|Home | About | Journals | Submit | Contact Us | Français|
Sister-chromatid cohesion is established during S phase when Eco1 acetylates cohesin. In budding yeast, Eco1 activity falls after S phase due to Cdk1-dependent phosphorylation, which triggers ubiquitination by SCF-Cdc4. We show here that Eco1 degradation requires the sequential actions of Cdk1 and two additional kinases, Cdc7–Dbf4 and the GSK-3 homolog Mck1. These kinases recognize motifs primed by previous phosphorylation, resulting in an ordered sequence of three phosphorylation events on Eco1. Only the latter two phosphorylation sites are spaced correctly to bind Cdc4, resulting in strict discrimination between phosphates added by Cdk1 and Cdc7. Inhibition of Cdc7 by the DNA damage response prevents Eco1 destruction, allowing establishment of cohesion after S phase. This elaborate regulatory system, involving three independent kinases and stringent substrate selection by a ubiquitin ligase, enables robust control of cohesion establishment during normal growth and following stress.
Protein kinases and ubiquitin ligases are key regulators of the cell division cycle. The Cyclin-Dependent Kinases (Cdks) are particularly important, phosphorylating hundreds of substrates involved in numerous cellular processes1,2. Cdk function is linked to another important regulatory mechanism, protein degradation, by the Skp1–Cullin–F-box (SCF) ubiquitin ligase3. The F-box subunit of SCF recruits substrates by interacting with specific sequence motifs (degrons), which often contain phosphorylation sites. Phosphodegrons created by Cdks and other kinases tie SCF to many important cell cycle processes. In yeast, for example, phosphorylation of several cell-cycle regulators by Cdk1 leads to their recognition by the F-box protein Cdc4 (Supplementary Table 1). The vertebrate Cdc4 ortholog, Fbw7, also targets many phosphoproteins involved in cell proliferation and tumorigenesis4.
The preferred phosphodegron for Cdc4 and Fbw7 has been studied in considerable detail. Key insights came from phosphopeptide binding experiments, which showed that Cdc4 prefers peptides containing a phospho-serine or -threonine followed by a proline and preceded by hydrophobic residues: I/L/P-I/L-pS/pT-P-RKY4 (where x refers to disfavored residues)5. This consensus overlaps with the consensus motif of Cdk1 (S/T*-P-x-K/R, where S/T* indicates the phosphorylated residue and K/R enhances affinity). It was subsequently found that Cdc4 has a higher affinity for peptides containing two phosphorylated sites6, and that local sequence context is less critical than diphosphorylation7. SCF dimerization might also enhance binding of a multiply phosphorylated substrate8, thereby increasing the processivity of ubiquitination9.
Among its many cell cycle functions, SCF-Cdc4 helps regulate the generation of cohesion between sister chromatids as they are synthesized during S phase10. Sister-chromatid cohesion is established during S phase when the conserved protein Eco1 (also known as Ctf7) acetylates the Smc3 subunit of cohesin (refs. 11–14). Eco1 levels drop after S phase due to an increased degradation rate that requires Cdc4 and Cdk1 phosphorylation sites in Eco1, suggesting that phosphorylation of Eco1 by Cdk1 results in the generation of phosphodegrons that interact with Cdc4 (ref. 10). The drop in Eco1 abundance prevents cohesion establishment after S phase. Cdk1 and Cdc4 therefore collaborate to regulate chromosome segregation by preventing excess chromatid cohesion.
Cohesion establishment can also occur after S phase in cells with DNA damage15–17, enabling efficient DNA repair18. DNA damage leads to phosphorylation of the cohesin subunit Scc1 (also called Mcd1) by Chk1 (ref. 19), which is thought to promote Scc1 acetylation at Lys84 and Lys 210 by Eco1 (ref. 20). The reactivation of cohesion establishment by DNA damage might also depend on the stabilization of Eco1 (ref. 10).
We set out to understand the mechanisms by which phosphorylation promotes the degradation of Eco1. We found that Eco1 degradation depends on a remarkable cascade of phosphorylation events involving Cdk1 and two additional kinases, Cdc7 and Mck1. We find that after priming by Cdk1 at one site, Cdc7 and Mck1 sequentially phosphorylate adjacent sites to create a diphosphodegron with high affinity for Cdc4, leading to degradation. We further show that Cdc7 inhibition upon DNA damage stabilizes Eco1, thereby reactivating cohesion establishment. Our experiments highlight the complex regulatory possibilities that can be achieved by multisite phosphorylation of key regulatory proteins.
Eco1 contains four potential Cdk1 phosphorylation sites in a region of predicted disorder (Fig. 1a). We showed previously that mutation of all four sites fully stabilized Eco1 (ref. 10). Here, we constructed Eco1 mutants lacking single phosphorylation sites and measured their degradation rates in metaphase-arrested cells treated with the translational inhibitor cycloheximide. We found that the central two sites (Thr94 and Ser99) are each necessary for Eco1 degradation (Fig. 1b). Mutation of Ser105 had little effect, while mutation of Ser67 promoted turnover (perhaps due to disruption of the nearby zinc finger).
Eco1 displays a mobility shift on gels supplemented with a phosphate-binding reagent10,21. Mutation of Ser99 abolished the mobility shift, suggesting that Ser99 phosphorylation is responsible for the shift (Fig. 1c). We showed previously that this shift occurs early in S phase, long before Eco1 is degraded, and also following DNA damage, when Eco1 is stabilized10. Thus, phosphorylation at Ser99 alone is not sufficient for Eco1 degradation.
Ser99 matches the full Cdk1 consensus sequence (S/T*-P-x-K/R), whereas Ser67, Thr94, and Ser105 are only minimal matches (S/T*-P). To assess Cdk1 activity toward these sites, we carried out kinase reactions in vitro with purified cyclin–Cdk1. Surprisingly, we found that the S99A mutation blocked all phosphate incorporation by Clb2–Cdk1, to the same degree as mutation of all four sites (Fig. 1d). Similar results were obtained with Cln2–Cdk1, Clb3–Cdk1, and Clb5–Cdk1, suggesting that Ser99 is the only efficient site for phosphorylation in vitro by most, if not all, forms of Cdk1 (data not shown).
To obtain an unbiased understanding of Eco1 phosphorylation, we identified sites of phosphorylation in vivo by mass spectrometry, using Eco1-Flag3His6 purified from yeast cells in which CDC4 was transcriptionally repressed. We identified peptides phosphorylated at Thr94 and Ser99, as well as peptides with phosphates on Thr94 and the non-Cdk1 site Ser98 (Fig. 2a, b; Supplementary Fig. 1). Two other non-Cdk1 sites, Ser89 and Thr90, were also identified. Mutation of these residues had little (T90A) or no effect (S89A) on Eco1 stability (data not shown), so we focused our efforts on the other non-Cdk1 site, Ser98.
Ser98 is the first serine in an SSP motif, which is found at both the third and fourth Cdk1 consensus sites (Ser99 and Ser105). Mutation of both upstream serines (S98A S104A) fully stabilized Eco1 (Fig. 2c). Stabilization was not due to impaired Cdk1 activity toward Ser99, as Eco1 S98A S104A still displayed a mobility shift in vivo (Fig. 2d) and was phosphorylated by Cdk1 in vitro (Fig. 1d). Single mutations demonstrated that Ser98 phosphorylation is particularly important for Eco1 degradation, although Ser104 might also contribute (Fig. 2c).
In summary, Eco1 degradation depends primarily on phosphorylation at Thr94, Ser98, and Ser99. Only Ser99 is phosphorylated by Cdk1 in vitro, raising the possibility that other protein kinases modify the other two sites.
The protein kinase Cdc7–Dbf4 is known to phosphorylate the first serine in SSP motifs that are phosphorylated at the second serine22–25. Cdc7 was therefore an appealing candidate for the kinase that phosphorylates Ser98 in Eco1.
Cdc7–Dbf4 is analogous to Cdk1–cyclin, as the catalytic subunit Cdc7 is active only when bound to Dbf4. Cdc7–Dbf4 is also called Dbf4-Dependent Kinase (DDK). Like those of Cdk1, Cdc7 levels are constant through the cell cycle, whereas Dbf4 oscillates, rising in late G1 and declining in mitosis26,27. DDK is required for the firing of replication origins, where its target is the Mcm replicative helicase28. It also has functions in mitosis27 and meiosis29.
DDK is essential for viability but can be deleted in the background of the suppressor mutation MCM5-P83L (ref. 30). To assess the role of DDK in Eco1 turnover, we measured Eco1 degradation in dbf4Δ MCM5-P83L cells arrested in metaphase. Deletion of DBF4 resulted in full stabilization of Eco1 (Fig. 3a). Thus, both Cdk1 and DDK are necessary, and neither is sufficient, for Eco1 destruction.
As Thr94 is a poor Cdk1 site in vitro (Fig. 1d), we pursued the possibility that another protein kinase acts at this site. Interestingly, phosphorylation of Ser98 by DDK is predicted to create a consensus priming site for GSK-3 (glycogen synthase kinase-3), which phosphorylates amino acids four residues upstream of a previously phosphorylated residue (consensus motif: S/T*–x–x–x–pS/pT, where “x” is often proline)31,32. GSK-3 family kinases are the only kinases in yeast primed by +4 phosphorylation24. Furthermore, multiple substrates of vertebrate SCF-Fbw7 contain diphosphodegrons in which one site is phosphorylated by GSK-3 after priming of another site at the +4 position4,33.
Among the four GSK-3 homologs of S. cerevisiae, Mck1 has been linked to the broadest range of functions34. Intriguingly, it has been implicated in the degradation of three SCF-Cdc4 substrates: Hsl1 (ref. 35), Rcn1 (ref. 36), and Cdc6 (ref. 37).
We analyzed Eco1 turnover in metaphase-arrested cells individually deleted for each GSK-3 homolog. Strikingly, Eco1 was stabilized in a mck1Δ strain to the same extent as the Eco1-T94A mutant (Fig. 3b). Deletions of other GSK-3 paralogs had no effect, revealing a high degree of specificity among the GSK-3 family members.
Our results suggest that phosphorylation of Ser99 by Cdk1 primes Eco1 for phosphorylation at Ser98 by Cdc7, which then primes the protein for phosphorylation at Thr94 by Mck1. We tested this model directly by measuring kinase activities toward Eco1 peptides carrying phosphates at the predicted priming sites. Synthetic phosphopeptides (containing residues 91–103 of Eco1) were used instead of full-length Eco1 to ensure complete phosphorylation of priming sites.
Clb2-Cdk1 efficiently phosphorylated an unmodified peptide but not a peptide with phosphate at Ser99 (Fig. 3c), consistent with our other evidence that Ser99 is the sole major Cdk1 site. Interestingly, a phosphate on Ser98 also blocked Cdk1 activity. As mutation of Ser99 prevents all Cdk1 activity toward Eco1 (Fig. 1d), we suspect that Ser98 phosphorylation interferes with phosphorylation at Ser99.
DDK had no activity toward unmodified peptide, but rapidly phosphorylated the peptide with a phosphate at Ser99 (Fig. 3c). Phosphorylation was blocked by an alanine at position 98, but not at position 94, consistent with our priming hypothesis. Similarly, Mck1 phosphorylated an Eco1 peptide only if it contained a phosphate at Ser98, and only if it contained Thr94 (Fig. 3c). These data strongly support our model for the ordered phosphorylation of Eco1 by the three kinases.
We next investigated the contribution of each phosphate in Eco1 to recognition by the Cdc4 subunit of SCF. Cdc4 displays a preference for diphosphorylated degrons, often grouped in clusters6,7,38,39. In those substrates with defined diphosphodegrons, phosphorylated residues are 2 or 3 residues apart, not the 4 residues between Thr94 and Ser99 in Eco1 (see Supplementary Table 1). Similarly, targets of vertebrate Fbw7 contain phosphorylation sites 3 residues apart4. We therefore hypothesized that phosphorylation of Ser98 brings the phosphate to within the ideal range for Cdc4 binding. We tested this idea by measuring the binding of various phosphopeptides to recombinant Cdc4–Skp1 in vitro. Phosphopeptides were synthesized, tagged with a fluorescent molecule (FITC), and incubated with increasing concentrations of Cdc4. Fluorescence polarization was used to measure bound peptide.
A peptide with phosphates at all three sites (“94-98-99”) bound Cdc4 with an equilibrium dissociation constant of 10.3 μM (Fig. 4a). Affinity was much lower in the absence of phosphate at Ser98, but not in the absence of phosphate at Ser99. We observed moderate binding of a peptide phosphorylated at Thr94 alone, likely because the sequence around Thr94 is similar to the consensus Cdc4 binding motif5. This binding seems insufficient to mediate degradation in vivo, however, as the S98A and S99A mutations completely stabilize Eco1, and monophosphorylation at Thr94 is unlikely to occur since Mck1 requires priming at Ser98. Single phosphates at Ser98 or Ser99 had no effect on binding to Cdc4. Thus, phosphorylated Thr94 is the primary binding site whose affinity is supplemented by phosphorylated Ser98, as seen in other Cdc4 substrates6,7,40.
These results suggest that the distance between phosphates is critical for Cdc4 recognition, as suggested by alignment of known Cdc4 and Fbw7 degrons (Supplementary Table 1 and ref. 4). To systematically test the importance of phosphate spacing, we measured Cdc4 binding to peptides containing variations in the distance between phosphates (Fig. 4b). Insertion of isoleucine (which should not alter Cdc4 binding5) before Ser98, moving the DDK site to the position of the Cdk1 site (peptide “94-98+I98”), removed the effect of the second phosphate. Removing a residue between Thr94 and Ser98 (peptide “94-99ΔN97”) made phosphorylation at Ser98 unnecessary. The slightly reduced binding of this peptide relative to “94-98” could be due to local sequence context. Thus, efficient Cdc4 binding requires a maximum of 3 residues between phosphates.
We next tested whether there is a minimum distance between phosphates. Removing Asn97 from a peptide with phosphate at Ser98 (peptide “94-98ΔN97”) did not reduce high affinity binding (Fig. 4b). However, deleting both Leu96 and Asn97, leaving only one residue between phosphates (peptide “94-98ΔLN”), blocked the effect of the second phosphate. Therefore, Cdc4 binds with high affinity only to substrates with 2 or 3 residues between phosphates.
We also tested the importance of phosphate spacing in vivo. Our model predicted that moving the sites closer by one residue (Eco1 N97Δ) should remove the requirement for DDK in Eco1 degradation. Indeed, Eco1 N97Δ was highly unstable in a dbf4Δ strain (Fig. 5a), even when Ser98 was mutated (Fig. 5b). The extreme instability of this mutant suggests that the activity of Cdk1 and Mck1 towards Eco1 is higher than that of DDK. Consistent with this explanation, the small amount of Eco1ΔN97 protein did not display a mobility shift in metaphase-arrested cells (Supplementary Fig. 2). We suspect that phosphorylation of this mutant by Cdk1 results in immediate degradation, whereas Cdk1-phosphorylated wild-type Eco1 is more stable due to limited phosphorylation by DDK.
As expected, Eco1 N97Δ was highly stabilized in a mck1Δ strain (Fig. 5c). Stabilization was not complete, perhaps indicating that deletion of Asn97 causes non-specific protein turnover. However, Eco1 N97Δ displays a robust mobility shift in mck1Δ cells, indicating that it is still targeted by Cdk1 in vivo (Supplementary Fig. 2). Moreover, additional deletions in this region (see below) fully stabilized the protein.
Eco1 was fully stabilized by insertion of isoleucine upstream of Ser98 (Eco1+I98) or by deletion of Leu96 and Asn97 (Eco1 L96Δ N97Δ) or all three intervening residues (Eco1 Δ96-98) (Fig. 5d). These results are generally consistent with our peptide binding results in vitro, with the exception of the Eco1 L96Δ N97Δ protein, in which phosphorylation sites are spaced at the same distance as the “94-98ΔN97” peptide that bound well to Cdc4 (Fig. 4b). Stabilization of this mutant is not due to reduced Ser99 phosphorylation by Cdk1 (Supplementary Fig. 2). Instead, stabilization is likely due to lack of phosphorylation by Mck1, as removal of Leu96 and Asn97 disrupts the correct spacing of the GSK-3 recognition motif32.
To further characterize the DDK-independent degradation of Eco1 N97Δ, we examined protein levels over the cell cycle. As seen previously10, wild-type Eco1 accumulated during S phase, declined in mitosis, and reappeared after mitosis (Fig. 6a). In contrast, Eco1 N97Δ levels declined in late G1, 20 minutes earlier than wild-type Eco1 (Fig. 6b). Destruction of Eco1 N97Δ began at the time that wild-type Eco1 displayed the Ser99-dependent mobility shift, but the mobility of Eco1 N97Δ did not change over the entire cycle. Eco1 N97Δ accumulated as cyclin levels dropped in mitosis. Thus, Eco1 N97Δ levels were inversely related to the amount of Clb–Cdk1 activity, suggesting that Cdk1 activity determines the timing of Eco1 N97Δ turnover and that Mck1 activity may be constitutive throughout the cell cycle. These results also suggest that the normal timing of Eco1 destruction in late S phase is determined by changes in DDK activity toward Ser98.
Our studies of the Eco1 N97Δ mutant suggested that the requirement for DDK provides a mechanism to delay Eco1 degradation until late S phase. We also considered a second rationale for regulation by DDK, based on our previous observation that Eco1 is stabilized by DNA damage10: DDK activity is thought to be reduced in yeast cells following DNA damage41,42 (although perhaps not towards all substrates26,30,43). DNA damage causes extensive phosphorylation of Dbf4, probably by the kinase Rad53 (refs. 42,44,45). Mutations in Dbf4 that reduce phosphorylation by Rad53 lead to inappropriate firing of late replication origins following DNA damage, suggesting that Dbf4 phosphorylation inhibits DDK function46,47.
To test if DDK inhibition prevents the degradation of Eco1 in a damage response, we used dbf4-m25 cells carrying a mutant Dbf4 that lacks Rad53 phosphorylation sites and is therefore refractory to damage-dependent inhibition46. Activation of a DNA damage response with hydroxyurea (HU) led to Eco1 stabilization (Fig. 7a), as seen previously10. However, Eco1 was rapidly degraded in dbf4-m25 cells treated with HU (Fig. 7a), similar to mec1Δ cells that lack the damage response. Maintenance of DDK activity in damage is thus sufficient to destabilize Eco1, supporting the role of Cdc7 in Eco1 turnover and indicating that Cdk1 and Mck1 activities toward Eco1 are not affected by the damage response. Also consistent with our model, DNA damage failed to stabilize the Eco1 N97Δ protein, which does not require DDK activity for turnover (Fig. 7b).
Discovering the molecular basis of Eco1 stabilization upon DNA damage allowed us to test whether it is required to establish new cohesion after damage. We used a variation of a previous protocol to measure the generation of damage-induced cohesion in metaphase17. In this experiment, cells establish normal cohesion in S phase using an Scc1 cohesin subunit containing a TEV protease site. Cells are arrested in metaphase, followed by treatment with a DNA-damaging agent and induction of Scc1 lacking the TEV site. TEV protease is then expressed to remove cohesion established during S phase. Only the cohesin expressed during the damage response is left to hold the LacI-GFP-tagged sister chromatids together.
In this assay, wild-type cells fail to establish new cohesion and about 50% of the cells contain separated LacI-GFP foci. Activation of establishment lowers this to about 30% (refs. 10,15,17,19). As seen previously, we found that DNA damage induced new cohesion in wild-type cells (Fig. 7c). This effect depended on lysines 84 and 210 in Scc1, which are likely Eco1 acetylation sites20. Damage did not induce new cohesion in dbf4-m25 cells, but this was rescued by mutation of the two DDK consensus sites in Eco1, Ser98 and Ser104 (Fig. 7c). Additionally, the DDK-independent Eco1 N97Δ protein did not establish new cohesion after damage. Thus, the reactivation of cohesion establishment upon DNA damage depends on the increased levels of Eco1 caused by reduced DDK activity.
We report the discovery of an intricate series of phosphorylation events that controls the generation of sister chromatid cohesion through the phosphorylation and ubiquitination of the cohesion-promoting enzyme Eco1 (Fig. 7d). Our data suggest that Cdk1-dependent phosphorylation of one Eco1 residue, Ser99, primes for phosphorylation at the adjacent site, Ser98, by the cell cycle-regulated kinase Cdc7–Dbf4 (DDK) in late S phase. This then promotes a third phosphorylation at Thr94 by the GSK-3 homolog Mck1. The result is a precisely spaced pair of phosphates at Ser98 and Thr94 that create an interaction site for the SCF ubiquitin ligase subunit Cdc4. Because Dbf4 is inhibited by the DNA damage response, full Eco1 phosphorylation is blocked following DNA damage, leading to high Eco1 levels and establishment of new cohesion after S phase.
The participation of Cdc7 in Eco1 degradation adds another layer of regulation to cohesion establishment and permits differential regulation in different cellular states. Regulation by DDK would be inconsequential, however, without the rigorous substrate discrimination we observed for Cdc4 and Mck1. The ability of Cdc4 to distinguish a phosphate attached by DDK from an immediately adjacent phosphate added by Cdk1 is remarkable. Likewise, the strict spacing of the GSK-3 consensus sequence means that the distance between phosphates in GSK-3 targets matches the ideal spacing for Cdc4 binding4. These features of Eco1 regulation together impose a set of strict conditions necessary to shut off cohesion establishment, enabling establishment activity to be correctly timed in the cell cycle and responsive to environmental perturbations.
This modularity and convergence of inputs is a common theme in biological regulatory systems. Regulatory proteins that respond only in the presence of multiple upstream inputs are called coincidence detectors and operate as logical ‘AND’ statements. Eco1 degradation represents an unusual case of a triple ‘AND’ statement integrating at least three separate inputs. Changes in any one input—oscillations in Cdk1 activity during the cell cycle or DDK inhibition after DNA damage, for example—dramatically influence the output, Eco1 degradation, and the biological process it controls, cohesion establishment.
Critical to the integration of the three Eco1 inputs is our finding that tight Cdc4 binding requires two or three amino acids between phosphates (Fig. 4). In crystallographic structures of Cdc4 bound to phosphopeptides with three residues between phosphates, the anchor phosphothreonine binds the center of the WD40 domain and the C-terminal phosphoserine interacts with Arg443, Ser464, Thr465, and Arg485, near the edge of the domain6,39. The C-terminal end of the peptide in these structures forms a small loop, allowing the second phosphoserine to contact the above residues. It is reasonable to predict that a degron with only two residues between phosphates could interact with the same side chains in Cdc4, but perhaps without making a loop. A peptide with four residues separating phosphates, however, may not be able to make the necessary contortions to align the second phosphate, and a degron with only a single intervening residue is probably not long enough to allow simultaneous interaction with both phosphate-binding sites.
Diphosphorylated Eco1 peptides have an affinity for Cdc4 that is similar to that of Sic1 phosphopeptides but weaker than those of the high-affinity substrates Tec1 and Ash1 (Supplementary Table 1). This moderate affinity could represent a selected trait that prevents Eco1 levels from getting too low after S phase, allowing cohesion establishment to be more easily reactivated. Another possibility is that the peptides do not fully recapitulate all of the binding contacts of the full-length substrate. Upstream residues could make extended interactions with Cdc4, as suggested by recent studies of long peptides from Sic1 (ref. 39).
It is also possible that additional phosphorylation sites in Eco1 contribute to Cdc4 binding. These sites might include the second SSP motif (Ser104-Ser105), as suggested by the slightly additive effect of the S98A and S104A mutations (Fig. 2c), or possibly the more N-terminal phosphates identified by mass spectrometry (Ser89 and Thr90). Mck1 may catalyze phosphorylation of Thr90, since Thr94 is at the correct priming distance and GSK-3 kinases are known to sequentially phosphorylate substrates by creating their own consensus sites31,32. Phosphates at Thr90 and Thr94 could create another Cdc4 degron like that in Ash1, which contains three phosphates that create two redundant overlapping Cdc4 binding sites40. Ser89 might be phosphorylated by DDK after priming at Thr90, but this likely does not affect Eco1 degradation. Ser104 is too far from Ser99 to constitute a diphosphodegron, but might bind to another molecule in a Cdc4 dimer, as suggested for the interaction of human Fbw7 with its substrate cyclin E (ref. 8).
Our observation that Cdk1 only phosphorylates Ser99 was unexpected, as it shows that some S/T-P sequences are not efficiently targeted by Cdk1, even when they appear accessible to another kinase. Since Cdk1 was unable to phosphorylate Thr94 even in a short peptide (Fig. 3c), this substrate discrimination could be due to the surrounding primary or secondary sequence, though the exact cause is unclear. Another surprising result from our peptide phosphorylation studies was that a phosphate at the -1 position blocked Cdk1 activity (Fig. 3c).
Our evidence suggests that the timing of Eco1 degradation in late S phase is determined by DDK (Fig. 6b). These results are puzzling because Cdc7 activity is thought to be determined solely by binding to Dbf4, which is present throughout S phase26,30. DDK activity toward Eco1 might be limited until after the kinase completes its critical functions in the firing of replication origins. Interestingly, overexpression of replication initiation factors, including Dbf4, causes late replication origins to fire earlier in S phase48,49, suggesting that Dbf4 levels are limiting for origin firing in late S phase. We found, however, that overexpression of DBF4 has little effect on the timing of Eco1 degradation (data not shown). Thus, DDK activity toward Eco1 appears to be restrained during S phase by some mechanism that is not simply dependent on limiting Dbf4 levels. Further studies of other DDK substrates will be required to determine the regulatory mechanism that controls the timing of DDK activity toward Eco1.
The adaptive value of GSK-3 involvement in Eco1 degradation is less clear than that of DDK, but its recognition of primed substrates enables it to serve as a sensor for the activity of other kinases. This prevents attachment of the anchor phosphate at Thr94 until DDK is active, preventing spurious low-level binding of Cdc4 to the Thr94 monophosphate. The requirement for Mck1 ensures that Eco1 binding to Cdc4 goes from zero (“99” and “98-99” peptides in Fig. 4) to maximal (“94-98-99” peptide) with no intermediate. The system would not have this property if Cdk1 phosphorylated Thr94. Additionally, Cdk1 phosphorylation of Thr94 might require a basic residue at the +3 position, which would weaken Cdc4 binding5. Given their mutually exclusive consensus motifs, it will be interesting to see how many of the anchor phosphates in Cdc4 degrons are attached by Cdk1. In contrast, the GSK-3 motif overlaps perfectly with that of Cdc4: both proteins prefer substrates with three residues between phosphorylation sites. Thus, GSK-3 is ideally suited to provide the N-terminal phosphate in a Cdc4 degron. Consistent with this idea, metazoan GSK-3 and SCF have many substrates in common4,33.
Mck1 is responsive to various cellular stresses34, all of which seem to increase Mck1 activity towards certain substrates, possibly via priming by stress-activated kinases. These stress response pathways may not impinge on cohesion regulation, however, since Mck1 activity does not seem to be limiting for Eco1 degradation (Fig. 6b). The activity profiles of yeast GSK-3 homologs have not been as well characterized as those of their metazoan counterparts, but there are likely to be differences since GSK-3 family members in other eukaryotes tend to be active in cellular resting states and inhibited by growth factors31, whereas Mck1 may be constitutively active37. Discovery of conditions that inhibit Mck1 activity would reveal new circumstances in which Eco1 could be reactivated, possibly indicating that DNA damage may not be the only stressor to modulate chromatid cohesion.
The results described in this study underscore the utility of post-translational modifications in precisely coordinating essential cell biological processes. To create an exact duplicate of an entire cell, dozens of diverse events must be properly organized and executed faithfully every generation. Just as multicellular development requires an extremely well-orchestrated series of interdependent events, so too must the events of the cell division cycle be exquisitely coordinated. An elegant regulatory system has thus evolved to ensure the fidelity of cell reproduction, and central to this is the accurate segregation of the genetic material.
Yeast strains are derivatives of the S288C strain BY4741 (see Supplementary Table 2 for a list of strains and plasmids). Genetic manipulations were performed using standard methods50. Cell cycle arrests were achieved using 15 μg/ml nocodazole, 15 μg/ml alpha-factor, 2 μM 4-nitroquinoline, or 200 mM hydroxyurea for 3 h at 30°C; all arrests were confirmed by flow cytometric analysis of DNA content. Budding index and binucleate formation were counted using cells fixed in 0.02% sodium azide and either light microscopy or epifluorescence after staining DNA with 1 μg/ml 4′,6-diamidino-2-phenylindole (DAPI). Protein degradation was analyzed using 100 μg/ml cycloheximide to inhibit protein synthesis, followed by cell lysis by bead beating in urea buffer (20 mM Tris-HCl pH 7.4, 7 M urea, 2 M thiourea, 4% CHAPS).
Primary antibodies were anti-myc (9E10, Covance, 1:1000 dilution), anti-PAP (P 1291, Sigma, 1:5000), and anti-Pgk1 (22C5D8, Invitrogen, 1:10,000). Secondary antibodies coupled to horseradish peroxidase from GE Healthcare were used at 1:10,000 dilution. To visualize Eco1-Ser99 phosphorylation, the resolving portions of SDS-polyacrylamide gels were supplemented with 10 μM Phos-tag reagent21 made using standard chemical techniques, 50 μM MnCl2, and two-fold excess ammonium persulfate and TEMED.
Eco1-Flag3His6 was purified from PGALS-HA3-CDC4 sic1Δ yeast cells grown overnight in 2% galactose, followed by dilution into 0.033% galactose. Following another day of growth, CDC4 expression was repressed by addition of 2% dextrose for 3.5 h at 30°C and 4°C for 45 min. Cells were harvested, frozen in liquid N2 and lysed by blending in lysis buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 0.1% NP-40 plus phosphatase and protease inhibitors). Lysates were spun at 40,000 rpm for 1 h at 4°C and the supernatant incubated with Flag antibody-coupled magnetic beads for 19 h at 4°C. Beads were washed three times in lysis buffer containing 300 mM NaCl. Bound protein was eluted in 1 M arginine-HCl (pH 3.5) at 4°C for 2 h, frozen in liquid N2 and stored at −80 °C.
Eluted fractions were diluted to 8 M urea in 100 mM Tris (hydroxyethylamine), pH 8.4, for denaturation and reduction of proteins with 5 mM Tris (2-carboxyethyl) phosphine for 30 min. Cysteine residues were acetylated with 10 mM iodoacetamide for 15 min in the dark. The sample was then diluted to 2 M urea with 100 mM Tris (hydroxyethylamine), pH 8.4, and 1 mM CaCl2, and 0.5 μg trypsin were added and incubated for 4 h at 37°C. The peptide sample was acidified to 5% formic acid and spun at 18,000 × g.
Tryptic phosphopeptides were analyzed by Multi-dimensional Protein Identification Technology (MudPIT) as previously described51, with the following modifications. 3-step MudPIT was performed where each step corresponds to 0, 25, and 100% buffer C (500 mM ammonium acetate) being run for 5 min at the beginning of a 2 h reverse phase gradient. Precursor scanning in the Orbitrap XL was performed from 400 to 2000 m/z with the following settings: 5 × 105 target ions, 50 ms maximum ion injection time, and 1 microscan. Data-dependent acquisitions of MS/MS spectra with the LTQ on the Orbitrap XL were performed with the following settings: MS/MS on the 8 most intense ions per precursor scan, 30K automatic gain control target ions, 100 ms maximum injection time, and 1 microscan. Multistage activation (MSA) fragmentation was performed at 35% normalized collision energy on the precursor and then on the expected neutral loss ions with reduced m/z’s of 32.70 (+3 ions), 49.00 (+2), and 98.00 (+1). Dynamic exclusion settings were as follows: repeat count, 1; repeat duration, 30 s; exclusion list size, 500; and exclusion duration, 60 s. Protein and phosphopeptide identification and phosphorylation analysis were done with Integrated Proteomics Pipeline (IP2, www.integratedproteomics.com). Tandem mass spectra were extracted to MS2 files from raw files using RawExtract 1.9.9 and were searched against the SGD S. cerevisiae protein database (1/5/2010) with reversed sequences using ProLuCID with the multistage search option52. The search space included all fully- and half-tryptic peptide candidates. The MSA search option in ProLuCID models both precursor fragment ions and neutral loss fragment ions in the XCorr calculation. Fragment ions from the phosphopeptide precursor were modeled with a mass increase of 79.9663 Da to the original peptide sequence for each phosphate present. The neutral loss fragment ions were modeled with a mass shift of −18.0106 from the loss of water from the original peptide sequence after the neutral loss of phosphate in the first activation step of MSA. Carbamidomethylation (+57.02146) of cysteine was considered as a static modification; phosphorylation (+79.9663) on serine, threonine, and tyrosine were considered as variable modifications. Peptide candidates were filtered to 0.1% FDR using DTASelect53,54.
Fig. 2a is the MSA CID fragmentation spectrum of the doubly phosphorylated peptide STGTIpTLNpSSPLKK. From initial unsuccessful attempts with CID fragmentation of the precursor ion only, we found that MSA (ref. 55) provided an adequate balance of sequence-specific and phosphorylation site-specific fragment ions for identification of phosphorylation sites on this complex phosphopeptide. The six potential sites of phosphorylation complicated phosphate localization and created multiple phosphate neutral losses (−98 Da) and water losses (−18 Da) during fragmentation. Fortunately, an adequate number of sequence-specific ions for peptide identification were identified with MSA, many with water losses. Additionally, phosphate molecules were detected intact on fragment ions y9 and y10, yet not on y5, unambiguously indicating that Thr94 and Ser98 were phosphorylated. Further, preferential cleavage N-terminal of the two proline residues within the sequence due to the proline effect56 conveniently aided generation of both sequence- and phosphorylation site-specific ions. Supplementary Fig. 1 provides additional support of both dual Thr94 and Ser98 phosphorylation and singly phosphorylated peptides. Due to co-migration and co-fragmentation of phosphopeptides with LC-MS/MS (ref. 57), we also found site-specific fragment ions for individual phosphorylation at Ser89, Thr90, and Ser99, as indicated in Fig. 2b and Supplementary Fig. 1. Ultimately, either the lower abundance of these phosphopeptides, or proline effect-favorable fragmentation pathways, may have hindered the identification of other phosphorylation sites.
Clb2-Cdk1 and Mck1 were purified from yeast cells overexpressing TAP-tagged versions, as described58. Cdc7–Dbf4-Flag from yeast cells was kindly provided by the lab of Stephen Bell. Cdk1 was incubated at room temperature with 1 μM GST-Eco1 (purified from bacteria) in a 30 μl reaction containing 25 mM HEPES (pH 8.0), 15 mM NaCl, 1 mM MgCl2, 1 mM DTT, 4 μCi γ-32P-ATP (3000 Ci/mmol), and 100 μM unlabeled ATP. Reaction products were analyzed by SDS-PAGE and autoradiography.
Peptide kinase assays were performed as above, in duplicate, using peptides synthesized by NeoBioSciences (Cambridge, MA) in 15 μl total volume with 10 μM peptide (Cdk1 reactions), 33 μM peptide (Cdc7 reactions), or 5 μM peptide (Mck1 reactions) for 45 min (Cdk1), 30 min (Mck1), or 15 min (Cdc7). Reactions were stopped by spotting on P81 phosphocellulose paper (Whatman), washed five times for 5 min each in 75 mM phosphoric acid, followed by a single 5 min wash in acetone. Peptides included an extra C-terminal lysine to enhance peptide binding to phosphocellulose. Air-dried papers were counted on a scintillation counter (Beckman Coulter LS 6500). Counts were converted to moles of total phosphate using a standard curve.
Phosphopeptides containing the Eco1 degron sequence plus a C-terminal tyrosine and cysteine to provide a means of absorbance detection and to conjugate a FITC fluorophore were synthesized by NeoBioSciences and purified by HPLC. Lyophilized peptides were resuspended in 1 ml 50 mM Tris-HCl (pH 7.4) to a final concentration of 1–3 mM and stored at −20°C. His6-Cdc4-LP6DelA GST-Skp1 was purified from bacteria as described7. Equilibrium binding of FITC-peptides to Cdc4 was measured by fluorescence polarization at room temperature in 96-well plates in triplicate (except the two highest Cdc4 concentrations, done in duplicate) in 40 μl 50 mM Tris-HCl (pH 7.4), 100 mM NaCl, 5 mM β-ME, 5% glycerol, 0.1 mg/ml BSA, plus 5 nM peptide and varying concentrations of Cdc4-Skp1. An LJL Biosystems Analyst AD plate reader was used with 495 nM excitation and 525 nM emission wavelengths, 10 reads per well, 100 ms integration time, and a z-height of 0.5 mm. Background fluorescence was normalized to a reaction lacking peptide. Data were analyzed using Prism software.
PSCC1-scc1(TEV268)-HA3:hphNT1:PSCC1:natNT2:PMET25-GFP-SCC1 trp1::TRP1:PGAL1-NLS-myc6-TEV-NLS2 ura3::lacO256:LEU2 his3::PCUP1-1-GFP-LacI:HIS3 cells were grown overnight in YEP media containing 2% raffinose, then arrested in mitosis with nocodazole for 3 h at 30°C. Cultures were transferred to synthetic media lacking methionine (to induce expression of wild-type Scc1 or Scc1 K84R K210R) and containing nocodazole plus either 333 μg/ml zeocin (first experiment) or 2 μM 4-NQO (second experiment) for 1.5 h to establish cohesion or not. Galactose was then added to 2% for 2 h to express TEV protease and cleave the cohesion established during S phase. 100 μM CuSO4 was added for the last 30 min to fully induce GFP-LacI expression, followed by fixation of cells with 4% paraformaldehyde for 8 min. Cells were washed, resuspended in dibasic sodium phosphate buffer, affixed to concanavalin A-coated slides, and visualized by epifluorescence microscopy.
We thank Stephen Bell (Massachusetts Institute of Technology, Cambridge, MA), David Toczyski (University of California, San Francisco, CA), Hiten Madhani (University of California, San Francisco, CA), and Kevan Shokat (University of California, San Francisco, CA) for strains and reagents, Mart Loog for advice with kinase assays, Amy Ikui for discussion of unpublished results, James Wohlschlegel for advice with mass spectrometry, Jeff Mugridge for assistance with fluorescence anisotropy, and Ellen Edenberg and Scott Foster for critical review of the manuscript. This work was supported by funding from the U.S. National Institute of General Medical Sciences (R01-GM069901 to D.O.M. and P41-GM103533 to J.R.Y.) and the National Center for Research Resources (P41-RR011823 to J.R.Y.).
AUTHOR CONTRIBUTIONSN.A.L. and D.O.M. conceived the experiments, N.A.L. conducted the biological and biochemical experiments, B.R.F. performed mass spectrometry, and B.R.F. and J.K.D. analyzed mass spectra under the guidance of J.R.Y. N.A.L. and D.O.M. wrote the manuscript.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.