Electron Cryo-Microscopy and Single-Particle Averaging of Rift Valley Fever Virus: Evidence for G N -G C Glycoprotein Heterodimers

JOURNAL OF VIROLOGY, Apr. 2009, p. 3762–3769 Vol. 83, No. 80022-538X/09/$08.00�0 doi:10.1128/JVI.02483-08Copyright © 2009, American Society for Microbiology. All Rights Reserved.

Electron Cryo-Microscopy and Single-Particle Averaging of Rift ValleyFever Virus: Evidence for GN-GC Glycoprotein Heterodimers�

Juha T. Huiskonen,1,2* Anna K. Overby,3 Friedemann Weber,3 and Kay Grunewald1

Department of Molecular Structural Biology, Max Planck Institute of Biochemistry, Am Klopferspitz 18, D-82152 Martinsried,Germany1; Institute of Biotechnology, University of Helsinki, P.O. Box 65, FI-00014 Helsinki, Finland2; and Department of

Virology, University of Freiburg, Hermann-Herder-Strasse 11, D-79008 Freiburg, Germany3

Received 3 December 2008/Accepted 23 January 2009

Rift Valley fever virus (RVFV) is a member of the genus Phlebovirus within the family Bunyaviridae. It is amosquito-borne zoonotic agent that can cause hemorrhagic fever in humans. The enveloped RVFV virions areknown to be covered by capsomers of the glycoproteins GN and GC, organized on a T�12 icosahedral lattice.However, the structural units forming the RVFV capsomers have not been determined. Conflicting biochemicalresults for another phlebovirus (Uukuniemi virus) have indicated the existence of either GN and GC ho-modimers or GN-GC heterodimers in virions. Here, we have studied the structure of RVFV using electroncryo-microscopy combined with three-dimensional reconstruction and single-particle averaging. The recon-struction at 2.2-nm resolution revealed the organization of the glycoprotein shell, the lipid bilayer, and a layerof ribonucleoprotein (RNP). Five- and six-coordinated capsomers are formed by the same basic structural unit.Molecular-mass measurements suggest a GN-GC heterodimer as the most likely candidate for this structuralunit. Both leaflets of the lipid bilayer were discernible, and the glycoprotein transmembrane densities were seento modulate the curvature of the lipid bilayer. RNP densities were situated directly underneath the transmem-brane densities, suggesting an interaction between the glycoprotein cytoplasmic tails and the RNPs. Thesuccess of the single-particle averaging approach taken in this study suggests that it is applicable in the studyof other phleboviruses, as well, enabling higher-resolution description of these medically important pathogens.

Rift Valley fever virus (RVFV) is a mosquito-transmittedvirus causing diseases in humans ranging from mild flu-likesymptoms to severe hemorrhagic fever (12). Since its first iso-lation in 1930 (9), several large outbreaks have been identifiedin Sub-Saharan Africa, where it is endemic, and also in neigh-boring regions, such as Egypt and the Arabian Peninsula (2, 5).Humans can acquire RVFV through mosquito bites or expo-sure to infected animal material. In cattle, RVFV causes abor-tions and high levels of neonatal mortality. Thus, RVFV is amedically and economically significant pathogen.

RVFV is a member of the genus Phlebovirus within thefamily Bunyaviridae. This family is divided into five genera:Orthobunyavirus, Phlebovirus, Hantavirus, Nairovirus, and To-spovirus (25). Common to all members of the family Bunya-viridae is the characteristic that the virions possess a single-stranded, tripartite RNA genome core enclosed by a lipidbilayer and a shell of glycoproteins (37). Bunyaviruses replicatein the cytoplasm, and nascent virions assemble by budding intothe lumen of the Golgi apparatus (35). Bunyaviruses encodefour structural proteins from the three genome segments: theL segment encodes the RNA-dependent RNA polymerase (L),the M segment encodes the two membrane glycoproteins (GN

and GC), and the S segment encodes the nucleocapsid protein(N) (37). The N protein coats the genome segments to formribonucleoproteins (RNPs). The two glycoproteins are trans-lated as a polyprotein, which is cleaved in the endoplasmic

reticulum. Formation of a heterodimeric complex of the gly-coproteins is required for targeting to the Golgi apparatus,since only GN has the Golgi localization signal (7, 14). GN andGC are type I transmembrane proteins. They both traverse thelipid bilayer, and the C-terminal tails are internal to the virion.Unlike other negative-stranded RNA viruses, bunyaviruseslack a matrix protein (37). Thus, the glycoprotein C-terminaltails are thought to interact directly with the RNPs duringbudding to ensure the proper packaging of the RNPs (29, 30).In RVFV, the M segment encodes two proteins in addition tothe two glycoproteins GN (54 kDa) and GC (56 kDa): thenonstructural protein NSm and an additional 78-kDa protein,which is a minor component of the virion (13).

The glycoproteins GN and GC have been shown to formheterodimers in cells infected with RVFV (13) and with otherphleboviruses, such as Uukuniemi virus (UUKV) (32) andPunta Toro virus (6). However, as reassociations of the glyco-proteins during virus maturation cannot be ruled out, the mul-timeric status of the glycoproteins in the virions is not wellestablished. In fact, there exist conflicting reports aboutisolation of both heterodimers (32) and homodimers (36)from Triton X-100-solubilized UUKV virions. The structureof RVFV virions studied recently using electron cryo-tomog-raphy revealed five- and six-coordinated glycoprotein capsom-ers organized on a T�12 lattice (10). However, the limitedresolution (6.1 nm) impeded determination of the composi-tions of the capsomers. It remained unclear whether a singlestructural unit composed of both GN and GC forms both thefive- and six-coordinated glycoprotein capsomers or whetherGN alone forms some capsomers and GC alone forms the othercapsomers (10). In addition, the organization of RNPs insidethe virions could not be addressed. Knowledge of the structural

* Corresponding author. Mailing address: Department of MolecularStructural Biology, Max Planck Institute of Biochemistry, Am Klopfer-spitz 18, D-82152 Martinsried, Germany. Phone: 49 89 8578-2632. Fax:49 89 8578-2641. E-mail: [email protected].

� Published ahead of print on 4 February 2009.

3762

on June 26, 2015 by guesthttp://jvi.asm

.org/D

ownloaded from

http://jvi.asm.org/

composition of RVFV glycoprotein capsomers and of the RNPorganization is crucial in understanding the assembly and in-fection mechanisms of this important pathogen.

Here, we have studied the structure of the RVFV virionusing electron cryo-microscopy combined with three-dimen-sional image reconstruction and single-particle averaging. Thestructure was solved to 2.2-nm resolution. The structure re-vealed 110 cylinder-shaped glycoprotein hexamers and 12 pen-tamers organized on an icosahedral T�12 lattice. Only onetype of structural unit was detected in the capsomers. Thisstructural unit most likely corresponds to a GN-GC het-erodimer. Six structural units form the hexamers, and five formthe pentamers. Transmembrane densities, each correspondingmost likely to several glycoprotein transmembrane helices,were seen to modulate the curvature of the bilayer. Further-more, the organization of the membrane-proximal RNP den-sity correlated with the positions of these transmembrane den-sities.

MATERIALS AND METHODS

Cell culture and virus purification. Vero cells were grown in Dulbecco’smodified Eagle’s medium supplemented with 10% fetal calf serum and main-tained in medium with 2% serum. The Vero cells were infected with RVFVstrain Clone 13 at a multiplicity of infection of 0.01, and the supernatant con-taining virions was collected 72 h postinfection. The supernatant was clarified bylow-speed centrifugation (6,000 rpm) for 10 min before the pH 6.0 or 7.4 wasadjusted with phosphate buffer (final concentration, 50 �M). The virus was fixedwith glutaraldehyde (final concentration, 0.5%) and concentrated through asucrose cushion (20% [wt/vol]) in TN buffer (50 mM Tris-HCl, pH 6.0 or 7.4, 100mM NaCl) at 100,000 � g for 1 h. The virus pellet was resuspended in TN buffer.

Electron cryo-microscopy and image processing. A 3-�l aliquot of freshlyprepared virus suspension at pH 6.0 or pH 7.4 was pipetted on a glow-dischargedholey carbon-coated electron microscopy grid (C-flat; Protochips), excess sus-pension was blotted with a filter paper, and the sample was vitrified by plungingit rapidly into liquid ethane. Electron cryo-microscopy was performed using aTecnai F20 microscope (FEI) operated at 200 keV and equipped with a4,000-by-4,000 charge-coupled-device camera (Eagle; FEI). Low-dose images(20 e�/ A2) were taken with SerialEM (23) and 1- to 3-�m underfocus at anominal magnification of �80,000, giving a calibrated pixel size of 0.13 nm.Images were computationally down-sampled by a factor of 3 to give a finalpixel size of 0.40 nm.

Images free from drift and astigmatism were used to extract views for three-dimensional reconstruction of the virion at pH 6.0 and pH 7.4. Virus particleswere automatically localized in the electron micrographs with ETHAN (20), andsubimages of single viral particles were boxed in EMAN (22). One hundred thirtyparticles extracted from 42 micrographs were used for the pH 6.0 reconstructionand 267 particles extracted from 111 micrographs for the pH 7.4 reconstruction.Bsoft was used in further image-processing tasks unless otherwise stated (17).Density values in the images were normalized to a common average value andstandard deviation (�). Contrast transfer function parameters were determined,and images were corrected for the contrast reversals by flipping the phases atappropriate spatial frequencies. Orientations and origins for particles were de-termined using a modified version of the original polar Fourier transform algo-rithm (1). Three-dimensional reconstructions were computed with a Fourier-Bessel algorithm (8, 11). The previously published model of UUKV (28) wasused as a starting model in model-based refinement of particle origins andorientations, and reconstruction processes were iterated until no improvementwas observed in the resolution. Icosahedral symmetry had been detected inRVFV particles previously (10) and was imposed here on the reconstructions.

The resolution of each reconstruction was evaluated using two test reconstruc-tions, both calculated from independent half-sets of particle views to 0.8-nmresolution. Fourier shell correlation was calculated between the glycoproteinshells of the two test reconstructions, and the resolution was determined using athreshold of 0.5. In addition, the radial resolution was determined for 4-nm-thickshells at every 0.4 nm using Fourier shell correlation and the same threshold. Thefinal pH 6.0 and pH 7.4 reconstructions were calculated to 2.3-nm and 2.0-nmresolutions, respectively. A difference map was calculated between the pH 6.0reconstruction and pH 7.4 reconstruction at 3.0-nm resolution.

For comparison, glycoprotein densities external to the membrane weremasked out from the pH 7.4 reconstruction. Glycoprotein cylinders were com-putationally rotated to align their cylindrical axes on the icosahedral twofold axisof symmetry. The rotations required for three glycoprotein cylinders situated onthe icosahedral-symmetry axis were known from the icosahedral symmetry (seeFig. 2 and 5, positions 2, 3, and 4), but the rotation required for the cylindersituated off the symmetry axis (see Fig. 2 and 5, position 1) had to be determined.Density extracted from the twofold position acted as a template (see Fig. 2 and5, position 2) in an orientation search based on cross-correlation of the twovolumes. Three-dimensional visualizations were done in UCSF Chimera (34).

To estimate the molecular masses of the glycoprotein capsomers, the RNP andmembrane were removed from the pH 7.4 reconstruction with a spherical mask(radius, 42 nm). The volume was calculated in EMAN (22). We tested differentthreshold values for the molecular surface (� above the mean density). Theaverage density for a solvated protein (1.35 g/ml) was assumed to calculate themass (24).

Density map accession numbers. The reconstructions have been submitted tothe EM Data Bank with accession codes EMD-1549 (pH 6.0) and EMD-1550(pH 7.4).

RESULTS

Electron microscopy of purified RVFV virions. Icosahedralsymmetry was detected in the outer glycoprotein shells ofUUKV particles in a recent study using electron cryo-tomog-raphy (28). This suggested that other phlebovirus virions mightalso be icosahedrally symmetric, rather than being pleomor-phic, as previously thought. Furthermore, it raised the possi-bility of exploiting a single-particle averaging approach, wheremultiple virions are computationally averaged together andicosahedral symmetry is imposed (1). When applicable, thisapproach is beneficial, since it typically yields a higher resolu-tion than is possible by electron cryo-tomography.

To study whether this averaging approach was suitable forRVFV, we first analyzed the morphology of purified, nega-tively stained RVFV virions using electron microscopy. Theparticles had nearly uniform sizes and a roughly roundshape in projection images (Fig. 1A). More strikingly, someof the RVFV particles displayed clear twofold, threefold,and fivefold symmetric views, arguing for the presence oficosahedral symmetry (Fig. 1B). These results were in agree-ment with a recently published tomographic reconstructionin which icosahedral symmetry was detected (10).

We used electron cryo-microscopy to visualize purified, un-stained RVFV virions (Fig. 1C). Electron cryo-microscopyprojection images revealed round particles of homogeneoussizes and shapes (Fig. 1C). This was in contrast to an earlierelectron cryo-microscopy study of La Crosse bunyavirus, whereround particles with heterogeneous sizes were observed (38).Fixation with 0.5% glutaraldehyde prior to purification wasrequired to reduce excessive aggregation of virus particlesoccurring in the unfixed sample (not shown). Fixation notonly preserves fragile complexes during purification, butalso changes the surface properties of the particles and thusaffects particle aggregation (19). Fixation likely also helpedto preserve RVFV particle morphology in negative staining(Fig. 1A), since unfixed RVFV particles are commonly seento be distorted (10).

Three-dimensional reconstruction of RVFV virions revealeda multilayered structure. We calculated a three-dimensionalreconstruction of the virion by averaging together particle pro-jection images and applying icosahedral symmetry (Fig. 2).Micrographs were recorded in those areas of the sample that

VOL. 83, 2009 RIFT VALLEY FEVER VIRUS STRUCTURE 3763


.org/D

ownloaded from

http://jvi.asm.org/

showed only mild particle aggregation (Fig. 1C). We included267 out of 472 particle images (57%) from 111 micrographs inthe reconstruction. The resolution of the reconstruction was2.2 nm, as determined for the glycoprotein shell by Fouriershell correlation, with a 0.5 cutoff (Fig. 3A). Sample aggrega-tion may distort some of the particles from perfect icosahedralsymmetry, thereby limiting the achievable resolution. Theshape of the virion is nearly spherical, with the diameter vary-ing only slightly from 101 nm to 106 nm, as measured fromicosahedron edge to edge and vertex to vertex, respectively.

The virion surface is entirely covered by a shell of 122 gly-coprotein capsomers. The capsomers resemble hollow cylin-ders when seen from the outside of the particle (Fig. 2A). Thecylinders are situated at the five-coordinated (Fig. 2A, penta-gon) and at three types of six-coordinated (Fig. 2A, positions 1,2, and 3) positions of a T�12 icosahedral lattice, as describedpreviously (10). Twelve cylinders are situated at the fivefold-symmetry axis of the icosahedral lattice (Fig. 2A, pentagon), 60cylinders are off the symmetry axis (Fig. 2A, hexagon 1), 30cylinders are at the twofold-symmetry axis (Fig. 2A, hexagon2), and 20 cylinders are at the threefold-symmetry axis (Fig.2A, hexagon 3). The cylinders at five-coordinated positions areslightly smaller (outer diameter, 14 nm) than the cylinders atthe six-coordinated positions (outer diameter, 16 nm).

Under the glycoprotein shell, two leaflets of the lipid bilayerare clearly discernible (Fig. 2B). Strikingly, the membrane isnot completely round but instead makes kinks toward the gly-coproteins (Fig. 2B). This is likely due to the presence oftransmembrane segments of glycoproteins at these positions,sometimes visible as transmembrane densities (Fig. 2C to E).Furthermore, this is reminiscent of many other membrane-containing viruses, such as dengue virus, in which transmem-brane domains modulate the curvature of the bilayer (18, 21,39). A layer of RNP is located proximal to the inner leaflet ofthe membrane and closely follows the shape of the membrane(Fig. 2B).

We calculated a radially averaged density distribution of theRVFV virion that clearly showed the glycoprotein shell, lipidbilayer, and RNP density (Fig. 3B). The glycoproteins protrude9 nm from the membrane surface and are situated at 42- to51-nm radii. The glycoprotein shell is divided into two peaks at42 to 46 nm and at 47 to 51 nm, reflecting the domain orga-nization of the glycoprotein capsomers (see below). The lipidbilayer is visible at the 35- to 42-nm radii, with the stronglyscattering phospholipid head groups likely causing the twodensity maxima at the 36- and 40-nm radii.

A plot of the resolution against the particle radius revealeda radially anisotropic resolution in the reconstruction (Fig.3B). This was due to variable degrees of order in the differentcomponents of the virion. While the outer part of the recon-struction—the glycoprotein shell and membrane—has thehighest resolution of 2.1 to 2.3 nm, the inner RNP density isunresolved. In conclusion, the RNP density is not icosahedrallyordered. However, the RNP density is not completely evenlydistributed either. Rather, there is a distinct layer proximal tothe membrane at 32- to 34-nm radius (Fig. 3B). This distribu-tion likely reflects interactions between the C-terminal tailsof the glycoproteins and RNPs, as suggested for UUKV (28,29). The distance between this membrane-proximal RNP layerand the membrane is 4.4 nm, as measured from peak to peak.

FIG. 1. Electron microscopy of RVFV. (A) Electron microscopyimage of purified virus particles stained with 2% uranyl acetate.(B) Examples of views along twofold (left), threefold (middle), andfivefold (right) axes of icosahedral symmetry for stained particles.(C) Electron cryo-microscopy image of plunge-frozen, unstained virusparticles. Scale bars, 100 nm.

3764 HUISKONEN ET AL. J. VIROL.


.org/D

ownloaded from

http://jvi.asm.org/

Structures of the glycoprotein capsomers. To compare thestructures of the four different classes of glycoprotein capsom-ers (Fig. 2A, pentagon and hexagons 1, 2, and 3), their respec-tive densities were extracted from the reconstruction (Fig. 4).Reliable demarcation of molecular boundaries between the

FIG. 2. Three-dimensional icosahedral reconstruction of RVFV.(A) Radially colored isosurface representation of the reconstruction.Glycoprotein cylinders are in blue, the glycoprotein base layer is ingreen, and the membrane and RNP density are in brown. One five-coordinated position (pentagon) and three different six-coordinatedpositions (hexagons 1, 2, 3) are indicated. In the lower left part of thepanel, we removed the front half of the reconstruction to reveal thecapsomers from the side (a single capsomer is indicated with a red

oval). The isosurface was rendered at 1.3� above mean density. Scalebar, 50 nm. (B) Central cross section through the density (black, highdensity; white, low density). The indicated radii correspond to theglycoprotein shell, the membrane, and the RNPs. A shell of RNPdensity, closely following the shape of the membrane, is indicated withan asterisk. The membrane is distorted from a round shape, makingkinks toward the bases of glycoprotein cylinders (some indicated witharrows). The same capsomer as in panel A is indicated with a red oval.Scale bar, 50 nm. (C to E) Close-ups of cross sections showing trans-membrane densities (indicated with arrows). Scale bar, 10 nm (C to E).The cross sections in panels B to E are 0.4 nm thick.

FIG. 3. Resolution and radial-density distribution of the RVFVreconstruction. (A) A Fourier shell correlation as a function of spatialfrequency was calculated between two test reconstructions, calculatedfrom two independent half-sets of the data. It shows correlation abovethe 0.5 threshold up to 2.2-nm resolution. (B) Radially averaged den-sity (solid line) and resolution (dotted line) are plotted as a function ofthe radius. The radii corresponding to the glycoprotein shell, the mem-brane, and the RNPs are indicated. The shell of RNP density proximalto the membrane is indicated with an asterisk.



.org/D

ownloaded from

http://jvi.asm.org/

capsomers was not possible due to the limited resolution, andthus, we used a cylindrical mask as an approximation. Densitiesextracted from all three six-coordinated positions were strik-ingly similar to each other, showing the same structural fea-tures arranged in quasi-sixfold symmetry (Fig. 4, hexagons 1, 2,and 3). Furthermore, similarly shaped subunits form the cap-somers at five-coordinated (Fig. 4, pentagon) and at six-coor-dinated (Fig. 4, hexagons 1, 2, and 3) positions.

Organization of the glycoprotein shell. The complex orga-nization of the RVFV glycoprotein shell is shown in Fig. 5.This shell can be divided conceptually into four different lay-ers: (i) glycoprotein cylinders (Fig. 5A, radii of 48 to 50 nm),(ii) intracylinder contacts (Fig. 5A, radii of 45 to 47 nm), (iii)triangular membrane connections (Fig. 5A, radii of 41 to 44nm), and (iv) transmembrane densities (Fig. 5A, radii of 39 to40 nm). In the first, outermost layer, the glycoprotein cylindersat six-coordinated positions (Fig. 5B, hexagons) show a quasi-sixfold symmetry, with a pattern of six two-lobed structuralunits (Fig. 5A, radii of 48 to 50 nm). The cylinders at five-coordinated positions (Fig. 5B, pentagon) are composed of fivesimilar two-lobed structural units. In some, but not in all,quasi-twofold positions, neighboring cylinders are connectedto each other via thin, bridging densities (Fig. 5A, 50 nm). Inthe second layer, the glycoprotein cylinders form extensiveintracylinder contacts at each of the quasi-twofold positions(Fig. 5A, radii of 45 to 47 nm, and B, rectangles). Two spokes,one from each cylinder, bind together in a parallel fashion.In the third layer, the densities from the glycoprotein cylindersand spokes come together from three neighboring capsomers andform triangular densities (Fig. 5A, radii of 41 to 44 nm, and B,triangles). These densities connect the capsomers to the under-lying membrane. In the fourth layer, the triangular membrane-proximal densities split into a punctate pattern of transmem-brane densities (Fig. 5A, radii of 39 to 40 nm) (see below).

The positions of transmembrane densities correlate withmembrane curvature and RNP organization. Triangular den-sities were seen to connect the glycoproteins to the underlyingmembrane (Fig. 5A, radius of 41 nm). Both of the glycopro-teins are expected to contain one most likely alpha-helicaltransmembrane segment (13). Thus, under each of the trian-gular membrane-proximal densities, transmembrane helicesare expected to be present. At 2.2-nm resolution, it was notpossible to identify single helices, which typically have a diam-eter of only 0.7 nm. However, visual inspection of the densitymap revealed several positions in which clear density was con-necting the two leaflets of lipid bilayer (Fig. 2C to E and 5A,radius of 39 nm and inset). These transmembrane densitieslikely correspond to two or more transmembrane helices.

As expected, the transmembrane densities are situated di-rectly underneath the triangular membrane-proximal densities(Fig. 5A, radius of 41 nm). Furthermore, the organization ofthe transmembrane densities correlated with the observedkinks in membrane curvature. The transmembrane densitiescreate a high membrane curvature by pulling the membranetoward the glycoprotein shell (Fig. 2C to E) and thus encirclemembrane areas with a smaller curvature, as shown for thebilayer outer leaflet in Fig. 5 (radius, 40 nm).

We correlated the positions of the transmembrane densitieswith the lateral distribution of membrane-proximal RNP den-sity. The membrane-proximal RNP density is unevenly distrib-uted. Strikingly, the RNP prefers areas directly under thetransmembrane densities (Fig. 5A, radius of 34 nm). Althoughglycoprotein-RNP connections were not resolved here, the co-localization of RNP and glycoprotein transmembrane densitiesargues for a specific interaction between the glycoprotein cy-toplasmic tails and RNP complexes.

Glycoprotein heterodimers form the glycoprotein shell.Given the similarity in the subunits of five- and six-coordinated

FIG. 4. Comparison of the external glycoprotein densities in different positions of the lattice. Shown are top views (top row) and side views(bottom row) of the densities extracted from the three quasi-sixfold positions (hexagons 1, 2, and 3) and from the fivefold position (pentagon). Acylindrical mask extending to halfway between neighboring densities was used in the extraction. The fivefold cylinder is composed of subunitsshaped similarly to subunits of the quasi-sixfold cylinders, suggesting that there is only one type of structural building block in the glycoprotein shell.The isosurface was rendered at 1.3� above mean density and radially colored, as in Fig. 2A. Scale bar, 10 nm.



.org/D

ownloaded from

http://jvi.asm.org/

FIG. 5. Molecular organization of the glycoprotein lattice. (A) A closed surface with a shape between a perfect sphere and an icosahedron wasfitted to traverse glycoprotein cylinders, and the density values were mapped on the surface (black, low density; white, high density). Approximateradii for the surfaces are indicated. The insets show close-up views of one five- and one neighboring six-coordinated position at each radius.Surfaces at radii of 48 to 50 nm show cylinders of glycoproteins at five-coordinated and six-coordinated positions of the T�12 lattice. At radii of45 to 47 nm, quasi-twofold contacts between the capsomers become apparent. At radii of 41 to 44 nm, densities from three neighboring capsomerscome together at quasi-threefold positions to form triangular densities, connecting the capsomers to the underlying membrane. Surfaces at radiiof 39 and 40 nm traverse the outer leaflet of the lipid bilayer and the space between the two leaflets, revealing the punctate pattern oftransmembrane densities. The relative positions of the transmembrane densities are indicated with red for surfaces at radii of 34 to 41 nm tocorrelate their positions with the other features of the reconstruction. One membrane patch of the outer leaflet with a low curvature andsurrounded by the transmembrane densities is indicated with an asterisk at 40 nm. Scale bar, 50 nm. (B) Part of the RVFV T�12 lattice isillustrated as a schematic in the same orientation as in panel A. A five-coordinated cylinder (pentagon) and six-coordinated cylinders (1, 2, and3) are indicated. The positions of some of the quasi-twofold contacts (rectangles) and quasithreefold contacts (triangles) are also indicated.

3767


.org/D

ownloaded from

http://jvi.asm.org/

capsomers (Fig. 4 and 5), we reject the hypothesis that one ofthe glycoproteins alone would form a certain type of capsomer(e.g., the five-coordinated capsomers) and the other glycopro-tein alone would form the rest of the capsomers (e.g., thesix-coordinated capsomers). Rather, there is only one basicstructural unit forming the glycoprotein shell. Six copies of thisunit form each of the 110 six-coordinated capsomers, and fivecopies form each of the 12 five-coordinated capsomers. Thus,the total number of structural units is 720.

The basic structural unit is most likely composed of both GN

and GC, since they are both major components of the virionand are expected to be present in an equimolar ratio, as shownfor UUKV particles (33). It is also likely that both GN and GC

are exposed on the RVFV surface, since monoclonal antibod-ies against both GN and GC can neutralize RVFV infection (3).How many copies of GN and GC contribute, then, to the basicstructural unit? The GN and GC external parts correspond to amolecular mass of �45 kDa and �51 kDa, respectively (4). Toassess how many copies of GN and GC can fit in the recon-struction, we estimated the total mass of the glycoprotein shellfrom the reconstruction (see Materials and Methods). Wecarried out the estimation at different thresholds (� abovethe mean density) for the molecular surface. Commonlyused thresholds for the molecular surface (1.3 to 2.0�) cor-responded to a molecular mass of 68 to 86 MDa. The masscalculated using a much lower threshold (0.8�) was 100MDa. This threshold is at the noise level of the reconstruc-tion and thus provides an upper estimate for the molecularmass. The structural unit of a single GN-GC heterodimerwould give a total calculated molecular mass of 69 MDa(720 heterodimers) for the glycoprotein external parts. Adimer of heterodimers would correspond to 138 MDa (1,440heterodimers), a much greater mass than our upper esti-mate. Thus, our mass measurement is compatible only withthe former assumption, and we conclude that 720 GN-GC

heterodimers form the glycoprotein shell of RVFV.

DISCUSSION

In this study, we have addressed the glycoprotein organiza-tion of the RVFV virion. The virion structure argues againstpossible heterodimer-to-homodimer transitions of viral glyco-proteins during assembly or maturation. Rather, GN and GC

are organized in the virion as heterodimers, just as they areexpressed (13). These heterodimers form pentons and hexonsand assemble on a T�12 icosahedral lattice. However, in theassembled glycoprotein shell, GN-GN and GC-GC contacts arealso likely to occur. For example, at quasi-twofold positions,where the intracylinder contacts occur, homodimeric contactsare present (Fig. 5A and B, rectangles). This could explain theseemingly contradictory results for UUKV, another phlebovi-rus, where both homodimers (36) and heterodimers (32) havebeen extracted from the virions. Extraction of membrane pro-teins from virions could stabilize protein-protein interactionsoccurring only in the assembled state.

Only two members of the family Bunyaviridae, RVFV andUUKV (both members of the genus Phlebovirus), have beenstructurally characterized using electron cryo-microscopy (10,28). While the overall architectures of the two are the same(T�12), the glycoprotein capsomers differ in their morpholo-

gies, RNP interactions, and pH sensitivities. RVFV capsomersresemble hollow cylinders and protrude 9 nm from the mem-brane, whereas UUKV capsomers resemble spikes protruding13 nm from the membrane (28). The glycoproteins of UUKVare considerably larger (GN � 75 kDa; GC � 65 kDa) thanthose of RVFV (GN � 54 kDa; GC � 56 kDa), which probablyexplains the larger glycoprotein capsomers of UUKV.

The RVFV membrane-proximal RNP density was more or-dered than the rest of the RNP. Although similar ordering wasobserved for UUKV, the RNPs seem to be less important forparticle assembly in UUKV than in RVFV. UUKV glycopro-teins are sufficient to form empty virus-like particles withoutthe nucleocapsid protein (31). In the case of RVFV, glycopro-teins alone are not sufficient for generating virus-like particles,but nucleocapsid protein, and potentially also RNA, arerequired (16). While the interaction between the RVFV Nprotein and RNA has not been investigated, in members ofthe genus Orthobunyavirus, the N protein is known to bindviral RNA, cRNA, and, to a lesser extent, also nonspecificcellular RNA (26, 27). Thus, it is possible that cellular RNAis also incorporated into RVFV virus-like particles. Consid-ering these findings, we suggest that the interaction betweenthe RNP and the glycoproteins stabilizes the RVFV virionsor drives their assembly. This hypothesis is supported by theresults of our study, which show that the positions of mem-brane-proximal RNP density and glycoprotein transmem-brane densities are correlated.

pH-dependent conformational changes have been suggestedto play an essential role during bunyavirus entry in the endo-some. These conformational changes have been reported tooccur at pH 6.4 in the GC of La Crosse virus (15) and betweenpH 6.0 and 6.2 in the GC of UUKV (36). We have recentlyobserved a conformational change of UUKV glycoprotein cap-somers in virions at pH 6.0, studied using electron cryo-tomog-raphy (31). To study if pH-dependent conformational changeswould also occur in RVFV, we analyzed samples prepared atpH 6.0, in addition to those prepared at pH 7.4 (see Materialsand Methods). Unexpectedly, the difference map calculatedfor the two reconstructions did not reveal any significant dif-ferences (not shown). Further studies over a broader pH rangeare required to gain insight into possible pH-dependent struc-tural changes and the fusion mechanism of RVFV.

The view that bunyaviruses are pleomorphic, lacking well-defined shape, is now changing rapidly for the members of thegenus Phlebovirus. Virions of this genus are likely to haveicosahedral symmetry, as suggested earlier for UUKV andmore recently for RVFV (10, 28). Here, we have demonstratedthat the single-particle averaging approach exploiting icosa-hedral symmetry is feasible for studying the RVFV struc-ture. The achieved 2.2-nm resolution allowed us to analyzethe viral glycoprotein organization to a level that was notpossible in an earlier tomographic reconstruction at a muchlower resolution (6.1 nm) (10). Through improvements insample purification and preservation, we expect the ap-proach taken here to also be applicable to other phlebovi-ruses at increasing resolutions. Electron cryo-microscopyand single-particle averaging, together with X-ray structuredetermination of phlebovirus glycoprotein structures, wouldenable more detailed understanding of the assembly and



.org/D

ownloaded from

http://jvi.asm.org/

infection mechanism of these negative-stranded RNA vi-ruses lacking a matrix protein.

ACKNOWLEDGMENTS

We acknowledge support from the Academy of Finland (114649 toJ.T.H.), from the European Molecular Biology Organization (ALTF820-2006 to J.T.H.), and from the Deutsche Forschungsgemeinschaft(GR1990/1-3 and -2-1 to K.G. and WE 2616/5-1 to F.W.).

REFERENCES

1. Baker, T. S., and R. H. Cheng. 1996. A model-based approach for determin-ing orientations of biological macromolecules imaged by cryoelectron mi-croscopy. J. Struct. Biol. 116:120–130.

2. Balkhy, H. H., and Z. A. Memish. 2003. Rift Valley fever: an uninvitedzoonosis in the Arabian Peninsula. Int. J. Antimicrob. Agents 21:153–157.

3. Besselaar, T. G., and N. K. Blackburn. 1991. Topological mapping of anti-genic sites on the Rift Valley fever virus envelope glycoproteins using mono-clonal antibodies. Arch. Virol. 121:111–124.

4. Bird, B. H., M. L. Khristova, P. E. Rollin, T. G. Ksiazek, and S. T. Nichol.2007. Complete genome analysis of 33 ecologically and biologically diverseRift Valley fever virus strains reveals widespread virus movement and lowgenetic diversity due to recent common ancestry. J. Virol. 81:2805–2816.

5. Centers for Disease Control and Prevention. 2007. Rift Valley fever out-break—Kenya, November 2006–January 2007. MMWR Morb. Mortal. Wkly.Rep. 56:73–76.

6. Chen, S. Y., and R. W. Compans. 1991. Oligomerization, transport, andGolgi retention of Punta Toro virus glycoproteins. J. Virol. 65:5902–5909.

7. Chen, S. Y., Y. Matsuoka, and R. W. Compans. 1991. Golgi complex local-ization of the Punta Toro virus G2 protein requires its association with theG1 protein. Virology 183:351–365.

8. Crowther, R. A., L. A. Amos, J. T. Finch, D. J. De Rosier, and A. Klug. 1970.Three dimensional reconstructions of spherical viruses by Fourier synthesisfrom electron micrographs. Nature 226:421–425.

9. Daubney, R., J. R. Hudson, and P. C. Garnham. 1931. Enzootic hepatitis orRift Valley fever. An undescribed virus disease of sheep cattle and man fromEast Africa. J. Pathol. Bacteriol. 34:545–579.

10. Freiberg, A. N., M. B. Sherman, M. C. Morais, M. R. Holbrook, and S. J.Watowich. 2008. Three-dimensional organization of Rift Valley fever virusrevealed by cryo-electron tomography. J. Virol. 82:10341–10348.

11. Fuller, S. D., S. J. Butcher, R. H. Cheng, and T. S. Baker. 1996. Three-dimensional reconstruction of icosahedral particles—the uncommon line. J.Struct. Biol. 116:48–55.

12. Geisbert, T. W., and P. B. Jahrling. 2004. Exotic emerging viral diseases:progress and challenges. Nat. Med. 10:S110–S121.

13. Gerrard, S. R., and S. T. Nichol. 2007. Synthesis, proteolytic processing andcomplex formation of N-terminally nested precursor proteins of the RiftValley fever virus glycoproteins. Virology 357:124–133.

14. Gerrard, S. R., and S. T. Nichol. 2002. Characterization of the Golgi reten-tion motif of Rift Valley fever virus G(N) glycoprotein. J. Virol. 76:12200–12210.

15. Gonzalez-Scarano, F. 1985. La Crosse virus G1 glycoprotein undergoes aconformational change at the pH of fusion. Virology 140:209–216.

16. Habjan, M., N. Penski, V. Wagner, M. Spiegel, A. K. Overby, G. Kochs, J. T.Huiskonen, and F. Weber. 19 January 2009, posting date. Efficient produc-tion of Rift Valley fever virus-like particles: the antiviral protein MxA caninhibit primary transcription of bunyaviruses. Virology. doi:10.1016/j.virol.2008.12.011.

17. Heymann, J. B., and D. M. Belnap. 2007. Bsoft: image processing andmolecular modeling for electron microscopy. J. Struct. Biol. 157:3–18.

18. Huiskonen, J. T., and S. J. Butcher. 2007. Membrane-containing viruses withicosahedrally symmetric capsids. Curr. Opin. Struct. Biol. 17:229–236.

19. Kastner, B., N. Fischer, M. M. Golas, B. Sander, P. Dube, D. Boehringer, K.Hartmuth, J. Deckert, F. Hauer, E. Wolf, H. Uchtenhagen, H. Urlaub, F.

Herzog, J. M. Peters, D. Poerschke, R. Luhrmann, and H. Stark. 2008.GraFix: sample preparation for single-particle electron cryomicroscopy. Nat.Methods 5:53–55.

20. Kivioja, T., J. Ravantti, A. Verkhovsky, E. Ukkonen, and D. Bamford. 2000.Local average intensity-based method for identifying spherical particles inelectron micrographs. J. Struct. Biol. 131:126–134.

21. Laurinmaki, P. A., J. T. Huiskonen, D. H. Bamford, and S. J. Butcher. 2005.Membrane proteins modulate the bilayer curvature in the bacterial virusBam35. Structure 13:1819–1828.

22. Ludtke, S. J., P. R. Baldwin, and W. Chiu. 1999. EMAN: semiautomatedsoftware for high-resolution single-particle reconstructions. J. Struct. Biol.128:82–97.

23. Mastronarde, D. N. 2005. Automated electron microscope tomography usingrobust prediction of specimen movements. J. Struct. Biol. 152:36–51.

24. Matthews, B. W. 1968. Solvent content of protein crystals. J. Mol. Biol.33:491–497.

25. Nichol, S. T., B. J. Beaty, R. M. Elliott, R. W. Goldbach, A. Plyusnin, andR. B. Tesh. 2005. The Bunyaviridae, p. 695–716. In C. M. Fauquet, M. A.Mayo, J. Maniloff, U. Desselberger, and L. A. Ball (ed.), Virus taxonomy.Eighth report of the International Committee on Taxonomy of Viruses.Elsevier Academic Press, London, United Kingdom.

26. Ogg, M. M., and J. L. Patterson. 2007. RNA binding domain of JamestownCanyon virus S segment RNAs. J. Virol. 81:13754–13760.

27. Osborne, J. C., and R. M. Elliott. 2000. RNA binding properties of buny-amwera virus nucleocapsid protein and selective binding to an element in the5� terminus of the negative-sense S segment. J. Virol. 74:9946–9952.

28. Overby, A. K., R. F. Pettersson, K. Grunewald, and J. T. Huiskonen. 2008.Insights into bunyavirus architecture from electron cryotomography ofUukuniemi virus. Proc. Natl. Acad. Sci. USA 105:2375–2379.

29. Overby, A. K., R. F. Pettersson, and E. P. Neve. 2007. The glycoproteincytoplasmic tail of Uukuniemi virus (Bunyaviridae) interacts with ribo-nucleoproteins and is critical for genome packaging. J. Virol. 81:3198–3205.

30. Overby, A. K., V. L. Popov, R. F. Pettersson, and E. P. Neve. 2007. Thecytoplasmic tails of Uukuniemi virus (Bunyaviridae) G(N) and G(C) glyco-proteins are important for intracellular targeting and the budding of virus-like particles. J. Virol. 81:11381–11391.

31. Overby, A. K., V. Popov, E. P. Neve, and R. F. Pettersson. 2006. Generationand analysis of infectious virus-like particles of Uukuniemi virus (Bunyaviri-dae), a useful system for studying bunyaviral packaging and budding. J. Virol.80:10428–10435.

32. Persson, R., and R. F. Pettersson. 1991. Formation and intracellular trans-port of a heterodimeric viral spike protein complex. J. Cell Biol. 112:257–266.

33. Pesonen, M., E. Kuismanen, and R. F. Pettersson. 1982. Monosaccharidesequence of protein-bound glycans of Uukuniemi virus. J. Virol. 41:390–400.

34. Pettersen, E. F., T. D. Goddard, C. C. Huang, G. S. Couch, D. M. Greenblatt,E. C. Meng, and T. E. Ferrin. 2004. UCSF Chimera—a visualization systemfor exploratory research and analysis. J. Comput. Chem. 25:1605–1612.

35. Pettersson, R. F., and L. Melin. 1996. Synthesis, assembly, and intracellulartransport of the Bunyaviridae membrane proteins, p. 159–188. In R. M.Elliott (ed.), The Bunyaviridae. Plenum Press, New York, NY.

36. Ronka, H., P. Hilden, C. H. Von Bonsdorff, and E. Kuismanen. 1995. Ho-modimeric association of the spike glycoproteins G1 and G2 of Uukuniemivirus. Virology 211:241–250.

37. Schmaljohn, C. S., and S. T. Nichol. 2007. Bunyaviridae, p. 1741–1789. InD. M. Knipe, P. M. Howley, D. E. Griffin, R. A. Lamb, M. A. Martin, B.Roizman, and S. E. Straus (ed.), Fields virology, 5th ed. Lippincott Williams& Wilkins, Philadelphia, PA.

38. Talmon, Y., B. V. Prasad, J. P. Clerx, G. J. Wang, W. Chiu, and M. J.Hewlett. 1987. Electron microscopy of vitrified-hydrated La Crosse virus.J. Virol. 61:2319–2321.

39. Zhang, W., P. R. Chipman, J. Corver, P. R. Johnson, Y. Zhang, S. Mukho-padhyay, T. S. Baker, J. H. Strauss, M. G. Rossmann, and R. J. Kuhn. 2003.Visualization of membrane protein domains by cryo-electron microscopy ofdengue virus. Nat. Struct. Biol. 10:907–912.



.org/D

ownloaded from

http://jvi.asm.org/

Documents

Electron Cryo-Microscopy and Single-Particle Averaging of Rift Valley Fever Virus: Evidence for G N -G C Glycoprotein Heterodimers