Involvement of the Saccharomyces cerevisiae HDF1 gene in DNA double-strand break repair and recombination.

The HDF1 protein of Saccharomyces cerevisiae shares biochemical properties and structural homology with the 70-kDa subunit of the human autoantigen Ku. The Ku protein, a heterodimer composed of a 70-kDa subunit and an 80-kDa subunit, has been identified as the regulatory subunit of the DNA-dependent protein kinase. This enzyme has recently been shown to be involved in DNA repair and recombination processes in mammalian cells. Here we show that hdf1-disrupted S. cerevisiae strains are strongly sensitive toward the radiomimetic antibiotic bleomycin. In addition, mating-type switching and rates of spontaneous mitotic recombination are strongly reduced. This phenotype is similar to that of mammalian cells lacking components of the DNA-dependent protein kinase holoenzyme, suggesting that HDF1 participates in and exerts equivalent functions in S. cerevisiae.

DNA double-strand breaks (DSBs) 1 are intermediates of recombination events in both pro-and eukaryotic cells. DSBs can also be induced by treatments with ionizing radiation and with the radiomimetic antibiotic bleomycin (1)(2)(3). In Saccharomyces cerevisiae, DSB repair is mediated by activities of the RAD52 epistasis group (4), which includes gene products RAD50-RAD57. A prototype of RAD50-57 activities is the product of the RAD52 gene, which has been suggested to be required for almost all DSB events (5). Specifically, mating-type switching, which is initiated by a double-strand cleavage by the HO endonuclease at the MAT locus, is highly reduced in rad52 cells (6,7). rad52 strains fail to undergo proper meiotic chromosome segregation and therefore are defective in meiotic recombination (8 -10). Additionally, rad52 strains are defective in most mitotic recombination events (7,11).
Homologues of S. cerevisiae RAD51, RAD52, and RAD54 proteins have been identified in various higher eukaryotes (12)(13)(14)(15), suggesting that they play a similar functional role. Another activity involved in recombination events of higher eukaryotes has recently been attributed to the so-called Ku autoantigen. Ku is a heterodimeric protein composed of 70-and 80-kDa subunits that has recently been identified as the regulatory subunit of the DNA-dependent protein kinase (DNA-Pk) (16,17). Ku protein also binds to double-strand DNA ends, nicks, and hairpins (18 -25), suggesting that Ku may be in-volved in DNA repair, recombination, or replication events (21,22). This view is supported by the observation that the 80-kDa subunit of Ku is not detectable in x-ray-sensitive xrs hamster cell lines and that these cells fail to support normal V(D)J recombination (26,27). Both mutant phenotypes are complemented by the human XRCC5 gene, which has been shown to be identical with the gene for the Ku p80 subunit (28,29). Furthermore, the catalytic subunit of DNA-Pk, p350, is a strong candidate for the afflicted gene in cells derived from severe combined immunodeficient (SCID) mice (30,31). Like xrs cells, SCID cells are sensitive to ionizing radiation and defective in V(D)J recombination (32). The SCID defect appears to affect the rejoining of certain types of DNA strand breaks (33)(34)(35)(36), revealing a significant overlap between the mechanisms of V(D)J recombination and the joining of DSBs during DNA repair (37). V(D)J recombination is a specific form of DNA rearrangement that does not require extensive sequence homology. Therefore, the rejoining activity for DSBs that is deficient in the SCID and in xrs mutant cells appears distinct from the recombination repair activities in S. cerevisiae and hence might represent a specialized function of higher eukaryotes (37). On the other hand, DNA-Pk activities have been detected recently in cultured cells from mouse, hamsters, Xenopus, and Drosophila. Although these activities are approximately 10 -50 times less abundant than in human cells (38), their presence suggests that DNA-Pk is ubiquitous in eukaryotic cells and that it may have an evolutionarily conserved basal function in DSB repair and recombination.
The most compelling argument for the ubiquitous existence of DNA-Pk is the recent description of the Ku-like HDF1 protein in S. cerevisiae that binds to DNA ends and shares homology with the 70-kDa subunit of Ku (39). Here we show that disruption of the HDF1 gene in S. cerevisiae affects DSB repair, mating-type switching, and spontaneous mitotic recombination. The phenotype of hdf1 mutants is very similar to that of rodent xrs and SCID cells. This observation, together with the analogies in biochemical properties and the structural homology between HDF1 and Ku p70, suggests a functional conservation of these activities in eukaryotic organisms.

MATERIALS AND METHODS
S. cerevisiae Strains and Media-Strains used in this study are shown and characterized in Table I. Only relevant genotypes are listed. Isogenic W303rad52-4D was generated by crossing W303␣ and XS560-1C-1D1. ␣ segregants were tested for ura3-1 allele and then twice back-crossed with W303-1A, resulting in W303rad52-4D. For the characterization of hdf1 rad52 double mutants, W303rad52-4D was crossed with W303␣U and spore clones of two complete tetratype tetrades (W303LU-5 to W303LU-8 and W303LU-17 to W303LU-20) were used for the assays.
Complementation analysis of hdf1 mutants have been performed with the single copy vector pRS316 carrying the genomic XhoI-EcoRI fragment of the HDF1 gene. Mitotic recombination experiments were performed with strains or their derivatives generously provided by Robert Malone (7). Strains GM21 and GM28 were obtained by a sequence of crosses: K65-3D and K49-3A-7A were first crossed with IL993-5C (40) to remove the HO allele. The ho-negative segregants were then back-crossed three times with RM37-5D and RM38-7D, respectively, to develop isogenic strains. RAD52 segregants out of these crosses were identified by their ability to grow on YPD (Bacto yeast extract, 2% Bacto peptone, 2% dextrose (Difco)) plates containing 0.02% methyl methanesulfonate (MMS) (Sigma). Diploid GM2128 cells, resulting of crosses between GM21 and GM28, were used to determine wild-type mitotic recombination frequencies. Disruptions of the HDF1 gene in GM21 and GM28, followed by crosses of the hdf1-disrupted strains GM21␣U and GM28aU, resulted in diploid strains GM2128␣UaU. These were employed for the determination of spontaneous mitotic recombination frequencies in homozygous hdf1/hdf1 strains. Strains GM7726, GM2122, and GM5255 and hdf1/hdf1 derivatives thereof were generated in the same way. Standard genetic techniques were used throughout. hdf1-disrupted strains were constructed by lithium acetate transformation (41) with plasmid pS/H-URA3 (39) digested with EcoRI and HindIII.
Bleomycin Treatment-Stock solutions of BLM (Serva) were prepared with sterile water. For survival experiments, solid YEDA medium containing BLM was prepared according to methods described previously (43). Briefly, BLM was added to cooled YEDA immediately after pouring and mixed gently. Plates were used within 24 h after preparation. BLM concentrations varied between 1.88 and 15 g/ml. For a comparison of hdf1, rad6, and rad52 strains with the corresponding doubly mutated strains (hdf1rad6; hdf1rad52), bleomycin concentrations of 0.47-2.35 g/ml were employed. Different wild-type strains (W303-1A, K2346a) corresponding to temperature-sensitive hdf1-disrupted strains and hdf1-deficient strains transformed with the complete HDF1 gene on a single copy vector were treated and tested with BLM.
Survival Tests-Cultures were grown in YED medium at 23°C, and cell samples in the mid-log phase of growth (n ϭ 10 7 cells/ml) were collected. Samples were diluted in YED medium to give 200 -400 viable cells per plate and were promptly spread on YEDA in the presence and the absence of BLM. Plating was performed in quadruplicate for each BLM concentration. Plates were incubated at 23°C for 8 -10 days. The data from three to six experiments are given, and 95% binomial confidence limits are included in the figures.
MMS Treatment-Assays were performed analogous to BLM treatment on solid medium with and without MMS (Sigma) with concentra-tions ranging from 0.001 to 0.036% MMS.
Mating-type Switch Assay-Strains were transformed with YCp50GAL-HO and selected on SD medium lacking uracil. Single colonies were grown in synthetic media containing raffinose as a carbon source. Exponential cultures were inoculated in synthetic medium containing galactose or dextrose with 1000 cells/ml. Aliquots of the cultures were taken every hour; cells were spread on YPD (1 ϫ 10 2 ) and incubated at 30°C for 3 days. Because divergence in the portion of matingtypes at different time points can also be due to decelerated generation times and hence just reflect a decelerated pass through the G 1 phase of the cell cycle, the number of living cells were determined for each time point. The data were taken only from cultures with similar generation times. Colonies were replica-plated on synthetic medium without amino acids with KL14-4A and IL993-5C as mating testers. After 5 days of incubation at 30°C, prototrophic colonies were counted. In cases of nonor double-mating colonies, single colony cross-outs were performed and mating-type tests repeated. YCp50GAL-HO is a single copy shuttle plasmid containing a transcriptional fusion between the GAL10 promoter and HO gene kindly supplied by K. Luper and I. Herskowitz.
Determination of Mitotic Recombination Frequencies-Single colonies from a freshly constructed diploid were picked into 1 ml of YPD medium, and cell concentration was determined by a hemocytometer count. Approximately 500 cells/ml were inoculated into 5 ml of YPD medium, and the culture was grown at 23°C to a cell concentration of about 2 ϫ 10 7 cells. Each culture was started from an independent colony. Cells were washed twice in an equal volume of sterile water and plated at various dilutions on complete medium and medium lacking the supplement for the auxotrophic marker tested. Colonies were counted after 3 days of incubation at 30°C.

RESULTS
hdf1 Deficiency Causes an Increase in Bleomycin Sensitivity-One major effect caused by bleomycin is the introduction of DSBs into DNA molecules, and xrs and scid cells are known to be x-ray-sensitive as well as bleomycin-sensitive. 2 Accordingly, we determined the level of bleomycin sensitivity in hdf1-deficient yeast cells. Survival assays were carried out on solid media with and without BLM (43).
Although bleomycin affected survival rates in haploid wildtype strains only slightly, we observed a marked reduction by 4 orders of magnitude of survival rates in haploid hdf1 strains at a bleomycin concentration of 15 g/ml (Fig. 1A). The decline is not as drastic as that observed for rad52 strains, which display the same response already at a concentration of 3.75 g/ml, 2 S. Jackson, personal communication.
Ref. 40 suggesting that the RAD52 activity is a somewhat stronger factor for repair of bleomycin-induced DNA damage than the HDF1 function. Bleomycin sensitivity of hdf1-deficient strains can be restored to wild-type levels by the introduction of a functional HDF1 gene expressed from a single copy vector. The observed bleomycin sensitivity of hdf1-deficient cells thus must be caused by loss of HDF1 gene function. Comparable bleomycin sensitivities were observed for diploid hdf1/hdf1 strains, whereas heterozygous hdf1/HDF1 diploids and HDF1/HDF1 wild-type diploids displayed a resistant phenotype (Fig. 1B).
Because the hdf1 deficiency causes temperature-sensitive growth at 37°C, bleomycin sensitivity assays were each performed at 23 and 30°C. The results of these experiments demonstrate that bleomycin sensitivity was not temperature-dependent. No effects discriminating between hdf1-deficient and wild-type strains were observed when bleomycin was replaced by either streptomycin, tetracycline, erythromycin, ampicillin, or chloramphenicol at concentrations of up to 75 g/ml. This indicates the specificity of the bleomycin effect on hdf1-deficient strains (data not shown).
hdf1 rad52 Double Mutants Show Increased Sensitivity for Bleomycin Treatment-In view of the bleomycin sensitivity of hdf1-deficient strains, we investigated whether the function of HDF1 protein was related to other S. cerevisiae activities causing bleomycin sensitivity or acting on repair of DNA DSBs. Several genes of S. cerevisiae, collectively known as the RAD52 epistasis group, have been shown to affect DNA DSB repair and recombination (4). Mutations in these genes cause x-ray sensitivity and cross-sensitivity to bleomycin (44). Therefore, double mutants of hdf1 and the major representative of this group, rad52, were generated and their bleomycin sensitivity was compared with strains carrying the single mutations.
Surprisingly hdf1 rad52 double mutants are significantly more sensitive to bleomycin than the isogenic rad52 single mutated clones (Fig. 2). This effect of hdf1 and rad52 deficiency on bleomycin sensitivity indicates that both gene products are involved in repair of DNA DSBs in a cumulative manner.
This hypersensitive phenotype of hdf1 rad52 double mutants could also be observed with the radiomimetic agent MMS, which is known to induce strand breaks in DNA (4) (data not shown). In contrast, double mutants of hdf1 with either the rad3 or the rad6 mutations do not show hypersensitivity for bleomycin or UV irradiation (data not shown). This indicates that the observed bleomycin hypersensitivity of the hdf1 rad52 double mutant is specific for the rad52 mutation.
hdf1 Strains Show a Significant Reduction in Mating-type Switch Events-Apart from the sensitivity for ionizing radiation, another feature of higher eukaryotic cells lacking DNA-Pk activity is a defect in V(D)J recombination. This recombination event is initiated by a DSB. Similarly, the mating-type switch in S. cerevisiae is initiated by a double-strand specific cleavage introduced by the HO endonuclease at the MAT locus, resulting in gene conversion of the MAT allele (6). We therefore investigated the ability of hdf1-deficient strain W303aL MATa to undergo mating-type switching. We used a galactose-inducible HO endonuclease gene on a single copy vector and ascertained the percentage of mating-types at different time points.
A significant reduction of mating-type switching events to 30% or less in comparison with wild-type was observed in hdf1-deficient strains after 5 h of induction of the HO endonuclease gene (Fig. 3). This indicates that the HDF1 activity represents an important function for mating-type switching but may not be as important as the RAD52 gene product, because it has been shown by others that the rad52 mutation prevents homothallic switching altogether (7).
Mitotic Recombination Is Reduced by hdf1 Deficiency-Because the hdf1 mutation affects the rate of mating-type switching, which is a site-specific recombination event, we also studied the effect of the hdf1 mutation on spontaneous mitotic recombination. Although the proportion of spontaneous mitotic recombination events initiated by DSBs is not known, circumstantial evidence indicates a coupling between DSBs and mitotic recombination. Treatment of cells with x-rays, which can generate DNA DSBs, stimulates mitotic recombination more than 1000-fold (11). Furthermore, stimulation of mitotic (and meiotic) recombination by DSB was shown with systems based on the insertion of HO recognition sites into heterologous DNA and galactose-inducible expression of the HO endonuclease (45,46). The rad52 mutation known to be involved in DSB repair also causes a substantial reduction of spontaneous mitotic recombination (7). To study the influence of hdf1 mutation on FIG. 1. hdf1 deficiency results in an increased sensitivity toward bleomycin. A, haploid wild-type W303-1A (छ), isogenic hdf1deficient W303aL (Ç), and W303aL complemented with the wild-type HDF1 gene (Ⅺ) were plated on solid YEDA media containing bleomycin in concentrations of 1.88 -15 mg/ml. Surviving cells were grown up as colonies and counted after incubation at 23°C for 8 days. The data of rad52-deficient strain W303rad52-4D (E) are shown for comparison. B, increased bleomycin sensitivity is detectable also in homozygous hdf1/ hdf1. Diploid wild-type W303a␣ (f), heterozygous hdf1/HDF1 W303aLwt (q), and homozygous hdf1/hdf1 W303aL␣U (å) were assayed for bleomycin sensitivity as described above .   FIG. 2. Haploid hdf1 rad52 double mutants exhibit hypersensitivity for bleomycin. Spore clones of complete tetratype tetrades out of the cross of hdf1-deficient W303␣U with rad52-deficient W303rad52-4D were assayed for bleomycin sensitivity as described above. The data are shown for one tetrade only. Equivalent results were obtained for the others. Ⅺ, W3031u-17: HDF1 RAD52; छ, W3031U-18: hdf1 RAD52; Ç, W303Lu-19: HDF1 rad52; E, W303LU-20: hdf1 rad52. mitotic recombination, we used the same S. cerevisiae strains introduced by Malone and Esposito (7) to quantify the influence of the rad52 mutation on spontaneous mitotic recombination. The detection of mitotic recombination events is based on the use of diploids heteroallelic for several auxotrophic loci with the number of newly generated prototrophs allowing to monitor and measure recombination events. rad52/rad52 diploids were used as reference.
Our data show a 10 -40-fold reduction of spontaneous mitotic recombination events in hdf1/hdf1 diploids with respect to wild-type strains (Table II), whereas rad52/rad52 diploids showed reduction rates of 50 -200-fold. Comparable values were obtained for four independent crosses of wild-type strains and four independent crosses of hdf1 mutants. DISCUSSION This paper describes experiments that try to understand and explain the phenotype of the hdf1 mutation by studying its effects on bleomycin sensitivity, mating-type switching, and mitotic recombination frequencies. Bleomycin is generally regarded as a radiomimetic antibiotic because it induces a variety of but not all types of DNA damages known to be caused by ionizing radiation (44,47). Bleomycin is particularly well known for its ability to generate DNA DSBs, but there are also types of DNA damages thought to be specifically induced by bleomycin. For example, bleomycin produces 5Ј-phosphate and 3Ј-phosphoglycolate termini in DNA molecules, leaving a onenucleotide gap (48 -51). Using haploid and diploid radiosensitive strains, it was shown that all those single rad mutants that are sensitive to bleomycin are also sensitive to x-ray radiation, with the exception of rad15 (44). Furthermore, six of seven complementation groups of mutants (not blm2) directly selected for bleomycin sensitivity (blm mutants) are also sensitive to ionizing radiation (52). In conclusion, bleomycin sensitivity can be taken as an indicator for x-ray radiation sensitivity, although this may not be true for all mutations. In a strict sense, our data for bleomycin sensitivity of hdf1 mutants thus do not prove an involvement of HDF1 activity in DSB repair, although this appears very likely. Bleomycin sensitivity is less pronounced in hdf1 than in rad52 strains, whereas double mutants display a cumulative effect. This could be ex-plained if it was assumed that the HDF1 activity is involved in the repair of bleomycin-specific DNA damages while the RAD52 activity is responsible for DSB repair. The observed effect, however, could also be due to one singular DSB repair pathway if both activities affected different steps of this pathway. Although the rad52 mutation causes drastic effects on DNA repair, homothallic mating-type switching and mitotic recombination, there still remains some residual activity (7). In vivo, the RAD52 protein interacts physically with the RAD51 protein (53,54). The recA-like activity of RAD51 catalyzes homologous DNA pairing and strand exchange, which is ATPdependent (55). On the other hand, the functional defect in V(D)J-recombination in Ku/DNA-Pk-deficient cells seems to impair the rejoining of the free DNA ends of the coding strands (33)(34)(35)(36). If the HDF1 protein had a comparable function in S. cerevisiae, it could act downstream of the strand exchange activity of the RAD51/RAD52 complex. Successful DSB repair and recombination in rad52 mutants would then still require the downstream acting HDF1 activity.
To investigate further phenotypic similarities between hdf1 and DNA-Pk deficient cells, we studied the effects of hdf1 deficiency on mating-type switching and spontaneous mitotic recombination. V(D)J recombination is initiated by a precise double-strand cleavage between DNA segments and their flanking heptamers (56). Similarly, the mating-type switch is initiated by a DSB introduced by the HO endonuclease at the MAT locus (6). Both recombination events only require a limited number of base pairs for site specificity. V(D)J recombination results in the deletion of spacing sequences, whereas the mating-type switch results in a gene conversion event maintaining the donor sequence. We used an expression system with the HO endonuclease gene under the inducible GAL10 promoter to study the effects of hdf1 deficiency on mating-type switching rates. The significant reduction of mating-type switches in hdf1 mutants indicates that the HDF1 activity elicits a phenotype similar to the rad52 mutants. Taken in conjunction with data reported for bleomycin hypersensitivity, this further suggests that HDF1 activity may be involved in processing of DSB events.
DNA-Pk has been suggested to affect rejoining of broken DNA strands (33)(34)(35)(36). It seems reasonable to assume that during mating-type switching the HDF1 product might affect rejoining of donor and acceptor DNA after HO endonuclease cleavage and/or rejoining of DNA strands after resolution of the crossover fork, which occurs after strand exchange within the MAT locus. hdf1 deficiency should then also affect spontaneous mitotic recombination. The data presented in Table II

HDF1 Gene Involvement in DNA Break Repair and Recombination
strate that this is indeed the case. Therefore, we suggest that the HDF1 activity displays a very basic function in genetic recombination events.
The phenotype of hdf1-deficient strains strongly resembles that of rad52-deficient strains investigated so far, although the effect of hdf1 deficiency is not as pronounced with respect to bleomycin sensitivity, mating-type switching, and spontaneous mitotic recombination. It remains to be established whether this is a genuine feature of HDF1 or whether there is a second gene whose activities overlap with HDF function in S. cerevisiae. The fact that hdf1 and rad52 deficiencies differ phenotypically in respect to growth and viability parameters does not clarify the issue. hdf1 deficiency results in the inability to grow at 37°C, whereas members of RAD50-57 group, e.g. rad51 and rad54, show only a modest reduction in growth rate and no obvious reduction in vegetative viability at the higher temperature (39,37). Furthermore, hdf1/hdf1 diploids have no apparent defects in sporulation, whereas rad52/rad52 fail to undergo sporulation (8 -10). Further studies are therefore required to elucidate the relationship between HDF1 and RAD50-57 activities.
The discussion above has been focussed on the supposition that HDF1 has a "classical" repair function like other RAD activities of S. cerevisiae and thus led to our attempts to assign its function to known repair pathways. However, an emphasis on the homology of HDF1 protein to the Ku p70 product and hence on its function as a regulatory subunit of a putative DNA-Pk in S. cerevisiae leads to a broader, generic view of HDF1 function (57). Specifically, in higher eukaryotic cells DNA-Pk together with its regulatory subunit might act as a checkpoint modulator, because it is able to phosphorylate p53 (58). p53, in turn, is required for activation of G 1 checkpoint mechanisms that arrest cell cycle progression, presumably to allow time to repair DNA damage before DNA is replicated (59). Cell cyle progression from G 1 to S is inhibited when p53 levels are elevated. Exposures to UV light or ␥-irradiation causes a transient increase in p53 stability, which in turn increases its nuclear concentration. Finally, mutations in Ser 15 of human p53, which is phosphorylated by DNA-Pk in vitro, produces a remarkable increase in p53 half-life (60). By regarding HDF1 activity as one essential element of this damage recognition and controlling machinery in S. cerevisiae, the pleiotropic effects on repair, recombination, and viability functions become more comprehensible. Finally, when DNA-Pk is considered as just one element in a concerted interaction of DNA damage-induced kinases (57), the intimate relationship of HDF1 to DNA-Pk might also explain the reduced effects of hdf1 deficiency as compared with those observed upon loss of RAD52 function.
In conclusion, the reduced ability to perform mating-type switching and a reduction in spontaneous mitotic recombination events definitively prove that HDF1 is involved in recombination processes. The sensitivity of hdf1-deficient strains for the radiomimetic drug bleomycin indicates that the activity of HDF1 is not restricted to mating-type switching and spontaneous mitotic recombination. In view of the homology of HDF1 with the 70-kDa subunit of Ku protein of mammalian cells and its Ku-like activity, i.e. its ability to bind sequence-independently to ends of double-strand DNA, we suggest an analogous molecular function for HDF1 in the processing of DSB events in S. cerevisiae. The phenotypic characteristics of hdf1 deficiency in S. cerevisiae and of Ku/DNA-Pk deficiency in mammalian cells strongly support the notion that S. cerevisiae cells contain a DNA-Pk activity involved in functions similar to those in higher eukaryotes. We also suggest that DNA-Pk activity in higher eukaryotes may not be restricted to DSB repair and V(D)J recombination and may actually affect other recombinational processes such as gene conversion and mitotic recombination.