Senescence-Associated Oxidative DNA Damage Promotes the Generation of Neoplastic Cells

Both in vitro and in vivo, as a result of time and cumulative divisions, normal cells enter senescence, characterized by an enlarged morphology, lipofuscin accumulation, increased autophagic activity, cell cycle arrest, and frequent polynucleation (1–5). It is accepted that senescence results from cumulative oxidative damage and telomere shortening, each probably acting to a different degree according to cell type or environmental conditions. Oxidative damage is due mainly to enhanced production of reactive oxygen species and concerns all macromolecules. Oxidation of proteins and lipids may explain accumulation of lipofuscin and other damaged components (6, 7), and oxidative DNA damage may be a signal for cell cycle arrest (8). Telomere shortening is due primarily to the end-replication problem. It leads to deprotected chromosome ends that behave like DNA breaks and signal for cell cycle arrest (9, 10). Because of its associated cohort of damage and irreversible cell cycle arrest, senescence has been viewed as a tumor-suppressing mechanism that stops proliferation of genetically altered cells (11). Consequently, it has been assumed that, to become tumoral, a cell has to bypass senescence. Yet, this assumption is questionable regarding in vivo data: the incidence of carcinomas in humans is 2- to 3-fold higher in the 60 to 79 age bracket than in the 40 to 59 age bracket; cancer is frequent in patients suffering from progeroid syndromes (12); and when ageing is delayed by caloric restriction, the incidence of cancer decreases (13). Hence, aging and tumorigenesis are positively linked, suggesting that senescence might precede and sustain tumorigenesis.

Here, after monitoring long-term cultures of human primary keratinocytes, we report the systematic and spontaneous emergence from senescence of cells displaying some transformed and tumorigenic characteristics, suggesting that senescence could indeed be a tumor-promoting state per se. We show that post-senescent–emerging cells potentially originate from all initial cells and not from a special subpopulation and that they have not bypassed senescence but have been formed, on the contrary, through division of cells with already senescent characteristics. We present evidence that the molecular switches necessary for emergence are set during senescence by reactive oxygen species accumulated with senescence. This supports the view that senescence-associated reactive oxygen species might be both a cause of senescence through their deleterious effects and a cause of emergence of pretumoral cells through their mutagenicity.

Cell culture and senescence-associated β-galactosidase assays. Normal human epidermal keratinocytes (NHEK) purchased from Clonetics were obtained from eight female donors: five Caucasians (ages 60, 31, 18, 37, and 19 years), one Black (age 33 years), and one Asian (age 40 years). They were grown in KGM-2 BulletKit medium consisting of modified MCBD153 with 0.15 mmol/L calcium, supplemented with bovine pituitary extract, epidermal growth factor, insulin, hydrocortisone, transferrin, and epinephrin (Clonetics). Such a serum-free low-calcium medium has been shown to minimize keratinocyte terminal differentiation (14). The number of population doublings was calculated as follows at each passage: population doubling = ln(number of collected cells / number of plated cells) / ln2. Senescence-associated β-galactosidase assays were done as initially described (15).

Western blotting. Cells were lysed in 27.5 mmol/L HEPES (pH 7.6), 1.1 mol/L urea, 0.33 mol/L NaCl, 0.1 mol/L EGTA, 2 mmol/L EDTA, 60 mmol/L KCl, 1 mmol/L DTT, and 1.1% NP-40. The total protein concentration was measured with the Bio-Rad protein assay. Proteins were resolved by SDS-PAGE and transferred to nitrocellulose membranes (Hybond-C extra; Amersham). Equal loading was checked by Ponceau red staining. Primary antibodies were mouse anti-rat proliferating cell nuclear antigen (DAKO), anti-human involucrin, anti-human cytokeratin 14 (Chemicon), anti-human E-cadherin (Transduction Labs), and anti-human actin (Santa Cruz). The secondary antibody was a peroxidase-conjugated anti-mouse IgG (Jackson ImmunoResearch Laboratories).

Karyotype analyses. Metaphase spreads were obtained using a standard method. Briefly, cells were incubated 1 h in Karyomax Colcemid (Invitrogen), trypsinized, and incubated in a 60 mmol/L KCl hypotonic buffer. Cells were fixed with methanol/acetic acid solution (3:1, v/v), spread onto frozen slides, and air-dried overnight. For MGG coloration, slides were incubated in 0.035% trypsin diluted in PBS 1 min 40 at 4°C, washed, and stained by 0.024% (w/v) Giemsa (Sigma) in Gurr's buffer 3 min 25 at room temperature. Metaphases were analyzed by the Cytovision software for G-banding. For multiplex fluorescence in situ hybridization, slides were fixed in 4% formaldehyde in PBS for 2 min, washed, and treated with pepsin (Sigma) at 1 mg/mL for 10 min at 37°C at pH 2.0. After a wash in PBS, formaldehyde fixation and washes were repeated and the slides were dehydrated with ethanol and air-dried. They were then hybridized with multiplex fluorescence in situ hybridization probes (MetaSystems) according to the manufacturer's recommendations.

Anatomopathologic analysis of tissue samples and in situ hybridization with human Alu sequence probes. Tissue samples were formalin-fixed, paraffin-embedded, sectioned, and processed for May-Grunwald-Giemsa stainings according to standard procedures. Images were recorded using an Axioplan2 Zeiss microscope using Axiovision Software. For Alu in situ hybridization, sections were treated with proteinase K and post-fixed. FITC-labeled Alu probe (BioGenex) was added and slides covered with sealed coverslips were heated to 90°C for 5 min and then to 37°C overnight. Post-hybridization washes were carried out at 40°C with 2× SSC/0.1% SDS for 2 × 5 min, 0.1× SSC for 10 min, and 2× SSC/0.1% SDS for 5 min. Unspecific binding sites were blocked with 3% bovine serum albumin in PBS + 0.1% Tween 20 for 1 h at room temperature followed by an avidin-biotin-blocker (Vector Laboratories). Probe detection was achieved by incubation with a biotinylated anti-FITC antibody (Vector Laboratories) followed by rhodamine RedX–conjugated streptavidin (Jackson ImmunoResearch) and nuclei counterstaining with Hoechst 33258 (40 ng/mL). Slides were examined under a Zeiss confocal microscope LSM70. Images were recorded using the software Zen.

Adenoviral vector encoding MnSOD. The human MnSOD cDNA, obtained after retrotranscription, was amplified by PCR and propagated in pcDNA3.1. The cDNA was then digested with EcoRI and inserted into the pAdCMV2 vector at the XbaI sites after filling with Klenow polymerase. Recombinant adenovirus vectors were obtained by homologous recombination in Escherichia coli BJ5183 as described previously (16). Viral stocks were amplified after infection of N52.E6 cells (17). Recombinant adenoviruses were purified with the ViraBind Adenovirus purification kit (Cell Biolabs) and titrated with the Adeno-X rapid titer kit (BD Biosciences Clontech). Cells were infected by adding virus stocks directly to the culture medium at an input multiplicity of 200 viral particles per cell.

Comet assays. Ten thousand cells were embedded in 80 μL of 0.5% low melting point agarose at 37°C, and the suspension was immediately laid onto a Trevigen cometslide. Agarose was allowed to solidify at 4°C for 30 min. The slides were then immersed in prechilled Lysis Solution [2.5 mol/L NaCl, 100 mmol/L EDTA, 10 mmol/L Tris, 1% Triton (pH 10)] at 4°C for 60 min and equilibrated in the electrophoresis buffer for 20 min at room temperature. The electrophoresis buffer was either 89 mmol/L Tris, 89 mmol/L boric acid, 2 mmol/L EDTA (pH 8) or 300 mmol/L NaOH, 1 mmol/L EDTA (pH 13). Migration was carried out at 1 V/cm for 20 min. After migration, the slides were neutralized with 0.4 mol/L Tris (pH 7.5) and stained with either SYBR Green I (Trevigen) according to the manufacturer's recommendations or propidium iodide (2.5 μg/mL). Tail moments were analyzed with the Tritek Comet Score freeware.

Immunofluorescence staining of 8-oxoguanines. Cells were fixed in 4% paraformaldehyde for 15 min at 4°C, dehydrated at −20°C in 70% and 95% methanol for 3 min each followed by 99% methanol for 30 min, and rehydrated for 3 min in 95% and 70% methanol at −20°C and three times in PBS. The anti–8-oxo-7,8-dihydroguanine (8-oxoG) antibody is from Trevigen. Because it recognizes 8-oxoG in both ribonucleotides and deoxyribonucleotides, we performed a RNase A treatment, which did not affect the percentage of positive cells or the intracellular localization of the signal (data not shown).

Keratinocytes spontaneously give rise to transformed and tumorigenic cells. We monitored the behavior in culture of NHEK from eight female adult donors of different ages and races. Such cells first divided over ∼3 weeks, making 15 to 20 population doublings, and then reached a plateau at which they display the characteristics of senescence, including increased cell size, polynucleation, accumulation of vacuoles and various damaged components, senescence-associated β-galactosidase activity, and decreased proliferating cell nuclear antigen expression (Fig. 1). After a few days to 3 weeks at this plateau, several clones of small cells appeared spontaneously and systematically in all cultures, whatever the donor, while most senescent cells died. These cells (henceforth called “post-senescent–emerging cells”) were found to have resumed expression of proliferating cell nuclear antigen and to grow again for 5 to 15 population doublings, after which they reached a second plateau from which we observed a second emergence. The second emerging cells appeared more transformed than the initial ones (Fig. 1B). We first believed these cells were immortalized and named them ImKs for immortal keratinocytes (numbered IMK, IMK2, IMK3, …, to identify the donor). ImKs, however, underwent up to 60 population doublings but then stopped and died (data not shown). Neither ImKs nor post-senescent–emerging cells showed any resumption of telomerase activity (Supplementary Fig. S1). We estimated the emergence frequency, that is, the number of emerging clones generated per cell at the plateau, by plating plateau cells at low density and counting the emerging clones. Depending on the experiment, the frequency of the first wave of emergence ranged from 10⁻⁵ to 10⁻² and that of second wave was 10⁻⁵ to 10⁻⁴. These frequencies are considerably higher than 10⁻⁷, the frequency of immortalization of SV40-transformed human fibroblasts (18, 19).

Post-senescent–emerging cells looked partly transformed, with some clones displaying a fibroblastoid morphology associated with a tendency to scatter (Supplementary Fig. S2A). The expression of involucrin and keratin 14, two markers of keratinocyte differentiation, increased with senescence and decreased again with emergence (Supplementary Fig. S2B). That of E-cadherin, involved in cell-cell interaction, also slightly decreased in post-senescent–emerging cells (Supplementary Fig. S2C). A transcriptomic analysis on DNA microarrays revealed that, on the 50 most up-regulated and the 50 most down-regulated genes in post-senescent–emerging cells, 15 turned out to be linked to adhesion or migration, 6 to cytoskeleton structure or dynamics, 9 to senescence, oxidative stress, or DNA damage, 9 to cell cycle progression or cell death, and 10 to diverse cancer-related pathways (Supplementary Table 1). Hence, ∼50% of the genes whose expression changes in emerging keratinocytes are relevant to transformation. We also investigated the karyotypic of post-senescent–emerging cells and ImKs by analyzing metaphases by G-banding and multiplex fluorescence in situ hybridization. Post-senescent–emerging cells displayed no karyotypic aberrations. In contrast, 100% of ImK metaphases displayed various aberrations, mainly translocations (Fig. 2).

Finally, we investigated the tumorigenic potential of emerging cells. Pre-senescent, post-senescent–emerging cells or ImKs were injected in the flank of nude mice. MDA-MB-231 and NIH-3T3 were used as controls. As expected, MDA-MB-231-injected mice showed significant tumors after 4 weeks. Tumors were also recorded in NIH-3T3–injected mice but only after 17 weeks. In mice injected with pre-senescent, post-senescent–emerging cells and ImKs, the xenograft poorly developed. However, from the 19th week onward, disseminated skin lesions appeared away from the injection site in 4 of 5 and 6 of 6 mice injected with post-senescent–emerging cells and ImKs, respectively (Fig. 3A). Macroscopically, these lesions resembled early nonmelanoma skin carcinomas (Fig. 3B). Anatomopathologic analyses indicated hyperplasia, hyperkeratotic plaques, and actinic keratosis as most frequent precancerous phenotypes. Important mastocytosis was recorded facing hyperplasia and hyperkeratotic plaques, particularly at sites of basal lamina ruptures (Fig. 3B). Because these lesions developed very lately and away from the injection site, it was necessary to prove they really derive from the injected cells. We therefore performed the detection of primate-specific Alu sequences by fluorescence in situ hybridization. The results showed the presence of human cells within the epidermis, at the lesion sites (Fig. 3C).

Taken together, these results suggest that the cell populations present at both growth plateaus are able to generate partially transformed and moderately tumorigenic cells able to disseminate, with more and more marked phenotypes from the first emergence wave to the second.

Post-senescent–emerging cells are formed from a few senescent cells by an unusual budding mitosis mechanism. One of our concerns was to elucidate the origin of post-senescent–emerging cells. Our first hypothesis was they might come from an initial subpopulation of already transformed cells present in the explants despite the healthy status of the donors. To test this hypothesis, we performed monoclonal cultures of young NHEKs, which we conducted to senescence and then monitored for emergence. Emerging clones appeared in ∼75% of the cultures (Supplementary Table 2), indicating that emerging clones are not the progeny of a restricted initial subpopulation but rather that almost all initial young cells have the potential to yield emerging cells. This invalidated our initial hypothesis. We then reasoned that emerging cells might be generated during senescence. To test this hypothesis, we sorted senescent cells by fluorescence-activated cell sorting from a pre-senescent population as cells with the highest forward and scatter factors, that is, the largest and most granular. Sorted cells were plated and stained with fluorigenic filiation tracers (Vybrant diI or Vybrant CFDA SE). After ∼1 week, emerging cells arose around some senescent cells and were stained by the fluorigenic tracers (Supplementary Fig. S3), proving that they are directly generated through division of fully senescent cells.

The fact that senescent cells can divide was surprising, because numerous studies have shown that proliferation of senescent cells is irreversibly impaired by their short telomeres. However, in the case of keratinocytes, it has been shown that telomerase reexpression alone is insufficient to bypass senescence (20), suggesting that telomere length is not a limiting factor in this cell type. To confirm this point, we examined NHEK telomeres at senescence. Southern blot analysis showed that the telomere length continuously decreased from ∼9 kb in young NHEKs to 6 kb at senescence and then 5 kb in emerging cells (Supplementary Fig. S4A). However, teloFISH analysis revealed that most cells at both plateaus still had substantial telomeres on all their chromosomes and that only a minority displayed some chromosomes with very short to undetectable telomeres likely to cause irreversible cell cycle arrest (Supplementary Fig. S4B and C). Senescent cells, including keratinocytes, have also been described as irreversibly arrested through induction of cyclin-dependent kinase inhibitors p16 and p21 (20–22). A quantitative reverse transcription-PCR analysis showed, as expected, that both cyclin-dependent kinase inhibitor mRNAs increased at the first plateau but decreased again in post-senescent–emerging cells (Supplementary Fig. S5), indicating that their up-regulation is only transient. Bromodeoxyuridine incorporation assays indicated, as expected, that most typical polynucleated senescent cells were bromodeoxyuridine-negative but revealed that some were bromodeoxyuridine-positive (Supplementary Fig. S6A). Moreover, we observed some typical large polynucleated senescent cells with one nucleus in metaphase after a colcemid treatment (Supplementary Fig. S6B). Taken together, these experiments indicate that, although overall growth of the culture is arrested at the senescence plateau, senescent keratinocytes have still a division potential regarding their telomere length and are cell cycle arrested but not irreversibly and some senescent cells actually divide.

We next wondered by what cell division mechanism a senescent cell, very enlarged and littered with damaged components, gives rise to small cells with a clear cytoplasm. Multiple microscopic examinations suggested that senescent cells generate emerging cells by an unusual asymmetric mitosis mechanism, remaining budding yeast cells. First, on examining trypsin-dissociated fluorescence-activated cell sorted senescent cells by phase-contrast microscopy, we observed large cells with one, two, or three attached buds (Supplementary Fig. S7A). Second, in routine cultures on plastic, emerging cells were almost always observed gathered around a large senescent cell, to which some appeared still attached by a pedicle (Supplementary Fig. S7B). To formally evidence this attachment, we analyzed by confocal microscopy the cytokeratin 14 network. Optical transverse sections revealed cytoskeleton continuity between the senescent cell and some surrounding emerging cells (Supplementary Fig. S7C). Finally, despite the challenge due to the low fraction of senescing cells actually producing emerging cells, we managed videomicroscopy and succeeded in capturing three sequences of images showing two slightly different mechanisms of budding mitosis. In two cases, a large multinucleated senescent cell generated several small daughter cells by budding cytokinesis (data not shown). In the third case, we observed a typical senescent cell with already several nuclei, among which three underwent an additional full mitosis generating a daughter cell budding out of the senescent mother (Supplementary Fig. S8 and Supplementary Video).

NF-κB > MnSOD > hydrogen peroxide pathway is causal in both senescence and emergence. We have shown previously that NHEK senescence arises in part through hydrogen peroxide (H₂O₂) accumulation. This accumulation was shown to result from an activation of NF-κB and ensuing up-regulation of MnSOD, responsible for the dismutation of O₂·⁻ to H₂O₂. Without any co-up-regulation of downstream H₂O₂-degrading enzymes, this leads to H₂O₂ accumulation (23). H₂O₂ being mutagenic, we hypothesized here that it might also contribute to emergence.

To test the involvement of the NF-κB > MnSOD > H₂O₂ pathway in emergence, we first treated young NHEKs with a concentration of H₂O₂ that we previously established as inducing premature senescence (23), monitored cells until they all displayed the senescent phenotype, and then stopped the treatment and waited for potential emergence. Emerging clones did appear ∼10 days later at a frequency of ∼10⁻⁴ (Fig. 4A). We then examined whether MnSOD overexpression might have the same effect. Young NHEKs were infected with an adenoviral vector encoding MnSOD. A premature senescence phenotype arose after ∼3 days followed 10 days later by emergence. We checked that both senescence and emergence occurred without any change in expression of several other antioxidant enzymes (Fig. 4B). We finally examined whether antioxidants or NF-κB inhibitors could inhibit emergence. As antioxidant, we used catalase, which specifically degrades H₂O₂ and was already shown to delay keratinocyte senescence (23). We also used N-tert-butyl-hydroxylamine, a more general antioxidant shown to target mitochondria and to reduce nuclear DNA damage (24). To inhibit NF-κB activity, we used sulfasalazine and gliotoxin, two weak inhibitors we have shown previously to delay senescence without inducing massive apoptosis (23). A few days after having treated pre-senescent cells with one of these drugs, emerging clones appeared in 85% of the control wells compared with 50% of those treated by catalase, 0% of the wells with (10 μmol/L) N-tert-butyl-hydroxylamine–treated cells, and 0% of the wells containing NF-κB inhibitors (Fig. 5).

Taken together, these results show that a H₂O₂ accumulation resulting from an activation of NF-κB and an unbalanced antioxidant enzyme expression is sufficient to induce a post-senescence emergence similar to the spontaneous one.

To further test the hypothesis that H₂O₂ induces emergence through its mutagenicity, we searched for mutagenic oxidative DNA damage in senescent cells. We first investigated DNA single-strand breaks (SSB), which are known to be induced by H₂O₂ (25, 26), by comet assays that unable to distinguish SSB from double-strand breaks (DSB). The results show that SSB were predominant, increasing ∼2-fold with senescence and affecting ∼20% of cells. Cells displaying DSB were rare, even at senescence. H₂O₂ did not induce any change in DSB but induced a dramatic increase in SSB (Fig. 6A). Accordingly, a catalase treatment decreased ∼2-fold the number of SSB per cell (Fig. 6B). H₂O₂ is also known to induce oxidation of bases in nuclear and mitochondrial DNA and in the nucleotide pool, the most common being 8-oxoG (27). By immunofluorescence, ∼20% of senescent cells were found to display 8-oxoG, in their cytoplasm and nucleus, compared with only 3% of young cells (Fig. 6C). When cells were treated with catalase, the percentage of affected senescent cells was reduced 4.25-fold (Fig. 6C). Conversely, treatment of young cells with H₂O₂ increased the percentage of 8-oxoG–positive cells ∼5-fold (Fig. 6C). Thus, at least two types of oxidative mutagenic damage, SSB and 8-oxoG, accumulate during senescence in correlation with the H₂O₂ level.

Numerous studies have reported that senescence is an irreversible growth arrest associated with telomere shortening. This has led to viewing senescence as a tumor-suppressing phenomenon. However, most of these studies were done with human fibroblasts, which are not the most relevant cell model for studying the molecular links between tumorigenesis and ageing, because sarcomas are rare in humans and their incidence does not depend on age (National Cancer Institute statistics). Using normal human keratinocytes, we show that cells with moderate transformed and tumorigenic characteristics systematically and spontaneously emerge from senescence. We have observed a similar emergence with epithelial mammary cells, as already described by others (28–30), although these emerging cells never gave rise to a second emergence (data not shown). Therefore, we propose this in vitro post-senescence emergence as a model for studying the very first steps of carcinogenesis. However, it is not yet clear whether this model can be generalized to all carcinomas, because we never observed any emergence with prostatic epithelial cells under standard culture conditions (data not shown).

It is often assumed that bypassing senescence is an obligatory step for tumorigenesis. We show here that emerging cells have not bypassed senescence but are instead generated from fully senescent cells via an unusual budding-mitosis mechanism. Although largely ignored by the scientific community, this kind of cell division, specific to senescent or DNA-damaged cells, has already been described by two groups (31, 32) and called “neosis.” We highlight here the importance of such a mechanism by showing that cells generated in this way are tumorigenic. We have ruled out the hypothesis that emerging cells come from an initial subpopulation of already transformed cells. On this last point, our results are in disagreement with those of Tlsty (33), suggesting, in the case of epithelial mammary cells, that emerging cells come from cells preexisting in explants and having a hypermethylated p16 promoter. We show here that the decrease in p16 expression in post-senescent–emerging keratinocytes is only transient, because p16 is again up-regulated at the second plateau. Only in ImKs does expression of p16 decrease to a very low level compatible with epigenetic extinction.

Through this study and a preceding one (23), we show that the NF-κB > MnSOD > H₂O₂ oxidative stress pathway is causal in both senescence and emergence. This conclusion is based on the observations that NF-κB inhibitors and antioxidants delay the occurrence of the senescence plateau and decrease the emergence frequency, whereas, conversely, NF-κB or MnSOD overexpression or H₂O₂ treatment induces premature senescence followed by emergence. We report the presence, in senescent cells, of at least two types of oxidative DNA damage, SSB and 8-oxoG, both potentially causing point mutations (27, 34, 35). That they play a causal role in emergence is supported by the consistent correlation between the percentage of cells affected by these damages and the emergence frequency: this percentage rises on H₂O treatment (which triggers senescence and emergence) and drops after senescence-delaying antioxidant treatments. Furthermore, the proportion of cells with SSB and 8-oxoG was always much higher than the proportion of emerging cells, making it statistically possible for these alterations to affect a favorable cocktail of oncogenes, tumor suppressor genes, and/or other crucial regulators of adhesion, migration, cell cycle, or cell death, as suggested by the transcriptomic changes observed in emerging cells. The stochastic nature of the events leading to emergence is supported by the fact that emergence is always multiclonal, each clone having its specificities as regards morphology, doubling time, and life span (data not shown). In accordance with this mutagenic motor role of oxidative stress for emergence, it was shown that the spontaneous immortalization of mouse embryonic fibroblasts after senescence is accompanied by a 3-fold increase in point mutations resulting from oxidative stress (36). It was also reported that a mutation in codon 61 of the Ha-ras gene spontaneously occurs in mouse keratinocytes that, similarly to human keratinocytes, form emerging foci at the senescence plateau (37). We checked for such a mutation in four clones of post-senescent–emergent NHEKs but did not find it (data not shown). This suggests that emergence cannot rely on a unique mutation, even of a major oncogene, but probably necessitates multiple genomic alterations.

In conclusion, the results presented here suggest that the initiating events in tumorigenesis may result from the mutagenicity of the oxidative stress to which senescing cells are subject. Hence, senescence and its associated oxidative stress might be viewed as endogenous carcinogens, this providing a molecular explanation of the link between advanced age and increased cancer incidence. The presence of cells with senescence markers has been evidenced in some premalignant lesions of young people, such as congenital naevi and benign prostate hyperplasia (38, 39). Thus, oxidative stress, whether it results from normal aging, from a special local hormonal environment as in the prostate, or from inherited oncogene activation as in congenital naevi, may generate (prematurely) senescent cells from which cancer-initiated cells have a high risk of emerging.

No potential conflicts of interest were disclosed.

Grant support: PPF Bioinformatique of Lille 1 University, Association pour la Recherche contre le Cancer, Ligue contre le Cancer, Conseil Régional NPdC, European Regional Development Fund, European Integrated Project RISC-RAD (FI6R-CT2003-508842), and Contract EDF V3-103. Institut Pasteur de Lille, Région NPdC, and Société Française du Cancer (K. Gosselin); Ministry of Research and FRM (S. Martien); Institut Pasteur de Lille (N. Malaquin); CEA (P. Ostoich); and Association pour la Recherche contre le Cancer (C. T'Kint de Roodenbeke).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Fabrice Nesslany for technical advice on comet assays, Nathalie Jouy and Brigitte Quatannens (FACS Facility of IMPRT-IFR114 and IBL, respectively), Didier Deslee, Elisabeth Werkmeister, and Antonino Bongiovanni (Microscopy Platform at the Institut Pasteur de Lille), and Géraldine Pottier and Debbie Williams for excellent technical help.