Geminate Structures of African Cassava Mosaic Virus

Two types of geminate structures were purified from Africancassava mosaic geminivirus (ACMV)-infected Nicotiana benthamianaplants and analyzed by electron cryomicroscopy and image reconstruction.After cesium sulfate density gradient centrifugation, they wereseparated into lighter top (T) and heavier bottom (B) components.T particles comigrated with host proteins, whereas B particleswere concentrated in a cesium density typical for complete virions.Both particles were composed of two incomplete icosahedra of11 capsomers each, but T particles were slightly larger (diameter,22.5 nm) and less dense in the interior than B particles (diameter,21.5 nm). T particles were frequently associated with smallglobules of $~$ 14 nm diameter of unknown origin. The overall structureof ACMV, a begomovirus transmitted by whiteflies, was similarto that of Maize streak virus (MSV), a mastrevirus transmittedby leafhoppers, although the vertices of the icosahedra wereless pronounced. Models of ACMV coat proteins based on Satellitetobacco necrosis virus support the exposure of parts of themolecule essential for transmission specificity by whitefliesand provide possible structural explanations for the smallerprotrusion of the ACMV capsid relative to MSV. The differencesof ACMV and MSV virion shapes are discussed with reference totheir different animal vectors.

INTRODUCTION

Top

Introduction

Geminiviruses are major plant pathogens in tropical and subtropicalcountries (21). They possess a unique capsid structure of 18by 30 nm in diameter (24). Their circular single-stranded DNA(ssDNA) is packaged into two incomplete T1 icosahedra, whichare formed by multiples of a single coat protein (CP). Earlyelectron microscopic investigations revealed a first detailedmodel for Chloris striate mosaic virus (8, 9, 14), which wasrecently refined by electron cryomicroscopy and image reconstructionfor Maize streak virus (MSV) (29). Both viruses belong to thegenus Mastrevirus of the family Geminiviridae, which comprisesviruses with a monopartite genome infecting monocotyledonoushosts and vectored by different species of leafhoppers (24).

In contrast, African cassava mosaic virus (ACMV) belongs tothe genus Begomovirus, most members of which possess bipartitegenomes, dicotyledonous hosts, and one whitefly species (Bemisiatabaci Genn.) as a vector. Transmission by insects is dependenton the CP (5, 16, 17), and therefore it is conceivable thatthe capsid structure might have been adapted to different receptorsof the particular insects. To analyze such an interaction, detailedstructural information on the capsid morphology is desirable.Therefore, electron cryomicroscopy was used in combination withimage reconstruction to get more detailed structural informationon ACMV in order to compare its structure with that of MSV (29).

MATERIALS AND METHODS

Top

Materials and Methods

Purification of virus particles. Nicotiana benthamiana plants were infected with cloned ACMVDNA, and virus particles were purified by using differentialand cesium sulfate-gradient centrifugation as described in detailelsewhere (27). Virus particle-containing fractions were identifiedby analyzing each sample by Southern blotting and hybridizationagainst ACMV DNA A, as well as by Western blotting with anti-ACMVCP antibodies (27). Fractions with viral ssDNA and CP were pooled,dialyzed overnight against 0.1 M sodium borate buffer (pH 8.0)at 4°C, concentrated, and washed with borate buffer in MicroconYM-10 filtration tubes according to the manufacturer's recommendations(Millipore, Bedford, Mass.). To remove residual glycerol thatmay interfere with cryoelectron microscopy, filtration tubeswere washed with borate buffer (30 min, 14,000 x g, 4°C)prior to application of the sample. The samples were appliedsequentially in 500-µl portions on the membrane, washedonce with borate buffer (40 min [each spin], 14,000 x g, 4°C),and the concentrated virus particles were eluted (1 min, 1,000x g, 4°C).

Electron microscopy and image processing. (i) Unstained samples. Samples were frozen in a modified controlled environment freezingapparatus (3) at room temperature. Two sponges soaked in hotwater were placed inside the environmental chamber of the deviceensuring a humidity of nearly 100%. The sample (2 µl)was applied to grids (400 mesh copper-rhodium; Maxtaform HR26R)coated with a perforated carbon film covered with an additionalthinner continuous carbon film. The additional carbon supportreduced charging effects and raised the number of particlesobserved in the field of view. To ensure a homogeneous spreadof the particle solution, the grids were treated by glow dischargein air prior to use.

For freezing the grids loaded with sample were mounted insidethe chamber of the freezing device. Most of the sample was removedby blotting for 15 s with two layers of filter paper (Whatmanno. 1). The grid was then plunged into liquid ethane, whichwas cooled by a surrounding bath of liquid nitrogen. A heatingdevice inside the ethane pot prevented freezing of the ethane.Prepared grids were stored in liquid nitrogen until used formicroscopy.

(ii) Stained samples. Samples stained with uranyl formate and sandwiched between twolayers of carbon were prepared as described in the supplementto reference 12. The grids were frozen within 10 min by dippingthem into liquid nitrogen. Frozen grids were stored in liquidnitrogen until used for microscopy.

The frozen grids of the unstained sample were transferred witha Gatan Cryo-Holder 626 into a Philips CM-200-FEG equipped witha field emission gun. The microscope was operated under lowdose conditions at a 200-kV accelerating voltage. Micrographswere taken on Kodak SO-163 film, at a nominal magnificationof x66,000 and with a defocus of 1.3 to 3.5 µm.

Stained samples were transferred into a CM-120-Biotwin equippedwith an LaB₆-filament. The microscope was operated at 100 kV.Micrographs were taken at a nominal magnification of x52,000and with a defocus of 200 to 800 nm. The exposed micrographswere developed for 10 min in full-strength Kodak D-19 developerat room temperature.

For image processing, suitable micrographs were scanned witha Zeiss SCAI scanner. For the unstained samples, a pixel sizeof 14 µm corresponding to 2.1 Å at the specimenlevel was chosen (stained particles were scanned with a pixelsize of 21 µm, 4 Å at the specimen level). Particleimages were selected interactively and were boxed off from themicrographs by using the MRC image processing programs (6).For the unstained particles, a box size for the individual particlesof 220 by 220 pixels (for stained particles, 128 by 128 pixels)was used. The individual particle images were corrected forthe contrast transfer function (ctf) of the microscope. In orderto do this, the defocus and the astigmatism of each micrographwere determined by using the program ctffind2 of the MRC package.For correction, the particle images were first floated to anaverage gray value of zero and then padded with zeros to a largerbox size of 756 by 756 pixels (512 by 512 pixels for stainedparticles). Amplitudes and phases were ctf corrected by usingctfapply, assuming a Wiener weighting factor of 0.7 and an amplitudecontrast of 0.06 (0.15 for stained particles). After correction,the box size was again reduced to 220 by 220 pixels (128 by128 for stained particles).

Further image processing was carried out by using the IMAGIC-5software package (28). Two-by-two pixels of the ctf-correctedparticle images were combined, giving a pixel size of 4.2 Åat the specimen level (for stained particles pixels were notcombined). For further processing, particle images were normalizedin their gray value distribution and band-pass filtered, includinginformation between 1/9 Å^–1 and 1/168 Å^–1(1/10 Å and 1/160 Å for stained particles). Furtherimage processing was carried out as outlined in the IMAGIC-5manual. For the three-dimensional reconstructions and the determinationof the Euler-angles, a D5 symmetry was assumed. To check whetherthe result depended on the imposed symmetry, we did severalrounds of refinement and image reconstruction (alignment, classification,sinogram correlation, and map reconstruction), assuming thelower C1 symmetry, which basically showed the same result butwith less resolution.

For estimation of the resolution, two three-dimensional mapswere calculated from half of the data each. Both maps were comparedby Fourier-shell correlation. The resolution was estimated eitheraccording to the spatial frequency where the correlation droppedto 0.5 or to where it cut the symmetry-corrected 3 ${sigma}$ (23) curve.

The initial band-pass filtering and the contrast transfer functionof the microscope lead to an underestimation of the low-resolutionamplitudes, causing dark fringes and an overestimation of theholes inside the virus particle. To compensate for this effectin the unstained particles, the amplitudes of the low spatialfrequencies in the final map were scaled as described earlier(20).

Sequence and structural alignments and molecular modeling. Standard (BLAST) and iterative (PSI-BLAST) sequence similaritysearches (1) were run at the web server of the National Centerfor Biotechnology Information of the National Institutes ofHealth (http://www.ncbi.nih.gov). For multiple sequence alignment,sequences were evaluated by the CLUSTAL W (15) server at theEuropean Bioinformatics Institute (http://www.ebi.ac.uk). Theresulting alignment of geminivirus CPs was compared to the seedalignment of the profile PF00844 at the PFAM (2) server of theSanger Institute (http://www.sanger.ac.uk). For structure predictions,we used MetaServer at http://BioInfo.pl, which derives and scoresconsensus predictions from the results of various primary structureprediction servers (11). Based on the consensus alignment ofthe ACMV CP to that of the Satellite tobacco necrosis virus(STNV), a pseudoatomic model of the ACMV CP was built with version6.2 of the program MODELLER (25). Version 2.1 of the programMolscript (19) was used to generate the graphics in Fig. 9.

View larger version (48K):
[in this window]
[in a new window]
FIG.9. (A) Surface representation of stained ACMV particle B component (yellow) superimposed on a three-dimensional reconstruction of paired globules of the T component (red). For clarity, only half of the map of the ACMV B-component particles is shown. For calculating thethree-dimensional reconstruction of the paired globules, D5 symmetry was assumed. (B) Superposition of an ACMV model capsid generated from the atomic model of the ACMV CP with the density map derived by electron microscopy. The C ${alpha}$ -chain is shown in yellow and was fitted into the three-dimensional density map (blue wireframe). (Left) View from the outside; (center) slice shows the inner surface of the particle; (top right) slice of the equatorial part of the geminate particle; (bottom right) slice through the lower part of the geminate particle. (C to E) Modeling of the ACMV CP. (C) Consensus alignment of the ACMV, BCTV, MSV, and SNTV CPs compiled from a multiple sequence alignment of geminivirus CPs and structural alignments of the ACMV and MSV CPs to that of STNV. The N-terminal ${alpha}$ -helix is boxed and major ß-strands are labeled as described earlier (28) and underlined. Residues in boldface are conserved among the three proteins. Major insertions in the geminivirus proteins are double underlined. (D) Atomic structure of a pentameric capsomer of STNV (as recorded in PDB entry 2STV). (E) Model of a single ACMV CP in an orientation similar to that of the white monomer in panel C. The ßD/ßE loop is shown in red, with the six-residue ACMV insertion in violet. The ßF/ßG loop is depicted in blue, with the four-residue ACMV insertion in cyan. The start of each major ß-strand was labeled as described earlier (29).

Fitting of the atomic model to the three-dimensional map. First, an icosahedral capsid was generated by applying icosahedralsymmetry to the atomic model of the ACMV CP. The same packingwas assumed as in the template used for modeling the structure(STNV). Comparison of the modeled capsid and the capsid observedin the three-dimensional maps showed that the modeled capsidwas 10 to 15% too small for matching the size of halve of ageminate particle. To allow for the difference in size, theatomic model of the ACMV CP was radially moved outward beforewe applied icosahedral symmetry, giving a capsid with a suitablediameter. For further analysis we took the 11 subunits formingthe asymmetrical unit in the geminate particles (one CP fromthe upper capsomer and one peripentonal and one equatorial capsomer).These were rotated and translated to the correct position inthe three-dimensional map. The position of the equatorial capsomerwas slightly adjusted by tilting it slightly around its pseudofivefold axis. This moved the top of the capsomer closer tothe equator. The complete geminate particle was then generatedfrom this unit by applying D5 symmetry.

RESULTS AND DISCUSSION

Top

Results and Discussion

ACMV was purified by differential and isopycnic density gradientcentrifugation in cesium salts. All of the gradient fractionswere analyzed by Southern blotting and hybridization againstviral probes (Fig. 1), as well as by Western blotting and anti-CPantibody detection (see references 27). As expected, most viralssDNA and CP accumulated in fractions with a density of ca.1.30 g/cm³ (Fig. 1), referred to as the bottom pool (B). Moreover,some of these materials accumulated in fractions where mostof the copurified host proteins usually band, at a density of1.23 g/cm³ (Fig. 1), referred to as top pool (T). Both poolswere freed from cesium salts and analyzed by conventional electronmicroscopy by using negative stain or by electron cryomicroscopywith or without staining.

View larger version (18K):
[in this window]
[in a new window]
FIG. 1. Southern blot analysis of all fractions of an isopycnic cesium sulfate gradient after centrifugation, with increasing densities from right (top) to left (bottom). Viral chromatin, including open circular (oc) and covalently closed (ccc) double-stranded DNA, is well separated from viral twin particles and solitary particles containing either genomic (ss) or subgenomic (ss/2) ssDNA. Fractions pooled for electron microscopy are indicated for the bottom pool (B) and the top pool (T). Marker (M) was linearized full-length ACMV DNA A (100 pg).

Particles from both pools spread evenly across the carbon layer(Fig. 2; see also Fig. 6) and had a bipartite shape with twoparts of equal size separated by a gap. The B pool containedmostly geminate particles with some contaminating viral minichromosomes.From 47 micrographs, such as the one shown in Fig. 2, a totalof 5,530 individual particle images were selected. These imageswere aligned in respect to each other and classified accordingto their similarity by using IMAGIC 5. Typical class averagesare shown in Fig. 3. The class averages show particles thatvary in their length but have roughly the same diameter. Thisvariation in length is caused by different orientations of theparticles rather than by variations in the shapes of the particlesthemselves.

View larger version (132K):
[in this window]
[in a new window]
FIG. 2. Electron micrograph of unstained, frozen hydrated ACMV particles. Particles appear as dark bipartite structures. Bar, 50 nm.

View larger version (74K):
[in this window]
[in a new window]
FIG. 6. Micrographs of negatively stained, frozen ACMV B component (A) and T component (B). Some ACMV particles are lined with a black frame. In the T component small globules appeared frequently (white circles), which were sometimes paired or were attached to ACMV particles (white arrow). Bar, 100 nm.

View larger version (133K):
[in this window]
[in a new window]
FIG. 3. Two-dimensional projection maps of ACMV particles. Aligned particle images were sorted according to their similarity and divided into 20 classes. Class members were averaged and are shown above. Bar, 20 nm.

For calculation of the three-dimensional map, the data was dividedinto 300 classes and the class averages were calculated. Foreach of the class averages the relative spatial orientationwas determined by sinogram correlation and assuming D5 symmetry,the same symmetry that was observed for the MSV (29). To confirmthe chosen symmetry, particle orientations and a three-dimensionalmap were also determined, assuming the lower C1 symmetry. Thesame packing of capsomers was observed justifying the choiceof the D5 symmetry (data not shown). The orientations of theclass averages covered the asymmetric unit evenly (Fig. 4),which is a prerequisite for an isotropic resolution. The finalmap included 250 of the 300 class averages, excluding mainlyclasses with a smaller number of members and thus being noisier.The Fourier shell correlation dropped to 0.5 at 1/19 Å^–1and cut the symmetry-corrected 3 ${sigma}$ -curve (23) at 1/16 Å^–1.Accordingly, the resolution of the final map was estimated tobe between 16 and 19 Å. Surface representations and threeorthogonal slices through the center of the map are shown inFig. 5. The diameter of the ACMV capsid was 22 nm, as foundfor MSV (29).

View larger version (29K):
[in this window]
[in a new window]
FIG. 4. Spatial distribution of Euler angles in the asymmetric unit. Each "+" symbol corresponds to the orientation determined for an observed class average. Only the out-of-plane rotational angles are shown.

View larger version (81K):
[in this window]
[in a new window]
FIG. 5. Representations of the three-dimensional map calculated from unstained, frozen hydrated ACMV particles. (A and C) Three orthogonal slices of the three-dimensional map; (B and D) three different views of the surface representation of the three-dimensional map. In panels C and D, low-resolution amplitudes were scaled to minimize negative densities. Thin arrows indicate low density in the connecting region, which leads to discontinuity in the surface representation. After appropriate amplitude scaling, the discontinuity disappears. Thick arrows mark the core density inside the capsids. Bar, 20 nm.

In the slices, the particle showed a high-density outer shellthat surrounded a lower-density core structure. The shell wasorganized into lower and upper halves with a connecting regionof lower density in between (Fig. 5A and B, thin arrows). Thisregion did not show in surface representation and, therefore,the two parts of the virus appeared to be disconnected. We thinkthat underestimation of the amplitudes of the data at low spatialfrequencies caused this discontinuity. The dampening of theamplitudes at low spatial frequencies is caused by the high-passfilter applied during image processing and the uncorrected amplitudemodulation introduced by the contrast transfer function of theelectron microscope. If the amplitudes at low spatial frequencieswere scaled to minimize negative densities in the map as describedpreviously (20), the connection between the two parts and somedensity inside the particles became visible (Fig. 5C and D).A comparable density distribution was observed (not shown) whenthree-dimensional maps were calculated with frealign (13), whichscales amplitudes according to the modulation of the contrasttransfer function and does not apply high-pass filters. Thisemphasized the validity of our scaling approach.

The density inside the particle (core density) was now muchmore visible and appeared as ring-like structures in the slices(Fig. 5A and C, wide arrows) or as spherical densities thatfilled each of the incomplete icosahedra in three-dimensionalmaps. The core density had a dense outer rim and was less solidin its center. The outer rim lined the inner surface of theprotein shell of the capsids. It is plausible that the coredensity accounts for the DNA of ACMV. A similar core densitywith an outer ring lining the inner surface of the outer shellis observed in surface representations of MSV (29).

As already shown for MSV (29) and in contrast to the earliermodel for Chloris striate mosaic virus (14), the halves of thegeminate particles are twisted to each other by 20° suchthat the capsomers of one half point into the gap between twocapsomers of the other half.

To compare the T- and B-pool particles, negatively stained sampleswere investigated with an cryoelectron microscope (Fig. 6).The staining approach was chosen because the T component tendedto aggregate and was too dilute for conventional cryoelectronmicroscopy. Similar to the unstained sample (Fig. 2), the negativelystained B component showed an even spread of ACMV twin particleswith few smaller globules in the background (Fig. 6A). The Tcomponent had fewer ACMV particles, to which frequently globuleswere attached (Fig. 6B, arrow), with various positions at theouter surface of the twin capsids. In the background, similarglobules with a diameter of 14 ± 2 nm were observed,which occasionally appear in pairs (Fig. 6B, white double circles).The distance between the centers of these globules was equalto the distance of the centers of the two incomplete icosahedraforming the ACMV twin particles (16.5 nm).

For further comparison of T and B ACMV particles, three-dimensionalimage reconstructions were calculated by using IMAGIC-5 as outlinedabove. Surface representations of the image reconstructionsare shown in Fig. 7. Although T (Fig. 7A) and B (Fig. 7B) particleslooked very similar in their overall architecture, the T componentwas slightly larger (4% larger in diameter). This became evenmore evident when slices through the density of the maps werecompared. These slices extended further for the T component(Fig. 8A) than for the B component (Fig. 8B). For differenceimaging, the maps were made binary by setting all gray valuesabove a certain threshold to 1 and setting all below this thresholdto 0. The difference between the binary maps (Fig. 8C) revealeddifferences either as black densities when only the T componentwas present or as white densities when only the B componentwas present. The difference image showed black fringes at theouter surface of the particle and white fringes at the innersurface. Both white and black fringes had similar widths, whichwas indicative of a swelling of the T-component particles withrespect to the B-component particles without significant alterationin the thickness of the protein shells.

View larger version (49K):
[in this window]
[in a new window]
FIG. 7. Surface representation of the stained ACMV particles of the B component (A) and T component (B). The left column shows fivefold views; the right column shows twofold views. The central column shows views straight onto one of the capsomers flanking the capsomer at the fivefold axis.

View larger version (36K):
[in this window]
[in a new window]
FIG. 8. Slices of the three-dimensional map of stained ACMV particles. (A) T component; (B) B component. Slices are 4 Å thick and are spaced by 20 Å. For difference imaging, the maps were normalized in their gray values and were binarized by using the same threshold value. Binarizing set all gray values above the chosen threshold to 1 and below the threshold to 0. The binary maps were subtracted (B component minus T component). The resulting difference map is shown in panel C. Where both binary maps matched, the difference was 0 (gray). Additional density in the T component appeared black, whereas additional density of the B component was white. Bar, 10 nm.

The nature of the smaller globules of the T component (Fig.5B) is still unclear. Since they had a diameter similar to thatof the core densities inside ACMV and in pairs had the samecenter-to-center distance as the core density, we thought thatglobules and twin particles might be related. To test this hypothesis,we calculated three-dimensional maps of the paired globulesassuming the same D5 symmetry that was applied to the core densityduring virus reconstruction. If the paired globules and thecore density inside the capsids of the B component were related,this approach should result in a comparable three-dimensionalreconstruction, independent of whether the chosen symmetry wascorrect. A superposition of such a reconstruction (Fig. 9A,red) with the reconstruction of the B capsid particles (Fig.9A, yellow) showed that the paired globules fitted neatly intothe capsid of ACMV similar to the core density. This observationwas rather persuasive evidence that the paired globules hadthe same nature as the core density, which we thought was theACMV DNA. If these assumptions were correct, the DNA must beable to fold into a defined structure with two spherical domains,which is stable and self-maintaining independent of the surroundingcapsid protein. Whether or not other protein components arepart of these structures is unknown.

In contrast to first intuition caused by the coincidence oftwin capsids and bipartite genomes, only one molecule of DNAA (930 kDa) or DNA B (910 kDa) is thought to fill up one twinparticle formed by 110 proteins (30.2 kDa), since ACM virionscontain 22% nucleic acids (4). Solitary globules may, therefore,represent subgenomic fragments of DNA A or DNA B present inthe bottom as well as in the top pool (Fig. 1, ss/2). This classof viral DNAs has been identified as defective-interfering DNA(26) and is packaged into solitary particles (10).

Whereas MSV has been prepared for structural analysis in bufferwith pH 4.8 (29), we used pH 8.0, assuming that this conditionis closer to the milieu of phloem (30) and plant-feeding insectsaliva (7). Nevertheless, the overall structure is very similar,emphasizing that the geminate configuration is stable at differentpHs. Minor difference in structural details, however, betweenthe model created here and that of Zhang et al. (29) might thereforebe attributed to the buffer conditions. One of the obvious structuraldifferences is in the size of the capsomers, which protrudeless from the shell in ACMV.

To account for this deviation, we analyzed an alignment of theACMV and MSV CPs and obtained structural predictions for bothproteins (Fig. 9C). The two CPs have diverged considerably,which precluded a direct alignment. However, iterative sequencesearches (1) produced pairwise alignments of these two proteinswith about 500 geminivirus CPs. No related sequence of knownstructure was found. A multiple sequence alignment of a selectednumber of sufficiently divergent sequences indicated that thelatter three quarters of the illustrated alignment of the ACMVand MSV CPs are reliable. Interestingly, an essentially identicalmultiple sequence alignment has been used to construct a sequenceprofile for geminivirus CPs (PFAM profile PF00844). We thensubmitted the two sequences to the Structure Prediction MetaServer(11), which in both cases yielded the structure of STNV CP (PDBentry 2STV, Fig. 9C) as the only template for structural modelingwith a significant score (80.29 and 57.50 for the ACMV and MSVsequences, respectively). The observation that these divergentsequences that could only be aligned in iterative sequence searchesnevertheless yielded the same unique template adds to the validityof the hit.

Notably, the same structure had already been proposed as a suitabletemplate for the MSV structure (29). For both proteins, severalhighly similar alignments to the STNV CP were suggested witha few variations concentrated in the loops between the mainsecondary structure elements. These structural alignments ofthe ACMV and MSV sequences with that of the STNV CP essentiallyagreed with the multiple sequence-based alignment of the geminivirussequences with the exception of the N-terminal helix and thefirst beta-strand. This agreement lends additional weight tothe structure prediction. The added sequence and structure informationaccounts for the deviations of our alignment from the pairwisealignment of Zhang et al. (29).

The resulting structural model (Fig. 9E) points to two interestingregions of insertion that are located near the tip of the capsomer.The six-residue insertion in the geminivirus ßD/ßE-loop(Fig. 9E, red loop) has been identified as essential for controllingwhitefly transmission (17, 18, 22). Its predicted exposed positionwould be compatible with such a role. The additional 14 residuesin the ßF/ßG-loop of the MSV CP, on theother hand, may well explain why the MSV capsomer protrudesmore from the shell than the ACMV capsomer (Fig. 9E, blue loop).The insertion might be a special adaptation to leafhopper transmission.To evaluate this interesting hypothesis, we included the CPsequence of Beet curly top virus (BCTV) in the alignment (Fig.9C) after structural alignment to STNV. BCTV infects dicotyledonoushosts as ACMV but is vectored by leafhoppers as MSV. Indeed,a similar insertion in the ßF/ßG-loop waspredicted for BCTV.

An atomic model of the ACMV capsid was generated as describedin Materials and Methods. In comparison to STNV the CP had tobe moved to larger radii to generate a capsid of proper sizeand to avoid clashes between the subunits in the capsomers.As a reference point for expansion, the C ${alpha}$ -Atom of Lys 175 wasused, which is located at the peripheral base of the capsomers.The radial translation was 10 Å compared to a translationof 31 Å in MSV (29). Expansion in the chosen directionalso led to an increase in the diameter of the capsomers, whichremoved the steric clashes between the subunits in the capsomers.The modeled capsomers matched the observed density well (Fig.9B). Only the N-terminal helices of the CPs, which point towardthe center of the incomplete icosahedra, projected out of thedensity. On the other hand, there is unaccounted density betweenthe capsomers, which could accommodate the N-terminal region.This would require a different intercapsomer interaction thanobserved in STNV. As mentioned above, for the N-terminal partSTNV is not a suitable structural template. Therefore, it isconceivable that in ACMV other than in STNV the N-terminal partcould mediate the intercapsomer contact similar to that proposedfor MSV (29). However, because there are not sufficient dataavailable on which the fold and the proposed interaction ofthe first 58 amino acids could be modeled, we left the positionof the helices unaltered.

In summary, we have shown that, although ACMV resembles MSVin its overall shape, there are key differences that possiblyaccount for the alternative transmission mode. The results providevaluable information for design mutagenesis experiments to unravelthe interaction of geminiviruses with their particular insectvectors.

ACKNOWLEDGMENTS

We thank John Stanley, Norwich, United Kingdom, for kindly providinginfectious clones of ACMV, and Werner Preiss for excellent technicalassistance.

H.C. is a postdoctoral fellow of the Fund for Scientific Research-Flanders.This study was supported by a grant from the Deutsche Forschungsgemeinschaft(Je 116/8-5).

FOOTNOTES

* Corresponding author. Mailing address: Structural and Computational and Biology Programme EMBL, Meyerhofstrasse 1, D-69117 Heidelberg, Germany. Phone: 49-6221-387-304. Fax: 49-6221-387-306. E-mail: boettcher{at}embl-heidelberg.de.

Present address: Division of Biochemistry, Katholieke UniversiteitLeuven, B-3000 Leuven, Belgium.

REFERENCES

Top