|Home | About | Journals | Submit | Contact Us | Français|
Nudaurelia capensis ω virus (NωV) is a T=4, icosahedral virus with a bi-partite, positive-sense RNA genome. Expression of the coat protein gene in a baculovirus system was previously shown to result in the formation of procapsids when purified at pH 7.6. Procapsids are round, porous particles (480-Å diameter) and have T=4 quasi-symmetry. Reduction of pH from 7.6 to 5.0 resulted in virus like particles (VLP5.0) that are morphologically identical to authentic virions, with an icosahedral-shaped capsid and a maximum dimension of 410 Å. VLP5.0 undergoes a maturation cleavage between residues N570 and F571, creating the covalently independent, γ peptide (residues 571-641) that remains associated with the particle. This cleavage also occurs in authentic virions and in each case it renders the morphological change irreversible (i.e. capsids do not expand when the pH is raised back to 7.6). However, a non-cleavable mutant, N570T, undergoes the transition reversibly (NT7.6↔NT5.0). We used electron cryo-microscopy and three-dimensional image reconstruction to study the icosahedral structures of NT7.6, NT5.0 , and VLP5.0 at about 8, 6 and 6 Å resolution respectively. We employed the 2.8Å X-ray model of the mature virus, determined at pH 7.0 (XR7.0), to establish (1) how and why procapsid and capsid structures differ, (2) why lowering pH drives the transition, and (3) why the non-cleaving NT5.0 is reversible. We show that procapsid assembly minimizes the differences in quaternary interactions in the particle. The two classes of 2-fold contacts in the T=4 surface lattice are virtually identical, both mediated by similarly positioned, but dynamic γ peptides. Furthermore, quasi and icosahedral 3-fold interactions are indistinguishable. Maturation results from neutralizing the repulsive negative charge at subunit interfaces with significant differentiation of quaternary interactions (one 2-fold becomes flat, mediated by a γ peptide, while the other is bent with the γ peptide disordered) and dramatic stabilization of the particle. The γ peptide at the flat contact remains dynamic when cleavage can't occur (NT5.0) but becomes totally immobilized by noncovalent interactions after cleavage (VLP5.0).
Particle maturation is an integral part of the assembly of complex viruses 1. Initiation of assembly occurs when virally encoded protein subunits interact tenuously with one another, and possibly with nucleic acid, to produce unstable, procapsid particles. Procapsid assembly is achieved through a process of self-correcting, intersubunit annealing that places capsid proteins into quasi-equivalent positions in an icosahedral surface lattice. Given an external cue, (often a change in pH) this procapsid is programmed by the subunit tertiary and particle quaternary structures to undergo reorganization into a more stable particle (“capsid”) that may also undergo auto-catalytic processes that render particles infectious. Here we investigate particle maturation in an RNA animal virus to address three questions: (1) How and why do procapsid and capsid structures differ? (2) What drives the transition between these states? (3) Why can capsids that have not undergone a maturation cleavage reversibly revert to the procapsid state, whereas cleaved capsids cannot?
The subject of this study was Nudaurelia capensis ω virus (NωV: family Tetraviridae), a non-enveloped virus with a bipartite, positive-sense RNA genome 2. RNA1 and RNA2 encode the RNA dependent RNA polymerase and the coat protein (CP) subunit, respectively. The capsid contains 240 chemically-identical copies of CP arranged in a T=4 icosahedral lattice (Fig. 1). Expression of the CP gene in a recombinant baculovirus system results in the spontaneous assembly of virus-like particles (VLPs) that package predominantly heterogeneous cellular RNA3. VLPs purified from baculovirus at pH 7.6 were found to be round, porous particles of diameter 480Å. The subunits reorganize into smaller, 410 Å diameter particles when the pH is lowered to 5.04. This pH-induced transition from procapsid to capsid in VLPs occurs in less than 100 milliseconds based on stop-flow experiments5. This process is initially reversible, until an auto-proteolytic cleavage (t1/2=~30 minutes) occurs in at least 15% of the subunits and the virus is then locked in the mature capsid conformation5. Analogy with other non-enveloped, icosahedral viruses such as poliovirus 6 and Flock House Virus 7, suggests that autolytic cleavage in NωV is essential for cell entry and its infectivity. In vivo, NωV is presumably first assembled as procapsid in the neutral pH condition inside the cell. Then acidification caused by apoptosis would trigger its maturation and release 8. With the added stability imparted by maturation, the capsid can then survive the harsh extra-cellular environment before the next infection. The maturation-dependent cleavage occurs between residues Asn 570 and Phe 571 and converts the 70 kDa CP precursor (α) into 62 (β) and 8 kDa (γ) peptides, both of which remain associated with the mature capsid 9. When Asn 570 was replaced by Thr to produce the N570T mutant (“NT”), the maturation cleavage was blocked and purified particles could be transformed reversibly between large and small particle forms with changes in pH (Fig. 1) 10.
The VLP capsid has the same morphology as the capsid seen in crystals of authentic, infectious virions (Munshi, Liljas et al. 1996). The crystal structure of NωV revealed that each 641 amino acid (a.a.) CP subunit is comprised of three domains (Fig. 1). These include an internal, helical domain, a central, canonical viral jelly-roll 11, and an outer, immunoglobulin-like (Ig-like) domain that is likely involved in interactions with host cell receptors. The asymmetric unit of the icosahedral T=4 capsid consists of four subunits, labeled A (blue), B (red), C(green), and D (yellow) (Fig. 1a). They adopt slightly different tertiary conformations because they occupy quasi-equivalent, quaternary structure environments. The virion crystal structure shows that the C and D γ peptides (a.a. 571-641) each contain an ordered helix (a.a. 626-640) and these switch helices serve as wedges that open the C-D dimer contact and create a flat interface with a dihedral angle of 180° (Fig. 1a). These helices are absent in the A and B subunits because the C-termini of the γ peptides in these subunits are disordered, which allows the A-B interface to adopt a bent conformation with a dihedral angle of 138° (Fig. 1a). Hence, the presence or absence of helices in the γ peptides provides the molecular switch that distinguishes the dimers.
Here we report on electron cryo-microscopy (cryoEM) and three-dimensional (3D) icosahedral image reconstruction analyses of the NT procapsid at pH 7.6 (NT7.6) and the NT capsid at pH 5.0 (NT5.0), both of which cannot undergo maturation cleavage, and a virus-like particle at pH 5.0 (VLP5.0) that mimics the wild-type, mature capsid and undergoes normal maturation and cleavage. Each of these structures was compared to and interpreted with the X-ray coordinates of the mature, infectious virion crystallized at pH 7.0 (XR7.0) 12. Each of these particles is described in Table I. Together these data explain the basis for the pH-dependent change in the size of the particle and why the pH 5-induced contraction of the uncleaved particle (NT) is reversible when the pH is raised to 7 whereas the VLP5.0 particle contraction is not reversible. We find that the cleavage in VLP5.0 allows the γ peptide to achieve many more noncovalent interactions than occur in the uncleaved NT5.0 and that these stabilize the compact form, preventing the reversible expansion after cleavage. Indeed, comparing the ordered density in the cleaved form of the γ peptide with the corresponding regions in the uncleaved form revealed that roughly 20 percent more of these residues were ordered in the cleaved form than in the uncleaved form. This explanation is different from that previously proposed by Taylor et. al., that was based on lower resolution structures of the procapsid and capsid 10.
Structural flexibility and instability of the NωV procapsid have likely contributed to failures at obtaining crystals suitable for X-ray crystallography. We therefore employed cryo-EM 3D reconstruction methods to examine the structure of the NT procapsid at pH 7.6 (NT7.6). As peptide cleavage only occurs in the mature capsid conformation, we anticipated that WT and NT procapsids would have the same structure, and this was indeed verified in cryo-reconstructions performed at 20 Å resolution (data not shown).
The 8 Å cryo-reconstruction of NT7.6 shows that the procapsid has a porous surface dominated by openings at the quasi 6-fold (icosahedral 2-fold) axes and closely similar A-B (blue-red) and C-D (green-yellow) subunit dimer contacts (Fig. 2). In each dimer, the Ig domains are separated by 13Å. The otherwise tenuous particle is held together by close ABC and DDD trimer contacts formed among the inner helical domains. Basic residues dominate the N-terminal portion of the subunits (residues 11-40) that are not visible in the X-ray structure. These residues must contribute to the particle stability (and probably assembly) by interacting with the packaged, heterologus RNA. We constructed a pseudo-atomic model of the procapsid by fitting the coordinates of each capsid subunit from the crystal structure of NωV (XR7.0) into the cryoEM density map for one icosahedral asymmetric unit, and then refined the four subunits as independent, rigid bodies. The dominant negative surface charge on the subunit interfaces must create significant electrostatic repulsion at pH 7.6 that is balanced by the nucleoprotein interactions to create the procapsid.
Differences in shape and size between procapsid and capsid are primarily attributable to variations in the A-B and C-D dimer interfaces. In capsids, the A-B and C-D interfaces are distinct and exhibit bent (A-B) and flat (C-D) contacts. In the NT procapsid both interfaces are closely similar to the flat, A-B contacts in the capsid (Fig 2c, d). The procapsid dimers were modeled starting with the C-D dimer from the capsid. The well ordered switch helices in the capsid are dynamic in the procapsid but, on average, must occupy both of the structurally equivalent dimer interfaces as shown in the pseudo atomic model (Fig. 3).
The X-ray structure of the authentic, mature virus (XR7.0) was determined with crystals grown at pH 7.0 12. We compared VLPs at pH 5 with XR7.0 coordinates to determine if any portion of the structure changed (particularly within the internal helical region) when fully mature particles were returned to pH 7, the condition where procapsids exist if they have not been exposed to acidic pH. Compact, VLP5.0 capsids, were obtained by incubating procapsids at pH 5.0 for a period of days. This transformation to a more compact structure activates the maturation cleavage that ultimately leads to the virus particle being locked in the capsid conformation (Fig. 1c). The 6-Å VLP5.0 cryo-reconstruction closely matched the XR7.0, atomic structure (Fig. 4). All of the density in the VLP5.0 structure is well accounted for by the X-ray model and indicates that the capsid structure remains essentially unchanged between pH 5 and 7 following maturation. The VLP reconstruction clearly exhibits a break in the density at the cleavage site, consistent with the X-ray model where Asn570 and Phe571 are separated by 8 Å. Thus, within the limits at which such features can be discerned in a reconstruction at 6Å resolution, there are no detectable differences between the X-ray model and the cryoEM density map. This suggests that conformational reversibility in uncleaved particles is not linked to pH-induced conformational changes in the γ-peptide as was previously hypothesized (see Discussion).
A cryo-reconstruction of the NT (uncleaved) capsid at pH 5 (NT5.0), was computed at an estimated resolution of 6 Å. As anticipated, the NT5.0 structure showed continuous density at the site where cleavage occurs in wild-type virions and where the density is interrupted in VLP5.0 capsids. The density for NT5.0 and VLP5.0 correlate well in the external, Ig-like domain and also in the middle, jelly-roll domain. However, there are significant differences in the bottom helical domain on the inner surface of the capsid. NT5.0 has essentially no density for the switch helices (Fig. 5a), which is in stark contrast to the well-ordered density seen in VLP5.0 (Fig. 5b), and implies there is high mobility or lack of defined structure for this region in NT5.0. On average, the region just below the C-D interface must be occupied by protein in NT because hinging is inhibited and the dimer contact is flat. However, the polypeptide in NT is not rigidly fixed as it is in VLP5.0.
Another region where NT5.0 and VLP5.0 differ, is at the pentamer interface (Fig. 5c, d). VLP5.0 density agrees well with the XR7.0 coordinates in this region. Here, γ peptide residues 571-599 of the A-subunit form a well-ordered helix that interacts with the first helix in the B subunit (a.a. 44-55) to form a 10-helix bundle about each pentamer axis. In contrast, NT5.0 shows density only for residues 571-590 and no density is seen for residues 44-55 of the first B-subunit helix. Thus, NT5.0 has five isolated helices that do not interact with each other, significantly reducing the buried surface area due to the missing helix contribution from the B subunits. Cleavage in VLP5.0 particles apparently permits conformational flexibility in the γ-peptides and leads to a substantial increase in intersubunit interactions, generating greater stability of mature particles relative to NT5.0. Thus, at neutral pH, electrostatic repulsions outweigh the modest stability that exists in NT5.0, allowing procapsids to reform. This transition is not possible in the cleaved, more stable VLP5.0 capsid.
The striking feature of the procapsid as observed in the NT7.6 reconstruction is the virtual identity of the dyad contacts between A-B and C-D dimers. These contacts occur at low radii, i.e. within the capsid shell, whereas those in the external portion of the particle are sparse owing to the presence of the large channels that traverse the capsid at the quasi 6-fold axes. Procapsid stability derives mostly from two sources: protein-nucleic acid interactions contributed by the basic region of the capsid subunit between residues 1 and 40, and the closely similar trimeric contacts in the internal, helical domain (corresponding to residues from two neighboring subunits 44-56 and 571-591) in the ABC and DDD trimers. These features are consistent with a mechanism of capsid assembly wherein reversible associations permit chemically identical monomers to organize in a T=4 lattice with minimal difference between quasi-equivalent interactions. The tenuous nature of the procapsid is a consequence of numerous, repulsive interactions between charged, acidic groups on the surfaces of the subunits. At lower pH, this charge is reduced and this activates the dramatic, programmed reorganization of the subunits, leading to closer, yet dissimilar contacts within the two classes of dimers. The distinction of these contacts is obvious in the VLP5.0 reconstruction and is not detectibly different from what was observed in the crystal structure12. Hence, following cleavage, the distinct A-B and C-D interactions persist even when the pH is raised by two units. If cleavage is blocked as occurs as in the NT mutant, the entire process of switching between bent and flat A-B contacts is reversible.
A major motivation for this study was to examine our previous hypothesis for the mechanism of reversibility of NT5.0. Based on procapsid and capsid cryoEM structures at 20Å resolution 4 it was hypothesized that the γ peptide underwent a pH-dependent helix-to-coil transition and that in turn allowed the procapsid to condense into the more compact capsid 10 . In other words, in the pH 7 (helix) form, the expanded state was maintained, whereas in the pH 5 (coil) form, the subunits could condense. Taylor et al. posited that, in the absence of cleavage and hence release of γ peptide, particles would re-expand if the pH were raised back to 7.0 and the region corresponding to the γ peptide reformed the helix. However, with normal cleavage the pH-dependent structural changes in γ were expected to be uncoupled and pH change would not cause expansion. In contrast to the expectations, the cryoEM data presented here clearly demonstrate that there is no significant difference in the γ peptide regions of VLP5.0 and XR7.0, and hence disproves the previous hypothesis.
The cryo-reconstruction of NT5.0 provides the key to understanding the reversible nature of its procapsid-to-capsid transition. In the absence of cleavage, the internal helical domains retain mobility as is evident by numerous regions in NT5.0 that have weak or no density when compared to VLP5.0 and XR7.0. Thus, cleavage in VLP5.0 provides the required flexibility in the helical regions to generate increased, non-covalent, interactions and buried surface area sufficient to withstand the electrostatic repulsion at neutral pH that permits NT5.0 capsids to expand into NT7.6 procapsids.
Preparation and purification of VLPs were carried out as described previously 13. Procapsids were fraction collected in pH 7.6 buffer (250 mM NaCl, 50 mM Tris HCl) from a clean band in the sucrose gradient. Procapsid to capsid conversion was carried out by overnight dialysis at room temperature against pH 5.0 buffer (250 mM NaCl, 50 mM sodium acetate) to allow the capsid to fully mature.
Electron cryo-microscopy was performed with ~1 mg/ml specimen samples and images recorded at 50,000 nominal magnification, under low-dose conditions, and at 200keV on an FEI CM 200 electron microscope as described14. Micrographs were digitized in a Zeiss SCAI scanner (ZI imaging) at 7 μm intervals and 2x2 bin-averaged to give an effective pixel size of 2.8Å at the specimen. Individual particle images were boxed from the micrographs with the RobEM program (http://cryoem.ucsd.edu/programDocs/runRobem.txt) and 3D reconstructions were computed with program suite AUTO3DEM15. 3D reconstructions of the VLP5.0 NT5.0, and NT7.6 particles were computed from 8576, 18205, and 2582 particles, respectively. The resolution limit of each cryo-reconstruction (6 Å for VLP5.0 and NT5.0, and 8 Å for NT7.6) was estimated on the basis of standard Fourier Shell Correlation criteria, using a 0.5 threshold 16.
A pseudo-atomic model of the procapsid was built using the atomic coordinates derived from the crystal structure of the capsid (XR7.0) as a starting model. We used a modified version 17 of program Rsref 18 to dock the X-ray model as a rigid-body into the NT7.6 cryo-reconstruction and then refine it. Programs O 19 and Chimera 20 were used to visualize and analyze the map and model data.
We thank Mr. Kevin Chiang for help with boxing particle images from micrographs. We thank Bob Sinkovits for computational assistance in upgrading AUTO3DEM. This work was supported by grants from the National Institutes of Health RO1 GM54076 (JEJ) and RO1 GM033050 (TSB).