Regulation of Ku-DNA Association by Yku70 C-terminal Tail and SUMO Modification*

Background: The Ku70-Ku80 ring encloses DNA ends to enable telomere protection and DNA repair. Results: Conditional sumoylation of Ku70 is reduced by mutating C-terminal lysines and this mutant shows decreased DNA interaction, shortened telomeres, and altered DNA repair. Conclusion: Our analyses suggest that sumoylation modulates Ku function by enhancing its DNA association. Significance: Ku is regulated by Ku70 C-terminal lysines and SUMO modification. The Ku70-Ku80 ring complex encloses DNA ends to facilitate telomere maintenance and DNA break repair. Many studies focus on the ring-forming regions of subunits Ku70 and Ku80. Less is known about the Ku70 C-terminal tail, which lies outside the ring. Our results suggest that this region is responsible for dynamic sumoylation of Yku70 upon DNA association in budding yeast. Mutating a cluster of five lysines in this region largely eliminates Yku70 sumoylation. Chromatin immunoprecipitation analyses show that yku70 mutants with these lysines replaced by arginines exhibit reduced Ku-DNA association at both telomeres and internal DNA breaks. Consistent with this physical evidence, Yku70 sumoylation deficiency is associated with impaired ability to block DNA end resection and suppression of multiple defects caused by inefficient resection. Correlating with these, yku70 mutants with reduced sumoylation levels exhibit shorter telomeres, increased G overhang levels, and altered levels of non-homologous end joining. We also show that diminution of sumoylation does not affect Yku70 protein levels or its interactions with protein and RNA partners. These results suggest a model whereby Yku70 sumoylation upon DNA association strengthens Ku-DNA interaction to promote multiple functions of Ku.

The Ku70-Ku80 ring complex encloses DNA ends to facilitate telomere maintenance and DNA break repair. Many studies focus on the ring-forming regions of subunits Ku70 and Ku80. Less is known about the Ku70 C-terminal tail, which lies outside the ring. Our results suggest that this region is responsible for dynamic sumoylation of Yku70 upon DNA association in budding yeast. Mutating a cluster of five lysines in this region largely eliminates Yku70 sumoylation. Chromatin immunoprecipitation analyses show that yku70 mutants with these lysines replaced by arginines exhibit reduced Ku-DNA association at both telomeres and internal DNA breaks. Consistent with this physical evidence, Yku70 sumoylation deficiency is associated with impaired ability to block DNA end resection and suppression of multiple defects caused by inefficient resection. Correlating with these, yku70 mutants with reduced sumoylation levels exhibit shorter telomeres, increased G overhang levels, and altered levels of non-homologous end joining. We also show that diminution of sumoylation does not affect Yku70 protein levels or its interactions with protein and RNA partners. These results suggest a model whereby Yku70 sumoylation upon DNA association strengthens Ku-DNA interaction to promote multiple functions of Ku.
The evolutionarily conserved Ku heterodimer, composed of Ku70 and Ku80, forms a ring structure that can bind DNA ends with high affinity and no sequence specificity (1). In Saccharomyces cerevisiae, Ku acts in both telomere maintenance and DNA double strand break (DSB) 4 repair. Upon association with DNA ends in either situation, Ku inhibits the accessibility and activity of nucleases (2)(3)(4)(5)(6)(7). Consequently, Ku blocks end resection and recombination, resulting in end stabilization at telomeres and potentiation of non-homologous end joining (NHEJ) repair at internal breaks (2)(3)(4)(5)(6)(7). Moreover, the Ku80 subunit interacts with partners such as the telomerase RNA (TLC1) and the Sir4 protein to provide additional functions required in telomere maintenance and DNA repair (8 -12).
How Ku dynamically interacts with diverse biomolecules and carries out different roles is complex and involves multiple levels of regulation, only some of which have been revealed. For example, Ku uses different domains to engage in different interactions. Some of these interactions are cooperative, as exemplified by the facilitation of Ku-DNA end association via Ku80 interaction with other NHEJ factors (10,13). Others are competitive, such as those between Ku80 and TLC1 versus Ku and DNA ends (12). There is also a division of labor between the two sides of the ring structure of Ku, with the one facing internal DNA acting in telomere maintenance and the one facing DNA ends in NHEJ (14). In addition to these mechanisms, Ku is subject to multiple forms of post-translational modification. This could pose yet another layer of regulation and is exemplified by mammalian Ku80 ubiquitination, which regulates the removal of Ku from DNA ends likely after DNA repair (15). The functional consequences for most of the other modifications of Ku are unclear.
Ku70 is sumoylated in budding yeast and humans (16,17). Sumoylation occurs via covalently linking SUMO to lysines of target proteins by SUMO E2 with the help of E3 (ligase) proteins (18). Sumoylation is known to affect substrate properties in a number of ways; however, the effects of sumoylation on Ku70 have not been well elucidated. Here, we show that in budding yeast, Ku70 sumoylation is largely eliminated by mutating a cluster of five lysines on the most C-terminal end of the protein, a region located outside of the ring structure (19).
Although not as extensively studied as the ring structure, the Ku70 C-terminal region is involved in DNA binding, exhibits a large shift upon DNA association, and contacts DNA in the front of the ring structure facing the internal side of DNA (19 -22). These properties suggest that the Ku70 C-terminal tail can regulate Ku migration on DNA, such as by influencing the inward movement of the Ku ring on DNA. This type of regulation likely acts in conjunction with other forms of regulation that also affect Ku-DNA association, including the interaction between the Ku ring structure with DNA and with other proteins. Because many functions of Ku require it to be at DNA ends, the C-terminal tail and its modification can conceivably make broad contributions to the functions of Ku.
We show that mutating a cluster of C-terminal lysines greatly reduces Yku70 sumoylation, without affecting Yku70 protein levels, interactions with several binding partners, and other general properties. Our genetic and physical analyses show that this mutant reduces DNA association and affects both telomere length and DNA repair. Our data further suggest that Yku70 sumoylation occurs after the Ku complex is bound to DNA. These observations support a model in which conditional sumoylation allows better retention of Ku at DNA ends, consequently facilitating functions of Ku at telomeres and DSB sites. These findings highlight the regulatory feature of the most C-terminal region of Ku70.

EXPERIMENTAL PROCEDURES
Strains, Yeast Techniques, and Statistics-Strains are listed in Table 1. Standard yeast protocols were used for strain construction, medium preparation, cell growth, and spot assays. To construct yku70 mutant strains, DNA fragments containing the mutations, an epitope tag, and a selection marker were generated by PCR. The fragments were then transformed into YKU70 cells and colonies containing the correct gene replacements were identified by PCR and sequencing. We note that yku70-RLQL, in which residues 599 -602 are mutated, also contains a K438R mutation generated by PCR error. Student's t tests were used for statistical analysis.  -1 can1-100 ura3-1 his3-11,15 leu2-3112 trp1-1 rad5-535). A single representative of each genotype is listed. The TAF tag is composed of the protein A (ProA) and FLAG modules (23). TAF constructs in the figures were labeled with ProA to indicate that only the ProA feature of the module was used. Plasmids pAB531 and pAB537 were described in Ref. 11.

Strain
Genotype Source NHEJ and DSB Resection Assays-Plasmid ligation assays or canonical NHEJ assays were performed similarly to described previously (23). Uncut and BamHI digested CEN plasmid pRS415 was purified, quantified, and transformed into yeast. Transformants were selected on synthetic media lacking leucine. The relative efficiency of plasmid ligation was calculated as the number of transformants obtained from the cut plasmids divided by those from the uncut plasmids. Three spore isolates of the same genotype were tested at the same time to calculate the average and S.D. Imprecise NHEJ assays were performed, and survival rates were calculated as described previously (24). The yku80 mutants were analyzed for DNA binding as described (11). For these experiments, strains T1367-4 and T1368-6 were transformed with CEN-LEU2 plasmids pVL1076, pVL1054 or pVL1860 bearing the yku80-3, yku80-4, or yku80-8 mutant alleles, respectively, or the empty vector control (25). DSB resection assays were performed as described previously (26).
Detection of Sumoylated Proteins and Protein Detection-Preparation of cell lysates under denaturing conditions, immunoprecipitation of tagged proteins, and detection of the sumoylated and unmodified forms of the proteins have been described (27). Western blotting was performed according to standard procedures using the following primary antibodies: anti-Myc (9E10; MSKCC Monoclonal Antibody Core Facility), anti-ProA (Sigma), and anti-SUMO (16). We note that the gel migration patterns of sumoylated Yku70 can differ due to different gel conditions and tags. Sumoylated Yku70 appears to be of low abundance and is not seen under short exposure using anti-tag antibodies but can be readily detected by anti-SUMO antibody. Standard methods for detecting protein levels in crude cell extracts and protein interactions in co-immunoprecipitation (co-IP) were used.
Telomere Length and G-overhang Measurements-Telomere length was measured by Southern blotting as described (27), using XhoI and PstI to digest the genomic DNA and a C 1-3 A/ TG 1-3 telomere DNA probe to detect telomeric repeats. G-overhang measurement by native gel was performed as described (27). Briefly, genomic DNA was digested with XhoI and run on 0.8% agarose gels. Gels were dried and hybridized to a CA oligonucleotide probe, and G-overhang signals were detected by exposing the hybridized gels to a PhosphorImager screen. The same gels were subsequently denatured and hybridized to check loading.
Chromatin Immunoprecipitation (ChIP) at Telomeres-This assay was performed as described in Ref. 11. In brief, 50 ml log-phase cultures were treated with 1% formaldehyde for 5 min. Cell pellets were washed and treated with zymolyase for 60 min at 28°C. Spheroplasts were pelleted, resuspended in 600 l of lysis buffer, and disrupted by vortexing with glass beads. Samples were then sonicated and anti-FLAG-conjugated beads (Sigma) were added. After overnight incubation, beads were washed, cross-links were reversed, and DNA was purified, precipitated, and resuspended. The samples were then brought to a volume of 300 l in 0.4 N NaOH/10 mM EDTA, denatured at 100°C for 10 min before dot blotting onto Zeta-Probe GT membrane (Bio-Rad). The blot was probed with a telomere-specific randomly labeled ( 32 P) probe. After washing, the blot was exposed to PhosphorImager screen for quantitation.
ChIP at DSBs-This assay was performed as described previously (3). In brief, an aliquot of YP-raffinose grown yeast culture was fixed in formaldehyde (0 min). To the remaining culture, galactose was added to 2% final to induce HO expression. At 60 and 120 min, aliquots were fixed in formaldehyde (60 and 120 min). Cell pellets were processed as for telomere ChIP samples except that quantitative PCR was performed to measure the amount of DSB chromatin (AB1692 and AB1693 as primers) relative to PRE1 (AB1696 and AB1697 as primers), an uncut chromosomal locus using ⌬⌬C t method. The enrichment of DSB chromatin was normalized to the amount of PRE1 chromatin immunoprecipitated. Quantitative PCR assay of input DNA was used to ensure equal cleavage at the MAT locus.
TLC1 Co-IP, RNA Isolation, and Quantitative RT-PCR-This assay was performed as described previously (11). In brief, 50 ml cultures (A 600 ϭ 1.0) were lysed in 400 l of TMG (10 mM Tris-HCl, pH 8.0, 1 mM MgCl 2 , 10% (v/v) glycerol, 0.1 mM DTT, 0.1 mM EDTA) and 50 mM NaCl with silica beads. Fifty-two l of ␣-FLAG M2 magnetic beads (Sigma) were added to 4 mg of total protein in 500 l of TMG-50 ϩ 0.5% Tween 20 and rotated at 4°C overnight. Beads were washed once with TMG-50 ϩ 0.5% Tween 20, three times with TMG-300 mM NaCl ϩ 0.5% Tween 20 and once with TMG-50. The washed beads and 50 l of input samples were added to 150 l of proteinase K solution and incubated at 37°C for 30 min. RNA was then isolated and DNase-treated using the RNeasy mini kit (Qiagen). One l of input RNA and 11 l of IP RNA was used as a template for cDNA synthesis using the Flex cDNA synthesis kit (Quanta Biosciences) followed by quantitative PCR as described (11).

RESULTS
Genetic Requirements for Yku70 Sumoylation-We first verified the sumoylation of budding yeast Ku70, also called Yku70, by several means. For example, anti-SUMO antibody detected a number of Yku70 bands that depended on a functional SUMO E2 (Ubc9) and shifted accordingly when a larger His 6 and FLAG-tagged SUMO (HF-Smt3) was used ( Fig. 1, a and b). Consistent with previous findings (16), sumoylated forms of Yku70 comprised a small proportion of total protein levels, which raises the possibility that Yku70 sumoylation requires specific conditions.
To test this idea, we asked whether Yku70 sumoylation required Ku DNA end binding and heterodimer formation. Arg-456 of Yku70 is a conserved residue that corresponds to a residue of human Ku located inside the ring structure in close proximity with DNA (19). Accordingly, R456E mutation reduces Ku-DNA interaction by up to 100-fold without affecting Ku80 interaction (11). We found that this mutation greatly reduced Yku70 sumoylation, suggesting that Yku70 sumoylation requires DNA binding (Fig. 1c). We extended this observation by querying Yku70 sumoylation status upon removal of Yku80, which is important for Ku70-DNA interaction (11,19). Consistent with the above notion, Yku70 sumoylation was abolished in yku80⌬ cells (Fig. 1d). The results that Yku70 sumoylation requires both Yku80 and DNA binding suggest that Yku70 is likely sumoylated after the Ku heterodimer loads onto DNA.
Yku70 Sumoylation Requires Five Lysines within its C Terminus-To study the function of Yku70 sumoylation, we first determined the lysines responsible for this modification. Because each of the three SUMO E3s, namely Mms21, Siz1, and Siz2, partially contributed to Yku70 sumoylation ( Fig. 1e; (16,28)), the SUMO E2, which targets the lysines within the KXE ( is a bulky hydrophobic residue (29)) motif, likely determines the conjugation sites. Yku70 contains one SUMO E2 recognition site ( 587 IKEE 590 ), located 14 amino acids from the C-terminal end of the protein (Fig. 2a). When we mutated Lys-588 within this site to arginine (K588R), a small reduction of Yku70 sumoylation was detected, suggesting the presence of additional sites (Fig. 2b).
Lys-588 is adjacent to four lysines that may fit within two reverse sumoylation consensus motifs (Fig. 2a). We reasoned that these four lysines may also contribute to Yku70 sumoylation. The presence of multiple Yku70 sumoylation forms is consistent with this notion, although the ladder could also be caused by SUMO chains, which are formed by conjugation of additional SUMOs to a SUMO molecule. To distinguish between these possibilities, we used a variant of SUMO that lacks the ability to form chains due to mutations of all of its lysines to arginine (smt3-KR) (30). smt3-KR reduced the number of Yku70 sumoylation forms to five, suggesting that Yku70 contains a minimum of five SUMO conjugation sites (Fig. 2c).
We then constructed a series of combinatorial mutations of four lysines surrounding Lys-588. Simultaneous mutation of five lysines (Lys-588, Lys-591, Lys-592, Lys-596, and Lys-597) to arginine on the chromosomal locus led to a dramatic reduction in Yku70 sumoylation (Fig. 2d). Using mutant alleles containing various combinations of Lys to Arg mutations at these sites, we determined that each of these five lysines contributes to sumoylation. Representative examples from this test are shown in Fig. 2d. For example, Lys-591 contributes to Yku70 sumoylation, as mutating this site further decreased sumoylation levels in a quadruple mutation background (Fig. 2d). Similar observations were made for the other four lysines (Fig. 2d and data not shown). We refer to the allele with the quintuple mutations (K588R, K591R, K592R, K596R, and K597R) as yku70-sd (SUMO defective). Upon obtaining the mutant that eliminates the bulk of Yku70 sumoylation, we used an array of molecular, functional, and genetic assays to examine how reducing sumoylation affects Ku-mediated processes, including telomere length regulation, DNA repair, and DNA end processing.
yku70-sd Cells Exhibit Shorter Telomeres and Increased Levels of G Overhangs-We first examined how yku70-sd affected telomeres. In budding yeast, telomeric sequences are composed of ϳ300 base pairs of TG 1-3 /CA 1-3 repeats followed by a short single-stranded 3Ј extension of TG 1-3 sequence, called the G-overhang (31). Ku is important both for telomere length maintenance and protection of the C 1-3 A 5Ј strand; yku70⌬ results in ϳ180 bp shortening of telomeres and a large increase in the amount of G-overhang due to increased 5Ј end resection, an effect more obvious at 37°C (Fig. 3, a and b) (6,32). We found that telomeres in yku70-sd cells were ϳ60 bp shorter than wild-type telomeres (Fig. 3a). In addition, yku70-sd cells exhibited more G-overhang DNA at both 30 and 37°C than wild-type cells (Fig. 3b). Both defects were not as severe as those in cells lacking the Ku complex (Fig. 3, a and b); thus, yku70-sd reduces but does not abolish Ku70 function at telomere ends.
To test the biological significance of the observed alterations at telomeres by yku70-sd, we examined whether this mutation affects the viability of cells lacking the telomerase catalytic subunit Est2. It was shown that yku70⌬ resulted in inviability when combined with est2⌬ at 30°C (33). We found that although yku70-sd and est2⌬ single mutants were viable at the second streak out from spore clones, the yku70-sd est2⌬ double mutant was inviable (Fig. 3c). These results show that the effects of yku70-sd on telomeres are deleterious when telomerase function is perturbed.
yku70-sd Alters the Levels of Plasmid Religation, but Does Not Affect Interactions with Yku80 and Two Other Binding Partners-We also examined how yku70-sd affects NHEJ repair. In a standard canonical NHEJ test, BamHI-digested plasmid DNA was purified and transformed into wild-type, yku70-sd, and yku70⌬ cells. Religation of the plasmid depends on canonical NHEJ and was largely abolished in yku70⌬ cells (Fig. 4a). yku70-sd exhibited ϳ70% of the level of ligation compared with the wild-type control (Fig. 4a, p Ͻ 0.01). This result suggests that yku70-sd reduces canonical NHEJ efficiency. Consistent with a role in DSB repair, sumoylation of Yku70 increased upon DSB induction by Zeocin (Fig. 4b). This extends our previous observation that Yku70 sumoylation is induced by treatment with replication blocking agents (27). The dual effects of yku70-sd on canonical NHEJ and at telomeres suggest that perhaps a function common to both processes is affected or some general protein properties are impaired. To examine these possibilities, we first evaluated Yku70 protein levels and interactions with known binding partners. We found that yku70-sd did not affect its protein levels (Fig. 4c) or Yku80 interaction, as assayed by co-IP (Fig. 4d). In addition, yku70-sd did not alter the levels of interaction with known binding partners, such as Sir4 and telomerase RNA, TLC1 (Fig. 4, e and f). Moreover, yku70-sd cells grew normally, even at 37°C when yku70⌬ is inviable (Fig. 3c and data not shown). Together, these results indicate that yku70-sd does not have a global or general affect on protein properties (see below).
We then considered whether Ku-DNA association, which is required for both telomere functions and DNA repair, is affected by decreasing Yku70 sumoylation. This possibility is consistent with the involvement of the YKu70 C-terminal tail in DNA interactions (19 -22).
yku70-sd Rescues Several Defects of sae2⌬ Cells-Ku-DNA end association can be tested in vivo by examining how yku70-sd affects sae2⌬ cells. The nucleases Sae2 and the Mre11-Rad50-Xrs2 complex carry out the initial DNA end resection at DSBs leading to recombinational repair (34,35). In their absence, Ku binding to DNA ends inhibits the access and activity of the Exo1 nuclease, which provides an alternative resection function for recombination (2)(3)(4)(5). As a result, removal of Ku in sae2⌬ mutants allows Exo1-mediated resection and repair and rescues sae2 DNA damage sensitivity (2)(3)(4)(5). If yku70-sd impairs DNA association, one would expect it could suppress sae2⌬ DNA damage sensitivity. Indeed, we found that yku70-sd, similar to yku70⌬, increased sae2⌬ cell resistance to radiation and the replication-blocking agent methyl methanesulfonate (Fig. 5a).
Removal of Ku also suppresses the synthetic sickness or lethality between sae2⌬ and deletion of DNA helicase Sgs1 and nuclease Rad27, likely due to removal of Exo1 inhibition (5). We confirmed that sae2⌬sgs1⌬ spore clones were either inviable or grew extremely slowly, and that sae2⌬rad27⌬ cells c  Fig. 1. . G-overhangs were detected by non-denaturing (native) in-gel hybridization to a telomeric CA probe and are indicated by the bracket. The same gel was denatured and hybridized to the CA probe as a control for total telomeric DNA (denature). c, yku70-sd and est2⌬ are synergistic. Spore clones from a diploid heterozygous for yku70-sd and est2⌬ were streaked out on YPD at 30°C.
were inviable (Fig. 5, b and c). yku70-sd rescued both defects: sae2⌬sgs1⌬yku70-sd spore clones were all viable and exhibited only a moderate growth defect, whereas sae2⌬rad27⌬yku70-sd spore clones were viable, albeit sick (Fig. 5, b and c). Together, these genetic results strongly suggest that yku70-sd is defective in DNA end protection. A simple interpretation of the above data is that reduction of sumoylation levels by yku70-sd leads to sae2⌬ suppression, although it is formally possible that the suppression is due to other alterations caused by the conservative amino acid changes in the Yku70 C-terminal tail. We addressed this issue by mutating C-terminal residues outside the region affecting sumoylation. We made conservative changes of three to four amino acids adjacent to this region ( 599 KFNI 602 to 599 RLQL 602 , and 600 FNI 602 to 600 LQL 602 ). These changes are similar to the K-R changes in terms of the degree of amino acid conservation, which can minimize disruption of the C-terminal tail and DNA interaction. As expected, neither mutation affected Yku70 sumoylation (Fig. 5d). Importantly, neither exhibited yku70-sd level of effects in sae2 suppression and telomere length (Fig. 5, a and e). These results are consistent with the idea that the yku70-sd phenotype is not simply due to changes in the C-terminal end of the protein independent of sumoylation.
yku70-sd Exhibits Increased Resection of DSBs in G 1 Phase-Suppression of sae2 defects suggests that yku70-sd is less effective in blocking DNA end resection. To directly test this idea, DNA end resection assays were performed in G 1 -arrested cells because influence of Ku on resection is most obvious in G 1 (2,3,5). Using a probe detecting a 0.7-kb fragment adjacent to the HO endonuclease cut site on chromosomal III, we found that resection in yku70-sd was faster than that in wild-type cells (Fig.  5f). This result is consistent with the sae2 suppression results and provides a plausible explanation for the suppression.
yku70-sd Alters the Levels of Imprecise NHEJ-Ku mutants with reduced DNA association but proficient for interaction with TLC1 and other proteins exhibit a unique phenotype that distinguishes them from general loss-of-function alleles (11). This was manifested in an imprecise NHEJ assay. When cells experience continuous HO-induced chromosomal breaks, survival depends on imprecise NHEJ, where limited resection or  Ϫsd. Yku70 and Yku70-sd tagged with ProA tag. Sir4 was tagged with Myc. Crude extract (CE) and immunoprecipitated (IP) products were examined by Western blots using antibodies that recognize ProA or Myc. f, Yku70sd does not affect association with telomerase RNA, TLC1 in RNA co-IP analysis (see "Experimental Procedures" for details). Quantification of three tests is depicted with average and S.E. is shown. Note that % of actin RNA precipitated is very low and not visible on the graph. The difference between Yku70 wild-type and Ϫsd mutant is not statistically significant (p Ͼ 0.1). a. a., amino acids.
other alterations occur at the break site before ligation (24). This is because cells lack homologous sequences for recombinational repair (hml⌬hmr⌬), and simple ligation of the ends would lead to the regeneration of the HO site; alteration at the break site before ligation precludes further HO-induced cleavage. Although Ku is required for imprecise NHEJ, DNA-binding mutants of Ku lead to increased survival (11). This is likely due to reduced Ku-DNA binding favoring limited resection thus imprecise NHEJ. We employed this assay to determine whether yku70-sd behaves similarly to other Ku mutants specifically defective in DNA binding. yku70-sd was examined both alone and in combination with yku80 alleles that show DNA binding defects (Fig.  6a). The use of yku80 mutants was necessary for detecting small effects of yku70 mutants with mild effects on DNA association (11). Indeed, yku70-sd exhibited increased survival in this assay, which was most obvious in the yku80 mutant background (Fig.  6a). Together, these results support the notion that yku70-sd does not cause general defects but specifically impairs DNA association.
yku70-sd Reduces Association with DNA at DSB Sites and Telomeric Sequences-Motivated by the above results, we directly measured Yku70-sd protein association at DSBs and telomeres. First, we used ChIP to examine protein association at HO nuclease-induced DSB breaks at the MAT locus on chromosome III. Following an established method (3), both wildtype and Yku70 sumoylation-deficient strains achieved close to complete HO cleavage. When using a primer pair located about 130 bp away from the HO site for quantitative PCR tests, yku70-sd yielded ϳ40% of wild-type level of recovery of this fragment at two time points post cleavage ( Fig. 6b; p Ͻ 0.01). This result indicates that Yku70-sd reduces binding at DSB sites.
Next, we used ChIP to examine Yku70-sd protein association at telomeric sequences, measuring telomeric DNA by dot blotting (11). We found that yku70-sd led to an ϳ15% reduction in protein association with these sequences (Fig. 6c, p Ͻ 0.01). These results are consistent with our genetic findings and provide direct evidence that Yku70-sd impairs DNA end association in vivo.

DISCUSSION
Ku is a central player in telomere metabolism and DNA repair and dynamically associates with DNA, telomerase RNA, and protein factors. Interaction of Ku with DNA ends supports several of its functions. Here, our data suggest that modification of Yku70 by SUMO could provide one means to promote Ku-DNA end association and contributes to the roles of Ku in both NHEJ and telomere maintenance.
Ku-DNA association is known to be regulated by other factors. The channel of Ku enclosing DNA contains loops that impede its diffusion on DNA (19). In addition, NHEJ proteins facilitate Ku association with DNA DSBs in vivo (10,13). In comparison, SUMO-based regulation targets a different domain of Ku and is dynamic.
Our results suggest that Ku70 sumoylation occurs after Ku binds to DNA ends (Fig. 1, c and d). We postulate that conformational changes in the Ku70 C-terminal region occurring upon DNA interaction (19,22) render it amenable for modification (Fig. 6d). Considering that the Ku70 C terminus is in front of the ring as it loads onto DNA (19,22), SUMO may serve as a door-stopper to retard Ku movement on DNA, either by steric hindrance or contacting DNA via the positively charged surface of SUMO (Fig. 6d). As reducing Yku70 sumoylation sustained protein levels and interactions with other binding partners as well as resulted in behavior similar to other DNA binding mutants, sumoylation most likely affects Ku-DNA association.
The generally mild phenotype of yku70-sd can be interpreted in several ways. For example, Yku80 sumoylation or residual Yku70 sumoylation may provide redundant functions. Also, sumoylation likely provides one of several means to promote or stabilize Ku-DNA association. As sumoylation appears to regulate some processes through collective, mild effects on multiple targets (26,27,36,37), future studies of additional sumoylation events on Ku and interacting proteins can provide a comprehensive picture of the role of sumoylation in Ku-mediated processes.
In terms of telomere metabolism, sumoylation appears to target multiple substrates and exert different effects depending on the experimental contexts (16,27,28,38,39). For example, defective sumoylation in budding yeast leads to longer or shorter telomeres in different genetic backgrounds, hinting at a complex role for sumoylation (16,27,28,39). Our mutagenesis analyses suggest that Yku70 sumoylation affects telomere length by enhancing DNA association (Fig. 3a). A previous study using LexA-Yku70-SUMO fusion suggests that SUMO facilitates telomere anchorage at the nuclear envelope (28). It is possible that multiple SUMOs on Yku70 have different roles.
Among the phenotypes examined, yku70-sd exerted the strongest effect on suppression of sae2⌬ related phenotypes   FIGURE 6. Yku70-sd protein is defective in localization to telomeres and DSB ends in vivo. a, yku70-sd increases imprecise NHEJ levels, most obviously when combined with yku80 mutants defective in DNA binding. Left, quantification of survival of continuous HO cleavage, after normalization to wild-type cells.
More than five independent trials were performed, and the difference was analyzed for statistical significance. Right, the indicated yku80 mutants were subjected to DNA binding assays as described (top; 11) and examined for Yku70 protein levels (middle) with loading control Pgk1 (bottom). b, Yku70-sd binding at DSBs is reduced as assayed by ChIP analysis. The diagram of HO cleavage site and primers used in ChIP is shown at the top. Quantification of several tests is depicted at the bottom with average and S.E. shown. The levels of immunoprecipitated DNA at DSBs were calibrated to those at an uncut site Pre1. Wild-type (YKU70) and sumoylation-defective Ku (yku70-sd) contain ProA tag. Untagged strains are also shown. c, Yku70-sd retention at telomeres is reduced. A representative dot blot and quantification of several trials of Yku70 and Yku70-sd association at telomeric regions are on the top and bottom, respectively. d, a model of how sumoylation regulates Ku functions.
( Fig. 5, a-c). As Yku70-sd protein exhibited reduced but still measurable DSB association (Fig. 6b), we postulate that the moderate DNA end association defect of Yku70-sd is sufficient to manifest the observed suppression when large numbers of endogenous or exogenous DNA lesions are present.
In summary, our data highlight the regulatory function of the C-terminal tail of Ku70 and suggest that sumoylation promotes protein-DNA end association. As two lysines identified in this study are conserved in human Ku70, this work may provide clues for investigating the roles of sumoylation of human Ku70.