Control of Translation by the 5'- and 3'-Terminal
http://www.100md.com
病菌学杂志 2005年第13期
Department of Microbiology
Center for Gene Research and Biotechnology, Oregon State University, Corvallis, Oregon 97331-3804
Division of Vector-Borne Infectious Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Public Health Service, U.S. Department of Health and Human Services, P.O. Box 2087, Fort Collins, Colorado 80522
ABSTRACT
The genomic RNAs of flaviviruses such as dengue virus (DEN) have a 5' m7GpppN cap like those of cellular mRNAs but lack a 3' poly(A) tail. We have studied the contributions to translational expression of 5'- and 3'-terminal regions of the DEN serotype 2 genome by using luciferase reporter mRNAs transfected into Vero cells. DCLD RNA contained the entire DEN 5' and 3' untranslated regions (UTRs), as well as the first 36 codons of the capsid coding region fused to the luciferase reporter gene. Capped DCLD RNA was as efficiently translated in Vero cells as capped GLGpA RNA, a reporter with UTRs from the highly expressed -globin mRNA and a 72-residue poly(A) tail. Analogous reporter RNAs with regulatory sequences from West Nile and Sindbis viruses were also strongly expressed. Although capped DCLD RNA was expressed much more efficiently than its uncapped form, uncapped DCLD RNA was translated 6 to 12 times more efficiently than uncapped RNAs with UTRs from globin mRNA. The 5' cap and DEN 3' UTR were the main sources of the translational efficiency of DCLD RNA, and they acted synergistically in enhancing translation. The DEN 3' UTR increased mRNA stability, although this effect was considerably weaker than the enhancement of translational efficiency. The DEN 3' UTR thus has translational regulatory properties similar to those of a poly(A) tail. Its translation-enhancing effect was observed for RNAs with globin or DEN 5' sequences, indicating no codependency between viral 5' and 3' sequences. Deletion studies showed that translational enhancement provided by the DEN 3' UTR is attributable to the cumulative contributions of several conserved elements, as well as a nonconserved domain adjacent to the stop codon. One of the conserved elements was the conserved sequence (CS) CS1 that is complementary to cCS1 present in the 5' end of the DEN polyprotein open reading frame. Complementarity between CS1 and cCS1 was not required for efficient translation.
INTRODUCTION
Dengue virus (DEN) is a member of the mosquito-borne group of flaviviruses that includes yellow fever, Japanese encephalitis, and West Nile (WNV) viruses. It is the causative agent of millions of human infections annually in the tropical regions of the world (14). The virus has a vertebrate host range limited to humans and monkeys, but replication and amplification also occur in mosquito vectors, principally those of the Aedes genus (4).
The DEN genome is an 11-kb positive-strand RNA with a m7GpppA cap at the 5' end but lacking a poly(A) tail at the 3' end (29). It encodes a single long polyprotein (see Fig. 1A) that is processed through the action of viral and host proteinases to generate 10 mature structural and nonstructural proteins. The maturation pathway, the cleavage sites, and the responsible proteinases have been well described (29). However, systematic studies of the roles of the untranslated regions (UTRs) of the viral RNA in the gene expression of DEN or other flaviviruses are lacking. The UTRs are expected to play an important early role during viral infection in coordinating viral gene expression and the onset of RNA replication.
Translation of the typical cellular mRNA relies on the recruitment of ribosomes by features in the UTRs at both ends of the RNA (11, 20, 31, 40, 41). The 5' cap is recognized by the cytoplasmic cap binding protein eIF4E, which contacts initiation factor eIF4G. This large protein acts as a scaffold for further protein-protein interactions that support initiation. One of those interactions contacts eIF3, a complex factor that is itself bound to the small ribosomal subunit. This chain of interactions thus recruits the ribosomal subunit to the 5' terminus of the RNA. The 3' regions of an mRNA also enhance translation. One mechanism by which this can occur is through an interaction between eIF4G and poly(A) binding protein that is bound to the poly(A) tail (47). The translation enhancing contributions of the cap and poly(A) tail are synergistic (10, 46), and the most actively translated mRNAs are thought to exist in a cyclized or closed-loop state that is considered to enhance initiation and allow for improved ribosome recycling while ensuring that truncated, damaged mRNAs are poorly translated (11, 20, 40). After initial loading at the 5' end of the RNA, the small ribosomal subunit scans in the 3' direction along the RNA in search of an initiation site (27, 38, 39). This is usually the 5'-most AUG triplet, at which initiation frequency is in part determined by the immediate sequence context, GCC(A/G)CCAUGG being the optimal context in mammalian cells (27).
Genomes of positive-strand RNA viruses often lack a 5' cap and/or a poly(A) tail yet must compete with cellular mRNAs for ribosomes to support robust viral gene expression. Viral RNAs that lack a 5' cap often possess an internal ribosome entry site in the 5' UTR that directs interaction with the small ribosomal subunit directly or through the binding of an initiation factor, such as eIF4G (8, 17, 39, 41). This can lead to initiation at an AUG triplet that is not 5' proximal, as occurs with the picornaviruses (19) and the members of two genera (Hepacivirus and Pestivirus) of the family Flaviviridae (29). Since the genomes of DEN and other members of the genus Flavivirus have a 5' cap and initiate translation at the 5'-most AUG triplet, standard cap-dependent initiation rather than internal ribosome entry has been considered to be the gene expression strategy used by these viruses (29).
The 3' UTRs of some viral RNAs that possess a 5' cap but lack a poly(A) tail promote translation to an extent similar to that of UTRs with a poly(A) tail, and through a similar synergistic interaction with the cap. This is true of rotavirus mRNA (49), whose 3'-terminal GACC sequence acts as a translational enhancer (5) by binding the viral protein NSP3, which competes with poly(A) binding protein for interaction with eIF4G (49). It is also true of some plant viral RNAs whose 3' UTRs terminate in a tRNA-like structure, like those of Tobacco mosaic virus, Brome mosaic virus (12, 13), and Turnip yellow mosaic virus (33), although in these cases, the molecular interactions have not been identified.
We report here a systematic analysis of the properties of the 5'- and 3'-terminal regions of the DEN serotype 2 genome in directing translation in Vero cells, using a luciferase (LUC) reporter system. Both 5' and 3' DEN sequences were shown to participate in the regulation of expression, principally by modulating translational efficiency rather than RNA stability.
MATERIALS AND METHODS
Clones and cell culture. Dengue virus serotype 2 (DEN2) sequences were derived from the pD2/IC-30P-A infectious cDNA clone of virus strain 16681 (GenBank accession no. U87411) (23). West Nile virus sequences were derived from genomic clones of a New York flamingo isolate (GenBank accession no. AF196835). Sindbis virus sequences were obtained from pToto54, a full-length genomic clone (GenBank accession no. NC_001547) provided by James Strauss (Caltech). Vero (African green monkey kidney) cells were grown in Dulbecco's modified Eagle's medium (DMEM) (Gibco) containing antibiotics and 10% fetal bovine serum at 37°C in an atmosphere containing 5% CO2.
Plasmid constructs. The basic plasmid for the in vitro transcription of reporter mRNA with DEN2 5' and 3' UTRs and encoding firefly LUC was constructed in the following steps to produce pDLD, shown in Fig. 1B. PCR amplification with the appropriate sequence-modifying primers was used to produce a DNA fragment comprising the DEN2 5' UTR fused to the T7 promoter and flanked by a NotI site at the 5' end and a PstI site at the 3' end. The PstI site was positioned to correspond to the third and fourth codons of the reporter open reading frame (ORF) (encoding Leu-Gln). The PCR product digested with PstI was cloned into the PstI and blunted HindIII sites of pUC18. The LUC ORF was modified during PCR amplification by the addition of an in-frame PstI site in front of the second codon of the LUC ORF (Glu) and by silent mutation of the last two codons to incorporate a HindIII site immediately before the termination codon. At the same time, a BamHI site was added on the 3' side of the LUC ORF, and the PCR product digested with PstI and BamHI was cloned into the same sites of the plasmid described above. Finally, the DEN2 3' UTR was added into the HindIII and SmaI sites after PCR amplification to yield the 3' UTR with a HindIII site placed immediately in front of the native DEN2 TAG stop codon and a SnaBI site appended to the 3' end.
Derivatives of pDLD (see Fig. 1B) were made by analogous PCR amplification of the desired 5' and 3' UTR fragments as described above and subcloning into the NotI-PstI sites and HindIII-KpnI sites, respectively (a KpnI site is present in the polylinker downstream of SnaBI). The UTRs of rabbit -globin mRNA (GenBank accession no. J00658) were made by annealing synthetic DNA fragments. Substitution mutation and deletion variants were made by PCR-mediated mutagenesis and subcloning. A minimal 3' UTR () was made by the removal of sequences between the BamHI and XhoI sites of an intermediate plasmid, resulting in a 15-nucleotide (nt) nonviral 3' UTR (sequence AAAUGGAUCUCGA). The sequences of all subcloned segments were verified by DNA sequencing. Table 1 lists the relevant sequences of clones reported here.
RNA transcription and transfection of cells. Plasmids of the pDLD family were linearized by cleavage with a restriction enzyme as indicated in Fig. 1B. RNAs were produced in vitro by transcription with T7 RNA polymerase and uniformly labeled with low levels of [-32P]CTP (0.2 μCi per 20 μl reaction mixture) to facilitate quantification by liquid scintillation counting of trichloroacetic acid-precipitable material; 5'-capped RNAs were produced in the presence of a sixfold excess of m7GpppG (Epicentre) over GTP (32). Template DNA was removed by digestion with RNase-free DNase, and RNA quality was analyzed via native 1% agarose gel electrophoresis.
In vitro-transcribed reporter RNA (1 pmol) was electroporated into 6 x 105 trypsinized Vero cells in electroporation buffer (25% DMEM and 75% "cytosalts," which contain 120 mM KCl, 0.15 mM CaCl2, 10 mM potassium phosphate, pH 7.6, 5 mM MgCl2) by using 0.2-cm cuvettes and a Gene Pulser Xcell device (Bio-Rad) as recommended by the manufacturer (the protocol can be found at the BTX website [http://www.btxonline.com/applications/protocols/mammalian/]). Electroporated cells were diluted with 4 volumes of DMEM containing 10% fetal bovine serum, transferred to tissue culture plates, and held at 37°C in an atmosphere containing 5% CO2. Cells were harvested at various time points by lysis with 50 μl of 1 x passive lysis buffer (Promega). A number of the experiments reported in this paper were also conducted by using lipofection to deliver RNAs into cells (as used in reference 18). Although broadly similar results were obtained, we have found electroporation to be more consistent, and it permits direct estimation of translational efficiency and functional RNA half-life.
Analysis of luciferase activity. Portions of each extract (10 μl) were loaded into the wells of clear-bottom, black, 96-well plates and mixed with 50 μl of luciferase assay reagent (Promega) for luminometry in a 1450 MicroBeta TriLux counter (Wallac). Luciferase activities were normalized to the total protein concentration determined with protein assay reagent (Bio-Rad) and expressed as relative light units per mg protein.
Assessment of physical stabilities of RNAs after transfection. RNA transcripts were uniformly radiolabeled in the presence of [-32P]CTP and delivered by electroporation as usual into Vero cells. At various time points after electroporation, RNAs were extracted by use of Trizol (Invitrogen), denatured by glyoxalation, and separated by electrophoresis in 1% agarose gels run in phosphate buffer (32). After being dried, gels were exposed to phosphorimager plates, and full-length RNA bands were quantitated after phosphorimagery.
In vitro translation and assessment of the relative specific activities of LUC variants with altered N termini. LUC reporter RNAs were translated in rabbit reticulocyte lysate (Promega) in the presence of [35S]methionine (Amersham). Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and dried gels were exposed to phosphorimager plates. Luciferase bands were quantitated after phosphorimagery. Dilutions of translation reactions were used for estimation of LUC activity in the standard assay. Taking into account the number of methionine residues in each form of LUC, LUC-specific activities were expressed as relative light units per pmol of LUC fusion protein.
RESULTS
Use of luciferase as a reporter for studying the effects of DEN RNA terminal regions on translational expression in primate cells. Firefly LUC has been widely used as a reporter in translational expression studies because of its sensitivity and ease of assay. We modified a cloned LUC coding region to place it under the control of a T7 promoter and to permit convenient cassette replacement of upstream and downstream control regions (Fig. 1B). In vitro transcription with T7 RNA polymerase of linearized reporter constructs produced defined mRNA transcripts that were electroporated into Vero (monkey kidney) cells for subsequent assay of LUC activity in cell lysates.
The basic LUC reporter construct with 5'- and 3'-terminal regions from DEN2 RNA, from which DLD RNA (DLD was named for DEN 5' UTR/LUC/DEN 3' UTR) can be transcribed, is shown in Fig. 1B. A unique NotI site was placed upstream of the T7 promoter, which is immediately adjacent to the 96-nt-long DEN2 5' UTR. The 5' UTR was fused to the first two codons of the polyprotein ORF in order to preserve the natural AUG context, and a unique PstI site was engineered in frame between these codons and the LUC ORF. To facilitate T7 transcription, the 5' A residue of the genomic sequence was replaced with a G. At the other end of the LUC ORF, the last two codons were modified with silent mutations to incorporate a HindIII site just upstream of the stop codon. The native DEN2 stop codon and entire 451-nt-long 3' UTR was placed between this HindIII site and a SnaBI site placed at the 3' end (Fig. 1B). Linearization of pDLD plasmid DNA with SnaBI permits the production of transcripts with only three nonnative nucleotides at the 3' termini. To produce RNAs with other UTRs, pDLD was modified by the replacement of NotI-PstI and HindIII-SnaBI fragments with various PCR-generated inserts.
We constructed pDCLD (DCLD was named for DEN/Capsid/LUC/DEN) (Fig. 1B) to investigate the influence of the 5'-proximal part of the DEN ORF, which includes the cCS1 element that permits RNA cyclization by hybridization with the CS1 element in the 3' UTR (Fig. 1A). pDCLD includes the first 36 codons of the DEN capsid coding region fused in front of the PstI site and placed in frame upstream of the LUC coding sequence (Fig. 1B). This results in the production of a capsid-LUC fusion protein. To compare translation driven by viral sequences to that driven by UTRs from a highly expressed cellular mRNA, we have made pGLGpA and pGCLGpA (Fig. 1B). These constructs include the 5' and 3' UTRs derived from rabbit -globin mRNA and a poly(A) tail 72 residues long. In pGLGpA, the first two codons of the -globin ORF are fused to the PstI site and in frame with the LUC coding sequence, while in pGCLGpA, the entire coding region is identical to that in pDCLD. RNAs with and without a poly(A) tail can be made from pGLGpA and pGCLGpA after linearization with Acc65I and XhoI, respectively.
The specific activities of all the forms of N-terminally modified LUC made by these and other RNAs studied in this paper have been determined and have been verified to be similar (Fig. 2). This was assessed by determination of the light yields from known amounts of [35S]methionine-labeled LUC proteins synthesized by in vitro translation. We have previously observed consistent specific activities for LUC forms with different N-terminal extensions (32).
Flavivirus UTRs support highly efficient translation in Vero cells. To assess the overall efficiency with which flavivirus UTRs are able to support translation in Vero cells, we studied LUC expression from reporter RNAs with DEN or WNV UTRs: DLD, DCLD, WLW (named for WNV 5' UTR/LUC/WNV 3' UTR), and WCLW (named for WNV/Capsid/LUC/WNV) RNAs (Fig. 1B). Expression was compared to that from GLGpA RNA and from SinLSin RNA, which have -globin and Sindbis virus UTRs, respectively (Fig. 1B). SinLSin RNA was included as a representative of a virus capable of rapid amplification and highly efficient gene expression (44). All RNAs contained 5' m7GpppG caps (Fig. 3). In time courses following LUC expression up to 9 h after RNA delivery by electroporation, DLD, WLW, and SinLSin RNAs supported the synthesis of considerably more LUC activity than did GLGpA RNA (Fig. 3A).
We derive three parameters relevant to UTR function from LUC expression time courses. Estimates of the maximum rate of increase in LUC activity (linear rate [see Fig. 3 to 7]), which is observed during an early phase of linear increase, are taken to reflect translational efficiencies (9, 32). The maximum accumulation of LUC activity reflects the expression capacity of each RNA. This capacity is influenced by the translational efficiency of the mRNA and by the stabilities of the mRNA and translated protein. The combined RNA and protein stabilities can be monitored by determining the longevity of continued LUC expression. From the time courses, we estimate the half-life of LUC expression as the time taken for the maximum rate of expression (discussed above) to fall to half that rate. Rates are determined as the slopes of the lines of best fit for the expression time courses. For reporter RNAs encoding the same protein, differences in LUC half-life reflect differences in RNA functional stability (the RNA's capacity for directing protein synthesis) (9, 32). In cases in which the N termini of the encoded LUC proteins are different, differences in protein stability may also be reflected in the LUC protein half-life. The expression half-lives of LUC expressed from DLD, WLW, SinLSin, and GLGpA RNAs were all similar (Fig. 3A). LUC synthesis from the RNAs with viral UTRs were expressed with two- to threefold higher translational efficiency than LUC expressed from GLGpA RNA (Fig. 3A, linear rate). The flaviviral reporter RNAs expressed LUC at least as efficiently as did SinLSin RNA. LUC expression from WLW and SinLSin RNAs consistently appeared earlier than from DLD RNA (Fig. 3A), suggesting differences in the ability to recruit ribosomes for the initial round of translation.
Variants of DLD and WLW that included the first 36 and 31 codons, respectively, of each viral capsid coding region (DCLD and WCLW RNAs [Fig. 1B]) were constructed to more closely represent the translation of natural viral RNAs. The presence of additional 5' viral sequences allows local folding into a predicted stem-loop that includes the initiation AUG (mfold [52]) and long-distance pairing between CS1 and cCS1. Further, under some circumstances, the proximal part of an ORF can affect the efficiency of its own translation (7, 26). The same fragment of the DEN capsid ORF was present in GCLGpA RNA as a control. The specific activities of the capsid-LUC fusion proteins varied little from those of the corresponding LUC proteins lacking the capsid ORF (Fig. 2B), while expression half-lives were marginally shorter (Fig. 3). The decrease in LUC activities after 6 h suggests a shorter half-life for the capsid-LUC fusion proteins.
The efficiency and maximal expression of LUC from DCLD and WCLW RNAs were two- to fivefold lower than those of LUC from DLD and WLW RNAs, respectively (Fig. 3B and A, respectively). Nevertheless, DCLD and WCLW RNAs supported strong LUC expression at levels similar to that of GLGpA RNA. The presence of DEN capsid coding sequences in GCLGpA RNA had little effect on LUC expression (Fig. 3, compare GCLGpA and GLGpA), in contrast to their effect on DLD RNA. Decreased expression is, thus, not a property of the encoded protein but likely a property of the RNA (such as different folding). The capsid-coding regions include cCS1, thereby permitting long-distance hybridization to the complementary sequence, CS1, that is present in the viral, but not globin, 3' UTRs. Together with additional long-distance 5'-3' pairing involving sequences just upstream of the initiation AUG as suggested recently (48), these RNA-RNA interactions may dampen the efficiency of ribosome access in the case of RNAs with viral 5' and 3' terminal regions.
Translational regulation by DEN 5' and 3' UTRs. Efficient translation of cellular mRNAs depends heavily on the 5' cap and 3' poly(A) tail, which enhance expression synergistically (10, 46). This response was confirmed for Vero cells by using transcripts from pGCLGpA. Detecting synergy requires study of the expression of capped and uncapped versions of an RNA, as well as variants with and without a poly(A) tail. In the presence of a 5' cap, the addition of a polyadenylated globin 3' UTR (3'-GpA) enhanced translational efficiency (linear rate) by 108-fold with reference to an RNA with a minimal vector-derived 3' UTR (Fig. 4A and C; compare GCLGpA and GCL). Far less enhancement (5.5-fold) in response to adding 3'-GpA when RNAs were uncapped was observed (Fig. 4B and C; compare GCLGpA and GCL). The ratio between these responses (108/5.52 = 19.6) represents synergistic enhancing effects between the 5' cap and polyadenylated 3' UTR, implying collaboration between these features. Note that the presence of 3'-GpA also increased the LUC expression half-life by threefold (82/27) (Fig. 4A). Since GCLGpA and GCL RNAs encode the same form of LUC, we can conclude that there is a threefold difference in the functional half-lives of these RNAs.
Similar analyses were conducted to study the contribution of the DEN 3' UTR to translational expression. Like the 3'-GpA, the DEN 3' UTR is clearly important for high-level expression of LUC (Fig. 4A, compare DCLD and DCL RNAs). The addition of the DEN 3' UTR to DCL RNA (to produce DCLD RNA) enhanced translational efficiency 32-fold in the presence of a 5' cap (Fig. 4A and C; compare DCLD and DCL) and 10.5-fold in the absence of a cap (Fig. 4B and C; compare DCLD and DCL). This indicates the existence of a synergistic interaction between the cap and DEN 3' UTR (synergy value, 32.3/10.5 [3.1]), although the interaction is considerably less strong than that between a cap and 3'-GpA.
Studies with the uncapped forms of RNAs, which support relatively low levels of LUC expression, emphasized the importance of a 5' cap. Among uncapped RNAs, however, elevated levels of LUC expression from RNAs with DEN UTRs were observed. This is obvious in the case of DCLD RNA in the time course shown in Fig. 4B. An enhancing effect of 5' DEN sequences can be seen for DCL RNA in comparison with GCL RNA, while an effect of the DEN 3' UTR is evident when comparing expression levels from GCLD and GCLGpA RNAs (Fig. 4B and C). The enhancing effect was stronger for the DEN 5' UTR, which supported the synthesis of LUC 6 to 12 times more efficiently (Fig. 4B and C, linear rates) than 5' G in RNA counterparts with the same 3' UTR. No such 5' DEN-dependent enhancement in the capped state was seen, since capped DCLD, GCLD, and GCLGpA RNAs were similarly expressed (Fig. 4A).
LUC expression half-life estimates indicated that the GpA and DEN 3' UTRs and the 5' cap all contribute to RNA stability. Capped DCLD, GCLD, and GCLGpA RNAs (which encode the same form of LUC) had similar functional half-lives, and the removal of the 3' UTR decreased the half-life by about threefold (Fig. 4A, compare DCL and GCL RNAs). Uncapped DCLD RNA had a half-life similar to that of capped DCL or GCL RNAs, evident from the early peak in LUC activity (see the DCLD time course in Fig. 4B). While RNA stability differences associated with the 5' cap and 3' UTR are important, they were much smaller in magnitude than the enhancements in translational efficiency provided by these features. Analyses of the physical stabilities of RNAs during expression time courses confirmed that RNA stability differences were relatively small among the RNAs studied in Fig. 4 (Fig. 4D). In these assays, 32P-labeled RNAs were electroporated and incubated in cells before being extracted and examined by electrophoresis (see Materials and Methods).
Expression from DCLGpA and GCLD RNAs was studied to assess whether there are any cooperative effects between DEN 5' and 3' UTRs. LUC expression from both of these capped RNAs was comparable to expression from capped DCLD and GCLGpA RNAs (Fig. 4A), clearly indicating that the DEN UTRs do not need to be coupled to support strong translation. The addition of the DEN 3' UTR to GCL RNA enhanced translational efficiency 174-fold in the presence of a 5' cap and 20.6-fold with no cap (Fig. 4C). Thus, the synergy between the 5' cap and DEN 3' UTR (synergy value, 174/20.6 [8.4]) for these RNAs with a globin 5' UTR is stronger than that observed for RNAs with the DEN 5' UTR (synergy value of 3.1; see above). As evident when comparing DCLD and DCL RNAs, the DEN 3' UTR in GCLD RNA enhanced the RNA functional half-life about threefold (Fig. 4A). We conclude that the DEN 3' UTR is a general enhancer of translation and RNA stability whose action does not require a matched DEN 5' UTR.
Complementarity between cyclization sequences CS1 and cCS1 is not needed for translation. The analyses already discussed failed to detect any codependency between 5' and 3' DEN sequences in regard to the control of translation. To explore this issue further, we wished to test directly for a translational role of the conserved cyclization sequences and their potential long-distance hybridization. The RNAs of Fig. 5A represent a set in which CS1 and cCS1 have replaced one another singly and together, resulting in DCmLD and DCLDm RNAs that have diminished potential for long-distance cCS1/CS1 interaction and double-mutant DCmLDm RNA, in which CS1 and cCS1 have been switched and cyclization is possible. Note that one nucleotide of each transposed conserved sequence was changed to avoid introducing a stop codon with CS1 placed in the 5' location. Expression from capped forms of these RNAs in Vero cells was studied.
Replacement of cCS1 with the modified CS1 sequence had little impact on LUC expression or expression half-life (Fig. 5, compare DCLD and DCmLD). By contrast, replacement of CS1 with the modified cCS1 decreased LUC expression efficiency to 19%, with only a small effect on expression half-life (Fig. 5, compare DCLD and DCLDm). CS1 and cCS1 thus appear to have differential roles in translational regulation. Significantly, the inability of the double mutation (DCmLDm) to rescue the depressed LUC expression level of DCLDm RNA (Fig. 5B and C) suggests that RNA cyclization via cCS1/CS1 interaction is not required for the translation of the reporter RNAs used here. The minimal differences in the half-lives of LUC expression from the RNAs in Fig. 5 indicate that these RNAs have similar stabilities.
Contribution of conserved elements in the 3' UTR to translational expression. The results described above indicated an involvement of CS1 in translation. In order to identify contributions from other elements in the 3' UTR, the conserved features SLA, SLB, DB1, and DB2, as well the upstream variable region (UVR) (Fig. 1A and 6A), were deleted in turn from the 3' UTR of capped DCLD RNA. The endpoints of deletions were chosen to cleanly delete the predicted structures as shown in Fig. 6A, and this was confirmed by mfold (52) predictions of the folding of the 3' UTRs. A large 3' truncation was made by transcription from a pDCLD template linearized with NcoI, which cleaves at DEN nucleotide 10470 (Fig. 6A). This produced a variant of DCLD RNA (Fig. 6, /NcoI) truncated near the upstream end of DB2 and thus lacking SLA, SLB, CS1, DB1, and most of DB2. LUC expression from capped RNAs in Vero cells was studied (Fig. 6).
The deletion of either of the two 3'-terminal stem-loops SLA and SLB decreased the efficiency of LUC translation to about half (Fig. 6B, linear rate). The deletion of DB1 resulted in a smaller, though significant, decrease in the efficiency of LUC translation, while the deletion of DB2 had little effect (Fig. 6C). The simultaneous deletions of DB1 and DB2 reduced the efficiency of translation to 25%, suggesting redundancy in the contributions of these quite similar features. The deletion of the UVR, just downstream of the stop codon, also resulted in a relative translational efficiency of 25% (Fig. 6D).
Translational enhancement provided by the 3' UTR thus seems to derive from several features, including contributions by each of the conserved features. The individual deletions of these features led to a partial loss of activity. Expression from each of the RNAs with a single element deleted was higher than that from RNA entirely lacking a DEN 3' UTR (Fig. 7A, DCL). Presence of the UVR feature alone (Fig. 7A, /NcoI RNA) increased expression only slightly from this basal level (Fig. 7A), emphasizing the importance of combined contributions to translational enhancement by the 3' UTR.
Interestingly, DCLG RNA, which has the 87-nucleotide-long globin 3' UTR but no poly(A) tail, supported the same basal level of LUC expression as DCL RNA (Fig. 7). High levels of expression were dependent on the poly(A) tail (Fig. 7A, DCLGpA RNA). This emphasizes the similar roles of the DEN 3' UTR and poly(A) tail in enhancing translation.
It is important to note that no noteworthy differences in functional RNA stability were observed for the RNAs with various partial deletions of the DEN 3' UTR, although most of these capped RNAs had somewhat shorter half-lives than capped DCLD RNA (Fig. 7B); this conclusion was supported by an analysis of the physical stabilities of these RNAs (not shown). The RNAs containing partial 3' deletions were all more stable than DCL RNA (Fig. 7B). It appears that, as for translational enhancement, RNA stabilization (roughly threefold) provided by the DEN 3' UTR (noted earlier) derives from multiple features in the UTR. The globin 3' UTR also appears to enhance RNA stability (twofold stabilization observed for DCLG RNA relative to DCL [Fig. 7B]), with little if any further stabilization provided by a poly(A) tail (Fig. 7B, DCLGpA RNA). The contributions to translational enhancement were just the opposite, being primarily derived from the poly(A) tail and not measurable from the 3' G sequence (Fig. 7A).
Although it has been observed that a 3' UTR can stabilize RNA and enhance translation by virtue of its length, in a sequence-nonspecific effect (45), this is unlikely to be a consideration among the 3' UTR domain deletions. Those deletions remove no more than 154 nt (DB1 + 2), leaving at least a 297-nt-long UTR, and, conversely, the addition of the 87-nt 3' G did not enhance translation (Fig. 7A). The nonspecific length effects were observed for 3' UTRs shorter than 27 nt (45).
DISCUSSION
Flaviviral 5'- and 3'-terminal genome sequences support highly efficient translational expression. Using a luciferase reporter system and comparison with reporter RNAs that support highly efficient translation, we have shown that the 5'- and 3'-terminal regions from the DEN2 and WNV genomes are able to support strong translation in Vero cells. Translational expression levels from 5'-capped DCLD and WCLW RNAs with 5' and 3' sequences from DEN and WNV, respectively, were similar to the level of expression from capped GLGpA RNA and only two to three times lower than the level of expression from SinLSin RNA (Fig. 3). GLGpA RNA has UTRs from the -globin mRNA, while SinLSin RNA has UTRs derived from Sindbis virus, an alphavirus that has been developed as a gene expression vector because of its ability to support potent translational expression (1).
The viral sequences present in DCLD and WCLW RNAs included part of the capsid coding region in addition to the 5' and 3' UTRs. The capsid coding regions were included because they contain the conserved cCS1 sequence element that permits potential cyclization of the RNA by annealing to its complement, CS1, present in the 3' UTR. A surprising finding was that the omission of the capsid coding regions (DLD and WLW RNAs) increased translational expression by two- to fourfold (Fig. 3). This is not due to an elongational block of translation associated with the capsid ORF, since LUC was similarly expressed from GLGpA and GCLGpA RNAs (Fig. 3). We have also verified that the LUC and capsid-LUC fusion proteins have similar light-emitting specific activities (Fig. 2). It seems most likely that the presence of the proximal part of the capsid ORF lessens translational efficiency by affecting ribosome initiation, perhaps by altering the folding of the 5' UTR through local interactions or as a consequence of long-distance hybridization with CS1.
In any case, it is clear that the capped flaviviral 5' UTRs are capable of directing very high translational expression in Vero cells. This was rather unexpected in view of the conserved secondary structure positioned immediately adjacent to the 5' end (2) and suboptimal consensus around the initiation AUG (UCUCUGAUGA for DEN and AUCUCGAUGU for WNV). Secondary structure at the 5' terminus is detrimental to cap-dependent translation initiation (25), and weak consensus nucleotides at the –3 and +4 positions (italicized respectively in the sequences above) can diminish translation by up to 10-fold (–3A or –3G and +4G is optimal [27]). Schrader and Westaway (42) have also commented on the efficient translation of flaviviral (Kunjin virus) RNA in Vero cells despite suboptimal AUG context.
The DEN 3' UTR enhances translation and RNA stability in a way similar to that of a poly(A) tail. The contribution of the DEN 3' UTR to translational expression was studied by determining the effect of replacing a short nonviral 3' UTR (DCL or GCL RNAs) with the DEN 3' UTR (DCLD and GCLD RNAs). For capped RNAs with either the DEN or globin 5' UTRs, the DEN 3' UTR significantly increased RNA stability (about a threefold increase in functional half-life) (Fig. 4A). Far greater increases in translational efficiency were observed, represented by the maximum LUC expression rate in our assays (Fig. 3 to 6, linear rates). In the presence of a DEN 5' UTR, the rate was increased 32.3-fold (Fig. 4C, compare DCLD and DCL), while in the presence of a globin 5' UTR, the increase was 174-fold (Fig. 4C, compare GCLD and GCL). These responses are very comparable to those provided by a polyadenylated 3' UTR. Capped GCLGpA RNA was three times as stable and was translated with 108 times the efficiency of capped GCL RNA (Fig. 4C), and capped DCLGpA RNA was translated with 70 times the efficiency of capped DCL RNA (Fig. 7A). The ability of the DEN 3' UTR to enhance translational efficiency and RNA stability in the presence of either DEN or globin 5' sequences demonstrates this to be a free-standing regulator that is not codependent on viral 5' sequences. Stimulation of the translation of LUC reporter mRNAs by the DEN2 3' UTR has also recently been reported by Holden and Harris (18). Extending those studies, our kinetic analyses have enabled estimation of the contributions of both increased RNA stability and translational efficiency to expression enhancement by the 3' UTR.
The poly(A) tail is known to enhance translation synergistically with a 5' cap (10, 46). Such synergy is established by determining the extent to which the addition of a poly(A) tail has a greater effect in the presence of a cap than in the absence of a cap. The 3'-GpA enhanced translational efficiency 19.6 (108/5.52) times more when a cap was present (Fig. 4C, GCLGpA and GCL RNAs), confirming the strong synergy that others have reported. Similar analyses of the role of the DEN 3' UTR also indicated the existence of synergy with a 5' cap, although this synergy was weaker than that observed for the case with a poly(A) tail and it differed for different 5' UTRs. Thus, the cap/3' DEN synergy value was 8.4 for RNAs with a globin 5' UTR (ratio for GCLD and GCL RNAs, 174/20.6 [Fig. 4C]) and 3.1 for RNAs with a DEN 5' UTR (ratio for DCLD and DCL RNAs, 32.3/10.5 [Fig. 4C]).
If cap/poly(A) synergy reflects the existence of bridging interactions that cyclize the mRNA as described in the introduction, we may think of the DEN 3' UTR recruiting proteins capable of interacting with translation initiation factors assembled around the 5' cap. Such a scenario has been demonstrated for rotaviral mRNAs, with the bridging interaction mediated by the viral protein NSP3 (49). For the DEN 3' UTR, however, the synergy we have observed does not involve a viral protein, since the only viral protein present in our assays is a fragment of the capsid protein.
The quite significant difference in synergy observed with globin or DEN 5' UTRs is an interesting observation. Both the GCLD and DCLD RNAs used in our synergy studies possess the cCS1 and CS1 elements and therefore have the potential to cyclize through long-distance hybridization. Although this annealing would be disrupted by elongating ribosomes passing through cCS1, it may influence the initiation of translation. By forming a closed loop through direct RNA-RNA interaction rather than though protein mediation, translation may potentially be enhanced. This type of effect has been observed with Barley yellow dwarf virus RNA through direct RNA interaction between 5' and 3' UTRs (15). Alternatively, cCS1/CS1 interaction may be inhibitory towards translational initiation. This is suggested by the level of LUC expression from DCLD RNA being lower than that from DLD (Fig. 3) or DCLGpA RNA (Fig. 4), which contain only CS1 or cCS1. Recent modeling of the RNA folding of flaviviral RNAs has suggested the existence of RNA pairing between part of the 5' UTR just upstream of the AUG initiation codon and the 3' UTR (48). For DCLD RNA but not GCLD RNA, this pairing would occur in addition to the cCS1/CS1 interaction. Further experiments are needed to investigate whether these combined interactions are responsible for the cap/3' UTR synergy observed for DCLD RNA being lower than that for GCLD RNA and also for the translational expression level from DCLD RNA being lower than that from DLD RNA (Fig. 3).
Complementarity between cCS1 and CS1 is not necessary for efficient translation driven by DEN UTRs. The speculation above explored a possible inhibitory interaction between 5' and 3' genomic sequences that has yet to be tested experimentally. We did, however, test the opposite effect, that is, whether cCS1/CS1 complementarity and these sequences themselves are required for efficient translation. Replacement of cCS1 with a modified version of CS1 had little influence on translational expression (Fig. 5C, DCmLD). By contrast, replacement of CS1 decreased translational efficiency to about one-fifth (Fig. 5C, DCLDm), and there was a further reduction when these mutations were combined to permit potential hybridization and RNA cyclization (Fig. 5C, DCmLDm). These results indicate that cCS1/CS1 complementarity is not an attribute that supports translation. Indeed, the marginally increased translational efficiency of DCmLD RNA over DCLD RNA and the lower efficiency of DCmLDm DNA relative to DCLDm RNA support the idea that the long-distance RNA-RNA interactions interfere with translation. Such long-distance interactions have been implicated in flaviviral RNA replication (6, 21, 30, 50, 51), and it is intriguing to consider that this interaction may play a role in regulating the transition from the translation to the replication phase of the infection. If the cCS1/CS1 interaction can be stabilized, for instance, by the influence of bound proteins, this interaction might simultaneously inhibit translation and enhance replication. We have recently described an analogous switch involving translation factor eEF1A binding to the 3'-terminal region of Turnip yellow mosaic virus RNA (33, 34).
The low translational expression from DCLDm RNA suggests an independent role for CS1 in supporting translation. Inhibition of such a role may help to explain the potent inhibition of DEN replication by an antisense molecule directed at CS1 (24). No such role in translation was revealed by the only previous study of conserved sequence involvement in translational expression, performed using a WNV reporting replicon (30). In that study, neither the mutation of cCS1 nor that of CS1 detectably influenced the yield of the LUC reporter protein during the initial phase of viral gene expression that was not dependent on RNA replication. Further studies involving a variety of CS1 sequence variants and a comparison of the DEN and WNV systems are needed to clarify the situation. There may also be sensitivity differences between the replicon assay and the mRNA reporter assays we have used.
Multiple features of the DEN 3' UTR enhance translational expression. CS1 is only one of a number of conserved elements present in the 3' UTRs of flaviviral genomes (Fig. 1). By deleting each of the other elements (SLA, SLB, DB1, and DB2) from DCLD RNA, we tested their contributions to the stimulatory effect of the DEN 3' UTR. Our results suggest, for the first time, that both RNA stability and translational enhancement are conferred by multiple features (Fig. 7). At this point, we cannot be certain that lost function is directly due to the absence of the deleted element. Indirect effects due to structural rearrangements in the deleted 3' UTRs could have occurred, even though folding predictions suggested that this was not the case. Using a reporting replicon, Lo et al. (30) recently reported that each of the conserved sequence elements in the 3' UTR of WNV RNA could be mutated or deleted without noticeably affecting translational expression in BHK cells. As noted above, these differences between the DEN reporter RNA and WNV replicon systems will need to be explored.
Among the features that contributed to fully efficient translation were the conserved 3'-terminal stems SLA and SLB. The deletion of either of these decreased translational efficiency to 40 to 50% of that of DCLD RNA (Fig. 7). Holden and Harris (18) have also recently observed that simultaneous deletions of DEN2 SLA and SLB ("3'SL" in their nomenclature) resulted in a similar decrease in translational expression in BHK cells from a reporter RNA like ours. In contrast, Brinton and colleagues have viewed SLA and SLB as negative regulators of translation in the WNV system. That conclusion arose from the observation that addition of WNV SLA and SLB to the 3' UTRs of reporter mRNAs suppressed translation considerably in vitro, though only marginally in BHK cells (28). Further investigation will be needed to explain these differences.
ACKNOWLEDGMENTS
We thank Andrew Regan and Daiki Matsuda for assistance with some experiments, James Strauss for providing pToto54, and George Rohrmann, Connie Bozarth, and Doug Leisy for critical reading of the manuscript.
Andrew Regan was the recipient of a summer undergraduate research fellowship funded by Howard Hughes Medical Institute Grant 57003741 to Oregon State University. We are grateful for the use of the Cell Culture facility and of a 96-well plate densitometer of the Environmental Health Sciences Center at Oregon State University, which is supported by grant number P30 ES00210 from the National Institute of Environmental Health Sciences, National Institutes of Health. These studies were supported by National Institutes of Health grant AI055569 awarded to T.W.D.
REFERENCES
Agapov, E. V., I. Frolov, B. D. Lindenbach, B. M. Pragai, S. Schlesinger, and C. M. Rice. 1998. Noncytopathic Sindbis virus RNA vectors for heterologous gene expression. Proc. Natl. Acad. Sci. USA 95:12989-12994.
Brinton, M. A., and J. H. Dispoto. 1988. Sequence and secondary structure analysis of the 5'-terminal region of flavivirus genome RNA. Virology 162:290-299.
Brinton, M. A., A. V. Fernandez, and J. H. Dispoto. 1986. The 3'-nucleotides of flavivirus genomic RNA form a conserved secondary structure. Virology 153:113-121.
Burke, D. S., and T. P. Monath. 2001. Flaviviruses, p. 1043-1125. In D. M. Knipe and P. M. Howley (ed.), Fields virology, 4th ed. Lippincott Williams & Wilkins, Philadelphia, Pa.
Chizhikov, V., and J. T. Patton. 2000. A four-nucleotide translation enhancer in the 3'-terminal consensus sequence of the nonpolyadenylated mRNAs of rotavirus. RNA 6:814-825.
Corver, J., E. Lenches, K. Smith, R. A. Robison, T. Sando, E. G. Strauss, and J. H. Strauss. 2003. Fine mapping of a cis-acting sequence element in yellow fever virus RNA that is required for RNA replication and cyclization. J. Virol. 77:2265-2270.
Frolov, I., and S. Schlesinger. 1996. Translation of Sindbis virus mRNA: analysis of sequences downstream of the initiating AUG codon that enhance translation. J. Virol. 70:1182-1190.
Gale, M., Jr., S. L. Tan, and M. G. Katze. 2000. Translational control of viral gene expression in eukaryotes. Microbiol. Mol. Biol. Rev. 64:239-280.
Gallie, D. R. 2002. The 5'-leader of tobacco mosaic virus promotes translation through enhanced recruitment of eIF4F. Nucleic Acids Res. 30:3401-3411.
Gallie, D. R. 1991. The cap and poly(A) tail function synergistically to regulate mRNA translational efficiency. Genes Dev. 5:2108-2116.
Gallie, D. R. 1998. A tale of two termini: a functional interaction between the termini of an mRNA is a prerequisite for efficient translation initiation. Gene 216:1-11.
Gallie, D. R., and M. Kobayashi. 1994. The role of the 3'-untranslated region of non-polyadenylated plant viral mRNAs in regulating translational efficiency. Gene 142:159-165.
Gallie, D. R., and V. Walbot. 1990. RNA pseudoknot domain of tobacco mosaic virus can functionally substitute for a poly(A) tail in plant and animal cells. Genes Dev. 4:1149-1157.
Gubler, D. J. 2002. Epidemic dengue/dengue hemorrhagic fever as a public health, social and economic problem in the 21st century. Trends Microbiol. 10:100-103.
Guo, L., E. M. Allen, and W. A. Miller. 2001. Base-pairing between untranslated regions facilitates translation of uncapped, nonpolyadenylated viral RNA. Mol. Cell 7:1103-1109.
Hahn, C. S., Y. S. Hahn, C. M. Rice, E. Lee, L. Dalgarno, E. G. Strauss, and J. H. Strauss. 1987. Conserved elements in the 3' untranslated region of flavivirus RNAs and potential cyclization sequences. J. Mol. Biol. 198:33-41.
Hellen, C. U., and P. Sarnow. 2001. Internal ribosome entry sites in eukaryotic mRNA molecules. Genes Dev. 15:1593-1612.
Holden, K. L., and E. Harris. 2004. Enhancement of dengue virus translation: role of the 3' untranslated region and the terminal 3' stem-loop domain. Virology 329:119-133.
Jackson, R. J., and A. Kaminski. 1995. Internal initiation of translation in eukaryotes: the picornavirus paradigm and beyond. RNA 1:985-1000.
Jacobson, A. 1996. Poly(A) metabolism and translation: the closed-loop model, p. 451-480. In J. Hershey, M. B. Mathews, and N. Sonenberg (ed.), Translational control. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
Khromykh, A. A., H. Meka, K. J. Guyatt, and E. G. Westaway. 2001. Essential role of cyclization sequences in flavivirus RNA replication. J. Virol. 75:6719-6728.
Khromykh, A. A., and E. G. Westaway. 1997. Subgenomic replicons of the flavivirus Kunjin: construction and applications. J. Virol. 71:1497-1505.
Kinney, R. M., S. Butrapet, G. J. Chang, K. R. Tsuchiya, J. T. Roehrig, N. Bhamarapravati, and D. J. Gubler. 1997. Construction of infectious cDNA clones for dengue 2 virus: strain 16681 and its attenuated vaccine derivative, strain PDK-53. Virology 230:300-308.
Kinney, R. M., C. Y. H. Huang, B. C. Rose, A. D. Kroeker, T. W. Dreher, P. L. Iversen, and D. A. Stein. 2005. Inhibition of dengue virus serotypes 1 to 4 in Vero cell cultures with morpholino oligomers. J. Virol. 79:5116-5128.
Kozak, M. 1989. Circumstances and mechanisms of inhibition of translation by secondary structure in eucaryotic mRNAs. Mol. Cell Biol. 9:5134-5142.
Kozak, M. 1990. Downstream secondary structure facilitates recognition of initiator codons by eukaryotic ribosomes. Proc. Natl. Acad. Sci. USA 87:8301-8305.
Kozak, M. 1999. Initiation of translation in prokaryotes and eukaryotes. Gene 234:187-208.
Li, W., and M. A. Brinton. 2001. The 3' stem loop of the West Nile virus genomic RNA can suppress translation of chimeric mRNAs. Virology 287:49-61.
Lindenbach, B. D., and C. M. Rice. 2001. Flaviviridae: the viruses and their replication, p. 991-1041. In D. M. Knipe and P. M. Howley (ed.), Fields virology, 4th ed. Lippincott Williams & Wilkins, Philadelphia, Pa.
Lo, M. K., M. Tilgner, K. A. Bernard, and P. Y. Shi. 2003. Functional analysis of mosquito-borne flavivirus conserved sequence elements within 3' untranslated region of West Nile virus by use of a reporting replicon that differentiates between viral translation and RNA replication. J. Virol. 77:10004-10014.
Mangus, D. A., M. C. Evans, and A. Jacobson. 2003. Poly(A)-binding proteins: multifunctional scaffolds for the post-transcriptional control of gene expression. Genome Biol. 4:223.
Matsuda, D., L. Bauer, K. Tinnesand, and T. W. Dreher. 2004. Expression of the two nested overlapping reading frames of turnip yellow mosaic virus RNA is enhanced by a 5' cap and by 5' and 3' viral sequences. J. Virol. 78:9325-9335.
Matsuda, D., and T. W. Dreher. 2004. The tRNA-like structure of Turnip yellow mosaic virus RNA in a 3'-translational enhancer. Virology 321:36-46.
Matsuda, D., S. Yoshinari, and T. W. Dreher. 2004. eEF1A binding to aminoacylated viral RNA represses minus strand synthesis by TYMV RNA-dependent RNA polymerase. Virology 321:47-58.
Molenkamp, R., E. A. Kooi, M. A. Lucassen, S. Greve, J. C. Thijssen, W. J. Spaan, and P. J. Bredenbeek. 2003. Yellow fever virus replicons as an expression system for hepatitis C virus structural proteins. J. Virol. 77:1644-1648.
Olsthoorn, R. C., and J. F. Bol. 2001. Sequence comparison and secondary structure analysis of the 3' noncoding region of flavivirus genomes reveals multiple pseudoknots. RNA 7:1370-1377.
Pang, X., M. Zhang, and A. I. Dayton. 2001. Development of dengue virus type 2 replicons capable of prolonged expression in host cells. BMC Microbiol. 1:18-24.
Pelletier, J., and N. Sonenberg. 1985. Insertion mutagenesis to increase secondary structure within the 5' noncoding region of a eukaryotic mRNA reduces translational efficiency. Cell 40:515-526.
Pestova, T. V., and V. G. Kolupaeva. 2002. The roles of individual eukaryotic translation initiation factors in ribosomal scanning and initiation codon selection. Genes Dev. 16:2906-2922.
Sachs, A. B., P. Sarnow, and M. W. Hentze. 1997. Starting at the beginning, middle, and end: translation initiation in eukaryotes. Cell 89:831-838.
Schneider, R. J., and I. Mohr. 2003. Translation initiation and viral tricks. Trends Biochem. Sci. 28:130-136.
Schrader, A. P., and E. G. Westaway. 1990. Successful competition in translation by the flavivirus Kunjin with poliovirus during co-infections in Vero cells. Arch. Virol. 114:75-89.
Shi, P. Y., M. A. Brinton, J. M. Veal, Y. Y. Zhong, and W. D. Wilson. 1996. Evidence for the existence of a pseudoknot structure at the 3' terminus of the flavivirus genomic RNA. Biochemistry 35:4222-4230.
Strauss, J. H., and E. G. Strauss. 1994. The alphaviruses: gene expression, replication, and evolution. Microbiol. Rev. 58:491-562.
Tanguay, R. L., and D. R. Gallie. 1996. The effect of the length of the 3'-untranslated region on expression in plants. FEBS Lett. 394:285-288.
Tarun, S. Z., Jr., and A. B. Sachs. 1995. A common function for mRNA 5' and 3' ends in translation initiation in yeast. Genes Dev. 9:2997-3007.
Tarun, S. Z., Jr., S. E. Wells, J. A. Deardorff, and A. B. Sachs. 1997. Translation initiation factor eIF4G mediates in vitro poly(A) tail-dependent translation. Proc. Natl. Acad. Sci. USA 94:9046-9051.
Thurner, C., C. Witwer, I. L. Hofacker, and P. F. Stadler. 2004. Conserved RNA secondary structures in Flaviviridae genomes. J. Gen. Virol. 85:1113-1124.
Vende, P., M. Piron, N. Castagne, and D. Poncet. 2000. Efficient translation of rotavirus mRNA requires simultaneous interaction of NSP3 with the eukaryotic translation initiation factor eIF4G and the mRNA 3' end. J. Virol. 74:7064-7071.
You, S., B. Falgout, L. Markoff, and R. Padmanabhan. 2001. In vitro RNA synthesis from exogenous dengue viral RNA templates requires long range interactions between 5'- and 3'-terminal regions that influence RNA structure. J. Biol. Chem. 276:15581-15591.
You, S., and R. Padmanabhan. 1999. A novel in vitro replication system for dengue virus. Initiation of RNA synthesis at the 3'-end of exogenous viral RNA templates requires 5'- and 3'-terminal complementary sequence motifs of the viral RNA. J. Biol. Chem. 274:33714-33722.
Zuker, M. 2003. Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res. 31:3406-3415.(Wei-Wei Chiu, Richard M. )
Center for Gene Research and Biotechnology, Oregon State University, Corvallis, Oregon 97331-3804
Division of Vector-Borne Infectious Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Public Health Service, U.S. Department of Health and Human Services, P.O. Box 2087, Fort Collins, Colorado 80522
ABSTRACT
The genomic RNAs of flaviviruses such as dengue virus (DEN) have a 5' m7GpppN cap like those of cellular mRNAs but lack a 3' poly(A) tail. We have studied the contributions to translational expression of 5'- and 3'-terminal regions of the DEN serotype 2 genome by using luciferase reporter mRNAs transfected into Vero cells. DCLD RNA contained the entire DEN 5' and 3' untranslated regions (UTRs), as well as the first 36 codons of the capsid coding region fused to the luciferase reporter gene. Capped DCLD RNA was as efficiently translated in Vero cells as capped GLGpA RNA, a reporter with UTRs from the highly expressed -globin mRNA and a 72-residue poly(A) tail. Analogous reporter RNAs with regulatory sequences from West Nile and Sindbis viruses were also strongly expressed. Although capped DCLD RNA was expressed much more efficiently than its uncapped form, uncapped DCLD RNA was translated 6 to 12 times more efficiently than uncapped RNAs with UTRs from globin mRNA. The 5' cap and DEN 3' UTR were the main sources of the translational efficiency of DCLD RNA, and they acted synergistically in enhancing translation. The DEN 3' UTR increased mRNA stability, although this effect was considerably weaker than the enhancement of translational efficiency. The DEN 3' UTR thus has translational regulatory properties similar to those of a poly(A) tail. Its translation-enhancing effect was observed for RNAs with globin or DEN 5' sequences, indicating no codependency between viral 5' and 3' sequences. Deletion studies showed that translational enhancement provided by the DEN 3' UTR is attributable to the cumulative contributions of several conserved elements, as well as a nonconserved domain adjacent to the stop codon. One of the conserved elements was the conserved sequence (CS) CS1 that is complementary to cCS1 present in the 5' end of the DEN polyprotein open reading frame. Complementarity between CS1 and cCS1 was not required for efficient translation.
INTRODUCTION
Dengue virus (DEN) is a member of the mosquito-borne group of flaviviruses that includes yellow fever, Japanese encephalitis, and West Nile (WNV) viruses. It is the causative agent of millions of human infections annually in the tropical regions of the world (14). The virus has a vertebrate host range limited to humans and monkeys, but replication and amplification also occur in mosquito vectors, principally those of the Aedes genus (4).
The DEN genome is an 11-kb positive-strand RNA with a m7GpppA cap at the 5' end but lacking a poly(A) tail at the 3' end (29). It encodes a single long polyprotein (see Fig. 1A) that is processed through the action of viral and host proteinases to generate 10 mature structural and nonstructural proteins. The maturation pathway, the cleavage sites, and the responsible proteinases have been well described (29). However, systematic studies of the roles of the untranslated regions (UTRs) of the viral RNA in the gene expression of DEN or other flaviviruses are lacking. The UTRs are expected to play an important early role during viral infection in coordinating viral gene expression and the onset of RNA replication.
Translation of the typical cellular mRNA relies on the recruitment of ribosomes by features in the UTRs at both ends of the RNA (11, 20, 31, 40, 41). The 5' cap is recognized by the cytoplasmic cap binding protein eIF4E, which contacts initiation factor eIF4G. This large protein acts as a scaffold for further protein-protein interactions that support initiation. One of those interactions contacts eIF3, a complex factor that is itself bound to the small ribosomal subunit. This chain of interactions thus recruits the ribosomal subunit to the 5' terminus of the RNA. The 3' regions of an mRNA also enhance translation. One mechanism by which this can occur is through an interaction between eIF4G and poly(A) binding protein that is bound to the poly(A) tail (47). The translation enhancing contributions of the cap and poly(A) tail are synergistic (10, 46), and the most actively translated mRNAs are thought to exist in a cyclized or closed-loop state that is considered to enhance initiation and allow for improved ribosome recycling while ensuring that truncated, damaged mRNAs are poorly translated (11, 20, 40). After initial loading at the 5' end of the RNA, the small ribosomal subunit scans in the 3' direction along the RNA in search of an initiation site (27, 38, 39). This is usually the 5'-most AUG triplet, at which initiation frequency is in part determined by the immediate sequence context, GCC(A/G)CCAUGG being the optimal context in mammalian cells (27).
Genomes of positive-strand RNA viruses often lack a 5' cap and/or a poly(A) tail yet must compete with cellular mRNAs for ribosomes to support robust viral gene expression. Viral RNAs that lack a 5' cap often possess an internal ribosome entry site in the 5' UTR that directs interaction with the small ribosomal subunit directly or through the binding of an initiation factor, such as eIF4G (8, 17, 39, 41). This can lead to initiation at an AUG triplet that is not 5' proximal, as occurs with the picornaviruses (19) and the members of two genera (Hepacivirus and Pestivirus) of the family Flaviviridae (29). Since the genomes of DEN and other members of the genus Flavivirus have a 5' cap and initiate translation at the 5'-most AUG triplet, standard cap-dependent initiation rather than internal ribosome entry has been considered to be the gene expression strategy used by these viruses (29).
The 3' UTRs of some viral RNAs that possess a 5' cap but lack a poly(A) tail promote translation to an extent similar to that of UTRs with a poly(A) tail, and through a similar synergistic interaction with the cap. This is true of rotavirus mRNA (49), whose 3'-terminal GACC sequence acts as a translational enhancer (5) by binding the viral protein NSP3, which competes with poly(A) binding protein for interaction with eIF4G (49). It is also true of some plant viral RNAs whose 3' UTRs terminate in a tRNA-like structure, like those of Tobacco mosaic virus, Brome mosaic virus (12, 13), and Turnip yellow mosaic virus (33), although in these cases, the molecular interactions have not been identified.
We report here a systematic analysis of the properties of the 5'- and 3'-terminal regions of the DEN serotype 2 genome in directing translation in Vero cells, using a luciferase (LUC) reporter system. Both 5' and 3' DEN sequences were shown to participate in the regulation of expression, principally by modulating translational efficiency rather than RNA stability.
MATERIALS AND METHODS
Clones and cell culture. Dengue virus serotype 2 (DEN2) sequences were derived from the pD2/IC-30P-A infectious cDNA clone of virus strain 16681 (GenBank accession no. U87411) (23). West Nile virus sequences were derived from genomic clones of a New York flamingo isolate (GenBank accession no. AF196835). Sindbis virus sequences were obtained from pToto54, a full-length genomic clone (GenBank accession no. NC_001547) provided by James Strauss (Caltech). Vero (African green monkey kidney) cells were grown in Dulbecco's modified Eagle's medium (DMEM) (Gibco) containing antibiotics and 10% fetal bovine serum at 37°C in an atmosphere containing 5% CO2.
Plasmid constructs. The basic plasmid for the in vitro transcription of reporter mRNA with DEN2 5' and 3' UTRs and encoding firefly LUC was constructed in the following steps to produce pDLD, shown in Fig. 1B. PCR amplification with the appropriate sequence-modifying primers was used to produce a DNA fragment comprising the DEN2 5' UTR fused to the T7 promoter and flanked by a NotI site at the 5' end and a PstI site at the 3' end. The PstI site was positioned to correspond to the third and fourth codons of the reporter open reading frame (ORF) (encoding Leu-Gln). The PCR product digested with PstI was cloned into the PstI and blunted HindIII sites of pUC18. The LUC ORF was modified during PCR amplification by the addition of an in-frame PstI site in front of the second codon of the LUC ORF (Glu) and by silent mutation of the last two codons to incorporate a HindIII site immediately before the termination codon. At the same time, a BamHI site was added on the 3' side of the LUC ORF, and the PCR product digested with PstI and BamHI was cloned into the same sites of the plasmid described above. Finally, the DEN2 3' UTR was added into the HindIII and SmaI sites after PCR amplification to yield the 3' UTR with a HindIII site placed immediately in front of the native DEN2 TAG stop codon and a SnaBI site appended to the 3' end.
Derivatives of pDLD (see Fig. 1B) were made by analogous PCR amplification of the desired 5' and 3' UTR fragments as described above and subcloning into the NotI-PstI sites and HindIII-KpnI sites, respectively (a KpnI site is present in the polylinker downstream of SnaBI). The UTRs of rabbit -globin mRNA (GenBank accession no. J00658) were made by annealing synthetic DNA fragments. Substitution mutation and deletion variants were made by PCR-mediated mutagenesis and subcloning. A minimal 3' UTR () was made by the removal of sequences between the BamHI and XhoI sites of an intermediate plasmid, resulting in a 15-nucleotide (nt) nonviral 3' UTR (sequence AAAUGGAUCUCGA). The sequences of all subcloned segments were verified by DNA sequencing. Table 1 lists the relevant sequences of clones reported here.
RNA transcription and transfection of cells. Plasmids of the pDLD family were linearized by cleavage with a restriction enzyme as indicated in Fig. 1B. RNAs were produced in vitro by transcription with T7 RNA polymerase and uniformly labeled with low levels of [-32P]CTP (0.2 μCi per 20 μl reaction mixture) to facilitate quantification by liquid scintillation counting of trichloroacetic acid-precipitable material; 5'-capped RNAs were produced in the presence of a sixfold excess of m7GpppG (Epicentre) over GTP (32). Template DNA was removed by digestion with RNase-free DNase, and RNA quality was analyzed via native 1% agarose gel electrophoresis.
In vitro-transcribed reporter RNA (1 pmol) was electroporated into 6 x 105 trypsinized Vero cells in electroporation buffer (25% DMEM and 75% "cytosalts," which contain 120 mM KCl, 0.15 mM CaCl2, 10 mM potassium phosphate, pH 7.6, 5 mM MgCl2) by using 0.2-cm cuvettes and a Gene Pulser Xcell device (Bio-Rad) as recommended by the manufacturer (the protocol can be found at the BTX website [http://www.btxonline.com/applications/protocols/mammalian/]). Electroporated cells were diluted with 4 volumes of DMEM containing 10% fetal bovine serum, transferred to tissue culture plates, and held at 37°C in an atmosphere containing 5% CO2. Cells were harvested at various time points by lysis with 50 μl of 1 x passive lysis buffer (Promega). A number of the experiments reported in this paper were also conducted by using lipofection to deliver RNAs into cells (as used in reference 18). Although broadly similar results were obtained, we have found electroporation to be more consistent, and it permits direct estimation of translational efficiency and functional RNA half-life.
Analysis of luciferase activity. Portions of each extract (10 μl) were loaded into the wells of clear-bottom, black, 96-well plates and mixed with 50 μl of luciferase assay reagent (Promega) for luminometry in a 1450 MicroBeta TriLux counter (Wallac). Luciferase activities were normalized to the total protein concentration determined with protein assay reagent (Bio-Rad) and expressed as relative light units per mg protein.
Assessment of physical stabilities of RNAs after transfection. RNA transcripts were uniformly radiolabeled in the presence of [-32P]CTP and delivered by electroporation as usual into Vero cells. At various time points after electroporation, RNAs were extracted by use of Trizol (Invitrogen), denatured by glyoxalation, and separated by electrophoresis in 1% agarose gels run in phosphate buffer (32). After being dried, gels were exposed to phosphorimager plates, and full-length RNA bands were quantitated after phosphorimagery.
In vitro translation and assessment of the relative specific activities of LUC variants with altered N termini. LUC reporter RNAs were translated in rabbit reticulocyte lysate (Promega) in the presence of [35S]methionine (Amersham). Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and dried gels were exposed to phosphorimager plates. Luciferase bands were quantitated after phosphorimagery. Dilutions of translation reactions were used for estimation of LUC activity in the standard assay. Taking into account the number of methionine residues in each form of LUC, LUC-specific activities were expressed as relative light units per pmol of LUC fusion protein.
RESULTS
Use of luciferase as a reporter for studying the effects of DEN RNA terminal regions on translational expression in primate cells. Firefly LUC has been widely used as a reporter in translational expression studies because of its sensitivity and ease of assay. We modified a cloned LUC coding region to place it under the control of a T7 promoter and to permit convenient cassette replacement of upstream and downstream control regions (Fig. 1B). In vitro transcription with T7 RNA polymerase of linearized reporter constructs produced defined mRNA transcripts that were electroporated into Vero (monkey kidney) cells for subsequent assay of LUC activity in cell lysates.
The basic LUC reporter construct with 5'- and 3'-terminal regions from DEN2 RNA, from which DLD RNA (DLD was named for DEN 5' UTR/LUC/DEN 3' UTR) can be transcribed, is shown in Fig. 1B. A unique NotI site was placed upstream of the T7 promoter, which is immediately adjacent to the 96-nt-long DEN2 5' UTR. The 5' UTR was fused to the first two codons of the polyprotein ORF in order to preserve the natural AUG context, and a unique PstI site was engineered in frame between these codons and the LUC ORF. To facilitate T7 transcription, the 5' A residue of the genomic sequence was replaced with a G. At the other end of the LUC ORF, the last two codons were modified with silent mutations to incorporate a HindIII site just upstream of the stop codon. The native DEN2 stop codon and entire 451-nt-long 3' UTR was placed between this HindIII site and a SnaBI site placed at the 3' end (Fig. 1B). Linearization of pDLD plasmid DNA with SnaBI permits the production of transcripts with only three nonnative nucleotides at the 3' termini. To produce RNAs with other UTRs, pDLD was modified by the replacement of NotI-PstI and HindIII-SnaBI fragments with various PCR-generated inserts.
We constructed pDCLD (DCLD was named for DEN/Capsid/LUC/DEN) (Fig. 1B) to investigate the influence of the 5'-proximal part of the DEN ORF, which includes the cCS1 element that permits RNA cyclization by hybridization with the CS1 element in the 3' UTR (Fig. 1A). pDCLD includes the first 36 codons of the DEN capsid coding region fused in front of the PstI site and placed in frame upstream of the LUC coding sequence (Fig. 1B). This results in the production of a capsid-LUC fusion protein. To compare translation driven by viral sequences to that driven by UTRs from a highly expressed cellular mRNA, we have made pGLGpA and pGCLGpA (Fig. 1B). These constructs include the 5' and 3' UTRs derived from rabbit -globin mRNA and a poly(A) tail 72 residues long. In pGLGpA, the first two codons of the -globin ORF are fused to the PstI site and in frame with the LUC coding sequence, while in pGCLGpA, the entire coding region is identical to that in pDCLD. RNAs with and without a poly(A) tail can be made from pGLGpA and pGCLGpA after linearization with Acc65I and XhoI, respectively.
The specific activities of all the forms of N-terminally modified LUC made by these and other RNAs studied in this paper have been determined and have been verified to be similar (Fig. 2). This was assessed by determination of the light yields from known amounts of [35S]methionine-labeled LUC proteins synthesized by in vitro translation. We have previously observed consistent specific activities for LUC forms with different N-terminal extensions (32).
Flavivirus UTRs support highly efficient translation in Vero cells. To assess the overall efficiency with which flavivirus UTRs are able to support translation in Vero cells, we studied LUC expression from reporter RNAs with DEN or WNV UTRs: DLD, DCLD, WLW (named for WNV 5' UTR/LUC/WNV 3' UTR), and WCLW (named for WNV/Capsid/LUC/WNV) RNAs (Fig. 1B). Expression was compared to that from GLGpA RNA and from SinLSin RNA, which have -globin and Sindbis virus UTRs, respectively (Fig. 1B). SinLSin RNA was included as a representative of a virus capable of rapid amplification and highly efficient gene expression (44). All RNAs contained 5' m7GpppG caps (Fig. 3). In time courses following LUC expression up to 9 h after RNA delivery by electroporation, DLD, WLW, and SinLSin RNAs supported the synthesis of considerably more LUC activity than did GLGpA RNA (Fig. 3A).
We derive three parameters relevant to UTR function from LUC expression time courses. Estimates of the maximum rate of increase in LUC activity (linear rate [see Fig. 3 to 7]), which is observed during an early phase of linear increase, are taken to reflect translational efficiencies (9, 32). The maximum accumulation of LUC activity reflects the expression capacity of each RNA. This capacity is influenced by the translational efficiency of the mRNA and by the stabilities of the mRNA and translated protein. The combined RNA and protein stabilities can be monitored by determining the longevity of continued LUC expression. From the time courses, we estimate the half-life of LUC expression as the time taken for the maximum rate of expression (discussed above) to fall to half that rate. Rates are determined as the slopes of the lines of best fit for the expression time courses. For reporter RNAs encoding the same protein, differences in LUC half-life reflect differences in RNA functional stability (the RNA's capacity for directing protein synthesis) (9, 32). In cases in which the N termini of the encoded LUC proteins are different, differences in protein stability may also be reflected in the LUC protein half-life. The expression half-lives of LUC expressed from DLD, WLW, SinLSin, and GLGpA RNAs were all similar (Fig. 3A). LUC synthesis from the RNAs with viral UTRs were expressed with two- to threefold higher translational efficiency than LUC expressed from GLGpA RNA (Fig. 3A, linear rate). The flaviviral reporter RNAs expressed LUC at least as efficiently as did SinLSin RNA. LUC expression from WLW and SinLSin RNAs consistently appeared earlier than from DLD RNA (Fig. 3A), suggesting differences in the ability to recruit ribosomes for the initial round of translation.
Variants of DLD and WLW that included the first 36 and 31 codons, respectively, of each viral capsid coding region (DCLD and WCLW RNAs [Fig. 1B]) were constructed to more closely represent the translation of natural viral RNAs. The presence of additional 5' viral sequences allows local folding into a predicted stem-loop that includes the initiation AUG (mfold [52]) and long-distance pairing between CS1 and cCS1. Further, under some circumstances, the proximal part of an ORF can affect the efficiency of its own translation (7, 26). The same fragment of the DEN capsid ORF was present in GCLGpA RNA as a control. The specific activities of the capsid-LUC fusion proteins varied little from those of the corresponding LUC proteins lacking the capsid ORF (Fig. 2B), while expression half-lives were marginally shorter (Fig. 3). The decrease in LUC activities after 6 h suggests a shorter half-life for the capsid-LUC fusion proteins.
The efficiency and maximal expression of LUC from DCLD and WCLW RNAs were two- to fivefold lower than those of LUC from DLD and WLW RNAs, respectively (Fig. 3B and A, respectively). Nevertheless, DCLD and WCLW RNAs supported strong LUC expression at levels similar to that of GLGpA RNA. The presence of DEN capsid coding sequences in GCLGpA RNA had little effect on LUC expression (Fig. 3, compare GCLGpA and GLGpA), in contrast to their effect on DLD RNA. Decreased expression is, thus, not a property of the encoded protein but likely a property of the RNA (such as different folding). The capsid-coding regions include cCS1, thereby permitting long-distance hybridization to the complementary sequence, CS1, that is present in the viral, but not globin, 3' UTRs. Together with additional long-distance 5'-3' pairing involving sequences just upstream of the initiation AUG as suggested recently (48), these RNA-RNA interactions may dampen the efficiency of ribosome access in the case of RNAs with viral 5' and 3' terminal regions.
Translational regulation by DEN 5' and 3' UTRs. Efficient translation of cellular mRNAs depends heavily on the 5' cap and 3' poly(A) tail, which enhance expression synergistically (10, 46). This response was confirmed for Vero cells by using transcripts from pGCLGpA. Detecting synergy requires study of the expression of capped and uncapped versions of an RNA, as well as variants with and without a poly(A) tail. In the presence of a 5' cap, the addition of a polyadenylated globin 3' UTR (3'-GpA) enhanced translational efficiency (linear rate) by 108-fold with reference to an RNA with a minimal vector-derived 3' UTR (Fig. 4A and C; compare GCLGpA and GCL). Far less enhancement (5.5-fold) in response to adding 3'-GpA when RNAs were uncapped was observed (Fig. 4B and C; compare GCLGpA and GCL). The ratio between these responses (108/5.52 = 19.6) represents synergistic enhancing effects between the 5' cap and polyadenylated 3' UTR, implying collaboration between these features. Note that the presence of 3'-GpA also increased the LUC expression half-life by threefold (82/27) (Fig. 4A). Since GCLGpA and GCL RNAs encode the same form of LUC, we can conclude that there is a threefold difference in the functional half-lives of these RNAs.
Similar analyses were conducted to study the contribution of the DEN 3' UTR to translational expression. Like the 3'-GpA, the DEN 3' UTR is clearly important for high-level expression of LUC (Fig. 4A, compare DCLD and DCL RNAs). The addition of the DEN 3' UTR to DCL RNA (to produce DCLD RNA) enhanced translational efficiency 32-fold in the presence of a 5' cap (Fig. 4A and C; compare DCLD and DCL) and 10.5-fold in the absence of a cap (Fig. 4B and C; compare DCLD and DCL). This indicates the existence of a synergistic interaction between the cap and DEN 3' UTR (synergy value, 32.3/10.5 [3.1]), although the interaction is considerably less strong than that between a cap and 3'-GpA.
Studies with the uncapped forms of RNAs, which support relatively low levels of LUC expression, emphasized the importance of a 5' cap. Among uncapped RNAs, however, elevated levels of LUC expression from RNAs with DEN UTRs were observed. This is obvious in the case of DCLD RNA in the time course shown in Fig. 4B. An enhancing effect of 5' DEN sequences can be seen for DCL RNA in comparison with GCL RNA, while an effect of the DEN 3' UTR is evident when comparing expression levels from GCLD and GCLGpA RNAs (Fig. 4B and C). The enhancing effect was stronger for the DEN 5' UTR, which supported the synthesis of LUC 6 to 12 times more efficiently (Fig. 4B and C, linear rates) than 5' G in RNA counterparts with the same 3' UTR. No such 5' DEN-dependent enhancement in the capped state was seen, since capped DCLD, GCLD, and GCLGpA RNAs were similarly expressed (Fig. 4A).
LUC expression half-life estimates indicated that the GpA and DEN 3' UTRs and the 5' cap all contribute to RNA stability. Capped DCLD, GCLD, and GCLGpA RNAs (which encode the same form of LUC) had similar functional half-lives, and the removal of the 3' UTR decreased the half-life by about threefold (Fig. 4A, compare DCL and GCL RNAs). Uncapped DCLD RNA had a half-life similar to that of capped DCL or GCL RNAs, evident from the early peak in LUC activity (see the DCLD time course in Fig. 4B). While RNA stability differences associated with the 5' cap and 3' UTR are important, they were much smaller in magnitude than the enhancements in translational efficiency provided by these features. Analyses of the physical stabilities of RNAs during expression time courses confirmed that RNA stability differences were relatively small among the RNAs studied in Fig. 4 (Fig. 4D). In these assays, 32P-labeled RNAs were electroporated and incubated in cells before being extracted and examined by electrophoresis (see Materials and Methods).
Expression from DCLGpA and GCLD RNAs was studied to assess whether there are any cooperative effects between DEN 5' and 3' UTRs. LUC expression from both of these capped RNAs was comparable to expression from capped DCLD and GCLGpA RNAs (Fig. 4A), clearly indicating that the DEN UTRs do not need to be coupled to support strong translation. The addition of the DEN 3' UTR to GCL RNA enhanced translational efficiency 174-fold in the presence of a 5' cap and 20.6-fold with no cap (Fig. 4C). Thus, the synergy between the 5' cap and DEN 3' UTR (synergy value, 174/20.6 [8.4]) for these RNAs with a globin 5' UTR is stronger than that observed for RNAs with the DEN 5' UTR (synergy value of 3.1; see above). As evident when comparing DCLD and DCL RNAs, the DEN 3' UTR in GCLD RNA enhanced the RNA functional half-life about threefold (Fig. 4A). We conclude that the DEN 3' UTR is a general enhancer of translation and RNA stability whose action does not require a matched DEN 5' UTR.
Complementarity between cyclization sequences CS1 and cCS1 is not needed for translation. The analyses already discussed failed to detect any codependency between 5' and 3' DEN sequences in regard to the control of translation. To explore this issue further, we wished to test directly for a translational role of the conserved cyclization sequences and their potential long-distance hybridization. The RNAs of Fig. 5A represent a set in which CS1 and cCS1 have replaced one another singly and together, resulting in DCmLD and DCLDm RNAs that have diminished potential for long-distance cCS1/CS1 interaction and double-mutant DCmLDm RNA, in which CS1 and cCS1 have been switched and cyclization is possible. Note that one nucleotide of each transposed conserved sequence was changed to avoid introducing a stop codon with CS1 placed in the 5' location. Expression from capped forms of these RNAs in Vero cells was studied.
Replacement of cCS1 with the modified CS1 sequence had little impact on LUC expression or expression half-life (Fig. 5, compare DCLD and DCmLD). By contrast, replacement of CS1 with the modified cCS1 decreased LUC expression efficiency to 19%, with only a small effect on expression half-life (Fig. 5, compare DCLD and DCLDm). CS1 and cCS1 thus appear to have differential roles in translational regulation. Significantly, the inability of the double mutation (DCmLDm) to rescue the depressed LUC expression level of DCLDm RNA (Fig. 5B and C) suggests that RNA cyclization via cCS1/CS1 interaction is not required for the translation of the reporter RNAs used here. The minimal differences in the half-lives of LUC expression from the RNAs in Fig. 5 indicate that these RNAs have similar stabilities.
Contribution of conserved elements in the 3' UTR to translational expression. The results described above indicated an involvement of CS1 in translation. In order to identify contributions from other elements in the 3' UTR, the conserved features SLA, SLB, DB1, and DB2, as well the upstream variable region (UVR) (Fig. 1A and 6A), were deleted in turn from the 3' UTR of capped DCLD RNA. The endpoints of deletions were chosen to cleanly delete the predicted structures as shown in Fig. 6A, and this was confirmed by mfold (52) predictions of the folding of the 3' UTRs. A large 3' truncation was made by transcription from a pDCLD template linearized with NcoI, which cleaves at DEN nucleotide 10470 (Fig. 6A). This produced a variant of DCLD RNA (Fig. 6, /NcoI) truncated near the upstream end of DB2 and thus lacking SLA, SLB, CS1, DB1, and most of DB2. LUC expression from capped RNAs in Vero cells was studied (Fig. 6).
The deletion of either of the two 3'-terminal stem-loops SLA and SLB decreased the efficiency of LUC translation to about half (Fig. 6B, linear rate). The deletion of DB1 resulted in a smaller, though significant, decrease in the efficiency of LUC translation, while the deletion of DB2 had little effect (Fig. 6C). The simultaneous deletions of DB1 and DB2 reduced the efficiency of translation to 25%, suggesting redundancy in the contributions of these quite similar features. The deletion of the UVR, just downstream of the stop codon, also resulted in a relative translational efficiency of 25% (Fig. 6D).
Translational enhancement provided by the 3' UTR thus seems to derive from several features, including contributions by each of the conserved features. The individual deletions of these features led to a partial loss of activity. Expression from each of the RNAs with a single element deleted was higher than that from RNA entirely lacking a DEN 3' UTR (Fig. 7A, DCL). Presence of the UVR feature alone (Fig. 7A, /NcoI RNA) increased expression only slightly from this basal level (Fig. 7A), emphasizing the importance of combined contributions to translational enhancement by the 3' UTR.
Interestingly, DCLG RNA, which has the 87-nucleotide-long globin 3' UTR but no poly(A) tail, supported the same basal level of LUC expression as DCL RNA (Fig. 7). High levels of expression were dependent on the poly(A) tail (Fig. 7A, DCLGpA RNA). This emphasizes the similar roles of the DEN 3' UTR and poly(A) tail in enhancing translation.
It is important to note that no noteworthy differences in functional RNA stability were observed for the RNAs with various partial deletions of the DEN 3' UTR, although most of these capped RNAs had somewhat shorter half-lives than capped DCLD RNA (Fig. 7B); this conclusion was supported by an analysis of the physical stabilities of these RNAs (not shown). The RNAs containing partial 3' deletions were all more stable than DCL RNA (Fig. 7B). It appears that, as for translational enhancement, RNA stabilization (roughly threefold) provided by the DEN 3' UTR (noted earlier) derives from multiple features in the UTR. The globin 3' UTR also appears to enhance RNA stability (twofold stabilization observed for DCLG RNA relative to DCL [Fig. 7B]), with little if any further stabilization provided by a poly(A) tail (Fig. 7B, DCLGpA RNA). The contributions to translational enhancement were just the opposite, being primarily derived from the poly(A) tail and not measurable from the 3' G sequence (Fig. 7A).
Although it has been observed that a 3' UTR can stabilize RNA and enhance translation by virtue of its length, in a sequence-nonspecific effect (45), this is unlikely to be a consideration among the 3' UTR domain deletions. Those deletions remove no more than 154 nt (DB1 + 2), leaving at least a 297-nt-long UTR, and, conversely, the addition of the 87-nt 3' G did not enhance translation (Fig. 7A). The nonspecific length effects were observed for 3' UTRs shorter than 27 nt (45).
DISCUSSION
Flaviviral 5'- and 3'-terminal genome sequences support highly efficient translational expression. Using a luciferase reporter system and comparison with reporter RNAs that support highly efficient translation, we have shown that the 5'- and 3'-terminal regions from the DEN2 and WNV genomes are able to support strong translation in Vero cells. Translational expression levels from 5'-capped DCLD and WCLW RNAs with 5' and 3' sequences from DEN and WNV, respectively, were similar to the level of expression from capped GLGpA RNA and only two to three times lower than the level of expression from SinLSin RNA (Fig. 3). GLGpA RNA has UTRs from the -globin mRNA, while SinLSin RNA has UTRs derived from Sindbis virus, an alphavirus that has been developed as a gene expression vector because of its ability to support potent translational expression (1).
The viral sequences present in DCLD and WCLW RNAs included part of the capsid coding region in addition to the 5' and 3' UTRs. The capsid coding regions were included because they contain the conserved cCS1 sequence element that permits potential cyclization of the RNA by annealing to its complement, CS1, present in the 3' UTR. A surprising finding was that the omission of the capsid coding regions (DLD and WLW RNAs) increased translational expression by two- to fourfold (Fig. 3). This is not due to an elongational block of translation associated with the capsid ORF, since LUC was similarly expressed from GLGpA and GCLGpA RNAs (Fig. 3). We have also verified that the LUC and capsid-LUC fusion proteins have similar light-emitting specific activities (Fig. 2). It seems most likely that the presence of the proximal part of the capsid ORF lessens translational efficiency by affecting ribosome initiation, perhaps by altering the folding of the 5' UTR through local interactions or as a consequence of long-distance hybridization with CS1.
In any case, it is clear that the capped flaviviral 5' UTRs are capable of directing very high translational expression in Vero cells. This was rather unexpected in view of the conserved secondary structure positioned immediately adjacent to the 5' end (2) and suboptimal consensus around the initiation AUG (UCUCUGAUGA for DEN and AUCUCGAUGU for WNV). Secondary structure at the 5' terminus is detrimental to cap-dependent translation initiation (25), and weak consensus nucleotides at the –3 and +4 positions (italicized respectively in the sequences above) can diminish translation by up to 10-fold (–3A or –3G and +4G is optimal [27]). Schrader and Westaway (42) have also commented on the efficient translation of flaviviral (Kunjin virus) RNA in Vero cells despite suboptimal AUG context.
The DEN 3' UTR enhances translation and RNA stability in a way similar to that of a poly(A) tail. The contribution of the DEN 3' UTR to translational expression was studied by determining the effect of replacing a short nonviral 3' UTR (DCL or GCL RNAs) with the DEN 3' UTR (DCLD and GCLD RNAs). For capped RNAs with either the DEN or globin 5' UTRs, the DEN 3' UTR significantly increased RNA stability (about a threefold increase in functional half-life) (Fig. 4A). Far greater increases in translational efficiency were observed, represented by the maximum LUC expression rate in our assays (Fig. 3 to 6, linear rates). In the presence of a DEN 5' UTR, the rate was increased 32.3-fold (Fig. 4C, compare DCLD and DCL), while in the presence of a globin 5' UTR, the increase was 174-fold (Fig. 4C, compare GCLD and GCL). These responses are very comparable to those provided by a polyadenylated 3' UTR. Capped GCLGpA RNA was three times as stable and was translated with 108 times the efficiency of capped GCL RNA (Fig. 4C), and capped DCLGpA RNA was translated with 70 times the efficiency of capped DCL RNA (Fig. 7A). The ability of the DEN 3' UTR to enhance translational efficiency and RNA stability in the presence of either DEN or globin 5' sequences demonstrates this to be a free-standing regulator that is not codependent on viral 5' sequences. Stimulation of the translation of LUC reporter mRNAs by the DEN2 3' UTR has also recently been reported by Holden and Harris (18). Extending those studies, our kinetic analyses have enabled estimation of the contributions of both increased RNA stability and translational efficiency to expression enhancement by the 3' UTR.
The poly(A) tail is known to enhance translation synergistically with a 5' cap (10, 46). Such synergy is established by determining the extent to which the addition of a poly(A) tail has a greater effect in the presence of a cap than in the absence of a cap. The 3'-GpA enhanced translational efficiency 19.6 (108/5.52) times more when a cap was present (Fig. 4C, GCLGpA and GCL RNAs), confirming the strong synergy that others have reported. Similar analyses of the role of the DEN 3' UTR also indicated the existence of synergy with a 5' cap, although this synergy was weaker than that observed for the case with a poly(A) tail and it differed for different 5' UTRs. Thus, the cap/3' DEN synergy value was 8.4 for RNAs with a globin 5' UTR (ratio for GCLD and GCL RNAs, 174/20.6 [Fig. 4C]) and 3.1 for RNAs with a DEN 5' UTR (ratio for DCLD and DCL RNAs, 32.3/10.5 [Fig. 4C]).
If cap/poly(A) synergy reflects the existence of bridging interactions that cyclize the mRNA as described in the introduction, we may think of the DEN 3' UTR recruiting proteins capable of interacting with translation initiation factors assembled around the 5' cap. Such a scenario has been demonstrated for rotaviral mRNAs, with the bridging interaction mediated by the viral protein NSP3 (49). For the DEN 3' UTR, however, the synergy we have observed does not involve a viral protein, since the only viral protein present in our assays is a fragment of the capsid protein.
The quite significant difference in synergy observed with globin or DEN 5' UTRs is an interesting observation. Both the GCLD and DCLD RNAs used in our synergy studies possess the cCS1 and CS1 elements and therefore have the potential to cyclize through long-distance hybridization. Although this annealing would be disrupted by elongating ribosomes passing through cCS1, it may influence the initiation of translation. By forming a closed loop through direct RNA-RNA interaction rather than though protein mediation, translation may potentially be enhanced. This type of effect has been observed with Barley yellow dwarf virus RNA through direct RNA interaction between 5' and 3' UTRs (15). Alternatively, cCS1/CS1 interaction may be inhibitory towards translational initiation. This is suggested by the level of LUC expression from DCLD RNA being lower than that from DLD (Fig. 3) or DCLGpA RNA (Fig. 4), which contain only CS1 or cCS1. Recent modeling of the RNA folding of flaviviral RNAs has suggested the existence of RNA pairing between part of the 5' UTR just upstream of the AUG initiation codon and the 3' UTR (48). For DCLD RNA but not GCLD RNA, this pairing would occur in addition to the cCS1/CS1 interaction. Further experiments are needed to investigate whether these combined interactions are responsible for the cap/3' UTR synergy observed for DCLD RNA being lower than that for GCLD RNA and also for the translational expression level from DCLD RNA being lower than that from DLD RNA (Fig. 3).
Complementarity between cCS1 and CS1 is not necessary for efficient translation driven by DEN UTRs. The speculation above explored a possible inhibitory interaction between 5' and 3' genomic sequences that has yet to be tested experimentally. We did, however, test the opposite effect, that is, whether cCS1/CS1 complementarity and these sequences themselves are required for efficient translation. Replacement of cCS1 with a modified version of CS1 had little influence on translational expression (Fig. 5C, DCmLD). By contrast, replacement of CS1 decreased translational efficiency to about one-fifth (Fig. 5C, DCLDm), and there was a further reduction when these mutations were combined to permit potential hybridization and RNA cyclization (Fig. 5C, DCmLDm). These results indicate that cCS1/CS1 complementarity is not an attribute that supports translation. Indeed, the marginally increased translational efficiency of DCmLD RNA over DCLD RNA and the lower efficiency of DCmLDm DNA relative to DCLDm RNA support the idea that the long-distance RNA-RNA interactions interfere with translation. Such long-distance interactions have been implicated in flaviviral RNA replication (6, 21, 30, 50, 51), and it is intriguing to consider that this interaction may play a role in regulating the transition from the translation to the replication phase of the infection. If the cCS1/CS1 interaction can be stabilized, for instance, by the influence of bound proteins, this interaction might simultaneously inhibit translation and enhance replication. We have recently described an analogous switch involving translation factor eEF1A binding to the 3'-terminal region of Turnip yellow mosaic virus RNA (33, 34).
The low translational expression from DCLDm RNA suggests an independent role for CS1 in supporting translation. Inhibition of such a role may help to explain the potent inhibition of DEN replication by an antisense molecule directed at CS1 (24). No such role in translation was revealed by the only previous study of conserved sequence involvement in translational expression, performed using a WNV reporting replicon (30). In that study, neither the mutation of cCS1 nor that of CS1 detectably influenced the yield of the LUC reporter protein during the initial phase of viral gene expression that was not dependent on RNA replication. Further studies involving a variety of CS1 sequence variants and a comparison of the DEN and WNV systems are needed to clarify the situation. There may also be sensitivity differences between the replicon assay and the mRNA reporter assays we have used.
Multiple features of the DEN 3' UTR enhance translational expression. CS1 is only one of a number of conserved elements present in the 3' UTRs of flaviviral genomes (Fig. 1). By deleting each of the other elements (SLA, SLB, DB1, and DB2) from DCLD RNA, we tested their contributions to the stimulatory effect of the DEN 3' UTR. Our results suggest, for the first time, that both RNA stability and translational enhancement are conferred by multiple features (Fig. 7). At this point, we cannot be certain that lost function is directly due to the absence of the deleted element. Indirect effects due to structural rearrangements in the deleted 3' UTRs could have occurred, even though folding predictions suggested that this was not the case. Using a reporting replicon, Lo et al. (30) recently reported that each of the conserved sequence elements in the 3' UTR of WNV RNA could be mutated or deleted without noticeably affecting translational expression in BHK cells. As noted above, these differences between the DEN reporter RNA and WNV replicon systems will need to be explored.
Among the features that contributed to fully efficient translation were the conserved 3'-terminal stems SLA and SLB. The deletion of either of these decreased translational efficiency to 40 to 50% of that of DCLD RNA (Fig. 7). Holden and Harris (18) have also recently observed that simultaneous deletions of DEN2 SLA and SLB ("3'SL" in their nomenclature) resulted in a similar decrease in translational expression in BHK cells from a reporter RNA like ours. In contrast, Brinton and colleagues have viewed SLA and SLB as negative regulators of translation in the WNV system. That conclusion arose from the observation that addition of WNV SLA and SLB to the 3' UTRs of reporter mRNAs suppressed translation considerably in vitro, though only marginally in BHK cells (28). Further investigation will be needed to explain these differences.
ACKNOWLEDGMENTS
We thank Andrew Regan and Daiki Matsuda for assistance with some experiments, James Strauss for providing pToto54, and George Rohrmann, Connie Bozarth, and Doug Leisy for critical reading of the manuscript.
Andrew Regan was the recipient of a summer undergraduate research fellowship funded by Howard Hughes Medical Institute Grant 57003741 to Oregon State University. We are grateful for the use of the Cell Culture facility and of a 96-well plate densitometer of the Environmental Health Sciences Center at Oregon State University, which is supported by grant number P30 ES00210 from the National Institute of Environmental Health Sciences, National Institutes of Health. These studies were supported by National Institutes of Health grant AI055569 awarded to T.W.D.
REFERENCES
Agapov, E. V., I. Frolov, B. D. Lindenbach, B. M. Pragai, S. Schlesinger, and C. M. Rice. 1998. Noncytopathic Sindbis virus RNA vectors for heterologous gene expression. Proc. Natl. Acad. Sci. USA 95:12989-12994.
Brinton, M. A., and J. H. Dispoto. 1988. Sequence and secondary structure analysis of the 5'-terminal region of flavivirus genome RNA. Virology 162:290-299.
Brinton, M. A., A. V. Fernandez, and J. H. Dispoto. 1986. The 3'-nucleotides of flavivirus genomic RNA form a conserved secondary structure. Virology 153:113-121.
Burke, D. S., and T. P. Monath. 2001. Flaviviruses, p. 1043-1125. In D. M. Knipe and P. M. Howley (ed.), Fields virology, 4th ed. Lippincott Williams & Wilkins, Philadelphia, Pa.
Chizhikov, V., and J. T. Patton. 2000. A four-nucleotide translation enhancer in the 3'-terminal consensus sequence of the nonpolyadenylated mRNAs of rotavirus. RNA 6:814-825.
Corver, J., E. Lenches, K. Smith, R. A. Robison, T. Sando, E. G. Strauss, and J. H. Strauss. 2003. Fine mapping of a cis-acting sequence element in yellow fever virus RNA that is required for RNA replication and cyclization. J. Virol. 77:2265-2270.
Frolov, I., and S. Schlesinger. 1996. Translation of Sindbis virus mRNA: analysis of sequences downstream of the initiating AUG codon that enhance translation. J. Virol. 70:1182-1190.
Gale, M., Jr., S. L. Tan, and M. G. Katze. 2000. Translational control of viral gene expression in eukaryotes. Microbiol. Mol. Biol. Rev. 64:239-280.
Gallie, D. R. 2002. The 5'-leader of tobacco mosaic virus promotes translation through enhanced recruitment of eIF4F. Nucleic Acids Res. 30:3401-3411.
Gallie, D. R. 1991. The cap and poly(A) tail function synergistically to regulate mRNA translational efficiency. Genes Dev. 5:2108-2116.
Gallie, D. R. 1998. A tale of two termini: a functional interaction between the termini of an mRNA is a prerequisite for efficient translation initiation. Gene 216:1-11.
Gallie, D. R., and M. Kobayashi. 1994. The role of the 3'-untranslated region of non-polyadenylated plant viral mRNAs in regulating translational efficiency. Gene 142:159-165.
Gallie, D. R., and V. Walbot. 1990. RNA pseudoknot domain of tobacco mosaic virus can functionally substitute for a poly(A) tail in plant and animal cells. Genes Dev. 4:1149-1157.
Gubler, D. J. 2002. Epidemic dengue/dengue hemorrhagic fever as a public health, social and economic problem in the 21st century. Trends Microbiol. 10:100-103.
Guo, L., E. M. Allen, and W. A. Miller. 2001. Base-pairing between untranslated regions facilitates translation of uncapped, nonpolyadenylated viral RNA. Mol. Cell 7:1103-1109.
Hahn, C. S., Y. S. Hahn, C. M. Rice, E. Lee, L. Dalgarno, E. G. Strauss, and J. H. Strauss. 1987. Conserved elements in the 3' untranslated region of flavivirus RNAs and potential cyclization sequences. J. Mol. Biol. 198:33-41.
Hellen, C. U., and P. Sarnow. 2001. Internal ribosome entry sites in eukaryotic mRNA molecules. Genes Dev. 15:1593-1612.
Holden, K. L., and E. Harris. 2004. Enhancement of dengue virus translation: role of the 3' untranslated region and the terminal 3' stem-loop domain. Virology 329:119-133.
Jackson, R. J., and A. Kaminski. 1995. Internal initiation of translation in eukaryotes: the picornavirus paradigm and beyond. RNA 1:985-1000.
Jacobson, A. 1996. Poly(A) metabolism and translation: the closed-loop model, p. 451-480. In J. Hershey, M. B. Mathews, and N. Sonenberg (ed.), Translational control. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
Khromykh, A. A., H. Meka, K. J. Guyatt, and E. G. Westaway. 2001. Essential role of cyclization sequences in flavivirus RNA replication. J. Virol. 75:6719-6728.
Khromykh, A. A., and E. G. Westaway. 1997. Subgenomic replicons of the flavivirus Kunjin: construction and applications. J. Virol. 71:1497-1505.
Kinney, R. M., S. Butrapet, G. J. Chang, K. R. Tsuchiya, J. T. Roehrig, N. Bhamarapravati, and D. J. Gubler. 1997. Construction of infectious cDNA clones for dengue 2 virus: strain 16681 and its attenuated vaccine derivative, strain PDK-53. Virology 230:300-308.
Kinney, R. M., C. Y. H. Huang, B. C. Rose, A. D. Kroeker, T. W. Dreher, P. L. Iversen, and D. A. Stein. 2005. Inhibition of dengue virus serotypes 1 to 4 in Vero cell cultures with morpholino oligomers. J. Virol. 79:5116-5128.
Kozak, M. 1989. Circumstances and mechanisms of inhibition of translation by secondary structure in eucaryotic mRNAs. Mol. Cell Biol. 9:5134-5142.
Kozak, M. 1990. Downstream secondary structure facilitates recognition of initiator codons by eukaryotic ribosomes. Proc. Natl. Acad. Sci. USA 87:8301-8305.
Kozak, M. 1999. Initiation of translation in prokaryotes and eukaryotes. Gene 234:187-208.
Li, W., and M. A. Brinton. 2001. The 3' stem loop of the West Nile virus genomic RNA can suppress translation of chimeric mRNAs. Virology 287:49-61.
Lindenbach, B. D., and C. M. Rice. 2001. Flaviviridae: the viruses and their replication, p. 991-1041. In D. M. Knipe and P. M. Howley (ed.), Fields virology, 4th ed. Lippincott Williams & Wilkins, Philadelphia, Pa.
Lo, M. K., M. Tilgner, K. A. Bernard, and P. Y. Shi. 2003. Functional analysis of mosquito-borne flavivirus conserved sequence elements within 3' untranslated region of West Nile virus by use of a reporting replicon that differentiates between viral translation and RNA replication. J. Virol. 77:10004-10014.
Mangus, D. A., M. C. Evans, and A. Jacobson. 2003. Poly(A)-binding proteins: multifunctional scaffolds for the post-transcriptional control of gene expression. Genome Biol. 4:223.
Matsuda, D., L. Bauer, K. Tinnesand, and T. W. Dreher. 2004. Expression of the two nested overlapping reading frames of turnip yellow mosaic virus RNA is enhanced by a 5' cap and by 5' and 3' viral sequences. J. Virol. 78:9325-9335.
Matsuda, D., and T. W. Dreher. 2004. The tRNA-like structure of Turnip yellow mosaic virus RNA in a 3'-translational enhancer. Virology 321:36-46.
Matsuda, D., S. Yoshinari, and T. W. Dreher. 2004. eEF1A binding to aminoacylated viral RNA represses minus strand synthesis by TYMV RNA-dependent RNA polymerase. Virology 321:47-58.
Molenkamp, R., E. A. Kooi, M. A. Lucassen, S. Greve, J. C. Thijssen, W. J. Spaan, and P. J. Bredenbeek. 2003. Yellow fever virus replicons as an expression system for hepatitis C virus structural proteins. J. Virol. 77:1644-1648.
Olsthoorn, R. C., and J. F. Bol. 2001. Sequence comparison and secondary structure analysis of the 3' noncoding region of flavivirus genomes reveals multiple pseudoknots. RNA 7:1370-1377.
Pang, X., M. Zhang, and A. I. Dayton. 2001. Development of dengue virus type 2 replicons capable of prolonged expression in host cells. BMC Microbiol. 1:18-24.
Pelletier, J., and N. Sonenberg. 1985. Insertion mutagenesis to increase secondary structure within the 5' noncoding region of a eukaryotic mRNA reduces translational efficiency. Cell 40:515-526.
Pestova, T. V., and V. G. Kolupaeva. 2002. The roles of individual eukaryotic translation initiation factors in ribosomal scanning and initiation codon selection. Genes Dev. 16:2906-2922.
Sachs, A. B., P. Sarnow, and M. W. Hentze. 1997. Starting at the beginning, middle, and end: translation initiation in eukaryotes. Cell 89:831-838.
Schneider, R. J., and I. Mohr. 2003. Translation initiation and viral tricks. Trends Biochem. Sci. 28:130-136.
Schrader, A. P., and E. G. Westaway. 1990. Successful competition in translation by the flavivirus Kunjin with poliovirus during co-infections in Vero cells. Arch. Virol. 114:75-89.
Shi, P. Y., M. A. Brinton, J. M. Veal, Y. Y. Zhong, and W. D. Wilson. 1996. Evidence for the existence of a pseudoknot structure at the 3' terminus of the flavivirus genomic RNA. Biochemistry 35:4222-4230.
Strauss, J. H., and E. G. Strauss. 1994. The alphaviruses: gene expression, replication, and evolution. Microbiol. Rev. 58:491-562.
Tanguay, R. L., and D. R. Gallie. 1996. The effect of the length of the 3'-untranslated region on expression in plants. FEBS Lett. 394:285-288.
Tarun, S. Z., Jr., and A. B. Sachs. 1995. A common function for mRNA 5' and 3' ends in translation initiation in yeast. Genes Dev. 9:2997-3007.
Tarun, S. Z., Jr., S. E. Wells, J. A. Deardorff, and A. B. Sachs. 1997. Translation initiation factor eIF4G mediates in vitro poly(A) tail-dependent translation. Proc. Natl. Acad. Sci. USA 94:9046-9051.
Thurner, C., C. Witwer, I. L. Hofacker, and P. F. Stadler. 2004. Conserved RNA secondary structures in Flaviviridae genomes. J. Gen. Virol. 85:1113-1124.
Vende, P., M. Piron, N. Castagne, and D. Poncet. 2000. Efficient translation of rotavirus mRNA requires simultaneous interaction of NSP3 with the eukaryotic translation initiation factor eIF4G and the mRNA 3' end. J. Virol. 74:7064-7071.
You, S., B. Falgout, L. Markoff, and R. Padmanabhan. 2001. In vitro RNA synthesis from exogenous dengue viral RNA templates requires long range interactions between 5'- and 3'-terminal regions that influence RNA structure. J. Biol. Chem. 276:15581-15591.
You, S., and R. Padmanabhan. 1999. A novel in vitro replication system for dengue virus. Initiation of RNA synthesis at the 3'-end of exogenous viral RNA templates requires 5'- and 3'-terminal complementary sequence motifs of the viral RNA. J. Biol. Chem. 274:33714-33722.
Zuker, M. 2003. Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res. 31:3406-3415.(Wei-Wei Chiu, Richard M. )