Research Article

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

See allHide authors and affiliations

Science  18 Aug 2020:
eabd5223
DOI: 10.1126/science.abd5223

Abstract

The spike (S) protein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is required for cell entry and is the major focus for vaccine development. Here, we combine cryo electron tomography, subtomogram averaging and molecular dynamics simulations to structurally analyze S in situ. Compared to recombinant S, the viral S was more heavily glycosylated and occurred mostly in the closed pre-fusion conformation. We show that the stalk domain of S contains three hinges, giving the head unexpected orientational freedom. We propose that the hinges allow S to scan the host cell surface, shielded from antibodies by an extensive glycan coat. The structure of native S contributes to our understanding of SARS-CoV-2 infection and the development of safe vaccines.

The spike surface protein (S) of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is required to initiate infection (1). It binds to the angiotensin-converting enzyme 2 (ACE2) (2, 3) to mediate viral entry. S also determines tissue- and cell tropism. Mutations may alter the host range of the virus and enable the crossing of species barriers (4, 5). Vaccine efforts focus on neutralizing antibodies that block infection by binding to S.

S is a trimeric class I viral fusion protein (6) with a club-like shape of ~20 nm in length. The ectodomain consists of a head, which has been extensively studied in vitro. It is connected to the membrane by a slender stalk. The three receptor binding domains (RBDs) of the S head are conformationally variable, which may relate to receptor binding. In the closed conformation, the RBDs are shielded by the N-terminal domains (NTDs). In the open conformation, one RBD is exposed upwards (2, 3). Previous studies resolved roughly two thirds of the predicted 22 N-linked glycans that are thought to shield S against antibodies (2, 3). It remains unknown if the distribution of the conformational states and the glycosylation pattern observed with recombinant protein in vitro are representative for the native state generated during viral assembly. Furthermore, little is known about the stalk of S and how its conformational variability within the virion may impact on the accessibility of epitopes for neutralizing antibodies and facilitate viral entry.

SARS-CoV-2 virions present pre-fusion spike proteins in an irregular pattern

To structurally analyze SARS-CoV-2 spike protein in situ, we passaged the virus through tissue culture cells and purified it from the inactivated supernatant by sucrose centrifugation (see methods for detail). We acquired a large scale cryo electron tomography data set consisting of 266 tilt-series covering >1000 viruses. Visual inspection of the tomographic reconstructions revealed a very high-quality data set in which individual protein domains are clearly visible (Fig. 1A and movie S1). On average 40 copies of S trimer resided on the surface. S proteins appeared to be randomly distributed on the viral surface without any significant tendency to cluster (Fig. 2A).

Fig. 1 Cryo electron tomography of SARS-CoV-2 virions.

(A) Slices through tomographic reconstructions of SARS-CoV-2 virions. Scale bar 30 nm. (B) Same as (A), but as a gallery highlighting distinct features of S. All domains, including the transmembrane part, are clearly resolved. Although the vast majority of S is reminiscent of the pre-fusion conformation, the particles framed in orange are strikingly similar to the post-fusion conformation as proposed by (7). (C) Immunoblot showing the loss of cleavage products of SARS-CoV-2 S with uncleaved S (180 kDa) remaining, within 5 passages through tissue culture (loading control using anti-N antibody).

Fig. 2 Subtomogram analysis of SARS-CoV-2 S protein.

(A) Distance (top) and cluster-size distributions (bottom) of S on the viral surface, with non-overlapping hard disks of 10 nm diameter as reference. (B) Subtomogram average of the ectodomain of S shown isosurface rendered and fitted with the previously published atomic model as determined by single particle EM (PDB ID 6VXX). Subtomogram average in transparent gray, secondary structure elements in red and glycosylation sites in blue. (C) Same subtomogram average but shown as slices through the reconstruction. Scale bar 5 nm. (D) Detail of the average of the symmetric unit of S. Scale bar 5 Å. (E) Distribution of the angular orientation and distance of the ectodomain with respect to the bilayer. (F) Based on the initial subtomogram averaging, positions of the spike head were classified according to their distance from the lipid bilayer (supplementary materials). Averages of the resulting classes are shown isosurface-rendered; distance increases from left to right. Examples of individual particles are shown as slices at the bottom panel (scale bar 30 nm). At an optimal distance, the stalk domain stretches out and is resolved.

S was mostly present in the pre-fusion conformation (Fig. 1B). Post-fusion conformations (7, 8) were very rare (<0.1%), which appears typical for Vero E6 host cells (9). Sanger sequencing and immunoblot analysis revealed that the furin site for proteolytic cleavage into the S1 and S2 fragments (5, 10) was lost during tissue culture passage (Fig. 1C and fig. S1) confirming previous studies (11, 12). However, the isolate contained the D614G allele (13, 14). Large-scale sequencing of RNA isolated from tissue culture supernatant confirmed both findings (supplementary materials).

Subtomogram averaging using NovaSTA (15) and STOPGAP (16) resulted in a cryo electron microscopy (cryo-EM) map of the S head at 7.9 Å resolution (fig. S2), in which secondary structure elements and individual glycosylation sites were clearly discernible (Fig. 2, B and C). Classification suggested that about half of S was present in the fully closed conformation. A considerable fraction of the remaining subtomograms had one RBD exposed (fig. S3, supplementary materials). Structural analysis of the asymmetric unit yielded an average map of the closed conformation at an overall resolution of 4.9 Å. In particular, the cluster of parallel helices in the center of the head was clearly resolved (Fig. 2D and fig. S4).

By contrast, the stalk connecting the S head to the viral membrane appears to be dynamic. While the head was fully contained in the tomographic map, only the top of the stalk domain was resolved. Emerging from the neck of the spike head, it contains an 11-residue Leu repeat sequence (L1141, L1145 and L1152) and adopts an unusual right-handed coiled coil, consistent with a recent single-particle structure of the S head (7). We subsequently refer to this part of the stalk domain as the “upper leg.” Right-handed trimeric coiled coils were long thought to be absent from the structural proteome (17), but can be seen in the post-fusion structure of S from the related mouse hepatitis virus (18).

A stalk with three flexible hinges connects S to the viral membrane

The tomographic images suggest the presence of flexible hinges in the stalk. Whereas stalks of individual S proteins are clearly visible in the tomograms (Fig. 1B), subsequent to averaging their density declines sharply at the end of the trimeric coiled-coil forming the upper leg (Fig. 2B). Moreover, the head exhibited large positional and orientational freedom. It was tilted up to ~90 degrees with respect to the normal at distances of 5-35 nm from the membrane (Fig. 2E). We grouped our subtomograms into four classes, according to their distance from the bilayer, and averaged them separately. At an intermediate distance, parts of the stalk and bilayer were resolved, suggesting a more defined conformation (Fig. 2F). We next sub-selected ~3,200 particles in which the head was oriented roughly perpendicular to the membrane. In the resulting average, the stalk domain was resolved (fig. S5A). Visual inspection of the respective subtomograms, in which the stalk domains are clearly observed, further corroborated the idea of a kinked stalk with potentially several hinges (Fig. 2F). Local refinement of the lower part of the stalk (subsequently referred to as the “lower leg”) resulted in a moderately resolved structure that would be consistent with the continuation of the coiled coil below a flexible hinge (subsequently referred to as the “knee,” fig. S5B).

Molecular dynamics (MD) simulations helped us to pinpoint the molecular origins of the flexibility seen in the tomograms. We performed a 2.5 μs long all-atom MD simulation of a 4.1-million atom system containing four glycosylated S proteins anchored into a patch of viral membrane and embedded in aqueous solvent (Fig. 3A). In the simulations, the S heads remained stable. The stalks, however, exhibited pronounced hinging motions at the junctions between S head and upper leg (“hip”), between upper and lower leg (knee), and between the lower leg and transmembrane domain (“ankle”). This observation was consistent with discrete leg segments seen in raw tomograms (Fig. 3, B and C). The hip joint flexed the least (16.5 ± 8.8 deg), followed by the ankle (23.0 ± 11.7 deg) and knee (28.4 ± 10.2 deg) (Fig. 3D and fig. S6). However, the limited sampling in 4 × 2.5 μs of MD may not cover the full range of motions (compare Fig. 2E and fig. S6D).

Fig. 3 MD simulations of SARS-CoV-2 S protein.

(A) Model of the S protein. The three individual chains of S are shown in shades of red, N-glycosylation in blue, lipids of the ER-like membrane in gray with phosphates in green; “hip,” “knee” and “ankle” mark positions of the three flexible hinges. (B) Examples of the hinges as seen in the de-convoluted tomograms. Cyan and orange arrowheads indicate upper and lower leg respectively, with their typical lengths indicated. Scale bar 10 nm. (C) Hinge flexibility in the MD simulation illustrated through backbone traces (gray) at 75 ns intervals with different parts of the S protein fixed (red). (D) Probability density functions for hinge bending angles at hip, knee and ankle.

Structures of S seen along the MD trajectory fit well into the tomographic density of S proteins protruding from the viral surface (Fig. 4A). In particular, the joints of the hip, knee and ankle of the MD snapshots align with kinks in the density seen by cryo-EM. For a more detailed view, we flexibly fitted suitable snapshots of the MD simulations into the subtomogram averages classified according to the distance of the head from the membrane (compare Fig. 2F to Fig. 4B). Hinge bending gives the S stalk the flexibility required to connect also heavily tilted S heads to the viral membrane.

Fig. 4 Fitting of molecular simulations into cryo electron tomograms.

(A) Slices through tomograms (left) and isosurface rendered tomograms with snapshots of respective MD simulations superimposed without flexible fitting (right). The hinges of the stalk domain predicted by structural modeling (orange arrowheads) are consistent with the tomographic data. Scale bar 5 nm. (B) Fit of snapshots of MD simulations into the classes obtained for different distances of the head from the membrane (1-4) as presented in Fig. 2F. Shorter distances are concomitant with a stronger bending of the hinges and a lateral displacement of the stalk. Average MD density filtered to a resolution comparable to the subtomogram averages is shown as isosurface render (right).

As a result of hinge bending, the stalk is diluted out in subtomogram averages focused on the head (Figs. 2, B, C, and F, and 4B). Stalks were visible if the heads were aligned with the membrane normal (fig. S5A), or if the stalk itself was averaged separately (fig. S5B). To test this interpretation, we calculated the electron density averaged over the whole MD trajectory with aligned S heads. Filtered to a comparable resolution, this calculated 3D map is highly similar to the subtomogram averages (Fig. 4B). In rare cases, the coiled-coil near the membrane appears to be unfolded in the original tomograms (fig. S5C) and continuous with the disordered loops of the MD model.

Extensive N-glycosylation covers the surface of S

The predicted N-glycosylation sites, many already annotated in single particle EM maps (2), were generally very pronounced in the subtomogram averages. The electron density of N-glycans averaged over the MD trajectory was highly consistent with the tomographic map (Fig. 5A). Clustered glycosylation sites were visible in the raw density before averaging, e.g., protruding from the lower part of the S head (Fig. 5B). Analysis of individual sites in subtomogram averages further supports the decoration of spikes with rather bulky glycan chains (Fig. 5C). Notably, a number of sequons were resolved with more pronounced branching than previously reported (19). By contrast, the two predicted O-glycosylation sites (20) lacked excess density (fig. S7A). Sequon N17LT, due to its location on the unstructured N terminus, was not localized in the density (fig. S7B) but elongated features protruding from the tip of the N-terminal domain (fig. S7B) suggested the presence of sequons N74GT and N149KS.

Fig. 5 Analysis of S protein glycosylation sites and epitopes.

(A) N-glycosylation sites are clearly discernible in the subtomogram average of the head. From left to right: Isosurface rendering of subtomogram average with an individual N-glycosylation site indicated (orange arrowhead); same superimposed with the MD-calculated density for all annotated N-glycosylation sites; superimposed with previous structural model of the head (PDB ID 6VXX); superimposed with a snapshot of the MD simulations. N-glycosylation sites are shown in blue. (B) Tomographic slice highlighting an N-glycosylation site (orange arrow heads) in the original data. Scale bar 5 nm. (C) Highlight of N-glycosylation positions 709 and 1134 of the MD simulations (top) and in a previous structural model (bottom; PDB ID 6VXX, EMDB 21452). The subtomogram average is shown superimposed at different isosurface thresholds (transparent gray). Extensive additional density is visible. (D to F) The stalk domain is heavily glycosylated at the hinges. (D) Exemplary tomographic slices with bulky density at the hinge positions (orange arrowheads). Scale bar 5 nm. (E) Superposition of the subtomogram averages (transparent gray isosurfaces) of the head (framed red) and the stalk domain (framed green) with a respective snapshot of the MD simulations emphasizing the glycosylation at the hinges. (F) Same as (E) but shown as maximum intensity projection through the subtomogram averages. (G) Fits of snapshots from MD simulations into the surface of a virion; tomogram shown isosurface rendered in transparent gray. The position of epitopes for neutralizing antibodies at the RBDs are indicated with cyan arrowheads. (H) Cartoon illustrating a hypothetical docking event in which the hinges facilitate the engagement of multiple instances of S with their receptors.

N-glycosylation is also predicted on the knee (N1158HT, N1173AS) and the ankle (N1194ES) in the region not previously resolved by single particle EM (Fig. 3A). We observed that these positions generally appeared bulkier in tomographic reconstructions than one might expect if they were not glycosylated (Figs. 1B and 5D). Additional density was very clearly observed in subtomogram averages (Fig. 5, E and F, and fig. S5, A and B) and consistent electron density calculated from the MD trajectory aligned on the lower leg (fig. S7C). N-glycosylation in this region of S might protect the functionally important hinges from antibody binding and help to keep them flexible.

Discussion

The two major structural analysis techniques combined in this study are complementary. Our MD simulations revealed three flexible hinges within the stalk, coined hip, knee and ankle, which are consistent with the tomographic data. One might speculate that the high degree of conformational freedom of S on the viral surface is important for the mechanical robustness of the virus or may facilitate motions that interfere with antibody access to the stalk. It might also allow the spike to engage the relatively flat surface of host cells with higher avidity (Fig. 5, G and H). Tomographic studies of actual infection events might further address this in the future. In contrast to the pre-fusion conformation of spike, the post-fusion conformation previously observed in vitro and in situ (7, 9) and in this study (Fig. 1B), is apparently inflexible. To the best of our knowledge, extensive flexibility that would be comparable to the pre-fusion S stalk has not been reported for other class I viral fusion proteins, including HIV env, Influenza HA or Ebola GP. However, influenza HA attaches to micelles with a short linker permitting up to 25 degree bending (21).

A particularly unusual feature masked at the edge of the resolved density of single-particle structures but well resolved in the subtomogram averages is the short right-handed coiled-coil at the top of the pre-fusion stalk. Being lost in the post-fusion structure as resolved for SARS-CoV (8), we speculate that the right-handed coiled-coil is only marginally stable, priming the protein for a large structural reorganization in a spring-loaded viral fusion mechanism. Indeed, all three hinges are disassembled in the transition to the post-fusion conformation and placed outside its structural core (7, 8).

Overall, the observed distribution of S on the surface of the virion and its conformers is highly consistent with other studies (9, 22, 23). Host cell-type dependent differences in the abundance of pre- and post-fusion conformation (9, 22) may depend on different levels of ACE2 and the serine protease TMPRSS2 (10). If the furin cleavage site may play a role here remains to be addressed. An interesting difference is the higher abundance of S observed in this study as compared to (22, 23).

The fully closed pre-fusion conformation of S was abundant in situ. This finding emphasizes that the highly-engineered, recombinant versions of S locked into this conformation (24, 25) may indeed be valuable tools for vaccine development, although there are also differences to the in situ structure. N-glycosylation sites appeared very bulky in the tomographic map as compared to previous single particle analysis, suggesting that decoration with sugars may indeed be more extensive during viral assembly as compared to the recombinant ectodomain modified during default vesicular transport. Our map is suggestive of additional N-glycosylation at the hinges of the stalk domain and possibly on the very tips of the S NTDs. The native glycosylation pattern defines the accessibility of epitopes on the crowded viral surface (19), where the NTD and stalk domains appear occluded by neighboring spikes (Fig. 5G). A lack of excess density at the predicted O-glycosylation sites indicates that N-glycosylation dominates.

Using cryo-electron tomography of intact viruses, we could resolve functionally important parts of S, including its connection to the viral membrane and its glycan coat, which were masked in studies of recombinant detergent-solubilized protein. Beyond S, our large-scale tomographic data set contains rich and high-resolution structural information on SARS-CoV-2 particles in their native context. The in situ structures of several key viral components, including the nucleocapsid and the M protein that is highly enriched in the membrane, remain enigmatic. Our data might thus be explored to resolve such features in the future. We further demonstrate that high resolution structural models can be fitted directly into the tomographic reconstructions, underlining the remarkable quality of the data. This strategy might be further explored to build structural models of entire virions.

Supplementary Materials

science.sciencemag.org/cgi/content/full/science.abd5223/DC1

Materials and Methods

Figs. S1 to S7

Tables S1 and S2

References (2660)

MDAR Reproducibility Checklist

Movie S1

https://creativecommons.org/licenses/by/4.0/

This is an open-access article distributed under the terms of the Creative Commons Attribution license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

References and Notes

Acknowledgments: The cryo electron tomography data was collected at the EMBL Heidelberg Cryo Electron Microscopy Service Platform. The genome sequencing was done at the Genomics core facility of EMBL Heidelberg. We thank EMBL (B.T., W.J.H.H., S.M., A.S., M.B.) and the Max Planck Society (B.T., M.S., S.W., F.E.C.B., S.v.B., M.G., S.M., R.C., G.H., M.B.) for support, and the Max Planck Computing Data Facility for providing computational resources. B.T. acknowledges William Wan (Vanderbilt University) for helpful discussions. J.K.L. acknowledges excellent support by Regina Eberle (PEI). R.C. acknowledges the support of the Frankfurt Institute for Advanced Studies. The authors are indebted to G. Dobler and R. Wölfel, Bundeswehr Institute for Microbiology, for providing SARS-CoV-2 strain MUC-IMB1. Funding: We acknowledge a generous SuperMUC-NG computing allocation at the Leibniz Supercomputing Centre (M.S., S.v.B., M.G., F.E.C.B., R.C., G.H.), the Human Frontier Science Program (RGP0026/2017; S.v.B., G.H.), the German Ministry of Health (C.S.), the German Center for Infection Research (C.H., M.D.M.) and the Loewe center DRUID from the Justus Liebig university Giessen (J.K.L.) for funding. M.S. acknowledges support from the Austrian Science Fund FWF (Schroedinger Fellowship, J4332-B28). Author contributions: B.T. experimental design, tomographic reconstruction, particle picking, subtomogram averaging, structural analysis, paper writing. M.S. modeling design, molecular dynamics simulations, structural analysis, paper writing. C.S. experimental design, virus purification, biochemical analysis and sequencing. W.J.H.H. experimental design, cryo-EM data acquisition, tomographic reconstruction. S.W. experimental design, sample preparation and screening, data analysis. F.E.C.B., S.v.B., M.G. and R.C. molecular dynamics simulations, structural analysis. K.B. experimental design and virus purification. C.H. experimental design and virus growth. G.v.Z. experimental design, supervision. J.L. sequencing. N.T.D.d.A. sequencing. S.M. subtomogram averaging. A.S. tomographic reconstruction, particle picking. M.D.M. experimental design, supervision. G.H. modeling design, data analysis, supervision, paper writing. J.K.L. experimental design, supervision, paper writing. M.B. experimental design, supervision, paper writing. Competing interests: None declared. Data and materials availability: The original tilt series have been deposited (EMPIAR-10453). Subtomogram averages were deposited into the electron microscopy database under accession numbers EMD-11222 (S-trimer RBDs closed), EMD-11347 (S-trimer with one fully open RBD) and EMD-11223 (asymmetric unit with closed RBD). The viral sequencing reads have been deposited in the European nucleotide archive repository under the accession ID PRJEB39737. All other data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials. This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. To view a copy of this license, visit https://creativecommons.org/licenses/by/4.0/. This license does not apply to figures/photos/artwork or other content included in the article that is credited to a third party; obtain authorization from the rights holder before using such material.
View Abstract

Stay Connected to Science

Subjects

Navigate This Article