The central role of CDE/CHR promoter elements in the regulation of cell cycle-dependent gene transcription

Authors


K. Engeland, Molecular Oncology, University of Leipzig, Semmelweisstr. 14, D-04103 Leipzig, Germany
Fax: +49 341 9723475
Tel.: +49 341 9725900
E-mail: engeland@medizin.uni-leipzig.de

Abstract

The cell cycle-dependent element (CDE) and the cell cycle genes homology region (CHR) control the transcription of genes with maximum expression in G2 phase and in mitosis. Promoters of these genes are repressed by proteins binding to CDE/CHR elements in G0 and G1 phases. Relief from repression begins in S phase and continues into G2 phase and mitosis. Generally, CDE sites are located four nucleotides upstream of CHR elements in TATA-less promoters of genes such as Cdc25C, Cdc2 and cyclin A. However, expression of some other genes, such as human cyclin B1 and cyclin B2, has been shown to be controlled only by a CHR lacking a functional CDE. To date, it is not fully understood which proteins bind to and control CDE/CHR-containing promoters. Recently, components of the DREAM complex were shown to be involved in CDE/CHR-dependent transcriptional regulation. In addition, the expression of genes regulated by CDE/CHR elements is mostly achieved through CCAAT-boxes, which bind heterotrimeric NF-Y proteins as well as the histone acetyltransferase p300. Importantly, many CDE/CHR promoters are downregulated by the tumor suppressor p53. In this review, we define criteria for CDE/CHR-regulated promoters and propose to distinguish two classes of CDE/CHR-regulated genes. The regulation through transcription factors potentially binding to the CDE/CHR is discussed, and recently discovered links to central pathways regulated by E2F, the pRB family and p53 are highlighted.

Abbreviations
CDE

cell cycle-dependent element

cdk

cyclin-dependent kinase

ChIP

chromatin immunoprecipitation

CHR

cell cycle genes homology region

cIAP2

DRS, downstream repression site

EMSA

electrophoretic mobility shift assay

MEF

mouse embryonic fibroblast

SV40

simian virus 40

Introduction

The cell division cycle is a fundamental process. It is regulated at different molecular levels. One central modification controlling the cell cycle is phosphorylation by complexes of cyclin-dependent kinases (cdks) and their corresponding cyclins. A prominent example of such a pair is cyclin B and cyclin-dependent kinase 1 (cdk1/Cdc2) controlling the checkpoint between G2 phase and mitosis (Fig. 1).

Figure 1.

 CDE/CHR-regulated genes controlling G2/M progression. The expression of many central players appearing in G2 phase and mitosis was shown to be regulated at the transcriptional level by CDE/CHR tandem elements. Tightly controlled gene expression, as well as rapid protein degradation, is required for cell cycle progression. Regulatory circuits also include control through p53. Cell cycle arrest can be mediated by p53 downregulating the transcription of central cell cycle regulators such as cyclin B, Cks1, Cdc2 and Cdc25C.

Cyclins were discovered by their cyclic appearance during the cell cycle [1]. In particular, the abrupt disappearance of the proteins was noticed in early reports and described to be regulated by ubiquitin-mediated proteolysis. Much later control of cyclin synthesis was investigated in more detail [2]. In mammals, two B-type cyclins form complexes with cdk1/Cdc2. Synthesis of proteins encoded by cyclin B1 and cyclin B2 genes is mostly regulated at the transcriptional level [3]. We and others then observed that transcription from both cyclin B genes is controlled by combinations of tandem sites called the cell cycle-dependent element (CDE) and the cell cycle genes homology region (CHR) [4–7].

The CDE was first observed in the Cdc25C promoter by in vivo footprinting as being protected in G0 cells. The Cdc25C gene is not expressed in resting cells or in cells in G1 phase. Only in G2 phase can strong transcription of the gene be detected. Mutation of the CDE in the Cdc25C promoter and analysis in reporter assays shows that this element is responsible for cell cycle-dependent expression of the gene. Surprisingly, deregulation of the promoter does not lead to loss of its activity but causes activation in resting cells and in G1 cells. Therefore, transcriptional repression is responsible for regulation through the CDE [8]. Shortly after this initial description, another report confirmed the CDE in the Cdc2 promoter as being differentially occupied by protein complexes during the cell cycle [9].

Mutation of nucleotides close to the CDE in the Cdc25C promoter, and analysis in reporter assays, yielded the first hints that there is another site relevant for cell cycle-dependent repression of this gene. Sequence comparison of the cyclin A and Cdc2 promoters with that of the Cdc25C gene, followed by promoter mutations analysed in reporter assays led us to identify a new type of site downstream from the CDE. Because of the high sequence conservation of this site among the three promoters, we named this type of element the CHR [10]. Transcriptional regulation through this new site appeared to be functionally identical to that of the CDE, with repression in resting cells and relief from downregulation later in the cell cycle. Mutation of the CHR led to derepression of transcription in G0 cells [10].

Genes cannot be regulated solely by repression: the activation of promoters is also required. To this end, CDE/CHR repressor sites are usually found in conjunction with two or three CCAAT-box elements through which NF-Y transcription factors activate the promoters. Activation by NF-Y generally contributes the largest part to promoter activity, which is then repressed through the CDE/CHR sites in the early phases of the cell cycle [11,12].

Other proteins binding to CDE/CHR promoters are E2F family members. It has been shown that CDEs are related to E2F sites and can, at least in some cases, also bind members of the E2F transcription factor family [13]. Since the discovery of the first three genes regulated by CDE/CHR tandem sites, many other important cell cycle-regulator genes have been reported to be controlled by this class of elements.

Promoters regulated by CDE and CHR sites

Genes regulated by CDE and CHR elements in their promoters generally encode proteins with functions in S, G2 or M phases (Table 1). In quiescent cells these genes are not expressed. CDE/CHR promoters usually lack a TATA-box and employ multiple transcriptional start sites [10,14]. Mutation of either a CDE or a CHR in a promoter leads to the activation of transcription in quiescent cells. A narrowly defined sequence consensus for CDEs, from which functional conclusions can be drawn, has not evolved.

Table 1. Class I and class II genes with their functions in the cell cycle.
Gene symbolGene nameFunction
AURKAaurora kinase AProtein kinase, regulates microtubule formation and stabilization at the spindle pole during chromosome segregation
AURKBaurora kinase BProtein kinase, key regulator of cytokinesis, mediates attachement of the mitotic spindle to the centromere, phosphorylates histone H3 during mitosis
B-MYB/MYBL2v-myb myeloblastosis viral oncogene homolog (avian)-like 2Transcription factor, involved in cell cycle progression, possesses both activator and repressor activities
CCNAcyclin ARegulatory subunit of CDC2 or CDK2 kinases, promotes both G1/S and G2/M transitions
CCNB1cyclin B1Regulatory subunit of mitosis promoting factor (MPF), regulates G2/M phase transition, co-localizes with microtubules
CCNB2cyclin B2Regulatory subunit of mitosis promoting factor (MPF), regulates G2/M phase transition, co-localizes with Golgi region
CDC2/CDK1cell division cycle 2/cyclin-dependent kinase 1Serine/threonine kinase, catalytic subunit of the mitosis promoting factor (MPF), controls G1/S and G2/M phase transitions
CDC25Ccell division cycle 25 homolog CTyrosine phosphatase, triggers entry into mitosis
CKS1CDC28 protein kinase regulatory subunit 1Binds to the catalytic subunit of cyclin-dependent kinases (CDK), essential for their biological function
MKLP1/KIF23mitotic kinesin-like protein 1/kinesin family member 23Kinesin-like protein, motor enzyme that moves antiparallel microtubules, localizes to the interzone of mitotic spindles
PLKpolo-like kinase 1Protein kinase, multiple function in cell cycle, activates CDC25, interacts with anaphase-promoting complex (APC)
TOME-1/CDCA3trigger of mitotic entry/cell division cycle associated 3F-box like protein, required for degradation of the CDK1 inhibitory tyrosine kinase WEE1, triggers entry into mitosis

After the initial description of CDE/CHR-dependent gene regulation, many promoters were described as being controlled by CDE or CHR sites [10]. However, one conclusion from these numerous reports is that sequence comparison alone does not suffice for genes to be designated as regulated by CDE and/or CHR elements. We would like to derive, from the many publications, functional requirements, sequence similarities and characteristics of general promoter structure for cell cycle-regulating sites to be regarded as bona fide CDE/CHR elements (Table 2).

Table 2.   Criteria for promoters controlled by CDE/CHR sites.
Class I
 Genes not expressed in G0 and G1 cells
 Genes encode proteins with functions in S, G2 or M phases
 CHR consensus similar to 5′-TTTGAA-3′
 CDE is a site rich in G and C found upstream of a CHR
 CDE positioned with a four-nucleotide spacer upstream of a  CHR
 Orientation with CHR proximal to the coding region
 Only one CDE/CHR per promoter
 TATA-less promoters, multiple transcriptional start sites
 Protein binding to the elements in G0 and G1 cells as monitored  by in vivo footprinting
 Mutation of either CDE or CHR leads to a substantial  deregulation of repression in G0
 Two or three CCAAT-boxes, spaced 31–33 bp apart, which bind  heterotrimeric NF-Y proteins
 NF-Y is the main activator of the genes
Class II
 The same as for class I, but no functional CDE site four  nucleotides upstream of a functional CHR

CDE sites represent special E2F-binding elements and thereby display sequence similarity to these sites. However, a requirement for functional CDEs, distinguishing them from E2F elements, is that they must be positioned with a four-nucleotide spacer upstream of a CHR. Consistent with our original description of the first CDE/CHR promoters [10], the CDE in the human cyclin A promoter was also identified as a variant E2F site [15]. Cell cycle-dependent protection of the CDE in the mouse cyclin A promoter was confirmed by in vivo footprinting and named CCRE [16]. Earlier, the CDE/CHR region from the human Cdc2 gene had been found to be responsible for 12-O-tetradecanoylphorbol-13-acetate (TPA)-dependent transcriptional repression and was termed the R box [17].

The best studied E2F site, located with a four-nucleotide spacer upstream from a CHR, is found in the B-Myb gene. This site was first identified without recognizing the adjacent element comprising CHR function. However, it was observed that the E2F element downregulates B-Myb transcription in G0 and that its mutation leads to derepression because it is observed with CDE sites [18]. Repressive protein complexes appear to occupy the CDE-related E2F site in G0 and G1 cells, as determined by in vivo footprinting. Site occupation during the cell cycle is lost precisely at the time when B-Myb becomes expressed [19]. After the E2F site was well established as regulating B-Myb expression, a CHR-like element, named the downstream repression site (DRS), was identified to regulate cell cycle-dependent transcription together with the E2F site [20,21]. The DRS/CHR in the B-Myb promoter deviates most from other CHR sequences with its two-nucleotide exchange from the CHR consensus (Fig. 2). Changing the distance between E2F and DRS sites in reporter constructs by the insertion of two or four nucleotides leads to derepression in G0 cells in reporter assays [22]. With only one nucleotide exchange compared to the mouse sequence, the E2F and DRS/CHR segment in the human B-Myb promoter is well conserved [23]. Recently, the E2F site of the B-Myb promoter was mutated in mice. Homozygous mutation of the element was found to lead to derepression of the B-Myb promoter in mouse embryonic fibroblasts (MEFs) derived from the animals. Furthermore, elevated expression of B-Myb mRNA, indicating a deregulation, is observed in brain cells carrying the mutant E2F site compared with the wild-type mice [24].

Figure 2.

 Experimentally validated CDE/CHR sites. Two classes of promoters can be distinguished. Class I genes require both sites for cell cycle-dependent repression. In contrast, class II genes do not have a functional CDE and are only regulated through a well-conserved CHR. Interestingly, some ortholog genes from mouse and human, such as cyclin B2 and Cdc25C, can be members of class I or class II depending on the species origin. The tandem element in the mouse B-myb promoter is an E2F site in combination with an element named the ‘downstream repression site’ (DRS or CHR).

In the human Cdc25C gene, CDE and CHR cooperate in cell cycle-dependent repression. They are of similar importance because their mutation leads to a comparable derepression in the cell cycle [10,25]. We designate such genes as class I CDE/CHR genes (Fig. 2). Moreover, orientation of the CDE/CHR in the general context of a promoter appears to be relevant because inversion of the site in the human Cdc25C promoter resulted in a deregulation of cell cycle-dependent transcription. Deregulation is also observed when the CDE alone is inverted [26].

Interestingly, regulation through the CDE/CHR is different with the mouse Cdc25C promoter. The timing of cell cycle-dependent expression from mouse and human promoters is identical. Also, essentially all promoter elements are conserved in the two genes except for the CDE. Mutational analysis of the region four nucleotides upstream from the CHR in mouse Cdc25C promoter-reporter assays leads to only a small deregulation when compared with changes in the CHR [27]. We suggest referring to genes that have a functional CHR but lack a site four nucleotides upstream from the CHR, which, when mutated, does not lead to any or to only a minor deregulation, as class II genes (Fig. 2).

Furthermore, some other properties of CHR elements were shown using the mouse Cdc25C gene as an example. The CHR in this promoter naturally lacking a CDE can cooperate with bona fide CDE, E2F or Sp1/3 sites introduced upstream of it, at least when tested in reporter assays [27].

Many other genes were initially reported to be controlled by both CDE and CHR elements. Examples are present within the cyclin B family. In mammals, three B-type cyclins are known. For the most recently discovered family member, mammalian cyclin B3, the exact function and kinase association partners are not known [28,29]. By contrast, cyclin B1 and cyclin B2 are central to the regulation of progress through the cell cycle (Fig. 1). Cyclins B1 and B2 appear in S phase and accumulate in G2 and mitosis before disappearing at the transition from metaphase to anaphase. Synthesis is controlled at the level of gene transcription [3]. Interest in control mechanisms of cyclin B1 and cyclin B2 cell cycle-dependent transcription began early [3,30–33]. When investigating the regulation of human cyclin B1 transcription, a potential CDE was tested and found to play only a limited role in cell cycle-dependent transcription [4]. Later, this finding on the CDE was confirmed and the major cell cycle-dependent regulation was attributed to a novel type of CHR site just next to the CDE. This CHR holds a change of one nucleotide compared with other elements of this type, which mostly follow the consensus 5′-TTTGAA-3′ [5]. As the putative CDE has, in contrast to the CHR, only a modest impact on cell cycle-dependent transcription, the human cyclin B1 gene is class II (Fig. 2).

Analysis of cyclin B2 cell cycle-dependent transcription offers some insights into the variability of CDEs regarding sequence and function. Initially, mouse cyclin B2 expression was shown to be regulated by a CDE/CHR tandem site and was therefore considered to be a class I promoter; however, the CDE in this promoter leads to a smaller deregulation than the CHR when mutated [6]. By contrast, the human cyclin B2 promoter does not require a CDE for cell cycle-dependent transcription. Mutation of the site in the human promoter that is equivalent to the CDE from the mouse cyclin B2 promoter does not result in a deregulation [7]. Therefore, human cyclin B2 is clearly a class II gene (Fig. 2). In a comparison of nucleotide sequences from both promoters, nine homologous regions stand out. Only one of them, the CDE in the mouse cyclin B2 promoter, is not perfectly conserved. A one-nucleotide change is found in the human promoter. The alteration appears to be sufficient to render the human cyclin B2 promoter resistant to deregulation through mutation of this region. Nevertheless, changes in the CHR lead to a complete deregulation of expression from the human cyclin B2 promoter [7,25]. It remains unclear why one CHR requires a CDE four nucleotides upstream, whereas another CHR, particularly in a very similar context as exemplified in the cyclin B2 promoters, can function without a CDE.

Such differences in sequence with identities in function are often found between mouse and human promoters. However, regulation through CDE/CHR sites found in the human Cdc25C, cyclin B1 and cyclin B2 promoters is fully conserved in nucleotide sequence and function in closely related organisms such as chimpanzee, orangutan and human [25].

Timing of gene expression during the cell cycle has been believed to be dependent on the exact nucleotide sequence of the CDE/CHR site. Expression from a cyclin A reporter usually precedes that of cyclin B2, as expected from the chromosomal expression [6]. In order to test whether this solely depends on the CDE/CHR, the CHR region and the element upstream from it were replaced in the human cyclin B2 promoter with the well-characterized CDE/CHR sites from the human Cdc25C and cyclin A promoters [10] and expression from the altered reporters was tested during the cell cycle compared with expression from the wild-type construct [7]. Timing of expression from the three promoters was similar, without a significant shift between cell cycle phases. This indicates that a promoter does not simply adopt the timing of expression from the other promoter as a result of replacing the cell cycle-regulatory elements. Thus, it is likely that cell cycle-dependent timing of expression is also determined by elements outside the CDE/CHR elements [7].

Furthermore, the effect of DNA methylation on CDE/CHR-dependent transcriptional regulation, possibly through mediating protein binding to the elements, was investigated. The CpG sites of the CDE in the cyclin B2 promoter were found to be partially methylated. However, quantitative methylation analysis did not show any alterations during the cell cycle, making it unlikely that protein binding to the CDE/CHR is affected by change in DNA methylation during different phases of cell division [34].

Another cell cycle gene, also relevant for mitosis, codes for the serine/threonine-specific Polo-like kinase 1 protein. A mutation in the CHR deregulates cell cycle-dependent transcription from the Plk1 promoter. Changing a putative CDE four nucleotides upstream from the CHR had almost no effect on the cell cycle-dependent regulation of the promoter [35]. Therefore, Plk1 was considered to be a class II gene (Fig. 2). Also, Cks1, a member of the cyclin-dependent kinase subunit family, reaches peak expression in S/G2 phases of the cell cycle. This expression pattern is dependent on transcriptional repression through both a CDE and a CHR in the Cks1 promoter [36].

Moreover, Tome-1 was reported as a CDE/CHR gene. Tome-1 mediates destruction of the mitosis-inhibitory kinase Wee1 via the E3 ligase SCF and becomes maximally expressed in G2 (Fig. 1). Human and mouse Tome-1 promoters were tested by mutating putative CDE and CHR sites separately in promoter assays. Both sites are required for cell cycle-dependent transcription. However, as in most other CDE/CHR promoters, mutation of the CHR results in a smaller remaining cell cycle-regulation than alteration of the CDE [37]. Interestingly, the core of the human Tome-1 promoter CDE/CHR has a sequence identical to the tandem element in the human Cdc25C promoter [10].

Recently, the mitosis-related genes Ect2, MgcRacGAP and MKLP1 were shown to be transcriptionally regulated during the cell cycle, being weakly expressed in G1 and strongly expressed in G2/M. Promoters became derepressed in the cell cycle when the CHRs were mutated and assayed in the interleukin-2-dependent Kit 225 T cells. Also, the interleukin-2-dependent derepression, usually seen in this system, was derepressed upon CHR mutation. The effects were very strong with the MKLP1 promoter. The MgcRacGAP CHR has the sequence ‘5-TTTCAA-3′ and thereby a reverse orientation to canonical CHRs. This may explain why the effect in this promoter is particularly small [38]. All three CHRs may be class II, although regions upstream from them were not tested for functional E2F or CDE sites (Fig. 2).

The gene for Aurora A, a serine-threonine kinase whose expression peaks in G2/M, was found to be regulated by a CDE/CHR site. The CHR has the unusual sequence 5′-CTTAAA-3′. In order to yield a high sequence similarity for the CDE with sites published at the time, the CDE and CHR were postulated to be located next to each other without a spacer [39]. However, considering the great variability in CDE nucleotide sequences, and the fact that functional assays with just one mutant promoter could not pinpoint this site exactly, we suggest that the site shifted upstream by five nucleotides is the CDE. The functional data would allow such a change in interpretation. The shift in exact position of the CDE would yield a spacer essential to define these elements as a CDE/CHR tandem site [10]. One other member of the Aurora kinase family was also found to be regulated by a typical CDE/CHR site. The Aurora B promoter is controlled by CDE and CHR sites separated by four nucleotides. As in many similar promoters, mutation of the CHR leads to a more pronounced deregulation than the alteration of the CDE [40].

Another gene tested for its cell cycle regulation is the cellular inhibitor of apoptosis protein 2 gene (cIAP2). It is induced by nuclear factor-κB and was shown to be expressed in a cell cycle-dependent manner, with low expression in G1 and reaching peak levels in G2/M. Mutational analysis of the promoter in HeLa cells synchronized by double-thymidine or nocodazole block showed that a CHR is responsible for the cell cycle-dependent expression. The sequence upstream of the CHR does not match any of the published CDEs. However, alteration of a putative CDE, which is more distant from the CHR than the usual four nucleotides, in addition to the CHR, yielded a further decrease in regulation. The CDE alone was not tested [41]. The CDE mutation that was assayed would also alter a putative CDE site with the standard distance of four nucleotides to the CHR. With the data presented it is not quite clear where exactly the CDE is located and what its contribution to cell cycle-dependent regulation is. Possibly the CHR constitutes a class II regulatory site.

Over the years numerous additional genes were reported to be regulated by CDE/CHR sites. Often sites were postulated only based on sequence similarity. Generally, functional assays are required to define relevant elements. Sometimes reported experiments do not yield a consistent picture. Survivin, also named Birc5, API4 or IAP4, functions as an apoptosis inhibitor and is expressed in G2/M. In an initial study, cell cycle-dependent regulation of about three-fold had been described for G2/M expression of the wild-type promoter-reporter compared with the expression level in G1. One site designated a CDE led to a partial deregulation when mutated. However, a putative CHR led to a deregulation upon mutation, although it is not part of a CDE/CHR tandem site. Furthermore, one experiment suggested a strong deregulation when an upstream Sp1 site is mutated [42]. Objections to most of these results were raised by a later study. A stronger cell cycle-dependent regulation of the wild-type construct was observed than in the first study. Although numerous CDEs were also postulated, the experiments finally yielded only one functional CDE close to the CHR [43]. In this report an alignment of CDE/CHR sites is displayed in which the CDE is moved downstream by two nucleotides to yield a better consensus with other CDEs. However, just one mutation, with a single nucleotide change, was analyzed. The mutant does not dictate such a shift [43]. The picture becomes even more complicated when considering another report on the human survivin gene. The article tries to correlate mutations or polymorphisms found in the survivin promoter to regulation through several possible CDE and CHR sites. When mutated the sites led only to a moderate deregulation of cell cycle-dependent transcription of the reporter. According to the results from this report, possible protein binding to the putative CDE appears stronger in G2/M than in G1 [44]. This contradicts repression through a complex in G1 and in vivo footprinting results in the original definition of the sites [10]. Taken together, the sites in the survivin promoter do not display properties of bona fide CDE/CHR elements. This notion is confirmed in a later report describing transforming growth factor-β responsiveness of the survivin promoter. In the experiments the putative CHR does not contribute to regulation [45].

Also, the BUB1B gene was implicated as a CDE/CHR-regulated gene. BubR1 is a protein important for spindle checkpoint activation. Expression of the BUB1B gene coding for BubR1 is undetectable in G1, but peaks in G2/M. Recently, the regulation of the BUB1B promoter was tested. The transcription factor hStaf/ZNF143 was found to be the main activator of the promoter. Furthermore, cell cycle-dependent regulation depends on two sites with similarity to CHRs and CDEs. Interestingly, in the BUB1B promoter the CHR is located upstream of the CDE-like site [46]. These observations and the fact that activation does not rely on CCAAT-boxes, NF-Y or Sp1 proteins, leave open the question of whether the BUB1B promoter represents a canonical CDE/CHR-regulated promoter.

Cell cycle-dependent transcription of the human CDC20/p55CDC/Fizzy promoter was reported to depend on a new element named SIRF (Cell-Cycle Site-Regulating p55Cdc/Fizzy-Transcription). E2F proteins are able to bind the promoter as analyzed by chromatin immunoprecipitations and can activate transcription of the promoter in transient transfection assays through the upstream part of the SIRF element. Mutational analysis of a putative CDE/CHR site in the human CDC20 promoter showed that this element has no significant impact on promoter cell cycle regulation [47]. Without reference to this earlier report, Kidokoro and coworkers postulated a CDE/CHR in a recent paper. They observed that CDC20 expression is downregulated when p53 is active. The mechanism was suggested to require p21WAF1/CIP1, which appears to regulate the CDC20 promoter through a site just downstream of the E2F-responsive part of SIRF [48]. The results of Kidokoro and colleagues have been put into question by a very recent report identifying a p53-binding element further upstream in the CDC20 promoter as the major regulatory site [49]. According to Banerjee et al., [49] p21WAF1/CIP1, the putative CDE/CHR and CCAAT-boxes suggested by Kidokoro et al. as relevant for p53-dependent downregulation, are not required when p53 is expressed at physiological levels. Another report suggests that the human and mouse RB2 (p130) genes are controlled by a CDE/CHR-like site. The element is occupied by protein, as measured by in vivo footprinting. Mutation of this site leads to derepression of the promoter in reporter assays. However, p130 expression does not oscillate significantly during the cell cycle. Therefore, its regulation may be related to, but appears to be different from, cell cycle-controlled CDE/CHR-dependent expression. E2F family proteins did not bind to the CDE-related site [50].

In addition, some more genes were postulated to be regulated through CDE and CHR sites during the cell cycle without experimental verification. Based on the mRNA expression pattern and a promoter sequence comparison, the centromeric histone H3 homolog CENP-A gene was postulated to contain a CDE/CHR site [51]. The gene coding for the kinesin-like protein RB6K was observed to be expressed similarly to cyclin B with RB6K lagging a little behind cyclin B expression. In the RB6K promoter a tandem element with similarity to known CDE/CHR sites was observed, but not assayed functionally [52]. Furthermore, numerous CDE and CHR sites were postulated for the human, mouse and rat cyclin A genes; however, without experimental verification [53].

In summary, all examples described here contribute to the definition of which sites can be regarded as CDE/CHR sites. They also help to define class I and class II genes. One clear conclusion from the studies is that just scanning a promoter for CDE- or CHR-like sequences is not sufficient to identify functional sites. Generally, sequence alignments yield numerous hits, among them many false positives, particularly when considering the not-very-restrictive consensus for CDEs. Therefore, functional analyses are required before naming a site a CDE or a CHR. One can conclude from the many experimentally confirmed CDEs, that this class of sites, in contrast to the CHRs, is much more variable in its sequence. The consensus for a CDE may just be a site rich in G and C found upstream of a CHR with a distance of four nucleotides. Additionally, CDEs always require a CHR positioned downstream with a spacer of four nucleotides (Fig. 2).

NF-Y is the main activator, and the distance between two CCAAT-boxes is 31, 32, or 33 bp

Already with the discovery of the first CDE/CHR genes it was recognized that these promoters were activated through CCAAT-boxes binding the transcriptional activator NF-Y.

Functional CCAAT-boxes are found in both orientations. Interestingly, from the first publications on NF-Y binding to cell cycle promoters it appeared that the protein complex is constitutively bound to the CCAAT-elements throughout the cell cycle when assayed by in vivo footprinting [10,54]. However, based on chromatin immunoprecipitation (ChIP) assays, a more recent report indicates that NF-Y is only bound to DNA when the promoter is activated [55]. As the identity of proteins occupying DNA cannot be solved by in vivo footprints, it has not been ruled out that the CCAAT-boxes are bound by other proteins in G0 and G1 with a shift to NF-Y in later cell cycle phases (Fig. 3).

Figure 3.

 Possible protein occupation on class I CDE/CHR promoters. The model for regulation is primarily based on results obtained using the Cdc2 and Cdc25C promoters. In G0, proteins appear to bind to the CDE/CHR, as monitored by in vivo footprinting. According to these early experiments all binding is lost in G2/M. In contrast, constitutive binding to the CCAAT-boxes is observed. Trimeric NF-Y binds to the CCAAT-boxes and stimulates gene expression in cooperation with the histone acetyltransferase p300 in S/G2/M phases. Nevertheless, CCAAT-boxes are occupied by proteins, as suggested by in vivo footprinting in G0 and G1. However, these proteins are probably different from NF-Y and p300. For efficient activation of the promoters, the distance between the CCAAT-boxes has to be 31 to 33 bp, probably to allow binding of the p300 co-activator. In G0 and G1 phases, transcription of these cell cycle genes is repressed by a complex of inhibitory proteins at the CDE/CHR. It was shown that E2F4 binds to the CDE and that Lin-54 binds to the CHR in the Cdc2 promoter in G0. It is probable that these proteins constitute part of the DREAM complex on these promoters because Lin-54 is a constitutive member of DREAM. Furthermore, in later cell cycle phases DREAM appears to activate promoters, whereas Lin-54 may be bound to sites other than the CHR. In order to activate in S/G2/M phases, the composition of DREAM is altered by replacing E2F4 and p107/p130 with B-Myb. Because in vivo footprints provided evidence that the CDE/CHR in G2/M cells is devoid of proteins, but DREAM components were detected at CDE/CHR promoters in G0/G1 as well as in S/G2/M phases, it is likely that the complex is able to bind alternative recognition sites outside the CDE/CHR. It still has to be established which proteins contact DNA at which sites during late phases of the cell cycle.

Many cell cycle genes were found to contain two or three CCAAT-boxes essential for promoter activity (e.g. the mouse cyclin B1 and cyclin B2 genes) [56,57]. A dominant-negative variant of the NF-YA subunit of the heterotrimeric complex NF-Y was instrumental in showing that the activating protein on multiple CCAAT-boxes is indeed NF-Y [27,57].

The multiple sites synergize. Individual mutations lead to a large drop in promoter activation, indicating cooperation between the two or three CCAAT-boxes of a gene [54,57]. Conspicuously, the distance between two functionally important CCAAT-boxes is always 31, 32 or 33 bp. Comparison of nucleotide sequences in promoters of ortholog genes from different organisms shows that not only the CCAAT-boxes themselves, but also their distance, is conserved [7,25,58]. The particular distance with approximately three turns of the DNA double helix yields binding of the two or three NF-Y complexes on the same side of the DNA. Conservation of spacing is required for optimal promoter activity because changing the distance leads to a loss of activation [58].

One reason for the specific spacing between CCAAT-boxes may be binding of the p300 histone acetyltransferase (HAT) to NF-Y heterotrimers. Association of NF-Y with HAT activity had been observed earlier. A complex consisting of the three NF-Y subunits and other proteins has been shown to possess histone acetyltransferase activity through physical association with the related GCN5 and P/CAF enzymes [59]. The p300 HAT enzyme was observed to bind to the human cyclin B1 promoter in vivo and is able to increase expression from the promoter when overexpressed [5]. p300 binding requires all three CCAAT-boxes and association of NF-Y with these elements for optimal transcriptional activation of the mouse cyclin B2 promoter. Changing the distance of the CCAAT-sites reduces p300-dependent activation [58]. Recruitment of p300 HAT on the cyclin A and Cdc2 promoters may also be in accordance with activation and histone H3 and H4 acetylation beginning in late G1, as observed by ChIP experiments [60].

Interestingly, NF-Y appears to form interactions also with other activating factors. The results of employing plasmid-based ChIP assays on the Cdc2 promoter indicate that E2F3 binding to the distal activating E2F site may require an intact CCAAT-box occupied by NF-Y [61]. NF-Y proteins bind to the human Cdc2, cyclin B1 and cyclin A2 promoters throughout the cell cycle, as determined using ChIP assays [61]. This is consistent with earlier genomic footprinting observations [9,10,54].

In addition to the few CDE/CHR promoters analyzed in detail to determine a role of NF-Y in their control, a number of such genes were implicated to be regulated through CCAAT-boxes. Cotransfection of dominant-negative NF-YA demonstrated that most of the Tome-1 promoter activity depends on NF-Y. One CCAAT-box had been tested for its role in activation by reporter assays comparing wild-type with mutant promoters. It was not further investigated whether another NF-Y-binding site at a specific distance may be required in conjunction with this first CCAAT-box. Furthermore, ChIP assays demonstrated that NF-Y can bind to the Tome-1 promoter in vivo [37].

Also, the Aurora B promoter had been established as being CDE/CHR-regulated. Similarly to other CDE/CHR promoters, the Aurora B gene does not contain a TATA-box, but two CCAAT-boxes with a distance of 33 bp to each other were found upstream of the CDE/CHR [40]. However, the two sites were not tested functionally.

Generally, to our knowledge, all promoters containing functional CDE/CHRs that were also tested for CCAAT-boxes were indeed found to be activated through their CCAAT-boxes. However, many CDE/CHR genes have not been assayed for CCAAT-box-dependent activation. Therefore, it appears likely, but is not proven, that CCAAT-boxes are required to activate CDE/CHR promoters.

p53 and repression through the CDE/CHR

Conspicuously, many CDE/CHR genes, such as Cdc2, cyclin B1, cyclin B2 and Cdc25C, are downregulated by the tumor suppressor p53 [27,62–66]. However, there are also many examples of genes repressed by p53 that are not regulated by CDE/CHR sites but have been tested for such elements (e.g. Cdc25A and Cks2) [67,68].

Evidence exists that DNA damage-dependent downregulation of Cdc2 transcription relies on intact CDE and CHR elements. A report implied p53 and p21WAF1/CIP1 in this downregulation by employing p53-positive or p53-negative cell lines [69]. For the Plk1 gene, coding for Polo-like kinase 1, downregulation by p21WAF1/CIP1 was also postulated to be controlled through CDE and CHR elements [70]. Experiments confirmed that p53 and p21WAF1/CIP1 regulate, in part, through the CDE and CHR sites. However, mutation of the CDE/CHR did not completely abrogate p53-dependent downregulation [71]. Earlier, it had been shown that the CHR in the Plk1 promoter was more relevant for cell cycle-dependent transcription than the CDE [35]. Furthermore, the topoisomerase IIα gene is downregulated by overexpression of p21WAF1/CIP1. Like the Plk1 gene, topoisomerase IIα was presented as downregulated through CDE/CHR sites upon p21WAF1/CIP1 overexpression [70]. However, a combination of a CDE and an adjacent CHR had been postulated only by sequence comparison [72] but was not confirmed by experiments [73].

The mouse cyclin B2 and human Cdc25C promoters are downregulated by p53 [65,66]. In order to pinpoint the site responsible for the repression, numerous cyclin B2 promoter mutants were tested. p53-dependent downregulation does not appear to be dependent on the CDE and CHR sites. Also, other regions or specific sites could not be conclusively established asresponsible for repression. One challenge investigating downregulation of the cyclin B2 promoter is that after deletion or mutation of constitutively activating sites, such as the destruction of CCAAT-boxes, small levels of reporter activity remain to analyze further repression through the CDE or CHR elements [25,65].

The CDE/CHR in the human Cdc25C promoter was implicated in the p53-dependent downregulation of the gene [74]. Initially, a potential p53-binding site was observed in the human Cdc25C promoter, which is able to bind p53 in electrophoretic mobility shift assays (EMSAs). When fused to a minimal promoter, the p53 site can function as a transcriptional activator from a reporter construct [75]. However, Cdc25C is repressed by p53. Also, mouse Cdc25C is downregulated by p53 but lacks the putative p53 site in its promoter [27]. By contrast, downregulation needs the CCAAT-boxes in the promoter and functional NF-Y transcription factors [66,76]. p53-dependent repression of the human Cdc25C promoter does not require the putative p53 site implicated earlier [75], or the CDE/CHR, but is lost when three CCAAT-boxes are deleted [66]. In a later report the CDE/CHR was implicated to be responsible, at least in part, for p53-dependent downregulation of the Cdc25C promoter. The results were based on transient expression experiments employing a 76 bp CDE/CHR-containing promoter fragment cloned upstream of a minimal adenovirus E1b promoter-driven luciferase reporter [74]. The same reporter system yielded activation through a putative p53-binding site that was later implicated in downregulation of Cdc25C by the same group [74,75]. This short Cdc25C promoter fragment lacks the CCAAT-boxes originally found to be responsible for most of the promoter activity [10,54]. Therefore, the promoter exerts an artificially low activity. Further downregulation measured with this short fragment may be unnatural. The partial downregulation, observed in this experimental system, through the Cdc25C CDE/CHR stands in contrast to earlier results. In these experiments, deletion of the Cdc25C CDE/CHR in the full promoter context had almost no effect on p53-dependent repression after overexpressing p53 in transient transfection assays [66]. Furthermore, no binding of p53 protein to the segment was observed for the proposed CDE/CHR-dependent repression mechanism [74].

By contrast, protein binding to promoters of genes such as cyclin B1 had been demonstrated upon p53 induction. Mannefeld et al. [77] describe, in a recent report, that in the presence of p53 the DREAM complex switches from containing activating B-Myb to repressing E2F4/p130. Although the report does not specify binding sites, other reports imply CDE/CHR sites for binding of DREAM (please see the later discussion on the binding of DREAM proteins to CDE/CHR elements). However, because for some CDE/CHR genes, such as mouse cyclin B2 and human Cdc25C, a function of the CDE and CHR sites in p53-dependent downregulation appeared unlikely, it will be of interest to establish the promoter sites to which DREAM complex components bind to participate in p53-dependent transcriptional repression.

Influence of viral proteins on CDE/CHR regulation

As it had been noted that CDE/CHR sites are related to E2F elements, it is a pertinent question whether viral proteins disturb regulation through the tandem element in a manner similar to viral oncoproteins interfering with the control by E2F and pRB-related pocket proteins. One example of a gene deregulated by viral proteins is Cdc2. The human Cdc2 promoter is upregulated upon the expression of simian virus 40 (SV40) T antigen. However, CCAAT-boxes were made responsible for the transcriptional activation by the viral protein, whereas interaction of T antigen with p53 or pRB did not appear to be essential [78].

Similarly, expression of the SV40 T oncogene resulted in deregulation of the Cdc25C promoter by destroying repression in G0 and G1. Expression of SV40 T antigen in promoter-reporter assays yielded deregulation, which was dependent on the CDE of the human Cdc25C promoter. Dimethyl sulfate footprinting of the CDE in the presence of SV40 large T indicated a loss of protein occupation on this site in vivo [79].Elevated expression of cyclin A, cyclin B, Cdc25C and Cdc2 had already been observed after expression of SV40 T antigen, which led to the disruption of mitotic checkpoints [80]. This information, combined with the change of protein occupation on the Cdc25C CDE, indicates that CDE/CHR sites which regulate cyclin A, cyclin B and Cdc2 promoters may lose binding of their regulator proteins and repression in G0/G1.

Deregulation by viral proteins was also tested using the mouse cyclin A promoter as an example. Polyomavirus T antigen has functions similar to those of adenovirus E1A, human papillomavirus E7 proteins and simian virus T antigen regarding the dissociation of pocket proteins from E2Fs [81]. Large T from polyomavirus was able to deregulate transcription, which was dependent on the CDE in the mouse cyclin A promoter. The CHR was not tested separately. However, protein complexes did not change as one would expect when pocket proteins dissociate from the complexes, and free E2F would remain bound to the site [81].

Proteins binding to the CDE and CHR elements

One central question stemming already from the early days of CDE/CHR research is which protein regulators bind to this tandem site. E2F proteins were implicated early on to regulate through the CDE. Analysis and identification of factors regulating through the CHR is particularly important because class II promoters lack functional E2F/CDE elements (Fig. 2).

Preliminary reports on proteins requiring a CHR for binding include a factor named CDF-1, which was observed to bind to the CDE/CHR elements in the Cdc25C and cyclin A promoters [82,83]. However, many attempts by several groups to clone and further characterize this factor failed. A protein called CHF has been observed to bind by EMSA to the CHR in the mouse cyclin A promoter. However, also this factor was not further characterized [84].

More information is available on the binding of E2F and pocket proteins to CDE/CHR promoters. There are several hints for a functional connection of E2F4 binding and CDE-dependent regulation. It was shown, by EMSAs, that E2F4 and p130, but not E2F1, E2F2, p107 or pRB, are able to bind to the CDE in the Cdc2 promoter in vitro [9]. Later, when ChIP was developed, binding of E2F4, p107 and p130 to the Cdc2 and B-Myb promoters was shown without specifying the particular site [85].

Similarly, E2F4/DP1 and p107 associate with the mouse B-Myb E2F site when assayed by EMSAs. This binding site is close to the CHR-related DRS element [20]. Interestingly, the core of this E2F element displays an identical sequence to the CDE from the human Cdc25C promoter (Fig. 2), but, in contrast to the B-Myb site, the Cdc25C element does not bind any E2F proteins in vitro [83]. Obviously, nucleotides outside the core sites are responsible for this differential binding, possibly the different CHRs. Probably, the distinct binding to the related elements, together with the different DRS/CHR, are responsible also for altered timing of gene expression as seen with B-Myb mRNA appearing earlier in the cell cycle than Cdc25C expression (Fig. 2). E2F binding-site occupation is reduced in p107−/− p130−/− MEFs compared with wild-type cells when assayed using in vivo footprinting. Loss of the two pocket proteins leads to deregulation of the B-Myb promoter during the cell cycle. Re-introduction of p107, and to some extent also p130, causes repression of the B-Myb promoter reporter, which is dependent on an intact E2F site. Regulation cannot be restored by pRB. Another important aspect related to protein binding to the E2F site in the B-Myb promoter is that ChIP assays on stably transfected promoter constructs reveal that in vivo binding of p107 and p130 to the B-Myb promoter is lost when the DRS is mutated [22]. By analogy, one is tempted to speculate that also bona fide CDE and CHR sites cooperate in protein binding. Similar results on the cooperative regulation of putative CHR-binding proteins with an E2F/CDE site were observed when studying Cdc2 promoter regulation. Mutation of the CHR abolished the interaction of E2F4 with the negative-regulating proximal E2F/CDE site, as measured by ChIP with transfected mutant human Cdc2-promoter plasmids [10,61]. Later, this cooperation was confirmed by Lin-54 binding to the CHR, cooperating with E2F4 binding to the CDE in EMSAs in vitro [86].

ChIP experiments on the Cdc2 gene also provided insights into the association of other cell cycle proteins with the promoter. E2F4 binding during the cell cycle coincides with the binding of p107 or p130 [61,85]. Another E2F site distal to the E2F/CDE acts as an activating element binding E2F1, -2 and -3. These activating E2Fs cooperate with NF-Y proteins binding to CCAAT-boxes and with Myb proteins associating with a distal Myb site in activating the Cdc2 promoter. Binding of the activating E2Fs during the cell cycle to the promoter alternates with binding of E2F4 in quiescent cells. Mutation of the distal E2F site allows E2F4 to associate with the Cdc2 promoter also in G2 cells, suggesting that E2F1, -2 and -3, although they bind to the distal E2F site, block binding of E2F4 to the CDE [61].

For other CDE/CHR promoters, E2F and pocket proteins were also implicated in binding. The CDE in the human Aurora B promoter bound DP2, E2F1 and E2F4, but not DP1, when HeLa nuclear extracts were employed for biotin–streptavidin pull-down assays followed by western blot analysis [40]. Furthermore, the CDE in the human cyclin A promoter, also identified as a ‘variant E2F site’, binds E2F complexes (including DP1 and p107) in EMSAs from nuclear extracts. The complexes appear to lack pRB, E2F1, E2F2 or E2F3 because antibodies directed against these proteins did not recognize any of the protein complexes formed on this site in the cyclin A promoter [10,15]. By contrast, another set of experiments showed that the cyclin A CDE can bind E2F1 and E2F3 from HeLa nuclear extracts, as well as recombinant E2F1 and DP1 glutathione S-transferase fusion-protein complexes in EMSAs. However, E2F4 was not shifted with this probe although the protein was present in the HeLa extracts [83]. In a large study employing ChIP on cell cycle promoters from human cells, the cyclin A promoter was described to bind E2F4 and p130 in G0 and early G1in vivo. In late G1, some E2F3 binding appears in the ChIP assays [60]. Similarly, ChIP experiments on the cyclin A, Cdc2 and cyclin B2 genes suggest that E2F4 and E2F6 bind to the promoters in G0 and to some extent also in G1 [55]. Furthermore, E2F4 binds to the promoters of cyclin B1 and cyclin B2 in human tissues according to ChIP, followed by identification of genes by genome-wide DNA-microarray hybridization [87]. However, the limitation of ChIP analyses is that the exact binding location is not identified in the experiments. Therefore, it is not clear whether the CDE/CHR is involved in binding these factors.

Consistent with a function of p107 and p130 in the control of CDE/CHR promoters is the loss of regulation observed for the expression of cyclin A, B-Myb and Cdc2 in p107−/− p130−/− double knockout cells. However, individual deletion of pocket proteins in p130−/− or p107−/− MEFs does not lead to deregulation, indicating that p107 and p130 can substitute for each other [88]. It is likely, although not proven, that regulation by p107 and p130 is executed through the E2F/CDE sites in these promoters. Interestingly, Cdc25C, the gene with the first identified CDE/CHR promoter, is not deregulated in Rb−/− or p107−/− p130−/− MEFs [88]. This implies that Cdc25C does not require pocket proteins for its regulation and is regulated differently from cyclin A, B-Myb or Cdc2.

The DREAM complex: a possible role in CDE/CHR-dependent regulation

E2F4 and the pocket protein p107 have been shown, in vitro by EMSA, to bind to the B-Myb promoter E2F site with its neighboring CHR-related DRS element. Importantly, p107 and p130 binding, measured using ChIP on stably transfected promoter constructs, is lost when the DRS/CHR is mutated [20,22]. Furthermore, E2F4 binds to the CDE in the Cdc2 gene assayed using ChIP in cells transfected with promoter-carrying plasmids. Importantly, this association also requires an intact CHR, indicating cooperation between proteins binding to both sites of the tandem element [61].

In recent years, E2F4 has been identified as a component of the DREAM complex. The ortholog dREAM complex was first identified in Drosophila [89]. Human DREAM was isolated by employing the MudPIT method with overexpression of an HA-tagged p130 protein followed by extraction via the tagged protein, chromatography and identification by MS [90,91]. In another approach the complex was affinity-purified, employing a stably expressed flag-tagged Lin-37 protein followed by MS. Composition of the protein complex was later confirmed by purifying the endogenous components by classical chromatography monitored through Lin-37 western blotting [92]. The complex contains mammalian homologs of Caenorhabditis elegans synMuvB proteins and is composed of p130, E2F4/5 and DP1/2 and a module containing the MuvB proteins Lin-37, Lin-52, Lin-54 and chromatin-associated Lin-9 and Lin-53/RBBP4/RbAp48. Furthermore, A-Myb and B-Myb were present in immunoprecipitations for Lin-9, Lin-37 and Lin-54 [90,92]. Composition of the DREAM complex changes with cell cycle phases. Lin proteins form a core complex that associates with different proteins during passage through the cell division cycle. It appears that E2F4/p107/p130 form part of the complex in G0 phase and B-Myb constitutes a component of the complex in S-phase. ChIP analyses have shown that DREAM complex components are able to bind to CDE/CHR-regulated promoters, mostly without specifying particular binding sites [20,22,61,90,92–96]. Very recently, it has been shown that Lin-54 can bind to two sites in the Cdc2 promoter in vitro. Lin-54 binds in EMSAs to a region in the upstream part as well as to the CHR of the Cdc2 promoter [86].

Cyclin A and human cyclin B1 are class I and class II CDE/CHR-controlled genes, respectively (Fig. 2). Both promoters become derepressed in G0 cells when CDE/CHR sites are mutated [5,10]. Supporting this observation, Litovchick et al. [90] showed an upregulation in the expression of cyclin B1 after knockdown of DREAM complex components in G0-arrested cells. Consistently, the opposite effect was shown with a slowdown in cell cycle progression after serum restimulation of G0 cells when p130, Lin-9, Lin-37 or Lin-54 are stably overexpressed.

In contrast to a possible function in cell cycle-dependent repression mediated by CDE/CHR sites in the G0 phase, Lin proteins contribute to activation of CDE/CHR promoters in cycling cells. It was shown that Lin-9 activates the cyclin A and cyclin B1 promoters [94]. Interestingly, functional mapping of the cyclin B1 promoter did not yield the class II CHR as the site through which Lin-9 exerts its control [94]. In another report, depletion of Lin-9, Lin-37, Lin-52 or Lin-54 by RNAi also led to a repression of genes, such as cyclin A and cyclin B1, in dividing cells [92]. Furthermore, depletion of Lin-9 and B-Myb in p53/ MEFs caused lowered expression of cyclin A, cyclin B and Cdc2 [94]. In similar experiments, knockdown of Lin-9 or Lin-54 was responsible for strong shifts in cell cycle distribution. A large cell fraction is found in G2 or appears to have an 8n DNA content. Many cell cycle-regulated genes are downregulated upon Lin-9 knockdown (e.g. Polo-like kinase, Aurora kinase B, cyclin A, Cdc2, cyclin B1 and cyclin B2) [93,96]. The shift in cell cycle-distribution towards the later phases would favor elevated expression of these genes. Without this effect, downregulation would be expected to be even lower, whereas an upregulation would indicate a function of Lin-9 in repression. This indicates that Lin-9 is involved in the activation of these genes in cycling cells.

In summary, it has been shown that the DREAM complex controls expression of some CDE/CHR-regulated genes. E2F4 can bind to certain CDE sites. Lin-54 can bind to the CHR in the Cdc2 promoter in vitro. Incorporating the results from different reports, it appears that a change in the composition of the DREAM complex shifts its properties from a repressor complex in G0 cells to an activator complex in S/G2/M cells. It is likely that these contrasting features are mediated by the shift from E2F4/p107/p130 to B-Myb (Fig. 3). It still has to be shown which elements in the promoter are bound by the ‘activating DREAM complex’ because the CDE/CHR tandem element is not occupied by proteins in later cell cycle phases and is not necessary for activation of the promoters.

Conclusions and perspectives

What defines CDE and CHR elements? A CHR usually has the sequence 5′-TTTGAA-3′. However, this sequence appears too frequently in promoters to clearly indicate a CHR with a role in cell cycle-dependent transcriptional repression (Table 2). Therefore, this site has to be confirmed in each case by functional assays. CDEs follow a loose consensus similar to that of E2F sites. What distinguishes CDEs from E2F elements? CDEs must always be located with a four-nucleotide spacer upstream of a CHR. Both CDEs and CHRs control cell cycle-dependent transcription as repressor sites in G0 cells. Classic E2F sites do not appear next to CHR elements and can either activate or repress transcription. In the genes analyzed to date, any promoter appears to hold only one CDE/CHR pair with the CDE upstream of the CHR. CDE/CHR promoters employ multiple start sites, do not bear functional TATA-boxes and are usually activated by two or three CCAAT-boxes with certain spacing. Do E2F proteins regulate CDE/CHR promoters? There is considerable evidence that E2F4 is involved in their control through the CDEs (Fig. 3).

Central questions still have to be addressed. Are all CDE/CHR promoters regulated by the same protein complex? Will changes in the sequence of a CDE/CHR tandem site lead to a recruitment of different proteins? Are class I and class II promoters bound by different protein complexes? Why are some promoters class I in mouse and class II in humans, and vice versa, as demonstrated for cyclin B genes? How are class II promoters, which do not contain a functional CDE, regulated? If the DREAM complex is the main regulator of CDE/CHR cell cycle genes, to which promoter sites does it bind during any particular phase of the cell cycle? What is the exact mechanism of p53-dependent downregulation of CDE/CHR genes? Is this mechanism the same for all promoters?

More than a decade after the initial observation of the tandem site, CDE/CHR-dependent regulation has been established for many central genes involved in cell cycle control. The CDE is demonstrated to be different from classical E2F elements. Class II promoters have been identified as controlled by just a CHR and to lack a functional CDE. Links have been discovered from CDE/CHR promoter control to pathways regulated by p53, E2F and the pRB family.

Acknowledgements

G.A.M. was the recipient of a graduate fellowship awarded by the Freistaat Sachsen. Research in our group was supported by the Bundesministerium für Bildung und Forschung (BMBF) through the Interdisciplinary Center for Clinical Research (IZKF) at the University of Leipzig and the Deutsche Forschungsgemeinschaft by grants SPP 314, EN 218/6-1 and 6-2 (to K. E.).

Ancillary