Sequence-specific Binding of Ku Autoantigen to Single-stranded DNA*

Glucocorticoid-induced transcription of mouse mammary tumor virus is repressed by Ku antigen/DNA-dependent protein kinase (DNA-PK) through a DNA sequence element (NRE1) in the viral long terminal repeat. Nuclear factors binding to the separated single strands of NRE1 have been identified that may also be important for transcriptional regulation through this element. We report the separation of the upper-stranded NRE1 binding activity in Jurkat T cell nuclear extracts into two components. One component was identified as Ku antigen. The DNA sequence preference for Ku binding to single-stranded DNA closely paralleled the sequence requirements of Ku for double-stranded DNA. Recombinant Ku bound the single, upper strand of NRE1 with an affinity that was 3–4-fold lower than its affinity for double-stranded NRE1. Sequence-specific single-stranded Ku binding occurred rapidly (t½ on = 2.0 min) and was exceptionally stable, with an off rate of t½= 68 min. While Ku70 cross-linked to the upper strand of NRE1 when Ku was bound to double-stranded and single-stranded DNAs, the Ku80 subunit only cross-linked to single-stranded NRE1. Intriguingly, addition of Mg2+ and ATP, the cofactors required for Ku helicase activity, induced the cross-linking of Ku80 to a double-stranded NRE1-containing oligonucleotide, without completely unwinding the two strands.

Mouse mammary tumor virus is a slow transforming retrovirus that causes mammary tumors in lactating mice (1,2). Transcription of MMTV 1 is strongly induced by steroid hormones through a promoter proximal regulatory region of the LTR that has been characterized in great detail (3)(4)(5)(6)(7)(8)(9)(10). A second region, at the 5Ј end of the LTR, mediates tissue-specific expression and responsiveness to prolactin (11,12). Recently, it has been demonstrated that the region of the viral LTR between Ϫ420 and Ϫ360 contains sequences that act to repress viral transcription in several cell types, but most notably in T cells (13)(14)(15)(16)(17)(18)(19)(20). These sequences appear to be important for restricting cellular transformation by MMTV to the mammary gland, as viruses containing deletions encompassing this region of the LTR induce T cell lymphoma in addition to mammary carcinoma (19,(21)(22)(23)(24)(25)(26).
In preliminary mapping experiments, we identified a 14base pair polypurine/polypyrimidine DNA sequence element (NRE1) within the negative regulatory region of the LTR of the GR strain of MMTV that was sufficient to repress glucocorticoid hormone-induced MMTV transcription (20). Subsequently, we demonstrated that this sequence functioned as a direct, sequence-specific DNA binding site for Ku autoantigen/ DNA-dependent protein kinase (DNA-PK) (27). Both Ku and the DNA-PK catalytic subunit (DNA-PK cs ) were found to be required for the transcriptional effects of NRE1 on MMTV expression (27).
Ku (p70/p80) is an unusual DNA-binding protein that functions as both a DNA binding subunit and an allosteric activator of the DNA-PK cs (28). Ku/DNA-PK cs are predominantly nuclear proteins that are involved in multiple aspects of cellular homeostasis. In particular, Ku/DNA-PK are required for V(D)J recombination of immunoglobulin genes and the correct repair of double-stranded DNA breaks by the nonhomologous DNA break-repair pathway (29 -36). Additionally, Ku/DNA-PK have been implicated in the regulation of transcription by RNA polymerases I and II (27,37,38), DNA replication (39,40), and control of progression through the cell cycle (35,36).
The many activities of Ku appear to depend on its prolific and unique ability to interact with multiple forms of DNA. Prior to the identification of Ku as a sequence-specific DNA binding protein, it was well established that Ku was a DNA end-binding protein that also recognized DNA nicks and virtually any double to single-stranded transition in DNA including DNA loops and cruciform structures (28,(41)(42)(43)(44)(45). However, unstructured single-stranded DNA was only poorly recognized by Ku (44 -46). Sequence-specific binding to NRE1 is preferred to DNA end binding (47). Remarkably, Ku bound to doublestranded DNA also has the ability to translocate along DNA from its entry point at NRE1 or DNA ends (28,41,42). Translocation of Ku is a process that appears to be facilitated by Mg 2ϩ , but does not require energy (27,42). Additionally, Ku has also been identified to be both an ATPase and the human HDH II DNA helicase (46,48). This helicase activity appears to be somewhat limited, as Ku is reported to only be effective in unwinding linear double-stranded DNAs containing extended single-stranded overhangs adjacent to the double-stranded sequence (46).
Although Ku/DNA-PK cs appear to be required for NRE1-mediated transcriptional repression (27), nuclear factor recognition of NRE1 is complex. In addition to the double-stranded NRE1 binding activity of Ku, we have also observed nuclear factors that specifically recognized the upper and lower single strands of NRE1-containing DNAs (20,49). The identity of these factors and their role in NRE1-mediated transcriptional regulation is not known. The potential for these strand-specific factors to interact with NRE1 in vivo is particularly intriguing in light of the helicase activity of Ku (46) and our previous observation that nuclear factor binding to double-stranded NRE1 induces Mg 2ϩ -dependent structural transitions in MMTV LTR flanking sequences that are sensitive to the single strand-specific agents, KMnO 4 and S1 nuclease (50).
In the present study we have initiated a characterization of the nuclear factors that bind specifically to the polypurine rich upper strand of NRE1. A single-stranded oligonucleotide affinity column separated NRE1 binding activity into two fractions. One of the factors was purified to homogeneity. Remarkably, this sequence-specific, single-stranded NRE1 binding activity was revealed to be Ku. Both purified Ku and recombinant Ku expressed in insect cells from baculovirus vectors bound specifically and stably to the single, upper strand of NRE1. While the affinity of Ku for single-stranded NRE1 was slightly lower than its affinity for the double-stranded element, the DNAsequence requirements for binding to the two forms of DNA overlapped closely. Cross-linking experiments indicated that while only the 70-kDa subunit of Ku (Ku70) appeared to contact double-stranded NRE1, Ku binding to the upper, single strand of NRE1 also involved direct participation of the 80-kDa Ku subunit (Ku80) with the DNA. Intriguingly, the addition of Mg 2ϩ and/or ATP to a binding reaction containing doublestranded NRE1 induced contact of Ku80 with the upper strand of NRE1 without completely unwinding the oligonucleotide tested. The implication of these findings for the regulation of MMTV transcription are discussed.

MATERIALS AND METHODS
Tissue Culture and Antibodies-Jurkat T cells were cultured in RPMI 1640 (Life Technologies Inc.) supplemented with 10% fetal bovine serum and incubated at 37°C in 5% CO 2 . The Spodoptera frugiperda cell line Sf9 was grown in TMN-FH medium (Invitrogen) at 27°C. Anti-Ku antibody Ab162 (51) was generously provided by W. Reeves.
Preparation of NRE1 DNA Affinity Matrices-To prepare a singlestranded DNA oligonucleotide affinity column for the purification of proteins binding to the upper strand of NRE1 (upNRE1), an 80-mer oligonucleotide containing four copies of the upNRE1 sequence 5Ј-(ACT-GAGAAAGAGAAAGACGA) 4 -3Ј was synthesized on a Beckman Oligo 1000 DNA synthesizer. This oligonucleotide was coupled to cyanogen bromide-activated Sepharose CL-2B as described previously (52,53).
Purification of an Upper-stranded NRE1 Binding Protein-Nuclear extracts were prepared from Jurkat T cells as described by Dignam et al. (54), except that KCl was substituted for NaCl. 6 mg of crude Jurkat nuclear extract was loaded onto a 2-ml upMTV-Sepharose column equilibrated to 100 mM KCl in buffer A (20 mM HEPES, pH 7.9, 20% glycerol, 0.2 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol, 0.1% Nonidet P-40). The column was washed with 3 column volumes of buffer A in 0.1 M KCl, then eluted with a stepwise gradient of 3 volumes of 0.2, 0.4, 0.6, and 0.8 M KCl in buffer A.
Double-stranded NRE1 affinity chromatography was accomplished by combining and diluting the 0.2 M KCl fractions from the upNRE1 column in buffer A to 0.1 M KCl and loading the sample onto a 1-ml double-stranded sMTV affinity column equilibrated to 0.1 M KCl with buffer A. The column was washed with 3 column volumes of buffer A, then eluted with a stepwise gradient of 3 volumes of 0.3, 0.5, and 1.0 M KCl in buffer A. To analyze the protein content of individual fractions, samples were fractionated on 10% SDS-polyacrylamide gels, and the proteins were visualized by silver staining (Bio-Rad).
Electrophoretic Mobility Shift Analysis of Binding to NRE1-All oligonucleotides employed in electrophoretic mobility shift assays were labeled at their 5Ј ends with ␥-[ 32 P]ATP using T 4 polynucleotide kinase. The labeled oligonucleotides employed in these experiments included the double-stranded MTV oligonucleotide, the upper and lower single strands of the MTV. Single-stranded oligonucleotides encoding the upper strands of the NRE1 element in the C3H strain of MMTV, the c-Myc plasmacytoma repressor element, an NRE1-like transcriptional repressor in the HTLV LTR, an octamer motif and a heat shock response element, that have been previously described (47), were also employed in some assays.
Oligonucleotide affinity column fractions were assayed for NRE1 binding activity by electrophoretic mobility shift assays (EMSA) using the 23-base pair double-stranded MTV oligonucleotide or the 23 nucleotide single, upper strand of MTV. Both probes were labeled. Protein samples were incubated for 20 min at 20°C in Buffer B (0.6ϫ Buffer A containing 60 mM KCl and 2 g of bovine serum albumin), and 2-20 ng of labeled specific probe and 100 ng to 1 g of highly sheared calf thymus DNA. For experiments examining binding to single-stranded oligonucleotides, the calf thymus DNA was denatured by heating to 95°C for 5 min and rapidly cooled on ice immediately prior to the binding assays. Electrophoresis was performed on native 4% polyacrylamide gels in 0.5ϫ Tris-borate-EDTA for 1.5 h at 150 V. Following electrophoresis, the gels were dried and exposed on NEF496 film (Dupont). In some experiments, antibodies were preincubated with the protein samples for 30 min prior to the addition of DNA to the binding reactions.
All experiments examining the DNA binding of recombinant Ku were performed under the binding conditions described above. Determination of the K d of Ku binding to the upper strand of NRE1 was performed exactly as described previously for double-stranded Ku binding (47). EMSA was performed with a constant amount of recombinant Ku incubated with an increasing concentration of upMTV in a volume of 20 l. The concentration of upMTV oligonucleotide was determined by spectrophotometry. The concentration of bound upMTV was calculated from the fraction of total upMTV as determined by phosphorimage analysis. K d was determined by Scatchard analysis in three independent trials and is expressed as the mean Ϯ S.E.
Kinetic analysis of recombinant Ku binding to the upper strand of NRE1 was also performed essentially as described previously (49). The on rate was determined by EMSAs in which binding reactions were allowed to progress for 1 to 60 min prior to electrophoresis. The percentage of equilibrium binding was determined by phosphorimage analysis and plotted as mean Ϯ S.E. The half-time (t1 ⁄2 ) of initial binding was calculated using linear regression of the data in the linear range. The off rate was determined by EMSAs in which NRE1 binding was allowed to equilibrate for 30 min and was subsequently monitored over a 24-h period following the addition of 200 ng of unlabeled upMTV competitor DNA. The percentage of initial binding was determined by phosphorimage analysis and plotted as mean Ϯ S.E. Half-time (t1 ⁄2 ) of binding was calculated using logarithmic regression of the data.
Footprinting of Protein Binding to Single-stranded DNA-Two reagents, KMnO 4 and DNase I, were used to footprint protein-DNA interactions over the MMTV LTR. 5-10 ng of a fragment from Ϫ421 to Ϫ106 of the MMTV promoter radiolabeled on the upper strand or lower strands was mixed with 100 ng of calf thymus DNA, heat denatured at 95°C for 5 min, then rapidly cooled on ice. Binding was performed in buffer B with crude nuclear extracts and purified factors as described for 20 min at 20°C. Treatment with 4 mM KMnO 4 was for 1 min as described previously (20,49). Treatment with DNase I was performed for 1 min in the presence of 4 mM MgCl 2 . DNase I treatment was stopped by adding SDS to 0.1% and Na 2 EDTA to 10 mM. Footprints were visualized following electrophoresis of the samples on DNA sequence gels. Cleavage was positioned relative to guanine and thymidine sequencing reactions.
Production of Recombinant Ku-Sf9 cells were infected with baculovirus vectors encoding Ku-80 and histidine-tagged Ku-70 as described previously (55). Ku heterodimer was purified over nickel-nitrilotriacetic acid Sepharose ® 6B (Novagen) as described previously (47,55). As Ku-70 is only poorly soluble in Sf9 cells as a monomer, effectively only the Ku heterodimer is obtained. The purity of the recombinant preparation and the Ku-70/Ku-80 ratio was verified by silver staining of SDS-polyacrylamide gels as shown previously (47).
UV Cross-linking-The upper strand of NRE1 was labeled by extension of a 6-nucleotide primer (5Ј-AACTGA-3Ј) hybridized to the loMTV with Klenow fragment in the presence of [␣-32 P]dATP to produce a double-stranded MMTV oligo radiolabeled throughout NRE1, on the upper strand. Denaturation of the double-stranded fragment mixed with 1 g of calf thymus DNA in buffer A was accomplished by heating at 95°C for 5 min, followed by rapid cooling on ice. Binding was performed in buffer B under conditions identical to those employed for EMSA and used comparable amounts of Jurkat purified or recombinant Ku. 10 mM MgCl 2 and 4 mM ATP were added as indicated. Cross-linking was performed in a Stratalinker 1800 (Stratagene) for 12 min at 4°C. Protein DNA complexes were resolved on 10% SDS-polyacrylamide gels

Nuclear Factor Binding to the Separated Upper and Lower
Strands of the MMTV LTR-We have previously demonstrated in EMSA and in single-stranded DNA footprinting experiments with the thymidine specific reagent KMnO 4 , that nuclear factors in human Jurkat T Cell nuclear extracts bound to both the upper and lower single strands of the MMTV LTR over NRE1 (20). In the KMnO 4 footprinting, however, nuclear factor binding to the upper strand of NRE1 was reflected only by the protection of a single thymidine adjacent to the polypurine sequence.
To increase our understanding of the interaction of nuclear factors with the separated single strands of the MMTV LTR in and around NRE1, we examined nuclear factor binding by single-stranded DNase I protection footprinting ( Fig. 1). An LTR fragment extending from Ϫ421 to Ϫ364 was labeled at the 5Ј end of the upper strand or the 3Ј end of the lower strand. Strand separation was accomplished as we have previously described for single-stranded KMnO 4 footprinting (20). Under these conditions, the strands from the LTR fragment remain single-stranded for the duration of the binding and footprinting reactions.
DNase I is considered a double-stranded endonuclease. However, the majority of DNA contacts made by DNase I occur on the DNA strand that is cleaved and all contacts occur on the same face of the DNA (56 -58). Further, DNase I cleaves DNA asymmetrically on only one strand in the presence of Mg 2ϩ (59) and has been observed to cleave single-stranded DNA (60).
The DNase I cleavage patterns in the absence of nuclear extract were markedly different for the double-stranded and single-stranded MMTV LTR fragments (Fig. 1, A and B, lanes 2 and 4). In general, cleavage of the single-stranded DNAs by DNase I occurred at both fewer and different sites. For example, the region over NRE1 on the lower, single strand, of the MMTV DNA fragment was strikingly sensitive to DNase I when single-stranded, but poorly cleaved on double-stranded DNA.
Incubation of the lower single strand with 7 g of Jurkat nuclear extract prior to DNase I digestion led to complete protection of the strong DNase I cleavage sites over NRE1 (Fig.  1A, lane 5). Further, as the most sensitive sequences in the fragment were protected, a more extensive DNase I cleavage pattern over the rest of the DNA was observed. The protection on the lower, single-strand (Ϫ400/Ϫ375), corresponded approximately to the boundaries of protection that we had previously observed with Jurkat nuclear extract on double-stranded DNA (20).
DNase I footprinting of upper-stranded LTR binding (Fig.  1B, lane 5), showed a central core of protection between Ϫ405 and Ϫ384 that was bordered on each side by regions of partial protection. This was similar to, but broader than, the footprint observed on the upper strand of double-stranded LTR DNA (Ϫ402/Ϫ380) with Jurkat nuclear extract (20). Further, only one of the three Ts within the central core of DNase I protection on the upper, single strand were protected from modification with KMnO 4 (20).
Upper-stranded NRE1 Binding Activity Contains Two Components-To begin to characterize the single-stranded NRE1 binding factors, Jurkat nuclear extract was fractionated by oligonucleotide affinity chromatography. We prepared an upper strand NRE1 oligonucleotide affinity column by coupling a synthetic single-stranded 80-mer containing four copies of the upNRE1 sequence to cyanogen bromide-activated Sepharose. Previously, this coupling procedure has been used successfully to prepare a single-stranded oligonucleotide affinity column for the purification of polypyrimidine tract binding protein, which binds to a single-stranded polypyrimidine tracts of DNA (53).
Unexpectedly, passage of Jurkat T cell nuclear extract over the upNRE1 column separated the upNRE1 binding activity into two components (Fig. 2). The first component eluted at 0.2 M KCl ( Fig. 2A, lanes 7 and 8). The second component, which comprised approximately 80% of the total upper-stranded NRE1 binding activity and bound strongly to the upper strand of the NRE1-containing oligonucleotide upMTV in EMSAs, was unable to bind the upNRE1-Sepharose column prepared by cyanogen bromide coupling. The presence of NRE1 binding activity in the flow-through fractions was not due to overloading of the column, as when these fractions were pooled and passed through the column a second time, no additional up-MTV binding activity was retained. 2 Subsequently, we found that an upNRE1 affinity matrix, prepared by linking a biotinylated oligonucleotide to streptavidin agarose beads, retains this second component, 2 suggesting that modification of the bases along upNRE1 by cyanogen bromide interfered with the binding of this factor to the original column. Further, characterization of this factor will be presented elsewhere.
Silver staining of an SDS-polyacrylamide gel of the 0.2 M fraction showed that the upNRE1 column had efficiently purified the upper-stranded NRE1 binding activity into three major components of 70, 85, and 110 kDa (Fig. 2B).
Interestingly, the mobility of the 70-and 85-kDa species were similar to the SDS-polyacrylamide gel electrophoresis staining pattern of the Ku antigen that we had purified previously as a double-stranded NRE1 binding protein (27). This prompted us to determine how Ku had fractionated over the upNRE1 affinity column (Fig. 3). First, we determined that the 0.2 M KCl fraction contained almost all of the double-stranded NRE1 binding activity present in the crude extract, in addition to upper-stranded NRE1 binding activity (Fig. 3A, lane 3). In contrast, little double-stranded NRE1 binding activity was observed in the flow-through from the column (lane 2), and no double-stranded binding activity was detected in other column fractions. 2 Second, addition of the Ku-specific antibody 162 to binding reactions containing the upMTV single-stranded oligonucleotide retarded the mobility of the complex formed with the 0.2 M KCl fraction (Fig. 3B, lanes 3 and 4), but had no discernible effect on the mobility of the upMTV binding activity in the flow-through fraction (lanes 1 and 2). In contrast, addition of a monoclonal antibody specific for the glucocorticoid receptor had no effect on upMTV binding. 2 These data implicated Ku as the upper-stranded NRE1 binding component that eluted from the upNRE1 column in the 0.2 M KCl fraction. It also excluded Ku as the second upper-stranded NRE1 binding factor.
Subsequent passage of the 0.2 M KCl fraction over the double-stranded NRE1 oligonucleotide affinity column that we have described previously, purified the 70-and 80-kDa Ku heterodimer to near homogeneity (Fig. 3C). Thus, these experiments also demonstrate a simple, two-step, protocol for the purification of Ku antigen from Jurkat T cell crude nuclear extracts that we expect will also allow the rapid and efficient purification of Ku from other sources. In our hands this protocol yielded approximately 30 g of Ku from 40 liters of cultured Jurkat cells.
Purified Ku Protects NRE1 in Single-stranded DNA Footprinting Assays-As crude Jurkat nuclear extracts fractionated into two upper-stranded NRE1 binding activities it was important to determine the contribution of Ku to footprints on the upper strand of the MMTV LTR obtained with crude extracts. We used KMnO 4 and DNase I to examine the binding of purified Ku to the same single-stranded DNA fragment from the upper strand of the MMTV LTR described above in experiments with crude nuclear extracts (Fig. 4).
Incubation of this fragment with unfractionated Jurkat extract leads to the protection of the thymidine at Ϫ395 from KMnO 4 at the 5Ј end of NRE1 (20). The same residue was protected from KMnO 4 modification by the purified Ku prepa-   Fig. 2A (lane 3). B, EMSA of the binding to the upper strand of NRE1 oligonucleotide upMTV of the flow-through from the upNRE1 affinity column represented in lane 2 from Fig. 2A  (lanes 1 and 2), and the fraction eluting from the upNRE1 affinity column with 0.2 M KCl represented in lane 7 from Fig. 2A (lanes 3 and  4). Preincubation in the presence (ϩ) or absence (Ϫ) of anti-Ku antibody Ab162 is indicated above the lanes. C, silver-stained SDS-polyacrylamide gel of the fraction containing double-stranded NRE1 binding factors sequentially purified over upMTV and double-stranded MTV (dsMTV) oligonucleotide affinity columns. The migration of molecular mass standards (kDa) is shown to the left. ration (Fig. 4A). The use of purified Ku in these experiments, however, also resulted in protections that were not observed with the crude extract. First, the subsequent thymidine in the LTR, at Ϫ399, was more than 50% protected from KMnO 4 modification, extending the 5Ј boundary of protection with KMnO 4 closer to that obtained with DNase I. Second, under the binding conditions of this experiment, the Ts at the 5Ј end of the LTR fragment were also protected from KMnO 4 modification. Thus it appears that Ku may also have some preference for binding to single-stranded DNA ends, in addition to its well known propensity for binding to double-stranded DNA ends.
In experiments with DNase I (Fig. 4B), the protection of sequences extending 5Ј from NRE1 by the purified Ku preparation was very similar to that observed with crude extracts, with the boundaries of protection mapping within 1-2 nucleotides in the two instances. With DNase I, however, no protection over the 5Ј end of the fragment was observed at the concentration of Ku employed. One possibility was that Ku interacted with DNA ends in a way that was not readily detectable with DNase I. Alternatively, these results may reflect the preference of Ku for NRE1 over DNA-ends that is also seen with double-stranded DNAs (27,47). In contrast, the DNase I footprint with the purified fraction was extended in the 3Ј direction compared with that observed with the crude extract (compare with Fig. 1B). Notably strong protection was observed to the very end of the polypurine stretch at Ϫ381. In addition, partial protection extended an additional 20 nucleotides in the 3Ј direction to Ϫ357. Together, these experiments demonstrate that the purified Ku specifically protected sequences on the upper, single strand of the MMTV LTR centered over the polypurine NRE1 element.
Properties of the Sequence-specific Binding of Recombinant Ku to the Upper Strand of NRE1-To confirm that no factors in the purified preparation in addition to Ku participated in binding to the upper strand of NRE1, we examined the ability of recombinant Ku expressed in insect cells and purified to near homogeneity, to bind specifically to the upper-strand of NRE1 (Fig. 5A). In this EMSA experiment, the incubation of recombinant Ku with either the double or upMTV oligonucleotides in the presence of the appropriate nonspecific competitor DNAs resulted in single shifted complexes of equal mobility. Binding was specific to the upper strand of NRE1, as no complex was observed with the complementary lower single-stranded NRE1 oligonucleotide.
Several sequence-specific double-stranded Ku binding sites have been proposed based on the results of EMSA and DNA footprinting experiments with linear DNA fragments (37,45,(61)(62)(63)(64)(65)(66)(67)(68). Recently, we demonstrated that only the subset of sequences with a obvious homology to NRE1, appear to serve as direct double-stranded DNA binding sites for Ku on covalently closed circular DNA (27,47). To begin to probe the sequence requirements for sequence-specific single-stranded DNA binding by Ku, we compared the binding of recombinant Ku to the upper-strand of NRE1 from the LTR of GR MMTV strain to binding to the polypurine-rich strands of several of oligonucleotides that we have previously assessed for direct recognition by Ku in double-stranded DNA microcircles (Fig. 5, B and C).
The binding of Ku to single-stranded, linear oligonucleotides closely paralleled our previous results with double-stranded DNA microcircles. First, Ku displayed the highest affinity for the NRE1-containing oligonucleotide from the GR strain of MMTV (Fig. 5C, lane 1). However, the 3 oligonucleotides encoding polypurine-rich sequences similar to NRE1 were also bound by Ku, albeit with what appeared to be, over the course of several experiments, a consistently 2-3-fold decrease in affinity (lanes 2-4). This result suggested that the substitutions in these NRE1-like sequences had a small, but reproducible, effect on their ability to be recognized by Ku. Further, the results obtained for the C3H MMTV LTR NRE1 element exactly parallels the difference in binding that we previously reported for the binding of Ku to double-stranded NRE1-containing DNA microcircles (47). By contrast, single-stranded oligonucleotides containing two sequences that we have previously shown not to be directly recognized by Ku under our binding conditions when double-stranded (47), also failed to be bound appreciably by Ku in our single-stranded EMSA experiments ( lanes 5 and 6), even in the presence of a 5-fold higher concentration of recombinant Ku. 3 To determine the affinity of Ku binding to the single, upper strand, of NRE1 under our binding conditions, we used EMSA to perform Scatchard analyses of the binding of a constant amount of recombinant Ku to an increasing concentration of the upMTV oligonucleotide (Fig. 6). Averaging of three independent experiments yielded a K d of 3.5 Ϯ 1.3 nM for sequencespecific single-stranded Ku binding. This result indicates that the affinity of recombinant Ku for single-stranded NRE1 was approximately 4-fold lower than the 0.84 Ϯ 0.24 nM K d that we have previously reported for the direct binding of recombinant Ku to double-stranded NRE1 under the same binding conditions. This lower value for the affinity of Ku for single-stranded NRE1 is consistent with our observation that the DNA ends in highly sheared calf thymus DNA begins to compete Ku binding to upMTV at lower concentrations than it does double-stranded NRE1 binding. 3 This is also consistent with our early results with crude Jurkat T cell nuclear extracts where we reported that the double-stranded MTV oligonucleotide competed 3-5fold more effectively for double-stranded NRE1 binding than did the upMTV sequence (20). However, competition of upMTV binding by DNA ends still requires a greater than 100-fold molar excess of DNA ends. 3 Ku binding to DNA ends has previously been described to be extremely stable when compared with the norm for transcription factor binding to DNA sequences (42,69). Therefore, to complete our analysis of the single-stranded NRE1 binding properties of Ku we examined the kinetics with which recombinant Ku bound to the upNRE1 oligonucleotide (Fig. 7). The binding of Ku to upMTV occurred rapidly, with a t1 ⁄2 ϭ 2 min (Fig. 7, A and B). This is comparable to the fastest rates that have been reported for transcription factor DNA binding in vitro. In contrast, the off rate of Ku from upMTV following the equilibration of binding was unusually slow, with a t1 ⁄2 of 68 min (Fig. 7, C and D). These data suggest, that Ku can rapidly access the single, upper strand of NRE1. However, once binding has occurred, Ku may remain stably associated with the upper strand of NRE1 for an extended period of time.
Upper-stranded NRE1 Binding Promotes the Contact of Ku80 with DNA-In DNA cross-linking assays with crude nuclear extracts, performed both before and after EMSA, we have shown that distinct nuclear factors are UV cross-linked to the upper strand of NRE1 when it was presented as doublestranded DNA or as single-stranded DNA (20,50). A 45-kDa protein-DNA complex formed on the lower strand of NRE1, but only when the DNA was single-stranded (20,50). While a factor in crude nuclear extracts migrating at 80 kDa cross-linked to double-stranded NRE1, two factors migrating at 80 and 95 kDa cross-linked to upNRE1. Further, while the total amount of cross-linking obtained varied considerably between nuclear extracts prepared from different cell types, the ratio of the 80 and 95 kDa bands was the same in all instances (20). Ku binds both of these forms of NRE1 and the Ku70 and Ku80 subunits cross-linked to DNA would be expected to yield protein-DNA products with mobilities similar to the protein DNA products obtained with the crude extracts. Therefore, these results suggested the possibility that the two Ku subunits differentially To investigate the contact of Ku70 and Ku80 with doubleand single-stranded NRE1-containing oligonucleotides, we performed protein-DNA cross-linking experiments with the purified recombinant Ku (Fig. 8). First, under standard binding conditions in the absence of Mg 2ϩ and ATP, a single, 80-kDa DNA-protein complex marked the cross-linking of Ku70 to the double-stranded NRE1-containing oligonucleotide (lane 2). In contrast, a second complex of 90 -95 kDa indicated the crosslinking of Ku80 in addition to Ku70 to the upper strand of NRE1 (lane 1). Interestingly, upon addition of Mg 2ϩ , the higher mobility, 90 -95-kDa Ku80-DNA complex became detectable with double-stranded oligonucleotide (lane 3). In the presence of both Mg 2ϩ and ATP the ratio of Ku70/Ku80 cross-linking decreased to that observed on the upper, single-stranded NRE1 oligonucleotide (lane 4). Similar results were obtained with Ku purified from Jurkat cells. 3 Together, these results suggested that the binding of Ku to double-and single-stranded NRE1containing DNAs was mediated through distinct regions of the heterodimer. Further, they indicated that the presence of Mg 2ϩ /ATP changed the nature of the Ku-NRE1 interaction on double-stranded DNA.

DISCUSSION
Fractionation of a Jurkat nuclear extract over consecutive single-and double-stranded NRE1 oligonucleotide affinity columns has revealed Ku antigen to be one of two factors with the ability to bind specifically and with high affinity to the single upper strand of NRE1. Ku binding to single-stranded NRE1 was rapid and occurred with an affinity only slightly lower than binding to double-stranded NRE1. Further, once bound to the upper strand of NRE1, Ku remained stably associated with the DNA for long periods of time. Our results also suggest that Ku may bind with at least a small degree of selectivity to single-stranded DNA ends. These results add two new dimensions to the DNA binding activities of a protein that may be the most versatile DNA binding protein characterized to date. They have implications for the regulation of transcription through NRE1, understanding the molecular basis for the induction of structural transitions in NRE1 flanking DNA, and offer a potential explanation of the need for an extended single-stranded DNA extension for Ku helicase activity.
DNase I footprinting experiments with single-stranded DNA fragments demonstrated that nuclear factor binding to both the upper and lower strands of the MMTV LTR protected very similar regions centered approximately over the polypurine/ polypyrimidine NRE1 core. On both strands, the protection observed overlapped extensively with the protection from DNase I that occurs on double-stranded DNA. These results suggest it unlikely that the double-and single-stranded binding factors could simultaneously occupy NRE1, unless the double-stranded DNA binding factor also accounted for at least some of the single-stranded binding. Several examples of sequence-specific binding proteins that bind both doublestranded DNA and one of the two single strands have been identified (70 -73). In these cases, single-stranded DNA binding has been perceived as either a direct mechanism of DNA recognition or the end point of initial recognition of a doublestranded template. Competitive binding of different factors to the double-and single-stranded forms of a single element has also been observed (74 -78).
Fractionation of Jurkat nuclear extract over an upperstranded NRE1 oligonucleotide affinity column separated the upper-stranded binding activity into two components that yielded EMSA complexes. Analysis of the upNRE1 binding factor that eluted from the upNRE1 column in the 0.2 M KCl fraction identified it as immunologically related to Ku antigen. Both purified Ku and recombinant Ku purified from insect cells bound specifically to both double-stranded NRE1 and to the single upper strand. These results establish Ku as a protein that recognizes its response element with high affinity in both double-and single-stranded configurations. Further, the DNA helicase activity of Ku (46) suggests an obvious mechanism for the introduction and maintenance of strand separation around NRE1 that would promote the open DNA conformation needed for the binding of the factor which recognizes the lower strand of NRE1.
Although the second upper-strand NRE1 binding factor that occurred in the flow through of the upNRE1 oligonucleotide column appeared to bind strongly and specifically to upNRE1 in EMSAs, the complete reproduction of the footprints on the upper strand of NRE1 by Ku alone makes it unclear as to the potential of this second factor to interact with NRE1 in unfractionated extracts. The binding of purified Ku to the upper strand of the MMTV over NRE1 encompassed and even extended somewhat the footprints obtained with crude nuclear extracts (20,50). Thus, the KMnO 4 footprint was extended 5Ј to include the partial protection of the T at Ϫ399. With DNase I, the area of strong protection extended 3Ј to Ϫ381 to completely cruciforms and other structures (28, 41-45, 79 -82), to date it has been reported to bind only single-stranded unstructured DNA with low affinity (44 -46, 81). Our KMnO 4 footprinting results also suggest that Ku may be able to bind to free singlestranded DNA ends. As Ku binding to single-stranded DNA ends was not verified by DNase I footprinting, it would appear that Ku bound only very weakly, albeit with at least some specificity, to the single-stranded DNA ends. However, it is clear from our EMSA experiments in which large excesses of highly sheared single-stranded calf thymus DNA were routinely included, that single-stranded NRE1 binding (K d ϭ 3.5 Ϯ 1.3 nM) was strongly preferred to sequence-independent singlestranded DNA binding, including DNA end binding. Nonetheless, this single-stranded DNA end binding activity may be important for the DNA helicase activity of Ku. Otherwise, given the high affinity of Ku binding to double-stranded DNA ends, it is difficult to explain the requirement for an extended single-stranded overhang for double-stranded DNA to be unwound by Ku (46). One intriguing possibility is that the binding of Ku to the upper strand of NRE1, or to the point of strand separation that exists in the known helicase templates, recruits a second molecule of Ku to the end of the single-stranded overhang. Ku is known to form DNA-dependent dimers and Ku helicase activity has to date been observed only under conditions where dimers can form (46,81,83,84).
One issue that is not resolved by our experiments is whether, when presented with the single upper strand of the MMTV LTR, Ku binds directly to NRE1 or accumulates over NRE1 from alternative entry points as a result of translocation. Ku translocates efficiently from DNA ends and NRE1 on doublestranded DNA in the presence of Mg 2ϩ (27,41). It is also suspected to accumulate following translocation over some double-stranded sequences to which it does not bind directly (47). While not conclusive, the present evidence favors direct binding to the single, upper strand, of NRE1. For example, the binding reactions in our single-stranded footprinting experiments were performed in the absence of Mg 2ϩ , and KMnO 4 footprinting was accomplished in the complete absence of Mg 2ϩ . Moreover, with the exception of the weak interaction that we observed with single-stranded DNA ends, we have not detected evidence of higher order Ku-DNA complexes with single-stranded DNA that resemble the multimeric Ku-DNA complexes that are readily observed on double-stranded DNA (41,42).
Under normal circumstances in the cell, Ku is most likely to initiate its interaction with NRE1 by binding to the double-stranded LTR DNA, as despite detailed study, no obvious strand separation has been detected in MMTV LTR chromatin (85,86). Although the sequence preferences for double-and single-stranded NRE1 binding appeared to be very similar, under the same DNA binding conditions, the affinity for singlestranded NRE1 (K d ϭ 3.5 Ϯ 1.3 nM), was 3-4-fold lower than the affinity of Ku binding to double-stranded NRE (K d ϭ 0.84 Ϯ 0.24 nM) (47). Despite the decreased affinity, the kinetics of binding of recombinant Ku to single-stranded NRE1 (t1 ⁄2 on ϭ 2 min, t1 ⁄2 off ϭ 68 min) appear to be highly similar to the kinetics of the binding of the Ku in Jurkat nuclear extracts to doublestranded NRE1 (49).
Recent reports indicate that multiple subdomains within Ku have the potential to mediate DNA end binding. In particular, it appears that either the amino or carboxyl terminus of Ku70 can pair with the carboxyl terminus of Ku80 to bind doublestranded DNA ends (79 -82). One intriguing feature of doublestranded NRE1 binding by Ku is that in the absence of Mg 2ϩ / ATP, only Ku70 appears to be in intimate contact with the DNA. Further, Ku70 only appears to contact the upper strand of NRE1 directly. Whether this contact is mediated by the NH 2 or COOH terminus of Ku70 is not yet known. However, upon addition of Mg 2ϩ /ATP, a change occurs in the interaction of Ku with NRE1 that is reflected by the cross-linking of Ku80 in addition to Ku70 to the upper strand of the double-stranded sequence. Thus the dynamics of Ku DNA binding appear to reflect a differential participation of individual DNA binding subdomains of Ku with the upper strand of NRE1.
While the cross-linking of Ku70/80 to double-stranded NRE1 in the presence of Mg 2ϩ /ATP mirrors the cross-linking of the two Ku subunits to the single, upper strand of NRE1, we have been unable to detect complete unwinding of the two strands of blunt-ended NRE1-containing oligonucleotides in DNA helicase assays performed with either purified Ku or crude nuclear extracts. 3 However, the contact of Ku80 with double-stranded NRE1 does correlate with the destabilization of the doublehelix in sequences flanking NRE1 that is reflected by the induction of sensitivity to modification by KMnO 4 and cleavage by S1 nuclease (50). Whether this destabilization is sufficient to allow the binding of the other single-stranded binding factors to NRE1 remains to be demonstrated. Together, our results to date suggest that the interaction of Ku with NRE1 occurs in two steps; initial contact of Ku70, followed by a Mg 2ϩ -dependent structural transition that leads to the contact of Ku80 with DNA. Interestingly, there is a recent report (81) that indicates that Ku70 alone has helicase activity.
The importance of the single-stranded NRE1 binding factors for the regulation of MMTV transcription by Ku/DNA-PK remains to be demonstrated. However, two results raise the expectation that these factors will also be required for the repression of MMTV transcription through NRE1. Previously, we have demonstrated that a truncated NRE1 element, which supports double-stranded nuclear factor binding only, is unable to repress transcription from the MMTV promoter proximal regulatory region, even when present in multiple copies (20). Results, to be submitted elsewhere, 4 demonstrate that this truncated element is indeed a direct internal double-stranded DNA binding site for Ku. In contrast, there is no detectable binding to the upper, single strand of this element by any factor including Ku.
Last, it has recently been shown that the XRCC4 gene product facilitates the DNA end binding activity of Ku and the recruitment of DNA-PK cs to DNA (87). It will be interesting to determine whether this factor also participates in the binding of Ku/DNA-PK to NRE1. Interestingly, with a predicted molecular mass of 38 kDa, XRCC4 is a candidate for the lowerstranded NRE1 binding factor which in cross-linking experiments yields a DNA-protein complex that migrates at 45 kDa.