The Role of Transcription Initiation Factor IIIB Subunits in Promoter Opening Probed by Photochemical Cross-linking*

The core transcription initiation factor (TF) IIIB recruits its conjugate RNA polymerase (pol) III to the promoter and also plays an essential role in promoter opening. TFIIIB assembled with certain deletion mutants of its Brf1 and Bdp1 subunits is competent in pol III recruitment, but the resulting preinitiation complex does not open the promoter. Whether Brf1 and Bdp1 participate in opening the promoter by direct DNA interaction (as ς subunits of bacterial RNA polymerases do) or indirectly by their action on pol III has been approached by site-specific photochemical protein-DNA cross-linking of TFIIIB-pol III-U6 RNA gene promoter complexes. Brf1, Bdp1, and several pol III subunits can be cross-linked to the nontranscribed strand of the U6 promoter at base pair −9/−8 and +2/+3 (relative to the transcriptional start as +1), respectively the upstream and downstream ends of the DNA segment that opens up into the transcription bubble. Cross-linking of Bdp1 and Brf1 is detected at 0 °C in closed preinitiation complexes and at 30 °C in complexes that are partly open, but also it is detected in mutant TFIIIB-pol III-DNA complexes that are unable to open the promoter. In contrast, promoter opening-defective TFIIIB mutants generate significant changes of cross-linking of polymerase subunits. The weight of this evidence argues in favor of an indirect mode of action of TFIIIB in promoter opening.

Bacterial RNA polymerase holoenzymes locate their cognate promoters through subunits that recognize specific DNA motifs located upstream of transcriptional start sites. subunits also participate in subsequent steps of the reaction sequence that generates an RNA chain-elongating transcription complex. For example: 1) 54 is directly involved in initiating promoter opening next to its Ϫ12 DNA-binding site (1); 2) 70 participates directly in initiating promoter opening and binds sequence-specifically to the nontranscribed strand of the upstream segment of the transcription bubble in the open promoter complex (2)(3)(4)(5); and 3) 70 also participates in promoterproximal pausing events that are required for assembly of promoter-specific termination-resistant elongation complexes (6 -8). In addition, 70 and 54 also mediate effects of proteins that activate transcription by allowing or accelerating promoter opening (9,10).
Eukaryotic nuclear RNA polymerases locate their promoters through DNA-bound complexes of core transcription initiation factors. Yeast RNA polymerase (pol) 1 III is brought to its promoters primarily by interacting with its transcription factor TFIIIB, which is composed of subunits TBP, Brf1, and Bdp1 (previously called BЉ or TFIIIB90). A homologous assembly operates at most human pol III promoters, whereas a paralogous assembly containing Brf2 (previously called BRFU or TFIIIB50) in place of Brf1 functions at a specialized subgroup of promoters that require participation of the SNAP C DNAbinding complex and its activators (reviewed in Refs. 11-13; see Ref. 14 for a new TFIIIB subunit nomenclature).
TFIIIB also participates in two post-recruitment steps of transcriptional initiation. The first evidence for this proteinlike role of TFIIIB came from analysis of the properties of certain Brf1 and Bdp1 mutants that are inactive in transcription of linear DNA, although they recruit pol III to the promoter (15,16): 1) A cluster of Bdp1 internal deletions interferes with initiation of DNA strand separation of the upstream end of the transcription bubble; this defect can be repaired by unpairing a short (3 or 5 bp) DNA segment at the corresponding location, 5-9 bp upstream of the transcriptional start site.
2) An N-terminal Brf1 deletion that removes its zinc-binding region interferes with propagation of DNA strand separation past the transcriptional start site; this defect is repaired by unpairing short DNA segments of the downstream end of the transcription bubble (17). Breaks in either DNA strand in the vicinity of the transcriptional start also restore some transcriptional activity to these defective TFIIIB assemblies in a strand break location-specific and mutation-specific pattern. Interpretation of these effects of DNA breaks leads to an explanation of how Brf1 and Bdp1 help to enforce the upstream 3 downstream polarity of promoter opening (18).
Although the demonstration of a role for TFIIIB in promoter opening exposes a fundamental parallel between the action of initiation factors in bacterial and eukaryotic transcription, the distinctive architecture of bacterial and eukaryotic promoters may dictate corresponding differences of initiation mechanism. In particular, DNA strand separation at bacterial promoters initiates within or immediately adjacent to the transcriptional start-proximal DNA-binding sites of their subunits. On the other hand, the DNA-anchoring site of TFIIIB is centered ϳ17 bp upstream of the upstream edge of the transcription bubble, and pol II as well as pol I promoter complexes are also DNAanchored at some distance from (and primarily upstream of) the transcription bubble. Thus, it is not clear whether the TFIIIB subunits should be expected to interact directly with the pol III transcription bubble, as 70 does, or whether they might act allosterically, perhaps through pol III subunits C82, * This work was supported by the National Institute of General Medical Sciences. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Div. of Biological Sciences and the Center for Molecular Genetics, University of California, San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0634. Fax: 858-534-7073; E-mail: gak@ucsd.edu. C34, or C31, which have a clear role in open complex formation and in initiation of transcription (19 -24). We report the outcome of experiments that approach this question by probing DNA proximity of TFIIIB in the DNA segment of the transcription bubble by site-specific photochemical cross-linking.

EXPERIMENTAL PROCEDURES
DNA Templates and Probes-The 198-bp (Ϫ60 to ϩ138) pU6 R boxBderived transcription template has been described (25). Fully duplex and heteroduplex bubble-containing 86-bp (Ϫ56 to ϩ30) photochemical cross-linking probes all differ from pU6 R boxB at three positions: an upstream (Ϫ29) A 3 G substitution to generate a TGTA mutant TATA box for TFIIIB binding in a single orientation with the TBP variant, TBPm3 (26) and two downstream (ϩ8 and ϩ9) A 3 T substitutions (Fig.  1A). These substitutions were generated by PCR with mutagenic primers using the above-specified 198-bp transcription template (25) and variants of this template that were previously used for making transcription templates with short heteroduplex bubbles (27). Bottom (transcribed) strands that served as templates for primer-directed incorporation of the photoreactive deoxyribonucleotide 5-(NЈ-[p-azidobenzoyl]-3-aminoallyl)-dUTP (28) with adjacent 32 P-labeled deoxyribonucleotides (essentially as described in Ref. 29) and heteroduplex formation were generated by PCR in which the primer for the bottom strand contained a ribonucleotide at its 3Ј end for subsequent alkaline hydrolysis and denaturing gel isolation (25). One half of each duplex photoreactive probe was denatured at 94°C in the presence of a 3-fold molar excess of bottom strands altered in sequence to form either a Ϫ9/Ϫ5 or a ϩ2/ϩ6 heteroduplex bubble (Fig. 1A), followed by annealing for 60 min at 63°C in 30 mM Tris-Cl, pH 8.0, 7 mM MgCl 2 , 50 mM KCl, and 2 mM ␤-mercaptoethanol. Both duplex and heteroduplex probes were purified on MDE-Hydrolink (BioWhittaker) gels (30).
Proteins-The purification and quantification of proteins (with the exception of FLAG epitope-tagged pol III) has been described (17). Quantities of pol III are specified as fmol of enzyme active for specific transcription (31); quantities of other proteins are specified as fmol of protein. Wild type TBP and full-length Bdp1 were estimated to be nearly 100% active, and full-length Brf1 was estimated to be 20% active in the formation of heparin-resistant TFIIIB-DNA complexes. Brf1⌬366 -409 (rtBrf1), the parental reference type for Brf1N⌬68, is 70% more active than Brf1 in specific transcription.
Saccharomyces cerevisiae strain NZ16 (32) (originally constructed by Nick Zecherle and Benjamin Hall, University of Washington) was generously provided by Blaine Bartholomew (Southern Illinois University at Carbondale). The strain contains a hexahistidine tag followed by four tandem FLAG epitope tags at the N terminus of the RET1 gene (encoding the C128 subunit of pol III). A rapid purification scheme involving tandem chromatography on BioRex 70, nickel-nitrilotriacetic acidagarose, and heparin-Sepharose, using an automated fast protein liquid chromatography system, yielded apparently homogeneous pol III (as judged by silver staining) with 40% of molecules active for specifically initiating transcription. The details of the procedure and chromatography program are available upon request.
Photochemical Cross-linking-Preinitiation complexes were prepared in two closely related ways. In procedure I (Figs. 1, B, C, and E, and 2) TFIIIB-DNA complexes were formed during a 60-min incubation at 20°C in 96 l of binding buffer (40 mM Tris-Cl, pH 8.0, 7 mM MgCl 2 , 60 mM NaCl, 3 mM ␤-mercaptoethanol, 5 g/ml poly(dG-dC):poly(dG-dC), 25 g/ml bovine serum albumin, 5% (v/v) glycerol) containing 1.6 pmol of TBPm3, 1.4 pmol of Brf1, or 1 pmol of Brf1N⌬68, 0.9 pmol of Bdp1, or Bdp1⌬355-372, and 24 fmol of photoreactive DNA probe. This was followed by the addition of 600 ng of sheared salmon sperm DNA in 4 l of diluent buffer (10 mM Tris-Cl, pH 8, 0.1 mM Na 3 EDTA) and 9.6 fmol of FLAG-tagged pol III in 24 l of buffer H (20 mM NaHEPES, pH 7.8, 20% (v/v) glycerol, 7 mM MgCl 2 , 0.1 mM Na 3 EDTA, 10 mM ␤-mercaptoethanol, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin and pepstatin) also containing 65 mM (NH 4 ) 2 SO 4 , 100 mM NaCl, and 100 g/ml bovine serum albumin for 20 min. Complexes were UVirradiated as described (33), loaded onto 20-l anti-FLAG M2 agarose columns equilibrated in buffer A (20 mM Tris-Cl, pH 8.0, 7 mM MgCl 2 , 100 mM NaCl, 10 mM ␤-mercaptoethanol, 20 g/ml bovine serum albumin, 5% (v/v) glycerol, 1 mM Na 3 EDTA, 0.1% Tween 20, and 0.5 mM phenylmethylsulfonyl fluoride), washed four times with 40-l aliquots of buffer A and eluted with 40 l of buffer A containing 100 g/ml 3ϫ FLAG peptide. Nuclease treatment followed (33). In procedure II ( A, constructs for analysis of cross-linking. U designates the photoactive T analogue ABdUMP, and lowercase letters designate ␣-32 P-labeled nucleotides. B, cross-linking to the nontranscribed strand at bp Ϫ8/Ϫ7. Each set of five lanes shows one analysis of a promoter complex assembled with TFIIIB only (no pol III; lanes 1-5); BЈ (Brf1 and Tbp1; no Bdp1) and pol III (lanes 6 -10); and TFIIIB together with pol III (lanes [11][12][13][14][15]. L is a duplicate of the reaction mixture applied to immobilized anti-FLAG antibody, FT is the flowthrough; W3 and W4 are wash fractions, and E is the eluate released with FLAG peptide. The same proportion of the total material in each fraction has been loaded on the gel. Cross-linked proteins are identified at the right side. The asterisks next to lanes 6 and 7 mark unknown materials whose mobilities do not match Bdp1 or C128. C, cross-linking to the nontranscribed strand at bp Ϫ8/Ϫ7 in a preformed bubble segment covering bp Ϫ9 to Ϫ5. Fractions and cross-linked proteins are indicated as in panel B, except that L represents one-half of the column load. D, cross-linking to the nontranscribed strand at bp Ϫ8/Ϫ7 at 30 and 0°C. TFIIIB-DNA complexes (lanes 1 and 3) or control reactions lacking TFIIIB (lanes 2 and 4) were formed as described under "Experimental Procedures" and transferred to 30°C (lanes 1 and 2) or 0°C (lanes 3 and 4) after the addition of nonspecific competitor (salmon sperm) DNA. Pol III was added to each reaction mixture 5 min later, and the incubation was continued for 20 min followed by UV irradiation at the same temperature. Each lane shows cross-linking in the fraction eluted from immobilized anti-FLAG antibody with FLAG peptide. E, cross-linking to the nontranscribed strand at bp ϩ2/ϩ3 in intact duplex DNA and in DNA with a preformed bubble segment covering bp ϩ2 to ϩ6. The fractions and cross-linked proteins are identified as in C.
(NH 4 ) 2 SO 4 and 100 g/ml bovine serum albumin for 20 more min prior to UV irradiation. Chromatography on anti-FLAG agarose proceeded as above. All of the experiments shown were replicated two or three times with essentially identical results.

Access of Brf1 and Bdp1 to the Opening DNA Bubble of the Transcription Initiation Complex-If
Bdp1 and Brf1 play a direct role in DNA melting at the transcriptional start, they should contact this DNA segment. Both TFIIIB subunits can be cross-linked to both DNA strands near bp ϩ1 in TFIIIB-DNA complexes (15, 32, 34 -36), but it remained to be determined whether the proximity of Bdp1 and Brf1 to the start site of transcription is maintained upon entry of pol III into the initiation complex or whether these subunits of TFIIIB are displaced. The following approach was used to examine the crosslinking of TFIIIB in the presence of DNA-bound pol III. Initiation complexes were assembled with FLAG-tagged pol III on duplex DNA as well as DNA unpaired at the upstream end (bp Ϫ9 to Ϫ5) or downstream end (bp ϩ2 to ϩ6) of the transcription initiation bubble. Duplex DNA and bp Ϫ9/Ϫ5 "bubble" DNA were photoactively tagged in the nontranscribed strand at bp Ϫ8 and Ϫ7 with ABdUMP. Duplex and bp ϩ2/ϩ6 bubble DNA were also photoactively tagged in the nontranscribed strand at bp ϩ2 and ϩ3. Each DNA probe contained a radioactive nucleotide next to its photoactive nucleotides (Fig. 1A). TFIIIB-pol III complexes assembled on these DNAs were UVirradiated, pol III-containing complexes were separated from TFIIIB-DNA complexes on immobilized anti-FLAG antibody, and both the antibody-retained and flow-through materials were analyzed to identify cross-linked proteins. Control promoter complexes were made with TFIIIB alone, with the omission of Bdp1 and with promoter opening-defective TFIIIB assembled with Bdp1⌬355-372 or Brf1N⌬68. The nearly symmetric U6 TATA box allows some TBP binding and resultant assembly of TFIIIB-pol III complexes in the reverse orientation, which complicates structural interpretation of crosslinking. This problem was circumvented by the use of the mutant TBPm3 in conjunction with a TGTA box promoter (TGTAAATA) (26,37). Truncating the upstream end of the DNA probe at bp Ϫ56 also eliminated the possibility of pol III binding to reverse-oriented TFIIIB-DNA complexes.
The principal result of the analysis of probes photoactively tagged at bp Ϫ8 and Ϫ7 was that both Bdp1 and Brf1 (the latter cross-linking more weakly) are located in the vicinity of DNA in a pol III-containing complex (Fig. 1B, lane 15). TFIIIB complexes were cleanly separated from TFIIIB-pol III-DNA complexes on the anti-FLAG column (lanes 1-5 and 11-15); omission of Bdp1 prevented pol III recruitment (lanes 6 -10). The C160, C128, C82, and C34 subunits of pol III were also cross-linked to DNA at this location.
Preopening bp Ϫ9 to Ϫ5 restores the transcriptional activity of promoter opening-defective Bdp1 internal deletion mutants. Completely opening bp Ϫ9 to Ϫ5 by unpairing clearly correlated with greatly increased cross-linking of the C82 pol III subunit in the presence of full-length Bdp1 (Fig. 1C, lane  7; compare the ratio of the C82:Brf1 signals with Fig. 1B, lane  15). Although this result appears to imply that cross-linking of the C82 subunit at positions Ϫ8/Ϫ7 is an inherent property of open complex formation, this may not be the case. Wild type TFIIIB-pol III-duplex DNA complexes formed and UVirradiated at 30°C (open) and 0°C (closed) both displayed equivalent, low levels of C82 cross-linking relative to Bdp1, Brf1, and the other subunits of pol III (Fig. 1D, compare lanes 1 and 3; only the FLAG peptide-eluted fractions are shown). The somewhat weaker cross-linking of all proteins in Fig. 1D (lane 3, relative to lane 1) was not reproducibly observed and probably reflects variations in the efficiency of pol III assembly at 0°C or in recovery of complexes from anti-FLAG agarose, because the relative efficiencies of all protein crosslinking (including C34; not shown) at 30 and 0°C were indistinguishable. Cross-linking to C160 and C128 (relative to Bdp1 and Brf1) also increased when the bp Ϫ9/Ϫ5 segment was preopened but did not change between 30 and 0°C in complexes formed with duplex DNA.
Cross-linking at the downstream segment of the transcription bubble was also examined. Preopening bp ϩ2 to ϩ6 restores transcriptional activity to TFIIIB assembled with the N-terminal deletion subunit Brf1N⌬68. Full-length Bdp1 and Brf1 were found to cross-link to the nontranscribed strand at bp ϩ2/ϩ3 in duplex DNA along with C128 (Fig. 1E, lane 10) and C34 (not shown). Cross-linking of the largest pol III subunit, C160, at this location was very weak. Opening the bp ϩ2/ϩ6 bubble greatly increased the efficiency of Brf1 crosslinking and also increased Bdp1 cross-linking relative to C128 (lane 5). Pol III entry into the bp ϩ2/ϩ6 bubble promoter changed the relative efficiencies of Brf1 and Bdp1 cross-linking (compare lanes 2 and 5). Cross-linking of Brf1 and Bdp1 to DNA at bp ϩ2/ϩ3 does not require promoter opening, because it was also observed in FLAG antibody-purified TFIIIB-pol III promoter complexes formed and UV-irradiated at 0°C and because the relative cross-linking efficiencies of C160, C128, Bdp1, Brf1, and C34 at 30 and 0°C were indistinguishable (data not shown).
The above results indicate that Bdp1 and Brf1 lie in close proximity to the upstream and downstream borders of the prospective and actual open complex. Because the photoreactive moiety of ABdUMP barely protrudes beyond the edge of the major groove (28) tive Bdp1 and Brf1 deletion mutants on the cross-linking of TFIIIB and pol III subunits near the start site of transcription was examined next (Fig. 2). Protein-DNA complexes containing both pol III and TFIIIB were affinity-purified after UV irradiation, as in the preceding experiment. Only the FLAG peptideeluted fractions are shown.
TFIIIB-pol III-DNA complexes assembled with Bdp1⌬355-372 displayed significantly diminished cross-linking of Bdp1, Brf1, and the C160, C128, and C82 subunits of pol III relative to the C34 subunit of pol III to bp Ϫ8/7 in duplex DNA (Fig. 2,  compare lanes 1 and 2). A similar diminution of C160 and C128 cross-linking was previously observed with Bdp1⌬372-387 in a mixture of TFIIIB-and TFIIIB-pol III-DNA complexes (15). Preopening the bp Ϫ9/Ϫ5 bubble, which allows a fully open complex to be formed with Bdp1⌬355-372, partially restored C160, C128, Bdp1, and Brf1 cross-linking relative to C34 (compare lanes 5 and 2). The effect of deleting Bdp1 amino acids 355-372 on cross-linking at the downstream end of the transcription bubble (probe ϩ2/ϩ3) was more pronounced, with a nearly complete loss of cross-linking of Bdp1, Brf1, and the C128 subunit of pol III but not of C34 cross-linking (compare lanes 7 and 8). The loss of pol III large subunit cross-linking to duplex DNA evidently reflects a subtle structural change, in light of a prior methidiumpropyl EDTA-Fe(II) footprinting analysis of the transcribed strand of TFIIIB-pol III-DNA complexes formed with wild type Bdp1 and Bdp1⌬372-387 (17) showing identical protection of duplex DNA in this region. Opening the bp ϩ2/ϩ6 bubble, which does not ameliorate the promoter opening defect of Bdp1⌬355-372 (17), partly restored C128, Brf1, and Bdp1 cross-linking (compare lanes 11 and 8).
Deletion of the N-terminal 68 amino acids of Brf1 had nearly the opposite effect, in that cross-linking of the C34 pol III subunit at bp Ϫ8/Ϫ7 in duplex DNA decreased significantly relative to that of Bdp1, Brf1, C160, and C128 (Fig. 2, compare  lanes 1 and 3). The most pronounced effect of the N⌬68Brf1 deletion on TFIIIB-pol III complexes was the loss of crosslinking of the C82 subunit of pol III (lane 3). Opening the bp Ϫ9/Ϫ5 bubble restored cross-linking to C82 and increased cross-linking to C34 (lane 6). The pol III complex assembled on bp Ϫ9/Ϫ5 bubble promoters with Brf1N⌬68 remains defective in downstream bubble propagation and transcription (17), so restoration of C82 cross-linking to bp Ϫ8 and Ϫ7 only reflects local strand separation rather than transcriptional competence. Brf1N⌬68 had little effect on the pattern of cross-linking at downstream segment of the transcription bubble (probe ϩ2/ ϩ3; lanes 9 and 12).
In summary, we conclude that Bdp1, Brf1, and four pol III subunits lie close to the upstream edge of the transcription bubble both in open and in closed promoter complexes. One signature of this segment of the open complex transcription bubble is cross-linking of the C82 pol III subunit, which increases when this region is preopened and dramatically decreases in complexes formed with promoter opening-defective Brf1N⌬68. Proximity of Brf1 and Bdp1 to DNA extends downstream of the transcriptional start site, in both open and closed pol III-promoter complexes. The characteristic of the downstream site is the loss of cross-linking of Bdp1, Brf1, and the C128 pol III subunit in complexes formed with Bdp1⌬355-372. The regions in Bdp1 and Brf1 that cross-link to the nontranscribed strand of the transcription bubble lie outside of the protein segments whose deletion prevents open complex formation (that is, outside Bdp1 amino acids 355-421 and Brf1 amino acids 1-68; data not shown).

DISCUSSION
Implications of the Complexity of Protein Cross-linking at 30 and 0°C-Four subunits of pol III (C160, C128, C82, and C34), Bdp1, and Brf1 all lie close to bp Ϫ8 and Ϫ7 on the nontranscribed strand at the upstream end of the open complex transcription bubble. At bp ϩ2 and ϩ3, at the downstream end of the open complex transcription bubble, cross-linking of C160 diminished significantly, and cross-linking of C82 was not detected. This pattern of cross-linking is similar to what has been observed previously on the SUP4 tRNA and 5 S rRNA genes, with the exception that cross-linking to Bdp1 and Brf1 was either barely detectable or absent downstream of bp ϩ1 in the presence of TFIIIC but could be detected at bp ϩ6 when TFIIIC was removed (Ref. 34; see Ref. 32 for a comprehensive survey of protein cross-linking between bp Ϫ17 and ϩ17). The proximity of these pol III and TFIIIB subunits to these sites was not strongly dependent on open complex formation; the same subunits were cross-linked at 30 and 0°C. The close proximity of the nontranscribed strand to multiple protein subunits near the start site of transcription has also been observed in human pol II initiation complexes: Rpb1, Rpb2, TFIIB, and the RAP74 subunit of TFIIF co-cross-link at the upstream edge of the prospective transcription bubble in both closed and open promoter complexes; Rpb1, Rpb2, Rpb5, TFIIE-␣, and the ERCC3 subunit of TFIIH co-cross-link at the downstream edge of the transcription bubble (38).
Two aspects of the cross-linking patterns are somewhat puzzling and deserve comment. First, it is possible to rationalize the cross-linking of six subunits of pol III and TFIIIB to the Ϫ8/Ϫ7 probe (or four subunits to the ϩ2/ϩ3 probe) in open promoter complexes. The nontranscribed strand of the pol III open complex is not constrained (18) and therefore might not be tethered to any particular site. However, the same proteins are also cross-linked to DNA at 0°C in the absence of DNA strand separation. Although the ABdUMP reactive side chain has some rotational flexibility, it barely projects beyond the DNA major groove, making it unlikely that two adjacent side chains would access six proteins. Heterogeneity of pol III placement at the start site of transcription by TFIIIB would account for the diversity of subunits cross-linked at 0°C. This possibility clearly applies to the previously analyzed TATA-less SUP4 tRNA gene, for which alternative placements of TFIIIB by TFIIIC lead to heterogeneity in selection of the transcriptional start site (39,40). However, this does not appear to be the case for the TATA-containing SNR6 (U6) gene; initiation occurs exclusively at bp ϩ1 on U6 DNA templates with an intact transcribed strand (25).
We suggest instead that the polymerase-DNA association in the TFIIIB-pol III-promoter closed complex is transient, like that of bacterial RNA polymerase closed complexes. Transient release and reassociation of duplex DNA with pol III would account for the diversity of cross-linking observed at 0°C. We note that the DNase I and methidiumpropyl EDTA-Fe(II) footprints of gel-purified TFIIIB-pol III-DNA complexes are less complete for pol III than for TFIIIB (Fig. 4c in Ref. 15, Fig. 7 in Ref. 27,and Fig. 8 in Ref. 17), an observation that would also be accounted for by a loosely held pol III-DNA interaction. Transient association of DNA may also partly account for the crosslinking of nine proteins to ABdUMP at bp Ϫ5 on the transcribed strand in pol II preinitiation complexes (Ref. 41; although partial complex assemblies may exist in this study, three subunits of the final member of this complex, TFIIH, cross-link to this position).
The second peculiarity concerns the absence of strong differences between the patterns of cross-linking at 0 and 30°C. One anticipates that the pattern of cross-linking at bp ϩ2/ϩ3 would change when the promoter opens because a fully opened promoter complex pulls an additional 4 -5 bp of downstream DNA into polymerase (25), and open complex formation should ac-cordingly shift the location of bp ϩ2 and ϩ3 relative to protein.
In contrast, we found no significant difference in the relative levels of cross-linking of C160, C128, Bdp1, Brf1, and C34 at 30 and 0°C (data not shown). However, preopening the downstream end of the transcription bubble between bp ϩ2 and ϩ6 significantly enhanced ϩ2/ϩ3 probe cross-linking of Brf1 relative to C128 (Fig. 1E). A prior analysis of the SUP4 tRNA and 5 S rRNA genes suggested that cross-linking of the C82 subunit of pol III might also depend on open complex formation (35). In that work, C82 was cross-linked to DNA containing the photoreactive nucleotide 4-S-thymidine upstream of the start site of transcription but not with ABdUMP at the same positions. Efficient cross-linking to 4-S-thymidine, like its RNA counterpart, 4-S-uridine, occurs at sites that become unpaired or helically distorted (42)(43)(44)(45)(46). Again, no difference in the relative amounts of C82 cross-linking at bp Ϫ8/Ϫ7 was observed between 0 and 30°C (Fig. 1D), yet preopening the upstream end of the transcription bubble increased the relative efficiency of C82 cross-linking ϳ4-fold (Fig. 1C).
The simplest interpretation of the above observations is that even at 30°C, the TFIIIB-pol III-DNA complex is predominantly closed. If the 4-fold difference in C82 cross-linking reflects the difference between closed and fully open complexes, then open complex formation at 30°C would need to be less than 20% to obscure detection of increased C82 cross-linking. Alternatively, the changed pattern of cross-linking that is generated by preopening different segments of the transcription bubble may result from a fraction of these templates that fail to enter the single-stranded binding channel of pol III. A singlestranded bubble lying at the surface of polymerase in the closed complex would have considerably greater reach and flexibility than duplex DNA.
An Indirect Role for TFIIIB in Open Complex Formation-Deletions and point mutations in the N-terminal domain containing the zinc ribbon motif and the adjacent sequence of both Brf1 and TFIIB display three similar properties: 1) a defect in binding their cognate polymerases, 2) alterations in transcriptional start site selection and 3) a defect in transcription at a step subsequent to polymerase recruitment (16,17,(47)(48)(49)(50)(51)(52). TFIIB has also been observed to cross-link to DNA within the transcription bubble region of the pol II initiation complex (38). It is likely that the post-recruitment roles of TFIIB and Brf1 are related mechanistically but not necessarily in open complex formation.
The cross-linking experiments shown here demonstrate an intimate proximity of Brf1 and Bdp1 to the DNA segment that becomes unwound in the pol III open complex and also to the pol III subunits likely to be involved in open complex formation (Fig. 1). This proximity is consistent with a direct involvement of TFIIIB in the DNA strand separation process during open complex formation. However, the effects of promoter openingdefective deletion mutations in Bdp1 and Brf1 on cross-linking do not provide additional evidence for the direct involvement of TFIIIB. In particular, the expectation that the segments of Brf1 and Bdp1 whose deletion prevents open complex formation would be proximal to the prospective or actual open complex transcription bubble is not realized (Fig. 2). Instead, the region of Bdp1 that cross-links to the bp Ϫ8 and Ϫ7 probe has been tentatively mapped between Met 425 and Cys 485 , using a technique similar to one that was described previously (Ref. 53 and data not shown).
The results of these experiments favor an indirect mode of action of TFIIIB in open complex formation. If promoter opening-defective deletion mutations in Bdp1 or Brf1 alter or prevent DNA/protein contacts by pol III that are required for pol III-mediated DNA strand opening, one would predict that these deletions might affect the cross-linking of pol III subunits. This is what is observed (Fig. 2). Promoter opening-defective Bdp1 mutations grossly alter the proximity of pol III to DNA, reducing cross-linking of C160, C128, and C82 to bp Ϫ8 and Ϫ7 and obliterating cross-linking of all pol III subunits, except for C34, to bp ϩ2 and ϩ3. Similarly, the promoter opening-defective Brf1N⌬68 affects cross-linking of the C82 and C34 pol III subunits to bp Ϫ8 and Ϫ7. Thus, although the Brf1 and Bdp1 subunits of TFIIIB are located close to the transcriptional start site in pol III-containing preinitiation complexes and can be cross-linked from the upstream and even the downstream edge of the DNA segment that forms the transcription bubble, the evidence of this analysis favors an indirect, pol III-mediated mode of action of TFIIIB in promoter opening rather than a bacterial -like direct participation in DNA strand separation.