 2B). We sought to determine if these antibodies originated from preexisting MBCs induced by prior exposures to naturally circulating HCoVs, which share up to 32% S amino acid identity with SARS-CoV and SARS-CoV-2. Accordingly, we assessed binding of the antibodies to recombinant S proteins of naturally circulating human alphacoronaviruses (HCoV-NL63 and HCoV-229E) and betacoronaviruses (HCoV-OC43 and HCoV-HKU1).
More than 80% of the low-affinity (K D Apps > 10 nM) SARS-CoV and SARS-CoV-2 crossreactive antibodies reacted with one or more of the HCoV S proteins, suggesting that SARS-CoV infection may have boosted a preexisting MBC response induced by circulating HCoVs (Fig. 2B). Consistent with this hypothesis, the broadly cross-reactive antibodies showed significantly higher levels of SHM and clonal expansion compared with those that only recognized SARS-CoV and SARS-CoV-2 ( Fig.  2, B to D). Furthermore, 72% of the broadly binding antibodies used VH1-69/VK2-30 germline gene pairing, suggesting germline-mediated recognition of a common antigenic site ( Fig. 2B  and fig. S2). Although we were unable to finely map the epitopes recognized by these antibodies, none of them bound to recombinant SARS-CoV-2 S1, suggesting that they likely target epitopes within the more conserved S2 subunit ( fig. S3). Index sorting analysis revealed that the majority of the broadly crossreactive antibodies were derived from IgA + MBCs, indicating a mucosal origin, whereas most of the SARS-CoV and SARS-CoV-2 crossreactive antibodies originated from IgG + MBCs (Fig. 2E). Finally, all of the broad binders lacked polyreactivity, demonstrating that their cross-binding is not due to nonspecific cross-reactivity ( fig. S4).
To investigate whether the above results were due to an original antigenic sin phenomenon, or rather simply due to avid binding of circulating HCoV-specific B cell receptors to the SARS-CoV-2 S tetramers used for cell sorting, we assessed whether similarly broadly binding antibodies were also present in SARS-CoV-and SARS-CoV-2-naïve donors that had been exposed to endemic HCoVs. We obtained peripheral blood mononuclear cell (PBMC) samples from three healthy adult donors with serological evidence of circulating HCoV exposure and no history of SARS-CoV or SARS-CoV-2 infection and stained the corresponding B cells with a fluorescently labeled SARS-CoV-2 S probe ( fig. S5A). Flow cytometric analysis revealed that between 0.06 and 0.12% of total B cells in the three naïve donors displayed SARS-CoV-2 reactivity ( fig. S5B). More than 350 SARS-CoV-2-reactive MBCs were sorted and amplified by single-cell RT-PCR, and 141 variable region of Ig heavy chain (V H )-variable region of Ig light chain (V L ) pairs were cloned and expressed as full-length IgGs. Although a limited number of SARS-CoV-2 S binding antibodies (3 to 22) were isolated from all three naïve donors, they displayed significantly lower levels of SHM, clonal expansion, and K D Apps for both SARS-CoV and SARS-CoV-2 S compared with the cross-reactive antibodies identified from donor 84 (Fig. 2, F and G, and fig. S5C). Altogether, these results suggest that SARS-CoV infection likely led to the activation and expansion of preexisting cross-reactive MBCs induced by circulating HCoV exposure in this donor.
To map the antigenic sites recognized by the SARS-CoV and SARS-CoV-2 cross-reactive antibodies isolated from donor 84, we performed binding experiments using a panel of recombinant S protein subunits and individual domains. Because of the inherent technical challenges associated with measuring binding of low-affinity antibodies to monomeric proteins, we analyzed only the 64 high-affinity binders (K D Apps < 10 nM) to SARS-CoV-2 S (Fig. 2, A and B). We first evaluated binding to recombinant SARS-CoV-2 S1 and S2 subunits and observed that 75% of the antibodies recognized epitopes within S1, whereas the remaining 25% bound to epitopes within S2 (Fig. 3A). Two of the S2-directed antibodies also showed strong reactivity with OC43 S, suggesting recognition of a conserved antigenic site ( fig. S6). We next evaluated the 49 S1-directed antibodies for reactivity with individual SARS-CoV-2 RBD and NTD proteins and found that 21 (43%) and 28 (57%) of the S1-specific antibodies recognized the RBD and NTD, respectively (Fig. 3A).
To further define the epitopes recognized by the 21 RBD-directed antibodies, we performed competitive binding studies with recombinant hACE2 and a previously described antibody, CR3022, that targets a conserved epitope that is distinct from the receptor binding site (Fig. 3B and fig. S7) (15). Six of the antibodies competed only with hACE2, three competed only with CR3022, four competed with both hACE2 and CR3022, and seven did not compete with hACE2 or CR3022 (Fig.  3B). Thus, these antibodies delineate at least four adjacent and potentially overlapping sites within the RBD. Most of the antibodies that competed with recombinant hACE2 binding to SARS-CoV-2 RBD in the biolayer interferometry (BLI) assay also interfered with binding of full-length SARS-CoV-2 S to endogenous ACE2 expressed on the surface of Vero E6 cells (Fig. 3C). The four antibodies (ADI-55951, Wec  ADI-55993, ADI-56000, and ADI-56035) that showed stronger competition in the BLI assay displayed weak binding affinities for SARS-CoV-2 S ( fig. S12), which likely explains their lower level of competition in the cell-surface assay. Thus, SARS-CoV infection elicited highaffinity cross-reactive antibodies to a range of antigenic sites within both the S1 and S2 subunits.
To evaluate the neutralization activities of the SARS-CoV-2 binding antibodies, we per-formed neutralization assays using both murine leukemia virus (MLV)-and vesicular stomatitis virus (VSV)-based pseudotype systems as well as authentic SARS-CoV-2. Because of the large number of antibodies, we first measured infection inhibition of authentic SARS-CoV-2 at a single concentration of purified IgG. Only 9 out of 200 antibodies displayed neutralizing activity at the 100 nM concentration tested, eight targeted the RBD, and the remaining one recognized the NTD (Fig. 3D). Similar results were observed in the VSV-based pseudovirus assay (fig. S8). Of the eight RBDdirected nAbs, four targeted epitopes overlapping with both the hACE2 and CR3022 epitopes and the other four recognized epitopes overlapping only the hACE2 epitope, suggesting the existence of two partially overlapping neutralizing epitopes within the RBD (Fig. 3B). Neutralization titration studies revealed that the median inhibitory concentrations (IC 50 s) of the RBD-directed nAbs ranged from 0.05 Wec   to 1.4 mg/ml against SARS-CoV-2 and 0.004 to 0.06 mg/ml against SARS-CoV in the MLV assay ( Fig. 3E and fig. S9). Comparable neutralization IC 50 s were observed in authentic SARS-CoV and SARS-CoV-2 neutralization assays ( Fig. 3E and fig. S9). By contrast, the VSV-SARS-CoV-2 neutralization IC 50 s were substantially lower (8-to 35-fold) than those observed for live SARS-CoV-2 (figs. S9 and S10). To assess the breadth of neutralization against representative preemergent SARS-like bat CoVs, we measured infection inhibition of authentic WIV1-CoV using a plaque reduction assay (16). All eight antibodies neutralized WIV1-CoV, with median plaque reduction neutralization titers (PRNT 50 s) ranging from 0.076 to 1.7 mg/ml, demonstrating their breadth of activity ( Fig. 3E and fig. S11). Crucially, none of the antibodies left an unneutralized viral fraction in any of the assays (figs. S9 and S11).
We observed little to no correlation between apparent binding affinity for wild-type (WT) SARS-CoV-2 cell surface S and neutralizing activity. For example, all of the S2-directed antibodies and a subset of NTD-directed antibodies bound with high avidity to both re-combinant and cell surface S, but none were neutralizing (Fig. 3F). Surprisingly, even within the group of hACE2-blocking nAbs, we did not observe a strong correlation between binding to cell surface-S or recombinant-S and neutralization, suggesting that antibody potency is governed at least in part by factors beyond binding affinity (Fig. 3F and figs. S12 and S13). To determine whether the hACE2 competitor antibodies neutralized by inducing S1 shedding and premature S triggering (17), we incubated human embryonic kidney (HEK)-293 cells expressing WT SARS-CoV-2 S with saturating concentrations of antibody and measured the median fluorescence intensity (MFI) of antibody binding over time by flow cytometry. Indeed, all of the hACE2-blocking antibodies showed substantially decreased binding over time, consistent with induced S1 dissociation, whereas antibodies recognizing the NTD, S2 stem, and RBD epitopes outside of the hACE2 binding site displayed either no change or an increase in binding over time (Fig. 3G). We conclude that SARS-CoV infection induces high-affinity cross-reactive antibodies targeting multiple distinct antigenic sites on the S pro-tein, but neutralizing activity is primarily restricted to RBD-directed antibodies that interfere with receptor binding and promote S1 dissociation.
To structurally characterize the epitopes recognized by the RBD-directed nAbs, we performed negative-stain electron microscopy (EM) to observe each of these Fabs bound to the SARS-CoV-2 S protein. Many of the twodimensional (2D) class averages that we obtained displayed obvious heterogeneity in the number of Fabs that were bound to a single S trimer, which is likely due to dynamic inaccessibility of RBD epitopes and substoichiometric binding of S at the low protein concentrations used to prepare grids (Fig. 4A) (5,18). The 3D reconstructions of these complexes support the results of our biophysical competition assays and show that the RBDdirected nAbs recognize a single region on the solvent-exposed surface of the RBD with overlapping footprints. ADI-55689, which potently neutralizes and competes with hACE2, appears to bind at the edge of the hACE2 binding site, close to the more structurally conserved core domain of the RBD, without overlapping with the CR3022 epitope (Fig.  4B). ADI-56046, which exemplifies the group of antibodies that compete with both hACE2 and CR3022, binds slightly farther away from the flexible tip of the RBD, and thus its epitope spans both the hACE2 binding site and the CR3022 epitope (Fig. 4C). Our structural analysis suggests that all of the nAbs recognize a single patch on the surface of the RBD with overlapping footprints. These antibodies potently cross-neutralize SARS-CoV, SARS-CoV-2, and WIV1, suggesting that this antigenic surface exhibits extensive conservation among the SARS-like coronaviruses.
The potent cross-neutralizing antibodies described here bind to conserved epitopes overlapping the hACE2 binding site, thus illuminating this antigenic surface as a promising target for the rational design of pan-sarbecovirus vaccines. For example, the RBD epitope(s) defined by this class of antibodies could be presented on conformationally stable protein scaffolds to focus the antibody response on this site, as previously demonstrated for the motavizumab epitope on respiratory syncytial virus F (19). Furthermore, the nAbs themselves, alone or in combination, represent promising candidates for prophylaxis or therapy of SARS, COVID-19, and potentially future diseases caused by new emerging SARS-like viruses.