Compensatory Energetic Relationships between Upstream Activators and the RNA Polymerase II General Transcription Machinery*

Activation of RNA polymerase II transcriptionin vivo and in vitro is synergistic with respect to increasing numbers of activator binding sites or increasing concentrations of activator. The Epstein-Barr virus ZEBRA protein manifests both forms of synergy during activation of genes involved in the viral lytic cycle. The synergy has an underlying mechanistic basis that we and others have proposed is founded largely on the energetic contributions of (i) upstream ZEBRA binding to its sites, (ii) the general pol II machinery binding to the core promoter, and (iii) interactions between ZEBRA and the general machinery. We hypothesize that these interactions form a network for which a minimum stability must be attained to activate transcription. One prediction of this model is that the energetic contributions should be reciprocal, such that a strong core promoter linked to a weak upstream promoter would be functionally analogous to a weak core linked to a strong upstream promoter. We tested this view by measuring the transcriptional response after systematically altering the upstream and core promoters. Our data provide strong qualitative support for this hypothesis and provide a theoretical basis for analyzing Epstein-Barr virus gene regulation.


Activation of RNA polymerase II transcription in vivo
and in vitro is synergistic with respect to increasing numbers of activator binding sites or increasing concentrations of activator. The Epstein-Barr virus ZEBRA protein manifests both forms of synergy during activation of genes involved in the viral lytic cycle. The synergy has an underlying mechanistic basis that we and others have proposed is founded largely on the energetic contributions of (i) upstream ZEBRA binding to its sites, (ii) the general pol II machinery binding to the core promoter, and (iii) interactions between ZEBRA and the general machinery. We hypothesize that these interactions form a network for which a minimum stability must be attained to activate transcription. One prediction of this model is that the energetic contributions should be reciprocal, such that a strong core promoter linked to a weak upstream promoter would be functionally analogous to a weak core linked to a strong upstream promoter. We tested this view by measuring the transcriptional response after systematically altering the upstream and core promoters. Our data provide strong qualitative support for this hypothesis and provide a theoretical basis for analyzing Epstein-Barr virus gene regulation.
A typical RNA polymerase II promoter contains upstream regulatory elements and a core region encompassing the TATA box, initiator, and downstream sequence elements (1). One of the key challenges in understanding RNA polymerase II gene regulation is deciphering the dynamics of interaction between the upstream and core promoters and how these interactions generate a transcriptional response. To address this issue, our studies have focused on a model system, which is based on a prototypic eukaryotic regulatory switch: the transition of Epstein-Barr virus (EBV) 1 from a latent to a lytic life cycle. The EBV switch from latent to lytic growth in B lymphocytes is initiated by a viral transactivating protein called ZEBRA, which is synthesized in response to extracellular cues and, in turn, activates the expression of downstream target genes to different levels, apparently in a temporally distinct manner (2,3). Results from our laboratory and others have shown that appearance of cytoplasmic viral mRNAs is highly synergistic with respect to ZEBRA concentration (2)(3)(4). 2 ZEBRA (also called Zta or EB-1) is a b-Zip family member bearing an amino-terminal non-acidic activation domain and a carboxyl-terminal basic zipper or coiled-coil domain (5)(6)(7)(8)(9)(10). ZEBRA was originally shown to bind to specific sites upstream of several early genes, including BRLF-1 (Rta, a transcriptional activator), BMLF-1 (Mta, a posttranscriptional activator), BMRF-1 (a polymerase accessory factor), BHLF-1 (a Bcl-2 homologue), and its own gene, BZLF-1 (4,8,(11)(12)(13)(14)(15)(16)(17)(18). Computer analysis of these and other ZEBRA-responsive genes revealed core promoters varying widely in sequence and upstream promoters differing in the number, position, and affinity of ZEBRA binding sites and occasionally, the presence of sites for other regulatory factors. We had previously hypothesized that different promoter geometries or architectures and the synergistic or greater than additive effects of multiple bound ZEBRA dimers are responsible for differential gene expression during the early lytic cycle (19), a hypothesis that we elaborate on in the present study.
The ZEBRA activation domain and its biochemical mechanism have been extensively characterized (5,9,10,20). The activation domain of ZEBRA can be subdivided into a series of uncharged modules rich in hydrophobic amino acids. These modules act cooperatively to enhance the potency of ZEBRA (5,21). Our current view is that ZEBRA stimulates transcription by binding to its upstream sites, either alone or in concert with other EBV (i.e. Rta) and cellular regulatory proteins (i.e. Sp-1); once bound, the activation domain engages in protein-protein interactions with components of the RNA polymerase II general transcription machinery, resulting in assembly of a transcription complex over the promoter. A key biochemical step affected by ZEBRA is recruitment of the general transcription factors TFIIA and TFIID to the core promoter to form the so-called "DA complex." DA complex recruitment correlates with the ability of ZEBRA to stimulate transcription and assembly of open complexes in vitro (22). ZEBRA also induces a conformational change or isomerization in TFIID that plays a key role in activated transcription (23).
In an effort to understand aspects of promoter architecture that govern the timing and extent of gene activation, we developed an in vitro model system that has allowed us to systematically define important aspects of promoter architecture. Previous studies had revealed that several parameters play * This study was supported by Grant MV-547 from the American Cancer Society (to M. C.) 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.
¶ dominant roles in promoter output. These parameters include the number of sites, ZEBRA concentration, TFIID concentration, and potency of the activation domain (5,19,22,23). The underlying theme in each of our previous studies was that the transcriptional response is a function of each parameter and that increasing numbers of promoter sites, ZEBRA concentration, or number of ZEBRA activation modules led to synergistic gene activation. These and other studies led to the view that the transcription complex is a network of DNA-protein and protein-protein interactions between multiple upstream activators and the general transcription machinery (Fig. 1). According to this view, the "activation potential" of a promoter can be defined minimally by three physically and energetically interconnected parameters, each represented by a distinct equilibrium constant (K) and free energy (⌬G): (i) the contribution of the upstream promoter, a function of the number, affinity, and location of activator binding sites; (ii) the contribution of the core promoter, a function of the interaction of TATA and surrounding sequences with components of the general machinery; and (iii) the affinity of activators for the general transcription machinery and the reciprocal effects that interaction has on binding of the activators (see Ref. 24). The total free energy or stability of the complex assembly reaction would be defined by the summed energetic contributions of each component.
A principal yet untested corollary of the aforementioned hypothesis is that the various energies should be compensatory, an assumption that raises two experimentally testable predictions: (i) a strong or potent core contribution would compensate for a weak upstream contribution, and this scenario would be energetically analogous to a weak core adjacent to a potent upstream promoter; and (ii) multiple upstream activators could compensate for a weaker core promoter. To test these predictions, we systematically altered the strength of the upstream promoters by varying either the affinity of ZEBRA sites or their number, while simultaneously sampling several different core promoters varying in their affinity for TFIID and, concurrently, in their basal level of transcription. We show that indeed, core and upstream promoters have compensatory effects on transcription, providing strong support for the hypothesis that transcription complex assembly and stability is defined largely by an entire network of interactions. The implication of this study is that transcription complex assembly can be a concerted process dictated by the concentrations of the interacting components and their affinities for one another.

EXPERIMENTAL PROCEDURES
Biochemical Analyses-Purification of recombinant ZEBRA, in vitro transcription, DNase I footprints, and gel shifts were performed as described previously (19). Gel shifts to determine core affinity for TFIID were done as described previously (21) with the following modifications. 0.5 fmol of 32 P-end-labeled restriction fragments encoding three or seven ZIIIB sites upstream of the E4, M, H, or Z core promoters was incubated in the presence of 67 ng of recombinant TFIIA and titrations of TFIID ranging from 50 to 200 ng. Half of the reactions included 2 ng of recombinant ZEBRA. Following a 30 min reaction, one-quarter of the reaction was loaded onto a 1.4% agarose gel containing 5 mM magnesium acetate and electrophoresed for 4 h at 50 V. The remainder of the reaction was subjected to DNase I footprinting analysis to confirm the specificity of the binding observed by gel shift. The gels were dried and visualized on a Molecular Dynamics PhosphorImager. The DNase I footprints confirmed that TFIID was indeed binding the sites as expected from the gel shift results.
Constructs-Three tandem copies of each ZEBRA-responsive element (ZRE) were cloned 22 bp upstream of the adenovirus E4 core promoter. In the initial step, a 45-b oligonucleotide bearing three tandem 7-bp sites separated by 2 flanking bp was synthesized along with a 15-bp oligonucleotide complementary to the 3Ј-terminus. The two oligonucleotides were annealed, and the single stranded DNA was repaired by the fill-in reaction using Klenow DNA polymerase. The double-stranded DNA fragments were subsequently gel purified and ligated into the HincII site of p⌬-38 (as described in Ref. 19). The resulting constructs, pZRE-n-E4T, were cleaved with HindIII and BamHI, releasing the fragment bearing the ZREs and ligated into HindIII-BamHI cleaved pE4TCAT (25), generating the pZRE-n-E4TCAT series of clones, or into the pMCAT, pHCAT, or pZCAT clones. pMCAT, pHCAT, and pZCAT were prepared by cloning the Ϫ42 to ϩ24 region of BMRF-1 (5Ј-TTCTGGGCATAAATTCTCCTGCCTGCCTCTG-CTCTCTGGTACGTTGGCTTCTGCTGCTTGTGGACT-3Ј), the Ϫ41 to ϩ42 region of BHLF-1 (5Ј-CCAAAAAGAGGATAAAAGAAGGCGAGC-CGGCCCGGCTCGCCAGCGTCGTCCAGACGCTCGGGGGGTGCAC-ACCTCCCAGCCGG-3Ј), and the Ϫ46 to ϩ26 region of BZLF-1 (5Ј-CC-TTGGCTTTAAAGGGGAGATGTTAGACAGGTAACTCACTAAACAT-TGCACCTTGCCGGCCACCTTTGCT-3Ј), respectively, into BamHI and KpnI-digested p⌬-38 (which removes the E4 core promoter). All DNAs were purified twice by ultracentrifugation in ethidium bromide-cesium chloride density gradients.
Transfections and CAT Assays-10 g of each reporter construct was transfected into 10 7 Akata cells using electroporation essentially as described (19). The transfected cells were resuspended in 7 ml of RPMI 1640 medium supplemented with 10% fetal calf serum. After 12 h at 37°C, the suspensions were treated with 700 g of anti-IgG (Sigma) to induce the lytic cycle, and cell extracts were harvested 12 h later. The cells were centrifuged at low speed to harvest the cells and washed three times in phosphate-buffered saline and once in TEN. The cells were then resuspended in 100 l of 0.25 M Tris, pH 7.5, freeze-thawed three times, and centrifuged at 10,000 ϫ g in a microcentrifuge to remove debris. Typical CAT assays were normalized by cell count or by Bradford protein assays and contained 25-50 l of extract, 0.01 Ci of 14 C-chloramphenicol, 25 g of acetyl-CoA in 0.25 M Tris, pH 7.5. After 12 h at 37°C, the mixtures were fractionated by thin layer chromatography. The autoradiographs were quantitated using a Molecular Dynamics PhosphorImager and ImageQuant software.
Quantitation: DNA Affinity Measurements-The 4.5% dried acrylamide gels were exposed and quantitated using a Molecular Dynamics PhosphorImager and ImageQuant software. The amount (volume) of 32 P probe present in each of the shifted complexes was measured and summed. The percentage of each complex relative to the total amount of probe was determined. Percentage of site occupancy was determined by multiplying the percentage of each complex by the fraction of sites occupied ( 1 ⁄3, 2 ⁄3, or 1) for each of the three shifted complexes. These were summed to reveal the final percentage of occupancy. A graph of percentage of probe shifted versus the amount of protein added was constructed. Determination of the protein amounts necessary to occupy 50, 40, 30, 20, and 10% of the probe ZREs was determined from the graphs (for some sites, data were not available for the higher percentage of shifts). The amount of protein necessary to obtain each percentage of shift was determined and compared with the amount of protein necessary to obtain a comparable shift on the ZIIIB site. The affinity of each site for ZEBRA relative to ZIIIB was determined by averaging the relative affinity determined at each of the five values for percentage of shift. Data from four experiments were obtained and averaged.
Quantitation of the affinity of the core promoters was performed by determining the percentage of probe shifted relative to the total amount of probe present at 2-fold increasing concentrations of TFIID ranging from 50 to 200 ngs. In vitro transcription activities were determined by quantitation of the primer extension products using a PhosphorImager. The amount of transcription product in fmol was determined based on a standard curve generated by 2-or 3-fold dilutions of the 32 P-labeled primer electrophoresed alongside each experiment. Results shown are the average of two to four experiments. Basal levels of transcription were subtracted from the amount of transcription activated in the presence of 100 ng of recombinant ZEBRA to take into account the differences in basal level transcription from different promoters. The amount of transcription from each promoter was then compared with the level of transcription elicited by the promoter containing the ZIIIB site. The results presented are the average of four separate experiments. CAT activities were quantitated by determining the percentage of acetylated 14 C-chloramphenicol relative to the total amount of chloramphenicol present in the reaction. The basal CAT activity obtained from the E4TCAT construct containing was subtracted from the activities of the constructs containing each of the different sites.

RESULTS
Experimental Design-Our hypothesis in Fig. 1 posits that within a reasonable range, the core and upstream promoters and the activator-general factor interactions contribute to the transcription complex assembly in a compensatory fashion. To test this hypothesis, we systematically altered the strength of the upstream and core promoters and quantitatively analyzed their output. To alter the contribution of the upstream promoter, we varied both the affinity and the number of ZEBRA binding sites. To alter the core promoter affinity for the general machinery, we compared four different core promoters. All four contained a unique TATA box and surrounding sequences and, as we will show, varied in their affinities for the DA complex in a manner that approximately correlated with basal transcription levels (7,8,26). In a previous study, ZEBRA potency was systematically altered, and we will consider these results under "Discussion." To vary the affinity of the ZEBRA binding sites, we first surveyed the literature to identify known sites. Footprinting and deletional analysis of early lytic genes, including BHLF-1, BMRF-1, BMLF-1, and BZLF-1, revealed seven different sites, ZREs, which generate the following consensus: 5Ј-T(T/G)(T/ A)G(T/C)(C/A)A-3Ј (27). We predicted that if all 16 members of this consensus were synthesized, they would display a range of affinities. Reporter templates bearing three tandem copies of each member, positioned 22 bp upstream of the adenovirus E4 core promoter TATA box (Fig. 2), were compared for i) their affinity for ZEBRA in vitro, and ii) their abilities to support transcriptional activation both in vitro in HeLa extracts and in vivo in B lymphocytes harboring EBV induced into the lytic cycle. To vary the number of sites, we placed 1, 3, or 7 copies of the high affinity ZIIIB site upstream of the different core promoters. We had previously employed the ZIIIB site for other mechanistic studies on ZEBRA action (5,19,21,23,28).
To obtain core promoters varying in affinity for the transcriptional machinery, we compared our standard adenovirus E4 template with the subcloned core regions of several well characterized ZEBRA-responsive genes, BHLF-1 (H), BMLF-1 (M), and BZLF-1 (Z). In all four cases, these core promoters were positioned adjacent to upstream promoters varying in affinity for ZEBRA or number of ZEBRA sites. The strength of a core promoter was measured by the promoter's affinity for the DA complex and by the basal transcription level.
Affinity of ZEBRA-responsive Elements-We first measured the affinity of the ZEBRA sites in vitro and then measured their ability to support a transcriptional response in vitro and in vivo. We found that the 16 ZREs displayed a 20-fold range of affinities for ZEBRA. 32 P-end-labeled DNA fragments bearing three tandem copies of each site were analyzed for binding to ZEBRA in a gel mobility shift assay. The fragments were in-cubated with 2-fold increasing concentrations ranging from 6.25 ng (3.75 nM) to 100 ng (60 nM) of purified recombinant ZEBRA, and the complexes were fractionated on native polyacrylamide gels. Three distinct complexes were observed, corresponding to occupancy of one, two, or all three ZREs. The percentage of probe in each of the ZEBRA-DNA complexes was quantitated by laser densitometry and PhosphorImager analysis and the affinity relative to ZIIIB, the highest affinity ZRE, was determined. Fig. 3 shows a chart summarizing the results from four experiments alongside representative autoradiographs. ZE-BRA does not apparently bind to these artificial promoters cooperatively, and thus, the occupancy of each of the sites is a direct measure of the affinity for a specific element. The high affinity sites (ZIIIB) demonstrated a significant shift at the lowest concentrations of protein used, 3.75 nM dimer, whereas the lowest affinity sites required the highest levels of protein added, 60 nM dimer, to observe a significant shift. We determined the apparent K d of the ZIIIB site to be 15 nM in this experiment compared with our previously measured value of 30 nM (19). This is only an apparent K d because specific oligonucleotide competition, which measures the active dimer concentration, generates a K d value of 0.1 nM. 3 We have not resolved whether this difference reflects the activity of the ZEBRA preparation or the amount of dimer; this issue is not relevant here. It was not possible to demonstrate 50% occupancy for the low affinity sites due to nonspecific binding of ZEBRA at high concentrations. However, by comparing the concentrations of protein necessary to achieve less than 50% occupancy and extrapolating, we estimate the K d of the lowest affinity sites to be 20-fold higher than that of ZIIIB. These results demonstrate that ZEBRA can bind to all 16 of the sites tested and suggest the involvement of a wide range of ZREs in initiation of the EBV lytic cycle. These values are not an artifact of the gel shift assay because DNase I footprints of representative sites, including ZIIIB, AP-1, and ZRE-3, result in relative affinities similar to those determined by gel shift (data not shown). We believe that this wide variation in affinity is a mechanism employed by the virus for differential gene regulation.
Analysis of the 16 ZEBRA-responsive promoters revealed that there is a correlation between the affinity of a ZRE and its ability to support activated transcription both in a HeLa cell 3 1. The threshold hypothesis. A typical promoter contains upstream activator binding sites (X) and a core promoter (Y). Activators stimulate transcription by binding to the upstream promoter and recruiting limiting components of the general pol II machinery (the general transcription complex (GTC)) to DNA. The final transcription complex has an affinity or free energy that reflects the protein-DNA interactions between the activators and its sites (⌬G X ) and between the general machinery and the core (⌬G Y ) and a protein-protein interaction energy between activators and the general machinery (⌬G coop ) that allows preinitiation complex assembly. We think that there is a maximum free energy (⌬G T ) or barrier required to assemble this complex and that energies below that barrier result in cooperative transcription complex assembly. nuclear extract and in vivo in an induced EBV-containing cell line. In vitro transcription reactions were performed in the presence of 2-fold increasing concentrations of ZEBRA protein ranging from 12.5 to 200 ng, and RNA products were analyzed by primer extension analysis. The chart in Fig. 4A illustrates the average of four independent experiments and shows representative results. We found that ZIIIB and ZRE-5, which were shown to be high affinity ZEBRA binding sites (apparent K d ϭ ϳ15 nM), support high levels of activation. For example, on our ZIIIB-responsive reporter template, the transcriptional stimulation was greater than 40 times the basal level at peak con-centrations (100 ng) of ZEBRA. ZRE-1 and AP-1, which were shown to be intermediate affinity sites, activated lower levels of transcription, whereas ZRE-3, ZRE-9, and ZRE-12, which were shown to be low affinity sites, barely activate above basal values. Note, however that ZRE-16 supports much more activity than ZIIIA, despite their very close values for affinities.
We also found that not only are the absolute levels of transcription directed by a specific site dependent upon the affinity of the site for ZEBRA, but the concentration of ZEBRA necessary to "activate" transcription also varies with the affinity of the site. For example, we observed activated transcription at the lowest level of protein added in the case of ZIIIB (12.5 ng), but it was not until the second highest concentration of protein (100 ng) added that we began to see activated transcription in the case of ZIIIA. The slight decrease in transcription observed in the last lane of each set of transcription reactions most likely is due to ZEBRA either binding nonspecifically to DNA or squelching (29).
This same phenomenon occurred in vivo. The 16 ZEBRAresponsive reporters were cloned upstream of the chloramphenicol acetyltransferase reporter gene and transfected into Akata cells, a Burkitt's lymphoma cell line, which harbors latent EBV (30). By transfecting the reporter templates into Akata cells and inducing the lytic cycle, we ensured that the levels of ZEBRA being used were identical to those present during the early lytic cycle. The relative activities of the different sites in vivo were similar to those seen in vitro. Again, a sharp distinction was observed between sites that were able to direct high levels of transcription and those that mediated low levels. A chart summarizing the average of two representative experiments is shown in Fig. 4B alongside the CAT assays performed on extracts obtained from representative transfections performed in the presence and absence of anti-IgG induction. These results demonstrate that the different activities of our collection of ZEBRA-responsive elements are clearly evident when activated by biologically relevant quantities of ZEBRA. Similar experiments were performed in Raji cells, another EBV-containing cell line (data not shown) and were shown to be consistent with the Akata data.
Core Promoter Affinity-To investigate the contribution of the core promoter to the transcriptional potential of a promoter, we determined the affinity of the DA complex for four different natural promoters. The core promoters of BMLF-1 (M), BHLF-1 (H), and BZLF-1 (Z) were cloned downstream of 7 ZIIIB sites and compared with our benchmark, E4 promoter. 32 P-labeled restriction fragments were incubated with increas-

FIG. 3. Electrophoretic mobility shift analysis of ZRE affinity.
The 30-l preincubation mixtures contained 32 P-labeled DNA fragments bearing the three tandem copies of each ZRE and 2-fold increasing concentrations of purified recombinant ZEBRA protein ranging from 6.25 to 100 ng. After 30 min at 23°C, the mixtures were fractionated on 4.5% native polyacrylamide gels electrophoresed in 0.5ϫ Tris-Borate-EDTA and 1% glycerol. An autoradiograph of a dried polyacrylamide gel is shown. Four representative results are shown, demonstrating the range of affinities for ZEBRA. The three arrows indicate the positions of the complexes generated when one, two, or three sites are filled; the free probe is indicated (Free). The summary at the top demonstrates the fractional affinity of all 16 ZREs relative to ZIIIB. Affinities were calculated by determining the percentage of sites filled at several different protein concentrations relative to ZIIIB and averaging the results (see "Experimental Procedures").
ing concentrations of immunoaffinity-purified TFIID (31) and TFIIA in the presence and absence of recombinant ZEBRA. The complexes were resolved on Mg 2ϩ -agarose gels, which are necessary to observe the large DA complexes. Fig. 5 shows a representative experiment revealing a reproducible 4-fold range of affinities in the absence of ZEBRA, with the core promoter from E4 having the highest affinity for TFIID (K d ϭ 4 ϫ 10 Ϫ8 M), and the BZLF-1 core promoter having the lowest affinity (K d ϭ 1.6 ϫ 10 Ϫ7 M). In the presence of ZEBRA, we observed roughly the same rank order of affinities but a nar- FIG. 4. Activity of ZREs upstream of the E4 core promoter. A, in vitro activities of ZREs. A template bearing three tandem copies of each ZRE upstream of the adenovirus E4 TATA and start site was transcribed for 1 h in a HeLa nuclear extract in the presence of 2-fold increasing concentrations of recombinant ZEBRA ranging from 12.5 to 200 ng. Transcription was assayed by primer extension, and the products were fractionated on denaturing 10% polyacrylamide gels electrophoresed in 1ϫ Tris-Borate-EDTA. A representative autoradiograph of the primer extension products is shown alongside a chart of more extensive results. The autoradiograph shows ZEBRA titrations on four representative sites, demonstrating the range of activities and activator responsiveness. Arrows indicate the position of the free primer (Pr) and the bands corresponding to the extension products (EPs). The summary at the left demonstrates the levels of transcription directed by each of the 16 ZREs relative to ZIIIB in the presence of 100 g of purified ZEBRA, the concentration that was demonstrated to produce maximum levels of transcription. Transcriptional activities, averaged from two separate experiments, were determined by quantitating the amount of primer extension product on a PhosphorImager and calculating the level relative to ZIIIB. B, in vivo activities of ZREs. 10 g of reporter templates bearing three tandem copies of each ZRE, 22 bp upstream of the adenovirus E4 core promoter, driving the CAT gene reporter, were transfected into Akata cells, a B-cell line that harbors latent EBV. 12 h following lytic activation, cell extracts were harvested and normalized for protein concentration using Bradford reagent, and overnight CAT assays were performed. Acetylated and unacetylated chloramphenicol were separated by thin layer chromatography. Autoradiographs of the CAT assays from four representative templates are shown. The spots corresponding to the acetylated chloramphenicol (Ac), unacetylated chloramphenicol (Ch), and the origin (Or) are indicated. Percentage of acetylated chloramphenicol was determined by comparing the intensity as determined by a PhosphorImager of the acetylated chloramphenicol spots to the total amount of label present in each reaction. This ratio was used as a measure of the strength of each promoter. The summary describes the strength of each promoter relative to ZIIIB averaged from two separate experiments. rower difference in the affinities of the E4 and H core promoters. A similar range of affinities was observed with core promoters bearing only three sites (data not shown). The DNase I footprint of the DA complex on all four promoters was also somewhat similar (data not shown). Although we found small but reproducible differences in the affinity of the core promoters for the DA complex, it is also possible that these core promoters have additional differences in affinity for other components of the transcription complex that contribute to transcriptional output. In preliminary experiments, we tested the effect of adding TFIIB to reactions containing TFIID, TFIIA, and ZEBRA but observed no significant differential effect.
ctivity of Different Core Promoters with Different Affinity ZEBRA Binding Sites-To investigate the combinatorial interactions between upstream and core promoters, we generated constructs in which high, medium, or low affinity sites were placed upstream of strong or weak core promoters. As our hypothesis predicts, high affinity ZEBRA binding sites were required to generate activated transcription from low affinity core promoters, and high affinity core promoters were required to generate activated transcription from low affinity ZEBRA binding sites.
Three tandem copies of a high affinity ZEBRA binding site, ZIIIB, a medium affinity site, ZRE-16, and two lower affinity sites, ZRE-3 and ZRE-14, were cloned upstream of the E4, M, H, and Z core promoters. In vitro transcription reactions were performed in the presence of 3-fold increasing concentrations of ZEBRA protein, ranging from 22 to 200 ng; RNA products were analyzed by primer extension analysis. The absolute amount of primer extension product in fmol was calculated based on a standard curve generated from the dilutions of the primer used in each experiment. Fig. 6 shows representative data on four templates differing in the affinity for their upstream ZEBRA binding sites and for their core promoters. We found that there was approximately a 6 -8-fold difference in the level of transcription between the core promoters when they were placed downstream of the high affinity ZIIIB or ZRE-16 sites versus the low affinity ZRE-14 and ZRE-3 sites. Furthermore, in agreement with the DA complex affinity studies, the higher affinity H and E4 core promoters exhibited higher basal transcription levels and supported significantly higher levels of transcription than the lower affinity M and Z core promoters.
The transcriptional effects were, however, greater than would be predicted based on the difference in affinity for TFIID. The high affinity ZIIIB sites, for example, produced 7-fold higher peak levels of transcription when placed upstream of the H core promoter than upstream of the Z core promoter. This may reflect differences in the affinity of these core promoters for other components of the transcriptional machinery or differences in other promoter features not identified here. Further, the H core appeared slightly more active transcriptionally than E4, although its affinity for the DA complex was slightly lower. The key result here, however, is the finding that the high affinity sites upstream can compensate for a low affinity core promoter and a high affinity core promoter can compensate for low affinity ZREs. Thus, for example, the ZIIIB sites upstream of the Z core promoter and ZRE-3 cloned upstream of the H core promoter produced 0.89 and 0.150 fmol of product, respectively.
Activity of Different Core Promoters with Different Numbers of ZEBRA Binding Sites-We further characterized the cooperative interactions between upstream and core promoters by Transcription was assayed by primer extension, and the products were fractionated on denaturing 10% polyacrylamide gels in 1ϫ Tris-Borate-EDTA. A PhosphorImager scan of the primer extension products is shown. ZEBRA titrations on four representative affinity sites upstream of the four different core promoters demonstrate the range of activities that were observed. Arrows indicate the position of the free primer (Pr) and the bands corresponding to the extension products (EPs). Transcription activities summarized in the text, averaged from either three or four separate experiments, were determined by quantitating the amount of primer extension product using a Phos-phorImager and calculating the number of fmol based on a standard curve of the free probe. changing the number of upstream binding sites rather than their affinity. Also consistent with our hypothesis, we found that fewer numbers of ZEBRA binding sites are required for activated transcription in the presence of a high affinity core promoter than are required for activated transcription upstream of a low affinity core promoter.
Templates bearing one, three, or seven high affinity ZIIIB sites were cloned upstream of the E4, M, H, or Z core promoters (data from E4 and M are not shown). Fig. 7 shows in vitro transcription reactions in HeLa nuclear extract performed in the presence of 3-fold increasing concentration of ZEBRA protein ranging from 7 to 200 ng. RNA products were analyzed by primer extension; fmol of primer extension products were determined based on a standard curve of the primer. We found that templates bearing three ZEBRA binding sites placed upstream of the high affinity H core promoter were activated at similar concentrations of ZEBRA protein and produced similar or even higher levels of transcription than templates with seven sites cloned upstream of the weak Z core promoter. Surprisingly, we were even able to detect low levels of activation with a single ZEBRA binding site upstream of the high affinity H core promoter, a phenomenon that was not observed on any core promoter previously tested (at physiological concentrations of TFIID). These data suggest that the increased free energy of the DA complex interaction with a high affinity core promoter overcomes the need for multiple ZEBRA proteins for activation, as observed previously (23). DISCUSSION The phenomenon of synergy or cooperativity has formed the framework for understanding differential gene expression in eukaryotic cells. We have been evaluating the mechanism of this phenomenon by systematically altering the properties of ZEBRA and its responsive upstream and core promoters. In this study, we posed the simple hypothesis that the transcriptional output of a ZEBRA-responsive reporter template is an equilibrium process that can be largely defined by a the energetic contributions of the upstream and core promoters. This hypothesis predicts compensatory energetic relationships among the different components of the complex due to the cooperative binding. Because the study was performed primarily in vitro, many of the phenomena do not involve sophisticated higher order chromatin, although such structures may enhance the effects and certainly contribute to thresholds and cooperativity in vivo and in certain in vitro systems (32,33).
Cooperativity-The concept of cooperativity in transcription complex assembly is based on two experimental observations: the synergistic effects of the number of activator sites and activator concentration. Both effects have been previously documented with ZEBRA (19). The synergistic effect of sites was observed under site-saturating conditions of ZEBRA, suggesting that the effect was not due solely to cooperative binding, an effect shown previously with truncated GAL4 derivatives (34). We proposed that multiple bound ZEBRA molecules simultaneously contacted the general machinery resulting in synergistic recruitment and assembly of the transcription complex (34). Similarly, increases in the concentration of ZEBRA also resulted in synergistic gene activation. This result can best be explained if the transcription machinery is promoting cooperative binding of ZEBRA to multiple sites when the concentration of ZEBRA is limiting. We think that the cooperative binding again is a function of several molecules of ZEBRA simultaneously contacting the general machinery in solution and the machinery having a reciprocal cooperative effect on ZEBRA binding. As mentioned above, other mechanisms involving chromatin and kinetic effects could be superimposed on this effect to increase the sensitivity (33,35). For example, one possibility is that there is a time-dependent isomerization necessary for efficient complex assembly and that each dissociation event resets this clock. In such a scenario, the kinetics of binding would become very important determinants of transcriptional regulation.
Although the arrangements and affinities of binding sites in a natural promoter may be somewhat varied, their simultaneous interactions with the general machinery may cause site filling and transcription complex assembly to be a highly concerted process, a hypothesis that we are currently testing. The implications of such a finding are that the strength of neither the core nor the upstream promoters is a relevant variable because it is the cooperative action of both that determines the timing and levels of promoter activity. The theme of synergy and thresholds has been repeated throughout eukaryotes, and such mechanisms have been shown to play a central role during development of Drosophila (36,37).
Energetic Reciprocity and Compensation in Gene Activation-A key prediction of the cooperativity hypothesis is that each of the various energetic components of the equation should, within a reasonable range, compensate for the others to achieve a threshold energy for transcription. We showed that an upstream promoter having a high affinity for ZEBRA can compensate for a promoter having a lower affinity for the general machinery. For example, constructs bearing the high affinity ZIIIB sites cloned upstream of the weak Z core promoter generated 0.090 fmol of product in an in vitro reaction, FIG. 7. Activity of different numbers of sites with different core promoters. Transcriptional activity of templates bearing one, three, or seven high affinity ZIIIB sites cloned upstream of the high affinity H or the low affinity Z core promoter were assayed in vitro in HeLa nuclear extract in the presence of 3-fold increasing concentrations of recombinant ZEBRA ranging from 7.4 to 200 ng. Transcription was assayed by primer extension. A PhosphorImager scan of the primer extension products is shown. Arrows indicate the position of the free primer (Pr) and the bands corresponding to the extension products (EPs). Transcription activities, summarized in the text, averaged from three or four separate experiments, were determined by quantitating the amount of primer extension product using a PhosphorImager and calculating the number of fmol based on a standard curve of the free probe.
whereas low affinity ZRE-14 sites cloned upstream of the high affinity H core promoter produced 0.166 fmol of product, only a 2-fold difference in promoter activities. Similarly, a high affinity core can compensate for a low affinity upstream promoter. For example, the low affinity ZRE-3 sites cloned upstream of the high affinity E4 core promoter produced 0.037 fmol of product, whereas the medium affinity ZRE-16 sites cloned upstream of the lower affinity M core promoter produced 0.032 fmol of product. A high affinity core can also compensate for the number of sites as well as the presence of low affinity upstream sites. For example, a single ZIIIB site upstream of the strong H core promoter produces similar or even slightly higher levels of transcription than seven ZIIIB sites upstream of the weak Z promoter. This is quite unusual and predicts that a strong TATA box can bypass the requirement for synergistic activation by multiple molecules of ZEBRA, a prediction that we confirmed by showing that increased TATA box-saturating concentrations of TFIID dramatically lower the synergistic effect of multiple upstream sites (23).
Consistent with the threshold hypothesis, we have previously shown that strong ZEBRA interactions with the general machinery can compensate for increasing numbers of promoter sites (5). The potency of ZEBRA was reduced by sequential deletion of the activation domain, subdividing the activation domain into four modules, which synergistically contributed to the potency of ZEBRA. The sites and modules were interchangeable such that differences in potencies of the different deletion mutants were greater on templates bearing three sites than on templates bearing five sites. For example, one derivative, Z(77-245), a deletion of the first 77 amino acids of the activation domain, was inactive (less than 10% of wild type ZEBRA) on templates bearing three sites but supported near saturating levels of transcription on templates bearing five sites. Studies on the activity of multimerized VP16 activation domains showed a conceptually similar effect (25, 38 -40).
Core Promoters-The differences between transcriptional activities of the core promoters is somewhat greater than the differences in affinity for the TFIID/TFIIA complex measured in gel shift assays. The E4 core promoter has a 4-fold higher affinity for the DA complex than the M core promoter but generates 6-fold or higher levels of transcription. Even more surprising is the observation that the H core promoter, which has a 2-fold lower affinity for DA than E4, produces as much as 20-fold higher levels of transcription under certain conditions. These data suggest that there are clearly other parameters, which influence the activity of a core promoter in addition to the DA complex affinity. These other parameters may include differential affinities for other components of the general machinery or specific sequence-induced conformational effects on the complex that either stabilize binding or promote open complex formation or elongation.
The findings presented in this paper suggest a mechanism for the precise regulation of different genes in response to a common activator. Our results demonstrate that the activity of a promoter can be varied greatly by altering either the the upstream or core promoter elements. The data suggest mechanisms for regulating both the timing and the levels of activation simultaneously. Promoters with low affinity sites upstream of a low affinity core promoter will be off when a promoter bearing these same low affinity sites upstream of a high affinity core promoter will be on. By varying the number and affinities of activator binding sites as well as the core promoters, a very wide range of ZEBRA responsiveness can be achieved in vivo. It is our goal to be able to apply our rules of upstream and core interaction to the precise gene regulation observed during the EBV early lytic cycle and ultimately to the extremely complex cascades of gene regulation that occur during growth and development.