Structures required for poly(A) tail-independent translation overlap with, but are distinct from, cap-independent translation and RNA replication signals at the 3′ end of Tobacco necrosis virus RNA

Abstract

Tobacco necrosis necrovirus (TNV) RNA lacks both a 5′ cap and a poly(A) tail but is translated efficiently, owing in part to a Barley yellow dwarf virus (BYDV)-like cap-independent translation element (BTE) in its 3′ untranslated region (UTR). Here, we identify sequence downstream of the BTE that is necessary for poly(A) tail-independent translation in vivo by using RNA encoding a luciferase reporter gene flanked by viral UTRs. Deletions and point mutations caused loss of translation that was restored by adding a poly(A) tail, and not by adding a 5′ cap. The two 3′-proximal stem-loops in the viral genome contribute to poly(A) tail-independent translation, as well as RNA replication. For all necroviruses, we predict a conserved 3′ UTR secondary structure that includes the BTE at one end of a long helical axis and the stem-loops required for poly(A) tail-independent translation and RNA replication at the other end. This work shows that a viral genome can harbor distinct cap- and poly(A) tail-mimic sequences in the 3′ UTR.

Introduction

Translational control is a major step of gene regulation for RNA viruses, oocytes, and other systems with little or no transcriptional control. Most translational control elements and features in mRNAs exist in the 5′ and 3′ untranslated regions (Hughes, 2006; Mignone et al., 2002; Pesole et al., 2001; Wilkie et al., 2003). On average, 3′ UTRs are substantially longer than 5′ UTRs (Pesole et al., 2001). Consequently, the 3′ UTR is a region with great regulatory potential. Many 3′ UTRs contain translational control elements and an array of diverse binding sites for regulatory proteins or RNAs (He and Hannon, 2004; Kuersten and Goodwin, 2003; Wickens et al., 2000; Wilkie et al., 2003).

A ubiquitous 3′ UTR element is the poly(A) tail. It regulates both stability and translational efficiency of mRNAs (Jacobson, 1996). The 5′ cap and poly(A) tail function synergistically to facilitate efficient translation initiation via circularization of mRNA-initiation factor complex (Gallie, 1991; Hentze, 1997; Preiss and Hentze, 1999; Preiss and Hentze, 1998; Tarun and Sachs, 1995; Wells et al., 1998). Eukaryotic initiation factor (eIF) 4E binds the 5′ cap and is associated with eIF4G. eIF4G also binds poly(A) binding protein (PABP), which binds to the poly(A) tail (Kahvejian et al., 2005). Thus mRNA is circularized (Gallie, 1991; Hentze, 1997; Preiss and Hentze, 1999; Preiss and Hentze, 1998; Tarun and Sachs, 1995; Wells et al., 1998). The mRNA circularization provides a framework to understand how elements within 3′ UTRs control translation.

Many positive strand viral RNAs have a cap structure but lack a poly(A) tail. Generally, specific sequences within the 3′ UTR replace the function of a poly(A) tail in translation (Gallie and Kobayashi, 1994; Gallie and Walbot, 1990). This region also contains the promoter for negative strand synthesis, which can be regulated by switchable repressor activity in many plant viruses (Dreher, 1999; Dreher and Hall, 1988; Matsuda et al., 2004; Olsthoorn et al., 1999; Petrillo et al., 2005; Pogany et al., 2003; Zhang et al., 2004a).

The RNAs of Tobacco mosaic virus (TMV) (Gallie et al., 1996; Gallie and Walbot, 1990), rotavirus (Piron et al., 1998; Varani and Allain, 2002), Turnip yellow mosaic virus (Gallie and Kobayashi, 1994; Matsuda and Dreher, 2004), and Brome mosaic virus (Gallie and Kobayashi, 1994) have functional alternatives for the poly(A) tail in the 3′ UTR. Histone mRNAs of metazoans, but not plants, also lack a poly(A) tail and harbor a stem-loop structure that functionally mimics a poly(A) tail (Gallie et al., 1996; Gorgoni et al., 2005). The 3′ UTR of Alfalfa mosaic virus (AMV) contains a series of stem-loops separated by AUGC motifs that bind viral coat protein to facilitate poly(A) tail-independent translation (Neeleman et al., 2001). Multiple domains in the 3′ UTR each contribute to the poly(A) tail-independent translation of dengue virus RNA (Chiu et al., 2005; Holden and Harris, 2004). While poly(A) tail-independent translation has been studied in capped viral RNAs, sequences that substitute for poly(A) tails have not been clearly identified on uncapped viral RNAs.

To investigate poly(A) tail-independent translation of a naturally uncapped viral RNA, we used Tobacco necrosis virus strain D (TNV-D) as a model system. TNV is the type member of genus Necrovirus in the Tombusviridae family. TNV-D has a positive sense single-stranded RNA genome of 3762 nt that has neither a 5′ cap (Lesnaw and Reichmann, 1970) nor a 3′ poly(A) tail (Coutts et al., 1991). TNV-D RNA encodes six open reading frames (ORFs) (Fig. 1A). We reported previously (Shen and Miller, 2004b) that TNV-D RNA has a BYDV-like cap-independent translation element (BTE) within the 3′ UTR that functionally mimics a 5′ cap (Fig. 1A). When located in the 3′ UTR (its natural location), the TNV BTE is predicted to base pair to the viral 5′ UTR, via kissing stem-loops, to function. The BTE also functions in the 5′ UTR when the AUG triplets within the BTE are altered to prevent initiation before the correct start codon. A 105 nt sequence (nts 3555–3659) is sufficient for cap-independent translation in wheat germ extract, but an undetermined amount of additional sequence downstream is required for efficient cap-independent translation in vivo (Shen and Miller, 2004b). The extent and role of the extra sequence is not clear.

Fig. 1.

Deletion mapping of 3′ UTR translational control signals in oat protoplasts. A. TNV genome organization. Open boxes represent open reading frames (ORFs) with encoded protein names in or beside boxes. Shaded box indicates the in vitro-defined BYDV-like cap-independent translation element (BTE). Bold black lines represent genomic RNA (gRNA) and subgenomic RNAs (sgRNAs). B. Left side: maps of TLucT deletion mutants (not to scale). Positions of deletions are indicated using the wild type TNV-D genomic RNA numbering. fLuc: firefly luciferase ORF. 5′ and 3′ UTRs are indicated by black bold lines with blank areas indicating deleted portions. Right side: relative luciferase activities of translation products of TLucT and deletion mutants obtained 4 hr after electroporation of protoplasts. Luciferase assays were performed in at least three independent experiments, each of which was in triplicate. Standard deviations are indicated.

Here, we report that a small portion of the extra sequence is necessary for cap-independent translation in vivo, but most of the extra sequence required for translation in vivo can be replaced by a 60 nt poly(A) tail, and not by a 5′ cap. This suggests that the extra sequence has a poly(A) tail-substitute function. The extra sequence includes two stem-loops well-known to serve as the promoter for initiation of negative strand RNA synthesis. These results show that the poly(A) tail-independent and cap-independent translation require different, but possibly overlapping portions of the viral 3′ UTR.

Results

A sequence within the TNV-D 3′ UTR functionally replaces a poly(A) tail

We showed previously that the in vitro-defined BTE of TNV has a 3′ boundary immediately upstream of nt 3659, but an unknown amount of additional downstream sequence is needed for full expression of capped and uncapped RNAs in vivo (Shen and Miller, 2004b). Because poly(A) tails are necessary for translation of plant mRNAs in vivo, but not in wheat germ extract (WGE), we speculated that it might be possible to replace the additional sequence with a poly(A) tail. To test the function of the 103 nucleotides downstream of the BTE needed for translation in vivo, we made a series of deletions and truncations in this region in reporter construct TLucT, which encodes the firefly luciferase ORF flanked by TNV-D 5′ and 3′ UTRs (Shen and Miller, 2004b). We examined the translatability of these mutated RNAs containing (i) both cap and a 60 nt poly(A) tail, (ii) cap only, (iii) poly(A) tail only, or (iv) neither a cap nor a poly(A) tail.

Addition of a 60 nt poly(A) tail to TLucT with the complete 3′ UTR increased luciferase expression about 2.5-fold (Fig. 1B). While this is significantly higher than expression in the absence of the tail, it is far less than the level of stimulation seen on cellular mRNAs that require a poly(A) tail (Gallie, 1991; Tarun and Sachs, 1995). Thus, significant translation of TLucT occurs in the absence of the poly(A) tail.

Deletions of portions of the 3′ end between nt 3720 and 3762 (the 3′ terminus) reduced translation to about 40% of that from TLucT (Fig. 1B, TLucT, D3720, D3748). Addition of a 5′ cap had no notable effect on translation. However, addition of a 60 nt poly(A) tail alone restored translation to a similar level as TLucT with a poly(A) tail. Addition of a 5′ cap and a poly(A) tail had a similar effect on translation as the poly(A) alone. We conclude that sequences between nts 3720 and 3745, and downstream of nt 3748 are necessary for poly(A) tail-independent translation, and are not required for cap-independent translation.

Truncations further upstream, from the 3′ end up to nt 3700 or 3680 reduced luciferase expression 14 to 20-fold relative to TLucT. Again, addition of a 5′ cap had no effect on translation of transcripts with or without a poly(A) tail. Presence of the 60 nt poly(A) tail increased luciferase expression about five- to 8-fold (D3680, D3700, Fig. 1B), but only to about half of the level obtained from nonpolyadenylated TLucT. Thus, full translation could not be restored on RNAs with these larger truncations, even in the presence of a 5′ cap and poly(A) tail.

Truncating still further upstream, to nt 3661, caused a 17-fold reduction in translation. Addition of a 60 nt poly(A) tail increased translation 4.5-fold. Addition of a 5′ cap increased translation 2.5-fold. Interestingly, a 37-fold increase in translation was observed in D3661 RNA by adding both a 5′ cap and a poly(A) tail. The translation level of capped, polyadenylated D3661 RNA was similar to capped, polyadenylated TLucT RNA (Fig. 1B, TLucT, D3661). Thus, when all sequence downstream of 3661 was deleted, addition of a cap and poly(A) had a strikingly synergistic effect.

Taken together, these data show that sequence between nts 3661 and 3680 is necessary for cap-independent translation in vivo, even though it was not needed in vitro. Sequence downstream of nt 3680 does not contribute to cap-independent translation. Why the capped, polyadenylated forms of the two RNAs with intermediate truncations (D3610 or D3700) translated at lower levels than the capped, polyadenylated RNA with even more sequence deleted (D3661) is unclear, but it is obvious that addition of a poly(A) tail increased translation at least five-fold. Thus, sequence downstream of nt 3681 contains element(s) required only for poly(A) tail-independent translation. At least a portion of this poly(A) tail-independent translation element (PITE) extends downstream of nt 3748.

Two stem-loops are required for poly(A) tail-independent translation

To guide investigation of the structure and function of nts 3681–3762, we used the MFOLD program (Zuker, 2003) to predict the secondary structure of this portion of the PITE. Nucleotides 3681–3762 are predicted to form a structure with two stem-loops separated by a 15 nt single-stranded pyrimidine-rich tract (nts 3725–3739) (Fig. 2A). Stem-loop II has two internal bulges. Based on the results in Figure 1B, deletions that truncate or remove stem-loop I (D3720, D3748) reduced poly(A) tail-independent translation by 50%, and deletions that removed both stem-loops (D3661, D3680, D3700) abolished translation in the absence of a poly(A) tail.

Fig. 2.

Effect of pyrimidine-rich tract deletion and SL-I mutations on TLucT translation and TNV-D replication. A. Predicted secondary structure of the 3′ end of the TNV genome showing mutated bases in bold, and boxed with corresponding base on opposite strand. Dash between bases indicates Watson-Crick complementarity. Adjacent mutated bases in loop L1 and pyrimidine-rich tract are also boxed. Names of constructs are shown beside boxes. B. Relative luciferase activity expressed from TLucT, pyrimidine-rich tract deletion (d3726/3738) and SL-I mutants, in oat protoplasts. C. Northern blot hybridization of total RNA accumulated in NT-1 protoplasts 24 hr after inoculation with TNV-D transcripts containing the indicated mutations. Probed with negative sense transcript complementary to the 3′ end of the TNV-D genomic RNA. Mobilities of viral genomic and subgenomic RNA are indicated at right. The bottom panel shows ethidium bromide staining of the gel prior to blotting, to indicate the amount of RNA loaded in each lane.

To determine the primary and secondary structures required for the PITE function, we introduced mutations into the stems and loops, and deleted the pyrimidine-rich tract (Figs. 2 and 3). Deletion of the pyrimidine rich tract (D3726–3738) decreased luciferase activity about 3-fold (Fig. 2B). The Loop I mutation (LI-mut) had little effect on translation. SL-I mutations C3753G and G3744C, predicted (by mfold (Zuker, 2003)) to weaken or disrupt the SL-I helix, reduced translation by 2- and 3-fold, respectively. The combination of these two mutations (SI-re), which is predicted to restore the stem, restored translation efficiency to the wild-type level. Presence of the 60 nt poly(A) tail caused all constructs to generate about 2- to 2.5-fold more luciferase than non-polyadenylated “wild type” TLucT RNA (Fig. 2B). Thus, the pyrimidine tract and the secondary structure of SL-I are important for poly(A) tail-independent translation.

Fig. 3.

Effects of mutations in SL-II and the 3′ terminus on translation and TNV-D replication. A. Predicted secondary structure of the 3′ end of the TNV genome. Mutated bases are in bold, and boxed with corresponding base on opposite strand, or with adjacent bases if in a predicted loop or bulge. Names of constructs are shown beside rectangles. Dashed arrow indicates potential base pairing between the terminal four bases and the bulge in SL-II, that is disrupted or weakened in constructs U3714A and A3759U and restored in the double mutant, AU. B. Relative activity of luciferase translated from TLucT transcripts harboring indicated mutations without (unshaded bars) or with (shaded bars) a poly(A) tail. C. Northern blot hybridization of TNV-D RNA accumulated in NT-1 protoplasts 24 hr after inoculation with TNV-D transcripts containing the indicated mutations. Mobilities of viral genomic and subgenomic RNA are at right. Ethidium bromide staining of the gel prior to blotting is below the blot. D. Relative luciferase activity translated from TLucT transcripts harboring indicated mutations at the 3′ end and in the four base bulge in SL-II.

To determine the biological relevance of these mutations, we subcloned them into full-length TNV-D genomic RNA and observed their effects on viral RNA accumulation in tobacco NT-1 protoplasts (Fig. 2C). Because translation is necessary for production of viral replicase, mutations that reduce translation should reduce replication. The pyrimidine tract deletion (Δ3762–3738) decreased the replication of TNV-D to an undetectable level. Stem I mutations (C3753G and G3744C) and Loop I mutation (LI-mut) dramatically reduced accumulation of TNV-D RNA. Remarkably, the compensatory mutation (SI-re) that restored SL-I, resulted in a massive increase in TNV-D RNA accumulation, many fold above that of wild-type TNV-D RNA. This is not due to a major change in stability of SL-I (wt ΔG = −9.8 kcal/mol, SI-re ΔG = −9.6 kcal/mol). Thus, RNA replication was more sensitive to alterations in the 3′ UTR than was translation. But, as expected, all mutations that reduced translation had a negative impact on RNA accumulation.

We next examined the effects of disrupting and restoring SL-II (Fig. 3A). Mutations in Loop II (LII-mut) and those that weakened or disrupted Stem II (G3697C and C3704G) reduced luciferase accumulation in protoplasts by 2 to 3-fold (Fig. 3B). The double mutant containing G3697C and C3704G, predicted to re-form the SL-II helix (SII-re), restored translation to the wild-type level. Addition of a poly(A) tail caused all mutants to translate at the level of polyadenylated TLucT RNA, which is about double that of nonpolyadenylated TLucT. LII-mut and Stem II mutation G3697C abolished the replication of TNV-D. Stem II mutation C3704G significantly decreased the replication of TNV-D. SII-re restored the accumulation of TNV-D to a level higher than wild type. Thus, as for the mutations in SL-I, effects of these mutations on replication were more extreme than on translation (Fig. 3C). In summary, the distal SL-II helix and the primary sequence of L-II are important for the function of the TNV-D PITE, as well as for TNV-D replication.

The upper bulge in SL-II, GGGU, can potentially base pair to the 3′ terminal four bases, ACCC, of the TNV genome (boxed sequence connected by dashed arrow, Fig. 3A). A similar interaction, but with five base pairs, is required for the Tomato bushy stunt tombusvirus (TBSV) and Turnip crinkle virus (TCV) replication silencer to down-regulate negative strand RNA synthesis (Pogany et al., 2003; Zhang et al., 2004a). We tested the effect of disrupting this interaction on translation by introducing point mutations in either the GGGU or ACCC sequence or both (Fig. 3D). (The predicted G-C pairs were not altered because the terminal three C’s comprise half of the Sma I site required for plasmid linearization prior to in vitro transcription.) In all cases, translation was reduced by only 36% to 44% in the disruptive mutants and remained at a similar lower level in the double mutant predicted to restore the base pairing (Fig. 3D). Thus, this pseudoknot interaction does not play a major role in poly(A) tail-independent translation in uninfected cells.

Mutations have little effect on stability of mRNA

The expression differences of the mutants could be caused by changes in translation efficiency and/or RNA stability. To distinguish between these possibilities, we assessed the functional stability of the lowest expressing mutant RNAs. Functional half-life, the time in which the rate of reporter protein accumulation halves, was determined. Time-course analyses showed that, while total luciferase accumulation from actively translated RNAs varied, all mutants had a similar functional half-life (Fig. 4). Thus, the mutations that reduced luciferase accumulation did not significantly affect the stability of the reporter mRNA. We conclude that the mutations that reduced luciferase expression did so by reducing the translation efficiency of the mRNA.

Fig. 4.

Functional stability assay of reporter constructs. A. Time course of luciferase activity accumulation from TLucT and selected mutant RNAs at indicated times after electroporation into oat protoplasts. Data represent the mean from three independent experiments. C. mRNA functional half-live calculated from data in panel A (see Materials and Methods).

Full length TNV-D 3′ UTR, but not the double stem-loop structure alone, is sufficient to replace the poly(A) tail

Having established that stem-loop I and at least the upper portion of SL-II contribute to poly(A) tail-independent translation, we next tested whether the 3′-terminal sequence harboring these two stem-loops and extending to the genomic 3′ end is sufficient to replace a poly(A) tail. Because the BTE is absent in most of these constructs, all constructs in this experiment were capped. We tested a luciferase reporter construct, VLucT122, in which the 5′UTR is from vector sequence and the 3′ UTR consists of the 3′-terminal 122 nt of TNV-D RNA (nts 3641–3762), which includes the double stem-loop sequence discussed above (V indicates vector-derived sequence, T indicates TNV-D sequence). We compared translation efficiencies of VLucT122, with and without a poly(A) tail, to those of constructs that harbored only vector-derived UTRs, VLucVAn and VLucV294, in oat protoplasts. The positive control, VLucVAn, has a 67 nt vector-derived 3′ UTR followed by a 60 nt poly(A), whereas VLucV294 has a 294 nt vector-derived sequence as its 3′UTR. VLucT122 and VLucV294 translated about 7% as efficiently as VLucVAn (Fig 5). Poor translation of VLucV294, shows that a long, non-poly(A) sequence is insufficient to stimulate poly(A) tail-independent translation, in contrast to a previous report that any long 3′ UTR enhances translation in vivo (Tanguay and Gallie, 1996). Addition of a poly(A) tail increased the translation of VLucT122 to the level of VLucVAn (Fig. 5, VLucVAn, VLucV294, and VLucT122). Thus, the TNV-D 122 nt sequence containing the 3′ terminus of TNV-D RNA, including the two stem-loops is not sufficient to replace a poly(A) tail.

Fig. 5.

Test of the ability of sequences from the TNV-D 3′UTR to replace a poly(A) tail. Left side: maps of constructs beside their names on the left. The sizes and sequence of TNV RNA fragments added to 3′ UTRs of reporter constructs are indicated. Right side: relative luciferase activity obtained from translation of each construct in oat protoplasts. Relative luciferase activities were normalized to the capped, polyadenylated construct with vector-derived UTRs and a 60 nt poly(A) tail (VLucVAn). Luciferase assays were performed in at least three independent experiments, each of which was in triplicate. Standard deviations are indicated.

The inability of the 122 nt sequence to replace a poly(A) tail in the above construct may be due to absence of the TNV-D 5′ UTR or additional upstream 3′ UTR sequence. To test these possibilities, we first replaced the 5′ UTR of VLucT122 and VLucV294 with the 5′ UTR of TNV-D to create constructs TLucT122 and TLucV294 and examined their translatability. No statistically significant difference between luciferase activity from constructs containing the vector 5′ UTR or the TNV-D 5′ UTR was detected (Fig. 5). Thus, the viral 5′ UTR in combination with the 3′-terminal 122 nt was insufficient to confer poly(A) tail-independent translation. Secondly, we replaced the 122 nt 3′ UTR of TLucT122 with the 171 nt closest to the TNV-D RNA 3′ end (nts 3592–3762) to construct TLucT171. We also constructed TLucTBF which is the same as TLucT but with a GAUC duplication in a Bam HI site that destroys the cap-independent translation function of the BTE (Shen and Miller, 2004b). The translation of TLucT171 was 4.2-fold higher than that of TLucT122, 3.2-fold higher than that of TLucV294, and was about 40% as efficiently translated as VLucVAn. Thus, the 3′-proximal 171 nt of TNV-D RNA contain significant but incomplete poly(A) tail-independent translation activity. Non-polyadenylated TLucTBF translated even more efficiently than the polyadenylated control, VLucVAn (Fig. 5). Thus, the full-length 3′ UTR of TNV-D can functionally replace a poly(A) tail and PITE is separable from the BTE.

Conserved secondary structure at the 3′ ends of necrovirus genomes

To begin to understand the structural basis of the above observations, we examined the secondary structure of the entire translational control region of the 3′ UTR. We found MFOLD predictions of secondary structure shared by all necroviruses. Stem-loops I and II and the potential pairing of the four terminal bases to a bulge in SL-II are phylogenetically conserved (Fig. 6). This provides strong support of the experimental data that the structures exist and serve an important function. Double stem-loop structures at the 3′ termini of all necroviruses can be grouped into two classes: TNV-D-like and TNV-A-like. In the TNV-D-like viruses (TNV-D, Leek white stripe and Beet black scorch), the bulge in stem-loop II that base pairs to the 3′ terminus consists only of the four bases complementary to the 3′ end, and there are no predicted single stranded bases on the opposite strand. In TNV-A-like necroviruses, TNV-A, Olive mild mosaic virus (OMMV) and Olive latent virus-1 (OLV-1), this bulge may be larger than for the other necroviruses as there are unpaired bases on both sides of the bulge, but stem-loop II may contain non-Watson-Crick pairing of a G.A/A.G tandem (Fig. 6) (Na and White, 2006). Also the four terminal bases involved in the pseudoknot interaction of TNV-D-like necroviruses consists of RCCC (R=purine), whereas it is CCCC in the TNV-A-like viruses, with an extra 3′-terminal A at the 3′ end of OLV-1 RNA.

Fig. 6.

Secondary structures of 3′ ends of TNV-D and other necrovirus genomes predicted using mfold. Circled region in TNV-D RNA represents the BTE that is fully functional in vitro. Arrows indicate the 3′-most boundary of sequence required for cap-independent translation in vivo (nt 3680), and the minimal region required for poly(A) tail-independent translation (upstream of BTE to downstream of nt 3748). Dots connect possible G.A/A.G tandem pairs in TNV-A, OMMV and OLV-1 genomes. Dashed lines indicate predicted, phylogenetically conserved pairing of the 3′ terminus with the SL-II bulge not demonstrated here, but known to affect replication in other Tombusviridae. Genbank accession numbers of sequences: Tobacco necrosis virus strain D, UK isolate (D_00942), Leek white stripe virus (NC_001822), Beet black scorch virus (NC_004452), Tobacco necrosis virus strain A (NC_001777), Olive mild mosaic virus (NC_006939), Olive latent virus-1 (NC_001721).

It is noteworthy that sequence upstream of SL-II needed for full poly(A) tail-independent translation forms a long bulged, branched stem-loop structure that includes the BTE. In TNV-D RNA, bases 3664–3679 that are downstream of the minimal in vitro-defined BTE (circled in Fig. 6), which ends at nt 3663 (Shen and Miller, 2004b), and are upstream of stem-loop II, are predicted to base pair to sequence upstream of the BTE to extend helix IV of the BTE. This extended stem-IV of the BTE is conserved among all necroviruses (BTE S-IV, Fig. 6).

Discussion

The sequence that substitutes for a poly(A) tail differs from the BTE

Here we report that the two 3′-proximal stem-loops and upstream sequence of TNV-D RNA 3′UTR are necessary for the 3′ end to functionally replace a poly(A) tail. Translation of mRNAs with 3′ deletions up to nt 3720 was restored fully by adding only a poly(A) tail, indicating that this region is unnecessary for cap-independent translation. 3′ truncations to nts 3700 or 3680 were somewhat puzzling because translation remained low even in the presence of a cap and a poly(A) tail. However it is clear that the region lacks a cap-mimic sequence because addition of a 5′ cap has no stimulatory effect. Either this region contains a sequence that counteracts a negative element further upstream, or the deletion causes the RNA to misfold to the extent that translation of even capped and polyadenylated transcripts is low.

Sequence between nts 3661 and 3680 is required for cap-independent translation in vivo. The cap and poly(A) tail synergistically enhance translation of the RNA truncated at nt 3661 (Fig. 1B, D3661). Thus 3′ boundary of in vivo-defined TNV BTE is between nts 3662–3680, compared to the in vitro-defined 3′ boundary at nt 3659 (Shen and Miller, 2004b). The full-length 3′ UTR of TNV-D is not only sufficient to replace a poly(A) tail, it allows 70% higher translation than a 60 base poly(A) tail, and increases translation more than 27-fold compared to a 294 nt vector-derived 3′ UTR sequence (Fig. 5). In summary, it is clear that the BTE and the PITE have different sequences, but the full PITE overlaps with the BTE.

The “boundaries of these elements are difficult to define because the deletions used to map them could alter important secondary structures in unpredictable ways. Nevertheless, these results reveal two important properties of the 3′ UTR. First, it harbors both a PITE and a cap-independent translation element sequence on a naturally uncapped, nonpolyadenylated mRNA. Many 3′ UTR elements have been identified that confer poly(A) tail-independent translation on capped viral RNAs (Chiu et al., 2005; Leathers et al., 1993; Neeleman et al., 2001; Piron et al., 1998), and many cap-independent translation elements have been characterized on polyadenylated mRNAs (Hellen and Sarnow, 2001; Kneller et al., 2006), but to our knowledge, this is the first poly(A) tail-independent translation element that has been characterized on an RNA that lacks both a cap and a poly(A) tail. Secondly, stem-loops required for the TNV-D PITE are also known to be required for polymerase recognition and initiation of minus strand synthesis in related viruses (Fabian et al., 2003; Panaviene et al., 2005; Stupina and Simon, 1997; Zhang et al., 2004a). Thus, translational control signals overlap with replication signals.

The location of both the cap- and poly(A) tail-substitute sequences in the 3′ UTR poses interesting question about RNA circularization. If translation factors and the 40S ribosomal subunit bind these elements, one might predict “circularization” within the 3′ UTR itself (i.e. two elements in the 3′ UTR interacting directly or indirectly with eIF4G). The complex would then require the BTE-mediated long-distance base pairing with the 5′ UTR to deliver the translation factors to the 5′ end. Alternatively to the above model, the PITE may interact with the 5′ UTR directly or via host proteins, to enhance the interaction with the 5′ UTR mediated by the BTE.

Conserved secondary structure of 3′ end containing the cap-independent, and poly(A) tail-independent translation elements

The conserved 3′ UTR secondary structures in Fig. 6 reveal an extension of stem IV of the BTE (BTE S-IV) that is longer than that on the minimal BTE necessary for function in vitro. Interestingly, it is the extra sequence needed for cap-independent translation in vivo (nts 3662–3680) that confers the extension to BTE S-IV. This extended helical region ends just two bases upstream of 3′ SL-II. Because helices have a tendency to stack coaxially, we drew the 3′ end differently than in the “standard” format for Tombusviridae (Fig 3A). Drawn as in Fig. 6 also allows placement of the 3′-terminal four bases in proximity with their conserved complementary bases in the bulge of 3′ SL II. These base pairs are shown as dashed lines, because they do not seem to contribute substantially to translation. Instead, this base pairing plays a role in replication, based on experimental results and phylogenetic comparisons of many other viruses in the Tombusviridae (Na and White, 2006; Pogany et al., 2003; Zhang et al., 2004a; Zhang et al., 2004b). The asymmetry of the bulge is expected to form a kink in the 3′ SL-II helix. Base pairing of the four terminal bases may reinforce the kink and stack coaxially with either the top or bottom half of the SL-II helix. The terminal base pairing may stack coaxially with (and between) both SL-I and the top of SL-II. This was proposed for tombusviruses (Na and White, 2006), but necroviruses lack unpaired bases in the SL-II that may be needed to allow such a structure to form.

TNV-A, OMMV and OLV-1 differ slightly from the other necroviruses in that there are unpaired bases on both sides of the bulge in 3′ SL-II, and the four terminal bases to which this bulge pairs are all cytosines. Na and White proposed that the G.A/A.G tandem in this bulge may base pair, extending SL-II (Na and White, 2006). Sheared non-Watson-Crick base pairing in a G.A/A.G tandem occurs, for example, in the cis-acting element required for incorporation of the amino acid selenocysteine during translation of selenoproteins (Walczak et al., 1998).

The pyrimidine-rich tract may be a spacer to provide length to allow the 3′ terminal four bases to reach the SL-II bulge. A sequence this long (15–16 nt) between the terminal stem-loops is present in all Tombusviridae RNAs, except that it is interrupted by a short stem-loop in genus Tombusvirus. and it is not pyrimidine-rich in other Tombusviridae.

Presence of a cap-independent translation element immediately upstream of the 3′ terminal structure needed for replication resembles the organization of other Tombusviridae, including dianthoviruses which also harbor a BTE (Mizumoto et al., 2003), as well as Tombusviridae that harbor non-BTE cap-independent translation elements, such as TBSV (Fabian and White, 2006), TCV (Qu and Morris, 2000), Panicum mosaic virus (Batten et al., 2006), and Maize necrotic streak virus (Scheets and Redinbaugh, 2006). These all differ from BYDV in which the BTE is located 800 nt upstream from the 3′ end, even though the 3′ structures required for replication bear some resemblence to those of the Tombusviridae (Koev et al., 2002).

The poly(A) tail overlaps with replication signals

The primary and secondary structural requirements for replication of viral RNAs in the Tombusviridae are well-characterized. Stem-loops of similar size and position as 3′ SL-I and 3′ SL-II, referred to as Pr and H5, respectively, in the carmoviruses, and as gPR and SL3, respectively, in the tombusviruses have been mutated exhaustively (Fabian et al., 2003; Pogany et al., 2003; Zhang et al., 2004a; Zhang and Simon, 2005). Stem-loop I is required for minus strand synthesis, and facilitates replicase assembly (Panviene et al., 2005). Also, the pseudoknot base pairing between the 3′ end and the SL-II bulge is predicted in eight of the nine Tombusviridae genera (Na and White, 2006). In this interaction there are usually five, rather than the four base pairs predicted in the necroviruses. In the above viruses, the short helix formed by the 3′ end-SL-II bulge base pairing is expected to stack with the adjacent helical domain in SL-II. A similar coaxially-stacked helix interaction occurs at the BYDV RNA 3′ end (Koev et al., 2002) which is a Tombusviridae-like luteovirus (Miller et al.,, 2002). In all cases, this interaction is predicted to make the 3′ end inaccessible to proteins such as the viral replicase. Consistent with this, the 3′ end-S-L II bulge base pairing attenuates minus strand synthesis (Pogany et al., 2003; Zhang et al., 2004a). This would allow preferential synthesis of plus strands which is the goal of these positive strand RNA viruses.

The role of the above elements in translation had not been investigated for Tombusviridae. The switch above may also serve as a switch between minus strand synthesis and translation. We show that the base pairing of the terminal four bases to the bulge does not appear to enhance or facilitate translation directly (Fig. 3). Due to the requirement for a 3′-terminal CCC for in vitro transcription, we could change only the fourth base from the 3′ end, A to U. It is possible that the three C-G pairs are sufficient. The lack of any significant change in activity when the disrupted U-A pair was replaced by an A and a U, respectively, suggests that this terminal pairing plays little role in translation in unifnected cells (Fig. 3D). Alternatively, an entirely different and unpredicted structure, rather than the restored helix, may form in the presence of the A and U substitutions expected to replace the U-A pair. Even if this were the case, the modest decrease in translation in all of the mutants at this site (>50% of wild type, Fig. 3D) do not suggest a major role in translation for this helix. Importantly, the role in translation of this base pairing of the four terminal bases to the SL-III bulge, remains to be determined in infected cells. It is possible that by interacting with the unpaired conformation of the 3′ end, replicase sequesters the RNA away from the translational machinery. Thus, the base paired conformation of the four 3′-terminal bases would enhance translation efficiency only in infected cells.

Control of translation and replication by the 3′ UTR has been investigated in other viruses. The tRNA-liked structures (TLS) of the Bromoviridae, Tymoviruses, Tobamoviruses facilitates both translation in the absence of a poly(A) tail (Gallie and Kobayashi, 1994; Matsuda and Dreher, 2004), and serves as the key replicase recognition site for minus strand synthesis. Dreher and colleagues showed that eIF1A binding blocked replicase access to the initiation site (Matsuda et al., 2004). This provides the same function as the molecular switch proposed above to favor either translation or plus strand synthesis. Similarly, the 3′ terminus of AMV RNA consists of a series of stem-loops which can refold into a pseudoknotted structure that resembles a TLS. The AMV coat protein binds the stem-loop series conformation, while the pseudoknotted version is proposed to serve as the active minus strand template (Vlot et al., 2001).

Advantage of viral RNA with a poly(A) tail-independent element instead of a poly(A) tail

Why have viral RNAs evolved a poly(A) tail-independent translation element instead of having a poly(A)? By lacking a poly(A) tail, nonpolyadenylated viral RNAs may avoid host translational regulation mechanisms. By obviating the need for PAPB, a PITE evades competition with cellular mRNAs for PAPB. For example, the Rotavirus NSP3 protein competes with PABP for binding eIF4F (Groft and Burley, 2002), shutting off translation of cellular polyadenylated mRNAs in favor of viral translation. Secondly, poly(A) tail-independent translation may help the viral RNA bypass the host’s antiviral defenses that shut down translation via negatively regulating translation factor activity. Thirdly, absence of a poly(A) tail allows the viral RNA to harbor key RNA replication control elements at the extreme 3′ terminus where minus strand synthesis initiates. It is possible that absence of a poly(A) tail provides no advantage for translation per se, but instead provides the replicative advantage of switchable replication control signals at the extreme 3′ terminus.

Materials and Methods

Plasmids and RNA constructs

All clones were verified by automated sequencing at the Iowa State University DNA Sequencing and Synthesis Facility. Plasmid pTNV-D is a full-length infectious clone of TNV-D, kindly provided by R. H. A. Coutts, Imperial College, London (Coutts et al., 1991). pTLucT is the template for TLucT, which has a firefly luciferase ORF as a reporter flanked by the 5′ and 3′ UTRs of TNV-D (Shen and Miller, 2004b). D3720, D3700, D3680, and d3726–3738 were constructed by replacing the 3′ UTR of TLucT with the respective shortened 3′ UTR of TNV-D generated by PCR. In D3720, nts 3721–3744 were deleted. In D3700, nts 3701–3759 were deleted. Nts 3681–3759 were deleted in D3680. D3748 and D3661 were truncations of TLucT at Bgl II₃₇₄₅ and Ssp I₃₆₅₉, respectively.

Mutants within the stem-loops were constructed by using standard PCR-mediated, site-directed mutagenesis as in (Guo et al., 2000; Guo et al., 2001). VLucVAn was described in Guo et al. (Guo et al., 2000), in which a 60 base poly(A) tail was inserted into the Stu I/Sal I site of pGEM-luc (Promega, Madison, WI) (Guo et al., 2000). VLucT122 was constructed by replacing the 3′ UTR of VLucVAn with a 122 nt sequence from TNV-D 3′ UTR (nts 3641–3762), which includes the PAM sequence. V indicates sequence from vector, and T indicates sequence from TNV-D. The template for in vitro transcription of VLucV294 was pGEMLUC linearized with Ssp I. TLucT122 and TLucV294 were constructed by replacing the 5′ UTR of VLucT122 and VLucV294 with the 5′ UTR of TNV-D. In all constructs with a vector-derived 5′ UTR, the 5′ UTR sequence is the complement of nts 1755–1803 in pGEM-luc (Genbank accession no. X65316). The 3′ UTR of VlucVAn is the complement of pGEM-luc nts 45–101, and the 3′ UTR of TLucV294 and VlucV294 is the comlement of nts 4739–101. TLucT171 was constructed by replacing the 3′ UTR of TLucT122 with the 171 nt (nts 3592–3762) sequence from the 3′ UTR of TNV-D. TLucTBF has the full length 3′ UTR of TNV-D with a GUAC duplication in a BamH I site. The cap-independent translation function of TE is destroyed by this duplication (Shen and Miller, 2004b).

In vitro transcription

Capped and uncapped RNAs were synthesized by in vitro transcription using the T7 mMESSAGE mMACHINE® and MegaScript® kits (Ambion, Austin, TX) as per manufacturer’s instructions, respectively. Templates for RNAs with a poly(A) tail were linerized with Vsp I. Template for VLucV294 was pGEMLUC linearized with Ssp I. Templates for D3661 and D3748 were pTLucT linearized with Ssp I and Bgl II, repectively. All other templates were digested with Sma I.

In vivo translation

Oat (Avena sativa cv. Stout) protoplasts were prepared and electroporated with RNA as described in (Dinesh-Kumar and Miller, 1993). Luciferase assays were done as in Shen and Miller (Shen and Miller, 2004b). We included a capped and polyadenylated renilla luciferase reporter as an internal control, and the Promega Stop-N-Glo^™ (Madison, WI) system was used to assay both luciferase activities. All luciferase assays were performed in triplicates in at least three independent experiments. Firefly luciferase activities were first normalized with renilla luciferase activity to minimize variation between samples. The luciferase activities of all constructs were then compared to TLucT, whose luciferase activity is defined as 100%.

Northern blot hybridization

For TNV-D replication assays, NT-1 protoplasts were used and incubated for 24 hr after electroporation. Total RNAs were extracted from these cells by using the Trizol reagent (Invitrogen, Carlsbad, CA) as per manufacturer’s instructions and analyzed by Northern blot analysis as described previously (Shen and Miller, 2004a). A ³²P-labled probe, complementary to the 107 nt TE of TNV-D, was used to detect TNV-D gRNA and sgRNAs.

Stability assay

Functional stability assays were done as described in (Danthinne et al., 1993). Protein accumulation (A) as a function of time (t) was analyzed by using the first order kinetics equation:

$$\text{A}(\text{t})={\text{A}}_0{\text{e}}^{-\text{kt}}.$$

A function y = a · ln(t) + b was achieved from the logarithmic trend line of curve protein accumulation (A) vs. time (t) by using Microsoft Excel. Constant k was calculated by giving an arbitrary time t = 60 min and the function. We then calculated the functional half time t _1/2 = ln(1/2)* (1/k). Data shown was the mean from three independent experiment.