Cryo-EM shows the polymerase structures and a nonspooled genome within a dsRNA virus

See allHide authors and affiliations

Science  18 Sep 2015:
Vol. 349, Issue 6254, pp. 1347-1350
DOI: 10.1126/science.aaa4938

Caught in the act of copying

The genomes of double-stranded RNA (dsRNA) viruses consist of about a dozen dsRNA segments enclosed by a protein coat. Inside the host cell, the coat remains intact, and the dsRNAs have to replicate within the coat. Liu and Cheng used cryo–electron microscopy of cypovirus particles to catch the dsRNAs in the act of being copied. The structures revealed that the RNA formed a liquid-crystalline array on which viral enzymes carry out multiple rounds of transcription to replicate the viral genome.

Science, this issue p. 1347


Double-stranded RNA (dsRNA) viruses possess a segmented dsRNA genome and a number of RNA-dependent RNA polymerases (RdRps) enclosed in a capsid. Until now, the precise structures of genomes and RdRps within the capsids have been unknown. Here we report the structures of RdRps and associated RNAs within nontranscribing and transcribing cypoviruses (NCPV and TCPV, respectively), using a combination of cryo–electron microscopy (cryo-EM) and a symmetry-mismatch reconstruction method. The RdRps and associated RNAs appear to exhibit a pseudo-D3 symmetric organization in both NCPV and TCPV. However, the molecular interactions between RdRps and the genomic RNA were found to differ in these states. Our work provides insight into the mechanisms of the replication and transcription in dsRNA viruses and paves a way for structural determination of lower-symmetry complexes enclosed in higher-symmetry structures.

The family Reoviridae causes disease in humans, livestock, insects, and plants. The virions have 10 to 12 segments of dsRNA enclosed in a single-, double-, or triple-layered capsid. The inner capsids (cores) remain intact after the viruses are delivered into the host cell’s cytoplasm, and the RNA-dependent RNA polymerases (RdRps) repeatedly transcribe RNA from the minus-strand RNA genome within the core (1, 2). Assembly of the reovirus cores requires encapsidation of the genomic RNA plus strands, along with a roughly equal number of RdRps. The maturation of the reoviruses is accompanied by RdRps-driven synthesis of RNA minus strands complementary to the plus strands, in turn forming genomic double-stranded RNA (dsRNA) segments within the mature virions (3, 4).

Although structures of viral capsids and isolated RdRp complexes have been studied extensively for more than two decades (313), the structures of genomes and RdRps within viral capsids have thus far evaded determination. In this study, we used cryo–electron microscopy (cryo-EM), in combination with our symmetry-mismatch reconstruction method, to report the structures of RdRps and associated RNAs for nontranscribing and transcribing cypoviruses (NCPV and TCPV, respectively) in the family Reoviridae.

Cypovirus particles were isolated and purified, and viral transcription was assayed (14, 15). We reconstructed the structures of the NCPV and TCPV without imposing any symmetry (see supplementary materials and methods). Our analysis of the NCPV showed that the genomic RNAs and RdRps are located inside the capsid within a region of 510 Å radius. The structure of the genomic RNAs is of spherical outline, is composed of regularly distributed layers that are formed by discontinuous dsRNA fragments running in parallel, and is associated with RdRps (Fig. 1, A to C; fig. S1; movie S1). Each RdRp is anchored at the inner surface of the capsid and surrounded by multiple layers of dsRNA (Fig. 1, B and C). The distance between two adjacent dsRNA fragments within the same layer is fixed at ~25 Å, whereas two adjacent layers are ~30 Å apart. The double helices of both dsRNA fragments located close to the inner capsid surface and interacting with the RdRps have a measured helix pitch of ~28 Å (Fig. 1C). The dsRNA fragment structures located closer to the spherical center are not as well resolved as the those at the periphery (fig. S1). Each RdRp density anchors to the inner surface of the capsid, slightly off-center from the fivefold axis (Fig. 1B) (16). These RdRps and the associated dsRNA fragments appear to exhibit a pseudo-D3 symmetric organization (Fig. 1A and figs. S1 to S3), allowing for 12 distinct locations of RdRps inside a viral capsid: Two groups containing three RdRps (threefold RdRps) each approach and are symmetrically arranged about the threefold axes on opposite sides of the virion, and three groups containing two RdRps (twofold RdRps) each approach and are symmetrically arranged about the twofold axes that encircle the center of the virion (fig. S4). Within the three-dimensional density maps, the average density value of the twofold RdRps amounts to approximately two-thirds of the average density value of the threefold RdRps. In contrast, the dsRNA densities surrounding the twofold and the threefold RdRps are all of similar intensity. We reason that this reflects six RdRps occupying the six positions of the threefold RdRps and only four RdRps occupying the six positions of the twofold RdRps (thus, two-thirds of the average density). Therefore, the total number of RdRps within the capsid is 10, in tentative agreement with the observation that each cypovirus genome contains only 10 RNA segments, with each genome segment being specifically associated with one RdRp (17). Our structural analysis also revealed that TCPV and NCPV have almost identical genome structures (figs. S2 and S5), except for those genome regions that interact with RdRps. Given the great variations of size and the encoded genes of the 10 different genomic RNA segments in each cypovirus, it is likely that the observed D3 symmetry in the dsRNA organization does not reflect the true organization of the RNA genome. The layers of the dsRNA fragment resemble the organization of the cholesteric liquid crystal (18) (fig. S1 and movie S2), which is consistent with earlier evidence that the dsRNA genome forms liquid crystalline arrays within the highly condensed capsid (5). The liquid crystalline model of genome packaging does not support the spool model of Reoviridae genomic organization based on icosahedrally averaged genome structures (1921).

Fig. 1 Structures of the RNA fragments and RdRps within the NCPV capsid.

(A) (Left) NCPV icosahedral capsid structure. (Right) Half of the icosahedral capsid (blue) is removed to show the structures of genomic dsRNA and RdRps (red). One threefold axis and three twofold axes of pseudo-D3 symmetry are indicated by a triangle and ellipses, respectively. (B) Central slice of the NCPV structure showing the genomic dsRNA and RdRps. (C) (Left) Zoomed-in view of the dsRNA. The RdRp has been removed for clarity. (Right) Zoomed-in view of a dsRNA fragment.

Using the presence of D3 symmetry of the RdRp structures, we further improved the resolution of RdRps in NCPV and TCPV from ~6 Å (Fig. 2, A and B) to ~5 Å (Fig. 2C, fig. S6, table S1, and movie S3) by applying this symmetry to the reconstruction. The well-defined densities of the two RdRp conformers in NCPV and TCPV allowed us to build Cα backbone models (Fig. 2, B and C, and figs. S7 and S8). The structure of the TCPV RdRp is very similar to that of the orthoreovirus RdRp λ3 (3). We therefore refer to the domains in the cypovirus RdRp according to the nomenclature of the orthoreovirus λ3. TCPV RdRp has four channels that connect to its central catalytic cavity (Fig. 2D); these channels are used for RNA template entry, nucleoside triphosphate (NTP) entry, template exit, or RNA transcript exit (3). An additional density, which we attributed to the homolog of μ2 protein in orthoreovirus, was observed attached to the surface region between the channels for template entry and NTP entry of each RdRp in both NCPV and TCPV (Fig. 2A and figs. S7A and S8A). Consistent with an earlier biochemical analysis suggesting that μ2 might be an NTP instead of an RNA 5′-triphosphatase (RTPase) (22), this observation implies that the μ2-homology protein is not an RTPase, as it does not have access to the 5′ end of the transcript.

Fig. 2 Structures of the RdRp complexes within NCPV and TCPV.

(A) Location of an RdRp within the NCPV capsid (partially shown). The RdRp density map (gray) is superimposed on its Cα model (magenta). (B) Zoomed-in view of the NCPV RdRp in (A). The μ2-homology protein is removed for clarity. (C) Partial view of the density map (mesh) of the RdRp superimposed on its Cα model. Several N-terminal α helix residues (left) are shown (National Center for Biotechnology Information accession number AF323782). K, Lys; H, His; F, Phe; R, Arg; E, Glu. (D) Cut-open view (cross-eye stereo view) showing the four channels in the TCPV RdRp. Asterisks denote the location of the switch loop. (E) Cα model of the NCPV RdRp (magenta) superimposed on that of the TCPV RdRp (blue).

A dsRNA fragment was observed to bind to each of the 12 RdRp bracelet (C-terminal) domains via nucleotides of the second turn downstream of the dsRNA end in NCPV (Fig. 2, A and B). A caplike structure visible at each of the dsRNA ends appears to be the 5′ end of the plus-strand RNA (i.e., the 3′ end of the minus-strand RNA) (Fig. 2B). In response to binding dsRNA, the bracelet domain of NCPV RdRp shifts compared with that of TCPV: One α helix moves into the transcript exit channel, and another two α helices move into the template exit channel, leading to partial blockage of the two channels (Figs. 2E and 3, A and B).

Fig. 3 Conformational changes between RdRps of NCPV and TCPV.

Cα models are shown in bold. (A) Zoomed-in view of the transcript exit channels in the NCPV and TCPV RdRps shown in Fig. 2E. The transcript exit channel in NCPV is blocked by an α helix from the bracelet domain. (B) Zoomed-in view of the template exit channels in NCPV and TCPV RdRps. The template exit channel in NCPV is blocked by two α helices from the bracelet domain.

In TCPV, the end of the dsRNA approaches the template entry channel of RdRp (Fig. 4, A and B). The dsRNA is unwound at the channel entrance, with one strand (presumably the transcriptional template; i.e., the 3′ end of the minus-strand RNA) inserting into the channel and reaching the RdRp active site, and the other strand (the capped 5′ end of plus-strand RNA) being tethered to the RdRp surface (Fig. 4, B and C). Fitting the orthoreovirus RdRp atomic model into the TCPV RdRp density reveals that TCPV RdRp’s plus-strand RNA binding site superimposes with the proposed RdRp cap recognition site (3) (Fig. 4D). No RNA density was observed inside the template exit channel.

Fig. 4 TCPV RdRp (gray) interacting with RNA (orange).

(A) Location of an RdRp within the TCPV capsid (partially shown). The RdRp density map (gray) is superimposed on its Cα model (magenta). (B) Zoomed-in view of the TCPV RdRp in (A). The dsRNA is shown unwound at the entrance of the RdRp template entry channel. (C) Cut-open view showing the inside of the green box indicated in (B). (D) Zoomed-in view of the cap binding site of the RdRp. The atomic model of orthoreovirus λ3 (purple) is superimposed on the density map of RdRp, showing that the cap recognition site in the orthoreovirus λ3 coincides with the RNA binding site of the RdRp. (E) Cut-open view of template RNA and an NTP. (F) The switch loop, which is presented in a retracted conformation in NCPV RdRp (blue), becomes extended in TCPV RdRp (magenta).

The initiation stage of the transcription is known to be less efficient than the elongation process (23). Thus, a considerable percentage of the RdRps in our sample are initiating transcription, which suggests that the TCPV RdRp structure might represent the initiation stage of transcription. We did not observe any auxiliary proteins at the entrance of the template entry channel, suggesting that this RdRp, like the polymerases from bacteriophage Embedded Image and bursal disease virus (13, 24), is able to separate the two RNA strands in the absence of additional helicases.

We also observed a number of density features in the active site of TCPV and NCPV RdRps. Some density features, which can be assigned to the template RNA and an NTP, were specifically observed in the active site of TCPV RdRp. Models of the first four nucleotides of the template RNA (3′-UCAU) and an NTP can be built into the densities (Fig. 4E and fig. S9). The channels for transcript and template exit, which are partially blocked in the NCPV RdRp, become open in the TCPV RdRp (Fig. 3). These conformational changes are likely required for the NCPV RdRp’s switch to transcription mode. In addition, we observed that a loop present in a retracted conformation in the NCPV RdRp becomes extended in TCPV (Fig. 4F), narrowing the template exit channel; this observation suggests that the transcript will collide with this loop during transcription (Figs. 2D and. 4E). Thus, the loop might prevent the transcript from entering the template exit channel and may instead direct it toward the transcript exit channel during transcription. Presumably, this loop (the “switch loop”) is also involved in replication, where it is in the retracted conformation, as shown for the NCPV RdRp, thus allowing nascent dsRNA to egress from the template exit channel.

On the basis of previous structural studies (3, 4, 15, 16, 25) and our RdRp and RNA structures, we propose a model for dsRNA virus replication and transcription. The replication process is initiated from the 3′ end of the plus-strand RNA (template), and the plus-strand RNA is pulled through the RdRp via the template entry channel (fig. S10A). The dsRNA product exits through the template exit channel (3) and finally binds to the bracelet domain of the RdRp (Fig. 2B and fig. S10B). In NCPV, the interaction between the dsRNA and RdRp keeps the 3′ end of minus-strand RNA in close proximity to the template entry channel so that this end will be ready to act as the transcriptional template (Fig. 2B and fig. S10). Upon transcription initiation, the end of the dsRNA is released from the binding site, which triggers the opening of the transcript and template exit channels, as well as the extension of the switch loop. The released dsRNA is unwound at the entrance of the template entry channel (Fig. 4B), a process that is likely to be catalyzed by the RdRp itself. The 3′ end of the minus strand inserts into the channel and serves as the RNA template, whereas the cap of the 5′ end of the plus strand binds to the RdRp (3). The nascent RNA duplex is split, and the transcript is then directed toward the transcript exit channel by an unknown mechanism, possibly controlled by the switch loop (3) (fig. S10C). The template RNA re-anneals with the plus-strand RNA tethered to the RdRp after it emerges from the template exit channel (3) (fig. S10D). This will ensure that the 3′ end of the template stays at the entrance of the template entry channel, facilitating efficient insertion of the 3′ end into the template channel during the next round of transcription (3). The transcript exit channel opens toward a peripentonal channel in the capsid shell, from which the transcript exits the RdRp before interaction with 5′ capping enzymes (3, 4, 15, 16, 25).

Supplementary Materials

Materials and Methods

Figs. S1 to S15

Table S1

References (2641)

Movies S1 to S3

References and Notes

  1. Acknowledgments: We thank H. Wang for critical discussion of the manuscript, B. Zhu and X. Li for assistance with computer programming, X. Gong for help with model refinement, T. Juelich for language editing, C. Yang for discussion on sample preparation, G. Ji and X. Huang for assistance with cryo-EM, and P. Zhu for support. This research was supported by the National Natural Science Foundation of China (91230116, 31170697, 31370736, and 31000333), the National Basic Research Program of China (2010CB912403), the Program for New Century Excellent Talents (NCET-13-0787), and the Hunan Provincial NSF (13JJ1017). All EM data were collected at the Center for Biological Imaging, Institute of Biophysics, Chinese Academy of Sciences. We declare no competing financial interests. The electron density maps and Cα models have been deposited in the EM Data Bank and the Protein Data Bank (PDB) under EM accession numbers EMD-6321 and EMD-6322 and PDB IDs 3JA4 and 3JA5, respectively. L.C. and H.L. designed the reconstruction algorithm, analyzed the data, and contributed to the writing of the manuscript.
View Abstract

Stay Connected to Science

Navigate This Article