Main

Recent studies have demonstrated that it is possible to reprogram somatic cells into pluripotent stem cells by expressing a specific combination of transcription factors.1, 2, 3, 4, 5 The typical mix of transcription factors used for this purpose includes the oncogenes c-Myc and Klf4. Interestingly, oncogenes have the potential to induce genomic instability,6 which raises the possibility that induced pluripotent stem (iPS) cells have aberrant genomes.

Given the excitement in the stem cell field by the ability to transform adult differentiated cells into pluripotent stem cells, it is not surprising that research has focused mainly on methods that enhance reprogramming efficiency, with less attention been paid to the genomic status of the generated iPS cells.7, 8, 9 However, the presence of genomic aberrations in iPS cells could be one of the reasons why the efficiency with which these cells produce live mice in tetraploid complementation assays is very low.10

The possibility that aberrant genomes are prevalent in iPS cells became more likely after the demonstration that p53 inactivation facilitates reprogramming.11, 12, 13, 14 One study suggested that the absence of p53 enhances reprogramming, because it enhances cell proliferation.15 However, p53 is a DNA damage response gene,16 raising the possibility that reprogramming is accompanied by DNA damage and genomic instability. Consistent with this interpretation, iPS cells generated from mouse embryo fibroblasts (MEFs) lacking p53 function, form malignant tumors when injected in donor mice.17

In this study, we examined directly the genomes of induced stem cells by array-based comparative genomic hybridization (cGH) analysis. Using three different experimental systems, we report genomic aberrations in induced stem cells. These aberrations were associated with oncogene-induced DNA replication stress.

Results

Genomic instability in induced mammary stem cells

As a first step in exploring whether stem cell reprogramming is accompanied by genomic instability, we developed a protocol for inducing mammary stem cells. The protocol involves expressing c-Myc, one of the four original stem cell reprogramming genes, in mouse mammary cells and examining whether these cells acquire stem cell properties, such as the ability to form mammospheres in vitro and to repopulate cleared mouse fat pads.18

We first attempted to reprogram nearly homogeneous populations of mammary progenitor cells. These cells, chosen because they lack stem cell properties, were obtained using a previously described PKH26-based label-retaining protocol.19 Briefly, primary mammary cells were pulse-labeled with the lipophilic, fluorescent-dye PKH26 and then cultured as mammospheres for two passages. At this time, PKH26-high cells (about 0.3% of all cells, representing stem cells) and PKH26-negative cells (about 30% of all cells, representing progenitor cells) were isolated by flow sorting.19 The PKH26-negative cells were then infected with a control lentivirus or a lentivirus-expressing MycER, a c-Myc protein containing a modified estrogen receptor hormone-binding domain at its C-terminus.20 Five thousand infected cells were cultured under non-adherent conditions to generate mammospheres, which were then passaged on a weekly basis. The control-infected progenitor cells formed mammospheres with very low efficiency even at the first passage and both the mammosphere number and the cumulative cell number declined to practically zero within a few passages (Figures 1a–c). In contrast, the MycER-expressing progenitor cells were reprogrammed into mammary stem cells, as ascertained by their ability to form mammospheres and repopulate a cleared mouse fat pad (Figure 1).

Figure 1
figure 1

Reprogramming of progenitor mammary cells to stem cells by MycER. (a and b) PKH26-negative cells, which are devoid of stem cells, were isolated from secondary mammospheres and infected with a lentivirus-expressing MycER or with a control virus (Empty Vector). The cells were cultured under non-adherent conditions and mammosphere number (a) and cumulative cell number (b) were determined at each passage. Results are presented as means±1 S.D. (c) Representative images of the cultures at passages 1 and 2. (d) Repopulation of a cleared mouse fat pad by MycER-expressing mammospheres originally prepared from PKH26-negative cells with no evidence of tumor development

Full size image

Having established that c-Myc can reprogram mammary progenitor cells into stem cells, we repeated these experiments bypassing the PKH26-sorting step. Control virus-infected mammospheres prepared from wild-type mice could be maintained only for a few passages in tissue culture, suggesting exhaustion of the stem cell population. In contrast, mammospheres infected with the virus-expressing MycER could be easily expanded with no signs of crisis or stem cell exhaustion and could also repopulate a cleared mouse fat pad (Figure 2).

Figure 2
figure 2

Stem cell reprogramming of primary mammary cells by MycER. (a) Primary mammary cells prepared from wild-type or p53−/− mice were infected with a lentivirus-expressing MycER or with a control virus (Empty Vector) and cultured under non-adherent conditions. In some cultures TAM was added to induce high levels of MycER activity. The number of mammospheres and the cumulative number of cells were determined at each passage. Results are presented as means±1 S.D. (b) Repopulation of a cleared mouse fat pad by MycER-expressing mammospheres prepared by lentiviral infection of primary mammary cells. There are no signs of tumor development in the replanted mammary gland

Full size image

The reprogramming of mammary progenitor cells into mammary stem cells described above was performed in the absence of 4-hydroxytamoxifen (TAM). When TAM was added to the media, nuclear MycER protein levels increased (Supplementary Figure 1a) and the MycER-expressing mammosphere cultures were exhausted within one passage (Figure 2a). This was associated with phosphorylation and stabilization of p53, expression of p53 target genes and apoptosis (Supplementary Figure 1b), suggesting that high levels of MycER protein induced a p53-dependent DNA damage response.16, 21, 22, 23 Consistent with this scenario, MycER-expressing mammospheres prepared from p53−/− mice could be expanded in the presence of TAM (Figure 2a). During stem cell reprogramming, MycER levels were lower and apparently insufficient to induce an overt DNA damage response (Supplementary Figure 1b), but whether they were low enough to prevent induction of genomic instability was not clear.

To address the issue of genomic instability, mammary cells, reprogrammed into stem cells by infecting them with the lentivirus-expressing MycER, were passaged for 9 weeks in the absence of tamoxifen. Then, the cells were serially diluted in 48-well plates, to obtain single stem cell clones, which were expanded for 3–6 weeks, again in the absence of tamoxifen, before preparing genomic DNA (Figure 3a). Control genomic DNA was prepared from non-infected primary mammospheres (passage 5). Eight randomly selected reprogrammed stem cell clones were subjected to cGH analysis using high-density arrays covering chromosomes 10–13 in their entirety and part of chromosomes 9 and 14, corresponding in total to a quarter of the mouse genome.

Figure 3
figure 3

Genomic instability in MycER-induced mammary stem cells. (a) Protocol for isolation of genomic DNA from wild-type mammary cells and from MycER-reprogrammed stem cell clones. (b) cGH analysis of four MycER-reprogrammed stem cell clones. Selected regions of mouse chromosomes 9, 13 and 14 are indicated. The results are expressed as log2 ratios of the intensity signals of the stem cell clone DNA to the control DNA. Each black bar represents the average of 10 probes with 1388-bp median spacing. The red lines represent the statistical average and show discontinuities, when a series of averaged probe data deviates in a statistically significant manner from its neighbors. Transcripts are indicated by blue lines and the names of select genes are indicated

Full size image

Four out of the eight examined clones had focal copy number changes (CNCs) (Figure 3b). The first clone had a deletion of about 100 kbp within the retinoic acid receptor-related orphan receptor A (Rora) gene in chromosome 9. In humans, the RORA gene maps to the common fragile site (CFS) FRA15A.24, 25 The second reprogrammed clone had a deletion of about 250 kbp within the phosphodiesterase 4D (Pde4D) gene in chromosome 13. Like Rora, Pde4D is a very large gene. The third clone had a deletion of about 200 kbp within the protein tyrosine phosphatase receptor type G (Ptprg) gene in chromosome 14. In humans, PTPRG maps within the CFS FRA3B,25 although most deletions targeting FRA3B involve the adjacent fragile histidine triad (FHIT) gene. The fourth clone had an amplification of the jumonji AT rich interactive domain 2 (Jarid2) and dystrobrevin-binding protein 1 (Dtnbp1) genes in chromosome 13. The protein product of Jarid2 associates with the Polycomb repressive complex 2 and regulates the self-renewal of embryonic stem cells,26, 27, 28 suggesting that the amplification of Jarid2 may have been selected in this clone. The observed frequency of genomic aberrations within just a quarter of the mouse genome indicates that reacquisition of stemness features in the reprogramming protocol is associated with the occurrence of genomic rearrangements.

Genomic instability in iPS cells

The analysis of the genomes of induced mammary stem cells, described above, prompted us to explore the presence of genomic instability in iPS cells prepared by standard protocols.1, 2, 3, 4, 5 We first focused on iPS cells reprogrammed using three factors. Cells prepared from six independent clones (from two independent experiments, three clones per experiment) expressed pluripotency markers and could be induced to differentiate and form embryoid bodies in vitro (Supplementary Figure 2). Further, all these clones were diploid, as determined by karyotype analysis (data not shown). cGH was performed for chromosomes 5–13 in their entirety and for part of chromosomes 4 and 14, corresponding in total to half the mouse genome. Only two clones showed signs of genomic instability. One clone had a small gain within chromosome 8 targeting the odd Oz/ten-m homolog 3 (Odz3) gene; a second clone also had a small gain, again within chromosome 8, targeting the cadherin 13 (Cdh13) gene (Figure 4). Both Cdh13 and Odz3 are large genes.

Figure 4
figure 4

Genomic instability in iPS cells induced by OSK. (a) Protocol for isolation of genomic DNA from iPS cells and from non-infected MEFs. (b) cGH analysis of six iPS cell clones. Two regions of mouse chromosome 8 are indicated. The results are expressed as log2 ratios of the intensity signals of the iPS cell DNA to the control DNA. Each black bar represents a single probe with 1388-bp median spacing. The red lines represent the statistical average and show discontinuities, when a series of averaged probe data deviates in a statistically significant manner from its neighbors. Transcripts are indicated by blue lines and the names of select genes are indicated

Full size image

We subsequently examined iPS cells that had been reprogrammed using four factors. Pluripotency was assessed by expression of pluripotency markers (including the endogenous Oct4 locus and alkaline phosphatase), the ability to grow in the presence of ERK and GSK3 kinase inhibitors (the most stringent conditions for the propagation of embryonic stem cells) and the ability to differentiate into embryoid bodies (Figure 5). The karyotypes of cells from four independent clones were examined; two clones were diploid and two aneuploid (data not shown). Cells from the diploid clones were subjected to cGH for half the mouse genome, as described above for the OSK iPS cells (Figure 6a). Cells from the first clone had an amplification in chromosome 7 that involved >20 genes, as well as a deletion in chromosome 8 targeting the fat mass and obesity associated (Fto) gene (Figure 6b), a large gene. In contrast, we could not identify any focal CNCs in cells from the second clone.

Figure 5
figure 5

Characterization of iPS cells induced by four factors (OSKC). (a) Morphology of iPS-OSKC clones 1 and 2 by phase-contrast microscopy. (b) Staining of iPS-OSKC clones 1 and 2 for alkaline phosphatase activity. (c) Expression of pluripotency markers in iPS-OSKC cells. Expression levels of SSEA-1 and Oct4-GFP in iPS-OSK clone 1 and 2 cells and in the MEFs from which these iPS cells were derived, as determined by flow cytometry. The percentages of cells expressing high or low levels of SSEA-1 and Oct4-GFP are indicated by purple-colored numbers. (d) Expression of pluripotency genes in iPS-OSKC cells. Levels of Oct4, Zfp296, Eras and Fgf4 in iPS-OSKC clones 1 and 2, in embryonic stem (ES) cells and in MEFs were determined by PCR. Nat1 expression serves as a standard. (e) Morphology of embryoid bodies formed after differentiation of iPS-OSKC clones 1 and 2 for 7 days in vitro, as revealed by hematoxylin–eosin staining

Full size image
Figure 6
figure 6

Genomic instability in iPS cells induced by OSKC. (a) Protocol for isolation of genomic DNA from iPS cells and from non-infected MEFs. (b) cGH analysis of two iPS cell clones. The array hybridization for clone 1 was performed twice and both replicates are shown. Selected regions of mouse chromosomes 7 and 8 are indicated. The results are expressed as log2 ratios of the intensity signals of the iPS cell DNA to the control DNA. Each black bar represents the average of 10 probes with 1388-bp median spacing. The red lines represent the statistical average and show discontinuities, when a series of averaged probe data deviates in a statistically significant manner from its neighbors. Transcripts are indicated by blue lines and the names of select genes are indicated

Full size image

Genomic lesions induced by DNA replication stress

Given the conservation of CFS in the mouse and human genomes,29, 30, 31 it appears that some of the CNCs identified in the induced stem cells map to CFS. Specifically, the deletions in the Rora and Ptprg genes map to the human CFS FRA15A and FRA3B, respectively. Several of the remaining identified CNCs targeted large genes, again suggesting DNA replication stress as the culprit.30, 31 To explore the spectrum of CNCs generated by DNA replication stress, we treated cells for 4 weeks with low doses of aphidicolin, a prototypical agent for inducing DNA replication stress, and then monitored CNCs by cGH (Figure 7a). For this experiment, we employed a mouse–human hybrid cell line, GM11713A, which contains a single copy of human chromosome 3, thus, facilitating the detection of deletions. This cell line was previously used to map aphidicolin-induced deletions within the FHIT gene.32

Figure 7
figure 7

Genomic instability induced by chronic DNA replication stress. (a) Protocol for isolation of genomic DNA from aphidicolin-treated GM11713A cells and untreated cells. (b) cGH analysis of six aphidicolin-treated clones. Selected regions of human chromosome 3 are indicated. The results are expressed as log2 ratios of the intensity signals of the aphidicolin-treated cell DNA to the control DNA. Each black bar represents the average of 10 probes with 475-bp median spacing. The red lines represent the statistical average and show discontinuities, when a series of averaged probe data deviates in a statistically significant manner from its neighbors. Transcripts are indicated by blue lines and the names of select genes are indicated. The fifth and six clones are related to each other and are referred to as subclones 5a and 5b. (c) List of the 17 largest genes of human chromosome 3, according to their size, indicating also the presence of identified genomic aberrations in the five aphidicolin-treated clones. The status column indicates the number of clones with a genomic aberration and the type of aberration. del, deletion

Full size image

Six aphidicolin-treated clones were isolated for cGH analysis spanning the entire chromosome 3. Two clones were essentially identical, except for one small deletion, suggesting that they arose from a common ancestor cell. The concordance of the findings for these two clones provided a validation of the quality of the cGH analysis, but reduced the number of independent clones from six to five. In addition to deletions targeting the FHIT gene in FRA3B in four out of the five clones, two clones had deletions within the PTPRG gene, also mapping to the FRA3B CFS (Figure 7b). Deletions were further found in the genes dedicator of cytokinesis 3 (DOCK3), cell adhesion molecule 2 (CADM2), limbic system-associated membrane protein (LSAMP) and N-acetylated alpha-linked acidic dipeptidase-like 2 (NAALADL2), none of which map within established CFS (Supplementary Figure 3). Finally, a gain was detected within the ELKS–RAB6-interacting/CAST family member 2 (ERC2) gene (Figure 7b). An interesting feature of this analysis is that all the identified CNCs mapped to large genes. In fact, the CNCs targeted 7 of the 17 largest genes of chromosome 3, whereas no CNCs were detected in any of the remaining annotated genes of chromosome 3 (Figure 7c). Further, while DNA replication stress is usually associated with deletions, one of the observed CNCs was a gain. Similar findings have been obtained by another group using a similar experimental approach.32, 33 Thus, based on the spectrum of CNCs induced by aphidicolin, we propose that of the eight CNCs observed in the induced stem cells (mammary stem cells and iPS cells), six may be linked to DNA replication stress.

Discussion

Previous analyses of mouse and human iPS cells has demonstrated the presence of aneuploid karyotypes in some clones.3, 34 However, it was assumed that the clones that had diploid karyotypes did not have genomic aberrations.3, 35 Our analysis and data accumulating in the literature suggests that this assumption was incorrect.36, 37, 38 Although it is difficult to be conclusive, the most likely mechanism underlying the observed genomic instability in induced stem cells is oncogene-induced DNA replication stress.6, 39, 40, 41, 42 The presence of CNCs targeting very large genes and the lower frequency of genomic aberrations in stem cells induced without c-Myc support this tentative conclusion. It is important to note that the mechanisms by which c-Myc and other oncogenes induce DNA replication stress and genomic instability are not well established. Yet, we know that genomic instability is not directly linked to the number of cell divisions or to proliferation rate.39 In some settings, oncogenes induce genomic instability within one cell cycle.43 Importantly, the induced stem cells we examined were all early passage and were all derived from very early passage mammary cells or MEFs.

A key question is whether a few genomic aberrations compromise the function and utility of reprogrammed stem cells. To answer this question, it will be important to determine whether the observed genomic aberrations are due to transient genomic instability during the reprogramming process or whether the genomic instability persists even after reprogramming. Our study does not address this point. We note that the induced mammary stem cells described here, despite harboring CNCs, were capable of repopulating cleared mouse fat pads and none of the mammary glands derived from these cells have become cancerous so far (in 20 reconstituted mice; albeit within the short time frame of 7 months after transplantation; data not shown). Also, it has been established that OSK and OSKC-iPS cells can form viable fertile mice; yet, the efficiency is variable, possibly reflecting the presence of genomic aberrations in the iPS cells.2, 3, 5, 10 Finally, mice derived from iPS cells often develop tumors and many of them die in utero, suggesting developmental abnormalities.2, 10

As considerable effort is being placed to develop more efficient protocols for inducing stem cells, care must be taken that increased efficiency is not achieved at the expense of genomic stability. For example, c-Myc enhances the efficiency of reprogramming,1, 4, 5 but in our study it promoted genomic instability. Inactivation of p53, which also increases the efficiency of reprogramming,11, 12, 13, 14 probably does so by allowing genomically unstable cells to escape apoptosis; thus explaining why iPS cells derived from p53-deficient fibroblasts form malignant tumors in mice.17 Finally, some of the chemicals used for reprogramming, such as 5-azacytidine, are known DNA damage-inducing agents.44 On a brighter note, if oncogene-induced DNA replication stress is a significant contributor to the genomic instability observed in induced stem cells, then modifications to the reprogramming protocols that mitigate oncogenic stress should improve the quality of the generated cells. Nevertheless, at this time, our results and the results of others,36, 37, 38 suggest that great caution should be exercised when planning human therapies using induced stem cells.

Materials and Methods

Induced mammary stem cells

To prepare mammosphere cultures, mammary tissues from 5-month-old virgin female FVB mice (Harlan) or p53−/− mice (in the C57/BL6J background, back-crossed from a 129v background) were collected and dissociated mechanically. Disaggregation of the tissues was completed by enzymatic digestion in DMEM supplemented with 100 U/ml Hyaluronidase (Sigma, St. Louis, MO, USA) and 200 U/ml Collagenase (Sigma) for approximately 3 h at 37 °C. The digested material was then centrifuged and filtered through 100, 70, 40 and 20 μm meshes. Red blood cells were lysed by incubating the cell suspension in 0.2% NaCl. The resulting cells were plated on ultralow adhesion plates (Falcon, BD Biosciences, San Jose, CA, USA) at 100 000 cells/ml in MEBM medium (BioWhittaker, Walkersville, MD, USA) supplemented with 5 μg/ml insulin, 0.5μg/ml hydrocortisone, 2% B27 (Invitrogen, Carlsbad, CA, USA), 20 ng/ml EGF and bFGF (BD Biosciences, San Jose, CA, USA) and 4 μg/ml heparin (Sigma). Primary mammospheres were allowed to form for 6 days. At passaging, the number of mammospheres was counted, the mammospheres were then dissociated in single cells and the number of cells counted. A total of 5000 cells were then plated in 24-well plates. At each passage, the number of mammospheres reflects the number of stem cells that had been plated, because a stem cell can form a mammosphere, whereas, by definition, progenitor cells cannot form mammospheres.

For distinguishing native stem cells from progenitor cells, primary mammary cells were incubated with the PKH26 dye (Sigma) for 5 min, as previously described.19 The reaction was blocked in 1% BSA and the cells were plated and passaged twice to obtain secondary mammospheres. Single cells obtained from disaggragated secondary mammospheres were sorted by flow cytometry on the basis of PKH26 fluorescence (FACS Vantage SE flow cytometer, Becton and Dickinson, Franklin Lakes, NJ, USA). The PKH26-high population was isolated as the most epifluorescent of the total population (about 0.3% of the cells). PKH26-negative cells were gated at 10 times less fluorescence units with respect to the PKH26-high population (about 30% of the cells). Since stem cells divide once or just a few times during each passage, they retain high levels of the dye, whereas the progenitor cells that divide many times, become PKH-negative. For non-infected mammospheres obtained from wild-type mice, the frequency of stem cells in the PKH26-high population is about 1 : 3, whereas the corresponding frequency in the PKH26-negative population is <1 : 80 000.19

Lentiviral infections were performed using the lentiviral vector pWPI (Addgene, Cambridge, MA, USA) carrying the GFP reporter gene and either no other insert (Empty Vector) or MycER as an insert. Disaggregated primary mammospheres or PKH26-negative cells were plated in Phoenix-generated viral supernatants and infections were carried out in three cycles of 6 h each. After 6 days in culture, secondary mammospheres were collected, disaggregated and FACS sorted for the expression of GFP. For serial passage experiments, 5000 cells from disaggregated mammospheres were plated in 24-well plates and, after 6 days, counted, and re-plated at the same density. Treatment with 500 nM TAM (Sigma) was performed continuously for 7 days during mammosphere formation or once on replating of disaggregated mammospheres on day 1 of culture.

For transplantation experiments, mammosphere cell suspensions were pelleted into Eppendorf benchtop microfuge tubes, counted and resuspended in PBS. In all, 30 μl of cell suspensions were transplanted in the cleared fat pad of 3-week-old virgin female FVB mice. The presence of positive outgrowths was evaluated 2 months later by whole mount analysis.

Immunoblot analysis of protein extracts from mammosphere cultures was performed using anti-vinculin (Sigma), anti-p53 (clone AI25, gift from K Helin), anti-phospho-serine15-p53 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-p21 (clone F5, Santa Cruz Biotechnology) and anti-cleaved caspase3 (clone D175, Cell Signaling Technology, Danvers, MA, USA) primary antibodies, HRP-conjugated secondary antibodies (Sigma) and the SuperSignal West Pico Substrate detection kit (Pierce, Rockford, IL, USA). For immunofluorescence, single cell suspensions were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 and blocked with 3% BSA. Cells were stained with anti-Myc antibody (provided by S Hann, Vanderbilt University School of Medicine, Nashville, TN, USA).

iPS cells induced by OSK

MEFs containing an Oct4-GFP transgene45 were infected with a lentiviral vector-expressing human OCT4, SOX2 and KLF4 (OSK) from an SFFV promoter as a single transcript with self-cleaving 2A sequences separating each gene. Colonies of reprogrammed cells were picked and expanded after 20–25 days. iPS cells were cultured on feeder layers in embryonic stem cell (ESC) medium (DMEM, 15% knock-out serum replacement supplement (Invitrogen), L-glutamine, penicillin–streptomycin, nonessential amino acids, β-mercaptoethanol, and 1000 U/ml LIF).

Expression of endogenous pluripotency genes, such as Nanog, Oct4 and Sox2, was measured in iPS cells and matched MEFs in triplicate by quantitative PCR. Expression levels were normalized to the levels present in embryonic stem (ES) cells. For embryoid body differentiation, the iPS cells were washed in IMDM medium supplemented with FBS, L-Glu and MonoThioGlycerol to remove LIF. Cells were then resuspended in methylcellulose (40%) containing IMDM medium and plated. GFP expression from the Oct4-GFP reporter was monitored for 10 days.

iPS cells induced by OSKC

MEFs containing an Oct4-GFP knock-in reporter allele46 were infected with a lentiviral vector-expressing mouse Oct4, Sox2, Klf4 and c-Myc (OSKC) from a doxycycline-regulated TetO-mini CMV promoter as a single transcript with self-cleaving 2A sequences separating each gene.47 After lentiviral infection, the cells were cultured in ESC medium and treated with doxycycline (1 μg/ml) for 23 days. iPS cell colonies were isolated 12 days after doxycycline withdrawal. Genomic DNA was extracted from passage 6 iPS cells.

To monitor expression of pluripotency markers, iPS cells were washed in PBS, stained with anti-SSEA-1 phycoerythrin-conjugated antibody (eBioscience, San Diego, CA, USA) and examined for SSEA1 and Oct4-GFP expression by flow cytometry (FACS Calibur flow cytometer, Becton and Dickinson). In addition, expression of the endogenous genes Oct4, Zfp296, Eras and Fgf4 was monitored in the iPS cells, matched MEFs and ES cells by PCR, as previously described.48 Embryoid body differentiation was induced, as described above for the iPS-OSK cells. 21 days later, embryoid bodies were collected, fixed and stained with hematoxylin and eosin.

Aphidicolin-induced DNA replication stress

The GM11713A cell line, a mouse–human hybrid cell line containing a single wild-type copy of human chromosome 3 (Coriell Cell Repository, Camden, NJ, USA), was subjected to DNA replication stress over a period of 30 days, by adding every 4 days to the tissue culture media 0.4 μ M aphidicolin. At the end of the 30-day period, the cells were trypsinized and hundred cells were plated on a 10 cm diameter plate. Two weeks later, single colonies were isolated and expanded for 3 weeks, at which time genomic DNA was prepared.

cGH analysis

Control DNA was labeled with Cy5, whereas DNA from the induced stem cells or aphidicolin-treated cells was labeled with Cy3. Equal amounts of Cy5- and Cy3-labeled DNA were hybridized against high-density tiling microarrays. For analysis of mouse genomic DNA the tiling arrays contained probes covering chromosomes 5–8 and parts of chromosomes 4 and 9 (Roche Nimblegen, Madison, WI, USA, array MM8 WG CGH2; Build 36) and/or chromosomes 10–13 and parts of chromosomes 9 and 14 (Roche Nimblegen array MM8 WG CGH3; Build 36). The median probe density on these arrays was 1388 bp. For analysis of human genomic DNA, the tiling arrays contained probes covering chromosome 3 (Roche Nimblegen array HG18 CHR3 FT B3734001-00-01; Build 36). The median probe density on this array was 475 bp.