Atomic model of CPV reveals the mechanism used by this single-shelled virus to economically carry out functions conserved in multi-shelled reoviruses

Summary

Unlike the multi-shelled viruses in the Reoviridae, cytoplasmic polyhedrosis virus (CPV) is single shelled, yet stable and fully capable of carrying out functions conserved within Reoviridae. Here, we report a 3.1-Å resolution cryo electron microscopy structure of CPV and derive its atomic model, consisting of 60 turret proteins (TP), 120 each of capsid shell proteins (CSP) and large protrusion proteins (LPP). Two unique segments of CSP contribute to CPV’s stability: an inserted protrusion domain interacting with neighboring proteins and an N-terminal anchor tying up CSPs together through strong interactions such as β-sheet augmentation. Without the need to interact with outer shell proteins, LPP retains only the N-terminal two-third region containing a conserved helix-barrel core and interacts exclusively with CSP. TP is also simplified, containing only domains involved in RNA capping. Our results illustrate how CPV proteins have evolved in a coordinative manner to economically carry out their conserved functions.

Introduction

Cytoplasmic polyhedrosis virus (CPV) belongs to the Cypovirus genus within the family Reoviridae. Viruses in this family have an inner core that is a self-competent molecular machine capable of endogenous RNA transcription/processing/releasing (Reviewed by Mertens et al., 2004; Zhou, 2008). Most members of the Reoviridae, such as orthoreoviruses, aquareoviruses, phytoreoviruses and rotaviruses, have a capsid with one to two outer layer(s) enclosing the inner core. In addition to conferring host specificity and mediating cell entry, these additional layers are believed to play important structural roles in maintaining the stability of the thin inner capsid shell (Lawton et al., 2000).

CPV is distinctive in having only a single-shelled capsid. Despite lacking protective outer shells, CPV exhibits striking capsid stability (Reviewed by Mertens et al., 2004; Zhou, 2008). While its infection of silkworms can have a negative economic impact in Asia, CPV is better recognized as a biocontrol agent, serving as an environmentally friendly pesticide in fruit and vegetable farming (Mertens et al., 2000).

120 capsid shell protein (CSP) molecules form the single CPV capsid shell, which is decorated by 12 turrets (the mRNA capping complexes) on the icosahedral vertices and by 120 molecular clamps (large protrusion proteins, LPPs), as opposed to the 150 copies of the structural homolog σ2 in orthoreovirus (Dryden et al., 1993; Hill et al., 1999; Zhang et al., 1999; Reinisch et al., 2000), that cement neighboring CSP molecules. CPV contains specialized spikes (A spikes) attached to the turrets (B spikes). The A spike is believed to function in mediating cell attachment and penetration. The CPV capsid is architecturally similar to the non-infectious inner cores of orthoreovirus and aquareovirus, and yet, is infectious (Payne and Harrap, 1977).

Many other Reoviridae viruses have been subjected to structural investigations by both cryo electron microscopy (cryoEM) (e.g. Prasad et al., 1990; Lawton et al., 1997; Zhou et al., 2001; Zhang et al., 2003; Zhou et al., 2003; Zhang et al., 2008; Cheng et al., 2010; Zhang et al., 2010) and x-ray crystallography (e.g. McClain et al.; Grimes et al., 1998; Reinisch et al., 2000). We have previously used cryoEM to obtain a CPV reconstruction at near-atomic-resolution (Yu et al., 2008). This reconstruction has permitted the construction of only backbone models (not full atom models) and only for portions of CSP and about 1/3 of TP. Notably, for structures that are directly responsible for capsid stability, including the cement protein LPP and the CPV-unique protrusion domain of CSP, even backbone models were not possible due to the limited resolution of the previous structure. As a result, the structural basis of CPV stability remains unclear.

Here, we have determined the three-dimension (3D) structure of CPV to 3.1 Å resolution by using improved image acquisition and data processing methods. The new map has allowed us to resolve all types of amino acids and enabled us to build full atom models for all the structural proteins of CPV, except for the non-icosahedrally arranged A spike protein. These atomic models reveal molecular interactions among CPV capsid proteins and explain the structural basis for the unparalleled stability of CPV (Zhang et al., 2002), as compared to multi-shelled members of the Reoviridae family. Our structures show that CPV has evolved in two opposite directions with respect to the complexities of structural proteins, leading to a stable yet fully functional single-shelled virus. CPV CSP has become more complex by acquiring an additional structural domain to strengthen its interactions with other capsid proteins. In addition, CPV CSP-B has a figure-8-like N-terminal anchor (N-anchor), which is dramatically different from its counterparts in multi-shelled viruses in the Reoviridae. Conversely, both LPP and TP are simplified. Without the need to mediate interactions with outer shell proteins, CPV cement protein LPP only retains the N-terminal region with a conserved helix-barrel core. The conserved helix-barrel core identified here is likely to play a key role in maintaining its structural integrity as a stabilizing cement protein.

Results

CryoEM structure determination

In an attempt to decipher the molecular mechanism underlying the extraordinary stability of the single-shelled CPV, we obtained a cryoEM reconstruction of the CPV at 3.1 Å resolution and derive full atom models for all the structural proteins that make up the CPV capsid shell. As detailed in Methods, we have made improvements in two main areas to obtain the improved structure. First, our data are now obtained on photographic film in a Titan Krios cryo electron microscope, one of the first of FEI’s new generation of high-resolution cryoEM instruments, with parallel beam illumination and minimized beam tilt. Second, we improved data processing by using our latest reconstruction program in IMIRS (Liang et al., 2002; Liu et al., 2008). We selected 646 from a total of 996 cryoEM micrographs (e.g., Figure 1A). All of the selected micrographs are free of specimen drift/charging and show visible rings of contrast transfer function (CTF) in their power spectra beyond a spatial frequency of 1/6 Å⁻¹.

Figure 1. CryoEM and 3D reconstruction of cytoplasmic polyhedrosis virus (CPV) at 3.1 Å resolution.

(A) A representative cryoEM image of CPV particles (white) embedded in vitreous ice, recorded on Kodak SO 163 film in an FEI Titan Krios cryo electron microscope operated at 300kV at liquid-nitrogen temperature.

(B) Radially colored, shaded surface representation of the CPV reconstruction at 3.1 Å resolution as viewed along a twofold axis. Three symmetry axes one twofold, one threefold, and one fivefold–are indicated by “2”, “3”, and “5”, respectively.

(C) A zoom-in view of the boxed region of

(B). See also Figure S1–4

The final map (Figure 1B and Figure S1A) was reconstructed by combining 28,993 particle images (Table 1). We estimate the effective resolution of the final reconstruction to be ~3.1 Å, based on the 0.143 Fourier shell correlation coefficient (FSC) criterion of Rosenthal and Henderson (Rosenthal and Henderson, 2003) (~3.5Å based on 0.5 FSC criterion; Supplementary Figure S1B). This resolution was further validated by structural features revealed in the cryoEM map. At 3.1 Å resolution, side chain densities of all types of amino acids, except glycine, should be visible (Figure 2 and Figure S1–S4) and the three types of aromatic amino acids should be distinguishable based on their characteristic side chain densities (Figure 2C). For example, compared to Phe, Tyr shows an extra narrowed tip density contributed by the OH group extended from the aromatic ring (Figure 2C). Furthermore, carboxyl oxygen atoms can be identified not only in strands and loops, but also in helices (Figure 2C). Using crystallographic modeling tool COOT (Emsley and Cowtan, 2004), we built an ab initio full atom model for CPV (Figure 2A). Our CPV atomic model contains two conformers of CSP, two conformers of LPP and one conformer of TP (Figure 2A–B and Figure S2–S4).

Table 1.

CryoEM Imaging and Data Processing Statistics

Films recorded	996
Films used	645
Particles boxed	73596
Particles included in the final reconstruction based on phase residue	28993
Defocus range (μm)	1.15–3.45
B-factor (Å2)	240
Resolution* (Å)	3.1

^*Based on Rosenthal and Henderson 0.143 FSC criterion (Rosenthal and Henderson, 2003).

Figure 2. Atomic model of CPV.

(A, B) Atomic model of CPV capsid (A) and its asymmetric unit (B), color-coded by protein subunits. TP is in red; CSP has two conformers: CSP-A (yellow) and CSP-B (magenta); and LPP also has two conformers: LPP-5 (cyan) and LPP-3 (blue).

(C) Density maps superimposed with atomic models, demonstrating the quality of our cryoEM map and that of the refined atomic model. (Top panel) Atomic models (stick) of an α helix from CSP-B (left) and a β strand from CSP-A (right) are superimposed in their density maps (mesh). (Bottom panel) Representative side chain densities (mesh) superimposed with their atomic models (stick). The three aromatic amino acids, Trp (from LPP-5, left), Phe (from LPP-3, middle), and Tyr (from CSP-B, right) are distinguishable. Three carboxyl oxygen atoms are indicated by arrows.

Atomic models of CSP-A and CSP-B

Each CPV capsid contains 120 CSP monomers in two conformations: CSP-A and CSP-B. Sixty copies of each of CSP-A and CSP-B conformers interdigitate one another to form a thin capsid shell. Five CSP-A molecules are organized around each five-fold axis and three CSP-B molecules surround each three-fold axis (Figure 2A–B).

Each CSP monomer contains four domains: an apical domain (residues 405–827, 960–1048), a middle domain (residues 145–404, 1049–1069, and 1239–1333), a dimerization domain (residues 1070–1238), and a small protrusion domain (SPD, residues 828–959) (Figure 3). At the sequence level, apical and dimerization domains can be thought of as inserts of the middle domain at its N-terminus and C-terminus, respectively. While the apical, middle and dimerization domains form a large, thin plate-like structure that is common in all members of the Reoviridae, the SPD is unique to CPV. SPD is an insert into the apical domain near its N-terminal end (Figure 3A, D) and has a fold of helices/sheet/helices sandwich (Figure 3C). The SPD of CSP-A interacts with its neighboring CSP-B (Figure 4), involving residues D828, S829 and V945 in the SPD of CSP-A with residues T643, V644 and T645 in the apical domain of CSP-B (Figure 4B).

Figure 3. Atomic models of CSP-A and CSP-B.

(A, B) Two orthogonal views (A: viewed from outside of the virus; B: 90° rotated from A) of CSP- A atomic model, color-coded by domains: apical (residues 405–827, 960–1048) is green, small protrusion (residues 828–959) is red, middle (residues 145–404, 1049–1069, 1239–1333) is goldenrod, and dimerization (residues 1070–1238) is cyan. The short N-terminal loop (residues 135–144) is colored in magenta.

(C) Zoom-in view of the CPV-unique small protrusion domain (SPD). SPD has a fold of helices/sheet/helices sandwich. The model is color-coded by residues from blue at N terminus through green and yellow to red at C terminus.

(D, E) Outside (D) and inside (E) views of CSP-B atomic model. Here, corresponding domains are colored the same as those in CSP-A shown in (A, B). The figure-8-like N-terminal anchor (residues 74–144) is colored in magenta.

(F) The structure of the N-anchor rotated to reveal the pointy turn (i.e., amino acids T95-V102) and β strands. The N-anchor is composed of four strands and five loops.

See also Figure S2

Figure 4. Interactions between the small protrusion domain (SPD) of CSP-A and the apical domain CSP-B.

(A) Overall view of SPD (red) from a CSP-A subunit (blue) interacting with a neighboring CSP-B subunit (green) viewed from outside.

(B) Zoom-in view of the boxed region in (A), showing the interactions between the SPD of a CSP-A and the apical domain of a neighboring CSP-B. The amino acids involved in the interactions are labeled.

The most striking structure of CSP-B is the N-terminal (residues 74–144) figure-8-like anchor (N-anchor), located underneath the capsid shell. Each N-anchor contains four β strands and five loops (Figure 3F). As shown in Figure 5A, the N-anchor of the green CSP-B interacts with its neighboring molecules on both the left and right sides of the green CSP-B (green in Figure 5A–B), spanning a wide region across one CSP-A (blue in Figure 5A–B) and three CSP-B molecules (Figure 5A–B).

Figure 5. Interactions among CPV capsid proteins around a three-fold axis.

(A) Overall inside view of three CSP-B subunits (green, magenta, and goldenrod), one CSP-A subunit (blue), and one LPP-3 subunit (red) involved in molecular interactions. The N-anchor from one CSP-B subunit (green) is displayed with thicker ribbon.

(B) Zoom-in view of the boxed region in (A), showing a CSP-B (green) N-anchor (thicker ribbon) interacting with regions of two neighboring CSP-B subunits (magenta and goldenrod), one CSP-A subunit (blue), and one LPP-3 subunit (red). The dotted ellipse indicates the N-anchor’s N-terminal loop fragment (residues 74–76) interacts with another N-anchor (goldenrod) at its loop region immediately adjacent to the turn region. The dotted circle indicates the N-anchor pointy turn, which pierces through the capsid shell and interacts with the LPP-3 (red) located above (see also Figure 3F).

(C) Rotated zoom-in view from the left dash box region in (C), showing the N-anchor β strand 1 (residues 78–82) is joined by three strands from the middle domain of its own CSP-B, thus firmly fix the N-anchor to its main body.

(D) Rotated zoom-in view from the middle dash box region in (C), showing the N-anchor β strand 2 and 4 (residues 112–114 and 133–136) are joined by a strand from the neighboring CSP-B middle domain (goldenrod) to form a three-strand β sheet through β augmentation.

(E) Rotated zoom-in view from the right dash box region in (C), showing the N-anchor β strand 3 (residues 120–124) is joined by a strand from the CSP-A dimerization domain (blue) to form a two-strand β sheet through β augmentation.

(F) Zoom-in view of a boxed region from (A), showing β augmentations among the three CSP-B molecules around the three-fold axis. Each CSP-B contributes two strands from its middle domain to form three identical two-strand β sheets through β augmentation (indicated by dash boxes). The threefold axis is indicated by a black triangle.

To the left, as shown in Figure 5A, the N-anchor of the green CSP-B uses a short loop fragment of four residues (74–77) to interact with the N-anchor (dotted ellipse in Figure 5B) of a neighboring CSP-B (magenta in Figure 5A–B). It then fastens itself to the main body of its own (green) CSP-B by running across the middle domain using residues 78–94 and forming a strand (i.e., β1, residues 78–82) of a 4-stranded β sheet (Figure 5B–C).

To the right, the rest (residues 95–144) of the N-anchor is involved in interactions with three neighboring molecules: one CSP-B, one CSP-A and one LPP-3 (Figure 5B). First, a loop region (residues 95–103, within the dotted circle in Figure 5B) forms a pointy turn. This pointy turn pierces through the shell and attaches residues V99 and D100 to LPP-3 located on the outside of the capsid (Figure 3F and Figure 5B). This segment also interacts with the middle domain of a neighboring CSP-B (yellow in Figure 5B). Second, two β strands (residues 112–114 of β2 and residues 133–136 of β4) interact with the same neighboring CSP-B (yellow in Figure 5A–B) by forming a three-stranded β sheet through β augmentation (Figure 5D). Third, strand β3 (residues 120–124) interacts with the dimerization domain of a neighboring CSP-A (blue in Figure 5A–B) by forming a two-stranded β sheet also through β augmentation (Figure 5E).

Amazingly, β augmentation is again used by the CSP-B molecules around the 3-fold axis (Figure 5F). There, six strands from the middle domains of the three CSP-B molecules, one on either side of each middle domain, form three identical two-stranded β sheets and join the 3-fold related CSP-B molecules together (Figure 5F).

Unlike the N-anchor of CSP-B, only a short N-terminal loop of 10 amino acids (residues 135–144) is resolved in CSP-A (Figure 3A). This short loop interacts with the apical domain of a neighboring CSP-B.

The structures of CPV CSP molecules exhibit significant differences from its structural homologs in multi-shelled turreted reoviruses. Prominent examples of the latter include the animal reovirus and aquareovirus, whose structures have been determined to atomic resolution by X-ray crystallography (Reinisch et al., 2000) and single-particle cryoEM (Zhang et al., 2010), respectively. As described above, CSP of CPV has an SPD that provides extra interactions with neighboring CSP and LPP-5 molecules to strengthen the capsid shell. This SPD is absent in λ 1 of animal reovirus and VP3 of aquarevirus, the structural homologs of CPV CSP in these multi-shelled reoviruses. Similarly, animal reovirus λ 1B and aquareovirus VP3B both lack the figure-8-like N-anchor of CSP-B, which forms an extended network of molecular interactions on the inner surface of the CPV capsid. The molecular interactions afforded by these additional structures in CPV provide the structural basis for the stability of the single-shelled capsid. Indeed, among all turreted members of the Reoviridae, CPV is the only known virus whose CSP alone can assemble into capsid-like particles in vitro (Hagiwara and Naitow, 2003).

Atomic model of LPP conformers and their interactions with CSP

Each CPV contains 120 LPP molecules in two distinct conformations (conformers): 60 LPP-3 conformers and 60 LPP-5 conformers. LPP-3 and LPP-5 lie near the three-fold and five-fold axis, respectively (Figure 2A). The two conformers have similar globular structures, with only minor differences, which are located mostly in loop regions (Figure 6B). The overall root mean squared deviation (RMSD) is 0.79 Å between the atomic models of LPP-3 and LPP-5.

Figure 6. Atomic models of CPV LPP-3 and LPP-5.

(A) Ribbon diagram of LPP-3 atomic model with secondary structural elements labeled. The model is color-coded by residues, from blue at N terminus through green and yellow to red at C terminus.

(B) Superimposition of LPP-3 (green) and LPP-5 (red), revealing the nearly identical structures of the two conformers of LPP.

(C) Amino acid sequence of LPP with secondary structural elements shown above, which are derived from the atomic model of LPP-3.

See also Figure S3

LPP-5 and LPP-3 interact with their underlying CSP molecules in specific, but non-equivalent ways. LPP-5 interacts with two CSP molecules: one CSP-A (blue in Figure 7A) and one CSP-B (green in Figure 7A). LPP-3 interacts with three CSP molecules: one CSP-A (blue in Figure 7B) and two CSP-Bs (green and yellow in Figure 7B). Despite these different contact surfaces on the CSP shell (compare the upper panels of Figure 7A with that of Figure 7B), the contact surfaces on the two LPP molecules are almost identical (compare the lower panel of Figure 7A with that of Figure 7B), but differ in the residues involved in these interactions (compare the highlighted residues between the lower right panels of Figure 7A and Figure 7B).

Figure 7. Interactions between LPP and CSP.

(A) Interaction between LPP-5 and CSPs. (Top left) Overall view of one LPP-5 subunit (red) interacting with one CSP-A (blue) and one CSP-B (green). (Top right inset) Zoom-in view of the boxed region with LPP-5 removed, showing the CSP-A (blue) and CSP-B (green) interaction regions (highlighted in red) with LPP-5. The CPV-unique SPDs from both CSP-A and CSP-B participate in interactions with LPP-5, as indicated by two dashed ellipses. (Bottom insets) LPP-5 and its flipped zoom-in views, showing interaction regions that are highlighted in blue (indicating interaction with the blue-colored CSP-A) and in green (indicating interaction with the green-colored CSP-B).

(B) Interaction between LPP-3 and CSPs. (Top left) Overall view of one LPP-3 (red) interacting with one CSP-A (blue) and two CSP-B molecules (goldenrod and green). (Top right inset) Zoom-in view of the boxed region with LPP-3 removed, showing the CSP-A (blue) and CSP-Bs (goldenrod and green) interaction regions (highlighted by red color) with LPP-3. (Bottom insets) The LPP-3 subunit and its flipped zoom-in views, showing interaction regions that are highlighted in blue (indicating interaction with the blue-colored CSP-A), green and goldenrod (indicating interactions with the green- and goldenrod-colored CSP-B subunits, respectively). For those residues interacting with both two CSP-B subunits are colored in cyan (arrow in the bottom right inset).

Comparison of LPP with the cement proteins of other reoviruses

The structural homologs of LPP include orthoreovirus σ2 protein (417 amino acids, residues 2–418) and aquareovirus VP6 protein (411 amino acids, residues 2–412). The atomic structures of σ2 and VP6 are almost identical, both significantly larger than the structure of LPP. Hagiwara et al have shown that the LPP is the cleavage product from a 448 amino-acid protein coded by genomic RNA segment 7 (Ref. Hagiwara and Matsumoto, 2000). Consistent with this biochemical study, our cryoEM map reveals an LPP structure with only 290 amino acids (residues 3–292).

Existing sequence alignment methods failed to identify any sequence similarities between LPP and σ2/VP6. However, careful 3D structural comparisons between LPP and VP6 revealed a fold of helix-barrel core that is conserved among these proteins. In LPP, this central helix- barrel core is surrounded by peripheral structures, including a 4-stranded β sheet, 4 α helices, and many loops (Figure 6). The helix-barrel core is formed by six helices (H2, H3, H4, H7, H9 and H10), four (H3, H4, H9 and H10) of which run roughly perpendicular to the capsid shell and the other two (H2 and H7) lie roughly parallel to the capsid shell (Figure 8B).

Figure 8. Structural comparison between CPV LPP and aquareovirus VP6.

(A) Atomic model of aquareovirus VP6B (Zhang et al., 2010). The N-terminal segment (residues 2–295) is color-coded by residues from blue at N terminus through green to yellow at C terminus. The C-terminal segment (residues 296–412) is colored in red, which is facing to the outside and is accessible to outer shell proteins. VP6B contains a helix-barrel core, composed of one helix-like coil (C1) and five helices of H1, H3, H5, H6 and H7 (Top right and middle right insets).

(B) Atomic model of CPV LPP-3, colored as in Figure 6A. CPV LPP-3 contains a helix-barrel core is composed of 6 helices (H2, H3, H4, H7, H9 and H10) (Bottom inset).

(C) Superposition of CPV LPP-3 helix-barrel core (cyan) and the aquareovirus VP6B helix-barrel core (green).

(D) Amino acid sequence of VP6 with secondary structural elements shown above, which are derived from the atomic model of VP6B. The sequence of the helix-like coil (C1) is shown in cyan. The C-terminal segment (residues 296–412, shown in red) of VP6 forms the outer-shell-binding structures (red in A) that are absent from CPV LPP.

The helix-barrel core in aquareovirus VP6 is composed of one coil and 5 helices (Figure 8A). Although the orientations of the helices in the barrel cores differ between CPV LPP and aquareovirus VP6, they align the best with each other after the aquareovirus VP6 structure is rotated about 130 degrees along its long axis, as indicated in Figure 8. The peripheral structures surrounding the helix-barrel core in CPV LPP and those in aquareovirus VP6 bear no similarities, consistent with the different specific interactions LPP and VP6 have with other capsid proteins. In CPV, the C-terminal half of LPP is cleaved off and not present in the capsid. The corresponding C-terminal segment in multi-shelled viruses interacts with outer shell proteins. In aquareovirus, the C-terminal segment (residues 296–412) of VP6 is exclusively located on the outside, making it accessible to the outer shell proteins (Figure 8A, C). It is conceivable that structural integrity is essential for a cement protein, whose role is to stabilize capsid structures. We suggest that the conserved helix-barrel core we identified here plays the role of maintaining structural integrity of the cement proteins in the Reoviridae. The peripheral structures surrounding the helix-barrel core vary across different viruses, reflecting their specific interactions with inner shell proteins, and as in multi-shelled reoviruses, with the outer layer proteins.

Atomic model of TP

At each icosahedral vertex, five TP monomers form a pentameric turret that serves as the mRNA capping complex of CPV. Each TP has four domains: a guanylyltransferase (GTase), two methylase domains (Methylase-1 and -2), and a bridge domain (Figure 9A).

Figure 9. Atomic model of the turret protein (TP).

(A) Atomic model of TP, color-coded by domain: guanylyltransferase (GTase) (residues 1–362) is green, bridge (residues 363–406, 727–828) is cyan, methylase-1 (residues 407–726) is yellow, and methylase-2 (residues 829–1057) is red.

(B) Atomic model of GTase, color-coded by sub-domain: the smaller N-terminal sub-domain 1 is colored in red, and the larger C-terminal sub-domain 2 is colored in green.

See also Figure S4

The N-terminal 1–362 residues form the GTase domain, which can be further divided into two sub-domains: sub-domain 1 and sub-domain 2 (Figure 9). Residues 1–159 form the smaller, sub-domain 1, which contains a two-stranded β-sheet and 7 short helices (Figure 9B). Residues 160–362 form the larger, sub-domain 2, which contains a five-stranded β-sheet and six long helices (Figure 9B). The two β sheets in the GTase domain line the cleft that separates the two sub-domains. Overall, the GTase domain has a ‘C’ shape with its bottom sitting on the apical domain of CSP-A. Its active center is located at the cleft between the two sub-domains, coupled to the nascent RNA releasing hole as described previously (Yu et al., 2008). This cleft is situated underneath the methyltransferase-1 domain of a neighboring TP subunit. The full atom model clearly reveals that the N terminal end is an integral part of the GTase. In fact, all residues starting from residue 1 to 1057 are well resolved in our cryoEM density map. Based on biochemical data of chimeric proteins containing various segments of TP N-terminal sequence, a “polyhedrin-binding domain” was first proposed to be within residues 179 (Ref. Ikeda et al., 2006) and subsequently narrowed to be within residues 42–93 (Ref. Mori et al., 2007; Ijiri et al., 2009). Our atomic model shows that residues 42–93 are located in the interior of the turret and are completely buried in the turret complex (Figure 9B), making the domain inaccessible to polyhedrin-binding. This atomic model represents a correction of our previous assignment of the hypothetical “polyhedrin-binding domain” based on a lower resolution map (Yu et al., 2008). The incorrectly assigned domain is now part of the bridge domain. In the lower resolution map, the density of that part was not clearly separated from the density of the N-terminal region of the GTase domain of a neighboring TP. This ambiguity, coupled with the published biochemical data regarding the hypothetical “polyhedrin-binding domain” (Mori et al., 2007; Ijiri et al., 2009), has led to the previous incorrect assignment. Consistent with this assignment, a recent 3D electron tomographic study shows that TP does not interact with polyhedrin inside polyhedra (Chen et al., 2011).

Residues 407–726 and residues 829–1057 form the methylase-1 domain and methylase- 2 domains, respectively. Both methylase domains have a fold of helices/7-strand sheet/helices sandwich (Zhou et al., 2003), which is similar to those of orthoreovirus λ2 (Reinisch et al., 2000).

The bridge domain, comprising residues 363–406 and residues 727–828, connects the GTase domain to the two methylase domains. Its center is composed of a 4-stranded β sheet, sandwiched by a group of helices on the outside of turret and a two-stranded β sheet facing the central chamber of turret.

The λ2 protein of orthoreovirus is composed of seven domains with three C-terminal immunoglobulin-like domains forming a flap on top of the turret (Reinisch et al., 2000). The CPV TP lacks this flap structure and is composed of only the four consecutive domains described above. For orthoreovirus, the flap structure has two functions: anchoring the cell attachment protein σ1 in virions and intermediate subviral particles (ISVPs) and serving as a gate to retard exit of the mRNA from the core (Reinisch et al., 2000). In CPV, these three functional entities (virion, ISVP and core) have become a single one, i.e., a three-in-one structure; therefore its TP has evolved to retain only the four domains related to the conserved function of mRNA capping. In the absence of the flap structure in CPV, two loop fragments (480–483 and 507–516) from methylase-1 of TP participate in anchoring the CPV-unique A-spike, which performs the dual functions as a possible cell attachment protein (similar to σ1) and as a gate to retard premature mRNA exit.

Discussion

By reconstructing the CPV structure to 3.1Å resolution, we built full atom models for all the three structural proteins that make up the capsid. Our structure suggests that CPV has evolved in two opposite directions with respect to the complexities of its structural proteins, leading to formation of a stable single-shelled capsid conserving functions carried out by multi-shelled members of the Reoviridae.

On one hand, compared to its counterparts in multi-shelled reoviruses, CPV’s capsid shell protein has gained a higher level of complexity, by acquiring an additional structural domain, SPD, and by adapting a novel figure-8-like N-terminal anchor. The extra interactions among CSPs and LPP involving SPD may contribute to stabilizing CPV’s capsid shell. The N-anchor of CSP-B is composed of only loops and strands, and lacks the helix-containing zinc finger motifs found in the N-terminal arms of CSP-B homologs in multi-shelled turreted reoviruses (Reinisch et al., 2000; Cheng et al., 2010; Zhang et al., 2010). The zinc finger motifs are believed to specifically interact with nucleic acid. Thus, the N-anchor of CSP-B seems to be functionally specialized to interact only with capsid proteins so as to form a more rigid capsid shell, which is likely critical to the stability of the single-shelled CPV. In CPV, both CSP-A and CSP-B monomers have a disordered N-terminal end (residues 173 for CSP-B, residues 1–134 for CSP-A) that might participate in interacting with viral RNA both during capsid assembly and in mature capsids.

On the other hand, the other two CPV capsid proteins are simplified as compared to their homologs in multi-shelled viruses. First, the C-terminal segments of the cement proteins in multi-shelled reoviruses are responsible for interacting with outer shell proteins (Cheng et al., 2010). Without the need to mediate interactions with outer shell proteins, the C-terminal segment of CPV LPP is cleaved off, and is not present in matured capsids (Figures 6–8). Second, because A-spike, which is the putative cell attachment and penetration protein in CPV, always plugs into the turret and can function to retard mRNA exit, there is no need for a flap structure in TP to perform these tasks in CPV. Indeed, TP has let go of such a flap structure and only retains those domains required for RNA capping.

Notably, the levels of molecular interactions among CPV capsid proteins are balanced in a coordinative way at the two regions with extensive molecular interactions, centered around LPP-3 and LPP-5, respectively. At the region around the fivefold axis, each LPP-5 is near the SPDs of both CSP-A and CSP-B and interacts with both SPDs (Figure 7A). Here, the SPD of CSP-A also interacts with a neighboring CSP-B (Figure 4). At the region around the 3-fold axis, each LPP-3 is more than 10-Å away from the SPDs of both CSP-A and CSP-B, thus having no direct interaction with an SPD. In addition, the SPD of CSP-B does not interact with its neighboring CSPs either. Accordingly, under this region, each N-anchor of CSP-B ties up four CSPs together on the internal surface of the capsid right underneath LPP-3. Therefore, the alternative utilization of different molecular interactions remarkably achieved a balance conducive of assembling a stable single-shelled capsid of CPV.

In summary, our structure reveals how the three structural proteins of CPV, a single-shelled member of the Reoviridae, have evolved in a coordinative manner to economically and efficiently carry out functions conserved in multi-shelled reoviruses.

Methods

Virus purification

CPV particles were purified from polyhedra as previously described (Yu et al., 2008). Briefly, purified polyhedra were treated with an alkaline solution of 0.2 M Na₂CO₃-NaHCO₃ (pH 10.8) for 1 hour. The suspension was centrifuged at 10,000g for 40 minutes. The supernatant was then collected and centrifuged at 80,000g for 60 minutes at 4°C to spin down CPV capsids. The capsids were resuspended in 10mM PBS (pH7.4), immediately flash-frozen on Quantifoil cryoEM grids, and stored in a liquid nitrogen dewar for future cryoEM imaging.

CryoEM imaging and 3D reconstruction

Previously frozen cryoEM grids were loaded into an FEI Titan Krios cryo electron microscope for cryoEM imaging at 300kV and liquid-nitrogen temperature. Before imaging, electron beam tilt was carefully minimized by coma-free alignment. CryoEM images were recorded on Kodak SO163 films at a dosage ~25 electrons/Ų and 59,000x nominal magnification with parallel beam illumination and an intended defocus value ranging from 1.5μm to 2.5μm. The films were digitized with a Nikon Super CoolScan 9000 ED scanner at a step size of 6.35 μm/pixel, corresponding to 1.076 Å/pixel at the sample level. The final pixel size was calibrated to be 1.104 Å/pixel by using tobacco mosaic virus as a standard.

We recorded a total of 996 micrographs and selected 645 micrographs that clearly showed signals beyond 1/6Å⁻¹ as judged by the visible CTF rings in their power spectra. Individual particle images (960×960 pixel) were first boxed out automatically by the autoBox program in the IMIRS package (Liang et al., 2002), followed by manual screening with the EMAN boxer program (Ludtke et al., 1999) to keep only well-separated, contamination-free, genome-containing intact particles.

The program CTFFIND (Mindell and Grigorieff, 2003) was used to determine the defocus value and astigmatism parameters for each micrograph. We determined particle orientation, center parameters and subsequent 3D reconstruction with the IMIRS package (Liang et al., 2002), enhanced by icosahedral symmetry-adapted functions for 3D reconstruction (Liu et al., 2008). We considered astigmatism during CTF correction in the steps of orientation/center refinement and 3D reconstruction.

We assessed the effective resolution with the reference-based Fourier shell correlation (FSC) coefficient as defined by Rosenthal and Henderson (Rosenthal and Henderson, 2003). The map was deconvolved by a temperature factor of 240Ų (conventional definition) to enhance higher resolution features. The final reconstruction was low-pass filtered to 3.1Šresolution.

Atomic model building, model refinement and 3D visualization

At 3.1 Å resolution, ab initio atomic model building is straightforward. First, we traced the chain and built Cα models based on the clear bumps for Cα atoms using the Baton_build utility in program COOT (Emsley and Cowtan, 2004). Second, amino-acid registration was accomplished solely based on the clear densities of side chains, which serve as ‘landmarks’. Third, we built coarse full-atom models in COOT with the help of REMO (Li and Zhang, 2009).

These coarse full-atom models were then refined in a pseudo-crystallographic manner using CNS (Brunger et al., 1998). This procedure only improves atomic models and does not modify the cryoEM density map. Briefly, densities for individual proteins were segmented, put in artificial crystal lattices, and then used to calculate their structure factors with the utility program em_map_to_hkl.inp in CNS (Brunger et al., 1998). The amplitudes and phases of these structure factors were used as pseudo-experimental diffraction data for model refinement by crystallographic conjugate gradient minimization and simulated annealing refinement in CNS, with group B-factors for amino acid backbones and side-chains refined also in CNS. To improve interactions between protein subunits, we put the refined structures of all five subunits in an asymmetric unit into a single coordinate file and pseudo-crystallographically refined them simultaneously with their non-crystallographic symmetry (NCS) ghosts governed by the icosahedral strict NCS constraint. This refinement uses pseudo-experimental diffraction data generated from the 3.1 Å cryoEM map of the full virus. Eventually, we obtained full-atom protein models that could match the cryoEM density well and also have “good geometry”, as defined by the force-field implemented in CNS. The final R-factor for the entire virus up to 3.1 Å resolution is 27.30 (R-free: 27.32).

CryoEM reconstruction was visualized and segmented using Chimera (Pettersen et al., 2004). Figures were prepared with Chimera (Pettersen et al., 2004) and Molscript (Kraulis, 1991).