Paclitaxel activation of the GADD153 promoter through a cellular injury response element containing an essential Sp1 binding site.

The GADD153 promoter is transcriptionally activated by paclitaxel-induced injury. Promoter deletion from −786 to −85 base pairs relative to the start of transcription had no significant effect on activation, but deletion to the TATA box abolished it. Placement of the 39 bases from −74 to the TATA box (cellular injury response element, CIRE) upstream of the adenovirus E4 TATA box conferred paclitaxel inducibility. The only consensus sequence present in the CIRE is an Sp1 site; mutation of this site inhibited paclitaxel activation. Paclitaxel failed to activate a SV40-driven luciferase construct containing five Sp1 sequences, and Sp1 sites further upstream in the GADD153 promoter were not essential for activation. Pure Sp1 and nuclear extracts from uninjured and paclitaxel-injured cells protected the same region from −62 to −48 bases on the noncoding strand and −74 to −53 on the coding strand. Nuclear extracts shifted the CIRE to the same extent as purified Sp1 but had no effect on a CIRE with a mutated Sp1 site in gel shift assays. Immunodepletion of Sp1 abolished the shift; antibody to Sp1 produced a supershift. These data indicate that paclitaxel activates the GADD153 promoter through a constitutively occupied Sp1 site at −61 bases.

The GADD153 promoter is transcriptionally activated by paclitaxel-induced injury. Promoter deletion from ؊786 to ؊85 base pairs relative to the start of transcription had no significant effect on activation, but deletion to the TATA box abolished it. Placement of the 39 bases from ؊74 to the TATA box (cellular injury response element, CIRE) upstream of the adenovirus E4 TATA box conferred paclitaxel inducibility. The only consensus sequence present in the CIRE is an Sp1 site; mutation of this site inhibited paclitaxel activation. Paclitaxel failed to activate a SV40-driven luciferase construct containing five Sp1 sequences, and Sp1 sites further upstream in the GADD153 promoter were not essential for activation. Pure Sp1 and nuclear extracts from uninjured and paclitaxel-injured cells protected the same region from ؊62 to ؊48 bases on the noncoding strand and ؊74 to ؊53 on the coding strand. Nuclear extracts shifted the CIRE to the same extent as purified Sp1 but had no effect on a CIRE with a mutated Sp1 site in gel shift assays. Immunodepletion of Sp1 abolished the shift; antibody to Sp1 produced a supershift. These data indicate that paclitaxel activates the GADD153 promoter through a constitutively occupied Sp1 site at ؊61 bases.
Treatment of mammalian cells with genotoxic agents causes an increase in the mRNA levels of a number of "damage response" genes (reviewed by Holbrook and Fornace (1)). Many of these genes are also inducible by phorbol ester treatment. Among those that do not respond to phorbol ester treatment, some can be activated by the tumor suppressor gene p53, such as WAF1 and GADD45 (2,3). However, there are also a number of DNA damage-inducible genes for which the activation signal is unknown, and GADD153 is one of these genes. GADD153 is of particular interest because the magnitude of the increase in its mRNA following cellular injury is greater than most other "damage response" genes.
GADD153 was originally cloned by subtractive hybridization of UV-treated versus proliferating Chinese hamster ovary cells. It was one of a subset of genes that was induced by UVradiation and other forms of DNA damage but not by heatshock or phorbol ester treatment (4). This subset of genes was found to be coordinately regulated by a number of agents that damage DNA or induce cell cycle arrest (5). The human homolog has now been cloned and localized to the 12q13.1-q13.2 region on chromosome 12 (6). This gene is highly conserved, with the human gene showing 78% nucleotide identity with the hamster gene.
The role of the GADD153 gene product has not been well defined, but there is some evidence that it plays a role in cell cycle control during the cellular injury response. When transfected into cells under the control of a constitutively active promoter, GADD153 expression blocks the ability of human ovarian carcinoma cells to proliferate and form colonies. 1 Zhan et al. (7) reported that co-transfection of a GADD153-expressing vector with a vector containing a selectable neo gene decreased the plating efficiency of H1299 and HeLa cells to 30% of that obtained with the neomycin plasmid alone. Barone et al. (8) microinjected an expression vector containing the GADD153 coding region or purified GADD153 protein into cells. Both the plasmid and the purified protein prevented the cells from exiting G 1 and entering the next S phase, as measured by bromodeoxyuridine incorporation. These results argue that the GADD153 gene product may be involved in cellular machinery that causes arrest at the G 1 -S boundary after DNA damage or other kinds of cellular injury.
We have previously demonstrated in cell lines and xenografts that the magnitude of the increase in GADD153 message is closely linked to the extent of cellular injury caused by the chemotherapeutic drugs cisplatin and paclitaxel, and that these two agents transcriptionally activate the GADD153 promoter via different signal transduction pathways (9). As an initial step toward defining these pathways in more detail, we have dissected the GADD153 promoter to identify the components essential to the transcriptional activation produced by paclitaxel. We report here the identification and initial characterization of a cellular injury response element (CIRE) 2 that is required for the paclitaxel activation of the GADD153 promoter.

MATERIALS AND METHODS
Chemicals-Paclitaxel was obtained from Calbiochem. Luciferin was obtained from Analytical Luminescence (San Diego, CA). DNase I was obtained from Sigma.
Cell Culture-The human ovarian carcinoma cell line 2008 (10) was carried as an exponentially growing monolayer in a humidified incubator at 37°C and 5% CO 2 in RPMI 1640 supplemented with 5% fetal calf serum and 2 mM glutamine.
Vector Construction-pGADD-LUC, a GADD153 promoter driven luciferase reporter construct, was created by ligating the ClaI/HindIII fragment of p9000 (a gift of Dr. N. J. Holbrook, National Institute on Aging, National Institutes of Health, Baltimore, MD) containing the hamster GADD153 promoter into the AccI/HindIII site of pB-LUC (11). pB-LUC contains the firefly luciferase gene ligated into the BamHI site of pBluescript KS Ϫ (a gift of Dr. Linda Quattrochi). Deletion mutations were made using a PCR-based strategy. Primers containing XhoI linkers and an antisense primer within the multicloning region of pBluescript were used to generate deletion fragments of the promoter. These fragments were cut with XhoI and HindIII and ligated into the XhoI/ HindIII sites of pB-LUC. Mutations in the Sp1 binding site were made by substituting TCC TAG ACC for TCC CGC CCC (underlined text, Sp1 binding site; boldface text, mutated bases) at position Ϫ59 relative to the start of transcription. Linker scanning mutations were made using primers containing the Sp1 mutation (as above) or a 3Ј mutation (substituting CCT CTC TAG for CCA AAA GAG (boldface text, mutated bases) at position Ϫ51 relative to the start of transcription) in both the sense and antisense orientation. Fragments were made from the mutation to the pBluescript cloning region and from the mutation to Ϫ185 (with XhoI linker) relative to the start of transcription. These fragments were mixed and used as template for a reaction with primers from Ϫ185 (with XhoI linker) to the pBluescript cloning region. These fragments were cut with XhoI and HindIII and ligated into the XhoI/HindIII sites of pB-LUC.
Luciferase Assay-The cells were transfected with the pGADD-LUC construct by a modification of the method described by Rose et al. (12). Cells were plated at 3 ϫ 10 5 cells per 35-mm dish, and then 18 h later, they were incubated at 37°C with 5 g plasmid DNA and 30 l liposomes in 1 ml RPMI 1640. After 3 h, the lipids were removed, and the cells were treated with paclitaxel for 24 h. The cells were lysed in 100 -500 l of lysis buffer (25 mM glycylglycine, pH 7.8, 15 mM MgSO 4 , 4 mM EGTA, 1% Triton X-100, and 1 mM dithiothreitol). Luciferase activity was measured by a modification of the method described by Brasier et al. (13). Fifty l of cell lysate were added to 200 l of reaction buffer (lysis buffer with 15 mM potassium phosphate, pH 7.8, and 2 mM ATP added). Light emission was measured after injection of 100 l of 1 mM luciferin into the lysate/reaction mixture using a MonoLight 2001 (Analytical Luminescence, San Diego, CA).
Nuclear Protein Extraction-Nuclear extracts were made by pelleting 10 8 cells and resuspending in buffer A (10 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, and 1 mM dithiothreitol). Cells were allowed to equilibrate for 10 min on ice and were collected by centrifugation for 5 min at 500 ϫ g. Cells were lysed in buffer B (buffer A with 0.2% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin) for 10 min on ice. Nuclei were collected by centrifugation for 10 min at 1000 ϫ g. Nuclei were suspended in buffer C (20 mM HEPES, pH 7.9, 20% glycerol, 100 mM KCl, 0.2 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin) and then lysed by bringing the final KCl concentration to 0.4 M, with rocking for 30 min at 4°C. Genomic DNA was pelleted by centrifugation for 40 min at 65,000 rpm (175,000 ϫ g) in a Beckman tabletop ultracentrifuge. The supernatant containing the nuclear proteins was recovered, and glycerol was added to a final concentration of 40%. Sp1 was immunodepleted from extracts by incubation for 1 h at 4°C with a rabbit polyclonal antibody raised against amino acids 520 -538 of the human Sp1 protein (PEP2; Santa Cruz Biotechnology, Inc., Santa Cruz, CA). The lysate was cleared using protein A-Sepharose (Sigma).
DNase I Footprinting-A fragment of DNA corresponding to bases Ϫ185 to ϩ21 relative to the start of transcription was generated by PCR. The fragment was cut with either XhoI or EcoRV to remove one of the primers, and the DNA was 32 P-labeled on a single strand by T4 polynucleotide kinase (Promega, Madison, WI), precipitated to remove unincorporated nucleotide, and resuspended. Ten g of nuclear extract, two footprinting units of purified Sp1 (Promega), or no protein controls were diluted to 25 l with buffer Z (25 mM Hepes, pH 7.5, 100 mM KCl, 12.5 mM MgCl 2 , 10 M ZnSO 4 , 20% glycerol, and 0.1% Nonidet P-40). Twenty-five l of a solution containing 10 l of 10% polyvinyl alcohol (Sigma), 1 mg poly(dI-dC), and Ն10,000 cpm (10 -25 fmol) of labeled DNA was added to the protein mixture and incubated 15 min at room temperature. Fifty l of 5 mM CaCl 2 /10 MgCl 2 mix was added and incubated for 1 min. Two l of DNase I (Sigma; 10 mg/ml; dilutions ranging from 1:2,000 to 1:100,000) was then added and incubated for another minute. The reaction was stopped by adding 90 l of stop solution (20 mM EDTA, pH 8.0, 1% SDS, 0.2 M NaCl, and 250 g/ml glycogen). Proteins were digested with 5 l of 10 mg/ml proteinase K and removed by phenol extraction. The DNA was precipitated and resuspended in 3 l of formamide loading buffer. The bands were separated on a 6% polyacrylamide/urea sequencing gel, dried, and exposed to Biomax MS film. Sequencing was done with the Promega fmol sequencing kit (Promega) using the protocol for end-labeled primers.
Gel Mobility Shift Assays-Oligonucleotides corresponding to the CIRE were constructed at the University of California, San Diego Molecular Biology Core facility, and annealed together by heating for 5 min at 65°C and incubating overnight at 37°C in 10 mM Tris, pH 7.8, 0.1 M NaCl, and 1 mM EDTA. The double-stranded CIRE (or Sp1mutated CIRE) was precipitated with ethanol, resuspended in H 2 O, and 32 P end-labeled with T4 polynucleotide kinase (Promega), precipitated to remove unincorporated nucleotide, and resuspended to 500 pg/l (ϳ50,000 cpm/l).
Supershift assays were performed as above with the addition of 0.05 g of anti-Sp1 antibody (PEP2; Santa Cruz Biotechnology, Inc.) or nonspecific rabbit antibody (anti-tubulin; Sigma) to the reaction mixture after the 30-min incubation and incubating for an additional 30 min.
Northern Blotting-Total cellular RNA was extracted, and Northern blots prepared using MagnaGraph nylon membranes (MSI, Westboro, MA) by standard techniques (14). The extent of hybridization was quantitated by the Molecular Imager System (Bio-Rad). The human GADD153 probe was a gift of Dr. N. J. Holbrook (National Institute on Aging, National Institutes of Health, Baltimore, MD). Lane loading differences were corrected for by comparison to the same blot hybridized with a ␤-actin probe.

RESULTS
Effect of Paclitaxel on the GADD153 Promoter and Endogenous GADD153 mRNA-The effect of paclitaxel on the GADD153 promoter-driven luciferase construct and endogenous GADD153 mRNA was studied in the 2008 ovarian carcinoma cell line. Cells were transfected with 5 g of pGADD-LUC and then exposed to paclitaxel for 24 h, at which point luciferase activity and endogenous GADD153 mRNA levels were determined. As shown in Fig. 1, paclitaxel induced a concentrationdependent increase in both GADD153 promoter activity and endogenous GADD153 message level that were of the same order of magnitude.
Deletions in the GADD153 promoter were created by PCR and are shown in Fig. 2. These constructs were transfected into 2008 cells and tested for inducibility following exposure to 70 nM paclitaxel relative to cells treated with vehicle alone. Deletions from Ϫ786 to Ϫ85 relative to the start of transcription had no significant effect on the activation of the GADD153 promoter by paclitaxel. Deletion to Ϫ74 base pairs increased the inducibility almost 2-fold, suggesting that the 11 bases 5Ј of Ϫ74 contain a transcriptional repressor. Further deletion of the promoter to Ϫ35 bases (leaving only the TATA box) decreased the paclitaxel inducibility to 2.2-fold. This suggests that the 39 base pairs from Ϫ74 to the TATA box contain an element that is responsible for the paclitaxel-induced increase in GADD153 promoter activity. The finding that the construct containing only the TATA box is activated 2-fold suggests that paclitaxelinduced injury has a slight positive effect on the basal transcription machinery.
Transfer of the Paclitaxel-responsive Element to a Heterologous Promoter-To determine whether this 39-base pair sequence from Ϫ74 bases to the TATA box was sufficient to confer paclitaxel responsiveness to another promoter, we transferred these bases into a promoter containing the adenovirus E4-TATA region. The relative position of these 39 bases and the TATA box was kept constant. As shown in Fig. 3, the 39 bases conferred paclitaxel inducibility on the E4-TATA region, and the magnitude of induction was similar to that produced by the Ϫ74 GADD-LUC construct. We have identified this 39-base pair region as a CIRE.
Mutation of the Sp1 Site within the CIRE-The CIRE region from Ϫ74 to Ϫ35 contains only one known transcription factor recognition sequence, an Sp1 binding site. We mutated this site from CCCGCCCC to CCTAGACC using a PCR-based system and assayed the activity of this construct. As shown at the bottom of Fig. 2, this mutation decreased the ability of paclitaxel to activate the promoter from 20.0-to 5.3-fold, indicating that the Sp1 site is a necessary part of the sequence that confers paclitaxel sensitivity.
Effect of Paclitaxel on Other Sp1-containing Promoters-Since the Sp1 site appears to be required for the maximal paclitaxel-induced increase in promoter activity, we wished to determine whether paclitaxel activates all promoters that contained Sp1 sites. We explored this using the SV40 immediateearly promoter that contains five functional Sp1 sites and the cytochrome p450 CYP1A1 promoter, which also contains five Sp1 sites (15). As shown in Fig. 4, paclitaxel increased the promoter activity in these constructs only 2-fold, which is the same increase as was observed in promoters containing only a TATA box. This indicates that although an Sp1 site is required for paclitaxel activation, it is not sufficient, or that the exact positioning of the Sp1 site relative to the TATA box is crucial.
DNase I Footprinting Analysis of the GADD153 Promoter-DNase I protection assays were performed to determine which bases of the CIRE are involved in the response to paclitaxel. The GADD153 promoter region from Ϫ185 to ϩ21 bases relative to the start of transcription was 32 P end-labeled and incubated with buffer alone or with two footprinting units of purified Sp1 protein, 10 g of nuclear extract from untreated cells, or 10 g of nuclear extract from cells treated with 70 nM paclitaxel for 24 h. As shown in Fig. 5, the pure Sp1 and nuclear extracts from both untreated and treated cells protected the same region from Ϫ62 to Ϫ48 bases on the noncoding strand. The nuclear extracts protected from Ϫ74 to Ϫ53 on the coding strand; purified Sp1 had a similar pattern of protection, protecting bases Ϫ71 to Ϫ53. The region protected on both strands includes the Sp1 site, which starts at Ϫ61 bases. These data suggest that Sp1, a protein complex that contains Sp1, or a protein that has the same binding characteristics as Sp1 is constitutively bound to the CIRE.
Linker Scanning Mutation of Sites Protected in Footprinting Assays-To determine whether the sites protected in the DNase protection assay are required for paclitaxel activation of the GADD153 promoter or simply protected in a nonspecific manner, we mutated the Sp1 site and the protected bases directly 3Ј of the Sp1 site using a linker scanning method. The sequence of the CIRE and the mutations constructed are shown in the top panel of Fig. 6. As shown in the bottom panel of Fig.  6, mutations in the Sp1 site decreased the inducibility of the construct to that of Ϫ35 pGADD-LUC, which contains only the TATA box. Mutation of the site directly 3Ј of the Sp1 site had no significant effect on the activation of the promoter by paclitaxel. Although the magnitude of the induction of the Ϫ185 Sp1 mutant is somewhat less than that observed for the Ϫ74 Sp1  (Fig. 2), it should be noted that the Ϫ185 Sp1 mutant contains the putative negative regulatory element between Ϫ85 and Ϫ74. The presence of this element and possible differences in tertiary structure of the constructs may account for the observed difference in activation after paclitaxel treatment. These data indicate that the Sp1 site, but not the protected site directly 3Ј of the Sp1 site, is required for activation of the GADD153 promoter by paclitaxel.
Gel Mobility Shift Assays-To further investigate the inter-action between the CIRE and nuclear proteins from untreated and paclitaxel-treated cells, we performed gel mobility shift assays. Nuclear extracts were incubated with 32 P-labeled CIRE and separated on a 0.5 ϫ TBE/4% polyacrylamide gel. As shown in Fig. 7, 5 and 10 g of nuclear extracts from untreated or paclitaxel-treated cells shifted the CIRE to the same extent as purified Sp1 protein (labeled I, lanes 2-6). However, no shift was observed when the proteins were incubated with a CIRE in which the Sp1 site had been mutated (Fig. 7, lanes 7-11). A nonspecific band of lower apparent molecular weight was observed when nuclear extracts were incubated with both the CIRE and the Sp1-mutant CIRE (labeled A in Figs. 7 and 8). This result indicates that the binding of the protein responsible for the gel shift of the CIRE is dependent upon the presence of functional Sp1 site within the CIRE. Since the shift in apparent molecular mass was dependent upon a functional Sp1 site, we sought to determine whether Sp1 was bound to the CIRE after incubation with nuclear extract. As shown in Fig. 8, co-incubation of the nuclear extract-exposed CIRE with a rabbit polyclonal antibody raised against amino acids 520 -538 of human Sp1 protein (PEP2) caused a supershift and increased the apparent molecular weight of the shifted complex (labeled II, lanes 4 and 6), whereas rabbit anti-␤-tubulin antibody (Sigma) had no effect on the mobility of the complex (labeled I, lanes 3 and 5). This indicates that Sp1 protein is part of the complex that shifts the CIRE. To determine whether Sp1 protein is required for the formation of the complex that shifts the CIRE, Sp1 was immunodepleted from the nuclear extract with the PEP2 anti-Sp1 antibody. The mock-depleted extract shifted the CIRE to the same extent as the untreated nuclear extract (Fig. 8, lane 7); however, immunodepletion of Sp1 abolished the shift (Fig. 8,  lane 8). Thus, Sp1 protein is required for formation of the complex that shifts the CIRE. However, Western blot analysis of 2008 whole-cell lysates 24 h after treatment with 70 nM paclitaxel demonstrated no change in Sp1 levels or mobility, The GADD153 promoter region from Ϫ185 to ϩ21 bases relative to the start of transcription was labeled with 32 P on either the coding or noncoding strand and incubated with buffer alone (No Protein), two footprinting units of purified Sp1 protein (Pure Sp1), 10 g of nuclear extract from untreated cells (Untreated), or 10 g of nuclear extract from cells treated with 70 nM paclitaxel for 24 h (Paclitaxel). The complex was incubated with DNase I for 1 min, and the DNA was phenol/chloroform extracted and separated on a 6% polyacrylamide/urea gel. Position was determined by dideoxy sequencing. The Sp1 consensus site is underlined. suggesting that activation of the promoter is mediated by an event other than a change in Sp1 protein levels. 3

DISCUSSION
The data presented in this report demonstrate that treatment of human ovarian carcinoma cells with the chemotherapeutic drug paclitaxel activates the GADD153 promoter through an element no larger than 39 bases positioned immediately 5Ј of the TATA box that we have identified as a CIRE. Three lines of evidence establish that the CIRE is necessary and sufficient for transcription activation of this promoter: 1) deletion of the CIRE eliminates the ability of paclitaxel to activate the GADD153 promoter above the level observed with a promoter containing only the TATA box; 2) mutation of the key Sp1 site within this element inhibits the ability of paclitaxel to activate the promoter; and 3) transfer of this element to a position 5Ј of a heterologous promoter confers paclitaxel inducibility.
Work by Sylvester et al. (16) suggested that transcriptional activation of the GADD153 promoter after treatment with lipopolysaccharide required the binding of transcription factors to the C/EBP element present in the promoter at Ϫ332 to Ϫ323 bases relative to the start of transcription (16). However, this element does not appear to be involved in the cellular response to paclitaxel. These data are consistent with our previous work and reports by others, showing that induction of GADD153 can be mediated by several pathways that differ depending on the type of growth-inhibitory stimulus (9,17).
The CIRE that mediates transcriptional activation by paclitaxel contains an Sp1 consensus sequence (CCCGCC) at Ϫ61 bases relative to the start of transcription. As shown most cogently by the linker scanning mutations, this site is required for activation of the promoter by paclitaxel, whereas the three other Sp1 sites further 5Ј in this promoter apparently are not required and cannot substitute for the Sp1 site in the CIRE. The specificity of the Sp1 site in the CIRE is further demonstrated by the fact that paclitaxel does not activate the SV40 early promoter, demonstrating that Sp1 sites alone are not sufficient for full activation.
The results reported here provide strong evidence that Sp1 or a very closely related protein is required for the activation of the GADD153 promoter by paclitaxel. In DNase I protection assays, the footprint of purified Sp1 was similar to that of the nuclear extract in that region of the promoter. A protein or proteins present in the nuclear extract bound to the CIRE and shifted the DNA to the same extent as purified Sp1 protein but did not shift the CIRE containing a mutation in the Sp1 site. An antibody specific for Sp1 supershifted the nuclear extract⅐CIRE complex, indicating that Sp1 is part of the complex. Sp1 is required for this binding since immunodepletion of Sp1 from the nuclear extract abolished the gel shift. These experiments very strongly suggest that Sp1 is the protein that binds to the CIRE. There is a small possibility that a protein which cross-reacts with the Sp1 antibody is the protein that binds to the CIRE. Recent evidence suggests that there is a family of proteins that can bind to the Sp1 consensus site (18 -21); however, the fact that the PEP2 anti-Sp1 antibody is highly specific and does not cross-react with other family members makes it unlikely that another known member of the family is involved.
Although Sp1 was originally thought to be a ubiquitous transcription factor involved solely in the constitutive expression of genes (22), recent evidence suggests that Sp1 is critical for regulated expression in some situations. Sp1 can act in concert with cell type-specific proteins to achieve tissue-specific expression (23)(24)(25). In addition, linker scanning mutations of the myeloid integrin CD11b promoter demonstrated that mutation of the Sp1 site abolished myeloid-specific promoter activity (26,27). Thus, there are now several examples of transcriptional activation of promoters where Sp1 is an essential component but, nevertheless, is involved in regulated rather than constitutive expression.
Despite the evidence that the Sp1 binding site in the CIRE is were end-labeled with 32 P and incubated with no protein (lanes 1 and 12), purified Sp1 (lanes 2 and 11), 10 g of nuclear extract from untreated cells (lanes 3 and 10), 10 g of nuclear extract from paclitaxeltreated cells (lanes 4 and 9), 5 g of nuclear extract from untreated cells (lanes 5 and 8), and 10 g of nuclear extract from paclitaxel-treated cells (lanes 6 and 7) before separation on a 4% polyacrylamide/0.5% TBE gel.
FIG. 8. Sp1 supershift and gel mobility shift assays of the CIRE using nuclear extracts immunodepleted of Sp1. The CIRE was end-labeled with 32 P as above and incubated with no protein (lanes 1 and 2), nuclear extract from untreated cells (lanes 3 and 4), or nuclear extract from paclitaxel-treated cells (lanes 5 and 6), and with rabbit ␣-tubulin (lanes 1, 3, and 5) or rabbit ␣-Sp1 antibody (lanes 2, 4, and 6). The CIRE was also incubated with mock-depleted nuclear extract (lane 7) and Sp1-depleted nuclear extract (lane 8). essential for paclitaxel-induced activation of the GADD153 promoter, we found that paclitaxel did not produce a change in apparent Sp1 binding to this consensus sequence, nor in Sp1 protein levels or mobility as measured by Western blot analysis. This suggests that changes in Sp1 protein levels are not responsible for GADD153 promoter activation after paclitaxel treatment. Sp1 was present in complexes formed with the CIRE by nuclear extracts from both untreated and paclitaxelexposed cells, and a footprint consistent with Sp1 was produced in in vitro DNase I protection by nuclear extracts from both types of cells. Other Sp1-dependent inducible systems show this type of constitutive binding. The retinoblastoma protein regulates the expression of a number of genes including c-fos, c-jun, c-myc, insulin-like growth factor II, and transforming growth factor ␤1 through a retinoblastoma control element (28,29). Sp1 protein binds to the retinoblastoma control element constitutively in the insulin-like growth factor II and the c-jun promoters (28,29). The amount of this binding is increased in the c-jun promoter in response to retinoblastoma expression, possibly through the interaction of RB protein and an inhibitor of Sp1 binding (Sp1-I) (29), but no change in binding of Sp1 as measured by gel shift or DNase I footprinting was observed in the insulin-like growth factor II promoter following retinoblastoma expression (28).
Apparently, in the case of the GADD153 promoter, formation of the complex for which the Sp1 site is essential and which contains Sp1 is not by itself sufficient for activation of the promoter, suggesting that activation requires additional events, such as phosphorylation of either Sp1 itself or one of the TAFs with which it associates. Recent data suggest that paclitaxel can activate a number of protein kinases (30 -33). Activation of one or many of these kinases that could be responsible for the activation of transcription from Sp1 constitutively bound to DNA. There is evidence that phosphorylation of Sp1 can, in fact, alter its transcriptional activity. Okadaic acid stimulates the human immunodeficiency virus long terminal repeat through Sp1 sites, yet the binding characteristics of Sp1 to the promoter are unchanged (34). Using Western blotting, it was observed that okadaic acid treatment resulted in the complete conversion of Sp1 from the hypophosphorylated state (95 kDa) to the hyperphosphorylated state (105 kDa). To determine whether Sp1 phosphorylation induced changes in the interaction with TBP or TBP-associated proteins (TAFs), the TATA box was exchanged with a non-TBP binding TATA box. This change strongly decreased the inducibility by okadaic acid. Vlach et al. postulated that phosphorylation of Sp1 increases interaction with TBP or TAFs, which results in increased transcriptional activity.
We postulate that Sp1 is constitutively bound to the Sp1 site at Ϫ61 bases relative to the start of transcription, and that after paclitaxel treatment, a posttranslation modification occurs that allows the Sp1 to activate transcription. Since we found that 24-h treatment with 70 nM paclitaxel had no significant effect upon Sp1 protein levels nor on the phosphorylationdependent electrophoretic mobility of Sp1, it is likely that if phosphorylation is involved, it is either one of the TAFs that is the essential target or that activation is the result of a change in the distribution of phosphorylated sites within Sp1 that does not change the electrophoretic mobility of Sp1. This change in phosphorylation would allow a TAF that interacts with Sp1, such as TAF II 110 (35), to bind to both Sp1 and the TATAbinding protein and initiate transcription. An alternative explanation is that paclitaxel treatment has no effect on Sp1 protein but modifies TAF II 110 or another of the TAFs in such a way that they can bind to Sp1 and TBP and increase transcription. As with other observations suggesting that Sp1 can act as a transcription factor for inducible genes, we do not know why transcriptional activation can occur from one Sp1 site but not from other Sp1 sites in the same promoter. Our results argue that the Sp1 site at Ϫ61 in the GADD153 promoter has some special features the elucidation of which will help solve this puzzle.