RNA Sequence and Two-dimensional Structure Features Required for Efficient Substrate Modification by the Saccharomyces cerevisiae RNA:Ψ-Synthase Pus7p*

The RNA:pseudouridine (Ψ) synthase Pus7p of Saccharomyces cerevisiae is a multisite-specific enzyme that is able to modify U13 in several yeast tRNAs, U35 in the pre-tRNATyr (GΨA), U35 in U2 small nuclear RNA, and U50 in 5 S rRNA. Pus7p belongs to the universally conserved TruD-like family of RNA:Ψ-synthases found in bacteria, archaea, and eukarya. Although several RNA substrates for yeast Pus7p have been identified, specificity of their recognition and modification has not been studied. However, conservation of a 7-nt-long sequence, including the modified U residue, in all natural Pus7p substrates suggested the importance of these nucleotides for Pus7p recognition and/or catalysis. Using site-directed mutagenesis, we designed a set of RNA variants derived from the yeast tRNAAsp(GUC), pre-tRNATyr(GΨA), and U2 small nuclear RNA and tested their ability to be modified by Pus7p in vitro. We demonstrated that the highly conserved U-2 and A+1 residues (nucleotide numbers refer to target U0) are crucial identity elements for efficient modification by Pus7p. Nucleotide substitutions at other surrounding positions (-4, -3, +2, +3) have only a moderate effect. Surprisingly, the identity of the nucleotide immediately 5′ to the target U0 residue (position -1) is not important for efficient modification. Alteration of tRNA three-dimensional structure had no detectable effect on Pus7p activity at position 13. However, our results suggest that the presence of at least one stem-loop structure including or close to the target U nucleotide is required for Pus7p-catalyzed modification.

RNA:pseudouridine (⌿)-synthases catalyze the post-transcriptional U to ⌿ conversion in RNAs. Extensive studies during the last 10 years allowed the discovery and characterization of almost the complete set of bacterial and yeast enzymes responsible for this modification in tRNAs, rRNAs, and small nuclear RNAs (snRNAs). 3 All known RNA:⌿-synthases share a set of conserved amino acid sequence motifs and can be grouped into six distinct families based on the degree of amino acid sequence homology: families related to TruA, TruB, RluA, RsuA, TruD, and PsuX, respectively (1)(2)(3). The Saccharomyces cerevisiae RNA:⌿-synthase Pus7p belongs to the recently described TruD-related family of proteins whose members are conserved in all kingdoms of life but have no apparent sequence homology with the other RNA:⌿-synthases. However, strong homologies were found at the level of the three-dimensional structure of the catalytic site (2). Pus7p was initially fished out in a high throughput screening aiming to characterize U2 snRNA-specific pseudouridylation activity. By this approach, ⌿ 35 -forming activity in U2 snRNA was attributed to the YOR243c open reading frame, designed as the PUS7 (pseudouridine synthase 7) gene (4). Later, by analysis of in vivo RNA modification defects resulting from deletion of the PUS7 gene and by in vitro tests using the recombinant protein, we identified other targets of the Pus7p enzyme (U 13 in several yeast cytoplasmic tRNAs and U 35 in the intron-containing pre-tRNA Tyr (G⌿A)) (5). Recently, the activity of Pus7p at position 50 in 5 S rRNA was also demonstrated (6).
The activity of the Escherichia coli homologue of the yeast Pus7p (TruD protein) (2) was also characterized. Since U 35 in the bacterial tRNA Tyr (GUA) is not converted to ⌿ and since snRNAs do not exist in these organisms, the activity of the bacterial TruD enzyme seems to be restricted to U 13 conversion in tRNAs (2). TruD-like proteins are also present in archaea. However, their activity and specificity have not been studied up to now.
The rules governing RNA substrate recognition by RNA:⌿synthases and RNA:modification enzymes in general have only been elucidated in a few cases (for a review, see Ref. 7). Some of the characterized RNA:⌿-synthases have a rather strict substrate specificity and are only able to modify one position in only one type of cellular RNA, such as, for instance, tRNAs or even in only one cellular RNA, like rRNA. This is the case for several of the characterized yeast RNA:⌿-synthases. Pus5p modifies a unique position in the mitochondrial 21 S rRNA (8), Pus6p modifies only position 31 in both cytoplasmic and mito-chondrial tRNAs (9), and Pus8p and Pus9p modify position 32 in the cytoplasmic and mitochondrial tRNAs, respectively (10). Finally, Pus4p, which converts the universally conserved U55 residue into a ⌿ residue in all elongator tRNAs (11), is the only one well studied example of RNA:⌿-synthase acting at a unique position in tRNAs. It was demonstrated that this strict substrate specificity depends on the universally conserved G 53 UUCNANNC 60 sequence in most tRNAs and on the particular three-dimensional structure of the T⌿C loop, which is stabilized by a reverse Hoogsteen interaction between residues U54 and A58. Mutations, introduced at any of the conserved positions, completely abolished or considerably reduced the modification efficiency (12). Very similar results were also obtained for bacterial TruB (13). The substrate specificity of other characterized yeast RNA:⌿-synthases, Pus5p, Pus6p, Pus8p, and Pus9p enzymes, was not studied in detail, so that the sequence and structure requirements for activity of these enzymes are not known.
In contrast to the yeast RNA:⌿-synthases Pus4p, Pus5p, Pus6p, Pus8p, and Pus9p, both Pus1p and Pus7p are peculiar in acting on different types of substrates: eight positions in tRNAs and position 44 in U2 snRNA for Pus1p (14 -16) and position 13 in cytoplasmic tRNAs, 35 in the pre-tRNA Tyr (G⌿A), 35 in U2 snRNA, and 50 in 5 S rRNA for Pus7p (4 -6). One can ask the question how RNA:⌿-synthase Pus7p can recognize such different substrates with a high degree of specificity. These substrate RNAs have different size, sequence, and also different two-and three-dimensional structures.
RNA substrate specificity of yeast Pus1p was not systematically studied, but, taken together, the previously published and unpublished observations indicate that Pus1p modifies multiple accessible uridines in a flexible segment between two RNA helices or even at the extremity of the helix (14 -16). In some cases, the preference for a purine residue immediately 3Ј to the modified U was observed, but this tendency seems not be general for all Pus1p substrates.
RNA regions modified by Pus7p also show no obvious common features at the level of the two-dimensional structure: U 13 in cytoplasmic tRNAs, U 35 in the pre-tRNA Tyr (G⌿A), and U 50 in 5 S rRNA are located in helical regions according to the two-dimensional structures proposed for these RNAs (Fig. 1) (17)(18)(19). In tRNAs, the stability of these helices are quite low. For yeast 5 S rRNA, several alternative two-dimensional structures have been proposed (see Fig. 1A) (20,21). The base-pairing patterns differ in the region covering ⌿ 50 , and this nucleotide appears to be unpaired in the unified two-dimensional structure proposed for eukaryotic 5 S rRNA (22). In contrast, U 35 in U2 snRNA is located in a single-stranded RNA region (23,24) (Fig. 1A). On the other hand, inspection of the Pus7p RNA targets in S. cerevisiae revealed a high conservation of a 7-nt-long sequence, including the target U residue (5). The newly identified Pus7p target in 5 S rRNA (U 50 ) is also surrounded by the same conserved sequence (6) (Fig. 1B).
Site-directed mutagenesis of the plant Arabidopsis thaliana pre-tRNA Tyr (G⌿A), which also contains an intron and is pseudouridylated at position 35 had revealed the importance of the U 33 N 34 U 35 A 36 Pu 37 sequence for efficient ⌿ 35 formation in a wheat germ and HeLa extracts (25).
To test whether the sequence recognition is a general feature of the eukaryal Pus7-like enzymes and if the conserved residues have an equal importance in different types of RNA substrates, we produced variants of each of the yeast Pus7p substrates (tRNA Asp (GUC), pre-tRNA Tyr (G⌿A), and U2 snRNA) with base substitutions at positions Ϫ4, Ϫ3, Ϫ2, Ϫ1, ϩ1, ϩ2, or ϩ3 and compared the activity of the recombinant Pus7p on these variant RNAs. Since the cytoplasmic tRNA Ala (ICG), which contains the conserved 7-nt long sequence including position 13, except for a C to U substitution at position Ϫ2, is not pseudouridylated at position 13 in vivo, we tested whether it can be modified in vitro after C11U substitution. Several mitochondrial tRNAs also have a sequence around the U 13 residue that fits to the consensus sequence of the Pus7p substrates. Therefore, we also tested the in vitro activity of recombinant Pus7p on these tRNAs. Finally, we produced several variants of the tRNA Asp (GUC), pre-tRNA Tyr (G⌿A) and U2 snRNA substrates of Pus7p that were designed in order to modify their two-and three-dimensional structures. Then we tested the activity of the recombinant Pus7p on these RNA variants. Altogether, the data obtained for all three types of Pus7p RNA substrates tested in this study demonstrate that two highly conserved residues around the target U are most important for Pus7p modification and that other conserved or semiconserved residues have a modulator effect on the activity, depending on the identity of the RNA. Furthermore, the presence of some stable RNA twodimensional structure elements was found to reinforce the activity, whereas a highly compact three-dimensional structure may decrease the action of Pus7p.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-The haploid yeast strain BY4742 MAT␣ his3⌬1 leu2⌬0 lys2⌬0 ura3⌬0 from the EUROSCARF collection was used as a source of genomic DNA for PCR amplification.
The plasmids used for in vitro transcription of tRNA Asp (GUC) and its mutants ⌬T⌿-SL, MC3, and MC4 were kindly provided by C. Florentz (IBMC, Strasbourg, France). The construct bearing the sequence of S. cerevisiae pre-tRNA Tyr (G⌿A) under the control of T7 promoter was described previously (5). Plasmid pT7U2 Sc , kindly provided by P. Fabrizio, was used for in vitro transcription of S. cerevisiae U2 snRNA fragment 1-139 (14). Other sequences of tRNA genes used in the study were PCR-amplified and inserted into the SmaI site of pUC19, along with the T7 promoter sequence at 5Ј and BstNI (MvaI) restriction site at 3Ј. Deletion of the 14-nt intron from pre-tRNA Tyr (G⌿A) construct was performed by PCR, using the following primers: 5Ј-ATCTTGAGATCGGG-CGTTCGACTCGCCCCCG-3Ј and 5Ј-TTACAGTCTTGCG-CCTTAAACCAACTTGG-3Ј. Amplified linear DNA was selfligated and used to transform competent E. coli cells. DNA matrices for in vitro transcription of minisubstrates and U2 snRNA truncated variants were produced by PCR or annealing of complementary DNA oligonucleotides. The forward primers used for PCR amplification generated the T7 RNA polymerase promoter. All variants of tRNA Asp (GUC), pre-tRNA Tyr (G⌿A), and U2 snRNA with point mutations were made by site-directed mutagenesis using the QuikChange kit (Stratagene). The sequences of all the generated recombinant plasmids were verified by DNA sequencing.
In vitro T7 RNA polymerase transcription with or without incorporation of the appropriate [␣-32 P]NTP and purification of the resulting RNA transcripts by electrophoresis on denaturing gel were performed as previously described (5,26).
Cloning, Expression, and Purification of the Recombinant His 6 -Pus7p-Preparation of the pET28-PUS7 plasmid used for expression of the recombinant S. cerevisiae Pus7p was described previously (5). N-terminally His 6 -tagged Pus7p protein was expressed in E. coli strain BL21(DE3)-RIL (Stratagene). The transformed cells were grown for 5 h in autoinduction ZYM-5052 medium (27) at 37°C followed by 20 h at 20°C. Cells where then harvested and disrupted in buffer A (20 mM Tris-HCl, pH 7.5, 1 M NaCl) by a 2 ϫ 2-min sonication at 4°C using a Branson-250 sonicator (50% duty cycle, Output 6). The lysate was cleared by centrifugation (20,000 ϫ g for 30 min) and was directly applied to a column of nickel-Sepharose (Amersham Biosciences) charged with Ni 2ϩ . The column was first washed with buffer A and then 50 mM imidazole in Buffer A, and the protein was eluted using 500 mM imidazole in the same buffer. Fractions containing the Pus7p were pooled, concentrated to 10 mg/ml in a buffer containing 10 mM Tris-HCl, pH 7.5, 300 mM NaCl using a Vivaspin YM-30 (Vivascience, France) concentration cell, and stored at Ϫ80°C in 50% glycerol.
In Vitro Activity Tests Using the Nearest Neighbor Approach-The enzymatic activity of the recombinant Pus7p was assayed in 100 mM Tris-HCl buffer, pH 8.0, containing 100 mM ammonium acetate, 5 mM MgCl 2 , 2 mM dithiothreitol, and 0.1 mM EDTA. All RNA substrates, diluted in 9 l of the reaction buffer, were renatured by incubation for 5 min at 80°C, followed by a 20-min incubation at room temperature. Two different ratios of [Pus7p]/[RNA] were used. In the first series of tests (excess of Pus7p), we used only 32 P-labeled RNA substrate (1 nM final concentration), and in the second series of tests (excess of RNA) we used the same amount of 32 P-labeled RNA but complemented by 10 pmol of nonradiolabeled RNA transcript (final concentration of 1 M). The reaction was initiated by the addition of purified Pus7p diluted in the reaction buffer: 1 M (conditions of Pus7p excess) or 10 nM (excess of RNA substrate). After 90 min of incubation at 30°C, the modified RNAs were phenol-extracted, ethanol-precipitated, and digested overnight by RNase T2 (0.01 units/l) in 50 mM ammonium acetate buffer, pH 4.6. The resulting 3Ј-nucleotide monophosphates were fractionated by two-dimensional chromatography on thin layer cellulose plates (Polygram CEL 400; Macherey-Nagel, Germany). An isobutyric acid, 25% ammonia, water (66:1:33, v/v/v) mixture was used for the first dimension, whereas the second dimension was done in 2-propanol, 37% HCl, water (68:17.6:14.4, v/v/v). Assignment of nucleotides was based on previously published maps (28). Radioactive spots were quantified on a PhosphorImager instrument (Typhoon 9410; Amersham Biosciences), using the Image-Quant software.

Importance of the Conserved Sequence in RNA Substrates for
Pus7p Activity in Vitro-The sequence alignment of the identified S. cerevisiae Pus7p substrates (Fig. 1B) revealed the strict conservation of a U residue at position Ϫ2 and an A residue at position ϩ1 (as referred to the modified uridine). A G or a C residue is found at position Ϫ3, a purine residue at both positions Ϫ4 and ϩ2, and a pyrimidine residue at position ϩ3. In contrast, any nucleotide can be found at position Ϫ1, suggesting that the identity of this residue is not important for Pus7p activity.
In order to test the influence of this conserved sequence on Pus7p activity, a set of tRNA Asp (GUC), pre-tRNA Tyr (G⌿A), and 5Ј-terminal regions of U2 snRNA variants were produced. The 5Ј-terminal region of U2 snRNA (positions 1-139), which contains all the functional elements of U2 snRNA, was previously shown to be sufficient for Pus7p activity (4). Versions of these three RNAs with any nucleotide at positions Ϫ4, Ϫ3, Ϫ2, ϩ1, and ϩ2 were produced by in vitro transcription. Variant RNAs with a pyrimidine to purine substitution at position ϩ3 were also generated. In addition, one variant with a base substitution at position Ϫ1 was used as a control.
The recombinant Pus7p was produced as previously described (5), and its activity on the variant and WT RNAs was tested both in the presence of a large excess of RNA substrate (1 M of RNA for 10 nM of enzyme) and in conditions where the enzyme was in large excess (about 1 nM labeled RNA alone for 1 M enzyme). The level of RNA modification by Pus7p was measured by the nearest neighbor approach after a long incubation period (90 min). This incubation time was selected based on the observation that complete modification of each of the WT RNAs was obtained after about 30 min of incubation under the same conditions. If some of the conserved residues in the RNA substrate were important for Pus7p activity, after their replacement, we expected to detect a decreased level of ⌿ formation after 90 min of incubation. For each variant RNA, the identity of the [␣-32 P]NTP used for transcription was defined according to the identity of the residue at position ϩ1 as referred to the modified uridine. After incubation with the enzyme, the RNAs were digested with RNase T2, and the released 3Ј-phosphate mononucleotides were fractionated by two-dimensional thin layer chromatography. The yield of U to ⌿ conversion was evaluated by measurement of the radioactivity in the fractionated 3Ј-phosphate mononucleotides, as previously described (5). Quantification of the formed ⌿ residue is less accurate for U2 snRNA as compared with tRNA Asp (GUC) and pre-tRNA Tyr (G⌿A), because of its length and a high occurrence of UA dinucleotides in this RNA. To verify that the observed ⌿ residue formed in the WT RNAs indeed corresponded to the expected position of U to ⌿ conversions, control experiments were performed with variants bearing U to C substitution of the target U (U13C tRNA Asp (GUC), U35C pre-tRNA Tyr (G⌿A), and U35C snRNA U2 5Ј-terminal region transcripts). As expected, no significant formation of ⌿ residue was detected in these mutant RNAs after a 90-min incubation with Pus7p (see supplemental material). These results confirmed the strict specificity of Pus7p for a single position within all of these RNA Recognition by the Yeast Pus7p FEBRUARY 27, 2009 • VOLUME 284 • NUMBER 9

JOURNAL OF BIOLOGICAL CHEMISTRY 5847
substrates; thus, we proceeded to definition of the identity elements required for this RNA recognition specificity.
Residues at Positions Ϫ2 and ϩ1 Play a Crucial Role for U to ⌿ Conversion by Pus7p-In vitro modification of the WT and variants of tRNA Asp (GUC) and pre-tRNA Tyr (G⌿A) by recombinant Pus7p was performed both in the excess of RNA substrate (Table 1) and enzyme (Pus7p) (Fig. 2). All tests were done in triplicate, and the mean values of the U to ⌿ conversion level, together with the S.D. value, for each RNA substrate are given in Table 1 and in Fig. 2. The autoradiograms of the representative two-dimensional TLC obtained in one of these three series of experiments are shown as examples in Fig. S1. Under both experimental conditions, a major observation is that the yield of ⌿ formation strongly depended on the position of the base substitution. Second, nearly identical results were obtained when using an excess of substrate or an excess of enzyme (Fig.  2). This might reflect the occurrence of structural heterogeneity of some of the variant RNAs, with only one part of the molecules being able to be modified. Since base substitutions in the tRNA Asp (GUC) were expected to alter the tRNA two-dimensional structure, we tested whether the presence of compensatory mutations at position 24 in the D-stem could compensate for the strong decrease of ⌿ formation found for base substitu-tions at position 11 (position Ϫ2 in the conserved sequence). As shown in Table 1, restoration of a Watson-Crick base pair between nucleotides 11 and 24 in the D-stem did not increase  (18,19,20,23,35). Alternative structures proposed for yeast 5 S rRNA are shown. Helices are labeled in Roman numerals for U2 snRNA and 5 S rRNA. Only the ⌿ residues formed by Pus7p (5,6,14) are shown in dark circles; the other post-transcriptional modifications of these RNAs are not represented. The intronic sequence in pre-tRNA Tyr (G⌿A) is in small characters. The residues at positions Ϫ4 to Ϫ1 and ϩ1 to ϩ3 as referred to the ⌿ residue formed by Pus7p (position 0) are indicated. In the yeast tRNA Asp (GUC) and pre-tRNA Tyr (G⌿A) transcripts, the A 1 -U 72 and C 1 -G 72 base pairs were both converted into a G 1 -C 72 in the RNA transcripts, in order to increase transcription efficiency. B, multiple sequence alignment of the Pus7p RNA substrates. The alignment of the sequences from positions 22-59 in U2 snRNA, 37-74 in 5 S rRNA, 22-45 in pre-tRNA Tyr (G⌿A), and 1-34 in cytoplasmic tRNAs is shown. The names of the RNAs are given in the first column with the anticodon sequence shown for tRNAs (3, mcm5s2U; unknown nucleotide; N, unknown modified uridine; I, inosine; 1, 5-methoxycarbonylmethyluridine; !, 5-carboxymethylaminomethyluridine). The conserved U Ϫ2 and A ϩ1 residues are indicated by black boxes, and the ⌿ residues formed by Pus7p are in black circles. The conserved sequence present in all Pus7p RNA substrates is shown at the bottom (R, purine; Y, pyrimidine; N, any nucleotide).
S. cerevisiae pre-tRNA Tyr (GΨA)  the Pus7p activity. Therefore, modification of the sequence and not of RNA structure is responsible for the observed effect. The levels of ⌿ formation in both tRNA transcripts bearing base substitutions at both positions Ϫ2 and ϩ1 were nearly identical to those formed in control molecules (U13C substitution in tRNA Asp (GUC) and U35C substitution in the pre-tRNA Tyr (G⌿A)). These data clearly indicated that base substitutions at positions Ϫ2 and ϩ1 in both tRNA Asp (GUC) and pre-tRNA Tyr (G⌿A) almost completely abolish U to ⌿ conversion by yeast Pus7p. The same strong negative effect of base substitutions at positions Ϫ2 and ϩ1 was also found in U2 snRNA (see supplemental material).
Hence, interaction with residues U Ϫ2 and A ϩ1 in any RNA substrate is probably required to achieve one of the steps in the overall Pus7p catalytic process. These data are in perfect agreement with the strict conservation of U Ϫ2 and A ϩ1 in all natural RNAs modified by Pus7p in vivo (Fig. 1B) and with previous data obtained for modification of the A. thaliana pre-tRNA Tyr (G⌿A) in a wheat germ extract (25).
Base substitutions at the four other positions, which are less strictly conserved (R Ϫ4 , G/C Ϫ3 , R ϩ2 , Y ϩ3 ), had much less dramatic effects on the yield of U to ⌿ conversion in all of the three substrates tested (Table 1 and Fig. S1, A-C). However, mutations at positions Ϫ3 and Ϫ4 had almost opposite effects in tRNA Asp (GUC) and pre-tRNA Tyr (G⌿A). This may be linked to their respective RNA two-dimensional structures. Indeed, the base Ϫ3 in tRNA Asp (GUC) is base-paired with the nucleotide in the D-stem, whereas nucleotide Ϫ3 in the pre-tRNA Tyr (G⌿A) is in the single-stranded region. On the other hand, position Ϫ4 in tRNA Asp (GUC) is not involved in the interaction with other nucleotides, whereas the base-pairing of the nucleotide at position Ϫ4 in the pre-tRNA Tyr (G⌿A) is expected to stabilize the RNA two-dimensional structure (18) (Fig. 1A). In consequence, substitution of a nucleotide at position Ϫ3 in tRNA Asp (GUC) had a greater effect on Pus7p activity than substitution at position Ϫ4, whereas a more important effect is observed when the nucleotide at position Ϫ4 was mutated in the pre-tRNA Tyr (G⌿A).
Interestingly, substitution of residue A ϩ2 had a marked negative effect in both the tRNA Asp (GUC) and pre-tRNA Tyr (G⌿A). In addition, in both substrates, the A to U substitution had the lowest negative effect, whereas the A to C substitution had the greatest negative effect (Table 1 and Fig. 2). As expected, mutation at position Ϫ1, where any one of the four possible nucleotides may be found in Pus7p RNA substrates, had no effect on the yield of modification of both the tRNA Asp (GUC) and pre-tRNA Tyr (G⌿A). This was also the case for the pyrimidine to purine substitution at position ϩ3 (Table  1 and Fig. 2).

Restoration of the Consensus Sequence Converts Some tRNAs into Pus7p
Substrates-Only three of the yeast cytoplasmic tRNAs that have a U residue at position 13 are not substrates of Pus7p (Fig. 3A). Interestingly, these three tRNAs (tRNA Ala -(IGC) and two isoacceptors of tRNA Arg (mcm 5 UCU)) (29) carry a C instead of a U residue at position 11 (position Ϫ2 in the consensus sequence of Pus7p substrates). Based on the above data, we expected that a C11U mutation in these tRNAs will convert them into Pus7p substrates. To test this hypothesis, we used the tRNA Ala (IGC) as a model. The activity of the recombinant Pus7p was tested both on a WT tRNA Ala (IGC) transcript and on a transcript with the C11U substitution. As illustrated in Fig. 3B, whereas an insignificant level of ⌿ formation was detected in the WT RNA transcript, the C11U variant was almost completely modified (U to ⌿ conversion of 95 Ϯ 3%). These data further demonstrated the requirement of a U residue at position Ϫ2 for Pus7p activity.
Recombinant Pus7p Can Modify Mitochondrial tRNAs with U Ϫ2 , U 0 , and A ϩ1 Residues-Interestingly, despite the fact that no ⌿ 13 residue was detected in yeast mitochondrial tRNAs (29), the sequence found for several of them around residue U 13 fits to the consensus sequence of Pus7p RNA substrates (Fig.  3A). The absence of U 13 pseudouridylation in mitochondria could therefore be due either to an inability of Pus7p to act on mitochondrial tRNAs, because of differences in two-dimensional structures as compared with cytoplasmic tRNAs, or simply due to the absence of Pus7p in mitochondria. To clarify this point, we tested the in vitro activity of Pus7p on one of the mitochondrial tRNAs having a U at position 13 surrounded by the sequence that is conserved in Pus7p substrates, namely tRNA Trp (UCA). A significant level of in vitro modification of this tRNA by Pus7p was obtained (77 Ϯ 2%), showing that the absence of ⌿ 13 modification in mitochondrial tRNAs is probably due to the absence of Pus7p in mitochondria. This hypothesis is in agreement with the very low probability for a mitochondrial localization of Pus7p (30), which was also found when using various computer programs (PSORT I and II, Sub-Loc, ESLPred, and YPLS) developed for the prediction of protein subcellular localization (31)(32)(33)(34).
⌿ 13 Formation in tRNA Asp (GUC) Does Not Depend on Correct tRNA Three-dimensional Structure but Is Increased by the Presence of RNA Helices-In the tRNA Asp (GUC), the residues Ϫ4 (A 9 ), ϩ1 (A 14 ), and ϩ2 (A 15 ) are involved in the stabilization of the RNA three-dimensional structure; A 9 is involved in the A 9 -A 23 -U 12 triple-base pair, A 14 forms the conserved U 8 -A 14 interaction, and A 15 forms the semiconserved purine-pyrimidine Levitt base pair A 15 -U 48 (17,35). The residue A 14 (ϩ1) is particularly important for efficiency of Pus7p modification. However, since the same requirement was also found for pre- tRNA Tyr (G⌿A) and U2 snRNA, the strong negative effect upon the replacement of residue ϩ1 is probably not related to the three-dimensional tRNA structure. In line with this, the mutation at position Ϫ4 (A 9 ) has almost no influence on the efficiency of modification, whereas its substitution affects tRNA three-dimensional folding. To verify that the tRNA three-dimensional structure is indeed not important for Pus7p activity and to test for a possible effect of the tRNA two-dimen-sional structure on this activity, we produced a large series of tRNA Asp (GUC) variants.
The three-dimensional tRNA structure was disrupted to a different extent in all of these variants; the most important perturbation was introduced in the variant ⌬T⌿-SL (Fig. 4A) where the T⌿-stem-loop, which interacts with the D-loop, was deleted. In the variants ⌬AC-SL and ⌬AA-SL (Fig. 4A) the anticodon stem-loop and acceptor stem, respectively, were missing. In MC3 and MC4 variants, the anticodon stem was either slightly stabilized (MC3) or slightly destabilized (MC4). For each of these RNAs, three independent series of in vitro modification experiments were performed in the excess of RNA as compared with Pus7p (Fig. 4A).
As already demonstrated above, point mutations that destabilize the tRNA D-stem have only a limited effect on the Pus7p activity (Table 1 and Fig. 2). To go one step further in the destabilization of this stem, the sequence of its 3Ј-strand was completely mutated in variant OD-SL (Fig. 4A). Finally, two minisubstrates corresponding to the D-stem-loop of tRNA Asp (GUC) and to the anticodon stem-loop containing the intron of the pre-tRNA Tyr (G⌿A), respectively, were produced (Fig. 4B).
Altogether, the results of these experiments with all eight RNA variants produced (Fig. 4) confirmed that the tRNA three-dimensional structure is not really required for Pus7p activity. The deletion of the acceptor stem (⌬AA-SL) had almost no effect on the activity, whereas elimination of one of the stem-loop structures other than the D-stem-loop (⌬AC-SL, ⌬T⌿-SL) or the destabilization of the anticodon stem loop (MC4) only slightly decreased the yield of modification.
A stronger negative effect was observed when the conformation of the D-stem-loop where modification occurs was altered (the level of ⌿ residue formed in OD-SL was divided by 2), and a quite low, but detectable, level of modification was obtained for the two minisubstrates (28 and 32%) (Fig. 4B). We noticed a slight increase of ⌿ formation upon stabilization of the anticodon stem (variant MC3). These data suggested that at least one stem-loop structure, containing the residue to be modified, is required for Pus7p activity. In addition, the presence of other stable stem-loops, which may facilitate the overall folding of the tRNA, probably also favors Pus7p activity.
⌿ 35 Can Be Formed in the Mature tRNA Tyr (G⌿A) in Vitro-Alignment of the sequence surrounding ⌿ 35 in the mature S. cerevisiae tRNA Tyr (G⌿A) with the sequences containing ⌿ 13 in Pus7p tRNA substrates reveals that intronless tRNA Tyr (G⌿A) fits almost perfectly to the consensus sequence of Pus7p RNA substrates (Fig. 5A). The only difference is the substitution of the pyrimidine at position ϩ3 by an A residue. As shown above, substitution of this pyrimidine at position ϩ3 by an A residue in both tRNA Asp (GUC) and pre-tRNA Tyr -(G⌿A) had no marked effect on the yield of U to ⌿ conversion by Pus7p in vitro. Therefore, one might expect that the recombinant Pus7p is able to modify the mature tRNA Tyr (G⌿A) in B.
C.   vitro. To verify this hypothesis, modification experiments were performed in parallel on the precursor and the mature tRNA Tyr (G⌿A), using the excess of RNA substrate. The yield of modification for both RNA transcripts was tested after incubation with either a 1 nM, 10 nM, or 1 M concentration of the recombinant Pus7p. Interestingly, whereas 81% of U 35 was converted into ⌿ 35 in the pre-tRNA Tyr (G⌿A) upon incubation with 10 nM Pus7p, no significant ⌿ 35 formation was detected in the mature RNA under these conditions (Fig. 5A). However, when the protein concentration was increased to 1 M, even mature tRNA Tyr (G⌿A) transcript was modified to a significant extent (63%). The modification occurred at position 35, since it was not observed in the same transcript bearing the U35C substitution (Fig. 5B). How can one explain the difference in modification of tRNA Asp (GUC) variants and mature tRNA Tyr (G⌿A)? Indeed, A ϩ3 variants of tRNA Asp (GUC) or pre-tRNA Tyr (G⌿A) are modified by 10 nM Pus7p to 76 -79% (Table 1), whereas mature tRNA Tyr (G⌿A), which has the same sequence surrounding U 35 , remains unmodified. One possible explanation could be the highly constrained three-dimensional structure of the anticodon loop containing U 35 in mature tRNA Tyr (G⌿A). Indeed, the residue U 33 (position Ϫ2) in the anticodon loop is known to be involved in a peculiar ribose-phosphate backbone folding that is present in some tRNA (U 33 -turn). The stability of this U 33 -turn has been attributed to the formation of three interactions involving U 33 : one hydrogen bond formed by the 2Ј-OH of residue U 33 and the N-7 of the purine residue at position 35, a second hydrogen bond formed between the N3-H of residue U 33 and the phosphate of the nucleotide N 36 , and the stacking of residue U 33 on the residue 35 (35). Therefore, residue U 33 , which is the important U Ϫ2 residue in the conserved sequence of Pus7p substrates, might have an inappropriate conformation in the anticodon loop of the mature tRNA Tyr (G⌿A). The other residue essential for Pus7 activity (position ϩ1 in the Pus7 conserved sequence, A 36 in mature tRNA Tyr (G⌿A)) might also be poorly accessible in the anticodon loop.
If this were the case, insertions of residues between positions 31 and 32 in the anticodon loop would abolish the three-dimensional structure constraints and therefore reinforce Pus7p activity. Indeed, we observed a significant ⌿ 35 formation in both ϩA and ϩAA variants of mature tRNA Tyr (G⌿A) upon their incubation in the presence of 10 nM Pus7p. These two variants are modified to 61 and 58%, respectively (Fig. 5C).
In contrast, the A38U substitution in the loop that restored the complete Pus7p consensus sequence did not allow ⌿ 35 formation in the mature tRNA Tyr (G⌿A) when the same enzymatic conditions were used (10 nM Pus7p enzyme) (Fig. 5C). These data strongly suggest that the three-dimensional structure of the anticodon loop in mature tRNA Tyr (G⌿A) strongly limits its modification by Pus7p.
The Presence of Two-dimensional Structural Motifs Is Required for Efficient Modification of U2 snRNA in Vitro-In the 5Ј-terminal region of yeast U2 snRNA, residue U 35 is flanked by two stem-loop structures, I and IIa. Stem-loop IIa is immediately followed by a third stem-loop structure, IIb, and a pseudoknot can be formed between the terminal stem-loop IIa and the single-stranded segment located downstream from stem-loop IIb (Fig. 6A). An additional potential target uridine surrounded by a consensus sequence is present in yeast U2 snRNA at position 56; however, this nucleotide is modified neither in vivo nor in vitro, using the recombinant Pus7p.
To test the importance of the U2 snRNA two-dimensional structure for Pus7p activity, we first eliminated the sequence downstream from position 85 in order to disrupt the pseudoknot structure (variant U2-(1-85)). Then the stem-loop structures I and IIb were individually or simultaneously deleted (variants U2-(1-85) ⌬I, U2-(1-85) ⌬IIb, and U2-(1-85) ⌬I⌬IIb), and all three stem-loop structures were eliminated in the RNA variant U2- (28 -47) (Fig. 6B). Disruption of the pseudoknot structure only slightly decreased the yield of ⌿ 35 formation. In agreement with previous data (4), the elimination of stem-loop structure I also had only a limited effect on the yield of ⌿ 35 formation, whereas deletion of stem-loop IIb alone or together with stem-loop I decreased the level of modification by a factor of about 2. In addition, Pus7p did not modify the RNA variant U2-(28 -47) without any double-stranded region. Therefore, altogether the data obtained on both U2 snRNA and the tRNA Asp (GUC) strongly suggested that the presence of stable two-dimensional structure motifs increases the level of in vitro activity of Pus7p.

DISCUSSION
In this study, we performed a deep characterization of the RNA sequence and structure requirements needed for modification of three distinct RNA substrates by the yeast RNA:⌿synthase Pus7p. The data obtained are especially important, since yeast Pus7p is presently the RNA:⌿-synthase with the largest spectrum of identified RNA substrates (tRNA, pre-tRNA Tyr (G⌿A), U2 snRNA, and 5 S rRNA). Our data obtained in vitro point out the importance of a given sequence in a flexible RNA segment for efficient Pus7p recognition and modification. However, both in vitro and in vivo, many potential U to ⌿ conversion sites are not modified by Pus7p. Even if the sequence of these RNA segments is suitable for recognition, their flexibility may not be sufficient for correct access of the target uridine into the Pus7p active site. In addition, such potential target uridines may by buried inside of stable RNA structures or covered by numerous associated proteins in vivo. The competition between the cognate and near cognate RNA substrates may also influence the outcome of the reaction in the cell. All of these parameters probably play an important role in the choice of RNA substrates by Pus7p.
Point Sequence Determinants Are Required for Pus7p Activity on Any of Its Substrates-Sequence alignments of all yeast Pus7p substrates reveals that the target U residue is located in a 7-nt-long sequence with three strictly conserved residues (the target U residue, U at position Ϫ2, and A at position ϩ1), the other residues being semiconserved or partially conserved, except the nucleotide at position Ϫ1. Here, we demonstrate that the three strictly conserved residues are absolutely required for Pus7p activity on three of its substrates (tRNA Asp (GUC), pre-tRNA Tyr (G⌿A), and U2 snRNA). Both purine-purine and pyrimidine-pyrimidine replacements at positions Ϫ2 and ϩ1, respectively, abolish the activity. Therefore, we show that the U 33 N 34 U 35 A 36 sequence previously found to be required for ⌿ 35 -forming activity in plant pre-tRNA Tyr (G⌿A) is needed for all Pus7p substrates. These residues may be necessary for stabilization of "active" RNA substrate conformation as well as for stable anchoring of the RNA substrate onto the enzyme or for catalytic activity. The residue at position ϩ2 in the consensus sequence is important for the yield of modification but is not essential as are residues U Ϫ2 and A ϩ1 . The identities of the other partially conserved residues of the 7-nt sequence only have a moderate effect on the modification yield, and these effects vary from one substrate to another, probably due to differences in their respective RNA two-dimensional structures. In the present in vitro tests, we compared the efficiency of variant RNA modification using an end point measurement. These differences in modification efficiency may be even higher in terms of the initial modification rate.
Remodeling of the tRNA Structure Is Expected to Occur for the ⌿ 13 Formation by Pus7p-In the L-shaped canonical three-dimensional structure of tRNAs, the essential U Ϫ2 and A ϩ1 residues in RNA substrates are not accessible for the enzyme. Both of them are involved in base pair interactions; U 11 (U Ϫ2 ) basepairs with A 24 in the D-stem, whereas A 14 (A ϩ1 ) forms a tertiary interaction with U 8 . Furthermore, they are buried into the molecule due to global tRNA three-dimensional structure. Thus, the isomerization of U to ⌿ at position 13 requires considerable structural remodeling. The three-dimensional structure of the complex formed by a tRNA and another RNA modification enzyme acting on a residue in the D-loop has been solved, namely the complex formed by the archaeal tRNA-guanine transglycosylase, which converts G at position 15 into archaeosine in many archaeal tRNAs and one of its tRNA substrates ( Fig. 7) (36). This tRNA carries a U 11 , U 13 , and A 14 residues like Pus7p substrates. In the three-dimensional structure, it fits into the enzyme active site in an open conformation called the -tRNA conformation (36). This conformation is characterized by the opening of the D-stem loop and the interaction of its 3Ј-strand with the variable loop (36). By inspection of yeast tRNA sequences, we noticed that 3-5 contiguous and noncontiguous base pairs can always be formed between the 3Ј-strand of the D-stem and the variable loop, so that all Pus7p tRNA substrates may be able to adopt a -shaped structure (data not shown). In this conformation, the 5Ј-strand of the D-stem containing U 11 and U 13 would be free of base pair interactions, and these two essential residues as well as residue A 14 would become easily accessible as compared with the canonical L-shaped tRNA structure (Fig. 7). Moreover, the conformation of the E. coli TruD active site observed in the crystal does not allow fitting the tRNA substrate in canonical L-form, whereas the alternative -form can be accommodated (37). Taking these arguments together, one can guess that by interaction with its tRNA substrate, Pus7p converts the L-shaped tRNA into a -like shaped tRNA so that the three identity elements (U 11 , U 13 , and A 14 ) become available, allowing U 13 to ⌿ 13 conversion by the catalytic Asp 256 residue.
The Intron in tRNA Tyr (G⌿A) Precursor May Avoid Threedimensional Structural Constraints Limiting Pus7p Activity on the Mature tRNA Tyr (G⌿A)-The presence on an intron in pre-tRNA Tyr (G⌿A) as well as the presence of a ⌿ 35 residue in mature tRNA Tyr (G⌿A) are general features of Eukarya. Previous work by the group of J. Abelson (38) demonstrated that the absence of the intervening sequence in the yeast RNA Tyr ochre suppressor gene (SUP6) leads to the absence of ⌿ 35 modification in the mature tRNA Tyr (G⌿A) in vivo. Later, studies performed by RNA microinjection in Xenopus laevis oocytes showed that ⌿ 35 could be formed in both X. laevis and Drosophila pre-tRNA Tyr (G⌿A), despite the strong sequence and length differences of the introns of these two precursors (39 -41). The conclusion of this study was that the tRNA:⌿ 35 -synthase needs the intron for tRNA Tyr (G⌿A) modification but that ⌿ formation is not dependent on the size and sequence of the interven-ing sequence. A similar observation was made for the plant pre-tRNA Tyr (G⌿A); the presence of an intron is required to obtain a modification, and only the identity of residues U 33 , A 36 , and A 37 have a strong influence on the activity (25). Further studies of ⌿ 35 formation in plant tRNA Tyr pointed out the moderate importance of C 32 (position Ϫ3) and the length of the intron; the minimal size still supporting modification was found to be 7 nt (42).
Here we show that the recombinant Pus7p can form ⌿ 35 residue in the mature tRNA Tyr (G⌿A) but only when used at rather high concentration (1 M), which is probably much superior to that found in vivo. Indeed, experimental measurements have indicated that Pus7p is moderately expressed in S. cerevisiae; its estimated amount is about 4000 molecules/cell (see Yeast GFP Fusion Localization Database, University of California). We could markedly increase the activity of Pus7p on the mature tRNA Tyr (G⌿A) by insertion of one or two A residues in the anticodon loop, suggesting that the low activity of Pus7p on the mature tRNA Tyr (G⌿A) is probably due to the inability of  Residues ⌿ 13 and A 14 are exposed at the surface of the -shaped tRNA observed in the crystal structure of archaeal tRNA-guanine transglycosylase (ArcTGT)-tRNA complex (B) (36) as compared with the canonical L-shaped tRNA structure (A). Ribbon three-dimensional structure models and the two-dimensional secondary structure of the -shaped tRNA Val (UAC) are shown. The region corresponding to the 7-nt conserved sequence found in all Pus7p tRNA substrates is colored in red. Positions of U 11 , ⌿ 13 , and A 14 are shown.