VIROLOGY
189,
1 O-20
(19%)
Three-Dimensional Reconstruction of Baculovirus Expressed Bluetongue Virus Core-like Particles by Cryo-Electron Microscopy ELIZABETH A. HEWAT,**’ TIMOTHY F. BOOTH,t PETER T. LOUDON,t
AND
POLLY ROYt+$
*Laboratoire de Biologie Structurale. CEA and CNRS URA 1333, DBMS/DSy CENG, 85X, 3804 1 Grenoble, France; tNERC Institute of Virology and Environmental Microbiology, Mansfield Road, Oxford OX 1 3SR, United Kingdom; *Laboratory of Molecular Biophysics, University of Oxford, South Parks Road, Oxford, OX1 3QU, United Kingdom; and SDepartment of Environmental Health Sciences, School of Public Health, The University of Alabama at Birmingham, University Station, Birmingham, Alabama 35294 Received January 3, 1992; accepted March 12, 1992 When the viral proteins VP3 and VP7 of bluetongue virus (BTV) are expressed simultaneously in the baculovirus system, core-like particles form spontaneously. The 3-D structure of these core-like particles, determined from cryoelectron micrographs, reveals an icosahedral structure 72.5 nm in diameter with 200 triangular spikes arranged on a T = 13,l lattice; The five spikes around each of the fivefold axes are absent. This is in contrast to the native BTV core particles which have a complete T = 13, I lattice of 280 spikes. The spikes, attributed to VP7 trimers appear as triangular columns 8.0 nm in height with distinct inner and outer domains. The inner shell of the core-like particles, or subcore-like particle, has a T = 1 lattice composed of 80 copies of VP3. The subcore-like particle is noticeably thicker around the fivefold positions. Pores in the subcore-like particle are situated near each of the local sixfold axes, below each six-membered ring of spikes. These pores could allow the passage of metabolites and RNA to and from the core for RNA transcription during infection. It is possible that the synthetic core-like particles have an incomplete complement of VP7 spikes because the ratio of VP7 to VP3 produced in the dual expression system is less than the 13:l required for complete core-like particles. Only the VP7 spikes which have the strongest affinity for the VP3 inner core and are involved in maintaining the structural integrity of the core-like particle are incorporated. The BTV core-like particle shows greater morphological similarity to the rotavirus than to the reovirus core particle. o 1992 Academic Press, Inc.
INTRODUCTION
1970, 1972). The outer viral capsid composed of VP2 and VP5 is degraded once the virion has entered the host cell to yield the core particle consisting of two major proteins (VP3 and VP7) and three minor proteins (VPl, VP4, and VP6) plus the dsRNA genome. The function of the three nonstructural proteins NSl, NS2, and NS3 in virus replication is unknown. NSl forms tubules and NS2 forms virus inclusion bodies (Lecatsas, 1968; Urakawa and Roy, 1988; Thomas et a/., 1990). The complete sequence of the BTV serotype-10 (BTV-10) genome and the general location of the major proteins in the virus capsid are known (Fukusho et al., 1989). Recently recombinant baculovirus expression systems using insect cells have been employed successfully to express up to five BTV proteins simultaneously (Loudon and Roy, 1991). This is a powerful technique for studying the sequence of events in virus assembly and the molecular interactions involved under conditions similar to those found in the natural host cell. When VP3 and VP7 of BTV are coexpressed in insect cells using a dual recombinant baculovirus, core-like particles (CLPs) are formed spontaneously (French and Roy, 1990). They resemble native BTV core particles in size and shape (as seen by conven-
The family Reoviridae is composed of viruses which have a double-stranded (ds) RNA genome consisting of from 10 to 12 segments. They have a double icosahedral capsid shell roughly 70 nm in diameter and no lipid envelope. While there are considerable differences in several steps in their replicative cycles the overall strategy of RNA transcription and replication is similar. Transcription involves synthesis of mRNA using a particle-associated transcriptase, with both strands of dsRNA remaining within the uncoated core. The three genera of Reoviridae which infect animals, including humans, are reovirus, orbivirus, and rotavirus. (For a review of Reoviridae see Fields et al., 1990.) Bluetongue virus (BTV) is the prototype virus of the Orbivirus genus. It is an arthropod-borne virus that is an economically important pathogen of ruminants (e.g., sheep) in many parts of the world. The BTV genome consists of 10 dsRNA segments each of which encodes at least one protein. Seven structural and three nonstructural proteins have been identified (Huismans, 1979; Martins eta/., 1973; Vet-woerd eta/., ’ To whom reprint requests should be addressed. 0042-6822192
$5.00
Copyright 0 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.
10
STRUCTURE
OF BTV CORE-LIKE 100
a0 60
40
20
010 0
Resolution
35 A-’
FIG. 1. The average common lines (continuous line) and cross common lines (dotted be) phase residual as a function of resolution for the 3.7 nm reconstructlon.
tional electron microscopy of negatively stained particles) and they have the same antigenicity although they do not contain the dsRNA genome. Dialysis of the CLPs against a low salt buffer removes their spikes leaving a smooth inner subcore-like particle (SCLP) composed of VP3 only (Loudon and Roy, 199 1). This adds to the evidence that the spikes are composed of VP7. It is noteable that SCLPs have not been detected in insect cells using a baculovirus expressing VP3 alone. This is probably because SCLPs are unstable under physiological conditions where VP7 may be necessary to stabilize the core-like particle structure (Loudon and Roy, 1991). While extensive effort has been focused on the biochemistry of BTV infection and assembly, the 3-D structure of the virus particle is less well characterized. In this paper we describe 3-D reconstructions of BTV10 core-like particles (including data to 6.5 and 3.7 nm) using image analysis of cr-yo-electron micrographs. The structural determination of the CLP yields information not available from the structure of the virus-derived core particle (Prasad et al., 1992), first because the contrast between the inner shell and the interior is higher for the CLP since it is empty, and second because it also provides information on the order of assembly of the core particle. These reconstructions reveal an icosahedral structure with 200 triangular spikes arranged on a T = 13,l lattice. The five spikes around each of the fivefold axes are absent, in contrast to the native core particles which have a complete T = 13,l lattice of 260 spikes. The spikes, attributed to VP7 trimers, appear as triangular columns 8.0 nm in height with distinct inner and outer domains. The inner shell of the core-like particles, the subcore-like particle, has a T
‘1
PARTICLES
= 1 lattice composed of 60 copies of VP3. Pores in the subcore-like particle are located near each local sixfold axis and possibly also at the fivefold axes. The significance of these structural features for the replication cycle of BTV is discussed and compared with the equivalent features of rotavirus and reovirus particles. Cryo-electron microscopy and image analysis have been used successfully to study the structure of a number of icosahedral viruses including rhesus rotavirus (Prasad et al., 1988, 1990; Yeager et al., 1990), reovirus (Metcalf et a/., 1991), adenovirus (Stewart et al., 1991); sindbis virus (Fuller, 1987), nonenveloped SV40 (Baker et al., 1988) and others. The major advantage of cryo-electron microscopy over conventional electron microscope techniques is that the biological molecules are observed in a frozen hydrated state in amorphous ice which closely resembles the native aqueous state. The problems associated with heavy metal stains, fixatives, dehydration, etc., are avoided. MATERIALS Growth
and purification
AND METHODS of BTV-10 CLPs and SCLPs
Suspension cultures of Spodoptera frugiperda cells were infected with a dual recombinant baculovirus expressing VP3 and VP7; after 3 days the cells were harvested and the CLPs purified by centrifugation on sucrose gradients as described in French and Roy (1990). SCLPs were prepared by dialysis of the CLPs against a low salt buffer for 6 days and purified from the capSomers by centrifugation as described in Loudon and Roy (1991). Preparation specimens
and observation
of frozen-hydrated
Frozen-hydrated specimens were prepared on holey carbon films as described by Dubochet et al., (1985). A modified Zeiss cryoworking unit was employed which allows manipulation of the specimen grid in gaseous rather than liquid nitrogen. The holey carbon films, supported on 400 mesh copper grids were glow-discharged prior to use. Samples of the capsid suspension (4 ~1)were applied to grids, blotted immediately with filter paper for l-2 set, and rapidly plunged into liquid ethane cooled by nitrogen gas at -175”. Specimens were observed at a temperature of -165 to -175” in a Zeiss 1OC electron microscope equipped with a top entry cold stage. The microscope was operated at 100 kV and images were recorded at 20,000~ magnification with a dose of less than 10 e-/A’ on Kodak SO1 63 electron image film which was developed in full-strength D19 for 12 min at room temperature.
HEWAT ET AL.
FIG. 2. Cryo-electron “spikes”
micrographs of CLPs at (a) 2 pm and (b) 6 pm underfocus. (arrowed) which are more highly contrasted.
The magnification was calibrated using the 4.0-nm first layer line in computed Fourier transforms of microtubule images (Amos and Klug, 1974; Chretien, 1991).
Conventional electron microscopy Specimens were prepared for conventional electron microscopy using one of the following negative stains: 2% w/v uranyl acetate, pH 4.4, 2% w/v PTA, pH 7.2, or 2% w/v sodium silicotungstate (SST), pH 7.2. These specimens were also observed in the Zeiss 1OC at room temperature at 100 kV under low dose conditions For heavy metal shadowing a suspension of 4 ~1 of specimen was applied to a glow-discharged carbon-
The CLPs aligned close to a threefold
axis have pair
coated grid and rapidly frozen in liquid ethane as described above. After sublimation at -65” in an Edwards vacuum evaporation unit for 1 to 3 hr they were shadowed at 45” from the surface normal with TaMI and observed in a Zeiss 1OC electron microscope operated at 100 kV at room temperature.
Image processing Preliminary examination of the micrographs was carried out using a CCD TV camera linked to a dedicated computer which produced the power spectrum of a 512 X 512 pixel image in 2 set (Tietz video and processing systems). Selected micrographs were digitized for further analysis using a CCD TV camera cou-
STRUCTURE
OF BTV CORE-LIKE
13
PARTICLES
tion was applied since all data lay inside the first zero of the contrast transfer function. Projected images of the CLP inner shell (SCLP) were simulated by setting the reconstructed CLP density to zero beyond a radius of 28.2 nm before projection. Surfaces of reconstructions were visualized using the Semper 6 Plus solid surface display facility.
RESULTS 10nm FIG. 3. The projected density of the 6.5nm 3-D reconstruction of the CLP (a) down the threefold axis and (b) close to the threefold axis. The highly contrasted parrs of “spikes” is characteristic of the r = 13 lattice with the five spikes missrng around each fivefold axis. Each of the “spikes” in this Instance are in fact aligned pairs of spikes. For a complete T= 13 lattice of spikes, all the projected “spikes” have a more unrform appearance (Prasad et a/., 1988, 1992).
Cryo-electron micrographs of CLPs at 2 and 6 pm underfocus are shown in Figs. 2a and 2b. Visual comparison of these images (or images of negatively stained specimens) with the corresponding images of native core particles (see French and Roy, 1990) do not suggest that there is any major structural difference between them although some CLPs appear to be in-
pled to a PC with an ADC card. A pixel size of 10.5 pm on the micrograph was employed, which for a magnification of 20,000X, corresponds to a pixel size of 5.2 A at the specimen. Image analysis was performed on a Vaxstation 3200 using the Semper 6 Plus image analysis package and a version of the MRC icosahedral reconstruction programs (Crowther et al., 1970a,b; Crowther, 1971) extensively modified by Fuller (1987) to improve the orientation refinement and analysis of the low contrast cryo-electron microscope virus images. The orientation and origin of each particle are determined by the method of common lines (Crowther, 197 1). In the higher resolution reconstruction the average cross common lines phases reached a value of 90” (i.e., random phases implying no correlation) at 3.5-nm resolution. Minimization of the phase residual between cross common lines in different images is used to obtain the best agreement between each image and the whole data set (Fig. 1). The Fourier transform data from a complete data set are then interpolated at regular intervals on a cylindrical coordinate system using the matrix inversion described by Crowther (197 1) with the axis along one of the fivefold icosahedral axes. For the higher resolution reconstruction including data to 3.7 nm all reciprocal eigen vectors for the normal equations were less than 1 .O and 99% less than 0.1 O/O. The density map is calculated by Fourier-Bessel inversion with only D5 symmetry and complete icosahedral symmetry is imposed by real space averaging (Fuller, 1987). Two separate reconstructions were performed to 6.5 and 3.7 nm on 14 and 12 particles, respectively. In each case no correction for the contrast transfer func-
FIG. 4. (a) Freeze etched, Ta/W shadowed CLP, vrewed from the side on whrch the metal was deposrted and highly defocused to show the surface symmetry The scale bars represent 50 nm. (b) As for (a) with dots markrng the brrghter regions. (c) The dots from (b) with the fivefold axes marked. Startrng at one 5fold axis and stepping along three sixfold axes, one then takes a left turn to reach the nearest frvefold axrs as shown In Fig. 5c.
HEWAT ET AL.
F IG. 5. Surface representations of the icosahedrally symmetrized reconstructed CLPs viewed along (a) a threefold and (b) a fivefold axes (for the 3.7.nm reconstruction) and(c) and (d) a twofold axis (forthe 6.5~nm reconstruction). The surface spikes attributed to VP7 trimers lie on a T= 13.1 lattice. The positions of spikes are marked with small dots, the spikes missing from the complete T = 13 lattice are marked with large tjots and the icosahedral fivefold axes and local sixfold axis are indicated (c). lcosahedral five-, three-, and twofold axes are marked on (d). The SItale bar represents 10 nm.
complete. It is only with hindsight that a close inspection of the cryo-electron microscope (EM) images reveals small differences. In particular the CLPs aligned close to a threefold axis have pairs of “spikes” which are more highly contrasted (Fig. 2, arrows). The projected density of the 3-D reconstruction of the CLP down the threefold axis is shown in Fig. 3 for comparison. Also, from cryo-EM images the diameter of the CLP is 72.5 nm whereas the diameter of the authentic core particle is 69 nm (Prasad et a/., 1990, and our
recent reconstruction of the authentic virus particle, Hewat et al., in preparation). The absolute hand of the core-like particle cannot be determined from the cryo-EM images which are projections of the structure. The hand was determined from the single surface structure observed on freeze etched, heavy metal shadowed CLPs (Figs. 4a, 4b, and 4~). It will be noted that the shadowed images are consistent with a T = 13, I lattice in which the five spikes around the fivefold axes are missing. Starting at one
STRUCTURE
OF BTV CORE-LIKE
FIG. 6. Sectrons of the reconstructed density map viewed down a twofold axts (a) through the distal spike domains showing the triangular section of the spokes and (b) through the particle ongin. The dtvtson of each spike into two domatns (Inner and outer) IS evident.
fivefold axis and stepping along three sixfold axes, one then takes a left turn to reach the nearest fivefold axis. Thus the lattice is left handed according to the convention of Caspar and Mug (1962). Surface representations of the icosahedrally symmetrized reconstructed CLP’s are shown in Figs. 5a and 5b viewed along the three- and fivefold axes. The surface spikes attributed to VP7 trimers lie on a T = 13,l lattice. The positions of spikes missing from the complete T = 13 lattice are indicated on Fig. 5c together with the local sixfold axes. The 6.5- and 3.7-nm reconstructions both show clearly the lack of spikes around the fivefold axes. The triangular section of the spikes is best seen in sections of the distal spike domains (Fig. 6a). The division of each spike into two
15
PARTICLES
domains is evident in the section shown in Fig. 6b. This section is oriented for direct comparison with the equivalent section of the rhesus rotavirus reconstruction (see Fig. 8 of Yeager et al., 1990) which has a very similar density for the core particle, except of course for the missing spikes. The inner shell (SCLP) is more clearly visible in the CLP reconstruction than in the core reconstruction partly because the interior is empty thus giving a higher contrast. A thickening of the inner shell on the fivefold positions is evident and a small pore on the fivefold axes is suggested by the lower density exactly on these axes. The exterior surface of the inner shell is almost spherical but it has an angular appearance because of the variations in thickness. The radial average of the section in Fig. 6b is shown in Fig. 7. It demonstrates the relatively thick inner shell and the two domains of the VP7 spikes. A further insight into the structure is obtained by setting the reconstructed density to zero beyond a chosen value of the radius. The surface contours of such truncated contour maps are shown in Figs. 8a-8c for a cutoff radius of 36.3, 31.9, and 28.2 nm. They show (a) the whole CLP, (b) the CLP minus the distal domains of the VP7 spikes, and (c) the inner shell only. While each VP7 spike is clearly divided into two domains the trimerit nature of the spikes is inferred by the triangular form of the spikes on local threefold axes. Given that VP7 has a molecular weight of 38.5 kDa it follows that, if it has roughly equal domains, each domain cannot be more than about 2.5 nm in diameter. This is too small to be resolved in the present reconstruction. In the surface representations shown so far the contour level has been chosen arbitrarily to give reasonable connectivity to the structure. However, because
0
93.75
187.5
281.3
375
Radius FIG. 7. The radral average of the sectton In Frg. 6b. Note the three dense shells at 26.6, 30.0, and 34.0 nm radtus (arrowed). We attrtbute the Inner shell to VP3 and the two outer shells to VP7.
16
HEWAT ET AL
of the limited resolution it is difficult to choose a contour level which correctly represents both globular domains and the holes in the structure: The globular domains are enlarged and the holes are filled in. A surface representation of the CLP with a higher contour level is shown in Figs. 9a-9d; large pores in the inner shell are located on the three local sixfold axes closest to each threefold axes and smaller pores are visible near the remaining local sixfold axes of the T = 13 lattice. The inner shell of the CLP (the SCLP) has only T = 1 symmetry. Since the density distribution in the reconstruction suggests that the basic unit of the inner shell is a monomer it follows that it is composed of 60 copies of VP3. This assignment of VP7 and VP3 gives a ratio of 10: 1 for VP7 to VP3 in the CLP. This is to be compared with the biochemically measured value of 15:2 using radioactive labeling for CLPs (French and Roy, 1990) and carbon 14 labeling for virus particles (Verwoerd et al., 1972). Our reconstruction-based model predicts a ratio of 13: 1 for the complete core particle supposing that all the spikes are VP7 trimers. Careful inspection of the cryoelectron micrographs reveals many CLPs which appear to have missing spikes, so one might expect the measured ratio of VP7 to VP3 to be less than the 10: 1. The existence of a few spikes around the fivefold positions cannot be completely ruled out but there is no trace of them in any of the CLP reconstructions. The inner shell is apparently composed of closely bonded pentamers of VP3 at each of the fivefold axes. Each VP3 monomer has two protrusions which link it to the neighboring VP3 pentamer. The pores in this shell are essentially spaces between VP3 molecules. At this resolution the molecular outline is largely speculative but the existence of a VP3 pentamer is given support by the electron microscope images of disintegrating SCLPs in PTA stain where pentamer-sized clumps of protein remain intact (data not shown). Images of the SCLPs in amorphous ice are shown in Fig. 10a. They are spherical with slight thickening of the walls seen in certain orientations. The appearance of SCLPs in negative stain depends strongly on the type of stain employed. Uranyl acetate (Loudon and Roy, 1991) gives images similar to those in amorphous ice but in PTA, pH 7, the SCLPs have an exaggerated angular appearance (see Fig. 1Ob and Huismans et al., 1987). This increased angularity of virus particles in
FIG. 8. Surface representations of truncated density maps for a cutoff radius of 36.3, 31.9, and 28.2 nm viewed down a twofold axis. They show (a) the whole CLP, (b) the CLP minus the distal domains of the VP7 spikes, and (c)the inner shell only. The threshold is higher in (c)than in (a) and (b) in order to show the general outline of the VP3 pentamer. The scale bar represents 10 nm.
STRUCTURE
OF BTV CORE-LIKE
PARTICLES
FIG.9. Surface representations of the CLP with high contour levels vrewed down a twofold axis from Inside the partrcle (a) and (b), from outside (c), and viewed down a threefold axis from outside the particle (d). In (a) the base of spikes on a threefold axes (large arrow) and adjacent to a twofold axis (small arrows) are vrsible. These spokes probably play an important role in maintaining the structural integrity of the CLP. In (b) a threefold axes IS marked and large pores in the inner shell are visible near the three adjacent local srxfold axes, There are apparently smaller pores (arrowed) near the remaining local sixfold axis of the T= 13 lattrce. None of the pores are located exactly on the local srxfold axes. The scale bar represents 10 nm.
negative stain has been observed for adenovirus (Stewart eta/., 1991). It is interesting to compare these SCLP images with the projections of the SCLP calculated from the CLP reconstruction (Fig. 1 1). The SCLPs clearly interact with the supporting carbon film and the majority are attached to the film on one of their five- or twofold axes. Only very occasionally are they attached at a threefold axis and practically never in an arbitrary orientation.
DISCUSSION The structure of the BTV-10 CLP presented here has interesting features not evident in the structure of the virus-derived core particle (Prasad et a/., 1992). The VP7 spikes which lie on a T = 13,l lattice are seen more clearly to consist of an inner and outer domain and their presumed trimeric nature, suggested in the core particle reconstruction, is supported by the triangular
18
HEWAT ET AL.
FIG. 11. Projections of the SCLP calculated from the CLP reconstruction viewed down a two-, three-, and fivefold axis are shown on the top row from left to right. PTA negatively stained SCLP’s in corresponding orrentations are shown on the bottom row. The scale bar represents 100 nm.
shown schematically in Fig. 12 (D. Stuart, private communication). The left handedness of the lattice has been established by shadowing experiments; this agrees with the hand determined by Prasad et al. (1992) using tilting experiments. In the core particle reconstruction the details of the inner shell are masked by the lack of contrast between the inner shell and the RNA/proteinfilled core, whereas in the CLP reconstruction, the empty core gives higher contrast and the pores in the inner shell are visible. There are possibly very small pores on each of the fivefold axes and two types of
a
FIG. 10. Images of the SCLPs (a) in amorphous ice are spherical with slight thickening of the walls seen in certain orientations (arrows) and (b) in PTA pH 7 the SCLPs have an exaggerated angular appearance. The scale bar represents 100 nm.
shape of the spikes on the CLP. Baculovirus expressed VP7 has recently been crystallized in two crystallographic forms with space groups P2, and P6,2,2 (Basak et al., 1991). In both crystal forms VP7 exists as a trimer. The molecular envelope of the 6 A map currently available shows a trimer with each VP7 monomer divided in one large and one small domain as
FIG. 12. Schematic twofold axes (arrow) distal domains have the CLPs. but when slightly causing the and the distal lobes core particle.
drawing of the two VP7 spikes either side of a showing that when the capsid is (a) empty the the mean separation of all the spikes as seen in it is full (b), the VP3 pentamers might bulge out junction between the pentamers to be flatter, of the spikes to become closer as seen in the
STRUCTURE
OF BTV CORE-LIKE
larger holes near each of the local sixfold axes. It is possible that the different pores may play a specialized role for synthesis of mRNA and/or in the passage of metabolites and mRNA into and out of the core particle during infection. The radial density plots of the CLP and core particle reconstructions differ in that the inner shell which has the highest density in the CLP reconstruction is barely visible in the core particle reconstruction. There are three shells of density in the CLP reconstruction and only two in the core particle. This led Prasad et a/. to attribute the outer shell to VP7 and the inner shell to VP3, whereas we attribute the two outer shells to VP7 and the extra inner shell to VP3. This attribution for VP3 is confirmed by the good agreement between the images of the SCLPs, which are known to be composed of VP3 only, (Loudon and Roy, 1991), and the projected densities of the reconstructed inner shell only. The inner shell has only T = 1 symmetry and hence 60 copies of VP3. The most striking difference between the core particle and the CLP that has been visualized in this study is the lack of five spikes around each of the fivefold axes on the CLP. Thus the CLP is composed of 200 trimeric spikes consisting of 600 copies of VP7 and an inner shell consisting of 60 copies of VP3. The core particle has 260 spikes which implies 780 copies of VP7 supposing that all the spikes are VP7 trimers. There are numerous possible explanations for the missing spikes, including (a) the missing spikes are a modified form of VP7; (b) they are a different protein, e.g., VP6; (c) post-translational processing of VP3 is necessary for the missing spikes to stick; (d) the presence of another viral protein is necessary for the spikes to stick; (e) the VP7 spikes fell off the capsid during purification; or (f) the ratio of expressed VP7 to VP3 is not sufficiently high (13: 1) to make completed CLPs. In the dual expression vector created to express VP7 and VP3 simultaneously both genes are under the control of a separate copy of the same promoter. It is possible that the ratio of expressed VP7 to VP3 is not high enough to produce complete CLPs. Only the VP7 spikes which have the strongest affinity for the VP3 inner core and/or are involved in maintaining the structural integrity of the CLP are incorporated. The VP7 spikes present on the CLPs are mostly those which are required to hold the VP3 pentamers together to form a sphere. The location of spikes on the edges of the star-shaped VP3 pentamers supports this hypothesis. This hypothesis may be tested by analyzing CLPs derived from expression vectors that produce a substantially larger quantity of VP7 than VP3. The diameter of the CLP (72.5 nm) is 5010 larger than that of the authentic core particle (diameter, 69 nm). The cause of this difference remains to be clarified.
19
PARTICLES
The variation in inter-spike spacing seen by Prasad et a/. in the core particle is not seen in the CLP; however, it is evident how such variation could occur. In particular the two spikes on either side of an icosahedral twofold axis are each located above a hole at the edge of adjacent VP3 pentamers and must play an important role in holding the pentamers together. They could be hinged together so that when the capsid is empty they have the mean separation of all the spikes, but when it is full, the pentamers bulge out slightly causing the junction between the pentamers to be flatter, and the distal lobes of the spikes to become closer as seen in the core particle (Fig. 12). It is interesting to compare the structure of the BTV10 to that of other members of the same family, the rhesus rotavirus (Prasad et a/., 1988, 1990; Yeager et a/., 1990), and the human reovirus (Metcalf et al., 1991). Each of these viruses has a double-shelled capsid with a T = 13,l lattice for the core particle. The outer shell of BTV has not been determined, but for the others it has also a T = 13,l lattice. The core particle of BTV and the rhesus rotavirus have very similar features except that the triangular spike motif is organized around each sixfold axis with the opposite relative rotation. The human reovirus has a remarkably different core particle with trumpet shaped projections around each of the fivefold axes and very little evidence of other spikes on the core particle. Both rhesus rotavirus and human reovirus have fine spike-like projections on their outer shell which have a hemagglutinating activity. So far no evidence for such a spike has been found in BTV. Clearly more detailed structural data are necessary to establish the relation between these representatives of the three genera of the Reoviridae family.
ACKNOWLEDGMENTS It IS a pleasure to thank C. Closse for expert technical assistance, R. H. Wade, Prof. D. Bishop and R. Rulgrok for their helpful comments on thrs manuscript, and S. Fuller for supplying the icosahedral reconstruction programs used in this work. This work was funded partly by NIH Grant Al 26879, AFRC Grant LRG43/535, and EMBO and NERC.
REFERENCES AMOS, L. A., and KLUG, A. (1974). Arrangement of subunits In flagellar microtubules. /. Cell SC/. 14, 523-549. BAKER, T. S., DRAC, J., and BINA, M. (1988). Reconstruction of the three-dimensional structure of simian virus 40 and visualization of the chromatln core. Proc. Nat/. Acad. SC;. USA 85, 422-426. BASAK, A. K., GRIMES, 1. BURROUGHS,J. N., MERTENS, P. P. C., ROY, P.. and STUART, D. (1991). Crystallographic studies on the structure of the blue tongue virus inner capsid. Submitted for publication. CASPAR, D. L. D., and KLUG, A. (1962). Physical principles of the construction of regular viruses. Cold Spring Harbor Symp. Quan?. Biol. 24, 1-24.
20
HEWAT ET AL.
CHR~IEN, D. (1991). Apports de la cryomicroscopie Blectronique B 1’6tude de microtubules assembl& in vitro. Thesis, Universitb Joseph-Fourier, Grenoble. CROWTHER, R. A. (1971). Procedures for the three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs. Phil. Trans. Roy. Sot. (London) B 261, 221230. CROWTHER, R. A., AMOS,L. A., FINCH,J. T., DE ROSIER,D. J., and KLUG, A. (1970a). Three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs. Nafure 226,42 l425. CROWTHER, R. A., DE ROSIER,D. J., and KLUG,A. (1970b). The reconstruction of three-dimensional structure from projections and its application to electron microscopy. Phi/. Trans. Roy. Sot. (London) A 317,319-340. DUBOCHET, J., ADRIAN,M., LEPAULT,J., and MCDOWALL,A. W. (1985). Cryo-electron microscopy of vitrified biological specimens. Trends Biochem. Sci 6, 143-l 46.
FIELDS,B. N., KNIPE,D. M., CHANOCLE,R. M., HIRSCH,M. S., MELNICK, J. L., MONATH,T. P., and ROIZMAN,B., Eds. (1990). “Virology,” 2nd ed. Raven Press, New York. FRENCH,T. J., and ROY,P. (1990). Synthesis of BTV corelike particles by a recombinant baculovirus expressing the two major structural core proteins of BTV. /. Viral. 64, 1530-l 536. FUKUSHO, A., Yu, Y ., YAMAGOUCHI, Y., and ROY,P. (1989). Completion of the sequence of bluetongue virus serotype 10 by the characterisation of structural protein VP6 and a non-structural protein, NS2. J. Gen. Viral. 70, 1677-1689. FULLER,S. D. (1987). The T = 4 envelope of sindbis virus is organized by interactions with a complementary T = 3 capsid. Cell 48, 923934. HUISMANS,H. (1979). Protein synthesis in bluetongue virus-infected cells. Virology 92, 385-396. HUISMANS,H., VAN DIJK,A. A., and ELS, H. J. (1987). Uncoating of parental blue-tongue virus to core and subcore particles in infected L cells. \/iro/ogy 157, 180-l 88. LECATSAS,G. (1968). Electron microscope study of the formation of bluetongue virus. Ondersfepoort J. Vet. Res. 35, 139-l 50.
LOUDON,P. T., and ROY, P. (1991). Assembly of five BN proteins expressed by recombinant baculovirus: Inclusion of the largest protein VP1 in the core-like and virus-like particles. L/iro/ogy 180, 798-802. MARTINS,S. A., Pm, D. M., and ZWEERINK, H. J. (1973). Studies on the topography of reovirus and bluetongue virus capsid proteins. J. Viral. 12, 194-198. METCALF,P. M., CYRKLAFF, M., and ADRIAN,M. (1991). The three-dimensional structure of reovjrus obtained by cryo-electron microscopy. EMBOJ. 10, 3129-3136. PRASAD,B. V., YAMAGOUCHI, S., and ROY, P. (1992). Three-dimensional structure of single shelled core particles of BTV. J. Viral. 66, 2135-2142. PRASAD,6. V. V., WANG,G. J., CLERX,J. P. M., and CHIU,W. (1988). Three-dimensional structure of rotavirus. J. Mol. Biol. 199, 269275. PRASAD,B. V. V., BURNS,1. W., MARIETTA,E., ESTES,M. K., and CHIU, W. (1990). Localization of VP4 neutralization sites in rotavirus by three-dimensional cryo-electron microscopy. Nature 343, 476479.
STEWART,P. L., BURNER, R. M., CYRCLAFF,M., and FULLER,S. D. (1991). Image reconstruction reveals the complex molecular organization of adenovirus. Ce// 67, 145-l 54. THOMAS,C. P., BOOTH,T. F., and ROY,P. (1990). Synthesis of bluetongueviral-coded phosphoprotein and formation of inclusion bodies by recombinant baculovirus in insect cells: It binds the single stranded RNA species. 1. Gen. Viral. 71, 2073-2083. URAKAWA,T.. and ROY,P. (1988). Bluetongue virus tubules made in insect cells by recombinant baculovirus: Expression of the NSl gene of bluetongue virus serotype 10. J. Viral. 62, 39 19-3927. VERWOERD, D. W., ELS, H. J., DE VILLIERS,E. M., and HUISMANS,H. (1972). Structure of the bluetongue virus capsid. J. Viral. 10, 783794. VERWOERD, D. W., Louw, H., and OELLERMANN, R. A. (1970). Characterization of bluetongue virus ribonucleic acid. J. l&o/. 5, l-7. YEAGER,M., DRYDEN,K. A., OLSEN,N. H., GREENBERG, H. B., and BAKER,T. S. (1990). Three-dimensional structure of rhesus rotavirus by cryoelectron microscopy and image reconstruction. J. Cell Biol. 110, 2133-2144.
189,
1 O-20
(19%)
Three-Dimensional Reconstruction of Baculovirus Expressed Bluetongue Virus Core-like Particles by Cryo-Electron Microscopy ELIZABETH A. HEWAT,**’ TIMOTHY F. BOOTH,t PETER T. LOUDON,t
AND
POLLY ROYt+$
*Laboratoire de Biologie Structurale. CEA and CNRS URA 1333, DBMS/DSy CENG, 85X, 3804 1 Grenoble, France; tNERC Institute of Virology and Environmental Microbiology, Mansfield Road, Oxford OX 1 3SR, United Kingdom; *Laboratory of Molecular Biophysics, University of Oxford, South Parks Road, Oxford, OX1 3QU, United Kingdom; and SDepartment of Environmental Health Sciences, School of Public Health, The University of Alabama at Birmingham, University Station, Birmingham, Alabama 35294 Received January 3, 1992; accepted March 12, 1992 When the viral proteins VP3 and VP7 of bluetongue virus (BTV) are expressed simultaneously in the baculovirus system, core-like particles form spontaneously. The 3-D structure of these core-like particles, determined from cryoelectron micrographs, reveals an icosahedral structure 72.5 nm in diameter with 200 triangular spikes arranged on a T = 13,l lattice; The five spikes around each of the fivefold axes are absent. This is in contrast to the native BTV core particles which have a complete T = 13, I lattice of 280 spikes. The spikes, attributed to VP7 trimers appear as triangular columns 8.0 nm in height with distinct inner and outer domains. The inner shell of the core-like particles, or subcore-like particle, has a T = 1 lattice composed of 80 copies of VP3. The subcore-like particle is noticeably thicker around the fivefold positions. Pores in the subcore-like particle are situated near each of the local sixfold axes, below each six-membered ring of spikes. These pores could allow the passage of metabolites and RNA to and from the core for RNA transcription during infection. It is possible that the synthetic core-like particles have an incomplete complement of VP7 spikes because the ratio of VP7 to VP3 produced in the dual expression system is less than the 13:l required for complete core-like particles. Only the VP7 spikes which have the strongest affinity for the VP3 inner core and are involved in maintaining the structural integrity of the core-like particle are incorporated. The BTV core-like particle shows greater morphological similarity to the rotavirus than to the reovirus core particle. o 1992 Academic Press, Inc.
INTRODUCTION
1970, 1972). The outer viral capsid composed of VP2 and VP5 is degraded once the virion has entered the host cell to yield the core particle consisting of two major proteins (VP3 and VP7) and three minor proteins (VPl, VP4, and VP6) plus the dsRNA genome. The function of the three nonstructural proteins NSl, NS2, and NS3 in virus replication is unknown. NSl forms tubules and NS2 forms virus inclusion bodies (Lecatsas, 1968; Urakawa and Roy, 1988; Thomas et a/., 1990). The complete sequence of the BTV serotype-10 (BTV-10) genome and the general location of the major proteins in the virus capsid are known (Fukusho et al., 1989). Recently recombinant baculovirus expression systems using insect cells have been employed successfully to express up to five BTV proteins simultaneously (Loudon and Roy, 1991). This is a powerful technique for studying the sequence of events in virus assembly and the molecular interactions involved under conditions similar to those found in the natural host cell. When VP3 and VP7 of BTV are coexpressed in insect cells using a dual recombinant baculovirus, core-like particles (CLPs) are formed spontaneously (French and Roy, 1990). They resemble native BTV core particles in size and shape (as seen by conven-
The family Reoviridae is composed of viruses which have a double-stranded (ds) RNA genome consisting of from 10 to 12 segments. They have a double icosahedral capsid shell roughly 70 nm in diameter and no lipid envelope. While there are considerable differences in several steps in their replicative cycles the overall strategy of RNA transcription and replication is similar. Transcription involves synthesis of mRNA using a particle-associated transcriptase, with both strands of dsRNA remaining within the uncoated core. The three genera of Reoviridae which infect animals, including humans, are reovirus, orbivirus, and rotavirus. (For a review of Reoviridae see Fields et al., 1990.) Bluetongue virus (BTV) is the prototype virus of the Orbivirus genus. It is an arthropod-borne virus that is an economically important pathogen of ruminants (e.g., sheep) in many parts of the world. The BTV genome consists of 10 dsRNA segments each of which encodes at least one protein. Seven structural and three nonstructural proteins have been identified (Huismans, 1979; Martins eta/., 1973; Vet-woerd eta/., ’ To whom reprint requests should be addressed. 0042-6822192
$5.00
Copyright 0 1992 by Academic Press, Inc. All rights of reproduction in any form reserved.
10
STRUCTURE
OF BTV CORE-LIKE 100
a0 60
40
20
010 0
Resolution
35 A-’
FIG. 1. The average common lines (continuous line) and cross common lines (dotted be) phase residual as a function of resolution for the 3.7 nm reconstructlon.
tional electron microscopy of negatively stained particles) and they have the same antigenicity although they do not contain the dsRNA genome. Dialysis of the CLPs against a low salt buffer removes their spikes leaving a smooth inner subcore-like particle (SCLP) composed of VP3 only (Loudon and Roy, 199 1). This adds to the evidence that the spikes are composed of VP7. It is noteable that SCLPs have not been detected in insect cells using a baculovirus expressing VP3 alone. This is probably because SCLPs are unstable under physiological conditions where VP7 may be necessary to stabilize the core-like particle structure (Loudon and Roy, 1991). While extensive effort has been focused on the biochemistry of BTV infection and assembly, the 3-D structure of the virus particle is less well characterized. In this paper we describe 3-D reconstructions of BTV10 core-like particles (including data to 6.5 and 3.7 nm) using image analysis of cr-yo-electron micrographs. The structural determination of the CLP yields information not available from the structure of the virus-derived core particle (Prasad et al., 1992), first because the contrast between the inner shell and the interior is higher for the CLP since it is empty, and second because it also provides information on the order of assembly of the core particle. These reconstructions reveal an icosahedral structure with 200 triangular spikes arranged on a T = 13,l lattice. The five spikes around each of the fivefold axes are absent, in contrast to the native core particles which have a complete T = 13,l lattice of 260 spikes. The spikes, attributed to VP7 trimers, appear as triangular columns 8.0 nm in height with distinct inner and outer domains. The inner shell of the core-like particles, the subcore-like particle, has a T
‘1
PARTICLES
= 1 lattice composed of 60 copies of VP3. Pores in the subcore-like particle are located near each local sixfold axis and possibly also at the fivefold axes. The significance of these structural features for the replication cycle of BTV is discussed and compared with the equivalent features of rotavirus and reovirus particles. Cryo-electron microscopy and image analysis have been used successfully to study the structure of a number of icosahedral viruses including rhesus rotavirus (Prasad et al., 1988, 1990; Yeager et al., 1990), reovirus (Metcalf et a/., 1991), adenovirus (Stewart et al., 1991); sindbis virus (Fuller, 1987), nonenveloped SV40 (Baker et al., 1988) and others. The major advantage of cryo-electron microscopy over conventional electron microscope techniques is that the biological molecules are observed in a frozen hydrated state in amorphous ice which closely resembles the native aqueous state. The problems associated with heavy metal stains, fixatives, dehydration, etc., are avoided. MATERIALS Growth
and purification
AND METHODS of BTV-10 CLPs and SCLPs
Suspension cultures of Spodoptera frugiperda cells were infected with a dual recombinant baculovirus expressing VP3 and VP7; after 3 days the cells were harvested and the CLPs purified by centrifugation on sucrose gradients as described in French and Roy (1990). SCLPs were prepared by dialysis of the CLPs against a low salt buffer for 6 days and purified from the capSomers by centrifugation as described in Loudon and Roy (1991). Preparation specimens
and observation
of frozen-hydrated
Frozen-hydrated specimens were prepared on holey carbon films as described by Dubochet et al., (1985). A modified Zeiss cryoworking unit was employed which allows manipulation of the specimen grid in gaseous rather than liquid nitrogen. The holey carbon films, supported on 400 mesh copper grids were glow-discharged prior to use. Samples of the capsid suspension (4 ~1)were applied to grids, blotted immediately with filter paper for l-2 set, and rapidly plunged into liquid ethane cooled by nitrogen gas at -175”. Specimens were observed at a temperature of -165 to -175” in a Zeiss 1OC electron microscope equipped with a top entry cold stage. The microscope was operated at 100 kV and images were recorded at 20,000~ magnification with a dose of less than 10 e-/A’ on Kodak SO1 63 electron image film which was developed in full-strength D19 for 12 min at room temperature.
HEWAT ET AL.
FIG. 2. Cryo-electron “spikes”
micrographs of CLPs at (a) 2 pm and (b) 6 pm underfocus. (arrowed) which are more highly contrasted.
The magnification was calibrated using the 4.0-nm first layer line in computed Fourier transforms of microtubule images (Amos and Klug, 1974; Chretien, 1991).
Conventional electron microscopy Specimens were prepared for conventional electron microscopy using one of the following negative stains: 2% w/v uranyl acetate, pH 4.4, 2% w/v PTA, pH 7.2, or 2% w/v sodium silicotungstate (SST), pH 7.2. These specimens were also observed in the Zeiss 1OC at room temperature at 100 kV under low dose conditions For heavy metal shadowing a suspension of 4 ~1 of specimen was applied to a glow-discharged carbon-
The CLPs aligned close to a threefold
axis have pair
coated grid and rapidly frozen in liquid ethane as described above. After sublimation at -65” in an Edwards vacuum evaporation unit for 1 to 3 hr they were shadowed at 45” from the surface normal with TaMI and observed in a Zeiss 1OC electron microscope operated at 100 kV at room temperature.
Image processing Preliminary examination of the micrographs was carried out using a CCD TV camera linked to a dedicated computer which produced the power spectrum of a 512 X 512 pixel image in 2 set (Tietz video and processing systems). Selected micrographs were digitized for further analysis using a CCD TV camera cou-
STRUCTURE
OF BTV CORE-LIKE
13
PARTICLES
tion was applied since all data lay inside the first zero of the contrast transfer function. Projected images of the CLP inner shell (SCLP) were simulated by setting the reconstructed CLP density to zero beyond a radius of 28.2 nm before projection. Surfaces of reconstructions were visualized using the Semper 6 Plus solid surface display facility.
RESULTS 10nm FIG. 3. The projected density of the 6.5nm 3-D reconstruction of the CLP (a) down the threefold axis and (b) close to the threefold axis. The highly contrasted parrs of “spikes” is characteristic of the r = 13 lattice with the five spikes missrng around each fivefold axis. Each of the “spikes” in this Instance are in fact aligned pairs of spikes. For a complete T= 13 lattice of spikes, all the projected “spikes” have a more unrform appearance (Prasad et a/., 1988, 1992).
Cryo-electron micrographs of CLPs at 2 and 6 pm underfocus are shown in Figs. 2a and 2b. Visual comparison of these images (or images of negatively stained specimens) with the corresponding images of native core particles (see French and Roy, 1990) do not suggest that there is any major structural difference between them although some CLPs appear to be in-
pled to a PC with an ADC card. A pixel size of 10.5 pm on the micrograph was employed, which for a magnification of 20,000X, corresponds to a pixel size of 5.2 A at the specimen. Image analysis was performed on a Vaxstation 3200 using the Semper 6 Plus image analysis package and a version of the MRC icosahedral reconstruction programs (Crowther et al., 1970a,b; Crowther, 1971) extensively modified by Fuller (1987) to improve the orientation refinement and analysis of the low contrast cryo-electron microscope virus images. The orientation and origin of each particle are determined by the method of common lines (Crowther, 197 1). In the higher resolution reconstruction the average cross common lines phases reached a value of 90” (i.e., random phases implying no correlation) at 3.5-nm resolution. Minimization of the phase residual between cross common lines in different images is used to obtain the best agreement between each image and the whole data set (Fig. 1). The Fourier transform data from a complete data set are then interpolated at regular intervals on a cylindrical coordinate system using the matrix inversion described by Crowther (197 1) with the axis along one of the fivefold icosahedral axes. For the higher resolution reconstruction including data to 3.7 nm all reciprocal eigen vectors for the normal equations were less than 1 .O and 99% less than 0.1 O/O. The density map is calculated by Fourier-Bessel inversion with only D5 symmetry and complete icosahedral symmetry is imposed by real space averaging (Fuller, 1987). Two separate reconstructions were performed to 6.5 and 3.7 nm on 14 and 12 particles, respectively. In each case no correction for the contrast transfer func-
FIG. 4. (a) Freeze etched, Ta/W shadowed CLP, vrewed from the side on whrch the metal was deposrted and highly defocused to show the surface symmetry The scale bars represent 50 nm. (b) As for (a) with dots markrng the brrghter regions. (c) The dots from (b) with the fivefold axes marked. Startrng at one 5fold axis and stepping along three sixfold axes, one then takes a left turn to reach the nearest frvefold axrs as shown In Fig. 5c.
HEWAT ET AL.
F IG. 5. Surface representations of the icosahedrally symmetrized reconstructed CLPs viewed along (a) a threefold and (b) a fivefold axes (for the 3.7.nm reconstruction) and(c) and (d) a twofold axis (forthe 6.5~nm reconstruction). The surface spikes attributed to VP7 trimers lie on a T= 13.1 lattice. The positions of spikes are marked with small dots, the spikes missing from the complete T = 13 lattice are marked with large tjots and the icosahedral fivefold axes and local sixfold axis are indicated (c). lcosahedral five-, three-, and twofold axes are marked on (d). The SItale bar represents 10 nm.
complete. It is only with hindsight that a close inspection of the cryo-electron microscope (EM) images reveals small differences. In particular the CLPs aligned close to a threefold axis have pairs of “spikes” which are more highly contrasted (Fig. 2, arrows). The projected density of the 3-D reconstruction of the CLP down the threefold axis is shown in Fig. 3 for comparison. Also, from cryo-EM images the diameter of the CLP is 72.5 nm whereas the diameter of the authentic core particle is 69 nm (Prasad et a/., 1990, and our
recent reconstruction of the authentic virus particle, Hewat et al., in preparation). The absolute hand of the core-like particle cannot be determined from the cryo-EM images which are projections of the structure. The hand was determined from the single surface structure observed on freeze etched, heavy metal shadowed CLPs (Figs. 4a, 4b, and 4~). It will be noted that the shadowed images are consistent with a T = 13, I lattice in which the five spikes around the fivefold axes are missing. Starting at one
STRUCTURE
OF BTV CORE-LIKE
FIG. 6. Sectrons of the reconstructed density map viewed down a twofold axts (a) through the distal spike domains showing the triangular section of the spokes and (b) through the particle ongin. The dtvtson of each spike into two domatns (Inner and outer) IS evident.
fivefold axis and stepping along three sixfold axes, one then takes a left turn to reach the nearest fivefold axis. Thus the lattice is left handed according to the convention of Caspar and Mug (1962). Surface representations of the icosahedrally symmetrized reconstructed CLP’s are shown in Figs. 5a and 5b viewed along the three- and fivefold axes. The surface spikes attributed to VP7 trimers lie on a T = 13,l lattice. The positions of spikes missing from the complete T = 13 lattice are indicated on Fig. 5c together with the local sixfold axes. The 6.5- and 3.7-nm reconstructions both show clearly the lack of spikes around the fivefold axes. The triangular section of the spikes is best seen in sections of the distal spike domains (Fig. 6a). The division of each spike into two
15
PARTICLES
domains is evident in the section shown in Fig. 6b. This section is oriented for direct comparison with the equivalent section of the rhesus rotavirus reconstruction (see Fig. 8 of Yeager et al., 1990) which has a very similar density for the core particle, except of course for the missing spikes. The inner shell (SCLP) is more clearly visible in the CLP reconstruction than in the core reconstruction partly because the interior is empty thus giving a higher contrast. A thickening of the inner shell on the fivefold positions is evident and a small pore on the fivefold axes is suggested by the lower density exactly on these axes. The exterior surface of the inner shell is almost spherical but it has an angular appearance because of the variations in thickness. The radial average of the section in Fig. 6b is shown in Fig. 7. It demonstrates the relatively thick inner shell and the two domains of the VP7 spikes. A further insight into the structure is obtained by setting the reconstructed density to zero beyond a chosen value of the radius. The surface contours of such truncated contour maps are shown in Figs. 8a-8c for a cutoff radius of 36.3, 31.9, and 28.2 nm. They show (a) the whole CLP, (b) the CLP minus the distal domains of the VP7 spikes, and (c) the inner shell only. While each VP7 spike is clearly divided into two domains the trimerit nature of the spikes is inferred by the triangular form of the spikes on local threefold axes. Given that VP7 has a molecular weight of 38.5 kDa it follows that, if it has roughly equal domains, each domain cannot be more than about 2.5 nm in diameter. This is too small to be resolved in the present reconstruction. In the surface representations shown so far the contour level has been chosen arbitrarily to give reasonable connectivity to the structure. However, because
0
93.75
187.5
281.3
375
Radius FIG. 7. The radral average of the sectton In Frg. 6b. Note the three dense shells at 26.6, 30.0, and 34.0 nm radtus (arrowed). We attrtbute the Inner shell to VP3 and the two outer shells to VP7.
16
HEWAT ET AL
of the limited resolution it is difficult to choose a contour level which correctly represents both globular domains and the holes in the structure: The globular domains are enlarged and the holes are filled in. A surface representation of the CLP with a higher contour level is shown in Figs. 9a-9d; large pores in the inner shell are located on the three local sixfold axes closest to each threefold axes and smaller pores are visible near the remaining local sixfold axes of the T = 13 lattice. The inner shell of the CLP (the SCLP) has only T = 1 symmetry. Since the density distribution in the reconstruction suggests that the basic unit of the inner shell is a monomer it follows that it is composed of 60 copies of VP3. This assignment of VP7 and VP3 gives a ratio of 10: 1 for VP7 to VP3 in the CLP. This is to be compared with the biochemically measured value of 15:2 using radioactive labeling for CLPs (French and Roy, 1990) and carbon 14 labeling for virus particles (Verwoerd et al., 1972). Our reconstruction-based model predicts a ratio of 13: 1 for the complete core particle supposing that all the spikes are VP7 trimers. Careful inspection of the cryoelectron micrographs reveals many CLPs which appear to have missing spikes, so one might expect the measured ratio of VP7 to VP3 to be less than the 10: 1. The existence of a few spikes around the fivefold positions cannot be completely ruled out but there is no trace of them in any of the CLP reconstructions. The inner shell is apparently composed of closely bonded pentamers of VP3 at each of the fivefold axes. Each VP3 monomer has two protrusions which link it to the neighboring VP3 pentamer. The pores in this shell are essentially spaces between VP3 molecules. At this resolution the molecular outline is largely speculative but the existence of a VP3 pentamer is given support by the electron microscope images of disintegrating SCLPs in PTA stain where pentamer-sized clumps of protein remain intact (data not shown). Images of the SCLPs in amorphous ice are shown in Fig. 10a. They are spherical with slight thickening of the walls seen in certain orientations. The appearance of SCLPs in negative stain depends strongly on the type of stain employed. Uranyl acetate (Loudon and Roy, 1991) gives images similar to those in amorphous ice but in PTA, pH 7, the SCLPs have an exaggerated angular appearance (see Fig. 1Ob and Huismans et al., 1987). This increased angularity of virus particles in
FIG. 8. Surface representations of truncated density maps for a cutoff radius of 36.3, 31.9, and 28.2 nm viewed down a twofold axis. They show (a) the whole CLP, (b) the CLP minus the distal domains of the VP7 spikes, and (c)the inner shell only. The threshold is higher in (c)than in (a) and (b) in order to show the general outline of the VP3 pentamer. The scale bar represents 10 nm.
STRUCTURE
OF BTV CORE-LIKE
PARTICLES
FIG.9. Surface representations of the CLP with high contour levels vrewed down a twofold axis from Inside the partrcle (a) and (b), from outside (c), and viewed down a threefold axis from outside the particle (d). In (a) the base of spikes on a threefold axes (large arrow) and adjacent to a twofold axis (small arrows) are vrsible. These spokes probably play an important role in maintaining the structural integrity of the CLP. In (b) a threefold axes IS marked and large pores in the inner shell are visible near the three adjacent local srxfold axes, There are apparently smaller pores (arrowed) near the remaining local sixfold axis of the T= 13 lattrce. None of the pores are located exactly on the local srxfold axes. The scale bar represents 10 nm.
negative stain has been observed for adenovirus (Stewart eta/., 1991). It is interesting to compare these SCLP images with the projections of the SCLP calculated from the CLP reconstruction (Fig. 1 1). The SCLPs clearly interact with the supporting carbon film and the majority are attached to the film on one of their five- or twofold axes. Only very occasionally are they attached at a threefold axis and practically never in an arbitrary orientation.
DISCUSSION The structure of the BTV-10 CLP presented here has interesting features not evident in the structure of the virus-derived core particle (Prasad et a/., 1992). The VP7 spikes which lie on a T = 13,l lattice are seen more clearly to consist of an inner and outer domain and their presumed trimeric nature, suggested in the core particle reconstruction, is supported by the triangular
18
HEWAT ET AL.
FIG. 11. Projections of the SCLP calculated from the CLP reconstruction viewed down a two-, three-, and fivefold axis are shown on the top row from left to right. PTA negatively stained SCLP’s in corresponding orrentations are shown on the bottom row. The scale bar represents 100 nm.
shown schematically in Fig. 12 (D. Stuart, private communication). The left handedness of the lattice has been established by shadowing experiments; this agrees with the hand determined by Prasad et al. (1992) using tilting experiments. In the core particle reconstruction the details of the inner shell are masked by the lack of contrast between the inner shell and the RNA/proteinfilled core, whereas in the CLP reconstruction, the empty core gives higher contrast and the pores in the inner shell are visible. There are possibly very small pores on each of the fivefold axes and two types of
a
FIG. 10. Images of the SCLPs (a) in amorphous ice are spherical with slight thickening of the walls seen in certain orientations (arrows) and (b) in PTA pH 7 the SCLPs have an exaggerated angular appearance. The scale bar represents 100 nm.
shape of the spikes on the CLP. Baculovirus expressed VP7 has recently been crystallized in two crystallographic forms with space groups P2, and P6,2,2 (Basak et al., 1991). In both crystal forms VP7 exists as a trimer. The molecular envelope of the 6 A map currently available shows a trimer with each VP7 monomer divided in one large and one small domain as
FIG. 12. Schematic twofold axes (arrow) distal domains have the CLPs. but when slightly causing the and the distal lobes core particle.
drawing of the two VP7 spikes either side of a showing that when the capsid is (a) empty the the mean separation of all the spikes as seen in it is full (b), the VP3 pentamers might bulge out junction between the pentamers to be flatter, of the spikes to become closer as seen in the
STRUCTURE
OF BTV CORE-LIKE
larger holes near each of the local sixfold axes. It is possible that the different pores may play a specialized role for synthesis of mRNA and/or in the passage of metabolites and mRNA into and out of the core particle during infection. The radial density plots of the CLP and core particle reconstructions differ in that the inner shell which has the highest density in the CLP reconstruction is barely visible in the core particle reconstruction. There are three shells of density in the CLP reconstruction and only two in the core particle. This led Prasad et a/. to attribute the outer shell to VP7 and the inner shell to VP3, whereas we attribute the two outer shells to VP7 and the extra inner shell to VP3. This attribution for VP3 is confirmed by the good agreement between the images of the SCLPs, which are known to be composed of VP3 only, (Loudon and Roy, 1991), and the projected densities of the reconstructed inner shell only. The inner shell has only T = 1 symmetry and hence 60 copies of VP3. The most striking difference between the core particle and the CLP that has been visualized in this study is the lack of five spikes around each of the fivefold axes on the CLP. Thus the CLP is composed of 200 trimeric spikes consisting of 600 copies of VP7 and an inner shell consisting of 60 copies of VP3. The core particle has 260 spikes which implies 780 copies of VP7 supposing that all the spikes are VP7 trimers. There are numerous possible explanations for the missing spikes, including (a) the missing spikes are a modified form of VP7; (b) they are a different protein, e.g., VP6; (c) post-translational processing of VP3 is necessary for the missing spikes to stick; (d) the presence of another viral protein is necessary for the spikes to stick; (e) the VP7 spikes fell off the capsid during purification; or (f) the ratio of expressed VP7 to VP3 is not sufficiently high (13: 1) to make completed CLPs. In the dual expression vector created to express VP7 and VP3 simultaneously both genes are under the control of a separate copy of the same promoter. It is possible that the ratio of expressed VP7 to VP3 is not high enough to produce complete CLPs. Only the VP7 spikes which have the strongest affinity for the VP3 inner core and/or are involved in maintaining the structural integrity of the CLP are incorporated. The VP7 spikes present on the CLPs are mostly those which are required to hold the VP3 pentamers together to form a sphere. The location of spikes on the edges of the star-shaped VP3 pentamers supports this hypothesis. This hypothesis may be tested by analyzing CLPs derived from expression vectors that produce a substantially larger quantity of VP7 than VP3. The diameter of the CLP (72.5 nm) is 5010 larger than that of the authentic core particle (diameter, 69 nm). The cause of this difference remains to be clarified.
19
PARTICLES
The variation in inter-spike spacing seen by Prasad et a/. in the core particle is not seen in the CLP; however, it is evident how such variation could occur. In particular the two spikes on either side of an icosahedral twofold axis are each located above a hole at the edge of adjacent VP3 pentamers and must play an important role in holding the pentamers together. They could be hinged together so that when the capsid is empty they have the mean separation of all the spikes, but when it is full, the pentamers bulge out slightly causing the junction between the pentamers to be flatter, and the distal lobes of the spikes to become closer as seen in the core particle (Fig. 12). It is interesting to compare the structure of the BTV10 to that of other members of the same family, the rhesus rotavirus (Prasad et a/., 1988, 1990; Yeager et a/., 1990), and the human reovirus (Metcalf et al., 1991). Each of these viruses has a double-shelled capsid with a T = 13,l lattice for the core particle. The outer shell of BTV has not been determined, but for the others it has also a T = 13,l lattice. The core particle of BTV and the rhesus rotavirus have very similar features except that the triangular spike motif is organized around each sixfold axis with the opposite relative rotation. The human reovirus has a remarkably different core particle with trumpet shaped projections around each of the fivefold axes and very little evidence of other spikes on the core particle. Both rhesus rotavirus and human reovirus have fine spike-like projections on their outer shell which have a hemagglutinating activity. So far no evidence for such a spike has been found in BTV. Clearly more detailed structural data are necessary to establish the relation between these representatives of the three genera of the Reoviridae family.
ACKNOWLEDGMENTS It IS a pleasure to thank C. Closse for expert technical assistance, R. H. Wade, Prof. D. Bishop and R. Rulgrok for their helpful comments on thrs manuscript, and S. Fuller for supplying the icosahedral reconstruction programs used in this work. This work was funded partly by NIH Grant Al 26879, AFRC Grant LRG43/535, and EMBO and NERC.
REFERENCES AMOS, L. A., and KLUG, A. (1974). Arrangement of subunits In flagellar microtubules. /. Cell SC/. 14, 523-549. BAKER, T. S., DRAC, J., and BINA, M. (1988). Reconstruction of the three-dimensional structure of simian virus 40 and visualization of the chromatln core. Proc. Nat/. Acad. SC;. USA 85, 422-426. BASAK, A. K., GRIMES, 1. BURROUGHS,J. N., MERTENS, P. P. C., ROY, P.. and STUART, D. (1991). Crystallographic studies on the structure of the blue tongue virus inner capsid. Submitted for publication. CASPAR, D. L. D., and KLUG, A. (1962). Physical principles of the construction of regular viruses. Cold Spring Harbor Symp. Quan?. Biol. 24, 1-24.
20
HEWAT ET AL.
CHR~IEN, D. (1991). Apports de la cryomicroscopie Blectronique B 1’6tude de microtubules assembl& in vitro. Thesis, Universitb Joseph-Fourier, Grenoble. CROWTHER, R. A. (1971). Procedures for the three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs. Phil. Trans. Roy. Sot. (London) B 261, 221230. CROWTHER, R. A., AMOS,L. A., FINCH,J. T., DE ROSIER,D. J., and KLUG, A. (1970a). Three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs. Nafure 226,42 l425. CROWTHER, R. A., DE ROSIER,D. J., and KLUG,A. (1970b). The reconstruction of three-dimensional structure from projections and its application to electron microscopy. Phi/. Trans. Roy. Sot. (London) A 317,319-340. DUBOCHET, J., ADRIAN,M., LEPAULT,J., and MCDOWALL,A. W. (1985). Cryo-electron microscopy of vitrified biological specimens. Trends Biochem. Sci 6, 143-l 46.
FIELDS,B. N., KNIPE,D. M., CHANOCLE,R. M., HIRSCH,M. S., MELNICK, J. L., MONATH,T. P., and ROIZMAN,B., Eds. (1990). “Virology,” 2nd ed. Raven Press, New York. FRENCH,T. J., and ROY,P. (1990). Synthesis of BTV corelike particles by a recombinant baculovirus expressing the two major structural core proteins of BTV. /. Viral. 64, 1530-l 536. FUKUSHO, A., Yu, Y ., YAMAGOUCHI, Y., and ROY,P. (1989). Completion of the sequence of bluetongue virus serotype 10 by the characterisation of structural protein VP6 and a non-structural protein, NS2. J. Gen. Viral. 70, 1677-1689. FULLER,S. D. (1987). The T = 4 envelope of sindbis virus is organized by interactions with a complementary T = 3 capsid. Cell 48, 923934. HUISMANS,H. (1979). Protein synthesis in bluetongue virus-infected cells. Virology 92, 385-396. HUISMANS,H., VAN DIJK,A. A., and ELS, H. J. (1987). Uncoating of parental blue-tongue virus to core and subcore particles in infected L cells. \/iro/ogy 157, 180-l 88. LECATSAS,G. (1968). Electron microscope study of the formation of bluetongue virus. Ondersfepoort J. Vet. Res. 35, 139-l 50.
LOUDON,P. T., and ROY, P. (1991). Assembly of five BN proteins expressed by recombinant baculovirus: Inclusion of the largest protein VP1 in the core-like and virus-like particles. L/iro/ogy 180, 798-802. MARTINS,S. A., Pm, D. M., and ZWEERINK, H. J. (1973). Studies on the topography of reovirus and bluetongue virus capsid proteins. J. Viral. 12, 194-198. METCALF,P. M., CYRKLAFF, M., and ADRIAN,M. (1991). The three-dimensional structure of reovjrus obtained by cryo-electron microscopy. EMBOJ. 10, 3129-3136. PRASAD,B. V., YAMAGOUCHI, S., and ROY, P. (1992). Three-dimensional structure of single shelled core particles of BTV. J. Viral. 66, 2135-2142. PRASAD,6. V. V., WANG,G. J., CLERX,J. P. M., and CHIU,W. (1988). Three-dimensional structure of rotavirus. J. Mol. Biol. 199, 269275. PRASAD,B. V. V., BURNS,1. W., MARIETTA,E., ESTES,M. K., and CHIU, W. (1990). Localization of VP4 neutralization sites in rotavirus by three-dimensional cryo-electron microscopy. Nature 343, 476479.
STEWART,P. L., BURNER, R. M., CYRCLAFF,M., and FULLER,S. D. (1991). Image reconstruction reveals the complex molecular organization of adenovirus. Ce// 67, 145-l 54. THOMAS,C. P., BOOTH,T. F., and ROY,P. (1990). Synthesis of bluetongueviral-coded phosphoprotein and formation of inclusion bodies by recombinant baculovirus in insect cells: It binds the single stranded RNA species. 1. Gen. Viral. 71, 2073-2083. URAKAWA,T.. and ROY,P. (1988). Bluetongue virus tubules made in insect cells by recombinant baculovirus: Expression of the NSl gene of bluetongue virus serotype 10. J. Viral. 62, 39 19-3927. VERWOERD, D. W., ELS, H. J., DE VILLIERS,E. M., and HUISMANS,H. (1972). Structure of the bluetongue virus capsid. J. Viral. 10, 783794. VERWOERD, D. W., Louw, H., and OELLERMANN, R. A. (1970). Characterization of bluetongue virus ribonucleic acid. J. l&o/. 5, l-7. YEAGER,M., DRYDEN,K. A., OLSEN,N. H., GREENBERG, H. B., and BAKER,T. S. (1990). Three-dimensional structure of rhesus rotavirus by cryoelectron microscopy and image reconstruction. J. Cell Biol. 110, 2133-2144.
