Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Cryo-electron microscopy structures of the SARS-CoV spike glycoprotein reveal a prerequisite conformational state for receptor binding

Cryo-electron microscopy structures of the SARS-CoV spike glycoprotein reveal a prerequisite... Cell Research (2017) 27:119-129. © 2017 IBCB, SIBS, CAS All rights reserved 1001-0602/17 $ 32.00 www.nature.com/cr ORIGINAL ARTICLE Cryo-electron microscopy structures of the SARS-CoV spike glycoprotein reveal a prerequisite conformational state for receptor binding 1, * 2, * 2 1 1 1 2, 3 Miao Gui , Wenfei Song , Haixia Zhou , Jingwei Xu , Silian Chen , Ye Xiang , Xinquan Wang Center for Global Health and Infectious Diseases, Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases, Beijing Advanced Innovation Center for Structural Biology, Department of Basic Medical Sciences, School of Medicine, Tsinghua University, Beijing 100084, China; The Ministry of Education Key Laboratory of Protein Science, Beijing Advanced Innovation Center for Structural Biology, Collaborative Innovation Center for Biotherapy, School of Life Sciences, Tsinghua University, Beijing 100084, China; Collaborative Innovation Center for Biotherapy, State Key Laboratory of Biotherapy and Cancer Center, West China Hospital, West China Medical School, Sichuan University, Chengdu, Sichuan 610041, China The global outbreak of SARS in 2002-2003 was caused by the infection of a new human coronavirus SARS- CoV. The infection of SARS-CoV is mediated mainly through the viral surface glycoproteins, which consist of S1 and S2 subunits and form trimer spikes on the envelope of the virions. Here we report the ectodomain structures of the SARS-CoV surface spike trimer in different conformational states determined by single-particle cryo-electron microscopy. The conformation 1 determined at 4.3 Å resolution is three-fold symmetric and has all the three recep- tor-binding C-terminal domain 1 (CTD1s) of the S1 subunits in “down” positions. The binding of the “down” CTD1s to the SARS-CoV receptor ACE2 is not possible due to steric clashes, suggesting that the conformation 1 represents a receptor-binding inactive state. Conformations 2-4 determined at 7.3, 5.7 and 6.8 Å resolutions are all asymmetric, in which one RBD rotates away from the “down” position by different angles to an “up” position. The “up” CTD1 exposes the receptor-binding site for ACE2 engagement, suggesting that the conformations 2-4 represent a recep- tor-binding active state. This conformational change is also required for the binding of SARS-CoV neutralizing anti- bodies targeting the CTD1. This phenomenon could be extended to other betacoronaviruses utilizing CTD1 of the S1 subunit for receptor binding, which provides new insights into the intermediate states of coronavirus pre-fusion spike trimer during infection. Keywords: SARS-CoV; spike glycoprotein; receptor-binding active state; cryo-EM Cell Research (2017) 27:119-129. doi:10.1038/cr.2016.152; published online 23 December 2016 HCoV-229E and HCoV-NL63 and lineage A betacoro- Introduction naviruses HCoV-OC43 and HCoV-HKU1 usually cause Coronaviruses are a large group of highly diverse, mild and self-limiting upper respiratory tract infection enveloped, positive-sense, single-stranded RNA viruses [1]. In 2002, the severe acute respiratory syndrome coro- that infect many mammalian and avian species. Current- navirus (SARS-CoV), a lineage B betacoronavirus, was ly, six coronavirus strains that are able to infect humans identified and infected more than 8 000 persons includ - have been identified. Among them, alphacoronaviruses ing nearly 800 related deaths worldwide in the 2002- 2003 SARS pandemic [2-4]. Ten years later, another highly pathogenic lineage C betacoronavirus named the *These two authors contributed equally to this work. Middle East respiratory syndrome coronavirus (MERS- a b Correspondence: Ye Xiang , Xinquan Wang CoV) emerged in Saudi Arabia in 2012 [5, 6]. Since its E-mail: [email protected] discovery, the MERS-CoV has infected 1 800 persons E-mail: [email protected] including 640 related deaths according to the WHO data Received 25 August 2016; revised 18 November 2016; accepted 30 No- vember 2016; published online 23 December 2016 in August, 2016. These two deadly coronaviruses have Cryo-EM structures of the SARS-CoV spike glycoprotein trimer been extensively studied in epidemiology, virology, clin- the S glycoprotein trimer. Structural comparisons further ical features and other aspects [7-10]. However, there are indicated that a “down” to “up” positional change of the still no approved antiviral drugs and vaccines to treat and CTD1 switches the S glycoprotein trimer from recep- prevent the infections of SARS-CoV and MERS-CoV. tor-binding inactive to active state, which is a prerequi- The spike (S) glycoprotein on the coronavirus en- site for the binding of SARS-CoV receptor ACE2 and for velope is responsible for host cell attachment, receptor the neutralization by monoclonal antibodies. binding, and for mediating host cell membrane and viral membrane fusion during infection. It is synthesized as a Results precursor single polypeptide chain of ~1 300 amino ac- ids and then cleaved by host furin-like proteases into an Structure determination amino (N)-terminal S1 subunit and a carboxyl (C)-termi- By using the Bac-to-Bac insect cell system, we ex- nal S2 subunit [7, 11]. The S1 subunit contains domains pressed and purified a mutant SARS-CoV S glycoprotein for host cell attachment by recognizing cell surface ectodomain, in which the residue Arg667 at the S1/S2 sugar molecules and binding to specific cellular recep- cleavage site was mutated to alanine to enhance sample tors . Therefore, the S1 subunit, especially its homogeneity and a strep tag was added to the C-terminus [12, 13] receptor-binding domain (RBD) is critical in determining to facilitate purification [27] (Figure 1A). The purified cell tropism, host range and zoonotic transmission of SARS-CoV S glycoprotein was subjected to cryo-EM coronaviruses [14, 15]. The S2 subunit contains a hydro- structural analysis using an FEI Titan Krios electron phobic fusion loop and two heptad repeat regions (HR1 microscope equipped with a Gatan K2 Summit direct and HR2), which suggest a coiled helix structure of the electron counting camera (Supplementary information, S2 subunit . Previous studies suggested that three S Figure S1A). Projected secondary structure features [7] monomers assemble to form homo-trimer spikes anchor- were clearly visible in the 2D classification analysis of ing on the outmost viral envelope . Binding of RBD the boxed particles (Supplementary information, Figure [16] to cellular receptors triggers conformational changes in S1B). After 3D classification and refinement, four major the S1 and S2 subunits, leading to the exposure of the fu- different conformational states (conformations 1-4) were sion loop and its insertion into target cell membrane [17]. determined (Supplementary information, Figure S2A). The HR1 and HR2 regions in the S glycoprotein trimer The SARS-CoV spike has an overall mushroom-like then form a six-helix bundle fusion core that bridges the shape. One of the conformation states is closely three- viral and host cell membranes into close apposition to fold symmetric, whereas the other three conformation facilitate fusion [17]. For the highly pathogenic SARS- states show significant asymmetric features in the mush- CoV and MERS-CoV, the RBD in the S1 subunit and the room head region. Further calculation and processing post-fusion core in the S2 subunit have been structurally were performed with C3 symmetry imposed for the and functionally studied as separate domains . A symmetric conformation 1 and without any symmetry [18-23] previous study of the SARS-CoV virions by single-par- imposed for the asymmetric conformations 2-4 (Supple- ticle cryo-electron microscopy (cryo-EM) reported the mentary information, Figures S2 and S3). structure of the S glycoprotein trimers on the virion at a The resolution of the final C3 symmetric conformation low resolution of 16.0 Å . Recently the pre-fusion 1 calculated with 34 152 particles was 4.3 Å (Supple- [16] structures of mouse hepatitis virus (MHV) and human mentary information, Figure S1). The atomic model of coronaviruses HKU1 and HCoV-NL63 S glycoprotein the SARS-CoV S was built based on the C3 symmetric trimers were determined by cryo-EM at 4.0, 4.0 and 3.4 density map (Supplementary information, Table S1A). Å resolutions, respectively . However, high-res- [24-26] olution structures of the highly pathogenic SARS-CoV SARS-CoV S glycoprotein trimer conformation 1 and MERS-CoV S glycoprotein trimers are still missing. In the SARS-CoV S glycoprotein, the β-strand-rich In addition, intermediate states of the coronavirus S gly- S1 subunit is composed of an N-terminal domain (NTD, coprotein trimer are also required for a better understand- residues 14-294) and three C-terminal domains (CTD1, ing of the molecular mechanisms underlying receptor residues 320-516; CTD2, residues 517-578 and CTD3, binding and membrane fusion. residues 579-663; Figure 1A and 1B). The CTD1 func- We report here the cryo-EM structure determination tions as the RBD of SARS-CoV S glycoprotein, which of the SARS-CoV S glycoprotein trimer in four different specifically binds to the cellular receptor ACE2 [18, 28]. conformations. Structural analyses revealed that these The CTD1 is immediately followed by the CTD2 and conformations are different in the position of one C-ter- CTD3, and the NTD is connected to the CTD1 through minal domain 1 (CTD1), which functions as the RBD of a long linker (residues 295-319; Figure 1A and 1B). The SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. Figure 1 Overall structure of the SARS-CoV S glycoprotein. (A) A schematic diagram showing the domain organization of the SARS-CoV S glycoprotein. SP: signal peptide; NTD: N-terminal domain; CTD1: C-terminal domain 1, cyan; Linker: the linker connecting NTD and CTD1, grey; CTD2: C-terminal domain 2, light green; CTD3: C-terminal domain 3, dark green; FP: fusion peptide, red; HR1: heptad repeat 1, pink; HR2: heptad repeat 2; TM: transmembrane domain; CT: cytoplasmic tail. The TM and CT regions are not included in the expression construct. The NTD and HR2 regions that are not resolved in the reconstruction are represented with diagonal stripes. (B) Ribbon diagrams showing the structure of one SARS-CoV S glyco- protein monomer with domains colored as the same as in A. (C) Ribbon diagrams showing the structure of the SARS-CoV S glycoprotein trimer. The three protomers are colored pink, yellow and cyan, respectively. The circles indicate the locations of the unmodeled NTD regions. (D) Surface shadowed diagrams showing the 4.3 Å resolution 3D density map of the SARS- CoV S trimer. The protomers are colored the same as in C. www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer NTDs of MHV and HKU1 S glycoproteins are structural- (Figure 2B-2D). The angles between the long axes of the “up” CTD1s and their projections on the horizontal plane ly similar and both adopt a galectin-like β-sandwich fold. The model to map correlation values are 0.86 for the have increased to 52, 64 and 70 degrees, respectively (Figure 2B-2D). Of note, the NTD and CTD2 around MHV NTD and 0.87 for the HKU1 NTD (Supplementary information, Figure S4A and S4B). Fitting of the MHV the rotated CTD1 do not have significant conformation - al changes (Supplementary information, Figure S5A). or HKU1 NTD structure into the corresponding SARS- CoV S glycoprotein density map gave relatively low Therefore, the rotation of the CTD1 is a hinge motion around the loops connecting NTD to CTD1 and CTD1 resolution to map correlation values (0.55 for MHV NTD and 0.52 for HKU1 NTD; Supplementary information, to CTD2 that are in close proximity (Supplementary in- formation, Figure S5B and S5C), suggesting that these Figure S4C and S4D), indicating that the SARS-CoV NTD has different local conformations, although it may two loops would play key roles in the conformational switch between the “down” and “up” positions of CTD1. still adopt the galectin-like β-sandwich fold. The rela- tively low resolution in this region also did not allow us Although the position of the CTD2 is not affected within the monomer where the CTD1 adopts an “up” confor- to perform ab initio model building of the NTD, there- fore only two strands and a short α-helix were built and mation, the CTD2 may become more liable to undergo conformational changes to expose the fusion loop under- residues 14-260 were not included in the atomic model (Figure 1A and 1B). The α-helix-rich S2 subunit begins neath. after the S1/S2 cleavage site at residue 667 (Figure 1A). The atomic model of the S2 subunit includes the func- Receptor-binding inactive and active states of the S gly- coprotein trimer tionally important fusion peptide (residues 798-815) and HR1 (residues 880-967; Figure 1A and 1B). The C-ter- As the RBD of SARS-CoV S glycoprotein trimer, the CTD1 specifically binds to the cellular receptor ACE2, minal HR2 (residues 1 154-1 183) was not built due to relatively poor density in this region (Figure 1A and 1B). which is a prerequisite for the host cell attachment of the virion and the subsequent membrane fusion [18, 28]. We In the conformation 1 with three-fold symmetry, the three S glycoprotein monomers intertwine around each superimposed the previously determined CTD1-ACE2 complex crystal structure onto one CTD1 of the S glyco- other to form a closely packed mushroom-shaped homo- trimer (Figure 1C and 1D). The triangular head of the tri- protein trimer (Figure 3A). For conformation 1 in which all CTD1s are in the “down” positions, numerous steric mer spike is composed of the NTDs and CTD1s of three S1 subunits. Three CTD1s locate in the center of the clashes were observed between ACE2 and the neighbor- ing CTD1, and the volume of the steric clashes reaches triangular head and are arranged around the 3-fold sym- metry axis (Figure 1C and 1D). Three NTDs locate at the 10 696 Å (Figure 3B and 3C). Therefore, the SARS- outside of the triangular head and each NTD interacts CoV S glycoprotein trimer with all its CTD1s in the with one CTD1 from the neighboring S1 subunit (Figure “down” positions would not be able to bind the cellular 1C and 1D). The stem of the trimer spike consists of a receptor ACE2, suggesting that the three-fold symmetric core helix bundle formed by the long helices of three S2 conformation 1 represents a receptor-binding inactive subunits, and the core helix bundle is further surrounded state of the spike. In contrast, similar structural superim- by CTD2s and CTD3s of three S1 subunits (Figure 1C positions showed that ACE2 binds the “up” CTD1 well and 1D). The three CTD1s in the head all lay on and cov- without steric clashes with other regions of the S glyco- er the top of the S2 subunits (Figures 1C, 1D and 2A). protein trimer (Figure 3D-3F, Supplementary informa- tion, Figure S6A and S6B), suggesting that spikes with SARS-CoV S glycoprotein trimer conformations 2-4 one CTD1 in the “up” position are able to bind receptor We observed three other conformations 2-4 showing ACE2 and asymmetric conformations 2-4 represent a asymmetric features of the CTD1 in the triangular head receptor-binding activate state of the spike. Notably, the (Figure 2B-2D). In the symmetric conformation 1, the “up” position of the CTD1 also exposes one of the cov- CTD1s in the head are all in a “down” position, cover- ered S2 subunits, leaving the space for the large-scale ing the S2 subunits in the stem (Figure 2A). The angle conformational change of the S2 subunit to expose and between the long axis of the CTD1 and its projection on insert the fusion peptide into the target cell membrane. the horizontal plane perpendicular to the 3-fold axis is ~19 degree (Figure 2A). In the conformations 2-4, two The “up” conformation of the CTD1 is also required for CTD1s still adopt the same “down” conformation as in the binding of neutralizing antibodies the conformation 1, whereas one CTD1 rotates outward Several neutralizing antibodies targeting the CTD1 to an “up” position and no longer covers the S2 subunit (m339, 80R and F26G19) have shown potent inhibition SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. Figure 2 Four different conformations of the SARS-CoV S glycoprotein trimer. Top: surface shadowed diagrams showing the four different conformations (conformations 1-4) of the S trimer. The CTD1s are colored pink. Bottom: ribbon diagrams showing S monomers with the semi-transparent CTD1 densities colored pink. The tilt angles of the CTD1s are defined by the angle between the long axis of the CTD1 (red cylinder) and its projection on the horizontal plane (grey ellipse). (A) Three-fold symmetric conformation 1 with all the three CTD1s in the “down” conformations. (B-D) Asymmetric conformations 2-4 with one CTD1 in the “up” conformation. activity against the cell infection of pseudo-typed or trimer is required not only for successful infections of the live SARS-CoV, and the antibody binding epitopes have SARS-CoV virions, by also for efficient neutralization been elucidated by crystal structure determination of the by antibodies targeting the CTD1. antibody-CTD1 complexes [29-31]. These antibodies binds to the receptor-binding site on the CTD1, thereby Discussion directly inhibiting the engagement of ACE2 receptor. Structural superimpositions showed that these antibodies As the known largest class I viral fusion protein, the all have steric clashes with other regions of the S glyco- coronavirus S glycoprotein trimer recognizes a variety protein trimer in the conformation 1 state (Figure 4A- of host cell receptors through the NTD or CTD1 of the 4C). In contrast, the S glycoprotein trimer conformations S1 subunit, and subsequently mediates the viral and cell 2-4 would allow the antibodies to bind the “up” CTD1 membrane fusion through the fusion peptide and two and inhibit its interaction with the ACE2 receptor (Figure heptad repeats of the S2 subunit. Since the discoveries of 4D-4F). These results indicated that the receptor-binding highly pathogenic SARS-CoV in 2002 and MERS-CoV inactive to active state transition of the S glycoprotein in 2012, most studies have been focused on the RBD www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer Figure 3 Models of the SARS-CoV S monomer and trimer bound with the receptor ACE2. (A) “Binding” of the receptor ACE2 (green) to one S monomer (pink) of the conformation 1 S trimer. The CTD1 is in the “down” conformation. (B) “Binding” of the receptor ACE2 (green) to the conformation 1 S trimer. Three CTD1s are all in the “down” conformations. The steric clashes between a neighboring CTD1 (grey) and ACE2 (green) are colored blue. (C) The same model as in A, with the S trimer den- sity map presented. Only the boundary profile of the “bound” ACE2 is shown (green lines) for a better view of the clashes (volume: 10 696 Å ). (D) “Binding” of the receptor ACE2 (green) with the conformation 3 S monomer (pink) of which the CTD1 is in the “up” conformation. (E) “Binding” of the receptor ACE2 (green) to the “up” CTD1 (pink) of the conformation 3 S trimer showing no steric clashes with any neighboring “down” CTD1 (grey). (F) The same model as in D with the S trimer 3D density map presented. All models are generated by superimposing the CTD1-ACE2 complex crystal structure onto the CTD1 of the corresponding SARS-CoV S monomer or trimer. The NTD models are not shown. (NTD or CTD1) of the S1 subunit because it binds to very limited. the host receptors and is also the main target of neutral- In the present study, we determined the structure of izing antibodies during infection. Our knowledge about SARS-CoV S glycoprotein trimer in four different con- the structures of complete SARS-CoV and MERS-CoV formational states. Recently reported cryo-EM structures spike trimers in pre-fusion and post-fusion states are still of the MHV and HKU1 S glycoprotein trimers are simi- SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. Figure 4 Structural superimpositions showing the “binding” of neutralization antibodies to the SARS-CoV S trimers. (A) Struc- tural superimposition of the CTD1-Fab m396 complex (PDB accession code: 2DD8) onto one CTD1 of the SARS-CoV S trimer in conformation 1, showing the “binding” of Fab m396 to the SARS-CoV S trimer. The EM densities of the S trimer are represented using shadowed surfaces in semi-transparent grey. (B-C) Similar structural superimpositions showing the “binding” of neutralization antibodies 80R (PDB accession code: 2GHW; B) and F26G19 (PDB accession code: 3BGF; C). The steric clashes with the “bound” Fab are colored blue and the corresponding volumes are shown in bracket. (D-F) Structural super- impositions of three CTD1-antibody complex structures with the SARS-CoV S trimer in conformation 3. No steric clashes be- tween the “bound” Fab and the S trimer. lar to the conformation 1 of the SARS-CoV S glycopro- coprotein trimers. The protein receptor of the HKU1 has tein trimer in the receptor-binding inactive state, in which not been identified and the NTD in the S1 subunit was all three CTD1s in the S1 subunit are in “down” posi- shown to bind O-acetylated sialic acids on host cells [32]. tions (Figure 2A). The SARS-CoV S glycoprotein trimer The RBD of MHV S glycoprotein is the NTD of the S1 conformations 2-4 in the receptor-binding active state subunit, and structural alignment revealed that the MHV (Figure 2B-2D), in which one CTD1 is in an “up” posi- protein receptor CEACAM1a binds to the NTD in the S tion, were not observed in the MHV and HKU1 S gly- glycoprotein trimer without steric clashes (Supplementary www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer information, Figure S6C) [33]. Therefore, the MHV and one CTD1 from “down” to “up” position would release HKU1 S glycoprotein trimer may not need the prereq- the steric clashes and enable the binding of one receptor uisite receptor-binding inactive to active state transition molecule (Figure 5). Whether the binding of the first re- for receptor binding. This intrinsic receptor-binding dif- ceptor molecule is enough, or the binding of the second ference among these coronavirus S glycoprotein trimers and third receptors is required to trigger conformational may be one of the reasons why the previous studies were changes in the S1 and S2 subunit necessary for mem- not able to reveal other conformations rather than confor- brane fusion will be an interesting question for future mation 1 in the MHV and HKU1 S glycoprotein trimers. studies. Nevertheless, the conformational change of CTD1 may still be needed for the subsequent exposure and confor- Materials and Methods mational changes of the S2 subunit for membrane fusion during the infection of MHV and HKU1. Protein expression and purification A human codon-optimized gene coding the SARS-CoV spike The protein receptor-binding site in the CTD1 has (S) glycoprotein ectodomain (NCBI Accession NP_828851.1) also been confirmed for MERS-CoV and bat corona- residues 1-1 195 with an R667A mutation to enhance sample ho- virus HKU4, which are lineage C betacoronaviruses mogeneity and a C-terminal strep tag for purification was cloned and have the same host receptor DPP4 [19, 20, 34-36]. and inserted into pFastBac-Dual vector (Invitrogen). The construct Structural superimposition of DPP4 complexed with the was transformed into bacterial DH10Bac competent cell and the CTD1 of MERS-CoV or HKU4 onto the SARS-CoV S extracted Bacmid was then transfected into Sf9 cells using Cell- glycoprotein trimer also showed that the “up” confor- fectin II Reagent (Invitrogen). The low-titer viruses were harvest- ed and then amplified to generate high-titer virus stock, which was mation of CTD1 is required for the binding of DPP4 by used to infect 2 L Sf9 cells at a density of 2 × 10 cells/ml. The these two betacoronaviruses (Supplementary informa- supernatant of cell culture containing the secreted SARS-CoV S tion, Figure S7). Therefore, we suggest that the recep- glycoprotein was harvested 60 h after infection, concentrated and tor-binding inactive to active state transition observed in buffer-exchanged to binding buffer (10 mM HEPES, pH 7.2, 500 the SARS-CoV spike would also occur in the spikes of mM NaCl). SARS-CoV S glycoprotein was captured by StrepTac- other betacoronaviruses that bind to their host receptors tin Sepharose High Performance (GE Healthcare) and eluted with through the CTD1. In receptor-binding inactive state, all 10 mM D-desthiobiotin in binding buffer. The eluted protein was then purified by gel filtration chromatography using the Superose three CTD1s in the “down” positions would not allow 6 column (GE Healthcare) pre-equilibrated with HBS buffer (10 efficient binding to receptors (Figure 5). The transition of Figure 5 A cartoon model showing the transition of the S trimer spikes from receptor-binding inactive to active and subse- quent fusogenic states. For betacoronavirus S glycoprotein utilizing the CTD1 as the receptor-binding domain, the state tran- sition of the S trimer with the “down” to “up” conformational change of CTD1 would allow receptor binding and may initiate subsequent conformational changes in the S2 subunits to mediate membrane fusion. SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. mM HEPES, 150 mM NaCl). Fractions containing SARS-CoV S 0.143 criterion. Local resolution variations were estimated using glycoprotein were pooled and concentrated for electron microsco- ResMap [42]. py analysis. Model building and structure refinement The crystal structure of the SARS-CoV S glycoprotein RBD Cryo-EM Aliquots of 3 µl purified SARS-CoV S glycoprotein were ap- (or CTD1, residues 324-502, PDB accession code: 3D0G) and the plied to glow-discharged holey carbon grids (Quantifoil, Cu 400 structures of HKU1 S glycoprotein (PDB accession code: 5i08), mesh, R1.2/1.3) or grids with a layer of continuous ultrathin car- CTD2 (residues 596-673), CTD3 (residues 674-771) and S2 do- bon film (with/without glow-discharge treatment) (Ted Pella, Inc.). main (residues 793-1147) were initially fitted into the SARS-CoV The protein concentration was ~0.34 mg/ml for holey carbon grids density map using UCSF Chimera [43]. Sequence alignment of and ~0.17 mg/ml for continuous carbon grids. The grids were blot- the SARS-CoV S and the fitted structures was performed using ted and then were plunged into liquid ethane using an FEI Vitro- DNAman (Lynnon Corporation, Quebec, Canada) and ClustalX bot. The grids were checked in an FEI F20 microscope operating at [44]. The fitted model was rebuilt using RosettaCM with C3 200 kV with a Gatan 895 4 k × 4 k CCD camera and a small data symmetry imposed and the best output model was selected ac- set was collected for generating the initial model. Data for final cording to the energy and fitness of the model to the EM density classification and refinement were collected on an FEI Titan Krios map [45]. The model of CTD2, CTD3 and S2 region was refined microscope operating at 300 kV using a K2 Summit camera (Gatan using RosettaRelax and PHENIX real-space refinement [46, 47]. Inc.) in super-resolution mode with a nominal magnification of 22 The crystal structure of the CTD1 domain was fitted into the EM 500× (yielding a calibrated pixel size of 1.32 Å). Each image was density and was refined as a rigid body in real space by using fractionated into 32 movie frames with a total exposure time of 8 s PHENIX. Then the two models were merged, manually adjusted and at a dose rate of ~8 counts per physical pixel per second (~4.7 in COOT and refined in PHENIX again with reference model re- electrons/Å /s). UCSFImage4 was used for all data collection [37]. strains, secondary structure restrains and geometry restrains [48]. Cross-validation of overfitting was performed following the pro- cedures described before [49, 50]. Briefly, the atom coordinates of Image processing The image processing procedures were summarized in Supple- the model (including CTD1, CTD2, CTD3 and S2) were randomly mentary information, Figure S2A. An initial map was generated displaced by 0.5 Å using PHENIX PDB tools. Then the displaced using the program EMAN2 with the images collected on the F20 model was refined against one of the two half maps in reciprocal microscope [38]. A total of 3 309 movie stacks were collected space by using PHENIX. FSC curves were calculated between the using the K2 camera. All movie frames were aligned using the refined model and the corresponding half map that was used for program motioncorr and the CTF parameters were determined the refinement (half1, FSC ) and between the refined model and work by CTFFIND4 [39, 40]. A total of 4 015 particles were manually the other half map (half2, FSC ) that was not used for the model test picked and 2D classifications were performed in RELION 1.3 [41]. refinement. No big gaps were observed between the work and test Four representative class averaged images were selected as refer- FSC curves (Supplementary information, Figure S1D), indicating ences for automatic particle picking with the whole data set using that the model was not overfitted. Molprobity was used to evaluate RELION 1.4. All 2D and 3D classifications and refinements were the final refined model. performed using RELION 1.4. Auto-picked particles were visu- Models of the three asymmetric conformations were built by ally inspected and then were selected by several rounds of refer- fitting the symmetric model of SARS-CoV S into the density maps ence-free 2D classifications. During the 2D classification, particles using UCSF Chimera. The fitted models were then refined as rigid from classes not showing clear secondary structure features were bodies in real space using PHENIX. For the subunit with the “up” deleted, finally yielding 210 129 selected particles. These selected CTD1, the subunit was split into two rigid bodies for refinement, particles were subjected to an initial 3D refinement with C3 sym- including the “up” CTD1 and the rest part of the subunit. metry imposed, followed by two runs of 3D classifications. A total Structural comparisons between different density maps and of 52 983 particles of the best class were selected and subjected to between density maps and models were performed using UCSF a 3D auto-refine. At the final stage of the refinement, 34 152 par - Chimera measure correlation, PHENIX get_cc_mtz_mtz and get_ ticles with a higher “loglikelicontribution” value were used. The cc_mtz_pdb. The cross correlation values were listed in Supple- resolution of the final C3 symmetric density map was 4.3 Å post mentary information, Figure S4A-S4D and Table S1B. processing in RELION. The SARS-CoV S RBD-receptor complex (PDB accession The initial 210 129 selected particles were also subjected to 3D code: 3R4D) and SARS-CoV S RBD-Fab complexes (PDB ac- refinements and classifications without any symmetry imposed. cession code: 2GHW, 2DD8 and 3BGF) were superimposed onto The central parts of the different classes were similar but most the models with the SARS-CoV S RBD structure as reference classes showed obvious asymmetric features with one of three using the “match” command in UCSF Chimera. To calculate the CTD1 domains protruding up. These asymmetric particles could “up” angles of the CTD1s (Figure 2), the horizontal plane of the be grouped into three major classes according to the tilt angle of S perpendicular to the 3-fold axis and the long axis of the CTD1 the “up” CTD1. Particles from these classes were separately sub- were generated using the UCSF Chimera “define” command and jected to 3D auto-refine. The resolutions of the final density maps then the angle between the axis and the plane was calculated using were 7.3, 5.7 and 6.8 Å for the three classes. the UCSF Chimera “angle” command. The overlap volumes be- The handedness of the density maps was verified by docking tween the fitted receptor and the S spike were also calculated using the crystal structure of SARS-CoV S glycoprotein CTD1 into the UCSF Chimera. Two atoms were considered to be overlapped if maps. Reported resolutions are based on the gold-standard FSC their overlap score > 0. The overlap score is defined as the sum of www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer two VDW (van der Waals) radii minus the distance between them 7 Du L, He Y, Zhou Y, Liu S, Zheng BJ, Jiang S. The spike and minus an allowance (0.4 Å) for potentially hydrogen-bonded protein of SARS-CoV--a target for vaccine and therapeutic pairs. All the overlap atoms were used for the generation of the development. Nat Rev Microbiol 2009; 7:226-236. steric clash area. The volume of the steric clashes was calculated 8 de Wit E, van Doremalen N, Falzarano D, Munster VJ. SARS using UCSF Chimera “measure volume” command. All figures and MERS: recent insights into emerging coronaviruses. Nat were generated with UCSF Chimera and Pymol (The PyMOL Mo- Rev Microbiol 2016; 14:523-534. lecular Graphics System, Version 1.7.0.5 Schrödinger, LLC.). 9 Chan JF, Lau SK, To KK, Cheng VC, Woo PC, Yuen KY. Middle East respiratory syndrome coronavirus: another zoo- notic betacoronavirus causing SARS-like disease. Clin Micro- Accession numbers The cryo-EM maps and related materials have been deposited biol Rev 2015; 28:465-522. to the EM Data Bank under accession codes EMD-6679, EMD- 10 Vijay R, Perlman S. Middle East respiratory syndrome and 6680, EMD-6681 and EMD-6682. The atomic coordinate has been severe acute respiratory syndrome. Curr Opin Virol 2016; deposited to the Protein Data Bank under accession code 5WRG. 16:70-76. 11 Millet JK, Whittaker GR. Host cell proteases: critical deter- minants of coronavirus tropism and pathogenesis. Virus Res Acknowledgments 2015; 202:120-134. 12 Li F. Receptor recognition mechanisms of coronaviruses: a We thank the Cryo-EM and Computing Platforms of the decade of structural studies. J Virol 2015; 89:1954-1964. Tsinghua University Branch of the National Center for Protein 13 Li F. Structure, function, and evolution of coronavirus spike Sciences (Beijing) for facility support. This work was supported proteins. Annu Rev Virol 2016; 3:237-261. by the National Key Plan for Scientific Research and Development 14 Lu G, Wang Q, Gao GF. Bat-to-human: spike features deter- of China (2016YFD0500307 and 2016YFA0501100), the National mining ‘host jump’ of coronaviruses SARS-CoV, MERS-CoV, Natural Science Foundation of China (81471929, U1405228, and beyond. Trends Microbiol 2015; 23:468-478. 31470721 and 81550001) and the Beijing Advanced Innovation 15 Graham RL, Baric RS. Recombination, reservoirs, and the Center for Structural Biology. modular spike: mechanisms of coronavirus cross-species transmission. J Virol 2010; 84:3134-3146. Author Contributions 16 Beniac DR, Andonov A, Grudeski E, Booth TF. Architecture MG, WS, YX and XW designed the experiments. WS designed of the SARS coronavirus prefusion spike. Nat Struct Mol Biol and cloned the protein constructs. WS and HZ carried out protein 2006; 13:751-752. expression, purification and biophysical characterization. MG, JX 17 Hofmann H, Pohlmann S. Cellular entry of the SARS corona- and SC performed cryo-EM sample preparation and data collec- virus. Trends Microbiol 2004; 12:466-472. tion. MG processed the cryo-EM data and built the atomic model. 18 Li F, Li W, Farzan M, Harrison SC. Structure of SARS coro- MG, WS, YX and XW analyzed the data. MG, WS, YX and XW navirus spike receptor-binding domain complexed with recep- wrote the manuscript with input from all authors. tor. Science 2005; 309:1864-1868. 19 Lu G, Hu Y, Wang Q, et al. Molecular basis of binding be- Competing Financial Interests tween novel human coronavirus MERS-CoV and its receptor The authors declare no competing financial interests. CD26. Nature 2013; 500:227-231. 20 Wang N, Shi X, Jiang L, et al. Structure of MERS-CoV spike References receptor-binding domain complexed with human receptor DPP4. Cell Res 2013; 23:986-993. 21 Gao J, Lu G, Qi J, et al. Structure of the fusion core and inhi- 1 Su S, Wong G, Shi W, et al. Epidemiology, genetic recombi- bition of fusion by a heptad repeat peptide derived from the nation, and pathogenesis of coronaviruses. Trends Microbiol S protein of Middle East respiratory syndrome coronavirus. J 2016; 24:490-502. Virol 2013; 87:13134-13140. 2 Ksiazek TG, Erdman D, Goldsmith CS, et al. A novel coro- 22 Lu L, Liu Q, Zhu Y, et al. Structure-based discovery of Mid- navirus associated with severe acute respiratory syndrome. N dle East respiratory syndrome coronavirus fusion inhibitor. Engl J Med 2003; 348:1953-1966. Nat Commun 2014; 5:3067. 3 Drosten C, Gunther S, Preiser W, et al. Identification of a 23 Supekar VM, Bruckmann C, Ingallinella P, Bianchi E, Pessi novel coronavirus in patients with severe acute respiratory A, Carfi A. Structure of a proteolytically resistant core from syndrome. N Engl J Med 2003; 348:1967-1976. the severe acute respiratory syndrome coronavirus S2 fusion 4 Peiris JS, Lai ST, Poon LL, et al. Coronavirus as a possible protein. Proc Natl Acad Sci USA 2004; 101:17958-17963. cause of severe acute respiratory syndrome. Lancet 2003; 24 Walls AC, Tortorici MA, Bosch BJ, et al. Cryo-electron mi- 361:1319-1325. croscopy structure of a coronavirus spike glycoprotein trimer. 5 Zaki AM, van Boheemen S, Bestebroer TM, Osterhaus Nature 2016; 531:114-117. AD, Fouchier RA. Isolation of a novel coronavirus from a 25 Kirchdoerfer RN, Cottrell CA, Wang N, et al. Pre-fusion man with pneumonia in Saudi Arabia. N Engl J Med 2012; structure of a human coronavirus spike protein. Nature 2016; 367:1814-1820. 531:118-121. 6 de Groot RJ, Baker SC, Baric RS, et al. Middle East respira- 26 Walls AC, Tortorici MA, Frenz B, et al. Glycan shield and tory syndrome coronavirus (MERS-CoV): announcement of epitope masking of a coronavirus spike protein observed by the Coronavirus Study Group. J Virol 2013; 87:7790-7792. SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. cryo-electron microscopy. Nat Struct Mol Biol 2016; 23:899- images by UCSFImage. J Struct Biol 2015; 192:174-178. 905. 38 Tang G, Peng L, Baldwin PR, et al. EMAN2: an extensible 27 Belouzard S, Chu VC, Whittaker GR. Activation of the SARS image processing suite for electron microscopy. J Struct Biol coronavirus spike protein via sequential proteolytic cleavage 2007; 157:38-46. at two distinct sites. Proc Natl Acad Sci USA 2009; 106:5871- 39 Li X, Mooney P, Zheng S, et al. Electron counting and 5876. beam-induced motion correction enable near-atomic-resolu- 28 Li W, Moore MJ, Vasilieva N, et al. Angiotensin-converting tion single-particle cryo-EM. Nat Methods 2013; 10:584-590. enzyme 2 is a functional receptor for the SARS coronavirus. 40 Rohou A, Grigorieff N. CTFFIND4: fast and accurate defocus Nature 2003; 426:450-454. estimation from electron micrographs. J Struct Biol 2015; 29 Hwang WC, Lin Y, Santelli E, et al. Structural basis of neu- 192:216-221. tralization by a human anti-severe acute respiratory syndrome 41 Scheres SH. RELION: implementation of a Bayesian ap- spike protein antibody, 80R. J Biol Chem 2006; 281:34610- proach to cryo-EM structure determination. J Struct Biol 34616. 2012; 180:519-530. 30 Prabakaran P, Gan J, Feng Y, et al. Structure of severe acute 42 Kucukelbir A, Sigworth FJ, Tagare HD. Quantifying the lo- respiratory syndrome coronavirus receptor-binding domain cal resolution of cryo-EM density maps. Nat Methods 2014; complexed with neutralizing antibody. J Biol Chem 2006; 11:63-65. 281:15829-15836. 43 Pettersen EF, Goddard TD, Huang CC, et al. UCSF Chimera 31 Pak JE, Sharon C, Satkunarajah M, et al. Structural insights — a visualization system for exploratory research and analy- into immune recognition of the severe acute respiratory syn- sis. J Comput Chem 2004; 25:1605-1612. drome coronavirus S protein receptor binding domain. J Mol 44 Larkin MA, Blackshields G, Brown NP, et al. Clustal W and Biol 2009; 388:815-823. Clustal X version 2.0. Bioinformatics 2007; 23:2947-2948. 32 Huang X, Dong W, Milewska A, et al. Human coronavirus 45 Song Y, DiMaio F, Wang RY, et al. High-resolution compar- HKU1 spike protein uses O-acetylated sialic acid as an attach- ative modeling with RosettaCM. Structure 2013; 21:1735- ment receptor determinant and employs hemagglutinin-es- 1742. terase protein as a receptor-destroying enzyme. J Virol 2015; 46 DiMaio F, Tyka MD, Baker ML, Chiu W, Baker D. Refine- 89:7202-7213. ment of protein structures into low-resolution density maps 33 Peng G, Sun D, Rajashankar KR, Qian Z, Holmes KV, Li F. using rosetta. J Mol Biol 2009; 392:181-190. Crystal structure of mouse coronavirus receptor-binding do- 47 Adams PD, Afonine PV, Bunkoczi G, et al. PHENIX: a com- main complexed with its murine receptor. Proc Natl Acad Sci prehensive Python-based system for macromolecular structure USA 2011; 108:10696-10701. solution. Acta Crystallogr D Biol Crystallogr 2010; 66:213- 34 Wang Q, Qi J, Yuan Y, et al. Bat origins of MERS-CoV sup- 221. ported by bat coronavirus HKU4 usage of human receptor 48 Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and CD26. Cell Host Microbe 2014; 16:328-337. development of Coot. Acta Crystallogr D Biol Crystallogr 35 Raj VS, Mou H, Smits SL, et al. Dipeptidyl peptidase 4 2010; 66:486-501. is a functional receptor for the emerging human coronavi- 49 Brown A, Long F, Nicholls RA, Toots J, Emsley P, Mur- rus-EMC. Nature 2013; 495:251-254. shudov G. Tools for macromolecular model building and re- 36 Yang Y, Du L, Liu C, et al. Receptor usage and cell entry finement into electron cryo-microscopy reconstructions. Acta of bat coronavirus HKU4 provide insight into bat-to-human Crystallogr D Biol Crystallogr 2015; 71:136-153. transmission of MERS coronavirus. Proc Natl Acad Sci USA 50 Amunts A, Brown A, Bai XC, et al. Structure of the yeast mi- 2014; 111:12516-12521. tochondrial large ribosomal subunit. Science 2014; 343:1485- 37 Li X, Zheng S, Agard DA, Cheng Y. Asynchronous data ac- 1489. quisition and on-the-fly analysis of dose fractionated cryoEM (Supplementary information is linked to the online version of the paper on the Cell Research website.) www.cell-research.com | Cell Research | SPRINGER NATURE http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Cell Research Pubmed Central

Cryo-electron microscopy structures of the SARS-CoV spike glycoprotein reveal a prerequisite conformational state for receptor binding

Gui, Miao; Song, Wenfei; Zhou, Haixia; Xu, Jingwei; Chen, Silian; Xiang, Ye; Wang, Xinquan
Cell Research , Volume 27 (1) – Dec 23, 2016
Download PDF
11 pages

Loading next page...
 
/lp/pubmed-central/cryo-electron-microscopy-structures-of-the-sars-cov-spike-glycoprotein-iEUfSm00VF

References (107)

Publisher
Pubmed Central
Copyright
Copyright © 2017 Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences
ISSN
1001-0602
eISSN
1748-7838
DOI
10.1038/cr.2016.152
Publisher site
See Article on Publisher Site

Abstract

Cell Research (2017) 27:119-129. © 2017 IBCB, SIBS, CAS All rights reserved 1001-0602/17 $ 32.00 www.nature.com/cr ORIGINAL ARTICLE Cryo-electron microscopy structures of the SARS-CoV spike glycoprotein reveal a prerequisite conformational state for receptor binding 1, * 2, * 2 1 1 1 2, 3 Miao Gui , Wenfei Song , Haixia Zhou , Jingwei Xu , Silian Chen , Ye Xiang , Xinquan Wang Center for Global Health and Infectious Diseases, Collaborative Innovation Center for Diagnosis and Treatment of Infectious Diseases, Beijing Advanced Innovation Center for Structural Biology, Department of Basic Medical Sciences, School of Medicine, Tsinghua University, Beijing 100084, China; The Ministry of Education Key Laboratory of Protein Science, Beijing Advanced Innovation Center for Structural Biology, Collaborative Innovation Center for Biotherapy, School of Life Sciences, Tsinghua University, Beijing 100084, China; Collaborative Innovation Center for Biotherapy, State Key Laboratory of Biotherapy and Cancer Center, West China Hospital, West China Medical School, Sichuan University, Chengdu, Sichuan 610041, China The global outbreak of SARS in 2002-2003 was caused by the infection of a new human coronavirus SARS- CoV. The infection of SARS-CoV is mediated mainly through the viral surface glycoproteins, which consist of S1 and S2 subunits and form trimer spikes on the envelope of the virions. Here we report the ectodomain structures of the SARS-CoV surface spike trimer in different conformational states determined by single-particle cryo-electron microscopy. The conformation 1 determined at 4.3 Å resolution is three-fold symmetric and has all the three recep- tor-binding C-terminal domain 1 (CTD1s) of the S1 subunits in “down” positions. The binding of the “down” CTD1s to the SARS-CoV receptor ACE2 is not possible due to steric clashes, suggesting that the conformation 1 represents a receptor-binding inactive state. Conformations 2-4 determined at 7.3, 5.7 and 6.8 Å resolutions are all asymmetric, in which one RBD rotates away from the “down” position by different angles to an “up” position. The “up” CTD1 exposes the receptor-binding site for ACE2 engagement, suggesting that the conformations 2-4 represent a recep- tor-binding active state. This conformational change is also required for the binding of SARS-CoV neutralizing anti- bodies targeting the CTD1. This phenomenon could be extended to other betacoronaviruses utilizing CTD1 of the S1 subunit for receptor binding, which provides new insights into the intermediate states of coronavirus pre-fusion spike trimer during infection. Keywords: SARS-CoV; spike glycoprotein; receptor-binding active state; cryo-EM Cell Research (2017) 27:119-129. doi:10.1038/cr.2016.152; published online 23 December 2016 HCoV-229E and HCoV-NL63 and lineage A betacoro- Introduction naviruses HCoV-OC43 and HCoV-HKU1 usually cause Coronaviruses are a large group of highly diverse, mild and self-limiting upper respiratory tract infection enveloped, positive-sense, single-stranded RNA viruses [1]. In 2002, the severe acute respiratory syndrome coro- that infect many mammalian and avian species. Current- navirus (SARS-CoV), a lineage B betacoronavirus, was ly, six coronavirus strains that are able to infect humans identified and infected more than 8 000 persons includ - have been identified. Among them, alphacoronaviruses ing nearly 800 related deaths worldwide in the 2002- 2003 SARS pandemic [2-4]. Ten years later, another highly pathogenic lineage C betacoronavirus named the *These two authors contributed equally to this work. Middle East respiratory syndrome coronavirus (MERS- a b Correspondence: Ye Xiang , Xinquan Wang CoV) emerged in Saudi Arabia in 2012 [5, 6]. Since its E-mail: [email protected] discovery, the MERS-CoV has infected 1 800 persons E-mail: [email protected] including 640 related deaths according to the WHO data Received 25 August 2016; revised 18 November 2016; accepted 30 No- vember 2016; published online 23 December 2016 in August, 2016. These two deadly coronaviruses have Cryo-EM structures of the SARS-CoV spike glycoprotein trimer been extensively studied in epidemiology, virology, clin- the S glycoprotein trimer. Structural comparisons further ical features and other aspects [7-10]. However, there are indicated that a “down” to “up” positional change of the still no approved antiviral drugs and vaccines to treat and CTD1 switches the S glycoprotein trimer from recep- prevent the infections of SARS-CoV and MERS-CoV. tor-binding inactive to active state, which is a prerequi- The spike (S) glycoprotein on the coronavirus en- site for the binding of SARS-CoV receptor ACE2 and for velope is responsible for host cell attachment, receptor the neutralization by monoclonal antibodies. binding, and for mediating host cell membrane and viral membrane fusion during infection. It is synthesized as a Results precursor single polypeptide chain of ~1 300 amino ac- ids and then cleaved by host furin-like proteases into an Structure determination amino (N)-terminal S1 subunit and a carboxyl (C)-termi- By using the Bac-to-Bac insect cell system, we ex- nal S2 subunit [7, 11]. The S1 subunit contains domains pressed and purified a mutant SARS-CoV S glycoprotein for host cell attachment by recognizing cell surface ectodomain, in which the residue Arg667 at the S1/S2 sugar molecules and binding to specific cellular recep- cleavage site was mutated to alanine to enhance sample tors . Therefore, the S1 subunit, especially its homogeneity and a strep tag was added to the C-terminus [12, 13] receptor-binding domain (RBD) is critical in determining to facilitate purification [27] (Figure 1A). The purified cell tropism, host range and zoonotic transmission of SARS-CoV S glycoprotein was subjected to cryo-EM coronaviruses [14, 15]. The S2 subunit contains a hydro- structural analysis using an FEI Titan Krios electron phobic fusion loop and two heptad repeat regions (HR1 microscope equipped with a Gatan K2 Summit direct and HR2), which suggest a coiled helix structure of the electron counting camera (Supplementary information, S2 subunit . Previous studies suggested that three S Figure S1A). Projected secondary structure features [7] monomers assemble to form homo-trimer spikes anchor- were clearly visible in the 2D classification analysis of ing on the outmost viral envelope . Binding of RBD the boxed particles (Supplementary information, Figure [16] to cellular receptors triggers conformational changes in S1B). After 3D classification and refinement, four major the S1 and S2 subunits, leading to the exposure of the fu- different conformational states (conformations 1-4) were sion loop and its insertion into target cell membrane [17]. determined (Supplementary information, Figure S2A). The HR1 and HR2 regions in the S glycoprotein trimer The SARS-CoV spike has an overall mushroom-like then form a six-helix bundle fusion core that bridges the shape. One of the conformation states is closely three- viral and host cell membranes into close apposition to fold symmetric, whereas the other three conformation facilitate fusion [17]. For the highly pathogenic SARS- states show significant asymmetric features in the mush- CoV and MERS-CoV, the RBD in the S1 subunit and the room head region. Further calculation and processing post-fusion core in the S2 subunit have been structurally were performed with C3 symmetry imposed for the and functionally studied as separate domains . A symmetric conformation 1 and without any symmetry [18-23] previous study of the SARS-CoV virions by single-par- imposed for the asymmetric conformations 2-4 (Supple- ticle cryo-electron microscopy (cryo-EM) reported the mentary information, Figures S2 and S3). structure of the S glycoprotein trimers on the virion at a The resolution of the final C3 symmetric conformation low resolution of 16.0 Å . Recently the pre-fusion 1 calculated with 34 152 particles was 4.3 Å (Supple- [16] structures of mouse hepatitis virus (MHV) and human mentary information, Figure S1). The atomic model of coronaviruses HKU1 and HCoV-NL63 S glycoprotein the SARS-CoV S was built based on the C3 symmetric trimers were determined by cryo-EM at 4.0, 4.0 and 3.4 density map (Supplementary information, Table S1A). Å resolutions, respectively . However, high-res- [24-26] olution structures of the highly pathogenic SARS-CoV SARS-CoV S glycoprotein trimer conformation 1 and MERS-CoV S glycoprotein trimers are still missing. In the SARS-CoV S glycoprotein, the β-strand-rich In addition, intermediate states of the coronavirus S gly- S1 subunit is composed of an N-terminal domain (NTD, coprotein trimer are also required for a better understand- residues 14-294) and three C-terminal domains (CTD1, ing of the molecular mechanisms underlying receptor residues 320-516; CTD2, residues 517-578 and CTD3, binding and membrane fusion. residues 579-663; Figure 1A and 1B). The CTD1 func- We report here the cryo-EM structure determination tions as the RBD of SARS-CoV S glycoprotein, which of the SARS-CoV S glycoprotein trimer in four different specifically binds to the cellular receptor ACE2 [18, 28]. conformations. Structural analyses revealed that these The CTD1 is immediately followed by the CTD2 and conformations are different in the position of one C-ter- CTD3, and the NTD is connected to the CTD1 through minal domain 1 (CTD1), which functions as the RBD of a long linker (residues 295-319; Figure 1A and 1B). The SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. Figure 1 Overall structure of the SARS-CoV S glycoprotein. (A) A schematic diagram showing the domain organization of the SARS-CoV S glycoprotein. SP: signal peptide; NTD: N-terminal domain; CTD1: C-terminal domain 1, cyan; Linker: the linker connecting NTD and CTD1, grey; CTD2: C-terminal domain 2, light green; CTD3: C-terminal domain 3, dark green; FP: fusion peptide, red; HR1: heptad repeat 1, pink; HR2: heptad repeat 2; TM: transmembrane domain; CT: cytoplasmic tail. The TM and CT regions are not included in the expression construct. The NTD and HR2 regions that are not resolved in the reconstruction are represented with diagonal stripes. (B) Ribbon diagrams showing the structure of one SARS-CoV S glyco- protein monomer with domains colored as the same as in A. (C) Ribbon diagrams showing the structure of the SARS-CoV S glycoprotein trimer. The three protomers are colored pink, yellow and cyan, respectively. The circles indicate the locations of the unmodeled NTD regions. (D) Surface shadowed diagrams showing the 4.3 Å resolution 3D density map of the SARS- CoV S trimer. The protomers are colored the same as in C. www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer NTDs of MHV and HKU1 S glycoproteins are structural- (Figure 2B-2D). The angles between the long axes of the “up” CTD1s and their projections on the horizontal plane ly similar and both adopt a galectin-like β-sandwich fold. The model to map correlation values are 0.86 for the have increased to 52, 64 and 70 degrees, respectively (Figure 2B-2D). Of note, the NTD and CTD2 around MHV NTD and 0.87 for the HKU1 NTD (Supplementary information, Figure S4A and S4B). Fitting of the MHV the rotated CTD1 do not have significant conformation - al changes (Supplementary information, Figure S5A). or HKU1 NTD structure into the corresponding SARS- CoV S glycoprotein density map gave relatively low Therefore, the rotation of the CTD1 is a hinge motion around the loops connecting NTD to CTD1 and CTD1 resolution to map correlation values (0.55 for MHV NTD and 0.52 for HKU1 NTD; Supplementary information, to CTD2 that are in close proximity (Supplementary in- formation, Figure S5B and S5C), suggesting that these Figure S4C and S4D), indicating that the SARS-CoV NTD has different local conformations, although it may two loops would play key roles in the conformational switch between the “down” and “up” positions of CTD1. still adopt the galectin-like β-sandwich fold. The rela- tively low resolution in this region also did not allow us Although the position of the CTD2 is not affected within the monomer where the CTD1 adopts an “up” confor- to perform ab initio model building of the NTD, there- fore only two strands and a short α-helix were built and mation, the CTD2 may become more liable to undergo conformational changes to expose the fusion loop under- residues 14-260 were not included in the atomic model (Figure 1A and 1B). The α-helix-rich S2 subunit begins neath. after the S1/S2 cleavage site at residue 667 (Figure 1A). The atomic model of the S2 subunit includes the func- Receptor-binding inactive and active states of the S gly- coprotein trimer tionally important fusion peptide (residues 798-815) and HR1 (residues 880-967; Figure 1A and 1B). The C-ter- As the RBD of SARS-CoV S glycoprotein trimer, the CTD1 specifically binds to the cellular receptor ACE2, minal HR2 (residues 1 154-1 183) was not built due to relatively poor density in this region (Figure 1A and 1B). which is a prerequisite for the host cell attachment of the virion and the subsequent membrane fusion [18, 28]. We In the conformation 1 with three-fold symmetry, the three S glycoprotein monomers intertwine around each superimposed the previously determined CTD1-ACE2 complex crystal structure onto one CTD1 of the S glyco- other to form a closely packed mushroom-shaped homo- trimer (Figure 1C and 1D). The triangular head of the tri- protein trimer (Figure 3A). For conformation 1 in which all CTD1s are in the “down” positions, numerous steric mer spike is composed of the NTDs and CTD1s of three S1 subunits. Three CTD1s locate in the center of the clashes were observed between ACE2 and the neighbor- ing CTD1, and the volume of the steric clashes reaches triangular head and are arranged around the 3-fold sym- metry axis (Figure 1C and 1D). Three NTDs locate at the 10 696 Å (Figure 3B and 3C). Therefore, the SARS- outside of the triangular head and each NTD interacts CoV S glycoprotein trimer with all its CTD1s in the with one CTD1 from the neighboring S1 subunit (Figure “down” positions would not be able to bind the cellular 1C and 1D). The stem of the trimer spike consists of a receptor ACE2, suggesting that the three-fold symmetric core helix bundle formed by the long helices of three S2 conformation 1 represents a receptor-binding inactive subunits, and the core helix bundle is further surrounded state of the spike. In contrast, similar structural superim- by CTD2s and CTD3s of three S1 subunits (Figure 1C positions showed that ACE2 binds the “up” CTD1 well and 1D). The three CTD1s in the head all lay on and cov- without steric clashes with other regions of the S glyco- er the top of the S2 subunits (Figures 1C, 1D and 2A). protein trimer (Figure 3D-3F, Supplementary informa- tion, Figure S6A and S6B), suggesting that spikes with SARS-CoV S glycoprotein trimer conformations 2-4 one CTD1 in the “up” position are able to bind receptor We observed three other conformations 2-4 showing ACE2 and asymmetric conformations 2-4 represent a asymmetric features of the CTD1 in the triangular head receptor-binding activate state of the spike. Notably, the (Figure 2B-2D). In the symmetric conformation 1, the “up” position of the CTD1 also exposes one of the cov- CTD1s in the head are all in a “down” position, cover- ered S2 subunits, leaving the space for the large-scale ing the S2 subunits in the stem (Figure 2A). The angle conformational change of the S2 subunit to expose and between the long axis of the CTD1 and its projection on insert the fusion peptide into the target cell membrane. the horizontal plane perpendicular to the 3-fold axis is ~19 degree (Figure 2A). In the conformations 2-4, two The “up” conformation of the CTD1 is also required for CTD1s still adopt the same “down” conformation as in the binding of neutralizing antibodies the conformation 1, whereas one CTD1 rotates outward Several neutralizing antibodies targeting the CTD1 to an “up” position and no longer covers the S2 subunit (m339, 80R and F26G19) have shown potent inhibition SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. Figure 2 Four different conformations of the SARS-CoV S glycoprotein trimer. Top: surface shadowed diagrams showing the four different conformations (conformations 1-4) of the S trimer. The CTD1s are colored pink. Bottom: ribbon diagrams showing S monomers with the semi-transparent CTD1 densities colored pink. The tilt angles of the CTD1s are defined by the angle between the long axis of the CTD1 (red cylinder) and its projection on the horizontal plane (grey ellipse). (A) Three-fold symmetric conformation 1 with all the three CTD1s in the “down” conformations. (B-D) Asymmetric conformations 2-4 with one CTD1 in the “up” conformation. activity against the cell infection of pseudo-typed or trimer is required not only for successful infections of the live SARS-CoV, and the antibody binding epitopes have SARS-CoV virions, by also for efficient neutralization been elucidated by crystal structure determination of the by antibodies targeting the CTD1. antibody-CTD1 complexes [29-31]. These antibodies binds to the receptor-binding site on the CTD1, thereby Discussion directly inhibiting the engagement of ACE2 receptor. Structural superimpositions showed that these antibodies As the known largest class I viral fusion protein, the all have steric clashes with other regions of the S glyco- coronavirus S glycoprotein trimer recognizes a variety protein trimer in the conformation 1 state (Figure 4A- of host cell receptors through the NTD or CTD1 of the 4C). In contrast, the S glycoprotein trimer conformations S1 subunit, and subsequently mediates the viral and cell 2-4 would allow the antibodies to bind the “up” CTD1 membrane fusion through the fusion peptide and two and inhibit its interaction with the ACE2 receptor (Figure heptad repeats of the S2 subunit. Since the discoveries of 4D-4F). These results indicated that the receptor-binding highly pathogenic SARS-CoV in 2002 and MERS-CoV inactive to active state transition of the S glycoprotein in 2012, most studies have been focused on the RBD www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer Figure 3 Models of the SARS-CoV S monomer and trimer bound with the receptor ACE2. (A) “Binding” of the receptor ACE2 (green) to one S monomer (pink) of the conformation 1 S trimer. The CTD1 is in the “down” conformation. (B) “Binding” of the receptor ACE2 (green) to the conformation 1 S trimer. Three CTD1s are all in the “down” conformations. The steric clashes between a neighboring CTD1 (grey) and ACE2 (green) are colored blue. (C) The same model as in A, with the S trimer den- sity map presented. Only the boundary profile of the “bound” ACE2 is shown (green lines) for a better view of the clashes (volume: 10 696 Å ). (D) “Binding” of the receptor ACE2 (green) with the conformation 3 S monomer (pink) of which the CTD1 is in the “up” conformation. (E) “Binding” of the receptor ACE2 (green) to the “up” CTD1 (pink) of the conformation 3 S trimer showing no steric clashes with any neighboring “down” CTD1 (grey). (F) The same model as in D with the S trimer 3D density map presented. All models are generated by superimposing the CTD1-ACE2 complex crystal structure onto the CTD1 of the corresponding SARS-CoV S monomer or trimer. The NTD models are not shown. (NTD or CTD1) of the S1 subunit because it binds to very limited. the host receptors and is also the main target of neutral- In the present study, we determined the structure of izing antibodies during infection. Our knowledge about SARS-CoV S glycoprotein trimer in four different con- the structures of complete SARS-CoV and MERS-CoV formational states. Recently reported cryo-EM structures spike trimers in pre-fusion and post-fusion states are still of the MHV and HKU1 S glycoprotein trimers are simi- SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. Figure 4 Structural superimpositions showing the “binding” of neutralization antibodies to the SARS-CoV S trimers. (A) Struc- tural superimposition of the CTD1-Fab m396 complex (PDB accession code: 2DD8) onto one CTD1 of the SARS-CoV S trimer in conformation 1, showing the “binding” of Fab m396 to the SARS-CoV S trimer. The EM densities of the S trimer are represented using shadowed surfaces in semi-transparent grey. (B-C) Similar structural superimpositions showing the “binding” of neutralization antibodies 80R (PDB accession code: 2GHW; B) and F26G19 (PDB accession code: 3BGF; C). The steric clashes with the “bound” Fab are colored blue and the corresponding volumes are shown in bracket. (D-F) Structural super- impositions of three CTD1-antibody complex structures with the SARS-CoV S trimer in conformation 3. No steric clashes be- tween the “bound” Fab and the S trimer. lar to the conformation 1 of the SARS-CoV S glycopro- coprotein trimers. The protein receptor of the HKU1 has tein trimer in the receptor-binding inactive state, in which not been identified and the NTD in the S1 subunit was all three CTD1s in the S1 subunit are in “down” posi- shown to bind O-acetylated sialic acids on host cells [32]. tions (Figure 2A). The SARS-CoV S glycoprotein trimer The RBD of MHV S glycoprotein is the NTD of the S1 conformations 2-4 in the receptor-binding active state subunit, and structural alignment revealed that the MHV (Figure 2B-2D), in which one CTD1 is in an “up” posi- protein receptor CEACAM1a binds to the NTD in the S tion, were not observed in the MHV and HKU1 S gly- glycoprotein trimer without steric clashes (Supplementary www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer information, Figure S6C) [33]. Therefore, the MHV and one CTD1 from “down” to “up” position would release HKU1 S glycoprotein trimer may not need the prereq- the steric clashes and enable the binding of one receptor uisite receptor-binding inactive to active state transition molecule (Figure 5). Whether the binding of the first re- for receptor binding. This intrinsic receptor-binding dif- ceptor molecule is enough, or the binding of the second ference among these coronavirus S glycoprotein trimers and third receptors is required to trigger conformational may be one of the reasons why the previous studies were changes in the S1 and S2 subunit necessary for mem- not able to reveal other conformations rather than confor- brane fusion will be an interesting question for future mation 1 in the MHV and HKU1 S glycoprotein trimers. studies. Nevertheless, the conformational change of CTD1 may still be needed for the subsequent exposure and confor- Materials and Methods mational changes of the S2 subunit for membrane fusion during the infection of MHV and HKU1. Protein expression and purification A human codon-optimized gene coding the SARS-CoV spike The protein receptor-binding site in the CTD1 has (S) glycoprotein ectodomain (NCBI Accession NP_828851.1) also been confirmed for MERS-CoV and bat corona- residues 1-1 195 with an R667A mutation to enhance sample ho- virus HKU4, which are lineage C betacoronaviruses mogeneity and a C-terminal strep tag for purification was cloned and have the same host receptor DPP4 [19, 20, 34-36]. and inserted into pFastBac-Dual vector (Invitrogen). The construct Structural superimposition of DPP4 complexed with the was transformed into bacterial DH10Bac competent cell and the CTD1 of MERS-CoV or HKU4 onto the SARS-CoV S extracted Bacmid was then transfected into Sf9 cells using Cell- glycoprotein trimer also showed that the “up” confor- fectin II Reagent (Invitrogen). The low-titer viruses were harvest- ed and then amplified to generate high-titer virus stock, which was mation of CTD1 is required for the binding of DPP4 by used to infect 2 L Sf9 cells at a density of 2 × 10 cells/ml. The these two betacoronaviruses (Supplementary informa- supernatant of cell culture containing the secreted SARS-CoV S tion, Figure S7). Therefore, we suggest that the recep- glycoprotein was harvested 60 h after infection, concentrated and tor-binding inactive to active state transition observed in buffer-exchanged to binding buffer (10 mM HEPES, pH 7.2, 500 the SARS-CoV spike would also occur in the spikes of mM NaCl). SARS-CoV S glycoprotein was captured by StrepTac- other betacoronaviruses that bind to their host receptors tin Sepharose High Performance (GE Healthcare) and eluted with through the CTD1. In receptor-binding inactive state, all 10 mM D-desthiobiotin in binding buffer. The eluted protein was then purified by gel filtration chromatography using the Superose three CTD1s in the “down” positions would not allow 6 column (GE Healthcare) pre-equilibrated with HBS buffer (10 efficient binding to receptors (Figure 5). The transition of Figure 5 A cartoon model showing the transition of the S trimer spikes from receptor-binding inactive to active and subse- quent fusogenic states. For betacoronavirus S glycoprotein utilizing the CTD1 as the receptor-binding domain, the state tran- sition of the S trimer with the “down” to “up” conformational change of CTD1 would allow receptor binding and may initiate subsequent conformational changes in the S2 subunits to mediate membrane fusion. SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. mM HEPES, 150 mM NaCl). Fractions containing SARS-CoV S 0.143 criterion. Local resolution variations were estimated using glycoprotein were pooled and concentrated for electron microsco- ResMap [42]. py analysis. Model building and structure refinement The crystal structure of the SARS-CoV S glycoprotein RBD Cryo-EM Aliquots of 3 µl purified SARS-CoV S glycoprotein were ap- (or CTD1, residues 324-502, PDB accession code: 3D0G) and the plied to glow-discharged holey carbon grids (Quantifoil, Cu 400 structures of HKU1 S glycoprotein (PDB accession code: 5i08), mesh, R1.2/1.3) or grids with a layer of continuous ultrathin car- CTD2 (residues 596-673), CTD3 (residues 674-771) and S2 do- bon film (with/without glow-discharge treatment) (Ted Pella, Inc.). main (residues 793-1147) were initially fitted into the SARS-CoV The protein concentration was ~0.34 mg/ml for holey carbon grids density map using UCSF Chimera [43]. Sequence alignment of and ~0.17 mg/ml for continuous carbon grids. The grids were blot- the SARS-CoV S and the fitted structures was performed using ted and then were plunged into liquid ethane using an FEI Vitro- DNAman (Lynnon Corporation, Quebec, Canada) and ClustalX bot. The grids were checked in an FEI F20 microscope operating at [44]. The fitted model was rebuilt using RosettaCM with C3 200 kV with a Gatan 895 4 k × 4 k CCD camera and a small data symmetry imposed and the best output model was selected ac- set was collected for generating the initial model. Data for final cording to the energy and fitness of the model to the EM density classification and refinement were collected on an FEI Titan Krios map [45]. The model of CTD2, CTD3 and S2 region was refined microscope operating at 300 kV using a K2 Summit camera (Gatan using RosettaRelax and PHENIX real-space refinement [46, 47]. Inc.) in super-resolution mode with a nominal magnification of 22 The crystal structure of the CTD1 domain was fitted into the EM 500× (yielding a calibrated pixel size of 1.32 Å). Each image was density and was refined as a rigid body in real space by using fractionated into 32 movie frames with a total exposure time of 8 s PHENIX. Then the two models were merged, manually adjusted and at a dose rate of ~8 counts per physical pixel per second (~4.7 in COOT and refined in PHENIX again with reference model re- electrons/Å /s). UCSFImage4 was used for all data collection [37]. strains, secondary structure restrains and geometry restrains [48]. Cross-validation of overfitting was performed following the pro- cedures described before [49, 50]. Briefly, the atom coordinates of Image processing The image processing procedures were summarized in Supple- the model (including CTD1, CTD2, CTD3 and S2) were randomly mentary information, Figure S2A. An initial map was generated displaced by 0.5 Å using PHENIX PDB tools. Then the displaced using the program EMAN2 with the images collected on the F20 model was refined against one of the two half maps in reciprocal microscope [38]. A total of 3 309 movie stacks were collected space by using PHENIX. FSC curves were calculated between the using the K2 camera. All movie frames were aligned using the refined model and the corresponding half map that was used for program motioncorr and the CTF parameters were determined the refinement (half1, FSC ) and between the refined model and work by CTFFIND4 [39, 40]. A total of 4 015 particles were manually the other half map (half2, FSC ) that was not used for the model test picked and 2D classifications were performed in RELION 1.3 [41]. refinement. No big gaps were observed between the work and test Four representative class averaged images were selected as refer- FSC curves (Supplementary information, Figure S1D), indicating ences for automatic particle picking with the whole data set using that the model was not overfitted. Molprobity was used to evaluate RELION 1.4. All 2D and 3D classifications and refinements were the final refined model. performed using RELION 1.4. Auto-picked particles were visu- Models of the three asymmetric conformations were built by ally inspected and then were selected by several rounds of refer- fitting the symmetric model of SARS-CoV S into the density maps ence-free 2D classifications. During the 2D classification, particles using UCSF Chimera. The fitted models were then refined as rigid from classes not showing clear secondary structure features were bodies in real space using PHENIX. For the subunit with the “up” deleted, finally yielding 210 129 selected particles. These selected CTD1, the subunit was split into two rigid bodies for refinement, particles were subjected to an initial 3D refinement with C3 sym- including the “up” CTD1 and the rest part of the subunit. metry imposed, followed by two runs of 3D classifications. A total Structural comparisons between different density maps and of 52 983 particles of the best class were selected and subjected to between density maps and models were performed using UCSF a 3D auto-refine. At the final stage of the refinement, 34 152 par - Chimera measure correlation, PHENIX get_cc_mtz_mtz and get_ ticles with a higher “loglikelicontribution” value were used. The cc_mtz_pdb. The cross correlation values were listed in Supple- resolution of the final C3 symmetric density map was 4.3 Å post mentary information, Figure S4A-S4D and Table S1B. processing in RELION. The SARS-CoV S RBD-receptor complex (PDB accession The initial 210 129 selected particles were also subjected to 3D code: 3R4D) and SARS-CoV S RBD-Fab complexes (PDB ac- refinements and classifications without any symmetry imposed. cession code: 2GHW, 2DD8 and 3BGF) were superimposed onto The central parts of the different classes were similar but most the models with the SARS-CoV S RBD structure as reference classes showed obvious asymmetric features with one of three using the “match” command in UCSF Chimera. To calculate the CTD1 domains protruding up. These asymmetric particles could “up” angles of the CTD1s (Figure 2), the horizontal plane of the be grouped into three major classes according to the tilt angle of S perpendicular to the 3-fold axis and the long axis of the CTD1 the “up” CTD1. Particles from these classes were separately sub- were generated using the UCSF Chimera “define” command and jected to 3D auto-refine. The resolutions of the final density maps then the angle between the axis and the plane was calculated using were 7.3, 5.7 and 6.8 Å for the three classes. the UCSF Chimera “angle” command. The overlap volumes be- The handedness of the density maps was verified by docking tween the fitted receptor and the S spike were also calculated using the crystal structure of SARS-CoV S glycoprotein CTD1 into the UCSF Chimera. Two atoms were considered to be overlapped if maps. Reported resolutions are based on the gold-standard FSC their overlap score > 0. The overlap score is defined as the sum of www.cell-research.com | Cell Research | SPRINGER NATURE Cryo-EM structures of the SARS-CoV spike glycoprotein trimer two VDW (van der Waals) radii minus the distance between them 7 Du L, He Y, Zhou Y, Liu S, Zheng BJ, Jiang S. The spike and minus an allowance (0.4 Å) for potentially hydrogen-bonded protein of SARS-CoV--a target for vaccine and therapeutic pairs. All the overlap atoms were used for the generation of the development. Nat Rev Microbiol 2009; 7:226-236. steric clash area. The volume of the steric clashes was calculated 8 de Wit E, van Doremalen N, Falzarano D, Munster VJ. SARS using UCSF Chimera “measure volume” command. All figures and MERS: recent insights into emerging coronaviruses. Nat were generated with UCSF Chimera and Pymol (The PyMOL Mo- Rev Microbiol 2016; 14:523-534. lecular Graphics System, Version 1.7.0.5 Schrödinger, LLC.). 9 Chan JF, Lau SK, To KK, Cheng VC, Woo PC, Yuen KY. Middle East respiratory syndrome coronavirus: another zoo- notic betacoronavirus causing SARS-like disease. Clin Micro- Accession numbers The cryo-EM maps and related materials have been deposited biol Rev 2015; 28:465-522. to the EM Data Bank under accession codes EMD-6679, EMD- 10 Vijay R, Perlman S. Middle East respiratory syndrome and 6680, EMD-6681 and EMD-6682. The atomic coordinate has been severe acute respiratory syndrome. Curr Opin Virol 2016; deposited to the Protein Data Bank under accession code 5WRG. 16:70-76. 11 Millet JK, Whittaker GR. Host cell proteases: critical deter- minants of coronavirus tropism and pathogenesis. Virus Res Acknowledgments 2015; 202:120-134. 12 Li F. Receptor recognition mechanisms of coronaviruses: a We thank the Cryo-EM and Computing Platforms of the decade of structural studies. J Virol 2015; 89:1954-1964. Tsinghua University Branch of the National Center for Protein 13 Li F. Structure, function, and evolution of coronavirus spike Sciences (Beijing) for facility support. This work was supported proteins. Annu Rev Virol 2016; 3:237-261. by the National Key Plan for Scientific Research and Development 14 Lu G, Wang Q, Gao GF. Bat-to-human: spike features deter- of China (2016YFD0500307 and 2016YFA0501100), the National mining ‘host jump’ of coronaviruses SARS-CoV, MERS-CoV, Natural Science Foundation of China (81471929, U1405228, and beyond. Trends Microbiol 2015; 23:468-478. 31470721 and 81550001) and the Beijing Advanced Innovation 15 Graham RL, Baric RS. Recombination, reservoirs, and the Center for Structural Biology. modular spike: mechanisms of coronavirus cross-species transmission. J Virol 2010; 84:3134-3146. Author Contributions 16 Beniac DR, Andonov A, Grudeski E, Booth TF. Architecture MG, WS, YX and XW designed the experiments. WS designed of the SARS coronavirus prefusion spike. Nat Struct Mol Biol and cloned the protein constructs. WS and HZ carried out protein 2006; 13:751-752. expression, purification and biophysical characterization. MG, JX 17 Hofmann H, Pohlmann S. Cellular entry of the SARS corona- and SC performed cryo-EM sample preparation and data collec- virus. Trends Microbiol 2004; 12:466-472. tion. MG processed the cryo-EM data and built the atomic model. 18 Li F, Li W, Farzan M, Harrison SC. Structure of SARS coro- MG, WS, YX and XW analyzed the data. MG, WS, YX and XW navirus spike receptor-binding domain complexed with recep- wrote the manuscript with input from all authors. tor. Science 2005; 309:1864-1868. 19 Lu G, Hu Y, Wang Q, et al. Molecular basis of binding be- Competing Financial Interests tween novel human coronavirus MERS-CoV and its receptor The authors declare no competing financial interests. CD26. Nature 2013; 500:227-231. 20 Wang N, Shi X, Jiang L, et al. Structure of MERS-CoV spike References receptor-binding domain complexed with human receptor DPP4. Cell Res 2013; 23:986-993. 21 Gao J, Lu G, Qi J, et al. Structure of the fusion core and inhi- 1 Su S, Wong G, Shi W, et al. Epidemiology, genetic recombi- bition of fusion by a heptad repeat peptide derived from the nation, and pathogenesis of coronaviruses. Trends Microbiol S protein of Middle East respiratory syndrome coronavirus. J 2016; 24:490-502. Virol 2013; 87:13134-13140. 2 Ksiazek TG, Erdman D, Goldsmith CS, et al. A novel coro- 22 Lu L, Liu Q, Zhu Y, et al. Structure-based discovery of Mid- navirus associated with severe acute respiratory syndrome. N dle East respiratory syndrome coronavirus fusion inhibitor. Engl J Med 2003; 348:1953-1966. Nat Commun 2014; 5:3067. 3 Drosten C, Gunther S, Preiser W, et al. Identification of a 23 Supekar VM, Bruckmann C, Ingallinella P, Bianchi E, Pessi novel coronavirus in patients with severe acute respiratory A, Carfi A. Structure of a proteolytically resistant core from syndrome. N Engl J Med 2003; 348:1967-1976. the severe acute respiratory syndrome coronavirus S2 fusion 4 Peiris JS, Lai ST, Poon LL, et al. Coronavirus as a possible protein. Proc Natl Acad Sci USA 2004; 101:17958-17963. cause of severe acute respiratory syndrome. Lancet 2003; 24 Walls AC, Tortorici MA, Bosch BJ, et al. Cryo-electron mi- 361:1319-1325. croscopy structure of a coronavirus spike glycoprotein trimer. 5 Zaki AM, van Boheemen S, Bestebroer TM, Osterhaus Nature 2016; 531:114-117. AD, Fouchier RA. Isolation of a novel coronavirus from a 25 Kirchdoerfer RN, Cottrell CA, Wang N, et al. Pre-fusion man with pneumonia in Saudi Arabia. N Engl J Med 2012; structure of a human coronavirus spike protein. Nature 2016; 367:1814-1820. 531:118-121. 6 de Groot RJ, Baker SC, Baric RS, et al. Middle East respira- 26 Walls AC, Tortorici MA, Frenz B, et al. Glycan shield and tory syndrome coronavirus (MERS-CoV): announcement of epitope masking of a coronavirus spike protein observed by the Coronavirus Study Group. J Virol 2013; 87:7790-7792. SPRINGER NATURE | Cell Research | Vol 27 No 1 | January 2017 Miao Gui et al. cryo-electron microscopy. Nat Struct Mol Biol 2016; 23:899- images by UCSFImage. J Struct Biol 2015; 192:174-178. 905. 38 Tang G, Peng L, Baldwin PR, et al. EMAN2: an extensible 27 Belouzard S, Chu VC, Whittaker GR. Activation of the SARS image processing suite for electron microscopy. J Struct Biol coronavirus spike protein via sequential proteolytic cleavage 2007; 157:38-46. at two distinct sites. Proc Natl Acad Sci USA 2009; 106:5871- 39 Li X, Mooney P, Zheng S, et al. Electron counting and 5876. beam-induced motion correction enable near-atomic-resolu- 28 Li W, Moore MJ, Vasilieva N, et al. Angiotensin-converting tion single-particle cryo-EM. Nat Methods 2013; 10:584-590. enzyme 2 is a functional receptor for the SARS coronavirus. 40 Rohou A, Grigorieff N. CTFFIND4: fast and accurate defocus Nature 2003; 426:450-454. estimation from electron micrographs. J Struct Biol 2015; 29 Hwang WC, Lin Y, Santelli E, et al. Structural basis of neu- 192:216-221. tralization by a human anti-severe acute respiratory syndrome 41 Scheres SH. RELION: implementation of a Bayesian ap- spike protein antibody, 80R. J Biol Chem 2006; 281:34610- proach to cryo-EM structure determination. J Struct Biol 34616. 2012; 180:519-530. 30 Prabakaran P, Gan J, Feng Y, et al. Structure of severe acute 42 Kucukelbir A, Sigworth FJ, Tagare HD. Quantifying the lo- respiratory syndrome coronavirus receptor-binding domain cal resolution of cryo-EM density maps. Nat Methods 2014; complexed with neutralizing antibody. J Biol Chem 2006; 11:63-65. 281:15829-15836. 43 Pettersen EF, Goddard TD, Huang CC, et al. UCSF Chimera 31 Pak JE, Sharon C, Satkunarajah M, et al. Structural insights — a visualization system for exploratory research and analy- into immune recognition of the severe acute respiratory syn- sis. J Comput Chem 2004; 25:1605-1612. drome coronavirus S protein receptor binding domain. J Mol 44 Larkin MA, Blackshields G, Brown NP, et al. Clustal W and Biol 2009; 388:815-823. Clustal X version 2.0. Bioinformatics 2007; 23:2947-2948. 32 Huang X, Dong W, Milewska A, et al. Human coronavirus 45 Song Y, DiMaio F, Wang RY, et al. High-resolution compar- HKU1 spike protein uses O-acetylated sialic acid as an attach- ative modeling with RosettaCM. Structure 2013; 21:1735- ment receptor determinant and employs hemagglutinin-es- 1742. terase protein as a receptor-destroying enzyme. J Virol 2015; 46 DiMaio F, Tyka MD, Baker ML, Chiu W, Baker D. Refine- 89:7202-7213. ment of protein structures into low-resolution density maps 33 Peng G, Sun D, Rajashankar KR, Qian Z, Holmes KV, Li F. using rosetta. J Mol Biol 2009; 392:181-190. Crystal structure of mouse coronavirus receptor-binding do- 47 Adams PD, Afonine PV, Bunkoczi G, et al. PHENIX: a com- main complexed with its murine receptor. Proc Natl Acad Sci prehensive Python-based system for macromolecular structure USA 2011; 108:10696-10701. solution. Acta Crystallogr D Biol Crystallogr 2010; 66:213- 34 Wang Q, Qi J, Yuan Y, et al. Bat origins of MERS-CoV sup- 221. ported by bat coronavirus HKU4 usage of human receptor 48 Emsley P, Lohkamp B, Scott WG, Cowtan K. Features and CD26. Cell Host Microbe 2014; 16:328-337. development of Coot. Acta Crystallogr D Biol Crystallogr 35 Raj VS, Mou H, Smits SL, et al. Dipeptidyl peptidase 4 2010; 66:486-501. is a functional receptor for the emerging human coronavi- 49 Brown A, Long F, Nicholls RA, Toots J, Emsley P, Mur- rus-EMC. Nature 2013; 495:251-254. shudov G. Tools for macromolecular model building and re- 36 Yang Y, Du L, Liu C, et al. Receptor usage and cell entry finement into electron cryo-microscopy reconstructions. Acta of bat coronavirus HKU4 provide insight into bat-to-human Crystallogr D Biol Crystallogr 2015; 71:136-153. transmission of MERS coronavirus. Proc Natl Acad Sci USA 50 Amunts A, Brown A, Bai XC, et al. Structure of the yeast mi- 2014; 111:12516-12521. tochondrial large ribosomal subunit. Science 2014; 343:1485- 37 Li X, Zheng S, Agard DA, Cheng Y. Asynchronous data ac- 1489. quisition and on-the-fly analysis of dose fractionated cryoEM (Supplementary information is linked to the online version of the paper on the Cell Research website.) www.cell-research.com | Cell Research | SPRINGER NATURE

Journal

Cell Research – Pubmed Central

Published: Dec 23, 2016

There are no references for this article.