Activation of RNA Polymerase I1 Transcription by the Specific DNA-binding Protein LSF INCREASED RATE OF BINDING OF THE BASAL PROMOTER FACTOR TFIIB*

While the components of the initiation complex at an RNA polymerase I1 basal promoter have been well characterized, few mechanistic studies have focused on how upstream DNA-binding, transcriptional activators influence protein assembly at the initiation site. Our analysis of basal transcription on both the simian virus 40 and adenovirus major late promoters demon- strates that two slow steps in initiation of transcription are the assembly of the general transcription factors TFIID and TFIIB onto the template DNA. On the simian virus 40 major late promoter, the rate of initiation complex formation is dramatically increased in the presence of the cellular transcriptional activator LSF. Direct analysis by band mobility shift assays demonstrates that LSF has no effect on the rate of binding, or the stability of TFIID on the promoter, predicting that LSF would not affect the template commitment step. Rather, kinetic analyses demonstrate that LSF reduces the lag in the rate of initiation complex for- mation attributable to the slow addition of TFIIB and suggest that LSF increases the rate of association of TFIIB with the committed template. In addition, LSF increases the total number of transcription complexes in long term assays, which is also consistent with LSF increasing the rate of association of TFIIB, where TFIIB is not saturating. These results indicate a mechanism for the activation of the initiation of RNA poly- merase I1 transcription by one upstream activating protein, LSF. This mechanism may also be applicable to other activators that function in cases where limiting concentrations

While the components of the initiation complex at an RNA polymerase I1 basal promoter have been well characterized, few mechanistic studies have focused on how upstream DNA-binding, transcriptional activators influence protein assembly at the initiation site. Our analysis of basal transcription on both the simian virus 40 and adenovirus major late promoters demonstrates that two slow steps in initiation of transcription are the assembly of the general transcription factors TFIID and TFIIB onto the template DNA. On the simian virus 40 major late promoter, the rate of initiation complex formation is dramatically increased in the presence of the cellular transcriptional activator LSF. Direct analysis by band mobility shift assays demonstrates that LSF has no effect on the rate of binding, or the stability of TFIID on the promoter, predicting that LSF would not affect the template commitment step. Rather, kinetic analyses demonstrate that LSF reduces the lag in the rate of initiation complex formation attributable to the slow addition of TFIIB and suggest that LSF increases the rate of association of TFIIB with the committed template. In addition, LSF increases the total number of transcription complexes in long term assays, which is also consistent with LSF increasing the rate of association of TFIIB, where TFIIB is not saturating. These results indicate a mechanism for the activation of the initiation of RNA polymerase I1 transcription by one upstream activating protein, LSF. This mechanism may also be applicable to other activators that function in cases where limiting concentrations of TFIIB in the cell dictate slow binding of TFIIB.
Promoter-specific initiation of mRNA synthesis in mammalian cells requires the assembly of RNA polymerase I1 (RNA pol 11)' and multiple additional proteins into an initiation complex at the transcriptional start site of the gene. These proteins, termed general transcription factors (TF), have thus far been resolved into the following components: TFIIA, TFIIB, TFIID, TFIIE, TFIIF, and TFIIG (1)(2)(3)(4)(5)(6). Of these six fractions, only TFIID (also called DB and BTF1) has been shown to bind DNA, specifically recognizing the TATA sequence element found in many RNA pol I1 promoters (7,8). Binding of TFIID is the first step in initiation complex formation and commits the template to transcription (9,10). The subsequent assembly of the remaining factors onto the committed template occurs in at least one defined order (TFIIA, TFIIB, RNA pol 11, TFIIE/TFIIF/TFIIG) ( l l ) , presumably via protein-protein interactions (reviewed in Ref. 12).
Specific initiation of transcription by the general factors requires only a template consisting of basal promoter elements. The well defined basal adenovirus major late promoter (Ad MLP) consists of a TATA sequence element and an initiator element (13), the latter defined as the initiation site and adjacent sequences. Transcription of the basal promoter by the general transcription factors is distinct from activated transcription, which requires additional DNA sequences and corresponding specific DNA-binding proteins. These activating transcription factors, and the DNA sequences to which they bind, have been the subject of extensive research, due to their key role in the control of initiation of gene-specific transcription. The potential mechanisms by which transcriptional activating proteins might influence the basal transcription machinery, however, are not understood. Previous studies that address this point suggest that either TFIID or, in one instance, TFIIB may be direct targets for the action of particular activating proteins (7, [14][15][16][17][18][19][20][21]. The experiments presented here use a combination of gel band mobility shift analysis and kinetics to determine the mechanism of stimulation by one transcriptional activator, LSF. LSF stimulates transcription from the simian virus 40 (SV40) major late promoter (SV MLP), which directs initiation of the late mode of SV40 transcription from its major in vivo start site, L325. The SV MLP consists of multiple upstream elements and at least one downstream element (22)(23)(24) and references therein). The promoter lacks a consensus TATA-box, but does contain a TATA-like sequence which has been shown to be important for late promoter activity in vitro (25). Activation of the initiation of transcription at the SV MLP occurs via LSF binding to a site centered at -45 with respect to the major late start site (24).* We have now determined that all of the general transcription factors are required for both basal and LSF-activated transcription of the SV MLP. In addition, kinetic analyses revealed the potential rate-limiting steps in complex assembly and demon-strated that LSF facilitates one of these steps in the initiation of transcription.

MATERIALS AND METHODS
Preparation of the General Transcription Factors-Nuclear extract (850 mg of protein in 91 ml) was prepared from 100 g of HeLa cells (26) and applied to a phosphocellulose column in buffer C (20 mM Tris-HC1, pH 7.9, 2 mM dithiothreitol, 0.1 mM EDTA, 20% glycerol) plus 0.1 M KCl. The column was sequentially eluted with 0.3 M KCl, 0.5 M KCl, and 1.0 M KC1 in buffer C as previously described (27,28). The phosphocellulose column eluates were further fractionated on DEAE-cellulose columns as described (27,28), except that the TFIIE/TFIIF/TFIIG fraction was eluted with a 0.3 M KC1 step in buffer C, rather than with a salt gradient. The resulting general factor preparations that also contained SV40 promoter DNA-binding proteins were further fractionated using oligonucleotide-affinity chromatography (29). LSF was removed from the TFIIE/TFIIF fraction using an LSF-280RS DNA-affinity column, as described (24). AP-1 (30) was removed from the TFIIB fraction used for Figs. 2,3,4A, and 6B by fractionation through a DNA-affinity column containing ligated AP-1 consensus oligonucleotides (a generous gift of C. R. Wobbe and K. Struhl, Harvard Medical School, Boston, MA). The TFIIB fraction used in Figs. 5 and 6C was purified through two chromatographic steps following the DEAE-cellulose chromatography: a singlestranded DNA-agarose column (27, this material was a generous gift of J. Workman, Massachusetts General Hospital, Boston, MA, and was found to be free of AP-1 DNA binding activity, data not shown), and a DNA-affinity column consisting of ligated copies of the SV40 nucleotide sequences from 318-341,5'-gatcAGAGGTTATTTCAGG-CCATGGTGC, which span the HIP1 binding site(s) (31). The DNAaffinity column step was performed in the absence of carrier DNA. TFIIB eluted at 0.18 M KCI. This latter column separated SV MLP initiator DNA-binding activity (31) from TFIIB. The level of activation of transcription by LSF was not affected by removal of initiator DNA-binding activity from the TFIIB fraction (data not shown). The TFIIA fraction was chromatographed through a second phosphocellulose column to remove contaminating TFIID. The TFIID fraction used for gel band mobility shift analyses (Fig. 4) was further fractionated on heparin-Sepharose as previously described (28). The resulting pools of general transcription factors, dialyzed against buffer C containing 0.1 M KCl, had the following protein concentrations: TFIIA, 11 mg/ml; TFIIB (containing HIPI), 5 mg/ml; TFIIB (lacking HIPl), e0.02 mg/ml; TFIIE/TFIIF/TFIIG, 3 mg/ml; TFIID (DEAE-fraction), 0.3 mg/ml; and TFIID (heparin fraction), cO.1 mg/ml. RNA polymerase I1 was isolated from calf thymus as previously described (32). The rat a factor (TSK phenyl 5-PW fraction) was isolated as described (33) and was a generous gift of J. and R. Conaway.
LSF protein was purified from the TFIIE/TFIIF/TFIIG fraction by successive fractionation over two LSF-280RS DNA-affinity columns as described (24). The resulting pool of LSF (3.5 units/pl) was free of all other SV40 specific DNA-binding proteins as assayed by DNase I footprinting (data not shown) (34). One unit is defined as the amount of protein required to bind 1 fmol of radiolabeled DNA fragment in a band mobility shift assay (24,351. Plasmid DNA Templates-The plasmid pSVS contains the wildtype SV40 genome cloned into a deletion derivative of pBR322 (36). The mutants of the SV MLP TATA region, pSVS3181, pSVS3182, and pSVS3184, were constructed by site-directed mutagenesis in recombinant MI3 phage containing SV40 DNA sequences and then subcloned into the pSVS plasmid background. The mutant pSVS3 was a generous gift of J. Brady (25). The Ad MLP plasmid, pML(C2AT)19A-50, contains adenovirus MLP sequences from -50 to +10 cloned upstream of the G-less cassette (7). Plasmid DNAs were isolated using a standard alkaline lysis procedure (37) and were purified by two successive bandings on CsCl density gradients. SV40 viral DNA was obtained from extracts of SV4O-infected CV-I cells (38) and similarly purified.
TFIIA, 2 p1 of TFIID, 0.2 pl of TFIIB, 0.4 pl of TFIIE/TFIIF/TFIIG), Transcription Assay-The general transcription factors (0.2 p1 of calf thymus RNA polymerase I1 (4 units, 1 unit of enzyme incorporates 1 pmol of UTP in 20 min at 37 "C), and LSF (amounts indicated in figure legends) were combined in 20-pl transcription reactions under the following final conditions: 25 mM HEPES, pH 7.9, 3 mM Tris-HC1, pH 7.9, 1 mM dithiothreitol, 4-6 mM MgClZ, 20 pg/ml DNA, 40-60 mM KCl, 3-5% glycerol. After preincubation of DNA with the factors indicated in the figure legends, for 30 min at 30 "C, transcription was initiated by the addition of the remaining proteins and nucleotides (60 p M GTP, ATP, and CTP, 2 mM creatine phosphate, 1 pM UTP, and 10-15 pCi of [cY-~'P]UTP at 800 Ci/mmol). Following incubation at 30 "C for the time indicated in the figure legends, additional nucleotides were added to concentrations of 330 p M ATP, CTP, GTP, and 1 mM UTP, for a final incubation period of 2 to 3 min. Reactions were terminated by the addition of 10 mM EDTA and 1% sodium dodecyl sulfate and extracted as described (39). Transcription reactions using pML(C2AT)19A-50 as template also contained 100 p M 3'-O-methyl-GTP (Pharmacia LKB Biotechnology Inc.) and 5 units of RNase T1 (Calbiochem).
Specifically initiated, radiolabeled SV40 RNAs were selected and trimmed by an RNase T1 protection assay as described (39,40). Transcription from the Ad MLP in pML(CzAT)19A-50 directly generated a 400-base specific RNase T1-resistant transcript. The resulting RNase-resistant transcripts from all templates were resolved by electrophoresis through denaturing 5% polyacrylamide gels (291 acrylamide/N, N"methylenebisacrylamide, 8 M urea, 89 mM Tris base, 89 mM boric acid, 2 mM EDTA) as described (39). Gels were dried and exposed to preflashed (41) XAR-5 film at -70 "C. Quantitation of specific transcript levels was performed by scanning autoradiographs using an LKB laser densitometer with Gelscan XL software or by directly analyzing radioactivity using a Betascope (Betagen) or a Phosphorimager (Molecular Dynamics) and subtracting the appropriate background values.
Gel Band Mobility Shift Assay-Radioactive, double-stranded DNA corresponding to SV40 nucleotides 257-342 was generated by enzymatic amplification (42) of the TATA consensus plasmid pSVS3181, using primers that were radiolabeled by T4 polynucleotide kinase and [y3'P]ATP according to standard procedures (37). The resulting DNA spanned the SV MLP to include the LSF-280 binding site, TATA sequence and the L325 initiation site. Complete binding reactions (IO ~1 ) contained 5 fmol of template DNA and the protein fractions indicated in the figure legend in the following amounts: TFIIA, 1.1 pg; TFIID, 0.6 pg; TFIIB, 1.0 pg; rat a, 0.1 p1 (a generous gift of J. and R. Conaway); and LSF, 2-3 units. Binding was performed in 10 mM Tris-HC1, pH 7.9, 60 mM KCl, 10% glycerol, 0.4 mg/ml bovine serum albumin, 10 pg/ml carrier poly[d(G-C)], 4 mM MgCIz, 1 mM dithiothreitol, and 0.1 mM EDTA. Except for time course protocols (Fig. 4, B and C), proteins were preincubated with carrier nucleic acid for 10 min at 30 "C followed by addition of template DNA. The reactions were further incubated at 30 "C for the time indicated in the figure legend.
The wt TATA-like oligonucleotide contains the TATA-like sequence (italicized) present in wild-type SV40 DNA. The 3181 TATA oligonucleotide contains the cunsensus TATA-box (italicized). The GC123 oligonucleotide contains the LSF-GC binding site and has been shown previously to bind efficiently to LSF (24). Protein-bound DNA uersus free DNA was resolved by electrophoresis through 4% polyacrylamide gels (301 acrylamide/bisacrylamide) in a buffer containing 25 mM Tris base, 192 mM glycine, 2 mM EDTA, 1.5% glycerol, and 0.5 mM dithiothreitol (final pH 8.3). Electrophoresis was carried out at 200 V for 3.5 h in a similar buffer lacking glycerol and EDTA. Gels were subsequently dried and exposed to preflashed Kodak XAR-5 x-ray film.

RESULTS
Transcription of the SV MLP Requires the General Transcription Factors TFIIA, TFIID, TFIIB, and TFIIEITFIIFI TFIIG-Promoter-specific RNA pol 11 transcription can be reconstituted in vitro from partially purified components of a HeLa cell nuclear extract. The chromatographic isolation of the necessary components (general transcription factors (TF) TFIIA, TFIIB, TFIIE, TFIIF, TFIIG, and TFIID) using phosphocellulose and DEAE-cellulose has been described previously (6,27,28). Although these transcription factors have been defined as essential for specific initiation of transcription from the Ad MLP, it is thought that they are required by all RNA pol I1 promoters. Indeed, we previously used all the partially purified general factors (DEAE-cellulose column fractions) to reconstitute basal transcription of the SV MLP and to demonstrate the activation of that promoter by the transcription factor LSF (24). Further detailed analysis of the mechanism of LSF action required that the general factor transcription system be free of endogenous LSF and other SV40 promoter binding proteins. An examination of the general factors using a DNase I footprint assay (34) revealed that the partially purified TFIIB was contaminated by both the SV40 promoter binding protein AP-1 (30) and an initiation site binding activity, termed HIP1 (31), and that the TFIIE/ TFIIF/TFIIG fraction was contaminated by LSF (data not shown). AP-1, HIP1, and LSF were subsequently removed from the indicated general factor preparations using DNAaffinity chromatography (29). The resulting transcription system retained a high level of transcriptional activity on the Ad MLP, and a minimal basal level of transcriptional activity on the SV MLP, and was responsive to the addition of SV40 transcriptional activating proteins (LSF, this study; and Spl, Ref. 43, data not shown). Due to the differences between the Ad MLP and the SV MLP, especially in the TATA region ( Fig. l ) , we were interested in determining whether transcription of the SV MLP, initiated at SV40 nucleotide 325 (SVL325), required in particular the TATA-binding factor TFIID, as well as all the other general factors.
Transcription of SV40 viral DNA performed with only the general factors resulted in the appearance of a faint, but discrete band representing transcription from SVL325 (Fig.   2, lane 1). In these experiments, addition of LSF to a relatively low concentration (see Fig. 2 legend) led to a %fold increase in the level of initiation from SVL325 (lane 7, upper band of doublet). The transcript originating from a downstream start site mapped to SV40 nucleotide position 355 and was also consistently elevated in response to LSF (lane 7, lower band of doublet, marked by arrowhead). This transcript maps close to a minor transcript initiated in uiuo (44); its appearance as an abundant product of transcription reactions using the reconstituted system is not fully understood. Omission of any single general transcription factor from the reaction abolished LSF-activated transcription from the SVL325 start site (Fig.  2, lanes 3-6, as compared to lane 2). Surprisingly, initiation at position SVL355 was not absolutely dependent on addition of the TFIIE/TFIIF/TFIIG or TFIID fractions (lanes 5 and 6). Omission of exogenous RNA pol I1 from the reaction had little effect on the level of transcription from either SVL325 or SVL355, due to the presence of endogenous HeLa RNA pol I1 in the TFIIE/TFIIF/TFIIG fraction (data not shown). These results demonstrate that LSF-activated initiation of transcription from the SV MLP requires all of the general transcription factors, including TFIID. Basal transcription from the SV MLP demonstrated identical requirements (data not shown).
Mutational Analysis of the S V MLP TATA Region-The   binding of TFIID to the SV MLP TATA-like sequence is weak in comparison to TFIID binding to the Ad MLP TATAbox (data not shown) which might explain the low levels of basal transcription observed for the SV MLP. A correlation between TFIID binding affinity and promoter efficiency has been demonstrated (45). Consequently, the function of LSF might be to stabilize the binding of TFIID, thus increasing the efficiency of the promoter. Therefore, mutations in the SV MLP TATA-like sequences ( Fig. 1) were designed (46) to determine first whether increasing the binding affinity of TFIID would correlate with an increase in basal transcription activity, and, second, whether the ability of LSF to activate the SV MLP might be abrogated.

P C G G G T G T T C C T~~G~~~~G G G T G~~C G C G T T C G T C C~C T C~C C G C
The different template DNAs were combined with general factors in the presence or absence of LSF and preincubated until complex assembly was maximal (empirically determined to be 30 min, data not shown). Under these conditions, the relative levels of transcription initiated upon addition of nucleotides would be proportional to the number of complete initiation complexes available at the time the nucleotides were added, and perhaps to the rate at which these transcription complexes initiate. The number of complexes present following the preincubation step reflects the overall stability of the promoter complex, one component of which is TFIID. However, the level of transcription in this experimental protocol would not reflect the relative rate of binding of factors to the DNA, which is analyzed in the kinetic assays described below. The effects of the TATA-like mutations on both basal (-LSF) and activated (+LSF) transcription are shown in Fig. 3. The triple point mutant pSVS3184 and the double point mutant pSVS3182 demonstrated little effect on basal transcription (Fig. 3A, lanes 2 and 3 ) . The mutant pSVS3 supported basal transcription 9-fold better than wild-type (Fig. 3A, lune 4 ) ) consistent with previous in vitro studies (25), and pSVS3181, the consensus TATA mutant, supported basal transcription at an 18-fold higher level than wild-type (Fig. 3A, lane 5). Direct gel mobility shift binding analysis comparing the wild- type TATA-like sequence and the consensus TATA sequence in pSVS3181 demonstrated a 17-fold difference in the efficiency of TFIID binding (data not shown). Thus, the levels of basal transcription correlated with the relative affinities of TFIID to the respective sequences.
The absolute levels of LSF-activated transcription on each mutant increased in parallel with the levels of basal transcription (Fig. 3A, lunes 6-10, as compared with lanes [1][2][3][4][5], such that the overall fold activation by LSF remained fairly constant, within the range of 12-to 20-fold (Fig. 3B). We note that in these experiments LSF was present during the entire incubation period, in contrast to the kinetic experiments described below, where LSF was added subsequent to the preincubation (Fig. 6). Therefore, increasing the efficiency of the basal promoter, due apparently to the increased binding affinity of TFIID, had little effect on the ability of LSF to activate transcription.

LSF Has No Effect on the OnlOff Rate of
TFIID Binding to the Promoter Sequences-Given the proximity of the LSF binding site relative to the SV MLP TATA-like sequence and previous reports that upstream activating proteins might target TFIID (7,(14)(15)(16)(17)(18)(19)(20), a direct analysis was undertaken of the effect of LSF on binding of TFIID to DNA. These studies were performed with TFIID purified from HeLa cells, as bacterially produced human TFIID will not support activation by LSF in vitro (data not shown). TFIID purified as described (see "Materials and Methods") bound DNA to form specific complexes that migrated upon nondenaturing gel electrophoresis in positions comparable to those observed by others (see below and Refs. 47 and 48). In addition, the purified HeLa TFIID produced a characteristic DNase I footprint on the Ad MLP, including protection of the TATA sequences as well as hypersensitive cleavages covering adjacent regions (8) (data not shown).
Incubation of TFIID and TFIIA with a radiolabeled DNA fragment from the SV MLP pSVS3181 TATA consensus mutant and analysis of the complexes by electrophoresis through nondenaturing gels resulted in a major species (Fig.  4 A , lunes 1 and 5 ) that was efficiently competed by an excess of consensus TATA sequences (3181 TATA, lune 7), minimally competed by SV40 wild-type TATA sequences (wt TATA-like, lune 6), and resistant to competition by an unrelated DNA sequence containing an LSF binding site (GC123, lune 8 ) . Addition to the binding reaction of TFIIB or a, the rat TFIIB analogue (33), resulted in the generation of new complexes DAB and DAa (lunes 2 and 3, respectively), which migrate more slowly than the DA complex. The relative migration of the DA and DAB complexes appears similar to that demonstrated previously using HeLa cell factors and the Ad MLP (48).
The LSF DNA complex migrated slightly slower than the DA complex ( Fig. 4 A , lunes 13-16,  complex. In addition, the amount of competitor GC123 DNA was not sufficient to compete for all of the LSF.DNA complexes (Fig. 4A, lunes 12 and 12'), resulting in a low level of LSF .DNA complex that was resolved from the DA complex (Fig. 4 A , lune 12'). The presence of the individual LSF. DNA and TFIID .TFIIA. DNA complexes in the same reaction in the absence of LSF. DA complexes and the nearly saturating amounts of LSF required to generate the quaternary complex suggest that there is no cooperativity of binding between LSF and TFIID.
With identification of these complexes, we could test directly whether LSF might activate transcription either by increasing the rate of binding of TFIID or by decreasing the dissociation rate of bound TFIID. The effect of LSF on the DNA-binding kinetics of TFIID are shown in Fig. 4, B and C. The binding of TFIID alone to the TATA containing SV40 promoter DNA was slow (Fig. 4B, lunes 1-8; DA complex) and began to plateau at about 90 min. In contrast, the rate of binding of LSF was extremely fast and complete by 5 min (Fig. 4B, lune 10). The effect of bound LSF on a subsequent TFIID binding event is demonstrated in Fig. 4B, lunes 9-16. Neither the rate nor amount of TFIID binding to the DNA appeared to be altered by the presence of high levels of LSF. Instead, the rate of accumulation of the LSF. DA DNA complex paralleled that of the DA . DNA complex.
To test the effect of LSF on the stability of TFIID TFIIA. DNA complexes, the rate of dissociation of bound TFIID. TFIIA was measured, both in the absence and presence of LSF. DA. DNA or LSF. DA-DNA complexes were formed by incubation of the proteins and template DNA for 120 min. Subsequently, the competitor double-stranded oligonucleotide containing the consensus TATA sequence was added in an amount empirically determined to prevent formation of new  DA.DNA complexes, but which would not affect complexes that had already formed (data not shown). A time course after the addition of competitor represents the dissociation of TFIID. The rate of dissociation was fairly slow (DA; Fig. 4C, lanes [1][2][3][4][5][6][7][8] and was unaffected by the presence of bound LSF (LSF.DA; Fig. 4C, lanes 9-16). These data, taken together, demonstrate that LSF has no detectable effect on the specific binding of TFIID to DNA, arguing that TFIID is not the functional target for LSF-mediated activation of transcription.
Kinetics of Initiation of the Adenovirus MLP-A kinetic approach was undertaken both to identify the slow steps involved in initiation complex assembly and to investigate the effects of LSF on these steps. Extensive kinetic analyses of transcription initiation complex assembly on bacterial promoters provide a model from which to initiate experiments on eukaryotic transcription complex assembly. A lag in the approach to a steady state rate of initiation complex assembly has been defined as the average time for Escherichia coli RNA polymerase to form an open-promoter complex (49). The time of the lag can represent one or a combination of slow steps in complex assembly and, under certain conditions, may reflect the rate-limiting step in complex formation. Lag time can be determined experimentally by extrapolation to zero product on a plot of product uersus time (49). We have performed a similar analysis on several promoters to define the slow or rate-limiting step(s) in RNA polymerase I1 initiation.
Previously, the basal RNA pol I1 transcription factors have been shown to be able to assemble onto promoter-containing DNA in the following order: TFIID/TFIIA, TFIIB, RNA pol 11, TFIIE/TFIIF (10,ll). Alternative pathways for the assembly may also be possible, with RNA pol I1 addition before or with TFIIB (28). This assembly results in a "rapid-start'' complex (50) capable of immediately incorporating nucleoside triphosphates into specifically initiated, nascent RNA. We began a kinetic analysis using the HeLa transcription factors on the well studied Ad MLP to establish whether the approach would distinguish separate steps in complex assembly. The Ad MLP template was initially incubated with different subsets of general factors, based on the sequential assembly indicated above (Fig. 6A). This preincubation was followed by the simultaneous addition of the complement of general factors and radiolabeled nucleotides a t low concentrations (Fig. 6A) and a subsequent incubation for short lengths of time. Finally, transcripts were fully elongated during an incubation with high concentrations of unlabeled nucleotides. Complete initiation complexes cannot form unless all of the general factors are present. Therefore, in reactions lacking one or more of the factors during the preincubation, the observed level of transcripts reflects the number of rapidstart complexes formed by the different time points, or the rate of completion of initiation complexes, following the addition of omitted factors. Thus, the level of transcription is correlated with the rate of assembly of the rapid-start complex. This is in contrast to the experiments of Fig. 3, where the level of transcription reflected the number of initiation complexes formed during a long preincubation step and correlated with the overall stability of the complex. When all general factors were preincubated with DNA, there was a rapid and linear transcriptional response to the addition of nucleotides (Fig. 5A, lanes 9-12 and open squares,  Fig. 5B). Extrapolation of the curve to zero product a t zero time demonstrates no lag in initiation on preformed complexes, as expected (28,49). When no factors were preincubated with the DNA, there was a significant lag in the rate of initiation complex assembly (Fig. 5A, lanes 13-16 and closed  circles, Fig. 5 B ) . Preincubation of the template DNA with TFIIA and TFIID reduced the lag slightly (Fig. 5A, lanes 1-4  and open circles, Fig. 5 B ) , indicating that the binding of TFIID is a slow process that can be overcome by prebinding. Additionally, preincubation of TFIIB substantially affected the linearity of the response (Fig. 5A, lanes 5-8 and closed squares, Fig. 5 B ) , indicating that another slow step had been overcome by the preincubation. These data could be interpreted in two ways: either the binding of TFIIB is an additional, slow step or TFIIB stabilizes the TFIID complexes formed on the DNA during the preincubation. To discriminate between these possibilities, another experiment was performed in which a (the rat analogue of TFIIB) was either preincubated with TFIID and TFIIA or added a t increasing concentrations subsequent to preincubation with TFIID and TFIIA and the DNA. The striking observation (data not shown) was that at sufficiently high levels of a added with nucleotides, the kinetics of product formation were identical with those observed when a was present during the entire incubation. The ability to abolish the slow step with high concentrations of highly purified a -" -" "  Fig. 6A, except that a 19-min rather than a 20-min sample was analyzed. Transcription reactions were performed using the Ad MLP containing plasmid, pML(C2AT)19A-50 as template (7), as described under "Materials and Methods." The preincubation reactions contained TFIIA and TFIID (lanes 1-4) indicates that rather than stabilizing the TFIID .DNA complex, the actual binding of a is the slow step under normal reaction conditions. Thus, binding of TFIIB under these conditions is slow and may be rate-limiting.
In studies of bacterial promoter initiation kinetics (49), reactions that received no preincubation eventually reached the same steady state rate of product formation as those that had been preincubated, i.e. the curves at late time points became parallel. In the experiments presented here, we have limited the analysis to early time points where the data are the most illuminating and, due to considerations of factor stability, are the most reliable. However, similar to what was observed in the experiments with E. coli RNA polymerase, the curves in Fig. 5B representing reactions that received an incomplete set of factors in the preincubations (closed sqwres, open circles, closed circles) are becoming parallel with the curve for the reaction that received all factors in the preincubation (Fig. 5B, open squares). Reinitiation, the repeated entry of RNA pol I1 into a rapid start complex on a single template, may be contributing somewhat to the level of transcripts detected in these assays. However, reinitiation levels would be low, especially at the early pulse time points due to the low concentration of nucleotides added during the pulse labeling stage of the reaction (9) and to the relatively slow rate a t which reinitiation takes place (51). The interpretation of these data do not require that the reaction be limited to a single round of initiation. The increase in slope of the curves a t later time points is evidence that the reactions are approaching a steady state rate of initiation, i.e. an average of the rates for all events related to production of signal (from formation of the rapid start complex through limited elongation, promoter clearance, and reinitiation). It has been demonstrated that after "promoter clearance'' by RNA pol 11, the only basal transcription factor detected at the promoter is TFIID (52). Therefore, in order for reinitiation to occur, the slow TFIIB binding step must be repeated. This event and subsequent completion of the rapid start complex are thus indistinguishable kinetically from the first round of initiation on that template molecule (in the case where TFIID is prebound).
These results demonstrate that the rate of completion of initiation complexes is sensitive to and depends upon which factors are present in the preincubation step and validates the approach for examining individual steps in RNA pol I1 initiation complex assembly.
LSF Increases the Rate of TFIIB Binding to the Committed Template-A similar kinetic analysis was performed on the basal SV MLP in order to investigate the generality of the rate-limiting steps in initiation at basal RNA pol I1 promoters. In parallel, LSF was added following preincubation of the different sets of factors to determine at which step LSF exerts its effect in the assembly of the rapid-start complex (Fig. 6 A ) . The effect of the strength of the TFIID binding site on the mechanism of LSF activation was analyzed by comparing the rates of complex assembly on the wild-type SV MLP (in pSVS, Fig. 6 B ) , to those on a mutant SV MLP containing the consensus TATA sequence (in pSVS3181, Fig. 6C).
Experiments performed in the absence of LSF delineated the slow steps in the basal initiation process on the SV40 late promoter. The TATA consensus mutant was transcribed at higher levels after preincubation with TFIIA and TFIID than after preincubation with no factors (closed circles, Fig. 6C, panel ii versus panel i ) , consistent with template commitment and preferential transcription due to the binding of TFIID (10). After preincubation of the TATA consensus mutant with TFIIA, TFIID, and TFIIB, there was a significant re-  (closed circles, Fig. 6C, panel iii), compared to preincubation with TFIIA and TFIID alone (closed circles, Fig. 6C, panel ii), which indicates the elimination of a slow step in complex assembly. As argued above for the Ad MLP, this slow step is due to binding of TFIIB. Preincubation with TFIID and TFIIA resulted in no detectable effect on the rate of basal transcription from the wild-type promoter (closed circles, Fig. 6B, panel ii versus panel i). Presumably, commitment on the wild-type template was not observed because of the 17-fold lower affinity of TFIID to this promoter. However, as with the TATA consensus mutant promoter, preincubation with TFIIA, TFIID, and TFIIB on the wild-type promoter (closed circles, Fig. 6B, panel iii) was sufficient to allow sub-stantial accumulation of the TFIIA. TFIID . TFIIB . DNA complexes and to significantly increase the rate of rapid-start complex assembly (greater than 9-fold and 6-fold transcriptional increases on the wild-type and the TATA consensus promoters, respectively, at 5 min; Fig. 6, B and C, panels iii versus panels ii). Thus, the slow assembly rates of basal initiation observed in panels i and ii were due, at least in part, to the slow binding of TFIIB.
The rates of basal complex assembly for reactions preincubated with TFIIA, TFIID, and TFIIB were very similar to rates of assembly observed when all general factors were present in the preincubation (closed circles, Fig. 6, B and C, panels iv versus panels iii; no difference on the wild-type promoter and 1.5-fold higher with all the general factors on the TATA consensus promoter). These data indicate that the slow steps in complex assembly were completed during the preincubation with TFIIA, TFIID, and TFIIB, and that the subsequent associations of RNA pol 11, TFIIE, TFIIF, and TFIIG were relatively fast.
When LSF was added along with the complement of general factors and nucleotides (Fig. 6A) to reactions preincubated with no factors (open circles, panel i of Fig. 6, B and C ) , with TFIIA and TFIID (open circles, panel ii of Figs. 6, B and C), or with TFIIA, TFIID, and TFIIB (open circles, panel iii of Fig. 6, B and C ) , the rates of initiation of transcription on both promoters were highly elevated, in comparison to reactions that received no LSF (e.g. a 30-fold increase on the wildtype promoter Fig. 6B, panel i, and a 10-fold increase on the TATA consensus promoter, Fig. 6C, panel i, at 10 min). The degree to which LSF stimulated transcription varied a t every time point in the reaction, as well as from reaction to reaction (see numbers below). Thus, LSF cannot be simply causing an increased rate of initiation from preformed transcription complexes. Instead, LSF must necessarily be dramatically increasing the rate at which the rapid-start complexes are formed and recruiting new complexes competent for initiation. Given that the basal promoter analysis established that both the TFIID and TFIIB assembly steps limit the rate of initiation complex assembly, the rapid activation observed in the presence of LSF indicated that LSF is accelerating the assembly of TFIID and/or TFIIB. Preincubation with TFIIA and TFIID did not affect the LSF-activated rate of initiation complex assembly on the wild-type promoter and only slightly increased it on the TATA consensus promoter (open circles, Fig. 6, B and C, panels ii uersus i). Since there was little reduction in the lag upon preincubation with TFIID (open circles, panel ii compared to panel i ) , it is unlikely that LSF is targeting TFIID alone. In agreement, the gel band mobility shift analyses of Fig. 4 provide overwhelming evidence that the binding of TFIID is unaffected by LSF.
Two significantly slow steps in the assembly of RNA pol I1 initiation complexes have been demonstrated: the binding of TFIID to the DNA and the assembly of TFIIB onto the TFIID TFIIA .DNA complex. By the kinetic analysis described above and by direct binding analysis (Fig. 4), LSF has been shown to have no detectable effect on the binding of TFIID. Therefore, by the process of elimination, in order to achieve high rates of initiation in the absence of preincubation, LSF must accelerate the subsequent slow step, the binding of TFIIB, to the TFIID . TFIIA. DNA complex. Consistent with this, preincubation with TFIIB significantly reduced the need for LSF in the production of relatively high levels of transcription (only 4-fold activation by LSF in Fig.  6B, panel iii, compared to 30-fold in Fig. 6B, panel ii, a t 10 min). Following preincubation with TFIIA, TFIID, and TFIIB, the rate of initiation complex assembly on the wildtype promoter in the presence of LSF still demonstrated a substantial lag, whereas on the TATA consensus mutant, the rate was strikingly linear after a minimal lag of 1-2 min (open circles, compare Fig. 6, B and C, panels iii). The long lag on the wild-type promoter presumably reflects the low levels of TFIIA. TFIID . TFIIB complexes formed on this promoter during the preincubation and is due to the slow formation, even in the presence of LSF, of the TFIID .DNA complex. On the TATA consensus promoter, the slow TFIID binding step (9, 10, 28) has presumably largely been completed, resulting in only a minimal lag. The data are consistent with LSF causing the rapid association of TFIIB to those relatively stable TFIIA. TFIID. DNA complexes on the TATA consen- sus promoter that were formed during the preincubation but that had not yet bound TFIIB (see Fig. 7). Thus, the substantial lag on the wild-type promoter (Fig. 6B, panel iii) again suggests that LSF does not have a stimulatory effect on the rate of binding of TFIID to the promoter.
Preincubation of both the wild-type and TATA consensus promoters with either all of the general factors (Fig. 6, B and C, panels iu), or with only TFIIA, TFIID, and TFIIB (Fig. 6, B and C, panels iii), resulted both in similar extents of activation by LSF and in similar shapes of the curves. This indicates that the major effect of LSF in increasing the rate of initiation complex formation is on steps prior to addition of TFIIE/TFIIF/TFIIG and RNA pol 11. The only minor difference was that preincubation of the wild-type promoter with all of the general factors (Fig. 6B, panel iu) rendered the reaction relatively unresponsive to the addition of LSF in the first 5 to 10 min of the pulse, compared to the reaction preincubated with TFIIA, TFIID, and TFIIB (Fig. 6B, panel  iii). However, the reaction in which template was preincubated with all general factors did respond to the addition of LSF upon a 20-min pulse, where a greater than 4-fold increase in transcription was observed over that in the reaction receiving no LSF (Fig. 6B, panel iv). In contrast, the TATA consensus promoter responded similarly to LSF when preincubated either with TFIIA, TFIID, and TFIIB (Fig. 6C, panel iii) or with all of the general factors (Fig. 6C, panel io). The apparent delayed response to LSF on the wild-type promoter may be due to the formation of nonproductive initiation complexes during the preincubation at cryptic TATA sequences on the wild-type SV MLP DNA template. These would be less likely to form on the TATA consensus template, which would instead form productive initiation complexes at the authentic high affinity TFIID binding site (TATA-box) in the SV MLP sequence of pSVS3181.

DISCUSSION
cis-and trans-Components of the SV MLP-In view of our demonstration that the general factor requirements of the SV MLP and Ad MLP are similar (Fig. 2), it might be expected that the two basal promoters would have similar functional elements. A comparison of the DNA sequences within the SV MLP and the Ad MLP, however, reveals no striking similarity (Fig. 1). The basal adenovirus promoter consists of a TATAbox to which TFIID binds (7,8) and an initiator element (Inr) (13,53) for which no known binding protein has been found. The sequences upstream of the TATA-box and downstream of the Inr element can be deleted without significantly affecting the level of basal transcription from the promoter. It has been suggested, however, that the G-rich sequences between the TATA-box and the Inr element may influence the efficiency of TATA function, presumably at the level of binding of TFIID (46, 54).
SV40 late promoter elements have been characterized both in vivo and in uitro and consist of multiple upstream regula-tory regions (for a discussion see Ref. 24) and at least one downstream element (22,23). Analyses of SV40 transcription have been performed either in vivo or in crude extracts which contain many SV40 regulatory proteins, thus the distinction between regulatory versus minimal or basal elements has not been made. The TATA-like region of the SV MLP is a presumed basal element (Fig. 1) and was shown to be important for SV40 late transcription in vitro in unfractionated cellular extracts (25). Due to i) the sequence analogy with the Ad MLP, ii) the efficiencies of binding of TFIID to the wildtype SV MLP TATA-like sequence and pSVS3181 TATA sequence in gel mobility shift assays correlating with efficiencies of wild-type SV MLP and pSVS3181 basal transcription, respectively, and iii) the requirement of SV MLP transcription for TFIID (Fig. 2), it can be concluded that the wild-type SV MLP does indeed contain a sequence that is functionally recognized by TFIID. In addition, the SV MLP contains a sequence at the SV40 L325 start site (Fig. 1) that is very similar to the initiator element from the dihydrofolate reductase (DHFR) promoter (31), which appears to operate similarly to the Ad MLP Inr (13) in specifying the start site of transcription. Thus, despite the obvious differences in sequence, the basal adenovirus and SV40 major late promoters appear to be functionally similar.
Mutational Analysis of the S V MLP TATA-box-Binding of TFIID to the promoter-template DNA is the first step in initiation complex assembly (9,ll) at the Ad MLP and serves to commit that template to specific transcription (9, 10). It is not known specifically how TFIID exerts its effect on initiation once it is bound to the DNA. The position dependence of the TATA-box suggests that the bound TFIID may serve as a molecular magnet to attract the remaining general factors and RNA polymerase I1 to the proper positions surrounding the start site of transcription. Since TFIID is the only general factor known to bind DNA specifically and TFIID binding to the TATA-box is the first step in complex formation, TFIID has been considered a likely target for the action of upstream activating proteins. Several studies have suggested that the binding of TFIID in particular may be affected. Cooperative interactions (7) and direct protein-protein interactions leading to both qualitative and quantitative changes in TFIID binding characteristics have been described for several activators (15)(16)(17)(18)(19)(20). In contrast, the transcriptional activator Spl was found to have no effect on the DNA-binding activity of yeast TFIID (55).
The results of mutational analysis of the SV MLP TATA region (Fig. 3) are consistent with previous reports (45) that basal promoter efficiency is dictated largely by TFIID binding affinity. Interestingly, increased TFIID binding affinity and, therefore, more efficient basal transcription did not abrogate the ability of LSF to activate transcription on the SV MLP.
Direct evidence that LSF has no detectable effect on the binding of TFIID was provided by measuring the rates of association and dissociation of TFIID and DNA in the absence and presence of LSF. No apparent effect of bound LSF on either measurement was evident. These results are consistent with our model (Fig. 7) that LSF has no detectable direct effect on the binding of TFIID. Subtle changes in the structure of the TFIID .DNA complex which could be significant in terms of complex function, however, cannot be ruled out completely. These might not be detectable in the gel band mobility shift assay.
Model for the Mechanism of LSF Transcriptional Actiuation- Fig. 7 summarizes our findings on both the SV MLP and the Ad MLP in a flow scheme of the molecular events leading to initiation of transcription. The binding of the TATA-box factor, TFIID, along with TFIIA, is relatively slow (9,55) (Figs. 4B and 7). Depending on the DNA sequence, the association is either stable (TATA consensus mutant SV MLP, Ad MLP) or unstable (wild-type SV MLP). The contribution of TFIIA to the formation of a complex of TFIID with the DNA is not clear, but it is likely to stabilize the complex (9,48): We find that the transcriptional activator LSF has little or no effect on this step in assembly of initiation complexes, because the rate of new complex assembly is slow on a promoter with a weak TFIID binding site, even in the presence of LSF (Fig. 6B, panel iii). In addition, prebinding of TFIID has little effect on transcriptional activation by LSF (Fig. 6, panels i versus ii).
Instead, we demonstrate that the subsequent step in initiation complex formation, the binding of TFIIB to the TFIIA .
TFIID DNA complex (11) (Fig. 7), is also a slow and perhaps rate-limiting step in initiation complex formation. Indeed, prebinding of TFIIB greatly enhanced the rate of complete initiation complex formation in the basal assay, indicating that the assembly of TFIIB onto the promoter is a prerequisite for high rates of complex assembly. LSF increases the rates of complex formation 30-fold in reactions that have not been preincubated with TFIIB. Thus, LSF allows the reaction to rapidly overcome the slow, but prerequisite TFIIB binding step. The data are consistent with a model (Fig. 7) whereby LSF activates transcription by facilitating TFIIB assembly into the initiation complex. In particular, the model invokes that LSF increases the rate of binding of TFIIB to the complex (k4 is greater than k2). By mass action arguments, driving the TFIIB binding reaction forward would also increase the number of TFIIA.
TFIID . TFIIB . DNA complexes (e.g. increase the overall association binding constant for the complex). Thus, it is totally consistent that LSF could both increase the rate of complex formation as evidenced by the kinetic assays (Fig. 6) and also increase the number of complexes formed in the assays that reflect complex stability (Fig. 3). Prebinding of TFIIB dramatically decreases activation by LSF (see legend to Fig. 6), but does not totally abolish the LSF effect because TFIIB is not totally bound in the preincubation reaction. A molecular interpretation of the kinetic studies suggests that LSF may exert its effect by directly interacting with TFIIB and/or with the TFIIA . TFIID . TFIIB .DNA complex. This could occur either by direct and continuous interaction of LSF with the complex or by induction of a conformational change through direct interaction with one or more of the proteins in the complex. If reversal of a new conformation in the complex were sufficiently slow, then the immediate requirement for LSF would be transient, as has been suggested for ATF (15,16). However, it is likely that LSF remains closely associated with the template. Binding of LSF to the template is relatively fast and is complete in less than 1 min (Fig. 4B)4; whereas the half-time for dissociation of the LSF.DNA complex is relatively long, being 8 min. 4 We have outlined the most logical, simple model (Fig. 7) that is consistent with all of the data presented. More complicated scenarios are possible, but it is striking that invoking an increased k4 versus k2 is sufficient to explain all of our observations. Of course, the interpretation is limited by the fact that the TFIIB used in most of the experiments is not homogeneous. However, to date, only one basal promoter activity has been isolated from a TFIIB fraction analogous to that used for these studies, although further purification steps were enlisted (27,48). In addition, we have successfully sub-D. Hawley, personal communication. ' R. Hung and R. Sundseth, unpublished observations. stituted a highly purified fraction of the rat analogue of TFIIB, a, for both basal and LSF-activated transcription (data not shown), which strongly argues that there is only one relevant factor in our TFIIB preparations. One prediction of the model is that the rate-limiting binding of TFIIB should be overcome by addition of saturating levels of TFIIB after the preincubation with TFIIA and TFIID alone, since the rate of a binding reaction is directly proportional to the concentration of substrate. Addition of saturating levels of rat (Y eliminated the lag in initiation on the Ad MLP due to the slow binding of TFIIB (a)(data not shown), demonstrating that the slow step detected by the kinetic assays is in fact the binding event and not a subsequent isomerization or conformational change, the rate of which would be independent of concentration. A kinetic analysis of initiation events on the Ad MLP using an approach analogous in design to those presented here also suggests that the binding of TFIIB is relatively slow (28). TFIIB and RNA pol I1 have been reported to interact weakly in solution (27); thus, the positioning of TFIIB in the initiation complex may be a key prior step in the recruitment of RNA pol I1 onto the promoter (11). Given that TFIIB does not appear to remain at the promoter after initiation (52), LSF may play an important role in stimulating reinitiation by recruiting TFIIB.
Attempts to directly measure the increased rate of binding of TFIIB by the presence of LSF were thwarted, because direct measurement of formation of the TFIID .TFIIA. TFIIB 'DNA complex by gel band mobility shift analysis requires the use of high levels of TFIIB that are effectively saturating in the binding reaction. Therefore, as would be predicted, the measured rate of TFIIB assembly onto the TFIID TFIIA complex was very fast (less than 1 min, data not shown and data cited in Ref. 48). This is not inconsistent with our transcription assay results, however, demonstrating slow assembly of TFIIB, as these experiments were done under conditions of limiting TFIIB where the rate of binding of TFIIB is expected to be slower. In vivo DNase I footprint analysis of some active promoters demonstrate that only TATA-box occupancy, presumably by TFIID, is detectable within the basal promoter (56-58), suggesting that in the steady state, the remaining initiation factors and RNA pol I1 are not bound. The reason for minimal promoter occupancy could be manyfold, but is likely to be the result of limiting amounts of factors in the cell. Thus, the ability of an activator such as LSF to nucleate complete initiation complex assembly would represent a sensitive mechanism for attracting these limiting factors ultimately leading to activation of the promoter.
Recently, a new class of proteins termed adaptors or coactivators has been theorized to physically link upstream activators to the RNA pol I1 general transcription factor machinery and to mediate their activation function (59-65). The possibility that activation of transcription by LSF may require an adaptor is not inconsistent with the results presented here. An "LSF adaptor" could simply mediate the ability of LSF to increase the rate of TFIIB binding and serve as a stabilizing force for the TFIIA TFIID . TFIIB . DNA complex.
The kinetics of initiation complex assembly on the basal Ad MLP and SV MLP are strikingly similar, suggesting that these basal promoters function by similar mechanisms. This is not surprising since the two promoters require the same set of general transcription factors, and, though different in sequence, appear to contain similarly functioning basal elements. We conclude that the mechanism for basal promoter function may be similar for many RNA pol I1 promoters, with the binding of TFIIB playing a major role in limiting the rate of initiation. The control point of initiation of basal transcription by transcriptional activating factors could be at any slow step along the pathway of assembly of the rapid-start complex. The SV MLP transcriptional activator, LSF, appears to activate this process at one of its slowest steps, thereby providing the means to rapidly and efficiently induce initiation of transcription from an otherwise inefficient promoter.