Significance
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a severe threat to public health and the global economy. Its spike protein is responsible for the membrane fusion and is thus a major target for vaccine and drug development. Our study presents the in situ structure of the spike protein in the postfusion state with higher resolution, giving further insights into the design of a viral entry inhibitor. Our observation of the oligomerization states of spikes on the viral membrane implies a possible mechanism of membrane fusion for viral infection.
Abstract
The spike protein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) mediates membrane fusion to allow entry of the viral genome into host cells. To understand its detailed entry mechanism and develop a specific entry inhibitor, in situ structural information on the SARS-CoV-2 spike protein in different states is urgent. Here, by using cryo-electron tomography, we observed both prefusion and postfusion spikes in β-propiolactone–inactivated SARS-CoV-2 virions and solved the in situ structure of the postfusion spike at nanometer resolution. Compared to previous reports, the six-helix bundle fusion core, the glycosylation sites, and the location of the transmembrane domain were clearly resolved. We observed oligomerization patterns of the spikes on the viral membrane, likely suggesting a mechanism of fusion pore formation.
Over the past two decades, several zoonotic coronavirus (CoV) diseases have emerged and posed a devastating threat to global public health and the economy, such as severe acute respiratory syndrome (SARS) (1), Middle East respiratory syndrome (MERS) (2), and COVID-19 (3). As of this writing, COVID-19 has more than 229 million confirmed cases and has caused 4.7 million deaths worldwide, with rapidly increasing numbers. This pneumonia epidemic was caused by a novel coronavirus named SARS coronavirus 2 (SARS-CoV-2), a β-coronavirus, with a genomic sequence that is closely related to SARS-CoV. SARS-CoV-2 is an enveloped, positive-sense single-stranded RNA virus with an ∼30-kb genome (4). Given the current pandemic situation, understanding the structure of SARS-CoV-2 as well as its infection process is very important for vaccine development and drug discovery.
The SARS-CoV-2 genome encodes three viral surface proteins: the spike (S) glycoprotein, envelope (E) protein, and membrane (M) protein. During the infection process, the trimeric S glycoprotein is cleaved by host proteases (4, 5) to produce two functional subunits: The N-terminal S1 subunit is responsible for receptor recognition, and the C-terminal S2 subunit is responsible for membrane fusion (6). Mediated by receptor binding and proteolytic activation, the S1 subunit falls off, and the S2 subunit undergoes extensive and irreversible conformational changes to insert its hydrophobic fusion peptide (FP) into the target cell membrane. Subsequently, two heptad repeat regions of the S2 subunit, heptad repeat 1 (HR1) and heptad repeat 2 (HR2), form a stable six-helix bundle (6-HB) fusion core to bring together the viral and cellular membranes, leading to colocalization of the FP and the transmembrane (TM) region at the same end to form the fusion pore (7). Thus, the S protein is one of the major targets for developing vaccines and antiviral drugs.
After the outbreak of COVID-19, the in vitro structures of SARS-CoV-2 S in the prefusion state were promptly solved using single-particle cryo-electron microscopy (cryo-EM) (8, 9) and X-ray crystallography (7, 10, 11). Soon afterward, the in situ structures of S in the prefusion state were revealed by cryo-electron tomography (cryo-ET) and cryo-subtomogram averaging (cryo-STA) (12⇓–14), uncovering the distribution of different conformational states as well as the native glycosylation sites. However, how the S protein is activated to induce membrane fusion with its host is less understood. The structure of S in the postfusion state would provide an important clue to investigate the fusion mechanism. The high-resolution structure of recombinant S in the postfusion state has been reported by Cai et al. (15), but this in vitro study failed to determine how the postfusion S proteins organize on the membrane. Previous in situ studies (12, 13, 16) explored this question but yielded limited information, due to the poor quality of the density map. In addition, we previously showed that the recombinant 6-HB fusion core of S in the postfusion state would be an effective target for the design of viral fusion inhibitors (7), which needs to be further validated by a higher-resolution structure and glycosylation information of in situ S in the postfusion state.
In the present work, we utilized cryo-ET and cryo-STA to study the structure of SARS-CoV-2 viruses that were inactivated by β-propiolactone (BPL). We solved the in situ structures of S in both the prefusion and postfusion states with resolutions of 12.9 and 10.9 Å, respectively. In addition to visualizing the TM region and glycosylation sites, we found that our previous crystal structure of the recombinant 6-HB fusion core fits well to the density map. In addition, we observed oligomerization of postfusion Ss on the viral membrane, suggesting a mechanism of S-induced membrane fusion. Our study will facilitate a better understanding of the SARS-CoV-2 fusion mechanism and be beneficial for viral entry inhibitor development.
Results
Cryo-ET Analysis of the Inactivated SARS-CoV-2 Virus.
We propagated SARS-CoV-2 virions into Vero cells and purified the viral particles in a biosafety level 3 (BSL-3) laboratory. The purified virus was inactivated with BPL and imaged by cryo-ET in a BSL-2 laboratory. In the reconstructed tomograms, we observed a typical coronavirus morphology of SARS-CoV-2 virions with diameters ranging from 80 nm to 120 nm (Fig. 1A). Inside each virion, the ribonucleoprotein complexes were tightly packed, with a diameter of ∼15 nm. From the deconvoluted tomograms using Warp (a computer software for cryo-EM data processing) (17), we could clearly visualize most Ss that were ready for subsequent particle picking. Both the prefusion and postfusion states of S were observed (Fig. 1 A and B), as reported previously (16), which was in line with the fact that cleavage of S had occurred during the sample preparation (SI Appendix, Fig. S1).
Cryo-ET of inactivated SARS-CoV-2 virions. (A) Slice view of tomographic reconstructions of BPL-inactivated SARS-CoV-2 virions. White arrows indicate Ss in the prefusion and postfusion states. (Scale bar, 20 nm.) (B) Selected slices of separate postfusion Ss. (C and D) Selected slices of oligomerized postfusion Ss with side-by-side (C) and branching (D) patterns. Dotted outlines indicate the adjacent postfusion Ss. All tomograms were deconvolved using Warp (17) and are displayed using IMOD (38). (E) Statistics of virion numbers per tomogram and numbers of postfusion Ss per virion. (F) Histograms of nearest pair distances for postfusion Ss in the experimental data and in the simulated data with random distributions.
Previous studies have argued that the percentages of S in the prefusion and postfusion states are related to viral inactivation methods. The majority of prefusion S was from formaldehyde-fixed samples (12), while a great portion of postfusion S appeared to be from the BPL-inactivated sample, with a ratio of up to 66 to 81.3% (16, 18). In our sample, we further investigated the ratio of prefusion to postfusion Ss. We picked all possible S particles by combining the template matching approach with the manual method. We averaged the maps of both prefusion and postfusion Ss to generate a reference for subsequent three-dimensional (3D) classification, which showed that 42% of the particles were classified into the prefusion state, and 48% were classified into the postfusion state (SI Appendix, Fig. S2). Thus, the populations of prefusion and postfusion Ss were similar in our BPL-inactivated sample, which was different from a previous report (16).
A recent study showed that prefusion S exhibits a flexible orientation with respect to the viral membrane, tilting from the vertical axis to the viral membrane at a range of 50° (12). This structural feature could help prefusion S seek and bind to the ACE2 receptor of the target cell. In contrast, by visual inspection, we found that a great portion of the postfusion Ss appeared perpendicular to the viral membrane, which suggested that the conformation of the postfusion S has a stable membrane proximal external region or a stable TM region. In addition to the dispersed postfusion Ss on the viral membrane (Fig. 1B), we also observed that some postfusion Ss oligomerized in parallel (Fig. 1C) or in branches (Fig. 1D). We statistically determined the postfusion S on the viral membrane and found that the distribution of the postfusion S was sparse, with three Ss per virion on average (Fig. 1E). However, by inspecting all pairs of postfusion Ss in the same virions and calculating the pair distances, we found that many pairs had distances of ∼20 nm or less (Fig. 1F), implying potential clustering behavior of postfusion Ss. In order to validate this observation, we generated a simulated dataset, in which the numbers of viruses and Ss on each virus were all kept the same as the experimental data, but the postfusion Ss were randomly distributed on sphere-shaped virus. Using the same calculation method as for the nearest pair distance, the randomly placed Ss had no clustering effect, showing a normal distribution pattern with a center of 60 nm (Fig. 1F). This suggested that the clustering peak of postfusion Ss in our experimental dataset was statistically significant.
Subtomogram Analysis of SARS-CoV-2 Postfusion S.
We then performed subtomogram analysis from a total of 15,525 selected S particles (SI Appendix, Figs. S2 and S3). After 3D classification and autorefinement, we obtained an in situ structural map of prefusion S with C3 symmetry at a resolution of 12.9 Ã… according to the gold standard Fourier shell correlation (FSC) coefficient at 0.143 (SI Appendix, Fig. S2). Our in situ structure of prefusion S was similar to those obtained in previous reports (12, 13). In the present study, we focused on postfusion S showing the nail shape on the viral membrane.
We utilized different approaches to align the particles of postfusion Ss by trying local or global searches of orientations with C1 or C3 symmetries. We found that only local searches with C3 symmetry with restriction of the Euler angles that had prior values during particle picking (SI Appendix, Fig. S2) could yield a high-resolution (10.9 Ã…) map according to the gold standard FSC coefficient at 0.143 (Fig. 2 A and B). From the averaged map of postfusion S, we clearly distinguished the head region (connector), stalk region (6-HB), and TM region. The three S protomers could also be distinguished from the map with a higher threshold (Fig. 2A). The directional FSC analysis confirmed that there was no preferred orientation in our dataset (Fig. 2C).
Nanometer-resolution in situ structure of the SARS-CoV-2 postfusion spike protein