In Prostate Cancer Cells the Interaction of C/EBPα with Ku70, Ku80, and Poly(ADP-ribose) Polymerase-1 Increases Sensitivity to DNA Damage*

Prostate cancer cell lines were examined for proteins that partnered with the transcription factor C/EBPα by use of a pull-down assay with S-tagged C/EBPα combined with matrix-assisted laser desorption ionization time-of-flight mass spectroscopy analysis. Ku70, Ku80, and poly(ADP-ribose) polymerase-1 (PARP-1) were identified as proteins that associated with C/EBPα. The physical interaction of C/EBPα with these partner proteins was further demonstrated by glutathione S-transferase (GST) pull-downs using purified protein expressed in Escherichia coli. The strongest binding was between C/EBPα and PARP-1. Immunoprecipitation of C/EBPα expressed in prostate cancer cells co-precipitated Ku70, Ku80, and PARP-1. Deletion analysis of C/EBPα indicated that the C terminus of C/EBPα was essential for the interaction of C/EBPα with Ku70, Ku80, and PARP-1. Functional analysis of the interaction between C/EBPα and the Ku proteins as well as PARP-1 showed that cells exhibiting these interactions had increased radiation sensitivity and decreased ability to repair double strand DNA breaks. Deficient DNA repair was dependent on the prostate cancer cell line tested, suggesting a complex process. We conclude that the association of C/EBPα with Ku proteins and PARP-1 raises the likelihood that C/EBPα-expressing prostate cancer cells may be more sensitive to DNA-damaging agents and may be important in the design of new prostate cancer therapies.

C/EBP␣ is a transcription factor belonging to a basic region-leucine zipper (Bzip) protein family (1)(2)(3). The intronless gene creates multiple protein isoforms with molecular weights of 42, 30, and 20 kDa by the differential use of multiple AUG initiation codons within the same open reading frame of a single mRNA (4). The 42-kDa isoform acts to inhibit cell growth and stimulate terminal differentiation. The truncated forms are generated through the mechanism of leaky ribosome scanning, and these isoforms may be important in aging and reaction to stress (5)(6)(7).
C/EBP␣ is expressed in many tissues including hepatocytes, adipose tissue, lung, small intestine, skin, mammary gland, adrenal gland, hematopoietic cells, and ovaries. C/EBP␣ plays essential roles in energy homeostasis and the differentiation of white adipose tissue and granulocytes (8 -11). Mutations in the C/EBP␣ gene in some patients with acute myeloid leukemia give rise to proteins that are dominant negative and that impair myeloid differentiation (12). In leukemias with the (8:21) translocation, AML1-ETO blocks granulocytic differentiation by down-regulation of CEBP␣ (13,14). Reduced expression of CEBP␣ has been observed in lung and skin cancer tissue, raising the possibility that in some tissues CEBP␣ may act as a tumor suppressor gene (15).
C/EBP␣ has been well demonstrated to act as an antiproliferation factor in a number of tissues. In animal models, the expression of C/EBP␣ is transiently decreased in regenerating liver after partial hepatectomy, and hepatocytes from C/EBP␣ knock-out mice manifest increased proliferation in culture. In several cancer cell lines, enforced expression of C/EBP␣ by transfection causes significant growth inhibition or arrest (16 -18). The mechanisms of the antiproliferation effects of C/EBP␣ depend upon its interactions with cell cycle-related proteins (19). In other cell types the antiproliferative effects of C/EBP␣ result from C/EBP␣ binding to and inhibiting E2F; in hepatocytes, both inhibition of E2F and stabilization of p21 contribute to the growth arrest. Growth arrest by C/EBP␣ involves the stabilization of the cyclin-dependent kinase inhibitor p21 (20,21), interaction with cdk2 and cdk4 (22,23), disruption of E2F protein complexes (24,25), and repression of c-Myc expression by interacting with the E2F binding site in the c-Myc promoter (26). Recent studies indicate that the inhibition of cell proliferation by C/EBP␣ is contingent upon its phosphorylation status, which when altered may allow C/EBP␣ to stimulate cell growth. In hepatocytes, activated phosphatidylinositol 3-kinase/ AKT led to the nuclear accumulation of protein phosphatase 2A, which dephosphorylated C/EBP␣ at . The dephosphorylated C/EBP␣ interacted with and sequestered the retinoblastoma protein (Rb), decreasing E2F-Rb complexes with a consequent acceleration of cell growth (27,28). Another potential regulatory point for the anti-proliferation properties of C/EBP␣ is suggested by the interaction of C/EBP␣ with the chromosome remodeling factor, SWI/SWF, which is necessary for C/EBP␣-mediated growth arrest (29).
In both normal and cancerous prostate epithelia the expression of C/EBP␣ has been verified at the RNA level (30,31). We have observed that C/EBP␣ is differentially expressed between normal and cancerous prostate epithelia, that in prostate cancer cells C/EBP␣ down-regulated transcription of prostate-specific antigen, a marker of prostate differentiation (32), and that overexpression of C/EBP␣ stimulated proliferation of prostate cancer cell lines (33). These results suggest that C/EBP␣ might have tissue-and cell-specific functions in prostate epithelium. To determine whether C/EBP␣ exhibited unique functions in prostate cells, we screened prostate cancer cell lines for proteins that bound to C/EBP␣ and identified three DNA repair proteins, Ku80, Ku70, and PARP-1 as C/EBP␣ partner proteins. The interaction of C/EBP␣ with PARP-1, Ku80, and Ku70 interfered with the non-homologous ending joining (NHEJ) 2 DNA repair and contributed to increased sensitivity of prostate cells to radiation and DNA damage from bleomycin and iron. under a LSUHSC Institutional Review Board-approved protocol from the Feist-Weiller Cancer Center tissue repository at LSUHSC. The tissue protein was extracted with NE-PER nuclear and cytoplasmic extraction reagents (Pierce).
RNA Isolation and Microarray Analysis-RNA was isolated from retrovirus-transduced cells with TRI reagent-RNA/DNA/protein isolation reagent (Molecular Research Center, Inc., Cincinnati, OH) according to the manufacturer's protocol. The expression analysis with Affymetrix gene chip was conducted on the human U95A array (Affymetrix Inc., Santa Clara, CA) using 10 g of total RNA. Synthesis of cRNA and subsequent hybridization was completed by the Research Core Facility at LSUHSC-S using the standard Affymetrix protocols. The human U95A array represents 12,256 oligonucleotides of known genes or expression tags. The raw data were collected and analyzed with the Affymetrix Microarray Suite with the scale set at 2500.
Western Blot Analysis-Whole cell extracts from prostate cancer cell lines were obtained with radioimmune precipitation assay buffer (phosphate-buffered saline, 1% Nonidet P-40, 0.25% sodium deoxycholate, 0.1% SDS) containing 1ϫ protease inhibitor mixture (Roche Applied Science). Protein concentration was determined by BCA protein assay kit (Pierce). Cell proteins were separated by electrophoresis on SDS-PAGE, transferred to Hybond TM ECL nitrocellulose membrane (Amersham Biosciences), and blocked with 5% nonfat milk in 1ϫ TBST (10 mM Tris-HCL, pH 8.0, 150 mM NaCl, 0.05% Tween 20). The blots were then incubated at room temperature with rabbit anti C/EBP␣ antibody for 2 h, washed, and incubated with peroxidase-conjugated secondary antibody. The signal was detected with SuperSignal Dura Substrate (Pierce).
NHEJ Assay-The NHEJ assay was conducted according to previous work with slight modifications (34 -36). Briefly, the nuclear extract was prepared as above with NE-PER nuclear extraction reagents and dialyzed against 20 mM Tris-HCL pH 8.0, 100 mM potassium acetate, 10% glycerol, 0.5 mM EDTA, and 1 mM dithiothreitol. Then 20 -25 g of nuclear protein was mixed with 400 ng of BamH1-and XhoI-digested pBlueKs plasmid DNA in 50 mM Tris-HCL, pH 8.0, 40 mM potassium acetate, 2 mM magnesium acetate, 0.5 mM EDTA, 1 mM ATP, 1 mM dithiothreitol, 200 mM dNTPs, and 100 mg/ml bovine serum albumin. The 50-l reactions were incubated at 37°C for 2 h followed by protease K treatment and extraction with phenol/chloroform. Onehalf of each reaction was separated on a 0.8% agarose gel and stained with Gelstar (Cambrex Corp., East Rutherford, NJ). The stained gels were scanned, and the percentage of end rejoining was calculated by dividing the sum of dimer and multimer by sum of monomer, dimer, and multimer.
S-tagged C/EBP␣ Protein Capture and MALDI-TOF-MS-S-tagged C/EBP␣ fusion proteins were expressed in pET-30A-C/EBP␣ in E. coli BL21-CodonPlus (DE3)-RIL cells after induction with 0.8 mM isopropyl 1-thio-␤-D-galactopyranoside and purified with S-tagged agarose beads. The fusion proteins bound to agarose beads were incubated overnight at 4°C with 400 l (about 1 mg of protein) of cell lysates from Du145 and PC3 cells, obtained by extraction with NE-PER nuclear and cytoplasmic extraction reagents. The beads were pelleted and washed, and the "pull-down" proteins were separated on SDS-PAGE and stained with GelCode Blue. The proteins bound to S-tagged C/EBP␣ fusion proteins were analyzed by MALDI-TOF-MS in the Research Core Facility of LSUHSC-Shreveport with the Voyager-DE TM PRO Biospectrometry TM Work station (Applied Biosystems, Foster City, CA). Briefly, the proteins captured by the S-tag C/EBP␣ fusion protein were separated by SDS-PAGE and stained with GelCode Blue. The protein bands of interested were sliced from the gel and subjected to peptide extraction and elution with C18 zip-tips. The peptides were analyzed with three different instrument settings: Proteomics_autosample.bic, Insulin_liner.bic, and peptide negative_reflector.bic. The analyses of spectra data was performed with Data Explore Software (version 3.2.1) for algorithms to correct the base line and remove the noise at 2 S.D. The match of peptide data was conducted against protein databases, NCBI and Genpept, using Auto MS-Fit software.
Radiation, Cell Proliferation, and Clonogenic Survival Assays-80 ϫ 10 5 cells grown on 35-mm plastic dishes were suspended in 2 ml of RPMI1640 medium in a 15-ml plastic tube and exposed at room temperature to radiation at doses of 4 and 8 Gy from a Varian 6-MV linear accelerator. For proliferation assays after exposure to radiation, cells were grown in 96-well plates at initial densities of about 1000 cells/well for 0 and 4 Gy and 3000 cells/well for 8 Gy exposure in RPMI 1640 containing 10% fetal bovine serum and 400 g/ml of Geneticin. The cell proliferation assay was conducted 6 days after radiation exposure with the CellTiter 96 AQ ueous non-radioactive cell proliferation assay kit (Promega, Madison, WI) according to the manufacturer's instruction. For colony formation assays cells were exposed to radiation and plated at density of 500 cells per well in 6-well plates in RPMI 1640 containing 10% fetal bovine serum and 400 g/ml Geneticin, and after 10 days the cells were stained with 0.05% crystal violet, and colonies enumerated under a dissecting microscope.

Screening for Proteins That Partner with C/EBP␣ in Prostate Cancer
Cells-Putative protein partners of C/EBP␣ were identified by using a pull-down assay in which an S-tagged C/EBP␣ fusion protein was incubated with lysates containing both cytoplasm and nuclear proteins from the prostate cancer cell lines PC3 and Du145 (Fig. 1). Analysis of the proteins that co-precipitated with C/EBP␣ by separation by SDS-PAGE identified four protein bands that appeared to be specifically pulled down by C/EBP␣. These protein bands were neither seen with incubation of the S-tag protein alone with cell lysates from PC3 and Du145 cells (Fig. 1, lanes 1 and 3) nor with the S-tag C/EBP␣ fusion protein without cell lysate (Fig. 1, lane 5). MALDI-TOF-MS analysis of each protein band demonstrated that the proteins of approximate molecular masses 113, 82, and 70 kDa were PARP-1, Ku80, and HSC, respectively.
Based on analysis with Auto MS-Fit, the MOWSE (MOlecular Weight SEarch) scores for all identified proteins were greater than 10 5 . In addition, a band staining with a molecular mass of about 20 kDa was not successfully identified.
The Further Identification of the Interaction of C/EBP␣ with Putative Partner Proteins-To confirm the physical interaction between C/EBP␣ and the proteins identified in the pull-down assay, we conducted an in vitro GST pull-down assay by incubating a GST-C/EBP␣ fusion protein with Ku80, HSC, and PARP-1 expressed in the E. coli BL21 system. Considering that Ku80 functions as a heterodimer with Ku70, we also examined the interaction between Ku70 and C/EBP␣. Among the four proteins, Ku80, Ku70, and PARP-1 were pulled down by GST-C/EBP␣ fusion protein ( Fig. 2A), whereas HSC did not show any interaction with C/EBP␣ in the GST pull-down assay (data not shown). Furthermore, by taking advantage of the ability to express PARP-1, Ku70, and Ku80 in E. coli, we could determine whether the addition of one of these proteins competed for the binding of the others to C/EBP␣ (Fig. 2B). The presence or absence of Ku70/80 had no effect on the binding of PARP-1 to C/EBP␣ (Fig. 2B, left panel). Likewise, in the absence of PARP-1 or in the presence of PARP-1 at two different concentrations, the binding of Ku70 or Ku80 to C/EBP␣ was not affected (Fig. 2B, right panel). Therefore, it is most likely that C/EBP␣ forms a complex with PARP-1, Ku70, and Ku80 proteins rather than forming dimers with each individual protein.
In all experiments a greater percentage of PARP-1 appeared bound to C/EBP␣ than seen with Ku70 and Ku80. For example, by densitometric scanning of the Western blots of the pull-down proteins and the corresponding inputs in Fig. 2A, the ratio of bound protein over input was 2.4 for PARP-1, 0.63-fold for Ku70, and 0.28fold for Ku 80. These data imply that PARP-1 binds more strongly to C/EBP␣ than the two Ku proteins, but it is not known if the differences in binding have any functional consequence. To determine whether the interactions between C/EBP␣ and Ku80, Ku70, and PARP-1 also occurred in vivo in intact cells, we first examined the expression levels of Ku80, Ku70, and PARP-1 in various prostate and non-prostate cell lines. As shown in Fig. 2C, Ku80 and Ku70 were expressed in all the cell lines examined. PARP-1 was expressed in all cell lines except U937 cells. These cell lines also expressed low levels of endogenous C/EBP␣ protein, although only the p42 isoform was detected (Fig. 2D, left panel). To determine whether the endogenous Ku80, Ku70, and PARP-1 proteins would interact with C/EBP␣, several prostate cell lines were constructed that overexpressed C/EBP␣. Clones of the prostate cancer cell lines LNCaP, Du145, and PC3 were generated through retrovirus transformation using the full-length rat C/EBP␣ cDNA inserted into the retrovirus vector PLNCX. The increased expression of C/EBP␣ was seen in all transduced cell lines. Although the p42 isoform was the predominant isoform, the p30 isoform was detected in transduced LNCaP cells (Fig. 2D, right panel). As shown in Fig. 2E, in all three lines Ku70, Ku80, and PARP-1 co-precipitated with C/EBP␣. Among the three transduced cell lines PC3 showed the weakest interaction of C/EBP␣ with Ku proteins and PARP-1 based on the ratio of co-precipitated proteins to the input in the cell lysates. However, the interaction could be strengthened by the induction of double-strand DNA breaks by exposure of PC3 cells to bleomycin and iron. After incubation of the PC3 cells with bleomycin and iron for 24 h, expression of C/EBP␣ increased by 3.4-fold, and the amount of Ku70, Ku80, and PARP-1 that co-precipitated with C/EBP␣ increased as well by 6.6, 2.7, and 1.7-fold, respectively. Ku70, Ku80, and PARP-1 also co-precipitated with endogenous C/EBP␣ as was demonstrated in LNCaP cells (Fig. 2F, top panel) and in human prostate cancer tissue (Fig. 2F, middle and lower panels). The expression of C/EBP␣ and the two Ku proteins was FIGURE 1. Identification of C/EBP␣-associated proteins in Du145 and PC3 cells. Purified S-tagged C/EBP␣, lysates from the prostate cancer cell lines Du145 and PC3, and the pull-down assays were performed as described under "Experimental Procedures." After extensive washing, the proteins absorbed to the S-tagged agarose beads were separated on SDS-PAGE and stained with GelCodeா blue. Shown is the GelCodeா bluestained polyacrylamide gel of the C/EBP␣-associated proteins pulled down by incubating non-denatured whole cell extract from PC3 cells (lanes 1 and 4) and Du145 (lanes 3 and 6) cells with purified S-peptide alone (lanes 1-3) or S-tagged C/EBP␣ (lanes 4 -6). Lanes 2 and 5 represent S-peptide and S-tagged C/EBP␣ alone without cell extract. The molecular weight markers (MW) are shown. The appropriate portions of the polyacrylamide gel containing the protein bands were analyzed by MALDI-TOF-MS as described under "Experimental Procedures." The arrows indicate the proteins identified by MALDI-TOF-MS with MOWSE score above 1.0Eϩ05. detected in all three samples of prostate cancer. However, PARP-1 was only detected in two of the samples (Fig. 2F, middle panel), and in a third sample, PARP-1 was detected as only small fragments, presumably resulting from proteolysis, of less than 50 kDa apparent molecular mass (data not shown). In the cancer tissue the interaction of C/EBP␣ with Ku70, Ku80, and PARP-1 was seen by co-immunoprecipitation (Fig. 2F, bottom panel). These in vitro and in vivo binding assays demonstrate that Ku70, Ku80, and PARP-1 proteins may universally interact with C/EBP␣. In the lung cancer cell line H358, Ku70 and Ku80 also coprecipitated with C/EBP␣ when C/EBP␣ was overexpressed (data not shown). These results also suggest that formation of a C/EBP␣-Ku70 -80-PARP-1 complex could depend on the specific cell type and perhaps the status of the cell.
The C Terminus of C/EBP␣ Is Essential for the Binding of Ku80, Ku70, and PARP-1-C/EBP␣ has three transactivation domains located in the N terminus of the protein and a basic leucine zipper domain in the C terminus. Both the N and C termini have been shown to have different roles in regulation of gene transcription, cell differentiation, and proliferation. To determine which domain is necessary for the binding of Ku80, Ku70, and PARP-1, we expressed several different polypeptides of C/EBP␣ (Fig. 3). These peptides included an N-terminal fragment with two of the transactivation domains intact (Fig. 3A, N), a truncated C/EBP␣ p30, which can be physiologically translated from an internal AUG codon in some cells (Fig. 3A, C1), and a C-terminal fragment containing the basic leucine zipper domain alone (Fig. 3A, C2). All three fragments were expressed as S-tagged proteins and purified with S-protein-agarose. The C/EBP␣ fragments (Fig. 3B, lower panel) were incubated with whole cell extracts from LNCaP cells, bound proteins were pulled-down, and the presence of Ku80, Ku70, and PARP-1 was examined by Western blot analysis with specific antibodies. Ku80, Ku70, and PARP-1 interacted with C/EBP␣ via the C-terminal region (Fig. 3B,  upper panel, C1 and C2). No interactions could be demonstrated with the N-terminal region of C/EBP␣ (Fig. 3B, upper panel, N).
Stable Expression of C/EBP␣ Increases the Sensitivity of Du145 and PC3 Cells to Radiation-It has been well known that Ku80, Ku70, and PARP-1 take part in the repair processes of DNA double-strand breaks.
To determine whether the interaction of C/EBP␣ with Ku80, Ku70, and PARP-1 affected DNA repair function in prostate cancer cells overexpressing C/EBP␣, we first examined expression of genes participating in DNA repair. Without inflicting any DNA damage onto LNCaP, Du145, or PC3 cells, the mRNA levels of Ku70, Ku80, and PARP-1 were not affected by expression of C/EBP␣ (data not shown). The expression of genes responsible for both homologous recombination and nonhomologous end joining was examined by microarray analysis of clones of PC3 and LNCaP cell lines expressing C/EBP␣ (Table 1). No significant changes were observed except for a greater than a 2-fold log ratio change of Rad50 mRNA levels in both C/EBP␣-expressing lines. Next, the protein levels of Ku70, Ku80, and PARP-1 and C/EBP␣ were examined by Western blot analysis 72 h after exposure of the cells to radiation at 0, 4, and 8 Gy to determine whether C/EBP␣ affected the expression of these proteins under conditions of DNA damage, especially DNA double-strand breaks. For these experiments, clones of PC3 and Du145 cell lines were generated from cells transduced by retrovirus transformation with the full-length rat C/EBP␣ cDNA inserted into the retro-  4) compared with the non-radiated cells (Fig. 4A, right panel, lane  7). In the Du145-H cells, C/EBP␣ expression increased 5.5-and 5.6-fold at 4 and 8 Gy, respectively (Fig. 4A, left panel, lanes 1 and 4) compared with the non-radiated cells (Fig. 4A, left panel, lane 7). No increase of C/EBP␣ expression was seen in the low C/EBP␣-expressing cell clones, PC3-L and Du145-L (Fig. 4A, left and right panels, lanes 2, 5, and 8). Radiation did not affect the expression of the Ku proteins (Fig. 4A) but did affect the expression of PARP-1 in the Du145 cells but not in the PC3 cells. The relative expression of PARP-1 in Du145-N, Du145-L, and Du145-H before and after exposure to radiation is given in Table 2. In Du145-H cells the expression of PARP-1 increased by 6.5-and 4.4-fold at radiation doses of 4 and 8 Gy, respectively, compared with Du145-N (Fig. 4A, left panel, lanes 4 and 1 compared with lanes 6 and 3 and Table  2). In the absence of radiation, Du145-H expressed a similar level of PARP-1 protein as Du145-N with a relative expression ratio of 1 when corrected for tubulin expression (Fig. 4A, left panel, lane 7 compared with lane 9 and Table 2). It appeared, too, that the exposure to radiation reduced the expression of PARP-1 protein in Du145-N and Du145-L cells. In the PC3-H cells, although the expression level of PARP-1 was not changed by C/EBP␣ expression or radiation, the migration of PARP-1 on SDS-PAGE was slower than in PC3-N and PC3-L cells (Fig.  4A, right panel, lanes 1, 4, and 7). A clonogenic survival assay was conducted to correlate the expression of C/EBP␣ with the radiation sensitivity of the various C/EBP␣-expressing clones. After radiation, colony formation was reduced in both the Du145 and PC3 L and H clones. The decrease was about 50% in PC3-L and PC3-H cells at 4 Gy and about 90% in PC3-L and 100% in PC3-H cells at 8 Gy compared with the PC3-N cells (Fig. 4B, bottom panel). In the Du145 cell line, the Du145-H cells exhibited a greater sensitivity to radiation than the Du145-L cells (Fig. 4B, upper panel). At 4 Gy Du145-H cells showed a 75% reduction in colony formation compared with Du145-N cells, whereas Du145-L cells did not show a significant change. At 8 Gy of exposure, colony formation was decreased by 57% in Du145-L cells and by 100% in Du145-H cells. A cell proliferation assay was performed with the Du145 N, L, and H clones after 4 and 8 Gy of radiation (Fig. 5) with a response pattern similar to that observed with the clonogenic assay. In the proliferation assay decreased proliferation was observed both of the Du145 L and H clones after 8 Gy, but only a decrease in the H clones was observed after 4 Gy (Fig. 5).

C/EBP␣ Expression Interferes with NHEJ in Vivo and in Vitro-Based
on the function of PARP-1 and Ku proteins in repair of DNA doublestrand breaks and the increased sensitivity of C/EBP␣-expressing cells to the radiation exposure, we hypothesized that the interaction of C/EBP␣ with Ku70, Ku80, and PARP-1 may interfere with the NHEJ process which plays a primary role in repairing DNA double-strand breaks. To test this hypothesis, the NHEJ assay was conducted with nuclear extracts from LNCaP and Du145 cells with high (LNCaP-H, Du145-H)) or no expression (LNCaP-N, Du145-N) of C/EBP␣ (Fig. 6A). Both LNCaP-H and DU145-H cells exhibited an ϳ60% inhibition of end rejoining in the NHEJ assay (Fig. 6A) compared with LNCaP-N and Du145-N. Interestingly, the PC3-H cells, although expressing C/EBP␣, did not show a significant decrease in end rejoining compared with PC3-N cells (data not shown). Based on the observation above that PC3-H clones exhibited a significant increase in radiation sensitivity (Fig. 4B, bottom panel), we simulated a condition that would generate DNA double-strand breaks by exposure of PC3-N, PC3-H, Du145-N, and Du145-H cells to 100 milliunits bleomycin plus 1 mM ferric chloride and 1 mM ferrous sulfate for 1.5 or 12 h (37). After treatment with bleomycin and iron, the end rejoining was reduced by about 3.5-fold in PC3-H cells for both the 12-and 1.5-h exposures to bleomycin and iron (Fig. 6B, lanes 2 and 4) and by 40-and 4-fold in Du145-H cells for the 12and 1.5-h exposures, respectively (Fig. 6B, lanes 6 and 8). The reduced end joining in the PC3-H cells after bleomycin treatment indicates that in the PC3-H clones C/EBP␣-related interference with NHEJ is DNA damage-dependent. This phenomenon is consistent with the previous observation that DNA damage increased binding of C/EBP␣ to Ku and PARP-1 protein (Fig. 2E).
However, because the C/EBP␣-expressing cells were clonally selected, we could not exclude that decreased NHEJ in C/EBP␣-expressing cells was caused by clonal selection, allowing for factors other than direct interaction of C/EBP␣ with DNA repair proteins to interfere with DNA repair. To address this question, we performed the NHEJ assay using purified C/EBP␣ protein expressed in E. coli strain BL21-condon plus-RIL. The addition of purified C/EBP␣ to the nuclear extracts from LNCaP, PC3, and Du145 cells resulted in a 33% decrease of end rejoining in LNCaP cells and a 75% decrease in Du145 cells (Fig. 7A). The NHEJ assay using PC3 cell lysates demonstrated no change in end rejoining, similar to the observation that in intact PC3 cells expression of C/EBP␣ interfered with NHEJ only under conditions of DNA damage. To determine the regions of C/EBP␣ necessary for interference with DNA repair, we observed the effect of C-terminal-and N-terminal-deleted C/EBP␣ peptides on end rejoining (Fig. 7B). Neither the C/EBP␣ polypeptides with a C-terminal or N-terminal deletion affected end rejoining, suggesting that the intact protein is needed for function in C/EBP␣-mediated inhibition of the NHEJ repair process.

DISCUSSION
The transcription factor C/EBP␣ is instrumental in promoting cell differentiation and decreased cell proliferation in many tissues. C/EBP␣ is known to be expressed in normal prostate epithelium; however, we have observed that overexpression of C/EBP␣ in prostate cancer cell

TABLE 1 Microarray analysis of RNA expression of genes responsible for nonhomologous end rejoining and homologous recombination in PC3 and LNCaP cells expressing C/EBP␣
As described under "Experimental Procedures," RNA was extracted from PC3 and LNCaP cells transduced with C/EBP␣ or transduced with viral vector alone, and gene expression was determined using the Affymetrix U95A chip. Shown are the results for genes known to be involved in DNA repair with an absolute value of 1 for log-fold change, representing a 2-fold change in expression of a gene in the C/EBP␣-expressing cells compared to the control cells. These changes are also coded as increased (I), decreased (D), and no change (NC) as determined by Affymetrix Microarray Suite.

Probe set ID
Gene symbol lines leads to increased proliferation of the cells instead of the expected decrease (33). Hence, we were interested in examining these prostate cancer cell lines for proteins that partnered with C/EBP␣ and that would account for this unexpected behavior. Using a cell-free pulldown assay with S-tagged C/EBP␣ protein as bait and lysates of the Du145 and PC3 prostate cancer cell lines as prey we identified with MALDI-TOF-MS analysis three proteins, Ku80, PARP-1, and HSC, as putative partners of C/EBP␣. Further analysis using GST-C/EBP␣ and purified proteins confirmed Ku80 and PARP-1 as binding to C/EBP␣. In addition, because Ku70 and Ku80 form heterodimers, purified Ku70 was also examined in the GST-pull down assay and was confirmed as a C/EBP␣ partner. The initial experiments searching for protein partners of C/EBP␣ used C/EBP␣ that had been synthesized in prokaryotic cells. It is possible that recombinant C/EBP␣ synthesized in bacteria may not have the same folding as in mammalian cells, and therefore, either some partners might be missed or others observed that would not interact with C/EBP␣ in eukaryotic systems. In addition, the strategy used might

FIGURE 6. Expression of C/EBP␣ in vivo inhibits end rejoining in an NHEJ assay.
A, C/EBP␣ expressing (H) and non-expressing (N) LNCaP and Du145 cells were grown in RPMI 1640 plus 10% fetal bovine serum. Nuclear extracts were prepared, and the NHEJ assay was conducted as described under "Experimental Procedures." The percentage end rejoining (calculated by the sum of dimer and multimer divided by the sum of monomer, dimer, and multimer) is shown for each lane. B, the NHEJ assay was conducted with cells exposed to bleomycin and iron. The PC3 and Du145 N and H cells were exposed to 100 milliunits of bleomycin, 1 mM ferric chloride, and 1 mM ferrous sulfate for 1.5 h (lanes 1, 2, 5, and 6) and 12 h (lanes 2, 4, 7, and 8) before obtaining nuclear extracts. not detect either weak interactions or the interaction with low concentration of proteins. We were able to exclude the potential artifact imposed by improper folding by demonstrating that Ku70, Ku80, and PARP-1 interacted with C/EBP␣ that was 1) overexpressed in prostate cancer cell lines, 2) endogenously expressed in LNCaP cells, and 3) endogenously expressed in human prostate cancer. Furthermore, that the interactions could be demonstrated to occur with a discrete region of C/EBP␣, namely the C-terminal basic leucine zipper domain, suggests that mis-folding of C/EBP␣ with expression in prokaryotes did not play a role in the detection of protein partners.
The Ku proteins were originally identified as autoantibodies in patients with rheumatic disorders (38). The two Ku proteins have been well demonstrated to dimerize and to function in repair of DNA doublestrand breaks, DNA telomere length maintenance, transcription regulation, and V(D)J recombination (39 -43). Recently, Ku proteins expressed on the cell surface were found to be associated with the extracellular matrix and to interact specifically with matrix metalloprotease 9, suggesting that the Ku proteins may be involved in tumor invasion (44,45). PARP-1 is a founding member of poly(ADP-ribose) polymerases whose superfamily consists of 18 proteins (46 -48). PARP-1 is a DNA damage-activated protein and a sensor for DNA breaks that binds to DNA breaks, catalyzes NAD ϩ hydrolysis, produces a polymer of ADP-ribose, and transfers the polymers to histones and other nuclear proteins. The poly ADP-ribosylation of histones and nuclear proteins facilitates the access of DNA repair enzymes to relaxed chromatin. In response to DNA damage, both PARP-1 and Ku proteins participate in the non-homologous end joining pathway by direct binding to the broken ends of damaged DNA. The close association of Ku proteins and PARP-1 has been demonstrated with co-immunoprecipitation (49), although recently the importance of PARP-1 in DNA double-strand break repair has been questioned (50,51).
Because Ku70, Ku80, and PARP-1 are expressed universally, we asked if the interaction of C/EBP␣ with Ku70, Ku80, and PARP-1 occurs in all cells or only in prostate cancer cells. Co-immunoprecipitation using C/EBP␣-transfected H358 cells, a lung cancer cell line, demonstrated that Ku70 and Ku80 were co-precipitated by antibodies to C/EBP␣, suggesting that the interaction between C/EBP␣ and Ku proteins and PARP-1 is not unique to prostate cancer cells. The association of Ku80 and PARP-1 with C/EBP␣ has not been previously described and may in our studies with prostate cancer cells have FIGURE 7. The addition of C/EBP␣ protein into the NHEJ reaction inhibits DNA end rejoining. A, the NHEJ assay was conducted as in Fig. 6 using nuclear extracts prepared from LNCaP, PC3, and Du145 cells not transfected with C/EBP␣. Purified His-tagged C/EBP␣ was added to the NHEJ reaction (lanes 3, 5, and 7). Lane 1, pBlueKS plasmid alone; lanes 2, 4, and 6, NHEJ reaction without adding C/EBP␣. The percentage of rejoining (calculated by the sum of dimer and multimer divided by the sum of monomer, dimer, and multimer) is shown with each lane. B, top panel, the NHEJ reaction was performed with nuclear extract from non-C/EBP␣-expressing DU145 cells as above with the His-tagged C/EBP␣ fragments (1 g), as described in Fig. 3 been the result of the strategy and methods used for screening with a fortuitous ratio of bait to prey that allowed visualization of the co-precipitating proteins. Both the Ku proteins and PARP-1 are abundantly expressed and, therefore, could be visualized by protein staining of the gel. Using a similar strategy, Ku70 and Ku80 have been shown to interact with the androgen receptor in prostate cells (52). The failure to identify Ku70 on the initial gels was the mischaracterization of the Ku70 (which migrates slightly slower than HSC) as an E. coli protein. Detection of the complexes may have been enhanced by the relatively strong binding of PARP-1 to GST-C/EBP␣ in the GST-pull-down assay. In addition, cell-specific patterns of protein-protein interactions could also have enhanced detection. We repeated the pull-down assay by incubation of GST-C/EBP␣ fusion protein with whole cell lysate from the prostate cancer cells, LNCaP, Du145, and PC3 and the erythroleukemia cell line, K562. Although Ku80 could be directly immunoprecipitated in all the cells lines, only in the prostate cell lines did C/EBP␣ also result in a dominant pull-down band of Ku80 in stained SDS-PAGE. In the K562 cell line GST-C/EBP␣ pulled down a distinct 50-kDa protein that was not present in the pull-down assay with the prostate cells and which was identified by MALDI-TOF-MS as elongation factor ␥ (data not show). On the other hand, in the LNCaP, Du145, and PC3 C/EBP␣ overexpressing cell lines, no interaction could be observed of C/EBP␣ with E2F, CDK2, and CDK4. When the Ku proteins were removed from LNCaP cell lysates with antibodies to Ku70 and Ku80, binding of C/EBP␣ to E2F was seen in the pull-down assay (data not shown), supporting our hypothesis that the spectrum of proteins interacting with C/EBP␣ depends on the cell type and the relative concentrations of C/EBP␣ and the C/EBP␣-interacting proteins.
In general, prostate cancer cells are more resistant to radiation-induced killing compared with other cancer cells. The death of prostate cancer cells after exposure to radiation may be predominantly by mechanisms other than apoptosis. Indeed, we have observed that after radiation of various prostate cell lines, apoptosis occurred in less than 5% of cells as measured with DNA electrophoresis, 4,6-diamidino-2-phenylindole staining, and labeling with annexin V-fluorescein isothiocyanate (data not shown). There are multiple reasons why a particular cell type might exhibit resistance to radiation including the ability to repair radiation-induced DNA double-strand breaks. The repair of DNA double-strand breaks occurs preferentially through NHEJ rather than through homologous recombination (53). Having demonstrated that C/EBP␣ expressed in prostate cancer cell lines can associate with Ku70, Ku80, and PARP-1, which are important participants in the repair of DNA double strand breaks, we next wanted to examine if the interactions in C/EBP␣-expressing cells affected radiation sensitivity and DNA repair ability. In keratinocytes, UVB radiation exposure increases expression of C/EBP␣ both at the RNA and protein levels via a p53mediated pathway, and knockdown of C/EBP␣ diminished the DNA damage G (1) checkpoint activity and increased sensitivity to UVBinduced apoptosis (54). In our experimental system exposure to radiation increased the protein level of C/EBP␣ by severalfold in PC3 and Du145 clones with the high C/EBP␣ expression. Although we have not yet determined the mechanism for the increased expression, as the transcription of C/EBP␣ in these cells was driven by the strong cytomegalovirus promoter, it is most likely that the increased expression reflects increased translation or decreased degradation. Because the C/EBP␣ expressing PC3 and Du145 cell lines demonstrated increased radiation sensitivity by both the clonogenic survival and cell proliferation assays, it was of interest to see if the interaction of C/EBP␣ with the DNA repair proteins, Ku70, Ku80, and PARP-1, could interfere with the repair of DNA double-strand breaks. This possibility was verified by a C/EBP␣-mediated decreased NHEJ in LNCaP and Du145 cell lines assayed both in cell lysates and intact cells. The inhibition of NHEJ by C/EBP␣ was dependent on a full-length C/EBP␣ protein. In addition, decreased NHEJ was seen in the PC3-H cells only when double-strand DNA breaks were induced by exposing cells to bleomycin plus iron. These observations support our hypothesis that overexpression of C/EBP␣ in prostate cancer cells causes increased sensitivity to double-strand DNA breaks via decreased DNA repair.
It was of interest that radiation induced different clonogenic survival assay responses in the PC3 and Du145 cell lines. In Du145 cell line, increased radiation sensitivity was more related to expression level of C/EBP␣ than in the PC3 cell line. In addition, the NHEJ assay showed no altered end rejoining in PC3-H cells without DNA damage, and the addition of purified C/EBP␣ protein into NHEJ reaction with nuclear extract from non-treated PC3 cells did not decrease the end rejoining. Even with bleomycin-iron induction of DNA damage, NHEJ in the PC3 cell line was weaker than that seen with the Du145 cells. Because the levels of Ku70 and Ku80 were similar in the two cell lines, it was of interest to examine the role of PARP-1 in mediating the different responses to radiation. Radiation induced a significant increase of PARP-1 expression in the Du145-H cells. Although PARP-1 is important for repair of DNA double-strand breaks, increased expression of PARP-1 will enhance cell death via activation of apoptosis-inducing factor released from the mitochondria with subsequent activation of caspase-independent apoptosis (53,54). Hence, the net result of PARP-1 activation will be the balance between DNA repair and the induction of apoptosis. In the PC3 cell line, slow migration of PARP-1 in the PC3-H cells raises the possibility that protein modification of PARP-1 could affect PARP-1 function. PARP-1 has previously been demonstrated to be phosphorylated by DNA-dependent protein kinase, although it is not certain if the phosphorylation inhibited PARP-1 activity or if the binding of the kinase to PARP-1 was responsible for decreased PARP-1 activity (49). The nature of the modification induced by overexpression of C/EBP␣ and how this modification affects PARP-1 activity in PC3 cells needs to be investigated. Additionally, a PARP-1dependent double-strand break end-joining activity may exist as an alternative route to non-homologous end-joining repair (55). Perhaps one or another repair route is favored in one cell type or under a particular condition. Also, the differences in DNA repair between the PC3-H and Du145-H cells may reflect even more complicated protein interactions than what we have found.
In addition to the apparent interference of C/EBP␣ with Ku-and PARP-1-mediated DNA repair, it is possible that the binding of Ku70, Ku80, and PARP-1 to C/EBP␣ could alter C/EBP␣ function. To address this possibility, we examined the effect of forced expression of Ku70, Ku80, and PARP-1 on transcription regulation by C/EBP␣ of the PSA promoter. In LNCaP cells, co-transfection of C/EBP␣ with Ku70, Ku80, or PARP-1 had no effect on the inhibition of the PSA promoter by C/EBP␣ (data not shown). It seems then that under our experimental conditions the interaction of C/EBP␣ with the Ku proteins and PARP-1 does not affect transcription regulation by C/EBP␣.
In summary, we report newly identified protein partners of C/EBP␣ and that these associated proteins contribute to increasing radiosensitivity and block NHEJ in prostate cancer cell lines, especially in the Du145 prostate cell line. These findings suggest a new avenue for prostate cancer therapy.