key: cord-323967-2mo915u1
authors: Miersch, Shane; Li, Zhijie; Saberianfar, Reza; Ustav, Mart; Blazer, Levi; Chen, Chao; Ye, Wei; Pavlenco, Alia; Subramania, Suryasree; Singh, Serena; Ploder, Lynda; Ganaie, Safder; Leung, Daisy; Chen, Rita E.; Case, James Brett; Novelli, Guiseppe; Matusali, Giulia; Colavita, Francesca; Copabianchi, Maria R.; Jain, Suresh; Gupta, J.B.; Amarasinghe, Gaya; Diamond, Michael; Rini, James; Sidhu, Sachdev S.
title: Tetravalent SARS-CoV-2 Neutralizing Antibodies Show Enhanced Potency and Resistance to Escape Mutations
date: 2020-11-01
journal: bioRxiv
DOI: 10.1101/2020.10.31.362848
sha: 
doc_id: 323967
cord_uid: 2mo915u1

Recombinant neutralizing antibodies (nAbs) derived from recovered patients have proven to be effective therapeutics for COVID-19. Here, we describe the use of advanced protein engineering and modular design principles to develop tetravalent synthetic nAbs that mimic the multi-valency exhibited by IgA molecules, which are especially effective natural inhibitors of viral disease. At the same time, these nAbs display high affinity and modularity typical of IgG molecules, which are the preferred format for drugs. We show that highly specific tetravalent nAbs can be produced at large scale and possess stability and specificity comparable to approved antibody drugs. Moreover, structural studies reveal that the best nAb targets the host receptor binding site of the virus spike protein, and thus, its tetravalent version can block virus infection with a potency that exceeds that of the bivalent IgG by an order of magnitude. Design principles defined here can be readily applied to any antibody drug, including IgGs that are showing efficacy in clinical trials. Thus, our results present a general framework to develop potent antiviral therapies against COVID-19, and the strategy can be readily deployed in response to future pathogenic threats.

To date, all clinically advanced candidate nAbs against SARS-CoV-2 infection have been 72 derived by cloning from B-cells of recovered COVID-19 patients or from other natural 73 sources 9,17,19-23 . Here, we applied an alternative strategy using in vitro selections with phage-74 displayed libraries of synthetic Abs built on a single human framework derived from the highly 75 validated drug trastuzumab. This approach enabled the rapid production of high affinity nAbs 76 with properties optimized for drug development. Moreover, the use of a highly stable framework 77 enabled facile and modular design of ultra-high affinity nAbs in tetravalent formats that retained 78 favorable drug-like properties and exhibited neutralization potencies that greatly exceeded those 79 of the bivalent IgG format. These methods provide a general means to rapidly improve the 80 potency of virtually any nAb targeting SARS-CoV-2 and its relatives, and thus, our strategy can be 81 applied to improve COVID-19 therapies and can be adapted in response to future pathogenic 82 threats. 83 84 85 RESULTS 86

Using a phage-displayed human antigen-binding fragment (Fab) library similar to the 88 highly validated library F 24 , we performed four rounds of selection for binding to the biotinylated 89 RBD of SARS-CoV-2 immobilized on streptavidin-coated plates. Screening of 384 clones for 90 binding to CoV-2 RBD, revealed 348 Fab-phage clones that bound to the RBD but not to 91 streptavidin. Fab-phage were screened by ELISA and those that exhibited >50% loss in binding 92 to RBD in the presence of 200 nM ACE2 were sequenced, revealing 34 unique clones (Fig. 1A) , 93 deemed to be potential nAbs and converted into the full-length human IgG1 format for 94 purification and functional characterization. 95

To estimate affinities, ELISAs were performed with serial dilutions of IgG protein binding 96 to biotinylated S protein trimer captured with immobilized streptavidin, and these assays showed 97 that three IgGs bound with EC50 values in the sub-nanomolar range (Fig. 1B,C and Table 1 ). ELISAs 98 also confirmed that each IgG could partially block the binding of biotinylated ACE2 to immobilized 99 S protein (Fig. 1D) . Moreover, similar to the highly specific IgG trastuzumab, ELISAs showed that 100 the three IgGs did not bind to seven immobilized proteins that are known to exhibit high non-101 specific binding to some IgGs, and lack of binding to these proteins has been shown to be 102 predictive of good pharmacokinetics in vivo (Fig. 1E) 25, 26 . We also used biolayer interferometry 103 (BLI) to measure binding kinetics and determine avidities more accurately, and all three 104 antibodies exhibited sub-nanomolar dissociation constants (Table 1, Fig. S1 ), in close accord with 105 the estimates determined by ELISA. IgG 15033 exhibited the highest avidity, which was mainly 106 due to a two-or seven-fold faster on-rate than IgG 15031 or 15032, respectively, and thus, we 107 focused further efforts on this Ab. 108

We took advantage of the precision design of our synthetic Ab library to rapidly improve 109 the affinity of Ab 15033. The synthetic library was designed with tailored diversification of key 110 positions in all three heavy chain complementarity-determining regions (CDRs) and the third CDR 111 of the light chain (CDR-L3). Consequently, we reasoned that the already high affinity of Ab 15033 112 could be further improved by recombining the heavy chain with a library of light chains with naïve 113 diversity in CDR-L3. Following selection for binding to the RBD, the light chain library yielded 114 numerous variants, 17 of which were purified in the IgG format and analyzed by BLI (Fig. S2) . 115   Several of the variant light chains resulted in IgGs with improved binding compared with IgG  116 15033, and in particular, IgG 15033-7 (Fig. 1B) exhibited significantly improved avidity (KD = 300 117 or 39 pM, respectively) due to an off-rate that was an order of magnitude slower (Table 1, Fig.  118   S2) . 119 120

To understand the molecular basis for antagonism of ACE2 binding, we solved the X-ray 122 crystal structures of the SARS-CoV-2 RBD in complex with Fab 15033 or 15033-7 at 3.2 or 3.0 Å 123 resolution, respectively ( Fig. 2A) . As expected, backbone superposition showed that the two 124 complexes were essentially identical (RMSD = 0.17 Å). However, there were differences in side 125 chain interactions due to sequence differences in the CDR-L3 loop, which explained the enhanced 126 affinity of Fab 15033-7 compared with Fab 15033 (Fig. 2B) . Although the side chains of Tyr 108L in 127

Fab 15033 and His 108L in Fab 15033-7 both make hydrogen bonds with the side chain of Tyr 473 in 128 the RBD, the bond mediated by His 108L is shorter, and thus, likely to be stronger. Moreover, in 129

Fab 15033-7, the side chain of His 108L also makes an intramolecular hydrogen bond with the side 130 chain of Thr 109L , which Tyr 108L and Arg 109L are incapable of making in Fab 15033, and this 131 interaction may stabilize the CDR-L3 loop of Fab 15033-7 in a conformation that is favorable for 132 antigen recognition. Thus, the crystal structures show that the two substitutions in the CDR-L3 133 loop of Fab 15033-7 relative to Fab 15033 act in a cooperative manner to mediate favorable 134 intermolecular contacts with the RBD, and also, intramolecular interactions that stabilize the loop 135 in a conformation that may be better positioned to interact with the RBD. 136

We next analyzed the structures to understand how the Abs could function as antagonists 137 of RBD binding to ACE2. Binding of Fab 15033-7 to the RBD involves an extensive interface, with 138 1130 and 1112 Å 2 of surface area buried on the epitope or paratope, respectively, and 59% or 139 41% of the structural paratope is formed by the light or heavy chain, respectively (Fig. 2C) . 140

Comparison of the Fab and ACE2 epitopes on the RBD revealed extensive overlap, with 79% or 141 69% of the Fab or ACE2 epitope occluded by the other ligand (Fig. 2C) . Thus, direct steric 142 hinderance explains the blockade of ACE2 binding by Fabs 15033 and 15033-7 (Fig. 1D) . 143 We also used cryogenic electron microscopy to visualize Fab 15033 in complex with the S 144 protein trimer (Fig. S3A) . This analysis revealed that all three RBDs in a single trimer were 145 positioned in an "up" conformation, which was similar to the conformation bound to ACE2, and 146 the three RBDs were bound to three Fab molecules. Notably, the C-termini of the three Fabs were 147 positioned close to each other and pointed away from the S protein, suggesting that a single IgG 148 may be able to present two Fabs in a manner that would enable simultaneous engagement of 149 two RBDs on a single S protein. Indeed, this was confirmed in single particle negative stain 150 electron micrographs of IgG 15033 and the S protein, which revealed that the two Fabs of a single 151 IgG bound two RBDs on a single S protein trimer with a pincer-like grip (Fig. S3B) . Taken together, 152 the X-ray crystallography and electron microscopy showed that Fabs 15033 and 15033-7 block 153 ACE2 binding to RBD by direct steric hinderance, and simultaneous binding of Fabs to multiple 154

RBDs on the S protein trimer enables the IgGs to inhibit ACE2 binding with enhanced potency due 155 to avidity. 156 157

Next, we explored whether we could further enhance the avidity of nAbs by taking 159 advantage of modular design strategies to engineer tetravalent formats. Each SARS-CoV-2 160 particle displays multiple S protein trimers, suggesting that multivalent Fab binding could 161 enhance avidity, especially since a single IgG 15033 molecule can utilize both Fab arms to bind a 162 single S protein trimer. We reasoned that additional Fab arms added to an IgG may further 163 enhance avidity by interacting with RBDs on S protein trimers close to the trimer engaged by the 164 core IgG. Thus, we designed tetravalent versions of 15033 and 15033-7 by fusing additional Fabs 165 to either the N-or C-terminus of the IgG heavy chain to construct molecules termed Fab-IgG or 166

IgG-Fab, respectively (Fig. 3A) . Consistent with our hypothesis, the tetravalent molecules 167 exhibited higher avidity, and consequently, greatly reduced off rates compared with their 168 bivalent counterparts, and dissociation constants were in the low single-digit picomolar range 169 ( Fig. 3B, Table 1 ). 170

Our ultimate aim was to produce therapeutic Abs that could be used to treat COVID-19 in 171 patients. Aside from high affinity and specificity, effective Ab drugs must also possess favorable 172 biophysical properties including high yields from recombinant expression in mammalian cells, 173 high thermodynamic stability, and lack of aggregation and excessive hydrophobic surface area. 174

All IgGs and tetravalent molecules were produced in high yields by transient expression in 175

Expi293F cells (160-200 mg/L, Table 1 ). All proteins were highly thermostable with melting 176 temperatures of the CH3/Fab domain ranging from 81-87 o C, which exceeded the melting 177 temperature of the trastuzumab Fab (79.5 o C, Table 1 ). Size exclusion chromatography revealed 178 that all IgGs eluted as a predominant monodisperse single peak with elution volumes nearly 179 identical to that of trastuzumab ( Fig. 3C and Table 1) , and the monomeric fraction was calculated 180 to be 91 to >95% ( Table 1) To explore neutralization of potential escape mutants, we generated HIV-gag-based 205 lentivirus-like particle (VLPs) pseudotyped with the SARS-CoV-2 S protein. We confirmed ACE2-206 dependent uptake of the pseudotyped VLPs by HEK-293 cells stably over-expressing exogenous 207 ACE2, and we showed that uptake was inhibited by either Fc-tagged RBD (RBD-Fc) or IgG 15033. 208

Within this system, we generated a panel of 44 pseudotyped VLP variants, each containing a 209 single alanine substitution at an RBD position within or close to the ACE2-binding site. Twenty of 210 these VLP variants exhibited a >4-fold reduction in internalization compared with the wild-type 211 (wt) VLP, suggesting that these wt side chains contributed favorably to the interaction between 212 the RBD and ACE2. The remaining 24 VLP variants were internalized with high efficiency, and 213 these represent good mimics of escape mutants, which maintain strong ACE2-mediated 214 infectivity but may potentially reduce binding to nAbs that compete directly with ACE2. 215

With the panel of 24 VLP variants that mimicked potential escape mutants, we surveyed 216 the effects on cellular uptake after treatment with various nAbs (Fig. 4B) . We defined as escape 217 mutants those VLP variants for which cellular uptake in the presence of 50 nM nAb was >5% of 218 the uptake in the absence of the nAb. Based on this definition, we found that 6 of the mutations 219 enabled escape from IgG 15033, whereas only three mutations enabled escape from IgG 15033-220 7. Presenting the 15033 paratope in tetravalent formats resulted in nAbs that could neutralize 221 more variants than IgG 15033, and most importantly, tetravalent nAbs containing the 15033-7 222 paratope strongly neutralized all variants except one. As expected, these results showed that 223 enhancing the avidity of the IgG paratope for the S protein enhanced both potency and resistance 224 to escape mutations. Moreover, similar enhancements were also achieved by the presentation 225 of paratopes in tetravalent rather than bivalent formats, and the most effective nAbs were those 226 that presented the optimized paratope in the tetravalent format. 227 228 229 DISCUSSION 230 SARS-CoV-2 has wreaked havoc on global health and economics, and along with its relatives 231 SARS-CoV and MERS, has shown that viral outbreaks and pandemics will continue to plague the 232 world in the future. Consequently, it is essential for the scientific community to adapt the most 233 advanced drug development technologies to combat not only COVID-19, but also, pathogenic 234 disease in general. In this context, we have deployed advanced synthetic antibody engineering 235 to rapidly develop human nAbs, which are potent therapeutic candidates in the natural IgG 236 format, and are even better neutralizing agents in the synthetic tetravalent formats that our 237 modular design strategies enable. Most importantly, the enhanced affinities and potencies 238 afforded by tetravalent nAbs are achieved without compromising any of the favorable 239 characteristics that make IgG molecules ideal drugs. Moreover, tetravalent nAbs resist potential 240 escape mutants, which further augments the power of these molecules as drugs to combat not 241 only SARS-CoV-2, but also, its relatives that may emerge in the future. 242

COVID-19 has also exposed the need for drug development to respond to viral outbreaks 243 15031  15032  15033  15034  15035  15036  15037  15038  15039  15040  15041  15042  15043  15044  15045  15046  15047  15048  15049  15050  15051  15052  15053  15054  15055  15056  15057  15058  15059  15060  15061  15062  15063  15064   15033 Ab   28  29  36  37  38  56  57  65  107  108  109  114  115  116  30  35  36  37  38  39  55  56  57  58  59  62  63  64  65  66  107  108  109  110  111 For the RBD, residues in the ACE2-binding site are also shown as colored surfaces, and the following color scheme was used: red, contacts with both Fab 15033-7 and ACE2; blue, contacts with Fab 15033-7 only; yellow, contacts with ACE2 only. Fab 15033-7 residues that contact the RBD are colored magenta or cyan if they reside in the light or heavy chain, respectively. The CDR-L3 residues that differ between 15033-7 and 15033 are shown as red spheres. The VLPs were treated with 50 nM of the indicated nAb (x-axis) and uptake by ACE2-expressing HEK-293 cells was measured in triplicate and results are representative of n=2 independent experiments. The heat map shows uptake normalized to uptake in the absence of nAb. Boxed cells indicate VLPs that represented escape mutants for a given nAb, as defined by >5% uptake with nAb treatment compared with untreated control (the percent uptake is shown in each cell). 

Remdesivir for the Treatment of Covid-19 -Preliminary 605

A Randomized Trial of Hydroxychloroquine as 608

Postexposure Prophylaxis for Covid-19

Convergent antibody responses to SARS-CoV-2 in 611 convalescent individuals

Clinical 613 and immunological assessment of asymptomatic SARS-CoV-2 infections

Convalescent Plasma Therapy on Time to Clinical Improvement in Patients With Severe 617 and Life-threatening COVID-19: A Randomized Clinical Trial

Treatment 619 of 5 Critically Ill Patients with COVID-19 with Convalescent Plasma

Effectiveness 622 of convalescent plasma therapy in severe COVID-19 patients

A Potently Neutralizing Antibody Protects Mice against SARS-CoV-2

A human 628 neutralizing antibody targets the receptor-binding site of SARS-CoV-2

Potent neutralization of severe acute respiratory syndrome 632 (SARS) coronavirus by a human mAb to S1 protein that blocks receptor association

Potent cross-reactive neutralization of SARS coronavirus isolates by 636 human monoclonal antibodies

Human monoclonal antibody as 639 prophylaxis for SARS coronavirus infection in ferrets

Prophylactic and postexposure efficacy of a 642 potent human monoclonal antibody against MERS coronavirus

SARS-CoV-2 structure and replication 646 characterized by in situ cryo-electron tomography

Structures and distributions of SARS-CoV-2 spike proteins on intact 649 virions

Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease 652

Studies in humanized mice and convalescent humans yield a SARS-CoV-2 655 antibody cocktail

Cross-neutralization of SARS-CoV-2 by a human monoclonal 658 SARS-CoV antibody

Neutralizing Antibodies against SARS-CoV-2 Identified by High-Throughput Single-Cell 661

Sequencing of Convalescent Patients' B Cells

Isolation of potent SARS-CoV-2 neutralizing antibodies and protection from 664 disease in a small animal model

IgG-Neutralizing Monoclonal Antibodies Block the SARS-CoV-2 Infection

A panel of human neutralizing mAbs targeting 670

SARS-CoV-2 spike at multiple epitopes

Broad neutralization of SARS-related viruses by 673 human monoclonal antibodies

diversity is not required for antigen recognition by synthetic antibodies

Polyreactivity increases the apparent affinity of 679 anti-HIV antibodies by heteroligation

Biophysical properties of the clinical-stage antibody landscape

Title: LY-CoV555, a rapidly isolated potent 685 neutralizing antibody, provides protection in a non-human primate model of SARS

A strategy to prevent future epidemics similar to the 2019-nCoV 688 outbreak

A review of studies on animal reservoirs of the SARS coronavirus

A high through-put platform for recombinant antibodies to folded 693 proteins

Simple piggyBac transposon-based mammalian 695 cell expression system for inducible protein production

Structure of bacteriophage T4 698 fibritin: a segmented coiled coil and the role of the C-terminal domain

Site-specific biotinylation of purified proteins using BirA

Immunogenicity and structures of a rationally designed 704 prefusion MERS-CoV spike antigen

Diversity Synthetic Antibody Libraries Yield Novel Anti-EGFR Antagonists

High-throughput 709 screening of formulations to optimize the thermal stability of a therapeutic monoclonal 710 antibody

Neutralizing Antibody and Soluble ACE2 Inhibition of a Replication-Competent VSV-713

SARS-CoV-2 and a Clinical Isolate of SARS-CoV-2

A highly conserved 715 cryptic epitope in the receptor binding domains of SARS-CoV-2 and SARS-CoV

Molecular characterization of SARS-CoV-2 from the first case of 719 COVID-19 in Italy

A simple method of estimating fifty percent endpoints -ScienceOpen

Crystal violet assay for determining viability of 724 cultured cells

Primary structure of the Streptomyces enzyme endo-beta-N-acetylglucosaminidase H -726

Phaser 730 crystallographic software

Features and development of Coot

Towards 735 automated crystallographic structure refinement with phenix.refine

IMGT unique numbering for immunoglobulin and T cell receptor variable 739 domains and Ig superfamily V-like domains