key: cord-0761886-aqx8luvb
authors: Zhang, Kaiming; Li, Shanshan; Pintilie, Grigore; Chmielewski, David; Schmid, Michael F.; Simmons, Graham; Jin, Jing; Chiu, Wah
title: A 3.4-Å cryo-EM structure of the human coronavirus spike trimer computationally derived from vitrified NL63 virus particles
date: 2020-08-11
journal: bioRxiv
DOI: 10.1101/2020.08.11.245696
sha: 82873573270d58dc8e770d6eb7634480be6dac06
doc_id: 761886
cord_uid: aqx8luvb

Human coronavirus NL63 (HCoV-NL63) is an enveloped pathogen of the family Coronaviridae that spreads worldwide and causes up to 10% of all annual respiratory diseases. HCoV-NL63 is typically associated with mild upper respiratory symptoms in children, elderly and immunocompromised individuals. It has also been shown to cause severe lower respiratory illness. NL63 shares ACE2 as a receptor for viral entry with SARS-CoV and SARS-CoV-2. Here we present the in situ structure of HCoV-NL63 spike (S) trimer at 3.4-Å resolution by single-particle cryo-EM imaging of vitrified virions without chemical fixative. It is structurally homologous to that obtained previously from the biochemically purified ectodomain of HCoV-NL63 S trimer, which displays a 3-fold symmetric trimer in a single conformation. In addition to previously proposed and observed glycosylation sites, our map shows density at other amino acid positions as well as differences in glycan structures. The domain arrangement within a protomer is strikingly different from that of the SARS-CoV-2 S and may explain their different requirements for activating binding to the receptor. This structure provides the basis for future studies of spike proteins with receptors, antibodies, or drugs, in the native state of the coronavirus particles.

3 coronaviruses (https://covid19.who.int/) (Cui et al. 2019; Zhou et al. 2020) . While beta-CoVs are more commonly associated with high pathogenesis and severe respiratory disease, alpha-CoVs are widely circulating viruses associated with cold-like symptoms and in rare cases more serious respiratory failure . HCoV-NL63 is estimated to be the causative agent of up to 10% of annual respiratory disease and is a major cause of bronchiolitis and pneumonia in newborns (Chiu et al. 2005; van der Hoek et al. 2004) .

Coronaviruses utilize large (~500-600 kDa) spike (S) homotrimers protruding from the viral membrane to engage cellular receptors and mediate fusion with host membranes (Pallesen et al. 2017 ; Thorp et al. 2006) . These spikes, numerous on the virus surface, constitute the primary target of neutralizing antibodies (NAbs) and are central to vaccine design and structural studies for drug optimization. Each S molecule contains two large regions: N-terminal S1 responsible for receptorbinding and C-terminal S2 responsible for type-I fusion, as well as an additional single-pass transmembrane helix that anchors the spikes to the viral envelope (Bosch et al. 2003; Zheng et al. 2006 ). Spikes are activated by protease cleavage at the S2' site, near the putative fusion peptide, allowing the transition from pre-to post-fusion conformation during virus entry (Belouzard et al. 2009 ).

Several single-particle cryo-electron microscopy (cryo-EM) structures of purified, ectodomains of pre-fusion S trimers from alpha-CoV and beta-CoV genera have been determined, illustrating conformational heterogeneity of S1, multiple glycosylation sites, and the interactions with soluble receptors and NAbs (Song et al. 2018; Walls et al. 2016 Walls et al. , 2020 Wrapp et al. 2020) . These S complexes lack the helical C-terminal stem that connects S2 to the viral envelope, and require residues mutated to proline at the loop between the first heptad repeat (HR) and the central helix to stabilize the construct (Kirchdoerfer et al. 2018; Pallesen et al. 2017; Park et al. 2019) . High resolution structure determinations of molecular components in pleomorphic virions are typically obtained by single particle cryo-EM or X-ray crystallography of exogenously expressed and purified components. Conversely, cryo-electron tomography (cryo-ET) has been used to reconstruct tomograms of the whole virus particles, from which subvolume density, corresponding to the feature of interest, would be further processed by classification and averaging (Schmid et al. 2012; Wan & Briggs 2016) . Here, we take the approach of utilizing single-particle cryo-EM imaging of purified and vitrified HCoV-NL63 virus particles, and then using computational 4 methods to extract the S trimer density for its high resolution structure determination. Our approach is similar to a recent preprint report for chemically fixed SARS-CoV-2 particles (Ke et al. 2020) . We report the 3.4-Å structure of the S trimer in its native, pre-fusion state, and expect our data will be of value in future drug and vaccine design for various strains of coronaviruses without the need to genetically construct and biochemically purify spike proteins.

Since the S timers protrude from the membrane surface of the virion, many of the crown-shaped S trimers can be easily identified, computationally extracted from the raw images, and treated as single particle images with a distinct particle orientation without the need of tilting the specimens ( Fig. 1a) . Based on the fact that single-particle cryo-EM is a more routine procedure for computing high-resolution structures compared to cryo-ET, we decided not to collect tilt series, and used single-particle cryo-EM analysis of the native S trimer on purified and vitrified virions. The 2D class averages showed a clear triangular shape for the S trimer and other distinct views (Fig. 1b) .

De novo building of the initial map, using the "Ab initial reconstruction" option in cryoSPARC without any symmetry applied, resulted in a three-dimensional (3D) structure with well-defined features. Further 3D refinement was performed with and without C3 symmetry to obtain 3.4 Å and 3.7 Å maps, respectively ( Fig. 1c-e and Fig. S1 ). The reconstructed map without imposed symmetry also displayed 3-fold symmetry, showing a high cross-correlation coefficient (CC = 0.9802) with the symmetric map (Fig. S1d) . Hereafter, we will present our structural analysis of the 3.4-Å map with C3 imposed symmetry after refinement from a subset of ~82,000 S trimer particles showing adequate orientation sampling (Fig. S1) . The local resolution varies in our map ( Fig. 1e) ; the densities at the utmost surface of the protein have a much lower resolution than the central region (Fig. 1e) , which could be attributed to the inherent flexibility of the distal ends of glycans. Compared to the map of the purified S ectodomain, our map has more extended density without recognizable secondary structure element features pointing toward the viral membrane ( Fig. 1e) . The detection of this feature likely benefits from the direct picking of S trimer anchored in the lipid envelope of native virions, stabilizing the stem region that otherwise disassembles in the purified S ectodomains.

Because of the high feature similarity of our map and that determined from the biochemically purified S ectodomain (Walls et al. 2016) , we fit the published model (PDB ID: 5SZS) to our 3.4-Å map and refined it including the glycans (see Methods). The quality of the refined model was validated by MolProbity (Chen et al. 2010) (Table S1) , correlation between the map and model, and Q-score analysis per residue (Pintilie et al. 2020) (Fig. 2) . The density map of our structure is better resolved at many regions relative to the previous structure (PDB ID: 5SZS), including glycans ( Fig. 3) , which were previously interpreted with accompanied mass spectroscopy measurements (Walls et al. 2016) . Superimposition of our model and 5SZS yields 804 pruned atom pairs matched with 1.28-Å RMSD, indicating their high structural similarity. Structural comparisons based on individual domains also show similar results. These domains include domain 0, domain A, domain B (also known as RBD), domain C, and domain D, with the RMSD ranging from 0.29 Å to 0.90 Å. As shown in Fig. 1 and 2 , each of the three receptor-binding domains (RBDs) within the S trimer is pointed downwards. Notably, heterogeneous refinement and 3D variability did not reveal alternative RBD conformations, implying that the native S trimer proteins on the virion are predominantly in a fully closed state, consistent with the previous report of purified HCoV-NL63 S ectodomain (Walls et al. 2016 ).

Since glycosylation of S plays an important role in the viral life cycle and immune-evasion (Casalino et al. 2020; Vigerust & Shepherd 2007; Watanabe et al. 2019 Watanabe et al. , 2020 , we further investigated the glycosylation sites in our map by computing a difference map between our map and the map of biochemically purified S (EMD-8331). All glycosylation sites previously identified by a combination of cryo-EM densities protruding from the expected amino acid side chains and mass spectroscopy ( Table 1 and Table S2 ) (Walls et al. 2016) were found in our map and annotated in yellow (Fig. 3) as in . Two other N-linked glycosylation sites (Asn24 and Asn203 in domain 0), predicted to exist from their sequence but not found by Mass Spectroscopy or in the map of (Walls et al. 2016) , as well as one O-linked glycosylation site (Ser496 in domain B), were newly identified in our density map (Fig. 3, Table 1 ). In addition, we noted extended densities at several glycosylation sites shown in the difference map (annotated in green, Fig. 3 ) between our structure and the previously published one, possibly due to the differences in glycosylation patterns across species (Walls et al. 2016) . In our study the virus was grown in mammalian cells, while the previous study used S ectodomain expressed in insect cells.

N-linked glycosylation in mammalian cells typically results in complex-type glycans with two to four branches extended from the tri-mannosyl core. In contrast, glycosylation in insect cells 6 typically yields truncated or paucimannosidic or oligomannosidic glycans, with few if any complex-type glycans (Tomiya et al. 2004) . Overall there is significant agreement between in-situ and biochemically purified S protein structure, with some notable differences in glycan densities.

Our structure cross-validates the conformation of the spike protein of HCoV-NL63 determined from either the biochemically purified specimen or the intact virions by cryo-EM.

Compared to the S1 protein of betacoronavirus (i.e. SARS-CoV-2), HCoV-NL63 S1 has an additional N-terminal domain (domain 0) that is assumed to be a gene duplication of domain A and is a canonical feature of alphacoronaviruses. Domain B (RBD) of SARS-CoV-2 transitions between upward (open) and downward (closed) conformations due to its inherent flexibility, the viral receptor ACE2 is able to sample the RBD's up conformation to initiate virus-receptor binding.

In contrast, domain B of HCoV-NL63 interacts with domain A, therefore being stabilized in a closed "circle" of HCoV-NL63 S1 (Fig. 4a) . The interface area between domain A and B is ~500 Å 2 and involves 22 hydrophobic interactions (Fig. 4b) according to PDBsum structure bioinformatics analysis (Laskowski et al. 2018) . The three receptor binding motifs (RBMs) on HCoV-NL63 RBD are consequently buried in the interface between domain A and domain B, and prevented from binding to the viral receptor ACE2 (Wu et al. 2009 ). There must exist a mechanism to induce some conformational change to release the RBMs from domain A to allow receptor binding. It was reported that HCoV-NL63 utilizes heparan sulfate proteoglycans as the attachment factor before binding to ACE2 for virus entry (Milewska et al. 2014 (Milewska et al. , 2018 . Whether heparan sulfate binding to HCoV-NL63 S1 or other factors induce conformational changes to release RBMs for ACE2 binding warrants further studies. S proteins of coronaviruses require protease cleavage to release the fusion peptide that inserts into the target membrane to initiate membrane fusion for virus entry. We compared HCoV-NL63 and SARS-CoV-2 S2 fusion machineries by superimposing their structures together and excellent agreement was observed (Fig. 4c) . In both structures, the S2' loop where the host protease cleavage site is located is almost perpendicular to the central helix, protruding at the peripheral of the S trimer and readily accessible to host proteases. The fusion machineries of HCoV-NL63 and SARS-CoV-2 are almost identical except that the S2' trigger loop of HCoV-NL63 forms a short alpha helix before looping back to connect to the fusion peptide while the corresponding loop of SARS-7 CoV-2 is long and flexible. The fusion machinery of coronaviruses provides a potential target for broad anti-coronavirus therapeutics development.

The feasibility of solving the near-atomic resolution structure of membrane-anchored S trimer on the purified viral particle opens a straightforward path for follow-up structural studies of coronaviruses with external reagents like receptors, antibodies or drugs. In addition, the single particle cryo-EM derived structures can be used together with subtomogram averages of the S trimer in the virion to understand complex and heterogeneous modes of interactions between S and various external reagents in different biochemical conditions, and the orientation in 3D of S with respect to the viral membrane, and, ultimately, the cell surface. Such a hybrid approach across different imaging protocols will be very useful for the discovery of authentic structural states of molecular components in a pleomorphic virus particle in the context of pathogenesis relevant to human health.

Vero E6 (ATCC CRL-1586) and MA104 (ATCC CRL-2378.1) cells were maintained at 37°C in a fully humidified atmosphere with 5% CO2 in DMEM (Invitrogen) and M199 (Gibco) medium respectively. All culture media were supplemented with penicillin and streptomycin and 10% FBS (Hyclone). HCoV-NL63 was obtained from BEI Resources, NIAID, NIH: Human Coronavirus, NL63, NR-470.

Monolayers of Vero E6 or MA-104 cells were infected with HCoV-NL63 at MOI 0.5. Culture supernatants were harvested when clear cytopathic effect (CPE) developed. After clarification of the supernatants through a 0.45 micron filter, the virus was pelleted down through a 20% sucrose cushion. Virus was then resuspended in NTE buffer (20 mM Tris, pH 8.0, 120 mM NaCl, 1 mM EDTA) before layered onto 2 ml of 60% OptiPrep (Sigma-Aldrich) and spun at 50,000 x g for 1.5 hour using a SW28 rotor. After ultracentrifugation, the supernatants were removed to leave 4 ml above the virus band. The remaining 4 ml supernatant, the virus band and the underlay of 2 ml of 60% OptiPrep were mixed to reach a final concentration of 20% OptiPrep. The mixture was spun 8 at 360,000 x g for 3.5 hours with a NVT65.2 rotor. The virus band was extracted and buffer exchanged to the NTE buffer using an Amicon Ultra-2 Centrifugal Filter Unit with Ultracel-100 membrane (Millipore).

Three microliters of purified HCoV-NL63 virions (4.5 mg/mL) were first applied to 300-mesh R2/1 + 2 nm C-film grids (Quantifoil) that had been glow-discharged for 15s PELCO easiGlow Motion-corrected micrographs were imported into cryoSPARC for image processing. The contrast transfer function (CTF) was determined using the "Patch CTF Estimation" option, and the micrographs with "CTF fit < 5 Å" were selected using the "Manually Curate Exposures". Then 157 particles were manually picked and subjected to 2D classification into 10 classes, 3 of which were selected as the template for the auto-particle picking, yielding 944,822 particles. Several rounds of 2D classification were then performed to remove the poor 2D class averages, and 300,236 particles were obtained. The initial map was built using ab-initio reconstruction without any symmetry applied. Next, three rounds of heterogeneous refinement were performed to further remove bad particles. The final 3D non-uniform refinement was performed using the selected 82,030 particles with or without C3 symmetry applied, and a 3.4-Å map and a 3.7-Å map were obtained, respectively. Resolution for the final maps was estimated with the 0.143 criterion of the Fourier shell correlation curve. Resolution map was calculated in cryoSPARC using the "Local Resolution Estimation" option. The figures were prepared using UCSF Chimera (Pettersen et al. 2004) or UCSF Chimera X (Goddard et al. 2018) .

One protomer was first computationally extracted (Pintilie et al. 2010) from the 3.4-Å cryo-EM map of the spike protein expressed on HCoV-NL63 coronavirus. The PDB coordinates of biochemically purified HCoV-NL63 spike protein (PDB ID: 5SZS) were then docked into the computationally extracted S protomer map. Residues 883-889 and 993-1000 that were previously unresolved were modeled using SWISS-MODEL (Waterhouse et al. 2018 ). The resultant model was refined using phenix.real_space_refine (Adams et al. 2012 ) and manually optimized with Coot (Emsley et al. 2010 ). The atomic model of protomer was then fitted into the cryo-EM density of the other protomers in the HCoV-NL63 spike trimer using Chimera (Pettersen et al. 2004) , followed by the optimization of the whole model with phenix.real_space_refine. The glycans in the PDB ID: 5SZS model were also used in our model for refinement. The final model was evaluated by MolProbity (Chen et al. 2010 ) and Q-scores (Pintilie et al. 2020) . Statistics of the map reconstruction and model optimization are shown in Table S1 . All figures were prepared using Chimera (Pettersen et al. 2004) or Chimera X (Goddard et al. 2018) . (e) α-helix. 

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

Activation of the SARS coronavirus spike protein via sequential proteolytic cleavage at two distinct sites

The Coronavirus Spike Protein Is a Class I Virus Fusion Protein: Structural and Functional Characterization of the Fusion Core Complex

Shielding and Beyond: The Roles of Glycans in SARS-CoV-2 Spike Protein. bioRxiv : The Preprint Server for Biology

MolProbity: allatom structure validation for macromolecular crystallography

Human Coronavirus

Infection and Other Coronavirus Infections in Children Hospitalized with Acute Respiratory Disease in Hong Kong, China. Clinical Infectious Diseases

Origin and evolution of pathogenic coronaviruses

Features and development of Coot

UCSF ChimeraX: Meeting modern challenges in visualization and analysis

Structures, conformations and distributions of SARS-CoV-2 spike protein trimers on intact virions

Stabilized coronavirus spikes are resistant to conformational changes induced by receptor recognition or proteolysis

PDBsum: Structural summaries of PDB entries

Entry of Human Coronavirus NL63 into the Cell

Human coronavirus NL63 utilizes heparan sulfate proteoglycans for attachment to target cells

Immunogenicity and structures 12 of a rationally designed prefusion MERS-CoV spike antigen

Structures of MERS-CoV spike glycoprotein in complex with sialoside attachment receptors

UCSF Chimera--a visualization system for exploratory research and analysis

Quantitative analysis of cryo-EM density map segmentation by watershed and scale-space filtering, and fitting of structures by alignment to regions

Measurement of atom resolvability in cryo-EM maps with Q-scores

Co-circulation of three camel coronavirus species and recombination of MERS-CoVs in Saudi Arabia

A tail-like assembly at the portal vertex in intact herpes simplex type-1 virions

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

Palmitoylations on murine coronavirus spike proteins are essential for virion assembly and infectivity

Comparing N-glycan processing in mammalian cell lines to native and engineered lepidopteran insect cell lines

Identification of a new human coronavirus

Virus glycosylation: role in virulence and immune interactions

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

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

Review of bats and SARS

Discovery of a subgenotype of human coronavirus NL63 associated with severe lower respiratory tract infection in China

Cryo-Electron Tomography and Subtomogram Averaging

Vulnerabilities in coronavirus glycan shields despite extensive glycosylation

Exploitation of glycosylation in enveloped virus pathobiology

SWISS-MODEL: homology modelling of protein structures and complexes

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

Crystal structure of NL63 respiratory coronavirus receptor-binding domain complexed with its human receptor

Core structure of S2 from the human coronavirus NL63 spike glycoprotein

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

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