Identification of Aurora-A as a Direct Target of E2F3 during G2/M Cell Cycle Progression*

Aurora-A is a centrosome kinase and plays a pivotal role in G2/M cell cycle progression. Expression of Aurora-A is cell cycle-dependent. Levels of Aurora-A mRNA and protein are low in G1/S, accumulate during G2/M, and decrease rapidly after mitosis. Previous studies have shown regulation of the Aurora-A protein level during the cell cycle through the ubiquitin-proteasome pathway. However, the mechanism of transcriptional regulation of Aurora-A remains largely unknown. Here, we demonstrated that E2F3 modulates Aurora-A mRNA expression during the cell cycle. Ectopic expression of E2F3 induces Aurora-A expression. Stable knockdown of E2F3 decreases mRNA and protein levels of Aurora-A and delays G2/M entry. Further, E2F3 directly binds to Aurora-A promoter and stimulates the promoter activity. Deletion and mutation analyses of the Aurora-A promoter revealed that a region located 96-bp upstream of the transcription initiation site is critical for the activation of the promoter by E2F3. In addition, expression of E2F3 positively correlates with the protein level of Aurora-A in human ovarian cancer examined. These results indicate for the first time that Aurora-A is transcriptionally regulated by E2F3 during the cell cycle and that E2F3 is a causal factor for up-regulation of Aurora-A in a subset of human ovarian cancer. Thus, the E2F3-Aurora-A axis could be an important target for cancer intervention.

Aurora family of serine/threonine protein kinases is evolutionally conserved from human to Drosophila and yeast (1,2). In mammals, the Aurora kinase family comprises three members: Aurora-A, Aurora-B, and Aurora-C. They share similar structures, with their highly conserved catalytic domains flanked by very short C-terminal tails and N-terminal domains of variable lengths. The N-terminal domains of the three Auroras share low sequence conservation, which may determine selectivity during protein-protein interactions. As a mitotic kinase, activated Aurora-A is required for mitotic entry, centrosome maturation, and chromosome segregation (1,2). Aurora-A protein localizes to centrosomes during interphase and to both spindle poles and spindle microtubules during early mitosis. Ectopic expression of Aurora-A leads to an increase in centrosome number, causes catastrophic loss or gain of chromosomes, and then results in malignant transformation. Aurora-A kinase activity is regulated by phosphorylation, dephosphorylation, and association with a number of proteins such as HEF1, TPX2, or Bora (3)(4)(5).
Aurora-A is located at chromosome 20q13.2, which is commonly amplified in various epithelial malignant tumors, including breast, colon, bladder, ovarian, and pancreatic cancers, and the levels of Aurora-A mRNA and protein are also increased in those tumors (6 -10). However, alterations of Aurora-A at mRNA and/or protein levels are much more common than gene amplification. Discrepancy between amplification and overexpression rates was also reported in breast cancer, gastric cancer, and ovarian cancer (9,11,12). Therefore, Aurora-A overexpression is likely to be regulated not only by gene amplification but also by other mechanisms such as transcriptional activation and/or suppression of protein degradation. Expression of Aurora-A is cell cycle-regulated (2). Aurora-A mRNA and protein levels begin to accumulate at G 2 to M phase. After metaphase, Aurora-A protein is rapidly degraded by the ubiquitin-mediated proteolysis, which is promoted by the hCdh1activated anaphase-promoting complex/cyclosome. However, transcriptional regulation of Aurora-A during the cell cycle is largely unknown.
It has been well documented that the E2F family of transcription factors plays an important role in the control of expression of genes required for DNA replication, as well as further cell cycle progression (13,14). In addition, previous studies have demonstrated that a substantial number of E2F-induced genes are normally regulated at G 2 of the cell cycle, encoding proteins known to function in mitosis (15)(16)(17)(18). Moreover, the studies in Drosophila have provided evidence for a connection between E2F activity and the control of mitotic activities (19). A recent report shows that both E2F1 and E2F3 are required for cells to enter the S phase from a quiescent state, whereas only E2F3 is necessary for the S phase in growing cells (20). The acute loss of E2F3 activity affects the expression of genes encoding DNA replication and mitotic activities, whereas the loss of E2F1 affects a limited number of genes that are distinct from those regulated by E2F3 (21). Here we report that E2F3 up-regulates Aurora-A at transcriptional level during G 2 and the onset of mitosis. Among several putative E2F3-responsive elements, the region between Ϫ96 to transcriptional initiation site is responsible for E2F3 induction of Aurora-A promoter activity. Knockdown of E2F3 reduces Aurora-A expression and delays mitotic cell cycle progression.
Synchronization and Cell Cycle Analysis-Cell synchronization at the G 1 /S phase was performed using a double-thymidine block (2). Briefly, HeLa cells were grown in 60-mm plates, and thymidine (Sigma) was added to the culture medium at a final concentration of 2 mM for 12 h. Following two washes with serum-free medium, the cells were released from the thymidine block by culture in fresh medium containing 24 mM 2-deoxycytidine. After 9 h of incubation, the second thymidine block was initiated and completed after 14 h. The cells were released from the block by washing in warm phosphate-buffered saline and replacing with complete culture medium. At different time points, the cells were fixed in 70% ethanol. The fixed cells were rinsed with phosphate-buffered saline and then stained with solution containing 50 g/ml propidium iodide, 0.05% Triton X-100, and 0.1 mg/ml RNase A (Sigma). Cell synchrony at various cell cycle stages was monitored by flow cytometry. To synchronize cells in the M phase, HeLa cells were incubated in Dulbecco's modified Eagle's medium containing 200 ng/ml nocodazole for 14 h. Whole cell lysates and total RNA were isolated from parallel experiments for Western and Northern blotting analyses.
Luciferase Assay-The cells were cultured in 24-well plates and transfected with pGL3-Aurora-A reporters, a ␤-galactosidase expressing internal control, and the effector plasmids indicated in the figure legend. The amount of DNA in each transfection was kept constant by the addition of empty vector. After 36 h of transfection, luciferase activity was measured using a luciferase kit (Promega). The ␤-galactosidase activity was measured by using Galato-Light (Tropix). Transfection efficiency was normalized by ␤-galactosidase expression. Luciferase activity was expressed as relative activity to ␤-galactosidase. Each experiment was repeated in triplicate.
Northern and Western Blotting Analyses-Total RNA was isolated from cells using TRIzol reagent (Invitrogen). Northern blot analysis was performed as previously described (24). Briefly, 20 g of total RNA from each sample was separated on 1.0% agarose. After transferring to GeneScreen Plus Membrane (PerkinElmer Life Sciences) and prehybridization, the membrane was hybridized with [ 32 P]dCTP-labeled Aurora-A cDNA probe in Express hybridization solution (Clontech). For Western blot, equal amounts of protein were separated by SDS-PAGE and electroblotted onto membranes. Following blocking in TBS-T containing 5% milk, the membranes were probed with appropriate antibodies indicated in the figure legends.
Small Hairpin RNA Lentiviral Infection-Knockdown of E2F3 was carried out by infection of HeLa cells with lentiviruses (pLKO.1-shRNA; Sigma) 2 expressing five different shRNAs of human E2F3, four of which recognize the amino acids coding sequence of E2F3 and the other the 3Ј-untranslation region. Briefly, HeLa cells were plated in 60-mm plates. Following culture for 24 h and the addition of polybrene, the cells were infected pLKO.1-shRNA/E2F3 viruses at multiplicity of infection of 100. After swirling the plate, the cell-viral particle mixture was incubated at 37°C overnight and then replaced with complete culture medium. For transient experiments, the cells were harvested after 48 h of infection and assayed by Western blot. Stable knockdown of E2F3 cell lines were established by selection with puromycin (5 g/ml).
Chromatin Immunoprecipitation (ChIP) Assay-ChIP assay was performed essentially as previously described (25). Solubilized chromatin was prepared from a total of 2 ϫ 10 7 asynchronously growing HEK293 cells that were transfected with E2F3. The chromatin solution was diluted 10-fold with ChIP dilution buffer (1.1% Triton X-100, 1.2 mM EDTA, 167 mM NaCl, 16.7 mM Tris-HCl, pH 8.1, 0.01% SDS, protease inhibitors) and precleared with protein A/G-agarose beads blocked with 2 g of sheared salmon sperm DNA and preimmune serum. The precleared chromatin solution was divided and utilized in immunoprecipitation assays with either an anti-E2F3a antibody (N-20; sc-879, Santa Cruz) or an anti-actin antibody. Following wash, the antibody-protein-DNA complex was eluted from the beads by incubating in 1% SDS, 0.1 M NaHCO 3 at room temperature for 20 min. The protein-DNA cross-linking was reversed, and protein and RNA were removed by incubation with 10 g of proteinase K and 10 g of RNase A at 67°C for 6 h to overnight. Purified DNA was subjected to PCR with primers specific for putative E2F3-binding sites within the Aurora-A promoter. The sequences of the PCR primers used are as follows: site A, sense 5Ј-GAATCCTGCCCAATCTACCGC-TCC-3Ј and antisense 5Ј-GAAAAGCAAGAGAGAGTGG-GACCG-3Ј; site B, sense 5Ј-GCTATCGATCGGTCCCAC-TCTCTC-3Ј and antisense 5Ј-CTTGAGTCGCGTTTAAG-GGCGCAC-3Ј; site CϩD, sense 5Ј-CGACGCGTTGGCTC-CACCACTTCCGG-3Ј and antisense 5Ј-CCAGGAG-CTCAGCCGTTAGAATTCAAAGG-3Ј.

Expression of the Members of E2F Family and Aurora-A dur-
ing the Cell Cycle-Although mRNA and protein levels of Aurora-A are elevated at G 2 /M phase, the underlying molecular mechanism remains elusive (2). E2F family members have been shown to play a critical role in cell cycle progression through transcriptional regulation of a number of cell cycleassociated genes. To examine whether Aurora-A is regulated by E2F family, we compared the expression patterns of E2F family proteins with Aurora-A during the cell cycle (Fig. 1A). HeLa cells were synchronized with double-thymidine blocking and released for various time points. The cell cycle progression was monitored by flow cytometry. Before releasing, over 60% of the total cell population was at S phase. About 6 h, the cells entered G 2 /M. After 16 h, the majority of the cells were in G 1 phase (Fig.  1, B and C). E2F1 and E2F2 protein levels were high in G 1 /S and decreased at G 2 /M phase (Fig. 1A). E2F4 and E2F5 proteins distributed to the whole cell cycle. Notably, E2F3 protein was elevated in S and maintained through G 2 /M phases and decreased in G 1 phase. Recent studies have shown that E2F3, but not other E2F family members, regulates the expression of genes that are involved in G 2 /M (21), implying that Aurora-A could be transcriptionally regulated by E2F3.
Ectopic Expression of E2F3 Increases and Knockdown of E2F3 Decreases Aurora-A Protein and mRNA Levels-To determine whether E2F3 transcriptionally regulates Aurora-A, we transfected E2F3 into three cell lines, which include two human epithelial cell lines HeLa and A2780S and a mouse fibroblast NIH3T3 ( Fig. 2A). Immunoblotting analysis showed that E2F3 up-regulated Aurora-A protein expression in a dose-dependent manner ( Fig. 2A). Moreover, mRNA levels of Aurora-A was also induced by ectopic expression of E2F3 (Fig.  2B). To further demonstrate that Aurora-A expression is controlled by E2F3, we knocked down E2F3 by infection of HeLa cells with the lentiviruses (pLKO.1-E2F3/shRNA) expressing five different E2F3 shRNAs, four of which are corresponding to the coding region and the other of which matches the 3Ј-untranslated region. Fig. 2C showed that E2F3 was considerably reduced in the cells infected with three individual shRNA as well as their combination. After selection with puromycin, stable E2F3 knockdown pool cells were obtained. Expression of Aurora-A was significantly reduced in these cells both at interphase and mitosis (Figs. 2D and 3B).
Knockdown of E2F3 Delays G 2 /M Entry and Reduces Aurora-A Expression during the Cell Cycle-Having demonstrated E2F3 transcriptional regulation of Aurora-A, we then examined the effects of knockdown of E2F3 on G 2 /M progression and Aurora-A expression during the cell cycle. pLKO.1-shRNA vector-infected and E2F3 stable knockdown HeLa cells were synchronized by double-thymidine block. Flow cytometry analysis revealed that cells with knockdown of E2F3 enter G 2 /M phase later as compared with the pLKO.1 vector-infected cells (Fig. 3A). Consistent with cell cycle change, the elevation of Aurora-A mRNA and the protein levels were ϳ2 or 3 h delayed by knockdown of E2F3. Moreover, the mRNA and protein levels of Aurora-A at G 2 /M phase were also reduced in E2F3 knockdown cells as compared with control shRNA-treated cells (Fig. 3, B and C). However, Aurora-A protein underwent degradation at late M phase in both cell lines (Fig. 3C). These data further suggest that Aurora-A is transcriptionally regulated by E2F3 during the cell cycle. E2F3 Binds to and Transactivates the Aurora-A Promoter during G 2 /M Phase-We next examined whether the Aurora-A promoter is transactivated by E2F3. HeLa cells were transfected with pGL3-Aurora-A/Ϫ1486/ϩ354-Luc reporter and increasing amounts of E2F3. Luciferase reporter assay revealed that Aurora-A promoter activity was induced by E2F3 in a dose-dependent manner (Fig. 4A). Moreover, basal level of Aurora-A promoter activity, especially at mitosis, was reduced in E2F3 knockdown cells as compared with pLKO.1 vector-infected HeLa cells (Fig. 4B).
Sequence analysis showed four putative E2F-binding elements (TT(C/G)GCGC(C/G)) within the promoter (Fig. 4C). To define the response region(s) of the promoter to E2F3, we created a series of deletion mutants of Aurora-A promoter. Reporter assay showed that a mutant with deletion from Ϫ1486 to Ϫ415 significantly deceased E2F3-induced promoter activity, whereas it still contains all four putative E2F-binding elements, implying that this region is of transactivation function. Moreover, the deletion of two distal E2F-binding sites (i.e. from Ϫ1486 to Ϫ189; A and B) further reduced the promoter activity. However, promoter activity of the mutants with additional deletion of either Ϫ189 to Ϫ124 or Ϫ189 to Ϫ96, both of which retain the 2 E2F response elements (C and D) proximal to the transcriptional starting site, was significantly induced by E2F3. The promoter activity was completely abrogated by further deletion of the proximal E2F response elements (Fig. 4C). These results suggest that the region from Ϫ189 to Ϫ96 contains a repression element(s) and that all four E2F-binding sites are responsive to E2F3. However, the two binding elements proximal to the transcriptional start site are sufficient for E2F3 transactivation of the Aurora-A promoter.
To determine whether E2F3 could directly bind to the E2F-binding sites of the Aurora-A promoter  in vivo and to further define the E2F3 response elements in the promoter, we carried out ChIP assay, which detects specific genomic DNA sequences that are associated with a particular transcription factors in intact cells. Unsynchronized HeLa cells were immunoprecipitated with an anti-E2F3 antibody. The E2F3 bound chromatin was subjected to PCR using three pairs of oligonucleotide primers that amplify regions spanning each distal and two proximal E2F3-binding sites within the Aurora-A promoter because two proximal E2F-binding sites are so close that they could not be separated by ChIP assay. As shown in Fig. 5 (A and B), the anti-E2F3 antibody pulled down all E2F-binding sites. In contrast, immunoprecipitation with an irrelevant antibody (e.g. anti-Actin) or with E2F3 antibody but primers without spanning E2F binding element resulted in the absence of bands in these sites. These results indicate that E2F3 directly binds to the Aurora-A promoter. By mutation of individual or combinational E2F3-binding sites (CG 3 AT) in Aurora-A promoter, we further demonstrated that the two E2F response elements proximal to the transcriptional start site are required for E2F3 transactivation of the Aurora-A promoter (Fig. 5, A and B).
Because mRNA and protein levels of Aurora-A are low at G 1 and gradually increase during the G 2 /M phase, we examined whether E2F3 binding to and activation of the Aurora-A promoter is cell cycle-dependent. HeLa cells were transfected with pGL3-Aurora-A or pGL3 vector. Following synchronization with double-thymidine block, ChIP and luciferase reporter assays were performed at different phases of the cell cycle. Fig. 5C shows that E2F3 barely interacts with Aurora-A promoter during the G 1 /S phase. The binding and transactivation activities of E2F3 toward the Aurora-A promoter were significantly increased upon the cell entering the G 2 /M phase. These data further support the notion that E2F3 plays a pivotal role in Aurora-A expression during the G 2 /M phase.
Correlation of Expression of E2F3 and Aurora-A-To further demonstrate E2F3 regulation of Aurora-A, we transfected GFP-E2F3 into HeLa cells. After 48 h of transfection, the cells were immunostained with anti-Aurora-A antibody. As shown in Fig. 6, unsynchronized cells expressing GFP-E2F3 exhibited higher density and clearer centrosome than the cells that did not, further indicating E2F3 up-regulation of Aurora-A.
We and others previously demonstrated amplification and overexpression of Aurora-A in ovarian cancers (9,26). However, the frequency of elevated Aurora-A protein and/or mRNA is much higher than its change at DNA level (e.g. ϳl5%), suggesting the mechanism of activating transcription and/or translation of Aurora-A in ovarian cancer cells. Although overexpression of E2F3 has been detected in different tumors (27)(28)(29), the E2F3 status in ovarian cancer has not been well documented. Thus, we reasoned that E2F3 could be elevated in ovarian cancer and might be a causal factor of up-regulation of Aurora-A. To this end, we examined protein levels of E2F3 and Aurora-A in human ovarian cancers. Immunoblotting analyses were performed in a total of eight ovarian cancer cell lines and 72 microdissected ovarian tumor specimens (Fig.  7, A and B). Elevated levels of E2F3 and Aurora-A were detected in 43 of 72 (60%) and 41 of 72 (57%) primary tumors, respectively, as well as the majority of ovarian cancer cell lines examined. Notably, 78% (32/41) of tumors with high levels of Aurora-A overexpress E2F3 (Fig. 7C). Co-up-regulation of E2F3 and Aurora-A seems to be predominantly observed in serous ovarian carcinomas and clear cell tumors, whereas the number of cases is relatively small (Fig. 7C). These data indicate that a large subset of ovarian tumors with elevated Aurora-A might result from expression of high level of E2F3 and further support the findings of biochemical and functional links between E2F3 and Aurora-A.
E2F3 Regulates G 2 /M Cell Cycle Progression through Aurora-A in Ovarian Cancer Cells-Having shown close correlation between expression of E2F3 and Aurora-A in ovarian cancer, we further examined whether E2F3 regulates cell cycle through Aurora-A in ovarian cancer cells. Because expression levels of E2F3 and Aurora-A are high in OVCAR3 and low in A2780S cells (Fig. 6A), we stably knocked down and expressed E2F3 in OVCAR3 and A2780S cells, respectively (Figs. 8A and 9A). Cells treated with vector alone were used as controls. Aurora-A expression levels were decreased and elevated upon knockdown and expression of E2F3, respectively (Figs. 8A and 9A).
After double-thymidine block, flow cytometry analysis revealed that knockdown of E2F3 delays G 2 /M phase in OVCAR3 (Fig. 8, C and D), whereas ectopic overexpression of E2F3 in A2780S cells accelerates G 2 /M progression (Fig. 9, B and C), further supporting the findings in HeLa cells (Fig. 3A).
We next determined whether Aurora-A is required for the E2F3 action in cell cycle of ovarian cancer cells. E2F3 knockdown OVCAR3 cells were transfected with Aurora-A (Fig. 8B). As shown in Fig. 8 (C  and D), expression of Aurora-A largely rescues G 2 /M delay induced by knockdown of E2F3. Moreover, knockdown of Aurora-A in E2F3transfected A2780S cells largely abrogated E2F3-promoted G 2 /M progression (Fig. 9C). These results provide additional evidence that Aurora-A is a direct target of E2F3 and mediates E2F3 function in G 2 /M progression.

DISCUSSION
In this report, we investigated the transcriptional regulation of Aurora-A by E2F3. Ectopic expression of E2F3 induced mRNA and protein levels of Aurora-A, whereas knockdown of E2F3 decreased Aurora-A expression and resulted in mitotic cell cycle delay, which resembles the mitotic arrest phenotypes in Aurora-A siRNA-treated cells (4). Notably, chromatin immunoprecipitation revealed that E2F3 bound to Aurora-A promoter in vivo, and the interaction is cell cycle-dependent, i.e. it primarily occurred during G 2 /M phase. Moreover, co-overexpression of Aurora-A and E2F3 was frequently detected in ovarian cancer. These findings are important for several reasons. First, they provide a mechanistic understanding of transcriptional regulation of the Aurora-A expression during the cell cycle. Second, a direct link between Aurora-A and E2F3 has now been established. Finally, this is the first demonstration of co-alteration of E2F3 and Aurora-A in human ovarian cancer, and elevated E2F3 could be a causal factor for deregulation of Aurora-A in this disease.
In mammalian cells, the E2F family is composed of 10 distinct gene products encoded by eight independent loci that can be divided into three subfamilies based on their sequence  homology: the E2F1-3 genes, the E2F4 and 5 genes, and the E2F6 -8 genes. The E2F1-3 genes have been shown to be tightly regulated during the cell cycle, whereas the E2F4 -8 genes are constitutively expressed. Functionally, E2F1-3 act as positive regulators of transcription whose accumulation is tightly regulated and in most cell types correlates with increased cell proliferation (30 -32), whereas E2F4 and E2F5, when bound to p130 or Rb, act as transcriptional repressors (33). The E2F6 -8 proteins appear to function as Rb-independent transcriptional repressors (34,35). In addition, the E2F3 locus expresses two distinct transcripts, full-length E2F3a and N-terminally truncated E2F3b transcribed from an intronic promoter within the E2F3 locus (36). E2F3a expression is cell cycle-regulated, whereas E2F3b is expressed equivalently in quiescent and proliferating cells and may have an opposing role to E2F3a in cell cycle control. Accumulated evidence shows that E2F3 (e.g. E2F3a) regulates S and G 2 /M cell cycle progression (15,16). Gene expression microarray analyses revealed that E2F3 regulates many of the DNA replication, mitotic, and cell cycle regulatory genes (15,21). A previous report has shown that E2F3 regulates cyclin B1, cyclin A2, and cdc2 transcription (18). We demonstrated, in the present study, that E2F3 directly binds to Aurora-A promoter and tightly regulates Aurora-A expression during the G 2 /M phase.
Previous studies have shown that Aurora-A is transcriptionally regulated by E4TF1, a member of the Ets family, and GABP, the Ets-related transcription factor GABP (37,38). E4TF1 and GABP bind to the same DNA-binding motif (CTTCCGG; Ϫ85 to Ϫ79) of the human Aurora-A promoter to induce Aurora-A promoter activity and transcription. The transactivation of Aurora-A by GABP is regulated through interaction with TRAP220/MED1, an evolutionarily conserved multisubunit coactivator that plays a central role in regulating transcription from protein-encoding genes (37,38). Tanaka et al. (37) cloned human Aurora-A promoter and identified two E2F-binding elements that correspond to binding sites A and B in Fig. 4C. The findings presented here show that although E2F3 binds to these two  sites, their mutations still respond to E2F3 (Fig. 5A). This led us to identify two more E2F-binding motifs proximate to transcriptional start site (Fig. 4C). ChIP and reporter assays show that they are required for the transcriptional activation of Aurora-A gene by E2F3 (Fig. 5, B and C).
Previous reports have demonstrated that E2F3 is frequently overexpressed in a variety of types of human malignancy, and alteration of E2F3 is associated with late stage and high grade tumors. However, alterations of E2F3 in human ovarian cancer have not been well documented, whereas a gene expression array study shows up-regulation of E2F3 in ovarian tumor (39). Previously, we have shown frequent overexpression of Aurora-A in ovarian cancer. In the present study, we demonstrated frequent co-existence of elevated levels of E2F3 and Aurora-A in human primary ovarian carcinoma. Strong association of elevated E2F3 expression and Aurora-A in ovarian tumors underscored the clinical significance of the E2F3-Aurora-A signaling axis. It is very likely that elevated E2F3 is one of the major transcriptional factors that contribute to up-regulation of Aurora-A in other human primary tumors. As described above, E4TF1 and GABP are known transcriptional regulators of Aurora-A (37, 38); however, their roles in Aurora-A induction in human cancer are unclear.
The importance of Aurora-A and E2F3 in oncogenesis has been well established by their alterations in human neoplasms and their capacity to induce cell transformation (27)(28)(29)40). Based on our current observations, we believe that Aurora-A, as a mitotic E2F3 target gene, could mediate E2F3 oncogenic function. Pharmacological agents that inhibit E2F3 or its downstream mediator Aurora-A may lead to inhibition of tumor development, underscoring the significance of the E2F3-Aurora-A axis as an attractive target for cancer therapy.