key: cord-0966309-ubm4au41
authors: Marczinke, Beate; Hagervall, Tord; Brierley, Ian
title: The Q-base of asparaginyl-tRNA is dispensable for efficient −1 ribosomal frameshifting in eukaryotes
date: 2000-01-14
journal: J Mol Biol
DOI: 10.1006/jmbi.1999.3361
sha: 5987eb1a08525b36ae6ea012ce899031c4b032cf
doc_id: 966309
cord_uid: ubm4au41

The frameshift signal of the avian coronavirus infectious bronchitis virus (IBV) contains two cis-acting signals essential for efficient frameshifting, a heptameric slippery sequence (UUUAAAC) and an RNA pseudoknot structure located downstream. The frameshift takes place at the slippery sequence with the two ribosome-bound tRNAs slipping back simultaneously by one nucleotide from the zero phase (U UUA AAC) to the −1 phase (UUU AAA). Asparaginyl-tRNA, which decodes the A-site codon AAC, has the modified base Q at the wobble position of the anticodon (5′ QUU 3′) and it has been speculated that Q may be required for frameshifting. To test this, we measured frameshifting in cos cells that had been passaged in growth medium containing calf serum or horse serum. Growth in horse serum, which contains no free queuine, eliminates Q from the cellular tRNA population upon repeated passage. Over ten cell passages, however, we found no significant difference in frameshift efficiency between the cell types, arguing against a role for Q in frameshifting. We confirmed that the cells cultured in horse serum were devoid of Q by purifying tRNAs and assessing their Q-content by tRNA transglycosylase assays and coupled HPLC-mass spectroscopy. Supplementation of the growth medium of cells grown either on horse serum or calf serum with free queuine had no effect on frameshifting either. These findings were recapitulated in an in vitro system using rabbit reticulocyte lysates that had been largely depleted of endogenous tRNAs and resupplemented with Q-free or Q-containing tRNA populations. Thus Q-base is not required for frameshifting at the IBV signal and some other explanation is required to account for the slipperiness of eukaryotic asparaginyl-tRNA.

Transfer RNAs have the highest density of modi-®ed nucleosides among cellular RNAs and more than 80 different modi®cations have been characterised (Limbach et al., 1994 (Limbach et al., , 1995 . The modi®cations are found at many sites in the tRNA, but they occur most commonly in the anticodon loop, especially at the wobble base (position 34) and the base immediately 3 H of the anticodon (position 37; Grosjean et al., 1995) . Several of these modi®cations have been shown to play a role in extending or restricting the decoding capacity of the tRNA and can in¯uence translational ef®ciency, ®delity, frame maintenance and codon choice (reviewed by Bjo È rk, 1992 Bjo È rk, , 1995 . Here, we investigate the potential involvement of one such modi®cation, the hypermodi®ed nucleoside queuosine (Q) at the wobble base of asparaginyl-tRNA (tRNA Asn ), in the process of À1 ribosomal frameshifting.

Programmed À1 ribosomal frameshifting is a translational event that allows the production of two (or more) proteins from a single mRNA and was ®rst described as the mechanism by which the Gag-Pol polyprotein of the retrovirus Rous sarcoma virus (RSV) is produced from the overlapping gag and pol genes (Jacks & Varmus, 1985) . The mRNA signals that specify frameshifting are comprised of two essential elements; a heptanucleotidè`s lippery'' sequence, where the ribosome changes reading frame, and a stimulatory region of RNA secondary structure, often in the form of an RNA pseudoknot, located a few nucleotides downstream (Jacks et al., 1988; Brierley et al., 1989; Ten Dam et al., 1990) . The heptanucleotide stretch that forms the slippery sequence contains two homopolymeric triplets and conforms in the vast majority of cases to the motif XXXYYYN. Frameshifting at this sequence is thought to occur by simultaneous slippage of two ribosome-bound tRNAs, presumably peptidyl and aminoacyl tRNAs, which are translocated from the zero (X XXY YYN) to the À1 phase (XXX YYY) (Jacks et al., 1988) .

At naturally occurring frameshift sites, of the codons that are decoded in the ribosomal A-site prior to tRNA slippage (XXXYYYN), only ®ve are represented in eukaryotes, AAC, AAU, UUA, UUC, UUU, and two in prokaryotes, AAA and AAG (Farabaugh, 1996) . These codons are decoded by tRNAs with a highly modi®ed base in the anticodon loop (see Hat®eld et al., 1992 and references therein). In tRNA Asn (AAC, AAU), the wobble base is Q, in tRNA Phe (UUC, UUU), wyebutoxine (Y) is present just 3 H of the anticodon, in tRNA Lys (AAA, AAG), the wobble base is 5-methylaminomethyl-2-thiouridine (mnm 5 s 2 U) (prokaryotes) and in tRNA Leu (UUA), 2-methyl-5-formylcytidine is present at the wobble position (Debarros et al., 1996) . Hat®eld and colleagues (Hat®eld et al., 1992) have suggested that hypomodi®ed variants of these tRNAs may exist that function as speci®c``shifty'' tRNAs, since such variants will have a considerably less bulky anticodon and be more free to move around at the decoding site. Experimentally, however, no consensus has emerged regarding a role for hypomodi®ed tRNAs in frameshifting. Indirect support for the hypothesis has come from an examination of the modi®cation status of the anticodons of the aminoacyl-tRNAs that are required for translation at and around the frameshift sites of human immunode®ciency virus type 1 (HIV-1), human T-cell leukaemia virus type I (HTLV-1) and bovine leukaemia virus (BLV) (Hat®eld et al., 1989) . It was found that in HIV-1-infected cells, most of the tRNA Phe lacked Y-base and in HTLV-1 and BLV-infected cells, most of the tRNA Asn lacked Q-base. In contrast, it has been shown that the frameshift ef®ciency of the HIV-1 signal in T-lymphoid cell lines remains unchanged in cells uninfected or chronically infected with HIV-1 (Cassan et al., 1994) and also during a time-course of HIV-1 infection of CD4-expressing 293 cells (Reil et al., 1994) . A comparison of the two sets of experiments is dif®cult, however, since in each case, only one parameter, either frameshift ef®ciency or modi®cation status of the cellular tRNA population, was assessed.

In this study, we have investigated the role of Q in frameshifting by expressing a frameshift site derived from that of the coronavirus infectious bronchitis virus (IBV; Brierley et al., 1987 Brierley et al., , 1989 in tissue culture cells unable to synthesise Q-base. The IBV signal, which is present at the overlap of the 1a and 1b ORFs of the virus genomic RNA, is well characterised (Brierley et al., 1991 (Brierley et al., , 1992 and comprises the slippery sequence UUUAAAC and a downstream RNA pseudoknot. We measured frameshifting in cos cells that had been passaged in growth medium containing calf serum or horse serum. Growth in horse serum, which contains no free queuine (q), eliminates Q from the cellular tRNA population (Langgut, 1993 (Langgut, , 1995 Langgut et al., 1993; Reisser et al., 1993) . Over ten cell passages, however, we found no signi®cant difference in frameshift ef®ciency between the cell types. We con®rmed that the cells cultured in horse serum were devoid of Q by purifying tRNAs and assessing their Q-content by tRNA transglycosylase (TGT) assays and coupled HPLC-mass spectroscopy. Supplementation of the growth medium of cells grown either on horse serum or calf serum with free queuine, the precursor of Q, had no affect on frameshifting either. These ®ndings were recapitulated in an in vitro system using rabbit reticulocyte lysates (RRL) that had been depleted of endogenous tRNAs and resupplemented with Qfree or Q-containing tRNA populations. Taken together, the results argue against a speci®c role for the Q modi®cation in frameshifting.

Our analysis of the role of Q in frameshifting required that the process be measured in tissue culture cells. Hitherto, studies of the IBV frameshift signal had been restricted to in vitro translation systems and expression in Escherichia coli (Brierley et al., 1997) . It was thus necessary to con®rm that the key requirements for frameshifting, namely the slippery sequence and RNA pseudoknot (Brierley et al., 1991 (Brierley et al., , 1992 , were retained in vivo. The starting point for this analysis was the frameshift reporter construct pAC74Z (Stahl et al., 1995) . This plasmid contains, under the control of the SV40 early promoter, b-galactosidase (b-gal) and luciferase (luc) reporter genes separated by a short linker into which candidate frameshift signals can be inserted. The minimal IBV frameshift signal (Brierley et al., 1992) was inserted into this intergenic region as a set of complementary oligonucleotides (see Materials and Methods) in such a way that the expression of luc, as a carboxy-terminal extension of the upstream b-gal, required a À1 frameshift at the IBV slippery sequence (pACFS1). In order to quantify the frameshift ef®ciency, aǹ`i n-frame'' version of pACFS1 was prepared in which the two reporter genes were in the same frame and luc expression was independent of a frameshift event (pACFS2). Indeed, every frameshift site variant tested was matched with an individual in-frame construct (see Table 1 ). Although this increased the number of constructs generated, it ensured that any amino acid changes introduced into the linker region, potentially changing the stability of the b-gal-luc fusion protein, were also present in the corresponding control construct, allowing a valid comparison to be made. The inframe controls contained two changes; the spacing distance was increased from six to seven nucleotides to align b-gal and luc in the same frame (by insertion of an appropriate nucleotide immediately downstream of the slippery sequence) and the slippery sequence was changed to UUUAAAG, a change known to greatly reduce frameshifting in RRL (Brierley et al., 1992) . Inactivation of the slippery sequence was important, since if this was not done, any ribosomes that frameshifted on the mRNA would terminate shortly after, decreasing b-gal-luc levels and introducing errors.

The minimal IBV frameshift signal (see the legend to Figure 1 ) and mutant variants were transfected into cos cells and subsequently, b-gal and luc levels measured 48 hours post-transfection ( Figure 1 and Table 1 ). In contrast to the very high levels of frameshifting seen with the minimal IBV pseudoknot in RRL (45 %; Napthine et al., 1999) , only about 9 % of ribosomes changed frame in vivo, a ®vefold reduction. However, the response to frameshift signal variants was broadly similar to that seen in earlier in vitro studies. Firstly, frameshifting was sensitive to changes in the last nucleotide of the slippery sequence. The hierarchy of frameshifting for UUUAAAN variants closely paralleled that seen in RRL (Brierley et al., 1992) with C the most ef®cient (9 %/45 % [cos/RRL]) then A (8 %/ 25 %), U (5.8 %/22 %) and ®nally G (0.5 %/1 %). Secondly, frameshifting was RNA pseudoknot speci®c. Complete removal of the pseudoknotdramatically reduced frameshifting (pACFS41; 0.05 %) and high levels of frameshifting were not supported by a large stem-loop structure of the same predicted size and nucleotide composition as the stacked stems of the pseudoknot (pACFS5, 1.6 %). The pattern of frameshifting observed for mutations within the pseudoknot closely paralleled that seen in the RRL, in that destabilisation of either stem of the pseudoknot reduced frameshifting ef®ciency dramatically (pACFS9, 13 in stem 1; pACFS 21 and 25 in stem 2) and compensatory mutations predicted to restore the structure increased frameshifting (pACFS17 in stem 1, 4 %; pACFS29 in stem 2, 6.5 %). A failure to restore fully the ef®ciency of frameshifting with compensatory changes in stem 1 has been documented in a number of studies of eukaryotic frameshifting (e.g. see Ten Dam et al., 1995) and may re¯ect a functional requirement for a particular pseudoknot The minimal IBV frameshift site in pACFS1 contains a slippery sequence (UUUAAAC) that is separated by a 6 nt spacer (UGAUAC) from the IBV minimal pseudoknot (PK) structure. Derivatives of the frameshift site (pACFS2-42) carried changes in the seventh nucleotide of the slippery sequence (bold), single nucleotide insertions into the spacer region (bold), or the pseudoknot was disrupted, deleted (no structure) or replaced by a stem-loop (SL) structure of the same predicted length and base-pair composition as the pseudoknot.

conformation that is imprecisely reproduced in some of the compensatory mutants. Indeed, in pACFS17, the substitution of two G residues at the 5 H -arm of stem 1 by C is predicted to have a detrimental effect on frameshifting (despite being paired), since in RRL, the IBV frameshift signal shows a preference for a G-rich stretch at the start of the 5 H -arm (Napthine et al., 1999) . The stem 2 compensatory mutant (pACFS29) reached 70 % of the wild-type level, consistent with earlier studies in RRL (Brierley et al., 1991) . These experiments con®rm for the ®rst time a requirement for the IBV pseudoknot for ef®cient À1 frameshifting in vivo.

Most animal sera contain q in concentrations varying between 50 nM and 0.3 mM, with the exception of horse serum (HS), which does not contain detectable amounts of the q-base (Kersten & Kersten, 1990) . It has been shown that growth of HeLa cells in HS can induce Q-de®ciency in the tRNA population (Langgut, 1993 (Langgut, , 1995 Langgut et al., 1993; Reisser et al., 1993) . We reasoned that a similar approach would allow us to examine the effect of Q-depletion of mammalian tRNAs on ribosomal frameshifting. cos cells were grown and passaged in fetal calf serum (FCS) or HS-containing medium and the frameshift ef®ciency of the IBV signal measured by transfection of the pACFS1 and two reporter plasmids over a series of ten passages. Each time the cells were split, 6 cm culture dishes were seeded with 3 Â 10 5 cells from each population and on the following day transfected in triplicate with pACFS1 or pACFS2 using the DEAE-dextran method (see Materials and Methods). During the transfection procedure, cells were maintained in the appropriate serum supplement. At 48 hours post-transfection, the cells were lysed for analysis of b-galactosidase and luciferase expression and the assays performed in duplicate for each dish. Mean values of the frameshift ef®ciency calculated at each cell passage are shown in Figure 2 (a). The frameshift ef®ciencies measured were found to be very similar in both cell populations, varying between 8 and 11 %. Continued passage in either q-containing or q-de®cient serum led to a slight reduction in frameshift ef®ciency, but no signi®cant difference was apparent between the two cell populations. This indicates that the modi®cation status of the wobble base of tRNA Asn , Qcontaining or hypomodi®ed, does not notably (1) Slippery sequence variants were made in which the last nucleotide was changed to A, G or U (bold).

(2) Complementary and compensatory changes were created within the pseudoknot region. In this representation of the pseudoknot, the stems are arranged vertically. For each stem, the two complementary changes (no base-pairing) and the compensatory change (base-pairing restored) are boxed and labelled with a mutant number. (3) A deletion mutant was tested (pACFS41) in which the entire pseudoknot was removed. (4) Finally, a construct was tested that formed a stem-loop structure rather than a pseudoknot. In plasmid pACFS5, the stem nucleotides are of the same length and nucleotide composition as the stacked stems of the pseudoknot in pACFS1. The frameshift ef®ciency (in cos cells) speci®ed by each construct is also shown. A brief description of each construct and its relevant in-frame partner (see the text) is shown in Table 1. in¯uence frameshifting at the IBV signal. However, it was conceivable that the FCS-fed cos cell tRNAs had low levels of Q to begin with, and that switching to HS would not lead to any sig-ni®cant further reduction in Q-content. We tested this by adding back free queuine to the growth medium. Previous studies have shown that this can fully revert the Q-de®ciency of HeLa cell Figure 2 . In¯uence of Q-base on À1 ribosomal frameshifting at the IBV signal in vivo. (a) cos cells were cultured over ten passages in medium containing either 10 % FCS or 10 % HS. The frameshift ef®cency of the minimal IBV frameshift signal was examined in the two cell populations using the frameshift reporter plasmids pACFS1 and 2 (see the text). Frameshifting was measured at each cell passage in three transient transfection experiments and the average values and range are shown. (b) cos cells were cultured over ten passages in medium containing either 10 % FCS or 10 % HS. After this time, a further ten passages were performed in the same medium, or in medium containing exogenous q-base (at 300 nM). Frameshifting was measured at each cell passage in three transient transfection experiments and the average values and range are shown. Two of the samples from passage 4 (*) were lost. (c) The queuosine-content of tRNAs puri®ed from passage 10 cos cells cultured in FCS or HS with (q) and without (Àq) added q was measured by the guanine exchange assay (see Materials and Methods). The assay was performed in triplicate using 30mg aliquots of tRNA and the average values and range are shown. Controls included commercial yeast and tyrosinyl tRNAs, and tRNAs extracted from baby hamster kidney (BHK) cells. tRNAs within 72 hours (Langgut & Reisser, 1995) . To this end, cells that had been passaged ten times in medium containing either HS or FCS were supplemented with free q-base at 300 nM and passaged a further ten times. To control for the effects of the q-supplementation, cos cell passage was also continued in unsupplemented FCS or HS-containing medium. Each time the cells were split, transfections with the frameshift reporter plasmids pACFS1 and pACFS2 were carried out as above and frameshift ef®ciency measured. As can be seen in Figure 2 (b) the frameshifting ef®ciencies in the four cos cell populations (FCS, FCS q, HS and HS q) were found to be very similar to each other and to those measured in the previous analysis (Figure 2 (a)), between 8 % and 12 %, with no noticeable effect of q-supplementation. Nevertheless, it was important to determine the Q-content of the various tRNA populations. To this end, cellular tRNAs were puri®ed and assayed for Q-content enzymically, using the E. coli tRNA guanine transglycosylase (TGT) assay and biophysically, by HPLC-combined mass spectrometry. The TGT assay exploits the inability of bacterial TGT to recognise q or Q as a substrate; the enzyme can use only a q precursor or guanine in an exchange reaction in the anticodon. When offered hypomodi®ed tRNAs of the Q-family (tRNA Asn , tRNA Asp , tRNA His , tRNA Tyr ) and a radioactive guanine isotope, the bacterial TGT introduces the radioactive isotope into the anticodon loop in a base exchange reaction (Okada et al., 1978) . Thus the incorporation of the radioactive isotope is a direct measure of the Q-content of tRNA.

The guanine exchange assay was set up with 30 mg of tRNA puri®ed from passage 10 cells from both the initial comparison of frameshifting in FCS or HS-fed cells (FCS, HS; Figure 2 (a) and the q-supplementation experiment (FCS q, HS q; Figure 2 (b)). As controls, the assay was performed with Q-de®cient yeast tRNAs, Q-containing E. coli tRNA Tyr and baby hamster kidney (BHK) cell tRNAs. All reactions were carried out in duplicate and the results are summarised in Figure 2 (c). The results obtained were entirely consistent with the predicted Q-content of the various tRNA species. Those tRNAs that were expected to be hypomodi®ed (yeast, HS) were good substrates for TGT. In contrast, those tRNAs expected to contain Q (tRNA Tyr , FCS, FCS q, HS q, BHK) showed very low levels of [ 14 C]guanine incorporation.

From this experiment, we concluded that growth in HS did eliminate Q from the cos cell tRNAs and that supplementation with free q restored the level of Q. Furthermore, the addition of q to cells grown in FCS did not alter the guanine exchange activity of the tRNAs, suggesting that the tRNAs already contained high levels of Q. Consistent with this is the observation that the level of guanine exchange seen with the tRNAs was similar to that seen with the hypermodi®ed tRNA Tyr . To con®rm these data biophysically, the puri®ed tRNAs were subjected to combined HPLC-mass spectrometry. As judged by HPLC analysis, the HS-derived tRNA was free of Q, and the FCSderived tRNA was Q-containing (Figure 3(a) and (b)). As expected, addition of q to HS medium restored the Q-modi®cation in tRNA (Figure 3(c) ), whereas addition of q to FCS medium did not signi®cantly increase the Q-content of the tRNA (Figure 3(d) ), indicating that tRNA from cells grown in FCS medium is fully modi®ed. The separation of manQ and galQ (glycosylated derivatives (mannose and galactose) of Q present in mammalian cells) from other compounds was not optimal using the gradient adapted for mass spectrometry. To detect the presence of manQ and galQ in tRNA, mass spectrometric analysis in the form of selected-ion recordings was carried out. The ions used for detection of Q was m/z 410 and for manQ/galQ m/z 572. The analyses revealed that tRNA from cells grown in FCS medium with or without added q and in HS medium with added q, contained Q, manQ and galQ (Figure 3 (e) and data not shown). No ion corresponding to Q or manQ/galQ could be detected in tRNA from cells grown in HS medium (data not shown). In conclusion, tRNA from cells grown in HS medium lack Q-derived nucleosides, while tRNA from cells grown in FCS medium is fully modi®ed with respect to these nucleosides.

We went on to test the in¯uence of Q on in vitro frameshifting, using a tRNA-dependent in vitro translation system (unpublished results). RRL was depleted of tRNAs by gel ®ltration through an ethanolamine DEAE-Sepharose column and fractions judged to be greater than 90 % depleted of tRNA, yet retaining all other nucleic acid species, were used in translation assays. The nucleic acid content of such a depleted fraction is shown in Figure 4 (Brierley et al., 1992) , which contains the same minimal IBV frameshift signal as used in pACFS1 cloned into the in¯uenza A/PR/8/34 PB2 gene (Brierley et al., 1992) at a unique BglII site. Transcription of BamHI-digested pFScass 6 in vitro using SP6 RNA polymerase generates a 2.4 kb mRNA that when translated in RRL yields a 19 kDa non-frameshifted species and a 28 kDa À1 frameshift product. Transfer RNAdepleted RRL was programmed with capped pFScass 6/BamHI mRNA and supplemented with either Q-containing (FCS-grown cos cells, BHK cells, q-supplemented cos-FCS and q-supplemented cos-HS cells) or Q-free tRNAs (HSgrown cos cells, yeast tRNAs). The translation products, along with the calculated frameshift ef®ciencies, are shown in Figure 4 (b). As controls, translations were also carried out with undepleted RRL, undepleted RRL that had been diluted 30 % to account for the dilution that occurs during chromatographic depletion of tRNAs, and depleted RRL that was supplemented with calf-liver tRNA (the tRNA used to supplement standard RRL translations (Jackson & Hunt 1983) ). The translation of the pFScass 6 mRNA in the undepleted lysates (Figure 4, lanes 2 and 4) yielded the expected 19 kDa non-frameshifted and 28 kDa À1 frameshifted products and the frameshift ef®ciency was about 40 %, the expected value for this particular signal (Brierley et al., 1992; Napthine et al., 1999) . Indeed, both products were observed in the translation reactions that contained mRNA, except for the translation reaction without tRNA supplement (lane 5). Thus all puri®ed tRNAs were able to restore translation and frameshift activity. The relative abundance of non-frameshifted and frameshifted products on the gels was estimated by scanning densitometry and adjusted to take into account the differential methionine content of the 19 kDa and 28 kDa products (11 and 12, respectively). Although the overall frameshift ef®ciency seen in translations with depleted RRL was slightly reduced (to 30-35 %), the type of tRNA employed as a supplement, whether Q-modi®ed or Q-de®cient, did not appear to in¯uence the degree of frameshifting. Hence the in vitro translations paralleled the in vivo studies and reinforce the conclusion that the Q-modi®cation of tRNA Asn is not required for À1 ribosomal frameshifting at the IBV signal.

The potential in¯uence of tRNA anticodon nucleoside modi®cations on the process of À1 ribosomal frameshifting was ®rst considered by Hat®eld and colleagues (Hat®eld et al., 1989) . These authors reasoned that as most slippery sequence-decoding tRNAs have a hypermodi®ed nucleoside in the anticodon, hypomodi®cation would result in a tRNA with a less bulky anticodon and be more free to move around during decoding thus increasing frameshifting. An alternative hypothesis proposed a different role for the hypermodi®ed base (Tsuchihashi, 1991; Tsuchihashi & Brown, 1992) . In this scenario, the modi®cation in¯uences frameshifting by modulating the strength of the codon-anticodon interaction. From studies of ribosomal frameshifting in E. coli, the latter hypothesis seems the likelier. Firstly, it has been shown that hypomodi®cation per se, does not necessarily lead to a stimulation of frameshifting. In E. coli, the IBV signal, or a variant with slippery sequence UUUAAAU, functions inef®ciently (1-2 %), whether tRNA Asn possesses the anticodon QUU or GUU (Brierley et al., 1997) . Secondly, there is evidence to support a role for the strength of the codon-anticodon interaction in frameshifting. Tsuchihashi & Brown (1992) were able to diminish frameshifting at the dnaX signal (slippery sequence AAAAAAG) by co-expressing a variant tRNA Lys in which the anticodon had been changed from UUmnm 5 s 2 U to UUC. Expression of this tRNA, an isoacceptor not present naturally in E. coli cells, was considered to have reduced frameshift-ing by recognising more strongly the AAG codon. It was subsequently shown that the natural anticodon of E. coli tRNA Lys has an unusual conformation; it can probably form only a weak wobble pair with the anticodon, perhaps accounting for the high frameshift ef®ciency in E. coli at slippery sequences ending in AAG (Watanabe et al., 1993; Agris, 1996; Agris et al., 1997) .

The relevance of these experiments to the eukaryotic frameshifting, however, is uncertain. All ef®cient À1 frameshift sites in E. coli employ tRNA Lys as the A-site decoding tRNA and, given Figure 4 . In¯uence of Q-base on À1 ribosomal frameshifting at the IBV signal in vitro. (a) Transfer-RNA-depleted RRL (depleted; 30 ml), which had been left unsupplemented (À) or had been supplemented with HS tRNA to 50 mg/ml () was analysed for nucleic acid content following extraction with phenol/chloroform and precipitation in ethanol. Aliquots (5 ml) of the ®nal nucleic acid preparation (which was dissolved in 10 ml of water) were analysed on a 3 % agarose gel and visualised by staining with ethidium bromide. Also analysed were 5 ml of the nucleic acid preparation from undepleted, RRL (0), 0.75 mg of puri®ed HS tRNA (HS) and 0.75 mg of commercial calf liver tRNA (liver). Molecular mass DNA markers (M; Life Technologies) were included. (b) In vitro translation of pFScass 6/BamHI mRNA (Brierley et al., 1992) in normal RRL (complete), normal RRL diluted to 70 % with water (diluted) or RRL depleted of endogenous tRNAs (depleted). The depleted RRL was either unsupplemented (À) or supplemented with tRNAs (at 50 mg/ ml) derived from either BHK cells (BHK), FCS-grown cos cells (FCS), FCS-grown cos cells supplemented with endogenous q (FCS q), HSgrown cos cells (HS), HS-grown cos cells supplemented with endogenous q (HS q), yeast cells (yeast) or calf liver cells (liver). Tracks labelled H 2 O represent water-programmed translations. The pFScass 6/BamHI template directs the synthesis of a 19 kDa non-frameshift (non-FS) and 28 kDa frameshift product (FS) (Brierley et al., 1992) . Products were labelled with [ 35 S]methionine, separated on an SDS/15 % polyacrylamide gel and detected by autoradiography. The frameshift ef®ciencies quoted above the relevant lanes are the average of two independent measurements and the range is shown in parentheses. that this tRNA has a very unusual anticodon structure, it may be the case that frameshifting in prokaryotic systems is distinct from that in eukaryotes and that tRNA Lys is a special case. In this study, we have performed a detailed examination of the in¯uence of the hypermodi®ed base Q on ribosomal frameshifting in eukaryotes, using both in vivo and in vitro systems. During all experiments, the predicted modi®cation status of puri®ed cellular tRNAs with respect to queuosine was con®rmed with an enzymatic guanine exchange assay and combined HPLC-mass spectrometry. Frameshifting was measured in cos cells that had been passaged in growth medium containing either FCS or HS, and in cos cells grown in FCS or HS supplemented with free queuine base. Addition of free queuine converts hypomodi®ed tRNA (in HS) back to the hypermodi®ed state (in FCS). The puri®ed cellular tRNAs of de®ned modi®cation status were also used in a tRNA-dependent in vitro translation system to study the effect of tRNA Asn hypomodi®cation on in vitro frameshifting at the IBV signal. The ef®ciency of the IBV frameshift signal was not in¯uenced by the nature of the tRNA population present in either system. The simplest interpretation of these observations is that both hypo-and hypermodi®ed tRNA Asn can support ef®cient ribosomal frameshifting. However, as we cannot rule out the possibility that the tRNAs puri®ed from FCS-grown cos cells (or FCS-grown cos cells supplemented with q) still retain low levels of hypo-modi®ed tRNAs, it is not possible to conclude categorically that frameshifting can be mediated by hypermodi®ed tRNAs alone. However, it is clear that hypomodi®ed tRNAs are fully functional.

How do these observations ®t with the view that anticodon loop modi®cations may in¯uence frameshifting by in¯uencing the strength of the codonanticodon interaction during decoding of the slippery sequence? The three-dimensional structure of queuosine 5 H -monophosphate, determined by X-ray crystallography, and incorporated into tRNA Tyr on the basis of the coordinates of yeast tRNA Phe , suggests that Q in the anticodon loop does not interfere with codon-anticodon interactions (Yokoyama et al., 1979) . The bulky 7-substituent cyclopentenediol group of Q is fully extended outwards away from the anticodon, and the bond length and bond angles of the pyrimidine moiety of the 7-deazaguanine ring are almost equal to those of the unmodi®ed guanosine base. Thus the conformational characteristics imply very similar decoding properties of Q and G-containing tRNA Asn , consistent with our observation that frameshifting ef®ciencies are similar with Q or Gcontaining tRNA Asn . However, measurements of the relative stability of anticodon-anticodon complexes suggest a slight reduction in stability when a G-C pair is replaced by Q-C pair at the wobble position in the same anticodon sequence; the G-Ccontaining complex was stable for 840 ms compared with 620 ms for Q-C (Grosjean et al., 1978) . So we might have expected a decrease in frame-shift ef®ciency in Q-free cells if the stability hypothesis is correct.

Our observation that supplementation of the tRNA-dependent RRL with Q-free tRNAs (HSderived tRNAs or yeast tRNAs) did not stimulate frameshifting at the IBV site is inconsistent with recent work by Carlson et al. (1999) , who studied frameshifting using in vitro translation reactions supplemented with hypomodi®ed variants of tRNA Asn or tRNA Phe . In these experiments, slippery sequence variants of the mouse mammary tumour virus gag/pol frameshift signal were translated in RRL in the presence or absence of the exogenous hypomodi®ed tRNAs. It was found that frameshifting at AAAUUUU was stimulated by addition of Y À tRNA Phe (puri-®ed from rabbit reticulocytes) and at AAAAAAC or AAAAAAU by addition of Q À tRNA Asn (puri-®ed from yeast strain 3950-1B H 2). Currently, it is not known why these data contradict those of the present study. In the analysis by Carlson et al. (1999) , standard, non-depleted RRL was employed. Although it is possible that the tRNA-depleted lysate employed in our study is also depleted of an additional component(s) that is required to mediate the stimulation afforded by hypomodi®ed, but not hypermodi®ed tRNAs, this seems unlikely. Although the frameshift ef®ciencies measured in the tRNA-depleted RRL are slightly reduced from those measured in standard RRL (38 to about 32 %), this is probably a consequence of a reduced processivity of translation in the re-supplemented reactions. This leads to an underestimation of the amount of frameshift product and, consequently, an apparent reduction in frameshifting. A more likely explanation is that the in¯uence of hypomodi®ed tRNAs is seen only at certain slippery sequences; i.e. at certain P and A-site tRNA combinations. Alternatively, the IBV site may represent aǹ`u pper limit'' of frameshift ef®ciency that cannot be stimulated further by hypomodi®ed tRNAs. We are currently testing these possibilities. Whatever the case, Q is not required for frameshifting at the IBV site in vivo.

The study of IBV frameshifting in eukaryotic cells has con®rmed for the ®rst time the requirement for a pseudoknot in vivo. Indeed, the pseudoknot was about six times more effective at promoting frameshifting than a related hairpinloop. However, the level of frameshifting observed was reduced some four to ®vefold in comparison to in vitro translation systems (RRL; wheat-germ extract, data not shown) and the reason for this is not known. In RRL, the translation process is physiologically detached from other post-transcriptional events such as splicing, polyadenylation and nuclear export of mRNA, all of which might in¯uence the kind of proteins associated with the mRNA, and hence potentially frameshifting. Alternatively, differences with respect to the ribosome load on mRNAs between the two systems cannot be ruled out. The molecular basis for this difference may be informative in terms of the mechanism of the frameshift process.

The pACFS series of plasmids were prepared in a two-step cloning strategy. Plasmid pAC74Z (Stahl et al., 1995) was digested sequentially with BclI and NheI (which removes the b-galactosidase gene) and wild-type or mutant IBV frameshift sequences inserted as three pairs of complementary, phosphorylated oligonucleotides. The design of the oligonucleotides was such that the zero reading frame of the inserted frameshift site was in the same frame as the deleted b-galactosidase gene, and the À1 reading frame was that of luciferase. The ends of the oligonucleotides restored the NheI site at the 5 H -site of the frameshift window but the BclI site and the termination codon at the 3 H -site were not reformed. Ligation reactions were transformed into E. coli DH5a and correct clones (pACIBV* series) identi®ed by DNA sequencing (Sanger et al., 1977) . Plasmids of the pACIBV series were digested with NheI and ligated with the b-galactosidase gene fragment isolated from pAC74Z by NheI restriction digestion. The resulting plasmids (pACFS series, see Table 1 ) were transformed into E. coli JM101 cells and plasmid DNA isolated by CsCl-EtBr density-gradient centrifugation.

Tissue culture transfection protocol cos cells were maintained in Dulbecco's modi®cation of Eagle's medium (DMEM) containing 10 % (v/v) fetal calf serum (FCS) or 10 % (v/v) horse serum (HS) (all from Sigma). Cells were passaged when they reached 70 % con¯uence. Where required, queuine (a generous gift from Dr Susumu Nishimura, Banyu Tsukuba Research Institute, Tsukuba, Japan) was added to 300 nM. Plasmid transfections were carried out using the DEAE-dextran method. Cells (3 Â 10 5 ) were seeded in 60 mm dishes and grown for 18-24 hours. The cells were washed consecutively with phosphate-buffered saline (PBS) and Optimem (Gibco), before overlaying with 0.5 ml of a transfection mix containing 3 mg of plasmid DNA in 0.25 ml of Optimem and 0.25 ml of 1 mg/ml DEAE-dextran in TBS (25 mM Tris-HCl (pH 7.4), 140 mM NaCl, 3 mM KCl). After 30 minutes, the mix was removed by aspiration and replaced with 5 ml of the relevant growth medium containing 60 mg/ml chloroquine. Incubation was continued for a further one to four hours before the medium was replaced with fresh growth medium without chloroquine. The cells were harvested 48 hours post-transfection and reporter gene expression determined as below.

Transfected cells were washed with PBS, drained and 450 ml of lysis buffer (25 mM glycyl-glycine (pH 7.8), 15 mM MgSO 4 , 4 mM EGTA, 0.1 % (v/v) Triton-X-100, 1 mM DTT) added. Following ten minutes agitation at room temperature, cell debris was removed by centrifugation (10,000 g, one minute), the supernatants transferred to fresh tubes on ice and 100 ml aliquots tested immediately for b-galactosidase and luciferase activity according to Sambrook et al. (1989) and de Wet et al. (1987) , respectively. Enzyme assays were performed in duplicate. Each plasmid was tested in at least three independent transfections.

Preparation of tRNA-guanine transglycosylase (TGT) E. coli TGT was prepared from the prokaryotic expression plasmid pTGT5 (provided by Dr George Garcia, University of Michigan, Ann Arbor, USA) essentially as described (Garcia et al., 1993) . Brie¯y, TGT was overexpressed in E. coli BL21 cells and puri®ed by selective ammonium sulphate precipitation followed by anionexchange chromatography. The product was judged to have a purity greater than 95 % and was free of contaminating nucleic acids. The speci®c activity of the preparation was determined by the guanine exchange assay (see below) and was 817 units/mg. This value was derived according to the de®nition (Garcia et al., 1993) , that one unit of TGT catalyses the incorporation of 1 mmol of [ 14 C]guanine into tRNA per minute at 100 mM yeast tRNA at 37 C per mg of TGT Â 10 5 . The speci®c activity of the TGT was almost identical with that prepared by others (804 units/mg ; Garcia et al., 1993) . TGT was stored at À70 C at 2 mg/ml in 100 mM Hepes (pH 7.7), 5 mM DTT, 0.3 M KCl.

The presence or absence of Q-base in tRNAs was assessed by the guanine exchange assay, which exploits the ability of the E. coli TGT enzyme to catalyse the exchange of a guanine base at the ®rst position of the anticodon of queuosine-unmodi®ed tRNAs with radiolabelled [ 14 C]guanine (46 mCi/mmol; Amersham, UK). The enzyme cannot use queuosine-modi®ed tRNAs as a substrate, thus the guanine acceptance is maximal when the tRNAs are Q-de®cient. Radiolabelled guanine incorporation is therefore a measure of queuosine content (Okada et al., 1978) . The assay was performed for one hour at 37 C in a total volume of 50 ml and contained 0.1 M Hepes KOH (pH 7.5), 20 mM MgCl 2 , 1.6 units of TGT, 0.2 mCi of [ 14 C]guanine (46 mCi/mmol) and 0-1 mg of tRNA. Aliquots were spotted onto GF/C ®lter disks, washed once with 3 ml of 10 % trichloroacetic acid (TCA), twice with 3 ml of 5 % TCA, once with 3 ml of 70 % ethanol, dried and counted in a PackardTri-Carb 1 1500 scintillation counter. Assays were performed in triplicate. Commercial yeast and tyrosinyl-tRNAs used in the assay were from Sigma.

Pellets of 2.5 Â 10 8 cells, previously washed in PBS, were resuspended in 4 ml of NTE lysis buffer (0.1 M NaCl, 10 mM Tris-HCl (pH 7.5), 1 mM EDTA, 0.5 % (v/v) NP40) and incubated on ice for two minutes. The lysates were cleared by centrifugation (6000 g, ®ve minutes), the supernatants made 1 % (w/v) in SDS and subjected to two rounds of extraction with phenol/ chloroform and one round with chloroform. Following precipitation in ethanol, the crude nucleic acid pellet was resuspended in 0.5 ml of water. Aliquots (100 ml) were mixed with an equal volume of formamide, boiled for two minutes and loaded onto a denaturing 15 % polyacrylamide gel. The tRNA-containing bands were ident-i®ed by UV shadowing, sliced out, chopped up and eluted in 0.5 M ammonium acetate, 10 mM magnesium acetate, 1 mM EDTA, 0.1 % (w/v) SDS, at 4 C overnight. The eluted tRNA was concentrated by precipitation in ethanol. Typically 0.6 to 0.8 mg of tRNA was obtained from 10 9 cells and was dissolved in water at a concentration of 2 mg/ml.

Transfer RNA was degraded to nucleosides using nuclease P 1 and alkaline phosphatase (Gehrke & Kuo, 1989) . The nucleoside mixture was analysed on a Waters TM System liquid chromatograph with a Waters TM 996 diode array UV detector (Waters TM Corporation, Milford, MA, USA), directly interfaced to a VG platform mass spectrometer equipped with an electrospray ionisation source (Fisons Instruments, Altrincham, UK). Separation of nucleosides by HPLC was achieved using a Supelcosil LC-18-S reverse-phase column (2.1 mm by 250 mm, particle diameter 5 mm) and a Supelguard LC-18-S, 2.1 mm by 20 mm guard column (Supelco, Bellefonte, PA, USA) held at 26 C, at a¯ow-rate of 2 ml/ minute. Nucleosides were eluted using the gradient described by Buck et al. (1983) but altered to accommodate a lower ammonium acetate concentration (5 mM), which is more compatible with electrospray ionisation. UV data were recorded continuously, and mass spectra were taken every 1.0 second during the 60 minutes of chromatography. The procedures and interpretation of data for qualitative LC-MS analysis of nucleosides in RNA hydrolysates have been described (Pomerantz & McCloskey, 1990) .

In vitro transcription of plasmid pFScass 6 (Brierley et al., 1992) employing the bacteriophage SP6 RNA polymerase, was carried out essentially as described by Melton et al. (1984) and included the synthetic cap structure 7meGpppG (New England Biolabs) to generate capped mRNA. Product RNA was recovered by a single extraction with phenol/chloroform/isoamyl alcohol (49:49:2, by vol.) followed by precipitation in ethanol in the presence of 2 M ammonium acetate. The RNA pellet was dissolved in water, and remaining unincorporated nucleotide triphosphates removed by Sephadex G-50 chromatography. RNA was recovered by precipitation in ethanol, dissolved in water and checked for integrity by electrophoresis on 1.5 % agarose gels containing 0.1 % SDS. In ribosomal frameshift assays, serial dilutions of puri®ed mRNAs were translated in RRL as described (Brierley et al., 1989) . Translation products were analysed on SDS/15 % (w/ v) polyacrylamide gels according to standard procedures (Hames, 1991) . The relative abundance of non-frameshifted or frameshifted products on the gels was determined by direct measurement of [ 35 S]-methionine incorporation using a Packard Instant Imager 2024. Frameshift ef®ciencies were calculated from those dilutions of RNA where translation was highly processive (RNA concentrations of 10 mg to 25 mg RNA/ml of reticulocyte lysate). The frameshift ef®ciencies quoted are the average of at least three independent measurements, which varied by less than 10 %, i.e. a measurement of 40 % frameshift ef®ciency was between 36 % and 44 %. The calculations of frameshift ef®ciency take into account the differential methionine content of the various products (19 kDa, 11; 28 kDa, 35). Transfer RNA-depleted RRL was prepared by ethanolamine-Sepharose column chromatography (unpublished results).

The importance of being modi®ed: roles of modi®ed nucleosides and Mg 2 in RNA structure and function

Unconventional structure of tRNA(Lys)SUU anticodon explains tRNA's role in bacterial and mammalian ribosomal frameshifting and primer selection by HIV-1

The role of modi®ed nucleosides in tRNA interactions

Genetic dissection of synthesis and function of modi®ed nucleosides in bacterial transfer RNA

An ef®cient ribosomal frame-shifting signal in the polymerase-encoding region of the coronavirus IBV

Characterisation of an ef®cient coronavirus ribosomal frameshifting signal: requirement for an RNA pseudoknot

Mutational analysis of the RNA pseudoknot component of a coronavirus ribosomal frameshifting signal

Mutational analysis of the``slippery sequence'' component of a coronavirus ribosomal frameshifting signal

Expression of a coronavirus ribosomal frameshift signal in Escherichia coli: in¯uence of tRNA anticodon modi®cation on frameshifting

Complete analysis of tRNA-modi®ed nucleosides by high performance liquid chromatography: the 29 modi®ed nucleosides of Salmonella typhimurium and Escherichia coli tRNA

Transfer RNA modi®cation status in¯uences retroviral ribosomal frameshifting

Translational frameshifting at the gag-pol junction of human immunode®ciency virus type 1 is not increased in infected T-lymphoid cells

(5)CM), a new modi-®ed nucleotide at the wobble position of 2 cytoplasmic tRNAs (Leu) (NAA) from bovine liver

Fire¯y luciferase gene: structure and expression in mammalian cells

Programmed translational frameshifting

tRNAguanine transglycosylase from Escherichia coli: overexpression, puri®cation and quaternary structure

Ribonucleoside analysis by reversed-phase high-performance liquid chromatography

On the physical basis for ambiguity in genetic coding interactions

Postranscriptionally modi®ed nucleosides in transfer RNA: their locations and frequencies

An introduction to polyacrylamide gel electrophoresis

Chromatographic analysis of the aminoacyl-tRNAs which are required for translation of codons at and around the ribosomal frameshift sites of HIV, HTLV-1 and BLV

Translational suppression in retroviral gene expression

Expression of the Rous sarcoma virus pol gene by ribosomal frameshifting

Signals for ribosomal frameshifting in the Rous sarcoma virus gag-pol region

Preparation and use of nuclease-treated rabbit reticulocyte lysates for the translation of eukaryotic messenger RNA

Biosynthesis and function of queuine and queuosine tRNA

Changes of phosphorylation of membrane-associated proteins following treatment of HeLa cells with the guanine analogue queuine

Regulation of signalling by receptor tyrosine kinases in HeLa cells involves the Q-base

Involvement of protein kinase C in the control of tRNA modi®cation with queuine in HeLa cells

Modulation of mammalian cell proliferation by a modi®ed tRNA base of bacterial origin

Summary: the modi®ed nucleosides of RNA

Structures from posttranscriptionally modi®ed nucleosides from RNA

Ef®cient in vitro synthesis of biologically active RNA and RNA hybridisation probes from plasmids containing a bacteriophage SP6 promoter

The role of RNA pseudoknot stem 1 length in the promotion of ef®cient À1 ribosomal frameshifting

Detection of unique tRNA species in tumor tissues by Escherichia coli guanine insertion enzyme

Analysis of RNA hydrolyzates by liquid chromatography-mass spectrometry

CD4 expressing human 293 cells as a tool for studies in HIV-1 replication: the ef®-ciency of translational frameshifting is not altered by HIV-1 infection

Mitogenic stimulation of HeLa cells increases the activity of the anoxic stress protein, LDH 6/k: suppresseion by queuine

Molecular Cloning: A Laboratory Manual

DNA sequencing with chain-terminating inhibitors

Versatile vectors to study recoding: conservation of rules between yeast and mammalian cells

RNA pseudoknots: translational frameshifting and readthrough on viral RNAs

gag-pro ribosomal frameshift signal: loop lengths and stability of the stem regions

Translational frameshifting in the Escherichia coli dnaX gene in vitro

Sequence requirements for ef®cient translational frameshifting in the Escherichia coli dnaX gene and the role of an unstable interaction between tRNA Lys and an AAG lysine codon

Unusual anticodon loop structure found in E. coli lysine tRNA

Three-dimensional structure of hyper-modi®ed nucleoside Q located in the wobbling position of tRNA

We thank Dr Guillaume Stahl and Professor Jean-Pierre Rousset for plasmid pAC74Z and advice on vector construction, Dr George Garcia for plasmid pTGT5, highly puri®ed TGT and advice on TGT puri-®cation, and Dr Susumu Nishimura for the kind gift of queuine. The skilful technical assistance of Kerstin Jacobsson and Gunilla Ja È ger is greatly acknowledged. This work was supported by the Biotechnology and Biological Sciences Research Council UK (project no. C07089, to IB), the Wellcome Trust UK (prize studentship to B.M.), the Swedish Cancer Society (project no. 680, to Glenn R. Bjo È rk) and the Swedish Natural Science Research Council (project no BBU2930-105, to Glenn R. Bjo È rk). We thank Dr Paul Digard for helpful comments on the manuscript. We are indebted to Dr Richard Jackson (Department of Biochemistry, University of Cambridge) for providing the RRL tRNA depletion methodology.