GA-binding Protein-dependent Transcription Initiator Elements

Many eukaryotic RNA polymerase II promoters contain initiator elements which direct accurate transcription in a TATA-independent manner. The PEA3/Ets-binding site (PEA3/EBS) is a common enhancer element in eukaryotic genes and is also found near the transcriptional start sites of many TATA-less promoters. We demonstrate that two PEA3/EBSs driving expression of the luciferase reporter gene, function as a minimal transcriptional initiator element. Maximal levels of transcription was achieved when two PEA3/EBSs, in either orientation, were located on the same face of the DNA helix, and the sites could be separated by up to three helical turns. In vitro transcription start sites directed by PEA3/EBS elements were clustered on either side of the upstream PEA3/EBS and were abolished by immunodepletion of GA-binding protein (GABP) from FM3A cell nuclear extracts. In vivo, co-transfection of GABPα and GABPβ expression vectors enhanced reporter gene expression driven from PEA3/EBS initiator elements. Like other initiator elements, the PEA3/EBS elements were activated synergistically by upstream Sp1-binding sites. Thus, our results establish GABP as both a transcriptional activator factor and as an initiator factor.

Most eukaryotic RNA polymerase II (RNPII) 1 promoters contain a TATA motif bound by the general transcription factor TFIID, which directs the transcription start site to a position 28 -31 nucleotides downstream (1)(2)(3)(4)(5)(6)(7). TFIID binds to the TATA motif during formation of the preinitiation complex, followed by binding of additional general transcription factors (TFIIA, TFIIB, TFIIE, etc.) and RNPII. This preinitiation complex is capable of initiating transcription at a basal level in the absence of transcriptional activator proteins bound to distal sites from the TATA box. In the presence of such activators, high levels of transcription initiation are induced through mechanisms which include, stabilization or facilitation of binding of one or more of the general transcription factors to the pro-moter, modification of chromatin structure to increase access of the general transcription factors and RNPII to the promoter, and direct activation of one or more activities (including phosphorylation) required for the activation of RNPII (7)(8)(9)(10)(11).
A subset of RNPII promoters lack a discernable TATA motif. These "TATA-less" promoters typically initiate transcription from one or several closely spaced sites and also direct high levels of transcription in the presence of transcriptional activators such as Sp1 (1,5,12). While TATA-less promoters vary substantially in the DNA sequence near the region of transcription initiation, many contain initiator elements (Inr) which have been grouped into several families including the terminal deoxynucleotidyltransferase, porphobilinogen deaminase, dihydrofolate reductase, ribosomal protein gene, and the adenoassociated virus p5 Inrs (1,6,(13)(14)(15)(16)(17)(18)(19)(20). Inr-dependent promoters require TFIID for accurate transcription to occur, although, unlike TATA-dependent promoters, TFIID binding to the preinitiation complex is not a rate-limiting step (1, 4, 6, 18 -20). The mechanisms by which Inr-binding factors such as YY1, TFII-I, and E2F, activate TATA-independent transcription have not been well characterized (13,17,21). The transcriptional activator Sp1 is capable of recruiting TFIID to the terminal deoxynucleotidyltransferase Inr, and has been shown to participate in transcriptional start site selection in the TATAless promoter of the hamster carbamoyl-phosphate synthase (glutamine-hydrolyzing)/aspartate carbamoyltransferase/dihydroorotase gene (22).
The GA-binding protein (GABP), which was originally identified as a transcription activator important in the regulation of herpes simplex virus immediate early gene expression, has been shown to activate transcription from the PEA3/EBS-containing initiator elements from the COXIV and COXVb genes, implying that GABP functions as both a transcriptional activator and initiator factor (25,30,41). GABP, also known as nuclear respiratory factor 2, is composed of two distinct sub-units, GABP␣ and GABP␤. GABP␣ is an Ets-related protein containing the 85-amino acid DNA-binding domain (Ets domain). GABP␤ contains four ankyrin repeats at its NH 2 terminus, which mediate dimerization with GABP␣, and a COOH-terminal leucine zipper which mediates GABP␤ homodimerization as well as GABP␣ 2 ␤ 2 tetramer formation (12,47,48).
The structures of PEA3/EBS Inr elements vary with respect to the orientation and spacing of the PEA3/EBS motifs. For example, the initiator elements from the COXIV and COXVb genes contain two PEA3/EBS motifs spaced two and three helical turns apart, respectively, their orientations are different, and they direct transcriptional initiation at different positions relative to the PEA3/EBS motifs. In addition, each initiator element contains a different complement of associated transcription factor-binding sites, including Sp1 and YY1, located upstream and between the two PEA3/EBS sites. To better characterize the role of the PEA3/EBS motifs in initiator activity, we constructed synthetic initiator elements which contained PEA3/EBS motifs only, upstream of the firefly luciferase (Luc) reporter gene (49), and measured their expression in mouse 3T3 cells. We demonstrate that two PEA3/EBS motifs function as a bidirectional initiator element, and that maximal activity is dependent on two PEA3/EBSs positioned on the same face of the DNA helix.

EXPERIMENTAL PROCEDURES
Cloning of Synthetic PEA3/EBS Initiator Constructs-PEA3/EBS motifs were cloned at the unique XhoI site in the multiple cloning sequence upstream of the promoter-less and enhancer-less firefly luciferase reporter gene in pGL2-Basic (Promega) ( Fig. 2A). Synthetic oligonucleotides were synthesized on an Applied Biosystems model 394 or 380B DNA synthesizer. Names of synthetic PEA3/EBS elements indicate the number of PEA3/EBSs, the orientation of the site relative to the orientation of the PEA3 motif in the polyomavirus late promoter (37,38), and the length of the spacer between PEA3/EBSs. Thus, the designation, PEA3-3N8, indicates three PEA3/EBS motifs in the normal (CTTCCT) orientation, separated by 8 bp.
The sequences of the synthetic oligonucleotides used in this study are as follows: Constructs were generated by annealing the appropriate complementary oligonucleotides (e.g. PEA3-1N and PEA3-1R) and cloning the resulting double-stranded fragments in both orientations, into the indicated site of pGL2-Basic ( Fig. 2A). Constructs containing higher multiples of PEA3/EBSs were generated by cloning multiple copies of the above oligonucleotides, in tandem. The resulting sequences are as follows: PEA3-4R8, TCGAGCAGGAAGTCTCGAGCAGGAAGTCTCGAG-CAGGAAGTCTCGAGCAGGAAGTC; PEA3-5R8, TCGAGCAGGAAG-TCTCGAGCAGGAAGTCTCGAGCAGGAAGTCTCGAGCAGGAAGTC-TCGAGCAGGAAGTC.
Cell Culture, Transfection, and Preparation of Cell Lysate-NIH 3T3 cells were seeded at a density of 5 ϫ 10 5 cells/90-mm diameter plates in Eagle's minimal essential medium supplemented with 10% fetal calf serum. After 44 h the cells were given fresh medium and transfected 4 h later (36,37). The test DNAs (20 g) and internal control DNA (pSV-␤-galactosidase expression vector, Promega) (10 g) were co-precipitated with calcium phosphate and added to the plates (12.5-ml medium final volume). After 4 h, the cultures were treated with 15% glycerol for 1 min followed by addition of fresh medium. Cultures were harvested 42 h post-transfection by scraping and the cells pelleted by centrifugation (800 ϫ g for 5 min). Cell pellets were resuspended in 100 l of extraction buffer (100 mM potassium phosphate, pH 7.8, 1 mM dithiothreitol) and lysed by three cycles of freezing and thawing at Ϫ70°C and 37°C, respectively. Cell debris was pelleted and the supernatants were collected and immediately analyzed for luciferase and ␤-galactosidase activities. All constructs were tested in at least three independent transfections, the mean and standard deviations determined for each construct, and the data expressed as the fold activation relative to luciferase activity expressed from the promoter-less and enhancer-less pGL2-Basic vector. All luciferase measurements were normalized for transfection efficiency to ␤-galactosidase expressed from the plasmid pSV-␤-galactosidase.
To determine the effect of expression of exogenous GABP␣ and GABP␤ on the activity of PEA3/EBS elements in 3T3 cells, GABP␣ and GABP␤ cDNAs (kindly provided by Catherine C. Thompson, Carnegie Institute of Washington) were cloned into the XbaI and EcoRI sites, respectively, of the eukaryotic expression vector pcDNAI(Amp). For exogenous GABP␣ and GABP␤ expression, 0.5 g of the pcDNAI(Amp)-GABP␣ and -GABP␤ constructs was co-transfected with the indicated PEA3/EBS initiator constructs and luciferase expression analyzed as described above.
Luciferase and ␤-Galactosidase Assays-To determine luciferase activity in cell extracts, 10 l of the cell lysate were added to 90 l of assay buffer (0.25 mM ATP, 10 mM MgCl 2 , and 100 mM potassium phosphate, pH 7.8) in a 12 ϫ 75-mm cuvette (Analytical Luminescence Laboratory). The cuvette was placed in a Turner model TD-20e luminometer and the reaction was initiated by the injection of 100 l of 1.0 mM luciferin. Luminescence was measured and the raw data converted to relative luciferase activity. ␤-Galactosidase assays were performed by adding 30 l of the cell extract to 270 l of assay buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 1 mM MgCl 2 , 50 mM ␤-mercaptoethanol, and 0.665 mg/ml o-nitrophenyl-␤-D-galactoside), followed by incubation at 37°C for 45 min. The reaction was terminated by addition of 500 l of 1.0 M Na 2 CO 3 and the absorbance was measured at 420 nm.
Nuclear Extract Preparation and Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were prepared essentially as described by Dignam et al. (51), with minor modifications described previously (37,42,52). Protein concentrations were determined by the Bio-Rad protein assay with bovine serum albumin as a standard. EMSA analysis was performed as described previously using FM3A nuclear extract (15 g) (37,42,52). The sequence of the PEA3m probe is 5Ј-TCGAG-CACCTTGAGGAAGTCTCGA-3Ј and the sequence of the dPEA3-0 probe is 5Ј-TCGAGCAGGAAGAGGAAGTCTCGA-3Ј. EMSA probes were prepared as described previously (51), and 0.25 ng of each probe (approximately 50,000 cpm) was used in each EMSA reaction. Recombinant GABP (rGABP) proteins were prepared as described previously (52). GABP␣-and GABP␤-specific antisera were the kind gift of Catherine C. Thompson (12).
In Vitro Transcription-Transcription reactions (50 l) contained 27.5 mM Hepes, pH 7.6, 60 mM potassium glutamate, 3.75 mM MgCl 2 , 0.03 mM EDTA, pH 8.0, 3% (v/v) glycerol, 0.5 mM each of four ribonucleoside triphosphates, 5 mM dithiothreitol, 1 g of DNA template, and 10 l of FM3A cell nuclear extract (50 -70 g of total protein). The transcription reaction was carried out at 30°C for 30 min and stopped by the addition of 100 l of stop buffer containing 20 mM EDTA, pH 8.0, 200 mM NaCl, 1% SDS, and 0.25 mg/ml carrier yeast RNA. 5 l of 2.5 mg/ml proteinase K was added and the mixture was incubated at room temperature. 300 l of 0.3 M sodium acetate, pH 5.5, was added and the solution was extracted once with phenol/chloroform/isoamyl alcohol (25:24:1, v/v/v) and once with chloroform/isoamyl alcohol (24:1, v/v). The resulting aqueous phase was combined with 0.1 pmol of 5Ј-32 P-labeled primer (0.1 pmol/l) and the nucleic acids were precipitated with 1 ml of ethanol. Following centrifugation, the resulting pellet was washed with 75% ethanol (v/v), dried, and dissolved in 10 l of 2 mM Tris-HCl, pH 7.8, 0.2 mM EDTA, pH 8.0, and 0.25 M KCl. This solution was heated at 75°C for 2 min, incubated at 58°C for 30 min, and allowed to cool down to room temperature. 40 l of extension buffer (62.5 mM Tris-HCl, pH 8.3, 1.25 mM MnCl 2 , 125 g/ml actinomycin D, 12.5 mM dithiothreitol) containing, 0.33 mM each of four deoxyribonucleoside 5Ј-triphosphates and 0.5 unit of avian myeloblastosis virus reverse transcriptase (U. S. Biochemical Corp.), were added followed by incubation at 42°C for 1 h. The reaction was stopped by the addition of 300 l of ethanol, the nucleic acids were ethanol precipitated, washed with 75% (v/v) ethanol, dried, and dissolved in 3 l of formamide loading buffer. The nucleic acids were denatured by addition of 1.5 l of 0.1 M NaOH, the solution was boiled for 3 min and loaded onto a 6% denaturing (8 M urea) polyacrylamide gel.

Effect of the Helical Spacing between PEA3/EBS Sites on
Initiator Activity-Close inspection of initiator sequences containing PEA3/EBSs reveals no obvious pattern of position, orientation, or spacing between the sites (Fig. 1). To characterize the effects of orientation and spacing between PEA3/EBSs on initiator activity, synthetic initiators containing one or more PEA3/EBSs were cloned upstream of the promoter-less and enhancer-less firefly luciferase gene in pGL2-BASIC ( Fig. 2A). Constructs with two PEA3/EBS motifs (arbitrarily separated by 8 bp), in either orientation, expressed significantly higher levels of luciferase activity than those with a single PEA3 motif. Constructs with three or more PEA3/EBS motifs were even more active, expressing luciferase activity more than 150fold over background levels (Fig. 2B). In contrast, constructs with two mutant PEA3/EBS motifs (Mt2N7), two octamer sites (Oct2), and the adenovirus major late promoter TATA (Ad-TATA) or Inr (AdInr) elements, failed to activate transcription significantly above background levels.
To determine whether the helical spacing between PEA3/ EBSs affects initiator activity, constructs with two PEA3/EBSs spaced from 0.5 to 3.0 helical turns (Fig. 3A), in both orientations, were tested for initiator activity (Fig. 3B). Maximal luciferase activities were observed in constructs with PEA3/EBSs spaced at 1.1, 2.0, and 3.0 helical turns and were independent of orientation, whereas constructs with two PEA3/EBSs spaced at 0.6, 1.5, and 2.6 helical turns were consistently less active (Fig. 3B). We may conclude from these results that two PEA3/ EBS motifs, in either orientation, are sufficient to function as a transcriptional initiator element, and that maximal transcriptional activity requires two PEA3/EBSs positioned on the same face of the DNA helix.
Mapping of PEA3/EBS-mediated Transcriptional Start Sites-Unlike the terminal deoxynucleotidyltransferase family of Inrs, naturally occurring PEA3/EBS containing initiator elements display no obvious pattern of transcription start site selection. The levels of luciferase transcripts expressed from the basal PEA3/EBS constructs in transfected 3T3 cells was insufficient for mapping of the transcription start sites. 2 Sp1binding sites are often associated with initiator elements, including the PEA3/EBS initiator elements in the murine COXIV and COXVb promoters (23,25,30,31). To enhance the levels of transcription initiation from synthetic PEA3/EBSs for mapping of transcription start sites, we cloned six Sp1-binding sites (SV40 21-bp repeats, SV21) upstream of selected PEA3/EBS initiator elements. Transcriptional activity of the SV21 constructs was determined in vivo by transient transfection into mouse 3T3 cells (Fig. 4). Inclusion of the SV21 sequence with any of the PEA3/EBS elements resulted in synergistic activa- tion of transcription (6 -37-fold) compared with levels in the absence of the SV21 sequence, although, these levels were also insufficient for mapping of the transcription start sites in vivo. 2 However, we were able to detect transcripts by in vitro transcription using FM3A nuclear extracts, and therefore mapped the in vitro transcription start sites by primer extension.
Transcription start sites were observed for each of the SV21-PEA3/EBS elements analyzed and in all cases were mapped to within or near the upstream (5Ј) PEA3/EBS motif (Fig. 5). The SV21-PEA3-2N16 and SV21-PEA3-2N26 elements displayed the strongest intensities, consistent with the basal levels (i.e. in the absence of the SV21 sequence) of these elements in vivo. While similar start sites were identified for the SV21-PEA3-2N10 and SV21-PEA3-2N22 elements, their intensities were much weaker than those observed in the SV21-PEA3-2N16 and SV21-PEA3-2N26 elements, also consistent with the basal activities of these elements in vivo. These results indicate that the synthetic PEA3/EBS initiator elements direct transcription initiation in a similar manner and to the same positions near the upstream PEA3/EBS motif, and that helical spacing between the PEA3/EBSs influence initiator activity.
Many of the PEA3/EBS initiators shown in Fig. 1 are associated with various upstream activator-binding sites, including, Sp1. While differences in the helical spacing between PEA3/EBS sites affected the level of transcription from the SV21-PEA3/EBS elements in vitro, helical spacing had little, or no effect on the position of the transcription start sites (Fig. 5). This is in contrast to the substantial differences in the positions of transcriptional start sites observed in naturally occurring PEA3/EBS initiator elements. These observations suggest that other factors may contribute to start site selection within the context of a particular PEA3/EBS element, possibly including the relative position of the upstream activator-binding site. To determine if the distance between the PEA3/EBS element and the upstream Sp1-binding sites affects transcriptional start site selection, we inserted a 20-bp linker into the KpnI site downstream from the SV21 sequence of the SV21-PEA3-2N26 construct. Increasing the distance between the Sp1 sites and the PEA3/EBS elements by 20 bp (two helical turns) resulted in the appearance of four additional start sites approximately 20 nucleotides upstream of the two major PEA3/EBS-specific start sites (Fig. 5, lane 6). The Sp1-specific start sites are positioned ϳ47 nucleotides downstream from the SV21 sequence, similar to the position of Sp1-directed transcription start sites reported in the hamster carbamoyl-phosphate synthase(glutaminehydrolyzing)/aspartate carbamoyltransferase/dihydroorotase gene promoter (22). These results suggest that Sp1 directs transcription initiation at sites ϳ47-49 nucleotides downstream and that the PEA3/EBS elements direct transcription initiation at sites within or near the upstream PEA3/EBS. PEA3/EBS Elements Are Activated by GABP-Of the dozens of known ETS proteins, only GABP, as a tetrameric complex, binds specifically to two PEA3/EBS motifs (12,47,48). Furthermore, GABP binds to the two PEA3/EBSs in the initiators of the murine COXIV and COXVb, and activates expression from these initiator elements (41,46). These observations strongly implicate GABP as the likely factor mediating initiator activity of the synthetic PEA3/EBS initiators. To confirm that GABP mediates initiator activity of the synthetic PEA3/ EBS initiators, we first measured GABP DNA binding activity in the FM3A nuclear extract used in the in vitro transcription reactions.
EMSA analysis of FM3A nuclear extract revealed three major complexes formed on the dPEA3-0 probe containing two PEA3/EBSs (Fig. 6, lanes 5-7). The fastest migrating complex co-migrated with the complex formed with recombinant GABP␣ (rGABP␣), and the two slower migrating complexes co-migrated with complexes formed by recombinant GABP␣␤ dimers and GABP␣ 2 ␤ 2 tetramers (Fig. 6, compare lanes 1 and  5). Only two complexes were formed on the PEA3m probe containing a single PEA3/EBS, of which one co-migrated with rGABP␣ monomer and the other co-migrated with GABP␣␤ dimer (Fig. 6, compare lanes 1 and 2). The complexes formed with FM3A nuclear extract which co-migrated with recombinant GABP␣␤ dimer and GABP␣ 2 ␤ 2 tetramer complexes were super-shifted by both GABP␣and GABP␤-specific antiserum, but not by preimmune serum. The complex that co-migrated with GABP␣ monomer was super-shifted only by the GABP␣specific antiserum, but not by GABP␤-specific or preimmune serum. Essentially identical results were reported previously using 3T3 nuclear extracts (52). The results presented here and previously (52) establish that the three major PEA3/EBS binding activities in 3T3 and FM3A cells consist of GABP␣-monomer, GABP␣␤-dimer, and GABP␣ 2 ␤ 2 -tetramer complexes.
To establish that GABP is involved in initiator activity, in vitro transcription reactions were performed using the SV21-PEA3-2N26 construct in the presence of GABP␣or GABP␤- specific antisera, or with preimmune sera, (Fig. 7). Addition of preimmune sera had little or no effect on transcription from the SV21-PEA3-2N26 construct, or from the control construct, SV21-TATA, containing the Ad2 major late TATA element downstream from the SV21 sequence. Addition of GABP␣or GABP␤-specific antisera, however, inhibited transcription initiation from the SV21-PEA3-2N26 construct, but had no effect on transcription from the SV21-TATA construct. Thus, inhibition of PEA3/EBS initiator activity by both GABP␣and GABP␤-specific antisera essentially excludes other Ets proteins as initiator factors, since GABP␤ fails to interact with any additional Ets proteins (47).
To demonstrate that GABP is able to activate PEA3/EBS initiator activity in vivo, mouse 3T3 cells were co-transfected with various PEA3/EBS initiator constructs and with pcDNAI(Amp) constructs encoding GABP␣ and GABP␤ cDNAs (Fig. 8). Co-expression of GABP␣ and GABP␤ resulted in a ϳ3-fold activation in the activities of the PEA3/EBS elements, PEA3-2R0, -2R6, -2R10, and -2R16 (Fig. 8), and essentially identical results were observed with PEA3/EBS elements in the "normal" orientation. 2 These results and previously reported results (46) firmly establish GABP as an initiator factor which activates PEA3/EBS containing initiator elements.

DISCUSSION
Many TATA-less promoters contain PEA3/EBSs near their transcriptional start sites, suggesting that factors that bind to these sites may be involved in directing transcription initiation, i.e. that the PEA3/EBS motif can function as an initiator element. Rather than clone and analyze a large number of naturally occurring initiator elements, we chose to systematically characterize the ability of the PEA3/EBS motif to function as an initiator element by constructing a series of synthetic initiators containing two PEA3/EBSs that vary in orientation and helical spacing. These synthetic initiators function in both orientations, initiate transcription from similar start sites relative to the upstream PEA3/EBS site, and are activated synergistically by Sp1. The initiator in the mouse COXIV gene contains two PEA3/EBSs and directs transcription from sites near the upstream PEA3/EBS site, similar to those identified in our studies (25,41).
Since PEA3/EBS-binding sites occur in natural initiator elements in both orientations, it is not surprising that the syn-thetic PEA3/EBS elements also function well in both orientations. It is likely, therefore, that in the absence of additional promoter or enhancer elements, the PEA3/EBS motif functions as a bidirectional initiator element. This is reminiscent of previous observations that the polarity of transcription relative to the TATA element is independent of its orientation in the absence of an upstream activator (53). These studies demonstrated that the major determinant for TATA polarity was the position of a proximal upstream activator-binding site, such as the octamer motif. Similar observations have demonstrated that transcription from the adenovirus 2 major late TATA may occur in either direction depending upon the position of cislinked GC motifs (19,20,54). In contrast to the bidirectional nature of the TATA box and PEA3/EBS elements, the terminal deoxynucleotidyltransferase type Inr sequences appear to only function uni-directionally (54).
Maximal basal transcription from PEA3/EBS elements was obtained when the two sites were positioned on the same face of the DNA helix, whereas substantially lower basal activity was obtained when the PEA3/EBSs were positioned on the opposite face of the DNA. Helical periodicity effects on DNA binding and/or transcriptional activation have been demonstrated previously in both prokaryotic and eukaryotic systems. Cooperative DNA binding by -repressor is a well studied phenomenon and one that is dependent on positioning of the binding sites on the same face of the DNA (55). This spacing is also critical for DNA looping to occur between repressor bound at two distant sites, a mechanism that has also been proposed to explain activation of eukaryotic transcription by activators bound to sites a significant distance from the basal promoters of many genes (4,5,19,20). Sp1, for example, has been shown to cooperatively bind to GC boxes separated by large distances through DNA looping (56). Helical spacing between activatorbinding sites has also been shown to greatly affect transcriptional activation of linked promoters in both prokaryotes (cAMP receptor protein) and eukaryotes (rat prolactin gene), although in the latter case, the effect is not mediated through cooperative DNA binding, rather, these effects seem to be modulated through protein-protein interactions between the activators and/or the basal transcription machinery positioned at the promoter (57-59).
PEA3/EBSs positioned on the same face of the DNA helix may facilitate binding of the GABP(␣␤) 2 tetramer complex, but when PEA3/EBSs are positioned on opposite sides of the DNA, binding of the GABP(␣␤) 2 tetramer complex may be reduced or its structure altered, resulting in reduced levels of transcription initiation. The recent solution structure of the related Ets-1 DNA binding domain-DNA complex revealed that the Ets domain contacts DNA on only one face of the DNA helix. Because the Ets domain is highly conserved in Ets proteins, it is likely that the GABP␣ protein within the GABP␣␤ dimer complex also interacts with DNA on only one face of the helix. Thus, protein-protein interactions necessary for optimal GABP(␣␤) 2 tetramer complex binding would be favored when two PEA3/EBSs align on the same face of the DNA helix (60). Consistent with this notion, preliminary studies show that the DNA-GABP tetramer complex is significantly more stable when two PEA3/EBSs are on the same face of the DNA helix than on opposite sides. 2 While the basal activity of PEA3/EBS elements, in vivo, exhibit a distinct dependence on the helical spacing between the PEA3/EBSs, no such effect is seen with the composite promoters containing six upstream Sp1 sites (SV21 constructs) (Fig. 4). In vitro, however, the SV21 composite constructs display the same dependence on helical spacing as do the basal PEA3/EBS elements in vivo (Fig. 5). Since Sp1 recruits TFIID to the terminal deoxynucleotidyltransferase-type Inr elements (2,18,61), it is possible that the highly active, integrally spaced PEA3/EBS elements also efficiently recruit TFIID, perhaps in a manner similar to Sp1. If this were the case, Sp1-binding sites would be functionally redundant in terms of TFIID recruitment and Sp1 effects would be limited only to transcriptional activation mediated through interactions with other components of the general transcription machinery (62). The less active, halfintegrally spaced PEA3/EBS elements, however, may recruit TFIID inefficiently, perhaps due to diminished GABP binding, but in the presence of bound Sp1, recruitment of TFIID by Sp1 compensates for this deficiency resulting in a greater fold enhancement of initiator activity.
The fact that the activation of PEA3/EBS activity by Sp1 diminished the helical spacing effect in vivo but not in vitro may also reflect the involvement of Sp1 in the de-repression of nucleosome-organized templates (63). The reorganization of the chromatin structure in vivo, by Sp1 binding, may change the chromatin structure in a way which facilitates the binding of GABP, or another factor, to the half-integrally spaced PEA3/ EBS elements, resulting in a greater fold-enhancement of transcription of these templates. The lack of nucleosome repression of the naked DNA templates in the in vitro transcription experiments would result in levels of transcription that are primarily dependent on the intrinsic activity of the basal PEA3/ EBS elements, thus maintaining transcriptional dependence on the helical spacing between PEA3/EBSs. Surprisingly, the helical spacing between PEA3/EBSs had little effect on the position of transcriptional start sites. In all cases, the major start sites were located on either side of the upstream PEA3/EBS site. Changes in the spacing between the upstream Sp1-binding sites and the PEA3/EBS elements in SV21-PEA3-2N26 (KpnI), however, resulted in the appearance of several additional start sites 45-47 nucleotides downstream from the SV21 sequence, but had no effect on the major transcriptional start sites bordering the upstream PEA3/EBS motif. These results suggest that GABP bound to the PEA3/EBSs directs initiation from the major start sites near the upstream PEA3/EBS site. The presence of start sites ϳ47 nucleotides downstream of the SV21 sequence is consistent with previous reports demonstrating that Sp1 is capable of directing transcription initiation in the hamster carbamoyl-phosphate synthase (glutamine-hydrolyzing)/aspartate carbamoyltrans-ferase/dihydroorotase gene promoter (22). Thus, the heterogeneity in the position of transcriptional start sites in the PEA3/ EBS initiators shown in Fig. 1, might be attributable to differences in the relative position(s) of activator-binding sites upstream of these elements. Additional analysis of the effect of other activators on GABP-directed transcription initiation will be needed to confirm whether this is a general effect of transcription activators, or a specific characteristic of Sp1.