Telomerase-mediated telomere elongation from human blastocysts to embryonic stem cells

ABSTRACT High telomerase activity is a characteristic of human embryonic stem cells (hESCs), however, the regulation and maintenance of correct telomere length in hESCs is unclear. In this study we investigated telomere elongation in hESCs in vitro and found that telomeres lengthened from their derivation in blastocysts through early expansion, but stabilized at later passages. We report that the core unit of telomerase, hTERT, was highly expressed in hESCs in blastocysts and throughout long-term culture; furthermore, this was regulated in a Wnt–&bgr;-catenin-signaling-dependent manner. Our observations that the alternative lengthening of telomeres (ALT) pathway was suppressed in hESCs and that hTERT knockdown partially inhibited telomere elongation, demonstrated that high telomerase activity was required for telomere elongation. We observed that chromatin modification through trimethylation of H3K9 and H4K20 at telomeric regions decreased during early culture. This was concurrent with telomere elongation, suggesting that epigenetic regulation of telomeric chromatin may influence telomerase function. By measuring telomere length in 96 hESC lines, we were able to establish that telomere length remained relatively stable at 12.02±1.01 kb during later passages (15–95). In contrast, telomere length varied in hESCs with genomic instability and hESC-derived teratomas. In summary, we propose that correct, stable telomere length may serve as a potential biomarker for genetically stable hESCs.


INTRODUCTION
Mammalian telomeres are composed of tandem repeats of (TTAGGG) n DNA sequences and telomere-associated proteins, which protect chromosome ends and maintain chromosomal stability (Blackburn, 2001;Palm and de Lange, 2008). The telomeres of somatic cells progressively shorten with each cell division due to the end-replication problem (Blasco, 2005). When telomeres reach a critically short length, cells exhibit genomic instability and undergo cell cycle arrest, replicative senescence or apoptosis (d'Adda di Fagagna et al., 2003). Telomere maintenance is principally associated with telomerase activity, which protects telomere integrity through the addition of single stranded TTAGGG repeats onto the 39-ends of chromosomes. Telomerase activity is generally silenced or found at very low levels in somatic cells, and telomerase expression is insufficient to maintain telomere length in most adult stem cells, which limits cell division (Redaelli et al., 2012). In contrast, most tumor cells, germline cells and embryonic stem cells (ESC) exhibit high levels of telomerase expression. An important complementary maintenance mechanism is the alternative lengthening of telomere (ALT) pathway, in which telomeres are elongated by homologous recombination at a later stage of replication. The ALT pathway has been found to be activated in some tumors (Henson et al., 2002) and during early embryo development (Liu et al., 2007).
Human ESCs (hESCs) are derived from the inner cell mass of human blastocysts. They are capable of indefinite division and are considered to be a candidate cell source for regenerative medicine. High telomerase activity is an important characteristic of ESCs. Our current knowledge of telomere maintenance in ESCs is largely derived from studies in mouse ESCs (mESCs): telomere elongation has been observed during in vitro derivation of mESCs (Varela et al., 2011); loss of telomerase activity in mESCs resulted in progressive telomere loss, genomic instability, aneuploidy and telomere fusion, eventually leading to reduced growth rates (Niida et al., 1998); telomerase overexpression enhanced self-renewal, improved resistance to apoptosis and increased proliferation in mESCs (Armstrong et al., 2005). Tetraploid embryo complementation experiments have shown that telomerase-deficient mESCs with short telomeres lose their ability to generate complete ESC-pups (Huang et al., 2011). In contrast, induced pluripotent stem cells (iPSC), which resemble ESCs in many ways, with long telomeres, generate chimeras at a higher efficiency than those with shorter telomeres (Huang et al., 2011).
In combination, these reports indicate that telomerase plays a key role in ESC proliferation. Furthermore, correct telomere function is required to maintain the pluripotency of ESCs and a correlation exits between telomere length and self-renewal. Although telomerase activation and restoration of telomere length are essential for cellular reprogramming (Marión et al., 2011), telomere maintenance and its biological significance have not been fully investigated in hESCs. In this study, we monitored telomere length and maintenance in 96 hESC lines in vitro, from their derivation in blastocysts throughout early and prolonged culture. Our data revealed that hESCs preserved telomerase activity and acquired elongated telomeres during early culture and these stabilized during later culture. Further investigation indicated that molecular genotyping in aberrant hESCs and teratomas might also be involved in telomere lengthening. These finding suggested that a correct, stable telomere length could be a biomarker for genomically stable hESCs.

Telomere elongation in hESCs occurs during the early expansion in vitro
Four hESC lines, chHES-8, chHES-20, chHES-22 and chHES-69, were prepared from blastocysts and set to an initial telomere length, as described in Materials and Methods. Cells at interphase were analyzed by telomere quantitative fluorescence in situ hybridization (Q-FISH) during early (#5) and later (.15) passages. Fig. 1A gives the mean fluorescence intensities as arbitrary units (a.u.) relative to LY-S cells. The results revealed that hESCs from later passages had the longest telomeres, and that hESCs from early passages had longer telomeres than those in blastocysts. To verify these observations, telomere length was measured in seven hESC lines at different passages by telomeric terminal restriction fragment (TRF) assays (Fig. 1B). The results confirmed that early-passaged hESCs (,15) had shorter telomeres than late-passaged hESCs (P,0.05). Furthermore, the elongated telomeres in late-passaged hESCs (15-95) were maintained at a relatively stable length (P.0.05; supplementary material Table S1; Fig. S1). These results demonstrated that telomeres in hESCs progressively elongate during early expansion in vitro, eventually reaching a relatively stable length once the cells have established.
Telomerase regulates telomere lengthening in hESCs Telomerase lengthens telomeres by adding telomeric repeats onto chromosome ends. Telomerase reverse transcriptase (hTERT) is the limiting factor for telomerase activity. We therefore compared hTERT expression levels in the inner cell mass of six blastocysts (B) TRF assay to determine the telomere length in hESCs at different passages was performed using a biotin-labeled telomeric DNA probe (left). Mean telomere lengths were calculated in seven cell lines 26,20,22,69,137 and 177) at different passages (right). P-values were determined using the independent samples t-test.
with those in four hESC lines at passage P2. The results showed a consistently high level of hTERT expression in both groups ( Fig. 2A,B) suggesting that hTERT may play a key role in the regulation of telomere length in hESCs. To determine whether hTERT expression continued to influence telomere elongation in hESCs during early culture, we measured telomere length in early-passaged cells after silencing hTERT expression by lentiviral shRNA interference. Quantitative real-time reversetranscription polymerase chain reaction (qRT-PCR) and telomeric repeat amplification protocol (TRAP) assay showed that both hTERT expression and telomerase activity decreased by ,70% (supplementary material Fig. S2A,B). In order to investigate the mechanism by which telomerase activity influenced telomere amplification, we measured telomere length in hESCs at different passages. Although telomere lengthening still occurred in hTERTdeficient cells, it was 50% slower than in wild-type hESCs ( Fig. 2C; supplementary material Fig. S2C). To exclude the possibility that the growth rate influenced telomere lengthening, we compared growth curves between wide-type and hTERTsilenced hESCs, and found no significant difference between the two cell types (Fig. 2D). These data suggested that telomerase activity is necessary for telomere elongation in hESCs during early expansion.
It has previously been suggested than telomere lengthening occurs in telomerase-negative cancers and during embryonic development through the ALT pathway, which is based on chromosome recombination . We therefore investigated the ALT pathway in hESCs by examining the levels of ALT-associated promyelocytic leukemia (PML) nuclear bodies (APB) and telomere sister chromatid exchange (T-SCE), which are the hallmarks of ALT pathway activation (Muntoni and Reddel, 2005). The data showed that APBs occurred in ,5% of U20S cells (an ALT cell line) but were absent in hESCs at all passage steps (Fig. 3A). In addition, a low rate of T-SCE was observed in hESCs from both initial and late passages compared to U20S cells (Fig. 3B). These results indicated that the primary mechanism for telomere elongation and maintenance in hESCs is telomerase dependent, and that the ALT pathway has lesser influence. Fig. 3. The ALT pathway is suppressed in hESCs. (A) ALT-associated promyelocytic leukemia nuclear bodies (APB) are absent in hESCs. Cells were stained with DAPI (blue) and PML (red) antibody combined with telomeric DNA-FISH (green). In the merged images yellow foci show APBs in ,5% of U20S cells but none in hESCs. (B) Low levels of telomere sister chromatid exchange (T-SCE) occur in hESCs. BrdU was incorporated into cells from the indicated passages during a single round of DNA replication (,20 hours) and T-SCE analysis was performed using a telomeric C-strand probe (green). The inset images show non-telomere exchange (two signals; chromosome A) and telomere exchange (three signals; chromosome B). The results were quantified; values are means 6 s.d. from three independent experiments. Statistical analysis was performed using the independent samples t-test.
To further understand the mechanisms that regulate telomerase activity in hESCs, we focused on the Wnt-b-catenin signaling pathway, which has previously been reported to play a major role in regulating TERT expression and telomere length in mESCs (Hoffmeyer et al., 2012). Real-time qRT-PCR ( Fig. 2A) and chromatin immunoprecipitation (ChIP) assays ( Fig. 4A) were employed to analyze b-catenin mRNA expression in hESCs from blastocysts and throughout long-term culture. The results identified the hTERT promoter transcription start sequence in bcatenin pull-downs from hESCs, suggesting that b-catenin binds directly to the hTERT promoter region. Therefore we investigated whether Wnt-b-catenin participated in the regulation of telomerase activity in hESCs undergoing Wnt-3a treatment. We found that hTERT expression was upregulated and telomerase activity was increased in hESCs after 6 hours of Wnt-3a treatment ( Fig. 4B,C, respectively). In contrast, knockdown of b-catenin led to decreased hTERT expression and decreased telomerase activity in hESCs. Taken together, these results indicated that the Wnt-b-catenin pathway was involved in maintaining telomerase activity in hESCs. It has also been reported that b-catenin may regulate telomerase activity in mESCs through the transcription factor KLF4. b-catenin and KLF4 co-immunoprecipitated from hESCs (supplementary material Fig. S3A), therefore we explored the interaction between b-catenin and KLF4 in transcriptional regulation of hTERT. We cloned the core region of the hTERT promoter, which harbors KLF4 binding sites (2211, +40), into PGL3 luciferase reporters and transfected these into hESCs. The data showed that overexpression of either KLF4 or b-catenin led to increased promoter activity in hESCs, whereas b-catenin knockdown led to inhibition of promoter activity; furthermore, this could not be rescued by KLF4 overexpression (supplementary material Fig.  S3B). These data supported the proposed role of b-catenin in the regulation of hTERT expression in hESCs.

Telomere elongation is associated with decreased histone methylation
Our results had revealed that telomere length in hESCs was relatively stable at later passages. Therefore we speculated that other factors were involved in regulating telomere length in addition to telomerase activity. Previous reports have demonstrated that specific chromatin modifications at telomere sites are involved in the regulation of telomere length. TRF1, a telomeric DNA binding protein, is a component of shelterin subunits associated with mammalian telomeres. It binds specifically to double-stranded telomeric DNA and has been reported to enhance telomere shortening in cancer cells by fastening to chromatin in telomeres, thereby reducing the affinity of telomerase to these regions (Seimiya et al., 2005). However, our data showed that TRF1 expression increased in hESCs at later passages (Fig. 5A). This suggested that different chromatin structures may exist in early and late-passaged hESCs, in agreement with a previous report (Blasco, 2007b). It is now accepted that the main histone modifications occur at the H3K9me3 and H4K20me3 sites in telomeric regions, and regulate telomere lengthening (Blasco, 2007a). To observe changes in chromatin structure during telomere elongation, we performed double immunofluorescence staining combined with confocal scanning to measure the extent of colocalization of H3K9me3 and H4K20me3 with telomeres. In agreement a previous report, we found that colocalization was ,13% in human fibroblasts (Varela et al., 2011). In addition, the colocalization signal for H3K9me3 and H4K20me3 with TRF1 was 27.9% and 16.7%, respectively, in early-passaged hESCs, reducing to 13.8% and 7.6%, respectively, in late-passaged hESCs (Fig. 5B). A ChIP assay not only showed that binding of these markers to telomeric DNA decreased in late-passaged hESCs ( Fig. 5C) but also that the binding level to telomeres was stable in these hESCs. These observations indicated that switching between heterochromatin states might be involved in telomere lengthening in hESCs.
Telomere elongation is not associated with three-germ-layer differentiation Recovery of pluripotency and restoration of telomere length are twin facets of somatic cell reprogramming (West and Vaziri, 2010). Therefore, to explore the biological significance of telomere lengthening, both early-passaged and late-passaged hESCs were analyzed using in vitro and in vivo differentiation assays. The data showed that there were no significant differences in expression levels of pluripotency markers, including AKP, OCT-4 SSEA-3, SSEA-4, TRA-1-60 and TRA-1-81 between the two passage groups (Fig. 6A). The hESCs from all the passages expressed the triploblastic markers NGN3, HNF4 (endoderm markers), Tbx4 (mesoderm marker) and KRT17 (ectoderm marker) in formed embryoid bodies, and developed b-tubulin-, AFP-and SMA-positive cells after 3 weeks in vitro differentiation (Fig. 6B,C). Early-and late-passaged hESCs expressed similar proliferation markers (Fig. 6D) and displayed similar levels of three-germ-layer differentiation potential following in vivo (B) Heterochromatic markers are reduced in late-passaged hESCs. Fluorescence images show that H3K9me3 (green) or H4K20me3 (green) colocalize with TRF1 (red) in hESCs at the indicated passages. Images were captured by confocal microscopy. The bar graph gives the mean 6 s.d. from three or more independent experiments; statistical analysis was performed using the independent samples t-test; n5number of cells. (C) Telomeric DNA-associated H3K9me3 or H4K20me3 in hESCs at different passages. The ChIP assay was used to analyze hESCs at initial or later passages using telomeric probes. Data are means 6 s.d. of three independent experiments. teratoma formation (Fig. 6E). These data suggested that hESCs maintain a pluripotent state throughout in vitro culture, and that telomere elongation during early-stage hESC culture may not be directly related to three-germ-layer differentiation.
Aberrant telomere length is associated with genomic instability and teratoma formation in hESCs As telomere length had been found to be stable after early culture, we measured telomere length after passage P15 in the 96 hESC lines, including the widely used H9 cell line, to determine the normal range of telomere length (Fig. 7A). The mean telomere length was 12.0261.01 kb and did not follow Gaussian distribution, with telomere lengths of 9.8-14.4 kb between the 2.5th and 97.5th percentiles (supplementary material Table S2). We noted that the chHES-3 line, which has an unstable genomic state and acquires progressive chromosomal abnormalities during in vitro culture (Yang et al., 2008;Yang et al., 2010), had significantly shorter telomeres (P,0.05) up to passage P60 (Fig. 7B). We also observed a distinct elongation of telomeres in teratomas compared with the original hESCs (Fig. 7C). Teratoma formation leads to a failure of ESC-derived cell therapy, even following successful engraftment and functional improvement (Bongso et al., 2008), therefore we proposed that the variation in telomere length reflected genomic instability in hESCs. To confirm this hypothesis, we performed copy number variation (CNV) analyses in hESCs and teratoma cells derived from corresponding hESCs by SNP 6.0 array using the Affymetrix Genotyping Console 4.1.4. Variations in copy number .100 kb between early and latepassaged hESCs or between hESCs and corresponding teratoma cells were identified (supplementary material Table S4) and their frequency was calculated. The data showed that the difference in the number of novel CNVs between early and late-passaged hESCs was not significant (Fig. 8A). The CNV data were analyzed further for the hESC lines that had been found to have variations in telomere length: aberrant chHES-3 cells, which have short telomeres, had a total of 44 different CNVs .100 kb after culture for several months, of which 13 CNVs were .1 Mb; in contrast, normal chHES-3 cells had only 8 CNVs .100 kb, none of which were .1 Mb. In addition, the number of novel CNVs in teratomas was greater than in corresponding hESCs (Fig. 8B). These data suggested that there is a correct range of telomere lengths in hESCs, implying that variations may correlate with genomic instability and the potential for teratoma formation in hESCs.

DISCUSSION
It has previously been established that hESCs derived from the inner cell mass of blastocysts have the capacity for three-germlayer differentiation and self-renewal through unlimited symmetric cell division in vitro (He et al., 2009). Telomere length and telomerase activity have also been associated with unlimited proliferation in cells. We have shown that a high level of telomerase activity was maintained in hESCs from initial derivation to long-term culture; however, the rate of telomere lengthening during the early expansions (,15 passages) stabilized in later passages and telomere length remained relatively constant (12.0261.01 kb) for at least 95 passages. Similar observations have been reported in relation to cell reprogramming: telomere elongation was accompanied by the capacity for restoration of self-renewal in iPS cells; and telomere length was restored in dyskeratosis-congenita-derived iPS cells, which harbor genomic defects in the telomerase RNA component (TERC) (Marión et al., 2009;Agarwal et al., 2010). These findings indicated that telomere length may be a specific indicator of the acquisition, restoration and maintenance of self-renewal capacity in hESC and iPS cells during their establishment in vitro.
Our findings also indicated that telomere elongation in hESCs during early culture was telomerase dependent, as reducing telomerase activity in hESCs led to a decrease in the rate of telomere lengthening and suppression of the ALT pathway.
The expression of Wnt family members in feeder cells was reported to enhance proliferation and improve survival in ESCs (Hasegawa et al., 2012). In addition, Wnt-b-catenin signaling has been linked to TERT expression in both mESCs and human carcinoma cells (Hoffmeyer et al., 2012). Therefore we investigated the mechanism by which Wnt-b-catenin signaling might regulate hTERT expression in hESCs. Our results indicated that b-catenin induced hTERT expression by binding to the hTERT promoter region, suggesting that Wnt-b-catenin signaling regulates telomere lengthening in hESCs by promoting hTERT expression. We also found that KLF4, which is closely associated with pluripotency, was involved in regulating telomerase expression in hESCs, further supporting a potential relationship between the regulation of telomere length and the ability for self-renewal.
Histone methyltransferase was reported to control telomere length in a knockout mouse model (García-Cao et al., 2004); and a decrease in the heterochromatic markers H3K9me3 and H4K20me3 in the telomeres of mESCs and iPSCs has been observed during telomere lengthening (Marión et al., 2011;Varela et al., 2011). These reports suggested that chromatin status may influence its affinity for telomerase and thereby control telomere length. By investigating this relationship we found a reduction in these methylation markers in the telomeres of hESCs during early passages; however, this was seen to stabilize during later passages, which was similar to the pattern observed for telomere lengthening. Therefore our data support the hypothesis that chromatin modification plays a role in controlling and maintaining telomere length, and that a telomerasemediated mechanism dependent on chromatin modification may control telomere length in hESCs.
Although it has been suggested that demethylation may increase activation of the ALT pathway in mESCs (Zalzman et al., 2010), the characteristic features of ALT activation were absent in this study. Furthermore, we observed telomere elongation in a parthenogenetic cell line (hESC-69), which lacks long telomeres. These findings support a previous study showing that ALT signaling can be suppressed by high levels of telomerase activity in hESCs (Ford et al., 2001).
Maintaining a correct telomere length is one of the key factors for successful cell replication. Studies on the relationship between telomere length and tumorigenesis have reported that a variation in telomere length may affect tumor progression and could be a potential indicator for patient survival (Zhou et al., 2012). Telomerase-deficient mESCs (Liu et al., 2000) exhibit telomere attrition, leading to genomic instability during prolonged proliferation. Although it is acknowledged that maintaining genetic stability affects indefinite hESC division, there has been little investigation to determine the correct length of telomeres. We found that the majority of hESCs in this study maintained a relatively stable telomere length during longterm culture, but abnormal shortening or elongation of telomeres was observed in aberrant chHES-3 cells and in hESC-derived teratomas. Genetic instability is acknowledged to be a major obstacle in hESC therapy against tumorigenesis in vivo (Bongso et al., 2008). However, there is little consensus on a standard method for measuring the early events of genetic instability in hESCs. Our study has demonstrated that a stable telomere state was obtained after several months of culture, in accordance with the time required for hESCs to establish in vitro. This further supported our suggestion that telomere length may indicate the level of genetic stability in hESCs. We therefore propose that a correct, stable telomere length could serve as a potential marker for the characterization of hESCs.

Sample preparation
The hESC lines were obtained from the established hESC bank at our center (Lin et al., 2009). The blastocysts used in this study were collected from patients receiving in vitro fertilization treatment. Written, informed consent was obtained from all patients donating surplus embryos for research purposes. The study was approved by the Reproductive & Stem Cell Engineering of Central South University and the Reproductive & Genetic Hospital of CITIC-Xiangya Ethics Committee.
The initial telomere length was set in the inner cell mass of five clinical unused blastocysts from which the hESCs were derived. The hESC lines were established by mechanically separating the inner cell mass of 6-day blastocysts and plating them onto feeder cells for culture (passage zero, P0). Cells were routinely passaged every 7 days by mechanical cutting.

Cell culture
The hESCs were cultured in serum-free K-SR medium containing KnockOut DMEM (Gibco, Carlsbad, CA, USA) supplemented with 15% serum replacement (Gibco), 0.1 mM b-mercaptoethanol (Sigma, St. Louis, MO, USA), 1% nonessential amino acids (Gibco), 2 mM L-glutamine (Gibco) and 4 ng/ml human recombinant basic fibroblast growth factor (bFGF; Gibco), at 37˚C in a humidified atmosphere of 5% CO 2 . For Wnt-3a treatment, 100 ng/ml Wnt-3a was added to feeder-free cultured hESCs in N2B27 medium (Gibco) with 10 ng/ml bFGF and cell samples were collected at different time points. To construct the cell growth curve, hESCs were seeded onto 12-well plates. Fresh medium was added every 24 hours and the cells were counted at 1-day intervals using a NucleoCounter YC-100 automated cell counter (ChemoMetec, Norway). LY-S, LY-R and U2OS cells were cultured in DMEM (Gibco) medium with 10% fetal bovine serum under standard conditions. For embryoid body formation, hESCs clones were cut and suspended in ES medium (enriched natural seawater medium) without bFGF and cultured for 14 days to obtain embryoid body cells; the medium was changed every 24 hours. For in vitro differentiation, embryoid bodies were attached to slides and cultured for at least 3 weeks in ES medium without bFGF. For teratoma formation, 1610 6 ESCs were embedded in 30% Matrigel and injected into the rear leg of a mouse. Tumors were collected after 6-8 weeks. All animal studies were authorized by the Xiang-Ya Institute of Animal Use and Care Committee.

Antibodies
The following antibodies and telomeric probes were used in this study. . FITC/CY3-labeled Telo-G/C probe was obtained from Dako (Denmark) and the biotin-labeled Telo-G probe from TaKaRa (Japan).

Plasmids
The coding sequence for b-catenin cDNA was blunt-end cloned into the backbone of EcoRI-and SpeI-digested pSin-EF2-SOX2-Pur vector (Addgene, Cambridge, MA, USA) to generate the pSin-EF2-CTNNB1-Pur vector. The TERT promoter was PCR amplified from hESC genomic DNA and the PCR product was digested by KpnI and HindIII and cloned into KpnI-HindIII sites of the pGL3-Basic reporter vector (Promega, Madison, WI, USA) to create Luc-Tert-250 bp vector. All primers and sequences are listed in supplementary material Table S3.

qRT-PCR
RNA extraction of inner cell mass and initial hESCs clones was performed using a RealTime Ready Cell Lysis Kit (Roche, Switzerland) according to the manufacturer's protocol. Real-time qRT-PCR was performed as previously described (Agarwal et al., 2010). All primers and sequences are listed in supplementary material Table S3.
Telomere sister chromatid exchange fluorescence in situ hybridization Cells were examined in metaphase using Q-FISH by culturing established hESCs for 3-4 days and stalling them at metaphase by incubation with 0.1 mg/l colcemid for 1.5 hours. T-SCE analysis was performed using chromosome orientation-FISH (CO-FISH) by adding 30 mM BrdU to the culture medium 18 hours before colcemid treatment, as described above. The cells were then collected, resuspended in 0.56% KCl buffer for 10 minutes and fixed three times in fixation buffer (methanol:acetic acid, 3:1). Metaphase spreads were prepared by dropping fixed cells onto glass microscope slides. Cells in interphase were fixed directly onto microscope slides before examination. The Q-FISH procedure was performed as previously described (Gonzalo et al., 2006;. In order to avoid non-specific signals, the procedure was optimized by increasing the temperature to 45˚C during the first step of probe washing. LY-S and LY-R cell lines were used to generate a calibration curve for each experiment. Telomere length was determined by fluorescence intensity and was integrated using TFl-TELO software. Terminal restriction fragment assay TRF analysis is highly reproducible and recognized as the gold standard for telomere length measurement (Allshire et al., 1989). Genomic DNA was isolated following standard procedures and digested with the HinfI endonuclease at 37˚C overnight. DNA samples were separated by pulsed field electrophoresis on a 1% agarose gel to obtain a linear separation of 6-20 kb. Alkaline transfer of the DNA was performed and the membranes were hybridized using a biotin-labeled telomere probe. A Chemiluminescent Nucleic Acid Detection Module (Thermo Fisher Scientific, Rockford, IL, USA) was used for signal detection. Data processing and telomere length analyses were performed using Quantity One software.

Immunofluorescence microscopy
Cells were pre-extracted in ice-cold PBS containing 0.5% Triton X-100 for 1 minute at 4˚C. The cells were fixed and blocked before being incubated with the specified primary antibodies at 37˚C for 60 minutes, then washed and incubated with the appropriate secondary antibodies for 45 minutes at 37˚C. The following antibodies were used: anti-PML, anti-TRF2, anti-b-TUBLIN, anti-AFP, anti-SMA and anti-rabbit IgG (Santa Cruz, CA, USA), anti-TRF1, anti-H4K20me3, anti-b-catenin (Sigma, MO, USA), anti-H3K9me3, all the secondary antibodies (Cell signaling, USA). For combined telomeric DNA-FISH analysis, the cells were immunostained, treated with 0.5 mg/ml RNase A, dehydrated and incubated in hybridization solution containing 0.3 mg/ml Tel-C PNA telomeric probes (Applied Biosystems, Carlsbad, CA, USA). The slides were washed, dehydrated and mounted using anti-fading DAPI medium (Dako, Carpinteria, CA, USA). The cells were then visualized by fluorescence (Zeiss, Germany) or confocal (Olympus, Japan) microscopy. For confocal microscopy, the images were captured as a series of z-sections separated by 0.27 mm, taken at a depth of 30-40 mm. Image data were acquired and analyzed using confocal software (Olympus).

Chromatin immunoprecipitation assay and telomere dot-blots
ChIP assays were carried out as previously described Hoffmeyer et al., 2012). Briefly, 2610 6 hESCs were used for each experimental condition. Cells were sonicated after cross-linking to obtain chromatin fragments ,0.5-1 kb in length. After pre-incubation, protein-A-G-Sepharose slurry and the specified antibodies were added to the samples followed by rotation for 2 hours at 4˚C. The pellets were washed and the chromatin was eluted from the beads before being hybridized with the Tel-G (TTAGGG) 9 probe or analyzed by real-time qRT-PCR.

Lentiviral shRNA constructs
The shRNAs were designed according to the sequences given in supplementary material Table S3, and cloned to the MISSIONH pLKO.1-puro lentiviral construct (Sigma). The viral particles were packaged and collected according to the manufacturer's instructions, added to hESC culture medium with 8 ng/ml polybrene and incubated overnight. The second infection was performed the following day. After infection, hESCs were selected and cultured with 1 mg/ml puromycin in ES medium and passaged as usual.
Western blot analysis and co-immunoprecipitation (co-IP) Western blot analyses and immunoprecipitation were carried out following standard procedures. Briefly, cell extracts were prepared in lysis buffer containing 50 mM Tris-HCl (pH 7.5), 120 mM NaCl and 1% NP-40 with protease inhibitors. Cellular protein (0.5 mg) was immunoprecipitated with the specified antibodies and protein A/G beads. The proteins were then separated and analyzed by western blotting.

Telomerase activity
Quantitative measurement of telomerase activity was performed using the TRAPeze Telomerase detection kit or TRAPeze RT Telomerase detection kit (Millipore, Bedford, MA, USA). Briefly, the cells were lysed on ice for 30 minutes in CHAPS buffer and then centrifuged at 12,000 g for 20 minutes at 4˚C. The supernatant was collected and the protein concentrations were determined by BCA protein assay. The TRAP assay was carried out according to the manufacturer's instructions, using a total protein extract of 200 ng in each reaction. Telomerase activity was determined by electrophoresis of the PCR products in a 12.5% nondenaturing polyacrylamide gel with silver staining, or by fluorometric detection and real-time quantification of the PCR products using an ABI Prism 7500.

Luciferase assay
ESCs were separated by adding 0.02% EDTA and then seeded on 24-well Matrigel-coated plates in mTeSR medium (BD Biosciences, San Jose, CA, USA). The cells were co-transfected with the plasmids after 48-72 hours using X-tremeGENE HP (Roche). Luciferase activity was then quantified in a Dual-Luciferase Reporter Assay System after 72 hours. The reporter activities were normalized to the internal control in order to minimize experimental variability due to differences in cell viability or transfection efficiency. All experiments were performed in duplicate and repeated at least three times.

Statistical analysis
The independent sample t-tests between groups were used to evaluate the statistical significance of mean values by using SPSS 18.0. The level of significance was set to P,0.05.