The Fundamental Ribosomal RNA Transcription Initiation Factor-IB (TIF-IB, SL1, Factor D) Binds to the rRNA Core Promoter Primarily by Minor Groove Contacts*

Acanthamoeba castellaniitranscription initiation factor-IB (TIF-IB) is the TATA-binding protein-containing transcription factor that binds the rRNA promoter to form the committed complex. Minor groove-specific drugs inhibit TIF-IB binding, with higher concentrations needed to disrupt preformed complexes because of drug exclusion by bound TIF-IB. TIF-IB/DNA interactions were mapped by hydroxyl radical and uranyl nitrate footprinting. TIF-IB contacts four minor grooves in its binding site. TIF-IB and DNA wrap around each other in a right-handed superhelix of high pitch, so the upstream and downstream contacts are on opposite faces of the helix. Dimethyl sulfate protection assays revealed limited contact with a few guanines in the major groove. This detailed analysis suggests significant DNA conformation dependence of the interaction.

Acanthamoeba castellanii transcription initiation factor-IB (TIF-IB) is the TATA-binding protein-containing transcription factor that binds the rRNA promoter to form the committed complex. Minor groove-specific drugs inhibit TIF-IB binding, with higher concentrations needed to disrupt preformed complexes because of drug exclusion by bound TIF-IB. TIF-IB/DNA interactions were mapped by hydroxyl radical and uranyl nitrate footprinting. TIF-IB contacts four minor grooves in its binding site. TIF-IB and DNA wrap around each other in a right-handed superhelix of high pitch, so the upstream and downstream contacts are on opposite faces of the helix. Dimethyl sulfate protection assays revealed limited contact with a few guanines in the major groove. This detailed analysis suggests significant DNA conformation dependence of the interaction.
Extensive in vitro and in vivo analyses have shown that efficient transcription of the rRNA gene by RNA polymerase I requires formation of a stable transcription factor-promoter DNA complex, the committed complex (for review, see Refs. 1 and 2). In most species, two regions of the rRNA promoter, the core promoter element (CPE) 1 and upstream promoter element (UPE), participate in expression of the rRNA gene and formation of the committed complex. The CPE is located between Ϫ50 and ϩ10 relative to the transcription start site (position ϩ1), is absolutely required in all organisms, and is sufficient in some for initiation. It is the primary binding site for the TBPcontaining transcription factor. The UPE is located farther upstream, from approximately Ϫ150 to Ϫ110, and acts primarily to stimulate transcription. It has a variable requirement between species; at one end of the spectrum, Acanthamoeba castellanii has no UPE discernible in vitro (3). We (4) and others (5) have shown recently that the CPE can be subdivided into an element that interacts intimately with TIF-IB and an element that functions like the initiator element (Inr) of some RNA polymerase II promoters, interacting with a specific TAF I .
The UPE and CPE are important because they serve as the binding sites for two RNA polymerase I transcription factors, UBF and TIF-IB (also known as SL1, factor D, Rib1, or corebinding factor) respectively. These two proteins interact, by mechanisms that are still unclear but apparently involve DNA looping (6), to form a stable initiation complex capable of directing specific RNA polymerase I recruitment through multiple rounds of transcription. The need for UBF is variable between species; UBF is required in human and Xenopus (7)(8)(9), but is dispensable in rat (10), mouse (11), yeast (12)(13)(14), and A. castellanii (15). Therefore, TIF-IB is, by elimination, the fundamental transcription factor for rRNA genes, recruiting RNA polymerase I in the next step of the initiation process (16).
TIF-IB has recently been purified to homogeneity from several eukaryotic species, including human (17), mouse (18), yeast (12,14), and A. castellanii. 2 Biochemical analysis has shown that TIF-IB consists of TBP and several TAF I factors. The RNA polymerase I TAFs vary in size and number depending on the species from which they were isolated. For example, human TAF I factors have molecular masses of 110, 63, and 48 kDa, whereas yeast TAF I factors are 102, 66, and 60 kDa. SDS-polyacrylamide gel electrophoresis analysis has revealed that homogeneous A. castellanii TIF-IB consists of TBP and four TAF I factors of 145, 99, 96, and 91 kDa, all of which specifically cross-link to various regions in the promoter (19). Curiously, this is one more TAF I than in yeast or vertebrates. The genes for the human, mouse, and yeast TAF I factors have been cloned (12,14,20,21). Unexpectedly, the lower and higher eukaryotic TAF I factors show no significant homology at either the nucleotide or amino acid level. This lack of similarity of RNA polymerase I TAF I factors from distantly related species is consistent with the poor conservation of rRNA promoter primary sequence and their species-specific behavior (22), mediated mainly through these fundamental transcription factors (7,9,23,24). However, rRNA promoters do exhibit a common variation from canonical B-DNA in groove width, base tilt, and axis linearity along their length (25), suggesting that promoter recognition could involve interactions other than primary sequence. Indeed, UBF does not bind a common primary sequence, but instead recognizes DNA structure (8,26).
Little is known about how the fundamental transcription factors for RNA polymerase I interact with DNA. In most cases, the affinity of TIF-IB alone for the CPE is too low to carry out detailed in vitro binding experiments. This is not true for A. castellanii TIF-IB, which binds the promoter with a K d of ϳ50 pM without the aid of additional proteins. 3 Therefore, we have taken advantage of this unique opportunity to investigate the interactions of TIF-IB with the wild-type rRNA promoter.
Several observations led us to hypothesize that TIF-IB might bind in a relatively sequence-independent manner, presumably by interaction with the minor groove. A. castellanii TIF-IB binds the rRNA core promoter from Ϫ67 to Ϫ17, as shown by previous DNase I and methidiumpropyl-EDTA⅐Fe(II) footprints (27,28). However, deletion and point mutagenesis (29,30) showed that sequences upstream of about Ϫ45 could be altered without measurably affecting binding, and 5Ј-deletions to Ϫ32 still allowed robust transcription. Surprisingly, 5Ј-deletions all the way to Ϫ6 allowed weak transcription with correct start site selection (4). Site-specific photocross-linking (19) revealed that TBP and all four TAF I factors were in close proximity to the DNA in the region upstream of Ϫ38, whereas only TAF I 96 interacted downstream of this position. Substitution of thymine residues at various locations throughout the TIF-IB binding site with the uracil derivative 5-[N-(p-azidobenzoyl)-3-aminoallyl]deoxyuracil, which protrudes markedly into the major groove, has little effect on TIF-IB binding (31). Preliminary dimethyl sulfate protection footprinting assays showed very limited protection of guanines in the major groove. 4 Unfortunately, none of these experiments were sensitive enough to determine the precise interactions made between TIF-IB and the DNA, such as which groove of the DNA is being contacted.
We show here using minor groove-binding drugs and discriminating chemical footprinting that TIF-IB recognizes the A. castellanii rRNA promoter primarily through the minor groove, not the major groove as is the case with most DNAbinding proteins. This minor groove binding occurs by TIF-IB recognizing either sequence-specific variations in the architecture of the minor groove backbone(s) or functional groups on the base pairs in the minor groove, and our data cannot distinguish between these two possibilities. This is the first detailed description of the DNA-binding mechanism for TIF-IB from any species and makes plausible the notion that these fundamental transcription factors may recognize their binding sites by the variation in DNA intrinsic structure, such as minor groove width, and not (exclusively) by primary sequence. This notion is supported by the recent finding of conserved intrinsic structures in a variety of eukaryotic rRNA promoters, despite their disparate primary sequences (25).

MATERIALS AND METHODS
Plasmids and Templates-A fragment of the A. castellanii core promoter (Ϫ96 to ϩ18, relative to the transcription start site, position ϩ1) was removed from pEBH10 (27) by partial digestion with EagI and isolated on a 3.5% Metaphor agarose gel (FMC Corp. BioProducts, Rockland, ME). This 114-base pair fragment was subcloned into the NotI site of pBluescript II SK Ϫ (Stratagene, La Jolla, CA). Two clones were identified by sequencing: plasmid pGG4C, with Ϫ96 oriented toward the SacI site, was used to label the template strand, and plasmid pGG17C, with Ϫ96 near the XbaI site, was used to label the RNA-like strand for footprinting.
The templates used in TIF-IB binding assays were made by digestion of pGG4C and pGG17C with BamHI, treatment with shrimp alkaline phosphatase (Amersham Corp.), and digestion with SacI to generate a fragment of 150 base pairs. Fragments were separated from linear plasmid on a 1% agarose gel, visualized by UV shadowing (to minimize nicking), and eluted with a Schleicher & Schuell Elutrap apparatus for 2 h in 1 ϫ TAE (32). The eluate was phenol/CHCl 3 -extracted and precipitated, and the DNA concentration was determined by absorbance at 260 nm. Both strands (100 ng) were 5Ј-end-labeled with T4 polynucleotide kinase (Amersham Corp.), purified by electrophoresis on a 5% native polyacrylamide gel, and eluted as described (19). For electrophoretic mobility shift assays (EMSAs), the labeled template strand from pGG4C was purified with a QIAEX kit (QIAGEN Inc., Chatsworth, CA) using the manufacturer's desalting protocol. Before purification, aliquots of labeled DNA were spotted on Whatman DE81 filter discs; washed five times with 5% Na 2 HPO 4 , H 2 O, ethanol, and ether; and counted in a liquid scintillation counter to determine specific activity.
Protein Preparations and Purification-TIF-IB was isolated from nuclear extracts (33,34) and purified through one round of promoter affinity chromatography as described (35). For some footprinting experiments, promoter affinity-purified TIF-IB was concentrated with a Microcon 10 microconcentrator (Amicon, Inc., Beverly, MA) as described in the manufacturer's protocol. Cloned and expressed A. castellanii TBP and recombinant human CREB were purified as described (36,37). For EMSA experiments, CREB was diluted 1:2000 in 50 mM Tris-HCl, pH 7.9, 100 mM KCl, 12.5 mM MgCl 2 , 1 mM EDTA, and 20% glycerol to a final concentration of 225 ng/ml.
Minor Groove-binding Drug Preparations-Distamycin A and mithramycin A were obtained from Sigma. Netropsin was purchased from Boehringer Mannheim. Stock solutions of 2, 2, and 10 mM, respectively, were made in ddH 2 O and filtered through 0.2-m filters. For EMSA experiments, serial dilutions were made, in ddH 2 0, from stock solutions. All solutions were stored at Ϫ20°C. The sites of binding of the drugs to the core promoter were determined by MPE-Fe footprinting (see Fig. 3).
EMSAs and Drug Inhibition Studies-The 150-bp BamHI/SacI fragment of pGG4C was 5Ј-end-labeled and used for all EMSAs, except those with CREB. TIF-IB binding conditions were 20 mM HEPES, pH 7.9, 10 mM MgCl 2 , 0.5 mM dithiothreitol, 0.1% Nonidet P-40, 50 g/ml bovine serum albumin, 10% glycerol, 1-2 nM pBR322 (as nonspecific competitor), and 2-5 fmol of labeled core promoter DNA (ϳ25,000 cpm) in a 20-l final volume. The amount of TIF-IB required to saturate the rRNA promoter was determined by titration in the absence of drug and used in all relevant EMSA experiments (data not shown). DNA was preincubated either with a minor groove-binding drug or with 5 fmol of TIF-IB for 20 min at 25°C. The second reagent was added 20 min after the first and incubated under the same conditions. Reactions were stopped on ice and loaded on 5% nondenaturing polyacrylamide gels as described (19). The polyacrylamide gel support medium GelBond (FMC Corp. BioProducts) was used for all EMSA polyacrylamide gels. Gels were run at 200 V for 1.5 h, dried, exposed to Eastman Kodak storage phosphor screens, visualized on a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA), and quantified using ImageQuant software (Version 3.3, Molecular Dynamics, Inc.).
EMSAs with the minor groove-binding protein TBP were performed with 0.8 g of recombinant TBP on the rRNA core promoter (pGG4C, BamHI/SacI). The sequence of the promoter from Ϫ7 to ϩ3 contains two overlapping consensus TATA boxes (TATATATA ϩ1 AA; see Fig. 10, upper panel). Binding reactions were essentially the same as for TIF-IB, except that the KCl concentration was changed to 60 mM (38) and competitor DNA was changed to 3 nM poly(dI-dC) to prevent formation of slower migrating complexes. Acanthamoeba TBP binding to this site under these conditions was verified by MPE-Fe footprinting (data not shown). Either TBP or drug incubations were carried out at 30°C for 30 min; then the other reagent was added, and the incubations were continued for 30 min. For EMSAs of TBP, MgCl 2 was added to both the gel and the running buffer to a final concentration of 5 mM (38). Gels were run at 150 V for 1.5 h, dried, and analyzed as described above.
EMSAs with CREB were performed in a 10-l reaction containing 50 mM KCl, 25 mM Tris-HCl, pH 7.9, 6 mM MgCl 2 , 0.5 mM EDTA, and 10% glycerol. The template used for these reactions was a 5Ј-end-labeled 77-bp DNA fragment (39) containing the consensus CRE. For reactions in which drug was preincubated with DNA, 1 l of the appropriate stock solution of drug was added to 5 l of labeled CRE (ϳ1 fmol). The reaction was incubated for 30 min, and then 5 l of CREB (ϳ1.1 ng) was added to a final volume of 11 l and incubated for an additional 30 min. The concentration of drug in the first incubation was nearly double that in the final incubation. For reactions in which CREB was prebound, 5 l of CREB was added directly to 5 l of labeled CRE. After a 30-min incubation, 1 l of drug was added to a final concentration of 100 M (for controls, 1 l of ddH 2 O was added instead of drug), and the incubation was continued for 30 min as described above. Samples were electrophoresed on 5% gels and analyzed as described for TIF-IB.
Uranyl Nitrate and MPE-Fe Footprinting-The binding conditions for footprinting were the same as for EMSAs of TIF-IB, except that binding was carried out at pH 7.5 to increase the efficiency of the UO 2 (NO 3 ) 2 -DNA cleavage reaction (40). Ten fmol of TIF-IB in HEG 10 buffer (20 mM HEPES, pH 7.5, 0.1 mM EDTA, and 10% glycerol) or buffer alone was added to the reaction and incubated for 20 min at 25°C. Conditions for the UO 2 (NO 3 ) 2 cutting reactions and precipitations were modified from published procedures (41,42). The final uranyl nitrate concentration was 1 mM. The reactions were exposed to 366 nm UV light (Model UVGL-58 lamp with 6-watt bulb, UVP Inc., San Gabriel, CA) held 4 cm above an open 1.5-ml Eppendorf tube for 30 min at room temperature. Reactions were stopped with 10 mM sodium citrate. For MPE-Fe footprinting experiments at pH 7.5, 1 l of 70 M MPE and 50 M (NH 4 ) 2 Fe(SO 4 ) 2 solution and 1 l of 0.5 M dithiothreitol were added to the 20-l reactions instead of uranyl reagents. MPE-Fe reactions were stopped with a final concentration of 10 mM bathophenanthroline (43,44). After the addition of 30 l of 1 mg/ml proteinase K and 0.1% SDS solution, samples were incubated at 50°C for 30 min. The DNA solution was made 0.3 M sodium acetate, pH 4.5; precipitated with 70% ethanol; resuspended in 98% formamide (with bromphenol blue and xylene cyanol used as tracking dyes); and analyzed on 8 or 10% sequencing gels (7 M urea and 1 ϫ Tris borate/EDTA). Gels were run at 1100 -2000 V for 2-3 h, depending on the percentage of acrylamide and the DNA strand being analyzed.
Hydroxyl Radical and MPE-Fe Footprinting-The concentrations of hydroxyl radical reagents and the basic design of experiments were modified from published methods (45,46). The final binding conditions were as described for EMSAs of TIF-IB, with several important modifications. The final volume of the prebinding reaction was reduced to 3 l while keeping the final concentrations of reagents the same, except that glycerol was reduced to 1.67% and the DNA and TIF-IB concentrations were ϳ6 times higher. The low reaction volume was required to achieve efficient TIF-IB binding in the low concentration of glycerol by effectively increasing DNA and protein concentrations. The addition of either 0.5 l of TIF-IB (10 -20 fmol) or HEG 10 buffer supplied the only glycerol in the reaction. Samples were incubated at 25°C for 20 min. Three aliquots (1.5 l each) of 10 mM sodium ascorbate, 0.3% H 2 O 2 , 1 mM (NH 4 ) 2 Fe(SO 4 ) 2 , and 2 mM EDTA were suspended as droplets on the side of the tube, but not mixed. All three solutions were freshly made just before use. Then 7.5 l of HE buffer (HEG 10 without glycerol) was used to mix all three reagents with the prebinding reaction mixture and to dilute the reagents to their final concentrations of 1 mM sodium ascorbate, 0.03% H 2 O 2 , 100 M (NH 4 ) 2 Fe(SO 4 ) 2 , and 200 M EDTA in a final reaction volume of 15 l. This step lowered the glycerol concentration in the footprinting reaction to 0.33% to prevent quenching of hydroxyl radicals and increased cutting efficiency. Reactions were incubated for 3 min at room temperature and stopped by adding 15 l of 10% glycerol. MPE-Fe control footprints were performed the same way, except that 0.5 l of 70 M MPE and 50 M (NH 4 ) 2 Fe(SO 4 ) 2 and 0.5 l of 0.5 M dithiothreitol were added as droplets to the side of the tube. Eleven l of HE buffer was added to mix the reaction and to dilute the reactions to 15 l as described above. The cutting reaction was performed at 25°C for 10 min and stopped with 15 l of 20 mM bathophenanthroline. Proteinase K treatment, precipitation, resuspension, and electrophoresis of samples were done exactly as described for uranyl footprinting.
Dimethyl Sulfate (DMS) Footprinting-DMS protection assays were modified from those of Shaw and Stewart (47). Protection assays were performed under saturating TIF-IB conditions, and protein-DNA complexes were not separated by EMSA prior to cutting with piperidine. Binding conditions were the same as for EMSAs of TIF-IB. DMS was added to preformed TIF-IB-promoter complexes to a final concentration of 130 M. Reactions were incubated at room temperature for 3 min and stopped with 180 l of DMS stop buffer (350 mM sodium acetate, 220 mM ␤-mercaptoethanol, and 83 g/ml linear polyacrylamide). Samples were extracted once with an equal volume of phenol/CHCl 3 , ethanolprecipitated, washed with 70% ethanol, and dried. The pellets were resuspended in 50 l of 10% piperidine (Sigma) and heated to 95°C for 30 min. Piperidine was evaporated under vacuum, and the pellet was suspended in 30 l of ddH 2 O, dried, resuspended in 20 l of ddH 2 O, and dried again. The DNA pellets were resuspended in 6 l of 98% formamide containing tracking dyes, heated to 95°C for 4 min, and electrophoresed on 8 -10% standard sequencing gels (see above).
Data Analysis-Kodak storage phosphor screens were exposed to EMSA and footprinting gels, and the radioactivity was visualized by scanning with a PhosphorImager. For drug inhibition studies, the amount of the complexed and free DNAs was estimated using Image-Quant software (Version 3.3), and the percent complex was calculated. The relative percent complex was determined by normalizing to the amount of the shifted complex when no drug is added. The values of relative percent complex were averaged from multiple experiments. IC 50 is the amount of drug required to decrease the relative percent complex by 50%.
Footprinting gel data were scanned as described above, but were analyzed with Phoretix Photometrics 1D-full software (Version 2.51, Non-Linear Dynamics Ltd). Using this program, the volume for each band (or "peak") in the DNA alone or footprinted lane was determined. To correct for nicks in the DNA used for footprinting, the volumes of the corresponding bands in the untreated DNA lane were subtracted from the volumes of the experimental lanes. In some experiments, the overall cutting efficiency in lanes with added TIF-IB was lower than in those without protein, probably because free protein quenches the footprinting reagents. Thus, to quantitatively compare individual bands in the unprotected DNA ladder and the footprinted lane, the two lanes were normalized using bands outside the footprinted region. All bands were first paired using the "matching" function of this program. This function aligns the peaks of parallel lanes based on the relative position of the band in the gel. Matched bands well outside the footprinted region of the experimental lane and in the no-protein DNA lane were normalized to 1 and used as reference bands. The volumes for every other band within each lane were automatically normalized to the designated reference band. This procedure also normalizes the lanes of different experiments even if they were exposed for different times or used variable amounts of labeled DNA and therefore have different intensities. Next, the normalized volumes for each band from nine independent experiments were averaged. Relative protection was calculated by dividing the mean value for each individual band in lanes with added TIF-IB by the sum of the mean volumes for the corresponding bands in lanes with and without TIF-IB. Therefore, an ideal unprotected band should have a value of 0.50. Theoretically, any values below this could be considered protected, and those above, enhanced. We defined any value below 0.35 as a protected base and those above 0.65 as a base with enhanced reactivity.

TIF-IB Binding to the rRNA Promoter Is Inhibited by Minor
Groove-binding Drugs-We used the minor groove-binding antibiotics netropsin, distamycin A, and mithramycin A to investigate the interactions of TIF-IB with the rRNA core promoter. The structurally related compounds netropsin and distamycin A bind AT-rich sequences, preferring poly(dA)⅐poly(dT) binding sites of at least 4 base pairs over those of alternating poly(dA-dT)⅐poly(dA-dT) (48). Mithramycin A, a member of the aureolic acid family of antibiotics, binds the minor groove with preference for dG-dC sequences of at least 2 base pairs (48 -50). However, the binding sites for mithramycin A usually cover at least 3 base pairs and depend somewhat on flanking sequence (51). We chose this particular set of drugs to ensure that both dA-dT-and dG-dC-rich minor groove-binding regions would be occupied by at least one of the reagents (see Fig. 10 (upper panel) for the sequence of the promoter). The ability of these minor groove-binding drugs to inhibit TIF-IB complex formation on the A. castellanii rRNA promoter was determined by electrophoretic mobility shift assay.
A typical EMSA inhibition experiment in which the concentration of distamycin A was varied is shown in Fig. 1. Inhibitor was added either before (lanes 1-11) or after (lanes 12-21) the formation of the committed complex. The data demonstrate that it required less distamycin A to prevent formation of the TIF-IB complex than it did to disrupt preformed committed complexes (compare lanes 1-11 with lanes 12-21). To accurately determine the concentration of inhibitor needed to reduce the relative complex to 50% (IC 50 ), data from distamycin A EMSAs were analyzed graphically ( Fig. 2A). The IC 50 was 3 M when distamycin A was prebound. By 12.5 M, inhibition of TIF-IB binding was complete (Fig. 1, lane 7). In contrast, the amount of distamycin A required to disrupt 50% of the preformed committed complex was 5-fold higher, or 15 M ( Fig.  2A), and the complex was not totally disrupted until 50 M (Fig.  1, lane 20). This occurs because TIF-IB occupies the minor groove sites that would normally be bound by the drug, and free and bound TIF-IB are not in rapid equilibrium (29).
The same analysis was repeated for the other minor groovebinding drugs. Netropsin inhibited TIF-IB binding with an IC 50 of 8 M for reactions in which the drug was prebound (Fig. 2B). Seventy-fold more netropsin (IC 50 ϭ 560 M) was required to inhibit the same amount of preformed committed complex, consistent with direct competition for the minor groove. Unlike distamycin A, netropsin never fully inhibited TIF-IB binding to the promoter (Fig. 2B).
Since there are limited dA-dT-rich binding sequences in the rRNA promoter for netropsin or distamycin A (see Fig. 10, upper panel), mithramycin A was used to probe minor groove interactions at dG-dC-rich sites. The differential binding preferences of mithramycin A versus netropsin and distamycin A were verified by MPE-Fe footprinting and aligned with the site of TIF-IB binding (Fig. 3). Clearly, the two classes of antibiotics have different and complementary binding preferences within the footprinted region (approximately Ϫ70 to Ϫ17) (compare lanes 5 and 9 with lane 13). Fig. 3 further shows the appearance of drug footprints in the concentration range at which they inhibit TIF-IB and TBP binding. In agreement with the other antibiotics, the IC 50 was 4 M if the drug were prebound to the promoter and 17.5 M when TIF-IB was bound first (Fig.  2C). The concentrations at which TIF-IB binding was totally inhibited were essentially the same as for distamycin A (see above and Fig. 2C). Therefore, two separate classes of minor groove-binding drugs inhibit TIF-IB binding.

Complex Formation by Minor Groove-binding (but Not Major Groove-binding) Proteins Is Inhibited by Minor Groove-binding
Drugs-TBP binds the minor groove of the TATA box (52)(53)(54)(55). The binding of recombinant A. castellanii TBP to a consensus TATA box was examined as a control for inhibition of minor groove binding. To keep the reaction conditions as close as possible to those of the other experiments, we used the TATA box consensus sequences located from Ϫ7 to ϩ3 of the rRNA promoter (see Fig. 10, upper panel). Inhibition of TBP binding was complete with 0.4 M netropsin (Fig. 4, lane 3) and 0.8 M distamycin A (lane 11). The graphically determined IC 50 values for netropsin and distamycin A were 0.3 and 0.6 M, respectively. These values are fairly close to published values determined using TATAAA-containing oligonucleotides (56,57). When the drug concentration was high enough to saturate the specific TBP binding site, a band of higher mobility was seen (lanes 6 and 13). We do not have an explanation for this, but this complex was not quantified as a legitimate TBP-promoter complex in the analysis. In contrast to TIF-IB, TBP prebinding  to the TATA site did not result in a measurable increase in the amount of netropsin or distamycin A required for inhibition (Fig. 5, lanes 1-14) (see "Discussion"). This is also consistent with published results and probably reflects the relatively weak interaction of TBP in the absence of accessory proteins (see "Discussion") (56). Under our conditions, it is not possible to saturate this DNA with TBP without formation of slower migrating complexes containing multiple TBPs (data not shown). Therefore, unlike the EMSAs with TIF-IB, all TBP experiments were done with subsaturating amounts of TBP, which affects the IC 50 . The IC 50 values for the dA-dT-binding drugs on a TBP-promoter complex were expected to be low relative to those for TIF-IB because both the drugs and TBP directly compete for the exact same binding site (verified by MPE-Fe footprinting of both the drugs and TBP (data not shown)), whereas TIF-IB binds an extended site relative to the drugs.
Mithramycin A, whose preferred binding site is dG-dC-rich, did not inhibit TBP binding to a TATA box as efficiently as the dA-dT-binding drugs. When DNA was preincubated with mithramycin A, the IC 50 was 4 M, 10-fold higher than that with netropsin and distamycin A with TBP (Fig. 4, lanes 15-22). Mithramycin A added to the preformed TBP-promoter complex did not inhibit, even at 3.1 M (Fig. 5, lanes 15-20). This emphasizes the importance of using the appropriate drugs for the binding sites that one wishes to examine. The fact that the binding sites of TBP and mithramycin A do overlap slightly has been verified by MPE-Fe footprinting of the drug (Fig. 3) and protein (Ϫ6 to ϩ2, data not shown).
CREB, a member of the leucine zipper family of transcription factors, binds the major groove (for review, see Ref. 58). The ability of minor groove-binding drugs to inhibit CREB binding to its consensus binding site (the CRE) was tested as an expected negative control. As expected, none of the drugs in concentrations up to 100 M inhibited CREB binding either in preformed factor-DNA complexes or when the drug was prebound to the DNA (data not shown).
The results of the drug inhibition studies of TIF-IB, TBP, and CREB support a minor groove-binding mechanism for TIF-IB. However, drug inhibition analysis of protein/DNA interactions does have potential weaknesses when used as the sole criterion for groove selection. For instance, a few major groovebinding proteins have been shown to be inhibited by minor groove-binding drugs, possibly because the drugs alter the structure of DNA upon binding (see "Discussion") (59). For this reason, we chose to evaluate minor and major groove interactions more directly by using chemical footprinting techniques.
Phosphate and Deoxyribose Contacts Are Consistent with TIF-IB Interacting with the Minor Groove-Minor or major  4. Inhibition of TBP-TATA box binding by prebound minor groove-binding drugs. Minor groove-binding drugs were preincubated with the promoter at the concentrations indicated for 30 min, 0.8 g of TBP was added, and incubation was continued for an additional 30 min .  Lanes 1-7, 8 -14, and 15-22 are a titration of netropsin, distamycin A, and mithramycin A, respectively. The no-protein lanes are 7, 14, and 22. The open arrow marks a complex seen when drug concentrations are relatively high. This may result from TBP binding to sites other than the TATA box (see "Results"). groove contacts can be distinguished using footprinting techniques that detect interactions with the phosphodiester backbone. Phosphate contacts can be analyzed by either uranyl nitrate protection or ethylnitrosourea interference footprinting. The uranyl ion complexes with the phosphodiester backbone, probably by bridging phosphate groups across the minor groove (40). Exposure to long UV light, between 300 and 420 nm, excites the uranyl complex, ultimately leading to cleavage of the DNA by oxidation of the adjacent sugar ring. Cutting only occurs when uranyl is bound to DNA (40,60). Phosphate groups bound by protein are not able to complex with the uranyl ion and appear as protected regions. We chose this technique over ethylation interference with ethylnitrosourea because the latter technique requires premodification of one phosphate group per DNA molecule to completely inhibit binding of the protein. The large size of the TIF-IB binding site (50 bp) and data from point mutation studies (30) suggest that modification of one phosphate group would be insufficient to inhibit binding.
Uranyl footprints of TIF-IB on the template and RNA-like strands are shown in lanes 1-5 of Fig. 6 and 7, respectively. Four regions are protected from uranyl on the template strand, centered around Ϫ19/Ϫ20, Ϫ31, Ϫ45/Ϫ46, and Ϫ56, and there is a significantly enhanced site at Ϫ26 (Fig. 6, compare lanes 4  and 5 with lane 3). Footprints on the RNA-like strand revealed three strongly protected regions centered around Ϫ52, Ϫ39, and Ϫ27/Ϫ28 and a weak protection centered at Ϫ63 (Fig. 7). In some experiments, a few bases around Ϫ22 of this strand were also protected.
The data from nine individual footprints similar to those shown in Figs. 6 and 7 were quantified as described under "Materials and Methods." Briefly, the intensity of lanes with and without TIF-IB was normalized by setting equal the intensity of a band outside the footprint. A relative protection value for each band was calculated as the PhosphorImager volume of the band in the footprinted lane divided by the sum of the volumes for the bands in the lanes with and without TIF-IB. Thus, unprotected bands have a relative protection value of 0.5; protected bands have a value below 0.5; and enhanced bands have a value above 0.5. After such normalization, relative protection values from different experiments can be averaged. Average relative protection was plotted versus the base pair position in Fig. 8. Bands in the upper regions of the gels could not always be resolved and were not quantified. Consequently, the number of data points for those bases in the nonresolved portion of the gel was Ͻ9. We arbitrarily defined any relative protection value below 0.35 as a protected base and those above 0.65 as enhanced (i.e. changes in intensity Ͼ30%).
The quantitative data shown in Fig. 8 corroborate what was seen visually in Figs. 6 and 7. The four protected regions on the template strand (lower panels) are depicted by the four minima of the graph (compare Fig. 8A (lower panel) with Fig. 6 (lane 4)).
The enhanced band at Ϫ26 was also qualitatively and quantitatively reproducible. Also detectable by our criteria were iso-FIG. 6. Footprinting of TIF-IB on the template strand of the rRNA core promoter. Uranyl nitrate, hydroxyl radical, and MPE-Fe footprinting reactions were performed on ϳ5 fmol of the BamHI/SacI fragment of pGG4C labeled on the template strand as described under "Materials and Methods." All three sets of reactions were performed in parallel and run on the same sequencing gel. To visualize the footprint for each cutting method, the limits for each set were adjusted with ImageQuant software. Footprinting reactions are as follows: lanes 1-5, uranyl nitrate; lanes 6 -9, hydroxyl radical (HR), and lanes 10 -12, MPE. Lane m is a G ϩ A sequencing ladder. Uncut DNA is shown in lane 1 (without TIF-IB) and lanes 2 and 6 (with TIF-IB). Lanes 3, 7, and 10 are the DNA without TIF-IB cut with uranyl, hydroxyl radical, and MPE-Fe, respectively. The footprints produced by 5 or 10 fmol of TIF-IB are shown in lanes 4 and 5, 8 and 9, and 11 and 12, respectively. lated protected base pairs located at Ϫ25 and Ϫ41 on the template strand. The raw footprinting data (e.g. Fig. 6, lanes 4 and 5) suggest that Ϫ25 was not protected, and the datum in Fig. 8A arises from an unavoidable error in quantification due to the large enhanced signal at Ϫ26. The base at Ϫ41 is located in an unprotected region of the template strand, but unlike Ϫ25, was reproducibility protected in footprinting assays (e.g. Fig. 6, lanes 4 and 5). The three strongly protected regions on the RNA-like strand seen in Fig. 7 are also obvious in Fig. 8A  (upper panel). The quantitative analysis also shows that the weak protections around Ϫ63 were reproducible, but were too weak to meet our arbitrary criterion of a protected base. No consistent footprint was seen with uranyl nitrate on the RNAlike strand in the far downstream region (Ϫ22) (Fig. 8A, upper  panel). Although TIF-IB does not make direct contact with DNA in this region by our criteria, it is in close proximity to the rRNA promoter based on MPE-Fe footprints and photocrosslinking experiments (see "Discussion"). Phosphates that fit our criteria for protection and enhancement (Fig. 8A) are indicated in yellow or purple, respectively, on a representation of the rRNA promoter (see Fig. 10, upper panel).
As a complementary approach, we used hydroxyl radical footprinting to investigate the deoxyribose contacts made by TIF-IB. Hydroxyl radical footprinting utilizes EDTA⅐Fe(II), H 2 O 2 , and a reducing agent to produce a diffusible hydroxyl radical (46,61). This short-lived radical oxidizes the deoxyribose ring, which produces single-strand nicks in unprotected DNA, thereby revealing protein interactions with the sugar moiety. Unlike uranyl footprinting, which relies upon complex formation with phosphoryl groups on the DNA, hydroxyl radical footprinting relies on a diffusible oxidative species (40). Both methods provide detailed (Ϯ1 base pair) analysis of protein/DNA interactions because they cleave DNA very close to the bound protein with no sequence preference. Footprints were quantitatively analyzed as for uranyl nitrate. The hy-droxyl radical results are summarized on the linear rRNA promoter in Fig. 10 (upper panel). Data from Fig. 8B were used to mark the regions protected by our definition with a solid red line.
Hydroxyl radical treatment of committed complexes made with the same TIF-IB and DNA preparations yielded similar but not identical protection patterns compared with uranyl nitrate (Figs. 6 and 7 (lanes 6 -9) and Fig. 8B). Hydroxyl radical-protected regions were centered at approximately the same regions, but were larger by 1-3 residues (Figs. 6 and 7, compare lane 4 with lane 8; and Fig. 10, upper panel). The one exception was the region centered at Ϫ31 on the template strand, which was 1 base pair smaller than the corresponding uranyl-protected site. In addition, strong protection from hydroxyl radical extended one helical turn farther upstream to Ϫ67 of the template strand (Fig. 8B). In uranyl footprints, only the RNA-like strand was weakly protected in this region (Fig. 8A).
Committed complex formation is not affected by the various footprinting conditions regardless of the buffer or reaction conditions used. MPE-Fe footprinting was used to verify correct complex formation under the various binding conditions. A large part of the template (Ϫ67 to Ϫ17) and RNA-like strands (Ϫ64 to Ϫ14) were protected from cleavage with MPE-Fe, with a single intercalation site located between bases Ϫ49 and Ϫ50 (Figs. 6 and 7, lanes 10 -12), consistent with the original published MPE-Fe footprints (15). The MPE-Fe footprint did not change when performed under uranyl or hydroxyl radical footprinting conditions (Figs. 6 and 7 and data not shown). This result shows that binding at pH 7.5 instead of the usual pH 7.9 and diluting the complex as in the hydroxyl radical experiments do not affect the structure of the complex.
We noticed that there was a consistent differential protection of the DNA upstream and downstream of Ϫ49 in the MPE-Fe footprinting. The more functionally important promoter sequence (determined by mutation analysis) between Ϫ49 and Ϫ17 was protected better than the upstream portion. This is also true of uranyl footprints (see above). The importance of this observation, if any, is not known, but probably reflects the strength of the protein/DNA interactions in this region.
Several other important controls are shown in Figs. 6 and 7. Lane 1 is uncut DNA showing the amount of nicked DNA present before the addition of cutting reagents. This is subtracted in the quantitative analysis (see "Materials and Methods"). To be sure that enhanced cleavage was not due to contaminating nucleases in the protein preparation, TIF-IB was added to lanes 2 and 6, without the addition of footprinting reagents. There was no effect of the addition of protein to the uncut DNA (Figs. 6 and 7, compare lanes 2 and 6 with lane 1). Both of these footprinting techniques produced data consistent with minor groove binding by TIF-IB (see "Discussion").

DMS Protection Assays Indicate That TIF-IB Makes Limited
Contact with the Major Groove-To assess major groove contacts more directly, we used DMS protection assays. N-7 of guanine lies in the major groove and is accessible to methylation by DMS except when blocked by interaction with protein in the major groove. Three independent DMS protections were performed under saturating conditions of TIF-IB to determine whether reactivities of any guanine residues were consistently affected. Only 6 of the 31 guanines present in the TIF-IB binding site were protected, and 4 were enhanced in their reactivity to DMS. Those guanines that fit the criteria described under "Materials and Methods" for protection (asterisks) or enhancements (dots) are marked in Fig. 9; these are shown as white (protected) and orange (enhanced) in Fig. 10. On the template strand, 4 guanine residues were protected at Ϫ62, Ϫ60, Ϫ46, and Ϫ32 and 4 were enhanced at Ϫ22, Ϫ27, Ϫ33, and Ϫ51 (Fig. 9, compare lane 3 with lane 4). In multiple experiments, G Ϫ63 did not fit our criterion for enhanced cleavage. Two guanine residues were consistently protected on the RNA-like strand at positions Ϫ24 and Ϫ25, and none were enhanced (Fig. 9, compare lane 7 with lane 8).
To determine if affected guanine residues are important to TIF-IB binding, we examined previous deletion and point mutation data for 9 of the 10 guanines (Ϫ25 was not mutated) listed above (30). Only mutation of Ϫ24, Ϫ32, and Ϫ33 had a large negative effect on activity. The overall low number of affected guanines (10 out of 31) in the DMS experiments, together with mutation analysis, supports a primarily minor groove-binding mechanism.

DISCUSSION
The detailed structure of RNA polymerase I transcription initiation complexes on rRNA genes is generally not known, but progress is being made on the A. castellanii committed complex. We have studied the structure of the complex between this fundamental transcription factor and the core promoter and, as suspected from a number of earlier results, have found that it binds mainly to the minor groove of the template. Since chemical groups distinguishing base pairs are deficient in the minor groove, this is somewhat unexpected, but may give us insight into the evolution of the eukaryotic rRNA promoter into its wide variety of primary sequences, but conserved helical shape anomalies (25).
Drug Inhibition Experiments Are Consistent with Minor Groove Binding-Two structurally related dA-dT-binding drugs, netropsin and distamycin A, and a dG-dC-binding drug, mithramycin, inhibit TIF-IB binding as well as the known minor groove-binding protein TBP, but do not inhibit binding of CREB, a major groove-binding transcription factor. The two dA-dT-specific drugs have IC 50 values that are not identical, but their relative inhibition matches their binding to this template. The structure of distamycin A allows it to bind more easily to dA-dT sequences interrupted by single dG-dC base pairs, whereas netropsin binding to similar interrupted sites is compromised (62). MPE-Fe footprints of these minor groovebinding drugs on the A. castellanii rRNA promoter showed that distamycin A binds at lower concentrations (Fig. 3), consistent with its lower IC 50 .
An alternative explanation for discrepancies in IC 50 values is that the antibiotics inhibit binding by altering the structure of DNA, thereby inhibiting indirectly. Each drug might change the structure of the DNA in a slightly different manner, resulting in different IC 50 values. Several groups have shown that distamycin A alters the activity of DNase I and Escherichia coli RNA polymerase (63)(64)(65), which they attribute to the drug's ability to change the long-range structure of the DNA. Inhibition of major groove binding by homeodomain peptides (antp HD and ftz HD (59)) has been attributed to DNA structural changes induced by distamycin A, but this inhibition is not generally observed (e.g. CREB, the major groove-binding protein used as a control in our experiments). Therefore, indirect inhibition due to structural alteration depends on the protein and/or binding sites being tested. Indeed, crystal and solution structures of the drug-DNA complexes suggest that they introduce relatively subtle changes in DNA structure such as widening of the minor groove by 0.5-2.0 Å and slight bending (8°) as the drug replaces the spine of hydration (62, 66 -68). We have no evidence to support significant long-range structural changes in the A. castellanii rRNA promoter upon drug binding. For example, no unusual cutting patterns were observed outside the drug-protected binding sites in MPE-Fe footprints of the drugs (Fig. 3). However, we could not totally eliminate the possibility that small undetectable drug-induced changes are responsible for the inhibition and the observed differences in IC 50 .
We also found that a 10-fold higher concentration of drug was required to inhibit binding of TIF-IB compared with binding of TBP, which binds to the TATA box minor groove. There are several reasons for this discrepancy. First, the K D for TBP is ϳ1-2 orders of magnitude lower than that for TIF-IB: 50 pM for TIF-IB 3 and 400 -2000 pM for TBP (69 -71). Second, the dA-dT-specific drugs netropsin and distamycin A directly compete with TBP for the small dA-dT-rich TATA box. In contrast, the TIF-IB 50-bp binding site allows it to make contact with alternative sites even if the dA-dT-rich sequences are blocked, and so it requires a higher concentration of drug to inhibit. The presence of the TAF I factors in TIF-IB also makes direct comparison of the two protein-DNA complexes difficult. In TIF-IB, the TAF I factors, and not TBP, make the majority of contacts with the rRNA promoter (19). In fact, we have shown that TBP in Acanthamoeba does not utilize its TATA-binding domain to bind the rRNA promoter (35).
Footprinting Experiments Demonstrate Mainly Minor Groove Contacts between TIF-IB and the Promoter-The results of three chemical footprinting studies of the committed complex are shown on linear and three-dimensional representations of the Acanthamoeba rRNA core promoter, from Ϫ70 to Ϫ11, in Fig. 10 (upper and lower panels, respectively). Reactivities of phosphoryl groups (asterisks) with uranyl nitrate, deoxyribose (red bars) with hydroxyl radicals, and guanine in the major groove with DMS were tested. In each successive representation in Fig. 10 (lower panel), the ribosomal RNA promoter is rotated away from the reader in 90°increments around the x axis, starting at the top and proceeding down.
Since there are several successive protected regions, one cannot determine by simple inspection whether the protections are of the minor or major groove because two successive minor groove protections also bracket a major groove. However, the overall pattern indicates that TIF-IB binds primarily, but not exclusively, by minor groove interactions. First, in the threedimensional reconstruction, the protections on one strand are directly across the minor groove from the protections on the other strand, whereas protections across the major grooves are significantly out of register (Fig. 10, lower panel). For example, compare the alignment of b and bЈ with bЈ and c. Second, consider the ideal situation in which protein protects 5 base pairs of each strand across a groove. If the minor groove is contacted, the protection of one strand exhibits a 4-base offset in the 3Ј-direction relative to the other strand. In contrast, major groove contacts have a 5-base pair offset in the opposite or 5Ј-direction. Thus, for successive minor groove protections on the same face of the helix, the spacing from the center of one protected region to the center of the next on the opposite strand alternates between 4 and 6 bp as one zigzags along the helix. If successive major grooves are protected, the same center-tocenter distances are an unvarying 5 bp. In our data (Fig. 10), the distances between the centers of the protected regions as one zigzags downstream (a to aЈ to b to bЈ, etc.) are 3, 7, 4, 7, 4, 9, 3, and 7 bp, respectively. These are clearly closer to ideal minor groove spacings (4,6,4,6,4,6,4,6) than to ideal major groove spacings (5, 5, 5, etc). From this, we conclude that the paired sets (a and aЈ, etc.) represent protections across successive minor grooves, with somewhat non-ideal behavior because of the wrapping of the protein-DNA complex in a right-handed superhelix (see below). This non-ideal behavior is particularly exaggerated between minor groove protection c/cЈ and d/dЈ.
Based on this analysis, TIF-IB contacts three successive minor grooves spaced about one helical turn apart, starting with the most upstream protection at Ϫ65 (a/aЈ) and ending at the region centered at Ϫ42 (c/cЈ). Our data suggest that TIF-IB interactions with the promoter are weaker in the Ϫ65 region than at downstream sites. The DNA is more accessible to uranyl nitrate binding as shown by the relative weak protection of the RNA-like strand and no protection of the template strand around Ϫ65 (Figs. 6, 7, and 8A). Additionally, MPE-Fe footprinting is weaker in this region relative to protection FIG. 9. TIF-IB bound to the core promoter enhances and protects a limited number of major groove guanines in a DMS protection assay. Lanes 1-4 are the template strand, and lanes 5-8 are the RNA-like strand. The G ϩ A markers for the respective strands are in lanes 1 and 5. DNA was not treated with DMS in lanes 2 and 6. All remaining lanes were treated with DMS and cleaved with piperidine, either without (lanes 3 and 7) or with (lanes 4 and 8) TIF-IB. Protected guanine residues, determined by the criteria stated under "Results," are marked with asterisks. Guanine residues that exhibit enhanced reactivity in the presence of protein are indicated by dots. The MPE footprinting region is marked with a black bar for reference. This experiment was repeated three separate times on the template strand and twice on the RNA-like strand.
downstream of Ϫ50 (Figs. 6 and 7). Finally, the few potential major groove contacts in this region revealed by DMS footprinting (Ϫ60 and Ϫ62) were found by mutagenesis not to be important for TIF-IB binding since the rRNA promoter sequence from Ϫ55 to Ϫ67 can be substituted with vector sequence without large effects on TIF-IB binding (72).
Additional mutation data further support limited primary sequence dependence and thus probable minor groove recognition upstream of Ϫ40. Iida et al. (29) showed that the upstream portion of the promoter sequence can be replaced by vector sequence down to Ϫ48 or Ϫ32 while still retaining full or partial promoter activity, respectively. Not until the promoter was deleted to Ϫ26 was the ability of the promoter to sequester TIF-IB completely abolished. Additionally, point mutations upstream of Ϫ40 that change dG-dC to dA-dT base pairs (Ϫ56, Ϫ55, Ϫ53, Ϫ52, Ϫ51, Ϫ50, Ϫ47, Ϫ46, and Ϫ44), including those guanines whose DMS susceptibility is altered by TIF-IB (Ϫ46 and Ϫ51), either have no effect or result in an increase in complex stability (30). The one exception was the transition mutation at Ϫ55, which was slightly protected from DMS (but does not fit our criterion for protection), which results in a moderate (33%) decrease in activity. In contrast, mutations in the same region that change dA-dT to dG-dC base pairs (Ϫ41 and Ϫ45) result in a loss of complex stability, suggesting that the introduction of guanine's C-2 amino group, which protrudes into the minor groove, interferes with TIF-IB/promoter interactions either directly or indirectly by changing the groove width.
The TIF-IB footprint gradually wraps around the DNA helix in a right-handed fashion, eventually protecting the opposite face of the helix (Fig. 10, lower panel). We recognize that it is equally plausible the DNA wraps around the protein in a righthanded helix of high pitch, or both the DNA and protein could wrap around each other. In the region near Ϫ30, the protein footprint wraps nearly completely around the DNA. It is noteworthy that, downstream of Ϫ38, only a single TAF I , TAF I 96, can be cross-linked to the DNA (19). This TAF I contacts DNA from Ϫ64 to at least Ϫ7 (4), suggesting that it binds in an extremely extended configuration or that several molecules of TAF I 96 are in the complex. The latter is less likely because the molecular weight of the complex determined by scanning transmission electron microscopy is incompatible, within experimental error, with more than one of each subunit of TIF-IB in the committed complex.
In the region just upstream and downstream of Ϫ30, altered reaction of guanines with DMS suggests that TIF-IB makes more extensive contact than upstream with the major groove as the protein wraps around the DNA. Two of the guanines with In the upper panel, phosphates protected from cleavage by uranyl are marked with yellow asterisks, whereas the single enhanced uranyl cleavage site is purple. Those nucleotides protected from hydroxyl radical cleavage are shown by the solid red line next to each DNA strand. Guanine residues that are protected from DMS modification are white, and those with enhanced DMS reactivity are orange. In the lower panel, the rRNA promoter is rotated around the x axis in three 90°increments, away from the reader, starting with the top and proceeding down. The color scheme is the same as in the upper panel; the atoms of the phosphoryl groups for the nucleotides protected (yellow) and with enhanced reactivity (purple) in uranyl footprinting and the atoms of the deoxyribose residues exhibiting hydroxyl radical protection (red) are shown. Guanine residues protected from (white) or with enhanced reactivity toward (orange) DMS are colored only on the atoms that lie in the major groove. altered DMS reactivity, one protected (Ϫ32) and one enhanced (Ϫ33), are located in this region (Fig. 10, upper panel and third row in lower panel). Kownin et al. (30) have shown that mutation of these residues decreases committed complex "stability" by 47 and 39%, respectively, suggesting that these interactions are specific and important. We propose that as TIF-IB wraps more extensively around the helix, it interacts with these 2 guanines in the major groove. Major groove contacts in this region must be relatively limited, however, since several other upstream guanines (Ϫ35, Ϫ36, and Ϫ37) do not experience a change in susceptibility to DMS modification, and their mutation has relatively little consequence (see Fig. 9 and Ref. 30). Near Ϫ30, TIF-IB again bridges and interacts mainly with the minor groove. The dA-dT-rich portion of this site, Ϫ31 to Ϫ28, was strongly protected in uranyl and hydroxyl radical footprinting. This segment is also the most important for committed complex stability; point mutations that change any of these dA-dT base pairs to dG-dC result in 62-77% decreases in complex stability. Furthermore, a transition mutation at Ϫ30 (from A to G) disrupts the TIF-IB/DNase I footprint, again emphasizing the importance of this minor groove site (30). The footprinting data downstream of Ϫ26 suggest that there are additional major groove interactions. DMS footprinting reveals 2 strongly protected guanine residues located at Ϫ24 and Ϫ25 of the RNA-like strand and 1 guanine with enhanced reactivity at Ϫ22 on the template strand (Fig. 9). Point mutation of the protected guanine located at Ϫ24 decreases TIF-IB complex stability by 35%, and mutation of the guanine with enhanced reactivity at Ϫ22 increases stability by 59%, indicating that interaction with these guanines is significant. In addition, a cluster of four point mutations in this region (CϪ22G, TϪ23C, TϪ26A, and CϪ27G) essentially abolishes TIF-IB binding (3,29). Hydroxyl radical and uranyl protections show that TIF-IB makes strong contact with the DNA backbone on the template strand, but not on the RNA-like strand near Ϫ20. This shows that TIF-IB is bridging the major groove (Figs. 6, 7, 8 (A and B), and 10 (upper and lower panels).
There are several lines of evidence to support small TIF-IBinduced changes in promoter DNA structure. (a) Uranyl nitrate footprinting reveals a strongly enhanced cutting site at Ϫ26 on the template strand (Figs. 6 and 8A; marked with purple asterisks in Fig. 10, upper panel). This phosphate is on the side of the helix opposite the (nearly surrounding) protein. Structural changes could increase the binding efficiency of the uranyl ion to the phosphoryl group at this site. Alternatively, the structural changes could render the deoxyribose either 3Ј or 5Ј to the bound phosphoryl more susceptible to oxidation (40). (b) Four guanines on the template strand are more reactive to DMS modification in the presence of TIF-IB. Either structural changes have positioned the guanine so that it is more accessible to DMS, or TIF-IB somehow attracts DMS specifically to these sites. (c) There is an enhanced intercalation site for MPE-Fe between Ϫ49 and Ϫ50 when TIF-IB is added that is not seen in DNA alone. We suspect that slight changes in structure, such as groove width and twist, alter how the reagents bind, modify, intercalate, or cleave DNA. (d) Previous data showed that TIF-IB induces a bend of 45°, centered around Ϫ23 (19). However, gross structural changes resulting from TIF-IB binding such as large bends or extensive wrapping or looping of the rRNA promoter have been ruled out for Acanthamoeba TIF-IB by scanning transmission electron microscopy and circular permutation analysis (19).
The primary sequences of RNA polymerase I transcriptional regulatory elements are not conserved (for review, see Refs. 73 and 74). Yet the positioning of functional elements of the promoter (CPE, UPE, spacer promoters, and enhancers) is surpris-ingly conserved. Furthermore, spacer promoters within a given species may have very little primary sequence similarity to the core promoter from the same organism, yet they function using the same transcription factors (75). These observations have led to the hypothesis that polymerase I factors recognize intrinsic structural features of DNA and not primary sequence information. To distinguish these possibilities, we considered an extended substitution of C for T and I for A throughout the promoter as was done to show that TBP binds the minor groove of the TATA box (52). Such substitutions change mainly functional groups in the major groove, with limited effects on the minor groove functional groups. This experiment will not work in our case, however. In this experiment, the only result that gives interpretable data is when one can make all the changes and there is no effect on binding. If such occurs, one can conclude that the significant changes in the major groove functional groups do not affect binding. However, if the changes do affect binding, there are two possible reasons: 1) the protein is making contacts with functional groups in the major groove (or, in fact, with the change at position 2 in the minor groove caused by the change from A to G in response to the alteration from T to C in the complementary strand); or 2) the protein requires the shape of the major and minor grooves to be that of the wild-type sequence, and the sequence alteration changes this shape. From previous point mutation studies by Kownin et al. (30), we already know that a number of alterations of T to C and C to T have an effect on the binding of TIF-IB in competition assays. Thus, we know that binding will be affected by the suggested experiment, so the result will not distinguish between the two possibilities for the mechanism of binding.
New evidence to support the hypothesis that the tertiary structure of rRNA promoters has been conserved appeared recently. Marilley and Pasero (25) have computer-analyzed rRNA promoters from 13 phylogenetically diverse species for conserved parameters such as helical curvature, twist, and double-helical stability. They identified several regions of highly conserved curvature in the rRNA promoter. Interestingly, one of these elements is an increased curvature around Ϫ23, where we have shown TIF-IB induces a 45°bend. They also determined that the 200 base pairs preceding the transcription start site exhibit an unusual helical twist angle that decreases in a sharp continuous manner. Both of these parameters help shape the minor groove width and three-dimensional path of the DNA helix. Unfortunately, the A. castellanii promoter was not included in the study, so we cannot relate our results to specific structural details of the promoter. However, it is clear that extensive recognition of the minor groove by TIF-IB would be favored by the existence of conserved anomalies in the groove widths and three-dimensional path of the helix axis.
This report is the first to describe a detailed binding mechanism for TIF-IB or for any native TBP⅐TAF complex from any polymerase system. Cumulative data from previous mutagenesis, cross-linking, and our footprinting studies support a primarily minor groove-binding mechanism. Most specific transcription factors have major groove-binding motifs (58). TIF-IB has utilized multiple minor groove contacts, perhaps coupled with a few specific major groove interactions, to confer specificity in binding. It is possible that specific interaction with the minor groove is achieved by recognition of distinct structural features such as width, degree of twist, and bends. This would explain sequence tolerance of Acanthamoeba TIF-IB and suggest that structural recognition is not limited to UBF in the RNA polymerase I system. Alternatively, TIF-IB may be making sequence-specific contacts with functional groups in the minor groove, but distinguishing this mechanism will require additional studies.