Structure of the influenza virus A H5N1 nucleoprotein: implications for RNA binding, oligomerization, and vaccine design

Abstract

The threat of a pandemic outbreak of influenza virus A H5N1 has become a major concern worldwide. The nucleoprotein (NP) of the virus binds the RNA genome and acts as a key adaptor between the virus and the host cell. It, therefore, plays an important structural and functional role and represents an attractive drug target. Here, we report the 3.3-Å crystal structure of H5N1 NP, which is composed of a head domain, a body domain, and a tail loop. Our structure resolves the important linker segments (residues 397–401, 429–437) that connect the tail loop with the remainder of the molecule and a flexible, basic loop (residues 73–91) located in an arginine-rich groove surrounding Arg150. Using surface plasmon resonance, we found the basic loop and arginine-rich groove, but mostly a protruding element containing Arg174 and Arg175, to be important in RNA binding by NP. We also used our crystal structure to build a ring-shaped assembly of nine NP subunits to model the miniribonucleoprotein particle previously visualized by electron microscopy. Our study of H5N1 NP provides insight into the oligomerization interface and the RNA-binding groove, which are attractive drug targets, and it identifies the epitopes that might be used for universal vaccine development.—Ng, A. K.-L., Zhang, H., Tan, K., Li, Z., Liu, J.-h., Chan, P. K.-S., Li, S.-M., Chan, W.-Y., Au, S. W.-N., Joachimiak, A., Walz, T., Wang, J.-H., Shaw, P.-C. Structure of the influenza virus A H5N1 nucleoprotein: implications for RNA binding, oligomerization, and vaccine design.

Influenza is a contagious respiratory illness causing annual epidemics and occasional pandemics. The death toll of influenza epidemics worldwide is in the range of 250,000 to 500,000 each year. The mortality in influenza pandemics is high due to the lack of vaccination and effective therapeutic drugs. The 1918–1919 Spanish flu, for instance, claimed the lives of around 50 million people and is widely cited as the most devastating pandemic in recorded world history (1). Since the 1997 outbreak of influenza A strain H5N1 in Hong Kong, avian flu has become the major threat for the next pandemic (2). Therefore, the development of new therapeutic agents or a cross-subtype, “universal” vaccine is of the highest priority in the battle against influenza.

Like the rabies, measles, and vesicular stomatitis viruses (VSV), the influenza A virus is a single-stranded, negative-sense RNA virus (NSRV). Its genome comprises 8 segments of RNA (vRNA) encoding 11 identified polypeptides. The two strain-defining surface glycoproteins, hemagglutinin (HA) and neuraminidase (NA), are currently the main vaccine targets (3). Both HA and NA exist in several different forms and have high mutation rates. The antigenic drift allows the influenza A virus to escape from established humoral immunity in humans (4). Vaccines must then be reformulated, specifically for serologically distinct viruses in each influenza season. The time delay in large-scale vaccine production against a suddenly appearing, new pandemic strain could cause a global disaster. By contrast, internal viral proteins, such as the nucleoprotein (NP), are substantially more conserved, and hence are ideal targets for T-cell-mediated immunity (5).

Similar to other NSRVs, the genome of the influenza virus is encapsidated by NP. The primary function of NP is to condense the segmented genomic RNA into a helical nucleocapsid and, together with three polymerase subunits, PA, PB1, and PB2, to form a ribonucleoprotein (RNP) particle for RNA transcription, replication, and packaging. The RNP, rather than the naked vRNA, is the template for transcription and replication (6). However, NP is not only a structural protein for RNA binding, it also acts as a multifunctional key adaptor molecule for interactions between virus and host cell (reviewed in ref. 7). NP is likely to be the major switching factor that determines whether genomic vRNA is transcribed into mRNA encoding viral proteins or used as template to synthesize cRNA for genome replication (8). NP also interacts with a number of viral and cellular proteins. In addition to encapsidating viral RNA (9), NP also forms homooligomers to maintain the RNP structure (10) and binds to the PB1 and PB2 subunits of the RNA polymerase (11). Cellular proteins with which NP interacts include importin, F-actin, and CRM1/exportin-1 (reviewed in ref. 7). These interactions mediate RNP trafficking in and out of the nucleus and are thus vital for viral propagation and assembly.

Its multiple functions and conserved protein sequence make NP an excellent drug target for all influenza A virus subtypes. In this context, an atomic resolution structure of NP is of crucial importance. So far, electron microscopy (EM) has produced a very low resolution structure of the mini-RNP (12, 13), and more recently, X-ray crystallography has yielded a 3.2-Å resolution structure of a homologous NP from human origin (A/WSN/1933; H1N1) (14). Here we report the 3.3-Å crystal structure of A/HK/483/97 (H5N1) NP (PDB code: 2Q06) from avian origin. We describe a number of novel structural features, discuss a possible RNA-binding mechanism, and propose a strategy for vaccine design.

MATERIALS AND METHODS

Protein expression and purification

The NP gene of influenza virus A/HK/483/97(H5N1), encoding 498 amino acid residues, was cloned into the pRHisMBP expression vector, and the MBP-tagged NP was expressed in Escherichia coli BL21(DE3)pLysS. The cells were lysed by sonication, and the lysate was passed through an amylose column (New England Biolabs, Ipswich, MA, USA). Bound protein was eluted with a 0–20 mM maltose gradient in 20 mM sodium phosphate and 0.15 M NaCl, pH 6.5. The eluate was incubated with thrombin (100 U) at 4°C overnight to remove MBP from NP and then passed through a heparin HP column (GE Healthcare, Waukesha, WI, USA). NP was eluted with a 0–1.5 M NaCl gradient in the same buffer. Site-directed mutagenesis was performed according to standard protocol. NP mutants were purified as described for the wild-type protein.

EM

A 51-nt RNA (5′-GGG AGA UUU UUU UUU UUU UUU UUU UUU UUU UUU UUU UUU UUU UUU UUU UUU-3′) was produced by in vitro transcription using the MEGAscript kit (Ambion, Austin, TX, USA), purified by phenol extraction, and mixed with NP at a molar ratio of 1:2 in buffer containing 0.1 M sodium phosphate and 0.15 M NaCl, pH 6.5. EM samples of NP with or without RNA were prepared by conventional negative staining as described (15). Briefly, 3.5-μl sample was applied to a glow-discharged carbon-coated EM grid. After 30 s incubation, the grid was washed with 2 drops of Milli-Q water (Millipore, Billerica, MA, USA) and stained with 2 drops of 0.75% (w/v) uranyl formate. Grids were examined with a Philips CM10 electron microscope (Philips Electronics, Mahwah, NJ, USA) equipped with a tungsten filament and operated at an acceleration voltage of 100 kV. Images were recorded with a Gatan 1 × 1 k CCD camera (Gatan, Inc., Pleasanton, CA, USA) at ×52,000 and a defocus value of ∼2 μm.

Static light scattering

RNase-treated NP and untreated NP, both at a concentration of 2.5 mg/ml, were subjected to static light scattering analysis using a miniDAWN triangle (45, 90, 135°) light-scattering detector (Wyatt Technology Corporation, Santa Barbara, CA, USA) connected to an Optilab DSP interferometric refractometer (Wyatt Technology Corporation). This system was connected to a Superdex 10/300 GL Column (GE Healthcare), controlled by an ÄKTAexplorer chromatography system (GE Healthcare). Before sample injection, the miniDAWN detector system was equilibrated with 20 mM sodium phosphate and 150 mM NaCl, pH 7.0, for at least 2 h to ensure a stable baseline signal. The flow rate was set to 0.7 ml/min, and the sample volume was 100 μl. The laser scattering (687 nm) and the refractive index (690 nm) of the respective protein solutions were recorded during the measurement processes. Wyatt software ASTRA was used to evaluate all data obtained.

Crystallization

RNase-treated NP in 20 mM 3-(N-morpholino) propanesulfonic acid and 100 mM NaCl, pH 7.0, was first crystallized by vapor diffusion using the sitting drop technique. Initial crystallization conditions were surveyed using Index, Crystal screen I/II, and Natrix kits (Hampton Research, Aliso Viejo, CA, USA). Small crystals were obtained in Natrix condition no. 5 [0.2 M KCl; 0.01 M MgCl₂; 0.05 M 2-(N-morpholino)ethane sulfonic acid, pH 5.6; and 5% PEG 8000]. After refinement of the crystallization conditions, tetrahedral crystals were grown in hanging drops in 0.15 M KCl, 0.01 M MgCl₂, 0.1 M cacodylate (pH 6.2), and 7% PEG 8000.

Structure solution and refinement

Crystals were rapidly frozen after stepwise soaking them in increasing concentrations of sucrose up to a concentration of 20% in the crystallization buffer. A 3.3-Å resolution data set from a single, frozen crystal was collected at 100 K at the 19ID beamline (with X-rays at a wavelength of 0.97845 Å) of the Structural Biology Center at the Advanced Photon Source of Argonne National Laboratory using the program SBCcollect. The data were processed and scaled with the HKL3000 suite (16) (Table 1). The crystals belonged to space group P2₁3, and an asymmetric unit contained two NP molecules. The structure was determined by molecular replacement, using the coordinates of H1N1 NP (PDB code: 2IQH) as the search model, and refined to an R_free of 27.9%. All calculations were done using the CCP4 suite (17), and model building was done in Coot (18) and CNS (19). Ramachandran plot statistics show that all residues were found in most favored/allowed regions. Figures of protein structures were prepared with the program PyMOL (20).

TABLE 1.

Data collection and refinement statistics

Statistic	NP crystal
Data collection
Space group	P213
Cell dimensions
a, b, c (Å)	153.578, 153.578, 153.578
α, β, γ (degrees)	90.00, 90.00, 90.00
Resolution (Å)	48.564–3.301 (3.38–3.30)
Rsym or Rmerge	14.1 (66.0)
I/σI	11.16 (2.1)
Completeness (%)	99.7 (99.1)
Redundancy	5.0 (4.1)
Refinement
Resolution (Å)	46.32–3.30
Reflections	18,379
Rwork/Rfree	0.202/0.279
Protein atoms	7343
B-factors	93.109
RMS deviation
Bond length (Å)	0.009
Bond angle (degrees)	1.236

One crystal was used. Values in parentheses are for the highest-resolution shell. 

Modeling a ring composed of nine NP subunits

After assembling nine NP monomers into a ring according to the EM structure of a mini-RNP (12), the linker regions were removed, and the tail loop was bent by ∼90° toward the body domain. The linker regions were then rebuilt using the program ArchPRED (21).

Surface plasmon resonance

Biotinylated 2′-O-methylated RNA oligonucleotide with the sequence 5′ UUU GUU ACA CAC ACA CAC GCU GUG 3′ was purchased (RiboBioscience, Guangzhou, China) and immobilized on an SA sensor chip (GE Healthcare) until the surface density reached 26–28 response units (RU), according to manufacturer’s instructions (BIAcore; GE Healthcare). Surface plasmon resonance measurements were carried out with a BIAcore 3000 (GE Healthcare) at 25°C. Data were analyzed with BIAevaluation v. 4.1 (BIAcore), and curves were fitted with the “1:1 binding with drifting baseline” model.

RESULTS AND DISCUSSION

Novel features revealed by the H5N1 NP structure

H5N1 NP was expressed as a MBP-fusion protein in E. coli and purified to >95% homogeneity (see Supplemental Fig. 1). EM of negatively stained samples showed that, after tag removal and RNase treatment, NP exists in solution as a mixture of trimers and tetramers (Fig. 1A). In the presence of RNA, NP assembles into higher-order oligomers (Fig. 1B). Light-scattering studies of RNase-treated NP in solution suggested a molecular mass of ∼193 kDa (Fig. 2), also corresponding to a mixture of trimers and tetramers, but only trimers were incorporated into the 3-dimensional crystals (see below).

Figure 1.

EM of NP in the presence and absence of RNA. A) RNase-treated NP exists in solution as a mixture of trimers and tetramers. B) Higher-order oligomers are observed when a 51-nt RNA is added to an RNase-treated NP sample. Scale bars = 50 nm.

Figure 2.

Light scattering of untreated and RNase-treated NP. Two distinct populations were found in an NP sample not treated with RNase (dotted trace). The peak at 586 kDa corresponds to higher-order oligomers; the peak at 193 kDa corresponds to the mixture of trimers and tetramers. After treatment with RNase, the high molecular mass population disappeared (solid trace). The RNase-treated sample was less polydispersed (0.5%) than the untreated sample (2%) and therefore more suitable for crystallization.

H5N1 NP formed small tetrahedral crystals with the cubic space group P2₁3. A diffraction data set was collected to 3.3-Å resolution using the synchrotron radiation source at Argonne National Laboratory. The crystal structure of H5N1 NP was determined by molecular replacement, using the H1N1 NP monomer (PDB code: 2IQH, Chain A) (14) as the search model (its partial tail loop was not included in the calculations). One crystallographic asymmetric unit comprises two H5N1 NP molecules, referred to as molecules A and B. Even in the initial 2Fo-Fc map, the entire tail loops of both molecules were clearly defined by continuous electron density at a 1σ contour level. The two linkers as well as some residues that were missing from the C-terminal region of the H1N1 model were, therefore, built based on difference maps. The structure was refined to R/R_free = 20.2%/27.9% (Table 1). Overall, 467 amino acids were modeled for molecule A (residues 22–78 and 87–496), and 465 amino acids were modeled for molecule B (residues 21–77 and 88–495) (Fig. 3A).

Figure 3.

Crystal structure of H5N1 NP, the crystallographic trimer and model of a ring formed by 9 NP subunits. A) An asymmetric unit contains two NP monomers, molecules A and B. The NP structure consists of a head domain and a body domain with a protruding element in between (orange, molecule A). A loop extends from near the C terminus (tail loop, red) and is used for oligomerization. NP can form different trimers due to different contacts between the helix-loop-helix motifs (blue) of neighboring protomers, which is made possible by the flexibility of the linker sequences (green). B) Both molecules in the asymmetric unit, molecules A and B, form crystallographic trimers with their symmetry-related counterparts. The trimer formed by molecule A is shown here. Each NP molecule in the trimer inserts its tail loop into the neighboring protomer. C) Model of a ring formed by nine NP molecules based on the previously determined EM structure of a mini-RNP (12) and the flexibility of the linkers. The N-terminal region of NP is shown in red.

H5N1 NP folds into two helical domains, the head and body domains, arranged in a banana-shaped configuration, very similar to H1N1 NP. H5N1 NP and H1N1 NP share 94% sequence identity, and after aligning 398 residues (residues 22–72, 92–202, 213–396, 438–489) the root-mean-square (RMS) deviation between the two crystal structures is 1.0 Å. Our H5N1 NP structure reveals, however, several novel and important features that have not previously been seen in the H1N1 NP structure. First, our map of H5N1 NP shows well-defined electron densities for residues 397–401 and 429–437 (see Supplemental Fig. 2), which were missing in the H1N1 NP model and now connect the extended tail loop (residues 402–428) to the main body. Second, seven additional C-terminal residues (residues 490–496) could be modeled in our H5N1 NP structure (see Supplemental Fig. 2) and were found to form coils and bends. These residues are located near the insertion site of the tail loop and are thought to play a regulatory role in NP oligomerization (22). Third, more residues (residues 72–78 and 87–92) could be built in a flexible, basic loop toward the concave region of the molecule postulated to be an RNA-binding groove (14). Finally, H5N1 and H1N1 show very interesting differences in the way they trimerize, as will be discussed in detail below.

H5N1 NP contains five cysteine residues. Remarkably, none of them form disulfide bonds with each other. Cys-44, Cys-164, and Cys-223 form hydrogen bonds with the side chains of other amino acids. Interestingly, the Cα atoms of Cys-275 and Cys-333 are only 5.6 Å apart, well within the range of disulfide bond formation, but their side chains point away from each other. A previous study suggested the formation of a transient disulfide bond during the conformational maturation of NP (23), and Cys residues 275 and 333 would be very good candidates to play a role in this process.

Comparison of NPs from different NSRVs

Influenza virus, rabies virus, and VSV are all NSRVs. The sequences of their NPs differ in length (influenza: 498 residues; rabies: 398 residues; VSV: 422 residues) and show very little homology. Nevertheless, as shown in Supplemental Fig. 3, the overall conformation of influenza virus A NP is similar to that of the rabies virus (24) and VSV (25). First, the three NPs are all composed of two helical domains in a banana-shaped configuration. A long tail loop, which is important for RNP assembly, reaches out from near the C-terminal end of the polypeptide chain and interacts with the neighboring molecule. Second, all NPs contain three loops that bridge the two domains and maintain the conformation of the protein by connecting the body and head domains. These segments in the rabies and VSV NPs are all located in the middle of the polypeptide chain but are spread throughout the sequence in the influenza virus A NP (residues 147–152, 265–278 and 450–463). Third, superimposition of the individual NP domains shows that the head domains are structurally more conserved than the body domains (26). Fourth, like VSV and rabies NPs, influenza virus A NP also contains a helix-loop-helix motif (residues 130–157), with the loop being one of the three passes between the two domains and being located at the center of the presumed RNA binding groove. These conservations in conformation among the NSRV NPs are thought to relate to their functional roles, as all of them are involved in genomic vRNA encapsidation.

The flexibility of the linkers allows NP to form different oligomers

In our H5N1 NP crystal, each of the two molecules in the asymmetric unit is part of a trimer about the crystallographic 3-fold axis, so that the intermolecular interactions within each trimer are identical (Fig. 3B). By contrast, in the H1N1 crystal, the asymmetric unit contains a single trimer with the three protomers forming slightly different interactions with one another. Although the trimers formed by H5N1 NP and H1N1 NP are very different and cannot be superimposed (Fig. 4A), the subunits in both trimers are connected by the same principle: the insertion of the long tail loop (residues 402–428) from one NP protomer into the body domain of a neighboring protomer (Fig. 4B, C). In the H5N1 NP, the long loop protrudes straight from one protomer into the neighboring protomer, but forms an angle of 70° in the H1N1 NP. These very different conformations suggest that the two linkers (residues 397–401 and 429–437) have a high degree of flexibility. It is conceivable that this linker flexibility is important for converting NP trimers to higher-order oligomers during virus packaging.

Figure 4.

H5N1 NP and H1N1 NP have identical interactions of the tail loops but form different trimers. A–C) Superimposition of one molecule (chain A) of the H5N1 NP trimer (in color) with a molecule (chain C) of the H1N1 NP trimer (gray) shows that the other two molecules do not superimpose with each other (A). Nevertheless, the interaction of the tail loop of one NP molecule with the neighboring protomer in H5N1 (B) and H1N1 (C) is virtually identical. D) The trimer formed by H5N1 NP molecule A is held together by a salt bridge formed between R422 in helix 1 of the helix-loop-helix motif in one protomer with E449 in helix 2 of the helix-loop-helix in the neighboring protomer. E) The trimer formed by H5N1 NP molecule B is held together by hydrogen bonds formed between the E434 residues of the three protomers (top panel). A salt bridge between residues K430 and E449 stabilizes the two helices of the helix-loop-helix motif in each protomer (bottom panel).

The formation of a trimer brings the three helix-loop-helix motifs near the C-terminus of each NP protomer into close contact (Fig. 3B). Helix 1 of the helix-loop-helix motif is formed by residues E421-K430 and helix 2 by residues D438-A451. Since the trimers formed by H5N1 NP and H1N1 NP are different, the molecular contacts between the helix-loop-helix motifs are also different in the two trimers. This distinction can even be seen in the trimers formed by molecules A and B in our H5N1 NP structure. In the trimer formed by molecule A, helices from two adjacent protomers are held together by a salt bridge between R422 in helix 1 of one protomer and E449 in helix 2 from the adjacent protomer (Fig. 4D). This interaction was not observed in the trimer formed by H1N1 NP. In the trimer formed by molecule B, the two helices of the helix-loop-helix motif are held close together by a salt bridge between K430 in helix 1 and E449 in helix 2. The helix-loop-helix motifs of neighboring protomers are linked through hydrogen bonds between the side chain of E434 in one protomer with the NH group of E434 in the neighboring protomer (Fig. 4E). Although residues 429–437 were not resolved in the crystal structure, H1N1 NP has a threonine at position 430 instead of a lysine. Therefore, the salt bridge linking protomers in the H5N1 NP trimer is unlikely to be formed in the H1N1 NP trimer, offering a possible explanation for the different configurations of the two trimers. It is worth noting that a survey of more than 300 influenza virus A NP sequences showed that only 25 had a lysine residue at position 430, and 20 of these were from influenza A viruses of the H5N1 subtype. Whether this H5N1-specific substitution bears any significance for virus assembly remains to be determined.

Because the H5N1 NP and H1N1 NP trimers differ substantially (Fig. 4A), it is unlikely that the trimer is the physiologically relevant RNA-binding unit in influenza A virus, as has previously been suggested (14). Instead, we propose a model in which nine NP molecules organize into a ring-like structure (Fig. 3C). This model is based on structures of nucleoprotein-RNA complexes of VSV (25) and rabies virus (24) and the low-resolution EM structure of a mini-RNP (12, 13). On the basis of our model, we can propose the surfaces of NP that may be involved in its interaction with other viral and cellular components. In our model, NP assembles into a ring using the same tail loop as it uses to form a trimer, since mutant viruses in which the tail loop of NP (residues 418–426) was deleted could not be rescued (27). To assemble nine protomers into a ring, the tail loop has to bend by ∼90° toward its body domain, compared to its conformation in the trimer (see Supplemental Fig. 4). Secondary structure analysis suggests that the linker regions are composed of random coil (data not shown). This notion is supported by the H1N1 NP crystals, in which the linker regions are disordered (14), indicating that these regions are indeed flexible.

The proposed 9-mer NP structure has the following characteristics. A ring composed of 9 NPs has inner and outer diameters of 75 and 190 Å, respectively. These dimensions agree with the EM structure (12) but are slightly larger than those of the rabies and VSV NP/RNA complexes (24, 25). Alignment of the influenza virus A H5N1 NP and VSV NP sequences shows similarities in the tail loop (data not shown), yet they mediate interactions between neighboring protomers in different ways. The loops of rabies NP and VSV NP only drape along the surface of the neighboring protomer. On the other hand, the tail of the influenza virus A NP (both H5N1 and H1N1) makes extensive interaction with the adjacent protomer by inserting deeply into its body domain, and this interaction may be retained in the 9-mer model.

In the rabies and VSV NP/RNA complexes, in addition to the tail loop, the N-terminal segments also make interactions with neighboring molecules, thus helping to maintain the integrity of the ring structure (24, 25). Their N-terminal 20 residues form an extended antiparallel β-ribbon that reaches to the neighboring protomer and establishes another tether. In both the H5N1 and H1N1 NP crystal structures, the 20-residue N-terminal segment is disordered, presumably pointing away from the trimer. Although this region seems to be pointing to a neighboring subunit in our model of the ring assembly (Fig. 3C), determining whether and how an interaction is formed will require further studies.

Finally, a consequence of our model of the ring-shaped NP oligomer would be that the arginine-rich RNA binding groove (see below) is exposed on the surface, which is consistent with the finding that the bases of the RNA in the RNP are susceptible to modification (28).

RNA binding by NP

The H5N1 NP features a protrusion (residues 167–186, green sequence in Fig. 6B). This protrusion, which is not present in NPs from VSV and rabies virus, extends significantly into a spacious groove between the head and body domains. The presence of many positively charged residues in the protrusion and nearby regions suggested a possible role of this region, referred to as NP-G1 (Fig. 6A), in RNA binding. To test this hypothesis, a series of NP mutants was generated in which multiple Arg and/or Lys residues were substituted by Ala residues, and their binding kinetics and affinities were assessed by surface plasmon resonance (SPR). This work identifies the important residues involved in the association of trimeric/tetrameric NP with RNA, which then results in their assembly into higher-order NP oligomers and the encapsidation of the RNA. First, different concentrations of wild-type NP were passed over a sensor chip with immobilized RNA (Fig. 5A and Table 2). The measured RNA binding affinity of 23.1 ± 0.8 nM was consistent with previous studies (around 20 nM) (reviewed in ref. 7). To assess the contribution of individual residues to RNA binding, mutant NPs at concentrations of 700 and 70 nM were analyzed by SPR (Fig. 5B and Table 2).

Figure 6.

NP regions tested for their involvement in RNA binding. A) Regions containing clusters of arginine residues are denoted as NP-G1 to NP-G4. Mutation of the arginines in NP-G1 and NP-G2 (magenta) resulted in a significant reduction in RNA binding, whereas mutation of the arginines in NP-G3 and NP-G4 (green) showed little effect. B) Close-up view of the arginine-rich groove in NP. The unique protrusion (residues 167–186) found in influenza NP, but not in rabies or VSV NP, is shown in green. The distance between the side chains of R174, R175, and R221, which point toward each other, is ∼8 Å, which is about the diameter of a single-strand RNA molecule. The flexible, basic loop (blue) is in proximity to both the protrusion and the NP-G2 region and may help to position the vRNA into the basic groove.

Figure 5.

Wild-type and mutant NP bind RNA with different affinities. A) Sensogram of the binding of wild-type NP to immobilized RNA. Different concentrations of purified wild-type NP were injected, and the signal (RU) was plotted as a function of time. The square boxes (□) represent the averages from three independent experiments, whereas the solid lines represent the fitted curves. B) Binding of wild-type and mutant NPs (700 nM) to immobilized RNA. Two controls are also shown: buffer (ctrl-buffer) and bovine serum albumin (ctrl-BSA).

TABLE 2.

Kinetic and affinity constants for NP-RNA interaction

Protein	Mutation/deletion sites	ka (M−1 s−1)	Fold change	kd (s−1)	Fold change	KD (nM)	Fold change	Relative association
								700 nM (RU)	70 nM (RU)
NP-WT		4.14 ± 0.06 × 105		9.58 ± 0.19 × 10−3		23.1 ± 0.8		210.2	95.0
NP-G1	R74A; R75A; R174A; R175A; R221A	ND		ND		ND		4.3	1.5
NP-G2	R150A; R152A; R156A; R162A	4.82 ± 0.08 × 104	−8.59	5.25 ± 0.13 × 10−3	−0.55	108.9 ± 4.5	−4.72	62.9	19.8
NP-G3	K90A; K91A; K113A; R117A; R121A	3.80 ± 0.06 × 105	−1.08	9.00 ± 0.19 × 10−3	+0.94	23.7 ± 0.9	−1.03	206.7	96.1
NP-G4	R382A; R384A	2.12 ± 0.03 × 105	−1.95	1.10 ± 0.02 × 10−2	+1.15	51.8 ± 1.7	−2.24	173.8	81.9
NP-Δ74–88	Δ74–88	7.84 ± 0.16 × 104	−5.28	1.16 ± 0.03 × 10−2	+1.21	148.0 ± 6.8	−6.41	136.1	46.6
Ctrl-BSA		ND		ND		ND		1.0	0.8
Ctrl-buffer		ND		ND		ND		0.4	0.5

K_D = k_d/k_a, where K_D is affinity, k_d is dissociation rate, and; k_a is association rate. Errors of k_a and k_d represent standard error; error of K_D represents the maximum error from each experiment. Fold change is relative to wild type: +, fold increase; –, fold decrease. Value of RU was taken from t = 345 s, when the NP-RNA binding reached equilibrium. ND, not determined. 

Residues R74, R75, R174, R175, and R221 in the NP-G1 region (Fig. 6A) were found to be essential for RNA binding. Mutation of all these arginines to alanine almost completely abolished RNA binding (Table 2). Whereas residues R174 and R175 are part of the protrusion, R74 and R75 belong to the flexible, basic loop (residues 73–91), which contains seven basic residues (ERRNRYLEEHPSAGKDPKK). Although the electron density for this loop is not very well defined, residues 73–78 appear to extend toward the protrusion essential for RNA binding. In particular, R74 and R75 of the basic loop are in proximity to residues R174 and R175 of the protrusion. When the loop was deleted (NP-Δ74–88), the affinity of NP for RNA is 6-fold reduced (Table 2), demonstrating that it contributes to the ability of NP to bind vRNA, probably through an induced fit.

Another region of NP contains four arginine residues, R150, R152, R156, and R162, and is also located along the groove between the head and the body domain (NP-G2 in Fig. 6A). It thus appeared to be another likely protein region involved in RNA binding. Changing all four arginines in the NP-G2 region to alanine caused the affinity of NP for RNA to drop by a factor of 5 (Table 2). Two other regions of NP also contain clusters of arginine residues: K90, K91, K113, R117, and R121 in the region denoted NP-G3, and R382 and R384 in NP-G4 (Fig. 6A). Mutations of either cluster of arginines to alanine had little effect on RNA binding (Table 2).

Both regions involved in RNA binding, NP-G1 and NP-G2, contain multiple arginine residues. This feature is consistent with a previous study that employed nonspecific chemical modification and found that arginine residues are more important in mediating NP-RNA contacts than lysine residues (29).

The positively charged arginine residues, thought to interact with the negatively charged phosphate backbone of the RNA molecules (29), are especially enriched along the groove between the head and body domains. In our model of the ring-shaped NP assembly, this arginine-rich groove is exposed on the surface of the ring, which agrees with the RNase sensitivity of influenza virus A RNPs (30) and the accessibility of the RNA bases to modifying agents (28). Moreover, a previous EM study has shown that the outside of the ring formed by influenza virus A NP interacts with RNA polymerase (12), which recognizes and interacts with the panhandle structure of the vRNA promoter (31). An NP encapsidation that exposes the RNA would thus favor such interactions with the RNA polymerase. This mode of encapsidation is, however, distinctly different from the RNPs formed by the rabies virus and VSV, in which the NPs sequester the RNA at the inside of their ring-shaped assemblies (24, 25). A high-resolution structure of influenza A NP-RNA oligomers will be needed to truly understand RNA encapsidation by influenza A NP.

With the new information provided by the H5N1 NP structure presented here, we can now propose a possible mechanism by which NP associates with RNA. First, the flexibility of the basic loop (residues 73–91) may allow it to sample the environment and capture RNA, explaining why deletion of the loop significantly reduces the affinity of NP for RNA. Because the loop is in close proximity to the two regions most important for RNA binding, NP-G1 and NP-G2, it could then deliver the captured RNA into the arginine-rich groove. Second, our experimental data show that the region centered around the protrusion is crucial for RNA binding and presumably is the major RNA binding site on NP. The side chains of the arginine residues in this region are pointing toward each other, suggesting that this region may clamp the RNA into the groove. Third, our data show that the NP-G2 region at the other end of the groove is also important for RNA-binding. Since 24–27 RNA nucleotides bind to an influenza virus A NP molecule (32), as opposed to only nine RNA nucleotides in rabies and VSV NPs, the RNA is expected to make further contacts with NP in addition to binding along the arginine-rich groove.

NP as a target for drug design and vaccine development

The structure of H5N1 NP should be of much value for the design of antiviral drugs and the development of a vaccine to fight the next potential pandemic. It suggests that the RNA-binding groove, which is exposed and highly accessible, and the tail domain, used for oligomerization of NP, would be attractive drug targets. Small molecules that effectively bind to these sites would interfere with proper RNP function and disrupt the transcription-replication processes, hence, decreasing the viability of the virus. Furthermore, a universal influenza vaccine has always been a long-term goal of biomedical research (33, 34). The presence of CD8⁺ T cells due to previous infection with influenza virus may reduce the severity of the pandemic (35) and age-associated decline in the T-cell repertoire diversity is correlated with poor heterosubtypic protection (36). NP is a conserved internal protein and thus a good target for eliciting cell-mediated immunity. At least 14 human NP peptides have been identified as epitopes of cytotoxic T lymphocytes (CTL) (37). With respect to the H5N1 NP structure, some of these epitopes, including NP_380–388, NP_383–391, and NP_418–426, are located on random coil regions and thus susceptible to mutations, explaining why they are hypervariable (38). The accumulation of mutations in these epitopes has been associated with escape from CTL immunity (39). Some other epitopes, however, are in regions that are structurally and functionally important. The NP_265–274 epitope, for instance, is buried inside the protein. It is part of one of the three peptide segments that span both the head and body domains and thus may be structurally important for maintaining the proper geometry of the RNA-binding groove. Another epitope, NP_174–184, is located at the putative RNA-binding groove, and would thus also be functionally important for the protein (see Supplemental Fig. 5). Because healthy human subjects were found to have a robust CD4+ T cell response against peptides of these two conserved epitopes (40), these epitopes may be candidates for providing partial immunity to the pandemic H5N1 strain.