key: cord-273891-7w334xgt
authors: Kirchdoerfer, Robert N.; Wang, Nianshuang; Pallesen, Jesper; Wrapp, Daniel; Turner, Hannah L.; Cottrell, Christopher A.; Corbett, Kizzmekia S.; Graham, Barney S.; McLellan, Jason S.; Ward, Andrew B.
title: Receptor binding and proteolysis do not induce large conformational changes in the SARS-CoV spike
date: 2018-03-31
journal: bioRxiv
DOI: 10.1101/292672
sha: 
doc_id: 273891
cord_uid: 7w334xgt

Severe acute respiratory syndrome coronavirus (SARS-CoV) emerged in 2002 as a highly transmissible pathogenic human betacoronavirus. The viral spike glycoprotein (S) utilizes angiotensin-converting enzyme 2 (ACE2) as a host protein receptor and mediates fusion of the viral and host membranes, making S essential to viral entry into host cells and host species tropism. As SARS-CoV enters host cells, the viral S undergoes two proteolytic cleavages at S1/S2 and S2’ sites necessary for efficient membrane fusion. Here, we present a cryo-EM analysis of the trimeric SARS-CoV S interactions with ACE2 and of the trypsin-cleaved S. Surprisingly, neither binding to ACE2 nor cleavage by trypsin at the S1/S2 cleavage site impart large conformational changes within S or expose the secondary cleavage site, S2’. These observations suggest that S2’ cleavage does not occur in the S prefusion conformation and that additional triggers may be required.

viral and host membranes, making S essential to viral entry into host cells and host species 23 tropism. As SARS-CoV enters host cells, the viral S undergoes two proteolytic cleavages at 24 S1/S2 and S2ʹ′ sites necessary for efficient membrane fusion. Here, we present a cryo-EM 25 analysis of the trimeric SARS-CoV S interactions with ACE2 and of the trypsin-cleaved S. 26 Surprisingly, neither binding to ACE2 nor cleavage by trypsin at the S1/S2 cleavage site impart 27 large conformational changes within S or expose the secondary cleavage site, S2´. These 28 observations suggest that S2´ cleavage does not occur in the S prefusion conformation and that 29 additional triggers may be required. 30 3 Severe acute respiratory syndrome coronavirus (SARS-CoV) emerged in humans in 2002 31 and rapidly spread globally causing 8,096 cases and 774 associated deaths in 26 countries 32 through July 2003 1 . SARS-CoV reappeared in a second smaller outbreak in 2004, but has since 33 disappeared from human circulation. However, closely related coronaviruses, such as WIV1, 34 currently circulate in bat reservoirs and are capable of utilizing human receptors to enter cells 2 . 35 The more recent emergence of Middle East respiratory syndrome coronavirus (MERS-CoV) 1 and 36 the likelihood of future zoonotic transmission of novel coronaviruses to humans from animal 37 reservoirs make understanding the coronavirus infection cycle of great importance to human 38 health. 39 Coronaviruses are enveloped viruses possessing large, trimeric spike glycoproteins (S) 40 required for the recognition of host receptors for many coronaviruses as well as the fusion of 41 viral and host cell membranes for viral entry into cells 3 . During viral egress from infected host 42 cells, some coronavirus S proteins are cleaved into S1 and S2 subunits. The S1 subunit is 43 responsible for host-receptor binding while the S2 subunit contains the membrane-fusion 44 machinery. During viral entry, the S1 subunit binds host receptors in an interaction thought to 45 expose a secondary cleavage site within S2 (S2´) adjacent to the fusion peptide for cleavage by 46 host proteases 4-7 . This S2´ proteolysis has been hypothesized to facilitate insertion of the fusion 47 peptide into host membranes after the first heptad repeat region (HR1) of the S2 subunit 48 rearranges into an extended α-helix 8-10 . Subsequent conformational changes in the second heptad 49 repeat region (HR2) of S2 form a six-helix bundle with HR1, fusing the viral and host 50 membranes and allowing for release of the viral genome into host cells. Coronavirus S is also the 51 target of neutralizing antibodies 11 , making an understanding of S structure and conformational 52 transitions pertinent for investigating S antigenic surfaces and designing vaccines.

The SARS-CoV S1 subunit is composed of two distinct domains: an N-terminal domain 54 (S1 NTD) and a receptor-binding domain (S1 RBD) also referred to as the S1 CTD or domain B. 55 Each of these domains have been implicated in binding to host receptors, depending on the 56 coronavirus in question. However, most coronaviruses are not known to utilize both the S1 NTD 57 and S1 RBD for viral entry 12 . SARS-CoV makes use of its S1 RBD to bind to the human 58 angiotensin-converting enzyme 2 (ACE2) as its host receptor 13,14 . 59 Recent examination using cryo-electron microscopy (cryo-EM) has illuminated the 60 prefusion structures of coronavirus spikes [15] [16] [17] [18] [19] [20] [21] [22] . Initial examination of HCoV-HKU1 S showed 61 that the receptor-binding site on the S1 RBD was occluded when the RBD was in a 'down' 62 conformation and it was hypothesized that conformational changes were required to access this 63 site 16 . Subsequent studies of the highly pathogenic human coronavirus S proteins of SARS-64 CoV 15,22 and MERS-CoV 17,22 showed that these viral S1 RBD do indeed sample an 'up' 65 conformation where the receptor-binding site is accessible. These structural studies also located 66 the positions of the S1/S2 and S2´ cleavage sites on the prefusion spike. The S1/S2 site lies 67 within a surface exposed loop in the second subdomain of S1 16 . However, the S2´ site lies closer 68 to base of the spike and though this region is located on the surface of the spike, cleavage at this 69 site is prevented by surrounding protein elements 17 . 70 To examine the hypothesized conformational transitions induced by proteolysis and 71 receptor binding, we used single-particle cryo-EM to determine structures of S in uncleaved, 72 S1/S2 cleaved and ACE2-bound states. Three-dimensional classification of the S1 RBD 73 positions and corresponding atomic protein models revealed that neither ACE2-binding nor 74 trypsin cleavage at the S1/S2 boundary induced substantial conformational changes in the CoV may use a distinct mechanism of FP2 membrane insertion. 110 As observed in the previous SARS-CoV and MERS-CoV S structures 15,22 , the trimeric S 111 adopts two distinct conformations related to each of the S1 RBD. The 'down' conformation caps 112 the S2 helices and makes extensive contacts with the S1 NTD. The 'up' conformation of the S1 113 RBD exposes the S1 RBD receptor-binding site. It has been previously reported that for wild-114 type SARS-CoV S, 56% of the particles contained three 'down' RBD conformations while 44% 115 contained a single 'up' S1 RBD conformation 22 . To examine the conformation of the S1 RBD 116 among our SARS-CoV S 2P ectodomains, we used a local masking and 3-D sorting strategy 17 to 117 more accurately classify the conformations as being either 'down' or 'up' at each of the three S1 118 RBD positions within the trimer. This analysis revealed that the majority of the S 2P proteins 119 were in the single-'up' conformation (58%) with lesser amounts of double-and triple-'up' 120 conformations (39% and 3% respectively) and with no all-'down' conformation observed. The 7 increased propensity to adopt the 'up' S1 RBD conformation may indicate a difference in the 122 coronavirus S containing the 2P mutations, however other differences in sample preparation 123 cannot be ruled out. 124 125 ACE2 and S1 C-terminal domains 126 To examine the structure of SARS-CoV S bound to its receptor, ACE2, we combined 127 SARS-CoV S 2P ectodomain with an excess of soluble human ACE2 with subsequent 128 purification by size-exclusion chromatography and immediate cryo-EM specimen preparation. 129 Initial sorting of particle heterogeneity indicated spikes could be split into ACE2-bound (45%) 130 and unbound (55%) classes. Using a similar masking and 3-D sorting strategy as above we sorted 131 the unbound S class further into classes with S1 conformations of one or two 'up' S1 RBDs (Fig   132   3 and Supplementary Tables 2-4 and Supplementary Fig. 2-4) . We did not observe an all-'down' 133 class nor a three 'up' S1 RBD class indicating a low prevalence of these conformations among 134 the unbound spikes. Expanding our 3D sorting strategy, we classified our ACE2-bound particles 135 at each S1 RBD position and identified single, double and triple ACE2-bound S. We were 136 further able to identify S1 RBD conformations at the non-ACE2 occupied RBD positions to 137 represent each population of S1 RBD conformations among ACE2-bound S. 138 As hypothesized by previous structural work 15-17,22 , the S1 RBD recognizes ACE2 with 139 an 'up' S1 RBD conformation. The proportion of total 'up' S1 RBD conformations within the 140 ACE2-bound and -unbound classes is nearly identical within this dataset (58% 'up' S1 RBD), 141 similar to the proportion of total 'up' S1 RBD in the SARS S 2P ectodomain dataset (48%). This 142 strongly suggests that binding of a single ACE2 receptor does not induce adjacent S1 RBDs to 143 transition from a 'down' to 'up' conformation. Hence, ACE2 is more likely to bind to an already 144 8 'up' S1 RBD rather than inducing the conformational changes that are required for the S1 RBD 145 to become accessible to ACE2. 146 It is noteworthy that despite prolonged co-incubation and an excess of ACE2, we had 147 difficulties in saturating the S1 RBD with ACE2 in the context of trimeric S ectodomain. This 148 poor saturation is illustrated by the small proportion of triple-bound ACE2 and the majority of 149 spikes that are unbound by receptor. In contrast, isolated recombinant S1 RBD easily binds 150 ACE2 and is capable saturating ACE2 on target cells to block S-mediated entry 14 . Our observed 151 sub-stoichiometric ACE2 binding to trimeric spikes is consistent with the difficulty in using 152 soluble ACE2 receptor to neutralize SARS-CoV S pseudotyped onto VSV 25 . The reduced 153 binding of ACE2 to trimeric spikes is likely due to the incomplete exposure and conformational 154 flexibility of the S1 RBD. Incomplete neutralization with soluble receptor was not encountered 155 for MHV which binds CEACAM1a via its S1 NTD, which does not undergo conformational 156 changes 19,26 . 157 Similar to recently published MERS-CoV S structures 17 , the ACE2-bound RBD adopts a 158 much more extended and rotated conformation compared to S1 RBD modeled in previous 159 SARS-CoV S structures 22 . This difference is likely due to poor density in the hinge regions 160 between the S1 RBD and subdomain 1 (SD-1) in these previous reconstructions 15,22 rather than 161 the presentation of a unique receptor-bound conformation. Indeed, the bound ACE2 receptor and 162 S1 RBD for all reconstructions here show poorer density quality than the less mobile regions of 163 the SARS-CoV S (Fig 4) . To improve the density for ACE2-bound S1 RBD, we used focused 164 refinement on this region to overcome the flexibility of these domains relative to the rest of S. 165 This yielded a 7.9 Å resolution reconstruction with improved local density quality (Fig 4b and c) . 166 We successfully placed the crystal structure of the SARS-CoV S1 RBD bound to ACE2 167 9 (2AJF.pdb 23 ) into this density as a rigid body indicating that the previously determined crystal 168 structure accurately recapitulates the conformation between the ACE2-bound S1 RBD in the 169 trimeric spike. 170 The ACE2-bound, S1 RBD extends upwards and rotates away from contacts with nearby 171 amino acids. Hence, any conformational changes induced by receptor binding to the S1 RBD are 172 more likely to be caused by the absence of the S1 RBD contacts in the 'up' conformation, rather 173 than the formation of additional contacts (Supplemental Figure 5) . This model provides a flexible 174 mechanism for how different coronavirus spikes can bind to different protein receptors with their 175 S1 RBD and facilitate fusion with host cells. Moreover, movements of the S1 RBD to the 'up' Fig. 6 ). Nearing the end of the time course additional lower molecular weight 214 bands are observed which we interpret to be degradation of the S1 subunit. Regardless of which 215 construct was used or whether ACE2 was bound to the S ectodomains, there is no prominent 216 band that corresponds to a S2ʹ′ cleavage product (approximately 52 kDa). 217 To analyze the cleavage products in detail, we performed cryo-EM analysis on the 218 trypsin-cleaved SARS-CoV S 2P ectodomain. Using all-particles and C3 symmetry yielded a 219 reconstruction at 3.3 Å resolution (Fig. 5, Supplementary Tables 1 and 4 and Supplementary Fig.   220 7). The short loop containing the S1/S2 cleavage site is disordered in the uncleaved spike 221 reconstruction and remains disordered in the trypsin cleaved reconstruction. Moreover, 222 examination of the structure models indicates no significant differences between the trypsin-223 cleaved and uncleaved SARS-CoV S (Fig. 5b) . Fine sorting of S1 RBD positions of the trypsin-224 cleaved S reveals a very similar distribution of 'up' S1 RBD conformations available for receptor 225 binding as in the uncleaved samples, although we additionally observe a small proportion of S1 226 RBD in the all-'down' conformation (Fig. 5c) . These results indicate that trypsin-cleavage at 227 S1/S2 does not impart large conformational changes on the SARS-CoV S and justifies the 228 removal of S1/S2 cleavage sites for the production of more homogeneous material as vaccine 229 immunogens. This suggests that although cleavage at S1/S2 may remove an obstacle for 230 conformational changes leading to fusion, S1/S2 cleavage alone does not produce significant 231 conformational changes. terminal helix of S2 HR1 (Fig 6) . Exposure of this site for cleavage may require remodeling of 237 this penultimate loop or HR1 beyond the conformation observed in the prefusion state. We 238 hypothesize that additional triggers beyond cleavage at the S1/S2 site or protein-receptor binding 239 are needed to transition the spike from its prefusion state to a yet to be observed intermediate. changes and that the S2′ proteolysis does not occur in the S prefusion state (Fig. 7) . This Grids were loaded onto a Titan Krios and data was collected using Leginon 32 at a total dose of 65 302 e -/Å 2 . Frames were aligned with MotionCor2 (UCSF) 33 implemented in the Appion workflow 34 . 303 Particles were selected using DoG picker 35 . Images were assessed and particle picks were 304 masked using EM Hole Punch 36 . The CTF for each image was estimated using Gctf 37 . 305 Electron microscopy data processing 306 Initial particle stacks were cleaned using multiple rounds of 2D classification in 307 RELION 38 . Good particles were selected as resembling prefusion coronavirus spikes. For the 308 SARS S 2P and trypsin-treated SARS S 2P, all particles from the clean stacks were used for 309 reconstruction with C3 symmetry. All datasets were extensively sorted using 3D classification to 310 examine heterogeneity in the S1 RBDs as described previously 17 . Briefly, 3D masks were 311 defined to encompass the possible heterogeneity at each S1 RBD position. The density within 312 these masks was then removed from unfiltered, unsharpened reconstructions. We then used 313 relion_project with image subtraction to create a particle stack containing only the signal arising 314 from the masked density. Finally, we used focused 3D classification to identify compositional 315 and conformational states at each S1 RBD position. All 3D reconstructions were produced with 316 RELION 38 and final refinements were performed with a six-pixel soft-edge solvent mask. Post-317 processing was applied to each reconstruction to apply B-factor sharpening and amplitude 318 corrections as well as to calculate local resolution maps. 319 Coordinate models were built for several of the high-resolution reconstructions using 320 5I08.pdb 16 , 2AJF.pdb 23 and 5X4S.pdb 22 as template models with reference to a recently 

SARS and MERS: recent 341 insights into emerging coronaviruses

Proceedings of 344 the National Academy of Sciences of the United States of America

Mechanisms of coronavirus cell 347 entry mediated by the viral spike protein

Two-step conformational changes in a coronavirus envelope 350 glycoprotein mediated by receptor binding and proteolysis

Receptor-bound porcine epidemic diarrhea virus spike 353 protein cleaved by trypsin induces membrane fusion

Proteolytic processing of Middle East respiratory syndrome coronavirus 356 spikes expands virus tropism

Inhibitors of cathepsin L prevent severe acute respiratory syndrome 359 coronavirus entry

The coronavirus spike protein is 362 a class I virus fusion protein: structural and functional characterization of the fusion core 363 complex

Structure of influenza 365 haemagglutinin at the pH of membrane fusion

Tectonic conformational changes of a coronavirus spike glycoprotein 368 promote membrane fusion

SARS Immunity and Vaccination

Recombination, reservoirs, and the modular spike: mechanisms 373 of coronavirus cross-species transmission

Angiotensin-converting enzyme 2 is a functional receptor for the SARS 376 coronavirus

A 193-amino acid fragment of the 378 SARS coronavirus S protein efficiently binds angiotensin-converting enzyme 2. The 379

Cryo-electron microscopy structures of the SARS-CoV spike glycoprotein 381 reveal a prerequisite conformational state for receptor binding

Pre-fusion structure of a human coronavirus spike protein

Immunogenicity and structures of a rationally designed prefusion MERS-386

Proceedings of the National Academy of Sciences of the United States

Cryo-EM structure of porcine delta coronavirus spike protein in the pre-389 fusion state

Cryo-electron microscopy structure of a coronavirus spike glycoprotein 391 trimer

Glycan shield and epitope masking of a coronavirus spike protein 393 observed by cryo-electron microscopy

Glycan shield and fusion activation of a deltacoronavirus spike glycoprotein 396 fine-tuned for enteric infections

Cryo-EM structures of MERS-CoV and SARS-CoV spike glycoproteins 398 reveal the dynamic receptor binding domains

Structure of SARS coronavirus spike receptor-401 binding domain complexed with receptor

Peptide Forms an Extended Bipartite Fusion Platform that Perturbs Membrane Order in a 405 Calcium-Dependent Manner

Vesicular stomatitis virus pseudotyped with severe acute respiratory 408 syndrome coronavirus spike protein

N-terminal domain of the murine coronavirus 411 receptor CEACAM1 is responsible for fusogenic activation and conformational changes 412 of the spike protein

Activation of the SARS coronavirus spike 414 protein via sequential proteolytic cleavage at two distinct sites

Protease-mediated 418 enhancement of severe acute respiratory syndrome coronavirus infection

Physiological and molecular triggers for SARS-CoV 422 membrane fusion and entry into host cells

Host cell proteases: Critical determinants of coronavirus 425 tropism and pathogenesis

Discovery of a rich gene pool of bat SARS-related coronaviruses provides new 428 insights into the origin of SARS coronavirus

Automated molecular microscopy: the new Leginon system

MotionCor2: anisotropic correction of beam-induced motion for improved 433 cryo-electron microscopy

Appion: an integrated, database-driven pipeline to facilitate EM image 435 processing

DoG Picker 437 and TiltPicker: software tools to facilitate particle selection in single particle electron 438 microscopy

EMHP: an accurate automated hole 440 masking algorithm for single-particle cryo-EM image processing

Real-time CTF determination and correction

Accelerated cryo-EM structure 445 determination with parallelisation using GPUs in RELION-2

Coot: model-building tools for molecular graphics

Atomic-accuracy models from 4.5-A cryo-electron microscopy data with 451 density-guided iterative local refinement

PHENIX: a comprehensive Python-based system for macromolecular 454 structure solution

Computational resources for electron microscopy 461 at The Scripps Research Institute are supported by NIH grant OD021634

466 processed electron microscopy data. R.N.K and C.A.C. built and refined atomic models

 457 We gratefully acknowledge Travis Nieusma, Charles Bowman, Jean-Christophe Ducom and Bill 458 Anderson for microscopy and computational support. We also thank Lauren Holden for a critical 459 reading of this manuscript. This work was supported by grants from NIH/NIAID to A.B.W and 460