Sp1 activates and inhibits transcription from separate elements in the proximal promoter of the human adenine nucleotide translocase 2 (ANT2) gene.

Expression of the adenine nucleotide translocator 2 (ANT2) gene is growth regulated. We report a feature of the ANT2 promoter that involves a novel regulatory function for the Sp1 transfactor. We show that expression from the ANT2 proximal promoter is modulated through three Sp1 elements, two of which activate and one of which partially inhibits transcription. The inhibitor site, box C, is juxtaposed to transcription start (nucleotides −7 to −2). Sp1 bound to box C decreases transcription initiation. This was demonstrated by introducing mutations in box C which (a) increased chloramphenicol acetyltransferase expression in the transient transfection assay and (b) inhibited binding of both purified Sp1 and Sp1 in crude nuclear extracts. The activating elements (A and B boxes) are located at adjacent sites in the distal region of the proximal promoter. Mutation of either box inhibits transfection by 90%, indicating that they act in a synergistic manner. Supershift experiments with crude nuclear extracts showed that only Sp1 was bound to the three GC boxes. The finding that Sp1 acts as an activator/inhibitor within the same promoter region was verified in NIH3T3, HeLa, JEG3, and COS-1, indicating that this dual effect of Sp1 is widely preserved. These data suggest a unique role for Sp1 and raise the possibility that growth activation of the ANT2 gene is regulated by the interaction of Sp1 on the A, B, and C boxes.

ANT2 is one of three genes that encodes mammalian adenine nucleotide translocator (ANT, ATP/ADP translocase) proteins (1)(2)(3)(4). The three genes are differentially expressed in a tissuespecific manner (5)(6)(7), and during cellular differentiation (6 -9). The ANT2 isoform is unique since it is expressed in a growthdependent fashion. ANT2 cDNA clones were first isolated as differentially expressed transcripts from growth-activated (10) and serum-stimulated (8) cells. Subsequently, growth-related expression of ANT2 was demonstrated in several cell lines (6,8) and in activated human peripheral lymphocytes (11). Induction of ANT2 transcripts is rapid (8,12); this, together with the conditions under which the original clones were isolated (10), suggests that it might belong to the group of early-immediate genes. This conclusion is supported by the finding that ANT2 transcripts are induced by mitogens such as platelet-derived growth factor and epidermal growth factor and that induction is not inhibited by cycloheximide (8).
To understand the growth-dependent control of ANT2 expression, we characterized the promoter region of the human gene. We report a unique role for the transactivating Sp1 protein. Sp1 is a ubiquitously expressed factor that is required for expression of a large number of constitutive and regulated genes. Sp1 has been described only as a transcription activator. However, Sp1 activation can be suppressed by a variety of mechanisms. For example, Sp3, a member of the Sp1 family (13,14), competes for Sp1 binding sites (15) but is not able to activate transcription (15). Activation by Sp1 can also be suppressed on formation of inactive (non-DNA-binding) complexes with other nuclear factors such as Sp1-I (16) and p107 (17). Thus, interference with Sp1 activation is an established mechanism by which gene transcription can be altered.
Our study raises the possibility of a unique mechanism by which Sp1 modulates gene expression. Here, we provide evidence that Sp1 can, itself, decrease transcription when bound to certain Sp1 elements. We show that the activity of the ANT2 proximal promoter is dependent on three Sp1-binding sites, two of which are adjacent to each other and interact synergistically to activate transcription. The third site lies adjacent to transcription start (nts 1 Ϫ7 to Ϫ2). Sp1 bound to this site lowers transcription initiation efficiency. The possible role of these elements in growth-regulated ANT2 expression is discussed.

EXPERIMENTAL PROCEDURES
Preparation of Clones and Oligonucleotides-The genomic clone of the human ANT2 gene (4) was a gift from Dr. J. Wurzel. A PstI/PstI fragment (nt Ϫ1237/ϩ46, relative to transcription start (4)) from the 5Ј-flanking region of the gene was cloned ahead of the CAT reporter gene in pCATbasic (Promega). Additional clones (nt Ϫ674/ϩ46, Ϫ253/ ϩ46, Ϫ87/ϩ46) were prepared by removing specific restriction fragments from pCAT(Ϫ1237/ϩ46) and religating the plasmid with DNA ligase. Fragments Ϫ64/ϩ46, Ϫ87/ϩ8, and Ϫ87/ϩ8 with mutations in the A, B, and C GC boxes (see Fig. 2) were prepared by polymerse chain reaction amplification. All of the above fragments were cloned into pCATbasic. All clones were verified by DNA sequencing. Oligonucleotides used as electrophoretic mobility shift and supershift assay probes were synthesized (Fig. 2).
Cell Growth and Transfection-Human JEG3, HeLa, mouse NIH3T3 fibroblasts, and monkey COS-1 cell lines were maintained at 37°C in DMEM supplemented with 10% fetal bovine serum (Flow Laboratories), 2 mM glutamine, 50 units of penicillin, and 50 g of streptomycin/ml. The cells were plated in 175-cm 2 plastic flasks and grown to subconfluence. * This study was supported by the Swedish Natural Sciences Research Council. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  Transfection was done by the calcium phosphate precipitate method. 2 ϫ 10 5 cells were grown for 24 h in 60-mm tissue culture dishes. Fresh medium was added 3 h before transfection. Cells were transfected for 24 h with 3 g of reporter plasmid and 0.6 g of pCH110 (Pharmacia Biotech Inc.) plasmid DNA. Medium was changed after the transfection period, and the cells were incubated for an additional 24 h. CAT and ␤-galactosidase activities (pCAT and pCH110, respectively) were measured according to the manufacturer's instructions.
Electrophoretic Mobility Shift and Supershift Assays-Nuclear extracts were prepared as by Dignam et al. (18). Electrophoretic mobility shift And supershift assay binding reactions were performed in 25 l of 15 mM Hepes (pH 7.1), 60 mM KCl, 7.3 mM MgCl 2 , 0.3 mM EDTA, 10% glycerol, 1.2 mM dithiothreitol, 40 M ZnCl 2 , and 20 g/ml calf thymus DNA. Competitor oligonucleotides were added to the binding buffer, followed by 5 g of nuclear extract and 0.1 pmol of labeled probe. The mixture was incubated for 30 min at 25°C and then resolved on prerun 5% native polyacrylamide gels containing 22.5 mM Tris-HCl, 22.5 mM boric acid, 0.5 mM EDTA (pH 8.0), and 5% glycerol. For supershift experiments, 1 g of specific antibody was added to the reaction mixture. The mixture was incubated for 15 min at room temperature followed by 10 min on ice. The antibodies used were Sp1 (Sc-059x), Sp3 (Sc-644x), and AP2 (Sc-184x), all from Santa Cruz Biotechnology, Santa Cruz, CA.

RESULTS
The ANT2 Promoter Is a Composite of Activation and Suppressor Regions-Transfection of deletion constructs of the human ANT2 promoter region (4) into human JEG3 choriocarcinoma cells, HeLa cells, and mouse NIH3T3 fibroblasts (Fig. 1) reveals a complex promoter structure. Removal of nt Ϫ1237 to Ϫ674 (relative to transcription start (4)) decreases CAT expression by 30 -50% in the three cell lines. More strikingly, removal of nt Ϫ674 to Ϫ235 results in reactivation of the CAT expression in the three cell lines. CAT expression supported by clone Ϫ235/ϩ46 is 2-6-fold higher than that supported by the fulllength promoter fragment (Ϫ1237/ϩ46) in the different cell lines tested and 3-13-fold higher than that supported by clone Ϫ674/ϩ46. These data suggest the presence of a negatively regulated element in the Ϫ674/Ϫ235 region which binds a factor(s) common to JEG3, HeLa, and NIH3T3 cells.
Experiments in Fig. 1 indicate that the proximal promoter region, nt Ϫ235/ϩ46, most probably contains the activating region of the promoter. This region, which provides maximum transcription activity, is characterized by the presence of 7 GC boxes ( Fig. 2A). Removal of nt Ϫ235 to Ϫ87, containing 4 GC boxes separated by 20 -40 nt, decreases CAT expression by approximately 50 -60% in JEG3, HeLa, and NIH3T3 cells (Fig.  1). The promoter activity of this shortened clone (clone Ϫ87/ ϩ46) was similar to or greater than that supported by clone Ϫ1237/ϩ46. Further removal of nt Ϫ87 to Ϫ62, which includes two adjacent core Sp1 elements (Fig. 2, boxes A and B), resulted in an 80 -95% decrease in promoter activity (relative to clone Ϫ235/ϩ46) in JEG3, NIH3T3, and HeLa cells. Thus, the activating region of the ANT2 promoter can be characterized as GC rich with six conserved Sp1 core elements between nt Ϫ235 and Ϫ70 (Fig. 2).
Three GC Boxes in the Proximal Promoter Region Define a Unique Activator/Suppressor Region-Of particular interest is the proximal activating region (to nt Ϫ87) which contains, in addition to the two adjacent GC boxes centered at nt Ϫ68 and Ϫ79 (Fig. 2, boxes A and B), a single GC box (box C) juxtaposed to transcription start (see Fig. 2). To test the roles of the A, B, and C boxes, a series of clones were prepared in which mutations were introduced into each of the boxes by polymerase chain reaction amplification. For the transfection experiments show in Fig. 3, the CCGCCC core elements in boxes A and B were replaced by CAGACC (Mut-3), and that in box C was replaced by CCACAC (Mut-1). Two important observations can be made from these experiments. First, mutating the C box results in activation of CAT expression in all cell lines tested (Fig. 3). As shown in a separate series of experiments (Table I), mutations at several sites in box C lead to activation. However, the Mut-2 sequence consistently gives the highest expression of CAT (Table I), i.e. between 5-and 11-fold in the different cell lines tested. These data suggest that the wild-type C box participates in the partial suppression of CAT expression from the ANT2 promoter. Furthermore, since several diverse cell lines respond in the same manner, suppression is probably caused by a factor(s) common to these cell lines. The requirements exhibited by Sp1 for binding to its core element (19) fit with box C mutation data in the present experiments.
A second conclusion drawn from the data in Fig. 3 is that expression of CAT activity is strongly dependent on the A and B boxes. Mutation of only one of the two boxes is sufficient to nearly eliminate promoter activity. Mutation of both boxes had little additional effect. Thus, the A and B boxes appear to cooperate with each other as reported earlier for adjacent GC boxes which support superactivation (20,21). The dependence of CAT expression on boxes A and B was even more pronounced The same three mutations were also placed in box C within the context of the Ϫ87/ϩ8 fragments (see Fig. 3). on transfection of clones carrying C box mutations. In these clones, mutation of either the A or B box decreased CAT expression by greater than 95 percent (Fig. 3). Like the suppressor function of box C, the activator functions of boxes A and B are also preserved in JEG3, NIH3T3, HeLa, and COS-1 cell lines (Fig. 3). Thus, the interaction of the A, B, and C boxes is preserved in divergent cell types, suggesting the involvement of a common, and preserved, mechanism.
Identification of Sp1 as Both an Activator and a Suppressor of the ANT2 Promoter-The above experiments suggest that Sp1 or an Sp1-related protein most probably binds to box C and partially inhibits transcription from the ANT2 proximal promoter. We have obtained four lines of evidence that support this idea. (a) Oligonucleotides carrying box C mutations (Mut-1 (Fig. 4), Mut-2 (Fig. 4, A and B), and Mut-3 (Fig. 4C)) all failed to complete for binding of a low mobility component to oligonucleotide wild-type Ϫ13/ϩ8 containing the wild-type box C. This result was obtained using nuclear extracts from HeLa (Fig. 4A), JEG3 (Fig. 4B), and NIH3T3 (Fig. 4C) cells. Formation of a higher mobility band is inhibited by the same concentrations of wild-type and mutated probes, indicating nonspecific binding. (b) The wild-type box C oligonucleotide binds purified human Sp1, whereas the oligonucleotides bearing a mutated box C do not (Fig. 5). Titration experiments showed that 0.25 ng of the Sp1 protein is sufficient to shift the wild-type probe, but 25 ng was still not sufficient to shift probes bearing mutated C boxes (not shown). (c) Nuclear extracts of HeLa cells (Fig. 5) and NIH 3T3 cells (not shown) shift the wild-type box C oligonucleotide, but not the mutated box C oligonucleotides, indicating that protein is unable to bind to the mutated oligonucleotide. (d) The low mobility band bound to the wild-type box C oligonucleotide is supershifted by antibodies against the Sp1 protein (Fig. 6). Antibodies against the Sp3 protein had no effect. The presence of Sp3 was tested since it binds with the same apparent affinity to the Sp1 element (13,14) but appears to suppress, rather than activate, transcription (15,21) This result was obtained in HeLa, JEG3, and NIH3T3 nuclear extracts (Fig. 6) The above experiments identify Sp1 as the major, and probably the sole, factor in several cell lines that binds to the C box. To test if Sp1 also binds to the upstream A and B boxes, supershift experiments were conducted using an oligonucleotide probe, nt Ϫ87/Ϫ58 (see Fig. 2), containing both boxes. The major, low mobility band obtained with this probe was shifted by antibodies against Sp1 but not Sp3 (Fig. 7). Antibodies against the human AP2 protein were included since AP2 also binds to a GC-rich element (22). AP2 antibodies had no effect. The same results were obtained with nuclear extracts from NIH3T3 and JEG3 cells (not shown). Thus, we conclude that all three GC boxes bind primarily, if not exclusively, Sp1 in several diverse cell lines.
Box C Is Not Juxtaposed to a Putative Ets Site-The sequence TTCGC at positions Ϫ12 to Ϫ8 in Fig. 2 was originally reported as TTCCG (4). TTCCG introduces an opposite-strand Ets binding site (GGAA) adjacent to box C. Ets-related proteins regulate expression from the proximal promoters of other nuclear encoded OXPHOS genes (23)(24)(25)(26). However, when we sequenced the ANT2 genomic clone received as a gift from J. Wurzel (4), TTCGC was found as shown in Fig. 2. This sequence was subsequently confirmed using polymerase chain reaction-amplified genomic DNA from human lymphocytes (not shown). The lack of an Ets binding site is consistent with studies reported above identifying Sp1 as the factor responsible for modulating C box activity. DISCUSSION Sp1 is a ubiquitously expressed nuclear factor that is required for activation of a large number of regulated and con- A, B, and C boxes were prepared by polymerase chain reaction amplification (see "Experimental Procedures"). The sequence of the wildtype and mutated boxes are shown in Fig.  2. XX, mutated GC boxes. Arrows, transcription start. The fragments were inserted ahead of the CAT gene in pCATbasic and then transfected into the indicated cell lines. CAT activity was corrected for transfection efficiency and then normalized to values obtained with the wild-type fragment in each cell line. All experimental points were run in triplicate. The means Ϯ S.E. (bars) are given for two to three individual experiments. stitutively expressed genes. To our knowledge, Sp1 is reported to function only as an activator of transcription. Data reported in this paper on the proximal promoter of the human ANT2 gene raise the possibility that Sp1 may also interfere with transcription. Furthermore, Sp1 appears to have a dual function as an activator/suppressor in this promoter. This unique response to Sp1 is mediated through an octanucleotide (CCGC-CCCG) organized as a distal, direct inverted repeat that is required for transcription activation (Fig. 2, boxes A and B), and as a single proximal element that binds Sp1 and decreases transcription efficiency (Fig. 2, box C). These three elements appear to define the proximal promoter.

TABLE I Effects of C box mutations on transcription initiation from the ANT2 proximal promoter
Transcription is strongly dependent on the inverted repeat A and B boxes. Replacing the CCGCCC core sequence with CA-GACC in either box decreased CAT expression by more than 90%, indicating that the two boxes interact in a synergistic manner. This result is consistent with earlier reports that adjacent GC boxes are used to synergize Sp1-dependent activation (20,21). DNA-bound Sp1 forms higher order homomeric complexes that provide interacting surfaces between complexes on adjacent elements (20). The Sp1 protein regions necessary for superactivation have been mapped, and a requirement for three domains, A, B, and the carboxyl-terminal D domain, has been demonstrated (20,(27)(28)(29). Although we have no direct proof that the same three domains of Sp1 participate in synergistic activation of the ANT2 promoter, we have shown by supershift experiments that Sp1 is the major, and most probably the only, protein bound to the A and B boxes. The presence of another candidate activating protein, Sp4 (13), was not tested. However, unlike Sp1, Sp4 does not synergize transcription from two adjacent Sp1-binding elements (21).
Several lines of evidence support the conclusion that box C is involved in the partial suppression of transcription from the ANT2 promoter. (a) Three mutations introduced into the core Sp1 element (CCGCCC) of box C were all found to enhance transcription severalfold in transient transfection experiments. (b) The same mutations also prevented binding of nuclear extracts and purified Sp1 in mobility shift experiments. (c) no new protein-binding sites were created by the mutations, thus the elimination of binding sites is responsible for activation. Although the involvement of additional, low abundant factors cannot be entirely eliminated, supershift and competition gel shift experiments demonstrate that Sp1 is the major, if not only, nuclear factor that binds to box C. Similar results were FIG. 4. The Sp1 core sequence in box C is required for nuclear factor binding. The gel mobility shift assay was run using the wild-type oligonucleotide (wild-type Ϫ13/ϩ8 (Fig. 2)) as the labeled probe, and nuclear extracts from HeLa (A), JEG3 (B), and NIH3T3 (C) cells. A, competitor oligonucleotide wild-type Ϫ13/ϩ8 was added in 5-, 10-, or 50-fold excess, and the mutated competitor oligonucleotides were added in 50-, 200-, 500-, or 1000-fold molar excess. B, all competitor oligonucleotides were added in 10-, 40-or 250-fold excess. C, all competitor oligonucleotides were added in 12.5-or 250-fold molar excess. The first lane in each series is always without added competitor oligonucleotide. Dot, nonspecifically bound probe. The mutated oligonucleotides (Mut-1, Mut-2, Mut-3) are shown in Fig. 2. The nonspecific oligonucleotide (NS) has the sequence 5Ј-GGAGGCCAAGATGGCGGCAGC-3Ј. obtained with all of the above assays using human, monkey, and mouse cell lines, indicating that the mechanisms of activation and suppression of the ANT2 proximal promoter are widely preserved, an observation consistent with the ubiqui-tous nature of Sp1.
The above experiments raise the possibility that Sp1 can, in special cases, down-regulate transcription via a mechanism that is fundamentally different from Sp1-related suppression mechanisms described previously. Sp3, a member of the Sp1 family (13,14), is a suppressor protein (15,21,30) that competes directly with Sp1 for binding to the Sp1 element (13). However, the A-, B-, and D-transactivating domains of Sp3 cannot replace the counterpart domains in Sp1 and appear to be nonfunctional (15). G10BP, a protein that inhibits Sp1 activation of the fibronectin gene, also acts via a similar mechanism, i.e. competition for Sp1 binding element (31). A second mechanism by which Sp1 function is suppressed has been described for the inhibitor proteins Sp1-I (16) and the riboblastoma-like p107 (17), both of which form complexes with Sp1 and prevent Sp1 binding to DNA. In the above cases, transcription suppression is caused by interference with Sp1 activation. Our data point to a third, unique mechanism in which bound Sp1 itself decreases transcription initiation. The mechanism by which this is achieved is not clear. An attractive explanation is that Sp1 bound at the transcription initiator site decreases the efficiency with which the transcription machinery is recruited and assembled. The TFIID complex protects an extended area on both sides of transcription start (32), and box C (nt Ϫ7 to Ϫ2) most probably lies within the initiator region.
Although our data demonstrate that Sp1 lowers transcription efficiency when bound to box C, it remains to be tested whether binding and release of Sp1 from the C box is a physiological mechanism for modulating ANT2 expression. ANT2 is actively expressed in growth-activated cell lines (6,7,8,9,11,12,33). Furthermore, activation of ANT2 expression is reportedly insensitive to cycloheximide (8), suggesting the involvement of a cell surface receptor-activated signal system. In line with this, phosphorylation of Sp1 has been reported in both human (34) and rat (35) nuclear extracts. Phosphorylation of HeLa cell Sp1 did not, however, alter Sp1-binding affinity or in  6. Supershift analysis identifies Sp1 as the box C binding protein in crude nuclear extracts. Supershift assays were as described under "Experimental Procedures," using oligonucleotide wildtype Ϫ13/ϩ8 as the probe. Nuclear extracts of HeLa, JEG3, and NIH3T3 cells were incubated with Sp1-or Sp3-specific antibodies. Dot, nonspecifically bound probe. FIG. 7. Supershift analysis identifies Sp1 as the major nuclear factor binding to the A and B boxes. An oligonucleotide bearing both the A and B boxes (nt Ϫ87 to nt Ϫ58 (Fig. 2)) was used as the mobility shift probe. The probe was incubated with HeLa cell nuclear extracts (NE) and Sp1-, Sp3-, or AP2-specific antibodies as detailed under "Experimental Procedures." Dot, nonspecifically bound probe.
vitro transcription rates (34). By contrast, phosphorylation of rat Sp1 decreases binding (35). In the context of a Sp1 binding/ release model suggested by our present findings, phosphorylation of Sp1 would be expected to decrease binding to the C box.
The ANT2 promoter is a complex structure with a suppressor region and multiple activating regions. The possibility that these regions can influence the interaction of Sp1 with the A, B, and C boxes remains to be studied. The GC rich Ϫ235/Ϫ88activating region is particularly important in this respect since electron microscopic investigations have revealed interactions between distal and proximal Sp1 multimeric complexes through DNA looping (36). Furthermore, widely separated Sp1 binding elements can also participate in superactivation (20). However, removal of the distal GC elements (Ϫ235/Ϫ88) from the ANT2 promoter decreases CAT expression by only 50%. Thus, although the distal GC elements contribute to transcription efficiency of the ANT2 gene, the data do not support their involvement in superactivation. Even so, future experiments should test whether Sp1 bound to a distal GC element can cooperate with either the A or B elements to induce superactivation, or influence Sp1 binding to the C box.
In summary, we demonstrate that Sp1 binding to three GC elements in the proximal promoter of the human ANT2 gene exerts a unique dual effect on transcription in which two of the elements are required for activation and one element partially suppresses transcription. This study provides the basis for future investigations on the possible involvement of these proximal promoter Sp1 elements in the growth-activated expression of the ANT2 gene.