Identification and characterization of an adenovirus 2 major late promoter CAP sequence DNA-binding protein.

DNase I footprint analysis of the core adenovirus 2 (Ad2) major late promoter (MLP) has revealed distinct patterns of protection corresponding to the assembly of transcription components during transcriptional initiation (VanDyke, M. W., Sawadogo, M., and Roeder, R. G. (1989) Mol. Cell Biol. 7, 3371-3379). By using partially purified transcription factors, DNase I protection over the TATA box element and the CAP sequence was attributed to the binding of a single factor, TFIID. We have determined, however, that protection of the CAP region results from the binding of a novel factor, designated CAP-site binding factor (CBF), which is chromatographically and functionally distinct from TFIID. DNase I footprint analysis and gel electrophoresis mobility shift competition assays confirm that distinct polypeptides bind to the Ad2 MLP upstream promoter sequence, TATA box, and CAP sequences. When the CAP sequence is mutated, transcriptional activity of the Ad2 MLP is reduced both in vitro and in vivo. The decrease in transcriptional activity correlates with decreased CBF binding activity. Nuclear extracts depleted of CBF also exhibit reduced Ad2 MLP transcriptional activity. The addition of DNA affinity purified CBF, free of TFIID or major late transcription factor, restores the activity to control levels.

The addition of DNA affinity purified CBF, free of TFIID or major late transcription factor, restores the activity to control levels.
An understanding of the molecular mechanisms underlying eukaryotic gene expression has in large part relied on the development of soluble in uitro transcription systems using exogenous DNA templates (l-3). These systems have allowed the identification of several general transcription factors, in addition to RNA polymerase II, that associate at the core promoter to allow basal levels of transcription (4-8). Additional trans-acting factors bind to specific upstream cis-acting DNA elements to stimulate basal activity and/or achieve temporal and tissue-specific activity. The characterization and purification of the various components involved in transcription initiation has been difficult, however, due to their relatively low abundance, instability, and the complexity of their interactions with the DNA template and/or other factors.
The adenovirus 2 (Ad2)' major late promoter (MLP) is * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ In partial fulfillment of the Ph.D. requirements at George Washington University.
A region of the Ad2 MLP that has not been as well characterized encompasses sequences surrounding the CAP site at +l, the point at which transcription begins. This region does not display extensive homology with most other promoters. Mutations to sequences at and just downstream of the Ad2 MLP CAP site, however, have been shown to affect the efficiency of transcription.
An A to T transversion at +l decreased the in vitro transcriptional efficiency to 50% of wild type (29). Transient in uiuo expression systems revealed that point mutations of the Ad2 MLP, at +l or -1, decreased promoter activity to 50 or 20% of control levels, respectively (16), whereas deletions of the wild-type promoter from +33 to +7, +5, or -2 resulted in an even greater reduction of promoter activity (17). These studies suggest that the sequences surrounding the Ad2 MLP CAP site may be required for accurate regulation of transcription, and thus, may represent an individual recognition site for a component of the transcriptional machinery.
In fact, several reports have suggested that protein may bind to the DNA sequences at the CAP site of the Ad2 MLP as well as other promoters (30)(31)(32)(33)(34)(35).
In this study, the protein-DNA interactions at the CAP site of the Ad2 MLP were examined. It has been suggested that a DNase I footprint spanning the Ad2 MLP TATA box and CAP sequence resulted from the binding of a single factor, TFIID (14,23,24). We show, however, that the CAP region represents a binding site for a novel factor, designated the CAP-site binding factor (CBF). This protein is shown to be required for optimal activity of the Ad2 MLP. In addition, CBF is chromatographically and functionally distinct from the known regulatory factors, TFIID and MLTF, which bind to the Ad2 MLP. The CAP-specific DNA affinity resin was first equilibrated with buffer D and was then pelleted by centrifugation (3,000 x g) for 2 min. An aliquot of extract (500 ~1) and the CAPspecific DNA affinity resin (250 ~1) were mixed by rotation at 4 "C for 30 min, followed by centrifugation (12,000 x g) for 2 min at 4 'C. transcription initiation site, over the CAP sequence, was also detected. This footprint on the coding strand was characterized by hypersensitive sites at the boundaries of the protected region, +l and +23, and a strong hypersensitive site at position +12. On the non-coding strand, a weak footprint appeared between +2 and +20 and was marked by a weak hypersensitive site at +lO.

Plasmids
To determine if distinct proteins were involved in the protection of these regions, DNA competitor fragments specific for the CAP region or the TATA box element (Fig. 1B) were added to the binding reaction. The TATA-specific DNA fragment abolished the footprint over the TATA box without disturbing the pattern over the CAP sequence (Fig. 2, lanes 5  and 10). Likewise, the CAP-specific competitor fragment competed for the footprint over the CAP sequence without affecting the protection over the TATA box sequence (lanes 4 and 9). Protein binding to the UPS on either the coding or noncoding strand was unaffected by the presence of either the TATA or CAP competitor fragments (lanes 3-5 and 8-10).
These observations were confirmed in mobility shift assays with crude nuclear extract and an Ad2 MLP CAP sequence probe (-14 to +21) (data not shown). Protein-DNA complexes formed that were specifically competed only by the CAP sequence competitor fragment. These data identify a protein in crude nuclear extract that recognizes the CAP sequence of the Ad2 MLP. We designate this protein the CBF. Oligonucleotide competition analysis distinguishes CBF from TFIID, which recognizes the TATA box element, and MLTF, which binds to the UPS.

The Chromatographic
Behavior of CBF Is Distinct from TFIID and MLTF-To compare further the properties of the Ad2 MLP DNA-binding proteins, CBF was partially purified from K562 nuclear extract. Sequential DEAE-anion exchange chromatography, phosphocellulose cation exchange chromatography, and DNA affinity chromatography were employed (Fig. 3A). CBF binding activity was detected by mobility shift assays in the PC 0.3 M KC1 phosphocellulose (PC) fraction (Fig. 3B). Specific CBF.DNA complexes were competed by the unlabeled CAP competitor fragment (compare lane I with lanes 2-4), but not by the same concentration of the TATA fragment (lanes 5-7) or a nonspecific DNA fragment (METI-(lanes 8-10). The PC 0.3 M KC1 fraction also generated a DNase I footprint over the Ad2 MLP CAP region that was identical to the footprint in crude nuclear extract (Fig. 3C, compare lane 3 with 4 and 8 with 9).
The PC 0.3 M KC1 fraction was then subjected to DNA affinity chromatography (see "Materials and Methods"). The 1.0 M KC1 fraction from the DNA affinity column generated a DNase I footprint over the Ad2 MLP CAP sequence (Fig.  3C, lanes 5 and 10) as well as specific protein-DNA complexes in mobility shift assays (data not shown). Competition experiments revealed that protection by DNA affinity purified CBF over the CAP sequence was competed only by the CAP sequence-specific competitor fragment (Fig. 30, lane 4). The same concentrations of competitor fragment specific for the TATA box element or the nonspecific METH sequence were without effect (lanes 5 and 6).
Since the DNA fragment used for the footprint assay encompassed the UPS and the TATA box element, it was possible to follow the independent segregation of MLTF and TFIID from CBF. A footprint over the TATA box element was generated in crude nuclear extract, but not in either the PC 0.3 M KC1 fraction or the 1.0 M DNA affinity fraction (Fig. 3C, lanes 3-5 and 8-10). Similarly, although MLTF copurified with CBF in the PC 0.3 M fraction (Fig. 3C, lanes 4 and 9), the footprint was significantly reduced in the 1.0 M DNA affinity fraction (Fig. 3C, lanes 5 and 10). These data illustrate that the binding activity of DNA affinity purified CBF is chromatographically distinct from that of MLTF or TFIID.

Identification
of CBF as an 85-95-kDa Polypeptide-To determine the size of the polypeptide responsible for CBF binding activity, the protein renaturation assay of Hager and Burgess (50) was performed. DNA affinity purified CBF was electrophoresed by SDS-PAGE along with molecular weight standards. Proteins contained in gel slices were eluted, denatured, renatured, and analyzed for CBF binding by mobility shift assays (Fig. 4). A specific CBF. DNA complex was generated by proteins eluting from the gel slice corresponding to the molecular mass range of 85-95 kDa (Fig. 4, lane 4).
Mutations within the Ad2 MLP CAP Sequence Affect CBF Binding Activity-To identify close contact points between CBF and the Ad2 MLP CAP sequence, a methylation interference assay was performed (Fig. 5A) 1B) were added into the binding reaction.
C, DNase I footprint assays were performed as in Fig. 2 DNA affinity purified CBF (4 pg) was electrophoresed by SDS-PAGE, and the proteins from individual gel slices were eluted, renatured, and assayed for DNA-binding activity as described under "Materials and Methods." Reactions contained the Ad2 MLP CAP sequence probe (-14 to +21). The load contained the 1.0 M KC1 DNA affinity fraction prior to SDS-PAGE (lane I). Molecular weights denoted above the gel refer to the location of the molecular weight protein standards relative to the gel slice.
guanine residues that interfered with protein binding are excluded from forming complexes. Following cleavage of modified residues, those residues critical for binding are underrepresented in the bound DNA as compared with the free DNA molecules. Methylation at the guanine residues at +ll on the coding strand (Fig. 5A, compare lanes 2 and 3) and at +lO and +12 on the non-coding strand (compare lanes 5 and 6) interfered with the binding of CBF to the Ad2 MLP.
Mutations within the CBF-binding domain were generated based on the above results ( Fig. 1C and "Materials and Methods"). The binding of CBF to the wild-type and the A10-12 and i13-16 mutated MLP fragments was first analyzed using DNase I protection assays (Fig. 5B). Digestion patterns showed that the wild-type CAP fragment bound CBF in crude nuclear extract and in the PC 0.3 M KC1 fraction (lanes 2 to 4). A footprint over the CAP sequence was not generated with these proteins on either the A10-12 or the i13-16 mutant fragments (Fig. 5B, lanes 5-10). By using wild-type and mutated CBF recognition sequences in mobility shift assays, CBF. DNA complexes were identified only on the wild-type CAP sequence probe (Fig. 5C, lanes l-3). Specific complexes were not formed on probes for the AlO-12 mutant (lanes 4-6), the d8-12 mutant (lanes 7-9) or the i13-16 mutant (lanes 10 and II). Similarly, competition experiments demonstrated that complexes formed by DNA affinity purified CBF on the wild-type CAP sequence probe, were efficiently competed by a fragment containing the wildtype CAP sequence, but not by as much as a loo-fold molar excess of fragments containing the AlO-12, d8-12, or i13-16 mutations within the CAP sequence (data not shown). tablish a functional role for CBF binding to the CAP sequence element, the transcriptional activity of wild-type and mutated promoters was analyzed in both in vitro and in uiuo transcription assays. The amount of run-off RNA transcribed from isolated Ad2 MLP fragments in uitro was compared in experiments using both HeLa whole cell extract and K562 nuclear extract over a range of template concentrations.
A typical autoradiogram of transcriptional activity is presented in Fig.  6A. The relative transcriptional efficiency at the various template concentrations of each mutation compared to wildtype is presented in Fig. 6B. Although the transcriptional activity of each template displayed a significant amount of variability in different extracts,* the average of three to five separate experiments at each template concentration revealed a consistent trend. The transcriptional activity from both the d8-12 mutant and the i13-16 mutant was 20-30% of wildtype levels. This reduction in activity correlated with the inability of CBF to bind to these mutant CAP sequences (Fig.  5). In contrast, the A10-12 mutant revealed only a minimal reduction in the transcriptional efficiency at any template concentration, with an overall activity of approximately 80% of wild-type levels. This result was inconsistent with the data from mobility shift and DNase I footprint assays, which demonstrated that CBF did not bind to the A10-12 mutation (Fig. 5). To resolve this contradiction, mobility shift assays were performed with wild-type and mutant CAP sequence probes under conditions in which the ionic strength was comparable to that in transcription reactions (Fig. 6C). At increasing KC1 concentrations a specific complex with the A10-12 probe was observed (lanes 5-8), whereas specific binding was not observed with either the d8-12 or i13-16 mutant probes (lanes 9-16). The upper band (Bound) represented the specific CBF.DNA complex as determined by competition experiments (data not shown). Thus, the higher binding affinity of the A10-12 mutation for CBF in comparison to the d8-12 and i13-16 mutations was consistent with its relatively weak effect on in vitro transcription. To assess the promoter strength in Go, wild-type and mutated promoters were fused to the chloramphenicol acetyltransferase gene of E. coli and enzymatic activity determined (Fig. 7). The wild-type Ad2 MLP chloramphenicol acetyltransferase construct was transfected in parallel reactions and for each mutant the conversion of unacetylated [i4C]chloramphenicol to the acetylated form was calculated relative to wild-type levels. A typical autoradiogram is shown in Fig. 7A, and the average of three independent experiments is presented in Fig. 7B. Chloramphenicol acetyltransferase expression was decreased to 52% of wild-type levels for the A10-12 mutant, 36% for d8-12, and to 69% for i13-16. These experiments demonstrate that the wild-type CBF-binding domain is required for optimal Ad2 MLP activity in vitro and in uiuo.
DNA Affinity Purified CBF Rescues TranscriptionalActivity to CBF-depleted Nuclear Extracts-To correlate directly CBF binding and functional activity, DNA affinity purified CBF was assayed in nuclear extracts that were depleted of CBF by the CAP-specific DNA affinity resin (see "Materials and Methods").
Mobility shift assays were used to monitor depletion of CBF (Fig. 8A). CBF.DNA complexes were generated in control extract (Bound, he I), but not with the same probe in depleted extract (lane 2). Protein-DNA complexes were generated in both control and depleted extracts with a UPS-specific probe (Fig. 8A, lanes 3 and 4). These results demonstrate that CBF was depleted by the CAP sequence DNA affinity resin, but other regulatory proteins, for example MLTF, were not.
The transcriptional activity from the wild-type Ad2 MLP template in CBF-depleted extracts showed a 3-fold reduction compared with the levels in control extracts (Fig. 8B, compare  lanes 1 and 2). The addition of DNA affinity purified CBF resulted in rescue of transcriptional activity to control levels (Fig. 8B, lanes 3-6). In contrast, transcriptional activity in an extract depleted by a DNA affinity resin containing nonspecific sequences unrelated to the Ad2 MLP was not rescued by the addition of CBF (Fig. 8B, lanes 7-10).
To eliminate the possibility that DNA affinity purified CBF used in the transcription rescue assay contained other stimulatory factors, such as MLTF, mobility shift assays were performed.
CBF. DNA complexes formed only on a probe specific for the CAP sequence; whereas, complexes did not form on UPS-specific probes (Fig. SC, lanes 1 and 2). Additionally, to test for the presence of TFIID in the DNA affinity purified CBF, a HTNE was prepared in which TFIID was inactivated (24). Specific transcription from the Ad2 MLP was abolished by this treatment (Fig. 80, compare lanes 1  and 2). The activity was efficiently rescued, however, by the addition of the PC 1.0 M KC1 fraction, containing TFIID (24) (Fig. 80, lanes 6-8). In contrast, DNA affinity purified CBF was not able to substitute for TFIID to rescue transcriptional activity to the heated extract (Fig. 80, lanes 3-5). DISCUSSION The transcriptional activity of the adenovirus 2 major late promoter has been shown to be regulated by trans-acting factors present in cellular extracts from uninfected cells. The core promoter of the Ad2 MLP has been described as consisting of the TATA box element and the associated CAP site sequences. These elements interact with components of the basic transcriptional machinery employed by most class II genes to determine the basal levels of promoter activity. A distinct protein, TFIID, binds to the TATA box element (8,(22)(23)(24). Further upstream is the upstream promoter sequence that is required for optimal transcription from the Ad2 MLP. The UPS is bound by the major late transcription factor, also A. B.
-_ -r-r---... A, in uitro transcription assays were performed as described under "Materials and Methods" with HeLa whole cell extract. All DNA templates contained Ad2 MLP sequences from -138 to +193, plus plasmid DNA and were purified by PICS (48). Wild-type and A10-12 mutant templates were 582-bp fragments and produced a 423-base transcript (Ad5 2anes I-II). The d8-12 and the i13-16 mutant templates were both 618-bp fragments, due to different cloning strategies, and produced 450-nucleotide run-off transcripts (Ad2, lanes 12-23). A 217-bp 32P-labeled DNA fragment (Standard) was added into each reaction following transcription to serve as an internal control for recovery of 32-P-labeled RNA product. B, the relative amounts of Ad2 MLP transcription as a function of template concentration. The amount of Ad2 MLP transcription, determined by densitometry, was normalized for variability in recovery using an internal standard. The results are expressed as a percentage of wild-type Ad2 MLP transcription. C, Labeled 39-bp Ad2 MLP probes (-14 to +21) for the wild-type CAP sequence (lanes Z-4) or the A10-12 (lanes 5-8), d8-12 (lanes g-12) or i13-16 (lanes 13-16) mutant CAP sequences were incubated with 6 Kg of the PC 0.3 M KC1 fraction in mobility shift assays. Binding reactions contained 5 pg of nonspecific herring sperm DNA and 20,30,40, or 50 mM KC1 where indicated. Specific CBF. DNA complexes (Bound), nonspecific complexes (AS), and free DNA (Unbound) are indicated. known as USF or UEF (14,(25)(26)(27)(28). These DNA sequences have been extensively characterized, and the factors that recognize them have been purified. Less information is available, however, regarding the role of the DNA at and just downstream of the transcription initiation site. In this report, a novel DNA element, the CAP sequence, has been identified downstream of the transcription initiation site. This element binds a previously unidentified protein, designated the CAPsite-binding factor, that is required for optimal transcriptional activity of the Ad2 MLP.
Mutations at and downstream of the CAP site show that these sequences are required for optimal activity (8, 16-20, 29,51,52). The adenovirus 5 Ela promoter required sequences downstream of the TATA box to +20 (53), whereas sequences to +53 were required for Ad2 early region EIII promoter activity (54). Mutations to DNA sequences between -10 to +7 of the mouse cu-globin gene (55), +5 to +15 of the herpes simplex virus thymidine kinase gene (56, 57), and -5 to +5 of the adenovirus Elb promoter (58), markedly decreased the efficiency of in vitro transcription. Further, mutations to the silk fibroin gene promoter (59) and the long t.erminal repeat of the avian sarcoma virus (60), demonstrated that efficient transcription required sequences with the 3' boundary at +6 for the silk fibroin gene and at +19 for the avian sarcoma virus long terminal repeat. Recently, novel DNA elements at the CAP site of the human gastrin gene (-17 to +57) and the terminal deoxynucleotidyl transferase gene (-6 to +ll) were identified (61, 62). The element within the gastrin gene appears to be a cell-specific regulatory element that may function in conjunction with upstream elements to bring about efficient transcription. The terminal deoxynucleotidyl transferase element (termed the initiator) shows strong sequence similarity to the Ad2 MLP initiation site and directs accurate basal transcription from the terminal deoxynucleotidyl transferase gene.
Although transcriptional efficiency of these promoters appears to depend on the sequences surrounding the initiation site, factor binding to CAP site sequences has been identified for only a limited number of promoters (31)(32)(33)(34). In this report, conventional DNase I footprinting and mobility shift assays have mapped the binding of CBF to DNA sequences immediately downstream of the Ad2 MLP initiation site. Compe- Ad2 MLP inserts (-138 to +33) containing wild-type and mutant CAP sequences were cloned into a vector previously described containing the bacterial chloramphenicol acetyltransferase gene (73). The inserts were cloned directly 5' to the CAT acne. and nlasmids &AT-CAPwt.
DCAT-CAPAlO-12.  were 'isolated in supercoiie'd form by PICS (38). Chloramphenicol acetyltransferase activity was assayed as described under "Materials and Methods." A, a typical autoradiogram demonstrating the conversion of unacetylated [Ylchloram-' phenicol (lower spot) to the two different monoacetylated forms (middle and upper spots) is presented for the wild-type and mutant templates. B, the spots were excised from the plates and counted by liquid scintillation to determine the level of chloramphenicol acetyltransferase expression of each mutant relative to wild-type levels. The average of three independent experiments is presented.
Unbund-CAP tition experiments using crude and partially purified CBF confirmed that the protein was distinct from MLTF, binding to the Ad2 MLP UPS, and from TFIID, binding to the TATA box element. The DNA-binding properties of the partially purified protein showed that only the CAP sequence competitor fragment specifically and efficiently competed for complexes in mobility shift assays or the DNase I footprint over the CAP sequence ( Fig. 3B and D). DNA-binding activities analyzed throughout fractionation demonstrated that MLTF copurified with CBF in the PC 0.3 M KC1 fraction; however, the proteins were separated following DNA affinity chromatography (Fig. 8C). Similarly, the segregation of TFIID and CBF was consistent with a recent report in which TFIID was purified (24). TFIID eluted from phosphocellulose columns in buffers containing 0.85 M KCl, whereas in our chromatography scheme CBF eluted in buffers containing 0.3 M KCl. In addition, a denaturation/renaturation assay determined that polypeptides with a molecular mass of approximately 90 kDa contained specific CBF binding activity. Thus, CBF appeared to be distinct from TFIID and MLTF, which have M, = 120,000 (8) and 43,000 (26,63), respectively.
It has been suggested that a footprint spanning the TATA box and distal CAP region of the Ad2 MLP (-45 to +35) was the result of TFIID binding (24). This hypothesis is open to question, however, for several reasons. Methidiumpropyl-EDTA-Fe(I1) footprinting revealed that specific TFIID interactions were limited to the TATA box element (-32 to -21) (14). Moreover, the downstream extension of the Ad2 MLP footprint (to +35) generated by a partially purified TFIID protein fraction was governed by nonspecific interactions (24). Footprint analysis of this fraction showed only weak to nondetectable interactions downstream of the initiation site of several other promoters (24). In addition, yeast TFIID, which functionally substitutes for mammalian TFIID, generates a DNase I footprint that is limited to the TATA box consensus sequence (64, 65). Interestingly, when TFIID was present in conjunction with the other components of a functional preinitiation complex, an extended footprint was ob- A, standard mobility shift assays were performed using probes specific for the Ad2 MLP CAP sequence (-14 to +21, lanes I and 2) or for the UPS (-91 to -41, lanes 3 and 4), and 10 pg of K562 nuclear extract (ctrl) or 10 mz of depleted K562 nuclear extract (depl). Bound identifies specific CBFIDNA complexes or MLTF.DNA complexes; unbound designates free DNA. B, in vitro transcription reactions were performed as described under "Materials and Methods" with 0.15 pmol of the wild-type Ad2 MLP template. Specific run-off transcripts were 423 nucleotides (Ad2). Reactions contained K562 nuclear extract (lanes 1 and 7), K562 nuclear extract depleted by the CAP sequence DNA affinity resin (lanes 24, or K562 nuclear extract depleted with a DNA affinity resin containing seauences unrelated to the Ad2 MLP (see "Materials and Methods") (&.cs 8-10). Depleted extracts were supplemented with the indicated amount of DNA affinity purified CBF. C, Mobility shift assays were performed as described under "Materials and Methods," using 0.1 pg of DNA affinity purified CBF and a CAP sequence-specific probe (-14 to +21) (lane I) or a UPS-specific probe (-91 to -41) (lane 2). Unbound and bound DNA molecules are shown. D. in vitro transcription was performed with the wild-type Ad2 MLP template as in panel B. Reactions contained K562 nuclear extract (ctrl, lane I) or heattreated nuclear extract (HTNE, lanes 2-8) prepared as described under "Materials and Methods." A labeled 217-bp fragment (standard) was added into each reaction to serve as an internal control for recovery of RNA products. DNA affinity purified CBF, 0.1 rg/rl (lanes 3-5) or the PC 1.0 M KC1 fraction, 0.58 pg/pl (lanes 6-8) were added into the reactions where indicated. Specific Ad2 MLP transcripts are shown (Ad2). served downstream of the initiation site of the MLP (23) and the adenovirus E4 promoter (66). A general model has been proposed, in which TFIID interacts with promoter specific upstream activator proteins, resulting in the sequence independent binding of TFIID to downstream sequences. Although it is likely that these protein-protein interactions occur to alter TFIID binding, it cannot be ruled out that an additional component of the transcription complex may directly interact with downstream promoter regions. In fact, the footprint on the E4 promoter generated by TFIID and the upstream activator protein, ATF, was altered when all the basic transcription components were present. Further, the nonspecific interaction of TFIID to downstream Ad2 MLP sequences may have masked the binding of a specific factor since adenoviral sequences extended only to +lO.
A functional role for CBF was demonstrated by analysis of templates mutated within the CBF-binding domain based on methylation interference assays. In uiuo chloramphenicol acetyltransferase assays showed a decrease in the transcriptional activity of three different mutant promoters. In uitro transcription assays showed that activity was significantly reduced from the d8-12 and i13-16 mutations, and only slightly reduced from the A10-12 mutation. Differences in the transcriptional efficiencies of each of the mutated promoters may be due to the inherent differences in each system. The in vitro data reflect the actual levels of RNA synthesis from isolated DNA fragments, whereas chloramphenicol acetyltransferase assays are a direct measure of stable enzyme activity after the mRNA has been transcribed, processed, transported, and translated.
Alternatively, the relative levels of transcription may reflect different concentrations of specific transcription factors present in the cell types used for each assay. The decreased transcriptional activity of the d8-12 and i13-16 mutations was consistent with the inability of these mutants to bind purified CBF. This establishes the sequence specificity of CBF binding and implies that CBF acts as a positive regulatory factor through its interaction with the CAP sequence element of the Ad2 MLP. Mobility shift assays performed under conditions that closely parallel those of the in vitro transcription system demonstrated a low affinity of CBF for the A10-12 mutant. This may explain the higher levels of activity from this mutant in the in uitro system. A more direct evaluation of CBF activity used extracts depleted of CBF by the CAP sequence DNA affinity resin.
Following depletion, transcriptional activity of the Ad2 MLP was reduced 3-fold. Addition of DNA affinity purified CBF rescued activity of CBF-depleted extracts to control levels.
Extracts depleted with a nonspecific DNA affinity resin or inactivated of TFIID activity showed a decrease in Ad2 MLP transcriptional activity. These results appear to be due to the depletion of rate-limiting factors other than CBF. Hence, the transcriptional activity in these extracts was not rescued by the addition of CBF (Fig. 8, B and D). Therefore, we conclude that DNA affinity purified CBF contains a novel protein with specific binding activity for the Ad2 MLP CAP sequence. This factor is required for optimal Ad2 MLP transcription. In addition, it is chromatographically and functionally distinct from TFIID and MLTF.
Although CBF is a cellular factor present in uninfected cells, whether other promoters utilize CBF is unknown. Most of the promoters that require wild-type sequences in the CAP region do not display any striking sequence similarity with either the Ad2 MLP CAP domain or with each other. This may suggest that a family of functionally related factors exists, with different sequence specificities for DNA binding. Alternatively, CAP-binding proteins may recognize a different feature of this region, such as secondary structure. Interestingly, a comparison of the CAP sequence of the human immunodeficiency virus type 1 (HIV-l) long terminal repeat and the Ad2 MLP illustrates these points. The HIV transactivating responsive region has been shown to contain two direct repeats, CTCTCTGG, that are involved in the binding of a cellular protein (35,(67)(68)(69). An imperfect homology to the HIV trans-activating responsive repeats is found in the Ad2 MLP, with CTCTCTTC between +2 and +9, and CTGTCTGC between +17 and +24 (underlined bases repre-&t matches). Although these direct repeats are part of the CBF-binding domain, our m&hylat.ion interference data dem-onstrated that these sequences were not critical contact points for CBF binding. Therefore, their role, if any, in Ad2 MLP function remains unknown at present. The activity of the HIV long terminal repeat has also been shown to be influenced by the maintenance of a stable stem-loop structure in the region between +l to +44 (69). The sequences required to maintain this secondary structure do not specify individual protein-binding domains. Both the Ad2 MLP (70, 71) as well as the human gastrin gene promoter (61) also contain potential stem-loop structures in their CAP regions; however, whether secondary structure at CAP sequences is a general phenomenon affecting promoter activity is unknown. Finally, since CAP sequences are contained in the primary transcripts, consideration must be given to the potential involvement of these sequences in transcript stability or attenuation in addition to their effects on promoter activity.