key: cord-355728-wivk0bm0
authors: Schoof, Michael; Faust, Bryan; Saunders, Reuben A.; Sangwan, Smriti; Rezelj, Veronica; Hoppe, Nick; Boone, Morgane; Billesbølle, Christian B.; Puchades, Cristina; Azumaya, Caleigh M.; Kratochvil, Huong T.; Zimanyi, Marcell; Deshpande, Ishan; Liang, Jiahao; Dickinson, Sasha; Nguyen, Henry C.; Chio, Cynthia M.; Merz, Gregory E.; Thompson, Michael C.; Diwanji, Devan; Schaefer, Kaitlin; Anand, Aditya A.; Dobzinski, Niv; Zha, Beth Shoshana; Simoneau, Camille R.; Leon, Kristoffer; White, Kris M.; Chio, Un Seng; Gupta, Meghna; Jin, Mingliang; Li, Fei; Liu, Yanxin; Zhang, Kaihua; Bulkley, David; Sun, Ming; Smith, Amber M.; Rizo, Alexandrea N.; Moss, Frank; Brilot, Axel F.; Pourmal, Sergei; Trenker, Raphael; Pospiech, Thomas; Gupta, Sayan; Barsi-Rhyne, Benjamin; Belyy, Vladislav; Barile-Hill, Andrew W.; Nock, Silke; Liu, Yuwei; Krogan, Nevan J.; Ralston, Corie Y.; Swaney, Danielle L.; García-Sastre, Adolfo; Ott, Melanie; Vignuzzi, Marco; Walter, Peter; Manglik, Aashish
title: An ultra-potent synthetic nanobody neutralizes SARS-CoV-2 by locking Spike into an inactive conformation
date: 2020-08-17
journal: bioRxiv
DOI: 10.1101/2020.08.08.238469
sha: 
doc_id: 355728
cord_uid: wivk0bm0

Without an effective prophylactic solution, infections from SARS-CoV-2 continue to rise worldwide with devastating health and economic costs. SARS-CoV-2 gains entry into host cells via an interaction between its Spike protein and the host cell receptor angiotensin converting enzyme 2 (ACE2). Disruption of this interaction confers potent neutralization of viral entry, providing an avenue for vaccine design and for therapeutic antibodies. Here, we develop single-domain antibodies (nanobodies) that potently disrupt the interaction between the SARS-CoV-2 Spike and ACE2. By screening a yeast surface-displayed library of synthetic nanobody sequences, we identified a panel of nanobodies that bind to multiple epitopes on Spike and block ACE2 interaction via two distinct mechanisms. Cryogenic electron microscopy (cryo-EM) revealed that one exceptionally stable nanobody, Nb6, binds Spike in a fully inactive conformation with its receptor binding domains (RBDs) locked into their inaccessible down-state, incapable of binding ACE2. Affinity maturation and structure-guided design of multivalency yielded a trivalent nanobody, mNb6-tri, with femtomolar affinity for SARS-CoV-2 Spike and picomolar neutralization of SARS-CoV-2 infection. mNb6-tri retains stability and function after aerosolization, lyophilization, and heat treatment. These properties may enable aerosol-mediated delivery of this potent neutralizer directly to the airway epithelia, promising to yield a widely deployable, patient-friendly prophylactic and/or early infection therapeutic agent to stem the worst pandemic in a century.

monoclonal antibodies disclosed to date. Our lead neutralizing molecule, mNb6-tri, blocks 106 SARS-CoV-2 entry in human cells at picomolar efficacy and withstands aerosolization, 107 lyophilization, and elevated temperatures. mNb6-tri provides a promising approach to deliver a 108 potent SARS-CoV-2 neutralizing molecule directly to the airways for prophylaxis or therapy. 109 110

Synthetic nanobodies that disrupt Spike-ACE2 interaction 112

To isolate nanobodies that neutralize SARS-CoV-2, we screened a yeast surface-displayed 113 library of >2x10 9 synthetic nanobody sequences. Our strategy was to screen for binders to the 114 full Spike protein ectodomain, in order to capture not only those nanobodies that would compete 115 by binding to the ACE2-binding site on the RBD directly but also those that might bind 116 elsewhere on Spike and block ACE2 interaction through indirect mechanisms. We used a 117 mutant form of SARS-CoV-2 Spike (Spike*,) as the antigen (15). Spike* lacks one of the two 118 activating proteolytic cleavage sites between the S1 and S2 domains and introduces two 119 mutations to stabilize the pre-fusion conformation. Spike* expressed in mammalian cells binds 120 ACE2 with a KD = 44 nM ( Supplementary Fig. 1) , consistent with previous reports (17). Next, we 121 labeled Spike* with biotin or with fluorescent dyes and selected nanobody-displaying yeast over 122 multiple rounds, first by magnetic bead binding and then by fluorescence-activated cell sorting 123 (Fig. 1A) . 124 125 Three rounds of selection yielded 21 unique nanobodies that bound Spike* and showed 126 decreased Spike* binding in the presence of ACE2. Closer inspection of their binding properties 127 revealed that these nanobodies fall into two distinct classes. One group (Class I) binds the RBD 128 and competes with ACE2 (Fig. 1B) . A prototypical example of this class is nanobody Nb6, which 129 binds to Spike* and to RBD alone with a KD of 210 nM and 41 nM, respectively ( Fig. 1C ; Table  130 1). Another group (Class II), exemplified by nanobody Nb3, binds to Spike* (KD = 61 nM), but 131 displays no binding to RBD alone (Fig. 1C, Table 1 ). In the presence of excess ACE2, binding of 132

Nb6 and other Class I nanobodies is blocked entirely, whereas binding of Nb3 and other Class II 133 nanobodies is decreased only moderately (Fig. 1B) . These results suggest that Class I 134 nanobodies target the RBD to block ACE2 binding, whereas Class II nanobodies target other 135 epitopes and decrease ACE2 interaction with Spike allosterically or through steric interference. 136

Indeed, surface plasmon resonance (SPR) experiments demonstrate that Class I and Class II 137 nanobodies can bind Spike* simultaneously (Fig. 1D) . 138 139 Analysis of the kinetic rate constants for Class I nanobodies revealed a consistently greater 140 association rate constant (ka) for nanobody binding to the isolated RBD than to full-length Spike* 141 (Table 1) , which suggests that RBD accessibility influences the KD. We next tested the efficacy 142 of our nanobodies, both Class I and Class II, to inhibit binding of fluorescently labeled Spike* to 143 ACE2-expressing HEK293 cells (Table 1, Fig. 1E ). Class I nanobodies emerged with highly 144 variable activity in this assay with Nb6 and Nb11 as two of the most potent clones with IC50 145 values of 370 and 540 nM, respectively (Table 1) To define the binding sites of Nb6 and Nb11, we determined their cryogenic electron 156 microscopy (cryo-EM) structures bound to Spike* ( Fig. 2A state RBDs only contacts a single RBD (Fig. 2D) . 175 176 Nb3 interacts with the Spike S1 domain external to the RBD 177

Our attempts to determine the binding site of Nb3 by cryo-EM proved unsuccessful. We 178 therefore turned to radiolytic hydroxyl radical footprinting to determine potential binding sites for 179

Nb3. Spike*, either apo or bound to Nb3, was exposed to 5-50 milliseconds of synchrotron X-ray 180 radiation to label solvent-exposed amino acids with hydroxyl radicals. Radical-labeled amino 181 acids were subsequently identified and quantified by mass spectrometry of trypsin/Lys-C or Glu-182 C protease digested Spike*(18). Two neighboring surface residues on the S1 domain of Spike 183 (M177 and H207) emerged as highly protected sites in the presence of Nb3 We assessed multivalent Nb6 binding to Spike* by SPR. Both bivalent Nb6 with a 15 amino acid 204 linker (Nb6-bi) and trivalent Nb6 with two 20 amino acid linkers (Nb6-tri) dissociate from Spike* 205 in a biphasic manner. The dissociation phase can be fitted to two components: a fast phase with 206 kinetic rate constants kd1 of 2.7x10 -2 s -1 for Nb6-bi and 2.9x10 -2 s -1 for Nb6-tri, which are of the 207 same magnitude as that observed for monovalent Nb6 (kd = 5.6x10 -2 s -1 ) and a slow phase that 208 is dependent on avidity (kd2 = 3.1x10 -4 for Nb6-bi and kd2 < 1.0x10 -6 s -1 for Nb6-tri, respectively) 209 ( Fig. 3A) . The relatively similar kd for the fast phase suggests that a fraction of the observed 210 binding for the multivalent constructs is nanobody binding to a single Spike* RBD. By contrast, 211 the slow dissociation phase of Nb6-bi and Nb6-tri indicates engagement of two or three RBDs. 212

We observed no dissociation for the slow phase of Nb6-tri over 10 minutes, indicating an upper 213 boundary for kd2 of 1x10 -6 s -1 and subpicomolar affinity. This measurement remains an upper-214 bound estimate rather than an accurate measurement because the technique is limited by the 215 intrinsic dissociation rate of Spike* from the chip imposed by the chemistry used to immobilize 216 Spike*. 217

We reasoned that the biphasic dissociation behavior could be explained by a slow 219 interconversion between up-and down-state RBDs, with conversion to the more stable down-220 state required for full trivalent binding. According to this view, a single domain of Nb6-tri 221 engaged with an up-state RBD would dissociate rapidly. The system would then re-equilibrate 222 as the RBD flips into the down-state, eventually allowing Nb6-tri to trap all RBDs in closed 223 Spike*. To test this notion directly, we varied the time allowed for Nb6-tri binding to Spike*. 224

Indeed, we observed an exponential decrease in the percent fast-phase with a t1/2 of 65 s ( Table 1 ). Nb6-tri shows a 2000-fold enhancement of inhibitory activity, with 237 an IC50 of 1.2 nM, whereas trimerization of Nb11 and Nb3 resulted in more modest gains of 40-238 and 10-fold (51 nM and 400 nM), respectively (Fig. 3C) . 239 240 We next confirmed these neutralization activities with a viral plaque assay using live SARS-241

Nb6-tri proved exceptionally potent, neutralizing SARS-CoV-2 with an average IC50 of 160 pM 243 (Fig. 3D ). Nb3-tri neutralized SARS-CoV-2 with an average IC50 of 140 nM (Fig. 3D) . 244 245

We further optimized the potency of Nb6 by selecting high-affinity variants. To this end, we 247 prepared a new library, starting with the Nb6 coding sequence, in which we varied each amino 248 acid position of all three CDRs by saturation mutagenesis (Fig. 4A) . After two rounds of 249 magnetic bead-based selection, we isolated a population of high-affinity clones. Sequencing 250 revealed two highly penetrant mutations: I27Y in CDR1 and P105Y in CDR3. We incorporated 251 these two mutations into Nb6 to generate matured Nb6 (mNb6), which binds with 500-fold 252 increased affinity to Spike* as measured by SPR (Fig. 4B) . As a monomer, mNb6 inhibits both 253 pseudovirus and live SARS-CoV-2 infection with low nanomolar potency, a ~200-fold 254 improvement compared to Nb6 ( Fig. 4I -J, Table 1 ). 255 256 A 2.9 Å cryo-EM structure of mNb6 bound to Spike* shows that, like the parent nanobody Nb6, 257 mNb6 binds to closed Spike (Fig. 4C, Supplementary Fig. 7) . The higher resolution map allowed 258 us to build a model with high confidence and determine the effects of the I27Y and P105Y 259 substitutions. mNb6 induces a slight rearrangement of the down-state RBDs as compared to 260 both previously determined structures of apo-Spike* and Spike* bound to Nb6, inducing a 9° 261 rotation of the RBD away from the central three-fold symmetry axis (Fig. 4H) (14, 15) . This 262 deviation likely arises from a different interaction between CDR3 and Spike*, which nudges the 263 RBDs into a new resting position. While the I27Y substitution optimizes local contacts between 264 CDR1 in its original binding site on the RBD, the P105Y substitution leads to a marked 265 rearrangement of CDR3 in mNb6 (Fig. 4F-G) . This conformational change yields a different set 266 of contacts between mNb6 CDR3 and the adjacent RBD (Fig. 4D) . Remarkably, an X-ray crystal 267 structure of mNb6 alone revealed dramatic conformational differences in CDR1 and CDR3 268 between free and Spike*-bound mNb6, suggestive of significant conformational heterogeneity 269 for the unbound nanobodies and induced-fit rearrangements upon binding to Spike* (Fig. 4E) . 270

The binding orientation of mNb6 is similar to that of Nb6, supporting the notion that our 272 multivalent design would likewise enhance binding affinity. Unlike Nb6-tri, trivalent mNb6 273 (mNb6-tri) bound to Spike with no observable fast-phase dissociation and no measurable 274 dissociation over ten minutes, yielding an upper bound for the dissociation rate constant kd of 275 1.0x10 -6 s -1 (t1/2 > 8 days) and a KD of <1 pM (Fig. 4B) . As above, more precise measurements 276 of the dissociation rate are precluded by the surface chemistry used to immobilize Spike*. 277 278 mNb6-tri displays further gains in potency in both pseudovirus and live SARS-CoV-2 infection 279 assays with IC50 values of 120 pM (5.0 ng/mL) and 54 pM (2.3 ng/mL), respectively (Fig. 4H-I,  280 Table 1). Given the sub-picomolar affinity observed by SPR, it is likely that these viral 281 neutralization potencies reflect the lower limit of the assays. mNb6-tri is therefore an 282 exceptionally potent SARS-CoV-2 neutralizing antibody, among the most potent molecules 283 disclosed to date. 284 285 Nb6, Nb6-tri, mNb6, and mNb6-tri are robust proteins 286

One of the most attractive properties that distinguishes nanobodies from traditional monoclonal 287 antibodies is their extreme stability (21). We therefore tested Nb6, Nb6-tri, mNb6, and mNb6-tri 288 for stability regarding temperature, lyophilization, and aerosolization. Temperature denaturation 289 experiments using circular dichroism measurements to assess protein unfolding revealed 290 melting temperatures of 66.9, 62.0, 67.6, and 61.4 °C for Nb6, Nb6-tri, mNb6 and mNb6-tri, 291 respectively ( Fig 5A) . Aerosolization and prolonged heating of Nb6, mNb6, and mNb6-tri for 1 292 hour at 50°C induced no loss of activity (Fig 5B) . Moreover, mNb6 and mNb6-tri were stable to 293 lyophilization and to aerosolization using a mesh nebulizer, showing no aggregation by size 294 exclusion chromatography and preserved high affinity binding to Spike* (Fig. 5C-D) . 295

There is a pressing need for prophylactics and therapeutics against SARS-CoV-2 infection. 298

Most recent strategies to prevent SARS-CoV-2 entry into the host cell aim at blocking the 299 ACE2-RBD interaction. High-affinity monoclonal antibodies, many identified from convalescent 300 patients, are leading the way as potential therapeutics (22-29). While highly effective in vitro, 301 these agents are expensive to produce by mammalian cell expression and need to be 302 intravenously administered by healthcare professionals (30). Moreover, large doses are likely to 303 be required for prophylactic viral neutralization, as only a small fraction of systemically 304 circulating antibodies cross the epithelial cell layers that line the airways (31). By contrast, single 305 domain antibodies (nanobodies) provide significant advantages in terms of production and 306 deliverability. They can be inexpensively produced at scale in bacteria (E. coli) or yeast (P. 307 pastoris). Furthermore, their inherent stability enables aerosolized delivery directly to the nasal 308 and lung epithelia by self-administered inhalation (32). 309 310 Monomeric mNb6 is among the most potent single domain antibodies neutralizing SARS-CoV-2 311 discovered to date. Multimerization of single domain antibodies has been shown to improve 312 target affinity by avidity (32, 33) . In the case of Nb6 and mNb6, however, our design strategy 313 enabled a multimeric construct that simultaneously engages all three RBDs, yielding profound 314 gains in potency. Furthermore, because RBDs must be in the up-state to engage with ACE2, 315 conformational control of RBD accessibility can serve as an added neutralization mechanism. 316 Indeed, our Nb6-tri and mNb6-tri molecules were designed with this functionality in mind. 

SARS-CoV-2 Seroconversion in Humans: A Detailed Protocol for a 435

Serological Assay, Antigen Production, and Test Setup

Trimeric SARS-CoV-2 Spike interacts with dimeric ACE2 with limited intra-438 Spike avidity. bioRxiv

Yeast surface display platform for rapid discovery of conformationally 440 selective nanobodies

Automated electron microscope tomography using robust prediction 442 of specimen movements

MotionCor2: anisotropic correction of beam-induced motion for 444 improved cryo-electron microscopy

cryoSPARC: algorithms for 446 rapid unsupervised cryo-EM structure determination

New tools for automated high-resolution cryo-EM structure 448 determination in RELION-3

Grigorieff, cisTEM, user-friendly software for single-particle image 450 processing

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the 452 ACE2 receptor

Structure of a nanobody-stabilized active state of the β(2) 454 adrenoceptor

RosettaES: a sampling 456 strategy enabling automated interpretation of difficult cryo-EM maps

Coot: model-building tools for molecular graphics

ISOLDE: a physically realistic environment for model building into low-461 resolution electron-density maps

PHENIX: a comprehensive Python-based system for macromolecular 463 structure solution

Allosteric nanobodies reveal the dynamic range and diverse 465 mechanisms of G-protein-coupled receptor activation

The Beamline X28C of the 467 Center for Synchrotron Biosciences: a national resource for biomolecular structure and 468 dynamics experiments using synchrotron footprinting

Fast quantitative analysis of timsTOF PASEF data with MSFragger and 471 IonQuant

MSstats: an R package for statistical analysis of quantitative mass 473 spectrometry-based proteomic experiments

Automatic processing of rotation diffraction data from crystals of initially 475 unknown symmetry and cell constants

Phaser crystallographic software

BUSTER version 1.10.0. . Cambridge, United 481 Kingdom: Global Phasing Ltd

Figure 2. Cryo-EM structures of Nb6 and Nb11 bound to Spike. A, Cryo-EM maps of Spike*-601

Nb6 complex in either closed (left) or open (right) Spike* conformation. B, Cryo-EM maps of 602

Spike*-Nb11 complex in either closed (left) or open (right) Spike* conformation. The top views 603 show receptor binding domain (RBD) up-or down-states. C, Nb6 straddles the interface of two 604 down-state RBDs, with CDR3 reaching over to an adjacent RBD. D, Nb11 binds a single RBD in 605 the down-state (displayed) or similarly in the up-state

Nb11 in either RBD up-or down-state. E, Comparison of RBD epitopes engaged by ACE2 607 (purple), Nb6 (red), or Nb11 (green)

Multivalency improves nanobody affinity and inhibitory efficacy. A, SPR of Nb6 611 and multivalent variants. Red traces show raw data and black lines show global kinetic fit for

Nb6 and independent fits for association and dissociation phases for Nb6-bi and Nb6-tri

Dissociation phase SPR traces for Nb6-tri after variable association time ranging from

Curves were normalized to maximal signal at the beginning of the dissociation phase. Percent 615 fast phase is plotted as a function of association time (right) with a single exponential fit. n = 3 616 independent biological replicates. C, Inhibition of pseudotyped lentivirus infection of ACE2

expressing HEK293T cells. n = 3 biological replicates for all but Nb11-tri (n = 2) D, Inhibition of 618 live SARS-CoV-2 virus. Representative biological replicate with n = 3 (right panel) or 4 (left 619 panel) technical replicates per concentration. n = 3 biological replicates for all but Nb3 and Nb3-620 tri (n = 2)

dissociation was observed for mNb6-tri over 10 minutes. C, Cryo-EM structure of Spike*-mNb6

Comparison of receptor binding domain (RBD) engagement by Nb6 and mNb6

demonstrating changes in Nb6 and 630 mNb6 position and the adjacent RBD. E, Comparison of mNb6 complementarity determining 631 regions in either the cryo-EM structure of the Spike*-mNb6 complex or an X-ray crystal structure 632 of mNb6 alone. F, CDR1 of Nb6 and mNb6 binding to the RBD. As compared to I27 in Nb6

Nb6 and mNb6 binding to the RBD demonstrating a large conformational 635 rearrangement of the entire loop in mNb6. H, Comparison of closed Spike* bound to mNb6 and 636

Rotational axis for RBD movement is highlighted. I, Inhibition of pseudotyped lentivirus 637 infection of ACE2 expressing HEK293T cells by mNb6 and mNb6-tri. n = 3 biological replicates 638

J, mNb6 and mNb6-tri inhibit SARS-CoV-2 infection of VeroE6 cells in a plaque assay

Representative biological replicate with n = 4 technical replicates per concentration. n = 3 640 biological replicates for all samples

Average values from n = 2 biological replicates for Nb12, Nb17, and Nb11-tri are presented

c Average values from n = 2 biological replicates for Nb3, Nb3-bi, and Nb3-tri. n = 3 biological 658 replicates for all others

NB -no binding 660 NC -no competition 661 NP -not performed 662

We thank the entire Walter and Manglik labs for facilitating the development and rapid execution 485 of this large-scale collaborative effort. We thank Sebastian Bernales and Tony De Fougerolles 486 for advice and helpful discussion, and Jonathan Weissman for input into the project and reagent 487 and machine use. We thank Jim Wells for providing the ACE2 ECD-Fc construct, Jason 488McLellan for providing Spike, RBD, and ACE2 constructs, and Florian Krammer for providing an 489 RBD construct. We thank Jesse Bloom for providing the ACE2 expressing HEK293T 

I 9.0x10 5 5.3x10 -1 5.9x10 -7 1.0x10 6 9.9x10 -1 9.7x10 -7 8.3x10 -6 (1.7x10 -6 ) NP NP Nb3 II 1.8x10 6 1.1x10 -1 6.1x10 -8 NB NC 3.9x10 -6 (7.9x10 -7 )3.0x10 -6 (3.2x10 -7 )Nb6 I 2.7x10 5 5.6x10 -2 2.1x10 -7 2.1x10 6 8.7x10 -2 4.1x10 -8 3.7x10 -7 (4.9x10 -8 ) 2.0x10 -6 (3.5x10 -7 )3.3x10 -6 (7.2x10 -7 )Nb8 I 1.4x10 5 8.1x10 -1 5.8x10 -6 6.6x10 5 3.3x10 -1 5.1x10 -7 4.8x10 -6 (4.9x10 -7 ) NP NP Nb11 I 1.2x10 6 1.6x10 -1 1.4x10 -7 3.2x10 6 2.4x10 -1 7.6x10 -8 5.4x10 -7 (1.2x10 -7 )2.4x10 -6 (5.4x10 -7 ) NP Nb12 I 1.2x10 2 2.0x10 -4 1.6x10 -6 Biphasic Biphasic Biphasic 2.5x10 -7 (5.5x10 -8 )1.2x10 -6 (9.0x10 -7 ) NP Nb15 I 1.7x10 5 2.3x10 -1 1.3x10 -6 6.0x10 5 2.2x10 -1 3.6x10 -7 2.2x10 -6 (2.5x10 -7 )6.7x10 -6 (3.6x10 -6 ) NP Nb16 I 1.1x10 5 1.3x10 -1 1.3x10 -6 NP 9.5x10 -7 (1.1x10 -7 ) NP NP Nb17 II 7.3x10 5 2.0x10 -1 2.7x10 -7 NB NC 7.6x10 -6 (1.0x10 -6 ) NP Nb18 II 1.4x10 5 6.4x10 -3 4.5x10 -8 NB 5.2x10 -5 (1.5x10 -5 ) NP NP Nb19 I 2.4x10 4 1.1x10 -1 4.5x10 -6 1.0x10 5 8.9x10 -2 8.8x10 -7 4.1x10 -6 (4.9x10 -7 )2.4x10 -5 (7.7x10 -6 ) NP Nb24 I 9.3x10 5 2.7x10 -1 2.9x10 -7 2.4x10 6 3.5x10 -1 1.5x10 -7 7.5x10 -7 (1.0x10 -7 ) NP NP ACE2 N/A 2.7x10 5 1.2x10 -2 4.4x10 -8 NP NP NP 1.7x10 -7 (6.6x10 -8 ) 6.2x10 -7 (1.7x10 -7 ) NP mNb6 I 1.0x10 6 4.5x10 -4 4.5x10 -10 1.1x10 6 6.4x10 -4 5.6x10 -10 1.3x10 -9 (4.1x10 -10 ) 6.3x10 -9 (1.6x10 -9 )