Structure of L-A virus: a specialized compartment for the transcription and replication of double-stranded RNA

Abstract

The genomes of double-stranded (ds)RNA viruses are never exposed to the cytoplasm but are confined to and replicated from a specialized protein-bound compartment-the viral capsid. We have used cryoelectron microscopy and three-dimensional image reconstruction to study this compartment in the case of L-A, a yeast virus whose capsid consists of 60 asymmetric dimers of Gag protein (76 kD). At 16-A resolution, we distinguish multiple domains in the elongated Gag subunits, whose nonequivalent packing is reflected in subtly different morphologies of the two protomers. Small holes, 10-15 A across, perforate the capsid wall, which functions as a molecular sieve, allowing the exit of transcripts and the influx of metabolites, while retaining dsRNA and excluding degradative enzymes. Scanning transmission electron microscope measurements of mass-per-unit length suggest that L-A RNA is an A-form duplex, and that RNA filaments emanating from disrupted virions often consist of two or more closely associated duplexes. Nuclease protection experiments confirm that the genome is entirely sequestered inside full capsids, but it is packed relatively loosely; in L-A, the center-to-center spacing between duplexes is 40-45 A, compared with 25-30 A in other double-stranded viruses. The looser packing of L-A RNA allows for maneuverability in the crowded capsid interior, in which the genome (in both replication and transcription) must be translocated sequentially past the polymerase immobilized on the inner capsid wall.

Figure 5

Spectra calculated from powder patterns of genome-revealing difference images of L-A virions. The positions and spread of the peaks in these spectra give statistically representative estimates of the predominant spacings between RNA filaments packed in L-A virus capsids. Because the shape of a spectrum depends on the defocus of the micrographs from which it was calculated, this analysis was repeated for micrographs covering a range of defocus values. In each case, difference images were calculated as shown in Fig. 4, their diffraction patterns calculated and azimuthally averaged, and the powder pattern obtained by summing over many such particles. The various spectra are indexed according to the serial numbers of the micrographs used. The closest-to-focus spectrum is at top, and the furthest-from-focus one is at bottom (Table I). Each spectrum shows a broad peak in the range of (47 Å)⁻¹ to (31 Å)⁻¹ (gray vertical band). Also shown is the control spectrum calculated from empty particles in micrograph #7261. It shows a much weaker peak in the same frequency range indicative of the defocus value at which this micrograph was recorded, and confirms that the strong peak in the full particle spectrum reflects scattering from the viral RNA. For reference, a line is shown at (26 Å)⁻¹, the average interduplex spacing for many other dsDNA and dsRNA viruses.

Figure 4

Computer-filtered images of encapsidated L-A genomes. Typical full and empty particles are shown in a and e, respectively. b and f show the same particles after low pass filtering at (28 Å)⁻¹. c and g were obtained by reprojecting the corresponding density maps in the appropriate viewing orientations, and reproduce the micrographs well, particularly in terms of peripheral detail. d and h are “difference images” calculated by subtracting c from a after nullifying the internal density of the “full” map, and g from b, respectively. d reveals the encapsidated genome. The control h simply represents background noise, and confirms that the shell is cleanly removed in this procedure. The two particles shown are from the highest defocus micrograph analyzed (2-μm defocus; first CTF zero at [27 Å]⁻¹). i–l show two more examples each of exposed genomes from four micrographs with different defocus values. Their first CTF zeros at spacings of 15.5, 18.5, 21, and 23 Å, respectively. In each case, the density map used to calculate the shell contribution to the virion image was calculated from the micrograph in question. These images illustrate the variable appearance of encapsidated genomes, and demonstrate that this variability does not arise from differences in defocus. Bar, 100 Å.

Figure 1

Cryoelectron micrographs of purified L-A virions at different defocus values. (a) ∼0.65 μm; (b) ∼0.9 μm; (c) ∼1.4 μm underfocus. One of the full particles exhibits a ringlike “swirl” pattern (c, large arrowhead). (d) At ∼2 μm underfocus; a “dimeric” capsid is shown (large arrowhead). Filamentous material, presumably RNA, is observed in all four micrographs (small arrowheads). Bar, 1,000 Å.

Figure 2

Three-dimensional density maps of L-A capsids at 16 Å resolution. (a) Stereo views of the outer (upper panels) and inner (lower panels) surfaces of the full capsid, viewed along a fivefold axis of symmetry. The internal contents were computationally removed to expose the inner surface. (b) Transverse central sections taken from the maps of empty (left) and full (right) capsids, viewed along a twofold axis. Darker shades represent higher densities (corresponding to protein and/or RNA). Black arrowheads, fivefold symmetry axes; black lollipops, threefold axes; and white arrowheads, twofold axes. The two shells are virtually identical. In the full capsid (right), the closest contacts between the inner surface of the protein shell and the underlying RNA appear to take place around the lateral twofolds. The white rings just inside and outside the shell represent interference fringes arising from phase contrast. Since the holes in the capsid wall are too small to admit an A-form duplex (see Discussion), it seems likely that the empty capsids lost their RNA through a hole created by the dislodging of one or a few Gag subunits. With the averaging that takes place in calculating a density map, such a loss would not significantly affect the average density at that lattice site, and thus may be reconciled with the essentially identical structures visualized for full and empty capsids. Bar, 50 Å.

Figure 3

Structural organization of the two nonequivalent Gag monomers in the L-A capsid. (a) Spherical sections through the density map of the L-A capsid, as viewed along a twofold axis. These sections correspond to radii of 214, 207, 201, 194, 188, 181, 175, and 168 Å, respectively (left to right, and top to bottom). Here, white tones correspond to high density, denoting protein and/or RNA (i.e., the contrast has been reversed relative to Fig. 2 b). (b) Schematic diagram showing arrangement of Gag subunits in the L-A surface lattice, as viewed along a fivefold axis (compare with Fig. 2 a). The A-subunits are green, and the B-subunits orange. (c) Diagram showing three pentons (each a Gag decamer) clustered around a threefold axis. Five- and twofold axes are also marked. (d) The outer crests of both subunits may be subdivided into three domains: B1, B2, B3, and A1, A2, A3, respectively. Two possible modes of association of neighboring A- and B-subunits into dimers are shown in the lower part of the panel. The one on the left appears to have more extensive intersubunit contacts and therefore may be favored. (e) Closeup views down a twofold axis from outside (left) and inside (right). A- and B-subunits are marked in both cases (A and B, respectively). On the left panel, the largest of the five kinds of holes that traverse the capsid wall are marked (arrows). On the right panel, the Y-shaped motifs seen on the subunits' undersides are apparent. For one A-subunit, this motif is tinged with green and for a B-subunit, with orange. Bars: (a) 100 Å; (e) 25 Å.

Figure 6

RNase III accessibility for L-A dsRNA. Purified RNA (lanes 1–3), dsRNA in intact full particles (lanes 4–6), and dsRNA from disrupted full particles (lanes 7 and 8) were treated with RNase III as described in Material and Methods, in the absence (lanes 2, 5, and 7) or presence (lanes 3, 6, and 8) of 10 mM MgCl₂. Lanes 1 and 4 are controls for purified dsRNA and dsRNA from intact full particles without RNase III treatment, respectively. On the left, molecular weight markers. After RNase treatment, RNA was separated in a 1.5% agarose gel and detected by ethidium bromide staining. The digestion products are diffusely spread out in lane 8, whereas in the other lanes the genome band is either present at the appropriate position or absent, having been digested to small fragments.

Figure 7

STEM dark-field images of unstained L-A RNA molecules. (a) Freeze-dried preparation of purified dsRNA, with a few tobacco mosaic virions (the rodlike particles), included as internal mass standards. At right is a histogram of mass-per-unit length measurements made from these micrographs (n = 53). (b) Purified L-A virions, with some free RNA; at right, the histogram of mass-per-unit length data (n = 275). These measurements were restricted to RNA segments visibly connected to capsids. The arrows indicate densities corresponding to integral numbers of strands per filament, assuming 2.8 kD/nm for an A-form RNA duplex (Results). Black arrows mark even numbers of strands, i.e., multiple duplexes. The precision of the measurements is insufficient to resolve the distribution into a set of well-defined peaks. In principle, these densities could be biased to higher values by residual salt deposits, as a consequence of insufficient grid washing. However, in this case, one would expect greater variability in the data, and the standard deviation of the measurements made on TMV particles to be considerably higher. In fact, the SDs were very similar, at 3.81% for the TMV in the experiments with purified RNA molecules, and 4.48% for those on capsid-connected RNA filaments. Bar, 1,000 Å.

Figure 8

Hypothetical model of a transcriptionally active L-A virion. The polymerase (Pol) is covalently attached to a Gag subunit, which we assume to be integrated into the icosahedral surface lattice, thus immobilizing Pol (Icho and Wickner, 1989). Pol (and the COOH terminus of Gag) are assumed to reside on the inner surface of the shell, to facilitate access to the substrate dsRNA. The shape assigned to Pol in this diagram is arbitrary. In this scenario, it is necessary for the genome to be propelled sequentially past Pol as transcription proceeds. Transcripts are hypothesized to be extruded directly through the largest holes in the capsid shell (Discussion).