Rotaviruses are enteric pathogens and a leading cause of severe diarrhea in infants and young children (
3). Like other segmented, double-stranded RNA (dsRNA) viruses of the family
Reoviridae, rotaviruses perform all stages of viral RNA synthesis within the confines of proteinaceous particles (
15,
16). The infectious virion consists of three concentric protein layers surrounding 11 unique dsRNA genome segments (
6,
20). The outermost virion layer is composed of the glycoprotein (VP7) and the spike protein (VP4), while the intermediate layer is made exclusively of VP6 (
6,
20). The innermost layer of the rotavirus particle, referred to as the core shell, is a T=1 icosahedron composed of VP2 (
9,
10,
12). Complexes containing the RNA-dependent RNA polymerase (VP1) and the capping enzyme (VP3) are located beneath VP2, in proximity of the 5-fold axes (
6,
20). The viral dsRNA genome, which codes for six structural and five or six nonstructural proteins, is thought to be organized as tubules that spool around the enzymes and pack tightly against the core shell (
3,
12). A VP1 monomer is dedicated to transcribing and replicating an associated genome segment, presumably while tethered to the VP2 interior (
5,
15). However, the role of VP2 extends beyond serving as a scaffold for the viral polymerase; it is also a critical cofactor that causes VP1 to initiate genome replication (dsRNA synthesis) (
19,
24).
The structural architecture of the rotavirus VP2 core shell has been determined at high resolution using cryo-electron microscopy and X-ray crystallography (
9,
10,
12,
27). The results show a relatively smooth icosahedron, ∼50 nm in diameter, comprised of 120 copies of VP2 organized into 12 decameric units (Fig.
1) (
9,
10,
12,
27). Each decamer is made up of two structurally distinct, but chemically identical, forms of the protein (type A and B VP2) (Fig.
1) (
9,
10,
12,
27). Five type A VP2 monomers converge around the 5-fold axis, while five type B monomers sit further back and intercalate between the type A molecules (Fig.
1) (
9,
10,
12,
27). The principal domain of rotavirus VP2 (residues ∼100 to 880) folds into a thin, comma-shaped plate, which can be structurally delineated into three subdomains (
12). The extreme amino-terminal residues of VP2 (residues ∼1 to 100 of type A and ∼1 to 80 of type B) are not fully resolved in any known structure. However, the X-ray crystallographic data for bovine rotavirus double-layered particles (DLPs) reveal a cylindrical feature underneath each 5-fold axis (
12). This internal protrusion, referred to as the 5-fold hub, is attributed to the first ∼100 residues of 10 abutting VP2 monomers. The 5-fold hub has been predicted to form a conduit for the exit of nascent plus-strand RNA transcripts following their synthesis by VP1 within transcriptionally active DLPs (
12).
Previous biochemical studies indicated that the VP2 5-fold hub is also important for interactions with VP1. Specifically, the deletion of residues 1 to 92 does not affect the capacity of recombinant VP2 to form core-like particles but does abrogate VP1 (and VP3) encapsidation into those assemblies (
9,
26). Likewise, VP2 lacking the amino-terminal 27 residues is not capable of supporting efficient VP1-mediated
in vitro dsRNA synthesis, suggesting that the 5-fold hub plays a role in polymerase activation (
19). The molar ratio of VP2 to VP1 for optimal
in vitro genome replication is 10:1, indicating that a VP2 decamer (or 5-fold hub thereof) triggers a VP1 monomer (
19,
24). In addition to VP1 binding, residues of the 5-fold hub have also been implicated in nonspecific interactions with nucleic acid (
8). Nonetheless, VP1 is capable of recognizing viral plus-strand RNA in the absence of VP2, indicating that the core shell requirement for enzyme activity is not merely a reflection of template recruitment (
11,
17,
24). In a previous study, we showed that VP1 proteins of genetically divergent rotavirus strains (SA11 and Bristol) are capable of mediating
in vitro dsRNA synthesis using each other's plus-strand RNA templates but that they require their cognate core shell proteins for enzymatic activity (
13). This result provides solid evidence that there is a direct and specific interaction between VP1 and VP2 during genome replication. VP2 is essential, along with plus-strand RNA, GTP, and Mg
2+, for the creation of a salt-stable VP1 initiation complex, which is a prerequisite for phosphodiester bond formation (
24). Based on the available data, it is hypothesized that an assembling VP2 core directly engages VP1 bound to one of 11 unique plus-strand RNA templates. This VP1-VP2 interaction induces structural changes in the polymerase that allow it to initiate dsRNA synthesis and create the viral genome (
11,
14,
23). The requirement for VP2 ensures that VP1 makes dsRNA only when inside a particle that is the morphogenic precursor to an infectious virion, thus serving to couple assembly, packaging, and genome replication. Still, the molecular mechanism by which VP2 activates VP1 remains very poorly understood.
In this study, we sought to elucidate rotavirus VP2 core shell regions critical for polymerase activation. Using amino acid sequence and structural predictions, we engineered mutant VP2 proteins and assayed them for the capacity to support VP1-mediated in vitro dsRNA synthesis. We found that although the VP2 amino terminus is important for VP1 activation, the residues that correlate with polymerase specificity are located on the inner face of the core shell principal domain. In fact, we identified multiple inner surface-exposed residues of VP2 that, when simultaneously mutated, abolished the capacity of the core shell protein to activate VP1. The results of this study enhance our understanding of how rotavirus VP2 binds to and regulates the function of the viral polymerase.
MATERIALS AND METHODS
Generation of VP1- and VP2-expressing baculoviruses.
The BaculoDirect expression system (Invitrogen) was used according to the manufacturer's protocol to create recombinant baculoviruses expressing the VP1 and VP2 proteins. Briefly, the VP1- and VP2-encoding genes were cloned into the entry vector pENTR-1A (see below) and then inserted into BaculoDirect C-Term linear DNA by recombination with LR clonase II. The baculovirus DNA was then transfected into
Spodoptera frugiperda (Sf9) cells, and recombinant virus was harvested from the medium. Baculoviruses expressing VP1 and VP2 proteins of rotavirus strains SA11 and Bristol were generated as described previously (
13).
The cDNAs for strain PO-13 VP1 (with a carboxy-terminal His tag), strain PO-13 VP2, and strain ETD VP2 were codon optimized and synthesized de novo by Geneart (Regensburg, Germany). These synthetic genes were then cloned into pENTR-1A using 5′ BamHI and 3′ NotI restriction sites. The amino acid sequences of PO-13 VP1, PO-13 VP2, and ETD VP2 exactly match sequences in the GenBank database (accession numbers AB009629, AB009630, and GQ479948, respectively).
To create pENTR-1A vectors for the generation of baculoviruses expressing SA11 VP2 with amino-terminal truncations (Δ10, Δ36, and Δ102), PCR was performed by using Platinum
Taq (Invitrogen) as the enzyme and pENTR-SA11-VP2 as the template (
13). The PCR products were cloned directly into pGEM-T Easy (Promega) and sequenced across the VP2-encoding region prior to subcloning into pENTR-1A using restriction sites (5′ EcoRI and 3′ NotI).
Blunt-end PCR was used to engineer pENTR-1A vectors for the generation of baculoviruses expressing chimeric VP2 proteins. For the subdomain chimeras (CHIMs) (CHIMs 1 to 4), the vector sequence was amplified by using outward PCR, nonphosphorylated primers, and pENTR-Bristol-VP2 as a template (
13). The insert sequence was amplified by using 5′-phosphorylated primers and pENTR-SA11-VP2 as a template (
13). For the multipoint (MP) mutants, a cDNA encoding Bristol VP2 in which nonconserved, inner surface-exposed residues were changed to those of SA11 VP2 (i.e., MP-VP2) was synthesized
de novo by Geneart (Regensburg, Germany). The synthetic gene was cloned into pENTR-1A using 5′ BamHI and 3′ NotI restriction sites to create pENTR-MP-VP2. PCR was used to create pENTR-1A vectors encoding Bristol VP2 that contained the SA11 mutations only in the apical subdomain (pENTR-MP-api), the central subdomain (pENTR-MP-cent), or both the apical and central subdomains (pENTR-MP-both). Specifically, the vector sequence was amplified by using outward PCR, nonphosphorylated primers, and pENTR-Bristol-VP2 as a template. The insert sequence was amplified by using 5′-phosphorylated primers and pENTR-MP-VP2 as a template. In all reactions, Accuprime
Pfx Supermix (Invitrogen) was used as the enzyme. The PCR products were treated with DpnI (New England BioLabs [NEB]) for 1 h at 37°C, and the cDNAs were gel purified prior to blunt-end ligation using T4 DNA ligase (NEB) for 1 h at 25°C. The final pENTR-1A vectors were sequenced across the entire VP2-encoding region to ensure the proper ligation of the insert and the absence of secondary mutations. Primer sequences used for cloning are available upon request.
Purification of recombinant VP1 and VP2.
To prepare His-tagged SA11, PO-13, and Bristol VP1 proteins, 4 × 107 adherent Sf9 cells were infected at a multiplicity of infection (MOI) of approximately 5 with the appropriate baculovirus and maintained in TNH-FH medium (Invitrogen) containing 10% fetal bovine serum (FBS) for 4 days at 20°C. Infected cells were washed twice with cold phosphate-buffered saline (PBS), resuspended in 10 ml of cold VP1 lysis buffer (25 mM NaHPO4, 200 mM NaCl [pH 7.8]), and sonicated. The insoluble fraction was removed by centrifugation at 15,000 × g for 10 min at 4°C, and His-tagged VP1 was recovered from the soluble fraction by incubation with cobalt resin (Talon) for 1.5 h at 4°C. The purified VP1 proteins bound to resin were washed using lysis buffer and then eluted using 300 μl lysis buffer containing 300 mM imidazole. Purified VP1 preparations were dialyzed against low-salt buffer (2 mM Tris-HCl [pH 7.5], 0.5 mM EDTA, 0.5 mM dithiothreitol) and stored at 4°C.
To prepare VP2 proteins, 2 × 108 spinner Sf9 cells were infected at an MOI of approximately 5 with the appropriate baculovirus and maintained in TNH-FH medium containing 10% FBS for 72 h at 28°C. Infected cells were washed twice with cold PBS and resuspended in 25 ml of cold low-salt buffer. Cells were lysed by the addition of deoxycholic acid to a 1% final concentration and by sonication. The VP2 proteins were pelleted through a 5-ml cushion of 35% (wt/vol) sucrose in low-salt buffer by centrifugation at 80,000 × g for 90 min at 12°C. Purified VP2 proteins were resuspended in 500 μl of low-salt buffer containing 1× complete protease inhibitor (Roche) and stored at 4°C.
The concentrations of the VP1 and VP2 proteins were determined by comparison with known amounts of standards electrophoresed in SDS-polyacrylamide gels and stained with PageBlue (Fermentas). The average purifications yielded approximately 2 to 10 μg of VP1 (per 4 × 107 cells) and 800 to 1,000 μg of VP2 (per 2 × 108 cells). Despite expressing equally well in Sf9 cells, the average yield of PO-13 VP1 following cobalt affinity purification was 3-fold less than that of SA11 or Bristol VP1.
Preparation of SA11 gene 8 RNA.
To generate cDNA templates containing authentic 5′ and 3′ ends for T7 promoter-driven
in vitro transcription, PCR was performed by using Accuprime
Pfx Supermix (Invitrogen) and pSP65g8R as a template (
13). The PCR-amplified cDNA was extracted twice with phenol-chloroform-isoamyl alcohol and once with chloroform and precipitated using isopropanol, prior to serving as a template for the T7 MEGAscript transcription system (Ambion) according to the manufacturer's instructions. The SA11 gene 8 products of transcription reactions were cleaned by using phenol-chloroform-isoamyl alcohol extraction and Quick Spin RNA minicolumns (Roche). The RNA quantity was determined by using a UV spectrophotometer (optical density at 260 nm [OD
260]), and RNA quality was assessed by electrophoresis in 7 M urea-5% polyacrylamide gels stained with ethidium bromide.
In vitro dsRNA synthesis assays.
Reaction mixtures contained 50 mM Tris-HCl (pH 7.1); 1.5% polyethylene glycol; 2 mM dithiothreitol; 1.5 units of RNasin (Promega); 20 mM magnesium acetate; 4 mM MnCl2; 1.25 mM each ATP, CTP, UTP, and GTP; 10 μCi of [α-32P]UTP (3,000 Ci/mmol); 8 pmol of SA11 gene 8 template; 1 to 2 pmol of VP1; and 10 to 20 pmol of VP2 (1:10 ratio of VP1 to VP2). Reactions proceeded at 37°C for 3 h, and the radiolabeled dsRNA was visualized following SDS-polyacrylamide gel electrophoresis and autoradiography. Figure images were generated by using Adobe Photoshop 9.0 and Adobe Illustrator 12.0.
Sequence and structural analyses.
The VP1 and VP2 gene sequences of murine rotavirus strain EB were determined according to a method described previously by Rippinger et al. (
22). Briefly, viral RNA was harvested from intestinal homogenates of EB-infected neonatal BALB/c mice (gift from H. Greenberg, Stanford University). The VP1- or VP2-encoding genes were amplified with the Superscript One-Step reverse transcription (RT)-PCR kit (Invitrogen) and sequenced by using the ABI Prism BigDye v3.1 terminator cycle sequencing kit and a 3730 DNA analyzer (Applied Biosystems).
Phylogenetic analyses and amino acid alignments were constructed by using MacVector 8.1.2. (Accelrys). Amino acid dendrograms were generated by using the neighbor-joining method (1,000 bootstrap repetitions) and the Poisson correction parameter. Amino acid alignments were constructed with MacVector 8.1.2 using ClustalW, BLOSUM series, with the defaults set (open-gap penalty of 10.0, extended-gap penalty of 0.05, and delay divergence of 40%). The GenBank accession numbers for all of the VP1 and VP2 sequences used in this study are listed in Table S1 in the supplemental material.
Structural analyses were performed by using the crystallographic data for the VP2 layer of bovine rotavirus strain UK (Protein Data Bank [PDB] accession number 3KZ4). Figure images were generated by using the UCSF Chimera molecular modeling system (
21).
Nucleotide sequence accession numbers.
The sequences deduced here were deposited into the GenBank database under accession numbers HQ540507 and HQ540508.
DISCUSSION
The rotavirus RNA-dependent RNA polymerase performs all stages of RNA synthesis in association with an assembled or assembling VP2 core particle (
15). Viral transcription occurs in the context of a DLP, whereby VP1 utilizes the minus strand of an associated dsRNA genome segment as a template for the creation of capped, nonpolyadenylated plus-strand RNAs (
6,
15). In addition to serving as templates for protein synthesis, plus-strand RNAs are selectively packaged into forming VP2 core shells and act as templates for dsRNA synthesis. While the order of protein and RNA interactions that occur during packaging and genome replication is not completely understood, cumulative biochemical and structural data suggest a working model (
4,
11,
18,
23,
24). First, individual copies of VP1 (with VP3) are thought to bind to the 3′ termini of the viral plus-strand RNAs, creating 11 different enzyme-RNA complexes (
4,
11,
18,
24). The polymerase interacts with viral plus-strand RNA in a sequence-specific manner but positions the template out-of-register with the catalytic site (
11). Thus, the VP1/VP3/plus-strand RNA complexes that first form in the viroplasm are thought to be catalytically inactive; the polymerases would have to undergo one or more conformational changes to synthesize dsRNA (
11,
23). The autoinhibited RNA-binding mechanism of VP1 may allow time for assortment to occur so that one of each of the 11 VP1/VP3/plus-strand RNAs is incorporated into an assembling core. Interactions among VP1, VP2, and VP3 with each other and with the inclusion-forming nonstructural proteins (NSP2 and NSP5) are also likely to be important for regulating the timing of core assembly (
1,
2). Following their assortment, VP2 would engage the polymerase component of VP1/VP3/plus-strand RNA complexes, thereby activating the enzymes to initiate minus-strand synthesis and produce dsRNA genome segments. Based on the observation that maximal
in vitro dsRNA synthesis occurs with a 10:1 molar ratio of VP2 to VP1, we hypothesize that a decamer of VP2 activates each VP1 monomer (
19,
24). Recombinant VP2 is heterogeneous in form, composed of assembly intermediates (dimers and decamers, etc.) and mixed multimers (bristly helix-like structures and sheet-like structures, etc.) in addition to T=1 core-like particles (
7,
25). Therefore, whether a decamer is sufficient to activate a VP1 monomer remains to be experimentally confirmed. Nonetheless, the notion that a fully formed icosahedral core is not essential for polymerase activation is supported by the observation that the dsRNA products of
in vitro genome replication are not encapsidated. In the process of dsRNA synthesis, however, the VP2 decamers must be brought together to form a complete, closed T=1 shell. In this model, the rotavirus core would be composed of 11 VP2 decamers with internally tethered VP1/VP3/plus-strand RNA complexes and a single empty vertex.
In the current study, we demonstrate that the VP2 core shell principal domain, rather than the amino terminus, correlates with polymerase activation specificity. This result was particularly surprising, as the VP2 amino terminus forms an internal hub beneath each 5-fold axis and has long been known to be important for VP1 interactions. Specifically, previous studies have shown that this region of the core shell protein is dispensable for particle formation but is required for (i) the encapsidation of VP1 and VP3, (ii) interactions with RNA, and (iii) efficient
in vitro dsRNA synthesis (
8,
19,
26). Consistent with those results, we found that even a minor truncation of the VP2 amino terminus had dramatic effects on the capacity of the protein to support VP1-mediated dsRNA synthesis
in vitro but did not prevent VP2/VP6 virus-like particle formation (data not shown). Moreover, chimeric VP2 proteins (SA:Br and Br:SA) that have their 5-fold hub residues replaced with those from a different strain supported lower levels of dsRNA synthesis by VP1. However, the reduced VP1 activation in the presence of the chimeric SA11 VP2 with a Bristol 5-fold hub (Br:SA) could be complemented by the addition of a Bristol dimer-forming subdomain (i.e., CHIM 4). This result was interesting, as the SA11 and Bristol VP2 amino termini are very divergent in sequence (<10% amino acid identity). It is possible that the 5-fold hub plays an indirect role in polymerase activation by helping to maintain the structural integrity of the principal domain via interactions with the dimer-forming subdomain. Deletion mutagenesis of the 5-fold hub or the substitution of a noncognate hub (without a matching dimer-forming subdomain) might have caused subtle defects in the presentation of the VP1-binding residues located in the apical and central subdomains. In support of this idea, cryo-electron microscopy studies of virus-like particles with amino-terminally truncated VP2 (bovine strain RF) revealed alterations in the vicinity of the 5-fold axis compared with the wild-type controls (
9). We are now generating reagents for the production of Bristol VP2/VP6 virus-like particles to determine whether the chimeras exhibit structural defects.
Aside from maintaining the integrity of the principal domain, the 5-fold hub might also play important roles in other stages of viral replication. For example, during core shell assembly, interactions of VP2 amino termini may serve to enhance the kinetics of decamer formation. Additionally, the hub could be required for the initial engagement of the VP1/VP3/plus-strand RNA complex, bringing the polymerase in proximity to the activation residues in the VP2 apical and central subdomains. Such a “two-step” manner of core shell-polymerase interactions would confer another level of temporal control to rotavirus packaging and genome replication. Ongoing experiments are aimed at determining whether VP2 amino-terminal residues are involved directly in core formation efficiency and/or VP1 encapsidation. However, it is important that the VP2 5-fold hub is composed of 10 abutting amino termini of 100 amino acids in length and has a molecular mass of nearly 100 kDa (
12). The diameter of the hub extends out to the first resolved helix (α0) of the type B VP2 monomers (Fig.
10) (
12). At this size, the 5-fold hub would presumably block access to the apical subdomains of the VP2 monomers, which are important for VP1 activation
in vitro. The internal 5-fold hub density attributed to VP2 amino termini has been detected only in the context of double-layered particles that have already undergone genome replication (
12). It is interesting to speculate that the amino terminus of VP2 does not form an organized hub-like structure until after enzyme activation and dsRNA synthesis. Alternatively, the fact that the 5-fold hub has been difficult to visualize in structural studies suggests that it is flexible and therefore might be capable of “bending” to accommodate VP1. It was also proposed previously that the VP2 5-fold hub serves as a conduit for the exit of plus-strand RNAs out of the core during transcription (
12). The dramatic sequence variation and high concentration of charged amino acids in the VP2 amino terminus are consistent with this idea.
To date, every rotavirus polymerase that we have tested exhibits
in vitro enzymatic activity that is dependent upon the presence of a core shell protein (
13). While the group A rotavirus VP2 proteins (categories I to III) could substitute for each other to various degrees in our experiments, the group C Bristol VP2 (category IV) was completely incapable of functioning in place of any group A VP2. The lack of obvious gene reassortment between group A and C rotavirus strains might be a reflection of such incompatible viral proteins. Furthermore, analysis of protein sequences from different strains revealed that the VP1 and VP2 trees have nearly identical topologies and phylogenetic groupings. This result indicates that the polymerase and core shell proteins have coevolved and function best when kept together. Indeed, the observation that the SA11, PO-13, and Bristol polymerases are most active in the presence of VP2 from the same strain suggests that the protein-protein interaction(s) leading to enzyme activation is fairly specific. Thus, amino acids that differ among the SA11, ETD, PO-13, and Bristol VP2 proteins must be involved in creating the binding site(s) for VP1. In an attempt to identify such residues, we created Bristol VP2 proteins in which nonconserved, inner surface residues of the apical and/or central subdomains were changed to the corresponding residue of SA11 VP2 (MP-api, MP-cent, and MP-both). Our goal was to create Bristol VP2 proteins that gained the capacity to activate SA11 VP1. Unfortunately, the mutant proteins that we generated did not support SA11-mediated dsRNA synthesis
in vitro, demonstrating that many more residues are involved. The Bristol VP2 mutants did, however, show reduced or no activation of cognate Bristol VP1
in vitro, suggesting that the identified residues are involved in this process. We cannot exclude the possibility that the introduced mutations caused gross structural defects in VP2. Due to the heterogeneous form of recombinant VP2, we are unable identify perturbations simply by visualizing the preparations using electron microscopy (
7,
25). Moreover, the biochemical insolubility of recombinant VP2 precludes classical protein-protein interaction assays (gel filtration, sucrose gradient analyses, and pulldowns, etc.) to elucidate the multimerization status or VP1-binding capacities of the mutants (
7,
25). However, because the targeted amino acids are all located on the core shell interior and have downward-pointing side chains, we think that it is likely that the mutations did directly affect the polymerase interaction(s). Given the size of VP1 (∼125 kDa; ∼70 Å in diameter), it is possible that its binding footprint (∼1,400 Å
2) spans several monomers of a VP2 decamer (Fig.
10). The quasi-equivalent positioning of type A and B VP2 monomers also implies that both the apical and central subdomains are involved in VP1 binding and activation. Nevertheless, future studies are needed to more precisely identify the core shell amino acids critical for multifaceted interactions with the rotavirus polymerase.