Author contributions: N.H.-N.: conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; T.N.: conception and design, financial support, collection and/or assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; M.M. and P.W.A.: financial support, provision of study materials, data analysis and interpretation, and final approval of manuscript; J.D.: provision of study materials, collection and/or assembly of data, and final approval of manuscript; P.J.G. and M.J.: collection and/or assembly of data, data analysis and interpretation, and final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS July 20, 2012.
Human embryonic stem cells (hESCs) tend to lose genomic integrity during long periods of culture in vitro and to acquire a cancer-like phenotype. In this study, we aim at understanding the contribution of point mutations to the adaptation process and at providing a mechanistic explanation for their accumulation. We observed that, due to the absence of p21/Waf1/Cip1, cultured hESCs lack proper cell cycle checkpoints and are vulnerable to the kind of DNA damage usually repaired by the highly versatile nucleotide excision repair (NER) pathway. In response to UV-induced DNA damage, the majority of hESCs succumb to apoptosis; however, a subpopulation continues to proliferate, carrying damaged DNA and accumulating point mutations with a typical UV-induced signature. The UV-resistant cells retain their proliferative capacity and potential for pluripotent differentiation and are markedly less apoptotic to subsequent UV exposure. These findings demonstrate that, due to deficient DNA damage response, the modest NER activity in hESCs is insufficient to prevent increased mutagenesis. This provides for the appearance of genetically aberrant hESCs, paving the way for further major genetic changes. Stem Cells2012;30:1901–1910
Human embryonic stem cells (hESCs) are derived from the inner cell mass of the blastocyst and genetically programmed for both self-renewal and pluripotency. Key determinant of these fates are transcription factors, including OCT4, SOX2, and NANOG, which activate a common set of genes required for self-renewal, while repressing genes important for differentiation [1, 2]. In culture, hESCs form a dynamic system, with subsets of cells that possess distinct clonogenic and differentiation capacities. The stage-specific embryonic antigen-3 (SSEA3) is a key surface marker to determine the state of hESCs [3–5]. SSEA3-positive (SSEA3+) cells are the major population in both normal and culture-adapted hESCs and differ from the SSEA3-negative (SSEA3−) subset in terms of gene expression patterns, cell cycle profile, and clonogenicity . In general, cells that lose SSEA3 expression appear to be about to commit to differentiation [4, 7]. Subsequently, expression of other cell surface markers, TRA-1-60, TRA-1-81, and SSEA4, is sequentially downregulated .
Ideally, while hESCs are in culture, genome integrity should be preserved to avoid abnormal development and tumorigenesis. However, these are rapidly proliferating cells, characterized by leaky mitotic checkpoints, as illustrated by the high frequency (30%–60%) of aneuploidy in preimplantation-stage embryos [8, 9]. In long-term cultures, hESCs undergo culture adaptation and acquire multiple genetic and epigenetic changes. Culture-adapted hESCs have abnormal karyotype, increased copy number variations, sites of loss of heterozygosity, and increased proliferation rate and resistance to stress-induced apoptosis, all characteristic features of tumor cells [10–12]. Moreover, low-passage number hESCs transferred into severe combined immunodeficiency (SCID) mice form teratoma, whereas the transfer of adapted hESCs results in teratocarcinoma, indicating that culture adaptation leads to cell transformation and increased malignancy [13–15].
It is possible that genomic instability of hESCs in culture begins with progressive accumulation of point mutations, prior to more substantial genetic changes and selection of faster proliferating cells. It is thought that uncontrolled cell cycle checkpoints and abnormal DNA damage response and repair are among the mechanisms contributing to hESCs adaptation and major karyotypic changes [16–18]. We lack, however, information on the generation of point mutations in hESCs and mechanisms contributing to this process.
Deficiencies in the nucleotide excision repair (NER) pathway are a well-established cause of point mutations and subsequent carcinogenic transformation. NER repairs helix-distorting DNA lesions, including UV-induced cyclobutane pyrimidine dimers (CPDs) and 6-4 pyrimidine pyrimidone photoproducts [(6-4)PPs], bulky chemical adducts caused by food contaminants and air pollutants, protein-DNA adducts, DNA cross-links, and even some forms of endogenous oxidative damage [19, 20]. Importantly, NER efficiency is known to vary widely between cell types [21–23]; for instance, we have reported that NER is downregulated in terminally differentiated cells [24–26] and in temporarily quiescent cells [27, 28], by a mechanism involving decreased ubiquitination of an NER factor [29, 30]. It was thus conceivable that NER could likewise be modulated in hESCs, possibly through a completely different mechanism than in the above cells but potentially explaining their propensity for genomic instability.
The best defined model to study NER is repair of UV-induced lesions. Aside from practical considerations, like the ability to inflict a precise amount of lesions at a precise time, UV allows analysis of NER on a very poor and a very good substrate in the same DNA sample. While (6-4)PPs are highly distorting lesions readily recognized and rapidly repaired by NER , CPDs remain capable of base-pairing with the opposite strand and are repaired slowly at the global genome level . Additionally, UV produces a typical mutational signature , which we exploited to appraise UV-induced mutagenesis .
UV irradiation initiates signaling pathways that couple NER and cell cycle checkpoints, culminating with cell cycle arrest that provides time for DNA repair and insures genomic integrity in surviving cells. In NER-proficient cells, it is believed that damage processing by NER induces phosphorylation of the Ataxia Telangectasia and Rad3 related (ATR) kinase and initiates signaling pathways that target specific substrates at different phases of the cell cycle, controlling the G1/S, intra-S, and G2/M checkpoints . The outcome of the DNA damage response is different in an NER-deficient background. Depending on the amount of damage, cells with low NER activity may either undergo apoptosis or resort to error-prone translesion synthesis (TLS) DNA polymerases to bypass DNA lesions during S-phase, resulting in the generation of point mutations and genomic instability . In this study, we analyze the relative efficiency of NER with respect to DNA damage response and cell cycle checkpoints and their contribution as a whole to the generation of point mutations in hESCs.
MATERIALS AND METHODS
Low-passage number H7S14 and Shef5 cells were maintained as described  on a feeder layer of inactivated mouse embryo fibroblasts (MEF), in knockout Dulbecco's modified Eagle's medium (DMEM) supplemented with 20% knockout serum and 4 ng/ml basic fibroblast growth factor (bFGF) (all from Gibco, Paisley, UK, www.invitrogen.com). In some cases, a final passage was done on plates coated with Matrigel (BD Biosciences, Oxford, UK, www.bdbiosciences.com) in MEF-conditioned medium.
Cells were irradiated (or not) with 5 J/m2 UV, dissociated with trypsin 5 days later, and single cells were plated on feeders at various densities. After 6 days, cells were fixed, permeabilized with 0.1% Triton X-100, stained with Hoechst 33258, and analyzed with an INCellAnalyzer scanning microscope (GE Healthcare, Hatfield, UK, www3.gehealthcare.co.uk). Colony irradiation: cells were dissociated with Accutase (Millipore, Watford, UK, www.millipore.com), plated as single cells either on MEF feeders with DMEM, 20% serum, and 4 ng/ml bFGF, or on Matrigel with mTeSR (Stemcells Technologies, Vancouver, Canada, www.stemcell.com), cultured until small colonies had formed, irradiated with 0, 5, or 10 J/m2 UV, fixed at various times after irradiation, and analyzed as above.
Cells were washed in phosphate buffered saline and irradiated with a 254 nm germicidal lamp at a fluency of 7 μW/cm2. For Cdc25A experiments, cells were preincubated 5 minutes with 25 μg/ml cycloheximide. For density labeling, cells were placed in medium containing 50 μM bromodeoxyuridine (BrdU), 50 μM deoxycytidine, and 1 μM 5-fluorouracil immediately after irradiation. Native- and hybrid-density DNA were separated by isopycnic centrifugation in cesium chloride for 14 hours at 37,000 rpm in a NVT65 rotor (110,000g).
Cells were irradiated with UV, harvested at various times after irradiation, stained with Annexin V antibody (BD Pharmingen, London, UK, www.bdbiosciences.com), propidium iodide, and Cy5-conjugated anti-SSEA3 antibodies (MC631), and analyzed by fluorescence-activated cell sorting (FACS) with CyAnADP O2 or MoFlo (Dako, Ely, UK, www.dako.com).
Cell Cycle Analysis
Cells were labeled with 10 μM BrdU (BD Pharmingen) after UV-irradiation, harvested at the indicated times, fixed, permeabilized, and stained with fluoresceine isothiocyanate (FITC)-conjugated anti-BrdU antibody and 7-AAD according to the manufacturer's instructions. Flow cytometry analyses were performed using FACSCalibur or FACSAria and FACSDiva or FlowJo software.
DNA was blotted on nitrocellulose membranes (1 μg for (6-4)PPs, 60 ng for CPDs). Lesions were detected with specific monoclonal antibodies (Cosmobio, Oxford, UK, www.cosmobio.co.jp) as previously described [26, 36]. Blots were then probed with 32P-labeled genomic DNA probes to control for DNA load. Chemiluminescence was quantified with a Fujifilm LAS-3000 Imager, 32P with a FLA-3000 Imager (Fujifilm, Bedford, UK. www.fujifilm.eu).
Cells were scraped in RIPA (250 mM NaCl, 2% Igepal, 0.5% sodium deoxycholate, 0.1% sodium dodecylsulfate, and 50 mM Tris-HCl pH 8.0). Lysates were passed five times through 20-gauge needles, incubated 20 minutes in ice, and cleared by centrifugation (13,000 rpm, 20 minutes). Protein content was determined by Bradford assay (Bio-Rad, Hemel, UK, www3.bio-rad.com). Samples of 20 μg proteins were run at 120–140 V on acrylamide/bisacrylamide (37.5:1) gels and transferred to nitrocellulose membranes (Bio-Rad) with a Criterion blotter (Bio-Rad) for 45 minutes at 100 V. Membranes were blocked with 2% bovine serum albumin (Sigma-Aldrich, London, UK, www.sigmaaldrich.com) or 5% skimmed milk in tris-buffered saline-Tween (TBS-T) (150 mM NaCl, 50 mM Tris-HCl pH 7.4, and 0.2% Tween 20 [Acros], London, UK, www.acros.com), incubated 2 hours at room temperature or overnight at 4°C with primary antibodies: Phospho-ATM S1981 (Epitomics, Burlingame, CA, www.epitomics.com; 1:5,000), ATM (Cell Signaling, Hitchin, UK, www.cellsignal.com; 1:1,000), Phospho-Chk1 S345 (Cell Signaling; 1:1,000), Chk1 (Cell Signaling; 1:1,000), Cdc25A (Santa Cruz, Bergheim, Germany, www.scbt.com; 1:200), Phospho-p53 S15 (Cell Signaling; 1:1,000), p53 (Cell Signaling; 1:1,000), p21 (Cell Signaling; 1:1,000), Beta-actin (Sigma Aldrich; 1:4,000), Caspase 3 (Abcam, Cambridge, UK, www.abcam.com; 1:500), DDB1 (Santa Cruz; 1:100), DDB2 (Santa Cruz; 1:200), and PCNA (Santa Cruz; 1:200). Blots were washed, incubated 1 hour at room temperature with peroxidase-coupled anti-rabbit (Abcam; 1:5,000) or anti-mouse (Amersham; 1:10,000) antibodies, washed again, treated with enhanced chemiluminescence (ECL)-reagent (Amersham, Hatfield, UK. www3.gehealthcare.co.uk), and imaged with a LAS-3000 Imager (Fujifilm).
As previously described in . Briefly, a fragment encompassing the potential mutation site was amplified with high-fidelity polymerase, purified on gel, and a fixed amount (600 ng) was digested with Tsp509I, which cuts at AATT sites. PCR primers for Sept4: 5′-TTGGTGCCTCTGTCCTCTCTC and 5′-AGGCAAGTCTGCCTGATTGTC, other primers were previously described in . Linkers with AATT cohesive ends were ligated, and two rounds of PCR amplification performed using one of the genomic primers and nested primers complementary to the linker. Products were analyzed on 12% acrylamide/bisacrylamide (29:1) gels stained with ethidium bromide.
High UV-Induced Apoptosis in hESCs with Rapid Proliferation of the Surviving Cells
Two low-passage number hESCs lines, H7-S14 (locally derived  from the H7 line ) and Shef5 , were irradiated with 5 J/m2 254 nm UV, a relatively mild dose inducing little apoptosis in other cell types, such as fibroblasts and lymphoblasts . By contrast, it caused massive apoptosis in hESCs, with less than 30% of the initial population remaining viable by 6 hours (Fig. 1A). Apoptosis was evident in both SSEA3+ and SSEA3− cells (Fig. 1B). Yet surviving cells rapidly resumed proliferation, and by 24 hours, the number of viable cells had markedly increased (Fig. 1, Supporting Information Fig. S1).
The high rate of apoptosis, followed with rapid resumption of proliferation, suggested that hESCs may not undergo cell cycle arrest after DNA damage. We therefore labeled newly synthesized DNA with BrdU at various times following UV irradiation and analyzed cell cycle profiles by double-staining for BrdU and DNA content. In both H7 and Shef5 cells, we observed a rapid decrease in both the fraction of cells incorporating BrdU and the level of BrdU incorporation in the S-phase cells after irradiation with 5 J/m2 UV (Fig. 2A). Contrarily to what is normally observed upon cell cycle arrest, cells in S-phase did not progress and accumulate in G2. In addition, BrdU labeling also revealed that, at all considered time points, some cells continued to enter S-phase, indicating that the G1/S checkpoint is not fully enforced in hESCs. Strikingly, the surviving cells fully recovered at later times, and 24 or 48 hours after irradiation, hESCs incorporated BrdU as vigorously as nonirradiated cells, even after 10 J/m2 UV (Fig. 2B, Supporting Information Fig. S2). Thus, hESCs show decreased incorporation of BrdU very soon after UV irradiation, followed by a progressive loss of S- and G2/M phase cells and a high level of apoptosis.
To investigate the relationship between the S-phase arrest and apoptosis, we pulse-labeled cells for 20 minutes with BrdU so as to follow them through the cell cycle and irradiated them at various stages, before appraising apoptosis by Caspase 3 cleavage. Although hESCs appear moderately more sensitive to UV during S-phase, apoptosis was induced in all phases of the cell cycle at similar rates (Supporting Information Fig. S3).
Defective DNA Damage Response Pathway and Late G1/S Checkpoint
We then examined the DNA damage response pathway to determine whether it could be responsible for impaired cell cycle arrest. UV-induced damage mostly activates the ATR-Chk1 pathway, although there can be a moderate activation of the ATM-Chk2 pathway, due to crosstalk between these pathways . Accordingly, we observed a rapid but transient phosphorylation of Chk1, beginning within 10 minutes of irradiation with a maximum at 45–60 minutes, together with a more modest phosphorylation of Ataxia Telangectasia Mutated (ATM), peaking at 60 minutes (Fig. 3A, 3B).
The S-phase checkpoint is triggered by phosphorylation of Cdc25A, causing its ubiquitination and rapid degradation. This prevents dephosphorylation of Cyclin-dependent kinase 2 (Cdk2) by Cdc25A and locks the Cdk2-cyclin E complex in an inactive state. In both hESCs lines, we observed rapid disappearance of Cdc25A within minutes of UV irradiation, although it was very transient and Cdc25A levels rapidly returned to the normal (Fig. 3C), making it difficult to appraise the contribution of Cdc25A to cell cycle checkpoints in hESCs.
The G1/S checkpoint involves activation of p53 via phosphorylation and its subsequent stabilization. Activated p53 induces transcription of p21/Waf1/Cip1, which inhibits Cdk2 and Cdk4. We observed p53 phosphorylation beginning within 1 hour of irradiation and peaking at 3 hours as well as a slight increase in total p53. Accordingly, quantitative RT-PCR revealed a robust induction of p21 mRNA in Shef5 cells, similar to that observed in our positive control, p21+/+ HCT116 cells (Supporting Information Fig. S4). However, p21 protein was undetectable by Western blotting in Shef5 cells (Fig. 3D), while barely visible and not inducible in H7 cells. This observation is in complete accordance with previous results obtained with different hESC lines [40, 41]. We thus conclude that the G1/S checkpoint is inefficient in hESCs, due to the absence of p21 protein, while a transient S-phase checkpoint is triggered early after irradiation consistent with the reduced incorporation of BrdU seen at that time.
Characterization of the Surviving Cells
We then focused on characterizing UV-resistant cells. A distinct possibility was that these cells might be differentiated or about to differentiate. However, we did not observe major changes in expression of cell surface markers such as SSEA3, SSEA4, TRA-1-60, or TRA-1-81 after irradiation (Supporting Information Fig. S5). Similarly, the surviving subpopulation did not markedly differ from nonirradiated cells in clonogenic potential (Supporting Information Fig. S6). Finally, UV-resistant hESCs were able to form embryoid bodies and differentiate into the three germ layers (Supporting Information Fig. S7). Thus, cells surviving UV-irradiation still present all the hallmarks of hESC, including pluripotency and capacity for self-renewal.
Yet the question remains, whether the surviving cells are phenotypically distinct from the original population or whether UV-survival is a stochastic event. To determine whether UV-resistance is a phenotypic characteristic of a subpopulation of hESCs, we attempted to select for this trait. H7 cells were irradiated with 5 J/m2 UV, allowed 1 week to recover and proliferate, and irradiated with the same dose again. Apoptosis was compared in the first and second batch of irradiated cells (Fig. 4A, Supporting Information Fig. S8). H7 cells were markedly more resistant to UV the second time around, suggesting that UV-resistance may indeed arise in a phenotypically distinct subpopulation. It is worth mentioning that, although the number of SSEA3− cells increased after the first irradiation (likely because SSEA3+ cells are more apoptotic than SSEA3− cells), both SSEA3+ and SSEA3− cells were more resistant to the second irradiation than to the first. Thus, loss of SSEA3 surface marker did not correlate with UV-resistance.
We then asked whether UV-resistance is present in a subpopulation of hESCs prior to irradiation or might be caused by selective UV-inactivation of a given subset of genes. We thus plated single cells and grew them to small colonies prior to irradiation. We reasoned that stochastic inactivation should result in a few UV-resistant cells in most colonies, whereas a pre-existing UV-resistance phenotype transferred to an entire colony should result in a few fully resistant colonies among a majority of fully apoptotic ones. Automated counting of cells and colonies revealed that colony size, rather than colony number, decreased after UV-irradiation (Fig. 4B). This is mostly compatible with the stochastic model, although it could be argued that a putative UV-resistant phenotype might be transient and lost rapidly during colony expansion in the absence of selective pressure.
NER Activity in Replicating hESCs
An obvious possibility was that DNA repair is more efficient in the UV-resistant population than in the cells that eventually undergo apoptosis, possibly because it occurred in the subset of transiently arrested cells. We thus appraised NER by measuring removal of the two main UV-induced lesions, CPDs and (6-4)PPs in Shef5 and H7 cells. In line with the notion that (6-4)PPs are among the best NER substrates, they were rapidly and efficiently repaired in both hESCs lines (Fig. 5A, 5B).
By contrast, because CPDs are poor NER substrates, their detection requires the specialized damaged DNA binding (DDB)  complex (dispensable for most other lesions), their repair is relatively slow, and generally not completed by 24 hours. We confirmed that the two main components of the DDB complex, DDB1 and DDB2/XPE, were expressed in hESCs (Supporting Information Fig. S9). Repair of CPDs in hESCs displayed considerable experimental variation in our hands, ranging from poor to decent. In general, repair was negligible at early time points (up to 6 hours), and analysis at 24 hours was complicated by the fact that, as hESCs rapidly resume proliferation, a significant amount of newly synthesized DNA dilutes the signal. To overcome this problem, we density-labeled cells with BrdU just after irradiation and separated parental DNA from replicated DNA by isopycnic centrifugation. In this way, we confirmed that repair of CPDs was reasonably proficient by 24 hours and, importantly, that NER efficiency was not higher in cells having divided than in those that did not divide (Fig. 5C).
Detection of Point Mutations in hESCs
While the observed rates of CPD repair would be considered low to borderline normal in other cells (e.g., fibroblasts or lymphoblasts), these undergo strong UV-induced cell cycle arrest and generally replicate little DNA by 24 hours [25, 39]. By contrast, we observed that hESCs resumed proliferation within 24 hours, when large amounts of CPDs lesions remain present, and density labeling experiments revealed that approximately 50% of total DNA had been replicated 24 hours after irradiation. UV-induced lesions constitute strong blocks for replication forks, but replication of damaged DNA is nevertheless possible by switching from replicative polymerases to specialized TLS polymerases. This switch is orchestrated by proliferating cells nuclear antigen (PCNA), which becomes monoubiquitinated upon DNA damage and triggers the recruitment of TLS polymerases . In both H7 and Shef5 cells, we observed PCNA ubiquitination as early as 1 hour after irradiation (Supporting Information Fig. S10).
TLS polymerases, lacking proof-reading activity, are low-fidelity enzymes and TLS is therefore an error-prone process . The main polymerase involved in bypassing UV-induced damage, Pol eta, strongly favors incorporating adenine opposite the 3′ nucleotide of a pyrimidine dimer . If the dimer lies in a TC or CC context, misincorporation results in a C to T transition, a pattern that is considered a signature UV-induced mutation. We recently developed a method to detect such mutations at very low frequencies, down to 5 × 10−6. It relies on the creation of a new Tsp509I restriction site by UV-induced mutation of an AATC sequence to AATT. A linker with AATT cohesive ends is ligated to the cut products, allowing selective amplification of mutated products with linker-specific PCR primers .
Using this technique, we observed that mutagenesis was elevated in both hESCs lines (Fig. 6), although not to the level observed in NER-deficient cells originating from a xeroderma pigmentosum patient (XP-E, GM01646) . In the silent myosin gene, mutation frequency was increased upon UV irradiation by a factor of 1.8 in H7 cells and 3.4 in Shef5. No such increase was observed in GM01953 control cells. In H7 cells, mutagenesis was increased 3.2-fold in the nontranscribed strand of the constitutively active gene DHFR and 1.8-fold in the transcribed strand of TP53. In BCL-6, a commonly mutated gene in B lymphoma, mutation frequency was increased 1.9-fold, whereas in SEPT4, a proapoptotic gene often mutated in leukemia, UV irradiation caused a 3.1-fold increase in mutation frequency.
Functional DNA damage response and repair are necessary to avoid cell death or mutagenesis. The two hESCs lines we tested, H7 and Shef5, proved highly apoptotic after irradiation with relatively mild (5 J/m2) doses of 254 nm UV light (Fig. 1). This behavior is not specific to UV, as we and others have observed high rates of apoptosis after exposure to replication inhibitors, ionizing radiation, and other DNA-damaging agents, in the same and other hESCs lines [18, 44]. It thus appears that high propensity to apoptosis is a common characteristic of hESCs, perhaps a strategy to ensure genomic integrity by favoring apoptosis over DNA repair.
This strategy, however, is not fully efficient, as we consistently observed that a small component of cells survived UV-irradiation and resumed vigorous proliferation within 24 hours. Chk1 was rapidly phosphorylated, although cell cycle analysis of UV-irradiated cells did not reveal a clear activation of DNA damage-induced checkpoints (Fig. 2). Instead there were heavy losses of both S- and G2/M-phase cells by 3 hours after irradiation. Nevertheless, by 24 hours, cell cycle profiles were restored to those found in unirradiated cells (Fig. 2). We and others also observed a deficient G1/S checkpoint after ionizing radiation . In the latter case, cells accumulated in G2 due to strong activation of the G2/M checkpoint by ionizing radiation. However, the G2/M checkpoint is only activated by very high UV doses (>20 J/m2)  and it is not surprising that the dose we used (5 J/m2) did not activate it. Conversely, the replication-dependent S-phase checkpoint is normally responsive to UV doses as low as 1 J/m2. The molecular steps leading to this activation are known to differ from those triggered by ionizing radiation but are unfortunately poorly understood. Possible contributions by the Dbf4-Cdc7 or the p38γ MAP kinase have been suggested but are largely based on indirect evidence . We were thus unable to identify the precise causes underlying the poor activation of the S-phase checkpoint in hESCs.
Activation of the late G1/S checkpoint through transcriptional induction of p21 by p53 was defective, as p21 protein did not increase after UV in either cell line, despite proper transcriptional induction (Supporting Information Fig. S4). We did not attempt to determine the cause of this discrepancy between mRNA and protein levels, but it could be explained by the well-known phenomena of p21 mRNA sequestration in cytoplasmic stress granules , binding of a translation-inhibiting factor to its 5-UTR , or very efficient degradation of the p21 protein . The stability of p21 is controlled by ubiquitination and requires its association with CDK/cyclin complexes . A short G1 phase and unphased expression of CDK/cyclins in hESCs might thus account for a quick degradation of p21, allowing rapid cycling of hESCs . Indeed, p21 is generally well expressed in G0-arrested cells and required to keep stem cells in quiescence, while downregulation of p21 is associated with cell cycle entry . Induction of p21 in response to DNA damage, such as UV-irradiation, ensures G1 arrest and protects cells from apoptosis . Significantly, p21−/− MEFs fail to properly arrest in G1 in response to DNA damage [53, 54], a phenotype similar to what we observed in hESCs.
Interestingly, cells surviving UV irradiation retained a typical hESC phenotype, with little change in clonogenic potential (Supporting Information Fig. S6) and ability to differentiate (Supporting Information Fig. S7). We did not observe major changes in cell surface markers after irradiation (Supporting Information Fig. S5), although SSEA3− cells consistently proved more resistant to UV than SSEA3+ cells. This often led to a slight enrichment of the surviving population in SSEA3− cells, although these always remained a minority. In addition, when we reirradiated cells having survived a prior irradiation, we found that both SSEA3− and SSEA3+ cells had become more resistant to UV-induced apoptosis (Fig. 4A). Thus, UV resistance, while a selectable trait, does not correlate with the degree of differentiation appraised by SSEA3 expression.
UV-resistance does not appear to be a pre-existing characteristic of a subpopulation of hESCs, as single-cell plating did not give rise to UV-resistant clones (Fig. 4B). It thus seems that UV-survival is a stochastic phenomenon, perhaps depending on which genes are inactivated by UV in a given cell or at which phase of the cell cycle a given cell is at the time of irradiation. The latter possibility is unlikely, however, as we did not observe major differences in caspase 3 cleavage between hESCs that were irradiated at different phases of the cell cycle (Supporting Information Fig. S3). Thus, as far as we can tell, UV-resistant hESCs do not differ from the original population, apart for their lower propensity to apoptosis. In particular, we verified that repair of UV-induced lesions was not more efficient in the surviving cells (Fig. 5C).
Repair of UV-induced lesions by NER in hESCs was characterized by efficient removal of (6-4)PPs and low, borderline normal removal of CPDs (Fig. 5). This is probably indicative of a slightly reduced NER efficiency, affecting the repair rate of the poorer substrates only, a phenomenon we have previously observed in quiescent cells .
Our results contrast with prior observations on mouse ESC (mESC), which were considered deficient in the repair of UV-induced lesions . This discrepancy may be due to species-specific differences, as mESCs were also found deficient in repair of double-strand breaks by nonhomologous end joining (NHEJ), whereas the hESC lines H1, H9, and CA1 were NHEJ-proficient . H9 and BG01 hESCs were also proficient in homologous recombination, appraised by appearance of Rad51 foci and disappearance of γH2AX foci . Finally, hESCs were proficient in base excision repair, displaying OGG1 activity levels similar to that of HEK293 cells, and induction of OGG1 and APE1 by oxidative damage greater than in WI-38 fibroblasts . Several groups reported increased expression of genes involved in various DNA repair pathways [58, 59]; however, such increases in mRNA levels do not necessarily translate into elevated protein levels or enzymatic activity. For instance, levels of DNA-PKcs and Ku70 were normal in hESCs  and so was OGG1 activity . Transcriptional induction may thus represent a feedback mechanism, in which expression of rate-limiting enzymes is increased in compensation for decreased efficiency of the pathway, a situation we previously encountered in a different cell type .
The reduction in NER efficiency we observed in hESCs may have adverse consequences, as these cells resume rapid replication while a significant amount of DNA lesions remains in their genome, and damage bypass by TLS polymerases is known to result in high mutagenesis . Indeed, we observed a twofold to threefold increase in mutation frequencies upon UV irradiation in hESCs. While these levels did not quite reach those observed in NER-deficient lymphoblasts (threefold to fivefold), they are clearly higher than in wild-type GM1953 lymphoblasts, where we did not detect any significant increase in mutagenesis after UV irradiation (Fig. 6) .
Such an increase in mutation frequency might be explained by the observed deficiency in p21 accumulation, as p21 might affect mutagenesis at several levels. First, the p21-dependent G1/S checkpoint allows time for cells to repair DNA before replicating it. Consequently, hESCs with low p21 have less time for repair, predisposing them to accumulate mutations and/or undergo massive apoptosis. Second, p21-deficient MEFs show a modest but statistically significant decrease in the repair of (6-4)PPs, demonstrating a stimulatory effect of p21 on NER efficiency . However, we did not observe such an effect in hESCs, which repaired (6-4)PPs very efficiently. Finally, p21 is postulated to be a negative regulator of TLS, by competing with, and impairing binding of polymerase eta to PCNA . Thus, cells with low amounts of p21 are more likely to resort to error-prone TLS polymerases to bypass DNA damage. Indeed, p21-deficient cells were shown to undergo increased TLS-associated mutagenesis , whereas p21−/− HCT116 cancer cells were twofold to threefold more UV-sensitive than their p21+/+ counterparts in clonogenic survival assays and displayed a twofold to threefold increase in frequency of spontaneously arising mutations in a 6-TG resistance assay .
Our results are important to consider in the perspective of ES-based replacement therapy. Obviously, cultured hESCs are unlikely to be exposed to UV, but we only used this agent as a convenient model, and NER repairs many other types of lesions, some of which hESCs are more likely to encounter. These include for instance oxidative damage caused by high oxygen tension in usual culture conditions (compared to intrauterine conditions), atmospheric pollutants, contaminants in culture media, or spontaneous endogenous damage . Their impact on mutagenesis will depend on their rate of repair, which is generally intermediary between that of (6-4)PPs and CPDs. Lesions remaining unrepaired when replication occurs will lead to increased mutagenesis, although we could not verify it, since they do not have specific mutational signatures. Our observation that cells surviving DNA damage become less prone to apoptosis (Fig. 4A) implies a positive selection for these and a progressive enrichment of the culture in damage-resistant cells that have accumulated point mutations.
Increased mutagenesis in silent genes (e.g., Myosin) may become an issue when hESCs are differentiated for therapeutic purposes. More disquietingly, we observed increased mutation frequencies in several cancer-related genes. BCL-6, a transcriptional repressor of B cell differentiation, is one of the most frequently mutated genes in B lymphoma . Point mutations in BCL-6 mostly occur at a hotspot in the 3′-UTR , the region we examined and found mutated. SEPT4 is a proapoptotic gene, encoding an inhibitor of antiapoptotic factors. Its expression is reduced in several types of human leukemia, and sept4−/− mice are leukemia-prone and display abnormally high proportions of hematopoietic stem cells in the bone marrow . Thus, elevated mutagenesis in hESCs may increase their risk of carcinogenic transformation by two mechanisms: activating an oncogene (e.g., BCL-6) or inactivating a tumor suppressor (e.g., SEPT4). When replacement therapy reaches the clinics, it will therefore be important to minimize DNA damage in hESCs, for instance by minimizing the time they are kept in culture prior to transplantation.
However, there is increasing evidence for the existence of a small population of cancer stem cells (CSCs) at the heart of many tumors, which seems to correlate with the aggressiveness of the cancer and are thought to be responsible for metastases  and resistance to therapy . Specific targeting of CSCs has become an important goal for cancer therapy, but their paucity makes them difficult to isolate and study. By contrast, ESCs are available in reasonably large amounts and may thus constitute a good substitute model for CSCs, especially for germ cell tumors. Even though ESCs and CSCs differ in their microenvironment and likely in the relative time they spend in quiescence, they share important homologies in patterns of gene expression, display the same key capacities for self-renewal and pluripotency, and (as we demonstrate here) a similar propensity for genomic instability .
Our results document, for the first time to our knowledge, an accumulation of damage-induced point mutations in cultured hESCs. The elevated mutation frequency we observed in several genes likely results from defects in the DNA damage signaling pathway, leading to incomplete arrest at cell cycle checkpoints. Combined with the rapid rate of proliferation of hESCs, this phenomenon severely limits the time available for DNA repair prior to replication. Under these conditions, the relatively modest NER capacity of hESCs reveals insufficient to maintain genomic integrity.
We thank Sue Newton and Kay Hopkinson for expert assistance with FACS analyses, Bernard Corfe for p21 PCR primers and control, and Sheila Blizzard for qRT-PCR analyses. This work was supported by grants from Yorkshire Cancer Research (PP005 to T.N., and S298 to M.M. and P.W.A.), the U.K. Medical Research Council (G0700785 to P.W.A.), and a fellowship from the Fonds de recherches du Québec—Santé (to J.D.).
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.