Lower genomic stability of induced pluripotent stem cells reflects increased non-homologous end joining

Background Induced pluripotent stem cells (iPSCs) and embryonic stem cells (ESCs) share many common features, including similar morphology, gene expression and in vitro differentiation profiles. However, genomic stability is much lower in iPSCs than in ESCs. In the current study, we examined whether changes in DNA damage repair in iPSCs are responsible for their greater tendency towards mutagenesis. Methods Mouse iPSCs, ESCs and embryonic fibroblasts were exposed to ionizing radiation (4 Gy) to introduce double-strand DNA breaks. At 4 h later, fidelity of DNA damage repair was assessed using whole-genome re-sequencing. We also analyzed genomic stability in mice derived from iPSCs versus ESCs. Results In comparison to ESCs and embryonic fibroblasts, iPSCs had lower DNA damage repair capacity, more somatic mutations and short indels after irradiation. iPSCs showed greater non-homologous end joining DNA repair and less homologous recombination DNA repair. Mice derived from iPSCs had lower DNA damage repair capacity than ESC-derived mice as well as C57 control mice. Conclusions The relatively low genomic stability of iPSCs and their high rate of tumorigenesis in vivo appear to be due, at least in part, to low fidelity of DNA damage repair.


Background
Embryonic stem cells (ESCs) are pluripotent and could differentiate into all types of somatic cells [1]. ESCc have enormous potential in the treatment of a variety of diseases, but their clinical application has been limited by ethical controversy. In 2006, Yamanaka and colleagues overexpressed four transcription factors (Oct4, Sox2, c-Myc and Klf4) in mouse somatic cells and obtained ESC-like pluripotent stem cells, termed induced pluripotent stem cells (iPSCs) [2]. iPSCs resemble ESCs in morphology, gene expression profile, epigenetic status and in vitro differentiation capacity. The development of iPSCs raises new hope for personalized clinical therapy [3][4][5].
The four transcription factors (Oct4, Sox2, c-Myc and Klf4) that are critical for the production of iPSCs are frequently overexpressed in various cancers, and mice derived from iPSCs are prone to develop tumors [6][7][8][9].
Although only a small population of transformed cells with genetic mutations is likely to develop into tumors [10], the genomic instability of iPSCs is a major concern that could produce huge impact on their eventual clinical use [11][12][13][14][15][16].
One possible explanation for the observed greater genomic instability of iPSCs is alterations in the fidelity of DNA repair pathways. Double-stranded DNA breaks, for example, can be repaired via homologous recombination (HR) with high fidelity, or via non-homologous end joining (NHEJ) with lower fidelity [17][18][19][20]. In the current study, we examined whether iPSCs differ from other types of pluripotent cells in their ability to perform these types of DNA repair. Briefly, ionizing radiation was used to induce double-stranded DNA breaks in the following cells: mouse iPSCs induced using lentivirus (lv-iPSCs) or chemically with CHR99021, Repsox and forskolin (ci-iPSCs) [21]; mouse ESCs; and mouse embryonic fibroblasts (MEFs) [22][23][24][25][26].
The experiments showed that lv-iPSCs are more likely than the other cell types to harbor genomic abnormalities, likely due to lower genomic fidelity of DNA damage repair. We also found greater genomic stability in ci-iPSCs than lv-iPSCs.

Irradiation
Cells were passaged 1 day before γ-irradiation (4 Gy) with a cobalt irradiator (Thermo Fisher Scientific). After the irradiation, cells were immediately returned to the incubator, and cultured for 4 h prior to analyses as described below.
In assays of histone modification, cells were re-suspended in 1-mL triton extraction buffer (TEB) containing 0.5% Triton X-100 and 2 mmol/L PMSF, and then lysed on ice for 10 min. The lysates were centrifuged at 1500g for 10 min at 4 °C. The pellet was washed with 1.5-mL TEB, re-suspended in 0.2 mol/L HCl, and incubated at 4 °C overnight. Samples were centrifuged at 6500g for 10 min, after which 200-µL supernatant was transferred to a new tube, and neutralized with 20-µL 2 mol/L NaOH.

Immunofluorescence labeling of γ-H 2 AX foci
Cells were passaged onto slides, exposed 24 h later to 4 Gy of γ-irradiation, and incubated at 37 °C for 4 h. Cells were washed with PBS, fixed with 4% paraformaldehyde for 10 min at room temperature, washed again with PBS, permeabilized for 10 min using 0.05% Triton X-100 and 0.5% NP-40, and then washed three times (5 min each) in PBS. The cells were blocked for 1 h with 2% bovine serum albumin (BSA), and then incubated for 1 h at room temperature with a mouse anti-γH 2 AX antibody (1:500; Millipore, Temecula, CA, USA). Cells were washed three times with PBS containing 0.05% Tween 20, and then incubated with a goat anti-mouse secondary antibody (1:800; Abcam) for 1 h in the dark at room temperature. Cells were counterstained with 0.2 mg/mL 4′,6-diamidino-2-phenylindole (DAPI, 1:2000; Sigma, Shanghai, China). Confocal images were acquired and analyzed using a TCS SP5 (Leica) microscope equipped with an HCX PL 63 × 1.4 CS oil-immersion objective lens.

DNA extraction
Three types of cells (lv-iPSCs, ci-iPSCs, ESCs) were digested with 0.25% trypsin and re-suspended in gelatin-coated dishes. After incubation at 37 °C for 15 min, supernatants were transferred to 15-mL centrifuge tubes, and cells were collected by centrifugation at 500g for 5 min at room temperature. DNA was extracted using a QIAamp DNA Mini Kit (Qiagen, Hilden, Germany).

Whole-genome re-sequencing
Whole-genome DNA libraries suitable for sequencing using an Illumina sequencing platform were generated from 1-µg genomic DNA. The DNA was sheared to approximately 300-500 bp using a Covaris S220 instrument (Life Technologies, Carlsbad, CA, USA). A total of 2× 101-bp paired-end reads were produced using the HiSeq 2000 DNA Sequencer.
The sequencing data were mapped to a reference mouse genomic sequence (mm9) using the Burrows-Wheeler alignment tool algorithm [31]. Unique alignment reads were retained for later analysis. Using the untreated cells as a control, single-nucleotide variations (SNVs) were collected using the "mpileup" tool in SAMTools as well as the UnifiedGenotyper in the GATK module [32,33]. Quality recalibration and local realignment were performed using GATK tools before variation calling was performed. The following criteria were applied for calling mutations using pairwise samples: (1) the minimum coverage of variant sites had to be greater than 20 and base quality greater than 15; (2) the frequency of mutant SNVs had to be 0 in control samples and 0.2 in irradiated samples; and (3) the variant sites had to be supported by at least two reads on the forward strand and two reads on the reverse strand.

RNA sequencing
Total RNA was extracted from each cell line using TRIzol reagent and enriched for mRNA using oligo (dT) magnetic beads. Approximately 1-µg mRNA was fragmented and electrophoresed to isolate mRNA fragments (200-250 bases). These fragments were subjected to end repair, 3′ terminal adenylation and adapter ligation, followed by cDNA synthesis. The resulting cDNAs were gel-electrophoresed to isolate 250-300 bp fragments, and were sequenced using a HiSeq 2000 system (Illumina).
Sequencing reads were aligned to a reference sequence (GRCm37/mm9) using TopHat alignment software [34,35]. Only uniquely aligned reads were used for transcript assembly, which was performed using Cufflinks software [36]. Read counts for each gene were calculated, and the expression levels of each gene were normalized using the "fragments per kilobase of exon model per million mapped" (FPKM) algorithm. Differentially expressed genes were filtered based on false discovery rate (FDR)-adjusted P < 0.05. The profile of differentially expressed genes was visualized and analyzed using the Bioconductor function "CummeRbund" in the R program [37]. Hierarchical clustering was performed using the "heatmap" package in R.

Generation of iPSC-and ESC-derived mice
Two cell-stage ICR embryos were electrofused to produce tetraploid embryos, and 10-15 iPSCs and ESCs were subsequently injected into the reconstructed tetraploid blastocysts. Embryos were cultured for 1 day prior to transplantation into the uterus of pseudo-pregnant mice. Caesarean sections were performed at E19.5, and the pups were fostered by lactating ICR mothers [38].

Comet assay
Mice derived from iPSCs or ESCs as well as C57 mice were treated with 4 Gy ionizing radiation. At 4 h later, bone marrow cells were isolated and re-suspended using PBS and concentrated by adding 150-μL molten 0.75% low-melting-point agarose. An aliquot of concentrated cells (60 μL) was then added to molten 0.8% normal-melting-point agarose on comet slides. The slides were incubated for 1-2 h at 4 °C with pre-chilled lysis buffer, stored in the dark at 4 °C for 20 min, then incubated with pre-chilled electrophoresis buffer (0.3 mol/L NaOH containing 0.5 mol/L EDTA, pH > 13.0). Gel electrophoresis was performed at 25 V for 20 min at 4 °C. Slides were incubated at 4 °C for 15 min with neutralization buffer (0.4 mol/L Tris, pH > 7.5), washed with 100% ethanol for 3-5 min and air-dried at room temperature. Diluted ethidium bromide (EB) solution (20-30 μL) was placed onto each dried agarose circle. Slides were then read at 100 cells/sample using a fluorescence microscope equipped with CASP DNA damage analysis software.

Similar gene expression profile between lv-iPSCs and ESCs
RNA-seq analysis showed that the gene expression profile of lv-iPSCs was similar to that of ESCs but not to that of MEFs (Fig. 1a), indicating iPSC pluripotency. Since genomic stability depends on DNA damage repair, we analyzed expression of the genes involved in DNA damage repair pathways. No significant differences in the expression of such genes were found between lv-iPSCs and ESCs (Fig. 1b). We further analyzed the expression of DNA repair genes that were identified during early reprogramming of iPSCs in our previous report [39] and confirmed the up-regulation of those genes at early iPSC stages (Fig. 1c). These results suggest that DNA damage repair pathways can be reprogrammed at early iPSC stages and become similar to pathways in ESCs as reprogramming continues [39].

More DNA mutations in lv-iPSCs than in other cell types after ionizing irradiation
We treated mouse lv-iPSCs, ESCs and MEFs with 4 Gy ionizing radiation to induce double-strand breaks. If not repaired properly, such breaks can result in genomic abnormalities, apoptosis and senescence [23,26,40]. Whole-genome DNA sequencing at 4 h after irradiation revealed more SNVs in lv-iPSCs than in the other cell types (Fig. 2a, Table 1), as well as more short indels (Fig. 2a, Table 2). MEFs showed a larger variety of copy number variations (CNVs) than the other cell types (Fig. 2a).

Similar gene expression profile in lv-iPSCs with or without ionizing radiation
To determine whether ionizing radiation alters the expression of certain genes in lv-iPSCs that may help explain the high mutation rate, RNA-seq analysis was conducted in irradiated versus control cells. The results indicated a similar gene expression profile with or without radiation (Fig. 3a). In fact, irradiation appeared to upregulate only 46 genes in ESCs and 30 genes in lv-iPSCs (Fig. 3b). In contrast to the genes in lv-iPSCs that radiation up-regulated, majority of the genes up-regulated in ESCs is implicated in cellular response to stress and cell cycle processes (Fig. 3c, d).
Expression levels of genes involved in DNA damage repair pathways were higher in lv-iPSCs and ESCs than in MEFs, and ionizing radiation did not substantially alter the expression of these genes (Fig. 3e). Thus the genomic instability of lv-iPSCs is unlikely to reflect changes in the expression level of genes involved in DNA damage repair.

Weaker DNA damage repair response to ionizing radiation in lv-iPSCs
The phosphorylated histone variant H2AX (γ-H2AX) is a marker of double-strand breaks. Ionizing radiation significantly increased the number of γ-H2AX foci in lv-iPSCs, ESCs and MEFs, but the magnitude of decrease was much smaller in lv-iPSCs (Fig. 4a), suggesting lower capacity to repair DNA damage.
Next we tested whether the lower genomic stability of lv-iPSCs reflects deficiency in the error-free HR repair pathway. Indeed, we found ATM phosphorylation to be defective in lv-iPSCs (Fig. 4b) [30,41]. We also found lower levels of H3K9me3, which recruits repair proteins to double-strand breaks, in irradiated lv-iPSCs than in irradiated ESCs or MEFs (Fig. 4c). All together, these findings may help explain the higher mutation rate of lv-iPSCs.

Lower genomic stability in lv-iPSCs than ci-iPSCs
Treatment with ionizing radiation led to higher levels of phosphorylated ATM in ci-iPSCs than in lv-iPSCs (Fig. 5a). This may help explain the higher genomic stability of ci-iPSCs [41]. Whole-genome re-sequencing at 4 h after irradiation revealed 1709 SNVs in the ci-iPSCs;   this was slightly more than in treated ESCs but far less than in lv-iPSCs (Fig. 5b). Similarly, the proportion of SNVs in coding sequences, introns, 5′ or 3′ UTRs and intergenic regions was slightly higher in ci-iPSCs than in ESCs, but much higher in lv-iPSCs (Fig. 5c, d). These results indicate greater genomic stability in ci-iPSCs than in lv-iPSCs, which is due at least in part to greater activity of the HR pathway of DNA damage repair.

lv-iPSCs can tolerate more genomic DNA variation
The abovementioned results led us to hypothesize that lv-iPSCs can survive with greater genomic variation than the other cell types. Consistent with this hypothesis, we found that lv-iPSCs indeed had more DNA variation than the other cell types, yet the percentage of apoptotic lv-iPSCs did not increase between 24 and 48 h after irradiation (Fig. 6a) and the rate of lv-iPSC proliferation was greater than that of ESCs or MEFs (Fig. 6b). When we analyzed whether irradiation arrested lv-iPSCs in the G2/M phase, we observed a high proportion of arrested cells at 24 h after irradiation, but a lower proportion at 48 h (Fig. 6c). We observed similar results with ESCs, showing an increased proportion of ESCs in G2/M phase at 24 h after irradiation and a lower radiation arrest at 48 h. These results suggest that lv-iPSCs tolerate greater genomic DNA variation than the other cell types.
lv-iPSCs are more susceptible to DNA damage Next we compared genomic stability in mice derived from lv-iPSCs versus ESCs. C57 mice were included as additional control. Irradiation of the mice led to a higher percentage of impaired bone marrow cells (Fig. 7a-c) and of tail DNA in bone marrow cells (Fig. 7d) in iPSCderived mice than in ESC-derived mice and C57 mice. These results suggest that mice derived from lv-iPSCs have lower DNA damage repair capability than ESCderived or C57 mice and are therefore more susceptible to DNA damage. Taken together, our in vitro and in vivo experiments suggest that lv-iPSCs are more sensitive to environmental stress than ci-iPSCs, ESCs or MEFs. Ionizing radiation induces higher genomic mutation rates in lv-iPSCs, which nevertheless better tolerate the resulting genomic alterations. Genomic mutations that accumulate in lv-iPSCs are passed onto the next generation, resulting in genomic instability (Fig. 8).

Discussion
Reprogramming to generate iPSCs more efficiently [29,[42][43][44][45][46][47][48][49][50][51] has been linked to the accumulation of genomic abnormalities [52][53][54][55][56][57][58][59]. This poses a problem for the use of iPSCs, since mice derived from such cells can tolerate the accumulation of somatic mutations for up to six generations [60]. In the present study, we used wholegenome sequencing to compare the genomic stability of iPSCs prepared using lentivirus or chemically, and to benchmark that stability against ESCs and MEFs. We found that ionizing irradiation led to the highest rate of somatic mutations and short indels in lv-iPSCs, and this correlated with low levels of ATM phosphorylation,       indicating low fidelity of DNA damage repair [41]. Experiments in vitro and in mice derived from lv-iPSCs showed that this type of pluripotent cell tolerates genomic mutations better than the other cell types evaluated. Although iPSCs resemble ESCs in morphology, gene expression profile and in vitro differentiation capacity, they differ substantially in genomic stability. The low fidelity of DNA repair observed in our study suggests that irradiation of lv-iPSCs induces a high rate of genomic abnormalities, which is less likely to trigger apoptosis in these cells and is therefore more likely to be tolerated, thus leading to a high rate of tumorigenesis in vivo. Compromised error-free HR pathway of DNA damage repair in lv-iPSCs may help explain the relatively high genomic instability in these cells. Indeed, inhibiting the HR pathway in iPSCs has been shown to destabilize the genome [61].
Our results suggest that the epigenetic status of iPSCs may contribute to, or modulate, their genomic instability. Variation in levels of H3K9me3 and phosphorylated ATM among iPSCs may mean that cells vary in their reliance on DNA damage repair pathways, which vary in their fidelity. Future studies should further examine the potential involvement of epigenetics and other factors in iPSC genomic instability.
Future work is also needed to clarify to what extent factors that are intrinsic or extrinsic to stem cells determine the risk of malignant transformation. Tomasetti et al. found that cancer risk in certain tissues correlated strongly with the number of divisions that the stem cells had undergone, suggesting that the accumulation of genomic mutations is primarily responsible for high risk of tumorigenesis [62]. Another study, in contrast, suggested that intrinsic factors account for only  10%-30% of cancer risk, with the majority of the risk due to extrinsic factors [63]. The results from the present study suggest that extrinsic factors induce more genomic mutations than intrinsic factors in lv-iPSCs. The high rate of tumorigenesis of iPSCs in vivo suggests that extrinsic factors strongly contribute to cancer risk and carcinogenesis.

Conclusions
The present study demonstrated a low level of DNA damage repair in iPSCs. Ionizing radiation induced more somatic mutations and short indels in iPSCs than in ESCs or MEFs. Genome stability was higher in iPSCs induced chemically than in iPSCs induced with lentivirus. The high genome instability of lv-iPSCs appears to reflect  Greater tolerance of genomic DNA variation in lv-iPSCs. a Flow cytometric analysis of apoptosis rate in lv-iPSCs, ESCs and MEFs following ionizing irradiation (IR) for the indicated periods. 7-AAD 7-amino-actinomycin. b Cell proliferation rate (based on BrdU incorporation) in lv-iPSCs, ESCs and MEFs exposed to ionizing irradiation for the indicated periods. Each point represents a mean of three replicates **P < 0.01. c Analysis of cell cycle distribution in lv-iPS, ESCs and MEFs exposed to ionizing radiation for the indicated periods Fig. 7 Genome stability of mice derived from lv-iPSCs or ESCs following exposure to ionizing radiation (IR). Controls were C57 mice. a Mice were generated from lv-iPSCs or ESCs through tetraploid embryo complementation. Representative results from three independent experiments are shown. b Examples of bones from the three types of mice, from which marrow cells were extracted. c Box plots showing the percentage of impaired bone marrow cells in each mouse strain. DNA damage was evaluated using single-cell gel electrophoresis **P < 0.01. d Box plots showing the percentage of Tail DNA in impaired cells as a measure of DNA damage. Tail DNA% = Tail DNA intensity/Cell DNA Intensity × 100%. CASP software was used to calculate tail moment based on 50-100 randomly selected cells per sample increased NHEJ and decreased HR pathways of DNA damage repair, and could contribute to the high rate of tumorigenesis in vivo.