Molecular stop signs: regulation of cell-cycle arrest by C/EBP transcription factors

One of the central problems in biology is to understand how organisms regulate cell division in response to developmental and environmental cues. Although cells must undergo controlled replication, they are also required to enter a quiescent state in a precisely regulated manner, such as during terminal differentiation, to maintain tissue homeostasis and in response to DNA damage or oncogenic stress (senescence). Indeed, for multicellular organisms, the decisions to exit the cell cycle are at least as important as those that drive proliferation. The general subject of cell-cycle control encompasses a vast literature and many reviews. By contrast, the scope of this Commentary is much more limited – to summarize the roles of CCAAT/enhancer-binding protein (C/EBP) transcription factors in regulating cell-cycle arrest. Because even this more-narrowly defined field now includes an extensive literature, this is not intended to be a comprehensive review of the area but, rather, focuses on recent discoveries and areas of controversy.

C/EBP proteins are a family of basic region leucine zipper (bZIP) transcription factors that includes six members, C/EBPα, C/EBPβ, C/EBPγ, C/EBPδ, C/EBPϵ and C/EBPζ (genetic nomenclature CEBPA, CEBPB, CEBPG, CEBPD, CEBPE and CEBPZ, respectively), with related sequences and functions (Fig. 1A). Except for C/EBPζ, which lacks a canonical basic region, each protein contains similar basic region and leucine zipper sequences at its C-terminus, which mediate DNA binding and dimerization, respectively. C/EBPs bind to palindromic DNA sites as homo- or heterodimers (Fig. 1B). The N-terminal portion of each protein contains effector domains that mediate transcriptional activation, repression and autoregulatory functions. C/EBPγ, owing to its small size (∼20 kDa), lacks any known functional domains outside the bZIP region. The C/EBPα and C/EBPβ transcripts also encode internally initiated translation products termed p30 and liver inhibitory protein (LIP), respectively, which retain the C-terminal DNA-binding domain (Descombes and Schibler, 1991; Lin et al., 1993; Ossipow et al., 1993). These proteins lack the major N-terminal transactivation domain and can dominantly inhibit the activating C/EBP isoforms.

C/EBPγ is ubiquitously expressed; C/EBPβ is also found in nearly all cells, whereas C/EBPα expression is more restricted but nonetheless evident in a wide range of cell types. C/EBPδ is primarily an inducible protein whose expression is activated by inflammatory and stress stimuli; C/EBPϵ is largely confined to granulocytic cells of the hematopoietic system. Most of the available data suggest that C/EBP proteins function as transcriptional activators, but it is becoming increasingly apparent that they also repress transcription of certain target genes and may even exert non-transcriptional effects.

C/EBPα is the founding member of the C/EBP family and is expressed predominantly in post-mitotic cells. Activation of a chimeric C/EBPα-estrogen-receptor (ER) protein by estrogen was shown to arrest 3T3-L1 pre-adipocytes in G0/G1 (Umek et al., 1991), providing the first evidence that C/EBPα can enforce cell-cycle exit and accounting for earlier unsuccessful attempts to generate hepatic cell lines stably expressing a CEBPA transgene (A. Friedman, personal communication). C/EBPα expression is also induced during adipogenesis and activates genes specifically expressed in differentiated fat cells (Christy et al., 1989). These and other observations (Friedman et al., 1989) led to the notion that a single transcription factor can control tissue-specific gene expression and proliferation arrest, thereby coordinately regulating two essential features of terminally differentiated cells. Subsequent studies demonstrated that forced expression of C/EBPα commits 3T3-L1 pre-adipoblasts to the adipocyte differentiation program and induces cell-cycle arrest (Freytag et al., 1994) and showed that the antiproliferation activity of C/EBPα extends to hepatocytes (Hep3B2) and Saos2 osteosarcoma cells (Hendricks-Taylor and Darlington, 1995). In vitro studies, analysis of knockout mice and examination of leukemic cells have since revealed that C/EBPα can induce cell-cycle exit in many contexts and significantly affects cellular differentiation and tumorigenesis.

C/EBPα-null mice show grossly normal embryonic development but neonates are hypoglycemic and die shortly after birth (Wang et al., 1995). Their viability can be extended for a few days by intravenous glucose administration. The underlying defect is a reduction in hepatic glycogen stores owing to impaired liver-specific expression of glycogen synthase and the gluconeogenic enzymes glucose-6-phosphatase and phosphoenolpyruvate carboxykinase. Although there is no clear evidence for hyperproliferation of C/EBPα^–/– hepatic cells in vivo, cultured primary hepatocytes from these embryos display increased DNA synthesis and a much greater propensity for immortalization (Soriano et al., 1998). C/EBPα^–/– neonatal livers contain increased levels of Jun and Myc transcripts and have elevated expression of proliferating cell nuclear antigen (PCNA), which indicates enhanced proliferative activity (Flodby et al., 1996); however, liver size is normal. Although the basis for the more severe growth deregulation in cultured C/EBPα^–/– hepatocytes has not been established, one can envision proliferative controls imposed by the tissue environment, such as limited availability of growth factors, that override the loss of growth constraints in vivo.

Terminal differentiation of granulocytes is also dramatically impaired in the absence of C/EBPα (Zhang et al., 1997). C/EBPα^–/– embryos exhibit a lack of mature neutrophils and eosinophils and increased numbers of blasts in blood and fetal liver (FL). This maturation defect results in part from reduced transcription of the gene encoding granulocyte colony-stimulating factor receptor (G-CSFR), whose promoter is a known target of C/EBPα (Smith et al., 1996). However, the granulopoietic defects in C/EBPα^–/– mice are more severe than those of animals lacking the G-CSFR gene (Liu et al., 1996), suggesting that there are additional functions and targets for C/EBPα in the regulation of terminal granulocytic differentiation. This conclusion is further supported by experiments in which dominant-negative C/EBP and the G-CSFR were co-expressed in a myeloid cell line (Wang and Friedman, 2002).

C/EBPα^–/– FL also lacks mature macrophages and macrophage progenitors (Heath et al., 2004). In addition, there are fewer bipotential granulocyte/macrophage precursors in C/EBPα^–/– FL but normal numbers of multipotent cells. Mutant FL cells transplanted into wild-type irradiated recipients fail to generate macrophages but give rise to very large, hyperproliferative spleen colonies, and the host mice develop myelodysplastic syndrome. This indicates that C/EBPα functions not only during terminal differentiation but also at an early stage of myelopoiesis. Interestingly, C/EBPα^–/– FL cells grown in liquid culture with hematopoietic growth factors such as interleukin (IL)-3 show unrestrained mitotic growth and increased self-renewal, whereas wild-type cells undergo terminal myeloid differentiation and cease to proliferate after a few weeks. The C/EBPα^–/– cells behave like an immortalized cell line and can be cultured indefinitely, although the cells remain dependent on hematopoietic growth factors for proliferation and survival (Heath et al., 2004). Thus, the existing data point to a critical role for C/EBPα in maturation of granulocytes and macrophages and its involvement in control of cell-cycle exit during terminal differentiation.

Recent experiments in which the CEBPA gene was inducibly deleted in adult mice demonstrate that C/EBPα is required for the common myeloid progenitor cell (CMP) to give rise to the more committed granulocyte/monocyte progenitor (GMP) (Zhang et al., 2004). In addition, deletion of C/EBPα leads to increased numbers of blast cells and enhances the competitive repopulation of hematopoietic stem cells (HSCs) relative to wild-type cells in transplantation experiments. The increased self-renewal potential of C/EBPα^–/– HSCs may, at least in part, be due to elevated expression of the polycomb gene Bmi-1. The latter two studies therefore reveal that C/EBPα functions much earlier in hematopoiesis than previously thought, providing a partial brake on proliferation even in stem cells and progenitor cells.

The potent antiproliferative activity of C/EBPα in myeloid cells suggested that it might function as a suppressor of leukemogenesis. Sequencing of CEBPA alleles from patients with acute myelogenous leukemia (AML) (Pabst et al., 2001b) revealed that 7.3% of those examined carry spontaneous heterozygous CEBPA mutations in tumor cell DNA. One class of mutations alters residues in the bZIP region, causing diminished DNA-binding activity. A second class, constituting half of these patients, causes truncation of the full-length protein and upregulated expression of an internally initiated translation product, p30 (Fig. 2). The p30 polypeptide lacks the N-terminal transactivation domain but retains the C-terminal DNA-binding domain and functions as a dominant inhibitor of C/EBP-mediated transcription (Lin et al., 1993; Ossipow et al., 1993).

In contrast to wild-type C/EBPα, the mutant proteins cannot induce neutrophil differentiation when expressed in bipotential myeloid precursor cells (Pabst et al., 2001b; Radomska et al., 1998). Interestingly, p30 strongly inhibits the ability of co-expressed full-length C/EBPα to bind to its cognate site in the G-CSFR gene. However, this inhibitory effect is highly dependent on the C/EBP-binding site tested (Cleaves et al., 2004). The mechanism for the dominant-negative effect of p30 on DNA binding remains obscure, but it probably accounts for the disruption of C/EBPα activity in heterozygotic leukemic cells containing p30 and the wild-type allele. Also puzzling is the observation that p30 inhibits differentiation of human myeloid progenitors but lacks this activity in mouse cells (Schwieger et al., 2004).

The expression of truncated or other mutant C/EBPα polypeptides probably facilitates AML development by imposing a differentiation block and disrupting normal cell-cycle exit. The hyper-proliferating mutant cells may be at increased risk of acquiring other oncogenic mutations that are necessary for transformation and leukemogenesis. Whether CEBPA mutations initiate leukemogenesis or arise during later stages of cancer development is unclear, and further experiments using animal models will be necessary to answer this question.

The studies of Pabst et al. lead to two important conclusions: first, C/EBPα has a role in regulating differentiation of human myeloid cells, as in mice; second, loss of C/EBPα function can contribute to human cancers. Experiments showing that reinstating C/EBPα expression in AML cells suppresses their proliferation in vitro and tumorigenicity in vivo support these conclusions (Truong et al., 2003). Although CEBPA mutations have not been reported for other cancers, C/EBPα expression is known to be downregulated in lung tumors (Halmos et al., 2002), hepatocarcinomas (Friedman et al., 1989; Xu et al., 1994) and squamous cell carcinomas (Oh and Smart, 1998; Shim et al., 2005). In addition, the oncogenic fusion proteins BCR-ABL and AML1-MDS1-EVI1 (AME), which are synthesized as a consequence of chromosomal translocations associated with myeloid leukemias, inhibit C/EBPα translation (Helbling et al., 2004; Perrotti et al., 2002), and AML-ETO blocks CEBPA transcription (Cilloni et al., 2003; Pabst et al., 2001a). Collectively, these observations indicate that `dialing down' C/EBPα activity by one of several mechanisms is necessary for the decreased differentiation and increased proliferative capacity of cancer cells, specifically those in which C/EBPα controls differentiation of the normal tissue. Hence, therapeutic compounds selected for their ability to reactivate C/EBPα expression (thus inducing cell-cycle arrest and differentiation) might be effective agents for treating cancers such as AML (Tenen, 2003).

Recently, Yoon and Smart identified another antiproliferative function of C/EBPα involving DNA-damage-induced growth arrest in epidermal keratinocytes. Exposure of keratinocytes to ultraviolet B (UVB) radiation causes DNA damage and activates a p53-dependent G1 checkpoint that arrests cell-cycle progression during DNA repair. C/EBPα expression is strongly induced by UVB irradiation in cultured keratinocytes and mouse skin (Yoon and Smart, 2004). Knocking down C/EBPα by RNA interference (RNAi) greatly reduces UVB-induced cell-cycle arrest; furthermore, C/EBPα induction is impaired in p53-mutant cells, and p53 inducibly associates with the CEBPA promoter in chromatin immunoprecipitation (ChIP) assays. These results suggest that CEBPA is a direct transcriptional target of p53 in keratinocytes and that C/EBPα functions as a critical effector of the p53-dependent growth arrest response. Thus, in addition to regulating terminal differentiation, C/EBPα can be induced by stress signals that inhibit cell proliferation during DNA repair, which could contribute to its role in tumor suppression. Whether C/EBPα has a widespread function in DNA repair responses or whether this is mainly restricted to epidermal keratinocytes, which are particularly susceptible to UVB-induced DNA damage, remains to be established.

Since C/EBPα induces cell-cycle arrest in a variety of physiological settings, an important objective has been to determine the molecular mechanism by which it inhibits cell proliferation. Below, I discuss several different published models of C/EBPα-induced growth arrest, and these are depicted schematically in Fig. 3.

The first studies to investigate the C/EBPα antiproliferative mechanism implicated the cyclin-dependent kinase (CDK) inhibitor p21. p21 binds to and inhibits the kinase activity of CDK4, CDK6 and CDK2 associated with their cyclin regulatory subunits. p21 can be induced by stress signals such as DNA damage that activate the p53 stress response pathway. Using a cell line containing an inducible CEBPA transgene, Timchenko et al. (Timchenko et al., 1996) observed that p21 levels increase nearly 20-fold upon expression of C/EBPα. This effect is partly transcriptional but mainly post-transcriptional, involving an association between p21 and C/EBPα that stabilizes p21. Levels of p21 are reduced in livers from newborn C/EBPα^–/– mice and in hepatocytes from regenerating liver, where C/EBPα is downregulated in proliferating cells during tissue renewal (Mischoulon et al., 1992; Timchenko et al., 1997). The ability of C/EBPα to block growth of Saos2 cells (Hendricks-Taylor and Darlington, 1995) indicates that another p21 regulator, p53, is dispensable for C/EBPα to induce proliferation arrest, because Saos2 cells lack functional p53.

Later studies argue against a significant role for p21 (Muller et al., 1999). Leutz and coworkers discovered that the HPV16 E7 oncoprotein disrupts the ability of C/EBPα to induce cell-cycle exit but affects neither its transcriptional activation function nor its differentiation-inducing activity in pre-adipocytes (Muller et al., 1999). These observations reveal that the growth arrest and differentiation functions of C/EBPα are separable activities. The authors also found that C/EBPα can induce cell-cycle exit in p21-deficient mouse embryo fibroblasts (MEFs), and that E7 blocks C/EBPα-induced growth arrest in p21-null cells. Thus, at least in fibroblasts, p21 is dispensable for the anti-mitotic activity of C/EBPα. The normal responses to C/EBPα in p21-deficient cells are difficult to reconcile with an important role for a C/EBPα-p21 mechanism in proliferation arrest.

Subsequent work has implicated CDK2 and CDK4 as targets of C/EBPα inhibition (Harris et al., 2001; Wang et al., 2001). Both play key roles in cell-cycle progression in G1-phase by phosphorylating specific substrates, the most important of which is the retinoblastoma (pRB) tumor suppressor. Phosphorylation inactivates pRB, causing its release from E2F transcription factors and de-repression of S-phase genes that are either required for DNA replication or regulate subsequent cell-cycle events. C/EBPα binds CDK2 and CDK4 in vitro and inhibits their ability to phosphorylate substrates such as histone H1. Moreover, deletion of the CDK-interacting sequences abrogates C/EBPα-induced growth arrest in transfected cells.

Wang et al. (Wang et al., 2001) mapped the CDK2/CDK4-binding region on C/EBPα to a sequence spanning residues 175-187; this group also observed that C/EBPα disrupts the association of CDKs with cyclins. By contrast, Harris et al. localized the CDK2-interacting region to residues 119-160 (Harris et al., 2001), a segment that also appears to be involved in p21 binding. Harris et al. suggested that C/EBPα stabilizes the CDK2-p21 inhibitory complex, whereas Wang et al. found that the activity of free CDK2/CDK4 can be inhibited by C/EBPα (Wang et al., 2001). The latter group also detected C/EBPα-CDK2/CDK4 interactions in cells by co-immunoprecipitation but did not observe p21 in this complex. In addition, they showed that over-expression of CDK2 overcomes C/EBPα-mediated proliferation arrest in HT1 fibrosarcoma cells. Clearly, these two reports contain discrepancies with respect to the CDK-binding domains identified and the precise target for inhibition by C/EBPα (CDKs alone or p21-CDK2 complexes), but there is no obvious explanation for the conflicting results.

The inhibition of CDK2/CDK4 clearly does not require DNA binding, and this conclusion is supported by the observation that DNA-binding-defective C/EBPα mutants still slow cell-cycle progression (Wang et al., 2001; Wang et al., 2003). Nevertheless, the underlying evidence is based on overexpression studies, which raises the possibility of experimental artifacts. High levels of C/EBPα in the cell could sequester and inactivate an interacting protein(s), which might not occur under physiological conditions. Therefore, a more definitive test of the CDK inhibition mechanism would be to use a gene-knockin approach to examine the effects of CDK-binding site mutations in vivo to circumvent spurious results resulting from supraphysiological levels or inappropriate timing of C/EBPα expression. Indeed, a C/EBPα mutant lacking the CDK-interaction region apparently produces no overt phenotype when integrated into the mouse germline (Nerlov, 2004).

The RB family of proteins (pRB, p107 and p30) plays a key role in controlling cell proliferation and suppressing tumorigenesis. RBs have overlapping as well as distinct functions that depend in part on the cellular context, and they are also differentially expressed during the cell cycle (Classon and Harlow, 2002). RB proteins function mainly through their interactions with E2F transcription factors, forming complexes that repress transcription of S-phase genes. p107 and p130 preferentially associate with E2F4 and E2F5, whereas pRB can bind to all the E2Fs (Classon and Harlow, 2002). Analysis of cells lacking various combinations of RB genes reveals that all three proteins contribute proliferative constraints in primary fibroblasts (Dannenberg et al., 2000; Sage et al., 2000).

Darlington and coworkers have proposed that C/EBPα directly modulates RB-E2F complexes involved in regulating cell-cycle progression. They first showed that C/EBPα disrupts E2F-p107 complexes that are present during S-phase and are associated with proliferating cells (Timchenko et al., 1999a). In wild-type day 18 mouse embryos, E2F-p107 complexes in liver that are present at earlier stages disappear, and these are also absent in newborn animals. However, in C/EBPα^–/– mice, E2F-p107 complexes remain detectable until birth. The addition of purified C/EBPα to liver nuclear extracts disrupts p107-containing E2F complexes, and a synthetic peptide corresponding to a small region of the C/EBPα transactivation domain that shares sequence similarity with E2F also displays this activity. C/EBPα might thus block the association between E2F and p107. In a second study, Darlington and coworkers found that ectopic C/EBPα expression in 3T3-L1 pre-adipocytes increases the abundance of E2F-p130 complexes (Timchenko et al., 1999b). They postulated that this response involves the previously reported induction/stabilization of p21, which inhibits CDK activity.

From these studies, the authors suggested that C/EBPα alters the pattern of RB-E2F associations in a manner that promotes cell-cycle withdrawal (i.e. reduced numbers of p107 complexes and increased numbers of p130 complexes). However, given current information, it seems unlikely that E2F-p107 stimulates growth or that disruption of this complex would decrease cell proliferation. Indeed, the opposite might be expected given the growth properties of cells lacking p107 and other RB-family members, which reveal antiproliferative functions for p107 (Dannenberg et al., 2000; Sage et al., 2000). To date, the E2F-p107 disruption mechanism for C/EBPα-mediated growth inhibition remains unconfirmed, although the idea that C/EBPα directly or indirectly promotes the formation of growth-inhibiting E2F-RB complexes or enhances their activity remains a viable hypothesis.

C/EBPα might also suppress cell proliferation through interactions with free E2F, as opposed to regulating RB-E2F complexes. Using an inducible C/EBPα system, Slomiany et al. (Slomiany et al., 2000) showed both that C/EBPα inhibits proliferation of murine fibroblast lines and that C/EBPα is present in a complex that binds to E2F sites in genes such as dihydrofolate reductase (DHFR) and E2F-1 that are upregulated during the G1-S transition. However, this binding appears to be indirect, since purified recombinant C/EBPα does not bind to E2F site probes. C/EBPα also represses transcription from reporter constructs containing the DHFR or E2F-1 promoters or an artificial E2F-driven promoter. A similar study demonstrated that C/EBPα can repress transcription from the Myc promoter, which also contains an E2F-binding element, and it was suggested again that C/EBPα acts indirectly through the E2F site (Johansen et al., 2001). C/EBPα might thus associate with E2F complexes and convert them into repressors capable of inhibiting S-phase gene transcription. In this model (as well as in the p21 and CDK2/CDK4 inhibition mechanisms), the ability of C/EBPα to bind DNA is irrelevant to its growth arrest functions.

Porse et al. (Porse et al., 2001) have provided further support for the E2F co-repression model, delineating C/EBPα sequences required for repression of E2F-driven transcription. Their analysis identified transactivation element I (TE-I) at the N-terminus as a critical inhibitory determinant. They also investigated basic region residues that are predicted to face away from DNA (on the basis of molecular modeling using crystal structures for the bZIP proteins Fos, Jun and GCN4), reasoning that these sequences might mediate protein-protein interactions that are important for C/EBPα activity. Two basic region mutants (BRMs), BRM-5 (Y285A) and BRM-2 (I294A/R297A), show decreased repression of E2F-driven transcription compared with wild-type C/EBPα. These also exhibit diminished antiproliferative activity in NIH 3T3 cells and reduced ability to induce differentiation of pre-adipocytes. A gene-knockin approach revealed that mice homozygous for BRM-2, BRM-5, or a third mutant (BRM-1) that shows unimpaired activity in cell-based assays, survive to adulthood and have normal levels of hepatic glycogen and gluconeogenic enzymes. However, the BRM-2- and BRM-5-knockin animals display significant reductions in white adipose tissue and an absence of differentiated neutrophils, whereas other hematopoietic lineages develop normally. EMSA analysis of liver nuclear proteins using a C/EBP site probe showed detectable C/EBPα DNA-binding activity in all genotypes. However, a C/EBPα-containing complex present in wild-type extracts that binds to an E2F site from the DHFR promoter is not observed in extracts from the BRM-2 and BRM-5 mutants. The basic region residues mutated in BRM-2 and BRM-5 thus might mediate binding of C/EBPα to E2F and this association could control the differentiation and cell-cycle-exit functions of C/EBPα. These results support the findings of Slomiany et al. and Johansen et al., which suggest that C/EBPα inhibits expression of E2F-regulated genes through direct interactions with E2F transcription factors (Slomiany et al., 2000; Johansen et al., 2001).

Other observations raise questions about this seemingly straightforward model. In the crystal structure of the C/EBPα bZIP domain bound to DNA (Miller et al., 2003), the DNA-contacting residues differ in several respects from those predicted by the molecular modeling of Porse et al. (Porse et al., 2001). In particular, Tyr285 (the residue mutated to alanine in BRM-5) makes critical contacts with other basic region residues that stabilize the DNA-protein interface (Fig. 4). Tyr285 also makes a phosphate contact with the DNA backbone. Importantly, the Y285A (BRM-5) protein exhibits strongly decreased affinity for C/EBP sites in vitro and significantly reduced ability to transactivate a C/EBP reporter construct (Miller et al., 2003). BRM-5 is thus a hypomorphic allele and the mutant protein has intrinsically lower DNA-binding activity than wild-type C/EBPα. The phenotypes of BRM-5 mice could therefore be explained by impaired interaction of C/EBPα with target promoters to which it binds directly.

Other studies have analyzed the ability of various C/EBPα mutants to induce granulocytic differentiation when expressed in myeloid precursor cells, as well as their effects on interactions with E2F-1. BRM-2 fails to promote myeloid differentiation, which is similar to the phenotypes of DNA-binding-defective basic region mutants and the p30 isoform lacking the N-terminal TAD sequences (Keeshan et al., 2003; Wang et al., 2003). However, BRM-2 is similar to wild-type C/EBPα in its ability to associate with E2F-1 in co-immunoprecipitation assays. This suggests that the defect in BRM-2 might not involve impaired binding to E2F (Keeshan et al., 2003). Although the latter finding does not rule out the C/EBPα-E2F interaction model, it does raise questions about the molecular mechanism by which C/EBPα communicates with E2F to regulate differentiation and cell-cycle arrest.

The BRM mutations might have effects on DNA binding that are not apparent from in vitro binding assays, especially if the major C/EBPα complexes detected in cell extracts are heterodimers containing other C/EBP proteins, such as C/EBPγ (Parkin et al., 2002). In such cases, one subunit would have normal binding affinity, partially overcoming the effect of the mutation on DNA binding. However, C/EBPα homodimers (which may have critical biological roles) would exhibit more severely impaired binding. Moreover, the fact that in vitro binding assays are generally performed under conditions of excess DNA probe might mask minor but important differences in DNA-binding affinities. In view of these potential complications, it will be instructive to determine whether the in vivo binding of BRM mutant proteins to chromatin targets is diminished (e.g. using ChIP assays).

No definitive evidence for C/EBPα-E2F complexes bound to E2F sites via the E2F moiety has been reported to date. Biochemical characterization of such a complex, including identification of the E2F-family members involved, would greatly strengthen the model, as would further mutational studies of the E2F and C/EBPα proteins to identify sequences that mediate their interaction. Also, it is possible that C/EBPα-binding sites are present in some S-phase gene promoters and these might mediate the antiproliferative effects of C/EBPs. We recently identified a C/EBP site in the murine DHFR promoter juxtaposed to the E2F element that is required for transcriptional repression by C/EBPβ in reporter assays (T. Sebastian, S. Thomas, J. Sage and P.F.J., unpublished). Further investigation should reveal whether other S-phase genes also contain previously unrecognized C/EBP-binding sites.

The C/EBPα-E2F interaction model predicts that RB proteins are dispensable for C/EBPα-induced proliferation arrest, whereas current dogma holds that RB-E2F complexes are the ultimate effectors of most, if not all, cellular pathways that regulate G1 arrest (Sage et al., 2000). Recent observations from our laboratory indicate that cell-cycle arrest induced by C/EBPα or C/EBPβ in primary fibroblasts requires RB-E2F complexes. MEFs lacking all three pocket proteins (i.e. triple-knockout, or TKO, cells) do not undergo growth arrest following ectopic expression of either C/EBP protein, whereas wild-type cells cease to proliferate (T. Sebastian, S. Thomas, J. Sage and P.F.J., unpublished). This is consistent with earlier studies showing that C/EBPβ physically and functionally interacts with pRB (Chen et al., 1996a; Chen et al., 1996b). Moreover, expression of a dominant-negative E2F-1 protein (i.e. lacking transactivation and RB-binding sequences) in wild-type MEFs similarly disrupts C/EBP-induced proliferation arrest. Interestingly, in cells lacking RB or E2F function, over-expression of C/EBP stimulates, rather than inhibits, cell proliferation. These observations demonstrate a clear requirement for RB proteins and E2F in C/EBP-induced cell-cycle arrest and, as such, raise questions about the C/EBP-E2F co-repressor model, at least in terms of an RB-independent mechanism. It is also possible that the RB-E2F-dependent C/EBPα effector pathway is cell specific and that different mechanisms of C/EBPα-mediated cell-cycle arrest operate in non-fibroblastic cells.

Additional clues as to the mechanism of C/EBPα-induced growth arrest have come from studies to examine cellular pathways required for cell-cycle exit. A combination of biochemical experiments and genetic studies indicates that the SWI/SNF chromatin-remodeling complex (Eberharter and Becker, 2004) is necessary for C/EBPα-regulated cell-cycle exit. Pedersen et al. (Pedersen et al., 2001) found that the TE-III region of C/EBPα mediates docking with the SWI/SNF complex and that SWI/SNF binding is necessary for C/EBPα to induce adipocyte differentiation and to activate expression of adipocyte- or myeloid-specific target genes. Subsequently, Iakova et al. (Iakova et al., 2003) reported that livers of old animals contain a high-molecular-weight C/EBPα complex that contains E2F4, Rb and the Brm (Brahma) subunit of the SWI/SNF complex. This complex is less abundant in livers of young mice and its appearance seems to correlate with the impaired regeneration of aging liver tissue following partial hepatectomy. The C/EBPα-E2F4-Rb-Brm complex is associated with the Myc promoter in vivo and may confer a stable form of repression on Myc and other growth-promoting genes that is extremely difficult to reverse. This might prevent hepatocytes in older animals from undergoing proliferation and regeneration following liver damage. Irreversible cell-cycle arrest is a defining feature of senescent cells, whose numbers increase during aging. Thus, it is tempting to speculate that C/EBPα plays an important role in establishing the senescent state in cells such as hepatocytes.

Muller et al. (Muller et al., 2004) expressed C/EBPα in human cell lines that lack endogenous Brm. These lines cannot undergo C/EBPα-induced cell-cycle exit; however, expression of Brm restores this ability. In addition, siRNA-induced ablation of endogenous Brm in a SWI/SNF-positive cell line blocks the antiproliferative effects of C/EBPα. These results, together with the previous studies, establish a clear requirement for the SWI/SNF chromatin-remodeling activity in C/EBPα-dependent cell-cycle arrest. The involvement of chromatin-remodeling activity tends to support a transcriptional mechanism for C/EBPα action. The antiproliferative activity of pRB likewise involves SWI/SNF function (Muchardt and Yaniv, 2001), lending further credence to the idea that C/EBPα acts in combination with RB proteins to impose G1 arrest.

Given its strong antiproliferative activity, C/EBPα is likely to be regulated dynamically in cells. Wang et al. (Wang et al., 2004) have observed that C/EBPα levels remain relatively high in proliferating hepatocytes following partial hepatectomy, as well as in liver tumor cells, indicating that the anti-mitotic effects of C/EBPα might be suppressed under certain conditions. A putative phosphorylation site in murine C/EBPα, Ser193 (Fig. 2), appears to be essential for its growth arrest activity. Mutational analysis showed that this residue is required for C/EBPα to bind to CDK2/CDK4 and also to Brm. Signaling through the phosphoinositide 3-kinase (PI3K)-Akt pathway induces dephosphorylation of Ser193, probably through increased nuclear accumulation of the PP2A phosphatase. Ser193 dephosphorylation thus disengages two major C/EBPα growth arrest effector pathways. A mutant containing the S193A substitution still activates transcription in reporter assays, again suggesting that the growth arrest and transcriptional activation functions of C/EBPα are separable activities. At first glance, these results appear to contradict the findings showing that SWI/SNF binding is required for C/EBPα to induce transcription of endogenous differentiation-specific genes. However, transcriptional activation of chromatin-embedded genes is likely to require chromatin remodeling, whereas this may be dispensable for transactivation in transient reporter assays. It will be informative to test whether the S193A mutant can induce differentiation in adipocytes and myeloid cells.

Previous work showed that insulin-mediated activation of the PI3K-Akt pathway in adipocytes induces C/EBPα dephosphorylation on Thr222 and Thr226, substrates for glycogen synthase kinase 3 (GSK3) (Ross et al., 1999). GSK3 activity is suppressed by engagement of the insulin-PI3K-Akt pathway, which also activates the PP1 and PP2A phosphatases. Both reduced GSK3 activity and enhanced phosphatase activity may therefore contribute to C/EBPα dephosphorylation; however, the functional consequences of phosphorylation/dephosphorylation on Thr222 and Thr226 are unknown. The possibility of cross-talk between the GSK3 sites and Ser193 is attractive, especially since the PI3K/Akt pathway impinges on both targets, but this idea lacks experimental evidence. Since the kinase that modifies Ser193 is unknown, identification of this protein will be an important advance in our understanding of the pathways that modulate C/EBPα activity.

Because PI3K-Akt signaling is associated with hepatocyte proliferation during liver regeneration and occurs in hepatoma cells, Wang et al. (Wang et al., 2004) have suggested that C/EBPα deactivation is a major target of this pathway in liver tissue. If this is the case, strategies aimed at reactivating C/EBPα by disrupting this post-translational inhibitory mechanism could be another promising avenue for cancer therapy.

Another phosphoacceptor in C/EBPα, Ser21, is a direct substrate of ERK1/2 (Ross et al., 2004). Phosphorylation of Ser21 blocks the ability of C/EBPα to induce granulopoiesis in bipotential myeloid progenitor cell lines but does not affect monocyte differentiation or adipogenesis. This modification also appears to induce a conformational shift in C/EBPα dimers. The effect of Ser21 phosphorylation on cell proliferation has not been described, and it will be interesting to determine its role, if any, in cell-cycle arrest and to elucidate the molecular mechanism by which it affects lineage-specific cellular differentiation.

Although C/EBPα has been the most intensively studied, several other family members also exhibit antiproliferative activity.

Forced expression of C/EBPβ in HepG2 hepatocarcinoma cells arrests the cells at or near the G1-S boundary (Buck et al., 1994). This effect requires both the bZIP domain and the N-terminal transactivation sequences, and is not reproduced by the transcriptionally inert LIP isoform. Moreover, C/EBPβ-knockout mice display a lymphoproliferative disorder, which suggests that C/EBPβ inhibits expansion of the lymphoid cell compartment (Screpanti et al., 1995). Several observations also indicate an antiproliferative function for C/EBPβ in epidermal keratinocytes: (1) expression of C/EBPβ in BALB/MK2 keratinocytes inhibits colony formation; (2) mice lacking C/EBPβ exhibit mild epidermal hyperplasia; and (3) keratinocytes derived from C/EBPβ mutant mice display partially defective Ca²⁺-induced growth arrest in vitro (Zhu et al., 1999). Expression of the differentiation markers keratin 1 (K1) and K10 is also reduced in these cells, indicating that C/EBPβ regulates proliferation arrest as well as differentiation-specific genes. Thus, C/EBPβ contributes to growth arrest and differentiation of keratinocytes, which is similar to the dual functions of C/EBPα in terminally differentiating cells.

Ectopic expression of C/EBPβ in primary fibroblasts induces cell-cycle exit through a mechanism requiring RB-E2F activity (T. Sebastian, S. Thomas, J. Sage and P.F.J., unpublished). C/EBPβ-arrested fibroblasts exhibit morphological features that are indicative of senescent cells. A similar senescent state is also provoked by over-expression of Ras^V12 and other oncogenes in primary, non-immortalized cells (Lin and Lowe, 2001; Serrano et al., 1997). Ras^V12-expressing primary cells enter an irreversible G1 arrest that requires induction of the p16^INK4a-RB and p19^ARF-p53 tumor suppressor pathways. This oncogene-induced, `premature' senescence is thought, like apoptosis, to provide a barrier to neoplastic transformation and cancer. Interestingly, Ras^V12-expressing C/EBPβ^–/– MEFs fail to senesce and instead continue to proliferate, although (in contrast to p19^ARF- or p53-deficient cells) they are not fully transformed (T. Sebastian, S. Thomas, J. Sage and P.F.J., unpublished). C/EBPβ seems to function in concert with RB-E2F to arrest cells at the G1-S boundary, probably by repressing S-phase genes. Thus, C/EBPβ is an important regulator of cell-cycle exit/senescence induced by Ras^V12 in primary cells. Post-translational activation of C/EBPβ by Ras-stimulated kinases (Hanlon and Sealy, 1999; Mo et al., 2004; Nakajima et al., 1993; Shuman et al., 2004) may be one pathway by which oncogenic Ras^V12 signaling is linked to premature senescence.

Growth arrest induced by C/EBPβ is highly context specific, because in several cases C/EBPβ displays growth-promoting activity. For example, mammary epithelial cells (MECs) from C/EBPβ^–/– female mice have a proliferation defect that leads to impaired ductal morphogenesis and a failure to lactate (Robinson et al., 1998; Seagroves et al., 1998). Conversely, ectopic C/EBPβ expression in a human MEC cell line (MCF10A) induces hyper-proliferation and the cells acquire a partially transformed phenotype (Bundy and Sealy, 2003). A growth-stimulatory effect of C/EBPβ is also observed in macrophage tumor cells (Wessells et al., 2004). Interestingly, deletion of the CEBPB gene renders mice totally resistant to carcinogen-induced skin tumor development (Zhu et al., 2002), and low levels of C/EBPβ expression enhance, rather than inhibit, Ras^V12-induced focus formation in NIH 3T3 cells (Shuman et al., 2004; Zhu et al., 2002). A data-mining approach has also associated C/EBPβ expression with cyclin-D1-dependent tumors (Lamb et al., 2003). Hence, C/EBPβ can function as a pro-oncogenic transcription factor that promotes proliferation and/or survival of some tumor cells. How it stimulates mitotic growth and why it elicits completely opposite effects on proliferation in different cellular contexts are intriguing questions for future investigation.

DeWille and coworkers have described a role for C/EBPδ in G0 arrest of MECs. Proliferating HC11 or COMMA D mammary cells express low levels of C/EBPδ, but when they undergo G0 arrest in response to serum withdrawal, C/EBPδ mRNA and protein levels increase significantly (O'Rourke et al., 1997; O'Rourke et al., 1999). The effect is specific to G0 MECs, since C/EBPδ levels do not increase in other cells, nor is induction observed when MECs arrest at other stages of the cell cycle. Expression of C/EBPδ antisense RNA inhibits endogenous C/EBPδ expression and prevents the cells from entering G0 upon serum withdrawal. Further experiments have implicated the transcription factor STAT3 in activating CEBPD gene transcription in response to low levels of serum (Hutt et al., 2000). A cytokine or related factor, acting in an autocrine manner, might therefore control CEBPD induction by activating Jak/STAT3 signaling. Indeed, oncostatin M induces growth arrest in MECs by a C/EBPδ-dependent mechanism (Hutt and DeWille, 2002).

C/EBPδ also regulates proliferation of MECs in vivo. C/EBPδ is induced during mouse mammary gland involution, which is the period following the end of lactation when extensive tissue remodeling and regression occur (Gigliotti and DeWille, 1998). Female C/EBPδ-knockout mice reproduce and lactate normally (Sterneck et al., 1998); however, nulliparous mutant animals display increased mammary ductal growth and branching, and show a higher epithelial bromodeoxyuridine (BrdU) labeling index than wild-type controls (Gigliotti et al., 2003). Involution and associated cellular apoptosis was not affected, although other work shows that these responses are delayed in C/EBPδ^–/– animals (E. Sterneck, personal communication). The different results might reflect the fact that C/EBPδ^–/– mice of different strain backgrounds were used in the two studies. The involvement of C/EBPδ in control of MEC proliferation raises the possibility that C/EBPδ functions as a tumor suppressor. It will therefore be interesting to determine whether tumor-associated C/EBPδ mutations exist and contribute to the transformed phenotype of cancer cells.

C/EBPϵ is expressed exclusively in hematopoietic cells and their progenitors. C/EBPϵ-null mice are viable but lack functional neutrophils and eosinophils, and eventually develop myelodysplasia (Verbeek et al., 2001; Yamanaka et al., 1997). Conversely, forced expression of C/EBPϵ induces differentiation of promyelocytic leukemia cells (Lekstrom-Himes, 2001; Truong et al., 2003). These findings indicate a role for C/EBPϵ in differentiation and proliferation arrest of myeloid progenitors. The antiproliferative activity of C/EBPϵ might involve E2F-RB, since interactions with E2F1 and pRB have been observed and C/EBPϵ can repress transcription of E2F targets such as Myc (Gery et al., 2004). C/EBPϵ thus mimics many of the functions of C/EBPα and may provide a second pathway by which terminal granulocytic differentiation is implemented (Zhang et al., 2002).

The antiproliferative role of C/EBPα is incontrovertible and considerable effort has been applied towards establishing the underlying mechanism(s). Nevertheless, a number of uncertainties and inconsistencies remain, such as whether DNA-binding activity is required. Some of the discrepancies could arise from the use of different protocols to assay cell growth arrest. For example, transient C/EBPα over-expression followed by colony formation assays could yield results different from those generated by experiments using inducible C/EBPα systems in which proliferation arrest is assayed by growth curves or analysis of DNA content. Another complication arises in interpreting cell-cycle arrest experiments because the cycle consists of a series of sequential, dependent events, which make it difficult to distinguish direct from indirect effects. For instance, if an over-expressed protein causes arrest by repressing S-phase genes, it may nevertheless reduce the activity of G1 cyclin-CDKs because the cells are prevented from progressing to the next cycle. The reverse can also occur – for example, C/EBPα might primarily function through direct inhibition of CDK activity, but the cells would still require RB to implement G1 arrest because RB is a major target of CDKs in the regulation of G1-S progression. Therefore, although over-expression experiments and analysis of cells lacking known cell-cycle regulators are useful for identifying the pathway(s) in which C/EBPα functions, these approaches do not necessarily reveal the primary target. Ultimately, a combination of biochemical, cellular and genetic experiments will be necessary to establish definitively the mechanism by which C/EBPα elicits cell growth arrest.

With the importance of C/EBP-family members in the control of cell proliferation now well established, the next challenge is to acquire a more comprehensive description of C/EBP targets and biological activities. The availability of increasingly sophisticated tools to identify target genes, binding sites in chromosomal DNA, protein-protein interactions and protein modifications, combined with powerful genetic and computational methods, should soon provide significant advances in our knowledge of how C/EBP proteins regulate cell proliferation and differentiation. Thus, we can anticipate the eventual characterization of `regulons' for each C/EBP-family member in different cell types and cellular states (e.g. proliferating versus arrested, normal versus neoplastic), as well as networks of protein interactions and regulatory pathways that control their activities. This information should yield answers to several important questions. Do different C/EBP-family members regulate distinct sets of target genes in the same cell and, if so, how is this specificity achieved? What combinatorial relationships exist between C/EBPs and other transcription factors to establish cell-specific transcriptional programs? What assemblies of co-activators or co-repressors are involved in determining positive or negative regulation of target genes? How do C/EBP proteins contribute to disease states such as cancer and inflammation? The challenge ahead for researchers in the field is to adopt these new research tools and technologies, and also to develop cooperative networks so that the large, complex data sets generated from these experiments can be shared and thus used to maximum effect.

The author is indebted to Alan Friedman, Thomas Sebastian, Hyung-Chan Suh and Esta Sterneck for critical comments on the manuscript.