Geminin is targeted for repression by the retinoblastoma tumor suppressor pathway through intragenic E2F sites.

The geminin protein is a critical regulator of DNA replication. It functions to control replication fidelity by blocking the assembly of prereplication complexes in the S and G(2) phases of the cell cycle. Geminin protein levels, which are low in G(0)/G(1) and increase at the G(1)/S transition, are controlled through coordinate transcriptional and proteolytic regulation. Here we show that geminin is regulated transcriptionally by the retinoblastoma tumor suppressor (RB)/E2F pathway. Initially, we observed that the activation of RB led to the repression of geminin transcription. Conversely, Rb-null mouse embryonic fibroblasts have enhanced the expression of geminin relative to wild type mouse embryonic fibroblasts. Similarly, an acute loss of Rb in mouse adult fibroblasts deregulated geminin RNA and protein levels. To delineate the responsible regulatory motifs, luciferase reporter constructs containing fragments of the geminin promoter were generated. An analysis of the critical regulatory cis-acting elements in the geminin promoter indicated that intragenic E2F sites down-stream of the first exon were responsible for RB-mediated repression of geminin. The direct analysis of the endogenous geminin promoter revealed that these intragenic E2F sites are occupied by E2F proteins, and the mutation of these sites eliminates responsiveness to RB. Together, these data link the expression of geminin to the RB/E2F pathway and represent the first promoter analysis of this important regulator of DNA replication.

The geminin protein is a critical regulator of DNA replication. It functions to control replication fidelity by blocking the assembly of prereplication complexes in the S and G 2 phases of the cell cycle. Geminin protein levels, which are low in G 0 /G 1 and increase at the G 1 /S transition, are controlled through coordinate transcriptional and proteolytic regulation. Here we show that geminin is regulated transcriptionally by the retinoblastoma tumor suppressor (RB)/E2F pathway. Initially, we observed that the activation of RB led to the repression of geminin transcription. Conversely, Rb-null mouse embryonic fibroblasts have enhanced the expression of geminin relative to wild type mouse embryonic fibroblasts. Similarly, an acute loss of Rb in mouse adult fibroblasts deregulated geminin RNA and protein levels. To delineate the responsible regulatory motifs, luciferase reporter constructs containing fragments of the geminin promoter were generated. An analysis of the critical regulatory cis-acting elements in the geminin promoter indicated that intragenic E2F sites downstream of the first exon were responsible for RB-mediated repression of geminin. The direct analysis of the endogenous geminin promoter revealed that these intragenic E2F sites are occupied by E2F proteins, and the mutation of these sites eliminates responsiveness to RB. Together, these data link the expression of geminin to the RB/E2F pathway and represent the first promoter analysis of this important regulator of DNA replication.
Geminin was identified originally as a protein that is degraded by mitotic Xenopus egg extracts, but not by interphase extracts (1), and concurrently during a screen to identify proteins that affect Xenopus development (2). Geminin is a small (25 kDa) protein expressed during the S and G 2 phases of the cell cycle but degraded in M phase at the metaphase/anaphase transition via the anaphase-promoting complex-mediated ubiquitination (3,4). Functional analyses demonstrated that geminin acts to prevent the relicensing of replication origins after they have fired once. This is accomplished by binding to CDT1 (5), a requisite factor for loading MCMs 1 into the prereplication complex. Beginning at the G 1 /S transition, geminin protein levels accumulate and become sufficient to inhibit CDT1 activity. As soon as the origins of replication have fired once in the S phase, this inhibition of CDT1 prevents the reloading of MCMs onto chromatin until the completion of mitosis when geminin is degraded. Although the levels of geminin mRNA have been shown to increase 2-3-fold at the G 1 /S transition (6), the mechanism of this transcriptional regulation remains to be elucidated.
RB has several emerging roles in the control of diverse processes outside of the G 1 /S transition. These include DNA repair, cell death, and DNA replication (7)(8)(9)(10). In the case of DNA replication, the activation of the RB pathway results in the repression of numerous target genes. Classically, this repression is achieved through assembling repressor complexes at the E2F family of transcription factor binding sites (11)(12)(13). For example, the known targets of RB-mediated repression include several genes that encode components of the MCM complex, which are required for the initiation of DNA synthesis (14 -16). DNA polymerase ␣ and PCNA, both required for DNA replication to proceed, also are repressed via the activation of RB (14 -16). As the noted replication factors all stimulate DNA replication, their negative regulation is consistent with the action of RB in the inhibition of the S phase. However, in a recent microarray screen, we identified geminin as a putative target for RB-mediated repression. This finding is difficult to reconcile with the classical cell cycle inhibitory role of RB. If geminin in fact is repressed via RB, this would represent a novel function because it would indicate that RB plays an important role not only in inhibiting replication but also in enabling replication to commence. In this study, we observed the repression of endogenous geminin by active RB as well as deregulation of geminin expression with a loss of RB signaling. A comparison of the geminin genes in mouse and rat revealed conserved E2F sites. We cloned the geminin promoter and identified critical regulatory elements. An analysis of the cisacting elements by reporter assays showed that the geminin promoter is repressed via active RB functioning through intragenic E2F sites.

EXPERIMENTAL PROCEDURES
Cell Lines and Culture-The Rat-16, A2-4, and A5-1 cell lines were derived and cultured as described previously (17). A2-4 and A5-1 express a phosphorylation site mutant (PSM) of the RB large pocket domain, PSM-RB, in a tetracycline-off-inducible manner. PSM-RB is unable to be phosphorylated by CDK/cyclin activity and is active constitutively. Twenty-four hours prior to harvest, these cultures were washed extensively with phosphate-buffered saline to remove doxycycline (dox) from the medium. The S phase synchronization was performed by culture in aphidicolin as described previously (17). The Rb exon 3 loxP/loxP mice were obtained from Dr. Tyler Jacks and have been described previously (18). Mouse adult fibroblasts isolated were cultured in Dulbecco's modified Eagle's medium supplemented with 15% * This work was supported by NCI, National Institutes of Health (to E. S. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Adenoviral Infections-Cells were infected with p16INK4A-or GFPencoding adenovirus as described previously (16). Adenovirus encoding cre-recombinase was obtained from the University of Iowa Vector Core.
Microarray Analysis-Total RNA harvested from cell cultures was utilized for microarray studies as described previously using Affymetrix Genechips U34A, B, and C (rat cell lines) or U74Av2 (mouse primary cells). RNA was harvested from Rb exon 3 loxP/loxP mouse adult fibroblasts using TRIzol (Invitrogen) at 72 and 144 h post-infection with an adenovirus-encoding cre-recombinase. RNA was harvested from passage 3 MEFs using TRIzol and following the manufacturer's suggested protocols. RNA was isolated from Rb ϩ/ϩ and Rb Ϫ/Ϫ MEFs after 24 h in the presence or absence of dox using TRIzol. Hybridizations were carried out up to three times for each condition and each cell line. The results from each chip were analyzed using Microarray Suite (Affymetrix) and GeneSpring (Silicon Genetics). Prior to the analysis, samples were verified for the induction of active RB (in the rat cell lines) or knock-out of Rb (in the Rb exon 3 loxP/loxP mouse fibroblasts) and RNA quality was evaluated by formaldehyde-agarose gel electrophoresis.
Immunoblotting-Immunoblotting was carried out using standard biochemical techniques. Equal total protein from cell lysates was loaded in each lane. Loading was verified by comparing to napthal black staining and CDK4 protein levels. Antibodies were from Santa Cruz Biotechnology against geminin (FL-209) and CDK4 (H-22). The RB antibody was a gift from Dr. Jean Wang.
RT-PCR-RNA was harvested from A5-1 cell cultures using TRIzol according to the manufacturer's suggested protocols. Superscript RT (Invitrogen) and 1 g of total RNA were used to generate cDNAs using random hexamer primers. PCR was carried out using the following primers: 5Ј-CCA TCG GAA GAG GAA GAC AC-3Ј and 5Ј-GAA GTG GCT GAG CAC GTA CA-3Ј (for rat Gmnn) and 5Ј-GGT CAT CAA TGG GAA ACC CAT CAC-3Ј and 5Ј-AGT ACT GGT GTC AGG TAC GGT AGT-3Ј (for glyceraldheyde-3-phosphate dehydrogenase. Analysis of Promoter Regions and Identification of E2F Sites-The genomic sequences for geminin were obtained from the NCBI data base. The sequences flanking NM_020567 in mouse, XM_214477 in rat, and NM_015895 in human were used for these studies. TraFaC (transcription factor binding site comparison) (19) was used to identify the transcriptional start sites, intron/exon boundaries, and E2F binding sites. Binding sites were determined based on the TRANSFAC (20) data base, version 6.0. Position weight matrices M00050 and M00024 were used, but only M00050 sites were found to be conserved in the regions we examined. 91 transcription factor binding sites were found to be conserved in parallel between mouse and rat within this region; however, the non-E2F binding sequences are not shown here for clarity. TraFaC also was used to compare and align the E2F promoter sequences into the image shown in Fig. 4D.
Cloning of the Geminin Promoter and Construction of Luciferase Reporter Constructs-Mouse genomic DNA was isolated from adult fibroblasts by phenol-chloroform extraction. The following primers were used to generate constructs containing pieces of the geminin promoter region: mGmnn (Ϫ520 to 44), 5Ј-AAA ATT CCT CAG CTT CAA CTT TTT-3Ј and 5Ј-AGA CAG CAA TGC TCA GTT CC-3Ј; mGmnn (1160 to FIG. 1. Geminin is repressed transcriptionally by active RB. A, rat-16, A2-4, and A5-1 cells were grown in the presence or absence of doxycycline for 24 h. Cell lysates were immunoblotted for RB and CDK4 as a loading control. The expression of PSM-RB is induced strongly in the absence of dox. B, Affymetrix GeneChips U34A, B, and C were probed using cRNA generated from the cell lines described previously. Rat-16, A2-4, and A5-1 were grown in the presence or absence of doxycycline for 24 h prior to analysis. Geminin was repressed strongly following the induction of PSM-RB. Error bars indicate the mean Ϯ S.E. between samples. C, RNA was harvested from A5-1 cells at time points between 0 and 24 h after the removal of doxycycline from the medium. RT-PCR was performed as described under "Experimental Procedures." C, cell lysates were immunoblotted for geminin and CDK4 as a loading control in Rat-16 cells ϩ or Ϫ doxycycline, A5-1 cells ϩ or Ϫ doxycycline, and A5-1 cells ϩ doxycycline infected with GFP-or p16INK4A-encoding adenovirus for 24 h. TA cloning for each fragment was carried out using the Invitrogen TA cloning kit and the manufacturer's suggested protocols. Promoter fragments then were subcloned into the KpnI and XhoI sites of pGL2-Basic (Promega), such that the cloned promoter fragment drives the expression of the firefly luciferase gene. All of the constructs were verified by sequencing.
Mutagenesis of E2F sites was carried out using the QuikChange site-directed mutagenesis kit (Stratagene) and the manufacturer's suggested protocol. The primers employed were as follows: 5Ј-GGT GAG GTC GGT TTT GGa tCG CTC TGG TGT TCG C-3Ј and 5Ј-GCG AAC ACC AGA GCG atC CAA AAC CGA CCT CAC C-3Ј; 5Ј-GTT TTG TTT TGT TTT TGG atA TCG GGT TTT ATT AAC TTG-3Ј and 5Ј-CAA GTT AAT AAA ACC CGA Tat CCA AAA ACA AAA CAA AAC-3Ј; and 5Ј-CGA CAT CTT GGT TGG Aat CAA AAG ATC TTG GGT TTT G-3Ј and 5Ј-CAA AAC CCA AGA TCT TTT Gat TCC AAC CAA GAT GTC G-3Ј. All of the mutations were confirmed by sequencing.
Transfections were carried out using calcium phosphate. Doxycycline and calcium phosphate were washed from the medium 16 h after transfection, and the cells then were allowed to grow in the presence or absence of doxycycline for 24 h before harvesting. Reporter vectors and cytomegalovirus-␤-galactosidase were cotransfected into A5-1 cells and then grown in the presence or absence of doxycycline for 24 h. To look at the activation of the reporter constructs during cell cycle progression (Fig. 5E), transfections were carried out in Rat1 cells. Luciferase and ␤-galactosidase activity were measured for each sample, and luciferase activity was normalized to ␤-galactosidase levels to control for transfection efficiency. Normalized activity for each reporter was set to 1 for the ϩdoxycycline condition, and the Ϫdoxycycline condition is relative to this condition. Error bars represent the mean Ϯ S.E. between the samples. Results are representative of three independent experiments. Flow cytometry (fluorescence-activated cell sorting) was carried out as described previously (21).
Chromatin Immunoprecipitation Assay-Chromatin immunoprecipitation assays were performed as described previously (22). A5-1 cells were cultured in 15-cm cultures plates in the presence of dox. Chromatin isolated from ϳ10 7 cells was incubated with 1 g of E2F4 (sc-1082X, Santa Cruz Biotechnology) and Dbf-4 (sc-11354, Santa Cruz Biotechnology) antibodies. DNA was purified using QIAquick PCR purification kit following the manufacturer's protocol (Qiagen). Promoter regions of geminin then were amplified using the following primer pairs: rat Gmnn

Geminin Is Repressed Transcriptionally by RB Activation-We have developed cells lines that inducibly express specific
Rb alleles under the regulation of tetracycline (17). As shown in Fig. 1A, the A5-1 and A2-4 cell lines express PSM-RB when the tetracycline analog dox is removed from the medium. This allele is constitutively active and is capable of inducing cell cycle arrest and transcriptional repression. Microarray analyses showed that geminin was a target of RB-mediated RNA attenuation (Fig. 1B). Similar repression was observed when the RB pathway was activated through the expression of p16INK4A, which leads to the dephosphorylation/activation of the endogenous RB (data not shown).
To verify the results of the microarray, we used the RT-PCR analysis. Specifically, A5-1 cells were grown in the presence or absence of dox for up to 24 h. Recovered RNAs were reversetranscribed and amplified for geminin or glyceraldheyde-3phosphate dehydrogenase as a control. Under these conditions, the induction of PSM-RB lead to the attenuation of geminin RNA levels with rapid kinetics (Fig. 1C). If the effect of RB on geminin is of functional significance, the protein levels also should be attenuated. By immunoblot analysis, no change in the geminin expression was observed in the parental Rat-16 cells (Fig. 1D, lanes 1 and 2) but geminin protein levels were attenuated in the presence of PSM-RB in A5-1 cells (Fig. 1D,  lanes 3 and 4). Similarly, the ectopic expression of p16INK4A resulted in significant attenuation of geminin protein levels (Fig. 1D, lanes 5 and 6). Thus, the activation of the RB pathway either through constitutively active alleles of RB or the action of p16INK4A resulted in the attenuation of geminin RNA and protein levels.
In principal, the attenuation of geminin could be attributed to indirect effects of RB on cell cycle position. To eliminate changes in cell cycle position, A5-1 cells were synchronized and held in the S phase with the DNA-polymerase inhibitor aphidicolin. Microarray analysis and immunoblotting then were repeated. Under these conditions, we observed the attenuation of geminin RNA ( Fig. 2A). Additionally, geminin protein levels also were attenuated (Fig. 2B). Therefore, the influence of RB upon geminin is not through indirect changes in cell cycle distribution.
Geminin Is Deregulated upon Loss of RB-Because we observed that geminin is repressed by RB, it would be expected that the loss of RB would lead to the deregulation of geminin expression. To carry out this analysis, we initially utilized MEFs harboring wild type or homozygously inactivated Rb. By microarray analyses, we observed a rather modest increase in geminin RNA levels in those cells deficient in RB (Fig. 3A, 1.6-fold change; p ϭ 0.01). However, protein levels were significantly higher in the Rb Ϫ/Ϫ MEFs relative to their wild type counterparts (Fig. 3C). Recently, it has become clear that compensation can develop in cells lacking RB (18); therefore, we exploited a model to acutely knock out Rb. Murine adult fibroblasts were isolated from mice with Rb exon 3 flanked by loxP sites. In culture, Rb can be ablated efficiently from these cells through the use of cre-recombinase delivered by recombinant adenoviruses. When compared with cells infected with GFPencoding adenovirus as a control, we observed that geminin RNA levels were induced in concert with the knock-out of Rb (Fig. 3B). Immunoblotting in this system also revealed elevated geminin protein levels following cre-infection (Fig. 3D). Thus, RB is required to maintain appropriate levels of geminin.
The Geminin Promoter Contains Several Conserved E2F Sites-As RB is known to function as a transcriptional repres- FIG. 4. Canonical promoter region of geminin is unaffected by RB activation. A, the genomic region of murine geminin from 520 bp upstream of the transcriptional start site to 44 bp downstream was cloned into the pGL2-Basic reporter vector and transfected into A5-1 cells. B, cells were cultured for 24 h with or without dox after transfection, and luciferase reporter assays were carried out as described under "Experimental Procedures." C, gene sequences for mouse and rat geminin were obtained from the NCBI data base along with the sequences 10 kb upstream and downstream. These were aligned and compared using the TraFaC utility. No conserved E2F sites were found Ͼ1 kb upstream of the transcriptional start site. E2F sites are indicated on each sequence, and homology between species is indicated by connections between the E2F sites. The locations of E2F sites are shown for the mouse and rat promoters in relation to the transcriptional start site (base pair number 1). D, E2F binding sequences and comparison to the E2F consensus sequence. The site at 1291 bp downstream of the transcriptional start site is in a reverse orientation. sor, we sought to determine how geminin RNA levels are modulated by RB. RB is known to exert transcriptional repression via E2F DNA binding elements. Initially, we focused on the canonical promoter region upstream of the transcriptional start site (1,2). Specifically, we cloned 520 base pairs upstream of the transcriptional start site to 44 base pairs downstream to generate a luciferase reporter construct. This fragment contained a putative E2F site at Ϫ306 bp upstream of the tran-FIG. 5. The first intron of geminin is critical for regulation by the RB pathway. Reporter constructs were generated by PCR amplification of mouse genomic DNA and cloning into the pGL2-Basic luciferase reporter vector. Exon 1 (clear boxes) is indicated along with the transcriptional start site (forward arrow), and the numbering of E2F sites (darkened boxes) indicates the distance from the transcriptional start site in base pairs. The beginning of coding sequence is indicated by the shaded box. A, reporter construct mGmnn (Ϫ521 to 1696) including the region cloned previously (Fig. 4) and the entire first intron. B, luciferase reporter activity for mGmnn (Ϫ521 to 1696). C, reporter constructs mGmnn (ϩ1160 to 1696, ϩ871 to 1696, and ϩ181 to 1696) containing deletions of the first intron. Construct mGmnn ⌬E2F contains a mutation in each of three E2F sites. D, reporter activity for these clones in the A5-1 cell line. E, rat1 cells were transfected with pGL2-mGmnn (ϩ181 to 1696) and E2F2 expression plasmids. Activity is relative to the reporter without E2F2 cotransfection. Error bars represent the mean Ϯ S.E. between samples. F, Rat1 cells were transfected with pGL2-mGmnn (ϩ181 to 1696) and held in low serum for 96 h. Cells were returned to 10% serum for 0 -18 h and harvested for reporter assays or fluorescence-activated cell sorter assays. Luciferase activity above the 0-h time point is shown in blue. The percentage of cells in the S phase calculated by flow cytometry is shown in red and represents 10 6 cells analyzed/time point. scriptional start site (Fig. 4A). The reporter was transfected into A5-1 cells that were then grown in the presence or absence of dox for 24 h. Repression by active RB was measured as the change in relative luciferase activity in the absence of doxycycline. Surprisingly, although this construct harbored significant luciferase activity, it was not repressed by active RB (Fig. 4B).
To extend our analysis for E2F-regulatory elements, we used the TraFaC program (trafac.chmcc.org) (19) to search for additional E2F motifs within the murine geminin genomic sequence. This analysis identified four E2F sites that were clustered around the first exon (Fig. 4, C and D). Many other putative binding sites for other transcription factors also were identified (data not shown), because the control of any given promoter is ultimately very complex. Because of our observations that RB is affecting geminin expression and because of the canonical role of E2F in RB-mediated repression, we have focused this analysis on E2F binding sites. TraFaC alignment of the genomic regions ϳ1 kb upstream and downstream of the first exon in mouse, rat, and human showed that the E2F sites identified in mouse corresponded to similar sites in rat (Fig.  4C) and in human (data not shown). A compositional similarity between species for the presence of E2F sites further supports an important role for the RB/E2F pathway in the control of geminin expression. Additionally, this analysis suggested that the intragenic E2F sites downstream of the transcriptional start site may play a significant role in geminin control.
Active RB Represses Geminin through Intragenic E2F Sites-To address the role of all of the identified E2F sites, we cloned a larger region of geminin sequence and generated a reporter construct containing all four of the clustered E2F sites in mouse (Fig. 5A). In the reporter assays, the activity of this reporter was repressed strongly by active RB (Fig. 5B). These results suggested that the intragenic E2F sites are responsible for the observed repression. We then proceeded to generate deletions of this reporter to identify critical E2F sites. Initially, we focused on a specific intragenic E2F site just upstream of the coding region (exon 2) and found that in isolation it exhibited a relatively weak repression (Fig. 5, C and D). Similarly, the two downstream E2F sites in combination were only weakly repressed by PSM-RB. In contrast, the combination of all three E2F sites found in intron 1 resulted in strong repression equivalent to the repression observed with the larger geminin fragment. This reporter was activated strongly by the ectopic expression of E2F2, emphasizing the importance of the intragenic E2F sites in the control of geminin expression (Fig.  5E). These results suggest that the E2F sites downstream of FIG. 6. E2F sites within the first intron of geminin are necessary for full activity and RB-mediated repression of the geminin promoter. A, schematic representation of the murine geminin promoter and the mutations within the pGL2-mGmnn ⌬E2F reporter construct. B, wild type E2F sequences and the mutations introduced by site-directed mutagenesis as outlined under "Experimental Procedures." C, rat1 cells were transfected with pGl2-mGmnn (ϩ181 to 1696), pGL2-mGmnn ⌬E2F, and PSM-RB expression plasmids as indicated. Luciferase activity is relative to the reporters without PSM-RB cotransfection. Error bars represent the mean Ϯ S.E. between samples.
the transcriptional start site are the predominant targets for RB-mediated repression.
To verify that our geminin reporter with intact E2F sites is regulated by endogenous RB during the course of the cell cycle, transfections were carried out in Rat1 fibroblasts. These cells were held for 96 h in serum-deficient (0.01%) medium to synchronize the majority of cells in G 1 . Cells were stimulated with 10% serum for 0 -18 h and harvested for flow cytometry and reporter assays. The strongest activation of the reporter occurred at 12 h after release from G 1 , just preceding the entry of most cells into S phase (Fig. 5F).
The Response of the Geminin Promoter to RB Is Dependent upon E2F-To investigate whether the response of the geminin promoter to RB is dependent on the E2F sites, we use sitedirected mutagenesis to disrupt the three sites as shown in Fig.  6, A and B. Rat1 cells then were transfected with either pGL2-mGmnn (ϩ181 to 1696), pGL2-mGmnn ⌬E2F, PSM-RB, or PSM-RB and E2F2 expression plasmids (Fig. 6C). The wild type reporter was active and repressed by PSM-RB. This repression was alleviated mostly by cotransfection with E2F2, indicating that the observed repression occurs via E2F. The ⌬E2F reporter was not significantly influenced by PSM-RB. These results demonstrate that the repression of the geminin promoter by active RB is mediated through these E2F sites.
The Geminin Promoter Is Occupied by E2F-To determine the ability of E2F proteins to occupy the geminin promoter, we used chromatin immunoprecipitation assay at the E2F sites in the rat promoter (Fig. 7A). The intragenic region of this promoter was repressed by RB in the A5-1 cell line similarly to the mouse promoter (Fig. 7B). The primers were designed to amplify pieces of the promoter containing the different E2F sites or to amplify a region ϳ9 kb upstream of the transcriptional start site, which we determined by sequence comparison to be nonconserved and devoid of transcription factor binding sites (data not shown). As seen in Fig. 7C, E2F4 occupied predomi-nantly E2F sites at ϩ12, ϩ324, and ϩ708 of the rat geminin sequence. These results are consistent with the findings in mouse reporter assays confirming that the intragenic E2F sites are responsible for modulating transcription. DISCUSSION Here we show that geminin is regulated transcriptionally by the RB/E2F pathway. In the presence of a constitutively active RB allele, geminin RNA levels are repressed strongly regardless of cell cycle position. Similar repression is observed when endogenous RB is activated by p16INK4A overexpression. Conversely, in Rb-deficient cells, geminin expression is deregulated. Therefore, geminin RNA and protein levels are regulated via RB. To delineate the mechanisms of regulation, we investigated E2F sites in the geminin promoter region. Typically, RB exerts its influence on gene expression through E2F sites relatively close to the transcriptional start site. Surprisingly, we failed to observe any repression with reporter constructs surrounding this conventional region of regulation. Cross-species analyses identified highly conserved E2F sites in proximity of the first exon of geminin, and reporter analyses confirmed that this region facilitated RB-mediated repression. Inactivation of endogenous RB at the G 1 /S transition resulted in the activation of a reporter construct, and the mutation of the E2F sites in the first intron prevented the repression of the reporter by PSM-RB. Additionally, chromatin immunoprecipitation confirms that E2F proteins are bound to these intragenic regions. These findings demonstrate that intragenic sequences play an important role in the regulation of geminin transcription.
The expression of geminin is controlled by the intricate interplay of transcriptional and proteolytic regulatory mechanisms. Degradation of geminin at the metaphase to anaphase transition has been shown to rely on the anaphase-promoting complex (1, 4). Therefore, the regulation of geminin at the transcriptional level by RB represents an additional control FIG. 7. E2F4 occupies the geminin promoter. A, schematic representation of the rat Gmnn gene and a reporter construct incorporating all of the potential E2F sites between the transcriptional start site and the beginning of the coding region in exon 2. B, a luciferase reporter construct containing the region of the rat Gmnn gene from 159 bp upstream of the transcriptional start site to 1328 bp downstream is repressed transcriptionally by PSM-RB similar to mouse. C, cross-linked rat chromatin was immunoprecipitated with DBF4 (as a negative control) or E2F4 antibodies. Pieces of the rat geminin promoter containing E2F sites or a piece of the promoter 9 kb upstream of exon 1 (a non-conserved region with no nearby E2F consensus sequences) were amplified by PCR using ␣-32 P as described under "Experimental Procedures." PCR products were resolved by PAGE. mechanism. RB is dephosphorylated and activated at the M/G 1 transition when geminin levels decrease, and likewise, RB is phosphorylated and inactivated at the G 1 /S transition when the levels of geminin transcript increase and the protein begins to accumulate. We propose that RB functions in concert with the anaphase-promoting complex to maintain low levels of geminin throughout G 1 . The biological importance of low geminin levels through G 1 has been indicated by several studies. McGarry and Kirschner (1) injected dividing Xenopus embryos with a stable mutant of Xenopus geminin lacking all 9 amino acids of the destruction box (geminin DEL ). Although cell division continued, the resulting daughter cells were anucleate. Quinn et al. (23) identify the Drosophila geminin homolog and observe that overexpression caused the cells of Drosophila embryos to enter mitosis without replicating their DNA, resulting in a metaphase arrest and apoptosis. Numerous studies have shown that exogenous geminin blocks DNA licensing and replication (1,5,(23)(24)(25)(26)(27)(28)(29). Furthermore, a recent study (30) shows that during the Xenopus embryonic cell cycle, 30 -60% of endogenous geminin protein was not degraded at anaphase. The remaining geminin was altered in some way to prevent its activity through G 1 . Transcriptional control by RB could limit de novo synthesis of geminin in G 1 , conceivably preventing the need to alter additional geminin after the transition into G 1 .
RB classically functions as an inhibitor of DNA replication. Consistent with this concept, RB mediates the repression of several ORC and MCM genes that are requisite components of the preRC that initiates DNA replication at origins (14 -16). Additionally, CDC6 and CDT1 involved in assembling the preRC are regulated similarly through the RB/E2F pathway (30,31). Therefore, it was in a way paradoxical that geminin, an inhibitor of preRC assembly and DNA replication, was also under RB control. One reason for RB-mediated regulation of geminin during cell division processes would be to allow reentry into the cell cycle. It has been demonstrated previously (18) that cells rapidly reenter the cell cycle after RB-mediated repression of target genes is relieved. In order for a cell to quickly take advantage of growth-permissive conditions, it is critical that inhibitors of DNA replication not be present. Such a model would argue that there must be kinetic distinctions between the rate of preRC assembly and the accumulation of geminin during progression through G 1 . Geminin levels presumably are checked by the activity of the anaphase-promoting complex, and this may be sufficient to allow preRC assembly as cells transit through G 1 .
The increased levels of geminin observed in Rb-deficient cells could be hypothesized to tightly regulate DNA replication. However, Rb-deficient cells clearly exhibit defects in S phase control. For example, the loss of RB compromises S phase checkpoints, allowing for DNA replication in the presence of damage. Additionally, rather than being resistant to rereplication, Rb-null MEFs and Rb-deficient tumor cell lines readily undergo aberrant rounds of DNA replication resulting in aneuploidy (31,32). Thus, deregulation of geminin expression by loss of RB is insufficient to prevent either normal or aberrant replication. Two possible explanations exist for this observation. First, that geminin, while being expressed copiously, is not active in Rb-deficient cells. This model is supported by the finding that inactive species of geminin, which fail to bind CDT1 or inhibit DNA replication, can be detected (29). Second, the enhanced geminin levels present in Rb-deficient cells are met by correspondingly higher levels of positive regulators of the preRC (i.e. CDT1, cdc6, and MCMs). Under these conditions, geminin could be rendered inert because of an excess of targets. As such, Rb-deficient cells would be prone to rereplication in a manner analogous to geminin-deficient cells.
In summary, we find that RB is a potent regulator of the critical licensing factor geminin. These analyses underscore the complexity of RB signaling both in mechanism of repression (through intragenic E2F sites) and diverse target spectrum (through the regulation of an inhibitor of DNA replication).