A Minimal Promoter for TFIIIC-dependent in Vitro Transcription of snoRNA and tRNA Genes by RNA Polymerase III*

The Saccharomyces cerevisiae SNR52 gene is unique among the snoRNA coding genes in being transcribed by RNA polymerase III. The primary transcript of SNR52 is a 250-nucleotide precursor RNA from which a long leader sequence is cleaved to generate the mature snR52 RNA. We found that the box A and box B sequence elements in the leader region are both required for the in vivo accumulation of the snoRNA. As expected box B, but not box A, was absolutely required for stable TFIIIC, yet in vitro. Surprisingly, however, the box B was found to be largely dispensable for in vitro transcription of SNR52, whereas the box A-mutated template effectively recruited TFIIIB; yet it was transcriptionally inactive. Even in the complete absence of box B and both upstream TATA-like and T-rich elements, the box A still directed efficient, TFIIIC-dependent transcription. Box B-independent transcription was also observed for two members of the tRNAAsn(GTT) gene family, but not for two tRNAPro(AGG) gene copies. Fully recombinant TFIIIC supported box B-independent transcription of both SNR52 and tRNAAsn genes, but only in the presence of TFIIIB reconstituted with a crude B″ fraction. Non-TFIIIB component(s) in this fraction were also required for transcription of wild-type SNR52. Transcription of the box B-less tRNAAsn genes was strongly influenced by their 5′-flanking regions, and it was stimulated by TBP and Brf1 proteins synergistically. The box A can thus be viewed as a core TFIIIC-interacting element that, assisted by upstream TFIIIB-DNA contacts, is sufficient to promote class III gene transcription.

RNA polymerase (Pol) 3 III synthesizes tRNA, 5 S rRNA, and a variety of other types of small nuclear and cytoplasmic RNAs. In general, the transcription of class III genes is under the control of internal control regions (ICR) characterized by discon-tinuous structures, with essential boxes separated by non-essential nucleotides (1). In the case of tRNA and 5 S rRNA genes, the ICRs are highly conserved and comprise the binding sites for the general transcription factor TFIIIC (box A and box B) and for the 5 S-specific factor TFIIIA (box C). Once assembled, TFIIIC recruits TFIIIB upstream of the transcription start site (TSS); TFIIIB in turn recruits Pol III for transcription initiation. The strong conservation of the ICRs likely reflects their dual function as both nucleation sites for transcription complex assembly and key determinants of tRNA and 5 S rRNA structure. An indication of the constraints imposed on ICR by their overlapping structural and functional roles comes from the variability of promoter organizations displayed by the minority of class III genes not coding for tRNA and 5 S rRNA. In some of these genes, the TFIIIC-interacting control regions (box A and box B) have been maintained within the transcribed region, and adapted to the structure of the small RNA without losing the transcriptional function. For example, in the Saccharomyces cerevisiae SCR1 gene, coding for the 7SL RNA, box A and box B are both intragenic and involved in transcription, yet their sequences display distinguishing features as compared with the consensus of the tDNA box A and box B (2). In the S. cerevisiae SNR6 gene, coding for the U6 snRNA, a non-canonical, yet transcriptionally relevant box A has been maintained within the transcribed region, whereas the box B is located downstream of the transcription termination site (3,4). Other class III genes entirely rely for transcription on upstream promoter elements, an organization that completely removes the need for the reciprocal adaptation of transcription control elements and RNA structure. One group of such genes, exemplified by the metazoan U6 snRNA and the human 7SK RNA genes, exploits upstream promoter elements similar to those of Pol II-transcribed genes (5). Other genes, such as RPR1, coding for the RNA component of S. cerevisiae RNase P (6), and the 7SL RNA genes of Trypanosomatidae (7,8), have adopted a promoter strategy that places a tRNA gene-like promoter, containing box A and box B, upstream of the mature 5Ј-end of the RNA. Such a strategy has recently been found to be employed for box C/D snoRNA expression in Arabidopsis thaliana and rice, where some snoRNAs are synthesized as tRNA-snoRNA precursors originated by Pol III transcription of tRNA-snoRNA dicistronic genes (9). Intriguingly, recent studies of the genome-wide localization of the Pol III transcription machinery in S. cerevisiae have identified the snoRNA-encoding SNR52 gene as a new class III gene, and sequence analysis has revealed the presence, upstream of the snoRNA coding sequence, of box A and box B elements that could together act as an external, tRNA gene-like promoter (10 -12).
TFIIIC can specifically associate to target genes characterized by very different promoter configurations. This is because of the remarkable adaptability of this DNA-binding protein, which is related to the complexity and modular organization of its molecular structure. In particular, TFIIIC can bind to, and promote transcription of, templates characterized by very different box A-box B distances, from 20 to several hundred bp (13)(14)(15). This property, together with the results of limited proteolysis (16) and electron microscopy studies (17), suggests that TFIIIC consists of two DNA binding modules separated by a flexible linker that can accommodate variously spaced A and B boxes. The TFIIIC-box B interaction gives the major contribution to the stability of the TFIIIC-DNA complex, while TFIIICbox A contacts promote TFIIIB assembly and precise transcription initiation (18). A completely different mode of DNA binding by TFIIIC is that involved in 5 S rRNA gene transcription. In this case, TFIIIC is recruited to the target gene essentially through protein-protein interaction with 5 S DNA-bound TFIIIA (19). Adding to the complexity of TFIIIC action, DNAbound TFIIIC has recently been shown to be involved in Pol III-facilitated recycling in the yeast system (15,20). In some cases, discrepancies have been observed between box A/box B requirements for class III gene transcription in vitro and in vivo. Such discrepancies could be attributed, at least in part, to the fact that suboptimally bound TFIIIC can efficiently recruit TFIIIB in vitro, but is impaired in its ability to counteract repressive chromatin assembly in vivo (2,21). In vitro, a defective interaction between TFIIIC and its split binding site can be tolerated and may result in only a moderate transcriptional impairment, provided that such interaction is stable enough to allow the formation of long lived (kinetically trapped) TFIIIB-DNA complexes (22,23). This mainly occurs for mutations of the box A that do not reduce dramatically the affinity of TFIIIC for the promoter (1). Though highly variable in sequence, the 5Ј-flanking region of A box-and B box-containing genes constitutes the binding site for TFIIIB, and may thus exert a strong influence on transcription efficiency (see Ref. 24 and references therein). In some eukaryotic genomes, the upstream regions of ICR-containing class III genes are characterized by the presence of a TATA box, starting at around Ϫ30, which strongly contributes to transcription efficiency as an essential component of the promoter (25,26). In some cases, the upstream TATA element can even allow for TFIIIC-independent in vitro transcription (27,28). Whatever the mode of TFIIIC assembly, this factor acts in concert with TFIIIB to promote transcription of both TATA-less and TATA-containing class III genes. On TATA-less tRNA genes, TFIIIC binding initiates a series of protein-protein interactions ultimately resulting in TFIIIB assembly onto upstream DNA, with a key interaction taking place between the Tfc4 subunit of TFIIIC and the Brf1 component of TFIIIB (29,30).
In this study, we have carried out an extensive in vitro an in vivo analysis of SNR52 promoter architecture. Unexpectedly, we found that box B is dispensable whereas box A is essential for in vitro transcription of SNR52. This finding prompted us to re-evaluate the contribution of TFIIIC-box B interaction to promoter strength and its actual requirement for tRNA gene transcription.

MATERIALS AND METHODS
Amplification and Cloning of DNA Templates-The S. cerevisiae SNR52 gene was PCR-amplified from yeast genomic DNA (strain S288C) using the high fidelity Pfu DNA polymerase (Promega) and gene-specific pairs of oligonucleotide primers: SNR52_fw 5Ј-CTTTGAAAAGATAATGTATGATTATGC and SNR52_rev 5Ј-GCGTTCCATACTGTCAGAGGT. The amplified fragment (415 bp) contained the SNR52 transcribed sequence (260 bp) plus 110 bp of 5Ј-flanking and 45 bp of 3Ј-flanking sequences. The SNR52 Adown, Bdown, Tdown mutant variants were obtained by recombinant PCR (31). Two overlapping PCR primary products were generated using the SNR52_fw oligonucleotide in combination with 5Ј-GCGCACCTTTAGGGCTAG-CCCAAGAAG and the SNR52_rev oligonucleotide in combination with the oligonucleotide 5Ј-CCTCCTGGGCTAGCCCT-AAAGGTGCGC (mutagenic bases 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 full-length secondary product. The same strategy was employed for the Bdown and the T down variants, using the following mutagenic oligonucleotides: 5Ј-GGAGAAGTTTCCAAC-GCCGAAACATGC and 5Ј-GCATGTTTCGGCGTTGGAAA-CTTCTCC for the Bdown mutation; 5Ј-GTAGGGTGTGAACG-AATGCGCACC and 5Ј-GGTGCGCATTCGTTCACACC-CTAC for the Tdown mutation. The resulting DNA fragments were cloned into the SmaI site of the pUC-derived plasmid pNEB193 (New England Biolabs). For yeast transformation, all the inserts were subcloned into the shuttle vector pFL45S (32). 5Ј-mutated forms of SNR52 were obtained through amplification using the following mutagenic oligonucleotides as forward primers: SNR52 5Ј⌬-58, 5Ј-TGATGTTTTCTTTCGAGTAT-ATAC; SNR52 5Ј⌬-58mut, 5Ј-TGATGTTCTCTCTCGAGT-ATGTGC; SNR52 5Ј⌬-7, 5Ј-CAACTCTAGATTTTGTAGT-GCCC; and SNR52_rev as reverse primer, and the proper plasmid template. 3Ј-Truncated forms of SNR52 constructs (3Ј⌬ϩ60) were obtained through PCR amplification using the oligonucleotide 5Ј-GAAAAAAAAATTGTAGGGTGTG as a reverse primer in combination with the different forward primers. The resulting DNA fragments were cloned into pNEB193. The plasmids containing the S. cerevisiae tRNA genes tN(GTT)CR, tN(GTT)NR, tP(AGG)CR, and tP(AGG)NR have been described (24). Their 3Ј⌬ϩ64 truncated versions were prepared using the following reverse primers: 5Ј-GGAAAAAAAACACGATCTTG-CGATTAAC for tDNA Asn (GTT) and 5Ј-GGAAAAAAAAA-CGGGACCTCC for tDNA Pro (AGG). All the constructs were sequence-verified by dideoxy chain termination sequencing.
DNA Binding Assays-For DNase I footprinting analysis, a 415-bp SNR52 fragment, 5Ј-end-labeled with [␥-32 P]ATP on the sense strand, was generated by PCR using 5Ј-labeled SNR52_fw and unlabeled SNR52_rev oligonucleotides as primers, and the wt, Adown, Bdown, or Tdown versions of SNR52 as templates. 4 fmol (ϳ20000 cpm) of the purified fragment were incubated with 0.6 g of TFIIIC, partially purified up to the DEAE Sephadex A-25 step following a previously described procedure (34), in 18-l binding reaction mixtures containing 160 mM KCl, 10 g/ml supercoiled pBlueScript-KS plasmid DNA, 0.1 mg/ml bovine serum albumin, 5% glycerol, 20 mM HEPES pH 8, 4 mM MgCl 2 , 0.5 mM dithiothreitol. The complex was treated with 0.5 ng (0.05 units) of pancreatic DNaseI (Amersham Biosciences, E2215Y type) for 1 min at 25°C. The digestion was stopped with 20 mM EDTA. Footprinting mixtures were phenol-extracted, ethanol-precipitated, and fractionated on a 5% polyacrylamide, 7 M urea sequencing gel, sided by products obtained from a dideoxy chain termination sequencing reaction conducted on wt SNR52 using the 5Ј-labeled SNR52_fw oligonucleotide. The gel was phosphorimaged with a Personal Imager FX (Bio-Rad). For gel retardation assays, 445-bp DNA fragments, containing the wt or mutated SNR52 transcription unit, were prepared by BamHI-SacI digestion of the corresponding pNEB193 plasmids. The fragments were end-labeled with [␣-32 P]dATP and the Klenow fragment of DNA polymerase I (Promega), gel-purified, and quantified by ethidium bromide staining. 3 fmol of DNA fragment (15,000 cpm) were then incubated 15 min at 25°C with 0.6 g of partially purified TFIIIC in a final volume of 15 l. The standard binding reaction mixture contained 90 mM (NH 4 ) 2 SO 4 , 13 g/ml supercoiled pBlueScript-KS plasmid DNA, 1.3 mg/ml bovine serum albumin, 10% glycerol, 10 mM Tris/HCl pH 8, 0.1 mM EDTA. In TFIIIB-containing reactions, DNA was first incubated with TFIIIC, then recombinant TFIIIB components (prepared as described below) were added in amounts ranging from 20 to 160 ng (see legend to Fig. 4). The complexes were loaded on a pre-equilibrated native polyacrylamide gel (4% acrylamide, 0,048% bisacrylamide) containing 20 mM Tris/HCl, pH 8, 1 mM EDTA, 5% glycerol. The running buffer contained 20 mM Tris/ HCl, pH 8, 1 mM EDTA, 5 mM ␤-mercaptoethanol. After loading, the gel was run 4 -6 h at 4°C, with frequent changes of running buffer. The gels were then dried and phosphorimaged.
In Vitro Transcription Assays-Transcription of class III genes was reconstituted in vitro essentially as described (15,33). All reactions contained 0.6 g of TFIIIC partially purified up to the DEAE Sephadex A-25 step (34), except that yeast nuclear extract (35) was used as a starting material; 40 ng of recombinant TBP and 80 ng of recombinant Brf1, both purified from overexpressing Escherichia coli cells (34); and 10 ng of highly purified RNA polymerase III (15). As a source of Bdp1 protein, required to reconstitute TFIIIB activity, we used either 1.5 g of BЉ fraction, partially purified from chromatin pellets generated during yeast nuclear extract preparation (36), or 30 ng of recombinant 8His-Bdp1 protein purified from baculovirus-infected cells (15,37). Fully recombinant TFIIIC was expressed and purified as described (38). For the preparation of recombinant Nhp6A protein used in the experiment in Fig. 7C, the S. cerevisiae NHP6A open reading frame was cloned into a modified version of pET28b containing an engineered PmeI restriction site into the polylinker to facilitate cloning of PCR products (39). The construct was transformed into BL21 Rosetta TM (DE3) E. coli cells (Novagen). Nhp6A expression was induced by adding isopropyl-1-thio-␤-D-galactopyranoside (1 mM) and incubating for 2 h at 37°C. The His 6 -Nhp6A protein in the soluble fraction was purified by chromatography on Ni-NTA resin (Qiagen) under non-denaturing conditions, following the manufacturer's instructions. The template of in vitro transcription reactions (100 -200 ng of class III gene-containing plasmid) was preincubated at 20°C for 20Ј in a 45-l reaction mixture (25 mM Tris/HCl, pH 7.9, 100 mM KCl, 5 mM MgCl 2 , 8% (v/v) glycerol, 8 units of SUPERase-In RNase inhibitor (Ambion)) in the presence of transcription proteins (except Pol III). Pol III was then added together with cold NTPs (500 M ATP, CTP, and GTP; 25 M UTP) and 10 Ci of [␣-32 P]UTP (800 Ci/mmol, Amersham Biosciences), and multiple rounds of transcription were allowed to take place for 20 min at 20°C. Radiolabeled transcripts were separated on 6% polyacrylamide/7 M urea gels, and visualized and quantified by phosphorimaging using a Personal Imager FX (Bio-Rad). Primer extension analysis of in vitro produced transcripts was performed essentially as described (28). In vitro transcription reactions were carried out under the above described conditions, except that radiolabeled UTP was not included. Reaction products were then purified and used for primer extension analysis with Superscript III RNase H Ϫ (Invitrogen) as a reverse transcriptase (200 units/reaction) and the 5Ј-end-radiolabeled oligonucleotide primer: 5Ј-GTATCAGAGATTGTTCACGCTAATG.
In Vivo RNA Analyses-The SNR52-null strain used for in vivo analyses is yM4585 (MATa his3⌬200 lys2-801 leu2-3,2-112 trp1-901 tyr1-501 URA3 ADE2 CAN S ⌬SNR52-HIS3) (40), in which the 92-bp sequence corresponding to the mature snR52 RNA, plus ϳ10 bp of upstream and downstream flanking sequences, have been deleted and substituted with the HIS3 marker gene. RNA extraction was performed as described (41) from cells grown to an A 600 of 0.8 -1 in the appropriate selective medium. 15 g of total RNA were fractionated on a 6% polyacrylamide/7 M urea minigel, then transferred to a positively charged nylon membrane (Gene Screen Plus, PerkinElmer Life Sciences). The filter was hybridized 15 h at 42°C with a 5Ј-end-radiolabeled oligonucleotide probe (5Ј-CAGAAGGAAGGCAACATAAGT) in 5ϫ SSC, 5ϫ Denhardt's solution, 0.1 mg/ml denatured salmon sperm DNA, 0.5% (w/v) SDS, followed by two 10-min washing at 30°C with 0.1% SDS and SSC 2ϫ and then SSC 1ϫ. The U3specific oligonucleotide probe used for normalization was 5Ј-CTTTGCCGTTGCATTTGTAGTTTTTTCCTTTGGAAGT.

RESULTS
Mutational Analysis of the SNR52 Promoter-The S. cerevisiae SNR52 gene has recently been identified as a novel class III transcription unit in which a snoRNA coding region is preceded by a leader region containing putative box A and box B control regions, as well as a TATA element located ϳ60-bp upstream of the box A (10). The sequence and organization of the SNR52 transcription unit are reported in Fig. 1A. Both the box A and box B upstream of the snoRNA coding sequence can be easily recognized by analysis with the Pol3scan program (42), and have been recently found to be conserved in hemiascomycetes (43). The distance between the two putative control elements (86 bp) is considerably higher than the average distance found in the yeast tRNA gene complement (37 bp). A curious feature of the transcription unit is the presence of a run of 6 consecutive T residues located a few bp downstream of the box A. Even though the T 6 sequence is a strong termination signal for Pol III, we have recently shown that it behaves as an unusually weak terminator in the SNR52 context (33). The T 6 element might be relevant for SNR52 expression, because it is also present in this position in the genomes of at least two other Saccharomyces species (S. paradoxus and S. mikatae). To start to characterize the SNR52 promoter, we amplified the SNR52 transcription unit from yeast genomic DNA, and we introduced mutations into the different putative control elements: box A, box B, and the T run between them. The wt and mutant templates were assayed for their ability to support Pol III transcription both in vitro and in vivo. For in vitro assays, we used a Pol III-specific in vitro transcription system containing balanced amounts of recombinant TBP and Brf1 proteins, partially purified BЉ and TFIIIC fractions and highly purified RNA polymerase III (34). For in vivo analysis, mutagenized SNR52 derivatives were inserted into the multicopy pFL45S vector (32) and transformed into a snr52::HIS3-null mutant strain (kindly provided by Lowe and Eddy (40)). The results of in vitro transcription analysis are shown in Fig. 1B. Transcription of the wt template produced a 250-nt long primary transcript and two smaller RNAs of 44 -45 nt (lane 1). The shorter RNAs correspond to transcripts terminated at the internal T 6 sequence; indeed, transcription of a mutant template in which the T run was interrupted produced higher levels of the full-length transcript, whereas the short RNAs disappeared (lane 4). Transcript quantification, conducted by taking into account the different number of radiolabeled U residues incorporated into RNAs of different sizes, revealed only a slight decrease in transcription of the T-run mutant with respect to wt. In contrast, a double point mutation in the box A (CC in place of the universally conserved GG at positions ϩ29, ϩ30) abolished specific in vitro transcription (lane 2). Surprisingly, a G to C replacement in the box B, predicted to abolish TFIIIC binding and transcription (see for example (2)) only produced a 2-fold reduction of in vitro transcription with respect to wt (cf. lanes 1 and 3). Very similar in vitro results were obtained using a crude yeast nuclear extract as a source of transcription machinery (data not shown). The in vivo effects of the same mutations were analyzed by Northern blot after introducing the SNR52 variants into the snr52-null mutant strain. The results are shown in Fig. 1C. A probe hybridizing with the mature snR52 snoRNA detected both the 90-nt mature product and the 250-nt precursor in the wt strain (lane 1). Box A and box B mutations both reduced the mature snR52 RNA to barely detectable levels (lanes 2 and 3; the precursor RNA was undetectable even after longer expo-sure), while the T run mutation did not significantly affect the levels of SNR52 transcription products (lane 4). The A box and B box are thus both required for SNR52 transcript accumulation in vivo, whereas in vitro the box B appears to be dispensable for transcription. Fig. 1D shows the results of primer extension analysis conducted on in vitro produced SNR52 transcripts. Transcription initiated at a single position 18 bp upstream of the box A. As to the identity of this position, the gel shown in Fig. 1D leaves some ambiguity between the T and the A at the center of a CTAG stretch. Because the general rule for Pol III is that it initiates transcription at a purine preceded by a pyrimidine residue (44,45), we assume that the A within the CTAG stretch is the actual TSS for SNR52. Fig. 1D further shows that the TSS is not altered by mutation of the B box (lane 3) nor by interruption of the intragenic T run (lane 4). Despite the very low levels of precursor transcript in vivo, we were also able to verify that the TSS observed in vitro is the same as in vivo (data not shown).
Binding of TFIIIC to the SNR52 Control Region-The observation that box B mutational inactivation does not alter drastically the in vitro transcription capacity of SNR52 prompted us to directly analyze the TFIIIC binding properties of wt and mutant SNR52 templates. Fig. 2A shows the results of a DNase I footprinting analysis. A clear protection over the box B, but not over the box A, was observed with the wt template (cf. lanes  1 and 2). The same pattern was observed with the T stretch mutant (lanes 7 and 8). The lack of box A protection might be a consequence of the unusually large distance between box A and box B. This DNase I protection pattern was abolished upon mutational inactivation of the B box (lanes 5 and 6), while mutation of the box A produced a less dramatic decrease of box B protection (lanes 3 and 4). The results of gel retardation experiments, shown in Fig. 2B, were in reasonable agreement with the footprinting analysis. As expected, mutation of the B box abolished TFIIIC binding to SNR52 (cf. lanes 1, 2 with lanes 5, 6), whereas T run disruption had no effect on TFIIIC binding under the same conditions (lanes 7 and 8). The box A mutation, causing a weakened box B protection in the footprinting assay, did not affect the amount of gel-retarded TFIIIC-DNA complex (lanes 3 and 4). Because TFIIIC and DNA concentrations were very similar in footprinting and gel-retardation assays, the partial discrepancy between the results of the two assays might be due to a higher sensitivity of the footprinting assay in revealing suboptimal TFIIIC-DNA interactions. From the results in Figs. 1 and 2, we conclude that SNR52 can be efficiently transcribed by Pol III in vitro even in the absence of stable TFIIIC binding,  1-4). Shown in lanes 5-8 are the results of sequencing reactions conducted using the same 5Ј-labeled oligonucleotide utilized for primer extension. The sequence of the non-transcribed DNA strand around the start site (ϩ1) is indicated on the right.
whereas box A mutation abolishes transcription without severely affecting TFIIIC binding. One possible explanation of such unexpected behavior could be that TFIIIC, though required for SNR52 expression in vivo, is dispensable for its transcription in vitro (in this case, the box A requirement for in vitro transcription would be explained by assuming the existence of essential interactions between the A box and unknown transcription components). The results reported in Fig. 2C, however, rule out this possibility, by showing that TFIIIC is absolutely required for in vitro transcription of both wt SNR52 and SNR52 variants mutated in the box B or at the internal T run (Bdown, Tdown). Therefore, in the case of the Bdown tem-plate, TFIIIC promotes transcription without stably interacting with the SNR52 promoter.
The apparent extension of tRNA gene control regions can be manipulated by the conditions used to measure transcription in vitro. In particular, the use of high concentrations of templates can overcome in vitro the requirement for part of the control region (46). The template concentration used for the transcription assay in Fig. 1 was 1 nM. In the experiment in Fig. 3, the wt and mutant SNR52 templates were transcribed in vitro at concentrations ranging from 0.1 nM to 2 nM, in the presence of constant amounts of transcription proteins. Lowering the template concentration produced a clear decrease of transcription output in all cases (Fig. 3B). Such a decrease, however, was most evident in the case of the B box mutant, whose transcription dropped by ϳ6-fold in going from 2 nM to 0.1 nM (cf. lanes 11 and 15 in Fig. 3A; see also Fig. 3B). By comparison, transcription of wt SNR52 was only 2.5-fold reduced at 0.1 nM with respect to 2 nM template (cf. lane 1 with 5; see Fig. 3B). Therefore, the affinity of the transcription machinery for SNR52 is significantly reduced by the B box mutation. However, in vitro transcription of this template was still well detectable at subnanomolar template concentrations (lanes [11][12][13]. The discrepancies between the results of gel retardation and in vitro transcription assays with the Adown and Bdown SNR52 mutants were still observed when the two assays were carried out under identical reaction conditions with respect to the concentration of salt and glycerol (data not shown). We thus asked, by gel retardation assays, whether TFIIIB is able to associate with, and/or stabilize the TFIIIC-DNA complex at the Bdown promoter, and whether the Adown template, which is transcriptionally inactive, is able to recruit TFIIIB. The results of this experiment are shown in Fig. 4. The complex between TFIIIC and the Adown template (lane 8) could be specifically supershifted by fully recombinant TFIIIB (lane 10), exactly as the wt complex did (lanes 2 and 4), while the Bdown template did not form any retarded complex, even in the presence of TFIIIB (lanes 13-18). We argue from these results that in vitro transcription of the Bdown mutant involves the formation of a  2, 4, 6, and 8) or in the absence (lanes 1, 3, 5, and 7) of TFIIIC, followed by electrophoretic fractionation of TFIIIC-bound and free DNA molecules on a native polyacrylamide gel. The percentages of bound DNA (with respect to total DNA) in each TFIIIC-containing reaction are reported below the gel image. C, indicated SNR52 template variants were transcribed in vitro either in the presence (lanes 2, 4, 6, and 8) or in the absence (lanes 1, 3, 5, and 7) of TFIIIC. The migration positions of the 250 nt, primary transcript (250) and of the short transcripts terminated at the internal T run (45) are indicated on the right (the two panels derive from the same gel image).
TFIIIC/TFIIIB-DNA complex that is able to productively recruit Pol III despite being much less stable than canonical TFIIIC-and TFIIIB-containing preinitiation complexes, and that transcriptional inactivation of the Adown mutant is not caused by reduced TFIIIB recruitment.
The SNR52 Box A Acts as an Autonomous Internal Promoter Element for Pol III Transcription-The SNR52 transcription unit might contain, downstream of box A, a pseudo-B box that can be used when the natural B box is mutated. To address this point, we constructed a mutant version of SNR52 (SNR52_3Ј⌬ϩ60) in which the transcription unit is truncated 24-bp downstream of box A by the insertion of a T 9 termination sequence (Fig. 5A). Because all the region downstream of box A, including box B, is absent from this template, the observed transcription output will be caused by autonomous promoter activity of box A. Fig. 5B, lane 2, shows that the truncated template containing a wt box A could be efficiently transcribed in vitro. Its transcription produced, as expected, a 45-nt transcript, corresponding to termination at the internal T 6 run (this transcript disappeared when the T run was interrupted; see lane 4), and a longer RNA (ϳ65 nt) corresponding to termination at the artificially inserted T 9 terminator (lanes 2 and 4). The reaction in lane 1 was programmed with full-length, wt SNR52 template. Quantification showed that SNR52_3Ј⌬ϩ60 was transcribed as efficiently as wt SNR52. Lane 3 in Fig. 5B further shows that the box A was absolutely required for transcription of the truncated template. To exclude the possibility that a pseudo-B box, present in the vector sequence downstream of the truncated transcription unit, might provide a TFIIIC anchoring site in the absence of the natural box B (47), we tested the transcription capacity of  both the truncated, SNR52_3Ј⌬ϩ60 template and the fulllength Bdown template (see Fig. 1) after linearization at a restriction site located, within the vector polylinker, just downstream of the T 9 terminator (in the case of SNR52_3Ј⌬ϩ60) or of the natural SNR52 terminator (in the case of wt and Bdown SNR52). As shown in Fig. 4C, plasmid linearization produced a general decrease in transcription efficiency with all the tested templates (wt SNR52, SNR52 Bdown and SNR52_3Ј⌬ϩ60; cf. lanes 1, 3, and 5 with lanes 2, 4, and 6, respectively). However, both the linearized, box B-defective templates were transcribed almost as efficiently as the linearized, wt SNR52 gene (cf. lane 2 with lanes 4 and 6). The results in Fig. 5 thus show that the SNR52 box A can autonomously (i.e. independently from the B box) mediate the productive association of the transcription machinery and accurate initiation at the SNR52 TSS. Transcription competition experiments, in which wt or truncated box B-less SNR52 templates were transcribed in vitro in the presence of increasing concentrations of a competitor tDNA Ile , further showed that box B-less SNR52 is significantly affected, with respect to the wt gene, in its ability to sequester the transcription machinery (data not shown).
It is well known that the 5Ј-flanking region of class III genes can strongly contribute to overall promoter strength (see Ref. 24 and references therein). It thus seemed likely that some features of the SNR52 5Ј-flanking region, such as the TATA element around position Ϫ40 and the T-rich region starting at Ϫ53, might influence SNR52 transcription, and possibly play an essential role, perhaps in synergy with the box A, in transcription of the Bdown or box B-less SNR52 templates. To test this possibility, we constructed variants of wt, Bdown and 3Ј⌬ϩ60 (box B-less) SNR52 templates in which the 5Ј-flanking region was partially or totally mutated. The 5Ј-flanking sequences of these constructs are reported in Fig. 6A. Fig. 6B (lanes  1-4) shows that transcription of wt SNR52 was not affected by disruption of the TATA and T-rich elements, while the complete replacement of the 5Ј-flanking sequence by vector sequence produced a ϳ2-fold reduction in transcription (see also Fig. 6C). The Bdown SNR52 template behaved similarly: its transcription was neither decreased by deletion of all the region upstream of Ϫ58 (lane 6), nor by disruption of both the TATA-like element at Ϫ40 and the T-rich element at Ϫ53 (lane 7), whereas the complete substitution of the 5Ј-flanking region with vector-derived sequence (lane 8) resulted in a 2-fold transcription decrease. A similar behavior was observed for the 3Ј⌬ϩ60 (box B-less) template (lanes 9 -10). The SNR52 box A thus behaves as a fully autonomous core promoter element for Pol III transcription.
Box B-independent in Vitro Transcription of tRNA Genes-Because the box B is generally considered an essential promoter element for tRNA gene transcription, we wondered whether the box B-independent transcription of SNR52 reflects some unusual features of this particular template, related for example to its box A. We thus constructed box B-less versions of four different tRNA genes: two members of the tDNA Asn (GTT) family, tN(GT-T)CR and tN(GTT)NR, and two members of the tDNA Pro (AGG) family, tP(AGG)CR and tP(AGG)NR. The tDNAs in each pair have identical ICRs but different 5Ј-flanking regions characterized by different affinities for TFIIIB (24). As shown in Fig. 7A, the A box region of the tDNA Asn pair is more similar to the SNR52 A box than the A box of the tDNA Pro pair. Box B-less, truncated versions  1 and 2), Bdown (lanes 3 and 4), or box B-less (lanes 5 and 6) SNR52 templates, either in a circular supercoiled form (lanes 1, 3,5) or in a EcoRI-linearized form (lanes 2, 4, 6), were used to program in vitro transcription reactions. The transcriptional output of each reaction, relative to the output (arbitrarily set to 100) of the wt, supercoiled template (lane 1), is reported below each lane (Txn).
of these tDNAs were constructed by inserting a terminator sequence (T 8 ) 25-bp downstream of the box A and deleting all the downstream portion of the tDNA. The in vitro transcription efficiencies of the box B-less, truncated tDNAs were then compared with those of the corresponding wt templates. As expected on the basis of a previous study (24), the wt tN(GTT)NR template was transcribed less efficiently than wt tN(GTT)CR, because of differences in the 5Ј-flanking region (Fig. 7B, cf. lanes 1-5 with lanes   11-15). Quite surprisingly, the box B-less versions of tN(GTT)CR (lanes 6 -10) and tN(GTT)NR (lanes 16 -20) both produced detectable levels of truncated tRNA Asn transcript. Remarkably, however, as revealed by quantification (taking into account the lower number of incorporated, 32 P-labeled U residues into the truncated transcripts), the truncated, box B-less tN(GTT)CR was as efficiently transcribed as wt tN(GTT)CR (cf. lanes 6 -10 with lanes 1-5), whereas the box B-less tN(GTT)NR was transcribed 4 -5 times less efficiently than the corresponding wt tDNA (cf. lanes  16 -20 with lanes 11-15). Therefore, at variance with SNR52 (see Fig. 6), the ability of tDNA Asn (GTT) box A to promote transcription independently from the box B is significantly influenced by the 5Ј-flanking region. The box B-independent transcription of tDNA Asn (GTT) was still observed after linearization of the templates just downstream of the artificially inserted terminator (data not shown), thus excluding that residual transcription is due to pseudo-B boxes possibly present in the vector sequence downstream of the A box (47). Because the tN(GTT)NR upstream region has been characterized as a weak TFIIIB binding site (24), we asked whether increasing the concentration of TFIIIB components would specifically enhance the transcription of box B-less tN(GTT)NR. In the experiment in Fig. 7C, individual recombinant TFIIIB components, as well as all different combinations of them, were provided in a 5-fold excess with respect to standard transcription conditions. The transcription of box B-less versions of both tN(GTT)CR and tN(GTT)NR was not stimulated by an excess of any individual TFIIIB component (lanes 1-4 and 10 -13), nor by supplementation of the high mobility group Nhp6 protein (lanes 5 and 14), which has been previously reported to stimulate SNR6 transcription by Pol III (48,49) and to influence transcription start site selection on yeast tRNA genes (50). When TBP and Brf1 proteins were supplemented together, however, ϳ4-fold and ϳ3fold stimulations were observed for box B-less tN(GTT)CR and tN(GTT)NR, respectively (lanes 6 and 15). Stimulations were slightly less effective when excess Bdp1 protein was also added together with TBP and Brf1 (lanes 9 and 18). TFIIIB component Ϫ58 (with respect to the TSS), was deleted. The upstream T-rich and TATA elements are in bold. The base substitutions that have been introduced to generate the 5Ј⌬-58mut template variants are underlined. The TSS (ϩ1) is in italic bold. In the case of the 5Ј⌬-7 variants, the plasmid-derived sequence that replaces the original 5Ј-flanking sequence is in italics. B, SNR52-derived templates containing wt (lanes 1-4), Bdown (lanes 5-8), or 3Ј⌬ϩ60 (lanes 9 and 10) transcribed regions in combination with wt (lanes 1, 5, 9), 5Ј⌬-58 (lanes 2 and 6), 5Ј⌬-58mut (lanes 3 and 7), or 5Ј⌬-7 (lanes 4, 8, and 10) upstream regions, were used to program in vitro transcription reactions. Reactions in lanes 1-8 contained standard amounts of TFIIIB components. Reactions in lanes 9 and 10 contained reduced amounts (0.8 g) of BЉ fraction, as higher amounts were found to be selectively inhibitory for the 3Ј⌬ϩ60/5Ј⌬-7 template. The migration positions of the internally terminated (45) and full-length (250) SNR52 transcripts, as well as of transcripts terminated at the T 9 terminator introduced to generate the B box-less templates (65) are indicated on the left and on the right. C, bar plot derives from phosphorimaging quantification of the experiments shown in B and of duplicate experiments conducted in parallel. The different templates are grouped, on the x-axis, according to the sequence of their transcribed region (wt, Bdown, or 3Ј⌬ϩ60). The average transcription levels (expressed as percentage of the levels observed, within each group, with the template having a wt upstream region) are reported on the y-axis. The different bar colors refer to the different 5Ј-flanking regions (wt, 5Ј⌬-58, 5Ј⌬-58mut, 5Ј⌬-7) as indicated in the inset. Error bars indicate the S.E. supplementation did not level the transcriptional difference between the two templates. Under all the conditions tested, the box B-less tN(GTT)NR template was always more than 10 times less transcribed than box B-less tN(GTT)CR. A completely different behavior, in box B-independent transcription, was observed in the case of tRNA Pro (AGG) genes. As shown in Fig. 7D, tP(AGG)NR was a better template than tP(AGG)CR, as it was very efficiently transcribed at the lowest concentration tested (0.1 nM, lane 11). When the isocoding tP(AGG)CR was used at the same concentration, no transcription was observed (lane 1), most likely due to the lower transcriptional strength of its upstream region (24). Despite the high affinity of tP(AGG)NR for the transcription machinery, however, its box B-less variant could not be transcribed (lanes 16 -20). The box B-less version of tP(AGG)CR was also transcriptionally inactive (lanes 6 -10). The ability to promote transcription complex assembly in vitro in a B boxindependent fashion is thus not a general property of tDNA box A elements.

SNR52 and Box B-independent Transcription in the Presence of Allrecombinant Transcription Factors-
A recent paper has reported the reconstitution of in vitro class III gene transcription from fully recombinant TFIIIB and TFIIIC (38). We thus asked whether SNR52 can be transcribed in a system reconstituted with all-recombinant factors, and whether fully recombinant TFIIIC and TFIIIB are sufficient to support box B-independent transcription. In the experiment in Fig. 8A . lanes 3, 7, and 11 with  lanes 4, 8 and 12, respectively). When fully recombinant TFIIIB was used, however, neither native nor recombinant TFIIIC could support transcription of any of the templates (lanes 1, 2, 5, 6, 9, and 10). The BЉ fraction thus appears to contain, in addition to Bdp1 protein, additional component(s) required for SNR52 transcription. Such components seem not to be required for TFIIIB assembly onto SNR52, because the SNR52-TFIIIC complex could be specifically supershifted by all-recombinant TFIIIB in gel-  5 and 14); and all the different combinations of TFIIIB components in the above reported amounts, as indicated above the lanes. D, increasing concentrations (from 0.1 nM to 2 nM) of either tP(AGG)CR, in its wt (lanes 1-5) or B box-less (lanes 6 -10) versions, or tP(AGG)NR, in its wt (lanes [11][12][13][14][15] or B box-less (lanes 16 -20) versions, were used to program in vitro transcription reactions. The migration position of the full-length, pre-tRNA Pro transcript is indicated on the left. retardation assays (see Fig. 4). We also tested the ability of different TFIIIC and TFIIIB combinations to transcribe wt and box B-less versions of tN(GTT)CR. Fig. 8B shows that, as expected (38), the wt tN(GTT)CR template could be transcribed in the presence of all-recombinant TFIIIC and TFIIIB, albeit at very low efficiency (lane 1). tN(GTT)CR transcription was strongly stimulated when BЉ fraction replaced recombinant Bdp1 (cf. lanes 1 and 2). Recombinant TFIIIC could also support transcription of the box B-less tN(GTT)CR template (lane 5). With this template, the requirement for non-Bdp1 component(s) in the BЉ fraction was even more evident, as no transcription was observed in the presence of fully recombinant factors (lane 4).

DISCUSSION
In this study, we show that SNR52 transcription by RNA polymerase III depends on a box A/box B combination within the 5Ј-leader region that binds TFIIIC and allows for efficient transcription initiation at a site located ϳ160-bp upstream of the mature snoRNA coding sequence. Both A and B box are required for in vivo accumulation of the mature snoRNA. Because the ICR sequences do not appear in the mature RNA product, the reduced snoRNA levels observed in vivo with Adown and Bdown SNR52 templates are most likely caused by impaired transcription, and not to reduced stability of the RNA products. Quite unexpectedly, in vitro analysis revealed that the SNR52 box B, though essential for TFIIIC binding, is dispensable for in vitro transcription, whereas the box A, giving only a minor contribution to TFIIIC binding affinity, is essential for transcription. The dispensable character of box B in vitro was also observed for two tRNA Asn (GTT) genes, whose box A behaved as an autonomous intragenic promoter element able to nucleate TFIIIC-dependent TFIIIB assembly. In contrast, the B box of two tRNA Pro (AGG) genes could not be removed without inactivating the templates. The B box-independent in vitro transcriptional activity of the tRNA Asn genes, but not of SNR52, was found to be strongly influenced by the 5Ј-flanking region.
The finding that tRNA gene transcription can take place in the absence of box B is not unprecedented. Early studies, analyzing transcription of different tRNA gene fragments in crude extracts, reported that box B-less, truncated tRNA genes, at high concentrations, can initiate transcription in vitro at their normal site, albeit very inefficiently (46, 51-53) (reviewed in Ref. 1)). Our data strengthen and extend these observations by showing that box B-independent transcription takes place at a remarkable efficiency in vitro even when the template concentration is in the subnanomolar range. The apparent equilibrium constant for yeast TFIIIC binding to the SUP4 tRNA Tyr gene has been estimated to be ϳ1.9 ϫ 10 9 M Ϫ1 , and to drop by 370fold, to 5 ϫ 10 6 M Ϫ1 , upon box B mutational inactivation (54). The fact that a box B-less tDNA, like tN(GTT)CR 3Ј⌬ϩ64, is transcribed at wild-type levels even at a template concentration of 1 ϫ 10 Ϫ10 M indicates that, in our in vitro transcription conditions, a TFIIIC-box A interaction, though very weak, is nevertheless sufficient to nucleate TFIIIB assembly. At variance with SNR6, whose TATA box-dependent, box B-independent transcription could not be observed in crude extracts, because of the presence of inhibitory chromatin components normally counteracted by TFIIIC (55), box B-independent transcription of both SNR52 and tDNA Asn (GTT) was also observed with yeast nuclear extract as a source of Pol III transcription machinery (data not shown). Our data thus suggest that, as previously pointed out (1), the stability of the TFIIIC-DNA complex, whose major contribution comes from interaction with the B box, is not absolutely required for transcription. On the other hand, what emerges from promoter analysis of SNR52 is the centrality of box A in directing productive TFIIIB assembly and transcription initiation. We found that the SNR52 box A is absolutely required for transcription not only in vivo but also  -4), Bdown (lanes 5-8), or 3Ј⌬ϩ60 SNR52 templates were transcribed in vitro in the presence of either 100 ng of recombinant TFIIIC (rTFIIIC; lanes 1, 3,5,7,9,11) or 0.6 g of partially purified native yeast TFIIIC (natC; lanes 2,4,6,8,10,12) in combination with either fully recombinant TFIIIB (lanes 1, 2, 5, 6,9,10) or BЉ-containing TFIIIB (lanes 3,4,7,8,11,12). The migration positions of the internally terminated (45) and full-length (250) SNR52 transcripts, as well as of transcripts terminated at the T 9 terminator introduced to generate the B box-less templates (65) are indicated on the left and on the right. B, tN(GTT)CR (lanes 1-3) and tN(GTT)CR 3Ј⌬ϩ64 (lanes 4 -6) templates were transcribed in vitro in the presence of either fully recombinant TFIIIC (rTFIIIC; lanes 1,2,4,5) or partially purified native yeast TFIIIC (natC; lanes 3 and 6) together with either fully recombinant TFIIIB (lanes 1 and 4) or BЉ-containing TFIIIB (lanes 2, 3, 5, 6). The migration positions of full-length or truncated tDNA Asn transcripts are indicated on the left and on the right, respectively. in vitro. This observation contrasts with all the previously reported box A mutations, whose effects on Pol III transcription were never reported to be as severe as in the case of Adown SNR52 (1). Accordingly, early studies of the effects on tRNA gene transcription of base substitutions in the ICRs led to the common notion that tRNA gene promoter strength is principally determined by the B box, while the TFIIIC-box A interaction plays a secondary role in favoring TFIIIB assembly and start site selection (47, 56 -58). A more recent analysis of promoter organization of the non-tRNA class III gene SCR1 confirmed the primary role of the box B in determining the promoter strength in vitro, but also pointed out the importance of TFIIICbox A interaction in counteracting repressive chromatin assembly in vivo (2). A chromatin antirepression role of the TFIIIC-box A interaction has also been highlighted for the S. cerevisiae SNR6 gene (21). In the case of SNR52, in the absence of other transcription factors, the B box appears to be the main determinant of TFIIIC binding affinity (TFIIIC binding is abolished by a B box point mutation, while it is only slightly affected by A box disruption), yet the box A is essential for transcription, and it can even work independently from box B in directing efficient in vitro transcription. The results of gel-retardation analysis suggest that the SNR52 box A mutation does not impair TFIIIC binding nor TFIIIB recruitment. Its drastic effect on transcription thus likely results either from TFIIIB misplacement, perhaps directed by other, non-optimal box As, of from an unexpected involvement of box A in the postrecruitment step of the transcription cycle.
With respect to box B-less tDNAs, the autonomous promoter activity of the box A was found to be sequence contextdependent. In particular, the analysis of two box B-less tDNA Asn (GTT) templates differing for their upstream regions revealed that, in the absence of stable TFIIIC binding mediated by the B box, the box A and 5Ј-flanking region act synergistically to promote the assembly of a functional preinitiation complex. Moreover, the sequence of the box A itself might be relevant for autonomous promoter activity, as suggested by the observation that the SNR52 and tDNA Asn (GTT) A boxes, both supporting box B-independent transcription, are more similar in sequence to each other than to the tDNA Pro (AGG) box A, which is unable to support box B-independent transcription (see Fig. 7). A synergistic action of box A and upstream sequence elements in transcription initiation has previously been observed. For example, it has been shown that SNR6 transcription initiation in vivo is insensitive to mutation of the upstream TATA box, unless the box A is simultaneously mutated (59). It is also known that TSS selection on tRNA genes is co-directed by a TFIIIB component, TBP, and by the most upstream, A box-interacting portion of TFIIIC (60). By contrast, the position of the B box downstream of box A does not affect TSS selection, and transcription efficiency is generally unaffected even at very large A box to B box distances (14,15,61). By underlining the partially dispensable character of box B for transcription in vitro, our data contribute to define the upstream moiety of tRNA genes, comprising the box A and the 5Ј-flanking sequence, as a minimal core promoter module where all the key interactions leading to transcription initiation can take place. With respect to this, it was interesting to find that box B-independent tDNA transcription could be strongly stimulated by simultaneously increasing the concentrations of both TBP and Brf1, but not of either component alone. Such a cooperative effect is in agreement with the fact that these two components interact with each other (62), with TFIIIC subunits (29,63) and with upstream DNA (64). In particular, by binding to both TBP and the 131 subunit of TFIIIC, Brf1 could contribute to generate an extension of the DNA binding surface of the box A-interacting TFIIIC domain, thus conferring to the transcription factor-DNA complex sufficient stability to perform Pol III recruitment. Within this scenario, the role of the B box in the canonical TFIIIC-DNA interaction would simply be to contribute additional stability to the transcription complex that assembles on the core promoter elements. It is interesting to note that such a minimal TFIIIC/ TFIIIB assembly module is strongly reminiscent of the composite organization of the RNA polymerase II core promoter. A Pol II core promoter usually comprises some combination of an upstream moiety, containing a TATA box preceded by a BRE (TFIIB-recognition element), an Initiator element surrounding the TSS, and a downstream promoter element (DPE), conserved from Drosophila to humans, which is located at ϳϩ30 with respect to the TSS (65). Correspondingly, in tRNA genes, an Initiator-like element surrounding the TSS (and characterized, in some genomes, by the presence of a CAA triplet (24)) is flanked immediately upstream by a TFIIIB-binding region often carrying a TATA-like element and, downstream, by an A box element whose conserved positions are concentrated in a region roughly extending from ϩ20 to ϩ30 with respect to the TSS. These similarities in core promoter organization of Pol II and Pol III transcription units, together with our finding that the 5Ј-flank and A box are sufficient to promote TFIIIC-dependent Pol III transcription, suggest a functional parallelism between TFIID, which contacts DNA both upstream and downstream of the TSS through interactions with the TATA, Initiator and DPE elements, and the DNA recognition module composed of TFIIIB and the box A-interacting portion of TFIIIC.
Finally, we note that the availability of fully recombinant TFIIIB and TFIIIC factors allowed us to reveal the absolute requirement for additional activities, different from the well characterized TFIIIC and TFIIIB components, both for SNR52 transcription and for box B-independent tDNA transcription. This observation adds to a growing list of reports suggesting that the activity of the yeast Pol III transcription system can be directly modulated by different effectors (15,35,38,48,50,66). In particular, the possibility is suggested that SNR52 transcription requires (a) novel gene-specific factor(s), which might be also involved in transcription of the other snoRNA genes by RNA polymerase II, thus providing a way for coordinate expression of these genes by two different RNA polymerases.