Crystallographic Analysis of Rotavirus NSP2-RNA Complex Reveals Specific Recognition of 5′ GG Sequence for RTPase Activity

Abstract

Rotavirus nonstructural protein NSP2, a functional octamer, is critical for the formation of viroplasms, which are exclusive sites for replication and packaging of the segmented double-stranded RNA (dsRNA) rotavirus genome. As a component of replication intermediates, NSP2 is also implicated in various replication-related activities. In addition to sequence-independent single-stranded RNA-binding and helix-destabilizing activities, NSP2 exhibits monomer-associated nucleoside and 5′ RNA triphosphatase (NTPase/RTPase) activities that are mediated by a conserved H225 residue within a narrow enzymatic cleft. Lack of a 5′ γ-phosphate is a common feature of the negative-strand RNA [(−)RNA] of the packaged dsRNA segments in rotavirus. Strikingly, all (−)RNAs (of group A rotaviruses) have a 5′ GG dinucleotide sequence. As the only rotavirus protein with 5′ RTPase activity, NSP2 is implicated in the removal of the γ-phosphate from the rotavirus (−)RNA. To understand how NSP2, despite its sequence-independent RNA-binding property, recognizes (−)RNA to hydrolyze the γ-phosphate within the catalytic cleft, we determined a crystal structure of NSP2 in complex with the 5′ consensus sequence of minus-strand rotavirus RNA. Our studies show that the 5′ GG of the bound oligoribonucleotide interacts extensively with highly conserved residues in the NSP2 enzymatic cleft. Although these residues provide GG-specific interactions, surface plasmon resonance studies suggest that the C-terminal helix and other basic residues outside the enzymatic cleft account for sequence-independent RNA binding of NSP2. A novel observation from our studies, which may have implications in viroplasm formation, is that the C-terminal helix of NSP2 exhibits two distinct conformations and engages in domain-swapping interactions, which result in the formation of NSP2 octamer chains.

INTRODUCTION

Rotavirus, a member of the Reoviridae family, causes severe life-threatening infantile gastroenteritis worldwide (7). It is a large nonenveloped icosahedral virus with three concentric capsid layers that enclose a genome consisting of 11 segments of double-stranded RNA (dsRNA) (16, 31). The rotavirus genome encodes six structural (VPs) and six nonstructural (NSPs) proteins (7). Transcription of the genome segments takes place within the intact double-layered particles (DLPs), which are produced by the removal of the outer capsid layer during cell entry (7). This process, which yields positive-strand RNAs [(+)RNA] that are capped at the 5′ end, is carried out by the viral polymerase (VP1) and the capping enzyme (VP3) located inside the DLPs. Subsequent processes, including synthesis of dsRNA segments from (+)RNAs by VP1 and their packaging into newly assembled DLPs, take place in specialized compartments called viroplasms (7, 23). During genome replication, a consensus sequence [3′CS(+)], 5′-UGUGACC-3′ at the 3′ end of the (+)RNA, conserved among group A rotaviruses, is specifically recognized by VP1 to produce minus-strand RNA [(−)RNA] with a complementary CS [5′CS(−)] that terminates with a highly conserved GG motif at the 5′ end (18, 35, 39, 40) (Fig. 1). A strikingly consistent feature of the rotavirus (−)RNAs is the absence of γ-phosphate at their 5′ ends (9, 20, 37). This highly conserved feature suggests that removal of the γ-phosphate has an obligatory function during virus replication. Among the rotavirus proteins, only NSP2 exhibits RNA 5′ triphosphatase (RTPase) activity (37).

Fig 1.

Diagram schematic of a rotavirus dsRNA. The top schematic represents a dsRNA with the open reading frame (ORF) shown in gray, the 5′ and 3′ untranslated regions (UTRs) shown in orange, and the consensus sequences (CS) shown in red. The lower table shows the representative sequences of the 5′ and 3′ CS from different groups, strains, and gene segments of rotavirus. The conserved “GG” dinucleotide is shown in red.

NSP2 is a multifunctional enzyme that is implicated in various replication-related activities. Along with NSP5, it is essential for the formation of viroplasms, sites of rotavirus genome replication and DLP assembly. As one of the components of early replication intermediates, NSP2 is suggested to directly interact with VP1 and partially replicated viral RNA (2, 3, 12, 13). Biochemical studies with recombinant NSP2 show that it possesses several properties that are consistent with its intricate role in genome replication/packaging. NSP2 binds to single-stranded-RNA (ssRNA) without any observed sequence specificity and has the ability to destabilize duplex RNA (32, 34). In addition to the 5′ RTPase activity mentioned above, NSP2 exhibits nucleotidyl triphosphatase (NTPase) (32) and nucleoside diphosphate kinase (NDP kinase) activities (14), all of which are mediated by a highly conserved H225 as the catalytic residue. The RNA-binding and helix-destabilizing activities of NSP2, neither of which depends upon its enzymatic activities, are thought to be used to relax secondary structures of mRNA templates for dsRNA synthesis by VP1 (34). Precisely how the enzymatic activities of NSP2 are utilized during genome replication/packaging needs to be elucidated, and yet their necessity is strongly supported by complementation assays with enzymatically defective NSP2, which show the abrogation of dsRNA synthesis without compromising the functional oligomeric state of NSP2 or viroplasm formation (33).

The structure of NSP2, which readily self-assembles into an octamer, has been characterized by both X-ray crystallography and electron cryomicroscopy (cryo-EM) techniques (10, 11). NSP2 subunits, related by 4-2-2 symmetry, form a donut-shaped octamer with a 35 Å wide central hole and four exterior grooves that are lined predominantly by basic residues. The NSP2 monomer has distinct N- and C-terminal domains separated by a deep cleft. Crystallographic studies of NSP2 with ATP analogs show that nucleoside substrates bind within the cleft, and that γ-phosphate hydrolysis, involving a phosphohistidine intermediate, is mediated by H225 (14). However, how NSP2 recognizes 5′ RNA, which is shown to be a better substrate than NTPs for γ-phosphate hydrolysis (37), how the narrow catalytic cleft accommodates the 5′ nucleotides of RNA, and how, in the context of rotavirus replication, NSP2 makes sequence-specific interactions with conserved 5′ GG dinucleotide remain unclear.

In this study, to understand the structural basis for how NSP2 recognizes rotavirus (−)RNA as a substrate for RTPase activity, we carried out crystallographic studies of NSP2 with RNA oligonucleotides, including one that corresponds to 5′CS(−). Our results show that the 5′CS(−) oligoribonucleotide fully engages the NSP2 catalytic cleft through extensive contacts between 5′ GG and highly conserved amino acid residues in the cleft. A novel observation from our crystallographic studies, which may be relevant in the context of viroplasm formation, is that the C-terminal helix of NSP2 adopts two different conformations and engages in a domain-swapping interaction to link neighboring NSP2 octamers. Surface plasmon resonance (SPR) experiments show that deletion of this basic C-terminal helix significantly reduces NSP2 affinity for RNA, which suggests that while residues in the cleft confer specificity for the GG dinucleotide, the C-terminal helix and other basic residues outside the cleft in the NSP2 octamer are involved in sequence-independent RNA binding by NSP2 and play a role in its helix-destabilizing activity.

MATERIALS AND METHODS

Protein expression, purification, and crystallization.

The NSP2 gene from simian rotavirus strain SA11 was cloned into the bacterial expression vector pQE60 (Qiagen) with a C-terminal His tag following a thrombin cleavage site and expressed in Escherichia coli strain M15(pREP4). This His-tagged protein was purified using Ni-nitrilotriacetic acid (Ni-NTA) affinity chromatography (Qiagen). After this step, the His tag was removed using thrombin (Hematologic Tech). Untagged NSP2 was further purified over Ni-NTA affinity and HiTrap SP HP cation-exchange chromatography columns (GE Healthcare), dialyzed into 10 mM Tris-HCl (pH 7.2)–50 mM NaCl–1 mM dithiothreitol, and concentrated to 20 to 30 mg/ml.

Mutagenesis of NSP2.

The NSP2 C-terminal deletion mutant (NSP2_ΔCT) was produced by PCR amplification. An expression vector of pQEg8/KRΔT encoding the C-terminal deletion and the K223A and R227A mutations was produced by PCR amplification using pQE60g8/ΔT as the template and the primers 5′-TAACAAGTTTGCTGTAATCACACATGGTGCGGGTCATTATGCAATTGTAAAGTATTCATCAGTTGCTAAT-3′and 5′-ATTAGCAACTGATGAATACTTTACAATTGCATAATGACCCGCACCATGTGTGATTACAGCAAACTTGTTA-3′. The bases in these primers used to change the codons in NSP2 gene are denoted in boldface.

Crystallization.

SA11 NSP2/RNA (5′-P-GGU CAC AU, octa-riboadenyl nucleotide, purchased from Dharmacon RNAi Technologies was cocrystallized by hanging-drop vapor diffusion. Drops were composed 2 μl of protein-RNA mixed with 2 μl of well solution (50 mM sodium cacodylate [pH 6.5], 70 mM NaCl, 1 mM EDTA, 10 to 12% PEG 4000). Crystals were transferred in a single step to the mother liquid supplemented with 20% glycerol for 1 min and then flash-frozen in liquid nitrogen for diffraction data collection.

Structure determination and refinement.

Data were collected at 100K at beamline 19-ID or 19-BM at the Advanced Photon Source and processed using either using HKL2000 (22), d*trek (25), or IMOSFLM (41). The structures of NSP2-RNA complexes were determined by molecular replacement (MR) method using NSP2 structure (Protein Data Bank [PDB] no. 1L9V) as a search model. The MR solution was obtained using the program PHASER (19). Model building was performed using COOT (6). Iterative rounds of refinement, which included simulated annealing, translation/liberation/screw (TLS), and noncrystallographic symmetry constrains, and model building were carried using PHENIX (1). In the NSP2-5′CS(−) complex, the ligand density was validated using composite and simulated annealing omit maps using either the PHENIX (1) or the CNS (4) packages. The final model of this complex has an R_work = 20.63% and an R_free = 17.21% with 97.7% of all residues within the most favored or allowed regions in the Ramachandran plot. The data collection and refinement statistics are summarized in Table 1. Ligand interactions were analyzed using COOT (6) and LIGPLOT (38) with donor to acceptor distances between 2.6 and 3.2 Å for hydrogen bonding interactions and C-C distances between 3.4 and 4.5 Å for hydrophobic interactions. The electrostatic potential surface of the NSP2 octamer was calculated by using PyMol (29). Figures were generated using Chimera (24).

Table 1.

Diffraction data collection and refinement statistics of NSP2/RNA complex and NSP2_ΔCT^a

Parameter	NSP2-5′CS(−)RNA	NSP2ΔCT
Data collection
Space group	P3121	P4
Cell dimensions
a, b, c (Å)	107.6, 107.6, 272.1	122.5, 122.5, 301.1
α, β, γ (°)	90.0, 90.0, 120.0	90.0, 90.0, 90.0
Resolution (Å)	35.0–2.10 (2.14–2.10)	20.0–3.40 (3.52–3.40)
Rmerge (%)	8.1 (60.5)	36.4 (66.6)
I/σI	19.4 (2.6)	19.4 (2.6)
Completeness (%)	99.8 (100.0)	100.0 (99.9)
Redundancy	4.5 (4.5)	5.5 (5.5)
Refinement
Resolution (Å)	32.5–2.1	20.0–3.4
No. of reflections	107,346	57,280
Rwork/Rfree	0.17/0.21	0.24/0.29
No. of atoms	11,531	23,600
Protein	10,173	23,600
Ligand/ion	264/4
Water	1,090
B factor
Protein	37.04	78.8
Water	39.66
RMSD
Bond length (Å)	0.007	0.011
Bond angle (°)	0.954	1.573

^aValues in parentheses are for the highest-resolution shell.

NSP2_ΔCT also crystallized under the same conditions as the NSP2-5′CS(−) complex. Diffraction data for these crystals were collected at the Baylor College of Medicine crystallography core facility using Rigaku FR-E+ SuperBright rotating anode. NSP2_ΔCT crystals diffracted to ∼3.4 Å, and data processing, using d*TREK (25), indicated a P4 space group. Further data analysis using SCALA in CCP4 (41) and X-triage in PHENIX (1) indicated anisotropic diffraction and twinning. MR using PHASER (19), with previously determined NSP2 monomeric subunit without the residues 294 to 317 as a search model, provided a unique solution with a high z-score. The crystallographic asymmetric unit consists of five tail-to-tail interacting dimers. Four of these constitute one octamer, and the remaining dimer in association with three other dimers related by crystallographic 4-fold symmetry forms another octamer. The structure refinement was carried out with appropriate twinning corrections using REFMAC5 (41). The data collection and refinement statistics are summarized in Table 1.

SPR analysis.

Biacore 3000 was used to analyze the binding kinetics of NSP2 to RNA oligonucleotides. In these studies, 5′ phosphorylated and 3′ biotinylated oligonucleotides—5′-P-GGU CAC AUA A-biotin-3′ (GG nucleotide) and 5′-P-CCU CAC AUA A-biotin-3′ (CC nucleotide)—purchased from IDT DNA Technologies were immobilized on an SA sensor chip with an analyte R_max in the 100- to 200-resonance unit (RU) range. Binding experiments were performed with HBS-P buffer (10 mM HEPES [pH 7.4], 150 mM NaCl, 5 mM MgCl₂, 0.05% surfactant P20). Ten concentrations (5 to 500 nM) of NSP2 were injected at 20°C with a flow rate of 15 μl/min. The kinetic parameters and the dissociation constants (K_d) were calculated using BIAevaluation 4.1.1(Biacore Life Sciences). Initial fitting of the sensograms of wild-type NSP2 using a 1:1 Langmuir binding model produced high χ² values and subsequently were fit using a heterogeneous ligand parallel reactions (HLPR; two-site binding) model. Sensograms of the NSP2_ΔCT and NSP2_ΔCT K223A/R227A mutant were fit to a 1:1 Langmuir binding model.

Protein structure accession numbers.

Coordinates and structure factors of the structures discussed here have been deposited in the PDB under accession numbers 4G0A (NSP2-RNA complex) and 4G0J (NSP2_ΔCT).

RESULTS

Untagged NSP2 construct and cocrystallization with oligoribonucleotides.

All previously reported structural studies were carried out with a C-terminal His-tagged NSP2 (10, 11, 14). To avoid any possible interference of the basic His tag with negatively charged RNA, a new construct was designed such that the His tag could be removed (see Materials and Methods). With this untagged NSP2, crystallization trials in the presence of oligonucleotides, 5′-GGUCACAU-3′, which corresponds to the 5′ CS of rotavirus (−)RNAs [5′CS(−)] and other oligoriboadenyl-nucleotides were carried out using Natrix, a commercially available screen for nucleic acids and protein-nucleic acid complexes. Structures were determined by molecular replacement (MR) using the previously determined NSP2 monomer structure (PDB no. 1L9V) as the phasing model. Although we were able to obtain crystals of NSP2 in the presence of each of the RNA nucleotides mentioned above, and these crystals diffracted to resolutions ranging from 2.2 to 3.2 Å, only cocrystals of NSP2-5′CS(−), diffracting to 2.1-Å resolution, showed strong interpretable density for the ligand. Despite several attempts, including various buffer conditions, pH, ionic strength, and ligand concentrations, no interpretable density for the ligand was observed with other oligoribonucleotides. The data collection and refinement statistics for NSP2 in complex with the 5′ CS(−) is presented in Table 1.

Structure of NSP2-RNA complex.

In contrast to His-tagged NSP2, which crystallized in the I422 space group with one molecule in the asymmetric unit (10, 14), analysis of the diffraction data from the cocrystal of untagged NSP2-5′CS(−) indicated a different space group, P3₁21, with four NSP2 molecules in the crystallographic asymmetric unit. In all the four subunits, both 2Fo-Fc and Fo-Fc maps clearly indicated extra density that was unaccounted for by NSP2 but could be interpreted as the terminal GGU of the 5′CS RNA. The modeled nucleotide was further validated by computing composite and simulated annealing omit maps (Fig. 2A). All four subunits in the crystallographic asymmetric unit show the same structural features with two distinct domains separated by a cleft. A striking difference, however, is that in one of the subunits, the C-terminal helix is moved away from the C-terminal domain (henceforth referred to as open conformation), and in the other three subunits it remains closer to this domain (closed conformation) as in the His-tagged apo-NSP2 structure (Fig. 2B and C). The GG of the 5′ CS (−) RNA binds deep within the cleft. Although partial density for the next nucleotide (U) could be observed at lower contour levels, there is no interpretable density for the remaining five bases as they emerge from the cleft, indicating that they are not constrained by any specific interactions with NSP2.

Fig 2.

Structure of NSP2-5′CS(−)RNA complex. (A) Simulated annealing composite omit map showing the 5′ terminal nucleotides in a (2|F_observed| − |F_calculated|) map calculated by using the protein only as the phasing model and contoured at ∼1.1 σ. Modeled RNA is shown in a ball-and-stick representation (yellow) inside the map, with the nitrogen and the oxygen atoms colored in blue and red, respectively. (B) Ribbon diagram representation of an NSP2 monomer in the “closed” conformation. (C) Ribbon diagram representation of an NSP2 monomer in the “open” conformation.

Octamer formation.

NSP2 subunits in the NSP2-5′CS(−) complex structure show octamer formation that is consistent with the previous structural and biochemical studies of NSP2, which indicate that NSP2 readily forms an octamer and this quaternary organization is the functional form (30, 32). In the previous crystallographic studies of the His-tagged NSP2 in the I422 space group, application of the crystallographic 4-2-2 symmetry leads to octamer formation consisting of two tail-to-tail interacting tetramers. In the present P3₁21 space group, the four NSP2 molecules in the asymmetric unit, which are related by a noncrystallographic 4-fold symmetry, form a tetramer with extensive interactions between the subunits, and two of these tetramers, related by the crystallographic 2-fold axis, form the typical NSP2 octamer (Fig. 3A and B). Unlike the I422 octamer, in which the subunits of each tetramer are equivalent because of the crystallographic 4-fold axis, in the P3₁21 tetramer, conformational changes in the C-terminal helix break this equivalence. Except for this difference, all of the other structural features in the P3₁21 octamer, including the 35-Å central hole (Fig. 3B) along the 4-fold axis and the basic grooves at the 2-fold axes (Fig. 3C and D), are the same as in the I422 octamer.

Fig 3.

NSP2 octamer formation. (A) NSP2 tetramer in the crystallographic asymmetric unit as viewed along the pseudo 4-fold axis of the tetramer with bound RNA shown in a sphere representation. (B) NSP2 octamer formed by the tetramers related by the crystallographic 2-fold axis, as viewed along the pseudo 4-fold axis of the octamer. (C) NSP2 octamer viewed along the 2-fold axes showing the RNA-binding cleft. (D) Electrostatic surface potential surface of the NSP2 octamer (calculated without bound nucleotides) showing the groove (dashed black line) and two distinct electropositive patches: one near the cleft (white star) and the other at either end of the groove corresponding to the locations of the C-terminal helices in open (dashed black arrow) and closed (black arrow) conformations. Positively and negatively charged residues are shown in blue and red, respectively.

NSP2-RNA interactions.

The binding of 5′CS(−) by NSP2 involves a network of hydrogen bond interactions, several hydrophobic contacts, and a distinct cation-π interaction between the 5′ GG dinucleotide and the amino acid residues within the NSP2 catalytic cleft (Fig. 4). The side chains of K223 and N248 form hydrogen bonds with the base of G1, and R227 makes bidentate hydrogen bonds with two of the phosphate oxygens of G1. In addition, the G1 residue of RNA is further stabilized by a cation-π interaction between the side chain of Arg104 and the pyrimidine ring of G1. Consistent with the suggestion that NSP2 uses the same catalytic mechanism for both RTPase and NTPase activities (37), the terminal G1 nucleotide of the 5′CS(−) binds within the catalytic cleft at the same location as nonhydrolyzable ATP analogs in previous cocrystal structures (14).

Fig 4.

(A) Stereo view of interactions between NSP2 and 5′ CS of (−)RNA. The hydrogen bond interactions between the NSP2 residues (gray) and RNA are displayed as dashed lines. Residues involved in hydrophobic contacts are shown as spheres. A red dashed line indicates the cation-π interaction between R104 and G1. (B) Detailed interactions between SA11 NSP2 and RNA, as determined using LIGPLOT (36). The amino acid residues of NSP2 and the RNA involved in the interactions are labeled. Water molecules mediating RNA recognition are represented by small spheres in cyan. The carbon, nitrogen, and oxygen atoms are shown in black, blue, and red colored circles, respectively. Hydrogen bonds are shown as green lines between the respective donor and acceptor atoms, along with the acceptor-donor distance. The NSP2 residues involved in the hydrophobic interactions are indicated by an arc with spokes radiating toward the RNA atoms that they make contact with. The contacted atoms are shown with spokes radiating back.

The G2 of 5′CS(−) also makes a significant number of contacts with the NSP2 residues in the cleft. The phosphate oxygen of G2 participates in direct hydrogen bonding interaction with the side chain of R240, and its base with the side chain of N236. In addition to these electrostatic interactions, the sugars and bases of GG contribute to hydrophobic interactions that involves NSP2 residues V105, S107, D239, A243, and T244 (Fig. 4A and B). The conformation of the bound GG and its interactions with NSP2 are almost identical between the NSP2 subunits in the tetramer irrespective of the conformational state of the C-terminal helix. Compared to the previously published apo-NSP2 structure, the binding of 5′CS(−) does not induce any significant changes in NSP2, except for the slight movement toward the bound RNA of an α-helix (amino acids [aa] 235 to 243) and a loop (aa 251 to 256) on one side of the basic cleft.

The C-terminal helix participates in domain-swapping interactions.

As mentioned above, the C-terminal helix of NSP2 adopts distinct “closed” and “open” conformations. The C-terminal helix in the open conformation engages in interoctamer interactions, akin to domain swapping. The pocket vacated by the C-terminal helix in one octamer is occupied by an open-conformation C-terminal helix from the adjacent symmetry-related octamer such that the exchanged C-terminal helix mimics a C-terminal helix in the closed conformation. With each octamer consisting of two C-terminal helices on either side in the open conformation, the domain-swapping interactions leads to the formation of a chain of NSP2 octamers in the crystal lattice (Fig. 5C). Changes in the backbone dihedral angles of M293, K294, and P295 lead to the C-terminal helix switching its conformation (Fig. 5A and B). The domain-swapping interactions bring the neighboring octamers in the chain to closely interact with one another through several strong hydrophobic contacts involving residues L191, P196, and I146 with a buried surface area of ∼1,300 Ų between them. Such a domain-swapping interaction, leading to strong interoctamer association, is likely an important factor in the context of viroplasm formation.

Fig 5.

Domain swapping of the C-terminal helix of NSP2. (A) Superposition of the NSP2 subunits with C-terminal helix in open (orange), and closed (blue) conformations. The dihedral angle changes in residues M293, K294, and P295 lead to the C-terminal helix switching its conformation. (B) Two NSP2 tetramers (green and red) interact with each other through swapped C-terminal helices. (C) Domain-swapping interactions of the C-terminal helices between the octamers lead to the formation of a chain of NSP2 octamers in the crystal lattice.

SPR analysis of NSP2-RNA binding.

To further understand the RNA-binding characteristics of NSP2, particularly in regard to RNA nucleotides that lack the 5′ GG, and to examine whether the C-terminal helix, which contains several basic residues, plays any role in RNA binding, we carried out binding experiments using SPR. Previous studies showed that mutating some of the basic residues that line the groove, particularly residues in the N-terminal domain, diminishes RNA binding (37). Electrostatic potential surface of the NSP2 octamer shows that in addition to a strong electropositive patch close to the cleft, the C-terminal helices of the 2-fold related NSP2 subunits form another strong positively charged patch at either end of the groove (Fig. 2D). We hypothesized that the basic C-terminal helix of NSP2, in addition to the basic residues in the N-terminal domain, is important for sequence-independent RNA binding of NSP2.

For our SPR experiments, we cloned, expressed, and purified a deletion construct, NSP2_ΔCT, which lacks residues 295 to 317. Size-exclusion chromatography of NSP2_ΔCT showed a very similar elution profile to that of wild-type NSP2_1-317aa, indicating that the C-terminal truncation did not affect the formation of NSP2 octamers (Fig. 6A and B). Although we were able to obtain diffracting crystals of this deletion construct, crystals diffracted to a lower resolution (∼3.4 Å) and showed anisotropic diffraction and twinning that resulted in relatively high R_merge values (Table 1). Despite these challenges, phasing by molecular replacement was successful. The structure refined with R_work and R_free values of 26 and 30%, respectively, and showed that NSP2_ΔCT forms native-like octamers (Fig. 6C) consistent with the results from size exclusion chromatography.

Fig 6.

NSP2_ΔCT forms octamers. (A) Size-exclusion chromatography of wide-type and mutant NSP2s using Superose 6 (GE Healthcare), with elution volumes for each peak shown in the figure. (B) SDS-PAGE analysis of purified NSP2, NSP2_ΔCT, and NSP2_{K223A/R227A/ΔCT}. (C) Crystal structure NSP2_ΔCT octamer determined to 3.4-Å resolution (above) compared to the previously published NSP2 I422 octamer (11) (below). Two NSP2 subunits related by the 2-fold axis and involved in the tail-to-tail interaction of two NSP2 tetramers are shown in green and gold, respectively. The rest subunits of the octamers are shown in tan.

For RNA-binding analysis by SPR, we used a biotinylated decaribonucleotide with a sequence corresponding to the 5′ CS(−) and another decaribonucleotide in which the 5′ GG was replaced with CC. The sensogram of RNA binding with wild-type NSP2 (wtNSP2) was best fitted using an HLPR (two-site) binding model (Fig. 7). The χ² value of fitting using HLPR was consistently and significantly lower than using 1:1 Langmuir model (one-site) for all of the concentrations of NSP2. This data analysis shows that although the binding affinities of both GG and CC nucleotides are similar, the GG nucleotide has a slightly higher binding affinity compared to the CC nucleotide and it exhibits significantly (∼2-fold) higher association for one of the sites (“site 1”) (Fig. 7 and Table 2). In light of our crystallographic analysis with the GG nucleotide and the previous cryo-EM studies of NSP2 with decariboadenylnucleotide showing an accumulation of extra density across the groove regions (including the C-terminal helix) (11), one qualitative interpretation is that while both GG and CC can bind to the electropositive grooves (including the C-terminal helix) in a sequence-independent manner (groove binding), the GG nucleotide additionally binds cleft region through sequence-specific interactions. Such a preferential binding of the GG nucleotide within the cleft may contribute to its higher binding affinity. Although the CC nucleotide can potentially bind NSP2 interacting with some basic residues inside the cleft, this interaction would lack the hydrogen bonding with N236 and N248, cation-π interaction, and some of the stabilizing hydrophobic contacts found in the NSP2-GG interaction.

Fig 7.

RNA-binding assays using SPR. (A) Representative sensograms for wild-type and mutant NSP2 at 50 nM (green), 100 nM (red), and 200 nM (black) binding to GG RNA. (B) Representative sensograms for wild-type and mutant NSP2 binding to CC RNA with the same color scheme shown in panel A. The association lasts 240 s and is followed by disassociation for 180 s. The deduced k_on, k_off, and apparent K_d values are shown in Table 2.

Table 2.

Binding of “GG” RNA and “CC” RNA to NSP2 and NSP2 mutants^a

Protein	RNA	kon1 (M−1 s−1)	koff1 (s−1)	kon2 (M−1 s−1)	koff2 (s−1)	Kd1 (nM)	Kd2 (nM)
Wild type	GG	1.50 × 106	3.80 × 10−2	1.98 × 105	1.99 × 10−3	25.3	10.1
	CC	7.30 × 105	2.60 × 10−2	1.76 × 105	2.33 × 10−3	35.6	13.2
ΔCT	GG	8.26 × 104	5.18 × 10−3			62.7
	CC	5.42 × 104	8.76 × 10−3			161.6
K223A/R227A/ΔCT	GG	3.56 × 104	3.33 × 10−3			93.5
	CC	1.81 × 104	6.41 × 10−3			354.1

^aAssociation rate constants (k_on), dissociation rate constants (k_off), and apparent dissociation constants (K_d) were obtained from SPR assays. Sensograms of wild-type NSP2 were fit to a heterogeneous ligand (parallel reactions), whereas sensograms of the ΔCT and K223A/R227A/ΔCT mutant NSP2 were fit to a 1:1 (Langmuir) binding model.

With NSP2_ΔCT, the sensogram fits well to a 1:1 Langmuir model, with low χ² values. The binding parameters clearly show that deletion of the C-terminal helix has a profound effect on RNA binding (Fig. 7, Table 2). In the absence of the C-terminal helix, the binding affinity for both GG and CC nucleotides is significantly reduced. Of particular note, the rate of association (k_on) is considerably reduced with NSP2_ΔCT versus wtNSP2, suggesting that the basic C-terminal helix also has an electrostatic steering effect on RNA binding to NSP2. After deletion of the C-terminal helix, the difference between GG and CC binding to NSP2 becomes more pronounced with the apparent binding affinity of the GG nucleotide significantly higher than of the CC nucleotide. Since the binding is isolated to the cleft and other nearby basic residues in the groove, preferential binding of the GG nucleotide within the cleft is clearly accentuated. This is further substantiated by a proportional reduction in the binding affinities of the GG and CC residues upon mutating basic residues K223 and R227 within the cleft to alanine (Fig. 7, Table 2). Despite these mutations, since the other residues involved in GG specific interactions such as R104 are not altered, GG is likely to exhibit residual cleft binding, contributing to the differential affinities between GG and CC nucleotides.

DISCUSSION

NSP2 exhibits both sequence-specific and sequence-independent RNA binding.

Our studies provide a structural rationale for the seemingly contradictory results of biochemical studies showing that RNA binding by NSP2 is sequence independent (13, 32) and yet it can target rotavirus (−)RNA with a highly conserved 5′ GG motif for RTPase activity (37). Our crystallographic studies, together with SPR analysis, indicate that NSP2 exhibits both types of RNA binding. When presented with an RNA that has a GG motif at the 5′ end, NSP2 can target it for RTPase activity through extensive sequence-specific interactions that involve highly conserved residues within the enzymatic cleft. NSP2 can also bind to RNA in a sequence-independent manner through the C-terminal helix and other basic residues that line the grooves of the octamer. A new finding from our SPR analysis is that the C-terminal helix has a profound effect on the RNA binding of NSP2. This helix with its ability to exist in two distinct conformations is likely involved in enhancing both types of RNA binding through electrostatic steering effects. It can be surmised that the quaternary organization of NSP2 subunits into an octamer allows the integration of both GG-specific RNA “cleft binding” with sequence-independent “groove binding.” The 4-2-2 octamer organization of the NSP2 subunits, which also leads to a 2-fold related disposition of the basic residues along the grooves, may be the basis for the observed helix-destabilizing activity of NSP2.

It is clear from both our SPR analysis and previous studies (32) that NSP2 binds to any single-stranded RNA. However, we have not been successful in visualizing crystallographically the binding of oligoribonucleotides other than the 5′ GG oligonucleotide. A reasonable explanation is that the binding of the non-GG nucleotides, which involve interactions between the phosphate groups of the RNA and basic residues that line the groove, is nonspecific and varies within and between the NSP2 octamers and thus becomes transparent in crystallographic analysis. It is for the same reason that the remaining five nucleotides of the 5′ CS(−) used in our crystallographic studies were not visualized. These nucleotides that emerge from the cleft can bind the octamer groove in multiple orientations.

NTPase and RTPase activities share the same catalytic mechanism.

The binding of 5′ CS(−) within the cleft and proximal to the catalytic residue H225, similar to that of ATP analogs in previous crystallographic studies (14), is consistent with biochemical studies (37) showing that both RTPase and NTPase activities are mediated by H225 and that they share a catalytic mechanism involving transient phosphorylation of H225. However, compared to the binding of the ATP analogs in which only the terminal phosphates and the sugar moiety of the nucleoside could be observed in the crystal structures (14), our crystallographic studies show that the entire terminal GG is firmly anchored to the cleft and makes substantial interactions with cleft residues. This observation is in accordance with previous biochemical studies that indicate 5′ RNA binds to NSP2 with tighter affinity than the NTPs and that the former is a better substrate for γ-phosphate hydrolysis (37). The observation that most residues involved in coordinating the binding of the 5′ CS GG motif are highly conserved among the group A rotaviruses strongly support the notion that a GG-specific interaction is relevant in the context of rotavirus replication. The hydrolysis of the γ-phosphate of rotavirus (−)RNAs may indeed be a consequence of this specific recognition. Although it is currently unclear as to which of the enzymatic activities of NSP2 predominates during rotavirus replication, a strong possibility is that NSP2 uses different combinations of these catalytic activities or switching from one to another during genome replication and packaging.

In the viroplasms, which are the exclusive sites for rotavirus genome/encapsidation, only RNA species with a 5′ GG sequence that can be substrates for 5′ RTPase activity are the nascent (−)RNAs, since the 5′ ends of (+)RNAs are capped. NSP2 does not exhibit capping or the cap-binding activity. A relevant question is whether NSP2 is the only protein that is responsible for the removal of γ-phosphate from these (−)RNAs. Currently, NSP2 is the only rotavirus protein that has been experimentally shown to have 5′ RTPase activity (37). Another candidate is rotavirus VP3, which is responsible for 5′ capping of the (+)RNA during endogenous transcription (5, 8, 17). The formation of 5′ caps on rotavirus (+)RNA requires removal of the γ-phosphate. However, conservation of the GG motif in the rotavirus (−)RNAs together with the highly conserved nature of the NSP2 residues that interact with this motif argues for NSP2 as the most likely candidate responsible for the γ-phosphate hydrolysis of (−)RNAs of rotavirus.

Although the absence of a γ-phosphate is a characteristic feature of rotavirus (−)RNAs, its significance during virus replication is unclear. Whether it is a countermeasure against host antiviral response triggered by triphosphate RNA such as activation of the RIG-I pathways (28, 42) or whether it is also necessary for a particular function during rotavirus genome replication/encapsidation requires further study. It is conceivable that the specific binding of the NSP2 to the 5′ CS of the (–)RNA by NSP2 is required to confer structural organization of the newly synthesized duplex RNA around VP1 for subsequent encapsidation. Each VP1-VP3 heterodimeric complex inside the DLPs may associate with one segment to allow simultaneous transcription of dsRNA segments (15, 26, 36). Based on previous biochemical studies, which strongly suggest the possibility of direct interaction between NSP2 and VP1 (3, 12), one model that can be envisioned is that the specific recognition of 5′ CS of (−)RNA and sequence-independent binding of the 3′ end of the (+)RNA of the duplex as it emerges from VP1 allows for proper structural organization of the genome segment around VP1. This might then allow for subsequent packaging into assembling core particles such that each segment is associated with one VP1 molecule inside the assembled particles, which then would be conducive for simultaneous transcription of the genome segments during endogenous transcription.

The C-terminal helix exhibits two distinct conformational states.

A novel observation from our studies with untagged NSP2 is that the C-terminal helix of NSP2 exhibits two distinct conformational states, referred to here as the closed and open conformations. In the previous X-ray crystallographic studies of His-tagged NSP2, in the presence or in the absence of NTP analogs, the C-terminal helix exists only in the closed conformation. It is possible that the addition of His₆ residues to the C terminus prevents this conformational change. What triggers this conformational change in only one subunit of the tetramer is puzzling. From the observation that all subunits irrespective of the conformational state of their C-terminal helices show density for the bound RNA, it is unlikely that the RNA-binding affects this switching. However, without further studies we cannot rule out subtle cooperativity between the subunits, i.e., binding of RNA initially to one subunit triggers the open conformation, along with small changes in the intersubunit interactions, thus allowing the remaining subunits to bind RNA but without causing a conformational change in the C-terminal helix. A simple explanation for this, however, is that the conformational change is the result of crystallization conditions and crystal packing forces. Nonetheless, the ability of the C-terminal helix to exist in distinct conformational states is clearly evident.

C-terminal helix is involved in domain-swapping interactions.

Another interesting and novel observation from our structural analysis is the involvement of the C-terminal helix in an open conformation in the domain-swapping interactions resulting in the formation of chain of NSP2 octamers. Two residues, K294 and P295, in the loop region that connects the C-terminal helix to the rest of the C-terminal domain provide a switch for the open and closed conformations. Proline residues are frequently associated with domain-swapping interactions in proteins (21, 27). Both K294 and P295 are highly conserved in the NSP2 of both group A and C rotavirus, which implies that these residues are functionally significant. Considering that NSP2 is critical for viroplasm formation and is one of the most abundant proteins in the viroplasm with concentrations similar to those found in crystals, it stands to reason that the domain-swapping interaction provides an underlying mechanism for viroplasm formation.

In conclusion, our studies have provided novel insights into RNA-binding characteristics of NSP2, a multifunctional protein that is critical for rotavirus replication. We have provided a structural basis for specific recognition of the 5′CS of rotavirus (−)RNA by NSP2 and how the symmetric organization of the NSP2 octamer facilitates both sequence-specific and sequence-independent RNA binding. A novel observation from our studies is that the C-terminal helix of NSP2 exhibits two distinct conformations and engages in domain-swapping interactions that may have functional implications during viroplasm formation. Our SPR analysis highlights the importance of the C-terminal helix in RNA binding of NSP2 and suggests that it plays a role in electrostatic steering of the RNA. Although these SPR studies have provided a simple qualitative picture of the RNA-binding characteristics of NSP2, more sophisticated binding analyses are required to fully comprehend the kinetics of this interaction in the context of the symmetric organization of the NSP2 octamer. Further structural and functional studies are required to understand how NSP2 with multiple binding partners uses its RNA-binding and enzymatic activities to coordinate the complex and dynamic processes of viroplasm formation, genome replication, and encapsidation in rotavirus.