A viral sequence in the 3′‐untranslated region mimics a 5′ cap in facilitating translation of uncapped mRNA

Shanping Wang, Karen S. Browning, W.Allen Miller

Author Affiliations

  1. Shanping Wang1,2,
  2. Karen S. Browning3 and
  3. W.Allen Miller*,1,4
  1. 1 Molecular, Cellular and Developmental Biology Program, 351 Bessey Hall, Iowa State University, Ames, IA, 50011, USA
  2. 2 Present address: Laboratory of Tumor Biology, Massachusetts General Hospital Cancer Center, Building 149, Harvard University, Charlestown, MA, 02129, USA
  3. 3 Department of Chemistry and Biochemistry, University of Texas, Austin, TX, 78712, USA
  4. 4 Departments of Plant Pathology, and Biochemistry & Biophysics, 351 Bessey Hall, Iowa State University, Ames, IA, 50011, USA
  1. *Corresponding author. E-mail: wamiller{at}
View Full Text


For recognition by the translational machinery, most eukaryotic cellular mRNAs have a 5′ cap structure [e.g. m7G(5′)ppp(5′)N]. We describe a translation enhancer sequence (3′TE) located in the 3′‐untranslated region (UTR) of the genome of the PAV barley yellow dwarf virus (BYDV‐PAV) which stimulates translation from uncapped mRNA by 30‐ to 100‐fold in vitro and in vivo to a level equal to that of efficient capped mRNAs. A four base duplication within the 3′TE destroyed the stimulatory activity. Efficient translation was recovered by addition of a 5′ cap to this mRNA. Translation of both uncapped mRNA containing the 3′TE in cis and capped mRNA lacking any BYDV‐PAV sequence was inhibited specifically by added 3′TE RNA in trans. This inhibition was reversed by adding initiation factor 4F (eIF4F), suggesting that the 3′TE, like the 5′ cap, mediates eIF4F‐dependent translation initiation. The BYDV‐PAV 5′UTR was necessary for the 3′TE to function, except when the 3′TE itself was moved to the 5′UTR. Thus, the 3′TE is sufficient for recruiting the translation factors and ribosomes, while the viral 5′UTR may serve only for the long distance 3′–5′ communication. Models are proposed to explain this novel mechanism of cap‐independent translation initiation facilitated by the 3′UTR.


Almost all eukaryotic cellular mRNAs contain a 5′ m7G(5′)ppp(5′)N cap structure which is required for efficient initiation of translation. According to the ribosome scanning model of eukaryotic translation initiation, initiation factor eIF4F specifically recognizes the 5′ cap structure and, with the help of other initiation factors such as eIF3, recruits the 43S ribosomal subunit initiation complex that then scans 5′ to 3′ along the mRNA (Kozak, 1989; Merrick and Hershey, 1996). When the first (or second in the case of leaky scanning) AUG codon is reached, the 60S ribosomal subunit joins and peptide elongation ensues.

Although the ribosome scanning model explains the mechanisms of various translational regulatory elements in the 5′‐untranslated regions (UTRs) of mRNAs, numerous examples exist of translational control elements in the 3′UTRs of mRNAs. The 3′ poly(A) tail, found on most eukaryotic cellular mRNAs, stimulates translation initiation and stabilizes mRNA (Jacobson, 1996). The 5′ cap and poly(A) tail act synergistically to stimulate initiation in vivo (Gallie, 1991) and in a yeast in vitro translation system (Tarun and Sachs, 1995). In the case of viral RNAs that lack a poly(A) tail, sequences in the 3′UTR, such as the pseudoknot‐rich domain of tobacco mosaic virus (TMV), appear to substitute functionally for a poly(A) tail to stimulate translation in conjunction with the 5′ cap (Gallie and Walbot, 1990; Gallie, 1991). Other cis‐acting elements in 3′UTRs control translation initiation either by modulating poly(A) tail length during embryo development (Sheets et al., 1995) or via binding of a specific regulatory protein that inhibits initiation (Standart and Jackson, 1994; Curtis et al., 1995; Dubnau and Struhl, 1996). In all these examples, the mRNA must have a 5′ cap in order for the 3′ element to function. Thus, in previously described mRNAs, cap recognition is likely to be an essential component in the communication between 3′ and 5′ ends (Tarun and Sachs, 1995; Hentze, 1997).

Mammalian eIF4F complex is comprised of eIF4E (the cap‐binding subunit), eIF4G (p220) and the more loosely associated eIF4A (helicase) (Merrick, 1994). Plant cells contain two forms of the 4F complex: eIF4F, consisting of 26 and 220 kDa subunits (homologous to eIF4E and eIF4G, respectively), and eIFiso4F, consisting of 28 and 86 kDa subunits (Browning, 1996). eIF4A probably has the same function in plants but does not co‐purify with eIF4F, so it is not considered a subunit of this complex (Browning, 1996). In defined cell‐free systems, capped mRNAs have a reduced requirement for eIF4F compared with uncapped mRNAs (Fletcher et al., 1990).

Many viral mRNAs lack a 5′ cap or a poly(A) tail, or both. They have evolved ways of ensuring efficient translation of their genes from uncapped mRNAs (Carrington and Freed, 1990; Tsukiyamakohara et al., 1992; Sarnow, 1995), often at the expense of the host cell. The most well‐documented example is the translation of picornavirus RNAs (Jackson and Kaminski, 1995). Picornaviral RNAs lack a 5′ cap structure and have an extremely long, highly structured 5′UTR, including many AUG codons upstream of the start codon of the main open reading frame (ORF). Rather than scanning from the 5′ end, ribosomes bind internally in this long leader at the internal ribosomal entry site (IRES) just upstream of the start codon, in a cap‐independent manner (Pelletier and Sonenberg, 1988; Sarnow, 1995). Although IRES‐mediated cap‐independent translation does not conform to the rule of 5′ cap–eIF4E recognition, it still employs the other canonical initiation factors including the eIF4G subunit of eIF4F (Pestova et al., 1996), and follows the scanning concept of ribosome binding to the 5′UTR followed by scanning in a 3′ direction until the appropriate start codon is reached for initiation of translation.

In contrast to the above examples, we showed previously that the genomic RNA of the PAV barley yellow dwarf virus (BYDV‐PAV) harbors a sequence in the 3′UTR that confers cap‐independent translation initiation at the 5′ AUG of the mRNA in wheat germ extracts (Wang and Miller, 1995). The BYDVs are ubiquitous, economically important viruses of small grains (D'Arcy and Burnett, 1995). BYDV‐PAV is a subgroup I luteovirus, whereas some other BYDVs are members of the very divergent subgroup II (Miller et al., 1995; Miller and Rasochova, 1997). The mechanism of translation initiation of BYDV‐PAV is remarkable because the 3′ translation enhancer (3′TE) that confers cap‐independent translation is located 5 kb downstream, and is separated by several large untranslated ORFs, from the initiation codon. Furthermore, unlike other translational control elements in the 3′UTR, such as poly(A) tails, the 3′TE controls initiation at the 5′ AUG in the absence of a 5′ cap. The only other mRNA known to resemble this, in function, is satellite tobacco necrosis virus (STNV) RNA (Danthinne et al., 1993; Timmer et al., 1993). This RNA differs from BYDV‐PAV in that its 3′ stimulatory element is located immediately 3′ of the only ORF on the RNA, and 600 nucleotides downstream of the initiator AUG. The stimulatory elements of BYDV‐PAV and STNV RNAs share no obvious sequence homology. For both RNAs, cap‐independent translation has been reported only in cell‐free extracts. Like capped mRNAs, STNV RNA translation depends on eIF4F or eIFiso4F, but the amount of factor needed for maximal translation is one‐tenth of that needed for capped mRNAs (Timmer et al., 1993).

Here, we (i) localize the 3′TE to a smaller region in the BYDV‐PAV genome than reported previously, (ii) show that it functions in vivo but requires more of the 3′UTR than in wheat germ extracts, (iii) provide evidence that the 3′TE facilitates initiation in an eIF4F‐dependent manner and (iv) show that the translation initiation and 3′–5′ communication functions can be uncoupled. These data shed light on possible new mechanisms for a 3′UTR in mediating translation initiation.


Localization of minimal 3′TE sequence

Previously, we showed that a 500 base region in the genomic RNA of BYDV‐PAV facilitated translation of uncapped mRNA in vitro (Wang and Miller, 1995). Here, we show the ability of a smaller portion of this region to stimulate translation of uncapped mRNA. The mRNAs are in vitro transcripts containing the Escherichia coli uidA (GUS) reporter gene flanked by various 5′ and 3′UTRs (Figure 1A). Uncapped GUS mRNA that contains the 109 nucleotide sequence spanning the intergenic region between BYDV‐PAV ORFs 5 and 6 in its 3′UTR (bases 4817–4925 of the BYDV‐PAV genome) was translated with the same efficiency in vitro as transcripts that contain the larger (500 nt) 3′TE sequence (Figure 1B, lanes 3–5). This construct yielded >50‐fold more GUS protein than transcripts lacking the intercistronic sequence (Figure 1B, compare lanes 2 and 3). Deletion of bases upstream of nucleotide 4837 (Wang and Miller, 1995) or downstream of nucleotide 4873 (data not shown) within the 3′TE abolished stimulatory activity. Therefore, a subset (bases 4817–4925) of the previously reported 500 base fragment is sufficient to confer translation enhancement of uncapped mRNA in a wheat germ extract. We refer to this 109 nucleotide sequence as the in vitro‐defined, or 109 nt, 3′TE.

Figure 1.

(A) Maps of transcripts. The genome organization of BYDV‐PAV RNA with numbered ORFs (Miller et al., 1988) is shown at the top. A bold black line beneath the ORFs (numbered boxes) indicates the 5677 nt genomic RNA. Maps below the genome depict transcripts coding for GUS (ORF not to scale) containing viral sequence (bold lines), vector or Ω sequence (thin lines) or poly(A) tails (A30) in their UTRs. The 109 nt 3′TE is the intergenic region (between dashed lines) between ORFs 5 and 6. BF transcripts are identical to those shown, except that they contain the four base duplication at the filled and re‐ligated BamHI site (B) within the 109 nt 3′TE. Abbreviations: Sc, ScaI4513; B, BamHI4837; P, PstI5009; Sm, SmaI5677 (numbered as in Miller et al., 1988); R1, EcoRI; ICR1, EcoICRI; X1, XbaI. (B and C) Wheat germ translation products of the indicated transcripts. Each uncapped mRNA (0.2 pmol) was translated and their products analyzed electrophoretically as in Wang and Miller (1995). The relative radioactivity in the GUS product, as determined with a Phosphorimager, is indicated below each lane. (B) Molecular weights (in kDa) of translation products of brome mosaic virus (BMV) RNA (lane 1) and GUS (68 kDa) are at the left. The translation efficiency in lane 5 was defined arbitrarily as 100%. Templates used in lanes 2–5 were uncapped RNAs prepared by run‐off transcription from plasmids linearized with the indicated restriction enzymes. (C) All transcripts were from EcoRI‐cut plasmids, except for lanes 1 and 2, which were from EcoICRI‐cut plasmid. Transcripts with names ending in A+ are polyadenylated. C, capped transcript; U, uncapped transcript.

To confirm the specificity of the stimulation by the 3′TE, a mutation was introduced by inserting a four base duplication (GAUC) within the BamHI4837 site in the 3′TE (PGUS109BF; all clones with the BF designation contain this BamHI fill‐in mutation). This duplication abolished the stimulatory activity of the 3′TE (Figure 1C, compare lanes 4 and 6). Addition of a 5′ cap rescued translation of these mutant transcripts to the level observed for uncapped, 3′TE‐containing PGUS109 (Figure 1C, lanes 4 and 5). The presence of a poly(A) tail had little effect on translation of any transcripts in wheat germ extracts (Figure 1C), which is consistent with previously reported observations (Gallie, 1991; Wang and Miller, 1995).

Activity of the 3′TE in vivo

To assess the activity of the 3′TE in vivo, GUS‐encoding transcripts with various 5′ and 3′ UTRs were electroporated into oat protoplasts, and translational efficiency was measured by assaying for GUS activity. An uncapped control transcript, containing a plasmid‐derived 5′UTR and a 30 nt poly(A) tail yielded no significant GUS activity (Figure 2A, VecGUSA+). When capped, this mRNA expressed GUS at 50‐fold above background, consistent with the essential role for the 5′ cap. Replacement of the plasmid‐derived 5′UTR with that from BYDV‐PAV stimulated GUS expression from capped mRNA 3‐fold further (Figure 2A, PGUSA+). Addition of the 109 nt 3′TE was not sufficient for translation of mRNA in vivo (Figure 2A, compare PGUS109 and PGUS). The presence of a poly(A) tail in addition to the 109 nt 3′TE resulted in a 10‐fold increase above background in GUS expression from uncapped mRNA (Figure 2A, compare uncapped PGUS109A+ with uncapped PGUS109). However, addition of a 5′ cap to this transcript stimulated GUS expression by another order of magnitude (Figure 2, compare capped and uncapped PGUS109A+). Construct PGUS109BFA+, which contains the four base GAUC duplication, yielded no GUS activity when lacking a cap, but was fully active when capped (Figure 2A). These results show that the 109 nt 3′TE only partially, but significantly, facilitated cap‐independent translation in vivo.

Figure 2.

(A and B) GUS expression in protoplasts electroporated with transcripts containing various 5′ and 3′ UTRs (see Figure 1A). GUS activity is measured in nmoles (nm) of methylumbelliferone (MU) produced per μg of cellular protein per minute in a 2 h reaction. Transcripts containing a poly(A) tail (A+) are from EcoRI‐linearized plasmids; others are from plasmids linearized with SmaI. Transcript PGUS is from EcoICRI‐linearized pPGUS109. In (A), data represent averages (± SD) from three separate experiments, each of which was performed in duplicate. In (B), data represent averages from two separate experiments, each of which was performed in duplicate. (C) Wheat germ translation products of transcripts used in (B) were analyzed and quantified as in Figure 1. C, capped; U, uncapped transcript.

We tested the effect of additional viral sequence in the 3′UTR on 3′TE function in vivo. PAV bases 4515–5677 (the 3′‐terminal 1162 bases of the genome) were placed in the 3′UTR of the GUS reporter construct. Uncapped transcript from this plasmid gave activity >100 times background (Figure 2A and B, PGUS1162). Addition of a cap stimulated expression no more than 2‐fold (Figure 2A and B, compare capped and uncapped PGUS1162). The four base duplication in the BamHI4837 site abolished expression from uncapped transcript both in vitro (Figure 2C, lanes 2 and 4) and in vivo (Figure 2B, compare uncapped PGUS1162BF with uncapped PGUS1162), verifying the specificity of the 3′TE effect. Furthermore, the four base duplication had no effect on expression of the capped transcript (Figure 2B, compare capped PGUS1162BF with capped PGUS1162). Interestingly, the stimulatory activity of the 1162 nt 3′ end of BYDV‐PAV sequence did not require a poly(A) tail (Figure 2B, compare PGUS1162 and PGUS1162A+). Thus, sequence(s) different from the 3′TE functionally substitutes for a poly(A) tail. In addition, full stimulation of cap‐independent translation requires more viral sequence in vivo than in vitro. Because of this difference, rather than define the 3′TE as a precise sequence, we functionally define it as the sequence that stimulates cap‐independent translation from the 3′UTR and is rendered non‐functional by mutation at the BamHI site.

All the results above were obtained from RNA transcripts harboring the 5′UTR from BYDV‐PAV (indicated by the ‘P’ preceding ‘GUS’ in the transcript name). The requirement for the 5′UTR from BYDV‐PAV RNA was examined by replacing it with the 5′UTR (Ω sequence) of TMV RNA. This sequence stimulates translation in vitro (with or without a cap) and in vivo in a highly cap‐dependent manner (Gallie, 1991; Sleat and Wilson, 1992). Substitution of the BYDV‐PAV 5′UTR with Ω permitted a low rate of translation of uncapped mRNA in vitro (Figure 2C, lane 8). However, Ω in place of the BYDV‐PAV 5′UTR abolished translation of uncapped mRNA in vivo, even in the presence of the 1162 nt virus‐derived 3′UTR (Figure 2B, ΩGUS1162), but permitted efficient translation of capped ΩGUS1162 RNA. Thus, Ω facilitates translation in a cap‐dependent manner only, and does not cooperate with the 3′TE to promote cap‐independent translation. Previously, we showed that a vector‐derived 5′UTR also failed to support 3′TE activity in vitro (Wang and Miller, 1995). Thus, the 3′TE probably requires at least a portion of the BYDV‐PAV 5′UTR for cap‐independent translation in these contexts, indicating specific interactions between the 3′ and 5′UTR.

The fact that expression of the capped, poly(A)+, Ω‐containing mRNA which is considered to be an optimal plant message is no higher than that of uncapped mRNA containing the full BYDV‐PAV 5′ and 3′UTRs (Figure 2B, compare uncapped PGUS1162 and capped ΩGUS1162A+), demonstrates the very high levels of expression that are conferred by this cap‐independent translation signal.

The 3′TE mimics a 5′ cap in its effect on RNA stability

It is possible that the 3′TE stimulates gene expression in vivo, at least in part, by increasing RNA stability. However, we showed previously that the 109 nt 3′TE did not affect RNA stability in wheat germ extracts (Wang and Miller, 1995). To determine the effect of the larger, in vivo‐defined 3′TE on RNA stability in vivo, RNA transcripts with a wild‐type or mutant (the four base duplication) 1162 nt 3′UTR were electroporated into oat protoplasts under the same conditions as the GUS gene expression experiments, and RNA degradation was monitored by Northern blot hybridization. The RNA transcripts with either the wild‐type or the defective 3′TE did not differ significantly in degradation rate (Figure 3A).

Figure 3.

Relative stability of mRNAs electroporated in oat protoplasts. (A) Northern blot hybridization detecting uncapped transcripts isolated from oat protoplasts at the indicated times after electroporation. The blot was hybridized with 32P‐labeled antisense GUS transcript as described in Materials and methods. (B) Kinetics of GUS enzymatic activity accumulation at the indicated times after electroporation of oat protoplasts. The GUS assay was performed as in Figure 2.

Northern blot hybridization detects only the physical stability of total cellular GUS mRNAs, and cannot discriminate between translatable and untranslatable RNAs. As a more accurate assay of RNA stability, the functional RNA half‐life was measured by the kinetics of GUS synthesis. The activity of GUS (a very stable enzyme) should stop increasing sooner for a functionally unstable mRNA than for a stable one. Of the three mRNAs for which GUS activity was detectable, none leveled off until at least 30 h after electroporation (Figure 3B). This is beyond the 20 h timepoint used in Figure 2. Uncapped PGUS1162 mRNA appeared to be slightly less stable than its capped counterpart, but showed a very similar stability (shape of curve) to the capped P1162BF mutant. Thus, the capped mRNA with the defective 3′TE (PGUS1162BF) and uncapped mRNA containing active 3′TE (PGUS1162) have similar ‘functional’ stabilities, suggesting that the 3′TE either has no role in stabilizing mRNA or that it contributes a similar amount to stability as does the presence of a 5′ cap. The similar physical stability of uncapped PGUS1162 and uncapped PGUS1162BF (Figure 3A), but dramatic differences in GUS gene expression (Figure 3B), indicate that the 3′TE functions at the translational level.

The 3′TE decreases the requirement for eIF4F for efficient translation of uncapped mRNA

One possible role for the 3′TE in enhancing translation of uncapped mRNA is to efficiently recruit essential initiation factor(s) that facilitate(s) binding of the ribosomal small subunit, in a manner analogous to that of the 5′ cap of mRNAs. The binding of eIF4F to the 5′ cap structure confers the selective translation of capped, as opposed to uncapped, mRNA (Browning, 1996). The translation‐enhancing sequence from STNV RNA dramatically decreased the amount of the rate‐limiting initiation factor eIF4F, required for maximal translation efficiency in vitro (Timmer et al., 1993). The eIF4F requirements of mRNAs containing or lacking the 3′TE were compared. The already efficient translation of capped 3′TE‐lacking mRNA and of uncapped mRNA containing the 109 nt 3′TE was not stimulated by exogenous eIF4F (Figure 4, lanes 1–10). Consequently, eIF4F was not rate‐limiting for these mRNAs. In contrast, the endogenous eIF4F levels were rate‐limiting for uncapped mRNA with the defective 3′TE, because translation efficiency of this mRNA increased correspondingly with the addition of exogenous eIF4F (Figure 4, lanes 11–15). As a control, bovine serum albumin (BSA) was added instead of eIF4F, and was observed to have no effect on translation of any mRNAs (data not shown). Thus, the 109 nt 3′TE appears to reduce the amount of eIF4F required for efficient translation, a property that is normally conferred by a 5′ cap (Browning, 1996).

Figure 4.

Effect of added eIF4F on translation of capped and uncapped mRNAs containing or lacking the functional 109 nt 3′TE in wheat germ extract. PGUS transcript was prepared by in vitro transcription from pPGUS109 that had been linearized with EcoICRI, which removes the 3′TE. PGUS109 was prepared from the same plasmid linearized with EcoRI. Transcript PGUS109BF is from the BamHI‐filled mutant of pPGUS109 (pPGUS109BF), linearized with EcoRI. Translation reactions were performed as for Figure 1. Amount of added initiation factor: lanes 1, 6 and 11, 0; lanes 2, 7 and 12, 0.25 μg; lanes 3, 8 and 13, 0.5 μg; lanes 4, 9 and 14, 1.0 μg; lanes 5, 10 and 15, 2.0 μg. The translation efficiency of each mRNA with no exogenous translation factor was defined as 1.

The 109 nt 3′TE inhibits translation of mRNAs in trans by a mechanism that is reversed by exogenous eIF4F

If the 3′TE enhances translation by efficiently recruiting eIF4F, either directly or via an unidentified trans‐acting factor, we predict that 3′TE‐dependent or cap‐dependent translation would be inhibited by the 3′TE in trans. The free 3′TE RNA should act as a competitor for eIF4F or other factors that mediate 3′TE stimulation in cis. To test this, a 109 nt transcript comprising only the in vitro‐defined 3′TE RNA was added in excess to wheat germ extracts containing various mRNAs. Indeed, translation of uncapped 3′TE‐plus mRNA (PGUS109) was inhibited by the addition of 3′TE RNA (Figure 5A, lanes 1–4). Addition of the defective 3′TE, which differs only by having the GAUC duplication (3′TEBF), had no inhibitory effect (Figure 5A, lanes 5–7). This shows that the inhibition by wild‐type 109 nt 3′TE was not simply a non‐specific effect of adding RNA to the wheat germ extract. Addition of eIF4F to the reaction reversed the trans inhibition by the 3′TE (Figure 5A, lanes 8–12). Thus, either the 3′TE‐mediated cap‐independent translation requires eIF4F or the mechanism is normally eIF4F‐independent and added eIF4F allows translation of PGUS109 to bypass the 3′TE‐mediated mechanism. To distinguish between these possibilities, the effect of the 3′TE on translation of capped mRNA lacking any BYDV‐PAV sequence was examined. Excess added wild‐type 3′TE RNA inhibited translation of capped mRNA containing Ω as its 5′UTR (Figure 5B, lanes 1–5). Again, the defective 3′TEBF RNA had no effect on the translation of this capped mRNA (Figure 5B, lanes 6–9), and addition of eIF4F restored translation of ΩGUS mRNA in the presence of inhibitory levels of 3′TE (Figure 5B, lanes 10–14). Thus, the 109 nt 3′TE competes with capped mRNA for factor(s) required for cap‐dependent translation and does not need to interact with the BYDV‐PAV 5′UTR to do so. The restoration of translatability of both mRNAs in the presence of inhibitory levels of 3′TE by eIF4F strongly implicates a role, either direct or indirect, for eIF4F in cap‐independent translation mediated by the 3′TE in cis.

Figure 5.

Inhibition of translation by the 109 nt 3′TE in trans, and restoration by eIF4F. Uncapped PGUS109 (A) or capped ΩGUS (B) mRNAs (0.1 pmol) were translated in wheat germ extracts as in Figure 1, with the indicated amounts of added transcript comprising the 109 nt 3′TE (3′TE) or mutant 3′TE with the GAUC duplication at the BamHI site (3′TEBF) as competitor RNA. Amount of eIF4F added: (A) lanes 1–8, none; lanes 9–12, 0.5, 1.0, 2.0 and 4.0 μg, respectively; (B) lanes 1–10, none; lanes 11–14, 0.25, 0.5, 1.0 and 2.0 μg, respectively. The 109 nt 3′TE RNA and 113 nt 3′TEBF RNA were synthesized in uncapped form from SmaI‐linearized p3′TE and p3′TEBF, respectively (Figure 1A). Translation was performed in wheat germ extracts as described in Materials and methods.

The 109 nt 3′TE functions in the 5′UTR

So far, by all tests, the 3′TE functionally mimics a 5′ cap. If this is the case, like the cap, the 3′TE should function in the 5′UTR. Hence, the 109 nt 3′TE was moved to the 5′ end of the GUS gene in place of the BYDV‐PAV 5′UTR (construct 109GUSA, Figure 1A). This mRNA was translated cap‐independently (Figure 6A, lanes 3 and 4). In contrast, the nearly identical construct, 109BFGUS, differing only by the GAUC duplication, was not translated in a cap‐independent manner (Figure 6A, compare lanes 6 and 4). Capped mRNAs, even with the non‐functional 3′TE at the 5′UTR, were still translated efficiently (Figure 6A, compare lanes 5 and 3), similar to the results seen when the mutant 3′TE was located at the 3′UTR (Figure 1C). Furthermore, in trans, free 109 nt 3′TE (but not the defective mutant form) inhibited translation of uncapped mRNA that contained the 109 nt 3′TE at the 5′UTR (Figure 6B).

Figure 6.

Translation of transcripts having the 109 nt 3′TE in place of the viral 5′UTR upstream of the GUS ORF. (A) In vitro translation in wheat germ extracts. Prior to transcription, templates for transcripts in lanes 3–6 were linearized with XbaI, and templates for transcripts in lanes 7–10 were linearized with EcoRI, resulting in the poly(A) tail indicated by A+. (B) Effect of addition of the 109 nt 3′TE or 113 nt 3′TEBF transcripts in trans on translation of uncapped 109GUS in wheat germ extracts. 109GUS RNA was transcribed from XbaI‐cut p109GUSA. (C) GUS activity from the indicated transcripts 20 h after electroporation into oat protoplasts, assayed as described in Materials and methods and Figure 2.

To test the function of the 5′‐located 3′TE in vivo, polyadenylated transcripts containing wild‐type or defective 3′TE in the 5′UTR were electroporated into oat protoplasts. The wild‐type 109 nt 3′TE, but not the mutant 3′TE, stimulated GUS expression to a level 30‐fold greater than background (Figure 6C, compare uncapped 109BFGUSA+ with uncapped 109GUSA+). Addition of a 5′ cap stimulated translation another 7‐fold, similar to what was observed when the 109 nt 3′TE was located in the 3′UTR (Figure 2A, PGUS109A+). As a control, a construct also lacking the BYDV‐PAV 5′UTR but with the 109 nt 3′TE in the 3′UTR (ΩGUS109A+) gave no GUS expression unless capped (Figure 6C). This is not surprising because, as mentioned previously (Figure 2B), a similar construct but with the full 1162 nt 3′UTR from BYDV‐PAV (ΩGUS1162) also showed no cap‐independent translation. The 1162 nt 3′UTR (full, in vivo‐defined 3′TE) could not be tested in the 5′UTR owing to the presence of numerous AUG codons. Interestingly, the 5′‐located 109 nt 3′TE stimulated translation of uncapped mRNA to a level that was nearly one‐half that of uncapped PGUS1162 in protoplasts (Figure 6C). When located in the 3′UTR, the 109 nt 3′TE (PGUS109A+) gave expression that was only about one‐eighth that of PGUS1162 (Figure 2A). Thus, the 3′TE may stimulate translation more efficiently from the 5′UTR. Most importantly, Figure 6 shows that interaction between the natural viral 5′UTR and 3′TE is not essential for recruitment of the translation machinery that gives cap‐independent translation.

A highly conserved portion of the 3′TE is present in other plant viral genomes

To determine how widespread this 3′ translation enhancer phenomonon might be, we performed a BLAST search of the GenBank database for homologous sequences. A 27 base sequence that spans the BamHI site and is absolutely conserved in all 11 sequenced BYDV‐PAV and BYDV‐MAV isolates (Chaloub et al., 1994) was used as the sequence for comparison. An 18 nt portion of this sequence was discovered to be conserved near the 5′ ends of the 3′UTRs of all subgroup I luteovirus genomes, RNAs 1 of the bipartite dianthoviruses, and tobacco necrosis virus RNA (TNV, the helper for STNV), but not STNV RNA (Figure 7). This suggests that these other viruses may employ this mechanism of translation. TNV RNA is known to lack a 5′ cap (Lesnaw and Reichmann, 1970), but RCNMV RNA1 reportedly is capped (Xiong and Lommel, 1989). As expected, BYDV‐PAV RNA also appears to lack a cap, and in vitro transcripts of a full‐length clone of the BYDV‐PAV genome are more infectious when lacking a 5′ cap (Mohan et al., 1995; Wang, 1996). The presence of a highly conserved sequence at similar locations in the viral genomes, combined with the fact that this sequence contains the BamHI site, mutation of which destroys 3′TE function, suggests that the 18 base sequence in Figure 7 is involved directly in the mechanism of cap‐independent translation.

Figure 7.

Alignments of portions of 3′UTRs of viral genomes showing their locations between the 3′‐proximal confirmed gene (encodes a structural protein in all cases) and the 3′ end of the genome. Underlined bases of consensus (bold) are complementary to the conserved sequence 5–10 bases from the 3′ end of 18S rRNA (Dams et al., 1988). Underlined bases in STNV RNA have been proposed to base‐pair to a nearby, non‐overlapping region of 18S rRNA (Danthinne et al., 1993). Vertical bars indicate locations of additional bases (below line) which are looped out in order to optimize alignments. In addition to the BYDV‐PAV sequence of the given accession number, 10 other BYDV‐PAV isolates share this sequence (Chaloub et al., 1994). Abbreviations not used in text: BYDV‐MAV, the MAV BYDV; TNV‐A and ‐D, the A and D strains of TNV; SCNMV1, RNA 1 of sweet clover necrotic mosaic virus; CRSV1, RNA 1 of carnation ringspot virus; Luteo I, subgroup I luteovirus; Necro, necrovirus; Diantho, dianthovirus.


The 3′TE mimics a 5′ cap, not a poly(A) tail

The 109 nt intergenic region (109 nt 3′TE) between ORFs 5 and 6 of the BYDV‐PAV genome confers a remarkably high (30‐ to 100‐fold) stimulation of translation of uncapped mRNA in wheat germ extracts. The complete abolition of the 3′TE stimulatory activity by a small mutation (four base duplication) within this sequence demonstrates the high sequence specificity of the interaction between the mRNA and some component of the translational apparatus. However, the 109 nt 3′TE sequence alone was insufficient for full stimulation in vivo, even in the presence of a poly(A) tail. Poly(A) tails have little influence on translation in wheat germ extracts, but interact with the 5′ cap to facilitate translation in vivo (Gallie, 1991; Jacobson, 1996; Hentze, 1997). The endogenous levels of GUS activity in oat cells obscured any low level of translation that might have been obtained from reporter mRNAs lacking both a 5′ cap and the 3′TE, or both a poly(A) tail and the 1162 nt viral 3′UTR. This large portion of the BYDV‐PAV genome that spans the smaller in vitro‐defined 3′TE was required to stimulate translation of uncapped mRNA to a level of at least one‐half that of its capped counterpart in vivo. Because this sequence obviated the need for a poly(A) tail, we propose that the large 3′UTR also contains a sequence that functionally substitutes for a poly(A) tail such as the pseudoknot‐rich domain of TMV (Leathers et al., 1993) and the 3′ stem–loop of histone mRNAs (Gallie et al., 1996). This poly(A) tail‐like activity is functionally distinct from the cap‐like activity of the 3′TE, as evidenced by the BamHI fill‐in mutant which had no 3′TE activity but had full translation activity in vivo when capped (Figure 2B, PGUS1162BF).

Our experiments to measure the RNA degradation rate in wheat germ (Wang and Miller, 1995) and in vivo (Figure 3) indicated that there is no significant difference in the stability of uncapped RNA containing the full, in vivo‐defined 3′TE (PGUS1162) and capped 3′TE mutant (PGUS1162BF) transcripts. In one of the main eukaryotic decay pathways, cellular mRNAs are deadenylated first, followed by decapping and degradation in a 5′ to 3′ direction (Muhlrad and Parker, 1994; Beelman and Parker, 1995). Because decapped mRNAs are degraded rapidly in vivo, the additional viral sequence needed for the 3′TE to function in vivo, but not in vitro, may serve to confer RNA stability. It may do this simply by recruiting translational machinery so efficiently that the mRNA is inaccessible to nucleases. mRNAs undergoing efficient translation are associated with poly(A)‐binding protein, eIF4F and other translation factors to form a circular mRNP structure (Tarun and Sachs, 1996; Hentze, 1997), and thus they should be more resistant to the 5′ exonuclease attack (Caponigro and Parker, 1995). The 3′–5′ communication required for 3′TE function (in its normal 3′UTR setting) may exploit the factors that mediate the poly(A) tail–5′ cap interactions (Tarun and Sachs, 1996) in such a way as to prevent recognition by the exonucleases that target uncapped or non‐polyadenylated mRNAs.

To summarize, the 3′TE of PAV RNA has all the following properties of a 5′ cap. (i) It eliminates the need for a 5′ cap for efficient translation in vitro and in vivo (in conjunction with the 5′UTR). (ii) It can be replaced by a 5′ cap. (iii) It cannot be replaced by a poly(A) tail. (iv) It requires a poly(A) tail or viral 3′UTR sequence for translation in vivo but not in wheat germ extracts. (v) It stabilizes mRNA to the same extent as a 5′ cap in vivo. (vi) It reduces the concentration of eIF4F needed for maximum translation. (vii) Its activity is inhibited by added free m7G (Wang and Miller, 1995) or free 109 nt 3′TE in trans (Figure 5). (viii) Free m7G (Fletcher et al., 1990) or free 109 nt 3′TE inhibit translation of capped mRNAs lacking any BYDV‐PAV sequence. (ix) All types of inhibition above are relieved by exogenous eIF4F. (x) The 109 nt 3′TE functions when located at the 5′ end of mRNA. This last observation makes the important distinction between the two functions that are mediated by the 3′TE: translation initiation (e.g ribosome recruitment) and 3′–5′ communication.

3′–5′ communication

When the 3′TE is located at the 3′ end of the mRNA, interaction between the 5′ and 3′ ends must occur, given that ribosomes scan 5′ to 3′ and peptide synthesis initiates at the 5′‐proximal AUG. The specific involvement of the BYDV‐PAV 5′UTR in this interaction was suggested by the findings that neither a vector‐derived 5′UTR (Wang and Miller, 1995) nor the translationally efficient 5′UTR (Ω) of TMV (Figure 2B) supported 3′TE function. However, we cannot rule out the possiblity that other 5′UTRs might also function. The interaction between the 3′TE and 5′UTR of BYDV‐PAV RNA could be either through long‐distance base pairing as proposed for STNV (Danthinne et al., 1993) or mediated via trans‐acting factor(s). The first possibility is unlikely for BYDV‐PAV. First, computer analysis revealed no strong, conserved Watson–Crick base pairing between the 5′ and 3′UTRs. Secondly, the 3′TE functions at widely varying distances from the 5′UTR and tolerates unrelated (GUS or viral) intervening sequences, which would be expected to affect such long‐distance base pairing. Thirdly, the 3′TE can trans‐inhibit translation of capped ΩGUS mRNA lacking any BYDV‐PAV sequence, arguing against specific base pairing as a mechanism for inhibition. Instead, the 3′TE most probably interacts with a factor (or the ribosome itself) required for translation of all capped mRNAs, in order to facilitate the interactions between 3′ and 5′UTRs. One possibility is that a trans‐acting factor first recognizes the 3′TE and then interacts with an initiation factor, such as eIF4F, which in turn interacts with the 5′ end. This would resemble the interaction between eIF4F and poly(A)‐binding protein as in normal poly(A) tail‐mediated translation initiation (Tarun and Sachs, 1996), with the crucial difference that a 5′ cap is not needed for 5′ end recognition. Other proteins are known to interact functionally with both ends of mRNAs (Schnierle et al., 1992; Dubnau and Struhl, 1996; Tanguay and Gallie, 1996), but also in a cap‐dependent manner.

Ribosome recruitment

Some of the same components proposed for 3′–5′ communication may also be involved in ribosome recruitment. The simplest model would be that eIF4F has a high affinity for the 3′TE (Figure 8A). A structure within the 3′TE may mimic a 5′ cap to recruit the eIF4E subunit of eIF4F, or the 3′TE may bind eIF4G. The latter case would resemble IRES‐mediated cap‐independent translation in which eIF4G was found to bind the functional IRES RNA with much higher affinity than defective mutant IRES RNAs or non‐IRES RNA (Pestova et al., 1996). Of course the IRES differs from the 3′TE by having the binding site located in the 5′UTR. eIF4F as well as eIF4B have been reported to bind poly(A) tails directly (Gallie and Tanguay, 1994) or via poly(A)‐binding protein (Tarun and Sachs, 1996) in order to facilitate initiation at the 5′ AUG. Although these interactions require a cap for 5′ end recognition, the same protein(s) may facilitate the 3′–5′ communication that allows the 3′TE and viral 5′UTR to facilitate initiation at the 5′ end in the absence of a cap.

Figure 8.

Possible models for ribosome recruitment by the 3′TE consistent with the data presented. (A) eIF4F (consisting of subunits eIF4E and eIF4G) specifically binds the 3′TE with high affinity, recruits the 40S subunit as it does in normal cap‐dependent translation and, perhaps with participation of other factors, recognizes the 5′UTR. (B) An unknown factor (X) specifically binds the 3′TE, and possibly the 5′ UTR, prior to recruiting eIF4F. Subsequent events occur as in (A). (C) A portion (underlined in Figure 7) of the conserved region in the 3′TE (mRNA) base‐pairs directly to the conserved bases 5–10 nucleotides upstream of the 3′ end of 18S rRNA. This base pairing may be facilitated by eIF4F (shown as its subunits eIF4E and eIF4G), or eIF4F may bind after the mRNA base‐pairs to the 18S RNA. Following this binding, the ribosome is delivered to the 5′ end of the mRNA by a mechanism not shown, perhaps by eIF4F. eIF4F (and other factors; Browning, 1996) then participates in scanning until the first AUG is detected, at which polypeptide synthesis initiates. The GAUC duplication at the BamHI site disrupts base pairing to 18S rRNA, perhaps by formation of the alternative secondary structure shown or by others (not shown). eIF4F either does not recognize the mutated 3′TE, or it does not bind because the mRNA never base‐pairs to the 18S RNA.

Our second model invokes an unknown factor that specifically binds the 3′TE and possibly also the 5′UTR prior to recruiting eIF4F (Figure 8B). The trans inhibition of translation by the 3′TE, as well as reversal of trans inhibition by eIF4F, could be explained by the 3′TE RNA interacting with the unknown trans‐acting factor which then binds eIF4F. This resembles yeast poly(A)‐binding protein Pab1p which binds eIF4F only in the presence of poly(A) (Tarun and Sachs, 1996). To shed light on these models, obvious objectives for future research are to determine the relative binding affinities of 3′TE and 3′TEBF RNAs for eIF4F and other proteins in the wheat germ extract.

A third, different possibility which can act in addition to, or instead of, one of the previous models, is that the 3′TE may base‐pair directly to 18S rRNA. This would resemble bacterial mRNA–ribosome binding (Shine–Dalgarno) interactions. The highly conserved 18 base tract in the 3′TE contains the sequence GAUCCU (underlined in Figure 7). This sequence has the potential to base‐pair to a region in 18S rRNA that is precisely the same number of bases from the 3′ end as is the prokaryotic message‐binding site (anti‐Shine–Dalgarno sequence) in 16S rRNA (Dams et al., 1988). Thus, ribosome binding may occur by a prokaryotic‐like mechanism, as proposed for ribosomal recognition of picornaviral IRESs (Pilipenko et al., 1992), ribosome jumping on adenovirus mRNAs (Yueh and Schneider, 1996) and ribosome recognition of STNV RNA (Danthinne et al., 1993) (Figure 8C). eIF4F may facilitate this or eIF4F may simply bind after the event to promote 5′ end recognition and 5′ to 3′ scanning to the initiator AUG. In support of this model, the lethal GAUC duplication includes four of the six bases in the potential rRNA‐binding sequence (Figure 7). The duplication may disrupt 3′TE–18S rRNA interactions, for example by altering secondary structure (Figure 8C). The homologous sequence in the apparently capped RNA1 of red clover necrotic mosaic virus (RCNMV; Xiong and Lommel, 1989) has a mismatch in this six base sequence. Deletion of this homologous sequence did not greatly reduce translation of uncapped RCNMV RNA1 transcript (data not shown). This supports our model but begs the question as to why this sequence is conserved in dianthoviral RNAs.

The 18 base sequence is conserved poorly if at all in the 3′UTR of STNV RNA, but computer alignment of part of the stimulatory region of the STNV 3′UTR with the above viral RNAs also aligns the proposed rRNA‐binding sequences (Danthinne et al., 1993) (Figure 7). However, the STNV sequence identified by Danthinne et al. (1993) would base‐pair to a region much further from the 3′ end of 18S rRNA and requires disruption of a very stable stem–loop in the rRNA (Dams et al., 1988). Danthinne et al. (1993) also proposed that the 3′ stimulatory sequence enhances translation by binding ribosomes that have completed translation to facilitate rapid recycling to the 5′ end. This mechanism is ruled out in the case of BYDV‐PAV RNA by the ability of the 3′TE to function when located in the 5′UTR and absent from the 3′UTR.

Whatever the mechanism, it is clear that resolving the mechanism(s) of cap‐independent translation initiation conferred by the BYDV‐PAV 3′TE will broaden our understanding of the possibilities for translation initiation in eukaryotes.

Materials and methods

Plasmid construction

All constructs were verified by automated DNA sequencing. pGUS1162, pGUSA and pVecGUSA [formerly pCIGUS, pGUSEA and pT7GUSA(+), respectively] were described by Wang and Miller (1995). To construct pGUS1162A, the A30‐containing, SmaI–PvuII fragment of pSP64poly(A) (Promega) was cloned into SmaI‐cut pGUS1162. pGUS109A was derived from pGUSA. First, the 109 nt sequence corresponding to BYDV‐PAV bases 4817–4925 was amplified from pGUS1162 by PCR using primers: 5′ GCGGAGCTCAGACAACACCACTAGCAC 3′ and 5′ GCGGAGCTCCATCGGCCAAACACAATAC 3′. The product was cut with SacI (underlined, an isoschizomer of EcoICRI) and ligated into SacI‐digested pGUSA. To construct pGUS109, the amplified 109 base fragment was inserted into SacI‐digested pGEM3Zf(+), giving rise to p3′TE. Then the SmaI–EcoRI fragment from p3′TE was ligated to EcoICRI–EcoRI‐digested pGUS109A. Plasmids containing the GATC duplication at the BamHI4837 site (BF) were constructed from corresponding ‘wild‐type’ plasmids by cutting, Klenow‐filling and re‐ligating the BamHI site.

pΩGUS1162A was constructed via a series of intermediate plasmids. A fragment corresponding to the Ω sequence and 5′ end of the GUS coding region was PCR amplified from pAGUS‐Tn2 (Skuzeski et al., 1990) with the primers: 5′ GCGGCGGCCGCTAATACGACTCACTATAGGTATTTTTACAAC 3′ and 5′ TCGCGATCCAGACTGAATGC 3′. Following digestion with NotI (underlined) and NcoI (internal), the amplified fragment was cloned into NotI/NcoI‐cut pSL1180 (Pharmacia), giving rise to pSLΩ2. Intermediate plasmid pSLΩGUSA was constructed by subcloning the GUS gene and poly(A) tail from pVecGUSA into the NcoI and EcoRI sites of pSLΩ2. The Csp45I–EcoRI fragment of pGUS1162A was cloned into similarly cut pSLΩGUSA, resulting in pΩGUS1162A. To construct pΩGUS109A, GUS along with the A30 fragment was cut out from pVecGUSA with NcoI and EcoRI, and cloned into NcoI–EcoRI‐cut pSLΩ2.

Plasmids having the 109 nt 3′TE at the 5′ end of the GUS gene in place of the wild‐type BYDV‐PAV 5′UTR are called p109GUSA and p109BFGUSA (Figure 1A). pT7GUS1 (Wang and Miller, 1995) was digested with NcoI at the GUS start codon, filled in with Klenow fragment and then digested with EcoICRI 3′ of the stop codon. The 1.8 kb fragment containing the GUS coding sequence was ligated to SmaI‐cut pTE or pTEBF, giving rise to p109GUS and p109BFGUS, respectively. The EcoRI site in these two constructs was mutated by Klenow fragment fill‐in. Then the Csp45I (in the coding sequence of GUS)–ScaI [in the ampicillin resistance (bla) gene of the vector] fragment was replaced with the corresponding fragment from pPGUS109A, giving rise to p109GUSA and p109BFGUSA, respectively, so that the poly(A) tail was added at the 3′ end of the GUS gene. The plasmids were digested with XbaI to synthesize the 109GUS and 109BFGUS RNAs. They were cut with EcoRI to synthesize the 109GUSA+ and 109BFGUSA+ RNAs.

RNA synthesis

The plasmids were digested at appropriate sites as indicated in Figure 1A. Uncapped transcripts were synthesized with Megascript kit (Ambion, TX). Capped transcripts were synthesized with T7 Message mMachine (Ambion, TX). RNA concentrations were determined spectrophotometrically. RNA integrity was verified by 1% agarose gel electrophoresis.

In vitro translation

Subsaturating amounts (0.2 pmol) of capped or uncapped transcripts were translated in wheat germ extracts (Promega) as described previously (Wang and Miller, 1995). Protein products were quantified using a Molecular Dynamics PhosphorImager and ImageQuant 3.0 software. In competition experiments, the competitor, either 3′TE RNA or 3′TEBF RNA, was first incubated with the wheat germ extract at room temperature for 5 min before the tested mRNA was added to the translation reaction.

RNA electroporation in protoplasts and GUS assay

Oat (Avena sativa, cv. Stout) protoplasts were prepared, electroporated with RNA and cultured as described by Dinesh‐Kumar and Miller (1993). GUS reporter gene assays were performed as in Dinesh‐Kumar and Miller (1993), except that 30 pmol (subsaturating amounts) of the indicated transcripts were used instead of DNA. Cells were assayed for GUS activity 20 h after electroporation except in the time course experiment in Figure 3.

Northern blot hybridization

RNA from electroporated oat protoplasts was isolated and 5 μg of RNA was subjected to electrophoresis as described in Mohan et al. (1995). 32P‐labeled antisense GUS gene RNA transcribed from pVecGUS using SP6 polymerase was used as probe.


The authors thank Chris Brown for valuable advice. This work was funded by the National Science Foundation, grant No. MCB9420806, and the Iowa Agricultural and Home Economics Experiment Station (Journal Paper J‐16857; Project 3330).


View Abstract