In this study, we sought a quantitative understanding of dynamic changes in Cdk1 specificity over the budding yeast cell cycle. We found that a gradual change of specificity is an intrinsic feature of the cyclin-Cdk1 system, and it seems to have evolved to prevent Cdk1 from prematurely triggering mitosis using the built-in delay mechanisms created by the weaker activity of the earlier cyclin-Cdk1 forms toward Clb2-specific mitotic targets. This weak activity, however, does not prevent the phosphorylation of G1 and S phase Cdk substrates, whose targeting is accomplished by RXL-hydrophobic patch interactions (Clb5, Clb3), by a hydrophobic LP docking site (Cln2), or by different consensus sites (Cln2) (A). Thus, as a general conclusion, we can state that Cdk1 specificity is periodically changing in the course of the cell cycle.
Dynamics of Cdk1 Substrate Specificity during the Cell Cycle
Throughout the eukaryotes, mitotic Cdk1 activity is regulated in part by inhibitory phosphorylation by the tyrosine kinase Wee1. The homolog of Wee1 in budding yeast, Swe1, has been shown to exhibit a differential ability to phosphorylate and inactivate different cyclin-Cdk1 complexes, with the highest inhibitory potency toward Clb2 and gradually lower for the earlier cyclin-Cdks (Hu and Aparicio, 2005; Keaton et al., 2007
). We determined that the relative differences in Swe1 specificity are about the same order of magnitude as the gradual change in the intrinsic specificities of Cdk1 (Figure S2
). These data reveal another remarkable gradual phenomenon in the cyclin-Cdk1 system in yeast. The identity of the cyclins in complex with Cdk1 is sensed by the Swe1 kinase, which then applies inhibitory pressure in proportion to the intrinsic specificity of the complex. In fact, the same structural elements may control the accessibility of the active site of different cyclin-Cdk1s for substrate and for the kinase domain of Swe1.
The cyclin specificity model is an alternative to the quantitative model of the cyclin response, according to which different levels of accumulating Cdk1 activity trigger different cell cycle events (Coudreuse and Nurse, 2010; Stern and Nurse, 1996
). The major weakness of a system behaving solely according to the quantitative model is that the temporal resolution of events depends entirely on the use of substrates with wide differences in specificities toward cyclin-Cdk1. According to this model, early events of the cell cycle would be switched on by optimal substrates that are extensively phosphorylated at low Cdk1 activity levels, while later events must be triggered by suboptimal substrates that are phosphorylated only at high kinase levels. For this model to work, the differences between the cyclin levels triggering S phase and M phase must be very large, with S phase triggered by a small fraction of the mitotic Cdk1 activity levels. This would apparently make S phase very vulnerable to even mild deviations and fluctuations of the cyclin signal, which could lead to premature initiation of later events.
While the quantitative model is apparently not sufficient to describe the function of cyclins, it also appears that the other extreme, according to which docking mechanisms are used throughout the cycle, is not correct either. Instead, each cyclin-Cdk1 complex in the sequence has improved intrinsic specificity culminating with the mitotic complex, which relies almost entirely on the intrinsic consensus-site specificity and minimally on docking sites. Thus, Cdk1 broadens its specificity gradually for wider and wider fractions of the proteome.
With several hundred Cdk1 targets in the cell, most of which contain multiple Cdk1 consensus sites, the total substrate concentration for Cdk1 could be hundreds of micromolar or even low millimolar. This target pool is unphosphorylated in early G1 phase, and because multiple targets of an enzyme act as competitive inhibitors relative to one another, the higher KM values prevent early cyclin-Cdk1 complexes from being inhibited by the pool of unphosphorylated targets, as illustrated in B. The highlighted inhibition term (1 + STOT/KM,TOT) in the modified Michaelis-Menten equation raises the apparent KM for any given substrate and thereby decreases its phosphorylation rate. The panel on the left side of B schematically describes a simplified system following the cyclin specificity model, with three cyclins synthesized in sequence. As the concentration of the bulk unphosphorylated Cdk1 substrate pool decreases in correlation with the KM,TOT values (phosphorylated residues have considerably lower affinity for the active site), the inhibition term is kept at a constant low level, allowing each complex to perform its specific function using the cyclin-specific docking sites, while being unhindered by the bulk substrate pool, at the point of the cycle where it is assigned. The panel on the right side of B shows the system behaving according to the quantitative model, based on accumulation of a single cyclin similar to mitotic cyclin 3 with respect to specificity. In such a system, a full phosphorylation rate of any target is achieved only after substantial accumulation of cyclin when a large part of the total Cdk substrate pool is phosphorylated and the inhibition term is minimal.
We speculate that cyclin specificity evolved as follows. Early eukaryotic cell cycle control depended on a single cyclin system, in which S phase was switched on by low levels of kinase activity, using substrates bearing the optimal consensus motifs (S/TPXK/R), after which higher kinase levels triggered M phase by phosphorylation of suboptimal motifs (S/TP). Due to a relatively low complexity of regulation, there were few substrates and little substrate competition. Eventually, evolution of cyclin docking sites (LP and/or RXL) provided additional affinity for S phase targets carrying cyclin docking motifs in addition to the optimal consensus phosphorylation site. At this stage, more regulatory complexity also evolved, requiring more substrate sites and therefore more competition and mutual inhibition by substrates in S phase. This competition limited the possibilities for further complexity, until multiple cyclins appeared with weaker earlier complexes to reduce the competition at the active site level. As earlier cyclins became weaker activators of Cdk1 (higher KM), early targets with docking motifs, like Sld2 and Sic1, would be more rapidly phosphorylated as competition from optimal substrates decreased. Thus, as a result of cyclin docking interactions, suboptimal intrinsic specificity became an advantage, allowing greater complexity in Cdk-triggered processes and control systems. Such a system can survive (with some difficulty) using a single mitotic cyclin, which will hit the S phase targets (containing S/TPXK/R motifs) earlier than the later targets (containing S/TP motifs), because the former ones have lower KMs and are able to outcompete the latter ones.
The most surprising outcome of our work is that both Clb- and Cln2-Cdk1 possess some strikingly different elements in their phosphorylation site consensus sequence. While the general activity of a protein kinase has been shown to be regulated in a wide variety of cases, the modulation of the primary structure specificity profile by a regulatory subunit has to our knowledge not been reported. Additionally, Cln2 specificity was found to be strongly enhanced by a previously unreported docking interaction involving a hydrophobic stretch in Sic1. While the precise structural motifs of this docking interaction are yet to be established, we speculate that a hydrophobic pocket on the cyclin could serve as a docking site for this motif.
The concept of dynamically changing Cdk1 specificity could be used to explain previously reported phenotypes obtained by genetic manipulations of cyclin genes in yeast. For example, it may explain why yeast cells lacking both Clb5 and Clb6 experience only an S phase delay (Schwob and Nasmyth, 1993
), while double mutants of Clb2 and Clb3 or Clb2 and Clb1 are inviable and arrest prior to mitosis (Mendenhall and Hodge, 1998
). In the first case, after a delay, the intermediate Clb3-Cdk1 and the intrinsically most potent Clb2-Cdk1 can phosphorylate the S phase targets at a reasonable rate. In the two last-mentioned cases, however, the poor ability of early B-type cyclins to promote the phosphorylation of Clb2-specific targets apparently makes the cells incapable of initiating mitosis. Perhaps for the same reason, partly stabilized Clb5 is unable to block mitotic exit (Wasch and Cross, 2002
), when compared with a strain overexpressing stabilized Clb2 (Surana et al., 1993
). Interestingly, our cyclin specificity model also fits well with the observations that fully stabilized Clb5 is still capable of mitotic exit but incapable of S phase initiation, which we propose is due to constant Clb5-specific phosphorylation of S phase targets of the preRC, preventing origin licensing (Sullivan et al., 2008
). We also propose that blocking Clb2 activity toward Clb5 targets may be important at certain stages of mitosis (e.g., for the dephosphorylation switch of “Fin1-like” targets [Woodbury and Morgan, 2007
]) and could be accomplished, again, by the inability of Clb2 to use the RXL-hp docking mechanism.
It will be important to determine if higher eukaryotes possess a similar dynamic specificity scheme. The general conservation of the gradual model remains to be shown, but our studies of budding yeast shed light on somewhat puzzling and unexpected cyclin knockout studies in mice: if a cyclin is deleted, the process it was meant to trigger is delayed until the activated Cdk, through the synthesis of other accumulating cyclins, reaches the levels where the net value of (kcat
[E] corresponds to the threshold of the trigger. If one deletes the weaker early cyclins, compensation by the later, stronger ones is more likely than the opposite situation, as mice lacking cyclins E and D have been shown to be viable, while cyclins B1 and A2, for example, are the most nonredundant of all cyclins and are required for embryo viability (Satyanarayana and Kaldis, 2009
In conclusion, we have shown that in the course of the cell cycle, different cyclins gradually change the substrate specificity of Cdk1 at the active-site level. This modulation of specificity, when combined with docking site interactions, reveals the dynamic nature of continuous specificity changes of Cdk1 in the course of the cell cycle and provides a wide range of selective switchpoints for different cell cycle transitions.