key: cord-0298004-pv0e9jae
authors: Wang, Beibei; Zhong, Changqing; Tieleman, D. Peter
title: The supramolecular organization of SARS-CoV and SARS-CoV-2 virions revealed by coarse-grained models of intact virus envelopes
date: 2021-09-20
journal: bioRxiv
DOI: 10.1101/2021.09.16.460716
sha: 00087521166e932bcd178b928c3f2eccaa5e7e51
doc_id: 298004
cord_uid: pv0e9jae

The coronavirus disease 19 (COVID-19) pandemic is causing a global health crisis and has already caused a devastating societal and economic burden. The pathogen, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has a high sequence and architecture identity with SARS-CoV, but far more people have been infected by SARS-CoV-2. Here, combining structural data from cryo-EM and structure prediction, we constructed bottom-up Martini coarse-grained models of intact SARS-CoV and SARS-CoV-2 envelopes. Microsecond molecular dynamics simulations were performed, allowing us to explore their dynamics and supramolecular organization. Both SARS-CoV and SARS-CoV-2 envelopes present a spherical morphology with structural proteins forming multiple string-like islands in the membrane and clusters between heads of spike proteins. Critical differences between the SARS-CoV and SARS-CoV-2 envelopes are the interaction pattern between spike proteins and the flexibility of spike proteins. Our models provide structural and dynamic insights in the SARS virus envelopes, and could be used for further investigation, such as drug design, and fusion and fission processes.

However, the toxicity and transmission capacity of SARS-CoV and SARS-CoV-2 are 23 significantly different. In light of the ongoing global health emergency, there is an urgent need to 24 clarify how the envelope of the CoVs fulfills its function and explore why the infection capacity is 25 different. New structures of SARS-CoV-2 proteins have been obtained by cryo-electron 26 microscopy (cryo-EM) nearly weekly (16) (17) (18) (19) (20) (21) . Computational approaches have also been used for 27 structural prediction of unresolved sections (22) , molecular dockings of different drug molecules 28 on virus proteins (23) (24) (25) (26) , and free energy calculations of the S-ACE2 binding process (27) (28) (29) (30) (31) (32) and 29 the down-up transition of the receptor binding domain (RBD) (33, 34) . The larger scale 30 spatio-temporal processes, such as virion assembly, virus architecture, fusion, and budding, are 31 still poorly understood, and remain challenging for experimental techniques as well as all-atom 32 molecular dynamics (MD) simulations (35) . Coarse-grained (CG) models have been proven to be powerful to probe spatio-temporal 34 large-scale process of complicated biomolecular systems (36) . Martini CG models have been 35 widely used to investigate protein-lipid interactions (PLIs) and protein-protein interactions (PPIs) 36 (37, 38) . In this study, we constructed Martini CG models of SARS-CoV and SARS-CoV- 2 37 envelopes, containing multiple copies of E, M and S proteins and thousands of different lipids; 38 then microsecond (μs) MD simulations were performed to equilibrate the CG models. Overall, our 1 simulations revealed structural and dynamic details of the virus morphology, conformations of 2 three structural proteins (E, M and S), and their PPIs and PLIs for both SARS-CoV and 3 SARS-CoV-2. Our results provide insight into structural and dynamic details of the critical 4 difference between SARS-CoV and SARS-CoV-2 envelopes. Coordinate files of the CG models 5 are available at GitHub 6 (https://github.com/ChangqingZhong/Martini-MD-of-SARS-CoV-and-SARS-CoV-2). 7 8

We constructed Martini CG models of SARS-CoV and SARS-CoV-2 envelopes with Es, Ss 10 and Ms inserted in a lipid vesicle (Fig. 1A) . The lipidomics of the vesicle is according to the 11 composition of the human endoplasmic reticulum (ER)-related membrane (39) , and is asymmetric 12 between the outer and inner leaflets (Table S1 ). Microsecond MD simulations were carried out to 13 equilibrate the models. Longer simulations (up to 12 μs) did not result in observable changes in 14 the virions, so some simulations were performed for 3 μs to limit the computational cost ( Table  15 S2). 16

The virion size varies in different reports (17, 21, 40) , in the range of 50-120 nm for the outer 18 diameter. In order to minimize the required computing resources, the Martini CG models 19 constructed here, with an average outer diameter of ~53 nm, have a total of ~7.3 million CG beads, 20

containing 90 transmembrane proteins imbedded in a membrane of 16,000 lipid molecules with 21 68% PC, 7% PE, 16% LPC, 5% CHOL, and 4% SM in the outer leaflet, and 41% PC, 37% PE, 22 9% PI, 5% PS, 5% CHOL, and 2% SM in the inner leaflet (Fig. 1A-C and Table S1 ). 23

There is no clear consensus on the stoichiometric composition of structural proteins of the 24 virus envelopes so far. The reported number of Ss per vision varies in the range of 20-40 (17, 21) , 25 so 30 Ss were firstly randomly inserted into the equilibrated vesicle. The prefusion conformation 26 was used for all Ss, because the cryo-EM results suggested that more than 95% of Ss on the 27 virions are in prefusion conformation. Then 15 Es, the minimum abundance, were located 28 randomly, and finally limited by the size of our models, as much as 45 Ms were placed, which 29 may be less than previously inferred (35, 41) . A total of 90 structural proteins were included, 30 occupying about 15% of the virion surface and resulting in a 1:180 protein-to-lipid ratio. These 31 structural models enable us to probe the nanoscale organization and dynamics of the SARS-CoV 32 and SARS-CoV-2 envelopes. 33

In all simulations, both the SARS-CoV and SARS-CoV-2 envelopes are stable and present 34 very similar morphology globally (Figs.1D-E and S4-S11). The shapes of both virus envelopes 35 maintain a relatively regular sphere. The spherical morphology has been observed in several 36 cryo-EM studies (17, 20, 21) . The distances between individual lipid heads and the center of mass 37 of the entire virion present slight fluctuations less than 1.5 nm. The landscapes of distance 38 fluctuations of the inner leaflet show almost the same pattern as that of the outer leaflet (Figs. 1 1 and S4-S11), indicating no significant variation of the bilayer thickness and curvature. The local 2 curvature induced by transmembrane proteins was observed in 500 ns atomistic MD simulations 3

and was conjectured to be the driving force of the deviation from an ideal sphere (42) . In our virus 4 models, however, no curvature changes in the membrane were observed in all simulations, which 5 may be the reason that the virions maintain a spherical shape. 

Overall, the homotrimer Ss organize into curvilinear strings with varying length (up to 11 Ss) 18 and a small number of five-membered rings spatially through extensive PPIs between the huge 19 heads (S1 domains), but without obvious aggregation (Figs. 1, 2A and S4-S11). The morphology 20 of spatial organization is consistent with recent cryo-EM results (17, 18, 21) . The SARS-CoV Ss 21 prefer the conformations of dimer and trimer, while SARS-CoV-2 Ss tend to connect into longer 

Visualizing the envelopes without the S ectodomains, it is clear that the TMs of different 5 structural proteins form string-like islands with variable size (up to 18 different proteins) in the 6 envelopes of both SARS-CoV and SARS-CoV-2 (Figs. 3A and S14-S21). The string-like islands 7 was also observed in previous simulations of several different complex membrane models (38, 44) . down is the same as that caused by the crowding of GPCRs (45) . 14 Es, in monomers in our models, have extensive contacts with Ms and Ss, but no tendency to 15 form pentamers was observed. It is noteworthy that more SARS-CoV Es distribute around M and 16 S than SARS-CoV-2 Es do (Fig. S23 ). 17

The recruitment of islands is mainly driven by extensive PPIs between E/M/S and M proteins, Previous morphological studies (41, 46) suggested that a part of Ms form dimers, which also 32 was confirmed in our simulations. The M dimer is tightly bound by the strong and extensive 33 interactions between the two interfaces ( Fig. 3D and 3E ). In the process of dimer formation, the The representative structures of 9 clusters accounting for more than 1% are shown in Fig. 3F . The 4 cryo-EM study has classified M structures into elongated and compact conformations according to 5 the length of the endodomain domain (41) . Our simulations clearly demonstrated these two 6

conformations. The endodomain has interactions with the lipid heads, resulting in the compact 7 conformation (~55%, clusters C2, C3, C5, C6, C7 and C9), while the elongated conformation 8 (~45%, clusters C1, C4 and C8) appears an extended endodomain domain, which may interact 9

with the RNA package (47) . Protein modulates its local lipid environment in a unique way and lipid−protein interactions 13 is regarded as unique fingerprints for membrane proteins (48) . We calculated the lipid 14 depletion-enrichment (D-E) index of the three structural proteins separately ( Table 1 ). The D-E 15 index greater than 1 means enrichment and less than 1 means depletion. Consistent with the 16 protein-protein interactions discussed above, the lipid environments of SARS-CoV and 17 SARS-CoV-2 models are consistent due to the highly similar TMs. shows an almost equally wide range (0.3 º to 100 º) and a densest population at about 35°. The tilt 1 angle distribution of SARS-CoV-2 Ss is consistent with some of the recent cryo-EM results, which 2 also showed a peak at about 60° (17) . The distribution of α indicates that SARS-CoV-2 Ss prefer a 3 more standing conformation, comparing with SARS-CoV Ss. The sequences of the stalks are 4 conserved between SARS-CoV and SARS-CoV-2, so the difference in distribution may result 5 from the different PPI patterns. The flexibility of S makes it easier to search for receptor proteins, 6 but excessive flexibility is not conducive to the formation of stable interactions. Therefore, it may 7 be one of the reasons why SARS-CoV-2 has a stronger infection ability. 8

Two conformations of Ss, "RBD down" and "one RBD up", were observed in our 9

simulations, same as the observation of the virions by cryo-EM (17, 18, 21) . We calculated the 10 distance (h) between residues of the receptor binding domain (RBD) tip (residues 470-490 of 11 SARS-CoV-2 and residues 457-477 of SARS-CoV) and residues at central helix (residues 12 986-996 of SARS-CoV-2 and residues 968-978 of SARS-CoV) (Fig. 4C) . The populations of h in 13 SARS-CoV systems is between 1.2 nm and 5.0 nm with a peak at 3.0 nm, while in SARS-CoV-2 14 systems between 1.3 nm and 6.0 nm with a peak at 3.8 nm (Fig. 4D) . The distributions clearly 15 reveal that SARS-CoV-2 has more RBD in the up state than SARS-CoV. Only Ss with RBD in the 16 up state can bind to ACE2 and infect cells (16) . So higher intrinsic flexibility may be another 17 factor that makes SARS-CoV-2 more infectious than SARS-CoV. oligomerization of E in the virus membrane is still uncertain. Monomer E was adopt in our models, 34 but some works suggested E may form homopentamer (6) , which may induce local membrane 35 curvature. However, the content of E is the least, so the influence on membrane curvature should 36 be limited. Therefore, the ellipsoidal morphology may be mainly caused by the crowding between 37

In the simulations, we are able to observe the forming of S clusters and supramolecular 1 islands by the transmembrane domains of different proteins and lipids in the envelope membrane. 2

The interaction patterns and flexibility of S stalks show agreements with the experimental 3 observations (17, 21, 40) and all-atom simulations (18, 49) , indicating the reliability of the 4 structural and dynamic details obtained from our simulations. The forming of S clusters can 5 restrict the orientation of S heads, while the supramolecular islands slow down the S diffusion in 6 the membrane. Both of these points favor the virus infection by stabilizing interactions between S 7 and ACE2. Previous experiments have confirmed that nanocluster is required for efficient 8 pathogen binding and internalization (50) . 9

Our models, of course, are just approximate to the virus envelopes. In particular, the The S protein has a dense coating of glycans, to evade the host immune system. Atomistic 25 simulations also demonstrated that the glycans play an essential role in modulating the 26 conformational transitions of the S protein (18, 33, 34, 49, 52) . Our models in this study did not 27 include glycans. Apparently, glycans are not involved in the PPIs of the transmembrane domains, 28 which form the heterogeneous protein islands in the virus membrane, while the glycosylation may 29 affect the PPIs between S heads and the size of S clusters. However, the PPI patterns in our 30 models are consistent with these observed in the cryo-EM, as well as these from all-atom 31 simulations, which contain four copies of glycosylated S proteins. It indicates that glycosylation 32 may have little effect on the interaction modes between S proteins. Anyway, to get a deeper Another possible limitation of our simulations is the use of Martini CG force field 2.2, which 37 may limit the protein conformational changes due to the elastic network (53) , such as the RBD 38 opening in our simulations, and may tend to excessive protein aggregation because of excessive 1 inter-protein interactions. Future simulations will try the latest Martini force fields 3.0 (54). Table  25 S1). The vesicle was equilibrated by a 2 μs MD simulation with its radius of gyration stabilizing at 26

The genes of the E and M proteins are conserved in SARS-CoV and SARS-CoV-2 (Fig. S1) , 28 so the same structures of M and E were used in both models. The oligomerization of E differs 29 from monomer (57, 58) to pentamer (6, 58) for van der Waals and electrostatic interactions with a switching function from 10 Å for van der 20

Waals. The systems were minimized for 15000 steps with the steepest descent method, and then 21 equilibrated by short 3.6 ns NVT simulations with incremental time steps of 2fs, 5fs 10fs, 15fs and 22

20fs. Finally, the production simulations were performed with the NPT ensemble and a time step 23 of 20 fs. Both simulation systems were carried out for multiple replicas, for a total of 60 μs (Table  24 S2). All the analyses were performed using VMD (66) , which is also used for visualization and 25 Movie S1. One of the SARS-CoV-2 simulations. The color scheme is the same as Figure 1 .

The authors declare no competing financial interest. The shade of red indicates the contribution. (F) The representative M structure of 9 clusters accounting for more than 1%, resulting from a RMSD-based clustering analysis. The structure was colored according to the physicochemical properties of amino acids: polar residues in green, basic residues in blue, acidic residues in red and nonpolar residues in white. 

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

The molecular biology of coronaviruses

Membrane binding proteins of coronaviruses

Hexamethylene amiloride blocks E protein ion channels and inhibits coronavirus replication

The coronavirus E protein: assembly and beyond. Viruses

Structure and drug binding of the SARS-CoV-2 envelope protein transmembrane domain in lipid bilayers

A conserved domain in the coronavirus membrane protein tail is important for virus assembly

Incorporation of spike and membrane glycoproteins into coronavirus virions

Envelope protein palmitoylations are crucial for murine coronavirus assembly

Genetic evidence for a structural interaction between the carboxy termini of the membrane and nucleocapsid proteins of mouse hepatitis virus

Mechanisms of coronavirus cell entry mediated by the viral spike protein

Distinct conformational states of SARS-CoV-2 spike protein

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

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

ACE2: from vasopeptidase to SARS virus receptor

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

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

In situ structural analysis of SARS-CoV-2 spike reveals flexibility mediated by three hinges

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

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

Molecular Architecture of the SARS-CoV-2 Virus

Protein Structure and Sequence Reanalysis of 2019-nCoV Genome Refutes Snakes as Its Intermediate Host and the Unique Similarity between Its Spike Protein Insertions and HIV-1

Identification of potential inhibitors of SARS-CoV-2 papain-like protease from tropane alkaloids from Schizanthus porrigens: A molecular docking study

Inhibition of SARS-CoV-2 main protease 3CLpro by means of alpha-ketoamide and pyridone-containing pharmaceuticals using in silico molecular docking

Targeting SARS-CoV-2 spike protein of COVID-19 with naturally occurring phytochemicals: an in silico study for drug development

Molecular dynamics simulation perception study of the binding affinity performance for main protease of SARS-CoV-2

Critical Differences between the Binding Features of the Spike Proteins of SARS-CoV-2 and SARS-CoV

Thermodynamics of the Interaction between the Spike Protein of Severe Acute Respiratory Syndrome Coronavirus-2 and the Receptor of Human Angiotensin-Converting Enzyme 2. Effects of Possible Ligands

Computational Alanine Scanning and Structural Analysis of the SARS-CoV-2 Spike Protein/Angiotensin-Converting Enzyme 2 Complex

Is the Rigidity of SARS-CoV-2 Spike Receptor-Binding Motif the Hallmark for Its Enhanced Infectivity? Insights from All-Atom Simulations

Enhanced receptor binding of SARS-CoV-2 through networks of hydrogen-bonding and interactions

Why Does the Novel Coronavirus Spike Protein Interact so Strongly with the Human ACE2? A Thermodynamic Answer

Free Energy Landscapes from SARS-CoV-2 Spike Glycoprotein Simulations Suggest that RBD Opening Can Be Modulated via Interactions in an Allosteric Pocket

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

A Multiscale Coarse-grained Model of the SARS-CoV-2 Virion

The power of coarse graining in biomolecular simulations

Emerging Diversity in Lipid-Protein Interactions

Computational Modeling of Realistic Cell Membranes

Lipids and Their Trafficking: An Integral Part of Cellular Organization

Structural analysis of full-length SARS-CoV-2 spike protein from an advanced vaccine candidate

A structural analysis of M protein in coronavirus assembly and morphology

Atoms to Phenotypes: Molecular Design Principles of Cellular Energy Metabolism

Curvature induction and membrane remodeling by FAM134B reticulon homology domain assist selective ER-phagy

Supramolecular assemblies underpin turnover of outer membrane proteins in bacteria

Organization and Dynamics of Receptor Proteins in a Plasma Membrane

Assembly of the coronavirus envelope: Homotypic interactions between the M proteins

Nucleocapsid-independent specific viral RNA packaging via viral envelope protein and viral RNA signal

Lipid-Protein Interactions Are Unique Fingerprints for Membrane Proteins

Computational epitope map of SARS-CoV-2 spike protein

The Neck Region of the C-type Lectin DC-SIGN Regulates Its Surface Spatiotemporal Organization and Virus-binding Capacity on Antigen-presenting Cells

Membrane Lipid Composition: Effect on Membrane and Organelle Structure, Function and Compartmentalization and Therapeutic Avenues

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

Combining an Elastic Network With a Coarse-Grained Molecular Force Field: Structure, Dynamics, and Intermolecular Recognition

Martini 3: a general purpose force field for coarse-grained molecular dynamics

Simulations Using the CHARMM36 Additive Force Field

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

Structure of a Conserved Golgi Complex-targeting Signal in Coronavirus Envelope Proteins

Structural model of the SARS coronavirus E channel in LMPG micelles

Cryo-EM structures of MERS-CoV and SARS-CoV spike glycoproteins reveal the dynamic receptor binding domains

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

GROMACS: Fast, flexible, and free

The MARTINI force field: Coarse grained model for biomolecular simulations

The MARTINI coarse-grained force field: Extension to proteins

Canonical sampling through velocity rescaling

Molecular-dynamics with coupling to an external bath

VMD: Visual molecular dynamics