|Home | About | Journals | Submit | Contact Us | Français|
Cell cycle progression in mammalian cells from G1 into S phase requires sensing and integration of multiple inputs, in order to determine whether to continue to cellular DNA replication and subsequently, to cell division. Passage to S requires transition through the restriction point, which at a molecular level consists of a bistable switch involving E2Fs and pRb family members. At the G1/S boundary, a number of genes essential for DNA replication and cell cycle progression are upregulated, promoting entry into S phase. Although the activating E2Fs are the most extensively characterized transcription factors driving G1/S expression, LSF is also a transcription factor essential for stimulating G1/S gene expression. A critical LSF target gene at this stage, Tyms, encodes thymidylate synthetase. In investigating how LSF is activated in a cell cycle-dependent manner, we recently identified a novel time delay mechanism for regulating its activity during G1 progression, which is apparently independent of the E2F/pRb axis. This involves inhibition of LSF in early G1 by two major proliferative signaling pathways: ERK and cyclin C/CDK, followed by gradual dephosphorylation during mid- to late-G1. Whether LSF and E2F act independently or in concert to promote G1/S progression remains to be determined.
Stringent regulation of the progression of mammalian cells from the quiescent state, G0, to cellular DNA replication is essential for normal development and tissue homeostasis. Exit from the resting state and subsequent cell cycle progression involve integration of multiple environmental and internal cues. The critical decision of whether or not to commit to DNA replication is made at the restriction point,1 based on this integration of signaling information. Recently the restriction point was demonstrated to be constituted by a bistable switch, centered on regulation of the Rb/E2F pathway by the G1 cyclin D-dependent kinases.2 Once this decision point is traversed, activating E2F family members (E2F1, E2F2, E2F3a)3,4,5 drive transcription of genes at the G1/S boundary, leading to expression of genes essential for continued cell cycle progression through S phase. E2F target genes include those encoding the DNA replication machinery, deoxynucleotide biosynthetic enzymes, and key cell cycle regulatory proteins.6,7,8
Although E2Fs are central players in the progression from G0 into S phase, they are not the sole transcription factors involved. Many of these other players either activate the Rb/E2F pathway (i.e. c-myc) or directly interact with specific E2Fs to regulate gene expression at particular E2F target promoters (i.e. TFE3).9,10 The focus of this perspective is on yet another transcription factor, LSF (also commonly named CP2 and encoded by the TFCP2 gene11), that is critical for appropriate progression through the G1/S transition, but whose regulation is seemingly independent of the Rb/E2F axis. We recently demonstrated that regulation of LSF activity from G0 to S surprisingly involves rapid, staged, inhibitory phosphorylation in early G1, followed by a gradual dephosphorylation through mid- to late-G1.12 These post-translational modifications do not involve the canonical G1 cyclin-dependent kinases (CDKs), but instead require the MEK/ERK pathway and cyclin C-driven early G1 CDK activity. Below, we discuss the ramifications of this novel regulatory mechanism during G1 that leads to a time delay of LSF activation. Furthermore, we suggest models in which activation of LSF represents either a parallel sensing pathway for efficiently transitioning from G1 into S phase, or a pathway that exhibits crosstalk with the pRb pathway downstream of the restriction point.
Several mammalian transcription factors involved in cell proliferation signaling, including both E2F (Adenovirus E2 promoter Factor)13 and LSF (Late SV40 Factor)14 were initially identified due to their abilities to stimulate transcription from promoters of DNA viruses. These “tumor viruses” are reliant on inducing host cell DNA replication, so that they can capitalize on host resources (e.g. nucleotides) and in some cases DNA synthesis machinery, to replicate their own viral genomes. Evolutionarily, it therefore is reasonable that the viruses would take advantage of host transcription factors activated during this induction of proliferation for efficient expression of viral mRNAs. With this in mind, we previously tested whether LSF was involved in activation of cellular genes after induction of cellular proliferation. We demonstrated that the LSF family of transcription factors plays a key role in activating the gene encoding thymidylate synthetase (Tyms) at the G1/S transition.15
Thymidylate synthetase catalyzes the conversion of dUMP to TMP, and is the rate limiting enzyme in biosynthesis of TTP.16 Given this central role of an LSF target gene in deoxynucleotide biosynthesis, we also tested and demonstrated an essential role for LSF in cell cycle progression from G1 through S phase, in both mouse fibroblasts and human prostate cancer cells. Inhibition of LSF activity by two distinct mechanisms resulted in either arrest at the G1/S transition,17 or in apoptosis after entry into S phase.15 How LSF responds to cell proliferation and cell cycle signaling in order to specifically activate Tyms at the G1/S transition has been a focus of recent studies in our laboratory. LSF appears not to be regulated by altering cellular expression at either the mRNA or protein level. LSF mRNA levels do not generally fluctuate, and its mRNA has even been suggested to be a normalization control in gene expression profiling.18 LSF is expressed ubiquitously in cell lines and in the developing mouse,19,20 and in particular for the point of this discussion, protein levels are unaltered during cell cycle progression.12 Consistent with its constant protein level, the turnover of LSF is not particularly rapid; following cycloheximide addition to block protein translation (Fig. 1), the observed halflife for LSF is intermediate between those for short-lived proteins (e.g. cyclin D1) and very stable proteins (e.g. β-actin). Finally, in almost all cell types examined, LSF is localized constitutively to the nucleus. Two reported exceptions occur in differentiation of specific cell lines in vitro: a mouse erythroid cell line21 and a rat neuronal cell line.22 However, in nondividing primary rat cortical neurons we have observed LSF to be predominantly nuclear (S.-Y. Oh, L. Owens, S. Russek, U. Hansen, C. Abraham, unpublished results). These general characteristics of LSF rule out the activation of LSF at the G1/S transition via upregulation of expression of the LSF gene (TFCP2), stabilization of LSF from degradation, or translocation of cytoplasmic LSF into the nucleus. As a consequence, our investigations on the regulation of LSF activity have focused instead on post-translational modifications in response to proliferative signaling transduction pathways.
The DNA-binding activity of the LSF/CP2 family of transcription factors is atypical, in that its DNA recognition sequence spans at least 15 bp, and comprises a directly repeated motif separated by a variable linker region.11 Furthermore, whereas LSF exists largely as a dimer in solution,23 tetramerization is required for stable binding to DNA.24,25 Although no structural data are presently available for the LSF/CP2 transcription factor family members, or for the related Grainyhead (Grh) family members that bind DNA as dimers, our recent evolutionary comparisons (Traylor-Knowles N, Hansen U, Dubuc T, Martindale MQ, Kaufman L, Finnerty JR, manuscript submitted) combined with de novo protein folding prediction of others26 have merged to provide a solid working model for the LSF domain structure (Fig. 2).
Previous phylogenetic comparison of members of the LSF/CP2 family demonstrated evolutionary conservation within two broad segments of the protein – a central and a C-terminal region.27,28 Functionally, these two conserved portions corresponded to the DNA-interaction and oligomerization regions, respectively, as defined by activity assays of LSF truncation mutants.24 Our recent phylogenetic analysis not only includes a broader range of species, but also compares the sequences on a finer scale, using a motif analysis algorithm (Fig. 2). Using this methodology, structural units can also be more readily discerned.
By de novo computational protein folding analysis and comparison with known protein structures, Wyrwicz and colleagues26 predicted three structural domains in LSF. Importantly, the region in LSF that interacts with DNA is predicted to resemble the structure of the p53 DNA-binding domain. The p53/p63/p73 family of proteins also bind DNA as homotetramers. The C-terminal region, designated as the dimerization domain due to its requirement for oligomerization and shared conserved peptide motifs with the dimeric Grh subfamily, is predicted to be a ubiquitin fold-like domain. Finally, the designated tetramerization domain, which is also required for oligomerization but is unique to the LSF subfamily, is predicted to be a SAM domain. SAM domains often comprise protein-protein interaction domains, consistent with this assignment.29 Strikingly, although no previous experimental data was used in derivation of the structural predictions, the structural model is completely consistent with all previous mutational analysis of LSF structure/function relationships. Confirmatory experimental data include the following. 1) Truncation mutants of LSF identified residues 448 to 502 to be involved in LSF binding DNA in dimeric units,24 which corresponds to a portion of the predicted structural ubiquitin-like fold of residues 424 to 502.26 2) C-terminal truncation mutants of LSF indicated partial loss of oligomerization between residues 383 and 403, and complete loss when the truncation was extended to residue 377. Truncation to 383 removes just 6 amino acids of the predicted SAM domain, consistent with a partial loss of function, whereas the extended truncation leading to complete loss of function substantially removes significantly more of this protein-protein interaction domain, which is predicted to correspond to residues 326 to 389. 3) N-terminal truncation mutants of LSF indicated loss of DNA-binding activity when residues between 64 and 144 were removed. Consistent with these data, the DNA-binding domain is predicted to extend from residues 67 to 260. 4) Finally, the dominant negative mutant of LSF that we generated consists of two amino acid substitutions in the DNA-binding domain, the critical one being a K to E alteration at amino acid 236.30,31 Remarkably, the K236 in LSF corresponds in the predicted structure to R273 in the p53 DNA-binding domain, which is one of a small group of residues that interacts with DNA in the major groove.26
Appreciation of the DNA-binding characteristics of LSF provide a basis on which to model how post-translation modifications could alter LSF activity. In particular, given the requirement for tetramerization of LSF to bind DNA, oligomerization from the dimeric to the tetrameric form provides one attractive target for regulation of LSF.
There is only limited knowledge to date regarding how LSF activates transcription. The N-terminal region of LSF is required for transcriptional activation of synthetic reporter constructs (data cited in refs. 20,32,11). This region interacts with basal transcription factors TBP and TFIIB (Q. Zhu and U. Hansen, unpublished), but no coactivator proteins have been identified so far. As is the case for many other transcription factors, protein modifications might regulate binding to coactivators, either directly or through induction of conformational changes.
In searching for how LSF is regulated such that it stimulates Tyms expression in late G1, we initially anticipated that LSF would be post-translationally modified in late G1, in order to turn on its activity. However, our studies on the regulation of LSF activity in growth-regulated mouse fibroblasts have generated a diametrically opposed model.12 Instead, LSF is phosphorylated by the MEK/ERK pathway immediately (within 5 minutes) after mitogenic stimulation of quiescent cells, at the G0/G1 boundary. The responsible kinase is ERK1 or ERK2, as these enzymes both directly phosphorylate LSF in vitro on the site targeted by this signaling pathway in vivo.33,34 Initially we were surprised that this phosphorylation is quantitative, as demonstrated by total cellular LSF shifting its mobility upon SDS-PAGE analysis to migrate more slowly.33,34 This change in electrophoretic mobility of LSF depends on both the presence of Serine 291 (S291) and the activity of MEK/ERK in the cells. The biological rationale for why all the LSF should be modified by this pathway was elusive, until we demonstrated that phosphorylation of S291 can reduce the affinity of LSF binding to target DNA sites, and can also diminish its transactivation potential.12 In order to efficiently inhibit the function of a cellular protein, it is necessary to target all of the active pool; this is in contrast to the activation of cellular proteins by ERK, where modification of a fraction of the total cellular pool can result in efficient stimulation of activity and downstream biological consequences.
Interestingly, whereas S291 phosphorylation dramatically decreased the binding affinity of LSF to some DNA-binding sites in vitro (e.g. the binding site in the HIV LTR), it only minimally, if at all, affected the affinity of LSF to a high affinity binding site (e.g. its binding site in the SV40 late promoter).35 This suggests the intriguing possibility that some LSF target genes will be more sensitive than others to inhibition by phosphorylation at S291. Once a larger set of target genes has been identified, this hypothesis can be tested.
We recently demonstrated yet another inhibitory modification of LSF in early G1, namely the phosphorylation of Serine 309 (S309). The kinase targeting LSF at this stage proved to be a cyclin C-dependent CDK.12 For some time, it was thought that the only relevant kinase stimulated by cyclin C in vivo was CDK8, which as a complex phosphorylates RNA polymerase II.36 More recently however, cyclin C was shown to have the additional function of activating CDK3 specifically in early G1, phosphorylating pRb and thereby stimulating rRNA sythesis.37 This was demonstrated in a human cell line. In contrast, most laboratory mouse strains and cell lines lack functional CDK3.38 Given the previously documented plasticity in the ability of multiple CDKs to associate with G1 cyclins, particularly in the absence of the cognate CDK,39,40 we hypothesized that another CDK would functionally replace CDK3 in mouse cells by associating with cyclin C in early G1. The closest CDK family member to CDK3 is CDK2,41 and we confirmed the predicted association between cyclin C and CDK2 by co-immunoprecipitation assays of endogenous proteins in staged mouse fibroblast extracts. In addition, coexpression of either CDK3 or CDK2 with cyclin C caused phosphorylation of LSF at S309. Finally, our data from CDK2 knockout mouse embryo fibroblasts suggest that CDK1 can also associate with cyclin C in early G1, but this has yet to be verified. Thus, our results extend the previously appreciated plasticity of CDK-G1 cyclin interactions. To our knowledge, LSF is only the second known substrate, other than pRb, for the early G1 cyclin C/CDK activity.
For LSF, the functional consequence of the early cyclin C-CDK activity, phosphorylation at S309, is inhibition of its ability to activate transcription.12 This parallels the inhibition of transactivation by phosphorylation of LSF at S291. Notably, both sites of phosphorylation lie between the DNA-interaction domain and the tetramerization domain of LSF (Fig. 2). Inhibition could result from decreased DNA-binding and/or decreased binding to coactivators, and may be mediated by conformational changes in the protein (U. Saxena, G.M. Cooper, U. Hansen, manuscript in preparation). Transcriptional inhibition was demonstrated both from synthetic, LSF-driven reporter constructs, and from the endogenous Tyms gene. In particular, when cyclin C and CDK3 were overexpressed from retroviral constructs, stimulation of Tyms expression at G1/S was significantly and specifically diminished. Expression of LSF S309A, which cannot be phosphorylated at S309, fully rescued this expression to normal levels. In contrast, the expression of E2F-regulated genes (Ccne and Mcm3) was unaffected by expression of either cyclin C-dependent kinase or LSF. Thus, in addition to showing the transcriptional consequence of cyclin C/CDK modification of LSF, this result also suggested the independence of G1/S regulation by LSF versus E2F family members (i.e. E2F1 and E2F3).
Our studies have therefore shown that instead of the simple prediction of LSF being able to stimulate expression of G1/S genes via its activation by phosphorylation in late G1, LSF is inhibited by phosphorylation at two sites in early G1, via two of the major mitogenic signal transduction pathways. During the proliferative response, dephosphorylation at these residues occurs in mid- to late-G1, thereby facilitating activation of LSF target gene(s) at the G1/S transition. We propose that these modifications provide a novel delay mechanism, ensuring that LSF is off in G1 to prevent robust induction of G1/S phase genes until the appropriate point. This finding raises the question of how the kinetics of dephosphorylating LSF are regulated. One hypothesis would be that S291 and S309 dephosphorylation occurs via activation of late G1-specific phosphatase(s). An alternative hypothesis is that the gradual dephosphorylation of LSF during G1 progression results from the balance of activities between relevant kinases and relevant phosphatases. The two LSF kinases, ERK and cyclin C/CDK2, both peak in activity in early G1, and then diminish to low or undetectable levels by mid- to late-G1; the LSF phosphatases have yet to be fully elucidated, but need not be regulated in this model. This latter model provides a time-delay mechanism for activation of LSF, depending on whether kinase levels immediately decline or whether their activities are sustained. Such a model would suggest that LSF may remain suppressed during cellular responses to signals known to result in sustained, rather than transient, ERK activation. One notable example of differences in ERK activation kinetics in response to different signaling molecules has been documented in the PC-12 neuronal cell line.42 In response to growth factors that induce proliferation (e.g. EGF), the ERK activation profile is only transient, however in response to NGF, which induces differentiation, the ERK activation profile is sustained. The sustained ERK activity is critical for driving cells down the differentiation, rather than proliferative, pathway.
Although dephosphorylation of LSF during G1 progression is necessary, it is not sufficient to efficiently activate Tyms expression. This conclusion was based on the inability of exogenous, high level expression of LSF S309A to activate Tyms expression prior to the G1/S transition.12 Furthermore, expression of LSF S291A/S309A also does not prematurely induce Tyms mRNA levels (R. Cacioppo and U. Hansen, unpublished observations). The additional molecular switch that allows LSF to activate Tyms expression only at G1/S could involve: 1) induction of additional activating molecules, either another obligatory transcription factor,9,10 or a coactivator,43 much like G1/S E2F transcription factors, 2) derepression by removal of chromatin modifications or a transcriptional repressor at the promoter, and/or 3) yet another modification or conformational change of LSF that is required to generate robust transcriptional activation. In order to determine the overall biological role of LSF during cellular responses to extracellular signaling molecules, it will be essential to determine what signal transduction pathways are required for this subsequent step of activation of LSF target genes at the G1/S transition.
At every turn in our examination of LSF regulation during the transition from G0 through mid- to late-G1 phase, it has been notable that the signaling pathways to which LSF responds are parallel to, rather than connected to, the Rb-E2F axis (Fig. 3). Furthermore, the timing of the regulation is quite distinct. From extensive analyses from many laboratories, it has been shown that the ability of activating E2F's to function during cell cycle progression is directly dependent, first, on expression of cyclin D1 (which requires the MEK/ERK pathway for elevated gene expression) and subsequently on expression of cyclin E.44 Through association with CDK4/6 and CDK2, respectively, active G1 cyclin-CDK complexes are formed that phosphorylate members of the pRb family of proteins.3,45 Phosphorylation of p130 leads to dissociation from repressing E2Fs (E2F4 and 5), localization of these E2Fs to the cytoplasm, and the resulting release of repression of E2F target genes. At the same time, phosphorylation of pRb relieves its direct inhibition of the activating E2Fs. The efficient transcriptional induction at G1/S also requires cell cycle-dependent interaction of activator E2Fs with the coactivator HCF-1.43
In contrast, our laboratory has demonstrated that LSF is negatively regulated by two of the major mitogenic signaling pathways shortly after the G0/G1 transition – the MEK/ERK pathway, and the cyclin C-dependent early G1 CDK2 or CDK3 activity (Fig. 3). LSF is not phosphorylated by cyclin D/CDK complexes at all in vitro.12 Maximal induction of LSF activity in late G1 then requires dephosphorylation at the S309 and S291 sites. The subsequent switch that fully turns on LSF target gene(s) at G1/S has yet to be identified, and may involve additional late G1 alterations in either LSF, collaborating transcription factors, which could include synergy on regulatory regions with E2F family members, or coactivators.
At the current time, the only known G1/S-regulated LSF target gene is Tyms.15 We, of course, anticipate that there are other G1/S-regulated LSF target genes, which need to be identified by global gene expression profiling after inhibition of LSF activity, and/or by global DNA association studies using LSF antibodies in chromatin immunoprecipitation assays. Overall, our current data suggest that the LSF target genes and E2F target genes include completely distinct gene sets, indicating that activation of both pathways should be required in all cells in order to achieve robust cell cycle progression. On other genes, LSF and E2Fs might either redundantly or synergistically regulate expression of overlapping sets of genes. We anticipate that the dependence on LSF may vary widely, depending on its particular target gene. Notably, relative contributions of the LSF and E2F pathways on a particular gene may also prove to be cell type-dependent, reflective of differing pools of signaling molecules and coactivators.
Despite the general lack of knowledge of the LSF gene set, the one characterized gene, Tyms, is critical for cell cycle progression. Indeed, thymidylate synthetase was one of the first, and still a significant, target for cancer chemotherapy. A major unanswered question, which must be addressed to understand the biological role of LSF in responding to extracellular signals, is whether the induction of the G1/S-regulated Tyms gene by LSF is truly independent of and parallel to the E2F induction of G1/S-regulated genes, or if LSF activation is interconnected with the Rb-E2F pathway. We propose that the dephosphorylation at S291 and S309, which constitute a time-delay for LSF activation, will not depend on the restriction point, given the gradual dephosphorylation during cell cycle progression. However, it is critical to determine whether the secondary LSF-dependent activation step at the Tyms regulatory regions is dependent or not on traversal through the restriction point. If not, regulation of LSF may respond to different environmental or intracellular cues than regulation of the Rb-E2F pathway, allowing entry into S phase only when multiple environmental and intracellular signals are appropriate. Alternatively, if LSF activation is downstream of the restriction point, the LSF pathway may represent a synergistic mechanism to that of Rb-E2F that fine-tunes the robustness of entry into S phase.
We are grateful to Geof Cooper for reading and critiquing this manuscript. This research was previously supported by CA-081157 to U.H.