Research Article 871
Structures of orbivirus VP7: implications for the role of this protein in the viral life cycle Ajit K Basak1†§, Jonathan M Grimes1‡§, Patrice Gouet1, Polly Roy2,3,4 and David I Stuart1,5*
Background: Bluetongue virus (BTV) is the prototypical virus of the genus Addresses: 1The Laboratory of Molecular orbivirus in the family Reoviridae and causes an economically important disease Biophysics, University of Oxford, Rex Richards Building, South Parks Road, Oxford OX1 3QU, UK, in domesticated animals, such as sheep. BTV is larger and more complex than 2NERC Institute of Virology and Environmental any virus for which comprehensive atomic level structural information is Microbiology, Mansfield Road, Oxford OX1 3SR, UK, available. Its capsid is made primarily from four structural proteins two of which, 3School of Public Health, University of Alabama at Birmingham, Birmingham, Alabama 35294, USA, VP3 and VP7, form a core which remains intact as the virus penetrates the host 4Department of Biochemistry, University of Oxford, cell. Each core particle contains 780 copies of VP7. The architecture of the South Parks Road, Oxford OX1 3QT, UK and trimeric VP7 molecule has been revealed by crystallographic analysis and is 5Oxford Centre for Molecular Sciences, New unlike other viral coat proteins reported to date. Chemistry Building, South Parks Road, Oxford OX1 3QT, UK.
Results: Two new crystal structures of VP7 have been solved, one (a cleavage Present addresses: †Birkbeck College, Malet product) at close to atomic resolution and the other at lower resolution. The Street, London WC1E 7HX, UK and ‡EMBL Grenoble Outstation, c/o ILL, BP156, 38042 VP7 subunit consists of two domains. The smaller, ‘upper’, domain is exposed Grenoble Cedex 9, France. on the core surface and has the β jelly-roll motif common to many capsid proteins. The second, ’lower’, domain is composed of a bundle of α helices. §These authors contributed equally to this work. The cleavage product comprises the upper domain, which forms a rigid *Corresponding author. invariant trimeric fragment. The lower resolution structure of the intact molecule E-mail: [email protected] indicates that the α-helical domain can rotate about the linker to the upper Key words: domain to adopt radically different orientations with respect to the threefold axis bluetongue virus, protein flexibility, virus structure, X-ray crystallography in the intact protein. Received: 25 February 1997 Conclusions: The crystal structures of VP7 reveal a remarkable mix of rigidity Revisions requested: 21 March 1997 Revisions received: 12 May 1997 and flexibility that may provide insights towards understanding how VP7 Accepted: 5 June 1997 interacts with the other capsid proteins of different stoichiometries. These results suggest that substantial conformational changes in VP7 occur at some Structure 15 July 1997, 5:871–883 http://biomednet.com/elecref/0969212600500871 stage in the viral life cycle. Such changes may be related to the central role that VP7 is likely to play in cell attachment and membrane penetration. © Current Biology Ltd ISSN 0969-2126
Introduction BTV-1 and BTV-10, [4,5]), the viral structure and assem- Bluetongue virus (BTV) is the representative member of bly pathway, as well as modes of cell entry and exit the orbivirus genus in the Reoviridae, a family of double- remain poorly understood. stranded RNA (dsRNA) viruses that also includes the rotavirus and orthoreovirus genera, both of which are im- The Reoviridae have relatively large non-enveloped cap- portant human pathogens [1–3]. Orbiviruses are respon- sids that are robust enough for their icosahedral symmetry sible for a number of economically important diseases in to be visualised by conventional electron microscopy. wild and domesticated animals, including sheep, cattle The BTV particle (~800Å in diameter) is composed of and horses. The orbivirus and coltivirus genera of the seven structural proteins (VP1–VP7), which encapsidate Reoviridae infect mammalian hosts via transmission by the genome (ten discrete segments of dsRNA, each of arthropod vectors, such as ticks, mosquitoes and gnats which encodes one polypeptide chain) [6,7]. In the capsid (which they also replicate in). There are at least 14 sero- each of the RNA segments is thought to be associated groups in the orbivirus genus (as defined by cross-reac- with VP1, VP4 and VP6 which, it is presumed, collec- tivity tests) which include, in addition to BTV, African tively form the viral RNA-directed RNA polymerase horsesickness virus (AHSV) and epizootic hemorrhagic complex involved in the initiation, elongation and methy- disease virus (EHDV) [4]. Each group contains a large lation of the messenger RNA (mRNA), as well as nega- number of serotypes, characterised by specific neutralisa- tive strand RNA synthesis during replication/assembly. tion tests. Although BTV is relatively well studied (the These nucleoprotein complexes are enclosed in a stable complete genome has been sequenced for two serotypes, core made up of VP3 and VP7. 872 Structure 1997, Vol 5 No 7
The outer surface of the BTV core particle displays icosahe- RGD motifs are known to be involved in the attachment of dral symmetry with triangulation number T=13, laevo [8]. foot-and-mouth disease virus (FMDV) and other biological The core particle has a bristly surface due to 260 capsomers, systems to cells via integrins [16]. Interestingly, BTV core which are trimers of VP7 that sit on all local and strict icosa- particles bind to mammalian cells, competing for a receptor hedral threefold axes. These trimers of the 38kDa VP7 sub- of FMDV (P Mellors and P Mertens, personal communica- units (composed of 349 residues) clothe a thin icosahedral tion), and although the cores are only poorly infectious to scaffold of the much larger VP3 (103kDa) whose copy num- mammalian cells [17] this may simply reflect incorrect in- ber is thought to be 120, from biochemical studies [9]. ternalisation of the particle. Orbiviruses have a dual host life cycle (infecting both mammalian and insect cells) and we The viral core is shrouded in an outer capsid composed of would expect the virus to use distinct receptors for the dif- VP2 and VP5 which has icosahedral symmetry, although the ferent hosts, perhaps with common motifs for attachment. precise symmetry of the lattice on which the proteins sit is As BTV core particles are almost as infectious for insect unclear [10]. This outer layer is stripped off as the virus cells as the intact virus [17], they must possess a distinct penetrates the cell membrane [11], leaving an intact core entry mechanism. within the cell cytoplasm which acts as a factory, synthesis- ing mRNA that is extruded through pores in its surface. In this paper we compare the structures of four crystal forms of VP7, three from BTV and one from another Three non-structural proteins (NS1–NS3) are found in orbivirus (AHSV). Two of these structures have already infected cells, these proteins are thought to be involved in been described: that of intact BTV VP7 at close to atomic viral assembly, translocation to the cell membrane and resolution, crystallised in a monoclinic unit cell (we shall egress from the cell [12,13]. refer to this crystal form as REF in the subsequent dis- cussion) [1]; and a cleavage product of the AHSV-4 VP7, It is clear from both biochemical evidence and cryo-elec- composed of the upper domain only (termed AFRAG) tron microscopy (cryoEM) reconstructions that the number [2]. The structure determination of the two remaining of molecules of VP7 in the BTV particle is much larger forms is described here. One form is an upper domain than the number of VP3, VP2 or VP5 molecules [8,10]. VP7 fragment of BTV VP7, crystallised in a monoclinic unit forms the outer layer of the core particle, making a bulky cell (termed BFRAG; resolution 2.5Å), and the other is proteinaceous filling in the sandwich between the inner an hexagonal crystal form of the intact molecule at a reso- VP3 layer of the core and the outer capsid of VP2 and VP5. lution of 5.4Å (termed HEX). All crystal forms were The quasi-symmetry of the VP7 layer (T=13, l) matches solved using single isomorphous replacement (SIR) and neither that of the underlying layer of VP3 (presumed T=1 density modification techniques. Interestingly, the prote- lattice) nor that of the outer capsid proteins VP2 and VP5 olytic cleavages that occurred in BTV and AHSV were at (presumed T=1 lattice). This poses the question as to how homologous points in the polypeptide chain giving rise to VP7 is able to form a rather precise T=13, l icosahedral strikingly similar fragments. We focus here upon the lattice, mediated by quasi-equivalent contacts between VP7 structural similarities and differences revealed, and dis- molecules, as well as successfully making these symmetry- cuss the possible implications for core architecture, as- mismatched interactions. sembly and infectivity, with particular reference to the known structure of the outer layer of the core as des- Although the details differ, there appear to be deep similar- cribed in the accompanying cryoEM paper. ities in the mechanisms of viral cell entry between many members of the Reoviridae. The viral particles attach to Results and discussion cells, via unidentified receptors, and are endocytosed [14, Historically, work on the determination of the structure of 15]. At some point in the endosomal/lysosomal pathway, VP7 has moved through a number of different crystal forms. the outer capsid is stripped off and the core particle pene- HEX was the first crystal form obtained and data were trates into the cytoplasm. The BTV core seems to enter the quickly collected to low resolution and analysed (see Mate- cytoplasm before lysosomal fusion, whereas reovirus pro- rials and methods). These crystals were, however, unstable gresses through to lysosomes, where the outer capsid is within the crystallisation drop and from their debris well largely removed by hydrolytic enzymes, and the core enters ordered BFRAG crystals grew. The focus of attention was the cytoplasm. In addition, some members of the Reoviridae therefore shifted to these new robust, well ordered crystals family seem to be able to ‘activate’ their capsids (in the case and the structure of BTV solved from this crystal form. of BTV by the partial proteolysis of VP2 [11]). Such activa- When it became clear that the BFRAG crystals only con- ted particles (known as infectious subviral particles, ISVPs) tained a portion of the complete protein, some effort was initiate viral replication much more quickly, suggesting that directed towards growing crystals in the presence of pro- they enter the cell by direct membrane penetration. In both tease inhibitors. The result was a third crystal form, REF, BTV and AHSV VP7 there is an Arg-Gly-Asp (RGD) tri- which yielded a high-resolution structure of the intact pro- peptide motif, that may play a role in cellular attachment. tein [1]. The a priori knowledge of the internal structure of Research Article Structures of VP7 Basak et al. 873
Table 1
Data collection and processing statistics.
† ‡ Data set No. of Maximum Total Unique Completeness Rmerge MFID* FM
HEX Native 3 5.4 31 418 1689 87 7.1 – – – AuCN I 1 5.4 7 945 1492 77 7.8 17.5 0.49(6.0 Å) 2.1 AuCN II 1 5.4 8 785 1597 83 10.0 22.0 2.0 BFRAG Native 2 2.4 250 537 34 443 99 7.4 – – – AuCN 1 2.7 38 127 18 138 74 4.5 9.8 0.43 (4 Å) 2.4 (4 Å) REF Native 3 2.6 314 113 57 697 81 10.5 – – – AuCN 4 3.0 398 191 44 434 96 8.6 11.34 (4 Å) 0.2 (3.5 Å) 0.7 (3.5 Å)
*MFID = mean fractional isomorphous difference. †FM = figure of merit (cosine of the mean phase error). ‡
High-resolution structure of the intact molecule (REF) REF is a monoclinic crystal form of VP7 containing two copies of the complete trimer in the asymmetric unit [1]. The model has been refined using all available data to 2.6Å Bragg spacings, applying strict noncrystallographic symmetry (NCS) constraints, to give a residual R factor of 18.4% with good model stereochemistry.
Description of the REF structure The crystal structure has previously been described [1] and we simply present here the key features to serve as a basis for subsequent discussion. VP7 forms a trimer comprising three elongated subunits; the overall length of the trimer is 85Å, along the threefold axis. In cross-section the trimer is roughly triangular with maximum dimension 65 Å. Each of the subunits consist of two distinct domains, the upper and lower domains, with respect to the position of VP7 in the The mainchain fold of BTV-10 VP7 (crystal form REF). One monomer core structure (where the centre of the particle is vertically of the trimer is colour ramped progressively from blue to red from the N to the C terminus. The other two monomers are highlighted in grey. For downwards; Figure 1). The upper (outer with respect to an assignment of the secondary structure elements see Grimes et al. the core) domain contains roughly the cenral third of the [1]. (Figure produced using BOBSCRIPT; R Esnouf, unpublished polypeptide chain, residues 121–249, folded into a β sand- program, extensively modified from MOLSCRIPT [43] and rendered wich with the well known jelly-roll topology [18]. The with RASTER3D [44,45].) 874 Structure 1997, Vol 5 No 7
Figure 2
Stereo images of the Cα traces of VP7 (a) 200 200 (a) 240 240 structures. The monomer of VP7 from the 140 140 REF crystal form (residues 1–349) with every 150 150 tenth residue numbered. (b) The low 190 190 160 170 160 170 resolution structure of VP7 (HEX form). The 210 210 complete monomer from REF is shown in 230 230 220 220 lighter grey, with the β-sandwich domains of 130 130 both structures superimposed. Every tenth
250180 250180 residue is numbered, from 1–349. (c) The fragment of VP7 (BFRAG form; residues 349 349 123–253). The same residues for the upper 300 300 120 120 domain of REF are superimposed in a lighter 340 290 340 290 grey. (Figure produced using BOBSCRIPT.) 270 260 270 260 280 310 280 310 330 330 80 80 320 70 110 320 70 110 1 1 9010 90 10 2060 2060 40 100 40 100 50 30 50 30
(b) 200200 200200 240240 240240 140140 140140 150150 150150 190190 190190 160160 170170 160160 170170 210210 210210 230230 230 220220 340 220220 230 130130 349 130130
250180180 250180180 250 280330 250 280330 270 270 349 80320 349 80320 300 20 300 20 300290120 300290120 340 290 310 60 340 290 310 60 260 260 270 260120 70 270 260120 70 280 310 50 280 310 50 330 10 330 10 80 90 4030 80 90 4030 110 110 110 110 320 70 1 320 70 1 1 100 1 100 10 10 90 60 90 2060 20 40 100 40 100 50 50 30 30
(c) 200 200
240 240
140 140 150 150
160190 160 190 170 170 210 210
230 230 220 220 130 130
250 180 250 180
123 123
domain of a molecular partner. The two domains are linked We have unambiguously and accurately placed the REF by two extended loops (residues 120– 128 and 248–255), structure for the complete trimer in the cryoEM recon- which lie against each other. A stereo image of the Cα trace structions of the BTV core particle (see adjoining paper); of the monomer is shown in Figure 2a (Figure 2 includes the molecular threefold axis is perpendicular to the core similar views for the structures described below) and the surface and the broader base of the trimer contacts the in- trimer is shown in Figure 3a. ternal network of VP3 (Figure 1). There is no doubt from Research Article Structures of VP7 Basak et al. 875
Figure 3
Stereo images of the VP7 trimer from the (a) α crystal forms REF and HEX. C trace of (a) 200 200 VP7 from the REF crystal (residues 1–349) 240 240 with every tenth residue numbered; one 140 140 150 monomer is highlighted in black and its 150 190 160190 α 160 170 170 trimeric partners are drawn in grey. (b) C 210 210 trace of VP7 from the HEX crystal, with the 230 230 220 220 upper domains orientated as in (a). Residues 130 130 1–349 are shown and every tenth residue is 250180 numbered; one monomer is highlighted in 250180 black and its trimeric partners are drawn in 349 349 300 300 grey. (Figure produced using BOBSCRIPT.) 120 120 340 290 340 290 270 260 270 260 280 310 280 310 330 330 80 80 320 70 110 320 70 110 1 1 10 10 60 90 90 20 2060 40 100 40 100 50 30 50 30
(b) 200 200 240 240 140 140 150 150 160190 170 160190 170 210 210 220 230 220 230 130 130
250180 280330 250180 280330 270 270 80320 80320 300290 20 300290 20 310 60 310 60 260 260 120 70 120 70 50 50 1090 1090 110 4030 110 4030 1 100 1 100
the cryoEM results that the REF structure is similar to function for the upper domain. This improved the map that of all of the VP7 molecules on an intact viral core. considerably in the region of the α helical lower domain, and the model for this domain was fitted unambiguously to Low-resolution structure of the intact molecule (HEX) this new map (Figure 4) in a radically different orientation HEX, a hexagonal crystal form of VP7, contains one mono- to that observed in the REF structure. Refinement of this mer of VP7 in the crystallographic asymmetric unit; this modified model was then performed by considering the two crystal form only gave useful diffraction to 5.4Å spacings. domains as rigid bodies; reducing the number of parameters Analysis of the crystal form was somewhat problematic and to be refined to 12. The final R factor of 43.9% on all data to was not attempted until a detailed model for the molecule Bragg spacings of 5.4Å, indicates a unique solution. had been determined from a different crystal system (the REF structure). At this stage, it was clear from SIR elec- Description of the HEX structure tron-density maps that in the HEX structure the compact As the analysis of these crystals is at limited resolution trimeric head containing the β sandwich was essentially we have not attempted to model changes in the internal unchanged from the REF structure; this domain was easily structures of the domains. The electron density (Figure 4) fitted into the electron density. The α helical domain had, demonstrates that the upper domain remains essentially however, moved substantially, relative to its position in the identical to that seen in the REF structure, in both tertiary REF crystal form. Unfortunately, the quality of this part of and quaternary structure, whereas the lower domain is the SIR/solvent flattened electron-density map was not swung about the linking peptides to produce a radically dif- good enough for a straightforward and unambiguous posi- ferent set of quaternary interactions (Figures 2b and 3b). tioning of the atomic model of this domain. As discussed in Inspection of the electron density (Figure 4) also reveals greater detail in the Materials and methods section, the that there are some relatively minor changes in tertiary quality of this map was improved using a cyclical real space structure. No attempt has been made to model these struc- map modification procedure to solvent flatten SIR maps tural changes, although it is clear that the electron density and impose our a priori knowledge of the electron density for some of the helices furthest away from the connection 876 Structure 1997, Vol 5 No 7
Figure 4
Stereo image cartoon of the fold of the final model of the VP7 monomer from the HEX crystal form. The model is coloured red, with the averaged/solvent-flattened electron- density map, calculated using data to a resolution of 6.0 Å, shown as a semi- transparent green surface. (Figure produced using BOBSCRIPT and RASTER3D [44,45].)
region between domains, is more smeared, indicating some domain; the quaternary arrangement of these domains is less well ordered structure. The aspect ratio of this trimer is also preserved in the fragmented molecule (Figure 3b). inverted compared to the REF trimer, the overall dimen- The model contains a third of the complete polypeptide sions of the HEX trimer being approximately 60 Å along the chain running continuously from Arg123 through to threefold axis by 80Å in the perpendicular direction. Met253. However, in 2.5Å maps, calculated with both averaged and refined phases, the electron density for one High-resolution structure of a fragment of the molecule loop (corresponding to four residues from Arg178–Pro181) (BFRAG) is very weak, and only becomes connected at below 1σ The BFRAG crystal form is monoclinic and contains six (see Figure 5a). Nevertheless, maps calculated at 4Å do fragments of the monomer in the asymmetric unit. Each show continuous density for this region at 1σ, and the fragment comprises the smaller, jelly-roll domain of VP7 residues for this less well ordered loop have been mod- and the fragments are arranged as two trimers within the elled using these low-resolution maps (see Figure 5b). crystallographic asymmetric unit. The structure was deter- Presumably, the higher resolution Fourier terms simply mined, initially from a single crystal, using heavy-atom contribute noise in this region, rendering the map less phases improved by the imposition of NCS constraints interpretable. In the structure of the intact molecule this and solvent flattening. The excellent quality of the elec- loop is stabilised by packing interactions with the lower tron-density map allowed a molecular model to be con- domain, and clear continuous density is seen in maps cal- structed by a novel semi-automatic procedure (see Mat- culated with either averaged or refined phases (Figure 5c). erials and methods section). This model was refined at a resolution of 2.5Å using all available data, applying strict Structural comparisons NCS constraints, to a residual of 21.9% with good model Upper domain stereochemistry. No water molecules have been modelled Together, the different crystal forms present 16 individ- in this structure. ual copies of the upper domain of VP7 (REF, six; BFRAG, six; AFRAG, three; and HEX, one; Table 2) and Description of the BFRAG structure comparisons of these models emphasise the stability and The refined model of BFRAG is extremely similar to the robustness of this part of the trimer. Program SHP [19], same domain in the model of the complete monomer, con- which allows an automatic definition of ‘equivalent’ resi- taining identical residues (Figure 2c). The structure of dues, detected 123 structural equivalencies (out of 130 res- BFRAG is also very similar to the structure of the homolo- idues; from Tyr126– Ser252) between REF and BFRAG gous fragment from AHSV (AFRAG). The building block with a root mean square (rms) difference in Cα coor- of the crystal is the trimeric fragment made up from this dinates of 0.32Å (Figure 2c). There is a break of four Research Article Structures of VP7 Basak et al. 877
Figure 5
Stereo images of the models and densities for (a) residues 175–183. This loop is ordered in REF but partially disordered in BFRAG. Averaged electron-density maps are shown in green for two stages in the phase refinement of BFRAG and for the final stages of phase refinement for REF; the loop is coloured red and the rest of the molecule is drawn as a coil. (a) The electron density in a 4.0 Å averaged/solvent-flattened map for BFRAG, derived during phase extension (i.e. calculated 176 176 with phases derived from map back- 182 182 transformation); the map is contoured at 1σ. The model shown is for the 2.5 Å refined structure. (b) An averaged/solvent-flattened 178 178 map for BFRAG calculated at a resolution of 2.5 Å and contoured at 1σ. The map was 180 180 obtained by phase extension and shows the break in the electron density for the mainchain; the model of the loop in the 2.5 Å refined structure is superimposed. (c) The (b) same loop in the 2.6 Å resolution averaged/solvent-flattened map of REF. The map shows continuous, clearly defined electron density for this region of the structure. (Figure produced using BOBSCRIPT.)
176 176 182 182
178 178 180 180
(c)
176 176 182 182 178 178 180 180
residues in the equivalencies, from 178–181, due to the of the upper domain in HEX (calculated using the initial differences in the modelling of the disordered loop (the SIR/solvent-flattened phases) gave a real-space correlation maximum deviation is 5.0Å at residue 179; Figure 2c). A coefficient of 83.5% (calculated with GAP using 41139 real-space comparison of the electron density of BFRAG pixels; JMG and DIS, unpublished program). Both this cor- (calculated using refined phases) with the electron density relation in real space and the paucity of differences in Cα 878 Structure 1997, Vol 5 No 7
Figure 6 Lower domain The maps for the low-resolution HEX structure indicated a radically different orientation of the α-helical domains, relative to the invariantly trimeric β-sandwich domain, when compared to the model of the complete trimer at high resolution. For the orientation of the lower domains observed in HEX, it is possible to conceive two possible models for the trimer, dependent on which of the top domains the lower domain is connected to. Although the maps are not in themselves unambiguous we are confident that we have identified the correct connection between domains. In the proposed model the lower domains of the trimer ‘unwind’ compared to REF, resulting in a massive repositioning of these domains: a rotation of κ=160° about φ=92° and ϕ=102° (in the coordinate frame of the hexag- onal crystal). The alternative model is one in which the trimer ‘winds up’, producing a reduced but still substantial domain movement. A number of lines of evidence point to the unwound model being correct. Firstly, ‘unwinding’ of the trimer seems sterically more feasible than a conforma- tional winding up. Examination of the domain interface in Cartoon highlighting the movements of the lower domains of VP7 the REF structure reveals that the two connecting loops around the threefold axis of the molecule; the three monomers are form half of a tight right-handed helical turn. If the trimer shown in red, blue and green. The arrows show the direction of movement for the lower domain in going from the REF to the HEX unwinds, this half-turn helix would flip open which is ster- conformation, if the top domain is kept fixed. The top pair of images ically more acceptable than a further tightening. Further- shows REF (left) and HEX (right) viewed from above, the lower pair of more, in the REF structure, present in viral cores, this images shows the corresponding side views. half-turn buries those residues where cleavage occurs in BFRAG and AFRAG [2]. Unwinding of the connection region would expose the cleavage site to solvent, a prereq- coordinates between REF and BFRAG highlight the uisite to the observed proteolysis. This domain arrange- structural integrity of the upper domain of the trimer in ment and the presumed conformational change is shown the different crystallographic environments. The crystal in Figures 2, 3 and 6. As the analysis of the structure is at structure of AFRAG, solved at 2.3Å [2], is essentially the low resolution we cannot, a priori, rule out the possibility same cleavage product as that in the BFRAG crystals. The that the radical movement of the lower domains in the AFRAG polypeptide chain runs from Gly127–Thr251 HEX structure reflects cleavage of the interdomain linkers. (BTV residue numbers) with Arg177–Ala181 again appear- Other considerations, however, disfavour this hypothesis. ing disordered in the electron-density maps. The BTV The molecules in the crystals pack head-to-head and feet- and AHSV fragments can be superimposed with an rms to-feet, and therefore if cleavage had occurred one would deviation of 1.2Å between 125 equivalent Cα atoms, again expect the alternative layers of upper and lower domains emphasising the robustness of this trimeric domain. In to disassociate from one another. This, however, is clearly fact, the Cα traces of this domain for the two serogroups not the case as the crystals are robust enough for native differ mainly at the molecular surface, and in particular in and derivative data to be collected over a 24 hour time regions close to areas where deletions or insertions are period from each of several crystals. Nevertheless, to ad- present. The RGD motif is located on the outer β sheet dress this possibility and to confirm our interpretation of of the upper domain on an exposed loop. It is essentially the structure, averaging/solvent flattening was resumed. identical in REF and BFRAG, with an rms deviation The calculated electron density was imposed for both in Cα position of 0.02Å. The RGD motif in AHSV, domains but a sphere of density of radius 14Å, centred on however, is located in a different position, lower down on the connection region, was allowed to float during this the β sheet on a highly flexible loop [2]; flexibility that is process. Whilst these maps do not unambiguously define also observed in BFRAG (Figure 5b). Unfortunately, in whether VP7 was cleaved in the HEX crystals, all the fea- AHSV VP7 the mainchain for only one of the monomers tures of the electron density are consistent with an un- in the trimer can be modelled, and it is modelled less pre- cleaved structure. The internal structure of the α-helical cisely than in BTV VP7. The rms deviation in Cα posi- domain is not compromised although it may well be less tions for the RGD motif residues between AHSV and rigid; certainly the electron density for this domain is not BTV VP7 is 1.0Å, reflecting the flexibility of this region as sharp as the density for the rigid β-sandwich domain. in AHSV VP7. This is perhaps not surprising as the intratrimeric packing Research Article Structures of VP7 Basak et al. 879
interactions between the α-helical domains are very much across the orbivirus serogroups that have been sequenced. reduced in the HEX crystal (the surface area lost by each As the surface residues of the orbivirus VP7 molecules monomer on trimer formation is 2200Å2 for HEX, as op- sequenced to date show only 25% identity it seems proba- posed to 5200Å2 in the REF structure). Flexibility in ble that there is a structural or biological imperative to the these helical domains may explain why these crystals only flexibility in the connection domain. At this stage it is diffract to relatively low resolution. The two models of the speculation to suggest that this function is cleavage prior complete structure of VP7 shows that the molecule can to membrane penetration, but there are intriguing paral- take up two radically different structures, and has the lels with other known receptor-binding/membrane fusion potential for some internal flexibility. Most of the confor- proteins, such as influenza virus haemagglutanin and the mational flexibility arises from changes in the two con- surface glycoprotein from tick born encephalitis [21,22]. In necting loops, in line with the observation that this region these cases, a protein cleavage event takes place prior to is particularly susceptible to proteolytic cleavage in two receptor binding. This cleavage converts the protein into a different orbiviruses. metastable state, from which a drop in pH in the endo- some can trigger a major conformational relaxation to Conclusions produce the state capable of fusing with cell membranes. The VP7 trimer is the unit from which the complex A possible explanation for the observed cleavage patterns surface lattice of the viral core of BTV is built up. Because in both BTV and AHSV VP7 is that a similar event takes of the nature of the symmetry and the requirement for place, subsequent to binding of VP7 to a specific receptor icosahedral closure of this lattice, it seems reasonable to which could prime VP7 for a conformational change facili- expect some degree of flexibility in the orientation of the tating membrane penetration. It is possible that this lower domains of the VP7 trimer, with respect to the priming event could take place in the endosome on re- robust upper domains. We have demonstrated flexing in moval of the outer capsid of VP2 and VP5 subunits. The the lower α-helical domain relative to the upper rigid potential role of the RGD peptide in this process, as a β-sandwich domain, which has the necessary structural target for recognition by an integrin, is apparent. The integrity to retain its trimeric form. As one moves from the known structure of the I domain (an insertion domain icosahedral threefold axes on the core, the packing of the present in some integrins) of an integrin can be readily VP7 trimers varies, from close to a true hexagonal array to docked onto the RGD motif in BTV VP7 (and also in a pentagonal arrangement at the icosahedral fivefold axes. AHSV VP7) if we allow some flexibility in the loop (as This could cause a slight rearrangement of the lower observed in the crystal structure); this is in agreement domains, from the proper threefold symmetry seen in our with the observation that orbivirus core particles block crystal. Such a rearrangement would be consistent with RGD-mediated attachment of FMDV. In both BTV and the cryoEM fit reported in the accompanying paper. Varia- AHSV, integrin attaches to a structure on the flank of the tion in the energetics of the associated trimers in the upper domain of VP7 that is less protuberant than most surface lattice is indicated by cryoEM reconstructions of integrin-binding RGD loops. It is therefore possible that core-like-particles (CLPs) where molecules of VP7 are integrin binding might exert some strain on the capsid (as missing (or present in greatly reduced copy numbers) is thought to occur in picornaviruses, such as poliovirus around the fivefold axis [20]. and the human rhinoviruses which bind the cell adhesion molecule ICAM1 [23]), leading to cleavage of the interdo- It has not been established whether the proteolytic cleav- main linker by either endogenous proteases or autoprote- age we have observed in VP7 trimers from two different olysis. It may be worth investigating whether such an serogroups has any importance in the viral life cycle, or event occurs on orbivirus core penetration of the cell. is simply an artifact of the flexibility between the two Although autoproteolysis is thought to be a mechanism of domains. The viral core particles experience changes in maturation for other viruses [24], we do not favour this physicochemical conditions, particularly pH, as the virus mechanism for orbiviruses as there is no obvious trigger attaches to the cell and the core penetrates into the cyto- for the proteolytic event during crystallisation; proteolysis plasm. As yet there is no evidence that the change in pH, occurring for different proteins in different space groups. due to the endosomal pathway of cell entry, causes any While the VP7 cleavage site does not appear to have the conformational change in the structure of VP7. Indeed, motif of a protease cleavage site [2] both in BTV and the intact structure which corresponds to the conformation AHSV VP7 a conserved acidic amino acid and a basic seen in the core particle (REF) was crystallised at a lower residue are located in the vicinity: Asp254 and Arg123 pH than the rearranged VP7 structure observed in the (lysine in AHSV), respectively. HEX crystal form (Table 1). Examination of the BTV and AHSV VP7 fragments indicates that they both appear to Biological implications have the same cleavage sites (which seem to occur before Bluetongue virus (BTV) is a member of the orbivirus and after Gly127 and Tyr250, respectively) and it is inter- genus in the Reoviridae family. Orbiviruses are responsi- esting to note that these residues are strictly conserved ble for a number of diseases in wild and domestic 880 Structure 1997, Vol 5 No 7
animals, such as sheep, cattle and horses. It is hoped rearrangements, but is also susceptible to cleavage. Fur- that an increased knowledge of the structure of BTV, thermore, the core particle is in itself infectious and the the pathway of viral assembly and the modes of cell VP7 proteins on the outer surface of many orbivirus entry will ultimately aid the development of treatments cores bear an exposed integrin-binding motif; an obvious for orbiviral infections. The BTV core particle comprises candidate for the cellular receptor attachment site. It 780 subunits of the trimeric VP7 capsid protein which seems very likely that VP7 plays a role in the penetra- are organised around fewer copies of a larger structural tion of the cell membrane by the viral core particle. It protein, VP3. The VP3 proteins in turn enclose a seg- remains to be determined, however, if the intrigu- mented double-stranded RNA genome and associated ing structural analogies between VP7 and the receptor- replicative complexes. Once viral core particles are pro- binding proteins of enveloped viruses do indeed reflect vided with nucleoside triphosphates in the cytoplasm of analogous biological functions. an infected cell they synthesise mRNA, which is thought to be extruded through pores at the fivefold axes of the Materials and methods icosahedral cores (by analogy with reovirus and cyto- HEX: the 5.4 Å hexagonal crystal form of VP7 plasmic polyhedrosis virus [25,26]). The VP7 monomer Analysis of the hexagonal structure was divided into two parts, namely work done before and after analysis of the monoclinic crystal forms. has a two domain architecture distinct from that seen Although presented consecutively here, the latter subsections post in other viral capsid proteins to date: a smaller ‘upper’ date the work on BFRAG and REF and draw upon results from those domain with a β-sandwich structure (jelly-roll motif), structure determinations. and a larger ‘lower’ domain that is comprised almost Crystallisation entirely of α helices. BTV-10 VP7 was expressed as described in Oldfield et al. [27]. Hexag- onal rods were grown at a temperature of 16°C, from 10–12% w/v From the study of VP7 in a total of 16 subunits from PEG 6000, in 10 mM Tris/HCl buffer at pH 7.5. The protein concentra- four different crystal forms and two different serogroups tion prior to crystallisation was 12–15mg/ml and the crystals belong to space group P6 22 with unit cell dimensions a = b = 95.2 Å, of the orbiviruses, a picture emerges of a molecule rigid 3 c = 181.0 Å, α = β = 90°, γ = 120°. After four weeks these crystals disin- in part and yet capable of enormous tertiary and quater- tegrated and the monoclinic VP7 fragment crystals (BFRAG) grew nary structural rearrangements. The β-sandwich domain from the debris [28]. is extremely rigid and this head of the molecule is robust enough to retain exactly the same mode of trimerisation Data collection Data were collected in-house on a Xentronics area detector and a on proteolytic cleavage. Nevertheless, the trimerisation Rigaku RU200H rotating-anode source. Native data were collected mode of the whole molecule is more complex; while from four crystals and although these crystals showed some evidence still retaining, at least approximately, the intramolecular of diffraction to 3.0 Å, data were only collected to 5.4 Å (the limit of the threefold axis, the lower domains can pivot around the diffraction from the derivative crystals). Useful derivative data were col- linkage point by almost 180° such that the helical swirl lected from two crystals soaked in 2 mM KAu(CN)2, 12% PEG 6000, 10 mM Tris/HCl buffer at pH 7.5 for 24 h. They were treated indepen- of the domains around the molecular threefold axis dently in merging, isomorphous scaling and heavy-atom refinement. inverts from right-handed to left-handed. We find that XENGEN 1.3 was used to process both native and derivative data [29] the flexibility is associated with susceptibility to prote- (see Table 1). olytic cleavage, a feature conserved in at least two orbi- viruses. Whilst the cleavage and the observed confor- Phasing and refinement The KAu(CN) difference Patterson synthesis was immediately inter- mational change of the lower domain, with respect to the 2 pretable by eye in terms of a single heavy atom site per crystallographic upper domain, have not been formally demonstrated to asymmetric unit, whose position was refined using PHASE/REFINE have a biological role, it seems likely that there is some (P Shaw, unpublished program). The SIR map was then subjected to functional imperative in allowing the connection be- cyclical solvent flattening with envelope recalculation [30,31], which tween the two domains to be flexible. It is unlikely that confirmed that VP7 was a trimer with a well-defined two domain archi- tecture. It was clear that one domain formed a tight trimeric interaction this flexibility is a prerequisite for successful closure of with its partners and contained two β sheets, stacked approximately VP7 onto the inner shell of VP3 because, as seen in the parallel. This was confirmed when BFRAG was solved and at that accompanying paper, any such conformational changes stage these SIR phases were improved with judicious use of our in VP7 are small. All viruses, however, must somehow knowledge of the structure of the upper domain, using the following achieve membrane translocation and it is clear from pre- real-space method. The heavy-atom site for each derivative was re- refined using MLPHARE and two, figure of merit weighted, SIR maps vious structural studies of enveloped virus proteins, that calculated and averaged. Cyclical averaging using the program GAP protein cleavage plays a major role in ‘priming’ recep- (JMG and DIS, unpublished program) was then used to improve the tor-binding capsid proteins. Cleavage of these capsid quality of the electron density. The asymmetric units of the two maps proteins primes them with potential energy which later were averaged and the solvent regions flattened. A 6 Å electron-density drives a large scale conformational change that causes map for BFRAG was then averaged onto the corresponding portion of this new hexagonal map (using an envelope appropriate for the trimeric cell-membrane penetration [21,22]. We have demon- upper domain). After 20 cycles of density modification and phase com- strated that one of the proteins of the non-enveloped bination (using SIGMAA) the reciprocal space R factor had dropped BTV, VP7, is not only capable of large conformational from 36.8% to 18.4%. Research Article Structures of VP7 Basak et al. 881
The quality of the map was much improved by this procedure and the envelopes defined by dividing the original averaging envelope in half, high-resolution models of both the β-sandwich and the α-helical domains and sixfold averaging commenced. Throughout this procedure the were fitted as rigid bodies without ambiguity. This model was refined, envelopes were improved by inspection of 2| Fo |–| Fc | and | Fo |–| Fc | treating each domain as a rigid body, against data between 15.0 Å and maps. After over 600 cycles data had been extended to 2.5 Å and 5.4 Å (the limit of the native data) with an applied solvent correction. All phase refinement converged with a final reciprocal space R factor of refinements, for this and other crystal forms, were performed using 18.8% and correlation coefficient of 92%. The quality of the resultant XPLOR [32] and, as appropriate, the stereochemical parameters of 2| Fo |–| Fc | map was exceptional and following the trace of the polypep- Engh and Huber [33]. A final R factor of 43.6% indicated a unique tide chain and correctly aligning the sequence was trivial. An auto- solution. In an effort to clarify the structure of the connection region, mated building of the structure was achieved by using the programs further rounds of real-space map modification were carried out. A map CALPHA [36] and MUTATE (R Esnouf, unpublished program), based on phases and amplitudes for the molecular model was aver- CA_TRACE (JMG and DIS, unpublished program) and the graphics φ aged onto the 2 | Fo |–| Fc | , SIR/DENSITY MOD map, using an envelope program O [37]. In brief, the program BONES [38] was used to gener- defined by the coordinates for the refined domains, with a sphere of ate a skeletonised version of the map and this skeleton was edited radius 12 Å centred on the connection region allowed to float free (i.e. using O. CA_TRACE then converted the pseudo-atoms of the skeleton with no contribution from the model based map). This procedure was to Cα coordinates. These crude Cα positions were then fed into used in a cyclic fashion, in order to further improve the phase esti- CALPHA, which uses a database of fragments from known well-refined mates. The R factor between | Fo | and | Fc |, derived from back transfor- structures, to build a polyalanine chain. This model was inspected with mation of the density-modified map, dropped from 25.2% to 23.2%, reference to the averaged map and one or two residues inserted in a even with the imposition of such severe real-space constraints, and the number of loops (ten residues in all). These Cα coordinates were then map contained salient features indicating our interpretation was valid. rerun through CALPHA and the ‘rebuilt’ polypeptide compared to the map. The positions were essentially correct and the sidechains were BFRAG: a monoclinic crystal of a fragment of VP7 automatically built using MUTATE (which uses a rotamer database). Crystallisation Using the RSR_rigid option in O, polypeptide segments, usually six to Crystals of the fragment grew from the debris of the hexagonal crystals ten residues long, were automatically refined into the density. This was
[28] and belonged to monoclinic space group P21 with cell dimensions repeated using individual residues as rigid bodies. Finally, sidechains a = 69.8 Å, b = 97.8 Å, c = 72.2 Å, α = γ = 90°, β = 109.02°. Because were refined using RSR_rotamer option. Manual intervention was under physiological conditions VP7 is a trimer, it was deduced from an required twice to correctly fit sidechains into density. empirical solvent calculation that there was one trimer in the asymmet- ric unit with a solvent content of 40%; a large single peak at κ = 120° This model was refined using XPLOR, applying strict NCS constraints. for a self-rotation function appeared to confirm this [34]. Initial positional refinement was followed by simulated annealing, using all data to 3.0 Å Bragg spacings. Then the NCS constraints were opti- Data collection mised by refining the six monomers as rigid bodies. All data from 8 Å Initially only one crystal was available but fortunately it was extremely out to 2.5 Å were then included in cycles of positional and individual stable in the X-ray beam. It was possible to collect three native and two isotropic B-factor refinement. A solvent mask was used and all data derivative data sets from this crystal on a Xentronics area detector at from 20 Å–2.5 Å to were included in the refinement. The model for the room temperature. Later a new crystal was used to collect a higher res- monomer was rebuilt several times, which required a number of olution native data set to 2.4 Å on a MAR research imaging plate. Data peptide flips and the rebuilding of several loops; no waters were collected on the Xentronics area detector were processed and merged included. The current model has 88% of its residues in the most using XENGEN 1.3 [29]. The data set collected on the MAR research favoured regions of the Ramachandran plot, with an R factor on all data imaging plate was processed using DENZO and merged using ROTA- between 20 Å and 2.5 Å of 21.9% (because of the high NCS we do VATA/AGROVATA with the Xentronics data [34]. From our studies on not quote R free values) (Table 2). HEX we knew that KAu(CN)2 derivatised VP7 at a specific residue. The crystal was therefore soaked in 3 mM KAu(CN)2, under similar condi- REF: a monoclinic crystal form of the intact VP7 tions to those used previously for the hexagonal form. Table 1 summa- As a brief synopsis of the experimental procedure used to solve REF rizes the data collection statistics. has already been given [1] and the strategy employed was similar to that used with BFRAG, only a brief description will be given. Phasing and refinement The mean fractional isomorphous differences were rather modest for Crystallisation this derivative (9.8%) and yet the rather complex Patterson function Crystallisation conditions have been described [1]. The crystals belong was interpretable, using GROPAT [35], in terms of six heavy-atom to space group P21 with unit cell dimensions of a= 83.6 Å, b = 110.1 Å, c sites, which could be described as two trimers with co-linear threefold = 129.9 Å, α = γ = 90° and β = 103.1°. The unit cell size was consistent axes. This was in line with our expectation of a trimer in the asymmetric with two trimers per asymmetric unit and a solvent content of some 45%. unit (although we had only expected a single site of derivatisation per subunit). These heavy atom positions were refined using the PHASE/ Data collection REFINE suite of programs (P Shaw, unpublished program). Phase In-house data collection and processing have been described [1]. Native refinement and extension then commenced using threefold averaging in data were collected from four crystals. Three crystals were used to real space starting at 4.0 Å resolution. Initial NCS operators were cal- collect data sets II, III and IV to a resolution of 3Å. The fourth crystal, culated from the heavy atom sites. The molecular unit envelopes for which was larger than the other three, was used to collect a data set to a each trimer were defined, using GAP, by a logical combination of the resolution of 2.6Å spacings. Processing of the data using DENZO [39] solvent envelope (assuming 40% solvent) and a series of planes, and subsequent merging of the four separate data sets with defined as perpendicular to, and at the midpoint of, the vector between SCALEPACK [39] was straightforward (Table 1). From the work the centre of the clusters of heavy atoms and their crystallographically- described above, it was known that gold cyanide, bound specifically to related neighbours. After 180 cycles of averaging the data had been Cys123. Three crystals were soaked in 3mM KAu(CN)2, 10% w/v PEG extended to 3.2 Å, with a reciprocal space correlation coefficient of 3500, 0.1 M sodium acetate, pH4.6 for 24 h. Data were collected from 90%. However, the quality of the maps were poor, although there were each crystal to Bragg spacings of 3.0Å and then merged (see Table 1). clear discernible pieces of secondary structure. On closer inspection of the maps it became clear that there were two trimers in the asym- Phasing and refinement metric unit, with their threefold molecular axes lying almost co-linear There are two trimers of VP7 in the asymmetric unit and the positions with each other. New operators were calculated, two new molecular of six heavy-atom sites were determined using GROPAT [35]. These 882 Structure 1997, Vol 5 No 7
Table 2
Statistics for the different structures described and compared.
Crystal form REF BFRAG HEX AFRAG
Space group P21 P21 P6322 I4 Unit cell a,b,c (Å) 83.6, 110.1, 129.9 69.8, 97.8, 72.2 95.2, 95.2, 181.0 157.2, 157.2, 57.7 β = 103.1° β = 109.02° γ= 120° α = β = γ = 90° Resolution (Å) 15.0–2.6 20.0–2.5 15.0–5.4 15.0–2.3 No. of reflections 54 361 31 689 1689 29 456 Completeness (%) 81 99 87 93 No. of subunits in asymmetric unit 6 6 1 3 NCS applied constraints constraints N/A none R factor (%) 18.4 21.9 43.6 21.0 No. of non-hydrogen atoms 2703 1026 N/A 2728 backbone (Å2)* 24 33 N/A 34 Rmsd bond lengths (Å) 0.016 0.013 N/A 0.018 Rmsd bond angles (°) 1.9 2.6 N/A 2.2
* backbone is the mean isotropic B factor for atoms comprising the polypeptide backbone. †Rmsd = root mean square deviation.
positions were then refined against all reflections between 15 Å and the most favoured region of the Ramachandran plot (Table 2 gives the 3.5 Å using MLPHARE [40], giving an overall mean figure of merit of refinement and model statistics). 0.2 for the acentric reflections and 0.37 for the centric. These poor SIR phases were initially refined by cyclic solvent flattening using GAP. A Accession numbers solvent content of 50% was assumed and after 20 cycles the final R The atomic coordinates for REF and AFRAG have been deposited at the factor between observed and calculated structure factors was 10.0%. Brookhaven Protein Data Bank with accession codes 1BVP and 1AHS, respectively; the coordinates for BFRAG and HEX will be deposited and Further phase refinement used a very similar procedure to that are freely available from the authors (e-mail: [email protected]). described for BFRAG, although SIR and averaged phases were com- bined using SIGMAA [41]. During the cyclic averaging the envelopes were improved by inspection of 2 | Fo |–| Fc | and | Fo |–| Fc | maps and Acknowledgements the molecular orientations were repeatedly refined. This procedure We are grateful to Peter Mertens and E Yvonne Jones for many helpful dis- worked better for trimer 1 than trimer 2. After 61 cycles of density mod- cussions. Stephen Lee for advice and practical help with imaging tech- ification the real-space correlation coefficient for trimer 1 was 90% but niques, to Robert Esnouf and Richard Bryan for help with computing and to for trimer 2 was only 85%. As the NCS operators were not fully opti- Elspeth Garman and Karl Harlos for support of in-house data collection mised, BFRAG was used as a rigid body in reciprocal space to facilities. This project was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) and the Oxford Centre for Molecular improve them. The two trimeric fragments were placed in the appropri- Science (OCMS) is supported by the BBSRC, Medical Research Council ate positions, using the heavy-atom sites as markers, and then refined (MRC) and Engineering and Physics Research Council (EPSRC). DIS is as rigid bodies at 3.5 Å using XPLOR [32], yielding coordinates from supported by the MRC. which new NCS operators were derived. Real space averaging was resumed and converged rapidly to give real-space correlation coeffi- cients of 93.8% and 93.5% for trimers 1 and 2, respectively, with a References reciprocal space R factor of 19.3%. The 3.5 Å maps produced by this 1. Grimes, J.M., Basak, A.K., Roy, P. & Stuart, D.I. (1995). The crystal procedure were excellent, the course of the polypeptide chain was structure of bluetongue virus VP7. Nature 373, 167–170. unambiguous and a model was quickly built, with no attempt to cor- 2. Basak, A.K., Gouet, P., Grimes, J.M., Roy, P. & Stuart, D.I. (1996). rectly align the amino acid sequence with the 3D structure. Crystal structure of the top domain of African horsesickness VP7. Comparisons with bluetongue virus VP7. J. Virol. 70, 3797–3806. The high degree of structural redundancy (when refinement was con- 3. Knudson, D.L. & Monath, T.P. (1990). Orbiviruses. In Virology. (Fields, strained by the NCS) allowed coarse grained phase extension in recip- B.N., eds), pp. 1405–1433, Raven Press, New York. 4. Holmes, I.H., et al., & Uyeda, I. (1995). Family Reoviridae. in Virus rocal space using XPLOR. The model was refined initially at 3.5 Å Taxonomy: Classification and Nomenclature of Viruses. (Murhy, FA., using standard protocols for simulated annealing and phase refinement et al., & Summers, MD., eds) pp. 208–237, Springer Verlag, Vienna, [32], followed by phase extension to 3.0 Å which led to an R factor of Austria. 29.5%. Rigid-body refinement of the six monomers reduced this to 5. Roy, P. (1989). Bluetongue virus genetics and genome structure. 27.1%, and improved the estimates of the NCS operators. Including all Virus Res. 13, 179–206. data from 15.0 Å to 2.6 Å in the refinement led to an R factor of 28.1%. 6. Verwoerd, D.W., Louw, H. & Oellermann, R.A. (1970). Averaging and solvent flattening was then resumed at 2.6 Å, using the Characterisation of bluetongue virus nucleic acid. J. Virol. 29, 1–7. optimised NCS operators (phase combination was not used), leading 7. Mertens, P.P.C., Brown, F. & Sanger, D.V. (1984). Assignment of the to a reciprocal space R factor of 12.6% and correlation coefficient of genome segments of bluetongue virus serotype 1 to the proteins which they encode. Virology 135, 207–217. 96%. The averaged 2 | Fo |–| Fc | map was of exceptional quality and 8. Prasad, B.V., Yamaguchi, S. & Roy, P. (1992). Three-dimensional building the structure with FRODO [42] was trivial. Model refinement in structure of single-shelled bluetongue virus. J. Virol. 66, 2135–2142. XPLOR was recommenced using strict NCS constraints. The use of a 9. Burroughs, J.N., Grimes, J.M., Mertens, P.P. & Stuart, D.I. (1995). solvent mask allowed all data from 15.0 Å to 2.6 Å to be used in the re- Crystallisation and preliminary X-ray analysis of the core particle of finement and led to the current model which has 90% of its residues in bluetongue virus. Virology 210, 217–220. Research Article Structures of VP7 Basak et al. 883
10. Hewat, E.A., Booth, T.F. & Roy, P. (1992). Structure of bluetongue virus methods for building protein models in electron density and the particles by cryo-electron microscopy. J. Struct. Biol. 189, 61–69. location of errors in these models. Acta Cryst. A 47, 110–119. 11. Mertens, P.P., Burroughs, J.N. & Anderson, J. (1987). Purification and 38. Greer, J. (1974). Three-dimensional pattern recognition: an approach properties of virus particles, infectious subviral particles, and cores of to automated interpretation of electron density maps of proteins. J. bluetongue virus serotypes 1 and 4. Virology 157, 375–386. Mol. Biol. 82, 279–301. 12. Eaton, B.T., Hyatt, A.D. & White, J.R. (1983). Localisation of the 39. Otwinoski, Z.O. (1993). Oscillation data reduction program. In nonstructural protein NS1 in bluetongue in infected cells and its Proceedings of CCP4 Study Weekend: Data Collection and presence in virus particles. Virology 163, 527–537. Reduction. (Saywer, L., Isaacs, N. & Bailey, S., eds) SERC Daresbury 13. Hyatt, A.D., Zhao, Y. & Roy, P. (1993). Release of bluetongue virus- Laboratory, Warrington, UK. like particles from insect cells is mediated by BTV nonstructural 40. Otwinoski, Z.O. (1991). Maximum likelihood refinement of heavy atom protein NS3/NS3A. Virology 193, 592–603. parameters. In Proceedings of the CCP4 Study Weekend: 14. Borsa, J., Morash, B.D., Sargent, M.D., Copps, T.P., Lievaart, P.A. & Isomorphous Replacement and Anomolous Scattering. (Wolf, W., Szekely, J.G. (1979). Two modes of entry of reovirus particles into L Evans, P.R. & Leslie, A.G.W., eds) SERC Daresbury Laboratory, cells. J. Gen. Virol. 45, 161–170. Warrington, UK. 15. Zarbl, H. & Millward, S. (1983). The reovirus multiplication cycle. In 41. Read, R. (1986). Improved Fourier coefficients for map phases from The Reoviridae. (Joklik, W., ed), pp. 107–196. Plenum Press, New partial structures with errors. Acta Cryst. A 42, 140–149. York. 42. Jones, T.A. (1985). Interactive computer graphics: FRODO. Methods 16. Logan, D., et al., & Stuart, D.I. (1993). Structure of a major Enzymol. 115, 157–171. immunogenic site on foot-and-mouth disease virus. Nature 362, 43. Kraulis, P.J. (1991). MOLSCRIPT: a program to produce both detailed 566–568. and schematic plots of protein structure. J. Appl. Cryst. 24, 946–950. 17. Mertens, P.P.C., et al., & Mellor, P.S. (1996). Enhanced infectivity of 44. Merrit, E.A. & Murphy, M. (1994). Raster3D version 2.0. A program for modified bluetongue virus particles for two insect cell lines and for two photorealistic molecular graphics. Acta Cryst. D 50, 869–873. Cuilicoides vector species. Virology 217, 582–593. 45. Bacon, D.J. & Anderson, W.F. (1988). A fast algorithm for rendering 18. Stuart, D.I. (1993). Viruses. Curr. Opin. Struct. Biol. 3, 167–174. space-filling molecule pictures. J. Mol. Graph. 6, 219–220. 19. Stuart, D.I., Levine, M., Muirhead, H. & Stammers, D.K. (1979). Crystal structure of cat muscle pyruvate kinase at resolution of 2.6 Å. J. Mol. Biol. 134, 109–142. 20. Hewat, E.A., Booth, T.F. & Roy, P. (1992). 3-D Reconstruction of bluetongue virus core-like particles by cryo-electron microscopy. Virology 189, 10–20. 21. Bullough, P.A., Hughson, F.M., Skehel, J.J. & Wiley, D.C. (1994). Structure of influenza haemagglutanin at the pH of membrane fusion. Nature 371, 37–43. 22. Rey, F.A., Heinz, F.X., Mandl, C., Kunz, C. & Harrison, S.C. (1995). The envelope glycoprotein from tick-borne encephalitis virus at 2.0 Å resolution. Nature 371, 291–298. 23. Mason, P., Reider, E. & Baxt, B. (1994). RGD sequence of foot-and- mouth disease virus is essential for infecting cells via the natural receptor but can be bypassed by an antibody-dependant enhancement pathway. Proc. Natl. Acad. Sci. USA 91, 1932–1936. 24. Zlotnick, A., et al., & Johnson, J.E. (1994). Capsid assembly in a family of animal viruses primes an autoproteolytic maturation that depends on a single aspartic acid residue. J. Biol. Chem. 269, 13680–13684. 25. Gillies, S., Bullivant, S. & Bellamy, A.R. (1971). Viral RNA polymerases: electron microscopy of reovirus reaction cores. Science 174, 694–696. 26. Yazaki, K. & Miura, K. (1980). Relation of the structure of cytoplasmic polyhedrosis virus and the synthesis of its messenger RNA. Virology 105, 467–479. 27. Oldfield, S., Adachi, A., Urakawa, T., Hirasawa, T. & Roy, P. (1990). Purification and characterization of the major group-specific core antigen VP7 of bluetongue virus synthesized by a recombinant baculovirus. J. Gen. Virol. 71, 2649–2656. 28. Basak, A.K., Stuart, D.I. & Roy, P. (1992). Preliminary crystallographic study of bluetongue virus capsid protein, VP7. J. Mol. Biol. 228, 687–689. 29. Howard, A.J., Gilliand, G.L., Finzel, B.C., Poulos, T.L., Ohlendorf, D.H. & Salemme, F.R. (1987). The use of an imaging proportional counter in macromolecular crystallography. J. Appl. Cryst. 20, 383–387. 30. Wang, B.C. (1985). Resolution of phase ambiguity in macromolecular crystallography. Methods Enzymol. 115, 90–117. 31. Stuart, D.I. & Jones, E.Y. (1991). A modification of the Wang algorithm for the automatic determination of a molecular envelope. Acta Cryst. A 47, 753–770. 32. Brünger, A.T., Kuriyan, J. & Karplus, M. (1987). Crystallographic R- factor refinement by molecular dynamics. Science 235, 458–460. 33. Engh, R.A. & Huber, R. (1991). Accurate bond and angle parameters for X-ray protein structure refinement. Acta Cryst. A 47, 392–400. 34. Collaborative Computational Project Number 4. (1994). The CCP4 suite: programs for protein crystallography. Acta Cryst. D 50, 760–763. 35. Jones, E.Y. & Stuart, D.I. (1991). Locating heavy atom sites by automatic Patterson search — GROPAT. In Proceedings of the CCP4 Study Weekend: Isomorphous Replacement and Anomolous Scattering. (Wolf, W., Evans, P.R. & Leslie, A.G.W., eds), pp. 39–48. 36. Esnouf, R. (1997). Polyalanine reconstruction from Cα positions using the program CALPHA can aid initial phasing of data by molecular replacement procedures. Acta Cryst. D, in press. 37. Jones, T.A., Zou, Y.J., Cowan, S.W. & Kjeldgaard, M. (1991). Improved