A tight control of Rif1 by Oct4 and Smad3 is critical for mouse embryonic stem cell stability

Prolonged culture of embryonic stem cells (ESCs) leads them to adopt embryonal carcinoma cell features, creating enormous dangers for their further application. The mechanism involved in ESC stability has not, however, been extensively studied. We previously reported that SMAD family member 3 (Smad3) has an important role in maintaining mouse ESC stability, as depletion of Smad3 results in cancer cell-like properties in ESCs and Smad3−/− ESCs are prone to grow large, malignant teratomas. To understand how Smad3 contributes to ESC stability, we performed microarray analysis to compare the transcriptome of wild-type and Smad3−/− ESCs. We found that Rif1 (RAP1-associated protein 1), a factor important for genomic stability, is significantly upregulated in Smad3−/− ESCs. The expression level of Rif1 needs to be tightly controlled in ESCs, as a low level of Rif1 is associated with ESC differentiation, but a high level of Rif1 is linked to ESC transformation. In ESCs, Oct4 activates Rif1, whereas Smad3 represses its expression. Oct4 recruits Smad3 to bind to Rif1 promoter, but Smad3 joining facilitates the loading of a polycomb complex that generates a repressive epigenetic modification on Rif1 promoter, and thus maintains the expression of Rif1 at a proper level in ESCs. Interestingly, Rif1 short hairpin RNA (shRNA)-transduced Smad3−/− ESCs showed less malignant properties than the control shRNA-transduced Smad3−/− ESCs, suggesting a critical role of Rif1 in maintaining the stability of ESCs during proliferation.

Embryonic stem cells (ESCs) can serve as a rich source of differentiated cells for cell-based therapy due to their pluripotency and unlimited self-renewal capacity. However, prolonged culture of ESCs results in ESCs accumulating numerous mutations, and they gradually adopt embryonal carcinoma cell features. [1][2][3] This prompts serious safety concerns with regard to ESC applications and also raises important questions regarding how ESCs maintain their genomic stability.
Transforming growth factor beta (TGF-β) signaling has an important role in development and homeostasis. It also functions in multiple diseases such as cancer, tissue fibrosis and diabetes. 4,5 Through their respective ligand receptors, TGF-β/Activin/Nodal activates SMAD family member (Smad) 2/Smad3. The activated Smads bind to Smad4 and translocate from cytoplasm to the nucleus to regulate the downstream genes. 6,7 TGF-β/Activin/Nodal signaling is crucial for maintaining self-renewal and pluripotency in human ESCs, but appears to be dispensable for the pluripotency of mouse ESCs. 8,9 Instead, the activation of Activin/Nodal signaling is required for the propagation of mouse ESCs. 10,11 Smad3 is a downstream factor of TGF-β/Activin/Nodal signaling. Although depletion of Smad3 leads to transient expression distortion of mesoderm markers during embryoid body (EB) formation, the final lineage formation is not affected, 12 as Smad3 knockout mice are viable and fertile. 13 This may be because Smad2, another downstream factor of TGF-β/Activin/Nodal signaling, has a redundant role of Smad3. It has been reported that Smad2 and Smad3 collaboratively regulate mesoderm formation during embryo development. 12,14 Previously, we found that activation of Smad3 is vitally important for ESCs to maintain their genetic integrity during propagation, as depletion of Smad3 leads mouse ESCs to adopt cancer cell properties. 12 To further illustrate how Smad3 contributes to ESC stability, we performed microarray assay to identify genes that show an obvious change after Smad3 depletion. Among the genes affected by Smad3 depletion, Rif1 (RAP1-associated protein 1), a factor closely associated with chromatin stability, shows the greatest upregulation.
Rif1 is first identified in budding yeast as a Rap1interacting factor. It is recruited to the telomere by Rap1 and implicated in maintaining telomere structure and homeostasis. 15,16 In mammalian cells, except the regulation of telomere homeostasis, 17,18 Rif1 mediates the ATM (ataxia telangiectasia mutated)/53BP1 (tumor suppressor p53-binding protein 1) signaling after DNA damage to repress break resection and promote the non-homologous end joining (NHEJ) mechanism in G1 phase. [19][20][21][22][23] In addition, Rif1 globally regulates the replication-timing program in both yeast fission and mammalian cells. [24][25][26][27] Rif1 localizes to the stalled replication forks in response to ATR activation and serves as a component of the DNA replication checkpoint. [28][29][30][31] Rif1 is also highly expressed in the pluripotent stem cells. [32][33][34] Knockdown of Rif1 by RNA interference in mouse ESCs leads to ESC differentiation. 35 In this study, we determine that Rif1 is an important contributor to ESC stability during its propagation. Rif1 expression level is tightly controlled by Smad3 and Oct4. Reduction of Rif1 by RNA interference leads Smad3 − / − ESCs to show less malignant properties than control shRNA knockdown Smad3 − / − ESCs, suggesting that upregulation of Rif1 is a key factor in the transformation of Smad3 − / − ESCs.

Results
Rif1 is a direct downstream target of Smad3. Previously, we reported that depletion of Smad3 in mouse ESCs produced cancer cell-like features. 12 To understand the underlying mechanism, cDNA microarray analysis was performed to compare the transcriptome between wild-type (WT) and Smad3 − / − ESCs. Genes with more than a 1.5fold difference between WT and Smad3 − / − ESCs were selected by Partek software to generate a heat map. On the basis of the microarray data, the expression of Smad3 and Lefty1 was markedly reduced in Smad3 − / − ESCs. Besides, validation of eight randomly picked genes by real-time PCR further suggests that the microarray result was reliable. Among the genes that show different expression after Smad3 depletion, Rif1 ranked as the highest upregulated gene in Smad3 − / − ESCs (Figure 1a and Supplementary Figure S1A). Real-time PCR and western blot analysis confirmed the upregulation of Rif1 at both mRNA and protein level in Smad3 − / − ESCs (Figures 1b and c). Furthermore, overexpression of Smad3 in Smad3 − / − ESCs could significantly downregulate Rif1 expression, but upregulate Lefty1 expression ( Figure 1d). As Smad3 is a downstream factor of the Activin pathway in mouse ESCs, 10 we treated ESCs with Activin A (25 ng/ml) and Activin A inhibitor SB431542 (10 μM), respectively, to examine the expression of Rif1. As expected, the expression of Rif1 was decreased by Activin A treatment, but increased by SB431542 treatment. The expression of Lefty1 and Lefty2 was regulated conversely, confirming that Lefty1 and Lefty2 are positively regulated by Activin/Smad3 pathway, whereas Rif1 is negatively regulated by this pathway (Figures 1e and f). On the basis of Mullen's chromatin immunoprecipitation (ChIP)-seq data, there are two Smad3-binding sites (SBS1 and SBS2) at the promoter region of Rif1. 4 Therefore, we designed primers to quantitate Rif1-1 and Rif1-2 regions that cover SBS1 and SBS2, respectively. Examining ChIP-enriched DNA by real-time PCR, we found that Smad3 specifically bound to the Rif1-1 and Rif1-2 regions (Figure 1g). To further examine whether Rif1 promoter activity was affected by Smad3 depletion, a luciferase assay was performed with the Rif1 promoter containing the Smad3-binding sites. The result showed that Rif1 promoter activity was enhanced in Smad3 − / − ESCs compared with WT ESCs (Figure 1h). Taken together, all these data demonstrated that Rif1 is a target of the Activin/ Smad3 pathway, and that Smad3 represses Rif1 expression in mouse ESCs.
Inverse expression profiles between Smad3 and Rif1. To further examine the correlation between Smad3 and Rif1, we examined the expression profiles of these two genes in mouse ESCs, mouse embryonic fibroblasts (MEFs), teratoma cells and mouse ESC-differentiated cells. Smad3 showed higher expression in MEFs and teratoma cells than in ESCs. Conversely, Rif1 was expressed at a lower level in MEFs and teratoma cells than in ESCs (Supplementary Figures S1B, S1C and S2A). In addition, we also traced the expression changes of Smad3 and Rif1 during ESC differentiation using monolayer culture and EB formation (Supplementary Figures S1D and S2B). ESC differentiation was accompanied by gradual downregulation of pluripotent markers and expression of lineage markers (Supplementary Figures S1E and S2C). During ESC differentiation, the expression of Smad3 mRNA was gradually increased, whereas the expression of Rif1 mRNA was decreased ( Supplementary Figures S1F and S2D). Interestingly, although the expression of Rif1 was decreased in Smad3 − / − ESC-formed EBs, the Rif1 level was always   Figure S2E). These expression profiles also confirm previous reports that Rif1 is a factor associated with pluripotency. 32,35 Oct4 is required for Smad3 to bind to Rif1 promoter. Multiple Oct4-bound genes are found to be co-occupied by Smad3 and respond to TGF-β signaling. 4 To find out whether Rif1 is among these genes, we first knocked down the expression of Pou5f1 by RNA interference. The expression level of Rif1 was significantly decreased after Pou5f1 knockdown ( Figure 2a). This is consistent with the luciferase activity of Rif1 promoter being reduced to 20% after Pou5f1 knockdown, suggesting that Rif1 is regulated by Oct4 ( Figure 2b). Furthermore, overexpression of Pou5f1 in mouse ESCs can enhance the expression of Rif1 at both mRNA and protein level (Figures 2c and d). To determine whether Oct4 co-binds with Smad3 on Rif1 promoter region, we performed ChIP assay with Oct4 antibody. As expected, Oct4 was highly enriched on the promoter regions of Rif1 where Smad3 binds ( Figure 2e). These data support previous reports that Oct4 positively regulates Rif1 expression in mouse ESCs. 33,35 Through co-immunoprecipitation and sequential ChIP experiments, Mullen et al. 4 discovered that Oct4 can form a complex with Smad3 and recruit Smad3 to a Lefty1 enhancer to regulate Lefty1 expression in mouse ESCs. Prompted by the opposed regulatory roles of Smad3 and Oct4 on Rif1, we first examined whether Smad3 is required for Oct4 to bind to Rif1 promoter. As the Oct4 protein level is similar in WT and Smad3 − / − ESCs, 12 we directly performed a ChIP assay with Oct4 antibody. We found that Oct4 enrichment on Lefty1, Lefty2 and Rif1 in WT and Smad3 − / − ESCs was not obviously affected by the depletion of Smad3 (Figure 2f). We then investigated whether Oct4 is required for recruiting Smad3 to bind to Rif1. We knocked down Pou5f1 by shRNA. After 1-day selection with puromycin, the mRNA level of Pou5f1 was significantly reduced, but the expression of Smad3, pluripotent marker Nanog, lineage markers Cdx2, Cxcr4 and T were not significantly changed and ESCs still maintained the colony morphology (Supplementary Figure S3A). Meanwhile, the Oct4 protein level was obviously decreased, whereas the Smad3 was not affected (Supplementary Figure S3B). We performed ChIP assay with Smad3 antibody using cells at this stage and found that the binding efficiency of Smad3 on Lefty1, Lefty2 and Rif1 was significantly reduced after Pou5f1 knockdown ( Figure 2g). To further confirm that Oct4 and Smad3 co-bind to Rif1 promoter, we performed re-ChIP assay with a Smad3 antibody after Oct4 ChIP and discovered that Smad3 and Oct4 do bind to Rif1 promoter simultaneously ( Figure 2h). These data suggested that Oct4 is required for Smad3 to bind to Rif1 promoter, but Smad3 is not required for Oct4 binding.
Rif1 promoter shows Smad3-dependent H3K27 methylation. Mullen et al. 4 reported that Activin could induce both upregulation and downregulation of the expression of Oct4 and Smad3 co-occupied genes, indicating that Oct4 and Smad3 regulate their targets by sophisticated regulatory mechanisms. To uncover how Oct4 and Smad3 regulate Rif1, we performed the ChIP assay with histone modification markers, H3K9me2, H3K9me3, H3K4Me3 and H3K27Me3. H3K9me2 and H3K9me3 label the heterochromatin and H3K4me3 and H3K27me3 are the bivalent markers that label genes related to pluripotency and differentiation. [36][37][38] Rif1 promoter was enriched by H3K4me3 and H3K27me3, but not by H3K9me2 and H3K9me3 (Figures 3a and d). Depletion of Smad3 did not affect H3K4me3 enrichment (Figure 3c), but seriously affected H3K27me3 enrichment. However, H3K27me3 level at Lefty1 and Lefty2 was not obviously affected after Smad3 depletion ( Figure 3d). This result implies that Rif1 expression is controlled by a specific epigenetic modification involving Smad3. As Suz12 is the key component of the polycomb repressive complex (PRC2) and contributes to H3K27me3, we performed a ChIP assay with a Suz12 antibody. We found that Suz12 enrichment on Rif1 is significantly reduced after Smad3 depletion, but its enrichment on Lefty1 and Lefty2 was stable ( Figure 3e). Taken together, these data suggest that, although Oct4 and Smad3 co-bind to Lefty1, Lefty2 and Rif1, the mechanism used to regulate Rif1 is different to that used to regulate Lefty1 and Lefty2. In the Rif1 regulatory complex, Smad3 has a critical role in loading PRC2 to regulate the expression of Rif1 through H3K27 methylation.
Smad3 − / − ESCs show higher cell proliferation and DNA repair capacities than WT ESCs after ultraviolet irradiation. Our previous studies showed that Smad3 depletion enhanced the anti-apoptosis capacity of ESCs. To further substantiate this observation, we set out to examine the response of WT and Smad3 − / − ESCs to DNA damage in detail. Before ultraviolet (UV) irradiation, WT and Smad3 − / − ESCs showed no obvious difference after propidium iodide (PI) staining. Further, labeling the cells with bromodeoxyuridine (BrdU) revealed that more Smad3 − / − ESCs showed active DNA replication than WT ESCs. WT and Smad3 − / − ESCs were then exposed to UV irradiation (40 mJ/cm 2 ) to induce DNA damage. Five hours later, the cells were pulse labeled with BrdU and cultured for another 30 min. Subsequently, the cells were collected and stained with PI and FITC-conjugated BrdU antibody for flow cytometry analysis. About 7% Smad3 − / − ESCs showed active DNA replication compared with only about 3% for the WT cells (Figures 4a  and b). This result demonstrated that Smad3 − / − ESCs have a higher cell proliferation capacity than WT ESCs after DNA damage.
In responding to the DNA damage after UV irradiation, ATR-Chk1 and ATM-Chk2 are activated to modulate checkpoint, DNA repair, apoptosis and cell senescence. Replication protein A (RPA) is an ssDNA-binding protein in eukaryotes and prevents ssDNA from forming hairpin structures or re-annealing when the DNA is under repair, replication or recombination. [39][40][41][42] RPA is upregulated after DNA damage and it is essential for the ATR-mediated DNA damage checkpoint. 43,44 Smad3 − / − ESCs expressed a significantly higher level of RPA than WT ESCs at 1 h and 3 h after UV irradiation (Figure 4c). Consistent with this, phosphorylated Chk1 was significantly higher in Smad3 − / − ESCs than WT ESCs, whereas phosphorylated Chk2 was only slightly increased in Smad3 − / − ESCs (Figure 4d). These data imply an elevated DNA damage response in Smad3 − / − ESCs.
H2AX, a histone H2A variant, is phosphorylated by ATM after DNA damage. It binds to the damaged DNA and attracts more proteins to join the DNA repair, thus it is a good indicator of DNA damage. [45][46][47][48] Significant amounts of H2AX were detected in both WT and Smad3 − / − ESCs 1 h after UV  The expression of Ccnd2 (cyclin-D2), which is increased in Smad3 − / − ESCs, 12 was significantly decreased by Rif1 reduction (Figure 5d). Besides, BrdU integration assay revealed that proliferating cell number in Rif1 knockdown Smad3 − / − ESCs was significantly reduced to about the WT ESC level (Figures 5e and f), suggesting that upregulation of The cell migration capacity of Smad3 − / − ESCdifferentiated cells is reduced by knockdown of Rif1. Smad3 − / − ESC-differentiated cells show enhanced cell migration. 12 To examine whether upregulation of Rif1 contributes to cell migration (Supplementary Figure S6A), we performed wound-healing and transwell assays. After induction of ESC differentiation by withdrawing 2i and leukemia inhibitory factor (LIF), a wound scratch was generated. Three hours after the scratch, there was no obvious difference in the wound gap between control and Rif1 shRNA-transduced Smad3 − / − ESC-differentiated cells. But at 12 h after the scratch, the wound gap of the control sample was narrower than Rif1 knockdown sample, suggesting a reduced cell migration capacity caused by Rif1 knockdown. This difference was more obvious at 24 h (Figure 6a). In addition, a transwell assay also revealed that Rif1 knockdown Smad3 − / − ESC-differentiated cells had lower migration capacity than the control (Supplementary Figure S6B). This is consistent with the cell migration markers Mmp2 and Mmp9 being significantly downregulated in Rif1 knockdown Smad3 − / − ESC-differentiated cells compared with the control knockdown samples (Supplementary Figure S6C).  (Figures 6e and f). Collectively, these results demonstrated that upregulation of Rif1 is one of the main factors in Smad3 − / − ESC transformation.

Discussion
In this study, we found that Rif1, a factor involved in genomic stability, is tightly regulated by Oct4 and Smad3 in mouse ESCs. Oct4 recruits Smad3 to the Rif1 promoter and facilitates the loading of the PRC2. To maintain Rif1 expression at a proper level in ESCs, Oct4 activates Rif1 expression, but Smad3 is involved in repressing it (Figure 7). Mullen et al. reported that Nanog, Oct4 and Sox2 form a complex and tend to co-bind with Smad3 on many genes. From the ChIP-seq data of their report, we found that Nanog, Oct4 and Sox2 all bind to the SBE of Rif1. 4 It is therefore likely that Rif1 is synergistically regulated by these core transcription factors in ESCs. From past reports and our own studies, it seems very important to keep Rif1 expressed at a suitable level to sustain ESC pluripotency and stability. Low level of Rif1 leads to mESC differentiation. 35 However, high level of Rif1 is also deleterious to ESCs by driving malignant transformation of Smad3 − / − ESCs. Comparison of the expression of Smad3 and Rif1 between teratoma cells and teratocarcinoma cell lines F9 and P19 revealed that Smad3 was significantly lower in teratocarcinoma cells than teratoma cells, whereas Rif1 is significantly higher in teratocarcinoma cells than teratoma cells ( Supplementary Figures S7A and S7B). These data support our findings and indicate that a disturbance in the expression of Smad3 and Rif1 may be one of the underlying mechanisms for teratocarcinoma formation. It is reported that Rif1 colocalizes with DNA double-strand breaks (DSBs) and is involved in DNA repair. 48 Recent studies revealed that Rif1 contributes to the inhibition of 5′ end resection of DSBs, the first step of homologous recombination (HR). As a result, NHEJ, an error-prone DNA repair is promoted with Rif1 presence. In the absence of Rif1, the level of misjoined chromosomes is significantly reduced. 20,21,23,49 NHEJ leads to more chromatin instabilities, such as deletions, translocations and amplifications, than HR. Chromatin instability is closely linked to cancer cell formation as it enables rapid evolution of cell subclones that show enhanced proliferation, migration and resistance to drug treatment. 50 Owing to the important role of Rif1 in NHEJ, it is not surprising that it acts as an anti-apoptosis factor and is linked with tumor formation. [51][52][53] The Rif1 level is found to be significantly increased in the breast cancer cells and depletion of Rif1 makes them more sensitive to drug treatment. 51 Here we observed that Rif1 is highly upregulated in Smad3 − / − ESCs, which also adopt some cancer cell-like properties. Upregulation of Rif1 leads to enhanced DNA repair, most likely through NHEJ. Therefore, the chromatin of Smad3 − / − ESCs may be more unstable than WT ESCs and knockdown of Rif1 in Smad3 − / − ESC may just slow down the pace of the cells from evolving into more malignant cells. It has been found that Rif1 depletion can sensitize cancer cells to drug treatment with enhanced apoptosis. 51 Recent study revealed that Rif1 is important in maintaining the telomere stability in ESCs as it can repress Zscan4, which can trigger hyper-telomere elongation and cell senescence from an elevated expression. 18 Interestingly, the Rif1 shRNAtransduced Smad3 − / − ESCs formed smaller teratomas than WT and Smad3 − / − ESCs. This might be because the constant expression of shRNA of Rif1 triggers apoptosis and cell senescence. On the basis of these results, it is worthwhile investigating whether control of Rif1 levels by drugs could benefit the treatment of teratocarcinoma.

Materials and methods
Cell culture and differentiation. Derivation of WT and Smad3 − / − mouse ESCs was described in previous reports. 12,54 The ESCs were maintained on feeders under the normal ESC medium, which is composed of DMEM with high glucose ( For ESC differentiation assay, 2i and LIF were removed and 1 μM retinoic acid (RA) was added in the culture medium to induce ESC differentiation. For EB formation assay, monolayer undifferentiated WT ESCs and Smad3 − / − ESCs were Plasmid constructs. Two shRNA constructs targeting Rif1 were generated according to previous reports with pSuper.puro vector (Addgene, Cambridge, MA, USA). 35 To generate lentiviral vector for express shRNA by lentivirus, Rif1 shRNA sequences together with the H1 promoter were cut from pSuper plasmid with EcoRI and ClaI and sub-cloned to pLVTH plasmid. To construct the Rif1 promoter reporter plasmid, a 2000 bp fragment, encompassing the Smad3-binding sites, was amplified by primers from the genomic DNA of mouse ESCs and cloned into a pGL3 vector (Promega, Madison, WI, USA) at the MluI and XhoI sites. To construct the pCAG-GFP, pCAG-Smad3 and pCAG-Pou5f1 plasmids, the ORF sequences of these genes were amplified from the cDNA of mouse ESCs, digested and inserted into pCAG-Flag vector (Addgene) at BglII-XhoI (GFP and Smad3) and MluI-XhoI (Pou5f1) sites. All the amplification primers have been added to Table 1 and these recombinant vectors have been sequenced.
Real-time PCR assay. Real-time PCR analysis was conducted using the ABI Prism 7900HT (Applied Biosystems, Foster City, CA, USA) analysis machine with SYBR Premix Ex Taq (TaKaRa, Shiga, Japan) according to the manufacturer's instructions. The cycle (C T ) values of target genes were first normalized against the C T value of an internal control (Actin gene) and then normalized against the C T value of corresponding transcripts of the control sample. The DNA primer sequences used for the real-time PCR assay are listed in Table 1. For each pair of the primer, only one correct size band and one peak were detected. All the real-time PCR assays comprised triplicate data with samples from three independent experiments. ChIP assay. A ChIP assay for mouse ESCs was carried out as described previously. 55 Briefly, 1% (w/v) formaldehyde was added to 3 × 10 7 cells and incubated at room temperature for 10 min, then inactivated by adding 125 mM glycine for 5 min. The cells were then lysed and the chromatin fragmented by a Bioruptor Sonicator (Bioruptor UCD-200, Diagenode Company, Liège, Belgium) to a size around 500 bp. Soluble chromatins were incubated at 4°C overnight with a Dynabead (Invitrogen) coupled anti-Smad3 antibody (06-920, Upstate), anti-Oct3/4 (N-19) (sc-8628; Santa Cruz Biotechnology), anti-H3K9Me2 antibody (Milipore), anti-H3K9Me3 antibody (Milipore), anti-H3K4Me3 antibody (Abcam), anti-H3K27Me3 antibody (Milipore), anti-SUZ12 antibody (Abcam) or a corresponding control IgG (Milipore). The antibody-enriched DNAs were decrosslinked and purified with phenol-chloroform (Ambion, Applied Biosystems), followed by ethanol precipitation. The precipitated DNA was dissolved in TE buffer and analyzed by real-time PCR using the ABI Prism 7900HT sequence detection system and SYBR Premix Ex Taq (TaKaRa). Fold enrichments of the enriched DNA were calculated according to ratios of the immunoprecipitated DNA to the input samples and then normalized against the DNA level at control regions. All the DNA primer sequences used for the ChIP-qPCR assay are listed in Table 1. For each pair of the primer, only one correct size band and one peak were detected. All the real-time PCR assays comprised triplicate data with samples from three independent experiments.
Luciferase assay. For luciferase assay, the Renilla plasmid (5 ng per well) was used as the internal transfection control, whereas the pGL3 empty vector (100 ng  Wound-healing assay and transwell assay. 4 × 10 5 WT and Smad3 − / − ESCs were seeded in six-well tissue culture plates, respectively, in 2i and LIF ESC culture media. After 2 days, the ESC media were changed to differentiation media, in which 2i and LIF were removed and 1 μM RA (Sigma) was added, for another day to promote quick ESC differentiation. When ESCs were completely differentiated into monolayer cells, autoclaved yellow pipette tips were used to generate scratches. After scratching, the detached cells were removed by washing twice with DPBS buffer, and then cultured in differentiation media for another 24 h under normal condition. Images were taken by a Zeiss microscope (Carl Zeiss, Jena, Germany) and analyzed, using Image J software (NIH, Washington, DC, USA), at 0 and 24 h. Triplicate independent assays were performed. The cell invasive assay was performed using the CytoSelect 24-Well Cell Invasion Assay Kit (Cat # CBA-110-COL, Cell Biolabs, Inc., San Diego, CA, USA) according to the manufacturer's instructions. Genotyping primers pLvth-Rif1 shRNA F 5′-CGCTGACGTCATCAACCCGCTCCAAGGA-3′ pLvth-Rif1 shRNA R 5′-CGTATAATGTATGCTATACGAAG-3′ The role of Rif1 in mouse ES cell stability P Li et al