LAUSR

Dear Editor, The ongoing coronavirus disease 2019 (COVID-19) pandemic caused by severe acute respiratory syndrome coronavirus-2 (SARSCoV-2) has resulted in an unprecedented public health crisis, galvanizing a global effort for rapidly developing new therapeutic strategies effective against COVID-19. Human monoclonal antibodies (mAbs) are promising therapeutic molecules that can be used for the prevention or treatment of viral infectious diseases, including COVID-19. For instance, ZMapp, a cocktail consisting of three different mAbs targeting the Ebola glycoprotein is one of the most successful antibody-based therapeutic for treating infections caused by Ebola virus. Notably, this cocktail combines the bestperforming neutralizing antibodies (NAbs) screened and developed using two separate approaches, one from humanized antibodies of origin and the other from human survivors. The success of these methods indicates critical roles played by NAb diversity in the design of antibody cocktails. Cocktail therapies are not only a source of ultra-potent neutralizing activities, but they also offer advantage in overcoming potential drug resistance issues arising out of the rapid mutation of viral pathogens, in particular when selective pressure is applied. Concerningly, the emerged and rapidly spreading SARS-CoV-2 variants have arisen in the United Kingdom (UK), South Africa (SA) and other regions, such as more recently reported B.1.1.7 (UK strain, variant of concern, VOC202012/01) and 501Y.V2 (SA strain, VOC501Y.V2). These variants contain multiple mutations in their spike proteins (S), some of which are key targets of NAbs, highlighting the tremendous potential of multiple antibody-based cocktails in treating SARS-CoV-2 infection. Numerous human NAbs against SARS-CoV-2 target the receptor-binding domain (RBD) of the S, primarily blocking the interactions between the S and its receptor ACE2. Recent studies have also reported antibodies that potently neutralize SARS-CoV-2 by binding to the N-terminal domain (NTD) of the S, raising the possibility of formulation of an antibody cocktail capable of targeting both the RBD and NTD. Here we describe parallel high-throughput efforts undertaken for the generation of a large collection of highly potent humanized/human NAbs capable of binding to both the RBD and NTD using mouse and human survivors. Amongst the antibodies identified during the screening, two humanized NAbs, H014 and HB27, and one fully human NAb, P17, target the RBD and confer effective protection against SARS-CoV-2 in animal models. It is worth noting that H014 exhibited cross-neutralization activity against SARS-CoV and SARS-CoV-2, while others were SARS-CoV-2 specific. Additionally, both HB27 and P17 exert the double-lock type of mechanism of neutralization by which they block receptor attachment and interfere with viral membrane fusion. FC05, an NTD-directing NAb, was demonstrated to magnify the neutralizing potency by ~500-fold with an IC50 value up to pM level when used in combination with individual RBD-targeting NAbs. Correlated with this, the NTD of SARS-CoV-2 has also been demonstrated to be involved in entry into host cells by targeting the high-density lipoprotein (HDL) scavenger receptor B type 1 (SR-B1) and tyrosine-protein kinase receptor UFO (AXL). The significantly enhanced neutralizing potency may result from the full occupancies of RBD and NTD by RBD-directing NAbs and FC05, respectively, which almost abolishes viral attachments mediated by ACE2 or AXL or SR-B1. Although several two-antibody cocktails have been reported, threeor four-antibody cocktails have not been thoroughly characterized as yet. To explore the possibility of the formulation of a cocktail containing three or four NAbs and test its effectiveness against SARS-CoV-2, we firstly examined the simultaneous binding of the three RBD-targeting NAbs to S trimer by competitive surface plasmon resonance (SPR). The CM5 sensor labeled with SARS-CoV2 S trimer was fully saturated with one antibody and flooded with the other two antibodies in the flow through. Full saturation of HB27 was revealed to occlude the attachment of P17 to the SARS-CoV-2 S trimer and vice versa (Supplementary information, Fig. S1). Interestingly, full occupancy of HB27 blocked the binding of H014 to the SARS-CoV-2 S trimer, whereas HB27 could still attach to the S trimer in the presence of excessive H014 (Supplementary information, Fig. S1), which is line with structural observations of 0–3 bindings of H014 to the S trimer. By contrast, H014 and P17 were demonstrated to simultaneously bind to the SARS-CoV-2 S trimer (Supplementary information, Fig. S2). As expected, the binding of the NTD-targeting FC05 does not affect the interactions between any of the three RBDspecific NAbs and the SARS-CoV-2 S trimer (Supplementary information, Fig. S3). Not surprisingly, competitive binding assays verified a rational three-antibody cocktail consisting of FC05, H014 and P17 that simultaneously target three distinct regions (Fig. 1a). In most cases, binding of one NAb causes no notable changes in the binding affinities for other NAbs (Supplementary information, Fig. S4). Interestingly, partial occupancy of the epitopes with H014 yielded ~100-fold enhanced binding affinity for HB27, indicating an allosteric mechanism by which the conformational alterations caused by H014 binding might facilitate the interaction with HB27 in a synergistic manner (Fig. 1b). Theoretically, HB27 might act as a fourth partner to constitute the multi-component cocktail under the condition of the partial saturation of the RBD with H014 or P17. These results highlight the potential of these antibodies in formulating threeand four-antibody cocktails where the antibodies work cooperatively. We previously reported atomic structures of the SARS-CoV-2 S trimer in complex with individual Fab fragment (FC05, H014, HB27, P17) and characterized their epitopes. The seemingly stochastic movements of the RBD give rise to two distinct conformational states referred to as the “closed” and “open” states. The corresponding epitopes in both open and closed RBDs are conditionally accessible to HB27 and P17, but accessible to H014 only in open RBDs. Interestingly, binding of H014, but not P17 or HB27, prevents closure of the RBD, leading to the significantly increased transition of the RBD from the “closed” to “open”. To precisely visualize the cooperativity of the potential threeand four-antibody cocktails, we performed cryo-EM analysis of the SARS-CoV-2 S trimer in complex with two sets of multiple-antibody cocktails (FC05–H014–P17; FC05–H014–P17–HB27) with overall