A folded viral noncoding RNA blocks host cell exoribonucleases through a conformationally dynamic RNA structure

Significance

Folded RNA elements are essential for diverse biological processes. Recently discovered examples include viral xrRNAs, which co-opt the cellular RNA decay machinery within a novel noncoding RNA production pathway. Here we characterize an xrRNA with no apparent evolutionary link or sequence homology to those described previously. Our results show that xrRNAs are an authentic class of functional RNAs that have arisen independently in different contexts, suggesting that they may be widespread. The detailed 3D structure of one of these xrRNAs reveals that an underlying structural topology may be the key feature that confers exoribonuclease resistance to diverse xrRNAs.

Abstract

Folded RNA elements that block processive 5′ → 3′ cellular exoribonucleases (xrRNAs) to produce biologically active viral noncoding RNAs have been discovered in flaviviruses, potentially revealing a new mode of RNA maturation. However, whether this RNA structure-dependent mechanism exists elsewhere and, if so, whether a singular RNA fold is required, have been unclear. Here we demonstrate the existence of authentic RNA structure-dependent xrRNAs in dianthoviruses, plant-infecting viruses unrelated to animal-infecting flaviviruses. These xrRNAs have no sequence similarity to known xrRNAs; thus, we used a combination of biochemistry and virology to characterize their sequence requirements and mechanism of stopping exoribonucleases. By solving the structure of a dianthovirus xrRNA by X-ray crystallography, we reveal a complex fold that is very different from that of the flavivirus xrRNAs. However, both versions of xrRNAs contain a unique topological feature, a pseudoknot that creates a protective ring around the 5′ end of the RNA structure; this may be a defining structural feature of xrRNAs. Single-molecule FRET experiments reveal that the dianthovirus xrRNAs undergo conformational changes and can use “codegradational remodeling,” exploiting the exoribonucleases’ degradation-linked helicase activity to help form their resistant structure; such a mechanism has not previously been reported. Convergent evolution has created RNA structure-dependent exoribonuclease resistance in different contexts, which establishes it as a general RNA maturation mechanism and defines xrRNAs as an authentic functional class of RNAs.

During eukaryotic cellular RNA decay, 5′ → 3′ hydrolysis by an XRN protein is important for degrading decapped messenger RNA (mRNA) and other 5′-monophosphorylated RNAs including fragments of endonucleolytic cleavage, ribosomal RNA (rRNA), and transfer RNA (tRNA). Xrn1 is the dominant cytoplasmic 5′ → 3′ exoribonuclease in most eukaryotic cells (1, 2), where its processive translocation-coupled unwinding of RNA helices allows it to efficiently hydrolyze structured RNA substrates without releasing partially degraded intermediates (3, 4). Xrn1 plays a central role in constitutive mRNA turnover and RNA quality control and has been implicated in degrading viral RNAs as part of the cell’s antiviral response (5). Despite the efficiency of Xrn1, some viruses have evolved RNA sequences that robustly block the enzyme’s progression. This ability is conferred by their folded 3D structures without the help of accessory proteins; thus, we refer to them as Xrn1-resistant RNAs (xrRNAs) (6, 7).

xrRNAs were originally identified in the positive-sense RNA genomes of mosquito-borne flaviviruses (e.g., Dengue virus, Zika virus, West Nile virus) where they protect the viral genome’s 3′ UTR from degradation, generating biologically active noncoding RNAs involved in cytopathic outcomes and pathogenicity during infection (8–17). Thus, these mosquito-borne flaviviruses usurp Xrn1’s powerful degradation activity as part of an elegant RNA maturation pathway that relies on a compact and unusual RNA fold. Specifically, these xrRNAs form an interwoven pseudoknot (PK) conformation centered on a conserved three-helix junction, creating a protective ring-like structure that wraps around the 5′ end of the xrRNA (7, 18). This unique topology acts as a mechanical block to Xrn1 when the ring-like structure braces against the surface of the enzyme (7, 19, 20). To date, this topology and resultant mechanism has been observed only in the flaviviruses.

Recruiting Xrn1 to a larger precursor RNA and then blocking the enzyme using a compact structured RNA element to generate a biologically active smaller RNA could represent a useful general mechanism for RNA maturation. Consistent with this, xrRNAs have been identified in a broad range of flaviviruses, including those that are tick-borne, those specific to insects, and those with no known arthropod vector (20). Based on secondary structural patterns, flaviviral xrRNAs can be grouped into two classes; all xrRNAs from mosquito-borne flaviviruses identified to date belong to class 1 xrRNAs, while others compose class 2 xrRNAs (20). The 3D structure of a class 2 flavivirus RNA has not been solved; thus, for simplicity, here we use the term “flavivirus xrRNA” to refer to the well-characterized class 1 flavivirus xrRNAs unless specified otherwise. In addition, RNA sequences that appear to resist 5′ → 3′ exoribonuclease degradation have been identified in a few other virus families (21–25). However, nothing is known about the molecular processes or structures of putative xrRNAs outside the flavivirus family. We do not know whether other candidate exoribonuclease–resistant RNAs operate with the help of protein factors or if they are stably structured elements sufficient to block processive degradation. If RNA structure plays a role in diverse putative xrRNAs, we do not know whether there is a universal structural feature that confers Xrn1 resistance or many structures that can block Xrn1. The degree to which different xrRNAs operate by creating a mechanical block for Xrn1 versus using some other means to stop the enzyme remains unexplored. Indeed, if viruses outside the flavivirus family use diverse sequences and structures to block exoribonucleases in a programmed way, this would indicate that this mechanism is a general pathway, and that xrRNAs are a true functional class of RNAs.

To address these questions, we investigated an unexplored potential xrRNA sequence from the 3′ UTR of plant-infecting dianthoviruses (26, 27) (SI Appendix, Fig. S1A). Dianthoviruses belong to the Tombusviridae family and have a bipartite positive-sense single-stranded RNA genome consisting of RNA1 and RNA2 (26). A recent study described the accumulation of a noncoding viral RNA, SR1f RNA, during dianthovirus infection, generated from the 3′ UTR of viral RNA1 through an unexplored mechanism of incomplete degradation (SI Appendix, Fig. S1A) (27). Using a combination of virology and biochemistry, we show that the dianthovirus 3′ UTRs contain a bona fide xrRNA sequence that inhibits 5′ → 3′ exoribonucleolytic decay in cis without the help of protein factors. X-ray crystallography and single-molecule FRET experiments reveal a unique RNA fold that operates to block 5′ → 3′ exoribonucleolytic degradation through dynamic changes in RNA structure that we term “codegradational” RNA remodeling. This work reveals RNA-dependent exoribonuclease resistance beyond the realm of flaviviruses, conferred by diverse structured RNA elements that nonetheless share a common underlying topology. Thus, convergent evolution has given rise to this mechanism at least twice, indicating it to be a generally useful pathway that may be widespread across biology.

Results

The Dianthovirus 3′ UTR Contains a Bona Fide xrRNA.

We verified that dianthovirus infection results in accumulation of a subgenomic RNA by infecting Nicotiana benthamiana with red clover necrotic mosaic virus (RCNMV) (28). Northern blot analysis of RNA from infected leaves with probes targeting the viral 3′ UTR revealed the presence of a discrete RNA species consistent with the previously reported SR1f RNA (Fig. 1A), which has been shown to be the result of 5′ → 3′ exoribonuclease resistance (27). To determine if RNAs with sequences from the dianthovirus 3′ UTR are sufficient to block 5′ → 3′ exoribonucleolytic decay without trans-acting proteins, we challenged in vitro-transcribed RNA from RCNMV with recombinant Xrn1 (SI Appendix, Fig. S1B). A 130-nt-long RNA from the RCNMV 3′ UTR blocked exoribonucleolytic decay with no protein cofactors, revealing that RCNMV contains an authentic xrRNA (Fig. 1B). We mapped the Xrn1 halt site using reverse transcription and capillary electrophoresis (SI Appendix, Fig. S1 D–F), observing that the enzyme halts at a specific point. We then determined the 3′ terminus of the resistant sequence using test RNAs truncated at their 3′ ends, identifying a 43-nt RNA element as the minimal RCNMV xrRNA (Fig. 1B). Thus, the RCNMV 3′ UTR has an authentic xrRNA contained in a discrete RNA sequence that can operate outside of its natural context.

Fig. 1.

Structure of an authentic xrRNA in dianthovirus 3′ UTRs. (A) Northern blot of total RNA from mock- or RCNMV-infected N. benthamiana. Probes are against viral genomic 3′ UTR and 5.8S rRNA. Full-length RNA1 and exoribonuclease-resistant degradation product (SR1f) are indicated. (B) In vitro Xrn1 degradation assay of ³²P-3′ end-labeled RCNMV 3′ UTR sequences. (C) In vitro Xrn1 degradation assay on minimal xrRNAs from RCNMV, SCNMV, and CRSV. Data are average (± SD) percent resistance from three individual experiments. (D) Secondary structure of the crystallized SCNMV RNA. Lowercase letters represent sequences altered to facilitate transcription. Non–Watson–Crick base pairs are in Leontis–Westhof annotation (41). The Xrn1 stop site is denoted by the labeled arrow. (E) Ribbon representation of the SCNMV xrRNA structure. Colors match those in D. The “single-stranded” RNA at the 3′ end was ordered due to intermolecular interactions in the crystal, described in Fig. 2 D–F and SI Appendix, Fig. S4.

The xrRNA sequence identified in RCNMV is highly conserved in all three known members of the dianthovirus family —RCNMV, sweet clover necrotic mosaic virus (SCNMV), and carnation ringspot virus (CRSV) (SI Appendix, Fig. S1C)—and, accordingly, all viral 3′ UTR sequences confer Xrn1 resistance in vitro (Fig. 1C). As the mosquito-borne flavivirus xrRNAs are able to block diverse exoribonucleases and thus act as general mechanical blocks to these enzymes (20), we tested the three dianthovirus xrRNAs for this ability. All three blocked two 5′ → 3′ exoribonucleases unrelated to Xrn1—yeast decapping and exoribonuclease protein 1 (Dxo1) and bacterial 5′ → 3′ exoribonuclease RNase J1—indicating that they function as general roadblocks against exoribonucleolytic degradation rather than through specific interactions with the enzyme (SI Appendix, Fig. S1 G–I). That this matches the behavior of flavivirus xrRNAs suggests potential similarity in the molecular mechanisms of these xrRNAs. This result also justifies the use of recombinant yeast Xrn1 in our experiments as a proxy for the larger family of eukaryotic XRN proteins, which is important because dianthoviruses infect plant cells whose major cytoplasmic 5′ → 3′ exoribonuclease is the evolutionarily related XRN protein Xrn4 (2).

Dianthovirus xrRNAs Adopt an Unexpected Structure.

Dianthovirus xrRNAs are significantly shorter than flaviviral xrRNAs and there is no readily identified sequence similarity between the two, suggesting that a different structure confers Xrn1 resistance. To understand the structural basis of Xrn1 resistance by the dianthovirus xrRNAs, we aimed to solve the structure of a dianthovirus xrRNA. Only the SCNMV xrRNA yielded diffracting crystals, and we solved the structure of the complete 43-nt SCNMV xrRNA to 2.9-Å resolution using X-ray crystallography (SI Appendix, Fig. S2 and Table S1). The SCNMV xrRNA adopts a global conformation with no immediate similarity to the mosquito-borne flaviviral xrRNA structures (7, 18). Overall, the SCNMV xrRNA comprises a stem-loop (SL) structure with a single-stranded 3′ tail (Fig. 1 D and E). The stem contains coaxially positioned helices P1 and P2, which are linked by stacking of bases and a noncanonical base pair in the L1 internal loop. Helix P2 is capped by L2, an atypical bipartite loop. The 5′ portion of the loop contains a run of single-stranded stacked bases (L2A), followed by a sharp turn in the backbone and a compact hairpin structure (L2B) embedded within the larger loop structure. This L2B element forms long-distance tertiary stacking interactions with the L1 internal loop, bringing L2B into close proximity to the stem, thus causing L2 to adopt an overall “tilted” conformation. Sequence conservation between the dianthoviral xrRNAs suggests that this structure is representative of all three known members (SI Appendix, Fig. S1C).

The Dianthoviral xrRNA Crystal Structure Is an RNA Folding Intermediate.

The SCNMV xrRNA structure did not immediately suggest a mechanism for Xrn1 resistance, but functional data provided insight. The Xrn1 stop site is at the base of the P1 stem (Fig. 1D). This is significant because the entry channel into Xrn1’s active site accommodates only single-stranded RNA, mandating that the helicase activity of the enzyme unwinds RNA so it can enter the enzyme’s interior. The distance from the surface of the enzyme to the buried active site requires that at least five or six nucleotides of RNA downstream of the stop site be single-stranded (Fig. 2A) (3, 4). Thus, the location of the Xrn1 stop site on the dianthovirus xrRNA mandates that P1 be unwound at the moment when Xrn1 halts (Fig. 2B).

Fig. 2.

Partial unfolding of SCNMV xrRNA contributes to Xrn1 resistance. (A, Left) Structure of Drosophila melanogaster Xrn1 [Protein Data Bank (PDB) ID code 2Y35] (3). Red spheres denote active site residues; black, 8-nt RNA substrate. Five single-stranded nucleotides span the distance from the active site to the enzyme’s surface. (A, Right) Secondary structure of SCNMV xrRNA, with the Xrn1 stop site indicated. Yellow denotes the portion of the P1 stem that must unwind. (B) Scheme of Xrn1 unwinding the P1 stem (red). (C) In vitro Xrn1 degradation assay on ³²P-3′ end-labeled WT and mutant SCNMV xrRNAs. Data are average (± SD) percent resistance from three individual experiments. (D) Scheme of putative PK interaction (SI Appendix, Fig. S4). (E) In vitro Xrn1 degradation assay on ³²P-3′ end-labeled WT and mutant SCNMV xrRNAs. Data are average (± SD) percent resistance from three individual experiments. (F) Northern blot of total RNA from two replicates of mock- or RCNMV-infected N. benthamiana. Probe: RCNMV 3′ UTR or 5.8S rRNA.

This finding raised the question of whether formation of the P1 stem is needed to block the enzyme. We tested this by measuring Xrn1 resistance of an RNA mutated to disrupt P1 base pairing (P1_MUT) (all mutants; SI Appendix, Fig. S3). This mutation significantly reduced, but did not abolish, Xrn1 resistance in vitro, while compensatory mutations (P1_COMP) that restore Watson–Crick base pairing completely rescued Xrn1 resistance (Fig. 2C). Thus, formation of the P1 stem is not essential for Xrn1 resistance but does contribute to full activity, yet P1 must be unwound during the exoribonuclease halting event. Therefore, we hypothesized that the dianthovirus xrRNA structure observed in the crystal (with the P1 helix formed) is an important folding intermediate, but the RNA must change conformation in a programmed way to block Xrn1.

An RNA PK Interaction Is Necessary for Xrn1 Resistance.

The crystal structure suggests what the aforementioned putative conformational change might entail. Within the crystal lattice, nucleotides C43, U44, and C45 at the single-stranded 3′ end of one RNA molecule form Watson–Crick base pairs with G20, A19, and G18 in the loop region (L2A) of an adjacent RNA molecule (SI Appendix, Figs. S2C and S4A). The formation of these pairs in trans in the crystal suggests that they could form in cis, creating a PK between the L2A and S2 regions (Fig. 2D). The potential to form this PK is present in all three dianthovirus xrRNA sequences (SI Appendix, Fig. S4B), and in fact this putative PK can be extended by one additional Watson–Crick base pair (C17-G46, not present in the crystallized RNA). Indeed, addition of G46 to the 3′ end of the SCNMV xrRNA increased Xrn1 resistance in vitro and prompted us to perform all subsequent experiments using this 44-nt-long xrRNA sequence (SI Appendix, Fig. S4C).

To determine the importance of the L2A-S2 PK, we mutated either L2A or S2 in SCNMV, RCNMV, and CRSV xrRNA and assessed Xrn1 resistance in vitro (Fig. 2E and SI Appendix, Fig. S4D). In all xrRNAs, disruption of the putative PK led to loss of Xrn1 resistance, whereas compensatory mutations (PK_COMP) partially rescued Xrn1 resistance. We tested the importance of the PK for the accumulation of dianthovirus SR1f RNA by infecting Nicotiana benthamiana with RNA transcripts generated from an RCNMV RNA-1 infectious clone with the WT, PK_S2, or PK_COMP xrRNA sequence. Matching the in vitro data, disruption of the PK caused a complete loss of SR1f RNA formation, whereas compensatory mutations partially rescued production (Fig. 2F). We also determined that the PK interaction does not function in trans as a dimer (SI Appendix, Fig. S5 A–C). These results show that the dianthovirus xrRNAs require the L2A-S2 PK interaction to block Xrn1.

xrRNA PK Formation Creates a Protective Loop Around the 5′ End.

To understand how L2A-S2 PK formation could lead to exoribonuclease resistance, we constructed a model with this interaction formed in cis and the P1 helix unwound, mimicking the structure at the moment when Xrn1 is halted (Fig. 3 A and B). Even though it is theoretically possible that the PK forms without unwinding P1, this would require a much more dramatic distortion of the overall structure. In addition, the experimentally determined Xrn1 stop site at the base of P1 indicates that the stem must unwind to allow RNA entry into the active site, and thus P1 is not part of the Xrn1-resistant state. In the modeled conformation, the 3′ end of the dianthovirus xrRNA encircles the 5′ end, thereby creating a topology similar to the protective ring-like fold of mosquito-borne flaviviral xrRNAs (Fig. 3 B and C). Thus, these xrRNAs appear to rely on a similar overarching strategy to block Xrn1, which is remarkable and unexpected given their very different sequences.

Fig. 3.

PK formation creates a protective ring around the 5′ end. (A) Crystallized SCNMV xrRNA. The arrow indicates the conformational change needed to form the PK interaction. (B) Modeled xrRNA PK conformation (all 3 nt of P1 unwound, PK formed). PK is in orange. The 3′ end (blue) encircles the 5′ end (red). (C) Structure of ZIKV xrRNA1 (PDB ID code 5TPY) with 5′ end (red), ring-like fold (blue), and PK (orange) matching the colors in A and B.

In both cases, an apical loop forms long-range base pairs with a downstream element, causing intervening sequences to loop protectively around the 5′ end (SI Appendix, Fig. S5D). In the flavivirus xrRNAs, the loop is formed by two intact helical elements, while in the dianthovirus xrRNAs, the P1 helix must be unwound. A comparison of these two xrRNAs shows the ring-like structure to be an emerging theme of RNA structure-based 5′ → 3′ exoribonuclease resistance, and convergent evolution can select different RNA sequences and structures to achieve this topology.

The SL conformation with P1 paired likely represents a folding intermediate to position S2 in such a way that when the PK forms it will encircle the 5′ end of the RNA rather than forming a ring with the 5′ end outside the loop. This interpretation suggests that the SL-to-PK switch itself is promoted by the tilted SL conformation created by tertiary interactions formed by the highly conserved sequence (SI Appendix, Fig. S6A). One such interaction is enabled by the conformation of L2B, which is essentially a smaller loop embedded in the larger L2 loop. The backbone of L2B is similar to the U-turn found in some RNA tetraloops, but the loop is closed by a reverse Watson–Crick base pair between U21 and A28, with the latter base extruded from the P2 helix (SI Appendix, Fig. S6B). The L2B structure creates a “docking point” for A7, which is extruded from the L1 internal loop to stack between A25 and A26, thereby inducing the overall tilted conformation (SI Appendix, Fig. S6C). A similar long-range hairpin–adenosine interaction is found in 16S and 18S rRNA (SI Appendix, Fig. S7). The importance of this interaction and thus the overall tilted SL conformation is revealed by an A7U mutant, which abolishes Xrn1 resistance in vitro and prevents SR1f RNA production during RCNMV infection (SI Appendix, Fig. S6 D and E).

Conformational Dynamics of Dianthovirus xrRNA.

Our model suggests that formation of the essential S2-L2A PK involves a conformational switch between the SL and PK states. To test this idea, we used single-molecule Förster resonance energy transfer (smFRET). Using the crystal structure and the PK model as a guide, we designed an RNA with dye modifications at sites that would result in high FRET if the PK was formed and a mid-FRET state if not (Fig. 4A and SI Appendix, Fig. S8A). We verified that SCNMV xrRNA labeled with acceptor dye (Cy5) at position U15 was Xrn1-resistant (SI Appendix, Fig. S8B), then annealed this RNA to a surface-immobilized DNA oligo labeled with a donor dye (Cy3) and monitored FRET efficiency by total internal reflection fluorescence (TIRF) microscopy.

Fig. 4.

Conformational dynamics of SCNMV xrRNA. (A) smFRET labeling strategy. (B) smFRET histograms of WT xrRNA. (Left) RNA construct. Average FRET values are shown. n, number of molecules in the histogram. 18% and 82% of molecules populate the mid-FRET state and high-FRET state, respectively. (C) Representative smFRET traces of individual WT xrRNAs. (Left) Dynamic population. (Right) Stable high-FRET population. (D) smFRET histograms of PK_S2 xrRNA. (Left) Scheme of the RNA construct. (E) Representative smFRET traces of two individual PK_S2 xrRNA molecules.

WT SCNMV xrRNA exists predominantly in two states with roughly 0.8 and 0.4 FRET efficiency (Fig. 4 B and C). To assign these high- and mid-FRET states to specific conformations, in smFRET experiments we used the PK_S2 mutant, which predominantly adopts the mid-FRET–associated conformation (Fig. 4 D and E). Surprisingly, >80% of the WT RNA molecules adopted the high-FRET state, indicating that the PK state is more prevalent in solution than the SL state. This was true even when the Mg²⁺ concentration was decreased from 10 mM to near-physiological 1 mM (SI Appendix, Fig. S9). smFRET traces following individual molecules over time demonstrate two types of WT SCNMV xrRNAs: RNAs “locked in” a high-FRET state and RNAs that undergo rapid conformational dynamics between the two states (Fig. 4C and SI Appendix, Fig. S8C). This suggests two configurations in the PK state, indistinguishable by FRET: an unstable PK, which rapidly transitions back into the SL state, and a stable PK that remains folded for a longer period (perhaps stabilized by additional interactions). A subset of PK_S2 molecules display a very rapid exchange between mid- and high-FRET states (Fig. 4E and SI Appendix, Fig. S8D), but none remain in the high-FRET state; these molecules may sample the PK conformation but revert back to the SL state.

Of note, P1_MUT xrRNA has the same fraction of RNA molecules in the high-FRET state as WT xrRNA, despite being less efficient in blocking Xrn1 (SI Appendix, Fig. S10 C and D). This suggests that in the absence of P1, the xrRNA can still adopt a PK state, but either one that is Xrn1-resistant or one that is misfolded and is not resistant (SI Appendix, Fig. S10E). The misfolded RNA would be indistinguishable from properly folded xrRNA by smFRET (SI Appendix, Fig. S10 A–D). A 50:50 distribution of properly folded and misfolded RNA would produce the 50% resistance efficiency observed for P1_MUT. Overall, these data are consistent with the idea that the formation of P1 acts as a folding intermediate to ensure that the 5′ end is encircled when the PK forms.

Codegradational Remodeling of Dianthovirus xrRNA Contributes to Xrn1 Resistance.

The Xrn1 resistance of dianthoviruses relies on the SL-to-PK conformational change, yet only the PK state can block Xrn1. These observations raise several questions: What happens if Xrn1 encounters an xrRNA in the SL conformation? Is it able to quickly degrade the RNA, or does the xrRNA refold in the presence of the exoribonuclease? Degradation-coupled unwinding of the P1 stem could potentially liberate the 3′ end of the RNA structure, favoring the formation of the PK and inducing Xrn1 resistance through codegradational RNA remodeling. If this were so, then this model would require that RNA remodeling be faster than Xrn1-induced RNA hydrolysis.

To test how an RNA in the SL conformation responds to Xrn1 degradation, we constructed an xrRNA with a 7-nt 5′ extension that captures the 3′ end of the RNA in an extended P1 stem (P1_EXT), preventing PK formation (Fig. 5A and SI Appendix, Fig. S10F). As expected, the P1_EXT xrRNA almost exclusively adopted a mid-FRET state corresponding to the SL conformation, indicating that the PK is not formed in this mutant (Fig. 5B). However, P1_EXT is Xrn1-resistant (Fig. 5C), suggesting that on this RNA, the SL-to-PK switch required for exoribonuclease resistance is triggered during degradation, likely through unwinding of P1. Consistent with this, treating P1_EXT xrRNA with Xrn1 and examining the resultant RNA product by smFRET showed a redistribution of molecules from a mid-FRET to a high-FRET state (Fig. 5D), demonstrating that the enzyme has shifted the xrRNA conformation from the SL state to the PK state. Treatment with a catalytically inactive Xrn1 did not result in this redistribution (Fig. 5E), indicating that the enzyme’s degradation-linked helicase activity is necessary for the observed conformational change of P1_EXT. Specifically, this result strongly suggests that unwinding of the P1 stem liberates the 3′ end of the structure to encircle the 5′ end and form the PK. These data show that Xrn1 can be co-opted into a “pseudochaperone” function to favor the structure that ultimately prevents enzyme progression (Fig. 5F).

Fig. 5.

Codegradational remodeling of SCNMV xrRNA structure. (A) Schematic of smFRET construct with an extended P1 stem (P1_EXT). (B) smFRET histograms of P1_EXT xrRNA. Average FRET values are shown. n, number of molecules in the histogram. (C) In vitro Xrn1 resistance assay of WT and P1_EXT xrRNA. (D and E) smFRET histograms of P1_EXT xrRNA pretreated with WT Xrn1 (D) or mutant Xrn1(D208A) (E). (F) Scheme of the conformational changes and codegradational remodeling of SCNMV xrRNA.

Discussion

Exoribonuclease resistance as a means of generating noncoding RNA from larger precursors is an emerging theme in virology (21–25, 27). We have found that exoribonuclease-blocking sequences from plant-infecting dianthoviruses are authentic xrRNAs, generating viral noncoding RNAs in a process dependent on a specific but dynamic 3D RNA fold. Furthermore, these xrRNAs can fold spontaneously in vitro using a defined conformational intermediate, and also are capable of using the RNA unwinding activity of the exoribonuclease to promote or stabilize their active structure in a form of codegradational remodeling. These findings establish xrRNAs as a distinct functional class of RNAs and suggest that RNA structure-dependent exoribonuclease resistance is a general mechanism of RNA maturation.

Flaviviruses and dianthoviruses belong to unrelated virus families and do not infect the same host species, and the xrRNA sequences have no similarity. Nonetheless, in both a folded xrRNA in the 3′ UTR of a longer viral RNA is used to produce shorter noncoding RNAs that then play multiple (but different) important roles during viral infection (8–17, 27). Thus, a similar RNA maturation strategy has emerged in very divergent viruses, attesting to its utility and potential pervasiveness. Xrn1 can degrade most cellular RNAs, even highly structured ones (2); the fact that discrete xrRNA elements can stop the enzyme suggests that they have characteristics distinct from most folded RNAs. Indeed, while the 3D structures of class 1 flaviviral xrRNAs and dianthoviral xrRNAs are different, they have similarities that suggest a shared overarching strategy. Both flaviviral and dianthoviral xrRNAs rely on a PK, a common RNA fold with many different configurations (29–31). However, in these xrRNAs, the PK creates a ring-like fold around the 5′ end of the structure, a topology not observed in other PK structures. The ring comprises sequences within the PK and downstream of the protected 5′ end; thus, exoribonuclease hydrolysis-coupled helicase activity is unable to access or unfold the structure when the ring braces against the enzyme’s surface (7). In addition, this shared topology suggests a way to confer directionality; the protective ring poses an obstacle to 5′ → 3′ decay, yet can be easily unwound by enzymes approaching from the 3′ side, such as the viral RNA-dependent RNA polymerase. These ring-like folds have not yet been found in other RNAs, and thus they might be a defining and necessary feature of xrRNAs.

Although flaviviral and dianthoviral xrRNAs share several similarities, there are important differences. In the flavivirus version, the fold requires a three-way junction, which interacts with the 5′ end directly to form a second PK interwoven with the first PK (7, 18). The stretch of RNA that forms the ring is almost entirely base-paired, and a number of other tertiary interactions, such as base triples and long-range noncanonical base pairs, stabilize the final fold (SI Appendix, Fig. S5C). The dianthoviral xrRNAs are substantially shorter, there is no three-way junction, and most of the RNA that forms the ring appears to be single-stranded as a result of the unwinding of the P1 stem (SI Appendix, Fig. S5C). These differences reveal how very divergent sequences and secondary structures can give rise to the ring-like topology. Because the crystal does not contain the dianthoviral xrRNA in the PK form, we do not yet know the full repertoire of tertiary interactions that stabilize this conformation; local remodeling of the structure could give rise to additional interactions. A specifically configured PK may be a common feature of all xrRNAs, but more structures are needed to test this and to identify other shared features.

The unusual nature of the xrRNA fold invites speculation as to how it forms correctly; RNA PKs are ubiquitous, but only xrRNAs have been observed to form a protective ring. This implies that the high levels of Xrn1 resistance observed in vitro require a mechanism to ensure that the ring always wraps around the 5′ end of the structure. In flaviviral xrRNAs, the structure shows that the 5′ end must interact directly with the three-way junction for the junction to fold and form the ring structure (19). In dianthoviral xrRNAs, the data suggest that the P1 helix initially forms to position the 5′ and 3′ ends correctly (SL state), and then a programmed conformational change leads to P1 helix unwinding and PK formation (PK state). This pathway could help ensure that PK formation establishes the ring around the 5′ end rather than forming a ring with the 5′ end outside (SI Appendix, Fig. S10E). Accordingly, mutations affecting the putative folding intermediate (i.e., mutations to the P1 stem) lead to an approximate 50% reduction in Xrn1 resistance; the most straightforward interpretation of this finding is that failure to form the P1 stem decouples PK and ring formation from 5′ end placement, and approximately 50% of the molecules are misfolded.

Additional evidence for the folding model described above is the fact that dianthoviral xrRNAs retain their activity even when the PK state is prevented from forming through an extended P1 stem. This observation and the shift toward the PK state when Xrn1 degrades the RNA reveals that RNA remodeling can occur codegradationally, and that the conformational switch between the SL and PK states is faster than the progression of the enzyme through the structure. Thus, the helicase activity of Xrn1 can be co-opted into a pseudochaperone function; by unfolding P1, Xrn1 liberates the nucleotides required for PK formation and helps shift the conformation to the resistant state.

Although smFRET shows that PK formation can occur spontaneously in vitro, within the cell and in the context of the full-length viral RNA, codegradational remodeling may be important. It could allow the xrRNA to respond to the RNA decay machinery while remaining sufficiently flexible for 3′ → 5′ unfolding by the viral RNA-dependent RNA polymerase. At the very least, the presence of the enzyme favors formation of the PK by unwinding a potentially competing structure. To our knowledge, there are no other known examples of codegradational RNA structure formation or remodeling, but our discovery suggests that a similar pseudochaperone function dependent on exoribonucleases or helicases may exist elsewhere.

Exoribonuclease resistance conferred by RNA structure may be a more common mechanism than hitherto anticipated, and recent studies have identified putative xrRNAs in several virus families. A conserved sequence that blocks 5′ → 3′ exoribonucleolytic decay was recently found in the 3′ UTR of plant-infecting benyviruses and cucumoviruses, both of which have multipartite positive-sense RNA genomes (23–25, 32). Moreover, Xrn1 is confounded by RNA elements in the 5′ UTRs of hepaciviruses and pestiviruses (21), as well as by a G-rich sequence structure in the N mRNA of Rift Valley fever virus and multiple RNA structures in ambisense-derived transcripts of arenaviruses (22). There are no obvious common sequence patterns, and most await detailed functional and structural characterization.

The range of biological functions of ncRNAs produced by xrRNAs (8–17, 23, 27), as well as their presence in diverse virus families, suggest that regulated exoribonuclease resistance could be a pathway for RNA maturation in cells. G-rich RNA sequences installed artificially into mRNAs have been shown to confound Xrn1 activity in yeast (presumably through G-quadruplex structures), but the same G-rich sequences appear unable to stall exoribonucleolytic decay in mammalian cells (33). Moreover, pentatricopeptide repeat proteins function as protein barriers for exoribonucleolytic decay and thus define mRNA termini in chloroplasts and mitochondria (34–36). Likewise, rRNA processing involves precise trimming by exoribonucleases (37, 38). Cellular xrRNAs might exist to generate biologically active 5′-truncated decay intermediates; indeed, many RNA-centered processes used by cells were first discovered in viruses, including RNA structures that block 3′ → 5′ decay (39, 40). Thus, RNA elements identified in viruses can inform the search for related molecular structures elsewhere in biology and motivate ongoing research to understand the enormous complexity and diversity of structured RNAs.

Materials and Methods

Detailed information on RNA and protein production, viral infections, Northern blot analysis, halt site mapping, smFRET, and RNA structure determination is provided in SI Appendix, Materials and Methods.