Autoinhibition and Autoactivation of the DNA Replication Checkpoint Kinase Cds1

Abstract

Cds1 is the ortholog of Chk2 and the major effector of the DNA replication checkpoint in Schizosaccharomyces pombe. Previous studies have shown that Cds1 is activated by a two-stage mechanism. In the priming stage, the sensor kinase Rad3 and the mediator Mrc1 function to phosphorylate a threonine residue, Thr¹¹, in the SQ/TQ domain of Cds1. In the autoactivation stage, primed Cds1 molecules dimerize via intermolecular interactions between the phosphorylated Thr¹¹ in one Cds1 and the forkhead-associated domain of the other. Dimerization activates Cds1, probably by promoting autophosphorylation. To define the mechanisms for the autoactivation of primed Cds1 and the regulation of this process, we carried out genetic and biochemical studies to identify phosphorylatable residues required for checkpoint activation. Our data indicate that dimerization of Cds1 promotes trans-autophosphorylation of a number of residues in the catalytic domain, but phosphorylation of a highly conserved threonine residue (Thr³²⁸) in the activation loop is the only covalent modification required for kinase activation in vitro and in vivo. Autophosphorylation of Thr³²⁸ and kinase activation in unprimed, monomeric Cds1 are strongly inhibited by the C-terminal 27-amino acid tail of the enzyme. This autoinhibitory effect may play an important role in preventing spontaneous activation of the replication checkpoint during normal cell cycles. The two-stage activation pathway and the autoinhibition mechanism, which are probably shared by other members of the Chk2 family, provide sensitivity, specificity, and noise immunity, properties required for the replication checkpoint.

DNA replication forks can be arrested or stalled by damage to DNA templates, depletion of deoxyribonucleotides, or inhibition of replisome enzymes (1). If undetected, arrested or stalled replication forks may undergo collapse, resulting in loss of genetic information, mutagenesis, or even cell death. To maintain the integrity of the genome, eukaryotes have evolved a surveillance mechanism called the “replication checkpoint” that can detect perturbations of DNA replication and elicit a number of cellular responses that serve to mitigate the effects of such perturbations. These cellular responses may include stabilization of replication forks, suppression of initiation of DNA replication, increased DNA repair activity, augmented production of deoxyribonucleotide precursors, and delay of mitosis. The replication checkpoint pathway is essential for cell survival under a variety of stressful conditions and has been conserved from yeast to humans (for reviews, see Refs. 1–3). Mutations in the pathway are also linked to cancer (4–6).

The replication checkpoint is a complex signal transduction pathway that can be separated conceptually into three functional components. Sensors detect the perturbed DNA replication forks; mediators transduce the checkpoint signal, whereas effectors regulate the cell cycle and promote cell survival. Genetic studies, especially those in the yeasts, have identified most, if not all, of the essential factors of the pathway. In the fission yeast Schizosaccharomyces pombe, six Rad proteins mediate the sensor function (for reviews, see Refs. 7 and 8). The protein kinase Rad3 (ATR in human cells) binds an essential co-factor Rad26 (ATRIP in human cells), and the complex associates with stalled replication forks. Rad9, Hus1, and Rad1 form the “9-1-1” ring structure similar to that of the replication processivity factor proliferating cell nuclear antigen. Rad17, in association with Rfc2-5, loads the 9-1-1 complex onto DNA at stalled forks. After detection of stalled forks by the sensor complexes, the mediator protein Mrc1 protein (Claspin in human cells) functions to facilitate the Rad3-dependent phosphorylation and activation of the effector protein kinase Cds1 (Chk2 in human cells) (9–11). Studies in Saccharomyces cerevisiae suggest that Mrc1 may be a component of the replisome (12, 13). A second mediator, Crb2 (BRCA1 in human cells) (14, 15), functions in response to DNA damage either within or outside of S phase. Crb2 facilitates the activation of a second effector kinase, Chk1.

We have previously reported that in S. pombe, the effector kinase of the replication checkpoint pathway, Cds1, is activated by a two-stage mechanism (11). In the first or priming stage, the sensor kinase Rad3 phosphorylates two functionally redundant Cds1-docking repeats in the middle of the mediator Mrc1. The phosphorylated docking repeats on Mrc1 recruit Cds1 to the stalled replication fork by a phospho-dependent interaction with the forkhead-associated (FHA)³ domain of Cds1. Once recruited to the proximity of the assembled sensor complex, Cds1 is phosphorylated by Rad3 at Thr¹¹. In the second or autoactivation stage, primed Cds1 molecules dimerize by two identical intermolecular interactions between phosphorylated Thr¹¹ and the FHA domain. Dimerization promotes autophosphorylation and activation of Cds1. This two-stage activation mechanism is supported by genetic studies (9, 16–18) and is probably similar to the activation pathway for mammalian Chk2 (19–23). Although many steps in the pathway are now understood, the precise biochemical mechanism of autoactivation of primed Cds1 has not been well defined.

Protein kinases can be activated by a variety of mechanisms. Although phosphorylation of the activation loop, usually by an upstream kinase of a signal transduction pathway, is the most common mechanism for kinase activation, some protein kinases can be activated by phosphorylation of residues outside the activation loop (for reviews, see Refs. 24 and 25). Other protein kinases can be activated without phosphorylation (e.g. by intermolecular interactions following dimerization) (26), by removal of an inhibitory element (27), or by binding to an activator (27, 28). Since the autoactivation of primed Cds1 requires dimerization, three possible activation mechanisms can be proposed. First, like many other protein kinases, Cds1 may be activated by phosphorylation of the activation loop (24). There are several known examples of kinase activation via trans-autophosphorylation of the activation loop. In these cases, the activation loop usually contains a consensus phosphorylation site of the kinase itself. This is not the case for Cds1 family kinases. A second possibility is that dimerization of Cds1 may allow intermolecular interactions that promote activation, as has been suggested for the epidermal growth factor receptor (26). Finally, activation of Cds1 may be a consequence of phosphorylation of residue(s) outside the activation loop. In the second and the third models, phosphorylation of the two essential threonine residues in the activation segment observed previously in mammalian Chk2 (22) and in the S. cerevisiae homologue Rad53 (29) would be a by-product, not a cause, of kinase activation.

Several previous observations have provided evidence in support of the possibility that activation of Cds1 requires autophosphorylation. First, Cds1 is a phosphoprotein, and hydroxyurea (HU) treatment of cells induces further phosphorylation that is partially dependent on the kinase activity of Cds1 itself.⁴ In the case of mammalian Chk2, the ortholog of Cds1, sites of phosphorylation have been mapped to the activation segment residues, Thr³⁸³ and Thr³⁸⁷ (22, 30), as well as to residues Ser³⁷⁹ (31), Ser⁵¹⁶ (30, 32), and Ser⁴⁵⁶ (33), which lie outside of the activation segment. Phosphorylation has also been mapped by mass spectrometry to sites within and outside of the activation segment of Rad53 (29), the S. cerevisiae homologue of Cds1. Second, genetic studies have shown that residues Thr³²⁸ and Thr³³² in the activation segment of Cds1 (corresponding to Thr³⁸³ and Thr³⁸⁷ of Chk2 and Thr³⁵⁴ and Thr³⁵⁸ of Rad53) are essential for kinase activity (11, 34). Third, phosphatase treatment of “activated Cds1” purified from HU-treated cells abolishes kinase activity (11). Finally, activation of induced Cds1 dimers in vitro is dependent upon ATP (11).

In this report, we describe experiments aimed at distinguishing among the three potential mechanisms for Cds1 activation. We show that there are only three phosphorylatable residues in the Cds1 kinase domain (Thr³²⁸, Thr³³², and Tyr³⁵²) that are essential for activation of the replication checkpoint in vivo and for enzyme activity in vitro. Of these three residues, Thr³²⁸ in the activation loop is a target of autophosphorylation, and its phosphorylation is the only covalent modification required for Cds1 activation. Autophosphorylation of Thr³²⁸ occurs in trans and only proceeds at an appreciable rate when the enzyme is at high local concentration. Presumably, one molecule in a Cds1 dimer transiently assumes an active conformation and phosphorylates the Thr³²⁸ in the activation loop of the other molecule. The activated molecule can then rapidly phosphorylate its dimeric partner. The second essential residue, Thr³³², which is also in the activation loop, is not phosphorylated and is likely required, directly or indirectly, for catalysis. The third essential residue Tyr³⁵² can be autophosphorylated in vitro with the Cds1 purified from S. pombe, and its phosphorylation is strongly stimulated by dimerization. However, Tyr³⁵² phosphorylation is not readily observed in vivo and is not required for Cds1 activation. Our data rule out the other two possible mechanisms for Cds1 activation: phosphorylation of sites outside of the activation segment and phosphorylation-independent conformational changes induced by dimerization. We also report that the C terminus of Cds1 is a cis-regulatory element that can dramatically suppress Cds1 autoactivation in vitro and in vivo. Taken together, our data explain how the replication checkpoint can be sensitive and specific and also possess a high threshold for spontaneous activation.

EXPERIMENTAL PROCEDURES

Yeast Strains, Point Mutation, and Drug Sensitivity

S. pombe strains and growth media were prepared by standard methods (35). Two S. pombe strains were specifically made for this study: YJ373 h⁺ cds1(D312E)-6his2HA(int) and YJ374 h⁺cds1–6his2HA(int). Mutations of the potential phosphorylation sites in Cds1 were made by QuikChange mutagenesis PCR using the high fidelity DNA polymerase pfu on plasmid pYJ294 (prom-cds1–6his1HA-LEU2). All sequences and mutations were confirmed by DNA sequencing. Vectors were introduced into the S. pombe strain GBY191: h⁺ Δcds1::ura4⁺ by electroporation (Gene Pulser II; Bio-Rad). All strains used in this study contain the auxotrophic markers leu1-32, ura4-D18, ade6-M210, or ade6-M216. To test the drug sensitivity, 2 × 10⁷ cells/ml of logarithmically growing S. pombe were diluted 5-fold and spotted onto YE6S plates containing HU or methyl methanesulfonate (MMS) at the indicated concentrations. After incubation at 30 °C for 3 days, plates were photographed.

Purification of Cds1 from S. pombe or Escherichia coli

Inactive wild-type Cds1–6his1HA and its mutant proteins were expressed on a plasmid under the cds1⁺ promoter in Δcds1::ura4⁺ or Δrad3::ura4⁺Δcds1::ura4⁺ cells. Cells from a 2-liter culture were harvested, washed once with ice-cold distilled H₂O, and resuspended in an equal volume (ml/g) of 2× lysis buffer (100 mm Tris/HCl, pH 7.5, 500 mm NaCl, 4 mm EDTA, 2 mm sodium pyrophosphate, 100 mm NaF, 120 mm β-glycerophosphate, 0.2% Nonidet P-40) containing protease inhibitors. Cell lysate was made by the coffee mill method and clarified by repeated centrifugations at 4 °C, 43,000 × g, for 20 min. Extracts were mixed with 1 ml of prewashed Talon resin (Clontech), rotated at 4 °C for 1 h, washed five times with the washing buffer (20 mm Tris/HCl, pH 7.5, 250 mm KCl, 0.05% Tween 20, 0.5 mm dithiothreitol, 10% glycerol, and 5 mm imidazole), and eluted in elution buffer (20 mm Tris/HCl, pH 7.5, 150 mm KCl, 0.5 mm dithiothreitol, and 10% glycerol) containing 150 mm imidazole. Eluted Cds1 was concentrated in a Centricon 10 concentrator (Millipore). Protein concentration was determined by silver staining. Occasionally, the inactive Cds1 was further purified by HA antibody beads, as described for purification of activated Cds1. To purify activated Cds1–6hisHA, cells were treated with 25 mm HU in EMM6S[leu−] medium at 30 °C for 4 h. Cell extract was similarly prepared and added to 500 μl of prewashed anti-HA antibody beads in a 10-ml Econo-Column (Bio-Rad). The column was rotated at 4 °C for 2 h, washed five times, and eluted in the elution buffer containing 1 mg/ml HA peptide.

To purify Cds1 and its mutant proteins from E. coli, Cds1 containing a cleavable His₆ tag at the N terminus was co-expressed with untagged λ-phosphatase on the same pET-28 vector. The proteins were expressed in BL21(DE3) cells at 25 °C for 4 h after adding isopropyl 1-thio-β-d-galactopyranoside to 0.4 mm. Cells were lysed by a cell disruptor (Avestin, Inc.) in 20 mm phosphate buffer containing 500 mm NaCl and 20 mm imidazole (pH 7.6). Cell lysate was clarified by centrifugation at 4 °C, 43,000 × g, for 30 min. The supernatant was loaded onto a HisTrap column (GE Healthcare). After extensive wash in the lysis buffer containing 20 mm imidazole, the protein was eluted by an imidazole gradient. The His₆ tag was cleaved off by PreScission protease (GE Healthcare) and removed by running through the HisTrap column one more time. Cds1 was then further purified by MonoQ column (GE Healthcare) and gel filtration. Protein concentrations were determined by a Bradford assay.

Cds1 Kinase Assay and Western Blotting

Two established assays were used to measure Cds1 activity. The first assay uses GST-Wee1 (amino acids 11–152) as the substrate essentially as described previously (36). Briefly, cell lysate was made from five OD cells by a mini-bead beater in the lysis buffer (50 mm Tris/HCl, 150 mm NaCl, 2 mm EDTA, 1 mm Na₃VO₄, 10 mm pyrophosphate, 50 mm NaF, 60 mm β-glycerophosphate, 0.1% Nonidet P-40, pH 7.6, and protease inhibitors). After centrifugation at 4 °C, 14,000 rpm, for 5 min, the supernatant was added to glutathione beads with bound GST-Wee1, rotated at 4 °C for 1 h. The beads were washed twice with ice-cold PBS-T buffer, mixed with 25 μl of kinase reaction buffer (20 mm Tris/HCl, pH 7.5, 5 mm MgCl₂, 1 mm dithiothreitol, 75 mm KCl, 50 μm [γ-³²P]ATP), and incubated at 30 °C for 20 min. After separation by SDS-PAGE, the gel was stained with Coomassie Blue to visualize GST-Wee1, and the kinase activity was measured by PhosphoImaging.

The second assay uses myelin basic protein (MyBP) as the substrate as described (37). Cds1–6hisHA in cell extract was immunoprecipitated with anti-HA antibody beads. After washes with ice-cold PBS-T, 20 μl kinase reaction buffer containing 5 μg MyBP was added to the beads, incubated at 30 °C for 20 min, and analyzed by SDS-PAGE. Cds1 in top half of the gel was revealed by silver staining or Western blotting using anti-HA antibody 12CA5 (Roche Applied Science). MyBP in the bottom half of the gel was visualized by Coomassie staining. ³²P incorporation was measured by phosphorimaging.

Phospho-specific Antibodies

The affinity-purified phospho-specific rabbit antibodies T328-P, T332-P, and Y352-P were prepared by Bethyl Laboratories, Inc. The three phosphopeptides T328-P (CTGTFLEpTFCGTM), T332-P (CLETFCGpTMGYLA), and Y352-P (CVNLDGGpYDDKVD) were conjugated to KLH and used as the immunogens. The reactivity ratios of phosphorylated titer versus the nonphosphorylated titer for all antibodies are ≥99:1 and were confirmed by Western blotting after the peptides and phosphorylated peptides were spotted on a nitrocellulose membrane (Fig. S3). To suppress the cross-reactivity between T328-P and T332-P antibodies, anti-T328-P antibody was incubated with the membrane in the presence of 1 μm T332-P peptide, whereas T332-P antibody was incubated with 1 μm T328-P peptide. After several washes in TBS-T, the bound primary antibodies were detected with horseradish peroxidase-conjugated goat against rabbit antibody followed by enhanced chemiluminescence (Pierce).

RESULTS

Three Potential Phosphorylation Sites in the Kinase Domain of Cds1 Are Essential for Activation of the Replication Checkpoint

We have previously shown that artificial dimerization of the C-terminal region of Cds1 (amino acids 152–460), containing the kinase domain, can promote autophosphorylation and kinase activation both in vitro and in vivo (11). As a first step toward defining the autoactivation mechanism, we carried out a large scale mutational analysis of the kinase domain to determine the potential autophosphorylation sites that are required for checkpoint function. As shown in the diagram of Fig. 1, there are 52 phosphorylatable residues (28 serines, 16 threonines, and 8 tyrosines) within this region of Cds1. Each serine or threonine was mutated to alanine, either individually or as part of a small cluster. Similarly, each tyrosine was individually mutated to phenylalanine. The resulting mutant proteins were each expressed in Δcds1 cells under control of the cds1⁺ promoter and tested for sensitivity to either HU or the DNA-damaging agent MMS.

FIGURE 1.

Three potential phosphorylation sites in the kinase domain of Cds1 are essential for activation of the replication checkpoint. The SQ/TQ, FHA, and kinase domains of Cds1 are shown in the diagram on the right. The 52 potential phosphorylation sites (28 serines, 16 threonines, and 8 tyrosines) and their relative locations in the C-terminal half of Cds1 (amino acids 152–460) are marked by green, red, and black ovals, respectively. The activation segment in the kinase domain is highlighted in yellow. The C-terminal tail of Cds1 (amino acids 434–460) is enlarged to reveal the two triple serine motifs (black bars) and the five other phosphorylatable sites. Wild-type Cds1 and various mutants were expressed from a plasmid under the control of the cds1⁺ promoter in Δcds1 cells. 5-fold dilutions of cells were spotted on YE6S plates (Control) or YE6S plates containing 2.5 mm HU or 0.012% MMS for 3 days at 30 °C. The two deletion mutants of the C-terminal tail are shown separately at the bottom.

Surprisingly, only three of the 52 phosphorylatable sites in the C-terminal region of Cds1, Thr³²⁸, Thr³³², and Tyr³⁵², proved to be essential for Cds1 function following treatment with HU or MMS (Fig. 1, left). The T328A and T332A mutants were as sensitive to HU as Cds1 null cells, whereas the Y352F mutant was slightly less sensitive, indicating that it retained a low level of checkpoint function (Fig. S1). The responses of the remaining mutants were identical to those of wild-type cells, indicating that the corresponding phosphorylatable residues are not essential for Cds1 activity or are functionally redundant. The similar pattern of sensitivity to HU and MMS suggests that Cds1 is activated by a similar mechanism in response to depletion of deoxyribonucleotides or DNA damage. Thr³²⁸ and Thr³³² reside in the activation segment of Cds1 along with three other nonessential phosphorylatable residues Thr³²², Thr³²⁴, and Tyr³³⁵ (see Fig. 3B and Ref. 24). Thr³²⁸ is in the activation loop, and Thr³³² is in the P+1 loop. Tyr³⁵² lies in the αEF/αF loop, which interacts with and probably stabilizes the activation loop (24, 38). Among the nine tyrosines in the full-length Cds1 molecule, Tyr³⁵² is the only one that is required for full checkpoint activity (data not shown).

FIGURE 3.

Thr³²⁸ is the only essential phosphorylatable residue in the Cds1 kinase domain purified from E. coli that undergoes autophosphorylation. A, Cds1(ΔC) was co-expressed with λ-phosphatase in E. coli and purified to apparent homogeneity. The purified protein (1 μm) was incubated with 20 μm ATP in standard kinase buffer at 30 °C. At each indicated time point, an aliquot was removed from the reaction and analyzed by SDS-PAGE (top) and Western blotting with phospho-specific antibodies against the three essential phosphorylation sites (Thr³²⁸, Thr³³², and Tyr³⁵²) (middle panels). The relative loading of Cds1(ΔC) in each lane was revealed by Ponceau S staining (panels below Western blots). The kinase activity of Cds1 preincubated with ATP was assayed by incubation of 50 ng of enzyme with MyBP as substrate in the presence of radiolabeled ATP for 20 min at 30 °C (bottom three panels; see “Experimental Procedures”). The relative enzyme activities were quantified in a PhosphorImager. Activity increased more than 1000-fold between 0 and 35 min. B, the sequences of the activation segments of protein kinases in Chk2 family. Three highly conserved phosphorylatable residues are marked by solid squares, whereas the less conserved residue Thr³²⁴ is marked by a dashed square. Among the four sites, Thr³²⁸ and Thr³³² are essential for Cds1 activation, whereas Thr³²⁴ and Tyr³³⁵ are not. The primary phosphorylation site Thr³²⁸ in Cds1 is shown in red. C, MS/MS spectrometry of the phosphopeptide 323–333 in the Cds1 activation loop. Cds1-(25–454) purified from E. coli was autophosphorylated as above, recovered by SDS-PAGE, and digested with subtilisin or trypsin. The resulting peptides were analyzed by MS/MS spectroscopy. Numbers in the table are the predicted ions from peptide 323–333, with the red numbers indicating those that have been detected and verified. The arrows in the spectrum mark the most characteristic ions that are sufficient to prove the structure of the phosphopeptide.

The catalytic domain of Cds1 is followed by a C-terminal “tail” of 27 residues that is rich in hydrophilic and phosphorylatable residues (7 serines and 4 threonines; see enlarged portion of Cds1 in the lower part of Fig. 1). Most of the serines are clustered in two triplets (marked by black bars). None of the mutations in this region, including a deletion of all 27 residues (Δ434–460), increased sensitivity to HU, indicating that the C-terminal tail is not essential for Cds1 activation (bottom of Fig. 1 and Fig. S6). However, as described in more detail below, we later found that the C-terminal tail is autoinhibitory for Cds1 activation (see Fig. 6). A deletion mutation (Δ443–460) that removes the two serine triplets in this domain can dramatically facilitate autoactivation of Cds1 in vitro and in vivo (see Fig. 6). For simplicity, we refer this mutant as Cds1(ΔC) in the following studies.

FIGURE 6.

The C-terminal tail suppresses Cds1 autoactivation in vitro and in vivo. A, unactivated full-length Cds1 (lanes 1–4) and the C-terminal deletion (Δ443–460) mutant Cds1(ΔC) (lanes 5–8) were purified from S. pombe and incubated with (+) or without (−) 50 μm ATP at 90 nm, 30 °C or on ice for 30 min. Activated Cds1 (lanes 9–12) purified from HU-treated S. pombe was included as a positive control. After incubation with ATP, Cds1 kinase activity was measured using MyBP (top three panels) or GST-Wee1 (middle two panels) as substrates. Aliquots of each reaction were analyzed by Western blotting using HA antibody to detect Cds1 or phospho-specific antibody to detect phosphorylated Thr³²⁸ (bottom two panels). B, wild-type Cds1, the kinase mutant Cds1(D312E), and Cds1(ΔC) were expressed from a pREP vector under the control of a weak or a medium strength nmt1 promoter (marked by a thin or a heavy line at the top, respectively) in Δcds1 cells. The cells were cultured in the absence of thiamine for 17 h at 30 °C to induce the nmt1 promoter and photographed.

Thr³²⁸ of Cds1 Is a Target of Autophosphorylation

To map the autophosphorylation site(s), wild-type Cds1 and Cds1(ΔC), which lacks the C-terminal autoinhibitory domain, were expressed from plasmids under the control of the cds1⁺ promoter in Δrad3Δcds1 cells and purified by affinity chromatography. For comparison, we also purified “activated” wild-type Cds1 from HU-treated cells. The purified proteins were incubated with [γ-³²P]ATP to label the autophosphorylation sites and then digested with trypsin. The resulting phosphopeptides were separated by thin layer electrophoresis and chromatography in two dimensions (39). As shown in Fig. 2A, the patterns of phosphopeptides derived from the three proteins were quite similar. Two phosphopeptides in the digest of wild-type Cds1, located between spots d and e, were absent from the digest of Cds1(ΔC), suggesting that they were derived from the C-terminal tail. The remaining phosphopeptides (spots a–g) were found in all three digests and presumably included the phosphorylation site(s) responsible for Cds1 activation, since incubation with ATP can activate the kinase activity of all three proteins (e.g. see Fig. 6A). Of these, phosphopeptide b was autophosphorylated to a greater extent in Cds1(ΔC) than in wild-type or “activated” Cds1. Since Cds1(ΔC) is more readily autoactivated in vitro than wild-type Cds1 (Fig. 6A), we suspected that phosphopeptide b contained the autophosphorylation site(s) necessary for activation of Cds1. This possibility was consistent with the observation that phosphopeptide b in the “activated” kinase was only weakly labeled, possibly because it had already been phosphorylated in vivo.

FIGURE 2.

Identification of Thr³²⁸ as an autophosphorylation site by two-dimensional phosphopeptide mapping. A, unactivated full-length Cds1, Cds1(ΔC), or activated Cds1 protein (marked in the upper right corner of each panel) purified from S. pombe was incubated in the presence of [γ-³²P]ATP as in Fig. 6A and digested with trypsin. The tryptic peptides were loaded on TLC plates and separated by electrophoresis at pH 1.9, followed by chromatography (39). The two right panels show mixtures of Cds1 with Cds1(ΔC) (fourth panel) and Cds1(ΔC) with activated Cds1 (fifth panel). The spots labeled a–g are phosphopeptides found in all three samples. B, phosphopeptide b in A was recovered from the TLC plate, hydrolyzed in acid, and loaded on a new TLC plate with a mixture of phosphoserine, phosphothreonine, and phosphotyrosine. The phosphoamino acids were separated by two-dimensional electrophoreses by standard methods (39). The plate was stained with nihydrin to locate the three phosphoamino acids (dashed circles) and then autoradiographed. C, the recovered phosphopeptide b was coupled to a Sequenlon membrane and loaded onto an ABI Procise 494 sequencer for 25 cycles of Edman degradation. The radioactivity recovered following each cycle was determined by scintillation counting and plotted versus the cycle number. The sequence of the peptide 318–342 is shown at the top.

To locate the phosphorylation site(s) in phosphopeptide b, we recovered the peptide from the TLC plate and subjected it to further analysis. Acid hydrolysis followed by standard two-dimensional electrophoreses (39) revealed that phosphopeptide b contained mainly phosphothreonine with a trace amount of phosphotyrosine (Fig. 2B). The phosphotyrosine probably represented contamination by the nearby phosphopeptide a, which was the only phosphopeptide containing phosphotyrosine (Fig. S2B; data not shown). The locus of the phosphorylation site(s) within phosphopeptide b was determined by automated Edman degradation (40). The results of this analysis showed that phosphopeptide b was phosphorylated at the 7th position from the N terminus and at a second site at the 10th or 11th position. Only four Cds1 tryptic peptides contain a phosphorylatable residue at position 7, and of these, the only possible candidate was the peptide containing residues 318–342, which has phosphorylatable residues at positions 7 (Thr³²⁴) and 11 (Thr³²⁸). The relatively high amount of radioactivity that eluted at cycle 10 was probably a result of contamination by phosphopeptide a (Fig. S2). The theoretical charge and R_F value of peptide 318–342 with two phosphates are 0 and 0.47, respectively, consistent with the electrophoretic and chromatographic properties of phosphopeptide b shown in Fig. 2A. Secondary digestion by chymotrypsin also indicated that phosphopeptide b contained residues 318–342 (Fig. S2A). Thus, our data indicate that Thr³²⁴ and Thr³²⁸, both of which reside in the activation loop of Cds1, are autophosphorylated in vitro. Autophosphorylation of Thr³²⁸ could be important for autoactivation of Cds1, since it is one of the three essential residues identified by the genetic analysis of Fig. 1. This possibility was directly confirmed in studies described below. Autophosphorylation of Thr³²⁴ is not required for activation of Cds1, since the T324A mutation has no effect on checkpoint function in vivo (Fig. 1). Unlike previous studies of the Cds1 homologue, Chk2 (22, 30), we found no evidence for phosphorylation of Thr³³², a residue that is essential for checkpoint activation (Fig. 1). We suggest that Thr³³² is required for substrate binding or catalysis, a hypothesis consistent with the location of the residue in the P+1 loop of the activation segment.

Analysis of phosphopeptide a by phosphoamino acid analysis, secondary digestion, and Edman degradation strongly suggested that it contained residues 343–382, with a phosphotyrosine residue at position 352 (Fig. S2C). Since the kinase-dead mutant, Cds1(D312E), prepared under identical conditions, did not undergo autophosphorylation, it is unlikely that Tyr³⁵² was phosphorylated by a contaminating tyrosine kinase from S. pombe, although this possibility cannot be completely ruled out. The Y352F mutant is defective in checkpoint function (Fig. 1), so it is possible that phosphorylation of Tyr³⁵² plays a role in activation of Cds1. Alternatively, Tyr³⁵² may be important for catalytic activity, and its phosphorylation under these conditions may simply be a byproduct of Cds1 activation. Subsequent experiments described below strongly supported the latter possibility (see Fig. 3A and “Discussion”).

Autophosphorylation of Thr³²⁸ Directly Activates Cds1 Purified from E. coli

To confirm the relationship between autophosphorylation and Cds1 activation, we analyzed the autoactivation of Cds1 purified from E. coli. When expressed in E. coli, Cds1 undergoes partial autoactivation, and the resulting kinase activity is extremely toxic to the organism (11). In order to purify large quantities of unactivated Cds1 (or the Cds1 kinase domain), we co-expressed Cds1 with λ-phosphatase (see “Experimental Procedures” for details). Most, if not all, of the Cds1 purified by this approach was inactive, largely unphosphorylated, and monomeric (Fig. 4A; data not shown). As shown below, the purified Cds1 (or the kinase domain) undergoes efficient autoactivation when incubated with ATP in vitro, suggesting that it is properly folded.

FIGURE 4.

trans-Autophosphorylation of Thr³²⁸ is the only covalent modification required for Cds1 activation. A, purified Cds1 kinase domain is largely monomeric. His₆-Cds1-(154–442) was affinity-purified on a nickel column and, after digestion with PreScission protease to remove the affinity tag, further purified by ion exchange chromatography on a MonoQ column. The resulting protein of 33 kDa was subjected to gel filtration chromatography in 25 mm HEPES, pH 7.5, 150 mm NaCl, and 1 mm dithiothreitol. Standard proteins (thyroglobulin, bovine γ-globulin, chicken ovalbumin, equine myoglobin, and vitamin B₁₂) were chromatographed under the same conditions (arrows). B, wild-type Cds1-(154–442) and the mutant proteins, Cds1-(154–442 T324V) and Cds1-(154–442 T324V S378A-S379A) were incubated with 5 mm ATP at 30 °C in standard kinase buffer. Aliquots removed at various times were analyzed by electrophoresis in a 12% native polyacrylamide gel followed by Coomassie Blue staining (upper panel of the top section) and autoradiography (lower panel of the top section). The apparent number of phosphates in each species is indicated at the right. A sample of each aliquot was also subjected to Western blotting with Thr³²⁸ phospho-specific antibody (two middle panels). The relative loading of the Cds1 in each lane was assessed by Ponceau S staining (panel above the Western blot). The protein kinase activity of 50 ng of protein in each aliquot was measured using MyBP as substrate (bottom two panels). C, Thr³²⁸ is autophosphorylated in trans by the unactivated and activated Cds1 kinase domain. The Cds1 kinase domain (154–442 T324V/S378A/S379A) was purified in its unactivated form from E. coli (lane 1). A sample of the protein was preincubated with ATP to activate the kinase by autophosphorylation of Thr³²⁸. The resulting activated Cds1-(154–442 T324V/S378A/S379A)-PO₄ was further purified. A His₆-tagged Cds1 kinase domain with an inactivating mutation (kinase-dead Cds1-(154–442 T324V/S378A/S379A D312E)) (lane 2) was used as the substrate. The unactivated (Cds1) or activated (Cds1-PO₄) kinase domains were incubated with substrate in the presence of ATP for 2 h at 30 °C. The enzyme and substrate proteins were separated by SDS-PAGE and transferred to a nitrocellulose membrane. Proteins were detected by Ponceau S staining (top). Phosphorylation of Thr³²⁸ was detected by Western blotting using the Thr³²⁸ phospho-specific antibody (bottom). D, activated full-length Cds1 can efficiently autophosphorylate Thr³²⁸ in trans. Wild-type Cds1 or the kinase-dead control Cds1(D312E), each tagged with an HA epitope, were purified from HU-treated S. pombe and incubated with untagged kinase-dead Cds1(D312E) purified from E. coli as substrate in the presence of 200 μm [γ-³²P]ATP. The ratio of enzyme to substrate in the reactions was 75 nm/925 nm. At the indicated time points, aliquots were removed and subjected to SDS-PAGE to separate the enzyme from the substrate. After transfer of the denatured proteins to a nitrocellulose membrane, the extent of Cds1 phosphorylation was examined by autoradiography (top). Site-specific phosphorylation at Thr³²⁸ was examined by Western blotting with phospho-specific antibody (second panel from the top). The relative amounts of the proteins purified from S. pombe or E. coli were assessed by Western blotting using anti-HA antibody (third panel from the top) and Ponceau S staining (bottom panel), respectively. The kinase-dead Cds1(D312E), purified from HU-treated S. pombe, served as a negative control (right).

To follow autophosphorylation during autoactivation in vitro or in vivo, we generated three phospho-specific antibodies against peptides containing phosphorylated forms of the essential residues Thr³²⁸, Thr³³², and Tyr³⁵², identified in the experiment of Fig. 1. An analysis of the specificities of the three antibodies is presented in the supplemental material (Fig. S3). Since residues Thr³²⁸ and Thr³³² are only four residues apart, antibodies raised against the corresponding phosphopeptides exhibited significant cross-reactivity. However, the cross-reactions could be completely suppressed without affecting the specific reactions by including the cross-reacting phosphopeptide in the reaction mixture (Fig. S3, B and C).

Cds1(ΔC) purified from E. coli was incubated with ATP at 30 °C as described under “Experimental Procedures.” Aliquots were removed from the reaction at the indicated times and analyzed by SDS-PAGE (Fig. 3A, top). Incubation with ATP caused a slight reduction in the mobility of Cds1 (compare 20 min with 0 min), consistent with autophosphorylation of most, if not all, of the molecules. No further change in mobility occurred after 20–35 min of incubation, suggesting that autophosphorylation was complete by that time. Western blotting with the phospho-specific antibodies revealed that only one of the three essential phosphorylatable residues, Thr³²⁸, was autophosphorylated significantly in vitro. Thr³³² was not phosphorylated, confirming the results shown in Fig. 2. Surprisingly, Tyr³⁵² was not autophosphorylated in Cds1 purified from E. coli, unlike Cds1 purified from S. pombe (Fig. 2 and Fig. S2). We suspect that the difference is due to the fact that autophosphorylation of Tyr³⁵² only occurs in stable dimers of Cds1 (see “Discussion”). Unlike Cds1 purified from E. coli, Cds1 purified from S. pombe contains a small number of primed molecules that can efficiently dimerize. Consistent with this possibility, we observed autophosphorylation of Tyr³⁵² after induced dimerization of a Cds1-FKBP fusion protein purified from E. coli (11).

To confirm the results obtained with phospho-specific antibodies, the autophosphorylated Cds1-(25–454) was also analyzed by MS/MS spectrometry after digestion with subtilisin (Fig. 3C) or trypsin. This analysis led to the identification of 16 potential autophosphorylation sites, and among those, seven sites (Ser⁹², Ser²⁰⁷, Thr²¹³, Thr³²⁴, Thr³²⁸, Ser³⁴³, and Ser⁴³⁷) were manually verified from MS/MS spectra (data not shown). The finding that Thr³²⁸, as well as Thr³²⁴, a nonessential residue in the activation loop, were autophosphorylated is in agreement with the results from two-dimensional phosphopeptide mapping and Edman degradation. Phosphorylation of Thr³³² or Tyr³⁵² was never observed by mass spectrometry, consistent with the results from Western blotting. Phosphorylation of Ser³⁵⁰ and Thr³⁵⁴ has also been observed in Rad53 (corresponding to Thr³²⁴ and Thr³²⁸ of Cds1) by mass spectrometry (29). Analogous to Cds1, substitution of Thr³⁵⁴ and Thr³⁵⁸ with alanines in Rad53 sensitizes the cell to HU and MMS (34).

Kinase assays with MyBP demonstrated that preincubation in the presence of ATP increased the activity of Cds1 by more than 1000-fold (Fig. 3A; data not shown). Moreover, the extent of autoactivation was closely correlated with the extent of phosphorylation of Thr³²⁸, suggesting that phosphorylation of this residue may directly activate Cds1. However, it remained possible that Thr³²⁸ phosphorylation was a by-product, not the cause, of Cds1 activation, since other autophosphorylated residues in the catalytic domain of Cds1 could contribute to autoactivation, although they are clearly not essential. To rule out this possibility, we made use of native gel analysis and mass spectrometry to identify other phosphorylated residues in Cds1-(154–442), which contains the kinase domain but lacks the N-terminal SQ/TQ and FHA domains and most of the C-terminal tail.

Autophosphorylation of the Cds1 kinase domain (amino acids 154–442) generated three major species that can be separated from the unphosphorylated protein by electrophoresis in a native gel (Fig. 4B). Measurement of the ratios of ³²P incorporation to mass of protein in the fast moving bands clearly indicated that they are the products of increasing levels of phosphorylation. Consistent with this observation, mutation of the nonessential autophosphorylation site Thr³²⁴, identified in Fig. 2 and verified in Fig. 3C, removed one of the three new bands (compare T324V with wild type). By mass spectrometry and point mutagenesis, we identified another site of phosphorylation at Ser³⁷⁹, the last residue of the triple serine motif (Ser³⁷⁷–Ser³⁷⁹) in the kinase domain. Mutation of this residue and Thr³²⁴ to alanines converted almost all of the faster moving species to a single band with a mobility that was slightly higher than the unphosphorylated Cds1-(154–442) (Fig. S4A), indicating that Ser³⁷⁹ is autophosphorylated in a substantial fraction of the molecules. When the adjacent residue Ser³⁷⁸ was also mutated to alanine, all of the fast moving Cds1 molecules collapsed into a single species (Fig. S4A, compare T324V/S378A/S379A with T324V and T324V/S379A). As expected, an S. pombe strain containing mutations in Thr³²⁴ and the triple serine motif (Ser³⁷⁷–Ser³⁷⁹) exhibited normal checkpoint function in vivo, confirming that none of these residues are required for Cds1 activation (Fig. 1 and Fig. S4B). In addition, we observed that a mutant Cds1 kinase domain (amino acids 154–442, T324V/S378A/S379A) lacking all known autophosphorylation sites except Thr³²⁸ underwent autoactivation in vitro to nearly the same extent as the wild-type kinase domain (Fig. 4B, bottom two panels). The extent of autophosphorylation of Thr³²⁸ was closely correlated with the kinase activity of the mutant protein (Fig. 4B, middle two panels). These results, together with those shown in Figs. 1 and 3, demonstrated unambiguously that autophosphorylation of Thr³²⁸ is the only covalent modification required for Cds1 activation.

Thr³²⁸ Can Be Autophosphorylated in trans by Both Unactivated and Activated Cds1

To determine whether Cds1 autophosphorylates Thr³²⁸ in trans, we carried out kinase assays with kinase-dead Cds1 as the substrate. In the first set of experiments we assayed the activity of the isolated Cds1 kinase domain (Cds1-(154–442 T324V/S378A/S379A)), expressed in E. coli and purified by affinity chromatography (Fig. 4C, lane 1). A sample of the purified Cds1 kinase domain was activated by preincubation with ATP as described above and further purified by chromatography on a MonoQ column. The unactivated and activated Cds1 kinase domains were then incubated with purified kinase-dead substrate (Cds1-(154–442 T324V/S378A/S379A D312E); Fig. 4C, lane 2) in the presence of ATP. After SDS-PAGE to separate enzyme from substrate, phosphorylation of Thr³²⁸ was assayed by immunoblotting with Thr³²⁸-phospho-specific antibody. We observed that both the unactivated (Fig. 4C, lane 3) and the activated (Fig. 4C, lane 4) Cds1 kinase domains phosphorylated Thr³²⁸ in the kinase-dead substrate. As expected, the efficiency of phosphorylation of Thr³²⁸ by the activated Cds1 kinase domain was somewhat greater than that of the unactivated kinase domain.

In the second set of experiments, we assayed the ability of activated wild-type full-length Cds1 (or control kinase-dead Cds1), purified from HU-treated S. pombe cells, to phosphorylate kinase-dead Cds1(D312E) as substrate. Since the activated enzyme was tagged with an HA epitope at the C terminus, it could be separated from the untagged substrate Cds1(D312E) by SDS-PAGE after the kinase reactions. As shown in Fig. 4D, activated wild-type Cds1, but not the control kinase-dead Cds1(D312E), efficiently phosphorylated the kinase-dead Cds1 substrate (top). Immunoblotting with phospho-specific antibody demonstrated, as above, that Thr³²⁸ was a target of phosphorylation (second panel from the top). Judging from the intensities of the bands on the immunoblot, the number of substrate molecules containing phosphorylated Thr³²⁸ greatly exceeded the number of activated enzyme molecules, indicating that the reaction involved multiple rounds of catalysis.

These results clearly demonstrated that Thr³²⁸ is autophosphorylated in trans by either unactivated or activated Cds1. Since the isolated kinase domain of Cds1 was sufficient to mediate the reaction, autophosphorylation in trans can occur without stable dimerization as long as the enzyme concentration is sufficiently high. We suggest that autoactivation of Cds1 is initiated when unactivated Cds1 molecules transiently assume an active state and phosphorylate Thr³²⁸ in other molecules. The resulting activated Cds1 molecules can efficiently phosphorylate additional unactivated molecules, thus accelerating the autoactivation process.

Autophosphorylation of Thr³²⁸ and Cds1 Activation in Vivo

To verify that phosphorylation of Thr³²⁸ occurs in vivo, we purified Cds1 from HU-treated S. pombe cells by immunoprecipitation followed by SDS-PAGE. Western blotting with phospho-specific antibodies demonstrated that Thr³²⁸, but not Thr³³², is phosphorylated during the course of HU treatment (Fig. 5A). Since no Thr³²⁸ phosphorylation was detected in cells expressing the kinase-dead mutant (D312E), phosphorylation of Thr³²⁸ in vivo is due to autophosphorylation. The level of kinase activity of the purified Cds1 was closely correlated with the extent of Thr³²⁸ phosphorylation (compare Thr³²⁸ phosphorylation and ³²P incorporation in MyBP). A very weak signal was observed with Tyr³⁵²-phospho-specific antibodies. The significance of this signal, if any, is unclear, given that the in vitro autoactivation experiments showed that phosphorylation of Tyr³⁵² is not essential for Cds1 activation (see “Discussion”).

FIGURE 5.

Thr¹¹ and FHA domain-dependent trans-autophosphorylation of Thr³²⁸ activates Cds1 in vivo. A, wild-type Cds1 and the kinase-dead Cds1(D312E) mutant, tagged with HA, were expressed from the endogenous cds1⁺ locus. Cells were cultured in medium containing 25 mm HU at 30 °C. At each time point (marked on the top), extracts were prepared from equal numbers of cells, and the enzyme was immunoprecipitated with anti-HA antibody. After SDS-PAGE, phosphorylated Thr³²⁸, Thr³³², and Tyr³⁵² were detected by Western blotting with the corresponding phospho-specific antibodies. The relative loading of the samples was assessed by Western blotting with anti-HA antibody (fourth panel from the top). Cds1 kinase activity was measured using MyBP as substrate (lower two panels). B, phosphorylation of Thr³²⁸ depends on all three functional domains of Cds1. Wild-type Cds1 and its various mutants (noted at the top) were immunoprecipitated with anti-HA antibody and separated in SDS-PAGE as in A. Phosphorylation of Thr¹¹ (upper panel) and Thr³²⁸ (middle panel) were detected by Western blotting with the corresponding phospho-specific antibody. The same blot was stripped and reprobed with anti-HA antibody to reveal the presence of Cds1 (lower panel).

Previous studies have shown that the essential phosphorylatable residue Thr¹¹ in the SQ/TQ domain of Cds1 is phosphorylated by Rad3 in response to HU treatment (11, 18). To verify the requirements for phosphorylation of the two critical Cds1 residues, Thr¹¹ and Thr³²⁸, in vivo, we made use of phospho-specific antibodies to determine the effects of various Cds1 mutations (Fig. 5B). Consistent with previous reports, phosphorylation of Thr¹¹ is dependent on the FHA domain but not on a functional kinase domain (16). Unlike Thr¹¹, phosphorylation of Thr³²⁸ is dependent on a functional kinase domain as well as the SQ/TQ and FHA domains. These results indicate that Thr³²⁸ is autophosphorylated in vivo and that autophosphorylation requires both a priming event (Thr¹¹ phosphorylation by Rad3) and FHA-dependent dimerization.

A weak signal of Thr³²⁸ phosphorylation was reproducibly observed for the Y352F mutant (Fig. 5B, middle). Detailed analysis showed that the Cds1(Y352F) mutant can survive low concentrations of HU better than the Cds1(T328A) and Cds1(T332A) mutants or the Cds1 null mutant, indicating that it retains residual checkpoint function (Fig. S1). This result is consistent with the hypothesis that phosphorylation of Thr³²⁸, but not Tyr³⁵², is absolutely required for Cds1 activation (see “Discussion”).

The C Terminus of Cds1 Negatively Regulates Cds1 Autoactivation

As noted above, the C-terminal domain of Cds1 is rich in hydrophilic and phosphorylatable residues (see Fig. 1 and Fig. S6). To explore its potential function, we analyzed the effects of a mutation, Cds1(ΔC), lacking the two triple serine motifs (Δ434–460). The mutation did not affect cell survival following treatment with HU or MMS, indicating that the C-terminal tail is not essential for Cds1-dependent checkpoint activation (Fig. 1, bottom). To determine whether the mutation affected the ATP-dependent autoactivation of Cds1, we performed two-stage autoactivation assays in vitro. For this purpose, unactivated Cds1(ΔC) and wild-type Cds1 proteins were purified from Δrad3Δcds1 S. pombe cells. For comparison, we also purified “activated” Cds1 from HU-treated wild-type S. pombe cells. The purified proteins were preincubated with ATP at 30 °C to allow autophosphorylation/autoactivation and then assayed for protein kinase activity using two established assays. In the first assay (37), [γ-³²P]ATP and MyBP were added to the reaction mixtures, and the incorporation of radioactivity into the substrate was determined by SDS-PAGE after a 20-min incubation (Fig. 6A, top three panels). In the second assay (36), a GST-Wee1 fusion protein bound to glutathione beads was added to cell extracts to pull down active Cds1. After incubation of the beads with [γ-³²P]ATP for 20 min in the kinase reaction buffer, the incorporation of radioactivity into GST-Wee1 was determined by SDS-PAGE followed by PhosphorImaging (Fig. 6A, middle two panels). A relatively low Cds1 concentration (∼100 nm), approximating the physiological concentration in S. pombe (11), was employed in this experiment. At this concentration, preincubation of inactive wild-type Cds1 with ATP resulted in only a modest increase in kinase activity over the background (compare lane 4 with lane 3). Preincubation of “activated” wild-type Cds1 also resulted in a small increase in activity (compare lane 12 with lane 11), indicating that HU treatment of S. pombe did not fully activate the kinase. In contrast, preincubation of inactive Cds1(ΔC) with ATP at 30 °C (but not 0 °C) resulted in a dramatic increase in kinase activity (14-fold) in both assays (compare lane 8 with lanes 6 and 7), indicating that the C-terminal domain inhibits the activity or the autoactivation of Cds1. To distinguish between the two possibilities, we examined the phosphorylation of Thr³²⁸ in Cds1 after preincubation with ATP (Fig. 6A, bottom two panels). The results showed that preincubation of inactive wild-type Cds1 with ATP slightly increased the phosphorylation of Thr³²⁸. In contrast, Thr³²⁸ phosphorylation was dramatically increased in Cds1(ΔC) under similar conditions, consistent with the possibility that the C terminus of Cds1 can suppress the autoactivation of Cds1. As expected, activated wild-type Cds1 contains phosphorylated Thr³²⁸, and incubation with ATP resulted in further phosphorylation of the protein.

To determine whether the C-terminal tail can inhibit Cds1 autoactivation in vivo, we expressed Cds1(ΔC) under the control of two promoters of different strengths, the “weak” nmt promoter in the plasmid pREP81X and the “intermediate” nmt promoter in the plasmid pREP41X (Fig. 6B). As previously reported (11, 36, 41), overexpression of wild-type Cds1 resulted in elongation of the cells, indicative of cell cycle delay (compare wild-type Cds1 with Cds1(D312E) or vector control). The extent of the observed elongation was correlated with the strength of the promoter, consistent with the hypothesis that partial activation of Cds1 occurs when the protein is expressed at higher levels. For both promoters, we consistently observed that the extent of elongation was greater for Cds1(ΔC) than for wild-type Cds1 (Fig. 6B and Fig. S5A), indicating a higher level of kinase activation. This increased kinase activation was observed although the Cds1(ΔC) was expressed at a level about 20% of that of wild-type Cds1, as judged by Western blotting (Fig. S5B). These results indicate that the C-terminal tail of Cds1 suppresses spontaneous activation of Cds1 and suggest that it may contribute to the noise immunity of the replication checkpoint during normal cellular growth.

DISCUSSION

We have previously shown that the effector kinase Cds1 of the replication checkpoint in S. pombe is activated by a two-stage mechanism (11). In the first or priming stage, the sensor kinase Rad3 functions cooperatively with the mediator Mrc1 to phosphorylate Cds1 at Thr¹¹ in the SQ/TQ domain. In the second or autoactivation stage, primed Cds1 molecules dimerize by two identical interactions between the phosphorylated Thr¹¹ and the FHA domain of another Cds1 molecule. Dimerization facilitates autophosphorylation of Cds1 leading to the activation of the kinase. Here, we show that in the second stage, Cds1 is activated by trans-autophosphorylation of a single residue, Thr³²⁸, in the activation loop. Phosphorylation of Thr³²⁸ is probably initiated by inactive Cds1 in a transient active state and is rapidly amplified by activated Cds1 in which Thr³²⁸ has been phosphorylated. We also show that the C-terminal tail of Cds1 suppresses the autophosphorylation of Thr³²⁸. This autoinhibitory mechanism, together with the requirement for phosphorylation-induced dimerization, ensures that spontaneous autophosphorylation occurs at a very slow rate in an unperturbed cell cycle (i.e. the system has high noise immunity). Like many other protein kinases whose activation requires phosphorylation of the activation loop, autophosphorylation of Thr³²⁸ presumably activates Cds1 by promoting a local conformational change in the catalytic center that allows proper binding to the substrates and/or correct disposition of catalytic groups. This two-stage model for Cds1 activation (see Fig. 7) provides specificity and sensitivity as well as noise immunity, properties required for the replication checkpoint. It is also consistent with all genetic data available so far and is probably shared by all other protein kinases in the Chk2 family.

FIGURE 7.

Proposed mechanisms for autoinhibition and autoactivation of Cds1. Under normal conditions, Cds1 is largely monomeric, and the trans-autophosphorylation of Thr³²⁸ in monomeric Cds1 is suppressed by the autoinhibitory function of the C-terminal tail. The activity of the replication checkpoint is kept low during the normal cell cycles by the low intracellular concentration of Cds1, the inactive conformation of the activation loop, and the fact that Thr³²⁸ is a poor substrate for the enzyme. When DNA replication is perturbed, Rad3 functions in cooperation with the mediator Mrc1 to catalyze the site-specific phosphorylation of Thr¹¹ in the SQ/TQ domain of Cds1. Phosphorylation of Thr¹¹ promotes dimerization of the Cds1 molecules by two identical intermolecular interactions between the phosphorylated Thr¹¹ of one Cds1 molecule and the FHA domain of the other. The resulting high local concentration of Cds1 overcomes the autoinhibitory function of the C-terminal tail so that the trans-autophosphorylation of Thr³²⁸ can proceed. It is proposed that phosphorylation of Thr³²⁸ is initiated by an inactive Cds1 (green) in a “transient” active state (purple) and then amplified by activated Cds1 molecules (red) in which Thr³²⁸ has been phosphorylated. Phosphorylation of Thr³²⁸ is believed to activate Cds1 by promoting a local conformational change in the catalytic center that facilitates substrate binding and/or aligns key catalytic residues.

trans-Autophosphorylation of Thr³²⁸ and Cds1 Activation

Our data provide strong evidence that trans-autophosphorylation of Thr³²⁸ is the only requirement for the activation of primed Cds1 in the second or autoactivation stage. First, substitution of Thr³²⁸ with alanine, glutamic acid, or aspartic acid (data not shown) abolished Cds1 activity both in vivo and in vitro. Second, phosphorylation of Thr³²⁸ was detected by two-dimensional phosphopeptide mapping and verified in vivo and in vitro using two different methodologies: Western blotting with phospho-specific antibody and MS/MS spectrometry. Third, the extent of Thr³²⁸ phosphorylation was closely correlated with the level of increased kinase activity of Cds1 both in vitro and in vivo. Finally, a mutant form of the Cds1 kinase domain, in which Thr³²⁸ was the only available phosphorylation site, was activated by autophosphorylation to the same extent and with the same kinetics as the wild-type Cds1 kinase domain.

Autophosphorylation can occur at other sites in Cds1. We showed that these sites are irrelevant for activation of the enzyme. Mutation of only two phosphorylatable residues in the Cds1 kinase domain, other than Thr³²⁸, give rise to checkpoint defects. Of these, Thr³³² is not phosphorylated and probably plays a catalytic role. A previous study has reported that Chk2 is autophosphorylated at Thr³⁸⁷ (22), the residue that corresponds to Thr³³² of Cds1. Although it is possible that Thr³³² phosphorylation is required for Chk2, but not Cds1, activation, a more plausible explanation is that the discrepancy is due to technical factors, perhaps cross-reactivity of the antibodies used to detect phospho-Thr³⁸⁷ in Chk2. We did not observe Thr³³² phosphorylation by any method, and the homologous residue is not phosphorylated in the crystal structures of active kinases that are dependent upon T-loop phosphorylation. Previous studies of Chk2 and Rad53 have also identified some additional sites of autophosphorylation outside of the activation segment, but consistent with our data, these sites have only small effects on Chk2 and Rad53 activation (29, 30, 32, 33).

We observed that Tyr³⁵², the third essential phosphorylatable residue, was autophosphorylated in vitro by Cds1 purified from S. pombe (Fig. S2C). However, we did not obtain any convincing evidence that phosphorylation of Tyr³⁵² occurs in vivo or that it is required for Cds1 autoactivation. The weak signal observed with Tyr³⁵²-phospho-specific antibody in S. pombe extracts (Fig. 5A) was due, at least in part, to cross-reactivity of the antibody with other phosphorylated tyrosine residues in Cds1 (data not shown). Thus, our data suggest that the phosphorylation of Tyr³⁵² occurs with an extremely low efficiency in vivo if at all. Moreover, our biochemical studies of Cds1 expressed in E. coli clearly demonstrated that efficient autoactivation of the enzyme occurs without any detectable Tyr³⁵² phosphorylation (Fig. 3). The phosphorylation of Tyr³⁵² observed in vitro with the enzyme purified from S. pombe may be due to a small fraction of the purified Cds1 molecules that had already been primed in vivo for dimerization and autophosphorylation. We conclude that the phosphorylation of Thr³²⁸, but not Tyr³⁵², is absolutely required for Cds1 activation. The precise mechanism by which the Y352F mutation reduces the activity of Cds1 is not yet completely clear. Tyr³⁵² is a highly conserved residue in protein kinases of the Cds1 family and is located in the αEF/αF loop, which interacts with and probably stabilizes the activation loop (24, 38). The mutation may cause a structural change that prevents the initial phosphorylation of Thr³²⁸ by unactivated Cds1 or decreases the catalytic activity of the enzyme after phosphorylation of Thr³²⁸ or both.

Mechanism of Thr³²⁸ Autophosphorylation

The consensus substrate sequence for activated Cds1 is similar to that of Chk2 (LXRXXS) (42), and, like Chk2, the enzyme preferentially phosphorylates serine residues (data not shown). The sequence surrounding Thr³²⁸ is a poor match to the consensus, suggesting that it would be a very unfavorable substrate for monomeric Cds1. This fact probably contributes to the noise immunity of the checkpoint pathway. Two general models have been suggested to explain how autophosphorylation might occur in the context of a dimer (11, 19, 43, 44). One possibility is that the activation loop in an inactive Cds1 molecule transiently assumes an “activated” state, allowing the enzyme to phosphorylate Thr³²⁸ of its dimeric partner. Since the transiently “activated” state would be energetically unfavorable and Thr³²⁸ is a poor substrate, a high local concentration of enzyme, mediated by dimerization, would be required for the reaction to occur at an appreciable rate. In this scenario, autophosphorylation and activation of one Cds1 molecule would rapidly lead to autophosphorylation of the other because of the much greater activity of the enzyme when Thr³²⁸ is phosphorylated. A second model of autoactivation is based upon the recent determination of a crystal structure for the catalytic domain of Chk2 (19, 43). In this model, dimerization promotes the exchange of activation loops between the two partners, producing an “active” conformation in both kinase molecules. It is proposed that trans-autophosphorylation of Chk2 Thr³⁸³ (corresponding to Cds1 Thr³²⁸) in an exchanged activation loop can readily occur because it is near the optimal site for phosphoryl transfer. The key feature of this model is that the exchange of activation loops simultaneously results in an active conformation of both catalytic sites and effective binding of what would otherwise be poor substrate. Our data do not directly distinguish between the two models. However, we have shown that largely monomeric Cds1 catalytic domains undergo efficient autophosphorylation in vitro, so the exchanged loop structure, if it occurs in solution, is transient and does not require stable dimerization. We have also shown that activated Cds1, in which the activation loop is presumably stabilized in its active conformation by the phosphorylation of Thr³²⁸, can efficiently phosphorylate Thr³²⁸ in inactive Cds1. In this case, an exchange of loops would not be required for the catalytic center to adopt the “active” state, and the exchanged loop structure would presumably be highly disfavored.

Regulation of Cds1 Autoactivation

We have observed that the C-terminal tail of Cds1 is a negative cis-regulatory element for Cds1 activation. The autoinhibitory function of the C terminus mirrors the function of the C-terminal domain of Chk1 (45), the effector kinase in the DNA damage checkpoint pathway. However, the kinase domain of inactive Chk1, unlike that of Cds1, is already in an active state (but blocked by the autoinhibitory C-terminal domain) (45). Phosphorylation of Chk1 by Rad3 removes the inhibitory effect of the C-terminal domain, allowing the kinase domain of Chk1, already in the active state, to bind and phosphorylate its substrates (46–48). In the case of inactive Cds1, the C-terminal tail inhibits Thr³²⁸ autophosphorylation in unprimed monomeric molecules, possibly by stabilizing the inactive state of the catalytic site of the kinase domain, thereby reducing the fraction of enzyme in the transiently active state. Phosphorylation of Thr³²⁸, facilitated by dimerization, locks Cds1 into the active state. We observed that deletion of the C-terminal tail does not make Cds1 constitutively active in vivo. Under physiologic conditions with a low intracellular Cds1 concentration, the Cds1(ΔC) mutant is still dependent upon priming and dimerization steps for activation.

The C-terminal tail contains a number of phosphorylatable residues, including two triple serine motifs that are confirmed substrates of the enzyme (Fig. S4A), which raises the possibility that autophosphorylation of these sites might modulate the autoinhibitory effect of the C-terminal tail and/or provide some fine tuning of the autoactivation process. However, simultaneous mutation of all phosphorylatable sites in this domain does not increase the drug sensitivity (see Fig. S6). Further work will be required to determine the precise mechanism of autoinhibition by the C-terminal tail.

Given the profound effect of the replication checkpoint on cell cycle progression, it would seem important to prevent adventitious activation of the pathway when DNA replication is unperturbed. As we have shown, several factors function to prevent spontaneous activation of Cds1. First, the intracellular concentration of the enzyme is quite low (∼80 nm), making dimerization strongly dependent upon phosphorylation of the SQ/TQ domain. Second, the transiently active form of the enzyme is presumably energetically unfavorable, so that dimerization is required for autophosphorylation to proceed at a reasonable rate. Third, the critical autophosphorylation site, Thr³²⁸, is a poor substrate because it resides in a noncanonical sequence context. Finally, the rate of autophosphorylation is reduced by the autoinhibitory effect of the C-terminal tail. Although these mechanisms contribute to high noise immunity, the two-stage mechanism for activation of Cds1, which involves a kinase cascade, also provides the opportunity for significant amplification of an initiating signal, such as a stalled replication fork (11, 20, 22, 49). It seems likely that both of these critical properties are conserved in other species.