2.1 The G2/M Checkpoint
Early studies implicated 14-3-3 as a critical integration point for many of the protein kinases and phosphatases that control the transition from G2 into M phase (). Normal mitotic entry from G2 is triggered by a rapidly enhanced activation of Cdk1/cyclin B (c.f. [5
] and references therein). Cdk1 is activated by constitutive phosphorylation of a threonine residue (T161) within the “T-loop” by CAK (Cdc2-activating kinase) but maintained in an inhibited state by phosphorylation of a tyrosine residue (Y15) within the ATP-binding “P-loop” by the tyrosine kinase Wee-1 [5
]. In addition, higher eukaryotes also have a membrane-bound kinase, Myt1, that phosphorylates both Y15 and the preceding threonine, T14, to ensure Cdk1 inhibition [5
]. Removal of these inhibitory P-loop phosphorylations on Cdk1 to promote entry into mitosis from G2 is achieved by the Cdc25 family of dual-specificity protein phosphatases [6
]. Mammalian cells express three different Cdc25 proteins (A, B, and C), all of which have been implicated in G2/M progression. At the G2/M boundary, Cdc25B appears to be the trigger phosphatase that initially activates Cdk1/cyclin B, which in turn phosphorylates Cdc25C to create a binding site for a second mitotic kinase, Plk1 [8
], which in turn further phosphorylates Cdc25C [10
]. This cascade of sequential phosphorylation events results in an autoactivation loop that subsequently drives entry into mitosis [7
]. While Cdc25A is predominantly recognized for its involvement in earlier phases of cell cycle progression (see below), it has been shown to also play a role at the G2/M transition [14
] and is the only Cdc25 family member that has an embryonic lethal phenotype in knockout mice [16
]. The relative importance of Cdc25A in G2/M progression in normal cells probably depends on the status of Cdc25B and C, which may be influenced by the prior history of the cells.
The role of 14-3-3 in G2/M checkpoint activation, maintenance, and release
Like Cdk1/cyclin B, the Cdc25 phosphatases are themselves regulated through activating and inhibitory phosphorylations. For Cdc25B and C, activating phosphorylations are generated by Cdk1 and Plk1 [10
] through the positive feedback loop described above, although data in Xenopus extracts suggests that MAPKs (particularly Erk1/2) may play a major role [20
]. Whether this is also true in mammalian cells is less clear. Inhibitory phosphorylation events during G2 probably arise from the detection of ongoing endogenous DNA damage from normal replication events during the preceding S-phase, or exogenous DNA damage that the cell has encountered during late S and G2, and are mediated by three DNA damage-responsive kinases, Chk1, Chk2, and MK2 [21
] (see below). The main mechanism by which these inhibitory phosphorylations block Cdc25B and C function is through the phosphorylation-dependent binding to 14-3-3. Phosphorylation of a single binding site on Cdc25B (S323 in human), or Cdc25C (S216 in human) [23
] is necessary and sufficient for 14-3-3 binding, although there are some indications that additional binding sites may also contribute to their regulation [25
] in ways that are somewhat mechanistically unclear.
There are likely to be multiple mechanisms through which 14-3-3 proteins inhibit Cdc25B/C function. First, 14-3-3 binding to Xenopus Cdc25C causes a direct two-fold reduction in its phosphatase activity when assayed in vitro using Myt1-phosphorylated 32
P-labelled Cdk1 as a substrate [23
]. A similar direct inhibition of Cdc25B catalytic activity is likely to occur based on pull-down data indicating that 14-3-3 binding to Cdc25B blocks Cdc25B binding to CyclinB/Cdk1 [25
]. Second, 14-3-3 binding leads to the cytoplasmic sequestration of Cdc25B/C [27
], likely blocking access to its nuclear substrate Cdk1/cyclin B [31
], required for mitotic entry. The mechanism by which 14-3-3 proteins regulate the subcellular localization of their bound targets involves the selective exposure or shielding of specific sequence elements. This has been most clearly experimentally elucidated for Cdc25C, although the identical mechanism almost certainly functions for Cdc25B. In between the N-terminal regulatory domain and the C-terminal catalytic domain of Cdc25C lies an intrinsic nuclear localization sequence (NLS) and a nuclear export sequence (NES) [28
]. These subcellular localization signals allow Cdc25C to shuttle between the nuclear and cytoplasmic compartments. The major 14-3-3 binding site in human and Xenopus
Cdc25C (S216 and S287, respectively) is located right next to the NLS [28
], and in Xenopus
, 14-3-3 binding to Cdc25C has been shown to block the binding of Cdc25C to the importin-α nuclear import receptor [29
]. Thus, 14-3-3-binding causes defective nuclear import in the face of unaffected nuclear export [33
], relocalizing most of Cdc25C to the cytoplasm.
Although all of the 14-3-3 isoforms except 14-3-3 σ can bind to Cdc25 isoforms in vitro [34
], it appears, on the basis of co-IP experiments using tagged overexpressed proteins that the ε and γ isoforms of 14-3-3 bind strongest to Cdc25C in cells [34
], while yeast 2-hybrid experiments suggest that 14-3-3 β, η, τ, and ζ bind to Cdc25B [36
]. Whether this marked 14-3-3 isoform exclusivity for the mitotic ‘initiator’ Cdc25B phosphatase versus the mitotic ‘sustainer’ Cdc25C phosphatase is biologically important remains unknown.
In addition to inhibiting the Cdc25 phosphatases that activate Cdk1, 14-3-3 proteins also bind to Wee1, the protein kinase that inhibits Cdk1. In this case, however, 14-3-3 binding serves to distribute Wee1 uniformly throughout the nucleus, and appears to increase the catalytic activity of both Xenopus and human Wee1 at least 5-fold [37
]. These two 14-3-3 dependent mechanisms of Wee1 activation work synergistically to enhance the phosphorylation and inactivation of nuclear Cdk1, and together with the 14-3-3-mediated nuclear exclusion and catalytic reduction of Cdc25B and C activity, acts as a ‘belt and suspenders’ mechanism to initiate a strong G2 checkpoint arrest (). Interference with either 14-3-3:Cdc25 or 14-3-3:Wee1 binding results in inappropriate mitotic entry [23
In contrast to the selectivity of Cdc25B/C for binding to 14-3-3 isoforms other than σ, it appears that Wee1 binds directly to14-3-3σ [38
], likely in addition to other 14-3-3 proteins. 14-3-3 σ, a unique isoform that appears to have first arisen in birds, appears to play an especially important role in G2/M checkpoint maintenance. Hermeking, Chan, Vogelstein and colleagues made the important finding that 14-3-3σ was potently up-regulated by p53 in colorectal carcinoma cells after ionizing radiation, where it functioned in parallel with p21 to ensure that DNA damaged cells maintained a G2 arrested state [40
]. Cells lacking 14-3-3 σ could initiate, but were unable to maintain a G2 arrest after irradiation [40
], while cells lacking both 14-3-3σ and p21 showed both accelerated cell death and enhanced sensitivity to doxorubicin-induced DNA damage and 5-fluorouracil treatment [41
]. Possible targets of 14-3-3σ may be Cyclin/Cdk1 and Cdk2 complexes themselves [40
], although neither of these targets was detected in an unbiased proteomic screen for 14-3-3σ ligands [39
]. Importantly, 14-3-3σ likely functions, at least in part, through enhancing the activity of p53 as part of a coherent feed-forward loop ([45
]; see below), as well as through enhancing the stability of the Cdk inhibitor protein p27Kip1 by blocking its AKT-mediated ubiquitination and degradation [46
]. 14-3-3 regulation of p27Kip1 is more complex, however, since 14-3-3 bound p27Kip1 becomes re-localized in the cytoplasm, where it is presumably nonfunctional in suppressing Cdk activity [47
]. This potentially allows tumor cells with high levels of PI 3-kinase/AKT activity to bypass G1 arrest (see below), while maintaining a sufficient total concentration of intracellular p27 to arrest the cell if DNA damage occurs subsequently.
14-3-3-binding sites on Cdc25, Wee-1, and a variety of other checkpoint modulators are generated through the action of DNA damage-responsive basophilic kinases, particularly the kinases Chk1, Chk2, and MK2 [22
]. It is thought that Chk1 predominantly responds to single strand breaks, lesions created during replication fork collapse, and bulky DNA adducts induced by UV irradiation, while Chk2 is thought to be the major kinase responding to double strand breaks, although these kinases appear to participate to varying extents in all three DNA damage responses [53
]. Intriguingly, in fission yeast Chk1 itself binds to 14-3-3, and this binding appears to be important for phosphorylating and activating Chk1, and targeting it to the nucleus after DNA damage [54
]. Recent data have now revealed that a third DNA damage kinase, the p38MAPK-activated kinase MK2, is critical for prolonged G2 arrest in tumor cells that are defective in p53 signaling [48
]. The major direct targets of MK2 appear to be the Cdc25B and C phosphatases, which MK2 directly phosphorylates on their 14-3-3-binding motifs [48
], and whose persistent 14-3-3-binding and nuclear exclusion requires persistent MK2 kinase activity. Importantly, the mechanism responsible for prolonged MK2 activity after DNA damage involves another example of a coherent feed-forward loop, in this case involving MK2-mediated phosphorylation of RNA binding proteins, which stabilize the production of the cell cycle regulator Gadd45α, which in turn binds to p38MAPK to maintain prolonged MK2 activity [55
]. The fact that this pathway is critical for preventing inappropriate checkpoint release and mitotic catastrophe in p53-defective cells suggests that it may be a useful therapeutic target to preferentially kill p53-defective tumor cells [49
2.2 G2/M Checkpoint Release
Re-entry into the mitosis from G2 upon completion of DNA repair appears to involve Cdk1 and Plk1-dependent inactivation of the four upstream kinase pathways responsible for 14-3-3-dependent inactivation of CyclinB/Cdk1 itself - the ATM-Chk2 pathway, the ATR-Chk1 pathway, the p38MAPK-MK2 pathway (all of which target Cdc25 for binding to 14-3-3), and the Wee1 pathway (). Inactivation of the ATR-Chk1 pathway occurs through Plk1 mediated phosphorylation of Claspin, an adaptor protein that facilitates ATR-dependent Chk1 activation. Plk1 phosphorylation releases Claspin from damaged chromatin [57
], and targets it for ubiquitin-mediated degradation [58
]. Inactivation of the ATM-Chk2 pathway involves the Cdk1 and Plk1 dependent inactivation of the adaptor protein 53BP1 (that facilitates sustained Chk2 activation by ATM as well as direct inactivation of Chk2 itself in both mammalian cells and yeast [61
]. Inactivation of the p38MAPK-MK2 pathway is less well understood, but likely involves the dephosphorylation and inactivation of p38MAPK by the phosphatase Wip1 [63
], which also dephosphorylates the DNA damage marker histone γH2AX along with other ATM/ATR and DNA-PK targets [65
]. Inactivation of the Wee1 pathway involves the Cdk1 and Plk1 phosphorylation of Wee1, which targets it for ubiquitin-mediated proteolysis [67
]. These mechanisms function together to block upstream signals that target Cdc25B/C and Wee1 for 14-3-3 binding, thus shutting off the G2/M checkpoint regulated by 14-3-3-mediated Cdc25 inactivation and Wee1 activation, and allowing cells to re-enter the cell cycle.
The process of checkpoint release is still under active investigation. This is particularly important since it appears that G2/M checkpoint bypass may be an early event that triggers the progression from carcinoma in situ to full blown cancer [69
]. In addition, the G2/M checkpoint appears to be an error-prone checkpoint, since both normal cells and tumor cells are released from G2 into M-phase despite the presence of a finite number of unrepaired DNA breaks [70
], resulting in genomic instability. Taken together, these findings suggest that modulation of 14-3-3 function to enhance G2 arrest may be a viable mechanism to limit tumor cell proliferation and progression.
An important observation made by Kornbluth and colleagues was that 14-3-3 binding to Xenopus
Cdc25 phosphorylated on S287 protected this mitotic phosphatase from premature dephosphorylation and activation [71
]. In addition, protein phosphatase 1 was required to dephosphorylate Xenopus
Cdc25 at S287 (the 14-3-3 binding site corresponding to S216 in human Cdc25C) [72
]. Interestingly, the removal of 14-3-3 from Cdc25 preceded S287 dephosphorylation, implying that a phosphatase-independent pathway is responsible for the initial 14-3-3 removal from Cdc25, followed only later by dephosphorylation and loss of the 14-3-3 binding site on Cdc25. These authors postulated that the accumulation of phosphorylated intermediate filament proteins during mitosis may compete with Cdc25C for 14-3-3 binding, with a shift in the equilibrium away from 14-3-3:Cdc25C as the abundance of phosphorylated forms of intermediate filament proteins accumulate in mitosis. These findings agree well with a prior report from Tzivion et al. [73
] that treatment of COS-7 cells with the phosphatase inhibitor Calyculin-A resulted in marked accumulation of 14-3-3:phosphovimentin complexes, displacing many of the other 14-3-3-bound ligands, and a series of strong papers from Omary and colleagues that showed binding of phosphorylated epithelial keratins K8/18 to 14-3-3 is increased during mitosis, where it may contribute to mitotic progression, at least in hepatocytes during liver regeneration [74
]. These studies raise the possibility that intermediate filament phosphorylation during mitosis may function to facilitate the release of specific ligands from 14-3-3 whose function is important during cell division by competitive binding, although additional mechanisms are likely to be operative during the initial stages of checkpoint release when mitotic kinase activity may not be high enough to produce an abundance of phosphokeratins and phosphovimentin. Importantly, the 14-3-3 strongest binding motif on Cdc25B and C corresponds to the sequence RSPS
MP, where the bolded highlighted Ser is the critical 14-3-3-binding phosphorylation site. Once cells have entered mitosis, Cdk activity results in phosphorylation of the preceding Ser residue (RS
PSMP), which renders these Cdc25 molecules incapable of binding to 14-3-3 proteins [76
], and thereby ensuring that mitotic progression is strictly unidirectional.
The fact that Cdk1 is intimately involved in nearly all of the G2 checkpoint release mechanisms discussed above shows that, like checkpoint establishment, checkpoint release also involves a feed-forward loop when viewed from a systems biology perspective.
2.3 Mitotic roles of 14-3-3 proteins
In addition to from blocking mitotic entry, 14-3-3 proteins appear to contribute substantially to enhancing progression once cells have entered mitosis through stimulating the process of cytokinesis. Parker and colleagues showed that 14-3-3 binding to protein kinase C epsilon (PKCε) during mitosis activates it in a lipid-independent manner, locking PKCε in an open active conformation near the actomyosin ring. The resulting PKCε activity is crucial during telophase to down-regulate RhoA activity at the midbody, and allow cells to complete abscission. Loss of the 14-3-3 binding sites on PKCε, loss of PKCε expression, expression of a catalytically dead form of PKCε or expression of a dominant negative form of 14-3-3 results in cytokinesis delay or failure and the production of multinucleated cells [78
]. The cancer relevance of this phenomenon is unclear, since PKCε is generally thought to be a potent oncogene and its expression is upregulated in certain tumors and tumor-derived cell lines [79
Our laboratory showed that 14-3-3σ plays an important role in cytokinesis by facilitating the mitotic switch from cap-dependent to cap-independent translation. Notably, 14-3-3σ is a tumor suppressor protein that is frequently lost in tumors of epithelial origin [80
], inhibiting the ability of such cells to efficiently complete cytokinesis. This ultimately leads to the production of tetraploid cells that may serve as early precursors of cancer [81
]. In normal epithelial cells, completion of normal cytokinesis occurs as a consequence of binding of the translation initiation factor eIF4B to 14-3-3σ during mitosis, impairing eIF4B function, and culminating in a switch from cap-dependent to cap-independent (IRES mediated) mRNA translation during mitosis. During mitosis, a variant form of Cdk11, termed p58-PITSLRE, is then translated in a cap-independent manner [82
]. This cap-independent form is specifically required to complete the final stages of cell division [83
]. In the absence of 14-3-3σ, only the cap-dependent interphase form of Cdk11, p110-PITSLRE is produced, rendering 14-3-3σ-defective cells unable to sever the cell-cell attachment at the terminal midbody stage (abscission) of cell division [81
]. These findings are particularly significant given the finding that the expression of 14-3-3σ is suppressed in normal breast tissue adjacent to the sites tumors arose (see below). In this case, loss of 14-3-3 probably contributes to ‘field cancerization’ by facilitating the development of pre-cancerous tetraploid lesions that subsequently undergo genomic instability during the tumorigenic process.
2.4 G1/S Checkpoint control
The remaining Cdc25 phosphatase that has not been discussed at length, Cdc25A, is best appreciated for its role in regulating the G1/S transition by controlling CyclinE and CyclinA/Cdk2 activity. Cdc25A has been generally accepted as being primarily regulated in response to DNA damage by Chk1-dependent phosphorylation and β-TrCP-mediated ubiquitination and proteosomal degradation [85
]. Other kinases however, including CK1, GSK-3β, and Plk3 have also been implicated in this process [88
]. In addition to its role in the G1/S transition, Cdc25A also appears to play a role in facilitating the G2/M transition by activating CyclinB/Cdk1 [91
]. Piwnica-Worms and colleagues showed that Cdc25A has two 14-3-3-binding sites, and that mutants of Cdc25A that are defective in 14-3-3 binding have an enhanced ability to activate Cyclin B/Cdk1 [92
]. One of the two 14-3-3-binding sites sits near the C-terminal region of Cdc25A, immediately adjacent to a putative Cyclin B docking site, suggesting that 14-3-3 binding disrupts the interaction of Cdc25A with its substrate. These data suggest that 14-3-3-binding to Cdc25A may be important for short-term control of Cyclin/Cdk complexes involved in both G1/S and G2/M progression, while ubiquitin-mediated destruction of Cdc25A may be a mechanism of long term inhibition of Cyclin/Cdk activation in response to larger amounts of genotoxic stress. Unfortunately, the structural basis for 14-3-3-mediated effects on Cdc25 phosphatases remains unclear, largely because the 14-3-3-binding sites are primarily localized in relatively unstructured regions of Cdc25 between the N-terminal regulatory domain and the C-terminal phosphates domain.
An additional role for 14-3-3 proteins in the G1/S transition has emerged from a recent finding that 14-3-3τ seems to facilitate the MDM2-dependent but ubiquitin-independent degradation of the Cdk inhibitor p21 to facilitate S-phase entry [93
] (). Thus 14-3-3 proteins appear to have roles in both suppressing and enhancing the G1/S transition, likely as a function of which client proteins are bound, depending on the relative levels of activity of different 14-3-3 motif-generating kinases.