Cell cycle-dependent phosphorylation of the DNA polymerase epsilon subunit, Dpb2, by the Cdc28 cyclin-dependent protein kinase.

DNA polymerase epsilon (Polepsilon), one of the three major eukaryotic replicative polymerases, is comprised of the essential catalytic subunit, called Pol2 in budding yeast, and three accessory subunits, only one of which, Dpb2, is essential. Polepsilon is recruited to replication origins during late G(1) phase prior to activation of replication. In this work we show that the budding yeast Dpb2 is phosphorylated in a cell cycle-dependent manner during late G(1) phase. Phosphorylation results in the appearance of a lower mobility species. The appearance of that species in vivo is dependent upon the Cdc28 cyclin-dependent protein kinase (CDK), which can directly phosphorylate Dpb2 in vitro. Either G(1) cyclin (Cln) or B-type cyclin (Clb)-associated CDK is sufficient for phosphorylation. Mapping of phosphorylation sites by mass spectrometry using a novel gel-based proteolysis protocol shows that, of the three consensus CDK phosphorylation sites, at least two, Ser-144 and Ser-616, are phosphorylated in vivo. The Cdc28 CDK phosphorylates only Ser-144 in vitro. Using site-directed mutagenesis, we show that Ser-144 is sufficient for the formation of the lower mobility form of Dpb2 in vivo. In contrast, Ser-616 appears not to be phosphorylated by Cdc28. Finally, inactivation of all three CDK consensus sites in Dpb2 results in a synthetic phenotype with the pol2-11 mutation, leading to decreased spore viability, slow growth, and increased thermosensitivity. We suggest that phosphorylation of Dpb2 during late G(1) phase at CDK consensus sites facilitates the interaction with Pol2 or the activity of Polepsilon

One of the crucial tasks of the cell division cycle is the faithful duplication of the genetic material once and only once per cell division cycle. Several mechanisms have evolved to ensure the fidelity of replication, including cell cycle-dependent regulation of replication proteins at the levels of the expression, degradation, subcellular localization, protein-protein interactions, and enzyme activity. Posttranslational modification plays a central role in regulating many of these processes. The cyclin-dependent protein kinases (CDKs) 1 and Cdc7 play prom-inent roles in the regulation of DNA replication, and many eukaryotic replication proteins are phosphorylated.
Genetic and biochemical assays with budding and fission yeasts and Xenopus egg extracts have identified two distinct phases in replication initiation. Assembly of pre-replication complexes (pre-RC) occurs in the absence of CDK activity and involves the association of Cdc6, Cdt1, and the MCM complex with the origin recognition complex (ORC) at replication origins. Subsequently, S phase-specific forms of the CDK and Cdc7-Dbf4 kinase activities lead to, and are required for, the recruitment of several proteins to replication origins, culminating in the recruitment of DNA polymerases ␣ and ⑀ (1)(2)(3)(4)(5).
In budding yeast, CDC7 and DBF4 (6) are essential genes but can be suppressed by a specific mutation in MCM5 (bob-1). This has been taken as evidence that MCM proteins are the only essential targets of Cdc7-Dbf4 kinase. Consistent with that observation, Mcm2, and to lesser extent Mcm4 and Mcm6, can be phosphorylated in vitro and in vivo by Cdc7-Dbf4. Phosphorylation may increase the affinity of the MCM complex for Cdc45 and its helicase activity (5,7).
CDKs, particularly the budding yeast Cdc28, have been shown to phosphorylate a number of DNA replication factors including components of the origin recognition complex, Orc1, Orc2 and Orc6, as well as Cdc6 and MCM proteins (8,9). Cdc6 is targeted for degradation via Cdc28-dependent phosphorylation (10). Phosphorylation of MCM proteins by Clb-Cdc28 causes nuclear exclusion of the MCM complex (11). Phosphorylation of human Mcm4 by CDK was reported to inhibit the DNA helicase activity of the MCM complex (12). Although phosphorylation of these proteins individually is not essential for replication, it has been shown in each case to negatively regulate their function. One potential regulatory role for those phosphorylations is to prevent reinitiation (13).
The only CDK-dependent phosphorylation event known to be essential for the initiation of DNA replication in yeast is the phosphorylation of Sld2/Drc1 (2). Only the phosphorylated form of Sld2 binds to Dpb11, and, in a strain expressing the non-phosphorylatable form of Sld2 initiation fails, presumably because the Dpb11-mediated loading of DNA polymerases ␣ and ⑀ is inefficient. Similarly, CDK-dependent phosphorylation regulates the interaction between the homologous fission yeast proteins Drc1 and Cut5 (14).
At least three DNA polymerases (Pol␣, ␦, and ⑀) are required for chromosomal DNA replication in eukaryotes (15). Pol␣ is involved in the initiation of replication and priming of Okazaki fragments. Pol␦ is believed to be the main polymerase in leading strand synthesis and lagging strand maturation. Conse-quently, Pol␦ may synthesize the bulk of chromosomal DNA. Budding yeast and human Pol⑀ consist of four subunits (16,17). The catalytic subunit and the second largest subunit are encoded by essential genes in both budding yeast (POL2 and DPB2) (18,19) and fission yeast (cdc20 and dpb2) (20 -22). The two smallest subunits, Dpb3 and Dpb4, are nonessential (23,24). Pol⑀ has been implicated in chromosomal DNA replication by both genetic and biochemical analysis (25,26). Its association and movement with replication forks has been well documented (27,28). However, its essential role remains unclear. Although Pol⑀ apparently functions in DNA replication (29), the catalytic domain is not required for DNA replication or viability, suggesting that Pol⑀ has another essential role (30 -32). A role for Pol⑀ in sister chromatid cohesion has also been suggested (33). However, it is clear that Pol⑀ is recruited to replication complexes as an early event in the initiation of DNA replication and that the intact protein is required for robust replication of chromosomes during S phase.
In this work we show that the budding yeast Pol⑀ subunit, Dpb2, is phosphorylated in a cell cycle-dependent manner. Furthermore, phosphorylation in vivo is dependent upon Cdc28, and Dpb2 can be directly phosphorylated by the Cdc28 CDK in vitro. We have identified at least one site of phosphorylation by Cdc28 as well as additional sites of in vivo phosphorylation. Inactivation of three CDK consensus sites in Dpb2, although not lethal, results in a synthetic phenotype with the pol2-11 allele of the catalytic subunit, culminating in decreased spore viability, decreased growth rate, and thermosensitivity. We suggest that phosphorylation of Dpb2 at CDK consensus sites facilitates the interaction with Pol2 or the activity of the Pol⑀ complex. Plasmids-CDK consensus phosphorylation sites in DPB2 were mutated by PCR using overlapping primers. YCplac33DPB2 (35) was used as a template. Mutation of site I (TP to AP) was marked with a PvuI restriction site, and sites II and III (SP to AP) were marked with a HaeII restriction site. Mutated PCR fragments were subcloned into a pRS416-DPB2-3xHA vector for mobility shift assays or into pRS406-DPB2 for integrating the mutated form into the chromosomal locus. Following integration, Ura ϩ transformants were grown on plates containing 5-fluoroorotic acid, and DPB2-CDK site mutants were identified from the Ura Ϫ clones by Southern blotting or PCR-based HaeII and PvuI restriction analysis. For overexpression and biochemical assays, DPB2-3xHA and corresponding CDK site mutants were cloned into a YCp-GAL1 vector. pRS306-CUP1-TAP-CDC28 was generated by replacing the degron cassette of pPW66R (36) with a TAP fragment (37). The TAP fragment for N-terminal tagging was a gift from S. Saitoh. The pKL73 plasmid for generating mcm2-td strains was provided by J. Diffley. Details of plasmids will be provided upon request.
Yeast Growth and Synchronization-Yeast cells were grown in either rich medium (YEPD; yeast extract, peptone, and dextrose) or, when selection for plasmid maintenance was needed, in synthetic medium supplemented with the appropriate carbon source. bar1⌬ cells were arrested in G 1 phase with 50 ng of ␣-factor per milliliter. The ␣-factor was removed by centrifugation, and cells were suspended in fresh medium. The BAR1 strains were synchronized with 10 g of ␣-factor per milliliter as described (39). Samples were prepared for flow cytom-etry as described (40) and analyzed by a FACScan analyzer (BD Biosciences).
In Vivo Labeling of Dpb2-HA-tagged Dpb2 and CDK site mutants were expressed from the GAL1 promoter in YEP-Gal (yeast extract, peptone, and galactose) medium. Preparation of medium and in vivo labeling of proteins with [ 32 P]phosphoric acid were as described (41). Dpb2-HA was immunoprecipitated under denaturing conditions (42), and incorporation of 32 P was detected by a PhosphorImager (Amersham Biosciences). Identical cultures were mock treated (no 32 P), and Dpb2 was analyzed by Western blotting.
In Vitro Phosphorylation of Dpb2-Cdc28 was expressed as an Nterminal fusion to the TAP tag (37) from the CUP1 promoter, bound to IgG-Sepharose, and released by tobacco etch virus (TEV) protease cleavage. Dpb2-HA substrates were purified by immunoprecipitation from overexpressing yeast strains. They were incubated at 70°C for 10 min, and kinase assays were done essentially as described (43) with or without added Cdc28.
Mass Spectrophotometric Identification of Phosphorylation Sites in Dpb2-Dpb2-HA overexpressed from the GAL1 promoter was immunoprecipitated from cells collected from two liters of culture with anti-HA antibodies coupled to Sepharose beads. The beads were washed sequentially with lysis buffer, water, and radioimmune precipitation buffer, and bound proteins were eluted with 8 M urea containing 1% SDS. Eluted proteins were precipitated with trichloroacetic acid, suspended in SDS sample buffer, and separated in SDS-polyacrylamide gel. The Dpb2 doublet was visualized by Coomassie staining, and the protein bands were cut from the gel. Each gel slice was split into thirds and subjected to digestion using three separate proteases, trypsin, elastase, and subtilisin to increase the number and variety of peptides generated. The mixture of peptides was separated by multidimensional liquid chromatography coupled with tandem mass spectrometry (MudPIT) as described (44,45). Each potential site of phosphorylation was detected with multiple overlapping peptide identifications and then manually evaluated for quality of identification.

Dpb2p Is Phosphorylated in a Cell Cycle-regulated Manner-
Analysis of Dpb2 in SDS-polyacrylamide gels reveals two forms of differing mobility. This is true for both the N-terminally and C-terminally tagged proteins as well as untagged Dpb2 ( Fig. 1A and data not shown). When Myc-tagged Dpb2p was immunoprecipitated, treated with lambda phosphatase, and analyzed by SDS-PAGE, the slower migrating form was lost (Fig. 1B). The band persisted if phosphatase inhibitors were included in the digest. Thus, the reduced mobility was due to phosphorylation (Fig. 1B).
Both the low and high mobility forms were present in an approximately equal ratio in asynchronously growing cells and cells treated with the DNA replication inhibitor hydroxyurea (HU), nocodazole, or methyl methanesulfonate (MMS) (Fig.  1A). Interestingly, cells arrested in G 1 phase with ␣-factor contained little or none of the lower mobility species. This suggested that Dpb2 phosphorylation might be regulated during the cell cycle. To evaluate that possibility, wild-type cells carrying Myc-tagged Dpb2 were arrested in G 1 with ␣-factor and then released into the cell cycle. Samples were taken at 15-min intervals and analyzed for DNA content and Dpb2 mobility shift (Fig. 1C). Consistent with cell cycle-regulated phosphorylation of Dpb2, the slower migrating form of Dpb2p appeared at 45 min, coinciding with entry into S phase. Although we did not observe the disappearance of the phosphorylated form, perhaps due to poor synchrony in the second cycle, we found that unbudded G 1 phase cells isolated by centrifugal elutriation also contained little or no phosphorylated Dpb2 (data not shown). Thus, Dpb2 appears to be phosphorylated as cells enter S phase, with disappearance of the modified form occurring sometime subsequent to mitosis or cytokinesis.
Although Dpb2 appears to be phosphorylated for a signifi-cant portion of the cell cycle, we have not observed a complete shift of Dpb2 to the slow migrating form under any conditions. This suggests that phosphorylation might be restricted to a subpopulation of Dpb2. We first considered the possibility that Dpb2 phosphorylation was associated with its capacity to interact with Pol2. To evaluate that possibility, Pol2, tagged with the HA epitope, was immunoprecipitated from cells expressing Dpb2-Myc, and the proteins in the immune complexes were evaluated by immunoblotting (Fig. 1A). Both forms of the Dpb2-Myc protein were detected in immunoprecipitates of HAtagged Pol2p, demonstrating that the modification is neither required for, nor restricts, incorporation into Pol⑀ holoenzyme complexes. Furthermore, modification does not require the catalytic activity of DNA polymerase ⑀, because both species can also be found associated with Pol2-16, which is devoid of catalytic domains (30). However, under certain conditions we did observe a modest enrichment of the lower mobility species of Dpb2 in immunoprecipitates of Pol2-HA when compared with the whole cell extract from which that immune complex was prepared (Fig. 1A), suggesting that there may be a preference for binding to the phosphorylated form of the protein.
Next, we asked whether loading of Dpb2 onto origins of replication was required for Dpb2 phosphorylation. Loading of Pol⑀ is mediated by Dpb11, whose recruitment requires intact MCM function (28). Hence, Dpb2 was tagged with a Myc epitope in a strain in which Mcm2 could be eliminated via a thermally inducible degron (36,39). The mobility of Dpb2-Myc was analyzed by in an mcm2-td strain following release from mating pheromone arrest at the restrictive temperature ( Fig.  2A). The phosphorylated low mobility species of Dpb2 accumulated following release from the G 1 arrest regardless of the functionality of Mcm2. Thus, Dpb2 phosphorylation is independent of the binding of Pol2 holoenzyme complexes to replication origins; in fact, the possibility remained that phosphorylation of Dpb2 is a prerequisite for the association of Pol⑀ with replication origins.
Dpb2 Phosphorylation Is Dependent upon Cdc28 CDK-We wished to identify the kinase responsible for Dpb2 phosphorylation. Because Dpb2 was modified upon entry into S phase, we concentrated on kinases known to function in S phase progression. The checkpoint kinases Mec1 and Rad53 were eliminated as candidates, because phosphorylation of Dpb2 was unaffected by the deletion of either gene (not shown). To evaluate the role of Cdc7, a kinase required for the initiation of DNA replication and known to phosphorylate a number of replication proteins (6), Dpb2 was tagged with a Myc epitope in a temperaturesensitive cdc7-4 mutant strain. Its modification was followed in cells released from ␣-factor block under either permissive or restrictive conditions (Fig. 2B). At a restrictive temperature, no DNA synthesis was observed by fluorescence-activated cell sorter analysis, confirming Cdc7 inactivation under these conditions. However, phosphorylated Dpb2 was detected at 30 min. The earlier appearance of this form was likely due to the enhanced rate of cell cycle progression at 37°C. Thus, Dpb2p is phosphorylated independently of Cdc7 function. Importantly, Dpb2 phosphorylation is observed in cdc7-4 and mcm2-td cells despite the failure of the cells to complete or even initiate chromosomal DNA replication (see "Discussion").
To evaluate the role of the Cdc28 CDK, we performed the same experiment in a strain carrying cdc28-13, a temperature sensitive allele of CDC28 (Fig. 2C). Unlike the cdc7 mutant, when cdc28-13 cells were released from ␣-factor block at the restrictive temperature, Dpb2p failed to be phosphorylated. In contrast, both species of Dpb2 were detected when cells were maintained at the permissive temperature upon release from the mating pheromone. Thus, phosphorylation of Dpb2 was dependent upon the Cdc28 protein kinase.
The Cdc28 CDK exists in a number of distinct complexes. These can be separated into B-type cyclin (Clb) CDK complexes and G 1 cyclin (Cln) CDK complexes. Although Cln/CDK complexes are required to progress from the mating pheromone block into a new cell cycle (46), it is the B-type cyclins Clb5 and Clb6 that associate with Cdc28 to drive cells into S phase (47,48). Activation of those kinases, but not activation of Cln/CDK, depends upon the activity of the Cdc34 ubiquitin-conjugating enzyme, which targets the Clb/CDK inhibitor, Sic1, for degradation (48). Therefore, to evaluate whether Cln/CDK or Clb/ CDK was required for Dpb2 phosphorylation, we evaluated the formation of the lower mobility species of Dpb2 in a cdc34-3 mutant strain released from mating pheromone arrest at the restrictive temperature (Fig. 2C). Cells that arrested subsequent to budding, but prior to entry into S phase, accumulated the phosphorylated form of Dpb2. These results suggest that Cln/CDK is sufficient for the phosphorylation of Dpb2p in vivo.
The results above indicate that G 1 cyclin-associated forms of the Cdc28 CDK are sufficient for formation of the lower mobility phosphorylated species of Dpb2. However, because Cln/ CDK is required for cell cycle progression, it is unclear whether the accumulation of the phosphorylated species depends directly upon Cln/CDK or rather is a consequence of cell cycle progression. In addition, we also observe phosphorylation late in the cell cycle when G 1 cyclins are not expressed. It is unclear whether that is a consequence of the stability of the phosphorylated species or whether Clb-associated CDK is also sufficient for Dpb2 phosphorylation.
To address those questions, we performed the following experiment. First, we evaluated whether phosphorylated Dpb2 persists in cells arrested in S phase using the ribonucleotide reductase inhibitor, hydroxyurea. Wild-type cells carrying Myc-tagged Dpb2p were pre-synchronized during G 1 phase with the mating pheromone and then released in the presence of hydroxyurea. After 3 h to allow synchronization in S phase, Dpb2p was found to be phosphorylated (Fig. 2D, lane A). Because G 1 cyclins are absent in hydroxyurea-arrested cells, we can conclude that Dpb2 phosphorylation persists in the absence of Cln/CDK. Next, to determine whether the phosphorylation of Dpb2, apparent in the arrested cells, is dependent upon the CDK, we synchronized temperature-sensitive cdc28-4 mutant cells in the same manner and then, in the presence of hydroxyurea, shifted them to the restrictive temperature to inactivate the cdc28-4 CDK (Fig. 2D, lane B). In that case, the lower mobility species of Dpb2 is lost, indicating that phosphorylation of Dpb2 outside of G 1 phase is dependent upon the persistence of Cdc28 activity and, therefore, upon Clb-associated CDK. Finally, the cultures were shifted to 37°C, and hydroxyurea was removed (Fig. 2D, lane C). Again, no lower mobility species of Dpb2 was observed. In each case, phosphorylation persisted in wild-type cells. Together, these results establish that Cdc28 kinase is required for Dpb2p phosphorylation and that either Cln/Cdc28 or Clb/Cdc28 kinases are sufficient.
Dpb2 Is Phosphorylated at a Site Conforming to the CDK Consensus-Analysis of the primary sequence of Dpb2p reveals two sites conforming to the full consensus for CDK phosphorylation, (S/T)PX(K/R) (I and III, Fig. 3A) and an additional site meeting the minimal criteria for a CDK site, (S/T)P (II, Fig.  3A). We considered those sites candidates for phosphorylation by the Cdc28 kinase. To evaluate whether those sites were responsible for the phosphorylation-dependent mobility shift we observed in Dpb2, the serine or threonine residues at these sites were mutated to alanine either individually or in various combinations. The mutant forms of Dpb2 tagged with the HAepitope were expressed from a plasmid in a wild-type strain and analyzed by immunoblotting (Fig. 3B). That analysis revealed that mutation of Ser-144 (site III) was sufficient to eliminate the lower mobility species of Dpb2-HA, whereas alanine substitution at Thr-272 (site I) or Ser-616 (site II) had no effect on the appearance of that species. We conclude that Dpb2 is phosphorylated in a Cdc28-dependent manner and that at least one site of phosphorylation, Ser-144, leads to an observable mobility shift. Although it was possible that the remaining sites were phosphorylated, phosphorylation at those sites did not result in a change in electrophoretic mobility under the gel conditions used.
In an attempt to refine the previous analysis, we analyzed the incorporation of isotopically labeled 32 P into Dpb2 in vivo. Cells carrying HA-tagged alleles of either wild-type Dpb2 or Dpb2 in which the putative CDK sites were mutated were grown in 32 P-orthophosphate, and the Dpb2-HA immunopre-cipitated from those cells was analyzed by gel electrophoresis and autoradiography (Fig. 3C). We observed that both the lower and higher mobility bands of Dpb2 were phosphorylated in vivo (Fig. 3C, lower panel) and that an approximately equivalent incorporation occurred in each of the species (Fig. 3D) despite more protein being represented in the higher mobility form. Thus, the lower mobility species is phosphorylated to a greater extent than the higher mobility species. The substitution of the serine of site III with alanine resulted in loss of the lower mobility species with a concomitant loss of ϳ40% of the total 32 P incorporation into Dpb2 (Fig. 3, C and D). The loss of that species appears to have been accompanied by a slight FIG. 2. Dpb2 phosphorylation is independent of DNA replication but dependent upon Cdc28 in vivo. A, Dpb2 phosphorylation is independent of initiation of DNA replication. Dpb2-Myc was expressed in cells rendered temperature-sensitive for S phase entry by the thermolabile mcm2-td allele in combination with GAL1:UBR1 (TAY74) and a control GAL1:UBR1 strain (TAY73). Dpb2-Myc was analyzed by immunoblotting in cells during and following release from arrest during G 1 phase by ␣-factor at a restrictive temperature for mcm2-td. The replication defect of mcm2-td was confirmed by flow cytometry. B, Dpb2 phosphorylation does not require the Cdc7 protein kinase. The temperature-sensitive mutant, cdc7-4, expressing Myc-tagged Dpb2 (TAY771) was arrested with ␣-factor and released at either the permissive (25°C) or restrictive temperature (37°C), and the mobility of Dpb2 was monitored by immunoblotting. Failure of the cdc7-4 mutant to enter S phase was confirmed by flow cytometry. C, Dpb2 phosphorylation is dependent upon the Cdc28 protein kinase. Temperature-sensitive cdc28-13 (TAY768) and cdc34-3 (TAY775) strains expressing Myc-tagged DPB2 were synchronized in G 1 phase with ␣-factor, cultures were split, and one half was released into cell cycle at 25°C and the other half at 37°C. Dpb2-Myc was detected by immunoblotting. D, a temperature-sensitive cdc28-4 strain (TAY774) and wild-type strain (TAY179), both expressing Myc-tagged Dpb2, were released from ␣-factor arrest at the permissive temperature (25°C) into medium containing 0.2 M HU. After 3 h, cultures were shifted to the restrictive temperature (37°C) for 30 min to inactivate cdc28-4. HU was removed from one half of each culture by centrifugation, and all cultures were incubated for another 2.5 h at 37°C. Dpb2-Myc was detected by immunoblotting at the times indicated.
FIG. 3. Analysis of Dpb2 CDK consensus phosphorylation sites. A, a schematic representation of the CDK consensus sites in Dpb2. B, the mobility shift of Dpb2 is dependent upon the phosphorylation of Ser-144. The serine or threonine residues at CDK sites were mutated to alanines individually and in combinations, and the resulting HA-tagged Dpb2 mutant proteins were expressed from a centromeric plasmid under control of the Dpb2 promoter. Dpb2-HA was analyzed by immunoblotting with anti-HA antibodies. C, Dpb2 is phosphorylated at CDK consensus sites and other sites in vivo. HA-tagged Dpb2 was immunoprecipitated using anti-HA antibodies from cells expressing wild-type (wt) Dpb2 or Dpb2 in which the CDKIII site or sites CDKI-III were mutated and grown in the presence of [ 32 P]orthophosphate. Proteins were separated by SDS-PAGE and analyzed by a PhosphorImager. D, quantification of 32 P incorporation into Dpb2 species. The incorporation of radiolabeled phosphate into Dpb2-HA in the experiment shown in panel C was quantitated by a Phospho-rImager. Gray bars represent the higher mobility form of Dpb2, and black bars represent the lower mobility species. Counts were normalized to a background band present in all samples. E, Dpb2 is phosphorylated by the Cdc28 protein kinase in vitro. HA-tagged Dpb2 from wild-type and CDK site mutants were purified by immunoprecipitation and phosphorylated using ␥-32 P-ATP in vitro in the presence or absence of TAP-purified Cdc28 protein kinase. Products were analyzed by gel electrophoresis and autoradiography. increase in the relative amount of phosphate in the higher mobility form, consistent with the presence of additional phosphorylated residues in the lower mobility species. Analysis of the CDK I-III mutant resulted in an additional loss of total phosphate, suggesting that one or both of those sites are phosphorylated along with additional, as yet unidentified, residues. Thus, the higher mobility species represents Dpb2 that is phosphorylated at CDK consensus sites as well as at sites that do not match that consensus.
Dpb2 Is Phosphorylated by Cdc28 in Vitro-To clarify the results of the in vivo labeling experiment and confirm that Cdc28 can directly phosphorylate Dpb2, we performed an in vitro kinase assay. HA-tagged wild-type Dpb2 and various CDK site mutants were purified by immunoprecipitation from yeast strains overexpressing those forms of Dpb2 and phosphorylated in vitro with soluble TAP-tagged Cdc28 CDK complexes purified from yeast (see "Experimental Procedures"). Analysis of the products of the kinase reactions by electrophoresis and autoradiography revealed that the wild-type Dpb2 and Dpb2 mutated at the CDK consensus site II were phosphorylated, whereas little or no incorporation of 32 P was observed into Dpb2 mutated at CDK consensus site III or the protein in which all three consensus sites had been mutated (Fig. 3E). Furthermore, all of the Dpb2 phosphorylated in vitro migrated with a mobility equivalent to the lower mobility form observed in vivo. Finally, phosphorylation was strictly dependent upon the addition of Cdc28. Thus, Cdc28 can directly phosphorylate Dpb2 in vitro. Although it is possible that only the lower mobility species is a substrate for Cdc28 in vitro, we favor the interpretation that phosphorylation in vitro by Cdc28 occurs only at site III (residue Ser-144), the site responsible for the shift to the lower mobility form. Indeed, based on Coomassie staining of the kinase assay products, the relative abundance of the lower mobility form of Dpb2 was higher in the samples incubated with Cdc28 than in samples without Cdc28, arguing that this shift was a consequence of the phosphorylation of site III by Cdc28.
Dpb2 Is Phosphorylated at Multiple Sites in Vivo-Phosphorylation of Dpb2 in vitro appears to occur only at Ser-144. However, in vivo labeling showed that the faster migrating form of Dpb2 is also phosphorylated. To determine what other sites in Dpb2 were phosphorylated, we mapped the in vivo phosphorylation sites of Dpb2 by mass spectrometry. Dpb2 was purified by preparative scale immunoprecipitation, and the slower and faster migrating forms were separated by SDS-PAGE and excised individually from the gel. Because the "triple digest" protocol reported in MacCoss et al. (44) was designed for use on mixtures of unseparated proteins, it was modified for use on gel bands. By splitting the sample in thirds and performing three parallel proteolytic digestions, we were able to realize many of these same benefits of increased protein coverage (79 and 87% sequence coverage for the slower and faster migrating species, respectively) and have confidence in assigning sites of phosphorylation. The previously established MudPIT separation of peptides, tandem mass-spectrometry, and data base searching proved more than adequate for analyzing this mixture of peptides. Moreover, we were able to analyze the two electrophoretically distinct forms of the protein separately and could thus directly observe sites that differed between the two forms.
The phosphorylated peptides detected in the higher and lower mobility species are presented in Table I. A representative spectrum for a phosphorylated peptide containing each phosphorylated Ser-144 and phosphorylated Ser-616 is presented in Fig. 4. Consistent with the analysis described above, Ser-144 appears to be the predominant phosphorylated residue in the lower mobility species, based upon the frequency of spectra representing Ser-144-containing phosphopeptides. In the faster migrating form, phosphopeptides representing the putative CDK consensus site II (Ser-616) were identified. The phosphorylation of these two sites is sufficient to explain the incorporation of 32 P into the higher mobility species in vivo. Although peptides were observed containing the third CDK consensus phosphorylation site, there was no evidence that the site was phosphorylated. Strikingly, there was no evidence of the phosphorylation of Ser-616 in the lower mobility species. Although it is not possible to establish the absence of phosphorylation at those sites based upon the mass spectrometric analysis, the frequency of spectra in the higher mobility form representing the Ser-616 peptide suggests that it would have been detected if present.
CDK Phosphorylation of Dpb2 Is Required for Vigor of pol2-11 Mutants-Preliminary analysis of the DPB2 CDK site mutants indicated that all were functional, based upon their capacity to complement the temperature sensitivity of the conditional dpb2-1 mutant when expressed on a low copy plasmid (data not shown). To evaluate their effect on cellular function, each of the mutant genes was integrated at single copy into the genomic DPB2 locus (see "Experimental Procedures"). Cells carrying each of the DPB2 CDK site mutations were found to be viable and showed no gross growth defects (data not shown). Similarly, alanine substitution of all three serine residues (dpb2-CDKI-III) had no apparent phenotypic consequence (Fig.  5A). For example, dpb2-CDKI-III mutants progressed through the cell cycle with wild-type kinetics following release from G 1 phase arrest by the mating pheromone (Fig. 5B). Furthermore, the mutant cells grew robustly at both 30 and 37°C, and all exhibited a wild-type level of sensitivity to hydroxyurea and MMS ( Fig. 5A and data not shown). Thus, phosphorylation of the CDK phosphorylation sites of Dpb2p is required neither for its essential function nor for the role of wild-type Pol⑀ in DNA replication and repair, at least as far as can be determined based upon these rather crude assays.
Despite the lack of a requirement for CDK phosphorylation of Dpb2, it appears that the hyperphosphorylated form associates preferentially with Pol2p based upon its enrichment in Pol2-HA immunoprecipitates (Fig. 1A). Consequently, we hypothesized that Dpb2, which cannot be phosphorylated by Cdc28, might exacerbate the phenotype of pol2 mutants, which are already partially compromised for Pol2 function. We tested this hypothesis by crossing the strains carrying the dpb2-CDKI-III mutation to those carrying the temperature sensitive pol2-11 mutation (26). Because POL2 and DPB2 are not linked, we expected approximately one segregant of four to carry both Cell Cycle-Dependent Phosphorylation of Dpb2 mutant genes. Spores with wild-type and pol2-11 genotypes gave rise to colonies at 100 and 84% of the expected frequency, respectively (Table II). dpb2-CDKI-III had lower viability (68%). pol2-11 dpb2-CDKI-III segregants were recovered at a much lower frequency, only 18% of the expected rate. Thus, the pol2 dpb2-CDKI-III mutants exhibited substantially reduced spore viability. In addition, those pol2-11 dpb2-CDKI-III clones that were viable grew very slowly (Fig. 5C). We also tested dpb2-CDKIϩIII and dpb2-CDKII for a synthetic interaction with pol2-11 and found that dpb2-CDKII caused no synthetic effect, whereas dpb2-CDKIϩIII pol2-11 had reduced spore viability but a growth rate comparable with the pol2-11 spores (not shown).
In addition to the effect on spore viability, both the dpb2-CDKIIϩIII mutant (having mutations only in the confirmed sites of phosphorylation) and the dpb2-CDKI-III mutant enhanced the temperature sensitivity of the pol2-11 mutant (Fig.  5D, glucose). In contrast, the other dpb2-CDK mutations had Gel slices corresponding to the higher and lower mobility forms of Dpb2 were excised and subjected to enzymatic digestion using a gel-based version of the protocol described by MacCoss et al. (44). Tandem mass spectrometry (MS/MS) spectra corresponding to peptides containing phosphorylated serine 114 (S144) and phosphorylated serine 616 (S616) are shown. Fragments corresponding to B ion series (red), Y ion series (blue), and losses of phosphoric acid (green) are indicated for each peptide spectrum.
little or no effect, suggesting that CDK sites II and III must be eliminated to give rise to a synthetic interaction. As predicted, that defect is suppressed by the expression of wild-type POL2 from the GAL1 promoter (Fig. 5D, galactose). A modest increase in sensitivity to HU was also observed in the dpb2-CDK mutants (data not shown). Finally, we have observed a modest decrease in the spore viability in dpb2-CDKI-III dpb11-1 dou-ble mutants (data not shown). Thus, although Dpb2 lacking CDK consensus sites appears to interfere with the outgrowth of spores following meiosis in cells compromised for Pol⑀ function, it does not appear to have an effect on vegetative growth of otherwise wild-type cells. The fact that it appears to exacerbate the phenotype of pol2-11 mutants, which are compromised for the function of the Pol2 C terminus, is consistent with the observation that Dpb2 interacts with the essential non-catalytic C-terminal domain of Pol2 and with our observation that phosphorylation appears to enhance the interaction of Pol2 and Dpb2.

DISCUSSION
This study, together with the recent proteomic screen for CDK substrates (9) provides the first evidence of phosphorylation of the DNA polymerase ⑀ holoenzyme. We demonstrate that Dpb2, the B subunit of Pol⑀, is a phosphoprotein and an in vivo substrate of the Cdc28 kinase. The Dpb2 subunit of Pol⑀, like the B subunits of Pol␣ and ␦, is essential and conserved FIG. 5. Genotypic analysis of DPB2 CDK site mutants. A, Dpb2 CDK site mutants are unaffected for growth and sensitivity to hydroxyurea and MMS. 5-fold serial dilutions of wild-type (CWY231), dpb2-CDKIII (TAY786), and dpb2-CDKI-III (TAY857) cells were spotted on YEPD plates or the same plates containing 0.2 M HU or 0.01% MMS. Plates were incubated for 3 days at the indicated temperature. B, dpb2-CDK site mutants progress through S phase with normal kinetics. Wild-type (CWY231) and dpb2-CDKI-III cells (TAY857) were arrested in G 1 phase with ␣-factor and released into the cell cycle. DNA content was measured by flow cytometry at the times indicated. C, dpb2-CDK site mutants exhibit a synthetic defect with the pol2-11 mutation. Tetrad analysis of a cross of a pol2-11 strain (TAY247) and a dpb2-CDKI-III strain (TAY857) was performed. Two representative tetrads are shown; the genotypes are as follows: 1, wild-type; 2, dpb2-CDKI-III; 3, pol2-11; and 4, dpb2-CDKI-III pol2-11. See also Table II. D, characterization of pol2-11 dpb2-CDK mutants. The growth of various pol2-11 dpb2-CDK mutants containing GAL-POL2 on a plasmid was compared by a spotting assay as described for panel A. The cells were plated at 30°either under conditions that induce GAL-POL2 (Galactose) or conditions under which it is repressed (Glucose). between species (49). These accessory subunits are apparently involved in the interaction of the catalytic subunit with other proteins. However, their precise regulatory significance remains obscure. We identified three putative CDK sites in Dpb2. However, it appears that the Cdc28 CDK does not phosphorylate all three of those sites. In vitro, Cdc28 incorporates phosphate only into the lower mobility form, suggesting that only Ser-144 (consensus site III) is phosphorylated. Although it is possible that the other putative CDK sites are phosphorylated in that species, that interpretation requires that Ser-144 be phosphorylated prior to or along with the other sites. We conclude that Ser-144 is the preferred site or, possibly, the only site of in vitro phosphorylation by Cdc28. The absence of phosphopeptides containing residues other than Ser-144 in our mass spectrometry analysis of the lower mobility species of Dpb2 lends further support for Ser-144 as the site of phosphorylation by Cdc28 in vitro. Although it is possible that phosphorylation of Dpb2 at Ser-144 in the in vitro reaction prevents phosphorylation at additional CDK consensus sites, that alone would be insufficient to explain the lack of phosphorylation of the higher mobility species.
Despite the evidence that Ser-144 is the only site phosphorylated by Cdc28 in vitro, both metabolic labeling and mass spectrometric analysis demonstrate the phosphorylation of Ser-616 (consensus site II) in the faster migrating form in vivo. This site is interesting because it is in a region highly conserved between budding yeast, fission yeast, and human homologs. A number of proteins are known to be phosphorylated by CDK kinases at SP or TP sites that are not associated with full CDK consensus sequences (9). However, our in vitro analysis suggests that Ser-616 may not be a bona fide CDK site but rather a site for another proline-directed protein kinase.
Although much of our analysis of Dpb2 phosphorylation was done using a standard complement of molecular genetic and biochemical assays coupled with in vivo and in vitro 32 P-labeling experiments, we also applied a novel mass spectrometrybased strategy to provide us with direct analysis of phosphorylated residues. This is an extension of a previously published triple digest protocol in which three separate proteolytic digestions are performed on the sample with the goal of increasing sequence coverage and also our confidence when assigning the modification to the correct residue. Although this gel-based triple digest protocol did not yield sequence coverage quite as extensive as that in other examples using protein samples in solution (44,50), 2 it proved quite useful for probing the phosphorylation of the different electrophoretic forms of Dpb2. This provides us with an important new tool for deciphering posttranslational modifications and their regulatory roles within the cell.
Like Pol⑀, other DNA polymerases are phosphorylated on accessory as well as catalytic subunits. Both the catalytic subunit and subunit B of Pol␣ are phosphorylated in vivo and are CDK substrates in vitro (9,(51)(52)(53). Phosphorylation of human Pol␣ has been proposed to regulate its in vivo activity (54). Interestingly, the B subunit of yeast Pol␣ is phosphorylated only if it associates with the catalytic subunit (55). That phosphorylation was shown to cause dissociation of the holoenzyme from chromatin (56). Both the catalytic subunit and the C subunit/p66 of human Pol␦ are also phosphorylated in a cell cycle and CDK-dependent manner (57,58). Phosphorylation of subunit C may prevent its binding to the proliferating cell nuclear antigen (PCNA). Chromatin fractionation experiments suggest that the binding of Dpb2 to chromatin is not regulated by phosphorylation (data not shown). Although phosphorylation at Ser-144 appears not to be required for that association (nor does the association promote phosphorylation at Ser-144), we have noticed a modest bias for association of the phosphorylated form of Dpb2 with Pol2.
Phosphorylation of Dpb2 Ser-144 coincides with entry of cells into S phase, consistent with the essential role for Dpb2 in DNA replication. The timing of phosphorylation suggests that it might play a role in the activation of Pol2 or its recruitment to replication origins. Although the phenotype of the phosphorylation site mutants suggests a role in facilitating Pol⑀ activity when Pol2 is compromised, it does not support a central role for phosphorylation in regulating the timing of replication. In fact, whereas a number of replication proteins are phosphorylated with similar kinetics, in most cases those modifications appear to have little impact on DNA replication. Sld2/Drc1 is the only replication protein known that must be phosphorylated for replication to occur (2). This is surprising, because both the Cdc28 and the Cdc7 protein kinases are essential for initiation of DNA replication. The critical steps in replication requiring CDK and Cdc7 action have been identified along with candidates for the relevant substrates, including the MCM proteins and Cdc45 (3)(4)(5).
In addition to its role in promoting initiation of S phase, phosphorylation of replication proteins by Cdc28 plays a role in limiting replication to once per cell cycle (13). This raises the possibility that phosphorylation of Dpb2 is associated with the block to re-replication. However, the timing of phosphorylation and the involvement of the G 1 cyclin-associated form of Cdc28 suggest that is not the case. Furthermore, the enrichment of the phosphorylated form of Dpb2 in Pol2 complexes and the synthetic phenotype observed when combining the dpb2-CDKI-III mutation with the pol2-11 mutation are also consistent with a positive role for Dpb2 phosphorylation.
In addition to the two CDK phosphorylation sites identified in our mass spectrometric analysis of Dpb2, we obtained phosphopeptides representing sites not matching the CDK consensus (data not shown). The most prominent of those was Ser-125, residing in a sequence matching the casein kinase II consensus (SXXE) (59). That analysis suggests that Ser-125 and Ser-144 are the main phosphoacceptor sites in Dpb2. Ser-125 appears to be present predominantly in the higher mobility species and, in contrast to Ser-144, to be phosphorylated in G 1 -arrested cells. The Ser-125 phosphorylation site has been conserved between budding and fission yeast Dpb2. Although there is no homologous site in human Dpb2, there are a number of other predicted CKII sites. Similarly, whereas the Ser-144 CDK consensus site is not conserved among Dpb2 homologs, those proteins contain other CDK consensus sites. Whether Dpb2 from those organisms is phosphorylated remains to be investigated. However, there is precedence for phosphorylation of other replication proteins by both CKII and CDK. Interestingly, human DNA ligase I is a substrate for both casein kinase II and Cdk2, which may regulate its association with proliferating cell nuclear antigen and replication foci (60 -62). To date, the significance of this type of regulation with regard to genomic DNA replication and chromosomal stability is not known.
CDK-dependent phosphorylation plays a central role in regulating critical cell cycle events and in the implementation of cell cycle checkpoints. It is clear from the analysis of DNA replication proteins that many are targets of the Cdc28 CDK. The phosphorylation of Dpb2 and other elements of the DNA replication machinery is likely to play an important but, as yet, poorly defined role in their regulation. Definition of those phos-phorylation events and their impact on protein function will lead to a full appreciation of the regulation of DNA replication and its integration with the larger system of cell cycle control.