Structure of Double-Shelled Rice Dwarf Virus

Abstract

Rice dwarf virus (RDV), a member of the Reoviridae family, is a double-stranded RNA virus. Infection of rice plants with RDV reduces crop production significantly and can pose a major economic threat to Southeast Asia. A 25-Å three-dimensional structure of the 700-Å-diameter RDV capsid has been determined by 400-kV electron cryomicroscopy and computer reconstruction. The structure revealed two distinctive icosahedral shells: a T=13l outer icosahedral shell composed of 260 trimeric clusters of P8 (46 kDa) and an inner T=1 icosahedral shell of 60 dimers of P3 (114 kDa). Sequence and structural comparisons were made between the RDV outer shell trimer and the two crystal conformations (REF and HEX) of the VP7 trimer of bluetongue virus, an animal analog of RDV. The low-resolution structural match of the RDV outer shell trimer to the HEX conformation of VP7 trimer has led to the proposal that P8 consists of an upper domain of β-sandwich motif and a lower domain of α helices. The less well fit REF conformation of VP7 to the RDV trimer may be due to the differences between VP7 and P8 in the sequence of the hinge region that connects the two domains. The additional mass density and the absence of a known signaling peptide on the surface of the RDV outer shell trimer may be responsible for the different interactions between plants and animal reoviruses.

Rice dwarf virus (RDV) is a member of the genus Phytoreovirus of the family Reoviridae, which also includes animal reovirus, orbivirus, and rotavirus. RDV replicates both in insects and in graminaceous plant cells, but it can be transmitted only by insects such as the leafhopper (Nephotettix cincticeps or Resilia dorsalis) (30, 40). Unlike other phytoreoviruses, RDV does not induce neoplasia. Instead, plants infected with RDV are stunted, develop characteristic chlorotic flecks, and fail to bear seeds. This virus is widespread among rice plants in southern China and other Asian countries, leading to a possible severe decrease in rice production.

Like all other phytoreoviruses, intact RDV is a double-shelled particle enclosing a double-stranded RNA genome (18, 20, 21, 33, 41, 42). This double-shelled arrangement of the particle in phytoreoviruses has been shown to differ from structure for other members of Reoviridae, such as rotavirus, which are triple shelled (37, 44). The RDV genome is composed of 12 double-stranded RNA segments, designated S1 to S12 in ascending order of their mobility on a polyacrylamide gel (12). The complete sequences of all of these segments have been determined from a Japanese isolate (43) and a Chinese isolate (45), which share more than 90% sequence identity. Seven of the RDV gene products are considered to be structural proteins (29). The P3 (114-kDa) and P8 (46-kDa) proteins account for ∼29 and ∼52%, respectively, of the total RDV protein mass (32).

The RDV particle has been crystallized in the cubic space group I23 with a = 789 Å (28), but its three-dimensional (3D) structure remains unsolved. We have used electron cryomicroscopy (cryoEM) and computer reconstruction to derive the low-resolution 3D structure of the RDV particle. By combining the primary sequence and crystal structures of the major capsid protein (VP7) of bluetongue virus (BTV), an animal analog of RDV, we were able to deduce a relatively high resolution structural motif for P8, the major outer shell protein of RDV.

MATERIALS AND METHODS

Virus purification and cryoEM.

The RDV Zhejiang isolate was maintained and propagated in rice seedlings that were inoculated by the viruliferous leafhopper (N. cincticeps Uhler). Diseased leaves were harvested 2 months later, and the virus was purified by sucrose density gradient centrifugation (33). The purified RDV specimen was embedded in a thin layer of vitreous ice on a holey carbon grid with carbon support film, using a standard quick-freezing procedure (9). The carbon support film was used to prevent the virus particles from aggregating toward the edge of holey carbon film. The frozen hydrated intact RDV particles were kept at −162°C and imaged at a magnification of ×30,000 in a JEOL 4000 electron cryomicroscope operated at 400 kV with a dose of 6 to 8 e/Ų per micrograph. A focal pair was imaged from each specimen area first at close to focus and then 1-μm farther underfocused.

Image processing, 3D reconstruction, and visualization.

The micrographs were digitized on a Perkin-Elmer 1010M microdensitometer. Figure 1a is an example of a close-to-focus micrograph. Individual virus particles were boxed out into particle images of 140 by 140 pixels with a step size of 6 Å/pixel. Quality of the micrographs was evaluated from the incoherently averaged Fourier transforms of particle images (47) prior to subsequent data processing. The first zeros of the contrast transfer functions of three close-to-focus images used in the final reconstruction were clearly seen at 1/23, 1/18, and 1/17 Å⁻¹, which correspond to defocus values of 3.4, 2.0, and 1.8 μm, respectively.

FIG. 1.

CryoEM image and shaded surface representation of the 25-Å structure of full RDV. (a) Typical area of one of the electron cryomicrographs of RDV particles embedded in a thin layer of vitreous ice, recorded at 400 kV under low-dose conditions. Indicated by dashed circles are two particles. (b) Shaded surface view of the RDV reconstruction as viewed along the icosahedral twofold axis. The numbers 5, 3, and 2 designate the icosahedral five-, three-, and twofold axes. Highlighted in color are a contiguous group of five trimers found in each asymmetric unit. (c) Blown-up view of the group of five trimers computationally extracted from panel b. These trimers at the distinct quasi-equivalent positions are designated P, Q, S, R, and T, using a convention set forth for BTV (14).

The particles from the strongly underfocused micrographs were used to assist the determination of the center and orientation parameters of the corresponding particles in the close-to-focus micrographs. The determination and refinement of center and orientation parameters were based on the common-lines search method by comparing particle images and projections from preliminary reconstructions (5, 46, 48). The initial models were reconstructed from particle images in the underfocused micrographs, and the corresponding particles in the close-to-focus micrographs were refined against computed projections. All of the close-to-focus particle images were then combined to generate a new model followed by an additional cycle of refinement with the new projections. In the last round of refinement, a global and simultaneous refinement of the five particle parameters was performed by minimizing the cross common-line phase residuals across all particle images (46). The effective resolution of the final map was assessed by calculating the phase difference and Fourier ring correlation coefficients of two independent reconstructions (49). The reconstruction was carried out only to the first zero of the contrast transfer function, and thus no correction of the contrast transfer function was made.

To determine the absolute handedness of RDV, we recorded pairs of 0° and 20° tilt image from the same specimen areas. The handedness of the virus capsid was determined by comparing particle images in 0° tilt and 20° tilt, using a procedure described previously (3, 38). Twenty particle image pairs were then compared to determine handedness.

Structural components of interest in the map were computationally extracted and visualized by using the Explorer software (NAG Inc.) with custom-designed modules (7a). The mass of a computationally extracted subunit is estimated by assuming a protein density of 1.3 g/cm³. An isosurface value of 1ς (standard deviation) away from the mean density was used for rendering the surface representations. Our data processing was carried out on Iris R4400, R8000, and R10000 workstations (Silicon Graphics, Inc.).

Sequence and structural comparison.

The sequence alignment of P8 of RDV and VP7 of BTV was carried out by using MSA with a PAM 250 weight matrix available from the National Center for Supercomputing Applications (17). The crystal structure coordinates of HEX and REF of VP7 (2, 13) were provided by J. Grimes and D. Stuart, and the secondary structure motif of VP7 was obtained from PDB SUM (23). Following this, the 3D structures of REF and HEX, rendered in Ribbons (6), were aligned with the computationally extracted and scaled RDV trimers. Refinement of the RDV-BTV density fitting was then done to maximize the total density match based on visual inspection. This fitting was done with the map at atomic resolution and also the map blurred to 25-Å resolution and modified by a contrast transfer function corresponding to that in the RDV data. These densities were then aligned to the RDV trimer. Interpretation of mismatched density was done by examining the α-carbon trace of the BTV trimers by using RASMOL (R. Sayle, Glaxo Wellcome Research and Development, Greenford, United Kingdom). Further analysis of the structural interpretation was done by threading the primary sequence of the P8 protein through a profile search (11). Any proteins or protein segments with a Z score of ∼5.0 or greater were considered likely candidate proteins with similar folds.

RESULTS

Outer shell structure (T=13 icosahedral lattice).

The 3D structure of ice-embedded RDV was determined to 25-Å resolution from 81 particle images (Fig. 1a) recorded in a 400-kV electron cryomicroscope. The outer shell of RDV, with a 700-Å diameter, exhibits a T=13 icosahedral lattice, as shown clearly in the reconstruction map (Fig. 1b), contrary to the previous T=9 lattice model (21, 42). The absolute handedness was determined to be left-handed (T=13l) from tilt experiments of the ice-embedded RDV particles. The mass densities around the fivefold axis protrude radially ∼15 Å further outward than the mass densities around the threefold axis, resulting in a polyhedral appearance. The most prominent features on the outer shell are the 260 “knobby” trimeric density clusters centered at the local and strict threefold axes (Fig. 1b). These trimeric clusters associate with each other through extensive contacts at a lower radius around the icosahedral fivefold and local sixfold axes. There are also 132 openings, 12 at the fivefold axes and 120 at the local sixfold axes, which traverse the entire 69-Å-thick outer shell, becoming narrower at a lower radius (Fig. 1b).

Group of five trimers and their interactions.

In a T=13 icosahedral particle, there are 13 quasi-equivalent positions per asymmetric unit (Fig. 2). Each asymmetric unit contains four and one-third unique trimers (Fig. 1b and c; Fig. 2), which are designated P, Q, S, R, and T. The lettering scheme, based on the nomenclature used for BTV (14), is indicative of the relative position of the trimer: one around the fivefold axis (P [peripentonal]), three around a local sixfold axis (Q, S, and R), and one at the icosahedral threefold axis (T). To compare the structures of these trimers in detail, the five RDV trimer types were computationally extracted from the capsid reconstruction and displayed in equivalent top and side views (Fig. 3). The top domain of each trimeric knob, which measures 60 Å in diameter, has a triangular donut shape with a dimple or, in some cases, a hole at its center (Fig. 3b). The appearance of the dimples is sensitive to the choice of density threshold, suggesting the densities at these regions are not as well defined as those in other regions of the trimers. Such an observation has also been made in BTV, where dimples begin to appear at a higher density threshold (29a). Each trimer is about 69 Å in height, shorter than the structures of other outer capsid proteins from double-shelled virus particles in the Reoviridae (8, 15, 36, 38). Previous biochemical and immunological studies have indicated that the major outer shell protein is P8 (26, 32). By drawing an analogy to other members of the Reoviridae, (8, 15, 36, 38), we propose that each RDV trimer is made up of three subunits of the major outer shell protein P8. Thus, there would be 780 copies of the P8 monomeric subunit in the outer shell. This proposition was substantiated by structural comparison with the BTV capsid protein VP7 (see below).

FIG. 2.

Schematic diagram of the trimeric subunit organization within a triangular unit on a T=13l lattice. The five trimers are labeled P, Q, S, R, and T.

FIG. 3.

Comparison of trimers at distinct quasi-equivalent locations. (a) Side views of the five computationally extracted trimers P, Q, S, R, and T. The numerical labeling on the S trimer identifies a monomer of P8. The numbering also indicates the putative domain of the RDV trimer, where 1 is the upper domain and 2 and 3 are both located in the lower domain. More specifically, 2 marks the leg region and 3 indicates the floor region of the lower domain. (b) Top views of the five computationally extracted trimers.

There is no apparent contact among the adjacent trimers in the top domains regardless of contour display level. As the trimers of the outer shell extend toward the inner shell, each knob branches into three “legs” toward the adjacent five- or sixfold axes. These legs connect with neighboring legs to form an interconnected floor density on the outer shell (Fig. 1c). Furthermore, there seems to be more variability in the floor region density distribution than in the knob region among the trimers. This floor density network appears to be essential for maintaining the outer shell capsid stability since these are the only regions of interactions between the neighboring trimers.

Structural motif of P8.

The most structurally defined protein in an animal reovirus is VP7 of BTV, which consists of an upper domain with the typical β sandwich of viral proteins and a lower α-helical domain, as seen in both the REF and HEX crystal forms (Fig. 4) (2, 13). The major difference between these two structural isoforms, REF and HEX, is the relative orientation between these two domains. We have examined both the primary sequence and the two crystal structures of VP7 of BTV in an attempt to deduce a higher-resolution structural model of P8 of RDV from our low-resolution map.

FIG. 4.

Structures of the BTV VP7 and RDV P8 trimers. (a) Ribbon diagram of the REF conformation of BTV VP7; (b) side view of the RDV P8 S trimer merged with the REF VP7 ribbon diagram; (c) top view of panel b; (d) ribbon diagram of the HEX conformation of VP7; (e) side view of the RDV P8 S trimer superimposed with HEX VP7 ribbon diagram; (f) top view of panel d.

High genome sequence homology exists between RDV and other phytoreoviruses such as rice gall dwarf virus, wound tumor virus, and Fijivirus (4, 41), but no sequence homology between RDV and its animal analogs has been reported. The initial primary sequence alignment of RDV P8 and BTV VP7 revealed 27.5% homology, with a fairly large number of identities, as well as several obvious gaps. Upon overlaying the secondary structure of VP7, derived from the crystal structure (2, 13), onto the primary sequence alignment, we found that the secondary structure motifs corresponded well with the regions of homology (Fig. 5). The two structural domains were evident from the alignment, where the middle region of the RDV primary sequence corresponded to the β-sandwich domain and the two terminal regions corresponded to the lower α-helical domain. To further quantify the sequence homology between VP7 of BTV and P8 of RDV, sequence alignments of the helical and the β-sandwich domains were done separately. These segregated alignments also yielded homologies that were similar to the overall sequence alignment (Table 1).

FIG. 5.

RDV P8 and BTV VP7 amino acid sequence alignment. Homologous residues are shaded; regions of insertion in P8 relative to VP7 are noted above the RDV sequence. Secondary structure motifs, based on the VP7 structure, are shown as arrows for β sheets, zigzag lines for α helices, and hairpins for β turns.

TABLE 1.

Primary sequence homologies between P8 of RDV and VP7 of BTV

Region	% Homology	No. (of homologous residues/total no. of residues)
Entire protein	27.52	117/425
β strand	28.72	54/188
α helix	26.58	63/237

The crystal structures of the BTV VP7 trimers from both the REF and HEX isoforms (Fig. 4a to b) were fitted to the RDV trimers computationally extracted and scaled from our low-resolution map to see whether any structural homology was present. Figures 4b and c and 4e and f show examples of the fittings of the two structural isoforms of BTV trimers (ribbon representation) with the S trimer of RDV (shaded surface representation). Additionally, fitting of the other four trimers showed a match similar to that found in the S trimer. Similar fits were also found with the trimers and the VP7 trimers that had been blurred to 25 Å.

In the top view (Fig. 4c and f), both the REF and HEX conformations of the BTV trimer appear to fit the triangular donut-shaped upper domain of RDV in similar orientations. This good match suggests that the upper domain of the P8 subunit in the RDV trimer would likely have the same β-sandwich motif as that of the VP7 in the BTV trimer. However, a small density in the upper domain is seen in the RDV trimer but not the BTV trimer. To investigate the cause of this extra density, we examined the sequences of both proteins and found that the amino acid sequence of P8 of RDV is 71 residues longer than that of BTV VP7 (Fig. 5). The extra residues are mostly, but not entirely, accommodated by gaps within the alignment both at the C-terminal end and within the putative β-sandwich domain. The insertional residues in the putative β-sandwich domain are located in the gaps between residues 202 and 238 (Fig. 5). Therefore, the extra density seen in the upper domain of RDV (represented by space-filling balls in Fig. 6) might correspond to a total of 25 insertional residues in the four gaps.

FIG. 6.

Potential regions of structural variability. (a) Side view of the VP7 REF monomer of BTV. The space-filling region (residues 200 to 215) marks the region of major insertion in the RDV-BTV sequence alignment. The solid-line boxes (residues 251 to 253) and dashed-line boxes (residues 120 to 122) enclose the putative flexible hinge domain. (b) Top view of the REF VP7 trimer, illustrating the spatial distribution of the major insertional region.

While the upper domains of the REF and HEX trimers seem to fit that of the RDV trimer, the differences in the overall shape of the BTV trimer are evident in the lower domain. The REF global conformation appears to be a cylinder 80 Å in length with clockwise twisting of the monomeric subunits, whereas the HEX conformation is about 69 Å in length with L-shaped monomers (Fig. 4a and d). By examining the side views (Fig. 4b and e), we found that the REF conformation is too long and too narrow to fit the RDV trimer density. However, the HEX trimer has dimensions that are nearly identical to those of the RDV trimer and appears to accommodate the helical leg densities extending from the side of the RDV trimer (labeled “2” in Fig. 3). This suggests that the lower domain of P8 may have the α-helical motif of the HEX conformation. Nevertheless, minor mismatches in the bottom regions, including helices 1 and 2 of VP7, are noticeable. A primary sequence alignment of this region revealed little sequence homology but similar hydrophobicity plots (22). It has been suggested that helices 1 and 2 are highly hydrophobic and interact with the inner shell proteins in BTV (13). Therefore, these regions in P8 may have similar types of interactions with the inner shell proteins.

To seek further evidence to support our structural motif assignment based on the primary sequence alignment and 3D fitting of the RDV and BTV trimers, we undertook a series of computational tests with Fischer and Eisenberg’s Fold Recognition server (http://fold.doe-mbi.ucla.edu); this computational method predicts structural motifs based on the amino acid sequence and their biophysical properties (11). The algorithm searches for the best fit between a known protein or protein segment in the Protein Data Bank and a queried protein segment of unknown structure. The results of this analysis with the entire P8 sequence as well as various segments of P8, corresponding to the putative top, middle, and lower domains, are summarized in Table 2. The highest score for the whole P8 sequence was 4.96 with bacteriophage φX174 capsid protein GP, which contains a β sandwich (27). When the putative β-sheet region of P8 (residues 128 to 315 [Fig. 5]) was used, the Z score was improved to 5.57, the profile being similar to that of tumor necrosis factor alpha, another β-sandwich protein (10). Upon removal of all insertions in the putative β-sheet regions in the aligned RDV sequence (Fig. 5), a better match was found with the top domain of BTV VP7 (a Z score of 12.57) and also the top domain of African horse sickness virus VP7, a β-sandwich-containing protein (1) (a Z score of 4.60). These data strongly support the presence of β-sandwich structure in P8. When the N- and C-terminal domains were examined separately, no sufficiently high scoring proteins or protein segments were found. As a control, the sequences corresponding to the lower α-helical domain motif of BTV VP7 were submitted to the Fold Recognition server. No other protein with a similar fold was found to have a Z score above 5.0.

TABLE 2.

RDV P8 and BTV VP7 profile search

Protein or protein segment	Resulta	Z score
RDV P8
Entire protein	φX174 capsid protein GP	4.96
N-terminal domain (residues 1–127)	None	None
C-terminal domain (residues 316–420)	None	None
Putative β-sheet region (residues 128–315)	TNF-α	5.57
β-sheet-region (residues 128–315) minus all insertions (residues 132–133, 139–141, 163–167, 203–237, 263–265, 291)	BTV VP7; AHSV VP7, top domain	12.57
		4.60
BTV VP7
Entire protein	BTV VP7; AHSV VP7, top domain	50.16
		5.45
N-terminal helices (residues 1–216)	BTV VP7	49.31
C-terminal helices (residues 262–349)	BTV VP7	40.53
β-sheet region (residues 217–261)	BTV VP7; AHSV VP7, top domain	47.56
		16.32

^aTNF-α, tumor necrosis factor alpha; AHSV, African horse sickness virus. 

Inner shell structure (T=1 icosahedral lattice).

The inner shell particle of RDV can be chemically purified (32) and has been studied by 100-kV cryoEM and computer reconstruction (25). The 3D density map of the purified inner shell particle shows that it is composed of a thin layer, 25 Å thick, with a closely packed mass density arranged on a smooth T=1 icosahedral sphere approximately 567 Å in diameter. We have used this diameter to computationally remove the outer shell density in the RDV map (Fig. 7), which confirms that the surface mass densities of this putative inner shell are relatively smooth and indeed arranged as a T=1 icosahedral lattice.

FIG. 7.

Inner shell structure and interaction with trimers on the outer shell. (a) Inner shell computationally extracted at 590-Å diameter. It exhibits a T=1 lattice. The dashed triangle designates one triangular face of the icosahedron. (b) Schematic diagram of fish-shaped density distribution within a triangle in a T=1 lattice. (c) Capsid computationally extracted at 604-Å diameter showing the interface between the outer shell (yellow) and inner shell (blue) proteins. Color bar shows the color coding as a function of radius. (d) Schematic diagram illustrating the interaction pattern of the trimers (yellow triangles) of the outer shell with the fish-shaped densities of the inner shell.

The inner shell consists of many small holes connecting the inside to the outside of the shell. For example, a hole with a diameter of about 20 Å is located at the fivefold axis, which is surrounded by five adjacent holes 20 Å in diameter. These holes, which line up with those seen in the outer shell, may serve as the pathway of transport of RNA during the assembly and disassembly process, as is the case in rotavirus (24). The most striking feature of the inner layer are the fish-shaped mass densities about 100 Å in length, 30 Å in width, and 25 Å in thickness (Fig. 7a). In a T=1 icosahedral lattice, there are two fish-shaped densities within each asymmetric unit, yielding a total of 120 copies per capsid (Fig. 7b).

The second-most-abundant protein in the RDV capsid is the 114-kDa protein, P3, which is known to be the major protein of the inner shell (19). We propose that each fish-shaped density consists of a single P3 molecule. It is interesting that a consensus sequence of RNA polymerase activity was found within the P3 sequence, indicating that P3 may function as a structural protein in the core, as well as directly participating in RNA replication (40). It is also noteworthy that a protein of similar size, VP2, which forms the 60 dimers in the inner core shell (35), exists in rotavirus. In aquareovirus, a protein of similar molecular mass, VP3 (126 kDa), has also been found to form the T=1 icosahedral inner shell (38).

There are two minor structural proteins of RDV (P1 and P5) that are proposed to reside internally. P1 is an RNA-dependent RNA polymerase (39) and is thus likely to be located within the internal core of RDV. P5 previously was thought to be an outer capsid shell protein but recently was demonstrated to have GTP binding activity, suggesting that it may actually be located internally (31). In our present analysis, we were unable to identify the exact locations of these two proteins.

Densities between the outer and inner shells.

To examine the interface densities between the outer and inner shells, the capsid was computationally extracted at a radius slightly larger (i.e., 597 Å in diameter) than that of the putative inner shell and displayed with different colors between the two shells (Fig. 7c). Though the resolution of the reconstruction is only ∼25 Å, at which the molecular boundaries cannot be unambiguously determined, the smooth surface density of the biochemically purified inner shell justifies our choice for the putative boundary between two shells. The trimer legs in the outer shell floor density comprise the domain of interaction between the outer and inner shells. The legs of each trimer of the outer shell join together at the boundary of adjacent fish-shaped densities in the inner shell (Fig. 7c). Because the 120 fish-shaped densities are arranged on a T=1 lattice, the number of available densities varies according to the location (Fig. 7d). For example, on the icosahedral threefold axis, there are exactly three fish-shaped densities for the three legs of the trimer to attach. However, there are only two adjacent fish-shaped densities available for the legs of the trimers surrounding fivefold axes. Instead of connecting with a third fish-shaped density, one leg of the peripentonal trimer of the outer shell is directly attached to a mass density surrounding each fivefold axis. The pentonal complex of the inner shell protrudes farther outward about 10 Å; thus, the leg of the peripentonal trimer extends farther radially, resulting in the polyhedral shape of the outer shell. Therefore, it is clear from our reconstruction map that the two RDV shells have mismatched lattice symmetries.

To accommodate the mismatched lattice symmetries of the outer and inner shells, a variable interface region may be required. One possibility is that the minor proteins (for example, P2) are present as linkers at the interface between the two shells. The second possibility is that the structural variations in the floor domains of P8 facilitate various types of intersubunit interactions not only with itself but also with the inner shell proteins. Though the chemical identities of the densities that connect the two shells remain uncertain, it is conceivable that the interactions are made of complementary types of hydrogen, ionic, and hydrophobic bonding.

DISCUSSION

Most animal virus members of the Reoviridae such as rotavirus (37), animal reovirus (8), and BTV (orbivirus) have triple-shelled capsids (15, 16), while RDV has been found to have a double-shelled capsid. The double-shelled particles of these viruses are similar in capsid size and identical in triangulation number and handedness, suggesting possible structural conservation. The structural match between the HEX form of VP7 trimer of BTV and the trimer of the outer shell protein of RDV strongly favors a model of P8 that consists of two-domain motifs, as in VP7. It is interesting that the P8 trimer of RDV matches better with the HEX conformation than with the REF conformation, whereas the REF conformation has been found to match well with the outer shell structure of the BTV (14). The conformational preference in RDV could be due to the presence of a particular sequence of residues in the regions of P8 (residues 251 to 253 and 120 to 122 [Fig. 6]), which join the upper and lower domains and act as a hinge, favoring the HEX conformation.

The information on primary sequence alignment and 3D fitting suggests that the upper domain of the RDV trimer contains a β-sandwich domain, while the lower domain is arranged as a helical network. Combined with threading, the proposed β-sandwich motif in the upper domain of RDV P8 is further substantiated. However, since the HEX coordinates with the side chains are not yet in the Protein Data Bank and the structural motif of the lower domain of VP7 is considered unique, the sequence threading analysis would not be a reliable method of predicting the structure of the lower domains of RDV. In addition to the unique nature of this fold in the database, the sequences for such helical regions may not be highly conserved, leading to inconclusive results from the threading analysis. Though the Z scores for these putative helical regions are not above the threshold, this does not discount the structural motif assignments based on the conventional sequence analysis and direct 3D structure fitting.

The quasi-equivalent trimers of RDV appear to be similar but not identical with respect to mass density distribution (Fig. 1 and 3). For instance, the helical leg adjacent to the fivefold axis in the P trimer has an elongated appearance. The central dimples or holes also vary among the five trimers (Fig. 3). These structural variations seen in the quasi-equivalent subunits are not observed in the BTV trimers (14), which indicates that the quasi-equivalence rule is not strictly applicable to RDV, as in the case of BTV (14). The breakdown of quasi-equivalence in RDV may be an important mechanism of maintaining the stability and functions of the capsid.

While the structural motif of outer shell protein appears to be conserved in reoviruses, virus-host interactions are not. Evident in the primary sequence alignment between P8 of RDV and VP7 of BTV is the lack of any known signaling peptide that mediates virus-host cell interaction. There is no RGD or DGE tripeptide in P8 of both the Chinese and Japanese isolates, contrary to VP7 of BTV and VP6 of rotavirus (1, 7). Therefore, different mechanisms of cell entry for the animal and plant reoviruses are likely. Recent studies have shown that RDV particles lacking the P2 protein neither enter nor infect insect vector cells (34). Therefore, it is likely that the host selectivity stems from the differences in surface residues (as shown in Fig. 6), alternative signaling peptides, or additional signaling proteins. Overall, it may be suggested that regardless of genus, members of the Reoviridae contain a structurally conserved outer shell that may contain specific regions or residues on the surface of the capsid that would mediate viral host specificity.