key: cord-0684084-wtvjjc7p
authors: Wrapp, Daniel; De Vlieger, Dorien; Corbett, Kizzmekia S.; Torres, Gretel M.; Van Breedam, Wander; Roose, Kenny; van Schie, Loes; Hoffmann, Markus; Pöhlmann, Stefan; Graham, Barney S.; Callewaert, Nico; Schepens, Bert; Saelens, Xavier; McLellan, Jason S.
title: Structural Basis for Potent Neutralization of Betacoronaviruses by Single-domain Camelid Antibodies
date: 2020-03-28
journal: bioRxiv
DOI: 10.1101/2020.03.26.010165
sha: 9ad2f17028295cf26f336f4d9fb32e4417af8c00
doc_id: 684084
cord_uid: wtvjjc7p

The pathogenic Middle East respiratory syndrome coronavirus (MERS-CoV), severe acute respiratory syndrome coronavirus (SARS-CoV-1) and COVID-19 coronavirus (SARS-CoV-2) have all emerged into the human population with devastating consequences. These viruses make use of a large envelope protein called spike (S) to engage host cell receptors and catalyze membrane fusion. Because of the vital role that these S proteins play, they represent a vulnerable target for the development of therapeutics to combat these highly pathogenic coronaviruses. Here, we describe the isolation and characterization of single-domain antibodies (VHHs) from a llama immunized with prefusion-stabilized coronavirus spikes. These VHHs are capable of potently neutralizing MERS-CoV or SARS-CoV-1 S pseudotyped viruses. The crystal structures of these VHHs bound to their respective viral targets reveal two distinct epitopes, but both VHHs block receptor binding. We also show cross-reactivity between the SARS-CoV-1 S-directed VHH and SARS-CoV-2 S, and demonstrate that this cross-reactive VHH is capable of neutralizing SARS-CoV-2 S pseudotyped viruses as a bivalent human IgG Fc-fusion. These data provide a molecular basis for the neutralization of pathogenic betacoronaviruses by VHHs and suggest that these molecules may serve as useful therapeutics during coronavirus outbreaks.

The pathogenic Middle East respiratory syndrome coronavirus (MERS-CoV), severe acute respiratory syndrome coronavirus (SARS-CoV-1) and COVID-19 coronavirus have all emerged into the human population with devastating consequences. These viruses make use of a large envelope protein called spike (S) to engage host cell receptors and catalyze membrane fusion. Because of the vital role that these S proteins play, they represent a vulnerable target for the development of therapeutics to combat these highly pathogenic coronaviruses.

Here, we describe the isolation and characterization of single-domain antibodies (VHHs) from a llama immunized with prefusion-stabilized coronavirus spikes. These VHHs are capable of potently neutralizing MERS-CoV or SARS-CoV-1 S pseudotyped viruses. The crystal structures of these VHHs bound to their respective viral targets reveal two distinct epitopes, but both VHHs block receptor binding. We also show cross-reactivity between the SARS-CoV-1 S-directed VHH and SARS-CoV-2 S, and demonstrate that this cross-reactive VHH is capable of neutralizing SARS-CoV-2 S pseudotyped viruses as a bivalent human IgG Fc-fusion. These data provide a molecular basis for the neutralization of pathogenic betacoronaviruses by VHHs and suggest that these molecules may serve as useful therapeutics during coronavirus outbreaks.

Coronaviruses are enveloped, positive-sense RNA viruses that are divided into four genera (α, β, 2 γ, δ) and infect a wide variety of host organisms (Woo et al., 2009) . There are at least seven 3 coronaviruses that can cause disease in humans, and four of these viruses (HCoV-HKU1, HCoV-4 OC43, HCoV-NL63 and HCoV-229E) circulate seasonally throughout the global population, 5 causing mild respiratory disease in most patients (Gaunt et al., 2010) . The three remaining 6 viruses, SARS-CoV-1, MERS-CoV and SARS-CoV-2, are zoonotic pathogens that have caused 7 epidemics or pandemics with severe and often fatal symptoms after emerging into the human 8 population (Chan et al., 2020; Huang et al., 2020; Ksiazek et al., 2003; Lu et al., 2020; Zaki et 9 al., 2012) . For these highly pathogenic betacoronaviruses, prophylactics and therapeutic 10 treatments are needed. 11

The surfaces of coronaviruses are decorated with a spike glycoprotein (S), a large class I 12 fusion protein (Bosch et al., 2003) . The S protein forms a trimeric complex that can be 13 functionally categorized into two distinct subunits, S1 and S2, that are separated by a protease 14 cleavage site. The S1 subunit contains the receptor-binding domain (RBD), which interacts with 15 a proteinaceous host-cell receptor to trigger membrane fusion. The S2 subunit contains the fusion 16 machinery, including the hydrophobic fusion peptide and the α-helical heptad repeats. The 17 functional host cell receptors for SARS-CoV-1 and MERS-CoV are angiotensin converting 18 enzyme 2 (ACE2) and dipeptidyl peptidase 4 (DPP4), respectively (Li et al., 2003; Raj et al., 19 2013) . The interactions between these receptors and their respective RBDs have been thoroughly 20 characterized, both structurally and biophysically (Li et al., 2005; Wang et al., 2013) . Recently, it 21 has been reported that SARS-CoV-2 S also makes use of ACE2 as a functional host-cell receptor 22

and several structures of this complex have already been reported (Hoffmann, 2020; Lan, 2020; 23 Wan et al., 2020; Yan, 2020; Zhou et al., 2020) . 24

Recent advances in cryo-EM have allowed researchers to determine high-resolution 25 structures of the trimeric spike protein ectodomains and understand how S functions as a 26 macromolecular machine (Kirchdoerfer et al., 2016; Li et al., 2005; Walls et al., 2016; Wang et 27 al., 2013) . Initial cryo-EM characterization of the SARS-CoV-1 spike revealed that the RBDs 28 adopted at least two distinct conformations. In the "up" conformation, the RBDs could be 29 observed jutting out away from the rest of S, such that they could easily engage ACE2 without 30 causing any steric clashes. In the "down" conformation, the RBDs were tightly packed against 31 the top of the S2 subunit, preventing binding by ACE2 (Gui et al., 2017) . Subsequent 32 experiments have corroborated this phenomenon and similar dynamics have been observed in 33 MERS-CoV S, SARS-CoV-2 S and in alphacoronavirus S proteins (Jones et al., 2019; 34 Kirchdoerfer et al., 2018; Pallesen et al., 2017; Walls, 2020; Wrapp et al., 2020; Yuan et al., 35 2017) . Due to the relatively low abundance of particles that can be observed by cryo-EM with 36 three RBDs in the up conformation, it is thought that this conformation may correspond to an 37 energetically unstable state (Kirchdoerfer et al., 2018; Pallesen et al., 2017) . These observations 38 led to the hypothesis that the CoV RBDs may be acting as molecular ratchets, wherein a 39 receptor-binding event would trap the RBD in the less stable up conformation, leading to gradual 40 destabilization until S is finally triggered to initiate membrane fusion. Recent experiments 41 characterizing RBD-directed anti-SARS-CoV-1 antibodies that trap the SARS-CoV-1 RBD in 42 the up conformation and lead to destabilization of the prefusion spike have lent support to this 43 hypothesis (Walls et al., 2019) . 44

Numerous anti-SARS-CoV-1 RBD and anti-MERS-CoV RBD antibodies have been 45 reported and their mechanisms of neutralization can be attributed to the occlusion of the 46 receptor-binding site and to trapping the RBD in the unstable up conformation, effectively acting 47 as a receptor mimic that triggers a premature transition from the prefusion-to-postfusion 48 conformation (Hwang et al., 2006; Walls et al., 2019; Wang et al., 2018; Wang et al., 2015) . 49

Heavy chain-only antibodies (HCAbs), present in camelids, contain a single variable domain 50 (VHH) instead of two variable domains (VH and VL) that make up the equivalent antigen-51 binding fragment (Fab) of conventional IgG antibodies (Hamers-Casterman et al., 1993) . This 52 single variable domain, in the absence of an effector domain, is referred to as a single-domain 53 antibody, VHH or Nanobody® and typically can acquire affinities and specificities for antigens 54 comparable to conventional antibodies. VHHs can easily be constructed into multivalent formats 55

and are known to have enhanced thermal stability and chemostability compared to most 56 antibodies (De Vlieger et al., 2018; Dumoulin et al., 2002; Govaert et al., 2012; Laursen et al., 57 2018; van der Linden et al., 1999) . Their advantageous biophysical properties have led to the 58 evaluation of several VHHs as therapeutics against common respiratory pathogens, such as 59 respiratory syncytial virus (RSV) (Detalle et al., 2016; Rossey et al., 2017) . The use of VHHs as 60 biologics in the context of a respiratory infection is a particularly attractive application, since the 61 highly stable VHHs can be nebulized and administered via an inhaler directly to the site of 62 infection (Respaud et al., 2015) . Moreover, due to their stability after prolonged storage, VHHs 63 could be stockpiled as therapeutic treatment options in case of an epidemic. Although 64 therapeutics against MERS-CoV and SARS-CoV-2 are sorely needed, the feasibility of using 65

VHHs for this purpose has not yet been adequately explored. Several MERS-CoV S-directed 66

To map the epitopes targeted by the neutralizing VHHs, we tested binding to recombinant 114 MERS-CoV S1, RBD, NTD and SARS-CoV-1 RBD and NTD by ELISA (Figure 1A and S. 115 Figure 3 ).The MERS-CoV S-specific VHHs strongly bound to MERS-CoV S1 and RBD in a 116 concentration-dependent manner, and no binding to the MERS-CoV NTD was observed. 117

Similarly, strong binding of SARS VHH-72 to the SARS-CoV-1 RBD protein but not the SARS-118

CoV-1 NTD protein was observed. Again, no binding of SARS VHH-44 to either the SARS-119

CoV-1 RBD or NTD protein was detected. These data demonstrate that the neutralizing VHHs 120 SARS VHH-72 and MERS VHH-55 target the RBDs. Based on the specificity and potent 121 neutralizing capacity of SARS VHH-72 and MERS VHH-55, we measured the affinities of these 122

VHHs by immobilizing recombinantly expressed VHH to an SPR sensorchip and determined the 123 binding kinetics for their respective RBDs. We found that both of these VHHs bound to their 124 targets with high affinity. SARS VHH-72 bound to its target with an affinity of 1.2 nM and 125 MERS VHH-55 bound to its target with an affinity of 79.2 pM, in part due to a very slow off-126 rate constant (kd = 8.2 x10 -5 s -1 ) ( Figure 1B) . 127

To investigate the molecular determinants that mediate potent neutralization and high-affinity 129 binding by MERS VHH-55, we solved the crystal structure of MERS VHH-55 bound to the 130 MERS-CoV RBD. Crystals grew in space group C2221 and diffracted X-rays to a resolution of 131 3.4 Å. After determining a molecular replacement solution and iterative building and refinement, 132 our structure reached an Rwork/Rfree of 21.4%/26.8% ( Table 2) . The asymmetric unit of this 133 crystal contained eight copies of the MERS VHH-55 + MERS-CoV RBD complex and had a 134 solvent content of ~58%. The electron density allowed unambiguous definition of the interface 135 between the RBD and VHH, with the three CDRs forming extensive binding contacts with the 136 RBD, burying 716 Å 2 of surface area by pinching the RBD between the CDR2 and CDR3. The 137 CDR3 of MERS VHH-55 is looped over the DPP4-binding interface, occluding DPP4 from 138 productively engaging the MERS-CoV RBD (Figure 2A-B) . 139

There are numerous contacts between the CDRs of MERS VHH-55 and the MERS-CoV 140 RBD, and the majority of these are confined to CDRs 2 and 3 ( Figure 2C-D) . The sole 141 interaction from the MERS VHH-55 CDR1 comes from Asp35 forming a salt bridge with 142

Arg542 from the RBD. CDR2 forms hydrogen bonds using Ser53 and Asp61 to engage RBD 143 residues Asp539 and Gln544. Furthermore, Asn58 from the CDR2 also engages Arg542 from the 144 RBD. Trp99 from the MERS VHH-55 CDR3 forms a salt bridge with Glu513 via the nitrogen 145 from its pyrrole ring. Finally, Glu95 from the MERS VHH-55 CDR3 also forms a salt bridge 146 with Arg542 from the MERS-CoV RBD, suggesting that Arg542 plays a critical role in MERS 147 VHH-55 binding since it is productively engaged by residues from all three CDRs. This arginine 148 has also been implicated in binding to the MERS-CoV receptor DPP4, and has previously been 149 identified as one of the twelve highly conserved amino acids that is crucial for high-affinity 150 receptor engagement (S. Figure 4A ) (Wang et al., 2013; Wang et al., 2014) . 151

In addition to forming a salt bridge with Glu513 from the MERS-CoV RBD, Trp99 of the 152 MERS VHH-55 CDR3 is positioned near a hydrophobic patch formed by Phe506 (S. Figure  153   4B ). This amino acid exhibits natural sequence variation in several MERS-CoV strains, such that 154 a Leu is occasionally observed at this position. To evaluate the extent to which this substitution 155 may impact MERS VHH-55 binding, we generated a F506L substitution and measured binding 156 by SPR (S. Figure 4C) . Surprisingly, this substitution resulted in a ~200-fold reduction in 157 MERS VHH-55 binding affinity. Despite this substantial reduction, the affinity of MERS VHH-158 55 to MERS-CoV RBD F506L remains high, with a KD = 16.5 nM. Other than the variability 159 that is observed at position 506 of the MERS-CoV RBD, the rest of the MERS VHH-55 epitope 160 is highly conserved across the 863 strains that are curated in the MERS-CoV Virus Variation 161 database (S. Figure 4A ). This high degree of epitope conservation suggests that VHH-55 would 162 broadly recognize MERS-CoV strains. 163

We also sought to discover the molecular determinants of binding between SARS VHH-164 72 and the SARS-CoV-1 RBD by determining the crystal structure of this complex. Crystals 165 grew in space group P3121 and diffracted X-rays to a resolution of 2.2 Å. We obtained a 166 molecular replacement solution and refined the structure to an Rwork/Rfree of 20.3%/23.6% 167 through iterative building and refinement ( Table 2 ). Our structure reveals that CDRs 2 and 3 168 contribute to the majority of the 834 Å 2 of buried surface area at the binding interface ( Figure  169 3A). This interface does not, however, overlap with the ACE2 binding interface. Rather, ACE2 170 would clash with the CDR-distal framework of SARS VHH-72 ( Figure 3B ). This clash would 171 only be enhanced by the presence of an N-linked glycan at Asn322 of ACE2, which is already 172 located within the VHH framework when the receptor-bound RBD is aligned to the VHH-bound 173 RBD ( Figure 3C) . 174 SARS VHH-72 binds to the SARS-CoV-1 RBD by forming an extensive hydrogen-bond 175 network via its CDRs 2 and 3 (Figure3D-E). Ser56 from the CDR2 simultaneously forms 176 hydrogen bonds with the peptide backbone of three residues from the SARS-CoV-1 RBD: 177

Leu355, Tyr356 and Ser358. The peptide backbone of Ser358 also forms a hydrogen bond with 178 the backbone of neighboring Thr57 from the CDR2. A salt bridge formed between CDR2 residue 179 Asp61 and RBD residue Arg426 tethers the C-terminal end of the CDR2 to the RBD. The N-180 terminus of the CDR3 forms a short β-strand that pairs with a β-strand from the SARS-CoV-1 181 RBD to bridge the interface between these two molecules. This interaction is mediated by 182 backbone hydrogen bonds from CDR3 residues Gly98, Val100 and Val100a to RBD residues 183 Cys366 and Phe364. Glu100c from the CDR3 forms hydrogen bonds with the sidechain 184 hydroxyls from both Ser362 and Tyr494 from the SARS-CoV-1 RBD. The neighboring CDR3 185 residue also engages in a sidechain-specific interaction by forming a salt bridge between the 186 pyrrole nitrogen of Trp100d and the hydroxyl group from RBD residue Thr363. Asp101 is 187 involved in the most C-terminal interaction from the CDR3 by forming a salt bridge with RBD 188 residue Lys365. The extensive interactions formed between CDRs 2 and 3 of SARS VHH-72 189 and the SARS-CoV-1 RBD explain the high-affinity binding that we observed between these 190 molecules. 191

Analysis of 10 available SARS-CoV-1 strain sequences revealed a high degree of conservation 193 in the residues that make up the SARS VHH-72 epitope, prompting us to explore the breadth of 194 SARS VHH-72 binding (S. Figure 5A ). WIV1-CoV is a betacoronavirus found in bats that is 195 closely related to SARS-CoV-1 and also utilizes ACE2 as a host-cell receptor (Ge et al., 2013). 196 Due to the relatively high degree of sequence conservation between SARS-CoV and WIV1- CoV, 197 we expressed the WIV1-CoV RBD and measured binding to SARS VHH-72 by SPR (S. Figure  198 5B). SARS VHH-72 also exhibits high-affinity binding to the WIV1-CoV RBD (7.4 nM), 199 demonstrating that it is cross-reactive between these two closely related coronaviruses (S. Figure  200 

Based on the high degree of structural homology that has been reported between SARS-202

CoV-1 S and SARS-CoV-2 S (Walls, 2020; Wrapp et al., 2020) , we also tested SARS VHH-72 203 for cross-reactivity against the SARS-CoV-2 RBD-SD1 by SPR (Figure 4) . The binding affinity 204 of SARS VHH-72 for the SARS-CoV-2 RBD-SD1 was ~39 nM. This diminished binding 205 affinity, compared to the binding of SARS VHH-72 to SARS-CoV-1 RBD, can primarily be 206 attributed to an increase in the dissociation rate constant of this interaction ( Figure 4A ). The 207 only variant residue on the SARS-CoV-1 RBD that makes direct contact with SARS VHH-72 is 208

Arg426, which is Asn439 in the SARS-CoV-2 RBD ( Figure 3C) . 209

As stated previously, the RBDs of MERS-CoV S, SARS-CoV-1 S and SARS-CoV-2 S undergo 211 dynamic conformational rearrangements that alternately mask and present their receptor-binding 212 interfaces and potential neutralizing epitopes to host molecules. By aligning the crystal structures and SARS VHH-72, we can conclude that these molecules would likely disrupt the RBD 230 dynamics in the context of a full-length S protein by trapping the up conformation. Because this 231 up conformation is unstable and leads to S protein triggering, it is possible that this 232 conformational trapping may at least partially contribute to the neutralization mechanisms of 233 these VHHs. 234

Based on our structural analysis, we hypothesized that another mechanism by which both 235 MERS VHH-55 and SARS VHH-72 neutralize their respective viral targets is by blocking the 236 interaction between the RBDs and their host-cell receptors. To test this hypothesis, we performed 237 a BLI-based assay in which the SARS-CoV-1, SARS-CoV-2 and MERS-CoV RBDs were 238 immobilized to biosensor tips, dipped into VHHs and then dipped into wells containing the 239 recombinant, soluble host cell receptors. We found that when tips coated in the MERS-CoV 240 RBD were dipped into MERS VHH55 before being dipped into DPP4, there was no increase in 241 response that could be attributed to receptor binding. When tips coated with the MERS-CoV 242 RBD were dipped into SARS VHH-72 and then DPP4, a robust response signal was observed, as 243 expected. Similar results were observed when the analogous experiments were performed using 244 the SARS-CoV-1 or SARS-CoV-2 RBDs, SARS VHH-72 and ACE2 ( Figure 5D ). These results 245 support our structural analysis that both MERS VHH-55 and SARS VHH-72 are capable of 246 neutralizing their respective viral targets by directly preventing host-cell receptor binding. 247

Despite the relatively high-affinity binding that was observed by SPR between SARS VHH-72 249 and the SARS-CoV-2 RBD, this binding could not be detected by ELISA nor was SARS VHH-250 72 capable of neutralizing SARS-CoV-2 S VSV pseudoviruses, possibly due to the high off-rate 251 constant, whereas SARS-CoV-1 pseudotypes were readily neutralized ( Figure 6A-D) . In an 252 attempt to overcome this rapid dissociation, we engineered two bivalent variants of SARS VHH-253 72. These included a tail-to-head fusion of two SARS VHH-72 molecules connected by a 254 Li et al., 2015; Wang et al., 2018; Wang et 286 al., 2015; Ying et al., 2015; Yu et al., 2015) (S. Figure 7A ). The epitope of SARS VHH-72 does 287 not significantly overlap with the epitopes of any previously described antibodies other than that 288 of the recently described CR3022, which is also capable of binding to the RBDs of both SARS-289

CoV-1 and SARS-CoV-2 S (Hwang et al., 2006; Pak et al., 2009; Prabakaran et al., 2006; Walls 290 et al., 2019; Yuan M, 2020) Figure 7B) . However, unlike SARS VHH-72, CR3022 does not 291 prevent the binding of ACE2 and it lacks neutralizing activity against SARS-CoV-2 (Tian et al., 292 2020; Yuan M, 2020) . Because SARS VHH-72 binds with nanomolar affinity to a portion of the 293 SARS-CoV-1 S RBD that exhibits low sequence variation, as demonstrated by its cross-294 reactivity toward the WIV1-CoV and SARS-CoV-2 RBDs, it may broadly bind all S proteins 295 from SARS-CoV-like viruses. We show that by engineering a bivalent VHH-72-Fc construct, we 296 are able to compensate for the relatively high off-rate constant of the monovalent SARS VHH-297 72. This bivalent molecule expresses well in transiently transfected ExpiCHO cells (~300 mg/L) 298 and is capable of potently neutralizing SARS-CoV-2 S pseudoviruses in vitro. 299

Due to the inherent thermostability and chemostability of VHHs, they have been 300 investigated as potential therapeutics against a number of diseases. Several HIV-and influenza-301 directed VHHs have been reported previously, and there are multiple RSV-directed VHHs that 302 have been evaluated (Detalle et al., 2016; Ibanez et al., 2011; Koch et al., 2017; Rossey et al., 303 2017) . The possibility of administering these molecules via a nebulized spray is particularly 304 attractive in the case of respiratory pathogens because the VHHs could theoretically be inhaled 305 directly to the site of infection in an effort to maximize bioavailability and function (Larios Mora 306 et al., 2018) . Due to the current lack of treatments for MERS, SARS and COVID-19 and the 307 devastating effects associated with pandemic coronavirus outbreaks, both prophylactic and 308 therapeutic interventions are sorely needed. It is our hope that due to their favorable biophysical 309

properties and their potent neutralization capacity, MERS VHH-55, SARS VHH-72 and VHH-310 72-Fc may serve as both useful reagents for researchers and as potential therapeutic candidates. 311

We thank members of the McLellan Laboratory for providing helpful comments on the 313 manuscript. We would like to thank Dr. John Ludes-Meyers for assistance with cell transfection 314 and protein production. This work was supported by a National Institutes of Health 315 were detected with anti-HA (1/2,000, MMS-101P Biolegend) mAb followed by horseradish 410 peroxidase (HRP)-linked anti-mouse IgG (1/2,000, NXA931, GE Healthcare). Periplasmic 411 fractions, for which the OD450 value of the antigen coated wells were at least two times higher than 412 the OD450 value of the BSA coated wells, were considered to be specific for the coated antigen and 413 selected for sequencing. The selected clones were grown in 3 mL of LB medium with 100 μg/mL 414 ampicillin. The DNA of the selected colonies was isolated using the QIAprep Spin Miniprep kit 415 (Qiagen) and sequenced using the MP057 primer (5'-TTATGCTTCCGGCTCGTATG-3'). 416 417

In order to express the MERS-and SARS-CoV VHHs in Pichia pastoris, the VHH encoding 419 sequences were cloned in the pKai61 expression vector (Schoonooghe et al., 2009 ). In the vector, 420 the VHH sequences contain a C-terminal 6x His-tag, are under the control of the methanol 421 inducible AOX1 promotor and in frame with a modified version of the S.cerevisae α-mating factor 422 prepro signal sequence. The vector contains a Zeocine resistance marker for selection in bacteria 423 as well as in yeast cells. The VHH encoding sequences were amplified by PCR using the following 424 forward and reverse primer (5'-425

and 427 cloned between the XhoI and SpeI sites in the pKai61 vector. The vectors were linearized by PmeI 428 and transformed in the Pichia pastoris strain GS115 using the condensed transformation protocol 429 described by Lin-Cereghino et al (Lin-Cereghino et al., 2005) . After transformation, the yeast cells 430

were plated on YPD plates (1% (w/v) yeast extract, 2% (w/v) peptone, 2% (w/v) dextrose and 2% 431 (w/v) agar) supplemented with zeocin (100 µg/mL) for selection. 432 433

To generate bivalent tandem tail-to-head VHH constructs, the VHH sequence was amplified by 435 PCR using the following forward (5'-GGGGTATCTCTCGAGAAAAGGCAGGTGC 

The transformed Pichia pastoris clones were first expressed in 2 mL cultures. On day 1, 4 clones 447 of each construct were inoculated in 2 mL of YPNG medium (2% pepton, 1% Bacto yeast extract, 448 1.34% YNB, 0.1 M potassium phosphate pH 6, 0.00004% biotin, 1% glycerol) with 100 μg/mL 449 Zeocin (Life Technologies) and incubated while shaking at 28 °C for 24 hours. The next day, the 450 cells were pelleted by centrifugation and the medium was replaced by YPNM medium (2% pepton, 451 

Pseudovirus neutralization assay methods have been previously described (Pallesen et al., 2017; 488 Wang et al., 2015) . Briefly, pseudoviruses expressing spike genes for MERS-CoV England1 489 (GenBank ID: AFY13307) and SARS-CoV-1 Urbani (GenBank ID: AAP13441.1) were 490 produced by co-transfection of plasmids encoding a luciferase reporter, lentivirus backbone, and 491 spike genes in 293T cells . Serial dilutions of VHHs were mixed with 492 pseudoviruses, incubated for 30 min at room temperature, and then added to previously-plated 493

Huh7.5 cells. 72 hours later, cells were lysed, and relative luciferase activity was measured. 494

cells transduced with only pseudovirus as 0% neutralization. IC50 titers were determined based 496 on sigmoidal nonlinear regression. 497

CoV-2 S and coding for GFP or firefly luciferase were generated as described previously (Berger 499 Rentsch and Zimmer, 2011; Hoffmann, 2020) . For the VSV pseudotype neutralization 500 experiments, the pseudoviruses were incubated for 30 min at 37 °C with different dilutions of 501 purified VHHs or with dilution series of culture supernatant of 293S cells that had been 502 transfected with plasmids coding for SARS VHH-72 fused to human IgG1 Fc (VHH-72-Fc) or 503 with GFP-binding protein (GBP: a VHH specific for GFP). The incubated pseudoviruses were 504 subsequently added to confluent monolayers of Vero E6 cells. Sixteen hours later, the 505 transduction efficiency was quantified by measuring the firefly luciferase activity in cell lysates 506 using the firefly luciferase substrate of the dual-luciferase reporter assay system (Promega) and a 507

Glowmax plate luminometer (Promega). 508 509

Mammalian expression plasmids encoding for SARS VHH72, MERS VHH55, residues 367-589 511 of MERS-CoV S (England1 strain), residues 320-502 of SARS-CoV-1 S (Tor2 strain), residues 512 307-510 of WIV1-CoV S, residues 319-591 of SARS-CoV-2 S, residues 1-281 of SARS-CoV-1 513 S (Tor2 strain), residues 1-351 of MERS-CoV S (England1 strain), residues 1-751 of MERS-514

CoV S (England1 strain), residues 1-1190 of SARS-CoV-1 S (Tor2 strain) with K968P and 515 V969P substitutions (SARS-CoV-1 S-2P), residues 1-1291 of MERS-CoV S (England1 strain) 516 with V1060P and L1061P substitutions (MERS S-2P), residues 1-1208 of SARS-CoV-2 S with 517 K986P and V987P substitutions (SARS-CoV-2 S-2P), residues 1-615 of ACE2 and residues 40-518 766 of DPP4 were transfected into FreeStyle293 cells using polyethylenimine (PEI). All of these 519 plasmids contained N-terminal signal sequences to ensure secretion into the cell supernatant. using PEI. Briefly, suspension-adapted and serum-free HEK 293S cells were seeded at 3 x 10 6 533 cells/mL in Freestyle-293 medium (ThermoFisher Scientific). Next, 4.5 µg of pcDNA3.3-534 VHH72-Fc plasmid DNA was added to the cells and incubated on a shaking platform at 37 °C 535 and 8% CO2, for 5 min. Next, 9 µg of PEI was added to the cultures, and cells were further 536 incubated for 5 h, after which an equal culture volume of Ex-Cell-293 (Sigma) was added to the 537 cells. Transfections were incubated for 4 days, after which cells were pelleted (10', 300g) and 538 supernatants were filtered before further use. The resulting data were double-reference subtracted and fit to a 1:1 binding model using the 561 Biacore X100 Evaluation software. 562 563

Plasmids encoding for MERS VHH-55 and residues 367-589 of MERS-CoV S with a C-terminal 565 HRV3C cleavage site and a monomeric human Fc tag were co-transfected into kifunensin-treated 566

FreeStyle 293F cells, as described above. After purifying the cell supernatant with Protein A 567 resin, the immobilized complex was treated with HRV3C protease and Endoglycosidase H to 568 remove both tags and glycans. The complex was then purified using a Superdex 75 column in 2 569 mM Tris pH 8.0, 200 mM NaCl and 0.02% NaN3. The purified complex was then concentrated 570 to 5.0 mg/mL and used to prepare hanging-drop crystallization trays. Crystals grown in 1.0 M 571 Na/K phosphate pH 7.5 were soaked in mother liquor supplemented with 20% ethylene glycol 572 and frozen in liquid nitrogen. Diffraction data were collected to a resolution of 3.40 Å at the SBC 573 beamline 19-ID (APS, Argonne National Laboratory) 574

Plasmids encoding for SARS VHH-72 and residues 320-502 of SARS-CoV-1 S with a C-575 terminal HRV3C cleavage site and a monomeric human Fc tag were co-transfected into 576 kifunensin-treated FreeStyle 293F cells, as described above. After purifying the cell supernatant 577

with Protein A resin, the immobilized complex was treated with HRV3C protease and 578

Endoglycosidase H to remove both tags and glycans. The processed complex was subjected to 579 size-exclusion chromatography using a Superdex 75 column in 2 mM Tris pH 8.0, 200 mM NaCl 580 and 0.02% NaN3. The purified complex was then concentrated to 10.0 mg/mL and used to 581 prepare hanging-drop crystallization trays. Crystals grown in 0.1 M Tris pH 8.5, 0.2 M LiSO4, 582 0.1 M LiCl and 8% PEG 8000 were soaked in mother liquor supplemented with 20% glycerol 583 and frozen in liquid nitrogen. Diffraction data were collected to a resolution of 2.20 Å at the SBC 584 beamline 19-ID (APS, Argonne National Laboratory) 585 586

Diffraction data for both complexes were indexed and integrated using iMOSFLM before being 588 scaled in AIMLESS (Battye et al., 2011; Evans and Murshudov, 2013) . The SARS-CoV-1 589 RBD+SARS VHH-72 dataset was phased by molecular replacement in PhaserMR using 590 coordinates from PDBs 2AJF and 5F1O as search ensembles (McCoy, 2007) . The MERS-CoV 591 RBD+MERS VHH-55 dataset was also phased by molecular replacement in PhaserMR using 592 coordinates from PDBs 4L72 and 5F1O as search ensembles. The resulting molecular 593 replacement solutions were iteratively rebuilt and refined using Coot, ISOLDE and Phenix 594 (Adams et al., 2002; Croll, 2018; Emsley and Cowtan, 2004) . Crystallographic software 595 packages were curated by SBGrid (Morin et al., 2013) . 596 597

Anti-human capture (AHC) tips (FortéBio) were soaked in running buffer composed of 10 mM 599 HEPES pH 7.5, 150 mM NaCl, 3 mM EDTA, 0.005% Tween 20 and 1 mg/mL BSA for 20 min 600 before being used to capture either Fc-tagged SARS-CoV-1 RBD, Fc-tagged SARS-CoV-2 601 RBD-SD1 or Fc-tagged MERS-CoV RBD to a level of 0.8 nm in an Octet RED96 (FortéBio). 602

Tips were then dipped into either 100 nM MERS VHH-55 or 100 nM SARS VHH-72. Tips were 603 next dipped into wells containing either 1 µM ACE2 or 100 nM DPP4 supplemented with the 604 nanobody that the tip had already been dipped into to ensure continued saturation. Data were 605 reference-subtracted and aligned to each other in Octet Data Analysis software v11.1 (FortéBio) 606 based on a baseline measurement that was taken before being dipped into the final set of wells 607 that contained either ACE2 or DPP4. and incubated at room temperature for 20 min before an additional 10 min incubation on ice. 631 Vero E6 cells grown at sub-confluency were detached by cell dissociation buffer (Sigma) and 632 trypsin treatment. After washing once with PBS the cells were blocked with 1% BSA in PBS on 633 ice. All remaining steps were also performed on ice. The mixtures containing RBD and tail-to-634 head bivalent VHHs or VHH-Fc fusions were added to the cells and incubated for one hour. 635

Subsequently, the cells were washed 3 times with PBS containing 0.5% BSA and stained with an 636 AF647 conjugated donkey anti-mouse IgG antibody (Invitrogen) for 1 hour. Following 637 additional 3 washes with PBS containing 0.5% BSA, the cells were analyzed by flow cytometry 638 using an BD LSRII flow cytometer (BD Biosciences). 639 shown as dark blue ribbons and the SARS-CoV-1 RBD is shown as a pink-colored molecular surface. The ACE2 binding interface on the SARS-CoV-1 RBD is colored red. B) The structure of ACE2 bound to the SARS-CoV-1 RBD (PDB: 2AJF) is aligned to the crystal structure of SARS VHH-72 bound to the SARS-CoV-1 RBD. ACE2 is shown as a red, transparent molecular surface. C) A simulated N-linked glycan containing an energy-minimized trimannosyl core (derived from PDB ID: 1HD4) is modeled as red sticks, coming from Asn322 in ACE2. ACE2 is shown as a red molecular surface, the SARS-CoV-1 RBD is shown as pink ribbons and SARS VHH-72 is shown as a dark blue, transparent molecular surface. D) A zoomed-in view of the panel from 3A is shown, with the SARS-CoV-1 RBD now displayed as pinkcolored ribbons. Residues that form interactions are shown as sticks, with nitrogen atoms colored dark blue and oxygen atoms colored red. Hydrogen bonds and salt bridges between SARS VHH-72 and the SARS-CoV-1 RBD are shown as black dots. E) The same view from 3D has been turned by 60°to show additional contacts. Residues that form interactions are shown as sticks, with nitrogen atoms colored dark blue and oxygen atoms colored red. Interactions between SARS VHH-72 and the SARS-CoV-1 RBD are shown as black dots. The MERS-CoV spike (PDB ID: 5W9H) is shown as a molecular surface, with each monomer colored either white, gray or tan. The tan and white monomers are bound by MERS VHH-55, shown as blue ribbons. The clash between MERS VHH-55 bound to the white monomer and the neighboring tan RBD is highlighted by the red ellipse. B) The SARS-CoV-1 spike (PDB ID: 5X58) is shown as a molecular surface, with each protomer colored either white, gray or pink. Every monomer is bound by a copy of SARS VHH-72, shown as dark blue ribbons. The clashes between copies of SARS VHH-72 and the two neighboring spike monomers are highlighted by the red circle. C) The SARS-CoV-2 spike (PDB ID: 6VXX) is shown as a molecular surface, with each protomer colored either white, gray or green. Every monomer is bound by a copy of SARS VHH-72, shown as dark blue ribbons. The clashes between copies of SARS VHH-72 and the two neighboring spike monomers are highlighted by the red circle. The SARS-CoV-2 trimer appears smaller than SARS-CoV-1 S due to the absence of flexible NTD-distal loops which could not be built during cryo-EM analysis. D) CoV VHHs prevent MERS-CoV RBD, SARS-CoV-1 RBD and SARS-CoV-2 RBD-SD1 from interacting with their receptors. The results of the BLI-based receptor-blocking experiment are shown. The legend lists the immobilized RBDs and the VHHs or receptors that correspond to each curve.

: VHH-72-Fc neutralizes SARS-CoV-2 S pseudoviruses. A) BLI sensorgram measuring apparent binding affinity of VHH-72-Fc to immobilized SARS-CoV-2 RBD-SD1. Binding curves are colored black, buffer-only blanks are colored gray and the fit of the data to a 1:1 binding curve is colored red. B) Time course analysis of VHH-72-Fc expression in ExpiCHO cells. Cell culture supernatants of transiently transfected ExpiCHO cells were removed on days 3-7 after transfection (or until cell viability dropped below 75%), as indicated. Two control mAbs were included for comparison, along with the indicated amounts of purified GBP-Fc as a loading control. C) SARS-CoV-2 S pseudotyped VSV neutralization assay. Monolayers of Vero E6 cells were infected with pseudoviruses that had been pre-incubated with the mixtures indicated by the legend. The VHH-72-Fc used in this assay was purified after expression in ExpiCHO cells (n = 4). VHH-23-Fc is an irrelevant control VHH-Fc (n = 3). NI cells were not infected. Luciferase activity is reported in counts per second (c.p.s.) ± SEM. 

PHENIX: building new software for automated crystallographic structure determination

iMOSFLM: a new graphical interface for diffraction-image processing with MOSFLM

A vesicular stomatitis virus replicon-based bioassay for the rapid and sensitive determination of multi-species type I interferon

The coronavirus spike protein is a class I virus fusion protein: structural and functional characterization of the fusion core complex

A familial cluster of pneumonia associated with the 2019 novel coronavirus indicating person-to-person transmission: a study of a family cluster

Human Neutralizing Monoclonal Antibody Inhibition of Middle East Respiratory Syndrome Coronavirus Replication in the Common Marmoset

ISOLDE: a physically realistic environment for model building into lowresolution electron-density maps

Single-Domain Antibodies and Their Formatting to Combat Viral Infections

Generation and Characterization of ALX-0171, a Potent Novel Therapeutic Nanobody for the Treatment of Respiratory Syncytial Virus Infection

Single-domain antibody fragments with high conformational stability

Coot: model-building tools for molecular graphics

How good are my data and what is the resolution?

Epidemiology and clinical presentations of the four human coronaviruses 229E, HKU1, NL63, and OC43 detected over 3 years using a novel multiplex real-time PCR method

Isolation and characterization of a bat SARS-like coronavirus that uses the ACE2 receptor

Dual beneficial effect of interloop disulfide bond for single domain antibody fragments

Cryo-electron microscopy structures of the SARS-CoV spike glycoprotein reveal a prerequisite conformational state for receptor binding

Naturally occurring antibodies devoid of light chains

The novel coronavirus 2019 (2019-nCoV) uses the SARS-coronavirus receptor ACE2 and the cellular protease TMPRSS2 for entry into target cells

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Structural basis of neutralization by a human anti-severe acute respiratory syndrome spike protein antibody, 80R

Nanobodies with in vitro neutralizing activity protect mice against H5N1 influenza virus infection

Iterative screen optimization maximizes the efficiency of macromolecular crystallization

Pre-fusion structure of a human coronavirus spike protein

Stabilized coronavirus spikes are resistant to conformational changes induced by receptor recognition or proteolysis

Selection of nanobodies with broad neutralizing potential against primary HIV-1 strains using soluble subtype C gp140 envelope trimers

A novel coronavirus associated with severe acute respiratory syndrome

Crystal structure of the 2019-nCoV spike receptor-binding domain bound with the ACE2 receptor

Delivery of ALX-0171 by inhalation greatly reduces respiratory syncytial virus disease in newborn lambs

Universal protection against influenza infection by a multidomain antibody to influenza hemagglutinin

Structure of SARS coronavirus spike receptorbinding domain complexed with receptor

Angiotensin-converting enzyme 2 is a functional receptor for the SARS coronavirus

A humanized neutralizing antibody against MERS-CoV targeting the receptor-binding domain of the spike protein

Condensed protocol for competent cell preparation and transformation of the methylotrophic yeast Pichia pastoris

Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding

Solving structures of protein complexes by molecular replacement with Phaser

Structure-based design of a fusion glycoprotein vaccine for respiratory syncytial virus

Collaboration gets the most out of software

Structural insights into immune recognition of the severe acute respiratory syndrome coronavirus S protein receptor binding domain

Immunogenicity and structures of a rationally designed prefusion MERS-CoV spike antigen

Structure of severe acute respiratory syndrome coronavirus receptor-binding domain complexed with neutralizing antibody

Dipeptidyl peptidase 4 is a functional receptor for the emerging human coronavirus-EMC

Nebulization as a delivery method for mAbs in respiratory diseases

Potent single-domain antibodies that arrest respiratory syncytial virus fusion protein in its prefusion state

Efficient production of human bivalent and trivalent anti-MUC1 Fab-scFv antibodies in Pichia pastoris

Chimeric camel/human heavy-chain antibodies protect against MERS-CoV infection

Potent binding of 2019 novel coronavirus spike protein by a SARS coronavirus-specific human monoclonal antibody

Comparison of physical chemical properties of llama VHH antibody fragments and mouse monoclonal antibodies

Cryo-electron microscopy structure of a coronavirus spike glycoprotein trimer

Unexpected Receptor Functional Mimicry Elucidates Activation of Coronavirus Fusion

Structure, function and antigenicity of the SARS-CoV-2 spike glycoprotein

Receptor recognition by novel coronavirus from Wuhan: An analysis based on decade-long structural studies of SARS

Importance of Neutralizing Monoclonal Antibodies Targeting Multiple Antigenic Sites on the Middle East Respiratory Syndrome Coronavirus Spike Glycoprotein To Avoid Neutralization Escape

Evaluation of candidate vaccine approaches for MERS-CoV

Structure of MERS-CoV spike receptor-binding domain complexed with human receptor DPP4

Bat origins of MERS-CoV supported by bat coronavirus HKU4 usage of human receptor CD26

Coronavirus diversity, phylogeny and interspecies jumping

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

Structural basis for the recognition of the SARS-CoV-2 by full-length human ACE2

Junctional and allele-specific residues are critical for MERS-CoV neutralization by an exceptionally potent germline-like antibody

Structural basis for the neutralization of MERS-CoV by a human monoclonal antibody MERS-27

A highly conserved cryptic epitope in the receptor-binding domains of SARS-CoV-2 and SARS-CoV

Cryo-EM structures of MERS-CoV and SARS-CoV spike glycoproteins reveal the dynamic receptor binding domains

Isolation of a novel coronavirus from a man with pneumonia in Saudi Arabia

A Novel Nanobody Targeting Middle East Respiratory Syndrome Coronavirus

Receptor-Binding Domain Has Potent Cross-Neutralizing Activity and Protective Efficacy against MERS-CoV

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

Figure 6: Bivalency overcomes the high off-rate constant of SARS VHH-72. A) SARS-CoV-1 S and B)

SARS-CoV-2 S VSV pseudoviruses were used to evaluate the neutralization capacity of SARS VHH-72

NI cells were not infected. C) Binding of bivalent VHHs was tested by ELISA against SARS-CoV-1 S and D) SARS-CoV-2 RBD-SD1. VHH-72-Fc refers to SARS VHH-72 fused to a human IgG1 Fc domain by a GS(GGGGS) 2 linker. VHH-72-Fc (S) is the same Fc fusion with a GS, rather than a GS(GGGGS) 2 , linker. GBP is an irrelevant GFP-binding protein. VHH-72-VHH-72 refers to the tail-tohead construct with two SARS VHH-72 proteins connected by a (GGGGS) 3 linker. VHH-23-VHH-23 refers to the two irrelevant VHHs linked via the same (GGGGS) 3 linker. E) SARS-CoV-1 S and F) SARS-CoV-2 S pseudoviruses were used to evaluate the neutralization capacity of bivalent VHH-72-Fc

CoV VHH immunization and panning. A) Schematic depicting the immunization strategy that was used to isolate both SARS-CoV-1 S and MERS-CoV S-directed VHHs from a single llama. The prefusion stabilized SARS-CoV-1 spike is shown in pink and the prefusion stabilized MERS-CoV spike is shown in tan. B) Phylogenetic tree of the isolated MERS-CoV and SARS-CoV S-directed VHHs, based on the neighbor joining method. C) Reactivity of MERS-CoV and SARS-CoV S-directed VHHs against the prefusion stabilized MERS-CoV S and SARS-CoV-1 S protein, respectively. A VHH against an irrelevant antigen (F-VHH) was included as a control. RBD is shown with SARS VHH-72 as dark blue ribbons and the RBD as a pink molecular surface. Amino acids that vary between SARS-CoV-1 and WIV1-CoV are colored teal. C) SPR sensorgram measuring the binding of SARS VHH-72 to the WIV1-CoV RBD. Binding curves are colored black and the fit of the data to a 1:1 binding model is colored red. Figure 6 : Engineering a functional bivalent VHH construct. A) Flow cytometry measuring the binding of the bivalent SARS VHH-72 tail-to-head fusion (VHH-72-VHH-72) to SARS-CoV-1 or SARS-CoV-2 S expressed on the cell surface. VHH-23-VHH-23, a bivalent tail-to-head fusion of an irrelevant nanobody, was included as a negative control. B) Binding of SARS-CoV-2 RBD-SD1 to Vero E6 cells is prevented by VHH-72-VHH-72 in a dose-dependent fashion. Binding of SARS-CoV-2 RBD-SD1 to Vero E6 cells was detected by flow cytometry in the presence of the indicated bivalent VHHs (n = 2 except VHH-72-VHH-72 and VHH-23-VHH-23 at 5 µg/ml, n = 5). C) Binding of SARS-CoV-2 RBD-SD1 to Vero E6 cells is prevented by bivalent VHH-72-Fc fusion proteins in a dose-dependent fashion. Binding of SARS-CoV-2 RBD-SD1-Fc to Vero E6 cells was detected by flow cytometry in the presence of the indicated constructs and amounts (n = 2 except no RBD, n = 4). D) Cell surface binding of SARS VHH-72 to SARS-CoV-1 S. 293T cells were transfected with a GFP expression plasmid together with a SARS-CoV-1 S expression plasmid. Binding of the indicated protein is expressed as the median fluorescent intensity (MFI), measured to detect the His-tagged MERS VHH-55 or SARS VHH-72 or the SARS VHH-72-Fc fusions, of the GFP positive cells divided by the MFI of the GFP negative cells. E) Cell surface binding of SARS VHH-72 to SARS-CoV-2. MFI was calculated using the same equation as S. Figure 6D . Figure 7 : Comparison of the CoV VHH epitopes with known RBD-directed antibodies.

A) The structure of MERS VHH-55 bound to the MERS-CoV RBD is shown with MERS VHH-55 as blue ribbons and the MERS-CoV RBD as a white molecular surface. Epitopes from previously reported crystal structures of the MERS-CoV RBD bound by RBD-directed antibodies are shown as colored patches on the MERS-CoV RBD surface. The LCA60 epitope is shown in yellow, the MERS S4 epitope is shown in green, the overlapping C2/MCA1/m336 epitopes are shown in red and the overlapping JC57-14/D12/4C2/MERS-27 epitopes are shown in purple. B) The structure of SARS VHH-72 bound to the SARS-CoV-1 RBD is shown with SARS VHH-72 as cyan ribbons and the SARS-CoV-1 RBD as a white molecular surface. Epitopes from previously reported crystal structures of the SARS-CoV-1 RBD bound by RBD-directed antibodies are shown as colored patches on the SARS-CoV-1 RBD surface. The 80R epitope is shown in blue, the S230 epitope is shown in yellow, and the overlapping m396/F26G19 epitopes are shown in red.