A 3′-end structure in RNA2 of a crinivirus is essential for viral RNA synthesis and contributes to replication-associated translation activity

The terminal ends in the genome of RNA viruses contain features that regulate viral replication and/or translation. We have identified a Y-shaped structure (YSS) in the 3′ terminal regions of the bipartite genome of Lettuce chlorosis virus (LCV), a member in the genus Crinivirus (family Closteroviridae). The YSS is the first in this family of viruses to be determined using Selective 2′-Hydroxyl Acylation Analyzed by Primer Extension (SHAPE). Using luciferase constructs/replicons, in vivo and in vitro assays showed that the 5′ and YSS-containing 3′ terminal regions of LCV RNA1 supported translation activity. In contrast, similar regions from LCV RNA2, including those upstream of the YSS, did not. LCV RNA2 mutants with nucleotide deletions or replacements that affected the YSS were replication deficient. In addition, the YSS of LCV RNA1 and RNA2 were interchangeable without affecting viral RNA synthesis. Translation and significant replication were observed for specific LCV RNA2 replicons only in the presence of LCV RNA1, but both processes were impaired when the YSS and/or its upstream region were incomplete or altered. These results are evidence that the YSS is essential to the viral replication machinery, and contributes to replication enhancement and replication-associated translation activity in the RNA2 replicons.

. LCV genome organization and layout of the 3′ non-coding region. (a) A schematic representation of the LCV genome. Open reading frames (ORFs; 1a -3 and 1-10 in LCV RNAs 1 and 2, respectively) encoding the following predicted viral proteins are as indicated: P-Pro, papain-like protease; MTR, methyltransferase; HEL, RNA helicase; RdRp, RNA-dependent RNA polymerase; HSP70h, heat shock protein 70 homolog; CP, major coat protein; CPm, minor coat protein; and proteins that are named according to their relative molecular masses (indicated by numbers preceded by "P"): P8, P23, P5.6, P6, P6.4, P60, P9, P27, and P4.8. Black bars below the genome map represent DIG-labeled riboprobes (II and VIII) complementary to the corresponding locations in the genomic RNAs. (b) Enlargement of the areas indicated by the dashed-circles (in Fig. 1a) representing the 3′ terminal region of LCV RNAs 1 (top) and 2 (bottom). Numbers above the arrows indicate the nucleotide positions on both RNAs.
Scientific RepoRts | 6:34482 | DOI: 10.1038/srep34482 7 bp to form S3 (Fig. 2a,b). Most of the unpaired nts in the SLs were moderately-highly reactive to BzCN modification (see methods). With the exception of 2 nts in the loop of SL1 in LCV RNA1, all nts in the loop of both SLs in LCV RNAs 1 and 2 were moderately-highly reactive to BzCN modification (Fig. 2a,b), suggesting that these nts were not constrained by interactions with other parts of the RNA. In RNA2, nts that form the the A-U basepair (position 8474 and 8479) and the U-A base-pair (position 8473 and 8480) were moderately reactive to BzCN modification, whereas in RNA1, nts that form similar base-pairs (A-U at nt position 8509 and 8514, and U-A at nt position 8508 and 8515) were unreactive to BzCN modification. This suggests that the loop in SL2 of RNA2 may be larger than that in SL2 of RNA1. The S3s of RNAs 1 and 2 appear very similar, with only one nt (A) at position 8490 in RNA1 being moderately reactive to BzCN (Fig. 2a,b).

Luciferase assays to determine the role of the YSS in translation.
To determine if the YSS contributes to viral translation, we performed in vivo translation assays using a series of F-Luc reporter constructs (see Methods), and translation efficiency was determined by the ratio of the F-Luc/R-Luc measurements (with R-Luc serving as an internal control). In constructs LUC-TMV, LUC-R1 and LUC-R2A, the F-Luc gene was flanked by the 5′ and 3′ NCRs of Tobacco mosaic virus (TMV) RNA, LCV RNA1 and LCV RNA2, respectively (Fig. 3a). Relative F-Luc activity was observed for LUC-TMV (Fig. 3b), demonstrating the TMV NCRs' role in translation. The relative F-Luc activity of LUC-R1 was lower than that of LUC-TMV but significantly higher than that of LUC-R2A, which in turn was not significantly different from that of the water control inoculations (Fig. 3b). These results suggest that 5′ and 3′ NCRs of LCV RNA1, but not those of LCV RNA2, support translation activity. We also assessed the luciferase activity of LUC-R2A(− ) (Fig. 3a), a modified LUC-R2A in which the 3′ NCR (98-nt) of LCV RNA2 was substituted with 98 non-viral nts (taken from the GFP coding sequence). The relative F-Luc activities of LUC-R2A(− ) and LUC-R2A were comparable (Fig. 3b), and both were not significantly different from that of the water control. These results suggest that the 5′ and 3′ NCRs of LCV RNA2 (the latter covering all but 9 nts of the YSS) do not contribute to translation of LCV RNA2. We next broaden the area of analysis by using LUC-R2B and LUC-R2C, both of which contained the complete 5′ and 3′ NCRs of LCV . RNA secondary structures were generated from the RNAstructure software, with benzoyl cyanide (BzCN) reactivity incorporated as pseudo-free energy constraints. Each colored nucleotide corresponds to the level of BzCN reactivity for that particular nucleotide, with black, yellow, and red representing unreactive (0-0.4), moderately reactive (0.4-0.85), and highly reactive (> 0.85), respectively. Four digit numbers placed next to the sequence represent the positions of the nts in the LCV genome. "SL1" and "SL2" denote the right and the left apical stem-loops, respectively, of the Y-shape structure (YSS) in WT LCV RNAs 1 and 2. "S3" denotes the basal, closing stem of the YSS. The stop codon (UAG) encoding P4.8 in RNA2 is labeled "Stop". Gray and blue nucleotides correspond to nucleotides for which information of BzCN reactivity was unavailable and nucleotides of the primer-binding site, respectively. The engineered 3′ linkers are as indicated.
RNA2 in addition to 9 nts (in LUC-R2B) and 302 nts (in LUC-R2C) taken from the region immediately upstream of the 3′ NCR. These extra nts enabled complete coverage of the YSS (LUC-R2B) as well as the YSS plus the entire P4.8 coding sequence and the intergenic region between the P4.8 and P27 ORFs (LUC-R2C) (Fig. 3a). However, the relative F-Luc activities of LUC-R2B and LUC-R2C were comparably as low as that of LUC-R2A and both were not significantly different from that of the water control (Fig. 3b). Studies of Sindbis virus and Saguaro cactus virus (SCV) have shown that incorporating into luciferase constructs different number of additional nts from the 5′ proximal ORF of Sindbis virus sgRNA and SCV gRNA, respectively 27,28 , can result in different levels of luciferase activities. This prompted us to extend the 5′ NCR sequence in LUC-R2A and LUC-R2C by adding to it 99 nts from the proximal 5′ end of the P5.6 ORF (Fig. 1a), yielding LUC-R2D and LUC-R2E, respectively (Fig. 3a). Still, the relative F-Luc activity of LUC-R2D and LUC-R2E were not significantly different from that of LUC-R2A and LUC-R2C, respectively (Fig. 3b). We next deleted SL1 or SL2 in LUC-R2C but the resulting constructs, LUC-R2CΔ SL1 and LUC-R2CΔ SL2 (Fig. 3a), showed no significant changes in relative luciferase activity compared to that of LUC-R2C (Fig. 3b), and to each other. , the 3′ NCR is replaced by 98 nts from the GFP gene (striped box), and in LUC-R2B and LUC-2C, the 3′ NCR is extended by adding 9 and 302 nts, respectively, from the immediate upstream region of the LCV RNA 2 3′ NCR; LUC-R2CΔ SL1 and LUC-R2CΔ SL2: essentially LUC-R2C with stem-loop (SL)1 and SL2, respectively, of the Y-shape structure (YSS) deleted; and LUC-R2D and LUC-R2E: essentially LUC-R2A and LUC-R2C, respectively, except that the 5′ NCR is extended by adding 99 nts from the proximal 5′ end of the P5.6 ORF of LCV RNA2. Complete YSS (black bar labeled "YSS"), partial YSS (unlabeled black bar), and YSS with deleted SL1 (Δ SL1) or SL2 (Δ SL2) are indicated. Light gray and dark gray boxes represent non-coding and coding sequences, respectively. Numbers above the vertical lines in the constructs, except for those at the Translation activity for all of the above F-Luc constructs was also determined using in vitro assays and the results were consistent with those of the in vivo assays (Supplementary Note S1; Supplementary Fig. S1). Collectively, these results demonstrate that the 5′ and 3′ NCRs of LCV RNA2 (including an extensive 400-nts 3′ region that encompasses the YSS) do not contribute to translation of the RNA of the F-Luc constructs. Translation activity was observed for specific F-Luc constructs only when the transcripts were co-inoculated with LCV RNA1 (see later).
Mapping regions in the 3′ end of LCV RNA2 involved in viral RNA synthesis. Results from the preceding section raise the possibility that LCV gRNA2 does not itself possess messenger activity, and beg the question of what role the YSS-containing 3′ NCR might play in the LCV infection process. Here, we used a series of LCV RNA2 3′ NCR deletion mutants to investigate whether it is involved in viral RNA synthesis. Given that the 3′ NCR of LCV RNA1 supports translation, it is not an ideal template for making these mutations since they could affect translation, viral RNA synthesis, or both, thus making negative results difficult to interpret.
Capped in vitro transcripts of each RNA2 mutant were co-inoculated with that of WT LCV RNA1 into tobacco protoplasts, and the effects of the deletions on viral RNA synthesis were determined. (− )-RNA and (+ )-RNA synthesis of 3′ Δ 4 were comparable to that of the WT (Fig. 4a,b). To determine whether additional nts located upstream of those deleted in 3′ Δ 4 (but downstream of those that form the YSS) are involved in viral RNA synthesis, we constructed and tested 3′ Δ 11 (Fig. 4a). This deletion was predicted by mfold to not disrupt the YSS. Viral ). Total RNA (2 μ g each) extracted from the inoculated protoplasts harvested at 24, 48, and 96 hpi (lanes 24, 48 and 96, respectively), and total RNA (2 μ g) from water (mock)-inoculated protoplasts harvested at 96 hpi (lane W) were analyzed using DIG-labeled RNA2 negative-or positive-sense specific riboprobe VIII (Fig. 1a). Hybridization signals of minus-and plus-strand genomic RNA2 [G2(− ) and G2(+ ), respectively] are indicated. Estimation of RNA sizes and methylene-blue stained 25s rRNA equal loading controls are as in Fig. 3.

Mutations targeting SL1 and SL2 in the YSS of LCV RNA2.
To determine if either SL1 or SL2 could support viral RNA synthesis or YSS formation, we made a series of mutations targeting both SLs of LCV RNA2, and co-inoculated the in vitro produced capped transcripts of each mutant with that of WT LCV RNA1 to tobacco protoplasts. Δ SL1 and Δ SL2 (Fig. 5a) failed to accumulate viral RNA (Fig. 5b,c) suggesting that one SL alone cannot support viral RNA synthesis. SHAPE analysis of the in vitro transcripts synthesized from structural cassette plasmids containing Δ SL1 or Δ SL2 revealed that the YSS was lost with the deletion of either SL1 or SL2. (a) Schematic diagrams of the secondary structure of SL mutants: Δ SL1 (with SL1 deleted), Δ SL2 (with SL2 deleted), SLD1-1 (with a 6-nucleotide substitution in the right arm of the lower stem of SL1 engineered to disrupt SL1), SLD1-2 (with a 6-nucleotide substitution in the left arm of the lower stem of SL1 engineered to disrupt SL1) and SLR1 (with a compensatory 6-nucleotide substitution in the left arm of the lower stem of SLD1-1 engineered to restore SL1). Light gray letters in Δ SL1 and Δ SL2 represent the nucleotides deleted from SL1 and SL2, respectively. Bold letters in SLD1-1, SLD1-2 and SLR1 represent non-viral nucleotides engineered to substitute the viral nucleotides. The locations of SL1, SL2 and S3 are as indicated. (b-f) Tobacco protoplasts were inoculated with the in vitro transcripts of wild type LCV RNA1 along with those of LCV RNA2 mutants Δ SL1 (b), Δ SL2 (c), SLD1-1 (d), SLD1-2 (e), SLR1 (f), or wild type LCV RNA2 (WT). Total RNA (2 μ g each) extracted from transcript-inoculated protoplasts 24, 48, and 96 hpi (lanes 24, 48 and 96, respectively), and total RNA (2 μ g) extracted from water (mock)-inoculated protoplasts (lane W; harvested at 96 hpi) were analyzed using DIG-labeled negative-or positive-sense specific riboprobe VIII (Fig. 1a). Hybridization signals of minusand plus-strand genomic RNA2 are indicated as G2(− ) and G2(+ ), respectively. Estimation of RNA sizes and methylene-blue stained 25s rRNA equal loading controls are as in Fig. 3.
However, the deletion of one SL did not eliminate the other SL, although the structure of the latter became slightly altered. For example, in Δ SL1, the basal stem, S3, of the YSS was incorporated into SL2, thereby extending it and forming an additional 1 nt internal loop (compare Fig. 2b,c). In Δ SL2, SL1 was extended due to the incorporation of the nts from the basal stem, S3, of the YSS, and it also formed an additional 5-nt internal loop (compare Fig. 2b,c). Consistent with the previous results, nts in the loop portion of the modified SL1 and SL2 remained reactive to modification (Fig. 2c,d), suggesting that these nts did not interact with other parts of the RNA.
The removal of either SL1 or SL2 could have affected the stability of the transcripts and in turn, directly impacted viral RNA synthesis. To address this possibility, the capped in vitro transcripts of LCV RNA2, Δ SL1, or Δ SL2 were each inoculated to protoplasts, and the transcript levels were determined at 3, 12 and 18 hpi by semi-quantitative RT-PCR. The results showed that there was no significant difference in the amount of RT-PCR products generated from the pCM2 (LCV RNA2), Δ SL1, or Δ SL2 inoculations at all time points ( Supplementary  Fig. S2), suggesting that RNA stability was not affected by the removal of either SLs.
LUC-R2 replicons support translation activity in the presence of LCV RNA1. In LCV RNA2, ORFs that are located downstream of the first ORF (encoding P5.6) (Fig. 1a) are expressed via a nested set of 3′ co-terminal sgRNAs generated from the (− )-RNA 12 . Because transcripts produced from the LCV RNA2 (R2) series of F-Luc constructs alone do not support translation in vivo, this begs the question of how might the cap-dependent translation of P5.6 be achieved. We took the first step towards addressing this question by determining whether or not in the presence of LCV RNA1, RNA transcripts produced from the R2 series of F-Luc constructs could function as replicons to support translation activity. The capped RNA transcripts of constructs: LUC-R2A, -R2B, -R2C, -R2CΔ SL1, -R2CΔ SL2, -R2D and -R2E were each co-inoculated with that of the internal control R-Luc construct and that of LCV RNA1 into tobacco protoplasts, and the cells were harvested at 72 hpi for analyses. (− )-RNA1 and (+ )-RNA1 of LCV were both observed in the Northern blot analyses of all test samples (data not shown). In addition, high levels of (− )-RNA and (+ )-RNA synthesis (determined using riboprobes specific to the F-Luc sequence) were observed in the LUC-R2C and -2E samples. In contrast, RNA synthesis in the LUC-R2A, -R2B, -R2CΔ SL1, -R2CΔ SL2, and -R2D samples were drastically reduced (Fig. 3c). Remarkably, the replication activity of LUC-R2C coincided with a high level of relative luciferase activity (ave. 6.67) (Fig. 3c), while the weak replication of LUC-R2A, -R2B, -R2CΔ SL1, -R2CΔ SL2, and -R2D coincided with weak or absence of relative luciferase activity (ave. 0.009, 0.01, 0.346, 0.163, and 0.007, respectively) (Fig. 3c). Interestingly, while replication activity in the LUC-R2E sample was as strong as that of the LUC-R2C sample, its relative luciferase activity (ave. 2.49) was reduced (by approx. 2.5 fold) relative to that of the latter (Fig. 3c). These results are discussed in the next section, but a general conclusion is that in the presence of LCV RNA1, the YSS-containing 3′ NCR of LCV RNA2 and its upstream 302 nts support RNA synthesis, and this is accompanied by distinct replicon translation activity.

Discussion
Successful proliferation within the host cell requires the genome of (+ )-RNA viruses to undergo translation and replication. In many cases, these processes are aided by RNA structures located near the termini of the gRNAs. For viruses in the family Closteroviridae, a thorough understanding of these processes has been challenging given their large genomes and complex genome expression strategies 9,29 . In this study, we have combined SHAPE with biochemical and biological assays to obtain novel insights into the RNA structure and function of a member of this family. SL1 and SL2 are constituents of a higher-order structure (the YSS) that has hitherto not been experimentally determined. Overall, the YSS is much conserved in LCV RNAs 1 and 2, and does not appear to form a pseudoknot as most of the loop nts of SL1 and SL2 can be readily modified by BzCN, suggesting that nts in the loops do not interact with other parts of the RNA (Fig. 2). However, the potentially larger loop in SL2 of RNA2 (reflected in the moderate activity of the A-U and U-A base-pairs) as compared to that in SL2 of RNA1 (Fig. 2a,b) indicates that the two YSSs are not completely identical, suggesting the possibility that their functions might not be entirely the same.
Results from luciferase assays of LUC-R1 suggested that the 5′ and 3′ NCRs of RNA1 are involved in translation. This is not surprising since RNA1 encodes the replicase needed for viral replication. By contrast, similar regions from RNA2 alone, including those upstream of the YSS-containing 3′ NCR, did not support translation in the R2 series of luciferase replicons (Fig. 3b). On this basis, it is likely that in LCV RNA2 alone, the YSS may not function in translation enhancement, unlike for the 3′ cap-independent translational enhancers of viruses that do not have a cap structure at the 5′ end of their genome 30 .
In Brome mosaic virus (BMV), TMV and other (+ )-RNA viruses 14,33-36 , the last three nts, CC(A/G), in the gRNA(s) are essential for (− )-RNA synthesis. In contrast for LCV, a deletion of as many as 24 nts from the 3′ terminus of RNA2 can be tolerated, although as more nts are deleted, the less efficient viral RNA synthesis becomes (Fig. 4). Based on mfold predictions, deleting the nts in 3′ Δ 11 does not affect the overall structure of the YSS, while deleting those in 3′ Δ 24 eliminates the right basal stem of the YSS and disrupts its overall structure but leaves both SL1 and SL2 intact (not shown). This suggests that the YSS tolerates a disruption of the basal stem (S3), but cannot tolerate any SL disruptions. Disruptions of either SLs (whether they are large deletions: 3′ Δ 38, 3′ Δ 48, and 5′ Δ 50; targeted deletions: Δ SL1 and Δ SL2; or disruptive substitutions: SLD1-1 and SLD1-2) abolished viral RNA synthesis (Figs 4 and 5). These results suggest that both SLs are needed for viral RNA synthesis.
In addition, SHAPE analysis shows that the deletion of one SL eliminates the YSS, but does not disrupt the overall structure of the other SL (Fig. 2). This suggests that SL1 does not serve as a structural support for SL2 and vice versa. Furthermore, deletion of either SL does not appear to affect the overall stability of the RNA ( Supplementary  Fig. S2). Because the reciprocal exchange of YSS between RNA2 and RNA1 did not affect viral RNA synthesis of the resulting chimeric RNAs (Fig. 6), this suggests that the YSS of RNA1 most likely also participates in viral RNA synthesis. This seems possible given the structural similarity of the YSS of R1 and R2 despite a 22.3% difference in Scientific RepoRts | 6:34482 | DOI: 10.1038/srep34482 sequence identity 12 . One issue to be addressed in the future is whether the YSS of RNA2 has any role in mediating the messenger activity of RNA1, and if the YSS of RNA1 can confer messenger activity on RNA2.
The experiments in which protoplasts were inoculated with LCV RNA1 and the transcripts of each of the LCV RNA2-based F-Luc constructs were aimed at addressing the lack of in vivo translation activity in protoplasts inoculated with the latter alone (Fig. 3b). Prior to these experiments, it was presumed that cap-dependent translation of LCV RNA2 alone was responsible for the expression of P5.6, the first ORF. However, the association of replication and translation for specific R2 replicons co-inoculated with LCV RNA1 (Fig. 3c) demonstrates that the expression strategy is probably more complex than is originally anticipated. This is a novel finding for criniviruses and has led to several new insights. First, replicon RNA synthesis for LUC-R2C and its Δ SL derivatives, i.e. LUC-R2CΔ SL1 and LUC-R2CΔ SL2, is not dependent on P5.6; even with its coding sequence having been replaced by that of F-Luc (Fig. 3a), RNA synthesis (and luciferase activity) was observed for these replicons (Fig. 3c). For many viruses in the alphavirus-like supergroup, proteins that are non-essential for replication are translated after viral replication has initiated 20,37-39 . P5.6 may well be expressed after the (− )-RNA synthesis of LCV RNA2, although the possibility that it is expressed at the same time as, or even before, viral RNA synthesis cannot be completely excluded. It remains possible that the YSS of RNA2 may play a dual role in the LCV infection cycle -that of supporting viral RNA synthesis while also contributing to the translation of P5.6 in the presence of LCV RNA1. However, because P5.6 is not required for replicon RNA synthesis, this lowers the likelihood that the reduction or loss of viral RNA synthesis in the 3′ NCR and/or YSS mutants of LCV RNA2 (Figs 4 and 5) was due to defective P5.6 expression. This means that the role of the YSS in translating P5.6, if at all exists, is possibly independent of that in mediating viral RNA synthesis. Altogether, the above insights lend support to the notion that the YSS of RNA2 contributes to viral RNA synthesis. Targeted mutations engineered in the P5.6 coding sequence of LCV RNA2 will facilitate future work aimed at determining whether it has any role in viral RNA synthesis. Second, RNA synthesis was only efficient for constructs engineered with the YSS-containing 3′ NCR and its 302-nt upstream region (LUC-R2Cand LUC-R2E [ Fig. 3c]). However, even with this upstream region present, RNA synthesis was drastically reduced when either of the YSS SLs was deleted (i.e. LUC-R2CΔ SL1 and LUC-R2CΔ SL2 [ Fig. 3c]). The impairment of RNA synthesis for LUC-R2CΔ SL1 and LUC-R2CΔ SL2 was similar to and consistent with the loss of viral RNA synthesis in the Δ SL1 and Δ SL2 mutants of LCV RNA2 (Fig. 5b,c). Together, this indicates that the YSS serves as an RNA synthesis enhancer in the context of replicon replication, and possibly also in LCV RNA2. Third, since the P4.8 ORF and its potential upstream regulatory region are incorporated in LUC-R2C (and its Δ SL derivatives) and in LUC-R2E, P4.8 is probably produced. However, genome expression analysis suggested that it is translated from a sgRNA transcribed from the (− )-RNA template 12 i.e. P4.8 is produced downstream of (− )-RNA production; therefore, it is not likely to be involved in (− )-RNA synthesis. Fourth, translation activity corresponded with the replication efficiency of all but one replicon: LUC-R2E, which replicated efficiently but was reduced in translation activity by 2.5 fold relative to LUC-R2C. Thus, it seems that the additional P5.6 coding sequences can have a regulatory effect on translation without interfering with the RNA synthesis enhancer activity of the YSS. The basis for the reduction in translation activity for LUC-R2E is unclear but not unprecedented. Luciferase activity for constructs incorporated with SCV non-coding sequences was reduced when additional nts from the 5′ region of the P26 ORF were incorporated into the 5′ sequence flanking the F-Luc gene 27 .
The simplest explanation for the association between replication and translation involving LCV RNA1 and the R2 replicons is that the YSS facilitates the replicase produced from RNA1 5,10,40 to synthesize (− )-RNA, which serves as the template for the production of (+ )-RNA from which F-Luc (or P5.6, in the context of LCV RNA2) is expressed. It is possible that the accumulated production of (+ )-RNA is needed for the increased translation of a weakly translating message. Our results are also reminiscent of the coupling between replication and translation reported for a number of (+ )-RNA viruses with single-component genomes and also for those with segmented genomes [41][42][43][44][45] . For example, in Flock house virus, a segmented-genome virus also with two gRNAs, cap-dependent translation of a functional CP requires the replication of the encoding RNA (RNA2) 44 . In the case of the bipartite Red clover necrotic mosaic virus, cap-independent translation of RNA2, which encodes the movement protein, is linked to its replication in trans by the RNA1-encoded replicase 45 . The question remains, how is P5.6 expressed? One possibility is that it can be translated from a sgRNA produced by the (− )-RNA template; a similar strategy has been described for the cap-dependent translation of the coat protein (CP) encoded by the first ORF in RNA2 of Tobacco rattle virus 46 . This hypothesis is attractive as it explains the lack of messenger activity in the R2 series of replicons in the absence of LCV RNA1 (Fig. 3b).
A comparison of mfold-predicted models generated using the full-length gRNA sequences of criniviruses revealed that structures similar to the YSS are strikingly conserved among other crinivirus genomes ( Supplementary Fig. S3). What makes the formation of the YSS possible in these genomes is the conservation of the 3′ NCR in both gRNAs of each crinivirus 8 . An exception to this pattern is found in LIYV, which contains an equivalent of the YSS in the 3′ NCR of RNA1 but not in that of RNA2 largely due to the low (< 31%) sequence identity between the two NCRs 31 . On the basis of its position in the crinivirus genomes and the results from our study, it seems likely that the YSS of criniviruses is involved in regulating some common functions associated with viral RNA synthesis. Our data also suggest a biological significance of the YSS in facilitating the similar temporal accumulation of LCV RNAs 1 and 2, where both RNAs were previously found to accumulate in tobacco protoplasts within approximately 12 hpi 5,12 . Since LIYV RNA2 lacks the predicted YSS, viral RNA synthesis may require additional enhancements. Indeed, a previous study identified the LIYV RNA1-encoded P34 to be a trans enhancer required for the efficient accumulation of LIYV RNA2 10,32 . Thus, the structural differences in the 3′ NCR of both LIYV RNAs and requirement of P34 for efficient LIYV RNA2 accumulation may be a basis underpinning the asynchronous RNA replication-accumulation kinetics of LIYV, where the accumulation of RNA2 is delayed by 24 hrs relative to that of RNA1 during the initial stages of infection prior to the production of P34 10 , and this is clearly different from what we see in LCV. Further studies are needed to delineate the regions/nts of the YSS involved in viral RNA synthesis enhancement, and also to investigate the mechanisms underlying the replication-associated translation of replicons, including issues on whether translation is coupled to de novo replicon replication, the possible involvement of trans-activating viral protein(s) or even trans-activator structural element(s) present in LCV RNA1 45,47,48 .

Methods
Constructs. Structure cassette plasmids used for SHAPE analysis were engineered with the full-length cDNA sequences of LCV RNA1 (pS-CM1), LCV RNA2 (pS-CM2), LCV RNA2 without SL1 (pS-Δ SL1) and LCV RNA2 without SL2 (pS-Δ SL2). The construction of these plasmids involved using two independent PCR-amplifications. Sequence information of all oligo-primers used for the PCRs is provided in Supplementary Table S1. The first amplification was performed using one of the following oligo-primer pairs and DNA templates: LCV-91-CN/ LCV-311-CM and template pCM1, the WT clone of LCV RNA1 (for constructing pS-CM1), LCV-161-CM/ LCV-288-CM and template pCM2, the WT clone of LCV RNA2 (for constructing pS-CM2) or pΔ SL2 (for constructing pS-Δ SL2), and LCV-161-CM/LCV-289-CM and template pΔ SL1 (for constructing pS-Δ SL1). The resulting products were gel purified and subsequently used for the second amplification in which further modifications were achieved using the reverse oligo-primer LCV-285-CM and one of the following two forward oligo-primers: LCV-91CN (for construction of pS-CM1) and LCV-161-CM (for construction pS-CM2, pΔ SL1, and pΔ SL2). The amplified products were gel-purified, adenylated, and cloned into the pGEM-T Easy vector. Restriction digestion using HpaI and NdeI (for pS-CM1) or AatII and NgoMIV (for pS-CM2, pΔ SL1, and pΔ SL2) was performed to release the DNA fragments from the resulting recombinant pGEM-T Easy vectors, and the released DNA fragments were subcloned into similarly digested pCM1 (for making pS-CM1) or pCM2 (for making pS-CM2, pΔ SL1, and pΔ SL2), resulting in the respective final products.
F-Luc reporter constructs for translation analyses were engineered using pCM1, pCM2 and TMV30BGFP 49 as templates. Constructs for the determination of viral RNA synthesis were engineered using pCM1 and/or pCM2 as templates 5 . All constructs were engineered using PCR-mediated approaches using the oligo-primers listed in Supplementary Table S1. Details of the making of individual constructs are given in Supplementary Note S1.
All PCR-amplifications were performed using Herculase II fusion DNA polymerase with high fidelity proofreading capability (Agilent Technologies), and all cloned products derived from PCR-amplification were sequenced in both directions. SHAPE analysis. 5 μ g each of pS-CM1, pS-CM2, pS-Δ SL1 or pS-Δ SL2 were linearized by restriction enzyme digestion using NgoMIV. The linearized DNA served as a template for the in vitro synthesis of RNA transcripts using the T3 mMessage mMACHINE kit (Life Technologies). 11 μ l of dH 2 O containing 2 pmol of the RNA transcript was heated at 95 °C for 2 min, and immediately placed on ice for 1 min. 6 μ l of 3.3x RNA folding mix and 1 μ l of 100 mM MgCl 2 was added to the transcript (100 mM HEPES, pH 8.0, 5 mM MgCl 2 , 100 mM NaCl in the final volume), and the mixture was divided into two 9 μ l portions and incubated at 37 °C for 10 min to allow the RNA to renature 50 . The first portion of the renatured transcript was treated with 1 μ l of 400 mM benzoyl cyanide (BzCN) (Sigma-Aldrich) (40 mM final concentration) made freshly in DMSO at 37 °C for 15 min. The second portion of the renatured transcript (the control treatment) was treated with 1 μ l of DMSO without BzCN under the same conditions. RNA was recovered by ethanol precipitation and resuspended in 10 μ l of 0.5x TE (5 mM Tris, 0.5 mM EDTA, pH 8) 50 . Primer extension reactions were according to the methods of McGinnis et al. 51 . Briefly, BzCN-and DMSO (control)-treated RNA was subjected to primer extension by SuperScript III ® reverse transcriptase (Invitrogen) under three consecutive sets of temperature and time conditions: 42 °C for 1 min, 52 °C for 25 min, 65 °C for 5 min using 5′ end-labeled-VIC and -NED 3′ -linker primers [5′ -GAACCGGACCGAAGCCCGATTTGC-3′ , nts complementary to the RT primer binding site of the structure cassette], respectively. Note that the structure cassettes contain a linker region that folds into stable SL structures at the 3′ termini of the transcript upon synthesis. The 3′ linker is designed for the binding of the 5′ end-labeled-VIC and -NED 3′ -linker primers and reverse transcriptase (Fig. 2) 52,53 . Two separate sequencing reactions, each containing 2 pmol of in vitro transcripts produced from the respective structural plasmids, pS-CM1, pS-CM2, pS-3′ Δ SL1, or pS-3′ Δ SL2, were performed under the same conditions as with the primer extension reaction, except that 1 μ l of 5 mM ddTTP was also included in each reaction. In addition, one of the sequencing reactions contained the VIC 3′ -linker primer while the other contained the NED 3′ -linker primer. The primer extension and sequencing reactions were quenched and combined in the following manner: the BzCN-treated sample with the sequencing sample containing the NED 3′ -linker primer, and the DMSO (control)-treated sample with the sequencing sample containing the VIC 3′ -linker primer. Following ethanol precipitation, cDNAs were resuspended in 10 μ l of deionized formamide and then resolved by capillary electrophoresis. The electropherogram data of each reaction was processed and analyzed by the Qushape software 52 . The reactivity of nts to BzCN modification was classified as: unreactive (0-0.4), moderately reactive (0.4-0.85), or highly reactive (> 0.85) 52 . The processed SHAPE reactivities were incorporated as pseudo free energy constraints in RNAstructure (version 5.4) 54 with slope and intercept value of 2.6 and − 0.8, respectively, according to Low and Weeks 55 . In addition, base pairing between nucleotides greater than 600 positions was disallowed 55 .
Luciferase reporter assays. F-Luc reporter constructs (5 μ g) were linearized by restriction enzyme digestion using NgoMIV or KpnI (in the case of LUC-TMV). The linearized DNA then served as a template for the in vitro synthesis of capped RNA transcripts using the T3 mMessage mMACHINE kit (Life Technologies). The Renilla luciferase (R-Luc) construct was linearized by restriction enzyme digestion using SmaI, and capped RNA transcripts were synthesized using the T7 mMessage mMACHINE kit 56 . To perform in vivo translation, 3 pmol of the in vitro produced capped transcript of each F-Luc reporter construct along with 1 pmol of the in vitro synthesized capped R-Luc RNA transcripts were inoculated to half a million N. tabacum var. Xanthi protoplasts. In parallel, 1 μ l of water was inoculated along with the R-Luc RNA transcripts as a negative control. Additionally, for inoculations that included LCV RNA 1, 2 μ g (0.7 pmol) of in vitro transcripts synthesized using NgoMIV-linearized pCM1 were used. After 72 hr incubation at 26 °C, cells were harvested and processed using the Dual-Luciferase Reporter Assay System (Promega). Procedures for in vitro translation are provided in Supplementary Method S1. Luciferase activity was measured using the Turner Biosystems 20/20n Luminometer (Promega). In vivo translation efficiency was determined by the taking the ratio of luminescence produced by F-Luc to that produced by the internal control R-Luc. Both the in vivo (Fig. 3b) and in vitro (Supplementary Fig. S1) experiments involving the R2 series of luciferase constructs were repeated three times and triplicates of each sample/inoculation were tested in each experiment. All in vivo translation experiments involving the R2 series of luciferase constructs and LCV RNA 1 (Fig. 3c) were repeated three times, with duplicate samples for each treatment (see the next section). Scientific graphing and statistical analyses, including two-tailed Student's t-tests, were performed using the GraphPad Prism 5 software (GraphPad Software, Inc., San Diego, CA).
In vitro transcription, protoplast inoculation, total RNA extraction, and Northern blot analysis.
2 μ g of the in vitro synthesized transcripts of WT RNAs 1 and 2 (or their engineered derivatives) were each inoculated to N. tabacum var. Xanthi protoplasts following the previously described procedure 5,57 . To normalize any differences arising from potential uneven handling and inoculation of protoplasts, the following procedure was adopted for inoculations using the WT and the mutants p3′ Δ 4, p3′ Δ 11, p3′ Δ 24, pR1-3′ R2, pR2-3′ R1, pSLD1-2, pSLR1, and pΔ SL2: triplicates of the each combination of in vitro transcripts were individually inoculated to 0.5 × 10 6 protoplasts, and the protoplasts from all three inoculations were combined into a 100 × 20 mm petri dish and incubated at 26 °C. 16 hours post-inoculation (hpi), the combined protoplasts were redistributed into three 60 × 20 mm petri dishes (10 ml per dish) and incubated at 26 °C until the first harvest at 24 hpi. The inoculated protoplasts were also harvested at 48 and 96 hpi. A similar approach was adopted for the in vivo translation experiments involving the R2 series of luciferase constructs and LCV RNA1 (Fig. 3c). Duplicates of each inoculated protoplast sample (10 ml each) were combined at 72 hpi and re-distributed into two 10 ml portions; one portion was processed for measuring F-Luc/R-Luc activity and the other portion was saved for Northern Blot analysis. Total RNA was extracted by the TRIzol ® method (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. Approximately 2 μ g of total RNA from each sample were analyzed by Northern Blot using DIG-labeled riboprobes II and VIII as previously described 12 (Fig. 1a). Signals of viral RNA accumulation on X-ray films were estimated by densitometry using the Scion Image software (Scion Corp) as previously described 12 , or by using the histogram function in Photoshop Element (Adobe Systems) to compute the intensity value and standard deviation using the mean and pixel values of a region of interest in each signal. DIG-labeled riboprobes specific to (+ )-or (− )-sense F-Luc RNA were used to determine RNA synthesis for the LCV RNA2 (R2) series of replicons (Fig. 3c). To generate these riboprobes, a recombinant plasmid, pFLuc, containing the cDNA corresponding to a specific location in the F-Luc coding sequence (not present in that of R-Luc) was first obtained. Specifically, using luciferase reporter construct LUC-R2A(− ) as template, a 425 nt region specific to F-Luc was amplified using oligonucleotide primer set LUC-005-JZ and LUC-006-JZ (Supplementary Table S1). The PCR amplified product was gel purified and cloned into the pGEM-T Easy vector, resulting in pFLuc. R2 (− )-sense replicons were detected using the (+ )-sense F-Luc DIG-labeled riboprobe, which was generated by linearizing pFLuc with SpeI and in vitro transcribed with T7 RNA polymerase (Roche Applied Science). R2 (+ )sense replicons were detected using the (− )-sense F-Luc DIG-labeled riboprobe, which was generated by linearizing pFLuc with NcoI and in vitro transcribed with Sp6 RNA polymerase (Roche Applied Science).
Stability assays for input LCV RNA2 transcripts in tobacco protoplasts. 2 μ g of the in vitro synthesized transcripts of pCM2, pΔ SL1 and pΔ SL2 were each inoculated to N. tabacum var. Xanthi protoplasts following the procedures as described above, including the normalization step, except the inoculated protoplasts were redistributed at 3 hpi and harvested at 3, 12, and 18 hpi. Approximately 5 μ g of the TRIzol ® -extracted total RNA from each sample were subjected to DNase treatment (DNA-free ™ Kit; Life Technologies). The treated RNA was run on 1% HEPES denaturing gel stained with ethidium bromide to quantitate and normalize the 18S rRNA for equal amounts of rRNA that will be used for cDNA synthesis. 2.5 μ g of treated RNA was used to synthesize the first-strand cDNA using Superscript III reverse transcriptase (Invitrogen) and gene-specific oligo-primer LCV-99-AC (Supplementary Table S1) according to the manufacturer's instruction. 4 μ l of the amplified cDNAs was then used for PCR with 2.5 μ l of 10X Taq buffer, 2.5 μ l of 25 mM MgCl 2 , 1 μ l of 10 mM dNTPs, 0.5 μ l of Taq DNA polymerase, and 10 μ M of oligo-primers LCV-70-PW and LCV-99-AC (Supplementary Table S1). The following PCR conditions were used: 94 °C for 2 min, followed by different number of cycles (10,15,20,25, and 30 cycles) of 94 °C for 45 sec, 54.6 °C for 45 sec, and 72 °C for 1 min, with a final extension at 72 °C for 10 min. For internal control, the same procedure was used to amplify the 18S rRNA as described above except oligo-primer NtUbiR (Supplementary Table S1) was used for reverse transcription, and primers NtUbiF and NtUbiR (Supplementary  Table S1) were used for PCR-amplification.