key: cord-0872056-icnbhg6e
authors: Liu, Yafei; Soh, Wai Tuck; Kishikawa, Jun-ichi; Hirose, Mika; Nakayama, Emi E.; Li, Songling; Sasai, Miwa; Suzuki, Tatsuya; Tada, Asa; Arakawa, Akemi; Matsuoka, Sumiko; Akamatsu, Kanako; Matsuda, Makoto; Ono, Chikako; Torii, Shiho; Kishida, Kazuki; Jin, Hui; Nakai, Wataru; Arase, Noriko; Nakagawa, Atsushi; Matsumoto, Maki; Nakazaki, Yukoh; Shindo, Yasuhiro; Kohyama, Masako; Tomii, Keisuke; Ohmura, Koichiro; Ohshima, Shiro; Okamoto, Toru; Yamamoto, Masahiro; Nakagami, Hironori; Matsuura, Yoshiharu; Kato, Takayuki; Okada, Masato; Standley, Daron M.; Shioda, Tatsuo; Arase, Hisashi
title: An infectivity-enhancing site on the SARS-CoV-2 spike protein targeted by antibodies
date: 2021-05-24
journal: Cell
DOI: 10.1016/j.cell.2021.05.032
sha: 1d90d76e095ddbda4015de15c72ad5ca68bf1688
doc_id: 872056
cord_uid: icnbhg6e

Antibodies against the receptor-binding-domain of the SARS-CoV-2 spike protein prevent SARS-CoV-2 infection. However, the effects of antibodies against other spike protein domains are largely unknown. Here, we screened a series of anti-spike monoclonal antibodies from COVID-19 patients, and found that some of antibodies against the N-terminal-domain (NTD) induced the open conformation of receptor binding domain (RBD) and thus enhanced the binding capacity of the spike protein to ACE2 and infectivity of SARS-CoV-2. Mutational analysis revealed that all the infectivity-enhancing antibodies recognized a specific site on the NTD. Structural analysis demonstrated that all the infectivity-enhancing antibodies bound to NTD in a similar manner. The antibodies against this infectivity-enhancing site were detected at high levels in severe patients. Moreover, we identified antibodies against the infectivity-enhancing site in uninfected donors, albeit at a lower frequency. These findings demonstrate that not only neutralizing antibodies but also enhancing antibodies are produced during SARS-CoV-2 infection.

SARS-CoV-2 is a novel coronavirus that causes coronavirus disease 2019 (Zhou et al., 2020b) . Although SARS-CoV-2 infection can result in severe symptoms and is associated with high mortality in some patients, most infected individuals do not exhibit such severe symptoms; therefore, additional factors are likely to facilitate the progression to severe COVID-19 (Tabata et al., 2020; Zhou et al., 2020a) . SARS-CoV-2, an enveloped positive-strand RNA virus, requires fusion with the host cell membrane for infection (V'Kovski et al., 2021) . The spike protein is the major envelope protein in SARS-CoV-2 and is composed of S1 and S2 subunits. The S1 subunit is further divided into an N-terminal domain (NTD) and a receptor-binding domain (RBD) (Cai et al., 2020; Wrapp et al., 2020) . The interaction between the RBD and the host cell receptor, ACE2, is responsible for SARS-CoV-2 infection of host cells (Hoffmann et al., 2020; Shang et al., 2020) .

COVID-19 patients produce antibodies against the RBD of the spike protein, blocking the SARS-CoV-2 infection (Brouwer et al., 2020; Robbiani et al., 2020; Zost et al., 2020) . Therefore, antibody production against the spike protein plays a pivotal role in host defense against SARS-CoV-2 infection (Brouwer et al., 2020; Robbiani et al., 2020; Zost et al., 2020) . However, antibodies against viruses are not always protective (Bournazos et al., 2020) . For example, antibodies against dengue virus protein can cause severe diseases mediated by the Fc receptor . In addition, feline infectious peritonitis coronavirus (FIPV) infection is exacerbated by vaccination of spike protein or adaptive transfer of the serum antibodies from FIPV-challenged animals (Vennema et al., 1990; Weiss and Scott, 1981) , suggesting the presence of antibodies that augment coronavirus infection, although the exact mechanism of the enhancement of FIPV infection by such antibodies has remained unclear (Olsen, 1993) . Therefore, understanding the function of antibodies produced during SARS-CoV-2 infection is quite important to elucidate the etiology of COVID-19. In this study, we examined various types of J o u r n a l P r e -p r o o f anti-spike antibodies, with particular emphasis on those that enhance ACE2 binding and SARS-CoV-2 infection.

We addressed the function of the antibodies against SARS-CoV-2 spike protein by generating a series of anti-spike monoclonal antibodies from COVID-19 patients (Brouwer et al., 2020; Chi et al., 2020; Robbiani et al., 2020; Zost et al., 2020) . We analyzed the effect of antibodies on the binding of recombinant ACE2 to cells expressing the full length spike protein ( Figure 1A) . In order to determine the specificity of antibodies under physiological conditions, we used transfectants expressing FLAG-tagged NTD, RBD and S2 subunit fused with the transmembrane-cytoplasmic domains of PILRα, a type I membrane protein (Saito et al., 2017; Satoh et al., 2008) (Figure S1A ). Although all of these domains were detected on the cell surface using anti-Flag antibody, cell surface expression levels of S2-TM were lower than those of NTD-TM or RBD-TM ( Figure S1B ). Some anti-spike antibodies that did not to bind to recombinant S1 spike protein in previous reports, such as 8D2 (Chi et al., 2020; Zost et al., 2020) , specifically recognized NTD expressed on the cell surface, suggesting that the antigenicity of the NTD expressed on the cell surface is different from that of the recombinant spike protein ( Figure   1B ; Figure S1B ). On the other hand, recombinant ACE2 bound to whole spike and RBD-TM transfectants but not to NTD-TM or mock transfectants, indicating that recombinant ACE2 specifically binds to the RBD of the spike protein ( Figure S2A ). As expected, most of the antibodies against the RBD domain blocked the binding of ACE2 to the spike protein, and antibodies against the S2 domain did not affect ACE2 binding ( Figure 1B) . However, a specific subset of antibodies-8D2 (Chi et al., 2020 ), 2210 , 2369 , 2490 , 2582 , and 2660 J o u r n a l P r e -p r o o f 2020)-targeting the NTD domain (denoted "enhancing antibodies") were found to enhance the binding of ACE2 to the spike protein ( Figure 1B; Figure S2A ). In contrast, other anti-NTD antibodies, such as 2016 (Zost et al., 2020) and 4A8 (Chi et al., 2020) , did not affect the binding of ACE2 to the spike protein, although they recognized the full-length spike protein and the NTD domain to a similar degree as the enhancing antibodies ( Figure 1B; Figure S1B ). The increased ACE2 binding to the spike protein in the presence of enhancing antibodies was not observed in RBD-TM or NTD-TM transfectants ( Figure S2A) . Furthermore, the increase in ACE2 binding to the spike protein in the presence of enhancing antibodies was dose-dependent ( Figure 2A) . These data suggested that antibody binding to NTD is involved in enhanced ACE2 binding to RBD. In order to analyze the relationship between antibody binding to NTD and the enhanced ACE2 binding, we compared the EC50 values of antibody binding and ACE2-binding by each enhancing antibody ( Figure S2B and S2C) . The EC50s of antibody-binding and ACE2-binding varied, suggesting the affinities of the enhancing antibodies differed. On the other hand, EC50 of ACE2-binding of each antibody was greater or equal to that of antibody-binding.

Most of recent SARS-CoV-2 viruses harbor the D614G mutation that exhibits higher infectivity than wild-type virus (Grubaugh et al., 2020; Hou et al., 2020; Korber et al., 2020; Plante et al., 2020; Yurkovetskiy et al., 2020) . Structural analysis of the mutant has revealed that the RBD is likely to have a more open conformation with the ACE2 binding interface more exposed than in wild-type spike protein Plante et al., 2020; Yurkovetskiy et al., 2020) . Therefore, we compared the effect of the enhancing antibodies with that of D614G mutation. Compatible with the findings above, the amount of recombinant ACE2 bound to the D614G spike protein was greater than that of the wild-type spike protein ( Figure 2B ), although the cell surface expressions of spike protein were comparable ( Figure S1C) . Notably, the magnitude of the effect of the enhancing antibodies on ACE2 binding to the wild-type spike protein was J o u r n a l P r e -p r o o f higher than that of the D614G mutation. Moreover, the enhancing antibodies also increased the binding of ACE2 to the D614G spike protein to a similar extent as to wild-type spike protein ( Figure 2B ).

Since anti-RBD antibodies that block the binding of ACE2 to the spike protein play a central role in antibody-mediated host defense against SARS-CoV-2 infection (Brouwer et al., 2020; Robbiani et al., 2020; Zost et al., 2020) , we analyzed how the enhancing antibodies affected the ACE2-blocking activity by neutralizing anti-RBD antibodies. We mixed the enhancing antibodies with different concentrations of a neutralizing anti-RBD antibody, C144 (Robbiani et al., 2020) .

We found that the binding of ACE2 to the spike protein increased in the presence of the enhancing antibodies, even upon addition of C144 at concentrations up to 1 µg/ml ( Figure 2C ). Similar results were obtained using other anti-RBD neutralizing antibodies, C009 and C135, that recognize epitopes different from C144 (Robbiani et al., 2020) (Figure S2D ). In addition, the enhancing effect was also observed when neutralizing antibodies were added before the enhancing antibodies ( Figure S2E ). These data suggested that the neutralization activity of the anti-RBD antibodies is indeed reduced in the presence of the enhancing antibodies.

The effect of enhancing antibodies on ACE2 binding to the spike protein suggested that the infectivity of SARS-CoV-2 would also be increased, similar to the effect of the D614G mutation (Hou et al., 2020; Korber et al., 2020; Plante et al., 2020) . To investigate this question, we utilized vesicular stomatitis virus (VSV)/∆G-GFP SARS-CoV-2 spike pseudovirus (SARS-CoV-2 PV) to quantify the effect of representative enhancing antibodies on SARS-CoV-2 infection. As expected, the enhancing antibodies increased the infectivity of SARS-CoV-2 PV to the ACE2-transfected HEK293 cells in an enhancing antibody dose-dependent manner ( Figure   J o u r n a l P r e -p r o o f 3A; Figure S1D ). In contrast, the anti-NTD antibody (4A2), which did not enhance ACE2 binding to the spike protein ( Figure 1B) , did not increase the infectivity. Moreover, the enhancement of infectivity in the presence of the antibodies was observed regardless of the amount of the virus ( Figure 3B ).

Next, we examined the effect of the enhancing antibodies using authentic SARS-CoV-2 virus.

The replication of the authentic SARS-CoV-2 virus in ACE2-transfected HEK293 cells increased more than four times in the presence of the enhancing antibodies ( Figure 3C ). SARS-CoV-2 infection of Huh7 cells, with express lower levels of ACE2, but are also susceptible to SARS-CoV-2 infection (Chu et al., 2020) , was also significantly enhanced by the enhancing antibodies ( Figure 3C ). These data indicate that the enhancing antibodies robustly augment the infectivity of the SARS-CoV-2 virus to ACE2-expressing cells.

We employed competitive binding assays to identify the epitopes of the infectivity-enhancing antibodies. Surprisingly, all the enhancing antibodies competed with each other for the spike protein binding each other, indicating that they recognize similar epitopes on the NTD ( Figure 4A ).

Next, we generated a series of alanine mutants of NTD with TM domain ( Figure S1A ) and analyzed their binding to the enhancing antibodies. Because the 8D2 antibody contained negatively charged amino acids in the heavy chain CDR3, we analyzed the binding of the 8D2 antibody to lysine-or arginine-mutated NTD. The R214A and K187A mutants were not recognized by the 8D2 antibody ( Figure 4B) . We then analyzed a series of NTD mutants in which amino acid residues structurally close to R214 or K187 were mutated to alanine, and found that the binding of the enhancing antibodies to the NTD was substantially decreased in W64A, H66A, K187A, V213A, and R214A mutants ( Figure 4B ). Similar observations were made with the full-length spike J o u r n a l P r e -p r o o f protein with mutations in these residues ( Figure 4C) . Moreover, the W64A-H66A and V213A-R214A double mutants reduced binding of some enhancing antibodies such as 2490 and 2369 more than single mutants. Furthermore, the quadruple mutant of W64, H66, V213, and R214 was not recognized by any enhancing antibody. Significantly, these mutations did not affect NTD recognition by a non-enhancing anti-NTD antibody (4A8), an anti-RBD antibody (C144) or an anti-S2 antibody (2454) (Figure S3A) , suggesting that the NTD mutations did not affect the overall structure of the spike protein. These data suggested that the binding of antibodies to a specific site on the NTD is involved in the enhanced binding of ACE2 to spike protein.

The recent B.1.1.7 variant that shows high infectivity possesses a deletion of H69 and V70, very close to the epitopes recognized by most of the enhancing antibodies. It has been reported that the H69 and V70 deletion enhances the infectivity of SARS-CoV-2 (Kemp et al., 2021) . We thus analyzed the effect of the enhancing antibodies on B.1.1.7 spike protein. Among six enhancing antibodies, all except 2369 bound to B.1.1.7 spike protein at the same level as to wild-type spike protein ( Figure S3B ). Although B.1.1.7 spike protein showed quite high ACE2-binding capacity, even in the absence of the enhancing antibodies, ACE2 binding was further augmented by the five enhancing antibodies that bound to B.1.1.7 ( Figure S3C ). These data suggest that the enhancing antibodies are functional against B.1.1.7 variants, but the effects of the enhancing antibodies might vary depending on the SARS-CoV-2 variants.

The epitopes for the infectivity-enhancing antibodies revealed by alanine mutations spanned a narrow area on the NTD surface ( Figure 5A and 5B) . Based on the observed epitopes, we generated docking models of the antibody-spike protein complexes and found that each antibody was predicted to bind similarly to the NTD ( Figure 5C ). To confirm these predictions, single J o u r n a l P r e -p r o o f particle cryo-EM analysis of apo spike protein as well as spike protein complexed with the Fab fragments of enhancing antibodies 8D2 and 2490 was employed. Data were analyzed by heterogenous refinement followed by homogenous refinement and rigid-body fitting of antibody models and known spike protein trimers ( Figure 5D ; Figure S4D ). A single apo protein structure with one RBD in the open conformation was observed. For the 8D2 Fab, five structures were obtained ( Figure S4E-S4I) , one of which was Fab-unbound and agreed well with the apo structure ( Figure S4E ), and four that contained a single 8D2 Fab fragment ( Figure 5C and 5D ). For the 2490 Fab, two structures were obtained, one containing two and one containing three bound Fab fragments, suggesting higher binding affinity for 2490 compared with 8D2 ( Figure S4J and S4K).

The binding modes of the 2490 Fab were similar to those of 8D2 ( Figure 5D ). These data suggested that structures predicted by the docking model agreed well with the cryo-EM densities.

Because all the enhancing antibodies bound to the spike protein in a similar manner, this mode of binding appears to be required for the infectivity-enhancing effect of the antibodies.

Since other anti-NTD antibodies that recognize other sites on the NTD did not enhance ACE2 binding, it is likely that binding of antibodies to a specific site on the NTD is involved in enhancing ACE2 binding to the spike protein. It has been reported that ACE2 preferentially binds to the open conformation of the RBD (Henderson et al., 2020; Lu et al., 2020) . Therefore, it is possible that the enhancing antibodies induce the open conformation of the RBD, although the open conformation of the RBD did not increase in the presence of the Fab fragments of the enhancing antibodies in Cryo-EM analysis ( Figure 5 ). To address this possibility, we analyzed the effect of the enhancing antibodies by using a unique antibody, H014, that recognizes a cryptic epitope exposed in the open RBD state (Lv et al., 2020) . In addition, we used anti-RBD neutralizing antibodies, C009, C114, J o u r n a l P r e -p r o o f and C135, which recognize different epitopes on the RBD but recognize both open and closed RBDs. (Robbiani et al., 2020) . The binding of H014 antibody was highly enhanced in the presence of the enhancing antibodies ( Figure 6A ; Figure S5A ). On the other hand, the binding of H014 antibody to spike protein was not augmented by the non-enhancing anti-NTD antibodies, 4A8 and 4A2. The enhancing antibodies did not obviously affect the binding of C009, C114 or C135 anti-RBD antibodies. Furthermore, there was a very strong correlation between the enhancement of ACE2-binding and H014 antibody-binding induced by the enhancing antibodies (R=0.91). These RBD and thus enhanced ACE2-binding similar to whole IgG antibody, whereas the Fab fragments did not. These data suggested that the binding of the monomeric enhancing antibodies to NTD alone is not enough to induce open conformation of RBD. We next analyzed whether it was possible for both arms to bind to two NTD domains on the same spike. However, we were unable to model such an binding mode using existing full-length IgG antibodies as templates. Subsequently, we investigated the possibility that a single IgG antibody bridges two spike proteins, and found that it was possible to generate such a conformation by re-modeling the V-C linker regions ( Figure 6D ).

Taken together, it appears that bridging the spike proteins by divalent antibodies is required to induce the open conformation of RBD. These data are compatible with our Cryo-EM analysis, in which the induction of the open RBD state was not observed in Fab-bound spike proteins ( Figure   5 ).

Because the enhancing antibodies boost SARS-CoV-2 infectivity, we speculated that their presence might correlate with disease severity. To investigate this possibility, we first compared the serum levels of the enhancing antibodies in COVID-19 patients to those of uninfected individuals. We utilized a competitive binding assay using recombinant spike protein and DyLight fluorescence-labeled 8D2 enhancing antibody to determine the serum levels of the enhancing antibodies ( Figure 7A ). Serum levels of the neutralizing antibodies were also determined by using fluorescence-labeled C144 anti-RBD neutralizing antibody. Binding of both enhancing antibody 8D2 and neutralizing antibody C144 to spike protein-coated beads decreased in the presence of serum from a COVID-19 patient but not from an uninfected donor ( Figure 7B) . We validated the specificity of the competitive binding assay for the enhancing antibodies by analyzing COVID-19 patient-derived anti-NTD and S2 antibodies used in Figure 1A . Although 8D2 binding was partially decreased in the presence of a few antibodies that do not enhance ACE2 binding, enhancing effects on ACE2-binding by antibodies were negatively correlated with 8D2 blocking activity (R = -0.77, p = 2.0 x 10 -12 ) ( Figure S6A ). In addition, anti-RBD antibodies did not affect 8D2 binding. Therefore, the levels of the enhancing antibodies measured by the 8D2 competition assay appeared to largely reflect serum levels of enhancing antibodies. C144 binding was also not significantly affected by most anti-NTD and S2 antibodies, although some anti-RBD antibodies blocked C144 binding as well as ACE2 binding ( Figure S6A ). On the other hand, there are several epitopes for anti-RBD neutralizing antibodies, suggesting that the C144 competition assay is not sufficient to detect all neutralizing antibodies. Therefore, we compared C144 antibody with other anti-RBD antibodies, C009 and C135, that recognize different epitopes on the RBD. C144 antibody binding to spike protein was blocked by serum antibodies of COVID-19 patients equally J o u r n a l P r e -p r o o f or more than C009 or C135 antibodies, suggesting that the competition assay using C144 Ab is more sensitive than C009 and C135 antibodies to estimate the serum levels of neutralizing antibodies ( Figure S6B ). These data suggested that the serum levels of enhancing and neutralizing antibodies could be roughly estimated by the competition assay using 8D2 and C144 antibodies, although there remains a possibility that unrelated antibodies affect the competitive binding assay.

We then measured the levels of the enhancing and neutralizing antibody titers in COVID-19 patients as well as uninfected individuals ( Figure 7C ). The enhancing and neutralizing antibodies were not detected in uninfected individuals, whereas both were detected in severe COVID-19

patients. The balance of the enhancing and neutralizing antibody titers were different depending on the patients. Because the effect of enhancing antibodies is affected by the level of neutralizing antibodies ( Figure 2C ), we quantified the overall effect by the difference in the titers (enhancing-neutralizing). The observed titer difference was higher in severe patients than in non-severe patients ( Figure 7D ). Although more detailed and extensive studies on COVID-19 patients are required, our data would suggest a correlation between enhancing antibodies and severe COVID-19.

It has been reported that a few uninfected individuals possess antibodies against the SARS-CoV-2 spike protein (Anderson et al., 2021; Ng et al., 2020) . Therefore, we next investigated whether uninfected individuals carry enhancing antibodies. Because the competitive binding assay used to analyze COVID-19 patients was not sensitive enough to detect low levels of the enhancing antibodies in uninfected individuals, we quantified levels of the enhancing antibodies by comparing the difference in serum antibody binding to wild-type NTD-transfectants with enhancing epitope deficient SARS-CoV-2 NTD-transfectants (W64A, H66A, K187A, V213A, and Figure 7E ). Interestingly, some donors, such as #13, possessed antibodies that recognized wild-type and mutant NTD to a similar degree, suggesting that they lacked enhancing antibodies ( Figure 7F ), whereas other donors, such as #18, possessed antibodies specific to wild-type NTD, suggesting that they carried enhancing antibodies. Some donors also possessed low levels of anti-RBD antibodies, but there was no correlation between the titers of the anti-RBD antibodies and the enhancing antibodies ( Figure 7G ). These data suggest that antibody responses against specific sites of the SARS-CoV-2 spike protein might vary significantly between individuals.

Antibody-dependent enhancement (ADE) of viral infection has been reported for some viruses such as dengue virus (Wan et al., 2020) , feline infectious peritonitis (FIP) coronavirus (Hohdatsu et al., 1998; Vennema et al., 1990) , SARS (Jaume et al., 2011; Kam et al., 2007) , and MERS (Wan et al., 2020) . Binding of the Fc receptor to anti-virus antibodies complexed with virions has been thought to be involved in ADE . However, Fc receptor-mediated ADE is restricted to the infection of Fc receptor-expressing cells such as monocytes or macrophages. In this study, we found a non-canonical, Fc receptor-independent ADE mechanism. The antibodies against a specific site on the NTD of the SARS-CoV-2 spike protein were found to directly augment the binding of ACE2 to the spike protein, consequently increasing SARS-CoV-2 infectivity. Although the ADE induced by the enhancing antibodies is relatively lower than the Fc-receptor mediated ADE observed in other viruses such as dengue virus, Fc receptors are not involved in this new type of ADE. Therefore, the enhancing antibodies could be involved in SARS-CoV-2 infection for a broad range of cells that do not express Fc-receptors. So far, the wild-type SARS-CoV-2 has been completely replaced by the D614G mutant SARS-CoV-2.

However, the ACE2 binding enhancement effect by the D614G mutation was smaller than that induced by the enhancing antibodies. This suggests that increased ACE2 binding capacity of the spike protein, even if small, plays an important role in SARS-CoV-2 infection.

It is interesting that all the observed enhancing antibodies recognize a specific site on the NTD.

Cryo-EM analysis and docking revealed that all of the enhancing antibodies bind to the spike protein in a very similar fashion that is distinct from anti-NTD neutralizing antibodies such as 4A8 (Chi et al., 2020) . This finding suggests that the specific binding mode is required for this non-canonical ADE. The spike protein RBD is quite flexible, and ACE2 preferentially binds to the open conformation of the RBD (Henderson et al., 2020; Lu et al., 2020) . and that antibodies must bind to a specific site on NTD to induce this state. Because ACE2 is a dimer, there is a possibility that multimerization of spike protein by antibodies could be involved in enhanced ACE2 binding simply by increasing avidity. However, multimerization of spike protein alone cannot explain why enhanced ACE2-binding is observed with specific antibodies that bind to the specific enhancing site. Furthermore, the F(ab') 2 fragments of the enhancing antibody, but not Because the affinity of monomeric Fab fragments is lower than that of divalent F(ab') 2 fragments, J o u r n a l P r e -p r o o f there is a possibility that Fab fragments did not bind sufficiently to the spike protein to induce the open RBD. However, flow cytometric analysis demonstrated that both Fab and F(ab') 2 fragments bound equally to the spike protein. Furthermore, Cryo-EM analysis revealed that 2490 Fab fragments remained attached to all three NTDs on most spike protein even after the Fab-spike complex was purified by HPLC. In addition, we were unable to generate structural models of a full-length IgG molecule with both arms bound to a single spike protein. Together, this suggested that the binding of monomeric enhancing antibodies to the NTD is not sufficient to induce the open conformation of the RBD, and that the enhancing site must be bridged by a divalent antibodies. In summary, the primary mechanism for enhancing antibodies to augment ACE2 binding seems to be the induction of an open conformation of the RBD by coupling the NTD domains of two spikes.

The RBD packs tightly with the NTD of a neighboring chain (Roy et al., 2020; Wrapp et al., 2020) . Recent experimental and computational structural studies have demonstrated that the RBD domains are held in the closed position by their interaction with neighboring NTD domains (Henderson et al., 2020; Mori et al., 2021) . Moreover, the spike proteins themselves are highly dynamic, owing to several hinges in their long stalks (Turonova et al., 2020) . It follows that, when NTDs located on different spike proteins are bridged by divalent antibodies, the intra-spike NTD-RBD interactions will be decoupled, allowing the RBD to assume the open conformation depending on the antibody binding site on the NTD. These studies are consistent with our mechanistic insights that antibodies that recognize a specific site on the NTD can induce the open RBD state. This implies that the structures around the epitopes recognized by the enhancing antibodies play an important role in controlling the conformation of the RBD. It is noteworthy that recent B.1.1.7 and B.1.135 variants contain H69-V70 deletion and D215G mutation, respectively, which are very close to the enhancing antibody binding site. Recent studies have suggested that H69-V70 deletion increases infectivity of SARS-CoV-2 (Kemp et al., 2021) . In fact, ACE2 bound J o u r n a l P r e -p r o o f the B.1.1.7 spike protein better than the wild type, even in the absence of the enhancing antibodies.

Furthermore, a comparison of closely-related viruses indicated that the NTD is much more diverse than the rest of the spike protein, implying intense molecular evolution (Saputri et al., 2020) . These observations suggest that further mutations around the epitopes of the enhancing site are expected and may affect the infectivity of SARS-CoV-2.

All the enhancing monoclonal antibodies used in this study were derived from COVID-19 patients, indicating that the enhancing antibodies are produced in COVID-19 patients. Using fluorescent-labeled enhancing antibodies, we developed a simple assay to detect enhancing antibodies in serum. With this new assay, we demonstrated that the enhancing and neutralizing antibodies are produced in both non-severe and severe patients, and the levels of enhancing and neutralizing antibodies were higher in severe patients. These observations are compatible with previous reports showing that anti-spike antibodies were produced at high levels in severe patients Lau et al., 2021) . Because the enhancing antibodies do not work when neutralizing antibodies are at high levels, the enhancing antibodies will not function in general.

However, in pre-infection or early infection, when the levels of neutralizing antibodies are still low, the levels of enhancing antibodies may affect the course of COVID-19 disease progression. In addition, a major epitope of neutralizing antibodies has been lost in the recent SARS-CoV-2 variants (Supasa et al., 2021; Wang et al., 2021) , implying that the effectiveness of the enhancing antibodies may vary among SARS-CoV-2 variants. Further large-scale clinical studies are needed to clarify the relationship between emerging variants and enhancing antibodies.

It is noteworthy that uninfected individuals possess antibodies that recognize the infectivity-enhancing site on the NTD, albeit at quite low frequency. Because the epitopes of the enhancing antibodies contain charged residues, it is possible that binding to the NTD is mediated by polyreactive antibodies. However, serum antibodies against the infectivity-enhancing site did J o u r n a l P r e -p r o o f not bind to the RBD transfectants, suggesting that the binding was specific to the infectivity-enhancing site. Recent studies showed that most anti-spike antibodies often detected in uninfected individuals are directed against the S2 subunit, which is relatively similar to that of seasonal human coronaviruses (Anderson et al., 2021; Ng et al., 2020) . The sequence homology of the NTD to common human coronaviruses is low compared to the S2 subunit. In particular, the antibody epitopes on the enhancing site are not conserved among other coronaviruses. The cause of the production of enhancing antibodies in some uninfected individuals is unknown; certain pathogens with proteins similar to the NTD of SARS-CoV-2 may be involved in the production of enhancing antibodies in some uninfected individuals. There is a possibility that the production of enhancing antibodies might be boosted by SARS-CoV-2 infection or vaccination. It follows that spike proteins that lack the epitopes containing the enhancing site would not stimulate production of enhancing antibodies in individuals with pre-existing enhancing antibodies. Transfusion of plasma from recovered COVID-19 patients has been proposed as a possible treatment for COVID-19 . However, the levels of neutralizing antibodies and enhancing antibodies in serum differ among COVID-19 patients. Therefore, plasma levels of the enhancing antibodies in the donor may affect the treatment's effectiveness. Further studies on the role of the enhancing antibodies in both COVID-19 patients and uninfected individuals would be important to understand the complicated pathogenesis of COVID-19.

In the present study, we analyzed the levels of the enhancing and neutralizing antibodies in serum by competitive binding assay using 8D2 and C144 antibodies, respectively. Although the competitive binding assay seem to reflect serum levels of the enhancing and neutralizing antibodies as we demonstrated, we cannot exclude the possibility that the levels of enhancing and J o u r n a l P r e -p r o o f neutralizing antibodies analyzed by the competitive binding assay may be affected by unrelated antibodies. Indeed, although most of the anti-spike antibodies did not affect the binding of 8D2 and C144 antibodies to spike protein, there were a few antibodies other than the enhancing and neutralizing antibodies that partially inhibited the binding of 8D2 and C144 antibodies. In addition, due to the presence of multiple epitopes for anti-RBD neutralizing antibodies, the levels of neutralizing antibodies detected by the competition assay using C144 antibody may not correctly reflect the total neutralizing antibody levels, although the C144 competitive assay was at least as sensitive as other anti-RBD antibodies. In addition, the effects of enhancing and neutralizing antibodies are expected to vary among SARS-CoV-2 variants, because the mutations on spike protein of the variants would affect the binding of these antibodies. Therefore, there is a possibility that the presence of the enhancing antibodies does not always correlate with disease outcome.

More extensive studies on enhancing antibodies, neutralizing antibodies, and infected SARS-CoV-2 variant types using a larger number of COVID-19 patient samples are required to clarify the exact function of the enhancing antibodies. 

Osaka Univ. and HuLA immune Inc. have filed a patent application on the method to detect the enhancing antibodies and the design of spike protein that does not induce the enhancing antibodies. 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Hisashi Arase (arase@biken.osaka-u.ac.jp).

J o u r n a l P r e -p r o o f Liu et al. Page 20 All materials generated in this study will be made available on request by the Lead Contact.

Cryo-EM density maps for SARS-CoV-2 spike protein complexed with the infectivity-enhancing antibodies are deposited at the EMDB under accession code EMD-30915, EMD-30916, EMD-30917, EMD-30918, EMD-30919, EMD-30920, EMD-30921 and EMD-30922, respectively. Molecular models fitted to Cryo-EM data were deposited to PDB under accession code 7DZW, 7DZX and 7DZY. The data that support the findings of this study are available from the Lead Contact on request.

and TMPRSS2-expressing VeroE6 cells (Japanese Collection of Research Bioresources Cell Bank, JCRB1819) were cultured in DMEM (Nacalai, Japan) supplemented with 10% FBS (Biological Industries, USA), penicillin (100 U/mL), and streptomycin (100 µg/mL) (Nacalai, Japan) and cultured at 37°C in 5% CO 2 . The Expi293 cells (Thermo) were cultured with the Expi293 medium. The cells were routinely checked for mycoplasma contamination.

(KNG19-020). The stock virus was amplified in TMPRSS2-expressing VeroE6 cells.

SARS-CoV-2 infection assay was carried out in a Biosafety Level 3 laboratory.

The collection and use of human sera and synovial tissues were approved by Osaka University (2020-10 and 19546), Kobe City Medical Center General Hospital (200924) , and Osaka South Hospital (2-28). Written informed consent was obtained from the participants according to the relevant guidelines of the institutional review board. The diagnoses of SARS-CoV-2 were PCR-based. The patients required for ventilator were considered as severe patients. The sera from uninfected humans were taken before June 2019 (George King Bio-Medical). All the sera of the SARS-CoV-2 patients were treated with 2% CHAPS for 30 min at room temperature to inactivate the remaining virus.

The SARS-CoV-2 spike gene (NC_045512.2) was prepared by gene synthesis (IDT). The sequences encoding the spike protein lacking C-terminal 19 amino acids (amino acids 1-1254) were cloned into pME18S expression vector. NTD (amino acids 14-333), RBD (amino acids 335-587), and S2 (amino acids 588-1219) were separately cloned into a pME18S expression vector containing a SLAM signal sequence and a PILRα transmembrane domain (Saito et al., 2017) . A series of alanine mutants and B.1.1.7 variants (del_70-69, del_144, N501Y, A570D, D614G, P681H, T716I, S982A, D1118H) were prepared from SARS-CoV-2 spike using the QuickChange mutagenesis kit or the QuickChange multi-mutagenesis kit (Agilent). The primers for mutagenesis were designed on Agilent's website (https://www.agilent.com/store/primerDesignProgram.jsp).

The cDNA encoding ACE2 (NM_021804.3) was cloned into a pMxs retrovirus vector. The mouse ACE2-IgG2a Fc fusion protein was prepared by cloning the sequence encoding the extracellular domain of Ace2 (amino acid residues 20-740) into the pCAGGS expression vector containing the J o u r n a l P r e -p r o o f SLAM signal sequence and the sequence encoding the mouse IgG2a Fc domain as previously described (Hirayasu et al., 2016) . The sequence encoding the spike protein's extracellular domain with a foldon and His-tag at the C-terminus (Cai et al., 2020) was cloned into a pcDNA3.4 expression vector containing the SLAM signal sequence. Also, mutations D614G, R686G R687S R689G, K986P, and V987P were introduced using a Quick change multi-mutagenesis kit (Agilent) for stabilization of recombinant spike protein (Yurkovetskiy et al., 2020) . The DNA sequences of these constructs were confirmed by sequencing (ABI3130xl).

A pME18S expression plasmid containing the full-length or subunit spike protein was transiently transfected into HEK293T cells using PEI max (Polysciences); the pMx-GFP expression plasmids were used as the marker of transfected cells. ACE2 was stably transfected into HEK293T cells using the pMxs retrovirus vector. Briefly, pMxs-ACE2 and amphotropic envelope vectors were co-transfected into PLAT-E packaging cells. The cell culture supernatants containing the retroviruses were harvested 48 hours later and premixed with DOTAP (Roche, Switzerland) before spin transfection at 2400 rpm and 32°C for 2 hours. Afterward, the ACE2-expressing HEK293T cells were purified using a cell sorter (SH800S, Sony).

The V regions of anti-SARS-CoV-2 spike antibodies from COVID-19 patients were synthesized according to the published sequence (IDT) (Brouwer et al., 2020; Chi et al., 2020; Robbiani et al., 2020; Zost et al., 2020) . The cDNA sequences of the variable regions of the heavy chain and light chain were cloned into a pCAGGS vector containing sequences that encode the human IgG1 or kappa constant region. The IgG with the murine IgG2a constant region was used for the J o u r n a l P r e -p r o o f competitive binding assay (Figure 4A ) or analysis of the open conformation of RBD (Figure 6A and 6B; Figure S5A ). The pCAGGS vectors containing sequences encoding the immunoglobulin heavy chain and light chain were co-transfected into HEK293T cells or Expi293 (Thermo) cells, and the cell culture supernatants were collected according to the manufacturer's protocols.

Recombinant IgG was purified from the culture supernatants using protein A Sepharose (GE healthcare) except for Figure 1B . The concentration of unpurified IgG in the cell culture supernatants used in Figure 1B was measured using the protein A-coupled latex beads (Thermo A37304) and APC-labeled anti-human IgG F(ab') 2 antibodies (Jackson) against IgG standards of known concentration. The concentration of purified IgG was measured at OD280. For competitive binding assay, protein A-purified 8D2 (Chi et al., 2020) , C144 (Robbiani et al., 2020) , C009 (Robbiani et al., 2020) and C135 (Robbiani et al., 2020) antibodies were labeled using DyLight 650 amine-reactive dye (Thermo) according to the manufacturer's protocol.

Mouse anti-human ACE2 mAb (R&D Systems, USA), rat anti-Flag mAb (L5, Biolegend), allophycocyanin (APC)-conjugated donkey anti-mouse IgG Fc fragment antibody, APC-conjugated anti-human IgG Fc fragment specific antibody, PE-conjugated anti-human IgG Fab fragment specific antibody, APC-conjugated anti-rat IgG specific antibody, and APC-conjugated streptavidin (Jackson ImmunoResearch, USA) were used. The pCAGGS vector containing a sequence that encodes the ACE2-mouse IgG2a Fc fusion protein was transfected into the HEK293T cells. Recombinant ACE2-Fc fusion protein was purified from the cell culture supernatants with the protein A Sepharose (GE Healthcare). The purified ACE2-mouse IgG2a Fc fusion protein was biotinylated with Sulfo-NHS-LC-Biotin (Thermo), followed by buffer exchange with a Zeba spin desalting column. The pcDNA3.4 expression vector containing the J o u r n a l P r e -p r o o f sequence that encodes the His-tagged extracellular domain of the spike protein was transfected into Expi293 cells; then, the His-tagged spike protein produced in the culture supernatants was purified with a Talon resin (Clontech). F(ab′) 2 fragment was prepared by incubating protein A purified 2490 antibodies with immobilized pepsin (Thermo fisher) in the 20mM sodium acetate pH 4.5 at 37℃ 4h.

Fab fragment was prepared by incubating protein A purified 8D2 or 2490 antibodies with immobized papain (100 units/ml, Sigma) in the presence of in the presence of 10 mM EDTA and 10 mM cysteine at room temperature for 3h. Immobilized pepsin or papain was precipitated by centrifugation, and the supernatant was loaded onto protein A Sepharose equilibrated with PBS.

Flow through fractions were purified with Superdex 200 Increase 10/300 GL gel filtration column using ÄKTA Pure 25 System (GE Healthcare Life Sciences). The purity of the purified F(ab′) 2 or Fab fragments were analyzed with SDS/PAGE and the gel was stained with Oriole Fluorescent dye (Bio-Rad) ( Figure S5B) . For Cryo-EM analysis, the same protein amount of Spike protein and Fab were incubated at 4℃ for 1~2 h, and the complex was separated on a Superose 6 Increase 10/300 GL column equilibrated with 20 mM Tris-HCl (pH 8.0), 0.25 M NaCl. The purified complex was used for Cryo-EM analysis.

The plasmid expressing the full-length SARS-CoV-2 spike protein, Flag-NTD-PILR-TM, Flag-RBD-PILR-TM, Flag-S2-PILR-TM, mutated spike proteins or B.1.1.7 variant spike protein were co-transfected with the GFP vector into HEK293T cells. The transfectants were incubated with the mAbs, followed by APC-conjugated anti-human IgG Ab. Then, the antibodies bound to the stained cells were analyzed using flow cytometers (Attune™, Thermo; FACSCalibur BD bioscience, LSR Fortessa X20 BD bioscience). Antibody binding to the GFP-positive cells were shown in the figures using a FlowJo software (BD bioscience).

Because SARS-CoV-2 spike protein contains intracellular retention signal at C-terminus (Hu et al., 2020; Johnson et al., 2020) , we used spike protein lacking C-terminal 19 amino acids for ACE2 binding assay. C-terminal deleted spike protein and GFP were cotransfected into HEK293T cells and the transfectants were mixed with various concentrations of anti-spike antibodies at 4°C for 30 min, followed by incubation with a biotinylated-ACE2-Fc fusion protein at 1µg / ml and 4°C for 30 min. In some experiments, enhancing antibodies and neutralizing antibodies were added simultaneously or sequentially to the spike transfectants. Thereafter, the ACE2 bound to the spike protein was detected using APC-conjugated streptavidin (2.5 µg/mL). The amount of ACE2 on GFP-positive cells was analyzed by a flow cytometer.

The HEK293T cells were transiently transfected with expression plasmids for the SARS-CoV-2 spike protein lacking C-terminal 19 amino acids (Hu et al., 2020; Johnson et al., 2020) .

Twenty-four hours post transfection, VSV-G-deficient VSV carrying a GFP gene complemented in trans with the VSV-G protein was added for a two-hour incubation. The cells were then carefully washed with DMEM media without FBS and incubated with DMEM with FBS at 37°C in 5% CO 2 for 48 hours. The supernatant containing the pseudotyped SARS-CoV-2 virions was harvested and aliquoted before storage at −80°C. The virus titers were determined using TMPRSS2-expressing 

The virus stock (6.25 log TCID50) was prediluted 1: 100 in DMEM supplemented with 2% FBS before being incubated with the cells at 1 × 10 5 cells/mL at 37°C in 5% CO 2 for 3 hours. The virus was then removed by washing the cells with DMEM; the cell culture supernatant was collected for control. The cells were incubated for another 2 days. The cell culture supernatant was collected for viral RNA extraction using the QIAamp viral RNA extraction kit (Qiagen, Germany) and subsequently quantified using real-time PCR using the N2 primer set (AGCCTCTTCTCGTTCCTCATCAC and CCGCCATTGCCAGCCATTC).

The binding of the enhancing antibodies containing a mouse IgG2a constant region (1 µg/ml) to full-length spike transfectants was analyzed in the presence of the human antibodies (10 µg/ml).

The effect of competitors on ACE2-Fc binding to the spike transfectants was also analyzed. The non-enhancing anti-S2 antibody, 2147, was used as a control. Relative mean fluorescence intensities observed in the presence of competitor antibodies were calculated.

A homology model of the SARS-CoV-2 prefusion spike trimer was built using a template-based method. The prefusion spike trimer structure (PDB ID: 7jji) in the all-RBD-down status was used as the main structural template. To model missing residues within the S2 domain, fragments were sampled from other SARS-CoV-2 spike trimer structures. The best fragment template was selected as the one with the lowest RMSD within flanking regions of the main template and the complete structure was constructed using MODELLER (Webb and Sali, 2016) . We investigated possible J o u r n a l P r e -p r o o f binding modes of the enhancing antibodies through antibody modelling followed by docking.

Antibody structures were modelled by Repertoire Builder (Schritt et al., 2019) and docked onto the NTD using the HADDOCK2.4 webserver (Ambrosetti et al., 2020) using observed epitope residues as constraints. The top-scoring docked poses were rendered on the full-length spike model using PyMOL. Residues identified by alanine scanning to affect enhancement were represented as a heatmap onto the surface of the spike model.

Spike protein and Fab complex was purified by A 2.5 µl protein solution of the spike complex (4.5mg/ml) or the complex of spike and antibody (2.5 mg/ml) was applied onto the cryo-grid and frozen in liquid ethane using a Vitrobot IV (FEI, 4°C and 100% humidity). In the case of the complex of the spike-2940 antibody, the graphen corted Quantifoil Au was used. In contrast, Quantifoil Au R0.6/1.0 holey carbon grids were used for the apo and the spike-8D2 antibody complex. Data collection of each sample was carried out on a Titan Krios (FEI, Netherland) equipped with a thermal field emission electron gun operated at 300 kV, an energy filter with a 20 eV slit width and a K3 direct electron detector camera (Gatan, USA) ( Figure S5 ). For automated data acquisition, SerialEM software was used to collect cryo-EM image data. Movie frames were recorded using the K3 camera at a calibrated magnification of × 81,000 corresponding to a pixel size of 0.88 Å with a setting defocus range from −0.8 to −2.0 µm. The data were collected with a total exposure of 3 s fractionated into 50 frames, with a total dose of ~ 50 electrons Å 2 in counting mode. A total number of movies were collected; 5,075 for the apo spike complex, 2,135 for the spike protein-8D2 antibody complex, and 5,177 for the spike protein-2940 antibody complex, respectively.

All of image processes were carried out on cryoSPARC software (Punjani et al., 2017) . After motion correction of movies and CTF parameter estimation, the particles were automatically picked by template-based particle picking algorithm. The detailed information is summarized in Table S1 . The picked particles were extracted into a box of 360 × 360 pixels for the apo spike complex, 440 × 440 pixels for the spike-8D2 complex, and 380 × 380 pixels for the spike-2940

complex. For the apo spike complex, after particle extraction, some rounds of heterogenous refinement with C1 symmetry resulted in one class (74,756 particles) that was further used as homogenous refinement with CTF refinement, and particles polishing. As the result, the density map for the apo spike complex was obtained at 3.36 Å resolution (EMDBID: 30915). For the spike-8D2 antibody complex, 643,616 particles were selected after several rounds of 2D classification. Then, the selected particles were applied to several rounds of heterogenous refinement with C1 symmetry. As the result, five distinguishable classes were obtained. Each of these classified particles (161,974, 196,524, 99,745, 74,399, and 137,974 particles) were applied to homogenous refinement with CTF refinement. The density maps from the refinement were obtained at 3.46, 3.53, 3.82, 4.16, and 3.74 Å resolution, respectively (EMDBID: 30916, 39017, 30918, 30919, and 30920). For the spike-2940 antibody complex, using extracted particles, heterogenous classification with C1 symmetry resulted in two distinguishable classes (104,069 and 33,092 particles). As the results of homogenous refinement with CTF refinement, the two maps of the spike-2940 antibody complex at 3.15 and 4.06 Å resolution (EMDBID: 30921 and 30922) was obtained. Local resolution of the obtained maps were estimated by Local resolution estimation job on cryoSPARC.

To elucidate the relative position between the spike complex and antibody, the homology models of the spike complex and antibody (Fab) were fitted into the obtained maps as rigid bodies J o u r n a l P r e -p r o o f by Chimera (Pettersen et al., 2004) . A bunch of 36 one-RBD-up trimeric spike complexes were collected from the PDB (dated as of November 5, 2020) and used as templates to construct homology models with alternative conformations as described in methods. Spike homology models were then fitted into the density maps, and the model with the highest correlation was selected. Fab models were fitted independently of the spike. Variable domains of the Fabs were modelled by Repertoire Builder (Schritt et al., 2019) and used as templates to create complete Fab models. The best matched PDB structure 5X8M was selected as the template for Fab constant domains. MODELLER (Webb and Sali, 2016) was used to build the complete Fab models from the variable and constant domain templates. After fitting the Fab models into the density maps, binding interface between spike NTD and Fab CDRs was examined, and NTD loops clashing the Fab were remodeled by the MODELLER (Webb and Sali, 2016) . Finally, refined models were deposited (PDBID: 7DZW for apo, 7DZX for the complex with 8D2, 7DZY for the complex with 2490).

Spike protein and GFP were cotransfected into HEK293T cells and the transfectants were mixed with 3 µg/mL of anti-NTD antibodies of which constant regions were replaced with mouse-IgG2a constant region at 4°C for 30 min, followed by incubation with 1 µg / mL of H014, C009, C135 and C144 anti-RBD antibodies with human IgG1 constant region at 4°C for 30 min. Thereafter, the anti-RBD antibodies bound to the spike protein were detected using APC-conjugated anti-human-IgG Fc specific antibodies that do not cross-react with mouse IgG2a (2.5 µg/mL).

The 4 µm aldehyde/sulfate latex beads (Thermo A37304) were coated with protein A-purified anti-S2 mAb (2454) (Zost et al., 2020) and then incubated with recombinant spike protein at 7 J o u r n a l P r e -p r o o f µg/ml to capture the recombinant SARS-CoV-2 spike protein on the latex beads. The latex beads containing recombinant SARS-CoV-2 spike were then incubated with 1:30 diluted serum from COVID-19 patients, uninfected individuals or anti-spike antibodies used in Figure 1B (1 µg/ml) and stained with DyLight 650-labeled 8D2 mAb at 3 µg/ml) or C144 mAb at 3 µg/ml. After fixation with 4% paraformaldehyde (Nacalai Tesque), the beads were analyzed using flow cytometry. We defined the high levels of the enhancing and neutralizing antibody titers in a COVID-19 patient's serum as 1,000 units and used the serum as a standard to calculate antibody titers of other serum samples.

The plasmids expressing the wild-type Flag-tagged NTD-TM, which was recognized by the enhancing antibodies, and the Flag tagged NTD-TM mutants (W64A, H66A, K187A,V213A, and R214A) not recognized by all enhancing antibodies were transfected into HEK293T cells with the GFP vector as a transfection marker. Cell surface expression levels of NTD on these transfectants were analyzed using anti-Flag antibody and cells expressing the same levels of NTD were used for comparion between wild-type and mutant NTD. The transfectants were mixed with 1:100 diluted serum from uninfected individuals, and the bound antibodies were detected with the APC-labeled anti-human IgG Fc antibody. The 4A8 anti-NTD antibody (Chi et al., 2020) was used as a standard to calculate the relative concentration of the antibody against wild-type NTD. The stained cells were analyzed by a flow cytometer. The level of serum antibodies specific to the wild-type or mutant NTD was calculated by subtracting the mean fluorescence intensity of the antibodies bound to the GFP-negative cells from those of the GFP-positive cells. The level of the enhancing antibodies was calculated by subtracting antibody levels against the mutant NTD from those against the wild-type NTD. Similarly, the relative level of the anti-RBD antibody was measured J o u r n a l P r e -p r o o f using RBD-TM transfectants and C144 as a standard. One µg/ml of 4A8 and C144 was defined as 1 unit.

FlowJo version 10.7 (BD Biosciences, USA) was used to analyze the flow cytometry data, and

Graphpad Prism version 7.0e was used for graph generation and statistical analysis. The binding of ACE2-Fc-fusion protein to full-length spike transfectants was analyzed in the presence of the indicated antibodies at 1 µg/ml (bottom column). The antibodies that enhanced ACE2-Fc binding to the spike transfectants by more than 1.9 times are indicated in red. See also Figure S1 and S2. (B) Relative antibody binding levels to a series of NTD mutants compared to wild-type NTD are shown. Non-enhancing anti-NTD antibody 4A8 was used as a control. The most affected residues were shown as red.

(C) The full-length mutant spike proteins were stained with the indicated enhancing antibodies (red line). Staining of wild-type spike were shown as shaded histogram. See also Figure S3 . (C) Each SARS-CoV-2 infectivity enhancing antibody was docked onto the spike protein as described in Methods.

(D) A spike model built from PDBID: 7KEB was found to fit the Apo density best, while another model built from PDBID: 7K8W fit the two other antibody-bound densities best. The spike subunit in the one RBD-"up" form is colored green. Antibodies are colored pink (heavy chain) and blue (light chain). The scale bar is 30 Å. See also Figure S4 and Table S1 . Figure 5D were used as a template for the models. See also Figure S5 . (F) The serum levels of antibodies in uninfected individuals against the wild-type NTD (blue bar) and mutant NTDs whose epitopes for the enhancing antibodies were mutated (red bar).

(G) SARS-CoV-2 infectivity-enhancing antibody titers were calculated by subtracting the antibody levels against the mutant NTD from those against the wild-type NTD in uninfected individuals (red bar). Anti-RBD antibody titers were analyzed using RBD-TM transfectants (blue bar). See also Figure S6 . Figure 1B was mixed with spike protein transfectants, and DyLight 647-labeled 8D2 or C144 antibody binding as well as ACE2-Fc binding was analyzed. Anti-NTD and S2 antibodies were plotted as black dots and anti-RBD antibodies were plotted as blue dots. Approximate line and correlation coefficient between ACE2 binding and 8D2 binding by anti-NTD and S2 antibodies were shown in the figure. (B) Comparison of competitive binding assay using C144 antibody with that using C009 and C135 antibody. Spike protein transfectants were mixed with serially diluted COVID-19 patient sera followed by the staining with DyLight 647-labeled C009, C144 and C135 anti-RBD antibodies 

A protocol for information-driven antibody-antigen modelling with the HADDOCK2

Seasonal human coronavirus antibodies are boosted upon SARS-CoV-2 infection but not associated with protection

The role of IgG Fc receptors in antibody-dependent enhancement

Distinct conformational states of SARS-CoV-2 spike protein

SARS-CoV-2 neutralizing antibody levels are correlated with severity of COVID-19 pneumonia

A neutralizing human antibody binds to the N-terminal domain of the Spike protein of SARS-CoV-2

Comparative tropism, replication kinetics, and cell damage profiling of SARS-CoV-2 and SARS-CoV with implications for clinical manifestations, transmissibility, and laboratory studies of COVID-19: an observational study

Making Sense of Mutation: What D614G Means for the COVID-19 Pandemic Remains Unclear

Controlling the SARS-CoV-2 spike glycoprotein conformation

Microbially cleaved immunoglobulins are sensed by the innate immune receptor LILRA2

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

Antibody-dependent enhancement of feline infectious peritonitis virus infection in feline alveolar macrophages and human monocyte cell line U937 by serum of cats experimentally or naturally infected with feline coronavirus

SARS-CoV-2 D614G variant exhibits efficient replication ex vivo and transmission in vivo

Development of cell-based pseudovirus entry assay to identify potential viral entry inhibitors and neutralizing antibodies against SARS-CoV-2

Anti-severe acute respiratory syndrome coronavirus spike antibodies trigger infection of human immune cells via a pH-and cysteine protease-independent FcγR pathway

Optimized Pseudotyping Conditions for the SARS-COV-2

Antibodies against trimeric S glycoprotein protect hamsters against SARS-CoV challenge despite their capacity to mediate FcγRII-dependent entry into B cells in vitro

SARS-CoV-2 evolution during treatment of chronic infection

Tracking Changes in SARS-CoV-2 Spike: Evidence that D614G Increases Infectivity of the COVID-19 Virus

Neutralizing antibody titres in SARS-CoV-2 infections

The Impact of Mutations in SARS-CoV-2 Spike on Viral Infectivity and Antigenicity

Convalescent plasma treatment of severe COVID-19: a propensity score-matched control study

Real-Time Conformational Dynamics of SARS-CoV-2 Spikes on Virus Particles

Structural basis for neutralization of SARS-CoV-2 and SARS-CoV by a potent therapeutic antibody

Elucidation of interactions regulating conformational stability and dynamics of SARS-CoV-2 S-protein

Preexisting and de novo humoral immunity to SARS-CoV-2 in humans

A review of feline infectious peritonitis virus: molecular biology, immunopathogenesis, clinical aspects, and vaccination

Spike mutation D614G alters SARS-CoV-2 fitness

cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination

Convergent antibody responses to SARS-CoV-2 in convalescent individuals

Dynamic Asymmetry Exposes 2019-nCoV Prefusion Spike

Immune evasion of Plasmodium falciparum by RIFIN via inhibitory receptors

Flexible, Functional, and Familiar: Characteristics of SARS-CoV-2 Spike Protein Evolution

PILRα is a herpes simplex virus-1 entry coreceptor that associates with glycoprotein B

Repertoire Builder: high-throughput structural modeling of B and T cell receptors

Cell entry mechanisms of SARS-CoV-2

B.1.1.7 variant by convalescent and vaccine sera

Clinical characteristics of COVID-19 in 104 people with SARS-CoV-2 infection on the Diamond Princess cruise ship: a retrospective analysis

In situ structural analysis of SARS-CoV-2 spike reveals flexibility mediated by three hinges

Coronavirus biology and replication: implications for SARS-CoV-2

Early death after feline infectious peritonitis virus challenge due to recombinant vaccinia virus immunization

Molecular Mechanism for Antibody-Dependent Enhancement of Coronavirus Entry

Antibody Resistance of SARS-CoV-2 Variants B.1.351 and B.1.1.7

IgG antibodies to dengue enhanced for FcγRIIIA binding determine disease severity

Comparative Protein Structure Modeling Using MODELLER

Antibody-mediated enhancement of disease in feline infectious peritonitis: comparisons with dengue hemorrhagic fever

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Structural and Functional Analysis of the D614G SARS-CoV-2 Spike Protein Variant

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

A pneumonia outbreak associated with a new coronavirus of probable bat origin

Rapid isolation and profiling of a diverse panel of human monoclonal antibodies targeting the SARS-CoV-2 spike protein