key: cord-0296229-hfff43hm
authors: Hsu, Yen-Pang; Mukherjee, Debopreeti; Shchurik, Vladimir; Makarov, Alexey; Mann, Benjamin F.
title: Structural remodeling of SARS-CoV-2 spike protein glycans reveals the regulatory roles in receptor binding affinity
date: 2021-08-30
journal: bioRxiv
DOI: 10.1101/2021.08.26.457782
sha: 9f1c165d1e65c3f9b9b4c5fd0748a7c333eb9d0f
doc_id: 296229
cord_uid: hfff43hm

Glycans of the SARS-CoV-2 spike protein are speculated to play functional roles in the infection processes as they extensively cover the protein surface and are highly conserved across the variants. To date, the spike protein has become the principal target for vaccine and therapeutic development while the exact effects of its glycosylation remain elusive. Experimental reports have described the heterogeneity of the spike protein glycosylation profile. Subsequent molecular simulation studies provided a knowledge basis of the glycan functions. However, there are no studies to date on the role of discrete glycoforms on the spike protein pathobiology. Building an understanding of its role in SARS-CoV-2 is important as we continue to develop effective medicines and vaccines to combat the disease. Herein, we used designed combinations of glycoengineering enzymes to simplify and control the glycosylation profile of the spike protein receptor-binding domain (RBD). Measurements of the receptor binding affinity revealed the regulatory effects of the RBD glycans. Remarkably, opposite effects were observed from differently remodeled glycans, which presents a potential strategy for modulating the spike protein behaviors through glycoengineering. Moreover, we found that the reported anti-SARS-CoV-(2) antibody, S309, neutralizes the impact of different RBD glycoforms on the receptor binding affinity. Overall, this work reports the regulatory roles that glycosylation plays in the interaction between the viral spike protein and host receptor, providing new insights into the nature of SARS-CoV-2. Beyond this study, enzymatic remodeling of glycosylation offers the opportunity to understand the fundamental role of specific glycoforms on glycoconjugates across molecular biology. Covert art Legends The glycosylation of the SARS-CoV-2 spike protein receptor-binding domain has regulatory effects on the receptor binding affinity. Sialylation or not determines the “stabilizing” or “destabilizing” effect of the glycans. (Protein structure model is adapted from Protein Data Bank: 6moj. The original model does not contain the glycan structure.) Significance Glycans extensively cover the surface of SARS-CoV-2 spike (S) protein but the relationships between the glycan structures and the protein pathological behaviors remain elusive. Herein, we simplified and harmonized the glycan structures in the S protein receptor-binding domain and reported their regulatory roles in human receptor interaction. Opposite regulatory effects were observed and were determined by discrete glycan structures, which can be neutralized by the reported S309 antibody binding to the S protein. This report provides new insight into the mechanism of SARS-CoV-2 S protein infection as well as S309 neutralization.

Glycans of the SARS-CoV-2 spike protein are speculated to play functional roles in the infection processes 18 as they extensively cover the protein surface and are highly conserved across the variants. To date, the 19 spike protein has become the principal target for vaccine and therapeutic development while the exact 20 effects of its glycosylation remain elusive. Experimental reports have described the heterogeneity of the 21 spike protein glycosylation profile. Subsequent molecular simulation studies provided a knowledge basis 22

of the glycan functions. However, there are no studies to date on the role of discrete glycoforms on the 23 spike protein pathobiology. Building an understanding of its role in SARS-CoV-2 is important as we 24

continue to develop effective medicines and vaccines to combat the disease. Herein, we used designed 25 combinations of glycoengineering enzymes to simplify and control the glycosylation profile of the spike 26 protein receptor-binding domain (RBD). Measurements of the receptor binding affinity revealed the 27 regulatory effects of the RBD glycans. Remarkably, opposite effects were observed from differently 28 remodeled glycans, which presents a potential strategy for modulating the spike protein behaviors 29 through glycoengineering. Moreover, we found that the reported anti-SARS-CoV-(2) antibody, S309, 30 neutralizes the impact of different RBD glycoforms on the receptor binding affinity. Overall, this work 31

reports the regulatory roles that glycosylation plays in the interaction between the viral spike protein and 32 host receptor, providing new insights into the nature of SARS-CoV-2. Beyond this study, enzymatic 33 remodeling of glycosylation offers the opportunity to understand the fundamental role of specific 34 glycoforms on glycoconjugates across molecular biology. 35

The glycosylation of the SARS-CoV-2 spike protein receptor-binding domain has regulatory effects on the 37 receptor binding affinity. Sialylation or not determines the "stabilizing" or "destabilizing" effect of the indicating effective camouflaging effects against antibody recognition. However, a notable exception was 63 found in the receptor-binding domain (RBD).(9) 64

The S protein RBD contains 222 residues (R319 to F541 residues) with two glycosylation sites located at 65 N331 and N343. Strong ACE2 interactions were found in the receptor-binding motif (S438 to G504 66 residues) through hydrogen bonds and salt bridges, as revealed by crystal structures.(20, 21) ( Figure 1B ) 67

Despite its crucial role in ACE2 binding, the RBD has the lowest coverage of glycan shielding among the 68 entire protein, which makes it vulnerable to immune recognization.(12, 13) One possible explanation to 69 this phenomenon is the existence of the "up and down" conformational change of the spike protein during 70 the cell entry processes, where the RBD remains buried by the heavily glycosylated S1 unit (the "down" 71 state) during trafficking for immune evasion until it engages ACE2 at the infection interface (turning into 72 the "up" state).(3, 22) This strategy minimizes the exposure time of RBD to the surrounding local 73 environment, reducing the probability of immune recognition. However, this explanation makes the roles 74 of the RBD glycans even more intriguing, especially for the glycans at N343 that is very close to the 75 receptor-binding motif (RBM).(20) We suspect that the RBD glycans could have roles additional to glycan 76 shielding in the S protein-ACE2 interactions. 77

In-depth probing of glycans' functions during viral infection is challenging, largely due to the lack of 78 strategies to control glycan structures and minimize their micro-heterogeneity.(23-26) As a result, 79 molecular dynamics (MD) simulation has become the predominant approach for studying glycan 80

biochemistry; yet support from experimental data is in great demand.(27) In this work, we report the 81 strategies to harmonize the glycans of SARS-CoV-2 S protein RBD into controlled glycoforms, as well as 82 subsequent binding affinity measurement between human ACE2 and the glycoengineered RBD. By the 83 designed combinations of glycoengineering enzymes that we have characterized, we successfully 84 transformed the S protein RBD glycans into 1) glycoforms with harmonized terminal glycan species; and 85

2) structure-defined single glycoforms.(28) This work reveals the regulatory roles of S protein RBD glycan 86 in receptor binding: a double-edged sword that can either stabilize or destabilize RBD-ACE2 interactions. 87

In combination with their roles in glycan shielding, these insights lay the foundations for modulating the 88 S protein's nature through glycan remodeling, which may create new strategies for vaccine and 89 therapeutic design. 90 91 92

Dissecting the glycoforms of SARS-CoV-2 S protein RBD 94 Glycans in the S protein RBD comprise N-glycan species.(6, 7) Their biosynthesis occurs in the endoplasmic 95 reticulum (ER) and continues in the Golgi in a species-, cell-, protein-, and site-specific manner. N-glycans 96 56.5% of the RBD glycan population contains at least one terminal sialic acid; 29.3% of the population exhibits at least one terminal galactose without having any sialic acid; 9.6% of the population terminating with GlcNAc without having any galactose and sialic acid.

share a common core structure made of two N-Acetylglucosamine (GlcNAc) and three mannose (Man)  97 residues. The core further extends into a myriad of glycoforms through the activity of glycosidases and 98

glycosyltransferases. It has been known that different glycoforms could lead to different regulatory effects 99 to protein-protein or cell-cell interactions. (29) 100 Given that glycan formation is sensitive to the expression conditions, we first analyzed the glycoforms of 101 the recombinant RBD (expressed from HEK293, GenScript Z03483) used in this study by isolating the 102 glycans using PNGase F and profiling them by liquid-chromatography mass spectrometry (LC-MS) analysis. 103

( Figure 1C ) Our substrate RBD exhibits heterogeneous glycoforms composed of complex-type species 104

(95.4%), with three antennae the highest number we observed, and small amount of high-mannose (1%) 105

and hybrid-type (3.6%) species. (SI-Data) As summarized in Figure 1D , GlcNAc, galactose, and sialic acid 106

(SA) appear as the terminal monosaccharide at the non-reducing ends of N-glycans with relative 107 abundances of 9.6%, 29.3%, and 56.5%, respectively. Over 99% of the RBD glycans contain the core fucose 108

and 18.6% of the glycoforms have additional fucose located at the complex-type glycan antennae, as 109 supported by glycosidase treatment experiments. (Figure S1 ) N-Acetylgalactosamine (GalNAc), the 110 epimer of GlcNAc, likely exists in the RBD glycans since N-Acetylglucosaminidase (GlcNAcase) alone was 111 not able to remove all the N-acetylhexosamine (HexNAc) residues, unless GalNAcase was also added. 112

( Figure S2 ) In addition, retention time comparison using glycan standards suggested that a small amount 113 of bisecting N-glycan species exist, as indicated by FA2B glycoform ([Man] 3[GlcNAc]5[Fuc]1). ( Figure S3 ) 114 We note that our analyses based on isolated glycans did not provide site-specific information for the 115 glycosidic linkages. The reported numbers here are averaged results from the glycan populations at N331 116 and N343. 117

The RBD glycans can stabilize RBD-ACE2 interaction 118

We used BLI to study the interaction between the S protein RBD and ACE2 expressed from HEK293. The 119

RBD with native glycoforms binds to ACE2 with an equilibrium dissociation constant (KD) of 99 ± 12 nM, 120 similar to earlier reports.(30) The terminal saccharides of mature glycans often regulate the biochemical 121

properties and functions of glycoconjugates. A good example comes from the human blood group 122 antigens that are classified by their terminal residue species.(31) Therefore, we aimed to harmonize the 123 RBD glycan termini and investigate their potential effect on the RBD-ACE2 interaction. First, we created 124 glycoengineered RBD glycoprotein with all N-glycans ending in terminal HexNAc by incubating the native 125 RBD with α2-3/6/8 neuraminidase, β1-4 galactosidase, and α1-2,3/4 fucosidases. (Figure 2 , Table S1 ) 126 Interestingly, we found that the resulting substrate, named tHexNAc-RBD, had improved binding affinity 127

to ACE2 with a KD value of 47 ± 8 nM. (Figure 3A -B) Direct comparison of ACE2 binding curves showed a 128 20% increase in binding response in tHexNAc-RBD compared to the native RBD. (Figure S4 ) Furthermore, 129

we prepared glycoengineered RBD bearing (i) the core glycan ( we introduced sialic acids to the terminus of the RBD glycans. The resulting tSA-RBD contains 159 heterogeneous glycoforms with over 99% of the population bearing at least one terminal sialic acid. 160

( Figure 3A , SI-Data) Unexpectedly, BLI measurement showed that the tSA-RBD has lower binding affinity 161 to ACE2 (KD = 130 ± 6 nM) compared to the Native and deglycosylated RBD. We reasoned this result could 162 be attributed to the electrostatic repulsion between the N343 RBD glycan and the ACE2 surface. 163

Computational calculation of electrostatic potential has shown that the ACE2 surface is predominantly 164 negative, including the area that N343 RBD glycan engages.(37) ( Figure 3C ) Therefore, sialylated glycoforms, 165

which are negatively charged, could cause electrostatic repulsion and destabilize RBD-ACE2 interactions. 166

Given that the native RBD has a heterogeneous glycan profile, the existence of the sialylated glycan 167 species possibly negates the stabilizing effects resulting from other glycoforms, which explains the weaker 168 ACE2 binding affinity in native RBD over non-sialylated glycoforms. Recent reports of site-specific glycan 169 mapping have also indicated that both native N343 RBD and N322 ACE2 glycosylation sites have a low content 170 of sialylated glycan species, further implying that sialylation may not be favored for RBD-ACE2  171 interactions.(7, 33) 172

Another possible explanation of the reduced ACE2 binding affinity found in tSA-RBD is the increased steric 173 hindrance caused by sialylation. Having over-constructed glycan structures nearby the receptor-binding 174 motif could prevent the S protein from approaching ACE2. To test this possibility, we constructed 175 harmonized mono-antennary glycans on RBD, as confirmed by LC-MS analysis ( Figure 3A, S7, SI-Data) . 176

The FA1G1-RBD bears an octa-saccharide with only one antenna connected to the α1-3-linked core 177

mannose. This was achieved by incubating the core-RBD with N-acetylglucosaminyltransferase I (also 178 known as MGAT1) and galactosyltransferase in one pot. Similarly, the FA1G1S1-RBD contains an additional 179 sialic acid and was prepared by introducing sialyltransferase to the reaction. BLI analyses showed that the 180 FA1G1-RBD has improved ACE2 binding affinity compared to native RBD; by contrast, FA1G1S1-RBD 181

showed weaker binding than native RBD, which is consistent with the data found earlier. (Figure 3B ) This 182 result suggested that steric hindrance is not the major cause of the ACE2-binding affinity reduction found 183 in the tSA-RBD. Instead, the electrostatic repulsion is more likely the reason. Our results, together, suggest 184 that the RBD glycans have regulatory effects on the RBD-ACE2 interactions: non-sialylated glycoforms 185 reinforce the binding while sialylated glycoforms destabilize it. 186

Fucose in the RBD glycans does not have apparent impacts on ACE2 binding 187

Glycosidase treatment provided direct evidence that the core structure of the complex-type RBD glycans 188 is fucosylated because afucosylated core was not detected. (Figure 3A , S1) The core fucose linked to the 189 reducing end GlcNAc has been a hot target of interest for pharmaceutical research because of its 190 regulatory effects on protein-protein interactions, such as immunoglobulin G (IgG) and its receptors.(38) 191

To investigate the role of fucosylation in the RBD-ACE2 interaction, we prepared partially defucosylated 192 RBD using α1-6 fucosidase to target the core fucose. A 10-day reaction converted approximately 45% 193 native glycans into non-fucosylated forms. (Figure S8 ) It is known that the activity of this enzyme is 194 glycoform-dependent, where lower structural complexity leads to higher enzyme activity.(28) Therefore, 195

we used the core-RBD as the substrate for the reaction and successfully remove fucose from over 90% of 196 the core glycans. (Figure S8 ) BLI measurement showed that both defucosylated RBD substrates have no 197 significant difference in ACE2 binding affinity compared to their parent species. (Figure 3B ) This result is 198 not surprising because the core fucose is shielded close to the RBD peptide backbone, which minimizes 199 its probability of interacting with ACE2 or ACE2 glycans. In addition to the core-fucose, approximately 200

18.6% RBD glycans contain α1-2 and α1-3/4 fucose which can be removed by corresponding fucosidases. 201 ( Figure S1 ). Their removal did not result in a significant difference in ACE2 binding affinity (KD = 110 ± 14 202 nM). values at nanomolar scale. However, no significant difference was found when the RBD glycan bore 214 different terminal saccharide species, so as the removal of the core fucose. (Figure 4B ) This result 215

suggested that the N343 RBD glycan is not essential for the RBD-S309 binding. The RBD-S309 interaction is 216 likely established on the amino acid epitope of the RBD. 217

Despite the N343 RBD glycan is not involved in the antibody binding, its structure has been sterically 218 hindered by the S309-RBD interaction, which could restrict its probability of interacting with ACE2 and/or 219 ACE2 glycans. (Figure 4C ) To test this possibility, we measured the ACE2 affinity of the RBD samples that 220 were bound to S309. As expected, comparable binding affinities (KD ≈ 90 nM) were observed on the native 221 RBD, tHexNAc-RBD and tSA-RBD that were saturated with S309. (Figure 4D ) Namely, S309 neutralizes the 222 regulatory effect of the RBD glycan. Given that the native RBD glycans tend to stabilize ACE2 interaction 223 owing to the low content of sialylated glycoforms, the RBD-S309 binding could compromise this stabilizing 224 effect and reduce the infectivity of the virus, which presents a new inhibitory mechanism in S309 225 neutralization. In addition, we found that the binding of S309 neutralizes the regulatory effects of the RBD glycan, 251

preventing the potential stabilization of ACE2 interaction through the glycan structure. Given that the 252 glycosylation sites of the RBD are highly conserved across its evolution reported to date, glycans could 253 provide new handles for developing strategies to combat the emerging virus variants ( Figure S9) . For 254 example, glycan remodeling could be applied to protein subunit vaccines to maximize the mimicry of viral 255 glycoproteins and improve the immune response. Generally speaking, this work demonstrates that it is 256 possible to efficiently screen the biological behavior of discrete protein glycoforms, an approach that will 257 greatly advance our fundamental knowledge of glycobiology and reinforce our ability to design targeted 258 strategies to defeat disease. 259

All data are available upon reasonable request to the corresponding authors. 261

Method development for glycan remodeling, binding affinity measurements, LC-MS analysis, and data 263 analysis was performed by Y-P. H. MALDI-ToF analysis was performed by D. M. and V. S. All the authors 264

were involved in the design of the research and manuscript drafting. 265

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a 276 Clinically Proven Protease Inhibitor

Functional assessment of cell entry and receptor usage for SARS-278

CoV-2 and other lineage B betacoronaviruses

Cell entry mechanisms of SARS-CoV-2

Structural and functional properties of SARS-CoV-2 282 spike protein: potential antivirus drug development for COVID-19

Structural features of coronavirus SARS-CoV-2 spike protein: Targets 285 for vaccination

Deducing the N-and O-glycosylation profile 287 of the spike protein of novel coronavirus SARS-CoV-2

Site-specific glycan analysis of the 289 SARS-CoV-2 spike

Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Analysis of the SARS-CoV-2 spike protein glycan 293 shield reveals implications for immune recognition

Exploitation of glycosylation in enveloped 295 virus pathobiology

The Highly Conserved Glycan at Asparagine 260 of HIV-1 gp120 Is 298 Indispensable for Viral Entry*

Beyond Shielding: The Roles of Glycans in the SARS-CoV-2 Spike Protein

Vulnerabilities in coronavirus glycan shields despite extensive glycosylation

Virus-Receptor Interactions of Glycosylated SARS-CoV-2 Spike and Human ACE2 304 Receptor

Structure and Immune Recognition of the HIV Glycan Shield

Trimeric HIV-1-Env Structures Define Glycan Shields from Clades A, B, 308 and G

Playing hide and seek: how glycosylation of the influenza virus hemagglutinin 310 can modulate the immune response to infection

Cryo-EM analysis of a feline coronavirus spike protein reveals a unique structure 312 and camouflaging glycans

Glycan shield and epitope masking of a coronavirus spike protein observed by 315 cryo-electron microscopy

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 317 receptor

Receptor binding and priming of the spike protein of SARS-CoV-2 for 319 membrane fusion

Distinct conformational states of SARS-CoV-2 spike protein

Protein Glycoengineering: An Approach for Improving Protein Properties

Protein glycosylation: new challenges and opportunities

Measurement of macro-and micro-heterogeneity of glycosylation in 327 biopharmaceuticals: a pharmacopeia perspective

Relating glycoprotein structural heterogeneity to function -329 insights from native mass spectrometry

Challenges and opportunities in connecting 331 simulations with experiments via molecular dynamics of cellular environments

Successive remodeling of IgG glycans using a solid-phase enzymatic platform

Glycosylation in health and disease

Structural basis of receptor recognition by SARS-CoV-2

Structures Common to Different Glycans

Hummer, Dual nature of human ACE2 glycosylation in binding to SARS-CoV-343 2 spike

Comprehensive characterization of N-and O-glycosylation of SARS-CoV-2 345 human receptor angiotensin converting enzyme 2

Letter to the Glyco-Forum: Substrate specificity of 347 Flavobacterium meningosepticum Endo F2 and Endo F3: purity is the name of the game

Identification of distinct endoglycosidase (endo) activities in 350 Flavobacterium meningosepticum: endo F1, endo F2, and endo F3. Endo F1 and endo H 351 hydrolyze only high mannose and hybrid glycans

Sialic Acid Receptors of Viruses

Spike Proteins of SARS-CoV and SARS-CoV-2 Utilize Different Mechanisms to Bind 355 With Human ACE2

Structural basis for improved efficacy of therapeutic antibodies on 357 defucosylation of their Fc glycans

Protective monotherapy against lethal Ebola virus infection by a potently 359 neutralizing antibody

Cross-neutralization of SARS-CoV-2 by a human monoclonal SARS-CoV antibody. 361

Chronic COVID syndrome: Need for an appropriate medical terminology for long-363 COVID and COVID long-haulers

Escape of SARS-CoV-2 501Y.V2 from neutralization by convalescent plasma

Discovery of a SARS-CoV-2 variant from the P.1 lineage harboring

K417T/E484K/N501Y mutations in Kofu

Early transmissibility assessment of the 369 N501Y mutant strains of SARS-CoV-2 in the United Kingdom

The anti-influenza virus drug, arbidol is an efficient inhibitor of SARS-CoV-2 in 372 vitro

The authors declare no competing interests. 267