Enhancement of Saccharomyces cerevisiae End-Joining Efficiency by Cell Growth Stage but Not by Impairment of Recombination

Corresponding author: Thomas E. Wilson, University of Michigan Medical School, 1301 Catherine Rd., M4214 Med Sci I, Box 0602, Ann Arbor, MI 48109-0602., wilsonte{at}umich.edu (E-mail)

Communicating editor: L. S. SYMINGTON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Cells can repair DNA double-strand breaks by both homologous and nonhomologous mechanisms. To explore the basis of pathway utilization, we developed both plasmid and chromosomal yeast repair assays in which breaks are created with restriction endonucleases so that nonhomologous end-joining (NHEJ) competes with the single-strand annealing (SSA) recombination pathway, which we show acts with high efficiency via terminal direct repeats of only 28 bp and with reduced but measurable efficiency at 10 bp. The chromosomal assay utilizes a novel approach termed suicide deletion in which the endonuclease cleaves its own gene from the chromosome, thereby ending the futile cleavage cycle that otherwise prevents detection of simple-religation events. Eliminating SSA as a possibility in either assay, either by removal of the direct repeat or by mutation of RAD52, increased the relative but not the absolute efficiency of NHEJ. In contrast, the apparent efficiency of NHEJ was specifically increased in the G1 stage of the haploid cell cycle, as well as by the glucose depletion-signaled transition to stationary phase. The combined results argue against a model in which pathway utilization is determined by a passive competition. Instead, they demonstrate an active regulation designed to optimize the likelihood of genome restoration based on cell state.

MUTAGENESIS is the process by which chemical lesions in DNA are transformed into heritable changes in base sequence. Cellular DNA repair processes ordinarily function to restore DNA to its predamaged state, but, somewhat paradoxically, mutagenesis also requires chemical resolution of the DNA lesion and thus the participation of DNA repair. A major issue is thus how the appropriateness of DNA repair is maintained to optimize the likelihood of genome restoration. Double-stranded chromosome breaks, if misrepaired, can lead to a variety of mutations, including rearrangements, small deletions and insertions, and chromosome loss. They represent an ideal case to explore the phenomenon of differential mutagenesis based on DNA repair pathway utilization, because all eukaryotic (and perhaps many prokaryotic; ARAVIND and KOONIN 2001 ; DOHERTY et al. 2001 ) cells possess two enzymatically distinct pathways for double-strand break (DSB) repair (reviewed in PAQUES and HABER 1999 ; JACKSON 2001 ). Repair by homologous recombination involves resection of the 5' ends to create 3' nucleoprotein filaments at the broken chromosome ends. RAD52, the defining member of the required epistasis group of genes, encodes a multimeric protein that possesses DNA-end-binding function (VAN DYCK et al. 1999 ; STASIAK et al. 2000 ). In true recombination, Rad52 cooperates with the Rad51 protein, a homolog of bacterial recA that catalyzes strand exchange with a homologous donor duplex to provide a template for resynthesis across the broken region. Rad52 and Rad59 can also act to catalyze the direct annealing of the two 3' nucleoprotein filaments independently of the other epistasis group members, however, to ultimately create a deletion in a pathway known as single-strand annealing (SSA; see Fig 1A; IVANOV et al. 1996 ; SHINOHARA et al. 1998 ; SUGAWARA et al. 2000 ). In contrast, nonhomologous end-joining (NHEJ) is characterized by the preservation of the DNA ends and lack of involvement of a homologous donor, with repair entailing engagement and likely end-to-end bridging by the Ku heterodimer (Yku70/Yku80 in budding yeast; see Fig 1A; JONES et al. 2001 ), and ultimately ligation by the DNA ligase IV/XRRC4 complex (Dnl4/Lif1 in yeast; WILSON et al. 1997 ; HERRMANN et al. 1998 ).

View larger version (28K): In this window In a new window Download PPT slide   Figure 1. Design and validation of the competitive plasmid assay. (A) At a double-strand break with a terminal direct repeat (shaded boxes, DR) and compatible overhangs, such as those generated by KpnI and I-SceI, repair may proceed via either NHEJ, in which the Ku heterodimer engages the intact ends, or SSA, in which both resection of the 5' strand and engagement by Rad52 are required. (B) Diagram of the competitive plasmid substrates. Derivatives vary in the presence and length of the direct repeat. (C) Yeast strain YW422, bearing a yku70 mutation to prevent NHEJ, was transformed with plasmids containing the indicated direct repeat lengths. The normalized transformation efficiency is plotted as a function of repeat length.

Because of these fundamental differences in their enzymology it is likely that at some step each DSB repair (DSBR) pathway becomes irreversibly committed after early steps that are common to both pathways, suggesting a need for regulation. For example, the Mre11-Rad50-Xrs2 complex (corresponding to the Mre11-Rad50-Nbs1 complex in human cells) has an end-tethering function and clear, although variable, defects in both pathways (PETRINI 1999 ; DE JAGER et al. 2001 ). Further, observations of complex "hybrid" NHEJ recombination repair events indicate that both pathways might potentially be active even at a single DSB (RICHARDSON and JASIN 2000 ; KRAUS et al. 2001 ). An initial insight into DSBR pathway regulation came from observations that budding yeast cells of the mating type a/ (typically diploid cells) repress their NHEJ pathway (ASTROM et al. 1999 ; LEE et al. 1999 ), which has more recently been shown to be due to transcriptional repression of the NHEJ-activating gene NEJ1 (FRANK-VAILLANT and MARCAND 2001 ; KEGEL et al. 2001 ; VALENCIA et al. 2001 ). Nej1 interacts with Lif1 (OOI et al. 2001 ), although how this leads to enhanced NHEJ is not clear. Moreover, there is evidence from higher eukaryotic cells that NHEJ is used preferentially in G1 (LEE et al. 1997 ; TAKATA et al. 1998 ), although studies of HO endonuclease-induced breaks in yeast cells paradoxically concluded that the NHEJ was more active in S/G2 (MOORE and HABER 1996 ).

In an effort to further understand the manner in which eukaryotic cells optimize DSBR, we have begun to systematically examine pathway utilization and its effect on chromosomal rearrangement in the yeast Saccharomyces cerevisiae. Here we describe the development of two related assays that allow for the quantitative assessment of pathway utilization in a competitive fashion, one of which is amenable to genetic screening. We find no evidence of a simple competition between the pathways that would be revealed by an increase in utilization of one pathway when the other is blocked. In contrast, our results suggest that NHEJ is activated most specifically in the G1 and especially G0/postdiauxic stages of the haploid cell cycle. These results are considered in the context of recent similar studies in other model systems, as well as the technical difficulties of measuring DSBR in stationary-phase cells.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Yeast strains:
Strain genotypes are listed in Table 1. Strains used for radiosensitivity analysis were described previously (WILSON et al. 1997 ). Strains used in the plasmid study were isogenic derivatives of YW388 (VANCE and WILSON 2001 ), and those used in the suicide deletion assay were isogenic derivatives of BY4742 (BRACHMANN et al. 1998 ). Genes were disrupted using a PCR-mediated one-step replacement technique (BRACHMANN et al. 1998 ). All disruptions were confirmed by PCR. The yku70::kanMX4 and rad52::kanMX4 alleles in the suicide deletion strains were introduced by mating and sporulation with strains from the Saccharomyces Genome Deletion Project array (WINZELER et al. 1999 ). Corresponding ADE2 ura3 control strains were derived from the ade2::SD0+::URA3 strains by selecting and purifying a white colony after DSB induction on galactose plates.

Yeast growth:
Synthetic defined medium was as described (WILSON and LIEBER 1999 ) with either 2% glucose or 2% galactose as the carbon source. Rich medium was YPA (1% yeast extract, 2% Bacto-peptone, 40 µg/ml adenine) plus 2% dextrose to make YPAD for routine cultures. When varying the carbon source, the dextrose was reduced to 0.1% (an amount intended to facilitate the transition to respiratory metabolism) with further supplementation with 2% potassium acetate, 3% ethanol, or 3% glycerol. All cultures were grown at 30°, with shaking at 280 rpm for liquid cultures. Cell growth in liquid culture was monitored by measuring the OD₆₀₀ in a spectrophotometer, correcting from dilutions so that actual readings were between 0.1 and 1.0. A different spectrophotometer was used for the ionizing radiation experiments than for the plasmid or suicide deletion experiments. A direct comparison of the instruments was not possible, but the former gave lower readings at comparable culture densities.

Exponentially dividing glucose cultures were obtained by diluting an overnight culture 25-fold into fresh media in the morning, followed by still further dilution in the evening and by overnight growth to achieve a culture the next morning with an OD₆₀₀ <2.0. Experiments in which the culture density was varied were performed in one of two ways. First, 50 ml exponentially dividing cultures were grown in 250-ml flasks and followed over several days, with aliquots taken as needed. Second, 25 ml serial dilutions of an exponentially dividing culture were made in 50-ml conical tubes such that the desired range of densities was obtained over the course of the following day. Similar results were obtained with either protocol.

Glucose determination:
Glucose remaining in the medium was determined using an assay kit from Sigma-Aldrich (#510-A) according to the manufacturer's instructions and using the starting medium as the reference sample.

Thermotolerance:
Cells were washed into water and heated to 55° in a water bath for 5 min followed by cooling on ice for 5 min. Cells were then diluted and plated to YPAD. Viability was determined by comparing colony counts to those obtained in parallel from samples treated identically but held at room temperature instead of 55°. Heat treatment for Table 5 was similar except that it was performed in the growth medium and was followed by a 2-hr recovery period at 30° prior to plating.

Synchronization:
Synchronization was achieved by adding either 5 µg/ml -factor (Sigma-Aldrich) or 10 µg/ml nocodazole (Sigma-Aldrich) to an exponentially dividing culture. After 2 hr of further shaking, the fraction of budded cells was 10 and 95%, respectively. Release from synchronization was achieved by pelleting the cells in a centrifuge, washing once, and resuspending in fresh medium.

Competitive plasmid assay:
Plasmid pES16 (WILSON et al. 1997 ) was modified as follows to create the competitive plasmid substrates. After destroying the KpnI site in the polylinker, a novel KpnI site was created just downstream of the second ADE2 codon by simultaneous ligation of two PCR products into NotI/BglII-cut plasmid as previously described (WILSON and LIEBER 1999 ). Next, a common "dummy" sequence of 18 bp was inserted upstream of the new KpnI site by ligation of a tailed PCR fragment into NotI/KpnI-cut plasmid. Finally, different length direct repeats were created by inserting 27-bp sequences downstream of the KpnI site by ligation of tailed PCR fragments into KpnI/BglII-cut plasmid. Note that all plasmids have the same number of inserted nucleotides (they vary only in the identity of the base pairs immediately downstream of the KpnI site) and that even large amino-terminal tags do not affect Ade2 function. The exact sequences around the ADE2 start codon are the following [the ADE2 codons and KpnI site (GGTACC) are in boldface type, and the repeats are underlined]:

• pK1827 (no repeat), ATGGATCGAGGATCGCTAGGTCGTGGTACCTTGTCAAGAATCTCTGACGCTGGACGATCT;

• pDR04 (4-bp repeat), ATGGATCGAGGATCGCTAGGTCGTGGTACCGTGTCAAGAATCTCTGACGCTGGACGATCT;

• pDR10 (10-bp repeat), ATGGATCGAGGATCGCTAGGTCGTGGTACCTAGGTCGTGATCTCTGACGCTGGACGATCT;

• pDR19 (19-bp repeat), ATGGATCGAGGATCGCTAGGTCGTGGTACCGAGGATCGCTAGGTCGTGGCTGGACGATCT;

• pDR29 (29-bp repeat), CAAGTATGGATCGAGGATCGCTAGGTCGTGGTACCAAGTATGGATCGAGGATCGCTAGGTCGTCT.

For routine transformations, cell growth, plasmid digestion, purification, and transformation were performed with DNA amounts in the linear range as previously described (WILSON and LIEBER 1999 ). Variations in cell growth are described above, but in all cases approximately the same number of cells were added to the transformation mix. The same preparations of DNA were used within any panel, but preparations did vary between panels with attendant small changes in the measured efficiencies. Results are expressed as the transformation efficiency with KpnI-digested plasmid divided by that obtained in parallel transformations with undigested pK1827. Importantly, it is reproducibly observed that the transformation of linearized plasmids in fact substantially exceeds that of supercoiled plasmids, and so the KpnI/uncut ratio can significantly exceed unity.

Verification of the plasmid repair mechanism was achieved by colony PCR using primers OW563 (5'-GGCAGGAGAATTTTCAGCATC) and OW750 (5'-GTGATTGACTCTTGCTGAC), which yielded a 326-bp product for NHEJ and a 292-bp product for SSA. NHEJ PCR products were additionally subjected to cleavage with KpnI, with 185- and 141-bp products resulting if a KpnI site was recreated during repair.

Ionizing radiation:
Strains were grown as needed and then washed into water to 5 x 10⁶ cells/ml. Following irradiation to the indicated doses by timed exposure to a ⁶⁰Co source, 10-fold serial dilutions were made and appropriate amounts plated to YPAD. Colonies were counted and survival scored relative to a parallel unirradiated control of the same strain.

Suicide deletion assay:
To construct the suicide deletion allele directly in the yeast chromosome, the nuclear-modified coding sequence of the mitochondrial I-SceI mega-endonuclease (PLESSIS et al. 1992 ; provided by Bernard Dujon through collaboration with Michael Lieber) was first amplified by PCR and ligated into the BamHI-SalI sites downstream of the GAL1 promoter and upstream of the URA3 gene in the vector pBM272 (JOHNSTON and DAVIS 1984 ) to create pTW334. Next, three types of PCR fragments were generated: (i) 573 bp of the ADE2 sequence upstream of and including the second codon, ending with an MluI site; (ii) the GAL1-I-SceI cassette and the adjacent URA3 marker gene of pTW334, with the different required combinations of frameshifting and repeat nucleotides introduced as primer tails, beginning and ending with MluI and BglII sites, respectively; and (iii) 699 bp of ADE2 sequence downstream of and including the third codon, beginning with a BamHI site. After amplification and cleavage with the appropriate restriction enzymes, the three PCR fragments were ligated (BamHI and BglII overhangs are compatible). Following agarose gel electrophoresis, the 4.6-kb ligated fragment corresponding to (i), (ii), and (iii) was readily identified, gel purified, and transformed into BY4742. Successful integrants were identified as being Ade^- and Ura⁺ in a galactose-sensitive manner. The exact sequences around the ADE2 start codon in the different suicide deletion alleles following I-SceI cleavage and simple-religation NHEJ are the following (the ADE2 codons and the sequence corresponding to the 4-bp I-SceI overhang are in boldface type, and the repeats are underlined):

• ade2::SD0-::URA3 (no repeat, NHEJ = ade2, SSA not possible), ATGGATACGCGTGTATTACCCTGTTATCCCTAGCGTCAGATCCTCT;

• ade2::SD0+::URA3 (28-bp repeat, NHEJ = ade2, SSA = ADE2) ATGGATACGCGTCCTAGCGTACTCAAACGTGTATTACCCTGTTATCCCTAGCGTACTCAAACGTGTATTACCCTACAGATCCTCT;

• ade2::SD2-::URA3 (no repeat, NHEJ = ADE2, SSA not possible), ATGGATAAACGCGTGTATTACCCTGTTATCCCTAGCGTCAGATCCTCT;

• ade2::SD2+::URA3 (28-bp repeat, NHEJ = ADE2, SSA = ade2), ATGGATAAACGCGTCCTAGCGTACTCAAACGTGTATTACCCTGTTATCCCTAGCGTACTCAAACGTGTATTACCCTACAGATCCTCT.

Verification of suicide deletion NHEJ events was achieved by demonstrating the reconstruction of a new I-SceI site in one of three ways. First, primers OW620 (5'-GCTACCAAATGACATTCTCTG) and OW603 (5'-CCTTAAGTTGAACGGAGTCC) were used to generate a PCR fragment whose size of 1.3 kb indicated in 23 of 24 cases that suicide deletion had occurred. The fragments were further subjected to cleavage with I-SceI (New England Biolabs, Beverly, MA), which generated 574- and 699-bp products if a site had been recreated. Second, pTW334 was introduced by transformation and the transformants replica plated to galactose medium, where either poor survival (no repeat alleles) or conversion from white to red (repeat alleles) indicated that an I-SceI site had been cleaved. Third, pTW334 was first transformed into an isogenic strain of opposite mating type and introduced by mating, with scoring of the diploids as above.

For quantification, strains bearing the suicide deletion or ADE2 alleles were grown as required in synthetic defined medium lacking uracil and containing 40 µg/ml adenine (with uracil added for ADE2 control strains). Routine assays were performed after cultures had achieved postdiauxic growth, typically between 24 and 48 hr. Ten-fold serial dilutions of the cells were made in water, and appropriate volumes plated to synthetic defined medium. Percentage survival on galactose plates is expressed relative to parallel platings to glucose.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The competitive plasmid assay: efficient SSA via short terminal repeats:
To quantitatively examine the outcome of competition between NHEJ and homologous recombination, it is necessary to monitor repair in a population of cells where each has suffered a DSB that can be repaired in a stable fashion by either pathway. Moreover, the repair pathway used by a given cell should be readily discernible so that large numbers can be assayed. As a first approximation of this, we transformed cells with derivatives of the plasmid pES16 (WILSON et al. 1997 ), which had been linearized with KpnI just downstream of the ADE2 start codon (Fig 1). Transformation to Ura⁺ requires recircularization, which could occur by the SSA subpathway of homologous recombination or simple-religation NHEJ. SSA was controlled by varying the presence and length of a terminal direct repeat through which the ends interact. NHEJ and SSA events either did or did not create an in-frame ADE2 gene, respectively. Because the strain is ade2, ADE2/white Ura⁺ colonies must represent repair by NHEJ. The high efficiency of recombination further ensures that ade2/red Ura⁺ colonies obtained with repeat-bearing plasmids are nearly all SSA events. A greatly reduced frequency of red colony formation is observed in the absence of the direct repeat, but this primarily represents imprecise NHEJ. These predictions were verified by PCR analysis of red and white colonies (Table 2) and by previous results (WILSON et al. 1997 ; WILSON and LIEBER 1999 ).

To explore the effects of terminal direct repeat length on SSA efficiency, a series of plasmids with different repeat lengths was transformed into yku70 mutant yeast, which ensured that transformants could not arise by an unanticipated NHEJ event between the shorter internal microhomologies. Somewhat surprisingly, we found that SSA was readily detectable above background at very short repeat lengths of as little as 10 bp, although a 4-bp repeat was not different from the control plasmid (Fig 1C). Transformation efficiency was nearly maximal at a repeat length of 29 bp. This efficiency in fact exceeded that of supercoiled plasmid, in this case by >100-fold, presumably due to differences in the extent of plasmid uptake.

Lack of competition between NHEJ and SSA: plasmid assay:
The logic we sought to test in these experiments was that a mutation or substrate alteration that decreased the efficiency of one repair pathway should also lead to an increase in the absolute efficiency of the remaining pathway. This logic assumes that repair is truly competitive, i.e., that engagement by one pathway makes a DSB unavailable to the other. Such an assessment is possible because transformation efficiencies are normalized to parallel transformations with uncut DNA and thus compared on an absolute reference scale.

First, the behavior of a subset of the repeat-bearing plasmids was tested in wild-type haploid cells (Fig 2). These NHEJ-proficient cells again failed to show a significant increase in ade2/red colony recovery with the 4-bp repeat that indicates that the yeast NHEJ apparatus is inefficient in its use of internal microhomologies, at least when simultaneously presented with a compatible 4-bp overhang. As above, the 29-bp repeat supported a 100-fold increase in ade2/red SSA events, which initially appeared to be coincident with a modest decrease in ADE2/white NHEJ events when both were counted from the same uracil-selective plates (Fig 2A). When the ADE2/white events were counted from NHEJ-selective plates lacking both uracil and adenine, however, there was no difference in their recovery comparing plasmids with and without repeats (Fig 2B). This phenomenon was reproducible and may partially reflect the fact that fewer ADE2/white colonies could be counted from plates that also allowed growth of the ade2/red SSA events. A larger effect is likely caused by uptake and repair of more than one plasmid per cell. Since SSA via the 29-bp repeat is 10-fold more efficient than NHEJ, some NHEJ-repaired plasmids may simply be lost when selection for them is not maintained. In any case, it is clear that the total repair rate was substantially decreased when the terminal repeat was absent.

View larger version (31K): In this window In a new window Download PPT slide   Figure 2. NHEJ efficiency in the plasmid assay is independent of the direct repeat and RAD52. (A) Wild-type haploid and diploid yeast strains were transformed with plasmids with no repeat (–) or 4- or 29-bp repeats and transformation efficiency was scored on synthetic medium lacking uracil. In A–C, solid and open circles represent the normalized efficiency of ade2/red/SSA and ADE2/white/NHEJ colony recovery, respectively. Strains were YW389 (haploid) and YW397 (diploid). (B) The same transformation mixes corresponding to the first and third columns of A were also plated to synthetic medium lacking uracil and adenine to select for NHEJ events. (C) Plasmids with and without the 29-bp repeat were transformed into yeast bearing the indicated deletion mutations and transformation efficiency scored on synthetic medium lacking uracil. Strains were YW474 (wt), YW422 (hdf1), YW426 (dnl4), and YW415 (rad52). Different preparations of DNA were used in A and C, and so the absolute efficiencies are slightly different. Throughout, points represent the mean ± standard deviation of at least four independent transformations.

A similar phenomenon was observed when SSA was prevented by mutation of RAD52. This mutation decreased SSA by 16-fold, but led to no correspondent increase in NHEJ (Fig 2C). Conversely, mutation of either YKU70 or DNL4, presumably early and late contributors to NHEJ, led to the expected large decrease in NHEJ efficiency but no apparent increase in SSA, although even if all NHEJ events had become SSA events the increase may not have been detectable within the error of the assay. Similar overall results were seen when the ends were digested with Asp 718, a 5' isoschizomer of KpnI, or when the recombination event was gene conversion with a chromosomal ade2 point mutant allele (not shown). In summary, manipulating the DNA ends and repair genes had predictable detrimental consequences on the NHEJ and SSA efficiencies, but in no case did we observe a compensatory increase in the remaining pathway.

Increased NHEJ efficiency in haploid G1: plasmid assay:
A different model of pathway utilization was suggested by the observation that NHEJ of all plasmids was decreased >20-fold in diploid as compared to haploid cells, independent of repeat length (Fig 2A). This observation has been documented by several other groups and shown to be a result of downregulation of NHEJ via repression of NEJ1 as a consequence of the MATa/MAT mating type (ASTROM et al. 1999 ; LEE et al. 1999 ; FRANK-VAILLANT and MARCAND 2001 ; KEGEL et al. 2001 ; VALENCIA et al. 2001 ). Our experiments add the observation that MAT status does not affect SSA of the same broken plasmid whose NHEJ is impaired. An explanation for this observation is that haploid yeast cells must spend a significant fraction of their cell cycle in a state where no correct homolog is present for directing recombination, and so NHEJ becomes advantageous (VALENCIA et al. 2001 ). Extending this logic, we hypothesized that NHEJ should be a greater contributor to repair during the G1 phase of the haploid cell cycle. To test this, an exponentially growing culture of wild-type cells was synchronized in G1 with -factor, released into fresh medium, and then transformed at various times over the ensuing one and a half divisions with the plasmid containing the 29-bp repeat (Fig 3). As the population began to divide, the frequency of NHEJ events did in fact drop nearly 10-fold, only to rise and then fall again with the second round of budding. SSA remained highly efficient throughout the cell cycle, however, and so this synchronization experiment did not produce the anticipated finding that NHEJ would largely replace recombination during G1. In fact, the NHEJ efficiency in all cell cycle stages was surprisingly lower than that observed in previous experiments, even in G1. The explanation for this seemingly paradoxical finding must come from some difference in the experimental technique and is addressed below.

View larger version (23K): In this window In a new window Download PPT slide   Figure 3. NHEJ efficiency in the plasmid assay fluctuates with the cell cycle. A wild-type yeast culture (YW474) dividing exponentially in YPAD was synchronized into G1 with -factor, released into fresh medium, and then transformed at the indicated times with the plasmid containing the 29-bp repeat. The ensuing divisions were monitored microscopically after LiAc treatment and prior to heat shock by microscopic evaluation. The time points at which primary and secondary buds appeared are indicated on the graph. Data are plotted as in Fig 2.

Increased NHEJ efficiency in postdiauxic/stationary phase: radiosensitivity:
Other than treatment with -factor, the primary technical difference between Fig 2 and Fig 3 was that, for synchronization, cells were grown at higher dilution to a more active exponential growth. Experiments with ionizing radiation suggested that this seemingly small difference might account for the altered NHEJ frequency (Fig 4). First, rad52 mutant cells, in which radiation-induced DSBs either are repaired by DNL4-dependent NHEJ or are lethal (WILSON et al. 1997 ), became increasingly resistant to ionizing radiation as culture density increased (Fig 4A). Moreover, only at high culture density was an incremental radiosensitivity conferred on rad52 cells by dnl4 mutation, suggesting that NHEJ was being induced as the cultures aged (Fig 4B). Also, a surprising and currently unexplained additional resistance that did not require RAD52, DNL4, or YKU70 was conferred at high culture density (Fig 4B and not shown). Similar findings have been reported by HERRMANN et al. 1998 . It could be that G1 cells join ends more efficiently, as above, and the fraction of S/G2/M cells simply decreased with culture density. Indeed, the fraction of budded cells did drop from 75 to 55% across experiments such as that in Fig 4A, consistent with the genetically programmed diauxic shift to respiratory metabolism that occurs upon depletion of glucose, after which cells continue to divide more slowly for several days (WERNER-WASHBURNE et al. 1993 ). To test the G1 hypothesis directly, exponential phase cultures of rad52 yeast were synchronized in G1 or G2 using -factor or nocodazole, respectively. Neither treatment increased the radioresistance significantly or induced a DNL4-dependent hypersensitivity (Fig 4C), demonstrating that more than just the fraction of G1 cells must change with culture density. Indeed, after several days cells more gradually undergo a second genetically programmed switch to a highly stress-resistant cell state known as stationary phase, or G0 (WERNER-WASHBURNE et al. 1993 ). The fact that the cultures showing the incremental dnl4 radiosensitivity had become as much as 50% tolerant to a 5-min incubation at 55° suggests that they had in fact initiated the stationary-phase transition.

View larger version (20K): In this window In a new window Download PPT slide   Figure 4. Sensitivity of rad52 mutant yeast to ionizing radiation depends on culture density. (A) Cultures of rad52 mutant yeast were obtained over a wide OD600 range from exponential to early stationary phase. This was achieved by starting cultures as serial dilutions from an exponentially dividing culture. Radiosensitivity was determined after 18 hr growth as the percentage survival following 20 krad ionizing radiation as compared to nonirradiated controls (open squares, right axis). Plotting final vs. initial densities on a log-log chart mimics a growth curve (solid squares, left axis) and allows identification of cultures that were past the diauxic shift (dashed line). (B) Duplicates of rad52 and rad52 dnl4 mutant yeast cultures were grown to the indicated optical densities prior to irradiation. The percentage survival relative to nonirradiated cultures is plotted as a function of dose (mean ± standard deviation). (C) Exponential cultures in YPAD of rad52 and rad52 dnl4 mutant strains were synchronized in G1 or G2 with -factor or nocodazole, respectively, and radiosensitivity determined as in B. For reference, dashed lines correspond to the position of the two rad52 curves from B. Throughout, strains were YW115 (rad52) and YW136 (rad52 dnl4).

Increased NHEJ efficiency in postdiauxic/stationary phase: plasmid assay:
We next tested the effect of culture density on transformation efficiency with the 29-bp repeat plasmid. Plotting the results of a single experiment revealed that the recovery of NHEJ events went from very low levels at low density to almost as high as SSA at an OD₆₀₀ of 9, a nearly 50-fold increase (Fig 5A). Paradoxically, the recovery of SSA colonies did not change significantly. This might again be due to multiple plasmid uptake allowing for more than one chance for repair by SSA. Still other difficulties are that high culture density causes a refractoriness to transformation and an increased variability and also might have disproportionate effects on uncut as compared to linear DNA transformation. To minimize the impact of these issues, we plotted a larger data set collected over six independent experiments as the percentage of colonies that had repaired the plasmid by NHEJ (Fig 5B). In every experiment there was a large increase in the fractional recovery of NHEJ at high densities, in some cases as high as 40%. The most prominent upward inflection in the curve did not occur until more than one OD₆₀₀ unit higher than the disappearance of the glucose, which, because cultures were dividing slowly at this point, represents a lag of many hours. However, the precise timing is difficult to establish with certainty due to the inherent variability in the measurement, and in some experiments the increase in NHEJ appeared to initiate soon after the diauxic shift. Conversely, high levels of thermotolerance were not observed at the OD₆₀₀ values achieved in Fig 5B, i.e., those at which the yeast were still transformable, but did eventually reach as high as 70% at an OD₆₀₀ of 14. Thus, an apparent increase in NHEJ efficiency occurred gradually during the period of postdiauxic growth leading into, but before, the stationary-phase transition.

View larger version (30K): In this window In a new window Download PPT slide   Figure 5. NHEJ efficiency in the plasmid assay is increased in postdiauxic/stationary phase. A series of cultures of the wild-type strain YW474 at varying optical densities were transformed with the 29-bp repeat plasmid and the results plotted as a function of OD600. (A) A single experiment in which NHEJ (open symbols) and SSA (solid symbols) are plotted as the normalized transformation efficiency. Circles represent single data points. Squares represent low and high points measured in quadruplicate (mean ± standard deviation). (B) A compilation of six independent experiments plotted as the percentage of all transformants that were Ade+, i.e., NHEJ (open circles). This means of expressing the data is independent of any decrease in the overall transformation efficiency or uncut/linear transformation ratio caused by increased culture density. Shown on the same graph are the percentage of glucose remaining (solid triangles) and percentage of thermotolerance (open squares) as measured in one of the experiments.

Effect of carbon source on NHEJ efficiency: plasmid assay:
Because the increase in NHEJ was temporally related to the diauxic shift, we next directly tested the effect of carbon source on NHEJ efficiency. Wild-type cells were pregrown to exponential phase in YPAD and then washed into fresh YPA medium containing dextrose, acetate, glycerol, or ethanol as the carbon source. The fraction of colonies that had been repaired by NHEJ was determined after 5 and 24 hr further growth. At 5 hr all cultures had adapted to the new carbon source and begun dividing again. At 24 hr all were still dividing and none were stationary (which would have prevented transformation); the glucose culture was early postdiauxic (OD₆₀₀ = 7.3). Each nonfermentable carbon source caused a clear increase in the fractional contribution of NHEJ in a time-dependent fashion as compared with glucose, with glycerol being the most effective (Fig 6). The transition to respiratory metabolism thus accelerated the development of increased NHEJ recovery.

View larger version (19K): In this window In a new window Download PPT slide   Figure 6. NHEJ efficiency in the plasmid assay as a function of carbon source. A culture of the wild-type yeast strain YW474 growing exponentially in YPAD was washed into YPA medium with the indicated carbon sources and incubated for 5 hr (open bars) or 24 hr (solid bars) prior to transformation with the 29-bp repeat plasmid. The percentage of transformants that were Ade+ (i.e., NHEJ) is graphed.

The suicide deletion competitive chromosomal assay:
We sought to verify our results with another, more physiological, assay in which a discrete break is created in a chromosome by expressing a mega-endonuclease that cleaves a single natural or engineered site. A limitation of this approach is that NHEJ, unlike most engineered recombination events, almost always leads to the restoration of the original cleavage site by simple religation. The restored site is recleaved and therefore unstable in the face of continued endonuclease expression. Only rare imprecise joints that are no longer recognized by the endonuclease are recovered (MOORE and HABER 1996 ; WILSON and LIEBER 1999 ). Others have sought to circumvent this problem by timed expression of the endonuclease (LEE et al. 1999 ), but this approach is limited in its ability to quantify repair on a per-break basis and is not suitable for high-throughput analysis. We developed an alternative approach in which the coding sequence of the mega-endonuclease I-SceI was placed under the control of the GAL1 promoter and inserted into the ADE2 gene just downstream of its start codon (Fig 7A). Two I-SceI cleavage sites flank the expression cassette, which also includes a URA3 marker gene to allow for positive selection and maintenance of the allele. On transfer to galactose medium, the induced I-SceI endonuclease is expected to cleave both sites, resulting in a 3.3-kb interstitial deletion of the chromosome and concomitant loss of the fragment containing the I-SceI expression cassette. The assay is thus termed "suicide deletion." The result is immediate termination of I-SceI expression, which we predicted would lead to stabilization of efficient simple-religation NHEJ events that linked the chromosome ends. Outside of the I-SceI cleavage sites the allele configuration mimics the plasmid assay, i.e., with ADE2 status indicating whether repair occurred by NHEJ via a 4-bp 3' overhang or SSA via a 28-bp terminal direct repeat.

View larger version (71K): In this window In a new window Download PPT slide   Figure 7. Design and validation of the suicide deletion competitive chromosomal assay. (A) The suicide deletion allele(s) that were constructed in the ADE2 locus on chromosome 15 are diagrammed. Variable portions are the direct repeat (dark shading, DR) and the two frameshifting nucleotides. Upon galactose induction, I-SceI enzyme will be expressed and cleave the two I-SceI recognition sites (staggered boxes), thereby excising its gene from the chromosome (arrows), leaving a simple DSB that can be repaired by NHEJ or SSA. Note the similarity of the DSB ends to the plasmid assay (Fig 1) once the I-SceI/URA3 cassette is excised. Small arrows indicate the position of PCR primers used to verify repaired alleles; note that the starting allele, SSA, and NHEJ all yield different size products. (B) Approximately equal numbers of wild-type yeast strains bearing the indicated suicide deletion alleles were plated to synthetic complete galactose medium. Insets show a higher magnification view of NHEJ colonies to show that they are frequently sectored.

Fig 7B shows the appearance of strains bearing the four basic suicide deletion alleles when plated to nonselective galactose medium. In addition to the observed toggling of colony colors conferred by frameshifting base pairs (compare the top and the bottom) and the dependence of the SSA events on the direct repeats (compare the left and the right), colonies whose color predicted that they resulted from NHEJ were readily recovered. Closer inspection revealed that many of these colonies were sectored. Consistent with this, the construct predicted to give ADE2 NHEJ events yielded a heterogeneous population of colony sizes when plated to galactose medium lacking adenine (not shown). We interpret this to indicate that a significant subset of NHEJ-repaired chromosomes are recleaved in the first few cell divisions, with secondary cell loss or repair by SSA, although the extent of this recleavage is clearly self-limited.

Extensive testing was performed to validate these predictions in wild-type strains. True suicide deletion events should lose the URA3 gene, and indeed <1% of both the predicted SSA and NHEJ events remained Ura⁺ (not shown). Also, all colonies were stable to repeated restreaking on galactose medium, with all now giving similar colony size and appearance regardless of their size on the primary plate. All events were therefore genetically stable, and the factors that gave rise to the initial NHEJ colony size heterogeneity did not persist. Direct examination of the ADE2 alleles in recovered colonies was performed by various means (summarized in Table 3). PCR using primers that flanked the cut sites (see Fig 7A) gave products in 23 of 24 NHEJ events whose size indicated that suicide deletion had in fact occurred and that could be recleaved in vitro by I-SceI. Reintroduction of a functioning I-SceI expression cassette into an independent set of 98 ADE2 NHEJ isolates, either by transformation with an expression plasmid or by mating to a strain bearing this plasmid, led to their conversion to ade2 98% of the time, which could occur only if the I-SceI site had in fact been recreated in an initial simple-religation NHEJ event. In all, there was again no difference between large and small NHEJ colonies or between strains containing or lacking the direct repeat.

Lack of competition between NHEJ and SSA: suicide deletion assay:
Table 4 shows the results of quantitative plating experiments comparing different suicide deletion alleles. Considering wild-type strains first, it was observed that 30% of cells survived by SSA when the direct repeat was provided. Most of the remaining 70% did not survive, which is most likely attributable to an imperfect efficiency of the SSA event. NHEJ accounted for the survival of 3%, however. The suicide deletion alleles mirrored the plasmid results in that the presence of the direct repeat had no effect on the NHEJ efficiency. Considering mutant strains, it was observed that the recovery of colonies predicted to arise by NHEJ was >84-fold decreased by a yku70 mutation. Similarly, SSA was 37-fold decreased by a rad52 mutation. This provides still further verification that the repair events were being executed by the predicted pathways. Although the predicted NHEJ events from the yku70 strain did not recreate an intact I-SceI site as uniformly as the wild-type strain, 90% nonetheless did, supporting previous plasmid results that some precise end rejoining is possible even in the absence of yku70, albeit with greatly reduced efficiency (Table 3; WILSON et al. 1997 ). The apparent lack of competition between NHEJ and SSA was again evident in that both yku70 and rad52 mutants gave no increase in the recovery of the still-functioning pathway (Table 4). In fact, yku70 and rad52 mutations caused a mild 1.5-fold decrease in SSA and NHEJ events, respectively, which is discussed further below.

Increased NHEJ efficiency in postdiauxic/stationary phase: suicide deletion assay:
Finally, we attempted to use the suicide deletion system to verify NHEJ enhancement in stationary-phase yeast. A difficulty is that growth on galactose is used to induce the DSB, which means that cells must either lose glucose repression or exit stationary phase prior to inducing the chromosome break. Some effect might nonetheless persist if the changes responsible for NHEJ enhancement are slow compared to the time required for activation of the GAL1-I-SceI-fusion gene. Cultures of wild-type yeast were plated to galactose from three growth conditions: exponential phase in YPAD, 2-day postdiauxic cultures, and 5-day stationary-phase cultures. In the last case, cultures were heated to 55° prior to plating to kill any cells that had not achieved stationary phase. It was not meaningful to normalize to glucose in this experiment because the inherent galactose plating efficiency, independent of any need for DSBR, is different for these source cultures. Results were therefore expressed as the percentage of colonies on galactose that had repaired the break by NHEJ. As seen in Table 5, the contribution of NHEJ was again lowest in exponential phase, rising in postdiauxic and stationary-phase cultures by three- and sixfold, respectively.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Repeat length and NHEJ fraction in the plasmid assay:
We have adapted our existing plasmid repair assay to examine the outcome of competition between simple-religation NHEJ and the SSA pathway of recombination. It was observed that a terminal repeat of 29 bp supported efficient SSA and a high rate of transformation that substantially exceeded that seen in the absence of a repeat. It is thus clear that most linear plasmids are not repaired when only NHEJ is possible. Coincidentally, SUGAWARA et al. 2000 also recently determined that 29 bp was sufficient to support a 0.17% frequency of SSA, but this is much lower than our highly efficient plasmid repair or suicide deletion SSA frequency of 30%. The critical difference is that our repeated sequences both lie within several nucleotides of a DSB end, while in SUGAWARA et al. 2000 one repeat was recessed. This suggests that longer homology tracts are of greater relative benefit when a more extensive search for the base-pairing partner is required; shorter repeats are presumably more likely to interact nonproductively with nonhomologous sequences as resection proceeds. Our data thus also establish the minimal length of homology that need be present for Rad52 to facilitate bridging of two ends in vivo, and it is surprisingly short, as little as 10 bp. This result injects a note of caution for those considering spontaneous chromosomal rearrangements with limited "microhomology" at the fusion point. Such events are generally attributed to NHEJ, but data here suggest that they might sometimes be created by SSA.

The suicide deletion assay for NHEJ:
FAIRHEAD et al. 1996 previously showed that two chromosomal I-SceI sites arranged in an inverted orientation can be cleaved by I-SceI and repaired in a stable fashion by NHEJ of the incompatible outside ends, resulting in an interstitial deletion. We have developed a variation of this experimental approach in which simple-religation NHEJ of tandemly repeated sites can also be readily detected. The essential features of this suicide deletion assay are (i) that the DSB and its subsequent repair terminate endonuclease expression, thereby severely limiting recleavage and preserving simple-religation NHEJ events, and (ii) that the chromosome has suffered a small interstitial deletion after repair, so that all events can be distinguished from the starting substrate. The feasibility of the assay was documented by the ready recovery of NHEJ events that had indeed recreated an I-SceI cleavage site in the context of the predicted interstitial deletion. A primary advantage of the suicide deletion system is that very little manipulation of the cells is required. In addition to eliminating many potential artifacts, this makes the assay amenable to high-throughput genetic analysis. Another advantage is the ease with which SSA and NHEJ can be selected by changing two frameshifting base pairs in otherwise identical alleles. A limitation is that the detected events are deletions and therefore mutagenic as opposed to restorative repair; it cannot be ruled out that the relative repair frequencies would be different at simple DSBs, although this seems unlikely. Almost certainly some underestimation of the true NHEJ frequency is due to limited recleavage, although this is minimized by the understanding that a colony with any NHEJ growth should be counted as a single NHEJ event. Indeed, the frequency of NHEJ detected using this system is considerably higher than appreciated in a great many early experiments, demonstrating that NHEJ is in fact a substantial contributor to chromosomal DSBR in budding yeast.

The competitive end-binding model of DSBR pathway utilization:
Using the above systems, as well as ionizing radiation sensitivity, we have tested specific hypotheses regarding the basis for DSBR pathway utilization. Recent findings that Rad52 possesses DNA-end-binding activity suggest that it is the relative engagement of DNA ends by Rad52 or Ku proteins that represents the critical committed step (VAN DYCK et al. 1999 ). A prediction of this model is that in the absence of these proteins more breaks should be repaired by the remaining pathway. Results from neither the plasmid nor the suicide deletion assays support this. Although the relative contribution of NHEJ and SSA was predictably increased by mutation of RAD52 and YKU70, respectively, the absolute level of neither pathway ever increased above wild type. Further, preventing SSA by removing the required homologous repeat sequence did not have any impact on the frequency of repair by NHEJ. This was evident in the plasmid assay, although confounded by the inferred problem of multiple plasmid uptake, and more convincingly in the suicide deletion assay. Thus, the commitment of a majority of DSB ends to recombination in yeast occurs independently of the ability to complete repair by that pathway. The implication is that the processes that expose the 3' strand and execute homology searching are irreversible. Indeed, this is thought to occur via 5' resection at DSB ends in a fashion that does not require, and is in fact delayed by, Rad52 (SUGAWARA and HABER 1992 ). It is this strand loss more than any end binding event that would be expected to eliminate the possibility for NHEJ by preventing Ku binding and end alignment (see Fig 1). This can be seen as the Achilles' heel of recombination; testing for the presence of a homologous donor duplex can be achieved only after irreversibly altering the DNA ends.

In the plasmid assay, it is impossible to say with certainty that the yku70 mutation did not increase SSA because of the relatively low contribution of NHEJ and the large standard error in the assay. In the suicide deletion assay, however, the mutation of YKU70 and RAD52 not only did not enhance the remaining pathway, but in fact led to an unexpectedly mild but reproducible reduction. The yku70 effect was surprising because this mutant shows increased rates of 5' resection (LEE et al. 1998 ). Although the biochemical basis of this effect is uncertain, it is consistent with the view that end binding by Ku acts as a barrier to 5' resection and thus recombination. Indeed, two recent studies explored this Ku/recombination relationship in mammalian cells and similarly concluded that Ku has an inhibitory effect on homology-directed repair (FUKUSHIMA et al. 2001 ; PIERCE et al. 2001 ), the opposite effect of what we see here. In addition to species differences, it is possible that our failure to detect a yku70-stimulated increase in SSA is technical, since increased 5' resection might paradoxically disadvantage the successful completion of SSA via short terminal repeats. Indeed, CLIKEMAN et al. 2001 did observe a slight but reproducible increase in recombination via longer recessed direct repeats in yku70 yeast. Regarding the mild loss of NHEJ that we observed in rad52 cells, we note that HEGDE and KLEIN 2000 have also observed a minor impairment of precise NHEJ of transformed plasmids in sgs1 and rad52 mutant cells. It is thus possible that these recombination proteins actually facilitate Ku-dependent NHEJ by an unknown mechanism. In the end, it is clear that interactions between the DSBR pathways require further investigation and that at present a simple competition model does not seem adequate to account for DSBR pathway utilization.

The cell-state-regulated model of pathway utilization:
Both recombination and NHEJ have assets and limitations. Although recombination is potentially highly accurate, the correct homologous donor must be readily available to prevent homeologous repair. In contrast, NHEJ uses less information in the repair of DSBs, being limited to that present in overhanging ends, and so has a higher potential to create small insertions and deletions. NHEJ is always feasible, however, since the only required substrate is the DSB itself. Given these considerations, it is predictable that cells should regulate the usage of each pathway so that they are active when they are most likely to promote accurate repair. Factors that favor recombination include a small nonrepetitive genome and the late S, G2, and M stages of the cell cycle when the sister chromatid is physically associated with the broken chromosome. In contrast, NHEJ may in fact become less prone to error in nuclei of G1 or nondividing cells with large and repetitive genomes where a homolog is not readily available. These predictions are borne out by the results presented here, as well as by others. It was necessary to go to great lengths to observe high fractional contributions of NHEJ in yeast, whereas higher eukaryotic cells are known to preferentially use NHEJ in repair of radiation-induced and other DSBs, at least in G1 (LEE et al. 1997 ; TAKATA et al. 1998 ). The factors that did increase NHEJ are a logical extension of the observation that NHEJ is activated in MATa or MAT (i.e., haploid) cells, as opposed to MATa/MAT (i.e., diploid cells; ASTROM et al. 1999 ; LEE et al. 1999 ; and Fig 2). The reason that NHEJ is activated in haploid cells is almost certainly that these cells will find themselves without a homologous donor chromosome. Our observation that NHEJ was periodically upregulated in G1 is entirely consistent with this, although it stands in contrast to observations that imprecise NHEJ of HO-mediated breaks occurred at a lower efficiency when expression of the enzyme was restricted to G1 (MOORE and HABER 1996 ). The explanation for this difference may be that we have studied the more efficient pathway of simple religation instead of imprecise NHEJ, which in turn suggests that the polymerases and/or nucleases involved in imprecise NHEJ may themselves show a cell-cycle dependence.

Most strikingly, the apparent efficiency of NHEJ was consistently induced severalfold when cells left the exponential phase of fermentative growth and made the transition to respiratory metabolism and stationary phase. This finding requires careful scrutiny as it initially seems at odds with studies that have seen as much as 30% NHEJ of DSBs created by transient galactose-induced HO expression in dividing cells, where only a threefold further increase seems possible (LEE et al. 1999 ; FRANK-VAILLANT and MARCAND 2001 ). In addition to possible differences based on strain and the chromosomal site of the DSB, an important point is that these other studies used pregrowth in nonrepressing carbon sources prior to HO induction, which we found itself stimulates NHEJ. The plasmid assay uses cells grown and assayed in glucose medium and so is free from this concern. In the suicide deletion assay we specifically compared late stage source cultures with exponential glucose source cultures and saw a difference in pathway utilization, but this experiment is confounded by the fact that even cells pregrown in glucose must relieve glucose repression in order to express I-SceI. The time between loss of glucose repression and DSB induction would clearly be reduced, however. Since the change in NHEJ efficiency associated with nutritional status develops slowly (Fig 5), this provides an explanation for why a growth stage effect on the repair ratio might still be observed. Another potential difference between our suicide deletion and the above-cited HO assays is the extent to which recleavage occurs prior to NHEJ event measurement. This parameter is very difficult to judge, but if the suicide deletion assay in fact suffered more recleavage it would be even more sensitive to changes in the NHEJ efficiency, and indeed the highest fraction of NHEJ we observed was still significantly <30%. Finally, it is clear that both of our site-specific repair assays depend on factors that might artifactually increase the apparent NHEJ frequency. Specifically, if the successful entry of undegraded plasmids into the nucleus and the rate of loss of the I-SceI gene and enzyme are both dependent on growth stage, this could account for our experimental results independently of a true change in the DSBR repair mechanisms. We think that the consistency of the observation across three independent assays, including the inability to detect an incremental effect of NHEJ deficiency on radiation sensitivity in exponential cells (Fig 4 and HERRMANN et al. 1998 ), argues against the likelihood of experimental artifact, but this possibility cannot be ruled out.

If the observed increase in NHEJ event recovery were to reflect a true increase in NHEJ efficiency it should be biologically meaningful. By one logic, the need for induction of NHEJ should be the same for both G0 and G1 cells since each have only one copy of the chromosomes. On the other hand, there are clear differences in the population dynamics of dividing and nondividing cultures. In dividing cultures, not only are cells less likely to suffer damage in the absence of a homologous donor, given that a minority of cells are in G1/early S, but also loss of G1 cells that suffer infrequent DSBs would have only a small impact on the population growth. In contrast, G0 cells are at much greater risk of suffering damage irreparable by homologous recombination simply because they must persist in this state for a potentially prolonged period of nutritional deprivation. An increased likelihood of breakage may further be a consequence of the same environmental factors that led to nutritional deprivation in the first place. Moreover, since all cells in the environment will also be stationary, the population need for an appropriate G0 repair response is greater. It is thus conceivable that an adaptive pressure existed to promote evolution of more efficient NHEJ in stationary phase independently of mating-type regulation. In this view, increased NHEJ would simply be part of the stress responses induced by nutritional deprivation (WERNER-WASHBURNE et al. 1993 ). Importantly, a similar conclusion was recently reached by FERREIRA and COOPER 2001 who showed that telomere bridge-fusion-breakage cycles initiated by NHEJ of uncapped telomeres in Schizosaccharomyces pombe occurred only when cells were put under nutrient stress to induce stationary phase.

In total, we believe that current evidence accumulated from several different experimental approaches here and elsewhere suggests an activation of NHEJ based on nutritional status that is consistent with the biology of the stationary phase, although the technical difficulties of examining DSBR efficiency in stationary phase make this conclusion preliminary. Clarification of this issue will require further elucidation of the nature of the regulatory proteins affecting NHEJ and their targets. One critical question will be to examine whether Nej1 is responsible for the G0 effect as it is for the mating-type effect, which would predict that NEJ1 should be responsive to nutritional status. Preliminary data from a comprehensive genetic screen executed using the suicide deletion approach argues against this, however, because mutants have been identified that fail stationary-phase induction of NHEJ and impair NHEJ additively with nej1, further lending support to the notion that mating type and growth status represent separable regulatory inputs to NHEJ (T. E. WILSON, unpublished results).

	ACKNOWLEDGMENTS

We thank Shakuntala Fathepure for technical assistance in the early phases of this work. This work was supported in part by the Pew Scholars Program in the Biomedical Sciences of the Pew Charitable Trusts and Public Health Service grant CA-90911 (T.E.W.).

Manuscript received February 21, 2002; Accepted for publication April 18, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ARAVIND, L. and E. V. KOONIN, 2001  Prokaryotic homologs of the eukaryotic DNA-end-binding protein Ku, novel domains in the Ku protein and prediction of a prokaryotic double-strand break repair system. Genome Res. 11:1365-1374.[Abstract/Free Full Text]

ASTROM, S. U., S. M. OKAMURA, and J. RINE, 1999  Yeast cell-type regulation of DNA repair. Nature 397:310.[Medline]

BRACHMANN, C. B., A. DAVIES, G. J. COST, E. CAPUTO, and J. LI et al., 1998  Designer deletion strains derived from Saccharomyces cerevisiae S288C: a useful set of strains and plasmids for PCR-mediated gene disruption and other applications. Yeast 14:115-132.[Medline]

CLIKEMAN, J. A., G. J. KHALSA, S. L. BARTON, and J. A. NICKOLOFF, 2001  Homologous recombinational repair of double-strand breaks in yeast is enhanced by MAT heterozygosity through yKU-dependent and -independent mechanisms. Genetics 157:579-589.[Abstract/Free Full Text]

DE JAGER, M., J. VAN NOORT, D. C. VAN GENT, C. DEKKER, and R. KANAAR et al., 2001  Human Rad50/Mre11 is a flexible complex that can tether DNA ends. Mol. Cell 8:1129-1135.[Medline]

DOHERTY, A. J., S. P. JACKSON, and G. R. WELLER, 2001  Identification of bacterial homologues of the Ku DNA repair proteins. FEBS Lett. 500:186-188.[Medline]

FAIRHEAD, C., B. LLORENTE, F. DENIS, M. SOLER, and B. DUJON, 1996  New vectors for combinatorial deletions in yeast chromosomes and for gap-repair cloning using ‘split-marker’ recombination. Yeast 12:1439-1457.[Medline]

FERREIRA, M. G. and J. P. COOPER, 2001  The fission yeast Taz1 protein protects chromosomes from Ku-dependent end-to-end fusions. Mol. Cell 7:55-63.[Medline]

FRANK-VAILLANT, M. and S. MARCAND, 2001  NHEJ regulation by mating type is exercised through a novel protein, Lif2p, essential to the ligase IV pathway. Genes Dev. 15:3005-3012.[Abstract/Free Full Text]

FUKUSHIMA, T., M. TAKATA, C. MORRISON, R. ARAKI, and A. FUJIMORI et al., 2001  Genetic analysis of the DNA-dependent protein kinase reveals an inhibitory role of Ku in late S-G2 phase DNA double-strand break repair. J. Biol. Chem. 276:44413-44418.[Abstract/Free Full Text]

HEGDE, V. and H. KLEIN, 2000  Requirement for the SRS2 DNA helicase gene in non-homologous end joining in yeast. Nucleic Acids Res. 28:2779-2783.[Abstract/Free Full Text]

HERRMANN, G., T. LINDAHL, and P. SCHAR, 1998  Saccharomyces cerevisiae LIF1: a function involved in DNA double-strand break repair related to mammalian XRCC4. EMBO J. 17:4188-4198.[Medline]

IVANOV, E. L., N. SUGAWARA, J. FISHMAN-LOBELL, and J. E. HABER, 1996  Genetic requirements for the single-strand annealing pathway of double-strand break repair in Saccharomyces cerevisiae.. Genetics 142:693-704.[Abstract]

JACKSON, S. P., 2001  Detecting, signalling and repairing DNA double-strand breaks. Biochem. Soc. Trans. 29:655-661.[Medline]

JOHNSTON, M. and R. W. DAVIS, 1984  Sequences that regulate the divergent GAL1-GAL10 promoter in Saccharomyces cerevisiae.. Mol. Cell. Biol. 4:1440-1448.[Abstract/Free Full Text]

JONES, J. M., M. GELLERT, and W. YANG, 2001  A Ku bridge over broken DNA. Structure 9:881-884.[Medline]

KEGEL, A., J. O. SJOSTRAND, and S. U. ASTROM, 2001  Nej1p, a cell type-specific regulator of nonhomologous end joining in yeast. Curr. Biol. 11:1611-1617.[Medline]

KRAUS, E., W. Y. LEUNG, and J. E. HABER, 2001  Break-induced replication: a review and an example in budding yeast. Proc. Natl. Acad. Sci. USA 98:8255-8262.[Abstract/Free Full Text]

LEE, S. E., R. A. MITCHELL, A. CHENG, and E. A. HENDRICKSON, 1997  Evidence for DNA-PK-dependent and -independent DNA double-strand break repair pathways in mammalian cells as a function of the cell cycle. Mol. Cell. Biol. 17:1425-1433.[Abstract]

LEE, S. E., J. K. MOORE, A. HOLMES, K. UMEZU, and R. D. KOLODNER et al., 1998  Saccharomyces Ku70, Mre11/Rad50 and RPA proteins regulate adaptation to G2/M arrest after DNA damage. Cell 94:399-409.[Medline]

LEE, S. E., F. PACQUES, J. SYLVAN, and J. E. HABER, 1999  Role of yeast SIR genes and mating type in directing DNA double-strand breaks to homologous and non-homologous repair paths. Curr. Biol. 9:767-770.[Medline]

MOORE, J. K. and J. E. HABER, 1996  Cell cycle and genetic requirements of two pathways of nonhomologous end-joining repair of double-strand breaks in Saccharomyces cerevisiae.. Mol. Cell. Biol. 16:2164-2173.[Abstract]

OOI, S. L., D. D. SHOEMAKER, and J. D. BOEKE, 2001  A DNA microarray-based genetic screen for nonhomologous end-joining mutants in Saccharomyces cerevisiae.. Science 294:2552-2556.[Abstract/Free Full Text]

PAQUES, F. and J. E. HABER, 1999  Multiple pathways of recombination induced by double-strand breaks in Saccharomyces cerevisiae.. Microbiol. Mol. Biol. Rev. 63:349-404.[Abstract/Free Full Text]

PETRINI, J. H., 1999  The mammalian Mre11-Rad50-Nbs1 protein complex: integration of functions in the cellular DNA-damage response. Am. J. Hum. Genet. 64:1264-1269.[Medline]

PIERCE, A. J., P. HU, M. HAN, N. ELLIS, and M. JASIN, 2001  Ku DNA end-binding protein modulates homologous repair of double-s