Cap-independent translation of tobacco etch virus is conferred by an RNA pseudoknot in the 5'-leader.

The tobacco etch virus (TEV) 5'-leader promotes cap-independent translation in a 5'-proximal position and promotes internal initiation when present in the intercistronic region of a dicistronic mRNA, indicating that the leader contains an internal ribosome entry site. The TEV 143-nucleotide 5'-leader folds into a structure that contains two domains, each of which contains an RNA pseudoknot. Mutational analysis of the TEV 5'-leader identified pseudoknot (PK) 1 within the 5'-proximal domain and an upstream single-stranded region flanking PK1 as necessary to promote cap-independent translation. Mutations to either stem or to loops 2 or 3 of PK1 substantially disrupted cap-independent translation. The sequence of loop 3 in PK1 is complementary to a region in 18 S rRNA that is conserved throughout eukaryotes. Mutations within L3 that disrupted its potential base pairing with 18 S rRNA reduced cap-independent translation, whereas mutations that maintained the potential for base pairing with 18 S rRNA had little effect. These results indicated that the TEV 5'-leader functionally substitutes for a 5'-cap and promotes cap-independent translation through a 45-nucleotide pseudoknot-containing domain.

Virtually all eukaryotic cellular mRNAs contain a 5Ј-cap structure (m 7 G(5Ј)ppp(5Ј)N) that, during translation, serves as the binding site for the eukaryotic initiation factor (eIF) 1 4E, a subunit of eIF4F that also contains eIF4G and eIF4A (1,2). eIF4F promotes binding of 40 S ribosomal subunits through the interaction of eIF4G and eIF3, the latter of which is associated with the 40 S ribosomal subunit. The interaction between eIF4G and eIF3 directs 40 S subunit binding at or close to the 5Ј terminus. Following binding, the 40 S subunit scans down the RNA in search of the first AUG codon in a good context at which translation will initiate. The rate of 40 S subunit binding and initiation can be slowed or prevented altogether by the presence of a highly stable secondary structure (3,4).
Because binding of eIF4E is critical for the subsequent binding of the 40 S subunit and assembly of a competent initiation complex, those mRNAs that naturally lack a 5Ј-cap must have evolved alternative mechanisms to promote translation. The only known mRNAs that naturally lack a cap are viral in origin. Poliovirus, encephalomyocarditis virus (EMCV), and foot-and-mouth disease virus are examples of animal picornaviruses whose polyadenylated genomic mRNA lacks a 5Ј-cap structure and instead contains a VPg, i.e. a viral protein genome linked to the 5Ј terminus that is removed prior to RNA recruitment into polysomes (5)(6)(7)(8)(9). The 5Ј-leader of picornaviruses is long (610 to Ͼ1200 nucleotides), highly structured, and contains multiple AUGs upstream of the initiation codon of the polyprotein coding region (10), features that should inhibit normal 40 S ribosomal scanning. Despite these apparent barriers to translation, picornaviral mRNAs are efficiently translated as a consequence of its 5Ј-leader, which contains an internal ribosome entry site (IRES) to which the 40 S ribosomal subunit is recruited (11)(12)(13). For EMCV, the IRES recruits eIF4G which in turn promotes internal binding of the 40 S ribosomal subunit through its interaction with the 40 S subunit-bound eIF3 (14,15). In addition to canonical initiation factors, several cellular RNA-binding proteins that stimulate viral IRES activity have been identified. Polypyrimidine tractbinding protein, poly(rC)-binding protein-2, and unr have been shown to stimulate IRES activity for some but not all IRESs perhaps by acting as RNA chaperones that maintain IRES structure required for 40 S subunit binding (10,16,17).
Tobacco etch virus (TEV) is a member of the picornavirus supergroup of positive-strand RNA viruses but uses plants as its host. Its genomic organization is highly similar to picornaviruses in that the genomic RNA functions as a monocistronic mRNA encoding a single polyprotein that, once produced, is processed by viral-encoded proteases into capsid and noncapsid proteins required for the viral life cycle (18,19). The genomic RNA of TEV is polyadenylated and naturally lacks a 5Ј-cap but is nevertheless efficiently translated. The 143-nt TEV 5Ј-leader is sufficient to confer cap-independent translation even in the absence of the 5Ј-terminal VPg (20,21). Two regions within the 143-nt TEV 5Ј-leader were identified previously as necessary to direct cap-independent translation, and their combinatorial effect was approximately multiplicative, suggesting that the two elements are two functional parts of a single regulatory region (22). The TEV 5Ј-leader functionally interacts with the poly(A) tail to promote optimal cap-independent translation and requires eIF4G and poly(A)-binding protein (22,23). Although the TEV 5Ј-leader evolved to function in a 5Ј-proximal position in a monocistronic mRNA, it also promoted translation of a second cistron in a dicistronic mRNA when the TEV leader sequence was introduced into the intercistronic region, indicating that the TEV 5Ј-leader contains an IRES (22). Although IRES elements from picornaviral RNAs are typically large and highly structured, the TEV element is notable for its small size, representing one of the most compact viral elements identified that can promote cap-independent translation. The precise sequences in the TEV 5Ј-leader required to promote cap-independent translation have not been identified. Moreover, the struc-ture of the TEV 5Ј-leader and its relationship with the function of the 5Ј-leader in promoting cap-independent translation is unknown. In this study, we show that a 5Ј-proximal, 45-nt RNA pseudoknot-containing domain within the TEV 5Ј-leader is required to promote cap-independent translation. Mutations that disrupted the pseudoknot reduced capindependent translation, including mutations to loop 3 that exhibits complementarity to a conserved region in eukaryotic 18 S rRNAs. Changes to the loop that maintained its potential for base pairing with 18 S rRNA had only a small effect, supporting the possibility that base pairing between the pseudoknot and the 18 S rRNA may be involved in promoting cap-independent translation.

EXPERIMENTAL PROCEDURES
Secondary Structure Analysis-The TEV 5Ј-leader sequence used in the experiments represents the 5Ј-proximal 143-nt of the TEV genome. The predicted secondary structure for the TEV 5Ј-leader was calculated through minimization of free energy using the Fold program of the GCG package (Genetics Computer Group, Inc., version 7, University of Wisconsin (24)). Prediction of ternary RNA structure employed the paRNAss and pknotsRG programs (25), and RNA structures were visualized using PSEUDOVIEWER (26). The experimental data were used to correct the predicted models.
RNA fragments were transcribed in vitro from pT7 constructs containing the appropriate TEV sequence. RNAs were 5Ј-end-labeled with [ 32 P]ATP, resuspended into probing buffer, and heated rapidly. RNA folding was promoted by slow renaturation under native conditions (i.e. in the presence of 2 mM MgCl 2 ) in order to refold the molecule into its native conformation and stabilize higher ordered structures. The labeled RNA was subjected to partial digestion with enzymatic probes in the appropriate buffer, and the RNA fragments were resolved on a denaturing TBE (Tris/Borate/EDTA)/urea 10% polyacrylamide gel. A ladder generated by alkaline hydrolysis of RNA was run on the same gel to help identify the cleavage site locations.
mRNA Constructs and in Vitro RNA Synthesis-TEV-luc-A 50 and Con 144 -luc-A 50 luciferase constructs that contain the 143-nt TEV leader sequence or a 144-nt control sequence and that terminate in a poly(A) 50 tract have been described previously (22). The control construct was designed to be 60% AT-rich and nearly the same length as the TEV 5Ј-leader. The free energy calculated by the FOLD algorithm for the control sequence leader is ⌬G ϭ Ϫ11.5 kcal/mol, which is approximately equal to the free energy of ⌬G ϭ Ϫ10.7 kcal/mol of the 5Ј-leader of the TEV-luc-A 50 mRNA construct (27).
Following linearization downstream of the poly(A) 50 tract, the DNA concentration was quantitated spectrophotometrically and brought to 0.5 mg/ml. Uncapped mRNAs were synthesized in vitro as described previously (28) by using 10 g of template DNA in 40 mM Tris-HCl, pH 7.5, 6 mM MgCl 2 , 100 g/ml bovine serum albumin, 0.5 mM each of ATP, CTP, UTP, and GTP, 10 mM dithiothreitol, 0.3 units/l RNasin (Promega), and 0.5 units/l T7 RNA polymerase.

RESULTS
Identification of Sequences within the TEV 5Ј-Leader Required for Cap-independent Translation-Two regions of the TEV 5Ј-leader (i.e. nt 28 -65 and nt 66 -118) that functioned together to promote cap-independent translation were identified previously, suggesting that the region from nucleotides 28 to 118 was required (22). To identify more precisely the functional regions within the TEV 5Ј-leader involved in promoting cap-independent translation, a more detailed deletion analysis was performed. The full-length TEV 5Ј-leader and deletion derivatives were introduced upstream of the luc coding region in a T7 promoter-based construct that enabled in vitro synthesis of uncapped mRNA terminating in a poly(A) 50 tail. The full-length TEV 5Ј-leader (i.e. TEV-(1-143)) promoted cap-independent translation 285-fold relative to the uncapped control luc mRNA containing a 5Ј-untranslated region of similar length and degree of secondary structure (i.e. Con 144 ) (Fig. 1), in good agreement with previous results (21,22). The truncated TEV-(1-118) and TEV-(1-97) constructs retained 90% of the translational activity of the full-length leader, whereas TEV-FIG. 1. Identification of sequences within the TEV 5-leader required for cap-independent translation. Deletions introduced by PCR are indicated by the gap in the thick black line. Each deletion mutant was tested for its ability to direct cap-independent translation of luc mRNA in vivo. luc mRNA with each TEV leader construct was synthesized in vitro as an uncapped, polyadenylated mRNA and delivered to carrot protoplasts. Expression from each mRNA construct was determined by measuring luciferase activity and reported as a percentage of the amount directed by the wild-type TEV 5Јleader (set at 100%). A 144-nt control (Con) sequence as indicated by the thin gray line with similar ⌬G served as the control.
(1-85) enhanced translation 114-fold (40% of the activity of the full-length leader). Further truncation to TEV-(1-65) reduced the translational activity to 9% of the full-length leader. The activity associated with TEV-(1-65) was reduced more than 4-fold relative to TEV-(1-85), demonstrating the importance of the region between nt 65 and 85. Deletion of the 5Ј-proximal 65 nt (i.e. TEV-(66 -143)) reduced the translational activity to 10% of the full-length leader, whereas deletion of the 5Ј-proximal 28 nt (i.e. TEV-(28 -143)) retained 90% of the translational activity of the full-length leader, demonstrating the importance of the region between nt 29 and 65. Because this analysis suggested the region between nt 29 and 85 was critical to promoting cap-independent translation, a series of deletions within this region was made. Deletion of nt 38 -43 resulted in just 20% of the activity of the full-length TEV leader, whereas deletion of nt 47-51 reduced cap-independent translation to 49% of the activity of the full-length TEV leader. Deletion of nt 54 -60 substantially reduced cap-independent translation to just 9% of the full-length TEV leader construct, a level of reduction similar to deleting the 5Ј-proximal 65 nt (i.e. TEV-(66 -143)), whereas deletion of nt 61-67 reduced cap-independent translation to 26%. Deletion of nt 67-75 reduced cap-independent translation to 40% of the full-length TEV leader. Deletion of the entire TEV 5Ј-leader was shown to reduce the translational efficiency of uncapped and polyadenylated luc mRNA in carrot protoplasts but did not affect its physical or functional stability (Gallie et al. (21)). Thus, these results suggest that sequences from nt 28 to 75 are most important in promoting cap-independent translation. These results also demonstrate that cap-independent translation by the TEV 5Јleader is not conferred by a single discrete sequence but rather by several regions throughout the leader, raising the possibility that a higher order structure may be required for translation.
Structural Analysis of the TEV RNA 5Ј-Leader-In order to further understand how multiple discrete regions in the TEV 5Ј-leader function together to promote cap-independent translation, the structure of the viral leader was determined. Prediction of the thermodynamically most stable secondary struc-ture of the TEV 5Ј-leader suggested the presence of several stem-loops (SL) between nucleotides 38 and 136 preceded by a 5Ј-proximal 37-nt unstructured region (data not shown). Prediction of ternary structure within the TEV 5Ј-leader suggested tertiary interactions between the loop of the stem-loops and sequence adjacent to each stem that would form three RNA PKs ( Fig. 2A).
PK1 is an H-type pseudoknot with two stems, S1 (6 bp) and S2 (5 bp), that are connected by loops L1 (3 nt), L2 (2 nt), and L3 (7 nt) ( Fig. 2A). S1 contains a 1-base bulge (C56) on the 3Ј-distal side of the stem. S1 and S2 are separated by the two nucleotides of L2, i.e. C52 and A53, which, to allow coaxialstacking, would need to bulge out of the helix but if not would produce a bend in the pseudoknot, thus preventing coaxial stacking (30 -32). PK1 is flanked by a 5Ј-proximal 37-nt sequence that is predicted to possess little secondary structure.
PK2 is an H-type pseudoknot composed of S1 (6 bp) and S2 (3 bp) that are connected by loops L1 (3 nt), L2 (2 nt), and L3 (4 nt) ( Fig. 2A) and is similar to PK1 in that S1 and S2 are separated by two nucleotides in L2, i.e. C90 and U91. An alternative structure to PK2 is composed of two stem-loops (Fig. 2B). PK3 is also an H-type pseudoknot composed of S1 (5 bp) and S2 (3 bp) that are connected by loops L1 (2 nt) and L2 (1 nt). In this pseudoknot, S1 and S2 are not separated and thus may coaxially stack.
To determine whether these predicted structures are supported experimentally, the structure of the TEV 5Ј-leader was examined in solution. The 5Ј-end-labeled, in vitro-synthesized, 143-nt TEV 5Ј-leader RNA was heat-denatured and renatured in the presence of 2 mM MgCl 2 to promote folding of the structure. Double-stranded regions were identified following partial digestion with RNase V1, which preferentially cleaves between nucleotides of double-stranded regions RNA or stacked singlestranded regions, generating 5Ј-phosphate fragments (33). Single-stranded regions were identified following partial digestion with RNase T1, which cleaves 3Ј of unpaired G residues as follows: RNase T2, a nonspecific 3Ј-cleaving endoribonuclease with a preference for A residues; RNase A, which cleaves 3Ј of unpaired C and U residues; or nuclease S1, which cleaves 5Ј of FIG. 2. The predicted structure of the TEV 5-leader. A, the entire TEV 5Ј-leader (nt 1-143) is displayed with the predicted structure of the three pseudoknots, i.e. PK1, PK2, and PK3. The stems (e.g. S1 or S2) and loops (e.g. L1, L2, or L3) for each pseudoknot are indicated. B, the TEV 5Ј-leader containing an alternative structure, i.e. SL1 and SL2, in place of PK2 is shown. This alternative structure is also consistent with the enzymatic and chemical probing data. The AUG at the 3Ј terminus represents the initiation codon of the TEV polyprotein-coding region.
Cleavage of the 5Ј-proximal 37 nucleotides by RNase A, RNase T2, and nuclease S1 (Figs. 3 and 4) indicated that this region is highly accessible and unstructured, an observation consistent with the predicted structure. Some residues between 7 and 13 nt were less accessible to cleavage by these enzymatic probes, suggesting possible involvement in structure.
RNase V1 cleavage occurred at G38, A39, C42, U58, and C60 (Figs. 3 and 4) supporting the formation of S1 in PK1. RNase T2 cleaved at the bulged C (i.e. C56), indicating that this C was indeed accessible. A57, which lies immediately 3Ј to the bulged C, was cleaved to a limited extent by RNase T2, indicating some local breathing. G38, which is predicted to be part of the first base pair at the base of the SL1 stem, was accessible as indicated by probing with RNase T1, but G38 and C60 were also cleaved by RNase V1. The accessibility of G38 may be explained by breathing at this end of SL1 where it is cleaved by RNase V1 when it is base-paired with C60 but cleaved by RNase T1 when it is not. G55 on left side of the S1 was largely resistant to RNase T1 cleavage (Figs. 3 and 4), consistent with its internal position in S1.
G47 was largely resistant to cleavage by RNase T1 as were C48 and U51 to RNase A cleavage (Figs. 3 and 4) and C48, A49, A50, and U51 to cleavage by RNase T2 and nuclease S1. The low accessibility G47, C48, A49, A50, and U51 is inconsistent with their presence in a single-stranded loop, which would be expected to be accessible to single-stranded specific RNases and nucleases. The inaccessibility of this region is consistent with cleavage at C44, A45, C48, A49, A50, C52, and A53 by RNase V1 and suggests that the loop is structured or involved in base pairing. A sequence within this region (i.e. 47 GCAAU 51 ) is complementary to a region downstream of S1 (i.e. 68 AUUGC 72 ), raising the possibility of base pairing between these two regions to form a pseudoknot. Cleavage at A68, U69, and U70 by RNase V1 supports the putative base pairing between these regions and formation of S2 in PK1. The sequence representing L3 ( 61 UACUUCU 67 ) also exhibited little accessibility to single-stranded specific enzymatic probes (Figs. 3 and 4), suggesting that L3 is sequestered in such a way that is not susceptible to cleavage. These data are consistent with the prediction of an H-type pseudoknot between nucleotides 38 and 72.
The cleavage data supported the presence of either a pseudoknot or two stem-loops in the region from nt 78 to 104. Strong cleavage by RNase A at C85 and RNase T2 at A86 was observed (Figs. 3D and 4), suggesting that this region is highly accessible and may form loop L1 in PK2 or the loop in SL1 (Fig.  2B). Significant cleavage by RNase V1 was observed at U80 as it was at U91-U93, supporting the presence of S1 in PK2 (Figs. 3D and 4). Cleavage by RNase T2 at A95 and A97 (Figs. 3D and 4) suggests that this region is unstructured, which supports the presence of the loop in SL2 (Fig. 2B). Strong cleavage by RNase T1 at G98 is consistent with this base being present in L3 of PK2 or in the loop of SL2. Cleavage by RNase V1 was observed at A101 and supports the presence of the stem in SL2. Therefore, these data support the formation of PK2 or SL1/SL2 which may be in equilibrium.
Cleavage by RNase V1 was observed at A106, A107, U108, U109, U110, A117, and U120 (Figs. 3D and 4), supporting the existence of S1 in PK3. Cleavage by RNase A at U111 is consistent with this base being present in L1 of PK3. Within the loop, the sequence 113 UGA 115 is complementary with 122 UCA 124 , suggesting tertiary base pairing and the formation of a pseudoknot. The lack of strong cleavage at G114 by RNase T1 supports the presence of S2 in PK3, although moderate cleavage by nuclease S1 at this same position as well as RNase A at C123 may indicate that the tertiary base pairing in PK3 may not be strong. Cleavage by RNase A at C126, U130, and U139; by RNase T2 at A127, U128, U129, U130, A131, A138, and A140; by RNase T1 at G133, G137, and G141; and by nuclease S1 at U130, A131, G133, C136, and G137 suggested that the region downstream of PK3 may be unstructured.
Stable RNA folding is sensitive to the presence of cations that promote folding by reducing the electrostatic repulsion between RNA phosphates (36). Tertiary interactions, where the proximity of RNA phosphates is increased over that in duplex RNA, are particularly stabilized by Mg 2ϩ as was first observed during tRNA folding (37)(38)(39). Stabilization by Mg 2ϩ reduces enzymatic cleavage by single-stranded nucleases as a consequence of stable base pairing and thus can serve as an indicator of the presence of tertiary interactions. To examine whether the proposed tertiary structures within the TEV 5Ј-leader are stabilized by Mg 2ϩ , the RNA was probed with RNase T1 in the presence of decreasing amounts of Mg 2ϩ . An increase in accessibility at the paired G38, G47, and G55 in the stems of PK1 was observed following removal of Mg 2ϩ and, although faint, at the paired G71 in S2 as well, whereas little difference in accessibility was observed at the unpaired G74 that lies just downstream of PK1 (Fig. 5). The accessibility of the unpaired G98 in the loop of SL1 (or L3 of PK2) actually decreased following removal of Mg 2ϩ (Fig. 5). The high level of accessibility of G98 in the presence of Mg 2ϩ may be a result of this nucleotide being highly exposed in the folded structure. An increase in accessibility at G104 in the stem of SL2 (or S2 of PK2) and G114 in S2 of PK3 was observed following removal of Mg 2ϩ (Fig. 5). Larger molecular weight bands did not resolve well in the presence of Mg 2ϩ , making it difficult to assess changes in the accessibility of G133, G137, and G141. These data indicate that Mg 2ϩ does affect the stability of higher order structures within the TEV leader and support the involvement of several guanosine residues in secondary and tertiary interactions.
Because PK1 is present in the region between nt 28 and 75 identified in Fig. 1 as functionally critical in promoting capindependent translation, the involvement of tertiary base pairing within PK1 was investigated further. The structure of TEV-(1-85), in which the entire PK1 is present, and TEV-(1-65), in which the absence of nt 66 -72 abolishes the potential for tertiary base pairing in PK1, were compared with the fulllength TEV-(1-143). Probing TEV-(1-143) with RNase T1 in 5 M urea, which disrupts base pairing within the structure thus rendering the RNA single-stranded, resulted in increased cleavage at G47, G55, and particularly at G71 (Fig. 6A, compare lane T1* with T1), supporting the notion that the guanosines at each of these positions are involved in base pairing. In contrast, no increase in cleavage was observed at G74, suggesting that it is not involved in base pairing, consistent with its position just downstream of PK1. No increase in cleavage was observed at G38 (Fig. 6A, compare lane T1* with T1), which in the wild-type structure is already subject to FIG. 4. Summary of the enzymatic and chemical probing data of the TEV 5-leader. The enzymatic and chemical probing data of the TEV 5Ј-leader from Fig. 3 are summarized. Bases that were highly or moderately accessible to the probes are indicated by large or small symbols, respectively. Unmarked bases represent those that were not accessible. Tertiary base pairing is indicated by the gray lines. extensive cleavage perhaps due to breathing at the base of S1 in PK1. At G98, a decrease in cleavage was observed when probed with RNase T1 in urea, consistent with the notion that this guanosine is present in a highly exposed position within PK2 (or SL2). An increase in cleavage was observed at G104 when probed in urea, consistent with its involvement in a base pair in S2 of PK2 or the stem of SL2. No increase in cleavage was observed at G114 (Fig. 6A, compare lane T1* with T1), consistent with the possibility that the tertiary base pairing in PK3 may be weak.
Similar increases in accessibility were observed for G47, G55, and G71 when TEV-(1-85) was probed with RNase T1 in urea with a small increase at G38 observed as well. The observed increase in cleavage at G74 when probed with RNase T1 in urea may be a result of the termination of the RNA at nt 85, which disrupts PK2 or SL1 and may allow ectopic base pairing to G74. A reduction in cleavage by RNase V1 was observed at A68, U69, and U70 in TEV-(1-85) compared with the extent of cleavage in TEV-(1-143). As cleavage at these positions in TEV-(1-143) supported the presence of tertiary interactions in PK1, the decreased cleavage in TEV-(1-85) may indicate some reduction of stability of PK1. In contrast to TEV-(1-143) and TEV-(1-85), when TEV-(1-65) was probed with RNase T1 in urea, no increase in cleavage was observed at G47 and only a slight increase in cleavage was observed at G55 (Fig. 6A, compare lane T1* with T1), consistent with the notion that, in the absence of the distal side of S2 in PK1, G47 is no longer base-paired, and the stability of the C42-G55 base pair is reduced.
Direct comparison of cleavage at G38, G47, and G55 by RNase T1 under conditions to promote formation of structure (i.e. in the absence of urea and presence of Mg 2ϩ ) revealed substantially more cleavage at G47 in TEV-(1-65) than in TEV-(1-143) or TEV-(1-85) (Fig. 6B) with little change in cleavage at G38 or G55, consistent with the involvement of G47 in base pairing in TEV-(1-143) or TEV-(1-85) but not in TEV-(1-65). A substantial increase in cleavage by nuclease S1 at G47 and a smaller increase in cleavage by nuclease S1 at G55 support the notion that G47 is no longer base-paired and a reduction in the stability of the C42-G55 base pair in TEV-(1-65). Several additional changes in accessibility of TEV-(1-65) were observed, most notably an increase in cleavage at A57, RNase T1 (T1), RNase T2 (T2), RNase A (A), RNase S1 (S1), and RNase V1 (V1). Lanes T1* represent RNA denatured in 5 M urea prior to partial digestion with RNase T1. Lane C, undigested RNA included as a control. Partially hydrolyzed RNA (lane L) served as an RNA ladder to identify base positions. Numbering is from the first nucleotide of the TEV 5Ј-leader. The regions representing PK1 as well as its stems and loops are indicated. B, the comparison of the accessibility of guanosines at positions 38, 47, and 55 within PK1 by RNase T1 (T1) in the full-length TEV 5Ј-leader (i.e. TEV-(1-143)) with that in the TEV 5Ј-proximal 85 nucleotides (i.e. TEV-(1-85)) and the TEV 5Ј-proximal 65 nucleotides (i.e. TEV-(1-65)). Tertiary base pairing is possible in TEV-(1-143) and TEV-(1-85) but not in TEV-(1-65). Just the region of the gel representing these guanosines is shown. U58, and U59 by RNase V1 and a decrease in cleavage by RNase T2 at these same positions relative to TEV-(1-143) (Fig.  6A), suggesting an increase in the stability of these base pairs of S1 in PK1 or that RNase V1 has greater access to S1 when not present in a pseudoknot.
Whether deletions within PK1 would affect the stability of the pseudoknot was then examined. Deletion of L3 resulted in decreased RNase T1 cleavage at G47 and G71, suggesting an increase in the stability of S2 (Fig. 7A). Note that deletion of L3 (i.e. nt 61-67) results in a 7-nucleotide shift in the gel at what was A68 in the wild-type so that it runs at nt 61 in the mutant. To make comparisons of specific nucleotides between mutants without confusion, the wild-type position designation is used for the deletion mutants as well. For example, as a result of the deletion in the ⌬L3 mutant (Fig. 7A), the position of G74 and G71 has shifted downward 7 nucleotides, but their designation as G74 and G71 has been maintained in reference to the wildtype sequence. An increase in cleavage at C52, C56, and C60 by RNase A; an increase in cleavage at G55 by RNase T1 and nuclease S1; and a decrease in cleavage at A57, U58, and U59 by RNase T2 in ⌬L3 relative to the wild-type suggested substantial changes in the stability or base pairing of S1 (Fig. 7A). Deletion of G71 and C72 in ⌬GC71,72 resulted in increased cleavage at U69 and C66 by RNase A; increased cleavage at U67, U69, and U70 by RNase T2 suggested a decrease in the stability of S2 of PK1, whereas a decrease in cleavage at A57, U58, U59, and C60 by RNase T2 (Fig. 7A) suggested an increase in the stability of S1. The one exception in S1 was G55 where an increase in cleavage by RNase T1 (Fig. 7A) suggested a decrease in the stability of this base pair in S1. The observed decrease in cleavage at G38 (Fig. 7A) supports the notion of an increase in the stability of the base of S1 in this mutant. These results suggest that deletions to one part of the pseudoknot affect the overall stability of the pseudoknot.
The effect on the stability of PK1 that reversing the entire S2 (construct S2-3) or reversing just the two CG base pairs within S2 (construct S2-6), which would maintain base pairing within S2, was then examined (Fig. 7B). Structural probing confirmed that base pairing in S2 was maintained as revealed by the poor accessibility of G72 (Fig. 7B). The accessibility of G48 was also reduced in construct S2-3, suggesting that not only was base pairing maintained within S2 but that the stability of S2 was increased. In contrast, an increase in accessibility of S1 was observed by the increase in RNase T2 cleavage in S1 of S2-3 and RNase T2 and nuclease S1 cleavage of S1 in S2-6 ( Fig. 7B), indicating a decrease in the stability of S1. These results support the conclusion that a pseudoknot is present between G38 and C72 and that the structure of the pseudoknot is maintained by its primary sequence and stability of the two stems.

PK1 Is Required to Confer Cap-independent Translation-
The results of the structural analysis provided a framework to understand the observation from the data in Fig. 1 that multiple discrete regions within the TEV 5Ј-leader were required to promote cap-independent translation. The sequence from nt 38 to 43 and from nt 54 to 60 that was identified as functionally important represents S1 of PK1, whereas nt 61-67 corresponds to L3. The sequence from nt 47 to 51 and from nt 67 to75 that was also identified includes S2 of PK1.
In order to examine the primary sequence and structural requirements of PK1 and its flanking regions involved in promoting cap-independent translation, mutational analysis was performed, and the effect on cap-independent translation was examined. Deletion of the 10-nt flanking region (i.e. 28 CAAAACAAAC 37 ) immediately upstream of PK1 reduced cap-independent translation to 31% of that provided by the wild-type TEV leader (Fig. 8), despite the fact that a similar FIG. 7. The effect of mutations within PK1 on pseudoknot stability. A, representative PAGE autoradiogram of enzymatic probing of the wild-type (WT) TEV leader, the deletion mutants, ⌬L3 (shaded region), and ⌬GC71,72 (shaded region). RNase T1 (T1), RNase T2 (T2), RNase A (A), RNase S1 (S1), and RNase V1 (V1). Numbering is from the 5Ј end of the TEV leader. The regions representing PK1 as well as its stems and loops are indicated. The wild-type position designation is used for all mutants as well despite the shift in migration following deletion of sequence in the mutants. Thus, the position of G74 and G71 has shifted seven nucleotides, but their designation as G74 and G71 has been maintained in reference to the wild-type sequence. For ⌬GC71,72, the deletion results in loss of G71 and a shift in migration of G74 by two nucleotides. B, representative PAGE autoradiogram of enzymatic probing of the wild-type (WT) TEV leader or S2-3 in which S2 is reversed (shaded region) or S2-6 in which the last two base pairs of S2 are reversed (shaded region). Partially hydrolyzed RNA (lane L) served as an RNA ladder to identify base positions.
but not identical sequence (i.e. 18 ACAACAUAUA 27 ) now flanked PK1 in the deletion mutant. The spacing between PK1 and the 5Ј terminus could be shortened without loss of function as deletion of the first 27 nt (i.e. TEV 5Јuntranslated region nt 28 -143) did not substantially reduce cap-independent translation (90% of the full-length TEV leader) (Fig. 1). Changing the native sequence of nt 28 -36 to poly(U) reduced cap-independent translation to just 20% of that provided by the wild-type TEV leader (Fig. 8,  CAA3 UUU). Changing C28 and C33 to adenosines (i.e. 28 AAAAAAAAAC 37 ) reduced cap-independent translation to 66% of the wild-type leader (Fig. 8, C3 A), whereas changing these same bases to guanosines (i.e. 28 GAAAAGAAAC 37 ) reduced cap-independent translation to 5.5% of the wild-type leader (Fig. 8, C3 G). Substitution of this region with its complementary sequence (i.e. 28 GUUUUGUUUC 37 ) resulted in a reduction in cap-independent translation to 3.5% of the wild-type leader (Fig. 8, CAA3 GUU) These data suggest that the primary sequence of nt 28 -37 and not just an unstructured region immediately upstream of PK1 or the spac-ing between PK1 and the 5Ј terminus was required for capindependent translation.
Deletion of nt 38 -43 reduced cap-independent translation to 20% of the full-length TEV leader (Fig. 1). However, as this represents the 5Ј-proximal side of S1 in PK1, its deletion may have altered the spatial relationship between the upstream flanking sequence and PK1. Therefore, to disrupt S1 while maintaining the upstream flanking region in its immediate proximal position, the 3Ј-half of S1 was substituted for the 5Ј-half of S1, which resulted in a reduction in cap-independent translation to 14% of the wild-type TEV leader (Fig. 9, S1-1). Substituting the 5Ј-half of S1 for the 3Ј-half of S1 resulted in a reduction in cap-independent translation to 27% of the wild-type TEV leader (Fig. 9, S1-2). Completely reversing S1 in PK1 such that the 5Ј-side of S1 became the 3Ј-side and the 3Ј-side became the 5Ј-side altered the primary sequence of S1, whereas it maintained its potential to base pair (Fig. 9, S1-3). Reversing S1 reduced capindependent translation to 7.4% of the wild-type TEV leader, suggesting that restoring base pairing within S1 of PK1 was not sufficient to restore cap-independent translation.
Deletion of nt 67-75, which included deletion of one nucleotide from L3, the 3Ј-half of S2, and three nucleotides of the downstream flanking region reduced cap-independent translation to 40% of the full-length TEV leader (Fig. 1). To examine whether base pairing within S2 of PK1 was required, the 5Јhalf of S2 was substituted for the 3Ј-half of S2 which reduced cap-independent translation to 39% of the wild-type TEV leader (Fig. 9, S2-1). Substituting the 3Ј-half of S2 for the 5Ј-half of S2 also reduced cap-independent translation (18% of the wild-type TEV leader) (Fig. 9, S2-2). Completely reversing S2 in PK1 (construct S2-3) altered the primary sequence of S2 while maintaining its base pairing as shown in Fig. 7B. Reversing S2 reduced cap-independent translation to 34% of the wildtype TEV leader (Fig. 9, S2-3), suggesting that restoring base pairing within S2 of PK1 was not sufficient to restore capindependent translation.
In a more conservative change to S2, changing G71 and C72 to the complementary nucleotides and thus disrupting base pairing with C48 and G47, respectively, reduced cap-independent translation to 14% of the wild-type TEV leader (Fig. 9, S2-4), whereas changing C48 and G47 to the complementary nucleotides reduced cap-independent translation to 17% of the wild-type TEV leader (Fig. 9, S2 -5). Completely reversing these 2 bp (i.e. C48 -G71 and G47-C72) restored base pairing as shown in Fig. 7B, and partial restoration of cap-independent translation was observed, although the level was still just 40% of the wild-type TEV leader (Fig. 9, S2-6). The lack of complete restoration of cap-independent translation may be due to partial destabilization of S1 as observed in Fig. 7B and/or due to a change in the primary sequence at these positions.
To examine further how mutations affecting tertiary base pairing within PK1 impacted cap-independent translation, additional mutations were made to S2 of PK1, and their effect on translation was examined. Deleting the last two nucleotides of the 3Ј-half of S2 (⌬GC71,72) reduced cap-independent translation to 24% of the wild-type level, whereas deletion of G71 or C72 reduced cap-independent translation to 33 or 10% of the wild-type level, respectively, and a G71C mutation reduced translation to 14% (Fig. 10A). Deletion of A49 and A50 in the 5Ј-half of S2 (⌬AA49,50) reduced translation to 12% of the wild-type level, whereas deletion of A49 alone only reduced translation to 48% (Fig. 10A). Deletion of G47 and C48 in this 5Ј-half of S2 (⌬GC47,48) reduced translation to 62% of the wild-type level, whereas deletion of G47 or C48 alone reduced translation to 51 or 37%, respectively (Fig. 10A). A C48U mutation, which should maintain base pairing with G71, reduced The effect of each mutant on the translation of in vitrosynthesized uncapped luc-A 50 mRNA was tested following delivery and translation in protoplasts. The level of expression from each mutant (using nt 28 -118) is calculated relative to the wild-type sequence (using nt 28 -118), which is set at 100%. translation to 22%. These results demonstrate that changes to the 3Ј-half of S2 are more disruptive than most changes to the 5Ј-half and that alterations to S2 significantly affect cap-independent translation.
Deletion of L3 in PK1 (nt 61-67) reduced cap-independent translation (Fig. 1), suggesting that L3 contributes to PK1 activity. Most interestingly, the sequence of loop 3 is complementary to a region in 18 S rRNA that is conserved throughout eukaryotes. Although it is not known whether L3 base pairs with 18 S rRNA, mutations were made in L3 that would either disrupt, maintain, or extend its potential base pairing with 18 S rRNA. The effect of each mutation on base pairing with 18 S rRNA is shown in Fig. 11. Mutating the last five nucleotides of the L3 sequence (i.e. 61 UACUUCU 67 ) to the complementary sequence (i.e. 61 UAGAAGA-67) reduced cap-independent translation to 14% of the wild-type TEV leader (Fig. 11,  Complementary). Changing C63 and C66 to adenosines (i.e. 61 UAAUUAU 67 ) reduced cap-independent translation to just 14% of the wild-type leader (Fig. 11, C3 A), whereas changing U64, U65, and U67 to adenosines (i.e. 61 UAAAAAA 67 ) reduced cap-independent translation to 20% of the wild-type leader (Fig. 11, U3 A). Changing U65 and U67 to cytidines (i.e. 61 UACUCCC 67 ), which should maintain base pairing with 18 S rRNA, only moderately reduced cap-independent translation (Fig. 11, UU/65,67/CC). Extending the potential for base pairing between L3 and 18 S rRNA by introducing additional se-quence that was complementary to the 18 S rRNA at either the 5Ј or 3Ј end of the L3 sequence reduced cap-independent translation to just 17 or 11.5%, respectively, of the wild-type leader (Fig. 11, 60 ACCA 61 and 67 CCCGGAAC, respectively).
Sequence Distal to PK1 May Affect Its Function in Promoting Translation-To examine whether specific sequences downstream of PK1 are important, mutations were made that altered the primary structure and disrupted base pairing of the structured elements distal to PK1 and their effect on capindependent translation examined. Mutating nt 84 -91 (i.e. 84 UCAUUUCU 91 ) to the complementary sequence (i.e. 84 AGUAAAGA 91 ) to disrupt base pairing in SL1 (or the alternative pseudoknot structure) resulted in an increase in capindependent translation to 119% of the wild-type level (Fig. 12). Changing nt 92-97 (i.e. 92 UUUAAA 97 ) to the complementary sequence (i.e. 92 AAAUUU 97 ) to disrupt base pairing in SL2 (or PK2) also increased cap-independent translation to 156% of the wild-type level, whereas mutating nt 100 -103 (i.e. 100 AAAA 103 ) of SL2 to its complementary sequence (i.e. 100 UUUU 103 ) had only a small effect if any (Fig. 12). Altering nt 104 -108 (i.e. 104 GCAAU 108 ) to its complementary sequence (i.e. 104 CGUUA 108 ), which might affect the stability of SL2 and disrupt S1 of PK3, increased cap-independent translation to 116% of the wild-type level (Fig. 12). Finally, changing nt 114 -116 (i.e. 114 GAA 116 ) to its complementary sequence (i.e. 114 CUU 116 ), which would disrupt S2 in PK3, increased capindependent translation to 212% of the wild-type level (Fig. 12).
These results indicate that changes to the primary sequence and disruption of the structural elements distal to PK1 do not reduce cap-independent translation and, in several cases, increase translation moderately. DISCUSSION Although a member of the picornavirus supergroup, the TEV 5Ј-leader differs considerably from those of picornaviruses in that it is considerably shorter in length, less structured, and FIG. 9. The effect of mutations within S1 or S2 of PK1 on TEV 5-leader-mediated cap-independent translation. The wild-type PK1 and its flanking regions are shown at top. Each mutant PK1 is shown with the base changes indicated by shading. The effect of each mutant on the translation of in vitro-synthesized uncapped luc-A 50 mRNA was tested following delivery and translation in protoplasts. The level of expression from each mutant (using nt 28 -143 for all mutants except S2-4, S2-5, and S2-6 for which nt 28 -118 was used) is calculated relative to the corresponding wild-type sequence (either nt 28 -143 or nt 28 -118), which is set at 100%. contains no upstream AUG triplets. Deletion analysis indicated that several disparate regions in the 5Ј-proximal 75 nucleotides were required for cap-independent translation and corresponded to PK1 and its upstream flanking region. Mutations in the leader outside these regions had little to no effect on capindependent translation, demonstrating that the 5Ј-proximal pseudoknot is specifically required for mediating cap-independent translation. Disruption of S1 or S2 of PK1 resulted in a substantial loss of cap-independent translation, whereas reversing either S1 or S2 to restore base pairing did not restore translation. However, the deleterious effect that disruption of the distal two GC base pairs in S2 had on PK1 activity could be partially reversed when base pairing was restored, suggesting that the primary sequence at these positions was not absolutely essential. As some alterations in pseudoknot stability could be detected even when the base pairing in the stems was maintained, the primary sequence of PK1 may be optimal for the correct formation of the pseudoknot. Of the three loops in PK1, L1 was the least critical to PK1 activity, whereas alteration of L2 reduced cap-independent translation substantially as did some changes to L3. The observation that changes to L1 and some changes to L3 had little effect on cap-independent translation demonstrated that the importance of the PK1 primary sequence was confined to specific regions within the structure. The flanking sequence immediately upstream of PK1 was also required for cap-independent translation. This sequence shares similarity with the translational enhancer sequence identified in ⍀, the 5Ј-leader of tobacco mosaic virus genomic mRNA (29,40,41). Mutations within structures in the TEV leader were not always correlated with reduced expression as we observed that disrupting the structures downstream of PK1 did not impair expression, suggesting that unfolding these structures does not reduce mRNA stability consistent with previous work showing that deletion of the entire TEV leader did not affect mRNA stability (21). The observation that mutations affecting the region distal to PK1 resulted in increased expression suggests that the sequence downstream of PK1 may serve as a negative regulator. The presence of sequences that promote and inhibit IRES activity within the same leader has been demonstrated previously, e.g. in the cold stress-induced Rbm3 mRNA (42,43). Whether the structured nature of the region downstream of PK1 serves to slow 40 S ribosomal subunit scanning will depend on whether 40 S subunits are recruited directly to the initiator AUG or scan through at least a portion of the leader. Deletion of nearly the entire region distal to PK1 (e.g. TEV-(1-85)) reduced cap-independent translation moderately, correlating with some reduction in the stability of PK1. The deletion also changed the spacing between PK1 and the initiator AUG, although replacing nt 66 -143 with an unrelated sequence of the same length increased the activity of nt 1-65 only marginally relative to nt 1-65 without the spacer sequence (22). It is also possible that there is a requirement for other cis-acting elements such as polypyrimidine tracts that bind polypyrimidine tract-binding protein and stimulate IRES activity in some viral RNAs such as rhinovirus (17). Several polypyrimidine tracts are present downstream of PK1, and at least one remained in the deletions introduced downstream of PK1 (Fig.  12), whereas all would have been absent following deletion of the entire region distal to PK1.
The role of pseudoknots in facilitating translation initiation from mRNAs from eubacteria and eukaryotes has been reported. An RNA pseudoknot in hepatitis C and classical swine fever virus is required to promote 40 S ribosomal subunit binding to the IRES and thus is required for IRES-mediated translation (44 -47). Appropriate spacing between the pseudoknot and the initiator AUG was required for translation (46,47). Up to three pseudoknots in the leader of the cricket paralysis-like virus, Plateau stall intestine virus (PPS), are required for IRES-mediated translation at a non-methionine initiation codon (48). In these cases, the pseudoknot may largely play a structural role in positioning the 40 S ribosomal subunit at the initiator AUG.
The sequence of L3 in PK1 is notable because it is complementary to nt 1117-1123 of 18 S rRNA that is conserved throughout eukaryotes. L3 is part of a region (i.e. 57-75) encompassing the distal half of S1, L3, the distal half of S2, and flanking sequence that is complementary to 16 of 19 bases of nt 1109 -1127 of eukaryotic 18 S rRNA. Mutating the two cytidines to adenosines in L3 or changing the three uridines to adenosines that would disrupt base pairing with 18 S rRNA reduced cap-independent translation up to 86%. In contrast, mutating two uridines to cytidines that would maintain potential base pairing with 18 S rRNA had the least effect on PK1 activity, consistent with the possibility of an interaction with 18 S rRNA. Increasing the length of L3 that is complementary to 18 S rRNA reduced cap-independent translation, demonstrating that the length and primary sequence of L3 are optimal for PK1 activity.
Although a direct interaction between L3 and 18 S rRNA remains to be demonstrated, base pairing between other sequence elements that function as translational enhancers and 18 S rRNA has been proposed. A 40-nt segment of the leader from the Gtx homeodomain mRNA cross-linked to 18 S rRNA, following 40 S ribosomal subunit binding to the mRNA and a 9-nt sequence within the 40-nt segment binds 18 S rRNA, was complementary to nt 1123-1131 within helix 26 of rat 18 S rRNAs and enhanced translation when the length of the complementary match was shortened to 7 nt (49 -51). A sequence that was complementary to the same region of 18 S rRNAs was selected from random oligonucleotides on the basis of its observed IRES activity (52). A second sequence that was complementary to nt 68 -76 of rat 18 S rRNA was also selected in the same screen for its IRES activity (52). A similar interaction was proposed for a 22-nt sequence within the leader of the cold stress-induced Rbm3 mRNA, which is complementary to nt 808 -819 of 18 S rRNA (42,43).  28 -143) is calculated relative to the wild-type sequence (using nt 28 -143), which is set at 100%. nt 1112-1121 of plant 18 S rRNA was identified previously by using murine 18 S rRNA as "clangor" segment number 10 (nt 1163-1172), so named for its potential involvement in stable intermolecular base pairing (53). A subsequent study showed that nt 1115-1124 of plant 18 S rRNA is exposed and accessible for intermolecular base pairing (54). Moreover, mRNAs containing a region with perfect complementarity to this region exhibited a high affinity for 40 S ribosomal subunits and substantially enhanced expression (54). The L3 of TEV PK1 has the potential to base pair to the same sequence of plant 18 S rRNA to which this sequence binds. Similar observations were made for a 10-nucleotide sequence complementary to nt 1105-1114 (54), suggesting that the region from at least 1105-1124 is available for stable intermolecular base pairing. These observations suggest that sequences within the leader of an mRNA that are able to interact within this region (i.e. nt 1105-1124) of 18 S rRNA may function to enhance translation through direct base pairing with the 18 S rRNA.
Previous work had demonstrated that the TEV 5Ј-leader specifically required eIF4G to promote cap-independent translation that was enhanced further by eIF4A, eIF4B, and the poly(A)-binding protein. That the TEV leader promotes capindependent translation is consistent with the notion that eIF4G may be recruited directly or indirectly to the leader. Whether eIF4G or other factors serve to promote 40 S ribosomal subunit binding to the 18 S rRNA complementary region within L3 of the TEV 5Ј-leader provides a testable model for future studies.