Bivalent binding of a neutralising antibody to a calicivirus involves the torsional flexibility of the antibody hinge1

J. Mol. Biol. (1997) 270, 238±246

Bivalent Binding of a Neutralising Antibody to a Calicivirus Involves the Torsional Flexibility of the Antibody Hinge Eric Thouvenin1, Sylvie Laurent2, Marie-FrancËoise Madelaine2 Denis Rasschaert2, Jean-FrancËois Vautherot2 and Elizabeth A. Hewat1* 1

Institut de Biologie Structurale Jean-Pierre Ebel, 41 av. des Martyrs, 38027, Grenoble France 2

Unite de Virologie et d'Immunologie MoleÂculaires INRA, 78350, Jouy en Josas France

The structure of a complex between rabbit haemorrhagic disease virus (RHDV) virus-like particles (VLPs) and a neutralising monoclonal antibody mAb-E3 has been determined at low resolution by cryo-electron microscopy and three-dimensional (3-D) reconstruction techniques. The atomic co-ordinates of an Fab were ®tted to the cryo-electron microscope density map to produce a binding model. The VLP has a T ˆ 3 icosahedral lattice consisting of a hollow spherical shell with 90 protruding arches. Each dimeric arch presents two mAb binding sites; however, steric hindrance between the variable domains of the Fabs prevents the occupation of both sites simultaneously. Thus the maximum mAb occupation is 50%. Once a mAb is bound to one site it may bind to either of two neighbouring sites related by a local 3-fold axis. The mAbs are bound bivalently on epitopes not related by a 2-fold symmetry axis. This binding geometry implies a torsional ¯exibility of the mAb hinge region, involving a 60 rotation of one Fab arm with respect to the other. Owing to extreme ¯exibility of the hinge region, the Fc domains occupy random orientations and are not visible in the reconstruction. The bivalent attachment of mAb-E3 to RHDV suggests that the neutralisation mechanism(s) involves inhibition of viral decapsidation and/or the inhibition of binding to the receptor. # 1997 Academic Press Limited

*Corresponding author

Keywords: antibody ¯exibility; cryo-electron microscopy; mAb-E3/RHDV; virus neutralisation

Introduction The family Caliciviridae contains only one genus, the caliviviruses. Members include human pathogens such as Norwalk virus, and animal pathogens such as primate and feline calicivirus and rabbit haemorrhagic disease virus (RHDV). Their genome consists of close to 7.5 kb of positive sense single-stranded (ss)RNA and is characterised by two to three open reading frames. In common with many plant viruses, but unlike most animal viruses, caliciviruses possess a single capsid protein of about 60 kDa. One hundred and eighty Present address: S. Laurent, D. Rasschaert and J.-F. Vautherot, Laboratoire de Virologie MoleÂculaire, Unite de Pathologie, Aviaire et Parasitologie, INRA de Tours, 37380 Nouzilly, France. Abbreviations used: RHDV, rabbit haemorrhagic disease virus; VLP, virus-like particle; mAb, monoclonal antibody; ss, single-stranded. 0022±2836/97/270238±09 $25.00/0/mb971095

capsid proteins associate to form the capsid ranging from 35 to 40 nm. in diameter. The structures of two caliciviruses determined by cryo-electron microscopy and image analysis, i.e. Norwalk (Prasad et al., 1994a) and primate calicivirus (Prasad et al., 1994b), are known. They exhibit a T ˆ 3 icosahedral lattice composed of 90 arch-like capsomers formed by dimers of the capsid protein. These structures show distinct similarity to ssRNA plant viruses such as tomato bushy stunt virus (Harrison et al., 1978; Prasad et al., 1994a). RHDV causes a highly contagious disease in rabbits and is thus an economically important pathogen for commercial rabbit production. Contaminated rabbits usually die within two to three days of infection. As for many caliciviruses, cell culture conditions for growing RHDV have not been discovered. Thus cloning of the genome and expression in the baculovirus system have opened new possibilities for the study of many aspects of these viruses (Laurent et al., 1994; Nagesha et al., # 1997 Academic Press Limited

RHDV/mAb-E3 Structure

1995). Also the virus-like particle (VLP), which self-assembles on expression of RHDV capsid viral protein (VP60), has proved to be an effective vaccine giving complete protection in 15 days (Laurent et al., 1994). In all tests done to date, the RHDV VLP appears structurally and immunologically identical to the native virus particle. A panel of monoclonal antibodies elicited against the native particle also recognise the VLP. The combination of cryo-electron microscopy and X-ray crystallographic data has recently proved very useful for the study of the interactions of neutralising monoclonal antibodies (mAbs) with several viruses (e.g. see Porta et al., 1994; Smith et al., 1993a,b; Hewat & Blaas 1996). These structural studies provide information not only about the location of different viral proteins (Prasad et al., 1990; Trus et al., 1992) and the neutralisation of viruses by mAbs, but also about the interactions and properties of mAbs in general, their ¯exibility and modes of attachment. For example, it was recently shown that the spanning distance for bivaÊ lent mAb attachment can be as large as 140 A Ê (Hewat & (Smith et al., 1993a,b) or as little as 60 A Blaas, 1996). Several examples of mAbs bound monovalently to a virus have also been studied (Wang et al. 1992; Wikoff et al., 1994; Hewat et al., 1997), but except in one case (Porta et al., 1994) only the virus/Fab complex was studied, since the aggregation of the virus by monovalently bound mAbs makes specimen preparation very dif®cult. Monovalent binding of mAbs to viruses occurs when the orientation and/or position of neighbouring Fabs does not allow both Fabs from one mAb to bind simultaneously despite the mAb ¯exibility. It is notable that a mAb that binds monovalently to two viruses and thus aggregates the virus particles demonstrates considerable ¯exibility of the hinge region, particularly since the viruses involved will not generally be related by symmetry. In this study we used cryo-electron microscopy and three-dimensional reconstruction techniques to determine the structure of a neutralising mAb-E3 complexed to RHDV VLPs. The shape and density distribution of the bound mAb are much more complicated than has previously been reported for a virus/antibody complex. Hence, modelling the mAb binding using the X-ray crystallographic structure of an Fab was necessary to interpret the cryo-electron microscope data. It appears that each mAb may be attached to any pair of the three neighbouring binding sites, and only 50% mAb occupancy is achieved, due to steric hindrance between neighbouring mAbs.

Results and Discussion Electron microscopy of RHDV VLPs and VLP/ mAb-E3 complexes Electron microscope images of RHDV VLPs in negative stain and in frozen hydrated thin ®lms (Figure 1a and b) show the basic similarity of the

239

Figure 1. Electron micrographs of RHDV VLPs stained with uranyl acetate (a), frozen hydrated RHDV VLP (b) and frozen-hydrated RHDV VLP/mAb-E3 complex (c). Ê. The scale bar represents 1000 A

RHDV capsid to that of other caliciviruses (Prasad et al., 1994a,b). They consist of roughly spherical shells with protruding capsomers. It is remarkable that the stain rarely penetrates the interior of the VLPs (white interior) while the cryo-electron microscope images show that the VLPs are empty. This ``hermetic'' property of RHDV VLP is not a general property of the VLPs of all viruses (Labbe et al., 1991; Hewat et al., 1992a). The species present in suspension are VLPs, very few partially formed

240

RHDV/mAb-E3 Structure

VLPs and many examples of VP60 as small oligomers. The exact state of oligomerisation of VP60 is not known. For reasons we do not understand, attempts to remove the background VP60 using exclusion columns were not always successful, whereas the excess mAb and VP60 was clearly removed from the RHDV VLP/mAb-E3 complex (Figure 1c). In cryo-electron microscope images of the VLP/mAb-E3 complexes, the VLPs appear well decorated with mAb. There is no appreciable aggregation of the VLP/mAb complexes, which implies that the mAb have a bivalent attachment to one virus particle. The RHDV VLP/mAb-E3 reconstructed density The three-dimensional reconstructed density of the VLP/mAb-E3 complex is not as simple to interpret as previously studied virus/mAb complexes. It is necessary to analyse in detail several different representations of the density. The average radial density of the reconstruction (Figure 2a) allows identi®cation of the different constituents of the complex: notably, the spherical shell, the viral arches and the variable and constant domains of the Fab. Density corresponding to the Fc domain is not evident. This is in accordance with previous virus/mAb reconstructions where the Fc domain is either absent or only weakly represented (Smith et al 1993b; Porta et al., 1994; Hewat & Blaas, 1996). The Fc domain is very mobile, ¯exing about the hinge region (Huber et al., 1976; Branden & Tooze, 1991) and hence is lost in the icosahedral averaging of the reconstruction. Sections through the reconstructed density are particularly informative (Figure 2b). The maximum densities in the viral shell and arches are essentially the same as each other. The maximum density in the regions thought to be the variable domains is 50% of the viral density and in the constant domains is only 40%. As the variable domains of the mAbs are bound to the viral surface at well de®ned positions, it follows that the 50% density implies a 50% occupancy of the mAb rather than movement of these domains. However, the even lower density of the constant domains must be explained by movement or multiple positions of these domains. This is discussed further in Materials and Methods. As for the reconstruction of the RHDV VLP alone (data not shown), the capsid shell has a dome shaped-protuberance on each of the 5-fold axes. This was not seen on the previous reconstructions of Norwalk and primate calicivirus. The variation in density of the different regions presents a problem for the representation of the data as an isodensity surface, since one threshold will not produce a correct representation of all domains. An isodensity surface representation of the capsid only (Figure 3a) was extracted from the icosahedrally symmetrised reconstructed density by setting to zero all density beyond a radius of Ê . The projecting arches on all the local and 205 A icosahedral 2-fold axes are visible. A local 3-fold

Figure 2. a, Spherically averaged density distribution of the reconstructed VLP/mAb-E3 complex. b, A section of the reconstructed VLP/mAb-E3 density. Protein density is shown in black. This section does not pass through the origin. Domes on the 5-fold axis are visible. The VLP spherical shell, the arches, the variable and constant domains are indicated by A, B, C and D at 133, Ê , respectively. The scale bar rep180, 234 and 268 A Ê. resents 50 A

axis of the icosahedral T ˆ 3 lattice is marked, and it is evident that the local 3-fold symmetry of the capsid is not perfect. This is associated with the dome on the icosahedral 5-fold axes. (Note: the local 3-fold axes of the icosahedral T ˆ 3 lattice, which do not lie on the icosahedral 3-fold axes, do not have 3-fold symmetry imposed by icosahedral symmetrisation. They are the icosahedral equivalent of non-crystallographic symmetry.) The isodensity surface representation of the capsid plus Fab variable domains only (Figure 3b) shows two variable domains bound to each dimeric arch. The asymmetric unit of the T ˆ 3 lattice consists of three monomers, and hence there are essentially three non equivalent binding sites labelled

RHDV/mAb-E3 Structure

241 I, II and III (Figure 4a). In the isodensity surface representation of the complete reconstruction (Figure 3c), the rather low density corresponding to the constant Fab domains is seen to deviate from the anticipated 3-fold symmetry on the local 3-fold axis.

Modelling the mAb-E3 binding on the VLP: fit of an X-ray crystallographic Fab structure into the cryo-electron microscopy density map

Figure 3. Isosurface representations of the reconstructed RHDV VLP/ mAb-E3 complex viewed down a 2-fold Ê is set to axis. a, The VLP only. Density for R > 205 A zero. b, The VLP plus Fab variable domains only (high threshold). c, The VLP plus Fab variable and constant domains (low threshold). The VLP is coloured blue Ê ) and the mAb-E3 is coloured grey (radius <205 A Ê (radius >205 A). There is no visible density that could

Modelling the mAb-E3 binding on the VLP must take into account the available data; namely: (1) bivalent binding of the mAb; (2) 50% occupancy of the mAb binding sites; (3) the observed reduction in density corresponding to the constant domains; and (4) its irregular diffuse distribution around the local 3-fold axis. The only model that takes into account the above criteria is shown in Figure 4. The ®tting of the variable domain density is relatively easy, since these positions are quite well de®ned (Figure 4b). The variable domains are related by a local 3-fold axis. Presumably when one Fab is attached to one of the binding sites, the other Fab may then bind to either of the other two neighbouring sites (Figure 4b). Thus, for each binding site there are two Fab positions. For example, the ``green'' Fab may be linked either to the ``red'' Fab or to the ``blue'' Fab. These different positions are possible because of the ¯exibility of the elbow and hinge regions of the mAb. The variation in position occurs largely in the constant domains through ¯exing of the elbow. This accounts for the observed diffuse low density corresponding to these domains. The dome on the 5-fold axis of the capsid distorts the local 3-fold symmetry. The deviation is small at the level of the capsid surface but is ampli®ed with distance from the shell and thus explains the large deviation from 3-fold symmetry of the constant domain density. The orientation of the Fabs (180 ) about the 2-fold axis that relates heavy and light chains (Figure 5) is determined by the need to place the C-terminal ends of the heavy chains in close proximity in pairs to form a mAb. In our model each Ê of pair of C-terminal ends are placed within 25 A each other. Alternative models in which the Fabs Á about the 2-fold axis and the are rotated by 180A mAbs are monovalently bound to the capsid have been discarded. The absolute hand of the reconstruction has not been determined. However, the ®t of the Fabs in both enantiomorphs was tested and we present here the enantiomorph that gave the best visual ®t.

be attributed to the Fc fragment. Only the front half of the virus/mAb complex is displayed. The scale bar Ê. represents 50 A

242

RHDV/mAb-E3 Structure

Figure 4. Visual ®t of the X-ray crystallographic structure of an Fab in the cryo-electron microscope reconstruction of the RHDV VLP/mAb-E3 complex. The electron microscope density map is depicted in blue. The Fab Ca chains in red and magenta are bound to type I sites, the Fabs represented in light and dark blue are bound to type III sites and the Fabs represented in light and dark green are bound to type II sites. a, b, c and d, Thick sections viewed down a local 3-fold axis. a is a section through the viral arches. The icosahedral symmetry axes are marked 2, 3 and 5 and the local 2 and 3-fold axes are marked with symbols. The type I, II and III epitopes are indicated. b, A thick section through the Fab variable domains. c and d, Thick sections through the constant domains. There are two Fab positions per binding site. e, f and g show thick sections of the three possible bound mAbs. Just below the Fabs the arches of the VLP are visible and at the bottom of each of e, f and g, part of the spherical shell of the VLP is visible. The secÊ thick. The scale bar represents 50 A Ê. tions are between 30 and 40 A

RHDV/mAb-E3 Structure

243

Figure 5. Schematic drawing of an antibody depicting the torsional ¯exibility of the hinge invoked in the current model. Regions of the heavy chain are shaded in grey and the light chain is white. The 2-fold axis of the whole antibody and the long pseudo-2-fold axes of the Fabs are marked.

Steric hindrance of neighbouring mAbs Each dimeric arch presents two antigenic sites; however, modelling with two Fabs bound to one arch reveals a slight overlap of the variable domains of the Fabs. Thus, steric hindrance will prevent occupation of both sites simultaneously (Figure 6a). Once one site is occupied the second is effectively masked. This explains the 50% density of the variable domains compared with 100% for the capsid. Since each mAb has a multiple choice for binding there exist a number of different binding patterns. There are a maximum of 45 mAbs bound to each capsid. One of the many possible arrangements of bound mAbs is shown in Figure 6b. In view of the arbitrary nature of the mAb binding pattern, it is possible that some mAbs have a monovalent attachment. However, since there is no appreciable aggregation of the VLP/mAb complexes, this cannot be very common. Each antigenic site may include residues from the VP60 dimer or from just one VP60 molecule. We cannot exclude either possibility at present. Icosahedral reconstruction with partial occupation of mAbs The 50% occupancy of mAbs, the two possible positions for each bound Fab and the different dis-

Figure 6. a, Schematic diagram of the E3 epitopes on the VLP. The icosahedral axes are marked 2, 3 and 5 and the local 2 and 3-fold axes are marked with symbols. The type I, II and III epitopes are indicated. Possible pairs of binding sites for one mAb are shown linked by dots, dashes or continuous lines . b, Schematic diagram of one of the many possible distributions of mAb-E3 on the VLP.

tributions of the mAbs on each viral particle are all factors that tend to make the icosahedral reconstruction more dif®cult and less precise. The threedimensional reconstruction of an icosahedral virus particle requires determination of the orientation of each particle. This orientation determination is limited by the degree of icosahedral symmetry displayed by the particle. Thus, partial occupancy of Fabs, which reduces the icosahedral symmetry, renders the icosahedral orientation determination less precise. The clearly de®ned icosahedral structure of the capsid alone must be one of the factors that makes this reconstruction possible. The icosahedral symmetry of the particles is thus suf®ciently high to allow determination of particle orientations. The 47 particles included in the ®nal reconstruction are well distributed in orientation over the unit triangle (Figure 7). There is often a preferred orientation for virus/antibody complexes because of a speci®c interaction between the complex and

244

RHDV/mAb-E3 Structure

mAb-E3 bound to RHDV lies between these extremes. Neutralisation of RHDV by the mAb-E3 Data on the precise mechanism(s) by which the mAbs neutralise the calicivirus are still lacking. The bivalent attachment of mAb-E3 suggests that neutralisation could result from the inhibition of viral decapsidation due to cross-linking of the capsid and/or from an impaired binding of the virus to its cell receptor. As for several other mAbs, E3 was found to inhibit the binding of calicivirus to a cell-membrane component on the human group O red blood cells. It was proposed both for the mAb17-1A bound to HRV14 (Smith et al., 1993a) and the mAb-8F5 bound to HRV2 (Hewat & Blaas, 1996) that cross-linking of the capsid pentamers inhibits virus decapsidation. Neutralisation by virus aggregation is only possible with a monovalently binding mAb and hence is not an important factor in this case. Other mechanisms may be involved, for example conformational modi®cation of the capsid protein. However, at the limited resolution of the current reconstruction no major modi®cations of the capsid are visible. Figure 7. Re®ned orientations of # and f for the 47 VLP/mAb-E3 particles used in the ®nal reconstruction.

the air/water interface. Here the lack of preferred orientation is probably due, not to lack of a speci®c interaction, but to the large number of possible con®gurations of the mAb/virus complex (see Figure 6b for one example).

Implications of this model for mAb flexibility In the binding model presented here the mAbs have a bivalent attachment but are not located on a 2-fold axis. This has not been seen before in virus/mAb complexes. All previously studied virus/mAb complexes were either monovalent or bivalent across a 2-fold axis. The current model implies a torsional ¯exibility of the mAb hinge region allowing one Fab arm to rotate by 60 with respect to the other about the 2-fold axis of the mAb (Figure 5). It also involves ¯exing and slight torsion of the Fab elbow regions. This high degree of mAb ¯exibility is in agreement with their role in binding to target antigens as seen in previous electron microscopic studies of virus/mAb complexes and, for example, in the case of an mAb directed against scorpion hemocyanin (Wade et al., 1989). The observed separation between binding sites of a Ê bivalently bound mAb ranges from about 140 A for the mAb-17-1A bound to HRV14 (Smith et al., Ê for the mAb-8F5 bound to HRV2 1993a) to 60 A (Hewat & Blaas, 1996). The separation for the

Materials and Methods Preparation and purification of RHDV VLPs and mAb-E3 RHDV VLPs were prepared as described (Laurent et al., 1994). mAb-E3 was selected from a panel of 28 different antiVp60 mAbs resulting from one fusion experiment. For that fusion, an eight-week old Balb/C mouse was immunised by intrapodal injection of 10 mg of puri®ed RHDV emulsi®ed in Freund's complete adjuvant. After 12 days the mouse was boost injected with the same antigen in the absence of adjuvant. Three days later the mouse was euthanised and the popliteal lymph nodes were collected. The lymphocytes were collected by perfusing the lymph node with RPMI 1640 medium (GIBCO-BRL France). The cells were washed once in the same medium at 1000 rpm for ®ve minutes in a Jouan B 3-11 centrifuge and mixed with Sp2O/Ag14 myeloma cells (a kind gift from Dr A. Roseto, INSERM) at a ratio of 1:5. The cells were then collected by centrifugation at 800 rpm in a Jouan B 3-11 centrifuge and the fusion was performed as previously described (Vautherot et al., 1992). Screening for hybridomas secreting anti-RHDV mAbs was performed by ELISA with puri®ed virus either coated directly on the solid phase or captured by polyclonal anti-RHDV antibodies from a convalescent animal that had survived a virulent RHDV challenge. The serum was collected 15 days post-infection. The subsequent steps of cloning and ascites ¯uid production were performed by standard procedures. mAb-E3 was puri®ed from crude ascites ¯uid by salting out with sodium sulphate,18 % (®nal) and then 12% ®nal) in a PBS (phosphate buffered saline) solution, followed by af®nity chromatography on Protein A/Sepharose (Pierce; Immunopure IgG puri®cation kit, ref. 44902). Neutralisation assay: owing to the absence of a cell system supporting the replication of RHDV, an in vivo

245

RHDV/mAb-E3 Structure test was developed. The challenge dose (1000 LD50) was diluted 1:1 with the ascites ¯uid or the serum to be tested, incubated for one hour at 37 C and injected into seronegative NZW rabbits maintained in an isolation unit. mAb-E3 was selected as the mAb that showed the same neutralising activity as a polyclonal hyperimmune antiserum ( data not shown). Preparation of RHDV VLP/mAb-E3 complexes Complexes of VLP and mAb-E3 were prepared essentially as described (Hewat & Blaas 1996). VLPs (44 mg) and mAb-E3 were incubated at a molar ratio of 1:170 in a total volume of 60 ml for 45 minutes at room temperature in the elution buffer of the mAb puri®cation kit from Pierce. Excess mAb was removed by passage through an exclusion column ( SeÂphacryl S300 Spun column from Pharmacia) that had been equilibrated with incubation buffer. In principle, the concentration of the complex was about 1 mg/ml and no additional concentration was necessary. Electron microscopy Negatively stained specimens of VLP were prepared in uranyl acetate by standard techniques (Hewat et al., 1992a) and observed at 80 kV in a ZEISS 10C electron microscope. Frozen-hydrated specimens were prepared on holey carbon grids as described (Hewat et al., 1992b). The holey carbon ®lms supported on 400 mesh grids were not glow discharged before use. Samples of the virus or complex suspension (4 ml) were applied to grids, blotted immediately with ®lter paper for one to two seconds and plunged into liquid ethane cooled by nitrogen gas at ÿ175 C. Specimens were maintained at a temperature close to ÿ180 C using a Gatan 626 cryo-holder in a Philips CM200 operating at 200 kV. Images were recorded Ê 2) at a nominal under low-dose conditions (<10 eÿ/A magni®cation of 27,500 times at 2.5 and 3.5 mm underfocus. The magni®cation was calibrated using tobacco mosaic virus in an independent experiment. Image analysis Preliminary selection of micrographs, digitisation and preparation of virus particle images for analysis were performed as described (Schoehn et al., 1996). Only isolated complex particles were selected. The Optronics microdensitometer pixel size of 12.5 mm on the micrograph Ê /pixel at corresponds to a nominal pixel size of 4.55 A the specimen. Images were reinterpolated to a pixel size of 1.5 times the original so that 128  128 ®les could be used. Further image analysis was performed on a DEC Alpha using modi®ed versions of the MRC icosahedral programs and model-based programs (Baker & Cheng, 1996) supplied by S. Fuller (Fuller, 1987; Fuller et al., 1996). The method of common lines (Crowther et al., 1970a,b; Crowther, 1971) was used for the determination of particle origins and orientations in the high defocus image only. The model-based programmes were used for all subsequent orientation and origin determinations. Particle orientations and origins were re®ned against each other using cross-common lines (Simplex program), and the 47 particles with the best phase residual were retained from an image at 2.8 mm defocus to give a reÊ resolution. construction including information to 32 A For this data set all inverse eigenvalues were less than

Ê ÿ1. 0.1, and the phase residual went to 90 at about 32 A The data set was divided into two groups to give two independent reconstructions. The radial cross correlation coef®cient (Saxton & Baumeister 1982) between the Fourier transforms of these two reconstructions also beÊ ÿ1. came, zero at about 32 A Isosurface representations of the reconstructed density were visualised using the program Explorer on a SGI computer. Density sections were visualised using the program Semper 6 Plus on a SGI. Modelling partial occupation of Fab sites Unexpectedly low density in the icosahedral reconstruction of a protein domain may be attributed either to partial occupation of the site and/or variation in position of the domain either dynamic or static. The low density in the reconstructed regions attributed to the Fab variable domains raises the question: to what extent is the reconstructed density map directly related to the partial occupancy and/or movement of the domain? A hollow sphere model decorated with small spheres in an icosahedral (general position) arrangement was used to examine the correlation between density and occupancy. Various proportions of the small spheres were removed in a random manner to represent partial occupancy. For conditions similar to those used for the VLP/mAb-E3 reconstruction (i.e. similar size for the domains and similar resolution) there is a direct correlation between the occupancy and the reconstructed density. Fitting an Fab X-ray structure to the cryo-electron microscope reconstructed density The X-ray structure of the Fab E3 is not known at present, so that of the Fab 8F5 (Tormo et al., 1994) was used instead. The Fab structure was ®tted by eye to the cryoelectron microscope density map using the program O on a SGI. No correction was made for the effect of the contrast transfer function on the electron microscope reconstruction.

Acknowledgements We thank J.-P. Eynard and F. Metoz for assistance in running the computers, S. D. Fuller for supplying his latest versions of the MRC icosahedral programs and R. H. Wade for support.

References Baker, T. & Cheng, R. H. (1996). A model-based approach for determining orientation of biological macromolecules imaged by cryoelectron microscopy. J Struct. Biol. 116, 120± 130. Branden, C. & Tooze, J. (1991). Introduction to Protein Structure, pp. 193± 194, Garland Publishing, Inc., New York and London. Crowther, R. A. (1971). Procedures for three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs. Phil. Trans. Roy. Soc. ser. B. 261, 221± 230. Crowther, R. A., Amos, L. A., Finch, J. T., De Rosier, D. J. & Klug, A. (1970a). Three-dimensional reconstruction of spherical viruses by Fourier synthesis from electron micrographs. Nature, 226, 421± 425.

246

RHDV/mAb-E3 Structure

Crowther, R. A., De Rosier, D. J. & Klug, A. (1970b). The reconstruction of three-dimensional structure from projections and its application to electron microscopy. Phil. Trans. Roy. Soc. ser. A, 317, 319 ± 340. Fuller, S. D. (1987). The T ˆ 4 envelope of Sindbis virus is organised by complementary interactions with a T ˆ 3 icosahedral capsid. Cell, 48, 923 ±934. Fuller, S. D., Butcher, S. J., Cheng, R. H. & Baker, T. S. (1996). Three-dimensional reconstruction of icosahedral particles. The uncommon line. J. Struct. Biol. 116, 48 ± 55. Harrison, S. C., Olson, A., Schutt, C. E., Winkler, F. K. & Bricogne, G. (1978). Tomato bushy stunt virus at Ê resolution. Nature, 276, 368 ± 373. 2.9 A Hewat, E. A. & Blaas, D. (1996). Structure of a neutralizing antibody bound bivalently to human rhinovirus 2. EMBO J, 15, 1515± 1523. Hewat, E. A., Booth, T. F., Loudon, P. T. & Roy, P. (1992a). Three dimensional reconstruction of baculovirus expressed bluetongue virus core-like particles by cryo-electron microscopy. Virology, 189, 10 ± 20. Hewat, E. A., Booth, T. F. & Roy, P. (1992b). Structure of bluetongue virus particles by cryoelectron microscopy. J. Struct. Biol. 109, 61 ± 69. Hewat, E. A., Verdaguer, N., Fita, I., Blakemore, W., Brooks, S., King, A., Newman, J., Domingo, E., Mateu, M. G. & Stuart, D. (1997). Structure of the complex of an Fab fragment of a neutralising antibody with foot-and-mouth disease virus: Positioning of a highly mobile antigenic loop. EMBO J. 16, 1492± 1500. Huber, R., Deisenhofer, J., Colman, P. M. & Matsushima, M. (1976). Crystallographic structure studies of an IgG molecule and an Fab fragment. Nature, 264, 415± 420. LabbeÂ, M., Charpilienne, A., Crawford, S. E., Estes, M. K. & Cohen, J. (1991). Expression of rotavirus VP2 produces empty core-like particles. J. Virol. 65, 2946± 2952. Laurent, S., Vautherot, J. F., Madelaine, M. F., Le Gall, G. & Rasschaert, D. (1994). Recombinant rabbit hemorrhagic disease virus capsid protein expressed in baculovirius self-assembles into virus like particles and induces protection. J. Virol. 68, 6794± 6798. Nagesha, H. S., Wang, L. F., Hyatt, A. D., Morissy, C. J., Lenghaus, C. & Westbury, H. A. (1995). Self-assembly, antigenicity, and immunogenicity of the rabbit haemorrhagic disease virus (Czechoslovakian strain V-351) capsid protein expressed in baculovirus. Arch. Virol. 140, 1095± 1108. Porta, C., Wang, G., Cheng, H., Chen, Z., Baker, T. S. & Johnson, J. E. (1994). Direct imaging of interactions between an icosahedral virus and congugate Fab fragments by cryoelectron microscopy and X-ray crystallography. Virology, 204, 777± 788.

Prasad, B. V. V., Burns, J. W., Marietta, E., Estes, M. K. & Chiu, W. (1990). Localization of VP4 neutralization sites in rotavirus by three-dimensional cryo-electron microscopy. Nature, 343, 476 ± 479. Prasad, B. V. V., Rothnagel, R., Jiang, X. & Estes, M. K. (1994a). Three dimensional structure of baculovirusexpressed Norwalk virus capsid. J. Virol. 68, 5117± 5125. Prasad, B. V. V., Matson, D. O. & Smith, A. W. (1994b). Three-dimensional structure of calicivirus. J. Mol. Biol. 240, 256± 264. Saxton, W. O. & Baumeister, W. (1982). The correlation averaging of a regularly arranged bacterial envelope protein. J. Microscopy, 127, 127 ±138. Schoehn, G., Fender, P., Chroboczek, J. & Hewat, E. A. (1996). Adenovirus 3 penton dodecahedron exhibits structural changes of the base on ®bre binding. EMBO J. 15, 6841± 6846. Smith, T. J., Olson, N. H., Cheng, R. H., Chase, E. S. & Baker, T. S. (1993a). Structure of a human rhinovirus -bivalently bound antibody complex: Implications for viral neutralization and antibody ¯exibility. Proc. Natl Acad. Sci. USA, 90, 7015± 7018. Smith, T. J., Olson, N. H., Cheng, R. H., Liu, H., Chase, E. S., Lee, W. M., Leippe, D. M., Mosser, A. G., Rueckert, R. R. & Baker, T. S. (1993b). Structure of human rhinovirus complexed with Fab fragments from a neutralizing antibody. J. Virol. 67, 1148± 1158. Tormo, J., Blaas, D., Parry, N. R., Rowlands, D., Stuart, D. & Fita, I. (1994). Crystal structure of a human rhinovirus neutralizing antibody complexed with a peptide derived from viral capsid protein. EMBO J. 13, 2247± 2256. Trus, B. L., Newcomb, W. W., Booy, F. P., Brown, J. C. & Steven, A. C. (1992). Distinct monoclonal antibodies separately label the hexons or the pentons of herpes simplex virus capsid. Proc. Natl Acad. Sci. USA, 89, 11508± 11512. Vautherot, J. F., Madelaine, M. F., Boireau, P. & Laporte, J. (1992). Bovine coronavirus peplomer glycoproteins : detailed antigenic analyses of S1, S2, and HE. J. Gen. Virol. 73, 1725± 1737. Wade, R. H., Taveau, J. C. & Lamy, J. N. (1989). Concerning the axial rotational ¯exibility of IgG molecules. J. Mol. Biol. 206, 349 ± 356. Wang, G. J., Porta, C., Chen, Z. G., Baker, T. S. & Johnson, J. E. (1992). Identi®cation of a Fab interaction footprint site on an icosahedral virus by cryoelectron microscopy and X-ray crystallography. Nature, 355, 275 ± 278. Wikoff, W. R., Wang, G. J., Parrish, C. R., Cheng, R. H., Strassheim, M. L., Baker, T. S. & Rossmann, M. G. (1994). The structure of a neutralized virus: canine parvovirus complexed with neutralizing antibody fragment. Structure, 2, 595 ± 607.

Edited by I. A. Wilson (Received 7 October 1996; received in revised form 27 March 1997; accepted 10 April 1997)