key: cord-0852396-zw26usmh
authors: Walter, Justin D.; Hutter, Cedric A.J.; Garaeva, Alisa A.; Scherer, Melanie; Zimmermann, Iwan; Wyss, Marianne; Rheinberger, Jan; Ruedin, Yelena; Earp, Jennifer C.; Egloff, Pascal; Sorgenfrei, Michèle; Hürlimann, Lea M.; Gonda, Imre; Meier, Gianmarco; Remm, Sille; Thavarasah, Sujani; van Geest, Gerrit; Bruggman, Rémy; Zimmer, Gert; Slotboom, Dirk J.; Paulino, Cristina; Plattet, Philippe; Seeger, Markus A.
title: Biparatopic sybody constructs neutralize SARS-CoV-2 variants of concern and mitigate emergence of drug resistance
date: 2021-08-30
journal: bioRxiv
DOI: 10.1101/2020.11.10.376822
sha: 23008acdbb62b2eff1e09868b826dfbbb45b60c5
doc_id: 852396
cord_uid: zw26usmh

The ongoing COVID-19 pandemic represents an unprecedented global health crisis. Here, we report the identification of a synthetic nanobody (sybody) pair (Sb#15 and Sb#68) that can bind simultaneously to the SARS-CoV-2 spike-RBD and efficiently neutralize pseudotyped and live-viruses by interfering with ACE2 interaction. Two spatially-discrete epitopes identified by cryo-EM translated into the rational design of bispecific and tri-bispecific fusions constructs, exhibiting up to 100- and 1000-fold increase in neutralization potency. Cryo-EM of the sybody-spike complex further revealed a novel up-out RBD conformation. While resistant viruses emerged rapidly in the presence of single binders, no escape variants were observed in presence of the bispecific sybody. The multivalent bispecific constructs further increased the neutralization potency against globally-circulating SARS- CoV-2 variants of concern. Our study illustrates the power of multivalency and biparatopic nanobody fusions for the development of clinically relevant therapeutic strategies that mitigate the emergence of new SARS-CoV-2 escape mutants.

The spike glycoprotein (S-protein) is the most prominent surface-exposed entity of severe acute 40 respiratory syndrome coronavirus 2 (SARS-CoV-2), and possesses the vital molecular machinery 41 required for recognition and fusion with host membranes [1] . To date, most authorized vaccines 42 against coronavirus disease 2019 (COVID-19) rely on exposure of patients solely to the S-protein [2] . 43

Similarly, the S-protein is the exclusive target of currently approved monoclonal antibody therapies 44

for . Unfortunately, recent months have seen the emergence and rapid spread of 45 mutant viral strains conferring amino acid changes in the S-protein which can attenuate 46 neutralization by many convalescent, vaccine-induced, and monoclonal antibodies [4, 5] . Therefore, 47 from a public health perspective, it is imperative to pursue the development of therapeutic strategies 48 that can withstand the continued emergence of SARS-CoV-2 escape mutants. 49

The S-protein mutations that cause increased virulence and immune evasion are predominantly 50 found in the receptor-binding domain (RBD) [5, 6] , which is specifically responsible for host 51 recognition via interaction with human angiotensin-converting enzyme 2 (ACE2) [7] . The RBD harbors 52 two hotspots for antibody recognition. One of these epitopes overlaps with the ACE2 binding 53 interface and is evolutionarily unique in SARS-CoV-2; the second so-called "cryptic" epitope is found 54 in a peripheral region that is conserved among RBDs from several characterized coronaviruses [8] . 55

While individually targeting either epitope with antibodies quickly results in the emergence of escape 56 mutants [9, 10] , there is growing evidence that simultaneous engagement of both epitopes via 57 polyvalent antibodies may mitigate viral escape [11, 12] . 58 8 19 patients [22, 27] . Nanobodies offer key advantages over conventional antibodies, in particular the 253 ease of multimerization, inexpensive production and high protein stability. The latter simplifies 254 logistics and facilitates development in an inhalable formulation [28, 29] , thereby not only enabling 255 direct delivery to nasal and lung tissues (two key sites of SARS-CoV-2 replication), but also offering 256 the potential of self-administration. 257

Our study focused on a pair of sybodies, Sb#15 and Sb#68, which recognize two non-overlapping 258 epitopes on the RBD. Both sybodies were found to compete with ACE2 binding. While the binding 259 epitope of Sb#15 directly overlaps with the one of ACE2, this is not the case for Sb#68, which 260 interferes with ACE2 through a steric clash at the sybody backside. Sb#15 and Sb#68 exhibited similar 261 neutralization efficiencies, as well as a moderate synergistic effect in the virus neutralization test 262 when both individual sybodies were mixed together. This synergy may be explained by the concerted 263 action of the sybodies to compete with ACE2 docking via epitope blockage and steric clashing. 264

Cryo-EM analyses confirmed distinct binding epitopes for the two sybodies Sb#15 and Sb#68. 265 Without sybodies, the spike protein predominantly assumes an equilibrium between the 3down and 266 the 1up2down conformation [1, 20] . Upon addition of Sb#15, the conformational equilibrium was 267 shifted towards an asymmetric 1up/1up-out/1down state, whereas addition of Sb#68 favored an 268 asymmetric state with RBDs adopting a 2up/1flexible conformation. When added together, the 269 sybodies synergized to stabilize two states: a predominant 3up state, as well as the asymmetric 270

1up/1up-out/1down state, thereby shifting the conformational equilibrium of the spike towards RBD 271 conformations competent for ACE2 binding. These structural findings are reminiscent of a recent 272 study, in which a pair of nanobodies isolated from immune libraries (VHH E and VHH V) was found to 273 bind to similar epitopes as Sb#15 and Sb#68 [11] . However, in contrast to Sb#15 stabilizing the 274 asymmetric 1up/1up-out/1down state when added alone, the corresponding VHH E nanobody is 275 exclusively bound to, and thereby stabilizes, the 3up conformation. Hence, what is unique for our 276 Sb#15/Sb#68 pair is its concerted action to shift the conformational equilibrium of the spike towards 277 the 3up state and its capability to trap the spike protein in an unusual 1up/1up-out/1down 278 conformation, which to the best of our knowledge has not been previously described. 279

Akin to the antibodies CR3022 and EY6A [19, 30] as well as a growing number of nanobodies [11, 31-280 33] , the Sb#15/Sb#68 pair stabilized spike conformations with 2up or 3up RBDs. Thereby, the spike 281 protein may be destabilized, resulting in the premature and unproductive transitions to the 282 irreversible post-fusion state. This mechanism was dubbed "receptor mimicry" in a study on a 283 neutralizing antibody S230, which only bound to up-RBDs and thereby triggered fusogenic 284 conformational changes of SARS-CoV-1 spike [34] . In elegant experiments, König et al. could 285 demonstrate that stabilization of the spike protein in its 3up conformation by the addition of 286 nanobodies VHH E and V indeed resulted in aberrant activation of the spike fusion machinery [11] . 287

Hence, it is plausible to assume that our sybody pair inhibits SARS-CoV-2 infection and or/entry via 288 such a receptor mimicry mechanism, in addition to blockage of ACE2 binding.

The binding epitope of Sb#68, also called the "cryptic" epitope [8] , is highly conserved between 290 SARS-CoV-1 and SARS-CoV-2. The same conserved epitope is also recognized by the human 291 antibodies CR3022 (isolated from a SARS-CoV-1 infected patient and showing cross-specificity against 292 SARS-CoV-2) and EY6A [19, 30] as well as the nanobody VHH-72, which had been originally selected 293 against SARS-CoV-1 but was shown to cross-react with SARS-CoV-2 [35] (Fig. 7) . Recent months have 294 brought about a growing number of other nanobodies from immune and synthetic libraries whose 295 epitopes overlap with Sb#68 [11, [31] [32] [33] 36, 37] , suggesting that the cryptic epitope constitutes a 296 preferred binding site for VHHs (Fig. 7C) . The cryptic epitope is unchanged in currently circulating 297 variants of concern, including the B.1.1.7 (alpha), B.1.351 (beta) and the B.1.617.2 (delta) lineages 298 ( Fig. 5A) and consequently, neutralization efficiency of Sb#68 is unaffected against these variants 299 ( Fig. 6B and C, Table 3 ). 300

Fusion of nanobodies via flexible linkers has emerged as a promising strategy to improve 301 neutralization efficiencies by exploiting avidity effects. This potency-boosting procedure has been 302 explored in the context of SARS-CoV-2 binders by either fusing up to three identical nanobodies 303 (multivalency) [28, 31, 38] , or by the structure-based design of biparatopic nanobodies [11, 39] . We 304 exploited our structural data to first fuse Sb#15 and Sb#68 into the biparatopic GS4 construct, which 305 resulted in a more than 100-fold gain of neutralization efficiency (Table 1 ). In a subsequent fusion 306 step, the bi-paratopic GS4 construct was equipped with a trimerization domain (Tripod-GS4r), 307 resulting in a further 10-fold boost in neutralization potential, thereby increasing the cumulative 308 neutralization potency by a factor of around 1000 when compared to the single constituent 309 nanobodies (Table 1) . To our knowledge, engineering of such a "trimer-of-dimers" construct has not 310 yet been attempted with anti-SARS-CoV-2 binder proteins. In addition to neutralizing virus entry via 311 competition with receptor-binding and inducing premature activation of the fusion machinery, the 312 unique multivalent structure of Tripod-GS4r may trigger clustering of neighboring spikes, thereby 313 eventually deactivating multiple viral entry machineries simultaneously. 314

The ability of enveloped RNA viruses such as SARS-CoV-2 to rapidly develop resistance mutations is a 315 crucial issue of consideration for the development of reliable therapeutics. Escape mutants indeed 316 rapidly emerged when in vitro selection experiments were carried with single monoclonal antibodies 317 or nanobodies targeting either the receptor-binding motif or the cryptic epitope alone [11, 22, 27] . It 318 is interesting to note that the Q493R escape mutation we isolated for Sb#15 has been recently 319 observed in a COVID-19 patient who received monoclonal antibody therapy [15] . Thus, an attractive 320 strategy to potentially suppress mutational escape is to employ a cocktail of neutralizing antibodies 321 binding to discrete epitopes of the spike. Accumulating evidence indeed demonstrate that rapid viral 322 escape is not observed when experiments are performed either in the presence of a combination of 323 neutralizing monoclonal antibodies/nanobodies or in the presence of biparatopic fusion constructs 324 [11, 22, 27] . Furthermore, in addition to efficiently suppress mutational escape, our biparatopic 325 molecules (e.g., GS4 and Tripod-GS4r) consistently retained their neutralization capacity against 326 pseudotyped VSV carrying spikes that harbored key mutations of common SARS-CoV-2 variants of 327 concern. In particular, the trimerized biparatopic construct (Tripod-GS4r) exhibited ultra-potent 328 neutralizing activity against all tested spike variants, with IC 50 values of low picomolar range. Hence, 329

our study provides evidence that combining multivalency and biparatopic nanobody fusion proteins 330 represent a promising strategy to potentially generate therapeutic molecules with clinical relevance. 331

In conclusion, the rapid selection of sybodies [16] and their swift biophysical, structural and 332 functional characterization, provide a foundation for the accelerated reaction to potential future 333 pandemics. In contrast to a number of synthetic or naïve SARS-CoV-2 nanobodies from other libraries 334 that required a post-selection maturation process to reach satisfactory affinities [28, [40] [41] [42] 

For initial sybody selection experiments and binding affinity measurements, a gene encoding SARS-349

CoV-2 residues Pro330-Gly526 (RBD, GenBank accession QHD43416. RBD escape mutants and sybodies, an RBD construct consisting of residues Arg319-Phe541, 357 downstream from the native N-terminal SARS-CoV-2 secretion signal (Met1-Ser13) and appended 358 with a C-terminal 10x-histidine tag (RBD-His), was cloned into a custom mammalian expression 359 vector derived from pcDNA3.1 (ThermoFisher). Escape mutations P384H, K417N, E484K, Q493R, and 360 N501Y were introduced into RBD-His using QuikChange site-directed mutagenesis. Expression 361 plasmids harboring the prefusion ectodomain of the SARS-CoV2 spike protein (Met1-Gln1208), 362 containing two or six stabilizing proline mutations (S-2P or S-6P, respectively) and a C-terminal foldon 363 trimerization motif, HRV 3C protease cleavage site, and twin-strep tag, were a generous gift from 364 

Suspension-adapted Expi293 cells (Thermo) were transiently transfected using Expifectamine 370 according to the manufacturer protocol (Thermo), and expression was continued for 3-5 days in a 371 humidified environment at 37°C, 8% CO 2 . Cells were pelleted (500g, 10 min), and culture supernatant 372 was filtered (0.2 µm mesh size) before being incubated with the appropriate affinity chromatography 373 matrix. For RBD-vYFP, NHS-agarose beads covalently coupled to the anti-GFP nanobody 3K1K[52] 374 were used for affinity purification, and RBD-vYFP was eluted with 0.1 M glycine, pH 2.5, into tubes 375 that were pre-filled with 1/10 vol 1M Tris (pH 9.0). Strep-Tactin®XT Superflow® (iba lifesciences) was 376 used to pull down twin-strep-tagged S-2P or S-6P from culture supernatant, followed by elution with 377 50mM biotin. Ni-NTA beads were used for affinity purification of RBD-His, which was eluted with 378 300mM imidazole. All affinity-purified SARS-CoV-2 proteins were also subjected to size-exclusion 379 chromatography using either a Superdex 200 Increase 10/300 GL column for RBD constructs, or a 380 Superose 6 Increase 10/300 GL column for spike proteins. 381

Sybody selections, entailing one round of ribosome display followed by two rounds of phage display, 383 were carried out as previously detailed with the three synthetic sybody libraries designated concave, 384 loop and convex [16] . All targets were chemically biotinylated using NHS-Biotin (ThermoFisher 385 #20217) according to the manufacturer protocol. Binders were selected against two different 386 constructs of the SARS-CoV-2 RBD; an RBD-vYFP fusion and an RBD-Fc fusion. MBP was used as 387 background control to determine the enrichment score by qPCR [16] . In order to avoid enrichment of 388 binders against the fusion proteins (YFP and Fc), we switched the two targets after ribosome display. 389

For the off-rate selections we did not use non-biotinylated target proteins as described[16] because 390 we did not have the required amounts of purified target protein. Instead, we employed a pool 391 competition approach. After the first round of phage display, all three libraries of selected sybodies, 392 for both target-swap selection schemes, were subcloned into the pSb_init vector (giving 393 approximately 10 8 clones) and expressed in E. coli MC1061 cells. The resulting three expressed pools 394 were subsequently combined, giving one sybody pool for each selection scheme. These two final 395 pools were purified by Ni-NTA affinity chromatography, followed by buffer exchange of the main 396 peak fractions using a desalting PD10 column in TBS pH 7.5 to remove imidazole. The pools were 397 eluted with 3.2 ml of TBS pH 7.5. These two purified pools were used for the off-rate selection in the 398 second round of phage display at concentrations of approximately 390 µM for selection variant 1 399 (competing for binding to RBP-Fc) and 450 µM for selection variant 2 (competing for binding to RBP-400 YFP). The volume used for off-rate selection was 500 µl, with 0.5% BSA and 0.05% Tween-20 added 12 to pools immediately prior to the competition experiment. Off-rate selections were performed for 3 402 minutes. For identification of binder hits, ELISAs were performed as described [16] . 47 single clones 403 were analyzed for each library of each selection scheme. Since the RBD-Fc construct was 404 incompatible with our ELISA format due to the inclusion of Protein A to capture an α-myc antibody, 405 ELISA was performed only for the RBD-vYFP (50 nM) and the M) and later on with the S-2P (25 nM). 406

Of note, the three targets were analyzed in three separate ELISAs. As negative control to assess 407 background binding of sybodies, we used biotinylated MBP (50 nM). 72 positive ELISA hits were 408 sequenced (Microsynth, Switzerland). 409 410

The 63 unique sybodies were expressed and purified as described [16] . In short, all 63 sybodies were 412 expressed overnight in E.coli MC1061 cells in 50 ml cultures. The next day the sybodies were 413 extracted from the periplasm and purified by Ni-NTA affinity chromatography (batch binding) 414 followed by size-exclusion chromatography using a Sepax SRT-10C SEC100 size-exclusion 415 chromatography (SEC) column equilibrated in TBS, pH 7.5, containing 0.05% (v/v) Tween-20 416 (detergent was added for subsequent kinetic measurements). Six out of the 63 binders (Sb#4, Sb#7, 417

Sb#18, Sb#34, Sb#47, Sb#61) were excluded from further analysis due to suboptimal behavior during 418 SEC analysis (i.e. aggregation or excessive column matrix interaction). 419 420

To generate the bispecific sybodies (Sb#15-Sb#68 fusion with variable glycine/serine linkers), Sb#15 422 was amplified from pSb-init_Sb#15 (Addgene #153523) using the forward primer 5'-ATA TAT GCT CTT 423 CAA GTC AGG TTC and the reverse primer 5'-TAT ATA GCT CTT CAA GAA CCG CCA CCG CCG CTA CCG 424 CCA CCA CCT GCG CTC ACA GTC AC, encoding 2x a GGGGS motif, followed by a SapI cloning site. 425

Sb#68 was amplified from pSb-init_Sb#68 (Addgene #153527) using forward primer 5'-ATA TAT GCT 426 CTT CTT CTG GTG GTG GCG GTA GCG GCG GTG GCG GTA GTC AAG TCC AGC TGG TGG combined with 427 the reverse primer 5'-TAT ATA GCT CTT CCT GCA GAA AC. The forward primers start with a SapI site 428 (compatible overhang to Sb#15 reverse primer), followed by 2x the GGGGS motif. The PCR product of 429 Sb#15 was cloned in frame with each of the three PCR products of Sb#68 into pSb-init using FX-430 cloning [47] , thereby resulting in three fusion constructs with linkers containing 4x GGGGS motives as 431 flexible linker between the sybodies (called GS4). The bispecific fusion construct GS4 was expressed 432 and purified the same way as single sybodies [16] . 433 434

In order to engineer a trivalent GS4 molecule, we fused the following elements (from N-to C-436 terminus): Sb#68-(GGGGS) 4 -Sb#15-(GGGGS) 4 -CC-GCNt-Foldon-TST. CC is the trimeric CDV F protein 437 13 (589-599) and contains two successive cysteine mutations (I595C and L596C) shown in the context of 438 soluble measles virus F protein to stabilize the prefusion state [53] . GCNt is a trimerization motif that 439 was previously described [25] . Foldon stems from fibritin [24] . TST denotes a C-terminal 440

His/TwinStrepTag sequence for purification purposes. The Tripod-GS4r expression plasmid (3 mg) 441 was sent to the Protein Production and Structure Core Facility of the EPFL (Switzerland) for 442 expression (7 days in ExpiCHO cells). Subsequently, the protein was purified from 1 L of supernatant 443 using a 5 mL StrepTtrapXT column (Cytavia) and eluted with 500 mM biotin (Cytivia). incubated for 10 min, followed by washing three times with 250 µl TBS-T per well. Finally, to detect 456 S-2P bound to the immobilized sybodies, 100 µl ELISA developing buffer (prepared as described 457

previously [16] ) was added to each well, incubated for 1 h (due to low signal) and absorbance was 458 measured at 650 nm. As a negative control, TBS-BSA-T devoid of protein was added to the 459 corresponding wells instead of a S-2P-sybody mixture. 460 461

Kinetic characterization of sybodies binding onto SARS-CoV-2 spike proteins was performed using GCI 463 on the WAVEsystem (Creoptix AG, Switzerland), a label-free biosensor. For the off-rate screening, 464 biotinylated RBD-vYFP and ECD were captured onto a Streptavidin PCP-STA WAVEchip 465 (polycarboxylate quasi-planar surface; Creoptix AG) to a density of 1300-1800 pg/mm 2 . Sybodies 466 were first analyzed by an off-rate screen performed at a concentration of 200 nM (data not shown) 467 to identify binders with sufficiently high affinities. The six sybodies Sb#14, Sb#15, Sb#16, Sb#42, 468 Sb#45, and Sb#68 were then injected at increasing concentrations ranging from 1.37 nM to 1 μM 469 (three-fold serial dilution, 7 concentrations) in 20 mM Tris pH7.5, 150 mM NaCl supplemented with 470 0.05 % Tween-20 (TBS-T buffer). Sybodies were injected for 120 s at a flow rate of 30 μl/min per 471 channel and dissociation was set to 600 s to allow the return to baseline. 472

In order to determine the binding kinetics of Sb#15 and Sb#68 against intact spike proteins, the 473 ligands RBD-vYFP, S-2P and S-6P were captured onto a PCP-STA WAVEchip (Creoptix AG) to a density 474 14 of 750 pg/mm 2 , 1100 pg/mm 2 and 850 pg/mm 2 respectively. Sb#15 and Sb#68 were injected at 475 concentrations ranging from 1.95 nM to 250 nM or 3.9 nM to 500 nM, respectively (2-fold 

The serial dilutions of control sera and samples were prepared in quadruplicates in 96-well cell 514 culture plates using DMEM cell culture medium (50 µL/well). To each well, 50 µL of DMEM containing 515 100 tissue culture infectious dose 50% (TCID 50 ) of SARS-CoV-2 (SARS-CoV-2/München-1.1/2020/929) 516 were added and incubated for 60 min at 37°C. Subsequently, 100 µL of Vero E6 cell suspension 517 (100,000 cells/mL in DMEM with 10% FBS) were added to each well and incubated for 72 h at 37 °C. 518

The cells were fixed for 1 h at room temperature with 4% buffered formalin solution containing 1% 519 crystal violet (Merck, Darmstadt, Germany). Finally, the microtiter plates were rinsed with deionized 520 water and immune serum-mediated protection from cytopathic effect was visually assessed. amino acid substitution R685G is located in the S1/S2 proteolytic cleavage site and has been shown 538 to reduce syncytia formation and to enhance virus titers [57] . 539

The chimeric virus was rescued following transfection of cDNA according to a previously described 

Freshly purified S-2P was incubated with a 1.3-fold molar excess of Sb#15 alone or with Sb#15 and 605

Sb#68 and subjected to size exclusion chromatography to remove excess sybody. In analogous way, 606 the sample of S-6P with Sb#68 was prepared. The protein complexes were concentrated to 0.7-1 mg 607 ml -1 using an Amicon Ultra-0.5 mL concentrating device (Merck) with a 100 kDa filter cut-off. 2.8 μl of 608 the sample was applied onto the holey-carbon cryo-EM grids (Au R1.2/1.3, 300 mesh, Quantifoil), 609 which were prior glow discharged at 5 -15 mA for 30 s, blotted for 1-2 s and plunge frozen into a 610 liquid ethane/propane mixture with a Vitrobot Mark IV (Thermo Fisher) at 15 °C and 100% humidity. 611

Samples were stored in liquid nitrogen until further use. Screening of the grid for areas with best ice 612 properties was done with the help of a home-written script to calculate the ice thickness (manuscript 613 in preparation). Cryo-EM data in selected grid regions were collected in-house on a 200-keV Talos 614

Arctica microscope (Thermo Fisher Scientifics) with a post-column energy filter (Gatan) in zero-loss 615 mode, with a 20-eV slit and a 100 μm objective aperture. Images were acquired in an automatic 616 manner with SerialEM on a K2 summit detector (Gatan) in counting mode at ×49,407 magnification 617

(1.012 Å pixel size) and a defocus range from −0.9 to −1.9 μm. During an exposure time of 9 s, 60 618 frames were recorded with a total exposure of about 53 electrons/Å 2 . On-the-fly data quality was 619 monitored using FOCUS [64] . state, while only Sb#15 is bound to the down RBD. The 3up class was further refined with C3 636 symmetry imposed. The final refinement, where a mask was included in the last iteration, provided a 637 map at 7.6 Å resolution. Six rounds of per-particle CTF refinement with beamtilt estimation and re-638 extraction of particles with a box size of 400 pixels improved resolution further to 3.2 Å. The particles 639 were then imported into cryoSPARC, where non-uniform refinement improved the resolution to 3 Å. 640

The asymmetrical 1up/1up-out/1down was refined in an analogous manner with no symmetry 641 imposed, resulting in a map at 7.8 Å resolution. Six rounds of per-particle CTF refinement with 642 beamtilt estimation improved resolution to 3.7 Å. A final round of non-uniform refinement in 643 cryoSPARC yielded a map at 3.3 Å resolution. Local resolution estimations were determined in 644 cryoSPARC. All resolutions were estimated using the 0.143 cut-off criterion [69] with gold-standard 645

Fourier shell correlation (FSC) between two independently refined half-maps [70] . The directional 646 resolution anisotropy of density maps was quantitatively evaluated using the 3DFSC web interface 647 (https://3dfsc.salk.edu) [71] . 648 A similar approach was performed for the image processing of the S-2P/Sb#15 complex. In short, 649 2,235 micrographs were recorded, and 1,582 used for image processing after selection. 66,632 650 particles were autopicked via crYOLO and subjected to 2D classification in cryoSPARC. 57,798 651 selected particles were used for subsequent 3D classification in RELION-3.0.8, where the symmetrical 652 3up map, described above, was used as initial reference. The best class comprising 22,055 particles 653 (38%) represented an asymmetrical 1up/1up-out/1down conformation with Sb#15 bound to each 654 RBD. Several rounds of refinement and CTF refinement yielded a map of 4.0 Å resolution. 655

For the dataset of the S-6P/Sb#68 complex, in total 5,109 images were recorded, with 4,759 used for 656 further image processing. 344,976 particles were autopicked via crYOLO and subjected to 2D 657 classification in cryoSPARC. 192,942 selected particles were imported into RELION-3.0.8 and used for 658 subsequent 3D classification, where the symmetrical 3up map, described above, was used as initial 659 reference. Two distinct classes of spike protein were found. One class (24,325 particles, 13%) 660 revealed a state in which two RBDs adopt an up conformation with Sb#68 bound, whereby the 661 density for the third RBD was poorly resolved representing an undefined state. Several rounds of 662 refinement and CTF refinement yielded a map of 4.8 Å resolution. Two other classes, comprising 663 44,165 particles (23%) and 84,917 particles (44%), were identical. They show a 1up/2down 664 configuration without Sb#68 bound to any of the RBDs. Both classes were processed separately, 665 whereby the class with over 80k particles yielded the best resolution of 3. Defocus range (μm) -0.9 to -1.9 -0.9 to -1.9 -0.9 to -1.9 -0.9 to -1.9 -0.9 to -1.9

Pixel size (Å) Sybodies Sb#15 and Sb#68 bind non-overlapping epitopes on the S-protein, and inhibit ACE2 binding. (A) Affinity determina on of Sb#15 and Sb#68 against the immobilized S-protein (S-2P) using GCI. The data were fi ed using a heterogeneous ligand model. (B) le , GCI epitope-binning experiment showing Sb#15 (blue), Sb#68 (red) and their combina on (black) against immobilized S-protein (S-2P). Since both sybodies were present at satura ng concentra ons, the increased amplitude is indica ve of simultaneous binding. Right, ELISA experiment confirming dual-binding of Sb#15 and Sb#68. Myc-tagged Sb#15 was immobilized on an an -myc an body-coated ELISA plate, followed by exposure of bio nylated RBD which was pre-mixed with tag-less sybodies (indicated on the x-axis). (C) Compe on of sybodies and ACE2 for S-protein binding, inves gated by GCI. Bio nylated S-protein was immobilized on the GCI sensor and then Sb#15 (200 nM, le ), or Sb#68 (200 nM, right) were injected alone or premixed with human ACE2 (100 nM). Sb#0 represents a nonrandomized control sybody. The bispecific construct GS4 mi gates emergence of novel escape mutants. (A) Structural context of RBD escape mutants resul ng from adapta on experiments in the presence of either Sb#15 or Sb#68 alone (salmon spheres). Globally-circula ng variants of concern are shown as blue spheres. (B) GCI-based kine c analysis of the purified RBD bearing either no muta on (WT), or iden fied escape muta ons Q493R or P384H. Le , middle, and right plots correspond to immobilized Sb#15, Sb#68, or GS4, respec vely. (C) Neutraliza on assay using VSVΔG pseudotyped with Spike-protein containing the Q493R or P384H muta on, respec vely. Rela ve infec vity in response to increasing binder concentra ons was determined. Error bars correspond to standard devia ons of three biological replicates.

Sb#68 Sb#68 Sb#68 GS4 GS4

Affinity and neutraliza on by sybody constructs for variants of concern. (A) Semi-quan ta ve GCI analysis of interac on between RBDs carrying the individual K417N/E484K/N501Y muta ons (le panels) or the combined triple KEN-RBD mutant (right panels). Sb#15, Sb#68 and GS4 were immobilized via a bio nylated Avi-tag and the RBD variants were injected. (B) Neutraliza on assay using VSVΔG pseudotyped with Spike-protein containing the triple KEN (beta) muta ons. Rela ve infec vity in response to increasing binder concentra ons was determined. Error bars correspond to standard devia ons of three biological replicates. Kine c characteriza on of sybodies by GCI. (A) RBD-vYFP and ECD were immobilized as indicated and the six top sybodies were injected at increasing concentra ons ranging from 1.37 nM to 1 μM. Data were fi ed using a Langmuir 1:1 model. (B) In depth affinity characteriza on of Sb#15 and Sb#68. RBD-vYFP and S-6P were immobilized as indicated and Sb#15 and Sb#68 were injected at concentra ons ranging from 1.95 nM to 250 nM for Sb#15 and 3.9 nM to 500 nM for Sb#68. For RBD, data were fi ed using a Langmuir 1:1 model. For S-6P, the data were fi ed with the heterogeneous ligand model, because the 1:1 model was clearly not appropriate to describe the experimental data. Corresponding data for S-2P is shown in main Fig. 1C . 

SDS-PAGE analysis of Tripod-GS4r. Purified Tripod-GS4r was loaded on a SDS-PAGE gel with and without incuba on of βmercaptoethanol (β-ME) and stained using coomassie blue. 

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

SARS-CoV-2 immunity: review and 749 applications to phase 3 vaccine candidates. The Lancet

Neutralizing monoclonal antibodies for treatment of COVID-19

Multiple SARS-CoV-2 variants escape neutralization by vaccine-753 induced humoral immunity

SARS-CoV-2 variants, spike mutations and immune escape

Mapping neutralizing and immunodominant sites on the SARS-CoV-2 spike 757 receptor-binding domain by structure-guided high-resolution serology

Receptor binding and priming of the spike protein of SARS-CoV-2 for 760 membrane fusion

A highly conserved cryptic epitope in the receptor binding domains of SARS-762

Complete mapping of mutations to the SARS-CoV-2 spike receptor-764 binding domain that escape antibody recognition. Cell host & microbe

Escape from neutralizing antibodies by SARS-CoV-2 spike protein 767 variants. Elife

Structure-guided multivalent nanobodies block SARS-CoV-2 infection and 769 suppress mutational escape

Bispecific IgG neutralizes SARS-CoV-2 variants and prevents escape in 771 mice

Synthetic single domain antibodies for the conformational trapping of 773 membrane proteins. Elife

The therapeutic potential of nanobodies

Bamlanivimab/Etesevimab Treatment for COVID-19

Generation of synthetic nanobodies against delicate proteins

Rapid Quantification of SARS-CoV-2-Neutralizing Antibodies Using 781

Propagation-Defective Vesicular Stomatitis Virus Pseudotypes. Vaccines (Basel)

Synthetic nanobody-SARS-CoV-2 receptor-binding domain structures 783 identify distinct epitopes

Structural basis for the neutralization of SARS-CoV-2 by an antibody from a 785 convalescent patient

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

LY6E impairs coronavirus fusion and confers immune control of viral 789 disease

Antibody cocktail to SARS-CoV-2 spike protein prevents rapid mutational 791 escape seen with individual antibodies

Deep Mutational Scanning of SARS-CoV-2 Receptor Binding Domain Reveals 793 Constraints on Folding and ACE2 Binding

Structure of RSV Fusion Glycoprotein Trimer Bound to a Prefusion-795 Specific Neutralizing Antibody

Structure of the parainfluenza virus 5 F protein in its metastable, prefusion 797 conformation

Reduced neutralization of SARS-CoV-2 B.1.617 by vaccine and convalescent 799 serum

The monoclonal antibody combination REGEN-COV protects against SARS-801

CoV-2 mutational escape in preclinical and human studies

An ultrapotent synthetic nanobody neutralizes SARS-CoV-2 by stabilizing 803 inactive Spike

Nanobodies (R)dagger as inhaled biotherapeutics for lung diseases

Neutralization of SARS-CoV-2 by Destruction of the Prefusion Spike

Nanobodies from camelid mice and llamas neutralize SARS-CoV-2 variants. 809 Nature

Potent neutralizing nanobodies resist convergent circulating variants of SARS-811

CoV-2 by targeting novel and conserved epitopes. bioRxiv

Nanobody cocktails potently neutralize SARS-CoV-2 D614G N501Y variant 813 and protect mice

Unexpected Receptor Functional Mimicry Elucidates Activation of 815 Coronavirus Fusion

Structural Basis for Potent Neutralization of Betacoronaviruses by Single-817

Domain Camelid Antibodies

A high-affinity RBD-targeting nanobody improves fusion partner's potency 819 against SARS-CoV-2

NeutrobodyPlex-monitoring SARS-CoV-2 neutralizing immune responses 821 using nanobodies

Versatile and multivalent nanobodies efficiently neutralize SARS-CoV-2

SPHIRE-crYOLO is a fast and accurate fully automated particle picker for 825 cryo-EM

Neutralizing nanobodies bind SARS-CoV-2 spike RBD and block interaction with 827 ACE2

Rapid generation of potent antibodies by autonomous hypermutation in 829 yeast

Directed evolution of potent neutralizing nanobodies against SARS-CoV-831 2 using CDR-swapping mutagenesis

Selection, biophysical and structural analysis of synthetic nanobodies 833 that effectively neutralize SARS-CoV-2

An alpaca nanobody neutralizes SARS-CoV-2 by blocking receptor interaction

Engineered peptide barcodes for in-depth analyses of binding protein 837 libraries

A highly efficient modified human serum albumin signal peptide to secrete 839 proteins in cells derived from different mammalian species

A versatile and efficient high-throughput cloning tool for 842 structural biology

X-ray structure of a calcium-activated TMEM16 lipid scramblase

A variant of yellow fluorescent protein with fast and efficient maturation for 846 cell-biological applications

One-step purification of recombinant proteins using a nanomolar-affinity 848 streptavidin-binding peptide, the SBP-Tag. Protein expression and purification

Structure-based design of prefusion-stabilized SARS-CoV-2 spikes

Modulation of protein properties in living cells using nanobodies

Structures of the prefusion form of measles virus fusion protein in 855 complex with inhibitors

Immunogenicity and structures of a rationally designed prefusion MERS-858 CoV spike antigen

A human monoclonal antibody blocking SARS-CoV-2 infection

A Vesicular Stomatitis Virus Replicon-Based Bioassay for the 862 Rapid and Sensitive Determination of Multi-Species Type I Interferon

A Replication-Competent Vesicular Stomatitis Virus for Studies of SARS-864

CoV-2 Spike-Mediated Cell Entry and Its Inhibition

A recombinant vesicular stomatitis virus replicon vaccine protects 866 chickens from highly pathogenic avian influenza virus (H7N1). Vaccine

Non-replicating vaccinia vector efficiently expresses 869 bacteriophage T7 RNA polymerase

Trimmomatic: a flexible trimmer for Illumina sequence 871 data

Minimap2: pairwise alignment for nucleotide sequences

The Sequence Alignment/Map format and SAMtools

Scaling accurate genetic variant discovery to tens of thousands of samples

Focus: The interface between data collection and data processing in cryo-879 EM

MotionCor2: anisotropic correction of beam-induced motion for improved 881 cryo-electron microscopy

CTFFIND4: Fast and accurate defocus estimation from electron 883 micrographs

cryoSPARC: algorithms for rapid unsupervised cryo-EM structure 885 determination

New tools for automated high-resolution cryo-EM structure determination 887 in RELION-3. Elife

Optimal determination of particle orientation, absolute 889 hand, and contrast loss in single-particle electron cryomicroscopy

Prevention of overfitting in cryo-EM structure determination

Addressing preferred specimen orientation in single-particle cryo-EM through 894 tilting