Summary
SARS-CoV-2 neutralizing antibodies (NAbs) protect against COVID-19, making them a focus of vaccine design. A safety concern regarding SARS-CoV-2 antibodies is whether they mediate disease enhancement. Here, we isolated potent NAbs against the receptor-binding domain (RBD) and the N-terminal domain (NTD) of SARS-CoV-2 spike protein from individuals with acute or convalescent SARS-CoV-2 or a history of SARS-CoV-1 infection. Cryo-electron microscopy of RBD and NTD antibodies demonstrated function-specific modes of antibody binding. Select RBD NAbs also demonstrated Fc receptor-γ (FcγR)-mediated enhancement of virus infection in vitro, while five non-neutralizing NTD antibodies mediated FcγR-independent in vitro infection enhancement. However, both in vitro neutralizing and infection-enhancing RBD or infection-enhancing NTD antibodies protected from SARS-CoV-2 challenge in non-human primates and mice. One of 30 monkeys infused with enhancing antibodies had lung pathology and bronchoalveolar lavage cytokine evidence suggestive of enhanced disease. Thus, these in vitro assessments of enhanced antibody-mediated infection do not necessarily indicate biologically relevant in vivo infection enhancement.
Introduction
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a global pandemic with over 43 million cases and 1.16 million deaths (https://coronavirus.jhu.edu/). Development of combinations of neutralizing antibodies for prevention or treatment of infection can significantly help to control the pandemic, while the ultimate solution to control the COVID-19 pandemic is a safe and effective vaccine (Graham, 2020; Sempowski et al., 2020).
Neutralizing antibodies (NAbs) that can block viral entry are crucial for controlling virus infections (Battles and McLellan, 2019; Corti and Lanzavecchia, 2013; Dashti et al., 2019). Previously reported SARS-CoV and MERS-CoV NAbs function by targeting the receptor-binding domain (RBD) or the N-terminal domain (NTD) of spike (S) protein to block receptor binding, or by binding to the S2 region of S protein to interfere with S2-mediated membrane fusion (Du and Jiang, 2010; Jiang et al., 2020; Xu et al., 2019). Importantly, prophylactic or therapeutic use of SARS-CoV-2 NAbs in non-human primates (Shi et al., 2020) or rodent models (Hassan et al., 2020; Rogers et al., 2020; Wu et al., 2020b) showed protection from SARS-CoV-2-induced lung inflammation and/or reduction in viral load. SARS-CoV-2 NAbs reported to date predominantly target the RBD region (Baum et al., 2020; Brouwer et al., 2020a; Cao et al., 2020; Chen et al., 2020; Hansen et al., 2020; Ju et al., 2020; Liu et al., 2020a; Pinto et al., 2020; Robbiani et al., 2020; Rogers et al., 2020; Shi et al., 2020; Wang et al., 2020a; Wrapp et al., 2020a; Wu et al., 2020b). In contrast, NTD antibodies that neutralize SARS-CoV-2 are rare and of modest neutralization potency (Brouwer et al., 2020b; Chi et al., 2020; Wec et al., 2020; Zost et al., 2020a; Zost et al., 2020b).
Antibody-dependent enhancement (ADE) of infection in vitro has been reported with a number of viruses. ADE has been associated with vaccination for respiratory syncytial virus (RSV), with vaccination for dengue virus, or with dengue virus infection (Arvin et al., 2020). ADE is often mediated by Fc receptors for IgG (FcγRs), complement receptors (CRs) or both, and is most commonly observed in cells of monocyte/macrophage and B cell lineages (Iwasaki and Yang, 2020; Ubol and Halstead, 2010). In vitro studies have demonstrated FcγR-mediated ADE of SARS-CoV-1 infection of ACE2-negative cells (Jaume et al., 2011; Kam et al., 2007; Wan et al., 2020; Wang et al., 2014; Yilla et al., 2005; Yip et al., 2016; Yip et al., 2014). One group demonstrated FcγR-independent infection enhancement of SARS-CoV-1 in Vero cells, and isolated a monoclonal antibody that was suggested to induce enhanced lung viral load and pathology in vivo (Wang et al., 2016). The ability of antibodies that bind the SARS-CoV-2 S protein to mediate infection enhancement in vivo is unknown but is a theoretical concern for COVID-19 vaccine development (Arvin et al., 2020; Bournazos et al., 2020; Iwasaki and Yang, 2020).
Here, we identified potent in vitro neutralizing RBD and NTD antibodies as well as in vitro infection-enhancing RBD and NTD antibodies from individuals infected with SARS-CoV-1 or SARS-CoV-2. Negative stain electron microscopy (NSEM) and cryo-electron microscopy (cryo-EM) revealed distinct binding patterns and the precise epitopes of infection-enhancing and neutralizing antibodies. In vitro studies using human FcγR-expressing or ACE2-expressing cell lines demonstrated that the RBD antibodies mediated FcγR-dependent infection enhancement, whereas the NTD antibodies induced FcγR-independent infection enhancement. However, using monkey and mouse models of SARS-CoV-2 infection, none of these infection-enhancing antibodies consistently augmented SARS-CoV-2 lung viral load, infectious virus in the lung, or lung disease pathology in vivo. Rather, one of 30 monkeys had lung pathology and bronchoalveolar lavage (BAL) cytokine levels suggestive of enhanced lung disease. In vitro infection-enhancing antibodies prevented or reduced virus replication in cynomolgus macaques or mice challenged with SARS-CoV-2, SARS-CoV-2 mouse-adapted (MA) viruses (Dinnon et al., 2020b; Leist et al., 2020a), or bat WIV1-coronavirus (CoV) (Menachery et al., 2016). Thus, the in vitro infection enhancing activity of SARS-CoV-2 RBD and NTD antibodies only rarely translated to in vivo relevance.
RESULTS
Isolation of infection-enhancing SARS-CoV-2 antibodies
SARS-CoV-2-reactive human monoclonal antibodies from plasmablasts or SARS-CoV-2-reactive memory B cells were isolated by flow cytometric sorting (Liao et al., 2009; Liao et al., 2013) from a SARS-CoV-2 infected individual 11, 15, 36, and 70 days post-onset of symptoms. Additional memory B cells were isolated from an individual infected with SARS-CoV-1 ∼17 years prior to sample collection (Figures 1A-B, S1 and S2). We isolated and characterized 1,737 antibodies, which bound to SARS-CoV-2 S and nucleocapsid (NP) proteins (Figure 1C; Table S1). We selected 187 antibodies for further characterization, and first examined neutralization of SARS-CoV-2 pseudovirus and replication-competent SARS-CoV-2. Forty-four of 81 recombinant RBD antibodies exhibited neutralization when assayed in 293T/ACE-2 cell pseudovirus, SARS-CoV-2 microneutralization, or SARS-CoV-2 plaque reduction assays (Figures S3A-F; Tables S2-S3).
Ten of the forty-one NTD antibodies neutralized in the 293T/ACE2 pseudovirus and plaque reduction assays, with the most potent antibody neutralizing pseudovirus with an IC50 of 39 ng/mL (Figures S3G-I; Tables S4-S5). In addition, 5 non-neutralizing NTD antibodies enhanced SARS-CoV-2 pseudovirus infection in 293T/ACE2 cells by 56% to 148% (Figure 1D). To determine whether this effect was applicable to replication-competent virus, we tested these five antibodies for enhancement of SARS-CoV-2 virus infection of Vero cells. Infection of replication-competent SARS-CoV-2 nano-luciferase virus (Hou et al., 2020) also increased in the presence of each of the 5 NTD antibodies (Figure 1E). Analysis of in vitro enhancement of NTD antibodies in ACE2-negative TZM-bl cells demonstrated no infection enhancement, suggesting enhancement-dependence on ACE2. Both ACE2-expressing 293T cells used for pseudovirus assays and Vero cells lack FcγR expression (Takada et al., 2007). Thus, NTD enhancement of SARS-CoV-2 infection was FcγR-independent.
Previous studies have demonstrated FcγR-mediated ADE of SARS-CoV-1 infection in ACE2-negative cells (Jaume et al., 2011; Kam et al., 2007; Wan et al., 2020; Wang et al., 2014; Yilla et al., 2005; Yip et al., 2016; Yip et al., 2014). Here, FcγR-dependent infection enhancement was determined by generating stable TZM-bl cell lines that expressed individual human FcγRs (FcγRI, FcγRIIa, FcγRIIb or FcγRIII). TZM-bl cells naturally lack ACE-2 and TMPRSS2 receptors, thus SARS-CoV-2 was unable to infect FcγR-negative TZM-bl cells (Figure 1F). One hundred S-reactive IgG1 antibodies selected from Tables S2-S7 were tested for their ability to facilitate SARS-CoV-2 infection of TZM-bl cells. Three of the antibodies enabled SARS-CoV-2 infection of TZM-bl cells expressing FcγRI, and five antibodies mediated infection of FcγRIIb-expressing TZM-bl cells (Figures 1F-J). The antigen-binding fragments (Fabs) of these antibodies did not mediate infection enhancement of TZM-bl cells expressing FcγRI or FcγRIIb (Figures 1K-L). FcγRI and FcγRIIb-dependent infection-enhancing antibodies were specific for the RBD of S protein, consistent with a recent finding by another group using COVID-19 patient sera or a recombinant antibody (Wu et al., 2020a). Thus, RBD antibodies can be either neutralizing in the 293T/ACE2 cell line, infection-enhancing in the TZM-bl-FcγR-expressing cell lines, or both (Figure 2A). NTD antibodies can either be neutralizing or infection-enhancing in the 293T/ACE2 cell line or Vero E6 cells (Figure 2A). Therefore, the repertoire of SARS-CoV-2 antibodies included potent neutralizing RBD and NTD antibodies, FcγR-dependent, infection-enhancing RBD antibodies, and FcγR-independent infection-enhancing NTD antibodies.
Characterization of infection-enhancing Spike recombinant antibodies
We compared the phenotypes and binding modes to S protein for five infection-enhancing RBD antibodies and three RBD antibodies that lacked infection enhancement to elucidate differences between these two types of antibodies. All selected RBD antibodies neutralized SARS-CoV-2 pseudovirus and/or replication-competent virus in ACE2-expressing cells (Figures 2A and S5), despite five of these antibodies mediating infection enhancement in ACE2-negative, FcγR-positive TZM-bl cells (Figures 1F-L, 2A,and S5). Both types of selected RBD antibodies blocked ACE2 binding to S protein. Fabs of four of the infection-enhancing RBD antibodies and two of the non-infection-enhancing RBD antibodies bound to S with high affinities ranging from 0.1 to 9 nM (Figures S6-8). Thus, the infection-enhancing or non-enhancing RBD antibodies showed similarities in ACE2 blocking, affinity, and neutralization of ACE2-dependent SARS-CoV-2 infection (Figure 2A).
We obtained NSEM reconstructions of Fabs in complex with stabilized S ectodomain trimer for six representative RBD antibodies (Hsieh et al., 2020). Infection-enhancing RBD antibodies DH1041 and DH1043 bound with a vertical approach (Figure 2B), parallel to the central axis of the S trimer, similar to non-infection-enhancing antibodies DH1042 and DH1044 (Figure 2C). The epitopes of antibodies DH1041, DH1042, and DH1043 overlapped with that of the ACE-2 receptor (Wang et al., 2020b), consistent with their ability to block ACE-2 binding to S protein (Figure 2A and S10). Their epitopes were similar to those of three recently described antibodies, P2B-2F6 (Ju et al., 2020), H11-H4, and H11-D4 (Figure S10) (Huo et al., 2020; Zhou et al., 2020). The epitope of another non-infection-enhancing RBD antibody DH1044 was only slightly shifted relative to DH1041, DH1042 and DH1043 (Figure 2C), but resulted in DH1044 not blocking ACE2 binding (Figure 2A and S10). The remaining two RBD antibodies, DH1045 and DH1047 cross-reacted with both SARS-CoV-1 and SARS-CoV-2 S (Figure 2A and S4). Although DH1047 mediated FcγR-dependent infection of TZM-bl cells and DH1045 did not, both antibodies bound to RBD-up S conformations with a more horizontal angle of approach (Figure 2B-C and S10) (Pak et al., 2009). Thus, epitopes and binding angles of RBD antibodies determined by NSEM did not discriminate between antibodies that mediated FcγR-dependent infection enhancement and those that did not.
We performed similar comparative analyses of five FcγR-independent, infection-enhancing NTD antibodies, and five non-infection-enhancing NTD antibodies (DH1048, DH1049, DH1051, DH1050.1 and DH1050.2) that neutralized SARS-CoV-2 (pseudovirus IC50 titers 39 − 520 ng/mL; SARS-CoV-2 plaque reduction IC50 titers 390 − 780 ng/mL) (Figure 2A and S5C-D). The Fabs of neutralizing antibodies DH1050.1 and DH1051 bound to stabilized S ectodomain with affinities of 16 and 19 nM respectively, whereas the infection-enhancing antibody DH1052 bound with 294 nM affinity (Figures S6-8). NSEM reconstructions obtained for nine of the ten NTD antibodies showed that the FcγR-independent, infection-enhancing NTD antibodies (DH1053-DH1056) bound to S with their Fab constant domains directed downward toward the virus membrane (Figure 2D), whereas the five neutralizing NTD-directed Abs (DH1048-DH1051) bound to S with the constant domain of the Fab directed upward away from the virus membrane (Figure 2E). Thus, S protein antibody epitopes and binding modes were associated with FcγR-independent, infection-enhancing activity of NTD antibodies. The five neutralizing antibodies bound the same epitope as antibody 4A8 (Wrapp et al., 2020a), with three of the five having the same angle of approach as 4A8 (Figure S11). Interestingly, the NTD antibodies with the same angle of approach as 4A8, were also genetically similar to 4A8, being derived from the same VH1-24 gene segment (Table S8), although their light chains were different (Wrapp et al., 2020a). These antibodies may constitute a neutralizing antibody class that can be reproducibly elicited upon SARS-CoV-2 infection.
Competition between infection-enhancing and non-infection enhancing antibodies
To determine whether infection-enhancing antibodies could compete with non-infection-enhancing antibodies for binding to S ectodomain, we performed surface plasmon resonance (SPR) competitive binding assays. RBD antibodies segregated into two clusters, where antibodies within a cluster blocked each other and antibodies in different clusters did not block each other (Figures 3A and S12). One cluster included antibodies DH1041, DH1043 and DH1044, and the other cluster included antibodies DH1046 and DH1047. NSEM reconstructions showed DH1041 and DH1047 Fabs bound simultaneously to different epitopes of the stabilized S trimer (Figure 3B). Similarly, DH1043 and DH1047 Fabs also bound simultaneously to different epitopes on the stabilized S protein (Figure 3B).
NTD antibodies also segregated into two clusters based on their ability to block each other (Figure 3A). Neutralizing NTD antibodies blocked each other and formed one cluster, while infection-enhancing/non-neutralizing NTD antibodies blocked each other and formed a second cluster (Figures 3A, 3C, S12 and S13). NSEM reconstruction of SARS-CoV-2 S trimer bound with Fabs of neutralizing NTD antibody DH1050.1 and infection-enhancing NTD antibody DH1052 confirmed that the two antibodies could simultaneously bind to distinct epitopes on a single SARS-CoV-2 S trimer (Figure 3D). DH1054 was unique as it was able to block both infection-enhancing and neutralizing NTD antibodies (Figures 3C and S13).
NTD antibodies did not compete with RBD antibodies for binding to S trimer (Figure 3A). This result gave rise to the notion that in a polyclonal mixture of antibodies, the SARS-CoV-2 S trimer could bind both RBD and NTD antibodies. To determine the potential for this complex to form, we liganded SARS-CoV-2 S trimer with Fabs of each type of antibody and visualized the complex using NSEM. NSEM showed that neutralizing RBD antibodies could also bind to the same S protomer as neutralizing NTD antibodies DH1050.1 or DH1051 (Figure 3E). Moreover, we found that a single S protomer could be simultaneously occupied by two RBD antibodies (DH1043 and DH1047) and an NTD antibody (DH1050.1) (Figure 3F). Thus, the S trimer could simultaneously bind to multiple RBD and NTD neutralizing antibody Fabs.
FcγR-independent infection-enhancement in the presence of neutralizing antibodies
The NSEM determination of antibody binding modes demonstrated that certain infection-enhancing antibodies and non-infection enhancing antibodies bound to distinct epitopes on the same S protomer (Figures 3A-F). Thus, we determined the functional outcome of infection-enhancing antibodies binding to S in the presence of neutralizing antibodies. We examined whether a FcγR-independent, infection-enhancing NTD antibody could inhibit RBD antibody neutralization of ACE-2-mediated SARS-CoV-2 pseudovirus infection of 293T/ACE2 cells in vitro. We hypothesized that the outcome would be dependent on which antibody was present at the highest concentration. We examined RBD antibody DH1041 neutralization in the presence of 132 and 1,325-fold excess of infection-enhancing NTD antibody DH1052. DH1041 neutralization activity was minimally decreased in the presence of 132-fold excess of DH1052. When DH1041 neutralization was assessed in the presence of 1,325-fold excess of antibody DH1052, infection enhancement was observed when DH1041 concentration was below 10 ng/mL (Figures 3G and S14). A nearly identical result was obtained when we examined neutralization by DH1043 (Figures 3H and S14). Thus, a ∼1000-fold excess of infection-enhancing NTD antibody was required to out-compete the effect of a potent RBD neutralizing antibody in vitro.
Cryo-EM structural determination of RBD and NTD-directed antibody epitopes
To visualize atomic level details of their interactions with the S protein, we selected representatives from the panels of RBD and NTD-directed antibodies for structural determination by cryo-EM. Of the RBD-directed antibodies, we selected two (DH1041 and DH1043) that most potently neutralized SARS-CoV-2 virus in the 293T/ACE2 pseudovirus assay and also enhanced infection in TZM-bl-FcγRI or −FcγRIIb cells. We also selected the RBD-directed antibody DH1047, which shared the infection enhancing and ACE-2 blocking properties of DH1041 and DH1043, but unlike DH1041 and DH1043, DH1047 also showed reactivity with the SARS-CoV-1 S protein. From the panel of NTD-directed antibodies, we selected one infection-enhancing NTD antibody, DH1052, and one neutralizing NTD antibody, DH1050.1, for higher resolution structural determination by cryo-EM. The stabilized SARS-CoV-2 S ectodomain “2P” (S-2P) (Wrapp et al., 2020b) was used for preparing complexes for structural determination.
For all three RBD-directed antibodies, the cryo-EM datasets revealed heterogeneous populations of S protein with at least one RBD in the “up” position (Figures 4, S15 and S16). We did not find any unliganded S or any 3-RBD-down S population, even though the unliganded S-2P consistently shows a 1:1 ratio of 1-RBD-up and 3-RBD-down populations (Henderson et al., 2020; Walls et al., 2020), suggesting that binding of the RBD-directed antibodies to S protein alters RBD dynamics. All S-2P trimers were stoichiometrically bound to 3 Fabs, with antibodies bound to both up and down RBDs in an S-2P trimer.
We observed that the primary epitopes of DH1041 and DH1043 were centered on the Receptor Binding Motif (RBM; residues 483-506) of the RBD (Figures 4A-B, S17 and S18), providing structural basis for the ACE-2 blocking phenotype of these antibodies. While DH1041 utilized its heavy chain complementarity determining regions (CDRs) to contact the RBM, the DH1043 paratope included both its heavy and light chains.
Unlike the RBM-focused epitope of DH1041 and DH1043, the epitope of antibody DH1047 was focused around the α2 and α3 helices and ß2 strand that are located outside the N-terminus of the RBM (Figures 4C and S19)(Ju et al., 2020). Contact is also made by RBD residues 500-506 that are also outside the RBM but at its C-terminal end, and stack against the N-terminal end of the α3 helix providing a continuous interactive surface for DH1047 binding. The DH1047 paratope included heavy chain complementarity determining regions HCDR2, HCDR3 and light chain LCDR1 and LCDR3. The HCDR3 stacks against and interacts with the residues in the ß2 strand. Interactions with the ß2 strand are also mediated by HCDR2. Similar to DH1041 and DH1043, the DH1047 interacted with an “up” RBD conformation from an adjacent protomer although these interactions were not well-characterized due to disorder in that region.
We next determined cryo-EM structures of the NTD-directed neutralizing antibodies, DH1050.1 (Figure 4D) and NTD-directed infection-enhancing antibody, DH1052 (Figure 4E), at 3.4 Å and 3.0 Å resolutions, respectively. Unlike the RBD-directed antibodies DH1041, DH1043 and DH1047, where we only observed spikes with at least one RBD in the up position, the cryo-EM datasets of DH1050.1- and DH1052-bound complexes showed antibody bound to both 3-RBD-down and 1-RBD-up S-2P spikes (Figures S20-S21). Consistent with the NSEM reconstructions, the neutralizing antibody DH1050.1 and the non-neutralizing, infection-enhancing antibody DH1052 bound opposite faces of the NTD, with the epitope for the neutralizing antibody DH1050.1 facing the host cell membrane and the epitope for the non-neutralizing, infection-enhancing antibody DH1052 facing the viral membrane. The dominant contribution to the DH1050.1 epitope came from NTD loop region 140-158 that stacks against the antibody HCDR3 and extends farther into a cleft formed at the interface of the DH1050.1 HCDR1, HCDR2 and HCDR3 loops. The previously described NTD antibody 4A8 interacts with the same epitope in a similar elongated HCDR3-dominated manner making similar contacts, although DH1050.1 and 4A8 (Chi et al., 2020) show a rotation about the stacked HCDR3 and NTD 140-158 loops, suggesting focused recognition of the elongated NTD loop by a class of antibodies sharing the same VH origin. Consistent with their diverse light chain gene origins, the light chains of DH1050.1 and 4A8 do not contact the S protein. The focused recognition of the NTD loop 140-158 by DH1050.1 is reminiscent of the interactions that HIV-1 fusion peptide-directed antibodies make with the HIV-1 Env, where recognition is focused on the conserved region of the flexible fusion peptide (Dingens et al., 2018; Xu et al., 2018). In a structurally conserved mechanism, such focused recognition is characterized by robust cryo-EM density at the recognition site but increased disorder away from the site of focused interactions (Figures 4E and S21).
The infection enhancing NTD-directed antibody DH1052 bound the NTD at an epitope facing the viral membrane and comprised of residues spanning 27-32, 59-62 and 211-218, with all the CDR loops of both heavy and light chains involved in contacts with the NTD. We also observed contact of the antibody with the glycan at position 603, as well as the conformationally invariant SD2 region.
Thus, we found that the RBD-directed antibodies isolated in this study influenced RBD dynamics and bound only to spike with at least one RBD in the up conformations, and in some cases, also induced the 2-RBD-up and 3-RBD-up spike conformations. In contrast, the NTD-directed antibodies bound to both the 3-RBD-down and 1-RBD-up spikes that are present in the unliganded S-2P. Our results provide a structural explanation for the ACE-2 blocking activity of RBD-directed antibodies as well as for the cross-reactivity of the DH1047 antibody. We observed two distinct orientations for the NTD-directed antibodies that are either neutralizing or non-neutralizing, with the former binding the NTD surface that faces away from the viral membrane and the latter binding the surface that faces the viral membrane.
NTD antibodies that mediate SARS-CoV-2 infection enhancement in vitro, do not necessarily enhance infection or disease in murine and macaque models of infection
To determine the biological relevance of in vitro infection enhancement by NTD antibodies, we assessed enhanced SARS-CoV-2 acquisition or disease severity in vivo in a COVID-19 disease mouse model of aged BALB/c mice challenged with the mouse-adapted SARS-CoV-2 MA10 strain (Leist et al., 2020a). The FcγR-independent, in vitro infection-enhancing antibody DH1052 or a control influenza antibody CH65 were given 12 hours prior to SARS-CoV-2 MA10 infection (Figure 5A). Throughout the four days of infection, DH1052-infused mice exhibited similar levels of body weight loss and higher survival than mice given negative control IgG (2/9 control mice died while 0/10 DH1052-treated mice died) (Figures 5B-C). In addition, DH1052-treated mice exhibited lower lung hemorrhagic scores, lower lung viral titers and lower lung tissue subgenomic RNA levels compared to control IgG-infused mice (Figures 5D-F). Overall, DH1052 treatment resulted in less severe disease and reduced viral replication rather than infection enhancement. Therefore, NTD antibodies that enhanced infection in vitro did not enhance infection or disease in vivo in the SARS-CoV-2 MA10 virus infection mouse model.
Next, we examined the effect of infusion of NTD antibody DH1052 on SARS-CoV-2 infection in monkeys (Leist et al., 2020b; Rockx et al., 2020). Cynomolgus macaques were infused with 10 mg of antibody per kilogram of body weight and then challenged intranasally and intratracheally with 105 plaque forming units of SARS-CoV-2 three days later (Figure 5G) (54). We compared the protective activity of in vitro infection-enhancing antibody DH1052, neutralizing NTD antibody DH1050.1, and negative control influenza antibody CH65 in separate groups of macaques. While no human antibody was detected pre-infusion (Figure 5H), antibody infusion resulted in human antibody concentrations ranging from 11 to 238 μg/mL in serum at day 2 post-challenge (Figure 5I and S26A-D). Sera with DH1050.1 neutralized SARS-CoV-2 pseudovirus at a mean ID50 titer of 19 (Figure 5J), and neutralized SARS-CoV-2 replication-competent virus at a mean ID50 titer of 192 (Figure 5K). In contrast, the presence of DH1052 or control antibody CH65 in serum did not neutralize SARS-CoV-2 (Figure 5J-K). Macaques that received CH65 or DH1052 had comparable lung inflammation four days after infection (Figures 5L, S22 and S23A). However, one macaque administered DH1052 showed increased perivascular mononuclear inflammation, perivascular and alveolar edema (Figure S23B), and multiple upregulated bronchoalveolar fluid (BAL) cytokines (Figure S24-S25) compared to either control antibody-infused animals or the four other monkeys in the DH1052-treated group. In contrast, macaques administered DH1050.1, a neutralizing NTD antibody, had lower lung inflammation than CH65-infused macaques (Figures 5L, S22 and S23A). Infusion of either neutralizing NTD DH1050.1 or in vitro infection-enhancing antibody DH1052 reduced viral nucleocapsid antigen in the lung (Figures 5M, S22 and S23A). Envelope (E) gene subgenomic RNA (sgRNA) and nucleocapsid (N) gene sgRNA in the BAL were also reduced in macaques that were administered DH1050.1 or DH1052 compared to macaques treated with negative control antibody (Figures 5N-O and S26G). In nasal swab fluid, macaques showed reduced E and N gene sgRNA when DH1050.1 or DH1052 was infused respectively, with the reduction by DH1050.1 being statistically significant (Figures 5P-Q, S26E-F and S26H-I). Thus, neutralizing NTD antibody DH1050.1 provided protection against SARS-CoV-2 infection in cynomolgus macaques, while the in vitro infection-enhancing antibody DH1052 showed no infection enhancement overall, but rather showed partial protection from SARS-CoV-2 infection.
FcγR-dependent, in vitro infection-enhancing RBD antibodies do not enhance SARS-CoV-2 infection in mice
We next tested RBD neutralizing antibodies that also mediated infection enhancement in TZM-bl cells expressing FcγRI or FcγRII in a SARS-CoV-2 acquisition mouse model (Figures 6A-B). Aged (12 months old) BALB/c mice were injected intraperitoneally with 300 μg of antibody, and challenged with a SARS-CoV-2 mouse-adapted (MA) isolate twelve hours later (Dinnon et al., 2020a). Mice received either FcγR-dependent, in vitro infection-enhancing antibody DH1041, non-infection enhancing antibody DH1050.1, or a combination of both antibodies. Administration of RBD NAb DH1041 alone or in combination with DH1050.1 protected all mice from detectable infectious virus in the lungs 48h after challenge (Figure 6A). In the setting of therapeutic treatment, administration of DH1041 alone (300 μg) or in combination with DH1050.1 (150 μg of each) twelve hours after SARS-CoV-2 challenge significantly reduced lung infectious virus titers, with half of the mice having undetectable infectious virus in the lung 48h after challenge (Figure 6B). Thus, while RBD antibody DH1041 could mediate FcγR-dependent, in vitro infection enhancement, it protected mice from SARS-CoV-2 infection when administered prophylactically or therapeutically.
DH1046 and DH1047 are RBD antibodies that neutralize SARS-CoV, SARS-CoV-2 and bat WIV1-CoV (Menachery et al., 2016) (Figure 2A, S4A-B, S5E-H and S27). Both of these RBD antibodies mediated FcγR-dependent, in vitro SARS-CoV-2 infection enhancement of TZM-bl cells that lacked ACE-2 expression (Figures 1F-L). To determine whether SARS-CoV-2 in vitro infection enhancement predicted in vivo infection enhancement by SARS-related bat coronaviruses, we assessed the ability of either DH1046 or DH1047 to enhance or protect against bat WIV1-CoV infection in HFH4-ACE2-transgenic mice. Mice were challenged either 12 hours before or 12 hours after intraperitoneal injection of antibody (Figures 6C-D). Mice administered DH1046 or DH1047 before challenge had no detectable infectious virus in the lung, whereas control IgG administered mice had a mean titer of 84,896 plaque forming units per lung lobe (Figure 6C). Administration of DH1047 after challenge eliminated detectable infectious virus in the lung in 3 of 5 mice (Figure 6D). Therapeutic administration of DH1046 reduced infectious virus titers 10-fold compared to negative control IgG (Figure 6D). Thus, FcγR-dependent in vitro infection-enhancing RBD antibodies DH1046 and DH1047 did not enhance infection in vivo, but rather protected mice from SARS-related bat coronavirus infection.
In vitro infection-enhancing RBD antibodies do not enhance SARS-CoV-2 infection in nonhuman primates
Finally, we assessed the in vivo relevance of RBD antibody infection enhancement in the cynomolgus macaque SARS-CoV-2 intranasal/intratracheal challenge model. We examined in vivo infection enhancement by RBD antibodies DH1041, DH1043, DH1046, and DH1047 that neutralized SARS-CoV-2 pseudovirus and replication-competent virus, but enhanced infection in vitro in FcγRI or FcγRIIb-expressing TZM-bl cells (Figures 1 and 6E). After antibody infusion at 10 mg of antibody per kg of macaque body weight, serum human IgG concentrations reached 11-228 μg/mL at day 2 post-challenge (Figure 6F-G and S26A-D). The same macaque serum containing the RBD antibodies exhibited a wide range of neutralization potency (ID50 titers) against SARS-CoV-2 pseudovirus or replication-competent virus, commensurate with the neutralization potency of each antibody (Figure 6H and 6I). Infusion of RBD antibody DH1041, DH1043, or DH1047 resulted in vivo protection from SARS-CoV-2 infection. In macaques administered DH1041, DH1043, or DH1047, lung inflammation was reduced and lung viral antigen was undetectable in all but one macaque (Figures 6J-K, S22 and S23A). E gene sgRNA and N gene sgRNA were significantly reduced in the upper and lower respiratory tract based on analyses of bronchoalveolar lavage fluid, nasal swabs, and nasal wash samples (Figures 6L-O and S26E-I). The RBD antibody DH1046 only protected a subset of infused monkeys. Four monkeys treated with RBD antibody DH1046 exhibited the same or lower levels of lung inflammation compared to monkeys that received control IgG (Figure 6J). One DH1046-infused monkey had an increased lung inflammation score compared to control antibody monkeys (Figures 6J, S22, and S23), but had not evidence of perivascular or alveolar edema or evidence of abnormal BAL cytokines (Figures S24-S25). Comparing the DH1046 group to the control IgG group, viral antigen in the lung was reduced overall (Figures 6K, S22 and S23A), however reductions in E gene and N gene sgRNA levels were not significant in BAL or nasal fluid at most timepoints (Figures 6L-O and S26E-I). These results further demonstrated that both FcγR-independent and FcγR-dependent in vitro SARS-CoV-2 infection enhancement did not translate to SARS-CoV-2 in vivo infection enhancement.
DISCUSSION
Antibody enhancement of virus infection in vitro has been reported for MERS-CoV and SARS-CoV (Gu et al., 2005; Liu et al., 2020a; Wan et al., 2020; Wang et al., 2016) and is a critical question facing the safe administration of antibody prophylaxis, antibody therapy, and antibody-based COVID-19 vaccines (Halstead and Katzelnick, 2020; Karthik et al., 2020; Lee et al., 2020; Ulrich et al., 2020). The effects of SARS-CoV-2 antibodies have been further questioned since individuals who seroconverted earlier tend to have more severe COVID-19 disease than later seroconverters (Ho et al., 2005). Here, we demonstrated that SARS-CoV-2 antibodies can mediate infection enhancement in vitro. However, this in vitro phenotype does not translate to enhanced SARS-CoV-2 or SARS-related virus infection in mice, nor does it usually translate to enhanced infection in a non-human primate model of SARS-CoV-2 infection. We assessed in vivo infection enhancement as severe lung immunopathology or increased respiratory tract virus replication. Neither of these hallmarks of infection enhancement was significantly present in mice. Two of 30 monkeys administered RBD or NTD antibodies that mediate infection enhancement in vitro had increased lung inflammation despite having low or undetectable lung viral antigen. One of these two macaques had lung histology and BAL cytokine levels suggestive of antibody-mediated enhanced pathology. However, these indicators of antibody-mediated enhancement were not present in all macaques that received the same antibody, thus host-specific differences may have influenced the pathology outcomes. These preclinical results indicate that SARS-CoV-2 antibody treatments or the induction of SARS-CoV-2 antibodies by vaccination have a low likelihood of exacerbating COVID-19 disease in humans.
It is of interest that a potent therapeutic Spike antibody, LY-CoV555 recently failed to demonstrate efficacy in the treatment of hospitalized patients with COVID-19 (ACTIV-3/TICO LY-CoV555 Study Group,(2020). We observed here that high levels (1,345-fold excess) of enhancing antibodies can overcome effects of neutralizing antibodies and enhance SARS-CoV-2 infection in vitro (Figure S14). It will be of interest to determine if the lack of efficacy was due to neutralizing effects of LY-CoV555 being outcompeted by infection-enhancing effects of host antibodies.
Previous studies with polyclonal serum antibodies against SARS-CoV have also shown in vitro FcγR-dependent infection enhancement, but no in vivo infection enhancement in hamsters (Kam et al., 2007). For the present study, there are two potential explanations for why there is a lack of congruence between in vitro infection enhancement assays and outcomes of passive antibody infusion/SARS-CoV-2 challenge studies. First, macrophages and other phagocytes are the target cells that uptake MERS-CoV when infection enhancement occurs (Hui et al., 2020; Wan et al., 2020; Zhou et al., 2014). In contrast, SARS-CoV and SARS-CoV-2 do not productively infect macrophages (Bournazos et al., 2020; Hui et al., 2020; Yip et al., 2016). Thus, while RBD antibodies may be able to mediate FcγR-dependent virus uptake of SARS-CoV-2 in vitro, in vivo FcγR-dependent virus uptake of SARS-CoV-2 may largely lead to abortive infection in macrophages (Cheung et al., 2005; Yip et al., 2016).
A second potential explanation for a lack of agreement between in vitro infection enhancement assays and in vivo studies is that in vivo SARS-CoV-2 antibodies may have the ability to combat SARS-CoV-2 replication through antibody effector functions. While circulating in vivo, antibodies can opsonize infected cells or virions and recruit effector immune cells to kill the virus or the infected cells through Fc-mediated mechanisms. A recent study in a SARS-CoV-2 mouse model of acquisition suggested that Fc effector functions contribute to the protective activity of SARS-CoV-2 neutralizing antibodies C104, C002, and C110 (Schafer et al., 2021). Thus, antibody effector functions may contribute to the outcome in vivo, but not be accounted for in SARS-CoV-2 neutralization assays in vitro. In support of this hypothesis, infection-enhancing, non-neutralizing NTD antibody DH1052 reduced infectious virus titers in the lungs of challenged mice compared to the negative control antibody. Consistent with previous findings for human IgG (Dekkers et al., 2017), we have demonstrated that DH1052 antibody can bind to murine FcγRs (Alam SM, Li D, Saunders KO, Haynes BF, unpublished), raising the hypothesis that Fc-mediated effector functions may be responsible for the reduction in infectious virus observed in our mouse challenge study. Future studies will investigate the Fc-mediated effector functions of DH1052 to discern their role in reducing infectious SARS-CoV-2 virus titer in vitro and in vivo.
Notably, we observed two different types of in vitro infection enhancement. First, RBD antibodies mediated classical antibody-dependent enhancement that required FcγRs for virus uptake (Lee et al., 2020). Previous studies have demonstrated that uptake of MERS-CoV or SARS-CoV has mostly been mediated by FcγRIIa on the surface of macrophages (Bournazos et al., 2020; Wan et al., 2020; Yip et al., 2016). In contrast to SARS-CoV and MERS-CoV infection enhancing antibodies, we identified SARS-CoV-2 RBD antibodies utilized FcγRIIb or FcγRI. Thus, different FcγRs can mediate SARS-CoV and MERS-CoV in vitro infection enhancement compared to SARS-CoV-2 in vitro infection enhancement. Second, non-neutralizing NTD antibodies mediated FcγR-independent infection enhancement in two different FcγR-negative, ACE-2-expressing cell types. The mechanism of this in vitro enhancement remains unclear, but one hypothesis is the possibility of antibody modulation of S protein conformation. In NSEM studies, NTD antibodies preferentially bound to S in a 3-RBD-down conformation. One study has reported that binding of select NTD antibodies to S enhances S binding to ACE2 (Liu et al., 2020b). Whether S protein has different entry kinetics or higher affinity for ACE-2 when liganded to infection-enhancing antibody DH1052 will be a focus of future studies.
There are currently over 100 COVID-19 vaccine candidates under development (Anderson et al., 2020; Corbett et al., 2020a; Corbett et al., 2020b; Folegatti et al., 2020; Jackson et al., 2020; Mulligan et al., 2020; Wrapp et al., 2020b). Both the Pfizer/BioNTech and Moderna mRNA-lipid nanoparticle (LNP) vaccine efficacy trials have completed and showed ∼95% vaccine efficacy (Jackson et al., 2020; Polack et al., 2020). The Moderna vaccine trial had 30 severe COVID-19 cases—all in the placebo group—indicating no disease enhancement was seen (www.fda.gov, 2020). It is also important to note that administration of COVID-19 convalescent sera to over 35,000 COVID-19 patients have demonstrated the treatment to be safe and is not associated with enhanced disease (Joyner et al., 2020). However, a recent study demonstrated that suboptimal neutralizing antibody level is a significant predictor of severity for SARS-CoV-2 (Garcia-Beltran et al., 2020). Thus it will be important to continue to monitor on-going COVID-19 vaccine efficacy trials for the possibility of vaccine associated enhanced disease (VAED) (Haynes et al., 2020).
Our results here indicate while SARS-CoV-2 enhancement occurred in SARS-CoV-2 infection assays in vitro, these assays did not always predict infection enhancement in mice or monkeys in vivo. Additionally, the SARS-CoV-2-infected individual from whom many of our antibodies were isolated did not progress to severe COVID-19 disease. Thus, these results suggest that antibody-induced enhancement of infection is a rare possibility but will not likely be a biologically relevant adverse effect following COVID-19 vaccination in humans.
METHODS
Detailed methods are provided in the supplemental online material.
AUTHOR CONTRIBUTIONS
D.L. designed and performed experiments, analyzed data, and wrote the manuscript. R.J.E., K.Mansouri, K.Manne, S.G., K.J., M.K., V.S. and P.A. performed structure experiments and analysis. R.J.E. and P.A. supervised structural studies and wrote the manuscript. S.M.A. and K.C. performed and analyzed the SPR work. A.S. and D.R.M. performed the mouse challenge work. D.R.M., L.V.T., T.D.S performed SARS-CoV-1 and bat CoV neutralization assays. K.W. performed bioinformatic analysis of antibody sequences. X.L. performed antibody isolation. R.P., V.G., and M.D. performed ELISA assay. L.L.S. oversaw the NHP study. C.T.D. and T.N.D oversaw the viral RNA assay. T.H.O. and G.D.S. performed SARS-CoV-2 virus neutralization assays. E.L., A.F., F.C., G.E.H., A.S., K.T. and C.J. prepared DNA and produced antibodies. L.G.P., C.M. and D.C.M. performed pseudovirus neutralization assays. M.B., T.V.H. and B.F.H. performed autoreactivity assays. C.W.W and E.P. oversaw setting up the MESSI donor cohort. A.M., E.S., and R.P. performed COVID-19 donor serology. M.A.M. oversaw the protein fluorescent labeling and flow cytometry sorting. D.W.C. designed and performed flow cytometry sorting. A.G.S., J.F., B.M.H, T.M.C., I.T., T.Z., P.D.K., J.M. and B.G. provided key reagents for this study. K.W.B., M.M., B.M.N., I.N.M., and R.S. oversaw or performed pathology experiments. H.A., and M.G.L. led and performed the non-human primate studies. R.S.B. supervised the SARS-CoV-1 and bat CoV neutralization assays and the mouse challenge studies. K.O.S. and B.F.H. conceived, designed and supervised the study, reviewed all data, and wrote the paper. All authors reviewed and commented on the manuscript.
DECLARATION OF INTERESTS
B.F.H., G.D.S. K.O.S., R.P., D.L. and X.L. have applied for patents concerning SARS-CoV-2 antibodies that are related to this work. All other authors declare no conflict of interest.
ACKNOWLEDGMENTS
We thank the COVID-19 donors enrolled in the Molecular and Epidemiological Study of Suspected Infection protocol (MESSI), the MESSI clinical support team, the DHVI Clinical Accessioning Unit Core and the DHVI Immunology and Virology Quality Assessment Core for sample procurement, processing and biobanking. We thank Joseph Gilmore, Steve Slater, Kwan-Ki Hwang, Ahmad Yousef Abuahmad, and the Duke Human Vaccine Institute Flow Cytometry Facility for flow cytometric sorting; Tyler Evangelous for purification of PCR product; Caroline Jones for production of fluorophore-labeled reagents; Kara Anasti and the Duke BMI CORE facility for SPR measurements of antibody affinities; Madison Berry and Sravani Venkatayogi for bioinformatics assistance; Paige Rawls, Lena Smith, Jemma Hwang, Beth Bryan, Jingjing Li, Haiyan Chen for protein and DNA production assistance; Nicole De Naeyer for RNA extraction assistance. We thank Drs. Matthew Gagne and Daniel C. Douek at VRC/NIH for the helpful discussions regarding the subgenomic RNA assays. We thank Kelly Soderberg, Elizabeth Donahue, Amelia Karlsson, Amanda Newman and Whitney Edwards for program management. COVID-19 donor sample processing, flow cytometric sorting, and replication-competent virus neutralization assays were performed in the Duke Regional Biocontainment Laboratory, which received partial support for construction from the National Institutes of Health, National Institute of Allergy and Infectious Diseases (UC6-AI058607). This work was supported by a contract from the State of NC funded by the Coronavirus Aid, Relief, and Economic Security Act (CARES Act); an administrative supplement to NIH R01 AI145687 for coronavirus research (P.A.); NIH grants R01AI157155 and U54CA260543 (R.S.B.); NIH NIAID U19AI142596 grant (B.F.H.); NIH R01 AI146779 and a Massachusetts Consortium on Pathogenesis Readiness (MassCPR) grant (A.G.S.); training grants: NIGMS T32 GM007753 (B.M.H. and T.M.C); T32 AI007245 (J.F.); and a cooperative agreement with DOD/DARPA (HR0011-17-2-0069; G.D.S). Cryo-EM data were collected at the National Center for Cryo-EM Access and Training (NCCAT) and the Simons Electron Microscopy Center located at the New York Structural Biology Center, supported by the NIH Common Fund Transformative High Resolution Cryo-Electron Microscopy program (U24 GM129539,) and by grants from the Simons Foundation (SF349247) and NY State. We thank Ed Eng, Carolina Hernandez and Daija Bobe for microscope alignments and assistance with cryo-EM data collection. This study utilized the computational resources offered by Duke Research Computing (http://rc.duke.edu; NIH 1S10OD018164-01) at Duke University. David R. Martinez is funded by an NIH F32 AI152296, a Burroughs Wellcome Fund Postdoctoral Enrichment Program Award, and was previously supported by an NIH NIAID T32 AI007151. We thank M. DeLong, C. Kneifel, M. Newton, V. Orlikowski, T. Milledge, and D. Lane from the Duke Office of Information Technology and Research Computing for assisting with setting up and maintaining the computing environment. We thank Laurent Pessaint, Anthony Cook, Alan Dodson, Katelyn Steingrebe and Bridget Bart at BIOQUAL for administrative and technical assistance with non-human primate studies.
Footnotes
↵* Co-First Authors