Role of the N-terminal Forkhead-associated Domain in the Cell Cycle Checkpoint Function of the Rad53 Kinase* 210

Forkhead-associated (FHA) domains are multifunctional phosphopeptide-binding modules and are the hallmark of the conserved family of Rad53-like checkpoint protein kinases. Rad53-like kinases, including the human tumor suppressor protein Chk2, play crucial roles in cell cycle arrest and activation of repair processes following DNA damage and replication blocks. Here we show that ectopic expression of the N-terminal FHA domain (FHA1) of the yeast Rad53 kinase causes a growth defect by arresting the cell cycle in G1. This phenotype was highly specific for the Rad53-FHA1 domain and not observed with the similar Rad53-FHA2, Dun1-FHA, and Chk2-FHA domains, and it was abrogated by mutations that abolished binding to a phosphothreonine-containing peptide in vitro. Furthermore, replacement of the RAD53 gene with alleles containing amino acid substitutions in the FHA1 domain resulted in an increased DNA damage sensitivity in vivo. Taken together, these data demonstrate that the FHA1 domain contributes to the checkpoint function of Rad53, possibly by associating with a phosphorylated target protein in response to DNA damage in G1.

Forkhead-associated (FHA) domains are multifunctional phosphopeptide-binding modules and are the hallmark of the conserved family of Rad53-like checkpoint protein kinases. Rad53-like kinases, including the human tumor suppressor protein Chk2, play crucial roles in cell cycle arrest and activation of repair processes following DNA damage and replication blocks. Here we show that ectopic expression of the N-terminal FHA domain (FHA1) of the yeast Rad53 kinase causes a growth defect by arresting the cell cycle in G 1 . This phenotype was highly specific for the Rad53-FHA1 domain and not observed with the similar Rad53-FHA2, Dun1-FHA, and Chk2-FHA domains, and it was abrogated by mutations that abolished binding to a phosphothreonine-containing peptide in vitro. Furthermore, replacement of the RAD53 gene with alleles containing amino acid substitutions in the FHA1 domain resulted in an increased DNA damage sensitivity in vivo. Taken together, these data demonstrate that the FHA1 domain contributes to the checkpoint function of Rad53, possibly by associating with a phosphorylated target protein in response to DNA damage in G 1 .
Checkpoint signaling pathways are crucial for maintaining genome integrity by delaying the cell cycle in response to DNA damage, replicational stress or mitotic spindle assembly defects, and are highly conserved throughout evolution (1,2). Chk2-like cell cycle checkpoint protein kinases are characterized by the presence of noncatalytic FHA 1 domains and play crucial roles in the cellular response to DNA damage and replication blocks (3,4). For example, the human tumor suppressor kinase Chk2 phosphorylates p53 (5-7), Cdc25C (8), and BRCA1 (9) in response to DNA double strand breaks, leading to cell cycle arrest in G 1 , G 2 /M, and S phase, respectively. Mutations within the Chk2-FHA domain or truncations that disrupt the kinase domain but leave the FHA domain intact are the cause of a subset of cases of the Li-Fraumeni syndrome (usually caused by loss of p53) where patients suffer from multiple independent primary tumors, typically at a young age (10).
Budding yeast, Saccharomyces cerevisiae, contains three FHA domain-containing protein kinases. Rad53 (11,12) and Dun1 (13) function in the mitotic checkpoint response to DNA damage and replication blocks, and Mek1 plays an important role in meiosis (14). In contrast to other family members that contain a single FHA domain in the N terminus, Rad53 is the only checkpoint kinase with an additional FHA domain in the C terminus. RAD53 arrests the cell cycle in G 1 and G 2 /M in response to DNA damage and in S phase in response to replication blocks (12,15). In addition, it regulates the transcriptional induction as well as the subcellular localization and activity of proteins involved in DNA damage repair (12,16,17). RAD53 also has an essential, DNA damage-independent function by regulating the levels and activity of ribonucleotide reductase, and thereby the availability of dNTPs, during S phase (18,19). Several RAD53-dependent checkpoint pathways, but few definitive kinase substrates, have been identified. Following DNA damage in G 1 , Rad53 phosphorylates the Swi6 transcription factor, thereby contributing to the repression of the CLN1 and CLN2 cyclin genes and delaying transition into S phase (20). In response to replication blocks and DNA damage in S phase, Rad53 inhibits the "firing" of late replication origins (21) by negatively regulating the Dbf4/Cdc7 kinase complex (22)(23)(24). As part of the G 2 /M DNA damage checkpoint, Rad53 prevents anaphase entry and mitotic exit by inhibition of the polo-like kinase Cdc5 (25). FHA domains are highly diverse protein-protein interaction modules characterized by a 55-75-residue motif with only 7 residues that are Ͼ65% conserved among 120 family members (26). However, intact FHA domains are much larger than the core motif detected in computer comparisons and contain up to 180 amino acid residues (27,28). FHA domains form an 11stranded ␤-sandwich that is similar to the fold of the C-terminal domain of the smad tumor suppressor/transcription factor family (27). In addition to checkpoint kinases, FHA domains are also found in several other signaling molecules including Forkhead-like transcription factors, from which their name originated. Recent reports indicate that FHA domains bind directly to threonine-phosphorylated residues in target proteins (29). In the case of Rad53, it has been demonstrated that association of the C-terminal FHA2 domain with phosphorylated Rad9 is essential for Rad53 activation and G 2 /M arrest following DNA damage (30). Interestingly, the Rad53-FHA2 domain can also bind directly to phosphotyrosine-containing peptides in vitro, indicating that FHA domains may be bifunctional phosphothreonine/phosphotyrosine-binding domains (31). In vitro, the N-terminal FHA1 domain of Rad53 also binds to phosphorylated Rad9 as well as to phosphothreonine-containing synthetic peptides (29), but its physiological function is presently unclear.
We have recently determined the Rad53 and Dun1 FHA domain boundaries and demonstrated that overexpression of the Dun1-FHA domain in yeast can compete the Dun1-depend-ent transcriptional induction of ribonucleotide reductase 2 (RNR2) following replication blocks (28). To gain insight into the function of the Rad53-FHA1 domain, we have used a similar approach in the present study. We show that overexpression of the FHA1 domain can mimic the DNA damage-dependent G 1 arrest function of the intact Rad53 kinase, and that this phenotype is abrogated by mutation of residues that are involved in phosphopeptide binding. Moreover, we show that introduction of amino acid changes, that disrupt this overexpression phenotype, into the RAD53 gene results in an increased sensitivity of yeast to DNA damage. These data indicate that FHA1 domain association with a threoninephosphorylated target protein may be involved in the G 1 checkpoint function of Rad53.

EXPERIMENTAL PROCEDURES
Generation of FHA Domain Constructs-FHA domain constructs were generated by polymerase chain reaction (PCR) using synthetic oligonucleotides (all oligonucleotide sequences are available as on-line supplemental materials) as described (28). For bacterial expression, cDNAs were cloned into the NcoI and BglII sites of pQE60 (Qiagen). For yeast expression, constructs were ligated to the Gal4 DNA binding domain for nuclear targeting by cloning into the constitutive expression vector pAS2 (CLONTECH), followed by ligation of the resulting Hin-dIII-SalI restriction fragments into the inducible vector p416GAL1 (32). Sequences of all PCR generated fragments were confirmed by automated DNA sequence analysis (ABI).
Site-directed and Random Mutagenesis-Site-directed FHA1 mutants were generated by PCR using synthetic oligonucleotides (see supplemental materials available in on-line version of this article) and mutations were confirmed by automated DNA sequence analysis. Random mutagenesis of the Rad53-FHA1 was performed by low fidelity PCR essentially as described (33), except that 20 M manganese chloride was found to give the highest yield of viable yeast colonies with only single amino acid substitutions. The resulting library in pAS2 consisted of ϳ1950 independent clones.
In synchronization experiments, cells were diluted to 0.2 optical density units and treated twice for 2 h with 20 g/ml ␣-factor. Cells were released from ␣-factor arrest by three washes in selective medium containing 2% sucrose and 4% galactose and resuspended in fresh medium.
Flow Cytometry-For cell cycle analyses, 1-ml aliquots were removed from yeast cultures and fixed in 70% ethanol. 200 l of cells were washed with 50 mM sodium citrate, pH 7, and incubated overnight with 0.1 mg/ml RNase A in 50 mM sodium citrate at 37°C. Cells were sonicated, treated with 4 g/ml propidium iodide, and analyzed using a Becton Dickinson FACScan and CellQuest software.
RAD53 Gene Disruptions and Allele Replacements-RAD53 gene disruptions and allele replacements were generated in the sml1⌬ strain that is viable without Rad53. RAD53 was disrupted using a PCR product containing the URA3 selectable marker flanked by a 45-base pair RAD53 genomic sequence (see supplemental information on-line).
RAD53 allele replacements (R70A, N107K, and K227A) were performed using a PCR-based strategy similar to the one described by Erdeniz et al. (37), except that the S. cerevisiae rather than the Kluyveromyces lactis URA3 gene was used as the tagged selectable marker and that fusion fragments corresponded only to the targeted domains rather than the entire coding sequence. Strains containing mutated rad53 alleles were identified by colony PCR, using primers outside the PCRgenerated targeting sequences, followed by hybridization with discriminatory oligonucleotides. The correct mutations and absence of secondary mutations were confirmed by automated DNA sequence analysis. For survival assays, aliquots were removed from log phase cultures immediately before and 3 h after addition of MMS, and spread onto YPD plates.

Overexpression of the Rad53-FHA1 Domain Causes a Growth
Defect in Yeast-We used a galactose-inducible overexpression system to identify phenotypes that might provide clues to the function of the N-terminal FHA domain (FHA1) of the yeast Rad53 kinase. Transformed yeast strains had normal growth properties when the expression of these constructs (fused to the Gal4-DNA binding domain for nuclear targeting) was repressed in media lacking galactose (Fig. 1A). However, strains overexpressing the entire Rad53 N terminus (residues 1-199 including the FHA1 domain) or expressing a construct containing only the biochemically defined FHA1 domain (residues 20 -164) had a dramatic growth defect and failed to form visible colonies on solid media containing galactose (Fig. 1A). This growth defect was highly specific for the Rad53-FHA1 domain as it was not observed with the related Rad53-FHA2, Dun1-FHA, and Chk2-FHA domains (Fig. 1A) that were expressed at similar protein levels (Fig. 1B). In addition, this phenotype was dependent on the integrity of the FHA1 domain as it was abrogated by doublealanine substitution of the highly conserved residues Ser-85 and His-88, as well as truncation of just 15 residues at either end of the proteolytically defined domain boundaries (Fig. 1A).
Although overexpression of the C-terminal Rad53-FHA2 domain did not affect normal cell growth, it caused a considerably increased DNA damage sensitivity and decreased viability on plates containing galactose and the radiomimetic compound MMS (Fig. 1A, right panel). This phenotype is consistent with genetic evidence that the Rad53-FHA2 domain is involved in the DNA damage-dependent activation of Rad53 (30). Likewise, the Dun1-FHA domain construct used here can suppress the Dun1-dependent transcriptional induction of RNR2 following replication blocks (28) and can also compete the activation of Dun1 kinase by replication blocks or DNA damage. 2 Therefore, the Rad53-FHA2 and Dun1-FHA domains are fully functional in this system, and the fact that only the FHA1 domain causes a growth defect underscores the specificity of this phenotype.
The FHA1 Growth Defect Is Unrelated to the Essential Function of RAD53-RAD53 has an essential DNA damage-independent function by regulating the availability of dNTPs during S phase, that can be suppressed by deletion of the RNR inhibitor SML1 (18). To test if the FHA1 domain causes the growth defect by competing with the RAD53 essential function, we analyzed its effect on the growth of a sml1⌬ strain (Fig. 2). In this experiment, expression of the FHA1 domain again inhibited colony formation in the sml1⌬ strain to the same extent as in the wild-type strain. Similar results were also obtained in a rad53⌬ strain (viable because of a sml1-1 mutation; Fig. 2). This indicates that the FHA1 overexpression phenotype is unrelated to the essential function of RAD53.
Another possible reason for the growth defect could be that FHA1 overexpression generates a checkpoint signal that results in cell cycle arrest. The FHA1 domain can bind to phosphorylated Rad9 in vitro (29). To test if the growth defect is the result of an interaction of the FHA1 domain with Rad9, overexpression experiments were repeated in a rad9 deletion strain. Fig. 2A shows that overexpression of the Rad53-FHA1 domain, but not the vector and double-mutant controls, again caused a dramatic growth defect in the rad9⌬ strain; the FHA1 phenotype must therefore be independent of its proposed interaction with Rad9. Similar results were also obtained in a strain lacking RAD17 (Fig. 2A), a component of an alternative DNA damage-sensor pathway for Rad53 activation (38), and in a strain lacking DUN1 (Fig. 2B), a Rad53 effector at the transducer level. The maintenance of the FHA1 overexpression phenotype in rad9⌬, rad17⌬, rad53⌬, and dun1⌬ strains indicates that the growth defect does not result from an FHA1-generated checkpoint signal at the sensor/transducer level, e.g. by displac-ing Rad53 from inhibitory complexes. This genetic evidence was confirmed by immunoblot analysis of the Rad53 activation state. Rad53 is activated by phosphorylation resulting in slower migrating bands in SDS-polyacrylamide gels (39). Fig.  2C shows that expression of the FHA1 domain did not result in shifted (i.e. activated) Rad53 under basal conditions, nor did it inhibit Rad53 activation by MMS-induced DNA damage or hydroxyurea-induced replication blocks.
FHA1 Overexpression Arrests the Cell Cycle in G 1 -To assess the FHA1 phenotype in more detail, overexpression effects were examined in liquid cultures. Fig. 3A shows that FHA1overexpressing cultures had considerably slower growth kinetics than the vector control, which became apparent after ϳ8 h in galactose-containing media. To determine whether overexpression of the FHA1 domain affects cell viability, aliquots of these cultures were plated on dextrose-containing plates and tested for their ability to form colonies. Colony formation by the FHA1-overexpressing strain was directly proportional to its optical density throughout the 24-h experiment, and colony numbers per optical density unit were very similar for the FHA1 and control cultures (Fig. 3A, inset). This demonstrates that the FHA1-dependent growth defect is reversible and not the result of increased lethality.
To examine if the FHA1 domain reduces cell growth in a specific phase of the cell cycle, aliquots of liquid cultures were analyzed by flow cytometry. Starting cultures were adjusted to give equal cell densities (ϳ0.5 optical density units) at the end of this experiment to minimize nonspecific effects on the cell cycle profile. In this experiment, the majority of vector control cells had a 2n DNA content (corresponding to G 2 /M) before, and 6 and 12 h after addition of galactose (Fig. 3B). The cell cycle profile of the FHA1 strain without, and 6 h after addition of galactose (i.e. at time points when the growth defect was not apparent) was essentially identical to the vector control. However, induction of FHA1 expression for 12 h resulted in a marked increase in cells with a 1n DNA content (Fig. 3B). To determine whether this accumulation of cells in G 1 was the result of a cell cycle arrest or just a slower G 1 -S transition, FHA1-overexpressing and vector control cultures were synchronized in G 1 using ␣-factor and then released into fresh galactose-containing medium. Fig. 3C shows that the majority FIG. 1. FHA domain overexpression  phenotypes. A, yeast strains containing vector-control, Rad53 N terminus, Rad53-FHA1, Rad53-FHA1 mutant (20 -164 with substitutions S85A/H88A), truncated Rad53-FHA1, Rad53-FHA2, Chk2-FHA, and the Dun1-FHA p416GAL1 constructs were plated on 2% sucrose (left), 2% sucrose ϩ 4% galactose to induce FHA domain expression (center), and on 2% sucrose ϩ 4% galactose ϩ 0.007% MMS (only the relevant half of the MMS plate is shown). B, immunoblot analysis of protein extracts of the different strains after 6-h induction with 4% galactose. FHA domain fragments were detected using an anti-Gal4 antibody, and samples were reprobed using an anti-actin antibody as a loading control. of the control culture had shifted to G 2 /M within 75 min after ␣-factor release. In contrast, the majority of FHA1-overexpressing cells were still in G 1 150 min after release from the pheromone (Fig. 3C). ␣-Factor arrest kinetics (before the release) were similar for the FHA1 and control culture (data not shown), indicating that other phases of the cell cycle were largely unaffected. Therefore, cell cycle arrest in the G 1 phase is the major reason for the inability of the FHA1-overexpressing strain to form visible colonies on solid medium.
Dependence of the FHA1 Phenotype on Its Phosphopeptidebinding Ability-Only seven residues are highly conserved among FHA domains (26). To find out if these highly conserved residues are important for FHA1 domain function, we performed an alanine-screening mutational analysis in a constitutive expression system. The top panel in Fig. 4A shows that the FHA1 phenotype could be reproduced in this system; expression of the wild-type FHA1 domain again prevented colony formation, whereas equal amounts of yeast transformed with equal amounts of vector or double-mutant (S85A/H88A) control DNAs gave rise to numerous colonies. Fig. 4A shows that single alanine substitutions of each of the Ͼ80% consensus residues (Gly-69, Arg-70, Ser-85, His-88) as well as Asn-107 and Gly-108, that are conserved in Ͼ65% of all FHA domains, abrogated the overexpression phenotype and resulted in numerous viable colonies. In contrast, alanine substitution of the Ͼ65% conserved residue Asn-112 did not affect the phenotype, indicating that this residue is functionally neutral in the context of the FHA1 domain. Immunoblot analyses of viable colonies demonstrated that three of the alanine substitutions (G69A, H88A, G108A) resulted in considerably reduced FHA1 protein levels, FIG. 3. Cell cycle analysis of the Rad53-FHA1 growth defect phenotype. A, yeast strains containing vectorcontrol (circles) and Rad53-FHA1 (squares) p416GAL1 constructs were grown for 24 h in 2% sucrose ϩ 4% galactose. At 2-h intervals, aliquots were removed and cell growth monitored by optical density measurements (filled symbols), and cell viability determined by plating on YPD medium (open symbols). In the inset, the same values are plotted as cell viability versus cell growth for a linear fit (vector-control, broken line; FHA1, solid line). B, 1-ml aliquots were removed from exponentially growing vector-control and Rad53-FHA1 containing cultures at the time points indicated, and analyzed by flow cytometry. C, vector-control and FHA1 p416GAL1 constructs induced for 10 h in 4% galactose were diluted to 0.2 optical density units, synchronized in G 1 by ␣-factor, and released into galactose medium. Aliquots removed at the times indicated were analyzed by flow cytometry.

FIG. 2. Maintenance of the FHA1 overexpression phenotype in checkpoint-defective strains and an extragenic suppressor strain of the RAD53 essential function.
A, vector-control, Rad53-FHA1, and Rad53-FHA1 mutant (S85A/H88A) p416GAL1 constructs in wild-type, rad9⌬, and rad17⌬ strains in the BY4741 background were plated on 2% sucrose (top), and on 2% sucrose ϩ 4% galactose. B, cultures of vector-control (V) and Rad53-FHA1 (F) constructs in the W303-1A background (wild-type, rad53⌬sml1-1, sml1⌬), and the K699 background (wildtype, dun1⌬) (from left to right), were adjusted to 0.05 optical density units and 2 l plated in a series of 10-fold dilutions on 2% sucrose (top) and on 2% sucrose ϩ 4% galactose. C, yeast strains induced to express vector-control, Rad53-FHA1, and Rad53- indicating that these residues may be important for proper domain folding (Fig. 4B). In contrast, FHA1 domains with substitutions of Arg-70, Ser-85, or Asn-107 were still expressed at high levels ( Fig. 4B) that were similar to the galactoseinducible wild-type FHA1 preventing cell growth (Fig. 4C). The abrogation of the growth defect therefore indicates that these residues may be directly involved in the binding to a target protein that is crucial for the G 1 arrest phenotype.
To test these hypotheses, His 6 -tagged recombinant FHA1 domains were expressed in E. coli. Immunoblots of fractionated bacterial lysates (Fig. 5A) revealed that the solubility of these recombinant proteins mimicked their expression levels in vivo. Mutants that were expressed at low levels in yeast (G69A, H88A, G108A) were almost exclusively insoluble in vitro, supporting the notion that these domains are misfolded. In contrast, the wild-type FHA1 domain and the mutants with high expression levels in vivo (R70A, S85A, N107A) were predominantly soluble in vitro. Although the N112A mutant had an unperturbed phenotype in vivo, indicating that it is properly folded in yeast, a large fraction was insoluble when expressed in bacteria. However, the relative solubility of this domain was much greater than that of the G69A, H88A, and G108A mutants, and once purified its solubility did not differ from the recombinant wild-type FHA1 domain (data not shown).
Soluble FHA1 domains, as well as the Dun1-FHA domain, were purified (Fig. 5B) and tested in an ELISA for their ability to bind to a phosphothreonine-containing peptide, that has previously been shown to interact with an extended Rad53-FHA1 domain-containing GST fusion protein (residues 2-279) (29). In this assay (Fig. 5C), the wild-type FHA1 domain bound to the phosphopeptide with high affinity (half-maximal binding at 6 g/ml, ϳ350 nM). Although the dose-response curve of the N112A mutant, that had an unperturbed phenotype in vivo, was shifted to about 1 order of magnitude higher concentrations (EC 50 ϳ 66 g/ml) it still bound reasonably well with a profile similar to the Dun1-FHA domain (EC 50 ϳ 33 g/ml). In contrast, the mutants with an abrogated in vivo phenotype bound either very poorly (S85A, 35% maximal binding at 1 mg/ml) or not at all (R70A, N107A).
Taken together, these data indicate that the ability to bind phosphopeptide target sequences is a critical determinant of the growth defect caused by the FHA1 domain. However, there are clearly additional specificity requirements for the FHA1induced G 1 arrest, as the Dun1-FHA domain had an in vitro phosphopeptide-binding profile similar to the N112A mutant without causing the same phenotype.
Identification of Crucial Nonconserved Residues in an in Vivo Random Mutagenesis Screen-The strong growth-defect phenotype of the overexpressed FHA1 domain provided us with an opportunity to identify additional functionally important residues in a random mutagenesis screen. For this purpose, yeast were transformed with a library of randomly mutagenized FHA1 domains. Relative to the plating efficiency of the doublemutant (S85A/H88A) control, ϳ8% of the FHA1 sequences in the library contained mutations that abrogated the phenotype and gave rise to viable colonies (data not shown). 264 viable clones were counterscreened by immunoblotting to select only full-length FHA1 domains for sequence analysis. A representative blot is shown in Fig. 6A. Only 47 of the clones screened expressed the full-length FHA1 domain, and 45 of these contained single point mutations, while the majority contained truncated domains, presumably due to the introduction of stopcodons or frameshifts during the mutagenesis step. Altogether, 13 different mutations with amino acid substitutions in four conserved (G69S, R70K, N107I, N107K, N107S, G108A) and six nonconserved residues (C34R, V36D, L78S, S105C, S105T, L124P, L143P) were identified (Fig. 6B). These mutants differed somewhat in their expression levels, but with the exception of L124P all were expressed at considerably higher levels than the random G108A mutant (Fig. 6B) that was expressed at low levels similar to its counterpart generated by site-directed mutagenesis (Fig. 4B).
Increased DNA Damage Sensitivity of rad53 Alleles with Amino Acid Substitutions in the FHA1 Domain-To test if mutations that abrogate the FHA1 overexpression phenotype are relevant for Rad53 function at physiological protein levels, one of the site-directed and one of the random mutations were each introduced into the RAD53 gene using a modified allelereplacement procedure. To avoid secondary effects, the R70A and N107K mutations were selected for this purpose because their high overexpression protein levels (Figs. 4 and 6) indicated that they would not compromise Rad53 folding and stability. For comparison, we also generated a kinase-defective rad53-K227A allele (40) and a rad53⌬ mutant. All of these alleles were generated in a strain containing the sml1⌬ mutation as an extragenic suppressor of the essential viability function of RAD53.
All rad53 strains constructed were viable and, except for the deletion strain, had Rad53 protein levels similar to wild-type RAD53 (Fig. 7). To test if the FHA1 domain contributes to the DNA damage response function of Rad53, logarithmically growing cultures were treated for 3 h with MMS and then plated on YPD medium to determine survival rates. In these experiments, both FHA1 mutations had almost identical effects on cell viability, and relative to the RAD53 allele led to a 2.5-4-fold increased lethality after DNA damage (Fig. 7B). This confirms that these FHA1 residues are important for the function of Rad53. DISCUSSION Our results show that overexpression of the Rad53-FHA1 domain causes a dramatic growth defect (Fig. 1) by arresting yeast in the G 1 phase of the cell cycle (Fig. 3). This phenotype is specific for the FHA1 and not observed with three similar domains (Rad53-FHA2, Dun1-FHA, Chk2-FHA; Fig. 1) and mutational analyses indicate that it involves binding to a phosphorylated protein (Figs. 4 and 5). The FHA1-induced growth defect was also apparent in strains containing the sml1 deletion as a suppressor of the essential function of RAD53, as well as in a sml1-1 strain lacking RAD53 altogether, demonstrating that the overexpression phenotype is not dominant-negative or competitive with Rad53 (Fig. 2). One of the numerous checkpoint functions of the Rad53 kinase is to arrest the cell cycle prior to START in response to DNA damage during G 1 (20). Therefore, FHA1 domain overexpression seems to mimic this established checkpoint function of intact Rad53 even in the absence of DNA damage. Interestingly, the same FHA1 point mutations that abolish the overexpression phenotype (Figs. 4 and 6) also lead to an increased DNA damage sensitivity in mutant rad53 alleles (Fig. 7). The simplest explanation to combine our observations is that the FHA1 domain is involved in the G 1 checkpoint function of Rad53, and that the overexpression phenotype is caused by the binding of the FHA1 to a protein it usually only interacts with in response to DNA damage. In this model, the FHA1 domain normally interacts with other parts of the intact kinase but upon Rad53 activation in response to DNA damage becomes accessible and tethers a protein that is required for transition into S phase. This interaction is terminated when intact Rad53 phosphorylates this target, which maintains the G 1 arrest by a different mechanism and at the same time lowers its affinity for FHA1, and as a result the transient FHA1/target complex dissociates. In the ectopic overexpression scenario, the target remains trapped as it can not be phosphorylated by Rad53, and the continued sequestration of this protein causes an extended G 1 arrest.
The increased DNA damage sensitivity of the rad53 alleles with point mutations in the FHA1 domain (Fig. 7) clearly demonstrates that this domain contributes to the DNA damage response function of Rad53. However, whereas the rad53-R70A and rad53-N107K mutations reduced cell survival after DNA damage by up to 4-fold, DNA damage sensitivities of the "kinase-dead" rad53-K227A allele and the deletion mutant were an order of magnitude higher (Fig. 7). The similar DNA damage sensitivity of kinase-defective and deletion mutants indicates that the protein kinase catalytic domain is the most critical component of the Rad53 effector function. Rad53 has complex DNA damage regulated functions with multiple effectors in all phases of the cell cycle (16,17,20,24,25,41). Possible reasons for the lower DNA damage sensitivity of the rad53-FHA1 mutants relative to the kinase-dead Rad53 could be that this domain functions only in a specific phase of the cell cycle, or that it targets the kinase domain only to a specific subset of Rad53 effectors. This functional restriction would be similar to the Rad53-FHA2 domain. Deletions of the FHA2 do not increase lethality after UV-dependent DNA damage (42), but point mutations that abolish its interaction with Rad9 almost completely abrogate the G 2 /M cell cycle arrest after cdc13-dependent DNA damage (30).
Our overexpression results are somewhat similar to earlier reports showing that overexpression of intact Rad53 results in slower cell growth (11) by delaying G 1 -S transition by ϳ60 min (42) and, when fused to green fluorescent protein, overex- pressed Rad53 can severely impair the ability of yeast to form colonies on solid media (43). The green fluorescent protein-Rad53 phenotype can be suppressed by simultaneous overexpression of the Ptc2 protein phosphatase (43). However, simultaneous overexpression of Ptc2 had no effect on our FHA1 overexpression phenotype (data not shown), indicating that the FHA1 domain causes the G 1 arrest by a different mechanism. Our results are also consistent with a previous report that deletion of large parts of the FHA1 domain from Rad53 (rad53-⌬51-165 and rad53-⌬99 -161) results in increased UV sensitivity (42). In that case, FHA1-deleted rad53 mutants expressed from a centromeric plasmid in a rad53⌬ background failed to protect from UV-dependent lethality in contrast to wild-type and FHA2-deleted constructs. The more moderate DNA damage sensitivity in our approach could have two reasons. First, the DNA damage mechanism of UV irradiation (preferentially DNA adducts) is different from MMS treatment (preferentially double-strand breaks). Second, single amino acid substitutions in residues unlikely to be involved in domain folding may be less disruptive to the overall Rad53 protein structure than partial domain deletions (one of the FHA1 deletion mutants resulted in clearly reduced Rad53 protein levels; Ref. 42), resulting in more specific phenotypes.
The strong growth-defect phenotype of the FHA1 domain ( Fig. 1) makes this system an ideal model for the structure-function analysis of individual FHA residues in vivo, particularly as it is abrogated by the same mutations that increase the DNA damage sensitivity of rad53 alleles under physiological expression conditions (Fig. 7). Limited proteolysis experiments recently demonstrated that the FHA1 domain is contained within residues 24 -156 of Rad53 (28). This assignment is confirmed by results of the present study, showing first that truncation of the FHA1 domain to residues 40 -141 does not elicit the growth defect of the intact domain (Fig. 1A), and second that mutation of Cys-34 or Leu-143 in the random-mutagenesis screen also abolishes the phenotype (Fig. 6). This assignment is consistent with the recently solved NMR and x-ray crystallographic structures of the FHA1 domain (44,45).
Critical residues characterized in this study are highlighted in the apo-structure of the FHA1 domain (44) in Fig. 8. This model shows that the six conserved residues whose mutation abrogates the phenotype (shown in red) are clustered in close proximity. Interestingly, Gly-69, His-88, and Gly-108 are mostly buried with little surface exposure, indicating a role in domain folding rather than target binding. This is consistent with the low expression levels of the respective alanine mutants in vivo and their insolubility in a recombinant expression system in vitro. In contrast, Arg-70, Ser-85, and Asn-107, are  (44). Conserved residues are shown in red if their mutation abrogates the FHA1 phenotype or in yellow if their mutation does not affect the phenotype. Residues shown in green abolished the phenotype in the random mutagenesis screen. Views A, B, and C are clockwise 90°r otations around a longitudinal axis. surface-exposed in the cluster of conserved residues, consistent with roles in directly binding to targets. Among these residues, Arg-70 seems to be most crucial, as even the highly conservative R70K substitution identified in the random mutagenesis screen disrupts the growth defect phenotype (Fig. 6B). Asn-112 (shown in yellow) is unique among the conserved residues as its substitution by alanine does not disrupt the phenotype. This is reflected in the structure where it is not part of the cluster of conserved residues. Interestingly, large parts of this residue are buried underneath adjacent residues in the structure, which explains why the alanine substitution of this residue leads to a partial solubility problem when expressed in bacteria (Fig. 5A).
Functionally important nonconserved residues identified in the random mutagenesis screen are highlighted in green in Fig.  8. Cys-34, Val-36, Leu-124, and Leu-143 are buried in the core of the FHA1 domain, indicating that they are crucial for the proper domain fold. The other two critical nonconserved residues, Leu-78 and Ser-105, are very interesting as they directly interact with the cluster of conserved residues. Although the main chain residues of Leu-78 are surface-exposed and could interact with targets, the major function of its side chain is probably to correctly position the surface-exposed conserved residue Arg-70 that is crucial for target-binding (Fig. 5C) and with which it closely interacts. One of the phenotype-abrogating substitutions of Ser-105 (S105T) is highly conservative, underscoring the crucial role of this residue for the function of the FHA1 domain. In contrast to the conserved residues in the cluster that should have similar functions across all FHA domains, i.e. most likely direct interaction with the phosphate group in their targets, the nonconserved Ser-105 may therefore be a major secondary interaction site involved in target discrimination and binding specificity.
The FHA1 target that is involved in the G 1 arrest overexpression phenotype is presently unclear. Activated Cdc28/Cln cyclin-dependent kinase complexes play a major role in the transition from G 1 into S phase (46). Although simultaneous overexpression of the CLN1 or CLN2 G 1 cyclin genes does not overcome the FHA1-dependent arrest phenotype (data not shown), Cdc28 or its modulator Cks1 remain potential components of the pathway affected by FHA1 overexpression. Our future work will be directed at identifying proteins involved in the FHA1 overexpression phenotype and evaluating their role as effectors in the G 1 checkpoint function of Rad53.