Sequence context effects on oligo(dT) termination signal recognition by Saccharomyces cerevisiae RNA polymerase III.

Eukaryotic RNA polymerase (Pol) III terminates transcription at short runs of T residues in the coding DNA strand. By genomic analysis, we found that T(5) and T(4) are the shortest Pol III termination signals in yeasts and mammals, respectively, and that, at variance with yeast, oligo(dT) terminators longer than T(5) are very rare in mammals. In Saccharomyces cerevisiae, the strength of T(5) as a terminator was found to be largely influenced by both the upstream and the downstream sequence context. In particular, the CT sequence, which is naturally present downstream of T(5) in the 3'-flank of some tDNAs, was found to act as a terminator-weakening element that facilitates translocation by reducing Pol III pausing at T(5). In contrast, tDNA transcription termination was highly efficient when T(5) was followed by an A or G residue. Surprisingly, however, when a termination-proficient T(5) signal was taken out from the tDNA context and placed downstream of a fragment of the SCR1 gene, its termination activity was compromised, both in vitro and in vivo. Even the T(6) sequence, acting as a strong terminator in tRNA gene contexts, was unexpectedly weak within the SNR52 transcription unit, where it naturally occurs. The observed sequence context effects reflect intrinsic recognition properties of Pol III, because they were still observed in a simplified in vitro transcription system only consisting of purified RNA polymerase and template DNA. Our findings strengthen the notion that termination signal recognition by Pol III is influenced in a complex way by the region surrounding the T cluster and suggest that read-through transcription beyond T clusters might play a significant role in the biogenesis of class III gene products.

Eukaryotic RNA polymerase (Pol) 1 III is unique among DNAdependent RNA polymerases in recognizing a simple run of T residues on the coding strand as a termination signal, in the apparent absence of accessory factors. The minimal signal thought to be sufficient to provoke Pol III termination varies among different eukaryotes. T 4 suffices for termination by Xenopus and human Pol III (1,2), T 4 /T 5 for Schizosaccharomyces pombe Pol III (3), and T 5 /T 6 for Saccharomyces cerevisiae Pol III (4). Pol III termination within T clusters is generally heterogeneous and progressive (5), and termination efficiency tends to increase with the length of the T run (4). Mechanistically, Pol III termination involves extensive pausing, dictated by the T cluster itself (5,6). Even during elongation, the addition of three UMP residues in succession is particularly slow (5). Pol III pausing at T clusters sets cycles of hydrolytic RNA chain retraction (7). Both pausing and the associated hydrolytic RNA cleavage are affected by mutations in the C160, C128, and C11 subunits of yeast Pol III (7)(8)(9). In the case of C128 and C11, the same mutations have also been shown to affect the termination properties of Pol III, thus suggesting a functional interplay between hydrolytic RNA cleavage and transcription termination (9,10). The termination-prone character of Pol III is consistent with class III genes being very short, a fact that reduces the probability of unwanted termination signals within the coding regions. At the same time, the high efficiency of Pol III termination, by favoring polymerase recycling and transcription reinitiation, is likely to contribute in a crucial way to the high level of transcript production typical of Pol III (11)(12)(13).
The facility of Pol III termination poses questions that have not yet been addressed exhaustively. One of them relates to the mechanism allowing for easy termination by Pol III. Pol I and Pol II transcription units very often contain long T clusters, yet these two RNA polymerases do not terminate at these sequences, although T clusters are thought to generally induce RNA polymerase pausing. On which molecular features thus resides the distinctive ability of Pol III to terminate efficiently at short T clusters? At least part of the Pol III termination proneness has been attributed to the presence in this enzyme of the Rpc11p subunit, which is endowed with RNA cleavagestimulating activity (9). Such a feature, however, seems not to be unique to Pol III, since Pol I and Pol II both contain subunits that share sequence homology with Rpc11p and might similarly be involved in termination (14). A recent study has further suggested that the termination-prone character of Pol III may also come from a marked propensity of the enzyme toward RNA release, such that every T-rich sequence at which Pol III pauses for a sufficiently long time would become a termination site (15). Such a termination propensity takes us to another important question, arising from the observation that Pol IIItranscribed genes often contain internal stretches of T residues that would be predicted to provoke transcription termination. Is intragenic termination impaired in such cases, and by which mechanisms? A previously characterized example related to this issue is that of a tRNA Lys gene from Xenopus laevis, in which two T 4 tracts, one located at a canonical downstream position and the other one within the gene itself, determine transcription termination to different extents, with the internal T 4 cluster behaving as a much weaker terminator than the external one (16). The idea that the sequence context can generally influence the efficiency of T clusters first emerged from early studies of Pol III termination (1) and was confirmed and extended by later studies of transcription termination on the adenovirus VA RNA genes (17,18). The surrounding sequences seem to influence Pol III termination especially at short T clusters. In Xenopus, for example, only T 4 terminators display a context-dependent strength, whereas clusters of five or more T residues are highly efficient as terminators independently from the context (1). The few documented cases of surrounding sequence effects can hardly be accommodated into general rules of sequence specificity, because the same flanking nucleotides were found to produce different (or even opposite) effects on the terminators of different class III templates (1,16,18). Further adding to the complexity of the sequence requirements for Pol III termination, several observations also suggested the existence of Pol III termination signals that do not conform to the general "T-cluster" rule (18 -20).
In this study, we show that termination at oligo(dT) tracts by S. cerevisiae Pol III can be influenced by the flanking sequences to a previously unsuspected extent, and we address the sequence requirements and molecular mechanisms of such effects.

EXPERIMENTAL PROCEDURES
Statistical Analysis of tDNA 3Ј-Flanking Regions-Coordinates for the tDNA downstream regions were predicted directly from the genomic sequences of S. cerevisiae, S. pombe, Mus musculus and Homo sapiens by using Pol3scan (21) and tRNAscan-SE (22). Predicted tRNA genes were considered for the analysis if the tRNA boundaries and anticodon type were found to be identical by the two methods. Although this criterion does not guarantee the identification of all the genomic tDNAs, it strongly limits the occurrence of false positive results. The portions corresponding to the 100 nucleotides immediately downstream to the mature tRNA coding sequence were extracted as FASTA files and used for the analysis of terminator signals. A Perl script was designed for statistical analysis of T-runs of arbitrary length, making use of the BioPerl library (23).
DNA Templates-For tRNA gene nomenclature, we refer to MIPS (Munich Information Center for Protein Sequences; available on the World Wide Web at mips.gsf.de/proj/yeast/rna/trna.html). The P(AG-G)CR, N(GTT)CR, N(GTT)KL, and N(GTT)OL tDNAs have been described (24). The V(AAC)MR2 tDNA is carried by the previously described pY7 plasmid (25). The Saccharomyces cerevisiae tDNA L(CAA)LR2 was PCR-amplified from yeast genomic DNA (strain S288C) by using the high fidelity Deep Vent DNA polymerase (New England Biolabs) and the oligonucleotide primers L_fw (5Ј-GCAAATA-AACGTTGGAAATTGGC) and L_rev (5Ј-GGTAGGAAAAAAAGATGCT-GCAC). The tDNAs S(AGA)EL and M(CAT)E were amplified by using the Pfu Turbo DNA polymerase (Stratagene) and the following pairs of oligonucleotide primers: S_fw (5Ј-GGGAACTCTCAGAACGGGGG) and S_rev (5Ј-CGAGTTTTTACCCATTTAGGAAAAG); M_fw (5Ј-GCGA-CATGGAGAGATAGAGATAG), and M_rev (5Ј-CGCAGAAGTTAAAGC-TTATAGTC). The 3Ј-mutant forms of N(GTT)CR and L(CAA)LR2 were obtained by PCR using the cloned tDNAs as templates and the appropriate mutagenic reverse primers. All of the amplification products were inserted into pUC-derived plasmid vectors and sequence-verified by dideoxy chain termination sequencing. The SCR1⌬_T7 construct is the previously described 3Ј⌬ϩ90 variant of SCR1 (26). For in vivo analysis, this SCR1 gene variant was subcloned into the YEp352 shuttle vector. The SCR1⌬_3ЈL template was obtained using the cloned wild type SCR1 gene as a template, the oligonucleotide 5Ј-TGATCAACT-TAGCCAGGACATCC as a forward primer, and the oligonucleotide 5Ј-AAAAAAAAAACTTCAAAATTTTCGGAAGGCAAAAACGTGCAAT-CCGG as a reverse primer. The amplified fragment, corresponding to the fusion of the first 90 bp of the SCR1 gene with the T 5 terminator plus 20 bp of 3Ј-flank of L(CAA)LR2 followed by a T 10 terminator, was cloned into the YEp352 vector. The 5ЈU6-L(CAA)LR2 and 5ЈU6-L_3ЈN templates were created by PCR using the corresponding cloned constructs as templates and the forward primer 5Ј-TTTCGGCTACTATA-AATAAATGTTTTTTTCGGAGCTCTTACAACAAATAAGTGGTTGT-TTGG, carrying the 5Ј-flanking sequence of the SNR6 gene of S. cerevisiae immediately followed by a SacI site (underlined). The amplified fragments were inserted into a modified pNEB193 plasmid whose polylinker SacI site was disrupted. The 5ЈU6-SCR1⌬_3ЈL construct was generated by PCR using the oligonucleotide 5Ј-TTTCGGCTACTATAAATAAATGTTTTTTTCGCAACTAGCTAGG-CTGTAATGGCTTTCTG, carrying the 5Ј sequence of the SNR6 gene, as a forward primer. In a different construct, the SacI restriction site was placed 5 bp upstream of SCR1⌬_3ЈL; the SCR1⌬_3ЈL variants were both inserted into the modified pNEB193. The SNR52 transcription unit was PCR-amplified from yeast genomic DNA (strain S288C) by using Pfu Turbo DNA polymerase and the following oligonucleotide primers: SNR52_fw (5Ј-TGATCAACTTAGCCAGGACATCC), SNR52_rev (5Ј-GTTCTAAGTATTCTCATTTTATCC). The SNR52_mutT mutant was constructed by recombinant PCR (27). Two overlapping PCR primary products were generated using the SNR52_fw oligonucleotide in combination with the mutT_rev primer (5Ј-GTAGGGTGTGAACGAAT-GCGCACC) and the SNR52_rev oligonucleotide in combination with the mutT_fw primer (5Ј-GGTGCGCATTCGTTCACACCCTAC). Mutated positions in mutT_rev and mutT_fw are underlined. After gel purification, primary amplification products were mixed and used as templates in a subsequent amplification reaction, employing SNR52_fw and SNR52_rev as "outside" primers, which yielded the desired fulllength secondary product. The 5ЈU6_SNR52 variant of SNR52, carrying the 5Ј-flanking sequence of SNR6, was generated by PCR using cloned wild type SNR52 as a template and the oligonucleotide 5Ј-TTTCGGC-TACTATAAATAAATGTTTTTTTCGCAACTA as a forward primer. All of the SNR52 variants were cloned into the pNEB193 plasmid vector.
In Vitro Transcription Analyses-In vitro transcription reaction conditions and procedures for RNA purification and analysis were essentially as described (13,28). Reactions contained 40 fmol of plasmidborne class III template and the following amounts of transcription proteins: 150 ng of TFIIIC purified up to the DEAE-Sephadex A-25 step (29); 40 ng of pure, recombinant TATA box-binding protein, and 80 ng of recombinant Brf1, both purified from overexpressing Escherichia coli cells (29); 0.5 g of BЉ fraction partially purified, up to the Bio-Rex 70 chromatography step, from chromatin pellets generated during yeast nuclear extract preparation (30). The last three components reconstitute a highly active TFIIIB factor. Unless otherwise indicated, the ATP, CTP, and GTP concentration was 500 M, and UTP concentration was 250 M. Transcripts were radiolabeled using [␣-32 P]UTP. Template DNA was first incubated with TFIIIC and TFIIIB for 20 min at 20°C, and then Pol III (10 ng) and NTPs were added, and transcription was allowed to proceed for 20 min at 20°C. In Fig. 7, A and B, transcription complexes were assembled either in the presence or in the absence of TFIIIC, using increased amounts of the different components of TFIIIB: 400 ng of TATA box-binding protein, 240 ng of Brf1 and either 0.7 g of BЉ fraction (Fig. 8B), or 20 ng of recombinant yeast Bdp1 (13) (Fig. 7A). For the factorless transcription experiment in Fig. 7B, 3Ј-overhanged restriction fragments were obtained by digesting the above described SacI site-containing templates with SacI, cutting 5 bp upstream of the natural transcription start site, and either BamHI (for L(CAA)LR2) or NdeI (for L_3ЈN and SCR1⌬_3ЈL), cutting downstream of the cloned insert in the plasmid polylinker. Restriction fragments were gel-purified, treated once with phenol/chloroform, and ethanol-precipitated prior to use in transcription assays. General transcription conditions were the same as in factor-directed transcription. The 3Ј-overhanged fragment (30 ng) was first incubated with 20 ng of purified Pol III and 400 M CpU dinucleotide primer in transcription buffer containing 160 g/ml bovine serum albumin for 15 min at 20°C in a total volume of 25 l. NTPs and [␣-32 P]UTP (10 Ci) were then added, and transcription was allowed to proceed for 30 min at 20°C. In all transcription experiments, radioactive transcripts were quantified by phosphorimaging. For termination efficiency measurements, the signals corresponding to RNAs of different length were normalized for the number of incorporated U residues.
In Vivo RNA Analysis-Yeast cells (YPH500 strain) were transformed with the different YEp352-SCR1⌬ constructs by the lithium acetate procedure (31), and the resulting transformants were selected for uracil auxotrophy. Total RNA was prepared according to a previously described procedure (32). RNA samples (10 g) were electrophoresed on 6% polyacrylamide, 7 M urea gels and transferred to a Gene-Screen Plus membrane (PerkinElmer Life Sciences), which was then probed with a 5Ј-labeled oligonucleotide (5Ј-CTGGCCGAGGAA-CAAATCCTTCCTCGCGGC) complementary to the coding region of SCR1 between positions ϩ43 and ϩ73. Hybridization was carried out overnight at 28°C in 5ϫ SSC solution, 5ϫ Denhardt's solution, 0.1 mg/ml denatured salmon sperm DNA, 0.5% (w/v) SDS, followed by one short washing in 2ϫ SSC solution containing 0.1% SDS and two short washings in 1ϫ SSC solution containing 0,1% SDS. Hybridization products were visualized and quantified by phosphorimaging.

Oligo(dT) Termination Signals in Different Eukaryotic
Genomes-The flanking regions of eukaryotic tRNA genes present conserved sequence patterns that diverge considerably in going from yeast to mammalian genomes (24). As a further element of divergence, yeast and human RNA polymerase III enzymes respond differently to oligo(dT) terminators (3). To gain insight into Pol III termination signals on a genome-wide scale, we thus decided to analyze and compare the tDNA transcription terminators in two yeast (S. cerevisiae and S. pombe) and two mammalian (M. musculus and H. sapiens) genomes. Using the complete tRNA gene inventories from these genomes, we initially searched for the presence of runs of 4 or more T residues within the first 100 bp of the 3Ј-flanking region, because T 4 is the shortest known oligo(dT) tract causing Pol III termination (33). In S. cerevisiae and S. pombe, T 4 was never found as the sole potential termination signal: a run of 5 or more T residues was always present, and it always started within the first 40 bp following the end of the mature tRNA coding sequence. 2 This was taken as an indication that runs of less than 5 T residues are not used as Pol III terminators in yeast and that Pol III termination signals appear within the first 40 bp of 3Ј-flanking sequence. In contrast, many mammalian tDNAs had a run of 4 T residues as the sole potential termination signal in the same region, in agreement with the reported ability of human Pol III to terminate efficiently at T 4 , and with the observation that yeast Pol III requires longer T stretches than vertebrate Pol III in order to terminate efficiently (3,33). Again, most termination signals occurred within 40 bp of the 3Ј-flanking region. Therefore, we considered T 5 and T 4 as the minimal termination signals for yeast and mammalian tDNA transcription, respectively. By limiting further analysis to the first 40 bp of 3Ј-flanking sequence, we then selected, from each tDNA set, the tDNAs having a single T run within that region. Such "single terminator" tDNAs represented 56, 47, 69, and 67% of all tDNAs in S. cerevisiae, S. pombe, H. sapiens, and M. musculus, respectively. We then calculated, for the subsets of single terminator tDNAs from each genome, the frequency distribution of T run length. The aim of such analysis was to establish whether the T run length in the different tDNA sets correlates with lineage-specific differences in termination signal recognition by Pol III. The tDNAs with more than one termination signal within the first 40 bp of 3Ј-flank were excluded from analysis, because the presence of a potential "back up" terminator could have reduced the selective pressure on the upstream terminator. The results of this analysis are shown in Fig. 1. In S. cerevisiae and S. pombe, the most frequent T run lengths are 6 and 7, and (as already mentioned) there is no single terminator shorter than T 5 . Remarkably, the tDNA terminators in fission yeast tend to be ϳ1 nt shorter than in S. cerevisiae, in agreement with a previous analysis showing 2 The only exceptions were represented by dimeric tDNAs, in which the upstream tDNA unit has no termination signal (45,58). that S. cerevisiae Pol III requires slightly longer T runs than the fission yeast enzyme in order to terminate efficiently (3).
The T-run length distribution in mammals is completely different: T 4 is the most frequent tDNA terminator, and terminators longer than T 5 are rare. Again, this finding is in agreement with the reported ability of human Pol III (and, more generally, of vertebrate Pols III) to terminate efficiently at T 4 sequences (1,3,18,33).
Variable Strength of the T 5 Terminator in S. cerevisiae tDNAs-Of the 148 "single terminator" tDNAs of S. cerevisiae, 22 bear a simple run of 5 Ts as a termination signal. For such tDNAs, the T 5 sequence is thus expected to be endowed with the ability to induce efficient Pol III termination. On the other hand, previous studies have shown that 5 consecutive T residues in the 3Ј-flanking sequence of yeast tDNAs are only partially effective in inducing Pol III transcription termination (3,4). To better define the termination properties of the T 5 sequence, we initially analyzed transcription termination on eight yeast tRNA genes all displaying T 5 as the first termination signal encountered in the 3Ј-flanking region. These genes are listed in Fig. 2A, together with the features of their 3Јflanks. Two of them (numbers 1 and 2) are classified as "single terminator" tDNAs. In the other tRNA genes, T 5 is followed by a back up terminator located 3-20 bp downstream. The presence of a downstream T run allows to easily monitor polymerases reading through the T 5 terminator by their production of longer transcripts terminated at the downstream T cluster. This is also true in the case of tDNA 1 (L(CAA)LR2), since a back-up T run is present starting at 66 bp downstream of T 5 . In the case of P(AGG)CR, where no back-up T run is present, the tDNA-containing plasmid was linearized 121 bp downstream of T 5 , so as to be able to reveal reading-through polymerases by their production of longer, run-off transcripts. All of these templates were used to program in vitro transcription reactions in a reconstituted system from S. cerevisiae (see "Experimental Procedures"). Transcription was performed in the presence of either a low UTP concentration (25 M; Fig. 2B, left panel) or a 10-fold higher UTP concentration (Fig. 2B, right panel) closer to the in vivo UTP concentration, estimated to be in the millimolar range (34,35). Fig. 2B shows that the strength of T 5 as a terminator varies considerably in the different contexts and that T 5 read-through by Pol III is in some cases strongly favored at the higher UTP concentration (cf., for example, lanes 3 and 4 of the left panel with the same lanes of the right panel), in other words under conditions that increase the polymerization rate and thus decrease the dwell time of the enzyme at T 5 . For six of the eight genes, the T 5 element acted as a strong terminator (lanes 1, 2, and 5-8). In contrast, for two tDNAs (lanes 3 and 4), a large fraction of polymerases read through the T 5 element and terminated at the back up terminator. The two T 5 elements behaving as weak terminators are both immediately followed by the sequence CT and, a few base pairs downstream, by a longer T run acting as a back up terminator (tDNA 3 and 4 in Fig. 2A). In contrast, the six T 5 elements acting as strong terminators are all immediately followed by A or G, with the exception of M(CAT)E (tDNA 5). The T 5 element of M(CAT)E is immediately followed by C and is characterized by a slightly lower strength as a terminator (cf. lane 5 with lanes 1, 2, 6, 7, and 8 in the right panel of Fig. 2B). It is interesting to note that T 5  can reasonably be attributed to the downstream sequence context.
Influence of the Downstream Sequence on Pol III Termination at T 5 -To investigate in more detail the downstream sequence effect on T 5 terminator strength, we focused on two of the tDNAs tested in the previous experiment: L(CAA)LR2 (tDNA 1), whose T 5 is one of the strongest terminators, and N(GTT)CR (tDNA 3), whose T 5 is the weakest terminator observed in Fig.  2 and is flanked by a back up terminator 7 bp downstream. We initially constructed and tested two hybrid templates (Fig. 3A): the L_3ЈN template, in which the sequence downstream of the L(CAA)LR2 T 5 terminator was replaced by the CT 4 CT 13 sequence naturally present downstream of T 5 in N(GTT)CR and the N_3ЈL template, in which the sequence downstream of the N(GTT)CR T 5 element was replaced by 20 bp of the sequence naturally present downstream of the L(CAA)LR2 T 5 terminator, followed by an artificially placed T 10 back-up terminator. In the L_3ЈN context, the T 5 terminator of L(CAA)LR2 turned out to be remarkably weakened, displaying (at 250 M UTP) a termination efficiency of 55% as compared with 98% in the case of the wild type L(CAA)LR2 gene (Fig. 3B, cf. lanes 1 and 2). On the other hand, when N(GTT)CR with its T 5 element was fused to the 3Ј-flanking sequence derived from L(CAA)LR2, the strength of T 5 as a terminator was greatly increased (Fig. 3B, cf. lanes 4 and 5). There are two possible explanations for the above observations. The T 5 element might be intrinsically weak as a terminator, and the particular downstream context in L(CAA)LR2 would potentiate its termination proficiency; in this case, T 5 weakening in L_3ЈN would be simply due to the removal, and T 5 strengthening in N_3ЈL to the insertion, of a terminator-strengthening sequence. Alternatively, the CT 4 CT 13 sequence naturally present 3Ј to the T 5 element in N(GTT)CR might directly act as (or contain) a terminatorweakening element. According to the second hypothesis, the replacement of the sequence downstream of T 5 in N(GTT)CR with an unrelated sequence should increase the T 5 strength as a terminator, whereas the replacement of the sequence down-stream of T 5 in L(CAA)LR2 with an unrelated sequence should not decrease the T 5 terminator efficiency. In the experiment in Fig. 3B, the termination properties of the unmodified L(CAA)LR2 and of the L_3ЈN construct were compared with those of a L(CAA)LR2 derivative (L_⌬3Ј) in which the 3Ј-flanking sequence downstream of T 5 was replaced by a plasmid sequence (lane 3). Similarly, the termination properties of unmodified N(GTT)CR and of the N_3ЈL construct were compared with those of a N(GTT)CR derivative (N_⌬3Ј) in which the T 5 3Ј-flank was replaced by the same plasmid sequence as in L_⌬3Ј (lane 6). Termination at T 5 appeared to be unaffected in the L_⌬3Ј construct with respect to unmodified L(CAA)LR2 (cf. lanes 1 and 3), whereas the strength of the N(GTT)CR T 5 element was greatly increased in the N_⌬3Ј construct (cf. lanes 4 and 6). (Since no rescue terminator is present downstream of T 5 in the L_⌬3Ј and N_⌬3Ј templates, they were linearized with BamHI at a position 16 bp downstream of the T 5 sequence). The CT 4 CT 13 sequence thus behaves as (or appears to contain) a terminator-weakening element. This conclusion is further supported by the observation that T 5 is weak in only two of the eight tDNAs tested in the experiment in Fig. 2 (N(GTT)CR and S(AGA)EL) and that in both cases T 5 is followed by similar CT-containing sequences (CT 4 CT 13 in the case of N(GTT)CR, CTCT 8 C in the case of S(AGA)EL).
Sequence and Positional Requirements for T 5 Terminator Weakening-To better understand the T 5 terminator weakening effect, we concentrated on the L_3ЈN chimeric template, whose downstream weakening sequence, CT 4 CT 13 , was subjected to mutational analysis. In particular, mutants were constructed in which the CT 4 C sequence was variously replaced, whereas the downstream T 13 tract was left unvaried (or shortened to T 10 ) and used as a back-up terminator. All of the mutant constructs, schematically illustrated in Fig. 4A, were tested in an in vitro reconstituted system in at least three independent experiments. The results are shown in Fig. 4B. Pol III transcriptional read-through at T 5 was first calculated on the basis of the ratio between T 5 -terminated and T 13 /T 10 -ter- minated transcripts. In each experiment, the extent of T 5 terminator read-through observed with the L_3ЈN template (lane 2) was arbitrarily set to 100, and the effect of the mutations was evaluated as relative T 5 read-through with respect to L_3ЈN. When the first 3 nucleotides downstream of T 5 (CTT) were changed to GCC, so as to render the 6 base pairs downstream of T 5 identical to the sequence found in the wild type L(CAA)LR2 tDNA, the weakening effect was strongly dimin- Only the sequence 3Ј to the T 5 terminator (in boldface type and underlined) is reported. The back-up terminator is in boldface type. Templates 1-3 are the same as in Fig. 3. B, the lower part shows the results of in vitro transcription assays of the templates listed in A (identified by the numbers above the lanes). The migration positions of the different tDNA transcription products, terminated at the first or the second T-run, are indicated on the right, together with the position of a 415-nt-long DNA used as a recovery marker (RM). Reported in the upper part is a bar plot deriving from quantification of three independent experiments. Pol III transcriptional read-through at the first terminator was calculated on the basis of the ratio between tDNA transcripts terminated at the first and second T-run. In each experiment, the transcriptional read-through observed with the L_3ЈN template (template 2) was set to 100, and the effect of the mutations was evaluated as relative T 5 read-through with respect to L_3ЈN and reported on the y axis, with error bars indicating the S.E. ished (ϳ5-fold) with respect to L_3ЈN (cf. lanes 2 and 4). Since the CTT to GCC substitution, which increases the predicted stability of the DNA double helix immediately downstream of T 5 , strongly affected its weakening potential, we wondered whether replacing the downstream sequence with other sequences characterized by a similar (or even lower) DNA duplex stability would preserve (or increase) the weakening effect, and thus T 5 read-through. To address this point, the T 4 stretch in CT 4 C was replaced by TATA, ATAT, or A 4 . As shown in Fig. 4B, lanes 6 -8, and in the bar plot above the gel, such substitutions led to a partial loss of the weakening effect: a 40 -50% loss in the case of A 4 and ATAT and only a 20% loss in the case of TATA. The prediction of DNA duplex stability on the basis on nearest neighbor thermodynamics gives A 4 ϭ T 4 Ͼ ATAT Ͼ TATA (36). Therefore, among the three substitutions, the one causing the strongest read-through (TATA) is also the one displaying the lesser DNA duplex stability. However, despite its higher predicted stability, the original T 4 sequence causes more read-through than TATA. We conclude that a low DNA duplex stability in the region immediately downstream of T 5 is not sufficient for full T 5 terminator weakening, even if it somehow contributes to the weakening effect. This conclusion is also supported by the comparison of the termination properties of tDNAs 3 and 7 in Fig. 2. The T 5 terminator of tDNA 7 is much stronger than the T 5 terminator of tDNA 3, yet the predicted stability of the immediately downstream 6-bp sequence (CT 4 C for tDNA 3, GAATAC for tDNA 7) is very similar (36). With another series of constructs (templates 5, 9, and 10 in Fig. 4) we then analyzed the positional requirements for T 5 terminator weakening. In all of these constructs, the T 13 element was replaced by T 10 , a sequence that works as efficiently as T 13 as a downstream terminator and does not interfere with the T 5 weakening effect (data not shown). In construct 5, the distance between the T 4 and T 10 stretches, which are separated by only 1 bp in the L_3ЈN template (lane 2), was increased to 11 bp. As shown in Fig. 4B, T 5 terminator weakening in this context was very similar to that observed with L_3ЈN (cf. lanes 2 and 5). The CT 4 C sequence thus appears to act as an autonomous, termination-weakening element. In contrast, the weakening effect was found to be compromised in constructs in which the distance between T 5 and the CT 4 C sequence was increased by 3 bp (construct 10, lane 10) or even by only 1 bp (construct 9, lane 9). The terminator weakening effect exerted by the CT 4 C sequence thus seems to depend in a critical way on its distance from the T 5 terminator. Having restricted the terminator-weakening function to the CT 4 C sequence, we constructed other mutants (constructs 11-15 in Fig. 4A) in order to identify the minimal sequence requirements for the terminator weakening effect. As shown in lane 13, the C residue separating the T 4 stretch from the downstream T 10 sequence could be mutated without any loss of terminator weakening (on the contrary, a 50% increase in T 5 read-through with respect to L_3ЈN was reproducibly observed with this mutant; cf. lanes 2 and 13). With templates 14 and 15, in which the length of the T 4 stretch was progressively reduced, the weakening effect was maintained (lanes 14 and 15). The mutation of the C residue, immediately following the T 5 element, to A reproducibly reduced the read-through efficiency by ϳ65% with respect to L_3ЈN (cf. lanes 2 and 11). The mutation of the same residue to G produced a more modest, but still significant effect, reducing T 5 read-through by ϳ40% with respect to L_3ЈN (cf. lanes 2 and 12). A C residue immediately downstream of T 5 is thus required for full weakening efficiency. The results of this mutational analysis, together with the data in Fig. 2, suggest that a CT dinucleotide immediately downstream of a T 5 terminator represents a minimal sequence element capable of producing a significant ter-minator weakening effect. That CT alone is sufficient to induce read-through is especially evident from the termination properties of tDNA 4 in Fig. 2 and template 8 in Fig. 4. This weakening element, however, appears to be only effective on the T 5 terminator. When the termination signal length was increased to T 6 , the weakening effect was lost (Fig. 4B, lane 16).
Mechanism of T 5 Terminator Weakening by Downstream Sequence-The oligo(dT) tracts acting as Pol III terminators have the ability to induce both RNA polymerase pausing and transcript release. These two features are functionally distinguishable and are characterized by different sequence requirements (6,15). Weakening of the T 5 terminator by a CT sequence placed immediately downstream may involve an impairment of both (or only one) of these abilities. To gain insight into the mechanism of T 5 terminator weakening, we first analyzed termination on the L_3ЈN template (construct 2 in Fig. 4A) and on the L_3ЈGT 4 template (template 12 of Fig. 4A, in which the C immediately downstream of T 5 is replaced by G) in the presence of a standard (500 M) and a 10-fold lower concentration of CTP. Lowering CTP concentration is expected to increase the dwell time of Pol III across the T 5 sequence in L_3ЈN, but not in L_3ЈGT 4 , by reducing the rate of incorporation of the first nucleotide downstream of T 5 . In the same way, lowering the GTP concentration should increase the dwell time of Pol III onto the T 5 element in L_3ЈGT 4 , but not in L_3ЈN. As shown in Fig. 5, in the presence of 50 M CTP (lane 5) the T 5 terminator in L_3ЈN turned out to be significantly stronger than with 500 M CTP (lane 2), with a ϳ3-fold reduction in the fraction of reading-through polymerases (from 45% in lane 2 to 14% in lane 5). Lowering the GTP concentration had instead no effect on T 5 read-through with this template (cf. lanes 2 and 8). With the L_3ЈGT 4 template, lowering the CTP concentration did not influence the extent of T 5 read-through, whereas the fraction of reading-through polymerases was slightly (but reproducibly) decreased at low GTP concentration (from 22% in lane 3 to 17% in lane 9). T 5 terminator weakening can thus be specifically counteracted by reaction conditions that favor polymerase pausing on T 5 .
Effect of the Upstream Sequence Context on T 5 Terminator Efficiency-Altogether the above data show that termination efficiency at T 5 can be strongly influenced by the downstream sequence context, in a manner that is independent from the specific tRNA coding sequence preceding the T 5 element. To further test for possible effects of the upstream flanking sequence on T 5 -dependent termination, we generated a construct (SCR1⌬_3ЈL) in which the L(CAA)LR2 T 5 terminator, plus 20 bp of its natural 3Ј-flanking sequence followed by T 10 , were fused to the first 90 bp of the SCR1 gene, coding for the S. cerevisiae 7SL RNA. Apart from the presence of A-and B-blocks allowing for efficient in vitro transcription by the Pol III system (26), this portion of the SCR1 gene is essentially unrelated to tRNA coding sequences. As shown in Fig.  6A, lane 1, the insertion of a T 7 terminator at position ϩ90 of SCR1 results in a template (SCR1⌬_T 7 ) whose transcription is efficiently terminated at T 7 , thus generating 90-nt-long transcripts; no read-through transcription products were observed when this template was linearized 25 bp downstream of T 7 (data not shown). Surprisingly, the L(CAA)LR2-derived T 5 element plus 3Ј-flank, which produced a 97% termination efficiency in the case of the original template (see, for example, Fig. 2, lane 1), behaved as a poor terminator in the SCR1⌬_3ЈL construct, allowing ϳ40% of the polymerases to read through and terminate at the downstream T 10 terminator in the presence of 250 M UTP (Fig. 6A, lane 2). Readthrough was significantly less in the presence of 25 M UTP (lane 4) and almost absent with 2.5 M UTP (lane 6), thus suggesting that the SCR1-derived upstream sequence reduces termination at T 5 by affecting Pol III pausing. Transcription termination with the SCR1⌬_3ЈL construct could also be monitored in vivo, because the 20-bp distance between the T 5 element and the downstream T 10 terminator allows for the production of transcripts that differ from each other in size significantly and that are characterized by comparable in vivo stabilities. 3 The SCR1⌬ T 7 and SCR1⌬_3ЈL templates were inserted into the high copy number YEp352 vector and transformed into the YPH500 S. cerevisiae strain. The RNAs derived from logarithmically growing cells were then subjected to Northern blot analysis with an SCR1-specific antisense oligonucleotide. As shown in Fig. 6B (lane 2) 55% of the plasmid-encoded, SCR1-derived transcripts were terminated at T 5 , whereas 45% were 20 nt longer and derived from T 5 read-through and termination at the downstream T 10 sequence. By comparison, the T 7 terminator in the SCR1⌬_T 7 construct appeared to be fully efficient in vivo (lane 1).
Distinguishing between Weak and Strong T 5 Terminator Sequences Is an Intrinsic Property of Pol III-The observed sequence context effects on Pol III termination at T 5 might in principle be mediated by TFIIIC, a transcription factor that usually covers a large part of the transcription unit, extending up to the terminator region of tRNA genes (37,38). TFIIICassociated activities have previously been proposed to influence Pol III termination in humans (39), and yeast TFIIIC has recently been shown to facilitate transcription reinitiation by Pol III on long transcription units (13). The role of TFIIIC is thus not restricted to preinitiation complex assembly. To test whether TFIIIC influences termination at T 5 in different sequence contexts, some of the templates in which the T 5 terminator is weakened were modified so as to allow for their TFIIIC-independent transcription. To this end, the transcribed portions of L(CAA)LR2, L_3ЈN and SCR1⌬_3ЈL were fused to the 5Ј-flanking region of the S. cerevisiae SNR6 gene, which is known to support TATA box-mediated TFIIIB assembly and transcription initiation in the absence of TFIIIC (13,40). As shown in Fig. 7A, a pronounced weakening of T 5 by both the 3ЈN downstream context and the SCR1⌬ upstream context was observed in the absence of TFIIIC (cf. lanes 5 and 6 with lane  4), thus demonstrating that the weakening effects do not depend on this factor. A comparison of the termination efficiencies in the presence (lanes 1-3) or absence (lanes 4 -6) of TFIIIC rather suggests that TFIIIC somehow facilitated T 5 termination signal recognition by Pol III, an effect that is especially evident with the 5ЈU6-L_3ЈN template (cf. lanes 2 and 5 in Fig.  7A). To test in a more direct way whether distinguishing between weak and strong T 5 termination signals is an intrinsic property of Pol III, we exploited the reported ability of Pol III to initiate transcription on linear templates containing a 3Јoverhanging end (41). Linear, 3Ј-overhanged versions of L(CAA)LR2, L_3ЈN and SCR1⌬_3ЈL templates were prepared as described under "Experimental Procedures" and transcribed in vitro with purified RNA polymerase III in the absence of any transcription factor. As shown in Fig. 7B, and in agreement with a previous study (41), purified Pol III specifically and efficiently transcribed these templates. Transcription initiation was made to occur at the 3Ј-overhanging end generated by SacI cleavage, at a position 5 bp upstream of the natural  1 and 2), 25 M UTP (lanes 3 and 4), or 2.5 M UTP (lanes 5 and 6). The migration positions of transcripts terminated at the first or second T-run of the SCR1-derived templates are indicated on the right. B, in vivo expression profiles of the SCR1⌬_T 7 and SCR1⌬_3ЈL templates. Total RNA was extracted from yeast cells (YPH500 strain) transformed with YEp352 carrying the template indicated above the lanes. Fractionated RNA samples (10 g each) were probed with a radiolabeled oligonucleotide annealing within the first 90 bp of the SCR1 RNA product. The migration positions of the transcripts derived from the chromosome-encoded SCR1 gene and from the two plasmid-borne SCR1-derived templates (1 st T and 2 nd T) are indicated on the right. transcription start site. Pol III terminated transcription at the same terminator region present in the original templates, in which a T 5 element of variable strength is followed by a back-up terminator (placed 67, 7, and 21 bp downstream of T 5 in L(CAA)LR2, L_3ЈN, and SCR1⌬_3ЈL, respectively). A small percentage of transcription complexes (10% at most) failed to recognize both the T 5 and back-up terminators, thus giving rise to run-off products (the DNA fragments used as templates ended 99, 238, and 249 bp downstream of T 5 in L(CAA)LR2, L_3ЈN, and SCR1⌬_3ЈL, respectively). Such a small amount of run-off transcripts probably derives from transcription complexes unable to displace the RNA from RNA/DNA hybrids, a reported artifact of initiating transcription at 3Ј-overhanged or dC-tailed templates (6,41). Run-off transcripts were thus not considered for quantification purposes. The T 5 signal in the wild type context of L(CAA)LR2 acted as a rather strong terminator, both at 25 M and at 250 M UTP, with only 19% of transcription complexes reading through and terminating at the T 7 back up terminator (Fig. 7B, lanes 1 and 4). In contrast, the T 5 termination signal was remarkably weakened in the L_3ЈN and SCR1⌬_3ЈL contexts, both at 25 M UTP (lanes 2 and 3) and at 250 M UTP (lanes 5 and 6). The weakening effect was dramatic in the case of SCR1⌬_3ЈL, for which no transcripts terminated at T 5 could be detected at 250 M UTP (lane 6). Weakening of T 5 by both the upstream and the downstream sequence contexts thus clearly reflect intrinsic recognition properties of Pol III. By comparing the extents of T 5 weakening observed at 250 M UTP in the absence of transcription factors with those observed under the same conditions with a complete transcription system (e.g. see Fig. 5, lane 2, and Fig. 6, lane 2), we note that the weakening effect is significantly enhanced in the factor-free system. Such a T 5 read-through enhancement can at least in part be attributed to the absence of TFIIIC (see above in reference to Fig. 7A).
Antitermination within a Natural Pol III Transcription Unit-Recent studies of the genome-wide localization of the Pol III transcription machinery have identified the snoRNA encoding SNR52 gene as a new class III gene (42)(43)(44). Northern blot and primer extension analyses suggested that the primary transcript of SNR52 is a ϳ250-nt precursor RNA from which a leader sequence of 160 nt is cleaved to generate the ϳ90-nt mature snoRNA product, and sequence analysis revealed the presence of putative A-and B-block elements within the leader sequence (42,43). An unusual feature of this sequence, however, is that a run of 6 T residues is present within the transcribed region, a few bp downstream of the putative A-block. The T 6 sequence is a highly efficient termination signal for yeast Pol III, as illustrated by the results shown in Fig. 4B (lane 16) and by the results of in vitro transcription experiments with several tDNAs bearing a T 6 terminator (data not shown). As shown in Fig. 8 (lane 1), in vitro transcription of SNR52, in the presence of 25 M UTP, produced two main transcripts. The longer transcript was identified as the SNR52 primary transcript on the basis of its size (250 nt). The shorter RNA had the size expected for a transcript terminated at the intragenic T 6 element. Indeed, transcription of a mutant SNR52 template, in which T 6 was changed to TTCGTT, produced higher levels of the longer transcript, whereas the shorter transcript disappeared (cf. lanes 1 and 2). Quantification of the transcripts in lane 1, corrected for the number of incorporated U residues, led us to estimate that as much as 70% of the polymerases read through the T 6 element in SNR52. T 6 thus behaves as an unusually weak terminator in the SNR52 context, even under conditions (low UTP concentration) that favor termination at T stretches. The inactive T 6 element of SNR52 is located just downstream of a putative A-block promoter element within the transcribed region. Since TFIIIC is expected to contact such region during transcription, this factor might be required in order for Pol III to read through the internal T 6 . To test this possibility, the SNR52 transcription unit was fused with the 5Ј-flanking region of the S. cerevisiae SNR6 gene to allow for TFIIIC-independent transcription of the resulting chimeric template. As shown in Fig. 8B, the presence or absence of TFIIIC on SNR52 had no influence on the termination behavior of Pol III, characterized by highly inef-  2 and 4). DISCUSSION This work demonstrates that the recognition of oligo(dT) termination signals by S. cerevisiae RNA polymerase III can be strongly influenced by the sequence context in which the T stretch is embedded. Such influence becomes evident with the shortest termination signals encountered in yeast class III genes, namely T 5 and T 6 , that together characterize ϳ40% of the yeast tDNAs possessing a single T run as a termination signal (see Fig. 1). We have been able to distinguish between two types of sequence context effect; one, that we have more extensively addressed, is due to the few base pairs immediately downstream of T 5 ; the other was observed when a non-tDNA sequence was fused upstream of a strong terminator region (composed of T 5 plus a particular 3Ј-flank) derived from L(CAA)LR2 tDNA. An extreme example of sequence context effect was observed with the SNR52 transcription unit, in which a T 6 element placed between the A-block and the B-block promoter elements was barely recognized as a terminator by Pol III.
The main conclusion emerging from the study of the downstream sequence effect is that T 5 , when associated with tRNA genes, is intrinsically strong as a terminator, because it induces highly efficient termination in most of the contexts analyzed; however, features of the downstream sequence exist that can significantly weaken the T 5 termination potential. A specific weakening element identified in this study seems to minimally consist of the CT dinucleotide placed immediately downstream of T 5 in the coding DNA strand. The tested natural tDNAs displaying such a feature were characterized by incomplete termination at T 5 . The 3Ј-flank of N(GTT)CR, one of these tDNAs, could be transplanted 3Ј to another tDNA, L(CAA)LR2, without losing its T 5 terminator weakening properties. By extensive mutagenesis, we were able to show that T 5 weakening absolutely requires a C residue immediately downstream of T 5 and is greatly favored by the presence of at least one T just after the C. With this respect, it is interesting to note that, in an early study of yeast Pol III termination, termination at T 5 was found to be incomplete, both in vitro and in vivo, for a template in which T 5 was followed by CT (4). T 5 terminator weakening might not be an exclusive property of the CT element, and other short weakening sequences may exist. However, neither natural tDNA analysis nor extensive mutagenesis of the N(GTT)CR-derived weakening element could reveal any other sequence with higher or similar weakening activity. By which mechanism does CT downstream of T 5 induce terminator read-through? Our data suggest that CT facilitates Pol III translocation by reducing its pause time at T 5 . Indeed, readthrough tends to be lesser under conditions (such as low UTP or CTP concentrations) that increase the dwell time of Pol III across the T 5 tract. We also tested the sensitivity of T 5 terminator weakening to variations in KCl concentration (50 -150 mM range) and Mg 2ϩ ion concentration (3-10 mM range), which have the potential to influence the RNA release step of intrinsic bacterial termination (34), but we did not observe any significant variation in termination efficiency at T 5 either in the L(CAA)LR2 or in the L_3ЈN contexts (data not shown). Possible mechanistic explanations of the effect of downstream DNA on translocation include the stability of the duplex that must be melted or NTP binding to a melted region of downstream DNA, which would facilitate translocation (46 -49). Our data tend to exclude a simple correlation between T 5 terminator weakening by downstream DNA and the ease of duplex melting. For example, template 6 in Fig. 4 displayed the same duplex stability as template 2 in the region immediately downstream of T 5 (CA 4 versus CT 4 ), yet the weakening effect was ϳ2-fold lower with CA 4 than with CT 4 . Also, the introduction, in template 9 of Fig.  4, of an A residue before CT 4 completely abolished the weakening effect of CT 4 without changing significantly the duplex stability of this region. Our data rather suggest that the CT element might increase the processivity of Pol III at the position immediately upstream, possibly by a mechanism involving the binding of the corresponding NTPs to an allosteric NTP binding site, as recently proposed to explain downstream DNA effects on elongation by both E. coli RNA polymerase (49) and yeast RNA polymerase II (48).
Having established that T 5 , followed by the sequence naturally present in the L(CAA)LR2 tDNA, behaves as a strong terminator for Pol III, even when fused 3Ј to a completely different tDNA (see Fig. 3), we were surprised to find that the same sequence becomes a much weaker terminator when fused 3Ј to an A-and B-block-containing segment of the SCR1 gene. Even more surprising was the finding that the T 6 sequence, generally producing highly efficient Pol III termination, was substantially inactive as a terminator in the context of the SNR52 gene, where it is naturally present between the A-block and B-block promoter elements preceding the snoRNA coding sequence. Although the sequence and positional requirements for both the SCR1 and the SNR52 effects were not investigated in detail, in both cases read-through was found to be favored at higher UTP concentrations, thus indicating that these effects involve an alteration of Pol III pausing. The results of experiments in which transcription initiation and termination were made to occur in the absence of TFIIIC further suggest that T 5 /T 6 read-through in these two contexts does not require TFIIIC. Rather, in the absence of TFIIIC, read-through tended to be more pronounced, an observation that might be explained by postulating that DNA-bound TFIIIC slightly slows RNA chain elongation, thus favoring termination at T 5 . DNA-bound TFIIIC has previously been shown to exert no significant effect on RNA chain elongation by Pol III in the sense direction of the SUP4 tRNA Tyr gene (41). It cannot be excluded, however, that in specific sequence contexts, TFIIIC contributes to Pol III pausing and termination. By in vitro transcription experiments conducted in the presence of purified Pol III and template DNA only, we showed that T 5 terminator weakening also occurs in the absence of any transcription factor and thus reflects intrinsic recognition properties of Pol III, a conclusion in agreement with that of an early study of termination by Xenopus Pol III (2). The T 5 weakening effect was especially evident when the strong L(CAA)LR2 terminator region was placed downstream of an A-and B-block-containing fragment of the SCR1 gene. In this case, T 5 read-through enhancement can only depend on particular features of the upstream DNA region. This region might influence the Pol III behavior at T 5 through polymerase-DNA interactions, but it might also play an active role in the termination process through the corresponding RNA product, as is well known for the stem-loop coding portion of intrinsic bacterial terminators (50,51).
The lack of termination at T 6 within the SNR52 transcription unit is of evident biological significance, since the snoRNA coding portion of SNR52 is entirely located downstream of the T 6 element. Curiously, such a sequence, which is potentially detrimental to SNR52 gene transcription, is also present in the genomes of at least two other Saccharomyces species (Saccharomyces paradoxus and Saccharomyces mikatae (52)), thus indicating a possible relevant role in SNR52 expression. In principle, the T 6 element might influence the accessibility and/or transcriptional activity of SNR52 in vivo, or it might play a role in pre-snoRNA processing. Less evident is the biological role of T 5 terminator weakening by short downstream sequences like CT in the case of tRNA genes. As a result of T 5 read-through, Pol III is expected to synthesize in vivo tRNA precursors with elongated 3Ј trailers, that might in principle stabilize the pre-tRNA from 3Ј exonucleolytic digestion. In organisms from yeast to humans, binding of La protein also protects the pre-tRNA 3Ј trailer from exonucleolytic digestion (reviewed in Ref. 53). La protein binds the UUU OH at the 3Ј end of nascent Pol III transcripts, and, by favoring the formation of correctly folded pre-tRNAs, it can affect pre-tRNA stability, processing, localization, and charging (53)(54)(55)(56). The 3Ј trailer extension caused by Pol III read-through might influence each of these processes either directly or by affecting La binding. In support of this view, it has recently been shown that an increase in 3Ј-terminal oligo(U) length of pre-tRNAs can influence association with La and tRNA maturation (57). The modulation of Pol III terminator readthrough might be of even higher significance in mammals. In the mouse and human genomes, T 4 is by far the most frequent termination signal for tRNA gene transcription, and in many cases, it is the sole potential termination signal encountered 3Ј to the tDNA. Since runs of 4 T residues are often present within the internal transcribed region of tRNA genes, there must be features of the sequence context capable of strongly affecting Pol III termination at T 4 . Such features, however, are not simply represented by conserved sequence elements flanking the T 4 terminator, as revealed by statistical analysis of such regions in mouse and human tDNA data sets. 4 Why did mammalian tDNA terminators evolve toward the shortest possible T runs? One possible answer is that, in the presence of weak T run terminators, more room is left for the modulation of termination by accessory factors, such as NF1, affecting transcription termination by human Pol III (39). Al-ternatively, the predominance of very short T runs downstream of mammalian tDNAs might reflect the importance of short 3Ј polyuridylate tails for the post-transcriptional fate of mammalian pre-tRNAs.