Polo-like kinases participate in several processes that collectively promote mitotic progression, including mitotic exit, early anaphase, APC activation, and sister chromatid separation 
. The discovery of an adaptation-defective allele of CDC5
suggested that this kinase also had a role in negatively regulating the DNA damage checkpoint 
, however the mechanistic details remained unknown. Our data suggest that Cdc5 does not inhibit formation of the Rad9-Rad53 complex and yet blocks the ability of the Mec1-primed Rad53 molecules to produce hyperphosphorylated Rad53 in vivo.
Adaptation to DNA damage begins to occur after approximately 6–8 h of cell cycle arrest if cells were unable to repair the damage. Loss of checkpoint signaling has been previously shown to correlate with the onset of adaptation 
. However, there could be multiple pathways converging on the checkpoint after an extended cell-cycle arrest. One of the advantages of the CDC5
overexpression approach taken here is that it has allowed us to isolate CDC5
-specific effects from those of other pathways. For example, Ptc2 and Ptc3 clearly have a role in Rad53 regulation, and deletion of these phosphatases causes an adaptation-defective phenotype 
. However, we found that checkpoint suppression caused by CDC5
overexpression occurred in the absence of both these phosphatases (), consistent with the model that at least two pathways work independently to promote adaptation. Ptc2 and Ptc3 may have important roles in recovery from the checkpoint once damage has been repaired. Cdc5 does not appear to have an essential role in the process of recovery, since cdc5-ad
mutants are able to reenter the cell cycle once damage is repaired 
and Rad53 dephosphorylation occurs in DNA-damaged cdc5-ad
strains if MEC1
activity is removed 
. Rad53 has recently been proposed to act in a negative-feedback loop, in which Rad53 phosphorylates Rad9 to prevent the BRCT-SCD domain-specific oligomerization of Rad9 that is required to maintain checkpoint signaling 
. While this negative-feedback loop may also feed into adaptation, our results showing that overproduced Cdc5 prevents in vivo Rad53 autophosphorylation suggest Cdc5 exerts its effect upstream of this loop.
We found that Cdc5 and Rad53 could interact both in vivo and in vitro, which could support the notion that Cdc5 directly inhibits Rad53. Cdc5 kinase activity was required to suppress Rad53 phosphorylation () and kinase-dead Cdc5 was co-immunoprecipitated with Rad53 (unpublished data), eliminating the mechanism of simple binding inhibition. Interestingly, hypophosphorylated Rad53 from Cdc5 overproducing cells retained its ability to trans-autophosphorylate by ISA (). A population of Rad53 that is capable of undergoing limited autophosphorylation in the ISA assay in the absence of a significant phospho-shift as measured by gel mobility assay has been observed previously in the checkpoint-defective rad53 FHA2-R605A mutant 
. These data and those presented here suggest that these two assays measure distinct aspects of Rad53 activation that are together required for its in vivo function and that Cdc5 may specifically act to counter one of these functions in vivo. In fission yeast, phosphorylation of Cds1 (the Rad53 homolog) by Rad3 (the Mec1 homolog) is thought to promote Cds1-Cds1 interactions required for autophosphorylation 
. Similarly, Rad53 autophosphorylation activity requires Mec1/Tel1 phosphorylation 
. Therefore, if Mec1/Tel1 were completely inhibited, Rad53 would not be active in the ISA assay, which we did not observe. Consistent with Rad53 maintaining its priming phosphorylation, Rad53 remained a tight doublet even after CDC5
overexpression caused loss of its hyperphosphorylation. This, along with our data demonstrating Rad9 is appropriately phosphorylated by Mec1/Tel1 despite CDC5
overexpression, would suggest that Mec1/Tel1 are active and can phosphorylate Rad53 enough to prime its activity.
Cdc5 was able to directly phosphorylate Rad53 in vitro. Cdc5 phosphorylation might affect the positioning of Rad53 with respect to either other Rad53 molecules or Rad9 so as to prevent proper Rad53 trans-autophosphorylation. Active and phosphorylated Rad53 must be released from Rad9 
, suggesting that these Rad9-bound hypophosphorylated Rad53 molecules could act dominantly to prevent further checkpoint activation, as does expression of the kinase-dead allele of RAD53 
Our demonstration that Cdc5 phosphorylation of recombinant Rad53 depends on both Rad53 FHA domains () is particularly intriguing. First, it suggests that this activity is quite specific. Moreover, it argues that Rad53 provides the binding specificity to allow Cdc5 to phosphorylate it, in contrast to the classic model in which polo-like kinases recognize a substrate via their phosphobinding polo-box domains and then subsequently phosphorylate the bound substrate 
. This mechanism is also different from how the human homologs, Chk2 and Plk1, are reported to interact 
. However, as both proteins contain phosphobinding motifs, mutual recognition between Cdc5 and Rad53 may be required in vivo. An alternative model of indirect inhibition is one in which Rad53 could bridge an interaction between Cdc5 and Rad9 and promote Cdc5 phosphorylation of Rad9. As a result, Cdc5-mediated phosphorylation could interfere with proper Rad53 autophosphorylation. This model has the benefit of targeting the checkpoint mediator responsible for activating the two parallel effector kinases Rad53 and Chk1, both shown to lose activity as cells adapt 
Cdc5 can now be added to the growing list of proteins that interact with the Rad53 FHA1 domain. Rad53 contains two FHA domains, one at each terminus, whereas homologous proteins such as human Chk2 and S. pombe
Cds1 contain only one N-terminal FHA domain. Although both Rad53 FHA domains contribute to its checkpoint function, the N-terminal FHA1 is more structurally similar to its homologous counterparts. This raises interesting prospects on how Rad53's FHA1 domain facilitates interactions with downstream targets including Dbf4, Asf1, Mdt1, Rad9, and other Rad53 molecules 
, as well as promote its own inactivation by interacting with Ptc2 
and, potentially, Cdc5.
Our results strongly suggest the polo-like kinase, Cdc5, can inhibit checkpoint signaling at the level of Rad53 hyperphosphorylation. Rad53 autoactivation provides an amplification step in which primed Rad53 can activate additional Rad53 molecules in a positive-feedback loop, thereby preventing premature or unnecessary checkpoint activation. The findings that both the in vivo interaction and the in vitro phosphorylation of Rad53 by Cdc5 imply that there is potential for a constitutive interaction, in agreement with human Chk2 and Plk1 data 
. While the biological significance for a constitutive interaction is not yet clear, it presents the opportunity for each kinase to inhibit the other and generate a switch-like decision to undergo adaptation. Indeed, Plk1 has been reported to be inhibited by the DNA damage checkpoint 
. This leads us to question, what can tip the balance of this potential inhibitory face-off: the activity of a third kinase such as CDK on either or both Rad53 and Cdc5, or the relative strength of their interaction compared to other substrates?
Adaptation can be considered as a final attempt at survival after yeast have exhausted all other repair options. However, as a consequence of promoting cell division in the presence of DNA damage, adaptation also results in increased genomic stability 
. Our study of adaptation, particularly our use of CDC5
overexpression, may provide valuable insights into the mechanisms of tumorigenesis. The human homologue PLK1
has been reported to be overexpressed in various tumors including non-small-cell lung cancer, melanoma, colorectal cancer, and non-Hodgkin lymphoma. In addition, the levels of PLK1
in a subset of tumor types may provide prognostic value 
. Our work implies that, if indeed parallel with adaptation, PLK1
overexpression could lead to checkpoint suppression, an enhanced rate of mutagenesis due to genomic instability, and ultimately carcinogenesis.