A very small number of embryonic stem (ES) cells gives rise to all tissues of the embryo proper. This means that ES cells should be equipped with highly efficient mechanisms to defend themselves against various stresses and to prevent or repair DNA damage. One of these mechanisms is a high activity of a verapamil-sensitive multidrug efflux pump. Because reactive oxygen species are a major source of DNA damage, we further tested the idea that murine ES cells might differ from their more differentiated counterparts by high levels of antioxidant defense and good DNA strand break repair capacity. This was confirmed by comparing cellular peroxide levels, total antioxidant capacity, and activity of radiation-induced strand break repair between murine ES cells and embryoid bodies or embryonic fibroblasts. Using microarrays and confirmation by reverse transcription–polymerase chain reaction, we identified several candidate antioxidant and stress-resistance genes that become downregulated during differentiation of ES cells into embryoid bodies.
Murine embryonic stem (mES) cells are pluripotent cells derived from the inner cell mass of blastocysts. There is only a very small number of ES cells that give rise to all tissues of the embryo proper. This small size of the population means that ES cells should be equipped with highly efficient mechanisms to prevent DNA damage, to repair it when it happens, and to counteract any propagation of mutations arising from it. Apoptosis is one efficient way to prevent proliferation of mutations, and, in fact, mES cells are known to be much more proficient in apoptosis than differentiated cells after UV-induced DNA damage [1–3]. Ionizing irradiation seems to be a less potent trigger of apoptosis in mES cells . Not much is known about the quality of primary stress defenses, specifically antioxidant defenses in mES cells, so far.
Reactive oxygen species (ROS) are a major source of DNA damage because they are continuously produced as a byproduct of normal metabolism. Most differentiated cells are sensitive to an increase in ambient oxygen pressure, because this accelerates the production of ROS from the mitochondrial respiration chain. For instance, by increasing the ambient oxygen partial pressure from 21%–40%, levels of intracellular peroxides  and protein carbonyls  increased, and lipofuscin accumulated at a faster rate in normal human fibroblasts [5,6]. At the same time, the replicative lifespan decreased to very few population doublings (PDs), and the rate of telomere shortening increased 4- to 10-fold . Differentiated murine somatic cells are even more sensitive to oxidative stress than human ones . As a result, murine fibroblasts experience replicative senescence early under standard in vitro conditions despite the fact that their telomeres are protected by being much longer than in human cells and by expression of telomerase. If oxidative stress is minimized, these cells can grow without obvious limits, similar to telomerase-expressing human fibroblasts under standard conditions . However, immortality is not equivalent to genomic stability. In fact, mutation frequencies and genomic instability in differentiated murine cells increase with time and with increasing oxidative stress [9,10].
We speculated that mES cells should have more efficient stress defenses than differentiated ones to maintain high levels of genomic stability at the single-cell level. We believe that insight into how ES cells are able to maintain the stability of their genomes over many cell divisions will help to understand aging of somatic stem cells in vivo and its impact on age-related disease in the adult organism. mES cells can be grown in culture without apparent limit  on gelatin-coated culture flasks in media containing leukemia inhibitory factor (LIF). Growth in media from which LIF is absent permits differentiation into cell types that have lost the characteristic of infinite growth, form embryoid bodies (EBs), and will eventually senesce. We and others have used this system to demonstrate that downregulation of telomerase reverse transcriptase (Tert), the catalytic subunit of telomerase, is one of the key events occurring early in ES cell differentiation . We now analyze the resistance of mES cells to oxidative stress and their capacity to repair γ-induced DNA strand breaks. mES cells are highly proficient in antioxidant defense, an ability that decreases during early steps of differentiation into EBs. ROS levels also remain high in mouse embryonic fibroblasts and 3T3 fibroblasts. Furthermore, mES cells perform more efficient DNA strand break repair than differentiated mouse 3T3 fibroblasts. Using microarrays and confirmation by reverse transcription–polymerase chain reaction (RT-PCR), we have identified several candidate antioxidant and stress-resistance genes that become downregulated during differentiation of ES cells into EBs.
Materials and Methods
The mES cells (CGR8) were routinely passaged and maintained in an undifferentiated state in a suitable medium (GMEM) with 0.25% NaHCO3, nonessential amino acids, 2 mM L-glutamine and 0.1 mM pyruvate, 0.1 mM 2-mercaptoethanol, and 10% fetal calf serum with LIF. To form EBs, ES cells were cultured in non–gelatin-coated flasks in the presence of LIF for 48 hours in a humidified 5% CO2 atmosphere. ES cell aggregates were harvested into a petri dish (103 aggregates per 10 ml ES cell culture medium lacking LIF) and allowed to differentiate for up to 6 days. Medium was replaced every 2 days. Hematopoietic colony-forming unit–granulocyte-erythroid-macrophage-megakaryocyte (CFU-GEMM) assays were performed each day to monitor differentiation of ES cells. Cells were subjected to chronically increased oxidative stress by culture under 40% normobaric hyperoxia in a three-gas incubator (Binder, U.K.) for 2–3 weeks. All medium conditions were identical to normoxic cell culture.
CFU-GEMM assays were performed according to manufacturer's instructions (Stem Cell Technologies, Vancouver, Canada). In brief, approximately 150 EBs in a total volume of 0.3 ml were added to 3 ml of Methocult medium, yielding triplicate cultures of 1.1 ml each. The concentration of cytokines in Methocult was as follows: interleukin IL-3, 10 ng/ml; IL-6, 10 ng/ml; stem cell factor, 50 ng/ml; and erythropoietin, 3 U/ml. Methocult-containing EBs were dispensed into three 35-mm dishes using a 16-g blunt-end needle. Two plates were then placed into a 100-mm covered dish containing a third uncovered dish containing 3 ml of sterile water. Cultures were then placed in a 37°C incubator maintained with 5% CO2 and >95% humidity. After 10–12 days in culture, the plates were assayed by counting the number of EBs surrounded by a corona or burst of differentiated cells. Such bursts represent hematopoietic colony-forming units, and the percentage of these increased markedly on day 4 of the differentiation protocol described above, which is indicative of the differentiation of the ES cells.
Cellular Peroxide Concentrations
The acetate groups of the nonfluorescent substrate 2′,7-dichloro-3′,5-dihydrofluorescein diacetate (DCF-DA) are cleaved in vivo by indigenous esterases, and subsequent oxidation by intracellular peroxides produces the fluorescent 2′,7-dichlorofluorescin, which can be detected by flow cytometry (FL1 detector). A stock solution of DCF-DA (Sigma, St. Louis) was prepared fresh for every experiment as a 2-mM solution in absolute ethanol. EBs at different time points (days 0, 2, 4, and 6) were disintegrated to a suspension of single cells using Trypsin (0.05%) EDTA (0.53 mM) for 10 minutes. Stem cells, mouse embryonal fibroblasts, and 3T3 cells were trypsinized only briefly to detach the cells from the plastic surface. ES and EB cells were collected in serum-free medium containing 80 μM DCF-DA and 5 μM verapamil (Sigma) for 30 minutes at 37°C. Verapamil at this concentration fully blocked the efflux of DCF from cells for at least 2 hours (Fig. 2B, C). Afterwards, the staining medium was removed and the cells were pelleted and resuspended in full medium plus 5 μM verapamil. Cells were analyzed within 1 hour in a flow cytometer (Partec, Münster, Germany) using blue excitation and the green emission channel (FL1). System alignment and FL1 gain setting were adjusted using fluorescent beads. Cells were gated in forward scatter/side scatter (FSC/SSC), and the median of the gated FL1 fluorescence peak was used as an estimate of the peroxide concentration.
DNA Damage Assay (Fluorescence-Detected Alkaline DNA Unwinding)
DNA damage was assessed using a recently developed semi-automated, fluorescence-detected alkaline DNA unwinding assay , modified from an assay previously described by Birnboim and Jevcak . The method detects single- and double-strand breaks by their effect on the rate of DNA denaturation in alkali, monitored by the fluorescence intensity of an intercalating dye. Cells were γ-irradiated using a Cs137 source as a cell suspension with a constant density of 106 cells in 200 μl of serum-containing medium. After irradiation, cells were divided in two identical aliquots; one was kept on ice for the measurement of initial damage, whereas the other was incubated at 37°C for 1 hour to allow DNA strand break repair to take place. Afterwards, cells from each sample were lysed in 0.25 M mesoinositol, 1 mM MgCl2, and 10 mM Na2PO4/NaH2PO4, pH 7.2, followed by transfer of 70 μ (7 × 104 cells) onto a 96-well plate and addition of 70- μl denaturation buffer (9 M urea, 10 mM NaOH, 25 mM CDTA [trans-1,2-diaminocyclohexane-N, N, N′N′-tetraacetic acid], 0.1% SDS) and 70-μl alkali solution (200 mM NaOH, 40% denaturation buffer). All additions were performed at 4°C, and each sample was analyzed in 8 to 10 replicates. Partial alkaline DNA unwinding, starting from DNA ends and strand breaks, was activated under controlled alkaline conditions for 40 minutes at 30°C and stopped by the addition of 140 μl neutralization solution (1 M glucose, 15 mM β-mercaptoethanol). Remaining double-stranded DNA was then stained by adding 470 μl of the intercalating dye SYBR Green (diluted 1:25,000 in 13 mM NaOH) per well and 7-minute incubation at ambient temperature. Fluorescence was measured in a fluorescence reader Spectrafluor Plus (Tecan, Crailsheim, Germany) at λex 492 nm and λem 520 nm. The fluorescence intensity is inversely correlated to the number of DNA strand breaks present at the time of lysis. DNA damage percentage was calculated as 100 × (P0 − PX)/P0, where P0 is fluorescence intensity of an unirradiated sample and PX is fluorescence intensity of the irradiated cell sample.
Total Antioxidant Capacity
The ability of cell lysates to inhibit metmyoglobin/H2O2-mediated free radical production was measured using a Total Antioxidant Status (TAS) kit (Randox, Crumlin, U.K.). Cells were washed in phosphate-buffered saline, counted, and stored at −80°C. A total of 2 × 106 cells were lysed by shearing for 10 minutes on ice in 60 μl 50 mM NaPO4 buffer containing 1.8 mM phenylmethylsulfonyl fluoride, 14 μl/ml aprotenin (Sigma), and 3 mg/ml diethylenetriamine-pentaacetic acid anhydride. Lysates were cleared by centrifugation, and protein concentration was determined using the Bradford assay (Bio-Rad, Hercules, CA). Measurements were performed at 600 nm according to the instructions of the manufacturer.
Identification of Candidate Genes Involved in ES Cell Resistance and Repair
Total RNA was extracted from ES cells and EBs at day 4 using Trizol (Invitrogen, Paisley, U.K.), and 10 μg of the RNA sample was used to make biotin-labeled cRNA (copy RNA) according to the manufacturer's instructions. After appropriate cleanup of the cRNA (IVT cRNA cleanup spin column as supplied with Affymetrix gene chips), biotin-labeled cRNA was quantified and fragmented by metal-induced hydrolysis according to the manufacturer's protocol (Affymetrix, Buckingham, U.K.). This converts full-length cRNA into fragments of 35–200 bp in length, which are suitable for hybridization. Biotin-labeled cRNA produced in this manner was hybridized to the mouse 430 target array and washed according to the manufacturer's protocol. The data generated from these were interrogated using the Gene Ontogeny mining tool (Affymetrix), which allowed us to identify 15 genes that have functions implicated in stress response. The changes in expression levels of these genes were confirmed by RT-PCR. Approximately 500 EBs were selected each day, and total RNA was extracted using RNA-zol (Biotecx Lab, Houston, TX) and reverse transcribed using AMV reverse transcriptase (Promega, Southamptom, U.K.) and random oligo hexamers following the manufacturer's instructions. The amount of cDNA in each sample was normalized using Gapdh as control. PCR primers and conditions used are shown in Table 1. PCR products were run on 2% agarose gels and stained with ethidium bromide. RNA controls were included to monitor genomic contamination.
Table Table 1.. Primer sets used to amplify candidate genes by reverse transcription–polymerase chain reaction
Annealing temperature (°C)
Product size (bp)
Murine embryonic fibroblasts (MEFs) are sensitive to ambient oxygen. Although they are able to grow essentially unrestricted in culture under physiological ambient oxygen concentration (3%), they enter a telomere length–independent senescence-like state after 10 to 20 PDs under 21% oxygen . An additional increase of the ambient oxygen partial pressure to 40% arrested MEF growth after less than 3 PDs (Fig. 1). Prolonged culture of MEFs under 21% oxygen, for instance in a 3T3 protocol, leads to spontaneous immortalization. This is accompanied by major genomic changes. Such immortalized 3T3 murine fibroblasts died by apoptosis within a few days under 40% hyperoxia (data not shown). In marked contrast, mES cells did grow fast and without limits under 21% oxygen. They even proliferated with nearly the same rate under 40% hyperoxia, and this ability to withstand chronically increased oxidative stress did not decline with age in culture (Fig. 1, data not shown). Frequencies of apoptosis did not increase when mES cells were grown under 40% oxygen (data not shown).
These data suggested that mES cells might have especially high antioxidant defense capacities, which might be lost during differentiation. To test this suggestion, mES cells were differentiated into EBs. Differentiation toward hematopoietic lineages was determined by progenitor colony assays (CFU-GEMM), in which different cytokines and growth factors are used to promote differentiation of early hematopoietic stem/progenitor cells from ES cells. The percentage of hematopoietic CFUs increased markedly on day 4 of the differentiation protocol (24.3%), and 46.2% of EBs were committed toward the hematopoietic lineage at day 6 (Fig. 2A). To investigate whether mES cells have a better capability to maintain low intracellular levels of oxygen free radicals, we stained ES cells, EB cells, and differentiated murine cells with DCF-DA. DCF-DA is trapped within cells by esterification and converted in its fluorescent form by reaction with peroxides. DCF is a substrate for P-glycoprotein, the multidrug efflux pump that is known to be active in a wide variety of committed stem cells . In mES cells, this activity leads to a loss of DCF fluorescence, which is nearly complete within 2 hours. However, this loss is completely abrogated by verapamil, a known inhibitor of P-glycoprotein (Fig. 2B). As Figure 2C shows, the activity of this verapamil-sensitive efflux pump decreases with differentiation and is essentially lost at day 6 of the differentiation protocol. To measure the steady-state levels of peroxides irrespective of P-glycoprotein activity, 5 μM verapamil was used in all subsequent DCF staining experiments. A significant increase of cellular DCF fluorescence with time in the differentiation protocol was found. Differentiation into EBs (day 6) resulted in an approximately fourfold increase in cellular DCF fluorescence, which was similar to values found in 3T3 murine fibroblasts and MEFs (Fig. 2D). DCF fluorescence of mES cells did not increase under hyperoxic culture conditions, whereas that of differentiated cells did (data not shown).
To establish whether the low peroxide content in mES cells was attributable to their improved antioxidant capacity, we measured the capability of ES and EB cell lysates to delay metmyoglobin/H2O2-mediated free radical generation (TAS). As Figure 3 shows, lysates from ES cells are in fact more proficient in antioxidant defense than those from EB cells.
Next, we examined whether the improved stress protection of ES cells would also include their DNA damage repair capabilities. As expected, initial DNA damage induced by ionizing radiation did not differ between ES cells and differentiated immortal mouse 3T3 fibroblasts with the sole exception of the highest radiation dose (Fig. 4). The capacity of 3T3 cells to repair DNA strand breaks decreased drastically with increasing radiation dose. In contrast, ES cells performed efficient DNA repair even after doses as high as 20 Gy (Fig. 4).
To identify candidate genes that might be involved in the differential regulation of stress defense and repair, we performed a comparative gene expression analysis of ES and day-4 EB cells using the Affymetrix Mouse 430A target array, which analyzes the expression level of approximately 14,000 well-characterized mouse genes. Eight hundred twenty-one genes were identified using twofold change in expression between the samples as cut-off criterion. Affymetrix Gene Ontology Mining tool was used to select 15 candidate genes already known to be involved in cellular response to stress and DNA repair (Table 2). Differential expression was confirmed for all of them by semiquantitative RT-PCR (Fig. 5). With the exception of only two genes (Prdx2 and Hif3a), all candidate genes were downregulated to varying degrees upon differentiation of ES cells to the day-4 EB stage. Seven of the identified genes are known to be directly involved in the regulation of the cellular redox state. In addition, four heat shock genes and one DNA repair gene were found to be downregulated during differentiation. This suggests overall a decreasing antioxidant and stress defense capacity with ongoing differentiation in accordance with the functional data shown above.
Table Table 2.. Genes whose expression levels change significantly upon differentiation of embryonic stem cells to 4-day embryoid bodies with particular relevance to oxidative stress resistance
aA total of three one-sided p values for each probe set is calculated using Wilcoxon's signed rank test, and these are combined to give the final p value. The p value ranges from 0.0 to 1.0. Values close to 0.0 indicate the likelihood for an increase in transcript expression level in the experiment array (day-4 embryoid bodies) compared with the baseline (embryonic stem cells), whereas values close to 1.0 indicate the likelihood for a decrease in transcript expression level between the two samples.
Mouse heat shock protein 1a (mouse heat shock protein 68)
Heat shock protein 1b (structure is similar to the DNA-dependent kinase molecular chaperone Hsp70)
Heat shock protein 9a (mortalin)
B-lymphoma insertion region 1
Excision repair cross-complementing repair deficiency protein 4 (xeroderma pig-mentosum complementation group F protein)
In this study, we have determined that mES cells are remarkably resistant to potential genotoxic stress. Our data suggest that multiple pathways may contribute to this resistance, including the activity of a verapamil-sensitive multidrug efflux pump, antioxidant defense, DNA strand break repair, and heat shock protein expression. Interestingly, antioxidant capacity and the ability to maintain low peroxide levels decrease very early during differentiation, and this is mirrored by significantly decreased expression levels of redox and chaperone genes. Thus, it appears that very high stress resistance is specifically maintained in pluripotent stem cells.
Specifically, the glutathione/thioredoxin system not only forms an important defense against the accumulation of ROS, but glutathione is considered to be the major thiol-disulphide redox buffer of the cell . This is important because the tertiary structure of many proteins relies on maintaining the equilibrium between reduced thiol groups and oxidized disulphides. Glutathione contributes to this process by reducing cellular disulphides and is thus itself oxidized to glutathione disulphide. Glutathione is regenerated by the enzyme complex Tgr (thioredoxin-glutathione reductase)  at the expense of NADPH, and it is this requirement that makes the system dependent on the pentose phosphate pathway. Our study indicated that Tgr undergoes sizable downregulation upon differentiation, in common with several other enzymes involved in the reaction of glutathione with ROS such as glutathione peroxidases 2, 3, and 4 (Gpx2/3/4)  and glutathione-S-transferase (Gsta3) , which also has a role in the removal of xenobiotic materials and repair of damaged macromolecular cell components . A similar activity decrease is demonstrated for SOD2 (Mn-dependent superoxide dismutase), which, in view of its mitochondrial location, is generally regarded as being the primary defense of the cell against ROS .
The large downregulation of Pdh2 expression, a component of the pyruvate dehydrogenase complex that is responsible for the conversion of pyruvate into acetyl coenzyme A [22,23], could imply that the pyruvate dehydrogenase complex is less active in differentiated cells in accordance with the work of Houghton et al. , who showed that during development to the blastocyst stage of mouse embryos, there is a significant shift from dependence on oxidative phosphorylation to ATP generation by glycolysis. This correlates with the decreasing oxygen concentration encountered by the embryo in its descent through the oviduct and implantation in utero and with the observed upregulation of hypoxia-inducible factor alpha 3 subunit (Hif3a), which is part of a transcription factor complex that activates genes that mediate adaptive responses to reduced oxygen availability [25,26]. However, it must be stressed that increased reliance on glycolysis alone cannot explain the decrease of glutathione-reducing activity because it is not known whether a greater or lesser proportion of glucose-6 phosphate will enter the pentose phosphate pathway and thus alter NADPH concentration.
This study has identified four heat shock proteins that undergo downregulation upon differentiation of murine ES cells. Hspb1  is known to positively regulate the expression level of SOD2 ; thus it is possible that the decrease in Hspb1 is a causative factor behind the downregulation of SOD2. Hspa1b is known to function in the resistance of cells to apoptosis , so its downregulation is expected to accompany reduced resistance to apoptosis in differentiated cells. The observed downregulation of Hspa1a (murine heat shock protein 68) is consistent with its known downregulation at the cell cleavage stage of preimplantation mouse embryos , when it seems to give way to other heat shock proteins that possess heat shock elements in their promoter sequences, making their expressions inducible in the event of cellular stress. Hspa9a, also known as Mortalin, has been implicated in a range of functions, including stress response [31,32], control of cell proliferation , and cell differentiation . Overexpression of Mortalin inactivates p53 , confers a survival advantage to cells , and seems to assist hTERT in cell immortalization .
The polycomb group repressor Bmi1  is a transcriptional regulator that has been implicated in the self-renewal of hematopoietic and neural stem cells . It is a repressor of checkpoint genes p16ink4a and p19Arf and of Hox group genes and a positive regulator of telomerase activity. Its downregulation upon ES cell differentiation might be a prerequisite for the initiation of senescence mechanisms that are characteristic for somatic cells.
ERCC4/XPF complexes with ERCC1 to form the DNA repair endonuclease responsible for the 5′ incision during DNA nucleotide excision repair . It is also involved in interstrand crosslink repair and in homologous recombination. Homologous recombination has been shown to be the prominent pathway for repair of DNA double-strand breaks induced by ionizing radiation . Homologous and nonhomologous recombination differentially affect DNA damage repair in mice . Although the human XPF phenotype is comparatively mild, inactivation of the XPF gene leads to severe growth retardation and early death in mice, suggesting a major importance of this complex in different DNA repair pathways in mice .
It is currently unknown whether the observed expression changes are causes or consequences of differentiation. It is also unclear whether the reduction of the glutathione/thioredoxin redox couple and the gene expression changes leading to decreased antioxidant defense capacity will prove to be a molecular switch common to all ES cell differentiation events. It is, however, tempting to speculate that a shift in the thiol-disulphide equilibrium maintained by the glutathione/thioredoxin redox couples during differentiation might serve to adjust basal ROS concentrations to a level appropriate to a role as second messengers in differentiation. Such a role has been well demonstrated . Moreover, it has been suggested by the evolutionary theory of aging  that the maintenance of a high level of defense and repair capabilities might come at a high cost. Thus, downregulation of maintenance as early as possible might be seen as a sound evolutionary strategy.
Work presented in this paper was supported by the Regional Developmental Agency (One North East), Life Knowledge Park, and a program grant (252) from Research into Ageing, U.K. We would like to thanks Ilka Wappler for performing the microarray analysis and Dr. Heiko Peters for useful discussions.
Gabriele Saretzki and Lyle Armstrong contributed equally to this work.