The Cyclin A1-CDK2 Complex Regulates DNA Double-Strand Break Repair

Abstract

Vertebrates express two A-type cyclins; both associate with and activate the CDK2 protein kinase. Cyclin A1 is required in the male germ line, but its molecular functions are incompletely understood. We observed specific induction of cyclin A1 expression and promoter activity after UV and γ-irradiation which was mediated by p53. cyclin A1^−/− cells showed increased radiosensitivity. To unravel a potential role of cyclin A1 in DNA repair, we performed a yeast triple hybrid screen and identified the Ku70 DNA repair protein as a binding partner and substrate of the cyclin A1-CDK2 complex. DNA double-strand break (DSB) repair was deficient in cyclin A1^−/− cells. Further experiments indicated that A-type cyclins activate DNA DSB repair by mechanisms that depend on CDK2 activity and Ku proteins. Both cyclin A1 and cyclin A2 enhanced DSB repair by homologous recombination, but only cyclin A1 significantly activated nonhomologous end joining. DNA DSB repair was specific for A-type cyclins because cyclin E was ineffective. These findings establish a novel function for cyclin A1 and CDK2 in DNA DSB repair following radiation damage.

The cell cycle is regulated by external signals, especially in multicellular organisms, where cell proliferation is central to growth and differentiation (34). Internal signals ensure that cell cycle transitions do not occur until all required molecular events have been completed (20). The internal signals become evident as cell cycle checkpoints upon an insult to the cell (e.g., DNA damage) which arrests the cell cycle while repair occurs (10, 14). Both internal and external signals modulate the activity of the cyclin-dependent kinases (CDKs) that catalyze the ordered transitions from one phase of the cell cycle to the next. Cyclin E and cyclin A2 both associate with CDK2, and cyclin A2 is essential for DNA replication and proliferation in somatic cells (8). Embryos with a homozygous deletion of cyclin A2 are not capable of proliferative growth (28). During G₁ and S phases, the kinase activity of CDK2 associates first with cyclin E and later with cyclin A2. The timing of these processes as well as the appropriate concentration of cyclin/CDK2 complexes is crucial for successful DNA replication (8).

Cyclin A1 is a second A-type cyclin that binds CDK2. Cyclin A1 is abundantly expressed in the testis and has previously been shown to be essential for entry into the metaphase of meiosis I in the male germ line in mice (38, 44). Cyclin A1 is expressed at low levels in most other tissues, but no phenotype other than male infertility has been reported for mice lacking the cyclin A1 gene (19, 42). The expression of cyclin A1 in hematopoietic progenitor cells and in acute myeloid leukemia is best characterized in somatic cells (46). Surprisingly, recent microarray data suggested that cyclin A1 was transcriptionally induced following p53 activation (21).

Thus, the physiological role of cyclin A1 in somatic cells remains unknown. This prompted us to analyze the mechanisms of cyclin A1 induction in somatic cells and the functional consequences. Our analyses revealed a novel function for p53-induced cyclin A1 and CDK2 in DNA double-strand break (DSB) repair.

MATERIALS AND METHODS

Analysis of cyclin A1 induction and radiosensitivity following γ-irradiation.

cyclin A1 knockout mice were generated by substituting the cyclin A1 coding sequence with the β-galactosidase gene (42). Six-week-old BALB/c mice were γ-irradiated with 5 Gy and sacrificed after 24 or 48 h. At least two mice were analyzed at each time point. Spleens were stained for β-galactosidase activity as described previously (3), and sections were prepared after paraffin-embedding. For reverse transcription-PCR, RNA was isolated from various organs. Colony formation assays were performed with bone marrow cells prepared from cyclin A1^−/− and matched wild-type mice. Bone marrow cells were γ-irradiated (1 Gy) in vitro before cells were plated in methylcellulose with appropriate growth factors as described (27). Colonies were counted on day 14. Experiments were performed in triplicates, and the box plots represent the results of three independent experiments.

The comet assay was performed as described previously with slight modifications (32). In brief, cells were mixed with low-melting-point agarose and layered onto slides. Nuclei were lysed and electrophoresed in alkaline electrophoresis solution for 30 min at 300 mA (14 V). The slides were neutralized and stained with ethidium bromide. A fluorescence microscope (Axioskop; Zeiss, Jena, Germany) connected with a digital camera control unit (Spot RT) was used to acquire images of 100 randomly selected cells. The frequency of comet tail migration as an indication of DNA DSB repair was analyzed.

Real-time quantitative reverse transcription-PCR.

Real-time quantitative reverse transcription-PCR was performed essentially as described (18, 22). Primer and probe sequences were as follows: mouse cyclin A1, forward (5′-TTTCCCCAATGCTGGTTGA) and reverse (5′-AACCAAAATCCGTTGCTTCCT) and probe (5′-CCCACCACCCATGCCCAGTCA; mouse glyceraldehhyde-3-phosphate dehydrogenase, forward (5′-TTGTGCAGTGCCAGCCTC) and reverse (5′-CCAATACGGCCAAATCCG) and probe (5′-TCCCGTAGACAAAATGGTGAAGGTCGGT; the primer and probe sequences for human cyclins and human glyceraldehhyde-3-phosphate dehydrogenase have been described (26). PCRs were followed in real time with an SDS7700 sequence detector (Applied Biosystems). Serial dilutions of cDNA were run on each plate as standard curves, and gene expression levels were calculated with regard to these standards. Expression levels were standardized to cDNA concentrations with glyceraldehhyde-3-phosphate dehydrogenase expression levels.

Cyclin A1 promoter constructs and luciferase assays.

The cloning of the human cyclin A1 promoter has been described (23). The 5′ deletions were generated with S1 exonuclease treatment. The GC boxes were point mutated with the transformer site-directed mutagenesis kit (Clontech) (26). For luciferase assays, NIH 3T3 cells were transfected with 2 μg of cyclin A1 luciferase reporter construct, the pRL-SV40 construct (100 ng) (Promega), and 2 μg of either pCMV-p53 or an empty cytomegalogivurs expression vector. Drosophila S2 cells were transfected with the cyclin A1 promoter construct with the two proximal Sp1 binding sites mutated along with the pRL-Null vector (Promega), the p53 expression vector, and the pAC-Sp1 expression plasmid (23). Transfections were performed with Superfect (QIAGEN), and cells were lysed after 48 h in passive lysis buffer (Promega). Activities of firefly and Renilla luciferases were determined with the dual-luciferase assay (Promega). Experiments were independently performed three times, and the data presented are means and standard errors.

Chromatin immunoprecipitation.

Chromatin immunoprecipitation was performed as described with slight modifications (37). PCRs were followed in real time with either the 5′Nuclease assay with TaqMan technology (cyclin A1 and β-globin promoter) or with SYBR Green (p21^CIP1 promoter). The following primers were used: cyclin A1 forward (5′-CTCTTAACCGCGATCCTCCAG) and cyclin A1 reverse (5′-CAATAAAAGATCCAGGGTACATGATTG) and cyclin A1 probe (5′-FAM-TCCCGTCGTTATCTTTCTATGTGTCCGG-TAMRA) and p21^CIP1 forward (CCGCTCGAGCCCTGTCGCAAGGATCC) and p21^CIP1 reverse (GGGAGGAAGGGGATGGTAG). β-:globin primers and probe were 5′-GTGAAGGCTCATGGCAAGAAAG-3′ (forward), 5′-CAGCTCACTCAGTGTGGCAAAG-3′ (reverse), and 5′-VIC-ATGGCCTGGCTCACCTGGACAACC-3′ (probe). The results are presented as the average of two independent experiments with each condition being analyzed in three separate PCRs.

Triple hybrid screen.

Full-length human cyclin A1 cDNA was cloned in frame with the Gal4 DNA binding domain into the EcoRI and SalI sites of the pBridge vector. (Clontech). Within the cyclin A1 cDNA, the destruction box was point-mutated (R81C) to ensure consistently high expression of the GAL4-binding domain-cyclin A1 fusion protein. Full-length human CDK2 cDNA was cloned into the NotI and BglII sites of pBridge so that expression was regulated by the conditional methionine promoter (P_Met25). This construct was used as the bait in a triple hybrid screen of a cDNA library from human testis, constructed in pACT2 (Clontech). Approximately 8 × 10⁶ colonies were screened for activation of the HIS3, ADE, and lacZ reporter genes with Saccharomyces cerevisiae strain AH109 in the absence of methionine. The cDNAs from positive colonies were amplified by PCR and sequenced with standard procedures.

Expression of cyclin A1, cyclin A2, cyclin E, and CDK2 in SF9 cells and in vitro kinase assay.

The cyclin A1, cyclin A2, and CDK2 cDNAs were individually cloned into pBacPAK8 (Invitrogen), and recombinant baculovirus was generated with the BaculoGold system (Pharmingen). Baculovirus with cyclin E was kindly provided by Heike Laman (Wolfson Institute, London, United Kingdom). Recombinant baculovirus was used to infect Sf9 cells. The multiplicity of infection was 10. Sixty hours after infection, cells were lysed in buffer (50 mM Tris-HCl, 150 mM NaCl, 1% Nonidet P-40, and 1× protease inhibitor cocktail, pH 7.5). For in vitro kinase assays, 1 μl of lysate (2 μg of total protein per μl) was used in a 36-μl kinase reaction (50 mM HEPES, 1 mM NaF, 0.1 mM NaVO₄, 10 mM MgCl₂, 1 mM dithiothreitol, pH 7.0) containing 10 μM ATP, 3 μCi of [γ-³²P]ATP, and 2 μg of glutathione S-transferase (GST) fusion proteins bound to glutathione beads in a 50% slurry. The reactions were incubated at 30°C for 20 min. Beads were washed in kinase buffer once, and the phosphorylated proteins were analyzed by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) and autoradiography.

In vitro binding assay.

The Ku70 and Ku80 cDNA or deletion constructs of Ku70 cDNA were expressed in Escherichia coli BL21 as GST fusion proteins; 1 μg of GST fusion protein or 1 μg of GST bound to Sepharose beads was incubated in 1 ml of binding buffer (20 mM HEPES, 150 mM NaCl, 10 mM NaF, 0.4 mM EDTA, 1% Triton X-100m and 1× protease inhibitor cocktail, pH 7.2) with 10 μg of Sf9 lysate with baculovirus-expressed proteins or with lysates from U937 leukemia cells.

Immunoprecipitation and immunoblotting.

U937 cells were lysed in radioimmunoprecipitation (RIPA) buffer; 300 μg of cell lysate was used in the immunoprecipitation assays with 3 μg of primary antibody in 500 μl of RIPA buffer for 2 h at 4°C; 50 μl of a 50% slurry of protein A/G-Plus agarose (Santa Cruz Biotechnology) was added for another 1 h, and beads were washed three times with RIPA buffer. Immunoblot analyzes were performed following standard procedures (27). The antibodies used for immunoblotting were antiactin (Sigma), anti-Ku70 (BD-Transduction), anti-GST (Santa Cruz), anti-cyclin A2 (Sigma), anti-cyclin A1 (C20) (45), and anti-cyclin A1 (BD-Pharmingen).

Generation of murine embryonic fibroblasts.

Mice with homozygous deletion of p53 were obtained from Jackson Laboratories. Murine embryonic fibroblasts (MEF) were established from E13.5 embryos derived from breeding of cyclin A1^+/− females and males. Murine embryonic fibroblasts from Ku80^−/− and Ku80^+/+ mice were kindly provided by André Nussenzweig. MEF cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum, and passage 4 to 6 cells were used for the experiments.

In vitro DNA repair assays.

Whole-cell and nuclear extracts were prepared as described previously (12). For whole-cell extracts, 10⁸ cells were used in each preparation to yield 0.5 ml of extract with a protein concentration of 8 to 12 mg/ml. For nuclear extracts, cells were incubated in lysis buffer (20 mM HEPES, 10 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, 1.5 mM MgCl₂, 0.25% NP-40, and 1× protease inhibitor cocktail [Roche, Mannheim, Germany]) for 20 min on ice. Nuclei were pelleted, resuspended in lysis buffer containing 0.4 M NaCl and 25% glycerol, and kept on ice for 30 min. The supernatant was dialyzed with dialysis buffer (50 mM Tris-HCl, 1 mM dithiothreitol, 1 mM EDTA, and 10% glycerol) for 4 h. Aliquots were snap frozen and stored at −80°C until use.

For DSB repair reactions, plasmid pEGFP-1 (Clontech) was linearized with SmaI and used as the substrate. The reactions were performed in a volume of 20 μl in 30 mM HEPES-KOH, pH 7.8, 7 mM MgCl₂, 50 μg of bovine serum albumin per ml, 1 mM ATP, 100 μM deoxynucleoside triphosphates, 40 mM phosphocreatine, and 0.5 U of creatine phosphokinase, pH 7.8. Cell lysate containing 50 to 100 μg of protein was incubated with 100 ng of linearized plasmid for 4 h at 25°C. In all experiments with addition of baculovirus-produced recombinant protein, control samples were incubated with equal amounts of wild-type virus-produced lysate. Following deproteinization, 10% of the reaction products were loaded onto a 0.8% agarose gel. Following electrophoresis, Southern blotting was performed with a digoxigenin-labeled enhanced green fluorescent protein (EGFP) RNA probe. To analyze the fidelity of in vitro repair, the rejoined ends were amplified with the primers 5′-TTCGCCACCTCTGACTTGA and 5′-ATGGCGCTCCTGGACGTA. PCR products were cloned with the Topo TA cloning kit (Invitrogen) and sequenced.

In vivo DNA repair assays.

Plasmid pGL3-control (Promega) was linearized with either HindIII (site between the promoter and the luciferase cDNA) or NcoI (site includes the ATG initiation codon of the luciferase cDNA). The plasmid DNA was cotransfected along with the pRL-SV40 Renilla luciferase plasmid into MEF cells from wild-type and cyclin A1^−/− mice with Lipofectamine (Invitrogen). After 24 h, cells were lysed in passive lysis buffer and analyzed by the dual luciferase assay (Promega). The data presented are means plus standard deviation of three independent experiments.

NIH 3T3 cells were similarly transfected with HindIII-linearized pGL3 control plasmid along with the pRL-SV40 vector and different amounts of plasmid pcDNA3.1-cyclin A1. Empty pcDNA3.1 was added to keep the amount of DNA constant. Homologous recombination repair and nonhomologous end joining (NHEJ) assays in KMV5 cells were performed as described (1).

RESULTS

Irradiation induces cyclin A1 expression.

Under physiological circumstances, cyclin A1 is expressed at considerable concentrations only in few cell types. Recent microarray data suggested cyclin A1 induction following p53 expression (21). Cyclin A1 is expressed in cell types undergoing DNA repair after homologous recombination in the germ line or in hematopoietic cells. This prompted us to analyze whether cyclin A1 plays a role in pathways associated with DNA damage and subsequent DNA DSB repair.

To investigate the effect of DNA-damaging γ-irradiation on cell cycle regulators, we determined expression levels of all proliferation-associated cyclins in the duodenum carcinoma cell line HuTu80 and the colon carcinoma cell line LoVo (Fig. 1A). Cyclin A1 was the only consistently induced cyclin, in striking contrast to the expression of cyclin A2, which decreased upon irradiation. Moreover, cyclins E1, E2, B1, and B2 decreased after irradiation. Expression of p21^CIP1 served as the control for radiation response in all experiments (Fig. 1A). Several additional cell lines induced expression of cyclin A1 mRNA dose-dependently after γ-irradiation (Fig. 1B) and UV irradiation (data not shown), whereas cyclin A2 expression remained unchanged or decreased. The increase in the cyclin A1 mRNA levels was only observed in p53 wild-type cell lines (HuTu80, LoVo, and SW837), but not in a leukemic cell line harboring a p53 mutation (U937) (Fig. 1B).

FIG. 1.

Irradiation induces cyclin A1 expression. (A) Expression analysis of proliferation-associated cyclins and p21^CIP1 after irradiation in HuTu80 and LoVo carcinoma cell lines by quantitative real-time reverse transcription-PCR. (B) Cyclin A1 and cyclin A2 mRNA expression levels were analyzed in different cell lines by quantitative real-time reverse transcription-PCR following γ-irradiation with different doses (24 h). (C) HuTu80 colon carcinoma cells were γ-irradiated (25 Gy) for the times indicated. Protein expression was analyzed in nuclear extracts. A nonspecific band produced by the anti-cyclin A1 antibody was used to verify equal loading. (D) HuTu80 colon carcinoma cells were UV-irradiated, and protein expression in nuclear extracts was analyzed with Western blotting. A nonspecific band produced by the anti-cyclin A1 antibody in nuclear extracts was used to indicate equal loading. (E) Expression levels of murine cyclin A1, cyclin A2, and p21^CIP1 mRNA upon γ-irradiation (5 Gy) in mouse organs. Shown are the means of expression of two wild-type mice for each time point following irradiation. Please note the different scales for expression levels in different organs. (F) γ-Irradiation of heterozygous mice led to induction of LacZ expression under control of the endogenous murine cyclin A1 promoter. Shown are histological sections of the spleen of a heterozygous mouse 48 h after irradiation (5 Gy) and of another heterozygous mouse that was not irradiated (control).

On the nuclear protein level, induction of cyclin A1 was observed following UV irradiation and γ-irradiation of p53 wild-type HuTu80 cells. These experiments indicated that DNA damage-induced cyclin A1 mRNA expression was accompanied by a subsequent increase in cyclin A1 protein levels (Fig. 1C and D).

To determine whether cyclin A1 expression was also induced in tissues after γ-irradiation, RNA was prepared from a range of tissues 24 or 48 h after wild-type mice were exposed to 5 Gy whole body γ-irradiation. RNA was isolated from control tissues at the same time. Induction of p21^CIP1 occurred following irradiation and served as a positive control. Cyclin A1 was significantly induced following γ-irradiation in several tissues (Fig. 1E). This effect was specific for cyclin A1, since cyclin A2 levels decreased following irradiation (Fig. 1E). Cyclin A1 expression was repressed after γ-irradiation in testis, an organ with physiologically high levels of cyclin A1. General repression of transcription and a lack of a p53 effect (see below) are possible explanations. These explanations are supported by the finding that p21^CIP1P was also not induced in the testis after γ-irradiation (Fig. 1E).

To examine whether the endogenous murine cyclin A1 promoter responded to γ-irradiation in vivo, we took advantage of a previously generated knockout mouse line in which a lacZ gene was placed under the control of the endogenous cyclin A1 promoter (42). cyclin A1 heterozygous mice were irradiated with 5 Gy, and β-galactosidase activity was analyzed after 24 and 48 h in various tissues.

Histological sections of the organs indicated that the murine cyclin A1 promoter was induced following γ-irradiation. Enhanced LacZ expression was found in brain and lung, while its induction was found in spleen (Fig. 1F and data not shown). These results indicated that γ-irradiation could induce the endogenous murine cyclin A1 promoter in the organs of mice in vivo. In addition, the human cyclin A1 promoter but not the cyclin A2 promoter was activated by UV and γ-irradiation in luciferase assays following transient transfections (data not shown).

Cyclin A1 deficiency increases radiosensitivity of hematopoietic progenitor cells and of MEF.

Irradiation-induced genes are often involved in the cellular response to DNA damage, and alterations of these genes can lead to changes in cellular sensitivity to radiation damage. We analyzed primary cells from bone marrow, an organ which expresses cyclin A1, for changes in radiosensitivity. Bone marrow from cyclin A1^−/− mice exhibited slightly decreased colony formation potential (median, 110 versus 101 CFU, P = 0.2, not significant). However, upon irradiation of bone marrow with 1 Gy, cyclin A1^+/+ bone marrow formed twice as many colonies as cyclin A1^−/− bone marrow (median, 47 versus 23.5 CFU, P < 0.001, Wilcoxon test) (Fig. 2A).

FIG. 2.

Enhanced radiosensitivity of cells from cyclin A1^−/− mice. (A) Bone marrow cells from cyclin A1^+/+ and cyclin A1^−/− mice were irradiated with 1 Gy and seeded in methylcellulose. Experiments were performed in triplicate, and the box plots represent the results of three independent experiments. Colony formation by cyclin A1-deficient mouse bone marrow was decreased by 9% without irradiation. However, γ-irradiated bone marrow cells from cyclin A1^−/− mouse bone marrow produced less than half the number of colonies produced by bone marrow from wild-type mice (P < 0.001). Box plots indicate the median numbers of colonies (line within boxes), and the boxes themselves indicate the range from 25 to 75% of the observed values. (B) The comet assay was used to analyze the radiosensitivity of MEF cells. MEF cells with either the cyclin A1^+/+ or cyclin A1^−/− genotype were irradiated with 10 Gy and analyzed after 3 and 24 h. The arrows indicate typical migration tails visualizing a high number of DNA double-strand breaks. (C) Migration tails were more prominent in cyclin A1^−/− MEF cells than cyclin A1^+/+ cells after 3 and 24 h. The bar diagrams represent the mean and standard error of the mean of three experiments with two independent batches of MEF cells. cyclin A1^−/− cells show higher numbers of damaged cells after 3 h (P < 0.05) and after 24 h (P < 0.05).

Radiosensitivity was also analyzed in MEF cells from cyclin A1^+/+ and cyclin A1^−/− mice by the comet assay. DSBs, as indicated by the comet tail, were more frequent in cyclin A1^−/− MEF after 3 h as well as after 24 h (Fig. 2B and C). These findings established that cyclin A1-CDK2 was induced by irradiation and that cyclin A1 deletion was associated with increased radiosensitivity.

p53 activates the cyclin A1 promoter.

Cyclin A1 mRNA has recently been described to be expressed following p53 expression (21). This prompted us to analyze whether p53 induced cyclin A1 on the transcriptional level. For this purpose, a 1,344-bp fragment of the cyclin A1 promoter was cloned upstream of a luciferase reporter gene. Transient transfections showed that p53 induced cyclin A1 promoter activity in several cell types, including NIH 3T3 cells (data not shown). Since the cyclin A1 promoter does not contain a consensus p53 binding site, 5′ deletions of the cyclin A1 promoter were used to identify the relevant cis-acting sites. A section of the cyclin A1 promoter sequence from −112 to +145 was fully responsive to p53, whereas a fragment from −37 to +145 was unresponsive in NIH 3T3 cells (Fig. 3A). This portion of the cyclin A1 promoter (−112 to −37) contains four functionally active GC boxes that can be bound by Sp1 and Sp3 proteins (23). Analyses of point mutated constructs (described in (23) indicated that GC boxes 3 and 4 mediated p53 responsiveness (Fig. 3B).

FIG. 3.

p53 induces the cyclin A1 promoter. (A) Analysis of 5′ deletions of the cyclin A1 promoter and activation by p53. The 5′ deletions of the cyclin A1 promoter-luciferase constructs were cotransfected with p53 or control vector along with a Renilla luciferase plasmid. Data are presented as mean and standard error of the mean of three independent experiments. (B) Cyclin A1 promoter constructs harboring GC box mutations (23) were transfected along with the p53 or control expression vector. The graphs depict the ratio of luciferase activities of p53- versus control vector-transfected cells. GC box mutations that cannot be bound by Sp1 or Sp3 proteins are indicated in black, and intact GC boxes are depicted in white. Data are presented as mean plus standard error of the mean of three independent experiments. (C) The human cyclin A1 promoter with mutations in GC boxes 1 and 2 was cotransfected with p53 or control vector into Sp1-deficient S2 Drosophila cells. p53 alone did not transactivate the cyclin A1 promoter, but p53 enhanced Sp1-dependent activation of the cyclin A1 promoter. Data are presented as mean and standard error of the mean of three independent experiments. (D) Increased p53 binding to the cyclin A1 promoter following irradiation. HuTu80 cells were irradiated, and DNA was cross-linked at the indicated time points. Chromatin immunoprecipitation was performed with anti-p53 or nonspecific isotype control antibody. DNA was amplified by real-time PCR with SybrGreen (p21^CIP1) or the 5′ nuclease assay (cyclin A1 and β-globin). The bars represent the average values of two independent experiments with the quantitative PCR performed three times for each experiment. Black bars indicate the anti-p53-specific chromatin immunoprecipitation, and the grey bars represent the isotype antibody control. Following irradiation, the β-globin gene, which is expressed independently of p53, was decreasingly precipitated by anti-p53 and served as an additional negative control.

The nucleotide sequence surrounding GC box 3 is highly conserved between the murine and the human cyclin A1 promoter but does not contain a putative p53 consensus binding site (data not shown). Previously, we have shown that the Sp1 protein bound to this site, and it has been reported that p53 interacted with Sp1 protein (5, 40). We examined the involvement of Sp1 in the p53 reactivity of the cyclin A1 promoter in S2 Drosophila cells that lack Sp1 activity (Fig. 3C). The cyclin A1 promoter was activated by p53 only in the presence of Sp1. Chromatin immunoprecipitation indicated that p53 could be found in proximity to the cyclin A1 promoter following irradiation (Fig. 3D). As a positive control, we used the p21^CIP1 promoter, which is a transcriptional target for p53 and is also dependent on Sp1 activity (15). Analyses were performed with real-time reverse transcription-PCR, which is a reliable method to quantitate protein-DNA interaction in vivo (11).

The induction of the cyclin A1 promoter via a p53-dependent pathway hints at a probable mechanism for cyclin A1 upregulation after DNA damage and corroborates its putative role in DNA damage response. Therefore, we searched for cyclin A1-interacting proteins associated with DNA repair pathways to reveal the molecular function of cyclin A1 and its catalytic partner CDK2 in response to irradiation.

Yeast triple hybrid screen.

A yeast triple hybrid screen was performed to identify proteins that interact with the cyclin A1/CDK2 complex (Fig. 4A). The bait vector constitutively expressed a human cyclin A1-GAL4 binding domain fusion protein and conditionally expressed human CDK2 in the absence of methionine. The yeast strain carrying this plasmid was transformed with a human testis cDNA library, and 10⁷ independent clones were screened for histidine and adenine autotrophy and lacZ expression. Altogether, 80 positive colonies were identified, and the library cDNAs from these colonies were amplified and sequenced. Most of these cDNAs encoded testis-specific transcripts and an unknown gene. These data will be presented elsewhere. We also found fragments of the repair protein Ku70. The Ku70/Ku80 heterodimer plays an important role in DNA double-strand break repair and is expressed in all cell types (17).

FIG. 4.

Triple hybrid screening: interaction of cyclin A1 with Ku70/Ku80. (A) A yeast triple hybrid screen was performed with a Gal4-binding domain-cyclin A1 fusion protein as the bait, while human CDK2 was conditionally expressed. Screening was performed with a human testis cDNA library. (B) Two of the identified interacting clones contained a fragment of the human Ku70 cDNA. (C) Interaction of cyclin A1 and Ku70 in vitro. Baculovirus-expressed cyclin A1 or cyclin A1/CDK2 was incubated with GST, GST-Ku70, or GST-Ku80. After GST pulldown, binding was detected with anti-cyclin A1 antibody or anti-CDK2 antibody. (D) Deletion mutants of Ku70 were tested for their interaction with cyclin A1 in GST pulldown assays. Cyclin A1 binding was detected with anti-cyclin A1 Western blotting. The C terminus of Ku70 was required for the interaction with cyclin A1. (E) Interaction of cyclin A1 and Ku70 in vivo. Immunoprecipitations were performed with either anti-cyclin A1 or control immunoglobulin G. Ku70 coimmunoprecipitated with cyclin A1 in U937 cell lysate. (F) GST-Ku70, GST, or GST-Ku80 was incubated with baculovirus-expressed cyclin A1, CDK2, or cyclin A1/CDK2 in the presence of [γ-³²P]ATP. Lysate from wild-type baculovirus-infected insect cells was used as a control. Following SDS-PAGE, the gel was autoradiographed. GST-Ku70 protein was clearly phosphorylated by cyclin A1/CDK2, whereas GST-Ku80 was not. The nonspecific higher band appeared in all three GST preparations (GST, GST-Ku70, and GST-Ku80).

Cyclin A1 interacts with Ku70 and Ku80 proteins.

The cDNA isolated with the triple hybrid screen contained the C-terminal 295 amino acids of Ku70 (residues 315 to 609) in frame with the GAL4 activation domain (Fig. 4B). Residues 315 to 420 of Ku70 form the majority of the central β-barrel domain that is involved in Ku70-Ku80 heterodimerization (43). This central β-barrel domain binds the C-terminal α-helical region of the opposite heterodimerization partner.

GST pulldown experiments confirmed the interaction of both Ku70 and Ku80 with cyclin A1 in vitro (Fig. 4C). Complex formation between Ku70 and cyclin A1 required the C terminus of the Ku70 protein, as shown by GST pulldown experiments with Ku70 deletion constructs fused to GST (Fig. 4D). An isolated fragment of Ku70 (amino acids 201 to 400) containing most of the central β-barrel domain did not bind to cyclin A1 (data not shown). To further investigate protein interaction in vivo, we took advantage of the high levels of expression of cyclin A1 in the leukemia cell line U937 (25). The endogenous Ku70 protein coimmunoprecipitated with cyclin A1, indicating protein-protein interaction in vivo (Fig. 4E). These experiments provided evidence that cyclin A1 interacts with Ku70 in vitro and in vivo.

Cyclin A1/CDK2 phosphorylates Ku70 in vitro.

Ku70 contains both potential CDK1/CDK2 phosphorylation sites and recruitment motifs. GST-Ku70 and GST-Ku80 were used as phosphorylation substrates for lysates of insect cells expressing cyclin A1 or CDK2 or both (Fig. 4F). Following the in vitro kinase reaction, GST proteins were resolved by SDS-PAGE, and dried gels were analyzed by autoradiography. The lysate prepared from cells expressing both cyclin A1 and CDK2 phosphorylated GST-Ku70 to a much greater extent than the others. GST-Ku80 was not phosphorylated, indicating the specificity of Ku70 phosphorylation.

Cyclin A1 functions in DNA repair in vitro and in vivo.

An important function of Ku70 is to mediate the repair of DSBs in DNA through the nonhomologous end joining pathway. The comet assay (Fig. 2B and C) already hinted at defective DNA DSB in of cyclin A1^−/− cells. We therefore analyzed DSB repair in MEF cells. In wild-type MEFs, cyclin A1 expression was detected by quantitative reverse transcription-PCR at levels comparable to that in bone marrow (Fig. 5A). Cyclin A1 mRNA was not detected in MEF cells derived from the cyclin A1 deletion mutants (not shown). The MEFs derived from cyclin A1^−/− embryos proliferated significantly more slowly than wild-type MEFs after several passages in culture (C. Müller-Tidow et al., unpublished data). Thus, cyclin A1 is present in MEFs at low but functionally significant levels.

FIG. 5.

cyclin A1^−/− cells are NHEJ deficient. (A) Cyclin A1 mRNA expression levels in MEFs compared to expression in murine testis, bone marrow, and spleen. MEFs from cyclin A1^−/− mice did not express cyclin A1 mRNA (not shown). (B) An EGFP expression plasmid was linearized and incubated in vitro with different amounts of MEF cell extract from wild-type mice and from mice with homozygous cyclin A1 deletion to analyze DNA repair. Plasmids were run on an agarose gel and visualized by Southern blotting. Religated, supercoiled plasmids migrated faster than linearized plasmid. (C) Densitometry was performed on six independent end-joining analyses as shown in panel B with 50 and 75 μg of lysate. Indicated is the percentage of rejoined plasmids as calculated by dividing the rejoined and nonrejoined plasmids. Differences at 50 μg were statistically significant (P = 0.015), whereas the difference was not statistically significant when a larger amount of lysate was used. (D) In vitro-rejoined plasmids were PCR amplified, and 20 single clones were sequenced for each genotype to analyze the fidelity of DNA DSB repair. The bar graphs indicate the percentages of correctly joined plasmids. (E) NHEJ repair in wild-type and knockout fibroblasts was analyzed by luciferase activity: MEF cells were transiently transfected with linearized luciferase plasmids. An intact Renilla luciferase construct was cotransfected to normalize for transfection efficiency. The mean and standard error of the mean of three independent experiments are shown.

Extracts were prepared from wild-type and cyclin A1^−/− MEFs and analyzed in in vitro DNA repair assays. First, we analyzed the rejoining of a blunt-end linearized plasmid. Following incubation with the lysates, the plasmids were purified and analyzed by Southern blotting (Fig. 5B). There was significantly less end-joining activity in the lysates prepared from cyclin A1^−/− MEFs. These experiments were performed with several MEF batches and consistently showed that extracts from cyclin A1^−/− MEFs were impaired in DNA end joining. Combined data from six experiments are presented in Fig. 5C. No differences in NHEJ activity were found between heterozygous and wild-type MEF-derived extracts. The fidelity of the end joining was determined by sequencing after PCR amplification of the junction. Lysates from cyclin A1^−/− MEFs displayed a minor reduction in fidelity with respect to the residual end joining activity without statistical significance (Fig. 5D).

To analyze DNA DSB repair in MEFs in vivo, cells were transfected with a luciferase reporter plasmid (pGL3-control) previously linearized by digestion with HindIII, which cuts between the luciferase promoter and coding sequence, or with NcoI, which cuts at the initiation codon of the luciferase coding sequence. Under both conditions, wild-type MEFs showed higher levels of luciferase activity than cyclin A1^−/− MEFs (Fig. 5E).

To unequivocally decide whether the reduction in the rate of NHEJ in the cyclin A1^−/− MEFs was attributable to the absence of cyclin A1, we performed in vitro assays with lysates from cyclin A1^−/− MEFs supplemented with increasing amounts of recombinant cyclin A1 and CDK2 (Fig. 6A) There was a dose-dependent effect on religation of the plasmid: low doses of cyclin A1/CDK2 stimulated religation, whereas high levels of cyclin A1/CDK2 inhibited religation. A comparable result was obtained with nuclear extracts prepared from HeLa cells (HeLa cells express very low levels of endogenous cyclin A1 (22) (Fig. 6B).

FIG. 6.

Cyclin A1 stimulates NHEJ and homologous recombination DSB repair. (A) Cell extracts from cyclin A1^−/− MEF cells were incubated with linearized plasmid in the presence of increasing amounts of baculovirus-produced cyclin A1/CDK2. A total amount of 1.5 μg of baculovirus lysate containing 3 to 5% cyclin A1/CDK2 (≈40 ng) was serially diluted. The addition of small amounts of cyclin A1 increased rejoining activity, whereas high levels of cyclin A1 reduced rejoining. (B) Similar experiments were performed with nuclear extracts from HeLa cells, which endogenously express very low levels of cyclin A1. (C) Analysis of cyclin A1-activated DNA repair by end joining in vivo: Transfection of linearized luciferase reporter plasmid into NIH 3T3 cells with increasing amounts of cyclin A1 expression plasmid led to a dose-dependent effect on DNA repair and subsequent luciferase activity of the rejoined plasmid. A Renilla luciferase plasmid was cotransfected to standardize for transfection efficiency. The data represent mean and standard error of the mean of three independent experiments. (D) Microhomology-mediated NHEJ is activated by cyclin A1. The EJ-EGFP plasmid was transfected along with 10 μg of SceI and either 10 μg of a control or cyclin A1 or cyclin A2 expression plasmid into KMV5 cells (1). Cyclin A1 significantly activated NHEJ. (E) Transient transfection analyses of cyclin A1-activated DNA DSB repair by homologous recombination. Myeloid KMV5 cells were electroporated with HR-EGFP/3′EGFP (10 μg), a plasmid that leads to cellular EGFP expression upon DSB repair by homologous recombination (1). An expression vector for SceI (10 μg) was cotransfected, as well as different amounts of pcDNA3.1-cyclin A1. The total DNA amount of 30 μg was reached by addition of empty pcDNA3.1 plasmid when necessary. Transfection of an EGFP plasmid served as a positive control. Recombination frequency as assessed, and the number of EGFP-expressing cells within the transfected cell fraction was set at 100% in control cells not transfected with pcDNA3.1-cyclin A1. (F) DSB repair by homologous recombination with genome-integrated substrate. These assays were performed in stably transfected HR-EGFP-KMV5 cells with 1 μg of SceI nuclease expression vector and 1 μg of either control, cyclin A1, or cyclin A2 plasmid (1). Repair frequency was analyzed after 72 h. Data represent six analyses each (mean and standard error of the mean) of two independently performed experiments.

A similar pattern of dose dependence was observed in in vivo experiments with NIH 3T3 cells. A constant amount of linearized luciferase reporter plasmid was cotransfected with titrated amounts of a cyclin A1 expression plasmid: small amounts of cyclin A1 enhanced religation. Differences in repair activity were statistically significant (P < 0.05, Mann-Whitney U test)) at 0.01, 0.05, and 1 μg but not at 5 μg of cyclin A1 expression vector in cotransfection (Fig. 6C).

To rule out that repair in these assays was mediated by alternative mechanisms, we employed a second in vivo system that is based on microhomology-mediated repair to assess NHEJ (1). The cotransfected SceI nuclease cuts the EJ-EGFP plasmid in vivo. Repair of the plasmid by NHEJ but not by other repair mechanisms enables transcription of a complete EGFP coding sequence. EGFP expression is visualized after 24 and 48 h by flow cytometry. With this assay, we tested the effects of transiently transfected cyclin A1 and compared it with that of cyclin A2. Cyclin A1 significantly activated NHEJ after 24 h (P = 0.028) and after 48 h (P = 0.028) (Mann-Whitney U test) (Fig. 6D). Cyclin A2 effects were smaller and statistically not significant compared to those of the control vector at 24 h (P = 0.35, not significant) and at 48 h (P = 0.08, not significant) (Fig. 6D). These data confirmed that cyclin A1 activated NHEJ.

We further analyzed whether cyclin A1 also activated the more complex homologous recombination repair pathway. For this purpose, another modified EGFP plasmid (HR-EGFP) was cut in vivo by cotransfection of an expression plasmid encoding the SceI restriction enzyme. Repair of the plasmid by homologous recombination in KMV5 cells in vivo led to EGFP expression which was quantitatively analyzed with flow cytometry (1). Transient cyclin A1 expression enhanced homology-directed DNA DSB repair by about 50% (Fig. 6E). We further analyzed homologous recombination repair with substrates with genomic integration of the HR-EGFP system. Both cyclin A1 (P = 0.028) and cyclin A2 (P = 0.046) significantly activated homologous recombination after 72 h. No difference was found between cyclin A1 and cyclin A2 (P = 0.6, not significant) (Fig. 6F). Taken together, cyclin A1 and cyclin A2 stimulate homologous recombination repair, while cyclin A1 specifically activates NHEJ.

DSB repair activation by cyclin A1 depends on CDK2 and Ku proteins.

We next determined whether the cyclin A1-interacting proteins CDK2 and the Ku proteins were essential for the stimulatory effects of cyclin A1 on DSB repair. First, we tested the importance of CDK2 kinase activity by adding the CDK2 inhibitor olomoucine to cyclin A1^−/− MEF extracts supplemented with cyclin A1 and CDK2. In these experiments, olomoucine-exposed extracts showed significantly lower rates of DNA DSB repair (Fig. 7A).

FIG. 7.

Cyclin A1-activated DNA repair depends on CDK2 and Ku proteins. (A) MEFs from cyclin A1^−/− cells were incubated with baculovirus-expressed cyclin A1-CDK2 in either the presence or absence of the CDK2 inhibitor olomoucine. The end-joining efficiency of the extracts was tested with linearized plasmid, followed by Southern blotting. Inhibition of CDK activity strongly inhibited DNA DSB repair. (B) Different cyclin-CDK2 complexes expressed in the baculovirus system were standardized for similar histone H1 kinase activity by in vitro kinase assays for histone H1. (C) Cyclins in combination with CDK2, CDK2 alone, or cyclin A1 alone was tested for DNA DSB repair. Cyclin/CDK2 concentrations correspond to the concentrations tested for histone H1 kinase activity in panel B. Cyclin A1 or CDK2 alone did not activate DNA DSB repair, whereas the combination dose did so dependently. Cyclin A2/CDK2 also activated DNA DSB repair, but the activity was weaker than that of cyclin A1/CDK2. This result was confirmed in several independent experiments (data not shown; compare Fig. 6D). Cyclin E/CDK2 complexes did not have significant effects on DNA DSB repair. (D) Ku70 protein is instable in the absence of Ku80 (29). To verify the low levels of Ku70 in MEFs from Ku80^−/− and wild-type mice, lysates were analyzed for Ku70 protein expression by Western blot. (E) Cell extracts from Ku-deficient cells as shown in panel D were tested for activation of DNA DSB repair by cyclin A1/CDK2. Cyclin A1/CDK2 activated DNA DSB repair in extracts from wild-type but not Ku80^−/− MEF cells. (F) Recombinant Ku70 and Ku80 proteins expressed in the baculovirus system were added to Ku-deficient extracts. Cyclin A1-CDK2 activated DNA end joining in Ku protein-supplemented extracts but not in extracts lacking Ku proteins.

Then, we examined which cyclin-CDK complex could activate DNA DSB repair. Cyclin A1-CDK2, cyclin A2-CDK2, and cyclin E-CDK2 were expressed in the baculovirus system, and kinase activities were standardized for histone H1 kinase activity (Fig. 7B). Equivalent amounts were used to investigate DNA DSB repair (Fig. 7C). Cyclin A2-CDK2 complexes also activated DNA DSB repair, although to a lesser degree than cyclin A1-CDK2, as suggested by several independent experiments (Fig. 7C and data not shown). These results confirmed the NHEJ experiments in vivo (compare Fig. 6D). Cyclin E-CDK2 complexes, cyclin A1 alone, or CDK2 alone did not significantly alter end-joining repair. These data indicated that A-type cyclins and especially cyclin A1 are able to activate DNA DSB repair when complexed with CDK2.

While these data evidenced a role for cyclin A1-CDK2 in DNA DSB repair, it remained unclear whether these effects depended on Ku proteins. Ku80^−/− cells do not express Ku80 and also, because of decreased protein stability, do not express significant levels of Ku70 protein (Fig. 7D) (7, 29). Extracts from these cells and from the corresponding wild-type cells were prepared and tested for DSB repair activation by cyclin A1. Cyclin A1-CDK2 activated end joining in the wild-type cells but not in the Ku-deficient cells (Fig. 7E). Addition of recombinant Ku protein expressed as His-tagged proteins in the baculovirus system reconstituted the ability of cyclin A1-CDK2 to activate DNA DSB repair (Fig. 7F). These data demonstrated that the cyclin A1-CDK2 effects on DNA DSB repair by NHEJ are mediated by Ku proteins.

DISCUSSION

Mammalian cells express several CDK2 binding cyclins. For some of these cyclins, the role in cell cycle progression is clearly established, while the functions of others in vivo remain unknown.

Cyclin A1 is (in the absence of DNA damage) a tissue-specific cyclin with high levels of expression in the testis and in acute myeloid leukemia (44, 46). In acute myeloid leukemia, cyclin A1 is induced by various mechanisms, including c-Myb and the promyelocytic leukemia-retinoic acid receptor alpha oncogene (24, 25). Here, we demonstrate that cyclin A1 is induced by γ-irradiation via a p53-mediated mechanism. This was a surprising finding given that the other CDK2-associated cyclins are inhibited and/or repressed following p53 activation. In combination with the finding of increased radiosensitivity of cyclin A1^−/− cells, these results suggested a potential role for cyclin A1-CDK2 in the cellular response to DNA damage.

In a triple hybrid screen involving cyclin A1-CDK2 as the bait, we identified the Ku70 protein as a specific interaction partner. The Ku70 and Ku80 proteins function in DNA replication, recombination in B and T cells, and DNA repair (41). DNA replication critically depends on cyclin A2 (8), and therefore the Ku70-cyclin A1 interaction was more likely to play a role in other processes, since neither Ku70 nor cyclin A1 deficiency precluded DNA replication (19, 31). Also, meiotic recombination e.g., during spermatogenesis, utilizes elaborate regulatory mechanisms of DNA repair to avoid genomic instability. Cyclin A1 is essential for spermatogenesis (19, 42). Consequently, we analyzed the involvement of cyclin A1-CDK2 in the regulation of DNA DSB repair. Ku70 as well as its heterodimerization partner Ku80 bound to cyclin A1, and Ku70 was specifically phosphorylated by cyclin A1-CDK2 complexes in vitro. Phosphorylation of Ku70 protein has not been described before, and our findings suggest that cyclin A1-dependent phosphorylation might play a role in DNA DSB repair. Nonetheless, the functional consequences of Ku70 phosphorylation remain incompletely understood.

DNA DSB repair plays an important role in genomic stability (9, 10). The CDK2-associated cyclin E is a known inducer of genomic instability and is associated with a poor prognosis for several types of cancer (36). In contrast, cyclin A2 overexpression was not associated with genomic instability. Thus, it is likely that the relative expression patterns of the CDK-associated cyclins regulate DNA replication or DNA DSB repair. Changes in the expression patterns of different CDK2-associated cyclins (following DNA damage) might influence not only proliferation but also DNA DSB repair.

Cyclin A1 is essential during meiosis I in the male mouse, indicating a possible role of cyclin A1 in homologous recombination. In hematopoietic cells, expression peaks in S and G₂ phases, consistent with a role in homologous recombination between sister chromatids. Still, it seems to be contradictory that cyclin A1 activates NHEJ and at the same time stimulates homology-directed DSB repair, two competing pathways in DSB repair that are differentially regulated by Ku (13, 33). Interestingly, however, Ku also carries functions in modulating the DNA protein kinase enzyme (6), which in turn suppresses homologous recombination (2). Moreover, Ku also affects ATM, which phosphorylates multiple proteins involved in homologous recombination (48). Thus, cyclin A1-dependent phosphorylation of Ku may function not only in promoting NHEJ but also in additional relevant processes. A coordinated response is compatible with the concept that homologous repair and NHEJ are not completely separable and that coupling of invasion of one chromosome end into homologous sequences and subsequent NHEJ of the originally broken chromosome effectively prevents rearrangements between different chromosomes (35). Thus, cyclin A1 is expected to play a role in the maintenance of genome stability under physiological conditions, whereas aberrant expression may cause deregulated DSB repair and therefore contribute to leukemogenesis.

How specific is the involvement of cyclin A1 in DNA DSB repair? In our in vitro experiments, cyclin A1 activated DNA DSB repair more strongly than cyclin A2. Accordingly, in NHEJ in vivo assays, cyclin A1 again was more active. Importantly, repair depended on CDK2, and cyclin E did not have a significant effect on DNA DSB repair. In homologous recombination repair, both cyclin A1 and cyclin A2 were active. These data indicate that with the experimental conditions used, A-type cyclins can activate DNA DSB repair. However, several points of evidence indicate that cyclin A1 might have a special role in DNA DSB repair. First, cyclin A1 was the only cyclin to be induced by γ-irradiation and p53, whereas cyclin A2 expression was consistently repressed. Thus, in circumstances that require DSB repair, cyclin A1 is induced, whereas cyclin A2 is downregulated. Second, cyclin A1-deficient MEF cells, which readily expressed cyclin A2, were impaired in DNA DSB repair. Third, increased radiosensitivity was observed in hematopoietic progenitor cells and MEFs of cyclin A1^−/− mice. The high levels of cyclin A2 expressed in these proliferating cells did not protect from the radiation damage, indicating a special function for cyclin A1. Nonetheless, our data do not rule out that cyclin A2 also plays a role in DNA DSB repair.

What is the relationship between cyclin A1 induction by DNA damage and p53 and cyclin A1-activated DNA DSB repair? The p53 tumor suppressor is known to influence NHEJ. However, conflicting data exist about the role of p53 in DNA DSB repair. Several authors reported that p53 repressed NHEJ activity (1, 4, 16, 30). On the other hand, it has been reported that p53 activates DNA DSB repair (39, 47). These divergent findings do not merge easily into a coherent picture. It is tempting to speculate that the effects of p53 on DNA DSB repair may partially depend on the presence or absence of cyclin A1 in a particular cell type. Also, the late induction of cyclin A1 by p53 suggest that the effects of p53 on DNA DSB repair might vary at different times after genotoxic damage.

Taken together, our data provide evidence for a novel pathway linking p53 activity with cyclin A1 induction and subsequent DNA DSB repair.