RNA Structure Protects the 5′ End of an Uncapped Tombusvirus RNA Genome from Xrn Digestion

ABSTRACT

One of the many challenges faced by RNA viruses is the maintenance of their genomes during infections of host cells. Members of the family Tombusviridae are plus-strand RNA viruses with unmodified triphosphorylated genomic 5′ termini. The tombusvirus Carnation Italian ringspot virus was used to investigate how it protects its RNA genome from attack by 5′-end-targeting degradation enzymes. In vivo and in vitro assays were employed to determine the role of genomic RNA structure in conferring protection from the 5′-to-3′ exoribonuclease Xrn. The results revealed that (i) the CIRV RNA genome is more resistant to Xrn than its sg mRNAs, (ii) the genomic 5′-untranslated region (UTR) folds into a compact RNA structure that effectively and independently prevents Xrn access, (iii) the RNA structure limiting 5′ access is formed by secondary and tertiary interactions that function cooperatively, (iv) the structure is also able to block access of RNA pyrophosphohydrolase to the genomic 5′ terminus, and (v) the RNA structure does not stall an actively digesting Xrn. Based on its proficiency at impeding Xrn 5′ access, we have termed this 5′-terminal structure an Xrn-evading RNA, or xeRNA. These and other findings demonstrate that the 5′UTR of the CIRV RNA genome folds into a complex structural conformation that helps to protect its unmodified 5′ terminus from enzymatic decay during infections.

IMPORTANCE The plus-strand RNA genomes of plant viruses in the large family Tombusviridae are not 5′ capped. Here, we explored how a species in the type genus Tombusvirus protects its genomic 5′ end from cellular nuclease attack. Our results revealed that the 5′-terminal sequence of the CIRV genome folds into a complex RNA structure that limits access of the 5′-to-3′ exoribonuclease Xrn, thereby protecting it from processive degradation. The RNA conformation also impeded access of RNA pyrophosphohydrolase, which converts 5′-triphosphorylated RNA termini into 5′-monophosphorylated forms, the preferred substrate for Xrn. This study represents the first report of a higher-order RNA structure in an RNA plant virus genome independently conferring resistance to 5′-end-attacking cellular enzymes.

INTRODUCTION

RNA viruses face the challenge of protecting their genomes from degradation during infection of host cells (1, 2). For plant RNA viruses, in addition to suppressing the host antiviral gene silencing system (3), they must also avoid canonical decay pathways designed to remove unwanted and defective cellular RNAs (4–6). Viruses with plus-strand RNA genomes are susceptible to exoribonucleases at both their 5′ and 3′ ends, and those that are uncapped and/or nonpolyadenylated are especially vulnerable. Viral genomes with unmodified 5′ ends are substrates for enzymes that digest RNA in the 5′-to-3′ direction. Xrn1 is a 5′-to-3′ exoribonuclease that functions in the cytosol of yeast and metazoans (7, 8). In plants, Xrn4 is the functional homolog of Xrn1, and both of these enzymes prefer RNA substrates with a monophosphorylated 5′-terminal nucleotide (9, 10). The termini of 5′-capped mRNAs fated for degradation become substrates for Xrn1/4 by decapping activities (2). For uncapped RNAs, RNA pyrophosphohydrolases (RppH) remove terminal γ- and β-phosphate groups from 5′-triphosphates, generating α-monophosphates (11). To initiate digestion of a 5′-monophosphorylated RNA substrate, Xrn1 requires two to three unpaired 5′-terminal nucleotides (12, 13). The enzyme’s activity is highly processive, yielding 5′-monophosphate nucleotides; however, its progression can be impeded by certain RNA structures (14, 15).

A rapidly growing field of research has emerged that focuses on viral RNA structures that can stall actively digesting Xrn, referred to as Xrn-resistant RNAs (xrRNAs). By folding into higher-order structures in viral RNAs, xrRNAs are able to resist Xrn degradation through specific RNA-mediated contacts with the enzyme (16, 17). This inhibition can generate functional degradation intermediates, such as subgenomic flavivirus RNAs and other small virus-derived RNAs, which correspond roughly to the 3′UTRs of viral genomes or subgenomic mRNAs (18–22). Conversely, large functionally relevant xrRNAs, generated by Xrn1 termination within genomic 5′UTRs, accumulate in both hepatitis C virus and bovine viral diarrhea virus infections (23). These diverse virus-derived RNA degradation products are able to inhibit Xrn activity and/or modulate cellular processes to facilitate infections (18–23).

Carnation Italian ringspot virus (CIRV) is a plant virus in the genus Tombusvirus (family Tombusviridae). Like all tombusvirids, its plus-strand RNA genome contains no 5′-cap or 3′-poly(A) tail, making its genomic termini particularly susceptible to exoribonucleases. The polycistronic CIRV genome contains five open reading frames (ORFs) that encode an assortment of viral proteins (Fig. 1). The 5′-proximal ORFs encode two proteins that are translated directly from the genome and are essential for viral RNA synthesis: the p36 accessory replication protein and its readthrough product, p95, the RNA-dependent RNA polymerase (24). Three viral proteins are encoded further downstream that are translated from two smaller subgenomic (sg) mRNAs transcribed during infections (Fig. 1). The coat protein (p41) is translated from sg mRNA1, and the movement (p22) and suppressor of gene silencing (p19) proteins are translated from sg mRNA2. During infections, CIRV translates its proteins by utilizing long-range RNA-RNA interactions between the 5′UTRs of its three species of viral mRNAs (i.e., genome and two sg mRNAs) and a 3′-cap-independent translational enhancer (3′CITE) element located within their 3′UTRs (Fig. 1) (25).

FIG 1.

Schematic representation of CIRV genomic and subgenomic mRNAs. Viral ORFs are displayed as gray boxes. The location of the 3′-cap-independent translation enhancer (3′CITE) in the 3′UTR is indicated by a bracket. sg mRNA promoters are shown below the genome as black arrows. Approximate nucleotide lengths are shown, in kilobases, on the right.

Currently, it is not known how the 5′-triphosphorylated RNA genome and sg mRNAs of tombusviruses resist attack from plant 5′-to-3′ exoribonucleases such as plant Xrn4. Using CIRV, we have investigated the role of viral 5′-terminal RNA sequences in conferring protection from Xrn activity, focusing on the genomic 5′UTR. RNA-mediated protection of 5′-terminal triphosphates from RppH activity was also examined. Our results reveal a major role for higher-order RNA structure in the 5′UTR of the CIRV genome in blocking 5′-end access to these enzymes.

RESULTS

Susceptibility of uncapped tombusvirid RNA genomes to Xrn1.

Four different tombusvirid RNA genomes possessing varying degrees of higher-order RNA structure in their 5′UTRs were selected for analysis. The RNA structures in the 5′UTRs of closely related tombusviruses and aureusviruses, such as CIRV and Cucumber leaf spot virus (CLSV), respectively, have been characterized extensively (26–29). These 5′UTRs begin with a T-shaped domain (TSD) followed by a downstream domain (DSD), which are separated by stem-loop-5 (SL5) (Fig. 2A). Also present is a mid-range interaction involving sequences in the loop of SL4 and the DSD that forms a pseudoknot (PK-TD1). The 5′-terminal sequences of the RNA genomes in both these genera are base-paired in S1, which is positioned adjacent to SL5, providing the potential for coaxial stacking (Fig. 2A). In contrast, the 5′ ends of other tombusvirids, such as Tobacco necrosis virus (TNV-D, genus Betanecrovirus) and Turnip crinkle virus (TCV, genus Betacarmovirus), are predicted to exist in small or weak terminal hairpins, respectively (Fig. 2B). Each of these four full-length tombusvirid RNA genomes were in vitro transcribed under conditions that favored incorporation of a monophosphorylated 5′-terminal nucleotide, the optimal 5′-phosphorylation state for Xrn1. The viral RNAs generated were then treated with purified yeast Xrn1. Xrn1 has been utilized previously as an effective surrogate for plant Xrn4 (21). All genomes showed some susceptibility to Xrn1 degradation, with remaining full-length levels ranging from ∼87% to ∼25% (Fig. 2C). Notably, the degree of resistance observed correlated roughly with the predicted higher-order stability of the 5′UTRs, suggesting a role for RNA structure in protecting these uncapped genomic 5′ ends from Xrn1 activity.

FIG 2.

In vitro Xrn1 treatment of tombusvirid genomic RNAs. (A) RNA secondary structures of CIRV (tombusvirus) and CLSV (aureusvirus) 5′UTRs. The TSD and DSD are delineated by brackets, and the sequences involved in the PK-TD1 pseudoknot interaction are highlighted in gray. The position of the upstream AUG (uAUG) in CIRV is highlighted in green. S1, SL4, and SL5 are labeled with yellow highlighting, while the p33 and p36 start codons for CIRV and CLSV, respectively, are highlighted in light green. (B) Predicted RNA secondary structures of betanecrovirus and betacarmovirus 5′UTRs. The locations of respective p22 and p28 start codons are highlighted in light green. (C) Ethidium bromide-stained agarose gel of products from in vitro Xrn1 assay of tombusvirid genomes, with the percentage of input RNA remaining shown below with standard errors of the means (SEM) (n = 3).

Susceptibility of CIRV sg mRNAs to Xrn1.

The tombusvirus CIRV was selected for further characterization, beginning with an assessment of the sensitivity of its two cognate sg mRNAs to Xrn1. The 5′ ends of both sg mRNAs are not predicted to have significant local RNA structure, although a small terminal hairpin is possible in sg mRNA2 (Fig. 3A). Treatment of transcripts of the two 5′-monophosphorylated sg mRNAs revealed that both were more sensitive to Xrn1 than their genomic counterpart (Fig. 3B). Interestingly, sg mRNA1, with the predicted linear 5′ end, was more stable than sg mRNA2. One possibility for this observation was that the 5′-terminal sequence of sg mRNA1 (Fig. 3A, red box) was paired with its partner sequence in the 3′CITE (Fig. 3A). This prospect was tested by creating a 3′-truncated sg mRNA1, Δ3C, that lacked the 3′CITE (Fig. 3C). Compared to wild-type (WT) sg mRNA1, treatment of the Δ3C RNA with Xrn1 showed moderately enhanced resistance (Fig. 3D). Thus, Xrn resistance of the 5′ end of sg mRNA1 is not dependent on an interaction with the 3′CITE.

FIG 3.

In vitro Xrn1 treatment of CIRV sg mRNAs. (A) RNA secondary structures of 5′UTRs of CIRV genome and subgenomic mRNAs. The 5′ partner sequences for the long-range RNA-RNA interaction with the 3′CITE are boxed in red. The start codon for each respective open reading frame is highlighted in light green. (B) Ethidium bromide-stained agarose gel of in vitro Xrn1 assay of CIRV subgenomic mRNAs. The percentage of input RNA remaining is shown below with SEM (n = 3). (C) Deletion of terminal region containing 3′CITE from sg mRNA1 in mutant Δ3C. The long-range RNA-RNA interaction between the 5′UTR and the 3′CITE in WT sg mRNA1 is shown as a dotted red arrow. (D) In vitro Xrn1 treatment of mutant Δ3C. Percentage of input RNA remaining is shown below with SEM (n = 3).

The CIRV 5′UTR alone blocks Xrn1 access.

In most tombusviruses, the 5′UTR is approximately 140 to 160 nucleotides (nt) in length and is positioned immediately upstream of the start codon for a 33-kDa accessory replication protein, p33. In the CIRV genome, in addition to an AUG at position 144 (that would produce an ∼33-kDa protein), an in-frame upstream AUG (uAUG) start codon is located at genome coordinate 78 (Fig. 4A, top), positioned within the 3′ half of SL5 (Fig. 4B, left). Translation initiated at this uAUG generates an N-terminally extended accessory protein and RdRp with molecular masses of 36 kDa and 95 kDa, referred to as p36 and p95; however, the N-terminal extensions are dispensable for viral replication (30). Therefore, to match the size and structure of the CIRV 5′UTR to those of other tombusviruses and to allow for modifications of the downstream structures (i.e., SL5 and PK-TD1) without altering any underlying protein coding, the uAUG in the CIRV genome was inactivated via nucleotide substitutions. The changes made were designed to convert the uAUG to AUC while maintaining pairing in SL5 with a corresponding C-to-G nucleotide exchange in the partner strand (Fig. 4B, right). The resulting genome, termed CIRV*, contained a 143-nt-long 5′UTR, akin to those in other tombusviruses, with corresponding p33 and p92 products translated from AUG¹⁴⁴ (Fig. 4A). Importantly, the modifications made in the stem of SL5 in CIRV* had negligible impact on Xrn1 resistance compared to wild-type CIRV (Fig. 4C).

FIG 4.

In vitro Xrn1 treatment of CIRV, CIRV*, DI-7, and DI-7*. (A) Schematic representation of CIRV, CIRV*, DI-7*, and its deletion derivatives. Open reading frames are represented by gray boxes. The relative location of the uAUG is shown at the top, along with its inactivating mutation in CIRV* and DI-7*, shown as a green asterisk. The genomic segments present in DI-7* are shown directly below the genome as thick black bars, and the corresponding regions are labeled below in white boxes. The location of the 5′UTR (i.e., RI) in DI-7* is highlighted in yellow. (B) Location of the uAUG in SL5 (green highlight) and the substitutions made (green nucleotides) to inactivate the uAUG are shown on the right in CIRV*. (C) In vitro Xrn1 treatment of CIRV* and DI-7* versus their WT counterparts. Percentage of input RNA remaining is shown with SEM (n = 3). (D) In vitro Xrn1 treatment of DI-7* and its truncations. Percentage of input RNA remaining is shown with SEM (n = 3).

Similarly, defective interfering RNA-7 (DI-7), a viral RNA replicon composed of three discontinuous segments of the CIRV genome (Fig. 4A, lower) and its SL5-modified version, DI-7*, exhibited equivalent levels of susceptibility to Xrn1, which were also comparable to those of their full-length genome counterparts (Fig. 4C). DI-7* was constructed and tested along with CIRV* in preparation for its use as a surrogate substrate for CIRV* in later experiments assessing 5′ resistance. Because region-I (RI) of DI-7* corresponded to the 143-nt-long genomic 5′UTR and adjacent AUG¹⁴⁴, the replicon also provided a convenient way to assess whether the 5′UTR on its own was capable of blocking Xrn1 access. When viral segments downstream of RI were removed sequentially and tested, the results revealed that the 5′UTR (RI*) is highly effective at independently inhibiting Xrn1 access (Fig. 4D). Thus, the genomic 5′UTR, in the absence of other viral or cellular proteins, is an efficient blocker of Xrn entry.

The CIRV 5′UTR is not an xrRNA.

The structural conformation adopted by the 5′UTR can effectively block Xrn1 from accessing 5′-terminal nucleotides (Fig. 4D). To test whether the 5′UTR is also a proficient xrRNA, that is, can efficiently stall an actively digesting Xrn1, different-sized 5′-monophosphorylated leader sequences were attached to the 5′ end of the CIRV genome to provide single-stranded access (Fig. 5A). The A-tract leader sequences appended were not predicted to interfere with 5′UTR folding based on secondary structure predictions (data not shown) (31). Compared to CIRV* (here termed WT), genomes containing a 5′ extension were considerably more susceptible to Xrn1 digestion, indicating that the 5′UTR is not an equally effective xrRNA (Fig. 5B). Digestion of the mutants genomes did, however, lead to the accumulation of an intermediate degradation product (Fig. 5B, arrow), suggesting that there are downstream segments of the genome that possess more notable xrRNA activity.

FIG 5.

In vitro Xrn1 treatment of CIRV* genomes with and without appended poly(A) leaders. (A) RNA secondary structure diagram of the 5′UTR, with leader sequences shown below in red. (B) In vitro Xrn1 treatment of CIRV*-based genomes with poly(A) leader sequences. An unknown intermediate degradation product is indicated with an arrow. Percentages of input RNA remaining and relative levels of RNA remaining compared to the wild type (set at 100) are shown with SEM (n = 3). (C) In vitro Xrn1 treatment of 5′UTRs with (30app-RI*) and without (RI*) poly(A) leader sequences. Reaction products were separated in a denaturing 10% polyacrylamide gel and stained with ethidium bromide. An arrowhead on the right points to a product produced by Xrn1 stalling just upstream of the 5′UTR. A representative gel is shown (n = 3).

Due to the large size of the viral genomes tested, it could not be determined if the bands corresponding to the remaining full-length genome contained or lacked the A-tracts (Fig. 5B). Accordingly, the 30-nt-long A-tract was 5′ appended to the 5′UTR only, creating 30app-RI*, which would allow for size differentiation between substrates containing the A-tract and Xrn1 products lacking it. Transcripts of RI* (WT 5′UTR) and 30app-RI* (WT 5′UTR with a 30A 5′ appendage) along with respective counterparts with a destabilized stem-1, R1*-S1b and 30app-RI*-S1b (Fig. 5A), were tested in an Xrn1 assay. As observed previously (Fig. 4D), RI* showed a high level of resistance to Xrn1, while RI*-S1b, with its destabilized S1, was notably susceptible (Fig. 5C). The addition of 30A-tracts to the WT and S1-destabilized 5′UTRs in 30app-RI* and 30app-RI*-S1b, respectively, resulted in elevated susceptibility of both to Xrn1 and the presence of a very faint band that migrated at the position of 5′UTR-only transcripts (i.e., RI* and RI*-S1b) (Fig. 5C, arrowhead). Although these minor bands likely represent very inefficient Xrn1 stalling in the A-tract just upstream of the 5′UTR structure, the generation of the band seen in 30app-RI*-S1b containing a destabilized S1, which in RI*-S1b greatly enhanced Xrn1 susceptibility, argues against the slight stalls being dependent on the previously defined higher-order RNA structure of the 5′UTR (26–28). Accordingly, these results indicate that the 5′UTR is not an xrRNA.

RNA secondary structures contribute to Xrn1 and RppH blocking activity.

Having established the 5′-entry-blocking activity of the 5′UTR, the next step was to assess the contribution of the different structural components within it. Given their proximity to the 5′ end, the importance of S1 and SL5 was examined first, via compensatory mutational analysis. Destabilization of S1, which includes the 5′ terminus, markedly increased susceptibility of the 5′-monophosphorylated CIRV genome to Xrn1, while restoring base pairing resulted in a notable recovery in resistance (Fig. 6B and C). The removal of the terminal γ- and β-phosphate groups from the 5′-terminal nucleotide triphosphate is a prerequisite for efficient Xrn1 digestion in cells. Thus, to measure accessibility of the 5′-terminal triphosphate to an RNA pyrophosphohydrolase (RppH), the same S1 modifications (Fig. 6B) were tested with Escherichia coli RppH in the smaller DI-7* context using a dephosphorylation assay. 5′-Triphosphorylated DI-7* and mutant transcripts treated with RppH were analyzed by PABLO (phosphorylation assay by ligation of oligonucleotides) to determine the phosphorylation state of their 5′-terminal nucleotide after RppH treatment. In this assay, only 5′-triphosphorylated DI-7* transcripts that were converted to 5′-monophosphates by RppH are ligated to a DNA oligonucleotide. This generates a larger product (or products, due to oligonucleotide concatemerization) that can be detected as covalent DI-7* RNA-DNA oligonucleotide hybrids via denaturing gel electrophoresis. Accordingly, the presence of slower-migrating bands is a positive indicator of RppH activity. DI-7* (WT), with its unmodified 5′UTR segment (i.e., RI*), provided obvious protection from RppH compared to mutants with S1 disruptions, and restoration of base pairing with transposed nucleotides notably restored resistance (Fig. 6D).

FIG 6.

In vitro Xrn1 and RppH treatment of S1 mutants. (A) RNA secondary structure diagram of 5′UTR showing S1 stem boxed in gray. (B) S1 mutant series, with modified nucleotides shown in red. (C) In vitro Xrn1 treatment of CIRV*-based S1 mutants. Percentages of input RNA remaining and relative levels of RNA remaining compared to the wild type (set at 100) are shown with SEM (n = 3). (D) In vitro RppH treatment of DI-7*-based S1 mutants. Percentages of unligated RNA and relative levels of unligated RNA compared to wild type (set at 100) are shown with SEM (n = 3).

Next, the S1-adjacent SL5 was examined in a similar manner to determine its importance for 5′-end protection. Disruption of the stem of SL5 enhanced susceptibility of the genome to Xrn1, while restoration of base pairing effectively recovered resistance to wild-type levels (Fig. 7B and C). An additional SL5 mutant, S5CG, was also tested in which the two noncanonical UG base pairs in SL5 were replaced with more stable CG pairs. When challenged with Xrn1, this mutant performed marginally better than the wild type (Fig. 7C). In the RppH assay, similar results were observed for the corresponding DI-7* SL5 mutants, where predicted base pairing stability correlated with protection (Fig. 7D). Overall, these findings implicate RNA secondary structures in the viral 5′UTR, specifically S1 and SL5, as key determinants of resistance to both Xrn1 and RppH.

FIG 7.

In vitro Xrn1 and RppH treatment of SL5 mutants. (A) RNA secondary structure diagram of 5′UTR showing S5 stem boxed in gray. (B) SL5 mutant series, with modified nucleotides shown in red. (C) In vitro Xrn1 treatment of CIRV*-based SL5 mutants. Percentages of input RNA remaining and relative levels of RNA remaining compared to wild type (set at 100) are shown with SEM (n = 3). (D) In vitro RppH treatment of DI-7*-based SL5 mutants. Percentages of unligated RNA and relative levels of unligated RNA compared to the wild type (set at 100) are shown with SEM (n = 3).

RNA tertiary structures contribute to Xrn1 and RppH blocking activity.

A possible explanation for how SL5 limits Xrn access is that it coaxially stacks on the adjacent S1 stem, forming a tertiary interaction that would further decrease accessibility of the 5′ end (Fig. 8A). To test this idea, one or more uridylates were inserted at the junction between the two SL structures that would impede coaxial stacking (Fig. 8A). The addition of one, two, or five residues at this position increased access of both Xrn1 and RppH (Fig. 8B and C). Curiously, mutant CS5 was less susceptible to RppH activity than either CS1 or CS2, while all three mutants were similarly susceptible to Xrn1, suggesting that RppH and Xrn1 differ somewhat in their accessibility requirements.

FIG 8.

In vitro Xrn1 and RppH treatment of CS mutants impairing coaxial stacking. (A) RNA secondary structure diagram of 5′UTR depicting potential coaxial stacking between S1 and S5. The S1-S5 junction is marked by a red arrowhead, and the nucleotides inserted at this position are depicted in red. (B) In vitro Xrn1 treatment of CIRV*-based CS mutants. Percentages of input RNA remaining and relative levels of RNA remaining compared to the wild type (set at 100) are shown with SEM (n = 3). (C) In vitro RppH treatment of DI-7*-based CS mutants. Percentages of unligated RNA and relative levels of unligated RNA compared to the wild type (set at 100) are shown with SEM (n = 3).

An additional tertiary interaction, PK-TD1, was also assessed (Fig. 9A). Destabilization of the pseudoknot interaction negatively affected the ability of the 5′UTR to block access to both Xrn1 and RppH (Fig. 9C and D). Collectively, these results are consistent with the idea that RNA tertiary interactions in the CIRV 5′UTR are important for attaining a higher-order RNA conformation that keeps 5′-proximal residues inaccessible to these two 5′-end-targeting enzymes. This notion was further supported by examining the electrophoretic mobility of 5′UTR-only fragments containing the above-described modifications in nondenaturing polyacrylamide gels. In all cases, 5′UTRs with different structurally destabilizing alterations exhibited slower migration, indicative of looser, less compact structural conformations (Fig. 10A and B).

FIG 9.

In vitro Xrn1 and RppH treatment of PK-TD1 mutants. (A) Secondary structure diagram of 5′UTR showing the PK-TD1 interaction. The location of the partner sequences involved in the pseudoknot are boxed in gray. (B) PK-TD1 mutant series, with modified nucleotides shown in red. (C) In vitro Xrn1 treatment of CIRV*-based PK-TD1 mutants. Percentages of input RNA remaining and relative levels of RNA remaining compared to wild type (set at 100) are shown with SEM (n = 3). (D) In vitro RppH treatment of DI-7*-based PK-TD1 mutants. Percentages of unligated RNA and relative levels of unligated RNA compared to the wild type (set at 100) are shown with SEM (n = 3).

FIG 10.

Gel mobility assay of 5′UTR (RI*) mutants. (A) Ethidium bromide (EtBr)-stained nondenaturing 10% polyacrylamide gel of S1 and SL5 series mutants. (B) EtBr-stained polyacrylamide gel of PK-TD1 and CS series mutants. A representative gel is shown (n = 2).

Evidence for 5′UTR-mediated protection in a plant extract.

The results presented thus far were carried out using yeast Xrn1, which is the functional homolog of the plant enzyme Xrn4. Due to its cytosolic location, Xrn4 is the most likely candidate for the 5′ targeting of cytosolic CIRV RNAs during infections (9). To gain evidence that Xrn1 was an appropriate surrogate, the effects of the 5′UTR modifications in a plant-based milieu were examined. To do so, a tobacco cell-free lysate (BYL) was employed to determine relative Xrn resistance of 5′-monophosphorylated DI-7* mutants. To confirm that the exoribonucleolytic activity observed was indeed the result of Xrn (32), the SCNMV xrRNA sequence (xrS), which forms a very efficient Xrn-stalling RNA structure (33, 34), was inserted into DI-7* (DI-7*xrS) and its mutants directly after the 5′UTR (i.e., RI*) (Fig. 11A). Stalling of Xrn4 or Xrn1 at xrS would create a readily detectable degradation product (xrS-P) and act as a positive indicator of Xrn activity (Fig. 11A). The results revealed similar trends in xrS-P accumulation from the different DI-7xrS mutants with purified Xrn1 or plant extract containing Xrn4 (Fig. 11B and C). The less intense banding patterns seen for the lysate treatments were likely due to the activity of other ribonucleases present in the plant extract. These findings suggest that the 5′UTR and its mutants respond similarly when confronted with either source of Xrn.

FIG 11.

DI-7*xrS 5′UTR mutants treated with BYL extract. (A) Schematic representation of DI-7*xrS construct and xrS-P product formation following Xrn treatment. The blue box indicates the location of the SCNMV xrRNA sequence that stalls Xrn progression. (B) 5′UTR mutations introduced into DI-7*xrS. Nucleotide substitutions are indicated in red. (C) Control untreated DI-7*xrS and mutants (left), BYL extract-treated DI-7*xrS and mutants (middle), and in vitro Xrn1-treated DI-7*xrS and mutants (right). The positions of DI-7*xrS and xrS-P are indicated on the left. Representative gels are shown (n = 2).

Effects of monophosphorylation and adenylate capping on viral genome accumulation in plant cells.

To test the effect of different 5′ modifications on genome accumulation during plant cell infections, CIRV genomes were transcribed in vitro to generate 5′ ends containing either a natural triphosphate (PPP), an adenylate-cap (A-), or a monophosphate (P) (Fig. 12A). The latter two modifications would be present in the in vitro-transcribed genomes used to transfect the protoplasts but would be converted to 5′-triphosphates in accumulating progeny genomes. Thus, the major effect of the alterations would occur at the beginning of the infection, immediately following transfection. It was anticipated that the A-cap would increase stability of input viral genomes by providing protection against RppH and Xrn4 activity, whereas a monophosphate would decrease stability by making input genomes immediate substrates for Xrn4. Comparing progeny accumulation for input genomes with wild-type 5′UTRs (WT) and triphosphate ends (PPP, set as 100%), corresponding progeny levels for A-capped genomes were lower (A-, 82%), and notably more so for those that were monophosphorylated (P, 9%) (Fig. 12C, far left). The latter result is consistent with input genomes being immediate substrates for Xrn4. The unanticipated lower level of accumulation of progeny genomes in A-cap-modified transfections was most likely due to a defect in replication caused by the added A-cap, because such 5′ modifications are known to negatively impact replication of normally uncapped viral genomes (35). Interestingly, corresponding transfections carried out with mutant sets containing 5′UTRs with destabilized S1, SL5, or PK-TD1 showed increased levels of progeny genome production in A-cap-modified transfections compared to those with triphosphates (∼2-fold relative increase) (Fig. 12C). These reciprocal effects, compared to those for WT transfections, suggest that the initial protective effect of the A-cap on input genomes was more beneficial to mutant 5′UTRs, which are more vulnerable to Xrn4. That is, for 5′UTR mutants, the negative effect of the A-cap on RNA replication was offset by its positive effect on RNA stability. Overall, these findings suggest that the low genome accumulation levels observed in mutants with destabilized 5′UTRs can be partially rescued by a protective 5′-modification (i.e., A-capping).

FIG 12.

Protoplast infections with 5′-modified CIRV* genomes. (A) 5′ modifications introduced into CIRV*. Phosphate groups are shown as yellow circles, and the 5′-linked adenylate is represented by a blue square. PPP, triphosphate; A-, adenylate cap; P, monophosphate. (B) CIRV* 5′UTR mutants selected for 5′ modification. (C) Northern blot analysis of protoplast infection, with corresponding 5′-modified CIRV* genomes indicated above the lanes. Relative levels of genome accumulation compared to PPP (set at 100) are shown with SEM (n = 3).

Finally, to examine the effects of different 5′UTR mutations during CIRV infections, protoplasts were transfected with 5′-triphosphorylated WT and variant genomes. Those with destabilized 5′UTRs accumulated much less efficiently than the wild type or genomes with restored structures (Fig. 13C). Notably, additional large virus-derived RNAs were observed in mutant infections with destabilized 5′UTRs that amassed at levels similar to or higher than those of their cognate genomes (Fig. 13C, arrow). These in vivo-generated truncated genomes, as well as those observed during in vitro Xrn1 treatment of CIRV genomes (Fig. 5C, arrow), will be investigated in detail in future studies. Nevertheless, their detection is consistent with increased 5′-end access of Xrn and implies the presence of stall sites (i.e., xrRNA structures) within the CIRV genome.

FIG 13.

Protoplast infections of CIRV* 5′UTR mutants. (A) RNA secondary structure diagram of 5′UTR showing S1, SL5, and PK-TD1. The p33 start codon is boxed in light green. (B) 5′UTR mutants transfected into protoplasts. Nucleotide substitutions are indicated in red, and the uAUG inactivating mutation is shown in green. (C) Northern blot analysis of protoplast infection of CIRV 5′UTR mutants. An arrow on the right indicates 5′-truncated genomes, and the arrowhead on the left denotes a 3′-derived viral fragment. Relative levels of genome accumulation compared to the wild type (set at 100) are shown with SEM (n = 3).

DISCUSSION

We have investigated the role of the 5′UTR of the tombusvirus CIRV in protecting its plus-strand RNA genome from 5′-terminal attack by 5′-targeting cellular enzymes. Although this genome is not 5′ capped or protein linked, its 5′UTR can fold into a structural conformation that confers resistance to terminal access by both Xrn and RppH. Here, we describe the unique features of this antidecay defense mechanism and compare and contrast the strategy with those from related studies.

Xrn and tombusvirids.

The plus-strand RNA genomes of tombusvirids have a 5′-triphosphorylated nucleotide at their 5′ terminus, which makes them susceptible to 5′-to-3′ exoribonuclease attack. In this regard, Xrn4 has been shown to be a host restriction factor for tombusviruses and is likely also restrictive to other tombusvirids (36). Our results from assessing members of two other genera in the family Tombusviridae, TNV-D (genus Betanecrovirus) and TCV (genus Betacarmovirus), are consistent with the latter notion, as both genomes were more sensitive to Xrn1 attack than CIRV’s genome (Fig. 2). The low level of predicted structure in the 5′UTR of the TCV genome and its high sensitivity to Xrn1 digestion suggests that it, and other viruses with less structured 5′ termini, employ alternate means of end protection during infections (e.g., protein binding). In contrast to TCV, the genome of CLSV (genus Aureusvirus), which contains a 5′UTR structurally comparable to that in CIRV (29), was highly resistant. This underscores the effectiveness of this shared RNA conformation and helps to explain, at least in part, why similarly folded 5′-terminal RNA structures are also present in satellite RNAs associated with viruses in these genera (37, 38).

CIRV’s sg mRNAs were more susceptible to Xrn1 than the genome (Fig. 3). For sg mRNA1, the possibility of stabilization via the 5′ terminus interacting with the 3′CITE was considered but discounted (Fig. 3D). Nonetheless, it remains possible that RNA-mediated stabilization of viral RNAs occurs through long-range RNA-RNA interactions with their 5′UTRs (39). Since sg mRNAs are produced transiently during infections and are not packaged, their resistance to Xrn4 may be less crucial than it is for the genome. However, it is worth noting that susceptibility is not always detrimental, as exemplified by the unique Xrn4/sg mRNA relationship that exists in the genus Umbravirus (family Tombusviridae), where functionally relevant coding sg mRNAs are generated via Xrn4-mediated digestion (19, 20).

Noncoding xrRNAs corresponding roughly to the viral genomic 3′UTR have been identified in certain tombusvirid infections; these include the dianthovirus red clover necrotic mosaic virus (33) and the betanecrovirus TNV-D (21). Similar 3′UTR-derived genomic fragments have also been observed in some tombusvirus infections (40), including this study (Fig. 12C and 13C, arrowhead). These 3′-terminal fragments have been implicated in modulating viral processes related to translation, genome replication, and/or packaging (41). In tombusvirus infections of plants where Xrn4 was overexpressed, viral genome levels were greatly reduced (42). In some cases, 5′-truncated genomes were generated and, astonishingly, one mutant with a 138-nt-long 5′ deletion was found to be capable of limited independent replication and movement in host plants. Precisely how this viral genome, missing nearly all of its 151-nt-long 5′UTR, was able to survive, albeit minimally, is unclear. Regardless, its extremely limited levels of accumulation and spread underscore the negative impact of Xrn4 overexpression and the importance of maintaining a complete 5′UTR for maximal fitness. The structure and properties of the 5′-truncated genomes that accumulated in our protoplast infections (Fig. 13C, arrow) and in vitro assays (Fig. 5B, arrow) remain to be investigated. However, based on their relative sizes, ∼3.3 kb, they would lack approximately 1.5 kb of 5′-proximal genomic sequence, including a significant amount of coding sequence, which would render them incapable of self-replication.

Structural features conferring Xrn blocking activity.

CIRV’s 5′UTR plays important roles in both genome replication and translation of 5′-proximally encoded viral proteins. Different sequence motifs and structures in the plus-strand of the 5′UTR and linear sequences in its complementary minus strand are involved in genome replication (26–28, 43–46), while efficient initiation of translation involves the 5′UTR base pairing with the 3′CITE (47). Accordingly, multiple functional elements involved in different viral processes are integrated within the complex higher-order RNA structure that confers resistance to Xrn decay.

Xrn enzymes require a single-stranded, monophosphorylated 5′-terminal sequence to initiate digestion of an RNA substrate. In vitro studies with Xrn1 have shown that as little as a 2-nt single-stranded 5′ extension can permit engagement by Xrn1 (13). One way for an RNA to achieve protection is to incorporate its 5′-terminal nucleotides within an SL structure, preferably one with significant overall stability and optimal base pairing strength at the bottom of the stem. The yeast 20S RNA narnavirus is an uncapped positive-strand RNA virus that protects its 5′ end from Xrn1 digestion by incorporating it in a large and stable SL structure (48). Notably, the 5′-terminal sequence of the 20S RNA genome is 5′-GGGG, which allows for tight binding with its partner sequence, CCCC, at the base of the stem and effective blocking of Xrn1 access.

In contrast, in CIRV and all tombusviruses and aureusviruses, the genomic 5′-terminal nucleotides are A/U rich due to promoter requirements, and this results in less-than-optimal stability of S1 (Fig. 6A). However, S1 stability would be bolstered if it were to coaxially stack on the adjacent helix of SL5 (Fig. 8A), and this tertiary interaction is supported by the observation that maximal Xrn-blocking activity requires both a stable SL5 (Fig. 7) and its directly adjacent placement with S1 (Fig. 8). Xrn access was also inhibited by formation of a second tertiary interaction, PK-TD1 (Fig. 9), which promoted formation of a more compact global 5′UTR structure (Fig. 10). Notably, modification of any one of the mutationally targeted structural features in the 5′UTR resulted in a more extended structural conformation (Fig. 10). This indicates that a highly cooperative mode of RNA folding is involved in achieving the higher-order structure that effectively blocks Xrn access. Since comparable results were obtained for bacterial RppH access (Fig. 6 to 9), the same RNA structure also may confer CIRV genome protection from plant RppH enzymes. During infections, RppH activity precedes Xrn4 activity in the degradation pathway. The 10-fold higher sensitivity to decay of transfected monophosphorylated CIRV genomes over triphosphorylated counterparts indicates that, in vivo, RppH activities are rate limiting (Fig. 12). Furthermore, because RppHs are gatekeepers for Xrn activity, they are also predicted to be restriction factors for CIRV, as has been demonstrated for hepatitis C virus (11, 49).

Defining the anti-Xrn activity of the 5′UTR.

For CIRV, the 5′UTR on its own was most resistant to Xrn1 (Fig. 4D), suggesting that sequences downstream from it interfere with proper folding of the inhibitory structure. Notably, although this RNA conformation was able to effectively prevent access of Xrn1, it was not able to stall an actively digesting Xrn1 (Fig. 5). The latter deficiency is not surprising, because as long as 5′-end generation in genome replication is reasonably accurate, there would be no selection pressure to adopt xrRNA properties. Thus, the primary activity and corresponding viral function of the 5′UTR structure is to limit access of Xrn to the viral genome and protect it from 5′-to-3′ exoribonuclease attack, respectively. Consequently, to distinguish the central nuclease-inhibiting property of this RNA structure from that of xrRNAs, we propose the term Xrn-evading RNA, or xeRNA.

Intramolecular or cis-xeRNA structures, as in 20S RNA narnavirus and tombusviruses, represent secondary and tertiary structural approaches, respectively, to protect vulnerable 5′-triphosphorylated plus-strand RNA genomes. A different but related tactic, the binding of miRNA-122 to its 5′UTR, is used by hepatitis C virus to guard against attack from both pyrophosphatases and Xrn1 (49–52). In vitro assays showed that miRNA binding alone confers genome protection from Xrn1, categorizing the bimolecular interaction as a trans-xeRNA structure (51). However, in vivo, association of the bound miRNAs with Ago likely further augments shielding of the genomic 5′ terminus, suggesting a multilayered approach to the enzyme blocking strategy (52).

Conclusions.

This study has revealed an important role for the 5′UTR of the CIRV genome in conferring protection from 5′-end-targeting enzymes, such as Xrn and RppH. The RNA sequence involved is composed of secondary and tertiary structural components that fold cooperatively into a compact higher-order RNA conformation that efficiently blocks enzyme access. The structure is also highly integrated with RNA elements involved in genome replication and translation initiation, and equivalent counterparts are present in aureusviruses and subviral satellite RNA replicons. As other tombusvirids face similar challenges in their coexistence with these 5′-end-targeting enzymes, it will be interesting to explore the strategies they use to protect their vulnerable genomes and sg mRNAs.

MATERIALS AND METHODS

Viral constructs.

The full-length cDNA clones used in this study have been reported previously for CIRV (53), cucumber leaf spot virus (CLSV) (54), tobacco necrosis virus-D (TNV-D) (55), turnip crinkle virus (TCV) (56), and DI-7 (24). CIRV subgenomic clones and transcripts were generated as described previously (57). Mutants were derived from CIRV* and DI-7* cDNA constructs, which lacked the upstream AUG (see Results). DI-7*xrS was generated by inserting the SCNMV xrRNA sequence downstream of the 5′UTR in DI-7* (19). Standard overlapping PCR-based mutagenesis was used to introduce mutations into the cDNA clones, which were then sequenced to ensure that only the intended modifications were incorporated.

In vitro transcription and viral RNAs.

All RNAs were in vitro transcribed from either SmaI-linearized clones or PCR-generated templates using a T7-FlashScribe transcription kit as described previously (21). Monophosphorylated RNAs were generated as described previously (21), and RNAs containing a 5′ApppG cap were transcribed using a 10-fold molar excess of cap analog over GTP. Transcript integrity was verified by agarose gel electrophoresis. All RNA structures were generated using UNAfold structure prediction software and arranged with RNA2Drawer (31, 58).

Protoplast infections and Northern blot analyses.

Cucumber protoplasts were isolated from 6-day-old cotyledon leaves as described previously (40). Following isolation, approximately 3 × 10⁵ protoplasts were transfected with 1 μg of genomic CIRV transcripts using polyethylene glycol (PEG)-CaCl₂. Transfected protoplasts were incubated at 22°C for 22 h. Total nucleic acid content was extracted using phenol-chloroform-isoamyl alcohol (PCI) and then ethanol precipitated, after which one-sixth of the total volume was separated in 1.4% agarose gels. Following transfer to nylon, Northern blot analysis was performed using a previously described ³²P-labeled DNA oligonucleotide probe, P9, which is complementary to the 3′ terminus of the CIRV genome (59). All protoplast infections were repeated at least three times.

In vitro Xrn1 digestion assay.

Viral RNAs were transcribed in the presence of excess monophosphorylated-GTP, resulting in its preferential incorporation as the 5′-proximal nucleotide during in vitro transcription with T7 polymerase (21). The Xrn1 assay conditions used yielded ∼70% remaining product for WT CIRV, which allowed for facile detection of increased (or decreased) digestion of different CIRV 5′UTR mutants as well as different viral genomes (e.g., CLSV, TNV-D, and TCV). In vitro Xrn1 degradation assays were performed as described previously using 2 μg of 5′-monophosphorylated in vitro-transcribed RNA and 0.06 U recombinant yeast Xrn1 (New England Biolabs [NEB]) (21). Xrn1 reaction mixtures were incubated at 37°C for 2 h. Following PCI extraction, samples were ethanol precipitated in the presence of glycogen. One-quarter of the recovered RNAs were separated in either 1.4% or 2% agarose gels for nondenaturing conditions or urea-polyacrylamide gels (8 M urea, 10% polyacrylamide) for denaturing conditions, followed by staining with ethidium bromide. Each experiment was repeated three times.

In vitro RppH treatment and PABLO analysis.

Dephosphorylation assays were performed in a manner similar to that of Xrn1 digestion assays, except that purified recombinant RppH from E. coli (NEB) was used. Two micrograms of in vitro-transcribed DI-7* RNA was treated with 5 U of RppH for 2 h at 37°C. Following PCI purification and ethanol precipitation, one-quarter of the isolated RNA was subjected to a modified PABLO (phosphorylation assay by ligation of oligonucleotides), which was described previously (60). Briefly, isolated RNAs were heat denatured at 95°C for 5 min with 20 pmol oligonucleotide X (CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCGAACAATATGAATGATAACTTG) and 4 pmol of oligonucleotide Y1 (TCAAATCCTGAGATATTTCCCAAGTTATCATTCATATTGTTC) or Y2 (for mutants S1a and S1c only; TCAAATCCTGACAAATATCCCAAGTTATCATTCATATTGTTC) and then cooled gradually to 30°C for 30 min. Ligation mixtures (1 μl T4 DNA ligase [NEB], 0.5 μl RNase inhibitor [40 U/μl], 2 μl 10× T4 DNA ligase buffer [NEB], 1 μl 20 mM ATP, 4.5 μl H₂O-diethyl pyrocarbonate [DEPC]) were then added to the reaction mixtures on ice, which were then incubated at 37°C for 2 h. Reactions were quenched with 10 μl of 10 mM EDTA, PCI extracted, and ethanol precipitated in the presence of glycogen. Washed pellets were then dissolved in 5 μl of H₂O-DEPC and combined with 15 μl of RNA loading buffer, after which one-quarter of the sample was separated in a 5% denaturing urea-polyacrylamide gel. All reactions were repeated three times.

RNA EMSA.

RNAs subjected to electrophoretic mobility shift assays (EMSAs) were in vitro transcribed from PCR templates as described above. A modified version of a previously described EMSA protocol (61) was used in the present study. Briefly, 10 pmol RNA was heated at 94°C for 2 min, followed by snap cooling on ice. Reactions were then combined with 1× RNA binding buffer (5 mM HEPES, 6 mM MgCl₂, 100 mM KCl, 3.8% glycerol) and incubated at 37°C for 30 min. The RNAs were then placed on ice, combined with 20% glycerol (10% glycerol final), and separated in a 10% nondenaturing polyacrylamide gel in the presence of 1 mM MgCl₂, followed by staining with ethidium bromide.

Xrn digestion using BYL cell extract.

Ten micrograms of ALiCE cell extract (Sigma) was incubated with 1 μg of DI-7*xrS RNA for 20 min at 25°C. RNAs were then PCI extracted and ethanol precipitated in the presence of glycogen. Once dissolved in DEPC-treated water, one-quarter of the reaction was separated in 2% agarose gels and stained with ethidium bromide. Each experiment was repeated twice, with similar results.