Regulation of the cyclin D3 promoter by E2F1.

We have previously demonstrated that ectopic expression of E2F1 is sufficient to drive quiescent cells into S phase and that E2F1 expression can contribute to oncogenic transformation. Key target genes in this process include master regulators of the cell cycle, such as cyclin E, which regulates G(1) progression, and cyclin A, which is required for the initiation of DNA synthesis. In the present work, we present novel evidence that a second G(1) cyclin, cyclin D3, is also potently activated by E2F1. First, an estrogen receptor-E2F1 fusion protein (ER-E2F1) potently activates the endogenous cyclin D3 mRNA upon treatment with 4-hydroxytamoxifen, which induces nuclear accumulation of the otherwise cytosolic fusion protein. Furthermore, trans-activation of cyclin D3 by ER-E2F1 occurs even in the presence of the protein synthesis inhibitor cycloheximide and thus appears direct. Second, all of the growth-stimulatory members of the E2F family (E2F1, -2, and -3A) potently activate a cyclin D3 promoter reporter, whereas growth-restraining members of the family (E2F4, -5, and -6) have little effect. Third, recombinant E2F1 binds with high affinity to the cyclin D3 promoter in vitro. Fourth, chromatin immunoprecipitation assays demonstrate that endogenous E2F1 is associated with the cyclin D3 promoter in vivo. Finally, mapping experiments localize the essential E2F regulatory element of the cyclin D3 promoter to a noncanonical E2F site in the promoter between nucleotides -143 and -135 relative to the initiating methionine codon. We conclude that in addition to cyclins E and A, E2F family members can also activate one member of the D-type cyclins, further contributing to the ability of the stimulatory E2F family members to drive cellular proliferation.

nases complexes permits the release of DNA-E2F-Rb-bound histone deacetylase proteins, which impart a dominant transcriptional repressing activity to promoter-bound Rb-E2F complexes (8). Loss of histone deacetylase-mediated inhibition allows a modest transcriptional activation of certain Rb-silenced genes, including cyclin E (9), but other E2F-Rb-regulated genes remain repressed (8). Once stimulated, cyclin E associates with cyclin-dependent kinase 2 and adds additional phosphate modifications to Rb later in G 1 (10). Upon its hyperphosphorylation by the cooperative efforts of the D-type cyclins and cyclin E, the Rb protein completely releases the E2F transcription factor. Liberated E2F3B and E2F4, which are the predominant E2Fs expressed (11)(12)(13)(14) and promoter-bound (15,16) in G 0 and early G 1 , then activate transcription (17)(18)(19)(20) of the more potent S phase-promoting E2Fs, E2F1, -2, and -3A (21). The combined activities of E2F1, -2, and -3A then lead to the potent activation of a large number of genes that are required for nucleotide biosynthesis, the firing of origins of replication, and the completion of replicative DNA synthesis (22)(23)(24). In this paradigm, E2F activates transcription of cyclin E, which then further activates E2F by stimulating the phosphorylation of Rb. This positive feedback loop leads to an irreversible commitment to entry into S phase at the restriction point (25).
In previous work, we showed that ectopic expression of E2F1 alone is sufficient to stimulate quiescent cells to enter cell cycle (26) and that stable expression of E2F1 can lead to oncogenic transformation (27). Subsequently, we utilized microarray technology to screen for genes that could account for the ability of E2F to induce cellular proliferation (24) and apoptosis (28,29). In addition to verifying several known E2F targets, our microarray screen suggested that E2F1 potently activates cyclin D3 transcription. Cyclins D1 and D2 were not activated by E2F1 expression in this screen. Here we verify that cyclin D3 is indeed a direct E2F-regulated transcript, and we localize the E2F regulatory element of the cyclin D3 promoter within a small and extremely GC-rich region in the vicinity of the major transcriptional start site. We conclude that in addition to cyclins E and A, E2F family members can also activate one member of the D-type cyclins, further contributing to the ability of E2F to drive the G 1 to S phase transition.
Cell Culture and Analysis-H1299 cells obtained from Dr. Jiandong Chen (Moffitt Cancer Center) were cultured in Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum. H1299 cells were transfected with pcDNA3-ER-E2F1 plasmid using the calcium phosphate method, and stable transfectants were selected with G418 (0.5 mg/ml). G418-resistant colonies were isolated after 2 weeks and expanded in the continued presence of 0.25 mg/ml G418. ER-E2F1positive lines were identified by Western blot using anti-hemagglutinin and anti-E2F1 antibodies. H1299-pcDNA3 stable cell lines were established by the same method and served as negative controls. 4-Hydroxytamoxifen (HT 1 ; 300 nM) was added to the medium to induce rapid ER-E2F1 nuclear accumulation of the ER-E2F1 protein. Cycloheximide (CHX) was used at a final concentration of 10 g/ml to block new protein synthesis. Apoptosis and cell cycle parameters were measured by flow cytometry using an Apo-BrdU Kit (DB PharMingen) as previously described (32).
Biochemical Assays-Total RNA was isolated from 5 ϫ 10 6 H1299 cells using the RNAeasy mini kit (Qiagen). RNase protection assays were carried out with the Riboquant hCYC1 (cyclin family) multiprobe templates (BD PharMingen). Briefly, the multiprobe templates were synthesized by in vitro transcription with incorporation of [ 32 P]dUTP and purified on Quick Spin RNA columns (Roche Applied Science). Labeled probe (1ϫ 10 6 cpm) was hybridized with 10 g of total RNA through a temperature gradient of 90 to 56°C over a 16-h period. Unprotected probe was removed by RNase digestion at 30°C for 1 h followed by separation of protected RNA fragments on a 5% polyacrylamide-urea gel and detection using autoradiography.
Transfections were performed using calcium phosphate with test DNAs totaling 20 g of DNA per 100-cm dish. Transfections included 300 ng of expression plasmid (pcDNA3-based vectors), 10 g of reporter firefly luciferase reporter plasmid (pGL3, Promega), 2 g of Renilla luciferase reporter plasmid (pRL-TK, Promega), and carrier DNA (sheared salmon sperm DNA) to equal 20 g of total DNA in each transfection. Cells were harvested 48 h after transfection, and luciferase assays were performed using the Dual-Luciferase Reporter Assay System following the manufacturer's protocol (Promega). Experiments were done in triplicate, and the relative activities and S.E. values were determined. To control for transfection efficiency, firefly luciferase values were normalized to the values for Renilla luciferase.

Generation of Cell Lines
Expressing ER-E2F1-Our previous microarray analysis utilized an adenovirus to express E2F1; thus, it was possible that the activation of the cyclin D3 promoter was a secondary consequence of E2F1 expression. To verify that activation of cyclin D3 by E2F1 was direct, we created a cell line that constitutively expresses a hemagglutinin-tagged estrogen receptor-E2F1 fusion protein (ER-E2F1). In this construct, E2F1 is fused to a transcriptionally inactive mutant of the murine estrogen receptor that is unable to bind estrogen yet retains high affinity for HT. In the absence of HT, the ER-E2F1 fusion protein is excluded from the nucleus and thus cannot directly affect transcription. Upon the addition of HT, however, the fusion protein rapidly enters the nucleus and induces transcription of E2F1 target genes. The advantage of this system is that, since the fusion protein is preexisting when HT is added, induction can occur in the absence of new protein synthesis. Therefore, promoters that are activated upon HT addition in the presence of the protein synthesis inhibitor CHX are most likely regulated by a direct mechanism.
ER-E2F1-expressing H1299 cell lines were derived by transfecting the pcDNA3-ER-E2F1 plasmid followed by selection in G418. After G418-resistant colonies were isolated and screened by Western blot with both hemagglutinin and E2F1 antibodies (data not shown), five colonies expressing detectable levels of the ER-E2F1 protein were chosen for further experiments. For negative controls, G418-resistant cell lines were also derived with the empty pcDNA3 plasmid. To identify the ER-E2F1 cell line with the tightest HT regulation, the five ER-E2F1-expressing lines were transfected with an E2F reporter plasmid in which the adenovirus E2 promoter was fused to firefly luciferase. Following transfection, HT (or solvent control) was added to induce nuclear accumulation of ER-E2F1. Fig. 1A shows that all five ER-E2F1 positive cell lines induced the reporter gene upon the addition of HT. Cell line H1299-ER-E2F1-15, which produced the most dramatic response (a 9-fold induction), was selected for subsequent experiments. As expected, H1299-pcDNA3 control cells did not activate the Ad E2 promoter in response to HT.
To determine whether the ER-E2F1 fusion protein in ER-E2F1-15 was expressed at sufficient levels to alter cell growth properties, cell cycle parameters were determined before and after the addition of HT. As demonstrated in Fig. 1B, a 48-h treatment with HT resulted in a 15% increase in the fraction of cells in S phase and a corresponding decrease in cells in G 0 /G 1 . In addition to its ability to induce S phase, E2F1 is also a potent inducer of apoptosis. Not surprisingly, Fig. 1B reveals that nearly 24% of H1299-ER-E2F1-15 cells underwent apoptosis after 48 h of HT treatment, whereas less than 1% of the cells were apoptotic in the absence of HT (Fig. 1B). Thus, the ER-E2F1 fusion protein expressed in H1299-ER-E2F1-15 cells functioned as expected in the presence of HT.
Cyclin D3 Expression Is Directly Activated by E2F1-To verify direct activation of the cyclin D3 promoter by E2F1, the cell line ER-E2F1-15 and a control line (pcDNA3) were treated with solvent only, CHX alone, HT alone, or a combination of both CHX and HT. Following 16 h of treatment, cyclin D3 message levels were measured by an RNase protection assay. Fig. 2A reveals that the cyclin D3 mRNA increased dramatically in the presence of HT alone and even more dramatically in the presence of both HT and cycloheximide. Since immediate early mRNAs such as cyclin D3 are stabilized in the presence of cycloheximide, the observed effects of cycloheximide alone or in combination with HT are expected (33). These results suggest that the induction of the cyclin D3 mRNA is a direct effect of the ER-E2F1 fusion protein and does not involve synthesis of another protein that is activated or repressed by E2F1. Cyclin D3 protein expression was also measured by Western blot. In Fig. 2B, 48 h after drug treatment, the cyclin D3 protein was also induced in the presence of HT, reflecting the increase in mRNA. Levels of cyclin D3 mRNA and protein did not change in response to HT in the control cell line, as expected. These results demonstrate that cyclin D3 is a direct target of E2F1. In contrast to cyclin D3, cyclin D1 and D2 were repressed following HT treatment (data not shown).
E2F1, -2, and -3A Trans-activate the Cyclin D3 Promoter-To determine whether the activation of cyclin D3 expression by E2F1 was mediated via the promoter, we used luciferase reporter constructs to examine the effect of E2F1 expression on the level of transcription from the cyclin D3 promoter. Previous analysis (30) has mapped the cyclin D3 promoter to a region from nucleotide position Ϫ1017 to ϩ6 relative to the ATG start codon. To identify E2F1-responsive elements, we tested luciferase reporter activity of three cyclin D3 promoter constructs (Ϫ1017/ϩ6, Ϫ362/ϩ6, and Ϫ202/ϩ6). As demonstrated in Fig.  3A, transient co-transfection of E2F1 resulted in strong activation of all three of the cyclin D3 promoter constructs. The smallest of these constructs (Ϫ202/ϩ6) was fully responsive to exogenous E2F, suggesting that this small region contains the E2F1-responsive element.
Previous studies using various E2F1 mutants have demonstrated that both the DNA-binding domain and the transcriptional activation/Rb-binding domains of E2F1 are required for trans-activation to occur. To determine whether both of these functional domains are required for cyclin D3 promoter activation, two previously described E2F1 mutant proteins were tested for the ability to activate the cyclin D3 promoter. The first was E2F1 E132 , which lacks the ability to bind DNA (34), and the second was E2F1  , which lacks the trans-activation domain (35,36). As expected, both E2F1 mutants had diminished ability to trans-activate the cyclin D3 promoter (Fig. 3B), demonstrating that activation of the cyclin D3 promoter is dependent on the ability of E2F1 to bind DNA and to activate transcription.
Based upon structural and functional relatedness and on cell cycle expression pattern, the E2F family members are classified into two major subfamilies (37): the growth-promoting E2Fs (including E2F1, -2, and -3A) and the growth-restraining E2Fs (including E2F3B, -4, -5, and -6). Growth-promoting E2Fs are primarily expressed at the G 1 /S boundary and are capable of driving quiescent fibroblasts into S phase upon overexpression. In contrast, the growth-restraining E2Fs are expressed constitutively during the cell cycle and induce S phase less efficiently or not at all when overexpressed. To determine which of the E2Fs most potently modulate the cyclin D3 promoter, several E2F family members were tested in luciferase reporter assays. Fig. 3C shows that the cyclin D3 promoter is most potently activated by the growth-promoting E2Fs (especially E2F1 and E2F2), whereas growth-restraining E2Fs have much smaller effects, if any. These results suggest that cyclin D3 is a target of multiple growth-promoting E2F1s and is not an E2F1-specific target.
E2F1 Binds to Cyclin D3 Promoter in Vivo and in Vitro-The experiments described above utilized ectopic E2F expression, which is subject to the concern that the observed regulation is not physiological. Thus, ChIP assays were performed on native H1299 cells to ascertain whether endogenous E2F1 binds to the cyclin D3 promoter in vivo. As shown in Fig. 4A, PCR primers that span the region of Ϫ202 to ϩ6 of the cyclin D3 promoter clearly detect cyclin D3 promoter DNA in ChIP samples generated using an E2F1 antibody. However, chromatin immunoprecipitations with either an irrelevant antibody (anti-RhoA) or no added antibody resulted in the absence of detectable cyclin D3 promoter DNA. PCR primers specific for the human actin promoter (negative control) did not detect actin promoter DNA in E2F1 ChIP samples, as expected from previous work demonstrating that E2F1 does not bind the actin promoter (15). These data demonstrate that E2F1 associates with the cyclin D3 promoter under physiological conditions. As a complementary approach to ChIP assays, an EMSA was performed to determine whether recombinant E2F1-DP1 would bind the cyclin D3 Ϫ202/ϩ6 promoter region in vitro. Fig. 4B demonstrates that recombinant E2F1-DP1 present in extracts of Sf9 cells co-infected with E2F1-and DP1-expressing baculoviruses (38) binds to the cyclin D3 Ϫ202/ϩ6 promoter fragment, whereas no such binding activity is detectable in uninfected Sf9 cell extracts. The addition of an excess of double-stranded oligonucleotides containing a consensus E2F1 binding site (WT) abolished formation of the E2F1-DP1 complex on the cyclin D3 promoter, whereas the addition of a mutated version (MT) of the consensus oligonucleotide had no effect. The addition of an E2F1 antibody (100 ng) resulted in the retarded mobility of the putative E2F1-DP1-cyclin D3 pro-FIG. 3. E2F1 trans-activates cyclin D3 promoter/reporter constructs dependent upon its DNA binding and trans-activation domains. H1299 cells were transfected with 10 g of each reporter construct and 300 ng of E2F expression plasmid or mock vector (pcDNA3). Carrier DNA was used to bring total DNA/transfection to 20 g. A, the luciferase activity of each reporter in absence of E2F1 expression is normalized to 1. B, the indicated E2F1 mutants were compared for the ability to induce the Ϫ202/ϩ6 promoter/reporter. E2F1 1-284 lacks the trans-activation domain, and E2F1 E132 lacks the ability to bind DNA. C, the indicated E2F family members (human cDNAs) were compared for the ability to induce the Ϫ202/ϩ6 promoter/ reporter. Each of these pcDNA3 vectors has been previously described and expresses comparably high levels of the E2F proteins (55). FIG. 5. The E2F1 regulatory element is located between ؊143 and ؊135 of the cyclin D3 promoter. A, luciferase reporter assays were used to compare the response of the indicated cyclin D3 promoter constructs with the expression of E2F1. B, EMSAs were used to compare the binding affinity of recombinant E2F1-DP1 to the indicated fragments of the cyclin D3 promoter. Probes were labeled and normalized, and equal counts of radioactivity were loaded in each panel (10,000 cpm/lane). Oligonucleotide competition and antibody supershift assays were performed as in Fig. 4. C, a series of 15-bp partially overlapping oligonucleotides spanning the E2F1 regulatory region of the cyclin D3 promoter were used in competition assays. WT and MT are the 25-bp oligonucleotides used in Fig. 4. C15 corresponds to a 15-bp double-stranded oligonucleotide containing a consensus E2F site, which serves as a positive control. Although C15 has the same consensus sequence as WT, it is 10 bp shorter, E2F1 Regulates Cyclin D3 16774 moter complex. Together, these data support the conclusion that E2F1 activates cyclin D3 via direct binding to the promoter in a manner dependent upon its transcriptional activation domain.
Identification of the E2F1 Regulatory Region of the Cyclin D3 Promoter-To further define the E2F1-responsive region of the cyclin D3 promoter, we generated several additional promoter constructs. Constructs Ϫ202/Ϫ146, Ϫ202/Ϫ134, and Ϫ202/ Ϫ112 possessed deletions from the 3Ј-end of the promoter. Fig.  5A reveals that deletion of sequences downstream of Ϫ134 did not abolish induction by E2F1; in fact, constructs Ϫ202/Ϫ134 and Ϫ202/Ϫ112 were both activated nearly 4-fold by E2F1. However further deletion to Ϫ146 (Ϫ202/Ϫ146) abolished the E2F1 response. Constructs Ϫ159/ϩ6 and Ϫ115/ϩ6 had additional nucleotides deleted from the 5Ј-end of the promoter. Fig.  5A reveals that deletion of sequences upstream of Ϫ159 (Ϫ159/ ϩ6) had little effect on the E2F1 response. In contrast, further deletion of sequences upstream of Ϫ115 (Ϫ115/ϩ6) abolished E2F1 responsiveness. Taken together, these results demonstrate that the critical E2F1-responsive element resides within a highly GC-rich region between Ϫ159 and Ϫ134.
Despite the fact that the Ϫ159 to Ϫ135 region clearly accounts for the E2F1 responsiveness of the cyclin D3 promoter, this region does not contain a consensus E2F binding site (TTTCGCGC). To demonstrate that this region nonetheless accounts for E2F binding to the D3 promoter, additional EM-SAs were performed using various cyclin D3 promoter fragments as probes. Fig. 5B demonstrates that all fragments containing the Ϫ159/Ϫ135 region (Ϫ202/ϩ6, 202/Ϫ134, Ϫ202/ Ϫ112, and Ϫ159/ϩ6) bind recombinant E2F1. On the other hand, fragment Ϫ202/Ϫ146 that lacks the Ϫ159/Ϫ135 sequence does not bind recombinant E2F1-DP1. Although fragment Ϫ115/ϩ6 binds E2F1 in EMSA, it is not responsive to E2F1 in luciferase reporter assays. This suggests that whereas there is at least one functional E2F1 binding site with the region encompassed by the Ϫ115/ϩ6 construct, this region is insufficient to mediate a transcriptional response in the absence of the Ϫ159/Ϫ135 region. Thus, the Ϫ159/Ϫ135 region of the cyclin D3 promoter contains the critical E2F regulatory element.
Since the GC-rich Ϫ159/Ϫ135 region has no consensus E2F binding site, a series of oligonucleotides spanning this short region were examined in competition assays to empirically determine the best E2F binding site within it. Each oligonucleotide in this series was 15 bp in length and overlapped its 5Ј and 3Ј neighbors by 10 bp and spanned from Ϫ146 to Ϫ111. These oligonucleotides were used as competitors in E2F electrophoretic mobility shift assays that also utilized recombinant E2F-DP1 as a source of DNA binding activity and a fragment of the dihydrofolate reductase promoter as a labeled probe. The data in Fig. 5C reveal that of the seven oligonucleotides tested, oligonucleotide 2 (Ϫ145 to Ϫ131) was by far the best competitor for E2F binding to the dihydrofolate reductase probe. In fact, oligonucleotide 2 was nearly as good of a competitor as the C15 oligonucleotide that contains a consensus E2F1 binding site (TTTCGCGC). Oligonucleotide 1, which overlaps oligonucleotide 2 by 10 bp, also competed for E2F binding but was clearly less active than oligonucleotide 2. Oligonucleotides 3-7 appeared to lack competitor activity even at the highest doses used. The sequence common to oligonucleotides 1 and 2 was 5Ј-GGCGCGCGCG-3Ј.
To determine that the DNA sequence common to oligonucleotides 1 and 2 was indeed the central E2F1 regulatory element of the cyclin D3 promoter, this putative binding site was removed, and the resulting construct (Ϫ159/ϩ6(del9)) that lacks Ϫ143 to Ϫ135 was tested for promoter activity. As expected, Fig. 5A reveals that this construct retained only a modest 2.3-fold induction by E2F1 as compared with a 6-fold induction for an identical construct Ϫ159/ϩ6 which contains the putative E2F element. To determine whether this region is sufficient in and of itself to support an E2F1 response, a final construct (Ϫ146/Ϫ111) that contains the E2F regulatory element and three transcriptional start sites was examined. Fig. 5A reveals that this region is not sufficient to mediate transcriptional activation by E2F1 although it is sufficient to bind recombinant E2F1 (see Fig. 5B). Thus, it appears that this element binds E2F and is essential for a majority of the E2F1 responsiveness of the cyclin D3 promoter; however, it is not sufficient to mediate the E2F response and probably cooperates with other promoter elements including other nonessential E2F binding sites.

DISCUSSION
The central role of E2F in cell growth control was firmly established 10 years ago by the demonstration that ectopic expression of E2F1 could drive quiescent fibroblasts into S phase (26). Subsequent work has shown that E2F2 and E2F3A are also potent inducers of S phase (21,31). Ironically, studies have found that certain members of the E2F family are also potent inducers of apoptosis (21,31,39,40), demonstrating that there is a delicate, although poorly understood, balance between cellular proliferation and survival. In an attempt to understand how members of the E2F family control cell fate, a number of recent studies have measured global gene expression following induction of various members of the E2F family (22-24, 41, 42). Not surprisingly, these studies implicate members of the E2F family in the regulation of many genes involved in the G 1 to S phase transition, including replication activities such as replication protein C, DNA primase, DNA topoisomerase, flap endonuclease, and DNA ligase.
In our own microarray work (24), we identified cyclin D3 as an activated target of E2F1. This observation was something of a surprise for three reasons. First, in most cell cycle models, the stimulation of cyclin D3 in early and mid-G 1 temporally precedes activation of the S phase-promoting E2Fs at the G 1 to S phase boundary. Second, previous work has carefully documented repression of the cyclin D1 promoter by E2F1 (43) and by Myc (44,45). Repression of cyclin D3 by E2F, and Myc has been understood to represent a negative feedback loop. Third, consensus E2F binding sites are not conserved in the cyclin D3 promoters of mice (46), rats (47), and humans (30), although the promoters show over 70% sequence identity within the 700-bp region upstream of their translational start sites.
Due to the essential role played by the D-type cyclins in cell growth control, we considered it of interest to validate our microarray data and to determine the mechanism by E2F1 activates cyclin D3 expression. Thus, we constructed an H1299 cell line that expresses a HT-inducible ER-E2F1 fusion protein accounting for the reduced affinity of E2F1 to C15 relative to WT. Oligonucleotides 1-7 are the individual sequences used as competitors at 80, 160, and 320 ng/l reaction, respectively, with increasing amounts indicated by the ramp above the lanes. D, data from Figs. 4 and 5 are summarized. In D, BA represents binding activity, LA represents luciferase reporter activity, and the stars indicate the major transcription start sites. In the bottom element, the boxed sequence represents the essential E2F1 regulatory element of the cyclin D3 promoter, and arrows indicate major transcription start sites (30). Comparison of the consensus E2F binding sequence and the essential E2F site of the cyclin D3 promoter reveal an identical CGCGC core but lack of TTT trinucleotide in the cyclin D3 promoter.
to verify that activation of cyclin D3 by E2F1 is direct and not an indirect consequence of E2F1 expression (Fig. 1). Experiments in which this ER-E2F1 fusion protein is induced in the presence and absence of cycloheximide (Fig. 2) strongly support the hypothesis that E2F1 directly activates the endogenous cyclin D3 promoter. Further experimentation determined that E2F1, as well as several other members of the E2F family, activates transcription of the cyclin D3 promoter (Fig. 3). Binding of endogenous E2F1 to the cyclin D3 promoter in vivo was confirmed by CHIP assay, and direct binding of recombinant E2F1-DP1 to the cyclin D3 promoter was confirmed by EMSA (Fig. 4). The E2F-responsive region of the cyclin D3 promoter by E2F1 was mapped to positions Ϫ143 to Ϫ135 within a GC-rich region of the promoter that contains several of the cyclin D3 transcription start sites (Fig. 5). The E2F1 binding sequence of the cyclin D3 promoter matches the GC core of the E2F consensus sequence perfectly, but it completely lacks the TTT of the TTTCGCGC consensus sequence (see Fig. 5D). Although this site lacks the TTT of the consensus E2F binding site, it nonetheless binds with comparable affinity as a consensus sequence (C15 in Fig. 5C) to recombinant E2F1-DP1. It may be noteworthy that the E2F regulatory elements of a growing number of well characterized promoters (48 -50), including the cyclin A promoter (51), also lack consensus E2F binding sites and are GC-rich. Thus, our finding that the cyclin D3 promoter is activated via a nonconsensus E2F binding site is not novel but does further support the premise that sequence analysis alone is not sufficient to identify E2F-responsive elements (52).
We have found no evidence that E2F1 increases the expression of either cyclin D1 or D2. In fact, in H1299-ER-E2F1-15 cells we find that HT treatment represses expression of both cyclins D1 and D2 (data not shown). This observation is not unprecedented, since Myc is known to repress the cyclin D1 promoter (44,45) but to activate the cyclin D2 promoter (53,54). The natural question that arises is "Why would E2F1 or Myc up-regulate one G 1 cyclin and repress another?" We suggest that these observations reflect the fact that different Dtype cyclins respond to different transcription factors and signaling pathways with the net outcome begin determined by the relative contributions of various signals. In this model, cyclin D3 is merely the D-type cyclin specifically wired to respond to positive feedback from E2F during the course of a cell cycle.
In conclusion, we have confirmed that cyclin D3 is indeed a direct target of E2F, thus validating the power of genome-wide technologies to identify individual E2F targets. These findings add cyclin D3 to the list of well characterized E2F-activated cyclins that includes cyclins A1 (51) and E1 (9). By analogy to the ability of E2F1 to activate cyclin E1, the most straightforward interpretation of this work is that E2F-mediated activation of cyclin D3 contributes to the phosphorylation of Rb, thereby contributing to the potent ability of E2F to drive the G 1 to S phase transition.