Nhp6 Is a Transcriptional Initiation Fidelity Factor for RNA Polymerase III Transcription in Vitro and in Vivo*

The binding of the RNA polymerase III (pol III) transcription factor TFIIIC to the box A intragenic promoter element of tRNA genes specifies the placement of TFIIIB on upstream-lying DNA. In turn, TFIIIB recruits pol III to the promoter and specifies transcription initiating 17-19 base pairs upstream of box A. The resolution of the pol III transcription apparatus into recombinant TFIIIB, highly purified TFIIIC, and pol III is accompanied by a loss of precision in specifying where transcription initiation occurs due to heterogeneous placement of TFIIIB. In this paper we show that Nhp6a, an abundant high mobility group B (HMGB) family, non-sequence-specific DNA-binding protein in Saccharomyces cerevisiae restores transcriptional initiation fidelity to this highly purified in vitro system. Restoration of initiation fidelity requires the presence of Nhp6a prior to TFIIIB-DNA complex formation. Chemical nuclease footprinting of TFIIIC- and TFIIIB-TFIIIC-DNA complexes reveals that Nhp6a markedly alters the TFIIIC footprint over box A and reduces the size of the TFIIIB footprint on upstream DNA sequence. Analyses of unprocessed tRNAs from yeast lacking Nhp6a and its closely related paralogue Nhp6b demonstrate that Nhp6 is required for transcriptional initiation fidelity of some but not all tRNA genes, in vivo.

The binding of the RNA polymerase III (pol III) transcription factor TFIIIC to the box A intragenic promoter element of tRNA genes specifies the placement of TFIIIB on upstream-lying DNA. In turn, TFIIIB recruits pol III to the promoter and specifies transcription initiating 17-19 base pairs upstream of box A. The resolution of the pol III transcription apparatus into recombinant TFIIIB, highly purified TFIIIC, and pol III is accompanied by a loss of precision in specifying where transcription initiation occurs due to heterogeneous placement of TFIIIB. In this paper we show that Nhp6a, an abundant high mobility group B (HMGB) family, non-sequencespecific DNA-binding protein in Saccharomyces cerevisiae restores transcriptional initiation fidelity to this highly purified in vitro system. Restoration of initiation fidelity requires the presence of Nhp6a prior to TFIIIB-DNA complex formation. Chemical nuclease footprinting of TFIIIC-and TFIIIB-TFIIIC-DNA complexes reveals that Nhp6a markedly alters the TFIIIC footprint over box A and reduces the size of the TFIIIB footprint on upstream DNA sequence. Analyses of unprocessed tRNAs from yeast lacking Nhp6a and its closely related paralogue Nhp6b demonstrate that Nhp6 is required for transcriptional initiation fidelity of some but not all tRNA genes, in vivo.
The sites at which RNA polymerase III (pol III) 2 initiates tRNA gene transcription are specified by sequence-specific interactions of its two transcription factors, TFIIIC and TFIIIB, and by pol III itself. TFIIIC is a 520-kDa protein, containing six distinct subunits arranged into two globular subdomains, A (subunits Tfc1, Tfc4, and Tfc7) and B (subunits Tfc3, Tfc6, and Tfc8), which are connected by a flexible linker (Tfc8; see Refs. 1-3 for reviews). TFIIIC binds to two intragenic sequence elements, the start site-proximal box A (binding A) and the distal box B (binding B), which are optimally separated by 30 -60 bp. The B-box B interaction contributes nearly all of the affinity of TFIIIC for tRNA genes (K D Ͻ1 nM) (4). Nevertheless, it is the weak A-box A interaction that primarily specifies initiation 17-19 bp upstream of box A (5,6). Specification is indirect; the A-box A interaction determines the placement of TFIIIB, which, in turn, recruits pol III. TFIIIB placement is mediated through an interaction between Tfc4 and the Brf1 subunit of TFIIIB. Brf1 forms a stable complex with TBP, and the major DNA contacts during TFIIIC-dependent assembly of TFIIIB upstream of the start site of transcription are mediated by TBP. Consistent with this role of TBP, the DNA segments occupied by TFIIIB (ϳbp Ϫ40 to bp Ϫ10, relative to predicted start sites of transcription) are uniformly AT-rich (Ͼ70% (7)(8)(9)).
The notion that preferential sequence selection by TBP can occur during TFIIIC-dependent assembly of TFIIIB comes from in vivo and in vitro studies in which a shift in placement of a TATA box or partial TATA box shifts the start site accordingly but retains dependence on TFIIIC for transcription (10,11). Pol III also contributes to start site selection in a sequence-dependent manner, since initiation only occurs at sites specifying a purine as the initiating nucleotide preceded by a pyrimidine in the non-transcribed strand (i.e. Ϫ1/ϩ1 is always YR). In vitro and in vivo studies on SUP4 tRNA Tyr and tRNA Leu-3 gene variants that shift the position of YR motifs indicate that spacings of 17 to 20 bp are optimal for transcription and that spacings of 16 bp or less and 22 bp or more are severely compromised (5,6). In this regard, an analysis of all 273 Saccharomyces cerevisiae tRNA genes (9) 3 shows that only 26 genes lack a YR motif for initiation between bp Ϫ19 and Ϫ17 upstream of box A. Twenty-three of these genes accommodate initiation with a 16-bp spacing; of the remaining three, two have perfect TATA boxes specifying initiation 13 bp upstream of box A (8), and one is known to function poorly for transcription in vitro (12).
The combined DNA sequence-specific interactions of TFIIIC, TFIIIB, and pol III are inadequate for specifying the start site in vitro: 1) substantial purification of TFIIIB from yeast increases the proportion of minor initiation events at bp ϩ4 and ϩ8 on the SUP4 tRNA gene in assays with TFIIIC and pol III purified to near homogeneity (cf. Ref. 10); 2) replacement of TFIIIB with recombinant subunits elevates bp ϩ4 and ϩ8 initiation events to the level at bp ϩ1 (13,14). A crude Bdp1 fraction was also found to contain an excess of a factor or factors that re-establishes the fidelity of initiation at bp ϩ1 on the SUP4 gene with recombinant TFIIIB. EMSA implicated a protein or proteins that interact with TFIIIC-and TFIIIB-TFIIIC-SUP4 DNA complexes (13), although modification of TFIIIB, TFIIIC, and/or pol III (e.g. by a protein kinase) could not be excluded.
The loss of transcriptional initiation fidelity when recombinant TFIIIB, purified TFIIIC, and pol III are used to form pre-initiation complexes on the SUP4 tRNA gene appears to reflect a heterogeneity in TFIIIB placement upstream of the start site of transcription by TFIIIC. This has hampered mapping regions within TFIIIB subunits that are proximal to specific DNA locations upstream of the start site of transcription (15). To alleviate this problem we decided to turn to the start site selectivity factor.
Here we show that Nhp6a, an abundant S. cerevisiae HMG1-related protein and a component of yFACT, a factor that facilitates pol II transcription of chromatin templates (16,17), is the factor that enforces the fidelity of transcriptional initiation of the SUP4 tRNA gene by pol III in vitro and in vivo.

EXPERIMENTAL PROCEDURES
DNA-Plasmids pTZ2 (SUP4), pLNG56 (containing a transcriptionally inactive SUP4 gene with a mutated box B), and pPC1 (tRNA-L CL) have been described (18). The 250-bp SUP4 DNA fragment used for EMSA and footprinting was generated by PCR from pTZ2 with a 5Ј-32 Plabeled upstream primer (5Ј: bp Ϫ77, relative to the start site of transcription as ϩ1) and an unlabeled downstream primer (5Ј: bp ϩ173) and purified by non-denaturing PAGE. Primers for reverse transcription were 5Ј-32 P-labeled; sequences are available upon request.
Proteins-The purification and quantification of pol III, Bdp1, Brf1, TBP, TFIIIC, and the "crude" BR␣ fraction have been described (18 -20). Nhp6a and Nhp6b were generously provided by R. Johnson (UCLA) and HU from Thermotoga maritima and Helicobacter pylori and S. cerevisiae HMO1 by A. Grove (Louisiana State University). Human HMG1 was purchased.
For Figs. 1 and 3B, transcription complexes were formed by adding TFIIIB (100 fmol of Bdp1, 100 fmol of TBP, and 80 fmol of Brf1) premixed with or without Nhp6a, HU, and HMO1 (as specified in the figures) in 4-l volume to 14 l of the above reaction buffer containing 100 ng of pTZ2 or pPC1 (50 fmol) and 50 fmol of TFIIIC and incubated for 40 min; 2 l of pol III (5 or 10 fmol) was then added with an additional 20-min incubation. Single-round transcription was initiated by addition of 2.5 l of reaction buffer containing a 2 mM concentration each of ATP and CTP and 250 M [␣-32 P]UTP (20 cpm/fmol) for 5 min followed by addition of 2.5 l of reaction buffer containing 2 mM GTP and 2 mg/ml heparin for 5 min. Transcription reactions for primer extension analysis used unlabeled UTP. Primer extension with avian myeloblastosis virus reverse transcriptase was performed as described (19) using intron-specific primers. Reaction mixtures were processed for denaturing gel electrophoresis as described (18). For reaction mixtures that contained the crude BR␣ fraction, 3.7 g of this fraction in the requisite volume of bovine serum albumin diluent was used in place of TFIIIB and bovine serum albumin diluent in place of TFIIIC and pol III in the protocol described above. The order of addition transcription analysis shown in Fig. 2 used the following variations of the above protocol: 1) the reaction mixture containing TFIIIC and pTZ2 was incubated for 5 min followed by the addition of Nhp6a and TFIIIB for 40 min; 2) the reaction mixture with pTZ2, TFIIIC, and TFIIIB (16 l in this case) was incubated for 40 min followed by the addition of Nhp6a in 2 l volume for 5 min; and 3) the reaction mixture with pTZ2 and Nhp6a (14 l in this case) was incubated for 5 min followed by the addition of TFIIIC and TFIIIB in 4 l for 40 min.
Protein-DNA complexes for the EMSA analysis shown in Fig. 3A were formed under the conditions used for transcription, except that 100 ng of pLNG56 (which differs at 2 bp within box B to knock out its function) replaced pTZ2 and 3 fmol of a labeled 250-bp pTZ2-derived probe was included. Following 60 min of incubation, EMSA was per-formed on a 4% polyacrylamide gel as described (22). Two-dimensional methidiumpropyl-EDTA (MPE)-Fe(II) footprinting was performed as follows. Protein-DNA complexes were formed for 60 min in 20 l of reaction buffer containing 100 ng of pLNG56 as nonspecific competitor, 200 fmol of TBP, 160 fmol of Brf1, 100 fmol of Bdp1, 200 ng of Nhp6a, and 75 fmol of TFIIIC (as indicated in Fig. 5) with 6 fmol of the 250-bp SUP4 probe (12 fmol for naked DNA and Nhp6a only). Partial DNA cleavage was achieved by sequential addition of 1.2 l of 23 mM sodium ascorbate and 1.2 l of 46 M MPE-Fe(II) (34 M for naked DNA and Nhp6a alone) for 2 min and quenched by adding 2 l of 60 M bathophenanthroline. ⅐ OH-treated complexes were separated by EMSA and footprints of individual complexes were resolved by denaturing gel electrophoresis of the extracted DNA as described previously (20).
5Ј-End Mapping of Start Sites in Vivo-RNA was isolated from exponentially growing S. cerevisiae strains DY150 (reference type) and DY2381 (nhp6⌬⌬) (kindly provided by D. Stillman, University of Utah (23)) using the guanidinium thiocyanate-hot phenol extraction method (24) and an SDS-phenol extraction method (25). Primer extension, with 3 or 6 g of total RNA, was performed as described (19).

RESULTS
Because the crude Bdp1 fraction (26,27) containing the start site selectivity factor also contains significant levels of DNA-binding proteins (13), we first examined whether a nonspecific DNA-binding protein might impose greater stringency on TFIIIC-dependent placement of TFIIIB. The HMG1-related Nhp6a was chosen to mimic the effect of nonspecific DNA-binding proteins present in crude Bdp1 fractions. Nhp6a and the closely related Nhp6b play a positive role in pol III transcription of SNR6 (U6 snRNA gene) by a proposed compacting of DNA between the grossly suboptimally spaced box A and box B TFIIIC binding sites of that gene. Nhp6a and Nhp6b were not found to play a positive role in the transcription of any tRNA gene tested (28 -30). However, under conditions of limiting pol III with recombinant TFIIIB and highly purified TFIIIC, Nhp6a and Nhp6b were seen to stimulate tRNA gene transcription by shielding partial TATA boxes that otherwise lead to TFIIIC-independent, TFIIIB-DNA complex formation, depleting the availability of pol III for specific transcription and increasing nonspecific transcription (31). A bacterial HU protein, which like Nhp6 binds to DNA non-specifically and bends DNA sharply, also stimulated tRNA gene transcription under these conditions. 4 Nhp6a Promotes Initiation Fidelity- Fig. 1 examines the effect of Nhp6a on start site selection on the SUP4 tRNA gene. In A, TFIIIB, with or without Nhp6a, was added to TFIIIC-DNA complexes to form TFIIIB-TFIIIC-DNA complexes; pol III and NTPs were subsequently added under conditions that limit transcription to a single round (see "Experimental Procedures"). Lanes 1 and 3 compare SUP4 transcription with a crude pol III transcription fraction (BR␣ (22)) and with purified TFIIIC, pol III, and recombinant TFIIIB, respectively. BR␣ generated a relatively sharp SUP4 transcript band upon denaturing gel electrophoresis, reflecting initiation occurring almost exclusively at bp ϩ1 with a trace of initiation at bp ϩ4 (primer extension 5Ј-end mapping in B, lane 1; profile of this lane in C). In contrast, the SUP4 transcript generated with purified components was diffuse (A, lane 3), reflecting high levels of initiation at bp ϩ1, ϩ4, and ϩ8 and weaker initiation at bp ϩ12 (B, lane 2; profile in C). This pattern reflects initiation at every YR motif preceding the mature tRNA 5Ј-end-specifying sequence of SUP4: CAA-CAATTAAATA (bp Ϫ1 to ϩ12). The diffuse nature of the SUP4 transcript band in lane 3 of A arises from heterogeneity of 5Ј-ends and heterogeneous termination within the run of 7 Ts that comprises the SUP4 terminator. Addition of Nhp6a together with TFIIIB to TFIIIC-DNA complexes substantially sharpened the band of SUP4 transcripts (A, compare lanes 4 -6 with lane 3), resulting from suppression of ini-tiation at bp ϩ4, ϩ8, and ϩ12 (B, lanes 3 and 4; C profile). The degree of fidelity of initiation with 50 ng of Nhp6a was comparable with that of BR␣ for bp ϩ4 initiation with only a trace residuum of bp ϩ8 and bp ϩ12 initiation retained (C, compare left and right profiles). Suppression of downstream initiation also occurred under conditions of multiple round transcription (data not shown), and the closely related Nhp6b likewise suppressed downstream initiation (D). Nhp6a and Nhp6b also reduced nonspecific transcription as noted previously (31).
Although transcriptional initiation downstream of bp ϩ1 with purified components probably results from TFIIIB complexes that are misplaced downstream of their canonical site, it remains conceivable that Nhp6a suppresses downstream initiation at a step subsequent to TFIIIB-DNA complex formation (e.g. by shielding the more confined space between downstream-shifted TFIIIB and the box A-bound A domain of TFIIIC from access by pol III). This possibility was addressed in the experiment shown in Fig. 2. When TFIIIC was allowed to bind to DNA first, followed by the simultaneous addition of TFIIIB and Nhp6a, suppression of downstream initiation was observed (compare lanes 2-4 with lane 1). When Nhp6a was allowed to bind to DNA first, followed by the simultaneous addition of TFIIIC and TFIIIB, downstream initiation was also suppressed (compare lanes 10 -12 with lane 9), but this did not happen when Nhp6a was added to preformed TFIIIB-TFIIIC-DNA complexes (compare lanes 6 -8 with lane 5). We conclude that Nhp6a must be present as the TFIIIB-DNA complex forms to suppress downstream initiation.
Restoration of Transcriptional Initiation Fidelity Is Specific to Nhp6a and Nhp6b-We originally speculated that any nonspecific DNA-binding protein might promote start site fidelity and, accordingly, analyzed the effects of four other nonspecific DNA-binding proteins: human HMG1 and yeast HMO1, which are HMG B class proteins like Nhp6a (32), and bacterial HU proteins from T. maritima and H. pylori, which, like HMG B class proteins, are non-sequence specific and bend DNA. Nhp6a can, in fact, partially substitute for the function of HU in Escherichia coli nucleioid compaction and chromosome segregation (33). Of these proteins, only HMO1 and T. maritima HU could be verified by EMSA to associate stably with the SUP4 gene under the conditions used for transcription (Fig. 3A). HMO1 did not substitute for Nhp6a in restoring start site fidelity (compare lanes 6 and 7 with lanes 3 and 4 in the primer extension analysis shown in Fig. 3B). HU from T. maritima also had no effect on start site selection (Fig. 1B, lane 5); doubling the concentration of HU greatly inhibited transcription but the residuum retained the same relative abundance of downstream initiation events (data not shown).  1 and 2) and with recombinant or purified components (lanes 3-6) was performed as described under "Experimental Procedures," in the absence (lanes 1 and 3) or presence (lanes 2 and 4 -6) of 50 -200 ng of Nhp6a as specified above each lane. The band corresponding to initiation at bp ϩ1 and the smear corresponding to initiation at bp ϩ4, ϩ8, and ϩ12 are identified at the left (along with a radioactive DNA recovery marker (r.m.)). The asterisk at the right identifies a TFIIIC-independent transcript derived from shared pGEM and pUC vector sequence. B, primer extension mapping of start sites with reverse transcriptase. Transcription was performed as described for A with unlabeled NTPs. The presence and quantity of Nhp6a or T. maritima HU is indicated above each lane. The locations of 5Ј-ends corresponding to the observed reverse transcription products are identified on the left and are based on dideoxy sequencing ladders derived from the same primer and DNA template on the same gel that are not shown. C, densitometric profiles of B, lanes 1 (left), 2 (middle), and 3 (right). D, single-round transcription of the SUP4 tRNA gene with recombinant or purified components was performed as described for A in the absence (lane 1) or presence of Nhp6a (lane 2) and Nhp6b (lanes 3-5) in amounts specified above each lane.  1-4), TFIIIB-TFIIIC-DNA ( lanes 5-8), or Nhp6-DNA complexes (lanes 10 -12) were allowed to form prior to addition of Nhp6a (lanes 2-4) and TFIIIB (lanes 1-4), Nhp6a (lanes 6 -8), or TFIIIC and TFIIIB (lanes 9 -12), respectively. Pol III and NTPs were subsequently added to generate a single round of transcription as described under "Experimental Procedures." The amount of Nhp6a added is shown above each lane. The band of transcripts initiating at ϩ1 and the smear generated by transcripts initiating ϩ4/ϩ8/ϩ12 are identified at the left. MARCH 17, 2006 • VOLUME 281 • NUMBER 11

Nhp6a and Nhp6b Are Required for Initiation Site Fidelity in Vivo-
The inability of nonspecific DNA-binding proteins similar to Nhp6a to promote initiation fidelity prompted us to examine whether Nhp6a has a role in start site selection in vivo. For this purpose, RNA was purified from logarithmically growing yeast deleted for the genes encoding Nhp6a and Nhp6b and from the otherwise isogenic strain. RNA 5Ј-ends were mapped by primer extension with reverse transcriptase, using a primer complementary to the intron of the wild type allele of SUP4 (tRNA-Y JR). The non-coding 5Ј-leaders of tRNA gene transcripts are rapidly processed, but unspliced intron-containing tRNAs are greatly enriched for pre-tRNAs retaining the 5Ј-leader. This intron-specific primer is complementary to two additional tyrosyl-tRNA genes (tRNA-Y DR and tRNA-Y OL; Fig. 4A, bottom). Relative to the position of the SUP4 bp ϩ1 reverse transcription product, YR motifs that place initiation sites within the optimal 17-20-bp spacing from box A in all three genes (bold and underlined in Fig. 4A) predict 5Ј-ends at bp Ϫ1, ϩ1, ϩ2, and ϩ3. This is what was observed with RNA from the reference type strain (lanes 2 and 3). In contrast, RNA isolated from cells lacking Nhp6a and Nhp6b displayed a relative diminution of the bp Ϫ1 and ϩ1 primer extension products and new products corresponding to initiation at bp ϩ4 and ϩ8 (lanes 4 and 5). The same outcome was observed in two additional RNA preparations using different purification procedures (data not shown).
We also examined the intron-containing tRNA Leu-3 genes with two intron-directed primers. One primer, complementary to the tRNA-L DR2 intron (with possible priming of tRNA-L LR2 also), showed that the fidelity of initiation of this tRNA gene was not dependent on Nhp6 (Fig. 4B, compare lanes 1 and 2). The second primer, complementary to the introns of 7 tRNA-L genes (AR, CL, GL1, GR2, KR, MR, and NL), showed that one or more of these tRNA genes was dependent on Nhp6a and/or Nhp6b for initiation fidelity (compare lanes 3 and 4). One of these genes (tRNA-L CL) was used as a template for in vitro transcription with recombinant TFIIIB, purified TFIIIC, and pol III or with the crude BR␣ transcription fraction: no significant difference in initiation fidelity was observed (data not shown). It is clear from these results that not every tRNA gene in yeast requires Nhp6a and/or Nhp6b. Given that tRNA Leu-3 genes have identical coding sequences (including the box A and box B promoter elements), as well as nearly identical introns and 5Јand 3Ј-non-coding transcribed sequences, sequences upstream of the start site of transcription are implicated in determining the need for Nhp6 to reinforce initiation fidelity.
Nhp6a Alters TFIIIC-Box A Interaction-The order-of-Nhp6a-addition experiment in Fig. 2 implied that Nhp6a must act prior to TFIIIB-DNA complex formation to prevent initiation of transcription downstream of bp ϩ1. Two-dimensional hydroxyl radical footprinting with MPE-Fe(II) was employed to gain insight into the process by which Nhp6a specifies stringency in TFIIIB placement (Fig. 5). Naked DNA or protein-DNA complexes containing combinations of Nhp6a, TFIIIC, and TFIIIB were treated with MPE-Fe(II), and the complex of interest was isolated by EMSA (A, inset), followed by denaturing gel electrophoresis. A compares the MPE-Fe(II) digestion profiles of a 250-bp SUP4 tRNA gene probe, 5Ј-end-labeled on the non-transcribed strand, as free DNA and as the Nhp6a-DNA complex formed with 200 ng of Nhp6a (0.9 M). This concentration of Nhp6a is substantially less than required to completely cover the DNA probe (Ն7 M Nhp6a, data not shown; Fig. 3). Weak Nhp6a-dependent protection was observed near the label-distal end of the probe (at the left in the profile shown in A) and directly upstream of the start site of transcription (ϳbp Ϫ7 to Ϫ32) in a region of alternating TA and TG residues believed to be somewhat preferred for Nhp6a binding due to inherently greater local flexibility (34). Most yeast tRNA genes have upstream flanking sequences that are rich in these motifs ((8, 12). 3 Nevertheless, we were surprised to detect little protection in gel-isolated Nhp6a-DNA complexes under single-hit conditions and no requirement for additional MPE-Fe(II) to achieve digestion levels comparable with naked DNA. It is possible that the equilibrium between bound and unbound DNA states favors the latter  2-7, respectively). Reverse transcription products corresponding to initiation at bp ϩ1, ϩ4, ϩ8, and ϩ12 are identified at the left.  1 and 2), or a primer specific to the introns of seven additional tRNA Leu-3 genes (lanes 3 and 4). The 5Ј-leader sequences of these genes, aligned with respect to their primer extension product size with the start site of tRNA-L CL designated as bp ϩ1, are shown below the figure and annotated as described for A. The sequence ladder of tRNA-L CL (data not shown) was used to identify the 5Ј-ends designated at the right. under transcription and footprinting conditions but the former at the low salt concentrations of EMSA. Weak protection by Nhp6a was also observed upstream of the start site when DNase I was used as the cutting agent (data not shown).
Nhp6a generated two effects on the interaction of the TFIIIC A domain with box A (Fig. 5B). First, it stabilized the interaction of TFIIIC with box A as indicated by increased protection from MPE-Fe(II) cleavage. Second, it generated an upstream extension of the box A footprint from bp ϩ15 to ϩ10. This extended protection may reflect a conformational change in the A domain of TFIIIC or may reflect a TFIIIC-dependent and site-specific interaction of Nhp6a with DNA upstream of box A. A single Nhp6a molecule is expected to shield ϳ10 bp of the DNA minor groove from MPE-Fe(II) access (34,35).
The effect of Nhp6a on TFIIIB assembly is shown in Fig. 5, C and D. As observed previously (18), addition of TFIIIB to the TFIIIC-DNA complex increased protection of box A (C, compare with B) and generated protection upstream of the start site of transcription attributable to bound TFIIIB. This footprint extends from approximately bp Ϫ7 to bp Ϫ38. The addition of Nhp6a reproducibly generated a decrease in TFIIIB protection between bp Ϫ7 and approximately bp Ϫ15 with increased protection between approximately bp Ϫ24 and bp Ϫ38 (C). (In two repetitions of this footprinting analysis, interpretation of the inside borders of these two regions of decreased/increased protection varied by Ϯ2-3 bp, and the degree in the loss of protection between bp Ϫ7 and Ϫ15 also varied, but the result as just stated was uniformly observed.) For D, TFIIIC-TFIIIB-DNA complexes formed in the presence or the absence of Nhp6a were treated with heparin prior to the addition of MPE-Fe(II). Heparin strips TFIIIC from DNA leaving only TFIIIB (36,37). Heparin also strips the bulk of Nhp6a from DNA (data not shown). Protection by TFIIIB was decreased between bp Ϫ4 and approximately bp Ϫ22 and was increased between approximately bp Ϫ30 and bp Ϫ38 in complexes formed in the presence of Nhp6a. Overall, Nhp6a generated a compaction of the TFIIIB footprint by decreasing the already weak protection at the start site-proximal end of the footprint without introducing additional protected DNA residues upstream of bp Ϫ38. This is consistent with the following interpretation: the TFIIIB footprint of the promoter complex formed in the absence of Nhp6a represents an ensemble of heterogeneously positioned TFIIIB complexes; Nhp6a limits TFIIIB to its most upstream site that specifies initiation at bp ϩ1.

DISCUSSION
The impetus for examining whether Nhp6a might suppress incorrect initiation downstream of bp ϩ1 on the SUP4 tRNA gene came from the observation that Nhp6a suppresses TFIIIC-independent formation of TFIIIB-DNA complexes at partial TATA boxes (31). We reasoned that the initiation events downstream of bp ϩ1 with recombinant TFIIIB and highly purified TFIIIC and pol III might result from interaction of the A subdomain with overlapping box A variants with residence times sufficient to nucleate TFIIIB assembly at correspondingly shifted positions. It was supposed that, if a nonspecific DNA-binding protein like Nhp6a suppresses TFIIIB binding at partial TATA boxes, it might also suppress TFIIIC interaction with partial box A sequences. In this regard, 17 of the 46 distinct mature tRNA sequences encoded in yeast contain downstream-overlapping, variant box A sequences that deviate from consensus (TRGYBYAR⅐YGGY; the ⅐ denotes an additional base pair in some tRNA genes; the overlap is underlined) at 2 or 3 positions (the SUP4 gene has a variant box A that deviates at two positions). Among the limited number of 2-3-bp substitutions within box A that have been analyzed in vitro, none decreased transcription by more than 60% (38 -40).
We have shown that Nhp6a does restore fidelity of initiation site selection to transcription with highly purified TFIIIC and pol III and recombinant TFIIIB (Fig. 1) and that this effect requires the intervention of Nhp6a prior to the formation of TFIIIB-DNA complexes (Fig. 2). Nhp6b also restores initiation fidelity (Fig. 1D). However, our original notion that Nhp6a might function solely as a nonspecific DNA-binding protein in suppressing TFIIIC interaction with variant, overlapping, downstream box A sequences appears to be invalid. 1) Enforcement of initiation fidelity is not strictly due to nonspecific binding to and bending of DNA by Nhp6, since two bacterial HU proteins and two other HMG B class proteins that bind DNA and bend it similarly (reviewed in Ref. 32) fail to restore initiation fidelity (Figs. 1B and 3 and data not shown). Instead, Nhp6 is a specific pol III initiation fidelity factor, since cells that lack both Nhp6a and Nhp6b lose initiation fidelity at some tRNA genes (Fig. 4). 2) Some, but not all, tRNA Leu-3 genes require Nhp6 for initiation fidelity despite the fact that they all contain identical, overlapping, downstream box A variants. The two tRNA Leu-3 genes that are known not to require Nhp6 for initiation fidelity (tRNA-L DR2 and CL) have extended AT-rich upstream sequences, similar to SUP4, that should be amenable to heterogeneous placement of TFIIIB. 3) If Nhp6a functions by disfavoring the binding of TFIIIC to overlapping downstream box A variants, one should observe a tightening of the TFIIICbox A footprint, diminishing or removing protection at the downstream end and enhancing already existing protection at the upstream end. Instead, the presence of Nhp6a extends the box A footprint in the upstream direction without observable loss of protection at the downstream end (Fig. 5, B and C).
As noted above, the sequence upstream of the start site of transcription appears to determine whether a tRNA gene requires Nhp6 for initiation fidelity. The presence of Nhp6a results in a narrowing of the TFIIIB footprint (Fig. 5) that is consistent with disfavoring the downstream placement of TFIIIB by TFIIIC. The resulting narrowed TFIIIB footprint more closely matches the MPE-Fe(II) footprint of TFIIIB on SNR6 (bp Ϫ12/Ϫ14 to bp Ϫ40 (20, 31), which does not require Nhp6a for initiation fidelity in vitro. Either Nhp6a affects TFIIIB placement directly (e.g. by limiting the DNA sites at which the TFIIIB complex can assemble), or it does so through TFIIIC (e.g. by limiting the conformational flexibility of Tfc4 that allows alternative sites of TFIIIB placement (10)). MPE-Fe(II) footprinting (Fig. 5) favors the latter alternative; although Nhp6a alone preferentially binds upstream of the start site of transcription (A), it does not preferentially shield DNA downstream of the normal site of TBP interaction with the crude BR␣ fraction (bp Ϫ30 to Ϫ23 (41)) that specifies initiation at bp ϩ1; rather, Nhp6a enhances the TFIIIC-dependent protection of box A and expands the footprint in the upstream direction from bp ϩ15 to ϩ10. This upstream extension of the box A footprint may reflect an Nhp6a-dependent alteration in TFIIIC-box A interaction or a TFIIIC-stabilized binding of Nhp6a to DNA directly upstream of box A. In this connection, an interaction between Nhp6b and TFIIIC has been inferred by TAP tag affinity purification of yeast protein complexes (42), but two hybrid genetic analyses failed to detect an interaction between Nhp6a or Nhp6b with any subunit of TFIIIC and TFIIIB (28). We have explored whether the enhanced protection and upstream extension of the box A footprint results from a TFIIIC-stabilized binding of Nhp6a to DNA upstream of box A by sitespecific protein-DNA cross-linking with a photoreactive thymidylate analog (36) positioned at bp ϩ14 and ϩ16 on the non-transcribed strand of SUP4. Rather than enhance the cross-linking of limiting Nhp6a to DNA, the presence of TFIIIC diminished Nhp6a cross-linking to bp ϩ14/ϩ16 (data not shown). Either Nhp6a is not directly responsible for the upstream extension of the box A footprint or the N-terminal segment of Nhp6a that most likely cross-links to the photoreactive sidechain extruding from the major groove (35) may be bound to, or displaced by, TFIIIC.
As already noted, the temperature sensitivity of yeast lacking Nhp6 stems from a defect in SNR6 (U6 snRNA) transcription, whereas the other pol III-transcribed genes that were examined are not affected (28 -30). However, it is not clear that downstream initiation of transcription of some tRNA genes would compound the defect, since truncation of the tRNA Tyr 5Ј-leader to 2 bases upstream of the site of RNase P processing has no effect on the catalytic cleavage rate in vitro (43), and 5Ј-leader processing typically precedes 3Ј-end processing (44,45).
The SUP4 start site selectivity factor was originally characterized as a component present in a crude Bdp1 fraction (13). In view of its requirement for initiation fidelity in vivo (Fig. 4), it is likely that Nhp6 is the selectivity factor present in this fraction. However, it is also possible that crude Bdp1 fractions contain additional factors that affect initiation or that are capable of substituting for Nhp6a and Nhp6b in promoting initiation fidelity in vitro. An additional factor of unknown composition, TFIIIE, is also found in yeast-derived Bdp1 fractions (46). This factor greatly stimulates SUP4 transcription when fully recombinant TFIIIB is used (47). Similarly, under conditions of limiting pol III, Nhp6a also substantially stimulates tRNA gene transcription with a concomitant reduction of nonspecific transcription (31). In summary, we conclude that Nhp6 is the start site selectivity factor and note the possibility that Nhp6 may also be a component of TFIIIE.