Repair of Endonuclease-Induced Double-Strand Breaks in Saccharomyces cerevisiae: Essential Role for Genes Associated with Nonhomologous End-Joining

Corresponding author: Michael A. Resnick, Laboratory of Molecular Genetics, National Institute of Environmental Health Sciences, 111 Alexander Dr., Research Triangle Park, NC 27709., resnick{at}niehs.nih.gov (E-mail)

Communicating editor: L. S. SYMINGTON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Repair of double-strand breaks (DSBs) in chromosomal DNA by nonhomologous end-joining (NHEJ) is not well characterized in the yeast Saccharomyces cerevisiae. Here we demonstrate that several genes associated with NHEJ perform essential functions in the repair of endonuclease-induced DSBs in vivo. Galactose-induced expression of EcoRI endonuclease in rad50, mre11, or xrs2 mutants, which are deficient in plasmid DSB end-joining and some forms of recombination, resulted in G2 arrest and rapid cell killing. Endonuclease synthesis also produced moderate cell killing in sir4 strains. In contrast, EcoRI caused prolonged cell-cycle arrest of recombination-defective rad51, rad52, rad54, rad55, and rad57 mutants, but cells remained viable. Cell-cycle progression was inhibited in excision repair-defective rad1 mutants, but not in rad2 cells, indicating a role for Rad1 processing of the DSB ends. Phenotypic responses of additional mutants, including exo1, srs2, rad5, and rdh54 strains, suggest roles in recombinational repair, but not in NHEJ. Interestingly, the rapid cell killing in haploid rad50 and mre11 strains was largely eliminated in diploids, suggesting that the cohesive-ended DSBs could be efficiently repaired by homologous recombination throughout the cell cycle in the diploid mutants. These results demonstrate essential but separable roles for NHEJ pathway genes in the repair of chromosomal DSBs that are structurally similar to those occurring during cellular development.

CHROMOSOMAL DNA double-strand breaks (DSBs) are formed after exposure of cells to various physical and chemical DNA-damaging agents such as ionizing radiation, bleomycin, and methylmethane sulfonate (MMS). DSBs are also produced endogenously through the action of intracellular enzymes and chemicals such as the highly reactive free radicals derived from oxygen metabolism. Recent studies have suggested that DSBs occur frequently in DNA containing specific at-risk sequence motifs (ARMs; e.g., in DNA containing trinucleotide repeats or large inverted repeats; NAG and KURST 1997 ; FREUDENREICH et al. 1998 ; GORDENIN and RESNICK 1998 ) and may also form after induction of replication arrest (SEIGNEUR et al. 1998 ). Most clastogenic agents (e.g., radiation, MMS, bleomycin, etc.) produce multiple types of DNA damage, but cause cell killing primarily through induction of unrepaired DSBs (RESNICK and MARTIN 1976 ; FRANKENBERG-SCHWAGER and FRANKENBERG 1990 ; OBE et al. 1992 ). For example, X-irradiation generates DNA-protein crosslinks, single-strand breaks, and damaged bases and sugars, but only DSBs are strongly correlated with loss of cell viability. DSBs are also generated during cell development. Such enzymatically induced DSBs typically retain undamaged complementary overhangs (e.g., at the ends of DSBs associated with meiotic recombination, mating type switching, and intron homing in yeast; HABER 1992 ; MUELLER et al. 1993 ; LIU et al. 1995 ; BELFORT and ROBERTS 1997 ). More complex termini have been observed at the ends of the defined DSBs that initiate V(D)J recombination in immunoglobulin and T-cell receptor genes during differentiation of cellular immunity in humans (JEGGO et al. 1995 ; RAMSDEN and GELLERT 1995 ; CHU 1997 ). Despite differences in structure and cleavage mechanism, all types of DSBs appear to exert similar effects on chromosome and cellular metabolism, including reduction of cell viability and induction of mutation, recombination, and aneuploidy (OBE et al. 1992 ; BRYANT and JOHNSTON 1993 ; STRATHERN et al. 1995 ; LEWIS et al. 1998 ). In addition, DSBs have been shown to arrest growth at DNA damage-responsive cell-cycle checkpoints and induce transformation of human and animal cells (WEINERT and HARTWELL 1988 ; OBE et al. 1992 ; NELSON and KASTAN 1994 ; SIEDE 1995 ).

The repair of DSBs is primarily a function of genes associated with two discrete pathways that are conserved in all eukaryotic organisms from yeast to humans. One pathway, involving repair by homologous recombination, was first described for the repair of radiation-induced DSBs over 20 years ago (RESNICK 1976 ; RESNICK and MARTIN 1976 ). This model for recombination in mitotic cells remains largely as originally presented (SZOSTAK et al. 1983 ; THALER and STAHL 1988 ; PETES et al. 1991 ), although additional classes of DSB repair events have been identified (KLEIN 1995 ; MALKOVA et al. 1996 ). The second major pathway of DSB repair, nonhomologous end-joining (NHEJ), provides for the direct rejoining of broken molecules and may be either precise or error prone (JEGGO et al. 1995 ; BOULTON and JACKSON 1996 ; MILNE et al. 1996 ; CHU 1997 ; HENDRICKSON 1997 ).

Genes specifically associated with recombinational repair of DSBs in mitotic yeast cells include RAD51, RAD52, RAD54, RAD55, RAD57, RAD59, and the RFA (single-stranded DNA binding protein) complex genes (HAYNES and KUNZ 1981 ; GAME 1993 ; FIRMENICH et al. 1995 ; HAYS et al. 1995 ; BAI and SYMINGTON 1996 ). rad52 mutants are profoundly deficient in recombination in many assays (RESNICK 1976 ; RATTRAY and SYMINGTON 1995 ; SUGAWARA et al. 1995 ; ZOU and ROTHSTEIN 1997 ), while mutations in RAD51 (a structural and functional homologue of the recA gene of Escherichia coli), RAD54, RAD55, RAD57, or RAD59 often produce less severe deficiencies. In addition, Rad51, Rad52, Rad54, Rad55, and Rad57 (involved in strand annealing and exchange) appear to form distinct protein:protein associations in vivo (BAI and SYMINGTON 1996 ; CLEVER et al. 1997 ; GOLUB et al. 1997 ; SUNG 1997A , SUNG 1997B ; BENSON et al. 1998 ; NEW et al. 1998 ; SHINOHARA and OGAWA 1998 ).

The efficiency of homologous recombination is influenced by several additional genes associated with DSB repair. For example, RAD50, MRE11, and XRS2 participate in NHEJ (described below) and are not required for mating type switching or interchromosomal recombination; however, each of these genes is required for resistance to ionizing radiation (SAEKI et al. 1980 ; HAYNES and KUNZ 1981 ; GAME 1993 ; HENDRICKSON 1997 ) and a functional RAD50 gene is necessary for efficient homologous intrachromosomal recombination in some assays (RATTRAY and SYMINGTON 1995 ; TRAN et al. 1995 ; ELIAS-ARNANZ et al. 1996 ). In addition, mutations within the genes encoding DNA helicases Hpr5(Srs2), Rad5, and Rdh54 (LAWRENCE 1994 ; MILNE et al. 1995 ; SCHILD 1995 ; CHANET et al. 1996 ; AHNE et al. 1997 ; KLEIN 1997 ; SHINOHARA et al. 1997 ), the exonuclease Exo1 (FIORENTINI et al. 1997 ), excision repair endonuclease Rad1/Rad10 (HABRAKEN et al. 1994 ; RODRIGUEZ et al. 1996 ; PAQUES and HABER 1997 ), mismatch repair proteins Msh2 and Msh3 (PAQUES and HABER 1997 ), or Cdc1 (HALBROOK and HOEKSTRA 1994 ) can exert strong effects on spontaneous and/or damage-induced recombination.

NHEJ repair has been detected in yeast by assessing the efficiency of recircularization of linear plasmid DNA after cellular transformation and through study of infrequent, homology-independent DSB repair events (SCHIESTL et al. 1994 ; BOULTON and JACKSON 1996 ; MILNE et al. 1996 ; MOORE and HABER 1996 ; TSUKAMOTO et al. 1997B ). Genes involved in recombination-independent NHEJ repair include HDF1(YKU70), YKU80(HDF2), RAD50, MRE11, XRS2, DNL4 (the yeast homologue of human DNA ligase IV), SIR2, SIR3, and SIR4 (BOULTON and JACKSON 1996 ; MILNE et al. 1996 ; HENDRICKSON 1997 ; SCHAR et al. 1997 ; TEO and JACKSON 1997 ; TSUKAMOTO et al. 1997A ; WILSON et al. 1997 ). The proteins encoded by most genes involved in NHEJ bind to the ends of linear double-stranded DNA. Past studies (including two-hybrid analyses and immunoprecipitation experiments) have identified Ku70:Ku80, Rad50:Mre11:Xrs2, and Sir2:Sir3:Sir4 complexes within cells (JOHZUKA and OGAWA 1995 ; DOLGANOV et al. 1996 ; CHU 1997 ; HENDRICKSON 1997 ; SHERMAN and PILLUS 1997 ; TSUKAMOTO et al. 1997A ). More recent reports have suggested that these complexes have different functions in NHEJ; e.g., Ku heterodimers are associated with bridging of the ends of broken DNA molecules while Rad50 and Mre11 appear to be involved in nucleolytic processing of DNA ends (SHARPLES and LEACH 1995 ; MOORE and HABER 1996 ; CARY et al. 1997 ; HENDRICKSON 1997 ; PANG et al. 1997 ).

Most yeast genes primarily involved in recombination (RAD51, RAD52, RAD54, RAD55, RAD57, RAD59) or NHEJ (HDF1, YKU80, RAD50, MRE11, SIR2, DNL4) have structural homologues in human cells. Protein:protein interactions have also been conserved, e.g., Rad51:Rad52, Rad51:Rad54, Ku70:Ku80, and Rad50:Mre11 associations have been identified in both yeast and human cells (JOHZUKA and OGAWA 1995 ; SHEN et al. 1996A ; CLEVER et al. 1997 ; GOLUB et al. 1997 ; HENDRICKSON 1997 ). Conservation of the function(s) of members of each pathway is suggested by biochemical and genetic studies and by reports that expression of human Ku70 and Rad54 can complement repair defects in yeast hdf1(yku70) and rad54 mutants, respectively (e.g., KANAAR et al. 1996 ; BARNES and RIO 1997 ; BEZZUBOVA et al. 1997 ). In addition to their roles in DSB repair, genes involved in recombination and NHEJ have also been implicated in normal cell development and in processes leading to human pathology. For example, human Ku70 and Ku80, in conjunction with an associated DNA-dependent protein kinase (DNA-PK), are required for V(D)J recombination events that occur in immunoglobulin and T-cell receptor genes during development of immunity (JEGGO et al. 1995 ; CHU 1997 ). In addition, deletion of the mouse RAD51 recombinase gene produces an embryonic lethal phenotype (LIM and HASTY 1996 ; TSUZUKI et al. 1996 ). Furthermore, recent experiments have established that human Rad51 interacts with the tumor suppressor proteins encoded by p53, BRCA1, and BRCA2 in vivo (SHEN et al. 1996A , SHEN et al. 1996B ; STURZBECHER et al. 1996 ; BUCHHOP et al. 1997 ; MIZUTA et al. 1997 ) and that sequence polymorphisms in human KU80 (XRCC5) and XRCC3 (a RAD51 homologue) are associated with tumorigenesis (PRICE et al. 1997 ).

Because traditionally employed clastogens (X rays, MMS, bleomycin, etc.) produce a variety of chromosomal DNA lesions in addition to DSBs, systems for studying the genetic and cytological consequences of expression of site-specific endonucleases (which produce complementary or blunt-ended DSBs only) have been developed in yeast and mammalian cells (e.g., OBE et al. 1992 ; BRYANT and JOHNSTON 1993 ; GODWIN et al. 1994 ; BENNETT et al. 1996 ; SARGENT et al. 1997 ; LIANG et al. 1998 ). Studies in yeast have analyzed the repair of DSBs induced by HO, EcoRI, and I-SceI (BARNES and RINE 1985 ; HABER 1992 ; PLESSIS et al. 1992 ; BARNES and RIO 1997 ; and see references in LEWIS et al. 1998 ).

We have recently employed yeast strains that permit modulation of GAL promoter induction kinetics to show that EcoRI-induced DSBs arrest cell growth at the G2/M transition and stimulate interchromosomal recombination, but do not cause more cell killing in rad52 mutants than in wild-type cells (LEWIS et al. 1998 ). However, expression of EcoRI was lethal in hdf1 and rad9 mutants, suggesting separable roles for end-joining, RAD52-mediated recombination, and DNA damage-responsive checkpoint pathways in the repair of endonuclease-induced DSBs.

In this study we demonstrate that many genes associated with NHEJ are essential for repair of chromosomal DSBs in vivo. Although several recombinational repair genes were required for progression of cells past the G2 checkpoint, only NHEJ genes were required for survival after EcoRI-induced cleavage of chromosomal DNA. In addition, analysis of cellular responses suggested that the DSB repair functions of several end-joining pathway genes, e.g., HDF1/YKU80 vs. RAD50/MRE11/XRS2 or SIR2/SIR3/SIR4, are genetically separable. The complementary-ended DSBs produced by the endonuclease are structurally similar to DSBs generated during normal cell development (i.e., during meiosis, mating type switching, and site-specific homing of intron DNA), which suggests that such DSBs might also be repairable by both recombination and NHEJ pathways.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Genetic methods and media:
Yeast growth media, including YPD, YPG, synthetic dropout, and sporulation plates were prepared as described (SHERMAN 1991 ). Transformation of yeast cells was accomplished using a modification of the lithium acetate procedure (GIETZ et al. 1995 ). For selection for G418 resistance, YPD plates containing 0.2 mg/ml G418 (Life Technologies) were prepared as described (WACH et al. 1994 ). 5-fluoroorotic acid (5-FOA)-resistant colonies were isolated using synthetic plate media containing 0.1% 5-FOA (American Biorganics, Niagara Falls, NY). All strains were propagated and assayed at 30° except hdf1 mutants, which were propagated at 25° and assayed for DNA repair at 30°. A ¹³⁷cesium source was employed at a dose rate of 2.75 krads/min for analyses of cellular growth responses to ionizing radiation.

Strain construction:
Yeast strains constructed for this study are listed in Table 1. All strains are derivatives of the A364-based strain 334 (HOVLAND et al. 1989 ). This strain contains the mutant allele reg1-501. GAL1 and GAL10 promoter activity in reg1-501 cells is low in glucose media, but addition of galactose induces the GAL1/10 promoters while cells continue using glucose as carbon source. We note that although GAL transcription is apparently quite low in reg1-501 strains in glucose media (HOVLAND et al. 1989 ; NIEDERACHER and ENTIAN 1991 ; LEWIS et al. 1998 ), basal levels of expression in reg1 host strains remain to be established. Strain 334 (and T334, the trp1 derivative used here) also contains a mutation in the GAL1 gene that blocks metabolism of exogenous galactose. Plasmids used for gene disruption included p52LEU (rad52::LEU2) and prad1Blast (rad1::hisG-URA3-hisG) obtained from E. Perkins, pNKY83 (rad50::hisG-URA3-hisG) obtained from N. Kleckner, pSL101 (rad54::hisG-URA3-hisG) and pSTL11 (rad55::LEU2) from L. Symington, pAM50 (rad51::LEU2) and pSM51 (rad57::LEU2) from D. Schild, and p21 (rad6::hisG-URA3-hisG) and pDM610.23 (sir4::LEU2) obtained from C. Bennett.

Deletions of rad2, rad5, rad27, rdh54, exo1, hpr5(srs2), xrs2, and mre11 were accomplished using PCR-mediated disruption as described (BAUDIN et al. 1993 ; WACH et al. 1994 ; LEWIS et al. 1998 ). rad50 rad52 and xrs2 rad52 double mutants were created by deleting RAD50 or XRS2 in YLKL351 (rad52::LEU2, his3::GAL1::EcoRI) cells containing the plasmid YcpRAD52 (RAD52, URA3). Gene disruption/deletions were confirmed by PCR analysis of genomic DNA and/or by genetic crosses except for rad1 and rad2 strains, which were identified by their sensitivity to ultraviolet light. Two or three independent isolates of each gene disruption were analyzed in the experiments described below. Nucleotide sequences of primers used for deletion and PCR confirmation are available upon request.

Endonuclease-induced changes in cell growth and viability:
Growth of cell cultures was monitored by counting cells using a hemacytometer. Cell survival after induction of endonuclease expression in galactose was calculated as the number of viable cells per milliliter observed on YPD plates divided by the number of cells per milliliter in the culture determined by hemacytometer. YPD broth containing 3% glucose was used for overnight cultures and for control time-course experiments. Earlier studies using the reg1-501 strain employed YPD containing 2% glucose (HOVLAND et al. 1989 ; LEWIS et al. 1998 ), but this was modified after determination that prolonged growth of rad9 strains containing integrated GAL1::EcoRI in standard YPD media (for 12–24 hr) led to glucose depletion and apparent modest induction of the GAL1 promoter (LEWIS et al. 1998 ). Inducing media consisted of YPD broth with 3% glucose and 2% galactose (YPD + Gal). Haploid cell assays were performed by shifting logarithmically growing cell cultures from YPD to YPD + Gal media at a density of 2–4 x 10⁵ cells per milliliter. For the analysis of wild-type and mutant diploid cells depicted in Table 2, eight independent cultures were used to calculate cell viabilities and standard deviations. Standard deviations were 15% of the mean for all other cell survival assays. Diploid strains used for the assays in Table 2 included YLKL350 x YLKL313 (Rad⁺), YLKL372 x YLKL397 (rad50), YLKL381 x YLKL439 (rad51), YLKL351 x YLKL423 (rad52), and YLKL407 x YLKL433 (mre11). Each of the diploid strains contains a single integrated his3::GAL1::EcoRI fusion.

Distribution of cells in G1, S, and G2/M:
Cell-cycle progression was monitored as described (LEWIS et al. 1998 ). Large-budded cells were defined as cells in which the bud was >50% of the size of the mother cell. For analysis of unbudded, small-budded, and large-budded cells after 12 hr of induction in galactose media, 300–400 cells in each culture were counted.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Survival of cells expressing EcoRI requires genes associated with NHEJ, but not recombinational repair:
We previously demonstrated that expression of EcoRI endonuclease inhibits growth and induces G2 arrest in rad52 strains, but does not produce significantly more killing than in wild-type cells (LEWIS et al. 1998 ). This work, like the previous study, has employed the reg1-501 strain T334 (HOVLAND et al. 1989 ; NIEDERACHER and ENTIAN 1991 ). GAL1-GAL10 promoter activity is repressed in glucose media in this strain, but is induced in media containing glucose plus galactose. The ability to induce GAL promoter activity without changing carbon source is advantageous because cell growth rates and the relative lengths of each cell-cycle phase are different in cells metabolizing glucose vs. galactose (BARFORD and HALL 1976 ).

The observation that RAD52 was required for progression of cells past the checkpoint at G2, but not for survival, indicated that repair of EcoRI-induced DSBs involves additional genes. The effects of EcoRI expression on cell viability in strains containing deletions in several additional RAD52 group genes are described in Figure 1A–C. All cells used in the experiments contain a chromosomal his3::GAL1::EcoRI fusion. Assays were performed using logarithmically growing cells as previously described (LEWIS et al. 1998 ) with slight modification (see MATERIALS AND METHODS). The rad52 strain YLKL351 (LEWIS et al. 1998 ) is included in Figure 1 for comparison. Strains containing deletions in genes specifically associated with strand annealing and exchange in mitotic cells (RAD51, RAD52, RAD54, RAD55, and RAD57) displayed high survival throughout the time course (Figure 1A).

View larger version (15K): In this window In a new window Download PPT slide   Figure 1. Effects of endonuclease-induced DSBs on survival of wild-type cells and strains deficient in homologous recombination and/or end-joining pathways of repair. (A and B) Logarithmically growing cells were transferred to YPD + galactose media and cell viability was assessed at the indicated times after induction of EcoRI expression. (C) Survival of end-joining-defective hdf1 and sir4 strains propagated in noninducing (Glu) and galactose-containing media (Glu + Gal). Standard deviation (SD) values were 15% of the mean for all cell survival assays for which SD is not shown (see MATERIALS AND METHODS).

rad50, mre11, and xrs2 mutants previously have been shown to be deficient in NHEJ repair using plasmid recircularization assays, but are not defective in homologous interchromosomal recombination or mating type switching (gene conversion) in mitotic cells (BOULTON and JACKSON 1996 ; MILNE et al. 1996 ; MOORE and HABER 1996 ; HENDRICKSON 1997 ). As shown in Figure 1B, continuous EcoRI-induced scission of DNA in these mutants caused extensive cell killing. Plating efficiency was reduced to 2–3% after 12 hr of endonuclease expression.

The SIR2, SIR3, and SIR4 genes, which are required for transcriptional silencing at MAT and in the telomeric regions of yeast chromosomes, are also involved in plasmid DSB repair (TSUKAMOTO et al. 1997A ). The ability of cells containing a deletion of SIR4 to repair chromosomal DSBs induced by EcoRI is depicted in Figure 1C. The mean surviving fraction was reduced to 8% in sir4 mutants, indicating that repair of the cohesive-ended DSBs is impaired in these strains. Also presented in the figure is a comparative analysis of cell survival and cell cycling in the hdf1 strain YLKL389. As noted earlier, we previously observed (using slightly different media; see MATERIALS AND METHODS) that expression of EcoRI is lethal in hdf1 cells in the T334 strain background. The kinetics of cell killing in Ku-deficient cells was similar to that of rad50, mre11, and xrs2 mutants and plating efficiency was reduced 30-fold after 12 hr (Figure 1C).

The effects of EcoRI-induced breakage of chromosomal DNA on cell-cycle progression are shown in Figure 2A–C. Recombination-defective mutants arrested growth as large-budded cells (Figure 2A), indicating that they were unable to progress past the G2/M boundary when EcoRI was continuously expressed. Most arrested cells were much larger in size than normal G2 cells. We previously demonstrated that this DSB-induced arrest corresponds to a RAD9-dependent, DNA damage-responsive checkpoint that is phenotypically similar to the damage-induced arrest described by WEINERT and HARTWELL 1988 . Wild-type cultures displayed a modest increase in large-budded cells during the time course. This prolongation of the G2 phase is transient and wild-type cells advance to early stationary phase at 24 and 48 hr (LEWIS et al. 1998 ; data not shown). All of the strains studied in Figure 1 and Figure 2 exhibited high survival and normal cell cycling when grown in YPD (noninducing) media (Figure 1C and data not shown).

View larger version (18K): In this window In a new window Download PPT slide   Figure 2. Analysis of cell-cycle progression after induction of EcoRI expression in wild-type cells and in mutants deficient in recombination and/or NHEJ. Cells were transferred to galactose-containing media and analyzed microscopically at the indicated times after induction. Most large-budded cells observed in endonuclease-producing strains corresponded to enlarged G2 phase cells (see text).

Interestingly, the growth of rad50, mre11, and xrs2 mutants was blocked, with most cells arrested at the G2 checkpoint (Figure 2B; ~70% large-budded cells after 12 hr), but this pause was not sufficient for cells to repair the induced strand breaks after synthesis of the endonuclease was repressed. Most sir4 cells eventually arrested growth in G2 phase, but hdf1 cells did not (Figure 2C). Growth of Ku-defective cells was blocked, but these cultures consisted primarily of a mixture of enlarged unbudded and large-budded cells, suggesting that arrest occurred in both G1 and G2 (discussed below).

Requirement for RAD3 and RAD6 group genes after EcoRI-induced scission of chromosomal DNA:
The potential involvement of RAD1 and RAD2 (nucleotide excision repair genes in the RAD3 epistasis group; HAYNES and KUNZ 1981 ; HABRAKEN et al. 1994 ) and RAD5 and RAD6 (RAD6 epistasis group genes involved in postreplication repair and mutagenesis; HAYNES and KUNZ 1981 ; LAWRENCE 1994 ) in the repair of EcoRI-induced DSBs has also been examined (Figure 3, A–D). rad5 and rad6 mutants are moderately sensitive to ionizing radiation and ultraviolet light, but their functions in the repair of single- or double-strand breaks remains obscure. Survival curves for rad5 and rad6 strains were comparable to those of wild-type cells (Figure 3A). However, growth of rad5 cells was strongly inhibited and these mutants displayed a strong cell-cycle arrest response that was similar to that of recombination-deficient RAD52 group mutants (Figure 3B). This result suggests a role for RAD5 in the recombinational repair pathway and is consistent with a recent report indicating that recombinational repair of gapped plasmid DNA is greatly reduced in cells lacking the Rad5 DNA helicase (AHNE et al. 1997 ). The response of rad6 cells was complex. These mutants grew slowly when EcoRI was expressed and accumulated enlarged G2 phase cells only at late time points (12 and 24 hr; Figure 3B and data not shown). These results suggest that repair of the induced DSBs was relatively efficient at earlier time points (4 and 8 hr) when fewer breaks are detectable by gel analysis (LEWIS et al. 1998 and data not shown), but that a subset of DSBs required processing by Rad6 at later time points.

View larger version (30K): In this window In a new window Download PPT slide   Figure 3. Effects of endonuclease-induced DSBs on survival and cell cycling in mutants deficient in nucleotide excision repair (RAD3 group) or mutagenesis and postreplication repair (RAD6 group). Quantitation of cell viability and formation of large-budded cells (mostly G2 cells) was performed as for Figure 1 and Figure 2.

The excision repair endonuclease encoded by RAD1/RAD10 plays a role in cleavage of recombination intermediates and in the processing of nonhomologous DSB ends (HABRAKEN et al. 1994 ; RODRIGUEZ et al. 1996 ; PAQUES and HABER 1997 ). Furthermore, 5' single-stranded overhangs similar to those produced by EcoRI have previously been shown to be substrates for cleavage by the Rad2 endonuclease in vitro (HABRAKEN et al. 1995 ). The possibility that Rad1/Rad10 or Rad2 might be involved in processing of the ends of EcoRI-induced DSBs has been assessed in Figure 3C and Figure D. Survival of rad2 mutants was not affected by EcoRI, but rad1 strains displayed a modest, linear decrease in viability during the time course (Figure 3C). Cell-cycle progression in rad2 mutants was essentially identical to that of wild-type cells. In contrast, rad1 mutants were moderately growth-inhibited and slowly accumulated a high proportion of G2 cells (70%) at 12 and 24 hr after induction (Figure 3D and data not shown). These results suggest that a subset of DSBs produced by EcoRI are processed by the Rad1/Rad10 endonuclease, but not by Rad2.

To test the possibility that RAD1/RAD10 or RAD2 nuclease processing of cohesive-ended DSBs becomes more significant in the absence of other repair pathways, i.e., recombination or end-joining, cell survival was also monitored in double mutants. The viability of rad1 rad52 and rad2 rad52 cells was reduced ~7-fold during the time course (Figure 3C). This effect was slightly greater than that seen in rad1 and rad2 single mutants, but cell killing did not approach levels observed in hdf1, rad50, mre11, or xrs2 mutants (~20–40-fold killing after 12 hr). The kinetics and magnitude of growth arrest as large-budded cells in each double mutant cell culture appeared similar to that of rad52 cells (Figure 3D).

The data presented in Figure 1 Figure 2 Figure 3 and in past studies (see LEWIS et al. 1998 and references within) support the hypothesis that DSB termini containing single-stranded complementary overhangs can be repaired by multiple pathways in vivo. In contrast, repair of damaged ends produced by ionizing radiation is largely dependent on recombinational repair. The negative effects of EcoRI- and radiation-induced DSBs in wild-type and repair-deficient cells have been compared directly in Figure 4. Growth of cells containing a deletion of RAD52 (recombination deficient), RAD50 or MRE11 (partially defective in recombination and NHEJ), and SIR4 (NHEJ deficient) is inhibited by EcoRI. However, gamma-irradiation (30 krads) blocked growth of the Rec^- strains (rad52, rad50, mre11), but not sir4 cells (Figure 4). Growth of excision-repair-defective rad1 and rad2 strains on plates was largely unaffected by either clastogen.

View larger version (96K): In this window In a new window Download PPT slide   Figure 4. Growth-inhibitory effects of DNA damage produced by EcoRI vs. ionizing radiation in wild-type, Rec- (rad52), NHEJ-/partial Rec- (rad50, mre11), NHEJ- (sir4), and excision repair-deficient (rad1, rad2) strains. T334 cells (his3::GAL1::EcoRI) were replica plated to media containing (A) YPD (B) YPD + 2% galactose, or (C) YPD exposed to gamma irradiation (30 krads).

Viability and checkpoint responses in end-joining and recombination-defective double mutant strains:
The experiments described above indicated that cells experiencing EcoRI-induced DSBs require several recombinational repair genes for progression past the G2 checkpoint, but not for survival. In contrast, Ku-deficient strains and RAD52 group mutants involved in end-joining repair (rad50, xrs2, and mre11) displayed extensive cell killing, but had dissimilar arrest phenotypes (compare Figure 2B vs. Figure 2C). The effects of EcoRI-induced cleavage of DNA in strains deficient in both end-joining and recombinational repair are presented in Figure 5. Surprisingly, growth arrest and cell killing kinetics in rad50 rad52 and xrs2 rad52 double mutants were similar to those of rad50 and xrs2 single mutants, respectively (~2–3% viable cells after 12 hr; Figure 5A and Figure 1B and Figure 2B). Thus, combination of a deficiency in NHEJ with elimination of almost all homologous recombination did not produce additional killing in this assay.

View larger version (34K): In this window In a new window Download PPT slide   Figure 5. Responses of single and double mutant strains deficient in homologous recombination and end-joining pathways to the induction of DSBs. (A and B) Effects of endonuclease expression on survival and cell-cycle progression in end-joining-defective double mutant cells (rad50 hdf1) or in strains deficient in both recombination and end-joining (rad50 rad52, xrs2 rad52). (C and D) Effects of deletion of cellular nuclease genes (RAD1, RAD2, or EXO1) on endonuclease-induced killing of end-joining-deficient rad50 mutants.

Interestingly, rad51 hdf1 and rad52 hdf1 double mutants could be constructed in the T334-derived strains used here, but the cells grew poorly and displayed several stress phenotypes in both glucose and galactose media. Characteristics included slow growth, reduced plating efficiency, elevated levels of petite mutants, and an increased fraction of dark (presumably lysed) cells upon microscopic examination (data not shown). These effects were not observed in any other single or double mutant created for these studies and their source remains unclear. We note, however, that expression of EcoRI in wild-type, rad52, or hdf1 strains does not produce elevated levels of petite mutants (BARNES and RINE 1985 and data not shown), suggesting that the phenomena are not due to altered GAL1::EcoRI expression. Furthermore, rad52 hdf1 double mutants have previously been reported to have significantly lower plating efficiency than either single mutant (SIEDE et al. 1996 ). Finally, past reports have demonstrated that DNA damage-inducible genes are constitutively activated in rad52 and hdf1 null mutants, presumably because of elevated levels of unrepaired DNA damage (MAGA et al. 1986 ; BARNES and RIO 1997 ). This effect may be exacerbated in hdf1 rad52 double mutants, which are defective in both major pathways of DSB repair. The results presented in Figure 5A also demonstrated extensive EcoRI-induced cell killing in rad50 hdf1 mutants that was similar in extent to that observed in rad50 and hdf1 single mutants (~30-fold killing). rad50 hdf1 cell cultures consisted primarily of unbudded and large-budded cells, suggesting that checkpoint activation occurred in both G1 and G2, as observed earlier in hdf1 cultures (Figure 1C and see below).

Past studies have suggested that some or all DNA end-binding proteins that function in NHEJ repair might affect degradation of DSB termini (e.g., GETTS and STAMATO 1994 ; MILNE et al. 1996 ; MOORE and HABER 1996 ). The absence of one or more such proteins might lead to cell killing because the DNA ends become susceptible to cleavage by specific cellular endo- or exonucleases. As an initial test of this hypothesis we constructed double mutant strains that are deficient in Rad50-mediated end-joining and that contain a deletion of an endonuclease or exonuclease gene previously implicated in DSB repair. Changes in cell viability and cell cycling in rad1 rad50, rad2 rad50, and exo1 rad50 mutants upon induction of EcoRI are presented in Figure 5C and Figure D. EXO1 was included because this exonuclease has been shown to be involved in the processing of DNA ends and in intrachromosomal recombination (FIORENTINI et al. 1997 ). Cell survival was reduced 20- to 40-fold in each double mutant, which is comparable to that seen in rad50 strains (Figure 5C). All double mutants displayed a strong growth arrest response that was analogous to that observed previously in rad50 single mutants. These data suggest that aberrant processing of the ends of DSBs by Rad1/Rad10, Rad2, or Exo1 is not responsible for the killing observed in end-joining defective rad50 mutants.

Distinct cell cycle checkpoint responses in recombination- and NHEJ-deficient mutants:
A detailed examination of the distribution of wild-type and mutant cells within the cell cycle in response to EcoRI-induced scission of chromosomal DNA is presented in Figure 6. The proportion of unbudded (G1 phase), small-budded (an approximation of cells undergoing S phase), and large-budded cells (G2/M phase) was quantitated after 12 hr of EcoRI expression. Cultures of all RAD52 group mutants tested, including recombination-defective rad52 strains and end-joining-deficient rad50 cells, consisted primarily of S and G2 phase cells with most cells enlarged and arrested in G2 (Figure 6 and data not shown). rad50 rad52 double mutants (severely deficient in both recombination and NHEJ) also arrested predominantly at G2 with almost all cells in S or G2. Cultures of hdf1 and hdf1 rad50 mutants consisted of approximately equal numbers of G1 and G2 cells (Figure 6). Thus, the unusual biphasic DNA damage response of Ku-defective strains is dominant to that of rad50 mutants (and by extension to mre11 and xrs2 mutants).

View larger version (37K): In this window In a new window Download PPT slide   Figure 6. Changes in the distribution of G1, S, and G2/M cells in wild-type strains and in single and double mutants after EcoRI-induced cleavage of chromosomal DNA. Cells were analyzed at t = 12 hr (see MATERIALS AND METHODS).

Effects of EcoRI expression on cell viability and checkpoint responses in wild-type and repair-deficient diploid cells:
Although both haploid and homozygous diploid RAD52 group mutants are hypersensitive to ionizing radiation-induced DSBs, differences in resistance due to ploidy have been observed. For example, diploid rad50 strains are slightly more radio resistant than their haploid counterparts (presumably due to increased recombinational repair of some induced DSBs), but rad51 and rad52 diploids are not (SAEKI et al. 1980 ; GAME 1993 ). The possibility that EcoRI-induced growth inhibition and/or cell killing observed in repair-deficient haploid strains might be altered in diploids (which can undergo recombinational repair throughout the cell cycle) was tested in wild-type cells and in homozygous rad50, rad51, rad52, and mre11 mutants. Analysis of the growth characteristics of diploid cells containing a single GAL1::EcoRI chromosomal fusion revealed significant differences from haploid cells (Figure 7). Wild-type diploid cells containing either his3::GAL1::EcoRI or lys2::GAL1::EcoRI fusions (LEWIS et al. 1998 and data not shown) did not exhibit a transient G2 arrest response as previously observed in haploid cells. This result suggests that EcoRI-induced DSBs in the diploids were repaired efficiently and did not activate the checkpoint response system. All homozygous mutant diploids arrested growth in G2 phase after induction of EcoRI expression (70–80% of cells after 8–12 hr) with most of the G2 cells greatly enlarged.

View larger version (53K): In this window In a new window Download PPT slide   Figure 7. Analysis of endonuclease-induced effects on cell cycling in diploid cells. Recombination-deficient (rad51 or rad52) and end-joining-defective mutants (rad50 or mre11) displayed a strong G2 arrest response, but the transient G2 arrest observed in Rad+ haploid cells was abolished in repair-proficient diploids.

Interestingly, uninduced, logarithmically growing rad50, rad51, rad52, and mre11 diploid cells in glucose media contained ~50% large-budded cells. This is apparent at the zero time point in Figure 7 and in control time-course experiments in which cells were grown in standard glucose media (data not shown). Haploid RAD52 group mutants displayed only a slight increase in G2 cells in glucose media. A previous study has reported that logarithmically growing rad51 diploid cell cultures, but not rad2, rad6, or rad9 diploids, contain elevated levels (52%) of G2 phase cells (FINGERHUT et al. 1984 ). The extended G2 phase suggests that spontaneously occurring DNA lesions are elevated in these cells and/or that the additional chromosomes in diploids increase the likelihood of a lesion triggering the checkpoint monitoring system.

The effects of EcoRI-induced DNA cleavage on cell survival in wild-type and mutant diploid strains are presented in Table 2. Survival of recombination-deficient rad51 and rad52 mutants was similar in both haploid and diploid cells. However, the approximately twofold decrease in viability observed in wild-type haploid strains was abolished in the diploids. The greatest change was observed in rad50 and mre11 cells. Although these strains displayed a strong G2 arrest response, the cell killing in haploid mutants was largely eliminated in diploids. This suggests that, in contrast to ionizing radiation-induced DSBs, the cohesive-ended DSBs could be efficiently repaired by recombination between homologous chromosomes throughout the cell cycle.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The experiments described here, in combination with our prior study (LEWIS et al. 1998 ), have established that many genes previously implicated in recombination-independent end-joining in yeast cells have an essential role in precise repair of DSBs in vivo. The approach presented here has permitted assessment of the repair of DSBs produced in nuclear, chromatin-associated DNA and the effects of such breaks on progression through the cell cycle. Although the NHEJ pathway was not identified in past analyses of cellular sensitivity to X rays and chemical clastogens, the current study has established that recombination and end-joining repair are both efficient processes in yeast and that this efficiency is critically dependent upon the nature of the DSB ends. Specifically, these results indicate that complementary-ended DSBs that are similar in structure to those produced by endonucleases during cellular development are rejoined by both major pathways of DSB repair. Furthermore, the combined analysis of DSB-induced effects on growth, cell cycling, and survival has led to the demonstration of separable roles for genes involved in the NHEJ pathway of repair. A summary of the results described here and in the previous study (LEWIS et al. 1998 ) is presented in Table 3.

Homologous recombination pathway genes are required for efficient progression past the G2/M checkpoint, but not for survival:
We have previously demonstrated that expression of EcoRI in yeast cells results in extensive breakage of cellular DNA and stimulates homologous recombination (LEWIS et al. 1998 ). These experiments included the surprising observation that continuous synthesis of EcoRI arrested the growth of rad52 mutants, but did not produce significantly greater killing than in wild-type cells. This result is in direct contrast to previous studies demonstrating that RAD52 is essential for repair of DSBs induced by ionizing radiation and various DNA-damaging chemicals (RESNICK and MARTIN 1976 ; SAEKI et al. 1980 ; MOORE 1982 ; GAME 1993 ). In this study we have also observed high survival in cells containing a deletion of the gene encoding the Rad51 recombinase (a structural and functional homologue of E. coli RecA) and in cells deficient in RAD54, RAD55, and RAD57 (summarized in Table 3). The similar responses of these mutants are consistent with recent models suggesting that Rad54 and the Rad55/Rad57 heterodimer work in conjunction with the strand annealing and strand exchange activities of Rad51:Rad52 to effect recombination (CLEVER et al. 1997 ; GOLUB et al. 1997 ; SUNG 1997A , SUNG 1997B ; BENSON et al. 1998 ; NEW et al. 1998 ; SHINOHARA and OGAWA 1998 ). Each of these mutants exhibits deficiency in most assays of spontaneous and induced recombination (recombination between some direct repeat sequences is an exception), though rad52 mutants are often more deficient than other single or double mutant strains (RATTRAY and SYMINGTON 1995 ; SUGAWARA et al. 1995 ; ZOU and ROTHSTEIN 1997 ).

The EXO1 gene encodes a 5'-3' exonuclease that has been implicated in both recombination and mismatch repair. A specific role for EXO1 in homologous recombination, but not in NHEJ repair pathways, is supported by a recent study of recombination between direct repeats (FIORENTINI et al. 1997 ) and by two additional outcomes described here: the phenotypic responses of exo1 mutants to EcoRI cleavage of chromosomal DNA are similar to those of rad52 mutants and the rapid cell killing observed in rad50 mutants is not altered in rad50 exo1 double mutants (Figure 5). In addition to EXO1, we have recently observed that strains containing a deletion of the putative DNA helicases encoded by SRS2 (HPR5) and RDH54 also exhibit strong DNA-damage responses when EcoRI is expressed, but do not display extensive cell killing (Table 3; K. LEWIS and M. RESNICK, unpublished results). This suggests that RDH54 (a homologue of RAD54 and a suggested member of the SWI2/SNF2 family of DNA helicases; KLEIN 1997 ; SHINOHARA et al. 1997 ) and SRS2 (previously implicated in recombinational repair events mediated by RAD51, RAD52, and RAD54 and involved in the processing of DSB termini; MILNE et al. 1995 ; SCHILD 1995 ; CHANET et al. 1996 ; PAQUES and HABER 1997 ) participate in the recombinational pathway of repair of EcoRI-induced DSBs.

Although the two principal pathways of DSB repair, NHEJ and homologous recombination, appear to be common to all eukaryotic organisms, the relative importance of each pathway varies across phylogeny. Past studies have suggested that recombinational mechanisms are dominant in S. cerevisiae and that homology-independent pathways account for most DSB repair in higher eukaryotes (e.g., ROTH and WILSON 1986 ; GODWIN et al. 1994 ; CHU 1997 ). However, a number of recent experiments have revealed that human and animal cells possess a substantial potential for recombinational repair of DSBs (e.g., CARROLL et al. 1994 ; BEZZUBOVA et al. 1997 ; SARGENT et al. 1997 ; LIANG et al. 1998 ; SONODA et al. 1998 ). Furthermore, recombinational repair pathways in higher eukaryotes are dependent on many of the same genes as yeast. For example, RAD51, RAD52, and RAD54 homologues have been identified in human and animal cells and shown to affect repair of DSBs induced by DNA-damaging agents (BEZZUBOVA et al. 1997 ; ESSERS et al. 1997 ; SONODA et al. 1998 ).

Essential, but separable, roles for genes associated with NHEJ:
In yeast cells the prominant role of RAD52-mediated recombination in repair of DSBs induced by radiation and by chemical DNA-damaging agents is well established (HAYNES and KUNZ 1981 ; GAME 1993 ). A number of past multigenic analyses of NHEJ in yeast have analyzed predominantly nonconservative rejoining events in vivo (i.e., restriction enzyme-mediated integration of DNA fragments, deletion formation in dicentric plasmids, and formation of insertions and deletions at MAT after HO endonuclease-induced cleavage; SCHIESTL et al. 1994 ; MOORE and HABER 1996 ; TSUKAMOTO et al. 1997B ). Other recent studies have suggested that conservative (precise) repair of chromosomal DSBs produced by HO and EcoRI is defective in cells lacking Ku70 or Ku80 (BARNES and RIO 1997 ; LEE et al. 1998 ; LEWIS et al. 1998 ). Finally, linear, cohesive-ended plasmid DNA is accurately and efficiently recircularized after transformation into wild-type and rad52 cells, but not in hdf1, yku80, rad50, dnl4, sir2, sir3, or sir4 mutants (BOULTON and JACKSON 1996 ; MILNE et al. 1996 ; SCHAR et al. 1997 ; TEO and JACKSON 1997 ; TSUKAMOTO et al. 1997A ; WILSON et al. 1997 ).

The multigenic analysis performed here revealed that all mutants previously implicated in NHEJ repair pathways (using the above assays) were hypersensitive to killing by EcoRI, but responses to the enzyme were not identical (see Table 3). Cell viability was reduced 20- to 40-fold in rad50, mre11, xrs2, and hdf1 mutants and was reduced ~10-fold in sir4 strains. Interestingly, growth of all end-joining mutants was blocked by EcoRI expression, but cell cycle arrest responses varied. rad50, mre11, and xrs2 mutants displayed a rapid, rad52-like increase in enlarged G2 cells after induction of EcoRI synthesis. sir4 mutants also accumulated G2 cells, but did so more slowly. In contrast, hdf1 mutants accumulated as distended G1 and G2 phase cells. This unusual phenotype has been described recently for Ku70-deficient cells after exposure to high temperature or endonuclease-induced DSBs (BARNES and RIO 1997 ). Temperature-sensitive mutants deficient in CDC28 (encoding a cyclin-dependent protein kinase), TOP2 (topoisomerase II), and CKA2 (coding for part of the catalytic subunit of casein kinase II) have also exhibited arrest at both G1/S and G2/M at their restrictive temperatures (HOLM et al. 1985 ; HANNA et al. 1995 ; ZARSOV et al. 1997 ). Potential correlations between the mechanisms responsible for biphasic arrest in Ku strains and in the other mutants await further studies. We also note that the similar kinetics of growth and cell-cycle arrest in recombination mutants (which remain viable) and NHEJ mutants (displaying progressive loss of viability) suggests that galactose induction of the endonuclease is comparable in the different mutant strains. This conclusion is also supported by the past demonstration of similar induction kinetics for GAL::HO-induced cleavage at MAT in recombination and end-joining mutants (e.g., see references in HABER 1992 ; SUGAWARA et al. 1995 ).

The functions of the Ku70:Ku80 heterodimer and Rad50:Mre11:Xrs2 proteins in NHEJ appear to be distinct. The Ku complex has been implicated in the joining reaction for broken DNA ends, but Rad50 and Mre11, which have significant sequence similarity to exoendonucleases encoded by the Escherichia coli SbcC and SbcD proteins, appear to be involved in nucleolytic processing of the ends (SHARPLES and LEACH 1995 ; MOORE and HABER 1996 ; CARY et al. 1997 ; HENDRICKSON 1997 ; PANG et al. 1997 ). Several phenotypes observed in Ku mutants are also found in rad50, mre11, and xrs2 mutants, but differences that may reflect their disparate roles in NHEJ have also been described. For example, hdf1 and rad50 mutants show equivalent deficiencies in plasmid end-joining, participate in the same pathway of MMS repair based on analysis of epistatic interactions, display unusual processing of dicentric plasmids, exhibit telomere-shortening, and are hypersensitive to killing by EcoRI endonuclease (this work; BOULTON and JACKSON 1996 ; MILNE et al. 1996 ; PORTER et al. 1996 ; SIEDE et al. 1996 ; KIRONMAI and MUNIYAPPA 1997 ; TSUKAMOTO et al. 1997B ). In addition, Ku and Rad50 proteins have been found associated with the ends of broken DNA in yeast and in human cells (FELDMANN and WINNACKER 1993 ; GETTS and STAMATO 1994 ; MILNE et al. 1996 ; NELMS et al. 1998 ). However, a number of phenotypes are unique to Ku-deficient cells. Thus, hdf1 mutants have been reported to be temperature-sensitive for growth, have increased chromosome ploidy, exhibit constitutively elevated transcription of the damage-inducible RNR2 and RNR3 genes, are found associated with replication origin-binding complexes, and have distinct damage-induced growth arrest phenotypes in yeast (this work; FELDMANN and WINNACKER 1993 ; SHAKIBAI et al. 1996 ; BARNES and RIO 1997 ). Furthermore, results obtained in this study have indicated that the repair deficiency of Ku mutants is dominant to that of rad50 mutants, as rad50 hdf1 double mutants responded similarly to hdf1 strains (Figure 5). Finally, it is worth noting that several phenotypes of Ku mutants are remarkably strain specific. For example, temperature-sensitive growth at 37°, MMS sensitivity, damage-induced arrest in G1 and G2, and increased radiation sensitivity in rad52 hdf1 double mutants have been described for some strains, but not for others (cited above). Although reports that Hdf1 associates with Sir4 and with replication origin complex proteins in vivo suggest that the phenotypic variability of hdf