Interaction of p53 and DNA-PK in Response to Nucleoside Analogues

INTRODUCTION

Nucleoside analogues are effective in the clinical treatment of certain types of cancer, including hematological malignancies and solid tumors (1, 2, 3, 4) . Disruption of DNA synthesis following their incorporation into DNA is believed to be the major mechanism by which nucleoside analogues exert their cytotoxic activity. 2′,2′-difluorodeoxycytidine (gemcitabine; Ref. 3 ) is an analogue of 2′-deoxycytidine, which, upon entering the cell, is converted into its triphosphate and subsequently incorporated into DNA during replication (5 , 6) . The incorporated analogue causes cessation of DNA strand elongation and resists excision by the 3′-5′ exonuclease activity associated with DNA polymerase ε (6) . The disruption of DNA replication results in the triggering of apoptosis (5 , 6) . Cells have evolved various mechanisms to maintain their genomic integrity by recognizing and repairing DNA damage and by triggering apoptosis in the damaged cells when repair fails. Recently, several proteins involved in the recognition of certain types of DNA damage have been identified (7 , 8) . However, the cellular molecules that detect the incorporated nucleoside analogues in DNA and initiate downstream cellular response are yet unknown. The nature of such sensor proteins and the mechanism by which they signal for DNA repair or apoptosis are important subjects of investigation.

A possible candidate as the sensor of nucleoside analogue-induced DNA strand disruption is DNA-PK3 /Ku, a member of the PI-3 kinase family. The DNA-PK/Ku complex is known to recognize DNA strand breaks and regulate the activity of a number of proteins that interact with DNA (9, 10, 11, 12) . The complex consists of a catalytic subunit, DNA-PKcs, and the Ku regulatory subunits (13 , 14) . The M_r 460,000 DNA-PKcs requires the presence of DNA ends for activation (15) . Ku is a heterodimer (Ku70/Ku86) capable of binding DNA strand breaks, nicks, gaps, and stem-loop structures (16 , 17) . Both subunits of Ku are required for stable binding to DNA (18, 19, 20, 21) . The Ku heterodimer appears to stabilize the binding of DNA-PKcs to the DNA ends. DNA-PK, upon activation, phosphorylates a number of proteins, including Ku, p53, replication protein A, topoisomerases, and RNA polymerases (13 , 14 , 22) .

The tumor suppressor protein p53 is one of the substrates of DNA-PK. This molecule has been shown to accumulate and become activated as a transcription factor in response to DNA-damaging agents via posttranscriptional and posttranslational mechanisms (23, 24, 25) . Phosphorylation of p53 is a key event that regulates the stability and transcriptional activity of this molecule in response to DNA damage (26 , 27) . DNA-PK has been shown to phosphorylate human p53 on Ser15 and Ser37 in vitro (28) . Phosphorylation of p53 on Ser15 decreases its interaction with MDM2 (a protein that targets p53 for proteolytic degradation) and thus stabilizes the protein (29) . Site-specific mutation of Ser15 to alanine appears to impair the transactivation function of p53 and to compromise the G₁-S cell cycle checkpoint mechanism (30) . Fibroblasts from scid mice with defective DNA-PKcs exhibit a delay in their p53 responsiveness to DNA-damaging agents (31) . It has been suggested that DNA-PK acts upstream of p53 in response to DNA damage and appears to be necessary for the DNA-binding activity of p53 in certain experimental systems (32) . However, the importance of DNA-PK in p53 phosphorylation and activation still remains controversial. For instance, phosphorylation of Ser15 and the p53-mediated events in response to DNA damage seem to be normal in murine scid (33, 34, 35, 36, 37) and slip (38) cells defective in DNA-PK, suggesting the presence of other kinases capable of phosphorylating p53 and regulating its function. It has been shown that the cell cycle checkpoint molecules CHK1 and CHK2 (39 , 40) , and ATM and ATR, two other PI-3 kinase family members, may play important roles in p53 phosphorylation and cell cycle regulation in response to DNA damage (41, 42, 43, 44) . Furthermore, several studies suggest that p53 protein itself may be directly involved in recognizing DNA damage through its COOH-terminal region, which contains a nonspecific DNA-binding domain with a preferential binding affinity to damaged DNA (45, 46, 47) .

Our recent studies showed that cells with wild-type p53 were more sensitive to the cytotoxic effects of nucleoside analogues. Treatment of ML-1 cells (wild-type p53) with gemcitabine caused an accumulation of p53 protein and its localization to the nucleus, which was associated with apoptosis induction (48) . In this study, we investigated the possibility that p53 and DNA-PK/Ku may interact and play a role in sensing the gemcitabine-induced DNA damage and in signaling apoptosis.

MATERIALS AND METHODS

Cells, Chemicals, and Reagents.

The human myeloid leukemia cell line ML-1 was maintained in exponential growth in RPMI 1640 suspension culture supplemented with 10% FBS. Gemcitabine (2′,2′-difluorodeoxycytidine) was kindly supplied by Dr. L. W. Hertel of Lilly Research Laboratories. The T4 polynucleotide kinase and the large fragment of Escherichia coli DNA Polymerase I (Klenow fragment) were obtained from United States Biochemicals Co. (Cleveland, OH). DNA polymerases α and ε were purified by serial chromatographic separations as described in Huang et al. (49) . [γ-³²P]ATP was from ICN Radiochemical (Irvine, CA). The single-stranded M13mp 18(+) DNA and the dNTPs were purchased from Amersham Pharmacia Biotech, Inc. (Piscataway, NJ). Synthetic oligomers containing either normal deoxynucleotides or analogue dFdCMP were prepared by Genosys (Woodland, TX). The anti-DNA-PK antibody and the anti-p53 antibody (Ab-6) were purchased from Oncogene Research Products (Cambridge, MA). Ku70 and Ku86 antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA). The antibody for p53 Ser15 was purchased from New England Biolabs (Beverly, MA). The FITC-conjugated anti-p53 antibody was purchased from Santa Cruz Biotechnology (Santa Cruz Biotechnology, CA).

EMSA for ADBA.

Protein extracts from ML-1 cells were first precipitated with 60% saturated (NH₄)₂SO₄ dissolved in buffer A (see below) and fractionated by FPLC through a Resource Q column (Amersham) under the following conditions: buffer A, 20 mm potassium phosphate (pH 7.5); buffer B, 700 mm NaCl in buffer A; flow rate, 1 ml/min; sample loading, 10 min in buffer A; washing, 100% buffer A for 5 min; elution, a linear gradient of 0→100% buffer B over 20 min followed by a 5-min elution in 100% buffer B. Samples were collected at 1 ml/fraction. The fractions were assayed by EMSA for activity that binds the gemcitabine-containing DNA, using the following ³²P-labeled DNA hybrid as the probe:

5′ ³²P-GTAAAACGACGGCCAGTDG

3′-CATTTTGCTGCCGGTCAGCCACACA

(25-mer template)

The letter “D” indicates the position of the drug molecule (dFdCMP). A nonradioactive DNA hybrid with an identical nucleotide sequence (except that “D” was replaced with normal dCMP) was used as the competing DNA. A typical protein/DNA-binding mixture (20 μl) contained 20 mm Tris-HCl (pH 7.5), 5 mm MgCl₂, 10 mm NaCl, 5% glycerol, 50 μg/ml BSA, 5 μl of protein fraction from FPLC, the ³²P-labeled dFdCMP-DNA probe, and 50-fold of competing DNA. After being incubated at room temperature for 30 min, the samples were analyzed on a 6% nondenaturing polyacrylamide gel. The bands in the dried gel were visualized by using autoradiography. The fractions containing the ADBA were pooled and further separated by FPLC through a Mono Q column under the following conditions: buffer C, 20 mm Tris-HCl (pH 7.8), 0.2 mm DTT; buffer D, 700 mm NaCl in buffer C; flow rate, 1 ml/min; sample loading, 10 min in buffer C; washing, 100% buffer C for 5 min; elution, a linear gradient of 0→100% buffer D over 20 min, followed by 5 min 100% buffer D. Samples were collected at 1 ml/fraction. The fractions were assayed for ADBA using EMSA as described above.

Immunoprecipitation and Immunoblot Analysis.

For immunoprecipitation experiments, ML-1 cells were exposed to a 10-Gy ionizing radiation and then incubated for 3 h to induce accumulation of p53 protein. Cell lysates were prepared in buffer containing 300 mm NaCl, 0.5% sodium deoxycholate, 25 mm HEPES (pH 7.5), 20 mm glycerol phosphate, 0.1% SDS, 1 mm orthovanadate, 0.5 mm DTT, 1.5 mm MgCl₂, 1.2 mm EDTA, 1% Triton X-100, and a mixture of protease inhibitors. The cell lysates (1 μg protein/μl) were incubated with the specific antibodies (1 μg antibody/1 mg protein) at 4°C overnight with gentle rotation, followed by incubation with protein A-agarose conjugate (for mouse IgG against human DNA-PK or p53) or protein G-agarose conjugate (for goat IgG against human Ku70 or Ku86) for an additional 2 h at 4°C. The pellets were collected and washed three times with ice-cold PBS. The supernatants and the precipitated protein were subjected to SDS-PAGE followed by immunoblotting for the specific proteins using the respective antibodies. For analysis of cellular p53 and DNA-PK, protein lysates were prepared from the control and drug-treated cells using a lysis buffer consisting of 50 mm Tris-HCl (pH 6.8), 10% glycerol, 2% SDS, 0.025% bromophenol blue, and 0.025% β-mercaptoethanol. The protein lysates were heated at 100°C for 5 min and then separated by SDS-PAGE. A 10% gel was used for separating p53 and Ku 70 and an 8% gel for DNA-PK and Ku 86. The proteins were transferred to a nitrocellulose membrane and probed with the respective antibodies.

Localization of p53 and DNA-PK Proteins by Immunocytochemistry.

Exponentially growing ML-1 cells were treated with gemcitabine for various times and then spun onto glass microscope slides, which were precoated with Histogrip (Zymed Laboratories, Inc., San Francisco, CA). The samples were immediately fixed with 5% formaldehyde in acetone for 20 min. Cells were permeabilized in 100% methanol for 10 min and rehydrated in 90% methanol for 30 min. The cells on the slides were incubated with the p53 antibody (Ab-6; Oncogene), and the signal was revealed using the Histostain SP kit as recommended by the manufacturer (Zymed). To detect p53 and DNA-PK simultaneously, the cells on the slides were first incubated with 5% Rabbit IgG (Zymed) in PBS to block all nonspecific sites. This was followed by coincubation with a FITC-conjugated p53 antibody (mouse IgG) and a DNA-PK antibody (goat IgG) at 37°C for 2 h, followed by incubation with a Rhodamine-labeled antigoat IgG secondary antibody (Zymed) at 37°C for 1 h. The fluorescent signals from p53 (green) and DNA-PK (red) were analyzed using a laser scanning confocal microscope (Carl Zeiss Inc., Thornwood, NY).

DNA Primer Extension and Nucleotide Excision Assays.

The in vitro DNA primer extension assay using purified DNA polymerases was performed as described previously (49) . The 17-base primer (5′-GTAAAACGACGGCCAGT-3′) was labeled at its 5′-terminus with [γ-³²P]ATP using the T4 polynucleotide kinase and then annealed to a 24-base synthetic DNA template (5′-TACACACACTGGCCGTCGTTTTAC-3′). This DNA hybrid was used as the primer/template for polymerization by human polymerases α and ε and by the E. coli pol I. The reaction mixtures contained 20 mm Tris-HCl (pH 7.5), 5 mm MgCl₂, 0.5 mm DTT, 10 mm NaCl, 20 μg BSA/ml, and 10 μm each of dATP, dCTP, dGTP, and dTTP. After being incubated at 37°C for 30 min, the reaction products were analyzed by electrophoresis through a 15% polyacrylamide sequencing gel.

The following DNA constructs were used as the substrates for measuring the ability of the p53/DNA-PK protein complex to remove normal nucleotides or the analogue from DNA.

(1) Matched normal nucleotides:

(2) Normal nucleotides with two mismatched nucleotides at the 3′-end:

(3) Analogue (dFdCMP) at the 3′-penultimate site and a matched nucleotide pair at the 3′-end:

The letter “D” indicates the position of the drug molecule. The reaction mixtures contained 20 mm Tris-HCl (pH 7.5), 5 mm MgCl2, 0.5 mm DTT, 10 mm NaCl, the DNA substrate, and protein from FPLC fractions. The reaction products were analyzed by electrophoresis through a 15% denaturing polyacrylamide gel as described previously (6 , 49) .

RESULTS

Identification of an ADBA Containing DNA-PK/Ku and p53.

Incorporation of therapeutic nucleoside analogues into DNA causes disruption of DNA replication and triggers apoptosis (5) . Thus, recognition of the incorporated analogues in DNA and the subsequent signaling events constitute major cellular responses to the drug exposure. We first used an EMSA with a DNA construct containing dFdCMP at position 18 (3′-penultimate; see “Materials and Methods”) as the probe to detect potential sensor molecules that interact with the analogue-containing DNA in preference to normal competing DNA. The DNA probe containing dFdCMP at the 3′-penultimate position of the primer was labeled with ³²P at the 5′-end. Protein extracts of ML-1 cells treated with gemcitabine (1 μm for 3 h) were fractionated by using FPLC through a Resource Q column, and the fractions were analyzed for the ADBA by EMSA. To minimize the effect of nonspecific DNA-binding proteins, a 50-fold competing oligonucleotide with a sequence identical to that of the ³²P-labeled probe, but with the dFdCMP substituted by a normal dCMP, was included as competing DNA in the assay. As illustrated in Fig. 1A⇓ , the EMSA revealed that fractions 22–26 contained ADBA that caused a shift in the migration of the drug-containing probe at several positions, suggesting that multiple protein components might be involved in binding the analogue-containing DNA.

Every fraction between 20–28 was further analyzed by EMSA and Western blot analyses in an attempt to identify the protein components with the ADBA. Consistent with Fig. 1A⇓ , fractions 22–27 exhibited the ADBA by EMSA (data not shown). These fractions also contained p53 protein, Ku70, and Ku86 (Fig. 1B)⇓ . The catalytic subunit DNA-PKcs was observed in fractions 22, 24, and 25, but for a yet unknown reason was not detected in fraction 23. These data suggest a possibility that DNA-PK/Ku and p53 may be the components of a complex that binds the analogue-containing DNA. Interestingly, in contrast to fraction 24, which contained all four proteins (p53, Ku70, Ku86, and DNA-PK), fraction 26 appeared to contain a second peak of p53, Ku70, and Ku86, but without DNA-PK. This suggests that these proteins may form various complexes containing different protein combinations. The DNA-PKcs subunit appeared to be in limited amount and was missing from fraction 26.

A binding competition assay was used to evaluate the relative binding affinity of the protein fractions to the analogue-containing DNA probe in comparison with the normal DNA probe. The protein fractions from the Resource Q column containing ADBA were pooled and further fractionated through a second FPLC column (Mono Q HR5/5). After analysis by EMSA (data not shown), fraction 20, which contained the highest ADBA, was used in the binding competition assay. As shown in Fig. 2⇓ , about 10-fold of normal competing DNA was required to reduce 50% of the binding of ADBA when the radioactive analogue DNA was used as the probe. In contrast, when normal ³²P-DNA was used as the probe, only a 1:1 ratio of competing DNA:probe was needed to reduce the binding activity of ADBA by 50%. These data suggest that the FPLC-enriched ADBA fraction had a higher affinity to the analogue-containing DNA. However, this specificity is only relative because a 100-fold competing DNA was able to reduce the binding to the analogue DNA by 85% (Fig. 2)⇓ .

Several approaches were used to further identify the molecules that constitute the protein complex that interacts with the analogue containing DNA. Protein fractions containing the ADBA as demonstrated by the EMSA were subjected to SDS-PAGE, and the protein bands were visualized by staining with Coomassie Blue. As shown in Fig. 3A⇓ , multiple protein bands were seen in the fraction from the first column (Resource Q; Fig. 3A⇓ , Lane 2). There were only three major protein bands with apparent molecular weights of M_r 70,000, 53,000, and 40,000, respectively, present in the fraction following a second round of purification through a Mono Q column (Fig. 3⇓ , Lane 3). In a second experiment, the gel slice containing the major shifted band (band II) in the EMSA (nondenaturing gel) was excised and directly loaded onto a 10% denaturing gel. After electrophoresis, the gel was silver-stained for protein. Two protein bands were revealed at M_r 70,000 and 53,000 positions, whereas the M_r 40,000 protein was absent (Fig. 3B⇓ , Lane 3). These results suggest that the M_r 70,000 and 53,000 proteins are closely associated with the ADBA and, taken together with the results from the immunoblot analysis (Fig. 1B)⇓ , suggest that Ku70 and p53 are likely the molecules that directly bind to the analogue DNA.

Complex Formation between DNA-PK/Ku and p53 in Vitro.

Immunoprecipitation was used to further test whether p53 might physically interact with DNA-PK/Ku. DNA-PK was first immunoprecipitated from whole cell lysates with DNA-PK antibody (mouse IgG; Oncogene), and the precipitate was tested for presence of p53 by immunoblot analysis using the Ab-6 antibody (mouse IgG; Oncogene). As seen in Fig. 4A⇓ , Lane 6, p53 can be coimmunoprecipitated with DNA-PK. The protein detected along with p53 (the band that migrates slightly slower than p53) is the IgG heavy chain (Fig. 4A⇓ , Lane 6). A separate immunoprecipitation experiment was performed to further confirm the association between p53 and DNA-PK. p53 was immunoprecipitated from the cell lysates using the Ab-6 p53 antibody (mouse IgG), washed, and assayed for Ku70 using anti-Ku70 antibody (goat IgG; Santa Cruz Biotechnology). The results indicate that p53 may physically interact with Ku70 (Fig. 4B⇓ , Lane 6). In a reverse experiment, immunoprecipitation of p53 followed by immunoblotting for DNA-PKcs also showed the association of these two molecules (data not shown). These data suggest that DNA-PK/Ku and p53 are able to form a complex, at least in vitro.

Phosphorylation of p53 and Its Colocalization with DNA-PK in Cells Treated with Gemcitabine.

Several approaches were used to investigate the change in p53 protein levels, its phosphorylation status, and its interaction with DNA-PK in whole cells after treatment with gemcitabine. Immunoblot analysis of the protein extracts from ML-1 demonstrated a time-dependent increase of DNA-PK and p53 protein after exposure to gemcitabine (1 μm). This was associated with a significant increase of phosphorylation of p53 at Ser-15 (Fig. 5A)⇓ . Earlier studies have demonstrated that Ser15 of p53 is phosphorylated by DNA-PK (28) .

We next analyzed the cellular localization of p53 protein in response to treatment with gemcitabine. Gemcitabine-treated (1 μm) ML-1 cells were immunostained for p53. As shown in Fig. 5B⇓ , the proportion of cells with positive nuclear p53 stain increased within 1 h following drug treatment. The proportion of p53-positive cells was further elevated as the exposure time to the drug was prolonged. For instance, the portion of p53-positive cells increased to ∼30–40% at 3 h after drug incubation. Under a higher magnification, it was observed that cells with high level of nuclear p53 showed apoptotic morphology, whereas cells lacking nuclear p53 appeared morphologically intact.

A double immunostaining technique was then used to determine whether DNA-PK might by colocalized with p53 following treatment with gemcitabine. Cells were stained with a FITC-labeled p53 antibody (mouse IgG, green fluorescence) and a DNA-PK antibody (goat IgG) in the presence of rhodamine-labeled antigoat IgG (red fluorescence). Confocal microscopic analysis showed that the control (untreated) ML-1 cells had low levels of p53 and DNA-PK signals. However, after cells were treated with gemcitabine (1 μm for 3 h), ∼30% of the cells showed strong DNA-PK and p53 signals in the nuclei of the same cells (Fig. 5C)⇓ . Overlay of the two individual images indicated that p53 and DNA-PK were colocalized in the nuclei, as evidenced by the yellow color on the superimposed image. This is consistent with the fact that there were ∼30% of ML-1 cells in the S-phase, which were active in DNA synthesis and capable of incorporating nucleoside analogues into DNA.

Effect of ADBA on DNA Replication and Nucleotide Excision.

A DNA primer extension assay was used to evaluate the effect of the FPLC-enriched ADBA on DNA synthesis in vitro. Human DNA polymerases α and ε and the E. coli polymerase I were used to catalyze the extension of a 17-base primer annealed to a 24-base template. As shown in Fig. 6A⇓ , in the presence of four dNTPs (10 μm each of dATP, dCTP, dGTP, and dTTP) each enzyme was able to effectively elongate the 17-base primer to the 24-base product (Fig. 6A⇓ , Lanes 2, 5, and 7). Because pol ε and pol I also possess a 3′-5′ exonuclease activity, excision products (shorter DNA bands below the 17-base position) were also detected (Fig. 6A⇓ , Lanes 5 and 8). Addition of ADBA to the reactions almost completely suppressed the DNA synthesis activity of pol α (Fig. 6A⇓ , Lane 3). The DNA polymerization activities of pol ε and pol I were also significantly inhibited (Fig. 6A⇓ , Lanes 6 and 8). Interestingly, the 3′-5′ exonuclease activity of pol ε was only moderately inhibited, as evidenced by the presence of the excision band (Fig. 6A⇓ , Lane 6). No inhibition of the 3′-5′ exonuclease activity of pol I was observed (Fig. 6A⇓ , Lane 8). The absence of the excision band in Lane 7 of Fig. 6A⇓ was due to its rapid extension to the 24-base product by the potent polymerization activity of pol I in presence of dNTPs and the absence of ADBA. It appears that ADBA might bind to the DNA ends with a higher affinity and thus block the chain elongation.

Because ADBA showed a preferential inhibition of DNA chain elongation with only moderate effect on the 3′-5′ exonuclease activity of pol ε, we further tested whether the p53 protein in the ADBA complex still retained its 3′-5′ exonuclease activity. Three DNA hybrids (one with all matched normal deoxynucleotides between the primer and template strands, one with two mismatched deoxynucleotides at the 3′-end of the primer strand, and one with an analogue dFdCMP at the 3′-penultimate position and a matched deoxynucleotide at 3′-end of the primer strand; see “Materials and Methods”) were used as the substrates for excision. As shown in Fig. 6B⇓ , the ADBA fraction revealed a 3′-5′ exonuclease activity that preferentially removed the mismatched nucleotides from DNA, as evidenced by the appearance of two DNA bands (18-mer and 17-mer) below the intact 19-base primer (Fig. 6B⇓ , Lanes 5 and 6). In contrast, no excision was seen in samples with the perfectly matched DNA construct (Fig. 6B⇓ , Lanes 1–3). This is consistent with the property of the p53 exonuclease that prefers mismatched nucleotides as its substrate (50, 51, 52) . Interestingly, when the DNA primer contained a “matched” dFdCMP at the penultimate position and a matched normal nucleotide at the 3′-end, no significant excision was observed (Fig. 6B⇓ , Lanes 7–9). These data suggest that although the 3′-5′ exonuclease of p53 in the ADBA complex remained active (Fig. 6B⇓ , Lanes 5–6), it was unable to remove the drug molecule (dFdCMP) from DNA.

DISCUSSION

Cells are equipped with complex molecular machinery that detects DNA damage and subsequently initiates DNA repair. It is also speculated that if repair fails, apoptosis is triggered by a p53-dependent pathway to eliminate the damaged cells (53) . These processes together ensure the stability of the genome. Many anticancer agents induce cellular cytotoxicity by causing DNA damage. Nucleoside analogues, including gemcitabine, are known to be converted to triphosphates intracellularly and get incorporated into replicating DNA (5 , 6) . This incorporation disrupts further DNA strand elongation and ultimately leads to the induction of apoptosis (5 , 6) . However, the molecules that recognize and respond to the DNA damage caused by the incorporated analogues in DNA largely remain to be characterized. This study has demonstrated that a complex consisting of DNA-PK/Ku and p53 was able to bind DNA containing the anticancer nucleoside analogue gemcitabine monophosphate in vitro, with a relatively high affinity compared with its binding to normal DNA. Upon treatment of ML-1 cells with gemcitabine, both DNA-PK and p53 colocalized to the nuclei, apparently in those cells in the active DNA synthesis phase, which allows the analogue to become incorporated into nuclear DNA. There was a simultaneous accumulation of p53 protein and its phosphorylation on Ser15, which presumably contributes to the stabilization of p53 protein in these cells. These events coincided with the induction of apoptosis. These data together suggest that the DNA-PK/Ku/p53 complex may function as a sensor of DNA strand disruption caused by incorporation of the drug molecules and subsequently signal to trigger apoptosis.

The DNA-PK protein has been shown to be involved in recognizing and binding to both single- and double-strand breaks in DNA (9 , 12) . Once Ku70 and Ku86 subunits of DNA-PK bind to DNA, they interact with DNA-PKcs and induce its kinase activity, resulting in the phosphorylation of a variety of protein substrates including those involved in DNA damage responses such as p53, RPA, and RNA polymerase (13 , 14 , 22) . The phosphorylation of p53 by DNA-PK on Ser 15 and Ser 37 has been demonstrated in vitro (28) , although its role in phosphorylating and activating p53 in whole cells remains uncertain (32 , 38) . The colocalization of p53 and DNA-PK to the nucleus along with the enhanced phosphorylation of p53 on Ser 15 following treatment with gemcitabine (Fig. 5)⇓ suggests that DNA-PK might play a role in phosphorylating p53 in whole cells after exposure to the nucleoside analogue. However, it should be noted that Ser 15 of p53 has also been implicated as the site for phosphorylation by other molecules such as ATM (41 , 42) , suggesting either a redundancy in function between these kinases or a distinct role for each in response to different types of DNA damage. Earlier reports have suggested that in response to irradiation, the ATM kinase activates p53 for cell cycle arrest whereas DNA-PK activates p53 for apoptosis (54) .

Phosphorylation of p53 protein results in the stabilization of this molecule and the subsequent transcriptional activation of several genes that are targets of p53 and are involved in cell cycle arrest or apoptosis. This functional relationship between DNA-PK and p53, together with the ability of p53 exonuclease activity to preferentially remove mismatched nucleotides from DNA, suggest that they may be able to interact and respond to certain types of DNA damage. The observations that treatment of cells with the gemcitabine caused the colocalization of DNA-PK and p53 in the nuclei (Fig. 5C)⇓ and an increase of p53 phosphorylation at Ser15 (Fig. 5A)⇓ are consistent with their sensing and signaling function. When protein extracts prepared from the control cells (without gemcitabine treatment) were used in the EMSA assay a lower but detectable ADBA was observed (data not shown). It is possible that the sensor complex consisting of DNA-PK/Ku and p53 may normally maintain a certain level of activity in response to background DNA damage, such as strand breaks induced by endogenous free radicals or those generated during repair or homologous recombination.

While p53 and the DNA-PK/Ku components all copurify with the ADBA (Fig. 1B)⇓ , it is not clear at the present time which component(s) directly interact(s) with the analogue-containing DNA site. Several lines of evidence suggest that Ku70 and p53 may be the molecules that interact directly with the DNA strand containing the incorporated analogue. First, the M_r 70,000 and 53,000 protein bands are consistently detected in the fractions that show high ADBA following fractionation through two consecutive FPLC columns (Fig. 3A)⇓ . Second, analysis of the proteins in the gel slice that contained the shifted band in the EMSA assay using the analogue DNA as the probe revealed the M_r 70,000 and 53,000 bands (Fig. 3B)⇓ . Third, immunoblotting using specific antibodies detected the presence of Ku70 and p53 in the fractions that showed binding activity to the analogue-containing DNA (Fig. 1B)⇓ . In addition, coimmunoprecipitation experiments demonstrated a physical association between p53, DNA-PKcs, and Ku proteins (Fig. 4)⇓ . However, it is yet unknown which subunit of the DNA-PK complex mediates this interaction with p53. In future studies, genetically altered cells that lack one of the components of the complex (e.g., DNA-PKcs−/−, Ku70−/−, or Ku80−/− cells) would be useful in further elucidating the detailed interaction between p53 and the DNA-PK/Ku complex.

It is of interest to note that the binding of the ADBA protein complex to DNA effectively suppressed DNA polymerization activity of pol α, pol ε, and pol I in vitro, whereas the 3′-5′ exonuclease activities associated with the DNA polymerases and p53 remained relatively active (Fig. 6)⇓ . This suggests a possibility that the complex might play a role in the recognition of mismatched nucleotides incorporated into DNA during replication. A role for p53 in the preferential removal of mismatched nucleotides from the 3′-end of DNA through its 3′-5′ exonuclease activity has been observed (50 , 51 , 55) . Binding of the DNA-PK/Ku/p53 complex to the replication site with a mismatched nucleotide would stall further strand elongation and may allow the 3′-5′ exonuclease activity of p53 to remove the unpaired nucleotide. It has been shown that the DNA replication complex pauses upon encountering a misincorporated nucleotide at the 3′-end, and, without dissociating from the DNA, allows the excision of the misincorporated nucleotide from the 3′-end before resuming further strand elongation (56) .

It is known that after dFdCMP is incorporated into DNA, the analogue causes a cessation of DNA strand elongation after one more nucleotide is added at the 3′-end (6) . It is possible that the termination of the daughter DNA strand might resemble a single-strand break or an abnormal DNA end. This may serve as a structural basis for activation of DNA-PK/Ku, which in turn phosphorylates p53 in whole cells. It has been reported that the Ku/DNA-PK complex is able to bind DNA strand breaks (double- and single-stranded), gaps, and stem-loop structures (10, 11, 12, 13 , 17 , 18) . Unlike the mismatched nucleotides that can be readily repaired, the analogue (dFdCMP) in DNA is resistant to excision by the 3′-5′ exonuclease activity of p53 (Fig. 6B)⇓ and pol ε (6) . Failure to remove the analogue would result in the persistence of the stalled DNA end and a prolonged induction of DNA-PK activity, which in turn phosphorylates p53 and activates the downstream pathways, leading to apoptosis. This may explain the accumulation of DNA-PK and p53 in the nuclei and the subsequent apoptosis of the same cells after gemcitabine incubation (Fig. 5, B and C)⇓ . The induction of apoptosis by gemcitabine and its possible association with p53 activation have been observed in various experimental systems (48 , 57) .

In summary, p53 and DNA-PK/Ku appear to be able to form a complex that recognizes nucleotide analogue-containing DNA. Binding of the protein complex to the analogue-damaged DNA blocked further DNA chain elongation but failed to remove the incorporated analogue from DNA. It appears that the continuous existence of the analogue-stalled DNA end induced the kinase activity of DNA-PK/Ku, which subsequently phosphorylated p53. This phosphorylation resulted in the stabilization and activation of p53, which possibly triggered downstream pathways that culminated in apoptotic cell death. In future studies, it would be important to determine whether p53 has a more global role in recognizing DNA damage caused by radiation, reactive oxygen species, and alkylating agents.