key: cord-0302381-nbghk20z
authors: Capraz, Tümay; Kienzl, Nikolaus F.; Laurent, Elisabeth; Perthold, Jan W.; Föderl-Höbenreich, Esther; Grünwald-Gruber, Clemens; Maresch, Daniel; Monteil, Vanessa; Niederhöfer, Janine; Wirnsberger, Gerald; Mirazimi, Ali; Zatloukal, Kurt; Mach, Lukas; Penninger, Josef M.; Oostenbrink, Chris; Stadlmann, Johannes
title: Structure-guided glyco-engineering of ACE2 for improved potency as soluble SARS-CoV-2 decoy receptor
date: 2021-08-31
journal: bioRxiv
DOI: 10.1101/2021.08.31.458325
sha: 414462e43c6cbdf89a19dc7121b0aa06b4012251
doc_id: 302381
cord_uid: nbghk20z

Infection and viral entry of SARS-CoV-2 crucially depends on the binding of its Spike protein to angiotensin converting enzyme 2 (ACE2) presented on host cells. Glycosylation of both proteins is critical for this interaction. Recombinant soluble human ACE2 can neutralize SARS-CoV-2 and is currently undergoing clinical tests for the treatment of COVID-19. We used 3D structural models and molecular dynamics simulations to define the ACE2 N-glycans that critically influence Spike-ACE2 complex formation. Engineering of ACE2 N-glycosylation by site-directed mutagenesis or glycosidase treatment resulted in enhanced binding affinities and improved virus neutralization without notable deleterious effects on the structural stability and catalytic activity of the protein. Importantly, simultaneous removal of all accessible N-glycans from recombinant soluble human ACE2 yields a superior SARS-CoV-2 decoy receptor with promise as effective treatment for COVID-19 patients.

The rapid spread of Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2), the 55 causative pathogen of human coronavirus disease 2019 (COVID- 19) , has resulted in an unprecedented pandemic and worldwide health crisis. Similar to the beta-coronaviruses SARS-CoV and Middle Eastern Respiratory Syndrome (MERS)-CoV, SARS-CoV-2 is highly transmissible and can lead to lethal pneumonia and multi-organ failure. 1 For infection and viral entry, the Spike surface protein of SARS-CoV-2 binds to angiotensin converting enzyme 2 60 (ACE2) on host cells. 2, 3 Recombinant soluble human ACE2 (rshACE2) has been shown to bind Spike, 4 can effectively neutralize SARS-CoV-2 infections, 5, 6 and the corresponding drug candidate APN01 has undergone a phase 2 clinical trial for the treatment of hospitalized cases of COVID-19 (ClinicalTrials.gov Identifier: NCT04335136). A first case study of its use in a patient has been reported recently. 7 Additionally, an aerosol formulation of APN01 has been 65 developed and is currently undergoing Phase I clinical studies.

Multiple other therapeutic strategies attempt to target the Spike-ACE2 interaction, e.g. by development of neutralizing antibodies blocking the ACE2-binding site 8 or lectins that bind to glycans on the Spike surface. 9, 10 Using soluble ACE2 as a decoy receptor for Spike is particularly attractive, as it minimizes the risk that variants of concern may evade the treatment 70 through mutations as has been observed for antibodies. [11] [12] [13] Furthermore, protein engineering has yielded ACE2 variants with substantially improved affinities for Spike. 14, 15 Hence, soluble ACE2 based therapeutics offer considerable advantages over other therapeutic formats that aim to hamper the Spike-ACE2 interaction sterically.

Modern structural biology has been amazingly fast to respond to this pandemic. A mere three 75 months after identification of SARS-CoV-2 as the etiologic agent of COVID-19, structures of the complex between ACE2 and the receptor binding domain (RBD) 4, 16, 17 and of the ectodomain of trimeric Spike 18, 19 were already solved by X-ray crystallography or cryo-electron microscopy.

While this provided unprecedented insight into the protein-protein interactions between Spike and ACE2, the structural impact of protein-bound glycans on the Spike-ACE2 interface could 80 not be assessed experimentally so far due to their compositional diversity and conformational flexibility. Here, in silico modeling of the glycans offers a powerful alternative to study the effects of individual Spike and ACE2 glycans on the molecular interactions between these two proteins.

The SARS-CoV-2 Spike protein is heavily glycosylated with both complex and oligo-85 mannosidic type N-glycans, 1, 10, 20, 21 thereby shielding a large portion of the protein surface. [22] [23] [24] Similarly, ACE2 is a glycoprotein with up to seven highly utilized sites of N-glycosylation. 21 Recent computational studies started to investigate protein glycosylation in the context of the interaction between Spike and ACE2. 21, 22, 25 Extensive all-atom molecular dynamics (MD) simulations indicated that Spike N-glycans attached to N165 and N234 could be important 90 stabilizers of the ligand-accessible conformation of the receptor binding domain (RBD). 22 Furthermore, it has been proposed that the N-glycan at position N343 acts as a gate facilitating RBD opening. 26 Other MD studies concluded that the glycans attached to N90 and N322 of ACE2 could be major determinants of Spike binding, 25 while yet other simulation works postulate that glycosylation does not affect the RBD-ACE2 interaction significantly. 27, 28 Genetic 95 or pharmacological blockade of N-glycan biosynthesis at the oligomannose stage in ACE2expressing target cells was found to dramatically reduce viral entry, 29 even though several glycoforms of ACE2 were found to display comparatively moderate variation with respect to Spike binding. 30 Hence, a detailed understanding on how individual glycans on both Spike and ACE2 influence their interaction and a comprehensive experimental validation of the MD 100 findings is crucial for the rational design of novel therapeutic soluble ACE2 variants with enhanced Spike binding affinity and the capacity to block viral entry more efficiently than the native enzyme. 21 The identification of the Spike glycans essential for efficient association with ACE2 will be also critical to guide rational design of improved SARS-CoV-2 vaccines.

We started our research by creating 3D models of the trimeric Spike in complex with human ACE2 (hACE2). The RBD of Spike exists in two distinct conformations, referred to as "up" and "down". 18, 19 The "up" conformation corresponds to the receptor-accessible state with the RBD of one monomer exposed. By superimposing the RBD from the RBD-hACE2 complex 17 with the single RBD in the "up" conformation (monomer 3) of the trimeric Spike, 18 an initial model was 110 obtained. To assess the impact of all seven individual N-glycosylation sites of hACE2 on its interaction with Spike, we first elucidated the entire glycome of rshACE2 (Fig. S1 ). This also provided information on the glycans attached to N690, a glycosylation site not covered in previous glycoproteomic studies of soluble hACE2. 21, 30 For recombinant trimeric Spike the glyco-analysis has been reported elsewhere. 10, 20, 21, 31 Based on the site-specific glycosylation 115 profiles we added complex or oligo-mannosidic glycan trees to the respective sites of Spike and ACE2 (Table S1 ). We hence constructed fully glycosylated atomistic models of the trimeric Spike glycoprotein, free dimeric ACE2 and of the Spike glycoprotein in complex with dimeric hACE2 (Fig. 1 ). Using these fully glycosylated structures, we performed molecular dynamics simulations of the Spike-ACE2 complex (Video S1), and of free hACE2. Inspection of the most 120 important interacting residues on Spike and ACE2, their average distances and the electrostatic potential of the interface area identified critical contact sites (Figs. S2 and S3). We next quantified the complete solvent-accessible surface area (SASA) of the Spike protein in complex with ACE2, both with and without glycans. The average accessible area of protein atoms for non-glycosylated and glycosylated Spike was 1395 nm² and 864 nm², respectively,

indicating that glycans shield about 38% of the protein surface of Spike, a value that is 135 comparable to what was previously found in simulations of Spike alone. 23, 32 The area of protein atoms that are shielded by the individual glycans are shown in Fig. 2a and Fig. S4 .

Further analysis showed that glycans at N122, N165 and N343 on Spike directly interact with ACE2 or its glycans (Fig. 1b, c, Fig. 2b, c) . It has been reported that Spike mutants lacking the glycans at N331 and N343 display reduced infectivity, while elimination of the glycosylation 140 motif at N234 results in increased resistance to neutralizing antibodies, without reducing infectivity of the virus. 33 The equilibrium between the "up" and "down" conformations of Spike involves various stabilizing and destabilizing effects, with possible roles for the glycans at N165, N234, N331 and N343. 22, 26, 34 Removing the glycans at N165, N234 and N343 was experimentally seen to reduce binding to ACE2 by 10%, 40% and 56%, respectively. 22, 26 In our 145 MD simulations, the glycan at position N343 interacts directly with ACE2 ( Fig. 2) , while the glycan at N331 interacts with a neighboring Spike monomer (Fig. 1c, Fig. S5 ), indicating that the N331 glycosylation site only indirectly affects the interaction of Spike with ACE2. In our model, the glycan at N234 also does not interact directly with ACE2, but seems to stabilize the "up" conformation. Its removal could favor the "down" conformation of the RBD, possibly explaining 150 the observed more effective shielding against neutralizing antibodies. In agreement with previous simulations 22 the Man9 glycan at N234 of Spike partially inserts itself into the vacant space in the core of the trimer that is created when the RBD of monomer 3 is in the "up" conformation ( Fig. S6 ). In our simulations, the free space created by the "up" conformation seems slightly smaller for Spike in complex with ACE2, suggesting that binding to ACE2 has a stabilizing 155 effect on the Spike monomer.

The N165Q mutant was experimentally found to be more sensitive to neutralization. 33 In our models, the glycan at N165 is positioned directly next to the RBD (Fig. 1b) and thus could shield important antigenic sites. These data highlight the complex impact of Spike glycosylation on the intramolecular interactions of the Spike monomers and, critically, the interaction with ACE2, 160 posing a challenge to design SARS-CoV-2 neutralizing moieties. Since our modeling clearly confirmed that ACE2 glycosylation plays a significant role in its binding to Spike (Fig. 1d ,e), we also determined the area of the Spike-ACE2 interface region, by subtracting the SASA of the complex from the SASA of the individual proteins and dividing by 170 two. The total interface area was 24.6 nm², with glycans accounting for up to 51% of the interface area, i.e. 12.6 nm², contributed by the four most relevant glycans at positions N53, N90, N322 and N546 of ACE2 (Fig. 3) . Furthermore, we scored the number of atoms of each ACE2 glycan in contact with Spike. A contact was defined as a distance of less than 0.4 nm between two atoms. This allowed us to identify the glycans at N53, N90, N322 and N546 as interacting 175 with Spike, with the glycan at position N53 having the weakest interaction. Notably, N546 interacted with Spike for a significant amount of time only in one of the two independent simulations. The degree of interaction correlated with the spatial proximity between the glycans and the RBD (Fig. 1e, f) . Assessing the number of hydrogen bonds that formed during the simulations, the glycans at N90 and N322 appear most prominent (Fig. 3b) . Interestingly, the 180 glycans at N90 and N322 interact directly with Spike protein atoms, while the glycan at N546 (red sticks in Fig. 1f ) interacts with the glycans at N122 and N165 of Spike (dark green and orange sticks in Fig. 1b) . These findings are in agreement with previously reported simulations of the complexes. 

Next, we assessed the conformational freedom of ACE2 glycans upon binding to Spike and compared their respective density maps in the simulations of free ACE2, and ACE2 in complex with Spike (Fig. 4) . The density map of the unbound ACE2 (Fig. 4a) shows a continuous density of glycans, largely covering the interface area. Formation of the ACE2-Spike complex 195 significantly reduces the conformational freedom of the glycans, in particular the ones at N90 and N322 (Fig. 4b) . We predict that the glycans at N90 and N322 hamper binding to Spike, either sterically or through an entropic penalty upon binding due to a loss of conformational freedom. These glycans have been implicated as being relevant for binding before, 21 , as well as the glycan at N53, 35 but no conclusions were drawn if they contribute positively or negatively to 200 binding. Mehdipour and Hummer predicted the glycan at N322 to contribute favorably to binding, because of the favorable interactions of this glycan with the Spike surface. 25 We did not observe a significantly more pronounced interaction with Spike for the glycan at N322, compared to the one at N90 (Fig. 3) . Based on conformational considerations, we therefore rather predict a negative impact on binding for both glycans (Fig. 4) . Since only the glycans at N90 and N322 directly interact with the protein atoms of the Spike proteins, while the glycan on N546 forms hydrogen bonds with glycans present on Spike, we set out to confirm the negative influence of N90 and N322 glycosylation on the interactions with 215 Spike experimentally. First, we ablated N-glycosylation at N90 and N322 individually using the ACE2-Fc fusion constructs ACE2-T92Q-Fc (ref. 15 ) and ACE2-N322Q-Fc. Note that ref. 15 indeed suggests that removal of the glycan at N90 through a mutation of T92 leads to enhanced interaction with Spike. The same data set, however, suggests that removal of the glycan at N322 through a mutation of T324 most likely leads to reduced affinity to Spike. However, T324 is 220 itself part of the interface with Spike (Fig. S7) , and any mutation of this residue could easily disrupt ACE2 -Spike binding directly, rather than through its effect on the N322 glycosite. We therefore decided to mutate N322 into glutamine to prevent glycosylation at this position.

The wild-type and mutant ACE2-Fc constructs were expressed in HEK293-6E cells and purified from the culture supernatants by protein A affinity chromatography to apparent homogeneity 225 36 The T m 1 midpoint transition temperatures of the ACE2-Fc glycomutants (53.3-54.0°C) were slightly higher than for the wildtype protein (52.2°C), while T m 2 and T m 3 remained unchanged (Fig. 5 ). This indicates that 235 removal of the N90 and N322 glycans does not compromise the structural integrity of ACE2. Figure 5 . Analysis of ACE2 variants by differential scanning calorimetry (DSC). Raw data (black) were smoothened (red) and then fitted using a non-two-state thermal unfolding model (grey). Data are presented as mean ± SEM of three independent experiments. Cp, heat capacitance; rshACE2, clinical-grade recombinant soluble 240 human ACE2; deglyco-rshACE2, enzymatically deglycosylated rshACE2; deglyco-ACE2-wt-Fc, enzymatically deglycosylated wild-type ACE2-Fc; desialo-ACE2-wt-Fc, enzymatically desialylated wild-type ACE2-Fc.

The Spike-binding properties of the purified ACE2-Fc variants were characterized by biolayer interferometry (BLI). For this, ACE2-wt-Fc, ACE2-T92Q-Fc and ACE2-N322Q-Fc were 245 biotinylated, immobilized on streptavidin biosensor tips and dipped into serial dilutions of trimeric Spike. Since we did not observe appreciable dissociation of ACE2-Fc/trimeric Spike complexes in our analyses (Fig. S10) , we evaluated the association rates (k obs ; Fig. 6a ). To determine equilibrium affinity constants (K D ), we analyzed the interactions between the immobilized ACE2-Fc constructs and monomeric RBD (Fig. 6b, Fig. S11 ). The BLI data are in 250 good agreement with our computational models, confirming that the removal of protein Nglycosylation at either N90 or N322 results in up to 2-fold higher binding affinities, when To investigate a potential additive effect of simultaneous elimination of N-glycosylation at N90 and N322, we generated a double mutant ACE2-T92Q-N322Q-Fc construct. We also digested ACE2-wt-Fc with peptide-N4-(N-acetyl-beta-glucosaminyl)asparagine amidase F (PNGase F) to 295 remove all accessible N-glycans (deglyco-ACE2-wt-Fc) and neuraminidase to release terminal sialic acid residues (desialo-ACE2-wt-Fc). Purity and homogeneity of these additional ACE2-Fc variants was ascertained by SDS-PAGE and SEC-MALS (Figs. S8 and S9). The absence of Nglycans attached to N90 and/or N322 in ACE2-T92Q-N322Q-Fc and the respective single mutants was demonstrated by LC-ESI-MS (Fig. S14 ). Quantitative release of sialic acids and 300 complete removal of N-glycans from all ACE2-wt-Fc N-glycosylation sites with the exception of N546 was also confirmed (Figs. S15 and S16). The glycans at N546 of ACE2-wt-Fc exhibited partial resistance (40%) to PNGase F treatment (Fig. S15 ). Combined introduction of the mutations T92Q and N322Q as well as enzymatic desialylation did not reduce the thermal stability of ACE2-Fc as assessed by DSC, while close-to-complete removal of N-glycans by 305

PNGase F led to a slightly decreased midpoint transition temperature of the ACE2 domain (Fig.   5 ). Studies of the interaction between ACE2-T92Q-N322Q-Fc and deglyco-ACE2-wt-Fc with RBD by BLI analysis yielded K D values similar to those determined for the single mutant ACE2-T92Q-Fc (ACE2-T92Q-N322Q-Fc: K D = 8.2 ± 0.2 nM; deglyco-ACE2-wt-Fc: K D = 7.6 ± 0.3 nM). The affinity of desialo-ACE2-wt-Fc for RBD (K D = 11.3 ± 0.4 nM) was also higher than 310 that of native ACE2-wt-Fc (Fig. 6b) . The increased affinities of these ACE2-Fc variants for Spike correlate well with their potencies to neutralize SARS-CoV-2, with deglyco-ACE2-wt-Fc followed by ACE2-T92Q-N322Q-Fc displaying the highest neutralization potencies (Figs. 7 and S13). The effect of desialo-ACE2-wt-Fc on SARS-CoV-2 infections of Vero E6 cells was less pronounced and comparable to that of the single mutant ACE2-N322Q-Fc (Fig. 7) , in good 315 agreement with the almost identical RBD-binding affinities of these two ACE2-Fc variants (Fig.  6b ). Taken together, these data identify critical glycans at position N90 and N322 of ACE2 that structurally and functionally interfere with Spike-ACE2 binding; ablation of these glycans via site-directed mutagenesis or enzymatic deglycosylation generated ACE2 variants with improved Spike-binding properties and increased neutralization strength. 320

The results presented above uncover the critical importance of N-glycans located at the ACE2- Paralleling our observations with deglyco-ACE2-wt-Fc, we found the binding affinity of 330 deglyco-rshACE2 to RBD (K D = 5.1 ± 0.5 nM) to be two times higher than for native rshACE2 (K D = 10.5 ± 0.4 nM; Fig. 6b ). Furthermore, deglyco-rshACE2 displayed improved SARS-CoV-2 neutralization properties in Vero E6 cell infection assays. At a final concentration of 200 µg/mL deglyco-rshACE2, we observed a significant reduction in SARS-CoV-2 replication when compared to treatment with the native form of the protein (Fig. 8) . 335 

Besides serving as a soluble decoy receptor to prevent SARS-CoV-2 infection of ACE2expressing host cells, rshACE2 also regulates blood pressure and protects multiple organs such as the heart, kidney and lung as well as blood vessels via enzymatic degradation of angiotensin II. 37 In contrast to other recently described ACE2 mutants displaying improved Spike binding 350 concomitant with inadvertently or intentionally impaired enzymatic activity, 14,15 the catalytic activities of ACE2-T92Q-Fc, ACE2-N322Q-Fc and ACE2-T92Q-N322Q-Fc were found to be only modestly reduced as compared to ACE2-wt-Fc (ACE2-T92Q-Fc: 65 ± 11 %; ACE2-N322Q-Fc: 69 ± 7 %; ACE2-T92Q-N322Q-Fc: 79 ± 11 %; Fig. 9 ). Interestingly, deglyco-ACE2-wt-Fc (149 ± 1 %) and desialo-ACE2-wt-Fc (160 ± 2 %) exhibited higher enzymatic activities than native ACE2-wt-Fc (Figs. 10 and S17). A similar observation was made for deglyco-rshACE2, although the enhancing effects of enzymatic deglycosylation on catalytic efficiency were less pronounced (113 ± 2 % as compared to native rshACE2; Fig. 10 ).

These results show that enzymatic removal of N-glycans from ACE2-Fc and clinical-grade 365 rshACE2 results in increased Spike binding and enhanced SARS-CoV-2 neutralization while preserving its potentially critical enzymatic activity. 

Hydrolytic activity is plotted as relative fluorescence units (RFU) over ACE2 concentration (in nM). All assays were performed in technical triplicates. One representative experiment out of two is shown.

Our data demonstrate that structure-guided glycoengineering is a powerful means to develop 375 ACE2 variants with improved SARS-CoV-2 neutralization properties without compromising the structural stability and catalytic activity of the enzyme. Our in silico models of the Spike-ACE2 complex combined with simulations of its spatial and temporal dynamics rationalized previously published data and led to predictions that were confirmed by in vitro binding studies and cell-based SARS-CoV-2 neutralization assays. It is of note that the moderately enhanced affinity of 380 ACE2 glycovariants for monomeric RBD observed in biolayer interferometry experiments relates to a far more pronounced increase of their virus-neutralization potency. This may be explained by multiple cooperative effects. First, a cooperative effect may be expected for the association of trimeric Spike molecules present in the viral envelope with membrane-bound ACE2 dimers. In this supramolecular setting, a subtle increase in the affinity of ACE2 for RBD 385 can lead to a dynamic equilibrium of binding and unbinding events with up to six potential interactions, leading to an overall much stronger avidity effect. Second, a slight advantage of the soluble ACE2 decoy receptor over endogenous native ACE2 may be sufficient to tip the balance between SARS-CoV-2 attachment and shedding of viral particles from the host cell surface.

Third, blocking of initial binding prevents viral propagation and hence spread of the virus to 390 surrounding cells. Finally, it is possible that the N-glycan moiety of ACE2 also modulates other aspects of viral entry besides promoting the docking of Spike to the cell surface. 29 It has been reported that the sialylation status of ACE2 affects its interactions with SARS-CoV-2

Spike. 30 We have found that enzymatic desialylation of ACE2 results in a reproducible increase of its affinity to RBD without detectable structural penalties. Importantly, desialylated ACE2 is 395 more efficient in neutralizing SARS-CoV-2 than its native counterpart. Molecular simulations suggest that the terminal sialic acids of the N-glycans attached to ACE2 residues N90 and N322 mask parts of the Spike-ACE2 interface and thus could interfere with Spike binding through steric clashes and/or electrostatic effects (Fig. S18 ). This provides a structural rationale how sialic acids present on ACE2 might dampen interactions with Spike during SARS-CoV-2 400 attachment to host cells. 38

In line with other reports, 15, 38 our results indicate that the elimination of the N-glycans attached to N90 is largely responsible for the improved Spike-binding properties of enzymatically deglycosylated ACE2. As proposed 14, 15 and corroborated by our mutational analysis, substitution of ACE2 residues N90 or T92 could indeed provide an alternative approach for the development 405 of ACE2 variants with improved SARS-CoV-2 sequestering properties. Our data indicate that ablation of N90 glycosylation could be combined with mutations of N322 and possibly other ACE2 N-glycosylation sites to achieve an even higher SARS-CoV-2 neutralizing potency.

However, expression of an ACE2 variant lacking all potential N-glycosylation sites in ACE2negative host cells led to reduced rather than enhanced susceptibility of the cells to SARS-CoV-2 410 as compared to transduction with wild-type ACE2. 38 This was attributed to the much lower cellular content of the mutant protein relative to the native enzyme, thus demonstrating that the importance of N-glycosylation for proper folding of glycoproteins during their biosynthesis 39 also applies to ACE2. Given the inferior expression yields of glycan-free ACE2 and the potential of unwanted immunological side effects when non-natural mutations are introduced into a 415 therapeutic glycoprotein, we believe that the clinical potential of enzymatically deglycosylated rshACE2 is superior to that of any of our ACE2 glycomutants. In our opinion, treatment of clinical-grade rshACE2 with deglycosylation enzymes such as PNGase F followed by a final polishing step represents a straightforward, Good Manufacturing Practice (GMP)-compliant and industrially feasible alternative to generate a potent therapeutic drug for the treatment of SARS-420

CoV-2 infected persons and patients.

To model the fully glycosylated SARS-CoV-2 Spike-human ACE2 (hACE2) complex, a protein model was created using partial experimental structures deposited in the protein databank (PDB). and NaNaF, respectively. These conformations were fitted onto the respective glycosylation site in the Spike-hACE2 complex using a superposition of the backbone of the asparagine residues and the non-bonded interaction energy between the glycan and protein atoms or previously added glycans was computed. The lowest energy conformation was retained. Topologies and initial conformations were generated using the gromos++ suite of pre-and post-MD tools. 44 450

Glycans were added to the complex sequentially, to avoid collisions between individual glycans.

A few modeled glycans were incompatible with loops of the Spike protein not resolved in the experimental structures. Loops involved in these structural incompatibilities (residues 141-165 and 471-490) were partially re-modeled in the fully glycosylated model using the RCD+ loop modeling server. 45, 46 The final model was energy-minimized with the GROMOS 54A8 protein 455 force-field, 47, 48 the GROMOS 53A6glyc glycan force-field 41, 49, 50 and the GROMOS simulation software using the steepest decent algorithm. 51

Molecular dynamics simulations were performed using the simulation package Gromacs (Version 2019.5) and the indicated force field parameters. hACE2 was reduced to residues 21 to 460 730 in the models, to reduce its overall size prior to simulation. The models were placed in rhombic dodecahedron simulation boxes and solvated by explicit SPC water molecules. 52 This resulted in simulation systems of 5.9 × 10 5 and 2.2 × 10 6 atoms for hACE2 and the Spike-hACE2 complex, respectively. Two independent 100-ns molecular dynamics simulations were performed for hACE2 and for the Spike-hACE2 complex each. The equations of motion were 465 integrated using a leapfrog integration scheme 53 with a time-step of 2 fs. Non-bonded interactions were calculated within a cutoff sphere of 1.4 nm and electrostatic interactions were computed using a particle-particle particle-mesh (P3M) approach. 54 Bond-lengths were constrained to their optimal values using the Lincs algorithm. 55 Temperature was maintained at a constant value using a velocity-rescaling algorithm 56 

After filtration through 0.45 μ m membrane filters (Merck Millipore, Germany), supernatants containing RBD or soluble Spike were concentrated and diafiltrated against 20 mM sodium 510 phosphate buffer containing 500 mM NaCl and 20 mM imidazole (pH 7.4) using a Labscale TFF system equipped with a 5 kDa cut-off Pellicon XL device (Merck Millipore). The His-tagged proteins were captured using a 5 mL HisTrap FF crude column connected to an ÄKTA pure chromatography system (both from Cytiva, United States). Bound proteins were eluted by applying a linear gradient of 20 to 500 mM imidazole over 20 column volumes. ACE2-Fc 515 variants were purified by affinity chromatography using a 5 mL HiTrap Protein A column (Cytiva) according to the manufacturer's instructions and 0.1 M glycine-HCl (pH 3.5) for elution. Eluate fractions were immediately neutralized using 2 M Tris (pH 12.0). Fractions containing the protein of interest were pooled, concentrated using Vivaspin 20 Ultrafiltration Units (Sartorius, Germany) and dialyzed against PBS (pH 7.4) at 4°C overnight using SnakeSkin 520 Dialysis Tubing (Thermo Fisher Scientific). The RBD was further purified by size exclusion chromatography (SEC) using a HiLoad 16/600 Superdex 200 pg column (Cytiva) eluted with PBS. All purified proteins were stored at -80°C until further use.

For deglycosylation of ACE2-wt-Fc and rshACE2, proteins (2 mg mL -1 ) were incubated with 525 180000 U mL -1 PNGase F (New England Biolabs, Unites States) in PBS (pH 7.4) for 24 h at 37°C. Desialylation of ACE2-wt-Fc was performed with 2500 U mL -1 neuraminidase (New England Biolabs) in 50 mM sodium citrate (pH 5.0) under otherwise identical conditions. The deglycosylated or desialylated ACE2 variants were purified by preparative SEC using a HiLoad 16/600 Superdex 200 pg column eluted in PBS. The extent of enzymatic deglycosylation and 530 desialylation was assessed by SDS-PAGE (Fig. S8) , SEC-MALS (Fig. S9) and ESI-LC-MS/MS (Figs. S15 and S16).

Interaction studies were performed on an Octet RED96e system using high precision streptavidin 

Size-exclusion chromatography combined with multi-angle light scattering was performed to determine the homogeneity and the native molecular mass of all proteins under study. Analyses were performed on an LC20 Prominence HPLC equipped with a refractive index detector RID- 

Enzymatic activity of ACE2 was determined and quantified as described previously, 37 using 100 µM 7-methoxycoumarin-4-yl-acetyl-Ala-Pro-Lys-2,4-dinitrophenyl (Bachem, Switzerland) as substrate.

All work with infectious SARS-CoV-2 was performed under BSL-3 conditions. Vero E6 cells 620 

Molecular models and simulation trajectories are available through the BioExcel COVID-19 Molecular Structure and Therapeutics Hub (https://covid.bioexcel.eu/simulations/). 715

A Novel Coronavirus from Patients with Pneumonia in China

A new coronavirus associated with human respiratory disease in China

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

Structural and Functional 735 Basis of SARS-CoV-2 Entry by Using Human ACE2

740 (2020) Inhibition of SARS-CoV-2 Infections in Engineered Human Tissues Using Clinical-Grade Soluble Human ACE2

Human soluble ACE2 improves the effect of remdesivir in SARS-CoV-2 infection

750 (2020) Human recombinant soluble ACE2 in severe COVID-19

SARS-CoV-2 755 neutralizing antibody structures inform therapeutic strategies

A molecularly 760 engineered, broad-spectrum anti-coronavirus lectin inhibits SARS-CoV-2 and MERS-CoV infection in vivo

Identification of lectin receptors for conserved SARS-CoV-2 glycosylation sites

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

SARS-CoV-2 Neutralizing Antibodies for COVID-19 Prevention and Treatment

Structural and functional ramifications of antigenic drift in recent SARS-CoV-2 variants

Engineered ACE2 receptor traps potently neutralize SARS-CoV-2

Engineering human ACE2 to optimize binding to the spike protein of SARS coronavirus 2

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

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

2020) Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Cryo-EM structure of the 2019-nCoV spike in the 800 prefusion conformation

Sitespecific glycan analysis of the SARS-CoV-2 spike

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

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

Computational epitope map of SARS-CoV-2 spike protein

SARS-CoV-2 simulations go exascale to predict dramatic 820 spike opening and cryptic pockets across the proteome

2021) Dual nature of human ACE2 glycosylation in binding to SARS-CoV-2 spike

A glycan gate controls opening of the SARS-CoV-2 spike protein

Anchor-Locker Binding Mechanism of the Coronavirus Spike Protein to Human ACE2: Insights from Computational Analysis

835 (2021) Molecular basis for higher affinity of SARS-CoV-2 spike RBD for human ACE2 receptor

Inhibition of SARS-CoV-2 viral entry upon blocking N-and O-glycan elaboration

Subtle Influence of ACE2 Glycan Processing on SARS-CoV-2 Recognition

Mass Spectrometry Analysis of Newly Emerging Coronavirus HCoV-19 Spike 845 Protein and Human ACE2 Reveals Camouflaging Glycans and Unique Post-Translational Modifications. Engineering (Beijing)

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

The Impact of Mutations in SARS-CoV-2 Spike on Viral Infectivity and Antigenicity

Elucidation of interactions regulating conformational stability and dynamics of SARS-CoV-2 S-protein

The flexibility of ACE2 in the 860 context of SARS-CoV-2 infection

Fcab-HER2 Interaction: a Menage a Trois. Lessons from X-Ray and Solution Studies

Hydrolysis of biological peptides by human angiotensin-converting enzyme-related carboxypeptidase

Host and viral determinants for efficient SARS-CoV-2 infection of the human lung

Glycosylation-directed quality control of protein folding

SWISS-MODEL: homology modelling of protein structures and complexes

Modeling of Oligosaccharides within Glycoproteins from Free-Energy Landscapes

Local elevation: A method for improving the searching properties of molecular dynamics simulations

Conformational properties of glycose-based disaccharides investigated using molecular dynamics simulations with local elevation umbrella sampling

GROMOS plus plus Software for the Analysis of Biomolecular Simulation Trajectories

Random Coordinate Descent with Spinor-matrices and 895 Geometric Filters for Efficient Loop Closure

RCD+: Fast loop modeling server

New Interaction Parameters 900 for Charged Amino Acid Side Chains in the GROMOS Force Field

Testing of the GROMOS Force-Field Parameter Set 54A8: Structural Properties of Electrolyte Solutions, Lipid Bilayers, and Proteins

GROMOS 53A6GLYC, an Improved GROMOS Force Field for Hexopyranose-Based Carbohydrates

Extension and validation of the 910 GROMOS 53A6(GLYC) parameter set for glycoproteins

Architecture, implementation and parallelisation of the GROMOS software for biomolecular simulation

Interaction models for water in relation to protein hydration

The potential calculations and some applications

LINCS: A 925 linear constraint solver for molecular simulations

Molecular-dynamics with coupling to an external bath

Canonical sampling through velocity rescaling

Polymorphic transitions in single crystals: A new molecular dynamic method

Constant pressure molecular dynamics for molecular systems

GROmarhos: A GROMACS-Based Toolset to Analyze Density Maps Derived from Molecular Dynamics Simulations

High-level and high-throughput recombinant protein production by transient transfection of suspension-growing human 293-EBNA1 cells

A serological assay to detect SARS-CoV-2 seroconversion in humans

SARS-CoV-2 Seroconversion in Humans: A Detailed Protocol for a Serological Assay, Antigen Production, and Test Setup

Ligand binding assays at equilibrium: validation and interpretation

Limit of blank, limit of detection and limit of quantitation

A detection and quantification label-free tool to 960 speed up downstream processing of model mucins

Determination of 50% endpoint titer using a simple formula

Synthetic recombinant bat SARS-like coronavirus is infectious in cultured cells and in mice