Ku Antigen-DNA Conformation Determines the Activation of DNA-Dependent Protein Kinase and DNA Sequence-Directed Repression of Mouse Mammary Tumor Virus Transcription

Abstract

Mouse mammary tumor virus (MMTV) transcription is repressed by DNA-dependent protein kinase (DNA-PK) through a DNA sequence element, NRE1, in the viral long terminal repeat that is a sequence-specific DNA binding site for the Ku antigen subunit of the kinase. While Ku is an essential component of the active kinase, how the catalytic subunit of DNA-PK (DNA-PK_cs) is regulated through its association with Ku is only beginning to be understood. We report that activation of DNA-PK_cs and the repression of MMTV transcription from NRE1 are dependent upon Ku conformation, the manipulation of DNA structure by Ku, and the contact of Ku80 with DNA. Truncation of one copy of the overlapping direct repeat that comprises NRE1 abrogated the repression of MMTV transcription by Ku–DNA-PK_cs. Remarkably, the truncated element was recognized by Ku–DNA-PK_cs with affinity similar to that of the full-length element but was unable to promote the activation of DNA-PK_cs. Analysis of Ku–DNA-PK_cs interactions with DNA ends, double- and single-stranded forms of NRE1, and the truncated NRE1 element revealed striking differences in Ku conformation that differentially affected the recruitment of DNA-PK_cs and the activation of kinase activity.

The long terminal repeat (LTR) of mouse mammary tumor virus (MMTV) includes a complex transcriptional regulatory region that is strongly responsive to steroid hormones and prolactin (5, 6, 13–15, 18, 37, 39, 67, 73, 83). Tumorigenesis is mediated through the insertional activation of cellular proto-oncogenes (12, 42, 84). In addition to the promoter-proximal hormone response element and distal prolactin-responsive region, MMTV also contains several DNA sequence elements in the central portion of the LTR that repress or limit virally induced transcription and the response of the virus to steroid hormones (3, 51, 55, 56, 76, 77, 90). At least some of these elements appear to function to restrict virally induced gene expression and tumorigenesis to the lactating mammary gland, where the effects of prolactin and steroid converge to overcome the negative regulatory elements to promote a strong induction of virally induced transcription. Deletion of portions of the viral LTR that include one or more of the negative regulatory sequences deregulates MMTV-induced transcription and leads to virally induced tumors at sites not normally observed with wild-type virus, most notably T-cell lymphoma (40, 77, 90, 91).

Studies by several groups have shown that one of the negative regulatory sequences in the MMTV LTR that acts to repress viral transcription occurs in a region between 350 and 400 bp upstream of the viral transcriptional initiation site (21, 40, 77, 85, 90). We previously delimited the repressor element within this region, NRE1, to 23 bp of DNA centered over a sequence containing an overlapping direct repeat of the sequence GAGAAAGA (31). NRE1 was shown to inhibit transcription from the MMTV promoter-proximal regulatory region in T cells and in transformed mammary fibroblasts derived from an MMTV-induced tumor (31). Deletion of sequences including this element from the viral LTR also led to increased viral transcription in other T cells and fibroblasts (30, 31, 90).

In subsequent studies, we showed that NRE1 was a direct, sequence-specific DNA binding site for the Ku antigen (p70/p80) DNA binding subunit of DNA-dependent protein kinase (DNA-PK) and that it supported the activation of the kinase catalytic subunit (DNA-PK_cs) (29, 30). Further, the repression of MMTV transcription correlated with the recruitment of DNA-PK_cs to NRE1, as MMTV transcription was unaffected by NRE1 in cells derived from the severe combined immunodeficient (SCID) mouse, in which DNA-PK_cs is mutated (30). However, whether DNA-PK_cs has a structural or enzymatic role in regulation remains to be established. Repression was also absent from cells containing inactivating mutations in the Ku80^−/− subunit but was recovered upon replenishment of Ku80 (30). Subsequently, we have reported that Ku also binds to the upper, single strand of NRE1 (upNRE1) in the MMTV LTR (78), which suggested that the effects of Ku–DNA-PK_cs at NRE1 may also involve the Ku-mediated induction of a structural transition in the DNA.

The DNA sequence-directed regulation of MMTV transcription by Ku–DNA-PK_cs is one activity for a protein complex that has recently been shown to play important roles in many cellular processes, including nonhomologous double-stranded DNA break repair (24, 36, 59, 65, 75), V(D)J recombination (8, 48, 72, 75, 94), the suppression of thymic lymphoblastic lymphoma (54), and, from several perspectives, transcription (22, 30, 49, 50, 66).

Ku is an unconventional DNA binding protein that, in addition to double-stranded NRE1 (dsNRE1) and single-stranded NRE1, binds double-stranded DNA ends and binds to DNA structures containing double- to single-strand transitions (7, 11, 19, 23, 33, 35, 63). Ku has also been reported to translocate linearly along double-stranded DNA from DNA ends and NRE1 (19, 63) and to have limited DNA helicase activity (82). The binding of Ku to DNA ends and hairpin structures appears to be required for the V(D)J recombination and nonhomologous double-stranded DNA break repair (16, 36, 44–46, 58, 62, 87, 94). Ku binding to NRE1, however, occurs in vitro with an affinity at least 10-fold higher than DNA end binding (29), which may reflect the increased affinity required to recognize internal DNA sequences in chromatin.

DNA end-bound Ku recruits and promotes the activation of DNA-PK_cs (33, 38, 92). DNA-PK_cs is also activated by Ku bound to DNA through NRE1 (29, 30). Further, the specific targeting of DNA-PK_cs to NRE1 and DNA ends through Ku is reflected by a striking preference for the phosphorylation of substrates linked to DNA-PK in cis on DNA (29, 30). This preference for cis phosphorylation, which may exceed 3 orders of magnitude, likely reflects the low affinity of DNA-PK_cs for substrate (K_m, ∼200 μM) compared to its much higher affinity for NRE1 and DNA ends (1, 29). By contrast, the binding of Ku to DNA nicks or structural features such as hairpins does not appear to promote the activation of DNA-PK_cs (33, 38, 71).

Significantly, Ku also is involved in many cellular processes, apparently independently of DNA-PK_cs. These include telomere maintenance (9, 34, 80), progression through the G₂/M transition in the cell cycle (57), and Sir4-mediated transcriptional silencing (80). Ku occurs in the cell at higher levels than DNA-PK_cs and does not interact with DNA-PK_cs in solution (33, 74). However, Ku has been found to interact in solution with several other factors. These include protein phosphatase 2A (53); p95^vav, a hematopoietic oncogene (69); Sir4 (80); TATA binding protein (26); REF1, a redox factor that regulates transcription through negative calcium response elements (17); HSF1, a heat shock gene regulatory factor (41); and the bromodomains of the GCN5/CBP/TAF_II250 transcriptional coactivator (4). Little is known of the effect of Ku-DNA binding on these interactions.

In the present study, we have begun to dissect the molecular basis for the repression of MMTV transcription by DNA-PK and the DNA sequence requirements for that repression. Our results indicate that DNA-PK_cs kinase activity plays a central role in the repression of MMTV transcription through NRE1 and support a model in which the activation of kinase activity correlates with the induction of a DNA structural transition adjacent to NRE1 by Ku. Moreover, our results reveal Ku to be a strikingly flexible protein that adopts multiple DNA-dependent conformations that differentially control the recruitment and activation of DNA-PK_cs and that may be expected to play a major role in regulating the interaction of Ku with other proteins.

MATERIALS AND METHODS

Oligonucleotides, microcircles, and plasmids.

Oligonucleotides were synthesized on a Beckman Oligo 1000 DNA synthesizer. The following sequences are those of the upper strands of all oligonucleotides: 23-mer NRE1, 5′-AACTGAGAAAGAGAAAGACGACA-3′; 58-mer NRE1, 5′-AGCTTGAGCTAGACCT CCTTGGTGTATGCTAACTGAGAAAGAGAAAGACGACATGAAA-3′; 17-mer MT upper, 5′-AACTGAGAAAGACGACA-3′; MT_mt, 5′-AACTGAGATAGACGACA-3′; 58-mer MT, 5′AGCTTGAGCTAGACCTCCTTGGTGTATGCTAACTGAGAAAGACGACATGAAACAACAG-3′; 39-mer nonspecific oligonucleotide, 5′-ACCCTACTGCAGTAATAGTGAACCTGCTGTGTTTTGCTC-3′; and a 58-mer nonspecific oligonucleotide, 5′-TCGAGGATCCTGAG CTCATGTACCCTTACGACGTGCCAGATTATGCATATGGTACCGT-3′. The oligonucleotides used for primer extension of KMnO₄-treated plasmids were 5′-GACTTAAATTGGGATAG-3′ (upstream of NRE1) and 5′-TGTTCTATCAGTCCAG-3′ (downstream).

Covalently closed microcircles were prepared free of nicks or other structural features exactly as described previously (30) from 223-, 240-, and 246-bp pBluescript fragments containing no insert or containing single copies of the MT, MT_mt, or NRE1 oligonucleotides, respectively, cloned into the SmaI site of pBluescript. Recircularized DNA was routinely digested with exonuclease III to remove DNA ligated on only one strand (nicked DNAs) and the remaining linearized DNA. Covalently closed microcircles were subsequently gel purified and verified for resistance to S1 nuclease, exonuclease III, and Bal31 exactly as we have described previously (30). In addition, the microcircles were also tested and were shown to be completely resistant to T4 and T7 endonucleases and to modification by a 20 mM concentration of the single-stranded-DNA-specific reagent KMnO₄. KMnO₄ treatment was performed as described below for covalently closed circular plasmid DNAs. Following modification, the DNA was restricted, cleaved with piperidine, and resolved on DNA sequencing gels against a KMnO₄ sequencing track of the lower, polypyrimidine-rich NRE1 strand prepared by treatment of the single strand of the parent fragment with 1 mM KMnO₄, as we have previously described (78).

Plasmid pHCMT was created by inserting one copy of the double-stranded 58-mer MT oligonucleotide into the HindIII site of plasmid pHC364. This resulted in the truncation of NRE1 to MT while maintaining the exact position of NRE1 and without otherwise affecting the LTR sequence to position −421. Plasmids pHC17 (MMTV sequence positions −421 to +125) and pHC364 (MMTV sequence positions −364 to +125) have been described previously (55). Covalently closed circular plasmids were prepared as described above for the microcircles except that gel purification was performed through agarose rather than polyacrylamide.

Recombinant proteins.

Expression and purification of Ku from insect cells was performed essentially as described by Ono et al. (61). Baculovirus expression vectors containing the p86 subunit (VBB2-Kup86) and a hexahistidine-tagged p70 subunit (VBB2-Kup70tH6) of human Ku antigen were coinfected into Sf9 cells. Three days postinfection, cells were harvested and lysed by sonication in 40 mM HEPES (pH 7.9), 1 mM EDTA, 2 mM dithiothreitol, 0.1% Nonidet P-40, 1 mg of leupeptin/ml, 1 mg of pepstatin/ml, and 1 mM phenylmethylsulfonyl fluoride. Ku heterodimers were then purified by using a Ni 21 affinity resin (His-Bind Resin; Novagen). Purity was assessed by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis as we have described previously (29, 30) and was estimated at over 95%. pGEX-2T-X568, encoding a glutathione S-transferase (GST)–glucocorticoid receptor (GR) fusion protein containing amino acids 407 to 568 of rat GR, was created by cloning the BamHI-EcoRI fragment of pSP64X568 (70) in frame into the BamHI-EcoRI sites of pGEX-2T (Pharmacia). GST-GR expression and purification were performed as we have described previously (29).

Tissue culture, transfections, nuclear extracts, and CAT assays.

V79 Chinese hamster fibroblast cells (93) were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum (Life Technologies, Inc.), while Jurkat T cells were cultured in similarly supplemented RPMI medium (Life Technologies, Inc.). Plasmid DNAs were grown in Escherichia coli DH5α and purified through two CsCl gradients. V79 cells (2 × 10⁶ per plate) were transfected with 4 μg of MMTV/chloramphenicol acetyltransferase (CAT) plasmids (pHC17, pHC364, or pHCMT), 2 μg of rat GR expression vector p6RGR (64), and 1 μg of Rous sarcoma virus–β-galactosidase (RSV–β-gal) by using DEAE-dextran (14). Sixteen hours after transfection, cells were treated with 2 × 10⁻⁷ M dexamethasome (dex) for 48 h. Jurkat cells (5 × 10⁶ in 0.8 ml) were transfected with 1 μg of CAT reporter plasmid, 200 ng of rat GR expression vector p6RGR (64), 500 g of RSV–β-gal, and 1.3 μg of pBluescript carrier DNA by using 20 μl of Lipofectamine (Life Technologies, Inc.) for 6 h and were then cultured overnight in complete medium. Treatment with dex (2 × 10⁻⁷ M) was for 2 h. CAT and β-gal assays were performed as described previously (14, 30). Transfections and assays were performed in duplicate on three to five separate occasions and were standardized to β-gal activity (± standard errors of the means). Nuclear extracts were prepared from Jurkat cell cultures (2 × 10⁸ to 2 × 10⁹ cells) essentially in accordance with standard protocol (20, 31).

EMSA and protease digestion.

Binding of purified Ku or DNA-PK (Promega) to 4 to 5 pmol of γ-³²P-labeled oligonucleotides (NRE1, MT, or nonspecific) or 4 to 5 pmol of pBlue, MT, MT_mt, and NRE1 containing microcircles was performed in a 20-μl final volume in a solution containing 12 mM HEPES (pH 7.9), 12% glycerol, 60 mM KCl, 0.12 mM EDTA, and 1 μg of bovine serum albumin for 20 min at 20°C. Binding to linear, nonspecific DNAs was performed in the presence of 100 ng of highly sheared calf thymus DNA, while sequence-specific binding was performed in the presence of 2 μg of single- or double-stranded highly sheared calf thymus DNA. The labeling efficiency of the microcircles was 100% as recirculization depended upon γ-³²P addition, while the efficiency of oligonucleotide labeling was carefully monitored by spectrometry and scintillation counting to ensure that equal amounts of radioactivity and oligonucleotide were added to each incubation. Where indicated, the DNA-PK_cs monoclonal antibody 25-4 or Ku antibody 111 (Neomarkers) was added to the reaction at the beginning of the incubations. DNA protein complexes were resolved on 4 and 10% polyacrylamide gels that were electrophoresed at between 225 and 600 V/h in 0.5× Tris borate buffer. Where indicated, following the equilibration of binding, samples were incubated for an additional 5 min prior to electrophoretic mobility shift assay (EMSA) with 2 to 5 μg of l-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin, 1 μg of chymotrypsin (Boehringer Mannheim), or 2 μg of Asp-N (Boehringer Mannheim).

Kinase assays.

Analysis of the phosphorylation of GST-GR by DNA-PK on covalently closed circular DNAs was performed essentially as previously described (30). Plasmids employed as cofactors were linearized with HindIII, recircularized with T4 DNA ligase, and nuclease treated, and the covalently closed circles were gel purified (29, 30). Kinase reactions were performed for 20 min at 30°C in kinase buffer (50 mM HEPES [pH 7.5], 100 mM KCl, 10 mM MgCl₂, 0.2 mM Na₂-EGTA) in the presence or absence of 10 ng of plasmid DNA (pHC364, pHCMT, or pHC17), 0.5 U of DNA-PK (Promega), 5 μCi of [γ-³²P]ATP (6,000 Ci/mmol; Amersham) and 2 to 6 μl of a 1:1 slurry of GST-GR bound to glutathione-Sepharose beads that had been prewashed twice in kinase buffer. Reactions were stopped by dilution in binding buffer lacking MgCl₂ and containing 10 mM Na₂-EDTA at 4°C. The beads were washed three times in kinase buffer to remove free [γ-³²P]ATP. The GST-GR was eluted by boiling in SDS sample buffer, and ³²P incorporation was determined by autoradiography and phosphorimage analysis (model 525 phosphorimager; Bio-Rad) of SDS–12% polyacrylamide gels.

DNA-PK phosphorylation of a p53 peptide was performed in 20 μl of kinase buffer by using 10 U of DNA-PK, 2.8 nmol of DNA-PK p53 peptide substrate (EPPLSQEAFADLWKK; Promega) (52), 1 nmol of unlabeled ATP, 5 μCi of [γ-³²P]ATP, and 2 ng of the oligonucleotides indicated. Following a 20-min incubation at 30°C, the reactions were stopped by the addition of 30% acetic acid. The p53 peptide was subsequently recovered on Whatman P-81 paper. After extensive washing, the incorporated radioactivity was determined by scintillation counting.

KMnO₄ hypersensitivity analysis.

Structural distortion of covalently closed circular double-stranded DNA template following specific factor binding was assessed by using the T-specific DNA-modifying agent KMnO₄ (25). Seven micrograms of Jurkat nuclear extract or 10 ng of recombinant Ku was incubated for 20 min with 10 ng of pHC17 or pHCMT under the same conditions as employed for EMSA (1 μg of competitor DNA). Samples were then treated with 5 mM KMnO₄ in the presence or absence of 10 mM MgCl₂ for 1 min at room temperature, extracted with phenol-chloroform, and precipitated before being subjected to piperidine cleavage at 90°C for 30 min. To identify sites of cleavage, primer extension of [γ-³²P]ATP-labeled oligonucleotides specific for pHC17 and pHCMT was performed, followed by electrophoresis of the samples on 8% polyacrylamide sequencing gels to resolve the cleavage sites. KMnO₄ and dimethylsulfate sequence reactions were performed as previously described (27) on the same plasmids in linear, denatured form, and primer extension was also used to position sites of KMnO₄ modification.

DNA-protein cross-linking.

To obtain double-stranded ³²P-labeled DNA probes, the upper strands of the 23-mer NRE1 and 17-mer MT or both strands of the 39-mer nonspecific oligonucleotides were labeled with [γ-³²P]ATP by using T4 polynucleotide kinase (New England Biolabs), annealed with their corresponding unlabeled lower strand, and recovered from polyacrylamide gels. Probes (20 pmol) were incubated with purified recombinant Ku by using standard EMSA binding conditions including 10 mM ATP and/or 10 mM MgCl₂ in the incubations as indicated. In the absence of additional protease digestion, following the equilibration of binding, samples were directly exposed to UV for 12 min at 4°C in a Stratalinker 1800 (Stratagene). Cross-linked products were resolved on SDS–10% polyacrylamide gels. In other experiments, protease digestion and EMSA were performed first and were followed by UV irradiation of polyacrylamide gels for 15 min. Gel slices including the Ku-DNA complexes were then excised from the gel, washed for 10 min three times in SDS sample buffer, heated to 95°C for 5 min, inserted into the wells of SDS–15% polyacrylamide gels, and electrophoresed to separate the cross-linked products. Both protocols are modifications of procedures reported previously (31, 78). In the protease experiments, the control undigested samples were also cross-linked in the gel.

RESULTS

Repression of MMTV transcription by DNA-PK correlates with the binding of Ku to single-stranded NRE1.

A distinguishing feature of the NRE1 element in the MMTV LTR is the overlapping direct repeat of the sequence GAGAAAGA. In preliminary experiments with Jurkat T cells, we found that the deletion of one copy of this repeat from NRE1 compromised the repression of transcription from the MMTV promoter-proximal regulatory region by a synthetic array of NRE1 elements without appearing to affect the recognition of double-stranded oligonucleotides in EMSA with crude nuclear extracts (31). By contrast, binding to upNRE1 was lost.

To examine the extent to which these observations reflected a change in Ku antigen binding to NRE1, we compared the binding of crude Jurkat nuclear extract and purified recombinant Ku antigen to various forms of NRE1 and MT by EMSA. In the first instance, a factor in Jurkat nuclear extract that we have previously demonstrated to be Ku (30) formed a specific complex with both the dsNRE1 element and the purine-rich upNRE1 (Fig. 1, lanes 1 and 2). Under these stringent binding conditions, which included the addition of 2 μg of highly sheared calf thymus DNA competitor DNA, no Ku binding was detected on the double-stranded DNA ends of a nonspecific oligonucleotide (27). Incubation of dsNRE1 and the upper, single strand with recombinant Ku resulted in the formation of a similar complex (lanes 5 and 6). The MT oligonucleotide was also recognized by Jurkat nuclear extract and recombinant Ku when double stranded (lanes 3 and 7). By contrast, no complex was detected on the upper, single strand of this sequence in either instance (lanes 4 and 8), and recombinant Ku also failed bind to the lower, single strand of either NRE1 or MT (lanes 9 and 10).

FIG. 1.

Differential binding of Ku to double- and single-stranded DNAs upon truncation of NRE1. EMSA on a 4% polyacrylamide gel of Ku binding to full-length and truncated NRE1 elements through incubation of Jurkat nuclear extract (lanes 1 to 4) or purified recombinant Ku expressed from baculovirus (lanes 5 to 10) with 4 to 5 pmol of double-stranded (ds) (lanes 1, 3, 5, and 7), upper-strand (up) (lanes 2, 4, 6, and 8), and lower-strand (lo) (lanes 9 and 10) 23-mer NRE1 and 17-mer MT oligonucleotides as indicated above the gel. Binding was performed in the presence of 2 μg of highly sheared calf thymus DNA.

To confirm the specificity of binding and to ensure that the binding of Ku to the double-stranded MT (dsMT) element did not involve the DNA ends of the oligonucleotide employed, we compared Ku binding to covalently closed DNA microcircles containing the MT motif, the full-length NRE1 sequence, and an oligonucleotide containing a single A-to-T substitution in the MT polypurine core (Fig. 2). Covalently closed microcircles were prepared free of nicks or structural features from pBluescript plasmids as previously described (30). The microcircles were confirmed to be completely resistant to digestion with S1 nuclease, Bal31, exonuclease III, and T4 and T7 endonucleases and to be resistant to chemical modification by the single-stranded-DNA-specific reagent KMnO₄ (28). Binding of crude Jurkat nuclear extract and recombinant Ku was again assessed in the presence of 2 μg of calf thymus DNA. Under these conditions, neither Jurkat nuclear extract nor recombinant Ku recognized a microcircle containing only the pBluescript plasmid sequences (lanes 1 to 3 and 13 to 15). However, the addition of the NRE1 sequence to the microcircle resulted in the formation of a shifted complex whose migration was retarded further by a Ku antibody (lanes 4 to 7 and 16 to 18). A microcircle containing the MT element yielded similar complexes that were supershifted by the Ku antibody (lanes 7 to 9 and 19 to 21). Thus, although truncation of NRE1 to MT eliminated binding to upNRE1, it had little effect on the binding of Ku to dsNRE1.

FIG. 2.

MT is a direct, sequence-specific, double-stranded DNA binding site for Ku. EMSA on a 4% polyacrylamide gel with recombinant Ku was performed with 4 to 5 pmol of covalently closed circular DNA microcircles comprised of a 223-bp DNA microcircle from pBluescript (lanes 1 to 3 and 13 to 15), a 246-bp microcircle containing the same pBluescript fragment with a 23-bp NRE1 insert (lanes 4 to 6 and 16 to 18), a 240-bp microcircle containing the pBluescript fragment with a 17-bp MT insert (lanes 7 to 9 and 19 to 21), and a 240-bp microcircle containing the pBluescript fragment with a 17-mer MT oligonucleotide containing the substitution GAGATAGA for the GAGAAAGA MT core sequence (MT_mt) (lanes 10 to 12 and 22 to 24) in the presence of 2 μg of highly sheared calf thymus DNA. Lanes 1 to 12 show the results obtained with Jurkat crude nuclear extract, and lanes 13 to 24 show results obtained with recombinant Ku. Ku binding was verified by inclusion of a Ku antibody in lanes 3, 6, 9, 12, 15, 18, 21, and 24. Inclusion of a nonspecific antibody had no effect on complex migration (28).

Ku binding to the MT and MMTV microcircles was sequence specific and not merely dependent upon random polypurine residues, as the control microcircle derived from pBluescript contained two polypurine-rich segments of 10 nucleotides (GAAAGGGGGA) and 9 nucleotides (GAGGGGGGG). Further, Ku binding was abrogated by inversion of an A/T base pair in the MT core to give the sequence GAGATAGA (lanes 10 to 12 and 21 to 24). Similarly, rearrangement of the full-length NRE1 or MT polypurine sequences while maintaining the same overall G and A content also abrogated Ku binding (28).

To examine whether truncation of NRE1 to MT within the MMTV LTR would abrogate the repression of MMTV transcription that we have shown to be dependent upon both Ku and DNA-PK_cs, we compared the transcriptional responsiveness of an MMTV LTR construct with an MT-NRE1 substitution introduced by site-directed mutagenesis (pHCMT) to that of a wild-type LTR construct (pHC17) and to that of a construct from which NRE1 had been completely deleted (pHC364) (Fig. 3). Transfections were performed in two cell lines: hamster ovary fibroblast V79 cells, in which we have previously reported DNA-PK-dependent repression of MMTV transcription (30) (Fig. 3A), and Jurkat cells, a mature T-cell line representing the cell type in which MMTV containing deletions over NRE1 leads to increased viral expression and cellular transformation (Fig. 3B).

FIG. 3.

Mutation of NRE1 to MT abrogates the repression of MMTV transcription by Ku. (A) Transient transfection analysis of glucocorticoid-induced (2 × 10⁻⁷ M dex) transcription from the MMTV LTR in V79 Chinese hamster ovary fibroblasts transfected with MMTV CAT reporter genes pHC364, which is truncated prior to NRE1 (MMTV sequences −364 to +125; lanes 1 and 2); pHCMT (−421 to +125; lanes 3 and 4), in which one copy of the GAGAAAGA repeat in NRE1 had been mutated; and pHC17 (MMTV sequences −421 to +125; lanes 5 and 6), containing wild-type NRE1. (B) Similar transient transfection analysis in Jurkat T cells. Results are expressed as percentages of the maximal activity of the pHC364 construct treated with dex. In both panels A and B, CAT values were corrected against a constitutive RSV–β-gal reporter construct to control for transfection. Results are the averages (± standard errors of the means) of three to five individual experiments performed in duplicate.

In both cell lines, extension of the viral LTR from positions −364 (pHC364) to −421 (pHC17) to include NRE1 resulted in a sharp decrease in the levels of CAT activity obtained in response to the addition of the synthetic glucocorticoid dex. In the V79 cells, hormone induction was reduced from 90-fold to approximately 20-fold, while in the Jurkat cells, the steroid response was reduced from 50-fold to 10-fold. By contrast, there was no significant effect on hormone-independent transcription.

Replacement of the NRE1 element by the MT element by site-directed mutagenesis of pHC17 (pHCMT) completely abrogated the NRE1-mediated repression of MMTV transcription in both cell lines. Thus, our results suggested that the repression of steroid-activated MMTV transcription by DNA-PK_cs was somehow related to the ability of Ku to bind to upNRE1.

MT is a sequence-specific DNA-PK binding site from which DNA-PK_cs is inactive.

To examine whether Ku binding to MT affected the activation of DNA-PK_cs, we compared the abilities of pHC17 and pHCMT to support the phosphorylation of a recombinant GR DNA binding domain fusion protein on the MMTV LTR (Fig. 4A). The phosphorylation of DNA-bound GST-GR substrate by DNA-PK in cis on covalently closed circular MMTV LTR-containing plasmids has been characterized in detail (29, 30). pHC364, the MMTV LTR plasmid truncated prior to NRE1, was unable to support the phosphorylation of GST-GR by DNA-PK when circular (lane 4). However, GST-GR was strongly phosphorylated by DNA-PK on pHC17 (lane 6). Further, DNA-PK also efficiently phosphorylated GST-GR and a GST-p53 fusion protein on covalently closed circular pBluescript plasmids containing the appropriate NRE1, GRE, and p53 binding site oligonucleotide inserts (28). Notably, however, closed circular pHCMT not only failed to support the phosphorylation of GST-GR by DNA-PK but actually appeared to decrease the background level of phosphorylation that resulted from low levels of DNA contamination in the DNA-PK preparation (lanes 2 and 5).

FIG. 4.

MT and single-stranded NRE1 are unable to activate DNA-PK in vitro. (A) Phosphorylation of GST-GR substrate by DNA-PK in the presence of covalently closed circular was determined for MMTV LTR plasmids pHC364 (lane 4), pHCMT (lane 5), and pHC17 (lane 6). Results of control reactions performed with pHC17 in the absence of DNA-PK and GST-GR are displayed in lanes 1 and 3, while phosphorylation of GST-GR in the absence of added DNA is shown in lane 2. Samples were resolved on 12% polyacrylamide gels. The molecular size markers are showed to the left, while the position of GST-GR is highlighted to the right. (B) p53 peptide phosphorylation by DNA-PK in the presence of 58-mer oligonucleotides encoding a nonspecific double-stranded DNA sequence (dsNS), the upper, single strand of the MMTV LTR centered over NRE1 (upNRE), or the same sequence in which mutations have been introduced into one copy of the NRE1 repeat (upMT). The results are expressed as counts per minute of ³²P incorporated, and the error bars represent the standard errors of the means for three independent experiments performed in duplicate.

To directly address whether DNA-PK_cs could be activated from upNRE1, we compared the abilities of double- and single-stranded linear oligonucleotides to support the phosphorylation of a p53 peptide that has been the classical substrate for measuring DNA-PK activity from DNA ends (52). In these experiments (Fig. 4B), double-stranded DNA end-dependent peptide phosphorylation by DNA-PK was strongly activated by a 58-mer double-stranded oligonucleotide of nonspecific sequence. By contrast, single-stranded oligonucleotides of the same length from the MMTV LTR containing either the full polypurine-rich NRE1 sequence or the truncated NRE1 MT element had no significant stimulatory effect on kinase activity.

While this result discounted the possibility of NRE1 serving as a sequence-specific, single-stranded DNA-PK activation site, it remained unclear whether our results reflected an inability of the DNA-PK_cs to associate with Ku bound to dsMT and upNRE1 or whether DNA-PK holoenzyme complexes might assemble on these motifs in a conformation in which kinase activity failed to be activated. To investigate this question, we examined the association of DNA-PK_cs with NRE1 and MT by EMSA (Fig. 5). First, when purified DNA-PK was incubated with the NRE1 containing the microcircle described above, a very slowly migrating complex was detected near the top of the polyacrylamide gel which was further retarded upon addition of a DNA-PK_cs antibody (Fig. 5A, lanes 5 to 8). By contrast, no binding to the pBluescript microcircle was detected (lanes 1 to 4), indicating that DNA-PK_cs binding was dependent upon Ku binding to NRE1.

FIG. 5.

Binding of DNA-PK to NRE1 and MT. (A) EMSA on a 4% polyacrylamide gel of the binding of recombinant Ku (lanes 2, 6, and 10) or purified DNA-PK (lanes 3, 4, 7, 8, 11, and 12) to ³²P-labeled, covalently closed microcircles containing no insert (lanes 1 to 4), the wild-type NRE1 element (lanes 5 to 8), or the MT element (lanes 9 to 12) in the presence of 2 μg of highly sheared calf thymus DNA. The incubations in lanes 4, 8, and 12 included a DNA-PK antibody. (B) EMSA of recombinant Ku (lanes 2 and 6) or purified DNA-PK (lanes 3, 4, 7, and 8) binding to single-stranded 58-mer oligonucleotides encoding upNRE1 in the presence of 2 μg of sheared, denatured calf thymus DNA (lanes 1 to 4) or dsNS (lanes 5 to 8) in the presence of 100 ng of highly sheared calf thymus DNA. A DNA-PK antibody (Ab) was added to the incubations in lanes 4 and 8. The positions of the Ku–DNA-PK complexes and the shift in migration by the DNA-PK antibody to just below the well are indicated to the right.

Remarkably, despite the MT sequence being unable to support kinase activity, DNA-PK_cs interacted with the MT-containing microcircle to the same extent as to NRE1 (lanes 9 to 12). By contrast, no DNA-PK complex was detected on the upper strand of NRE1 in a 58-mer upper-stranded MMTV LTR oligonucleotide under the same binding conditions (Fig. 5B, lanes 2 to 4). By comparison, a DNA-PK complex sensitive to DNA-PK_cs antibody was readily detected on a double-stranded linear oligonucleotide of nonspecific DNA sequence of the same length (lanes 6 to 8) when the highly sheared calf thymus competitor DNA added to the incubations was decreased to 100 ng from the 2 μg used with the specific sequences. Thus, while neither dsMT nor upNRE1 supported the activation of DNA-PK_cs, these two Ku binding sites were distinguished by the ability of DNA-PK_cs to associate with dsMT-bound Ku.

MT is a static Ku binding site contacted only by Ku70.

The inability of DNA-PK to be activated on MT-containing DNAs despite the apparent similarity in the binding of Ku–DNA-PK_cs to MT and NRE1 suggested that the interaction of Ku with dsMT might differ from its interaction with NRE1 in a way that precluded the activation of DNA-PK_cs. To test this hypothesis, we compared the binding of Ku to NRE1 and MT in several assays.

First (Fig. 6), Ku is a protein that is known to translocate along DNA from its binding site (19, 63). Translocation from NRE1 is facilitated by Mg²⁺ (30). Thus, the inclusion of Mg²⁺ in electrophoretic mobility shift binding assays leads to the formation of slower-migrating Ku-DNA complexes, such as that shown in lane 4 of Fig. 6, which is indicative of two molecules of Ku bound to the NRE1 microcircle. In this experiment, the addition of Ku antibody 111 prevented Ku translocation. By contrast, the addition of Mg²⁺ to incubations with the MT-containing microcircle failed to promote the formation of a slower Ku-DNA complex (lane 9), indicating that the ability of Ku to translocate from MT was at least decreased and possibly absent.

FIG. 6.

Decreased translocation of Ku from MT. EMSA on 4% polyacrylamide gel of recombinant Ku binding to closed 240- and 246-bp microcircles containing the wild-type NRE1 element (lanes 1 to 5) or the MT element (lanes 6 to 10) in the presence of 2 μg of highly sheared calf thymus DNA. The microcircles were coincubated with Ku, 10 mM MgCl₂, and Ku antibody (Ab) as summarized above the autoradiograph.

A second property of Ku binding to NRE1 that we have previously demonstrated as being coincident with its ability to translocate from NRE1 is the Mg²⁺/ATP-sensitive cross-linking of Ku80 to the upper strand of NRE1 by UV irradiation (78). On both single- and double-stranded DNAs, Ku appears to contact only the upper strand of NRE1 in a manner that is detectable by UV cross-linking (78). Further, in the absence of Mg²⁺/ATP, only Ku70 was cross-linked to the upper strand of dsNRE1 (Fig. 7, lane 2). By contrast, both Ku70 and Ku80 were cross-linked to the upper strand of NRE1 when the DNA was single stranded (Fig. 7, lane 1). Upon addition of Mg²⁺, a band representing the cross-linking of Ku80 became detectable with the double-stranded template (lane 3). Ku80 cross-linking to dsNRE1 became prominent when ATP was also included in the incubation (lane 4).

FIG. 7.

Mg²⁺/ATP-dependent contact of Ku80 with DNA is specific for NRE1. A 23-mer NRE1 nucleotide and a 17-mer MT oligonucleotide were ³²P-labeled on the upper, purine-rich strands, while a 39-mer nonspecific oligonucleotide was labeled on both strands. Following the incubation of recombinant Ku with the purine-rich upNRE1 (lane 1), dsNRE1 (lanes 2 to 4), or dsMT (lanes 5 to 7) or the dsNS oligonucleotide (lanes 8 to 10) under standard binding conditions in the presence of 10 mM MgCl₂ and/or 10 mM ATP as indicated above the autoradiograph, the samples were irradiated with UV. Following irradiation, the samples were electrophoresed through an SDS–12% polyacrylamide gel to resolve the cross-linked products.

Similar to NRE1, in the absence of added Mg²⁺/ATP, Ku70 alone was cross-linked to the upper strand of the dsMT oligonucleotide (lane 5). However, in contrast to NRE1, Ku80 contact of MT was not induced by the addition of Mg²⁺/ATP (lanes 6 and 7).

Cross-linking experiments performed by other groups have suggested that only Ku70 contacts the DNA when Ku is bound to DNA ends (33). While these experiments appear to have been performed in the presence of at least a small amount of Mg²⁺, the influence of ATP on Ku-DNA contacts was not evaluated. In our experiments, we confirmed that at Mg²⁺/ATP concentrations at which Ku80 was readily crosslinked to dsNRE1, only Ku70 was cross-linked to DNA when bound at the DNA end (lanes 8 to 10).

These results indicated that the addition of Mg²⁺/ATP induced a change in the contact of Ku80 with NRE1 that was accomplished on the full-length element but not on MT or at the DNA end. While this change resembled the contact of Ku with single-stranded NRE1, this event was unlikely to include full strand separation, as DNA-PK_cs failed to associate with Ku on single-stranded NRE1. Moreover, although Ku has been shown to have DNA helicase activity in the presence of an extended single-stranded DNA overhang, it has been reported to be unable to unwind duplex DNA with blunt ends (82). Similarly, we have not detected complete unwinding of the MTV oligonucleotide by Ku, even in the presence of Mg²⁺/ATP (28).

Ku induces a structural transition in the MMTV LTR upstream of NRE1.

We have, however, previously reported that sequences upstream from NRE1 in the MMTV LTR between positions −395 and −420 undergo a Mg²⁺-dependent structural transition in the presence of crude nuclear T-cell extract that is detected by single-stranded-DNA-specific cleavage agents, including KMnO₄ (27). KMnO₄ is a reagent that specifically modifies T’s in single-stranded DNA or DNAs containing double- to single-strand transitions (25).

To determine the contribution of Ku to the structural transition observed with crude extracts and the dependence of this transition on full-length NRE1, we examined covalently closed circular MMTV LTR plasmids containing the wild-type NRE1 element (pHC17) or the MT motif (pHCMT) for sensitivity to KMnO₄ following incubation with crude Jurkat nuclear extract or recombinant Ku (Fig. 8). Mg²⁺-dependent KMnO₄ modification was detected upstream of NRE1 in the presence of Jurkat nuclear extract (Fig. 8). This sensitivity was not enhanced further by the addition of exogenous ATP (28). On the upper strand, the T at position −399 was strongly sensitive to KMnO₄ while T’s at positions −410 and −413 were modified to a lesser extent (Fig. 8A, lanes 2 to 6). By contrast, on the lower strand, T’s at positions −416 and −418 were strongly modified while the T’s at −403 and −398 were weakly modified (Fig. 8B, lanes 2 to 6). Modification was strictly dependent upon the presence of Mg²⁺/ATP and was not observed in the absence of nuclear extract. Although these cleavage sites overlap closely with those previously reported on linear DNA, differences in the exact positions of modification from the experiments with linear DNA are likely due to the proximity of the ends of the DNA fragment employed in the previous study. Interestingly, KMnO₄ sensitivity was strictly localized upstream of NRE1 and was not detected within the core polypurine-polypyrimidine sequence or downstream of the element. Also, it appears unlikely that this modification reflects the complete separation of the two LTR strands, as four T’s on the upper strand of the LTR between positions −399 and −410 remained unmodified. The lack of modification of T’s at positions −402 and −404 on the upper strand is particularly curious, since the T at −403 was modified on the lower strand.

FIG. 8.

Ku induces an NRE1-dependent structural transition in the MMTV LTR immediately upstream of NRE1. (A) Upper LTR strand. (B) Lower LTR strand. Covalently closed circular relaxed pHC17 and pHC17MT plasmids were incubated with Jurkat nuclear extract (lanes 2 to 6 and 15 to 20) or purified recombinant Ku (lanes 8 to 12 and 22 to 27). Following the equilibration of binding, samples were treated with Mg²⁺ and KMnO₄ as indicated. KMnO₄-modified bases were identified following piperidine treatment by linear PCR from oligonucleotides extended downstream from the plasmid backbone (upper strand, panel A), or upstream from the MMTV LTR at position −289, and visualized following electrophoresis on 8% (8 M urea) polyacrylamide gels. The short arrows beside each gel highlight the positions of the modified T’s in the LTR sequences as determined by comparison to KMnO₄-generated T sequencing tracks generated with single-stranded DNAs (lanes 1, 7, 13, 14, 21, and 28). The long arrows parallel to the sequences highlight the positions of the GAGAAGA repeat units that occur twice in NRE1 (left) but only once in the MT substitution. (C) Summary of Mg²⁺-dependent, KMnO₄-mediated thymidine modifications in the MMTV LTR. The overlapping direct repeat of NRE1 is highlighted by the overlapping arrows parallel to the DNA sequence, while the positions of the modification are indicated by the perpendicular arrows. The sizes of the arrows are approximately proportional to the intensity of cleavage.

Remarkably, the relative intensity and the exact positions of KMnO₄ modification were exactly reproduced by recombinant Ku alone (Fig. 8A and B, lanes 8 to 13). Therefore, Ku appeared to be sufficient for the Mg²⁺-dependent structural transitions upstream of NRE1 in the MMTV LTR. By contrast, the covalently closed circular MMTV LTR plasmid containing the NRE1-to-MT mutation (pHCMT) was completely resistant to modification by KMnO₄, regardless of the addition of Mg²⁺ to crude extracts or recombinant Ku (Fig. 8A and B, lanes 15 to 20 and 22 to 27).

Further, the binding of Ku to NRE1 was sufficient to promote a structural transition in flanking DNA independent of the flanking DNA sequence. A similar Ku- and Mg²⁺-dependent, KMnO₄-sensitive structural transition was observed upstream of an NRE1 oligonucleotide cloned into pBluescript (28). Thus, DNA-PK_cs kinase activity and the repression of MMTV transcription from NRE1 correlated with Ku80 DNA contact concomitant with the induction of a structural transition upstream of NRE1.

Ku conformation is dependent upon the form of DNA to which it is bound.

Since Ku binding to different DNA sites resulted in the differential DNA contact of Ku80, the conformation of the individual Ku-DNA complexes might be expected to be distinct. Close examination of the mobility of the Ku-DNA complexes by EMSAs on covalently closed microcircle DNA consistently indicated that the mobility of the Ku-MT complex was slightly, but reproducibly, slower than that of the Ku-NRE1 complex, even though the MT microcircle was 6 bp shorter than the NRE1 microcircle (Fig. 2). A small difference was also observed between Ku bound to DNA ends and the upper strand of NRE1 (Fig. 5B).

To examine more closely whether the binding of Ku to different forms of DNA resulted in different conformations for the Ku-DNA complexes, we increased the percentage of the polyacrylamide in the gels in EMSAs from 4 to 10% and reexamined the binding of Ku to short oligonucleotides (Fig. 9A). Four modes of Ku-DNA binding were compared: to double-stranded NRE1, single-stranded NRE1, double-stranded MT, and double-stranded DNA ends on an oligonucleotide of nonspecific sequence. DNA end binding was again accomplished by reducing the amount of the competitor DNA added to the binding reaction to 100 ng, while sequence-specific binding to the specific oligonucleotides was ensured through the inclusion of 2 μg of highly sheared calf thymus DNA to compete for DNA end binding. Under these modified electrophoresis conditions, striking differences in the mobilities of the four Ku-DNA complexes were revealed (lanes 1 to 4). dsNRE1 binding resulted in the complex with greatest mobility. Ku binding to upNRE1 resulted in complexes with a slightly slower mobility that was very close to the mobility of DNA end-bound Ku complexes. It should be noted that the difference in mobility of the Ku complexes on dsNRE1 and that of the dsNS oligonucleotide validate our binding conditions as being specific for DNA ends and NRE1, respectively. Remarkably, under these conditions, the Ku-MT complex exhibited a markedly slower mobility than the other Ku-DNA complexes (lane 3). Interestingly, no obvious change in the mobility of these complexes was detected when the binding of Ku to these oligonucleotides was performed in the presence of Mg²⁺/ATP (28).

FIG. 9.

EMSA of Ku conformation and protease sensitivity when bound to different DNA forms. (A) EMSA on 10% polyacrylamide gel of Ku binding to 23-mer dsNRE1, 17-bp dsMT oligonucleotides, a 39-mer dsNS oligonucleotide, or upNRE1 oligonucleotide on a nondenaturing 10% polyacrylamide gel with or without a 5-min incubation with 2 or 5 μg of trypsin following the equilibration of binding. dsNRE1 and dsMT binding were performed in the presence of 2 μg of highly sheared calf thymus DNA, DNA end binding was performed in the presence of 100 ng of the same DNA, and binding to upNRE1 was performed in the presence of 2 μg of heat-denatured calf thymus DNA. (B) EMSA results obtained as described for panel A, except that protease digestion was performed with 2 μg of trypsin (Tryp), 1 μg of chymotrypsin (Chymo), or 2 μg of Asp-N as indicated, prior to electrophoresis. (C) Competition of Ku from DNA following trypsin digestion. Following the incubation of Ku with the DNAs described for panel A, trypsin was added to the incubations as indicated for 5 min. EMSA was performed following a further 5-min incubation in the presence or absence of a 100-fold excess of unlabeled 23-mer dsNRE1 DNA.

Differences in protein conformation can often be revealed by protease digestion. We performed two experiments to evaluate the sensitivity of Ku to protease digestion when bound to the different DNA forms. First, following the equilibration of Ku-DNA binding, we digested the samples with increasing concentrations of trypsin (Fig. 9A, lanes 5 to 16). Digestion of Ku bound to dsNRE1 resulted in an increase in mobility of the Ku-DNA complex at a low trypsin concentration that was highly resistant to the addition of higher concentrations of enzyme, suggesting the presence of a tightly packaged NRE1 binding core for Ku. Interestingly, digestion of Ku bound to the single-stranded NRE1, MT, and DNA ends with trypsin converted the mobility of each of these Ku-DNA fragments to mobilities indistinguishable from that of the core NRE1-bound Ku fragment.

Digestion of Ku-DNA complexes with two additional proteases with different cleavage specificities, chymotrypsin and Asp-N, continued to suggest similarity in the core Ku fragments bound to dsNRE1, dsMT, and dsNS (Fig. 9B). However, it also revealed a difference in the core single-stranded NRE1-Ku complex that was reflected by an increased mobility of the upNRE1-Ku complex upon chymotrypsin digestion (lane 9). Further, Asp-N digestion of MT-bound Ku resulted in a striking increase in complex mobility (lane 15 versus lane 12) to match the motility of Ku on NRE1 in the absence of digest (compare to lane 2). By contrast, Asp-N did not appear to affect the mobility of the other three Ku-DNA complexes. Finally, resistance to protease digestion appeared to be strictly a property of DNA-bound Ku, as the preincubation of Ku with trypsin prior to the addition of dsNRE1 completely prevented Ku-DNA binding (28).

The protease-resistant nature of the core Ku-DNA complexes suggested that protease treatment had little effect on Ku-DNA binding. One feature of Ku-DNA binding is its stability (19, 63, 68). In one test of whether digestion of Ku to the core fragment affected the parameters of Ku-DNA binding, we compared the off rates of Ku from the four DNA forms before and after trypsin digestion (Fig. 9C). Prior to trypsin digestion, the addition of a 100-fold excess of unlabeled dsNRE1 oligonucleotide for 10 min prior to EMSA had no significant effect on the amount of Ku prebound to any of the four DNA forms. After trypsin digestion, however, the off rates of Ku from all four Ku-DNA complexes were markedly accelerated, such that competition was now complete within 5 min. However, this striking increase in the off rate of Ku from DNA also suggests that the affinities of the three sequence-specific Ku binding activities relative to DNA end binding remained similar. All of the incubations with sequence-specific oligonucleotides were performed in the presence of a large excess of double-stranded DNA ends in the form of the highly sheared calf thymus DNA that otherwise would have been expected to compete for the more dynamic DNA binding of the core Ku fragments.

Lastly, to directly visualize the nature of the Ku peptides associated with the four DNAs following protease digestion, UV cross-linking of Ku-DNA complexes was performed in the gel following the electrophoresis of protease-digested samples (Fig. 10). In-gel cross-linking of undigested Ku-DNA complexes yielded exactly the same cross-linked products observed when cross-linking was performed in the tube (28) (Fig. 6). In particular, in the absence of Mg²⁺, Ku80 was cross-linked only to upNRE1 (28).

FIG. 10.

In-gel cross-linking of protease-digested Ku-DNA complexes reveals a difference in Ku binding to dsNRE1 and DNA ends. SDS-PAGE analysis of Ku cross-linked to dsNRE1, upNRE1, dsMT, and a 39-mer dsNS oligonucleotide following EMSA of protease-treated Ku-DNA complexes. Ku binding to the dsNS oligonucleotide was performed in the presence of 100 ng of calf thymus DNA, while binding to dsNRE1, dsMT, and upNRE1 was performed in the presence of 2 μg of double-stranded or denatured calf thymus DNA. Protease digestion was performed with chymotrypsin (Chymo) (lanes 1 to 4), Asp-N (lanes 5 to 8), and trypsin (Tryp) (lanes 9 to 12). Following EMSA, the wet gel was irradiated with UV. Polyacrylamide slices containing Ku-DNA complexes were excised from the gel, washed and heated in SDS sample buffer, and then electrophoresed through an SDS–20% polyacrylamide gel and subjected to autoradiography. The asterisks highlight two bands that migrated within the salt front on the gel and which had mobilities identical to those of the ³²P-labeled free oligonucleotides cut from the EMSA gel and electrophoresed in parallel with the Ku-DNA complexes (28). Complexes specific for individual DNAs and/or protease treatments are highlighted by the numeric labels to the right of the autoradiograph.

By contrast, following protease digestion, no intact Ku70 or Ku80 was found to be cross-linked to any of the DNAs (28). However, at least four peptide-DNA complexes (labeled 1 to 4) that migrated between 24 and 33 kDa were observed (Fig. 10). It is important to recognize that the mobility of these complexes reflected the sum of the amino acid and nucleotide components. With full-length Ku, this increased the apparent size of the Ku70- and Ku80-DNA complexes by approximately 10 kDa, allowing for a rough estimate of the peptide component of complexes 1 to 4 of between 14 and 25 kDa.

A first observation is that Ku70 binding to all four DNAs was resistant to cleavage of the Ku70 by at least three different proteases, two of which were cleaved in at least two places. This was evidenced by the presence of two cross-linked peptides in most lanes containing samples treated with trypsin and Asp-N (complexes 3 and 4). Two additional bands were derived from the migration of the oligonucleotides with the salt front (28). An additional band at 22 kDa was of uncertain origin, as it was not observed with the oligonucleotides alone (28) but appeared in each protease-digested lane.

A second observation is that following cleavage with each of the three proteases, the peptides cross-linked to the dsNRE1, upNRE1, and dsMT were electrophoretically indistinguishable. Thus, the expectation of an additional, Ku80-derived protein-DNA product on upNRE1 was not realized. Thus, while Ku70-DNA contact resisted protease digestion, it may have compromised Ku80 contact with upNRE1.

Digestion with chymotrypsin yielded peptide-DNA complexes of 21 kDa for all four oligonucleotides (complex 1), while trypsin digestion resulted in doublets of 29 and 33 kDa. However, while Asp-N digestion of upNRE1-, dsNRE1-, and dsMT-bound Ku yielded 29- and 33-kDa doublets similar in mobility to the tryptic peptide-DNA products, a unique peptide-DNA complex with a mobility equivalent to 27 kDa was observed following Asp-N digestion of DNA end-bound Ku. Thus while the mobilities of dsNRE1- and DNA end-bound Ku were highly similar in EMSAs following protease digestion, a clear difference in protease sensitivities and/or DNA contact by Ku70 was apparent in these cross-linking experiments. Therefore, the differences in migration of the four Ku-DNA complexes reflected differences in Ku conformation on all four DNA forms that were reflected by specific differences in the protease sensitivities of the DNA-bound Ku.

DISCUSSION

In this study, we have distinguished the requirements for sequence-specific Ku-DNA binding from the activation of DNA-PK_cs at NRE1 and DNA ends. Our results support the proposal that the repression of steroid-induced transcription at the MMTV promoter through NRE1 is dependent upon the phosphorylation of target proteins by DNA-PK_cs. Further, the activation of DNA-PK_cs from NRE1 was not simply dependent upon the recruitment of DNA-PK_cs to DNA by Ku but appeared to be determined by the introduction of torsional stress 5′ to the Ku binding site in the MMTV LTR and by the entry of Ku80 into contact with double-stranded DNA. By contrast, the activation of DNA-PK_cs at DNA ends appeared to be mediated by a distinct conformation of Ku and occurred in the apparent absence of Ku80-DNA contact.

NRE1 appears to function in the MMTV LTR primarily to repress or blunt the induction of viral transcription in response to glucocorticoids (40, 77, 90). Previous results comparing the induction of MMTV transcription in SCID cells to that observed in normal cells have suggested that the recruitment of DNA-PK_cs to NRE1 plays a key role in the transcriptional effects mediated through NRE1 (30). Our present results with the MT element, which failed to repress MMTV transcription, argue strongly that the activation of DNA-PK kinase activity at NRE1 is the determining factor in the repression of MMTV transcription by NRE1.

Potential phosphorylation targets of DNA-PK at the MMTV promoter include GR and the octamer transcription factors that mediate the steroid-dependent induction of MMTV transcription (30). Other potential targets, such as nuclear factor 1 and the basal transcription machinery, would seem less likely candidates at this time, as NRE1 does not appear to influence hormone-independent MMTV transcription in our system. While nuclear factor 1 has been implicated in the hormone responsiveness of MMTV on chromatin templates (18, 79), in transient transfection experiments, its effects on transcription are hormone independent (2, 10). The recent report that DNA-PK might down regulate the histone acetyltransferase activity of some transcriptional coactivators (4) suggests the additional possibility that the recruitment and activation of DNA-PK_cs at NRE1 represses the positive effects of transcriptional coactivators recruited to the MMTV promoter in response to steroid.

Ku and DNA-PK_cs are known to form a stable complex only on DNA (33). Two possibilities for the association of Ku and DNA-PK_cs on DNA have been recognized. In the first instance, the simultaneous interaction of Ku and DNA-PK_cs with DNA would stabilize a transient or low-affinity protein-protein interaction (38, 88). However, it has also been hypothesized the binding of Ku to DNA induces a change in Ku conformation that is required for the presentation of a high-affinity interface for subsequent contact with DNA-PK_cs (43, 46). In this study, we have compared the binding of Ku and the assembly of Ku–DNA-PK complexes to four different Ku binding sites. Our results support the latter hypothesis—that Ku-DNA binding induces a conformational change in Ku that promotes association with DNA-PK_cs. In particular, Ku binding to dsNRE1 resulted in a resistance to trypsin that was not observed in the absence of DNA, supporting the prospects for DNA-dependent changes in Ku conformation. Unexpectedly, however, Ku was found to adopt multiple conformations on DNA, each specific for a particular DNA form. Moreover, our results appear to distinguish requirements for the association of Ku with DNA-PK_cs from the activation of kinase activity.

The similarity of the Ku peptides cross-linked to dsNRE1 and upNRE1 suggests that the difference in mobility and chymotrypsin sensitivity of Ku-upNRE1 complexes by EMSA were likely due to a difference in the conformation and DNA contacts of the Ku80 subunit. Interestingly, Ku bound to upNRE1 failed to form a complex with DNA-PK_cs in EMSA. As DNA-PK_cs has been reported to bind to unstructured single-stranded DNA (38), this result would seem to emphasize the importance of Ku conformation for its association with DNA-PK_cs on DNA under stringent DNA binding conditions.

Binding of Ku to the MT element was distinguished from binding to the other DNAs by a markedly decreased mobility of the shifted complex and a unique sensitivity to Asp-N in EMSA. The lack of Mg²⁺/ATP-dependent cross-linking of Ku80 to dsMT clearly distinguished Ku80 DNA contact on MT from binding to dsNRE1. This, together with the similarity in dsMT-cross-linked Ku70 peptides following Asp-N treatment, points towards a distinct Ku80 conformation on MT. Intriguingly, these differences in Ku-DNA association had no obvious effect on the recruitment of DNA-PK_cs to the MT sequence compared to the full-length NRE1 element. However, DNA-PK_cs failed to be activated from dsMT. This could be accounted for directly by the differences noted in Ku conformation but may also have been linked to an alteration in either DNA-PK_cs conformation or the DNA contact of DNA-PK_cs, two additional possibilities that remain to be evaluated.

Lastly, the binding of Ku to dsNRE1 and DNA ends also was reflected by differences in Ku-DNA complex mobility and protease sensitivity. While the sensitivities of the two Ku-DNA complexes to the three proteases employed were similar in EMSA, there was a striking difference in the size of the Asp-N-digested Ku peptides in direct contact with dsNRE1 and DNA ends. These results are supportive of a difference in Ku70 conformation on NRE1 and DNA ends. However, the interaction of Ku with DNA is complex and may involve up to three separate Ku domains (47, 86, 89). Thus, it is also possible that the difference in cross-linked peptides reflects a difference in the region of Ku70 in contact with the two DNAs in addition to an alteration in Ku conformation on the two DNA forms.

Unexpectedly, the activation of DNA-PK_cs from DNA ends was accomplished through a process that was clearly distinguished from activation at NRE1 by the apparent lack of a requirement for Ku80 contact with the DNA. Therefore, it is possible that DNA-PK_cs kinase activity from NRE1 and DNA ends will be found to have distinct properties. However, one simpler explanation may be that Ku80 plays an additional role at NRE1 that reflects the increase in energy required to induce DNA structural transitions important for DNA-PK_cs kinase activity on an internal DNA sequence as compared to that on DNA ends.

Most current models for the role of Ku association in promoting the activation of DNA-PK_cs from DNA ends do not account for the potential contribution of the Ku ATPase and helicase activities in regulating DNA-PK_cs. Our results, in which cross-linking and DNA structural analysis were performed in both the presence and absence of Mg²⁺/ATP, indicate that the activation of DNA-PK_cs at NRE1 correlates with a Ku-dependent change in DNA structure that occurred coincidentally with the contact of Ku80 with upNRE1. However, the use of the term helicase to describe Ku in the context of the activation of DNA-PK_cs from NRE1 is likely an overstatement of the nature of the structural transition occurring at NRE1. First, the pattern of KMnO₄ modification is inconsistent with extended strand separation. Second, consistent with other reports (60, 82), we have no evidence that Ku is able to completely unwind even short double-stranded oligonucleotides with blunt ends regardless of the presence of NRE1 (28). Third, full unwinding of NRE1-containing oligonucleotides would yield the single-stranded NRE1 oligonucleotide to which DNA-PK_cs was unable to bind.

However, it does seem likely that some limited structural transition surrounding NRE1 is important for the activation of DNA-PK_cs. Two possibilities seem plausible at this time. First, the structural transition may be required to allow for the contact of Ku80 with DNA, which may complete the conformational change in Ku that is required for the activation of DNA-PK_cs at NRE1. Alternatively, it may be that the transition in DNA structure is directly important for the binding of DNA-PK_cs to DNA in an active conformation. In this regard, it is interesting to note recent reports that the treatment of DNA with the cross-linking agent cis-diamminedichloroplatinum(II) inhibits DNA-PK kinase activity without affecting the affinity of Ku for DNA ends (81). Moreover, we have determined that ethidium bromide, which is a potent inhibitor of the Ku ATPase and helicase activities, is also a potent inhibitor of DNA-PK kinase activity from both DNA ends and NRE1 (32).

A schema summarizing what presently appear to be the most likely expectations for the interaction of Ku–DNA-PK_cs with the four DNA forms examined in this study is presented in Fig. 11. In solution, DNA-PK_cs occurs in an inactive conformation and does not associate with Ku (Fig. 11A). Ku binds DNA in multiple configurations, depending on the nature of the DNA binding site, and at least on the dsNRE1, and induces a structural transition in DNA that is coincident with the DNA contact of Ku80. Ku80 also appears to directly contact the single-stranded NRE1 element, but the resultant conformation of the DNA-bound Ku is distinct from that on dsNRE1. By contrast, on MT, no transition in DNA structure is induced, and contact of Ku80 with the DNA does not occur. DNA-PK_cs can associate with Ku in three of these four configurations (Fig. 11B). However, DNA-PK_cs association leads to the activation of kinase activity from only the two DNAs, NRE1 and double-stranded DNA ends, on which appropriate Ku–DNA-PK_cs conformations are achieved.

FIG. 11.

Schematic summary of the association of Ku and DNA-PK_cs with the DNAs employed in this study. (A) In solution, DNA-PK_cs occurs in an inactive conformation (PK_I) that is attracted to Ku bound to dsNRE1 (upper left), dsMT (upper right), and double-stranded DNA ends (lower left). It does not appear to interact with Ku bound to the single, upper strand of NRE1 (ssNRE1) (lower right). The association of Ku with each of these DNAs is reflected by differences in Ku conformation, DNA conformation, and Ku-DNA contacts. In this pictogram, differences in conformation of the Ku heterodimer on DNA, as reflected by differences in protease sensitivity in EMSA and crosslinking experiments, are illustrated through differences in the shape of the Ku subunits. The Mg²⁺-dependent structural transition upstream of NRE1 is indicated by the dotted lines in the absence of definitive information on the exact structure of the DNA. The relative positioning of the Ku subunits with respect to the DNAs has been assigned arbitrarily. The DNA bound by Ku at the end is shown in grey to illustrate that it is not known whether a transition in DNA end structure is also important for the activation of DNA-PK_cs from DNA ends. (B) The association of DNA-PK_cs with Ku and DNA at NRE1 and DNA ends induces an allosteric change in the kinase that activates catalytic activity (PK_A). By contrast, when associated with Ku on the MT element, DNA-PK_cs remains in an inactive conformation. The positioning of DNA-PK_cs over the structural transition upstream of NRE1 reflects our anticipation that this change in DNA structure may be in some way directly important for activation of the kinase at NRE1. The positioning of DNA-PK_cs at the double-stranded DNA end, with Ku70 moved internally on the DNA, reflects the findings and model of Hammarsten and Chu for the assembly of Ku on DNA ends (38), with the refinement that Ku80 may not directly contact the DNA.

Finally, Ku is present in the cell in considerable excess of DNA-PK_cs and has recently been reported to associate with a variety of other cellular proteins. These newly identified Ku binding proteins include p95^vav, a hematopoietic oncogene (69); REF1, a redox factor that regulates transcription through negative calcium response elements (17); Sir4, a factor implicated in transcriptional silencing (80); TATA binding protein (26); and the bromodomains of the GCN5/CBP/TAF_II250 transcriptional coactivator proteins (4). However, the consequences of these interactions for the activity of Ku and these binding partners generally remain to be established. Moreover, by contrast to the DNA-dependent association of Ku with DNA-PK_cs, the majority of these newly defined interactions with Ku have been demonstrated in solution only. Indeed, most of these interactions were identified in yeast two-hybrid screens in which the Ku subunits employed or detected did not directly contact DNA. One exception is the interaction of Ku with the double-stranded DNA repair factor XRCC4, which appears to increase or facilitate the binding of Ku to double-stranded DNA ends. It will be interesting to determine whether this factor also promotes sequence-specific Ku binding. Since Ku is predominantly DNA bound in the nucleus, understanding how these factors may interact with DNA-bound Ku is likely to be essential to understanding the molecular basis for the physiological consequences of these protein-protein interactions.