Transforming growth factor beta activates the promoter of cyclin-dependent kinase inhibitor p15INK4B through an Sp1 consensus site.

Transforming growth factor β (TGF-β) causes growth arrest in the G1 phase in many cell types. One probable pathway for this growth inhibition is through the TGF-β-mediated up-regulation of the cyclin-dependent kinase (CDK) inhibitor p15INK4B, which specifically inhibits the enzymatic activities of CDK4 and CDK6. An active cyclin D-CDK4/6 complex is required for pRb phosphorylation to allow the cell cycle to progress from G1 to S phase. To study the molecular mechanism of the p15INK4B induction by TGF-β, we isolated a 780-base pair promoter sequence of the human p15 gene and inserted this fragment upstream of a luciferase reporter gene. When this construct was transiently transfected into HaCaT cells, luciferase activity was induced more than 10-fold upon TGF-β treatment, indicating that the induction of p15INK4B expression by TGF-β is partly exerted at the transcription level. Promoter deletion analysis revealed that the sequence from −110 to −40 relative to the transcription start site is capable of conferring the 10-fold induction by TGF-β. Within this region there are three Sp1 consensus sites. Mutation of one of these sites, GGGGCGGAG, substantially reduced both the induction by TGF-β and the basal promoter activity, whereas mutations in the other two Sp1 sites and the spacer sequences had little effect. In addition, gel mobility shift assay indicates that the transcription factors Sp1 and Sp3 bind to this Sp1 site. Taken together, these data suggest that a specific Sp1 consensus site is involved in the mediation of TGF-β induction as well as the basal promoter activity of the p15 gene and that Sp1 and Sp3 transcription factors might be involved in this regulation.

Transforming growth factor ␤s (TGF-␤s) 1 represent a large family of cytokines with diverse activities in the regulation of cell growth, differentiation, and morphogenesis (1)(2)(3). TGF-␤ causes growth inhibition of most epithelial, endothelial, fibroblast, neuronal, lymphoid, and hematopoietic cell types (3). TGF-␤ treatment induces growth arrest in the G 1 phase of the cell cycle, and this effect has been attributed largely to an inhibition of phosphorylation of the retinoblastoma susceptibility gene product, pRb (4). Progression through the G 1 phase of the cell cycle requires phosphorylation of pRb by G 1 cyclin-dependent kinase (CDK) complexes, particularly the cyclin D-CDK4 and cyclin D-CDK6 complexes (5). Phosphorylation of pRb releases transcription factors, including members of the E2F transcription factor family, required for the G 1 to S phase transition of the cell cycle (6).
Two distinct families of CDK inhibitors, represented by p16 and p21, have been identified recently and shown to be capable of binding to and inhibiting the activities of various CDK enzymes (for a recent review, see Ref. 5). The p16 INK4 family of CDK inhibitors specifically interacts with two closely related CDK proteins, CDK4 and CDK6, both of which have been strongly implicated as the physiological pRb kinases. One member of this family, p15 INK4B , was specifically up-regulated by TGF-␤ in human keratinocyte HaCaT cells (7). The steadystate level of p15 INK4B mRNA was induced 30-fold upon TGF-␤ treatment, implicating p15 INK4B as a primary effector of the TGF-␤-mediated cell cycle arrest (7). Previously it was shown that treatment of HaCaT cells with TGF-␤ caused rapid transcriptional induction of the p21 gene (also known as cip1/ WAF1/sdi1) through a p53-independent pathway, suggesting that p21 is also involved in mediating the cell cycle arrest caused by TGF-␤ (8).
The signaling pathways downstream of the TGF-␤ receptor complex that lead to the inhibition of cell cycle progression are still poorly understood. Since both CDK inhibitors, p15 and p21, are up-regulated by TGF-␤ treatment, they could be coordinately regulated through a similar mechanism. In this study, we attempt to elucidate the mechanism through which TGF-␤ specifically up-regulates the expression of p15 INK4B by a detailed analysis of the p15 INK4B promoter sequences.

EXPERIMENTAL PROCEDURES
Isolation of p15 Genomic DNA-A human placenta genomic library cloned in FIX II (Stratagene, La Jolla, CA) was screened with the full-length human p16 cDNA (9). Of 1.2 million phage plaques screened, eight positives were isolated. Two oligonucleotide primers, 5Ј primer (5Ј AGGATCCATGGTGATGATGGGCAGCGCCCGC 3Ј) and 3Ј primer (5Ј GAAGCTTGGGTAAGAAAATAAAGTCGTTG 3Ј), specific to p15 cDNA were designed based on the previously published MTS2 genomic sequence (10) and used in polymerase chain reaction amplification to distinguish p15 from p16. Three p15 genomic clones were obtained, and clone G8 was confirmed by DNA sequencing to correspond to the p15 gene as compared with the p15 cDNA sequence (7,11). A 1.2-kilobase pair DNA fragment containing a 440-bp exon 1 and 780-bp sequence upstream of 5Ј-cDNA was isolated from clone G8 and completely sequenced by Sanger's method (12).
Gel Mobility Shift Assay-Complementary oligonucleotides from Ϫ68 to Ϫ82 on the p15 promoter were annealed and labeled with [␥-32 P]ATP by T4 polynucleotide kinase. HaCaT cells were treated with or without TGF-␤ for 20 h, and the nuclear extract was prepared according to Dignam et al. (14). 1 g of total nuclear protein was incubated with 0.5 g of poly(dI⅐dC) as well as the appropriate oligonucleotide competitors or antisera (as indicated in the figure legend) for 20 min on ice in a 20-l reaction containing 20 mM Hepes (pH 7.5), 5 mM MgCl 2 , 60 mM KCl, 1 mM dithiothreitol, 0.1% Triton X-100, and 6% glycerol. 0.2 ng of the end-labeled probe (2 ϫ 10 5 cpm) were then added, and the incubation was continued for 20 min at 30°C. The protein-DNA complexes were resolved on a 4% non-denaturing polyacrylamide gel. The gel was dried and exposed to x-ray film. Rabbit anti-human Sp1 and Sp3 antibodies were generous gifts from Dr. J. Horowitz.
Luciferase Assays-HaCaT cells were plated onto 6-well plates at a density of 100,000 cells/well. Cells were grown for 48 h and transfected with 6 g of plasmid DNA per well with the DEAE-dextran method as described elsewhere (8). Human TGF-␤1 was added to a concentration of 100 pM, and luciferase activity was assayed 20 h later as described (8). For each transfected plasmid, two duplicates were assayed under the same conditions, and the mean relative light units (RLU) were used in all the figures.

RESULTS AND DISCUSSION
A 780-bp genomic DNA fragment, which contains sequences upstream of the previously reported 5Ј-ends of the p15 INK4B cDNA (7, 11), was cloned from a human genomic library. The 5Ј-end of the p15 mRNA was mapped to the adenosine in the sequence CCCCACTCT as shown in Fig. 3A by S1 nuclease protection assay (data not shown). Thus, the cloned 780-bp DNA fragment largely contains the p15 promoter sequence. The sequence around the initiation site matches the initiator sequence as defined by Smale and Baltimore (15). No apparent TATA sequence was found around the Ϫ25 to Ϫ30 region. Therefore, the p15 promoter may be defined as a TATA-less/ initiator promoter. The p15 promoter sequences are highly GC-rich (70% G ϩ C from Ϫ200 to Ϫ1).
To determine its inducibility by TGF-␤, the 750-bp p15 promoter sequence was inserted upstream of a luciferase reporter gene in the vector pGL2-basic (Fig. 1). When the resultant construct, p15P751-luc, was transiently transfected into Ha-CaT cells, a 10 -15-fold induction of luciferase activity was

Sp1 Site Mediates TGF-␤ Induction of p15 INK4B
routinely observed upon TGF-␤ treatment as measured by RLU (Fig. 1). Thus, the p15 promoter is capable of being induced by TGF-␤, and the transcription activation is at least partly responsible for the accumulation of p15 mRNA upon TGF-␤ treatment (7).
To identify specific promoter elements that confer the TGF-␤ induction, a series of 5Ј processive promoter deletion constructs were generated (Fig. 1). These deletion constructs were transiently transfected into HaCaT cells and assayed for luciferase activities in the absence or presence of TGF-␤. Fig. 1 shows that deletions up to the position of Ϫ110 relative to the initiation site did not change the fold of induction by TGF-␤ although the overall promoter activities in the presence of TGF-␤ dropped about 3-fold. Deleting sequences from Ϫ110 to Ϫ30, however, abolished TGF-␤ induction (Fig. 1), indicating that a TGF-␤-responsive element is located in this region.
There are three potential Sp1 binding sites within the Ϫ110 to Ϫ30 sequences upstream of the transcription initiation site of the p15 gene. Previously, a GC-rich sequence (GCCTCC) capable of binding to Sp1 and Sp3 was shown to be responsible for the induction of p21 gene by TGF-␤ treatment (30). Interestingly, this sequence is identical to the first Sp1 consensus site in the Ϫ110 to Ϫ30 fragment of the p15 promoter. To test the possibility that this sequence, or the other two Sp1 consensus sites, may be involved in the mediation of p15 promoter induction by TGF-␤, we generated more 5Ј promoter deletion constructs to delete either one (p15P97-luc), two (p15P69-luc), or all three Sp1 consensus sites (p15P47-luc) and assayed for their luciferase activities in the absence or presence of TGF-␤ (Fig. 2). Fig. 2 shows that deletion of the first Sp1 consensus site had little effect on either the induction fold by TGF-␤ or the basal promoter activity of the p15 luciferase construct, whereas deletion up to the second Sp1 consensus site reduced TGF-␤ induction dramatically from 19fold of the wild type promoter to 2.8-fold. Deletion of all three Sp1 consensus sites reduced TGF-␤ induction only slightly further to 1.5-fold near the background level. These data suggest that the second Sp1 consensus site is the most critical sequence for either the induction by TGF-␤ or the basal promoter activity of the human p15 promoter.
To confirm the importance of the second Sp1 consensus site in conferring the transcription inducibility of the p15 promoter by TGF-␤, a series of scanning mutation constructs were made in the promoter context of p15P113, which contains 113 base pairs upstream of the transcription initiation site and is fully capable of being induced by TGF-␤ (Fig. 3A). The scanning constructs were transiently transfected into HaCaT cells and assayed for their luciferase activities in the absence or presence of TGF-␤1. With the exception of p15P113-LS3, all mutants did not change significantly the fold of induction upon TGF-␤ treatment or the basal transcription activity (Fig. 3B). Mutant p15P113-LS3, which contains a mutation in the second Sp1 site within the Ϫ110 to Ϫ30 region, decreased the fold of induction from 12-fold of the wild type promoter to 4-fold (Fig. 3B). In addition, the basal promoter activity of p15P113-LS3 was also reduced significantly, approaching the background level (Fig.  3B). Together with the promoter deletion analysis, these data suggest that a specific Sp1 consensus site is important for the mediation of TGF-␤ induction of the p15 gene as well as its basal promoter activity.
To identify protein factors interacting with the second Sp1 site, we used DNA sequences from Ϫ68 to Ϫ82 on the p15 promoter covering the second Sp1 site in the gel mobility shift assay. As shown in Fig. 4, lane 2, three distinctive protein complexes (I, II, and III) were observed when this probe was incubated with nuclear extract prepared from HaCaT cells. All three complexes are specific to the probe since they were readily competed by an excess of cold homologous competitor (Fig. 4, lanes 9 and 10) but were resistant to the competition of nonspecific oligonucleotides (Fig. 4, lanes 7 and 8). Complex I was abolished when antibody against the human Sp1 transcription factor was included in the binding reaction and therefore represents the complex formed between Sp1 and the second Sp1 site on the p15 promoter (Fig. 4, lane 4). Similarly, complexes II and III represent the complexes formed between Sp3 and the second Sp1 site on the p15 promoter since both complexes were abolished when the binding reaction includes antibody against the transcription factor Sp3 (Fig. 4, lane 6). Together with the promoter mutation analysis, these data show that the second Sp1 site within the Ϫ110 to Ϫ30 region on the p15 promoter, which is capable of conferring the TGF-␤ inducibility of the p15 promoter, binds to the transcription factors Sp1 and Sp3.
Sp1 consensus sites are present in numerous gene promoters including many housekeeping genes and cellular proto-oncogenes (16 -18). The well characterized transcription factor Sp1 binds to its cognate binding site and activates transcription FIG. 3. Scanning mutation analysis of the p15 promoter. A, scanning mutation constructs, p15P113-LS1 through p15P113-LS6, are shown with the mutated sequences. Sequences from Ϫ113 to ϩ5 of the wild type promoter, p15P113, are also shown and the three Sp1 consensus sites underlined. The transcription initiation site is indicated by an arrow. Fold induction by TGF-␤ treatment as measured in panel B is shown for each construct. B, the scanning mutants were assayed for luciferase activities as described in the legend of Fig. 1. presumably through its interaction with the TBP-associated factor 110 (TAF II 110) (19). Recent studies suggest that Sp1 and a member of the Sp1 family, Sp3, are involved in the regulation of many growth-related cellular genes, including c-fos, c-myc, TGF-␤1, and TGF-␤3 genes, through a cis-acting element termed the pRb control element (16, 20 -22). A model has been proposed to postulate that the functional interaction between pRb and Sp1 in vivo results in the "superactivation" of Sp1mediated transcription (16). Here we show that a specific Sp1 consensus site is involved in the mediation of TGF-␤ induction of the p15 gene and that the Sp1 and Sp3 proteins could bind to this specific Sp1 site. The same site is also responsible for the basal promoter activity. Evidence is accumulating that basal transcription machinery is also capable of being regulated in cells, presumably through the action of TBP-associated factors (TAFs). For example, p53 transcription activation is mediated by TAF II 40 and TAF II 60 (23) and the presence of TAF II 150 and TAF II 250 stabilized the preinitiation complex assembled on an initiator-containing promoter while it destabilized the complex on an initiator-less promoter (24). A temperature-sensitive mutation in TAF II 250 has been shown to cause cell cycle arrest (25,26). It is possible that some TAFs could relay the cues received from growth signals and affect the transcription preinitiation complex assembled on certain growth-related genes.
A GC-rich sequence in the p21 promoter capable of binding to the Sp1 and Sp3 proteins was shown to be responsible for the induction of the p21 gene by TGF-␤ in HaCaT cells (30). Interestingly, this GC-rich sequence, termed T ␤ RE for TGF-␤-responsive element, was capable of conferring TGF-␤ inducibility when inserted into an exogenous promoter (30). Based on these studies, it is conceivable that the same GC-box binding factor(s) may be involved in the coordinate regulation of both p15 and p21 genes. Several proteins capable of binding to those GC-rich sequences including the Sp1 consensus sites have been identified. Some of these factors have extensive homologies with Sp1 and thus belong to the Sp1 transcription factor family, such as Sp2, Sp3 and ⌬Sp3 (16), whereas others are entirely different from Sp1, such as ETF and GC-box binding protein (GCF) (27,28). Notably, Sp3 and GC-box binding protein have been shown to bind to the GC-boxes, including Sp1 consensus sites, and repress transcription from certain genes (27,29). More experiments are needed to determine if Sp1, its family members, or other novel GC-box binding proteins are involved in the regulation of p15 and p21 genes by TGF-␤. It is possible that the subtle sequence variations and the optimal spacing between binding sites for various transcription factors could alter the balance between the positive and negative transcription regulators and consequently exert a different mode of transcription regulation. FIG. 4. Factors binding to the second Sp1 site on the p15 promoter. Complementary oligonucleotides covering the second Sp1 site on the p15 promoter were labeled with [␥-32 P]ATP and used in the gel mobility shift assay. Three protein-DNA complexes (I, II, and III), as indicated by arrows, were observed when the probe was incubated with 1 g of nuclear protein (lane 2). In lanes 3 and 4, preimmune sera or antibody against human Sp1 were included in the binding reaction, respectively. In lanes 5 and 6, preimmune sera or antibody against human Sp3 were included in the binding reaction. In lanes 7 and 8, a 50-or 250-fold excess of nonspecific oligonucleotides was included in the binding reaction. In lanes 9 and 10, a 50-or 250-fold excess of the homologous oligonucleotides was included in the binding reaction. Lane 1 is probe alone.