Complex cartography: E2F transcriptional control by Cyclin F and ubiquitin

Abstract

The E2F-family of transcriptional regulators sit at the center of cell cycle gene expression and play vital roles in normal and cancer cell cycles. Whereas control of E2Fs by the retinoblastoma family of proteins is well established, much less is known about their regulation by ubiquitin pathways. Recent studies placed the SCF family of E3 ligases with the F-box protein Cyclin F at the center of E2F regulation, demonstrating temporal proteolysis of both activator and atypical repressor E2Fs. Importantly, these E2F members, in particular activator E2F1 and repressors E2F7 and E2F8, form a feedback circuit at the crossroads of cell cycle and cell death. Moreover, Cyclin F functions in a reciprocal circuit with the cell cycle E3 ligase APC/C, which also controls E2F7 and E2F8. This review focuses on the complex contours of feedback within this circuit, highlighting the deep crosstalk between E2F, SCF-Cyclin F and APC/C in regulating the oscillator underlying human cell cycles.

Control of the eukaryotic cell cycle

The molecular basis of cell proliferation is the cell cycle, a highly regulated series of events during which the genetic information is duplicated and partitioned into two daughter cells. The core components of the cell cycle machinery were originally discovered through genetic studies in yeast and biochemical studies in marine creatures and cell free systems [1–6]. Since these initial studies, it has become well-established that cell cycle progression is driven by the oscillating activity of Cyclin-dependent kinases (CDKs) which phosphorylate a myriad of substrates to coordinate the key molecular events of the cell cycle [7]. The activity of CDKs is positively regulated by “Cyclins” and negatively regulated by Cyclin kinase inhibitors (CKIs) through specific binding interactions. Cyclins and CKIs oscillate in abundance throughout the cell cycle, giving rise to temporal CDK activation profiles and ensuring unidirectionality of the cell cycle. Notably, the expression of Cyclins and CKIs is controlled by both transcription and degradation. Thus, excluding early embryonic cell divisions, the core cell cycle oscillator is timed and regulated by both transcriptional and degradative machinery.

Despite half a century of molecular research, important features of the cell cycle regulatory apparatus continue to emerge, with relevance to cancer biology and treatment. These discoveries reveal how cell cycle effectors are controlled in time and space to execute orderly cell cycle progression and genome maintenance. Here, we highlight recent studies that implicate the Skp1-Cul1-F-box-protein (SCF) ubiquitin ligase family as a key factor in the control of the E2F family of cell cycle transcriptional regulators. We discuss the feedback mechanisms connecting orderly proteolysis and transcription and the implications these discoveries have on the complex molecular circuits underlying the cell cycle.

Counting on cell cycle transcription: E2F1, 2, 3, 4, 5, 6, 7, 8

About one-tenth of the protein-coding transcriptome exhibits oscillatory mRNA expression during the human somatic cell cycle [8–13]. Many of the genes whose expression is required for G1 and S-phase are controlled directly by the E2F family of transcription factors [14–17] . There are eight human E2F family genes, E2F1-E2F8, which encode ten protein products (the E2F3 gene encodes two proteins with distinct functions via alternative promoters [18,19] and two E2F7 isoforms result from alternative splicing [20]). E2F proteins rely on a conserved DNA-binding domain (DBD) to engage their target gene promoters. E2F1-E2F5 contain a transactivation domain that mediates recruitment of transcriptional co-activators or one of three pocket proteins, termed Rb, p107, and p130, which repress E2F activity by masking their transactivation domain [16].

E2F family members are classified as either transcriptional activators or repressors. E2F1, E2F2 and E2F3a are well-established transcriptional activators that drive gene expression programs which dictate cell cycle commitment, orderly transition from G1 to S-phase, and enhanced cellular proliferation [16,21,22]. The importance of activator E2F proteins in proliferation is typified by their control of Cyclin E and Cyclin A [23–25], which promote entry into and progression through S-phase, respectively. Cyclin A/CDK, in turn, forms a negative feedback loop through inhibitory phosphorylation of activator E2Fs beginning in S-phase, thereby preventing their transactivation and association with DNA [26–28]. Thus, Cyclin A/CDK-mediated phosphorylation of activator E2Fs represents one of several reported mechanisms for downregulating E2F-mediated gene expression following entry into S-phase. Interestingly, high levels of E2F1 also promotes apoptosis [29,30]. Thus, the E2F family sits at the crossroads of cell proliferation and death [31].

E2F3b and E2F4-E2F6 belong to the ‘repressive’ E2F branch as they repress the transcription of E2F target genes during quiescence (G0) and early G1 phase (when associated with a pocket protein) [16]. E2F7 and E2F8 are also categorized as ‘repressive’ because of their ability to repress transcription of E2F1 and its target cell cycle genes [20,32,33].

E2F proteins are also subdivided into ‘classical’ (E2F1–6) and ‘atypical’ (E2F7–8) groups based on their structural features [16]. ‘Classical’ E2F members are activated by dimerization partners (DP1, DP2, DP3) [34,35]. ‘Atypical’ E2F members lack the DP dimerization-domain, as well as the transactivation, pocket-protein, and Cyclin A binding-domains. Instead of binding to DP proteins, E2F7 and E2F8 homo- or hetero-dimerize to facilitate DNA binding [20,32,33,36,37].

Unlike classical E2F repressors, expression of the atypical repressors E2F7 and E2F8 correlates with a high rate of proliferation [38,39]. In fact, E2F7/8 are both direct targets of E2F1, and thus, their mRNA and protein products oscillate throughout the cell cycle, beginning in late G1 and early S-phase [20,32,33,37]. E2F7/8 subsequently bind to E2F gene promoters in late S-phase, restraining E2F1 output in the latter phases of the cell cycle. In addition to phosphorylation by Cyclin A/CDK, the expression of E2F7/8 represents a second negative feedback loop whereby E2F1 activates its own inhibitors which ultimately extinguish E2F1-dependent transcription in late S-phase. More globally, the E2F1-E2F7/8 negative feedback balances both pro-proliferative and pro-apoptotic activities of E2F1 in regulating cell cycle and cell death [40–43].

In addition to the E2F family, the B-Myb and FoxM1 transcription factors play an important role in late S-phase and G2/M, where they cooperate to control the expression of key genes needed for entry and progression through mitosis [12,44–48]. Both B-Myb and FoxM1 interact with the MuvB subcomplex to form an active transcription factor unit, which is required for transcription of their target genes. Interestingly, MuvB also interacts with E2F4 and E2F5 during quiescence to repress transcription of E2F, B-Myb, and FoxM1 target genes, inhibiting their transcription. Thus, MuvB plays an important role in both promoting and restraining cell cycle progression, and it is intrinsically linked to the functions of E2F, B-Myb, and FoxM1 transcription factors [49,50].

Altogether, E2F, B-Myb and FoxM1 coordinate the oscillating expression of hundreds of genes required for controlled progression through the cell cycle and the maintenance of genome integrity. For example, the expression of Cyclin E [23,24] and Cyclin A [25] is upregulated by E2F1 in mid-G1-phase and the expression of Cyclin B is upregulated by B-Myb and FoxM1 in G2-phase [9,45]. This tight transcriptional control of gene expression is essential for normal cell cycle progression. Moreover, because these transcription factors play a major role in promoting proliferation, their dysregulation is often associated with cancer etiology. While beyond the scope of this review, numerous other reviews focus on aberrant expression of cell cycle genes in cancer [51–53].

The alphabet of cell cycle ubiquitination: A-P-C-C-S-C-F….

Coupled to the tightly regulated expression of cell cycle genes is the targeted degradation of hundreds of cell cycle proteins by the ubiquitin proteasome system. Ubiquitin is a small protein that is post-translationally conjugated to target proteins via an enzymatic cascade that relies on three enzymes termed E1, E2 and E3 [54,55]. Substrate specificity is designated by E3 ubiquitin ligases, which recognize targets to facilitate their ubiquitination, resulting in the covalent attachment of ubiquitin moieties onto a lysine residue in a substrate protein. The formation of polyubiquitin chains on substrates often serves as a signal that targets substrates to the proteasome, triggering their degradation [56].

The anaphase-promoting complex/cyclosome (APC/C), discovered twenty-five years ago [5,6], is perhaps the most well-characterized E3 ligase involved in eukaryotic cell cycle control. This multi-subunit APC/C ubiquitin ligase becomes active during mitotic metaphase, and it promotes anaphase by triggering the degradation of Cyclin B and Securin [57–61], resulting in a sharp reduction in mitotic CDK activity and dissolution of Cohesin, the molecular glue that holds sister chromatids together. Additionally, APC/C regulates the destruction of dozens of other proteins involved in mitotic signaling, spindle assembly, cytokinesis, transcription, and DNA replication [62–68].

APC/C utilizes substrate receptors that bind target proteins to facilitate their recruitment to the E3 and their subsequent ubiquitination. There are two related APC/C substrate receptors: Cdc20/Fizzy and Cdh1/Fzr1 [69–74]. Cdc20-bound APC/C complexes (APC/C^Cdc20; superscripts denote substrate receptor being used) are active from mitotic metaphase until late anaphase. APC/C^Cdh1 is active from mitotic telophase through the end of G1. The switch from APC/C^Cdc20 to APC/C^Cdh1 depends on a series of coordinated molecular events, beginning with APC/C^Cdc20-mediated degradation of Cyclin B and the resulting downregulation of Cyclin B/Cdk1 activity. This event, in combination with phosphatase activities, releases Cdc20 from the APC/C and allows Cdh1 to bind to the APC/C complex [75–79]. Finally, Cdc20 is ubiquitinated and degraded by APC/C^Cdh1, completing the APC/C^Cdc20 to APC/C^Cdh1 transition and enforcing irreversible mitotic exit [80]. APC/C^Cdh1 remains active throughout G1 phase, where it restrains entry into S-phase [81,82]. Thus, APC/C must be inactivated for cells to begin DNA replication. The inactivation relies on several events including expression of the APC/C inhibitor protein Emi1, degradation of the APC/C E2 enzymes, and phosphorylation and ubiquitination of the substrate receptor/coactivator Cdh1 [77,83–90]. We point interested readers to many excellent reviews on the regulation of APC/C and mitotic exit [83,91–95].

Whereas APC/C activity is restricted to the M and G1-phases, the activity of SCF family E3 ligases is more ubiquitous throughout the cell cycle, playing roles in the G1/S and G2/M transitions [96,97]. The SCF complex consists of an invariable core catalytic module (Skp1-Cul1) which binds one of approximately 70 F-box protein family members in humans [98]. Like Cdc20 and Cdh1, the F-box proteins act as substrate receptors, bridging substrates to the ubiquitin machinery and providing substrate specificity for the SCF ligases [99].

The degradation of many substrates mediated by SCF plays a major role in key cell cycle events. Perhaps the clearest example is the role of SCF in contributing to the activity of CDKs. In humans, SCF complexes bound to the F-box protein Skp2 (SCF^Skp2) ubiquitinate the CKI proteins p21 and p27, resulting in their destruction [100–102]. The degradation of p21 and p27 is particularly important in late G1-phase because it promotes the activation of CDK2 and entry into S-phase [103]. In addition, SCF^Fbw7 serves as a tumor suppressor by negatively regulating the G1/S transition through degradation of Cyclin E [104–107] and other pro-proliferative and oncogenic proteins, including Myc [108,109], Notch [110,111] and Jun [112]. Beyond G1/S, SCF^β-TRCP1/2 contributes to mitotic progression by degrading the APC/C inhibitor Emi1 in G2-phase, allowing for activation of APC/C in mitosis [113,114].

Cyclin F is the founding member of the F-box family of SCF substrate receptors and plays key roles in regulation of the cell cycle [115,116]. Like other cyclins, Cyclin F contains a cyclin domain, and Cyclin F mRNA and protein abundance oscillate throughout the cell cycle [115]. Additionally, like Cyclins A and B, Cyclin F is itself marked for degradation in mitosis and G1-phase by the APC/C, highlighting crosstalk between ubiquitinating enzymes [87]. However, unlike the canonical Cyclins (A, B, E, and D), Cyclin F does not activate a CDK and plays no known role in phosphorylating substrates [117]. Instead, like other F-box proteins, it functions as part of an SCF ubiquitin ligase complex, recruiting substrates to the E3 to enable substrate ubiquitination and degradation. To date, nearly all reported Cyclin F substrates are involved in cell cycle regulation, including centrosome homeostasis (CP110), DNA repair and synthesis (RRM2 and Exo1), DNA replication (CDC6), and mitotic spindle formation (NuSAP1) [118–122]. Importantly, Cyclin F also contributes to APC/C inactivation through degradation of Cdh1, providing an additional layer of regulation to the G1/S transition [87]. This also implies that Cyclin F exists in a double negative feedback loop with APC/C^Cdh1, where Cyclin F is targeted for degradation by APC/C in G1, and Cdh1 is targeted by SCF^{Cyclin F} in S-phase [123]. Taken together, these data highlight the importance of APC/C and SCF ubiquitin ligases in cell cycle control and the role of feedback circuits in shaping E3 activity, substrate receptor abundance, and the ability of specific enzymes to remodel the protein landscape.

Targeted proteolysis in cell cycle transcriptional control: the _SCF-_CyclinF-_E2F words

Understanding the role of E3 ubiquitin ligases in cell physiology requires identification of specific E3 ligase-substrate pairs. Accordingly, several groups, including our own, have attempted to map targets of SCF^{Cyclin F} ubiquitination. In doing so, recent studies have brought to light a role for Cyclin F as a deeply interconnected player in cell cycle transcriptional dynamics.

Klein et al. recovered Cyclin F in a RNAi screen for ubiquitin-related proteins required for DNA damage checkpoint arrest [124]. They showed that depletion of Cyclin F impairs the DNA damage induced checkpoint arrest and permits cells to enter mitosis despite the presence of damage. Mechanistically, they found that in response to ionizing radiation induced DNA damage, Cyclin F binds to the cell cycle transcription factor B-Myb. Surprisingly, Cyclin F restrains the ability of B-Myb to promote the expression of cell cycle genes that drive mitotic entry, and does so independent of B-Myb degradation, implying a non-catalytic mechanism for the role of Cyclin F in the maintenance, but not initiation, of a DNA damage induced checkpoint.

They further demonstrated that B-Myb binding to Cyclin F occludes B-Myb binding to Cyclin A; thus, depletion of Cyclin F results in an increase in B-Myb/Cyclin A binding. Since B-Myb is activated by Cyclin A/CDK2-dependent phosphorylation, the depletion of Cyclin F increases i) Cyclin A binding to B-Myb; ii) B-Myb phosphorylation and activation; and iii) expression of B-Myb target genes involved in cell cycle progression (e.g. Plk1, Cyclin A, Cyclin B). Therefore, their data led to a model where in response to DNA damage, Cyclin F antagonizes B-Myb by restraining its binding to Cyclin A and repressing the expression of genes that promote cell cycle progression.

The next hint of a role for Cyclin F in controlling the synthesis of proteins important for cell cycle came from Dankert et al. They discovered a role for Cyclin F in catalyzing degradation of the stem-loop binding protein (SLBP), a critical regulator of histone biogenesis and thus an important player in the synthesis of histone proteins in S-phase [125]. While not directly involved in histone transcription, they found that Cyclin F triggers the degradation of SLBP in G2 phase, after the completion of DNA replication, when new histone synthesis is no longer needed. Cyclin F bound a previously characterized site on SLBP that, like B-Myb, also binds to Cyclin A. Interestingly, it had previously been suggested that this binding site played a key role in SLBP degradation due to the binding of Cyclin A/CDK1 and the subsequent CDK1-dependent phosphorylation of SLBP in G2 [126]. Competition between, and relative contributions of Cyclin A/CDK1 and SCF^{Cyclin F} in mediating SLBP degradation remain unknown.

More recently, a string of papers identified a more direct role for SCF^{Cyclin F} in controlling cell cycle gene expression. A pair of studies from Clijsters et al. and Burdova et al. uncovered a role for SCF^{Cyclin F} in promoting the degradation of E2F1 once cells finish S-phase and enter G2 [127,128]. To determine how E2F activity is extinguished at the end of S-phase, Clijsters et al. analyzed binding between E2F3A and a panel of nine F-box proteins and found Cyclin F was the only one that bound E2F3A. They showed that Cyclin F could also bind E2F1 and E2F2, and that SCF^{Cyclin F} promoted the ubiquitination of all three activator E2Fs. Importantly, they identified mutant alleles in E2F1, E2F2 and E2F3a that could not bind to Cyclin F and which are resistant to degradation in G2-phase. When they expressed degradation-resistant activator E2Fs in U2OS cells, their enhanced stability resulted in increased expression of cell cycle genes which are known to be regulated by activator E2Fs, including RRM2, Cyclin E and CDC6 [127]. Notably, the Cyclin F and Cyclin A binding sites on E2F1 are the same [129]. It was previously reported that Cyclin A contributes to E2F1 inactivation by both directing CDK-dependent phosphorylation of E2F1 [27] and also by promoting E2F1 degradation [126]. The relative contributions and degree of redundancy between Cyclin A and Cyclin F in mediating E2F1 inactivation is unknown.

In a separate study, Burdova et al. identified E2F1 as a SCF^{Cyclin F} substrate based on an elegant drug sensitivity screen comparing control and Cyclin F knockout HeLa cells in combination with mass spectrometry analysis of Cyclin F interactors [128]. Interestingly, their data suggest that the loss of Cyclin F, which leads to an increase in E2F1 levels during G2-phase, is particularly deleterious in response to inhibition of the DNA damage checkpoint kinase Chk1. Burdova et al. performed whole transcriptome analysis using RNA-sequencing in Cyclin F knockout HeLa cells. Gene set enrichment analysis revealed an increase in the expression of E2F genes in Cyclin F knockout cells. Interestingly, changes in the expression of other Hallmark Gene Sets, including the Hedgehog and UV damage response pathways, were more significantly altered between control and Cyclin F knockout cells (our own re-analysis), suggesting a potential role for Cyclin F in other aspects of cell physiology and transcriptional control. Together with prior studies pointing to a role for Cyclin F in controlling B-Myb and SLBP [124,125], these two reports suggest that Cyclin F suppresses cell cycle transcription via E2F1 degradation and is therefore inextricably linked to the temporal patterning of gene expression throughout the cell cycle.

Two additional recent studies from Yuan et al. and Wasserman et al. further implicated Cyclin F in cell cycle transcriptional control [130,131]. Prior studies from both labs had analyzed E2F7/8 dynamics in the cell cycle and shown that they are controlled in G1-phase by the APC/C ubiquitin ligase [38,132]. Following these observations, both groups examined the dynamics of these atypical repressor E2Fs later during the cell cycle. Wasserman et al. identified mechanisms underlying E2F7/8 dynamics throughout the entire cell cycle, including roles for CDK-dependent phosphorylation in controlling E2F7/8 ubiquitination and degradation by APC/C. Both groups also sought to define the mechanisms which downregulate E2F7/8 prior to mitosis. Based on the importance of SCF ubiquitin ligases in cell cycle, and an absence of degron motifs for other SCFs that are involved in cell cycle, Wasserman et al. hypothesized a potential role for Cyclin F. They showed that Cyclin F bound to E2F8, controlled its stability, and was required for its degradation in G2-phase. Likewise, Yuan et al. identified a role for the Cullin RING ligase E3 family in E2F7/8 degradation and predicted a role for Cyclin F based on its high activity during G2-phase. They similarly found that E2F7 and E2F8 are controlled by Cyclin F during G2-phase. Thus, Cyclin F downregulates the activator E2Fs as well as the atypical repressor E2Fs via ubiquitination during G2-phase.

In addition to Cyclin F regulating cell cycle transcriptional programs through E2F family proteins, northern blot analysis and modern genomic approaches have found that Cyclin F mRNA expression is itself cell cycle regulated [8–12,115]. In addition, large-scale chromatin immunoprecipitation (ChIP) experiments indicate that Cyclin F expression is likely controlled by FoxM1 in G2-phase, although this has not been directly tested [46]. Moreover, the APC/C substrate receptor Cdh1 is also likely a target of FoxM1 [46]. If true, then FoxM1, and possibly B-Myb, might control mRNA expression of both Cyclin F and Cdh1. In addition, Cyclin F regulates Cdh1, B-Myb, E2F1/2/3 and E2F7/8 via ubiquitination and degradation. Further, APC/C^Cdh1 promotes the degradation of Cyclin F, FoxM1, and E2F7/8. And finally, E2F1 and E2F7/8 co-regulate each other. It would be difficult to overstate the remarkable complexity and interconnectedness of this circuit (see figure).

Figure 1.

A schematic illustration of relationships and feedback mechanisms between the E2F1-E2F7/8 transcriptional circuit and the SCF^{Cyclin F}-APC/C degradational circuit across cell cycle milestones. In G1-phase (top-left), APC/C and E2F1 are active. E2F1 drives the expression of cell cycle genes that ultimately promote S-phase entry, and also E2F7/8 which will ultimately extinguish its own activity. APC/C promotes the ubiquitination and degradation of many proteins, including the cell cycle transcription factor FoxM1, the SCF substrate receptor Cyclin F, and E2F7/8. In S-phase, APC/C is inactivated by several mechanisms, including the degradation of Cdh1 by SCF-Cyclin F. Since APC/C is inactive, E2F7/8 can inhibit the expression of E2F1 target genes. Cyclin F also inhibits B-Myb in response to DNA damage, although this inhibition is not through degradation. In G2-phase, Cyclin F protein levels peak and SCF-Cyclin F triggers the destruction of E2F1 (and E2F2/3) and E2F7/8. B-Myb and FoxM1 transcription factors are both fully active and promote the expression of Cyclin F and Cdh1. Finally, cells progress through and exit mitosis, APC/C becomes active and the system is reset.

Since SCF^{Cyclin F} promotes the ubiquitination and degradation of activator and repressor E2F proteins, which function in a feedback circuit [127,128,130,131], and Cyclin F also non-proteolytically controls B-Myb after DNA damage induced by ionizing radiation [124], it is important to understand its overall effect on cell cycle gene expression. To examine the role of Cyclin F in controlling cell cycle gene expression, Yuan et al. and Burdova et al. performed transcriptome analysis on Cyclin F depleted cells [128,130]. Consistent with their observation that Cyclin F triggers degradation of E2F1, Burdova et al. observed an overall increase in E2F gene expression following Cyclin F knockout based on gene set enrichment analysis. Klein et al. had previously monitored the mRNA expression of Cyclin A, Cyclin B, Plk1, and Aurora A by direct quantitative PCR analysis, and found that all were increased following Cyclin F depletion, which they attributed to an effect of Cyclin F on regulation of B-Myb [124]. Similarly, Yuan et al. observed an increase in the mRNA expression of the same genes, Cyclin A, Cyclin B, Plk1, and Aurora A, using whole transcriptome analysis in cells depleted for Cyclin F [130]. In addition, Yuan et al. also noted a decrease in some genes involved in DNA replication and repair [124]. However, whereas Yuan et al. and Klein et al. observed an increase in the expression Aurora A, Cyclin B1 and Plk1, none of these genes were changed in a statistically significant way in the transcriptome analysis performed by Burdova et al, and Cyclin A changed only minimally. The reason for these differences is unknown. It is possible that differences in the cell lines used in these studies (U2OS vs HeLa) contributed to differential responses to Cyclin F loss. Alternatively, the difference could be due to the starting levels of each of the individual genes or proteins which could shape how this signaling circuit is organized. Another possibility could relate to the very specific details of how experiments were performed. Finally, the complexity of the circuit could produce differential outcomes in response to the same perturbation based on external factors that determine how information moves through the circuit (discussed in detail below).

Finally, some studies have also pointed to a role for APC/C in controlling E2F1 and E2F3 during normal cell cycles or in quiescence [133–136]. However, E2F1 levels rise in mid G1-phase when APC/C^Cdh1 activity is at its peak, and unlike E2F7/8, both E2F1 and E2F3 are stable in APC/C-active cell free systems, suggesting a need to further examine the importance of this potential regulation ([38] and unpublished data).

Synthesizing complexity

The data above paints a remarkably complex portrait of the role that SCF^{Cyclin F} and APC/C^Cdh1 play in cell cycle transcriptional control (Figure 1). Beginning in early G1-phase, E2F1 becomes active and promotes expression of numerous cell cycle genes, including Cyclin E and CDC6, which are required for S-phase entry. Since E2F1 also promotes E2F7/8 expression, these atypical repressors begin to accumulate at the mRNA level during G1-phase. However, since APC/C is also active in G1-phase and promotes the degradation of E2F7/8, their protein levels remain low. At the G1/S boundary APC/C is inactivated, in part, through the degradation of its substrate receptor Cdh1 via the SCF^{Cyclin F} ligase. The inactivation of APC/C allows E2F7/8 proteins to accumulate throughout S-phase. In mid to late S-phase, E2F7/8 inhibit expression of the E2F1 gene expression program. Then, all of the activator and atypical repressor E2F proteins are targeted for degradation by SCF^{Cyclin F} in G2. Their ubiquitination and degradation presumably depend upon the increasing expression of Cyclin F, which is likely controlled by FoxM1. Once cells enter mitosis, Cdh1, which is also controlled by FoxM1, binds APC/C, promoting the degradation of myriad cell cycle proteins, including Cyclin F, FoxM1and E2F7/8 (see Figure).

This E2F-Cyclin F-APC/C circuit has important implications in considering how complex biological systems can be organized to provide robustness. It also illustrates cautionary truths for interpreting particular experimental approaches common in the study of E3-substrate relationships.

The use of non-degradable mutants has been essential in studying the biological significance of targeted proteolysis. For example, Clijsters et al. showed that expression of a non-degradable E2F1 in cells resulted in an increase in the expression of E2F1 target genes and caused E2F1-dependent apoptosis. While this confirmed that altering E2F1 degradation resulted in a phenotype consistent with the substrates upregulation, it left hidden the enzyme’s more complicated functional role in triggering degradation of the repressor E2F proteins. Notably, both E2F1/2/3 and E2F7/8 carry putative CY/RxL recognition motifs (degrons) for SCF^{Cyclin F}. While the RxL-degron in E2F1 was mapped [127,128], equivalent attempts could not unequivocally determine their identity in E2F7/8 [130,131]. Thus, the physiological role of Cyclin F-dependent degradation of the atypical, repressor E2F7/8 remains speculative.

The importance of E3-substrate relationships can also be analyzed by manipulating the E3. What then would be the prediction of experimental perturbations to the E2F-Cyclin F-APC/C circuit? Since the loss of Cyclin F increases the abundance of E2F1/2/3 and activates B-Myb (during DNA damage), one might anticipate that Cyclin F depletion would result in an increase in the abundance of cell cycle regulated gene transcripts. However, increased expression of E2F1 elevates expression of E2F7/8, and since E2F7/8 inhibits the expression of E2F1 and its targets, this would undoubtedly dampen the effect of upregulating E2F1. Confoundingly, Cyclin F depletion also directly stabilizes E2F7/8 which in turn represses E2F1 activity, further off-setting any potential increase in E2F target gene expression. Moreover, Cyclin F depletion stabilizes Cdh1. Because E2F7/8 are also APC/C^Cdh1 substrates, Cyclin F depletion could activate the APC/C and lead to E2F7/8 degradation.

The complexity of this network and the knowledge gap related to other inputs, makes it difficult to predict how perturbing Cyclin F might affect the abundance of specific proteins or alter cell cycle transcriptional programs. Indeed, independent studies observed different outputs after Cyclin F downregulation. Moreover, due to the highly interconnected nature of the network, perturbing other nodes would be expected to cause cascading and similarly unpredictable responses. Notably, the analysis of enzyme substrate relationships in vitro and the use of cell free systems that recapitulate specific physiologic cell states represent indispensable tools in defining these relationships. In fact, direct assays in G1 extracts led Cohen et al., to the discovery that both E2F7 and E2F8 are APC/C^Cdh1 targets [38]. Cell-free systems were vital in early cell cycle studies and continue to be an important tool that should complement many modern in vivo based approaches.

Two important predictions flow from the observations. First, the complex feedback built into the E2F-Cyclin F-APC/C circuit buffers the ultimate output against molecular noise resulting from stochastic variations in the expression or activity of critical hubs in the circuit (Cyclin F, E2F1/2/3, E2F7/8, Cdh1, etc.). Among a population of cells there is likely to be a distribution in the abundance of any given protein. Since many of the cell cycle genes are co-regulated by the same transcription factor, for example Cyclin F and Cdh1, their abundance would roughly correlate. However, in a given cell, transcriptional bursts at specific promoters combined with amplification of those differences through translation, can result in a greater or smaller number of molecules of any given protein [137]. We predict that contours and feedback connections in this circuit will buffer the ultimate output against this molecular noise. For example, a cell with an above (or below) average amount of Cyclin F would still be able to downregulate the G1/S gene expression program in late S and G2-phase, since Cyclin F controls both the activator and atypical inhibitor E2F proteins. Likewise, an over accumulation of Cdh1 could in principle lead to spurious APC/C activation in late S-phase or G2, degradation of E2F7/8 and deregulated expression of E2F1 target genes. However, this would be buffered by the fact that Cyclin F and Cdh1 would likely both be upregulated by the same transcription factor and because Cyclin F can promote the degradation of Cdh1.

A second prediction is that external cues or modifiers will regulate the flow of information through the E2F-Cyclin F-APC/C circuit and determine overall circuit output and its response to experimental or physiologic perturbation. We envision this circuit being affected by many proteins and enzymes which differentially regulate the abundance, activity, and interconnections of circuit components. Importantly, APC/C reactivation could depend significantly on the activity of CDKs, which phosphorylate Cdh1 and preclude its binding to the APC/C [77,92]. Similarly, expression levels of Emi1, which is transcribed by E2F and inactivates APC/C, might play a critical role. In addition, CDK phosphorylates E2F7/8 during mitosis and we recently showed a role for this in downregulating APC/C dependent degradation [131]. However, it remains unknown if phosphorylation of E2F7/8, or E2F1, which is also bound and phosphorylated by CDKs, alter recognition and ubiquitination by SCF^{Cyclin F}. We also recently showed that AKT phosphorylates and activates Cyclin F, and others pointed to a role for casein kinases in controlling Cyclin F abundance [138,139]. These too could alter how information flows through and ultimately alter the outputs of the circuit, particularly in response to perturbation. Circuit output could also depend on the activity of other key regulators, including, but not limited to, the Rb-tumor suppressor pathway, which directly binds and inhibits E2F in quiescent and G1 cells and also following stress or damage [140]. Likewise, the activity of other key proliferative regulators could sharpen, dampen or reverse the response to circuit perturbation. For example, the CKIs, which are controlled at the level of transcription and degradation, and which influence CDK activity, could influence information flow through the circuit, as could their own E3s and transcription factors. Another potential candidate is p53, which influences the expression of the CDK inhibitor p21, among myriad other genes involved in cell cycle control [141]. Ultimately, we anticipate that the effects of these and other circuit modifiers, will determine the overall influence of specific components and consequently, the overall output of the circuit.

Finally, Cyclin F, CKIs, APC/C and its regulators, Rb, Akt, and p53 are profoundly linked to tumorigenesis, cancer development and survival. It is reasonable to imagine that the E2F-Cyclin F-APC/C circuit is a molecular hub through which cancer cells evolve to foster proliferation and through which medical treatment could be developed.

Concluding Remarks

This review highlights the role of ubiquitin signaling networks in monitoring and coordinating the transcriptional program of the cell cycle. The studies summarized above identify the role of ubiquitination in E2F control, and we predict that the role of ubiquitin and the highly interconnected feedback observed for E2F, represent features that are likely to be shared by many other transcriptional networks. Second, we illustrate the importance of exercising caution when interpreting in vivo response to ubiquitin ligase manipulation, let alone in instances where the proteins being manipulated could be part of a complex circuit controlled by degradation and transcription via multiple mechanisms. This is particularly noteworthy since manuscript reviewers and journals are often intent on seeing an intuitive and straightforward phenotype resulting from ligase manipulation that matches upregulation of the reported substrate. However, these four studies highlight the fact that complex biological systems don’t always behave as predictably or neatly as one would expect. Third, it highlights the fact that studying the phenotypes of mutant alleles in substrates, which are made resistant to a ubiquitin ligase, can obscure the totality of that E3s broader function. In this case, analysis of the E2F1/2/3 mutant alleles, which enhanced protein stability, showed a predicted phenotype, but otherwise masked the more global role of Cyclin F in cell cycle gene expression. Finally, we believe that these observations together highlight the critical importance of collaborations between cell, molecular, and computational biologists in defining and modeling complex cellular circuits, allowing unforeseen predictions to be made and then tested.

Highlights.

• Cyclin F is a non-canonical cyclin that targets proteins for ubiquitination and proteasomal degradation via the SCF-family of E3 ligases

• Transcriptional activators E2F1, E2F2, and E2F3 are substrates of the SCF^{Cyclin F}

• Transcriptional repressors E2F7 and E2F8 are substrates of both SCF^{Cyclin F} and APC/C^Cdh1

• E2F1 and E2F7/8, and SCF^{Cyclin F} and APC/C^Cdh1 regulate each other via negative feedback, highlighting a complex circuit regulating cell cycle transcription

Outstanding Questions Box.

1. Mechanisms of Cyclin F-mediated degradation. Many SCF ligases recognize phosphorylated substrates [142,143]. However, there is no evidence among current Cyclin F substrates that it recognizes a phospho-degron. This suggests that other, hitherto undefined, mechanisms determine when SCF^{Cyclin F} substrates are degraded. Cyclin F begins to accumulate during S-phase when it promotes the degradation of Cdh1, RRM2 and NuSAP1 [119,120,123]. However, other substrates, including activator and repressor E2F proteins, are not degraded until G2 [127,128,130,131]. Learning from the APC/C, there are several possible mechanisms that can account for these differences, including binding affinities to Cyclin F, processivity of ubiquitination, antagonizing deubiquitinases, or subcellular localization [144,145].

2. Substrate ordering within the E2F world. Four independent studies converged on the notion that E2F1, E2F2, E2F3a, E2F7 and E2F8 are SCF^{Cyclin F} substrates at or near the completion of S-phase. It is unknown if all five E2F proteins, together with SLBP and perhaps other Cyclin F substrates, are degraded simultaneously, or in a specific order (like Cyclins A and B and Cdc20 in mitosis). It is therefore unknown if their potential ordering has biological significance. Relatedly, it is unknown if and how Cyclin A and Cyclin F coordinate to temporally inactivate E2F1/2/3 through phosphorylation and ubiquitination, respectively.

3. Inter-dynamics of the E2F1-E2F7/8 circuit across cell cycle milestones. Single cell analysis of protein dynamics highlights discrete cellular properties obfuscated by bulk measurements [146]. Inter-dynamics of the native E2F1-E2F7/8 network has yet to be described at single cell resolution. The introduction of protein markers at endogenous loci is now feasible but may be challenging for E2F proteins due to potential structural constraints associated with dimerization and DNA binding. Protein dynamics, however, can also be extracted from fixed cells using quantitative immunofluorescence and analysis of population dynamics by rate equations [147]. This requires highly specific antibodies for endogenous proteins or, alternatively, small immunodetectable tags integrated at genomic loci. In combination with cell cycle markers whose dynamics are regulated by APC/C and/or SCF, inter-dynamics of the E2F1-E2F7/8 circuit across cell cycle milestones can be uncovered.

4. Computational modeling of the Cyclin F-E2F-APC/C circuit. Complex circuits can be interpreted by collaborations between cell and computational biologists. Such analysis would enable predictions of how the circuit would respond to perturbation and would include information related to circuit modifiers, including kinases like AKT, Casein Kinase and CDK. This is essential to understanding the flow of information through the circuit.