Crystal structure of the Lassa virus nucleoprotein–RNA complex reveals a gating mechanism for RNA binding

Abstract

Arenaviruses cause disease in industrialized and developing nations alike. Among them, the hemorrhagic fever virus Lassa is responsible for ∼300,000–500,000 infections/y in Western Africa. The arenavirus nucleoprotein (NP) forms the protein scaffold of the genomic ribonucleoprotein complexes and is critical for transcription and replication of the viral genome. Here, we present crystal structures of the RNA-binding domain of Lassa virus NP in complex with ssRNA. This structure shows, in contrast to the predicted model, that RNA binds in a deep, basic crevice located entirely within the N-terminal domain. Furthermore, the NP-ssRNA structures presented here, combined with hydrogen-deuterium exchange/MS and functional studies, suggest a gating mechanism by which NP opens to accept RNA. Directed mutagenesis and functional studies provide a unique look into how the arenavirus NPs bind to and protect the viral genome and also suggest the likely assembly by which viral ribonucleoprotein complexes are organized.

The arenavirus family has a worldwide distribution and contains significant human pathogens such as Lassa (LASV), Machupo, Junin, Lujo (1, 2), and lymphyocytic choriomeningitis virus. Of these arenaviruses, LASV carries the largest disease burden, causing 300,000 to 500,000 infections per year in Western Africa. It is also the hemorrhagic fever most frequently transported out of Africa to the United States and Europe (2–4).

Arenaviruses have a bisegmented, negative-sense, single-stranded RNA genome with a unique ambisense coding strategy that produces just four known proteins: a glycoprotein, a nucleoprotein (NP), a matrix protein (Z), and a polymerase (L) (2). Of these proteins, NP is the most abundant in an infected cell. NP associates with L to form the ribonucleoprotein (RNP) core for RNA replication and transcription (5) and the matrix protein Z for viral assembly (6–8). The arenavirus NP also plays an important role in the suppression of the innate immune system (9–11).

Genome and antigenome RNAs of negative-strand RNA viruses (NSV) do not exist as naked RNA, but rather as a RNP complex in which the RNA is encapsidated by the viral nucleoprotein. During replication of many negative-strand RNA viruses, the nascent nucleoprotein (usually termed N) is bound by a polymerase cofactor (often a phosphoprotein, termed P), which prevents polymerization of N and nonspecific encapsidation of host cell RNAs (12–15). The resulting complex is termed N⁰-P, in which N⁰ denotes RNA-free N. The arenavirus, orthomyxovirus (flu), and bunyavirus (Hanta, Rift Valley Fever) families (i.e., segmented NSV) do not encode an analogous P protein, and the mechanism by which the nucleoprotein controls RNA binding during virus infection is not yet understood.

The arenavirus nucleoprotein (termed NP instead of N) has distinct N- and C-terminal domains connected by a flexible linker (16–19). The C-terminal domain functions as an exonuclease (16, 17) specific for dsRNA (17) and linked to antagonism of type I IFN (16, 17). A structure of LASV NP, in the absence of RNA, predicted the presence of a cap-binding site in the N-terminal domain and an RNA-binding site between the N- and C-terminal domains (16). Here we present two structures of the N-terminal domain of LASV NP, now in complex with ssRNA.

Results

Structure of LASV NP in Complex with RNA.

The recombinant N-terminal domain of LASV NP (NP_1–340) is monomeric in nature and is stably and irreversibly bound to ssRNA derived from the expression host (SI Appendix, SI Materials and Methods, and Fig. S1A). Crystals of NP_1–340 belong to the space group P6₁, contain six molecules in the asymmetric unit, and diffract to 3 Å resolution. Treatment of NP_1–340 with pepsin (NP_pep) removes a flexible polypeptide corresponding to residues 126–143 (SI Appendix, Figs. S1B and S2). Crystals of NP_pep belong to the space group P2₁2₁2, contain one molecule in the asymmetric unit, and diffract to 1.8 Å resolution (SI Appendix, Table S1).

In the structure of NP_pep, electron density is well defined for residues 8–112 and 163–339, indicating that residues 113–127 and 146–162, which flank the pepsin-deleted region, are indeed disordered as suggested by hydrogen/deuterium exchange (DX)/MS (SI Appendix, Fig. S2A). The undigested NP_1–340 contains an additional α-helix (α6) that corresponds to residues 130–144, as well as an additional two to three residues resolved on either side of this helix depending on the peplomer. In each of the six copies of NP_1–340 in the asymmetric unit, α6 extends from the NP core in a different location and orientation (SI Appendix, Fig. S3).

Consistent with the previously reported RNA-free structures of LASV NP (16, 19), NP_1–340 has a compact, mostly α-helical structure consisting of head and body regions that contain four and eight helices, respectively. The head region is formed by residues 8–24, 83–122, and 261–340, but the body is formed by residues 25–82 and 123–260 (Fig. 1A). The fragments released through pepsin digestion are separate collinear sections that each contain part of the head and part of the body (Fig. 1B).

Fig. 1.

Structure and RNA binding of the N-terminal domain of LASV NP. (A) Cartoon representation of NP_1–340, colored by head (tan) and body (blue) regions of LASV NP_1–340. Residues 113–127 are disordered. (B) NP_pep colored by 21-kDa N-terminal (yellow) and 14-kDa C-terminal sections that result from pepsin digestion (teal). Note that the head and body regions are made by interweaving of the N- and C-terminal sections of polypeptide. Residues 112–126 and 144–166 are disordered, and residues 124–143 were removed by digestion with pepsin. In A and B, for clarity, only the protein portion of the complex is illustrated. (C) Electrostatic surface potential of NP_pep calculated using APBS (34) shows a deep basic groove through which the single-stranded RNA channels. The tunnel is between 6 and 16 Å wide and at position Ar3, is recessed 10 Å from the protein surface. Positive surface is colored blue; negative surface is colored red with limits ± 10 k_bT/e_c. (D) The side chains of positively charged and other polar residues that interact with the phosphate backbone and bases of the single-stranded RNA are labeled. The eight nucleotides are colored from red (3′; Left) to blue (5′; Right). Residues and secondary structural elements of the head domain are colored tan; residues and secondary structural elements of the body domain are colored blue. Although many of the residues are nonspecific and have thus been built as uridine or cytosine, residues 3 and 8 are clearly purines and have been built as adenosine.

Six RNA nucleotides, corresponding to bases 2–7, are resolved in NP_pep (SI Appendix, Fig. S4 A and B). Bases 2–7 are resolved in NP_1–340 as well as bases 1 and 8, depending on the peplomer (SI Appendix, Fig. S4 C and D). Bases 1–4 are nearly identical in all NP peplomers, but bases 5–8 show small deviations in their relative positions. Although NP was predicted to bind RNA between the N- and C-terminal domains, this RNA-bound structure indicates that instead, the RNA is bound by the N-terminal domain in a deep, basic crevice that channels between its head and body regions (Fig. 1C).

Although each NP-RNA monomer binds random RNA from the expression host, position 3 in each RNA strand is clearly a purine (SI Appendix, Fig. S4 A and B), and is anchored between R300 and Y308. In addition, when visible, position 8 in all but one RNA strand can also be built as purine. Hence, although the overall binding of RNA by NP is thought to be nonspecific in nature, NP may have some unexpected, partial sequence specificity. However, whether LASV prefers this motif in the actual genome, or simply in static binding of ssRNA in the expression host, is as yet unclear.

An extensive network of interactions anchors the RNA phosphate backbone and the rest of the bases to the protein (Fig. 1D). Bases 2–4 demonstrate the highest level of interaction with the nucleoprotein through key residues on α12, α14, and η2. These α-helices contain all of the arginine and lysine residues responsible for binding the RNA backbone, as well as Y308, which stacks against Ar₃, forming a strong π-interaction with the six-membered ring of this base. RNA residues 1 and 5–8 show more modest interactions, primarily through a hydrogen-bond network between the phosphate backbone and several threonine and serine residues of NP.

RNA Binding Is Controlled Through a Gating Mechanism.

The previously determined RNA-free structure of LASV NP suggested that the basic crevice in the N-terminal domain was responsible for binding to the m7GTP cap of mRNA (16). Attempts to cocrystallize NP with an m7GTP analog were unsuccessful, but the single nucleotides dUTP and dTTP could be soaked in and visualized in the basic crevice (16). Notably, the location of each of these nucleotides is essentially equivalent to that of Ur₂ in the ssRNA-bound structure, suggesting that the soaked-in dUTP and TTP may not be mimicking cap, but rather, are reflecting the positions of individual nucleotides within the RNA strand (SI Appendix, Fig. S5). Indeed, there is as yet no direct evidence that LASV NP binds cap and our attempts to bind various NPs to cap-conjugated agarose beads were unsuccessful (SI Appendix, Fig. S6). Furthermore, recent biochemical analysis of several residues previously proposed to be involved in cap-binding demonstrates that mutation of these residues results in defects in antigenome synthesis rather than defects in mRNA levels (19). Hence, the binding of ssRNA observed here combined with the positional homology of the individual, soaked-in nucleotides indicates that this site is likely not a binding site for cap, but rather is a binding site for the viral genome.

The RNA-free structure also includes the C-terminal domain of NP. In this context, LASV NP forms a trimer with N- and C-terminal domains oriented in a head-to-tail fashion, forming an RNA-free ring of NP monomers. In the RNA-free structure, helix α5 is extended by 10 residues (residues 112–122) and both the extended helix α5 and helix α6 lie across the RNA-binding crevice, occluding access to RNA. The C-terminal domain interacts with these helices and the loop connecting them to stabilize this “closed,” trimeric form of RNA-free NP (SI Appendix, Fig. S7). In contrast, in the RNA-containing NP_1–340 structure, residues 112–122 are mobile and disordered. As a result, helix α5 is shorter and terminates before the crevice. In addition, helix α6 is rotated away from the crevice. Furthermore, the loop connecting α5 and α6 (residues 232–243) is shifted outward in the presence of RNA and appears to form a “gate” (Fig. 2). When in the “open” conformation, several residues in this loop interact with the RNA backbone and sugar moieties. However, in the closed conformation, this loop would clash with bound RNA. Thus, it appears the trimeric form of Lassa NP would be unable to bind to RNA.

Fig. 2.

Comparison of the RNA-free and -containing structures of the LASV NP N-terminal domain. RNA binding seems to be regulated through conformational changes that open and close the RNA-binding pocket. In the RNA-free conformation (gray), the RNA gate is closed. Helix α5 is extended across the RNA-binding pocket and α6 is positioned on top of the pocket, preventing RNA from binding. When RNA is bound (blue), residues 112–122 are disordered, α6 shifts away from the pocket, and the RNA gate opens to accommodate the ssRNA.

To determine if the trimeric ring arrangement causes the C-terminal domain to stabilize α5 and α6 in the closed form of LASV NP, we made the mutations S45R and K189E to residues that lie at the NP–NP interface in the trimeric structure (SI Appendix, Fig. S8A). Correspondingly, although WT NP elutes from a size-exclusion column as a monomer, trimer, and hexamer, mutant NP_S45R,K189E elutes as a dimer and tetramer (SI Appendix, Fig. S8B). To further investigate the potential structural changes resulting from this change in oligomeric state, we performed DX/MS analysis on both the WT and mutant NP_S45R,K189E. WT NP demonstrates low-moderate exchange rates for α5, α6, and their connecting loop (SI Appendix, Fig. S9A). In contrast, NP_S45R,K189E shows very high exchange rates for this region, similar to those rates of NP_1–340 from which the C-terminal domain is absent (SI Appendix, Figs. S9B and S2A, respectively). Furthermore, residues of the C-terminal domain that interact with either the N-terminal domain of the same NP (α17), or the N-terminal domain of the next NP in the ring (α22, and β10 and β11), also show alterations in the amide hydrogen exchange rates when mutated. Additionally, although recombinant WT NP exhibits a A260/280 ratio of 0.95, NP_S45R,K189E exhibits a ratio of 1.3, suggesting that more RNA is bound to mutant NP in which the trimeric interface is disrupted than to WT NP. Taken together, these results indicate that the trimeric form of NP, as previously crystallized, constitutes and stabilizes a closed conformation. In order for RNA to bind, NP must undergo structural changes that may ultimately result in reorganization of the oligomer.

Residues Within the RNA-Binding Pocket and at NP Interfaces Are Important for Transcription and Replication.

To characterize the role of residues involved in binding of RNA, we examined the ability of a panel of NPs with mutations to the RNA-binding crevice to promote expression of a CAT reporter gene within a LASV minigenome (MG) (Fig. 3). Mutations to residues that contact RNA bases 1–4 essentially abrogated MG transcriptional activity compared with WT NP. Mutations to residues that contact RNA bases 5–8 show less pronounced effects. S247A, which contacts base 5, and Y213A, which contacts base 8, result in ∼40% the transcriptional activity of WT, but mutations to residues that contact bases 6 and 7 have no effect on MG transcription. Overall, expression levels of NP with these mutations is similar to that of WT and all of these mutants are recognized by conformational anti-Lassa NP antibodies (SI Appendix, Fig. S10). Hence, loss of MG activity is likely a result of loss of RNA binding rather than defects in protein expression or structural integrity, confirming the importance of the observed RNA-binding site to NP function.

Fig. 3.

RNA binding, N–C interface, and NP–NP interface residues and their roles in viral RNA transcription. (A) Locations of mutations made to residues in three functional sites in the full-length NP: within the RNA-binding crevice, at the N–C interface, which is between the N- and C-terminal domains of one NP monomer, and at the NP–NP interface between two different NP monomers. (B) Ability of this panel of NP mutants to transcribe a CAT reporter gene in a LASV MG assay. The large array of mutations made in the RNA-binding crevice is subdivided into those mutations that contact RNA bases 1–4, those mutations that contact bases 5–8, others deep in the crevice, and a key glycine at the base of the RNA-binding gate.

In contrast, mutation of W164 or F176 to alanine also prevents MG activity, but the resulting mutant NPs are not recognized by conformational anti-NP antibodies. These residues form a hydrophobic pocket adjacent to bases 1 and 2. Although it was previously proposed that these residues formed a portion of the cap-binding pocket (16), it is possible that these residues are instead important for the structural integrity of the protein. Other mutations to residues that lie within a deep pocket of the RNA-binding crevice (R17, Y209, E266, and Y319), yet do not directly contact the RNA itself, also prevent CAT expression. These residues coordinate two water molecules and appear to be important for the structural integrity of NP as overall expression levels of these mutants are greatly reduced compared with WT NP, and the E266A and Y319A mutants are not recognized by conformational anti-NP antibodies. Finally, a mutation to proline of G243, a residue that lies at the base of the RNA gate, also eliminates MG-derived CAT expression. It is likely that mutation of this flexible glycine residue to a more rigid proline locks the gate in a closed position, preventing RNA from binding. G243P NP expresses to levels similar to WT NP and G243P is recognized by conformational anti-NP antibodies. Hence, the experimentally visualized RNA-binding site and the gate that opens it are both critical for MG replication and transcription.

We also sought to characterize the importance of interactions between the N- and C-terminal domains of a single NP, and of one NP with its neighboring NP in RNA synthesis (Fig. 3). At the interface between N- and C-terminal domains, some mutations block activity and others have little effect. Notably, it is those residues that are involved in protein-protein hydrogen bonds between the two domains that appear important for MG activity, but other basic residues that are not involved in protein-protein hydrogen bonds do not seem important. Specifically, K110A, R115A, R118A, and W331A mutations prevent or greatly reduce CAT expression. Each of the basic residues K110, R115, and R118 form hydrogen bonds with neighboring residues or the carbonyl backbone, but W331 stacks against the aliphatic chain of K110. Although a K110A mutation blocks CAT activity, a K110E mutation in which the basic Lys is replaced by an acidic Glu unexpectedly doubles CAT expression to 200% of WT. A K110E/S111R double mutation has ∼40% of WT CAT expression but a K110E/S111E double-mutation has no MG activity, suggesting the additional importance of residue 111.

In contrast, mutations to other basic residues within the N–C interface that do not specifically form salt-bridges or hydrogen bonds to the other protein domain (K167A, R556A, and R561A) show only a modest reduction in CAT expression (65–75% the activity of WT). These results suggest that interaction between the N- and C-terminal domains is important for replication and transcription but do not necessarily rule out the possibility of a secondary RNA-binding site between the two NP domains.

Other sites, beyond specific RNA contacts are also important for NP function. Outside of a single NP peplomer, mutations that prevent NP–NP interactions (S45R, K189E, R55A/K56E, D437R, D500R) also eliminate CAT expression. Hence, NP–NP association is critical for replication and transcription, perhaps to allow processive movement of the polymerase L.

Discussion

Crystal structures of the nucleoprotein-RNA complexes from the nonsegmented negative-strand RNA viruses rabies (RABV), vesicular stomatitis (VSV), and respiratory synsyctial viruses (RSV) demonstrate that RNA binds nonspecifically in a basic groove between the N- and C-terminal domains of N (20–22). In contrast, for LASV, a segmented RNA virus, RNA binds within the compact N-terminal domain.

Structures of nucleoproteins from other segmented viruses (influenza and Rift Valley fever) have been determined in the absence of RNA (23–25), but this work represents a unique structure of an NP from a segmented virus in complex with RNA. Comparison of the structures reveals some organizational features that appear to be in common among the NP and RNP complexes of these segmented viruses. The Ns of influenza virus and Rift Valley fever virus and the N-terminal domain of LASV NP are compact, rather than more extended like the Ns of the nonsegmented RABV, VSV, RSV, and Borna disease virus. Furthermore, electron microscopy of RNPs of the segmented Pichinde arenavirus, influenza, and Rift Valley fever viruses demonstrates the RNP complexes of segmented viruses are less helical in nature than those of the nonsegmented viruses (26–28). Furthermore, LASV NP could have some partial sequence specificity or preferences, but no specificity was observed in N-RNA complexes of the nonsegmented viruses RSV, RABV, and VSV.

Comparison of the RNA-free and RNA-containing structures and accompanying DX/MS analysis of LASV NP indicate that RNA binding is controlled through a gating mechanism of conformational changes within the N-terminal domain. Taken together with mutational analysis of LASV NP, these results suggest a possible model for how the RNP might be organized. In this model, the trimeric form of NP could represent “NP⁰,” in which trimerization performs the equivalent function of the P protein of other negative-strand viruses, and therefore prevents RNA binding (Fig. 4 A and B). Upon RNA binding, which could be triggered by an as-yet identified factor or perhaps the viral genome itself, the C-terminal domain of NP rotates slightly away from its position in the RNA-free trimer, allowing helices α5 and α6 to open away from the RNA-binding crevice (Fig. 4C). With RNA bound, NP will no longer be able to form the trimeric ring and instead may form another arrangement such as a linear chain of NP molecules that line the ssRNA, with NP–NP interactions mediated through the N-terminal portion of one NP and the C-terminal portion of another (Fig. 4D). Such an assembly would create a continuous head-to-tail polymer attached to the RNA, in which only the RNA bound within the crevice of the N-terminal domain is resistant to RNase attack. We indeed note that only the portion of RNA specifically bound by each N-terminal domain is resistant to RNases, with intervening nucleic acid easily clipped. A continuous head-to-tail polymer of NP would presumably also help the polymerase move along the template by transiently displacing NP (to copy the template), and repositioning NP on the template once the polymerase has passed. An alternative model in which the RNA binds both inside the N-terminal crevice and between the N- and C-terminal domains is also possible as this interface does contain several basic residues. However, the majority of these basic residues are involved in hydrogen bonds to the other NP domain. Mutation of other basic residues at this interface that are not involved in protein–protein interactions results in only modest effects on replication and transcription, suggesting that NP likely binds RNA without the strict requirement of basic residues between the N- and C-terminal domains.

Fig. 4.

A model for arenavirus RNP organization. (A) Organization of the trimeric, RNA-free LASV NP (Protein Data Bank ID code 3MWP). (B) The N-terminal domain of LASV NP colored by electrostatic surface potential and the C-terminal domain modeled as a cartoon (green) show the closed form of the NP structure. In this conformation, the RNA-binding crevice is not available to accept ssRNA. (C) To bind the viral genome, the C-terminal domain must shift away from the RNA-binding crevice to allow RNA to enter. This shift could be initiated by binding of NP by an as yet unidentified cofactor or perhaps the viral genome itself. (D) When bound to ssRNA, the trimer of NP will not form. Instead, monomers of NP line the ssRNA backbone. Each N-terminal domain of NP interacts with the adjacent C-terminal domain of a neighboring NP.

How is the RNA contained in the RNP read by polymerase during RNA synthesis? The narrow crevice through which ssRNA tunnels through the N-terminal domain of NP and the nuclease-resistance conferred by NP binding together suggest that the bound portion of RNA is not available to form duplexes during replication. Furthermore, the depth at which the ssRNA is bound likely renders it unable to form a base pair with a complementary strand when in complex with NP. A scenario that fits with our crystallographic, DX/MS, and functional data is a viral polymerase-induced conformational change within NP that transiently exposes the RNA for replication and transcription. The ability of mutants that block NP–NP interactions to also block viral RNA synthesis suggests that the polymerase may need to interact with the N- and C-terminal domains of adjacent NP molecules simultaneously to induce this conformational change or to move along the RNP. Residues involved in NP–L interactions, and the role, if any, that L plays in mediating NP–NP interactions are unknown. However, DX/MS of NP, purified in the absence of L, does support the existence and importance of NP–NP interactions.

The tight sequestering of RNA and likely conformational change needed for replication and transcription makes NP an excellent target for small molecules that either prevent NP binding of RNA or that stabilize the RNP to prevent conformational changes induced by the polymerase. The arenavirus NPs are well conserved, with a mean 73% similarity overall calculated for six family members spanning both the Old and New World groups (SI Appendix, Fig. S11A). Importantly, each residue that makes contact with the RNA is 100% conserved across the family (SI Appendix, Fig. S11B). The depth and intimate interactions by which the purine at position 3 packs in between Y308 and R300, coupled with the conservation of these two amino acids across the arenavirus family, makes this region a particularly attractive target for inhibitors of arenavirus replication.

The work presented here also provides a structural template by which we may determine the precise role of specific amino acids of NP in their functional interactions with the arenavirus polymerase and explore the relative roles of NP assemblies in the viral life cycle, in this and other families of pathogens that threaten human health.

Materials and Methods

Crystallization of NP_1–340 and NP_pep.

Crystals of NP_1–340 and NP_pep were obtained by hanging drop vapor diffusion using a 1:1 ratio of well solution to protein at 10–12 mg/mL. Data for NP_1–340 were collected at the Stanford Synchrotron Radiation Lightsource, Beamline 12.2. Data for native and SeMet-derivatized NP_pep crystals were at the Advanced Light Source, Beamline 8.3.1.

Structure Determination.

Structure determination of NP_pep using SIRAS (29) was carried out by AutoRickshaw (30). One copy of NP_pep was found in the asymmetric unit. Molecular replacement for the structure of the complete, undigested NP_1–340 was performed with PHENIX (31) using just the protein portion of the model derived from the 1.8Å NP_pep structure. Six copies of NP_1–340 were found in the asymmetric unit.

m7GTP Pull Downs.

Two hundred micrograms of each protein were incubated with 15-μL m7GTP-conjugated agarose beads for 4 h at room temperature. Beads were washed three times with 50 mM Tris pH 8, 300 mM NaCl. Bound protein was eluted with SDS/PAGE loading buffer. Samples were analyzed by Western blot.

LASV MG Transcription Assay.

The 293T cells were cotransfected with a plasmid encoding a LASV MG under control of the T7 RNA polymerase, pol-II expression plasmids directing expression of the virus L and T7RP, and the corresponding LASV-NP mutants, as previously described (32, 33). At 60 h posttransfection, cell lysates were prepared and assessed for levels of CAT protein using a CAT ELISA kit.