Biochemical and Structural Evidence in Support of a Coherent Model for the Formation of the Double-Helical Influenza A Virus Ribonucleoprotein

Because influenza viral RNA synthesis occurs in the nucleus (1), newly synthesized NP in the cytoplasm is not expected to be associated with vRNAs or cRNAs. In addition, although NP is a nonspecific RNA-binding protein (28), it should not bind to a significant extent to cellular or viral mRNAs, because mRNAs are associated with a large array of proteins that compete with NP for binding (29). To determine whether NP self-associates efficiently in the absence of RNA, freshly purified H1N1 WSN NP protein at two different concentrations (18 and 300 µM) was equilibrated for 5 days at 4°C and then analyzed by gel filtration (Fig. 1a). The amounts of trimers and monomers were quantitated, and by assuming the dissociation constant (K_d) for the reversible conversion between NP₃ and 3 NPs, i.e., K_d is equal to [NP]³/[NP₃], we estimated a K_d of 8.5 × 10⁻³ M². Such a large K_d indicates that NP has only a weak self-association activity in the absence of RNA. Consequently, NP likely exists primarily as monomers in the cytoplasm before it is imported to the nucleus and participates in the assembly of viral RNPs. We conclude that it is probably unnecessary to invoke the action of another protein to prevent NP oligomerization in the cytoplasm.

We previously showed that NP-NP interactions are largely mediated by a tail loop consisting of amino acid residues from 402 to 429 (10). Because cRNA and vRNA synthesis is coupled with RNP assembly (23, 24), we considered it likely that NP proteins lacking a functional tail loop would be inactive in viral RNA replication. To determine whether this is the case, we assayed such NP mutant proteins for their ability to support viral RNA synthesis using minigenome assays and a vRNA-sense luciferase reporter, as described in Materials and Methods. As shown in Fig. 2a, a WSN NP mutant that lacks the tail loop (Δ402–429NP), as well as NP proteins that contain the R416A or E339A mutation, which disrupts a critical intersubunit salt bridge, have completely lost the ability to support viral RNA synthesis. These results are consistent with results from earlier mutagenesis analysis (30, 31). In contrast, the ability of NP to interact with polymerase proteins, as assayed in cotransfection assays, was not affected by such mutations (Fig. 2b).

During viral RNA synthesis, a viral template RNA (vRNA or cRNA) has to dissociate from several, if not many, NP molecules in order to interact with the large trimeric polymerase. The NP scaffold in the RNA-free region remained intact only if NP oligomers are kinetically stable. To measure the stability of RNA-free NP oligomers, purified recombinant NPs lacking RNA, which are in the form of oligomers (i.e., mostly trimers), were incubated at 37°C. These NP oligomers broke down to monomers in solution very slowly. For the NP of H1N1 WSN virus, gel filtration analysis of purified NP trimers showed that they dissociated to monomers via an approximate first-order reaction with a half-life (t_1/2) of ≈10 h (Fig. 1b). A large t_1/2 was also found for the dissociation of the NP trimers of other influenza A virus strains, namely, an H3N2 virus (influenza A/Udorn/72 virus), Hong Kong (HK)/H5N1 virus, and an H9N2 virus (influenza A/HK/1074/99). Consequently, these NP-NP interactions have extremely low dissociation rates, indicating that the NP scaffold in the RNP complex likely remains intact during viral RNA synthesis, when template RNA has to dissociate from some NPs in order to interact with the polymerase. Consistent with this conclusion, RNPs have been found to retain their rod-like structure even after RNA was stripped away (25, 26).

When analyzed by gel filtration, the Δ402–429NP mutant of WSN virus, as expected, no longer forms homo-trimers or larger homo-oligomers (Fig. 2c), as is also the case for the R416A or E339A mutant NP protein (data not shown). Unexpectedly, the NP tail loop mutant not only forms monomers but also forms a species at the ~110-kDa position (Fig. 2c), which corresponds to the molecular mass of an NP dimer containing two ~55-kDa NP monomers. This dimer species was also observed with the tail loop mutant of HK H5N1 NP (Fig. 2c). This formation of homo-dimers indicates that NP is able to dimerize presumably through a new interface that does not involve the tail loop. In fact, it appears that native NP is also capable of forming dimers. As shown in Fig. 1b, a small population of dimer species was clearly visible in the native NP sample after most of the NP trimers had dissociated to monomers (Fig. 1b).

To identify the new dimer interface, we determined the crystal structure of the WSN Δ402–429NP dimer to 2.8-Å resolution (Table 1; Fig. 3). We also attempted to crystallize the Δ402–429NP monomer but without success. In the NP dimer crystal, a dimer is found in each asymmetric unit, and it appears to be the only stable dimer based on buried surfaces (~2,100 Å²). As is the case in the NP trimer, each NP subunit in the dimer adopts a crescent-like shape with a head and a body domain. The two NP molecules in the dimer pack against each other in an antiparallel fashion, with the putative RNA-binding groove fully exposed in both subunits (Fig. 3a). The two NP subunits are essentially identical and can be superimposed with a root mean square difference (RMSD) of 1.1 Å for 401 common C-α atoms. By comparison, superimposing subunit B of the Δ402–429NP dimer with B subunit of the WSN NP trimer yielded an RMSD of 1.4 Å for 385 common C-α atoms, suggesting that the removal of the tail loop did not cause a large overall conformational change in the tertiary structure of NP in the dimer form. Indeed, when the RNA-binding activity was measured using a 20-nt RNA oligonucleotide, the Δ402–429NP dimer exhibited a binding affinity that was indistinguishable from that of the wild-type NP trimer (K_d was equal to 3.9 ± 0.2 nM for the Δ491–498NP dimer versus 3.6 ± 0.5 nM for the wild type).

Because of its higher resolution (i.e., 2.8 Å), the crystal structure of Δ402–429NP reveals more-ordered regions than the two previously published influenza A virus NP structures. Two of the newly defined regions are residues 73 to 91 and 490 to 498, both of which are disordered in the original trimeric H1N1 NP structure (Fig. 3b). Residues 73 to 91 emanate from the body domain and form the so-called loop L1, which folds back into the putative RNA-binding groove and interacts with loop L2 (Fig. 3b). The other two loops in and around the RNA-binding groove are L3 and L4, numbered sequentially based on their primary sequence. Loops L1 and L2 have been found to play a critical role in NP RNA binding (11). In particular, loop L1 has the sequence ⁷³ERRNKYLE EHPSAGKDPK K⁹¹, with 6 basic amino acid residues (boldface), and deletion of residues 74 to 88 resulted in a 6-fold decrease in the RNA-binding activity of the HK H5N1 NP (11). Loop L3 does not directly participate in RNA binding but is found to mediate the NP-polymerase interaction that is required for viral RNA synthesis (17).

In both subunits of the Δ402–429NP dimer, the C-terminal tail, which contains amino acids 490 to 498, extends from the back of the body domain and winds into the putative RNA-binding groove at the front of the NP molecule (Fig. 3c). This C-terminal tail contains four acidic amino acid residues (D491, E494, E495, and D497) and two asparagines (N492 and N498), and therefore its electrostatic property is opposite to that of the positively charged surface of the putative RNA-binding groove. The acidic nature of the C-terminal tail is conserved across all members of the Orthomyxoviridae family. In one extreme case, the infectious salmon anemia virus (ISAV), an isavirus that is distantly related to the influenza A virus, has 12 acidic residues in the last 16 residues of its NP sequence (32).

Comparison of the trimeric and dimeric NP interaction interfaces shows that two different sets of residues are used for these two types of interactions (Fig. 4). The tail loop interaction is responsible for NP oligomerization and involves residues in and around the tail pocket at the back of the NP molecule (Fig. 4b). In contrast, dimer interactions are mediated mostly by two polypeptide regions: a helix-turn-helix motif comprised of residues 149 to 167 and the C terminus from residues 482 to the end (Fig. 4a). Near the 2-fold symmetry axis, the side chains of residues 487 to 489 close to the C-terminal tail are rearranged compared to those of the NP trimer. These movements allow the side chains of F489 from the two subunits to stack against each other. Overall, the intersubunit interface buried in the trimer is more extensive (~8,331 Å²) than that in the Δ402–429NP dimer (~2,122 Å²).

To verify that the dimer interface is biologically relevant, we tried to identify amino acid residues that are critical for NP dimerization but do not affect oligomerization or RNA binding. We first determined which mutation in the NP mutant with the loop deleted eliminated dimer formation. After testing various mutations, we showed that introducing a D491A mutation eliminated dimer formation (Fig. 5a and b). Although D491 is not on the dimer interface, its mutation apparently disturbs the conformation of the C-terminal tail, eliminating dimer formation. It was also confirmed that the D491A mutation did not have any effect on regular NP oligomerization/trimerization through the tail loop (Fig. 5b). In addition to identifying D491A, we identified two other mutations, F487D and F487R, that were able to affect dimer interaction but not trimer formation (Fig. 5a and data not shown). Of the other dimer mutations that we have made, R152A, D160A, R261A, and F489A result in partial but not complete disruption of NP dimer formation. The Δ491–498 mutation completely disrupted NP dimer formation, but considering the extensive nature of the deletion, functional characterization of this mutant was not further pursued.

The ability of these three single-residue NP dimerization mutants to support viral RNA synthesis was then tested in minigenome assays using both a vRNA-sense luciferase template and a cRNA-sense luciferase template (Fig. 5c and d). The Y487R NP mutant did not show any activity in either assay. Although the Y487D and D491A mutants showed approximately 10 to 20% activity in the vRNA luciferase assay, we did not detect any mRNA synthesis real time-PCR (RT-PCR). These two mutants showed similarly low activities in the cRNA assay, but, as shown, no vRNA-sense RNA was seen with 25 cycles of RT-PCR. Likewise, no band was observed with 30 cycles either. Additionally, we did not detect smaller RNAs with the three dimerization mutants. Our results thus demonstrate the biological importance of the NP dimer interaction.

We have identified three different molecular forms of NP, namely, trimers, dimers, and monomers. The NP molecules that form trimers oligomerize by interacting with each other largely via the tail loop and contain two structural domains, as previously described (10, 11). The NP dimer lacking the tail loop consists of two NP subunits that pack against each other side-by-side in an antiparallel fashion through mostly hydrophobic interactions and cannot further polymerize. The NP monomer, which likely has a structure similar to that of trimeric NP, is in dynamic equilibrium with NP oligomers, but these NP-NP interactions are weak in the absence of any associated RNA.

NP residues that mediate dimer formation are mostly different from those used for tail loop interactions, suggesting that these two types of interactions are nonexclusive and can occur simultaneously within a higher-order complex. The NP dimer crystal structure shows that two polypeptide regions constitute most of the dimer interface: the helix-turn-helix motif comprised of residues 149 to 167 and the C-terminal tail from residue 482 to the end. The critical role of the C-terminal tail of NP in NP dimer formation is evident from our mutagenesis data, as the disruption of the structural conformation of the C-terminal tail alone resulted in NP mutants that were not able to form any dimers. Interestingly, residues 490 to 498 are disordered in the original NP trimer crystal structure, and as expected, the deletion of residues 491 to 498 had no effect on NP’s ability to form trimers and larger oligomers (data not shown). In addition to mediating dimer interaction, the C-terminal tail of NP may help to regulate NP RNA-binding activity and adopt different conformations in different functional states (e.g., monomeric versus oligomeric NP, RNA-bound versus RNA-free NP). We found that the RNA-binding activity of the Δ491–498 mutant was enhanced by ~2-fold compared to that of the wild-type NP protein (K_d was equal to 1.7 ± 0.1 nM for the Δ491–498 mutant versus 3.6 ± 0.5 nM for the wild type).

While the tail loop functions in maintaining interactions between adjacent NP molecules in influenza virus RNPs, the NP dimer interaction has the potential to organize the RNP into its double-helical structure (Fig. 6). In any double-helical structures, two sets of stabilizing forces are required. The first set is between adjacent molecules on the same polymer strand, and the second is between molecules from the two opposing polymer strands of the double-helical structure. As shown in Fig. 6, our double-helical RNP model entails two types of NP interactions. One is the tail loop interaction, which is used to maintain connection between adjacent NP molecules that point in the same direction and bind to the same RNA strand. The other is the dimer interaction made between antiparallel NP molecules that are associated with two antiparallel RNA strands. Disruption of the NP dimer interface leads to the unwinding of the RNP, resulting in the dissociation of the two antiparallel RNA strands, as previously observed under high-salt and low-salt buffer conditions (7). Because the RNA-binding groove is located at the exterior of the NP dimer structure, bound RNAs are likely to be solvent exposed in the influenza virus RNPs. Since NP forms the internal protein scaffold of RNP based on our model, the overall structure of RNP is expected to remain intact after bound RNA is dissociated, consistent with an earlier observation (25). We speculate that the NPs of nonsegmented, negative-strand RNA viruses lack such a dimer interface, explaining why the RNPs of these viruses are linear rather than double-helical. Interestingly, the RNPs of another group of segmented, negative-strand RNA viruses, bunyaviruses, also appear to be double-helical (33), suggesting that a dimer interface similar to that described here for the influenza virus NP may be shared by other members of this group of RNA viruses.

We have shown that NP mutants that cannot dimerize also cannot support viral RNA synthesis, as determined in minigenome assays. In these assays, the vRNAs and cRNAs that are produced are presumably in the form of RNPs, as is the case in infected cells. These assays measure the steady-state levels of vRNPs or complementary RNPs (cRNPs) that are produced. Consequently, the absence of newly synthesized vRNPs and cRNPs are consistent with two possibilities, which are not mutually exclusive: (i) elongation of vRNA and cRNA chains stops at the point at which the normal double-helical region of the RNPs is reached, and the resulting small single-stranded RNPs are unstable, and/or (ii) partial or complete single-stranded RNPs are produced but are inherently unstable or cannot be used as the template for further RNA synthesis. In any case, these results indicate that the ability to form the NP-NP interface needed for the formation of a double-helical structure is biologically important.

Our results support the model for NP trafficking and functions in infected cells outlined in Fig. 6b. Newly synthesized NP exists as monomers in the cytoplasm due to the weak NP-NP association in the absence of bound RNA. After importation into the nucleus, NP molecules are added to the growing end of the assembling RNP complex, where monomeric NP polymerizes with RNA-bound NP through the tail loop interaction. Fully assembled RNP adopts a double-helical structure, mediated by the dimer interface revealed in the Δ402–429NP dimer structure described here.

During viral RNA transcription and replication, it is necessary for vRNA (and cRNA) to dissociate from at least part of the NP scaffold to access the polymerase active site in order to function as a template. This dissociation does not require the disruption of the NP scaffold because the vRNA is exposed on the outer surface of the RNP. Our results suggest that the structure of RNP is maintained primarily by NP protein-protein contacts (i.e., the tail loop between adjacent NP molecules and the dimer interaction between strands). Because we have shown that NP-NP interaction has a low dissociation rate constant, the NP scaffold is likely to remain intact when the vRNA template locally dissociates from the RNP during RNA synthesis, consistent with a previous observation that the overall structure of RNP is maintained even after vRNA is removed (25). After a vRNA region is copied by the polymerase, it can then readily reassociate with the NP scaffold. The vRNP and cRNP thus maintain their structural stability during the course of infection once they are properly assembled.

NP genes from A/WSN/1933 (H1N1) and A/HK/483/97 (H5N1) were cloned into pET28a(+) vectors and expressed in E. coli Rosetta 2(DE3) Singles competent cells (Novagen) (10). A C-terminal 6×His tag was engineered to facilitate purification by affinity chromatography. Nontagged NPs from both the H1N1 and H5N1 viruses were also expressed for control purposes. NP mutants were constructed using standard oligonucleotide mutagenesis methods and were expressed using the same system. E. coli cells were grown in Luria broth medium at 37°C to an optical density at 590 nm (OD₅₉₀) of 0.6 and induced by 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 6 h at 28°C.

For insect cell expression, constructs of untagged and C-terminally His-tagged NPs and their mutants were cloned into the baculovirus vector pFastBac and expressed in Sf21 insect cells using the Bac-to-Bac baculovirus expression system (Invitrogen). Baculoviruses expressing the three polymerase protein subunits were constructed, coexpressed, and purified as previously described (19). Sf21 insect cells were grown in Hink’s TNM-FH insect medium (JRH Biosciences) supplemented with 10% (vol/vol) fetal bovine serum (Valley Biomedical) and harvested 48 h postinfection.

Purification of NP, overexpressed in either E. coli or insect cells, was performed as previously described (10). Briefly, clarified cell lysate was initially subjected to a Ni-nitrilotriacetic acid (NTA) column (for His-tagged proteins) or to an ammonium sulfate precipitation step (for nontagged proteins). Fractions containing NP were subsequently purified by heparin affinity and gel filtration chromatography. The final yield was about 1.5 mg of NP dimer and 5 mg of NP monomer per liter of cells. The purified protein (~95% pure) was concentrated to 5 mg/ml using a Centricon tube (Millipore) and stored at –80°C.

Crystallization conditions were initially screened using a Hydra II Plus One microdispenser robot (Apogent Discoveries) with ready-to-use crystal screen kits (Hampton Research, Qiagen). Rectangular crystals (~10 by 20 by 80 µm³) of the Δ402–429NP dimer were grown at room temperature by vapor diffusion with hanging drops with a ratio of protein to well solution of 1:1. The well solution contained 100 mM Tris-HCl, pH 8.5, 18% polyethylene glycol 3350 (PEG 3350), 10% glycerol, and 10 mM dithiothreitol (DTT). The Δ402–429NP dimer crystallized as small rectangles. Crystals were transferred to stabilizing solution (100 mM Tris-HCl, pH 8.5, 20% PEG 3500, 10% glycerol, 20% PEG 400) and flash frozen in liquid nitrogen. Data were collected at the Cornell High Energy Synchrotron Source (CHESS) with beamline F1. All data were processed with HKL-2000 (Table 1) (34).

The structure of the Δ402–429NP dimer was solved by molecular replacement to 2.8-Å resolution using Phaser (35) and CCP4 (36). The A/WSN/33 (H1N1) NP trimer structure (10) was used as the initial phasing model, and it was found that there were two molecules per crystallographic asymmetric unit. The initial map was built using program O (37). To reduce model bias, composite omit maps and density modification were used, and the model was subjected to multiple cycles of simulated annealing with the Crystallography & NMR System (CNS) (38). The two noncrystallographic symmetry (NCS)-related NP molecules are independently refined without any restraints/constraints. After position and thermal parameter refinement, the R_work and R_free converged to 0.2415 and 0.2978, respectively. In the final model, subunit B contains 424 out of 470 amino acids (excluding the 28 residues in the tail loop), including residues 21 to 389, 438 to 450, and 460 to 498; subunit A is less ordered and contains 18 fewer residues. Specifically, residues 390 to 392 are ordered in subunit A, but residues 83 to 89, 202 to 212, and 451 to 453 are disordered. The Ramachandran plot showed that 83.4 and 16.6% of residues were located in the most favored regions and additional allowed regions, respectively.

Fluorescence polarization (FP) was measured using a PanVera/Beacon 2000 FP system at room temperature. Samples were excited at 490 nm, and emission was measured at 520 nm. The reaction buffer used for all experiments contained 20 mM Tris, pH 7.5, 200 mM NaCl, 5 mM 2-mercaptoethanol, and 1 mM EDTA. A 5′-fluoresceinated 20-mer RNA oligonucleotide (5′ AGUAGAACAGGGUGACAAAG 3′ with the conserved 5′ vRNA sequence) (12) was chemically synthesized (Dharmacon). For each experiment, protein was serially titrated into a 0.99-ml reaction buffer containing 0.2 nM fluoresceinated oligonucleotide until the reaction reached saturation. Each protein titration was repeated in triplicate. The data were fitted to a sigmoidal curve using Igor Pro (WaveMetrics, Portland, OR).

C-terminally 6×His-tagged wild-type NP and Δ402–429NP were coexpressed with untagged PB1 and PB2, respectively, using the Bac-to-Bac baculovirus expression system (Invitrogen). After sonication and centrifugation, the cell supernatant was incubated with Ni-NTA resin (Qiagen), washed, and eluted as previously described. Proteins were analyzed by 9% SDS-PAGE and Western blotting. Anti-PB1, anti-PB2, and anti-tetra-His (Qiagen) were used as primary antibodies, depending on which proteins were coexpressed. Goat anti-mouse IgG and anti-rabbit IgG conjugated with alkaline phosphate (Pierce) were used as secondary antibodies. The blot was developed with SigmaFAST 5-bromo-4-chloro-3-indolylphosphate (BCIP)-nitroblue tetrazolium (NBT).

Dual-luciferase reporter assays for viral RNA synthesis were carried out as previously described (17). The vRNA template was provided by a pHH21 plasmid expressing an RNA containing the 26 5′- and 3′-terminal bases of the Ud NS vRNA sandwiching a negative-sense firefly luciferase reading frame. Alternatively, a cRNA template was provided by a pHH21 plasmid expressing the 5′- and 3′-terminal bases of Ud NS cRNA sandwiching a positive-sense firefly luciferase reading frame. To monitor transfection efficiencies, a polymerase II-driven plasmid expressing Renilla luciferase was also transfected. The firefly/Renilla luciferase ratio measures the level of viral RNA synthesis. The amount of viral RNA produced (cRNA or vRNA) in these assays was also directly measured by semiquantitative RT-PCR.

Crystal structure coordinates have been deposited in the Protein Data Bank (PDB; identifier, 3MIR). All ribbon diagrams were made using the program PyMOL (W. L. Delano; http://www.pymol.org).