We have identified five sites of Rad53-mediated phosphorylation on Swi6 by using both recombinant Rad53 and the kinase immunoprecipitated out of yeast cells undergoing DNA damage. Importantly, Rad53 from both of these sources produces nearly identical profiles of phosphorylation on Swi6. This suggests that Rad53 alone and not a copurifying yeast kinase is responsible for Swi6 phosphorylation in our assays. Moreover, it indicates that DNA damage does not alter the site specificity of Rad53 but only increases kinase activity. This result is consistent with our previous observation that Rad53 overproduced in undamaged yeast cells can also phosphorylate Swi6 in vitro.
At least one of the identified sites, S547, is phosphorylated in vivo in an MMS-inducible, Rad53-dependent manner, suggesting that Rad53 recognizes Swi6 as one of its substrates in a cell undergoing DNA damage. Phosphorylation of this site is substoichiometric, suggesting that only a subpopulation of Swi6 molecules can be phosphorylated. For example, only DNA-bound or Swi4-bound fractions of Swi6 could be targeted by Rad53. We did not detect in vivo phosphorylation of the other sites mapped in vitro on Swi6. These sites may not be accessible, or they may be phosphorylated in a even more transient or substochiometric manner. Alternatively, there may be inducible phosphopeptides that are obscured by overlapping constitutive phosphopeptides on the peptide map.
Alignment of the five Rad53 phosphorylation sites of Swi6 reveals a remarkable degree of conservation, particularly at −2 and +2 positions (Fig. ). Both of these positions are occupied only by hydrophobic aliphatic amino acids, and conservative substitution of these residues to alanine prevents phosphorylation by Rad53. Position −3 shows a preference for basic or hydrophilic residues, and proline may be unfavorable at +1. There may also be a bias toward basic residues at the −6 position. Finally, it appears that either position 0 or +1 can be phosphorylated within a given site, if occupied by serine or threonine. Interestingly, comparison of the Swi6 sites to the in vivo phosphorylation site recognized by Chk2, the mammalian homolog of Rad53 (36
), on human BRCA1 (S988) (33
) shows exactly the same sequence preference at −2 and +2, as the one identified in the present study (Fig. ). Another in vivo Chk2-recognized site, S20 of p53 (8
), a reportedly less-than-optimal site, has a leucine only in the +2 position. This is consistent with our finding that “half” sites can be phosphorylated by Rad53 in vitro, albeit weakly. However, the other reported in vivo Chk2 phosphorylation site, S123 of CDC25A (15
), has no obvious similarity to the −2/+2 consensus (Fig. ).
The specificities of the Chk1 and Chk2 kinases have been studied by using peptide libraries (41
). The peptide substrate preferences derived for human Chk2 are similar to our consensus in that they include hydrophobic residues at positions following the phosphorylatable residue, and a basic residue, arginine, at position −3. However, there is no preference seen for the −2 position in the peptide study. These differences may be attributed to the use of peptide versus protein substrates or to the necessarily limited array of sequences present in any peptide library. It is also possible that human Chk2 and yeast Rad53 have somewhat different substrate preferences or that the identified Swi6 sites are not the highest affinity sites for Rad53 phosphorylation, just as S988 in BRCA1 is not the best Chk2 substrate (41
). The in vivo targets of checkpoint kinases may carry multiple suboptimal sites as a way of transforming their phosphorylation into an on-off switch actuated by a substantial increase of the kinase activity, as is the case for Sic1 phosphorylation (39
The −2/+2 consensus allows us to derive insights into Rad53 activity and regulation. All of the proteins implicated as targets or binding partners of Rad53 (Crt1, Dbf4, Dun1, Cdc5, Rad55, Bfa1, Asf1, Rad9, and Mrc1 [1
]) carry one to nine matches to the −2/+2 consensus. In Rad53 itself there are eight potential −2/+2 sites, most of which are within 30 amino acids of S/TQ sites, which are potential sites for phosphorylation by the upstream checkpoint kinases Mec1 and Tel1. The association of potential Rad53 and Mec1/Tel1 sites suggests that both transcatalytic (via Mec1 and Tel1) and autocatalytic pathways may activate Rad53 in response to damage (20
In addition, three −2/+2 sites are located between the β-sheets of the FHA2 domain of Rad53 (12
). FHA1 and FHA2 are protein domains that bind to phosphorylated threonine residues embedded within the FHA1- or FHA2-binding motifs (TxxD or TxxL/I, Fig. [11
]). Autophosphorylation of Rad53 on these sites could change the properties of the FHA2 domain. For example, the phosphorylated FHA2 domain of Rad53 may no longer associate with Rad9. This could provide a mechanistic explanation for the observed release of Rad53 from the complex with Rad9 after autophosphorylation (20
FIG. 8. (A) −2/+2 sites can overlap FHA-binding motifs. FHA1- and FHA2-binding motifs identified by using peptide libraries (12) were aligned with the known FHA1-binding motifs of p53 (10) and Rad9 (62), the known FHA2-binding motifs of Rad9 ( (more ...)
Interestingly, four of five Swi6 −2/+2 sites overlap or adjoin sequences that match the FHA2-binding motif (Fig. ). This is analogous to the in vivo Chk2 phosphorylation site S20 in p53, which overlaps the FHA1-binding motif that binds the FHA1 domain of Rad53 (10
). In Rad9, there are three −2/+2 sites within the cluster of the FHA-binding motifs that were shown to be necessary for Rad53 binding and activation in vivo (47
). Among these, two −2/+2 sites completely overlay both the putative Mec1/Tel1 phosphorylation sites (S/TQ) and the FHA-binding motifs. The third −2/+2 site in Rad9 adjoins the FHA-binding motif (Fig. ). It is tempting to speculate that Rad53 phosphorylation adjacent to FHA-binding motifs affects their ability to bind to FHA domains.
Our search for consensus sites for Rad53 phosphorylation in other proteins revealed a number of potentially interesting new candidate Rad53 targets among cell cycle and DNA metabolism proteins, in particular, the cohesin complex. This complex is known to be targeted by the DNA damage checkpoint in G2
both in yeast (60
) and in mammals (30
). We have shown that one of the cohesin complex subunits, Scc1, undergoes DNA damage-induced phosphorylation in vivo. We also established that at least one of the damage-inducible phosphorylation events on Scc1 is largely Rad53 dependent. Although an in vitro kinase assay is needed to prove that Scc1 is indeed a direct Rad53 target, our data are certainly consistent with this possibility.
Given the fact that the DNA damage checkpoint pathways in budding yeast (19
) consist of multiple parallel branches, it is not surprising that elimination of the Rad53 phosphorylation sites in Swi6 has only a minor effect on the length of the checkpoint-mediated delay in the G1
-to-S transition after damage. It is possible that Rad53 acts upon other components of the late-G1
transcription machinery, e.g., Swi4 and Mbp1, since both of these proteins carry multiple Rad53 consensus sites. Other kinases, such as Dun1, Chk1, and Mec1, are also activated and could phosphorylate Swi6 and additional G1
targets. This is consistent with our findings that SWI6
-null mutants have a G1
checkpoint defect and that the rad53-11 swi6Δ
double mutant has a more extreme G1
checkpoint-deficient phenotype than either mutation alone.
It is unclear at present what biochemical activities of Swi6 are affected by Rad53 phosphorylation and DNA damage in general. MMS-induced damage does not abolish Swi6 nuclear localization in G1
. Nor does it prevent Swi6 from binding to the CLN1
promoter in vivo in RAD+
cells (J. M. Sidorova, unpublished data). Loss of Sin3 repressor, or of Stb1, which may connect Swi6 to the Sin3 repressor complex (24
), does not give rise to a G1
/S checkpoint defect (Sidorova, unpublished). Perhaps other, yet-unidentified interacting factors may confer the effects of Swi6 phosphorylation by Rad53 on the G1
Checkpoints are likely to include multiple layers of control imposed on key events, with dozens of substrates targeted by more than one kinase and phosphorylated either in an additive or competitive fashion. A network thus organized would be highly responsive to changing conditions and failsafe. Understanding of the Rad53 phosphorylation site preferences provides a new tool for dissecting the complexity of checkpoint controls in budding yeast.