Common Mechanism for RNA Encapsidation by Negative-Strand RNA Viruses

ABSTRACT

The nucleocapsid of a negative-strand RNA virus is assembled with a single nucleocapsid protein and the viral genomic RNA. The nucleocapsid protein polymerizes along the length of the single-strand genomic RNA (viral RNA) or its cRNA. This process of encapsidation occurs concomitantly with genomic replication. Structural comparisons of several nucleocapsid-like particles show that the mechanism of RNA encapsidation in negative-strand RNA viruses has many common features. Fundamentally, there is a unifying mechanism to keep the capsid protein protomer monomeric prior to encapsidation of viral RNA. In the nucleocapsid, there is a cavity between two globular domains of the nucleocapsid protein where the viral RNA is sequestered. The viral RNA must be transiently released from the nucleocapsid in order to reveal the template RNA sequence for transcription/replication. There are cross-molecular interactions among the protein subunits linearly along the nucleocapsid to stabilize its structure. Empty capsids can form in the absence of RNA. The common characteristics of RNA encapsidation not only delineate the evolutionary relationship of negative-strand RNA viruses but also provide insights into their mechanism of replication.

IMPORTANCE What separates negative-strand RNA viruses (NSVs) from the rest of the virosphere is that the nucleocapsid of NSVs serves as the template for viral RNA synthesis. Their viral RNA-dependent RNA polymerase can induce local conformational changes in the nucleocapsid to temporarily release the RNA genome so that the viral RNA-dependent RNA polymerase can use it as the template for RNA synthesis during both transcription and replication. After RNA synthesis at the local region is completed, the viral RNA-dependent RNA polymerase processes downstream, and the RNA genome is restored in the nucleocapsid. We found that the nucleocapsid assembly of all NSVs shares three essential elements: a monomeric capsid protein protomer, parallel orientation of subunits in the linear nucleocapsid, and a (5H + 3H) motif that forms a proper cavity for sequestration of the RNA. This observation also suggests that all NSVs evolved from a common ancestor that has this unique nucleocapsid.

INTRODUCTION

All viruses contain a protein capsid that encapsidates the genomic polynucleotide. The capsid is assembled with multiple copies of one or a few types of protein subunits following certain symmetry. The most commonly studied symmetry is that of an icosahedron, which leads to spherical or prolate virus particles (1–3). In contrast, helical symmetry is used for assembly of filamentous capsids (4). The basic fold of the capsid protein subunit is found to be the same for a large number of virus families, even though they may not be related to each other on a genomic basis. The quintessential example of this is that of the β-barrel fold, which was first found in small plant RNA viruses yet now has been discovered in at least 15 different viral families (5). The architecture of the nucleocapsid is closely tied with the mechanism of replication in view of the fact that the assembly of the capsid packages the viral genome.

For negative-strand RNA viruses (NSVs), eight families have been recognized by the International Committee on Taxonomy of Viruses (ICTV). The unique feature that distinguishes NSVs from the rest of the virosphere is that the nucleocapsid, instead of the naked genome, is used as the template for viral nucleotide synthesis. It is indubitable that the assembly of the NSV nucleocapsid is related to the unique mechanism of its viral RNA synthesis. In each of the NSVs, the nucleocapsid is packaged inside a lipid envelope. The appearance of the nucleocapsid inside the envelope is different from virus to virus. In rhabdoviruses, the nucleocapsid adopts a characteristic bullet shape (6). In paramyxoviruses, the nucleocapsid is filamentous or herringbone-like (7). In orthomyxoviruses, the nucleocapsid has a double-helical structure (8, 9). When the nucleocapsids are released from the virion, they all have the appearance of a coil (10). The genomic RNA encapsidated in the nucleocapsid is protected from RNA nucleases to various degrees, depending on the structure. This protective structure renders the RNA not readily accessible to the viral polymerase. Thus, the viral polymerase must gain access to the encapsidated RNA in order to carry out viral RNA synthesis. Since this is a common mechanism of all NSVs, it is likely that the nucleocapsid of NSVs bears characteristic elements shared by all NSV families. Recognition of these elements helps in defining essential viral functions and revealing the underlying mechanism for NSV replication.

By systematic analyses of the known structures of NSV nucleocapsids, we discovered the common mechanism for genomic RNA encapsidation during replication. This unifying mechanism suggests a common origin of NSV families and presents a clear picture for the functions of the nucleocapsid protein.

MATERIALS AND METHODS

Coordinates were retrieved from the PDB with accession numbers 3PU4 (vesicular stomatitis virus [VSV] N protein [VSVN]), 2WJ8 (respiratory syncytial virus [RSV] N protein [RSVN]), 4H5P (Rift Valley fever virus [RVFV] N protein [RVFVN]), and 4BHH (La Crosse virus [LACV] N protein [LACVN]) (11–14). Superposition was carried out with Fr-TM-align (15). The results are summarized in Table 1.

TABLE 1.

Superposition of N proteins

Protein comparisona	No. of residues aligned (no. of residues in reference structure)	RMSD (Å2)b	TM score
RSVN(2–375) vs VSVN(2–422)	292 (421)	4.97	0.52
RVFVN(49–126) vs VSVN(131–206)	53 (76)	3.13	0.44
LACVN(35–124) vs VSVN(131–206)	50 (76)	4.08	0.35
LACVN(35–124) vs RVFVN(49–126)	55 (78)	3.88	0.40

^aResidue numbers used for each protein are noted in parentheses.

^bRMSD, root mean square deviation.

The N_A320-324-2P protein complex was produced in Escherichia coli as reported previously (16). N_A320-324 is a mutant of VSV N harboring five alanine mutations, at residues 320 to 324. X-ray scattering data on VSV N_A320-324-2P samples were collected on the SIBYLS beamline at the Advanced Light Source (Fig. 1A). Scattering curves were processed with PRIMUS (17). Protein samples ranging from 0.75 to 4.3 mg/ml showed no sign of aggregation by Guinier plot analysis (Fig. 1B). Radius of gyration (R_g), maximum size (D_max), and Porod volumes were calculated with the ATSAS package (Table 2) (18). Bead models (10 in total) were generated from the scattering data with DAMMIN (Fig. 2) (19), and an averaged bead model was calculated with DAMAVER (20). The average χ² value for the bead models was 0.986. EOM (RANCH and GAJOE) was used to build a multidomain model of the N_A320-324-2P complex against the small-angle X-ray scattering (SAXS) data (18, 21). Structures of the VSV P N terminus (P_NT) (amino acids 6 to 34) (22), the dimeric P oligomerization domain (P_OD) (23), the P C-terminal domain (P_CTD)/N protein complex with bound P_NT (24), and an additional unbound P_CTD were used as rigid bodies. The domain structures were derived from the structures reported with PDB accession numbers 2FQM, 3HHW, and 3PMK. No partial restraints were imposed on the individual rigid bodies. Fits of the bead and the multidomain models to the experimental data are shown in Fig. 2C and D. Inspection of the fitted curves showed a dip in the calculated curve at low q values, suggesting that a larger organized species was in solution. Two copies of the EOM model were fit to the experimental curve with SASREF (25). This fit to the curve was a remarkable match, yielding a χ value of 1.581 (single fit, 12.865) (Fig. 2E). The resulting models were superimposed onto the bead models from DAMMIN with SUPCOMB (19, 26). The fit is shown in Fig. 2A and B.

FIG 1.

SAXS analysis of the N⁰-P complex from VSV. (A) Experimental SAXS profile for three concentrations of the protein complex, 0.75, 2.4, and 4.3 mg/ml. Scattering curves were scaled with PRIMUS (17). The concentrations and associated coloring schemes are used in all panels. (B) Guinier plot of the experimental SAXS profile with the fit shown. Plots are shown on a relative scale. (C) Kratky plot. (D) Pair-distribution function calculated with GNOM (54).

TABLE 2.

Radius of gyration, maximum size, and Porod volume calculated from the SAXS curves for the VSV N⁰-P complex^a

Concn (mg/ml)	Rg (Å)	Dmax (Å)	Vp (Å3)
0.75	60.94	209	5.82 × 105
2.4	65.07	227.5	6.27 × 105
4.3	66	229	6.52 × 105

^aR_g was calculated with AUTORG, D_max was calculated with DATGNOM, and Porod volume (V_p) was calculated with DATPOROD, all of which are from the ATSAS package (18).

FIG 2.

Ab initio shape modeling of the VSV N⁰-P complex. (A) DAMMIN (19) was used to produce 10 bead models. These 10 models were used to produce an averaged model with DAMAVER (20). The averaged model (white spheres) is shown with the aligned “dimer” of the N⁰-P complex (red and blue). The dimer was determined with SASREF (25) and superimposed with SUPCOMB (26). The 90° rotations relating each orientation are noted. (B) A representative single-bead model from DAMMIN is shown in beads of cyan. The orientations are the same as those in panel A. The protein model is not adjusted to fit the single DAMMIN bead model. (C) Plot of a single DAMMIN model against the experimental scattering curve. (D) Fit of the N⁰-P model shown in Fig. 3 to the experimental SAXS curve. (E) SASREF (25) was used to rigid-body fit two copies of the N⁰-P complex against the SAXS data. The fit of the SASREF-derived dimer is shown.

RESULTS AND DISCUSSION

Capsid protomer.

The atomic structures of nucleocapsid-like particles (NLPs) have been reported for three NSV families: Rhabdoviridae, Paramyxoviridae (genus Pneumovirus), and Bunyaviridae (genera Phlebovirus and Orthobunyavirus) (11–14, 27–31). A comparison of representative structures from each genus was performed. Since the structures in the same genus are highly homologous, vesicular stomatitis virus (VSV) (11) was selected to represent rhabdoviruses, respiratory syncytial virus (RSV) (12) was selected to represent pneumoviruses, Rift Valley fever virus (RVFV) (13) was selected to represent phleboviruses, and La Crosse virus (LACV) (14) was selected to represent orthobunyaviruses. For each of these structures, it was found that the encapsidated RNA is sequestered in a protein cavity. The capsid protein (N, also known as nucleocapsid protein or nucleoprotein) is first synthesized as a monomeric protein (named N⁰, the capsid protein protomer) before being incorporated in the nucleocapsid. N⁰ remains monomeric through a number of different ways. However, the fundamental requirement to support viral replication is to prevent N⁰ from oligomerization before encapsidation of viral RNA. The historic description that N⁰ is prevented from RNA binding has been proven incorrect (16, 32). N⁰ is not an RNA binding protein but rather a capsid protein that assembles a capsid to accommodate any RNA sequence. Any reported in vitro RNA binding measurements are mainly for nonspecific electrostatic interactions between the negative charges of RNA phosphate groups and the positively charged residues in N⁰. The nature of these nonspecific interactions has no difference from that between any other positively charged protein, for instance, the matrix protein of influenza virus, and RNA.

For rhabdoviruses, the N⁰ form is kept monomeric by forming a complex with the phosphoprotein (P). To support viral replication continuously, it is required that the N and P proteins are expressed in a 1:1 molar ratio (33, 34). An N⁰-P complex was isolated from insect cell expression of rabies virus (RABV) N and P proteins, which contains an N subunit and a dimer of P (35). The same N⁰-P complex was also isolated for a mutant of the VSV N protein when the mutant N protein was coexpressed with P in E. coli (16). The mutations in the VSV N protein changed a stretch of 5 residues to Ala (N_A320-324) and prevented the formation of NLPs. Analytical ultracentrifugation experiments showed that the complex, like the insect-derived complex, has a 1:2 (N-to-P) stoichiometry. This complex was studied by small-angle X-ray scattering (SAXS) techniques (Fig. 1 and Table 2). The shape of the complex was determined by the Ab initio method in DAMMIN and is shown in Fig. 2 (19). Independently, previously determined crystal structures of N and domains of P were modeled against the scattering curves with EOM (18, 21). In this method, the linkers between P protein domains were also modeled by an Ab initio approach, yielding a complete model of the N⁰-P₂ protein complex. A complete model is presented in Fig. 3. The model suggests that dimeric P is associated with N⁰ through multiple sites of interactions. The observation that different parts from each monomer of the P dimer contribute to N binding helps to explain studies where functions of the N-terminally truncated P protein mutant can be restored when a C-terminally truncated P protein mutant is provided in trans (36).

FIG 3.

Maintaining a monomeric N protein. A complete multidomain model of the N⁰-P complex from VSV, as determined by EOM, is presented. The N protein is shown in red, while the monomers of the P protein dimer are shown in yellow and green. Previously determined domains are shown in a cartoon representation, while Ab initio-generated loops are shown as C-α ribbons. The three orange spheres represent positions of Ser60, Thr62, and Ser64 (sites of phosphorylation) of the P protein. These residues have a proximity to the negatively charged patch formed by residues Lys398, Arg399, Lys414, and Lys417 of the N protein (shown in dark blue on the surface). This region is circled and denoted the “PO binding site.” All cartoon drawings in this and the following figures were prepared with PyMol (55).

The P protein has a modular structure, with a flexible N-terminal region, a structured oligomerization domain in the middle, and a compact C-terminal domain (P_CTD). Both the N- and C-terminal regions of P bind to the N protein. In one study, a complex was generated by binding the N-terminal 60 residues of VSV P to the ring structure of an N mutant in which the N-terminal 21 residues were deleted (22). The N terminus of P interacts with the back of the C-terminal half (C-lobe) of N. An α-helix corresponding to residues 17 to 31 of P was shown to occupy the RNA cavity in the structure. Subsequent studies by Chen et al. (37), using a P3A mutant (triple mutations of Ser or Thr to Ala), showed that phosphorylation of P at Ser60, Thr62, and Ser64 is required to prevent N⁰ from encapsidating cellular RNA, which leads to a dead-end product of N and diminishes viral replication. In the SAXS model, Ser60, Thr62, and Ser64 were notably positioned to form charge interactions with Lys398, Arg399, Lys414, and Lys417 of N (here named the “PO binding site”). The crystal structure of the VSV NLP in complex with P_CTD showed that P binds to the C-lobe of N, primarily the C-terminal extended loop and α-helix 13 (24). Combined, the results of SAXS and crystallography suggest that P forms a stable complex with the monomeric N subunit involving interactions with the C-terminal loop, a positively charged patch at the PO binding site, the RNA cavity, and the back of the C-lobe of N. The oligomerization of N is prevented by binding of P, whereby the binding site for the N terminus of P overlaps that for the N terminus of N in the nucleocapsid. The region of P bound at the PO binding site can block the side-by-side contact between N subunits found in the nucleocapsid. Mutation studies showed that the loss of any of these interaction sites could diminish N oligomerization and, thus, RNA encapsidation (16). Upon assembly of the nucleocapsid, N⁰-bound P is removed. N begins to oligomerize and encapsidate viral RNA. The P protein is then recycled to bind another N⁰ subunit.

For RSV, there is no structural report of an N-P complex. However, it was shown that the N-terminal 119 residues precede the oligomerization domain of P in primary sequence (38). The N terminus of RSVP is required for P binding to N⁰ (39). The N-terminal region of RSVP was predicted to be intrinsically disordered, but residues 15 to 25 are predicted to be an α-helix. A homologous structure was recently reported for a complex between the N terminus of P and N of Nipah virus (40). The N terminus of P binds the C-terminal domain of N, where it may exclude binding of the C terminus of a neighboring N subunit to this surface to prevent N oligomerization. RSVP could bind to a similar site in RSVN or another site of N which is required for binding the N terminus of a neighboring N subunit. The exact binding site for the N terminus of RSVP remains to be determined for RSVN.

For RVFV, the structure of the N protein has been solved in two forms: an apo-structure without RNA and an NLP. The first form of RVFVN has its N-terminal helix associated with its own RNA encapsidation cavity (41). Residues 20 to 26 form a helix (α2). Within this helix, the hydrophobic side chain of Trp24 fits into the pocket where the bases of the RNA are sequestered in the nucleocapsid, as observed for the RVFV NLP. In the oligomeric structure of the RVFV NLP, the N terminus of N is moved out of the RNA cavity to interact with the neighboring N subunit (13). These observations suggest that RVFVN is kept RNA-free by sequestering its own N terminus. During assembly, the N terminus of RVFVN rearranges to support capsid formation and RNA encapsidation. In another study, the structures of RVFV empty capsids were also solved (42). The empty capsids maintain the architecture of capsids with encapsidated RNA. The reason for the formation of empty capsids in this case is not understood. It is possible that the N protein is not stable in the overexpressed bacterial milieu or that an unknown protein, either viral or cellular, is necessary to help keep N from self-assembling. Finally, this could be an artifact resulting from production of the protein with an N-terminal thioredoxin fusion.

For LACVN, several structures have been produced. The N-terminal 17 residues of N are highly flexible and could be in a fold-back conformation in the N monomer, based on structures of the N protein complex after RNA was removed (14). The fold-back conformation is similar to the sequestering of the N terminus in N of phleboviruses but to a lesser extent. In the LACV NLP, the N terminus of N assumes a conformation that extends to its neighboring N subunit. The extended N terminus has interactions with the bases and backbone of the encapsidated RNA as well as amino acid interactions with its neighboring N subunit. The C terminus (residues 218 to 235) is extended from the core of N and shows some conformational flexibility but not as much as the N terminus. The elements required for keeping the orthobunyavirus N proteins in a monomeric form could be similar to those for N proteins of phleboviruses.

The essential requirement for N to be competent in encapsidating viral RNA is to remain monomeric. Different strategies may be used to achieve this, including occupying the sites required for oligomerization by P binding, such as in the case of rhabdoviruses and pneumoviruses, or sequestering the N terminus by N itself, such as in the case of bunyaviruses. In rhabdoviruses, the proximity of P binding to the RNA cavity is purely coincidental and only for the formation of a stable N⁰-P complex. This complex is essential to prevent premature N oligomerization rather than to prevent RNA binding. This concept is supported by results reported by Chen et al., who showed that a phosphorylation-deficient triple mutant of P (P3A) that is mutated outside the cavity binding region could not prevent encapsidation of cellular RNA by VSVN (37).

Architecture of the nucleocapsid.

The N protein oligomerizes to encapsidate the single-strand viral RNA. The N subunits are associated with each other in a parallel orientation. A linear nucleocapsid is assembled with the single-strand viral RNA sequestered in the center. In rhabdoviruses, the nucleocapsid is a random coil when it is isolated from the virion (43). The nucleocapsid is packaged into a superhelical structure in the bullet-shaped virion (6). The supersymmetry is imposed on the nucleocapsid by the matrix protein, which has a 1:1 interaction with the N protein in the virion. The matrix protein subunits have direct contacts among themselves to form the two-dimensional (2D) helical mesh under the viral membrane envelope. The same helical symmetry is adopted by the nucleocapsid when packaged into the virion. The fact that the single-strand viral RNA is completely encapsidated prior to being packaged into the virion defines that the true symmetry of the nucleocapsid is linear. In most NSVs, the nucleocapsid appears to be a random coil (10). The nucleocapsid in members of the Paramyxovirinae contains a number of helical segments, but helical symmetry is not strict in terms of pitch and rotation of the N subunits (44). The Sendai virus nucleocapsid exists in at least four different helical states (45). Since the repeating unit in the nucleocapsid is linear along the encapsidated RNA, and helical symmetry is not required for RNA encapsidation, the exact symmetry of the nucleocapsid of members of the Paramyxovirinae, strictly speaking, must be considered linear. In influenza virus (IFV), the linear nucleocapsid is twisted into a rough double helix, with interactions between the two associated strings (8, 9). The helical superstructure is a way to condense the nucleocapsid for packaging into the virion but is not essential for viral RNA encapsidation.

The forces that stabilize the nucleocapsid involve extensive cross-molecular interactions among the N subunits. There are side-by-side interactions between N subunits aligned in parallel. These interactions are critical for capsid formation. The N protein can no longer assemble the nucleocapsid if these interactions are disrupted by mutation (16). The extent of the side-by-side interactions is different from virus to virus, ranging from <200 Ų for LACVN to up to ∼2,200 Ų for VSV. There is also a disparity in the amount of buried surface between adjacent C-terminal domains versus N-terminal domains of each viral N protein (11). The contact areas between neighboring domains in different NLPs are listed in Table 3. The contact areas must have plasticity because these calculated contact areas are derived from an artificial ring structure. In the authentic linear nucleocapsid, these contact areas are likely to be different. The difference between the two halves of the N protein may have some functional implications because the N protein needs to undergo a conformational change to reveal the sequestered template RNA during viral RNA synthesis. It is likely that the domain that has a smaller contact area with the neighboring domains should be opened to reveal the RNA. As shown in Table 3, the N-terminal domain of VSVN has a smaller contact area and is likely to be open during viral RNA synthesis, while the C-terminal domain remains associated to maintain the integrity of the nucleocapsid. Rearrangement of the N-terminal domain to open the cavity is further supported in lieu of the greater association between the C-terminal domains due to additional interactions, as noted below.

TABLE 3.

Areas of contacts between N proteins and numbers of nucleotides bound in the N protein

Virus	Surface area/protomer (Å2)	Area of contact (Å2)				No. of nucleotides in core
		Buried interface complex	Buried interface core N-lobe	Buried interface core C-lobe	Buried interface arms/loops
VSV	23,093.9	5,648.0	552.4	1,572.4	3,523.2	8
RSV	20,134.4	5,117.6	997.1	248.0	3,872.5	6
RVFV	13,884.2	3,576.6	344.6		3,232.0	4
LACV	13,775.4	1,982.7	165.0	63.9	1,753.8	10

In addition to the side-by-side contacts, the most obvious interactions between the N subunits are from the structural elements extended from the core of the N subunit. The core of N and the nucleotides associated with the core act as one structural unit. The number of nucleotides accommodated in the core is listed in Table 3. The registration of each N subunit may be defined by choosing one subunit as the origin (position 0). If the access to the RNA cavity faces away from the reader, −1 refers to the subunit on the left of the origin subunit (5′ end of the encapsidated RNA), and +1 refers to the subunit on the right (3′ end of the encapsidated RNA). In rhabdoviruses, the N-terminal arm (21 residues) extends away from the core to interact with the −1 N-terminal domain (N-lobe), whereas an extended loop in the C-lobe interacts with the +1 C-lobe. The N-terminal arm also interacts with the extended loop in the −2 C-lobe. In RSVN, the N-terminal arm (28 residues) interacts with both the N- and C-terminal domains of the −1 subunit, whereas the extended C-terminal arm, rather than a loop in this case, interacts with the C-terminal domain of the +1 N subunit. In RVFVN, only the N-terminal arm (32 residues) interacts with the −1 N-terminal domain. In LACVN, the N-terminal arm (17 residues) interacts with the −1 N-terminal domain involving only the first 8 residues. The remaining residues interact with the encapsidated RNA. The C-terminal arm of LACVN (18 residues) contains an α-helix that interacts with the C-terminal domain of the +1 subunit. Residues that are involved in capsid formation are illustrated in Fig. 4.

FIG 4.

Oligomerization of the N subunits in the nucleocapsid. Surface representations of the nucleocapsid proteins of VSV (A), RSV (B), RVFV (C), and LACV (D) are shown for three subunits. Each subunit has been radially spaced to expose the surfaces that contribute to subunit interactions. The monomers are shown in red, green, and yellow, while residues that contact the adjacent protomers are shaded in black.

In IFVN, there is also an extended loop in the C-terminal region that interacts with the C-terminal domain in the +1 N subunit. It is not clear if the N-terminal region is involved in the interactions with the neighboring subunits in the nucleocapsid. An atomic model of the nucleocapsid or an NLP is needed to fully address this. It is also not clear how IFVN is maintained in a monomeric form prior to RNA encapsidation. There is no report of any viral protein that is involved in this function. In the reported structure of IFVN, the N terminus of N could be in a sequestered conformation. If this is the case, IFVN may employ a method similar to that used by RVFVN and LACVN to remain monomeric. The extended loop in the C-terminal domain folds back on the surface of its own monomer, as shown by the structure of an obligate monomeric mutant of IFVN (46). This folded conformation of the C-terminal extended loop should also contribute to maintaining a monomeric mode of IFVN prior to RNA encapsidation.

The surface contact and domain swap interactions between neighboring N subunits are essential for the assembly of NSV nucleocapsids. Similar interactions are also common in the nucleocapsids of other viruses, such as picornaviruses and adenovirus (47, 48). The only unique feature of the NSV nucleocapsid is that these interactions are linear along the encapsidated single-strand viral RNA.

Sequestered RNA in the cavity.

The nucleocapsid of NSVs needs to be capable of encapsidating all possible RNA sequences. The mechanism for specific encapsidation of viral RNA may be that viral RNA encapsidation is concomitant with viral replication, which is likely to be at the site of the viral RNA replication complex (24). At this point, the monomeric N protein assembles the nucleocapsid simultaneously with encapsidation of the genome. No consensus interactions with the bases of the encapsidated RNA were identified among all the reported NLP structures, including the cases in which homogeneous sequences were encapsidated in the NLP (49). Interactions of the backbone phosphate groups with positively charged side chains were found for a fraction of the sequestered phosphate groups, but there is not a conserved pattern among NLPs of different viruses. This holds true even for the closely related NLPs of VSV and RABV (32) and suggests that the positively charged residues in the RNA cavity are not the key factor responsible for RNA encapsidation.

What is unique about the viral RNA in the NSV nucleocapsid is that the sequestered bases are stacked to form a motif similar to one-half of the A-form double helix of RNA (Fig. 5A). In rhabdoviruses, the bases of nucleotides 1 to 4 are stacked and face the solvent side of the RNA cavity, whereas the bases of nucleotides 5, 7, and 8 are stacked and face the interior of the N subunit. Similarly in RSV, the bases of nucleotides 2 to 4 are stacked and face the solvent side of the RNA cavity, whereas the bases of nucleotides 5 to 7 are stacked and face the interior of the N subunit. There is an additional “linker” nucleotide between the N subunits in rhabdoviruses and RSV. This nucleotide is still sequestered from solvent accessibility when N subunits oligomerize in the nucleocapsid, but it may allow some flexibility between the N subunits. The base of this linker nucleotide can be stacked with the other bases, or an aromatic side chain will take its place to stack with the other bases when the linker nucleotide is in a transitional conformation (11, 12). In the RVFV NLP, four bases are sequestered by the core domains, each facing the interior of the N subunit. Among these four bases, the two in the center are stacked. The bases of the three “linker” nucleotides are also in a stacked conformation and protected by the capsid structure. In the LACV NLP, seven bases near the 5′ end (nucleotides −11 and 1 to 6) of the RNA strand are stacked and face the entrance to the RNA cavity. The aromatic side chain of tyrosine 177 is intercalated between the bases of nucleosides 7 and 8, and the base of nucleotide 8 is further stacked with the base of nucleotide 9. The bases in this triple stacking face the RNA cavity. Finally, linker nucleotide 10 is sequestered in the RNA cavity, but its base is not stacked. The unique stacking arrangements observed in NSVs allow maximized packaging of the RNA in the capsid.

FIG 5.

Comparisons of RNA cavities. (A) Ribbon drawings showing the N subunit with encapsidated RNA for VSV, RSV, RVFV, and LACV. The rainbow color code of the polypeptide is blue for the N terminus to red for the C terminus. The encapsidated RNA is shown again below each N subunit as stick-and-ribbon models. The nucleotides are numbered 5′ to 3′. The number 1′ for RVFV indicates that this nucleotide is the equivalent of nucleotide 1 in the next N subunit in the NLP. (B) Superposition of the N proteins. C-α tracing of the aligned coordinates is shown. In the first three panels, the C-α tracing for VSVN is shown in gray. The superimposed C-α tracing is shown in rainbow coloring from the N terminus (blue) to the C terminus (red). In the right panel, the C-α tracing for LACVN is shown in sky blue; RVFVN is shown in rainbow coloring. Only the aligned portions are shown (Table 1). (C) Topological cartoons representing the helices in the N protein core. Each circle represents a helix that is labeled according to its secondary structure assignment reported previously (11–14).

Base stacking is stabilized only when the viral RNA is in the cavity of the nucleocapsid. The N subunits can form a stable empty capsid without RNA (16). The encapsidated RNA is a resident in the nucleocapsid, perhaps also contributing to the overall stability of the nucleocapsid. The shape and the electrostatic environment of the cavity define how base stacking occur. The stability of the base-stacking motifs in the RNA cavity is dependent on the RNA sequence, with poly(rA) being the most stable and poly(rU) being the least stable (49). Specific sequences in the sequestered RNA genome regulate many viral functions, such as transcriptional initiation and termination. In terms of transcriptional termination, there is a highly conserved U7 track at the end of each coding region in rhabdoviruses. The structure in this region is the least stable in the nucleocapsid. It is conceivable that such instability could promote dissociation of the viral transcription complex. Mutations in the N protein have been shown to alter the level of mRNA synthesis. In this case, genomic mutations have subsequently occurred, resulting in an extension of the U track to U8 in rescued VSV (50). U-rich sequences are also found in the intergenic regions in other NSV genomes but not in all of them.

The RNA cavity is formed by secondary structural elements from the N and C domains in the N protein. A highly conserved (5H + 3H) motif consisting of 5 helices from the N domain and 3 helices from the C domain has been identified to constitute the RNA cavity (51). When the structures of VSVN and RSVN are superimposed using the Fr-TM-align method (15), the (5H + 3H) motif from each N protein (α4 to α8 in the N domain and α9 to α11 in the C domain of VSVN [11] and αN3, αN5 to αN8, αC1, η2, and αC3 of RSVN [12]) could be superimposed as a rigid body (Fig. 5B and C and Table 1). η indicates a 3₁₀ helix. In addition, two helices following the (5H + 3H) motif in the C domain could also be superimposed as part of the core in these two N proteins (Fig. 5B and C). In the case of RVFVN, only the first four helices (α3 and α5 to α7 in RVFVN [13]) in the (5H + 3H) motif may be aligned with those in VSVN or RSVN (Fig. 5B and C and Table 1). The fifth helix of the motif is present but moved away from the RNA cavity, inducing reorientation of the following three helices in the (5H + 3H) motif as well as helices following the motif. As a result, this RNA cavity is almost completely collapsed compared to that in VSVN or RSVN. This small RNA cavity could accommodate the bases of only four nucleotides while excluding the ribose-phosphate backbone. LACVN (14) has the same topology as RVFVN (Fig. 5B and C and Table 1). The first four helices of LACVN in the (5H + 3H) motif (α2, η1, α3, and α4) correspond to those of RVFVN. However, there are two noticeable differences between the two structures. One is that the linker between the first and second helices is changed from a helix (α4) to a β-hairpin (β1-4). The other is that the third helix in RVFVN (α4) is much shorter than that in LACVN (α3). The three helices in the C-terminal domain of the (5H + 3H) motif are the same in RVFVN and LACVN, as are three additional helices following the motif, in terms of topology. In addition, LACVN has an extra helix at the C terminus (α11). This helix is involved in interactions with the neighboring N subunit on the right side (+1), as described above.

Summary.

The nucleocapsid of NSVs is assembled by the universal assembly principle of all viral nucleocapsids. Oligomerization of the protein subunits forms the capsid that encapsidates the viral polynucleotide. Extensive interactions cross protein subunits, including domain swaps. These interactions are involved in the stabilization of the nucleocapsid. What is unique about NSVs is that their nucleocapsid has a linear symmetry. The viral RNA is a string in the center of the nucleocapsid ribbon. The nucleocapsid assembly of all NSVs shares three essential elements: a monomeric capsid protein protomer, parallel orientation of subunits in the nucleocapsid, and the (5H + 3H) motif that forms a proper cavity for sequestration of the RNA. During viral RNA synthesis, the nucleocapsid does not disassemble to completely release the viral RNA template. Instead, conformational changes must occur in protein domains to temporarily release the RNA template on a local rather than a global basis. The viral polymerase complex is a likely candidate to induce this conformational change in order to gain access to the sequestered bases, some of which face the interior of the N subunits. Once revealed, the RNA template is available for transcription and genomic replication. After the viral polymerase complex passes, the RNA is tidily repositioned in the encapsidation cavity, and the integrity of the nucleocapsid is restored.

This new mechanism of RNA encapsidation allows the N protein to play a role in viral replication and transcription. Mutations in the N protein may increase or decrease the level of viral replication (52, 53). More interestingly, some of these mutations in the N protein did not change the level of viral replication but changed the level of transcription. This suggests that the structure of the nucleocapsid itself has a role in viral transcription that is independent of replication. For viral replication, the viral polymerase complex needs to load at the 3′ end of the genome and process through the nucleocapsid. For viral transcription, on the other hand, the viral polymerase complex needs to recognize the promoter and the transcription termination sequences, in addition to loading and processivity. There may be a unique structural feature associated with the regions where the promoter or the transcription termination sequence is located in the nucleocapsid to allow such recognition.