supporting this possibility. First, EM images
of ab-propiolactone–inactivated SARS-CoV-2
virus preparation purified by a potassium
tartrate–glycerol density gradient appeared
to have lost all S1 subunits, leaving only the
postfusion S2 on the virion surfaces ( 43 ). Like-
wise, EM images of ab-propiolactone–inactivated
SARS-CoV-2 virus vaccine candidate (PiCoVacc)
also showed needle-like spikes on its surfaces
( 44 ). Second, spontaneous shedding of SARS-
CoV-2 S1 from pseudoviruses in the absence of
ACE2 has been reported ( 39 ). Third, binding
antibodies against S2 are readily detectable
in COVID-19 patients ( 45 ), suggesting that S2
is more exposed to the host immune system
than indicated by the unprotected surfaces on
the prefusion structures ( 22 , 23 ) (Fig. 2). We
therefore suggest that postfusion S2 trimers
may have a protective function by constituting
part of the crown on the surface of mature and
infectious SARS-CoV-2 virion (Fig. 5). The
postfusion S2 spikes are probably formed after
spontaneous dissociation of S1 independently
of the target cells.
Membrane fusion
We identify a structure near the fusion pep-
tide, the FPPR, that may play a critical role
in the fusogenic structural rearrangements of
the S protein. There appears to be cross-talk
between the RBD and the FPPR, mediated by
CTD1, because a structured FPPR clamps down
the RBD whereas an RBD-up conformation
disorders the FPPR. Moreover, the FPPR is
close to the S1/S2 boundary and the S2' cleav-
age site and thus might be the center of ac-
tivities relevant to conformational changes in
S. One possibility is that one FPPR occasion-
ally flips out of position due to intrinsic pro-
tein dynamics, allowing the RBDs to sample
the up conformation. A fluctuation of this kind
would loosen the entire S trimer, as observed in
modified soluble S trimer constructs ( 22 , 23 ).
Once an RBD is fixed in the up position by
binding to ACE2 on the surface of a target cell,
a flexible FPPR may enable exposure of the S2'
cleavage site immediately upstream of the ad-
jacent fusion peptide. The phenotype of the
D614G mutation appears to be consistent
with the notion that the FPPR is involved in
membrane fusion ( 39 , 40 ). Cleavage at the
S2' site releases the structural constraints on
the fusion peptide, which may initiate a cas-
cade of refolding events in S2, including for-
mation of the long, central, three-stranded
coiled coil; folding back of HR2; and ultimately
membrane fusion. Cleavage at the S1/S2 site
allows complete dissociation of S1, which may
also facilitate S2 refolding.
Questions regarding membrane fusion re-
main because the regions near the viral mem-
brane are still not visible in the reconstructions.
However, these regions all play critical struc-
tural and functional roles. For example, the
conserved hydrophobic region immediately
preceding the TM domain, and possibly the
TM itself, have been shown to be crucial for
S protein trimerization and membrane fu-
sion ( 31 ). The cytoplasmic tail, containing a
palmitoylated, cysteine-rich region, is believed
to be involved in viral assembly and cell-cell
fusion ( 32 – 35 ). Whether other viral proteins,
such as M protein, may help to stabilize the
spike by interacting with HR2 remains an open
question. Thus, we still need a high-resolution
structureofanintactSproteininthecontext
of the membrane and other viral components
to answer such questions.Considerations for vaccine development
Asafeandeffectivevaccineistheprimarymed-
ical option to reduce or eliminate the threat
posed by SARS-CoV-2. The first round of vac-
cine candidates with various forms of the S
protein of the virus are passing rapidly through
preclinical studies in animal models and clin-
ical trials in humans. Our study raises several
potential concerns about the current vaccine
strategies. First, vaccines using the full-length
wild-typesequenceoftheSproteinmayproduce
the various forms in vivo that we have observed
here. The postfusion conformations could ex-
pose immunodominant, non-neutralizing epi-
topes that distract the host immune system, as
documented for other viruses such as HIV-1
and RSV ( 46 , 47 ). Second, the approach to sta-
bilizing the prefusion conformation by intro-
ducing proline mutations at residues 986 and
987 may not be optimal because the K986P
mutation may break a salt bridge between
protomers that contributes to trimer stability.
The resulting S trimer structure with a relaxed
apex may induce antibodies that could not
efficiently recognize S trimer spikes on the
virus,althoughitmaybemoreeffectiveinin-
ducing anti-RBD–neutralizing responses than
the closed form. Third, considering the pos-
sibility that the postfusion S2 is present on
infectious virions, vaccines usingb-propiolactone–
inactivated viruses may require additional
quality control tests. Although the PiCoVacc
appears to provide protection against challenges
in nonhuman primates after three immuniza-
tions ( 44 ), it is unclear how to minimize the
number of the postfusion S2 trimers to avoid
batch variations. Structure-guided immunogen
designmaybeparticularlycriticalifSARS-CoV-2
becomes seasonal and returns with antigenic
drift, as do influenza viruses ( 48 ).REFERENCESANDNOTES- E. de Wit, N. van Doremalen, D. Falzarano, V. J. Munster,
Nat. Rev. Microbiol. 14 , 523–534 (2016). - N. S. Zhonget al.,Lancet 362 , 1353–1358 (2003).
- B. Hijawiet al.,East. Mediterr. Health J. 19 , S12–S18 (2013).
- V. M. Corman, D. Muth, D. Niemeyer, C. Drosten,Adv. Virus
Res. 100 , 163–188 (2018). - Y. Guanet al.,Science 302 , 276–278 (2003).
- B. Huet al.,PLOS Pathog. 13 , e1006698 (2017).
- H. A. Mohd, J. A. Al-Tawfiq, Z. A. Memish,Virol. J. 13 ,87(2016).
8. A. Banerjee, K. Kulcsar, V. Misra, M. Frieman, K. Mossman,
Viruses 11 , 41 (2019).
9. P. Zhouet al.,Nature 579 , 270–273 (2020).
10. S. Belouzard, J. K. Millet, B. N. Licitra, G. R. Whittaker,Viruses
4 , 1011–1033 (2012).
11. R. P. Rand, V. A. Parsegian,Can. J. Biochem. Cell Biol. 62 ,
752 – 759 (1984).
12. V. A. Parsegian, N. Fuller, R. P. Rand,Proc. Natl. Acad. Sci. U.S.A.
76 , 2750–2754 (1979).
13. S. C. Harrison,Virology479-480, 498–507 (2015).
14. M. Kielian,Annu. Rev. Virol. 1 , 171–189 (2014).
15. W. Weissenhornet al.,Mol. Membr. Biol. 16 ,3–9 (1999).
16. L. Duet al.,Nat. Rev. Microbiol. 7 , 226–236 (2009).
17. B. J. Bosch, R. van der Zee, C. A. de Haan, P. J. Rottier,J. Virol.
77 , 8801–8811 (2003).
18. M. Hoffmannet al.,Cell 181 , 271–280.e8 (2020).
19. J. K. Millet, G. R. Whittaker,Proc. Natl. Acad. Sci. U.S.A. 111 ,
15214 – 15219 (2014).
20. M. A. Tortorici, D. Veesler,Adv. Virus Res. 105 , 93–116 (2019).
21. F. Wuet al.,Nature 579 , 265–269 (2020).
22. D. Wrappet al.,Science 367 , 1260–1263 (2020).
23. A. C. Wallset al.,Cell 181 , 281–292.e6 (2020).
24. R. Hendersonet al., bioRxiv 2020.05.18.102087 [Preprint].
18 May 2020. https://doi.org/10.1101/2020.05.18.102087.
25. J. Lanet al.,Nature 581 , 215–220 (2020).
26. R. Yanet al.,Science 367 , 1444–1448 (2020).
27. J. Shanget al.,Nature 581 , 221–224 (2020).
28. Q. Wanget al.,Cell 181 , 894–904.e9 (2020).
29. A. C. Wallset al.,Proc. Natl. Acad. Sci. U.S.A. 114 , 11157– 11162
(2017).
30. W. Song, M. Gui, X. Wang, Y. Xiang,PLOS Pathog. 14 ,
e1007236 (2018).
31. B. Schroth-Diezet al.,Biosci. Rep. 20 , 571–595 (2000).
32. B. J. Bosch, C. A. de Haan, S. L. Smits, P. J. Rottier,Virology
334 , 306–318 (2005).
33. E. Lontok, E. Corse, C. E. Machamer,J. Virol. 78 , 5913– 5922
(2004).
34. C. M. Petitet al.,Virology 341 , 215–230 (2005).
35. R. Ye, C. Montalto-Morrison, P. S. Masters,J. Virol. 78 ,
9904 – 9917 (2004).
36. S. M. Hurtley, D. G. Bole, H. Hoover-Litty, A. Helenius,
C. S. Copeland,J. Cell Biol. 108 , 2117–2126 (1989).
37. S. H. Scheres,J. Struct. Biol. 180 , 519–530 (2012).
38. R. N. Kirchdoerferet al.,Sci. Rep. 8 , 15701 (2018).
39. L. Zhanget al., bioRxiv 2020.06.12.148726 [Preprint].
12 June 2020. https://doi.org/10.1101/2020.06.12.148726.
40. Z. Daniloskiet al., bioRxiv 10.1101/2020.06.14.151357 [Preprint].
7 July 2020. https://doi.org/10.1101/2020.06.14.151357.
41. X. Fan, D. Cao, L. Kong, X. Zhang,Nat. Commun. 11 , 3618 (2020).
42. T. Danieli, S. L. Pelletier, Y. I. Henis, J. M. White,J. Cell Biol.
133 , 559–569 (1996).
43. C. Liuet al., bioRxiv 2020.03.02.972927 [Preprint].
5 March 2020. https://doi.org/10.1101/2020.03.02.972927.
44. Q. Gaoet al.,Science 369 , 77–81 (2020).
45. F. Wuet al., medRxiv 2020.03.30.20047365 [Preprint].
20 April 2020. https://doi.org/10.1101/2020.03.30.20047365.
46. J. S. McLellanet al.,Science 342 , 592–598 (2013).
47. G. Freyet al.,Nat. Struct. Mol. Biol. 17 , 1486–1491 (2010).
48. J. J. Skehel, D. C. Wiley,Annu. Rev. Biochem. 69 , 531–569 (2000).
ACKNOWLEDGMENTS
We thank M. Liao for generous advice; the SBGrid team for
technical support; and S. Harrison, M. Liao, A. Carfi, and D. Barouch
for critical reading of the manuscript. Negative-stain and cryo-EM data
were collected at the Molecular Electron Microscopy Suite and the
Harvard Cryo-EM Center for Structural Biology, respectively, at
Harvard Medical School.Funding:This work was supported by NIH
grants AI147884 (to B.C.), 3R01AI147884-01A1S1 (to B.C), AI141002
(to B.C.), and AI127193 (to B.C. and James Chou), as well as a
COVID-19 Award by Massachusetts Consortium on Pathogen
Readiness (MassCPR; to B.C.).Author contributions:B.C. and Y.C.
conceived the project. Y.C. and H.P. expressed and purified the full-
length S protein. T.X. expressed and purified soluble ACE2 and
performed SPR and cell-cell fusion experiments. Y.C. and J.Z.
performed negative-stain EM analysis. J.Z. prepared cryo-EM grids
and performed EM data collection with contributions from S.M.S. and
R.M.W. J.Z. processed the cryo-EM data and built and refined the
atomic models for the prefusion S trimer and the postfusion
S2 trimer. Y.C. processed the S1 data. S.R. contributed to data
processing for S1 and provided computational support. S.R.-V.
contributed to cell culture and protein production. All authors
analyzed the data. B.C., Y.C., J.Z., and T.X. wrote the manuscript with
input from all other authors.Competing interests:The authors
SCIENCEsciencemag.org 25 SEPTEMBER 2020•VOL 369 ISSUE 6511 1591
RESEARCH | RESEARCH ARTICLES