key: cord-323828-ug2duzw1
authors: Ni, Dongchun; Lau, Kelvin; Lehmann, Frank; Fränkl, Andri; Hacker, David; Pojer, Florence; Stahlberg, Henning
title: Structural investigation of ACE2 dependent disassembly of the trimeric SARS-CoV-2 Spike glycoprotein
date: 2020-10-12
journal: bioRxiv
DOI: 10.1101/2020.10.12.336016
sha: 
doc_id: 323828
cord_uid: ug2duzw1

The human membrane protein Angiotensin-converting enzyme 2 (hACE2) acts as the main receptor for host cells invasion of the new coronavirus SARS-CoV-2. The viral surface glycoprotein Spike binds to hACE2, which triggers virus entry into cells. As of today, the role of hACE2 for virus fusion is not well understood. Blocking the transition of Spike from its prefusion to post-fusion state might be a strategy to prevent or treat COVID-19. Here we report a single particle cryo-electron microscopy analysis of SARS-CoV-2 trimeric Spike in presence of the human ACE2 ectodomain. The binding of purified hACE2 ectodomain to Spike induces the disassembly of the trimeric form of Spike and a structural rearrangement of its S1 domain to form a stable, monomeric complex with hACE2. This observed hACE2 dependent dissociation of the Spike trimer suggests a mechanism for the therapeutic role of recombinant soluble hACE2 for treatment of COVID-19.

Coronavirus is a family of single-stranded RNA viruses, many of which can infect animals and humans 41 (MacLachlan and Dubovi, 2017; Monto, 1984) . The symptoms of coronavirus-related diseases can be mild 42 and mainly occur in respiratory tract. For example, roughly 15%-30% cases of the common cold are caused 43 by human coronaviruses (Mesel-Lemoine et al., 2012) . A coronavirus infection sometimes can develop serious 44 illnesses, such as SARS (severe acute respiratory syndrome), MERS (Middle East respiratory syndrome) and 45 also the current pandemic COVID-19 (coronavirus disease 2019) (Tang et al., 2020) . SARS-CoV is a beta-46 coronavirus that caused a pandemic in 2002. SARS-CoV-2 is a novel coronavirus that is genetically similar to 47 the previous SARS-CoV. SARS-CoV-2 causes the ongoing pandemic COVID-19 and has been spreading 48 globally since the first quarter of this year (Ciotti et al., 2020) . The symptoms of COVID-19 vary from person 49 to person. In some cases, the illness is very serious, in particular for the elderly (Pascarella et al., 2020) . As of 50 today, no specific antiviral drugs were approved for use against COVID-19 and vaccine development is still 51 at the phase of clinical testing.

Cryogenic electron microscopy (Cryo-EM) is a technique for structure determination of biomacromolecules, 53 which has been particularly successful for studying high molecular-weight proteins. Cryo-EM does not require 54 crystallization of the target protein. COVID-19 related protein structures have been widely investigated since 55 the spread of SARS-CoV-2 in this year, using cryo-EM single particle analysis (SPA). The protein nicknamed 56 Spike is with its 180kDa monomeric molecular weight the largest viral surface protein of SARS-CoV-2. It 57 consists of two domains S1 and S2 that are connected by a short linker. Spike forms stable trimers on the virus 58 surface that are attached to the virus membrane. This Spike trimer is the key molecule for host cells receptor 59 binding and invasion of the host cells. The cryo-EM structure of the entire Spike homotrimer was determined 60 recently, showing a mushroom sharped overall architecture (Walls et al., 2020; Wrapp et al., 2020a) . As also 61 for the SARS-CoV, the viral fusion bridge from SARS-CoV-2 to the host cell is formed by Spike and the 62 ectodomain of the human Angiotensin-converting enzyme 2 (hACE2), which is the virus receptor on the host 63 cell that triggers virus entry. In vitro studies have shown that the Spike receptor binding domains (RBDs) from 64 SARS-CoV as well as SARS-CoV-2 can both bind to the ectodomain of hACE2 with comparable binding 65 affinities in low nanomolar levels (Lan et al., 2020a) . However, the new SARS-CoV-2 exhibits a more potent 66 capacity of host cells adhesion, as well as a larger virus-entry efficiency than other beta-coronaviruses (Shang 67 et al., 2020) .

The membrane-attached hACE2 is known to be the key molecule for the infection by several viruses, including 69 SARS-CoV, Human coronavirus NL63 (HCoV-NL63) and SARS-CoV-2. The infection process primarily 70 involves virus adhesion and fusion (Bao et al., 2020; Fehr and Perlman, 2015; Kuba et al., 2005; Sia et al., 71 2020). Interestingly, hACE2 may not only serve as a drug target to prevent SARS-Co-2 infection, but hACE2 72 itself may also be considered as a potential therapeutic drug candidate for the usage against COVID-19 or 73 other beta-coronavirus related diseases. The clinical-grade soluble form of hACE2 has been reported to be a 74 potential novel therapeutic approach for reducing the infection of SARS-CoV-2 (Monteil et al., 2020) by preventing the viral Spike from interacting with other hACE2 present on human cells. Recently, researchers 76 have also characterized the entire architecture of the inactivated authentic virions from SARS-CoV-2 using 77 cryo-electron tomography, observing that post-fusion S2 trimers are distributed on the surface of SARS-CoV-78 2 virions (Ke et al., 2020; Turonova et al., 2017) . The exact role of hACE2 so far is not yet fully understood 79 in terms of its interaction with full-length Spike protein.

In this study, we present a cryo-electron microscopy (cryo-EM) study of the SARS-CoV-2 Spike protein in 81 complex with hACE2. Our analysis reveals a monomeric complex of Spike S1 domain with hACE2, requiring 82 a large structural rearrangement in S1 compared to its isolated structure. Our data show that hACE2 binding 83 induces a conformational change in Spike, leading to Spike trimer dissociation.

Spike and hACE2 production and its complex assembly.

The prefusion Spike 2P ectodomain was expressed in ExpiCHO cells and affinity purified via its twin Strep-87 tag. SDS-PAGE analysis showed the presence of pure full-length Spike protein, consisting of both, the S1 and 88 S2 domains at the expected molecular weight of 180 kDa for the Spike monomer. (Suppl. Fig. S1a ). The

purified Spike sample in PBS buffer was imaged as negatively stained preparations by transmission electron 90 microscopy (TEM). This revealed the expected trimeric shape, and 2D class averages of selected particles in 91 negative stain TEM images showed the typical, mushroom-shaped particles (Suppl. Fig. S2a .c), in accordance 92 with the expected structure of the SARS-CoV-2 Spike in the pre-fusion state.

Human ACE2 ectodomain was expressed in HEK293 cells and purified via a poly-histidine immobilized metal 94 affinity chromatography (IMAC) with a Fastback Ni2+ column, followed by another anion exchange column 95 (Suppl. Fig. S1b ). For details, see Methods.

Purified Spike protein was mixed with excess hACE2 (molar ratio Spike:hACE2 of 1:5) and incubated for 12 97 hours at 4º C. Samples were prepared by negative staining and imaged by TEM. Unexpectedly, the observed 98 particle features were largely different from the typical Spike trimers in shape. A 2D analysis of 5'854 picked 99 negatively stained particles revealed in class averages that the majority of the particles were smaller in size 100 and asymmetrical, compared to the non-incubated Spike trimeric samples (Suppl. Fig. S2b,d) . This suggests 101 that the prolonged incubation with hACE2 led to Spike trimer dissociation. A similar observation for SARS-

CoV-ACE2 complex was recently reported by Song et al (Song et al., 2018) . Due to particle heterogeneity, we 103 decided to further purify the complex by size exclusion chromatography (SEC) and indeed the SEC profile 104 showed three distinct peaks, called Peak1, Peak2 and Peak3 (Suppl. Fig. S1c,d) .

By analyzing the 3 peaks by SDS gel and negative stain EM, we could clearly differentiate non-structured 106 aggregates of full-length Spike and hACE2 in Peak1, that did not allow further structural analysis (data not 107 shown), to the excess of unbound hACE2 in Peak3 (Suppl. Fig. S1c,d) . The homogenous Peak2 that contained 108 full-length Spike in complex with hACE2, was further analyzed by cryo-EM. hACE2 binding can induce disassembly of Spike homotrimer 110 Peak2 (Spike:hACE2 at molar ratio 1:5 after overnight incubation at 4°C) was vitrified and frozen grids were 111 loaded into a Thermo Fisher Scientific (TFS) Titan Krios cryo-EM instrument, operated at 300kV acceleration 112 voltage, and equipped with a Gatan Quantum-LS energy filter equipped with K2 direct electron detector 113 (Suppl. Fig. S3 ). 8'927 dose-fractionated images (i.e., movies) were recorded (Suppl. Fig. S4 ), from which 114 ~1.7 million particles were extracted and subjected to image processing and 3D reconstruction. The final 3D 115 reconstruction from 72'446 particles at 5.1Å overall resolution showed a density map corresponding to a single, 116 monomeric Spike protein in complex with hACE2 (Fig. 1a) . The map allowed docking with available 117 structures for S1 and hACE2 taken from the previously reported structures (Spike PDB ID 6VYB and Spike 118 RBD-ACE2 6M0J), revealing a structural rearrangement of the C-terminal domain (CTD) and N-terminal 119 domain (NTD) of S1 compared to a monomer from that Spike structure in the RBD up conformation. The 120 interaction between the S1 RBD and hACE2 is in agreement with several other reported structures of the RBD-121 ACE2 complex (PDB ID 6M0J, 6VW1 or 6LZG). No additional density for S2 or a fragment of S2 was 122 detected in the reconstruction.

We tested a shorter incubation time by mixing Spike:hACE2 (molar ratio of 1:3) and let it incubate for 3 hours 124 at 4ºC, instead as overnight. No further SEC purification was performed. Subsequently, cryo-EM grids of this 125 sample were prepared and subjected to cryo-EM analysis (Suppl. Fig. S5 ). From 7'045 recorded movies, 126 615'348 particles were extracted and subjected to classification and 3D analysis. This revealed a small sub-set 127 of 47'901 particles corresponding to the prefusion Spike trimer, which allowed a 3D reconstruction at 4.2Å 128 overall resolution (no symmetry was applied), while some regions of the 3D map showed lower resolution, 129 presumably due to increased flexibility of these areas (Suppl. Fig. S6) . A resolution-limited map at 9Å 130 resolution ( Fig. 1b) allowed clear docking the models of Spike S1 and S2 and hACE2, which showed that the 131 complex is composed of Spike and hACE2 in a molar ratio of 3:3 (Spike:hACE2). Three hACE2 molecules 132 were observed to attach to the RBDs of Spike. All three RBDs were in the RBD up conformation and slightly 133 shifted away from the central trimer axis (Fig. 1b) . A similar arrangement was also recently observed by 140 S1-hACE2 model to that of the trimeric Spike-hACE2 (3:3) showed that a ~30º rotation of the C-terminal and 141 N-terminal sub-domains of Spike S1 were required to bring the Spike S1 protein into the monomeric 142 arrangement with hACE2. After such re-arrangement, the domains of hACE2 and the RBDs of the S1 protein 143 are in good agreement with a reported crystal structure (REF Lan et al.) (Fig. 2b) .

When comparing the docked model of the monomeric S1-hACE2 complex with that of the trimeric Spike-145 hACE2 complex (Fig. 2e) , a considerable number of stearic clashes at the interface between Spike S1 (CTD) 146 and its neighboring region from the S2 polypeptide chain was obvious. The docked monomeric S1-hACE2 147 complex is further structurally incompatible with the observed trimeric arrangement.

The stoichiometric ratio of the complex of Spike:hACE2 on the host cell upon virus entry is not well 150 established. Nevertheless, one hACE2 molecule per Spike trimer is likely sufficient for binding and initializing 151 virus fusion with the host cell (Song et al., 2018) . Even though the structure of a post-fusion S2 trimer has 152 recently been determined (Cai et al., 2020), it is not clear how membrane fusion during virus entry is 153 coordinated upon release of the S1-hACE2 caps (Fig. 3) . We here report the cryo-EM structure of a stable 154 monomeric S1 Spike-hACE2 (1:1) complex. Even though size-wise it would have been detectable, our particle 155 classification did not reveal any particle class corresponding to an isolated S1 fragment in addition to the 156 observed S1 Spike-hACE2 (1:1) particles (Suppl. Fig. S3c, S4) . Knowledge of the mechanism how the S1 157 fragment might be detached from the hACE2 receptors after virus entry would be relevant for understanding 158 its mechanism of infection and pathogenicity. The fact that we did not observe any free S1 fragments suggests 159 that the S1-hACE2 complex is rather stable, at least under our in vitro conditions.

Secondly, the S2 domain was not detected in the obtained structure of the Spike-hACE2 monomeric complex

(1:1), even though the SDS-PAGE analysis showed that S2 was present as full length in the sample (Suppl.

Fig. S1d). The S2 domain is expected to be connected to the S1 domain via a short loop between S1 and S2,

where a Furin protease cleavage site is expected (Belouzard et al., 2009; Haan et al., 2004; Hoffmann et al., 164 2020). However, in the absence of stable trimers, the loop between S1 and S2 is likely very flexible, possibly 165 making the S2 domain undetectable by cryo-EM maps. Our cryo-EM analysis that didn't show the S2 domain 166 in the averaged 3D reconstruction therefore likely failed to align the S2 domains either due to their flexibility, 167 or due to a denaturation of S2 during sample preparation.

An early study presented a potential dose-dependent inhibition of SARS-CoV-2 infection by a recombinant 169 soluble form of hACE2 (Monteil et al., 2020) . The mechanism, how the soluble forms of hACE2 would be 170 able to neutralize the virus, is not known. One possible mechanism could be a direct competition between the 171 soluble hACE2 and the host cell hACE2 receptor, so that Spike proteins saturated with soluble hACE2 domains 172 render them unable to interact with host cell hACE2. Here, however, we report that the soluble forms of hACE2 173 induce the opening and disassembly of the trimeric Spike structure to create the stable Spike S1-hACE2 174 complex (Fig. 3) . We propose a mechanism by which the formation of the Spike-hACE2 (3:3) complex induces 175 a high structural flexibility in the Spike trimer, allowing a conformational re-arrangement of the S1 C-and N-176 terminal domains when interacting with hACE2. In consequence, the new S1-hACE2 complex is incompatible 177 with a trimeric arrangement, causing the dissociation of the trimeric complex (Fig. 3) .

This hypothesis is supported by the recent manuscript deposited in bioRxiv.org, which describes a similar 179 effect triggered by engineered DARPin molecules (Walser et al., 2020) . Therefore, we suppose that the soluble 180 forms of hACE2 may not only block the infection and replication of SARS-CoV-2, but also destroy the trimeric 181 Spike adaptors that are responsible for viral host membrane fusion. This mechanism suggests a novel 182 therapeutic strategy for the treatment of COVID-19, by adding soluble hACE2 to dissociate the Spike trimer 183 of approaching viruses.

Protein production and purification 

Two different preparations of cryo-EM grids were performed. Purified Spike and hACE2 proteins were mixed 211 at the molar ratio of 1:5 and incubated for 12 hours at 4ºC. After that, the sample was subjected to size exclusion chromatography (SEC) with a Superose 6 increase (10/300) column, and the fractions from Peak2 were pooled 213 and concentrated in 100 kDa centrifugal concentrators (Millipore).

Alternatively, purified Spike and hACE2 were mixed at the molar ratio of 1:3 (Spike:hACE2) an incubated 215 for 3 hours at 4ºc, without further purification via sec.

For both samples, the concentration was adjusted to 0.5mg/ml. Cryo-EM grids were prepared with a Vitrobot 217 Mark IV (Thermo Fisher), using a temperature of 4°C and 100% humidity. 4 μL of sample was applied onto 

For both samples, dose-fractionated images (i.e., movies) were recorded with a Titan Krios (Thermo Fisher), 221 operated at 300kV, and equipped with a Gatan Quantum-LS energy filter (20 eV zero-loss energy filtration) 222 followed by a Gatan K2 Summit direct electron detector. The data collection statistics is presented in Table 1 . 

For the trimeric Spike-hACE2 complex, image processing was performed similarly. The final reconstruction 239 produced a 3D map at 4.2Å overall resolution.

Protein models were generated from reported structures (Spike: PDB ID 6VYB; ACE2-RBD: PDB ID 6M0J) 242 (Lan et al., 2020b; Walls et al., 2020) . For the S1-hACE2 structure, the model was manually docked into the 243 EM density with the program Chimera (Pettersen et al., 2004) and further refined using rigid-body fitting in 244 COOT (Emsley et al., 2010) . For the Spike-hACE2 trimer, the density corresponding to hACE2 was relatively weak, so that low-pass filtration to 9Å resolution was applied to the map before proceeding with the docking 246 of hACE2 as described above. Figure 1 Cryo-EM maps of SARS-CoV-2 Spike-hACE2 complexes and fitted models. 396 a. The 3D reconstruction of SARS-CoV-2 Spike and human ACE2 (mixed at a molar ratio 1:5) 397

incubated for 12hrs and further purified by SEC shows a structure corresponding to monomeric 398

Spike S1 protein in complex with hACE2. No density for S2 is observed. The N-and C-terminal 399 domains of S1 had to be re-arranged to fit into the map. Left: Side-view, center: 90º rotated view. 400

The structure is colored as follows: hACE2 ectodomain green, Spike S1-RBD yellow, Spike S1-CTD 401 slate blue, Spike S1-NTD salmon. Right: The bottom image shows a representative 2D class average 402 of the S1-hACE2 complex. The upper cartoon is its interpretation. b. 3D reconstruction of Spike and 403 hACE2 (ratio 1:3) incubated for 3 hours shows a trimeric map allowing docking of three hACE2 404 molecules and three S1 and three S2 molecules, all forming a trimeric complex. superimposed. b. Structural comparison of S1-hACE2 regions. The movement is presented as two 418

colors: SARS-CoV2 trimeric Spike-hACE2 complex (green) and S1-hACE2 (blue) c,d. The close-up 419 views show the proposed local structural rearrangements. e. Superposition of S1-hACE2(1:1) with 420 the trimeric form of Spike-hACE2 complex. The RBDs have been superimposed. The rearranged 421 structure of Spike-hACE2 is no longer compatible with formation of trimers so that it dissociates. 422

The right panel is a predicted structure of the dissociated Spike-hACE2 monomer. 423 424 425 Spike. In this model, one hACE2 molecule binds to one Spike S1 monomer and induces the 428 conformational changes in the trimeric Spike. Subsequently, a post-fusion S2 trimer is formed. The 429 lower row shows a novel proposed pathway leading to Spike trimer disassembly by hACE2. In 430 presence of a high concentration of hACE2 molecules, a Spike-hACE2 (3:3) complex is formed. 431

Structural clashes between the three Spike-hACE2 elements lead to their dissociation. This induces 432 the formation of monomeric Spike-hACE2 complexes. average classes. d. Model fitted into the S2 trimeric core (The map was low-pass filtered). e. 472

Distribution of particle orientations. f. Local resolution level at the best resolved regions of the 473 trimeric form of Spike-hACE2 complex (bottom view, MonoRes). g. The low-pass filtered EM map at 474

9Å resolution (For model generation). Scale bar in a is 50nm and in c is 3nm. 475 477 478 479 Figure S6 Processing workflows for the trimeric Spike-hACE2 complex. 480 481

Using Prediction of Specimen Position

A Human Coronavirus Responsible for the Common Cold Massively Kills Dendritic Cells 331 but Not Monocytes

Inhibition of SARS-CoV-2 Infections in Engineered Human 335

Tissues Using Clinical-Grade Soluble Human ACE2

Coronaviruses. Viral Infections of Humans: Epidemiology and Control

COVID-19 339 diagnosis and management: a comprehensive review

Chimera--a visualization system for exploratory research and analysis

cryoSPARC: algorithms for rapid unsupervised 345 cryo-EM structure determination

Structural basis of 347 receptor recognition by SARS-CoV-2

Pathogenesis and transmission of SARS-CoV-2 in golden 350 hamsters

Cryo-EM structure of the SARS coronavirus spike glycoprotein in 352 complex with its host cell receptor ACE2

The hallmarks of COVID-19 disease

Efficient 3D-CTF correction for cryo-electron 357 tomography using NovaCTF improves subtomogram averaging resolution to 3.4A

Structure, Function, and 360 Antigenicity of the SARS-CoV-2 Spike Glycoprotein